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)");
358 static const arg_def_t gf_cbr_boost_pct = ARG_DEF(
359 NULL, "gf-cbr-boost", 1, "Boost for Golden Frame in CBR mode (pct)");
361 #if CONFIG_VP8_ENCODER
362 static const arg_def_t cpu_used_vp8 =
363 ARG_DEF(NULL, "cpu-used", 1, "CPU Used (-16..16)");
364 static const arg_def_t token_parts =
365 ARG_DEF(NULL, "token-parts", 1, "Number of token partitions to use, log2");
366 static const arg_def_t screen_content_mode =
367 ARG_DEF(NULL, "screen-content-mode", 1, "Screen content mode");
368 static const arg_def_t *vp8_args[] = { &cpu_used_vp8,
381 &screen_content_mode,
383 static const int vp8_arg_ctrl_map[] = { VP8E_SET_CPUUSED,
384 VP8E_SET_ENABLEAUTOALTREF,
385 VP8E_SET_NOISE_SENSITIVITY,
387 VP8E_SET_STATIC_THRESHOLD,
388 VP8E_SET_TOKEN_PARTITIONS,
389 VP8E_SET_ARNR_MAXFRAMES,
390 VP8E_SET_ARNR_STRENGTH,
394 VP8E_SET_MAX_INTRA_BITRATE_PCT,
395 VP8E_SET_GF_CBR_BOOST_PCT,
396 VP8E_SET_SCREEN_CONTENT_MODE,
400 #if CONFIG_VP9_ENCODER
401 static const arg_def_t cpu_used_vp9 =
402 ARG_DEF(NULL, "cpu-used", 1, "CPU Used (-8..8)");
403 static const arg_def_t tile_cols =
404 ARG_DEF(NULL, "tile-columns", 1, "Number of tile columns to use, log2");
405 static const arg_def_t tile_rows =
406 ARG_DEF(NULL, "tile-rows", 1,
407 "Number of tile rows to use, log2 (set to 0 while threads > 1)");
408 static const arg_def_t lossless =
409 ARG_DEF(NULL, "lossless", 1, "Lossless mode (0: false (default), 1: true)");
410 static const arg_def_t frame_parallel_decoding = ARG_DEF(
411 NULL, "frame-parallel", 1, "Enable frame parallel decodability features");
412 static const arg_def_t aq_mode = ARG_DEF(
414 "Adaptive quantization mode (0: off (default), 1: variance 2: complexity, "
415 "3: cyclic refresh, 4: equator360)");
416 static const arg_def_t alt_ref_aq = ARG_DEF(NULL, "alt-ref-aq", 1,
417 "Special adaptive quantization for "
418 "the alternate reference frames.");
419 static const arg_def_t frame_periodic_boost =
420 ARG_DEF(NULL, "frame-boost", 1,
421 "Enable frame periodic boost (0: off (default), 1: on)");
422 static const arg_def_t max_inter_rate_pct =
423 ARG_DEF(NULL, "max-inter-rate", 1, "Max P-frame bitrate (pct)");
424 static const arg_def_t min_gf_interval = ARG_DEF(
425 NULL, "min-gf-interval", 1,
426 "min gf/arf frame interval (default 0, indicating in-built behavior)");
427 static const arg_def_t max_gf_interval = ARG_DEF(
428 NULL, "max-gf-interval", 1,
429 "max gf/arf frame interval (default 0, indicating in-built behavior)");
431 static const struct arg_enum_list color_space_enum[] = {
432 { "unknown", VPX_CS_UNKNOWN },
433 { "bt601", VPX_CS_BT_601 },
434 { "bt709", VPX_CS_BT_709 },
435 { "smpte170", VPX_CS_SMPTE_170 },
436 { "smpte240", VPX_CS_SMPTE_240 },
437 { "bt2020", VPX_CS_BT_2020 },
438 { "reserved", VPX_CS_RESERVED },
439 { "sRGB", VPX_CS_SRGB },
443 static const arg_def_t input_color_space =
444 ARG_DEF_ENUM(NULL, "color-space", 1, "The color space of input content:",
447 #if CONFIG_VP9_HIGHBITDEPTH
448 static const struct arg_enum_list bitdepth_enum[] = {
449 { "8", VPX_BITS_8 }, { "10", VPX_BITS_10 }, { "12", VPX_BITS_12 }, { NULL, 0 }
452 static const arg_def_t bitdeptharg = ARG_DEF_ENUM(
454 "Bit depth for codec (8 for version <=1, 10 or 12 for version 2)",
456 static const arg_def_t inbitdeptharg =
457 ARG_DEF(NULL, "input-bit-depth", 1, "Bit depth of input");
460 static const struct arg_enum_list tune_content_enum[] = {
461 { "default", VP9E_CONTENT_DEFAULT },
462 { "screen", VP9E_CONTENT_SCREEN },
466 static const arg_def_t tune_content = ARG_DEF_ENUM(
467 NULL, "tune-content", 1, "Tune content type", tune_content_enum);
469 static const arg_def_t target_level = ARG_DEF(
470 NULL, "target-level", 1,
471 "Target level (255: off (default); 0: only keep level stats; 10: level 1.0;"
472 " 11: level 1.1; ... 62: level 6.2)");
474 static const arg_def_t row_mt =
475 ARG_DEF(NULL, "row-mt", 1,
476 "Enable row based non-deterministic multi-threading in VP9");
479 #if CONFIG_VP9_ENCODER
480 static const arg_def_t *vp9_args[] = { &cpu_used_vp9,
495 &frame_parallel_decoding,
498 &frame_periodic_boost,
506 #if CONFIG_VP9_HIGHBITDEPTH
509 #endif // CONFIG_VP9_HIGHBITDEPTH
511 static const int vp9_arg_ctrl_map[] = { VP8E_SET_CPUUSED,
512 VP8E_SET_ENABLEAUTOALTREF,
514 VP8E_SET_STATIC_THRESHOLD,
515 VP9E_SET_TILE_COLUMNS,
517 VP8E_SET_ARNR_MAXFRAMES,
518 VP8E_SET_ARNR_STRENGTH,
522 VP8E_SET_MAX_INTRA_BITRATE_PCT,
523 VP9E_SET_MAX_INTER_BITRATE_PCT,
524 VP9E_SET_GF_CBR_BOOST_PCT,
526 VP9E_SET_FRAME_PARALLEL_DECODING,
529 VP9E_SET_FRAME_PERIODIC_BOOST,
530 VP9E_SET_NOISE_SENSITIVITY,
531 VP9E_SET_TUNE_CONTENT,
532 VP9E_SET_COLOR_SPACE,
533 VP9E_SET_MIN_GF_INTERVAL,
534 VP9E_SET_MAX_GF_INTERVAL,
535 VP9E_SET_TARGET_LEVEL,
540 static const arg_def_t *no_args[] = { NULL };
542 void usage_exit(void) {
544 const int num_encoder = get_vpx_encoder_count();
546 fprintf(stderr, "Usage: %s <options> -o dst_filename src_filename \n",
549 fprintf(stderr, "\nOptions:\n");
550 arg_show_usage(stderr, main_args);
551 fprintf(stderr, "\nEncoder Global Options:\n");
552 arg_show_usage(stderr, global_args);
553 fprintf(stderr, "\nRate Control Options:\n");
554 arg_show_usage(stderr, rc_args);
555 fprintf(stderr, "\nTwopass Rate Control Options:\n");
556 arg_show_usage(stderr, rc_twopass_args);
557 fprintf(stderr, "\nKeyframe Placement Options:\n");
558 arg_show_usage(stderr, kf_args);
559 #if CONFIG_VP8_ENCODER
560 fprintf(stderr, "\nVP8 Specific Options:\n");
561 arg_show_usage(stderr, vp8_args);
563 #if CONFIG_VP9_ENCODER
564 fprintf(stderr, "\nVP9 Specific Options:\n");
565 arg_show_usage(stderr, vp9_args);
568 "\nStream timebase (--timebase):\n"
569 " The desired precision of timestamps in the output, expressed\n"
570 " in fractional seconds. Default is 1/1000.\n");
571 fprintf(stderr, "\nIncluded encoders:\n\n");
573 for (i = 0; i < num_encoder; ++i) {
574 const VpxInterface *const encoder = get_vpx_encoder_by_index(i);
575 const char *defstr = (i == (num_encoder - 1)) ? "(default)" : "";
576 fprintf(stderr, " %-6s - %s %s\n", encoder->name,
577 vpx_codec_iface_name(encoder->codec_interface()), defstr);
579 fprintf(stderr, "\n ");
580 fprintf(stderr, "Use --codec to switch to a non-default encoder.\n\n");
585 #define mmin(a, b) ((a) < (b) ? (a) : (b))
587 #if CONFIG_VP9_HIGHBITDEPTH
588 static void find_mismatch_high(const vpx_image_t *const img1,
589 const vpx_image_t *const img2, int yloc[4],
590 int uloc[4], int vloc[4]) {
591 uint16_t *plane1, *plane2;
592 uint32_t stride1, stride2;
593 const uint32_t bsize = 64;
594 const uint32_t bsizey = bsize >> img1->y_chroma_shift;
595 const uint32_t bsizex = bsize >> img1->x_chroma_shift;
597 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
599 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
602 yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
603 plane1 = (uint16_t *)img1->planes[VPX_PLANE_Y];
604 plane2 = (uint16_t *)img2->planes[VPX_PLANE_Y];
605 stride1 = img1->stride[VPX_PLANE_Y] / 2;
606 stride2 = img2->stride[VPX_PLANE_Y] / 2;
607 for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
608 for (j = 0; match && j < img1->d_w; j += bsize) {
610 const int si = mmin(i + bsize, img1->d_h) - i;
611 const int sj = mmin(j + bsize, img1->d_w) - j;
612 for (k = 0; match && k < si; ++k) {
