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 *rc_twopass_args[] = { &bias_pct, &minsection_pct,
325 &maxsection_pct, NULL };
327 static const arg_def_t kf_min_dist =
328 ARG_DEF(NULL, "kf-min-dist", 1, "Minimum keyframe interval (frames)");
329 static const arg_def_t kf_max_dist =
330 ARG_DEF(NULL, "kf-max-dist", 1, "Maximum keyframe interval (frames)");
331 static const arg_def_t kf_disabled =
332 ARG_DEF(NULL, "disable-kf", 0, "Disable keyframe placement");
333 static const arg_def_t *kf_args[] = { &kf_min_dist, &kf_max_dist, &kf_disabled,
336 static const arg_def_t noise_sens =
337 ARG_DEF(NULL, "noise-sensitivity", 1, "Noise sensitivity (frames to blur)");
338 static const arg_def_t sharpness =
339 ARG_DEF(NULL, "sharpness", 1, "Loop filter sharpness (0..7)");
340 static const arg_def_t static_thresh =
341 ARG_DEF(NULL, "static-thresh", 1, "Motion detection threshold");
342 static const arg_def_t auto_altref =
343 ARG_DEF(NULL, "auto-alt-ref", 1, "Enable automatic alt reference frames");
344 static const arg_def_t arnr_maxframes =
345 ARG_DEF(NULL, "arnr-maxframes", 1, "AltRef max frames (0..15)");
346 static const arg_def_t arnr_strength =
347 ARG_DEF(NULL, "arnr-strength", 1, "AltRef filter strength (0..6)");
348 static const arg_def_t arnr_type = ARG_DEF(NULL, "arnr-type", 1, "AltRef type");
349 static const struct arg_enum_list tuning_enum[] = {
350 { "psnr", VP8_TUNE_PSNR }, { "ssim", VP8_TUNE_SSIM }, { NULL, 0 }
352 static const arg_def_t tune_ssim =
353 ARG_DEF_ENUM(NULL, "tune", 1, "Material to favor", tuning_enum);
354 static const arg_def_t cq_level =
355 ARG_DEF(NULL, "cq-level", 1, "Constant/Constrained Quality level");
356 static const arg_def_t max_intra_rate_pct =
357 ARG_DEF(NULL, "max-intra-rate", 1, "Max I-frame bitrate (pct)");
359 #if CONFIG_VP8_ENCODER
360 static const arg_def_t cpu_used_vp8 =
361 ARG_DEF(NULL, "cpu-used", 1, "CPU Used (-16..16)");
362 static const arg_def_t token_parts =
363 ARG_DEF(NULL, "token-parts", 1, "Number of token partitions to use, log2");
364 static const arg_def_t screen_content_mode =
365 ARG_DEF(NULL, "screen-content-mode", 1, "Screen content mode");
366 static const arg_def_t *vp8_args[] = {
367 &cpu_used_vp8, &auto_altref, &noise_sens, &sharpness,
368 &static_thresh, &token_parts, &arnr_maxframes, &arnr_strength,
369 &arnr_type, &tune_ssim, &cq_level, &max_intra_rate_pct,
370 &screen_content_mode, NULL
372 static const int vp8_arg_ctrl_map[] = { VP8E_SET_CPUUSED,
373 VP8E_SET_ENABLEAUTOALTREF,
374 VP8E_SET_NOISE_SENSITIVITY,
376 VP8E_SET_STATIC_THRESHOLD,
377 VP8E_SET_TOKEN_PARTITIONS,
378 VP8E_SET_ARNR_MAXFRAMES,
379 VP8E_SET_ARNR_STRENGTH,
383 VP8E_SET_MAX_INTRA_BITRATE_PCT,
384 VP8E_SET_SCREEN_CONTENT_MODE,
388 #if CONFIG_VP9_ENCODER
389 static const arg_def_t cpu_used_vp9 =
390 ARG_DEF(NULL, "cpu-used", 1, "CPU Used (-8..8)");
391 static const arg_def_t tile_cols =
392 ARG_DEF(NULL, "tile-columns", 1, "Number of tile columns to use, log2");
393 static const arg_def_t tile_rows =
394 ARG_DEF(NULL, "tile-rows", 1,
395 "Number of tile rows to use, log2 (set to 0 while threads > 1)");
396 static const arg_def_t lossless =
397 ARG_DEF(NULL, "lossless", 1, "Lossless mode (0: false (default), 1: true)");
398 static const arg_def_t frame_parallel_decoding = ARG_DEF(
399 NULL, "frame-parallel", 1, "Enable frame parallel decodability features");
400 static const arg_def_t aq_mode = ARG_DEF(
402 "Adaptive quantization mode (0: off (default), 1: variance 2: complexity, "
403 "3: cyclic refresh, 4: equator360)");
404 static const arg_def_t alt_ref_aq = ARG_DEF(NULL, "alt-ref-aq", 1,
405 "Special adaptive quantization for "
406 "the alternate reference frames.");
407 static const arg_def_t frame_periodic_boost =
408 ARG_DEF(NULL, "frame-boost", 1,
409 "Enable frame periodic boost (0: off (default), 1: on)");
410 static const arg_def_t gf_cbr_boost_pct = ARG_DEF(
411 NULL, "gf-cbr-boost", 1, "Boost for Golden Frame in CBR mode (pct)");
412 static const arg_def_t max_inter_rate_pct =
413 ARG_DEF(NULL, "max-inter-rate", 1, "Max P-frame bitrate (pct)");
414 static const arg_def_t min_gf_interval = ARG_DEF(
415 NULL, "min-gf-interval", 1,
416 "min gf/arf frame interval (default 0, indicating in-built behavior)");
417 static const arg_def_t max_gf_interval = ARG_DEF(
418 NULL, "max-gf-interval", 1,
419 "max gf/arf frame interval (default 0, indicating in-built behavior)");
421 static const struct arg_enum_list color_space_enum[] = {
422 { "unknown", VPX_CS_UNKNOWN },
423 { "bt601", VPX_CS_BT_601 },
424 { "bt709", VPX_CS_BT_709 },
425 { "smpte170", VPX_CS_SMPTE_170 },
426 { "smpte240", VPX_CS_SMPTE_240 },
427 { "bt2020", VPX_CS_BT_2020 },
428 { "reserved", VPX_CS_RESERVED },
429 { "sRGB", VPX_CS_SRGB },
433 static const arg_def_t input_color_space =
434 ARG_DEF_ENUM(NULL, "color-space", 1, "The color space of input content:",
437 #if CONFIG_VP9_HIGHBITDEPTH
438 static const struct arg_enum_list bitdepth_enum[] = {
439 { "8", VPX_BITS_8 }, { "10", VPX_BITS_10 }, { "12", VPX_BITS_12 }, { NULL, 0 }
442 static const arg_def_t bitdeptharg = ARG_DEF_ENUM(
444 "Bit depth for codec (8 for version <=1, 10 or 12 for version 2)",
446 static const arg_def_t inbitdeptharg =
447 ARG_DEF(NULL, "input-bit-depth", 1, "Bit depth of input");
450 static const struct arg_enum_list tune_content_enum[] = {
451 { "default", VP9E_CONTENT_DEFAULT },
452 { "screen", VP9E_CONTENT_SCREEN },
456 static const arg_def_t tune_content = ARG_DEF_ENUM(
457 NULL, "tune-content", 1, "Tune content type", tune_content_enum);
459 static const arg_def_t target_level = ARG_DEF(
460 NULL, "target-level", 1,
461 "Target level (255: off (default); 0: only keep level stats; 10: level 1.0;"
462 " 11: level 1.1; ... 62: level 6.2)");
465 #if CONFIG_VP9_ENCODER
466 static const arg_def_t *vp9_args[] = { &cpu_used_vp9,
481 &frame_parallel_decoding,
484 &frame_periodic_boost,
491 #if CONFIG_VP9_HIGHBITDEPTH
494 #endif // CONFIG_VP9_HIGHBITDEPTH
496 static const int vp9_arg_ctrl_map[] = { VP8E_SET_CPUUSED,
497 VP8E_SET_ENABLEAUTOALTREF,
499 VP8E_SET_STATIC_THRESHOLD,
500 VP9E_SET_TILE_COLUMNS,
502 VP8E_SET_ARNR_MAXFRAMES,
503 VP8E_SET_ARNR_STRENGTH,
507 VP8E_SET_MAX_INTRA_BITRATE_PCT,
508 VP9E_SET_MAX_INTER_BITRATE_PCT,
509 VP9E_SET_GF_CBR_BOOST_PCT,
511 VP9E_SET_FRAME_PARALLEL_DECODING,
514 VP9E_SET_FRAME_PERIODIC_BOOST,
515 VP9E_SET_NOISE_SENSITIVITY,
516 VP9E_SET_TUNE_CONTENT,
517 VP9E_SET_COLOR_SPACE,
518 VP9E_SET_MIN_GF_INTERVAL,
519 VP9E_SET_MAX_GF_INTERVAL,
520 VP9E_SET_TARGET_LEVEL,
524 static const arg_def_t *no_args[] = { NULL };
526 void usage_exit(void) {
528 const int num_encoder = get_vpx_encoder_count();
530 fprintf(stderr, "Usage: %s <options> -o dst_filename src_filename \n",
533 fprintf(stderr, "\nOptions:\n");
534 arg_show_usage(stderr, main_args);
535 fprintf(stderr, "\nEncoder Global Options:\n");
536 arg_show_usage(stderr, global_args);
537 fprintf(stderr, "\nRate Control Options:\n");
538 arg_show_usage(stderr, rc_args);
539 fprintf(stderr, "\nTwopass Rate Control Options:\n");
540 arg_show_usage(stderr, rc_twopass_args);
541 fprintf(stderr, "\nKeyframe Placement Options:\n");
542 arg_show_usage(stderr, kf_args);
