Refactor intra block prediction and reconstruction process
[aom.git] / vpxenc.c
blobcb9b8fecddcd88b65b823473aa1a41ae4950a407
1 /*
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.
9 */
11 #include "./vpxenc.h"
12 #include "./vpx_config.h"
14 #include <assert.h>
15 #include <limits.h>
16 #include <math.h>
17 #include <stdarg.h>
18 #include <stdio.h>
19 #include <stdlib.h>
20 #include <string.h>
22 #if CONFIG_LIBYUV
23 #include "third_party/libyuv/include/libyuv/scale.h"
24 #endif
26 #include "vpx/vpx_encoder.h"
27 #if CONFIG_DECODERS
28 #include "vpx/vpx_decoder.h"
29 #endif
31 #include "./args.h"
32 #include "./ivfenc.h"
33 #include "./tools_common.h"
35 #if CONFIG_VP8_ENCODER || CONFIG_VP9_ENCODER
36 #include "vpx/vp8cx.h"
37 #endif
38 #if CONFIG_VP8_DECODER || CONFIG_VP9_DECODER
39 #include "vpx/vp8dx.h"
40 #endif
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"
48 #if CONFIG_WEBM_IO
49 #include "./webmenc.h"
50 #endif
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,
55 FILE *stream) {
56 return fread(ptr, size, nmemb, stream);
58 #define fread wrap_fread
60 static size_t wrap_fwrite(const void *ptr, size_t size, size_t nmemb,
61 FILE *stream) {
62 return fwrite(ptr, size, nmemb, stream);
64 #define fwrite wrap_fwrite
67 static const char *exec_name;
69 static void warn_or_exit_on_errorv(vpx_codec_ctx_t *ctx, int fatal,
70 const char *s, va_list ap) {
71 if (ctx->err) {
72 const char *detail = vpx_codec_error_detail(ctx);
74 vfprintf(stderr, s, ap);
75 fprintf(stderr, ": %s\n", vpx_codec_error(ctx));
77 if (detail)
78 fprintf(stderr, " %s\n", detail);
80 if (fatal)
81 exit(EXIT_FAILURE);
85 static void ctx_exit_on_error(vpx_codec_ctx_t *ctx, const char *s, ...) {
86 va_list ap;
88 va_start(ap, s);
89 warn_or_exit_on_errorv(ctx, 1, s, ap);
90 va_end(ap);
93 static void warn_or_exit_on_error(vpx_codec_ctx_t *ctx, int fatal,
94 const char *s, ...) {
95 va_list ap;
97 va_start(ap, s);
98 warn_or_exit_on_errorv(ctx, fatal, s, ap);
99 va_end(ap);
102 static int read_frame(struct VpxInputContext *input_ctx, vpx_image_t *img) {
103 FILE *f = input_ctx->file;
104 y4m_input *y4m = &input_ctx->y4m;
105 int shortread = 0;
107 if (input_ctx->file_type == FILE_TYPE_Y4M) {
108 if (y4m_input_fetch_frame(y4m, f, img) < 1)
109 return 0;
110 } else {
111 shortread = read_yuv_frame(input_ctx, img);
114 return !shortread;
117 static int file_is_y4m(const char detect[4]) {
118 if (memcmp(detect, "YUV4", 4) == 0) {
119 return 1;
121 return 0;
124 static int fourcc_is_ivf(const char detect[4]) {
125 if (memcmp(detect, "DKIF", 4) == 0) {
126 return 1;
128 return 0;
131 static const arg_def_t debugmode = ARG_DEF(
132 "D", "debug", 0, "Debug mode (makes output deterministic)");
133 static const arg_def_t outputfile = ARG_DEF(
134 "o", "output", 1, "Output filename");
135 static const arg_def_t use_yv12 = ARG_DEF(
136 NULL, "yv12", 0, "Input file is YV12 ");
137 static const arg_def_t use_i420 = ARG_DEF(
138 NULL, "i420", 0, "Input file is I420 (default)");
139 static const arg_def_t use_i422 = ARG_DEF(
140 NULL, "i422", 0, "Input file is I422");
141 static const arg_def_t use_i444 = ARG_DEF(
142 NULL, "i444", 0, "Input file is I444");
143 static const arg_def_t use_i440 = ARG_DEF(
144 NULL, "i440", 0, "Input file is I440");
145 static const arg_def_t codecarg = ARG_DEF(
146 NULL, "codec", 1, "Codec to use");
147 static const arg_def_t passes = ARG_DEF(
148 "p", "passes", 1, "Number of passes (1/2)");
149 static const arg_def_t pass_arg = ARG_DEF(
150 NULL, "pass", 1, "Pass to execute (1/2)");
151 static const arg_def_t fpf_name = ARG_DEF(
152 NULL, "fpf", 1, "First pass statistics file name");
153 #if CONFIG_FP_MB_STATS
154 static const arg_def_t fpmbf_name = ARG_DEF(
155 NULL, "fpmbf", 1, "First pass block statistics file name");
156 #endif
157 static const arg_def_t limit = ARG_DEF(
158 NULL, "limit", 1, "Stop encoding after n input frames");
159 static const arg_def_t skip = ARG_DEF(
160 NULL, "skip", 1, "Skip the first n input frames");
161 static const arg_def_t deadline = ARG_DEF(
162 "d", "deadline", 1, "Deadline per frame (usec)");
163 static const arg_def_t best_dl = ARG_DEF(
164 NULL, "best", 0, "Use Best Quality Deadline");
165 static const arg_def_t good_dl = ARG_DEF(
166 NULL, "good", 0, "Use Good Quality Deadline");
167 static const arg_def_t rt_dl = ARG_DEF(
168 NULL, "rt", 0, "Use Realtime Quality Deadline");
169 static const arg_def_t quietarg = ARG_DEF(
170 "q", "quiet", 0, "Do not print encode progress");
171 static const arg_def_t verbosearg = ARG_DEF(
172 "v", "verbose", 0, "Show encoder parameters");
173 static const arg_def_t psnrarg = ARG_DEF(
174 NULL, "psnr", 0, "Show PSNR in status line");
176 static const struct arg_enum_list test_decode_enum[] = {
177 {"off", TEST_DECODE_OFF},
178 {"fatal", TEST_DECODE_FATAL},
179 {"warn", TEST_DECODE_WARN},
180 {NULL, 0}
182 static const arg_def_t recontest = ARG_DEF_ENUM(
183 NULL, "test-decode", 1, "Test encode/decode mismatch", test_decode_enum);
184 static const arg_def_t framerate = ARG_DEF(
185 NULL, "fps", 1, "Stream frame rate (rate/scale)");
186 static const arg_def_t use_webm = ARG_DEF(
187 NULL, "webm", 0, "Output WebM (default when WebM IO is enabled)");
188 static const arg_def_t use_ivf = ARG_DEF(
189 NULL, "ivf", 0, "Output IVF");
190 static const arg_def_t out_part = ARG_DEF(
191 "P", "output-partitions", 0,
192 "Makes encoder output partitions. Requires IVF output!");
193 static const arg_def_t q_hist_n = ARG_DEF(
194 NULL, "q-hist", 1, "Show quantizer histogram (n-buckets)");
195 static const arg_def_t rate_hist_n = ARG_DEF(
196 NULL, "rate-hist", 1, "Show rate histogram (n-buckets)");
197 static const arg_def_t disable_warnings = ARG_DEF(
198 NULL, "disable-warnings", 0,
199 "Disable warnings about potentially incorrect encode settings.");
200 static const arg_def_t disable_warning_prompt = ARG_DEF(
201 "y", "disable-warning-prompt", 0,
202 "Display warnings, but do not prompt user to continue.");
204 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
205 static const arg_def_t test16bitinternalarg = ARG_DEF(
206 NULL, "test-16bit-internal", 0, "Force use of 16 bit internal buffer");
207 #endif
209 static const arg_def_t *main_args[] = {
210 &debugmode,
211 &outputfile, &codecarg, &passes, &pass_arg, &fpf_name, &limit, &skip,
212 &deadline, &best_dl, &good_dl, &rt_dl,
213 &quietarg, &verbosearg, &psnrarg, &use_webm, &use_ivf, &out_part, &q_hist_n,
214 &rate_hist_n, &disable_warnings, &disable_warning_prompt,
215 NULL
218 static const arg_def_t usage = ARG_DEF(
219 "u", "usage", 1, "Usage profile number to use");
220 static const arg_def_t threads = ARG_DEF(
221 "t", "threads", 1, "Max number of threads to use");
222 static const arg_def_t profile = ARG_DEF(
223 NULL, "profile", 1, "Bitstream profile number to use");
224 static const arg_def_t width = ARG_DEF("w", "width", 1, "Frame width");
225 static const arg_def_t height = ARG_DEF("h", "height", 1, "Frame height");
226 #if CONFIG_WEBM_IO
227 static const struct arg_enum_list stereo_mode_enum[] = {
228 {"mono", STEREO_FORMAT_MONO},
229 {"left-right", STEREO_FORMAT_LEFT_RIGHT},
230 {"bottom-top", STEREO_FORMAT_BOTTOM_TOP},
231 {"top-bottom", STEREO_FORMAT_TOP_BOTTOM},
232 {"right-left", STEREO_FORMAT_RIGHT_LEFT},
233 {NULL, 0}
235 static const arg_def_t stereo_mode = ARG_DEF_ENUM(
236 NULL, "stereo-mode", 1, "Stereo 3D video format", stereo_mode_enum);
237 #endif
238 static const arg_def_t timebase = ARG_DEF(
239 NULL, "timebase", 1, "Output timestamp precision (fractional seconds)");
240 static const arg_def_t error_resilient = ARG_DEF(
241 NULL, "error-resilient", 1, "Enable error resiliency features");
242 static const arg_def_t lag_in_frames = ARG_DEF(
243 NULL, "lag-in-frames", 1, "Max number of frames to lag");
245 static const arg_def_t *global_args[] = {
246 &use_yv12, &use_i420, &use_i422, &use_i444, &use_i440,
247 &usage, &threads, &profile,
248 &width, &height,
249 #if CONFIG_WEBM_IO
250 &stereo_mode,
251 #endif
252 &timebase, &framerate,
253 &error_resilient,
254 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
255 &test16bitinternalarg,
256 #endif
257 &lag_in_frames, NULL
260 static const arg_def_t dropframe_thresh = ARG_DEF(
261 NULL, "drop-frame", 1, "Temporal resampling threshold (buf %)");
262 static const arg_def_t resize_allowed = ARG_DEF(
263 NULL, "resize-allowed", 1, "Spatial resampling enabled (bool)");
264 static const arg_def_t resize_width = ARG_DEF(
265 NULL, "resize-width", 1, "Width of encoded frame");
266 static const arg_def_t resize_height = ARG_DEF(
267 NULL, "resize-height", 1, "Height of encoded frame");
268 static const arg_def_t resize_up_thresh = ARG_DEF(
269 NULL, "resize-up", 1, "Upscale threshold (buf %)");
270 static const arg_def_t resize_down_thresh = ARG_DEF(
271 NULL, "resize-down", 1, "Downscale threshold (buf %)");
272 static const struct arg_enum_list end_usage_enum[] = {
273 {"vbr", VPX_VBR},
274 {"cbr", VPX_CBR},
275 {"cq", VPX_CQ},
276 {"q", VPX_Q},
277 {NULL, 0}
279 static const arg_def_t end_usage = ARG_DEF_ENUM(
280 NULL, "end-usage", 1, "Rate control mode", end_usage_enum);
281 static const arg_def_t target_bitrate = ARG_DEF(
