BUG FIX: Remove counts param
[aom.git] / vpxenc.c
blobbe9b8b61ddb85e80f5d3bd7d256ee199d64a2a3b
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 cpu_used_vp8 = ARG_DEF(
334 NULL, "cpu-used", 1, "CPU Used (-16..16)");
335 static const arg_def_t cpu_used_vp9 = ARG_DEF(
336 NULL, "cpu-used", 1, "CPU Used (-8..8)");
337 static const arg_def_t auto_altref = ARG_DEF(
338 NULL, "auto-alt-ref", 1, "Enable automatic alt reference frames");
339 static const arg_def_t arnr_maxframes = ARG_DEF(
340 NULL, "arnr-maxframes", 1, "AltRef max frames (0..15)");
341 static const arg_def_t arnr_strength = ARG_DEF(
342 NULL, "arnr-strength", 1, "AltRef filter strength (0..6)");
343 static const arg_def_t arnr_type = ARG_DEF(
344 NULL, "arnr-type", 1, "AltRef type");
345 static const struct arg_enum_list tuning_enum[] = {
346 {"psnr", VP8_TUNE_PSNR},
347 {"ssim", VP8_TUNE_SSIM},
348 {NULL, 0}
350 static const arg_def_t tune_ssim = ARG_DEF_ENUM(
351 NULL, "tune", 1, "Material to favor", tuning_enum);
352 static const arg_def_t cq_level = ARG_DEF(
353 NULL, "cq-level", 1, "Constant/Constrained Quality level");
354 static const arg_def_t max_intra_rate_pct = ARG_DEF(
355 NULL, "max-intra-rate", 1, "Max I-frame bitrate (pct)");
356 static const arg_def_t max_inter_rate_pct = ARG_DEF(
357 NULL, "max-inter-rate", 1, "Max P-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 static const arg_def_t screen_content_mode = ARG_DEF(NULL, "screen-content-mode", 1,
362 "Screen content mode");
364 #if CONFIG_VP8_ENCODER
365 static const arg_def_t token_parts = ARG_DEF(
366 NULL, "token-parts", 1, "Number of token partitions to use, log2");
367 static const arg_def_t *vp8_args[] = {
368 &cpu_used_vp8, &auto_altref, &noise_sens, &sharpness, &static_thresh,
369 &token_parts, &arnr_maxframes, &arnr_strength, &arnr_type,
370 &tune_ssim, &cq_level, &max_intra_rate_pct, &screen_content_mode,
371 NULL
373 static const int vp8_arg_ctrl_map[] = {
374 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
375 VP8E_SET_NOISE_SENSITIVITY, VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
376 VP8E_SET_TOKEN_PARTITIONS,
377 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
378 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
379 VP8E_SET_SCREEN_CONTENT_MODE,
382 #endif
384 #if CONFIG_VP9_ENCODER
385 static const arg_def_t tile_cols = ARG_DEF(
386 NULL, "tile-columns", 1, "Number of tile columns to use, log2");
387 static const arg_def_t tile_rows = ARG_DEF(
388 NULL, "tile-rows", 1, "Number of tile rows to use, log2");
389 static const arg_def_t lossless = ARG_DEF(
390 NULL, "lossless", 1, "Lossless mode");
391 static const arg_def_t frame_parallel_decoding = ARG_DEF(
392 NULL, "frame-parallel", 1, "Enable frame parallel decodability features");
393 static const arg_def_t aq_mode = ARG_DEF(
394 NULL, "aq-mode", 1,
395 "Adaptive quantization mode (0: off (default), 1: variance 2: complexity, "
396 "3: cyclic refresh)");
397 static const arg_def_t frame_periodic_boost = ARG_DEF(
398 NULL, "frame-boost", 1,
399 "Enable frame periodic boost (0: off (default), 1: on)");
401 static const struct arg_enum_list color_space_enum[] = {
402 { "unknown", VPX_CS_UNKNOWN },
403 { "bt601", VPX_CS_BT_601 },
404 { "bt709", VPX_CS_BT_709 },
405 { "smpte170", VPX_CS_SMPTE_170 },
406 { "smpte240", VPX_CS_SMPTE_240 },
407 { "bt2020", VPX_CS_BT_2020 },
408 { "reserved", VPX_CS_RESERVED },
409 { "sRGB", VPX_CS_SRGB },
410 { NULL, 0 }
413 static const arg_def_t input_color_space = ARG_DEF_ENUM(
414 NULL, "color-space", 1,
415 "The color space of input content:", color_space_enum);
417 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
418 static const struct arg_enum_list bitdepth_enum[] = {
419 {"8", VPX_BITS_8},
420 {"10", VPX_BITS_10},
421 {"12", VPX_BITS_12},
422 {NULL, 0}
425 static const arg_def_t bitdeptharg = ARG_DEF_ENUM(
426 "b", "bit-depth", 1,
427 "Bit depth for codec (8 for version <=1, 10 or 12 for version 2)",
428 bitdepth_enum);
429 static const arg_def_t inbitdeptharg = ARG_DEF(
430 NULL, "input-bit-depth", 1, "Bit depth of input");
431 #endif
433 static const struct arg_enum_list tune_content_enum[] = {
434 {"default", VP9E_CONTENT_DEFAULT},
435 {"screen", VP9E_CONTENT_SCREEN},
436 {NULL, 0}
439 static const arg_def_t tune_content = ARG_DEF_ENUM(
440 NULL, "tune-content", 1, "Tune content type", tune_content_enum);
442 static const arg_def_t *vp9_args[] = {
443 &cpu_used_vp9, &auto_altref, &sharpness, &static_thresh,
444 &tile_cols, &tile_rows, &arnr_maxframes, &arnr_strength, &arnr_type,
445 &tune_ssim, &cq_level, &max_intra_rate_pct, &max_inter_rate_pct,
446 &gf_cbr_boost_pct, &lossless,
447 &frame_parallel_decoding, &aq_mode, &frame_periodic_boost,
448 &noise_sens, &tune_content, &input_color_space,
449 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
450 &bitdeptharg, &inbitdeptharg,
451 #endif
452 NULL
454 static const int vp9_arg_ctrl_map[] = {
455 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
456 VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
457 VP9E_SET_TILE_COLUMNS, VP9E_SET_TILE_ROWS,
458 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
459 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
460 VP9E_SET_MAX_INTER_BITRATE_PCT, VP9E_SET_GF_CBR_BOOST_PCT,
461 VP9E_SET_LOSSLESS, VP9E_SET_FRAME_PARALLEL_DECODING, VP9E_SET_AQ_MODE,
462 VP9E_SET_FRAME_PERIODIC_BOOST, VP9E_SET_NOISE_SENSITIVITY,
463 VP9E_SET_TUNE_CONTENT, VP9E_SET_COLOR_SPACE,
466 #endif
468 static const arg_def_t *no_args[] = { NULL };
470 void usage_exit(void) {
471 int i;
472 const int num_encoder = get_vpx_encoder_count();
474 fprintf(stderr, "Usage: %s <options> -o dst_filename src_filename \n",
475 exec_name);
477 fprintf(stderr, "\nOptions:\n");
478 arg_show_usage(stderr, main_args);
479 fprintf(stderr, "\nEncoder Global Options:\n");
480 arg_show_usage(stderr, global_args);
481 fprintf(stderr, "\nRate Control Options:\n");
482 arg_show_usage(stderr, rc_args);
483 fprintf(stderr, "\nTwopass Rate Control Options:\n");
484 arg_show_usage(stderr, rc_twopass_args);
485 fprintf(stderr, "\nKeyframe Placement Options:\n");
486 arg_show_usage(stderr, kf_args);
487 #if CONFIG_VP8_ENCODER
488 fprintf(stderr, "\nVP8 Specific Options:\n");
489 arg_show_usage(stderr, vp8_args);
490 #endif
491 #if CONFIG_VP9_ENCODER
492 fprintf(stderr, "\nVP9 Specific Options:\n");
493 arg_show_usage(stderr, vp9_args);
494 #endif
495 fprintf(stderr, "\nStream timebase (--timebase):\n"
496 " The desired precision of timestamps in the output, expressed\n"
497 " in fractional seconds. Default is 1/1000.\n");
498 fprintf(stderr, "\nIncluded encoders:\n\n");
500 for (i = 0; i < num_encoder; ++i) {
501 const VpxInterface *const encoder = get_vpx_encoder_by_index(i);
502 const char* defstr = (i == (num_encoder - 1)) ? "(default)" : "";
503 fprintf(stderr, " %-6s - %s %s\n",
504 encoder->name, vpx_codec_iface_name(encoder->codec_interface()),
505 defstr);
507 fprintf(stderr, "\n ");
508 fprintf(stderr, "Use --codec to switch to a non-default encoder.\n\n");
510 exit(EXIT_FAILURE);
513 #define mmin(a, b) ((a) < (b) ? (a) : (b))
515 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
516 static void find_mismatch_high(const vpx_image_t *const img1,
517 const vpx_image_t *const img2,
518 int yloc[4], int uloc[4], int vloc[4]) {
519 uint16_t *plane1, *plane2;
520 uint32_t stride1, stride2;
521 const uint32_t bsize = 64;
522 const uint32_t bsizey = bsize >> img1->y_chroma_shift;
523 const uint32_t bsizex = bsize >> img1->x_chroma_shift;
524 const uint32_t c_w =
525 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
526 const uint32_t c_h =
527 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
528 int match = 1;
529 uint32_t i, j;
530 yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
531 plane1 = (uint16_t*)img1->planes[VPX_PLANE_Y];
532 plane2 = (uint16_t*)img2->planes[VPX_PLANE_Y];
533 stride1 = img1->stride[VPX_PLANE_Y]/2;
534 stride2 = img2->stride[VPX_PLANE_Y]/2;
535 for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
536 for (j = 0; match && j < img1->d_w; j += bsize) {
537 int k, l;
538 const int si = mmin(i + bsize, img1->d_h) - i;
539 const int sj = mmin(j + bsize, img1->d_w) - j;
