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 debugmode =
127 ARG_DEF("D", "debug", 0, "Debug mode (makes output deterministic)");
128 static const arg_def_t outputfile =
129 ARG_DEF("o", "output", 1, "Output filename");
130 static const arg_def_t use_yv12 =
131 ARG_DEF(NULL, "yv12", 0, "Input file is YV12 ");
132 static const arg_def_t use_i420 =
133 ARG_DEF(NULL, "i420", 0, "Input file is I420 (default)");
134 static const arg_def_t use_i422 =
135 ARG_DEF(NULL, "i422", 0, "Input file is I422");
136 static const arg_def_t use_i444 =
137 ARG_DEF(NULL, "i444", 0, "Input file is I444");
138 static const arg_def_t use_i440 =
139 ARG_DEF(NULL, "i440", 0, "Input file is I440");
140 static const arg_def_t codecarg = ARG_DEF(NULL, "codec", 1, "Codec to use");
141 static const arg_def_t passes =
142 ARG_DEF("p", "passes", 1, "Number of passes (1/2)");
143 static const arg_def_t pass_arg =
144 ARG_DEF(NULL, "pass", 1, "Pass to execute (1/2)");
145 static const arg_def_t fpf_name =
146 ARG_DEF(NULL, "fpf", 1, "First pass statistics file name");
147 #if CONFIG_FP_MB_STATS
148 static const arg_def_t fpmbf_name =
149 ARG_DEF(NULL, "fpmbf", 1, "First pass block statistics file name");
151 static const arg_def_t limit =
152 ARG_DEF(NULL, "limit", 1, "Stop encoding after n input frames");
153 static const arg_def_t skip =
154 ARG_DEF(NULL, "skip", 1, "Skip the first n input frames");
155 static const arg_def_t deadline =
156 ARG_DEF("d", "deadline", 1, "Deadline per frame (usec)");
157 static const arg_def_t best_dl =
158 ARG_DEF(NULL, "best", 0, "Use Best Quality Deadline");
159 static const arg_def_t good_dl =
160 ARG_DEF(NULL, "good", 0, "Use Good Quality Deadline");
161 static const arg_def_t rt_dl =
162 ARG_DEF(NULL, "rt", 0, "Use Realtime Quality Deadline");
163 static const arg_def_t quietarg =
164 ARG_DEF("q", "quiet", 0, "Do not print encode progress");
165 static const arg_def_t verbosearg =
166 ARG_DEF("v", "verbose", 0, "Show encoder parameters");
167 static const arg_def_t psnrarg =
168 ARG_DEF(NULL, "psnr", 0, "Show PSNR in status line");
170 static const struct arg_enum_list test_decode_enum[] = {
171 { "off", TEST_DECODE_OFF },
172 { "fatal", TEST_DECODE_FATAL },
173 { "warn", TEST_DECODE_WARN },
176 static const arg_def_t recontest = ARG_DEF_ENUM(
177 NULL, "test-decode", 1, "Test encode/decode mismatch", test_decode_enum);
178 static const arg_def_t framerate =
179 ARG_DEF(NULL, "fps", 1, "Stream frame rate (rate/scale)");
180 static const arg_def_t use_webm =
181 ARG_DEF(NULL, "webm", 0, "Output WebM (default when WebM IO is enabled)");
182 static const arg_def_t use_ivf = ARG_DEF(NULL, "ivf", 0, "Output IVF");
183 static const arg_def_t out_part =
184 ARG_DEF("P", "output-partitions", 0,
185 "Makes encoder output partitions. Requires IVF output!");
186 static const arg_def_t q_hist_n =
187 ARG_DEF(NULL, "q-hist", 1, "Show quantizer histogram (n-buckets)");
188 static const arg_def_t rate_hist_n =
189 ARG_DEF(NULL, "rate-hist", 1, "Show rate histogram (n-buckets)");
190 static const arg_def_t disable_warnings =
191 ARG_DEF(NULL, "disable-warnings", 0,
192 "Disable warnings about potentially incorrect encode settings.");
193 static const arg_def_t disable_warning_prompt =
194 ARG_DEF("y", "disable-warning-prompt", 0,
195 "Display warnings, but do not prompt user to continue.");
197 #if CONFIG_VP9_HIGHBITDEPTH
198 static const arg_def_t test16bitinternalarg = ARG_DEF(
199 NULL, "test-16bit-internal", 0, "Force use of 16 bit internal buffer");
202 static const arg_def_t *main_args[] = { &debugmode,
223 &disable_warning_prompt,
227 static const arg_def_t usage =
228 ARG_DEF("u", "usage", 1, "Usage profile number to use");
229 static const arg_def_t threads =
230 ARG_DEF("t", "threads", 1, "Max number of threads to use");
231 static const arg_def_t profile =
232 ARG_DEF(NULL, "profile", 1, "Bitstream profile number to use");
233 static const arg_def_t width = ARG_DEF("w", "width", 1, "Frame width");
234 static const arg_def_t height = ARG_DEF("h", "height", 1, "Frame height");
236 static const struct arg_enum_list stereo_mode_enum[] = {
237 { "mono", STEREO_FORMAT_MONO },
238 { "left-right", STEREO_FORMAT_LEFT_RIGHT },
239 { "bottom-top", STEREO_FORMAT_BOTTOM_TOP },
240 { "top-bottom", STEREO_FORMAT_TOP_BOTTOM },
241 { "right-left", STEREO_FORMAT_RIGHT_LEFT },
244 static const arg_def_t stereo_mode = ARG_DEF_ENUM(
245 NULL, "stereo-mode", 1, "Stereo 3D video format", stereo_mode_enum);
247 static const arg_def_t timebase = ARG_DEF(
248 NULL, "timebase", 1, "Output timestamp precision (fractional seconds)");
249 static const arg_def_t error_resilient =
250 ARG_DEF(NULL, "error-resilient", 1, "Enable error resiliency features");
251 static const arg_def_t lag_in_frames =
252 ARG_DEF(NULL, "lag-in-frames", 1, "Max number of frames to lag");
254 static const arg_def_t *global_args[] = { &use_yv12,
270 #if CONFIG_VP9_HIGHBITDEPTH
271 &test16bitinternalarg,
276 static const arg_def_t dropframe_thresh =
277 ARG_DEF(NULL, "drop-frame", 1, "Temporal resampling threshold (buf %)");
278 static const arg_def_t resize_allowed =
279 ARG_DEF(NULL, "resize-allowed", 1, "Spatial resampling enabled (bool)");
280 static const arg_def_t resize_width =
281 ARG_DEF(NULL, "resize-width", 1, "Width of encoded frame");
282 static const arg_def_t resize_height =
283 ARG_DEF(NULL, "resize-height", 1, "Height of encoded frame");
284 static const arg_def_t resize_up_thresh =
285 ARG_DEF(NULL, "resize-up", 1, "Upscale threshold (buf %)");
286 static const arg_def_t resize_down_thresh =
287 ARG_DEF(NULL, "resize-down", 1, "Downscale threshold (buf %)");
288 static const struct arg_enum_list end_usage_enum[] = { { "vbr", VPX_VBR },
293 static const arg_def_t end_usage =
294 ARG_DEF_ENUM(NULL, "end-usage", 1, "Rate control mode", end_usage_enum);
295 static const arg_def_t target_bitrate =
296 ARG_DEF(NULL, "target-bitrate", 1, "Bitrate (kbps)");
297 static const arg_def_t min_quantizer =
298 ARG_DEF(NULL, "min-q", 1, "Minimum (best) quantizer");
299 static const arg_def_t max_quantizer =
300 ARG_DEF(NULL, "max-q", 1, "Maximum (worst) quantizer");
301 static const arg_def_t undershoot_pct =
302 ARG_DEF(NULL, "undershoot-pct", 1, "Datarate undershoot (min) target (%)");
303 static const arg_def_t overshoot_pct =
304 ARG_DEF(NULL, "overshoot-pct", 1, "Datarate overshoot (max) target (%)");
305 static const arg_def_t buf_sz =
306 ARG_DEF(NULL, "buf-sz", 1, "Client buffer size (ms)");
307 static const arg_def_t buf_initial_sz =
308 ARG_DEF(NULL, "buf-initial-sz", 1, "Client initial buffer size (ms)");
309 static const arg_def_t buf_optimal_sz =
310 ARG_DEF(NULL, "buf-optimal-sz", 1, "Client optimal buffer size (ms)");
311 static const arg_def_t *rc_args[] = {
312 &dropframe_thresh, &resize_allowed, &resize_width, &resize_height,
313 &resize_up_thresh, &resize_down_thresh, &end_usage, &target_bitrate,
314 &min_quantizer, &max_quantizer, &undershoot_pct, &overshoot_pct,
315 &buf_sz, &buf_initial_sz, &buf_optimal_sz, NULL
318 static const arg_def_t bias_pct =
319 ARG_DEF(NULL, "bias-pct", 1, "CBR/VBR bias (0=CBR, 100=VBR)");
320 static const arg_def_t minsection_pct =
321 ARG_DEF(NULL, "minsection-pct", 1, "GOP min bitrate (% of target)");
322 static const arg_def_t maxsection_pct =
323 ARG_DEF(NULL, "maxsection-pct", 1, "GOP max bitrate (% of target)");
324 static const arg_def_t corpus_complexity =
325 ARG_DEF(NULL, "corpus-complexity", 1, "corpus vbr complexity midpoint");
326 static const arg_def_t *rc_twopass_args[] = {
327 &bias_pct, &minsection_pct, &maxsection_pct, &corpus_complexity, NULL
330 static const arg_def_t kf_min_dist =
331 ARG_DEF(NULL, "kf-min-dist", 1, "Minimum keyframe interval (frames)");
332 static const arg_def_t kf_max_dist =
333 ARG_DEF(NULL, "kf-max-dist", 1, "Maximum keyframe interval (frames)");