282 NULL, "target-bitrate", 1, "Bitrate (kbps)");
283 static const arg_def_t min_quantizer = ARG_DEF(
284 NULL, "min-q", 1, "Minimum (best) quantizer");
285 static const arg_def_t max_quantizer = ARG_DEF(
286 NULL, "max-q", 1, "Maximum (worst) quantizer");
287 static const arg_def_t undershoot_pct = ARG_DEF(
288 NULL, "undershoot-pct", 1, "Datarate undershoot (min) target (%)");
289 static const arg_def_t overshoot_pct = ARG_DEF(
290 NULL, "overshoot-pct", 1, "Datarate overshoot (max) target (%)");
291 static const arg_def_t buf_sz = ARG_DEF(
292 NULL, "buf-sz", 1, "Client buffer size (ms)");
293 static const arg_def_t buf_initial_sz = ARG_DEF(
294 NULL, "buf-initial-sz", 1, "Client initial buffer size (ms)");
295 static const arg_def_t buf_optimal_sz = ARG_DEF(
296 NULL, "buf-optimal-sz", 1, "Client optimal buffer size (ms)");
297 static const arg_def_t *rc_args[] = {
298 &dropframe_thresh, &resize_allowed, &resize_width, &resize_height,
299 &resize_up_thresh, &resize_down_thresh, &end_usage, &target_bitrate,
300 &min_quantizer, &max_quantizer, &undershoot_pct, &overshoot_pct, &buf_sz,
301 &buf_initial_sz, &buf_optimal_sz, NULL
305 static const arg_def_t bias_pct = ARG_DEF(
306 NULL, "bias-pct", 1, "CBR/VBR bias (0=CBR, 100=VBR)");
307 static const arg_def_t minsection_pct = ARG_DEF(
308 NULL, "minsection-pct", 1, "GOP min bitrate (% of target)");
309 static const arg_def_t maxsection_pct = ARG_DEF(
310 NULL, "maxsection-pct", 1, "GOP max bitrate (% of target)");
311 static const arg_def_t *rc_twopass_args[] = {
312 &bias_pct, &minsection_pct, &maxsection_pct, NULL
316 static const arg_def_t kf_min_dist = ARG_DEF(
317 NULL, "kf-min-dist", 1, "Minimum keyframe interval (frames)");
318 static const arg_def_t kf_max_dist = ARG_DEF(
319 NULL, "kf-max-dist", 1, "Maximum keyframe interval (frames)");
320 static const arg_def_t kf_disabled = ARG_DEF(
321 NULL, "disable-kf", 0, "Disable keyframe placement");
322 static const arg_def_t *kf_args[] = {
323 &kf_min_dist, &kf_max_dist, &kf_disabled, NULL
327 static const arg_def_t noise_sens = ARG_DEF(
328 NULL, "noise-sensitivity", 1, "Noise sensitivity (frames to blur)");
329 static const arg_def_t sharpness = ARG_DEF(
330 NULL, "sharpness", 1, "Loop filter sharpness (0..7)");
331 static const arg_def_t static_thresh = ARG_DEF(
332 NULL, "static-thresh", 1, "Motion detection threshold");
333 static const arg_def_t auto_altref = ARG_DEF(
334 NULL, "auto-alt-ref", 1, "Enable automatic alt reference frames");
335 static const arg_def_t arnr_maxframes = ARG_DEF(
336 NULL, "arnr-maxframes", 1, "AltRef max frames (0..15)");
337 static const arg_def_t arnr_strength = ARG_DEF(
338 NULL, "arnr-strength", 1, "AltRef filter strength (0..6)");
339 static const arg_def_t arnr_type = ARG_DEF(
340 NULL, "arnr-type", 1, "AltRef type");
341 static const struct arg_enum_list tuning_enum[] = {
342 {"psnr", VP8_TUNE_PSNR},
343 {"ssim", VP8_TUNE_SSIM},
344 {NULL, 0}
346 static const arg_def_t tune_ssim = ARG_DEF_ENUM(
347 NULL, "tune", 1, "Material to favor", tuning_enum);
348 static const arg_def_t cq_level = ARG_DEF(
349 NULL, "cq-level", 1, "Constant/Constrained Quality level");
350 static const arg_def_t max_intra_rate_pct = ARG_DEF(
351 NULL, "max-intra-rate", 1, "Max I-frame bitrate (pct)");
353 #if CONFIG_VP8_ENCODER
354 static const arg_def_t cpu_used_vp8 = ARG_DEF(
355 NULL, "cpu-used", 1, "CPU Used (-16..16)");
356 static const arg_def_t token_parts = ARG_DEF(
357 NULL, "token-parts", 1, "Number of token partitions to use, log2");
358 static const arg_def_t screen_content_mode = ARG_DEF(
359 NULL, "screen-content-mode", 1, "Screen content mode");
360 static const arg_def_t *vp8_args[] = {
361 &cpu_used_vp8, &auto_altref, &noise_sens, &sharpness, &static_thresh,
362 &token_parts, &arnr_maxframes, &arnr_strength, &arnr_type,
363 &tune_ssim, &cq_level, &max_intra_rate_pct, &screen_content_mode,
364 NULL
366 static const int vp8_arg_ctrl_map[] = {
367 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
368 VP8E_SET_NOISE_SENSITIVITY, VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
369 VP8E_SET_TOKEN_PARTITIONS,
370 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
371 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
372 VP8E_SET_SCREEN_CONTENT_MODE,
375 #endif
377 #if CONFIG_VP9_ENCODER
378 static const arg_def_t cpu_used_vp9 = ARG_DEF(
379 NULL, "cpu-used", 1, "CPU Used (-8..8)");
380 static const arg_def_t tile_cols = ARG_DEF(
381 NULL, "tile-columns", 1, "Number of tile columns to use, log2");
382 static const arg_def_t tile_rows = ARG_DEF(
383 NULL, "tile-rows", 1, "Number of tile rows to use, log2");
384 static const arg_def_t lossless = ARG_DEF(
385 NULL, "lossless", 1, "Lossless mode");
386 static const arg_def_t frame_parallel_decoding = ARG_DEF(
387 NULL, "frame-parallel", 1, "Enable frame parallel decodability features");
388 static const arg_def_t aq_mode = ARG_DEF(
389 NULL, "aq-mode", 1,
390 "Adaptive quantization mode (0: off (default), 1: variance 2: complexity, "
391 "3: cyclic refresh)");
392 static const arg_def_t frame_periodic_boost = ARG_DEF(
393 NULL, "frame-boost", 1,
394 "Enable frame periodic boost (0: off (default), 1: on)");
395 static const arg_def_t gf_cbr_boost_pct = ARG_DEF(
396 NULL, "gf-cbr-boost", 1, "Boost for Golden Frame in CBR mode (pct)");
397 static const arg_def_t max_inter_rate_pct = ARG_DEF(
398 NULL, "max-inter-rate", 1, "Max P-frame bitrate (pct)");
399 static const arg_def_t min_gf_interval = ARG_DEF(
400 NULL, "min-gf-interval", 1,
401 "min gf/arf frame interval (default 0, indicating in-built behavior)");
402 static const arg_def_t max_gf_interval = ARG_DEF(
403 NULL, "max-gf-interval", 1,
404 "max gf/arf frame interval (default 0, indicating in-built behavior)");
406 static const struct arg_enum_list color_space_enum[] = {
407 { "unknown", VPX_CS_UNKNOWN },
408 { "bt601", VPX_CS_BT_601 },
409 { "bt709", VPX_CS_BT_709 },
410 { "smpte170", VPX_CS_SMPTE_170 },
411 { "smpte240", VPX_CS_SMPTE_240 },
412 { "bt2020", VPX_CS_BT_2020 },
413 { "reserved", VPX_CS_RESERVED },
414 { "sRGB", VPX_CS_SRGB },
415 { NULL, 0 }
418 static const arg_def_t input_color_space = ARG_DEF_ENUM(
419 NULL, "color-space", 1,
420 "The color space of input content:", color_space_enum);
422 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
423 static const struct arg_enum_list bitdepth_enum[] = {
424 {"8", VPX_BITS_8},
425 {"10", VPX_BITS_10},
426 {"12", VPX_BITS_12},
427 {NULL, 0}
430 static const arg_def_t bitdeptharg = ARG_DEF_ENUM(
431 "b", "bit-depth", 1,
432 "Bit depth for codec (8 for version <=1, 10 or 12 for version 2)",
433 bitdepth_enum);
434 static const arg_def_t inbitdeptharg = ARG_DEF(
435 NULL, "input-bit-depth", 1, "Bit depth of input");
436 #endif
438 static const struct arg_enum_list tune_content_enum[] = {
439 {"default", VP9E_CONTENT_DEFAULT},
440 {"screen", VP9E_CONTENT_SCREEN},
441 {NULL, 0}
444 static const arg_def_t tune_content = ARG_DEF_ENUM(
445 NULL, "tune-content", 1, "Tune content type", tune_content_enum);
447 static const arg_def_t *vp9_args[] = {
448 &cpu_used_vp9, &auto_altref, &sharpness, &static_thresh,
449 &tile_cols, &tile_rows, &arnr_maxframes, &arnr_strength, &arnr_type,
450 &tune_ssim, &cq_level, &max_intra_rate_pct, &max_inter_rate_pct,
451 &gf_cbr_boost_pct, &lossless,
452 &frame_parallel_decoding, &aq_mode, &frame_periodic_boost,
453 &noise_sens, &tune_content, &input_color_space,
454 &min_gf_interval, &max_gf_interval,
455 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
456 &bitdeptharg, &inbitdeptharg,
457 #endif
458 NULL
460 static const int vp9_arg_ctrl_map[] = {
461 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
462 VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
463 VP9E_SET_TILE_COLUMNS, VP9E_SET_TILE_ROWS,
464 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
465 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
466 VP9E_SET_MAX_INTER_BITRATE_PCT, VP9E_SET_GF_CBR_BOOST_PCT,
467 VP9E_SET_LOSSLESS, VP9E_SET_FRAME_PARALLEL_DECODING, VP9E_SET_AQ_MODE,
468 VP9E_SET_FRAME_PERIODIC_BOOST, VP9E_SET_NOISE_SENSITIVITY,
469 VP9E_SET_TUNE_CONTENT, VP9E_SET_COLOR_SPACE,
470 VP9E_SET_MIN_GF_INTERVAL, VP9E_SET_MAX_GF_INTERVAL,
473 #endif
475 static const arg_def_t *no_args[] = { NULL };
477 void usage_exit(void) {
478 int i;
479 const int num_encoder = get_vpx_encoder_count();
481 fprintf(stderr, "Usage: %s <options> -o dst_filename src_filename \n",
482 exec_name);
484 fprintf(stderr, "\nOptions:\n");
485 arg_show_usage(stderr, main_args);
486 fprintf(stderr, "\nEncoder Global Options:\n");
487 arg_show_usage(stderr, global_args);
488 fprintf(stderr, "\nRate Control Options:\n");
489 arg_show_usage(stderr, rc_args);
490 fprintf(stderr, "\nTwopass Rate Control Options:\n");
491 arg_show_usage(stderr, rc_twopass_args);
492 fprintf(stderr, "\nKeyframe Placement Options:\n");
493 arg_show_usage(stderr, kf_args);
494 #if CONFIG_VP8_ENCODER
495 fprintf(stderr, "\nVP8 Specific Options:\n");
496 arg_show_usage(stderr, vp8_args);
497 #endif
498 #if CONFIG_VP9_ENCODER
499 fprintf(stderr, "\nVP9 Specific Options:\n");
500 arg_show_usage(stderr, vp9_args);
501 #endif
502 fprintf(stderr, "\nStream timebase (--timebase):\n"