613 for (l = 0; match && l < sj; ++l) {
614 if (*(plane1 + (i + k) * stride1 + j + l) !=
615 *(plane2 + (i + k) * stride2 + j + l)) {
618 yloc[2] = *(plane1 + (i + k) * stride1 + j + l);
619 yloc[3] = *(plane2 + (i + k) * stride2 + j + l);
628 uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
629 plane1 = (uint16_t *)img1->planes[VPX_PLANE_U];
630 plane2 = (uint16_t *)img2->planes[VPX_PLANE_U];
631 stride1 = img1->stride[VPX_PLANE_U] / 2;
632 stride2 = img2->stride[VPX_PLANE_U] / 2;
633 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
634 for (j = 0; match && j < c_w; j += bsizex) {
636 const int si = mmin(i + bsizey, c_h - i);
637 const int sj = mmin(j + bsizex, c_w - j);
638 for (k = 0; match && k < si; ++k) {
639 for (l = 0; match && l < sj; ++l) {
640 if (*(plane1 + (i + k) * stride1 + j + l) !=
641 *(plane2 + (i + k) * stride2 + j + l)) {
644 uloc[2] = *(plane1 + (i + k) * stride1 + j + l);
645 uloc[3] = *(plane2 + (i + k) * stride2 + j + l);
654 vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
655 plane1 = (uint16_t *)img1->planes[VPX_PLANE_V];
656 plane2 = (uint16_t *)img2->planes[VPX_PLANE_V];
657 stride1 = img1->stride[VPX_PLANE_V] / 2;
658 stride2 = img2->stride[VPX_PLANE_V] / 2;
659 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
660 for (j = 0; match && j < c_w; j += bsizex) {
662 const int si = mmin(i + bsizey, c_h - i);
663 const int sj = mmin(j + bsizex, c_w - j);
664 for (k = 0; match && k < si; ++k) {
665 for (l = 0; match && l < sj; ++l) {
666 if (*(plane1 + (i + k) * stride1 + j + l) !=
667 *(plane2 + (i + k) * stride2 + j + l)) {
670 vloc[2] = *(plane1 + (i + k) * stride1 + j + l);
671 vloc[3] = *(plane2 + (i + k) * stride2 + j + l);
682 static void find_mismatch(const vpx_image_t *const img1,
683 const vpx_image_t *const img2, int yloc[4],
684 int uloc[4], int vloc[4]) {
685 const uint32_t bsize = 64;
686 const uint32_t bsizey = bsize >> img1->y_chroma_shift;
687 const uint32_t bsizex = bsize >> img1->x_chroma_shift;
689 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
691 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
694 yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
695 for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
696 for (j = 0; match && j < img1->d_w; j += bsize) {
698 const int si = mmin(i + bsize, img1->d_h) - i;
699 const int sj = mmin(j + bsize, img1->d_w) - j;
700 for (k = 0; match && k < si; ++k) {
701 for (l = 0; match && l < sj; ++l) {
702 if (*(img1->planes[VPX_PLANE_Y] +
703 (i + k) * img1->stride[VPX_PLANE_Y] + j + l) !=
704 *(img2->planes[VPX_PLANE_Y] +
705 (i + k) * img2->stride[VPX_PLANE_Y] + j + l)) {
708 yloc[2] = *(img1->planes[VPX_PLANE_Y] +
709 (i + k) * img1->stride[VPX_PLANE_Y] + j + l);
710 yloc[3] = *(img2->planes[VPX_PLANE_Y] +
711 (i + k) * img2->stride[VPX_PLANE_Y] + j + l);
720 uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
721 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
722 for (j = 0; match && j < c_w; j += bsizex) {
724 const int si = mmin(i + bsizey, c_h - i);
725 const int sj = mmin(j + bsizex, c_w - j);
726 for (k = 0; match && k < si; ++k) {
727 for (l = 0; match && l < sj; ++l) {
728 if (*(img1->planes[VPX_PLANE_U] +
729 (i + k) * img1->stride[VPX_PLANE_U] + j + l) !=
730 *(img2->planes[VPX_PLANE_U] +
731 (i + k) * img2->stride[VPX_PLANE_U] + j + l)) {
734 uloc[2] = *(img1->planes[VPX_PLANE_U] +
735 (i + k) * img1->stride[VPX_PLANE_U] + j + l);
736 uloc[3] = *(img2->planes[VPX_PLANE_U] +
737 (i + k) * img2->stride[VPX_PLANE_U] + j + l);
745 vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
746 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
747 for (j = 0; match && j < c_w; j += bsizex) {
749 const int si = mmin(i + bsizey, c_h - i);
750 const int sj = mmin(j + bsizex, c_w - j);
751 for (k = 0; match && k < si; ++k) {
752 for (l = 0; match && l < sj; ++l) {
753 if (*(img1->planes[VPX_PLANE_V] +
754 (i + k) * img1->stride[VPX_PLANE_V] + j + l) !=
755 *(img2->planes[VPX_PLANE_V] +
756 (i + k) * img2->stride[VPX_PLANE_V] + j + l)) {
759 vloc[2] = *(img1->planes[VPX_PLANE_V] +
760 (i + k) * img1->stride[VPX_PLANE_V] + j + l);
761 vloc[3] = *(img2->planes[VPX_PLANE_V] +
762 (i + k) * img2->stride[VPX_PLANE_V] + j + l);
772 static int compare_img(const vpx_image_t *const img1,
773 const vpx_image_t *const img2) {
774 uint32_t l_w = img1->d_w;
775 uint32_t c_w = (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
777 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
781 match &= (img1->fmt == img2->fmt);
782 match &= (img1->d_w == img2->d_w);
783 match &= (img1->d_h == img2->d_h);
784 #if CONFIG_VP9_HIGHBITDEPTH
785 if (img1->fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
791 for (i = 0; i < img1->d_h; ++i)
792 match &= (memcmp(img1->planes[VPX_PLANE_Y] + i * img1->stride[VPX_PLANE_Y],
793 img2->planes[VPX_PLANE_Y] + i * img2->stride[VPX_PLANE_Y],
796 for (i = 0; i < c_h; ++i)
797 match &= (memcmp(img1->planes[VPX_PLANE_U] + i * img1->stride[VPX_PLANE_U],
798 img2->planes[VPX_PLANE_U] + i * img2->stride[VPX_PLANE_U],
801 for (i = 0; i < c_h; ++i)
802 match &= (memcmp(img1->planes[VPX_PLANE_V] + i * img1->stride[VPX_PLANE_V],
803 img2->planes[VPX_PLANE_V] + i * img2->stride[VPX_PLANE_V],
809 #define NELEMENTS(x) (sizeof(x) / sizeof(x[0]))
810 #if CONFIG_VP9_ENCODER
811 #define ARG_CTRL_CNT_MAX NELEMENTS(vp9_arg_ctrl_map)
813 #define ARG_CTRL_CNT_MAX NELEMENTS(vp8_arg_ctrl_map)
817 typedef int stereo_format_t;
818 struct WebmOutputContext {
823 /* Per-stream configuration */
824 struct stream_config {
825 struct vpx_codec_enc_cfg cfg;
827 const char *stats_fn;
828 #if CONFIG_FP_MB_STATS
829 const char *fpmb_stats_fn;
831 stereo_format_t stereo_fmt;
832 int arg_ctrls[ARG_CTRL_CNT_MAX][2];
835 #if CONFIG_VP9_HIGHBITDEPTH
836 // whether to use 16bit internal buffers
837 int use_16bit_internal;
841 struct stream_state {
843 struct stream_state *next;
844 struct stream_config config;
846 struct rate_hist *rate_hist;
847 struct WebmOutputContext webm_ctx;
848 uint64_t psnr_sse_total;
849 uint64_t psnr_samples_total;
850 double psnr_totals[4];
853 vpx_codec_ctx_t encoder;
854 unsigned int frames_out;
858 #if CONFIG_FP_MB_STATS
859 stats_io_t fpmb_stats;
861 struct vpx_image *img;
862 vpx_codec_ctx_t decoder;
866 static void validate_positive_rational(const char *msg,
867 struct vpx_rational *rat) {
873 if (rat->num < 0) die("Error: %s must be positive\n", msg);
875 if (!rat->den) die("Error: %s has zero denominator\n", msg);
878 static void parse_global_config(struct VpxEncoderConfig *global, char **argv) {
881 const int num_encoder = get_vpx_encoder_count();
883 if (num_encoder < 1) die("Error: no valid encoder available\n");
885 /* Initialize default parameters */
886 memset(global, 0, sizeof(*global));
887 global->codec = get_vpx_encoder_by_index(num_encoder - 1);
889 global->color_type = I420;
890 /* Assign default deadline to good quality */
891 global->deadline = VPX_DL_GOOD_QUALITY;
893 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
896 if (arg_match(&arg, &codecarg, argi)) {
897 global->codec = get_vpx_encoder_by_name(arg.val);
899 die("Error: Unrecognized argument (%s) to --codec\n", arg.val);
900 } else if (arg_match(&arg, &passes, argi)) {
901 global->passes = arg_parse_uint(&arg);
903 if (global->passes < 1 || global->passes > 2)