543 #if CONFIG_VP8_ENCODER
544 fprintf(stderr, "\nVP8 Specific Options:\n");
545 arg_show_usage(stderr, vp8_args);
547 #if CONFIG_VP9_ENCODER
548 fprintf(stderr, "\nVP9 Specific Options:\n");
549 arg_show_usage(stderr, vp9_args);
552 "\nStream timebase (--timebase):\n"
553 " The desired precision of timestamps in the output, expressed\n"
554 " in fractional seconds. Default is 1/1000.\n");
555 fprintf(stderr, "\nIncluded encoders:\n\n");
557 for (i = 0; i < num_encoder; ++i) {
558 const VpxInterface *const encoder = get_vpx_encoder_by_index(i);
559 const char *defstr = (i == (num_encoder - 1)) ? "(default)" : "";
560 fprintf(stderr, " %-6s - %s %s\n", encoder->name,
561 vpx_codec_iface_name(encoder->codec_interface()), defstr);
563 fprintf(stderr, "\n ");
564 fprintf(stderr, "Use --codec to switch to a non-default encoder.\n\n");
569 #define mmin(a, b) ((a) < (b) ? (a) : (b))
571 #if CONFIG_VP9_HIGHBITDEPTH
572 static void find_mismatch_high(const vpx_image_t *const img1,
573 const vpx_image_t *const img2, int yloc[4],
574 int uloc[4], int vloc[4]) {
575 uint16_t *plane1, *plane2;
576 uint32_t stride1, stride2;
577 const uint32_t bsize = 64;
578 const uint32_t bsizey = bsize >> img1->y_chroma_shift;
579 const uint32_t bsizex = bsize >> img1->x_chroma_shift;
581 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
583 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
586 yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
587 plane1 = (uint16_t *)img1->planes[VPX_PLANE_Y];
588 plane2 = (uint16_t *)img2->planes[VPX_PLANE_Y];
589 stride1 = img1->stride[VPX_PLANE_Y] / 2;
590 stride2 = img2->stride[VPX_PLANE_Y] / 2;
591 for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
592 for (j = 0; match && j < img1->d_w; j += bsize) {
594 const int si = mmin(i + bsize, img1->d_h) - i;
595 const int sj = mmin(j + bsize, img1->d_w) - j;
596 for (k = 0; match && k < si; ++k) {
597 for (l = 0; match && l < sj; ++l) {
598 if (*(plane1 + (i + k) * stride1 + j + l) !=
599 *(plane2 + (i + k) * stride2 + j + l)) {
602 yloc[2] = *(plane1 + (i + k) * stride1 + j + l);
603 yloc[3] = *(plane2 + (i + k) * stride2 + j + l);
612 uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
613 plane1 = (uint16_t *)img1->planes[VPX_PLANE_U];
614 plane2 = (uint16_t *)img2->planes[VPX_PLANE_U];
615 stride1 = img1->stride[VPX_PLANE_U] / 2;
616 stride2 = img2->stride[VPX_PLANE_U] / 2;
617 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
618 for (j = 0; match && j < c_w; j += bsizex) {
620 const int si = mmin(i + bsizey, c_h - i);
621 const int sj = mmin(j + bsizex, c_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 uloc[2] = *(plane1 + (i + k) * stride1 + j + l);
629 uloc[3] = *(plane2 + (i + k) * stride2 + j + l);
638 vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
639 plane1 = (uint16_t *)img1->planes[VPX_PLANE_V];
640 plane2 = (uint16_t *)img2->planes[VPX_PLANE_V];
641 stride1 = img1->stride[VPX_PLANE_V] / 2;
642 stride2 = img2->stride[VPX_PLANE_V] / 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 vloc[2] = *(plane1 + (i + k) * stride1 + j + l);
655 vloc[3] = *(plane2 + (i + k) * stride2 + j + l);
666 static void find_mismatch(const vpx_image_t *const img1,
667 const vpx_image_t *const img2, int yloc[4],
668 int uloc[4], int vloc[4]) {
669 const uint32_t bsize = 64;
670 const uint32_t bsizey = bsize >> img1->y_chroma_shift;
671 const uint32_t bsizex = bsize >> img1->x_chroma_shift;
673 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
675 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
678 yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
679 for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
680 for (j = 0; match && j < img1->d_w; j += bsize) {
682 const int si = mmin(i + bsize, img1->d_h) - i;
683 const int sj = mmin(j + bsize, img1->d_w) - j;
684 for (k = 0; match && k < si; ++k) {
685 for (l = 0; match && l < sj; ++l) {
686 if (*(img1->planes[VPX_PLANE_Y] +
687 (i + k) * img1->stride[VPX_PLANE_Y] + j + l) !=
688 *(img2->planes[VPX_PLANE_Y] +
689 (i + k) * img2->stride[VPX_PLANE_Y] + j + l)) {
692 yloc[2] = *(img1->planes[VPX_PLANE_Y] +
693 (i + k) * img1->stride[VPX_PLANE_Y] + j + l);
694 yloc[3] = *(img2->planes[VPX_PLANE_Y] +
695 (i + k) * img2->stride[VPX_PLANE_Y] + j + l);
704 uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
705 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
706 for (j = 0; match && j < c_w; j += bsizex) {
708 const int si = mmin(i + bsizey, c_h - i);
709 const int sj = mmin(j + bsizex, c_w - j);
710 for (k = 0; match && k < si; ++k) {
711 for (l = 0; match && l < sj; ++l) {
712 if (*(img1->planes[VPX_PLANE_U] +
713 (i + k) * img1->stride[VPX_PLANE_U] + j + l) !=
714 *(img2->planes[VPX_PLANE_U] +
715 (i + k) * img2->stride[VPX_PLANE_U] + j + l)) {
718 uloc[2] = *(img1->planes[VPX_PLANE_U] +
719 (i + k) * img1->stride[VPX_PLANE_U] + j + l);
720 uloc[3] = *(img2->planes[VPX_PLANE_U] +
721 (i + k) * img2->stride[VPX_PLANE_U] + j + l);
729 vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
730 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
731 for (j = 0; match && j < c_w; j += bsizex) {
733 const int si = mmin(i + bsizey, c_h - i);
734 const int sj = mmin(j + bsizex, c_w - j);
735 for (k = 0; match && k < si; ++k) {
736 for (l = 0; match && l < sj; ++l) {
737 if (*(img1->planes[VPX_PLANE_V] +
738 (i + k) * img1->stride[VPX_PLANE_V] + j + l) !=
739 *(img2->planes[VPX_PLANE_V] +
740 (i + k) * img2->stride[VPX_PLANE_V] + j + l)) {
743 vloc[2] = *(img1->planes[VPX_PLANE_V] +
744 (i + k) * img1->stride[VPX_PLANE_V] + j + l);
745 vloc[3] = *(img2->planes[VPX_PLANE_V] +
746 (i + k) * img2->stride[VPX_PLANE_V] + j + l);
756 static int compare_img(const vpx_image_t *const img1,
757 const vpx_image_t *const img2) {
758 uint32_t l_w = img1->d_w;
759 uint32_t c_w = (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
761 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
765 match &= (img1->fmt == img2->fmt);
766 match &= (img1->d_w == img2->d_w);
767 match &= (img1->d_h == img2->d_h);
768 #if CONFIG_VP9_HIGHBITDEPTH
769 if (img1->fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
775 for (i = 0; i < img1->d_h; ++i)
776 match &= (memcmp(img1->planes[VPX_PLANE_Y] + i * img1->stride[VPX_PLANE_Y],
777 img2->planes[VPX_PLANE_Y] + i * img2->stride[VPX_PLANE_Y],
780 for (i = 0; i < c_h; ++i)
781 match &= (memcmp(img1->planes[VPX_PLANE_U] + i * img1->stride[VPX_PLANE_U],
782 img2->planes[VPX_PLANE_U] + i * img2->stride[VPX_PLANE_U],
785 for (i = 0; i < c_h; ++i)
786 match &= (memcmp(img1->planes[VPX_PLANE_V] + i * img1->stride[VPX_PLANE_V],
787 img2->planes[VPX_PLANE_V] + i * img2->stride[VPX_PLANE_V],
793 #define NELEMENTS(x) (sizeof(x) / sizeof(x[0]))
794 #if CONFIG_VP9_ENCODER
795 #define ARG_CTRL_CNT_MAX NELEMENTS(vp9_arg_ctrl_map)
797 #define ARG_CTRL_CNT_MAX NELEMENTS(vp8_arg_ctrl_map)
801 typedef int stereo_format_t;
802 struct WebmOutputContext {
807 /* Per-stream configuration */
808 struct stream_config {
809 struct vpx_codec_enc_cfg cfg;
811 const char *stats_fn;
812 #if CONFIG_FP_MB_STATS
813 const char *fpmb_stats_fn;
815 stereo_format_t stereo_fmt;
816 int arg_ctrls[ARG_CTRL_CNT_MAX][2];