503 " The desired precision of timestamps in the output, expressed\n"
504 " in fractional seconds. Default is 1/1000.\n");
505 fprintf(stderr, "\nIncluded encoders:\n\n");
507 for (i = 0; i < num_encoder; ++i) {
508 const VpxInterface *const encoder = get_vpx_encoder_by_index(i);
509 const char* defstr = (i == (num_encoder - 1)) ? "(default)" : "";
510 fprintf(stderr, " %-6s - %s %s\n",
511 encoder->name, vpx_codec_iface_name(encoder->codec_interface()),
512 defstr);
514 fprintf(stderr, "\n ");
515 fprintf(stderr, "Use --codec to switch to a non-default encoder.\n\n");
517 exit(EXIT_FAILURE);
520 #define mmin(a, b) ((a) < (b) ? (a) : (b))
522 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
523 static void find_mismatch_high(const vpx_image_t *const img1,
524 const vpx_image_t *const img2,
525 int yloc[4], int uloc[4], int vloc[4]) {
526 uint16_t *plane1, *plane2;
527 uint32_t stride1, stride2;
528 const uint32_t bsize = 64;
529 const uint32_t bsizey = bsize >> img1->y_chroma_shift;
530 const uint32_t bsizex = bsize >> img1->x_chroma_shift;
531 const uint32_t c_w =
532 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
533 const uint32_t c_h =
534 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
535 int match = 1;
536 uint32_t i, j;
537 yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
538 plane1 = (uint16_t*)img1->planes[VPX_PLANE_Y];
539 plane2 = (uint16_t*)img2->planes[VPX_PLANE_Y];
540 stride1 = img1->stride[VPX_PLANE_Y]/2;
541 stride2 = img2->stride[VPX_PLANE_Y]/2;
542 for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
543 for (j = 0; match && j < img1->d_w; j += bsize) {
544 int k, l;
545 const int si = mmin(i + bsize, img1->d_h) - i;
546 const int sj = mmin(j + bsize, img1->d_w) - j;
547 for (k = 0; match && k < si; ++k) {
548 for (l = 0; match && l < sj; ++l) {
549 if (*(plane1 + (i + k) * stride1 + j + l) !=
550 *(plane2 + (i + k) * stride2 + j + l)) {
551 yloc[0] = i + k;
552 yloc[1] = j + l;
553 yloc[2] = *(plane1 + (i + k) * stride1 + j + l);
554 yloc[3] = *(plane2 + (i + k) * stride2 + j + l);
555 match = 0;
556 break;
563 uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
564 plane1 = (uint16_t*)img1->planes[VPX_PLANE_U];
565 plane2 = (uint16_t*)img2->planes[VPX_PLANE_U];
566 stride1 = img1->stride[VPX_PLANE_U]/2;
567 stride2 = img2->stride[VPX_PLANE_U]/2;
568 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
569 for (j = 0; match && j < c_w; j += bsizex) {
570 int k, l;
571 const int si = mmin(i + bsizey, c_h - i);
572 const int sj = mmin(j + bsizex, c_w - j);
573 for (k = 0; match && k < si; ++k) {
574 for (l = 0; match && l < sj; ++l) {
575 if (*(plane1 + (i + k) * stride1 + j + l) !=
576 *(plane2 + (i + k) * stride2 + j + l)) {
577 uloc[0] = i + k;
578 uloc[1] = j + l;
579 uloc[2] = *(plane1 + (i + k) * stride1 + j + l);
580 uloc[3] = *(plane2 + (i + k) * stride2 + j + l);
581 match = 0;
582 break;
589 vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
590 plane1 = (uint16_t*)img1->planes[VPX_PLANE_V];
591 plane2 = (uint16_t*)img2->planes[VPX_PLANE_V];
592 stride1 = img1->stride[VPX_PLANE_V]/2;
593 stride2 = img2->stride[VPX_PLANE_V]/2;
594 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
595 for (j = 0; match && j < c_w; j += bsizex) {
596 int k, l;
597 const int si = mmin(i + bsizey, c_h - i);
598 const int sj = mmin(j + bsizex, c_w - j);
599 for (k = 0; match && k < si; ++k) {
600 for (l = 0; match && l < sj; ++l) {
601 if (*(plane1 + (i + k) * stride1 + j + l) !=
602 *(plane2 + (i + k) * stride2 + j + l)) {
603 vloc[0] = i + k;
604 vloc[1] = j + l;
605 vloc[2] = *(plane1 + (i + k) * stride1 + j + l);
606 vloc[3] = *(plane2 + (i + k) * stride2 + j + l);
607 match = 0;
608 break;
615 #endif
617 static void find_mismatch(const vpx_image_t *const img1,
618 const vpx_image_t *const img2,
619 int yloc[4], int uloc[4], int vloc[4]) {
620 const uint32_t bsize = 64;
621 const uint32_t bsizey = bsize >> img1->y_chroma_shift;
622 const uint32_t bsizex = bsize >> img1->x_chroma_shift;
623 const uint32_t c_w =
624 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
625 const uint32_t c_h =
626 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
627 int match = 1;
628 uint32_t i, j;
629 yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
630 for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
631 for (j = 0; match && j < img1->d_w; j += bsize) {
632 int k, l;
633 const int si = mmin(i + bsize, img1->d_h) - i;
634 const int sj = mmin(j + bsize, img1->d_w) - j;
635 for (k = 0; match && k < si; ++k) {
636 for (l = 0; match && l < sj; ++l) {
637 if (*(img1->planes[VPX_PLANE_Y] +
638 (i + k) * img1->stride[VPX_PLANE_Y] + j + l) !=
639 *(img2->planes[VPX_PLANE_Y] +
640 (i + k) * img2->stride[VPX_PLANE_Y] + j + l)) {
641 yloc[0] = i + k;
642 yloc[1] = j + l;
643 yloc[2] = *(img1->planes[VPX_PLANE_Y] +
644 (i + k) * img1->stride[VPX_PLANE_Y] + j + l);
645 yloc[3] = *(img2->planes[VPX_PLANE_Y] +
646 (i + k) * img2->stride[VPX_PLANE_Y] + j + l);
647 match = 0;
648 break;
655 uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
656 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
657 for (j = 0; match && j < c_w; j += bsizex) {
658 int k, l;
659 const int si = mmin(i + bsizey, c_h - i);
660 const int sj = mmin(j + bsizex, c_w - j);
661 for (k = 0; match && k < si; ++k) {
662 for (l = 0; match && l < sj; ++l) {
663 if (*(img1->planes[VPX_PLANE_U] +
664 (i + k) * img1->stride[VPX_PLANE_U] + j + l) !=
665 *(img2->planes[VPX_PLANE_U] +
666 (i + k) * img2->stride[VPX_PLANE_U] + j + l)) {
667 uloc[0] = i + k;
668 uloc[1] = j + l;
669 uloc[2] = *(img1->planes[VPX_PLANE_U] +
670 (i + k) * img1->stride[VPX_PLANE_U] + j + l);
671 uloc[3] = *(img2->planes[VPX_PLANE_U] +
672 (i + k) * img2->stride[VPX_PLANE_U] + j + l);
673 match = 0;
674 break;
680 vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
681 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
682 for (j = 0; match && j < c_w; j += bsizex) {
683 int k, l;
684 const int si = mmin(i + bsizey, c_h - i);
685 const int sj = mmin(j + bsizex, c_w - j);
686 for (k = 0; match && k < si; ++k) {
687 for (l = 0; match && l < sj; ++l) {
688 if (*(img1->planes[VPX_PLANE_V] +
689 (i + k) * img1->stride[VPX_PLANE_V] + j + l) !=
690 *(img2->planes[VPX_PLANE_V] +
691 (i + k) * img2->stride[VPX_PLANE_V] + j + l)) {
692 vloc[0] = i + k;
693 vloc[1] = j + l;
694 vloc[2] = *(img1->planes[VPX_PLANE_V] +
695 (i + k) * img1->stride[VPX_PLANE_V] + j + l);
696 vloc[3] = *(img2->planes[VPX_PLANE_V] +
697 (i + k) * img2->stride[VPX_PLANE_V] + j + l);
698 match = 0;
699 break;
707 static int compare_img(const vpx_image_t *const img1,
708 const vpx_image_t *const img2) {
709 uint32_t l_w = img1->d_w;
710 uint32_t c_w =
711 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
712 const uint32_t c_h =
713 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
714 uint32_t i;
715 int match = 1;
717 match &= (img1->fmt == img2->fmt);
718 match &= (img1->d_w == img2->d_w);
719 match &= (img1->d_h == img2->d_h);
720 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
721 if (img1->fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
722 l_w *= 2;
723 c_w *= 2;
725 #endif
727 for (i = 0; i < img1->d_h; ++i)
728 match &= (memcmp(img1->planes[VPX_PLANE_Y] + i * img1->stride[VPX_PLANE_Y],
729 img2->planes[VPX_PLANE_Y] + i * img2->stride[VPX_PLANE_Y],
730 l_w) == 0);
732 for (i = 0; i < c_h; ++i)
733 match &= (memcmp(img1->planes[VPX_PLANE_U] + i * img1->stride[VPX_PLANE_U],
734 img2->planes[VPX_PLANE_U] + i * img2->stride[VPX_PLANE_U],
735 c_w) == 0);
737 for (i = 0; i < c_h; ++i)
738 match &= (memcmp(img1->planes[VPX_PLANE_V] + i * img1->stride[VPX_PLANE_V],
739 img2->planes[VPX_PLANE_V] + i * img2->stride[VPX_PLANE_V],
740 c_w) == 0);
742 return match;
746 #define NELEMENTS(x) (sizeof(x)/sizeof(x[0]))
747 #define MAX(x,y) ((x)>(y)?(x):(y))
748 #if CONFIG_VP8_ENCODER && !CONFIG_VP9_ENCODER
749 #define ARG_CTRL_CNT_MAX NELEMENTS(vp8_arg_ctrl_map)
750 #elif !CONFIG_VP8_ENCODER && CONFIG_VP9_ENCODER
751 #define ARG_CTRL_CNT_MAX NELEMENTS(vp9_arg_ctrl_map)
752 #else
753 #define ARG_CTRL_CNT_MAX MAX(NELEMENTS(vp8_arg_ctrl_map), \
754 NELEMENTS(vp9_arg_ctrl_map))
755 #endif
757 #if !CONFIG_WEBM_IO
758 typedef int stereo_format_t;
759 struct EbmlGlobal { int debug; };
760 #endif
762 /* Per-stream configuration */
763 struct stream_config {
764 struct vpx_codec_enc_cfg cfg;
765 const char *out_fn;
766 const char *stats_fn;
767 #if CONFIG_FP_MB_STATS
768 const char *fpmb_stats_fn;
769 #endif
770 stereo_format_t stereo_fmt;
771 int arg_ctrls[ARG_CTRL_CNT_MAX][2];
772 int arg_ctrl_cnt;
773 int write_webm;
774 int have_kf_max_dist;
775 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
776 // whether to use 16bit internal buffers
777 int use_16bit_internal;
778 #endif
782 struct stream_state {
783 int index;
784 struct stream_state *next;
785 struct stream_config config;
786 FILE *file;
787 struct rate_hist *rate_hist;
788 struct EbmlGlobal ebml;
789 uint64_t psnr_sse_total;
790 uint64_t psnr_samples_total;
791 double psnr_totals[4];
792 int psnr_count;
793 int counts[64];
794 vpx_codec_ctx_t encoder;
795 unsigned int frames_out;
796 uint64_t cx_time;
