Fix generic-gnu target build
[aom.git] / vpxenc.c
blob8e8ed23448f6c6067b2b03529215850668d827ed
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 #include "vpx/vpx_encoder.h"
23 #if CONFIG_DECODERS
24 #include "vpx/vpx_decoder.h"
25 #endif
27 #include "third_party/libyuv/include/libyuv/scale.h"
28 #include "./args.h"
29 #include "./ivfenc.h"
30 #include "./tools_common.h"
32 #if CONFIG_VP8_ENCODER || CONFIG_VP9_ENCODER
33 #include "vpx/vp8cx.h"
34 #endif
35 #if CONFIG_VP8_DECODER || CONFIG_VP9_DECODER
36 #include "vpx/vp8dx.h"
37 #endif
39 #include "vpx/vpx_integer.h"
40 #include "vpx_ports/mem_ops.h"
41 #include "vpx_ports/vpx_timer.h"
42 #include "./rate_hist.h"
43 #include "./vpxstats.h"
44 #include "./warnings.h"
45 #include "./webmenc.h"
46 #include "./y4minput.h"
48 /* Swallow warnings about unused results of fread/fwrite */
49 static size_t wrap_fread(void *ptr, size_t size, size_t nmemb,
50 FILE *stream) {
51 return fread(ptr, size, nmemb, stream);
53 #define fread wrap_fread
55 static size_t wrap_fwrite(const void *ptr, size_t size, size_t nmemb,
56 FILE *stream) {
57 return fwrite(ptr, size, nmemb, stream);
59 #define fwrite wrap_fwrite
62 static const char *exec_name;
64 static void warn_or_exit_on_errorv(vpx_codec_ctx_t *ctx, int fatal,
65 const char *s, va_list ap) {
66 if (ctx->err) {
67 const char *detail = vpx_codec_error_detail(ctx);
69 vfprintf(stderr, s, ap);
70 fprintf(stderr, ": %s\n", vpx_codec_error(ctx));
72 if (detail)
73 fprintf(stderr, " %s\n", detail);
75 if (fatal)
76 exit(EXIT_FAILURE);
80 static void ctx_exit_on_error(vpx_codec_ctx_t *ctx, const char *s, ...) {
81 va_list ap;
83 va_start(ap, s);
84 warn_or_exit_on_errorv(ctx, 1, s, ap);
85 va_end(ap);
88 static void warn_or_exit_on_error(vpx_codec_ctx_t *ctx, int fatal,
89 const char *s, ...) {
90 va_list ap;
92 va_start(ap, s);
93 warn_or_exit_on_errorv(ctx, fatal, s, ap);
94 va_end(ap);
97 int read_frame(struct VpxInputContext *input_ctx, vpx_image_t *img) {
98 FILE *f = input_ctx->file;
99 y4m_input *y4m = &input_ctx->y4m;
100 int shortread = 0;
102 if (input_ctx->file_type == FILE_TYPE_Y4M) {
103 if (y4m_input_fetch_frame(y4m, f, img) < 1)
104 return 0;
105 } else {
106 shortread = read_yuv_frame(input_ctx, img);
109 return !shortread;
112 int file_is_y4m(const char detect[4]) {
113 if (memcmp(detect, "YUV4", 4) == 0) {
114 return 1;
116 return 0;
119 int fourcc_is_ivf(const char detect[4]) {
120 if (memcmp(detect, "DKIF", 4) == 0) {
121 return 1;
123 return 0;
126 static const arg_def_t debugmode = ARG_DEF("D", "debug", 0,
127 "Debug mode (makes output deterministic)");
128 static const arg_def_t outputfile = ARG_DEF("o", "output", 1,
129 "Output filename");
130 static const arg_def_t use_yv12 = ARG_DEF(NULL, "yv12", 0,
131 "Input file is YV12 ");
132 static const arg_def_t use_i420 = ARG_DEF(NULL, "i420", 0,
133 "Input file is I420 (default)");
134 static const arg_def_t codecarg = ARG_DEF(NULL, "codec", 1,
135 "Codec to use");
136 static const arg_def_t passes = ARG_DEF("p", "passes", 1,
137 "Number of passes (1/2)");
138 static const arg_def_t pass_arg = ARG_DEF(NULL, "pass", 1,
139 "Pass to execute (1/2)");
140 static const arg_def_t fpf_name = ARG_DEF(NULL, "fpf", 1,
141 "First pass statistics file name");
142 static const arg_def_t limit = ARG_DEF(NULL, "limit", 1,
143 "Stop encoding after n input frames");