540 for (k = 0; match && k < si; ++k) {
541 for (l = 0; match && l < sj; ++l) {
542 if (*(plane1 + (i + k) * stride1 + j + l) !=
543 *(plane2 + (i + k) * stride2 + j + l)) {
544 yloc[0] = i + k;
545 yloc[1] = j + l;
546 yloc[2] = *(plane1 + (i + k) * stride1 + j + l);
547 yloc[3] = *(plane2 + (i + k) * stride2 + j + l);
548 match = 0;
549 break;
556 uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
557 plane1 = (uint16_t*)img1->planes[VPX_PLANE_U];
558 plane2 = (uint16_t*)img2->planes[VPX_PLANE_U];
559 stride1 = img1->stride[VPX_PLANE_U]/2;
560 stride2 = img2->stride[VPX_PLANE_U]/2;
561 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
562 for (j = 0; match && j < c_w; j += bsizex) {
563 int k, l;
564 const int si = mmin(i + bsizey, c_h - i);
565 const int sj = mmin(j + bsizex, c_w - j);
566 for (k = 0; match && k < si; ++k) {
567 for (l = 0; match && l < sj; ++l) {
568 if (*(plane1 + (i + k) * stride1 + j + l) !=
569 *(plane2 + (i + k) * stride2 + j + l)) {
570 uloc[0] = i + k;
571 uloc[1] = j + l;
572 uloc[2] = *(plane1 + (i + k) * stride1 + j + l);
573 uloc[3] = *(plane2 + (i + k) * stride2 + j + l);
574 match = 0;
575 break;
582 vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
583 plane1 = (uint16_t*)img1->planes[VPX_PLANE_V];
584 plane2 = (uint16_t*)img2->planes[VPX_PLANE_V];
585 stride1 = img1->stride[VPX_PLANE_V]/2;
586 stride2 = img2->stride[VPX_PLANE_V]/2;
587 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
588 for (j = 0; match && j < c_w; j += bsizex) {
589 int k, l;
590 const int si = mmin(i + bsizey, c_h - i);
591 const int sj = mmin(j + bsizex, c_w - j);
592 for (k = 0; match && k < si; ++k) {
593 for (l = 0; match && l < sj; ++l) {
594 if (*(plane1 + (i + k) * stride1 + j + l) !=
595 *(plane2 + (i + k) * stride2 + j + l)) {
596 vloc[0] = i + k;
597 vloc[1] = j + l;
598 vloc[2] = *(plane1 + (i + k) * stride1 + j + l);
599 vloc[3] = *(plane2 + (i + k) * stride2 + j + l);
600 match = 0;
601 break;
608 #endif
610 static void find_mismatch(const vpx_image_t *const img1,
611 const vpx_image_t *const img2,
612 int yloc[4], int uloc[4], int vloc[4]) {
613 const uint32_t bsize = 64;
614 const uint32_t bsizey = bsize >> img1->y_chroma_shift;
615 const uint32_t bsizex = bsize >> img1->x_chroma_shift;
616 const uint32_t c_w =
617 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
618 const uint32_t c_h =
619 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
620 int match = 1;
621 uint32_t i, j;
622 yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
623 for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
624 for (j = 0; match && j < img1->d_w; j += bsize) {
625 int k, l;
626 const int si = mmin(i + bsize, img1->d_h) - i;
627 const int sj = mmin(j + bsize, img1->d_w) - j;
628 for (k = 0; match && k < si; ++k) {
629 for (l = 0; match && l < sj; ++l) {
630 if (*(img1->planes[VPX_PLANE_Y] +
631 (i + k) * img1->stride[VPX_PLANE_Y] + j + l) !=
632 *(img2->planes[VPX_PLANE_Y] +
633 (i + k) * img2->stride[VPX_PLANE_Y] + j + l)) {
634 yloc[0] = i + k;
635 yloc[1] = j + l;
636 yloc[2] = *(img1->planes[VPX_PLANE_Y] +
637 (i + k) * img1->stride[VPX_PLANE_Y] + j + l);
638 yloc[3] = *(img2->planes[VPX_PLANE_Y] +
639 (i + k) * img2->stride[VPX_PLANE_Y] + j + l);
640 match = 0;
641 break;
648 uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
649 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
650 for (j = 0; match && j < c_w; j += bsizex) {
651 int k, l;
652 const int si = mmin(i + bsizey, c_h - i);
653 const int sj = mmin(j + bsizex, c_w - j);
654 for (k = 0; match && k < si; ++k) {
655 for (l = 0; match && l < sj; ++l) {
656 if (*(img1->planes[VPX_PLANE_U] +
657 (i + k) * img1->stride[VPX_PLANE_U] + j + l) !=
658 *(img2->planes[VPX_PLANE_U] +
659 (i + k) * img2->stride[VPX_PLANE_U] + j + l)) {
660 uloc[0] = i + k;
661 uloc[1] = j + l;
662 uloc[2] = *(img1->planes[VPX_PLANE_U] +
663 (i + k) * img1->stride[VPX_PLANE_U] + j + l);
664 uloc[3] = *(img2->planes[VPX_PLANE_U] +
665 (i + k) * img2->stride[VPX_PLANE_U] + j + l);
666 match = 0;
667 break;
673 vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
674 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
675 for (j = 0; match && j < c_w; j += bsizex) {
676 int k, l;
677 const int si = mmin(i + bsizey, c_h - i);
678 const int sj = mmin(j + bsizex, c_w - j);
679 for (k = 0; match && k < si; ++k) {
680 for (l = 0; match && l < sj; ++l) {
681 if (*(img1->planes[VPX_PLANE_V] +
682 (i + k) * img1->stride[VPX_PLANE_V] + j + l) !=
683 *(img2->planes[VPX_PLANE_V] +
684 (i + k) * img2->stride[VPX_PLANE_V] + j + l)) {
685 vloc[0] = i + k;
686 vloc[1] = j + l;
687 vloc[2] = *(img1->planes[VPX_PLANE_V] +
688 (i + k) * img1->stride[VPX_PLANE_V] + j + l);
689 vloc[3] = *(img2->planes[VPX_PLANE_V] +
690 (i + k) * img2->stride[VPX_PLANE_V] + j + l);
691 match = 0;
692 break;
700 static int compare_img(const vpx_image_t *const img1,
701 const vpx_image_t *const img2) {
702 uint32_t l_w = img1->d_w;
703 uint32_t c_w =
704 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
705 const uint32_t c_h =
706 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
707 uint32_t i;
708 int match = 1;
710 match &= (img1->fmt == img2->fmt);
711 match &= (img1->d_w == img2->d_w);
712 match &= (img1->d_h == img2->d_h);
713 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
714 if (img1->fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
715 l_w *= 2;
716 c_w *= 2;
718 #endif
720 for (i = 0; i < img1->d_h; ++i)
721 match &= (memcmp(img1->planes[VPX_PLANE_Y] + i * img1->stride[VPX_PLANE_Y],
722 img2->planes[VPX_PLANE_Y] + i * img2->stride[VPX_PLANE_Y],
723 l_w) == 0);
725 for (i = 0; i < c_h; ++i)
726 match &= (memcmp(img1->planes[VPX_PLANE_U] + i * img1->stride[VPX_PLANE_U],
727 img2->planes[VPX_PLANE_U] + i * img2->stride[VPX_PLANE_U],
728 c_w) == 0);
730 for (i = 0; i < c_h; ++i)
731 match &= (memcmp(img1->planes[VPX_PLANE_V] + i * img1->stride[VPX_PLANE_V],
732 img2->planes[VPX_PLANE_V] + i * img2->stride[VPX_PLANE_V],
733 c_w) == 0);
735 return match;
739 #define NELEMENTS(x) (sizeof(x)/sizeof(x[0]))
740 #define MAX(x,y) ((x)>(y)?(x):(y))
741 #if CONFIG_VP8_ENCODER && !CONFIG_VP9_ENCODER
742 #define ARG_CTRL_CNT_MAX NELEMENTS(vp8_arg_ctrl_map)
743 #elif !CONFIG_VP8_ENCODER && CONFIG_VP9_ENCODER
744 #define ARG_CTRL_CNT_MAX NELEMENTS(vp9_arg_ctrl_map)
745 #else
746 #define ARG_CTRL_CNT_MAX MAX(NELEMENTS(vp8_arg_ctrl_map), \
747 NELEMENTS(vp9_arg_ctrl_map))
748 #endif
750 #if !CONFIG_WEBM_IO
751 typedef int stereo_format_t;
752 struct EbmlGlobal { int debug; };
753 #endif
755 /* Per-stream configuration */
756 struct stream_config {
757 struct vpx_codec_enc_cfg cfg;
758 const char *out_fn;
759 const char *stats_fn;
760 #if CONFIG_FP_MB_STATS
761 const char *fpmb_stats_fn;
762 #endif
763 stereo_format_t stereo_fmt;
764 int arg_ctrls[ARG_CTRL_CNT_MAX][2];
765 int arg_ctrl_cnt;
766 int write_webm;
767 int have_kf_max_dist;
768 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
769 // whether to use 16bit internal buffers
770 int use_16bit_internal;
771 #endif
775 struct stream_state {
776 int index;
777 struct stream_state *next;
778 struct stream_config config;
779 FILE *file;
780 struct rate_hist *rate_hist;
781 struct EbmlGlobal ebml;
782 uint64_t psnr_sse_total;
783 uint64_t psnr_samples_total;
784 double psnr_totals[4];
785 int psnr_count;
786 int counts[64];
787 vpx_codec_ctx_t encoder;
788 unsigned int frames_out;
789 uint64_t cx_time;
790 size_t nbytes;
791 stats_io_t stats;
792 #if CONFIG_FP_MB_STATS
793 stats_io_t fpmb_stats;
794 #endif
795 struct vpx_image *img;
796 vpx_codec_ctx_t decoder;
797 int mismatch_seen;
801 static void validate_positive_rational(const char *msg,
802 struct vpx_rational *rat) {
803 if (rat->den < 0) {
804 rat->num *= -1;
805 rat->den *= -1;
808 if (rat->num < 0)
809 die("Error: %s must be positive\n", msg);
811 if (!rat->den)
812 die("Error: %s has zero denominator\n", msg);
816 static void parse_global_config(struct VpxEncoderConfig *global, char **argv) {
817 char **argi, **argj;
818 struct arg arg;
819 const int num_encoder = get_vpx_encoder_count();
821 if (num_encoder < 1)
822 die("Error: no valid encoder available\n");