797 size_t nbytes;
798 stats_io_t stats;
799 #if CONFIG_FP_MB_STATS
800 stats_io_t fpmb_stats;
801 #endif
802 struct vpx_image *img;
803 vpx_codec_ctx_t decoder;
804 int mismatch_seen;
808 static void validate_positive_rational(const char *msg,
809 struct vpx_rational *rat) {
810 if (rat->den < 0) {
811 rat->num *= -1;
812 rat->den *= -1;
815 if (rat->num < 0)
816 die("Error: %s must be positive\n", msg);
818 if (!rat->den)
819 die("Error: %s has zero denominator\n", msg);
823 static void parse_global_config(struct VpxEncoderConfig *global, char **argv) {
824 char **argi, **argj;
825 struct arg arg;
826 const int num_encoder = get_vpx_encoder_count();
828 if (num_encoder < 1)
829 die("Error: no valid encoder available\n");
831 /* Initialize default parameters */
832 memset(global, 0, sizeof(*global));
833 global->codec = get_vpx_encoder_by_index(num_encoder - 1);
834 global->passes = 0;
835 global->color_type = I420;
836 /* Assign default deadline to good quality */
837 global->deadline = VPX_DL_GOOD_QUALITY;
839 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
840 arg.argv_step = 1;
842 if (arg_match(&arg, &codecarg, argi)) {
843 global->codec = get_vpx_encoder_by_name(arg.val);
844 if (!global->codec)
845 die("Error: Unrecognized argument (%s) to --codec\n", arg.val);
846 } else if (arg_match(&arg, &passes, argi)) {
847 global->passes = arg_parse_uint(&arg);
849 if (global->passes < 1 || global->passes > 2)
850 die("Error: Invalid number of passes (%d)\n", global->passes);
851 } else if (arg_match(&arg, &pass_arg, argi)) {
852 global->pass = arg_parse_uint(&arg);
854 if (global->pass < 1 || global->pass > 2)
855 die("Error: Invalid pass selected (%d)\n",
856 global->pass);
857 } else if (arg_match(&arg, &usage, argi))
858 global->usage = arg_parse_uint(&arg);
859 else if (arg_match(&arg, &deadline, argi))
860 global->deadline = arg_parse_uint(&arg);
861 else if (arg_match(&arg, &best_dl, argi))
862 global->deadline = VPX_DL_BEST_QUALITY;
863 else if (arg_match(&arg, &good_dl, argi))
864 global->deadline = VPX_DL_GOOD_QUALITY;
865 else if (arg_match(&arg, &rt_dl, argi))
866 global->deadline = VPX_DL_REALTIME;
867 else if (arg_match(&arg, &use_yv12, argi))
868 global->color_type = YV12;
869 else if (arg_match(&arg, &use_i420, argi))
870 global->color_type = I420;
871 else if (arg_match(&arg, &use_i422, argi))
872 global->color_type = I422;
873 else if (arg_match(&arg, &use_i444, argi))
874 global->color_type = I444;
875 else if (arg_match(&arg, &use_i440, argi))
876 global->color_type = I440;
877 else if (arg_match(&arg, &quietarg, argi))
878 global->quiet = 1;
879 else if (arg_match(&arg, &verbosearg, argi))
880 global->verbose = 1;
881 else if (arg_match(&arg, &limit, argi))
882 global->limit = arg_parse_uint(&arg);
883 else if (arg_match(&arg, &skip, argi))
884 global->skip_frames = arg_parse_uint(&arg);
885 else if (arg_match(&arg, &psnrarg, argi))
886 global->show_psnr = 1;
887 else if (arg_match(&arg, &recontest, argi))
888 global->test_decode = arg_parse_enum_or_int(&arg);
889 else if (arg_match(&arg, &framerate, argi)) {
890 global->framerate = arg_parse_rational(&arg);
891 validate_positive_rational(arg.name, &global->framerate);
892 global->have_framerate = 1;
893 } else if (arg_match(&arg, &out_part, argi))
894 global->out_part = 1;
895 else if (arg_match(&arg, &debugmode, argi))
896 global->debug = 1;
897 else if (arg_match(&arg, &q_hist_n, argi))
898 global->show_q_hist_buckets = arg_parse_uint(&arg);
899 else if (arg_match(&arg, &rate_hist_n, argi))
900 global->show_rate_hist_buckets = arg_parse_uint(&arg);
901 else if (arg_match(&arg, &disable_warnings, argi))
902 global->disable_warnings = 1;
903 else if (arg_match(&arg, &disable_warning_prompt, argi))
904 global->disable_warning_prompt = 1;
905 else
906 argj++;
909 if (global->pass) {
910 /* DWIM: Assume the user meant passes=2 if pass=2 is specified */
911 if (global->pass > global->passes) {
912 warn("Assuming --pass=%d implies --passes=%d\n",
913 global->pass, global->pass);
914 global->passes = global->pass;
917 /* Validate global config */
918 if (global->passes == 0) {
919 #if CONFIG_VP9_ENCODER
920 // Make default VP9 passes = 2 until there is a better quality 1-pass
921 // encoder
922 if (global->codec != NULL && global->codec->name != NULL)
923 global->passes = (strcmp(global->codec->name, "vp9") == 0 &&
924 global->deadline != VPX_DL_REALTIME) ? 2 : 1;
925 #else
926 global->passes = 1;
927 #endif
930 if (global->deadline == VPX_DL_REALTIME &&
931 global->passes > 1) {
932 warn("Enforcing one-pass encoding in realtime mode\n");
933 global->passes = 1;
938 static void open_input_file(struct VpxInputContext *input) {
939 /* Parse certain options from the input file, if possible */
940 input->file = strcmp(input->filename, "-")
941 ? fopen(input->filename, "rb") : set_binary_mode(stdin);
943 if (!input->file)
944 fatal("Failed to open input file");
946 if (!fseeko(input->file, 0, SEEK_END)) {
947 /* Input file is seekable. Figure out how long it is, so we can get
948 * progress info.
950 input->length = ftello(input->file);
951 rewind(input->file);
954 /* Default to 1:1 pixel aspect ratio. */
955 input->pixel_aspect_ratio.numerator = 1;
956 input->pixel_aspect_ratio.denominator = 1;
958 /* For RAW input sources, these bytes will applied on the first frame
959 * in read_frame().
961 input->detect.buf_read = fread(input->detect.buf, 1, 4, input->file);
962 input->detect.position = 0;
964 if (input->detect.buf_read == 4
965 && file_is_y4m(input->detect.buf)) {
966 if (y4m_input_open(&input->y4m, input->file, input->detect.buf, 4,
967 input->only_i420) >= 0) {
968 input->file_type = FILE_TYPE_Y4M;
969 input->width = input->y4m.pic_w;
970 input->height = input->y4m.pic_h;
971 input->pixel_aspect_ratio.numerator = input->y4m.par_n;
972 input->pixel_aspect_ratio.denominator = input->y4m.par_d;
973 input->framerate.numerator = input->y4m.fps_n;
974 input->framerate.denominator = input->y4m.fps_d;
975 input->fmt = input->y4m.vpx_fmt;
976 input->bit_depth = input->y4m.bit_depth;
977 } else
978 fatal("Unsupported Y4M stream.");
979 } else if (input->detect.buf_read == 4 && fourcc_is_ivf(input->detect.buf)) {
980 fatal("IVF is not supported as input.");
981 } else {
982 input->file_type = FILE_TYPE_RAW;
987 static void close_input_file(struct VpxInputContext *input) {
988 fclose(input->file);
989 if (input->file_type == FILE_TYPE_Y4M)
990 y4m_input_close(&input->y4m);
993 static struct stream_state *new_stream(struct VpxEncoderConfig *global,
994 struct stream_state *prev) {
995 struct stream_state *stream;
997 stream = calloc(1, sizeof(*stream));
998 if (stream == NULL) {
999 fatal("Failed to allocate new stream.");
1002 if (prev) {
1003 memcpy(stream, prev, sizeof(*stream));
1004 stream->index++;
1005 prev->next = stream;
1006 } else {
1007 vpx_codec_err_t res;
1009 /* Populate encoder configuration */
1010 res = vpx_codec_enc_config_default(global->codec->codec_interface(),
1011 &stream->config.cfg,
1012 global->usage);
1013 if (res)
1014 fatal("Failed to get config: %s\n", vpx_codec_err_to_string(res));
1016 /* Change the default timebase to a high enough value so that the
1017 * encoder will always create strictly increasing timestamps.
1019 stream->config.cfg.g_timebase.den = 1000;
1021 /* Never use the library's default resolution, require it be parsed
1022 * from the file or set on the command line.
1024 stream->config.cfg.g_w = 0;
1025 stream->config.cfg.g_h = 0;
1027 /* Initialize remaining stream parameters */
1028 stream->config.write_webm = 1;
1029 #if CONFIG_WEBM_IO
1030 stream->config.stereo_fmt = STEREO_FORMAT_MONO;
1031 stream->ebml.last_pts_ns = -1;
1032 stream->ebml.writer = NULL;
1033 stream->ebml.segment = NULL;
1034 #endif
1036 /* Allows removal of the application version from the EBML tags */
1037 stream->ebml.debug = global->debug;
1039 /* Default lag_in_frames is 0 in realtime mode */
1040 if (global->deadline == VPX_DL_REALTIME)
1041 stream->config.cfg.g_lag_in_frames = 0;
1044 /* Output files must be specified for each stream */
1045 stream->config.out_fn = NULL;
1047 stream->next = NULL;
1048 return stream;
1052 static int parse_stream_params(struct VpxEncoderConfig *global,
1053 struct stream_state *stream,
1054 char **argv) {
1055 char **argi, **argj;
1056 struct arg arg;
1057 static const arg_def_t **ctrl_args = no_args;
1058 static const int *ctrl_args_map = NULL;
1059 struct stream_config *config = &stream->config;
1060 int eos_mark_found = 0;
1061 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1062 int test_16bit_internal = 0;
1063 #endif
1065 // Handle codec specific options
1066 if (0) {
1067 #if CONFIG_VP8_ENCODER
1068 } else if (strcmp(global->codec->name, "vp8") == 0) {
1069 ctrl_args = vp8_args;
1070 ctrl_args_map = vp8_arg_ctrl_map;
1071 #endif
1072 #if CONFIG_VP9_ENCODER
1073 } else if (strcmp(global->codec->name, "vp9") == 0) {
1074 ctrl_args = vp9_args;
1075 ctrl_args_map = vp9_arg_ctrl_map;
1076 #endif
1079 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
1080 arg.argv_step = 1;
1082 /* Once we've found an end-of-stream marker (--) we want to continue
1083 * shifting arguments but not consuming them.