819 #if CONFIG_VP9_HIGHBITDEPTH
820 // whether to use 16bit internal buffers
821 int use_16bit_internal;
825 struct stream_state {
827 struct stream_state *next;
828 struct stream_config config;
830 struct rate_hist *rate_hist;
831 struct WebmOutputContext webm_ctx;
832 uint64_t psnr_sse_total;
833 uint64_t psnr_samples_total;
834 double psnr_totals[4];
837 vpx_codec_ctx_t encoder;
838 unsigned int frames_out;
842 #if CONFIG_FP_MB_STATS
843 stats_io_t fpmb_stats;
845 struct vpx_image *img;
846 vpx_codec_ctx_t decoder;
850 static void validate_positive_rational(const char *msg,
851 struct vpx_rational *rat) {
857 if (rat->num < 0) die("Error: %s must be positive\n", msg);
859 if (!rat->den) die("Error: %s has zero denominator\n", msg);
862 static void parse_global_config(struct VpxEncoderConfig *global, char **argv) {
865 const int num_encoder = get_vpx_encoder_count();
867 if (num_encoder < 1) die("Error: no valid encoder available\n");
869 /* Initialize default parameters */
870 memset(global, 0, sizeof(*global));
871 global->codec = get_vpx_encoder_by_index(num_encoder - 1);
873 global->color_type = I420;
874 /* Assign default deadline to good quality */
875 global->deadline = VPX_DL_GOOD_QUALITY;
877 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
880 if (arg_match(&arg, &codecarg, argi)) {
881 global->codec = get_vpx_encoder_by_name(arg.val);
883 die("Error: Unrecognized argument (%s) to --codec\n", arg.val);
884 } else if (arg_match(&arg, &passes, argi)) {
885 global->passes = arg_parse_uint(&arg);
887 if (global->passes < 1 || global->passes > 2)
888 die("Error: Invalid number of passes (%d)\n", global->passes);
889 } else if (arg_match(&arg, &pass_arg, argi)) {
890 global->pass = arg_parse_uint(&arg);
892 if (global->pass < 1 || global->pass > 2)
893 die("Error: Invalid pass selected (%d)\n", global->pass);
894 } else if (arg_match(&arg, &usage, argi))
895 global->usage = arg_parse_uint(&arg);
896 else if (arg_match(&arg, &deadline, argi))
897 global->deadline = arg_parse_uint(&arg);
898 else if (arg_match(&arg, &best_dl, argi))
899 global->deadline = VPX_DL_BEST_QUALITY;
900 else if (arg_match(&arg, &good_dl, argi))
901 global->deadline = VPX_DL_GOOD_QUALITY;
902 else if (arg_match(&arg, &rt_dl, argi))
903 global->deadline = VPX_DL_REALTIME;
904 else if (arg_match(&arg, &use_yv12, argi))
905 global->color_type = YV12;
906 else if (arg_match(&arg, &use_i420, argi))
907 global->color_type = I420;
908 else if (arg_match(&arg, &use_i422, argi))
909 global->color_type = I422;
910 else if (arg_match(&arg, &use_i444, argi))
911 global->color_type = I444;
912 else if (arg_match(&arg, &use_i440, argi))
913 global->color_type = I440;
914 else if (arg_match(&arg, &quietarg, argi))
916 else if (arg_match(&arg, &verbosearg, argi))
918 else if (arg_match(&arg, &limit, argi))
919 global->limit = arg_parse_uint(&arg);
920 else if (arg_match(&arg, &skip, argi))
921 global->skip_frames = arg_parse_uint(&arg);
922 else if (arg_match(&arg, &psnrarg, argi))
923 global->show_psnr = 1;
924 else if (arg_match(&arg, &recontest, argi))
925 global->test_decode = arg_parse_enum_or_int(&arg);
926 else if (arg_match(&arg, &framerate, argi)) {
927 global->framerate = arg_parse_rational(&arg);
928 validate_positive_rational(arg.name, &global->framerate);
929 global->have_framerate = 1;
930 } else if (arg_match(&arg, &out_part, argi))
931 global->out_part = 1;
932 else if (arg_match(&arg, &debugmode, argi))
934 else if (arg_match(&arg, &q_hist_n, argi))
935 global->show_q_hist_buckets = arg_parse_uint(&arg);
936 else if (arg_match(&arg, &rate_hist_n, argi))
937 global->show_rate_hist_buckets = arg_parse_uint(&arg);
938 else if (arg_match(&arg, &disable_warnings, argi))
939 global->disable_warnings = 1;
940 else if (arg_match(&arg, &disable_warning_prompt, argi))
941 global->disable_warning_prompt = 1;
947 /* DWIM: Assume the user meant passes=2 if pass=2 is specified */
948 if (global->pass > global->passes) {
949 warn("Assuming --pass=%d implies --passes=%d\n", global->pass,
951 global->passes = global->pass;
954 /* Validate global config */
955 if (global->passes == 0) {
956 #if CONFIG_VP9_ENCODER
957 // Make default VP9 passes = 2 until there is a better quality 1-pass
959 if (global->codec != NULL && global->codec->name != NULL)
960 global->passes = (strcmp(global->codec->name, "vp9") == 0 &&
961 global->deadline != VPX_DL_REALTIME)
969 if (global->deadline == VPX_DL_REALTIME && global->passes > 1) {
970 warn("Enforcing one-pass encoding in realtime mode\n");
975 static void open_input_file(struct VpxInputContext *input) {
976 /* Parse certain options from the input file, if possible */
977 input->file = strcmp(input->filename, "-") ? fopen(input->filename, "rb")
978 : set_binary_mode(stdin);
980 if (!input->file) fatal("Failed to open input file");
982 if (!fseeko(input->file, 0, SEEK_END)) {
983 /* Input file is seekable. Figure out how long it is, so we can get
986 input->length = ftello(input->file);
990 /* Default to 1:1 pixel aspect ratio. */
991 input->pixel_aspect_ratio.numerator = 1;
992 input->pixel_aspect_ratio.denominator = 1;
994 /* For RAW input sources, these bytes will applied on the first frame
997 input->detect.buf_read = fread(input->detect.buf, 1, 4, input->file);
998 input->detect.position = 0;
1000 if (input->detect.buf_read == 4 && file_is_y4m(input->detect.buf)) {
1001 if (y4m_input_open(&input->y4m, input->file, input->detect.buf, 4,
1002 input->only_i420) >= 0) {
1003 input->file_type = FILE_TYPE_Y4M;
1004 input->width = input->y4m.pic_w;
1005 input->height = input->y4m.pic_h;
1006 input->pixel_aspect_ratio.numerator = input->y4m.par_n;
1007 input->pixel_aspect_ratio.denominator = input->y4m.par_d;
1008 input->framerate.numerator = input->y4m.fps_n;
1009 input->framerate.denominator = input->y4m.fps_d;
1010 input->fmt = input->y4m.vpx_fmt;
1011 input->bit_depth = input->y4m.bit_depth;
1013 fatal("Unsupported Y4M stream.");
1014 } else if (input->detect.buf_read == 4 && fourcc_is_ivf(input->detect.buf)) {
1015 fatal("IVF is not supported as input.");
1017 input->file_type = FILE_TYPE_RAW;
1021 static void close_input_file(struct VpxInputContext *input) {
1022 fclose(input->file);
1023 if (input->file_type == FILE_TYPE_Y4M) y4m_input_close(&input->y4m);
1026 static struct stream_state *new_stream(struct VpxEncoderConfig *global,
1027 struct stream_state *prev) {
1028 struct stream_state *stream;
1030 stream = calloc(1, sizeof(*stream));
1031 if (stream == NULL) {
1032 fatal("Failed to allocate new stream.");
1036 memcpy(stream, prev, sizeof(*stream));
1038 prev->next = stream;
1040 vpx_codec_err_t res;
1042 /* Populate encoder configuration */
1043 res = vpx_codec_enc_config_default(global->codec->codec_interface(),
1044 &stream->config.cfg, global->usage);
1045 if (res) fatal("Failed to get config: %s\n", vpx_codec_err_to_string(res));
1047 /* Change the default timebase to a high enough value so that the
1048 * encoder will always create strictly increasing timestamps.
1050 stream->config.cfg.g_timebase.den = 1000;
1052 /* Never use the library's default resolution, require it be parsed
1053 * from the file or set on the command line.