334 static const arg_def_t kf_disabled =
335 ARG_DEF(NULL, "disable-kf", 0, "Disable keyframe placement");
336 static const arg_def_t *kf_args[] = { &kf_min_dist, &kf_max_dist, &kf_disabled,
339 static const arg_def_t noise_sens =
340 ARG_DEF(NULL, "noise-sensitivity", 1, "Noise sensitivity (frames to blur)");
341 static const arg_def_t sharpness =
342 ARG_DEF(NULL, "sharpness", 1, "Loop filter sharpness (0..7)");
343 static const arg_def_t static_thresh =
344 ARG_DEF(NULL, "static-thresh", 1, "Motion detection threshold");
345 static const arg_def_t auto_altref =
346 ARG_DEF(NULL, "auto-alt-ref", 1, "Enable automatic alt reference frames");
347 static const arg_def_t arnr_maxframes =
348 ARG_DEF(NULL, "arnr-maxframes", 1, "AltRef max frames (0..15)");
349 static const arg_def_t arnr_strength =
350 ARG_DEF(NULL, "arnr-strength", 1, "AltRef filter strength (0..6)");
351 static const arg_def_t arnr_type = ARG_DEF(NULL, "arnr-type", 1, "AltRef type");
352 static const struct arg_enum_list tuning_enum[] = {
353 { "psnr", VP8_TUNE_PSNR }, { "ssim", VP8_TUNE_SSIM }, { NULL, 0 }
355 static const arg_def_t tune_ssim =
356 ARG_DEF_ENUM(NULL, "tune", 1, "Material to favor", tuning_enum);
357 static const arg_def_t cq_level =
358 ARG_DEF(NULL, "cq-level", 1, "Constant/Constrained Quality level");
359 static const arg_def_t max_intra_rate_pct =
360 ARG_DEF(NULL, "max-intra-rate", 1, "Max I-frame bitrate (pct)");
361 static const arg_def_t gf_cbr_boost_pct = ARG_DEF(
362 NULL, "gf-cbr-boost", 1, "Boost for Golden Frame in CBR mode (pct)");
364 #if CONFIG_VP8_ENCODER
365 static const arg_def_t cpu_used_vp8 =
366 ARG_DEF(NULL, "cpu-used", 1, "CPU Used (-16..16)");
367 static const arg_def_t token_parts =
368 ARG_DEF(NULL, "token-parts", 1, "Number of token partitions to use, log2");
369 static const arg_def_t screen_content_mode =
370 ARG_DEF(NULL, "screen-content-mode", 1, "Screen content mode");
371 static const arg_def_t *vp8_args[] = { &cpu_used_vp8,
384 &screen_content_mode,
386 static const int vp8_arg_ctrl_map[] = { VP8E_SET_CPUUSED,
387 VP8E_SET_ENABLEAUTOALTREF,
388 VP8E_SET_NOISE_SENSITIVITY,
390 VP8E_SET_STATIC_THRESHOLD,
391 VP8E_SET_TOKEN_PARTITIONS,
392 VP8E_SET_ARNR_MAXFRAMES,
393 VP8E_SET_ARNR_STRENGTH,
397 VP8E_SET_MAX_INTRA_BITRATE_PCT,
398 VP8E_SET_GF_CBR_BOOST_PCT,
399 VP8E_SET_SCREEN_CONTENT_MODE,
403 #if CONFIG_VP9_ENCODER
404 static const arg_def_t cpu_used_vp9 =
405 ARG_DEF(NULL, "cpu-used", 1, "CPU Used (-8..8)");
406 static const arg_def_t tile_cols =
407 ARG_DEF(NULL, "tile-columns", 1, "Number of tile columns to use, log2");
408 static const arg_def_t tile_rows =
409 ARG_DEF(NULL, "tile-rows", 1,
410 "Number of tile rows to use, log2 (set to 0 while threads > 1)");
411 static const arg_def_t lossless =
412 ARG_DEF(NULL, "lossless", 1, "Lossless mode (0: false (default), 1: true)");
413 static const arg_def_t frame_parallel_decoding = ARG_DEF(
414 NULL, "frame-parallel", 1, "Enable frame parallel decodability features");
415 static const arg_def_t aq_mode = ARG_DEF(
417 "Adaptive quantization mode (0: off (default), 1: variance 2: complexity, "
418 "3: cyclic refresh, 4: equator360)");
419 static const arg_def_t alt_ref_aq = ARG_DEF(NULL, "alt-ref-aq", 1,
420 "Special adaptive quantization for "
421 "the alternate reference frames.");
422 static const arg_def_t frame_periodic_boost =
423 ARG_DEF(NULL, "frame-boost", 1,
424 "Enable frame periodic boost (0: off (default), 1: on)");
425 static const arg_def_t max_inter_rate_pct =
426 ARG_DEF(NULL, "max-inter-rate", 1, "Max P-frame bitrate (pct)");
427 static const arg_def_t min_gf_interval = ARG_DEF(
428 NULL, "min-gf-interval", 1,
429 "min gf/arf frame interval (default 0, indicating in-built behavior)");
430 static const arg_def_t max_gf_interval = ARG_DEF(
431 NULL, "max-gf-interval", 1,
432 "max gf/arf frame interval (default 0, indicating in-built behavior)");
434 static const struct arg_enum_list color_space_enum[] = {
435 { "unknown", VPX_CS_UNKNOWN },
436 { "bt601", VPX_CS_BT_601 },
437 { "bt709", VPX_CS_BT_709 },
438 { "smpte170", VPX_CS_SMPTE_170 },
439 { "smpte240", VPX_CS_SMPTE_240 },
440 { "bt2020", VPX_CS_BT_2020 },
441 { "reserved", VPX_CS_RESERVED },
442 { "sRGB", VPX_CS_SRGB },
446 static const arg_def_t input_color_space =
447 ARG_DEF_ENUM(NULL, "color-space", 1,
448 "The color space of input content:", color_space_enum);
450 #if CONFIG_VP9_HIGHBITDEPTH
451 static const struct arg_enum_list bitdepth_enum[] = {
452 { "8", VPX_BITS_8 }, { "10", VPX_BITS_10 }, { "12", VPX_BITS_12 }, { NULL, 0 }
455 static const arg_def_t bitdeptharg = ARG_DEF_ENUM(
457 "Bit depth for codec (8 for version <=1, 10 or 12 for version 2)",
459 static const arg_def_t inbitdeptharg =
460 ARG_DEF(NULL, "input-bit-depth", 1, "Bit depth of input");
463 static const struct arg_enum_list tune_content_enum[] = {
464 { "default", VP9E_CONTENT_DEFAULT },
465 { "screen", VP9E_CONTENT_SCREEN },
466 { "film", VP9E_CONTENT_FILM },
470 static const arg_def_t tune_content = ARG_DEF_ENUM(
471 NULL, "tune-content", 1, "Tune content type", tune_content_enum);
473 static const arg_def_t target_level = ARG_DEF(
474 NULL, "target-level", 1,
476 " 255: off (default)\n"
477 " 0: only keep level stats\n"
478 " 1: adaptively set alt-ref "
479 "distance and column tile limit based on picture size, and keep"
481 " 10: level 1.0 11: level 1.1 "
482 "... 62: level 6.2");
484 static const arg_def_t row_mt =
485 ARG_DEF(NULL, "row-mt", 1,
486 "Enable row based non-deterministic multi-threading in VP9");
489 #if CONFIG_VP9_ENCODER
490 static const arg_def_t *vp9_args[] = { &cpu_used_vp9,
505 &frame_parallel_decoding,
508 &frame_periodic_boost,
516 #if CONFIG_VP9_HIGHBITDEPTH
519 #endif // CONFIG_VP9_HIGHBITDEPTH
521 static const int vp9_arg_ctrl_map[] = { VP8E_SET_CPUUSED,
522 VP8E_SET_ENABLEAUTOALTREF,
524 VP8E_SET_STATIC_THRESHOLD,
525 VP9E_SET_TILE_COLUMNS,
527 VP8E_SET_ARNR_MAXFRAMES,
528 VP8E_SET_ARNR_STRENGTH,
532 VP8E_SET_MAX_INTRA_BITRATE_PCT,
533 VP9E_SET_MAX_INTER_BITRATE_PCT,
534 VP9E_SET_GF_CBR_BOOST_PCT,
536 VP9E_SET_FRAME_PARALLEL_DECODING,
539 VP9E_SET_FRAME_PERIODIC_BOOST,
540 VP9E_SET_NOISE_SENSITIVITY,
541 VP9E_SET_TUNE_CONTENT,
542 VP9E_SET_COLOR_SPACE,
543 VP9E_SET_MIN_GF_INTERVAL,
544 VP9E_SET_MAX_GF_INTERVAL,
545 VP9E_SET_TARGET_LEVEL,
550 static const arg_def_t *no_args[] = { NULL };
552 void usage_exit(void) {
554 const int num_encoder = get_vpx_encoder_count();
556 fprintf(stderr, "Usage: %s <options> -o dst_filename src_filename \n",
559 fprintf(stderr, "\nOptions:\n");
560 arg_show_usage(stderr, main_args);
561 fprintf(stderr, "\nEncoder Global Options:\n");
562 arg_show_usage(stderr, global_args);
563 fprintf(stderr, "\nRate Control Options:\n");
564 arg_show_usage(stderr, rc_args);
565 fprintf(stderr, "\nTwopass Rate Control Options:\n");
566 arg_show_usage(stderr, rc_twopass_args);
567 fprintf(stderr, "\nKeyframe Placement Options:\n");
568 arg_show_usage(stderr, kf_args);
569 #if CONFIG_VP8_ENCODER
570 fprintf(stderr, "\nVP8 Specific Options:\n");
571 arg_show_usage(stderr, vp8_args);
573 #if CONFIG_VP9_ENCODER
574 fprintf(stderr, "\nVP9 Specific Options:\n");
575 arg_show_usage(stderr, vp9_args);
578 "\nStream timebase (--timebase):\n"
579 " The desired precision of timestamps in the output, expressed\n"
580 " in fractional seconds. Default is 1/1000.\n");
581 fprintf(stderr, "\nIncluded encoders:\n\n");
583 for (i = 0; i < num_encoder; ++i) {
584 const VpxInterface *const encoder = get_vpx_encoder_by_index(i);
585 const char *defstr = (i == (num_encoder - 1)) ? "(default)" : "";
586 fprintf(stderr, " %-6s - %s %s\n", encoder->name,
587 vpx_codec_iface_name(encoder->codec_interface()), defstr);
589 fprintf(stderr, "\n ");
590 fprintf(stderr, "Use --codec to switch to a non-default encoder.\n\n");
595 #define mmin(a, b) ((a) < (b) ? (a) : (b))