904 die("Error: Invalid number of passes (%d)\n", global->passes);
905 } else if (arg_match(&arg, &pass_arg, argi)) {
906 global->pass = arg_parse_uint(&arg);
908 if (global->pass < 1 || global->pass > 2)
909 die("Error: Invalid pass selected (%d)\n", global->pass);
910 } else if (arg_match(&arg, &usage, argi))
911 global->usage = arg_parse_uint(&arg);
912 else if (arg_match(&arg, &deadline, argi))
913 global->deadline = arg_parse_uint(&arg);
914 else if (arg_match(&arg, &best_dl, argi))
915 global->deadline = VPX_DL_BEST_QUALITY;
916 else if (arg_match(&arg, &good_dl, argi))
917 global->deadline = VPX_DL_GOOD_QUALITY;
918 else if (arg_match(&arg, &rt_dl, argi))
919 global->deadline = VPX_DL_REALTIME;
920 else if (arg_match(&arg, &use_yv12, argi))
921 global->color_type = YV12;
922 else if (arg_match(&arg, &use_i420, argi))
923 global->color_type = I420;
924 else if (arg_match(&arg, &use_i422, argi))
925 global->color_type = I422;
926 else if (arg_match(&arg, &use_i444, argi))
927 global->color_type = I444;
928 else if (arg_match(&arg, &use_i440, argi))
929 global->color_type = I440;
930 else if (arg_match(&arg, &quietarg, argi))
932 else if (arg_match(&arg, &verbosearg, argi))
934 else if (arg_match(&arg, &limit, argi))
935 global->limit = arg_parse_uint(&arg);
936 else if (arg_match(&arg, &skip, argi))
937 global->skip_frames = arg_parse_uint(&arg);
938 else if (arg_match(&arg, &psnrarg, argi))
939 global->show_psnr = 1;
940 else if (arg_match(&arg, &recontest, argi))
941 global->test_decode = arg_parse_enum_or_int(&arg);
942 else if (arg_match(&arg, &framerate, argi)) {
943 global->framerate = arg_parse_rational(&arg);
944 validate_positive_rational(arg.name, &global->framerate);
945 global->have_framerate = 1;
946 } else if (arg_match(&arg, &out_part, argi))
947 global->out_part = 1;
948 else if (arg_match(&arg, &debugmode, argi))
950 else if (arg_match(&arg, &q_hist_n, argi))
951 global->show_q_hist_buckets = arg_parse_uint(&arg);
952 else if (arg_match(&arg, &rate_hist_n, argi))
953 global->show_rate_hist_buckets = arg_parse_uint(&arg);
954 else if (arg_match(&arg, &disable_warnings, argi))
955 global->disable_warnings = 1;
956 else if (arg_match(&arg, &disable_warning_prompt, argi))
957 global->disable_warning_prompt = 1;
963 /* DWIM: Assume the user meant passes=2 if pass=2 is specified */
964 if (global->pass > global->passes) {
965 warn("Assuming --pass=%d implies --passes=%d\n", global->pass,
967 global->passes = global->pass;
970 /* Validate global config */
971 if (global->passes == 0) {
972 #if CONFIG_VP9_ENCODER
973 // Make default VP9 passes = 2 until there is a better quality 1-pass
975 if (global->codec != NULL && global->codec->name != NULL)
976 global->passes = (strcmp(global->codec->name, "vp9") == 0 &&
977 global->deadline != VPX_DL_REALTIME)
985 if (global->deadline == VPX_DL_REALTIME && global->passes > 1) {
986 warn("Enforcing one-pass encoding in realtime mode\n");
991 static void open_input_file(struct VpxInputContext *input) {
992 /* Parse certain options from the input file, if possible */
993 input->file = strcmp(input->filename, "-") ? fopen(input->filename, "rb")
994 : set_binary_mode(stdin);
996 if (!input->file) fatal("Failed to open input file");
998 if (!fseeko(input->file, 0, SEEK_END)) {
999 /* Input file is seekable. Figure out how long it is, so we can get
1002 input->length = ftello(input->file);
1003 rewind(input->file);
1006 /* Default to 1:1 pixel aspect ratio. */
1007 input->pixel_aspect_ratio.numerator = 1;
1008 input->pixel_aspect_ratio.denominator = 1;
1010 /* For RAW input sources, these bytes will applied on the first frame
1013 input->detect.buf_read = fread(input->detect.buf, 1, 4, input->file);
1014 input->detect.position = 0;
1016 if (input->detect.buf_read == 4 && file_is_y4m(input->detect.buf)) {
1017 if (y4m_input_open(&input->y4m, input->file, input->detect.buf, 4,
1018 input->only_i420) >= 0) {
1019 input->file_type = FILE_TYPE_Y4M;
1020 input->width = input->y4m.pic_w;
1021 input->height = input->y4m.pic_h;
1022 input->pixel_aspect_ratio.numerator = input->y4m.par_n;
1023 input->pixel_aspect_ratio.denominator = input->y4m.par_d;
1024 input->framerate.numerator = input->y4m.fps_n;
1025 input->framerate.denominator = input->y4m.fps_d;
1026 input->fmt = input->y4m.vpx_fmt;
1027 input->bit_depth = input->y4m.bit_depth;
1029 fatal("Unsupported Y4M stream.");
1030 } else if (input->detect.buf_read == 4 && fourcc_is_ivf(input->detect.buf)) {
1031 fatal("IVF is not supported as input.");
1033 input->file_type = FILE_TYPE_RAW;
1037 static void close_input_file(struct VpxInputContext *input) {
1038 fclose(input->file);
1039 if (input->file_type == FILE_TYPE_Y4M) y4m_input_close(&input->y4m);
1042 static struct stream_state *new_stream(struct VpxEncoderConfig *global,
1043 struct stream_state *prev) {
1044 struct stream_state *stream;
1046 stream = calloc(1, sizeof(*stream));
1047 if (stream == NULL) {
1048 fatal("Failed to allocate new stream.");
1052 memcpy(stream, prev, sizeof(*stream));
1054 prev->next = stream;
1056 vpx_codec_err_t res;
1058 /* Populate encoder configuration */
1059 res = vpx_codec_enc_config_default(global->codec->codec_interface(),
1060 &stream->config.cfg, global->usage);
1061 if (res) fatal("Failed to get config: %s\n", vpx_codec_err_to_string(res));
1063 /* Change the default timebase to a high enough value so that the
1064 * encoder will always create strictly increasing timestamps.
1066 stream->config.cfg.g_timebase.den = 1000;
1068 /* Never use the library's default resolution, require it be parsed
1069 * from the file or set on the command line.
1071 stream->config.cfg.g_w = 0;
1072 stream->config.cfg.g_h = 0;
1074 /* Initialize remaining stream parameters */
1075 stream->config.write_webm = 1;
1077 stream->config.stereo_fmt = STEREO_FORMAT_MONO;
1078 stream->webm_ctx.last_pts_ns = -1;
1079 stream->webm_ctx.writer = NULL;
1080 stream->webm_ctx.segment = NULL;
1083 /* Allows removal of the application version from the EBML tags */
1084 stream->webm_ctx.debug = global->debug;
1086 /* Default lag_in_frames is 0 in realtime mode CBR mode*/
1087 if (global->deadline == VPX_DL_REALTIME &&
1088 stream->config.cfg.rc_end_usage == 1)
1089 stream->config.cfg.g_lag_in_frames = 0;
1092 /* Output files must be specified for each stream */
1093 stream->config.out_fn = NULL;
1095 stream->next = NULL;
1099 static int parse_stream_params(struct VpxEncoderConfig *global,
1100 struct stream_state *stream, char **argv) {
1101 char **argi, **argj;
1103 static const arg_def_t **ctrl_args = no_args;
1104 static const int *ctrl_args_map = NULL;
1105 struct stream_config *config = &stream->config;
1106 int eos_mark_found = 0;
1107 #if CONFIG_VP9_HIGHBITDEPTH
1108 int test_16bit_internal = 0;
1111 // Handle codec specific options
1113 #if CONFIG_VP8_ENCODER
1114 } else if (strcmp(global->codec->name, "vp8") == 0) {
1115 ctrl_args = vp8_args;
1116 ctrl_args_map = vp8_arg_ctrl_map;
1118 #if CONFIG_VP9_ENCODER
1119 } else if (strcmp(global->codec->name, "vp9") == 0) {
1120 ctrl_args = vp9_args;
1121 ctrl_args_map = vp9_arg_ctrl_map;
1125 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
1128 /* Once we've found an end-of-stream marker (--) we want to continue
1129 * shifting arguments but not consuming them.