1055 stream->config.cfg.g_w = 0;
1056 stream->config.cfg.g_h = 0;
1058 /* Initialize remaining stream parameters */
1059 stream->config.write_webm = 1;
1061 stream->config.stereo_fmt = STEREO_FORMAT_MONO;
1062 stream->webm_ctx.last_pts_ns = -1;
1063 stream->webm_ctx.writer = NULL;
1064 stream->webm_ctx.segment = NULL;
1067 /* Allows removal of the application version from the EBML tags */
1068 stream->webm_ctx.debug = global->debug;
1070 /* Default lag_in_frames is 0 in realtime mode CBR mode*/
1071 if (global->deadline == VPX_DL_REALTIME &&
1072 stream->config.cfg.rc_end_usage == 1)
1073 stream->config.cfg.g_lag_in_frames = 0;
1076 /* Output files must be specified for each stream */
1077 stream->config.out_fn = NULL;
1079 stream->next = NULL;
1083 static int parse_stream_params(struct VpxEncoderConfig *global,
1084 struct stream_state *stream, char **argv) {
1085 char **argi, **argj;
1087 static const arg_def_t **ctrl_args = no_args;
1088 static const int *ctrl_args_map = NULL;
1089 struct stream_config *config = &stream->config;
1090 int eos_mark_found = 0;
1091 #if CONFIG_VP9_HIGHBITDEPTH
1092 int test_16bit_internal = 0;
1095 // Handle codec specific options
1097 #if CONFIG_VP8_ENCODER
1098 } else if (strcmp(global->codec->name, "vp8") == 0) {
1099 ctrl_args = vp8_args;
1100 ctrl_args_map = vp8_arg_ctrl_map;
1102 #if CONFIG_VP9_ENCODER
1103 } else if (strcmp(global->codec->name, "vp9") == 0) {
1104 ctrl_args = vp9_args;
1105 ctrl_args_map = vp9_arg_ctrl_map;
1109 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
1112 /* Once we've found an end-of-stream marker (--) we want to continue
1113 * shifting arguments but not consuming them.
1115 if (eos_mark_found) {
1118 } else if (!strcmp(*argj, "--")) {
1123 if (arg_match(&arg, &outputfile, argi)) {
1124 config->out_fn = arg.val;
1125 } else if (arg_match(&arg, &fpf_name, argi)) {
1126 config->stats_fn = arg.val;
1127 #if CONFIG_FP_MB_STATS
1128 } else if (arg_match(&arg, &fpmbf_name, argi)) {
1129 config->fpmb_stats_fn = arg.val;
1131 } else if (arg_match(&arg, &use_webm, argi)) {
1133 config->write_webm = 1;
1135 die("Error: --webm specified but webm is disabled.");
1137 } else if (arg_match(&arg, &use_ivf, argi)) {
1138 config->write_webm = 0;
1139 } else if (arg_match(&arg, &threads, argi)) {
1140 config->cfg.g_threads = arg_parse_uint(&arg);
1141 } else if (arg_match(&arg, &profile, argi)) {
1142 config->cfg.g_profile = arg_parse_uint(&arg);
1143 } else if (arg_match(&arg, &width, argi)) {
1144 config->cfg.g_w = arg_parse_uint(&arg);
1145 } else if (arg_match(&arg, &height, argi)) {
1146 config->cfg.g_h = arg_parse_uint(&arg);
1147 #if CONFIG_VP9_HIGHBITDEPTH
1148 } else if (arg_match(&arg, &bitdeptharg, argi)) {
1149 config->cfg.g_bit_depth = arg_parse_enum_or_int(&arg);
1150 } else if (arg_match(&arg, &inbitdeptharg, argi)) {
1151 config->cfg.g_input_bit_depth = arg_parse_uint(&arg);
1154 } else if (arg_match(&arg, &stereo_mode, argi)) {
1155 config->stereo_fmt = arg_parse_enum_or_int(&arg);
1157 } else if (arg_match(&arg, &timebase, argi)) {
1158 config->cfg.g_timebase = arg_parse_rational(&arg);
1159 validate_positive_rational(arg.name, &config->cfg.g_timebase);
1160 } else if (arg_match(&arg, &error_resilient, argi)) {
1161 config->cfg.g_error_resilient = arg_parse_uint(&arg);
1162 } else if (arg_match(&arg, &end_usage, argi)) {
1163 config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
1164 } else if (arg_match(&arg, &lag_in_frames, argi)) {
1165 config->cfg.g_lag_in_frames = arg_parse_uint(&arg);
1166 if (global->deadline == VPX_DL_REALTIME &&
1167 config->cfg.rc_end_usage == VPX_CBR &&
1168 config->cfg.g_lag_in_frames != 0) {
1169 warn("non-zero %s option ignored in realtime CBR mode.\n", arg.name);
1170 config->cfg.g_lag_in_frames = 0;
1172 } else if (arg_match(&arg, &dropframe_thresh, argi)) {
1173 config->cfg.rc_dropframe_thresh = arg_parse_uint(&arg);
1174 } else if (arg_match(&arg, &resize_allowed, argi)) {
1175 config->cfg.rc_resize_allowed = arg_parse_uint(&arg);
1176 } else if (arg_match(&arg, &resize_width, argi)) {
1177 config->cfg.rc_scaled_width = arg_parse_uint(&arg);
1178 } else if (arg_match(&arg, &resize_height, argi)) {
1179 config->cfg.rc_scaled_height = arg_parse_uint(&arg);
1180 } else if (arg_match(&arg, &resize_up_thresh, argi)) {
1181 config->cfg.rc_resize_up_thresh = arg_parse_uint(&arg);
1182 } else if (arg_match(&arg, &resize_down_thresh, argi)) {
1183 config->cfg.rc_resize_down_thresh = arg_parse_uint(&arg);
1184 } else if (arg_match(&arg, &end_usage, argi)) {
1185 config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
1186 } else if (arg_match(&arg, &target_bitrate, argi)) {
1187 config->cfg.rc_target_bitrate = arg_parse_uint(&arg);
1188 } else if (arg_match(&arg, &min_quantizer, argi)) {
1189 config->cfg.rc_min_quantizer = arg_parse_uint(&arg);
1190 } else if (arg_match(&arg, &max_quantizer, argi)) {
1191 config->cfg.rc_max_quantizer = arg_parse_uint(&arg);
1192 } else if (arg_match(&arg, &undershoot_pct, argi)) {
1193 config->cfg.rc_undershoot_pct = arg_parse_uint(&arg);
1194 } else if (arg_match(&arg, &overshoot_pct, argi)) {
1195 config->cfg.rc_overshoot_pct = arg_parse_uint(&arg);
1196 } else if (arg_match(&arg, &buf_sz, argi)) {
1197 config->cfg.rc_buf_sz = arg_parse_uint(&arg);
1198 } else if (arg_match(&arg, &buf_initial_sz, argi)) {
1199 config->cfg.rc_buf_initial_sz = arg_parse_uint(&arg);
1200 } else if (arg_match(&arg, &buf_optimal_sz, argi)) {
1201 config->cfg.rc_buf_optimal_sz = arg_parse_uint(&arg);
1202 } else if (arg_match(&arg, &bias_pct, argi)) {
1203 config->cfg.rc_2pass_vbr_bias_pct = arg_parse_uint(&arg);
1204 if (global->passes < 2)
1205 warn("option %s ignored in one-pass mode.\n", arg.name);
1206 } else if (arg_match(&arg, &minsection_pct, argi)) {
1207 config->cfg.rc_2pass_vbr_minsection_pct = arg_parse_uint(&arg);
1209 if (global->passes < 2)
1210 warn("option %s ignored in one-pass mode.\n", arg.name);
1211 } else if (arg_match(&arg, &maxsection_pct, argi)) {
1212 config->cfg.rc_2pass_vbr_maxsection_pct = arg_parse_uint(&arg);
1214 if (global->passes < 2)
1215 warn("option %s ignored in one-pass mode.\n", arg.name);
1216 } else if (arg_match(&arg, &kf_min_dist, argi)) {
1217 config->cfg.kf_min_dist = arg_parse_uint(&arg);
1218 } else if (arg_match(&arg, &kf_max_dist, argi)) {
1219 config->cfg.kf_max_dist = arg_parse_uint(&arg);
1220 } else if (arg_match(&arg, &kf_disabled, argi)) {
1221 config->cfg.kf_mode = VPX_KF_DISABLED;
1222 #if CONFIG_VP9_HIGHBITDEPTH
1223 } else if (arg_match(&arg, &test16bitinternalarg, argi)) {
1224 if (strcmp(global->codec->name, "vp9") == 0) {
1225 test_16bit_internal = 1;
1230 for (i = 0; ctrl_args[i]; i++) {
1231 if (arg_match(&arg, ctrl_args[i], argi)) {
1235 /* Point either to the next free element or the first
1236 * instance of this control.