597 #if CONFIG_VP9_HIGHBITDEPTH
598 static void find_mismatch_high(const vpx_image_t *const img1,
599 const vpx_image_t *const img2, int yloc[4],
600 int uloc[4], int vloc[4]) {
601 uint16_t *plane1, *plane2;
602 uint32_t stride1, stride2;
603 const uint32_t bsize = 64;
604 const uint32_t bsizey = bsize >> img1->y_chroma_shift;
605 const uint32_t bsizex = bsize >> img1->x_chroma_shift;
607 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
609 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
612 yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
613 plane1 = (uint16_t *)img1->planes[VPX_PLANE_Y];
614 plane2 = (uint16_t *)img2->planes[VPX_PLANE_Y];
615 stride1 = img1->stride[VPX_PLANE_Y] / 2;
616 stride2 = img2->stride[VPX_PLANE_Y] / 2;
617 for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
618 for (j = 0; match && j < img1->d_w; j += bsize) {
620 const int si = mmin(i + bsize, img1->d_h) - i;
621 const int sj = mmin(j + bsize, img1->d_w) - j;
622 for (k = 0; match && k < si; ++k) {
623 for (l = 0; match && l < sj; ++l) {
624 if (*(plane1 + (i + k) * stride1 + j + l) !=
625 *(plane2 + (i + k) * stride2 + j + l)) {
628 yloc[2] = *(plane1 + (i + k) * stride1 + j + l);
629 yloc[3] = *(plane2 + (i + k) * stride2 + j + l);
638 uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
639 plane1 = (uint16_t *)img1->planes[VPX_PLANE_U];
640 plane2 = (uint16_t *)img2->planes[VPX_PLANE_U];
641 stride1 = img1->stride[VPX_PLANE_U] / 2;
642 stride2 = img2->stride[VPX_PLANE_U] / 2;
643 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
644 for (j = 0; match && j < c_w; j += bsizex) {
646 const int si = mmin(i + bsizey, c_h - i);
647 const int sj = mmin(j + bsizex, c_w - j);
648 for (k = 0; match && k < si; ++k) {
649 for (l = 0; match && l < sj; ++l) {
650 if (*(plane1 + (i + k) * stride1 + j + l) !=
651 *(plane2 + (i + k) * stride2 + j + l)) {
654 uloc[2] = *(plane1 + (i + k) * stride1 + j + l);
655 uloc[3] = *(plane2 + (i + k) * stride2 + j + l);
664 vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
665 plane1 = (uint16_t *)img1->planes[VPX_PLANE_V];
666 plane2 = (uint16_t *)img2->planes[VPX_PLANE_V];
667 stride1 = img1->stride[VPX_PLANE_V] / 2;
668 stride2 = img2->stride[VPX_PLANE_V] / 2;
669 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
670 for (j = 0; match && j < c_w; j += bsizex) {
672 const int si = mmin(i + bsizey, c_h - i);
673 const int sj = mmin(j + bsizex, c_w - j);
674 for (k = 0; match && k < si; ++k) {
675 for (l = 0; match && l < sj; ++l) {
676 if (*(plane1 + (i + k) * stride1 + j + l) !=
677 *(plane2 + (i + k) * stride2 + j + l)) {
680 vloc[2] = *(plane1 + (i + k) * stride1 + j + l);
681 vloc[3] = *(plane2 + (i + k) * stride2 + j + l);
692 static void find_mismatch(const vpx_image_t *const img1,
693 const vpx_image_t *const img2, int yloc[4],
694 int uloc[4], int vloc[4]) {
695 const uint32_t bsize = 64;
696 const uint32_t bsizey = bsize >> img1->y_chroma_shift;
697 const uint32_t bsizex = bsize >> img1->x_chroma_shift;
699 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
701 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
704 yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
705 for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
706 for (j = 0; match && j < img1->d_w; j += bsize) {
708 const int si = mmin(i + bsize, img1->d_h) - i;
709 const int sj = mmin(j + bsize, img1->d_w) - j;
710 for (k = 0; match && k < si; ++k) {
711 for (l = 0; match && l < sj; ++l) {
712 if (*(img1->planes[VPX_PLANE_Y] +
713 (i + k) * img1->stride[VPX_PLANE_Y] + j + l) !=
714 *(img2->planes[VPX_PLANE_Y] +
715 (i + k) * img2->stride[VPX_PLANE_Y] + j + l)) {
718 yloc[2] = *(img1->planes[VPX_PLANE_Y] +
719 (i + k) * img1->stride[VPX_PLANE_Y] + j + l);
720 yloc[3] = *(img2->planes[VPX_PLANE_Y] +
721 (i + k) * img2->stride[VPX_PLANE_Y] + j + l);
730 uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
731 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
732 for (j = 0; match && j < c_w; j += bsizex) {
734 const int si = mmin(i + bsizey, c_h - i);
735 const int sj = mmin(j + bsizex, c_w - j);
736 for (k = 0; match && k < si; ++k) {
737 for (l = 0; match && l < sj; ++l) {
738 if (*(img1->planes[VPX_PLANE_U] +
739 (i + k) * img1->stride[VPX_PLANE_U] + j + l) !=
740 *(img2->planes[VPX_PLANE_U] +
741 (i + k) * img2->stride[VPX_PLANE_U] + j + l)) {
744 uloc[2] = *(img1->planes[VPX_PLANE_U] +
745 (i + k) * img1->stride[VPX_PLANE_U] + j + l);
746 uloc[3] = *(img2->planes[VPX_PLANE_U] +
747 (i + k) * img2->stride[VPX_PLANE_U] + j + l);
755 vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
756 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
757 for (j = 0; match && j < c_w; j += bsizex) {
759 const int si = mmin(i + bsizey, c_h - i);
760 const int sj = mmin(j + bsizex, c_w - j);
761 for (k = 0; match && k < si; ++k) {
762 for (l = 0; match && l < sj; ++l) {
763 if (*(img1->planes[VPX_PLANE_V] +
764 (i + k) * img1->stride[VPX_PLANE_V] + j + l) !=
765 *(img2->planes[VPX_PLANE_V] +
766 (i + k) * img2->stride[VPX_PLANE_V] + j + l)) {
769 vloc[2] = *(img1->planes[VPX_PLANE_V] +
770 (i + k) * img1->stride[VPX_PLANE_V] + j + l);
771 vloc[3] = *(img2->planes[VPX_PLANE_V] +
772 (i + k) * img2->stride[VPX_PLANE_V] + j + l);
782 static int compare_img(const vpx_image_t *const img1,
783 const vpx_image_t *const img2) {
784 uint32_t l_w = img1->d_w;
785 uint32_t c_w = (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
787 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
791 match &= (img1->fmt == img2->fmt);
792 match &= (img1->d_w == img2->d_w);
793 match &= (img1->d_h == img2->d_h);
794 #if CONFIG_VP9_HIGHBITDEPTH
795 if (img1->fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
801 for (i = 0; i < img1->d_h; ++i)
802 match &= (memcmp(img1->planes[VPX_PLANE_Y] + i * img1->stride[VPX_PLANE_Y],
803 img2->planes[VPX_PLANE_Y] + i * img2->stride[VPX_PLANE_Y],
806 for (i = 0; i < c_h; ++i)
807 match &= (memcmp(img1->planes[VPX_PLANE_U] + i * img1->stride[VPX_PLANE_U],
808 img2->planes[VPX_PLANE_U] + i * img2->stride[VPX_PLANE_U],
811 for (i = 0; i < c_h; ++i)
812 match &= (memcmp(img1->planes[VPX_PLANE_V] + i * img1->stride[VPX_PLANE_V],
813 img2->planes[VPX_PLANE_V] + i * img2->stride[VPX_PLANE_V],
819 #define NELEMENTS(x) (sizeof(x) / sizeof(x[0]))
820 #if CONFIG_VP9_ENCODER
821 #define ARG_CTRL_CNT_MAX NELEMENTS(vp9_arg_ctrl_map)
823 #define ARG_CTRL_CNT_MAX NELEMENTS(vp8_arg_ctrl_map)
827 typedef int stereo_format_t;
828 struct WebmOutputContext {
833 /* Per-stream configuration */
834 struct stream_config {
835 struct vpx_codec_enc_cfg cfg;
837 const char *stats_fn;
838 #if CONFIG_FP_MB_STATS
839 const char *fpmb_stats_fn;
841 stereo_format_t stereo_fmt;
842 int arg_ctrls[ARG_CTRL_CNT_MAX][2];
845 #if CONFIG_VP9_HIGHBITDEPTH
846 // whether to use 16bit internal buffers
847 int use_16bit_internal;
851 struct stream_state {
853 struct stream_state *next;
854 struct stream_config config;
856 struct rate_hist *rate_hist;
857 struct WebmOutputContext webm_ctx;
858 uint64_t psnr_sse_total;
859 uint64_t psnr_samples_total;
860 double psnr_totals[4];
863 vpx_codec_ctx_t encoder;
864 unsigned int frames_out;
868 #if CONFIG_FP_MB_STATS
869 stats_io_t fpmb_stats;
871 struct vpx_image *img;
872 vpx_codec_ctx_t decoder;
876 static void validate_positive_rational(const char *msg,
877 struct vpx_rational *rat) {
883 if (rat->num < 0) die("Error: %s must be positive\n", msg);
885 if (!rat->den) die("Error: %s has zero denominator\n", msg);
888 static void parse_global_config(struct VpxEncoderConfig *global, char **argv) {
891 const int num_encoder = get_vpx_encoder_count();
893 if (num_encoder < 1) die("Error: no valid encoder available\n");
895 /* Initialize default parameters */
896 memset(global, 0, sizeof(*global));