1085 if (eos_mark_found) {
1086 argj++;
1087 continue;
1088 } else if (!strcmp(*argj, "--")) {
1089 eos_mark_found = 1;
1090 continue;
1093 if (arg_match(&arg, &outputfile, argi)) {
1094 config->out_fn = arg.val;
1095 } else if (arg_match(&arg, &fpf_name, argi)) {
1096 config->stats_fn = arg.val;
1097 #if CONFIG_FP_MB_STATS
1098 } else if (arg_match(&arg, &fpmbf_name, argi)) {
1099 config->fpmb_stats_fn = arg.val;
1100 #endif
1101 } else if (arg_match(&arg, &use_webm, argi)) {
1102 #if CONFIG_WEBM_IO
1103 config->write_webm = 1;
1104 #else
1105 die("Error: --webm specified but webm is disabled.");
1106 #endif
1107 } else if (arg_match(&arg, &use_ivf, argi)) {
1108 config->write_webm = 0;
1109 } else if (arg_match(&arg, &threads, argi)) {
1110 config->cfg.g_threads = arg_parse_uint(&arg);
1111 } else if (arg_match(&arg, &profile, argi)) {
1112 config->cfg.g_profile = arg_parse_uint(&arg);
1113 } else if (arg_match(&arg, &width, argi)) {
1114 config->cfg.g_w = arg_parse_uint(&arg);
1115 } else if (arg_match(&arg, &height, argi)) {
1116 config->cfg.g_h = arg_parse_uint(&arg);
1117 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1118 } else if (arg_match(&arg, &bitdeptharg, argi)) {
1119 config->cfg.g_bit_depth = arg_parse_enum_or_int(&arg);
1120 } else if (arg_match(&arg, &inbitdeptharg, argi)) {
1121 config->cfg.g_input_bit_depth = arg_parse_uint(&arg);
1122 #endif
1123 #if CONFIG_WEBM_IO
1124 } else if (arg_match(&arg, &stereo_mode, argi)) {
1125 config->stereo_fmt = arg_parse_enum_or_int(&arg);
1126 #endif
1127 } else if (arg_match(&arg, &timebase, argi)) {
1128 config->cfg.g_timebase = arg_parse_rational(&arg);
1129 validate_positive_rational(arg.name, &config->cfg.g_timebase);
1130 } else if (arg_match(&arg, &error_resilient, argi)) {
1131 config->cfg.g_error_resilient = arg_parse_uint(&arg);
1132 } else if (arg_match(&arg, &lag_in_frames, argi)) {
1133 config->cfg.g_lag_in_frames = arg_parse_uint(&arg);
1134 if (global->deadline == VPX_DL_REALTIME &&
1135 config->cfg.g_lag_in_frames != 0) {
1136 warn("non-zero %s option ignored in realtime mode.\n", arg.name);
1137 config->cfg.g_lag_in_frames = 0;
1139 } else if (arg_match(&arg, &dropframe_thresh, argi)) {
1140 config->cfg.rc_dropframe_thresh = arg_parse_uint(&arg);
1141 } else if (arg_match(&arg, &resize_allowed, argi)) {
1142 config->cfg.rc_resize_allowed = arg_parse_uint(&arg);
1143 } else if (arg_match(&arg, &resize_width, argi)) {
1144 config->cfg.rc_scaled_width = arg_parse_uint(&arg);
1145 } else if (arg_match(&arg, &resize_height, argi)) {
1146 config->cfg.rc_scaled_height = arg_parse_uint(&arg);
1147 } else if (arg_match(&arg, &resize_up_thresh, argi)) {
1148 config->cfg.rc_resize_up_thresh = arg_parse_uint(&arg);
1149 } else if (arg_match(&arg, &resize_down_thresh, argi)) {
1150 config->cfg.rc_resize_down_thresh = arg_parse_uint(&arg);
1151 } else if (arg_match(&arg, &end_usage, argi)) {
1152 config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
1153 } else if (arg_match(&arg, &target_bitrate, argi)) {
1154 config->cfg.rc_target_bitrate = arg_parse_uint(&arg);
1155 } else if (arg_match(&arg, &min_quantizer, argi)) {
1156 config->cfg.rc_min_quantizer = arg_parse_uint(&arg);
1157 } else if (arg_match(&arg, &max_quantizer, argi)) {
1158 config->cfg.rc_max_quantizer = arg_parse_uint(&arg);
1159 } else if (arg_match(&arg, &undershoot_pct, argi)) {
1160 config->cfg.rc_undershoot_pct = arg_parse_uint(&arg);
1161 } else if (arg_match(&arg, &overshoot_pct, argi)) {
1162 config->cfg.rc_overshoot_pct = arg_parse_uint(&arg);
1163 } else if (arg_match(&arg, &buf_sz, argi)) {
1164 config->cfg.rc_buf_sz = arg_parse_uint(&arg);
1165 } else if (arg_match(&arg, &buf_initial_sz, argi)) {
1166 config->cfg.rc_buf_initial_sz = arg_parse_uint(&arg);
1167 } else if (arg_match(&arg, &buf_optimal_sz, argi)) {
1168 config->cfg.rc_buf_optimal_sz = arg_parse_uint(&arg);
1169 } else if (arg_match(&arg, &bias_pct, argi)) {
1170 config->cfg.rc_2pass_vbr_bias_pct = arg_parse_uint(&arg);
1171 if (global->passes < 2)
1172 warn("option %s ignored in one-pass mode.\n", arg.name);
1173 } else if (arg_match(&arg, &minsection_pct, argi)) {
1174 config->cfg.rc_2pass_vbr_minsection_pct = arg_parse_uint(&arg);
1176 if (global->passes < 2)
1177 warn("option %s ignored in one-pass mode.\n", arg.name);
1178 } else if (arg_match(&arg, &maxsection_pct, argi)) {
1179 config->cfg.rc_2pass_vbr_maxsection_pct = arg_parse_uint(&arg);
1181 if (global->passes < 2)
1182 warn("option %s ignored in one-pass mode.\n", arg.name);
1183 } else if (arg_match(&arg, &kf_min_dist, argi)) {
1184 config->cfg.kf_min_dist = arg_parse_uint(&arg);
1185 } else if (arg_match(&arg, &kf_max_dist, argi)) {
1186 config->cfg.kf_max_dist = arg_parse_uint(&arg);
1187 config->have_kf_max_dist = 1;
1188 } else if (arg_match(&arg, &kf_disabled, argi)) {
1189 config->cfg.kf_mode = VPX_KF_DISABLED;
1190 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1191 } else if (arg_match(&arg, &test16bitinternalarg, argi)) {
1192 if (strcmp(global->codec->name, "vp9") == 0) {
1193 test_16bit_internal = 1;
1195 #endif
1196 } else {
1197 int i, match = 0;
1198 for (i = 0; ctrl_args[i]; i++) {
1199 if (arg_match(&arg, ctrl_args[i], argi)) {
1200 int j;
1201 match = 1;
1203 /* Point either to the next free element or the first
1204 * instance of this control.
1206 for (j = 0; j < config->arg_ctrl_cnt; j++)
1207 if (ctrl_args_map != NULL &&
1208 config->arg_ctrls[j][0] == ctrl_args_map[i])
1209 break;
1211 /* Update/insert */
1212 assert(j < (int)ARG_CTRL_CNT_MAX);
1213 if (ctrl_args_map != NULL && j < (int)ARG_CTRL_CNT_MAX) {
1214 config->arg_ctrls[j][0] = ctrl_args_map[i];
1215 config->arg_ctrls[j][1] = arg_parse_enum_or_int(&arg);
1216 if (j == config->arg_ctrl_cnt)
1217 config->arg_ctrl_cnt++;
1221 if (!match)
1222 argj++;
1225 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1226 if (strcmp(global->codec->name, "vp9") == 0) {
1227 config->use_16bit_internal = test_16bit_internal |
1228 (config->cfg.g_profile > 1);
1230 #endif
1231 return eos_mark_found;
1235 #define FOREACH_STREAM(func) \
1236 do { \
1237 struct stream_state *stream; \
1238 for (stream = streams; stream; stream = stream->next) { \
1239 func; \
1241 } while (0)
1244 static void validate_stream_config(const struct stream_state *stream,
1245 const struct VpxEncoderConfig *global) {
1246 const struct stream_state *streami;
1247 (void)global;
1249 if (!stream->config.cfg.g_w || !stream->config.cfg.g_h)
1250 fatal("Stream %d: Specify stream dimensions with --width (-w) "
1251 " and --height (-h)", stream->index);
1253 // Check that the codec bit depth is greater than the input bit depth.
1254 if (stream->config.cfg.g_input_bit_depth >
1255 (unsigned int)stream->config.cfg.g_bit_depth) {
1256 fatal("Stream %d: codec bit depth (%d) less than input bit depth (%d)",
1257 stream->index, (int)stream->config.cfg.g_bit_depth,
1258 stream->config.cfg.g_input_bit_depth);
1261 for (streami = stream; streami; streami = streami->next) {
1262 /* All streams require output files */
1263 if (!streami->config.out_fn)
1264 fatal("Stream %d: Output file is required (specify with -o)",
1265 streami->index);
1267 /* Check for two streams outputting to the same file */
1268 if (streami != stream) {
1269 const char *a = stream->config.out_fn;
1270 const char *b = streami->config.out_fn;
1271 if (!strcmp(a, b) && strcmp(a, "/dev/null") && strcmp(a, ":nul"))
1272 fatal("Stream %d: duplicate output file (from stream %d)",
1273 streami->index, stream->index);
1276 /* Check for two streams sharing a stats file. */
1277 if (streami != stream) {
1278 const char *a = stream->config.stats_fn;
1279 const char *b = streami->config.stats_fn;
1280 if (a && b && !strcmp(a, b))
1281 fatal("Stream %d: duplicate stats file (from stream %d)",
1282 streami->index, stream->index);
1285 #if CONFIG_FP_MB_STATS
1286 /* Check for two streams sharing a mb stats file. */
1287 if (streami != stream) {
1288 const char *a = stream->config.fpmb_stats_fn;
1289 const char *b = streami->config.fpmb_stats_fn;
1290 if (a && b && !strcmp(a, b))
1291 fatal("Stream %d: duplicate mb stats file (from stream %d)",
1292 streami->index, stream->index);
1294 #endif
1299 static void set_stream_dimensions(struct stream_state *stream,
1300 unsigned int w,
1301 unsigned int h) {
1302 if (!stream->config.cfg.g_w) {
1303 if (!stream->config.cfg.g_h)
1304 stream->config.cfg.g_w = w;
1305 else
1306 stream->config.cfg.g_w = w * stream->config.cfg.g_h / h;
1308 if (!stream->config.cfg.g_h) {
1309 stream->config.cfg.g_h = h * stream->config.cfg.g_w / w;
1314 static void set_default_kf_interval(struct stream_state *stream,
1315 struct VpxEncoderConfig *global) {
1316 /* Use a max keyframe interval of 5 seconds, if none was
1317 * specified on the command line.