1238 for (j = 0; j < config->arg_ctrl_cnt; j++)
1239 if (ctrl_args_map != NULL &&
1240 config->arg_ctrls[j][0] == ctrl_args_map[i])
1244 assert(j < (int)ARG_CTRL_CNT_MAX);
1245 if (ctrl_args_map != NULL && j < (int)ARG_CTRL_CNT_MAX) {
1246 config->arg_ctrls[j][0] = ctrl_args_map[i];
1247 config->arg_ctrls[j][1] = arg_parse_enum_or_int(&arg);
1248 if (j == config->arg_ctrl_cnt) config->arg_ctrl_cnt++;
1255 #if CONFIG_VP9_HIGHBITDEPTH
1256 if (strcmp(global->codec->name, "vp9") == 0) {
1257 config->use_16bit_internal =
1258 test_16bit_internal | (config->cfg.g_profile > 1);
1261 return eos_mark_found;
1264 #define FOREACH_STREAM(func) \
1266 struct stream_state *stream; \
1267 for (stream = streams; stream; stream = stream->next) { \
1272 static void validate_stream_config(const struct stream_state *stream,
1273 const struct VpxEncoderConfig *global) {
1274 const struct stream_state *streami;
1277 if (!stream->config.cfg.g_w || !stream->config.cfg.g_h)
1279 "Stream %d: Specify stream dimensions with --width (-w) "
1280 " and --height (-h)",
1283 // Check that the codec bit depth is greater than the input bit depth.
1284 if (stream->config.cfg.g_input_bit_depth >
1285 (unsigned int)stream->config.cfg.g_bit_depth) {
1286 fatal("Stream %d: codec bit depth (%d) less than input bit depth (%d)",
1287 stream->index, (int)stream->config.cfg.g_bit_depth,
1288 stream->config.cfg.g_input_bit_depth);
1291 for (streami = stream; streami; streami = streami->next) {
1292 /* All streams require output files */
1293 if (!streami->config.out_fn)
1294 fatal("Stream %d: Output file is required (specify with -o)",
1297 /* Check for two streams outputting to the same file */
1298 if (streami != stream) {
1299 const char *a = stream->config.out_fn;
1300 const char *b = streami->config.out_fn;
1301 if (!strcmp(a, b) && strcmp(a, "/dev/null") && strcmp(a, ":nul"))
1302 fatal("Stream %d: duplicate output file (from stream %d)",
1303 streami->index, stream->index);
1306 /* Check for two streams sharing a stats file. */
1307 if (streami != stream) {
1308 const char *a = stream->config.stats_fn;
1309 const char *b = streami->config.stats_fn;
1310 if (a && b && !strcmp(a, b))
1311 fatal("Stream %d: duplicate stats file (from stream %d)",
1312 streami->index, stream->index);
1315 #if CONFIG_FP_MB_STATS
1316 /* Check for two streams sharing a mb stats file. */
1317 if (streami != stream) {
1318 const char *a = stream->config.fpmb_stats_fn;
1319 const char *b = streami->config.fpmb_stats_fn;
1320 if (a && b && !strcmp(a, b))
1321 fatal("Stream %d: duplicate mb stats file (from stream %d)",
1322 streami->index, stream->index);
1328 static void set_stream_dimensions(struct stream_state *stream, unsigned int w,
1330 if (!stream->config.cfg.g_w) {
1331 if (!stream->config.cfg.g_h)
1332 stream->config.cfg.g_w = w;
1334 stream->config.cfg.g_w = w * stream->config.cfg.g_h / h;
1336 if (!stream->config.cfg.g_h) {
1337 stream->config.cfg.g_h = h * stream->config.cfg.g_w / w;
1341 static const char *file_type_to_string(enum VideoFileType t) {
1343 case FILE_TYPE_RAW: return "RAW";
1344 case FILE_TYPE_Y4M: return "Y4M";
1345 default: return "Other";
1349 static const char *image_format_to_string(vpx_img_fmt_t f) {
1351 case VPX_IMG_FMT_I420: return "I420";
1352 case VPX_IMG_FMT_I422: return "I422";
1353 case VPX_IMG_FMT_I444: return "I444";
1354 case VPX_IMG_FMT_I440: return "I440";
1355 case VPX_IMG_FMT_YV12: return "YV12";
1356 case VPX_IMG_FMT_I42016: return "I42016";
1357 case VPX_IMG_FMT_I42216: return "I42216";
1358 case VPX_IMG_FMT_I44416: return "I44416";
1359 case VPX_IMG_FMT_I44016: return "I44016";
1360 default: return "Other";
1364 static void show_stream_config(struct stream_state *stream,
1365 struct VpxEncoderConfig *global,
1366 struct VpxInputContext *input) {
1367 #define SHOW(field) \
1368 fprintf(stderr, " %-28s = %d\n", #field, stream->config.cfg.field)
1370 if (stream->index == 0) {
1371 fprintf(stderr, "Codec: %s\n",
1372 vpx_codec_iface_name(global->codec->codec_interface()));
1373 fprintf(stderr, "Source file: %s File Type: %s Format: %s\n",
1374 input->filename, file_type_to_string(input->file_type),
1375 image_format_to_string(input->fmt));
1377 if (stream->next || stream->index)
1378 fprintf(stderr, "\nStream Index: %d\n", stream->index);
1379 fprintf(stderr, "Destination file: %s\n", stream->config.out_fn);
1380 fprintf(stderr, "Encoder parameters:\n");
1388 SHOW(g_input_bit_depth);
1389 SHOW(g_timebase.num);
1390 SHOW(g_timebase.den);
1391 SHOW(g_error_resilient);
1393 SHOW(g_lag_in_frames);
1394 SHOW(rc_dropframe_thresh);
1395 SHOW(rc_resize_allowed);
1396 SHOW(rc_scaled_width);
1397 SHOW(rc_scaled_height);
1398 SHOW(rc_resize_up_thresh);
1399 SHOW(rc_resize_down_thresh);
1401 SHOW(rc_target_bitrate);
1402 SHOW(rc_min_quantizer);
1403 SHOW(rc_max_quantizer);
1404 SHOW(rc_undershoot_pct);
1405 SHOW(rc_overshoot_pct);
1407 SHOW(rc_buf_initial_sz);
1408 SHOW(rc_buf_optimal_sz);
1409 SHOW(rc_2pass_vbr_bias_pct);
1410 SHOW(rc_2pass_vbr_minsection_pct);
1411 SHOW(rc_2pass_vbr_maxsection_pct);
1417 static void open_output_file(struct stream_state *stream,
1418 struct VpxEncoderConfig *global,
1419 const struct VpxRational *pixel_aspect_ratio) {
1420 const char *fn = stream->config.out_fn;
1421 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1423 if (cfg->g_pass == VPX_RC_FIRST_PASS) return;
1425 stream->file = strcmp(fn, "-") ? fopen(fn, "wb") : set_binary_mode(stdout);
1427 if (!stream->file) fatal("Failed to open output file");
1429 if (stream->config.write_webm && fseek(stream->file, 0, SEEK_CUR))
1430 fatal("WebM output to pipes not supported.");
1433 if (stream->config.write_webm) {
1434 stream->webm_ctx.stream = stream->file;
1435 write_webm_file_header(&stream->webm_ctx, cfg, stream->config.stereo_fmt,
1436 global->codec->fourcc, pixel_aspect_ratio);
1439 (void)pixel_aspect_ratio;
1442 if (!stream->config.write_webm) {
1443 ivf_write_file_header(stream->file, cfg, global->codec->fourcc, 0);
1447 static void close_output_file(struct stream_state *stream,
1448 unsigned int fourcc) {
1449 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1451 if (cfg->g_pass == VPX_RC_FIRST_PASS) return;
1454 if (stream->config.write_webm) {
1455 write_webm_file_footer(&stream->webm_ctx);
1459 if (!stream->config.write_webm) {
1460 if (!fseek(stream->file, 0, SEEK_SET))
1461 ivf_write_file_header(stream->file, &stream->config.cfg, fourcc,
1462 stream->frames_out);
1465 fclose(stream->file);
1468 static void setup_pass(struct stream_state *stream,
1469 struct VpxEncoderConfig *global, int pass) {
1470 if (stream->config.stats_fn) {
1471 if (!stats_open_file(&stream->stats, stream->config.stats_fn, pass))
1472 fatal("Failed to open statistics store");
1474 if (!stats_open_mem(&stream->stats, pass))
1475 fatal("Failed to open statistics store");
1478 #if CONFIG_FP_MB_STATS
1479 if (stream->config.fpmb_stats_fn) {
1480 if (!stats_open_file(&stream->fpmb_stats, stream->config.fpmb_stats_fn,
1482 fatal("Failed to open mb statistics store");
1484 if (!stats_open_mem(&stream->fpmb_stats, pass))
1485 fatal("Failed to open mb statistics store");
1489 stream->config.cfg.g_pass = global->passes == 2
1490 ? pass ? VPX_RC_LAST_PASS : VPX_RC_FIRST_PASS
1493 stream->config.cfg.rc_twopass_stats_in = stats_get(&stream->stats);
1494 #if CONFIG_FP_MB_STATS
1495 stream->config.cfg.rc_firstpass_mb_stats_in =
1496 stats_get(&stream->fpmb_stats);
1500 stream->cx_time = 0;