824 /* Initialize default parameters */
825 memset(global, 0, sizeof(*global));
826 global->codec = get_vpx_encoder_by_index(num_encoder - 1);
827 global->passes = 0;
828 global->color_type = I420;
829 /* Assign default deadline to good quality */
830 global->deadline = VPX_DL_GOOD_QUALITY;
832 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
833 arg.argv_step = 1;
835 if (arg_match(&arg, &codecarg, argi)) {
836 global->codec = get_vpx_encoder_by_name(arg.val);
837 if (!global->codec)
838 die("Error: Unrecognized argument (%s) to --codec\n", arg.val);
839 } else if (arg_match(&arg, &passes, argi)) {
840 global->passes = arg_parse_uint(&arg);
842 if (global->passes < 1 || global->passes > 2)
843 die("Error: Invalid number of passes (%d)\n", global->passes);
844 } else if (arg_match(&arg, &pass_arg, argi)) {
845 global->pass = arg_parse_uint(&arg);
847 if (global->pass < 1 || global->pass > 2)
848 die("Error: Invalid pass selected (%d)\n",
849 global->pass);
850 } else if (arg_match(&arg, &usage, argi))
851 global->usage = arg_parse_uint(&arg);
852 else if (arg_match(&arg, &deadline, argi))
853 global->deadline = arg_parse_uint(&arg);
854 else if (arg_match(&arg, &best_dl, argi))
855 global->deadline = VPX_DL_BEST_QUALITY;
856 else if (arg_match(&arg, &good_dl, argi))
857 global->deadline = VPX_DL_GOOD_QUALITY;
858 else if (arg_match(&arg, &rt_dl, argi))
859 global->deadline = VPX_DL_REALTIME;
860 else if (arg_match(&arg, &use_yv12, argi))
861 global->color_type = YV12;
862 else if (arg_match(&arg, &use_i420, argi))
863 global->color_type = I420;
864 else if (arg_match(&arg, &use_i422, argi))
865 global->color_type = I422;
866 else if (arg_match(&arg, &use_i444, argi))
867 global->color_type = I444;
868 else if (arg_match(&arg, &use_i440, argi))
869 global->color_type = I440;
870 else if (arg_match(&arg, &quietarg, argi))
871 global->quiet = 1;
872 else if (arg_match(&arg, &verbosearg, argi))
873 global->verbose = 1;
874 else if (arg_match(&arg, &limit, argi))
875 global->limit = arg_parse_uint(&arg);
876 else if (arg_match(&arg, &skip, argi))
877 global->skip_frames = arg_parse_uint(&arg);
878 else if (arg_match(&arg, &psnrarg, argi))
879 global->show_psnr = 1;
880 else if (arg_match(&arg, &recontest, argi))
881 global->test_decode = arg_parse_enum_or_int(&arg);
882 else if (arg_match(&arg, &framerate, argi)) {
883 global->framerate = arg_parse_rational(&arg);
884 validate_positive_rational(arg.name, &global->framerate);
885 global->have_framerate = 1;
886 } else if (arg_match(&arg, &out_part, argi))
887 global->out_part = 1;
888 else if (arg_match(&arg, &debugmode, argi))
889 global->debug = 1;
890 else if (arg_match(&arg, &q_hist_n, argi))
891 global->show_q_hist_buckets = arg_parse_uint(&arg);
892 else if (arg_match(&arg, &rate_hist_n, argi))
893 global->show_rate_hist_buckets = arg_parse_uint(&arg);
894 else if (arg_match(&arg, &disable_warnings, argi))
895 global->disable_warnings = 1;
896 else if (arg_match(&arg, &disable_warning_prompt, argi))
897 global->disable_warning_prompt = 1;
898 else
899 argj++;
902 if (global->pass) {
903 /* DWIM: Assume the user meant passes=2 if pass=2 is specified */
904 if (global->pass > global->passes) {
905 warn("Assuming --pass=%d implies --passes=%d\n",
906 global->pass, global->pass);
907 global->passes = global->pass;
910 /* Validate global config */
911 if (global->passes == 0) {
912 #if CONFIG_VP9_ENCODER
913 // Make default VP9 passes = 2 until there is a better quality 1-pass
914 // encoder
915 if (global->codec != NULL && global->codec->name != NULL)
916 global->passes = (strcmp(global->codec->name, "vp9") == 0 &&
917 global->deadline != VPX_DL_REALTIME) ? 2 : 1;
918 #else
919 global->passes = 1;
920 #endif
923 if (global->deadline == VPX_DL_REALTIME &&
924 global->passes > 1) {
925 warn("Enforcing one-pass encoding in realtime mode\n");
926 global->passes = 1;
931 static void open_input_file(struct VpxInputContext *input) {
932 /* Parse certain options from the input file, if possible */
933 input->file = strcmp(input->filename, "-")
934 ? fopen(input->filename, "rb") : set_binary_mode(stdin);
936 if (!input->file)
937 fatal("Failed to open input file");
939 if (!fseeko(input->file, 0, SEEK_END)) {
940 /* Input file is seekable. Figure out how long it is, so we can get
941 * progress info.
943 input->length = ftello(input->file);
944 rewind(input->file);
947 /* For RAW input sources, these bytes will applied on the first frame
948 * in read_frame().
950 input->detect.buf_read = fread(input->detect.buf, 1, 4, input->file);
951 input->detect.position = 0;
953 if (input->detect.buf_read == 4
954 && file_is_y4m(input->detect.buf)) {
955 if (y4m_input_open(&input->y4m, input->file, input->detect.buf, 4,
956 input->only_i420) >= 0) {
957 input->file_type = FILE_TYPE_Y4M;
958 input->width = input->y4m.pic_w;
959 input->height = input->y4m.pic_h;
960 input->framerate.numerator = input->y4m.fps_n;
961 input->framerate.denominator = input->y4m.fps_d;
962 input->fmt = input->y4m.vpx_fmt;
963 input->bit_depth = input->y4m.bit_depth;
964 } else
965 fatal("Unsupported Y4M stream.");
966 } else if (input->detect.buf_read == 4 && fourcc_is_ivf(input->detect.buf)) {
967 fatal("IVF is not supported as input.");
968 } else {
969 input->file_type = FILE_TYPE_RAW;
974 static void close_input_file(struct VpxInputContext *input) {
975 fclose(input->file);
976 if (input->file_type == FILE_TYPE_Y4M)
977 y4m_input_close(&input->y4m);
980 static struct stream_state *new_stream(struct VpxEncoderConfig *global,
981 struct stream_state *prev) {
982 struct stream_state *stream;
984 stream = calloc(1, sizeof(*stream));
985 if (stream == NULL) {
986 fatal("Failed to allocate new stream.");
989 if (prev) {
990 memcpy(stream, prev, sizeof(*stream));
991 stream->index++;
992 prev->next = stream;
993 } else {
994 vpx_codec_err_t res;
996 /* Populate encoder configuration */
997 res = vpx_codec_enc_config_default(global->codec->codec_interface(),
998 &stream->config.cfg,
999 global->usage);
1000 if (res)
1001 fatal("Failed to get config: %s\n", vpx_codec_err_to_string(res));
1003 /* Change the default timebase to a high enough value so that the
1004 * encoder will always create strictly increasing timestamps.
1006 stream->config.cfg.g_timebase.den = 1000;
1008 /* Never use the library's default resolution, require it be parsed
1009 * from the file or set on the command line.
1011 stream->config.cfg.g_w = 0;
1012 stream->config.cfg.g_h = 0;
1014 /* Initialize remaining stream parameters */
1015 stream->config.write_webm = 1;
1016 #if CONFIG_WEBM_IO
1017 stream->config.stereo_fmt = STEREO_FORMAT_MONO;
1018 stream->ebml.last_pts_ns = -1;
1019 stream->ebml.writer = NULL;
1020 stream->ebml.segment = NULL;
1021 #endif
1023 /* Allows removal of the application version from the EBML tags */
1024 stream->ebml.debug = global->debug;
1026 /* Default lag_in_frames is 0 in realtime mode */
1027 if (global->deadline == VPX_DL_REALTIME)
1028 stream->config.cfg.g_lag_in_frames = 0;
1031 /* Output files must be specified for each stream */
1032 stream->config.out_fn = NULL;
1034 stream->next = NULL;
1035 return stream;
1039 static int parse_stream_params(struct VpxEncoderConfig *global,
1040 struct stream_state *stream,
1041 char **argv) {
1042 char **argi, **argj;
1043 struct arg arg;
1044 static const arg_def_t **ctrl_args = no_args;
1045 static const int *ctrl_args_map = NULL;
1046 struct stream_config *config = &stream->config;
1047 int eos_mark_found = 0;
1048 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1049 int test_16bit_internal = 0;
1050 #endif
1052 // Handle codec specific options
1053 if (0) {
1054 #if CONFIG_VP8_ENCODER
1055 } else if (strcmp(global->codec->name, "vp8") == 0) {
1056 ctrl_args = vp8_args;
1057 ctrl_args_map = vp8_arg_ctrl_map;
1058 #endif
1059 #if CONFIG_VP9_ENCODER
1060 } else if (strcmp(global->codec->name, "vp9") == 0) {
1061 ctrl_args = vp9_args;
1062 ctrl_args_map = vp9_arg_ctrl_map;
1063 #endif
1066 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
1067 arg.argv_step = 1;
1069 /* Once we've found an end-of-stream marker (--) we want to continue
1070 * shifting arguments but not consuming them.