1319 if (!stream->config.have_kf_max_dist) {
1320 double framerate = (double)global->framerate.num / global->framerate.den;
1321 if (framerate > 0.0)
1322 stream->config.cfg.kf_max_dist = (unsigned int)(5.0 * framerate);
1326 static const char* file_type_to_string(enum VideoFileType t) {
1327 switch (t) {
1328 case FILE_TYPE_RAW: return "RAW";
1329 case FILE_TYPE_Y4M: return "Y4M";
1330 default: return "Other";
1334 static const char* image_format_to_string(vpx_img_fmt_t f) {
1335 switch (f) {
1336 case VPX_IMG_FMT_I420: return "I420";
1337 case VPX_IMG_FMT_I422: return "I422";
1338 case VPX_IMG_FMT_I444: return "I444";
1339 case VPX_IMG_FMT_I440: return "I440";
1340 case VPX_IMG_FMT_YV12: return "YV12";
1341 case VPX_IMG_FMT_I42016: return "I42016";
1342 case VPX_IMG_FMT_I42216: return "I42216";
1343 case VPX_IMG_FMT_I44416: return "I44416";
1344 case VPX_IMG_FMT_I44016: return "I44016";
1345 default: return "Other";
1349 static void show_stream_config(struct stream_state *stream,
1350 struct VpxEncoderConfig *global,
1351 struct VpxInputContext *input) {
1353 #define SHOW(field) \
1354 fprintf(stderr, " %-28s = %d\n", #field, stream->config.cfg.field)
1356 if (stream->index == 0) {
1357 fprintf(stderr, "Codec: %s\n",
1358 vpx_codec_iface_name(global->codec->codec_interface()));
1359 fprintf(stderr, "Source file: %s File Type: %s Format: %s\n",
1360 input->filename,
1361 file_type_to_string(input->file_type),
1362 image_format_to_string(input->fmt));
1364 if (stream->next || stream->index)
1365 fprintf(stderr, "\nStream Index: %d\n", stream->index);
1366 fprintf(stderr, "Destination file: %s\n", stream->config.out_fn);
1367 fprintf(stderr, "Encoder parameters:\n");
1369 SHOW(g_usage);
1370 SHOW(g_threads);
1371 SHOW(g_profile);
1372 SHOW(g_w);
1373 SHOW(g_h);
1374 SHOW(g_bit_depth);
1375 SHOW(g_input_bit_depth);
1376 SHOW(g_timebase.num);
1377 SHOW(g_timebase.den);
1378 SHOW(g_error_resilient);
1379 SHOW(g_pass);
1380 SHOW(g_lag_in_frames);
1381 SHOW(rc_dropframe_thresh);
1382 SHOW(rc_resize_allowed);
1383 SHOW(rc_scaled_width);
1384 SHOW(rc_scaled_height);
1385 SHOW(rc_resize_up_thresh);
1386 SHOW(rc_resize_down_thresh);
1387 SHOW(rc_end_usage);
1388 SHOW(rc_target_bitrate);
1389 SHOW(rc_min_quantizer);
1390 SHOW(rc_max_quantizer);
1391 SHOW(rc_undershoot_pct);
1392 SHOW(rc_overshoot_pct);
1393 SHOW(rc_buf_sz);
1394 SHOW(rc_buf_initial_sz);
1395 SHOW(rc_buf_optimal_sz);
1396 SHOW(rc_2pass_vbr_bias_pct);
1397 SHOW(rc_2pass_vbr_minsection_pct);
1398 SHOW(rc_2pass_vbr_maxsection_pct);
1399 SHOW(kf_mode);
1400 SHOW(kf_min_dist);
1401 SHOW(kf_max_dist);
1405 static void open_output_file(struct stream_state *stream,
1406 struct VpxEncoderConfig *global,
1407 const struct VpxRational *pixel_aspect_ratio) {
1408 const char *fn = stream->config.out_fn;
1409 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1411 if (cfg->g_pass == VPX_RC_FIRST_PASS)
1412 return;
1414 stream->file = strcmp(fn, "-") ? fopen(fn, "wb") : set_binary_mode(stdout);
1416 if (!stream->file)
1417 fatal("Failed to open output file");
1419 if (stream->config.write_webm && fseek(stream->file, 0, SEEK_CUR))
1420 fatal("WebM output to pipes not supported.");
1422 #if CONFIG_WEBM_IO
1423 if (stream->config.write_webm) {
1424 stream->ebml.stream = stream->file;
1425 write_webm_file_header(&stream->ebml, cfg,
1426 &global->framerate,
1427 stream->config.stereo_fmt,
1428 global->codec->fourcc,
1429 pixel_aspect_ratio);
1431 #endif
1433 if (!stream->config.write_webm) {
1434 ivf_write_file_header(stream->file, cfg, global->codec->fourcc, 0);
1439 static void close_output_file(struct stream_state *stream,
1440 unsigned int fourcc) {
1441 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1443 if (cfg->g_pass == VPX_RC_FIRST_PASS)
1444 return;
1446 #if CONFIG_WEBM_IO
1447 if (stream->config.write_webm) {
1448 write_webm_file_footer(&stream->ebml);
1450 #endif
1452 if (!stream->config.write_webm) {
1453 if (!fseek(stream->file, 0, SEEK_SET))
1454 ivf_write_file_header(stream->file, &stream->config.cfg,
1455 fourcc,
1456 stream->frames_out);
1459 fclose(stream->file);
1463 static void setup_pass(struct stream_state *stream,
1464 struct VpxEncoderConfig *global,
1465 int pass) {
1466 if (stream->config.stats_fn) {
1467 if (!stats_open_file(&stream->stats, stream->config.stats_fn,
1468 pass))
1469 fatal("Failed to open statistics store");
1470 } else {
1471 if (!stats_open_mem(&stream->stats, pass))
1472 fatal("Failed to open statistics store");
1475 #if CONFIG_FP_MB_STATS
1476 if (stream->config.fpmb_stats_fn) {
1477 if (!stats_open_file(&stream->fpmb_stats,
1478 stream->config.fpmb_stats_fn, pass))
1479 fatal("Failed to open mb statistics store");
1480 } else {
1481 if (!stats_open_mem(&stream->fpmb_stats, pass))
1482 fatal("Failed to open mb statistics store");
1484 #endif
1486 stream->config.cfg.g_pass = global->passes == 2
1487 ? pass ? VPX_RC_LAST_PASS : VPX_RC_FIRST_PASS
1488 : VPX_RC_ONE_PASS;
1489 if (pass) {
1490 stream->config.cfg.rc_twopass_stats_in = stats_get(&stream->stats);
1491 #if CONFIG_FP_MB_STATS
1492 stream->config.cfg.rc_firstpass_mb_stats_in =
1493 stats_get(&stream->fpmb_stats);
1494 #endif
1497 stream->cx_time = 0;
1498 stream->nbytes = 0;
1499 stream->frames_out = 0;
1503 static void initialize_encoder(struct stream_state *stream,
1504 struct VpxEncoderConfig *global) {
1505 int i;
1506 int flags = 0;
1508 flags |= global->show_psnr ? VPX_CODEC_USE_PSNR : 0;
1509 flags |= global->out_part ? VPX_CODEC_USE_OUTPUT_PARTITION : 0;
1510 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1511 flags |= stream->config.use_16bit_internal ? VPX_CODEC_USE_HIGHBITDEPTH : 0;
1512 #endif
1514 /* Construct Encoder Context */
1515 vpx_codec_enc_init(&stream->encoder, global->codec->codec_interface(),
1516 &stream->config.cfg, flags);
1517 ctx_exit_on_error(&stream->encoder, "Failed to initialize encoder");
1519 /* Note that we bypass the vpx_codec_control wrapper macro because
1520 * we're being clever to store the control IDs in an array. Real
1521 * applications will want to make use of the enumerations directly
1523 for (i = 0; i < stream->config.arg_ctrl_cnt; i++) {
1524 int ctrl = stream->config.arg_ctrls[i][0];
1525 int value = stream->config.arg_ctrls[i][1];
1526 if (vpx_codec_control_(&stream->encoder, ctrl, value))
1527 fprintf(stderr, "Error: Tried to set control %d = %d\n",
1528 ctrl, value);
1530 ctx_exit_on_error(&stream->encoder, "Failed to control codec");
1533 #if CONFIG_DECODERS
1534 if (global->test_decode != TEST_DECODE_OFF) {
1535 const VpxInterface *decoder = get_vpx_decoder_by_name(global->codec->name);
1536 vpx_codec_dec_init(&stream->decoder, decoder->codec_interface(), NULL, 0);
1538 #endif
1542 static void encode_frame(struct stream_state *stream,
1543 struct VpxEncoderConfig *global,
1544 struct vpx_image *img,
1545 unsigned int frames_in) {
1546 vpx_codec_pts_t frame_start, next_frame_start;
1547 struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1548 struct vpx_usec_timer timer;
1550 frame_start = (cfg->g_timebase.den * (int64_t)(frames_in - 1)
1551 * global->framerate.den)
1552 / cfg->g_timebase.num / global->framerate.num;
1553 next_frame_start = (cfg->g_timebase.den * (int64_t)(frames_in)
1554 * global->framerate.den)
1555 / cfg->g_timebase.num / global->framerate.num;
1557 /* Scale if necessary */
1558 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1559 if (img) {
1560 if ((img->fmt & VPX_IMG_FMT_HIGHBITDEPTH) &&
1561 (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1562 if (img->fmt != VPX_IMG_FMT_I42016) {
1563 fprintf(stderr, "%s can only scale 4:2:0 inputs\n", exec_name);
1564 exit(EXIT_FAILURE);
1566 #if CONFIG_LIBYUV
1567 if (!stream->img) {
1568 stream->img = vpx_img_alloc(NULL, VPX_IMG_FMT_I42016,
1569 cfg->g_w, cfg->g_h, 16);
1571 I420Scale_16((uint16*)img->planes[VPX_PLANE_Y],
1572 img->stride[VPX_PLANE_Y]/2,
1573 (uint16*)img->planes[VPX_PLANE_U],
1574 img->stride[VPX_PLANE_U]/2,
1575 (uint16*)img->planes[VPX_PLANE_V],
1576 img->stride[VPX_PLANE_V]/2,
1577 img->d_w, img->d_h,
1578 (uint16*)stream->img->planes[VPX_PLANE_Y],
1579 stream->img->stride[VPX_PLANE_Y]/2,
1580 (uint16*)stream->img->planes[VPX_PLANE_U],
1581 stream->img->stride[VPX_PLANE_U]/2,
1582 (uint16*)stream->img->planes[VPX_PLANE_V],
1583 stream->img->stride[VPX_PLANE_V]/2,
1584 stream->img->d_w, stream->img->d_h,
1585 kFilterBox);
1586 img = stream->img;
1587 #else
1588 stream->encoder.err = 1;
1589 ctx_exit_on_error(&stream->encoder,
1590 "Stream %d: Failed to encode frame.\n"
1591 "Scaling disabled in this configuration. \n"
1592 "To enable, configure with --enable-libyuv\n",
1593 stream->index);
1594 #endif
1597 #endif
1598 if (img && (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1599 if (img->fmt != VPX_IMG_FMT_I420 && img->fmt != VPX_IMG_FMT_YV12) {
1600 fprintf(stderr, "%s can only scale 4:2:0 8bpp inputs\n", exec_name);
1601 exit(EXIT_FAILURE);
1603 #if CONFIG_LIBYUV
1604 if (!stream->img)
1605 stream->img = vpx_img_alloc(NULL, VPX_IMG_FMT_I420,
1606 cfg->g_w, cfg->g_h, 16);
1607 I420Scale(img->planes[VPX_PLANE_Y], img->stride[VPX_PLANE_Y],
1608 img->planes[VPX_PLANE_U], img->stride[VPX_PLANE_U],
1609 img->planes[VPX_PLANE_V], img->stride[VPX_PLANE_V],
1610 img->d_w, img->d_h,
1611 stream->img->planes[VPX_PLANE_Y],
1612 stream->img->stride[VPX_PLANE_Y],
1613 stream->img->planes[VPX_PLANE_U],
1614 stream->img->stride[VPX_PLANE_U],
1615 stream->img->planes[VPX_PLANE_V],
1616 stream->img->stride[VPX_PLANE_V],
1617 stream->img->d_w, stream->img->d_h,
1618 kFilterBox);
1619 img = stream->img;
1620 #else
1621 stream->encoder.err = 1;
1622 ctx_exit_on_error(&stream->encoder,
1623 "Stream %d: Failed to encode frame.\n"
1624 "Scaling disabled in this configuration. \n"
1625 "To enable, configure with --enable-libyuv\n",
1626 stream->index);
1627 #endif
1630 vpx_usec_timer_start(&timer);
1631 vpx_codec_encode(&stream->encoder, img, frame_start,
1632 (unsigned long)(next_frame_start - frame_start),
1633 0, global->deadline);
1634 vpx_usec_timer_mark(&timer);
1635 stream->cx_time += vpx_usec_timer_elapsed(&timer);
1636 ctx_exit_on_error(&stream->encoder, "Stream %d: Failed to encode frame",
1637 stream->index);
1641 static void update_quantizer_histogram(struct stream_state *stream) {
1642 if (stream->config.cfg.g_pass != VPX_RC_FIRST_PASS) {
1643 int q;