897 global->codec = get_vpx_encoder_by_index(num_encoder - 1);
899 global->color_type = I420;
900 /* Assign default deadline to good quality */
901 global->deadline = VPX_DL_GOOD_QUALITY;
903 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
906 if (arg_match(&arg, &codecarg, argi)) {
907 global->codec = get_vpx_encoder_by_name(arg.val);
909 die("Error: Unrecognized argument (%s) to --codec\n", arg.val);
910 } else if (arg_match(&arg, &passes, argi)) {
911 global->passes = arg_parse_uint(&arg);
913 if (global->passes < 1 || global->passes > 2)
914 die("Error: Invalid number of passes (%d)\n", global->passes);
915 } else if (arg_match(&arg, &pass_arg, argi)) {
916 global->pass = arg_parse_uint(&arg);
918 if (global->pass < 1 || global->pass > 2)
919 die("Error: Invalid pass selected (%d)\n", global->pass);
920 } else if (arg_match(&arg, &usage, argi))
921 global->usage = arg_parse_uint(&arg);
922 else if (arg_match(&arg, &deadline, argi))
923 global->deadline = arg_parse_uint(&arg);
924 else if (arg_match(&arg, &best_dl, argi))
925 global->deadline = VPX_DL_BEST_QUALITY;
926 else if (arg_match(&arg, &good_dl, argi))
927 global->deadline = VPX_DL_GOOD_QUALITY;
928 else if (arg_match(&arg, &rt_dl, argi))
929 global->deadline = VPX_DL_REALTIME;
930 else if (arg_match(&arg, &use_yv12, argi))
931 global->color_type = YV12;
932 else if (arg_match(&arg, &use_i420, argi))
933 global->color_type = I420;
934 else if (arg_match(&arg, &use_i422, argi))
935 global->color_type = I422;
936 else if (arg_match(&arg, &use_i444, argi))
937 global->color_type = I444;
938 else if (arg_match(&arg, &use_i440, argi))
939 global->color_type = I440;
940 else if (arg_match(&arg, &quietarg, argi))
942 else if (arg_match(&arg, &verbosearg, argi))
944 else if (arg_match(&arg, &limit, argi))
945 global->limit = arg_parse_uint(&arg);
946 else if (arg_match(&arg, &skip, argi))
947 global->skip_frames = arg_parse_uint(&arg);
948 else if (arg_match(&arg, &psnrarg, argi))
949 global->show_psnr = 1;
950 else if (arg_match(&arg, &recontest, argi))
951 global->test_decode = arg_parse_enum_or_int(&arg);
952 else if (arg_match(&arg, &framerate, argi)) {
953 global->framerate = arg_parse_rational(&arg);
954 validate_positive_rational(arg.name, &global->framerate);
955 global->have_framerate = 1;
956 } else if (arg_match(&arg, &out_part, argi))
957 global->out_part = 1;
958 else if (arg_match(&arg, &debugmode, argi))
960 else if (arg_match(&arg, &q_hist_n, argi))
961 global->show_q_hist_buckets = arg_parse_uint(&arg);
962 else if (arg_match(&arg, &rate_hist_n, argi))
963 global->show_rate_hist_buckets = arg_parse_uint(&arg);
964 else if (arg_match(&arg, &disable_warnings, argi))
965 global->disable_warnings = 1;
966 else if (arg_match(&arg, &disable_warning_prompt, argi))
967 global->disable_warning_prompt = 1;
973 /* DWIM: Assume the user meant passes=2 if pass=2 is specified */
974 if (global->pass > global->passes) {
975 warn("Assuming --pass=%d implies --passes=%d\n", global->pass,
977 global->passes = global->pass;
980 /* Validate global config */
981 if (global->passes == 0) {
982 #if CONFIG_VP9_ENCODER
983 // Make default VP9 passes = 2 until there is a better quality 1-pass
985 if (global->codec != NULL && global->codec->name != NULL)
986 global->passes = (strcmp(global->codec->name, "vp9") == 0 &&
987 global->deadline != VPX_DL_REALTIME)
995 if (global->deadline == VPX_DL_REALTIME && global->passes > 1) {
996 warn("Enforcing one-pass encoding in realtime mode\n");
1001 static void open_input_file(struct VpxInputContext *input) {
1002 /* Parse certain options from the input file, if possible */
1003 input->file = strcmp(input->filename, "-") ? fopen(input->filename, "rb")
1004 : set_binary_mode(stdin);
1006 if (!input->file) fatal("Failed to open input file");
1008 if (!fseeko(input->file, 0, SEEK_END)) {
1009 /* Input file is seekable. Figure out how long it is, so we can get
1012 input->length = ftello(input->file);
1013 rewind(input->file);
1016 /* Default to 1:1 pixel aspect ratio. */
1017 input->pixel_aspect_ratio.numerator = 1;
1018 input->pixel_aspect_ratio.denominator = 1;
1020 /* For RAW input sources, these bytes will applied on the first frame
1023 input->detect.buf_read = fread(input->detect.buf, 1, 4, input->file);
1024 input->detect.position = 0;
1026 if (input->detect.buf_read == 4 && file_is_y4m(input->detect.buf)) {
1027 if (y4m_input_open(&input->y4m, input->file, input->detect.buf, 4,
1028 input->only_i420) >= 0) {
1029 input->file_type = FILE_TYPE_Y4M;
1030 input->width = input->y4m.pic_w;
1031 input->height = input->y4m.pic_h;
1032 input->pixel_aspect_ratio.numerator = input->y4m.par_n;
1033 input->pixel_aspect_ratio.denominator = input->y4m.par_d;
1034 input->framerate.numerator = input->y4m.fps_n;
1035 input->framerate.denominator = input->y4m.fps_d;
1036 input->fmt = input->y4m.vpx_fmt;
1037 input->bit_depth = input->y4m.bit_depth;
1039 fatal("Unsupported Y4M stream.");
1040 } else if (input->detect.buf_read == 4 && fourcc_is_ivf(input->detect.buf)) {
1041 fatal("IVF is not supported as input.");
1043 input->file_type = FILE_TYPE_RAW;
1047 static void close_input_file(struct VpxInputContext *input) {
1048 fclose(input->file);
1049 if (input->file_type == FILE_TYPE_Y4M) y4m_input_close(&input->y4m);
1052 static struct stream_state *new_stream(struct VpxEncoderConfig *global,
1053 struct stream_state *prev) {
1054 struct stream_state *stream;
1056 stream = calloc(1, sizeof(*stream));
1057 if (stream == NULL) {
1058 fatal("Failed to allocate new stream.");
1062 memcpy(stream, prev, sizeof(*stream));
1064 prev->next = stream;
1066 vpx_codec_err_t res;
1068 /* Populate encoder configuration */
1069 res = vpx_codec_enc_config_default(global->codec->codec_interface(),
1070 &stream->config.cfg, global->usage);
1071 if (res) fatal("Failed to get config: %s\n", vpx_codec_err_to_string(res));
1073 /* Change the default timebase to a high enough value so that the
1074 * encoder will always create strictly increasing timestamps.
1076 stream->config.cfg.g_timebase.den = 1000;
1078 /* Never use the library's default resolution, require it be parsed
1079 * from the file or set on the command line.
1081 stream->config.cfg.g_w = 0;
1082 stream->config.cfg.g_h = 0;
1084 /* Initialize remaining stream parameters */
1085 stream->config.write_webm = 1;
1087 stream->config.stereo_fmt = STEREO_FORMAT_MONO;
1088 stream->webm_ctx.last_pts_ns = -1;
1089 stream->webm_ctx.writer = NULL;
1090 stream->webm_ctx.segment = NULL;
1093 /* Allows removal of the application version from the EBML tags */
1094 stream->webm_ctx.debug = global->debug;
1096 /* Default lag_in_frames is 0 in realtime mode CBR mode*/
1097 if (global->deadline == VPX_DL_REALTIME &&
1098 stream->config.cfg.rc_end_usage == 1)
1099 stream->config.cfg.g_lag_in_frames = 0;
1102 /* Output files must be specified for each stream */
1103 stream->config.out_fn = NULL;
1105 stream->next = NULL;
1109 static int parse_stream_params(struct VpxEncoderConfig *global,
1110 struct stream_state *stream, char **argv) {
1111 char **argi, **argj;
1113 static const arg_def_t **ctrl_args = no_args;
1114 static const int *ctrl_args_map = NULL;
1115 struct stream_config *config = &stream->config;
1116 int eos_mark_found = 0;
1117 #if CONFIG_VP9_HIGHBITDEPTH
1118 int test_16bit_internal = 0;
1121 // Handle codec specific options
1123 #if CONFIG_VP8_ENCODER
1124 } else if (strcmp(global->codec->name, "vp8") == 0) {
1125 ctrl_args = vp8_args;
1126 ctrl_args_map = vp8_arg_ctrl_map;
1128 #if CONFIG_VP9_ENCODER
1129 } else if (strcmp(global->codec->name, "vp9") == 0) {
1130 ctrl_args = vp9_args;
1131 ctrl_args_map = vp9_arg_ctrl_map;
1135 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
1138 /* Once we've found an end-of-stream marker (--) we want to continue
1139 * shifting arguments but not consuming them.