1072 if (eos_mark_found) {
1073 argj++;
1074 continue;
1075 } else if (!strcmp(*argj, "--")) {
1076 eos_mark_found = 1;
1077 continue;
1080 if (arg_match(&arg, &outputfile, argi)) {
1081 config->out_fn = arg.val;
1082 } else if (arg_match(&arg, &fpf_name, argi)) {
1083 config->stats_fn = arg.val;
1084 #if CONFIG_FP_MB_STATS
1085 } else if (arg_match(&arg, &fpmbf_name, argi)) {
1086 config->fpmb_stats_fn = arg.val;
1087 #endif
1088 } else if (arg_match(&arg, &use_webm, argi)) {
1089 #if CONFIG_WEBM_IO
1090 config->write_webm = 1;
1091 #else
1092 die("Error: --webm specified but webm is disabled.");
1093 #endif
1094 } else if (arg_match(&arg, &use_ivf, argi)) {
1095 config->write_webm = 0;
1096 } else if (arg_match(&arg, &threads, argi)) {
1097 config->cfg.g_threads = arg_parse_uint(&arg);
1098 } else if (arg_match(&arg, &profile, argi)) {
1099 config->cfg.g_profile = arg_parse_uint(&arg);
1100 } else if (arg_match(&arg, &width, argi)) {
1101 config->cfg.g_w = arg_parse_uint(&arg);
1102 } else if (arg_match(&arg, &height, argi)) {
1103 config->cfg.g_h = arg_parse_uint(&arg);
1104 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1105 } else if (arg_match(&arg, &bitdeptharg, argi)) {
1106 config->cfg.g_bit_depth = arg_parse_enum_or_int(&arg);
1107 } else if (arg_match(&arg, &inbitdeptharg, argi)) {
1108 config->cfg.g_input_bit_depth = arg_parse_uint(&arg);
1109 #endif
1110 #if CONFIG_WEBM_IO
1111 } else if (arg_match(&arg, &stereo_mode, argi)) {
1112 config->stereo_fmt = arg_parse_enum_or_int(&arg);
1113 #endif
1114 } else if (arg_match(&arg, &timebase, argi)) {
1115 config->cfg.g_timebase = arg_parse_rational(&arg);
1116 validate_positive_rational(arg.name, &config->cfg.g_timebase);
1117 } else if (arg_match(&arg, &error_resilient, argi)) {
1118 config->cfg.g_error_resilient = arg_parse_uint(&arg);
1119 } else if (arg_match(&arg, &lag_in_frames, argi)) {
1120 config->cfg.g_lag_in_frames = arg_parse_uint(&arg);
1121 if (global->deadline == VPX_DL_REALTIME &&
1122 config->cfg.g_lag_in_frames != 0) {
1123 warn("non-zero %s option ignored in realtime mode.\n", arg.name);
1124 config->cfg.g_lag_in_frames = 0;
1126 } else if (arg_match(&arg, &dropframe_thresh, argi)) {
1127 config->cfg.rc_dropframe_thresh = arg_parse_uint(&arg);
1128 } else if (arg_match(&arg, &resize_allowed, argi)) {
1129 config->cfg.rc_resize_allowed = arg_parse_uint(&arg);
1130 } else if (arg_match(&arg, &resize_width, argi)) {
1131 config->cfg.rc_scaled_width = arg_parse_uint(&arg);
1132 } else if (arg_match(&arg, &resize_height, argi)) {
1133 config->cfg.rc_scaled_height = arg_parse_uint(&arg);
1134 } else if (arg_match(&arg, &resize_up_thresh, argi)) {
1135 config->cfg.rc_resize_up_thresh = arg_parse_uint(&arg);
1136 } else if (arg_match(&arg, &resize_down_thresh, argi)) {
1137 config->cfg.rc_resize_down_thresh = arg_parse_uint(&arg);
1138 } else if (arg_match(&arg, &end_usage, argi)) {
1139 config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
1140 } else if (arg_match(&arg, &target_bitrate, argi)) {
1141 config->cfg.rc_target_bitrate = arg_parse_uint(&arg);
1142 } else if (arg_match(&arg, &min_quantizer, argi)) {
1143 config->cfg.rc_min_quantizer = arg_parse_uint(&arg);
1144 } else if (arg_match(&arg, &max_quantizer, argi)) {
1145 config->cfg.rc_max_quantizer = arg_parse_uint(&arg);
1146 } else if (arg_match(&arg, &undershoot_pct, argi)) {
1147 config->cfg.rc_undershoot_pct = arg_parse_uint(&arg);
1148 } else if (arg_match(&arg, &overshoot_pct, argi)) {
1149 config->cfg.rc_overshoot_pct = arg_parse_uint(&arg);
1150 } else if (arg_match(&arg, &buf_sz, argi)) {
1151 config->cfg.rc_buf_sz = arg_parse_uint(&arg);
1152 } else if (arg_match(&arg, &buf_initial_sz, argi)) {
1153 config->cfg.rc_buf_initial_sz = arg_parse_uint(&arg);
1154 } else if (arg_match(&arg, &buf_optimal_sz, argi)) {
1155 config->cfg.rc_buf_optimal_sz = arg_parse_uint(&arg);
1156 } else if (arg_match(&arg, &bias_pct, argi)) {
1157 config->cfg.rc_2pass_vbr_bias_pct = arg_parse_uint(&arg);
1158 if (global->passes < 2)
1159 warn("option %s ignored in one-pass mode.\n", arg.name);
1160 } else if (arg_match(&arg, &minsection_pct, argi)) {
1161 config->cfg.rc_2pass_vbr_minsection_pct = arg_parse_uint(&arg);
1163 if (global->passes < 2)
1164 warn("option %s ignored in one-pass mode.\n", arg.name);
1165 } else if (arg_match(&arg, &maxsection_pct, argi)) {
1166 config->cfg.rc_2pass_vbr_maxsection_pct = arg_parse_uint(&arg);
1168 if (global->passes < 2)
1169 warn("option %s ignored in one-pass mode.\n", arg.name);
1170 } else if (arg_match(&arg, &kf_min_dist, argi)) {
1171 config->cfg.kf_min_dist = arg_parse_uint(&arg);
1172 } else if (arg_match(&arg, &kf_max_dist, argi)) {
1173 config->cfg.kf_max_dist = arg_parse_uint(&arg);
1174 config->have_kf_max_dist = 1;
1175 } else if (arg_match(&arg, &kf_disabled, argi)) {
1176 config->cfg.kf_mode = VPX_KF_DISABLED;
1177 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1178 } else if (arg_match(&arg, &test16bitinternalarg, argi)) {
1179 if (strcmp(global->codec->name, "vp9") == 0) {
1180 test_16bit_internal = 1;
1182 #endif
1183 } else {
1184 int i, match = 0;
1185 for (i = 0; ctrl_args[i]; i++) {
1186 if (arg_match(&arg, ctrl_args[i], argi)) {
1187 int j;
1188 match = 1;
1190 /* Point either to the next free element or the first
1191 * instance of this control.
1193 for (j = 0; j < config->arg_ctrl_cnt; j++)
1194 if (ctrl_args_map != NULL &&
1195 config->arg_ctrls[j][0] == ctrl_args_map[i])
1196 break;
1198 /* Update/insert */
1199 assert(j < (int)ARG_CTRL_CNT_MAX);
1200 if (ctrl_args_map != NULL && j < (int)ARG_CTRL_CNT_MAX) {
1201 config->arg_ctrls[j][0] = ctrl_args_map[i];
1202 config->arg_ctrls[j][1] = arg_parse_enum_or_int(&arg);
1203 if (j == config->arg_ctrl_cnt)
1204 config->arg_ctrl_cnt++;
1208 if (!match)
1209 argj++;
1212 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1213 if (strcmp(global->codec->name, "vp9") == 0) {
1214 config->use_16bit_internal = test_16bit_internal |
1215 (config->cfg.g_profile > 1);
1217 #endif
1218 return eos_mark_found;
1222 #define FOREACH_STREAM(func) \
1223 do { \
1224 struct stream_state *stream; \
1225 for (stream = streams; stream; stream = stream->next) { \
1226 func; \
1228 } while (0)
1231 static void validate_stream_config(const struct stream_state *stream,
1232 const struct VpxEncoderConfig *global) {
1233 const struct stream_state *streami;
1234 (void)global;
1236 if (!stream->config.cfg.g_w || !stream->config.cfg.g_h)
1237 fatal("Stream %d: Specify stream dimensions with --width (-w) "
1238 " and --height (-h)", stream->index);
1240 // Check that the codec bit depth is greater than the input bit depth.