1502 stream->frames_out = 0;
1505 static void initialize_encoder(struct stream_state *stream,
1506 struct VpxEncoderConfig *global) {
1510 flags |= global->show_psnr ? VPX_CODEC_USE_PSNR : 0;
1511 flags |= global->out_part ? VPX_CODEC_USE_OUTPUT_PARTITION : 0;
1512 #if CONFIG_VP9_HIGHBITDEPTH
1513 flags |= stream->config.use_16bit_internal ? VPX_CODEC_USE_HIGHBITDEPTH : 0;
1516 /* Construct Encoder Context */
1517 vpx_codec_enc_init(&stream->encoder, global->codec->codec_interface(),
1518 &stream->config.cfg, flags);
1519 ctx_exit_on_error(&stream->encoder, "Failed to initialize encoder");
1521 /* Note that we bypass the vpx_codec_control wrapper macro because
1522 * we're being clever to store the control IDs in an array. Real
1523 * applications will want to make use of the enumerations directly
1525 for (i = 0; i < stream->config.arg_ctrl_cnt; i++) {
1526 int ctrl = stream->config.arg_ctrls[i][0];
1527 int value = stream->config.arg_ctrls[i][1];
1528 if (vpx_codec_control_(&stream->encoder, ctrl, value))
1529 fprintf(stderr, "Error: Tried to set control %d = %d\n", ctrl, value);
1531 ctx_exit_on_error(&stream->encoder, "Failed to control codec");
1535 if (global->test_decode != TEST_DECODE_OFF) {
1536 const VpxInterface *decoder = get_vpx_decoder_by_name(global->codec->name);
1537 vpx_codec_dec_init(&stream->decoder, decoder->codec_interface(), NULL, 0);
1542 static void encode_frame(struct stream_state *stream,
1543 struct VpxEncoderConfig *global, struct vpx_image *img,
1544 unsigned int frames_in) {
1545 vpx_codec_pts_t frame_start, next_frame_start;
1546 struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1547 struct vpx_usec_timer timer;
1550 (cfg->g_timebase.den * (int64_t)(frames_in - 1) * global->framerate.den) /
1551 cfg->g_timebase.num / global->framerate.num;
1553 (cfg->g_timebase.den * (int64_t)(frames_in)*global->framerate.den) /
1554 cfg->g_timebase.num / global->framerate.num;
1556 /* Scale if necessary */
1557 #if CONFIG_VP9_HIGHBITDEPTH
1559 if ((img->fmt & VPX_IMG_FMT_HIGHBITDEPTH) &&
1560 (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1561 if (img->fmt != VPX_IMG_FMT_I42016) {
1562 fprintf(stderr, "%s can only scale 4:2:0 inputs\n", exec_name);
1568 vpx_img_alloc(NULL, VPX_IMG_FMT_I42016, cfg->g_w, cfg->g_h, 16);
1571 (uint16 *)img->planes[VPX_PLANE_Y], img->stride[VPX_PLANE_Y] / 2,
1572 (uint16 *)img->planes[VPX_PLANE_U], img->stride[VPX_PLANE_U] / 2,
1573 (uint16 *)img->planes[VPX_PLANE_V], img->stride[VPX_PLANE_V] / 2,
1574 img->d_w, img->d_h, (uint16 *)stream->img->planes[VPX_PLANE_Y],
1575 stream->img->stride[VPX_PLANE_Y] / 2,
1576 (uint16 *)stream->img->planes[VPX_PLANE_U],
1577 stream->img->stride[VPX_PLANE_U] / 2,
1578 (uint16 *)stream->img->planes[VPX_PLANE_V],
1579 stream->img->stride[VPX_PLANE_V] / 2, stream->img->d_w,
1580 stream->img->d_h, kFilterBox);
1583 stream->encoder.err = 1;
1584 ctx_exit_on_error(&stream->encoder,
1585 "Stream %d: Failed to encode frame.\n"
1586 "Scaling disabled in this configuration. \n"
1587 "To enable, configure with --enable-libyuv\n",
1593 if (img && (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1594 if (img->fmt != VPX_IMG_FMT_I420 && img->fmt != VPX_IMG_FMT_YV12) {
1595 fprintf(stderr, "%s can only scale 4:2:0 8bpp inputs\n", exec_name);
1601 vpx_img_alloc(NULL, VPX_IMG_FMT_I420, cfg->g_w, cfg->g_h, 16);
1603 img->planes[VPX_PLANE_Y], img->stride[VPX_PLANE_Y],
1604 img->planes[VPX_PLANE_U], img->stride[VPX_PLANE_U],
1605 img->planes[VPX_PLANE_V], img->stride[VPX_PLANE_V], img->d_w, img->d_h,
1606 stream->img->planes[VPX_PLANE_Y], stream->img->stride[VPX_PLANE_Y],
1607 stream->img->planes[VPX_PLANE_U], stream->img->stride[VPX_PLANE_U],
1608 stream->img->planes[VPX_PLANE_V], stream->img->stride[VPX_PLANE_V],
1609 stream->img->d_w, stream->img->d_h, kFilterBox);
1612 stream->encoder.err = 1;
1613 ctx_exit_on_error(&stream->encoder,
1614 "Stream %d: Failed to encode frame.\n"
1615 "Scaling disabled in this configuration. \n"
1616 "To enable, configure with --enable-libyuv\n",
1621 vpx_usec_timer_start(&timer);
1622 vpx_codec_encode(&stream->encoder, img, frame_start,
1623 (unsigned long)(next_frame_start - frame_start), 0,
1625 vpx_usec_timer_mark(&timer);
1626 stream->cx_time += vpx_usec_timer_elapsed(&timer);
1627 ctx_exit_on_error(&stream->encoder, "Stream %d: Failed to encode frame",
1631 static void update_quantizer_histogram(struct stream_state *stream) {
1632 if (stream->config.cfg.g_pass != VPX_RC_FIRST_PASS) {
1635 vpx_codec_control(&stream->encoder, VP8E_GET_LAST_QUANTIZER_64, &q);
1636 ctx_exit_on_error(&stream->encoder, "Failed to read quantizer");
1637 stream->counts[q]++;
1641 static void get_cx_data(struct stream_state *stream,
1642 struct VpxEncoderConfig *global, int *got_data) {
1643 const vpx_codec_cx_pkt_t *pkt;
1644 const struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1645 vpx_codec_iter_t iter = NULL;
1648 while ((pkt = vpx_codec_get_cx_data(&stream->encoder, &iter))) {
1649 static size_t fsize = 0;
1650 static int64_t ivf_header_pos = 0;
1652 switch (pkt->kind) {
1653 case VPX_CODEC_CX_FRAME_PKT:
1654 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1655 stream->frames_out++;
1658 fprintf(stderr, " %6luF", (unsigned long)pkt->data.frame.sz);
1660 update_rate_histogram(stream->rate_hist, cfg, pkt);
1662 if (stream->config.write_webm) {
1663 write_webm_block(&stream->webm_ctx, cfg, pkt);
1666 if (!stream->config.write_webm) {
1667 if (pkt->data.frame.partition_id <= 0) {
1668 ivf_header_pos = ftello(stream->file);
1669 fsize = pkt->data.frame.sz;
1671 ivf_write_frame_header(stream->file, pkt->data.frame.pts, fsize);
1673 fsize += pkt->data.frame.sz;
1675 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1676 const int64_t currpos = ftello(stream->file);
1677 fseeko(stream->file, ivf_header_pos, SEEK_SET);
1678 ivf_write_frame_size(stream->file, fsize);
1679 fseeko(stream->file, currpos, SEEK_SET);
1683 (void)fwrite(pkt->data.frame.buf, 1, pkt->data.frame.sz,
1686 stream->nbytes += pkt->data.raw.sz;
1690 if (global->test_decode != TEST_DECODE_OFF && !stream->mismatch_seen) {
1691 vpx_codec_decode(&stream->decoder, pkt->data.frame.buf,
1692 (unsigned int)pkt->data.frame.sz, NULL, 0);
1693 if (stream->decoder.err) {
1694 warn_or_exit_on_error(&stream->decoder,
1695 global->test_decode == TEST_DECODE_FATAL,
1696 "Failed to decode frame %d in stream %d",
1697 stream->frames_out + 1, stream->index);
1698 stream->mismatch_seen = stream->frames_out + 1;
1703 case VPX_CODEC_STATS_PKT:
1704 stream->frames_out++;
1705 stats_write(&stream->stats, pkt->data.twopass_stats.buf,
1706 pkt->data.twopass_stats.sz);
1707 stream->nbytes += pkt->data.raw.sz;
1709 #if CONFIG_FP_MB_STATS
1710 case VPX_CODEC_FPMB_STATS_PKT:
1711 stats_write(&stream->fpmb_stats, pkt->data.firstpass_mb_stats.buf,
1712 pkt->data.firstpass_mb_stats.sz);
1713 stream->nbytes += pkt->data.raw.sz;
1716 case VPX_CODEC_PSNR_PKT:
1718 if (global->show_psnr) {
1721 stream->psnr_sse_total += pkt->data.psnr.sse[0];
1722 stream->psnr_samples_total += pkt->data.psnr.samples[0];
1723 for (i = 0; i < 4; i++) {
1725 fprintf(stderr, "%.3f ", pkt->data.psnr.psnr[i]);
1726 stream->psnr_totals[i] += pkt->data.psnr.psnr[i];
1728 stream->psnr_count++;
1737 static void show_psnr(struct stream_state *stream, double peak) {
1741 if (!stream->psnr_count) return;
1743 fprintf(stderr, "Stream %d PSNR (Overall/Avg/Y/U/V)", stream->index);