1141 if (eos_mark_found) {
1144 } else if (!strcmp(*argj, "--")) {
1149 if (arg_match(&arg, &outputfile, argi)) {
1150 config->out_fn = arg.val;
1151 } else if (arg_match(&arg, &fpf_name, argi)) {
1152 config->stats_fn = arg.val;
1153 #if CONFIG_FP_MB_STATS
1154 } else if (arg_match(&arg, &fpmbf_name, argi)) {
1155 config->fpmb_stats_fn = arg.val;
1157 } else if (arg_match(&arg, &use_webm, argi)) {
1159 config->write_webm = 1;
1161 die("Error: --webm specified but webm is disabled.");
1163 } else if (arg_match(&arg, &use_ivf, argi)) {
1164 config->write_webm = 0;
1165 } else if (arg_match(&arg, &threads, argi)) {
1166 config->cfg.g_threads = arg_parse_uint(&arg);
1167 } else if (arg_match(&arg, &profile, argi)) {
1168 config->cfg.g_profile = arg_parse_uint(&arg);
1169 } else if (arg_match(&arg, &width, argi)) {
1170 config->cfg.g_w = arg_parse_uint(&arg);
1171 } else if (arg_match(&arg, &height, argi)) {
1172 config->cfg.g_h = arg_parse_uint(&arg);
1173 #if CONFIG_VP9_HIGHBITDEPTH
1174 } else if (arg_match(&arg, &bitdeptharg, argi)) {
1175 config->cfg.g_bit_depth = arg_parse_enum_or_int(&arg);
1176 } else if (arg_match(&arg, &inbitdeptharg, argi)) {
1177 config->cfg.g_input_bit_depth = arg_parse_uint(&arg);
1180 } else if (arg_match(&arg, &stereo_mode, argi)) {
1181 config->stereo_fmt = arg_parse_enum_or_int(&arg);
1183 } else if (arg_match(&arg, &timebase, argi)) {
1184 config->cfg.g_timebase = arg_parse_rational(&arg);
1185 validate_positive_rational(arg.name, &config->cfg.g_timebase);
1186 } else if (arg_match(&arg, &error_resilient, argi)) {
1187 config->cfg.g_error_resilient = arg_parse_uint(&arg);
1188 } else if (arg_match(&arg, &end_usage, argi)) {
1189 config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
1190 } else if (arg_match(&arg, &lag_in_frames, argi)) {
1191 config->cfg.g_lag_in_frames = arg_parse_uint(&arg);
1192 if (global->deadline == VPX_DL_REALTIME &&
1193 config->cfg.rc_end_usage == VPX_CBR &&
1194 config->cfg.g_lag_in_frames != 0) {
1195 warn("non-zero %s option ignored in realtime CBR mode.\n", arg.name);
1196 config->cfg.g_lag_in_frames = 0;
1198 } else if (arg_match(&arg, &dropframe_thresh, argi)) {
1199 config->cfg.rc_dropframe_thresh = arg_parse_uint(&arg);
1200 } else if (arg_match(&arg, &resize_allowed, argi)) {
1201 config->cfg.rc_resize_allowed = arg_parse_uint(&arg);
1202 } else if (arg_match(&arg, &resize_width, argi)) {
1203 config->cfg.rc_scaled_width = arg_parse_uint(&arg);
1204 } else if (arg_match(&arg, &resize_height, argi)) {
1205 config->cfg.rc_scaled_height = arg_parse_uint(&arg);
1206 } else if (arg_match(&arg, &resize_up_thresh, argi)) {
1207 config->cfg.rc_resize_up_thresh = arg_parse_uint(&arg);
1208 } else if (arg_match(&arg, &resize_down_thresh, argi)) {
1209 config->cfg.rc_resize_down_thresh = arg_parse_uint(&arg);
1210 } else if (arg_match(&arg, &end_usage, argi)) {
1211 config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
1212 } else if (arg_match(&arg, &target_bitrate, argi)) {
1213 config->cfg.rc_target_bitrate = arg_parse_uint(&arg);
1214 } else if (arg_match(&arg, &min_quantizer, argi)) {
1215 config->cfg.rc_min_quantizer = arg_parse_uint(&arg);
1216 } else if (arg_match(&arg, &max_quantizer, argi)) {
1217 config->cfg.rc_max_quantizer = arg_parse_uint(&arg);
1218 } else if (arg_match(&arg, &undershoot_pct, argi)) {
1219 config->cfg.rc_undershoot_pct = arg_parse_uint(&arg);
1220 } else if (arg_match(&arg, &overshoot_pct, argi)) {
1221 config->cfg.rc_overshoot_pct = arg_parse_uint(&arg);
1222 } else if (arg_match(&arg, &buf_sz, argi)) {
1223 config->cfg.rc_buf_sz = arg_parse_uint(&arg);
1224 } else if (arg_match(&arg, &buf_initial_sz, argi)) {
1225 config->cfg.rc_buf_initial_sz = arg_parse_uint(&arg);
1226 } else if (arg_match(&arg, &buf_optimal_sz, argi)) {
1227 config->cfg.rc_buf_optimal_sz = arg_parse_uint(&arg);
1228 } else if (arg_match(&arg, &bias_pct, argi)) {
1229 config->cfg.rc_2pass_vbr_bias_pct = arg_parse_uint(&arg);
1230 if (global->passes < 2)
1231 warn("option %s ignored in one-pass mode.\n", arg.name);
1232 } else if (arg_match(&arg, &minsection_pct, argi)) {
1233 config->cfg.rc_2pass_vbr_minsection_pct = arg_parse_uint(&arg);
1235 if (global->passes < 2)
1236 warn("option %s ignored in one-pass mode.\n", arg.name);
1237 } else if (arg_match(&arg, &maxsection_pct, argi)) {
1238 config->cfg.rc_2pass_vbr_maxsection_pct = arg_parse_uint(&arg);
1240 if (global->passes < 2)
1241 warn("option %s ignored in one-pass mode.\n", arg.name);
1242 } else if (arg_match(&arg, &corpus_complexity, argi)) {
1243 config->cfg.rc_2pass_vbr_corpus_complexity = arg_parse_uint(&arg);
1245 if (global->passes < 2)
1246 warn("option %s ignored in one-pass mode.\n", arg.name);
1247 } else if (arg_match(&arg, &kf_min_dist, argi)) {
1248 config->cfg.kf_min_dist = arg_parse_uint(&arg);
1249 } else if (arg_match(&arg, &kf_max_dist, argi)) {
1250 config->cfg.kf_max_dist = arg_parse_uint(&arg);
1251 } else if (arg_match(&arg, &kf_disabled, argi)) {
1252 config->cfg.kf_mode = VPX_KF_DISABLED;
1253 #if CONFIG_VP9_HIGHBITDEPTH
1254 } else if (arg_match(&arg, &test16bitinternalarg, argi)) {
1255 if (strcmp(global->codec->name, "vp9") == 0) {
1256 test_16bit_internal = 1;
1261 for (i = 0; ctrl_args[i]; i++) {
1262 if (arg_match(&arg, ctrl_args[i], argi)) {
1266 /* Point either to the next free element or the first
1267 * instance of this control.
1269 for (j = 0; j < config->arg_ctrl_cnt; j++)
1270 if (ctrl_args_map != NULL &&
1271 config->arg_ctrls[j][0] == ctrl_args_map[i])
1275 assert(j < (int)ARG_CTRL_CNT_MAX);
1276 if (ctrl_args_map != NULL && j < (int)ARG_CTRL_CNT_MAX) {
1277 config->arg_ctrls[j][0] = ctrl_args_map[i];
1278 config->arg_ctrls[j][1] = arg_parse_enum_or_int(&arg);
1279 if (j == config->arg_ctrl_cnt) config->arg_ctrl_cnt++;
1286 #if CONFIG_VP9_HIGHBITDEPTH
1287 if (strcmp(global->codec->name, "vp9") == 0) {
1288 config->use_16bit_internal =
1289 test_16bit_internal | (config->cfg.g_profile > 1);
1292 return eos_mark_found;
1295 #define FOREACH_STREAM(func) \
1297 struct stream_state *stream; \
1298 for (stream = streams; stream; stream = stream->next) { \
1303 static void validate_stream_config(const struct stream_state *stream,
1304 const struct VpxEncoderConfig *global) {
1305 const struct stream_state *streami;
1308 if (!stream->config.cfg.g_w || !stream->config.cfg.g_h)
1310 "Stream %d: Specify stream dimensions with --width (-w) "
1311 " and --height (-h)",
1314 // Check that the codec bit depth is greater than the input bit depth.
1315 if (stream->config.cfg.g_input_bit_depth >
1316 (unsigned int)stream->config.cfg.g_bit_depth) {
1317 fatal("Stream %d: codec bit depth (%d) less than input bit depth (%d)",
1318 stream->index, (int)stream->config.cfg.g_bit_depth,
1319 stream->config.cfg.g_input_bit_depth);
1322 for (streami = stream; streami; streami = streami->next) {
1323 /* All streams require output files */
1324 if (!streami->config.out_fn)
1325 fatal("Stream %d: Output file is required (specify with -o)",
1328 /* Check for two streams outputting to the same file */
1329 if (streami != stream) {
1330 const char *a = stream->config.out_fn;
1331 const char *b = streami->config.out_fn;
1332 if (!strcmp(a, b) && strcmp(a, "/dev/null") && strcmp(a, ":nul"))
1333 fatal("Stream %d: duplicate output file (from stream %d)",
1334 streami->index, stream->index);
1337 /* Check for two streams sharing a stats file. */
1338 if (streami != stream) {
1339 const char *a = stream->config.stats_fn;
1340 const char *b = streami->config.stats_fn;
1341 if (a && b && !strcmp(a, b))
1342 fatal("Stream %d: duplicate stats file (from stream %d)",
1343 streami->index, stream->index);
1346 #if CONFIG_FP_MB_STATS
1347 /* Check for two streams sharing a mb stats file. */
1348 if (streami != stream) {
1349 const char *a = stream->config.fpmb_stats_fn;
1350 const char *b = streami->config.fpmb_stats_fn;
1351 if (a && b && !strcmp(a, b))
1352 fatal("Stream %d: duplicate mb stats file (from stream %d)",
1353 streami->index, stream->index);
1359 static void set_stream_dimensions(struct stream_state *stream, unsigned int w,
1361 if (!stream->config.cfg.g_w) {
1362 if (!stream->config.cfg.g_h)
1363 stream->config.cfg.g_w = w;
1365 stream->config.cfg.g_w = w * stream->config.cfg.g_h / h;
1367 if (!stream->config.cfg.g_h) {
1368 stream->config.cfg.g_h = h * stream->config.cfg.g_w / w;
1372 static const char *file_type_to_string(enum VideoFileType t) {
1374 case FILE_TYPE_RAW: return "RAW";
1375 case FILE_TYPE_Y4M: return "Y4M";
1376 default: return "Other";
1380 static const char *image_format_to_string(vpx_img_fmt_t f) {
1382 case VPX_IMG_FMT_I420: return "I420";
1383 case VPX_IMG_FMT_I422: return "I422";
1384 case VPX_IMG_FMT_I444: return "I444";
1385 case VPX_IMG_FMT_I440: return "I440";
1386 case VPX_IMG_FMT_YV12: return "YV12";
1387 case VPX_IMG_FMT_I42016: return "I42016";
1388 case VPX_IMG_FMT_I42216: return "I42216";