1241 if (stream->config.cfg.g_input_bit_depth >
1242 (unsigned int)stream->config.cfg.g_bit_depth) {
1243 fatal("Stream %d: codec bit depth (%d) less than input bit depth (%d)",
1244 stream->index, (int)stream->config.cfg.g_bit_depth,
1245 stream->config.cfg.g_input_bit_depth);
1248 for (streami = stream; streami; streami = streami->next) {
1249 /* All streams require output files */
1250 if (!streami->config.out_fn)
1251 fatal("Stream %d: Output file is required (specify with -o)",
1252 streami->index);
1254 /* Check for two streams outputting to the same file */
1255 if (streami != stream) {
1256 const char *a = stream->config.out_fn;
1257 const char *b = streami->config.out_fn;
1258 if (!strcmp(a, b) && strcmp(a, "/dev/null") && strcmp(a, ":nul"))
1259 fatal("Stream %d: duplicate output file (from stream %d)",
1260 streami->index, stream->index);
1263 /* Check for two streams sharing a stats file. */
1264 if (streami != stream) {
1265 const char *a = stream->config.stats_fn;
1266 const char *b = streami->config.stats_fn;
1267 if (a && b && !strcmp(a, b))
1268 fatal("Stream %d: duplicate stats file (from stream %d)",
1269 streami->index, stream->index);
1272 #if CONFIG_FP_MB_STATS
1273 /* Check for two streams sharing a mb stats file. */
1274 if (streami != stream) {
1275 const char *a = stream->config.fpmb_stats_fn;
1276 const char *b = streami->config.fpmb_stats_fn;
1277 if (a && b && !strcmp(a, b))
1278 fatal("Stream %d: duplicate mb stats file (from stream %d)",
1279 streami->index, stream->index);
1281 #endif
1286 static void set_stream_dimensions(struct stream_state *stream,
1287 unsigned int w,
1288 unsigned int h) {
1289 if (!stream->config.cfg.g_w) {
1290 if (!stream->config.cfg.g_h)
1291 stream->config.cfg.g_w = w;
1292 else
1293 stream->config.cfg.g_w = w * stream->config.cfg.g_h / h;
1295 if (!stream->config.cfg.g_h) {
1296 stream->config.cfg.g_h = h * stream->config.cfg.g_w / w;
1301 static void set_default_kf_interval(struct stream_state *stream,
1302 struct VpxEncoderConfig *global) {
1303 /* Use a max keyframe interval of 5 seconds, if none was
1304 * specified on the command line.
1306 if (!stream->config.have_kf_max_dist) {
1307 double framerate = (double)global->framerate.num / global->framerate.den;
1308 if (framerate > 0.0)
1309 stream->config.cfg.kf_max_dist = (unsigned int)(5.0 * framerate);
1313 static const char* file_type_to_string(enum VideoFileType t) {
1314 switch (t) {
1315 case FILE_TYPE_RAW: return "RAW";
1316 case FILE_TYPE_Y4M: return "Y4M";
1317 default: return "Other";
1321 static const char* image_format_to_string(vpx_img_fmt_t f) {
1322 switch (f) {
1323 case VPX_IMG_FMT_I420: return "I420";
1324 case VPX_IMG_FMT_I422: return "I422";
1325 case VPX_IMG_FMT_I444: return "I444";
1326 case VPX_IMG_FMT_I440: return "I440";
1327 case VPX_IMG_FMT_YV12: return "YV12";
1328 case VPX_IMG_FMT_I42016: return "I42016";
1329 case VPX_IMG_FMT_I42216: return "I42216";
1330 case VPX_IMG_FMT_I44416: return "I44416";
1331 case VPX_IMG_FMT_I44016: return "I44016";
1332 default: return "Other";
1336 static void show_stream_config(struct stream_state *stream,
1337 struct VpxEncoderConfig *global,
1338 struct VpxInputContext *input) {
1340 #define SHOW(field) \
1341 fprintf(stderr, " %-28s = %d\n", #field, stream->config.cfg.field)
1343 if (stream->index == 0) {
1344 fprintf(stderr, "Codec: %s\n",
1345 vpx_codec_iface_name(global->codec->codec_interface()));
1346 fprintf(stderr, "Source file: %s File Type: %s Format: %s\n",
1347 input->filename,
1348 file_type_to_string(input->file_type),
1349 image_format_to_string(input->fmt));
1351 if (stream->next || stream->index)
1352 fprintf(stderr, "\nStream Index: %d\n", stream->index);
1353 fprintf(stderr, "Destination file: %s\n", stream->config.out_fn);
1354 fprintf(stderr, "Encoder parameters:\n");
1356 SHOW(g_usage);
1357 SHOW(g_threads);
1358 SHOW(g_profile);
1359 SHOW(g_w);
1360 SHOW(g_h);
1361 SHOW(g_bit_depth);
1362 SHOW(g_input_bit_depth);
1363 SHOW(g_timebase.num);
1364 SHOW(g_timebase.den);
1365 SHOW(g_error_resilient);
1366 SHOW(g_pass);
1367 SHOW(g_lag_in_frames);
1368 SHOW(rc_dropframe_thresh);
1369 SHOW(rc_resize_allowed);
1370 SHOW(rc_scaled_width);
1371 SHOW(rc_scaled_height);
1372 SHOW(rc_resize_up_thresh);
1373 SHOW(rc_resize_down_thresh);
1374 SHOW(rc_end_usage);
1375 SHOW(rc_target_bitrate);
1376 SHOW(rc_min_quantizer);
1377 SHOW(rc_max_quantizer);
1378 SHOW(rc_undershoot_pct);
1379 SHOW(rc_overshoot_pct);
1380 SHOW(rc_buf_sz);
1381 SHOW(rc_buf_initial_sz);
1382 SHOW(rc_buf_optimal_sz);
1383 SHOW(rc_2pass_vbr_bias_pct);
1384 SHOW(rc_2pass_vbr_minsection_pct);
1385 SHOW(rc_2pass_vbr_maxsection_pct);
1386 SHOW(kf_mode);
1387 SHOW(kf_min_dist);
1388 SHOW(kf_max_dist);
1392 static void open_output_file(struct stream_state *stream,
1393 struct VpxEncoderConfig *global) {
1394 const char *fn = stream->config.out_fn;
1395 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1397 if (cfg->g_pass == VPX_RC_FIRST_PASS)
1398 return;
1400 stream->file = strcmp(fn, "-") ? fopen(fn, "wb") : set_binary_mode(stdout);
1402 if (!stream->file)
1403 fatal("Failed to open output file");
1405 if (stream->config.write_webm && fseek(stream->file, 0, SEEK_CUR))
1406 fatal("WebM output to pipes not supported.");
1408 #if CONFIG_WEBM_IO
1409 if (stream->config.write_webm) {
1410 stream->ebml.stream = stream->file;
1411 write_webm_file_header(&stream->ebml, cfg,
1412 &global->framerate,
1413 stream->config.stereo_fmt,
1414 global->codec->fourcc);
1416 #endif
1418 if (!stream->config.write_webm) {
1419 ivf_write_file_header(stream->file, cfg, global->codec->fourcc, 0);
1424 static void close_output_file(struct stream_state *stream,
1425 unsigned int fourcc) {
1426 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1428 if (cfg->g_pass == VPX_RC_FIRST_PASS)
1429 return;
1431 #if CONFIG_WEBM_IO
1432 if (stream->config.write_webm) {
1433 write_webm_file_footer(&stream->ebml);
1435 #endif
1437 if (!stream->config.write_webm) {
1438 if (!fseek(stream->file, 0, SEEK_SET))
1439 ivf_write_file_header(stream->file, &stream->config.cfg,
1440 fourcc,
1441 stream->frames_out);
1444 fclose(stream->file);
1448 static void setup_pass(struct stream_state *stream,
1449 struct VpxEncoderConfig *global,
1450 int pass) {
1451 if (stream->config.stats_fn) {
1452 if (!stats_open_file(&stream->stats, stream->config.stats_fn,
1453 pass))
1454 fatal("Failed to open statistics store");
1455 } else {
1456 if (!stats_open_mem(&stream->stats, pass))
1457 fatal("Failed to open statistics store");
1460 #if CONFIG_FP_MB_STATS
1461 if (stream->config.fpmb_stats_fn) {
1462 if (!stats_open_file(&stream->fpmb_stats,
1463 stream->config.fpmb_stats_fn, pass))
1464 fatal("Failed to open mb statistics store");
1465 } else {
1466 if (!stats_open_mem(&stream->fpmb_stats, pass))
1467 fatal("Failed to open mb statistics store");
1469 #endif
1471 stream->config.cfg.g_pass = global->passes == 2
1472 ? pass ? VPX_RC_LAST_PASS : VPX_RC_FIRST_PASS
1473 : VPX_RC_ONE_PASS;
1474 if (pass) {
1475 stream->config.cfg.rc_twopass_stats_in = stats_get(&stream->stats);
1476 #if CONFIG_FP_MB_STATS
1477 stream->config.cfg.rc_firstpass_mb_stats_in =
1478 stats_get(&stream->fpmb_stats);
1479 #endif
1482 stream->cx_time = 0;
1483 stream->nbytes = 0;
1484 stream->frames_out = 0;
1488 static void initialize_encoder(struct stream_state *stream,
1489 struct VpxEncoderConfig *global) {
1490 int i;
1491 int flags = 0;
1493 flags |= global->show_psnr ? VPX_CODEC_USE_PSNR : 0;
1494 flags |= global->out_part ? VPX_CODEC_USE_OUTPUT_PARTITION : 0;
1495 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1496 flags |= stream->config.use_16bit_internal ? VPX_CODEC_USE_HIGHBITDEPTH : 0;
1497 #endif
1499 /* Construct Encoder Context */
1500 vpx_codec_enc_init(&stream->encoder, global->codec->codec_interface(),
1501 &stream->config.cfg, flags);
1502 ctx_exit_on_error(&stream->encoder, "Failed to initialize encoder");
1504 /* Note that we bypass the vpx_codec_control wrapper macro because