1131 if (eos_mark_found) {
1134 } else if (!strcmp(*argj, "--")) {
1139 if (arg_match(&arg, &outputfile, argi)) {
1140 config->out_fn = arg.val;
1141 } else if (arg_match(&arg, &fpf_name, argi)) {
1142 config->stats_fn = arg.val;
1143 #if CONFIG_FP_MB_STATS
1144 } else if (arg_match(&arg, &fpmbf_name, argi)) {
1145 config->fpmb_stats_fn = arg.val;
1147 } else if (arg_match(&arg, &use_webm, argi)) {
1149 config->write_webm = 1;
1151 die("Error: --webm specified but webm is disabled.");
1153 } else if (arg_match(&arg, &use_ivf, argi)) {
1154 config->write_webm = 0;
1155 } else if (arg_match(&arg, &threads, argi)) {
1156 config->cfg.g_threads = arg_parse_uint(&arg);
1157 } else if (arg_match(&arg, &profile, argi)) {
1158 config->cfg.g_profile = arg_parse_uint(&arg);
1159 } else if (arg_match(&arg, &width, argi)) {
1160 config->cfg.g_w = arg_parse_uint(&arg);
1161 } else if (arg_match(&arg, &height, argi)) {
1162 config->cfg.g_h = arg_parse_uint(&arg);
1163 #if CONFIG_VP9_HIGHBITDEPTH
1164 } else if (arg_match(&arg, &bitdeptharg, argi)) {
1165 config->cfg.g_bit_depth = arg_parse_enum_or_int(&arg);
1166 } else if (arg_match(&arg, &inbitdeptharg, argi)) {
1167 config->cfg.g_input_bit_depth = arg_parse_uint(&arg);
1170 } else if (arg_match(&arg, &stereo_mode, argi)) {
1171 config->stereo_fmt = arg_parse_enum_or_int(&arg);
1173 } else if (arg_match(&arg, &timebase, argi)) {
1174 config->cfg.g_timebase = arg_parse_rational(&arg);
1175 validate_positive_rational(arg.name, &config->cfg.g_timebase);
1176 } else if (arg_match(&arg, &error_resilient, argi)) {
1177 config->cfg.g_error_resilient = arg_parse_uint(&arg);
1178 } else if (arg_match(&arg, &end_usage, argi)) {
1179 config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
1180 } else if (arg_match(&arg, &lag_in_frames, argi)) {
1181 config->cfg.g_lag_in_frames = arg_parse_uint(&arg);
1182 if (global->deadline == VPX_DL_REALTIME &&
1183 config->cfg.rc_end_usage == VPX_CBR &&
1184 config->cfg.g_lag_in_frames != 0) {
1185 warn("non-zero %s option ignored in realtime CBR mode.\n", arg.name);
1186 config->cfg.g_lag_in_frames = 0;
1188 } else if (arg_match(&arg, &dropframe_thresh, argi)) {
1189 config->cfg.rc_dropframe_thresh = arg_parse_uint(&arg);
1190 } else if (arg_match(&arg, &resize_allowed, argi)) {
1191 config->cfg.rc_resize_allowed = arg_parse_uint(&arg);
1192 } else if (arg_match(&arg, &resize_width, argi)) {
1193 config->cfg.rc_scaled_width = arg_parse_uint(&arg);
1194 } else if (arg_match(&arg, &resize_height, argi)) {
1195 config->cfg.rc_scaled_height = arg_parse_uint(&arg);
1196 } else if (arg_match(&arg, &resize_up_thresh, argi)) {
1197 config->cfg.rc_resize_up_thresh = arg_parse_uint(&arg);
1198 } else if (arg_match(&arg, &resize_down_thresh, argi)) {
1199 config->cfg.rc_resize_down_thresh = arg_parse_uint(&arg);
1200 } else if (arg_match(&arg, &end_usage, argi)) {
1201 config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
1202 } else if (arg_match(&arg, &target_bitrate, argi)) {
1203 config->cfg.rc_target_bitrate = arg_parse_uint(&arg);
1204 } else if (arg_match(&arg, &min_quantizer, argi)) {
1205 config->cfg.rc_min_quantizer = arg_parse_uint(&arg);
1206 } else if (arg_match(&arg, &max_quantizer, argi)) {
1207 config->cfg.rc_max_quantizer = arg_parse_uint(&arg);
1208 } else if (arg_match(&arg, &undershoot_pct, argi)) {
1209 config->cfg.rc_undershoot_pct = arg_parse_uint(&arg);
1210 } else if (arg_match(&arg, &overshoot_pct, argi)) {
1211 config->cfg.rc_overshoot_pct = arg_parse_uint(&arg);
1212 } else if (arg_match(&arg, &buf_sz, argi)) {
1213 config->cfg.rc_buf_sz = arg_parse_uint(&arg);
1214 } else if (arg_match(&arg, &buf_initial_sz, argi)) {
1215 config->cfg.rc_buf_initial_sz = arg_parse_uint(&arg);
1216 } else if (arg_match(&arg, &buf_optimal_sz, argi)) {
1217 config->cfg.rc_buf_optimal_sz = arg_parse_uint(&arg);
1218 } else if (arg_match(&arg, &bias_pct, argi)) {
1219 config->cfg.rc_2pass_vbr_bias_pct = arg_parse_uint(&arg);
1220 if (global->passes < 2)
1221 warn("option %s ignored in one-pass mode.\n", arg.name);
1222 } else if (arg_match(&arg, &minsection_pct, argi)) {
1223 config->cfg.rc_2pass_vbr_minsection_pct = arg_parse_uint(&arg);
1225 if (global->passes < 2)
1226 warn("option %s ignored in one-pass mode.\n", arg.name);
1227 } else if (arg_match(&arg, &maxsection_pct, argi)) {
1228 config->cfg.rc_2pass_vbr_maxsection_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, &kf_min_dist, argi)) {
1233 config->cfg.kf_min_dist = arg_parse_uint(&arg);
1234 } else if (arg_match(&arg, &kf_max_dist, argi)) {
1235 config->cfg.kf_max_dist = arg_parse_uint(&arg);
1236 } else if (arg_match(&arg, &kf_disabled, argi)) {
1237 config->cfg.kf_mode = VPX_KF_DISABLED;
1238 #if CONFIG_VP9_HIGHBITDEPTH
1239 } else if (arg_match(&arg, &test16bitinternalarg, argi)) {
1240 if (strcmp(global->codec->name, "vp9") == 0) {
1241 test_16bit_internal = 1;
1246 for (i = 0; ctrl_args[i]; i++) {
1247 if (arg_match(&arg, ctrl_args[i], argi)) {
1251 /* Point either to the next free element or the first
1252 * instance of this control.
1254 for (j = 0; j < config->arg_ctrl_cnt; j++)
1255 if (ctrl_args_map != NULL &&
1256 config->arg_ctrls[j][0] == ctrl_args_map[i])
1260 assert(j < (int)ARG_CTRL_CNT_MAX);
1261 if (ctrl_args_map != NULL && j < (int)ARG_CTRL_CNT_MAX) {
1262 config->arg_ctrls[j][0] = ctrl_args_map[i];
1263 config->arg_ctrls[j][1] = arg_parse_enum_or_int(&arg);
1264 if (j == config->arg_ctrl_cnt) config->arg_ctrl_cnt++;
1271 #if CONFIG_VP9_HIGHBITDEPTH
1272 if (strcmp(global->codec->name, "vp9") == 0) {
1273 config->use_16bit_internal =
1274 test_16bit_internal | (config->cfg.g_profile > 1);
1277 return eos_mark_found;
1280 #define FOREACH_STREAM(func) \
1282 struct stream_state *stream; \
1283 for (stream = streams; stream; stream = stream->next) { \
1288 static void validate_stream_config(const struct stream_state *stream,
1289 const struct VpxEncoderConfig *global) {
1290 const struct stream_state *streami;
1293 if (!stream->config.cfg.g_w || !stream->config.cfg.g_h)
1295 "Stream %d: Specify stream dimensions with --width (-w) "
1296 " and --height (-h)",
1299 // Check that the codec bit depth is greater than the input bit depth.