1389 case VPX_IMG_FMT_I44416: return "I44416";
1390 case VPX_IMG_FMT_I44016: return "I44016";
1391 default: return "Other";
1395 static void show_stream_config(struct stream_state *stream,
1396 struct VpxEncoderConfig *global,
1397 struct VpxInputContext *input) {
1398 #define SHOW(field) \
1399 fprintf(stderr, " %-28s = %d\n", #field, stream->config.cfg.field)
1401 if (stream->index == 0) {
1402 fprintf(stderr, "Codec: %s\n",
1403 vpx_codec_iface_name(global->codec->codec_interface()));
1404 fprintf(stderr, "Source file: %s File Type: %s Format: %s\n",
1405 input->filename, file_type_to_string(input->file_type),
1406 image_format_to_string(input->fmt));
1408 if (stream->next || stream->index)
1409 fprintf(stderr, "\nStream Index: %d\n", stream->index);
1410 fprintf(stderr, "Destination file: %s\n", stream->config.out_fn);
1411 fprintf(stderr, "Encoder parameters:\n");
1419 SHOW(g_input_bit_depth);
1420 SHOW(g_timebase.num);
1421 SHOW(g_timebase.den);
1422 SHOW(g_error_resilient);
1424 SHOW(g_lag_in_frames);
1425 SHOW(rc_dropframe_thresh);
1426 SHOW(rc_resize_allowed);
1427 SHOW(rc_scaled_width);
1428 SHOW(rc_scaled_height);
1429 SHOW(rc_resize_up_thresh);
1430 SHOW(rc_resize_down_thresh);
1432 SHOW(rc_target_bitrate);
1433 SHOW(rc_min_quantizer);
1434 SHOW(rc_max_quantizer);
1435 SHOW(rc_undershoot_pct);
1436 SHOW(rc_overshoot_pct);
1438 SHOW(rc_buf_initial_sz);
1439 SHOW(rc_buf_optimal_sz);
1440 SHOW(rc_2pass_vbr_bias_pct);
1441 SHOW(rc_2pass_vbr_minsection_pct);
1442 SHOW(rc_2pass_vbr_maxsection_pct);
1443 SHOW(rc_2pass_vbr_corpus_complexity);
1449 static void open_output_file(struct stream_state *stream,
1450 struct VpxEncoderConfig *global,
1451 const struct VpxRational *pixel_aspect_ratio) {
1452 const char *fn = stream->config.out_fn;
1453 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1455 if (cfg->g_pass == VPX_RC_FIRST_PASS) return;
1457 stream->file = strcmp(fn, "-") ? fopen(fn, "wb") : set_binary_mode(stdout);
1459 if (!stream->file) fatal("Failed to open output file");
1461 if (stream->config.write_webm && fseek(stream->file, 0, SEEK_CUR))
1462 fatal("WebM output to pipes not supported.");
1465 if (stream->config.write_webm) {
1466 stream->webm_ctx.stream = stream->file;
1467 write_webm_file_header(&stream->webm_ctx, cfg, stream->config.stereo_fmt,
1468 global->codec->fourcc, pixel_aspect_ratio);
1471 (void)pixel_aspect_ratio;
1474 if (!stream->config.write_webm) {
1475 ivf_write_file_header(stream->file, cfg, global->codec->fourcc, 0);
1479 static void close_output_file(struct stream_state *stream,
1480 unsigned int fourcc) {
1481 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1483 if (cfg->g_pass == VPX_RC_FIRST_PASS) return;
1486 if (stream->config.write_webm) {
1487 write_webm_file_footer(&stream->webm_ctx);
1491 if (!stream->config.write_webm) {
1492 if (!fseek(stream->file, 0, SEEK_SET))
1493 ivf_write_file_header(stream->file, &stream->config.cfg, fourcc,
1494 stream->frames_out);
1497 fclose(stream->file);
1500 static void setup_pass(struct stream_state *stream,
1501 struct VpxEncoderConfig *global, int pass) {
1502 if (stream->config.stats_fn) {
1503 if (!stats_open_file(&stream->stats, stream->config.stats_fn, pass))
1504 fatal("Failed to open statistics store");
1506 if (!stats_open_mem(&stream->stats, pass))
1507 fatal("Failed to open statistics store");
1510 #if CONFIG_FP_MB_STATS
1511 if (stream->config.fpmb_stats_fn) {
1512 if (!stats_open_file(&stream->fpmb_stats, stream->config.fpmb_stats_fn,
1514 fatal("Failed to open mb statistics store");
1516 if (!stats_open_mem(&stream->fpmb_stats, pass))
1517 fatal("Failed to open mb statistics store");
1521 stream->config.cfg.g_pass = global->passes == 2
1522 ? pass ? VPX_RC_LAST_PASS : VPX_RC_FIRST_PASS
1525 stream->config.cfg.rc_twopass_stats_in = stats_get(&stream->stats);
1526 #if CONFIG_FP_MB_STATS
1527 stream->config.cfg.rc_firstpass_mb_stats_in =
1528 stats_get(&stream->fpmb_stats);
1532 stream->cx_time = 0;
1534 stream->frames_out = 0;
1537 static void initialize_encoder(struct stream_state *stream,
1538 struct VpxEncoderConfig *global) {
1542 flags |= global->show_psnr ? VPX_CODEC_USE_PSNR : 0;
1543 flags |= global->out_part ? VPX_CODEC_USE_OUTPUT_PARTITION : 0;
1544 #if CONFIG_VP9_HIGHBITDEPTH
1545 flags |= stream->config.use_16bit_internal ? VPX_CODEC_USE_HIGHBITDEPTH : 0;
1548 /* Construct Encoder Context */
1549 vpx_codec_enc_init(&stream->encoder, global->codec->codec_interface(),
1550 &stream->config.cfg, flags);
1551 ctx_exit_on_error(&stream->encoder, "Failed to initialize encoder");
1553 /* Note that we bypass the vpx_codec_control wrapper macro because
1554 * we're being clever to store the control IDs in an array. Real
1555 * applications will want to make use of the enumerations directly
1557 for (i = 0; i < stream->config.arg_ctrl_cnt; i++) {
1558 int ctrl = stream->config.arg_ctrls[i][0];
1559 int value = stream->config.arg_ctrls[i][1];
1560 if (vpx_codec_control_(&stream->encoder, ctrl, value))
1561 fprintf(stderr, "Error: Tried to set control %d = %d\n", ctrl, value);
1563 ctx_exit_on_error(&stream->encoder, "Failed to control codec");
1567 if (global->test_decode != TEST_DECODE_OFF) {
1568 const VpxInterface *decoder = get_vpx_decoder_by_name(global->codec->name);
1569 vpx_codec_dec_init(&stream->decoder, decoder->codec_interface(), NULL, 0);
1574 static void encode_frame(struct stream_state *stream,
1575 struct VpxEncoderConfig *global, struct vpx_image *img,
1576 unsigned int frames_in) {
1577 vpx_codec_pts_t frame_start, next_frame_start;
1578 struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1579 struct vpx_usec_timer timer;
1582 (cfg->g_timebase.den * (int64_t)(frames_in - 1) * global->framerate.den) /
1583 cfg->g_timebase.num / global->framerate.num;
1585 (cfg->g_timebase.den * (int64_t)(frames_in)*global->framerate.den) /
1586 cfg->g_timebase.num / global->framerate.num;
1588 /* Scale if necessary */
1589 #if CONFIG_VP9_HIGHBITDEPTH
1591 if ((img->fmt & VPX_IMG_FMT_HIGHBITDEPTH) &&
1592 (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1593 if (img->fmt != VPX_IMG_FMT_I42016) {
1594 fprintf(stderr, "%s can only scale 4:2:0 inputs\n", exec_name);
1600 vpx_img_alloc(NULL, VPX_IMG_FMT_I42016, cfg->g_w, cfg->g_h, 16);
1603 (uint16 *)img->planes[VPX_PLANE_Y], img->stride[VPX_PLANE_Y] / 2,
1604 (uint16 *)img->planes[VPX_PLANE_U], img->stride[VPX_PLANE_U] / 2,
1605 (uint16 *)img->planes[VPX_PLANE_V], img->stride[VPX_PLANE_V] / 2,
1606 img->d_w, img->d_h, (uint16 *)stream->img->planes[VPX_PLANE_Y],
1607 stream->img->stride[VPX_PLANE_Y] / 2,
1608 (uint16 *)stream->img->planes[VPX_PLANE_U],
1609 stream->img->stride[VPX_PLANE_U] / 2,
1610 (uint16 *)stream->img->planes[VPX_PLANE_V],
1611 stream->img->stride[VPX_PLANE_V] / 2, stream->img->d_w,
1612 stream->img->d_h, kFilterBox);
1615 stream->encoder.err = 1;
1616 ctx_exit_on_error(&stream->encoder,
1617 "Stream %d: Failed to encode frame.\n"
1618 "Scaling disabled in this configuration. \n"
1619 "To enable, configure with --enable-libyuv\n",
1625 if (img && (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1626 if (img->fmt != VPX_IMG_FMT_I420 && img->fmt != VPX_IMG_FMT_YV12) {
1627 fprintf(stderr, "%s can only scale 4:2:0 8bpp inputs\n", exec_name);
1633 vpx_img_alloc(NULL, VPX_IMG_FMT_I420, cfg->g_w, cfg->g_h, 16);
1635 img->planes[VPX_PLANE_Y], img->stride[VPX_PLANE_Y],
1636 img->planes[VPX_PLANE_U], img->stride[VPX_PLANE_U],
1637 img->planes[VPX_PLANE_V], img->stride[VPX_PLANE_V], img->d_w, img->d_h,
1638 stream->img->planes[VPX_PLANE_Y], stream->img->stride[VPX_PLANE_Y],
1639 stream->img->planes[VPX_PLANE_U], stream->img->stride[VPX_PLANE_U],
1640 stream->img->planes[VPX_PLANE_V], stream->img->stride[VPX_PLANE_V],
1641 stream->img->d_w, stream->img->d_h, kFilterBox);
1644 stream->encoder.err = 1;
1645 ctx_exit_on_error(&stream->encoder,
1646 "Stream %d: Failed to encode frame.\n"
1647 "Scaling disabled in this configuration. \n"
1648 "To enable, configure with --enable-libyuv\n",
1653 vpx_usec_timer_start(&timer);
1654 vpx_codec_encode(&stream->encoder, img, frame_start,
1655 (unsigned long)(next_frame_start - frame_start), 0,
1657 vpx_usec_timer_mark(&timer);
1658 stream->cx_time += vpx_usec_timer_elapsed(&timer);
1659 ctx_exit_on_error(&stream->encoder, "Stream %d: Failed to encode frame",
1663 static void update_quantizer_histogram(struct stream_state *stream) {
1664 if (stream->config.cfg.g_pass != VPX_RC_FIRST_PASS) {
1667 vpx_codec_control(&stream->encoder, VP8E_GET_LAST_QUANTIZER_64, &q);
1668 ctx_exit_on_error(&stream->encoder, "Failed to read quantizer");
1669 stream->counts[q]++;
1673 static void get_cx_data(struct stream_state *stream,
1674 struct VpxEncoderConfig *global, int *got_data) {
1675 const vpx_codec_cx_pkt_t *pkt;
1676 const struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1677 vpx_codec_iter_t iter = NULL;