1744 ovpsnr = sse_to_psnr((double)stream->psnr_samples_total, peak,
1745 (double)stream->psnr_sse_total);
1746 fprintf(stderr, " %.3f", ovpsnr);
1748 for (i = 0; i < 4; i++) {
1749 fprintf(stderr, " %.3f", stream->psnr_totals[i] / stream->psnr_count);
1751 fprintf(stderr, "\n");
1754 static float usec_to_fps(uint64_t usec, unsigned int frames) {
1755 return (float)(usec > 0 ? frames * 1000000.0 / (float)usec : 0);
1758 static void test_decode(struct stream_state *stream,
1759 enum TestDecodeFatality fatal,
1760 const VpxInterface *codec) {
1761 vpx_image_t enc_img, dec_img;
1763 if (stream->mismatch_seen) return;
1765 /* Get the internal reference frame */
1766 if (strcmp(codec->name, "vp8") == 0) {
1767 struct vpx_ref_frame ref_enc, ref_dec;
1770 width = (stream->config.cfg.g_w + 15) & ~15;
1771 height = (stream->config.cfg.g_h + 15) & ~15;
1772 vpx_img_alloc(&ref_enc.img, VPX_IMG_FMT_I420, width, height, 1);
1773 enc_img = ref_enc.img;
1774 vpx_img_alloc(&ref_dec.img, VPX_IMG_FMT_I420, width, height, 1);
1775 dec_img = ref_dec.img;
1777 ref_enc.frame_type = VP8_LAST_FRAME;
1778 ref_dec.frame_type = VP8_LAST_FRAME;
1779 vpx_codec_control(&stream->encoder, VP8_COPY_REFERENCE, &ref_enc);
1780 vpx_codec_control(&stream->decoder, VP8_COPY_REFERENCE, &ref_dec);
1782 struct vp9_ref_frame ref_enc, ref_dec;
1786 vpx_codec_control(&stream->encoder, VP9_GET_REFERENCE, &ref_enc);
1787 enc_img = ref_enc.img;
1788 vpx_codec_control(&stream->decoder, VP9_GET_REFERENCE, &ref_dec);
1789 dec_img = ref_dec.img;
1790 #if CONFIG_VP9_HIGHBITDEPTH
1791 if ((enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) !=
1792 (dec_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH)) {
1793 if (enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1794 vpx_img_alloc(&enc_img, enc_img.fmt - VPX_IMG_FMT_HIGHBITDEPTH,
1795 enc_img.d_w, enc_img.d_h, 16);
1796 vpx_img_truncate_16_to_8(&enc_img, &ref_enc.img);
1798 if (dec_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1799 vpx_img_alloc(&dec_img, dec_img.fmt - VPX_IMG_FMT_HIGHBITDEPTH,
1800 dec_img.d_w, dec_img.d_h, 16);
1801 vpx_img_truncate_16_to_8(&dec_img, &ref_dec.img);
1806 ctx_exit_on_error(&stream->encoder, "Failed to get encoder reference frame");
1807 ctx_exit_on_error(&stream->decoder, "Failed to get decoder reference frame");
1809 if (!compare_img(&enc_img, &dec_img)) {
1810 int y[4], u[4], v[4];
1811 #if CONFIG_VP9_HIGHBITDEPTH
1812 if (enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1813 find_mismatch_high(&enc_img, &dec_img, y, u, v);
1815 find_mismatch(&enc_img, &dec_img, y, u, v);
1818 find_mismatch(&enc_img, &dec_img, y, u, v);
1820 stream->decoder.err = 1;
1821 warn_or_exit_on_error(&stream->decoder, fatal == TEST_DECODE_FATAL,
1822 "Stream %d: Encode/decode mismatch on frame %d at"
1823 " Y[%d, %d] {%d/%d},"
1824 " U[%d, %d] {%d/%d},"
1825 " V[%d, %d] {%d/%d}",
1826 stream->index, stream->frames_out, y[0], y[1], y[2],
1827 y[3], u[0], u[1], u[2], u[3], v[0], v[1], v[2], v[3]);
1828 stream->mismatch_seen = stream->frames_out;
1831 vpx_img_free(&enc_img);
1832 vpx_img_free(&dec_img);
1835 static void print_time(const char *label, int64_t etl) {
1842 etl -= hours * 3600;
1847 fprintf(stderr, "[%3s %2" PRId64 ":%02" PRId64 ":%02" PRId64 "] ", label,
1850 fprintf(stderr, "[%3s unknown] ", label);
1854 int main(int argc, const char **argv_) {
1857 #if CONFIG_VP9_HIGHBITDEPTH
1858 vpx_image_t raw_shift;
1859 int allocated_raw_shift = 0;
1860 int use_16bit_internal = 0;
1861 int input_shift = 0;
1863 int frame_avail, got_data;
1865 struct VpxInputContext input;
1866 struct VpxEncoderConfig global;
1867 struct stream_state *streams = NULL;
1868 char **argv, **argi;
1869 uint64_t cx_time = 0;
1873 memset(&input, 0, sizeof(input));
1874 exec_name = argv_[0];
1876 if (argc < 3) usage_exit();
1878 /* Setup default input stream settings */
1879 input.framerate.numerator = 30;
1880 input.framerate.denominator = 1;
1881 input.only_i420 = 1;
1882 input.bit_depth = 0;
1884 /* First parse the global configuration values, because we want to apply
1885 * other parameters on top of the default configuration provided by the
1888 argv = argv_dup(argc - 1, argv_ + 1);
1889 parse_global_config(&global, argv);
1891 switch (global.color_type) {
1892 case I420: input.fmt = VPX_IMG_FMT_I420; break;
1893 case I422: input.fmt = VPX_IMG_FMT_I422; break;
1894 case I444: input.fmt = VPX_IMG_FMT_I444; break;
1895 case I440: input.fmt = VPX_IMG_FMT_I440; break;
1896 case YV12: input.fmt = VPX_IMG_FMT_YV12; break;
1900 /* Now parse each stream's parameters. Using a local scope here
1901 * due to the use of 'stream' as loop variable in FOREACH_STREAM
1904 struct stream_state *stream = NULL;
1907 stream = new_stream(&global, stream);
1909 if (!streams) streams = stream;
1910 } while (parse_stream_params(&global, stream, argv));
1913 /* Check for unrecognized options */
1914 for (argi = argv; *argi; argi++)
1915 if (argi[0][0] == '-' && argi[0][1])
1916 die("Error: Unrecognized option %s\n", *argi);
1918 FOREACH_STREAM(check_encoder_config(global.disable_warning_prompt, &global,
1919 &stream->config.cfg););
1921 /* Handle non-option arguments */
1922 input.filename = argv[0];
1924 if (!input.filename) usage_exit();
1926 /* Decide if other chroma subsamplings than 4:2:0 are supported */
1927 if (global.codec->fourcc == VP9_FOURCC) input.only_i420 = 0;
1929 for (pass = global.pass ? global.pass - 1 : 0; pass < global.passes; pass++) {
1930 int frames_in = 0, seen_frames = 0;
1931 int64_t estimated_time_left = -1;
1932 int64_t average_rate = -1;
1933 int64_t lagged_count = 0;
1935 open_input_file(&input);
1937 /* If the input file doesn't specify its w/h (raw files), try to get
1938 * the data from the first stream's configuration.
1940 if (!input.width || !input.height) {
1942 if (stream->config.cfg.g_w && stream->config.cfg.g_h) {
1943 input.width = stream->config.cfg.g_w;
1944 input.height = stream->config.cfg.g_h;
1950 /* Update stream configurations from the input file's parameters */
1951 if (!input.width || !input.height)
1953 "Specify stream dimensions with --width (-w) "
1954 " and --height (-h)");
1956 /* If input file does not specify bit-depth but input-bit-depth parameter
1957 * exists, assume that to be the input bit-depth. However, if the
1958 * input-bit-depth paramter does not exist, assume the input bit-depth
1959 * to be the same as the codec bit-depth.
1961 if (!input.bit_depth) {
1963 if (stream->config.cfg.g_input_bit_depth)
1964 input.bit_depth = stream->config.cfg.g_input_bit_depth;
1966 input.bit_depth = stream->config.cfg.g_input_bit_depth =
1967 (int)stream->config.cfg.g_bit_depth;
1969 if (input.bit_depth > 8) input.fmt |= VPX_IMG_FMT_HIGHBITDEPTH;
1972 { stream->config.cfg.g_input_bit_depth = input.bit_depth; });
1975 FOREACH_STREAM(set_stream_dimensions(stream, input.width, input.height));
1976 FOREACH_STREAM(validate_stream_config(stream, &global));
1978 /* Ensure that --passes and --pass are consistent. If --pass is set and
1979 * --passes=2, ensure --fpf was set.
1981 if (global.pass && global.passes == 2)
1983 if (!stream->config.stats_fn)
1984 die("Stream %d: Must specify --fpf when --pass=%d"
1985 " and --passes=2\n",
1986 stream->index, global.pass);
1991 if (stream->config.write_webm) {
1992 stream->config.write_webm = 0;
1994 "vpxenc was compiled without WebM container support."
1995 "Producing IVF output");
2000 /* Use the frame rate from the file only if none was specified
2001 * on the command-line.