1300 if (stream->config.cfg.g_input_bit_depth >
1301 (unsigned int)stream->config.cfg.g_bit_depth) {
1302 fatal("Stream %d: codec bit depth (%d) less than input bit depth (%d)",
1303 stream->index, (int)stream->config.cfg.g_bit_depth,
1304 stream->config.cfg.g_input_bit_depth);
1307 for (streami = stream; streami; streami = streami->next) {
1308 /* All streams require output files */
1309 if (!streami->config.out_fn)
1310 fatal("Stream %d: Output file is required (specify with -o)",
1313 /* Check for two streams outputting to the same file */
1314 if (streami != stream) {
1315 const char *a = stream->config.out_fn;
1316 const char *b = streami->config.out_fn;
1317 if (!strcmp(a, b) && strcmp(a, "/dev/null") && strcmp(a, ":nul"))
1318 fatal("Stream %d: duplicate output file (from stream %d)",
1319 streami->index, stream->index);
1322 /* Check for two streams sharing a stats file. */
1323 if (streami != stream) {
1324 const char *a = stream->config.stats_fn;
1325 const char *b = streami->config.stats_fn;
1326 if (a && b && !strcmp(a, b))
1327 fatal("Stream %d: duplicate stats file (from stream %d)",
1328 streami->index, stream->index);
1331 #if CONFIG_FP_MB_STATS
1332 /* Check for two streams sharing a mb stats file. */
1333 if (streami != stream) {
1334 const char *a = stream->config.fpmb_stats_fn;
1335 const char *b = streami->config.fpmb_stats_fn;
1336 if (a && b && !strcmp(a, b))
1337 fatal("Stream %d: duplicate mb stats file (from stream %d)",
1338 streami->index, stream->index);
1344 static void set_stream_dimensions(struct stream_state *stream, unsigned int w,
1346 if (!stream->config.cfg.g_w) {
1347 if (!stream->config.cfg.g_h)
1348 stream->config.cfg.g_w = w;
1350 stream->config.cfg.g_w = w * stream->config.cfg.g_h / h;
1352 if (!stream->config.cfg.g_h) {
1353 stream->config.cfg.g_h = h * stream->config.cfg.g_w / w;
1357 static const char *file_type_to_string(enum VideoFileType t) {
1359 case FILE_TYPE_RAW: return "RAW";
1360 case FILE_TYPE_Y4M: return "Y4M";
1361 default: return "Other";
1365 static const char *image_format_to_string(vpx_img_fmt_t f) {
1367 case VPX_IMG_FMT_I420: return "I420";
1368 case VPX_IMG_FMT_I422: return "I422";
1369 case VPX_IMG_FMT_I444: return "I444";
1370 case VPX_IMG_FMT_I440: return "I440";
1371 case VPX_IMG_FMT_YV12: return "YV12";
1372 case VPX_IMG_FMT_I42016: return "I42016";
1373 case VPX_IMG_FMT_I42216: return "I42216";
1374 case VPX_IMG_FMT_I44416: return "I44416";
1375 case VPX_IMG_FMT_I44016: return "I44016";
1376 default: return "Other";
1380 static void show_stream_config(struct stream_state *stream,
1381 struct VpxEncoderConfig *global,
1382 struct VpxInputContext *input) {
1383 #define SHOW(field) \
1384 fprintf(stderr, " %-28s = %d\n", #field, stream->config.cfg.field)
1386 if (stream->index == 0) {
1387 fprintf(stderr, "Codec: %s\n",
1388 vpx_codec_iface_name(global->codec->codec_interface()));
1389 fprintf(stderr, "Source file: %s File Type: %s Format: %s\n",
1390 input->filename, file_type_to_string(input->file_type),
1391 image_format_to_string(input->fmt));
1393 if (stream->next || stream->index)
1394 fprintf(stderr, "\nStream Index: %d\n", stream->index);
1395 fprintf(stderr, "Destination file: %s\n", stream->config.out_fn);
1396 fprintf(stderr, "Encoder parameters:\n");
1404 SHOW(g_input_bit_depth);
1405 SHOW(g_timebase.num);
1406 SHOW(g_timebase.den);
1407 SHOW(g_error_resilient);
1409 SHOW(g_lag_in_frames);
1410 SHOW(rc_dropframe_thresh);
1411 SHOW(rc_resize_allowed);
1412 SHOW(rc_scaled_width);
1413 SHOW(rc_scaled_height);
1414 SHOW(rc_resize_up_thresh);
1415 SHOW(rc_resize_down_thresh);
1417 SHOW(rc_target_bitrate);
1418 SHOW(rc_min_quantizer);
1419 SHOW(rc_max_quantizer);
1420 SHOW(rc_undershoot_pct);
1421 SHOW(rc_overshoot_pct);
1423 SHOW(rc_buf_initial_sz);
1424 SHOW(rc_buf_optimal_sz);
1425 SHOW(rc_2pass_vbr_bias_pct);
1426 SHOW(rc_2pass_vbr_minsection_pct);
1427 SHOW(rc_2pass_vbr_maxsection_pct);
1433 static void open_output_file(struct stream_state *stream,
1434 struct VpxEncoderConfig *global,
1435 const struct VpxRational *pixel_aspect_ratio) {
1436 const char *fn = stream->config.out_fn;
1437 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1439 if (cfg->g_pass == VPX_RC_FIRST_PASS) return;
1441 stream->file = strcmp(fn, "-") ? fopen(fn, "wb") : set_binary_mode(stdout);
1443 if (!stream->file) fatal("Failed to open output file");
1445 if (stream->config.write_webm && fseek(stream->file, 0, SEEK_CUR))
1446 fatal("WebM output to pipes not supported.");
1449 if (stream->config.write_webm) {
1450 stream->webm_ctx.stream = stream->file;
1451 write_webm_file_header(&stream->webm_ctx, cfg, stream->config.stereo_fmt,
1452 global->codec->fourcc, pixel_aspect_ratio);
1455 (void)pixel_aspect_ratio;
1458 if (!stream->config.write_webm) {
1459 ivf_write_file_header(stream->file, cfg, global->codec->fourcc, 0);
1463 static void close_output_file(struct stream_state *stream,
1464 unsigned int fourcc) {
1465 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1467 if (cfg->g_pass == VPX_RC_FIRST_PASS) return;
1470 if (stream->config.write_webm) {
1471 write_webm_file_footer(&stream->webm_ctx);
1475 if (!stream->config.write_webm) {
1476 if (!fseek(stream->file, 0, SEEK_SET))
1477 ivf_write_file_header(stream->file, &stream->config.cfg, fourcc,
1478 stream->frames_out);
1481 fclose(stream->file);
1484 static void setup_pass(struct stream_state *stream,
1485 struct VpxEncoderConfig *global, int pass) {
1486 if (stream->config.stats_fn) {
1487 if (!stats_open_file(&stream->stats, stream->config.stats_fn, pass))
1488 fatal("Failed to open statistics store");
1490 if (!stats_open_mem(&stream->stats, pass))
1491 fatal("Failed to open statistics store");
1494 #if CONFIG_FP_MB_STATS
1495 if (stream->config.fpmb_stats_fn) {
1496 if (!stats_open_file(&stream->fpmb_stats, stream->config.fpmb_stats_fn,
1498 fatal("Failed to open mb statistics store");
1500 if (!stats_open_mem(&stream->fpmb_stats, pass))
1501 fatal("Failed to open mb statistics store");
1505 stream->config.cfg.g_pass = global->passes == 2
1506 ? pass ? VPX_RC_LAST_PASS : VPX_RC_FIRST_PASS
1509 stream->config.cfg.rc_twopass_stats_in = stats_get(&stream->stats);
1510 #if CONFIG_FP_MB_STATS
1511 stream->config.cfg.rc_firstpass_mb_stats_in =
1512 stats_get(&stream->fpmb_stats);
1516 stream->cx_time = 0;
1518 stream->frames_out = 0;
1521 static void initialize_encoder(struct stream_state *stream,
1522 struct VpxEncoderConfig *global) {
1526 flags |= global->show_psnr ? VPX_CODEC_USE_PSNR : 0;
1527 flags |= global->out_part ? VPX_CODEC_USE_OUTPUT_PARTITION : 0;
1528 #if CONFIG_VP9_HIGHBITDEPTH
1529 flags |= stream->config.use_16bit_internal ? VPX_CODEC_USE_HIGHBITDEPTH : 0;
1532 /* Construct Encoder Context */
1533 vpx_codec_enc_init(&stream->encoder, global->codec->codec_interface(),
1534 &stream->config.cfg, flags);
1535 ctx_exit_on_error(&stream->encoder, "Failed to initialize encoder");
1537 /* Note that we bypass the vpx_codec_control wrapper macro because
1538 * we're being clever to store the control IDs in an array. Real