1645 vpx_codec_control(&stream->encoder, VP8E_GET_LAST_QUANTIZER_64, &q);
1646 ctx_exit_on_error(&stream->encoder, "Failed to read quantizer");
1647 stream->counts[q]++;
1652 static void get_cx_data(struct stream_state *stream,
1653 struct VpxEncoderConfig *global,
1654 int *got_data) {
1655 const vpx_codec_cx_pkt_t *pkt;
1656 const struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1657 vpx_codec_iter_t iter = NULL;
1659 *got_data = 0;
1660 while ((pkt = vpx_codec_get_cx_data(&stream->encoder, &iter))) {
1661 static size_t fsize = 0;
1662 static int64_t ivf_header_pos = 0;
1664 switch (pkt->kind) {
1665 case VPX_CODEC_CX_FRAME_PKT:
1666 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1667 stream->frames_out++;
1669 if (!global->quiet)
1670 fprintf(stderr, " %6luF", (unsigned long)pkt->data.frame.sz);
1672 update_rate_histogram(stream->rate_hist, cfg, pkt);
1673 #if CONFIG_WEBM_IO
1674 if (stream->config.write_webm) {
1675 write_webm_block(&stream->ebml, cfg, pkt);
1677 #endif
1678 if (!stream->config.write_webm) {
1679 if (pkt->data.frame.partition_id <= 0) {
1680 ivf_header_pos = ftello(stream->file);
1681 fsize = pkt->data.frame.sz;
1683 ivf_write_frame_header(stream->file, pkt->data.frame.pts, fsize);
1684 } else {
1685 fsize += pkt->data.frame.sz;
1687 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1688 const int64_t currpos = ftello(stream->file);
1689 fseeko(stream->file, ivf_header_pos, SEEK_SET);
1690 ivf_write_frame_size(stream->file, fsize);
1691 fseeko(stream->file, currpos, SEEK_SET);
1695 (void) fwrite(pkt->data.frame.buf, 1, pkt->data.frame.sz,
1696 stream->file);
1698 stream->nbytes += pkt->data.raw.sz;
1700 *got_data = 1;
1701 #if CONFIG_DECODERS
1702 if (global->test_decode != TEST_DECODE_OFF && !stream->mismatch_seen) {
1703 vpx_codec_decode(&stream->decoder, pkt->data.frame.buf,
1704 (unsigned int)pkt->data.frame.sz, NULL, 0);
1705 if (stream->decoder.err) {
1706 warn_or_exit_on_error(&stream->decoder,
1707 global->test_decode == TEST_DECODE_FATAL,
1708 "Failed to decode frame %d in stream %d",
1709 stream->frames_out + 1, stream->index);
1710 stream->mismatch_seen = stream->frames_out + 1;
1713 #endif
1714 break;
1715 case VPX_CODEC_STATS_PKT:
1716 stream->frames_out++;
1717 stats_write(&stream->stats,
1718 pkt->data.twopass_stats.buf,
1719 pkt->data.twopass_stats.sz);
1720 stream->nbytes += pkt->data.raw.sz;
1721 break;
1722 #if CONFIG_FP_MB_STATS
1723 case VPX_CODEC_FPMB_STATS_PKT:
1724 stats_write(&stream->fpmb_stats,
1725 pkt->data.firstpass_mb_stats.buf,
1726 pkt->data.firstpass_mb_stats.sz);
1727 stream->nbytes += pkt->data.raw.sz;
1728 break;
1729 #endif
1730 case VPX_CODEC_PSNR_PKT:
1732 if (global->show_psnr) {
1733 int i;
1735 stream->psnr_sse_total += pkt->data.psnr.sse[0];
1736 stream->psnr_samples_total += pkt->data.psnr.samples[0];
1737 for (i = 0; i < 4; i++) {
1738 if (!global->quiet)
1739 fprintf(stderr, "%.3f ", pkt->data.psnr.psnr[i]);
1740 stream->psnr_totals[i] += pkt->data.psnr.psnr[i];
1742 stream->psnr_count++;
1745 break;
1746 default:
1747 break;
1753 static void show_psnr(struct stream_state *stream, double peak) {
1754 int i;
1755 double ovpsnr;
1757 if (!stream->psnr_count)
1758 return;
1760 fprintf(stderr, "Stream %d PSNR (Overall/Avg/Y/U/V)", stream->index);
1761 ovpsnr = sse_to_psnr((double)stream->psnr_samples_total, peak,
1762 (double)stream->psnr_sse_total);
1763 fprintf(stderr, " %.3f", ovpsnr);
1765 for (i = 0; i < 4; i++) {
1766 fprintf(stderr, " %.3f", stream->psnr_totals[i] / stream->psnr_count);
1768 fprintf(stderr, "\n");
1772 static float usec_to_fps(uint64_t usec, unsigned int frames) {
1773 return (float)(usec > 0 ? frames * 1000000.0 / (float)usec : 0);
1776 static void test_decode(struct stream_state *stream,
1777 enum TestDecodeFatality fatal,
1778 const VpxInterface *codec) {
1779 vpx_image_t enc_img, dec_img;
1781 if (stream->mismatch_seen)
1782 return;
1784 /* Get the internal reference frame */
1785 if (strcmp(codec->name, "vp8") == 0) {
1786 struct vpx_ref_frame ref_enc, ref_dec;
1787 int width, height;
1789 width = (stream->config.cfg.g_w + 15) & ~15;
1790 height = (stream->config.cfg.g_h + 15) & ~15;
1791 vpx_img_alloc(&ref_enc.img, VPX_IMG_FMT_I420, width, height, 1);
1792 enc_img = ref_enc.img;
1793 vpx_img_alloc(&ref_dec.img, VPX_IMG_FMT_I420, width, height, 1);
1794 dec_img = ref_dec.img;
1796 ref_enc.frame_type = VP8_LAST_FRAME;
1797 ref_dec.frame_type = VP8_LAST_FRAME;
1798 vpx_codec_control(&stream->encoder, VP8_COPY_REFERENCE, &ref_enc);
1799 vpx_codec_control(&stream->decoder, VP8_COPY_REFERENCE, &ref_dec);
1800 } else {
1801 struct vp9_ref_frame ref_enc, ref_dec;
1803 ref_enc.idx = 0;
1804 ref_dec.idx = 0;
1805 vpx_codec_control(&stream->encoder, VP9_GET_REFERENCE, &ref_enc);
1806 enc_img = ref_enc.img;
1807 vpx_codec_control(&stream->decoder, VP9_GET_REFERENCE, &ref_dec);
1808 dec_img = ref_dec.img;
1809 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1810 if ((enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) !=
1811 (dec_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH)) {
1812 if (enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1813 vpx_img_alloc(&enc_img, enc_img.fmt - VPX_IMG_FMT_HIGHBITDEPTH,
1814 enc_img.d_w, enc_img.d_h, 16);
1815 vpx_img_truncate_16_to_8(&enc_img, &ref_enc.img);
1817 if (dec_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1818 vpx_img_alloc(&dec_img, dec_img.fmt - VPX_IMG_FMT_HIGHBITDEPTH,
1819 dec_img.d_w, dec_img.d_h, 16);
1820 vpx_img_truncate_16_to_8(&dec_img, &ref_dec.img);
1823 #endif
1825 ctx_exit_on_error(&stream->encoder, "Failed to get encoder reference frame");
1826 ctx_exit_on_error(&stream->decoder, "Failed to get decoder reference frame");
1828 if (!compare_img(&enc_img, &dec_img)) {
1829 int y[4], u[4], v[4];
1830 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1831 if (enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1832 find_mismatch_high(&enc_img, &dec_img, y, u, v);
1833 } else {
1834 find_mismatch(&enc_img, &dec_img, y, u, v);
1836 #else
1837 find_mismatch(&enc_img, &dec_img, y, u, v);
1838 #endif
1839 stream->decoder.err = 1;
1840 warn_or_exit_on_error(&stream->decoder, fatal == TEST_DECODE_FATAL,
1841 "Stream %d: Encode/decode mismatch on frame %d at"
1842 " Y[%d, %d] {%d/%d},"
1843 " U[%d, %d] {%d/%d},"
1844 " V[%d, %d] {%d/%d}",
1845 stream->index, stream->frames_out,
1846 y[0], y[1], y[2], y[3],
1847 u[0], u[1], u[2], u[3],
1848 v[0], v[1], v[2], v[3]);
1849 stream->mismatch_seen = stream->frames_out;
1852 vpx_img_free(&enc_img);
1853 vpx_img_free(&dec_img);
1857 static void print_time(const char *label, int64_t etl) {
1858 int64_t hours;
1859 int64_t mins;
1860 int64_t secs;
1862 if (etl >= 0) {
1863 hours = etl / 3600;
1864 etl -= hours * 3600;
1865 mins = etl / 60;
1866 etl -= mins * 60;
1867 secs = etl;
1869 fprintf(stderr, "[%3s %2"PRId64":%02"PRId64":%02"PRId64"] ",
1870 label, hours, mins, secs);
1871 } else {
1872 fprintf(stderr, "[%3s unknown] ", label);
1877 int main(int argc, const char **argv_) {
1878 int pass;
1879 vpx_image_t raw;
1880 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1881 vpx_image_t raw_shift;
1882 int allocated_raw_shift = 0;
1883 int use_16bit_internal = 0;
1884 int input_shift = 0;
1885 #endif
1886 int frame_avail, got_data;
1888 struct VpxInputContext input;
1889 struct VpxEncoderConfig global;
1890 struct stream_state *streams = NULL;
1891 char **argv, **argi;
1892 uint64_t cx_time = 0;
1893 int stream_cnt = 0;
1894 int res = 0;
1896 memset(&input, 0, sizeof(input));
1897 exec_name = argv_[0];
1899 if (argc < 3)
1900 usage_exit();
1902 /* Setup default input stream settings */
1903 input.framerate.numerator = 30;
1904 input.framerate.denominator = 1;
1905 input.only_i420 = 1;
1906 input.bit_depth = 0;
1908 /* First parse the global configuration values, because we want to apply
1909 * other parameters on top of the default configuration provided by the
1910 * codec.
1912 argv = argv_dup(argc - 1, argv_ + 1);
1913 parse_global_config(&global, argv);
1915 switch (global.color_type) {
1916 case I420:
1917 input.fmt = VPX_IMG_FMT_I420;
1918 break;
1919 case I422:
1920 input.fmt = VPX_IMG_FMT_I422;
1921 break;
1922 case I444:
1923 input.fmt = VPX_IMG_FMT_I444;
1924 break;
1925 case I440:
1926 input.fmt = VPX_IMG_FMT_I440;
1927 break;
1928 case YV12:
1929 input.fmt = VPX_IMG_FMT_YV12;
1930 break;
1934 /* Now parse each stream's parameters. Using a local scope here
1935 * due to the use of 'stream' as loop variable in FOREACH_STREAM
1936 * loops
1938 struct stream_state *stream = NULL;
1940 do {
1941 stream = new_stream(&global, stream);
1942 stream_cnt++;
1943 if (!streams)
1944 streams = stream;
1945 } while (parse_stream_params(&global, stream, argv));
1948 /* Check for unrecognized options */
1949 for (argi = argv; *argi; argi++)
1950 if (argi[0][0] == '-' && argi[0][1])
1951 die("Error: Unrecognized option %s\n", *argi);
1953 FOREACH_STREAM(check_encoder_config(global.disable_warning_prompt,
1954 &global, &stream->config.cfg););
1956 /* Handle non-option arguments */
1957 input.filename = argv[0];
1959 if (!input.filename)
1960 usage_exit();
1962 /* Decide if other chroma subsamplings than 4:2:0 are supported */
1963 if (global.codec->fourcc == VP9_FOURCC)
1964 input.only_i420 = 0;
1966 for (pass = global.pass ? global.pass - 1 : 0; pass < global.passes; pass++) {
1967 int frames_in = 0, seen_frames = 0;
1968 int64_t estimated_time_left = -1;
1969 int64_t average_rate = -1;
1970 int64_t lagged_count = 0;
1972 open_input_file(&input);
1974 /* If the input file doesn't specify its w/h (raw files), try to get
1975 * the data from the first stream's configuration.
1977 if (!input.width || !input.height) {
1978 FOREACH_STREAM({
1979 if (stream->config.cfg.g_w && stream->config.cfg.g_h) {
1980 input.width = stream->config.cfg.g_w;
1981 input.height = stream->config.cfg.g_h;
1982 break;
1987 /* Update stream configurations from the input file's parameters */
1988 if (!input.width || !input.height)
1989 fatal("Specify stream dimensions with --width (-w) "
1990 " and --height (-h)");
1992 /* If input file does not specify bit-depth but input-bit-depth parameter
1993 * exists, assume that to be the input bit-depth. However, if the
1994 * input-bit-depth paramter does not exist, assume the input bit-depth
1995 * to be the same as the codec bit-depth.