1505 * we're being clever to store the control IDs in an array. Real
1506 * applications will want to make use of the enumerations directly
1508 for (i = 0; i < stream->config.arg_ctrl_cnt; i++) {
1509 int ctrl = stream->config.arg_ctrls[i][0];
1510 int value = stream->config.arg_ctrls[i][1];
1511 if (vpx_codec_control_(&stream->encoder, ctrl, value))
1512 fprintf(stderr, "Error: Tried to set control %d = %d\n",
1513 ctrl, value);
1515 ctx_exit_on_error(&stream->encoder, "Failed to control codec");
1518 #if CONFIG_DECODERS
1519 if (global->test_decode != TEST_DECODE_OFF) {
1520 const VpxInterface *decoder = get_vpx_decoder_by_name(global->codec->name);
1521 vpx_codec_dec_init(&stream->decoder, decoder->codec_interface(), NULL, 0);
1523 #endif
1527 static void encode_frame(struct stream_state *stream,
1528 struct VpxEncoderConfig *global,
1529 struct vpx_image *img,
1530 unsigned int frames_in) {
1531 vpx_codec_pts_t frame_start, next_frame_start;
1532 struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1533 struct vpx_usec_timer timer;
1535 frame_start = (cfg->g_timebase.den * (int64_t)(frames_in - 1)
1536 * global->framerate.den)
1537 / cfg->g_timebase.num / global->framerate.num;
1538 next_frame_start = (cfg->g_timebase.den * (int64_t)(frames_in)
1539 * global->framerate.den)
1540 / cfg->g_timebase.num / global->framerate.num;
1542 /* Scale if necessary */
1543 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1544 if (img) {
1545 if ((img->fmt & VPX_IMG_FMT_HIGHBITDEPTH) &&
1546 (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1547 if (img->fmt != VPX_IMG_FMT_I42016) {
1548 fprintf(stderr, "%s can only scale 4:2:0 inputs\n", exec_name);
1549 exit(EXIT_FAILURE);
1551 #if CONFIG_LIBYUV
1552 if (!stream->img) {
1553 stream->img = vpx_img_alloc(NULL, VPX_IMG_FMT_I42016,
1554 cfg->g_w, cfg->g_h, 16);
1556 I420Scale_16((uint16*)img->planes[VPX_PLANE_Y],
1557 img->stride[VPX_PLANE_Y]/2,
1558 (uint16*)img->planes[VPX_PLANE_U],
1559 img->stride[VPX_PLANE_U]/2,
1560 (uint16*)img->planes[VPX_PLANE_V],
1561 img->stride[VPX_PLANE_V]/2,
1562 img->d_w, img->d_h,
1563 (uint16*)stream->img->planes[VPX_PLANE_Y],
1564 stream->img->stride[VPX_PLANE_Y]/2,
1565 (uint16*)stream->img->planes[VPX_PLANE_U],
1566 stream->img->stride[VPX_PLANE_U]/2,
1567 (uint16*)stream->img->planes[VPX_PLANE_V],
1568 stream->img->stride[VPX_PLANE_V]/2,
1569 stream->img->d_w, stream->img->d_h,
1570 kFilterBox);
1571 img = stream->img;
1572 #else
1573 stream->encoder.err = 1;
1574 ctx_exit_on_error(&stream->encoder,
1575 "Stream %d: Failed to encode frame.\n"
1576 "Scaling disabled in this configuration. \n"
1577 "To enable, configure with --enable-libyuv\n",
1578 stream->index);
1579 #endif
1582 #endif
1583 if (img && (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1584 if (img->fmt != VPX_IMG_FMT_I420 && img->fmt != VPX_IMG_FMT_YV12) {
1585 fprintf(stderr, "%s can only scale 4:2:0 8bpp inputs\n", exec_name);
1586 exit(EXIT_FAILURE);
1588 #if CONFIG_LIBYUV
1589 if (!stream->img)
1590 stream->img = vpx_img_alloc(NULL, VPX_IMG_FMT_I420,
1591 cfg->g_w, cfg->g_h, 16);
1592 I420Scale(img->planes[VPX_PLANE_Y], img->stride[VPX_PLANE_Y],
1593 img->planes[VPX_PLANE_U], img->stride[VPX_PLANE_U],
1594 img->planes[VPX_PLANE_V], img->stride[VPX_PLANE_V],
1595 img->d_w, img->d_h,
1596 stream->img->planes[VPX_PLANE_Y],
1597 stream->img->stride[VPX_PLANE_Y],
1598 stream->img->planes[VPX_PLANE_U],
1599 stream->img->stride[VPX_PLANE_U],
1600 stream->img->planes[VPX_PLANE_V],
1601 stream->img->stride[VPX_PLANE_V],
1602 stream->img->d_w, stream->img->d_h,
1603 kFilterBox);
1604 img = stream->img;
1605 #else
1606 stream->encoder.err = 1;
1607 ctx_exit_on_error(&stream->encoder,
1608 "Stream %d: Failed to encode frame.\n"
1609 "Scaling disabled in this configuration. \n"
1610 "To enable, configure with --enable-libyuv\n",
1611 stream->index);
1612 #endif
1615 vpx_usec_timer_start(&timer);
1616 vpx_codec_encode(&stream->encoder, img, frame_start,
1617 (unsigned long)(next_frame_start - frame_start),
1618 0, global->deadline);
1619 vpx_usec_timer_mark(&timer);
1620 stream->cx_time += vpx_usec_timer_elapsed(&timer);
1621 ctx_exit_on_error(&stream->encoder, "Stream %d: Failed to encode frame",
1622 stream->index);
1626 static void update_quantizer_histogram(struct stream_state *stream) {
1627 if (stream->config.cfg.g_pass != VPX_RC_FIRST_PASS) {
1628 int q;
1630 vpx_codec_control(&stream->encoder, VP8E_GET_LAST_QUANTIZER_64, &q);
1631 ctx_exit_on_error(&stream->encoder, "Failed to read quantizer");
1632 stream->counts[q]++;
1637 static void get_cx_data(struct stream_state *stream,
1638 struct VpxEncoderConfig *global,
1639 int *got_data) {
1640 const vpx_codec_cx_pkt_t *pkt;
1641 const struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1642 vpx_codec_iter_t iter = NULL;
1644 *got_data = 0;
1645 while ((pkt = vpx_codec_get_cx_data(&stream->encoder, &iter))) {
1646 static size_t fsize = 0;
1647 static int64_t ivf_header_pos = 0;
1649 switch (pkt->kind) {
1650 case VPX_CODEC_CX_FRAME_PKT:
1651 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1652 stream->frames_out++;
1654 if (!global->quiet)
1655 fprintf(stderr, " %6luF", (unsigned long)pkt->data.frame.sz);
1657 update_rate_histogram(stream->rate_hist, cfg, pkt);
1658 #if CONFIG_WEBM_IO
1659 if (stream->config.write_webm) {
1660 write_webm_block(&stream->ebml, cfg, pkt);
1662 #endif
1663 if (!stream->config.write_webm) {
1664 if (pkt->data.frame.partition_id <= 0) {
1665 ivf_header_pos = ftello(stream->file);
1666 fsize = pkt->data.frame.sz;
1668 ivf_write_frame_header(stream->file, pkt->data.frame.pts, fsize);
1669 } else {
1670 fsize += pkt->data.frame.sz;
1672 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1673 const int64_t currpos = ftello(stream->file);
1674 fseeko(stream->file, ivf_header_pos, SEEK_SET);
1675 ivf_write_frame_size(stream->file, fsize);
1676 fseeko(stream->file, currpos, SEEK_SET);
1680 (void) fwrite(pkt->data.frame.buf, 1, pkt->data.frame.sz,
1681 stream->file);
1683 stream->nbytes += pkt->data.raw.sz;
1685 *got_data = 1;
1686 #if CONFIG_DECODERS
1687 if (global->test_decode != TEST_DECODE_OFF && !stream->mismatch_seen) {
1688 vpx_codec_decode(&stream->decoder, pkt->data.frame.buf,
1689 (unsigned int)pkt->data.frame.sz, NULL, 0);
1690 if (stream->decoder.err) {
1691 warn_or_exit_on_error(&stream->decoder,
1692 global->test_decode == TEST_DECODE_FATAL,
1693 "Failed to decode frame %d in stream %d",
1694 stream->frames_out + 1, stream->index);
1695 stream->mismatch_seen = stream->frames_out + 1;
1698 #endif
1699 break;
1700 case VPX_CODEC_STATS_PKT:
1701 stream->frames_out++;
1702 stats_write(&stream->stats,
1703 pkt->data.twopass_stats.buf,
1704 pkt->data.twopass_stats.sz);
1705 stream->nbytes += pkt->data.raw.sz;
1706 break;
1707 #if CONFIG_FP_MB_STATS
1708 case VPX_CODEC_FPMB_STATS_PKT:
1709 stats_write(&stream->fpmb_stats,
1710 pkt->data.firstpass_mb_stats.buf,
1711 pkt->data.firstpass_mb_stats.sz);
1712 stream->nbytes += pkt->data.raw.sz;
1713 break;
1714 #endif
1715 case VPX_CODEC_PSNR_PKT:
1717 if (global->show_psnr) {
1718 int i;
1720 stream->psnr_sse_total += pkt->data.psnr.sse[0];
1721 stream->psnr_samples_total += pkt->data.psnr.samples[0];
1722 for (i = 0; i < 4; i++) {
1723 if (!global->quiet)
1724 fprintf(stderr, "%.3f ", pkt->data.psnr.psnr[i]);
1725 stream->psnr_totals[i] += pkt->data.psnr.psnr[i];
1727 stream->psnr_count++;
1730 break;
1731 default:
1732 break;
1738 static void show_psnr(struct stream_state *stream, double peak) {
1739 int i;
1740 double ovpsnr;
1742 if (!stream->psnr_count)
1743 return;
1745 fprintf(stderr, "Stream %d PSNR (Overall/Avg/Y/U/V)", stream->index);
1746 ovpsnr = sse_to_psnr((double)stream->psnr_samples_total, peak,
1747 (double)stream->psnr_sse_total);
1748 fprintf(stderr, " %.3f", ovpsnr);
1750 for (i = 0; i < 4; i++) {
1751 fprintf(stderr, " %.3f", stream->psnr_totals[i] / stream->psnr_count);
1753 fprintf(stderr, "\n");