1539 * applications will want to make use of the enumerations directly
1541 for (i = 0; i < stream->config.arg_ctrl_cnt; i++) {
1542 int ctrl = stream->config.arg_ctrls[i][0];
1543 int value = stream->config.arg_ctrls[i][1];
1544 if (vpx_codec_control_(&stream->encoder, ctrl, value))
1545 fprintf(stderr, "Error: Tried to set control %d = %d\n", ctrl, value);
1547 ctx_exit_on_error(&stream->encoder, "Failed to control codec");
1551 if (global->test_decode != TEST_DECODE_OFF) {
1552 const VpxInterface *decoder = get_vpx_decoder_by_name(global->codec->name);
1553 vpx_codec_dec_init(&stream->decoder, decoder->codec_interface(), NULL, 0);
1558 static void encode_frame(struct stream_state *stream,
1559 struct VpxEncoderConfig *global, struct vpx_image *img,
1560 unsigned int frames_in) {
1561 vpx_codec_pts_t frame_start, next_frame_start;
1562 struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1563 struct vpx_usec_timer timer;
1566 (cfg->g_timebase.den * (int64_t)(frames_in - 1) * global->framerate.den) /
1567 cfg->g_timebase.num / global->framerate.num;
1569 (cfg->g_timebase.den * (int64_t)(frames_in)*global->framerate.den) /
1570 cfg->g_timebase.num / global->framerate.num;
1572 /* Scale if necessary */
1573 #if CONFIG_VP9_HIGHBITDEPTH
1575 if ((img->fmt & VPX_IMG_FMT_HIGHBITDEPTH) &&
1576 (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1577 if (img->fmt != VPX_IMG_FMT_I42016) {
1578 fprintf(stderr, "%s can only scale 4:2:0 inputs\n", exec_name);
1584 vpx_img_alloc(NULL, VPX_IMG_FMT_I42016, cfg->g_w, cfg->g_h, 16);
1587 (uint16 *)img->planes[VPX_PLANE_Y], img->stride[VPX_PLANE_Y] / 2,
1588 (uint16 *)img->planes[VPX_PLANE_U], img->stride[VPX_PLANE_U] / 2,
1589 (uint16 *)img->planes[VPX_PLANE_V], img->stride[VPX_PLANE_V] / 2,
1590 img->d_w, img->d_h, (uint16 *)stream->img->planes[VPX_PLANE_Y],
1591 stream->img->stride[VPX_PLANE_Y] / 2,
1592 (uint16 *)stream->img->planes[VPX_PLANE_U],
1593 stream->img->stride[VPX_PLANE_U] / 2,
1594 (uint16 *)stream->img->planes[VPX_PLANE_V],
1595 stream->img->stride[VPX_PLANE_V] / 2, stream->img->d_w,
1596 stream->img->d_h, kFilterBox);
1599 stream->encoder.err = 1;
1600 ctx_exit_on_error(&stream->encoder,
1601 "Stream %d: Failed to encode frame.\n"
1602 "Scaling disabled in this configuration. \n"
1603 "To enable, configure with --enable-libyuv\n",
1609 if (img && (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1610 if (img->fmt != VPX_IMG_FMT_I420 && img->fmt != VPX_IMG_FMT_YV12) {
1611 fprintf(stderr, "%s can only scale 4:2:0 8bpp inputs\n", exec_name);
1617 vpx_img_alloc(NULL, VPX_IMG_FMT_I420, cfg->g_w, cfg->g_h, 16);
1619 img->planes[VPX_PLANE_Y], img->stride[VPX_PLANE_Y],
1620 img->planes[VPX_PLANE_U], img->stride[VPX_PLANE_U],
1621 img->planes[VPX_PLANE_V], img->stride[VPX_PLANE_V], img->d_w, img->d_h,
1622 stream->img->planes[VPX_PLANE_Y], stream->img->stride[VPX_PLANE_Y],
1623 stream->img->planes[VPX_PLANE_U], stream->img->stride[VPX_PLANE_U],
1624 stream->img->planes[VPX_PLANE_V], stream->img->stride[VPX_PLANE_V],
1625 stream->img->d_w, stream->img->d_h, kFilterBox);
1628 stream->encoder.err = 1;
1629 ctx_exit_on_error(&stream->encoder,
1630 "Stream %d: Failed to encode frame.\n"
1631 "Scaling disabled in this configuration. \n"
1632 "To enable, configure with --enable-libyuv\n",
1637 vpx_usec_timer_start(&timer);
1638 vpx_codec_encode(&stream->encoder, img, frame_start,
1639 (unsigned long)(next_frame_start - frame_start), 0,
1641 vpx_usec_timer_mark(&timer);
1642 stream->cx_time += vpx_usec_timer_elapsed(&timer);
1643 ctx_exit_on_error(&stream->encoder, "Stream %d: Failed to encode frame",
1647 static void update_quantizer_histogram(struct stream_state *stream) {
1648 if (stream->config.cfg.g_pass != VPX_RC_FIRST_PASS) {
1651 vpx_codec_control(&stream->encoder, VP8E_GET_LAST_QUANTIZER_64, &q);
1652 ctx_exit_on_error(&stream->encoder, "Failed to read quantizer");
1653 stream->counts[q]++;
1657 static void get_cx_data(struct stream_state *stream,
1658 struct VpxEncoderConfig *global, int *got_data) {
1659 const vpx_codec_cx_pkt_t *pkt;
1660 const struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1661 vpx_codec_iter_t iter = NULL;
1664 while ((pkt = vpx_codec_get_cx_data(&stream->encoder, &iter))) {
1665 static size_t fsize = 0;
1666 static FileOffset ivf_header_pos = 0;
1668 switch (pkt->kind) {
1669 case VPX_CODEC_CX_FRAME_PKT:
1670 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1671 stream->frames_out++;
1674 fprintf(stderr, " %6luF", (unsigned long)pkt->data.frame.sz);
1676 update_rate_histogram(stream->rate_hist, cfg, pkt);
1678 if (stream->config.write_webm) {
1679 write_webm_block(&stream->webm_ctx, cfg, pkt);
1682 if (!stream->config.write_webm) {
1683 if (pkt->data.frame.partition_id <= 0) {
1684 ivf_header_pos = ftello(stream->file);
1685 fsize = pkt->data.frame.sz;
1687 ivf_write_frame_header(stream->file, pkt->data.frame.pts, fsize);
1689 fsize += pkt->data.frame.sz;
1691 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1692 const FileOffset currpos = ftello(stream->file);
1693 fseeko(stream->file, ivf_header_pos, SEEK_SET);
1694 ivf_write_frame_size(stream->file, fsize);
1695 fseeko(stream->file, currpos, SEEK_SET);
1699 (void)fwrite(pkt->data.frame.buf, 1, pkt->data.frame.sz,
1702 stream->nbytes += pkt->data.raw.sz;
1706 if (global->test_decode != TEST_DECODE_OFF && !stream->mismatch_seen) {
1707 vpx_codec_decode(&stream->decoder, pkt->data.frame.buf,
1708 (unsigned int)pkt->data.frame.sz, NULL, 0);
1709 if (stream->decoder.err) {
1710 warn_or_exit_on_error(&stream->decoder,
1711 global->test_decode == TEST_DECODE_FATAL,
1712 "Failed to decode frame %d in stream %d",
1713 stream->frames_out + 1, stream->index);
1714 stream->mismatch_seen = stream->frames_out + 1;
1719 case VPX_CODEC_STATS_PKT:
1720 stream->frames_out++;
1721 stats_write(&stream->stats, pkt->data.twopass_stats.buf,
1722 pkt->data.twopass_stats.sz);
1723 stream->nbytes += pkt->data.raw.sz;
1725 #if CONFIG_FP_MB_STATS
1726 case VPX_CODEC_FPMB_STATS_PKT:
1727 stats_write(&stream->fpmb_stats, pkt->data.firstpass_mb_stats.buf,
1728 pkt->data.firstpass_mb_stats.sz);
1729 stream->nbytes += pkt->data.raw.sz;
1732 case VPX_CODEC_PSNR_PKT:
1734 if (global->show_psnr) {
1737 stream->psnr_sse_total += pkt->data.psnr.sse[0];
1738 stream->psnr_samples_total += pkt->data.psnr.samples[0];
1739 for (i = 0; i < 4; i++) {
1741 fprintf(stderr, "%.3f ", pkt->data.psnr.psnr[i]);
1742 stream->psnr_totals[i] += pkt->data.psnr.psnr[i];
1744 stream->psnr_count++;
1753 static void show_psnr(struct stream_state *stream, double peak) {
1757 if (!stream->psnr_count) return;
1759 fprintf(stderr, "Stream %d PSNR (Overall/Avg/Y/U/V)", stream->index);
1760 ovpsnr = sse_to_psnr((double)stream->psnr_samples_total, peak,
1761 (double)stream->psnr_sse_total);
1762 fprintf(stderr, " %.3f", ovpsnr);
1764 for (i = 0; i < 4; i++) {
1765 fprintf(stderr, " %.3f", stream->psnr_totals[i] / stream->psnr_count);
1767 fprintf(stderr, "\n");
1770 static float usec_to_fps(uint64_t usec, unsigned int frames) {
1771 return (float)(usec > 0 ? frames * 1000000.0 / (float)usec : 0);