1680 while ((pkt = vpx_codec_get_cx_data(&stream->encoder, &iter))) {
1681 static size_t fsize = 0;
1682 static FileOffset ivf_header_pos = 0;
1684 switch (pkt->kind) {
1685 case VPX_CODEC_CX_FRAME_PKT:
1686 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1687 stream->frames_out++;
1690 fprintf(stderr, " %6luF", (unsigned long)pkt->data.frame.sz);
1692 update_rate_histogram(stream->rate_hist, cfg, pkt);
1694 if (stream->config.write_webm) {
1695 write_webm_block(&stream->webm_ctx, cfg, pkt);
1698 if (!stream->config.write_webm) {
1699 if (pkt->data.frame.partition_id <= 0) {
1700 ivf_header_pos = ftello(stream->file);
1701 fsize = pkt->data.frame.sz;
1703 ivf_write_frame_header(stream->file, pkt->data.frame.pts, fsize);
1705 fsize += pkt->data.frame.sz;
1707 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1708 const FileOffset currpos = ftello(stream->file);
1709 fseeko(stream->file, ivf_header_pos, SEEK_SET);
1710 ivf_write_frame_size(stream->file, fsize);
1711 fseeko(stream->file, currpos, SEEK_SET);
1715 (void)fwrite(pkt->data.frame.buf, 1, pkt->data.frame.sz,
1718 stream->nbytes += pkt->data.raw.sz;
1722 if (global->test_decode != TEST_DECODE_OFF && !stream->mismatch_seen) {
1723 vpx_codec_decode(&stream->decoder, pkt->data.frame.buf,
1724 (unsigned int)pkt->data.frame.sz, NULL, 0);
1725 if (stream->decoder.err) {
1726 warn_or_exit_on_error(&stream->decoder,
1727 global->test_decode == TEST_DECODE_FATAL,
1728 "Failed to decode frame %d in stream %d",
1729 stream->frames_out + 1, stream->index);
1730 stream->mismatch_seen = stream->frames_out + 1;
1735 case VPX_CODEC_STATS_PKT:
1736 stream->frames_out++;
1737 stats_write(&stream->stats, pkt->data.twopass_stats.buf,
1738 pkt->data.twopass_stats.sz);
1739 stream->nbytes += pkt->data.raw.sz;
1741 #if CONFIG_FP_MB_STATS
1742 case VPX_CODEC_FPMB_STATS_PKT:
1743 stats_write(&stream->fpmb_stats, pkt->data.firstpass_mb_stats.buf,
1744 pkt->data.firstpass_mb_stats.sz);
1745 stream->nbytes += pkt->data.raw.sz;
1748 case VPX_CODEC_PSNR_PKT:
1750 if (global->show_psnr) {
1753 stream->psnr_sse_total += pkt->data.psnr.sse[0];
1754 stream->psnr_samples_total += pkt->data.psnr.samples[0];
1755 for (i = 0; i < 4; i++) {
1757 fprintf(stderr, "%.3f ", pkt->data.psnr.psnr[i]);
1758 stream->psnr_totals[i] += pkt->data.psnr.psnr[i];
1760 stream->psnr_count++;
1769 static void show_psnr(struct stream_state *stream, double peak) {
1773 if (!stream->psnr_count) return;
1775 fprintf(stderr, "Stream %d PSNR (Overall/Avg/Y/U/V)", stream->index);
1776 ovpsnr = sse_to_psnr((double)stream->psnr_samples_total, peak,
1777 (double)stream->psnr_sse_total);
1778 fprintf(stderr, " %.3f", ovpsnr);
1780 for (i = 0; i < 4; i++) {
1781 fprintf(stderr, " %.3f", stream->psnr_totals[i] / stream->psnr_count);
1783 fprintf(stderr, "\n");
1786 static float usec_to_fps(uint64_t usec, unsigned int frames) {
1787 return (float)(usec > 0 ? frames * 1000000.0 / (float)usec : 0);
1790 static void test_decode(struct stream_state *stream,
1791 enum TestDecodeFatality fatal,
1792 const VpxInterface *codec) {
1793 vpx_image_t enc_img, dec_img;
1795 if (stream->mismatch_seen) return;
1797 /* Get the internal reference frame */
1798 if (strcmp(codec->name, "vp8") == 0) {
1799 struct vpx_ref_frame ref_enc, ref_dec;
1802 width = (stream->config.cfg.g_w + 15) & ~15;
1803 height = (stream->config.cfg.g_h + 15) & ~15;
1804 vpx_img_alloc(&ref_enc.img, VPX_IMG_FMT_I420, width, height, 1);
1805 enc_img = ref_enc.img;
1806 vpx_img_alloc(&ref_dec.img, VPX_IMG_FMT_I420, width, height, 1);
1807 dec_img = ref_dec.img;
1809 ref_enc.frame_type = VP8_LAST_FRAME;
1810 ref_dec.frame_type = VP8_LAST_FRAME;
1811 vpx_codec_control(&stream->encoder, VP8_COPY_REFERENCE, &ref_enc);
1812 vpx_codec_control(&stream->decoder, VP8_COPY_REFERENCE, &ref_dec);
1814 struct vp9_ref_frame ref_enc, ref_dec;
1818 vpx_codec_control(&stream->encoder, VP9_GET_REFERENCE, &ref_enc);
1819 enc_img = ref_enc.img;
1820 vpx_codec_control(&stream->decoder, VP9_GET_REFERENCE, &ref_dec);
1821 dec_img = ref_dec.img;
1822 #if CONFIG_VP9_HIGHBITDEPTH
1823 if ((enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) !=
1824 (dec_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH)) {
1825 if (enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1826 vpx_img_alloc(&enc_img, enc_img.fmt - VPX_IMG_FMT_HIGHBITDEPTH,
1827 enc_img.d_w, enc_img.d_h, 16);
1828 vpx_img_truncate_16_to_8(&enc_img, &ref_enc.img);
1830 if (dec_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1831 vpx_img_alloc(&dec_img, dec_img.fmt - VPX_IMG_FMT_HIGHBITDEPTH,
1832 dec_img.d_w, dec_img.d_h, 16);
1833 vpx_img_truncate_16_to_8(&dec_img, &ref_dec.img);
1838 ctx_exit_on_error(&stream->encoder, "Failed to get encoder reference frame");
1839 ctx_exit_on_error(&stream->decoder, "Failed to get decoder reference frame");
1841 if (!compare_img(&enc_img, &dec_img)) {
1842 int y[4], u[4], v[4];
1843 #if CONFIG_VP9_HIGHBITDEPTH
1844 if (enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1845 find_mismatch_high(&enc_img, &dec_img, y, u, v);
1847 find_mismatch(&enc_img, &dec_img, y, u, v);
1850 find_mismatch(&enc_img, &dec_img, y, u, v);
1852 stream->decoder.err = 1;
1853 warn_or_exit_on_error(&stream->decoder, fatal == TEST_DECODE_FATAL,
1854 "Stream %d: Encode/decode mismatch on frame %d at"
1855 " Y[%d, %d] {%d/%d},"
1856 " U[%d, %d] {%d/%d},"
1857 " V[%d, %d] {%d/%d}",
1858 stream->index, stream->frames_out, y[0], y[1], y[2],
1859 y[3], u[0], u[1], u[2], u[3], v[0], v[1], v[2], v[3]);
1860 stream->mismatch_seen = stream->frames_out;
1863 vpx_img_free(&enc_img);
1864 vpx_img_free(&dec_img);
1867 static void print_time(const char *label, int64_t etl) {
1874 etl -= hours * 3600;
1879 fprintf(stderr, "[%3s %2" PRId64 ":%02" PRId64 ":%02" PRId64 "] ", label,
1882 fprintf(stderr, "[%3s unknown] ", label);
1886 int main(int argc, const char **argv_) {
1889 #if CONFIG_VP9_HIGHBITDEPTH
1890 vpx_image_t raw_shift;
1891 int allocated_raw_shift = 0;
1892 int use_16bit_internal = 0;
1893 int input_shift = 0;
1895 int frame_avail, got_data;
1897 struct VpxInputContext input;
1898 struct VpxEncoderConfig global;
1899 struct stream_state *streams = NULL;
1900 char **argv, **argi;
1901 uint64_t cx_time = 0;
1905 memset(&input, 0, sizeof(input));
1906 exec_name = argv_[0];
1908 if (argc < 3) usage_exit();
1910 /* Setup default input stream settings */
1911 input.framerate.numerator = 30;
1912 input.framerate.denominator = 1;
1913 input.only_i420 = 1;
1914 input.bit_depth = 0;
1916 /* First parse the global configuration values, because we want to apply
1917 * other parameters on top of the default configuration provided by the
1920 argv = argv_dup(argc - 1, argv_ + 1);
1921 parse_global_config(&global, argv);
1923 switch (global.color_type) {
1924 case I420: input.fmt = VPX_IMG_FMT_I420; break;
1925 case I422: input.fmt = VPX_IMG_FMT_I422; break;
1926 case I444: input.fmt = VPX_IMG_FMT_I444; break;
1927 case I440: input.fmt = VPX_IMG_FMT_I440; break;
1928 case YV12: input.fmt = VPX_IMG_FMT_YV12; break;
1932 /* Now parse each stream's parameters. Using a local scope here
1933 * due to the use of 'stream' as loop variable in FOREACH_STREAM
1936 struct stream_state *stream = NULL;
1939 stream = new_stream(&global, stream);
1941 if (!streams) streams = stream;
1942 } while (parse_stream_params(&global, stream, argv));
1945 /* Check for unrecognized options */
1946 for (argi = argv; *argi; argi++)
1947 if (argi[0][0] == '-' && argi[0][1])
1948 die("Error: Unrecognized option %s\n", *argi);
1950 FOREACH_STREAM(check_encoder_config(global.disable_warning_prompt, &global,
1951 &stream->config.cfg););
1953 /* Handle non-option arguments */
1954 input.filename = argv[0];
1956 if (!input.filename) usage_exit();
1958 /* Decide if other chroma subsamplings than 4:2:0 are supported */
1959 if (global.codec->fourcc == VP9_FOURCC) input.only_i420 = 0;
1961 for (pass = global.pass ? global.pass - 1 : 0; pass < global.passes; pass++) {
1962 int frames_in = 0, seen_frames = 0;
1963 int64_t estimated_time_left = -1;
1964 int64_t average_rate = -1;
1965 int64_t lagged_count = 0;
1967 open_input_file(&input);
1969 /* If the input file doesn't specify its w/h (raw files), try to get
1970 * the data from the first stream's configuration.
1972 if (!input.width || !input.height) {
1974 if (stream->config.cfg.g_w && stream->config.cfg.g_h) {
1975 input.width = stream->config.cfg.g_w;
1976 input.height = stream->config.cfg.g_h;
1982 /* Update stream configurations from the input file's parameters */
1983 if (!input.width || !input.height)
1985 "Specify stream dimensions with --width (-w) "
1986 " and --height (-h)");
1988 /* If input file does not specify bit-depth but input-bit-depth parameter
1989 * exists, assume that to be the input bit-depth. However, if the
1990 * input-bit-depth paramter does not exist, assume the input bit-depth
1991 * to be the same as the codec bit-depth.