1757 static float usec_to_fps(uint64_t usec, unsigned int frames) {
1758 return (float)(usec > 0 ? frames * 1000000.0 / (float)usec : 0);
1761 static void test_decode(struct stream_state *stream,
1762 enum TestDecodeFatality fatal,
1763 const VpxInterface *codec) {
1764 vpx_image_t enc_img, dec_img;
1766 if (stream->mismatch_seen)
1767 return;
1769 /* Get the internal reference frame */
1770 if (strcmp(codec->name, "vp8") == 0) {
1771 struct vpx_ref_frame ref_enc, ref_dec;
1772 int width, height;
1774 width = (stream->config.cfg.g_w + 15) & ~15;
1775 height = (stream->config.cfg.g_h + 15) & ~15;
1776 vpx_img_alloc(&ref_enc.img, VPX_IMG_FMT_I420, width, height, 1);
1777 enc_img = ref_enc.img;
1778 vpx_img_alloc(&ref_dec.img, VPX_IMG_FMT_I420, width, height, 1);
1779 dec_img = ref_dec.img;
1781 ref_enc.frame_type = VP8_LAST_FRAME;
1782 ref_dec.frame_type = VP8_LAST_FRAME;
1783 vpx_codec_control(&stream->encoder, VP8_COPY_REFERENCE, &ref_enc);
1784 vpx_codec_control(&stream->decoder, VP8_COPY_REFERENCE, &ref_dec);
1785 } else {
1786 struct vp9_ref_frame ref_enc, ref_dec;
1788 ref_enc.idx = 0;
1789 ref_dec.idx = 0;
1790 vpx_codec_control(&stream->encoder, VP9_GET_REFERENCE, &ref_enc);
1791 enc_img = ref_enc.img;
1792 vpx_codec_control(&stream->decoder, VP9_GET_REFERENCE, &ref_dec);
1793 dec_img = ref_dec.img;
1794 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1795 if ((enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) !=
1796 (dec_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH)) {
1797 if (enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1798 vpx_img_alloc(&enc_img, enc_img.fmt - VPX_IMG_FMT_HIGHBITDEPTH,
1799 enc_img.d_w, enc_img.d_h, 16);
1800 vpx_img_truncate_16_to_8(&enc_img, &ref_enc.img);
1802 if (dec_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1803 vpx_img_alloc(&dec_img, dec_img.fmt - VPX_IMG_FMT_HIGHBITDEPTH,
1804 dec_img.d_w, dec_img.d_h, 16);
1805 vpx_img_truncate_16_to_8(&dec_img, &ref_dec.img);
1808 #endif
1810 ctx_exit_on_error(&stream->encoder, "Failed to get encoder reference frame");
1811 ctx_exit_on_error(&stream->decoder, "Failed to get decoder reference frame");
1813 if (!compare_img(&enc_img, &dec_img)) {
1814 int y[4], u[4], v[4];
1815 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1816 if (enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1817 find_mismatch_high(&enc_img, &dec_img, y, u, v);
1818 } else {
1819 find_mismatch(&enc_img, &dec_img, y, u, v);
1821 #else
1822 find_mismatch(&enc_img, &dec_img, y, u, v);
1823 #endif
1824 stream->decoder.err = 1;
1825 warn_or_exit_on_error(&stream->decoder, fatal == TEST_DECODE_FATAL,
1826 "Stream %d: Encode/decode mismatch on frame %d at"
1827 " Y[%d, %d] {%d/%d},"
1828 " U[%d, %d] {%d/%d},"
1829 " V[%d, %d] {%d/%d}",
1830 stream->index, stream->frames_out,
1831 y[0], y[1], y[2], y[3],
1832 u[0], u[1], u[2], u[3],
1833 v[0], v[1], v[2], v[3]);
1834 stream->mismatch_seen = stream->frames_out;
1837 vpx_img_free(&enc_img);
1838 vpx_img_free(&dec_img);
1842 static void print_time(const char *label, int64_t etl) {
1843 int64_t hours;
1844 int64_t mins;
1845 int64_t secs;
1847 if (etl >= 0) {
1848 hours = etl / 3600;
1849 etl -= hours * 3600;
1850 mins = etl / 60;
1851 etl -= mins * 60;
1852 secs = etl;
1854 fprintf(stderr, "[%3s %2"PRId64":%02"PRId64":%02"PRId64"] ",
1855 label, hours, mins, secs);
1856 } else {
1857 fprintf(stderr, "[%3s unknown] ", label);
1862 int main(int argc, const char **argv_) {
1863 int pass;
1864 vpx_image_t raw;
1865 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1866 vpx_image_t raw_shift;
1867 int allocated_raw_shift = 0;
1868 int use_16bit_internal = 0;
1869 int input_shift = 0;
1870 #endif
1871 int frame_avail, got_data;
1873 struct VpxInputContext input;
1874 struct VpxEncoderConfig global;
1875 struct stream_state *streams = NULL;
1876 char **argv, **argi;
1877 uint64_t cx_time = 0;
1878 int stream_cnt = 0;
1879 int res = 0;
1881 memset(&input, 0, sizeof(input));
1882 exec_name = argv_[0];
1884 if (argc < 3)
1885 usage_exit();
1887 /* Setup default input stream settings */
1888 input.framerate.numerator = 30;
1889 input.framerate.denominator = 1;
1890 input.only_i420 = 1;
1891 input.bit_depth = 0;
1893 /* First parse the global configuration values, because we want to apply
1894 * other parameters on top of the default configuration provided by the
1895 * codec.
1897 argv = argv_dup(argc - 1, argv_ + 1);
1898 parse_global_config(&global, argv);
1900 switch (global.color_type) {
1901 case I420:
1902 input.fmt = VPX_IMG_FMT_I420;
1903 break;
1904 case I422:
1905 input.fmt = VPX_IMG_FMT_I422;
1906 break;
1907 case I444:
1908 input.fmt = VPX_IMG_FMT_I444;
1909 break;
1910 case I440:
1911 input.fmt = VPX_IMG_FMT_I440;
1912 break;
1913 case YV12:
1914 input.fmt = VPX_IMG_FMT_YV12;
1915 break;
1919 /* Now parse each stream's parameters. Using a local scope here
1920 * due to the use of 'stream' as loop variable in FOREACH_STREAM
1921 * loops
1923 struct stream_state *stream = NULL;
1925 do {
1926 stream = new_stream(&global, stream);
1927 stream_cnt++;
1928 if (!streams)
1929 streams = stream;
1930 } while (parse_stream_params(&global, stream, argv));
1933 /* Check for unrecognized options */
1934 for (argi = argv; *argi; argi++)
1935 if (argi[0][0] == '-' && argi[0][1])
1936 die("Error: Unrecognized option %s\n", *argi);
1938 FOREACH_STREAM(check_encoder_config(global.disable_warning_prompt,
1939 &global, &stream->config.cfg););
1941 /* Handle non-option arguments */
1942 input.filename = argv[0];
1944 if (!input.filename)
1945 usage_exit();
1947 /* Decide if other chroma subsamplings than 4:2:0 are supported */
1948 if (global.codec->fourcc == VP9_FOURCC)
1949 input.only_i420 = 0;
1951 for (pass = global.pass ? global.pass - 1 : 0; pass < global.passes; pass++) {
1952 int frames_in = 0, seen_frames = 0;
1953 int64_t estimated_time_left = -1;
1954 int64_t average_rate = -1;
1955 int64_t lagged_count = 0;
1957 open_input_file(&input);
1959 /* If the input file doesn't specify its w/h (raw files), try to get
1960 * the data from the first stream's configuration.
1962 if (!input.width || !input.height) {
1963 FOREACH_STREAM({
1964 if (stream->config.cfg.g_w && stream->config.cfg.g_h) {
1965 input.width = stream->config.cfg.g_w;
1966 input.height = stream->config.cfg.g_h;
1967 break;
1972 /* Update stream configurations from the input file's parameters */
1973 if (!input.width || !input.height)
1974 fatal("Specify stream dimensions with --width (-w) "
1975 " and --height (-h)");
1977 /* If input file does not specify bit-depth but input-bit-depth parameter
1978 * exists, assume that to be the input bit-depth. However, if the
1979 * input-bit-depth paramter does not exist, assume the input bit-depth
1980 * to be the same as the codec bit-depth.
1982 if (!input.bit_depth) {
1983 FOREACH_STREAM({
1984 if (stream->config.cfg.g_input_bit_depth)
1985 input.bit_depth = stream->config.cfg.g_input_bit_depth;
1986 else
1987 input.bit_depth = stream->config.cfg.g_input_bit_depth =
1988 (int)stream->config.cfg.g_bit_depth;
1990 if (input.bit_depth > 8) input.fmt |= VPX_IMG_FMT_HIGHBITDEPTH;
1991 } else {
1992 FOREACH_STREAM({
1993 stream->config.cfg.g_input_bit_depth = input.bit_depth;
1997 FOREACH_STREAM(set_stream_dimensions(stream, input.width, input.height));
1998 FOREACH_STREAM(validate_stream_config(stream, &global));
2000 /* Ensure that --passes and --pass are consistent. If --pass is set and
2001 * --passes=2, ensure --fpf was set.
2003 if (global.pass && global.passes == 2)
2004 FOREACH_STREAM( {
2005 if (!stream->config.stats_fn)
2006 die("Stream %d: Must specify --fpf when --pass=%d"
2007 " and --passes=2\n", stream->index, global.pass);
2010 #if !CONFIG_WEBM_IO
2011 FOREACH_STREAM({
2012 stream->config.write_webm = 0;
2013 warn("vpxenc was compiled without WebM container support."
2014 "Producing IVF output");
2016 #endif
2018 /* Use the frame rate from the file only if none was specified
2019 * on the command-line.