1774 static void test_decode(struct stream_state *stream,
1775 enum TestDecodeFatality fatal,
1776 const VpxInterface *codec) {
1777 vpx_image_t enc_img, dec_img;
1779 if (stream->mismatch_seen) return;
1781 /* Get the internal reference frame */
1782 if (strcmp(codec->name, "vp8") == 0) {
1783 struct vpx_ref_frame ref_enc, ref_dec;
1786 width = (stream->config.cfg.g_w + 15) & ~15;
1787 height = (stream->config.cfg.g_h + 15) & ~15;
1788 vpx_img_alloc(&ref_enc.img, VPX_IMG_FMT_I420, width, height, 1);
1789 enc_img = ref_enc.img;
1790 vpx_img_alloc(&ref_dec.img, VPX_IMG_FMT_I420, width, height, 1);
1791 dec_img = ref_dec.img;
1793 ref_enc.frame_type = VP8_LAST_FRAME;
1794 ref_dec.frame_type = VP8_LAST_FRAME;
1795 vpx_codec_control(&stream->encoder, VP8_COPY_REFERENCE, &ref_enc);
1796 vpx_codec_control(&stream->decoder, VP8_COPY_REFERENCE, &ref_dec);
1798 struct vp9_ref_frame ref_enc, ref_dec;
1802 vpx_codec_control(&stream->encoder, VP9_GET_REFERENCE, &ref_enc);
1803 enc_img = ref_enc.img;
1804 vpx_codec_control(&stream->decoder, VP9_GET_REFERENCE, &ref_dec);
1805 dec_img = ref_dec.img;
1806 #if CONFIG_VP9_HIGHBITDEPTH
1807 if ((enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) !=
1808 (dec_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH)) {
1809 if (enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1810 vpx_img_alloc(&enc_img, enc_img.fmt - VPX_IMG_FMT_HIGHBITDEPTH,
1811 enc_img.d_w, enc_img.d_h, 16);
1812 vpx_img_truncate_16_to_8(&enc_img, &ref_enc.img);
1814 if (dec_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1815 vpx_img_alloc(&dec_img, dec_img.fmt - VPX_IMG_FMT_HIGHBITDEPTH,
1816 dec_img.d_w, dec_img.d_h, 16);
1817 vpx_img_truncate_16_to_8(&dec_img, &ref_dec.img);
1822 ctx_exit_on_error(&stream->encoder, "Failed to get encoder reference frame");
1823 ctx_exit_on_error(&stream->decoder, "Failed to get decoder reference frame");
1825 if (!compare_img(&enc_img, &dec_img)) {
1826 int y[4], u[4], v[4];
1827 #if CONFIG_VP9_HIGHBITDEPTH
1828 if (enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1829 find_mismatch_high(&enc_img, &dec_img, y, u, v);
1831 find_mismatch(&enc_img, &dec_img, y, u, v);
1834 find_mismatch(&enc_img, &dec_img, y, u, v);
1836 stream->decoder.err = 1;
1837 warn_or_exit_on_error(&stream->decoder, fatal == TEST_DECODE_FATAL,
1838 "Stream %d: Encode/decode mismatch on frame %d at"
1839 " Y[%d, %d] {%d/%d},"
1840 " U[%d, %d] {%d/%d},"
1841 " V[%d, %d] {%d/%d}",
1842 stream->index, stream->frames_out, y[0], y[1], y[2],
1843 y[3], u[0], u[1], u[2], u[3], v[0], v[1], v[2], v[3]);
1844 stream->mismatch_seen = stream->frames_out;
1847 vpx_img_free(&enc_img);
1848 vpx_img_free(&dec_img);
1851 static void print_time(const char *label, int64_t etl) {
1858 etl -= hours * 3600;
1863 fprintf(stderr, "[%3s %2" PRId64 ":%02" PRId64 ":%02" PRId64 "] ", label,
1866 fprintf(stderr, "[%3s unknown] ", label);
1870 int main(int argc, const char **argv_) {
1873 #if CONFIG_VP9_HIGHBITDEPTH
1874 vpx_image_t raw_shift;
1875 int allocated_raw_shift = 0;
1876 int use_16bit_internal = 0;
1877 int input_shift = 0;
1879 int frame_avail, got_data;
1881 struct VpxInputContext input;
1882 struct VpxEncoderConfig global;
1883 struct stream_state *streams = NULL;
1884 char **argv, **argi;
1885 uint64_t cx_time = 0;
1889 memset(&input, 0, sizeof(input));
1890 exec_name = argv_[0];
1892 if (argc < 3) usage_exit();
1894 /* Setup default input stream settings */
1895 input.framerate.numerator = 30;
1896 input.framerate.denominator = 1;
1897 input.only_i420 = 1;
1898 input.bit_depth = 0;
1900 /* First parse the global configuration values, because we want to apply
1901 * other parameters on top of the default configuration provided by the
1904 argv = argv_dup(argc - 1, argv_ + 1);
1905 parse_global_config(&global, argv);
1907 switch (global.color_type) {
1908 case I420: input.fmt = VPX_IMG_FMT_I420; break;
1909 case I422: input.fmt = VPX_IMG_FMT_I422; break;
1910 case I444: input.fmt = VPX_IMG_FMT_I444; break;
1911 case I440: input.fmt = VPX_IMG_FMT_I440; break;
1912 case YV12: input.fmt = VPX_IMG_FMT_YV12; break;
1916 /* Now parse each stream's parameters. Using a local scope here
1917 * due to the use of 'stream' as loop variable in FOREACH_STREAM
1920 struct stream_state *stream = NULL;
1923 stream = new_stream(&global, stream);
1925 if (!streams) streams = stream;
1926 } while (parse_stream_params(&global, stream, argv));
1929 /* Check for unrecognized options */
1930 for (argi = argv; *argi; argi++)
1931 if (argi[0][0] == '-' && argi[0][1])
1932 die("Error: Unrecognized option %s\n", *argi);
1934 FOREACH_STREAM(check_encoder_config(global.disable_warning_prompt, &global,
1935 &stream->config.cfg););
1937 /* Handle non-option arguments */
1938 input.filename = argv[0];
1940 if (!input.filename) usage_exit();
1942 /* Decide if other chroma subsamplings than 4:2:0 are supported */
1943 if (global.codec->fourcc == VP9_FOURCC) input.only_i420 = 0;
1945 for (pass = global.pass ? global.pass - 1 : 0; pass < global.passes; pass++) {
1946 int frames_in = 0, seen_frames = 0;
1947 int64_t estimated_time_left = -1;
1948 int64_t average_rate = -1;
1949 int64_t lagged_count = 0;
1951 open_input_file(&input);
1953 /* If the input file doesn't specify its w/h (raw files), try to get
1954 * the data from the first stream's configuration.
1956 if (!input.width || !input.height) {
1958 if (stream->config.cfg.g_w && stream->config.cfg.g_h) {
1959 input.width = stream->config.cfg.g_w;
1960 input.height = stream->config.cfg.g_h;
1966 /* Update stream configurations from the input file's parameters */
1967 if (!input.width || !input.height)
1969 "Specify stream dimensions with --width (-w) "
1970 " and --height (-h)");
1972 /* If input file does not specify bit-depth but input-bit-depth parameter
1973 * exists, assume that to be the input bit-depth. However, if the
1974 * input-bit-depth paramter does not exist, assume the input bit-depth
1975 * to be the same as the codec bit-depth.
1977 if (!input.bit_depth) {
1979 if (stream->config.cfg.g_input_bit_depth)
1980 input.bit_depth = stream->config.cfg.g_input_bit_depth;
1982 input.bit_depth = stream->config.cfg.g_input_bit_depth =
1983 (int)stream->config.cfg.g_bit_depth;
1985 if (input.bit_depth > 8) input.fmt |= VPX_IMG_FMT_HIGHBITDEPTH;
1988 { stream->config.cfg.g_input_bit_depth = input.bit_depth; });
1991 FOREACH_STREAM(set_stream_dimensions(stream, input.width, input.height));
1992 FOREACH_STREAM(validate_stream_config(stream, &global));
1994 /* Ensure that --passes and --pass are consistent. If --pass is set and
1995 * --passes=2, ensure --fpf was set.
1997 if (global.pass && global.passes == 2)
1999 if (!stream->config.stats_fn)
2000 die("Stream %d: Must specify --fpf when --pass=%d"
2001 " and --passes=2\n",
2002 stream->index, global.pass);
2007 if (stream->config.write_webm) {
2008 stream->config.write_webm = 0;
2010 "vpxenc was compiled without WebM container support."
2011 "Producing IVF output");
2016 /* Use the frame rate from the file only if none was specified
2017 * on the command-line.