1993 if (!input.bit_depth) {
1995 if (stream->config.cfg.g_input_bit_depth)
1996 input.bit_depth = stream->config.cfg.g_input_bit_depth;
1998 input.bit_depth = stream->config.cfg.g_input_bit_depth =
1999 (int)stream->config.cfg.g_bit_depth;
2001 if (input.bit_depth > 8) input.fmt |= VPX_IMG_FMT_HIGHBITDEPTH;
2004 { stream->config.cfg.g_input_bit_depth = input.bit_depth; });
2007 FOREACH_STREAM(set_stream_dimensions(stream, input.width, input.height));
2008 FOREACH_STREAM(validate_stream_config(stream, &global));
2010 /* Ensure that --passes and --pass are consistent. If --pass is set and
2011 * --passes=2, ensure --fpf was set.
2013 if (global.pass && global.passes == 2)
2015 if (!stream->config.stats_fn)
2016 die("Stream %d: Must specify --fpf when --pass=%d"
2017 " and --passes=2\n",
2018 stream->index, global.pass);
2023 if (stream->config.write_webm) {
2024 stream->config.write_webm = 0;
2026 "vpxenc was compiled without WebM container support."
2027 "Producing IVF output");
2032 /* Use the frame rate from the file only if none was specified
2033 * on the command-line.
2035 if (!global.have_framerate) {
2036 global.framerate.num = input.framerate.numerator;
2037 global.framerate.den = input.framerate.denominator;
2038 FOREACH_STREAM(stream->config.cfg.g_timebase.den = global.framerate.num;
2039 stream->config.cfg.g_timebase.num = global.framerate.den);
2042 /* Show configuration */
2043 if (global.verbose && pass == 0)
2044 FOREACH_STREAM(show_stream_config(stream, &global, &input));
2046 if (pass == (global.pass ? global.pass - 1 : 0)) {
2047 if (input.file_type == FILE_TYPE_Y4M)
2048 /*The Y4M reader does its own allocation.
2049 Just initialize this here to avoid problems if we never read any
2051 memset(&raw, 0, sizeof(raw));
2053 vpx_img_alloc(&raw, input.fmt, input.width, input.height, 32);
2055 FOREACH_STREAM(stream->rate_hist = init_rate_histogram(
2056 &stream->config.cfg, &global.framerate));
2059 FOREACH_STREAM(setup_pass(stream, &global, pass));
2061 open_output_file(stream, &global, &input.pixel_aspect_ratio));
2062 FOREACH_STREAM(initialize_encoder(stream, &global));
2064 #if CONFIG_VP9_HIGHBITDEPTH
2065 if (strcmp(global.codec->name, "vp9") == 0) {
2066 // Check to see if at least one stream uses 16 bit internal.
2067 // Currently assume that the bit_depths for all streams using
2068 // highbitdepth are the same.
2070 if (stream->config.use_16bit_internal) {
2071 use_16bit_internal = 1;
2073 if (stream->config.cfg.g_profile == 0) {
2076 input_shift = (int)stream->config.cfg.g_bit_depth -
2077 stream->config.cfg.g_input_bit_depth;
2086 while (frame_avail || got_data) {
2087 struct vpx_usec_timer timer;
2089 if (!global.limit || frames_in < global.limit) {
2090 frame_avail = read_frame(&input, &raw);
2092 if (frame_avail) frames_in++;
2094 frames_in > global.skip_frames ? frames_in - global.skip_frames : 0;
2096 if (!global.quiet) {
2097 float fps = usec_to_fps(cx_time, seen_frames);
2098 fprintf(stderr, "\rPass %d/%d ", pass + 1, global.passes);
2100 if (stream_cnt == 1)
2101 fprintf(stderr, "frame %4d/%-4d %7" PRId64 "B ", frames_in,
2102 streams->frames_out, (int64_t)streams->nbytes);
2104 fprintf(stderr, "frame %4d ", frames_in);
2106 fprintf(stderr, "%7" PRId64 " %s %.2f %s ",
2107 cx_time > 9999999 ? cx_time / 1000 : cx_time,
2108 cx_time > 9999999 ? "ms" : "us", fps >= 1.0 ? fps : fps * 60,
2109 fps >= 1.0 ? "fps" : "fpm");
2110 print_time("ETA", estimated_time_left);
2116 if (frames_in > global.skip_frames) {
2117 #if CONFIG_VP9_HIGHBITDEPTH
2118 vpx_image_t *frame_to_encode;
2119 if (input_shift || (use_16bit_internal && input.bit_depth == 8)) {
2120 assert(use_16bit_internal);
2121 // Input bit depth and stream bit depth do not match, so up
2122 // shift frame to stream bit depth
2123 if (!allocated_raw_shift) {
2124 vpx_img_alloc(&raw_shift, raw.fmt | VPX_IMG_FMT_HIGHBITDEPTH,
2125 input.width, input.height, 32);
2126 allocated_raw_shift = 1;
2128 vpx_img_upshift(&raw_shift, &raw, input_shift);
2129 frame_to_encode = &raw_shift;
2131 frame_to_encode = &raw;
2133 vpx_usec_timer_start(&timer);
2134 if (use_16bit_internal) {
2135 assert(frame_to_encode->fmt & VPX_IMG_FMT_HIGHBITDEPTH);
2137 if (stream->config.use_16bit_internal)
2138 encode_frame(stream, &global,
2139 frame_avail ? frame_to_encode : NULL, frames_in);
2144 assert((frame_to_encode->fmt & VPX_IMG_FMT_HIGHBITDEPTH) == 0);
2145 FOREACH_STREAM(encode_frame(stream, &global,
2146 frame_avail ? frame_to_encode : NULL,
2150 vpx_usec_timer_start(&timer);
2151 FOREACH_STREAM(encode_frame(stream, &global, frame_avail ? &raw : NULL,
2154 vpx_usec_timer_mark(&timer);
2155 cx_time += vpx_usec_timer_elapsed(&timer);
2157 FOREACH_STREAM(update_quantizer_histogram(stream));
2160 FOREACH_STREAM(get_cx_data(stream, &global, &got_data));
2162 if (!got_data && input.length && streams != NULL &&
2163 !streams->frames_out) {
2164 lagged_count = global.limit ? seen_frames : ftello(input.file);
2165 } else if (input.length) {
2170 const int64_t frame_in_lagged = (seen_frames - lagged_count) * 1000;
2172 rate = cx_time ? frame_in_lagged * (int64_t)1000000 / cx_time : 0;
2173 remaining = 1000 * (global.limit - global.skip_frames -
2174 seen_frames + lagged_count);
2176 const int64_t input_pos = ftello(input.file);
2177 const int64_t input_pos_lagged = input_pos - lagged_count;
2178 const int64_t limit = input.length;
2180 rate = cx_time ? input_pos_lagged * (int64_t)1000000 / cx_time : 0;
2181 remaining = limit - input_pos + lagged_count;
2185 (average_rate <= 0) ? rate : (average_rate * 7 + rate) / 8;
2186 estimated_time_left = average_rate ? remaining / average_rate : -1;
2189 if (got_data && global.test_decode != TEST_DECODE_OFF)
2190 FOREACH_STREAM(test_decode(stream, global.test_decode, global.codec));
2194 if (!global.quiet) fprintf(stderr, "\033[K");
2197 if (stream_cnt > 1) fprintf(stderr, "\n");
2199 if (!global.quiet) {
2200 FOREACH_STREAM(fprintf(
2201 stderr, "\rPass %d/%d frame %4d/%-4d %7" PRId64 "B %7" PRId64
2202 "b/f %7" PRId64 "b/s"
2203 " %7" PRId64 " %s (%.2f fps)\033[K\n",
2204 pass + 1, global.passes, frames_in, stream->frames_out,
2205 (int64_t)stream->nbytes,
2206 seen_frames ? (int64_t)(stream->nbytes * 8 / seen_frames) : 0,
2208 ? (int64_t)stream->nbytes * 8 * (int64_t)global.framerate.num /
2209 global.framerate.den / seen_frames
2211 stream->cx_time > 9999999 ? stream->cx_time / 1000 : stream->cx_time,
2212 stream->cx_time > 9999999 ? "ms" : "us",
2213 usec_to_fps(stream->cx_time, seen_frames)));
2216 if (global.show_psnr) {
2217 if (global.codec->fourcc == VP9_FOURCC) {
2219 show_psnr(stream, (1 << stream->config.cfg.g_input_bit_depth) - 1));
2221 FOREACH_STREAM(show_psnr(stream, 255.0));
2225 FOREACH_STREAM(vpx_codec_destroy(&stream->encoder));
2227 if (global.test_decode != TEST_DECODE_OFF) {
2228 FOREACH_STREAM(vpx_codec_destroy(&stream->decoder));
2231 close_input_file(&input);
2233 if (global.test_decode == TEST_DECODE_FATAL) {
2234 FOREACH_STREAM(res |= stream->mismatch_seen);
2236 FOREACH_STREAM(close_output_file(stream, global.codec->fourcc));
2238 FOREACH_STREAM(stats_close(&stream->stats, global.passes - 1));
2240 #if CONFIG_FP_MB_STATS
2241 FOREACH_STREAM(stats_close(&stream->fpmb_stats, global.passes - 1));
2244 if (global.pass) break;
2247 if (global.show_q_hist_buckets)
2249 show_q_histogram(stream->counts, global.show_q_hist_buckets));
2251 if (global.show_rate_hist_buckets)
2252 FOREACH_STREAM(show_rate_histogram(stream->rate_hist, &stream->config.cfg,
2253 global.show_rate_hist_buckets));
2254 FOREACH_STREAM(destroy_rate_histogram(stream->rate_hist));
2256 #if CONFIG_INTERNAL_STATS
2257 /* TODO(jkoleszar): This doesn't belong in this executable. Do it for now,
2258 * to match some existing utilities.
2260 if (!(global.pass == 1 && global.passes == 2))
2262 FILE *f = fopen("opsnr.stt", "a");
2263 if (stream->mismatch_seen) {
2264 fprintf(f, "First mismatch occurred in frame %d\n",
2265 stream->mismatch_seen);
2267 fprintf(f, "No mismatch detected in recon buffers\n");
2273 #if CONFIG_VP9_HIGHBITDEPTH
2274 if (allocated_raw_shift) vpx_img_free(&raw_shift);
2279 return res ? EXIT_FAILURE : EXIT_SUCCESS;