2021 if (!global.have_framerate) {
2022 global.framerate.num = input.framerate.numerator;
2023 global.framerate.den = input.framerate.denominator;
2026 FOREACH_STREAM(set_default_kf_interval(stream, &global));
2028 /* Show configuration */
2029 if (global.verbose && pass == 0)
2030 FOREACH_STREAM(show_stream_config(stream, &global, &input));
2032 if (pass == (global.pass ? global.pass - 1 : 0)) {
2033 if (input.file_type == FILE_TYPE_Y4M)
2034 /*The Y4M reader does its own allocation.
2035 Just initialize this here to avoid problems if we never read any
2036 frames.*/
2037 memset(&raw, 0, sizeof(raw));
2038 else
2039 vpx_img_alloc(&raw, input.fmt, input.width, input.height, 32);
2041 FOREACH_STREAM(stream->rate_hist =
2042 init_rate_histogram(&stream->config.cfg,
2043 &global.framerate));
2046 FOREACH_STREAM(setup_pass(stream, &global, pass));
2047 FOREACH_STREAM(open_output_file(stream, &global));
2048 FOREACH_STREAM(initialize_encoder(stream, &global));
2050 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
2051 if (strcmp(global.codec->name, "vp9") == 0) {
2052 // Check to see if at least one stream uses 16 bit internal.
2053 // Currently assume that the bit_depths for all streams using
2054 // highbitdepth are the same.
2055 FOREACH_STREAM({
2056 if (stream->config.use_16bit_internal) {
2057 use_16bit_internal = 1;
2059 if (stream->config.cfg.g_profile == 0) {
2060 input_shift = 0;
2061 } else {
2062 input_shift = (int)stream->config.cfg.g_bit_depth -
2063 stream->config.cfg.g_input_bit_depth;
2067 #endif
2069 frame_avail = 1;
2070 got_data = 0;
2072 while (frame_avail || got_data) {
2073 struct vpx_usec_timer timer;
2075 if (!global.limit || frames_in < global.limit) {
2076 frame_avail = read_frame(&input, &raw);
2078 if (frame_avail)
2079 frames_in++;
2080 seen_frames = frames_in > global.skip_frames ?
2081 frames_in - global.skip_frames : 0;
2083 if (!global.quiet) {
2084 float fps = usec_to_fps(cx_time, seen_frames);
2085 fprintf(stderr, "\rPass %d/%d ", pass + 1, global.passes);
2087 if (stream_cnt == 1)
2088 fprintf(stderr,
2089 "frame %4d/%-4d %7"PRId64"B ",
2090 frames_in, streams->frames_out, (int64_t)streams->nbytes);
2091 else
2092 fprintf(stderr, "frame %4d ", frames_in);
2094 fprintf(stderr, "%7"PRId64" %s %.2f %s ",
2095 cx_time > 9999999 ? cx_time / 1000 : cx_time,
2096 cx_time > 9999999 ? "ms" : "us",
2097 fps >= 1.0 ? fps : fps * 60,
2098 fps >= 1.0 ? "fps" : "fpm");
2099 print_time("ETA", estimated_time_left);
2102 } else
2103 frame_avail = 0;
2105 if (frames_in > global.skip_frames) {
2106 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
2107 vpx_image_t *frame_to_encode;
2108 if (input_shift || (use_16bit_internal && input.bit_depth == 8)) {
2109 assert(use_16bit_internal);
2110 // Input bit depth and stream bit depth do not match, so up
2111 // shift frame to stream bit depth
2112 if (!allocated_raw_shift) {
2113 vpx_img_alloc(&raw_shift, raw.fmt | VPX_IMG_FMT_HIGHBITDEPTH,
2114 input.width, input.height, 32);
2115 allocated_raw_shift = 1;
2117 vpx_img_upshift(&raw_shift, &raw, input_shift);
2118 frame_to_encode = &raw_shift;
2119 } else {
2120 frame_to_encode = &raw;
2122 vpx_usec_timer_start(&timer);
2123 if (use_16bit_internal) {
2124 assert(frame_to_encode->fmt & VPX_IMG_FMT_HIGHBITDEPTH);
2125 FOREACH_STREAM({
2126 if (stream->config.use_16bit_internal)
2127 encode_frame(stream, &global,
2128 frame_avail ? frame_to_encode : NULL,
2129 frames_in);
2130 else
2131 assert(0);
2133 } else {
2134 assert((frame_to_encode->fmt & VPX_IMG_FMT_HIGHBITDEPTH) == 0);
2135 FOREACH_STREAM(encode_frame(stream, &global,
2136 frame_avail ? frame_to_encode : NULL,
2137 frames_in));
2139 #else
2140 vpx_usec_timer_start(&timer);
2141 FOREACH_STREAM(encode_frame(stream, &global,
2142 frame_avail ? &raw : NULL,
2143 frames_in));
2144 #endif
2145 vpx_usec_timer_mark(&timer);
2146 cx_time += vpx_usec_timer_elapsed(&timer);
2148 FOREACH_STREAM(update_quantizer_histogram(stream));
2150 got_data = 0;
2151 FOREACH_STREAM(get_cx_data(stream, &global, &got_data));
2153 if (!got_data && input.length && streams != NULL &&
2154 !streams->frames_out) {
2155 lagged_count = global.limit ? seen_frames : ftello(input.file);
2156 } else if (input.length) {
2157 int64_t remaining;
2158 int64_t rate;
2160 if (global.limit) {
2161 const int64_t frame_in_lagged = (seen_frames - lagged_count) * 1000;
2163 rate = cx_time ? frame_in_lagged * (int64_t)1000000 / cx_time : 0;
2164 remaining = 1000 * (global.limit - global.skip_frames
2165 - seen_frames + lagged_count);
2166 } else {
2167 const int64_t input_pos = ftello(input.file);
2168 const int64_t input_pos_lagged = input_pos - lagged_count;
2169 const int64_t limit = input.length;
2171 rate = cx_time ? input_pos_lagged * (int64_t)1000000 / cx_time : 0;
2172 remaining = limit - input_pos + lagged_count;
2175 average_rate = (average_rate <= 0)
2176 ? rate
2177 : (average_rate * 7 + rate) / 8;
2178 estimated_time_left = average_rate ? remaining / average_rate : -1;
2181 if (got_data && global.test_decode != TEST_DECODE_OFF)
2182 FOREACH_STREAM(test_decode(stream, global.test_decode, global.codec));
2185 fflush(stdout);
2186 if (!global.quiet)
2187 fprintf(stderr, "\033[K");
2190 if (stream_cnt > 1)
2191 fprintf(stderr, "\n");
2193 if (!global.quiet) {
2194 FOREACH_STREAM(fprintf(stderr,
2195 "\rPass %d/%d frame %4d/%-4d %7"PRId64"B %7"PRId64"b/f %7"PRId64"b/s"
2196 " %7"PRId64" %s (%.2f fps)\033[K\n",
2197 pass + 1,
2198 global.passes, frames_in, stream->frames_out, (int64_t)stream->nbytes,
2199 seen_frames ? (int64_t)(stream->nbytes * 8 / seen_frames) : 0,
2200 seen_frames ? (int64_t)stream->nbytes * 8 *
2201 (int64_t)global.framerate.num / global.framerate.den /
2202 seen_frames : 0,
2203 stream->cx_time > 9999999 ? stream->cx_time / 1000 : stream->cx_time,
2204 stream->cx_time > 9999999 ? "ms" : "us",
2205 usec_to_fps(stream->cx_time, seen_frames)));
2208 if (global.show_psnr) {
2209 if (global.codec->fourcc == VP9_FOURCC) {
2210 FOREACH_STREAM(
2211 show_psnr(stream, (1 << stream->config.cfg.g_input_bit_depth) - 1));
2212 } else {
2213 FOREACH_STREAM(show_psnr(stream, 255.0));
2217 FOREACH_STREAM(vpx_codec_destroy(&stream->encoder));
2219 if (global.test_decode != TEST_DECODE_OFF) {
2220 FOREACH_STREAM(vpx_codec_destroy(&stream->decoder));
2223 close_input_file(&input);
2225 if (global.test_decode == TEST_DECODE_FATAL) {
2226 FOREACH_STREAM(res |= stream->mismatch_seen);
2228 FOREACH_STREAM(close_output_file(stream, global.codec->fourcc));
2230 FOREACH_STREAM(stats_close(&stream->stats, global.passes - 1));
2232 #if CONFIG_FP_MB_STATS
2233 FOREACH_STREAM(stats_close(&stream->fpmb_stats, global.passes - 1));
2234 #endif
2236 if (global.pass)
2237 break;
2240 if (global.show_q_hist_buckets)
2241 FOREACH_STREAM(show_q_histogram(stream->counts,
2242 global.show_q_hist_buckets));
2244 if (global.show_rate_hist_buckets)
2245 FOREACH_STREAM(show_rate_histogram(stream->rate_hist,
2246 &stream->config.cfg,
2247 global.show_rate_hist_buckets));
2248 FOREACH_STREAM(destroy_rate_histogram(stream->rate_hist));
2250 #if CONFIG_INTERNAL_STATS
2251 /* TODO(jkoleszar): This doesn't belong in this executable. Do it for now,
2252 * to match some existing utilities.
2254 if (!(global.pass == 1 && global.passes == 2))
2255 FOREACH_STREAM({
2256 FILE *f = fopen("opsnr.stt", "a");
2257 if (stream->mismatch_seen) {
2258 fprintf(f, "First mismatch occurred in frame %d\n",
2259 stream->mismatch_seen);
2260 } else {
2261 fprintf(f, "No mismatch detected in recon buffers\n");
2263 fclose(f);
2265 #endif
2267 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
2268 if (allocated_raw_shift)
2269 vpx_img_free(&raw_shift);
2270 #endif
2271 vpx_img_free(&raw);
2272 free(argv);
2273 free(streams);
2274 return res ? EXIT_FAILURE : EXIT_SUCCESS;