2003 if (!global.have_framerate) {
2004 global.framerate.num = input.framerate.numerator;
2005 global.framerate.den = input.framerate.denominator;
2006 FOREACH_STREAM(stream->config.cfg.g_timebase.den = global.framerate.num;
2007 stream->config.cfg.g_timebase.num = global.framerate.den);
2010 /* Show configuration */
2011 if (global.verbose && pass == 0)
2012 FOREACH_STREAM(show_stream_config(stream, &global, &input));
2014 if (pass == (global.pass ? global.pass - 1 : 0)) {
2015 if (input.file_type == FILE_TYPE_Y4M)
2016 /*The Y4M reader does its own allocation.
2017 Just initialize this here to avoid problems if we never read any
2019 memset(&raw, 0, sizeof(raw));
2021 vpx_img_alloc(&raw, input.fmt, input.width, input.height, 32);
2023 FOREACH_STREAM(stream->rate_hist = init_rate_histogram(
2024 &stream->config.cfg, &global.framerate));
2027 FOREACH_STREAM(setup_pass(stream, &global, pass));
2029 open_output_file(stream, &global, &input.pixel_aspect_ratio));
2030 FOREACH_STREAM(initialize_encoder(stream, &global));
2032 #if CONFIG_VP9_HIGHBITDEPTH
2033 if (strcmp(global.codec->name, "vp9") == 0) {
2034 // Check to see if at least one stream uses 16 bit internal.
2035 // Currently assume that the bit_depths for all streams using
2036 // highbitdepth are the same.
2038 if (stream->config.use_16bit_internal) {
2039 use_16bit_internal = 1;
2041 if (stream->config.cfg.g_profile == 0) {
2044 input_shift = (int)stream->config.cfg.g_bit_depth -
2045 stream->config.cfg.g_input_bit_depth;
2054 while (frame_avail || got_data) {
2055 struct vpx_usec_timer timer;
2057 if (!global.limit || frames_in < global.limit) {
2058 frame_avail = read_frame(&input, &raw);
2060 if (frame_avail) frames_in++;
2062 frames_in > global.skip_frames ? frames_in - global.skip_frames : 0;
2064 if (!global.quiet) {
2065 float fps = usec_to_fps(cx_time, seen_frames);
2066 fprintf(stderr, "\rPass %d/%d ", pass + 1, global.passes);
2068 if (stream_cnt == 1)
2069 fprintf(stderr, "frame %4d/%-4d %7" PRId64 "B ", frames_in,
2070 streams->frames_out, (int64_t)streams->nbytes);
2072 fprintf(stderr, "frame %4d ", frames_in);
2074 fprintf(stderr, "%7" PRId64 " %s %.2f %s ",
2075 cx_time > 9999999 ? cx_time / 1000 : cx_time,
2076 cx_time > 9999999 ? "ms" : "us", fps >= 1.0 ? fps : fps * 60,
2077 fps >= 1.0 ? "fps" : "fpm");
2078 print_time("ETA", estimated_time_left);
2084 if (frames_in > global.skip_frames) {
2085 #if CONFIG_VP9_HIGHBITDEPTH
2086 vpx_image_t *frame_to_encode;
2087 if (input_shift || (use_16bit_internal && input.bit_depth == 8)) {
2088 assert(use_16bit_internal);
2089 // Input bit depth and stream bit depth do not match, so up
2090 // shift frame to stream bit depth
2091 if (!allocated_raw_shift) {
2092 vpx_img_alloc(&raw_shift, raw.fmt | VPX_IMG_FMT_HIGHBITDEPTH,
2093 input.width, input.height, 32);
2094 allocated_raw_shift = 1;
2096 vpx_img_upshift(&raw_shift, &raw, input_shift);
2097 frame_to_encode = &raw_shift;
2099 frame_to_encode = &raw;
2101 vpx_usec_timer_start(&timer);
2102 if (use_16bit_internal) {
2103 assert(frame_to_encode->fmt & VPX_IMG_FMT_HIGHBITDEPTH);
2105 if (stream->config.use_16bit_internal)
2106 encode_frame(stream, &global,
2107 frame_avail ? frame_to_encode : NULL, frames_in);
2112 assert((frame_to_encode->fmt & VPX_IMG_FMT_HIGHBITDEPTH) == 0);
2113 FOREACH_STREAM(encode_frame(stream, &global,
2114 frame_avail ? frame_to_encode : NULL,
2118 vpx_usec_timer_start(&timer);
2119 FOREACH_STREAM(encode_frame(stream, &global, frame_avail ? &raw : NULL,
2122 vpx_usec_timer_mark(&timer);
2123 cx_time += vpx_usec_timer_elapsed(&timer);
2125 FOREACH_STREAM(update_quantizer_histogram(stream));
2128 FOREACH_STREAM(get_cx_data(stream, &global, &got_data));
2130 if (!got_data && input.length && streams != NULL &&
2131 !streams->frames_out) {
2132 lagged_count = global.limit ? seen_frames : ftello(input.file);
2133 } else if (input.length) {
2138 const int64_t frame_in_lagged = (seen_frames - lagged_count) * 1000;
2140 rate = cx_time ? frame_in_lagged * (int64_t)1000000 / cx_time : 0;
2141 remaining = 1000 * (global.limit - global.skip_frames -
2142 seen_frames + lagged_count);
2144 const int64_t input_pos = ftello(input.file);
2145 const int64_t input_pos_lagged = input_pos - lagged_count;
2146 const int64_t limit = input.length;
2148 rate = cx_time ? input_pos_lagged * (int64_t)1000000 / cx_time : 0;
2149 remaining = limit - input_pos + lagged_count;
2153 (average_rate <= 0) ? rate : (average_rate * 7 + rate) / 8;
2154 estimated_time_left = average_rate ? remaining / average_rate : -1;
2157 if (got_data && global.test_decode != TEST_DECODE_OFF)
2158 FOREACH_STREAM(test_decode(stream, global.test_decode, global.codec));
2162 if (!global.quiet) fprintf(stderr, "\033[K");
2165 if (stream_cnt > 1) fprintf(stderr, "\n");
2167 if (!global.quiet) {
2168 FOREACH_STREAM(fprintf(
2169 stderr, "\rPass %d/%d frame %4d/%-4d %7" PRId64 "B %7" PRId64
2170 "b/f %7" PRId64 "b/s"
2171 " %7" PRId64 " %s (%.2f fps)\033[K\n",
2172 pass + 1, global.passes, frames_in, stream->frames_out,
2173 (int64_t)stream->nbytes,
2174 seen_frames ? (int64_t)(stream->nbytes * 8 / seen_frames) : 0,
2176 ? (int64_t)stream->nbytes * 8 * (int64_t)global.framerate.num /
2177 global.framerate.den / seen_frames
2179 stream->cx_time > 9999999 ? stream->cx_time / 1000 : stream->cx_time,
2180 stream->cx_time > 9999999 ? "ms" : "us",
2181 usec_to_fps(stream->cx_time, seen_frames)));
2184 if (global.show_psnr) {
2185 if (global.codec->fourcc == VP9_FOURCC) {
2187 show_psnr(stream, (1 << stream->config.cfg.g_input_bit_depth) - 1));
2189 FOREACH_STREAM(show_psnr(stream, 255.0));
2193 FOREACH_STREAM(vpx_codec_destroy(&stream->encoder));
2195 if (global.test_decode != TEST_DECODE_OFF) {
2196 FOREACH_STREAM(vpx_codec_destroy(&stream->decoder));
2199 close_input_file(&input);
2201 if (global.test_decode == TEST_DECODE_FATAL) {
2202 FOREACH_STREAM(res |= stream->mismatch_seen);
2204 FOREACH_STREAM(close_output_file(stream, global.codec->fourcc));
2206 FOREACH_STREAM(stats_close(&stream->stats, global.passes - 1));
2208 #if CONFIG_FP_MB_STATS
2209 FOREACH_STREAM(stats_close(&stream->fpmb_stats, global.passes - 1));
2212 if (global.pass) break;
2215 if (global.show_q_hist_buckets)
2217 show_q_histogram(stream->counts, global.show_q_hist_buckets));
2219 if (global.show_rate_hist_buckets)
2220 FOREACH_STREAM(show_rate_histogram(stream->rate_hist, &stream->config.cfg,
2221 global.show_rate_hist_buckets));
2222 FOREACH_STREAM(destroy_rate_histogram(stream->rate_hist));
2224 #if CONFIG_INTERNAL_STATS
2225 /* TODO(jkoleszar): This doesn't belong in this executable. Do it for now,
2226 * to match some existing utilities.
2228 if (!(global.pass == 1 && global.passes == 2))
2230 FILE *f = fopen("opsnr.stt", "a");
2231 if (stream->mismatch_seen) {
2232 fprintf(f, "First mismatch occurred in frame %d\n",
2233 stream->mismatch_seen);
2235 fprintf(f, "No mismatch detected in recon buffers\n");
2241 #if CONFIG_VP9_HIGHBITDEPTH
2242 if (allocated_raw_shift) vpx_img_free(&raw_shift);
2247 return res ? EXIT_FAILURE : EXIT_SUCCESS;