1997 if (!input.bit_depth) {
1998 FOREACH_STREAM({
1999 if (stream->config.cfg.g_input_bit_depth)
2000 input.bit_depth = stream->config.cfg.g_input_bit_depth;
2001 else
2002 input.bit_depth = stream->config.cfg.g_input_bit_depth =
2003 (int)stream->config.cfg.g_bit_depth;
2005 if (input.bit_depth > 8) input.fmt |= VPX_IMG_FMT_HIGHBITDEPTH;
2006 } else {
2007 FOREACH_STREAM({
2008 stream->config.cfg.g_input_bit_depth = input.bit_depth;
2012 FOREACH_STREAM(set_stream_dimensions(stream, input.width, input.height));
2013 FOREACH_STREAM(validate_stream_config(stream, &global));
2015 /* Ensure that --passes and --pass are consistent. If --pass is set and
2016 * --passes=2, ensure --fpf was set.
2018 if (global.pass && global.passes == 2)
2019 FOREACH_STREAM( {
2020 if (!stream->config.stats_fn)
2021 die("Stream %d: Must specify --fpf when --pass=%d"
2022 " and --passes=2\n", stream->index, global.pass);
2025 #if !CONFIG_WEBM_IO
2026 FOREACH_STREAM({
2027 stream->config.write_webm = 0;
2028 warn("vpxenc was compiled without WebM container support."
2029 "Producing IVF output");
2031 #endif
2033 /* Use the frame rate from the file only if none was specified
2034 * on the command-line.
2036 if (!global.have_framerate) {
2037 global.framerate.num = input.framerate.numerator;
2038 global.framerate.den = input.framerate.denominator;
2041 FOREACH_STREAM(set_default_kf_interval(stream, &global));
2043 /* Show configuration */
2044 if (global.verbose && pass == 0)
2045 FOREACH_STREAM(show_stream_config(stream, &global, &input));
2047 if (pass == (global.pass ? global.pass - 1 : 0)) {
2048 if (input.file_type == FILE_TYPE_Y4M)
2049 /*The Y4M reader does its own allocation.
2050 Just initialize this here to avoid problems if we never read any
2051 frames.*/
2052 memset(&raw, 0, sizeof(raw));
2053 else
2054 vpx_img_alloc(&raw, input.fmt, input.width, input.height, 32);
2056 FOREACH_STREAM(stream->rate_hist =
2057 init_rate_histogram(&stream->config.cfg,
2058 &global.framerate));
2061 FOREACH_STREAM(setup_pass(stream, &global, pass));
2062 FOREACH_STREAM(open_output_file(stream, &global,
2063 &input.pixel_aspect_ratio));
2064 FOREACH_STREAM(initialize_encoder(stream, &global));
2066 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
2067 if (strcmp(global.codec->name, "vp9") == 0) {
2068 // Check to see if at least one stream uses 16 bit internal.
2069 // Currently assume that the bit_depths for all streams using
2070 // highbitdepth are the same.
2071 FOREACH_STREAM({
2072 if (stream->config.use_16bit_internal) {
2073 use_16bit_internal = 1;
2075 if (stream->config.cfg.g_profile == 0) {
2076 input_shift = 0;
2077 } else {
2078 input_shift = (int)stream->config.cfg.g_bit_depth -
2079 stream->config.cfg.g_input_bit_depth;
2083 #endif
2085 frame_avail = 1;
2086 got_data = 0;
2088 while (frame_avail || got_data) {
2089 struct vpx_usec_timer timer;
2091 if (!global.limit || frames_in < global.limit) {
2092 frame_avail = read_frame(&input, &raw);
2094 if (frame_avail)
2095 frames_in++;
2096 seen_frames = frames_in > global.skip_frames ?
2097 frames_in - global.skip_frames : 0;
2099 if (!global.quiet) {
2100 float fps = usec_to_fps(cx_time, seen_frames);
2101 fprintf(stderr, "\rPass %d/%d ", pass + 1, global.passes);
2103 if (stream_cnt == 1)
2104 fprintf(stderr,
2105 "frame %4d/%-4d %7"PRId64"B ",
2106 frames_in, streams->frames_out, (int64_t)streams->nbytes);
2107 else
2108 fprintf(stderr, "frame %4d ", frames_in);
2110 fprintf(stderr, "%7"PRId64" %s %.2f %s ",
2111 cx_time > 9999999 ? cx_time / 1000 : cx_time,
2112 cx_time > 9999999 ? "ms" : "us",
2113 fps >= 1.0 ? fps : fps * 60,
2114 fps >= 1.0 ? "fps" : "fpm");
2115 print_time("ETA", estimated_time_left);
2118 } else
2119 frame_avail = 0;
2121 if (frames_in > global.skip_frames) {
2122 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
2123 vpx_image_t *frame_to_encode;
2124 if (input_shift || (use_16bit_internal && input.bit_depth == 8)) {
2125 assert(use_16bit_internal);
2126 // Input bit depth and stream bit depth do not match, so up
2127 // shift frame to stream bit depth
2128 if (!allocated_raw_shift) {
2129 vpx_img_alloc(&raw_shift, raw.fmt | VPX_IMG_FMT_HIGHBITDEPTH,
2130 input.width, input.height, 32);
2131 allocated_raw_shift = 1;
2133 vpx_img_upshift(&raw_shift, &raw, input_shift);
2134 frame_to_encode = &raw_shift;
2135 } else {
2136 frame_to_encode = &raw;
2138 vpx_usec_timer_start(&timer);
2139 if (use_16bit_internal) {
2140 assert(frame_to_encode->fmt & VPX_IMG_FMT_HIGHBITDEPTH);
2141 FOREACH_STREAM({
2142 if (stream->config.use_16bit_internal)
2143 encode_frame(stream, &global,
2144 frame_avail ? frame_to_encode : NULL,
2145 frames_in);
2146 else
2147 assert(0);
2149 } else {
2150 assert((frame_to_encode->fmt & VPX_IMG_FMT_HIGHBITDEPTH) == 0);
2151 FOREACH_STREAM(encode_frame(stream, &global,
2152 frame_avail ? frame_to_encode : NULL,
2153 frames_in));
2155 #else
2156 vpx_usec_timer_start(&timer);
2157 FOREACH_STREAM(encode_frame(stream, &global,
2158 frame_avail ? &raw : NULL,
2159 frames_in));
2160 #endif
2161 vpx_usec_timer_mark(&timer);
2162 cx_time += vpx_usec_timer_elapsed(&timer);
2164 FOREACH_STREAM(update_quantizer_histogram(stream));
2166 got_data = 0;
2167 FOREACH_STREAM(get_cx_data(stream, &global, &got_data));
2169 if (!got_data && input.length && streams != NULL &&
2170 !streams->frames_out) {
2171 lagged_count = global.limit ? seen_frames : ftello(input.file);
2172 } else if (input.length) {
2173 int64_t remaining;
2174 int64_t rate;
2176 if (global.limit) {
2177 const int64_t frame_in_lagged = (seen_frames - lagged_count) * 1000;
2179 rate = cx_time ? frame_in_lagged * (int64_t)1000000 / cx_time : 0;
2180 remaining = 1000 * (global.limit - global.skip_frames
2181 - seen_frames + lagged_count);
2182 } else {
2183 const int64_t input_pos = ftello(input.file);
2184 const int64_t input_pos_lagged = input_pos - lagged_count;
2185 const int64_t limit = input.length;
2187 rate = cx_time ? input_pos_lagged * (int64_t)1000000 / cx_time : 0;
2188 remaining = limit - input_pos + lagged_count;
2191 average_rate = (average_rate <= 0)
2192 ? rate
2193 : (average_rate * 7 + rate) / 8;
2194 estimated_time_left = average_rate ? remaining / average_rate : -1;
2197 if (got_data && global.test_decode != TEST_DECODE_OFF)
2198 FOREACH_STREAM(test_decode(stream, global.test_decode, global.codec));
2201 fflush(stdout);
2202 if (!global.quiet)
2203 fprintf(stderr, "\033[K");
2206 if (stream_cnt > 1)
2207 fprintf(stderr, "\n");
2209 if (!global.quiet) {
2210 FOREACH_STREAM(fprintf(stderr,
2211 "\rPass %d/%d frame %4d/%-4d %7"PRId64"B %7"PRId64"b/f %7"PRId64"b/s"
2212 " %7"PRId64" %s (%.2f fps)\033[K\n",
2213 pass + 1,
2214 global.passes, frames_in, stream->frames_out, (int64_t)stream->nbytes,
2215 seen_frames ? (int64_t)(stream->nbytes * 8 / seen_frames) : 0,
2216 seen_frames ? (int64_t)stream->nbytes * 8 *
2217 (int64_t)global.framerate.num / global.framerate.den /
2218 seen_frames : 0,
2219 stream->cx_time > 9999999 ? stream->cx_time / 1000 : stream->cx_time,
2220 stream->cx_time > 9999999 ? "ms" : "us",
2221 usec_to_fps(stream->cx_time, seen_frames)));
2224 if (global.show_psnr) {
2225 if (global.codec->fourcc == VP9_FOURCC) {
2226 FOREACH_STREAM(
2227 show_psnr(stream, (1 << stream->config.cfg.g_input_bit_depth) - 1));
2228 } else {
2229 FOREACH_STREAM(show_psnr(stream, 255.0));
2233 FOREACH_STREAM(vpx_codec_destroy(&stream->encoder));
2235 if (global.test_decode != TEST_DECODE_OFF) {
2236 FOREACH_STREAM(vpx_codec_destroy(&stream->decoder));
2239 close_input_file(&input);
2241 if (global.test_decode == TEST_DECODE_FATAL) {
2242 FOREACH_STREAM(res |= stream->mismatch_seen);
2244 FOREACH_STREAM(close_output_file(stream, global.codec->fourcc));
2246 FOREACH_STREAM(stats_close(&stream->stats, global.passes - 1));
2248 #if CONFIG_FP_MB_STATS
2249 FOREACH_STREAM(stats_close(&stream->fpmb_stats, global.passes - 1));
2250 #endif
2252 if (global.pass)
2253 break;
2256 if (global.show_q_hist_buckets)
2257 FOREACH_STREAM(show_q_histogram(stream->counts,
2258 global.show_q_hist_buckets));
2260 if (global.show_rate_hist_buckets)
2261 FOREACH_STREAM(show_rate_histogram(stream->rate_hist,
2262 &stream->config.cfg,
2263 global.show_rate_hist_buckets));
2264 FOREACH_STREAM(destroy_rate_histogram(stream->rate_hist));
2266 #if CONFIG_INTERNAL_STATS
2267 /* TODO(jkoleszar): This doesn't belong in this executable. Do it for now,
2268 * to match some existing utilities.
2270 if (!(global.pass == 1 && global.passes == 2))
2271 FOREACH_STREAM({
2272 FILE *f = fopen("opsnr.stt", "a");
2273 if (stream->mismatch_seen) {
2274 fprintf(f, "First mismatch occurred in frame %d\n",
2275 stream->mismatch_seen);
2276 } else {
2277 fprintf(f, "No mismatch detected in recon buffers\n");
2279 fclose(f);
2281 #endif
2283 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
2284 if (allocated_raw_shift)
2285 vpx_img_free(&raw_shift);
2286 #endif
2287 vpx_img_free(&raw);
2288 free(argv);
2289 free(streams);
2290 return res ? EXIT_FAILURE : EXIT_SUCCESS;