2019 if (!global.have_framerate) {
2020 global.framerate.num = input.framerate.numerator;
2021 global.framerate.den = input.framerate.denominator;
2022 FOREACH_STREAM(stream->config.cfg.g_timebase.den = global.framerate.num;
2023 stream->config.cfg.g_timebase.num = global.framerate.den);
2026 /* Show configuration */
2027 if (global.verbose && pass == 0)
2028 FOREACH_STREAM(show_stream_config(stream, &global, &input));
2030 if (pass == (global.pass ? global.pass - 1 : 0)) {
2031 if (input.file_type == FILE_TYPE_Y4M)
2032 /*The Y4M reader does its own allocation.
2033 Just initialize this here to avoid problems if we never read any
2035 memset(&raw, 0, sizeof(raw));
2037 vpx_img_alloc(&raw, input.fmt, input.width, input.height, 32);
2039 FOREACH_STREAM(stream->rate_hist = init_rate_histogram(
2040 &stream->config.cfg, &global.framerate));
2043 FOREACH_STREAM(setup_pass(stream, &global, pass));
2045 open_output_file(stream, &global, &input.pixel_aspect_ratio));
2046 FOREACH_STREAM(initialize_encoder(stream, &global));
2048 #if CONFIG_VP9_HIGHBITDEPTH
2049 if (strcmp(global.codec->name, "vp9") == 0) {
2050 // Check to see if at least one stream uses 16 bit internal.
2051 // Currently assume that the bit_depths for all streams using
2052 // highbitdepth are the same.
2054 if (stream->config.use_16bit_internal) {
2055 use_16bit_internal = 1;
2057 if (stream->config.cfg.g_profile == 0) {
2060 input_shift = (int)stream->config.cfg.g_bit_depth -
2061 stream->config.cfg.g_input_bit_depth;
2070 while (frame_avail || got_data) {
2071 struct vpx_usec_timer timer;
2073 if (!global.limit || frames_in < global.limit) {
2074 frame_avail = read_frame(&input, &raw);
2076 if (frame_avail) frames_in++;
2078 frames_in > global.skip_frames ? frames_in - global.skip_frames : 0;
2080 if (!global.quiet) {
2081 float fps = usec_to_fps(cx_time, seen_frames);
2082 fprintf(stderr, "\rPass %d/%d ", pass + 1, global.passes);
2084 if (stream_cnt == 1)
2085 fprintf(stderr, "frame %4d/%-4d %7" PRId64 "B ", frames_in,
2086 streams->frames_out, (int64_t)streams->nbytes);
2088 fprintf(stderr, "frame %4d ", frames_in);
2090 fprintf(stderr, "%7" PRId64 " %s %.2f %s ",
2091 cx_time > 9999999 ? cx_time / 1000 : cx_time,
2092 cx_time > 9999999 ? "ms" : "us", fps >= 1.0 ? fps : fps * 60,
2093 fps >= 1.0 ? "fps" : "fpm");
2094 print_time("ETA", estimated_time_left);
2100 if (frames_in > global.skip_frames) {
2101 #if CONFIG_VP9_HIGHBITDEPTH
2102 vpx_image_t *frame_to_encode;
2103 if (input_shift || (use_16bit_internal && input.bit_depth == 8)) {
2104 assert(use_16bit_internal);
2105 // Input bit depth and stream bit depth do not match, so up
2106 // shift frame to stream bit depth
2107 if (!allocated_raw_shift) {
2108 vpx_img_alloc(&raw_shift, raw.fmt | VPX_IMG_FMT_HIGHBITDEPTH,
2109 input.width, input.height, 32);
2110 allocated_raw_shift = 1;
2112 vpx_img_upshift(&raw_shift, &raw, input_shift);
2113 frame_to_encode = &raw_shift;
2115 frame_to_encode = &raw;
2117 vpx_usec_timer_start(&timer);
2118 if (use_16bit_internal) {
2119 assert(frame_to_encode->fmt & VPX_IMG_FMT_HIGHBITDEPTH);
2121 if (stream->config.use_16bit_internal)
2122 encode_frame(stream, &global,
2123 frame_avail ? frame_to_encode : NULL, frames_in);
2128 assert((frame_to_encode->fmt & VPX_IMG_FMT_HIGHBITDEPTH) == 0);
2129 FOREACH_STREAM(encode_frame(stream, &global,
2130 frame_avail ? frame_to_encode : NULL,
2134 vpx_usec_timer_start(&timer);
2135 FOREACH_STREAM(encode_frame(stream, &global, frame_avail ? &raw : NULL,
2138 vpx_usec_timer_mark(&timer);
2139 cx_time += vpx_usec_timer_elapsed(&timer);
2141 FOREACH_STREAM(update_quantizer_histogram(stream));
2144 FOREACH_STREAM(get_cx_data(stream, &global, &got_data));
2146 if (!got_data && input.length && streams != NULL &&
2147 !streams->frames_out) {
2148 lagged_count = global.limit ? seen_frames : ftello(input.file);
2149 } else if (input.length) {
2154 const int64_t frame_in_lagged = (seen_frames - lagged_count) * 1000;
2156 rate = cx_time ? frame_in_lagged * (int64_t)1000000 / cx_time : 0;
2157 remaining = 1000 * (global.limit - global.skip_frames -
2158 seen_frames + lagged_count);
2160 const int64_t input_pos = ftello(input.file);
2161 const int64_t input_pos_lagged = input_pos - lagged_count;
2162 const int64_t limit = input.length;
2164 rate = cx_time ? input_pos_lagged * (int64_t)1000000 / cx_time : 0;
2165 remaining = limit - input_pos + lagged_count;
2169 (average_rate <= 0) ? rate : (average_rate * 7 + rate) / 8;
2170 estimated_time_left = average_rate ? remaining / average_rate : -1;
2173 if (got_data && global.test_decode != TEST_DECODE_OFF)
2174 FOREACH_STREAM(test_decode(stream, global.test_decode, global.codec));
2178 if (!global.quiet) fprintf(stderr, "\033[K");
2181 if (stream_cnt > 1) fprintf(stderr, "\n");
2183 if (!global.quiet) {
2184 FOREACH_STREAM(fprintf(
2185 stderr, "\rPass %d/%d frame %4d/%-4d %7" PRId64 "B %7" PRId64
2186 "b/f %7" PRId64 "b/s"
2187 " %7" PRId64 " %s (%.2f fps)\033[K\n",
2188 pass + 1, global.passes, frames_in, stream->frames_out,
2189 (int64_t)stream->nbytes,
2190 seen_frames ? (int64_t)(stream->nbytes * 8 / seen_frames) : 0,
2192 ? (int64_t)stream->nbytes * 8 * (int64_t)global.framerate.num /
2193 global.framerate.den / seen_frames
2195 stream->cx_time > 9999999 ? stream->cx_time / 1000 : stream->cx_time,
2196 stream->cx_time > 9999999 ? "ms" : "us",
2197 usec_to_fps(stream->cx_time, seen_frames)));
2200 if (global.show_psnr) {
2201 if (global.codec->fourcc == VP9_FOURCC) {
2203 show_psnr(stream, (1 << stream->config.cfg.g_input_bit_depth) - 1));
2205 FOREACH_STREAM(show_psnr(stream, 255.0));
2209 FOREACH_STREAM(vpx_codec_destroy(&stream->encoder));
2211 if (global.test_decode != TEST_DECODE_OFF) {
2212 FOREACH_STREAM(vpx_codec_destroy(&stream->decoder));
2215 close_input_file(&input);
2217 if (global.test_decode == TEST_DECODE_FATAL) {
2218 FOREACH_STREAM(res |= stream->mismatch_seen);
2220 FOREACH_STREAM(close_output_file(stream, global.codec->fourcc));
2222 FOREACH_STREAM(stats_close(&stream->stats, global.passes - 1));
2224 #if CONFIG_FP_MB_STATS
2225 FOREACH_STREAM(stats_close(&stream->fpmb_stats, global.passes - 1));
2228 if (global.pass) break;
2231 if (global.show_q_hist_buckets)
2233 show_q_histogram(stream->counts, global.show_q_hist_buckets));
2235 if (global.show_rate_hist_buckets)
2236 FOREACH_STREAM(show_rate_histogram(stream->rate_hist, &stream->config.cfg,
2237 global.show_rate_hist_buckets));
2238 FOREACH_STREAM(destroy_rate_histogram(stream->rate_hist));
2240 #if CONFIG_INTERNAL_STATS
2241 /* TODO(jkoleszar): This doesn't belong in this executable. Do it for now,
2242 * to match some existing utilities.
2244 if (!(global.pass == 1 && global.passes == 2))
2246 FILE *f = fopen("opsnr.stt", "a");
2247 if (stream->mismatch_seen) {
2248 fprintf(f, "First mismatch occurred in frame %d\n",
2249 stream->mismatch_seen);
2251 fprintf(f, "No mismatch detected in recon buffers\n");
2257 #if CONFIG_VP9_HIGHBITDEPTH
2258 if (allocated_raw_shift) vpx_img_free(&raw_shift);
2263 return res ? EXIT_FAILURE : EXIT_SUCCESS;