144 static const arg_def_t skip = ARG_DEF(NULL, "skip", 1,
145 "Skip the first n input frames");
146 static const arg_def_t deadline = ARG_DEF("d", "deadline", 1,
147 "Deadline per frame (usec)");
148 static const arg_def_t best_dl = ARG_DEF(NULL, "best", 0,
149 "Use Best Quality Deadline");
150 static const arg_def_t good_dl = ARG_DEF(NULL, "good", 0,
151 "Use Good Quality Deadline");
152 static const arg_def_t rt_dl = ARG_DEF(NULL, "rt", 0,
153 "Use Realtime Quality Deadline");
154 static const arg_def_t quietarg = ARG_DEF("q", "quiet", 0,
155 "Do not print encode progress");
156 static const arg_def_t verbosearg = ARG_DEF("v", "verbose", 0,
157 "Show encoder parameters");
158 static const arg_def_t psnrarg = ARG_DEF(NULL, "psnr", 0,
159 "Show PSNR in status line");
161 static const struct arg_enum_list test_decode_enum[] = {
162 {"off", TEST_DECODE_OFF},
163 {"fatal", TEST_DECODE_FATAL},
164 {"warn", TEST_DECODE_WARN},
165 {NULL, 0}
167 static const arg_def_t recontest = ARG_DEF_ENUM(NULL, "test-decode", 1,
168 "Test encode/decode mismatch",
169 test_decode_enum);
170 static const arg_def_t framerate = ARG_DEF(NULL, "fps", 1,
171 "Stream frame rate (rate/scale)");
172 static const arg_def_t use_ivf = ARG_DEF(NULL, "ivf", 0,
173 "Output IVF (default is WebM if WebM IO is enabled)");
174 static const arg_def_t out_part = ARG_DEF("P", "output-partitions", 0,
175 "Makes encoder output partitions. Requires IVF output!");
176 static const arg_def_t q_hist_n = ARG_DEF(NULL, "q-hist", 1,
177 "Show quantizer histogram (n-buckets)");
178 static const arg_def_t rate_hist_n = ARG_DEF(NULL, "rate-hist", 1,
179 "Show rate histogram (n-buckets)");
180 static const arg_def_t disable_warnings =
181 ARG_DEF(NULL, "disable-warnings", 0,
182 "Disable warnings about potentially incorrect encode settings.");
183 static const arg_def_t disable_warning_prompt =
184 ARG_DEF("y", "disable-warning-prompt", 0,
185 "Display warnings, but do not prompt user to continue.");
186 static const arg_def_t experimental_bitstream =
187 ARG_DEF(NULL, "experimental-bitstream", 0,
188 "Allow experimental bitstream features.");
191 static const arg_def_t *main_args[] = {
192 &debugmode,
193 &outputfile, &codecarg, &passes, &pass_arg, &fpf_name, &limit, &skip,
194 &deadline, &best_dl, &good_dl, &rt_dl,
195 &quietarg, &verbosearg, &psnrarg, &use_ivf, &out_part, &q_hist_n,
196 &rate_hist_n, &disable_warnings, &disable_warning_prompt,
197 NULL
200 static const arg_def_t usage = ARG_DEF("u", "usage", 1,
201 "Usage profile number to use");
202 static const arg_def_t threads = ARG_DEF("t", "threads", 1,
203 "Max number of threads to use");
204 static const arg_def_t profile = ARG_DEF(NULL, "profile", 1,
205 "Bitstream profile number to use");
206 static const arg_def_t width = ARG_DEF("w", "width", 1,
207 "Frame width");
208 static const arg_def_t height = ARG_DEF("h", "height", 1,
209 "Frame height");
210 static const struct arg_enum_list stereo_mode_enum[] = {
211 {"mono", STEREO_FORMAT_MONO},
212 {"left-right", STEREO_FORMAT_LEFT_RIGHT},
213 {"bottom-top", STEREO_FORMAT_BOTTOM_TOP},
214 {"top-bottom", STEREO_FORMAT_TOP_BOTTOM},
215 {"right-left", STEREO_FORMAT_RIGHT_LEFT},
216 {NULL, 0}
218 static const arg_def_t stereo_mode = ARG_DEF_ENUM(NULL, "stereo-mode", 1,
219 "Stereo 3D video format", stereo_mode_enum);
220 static const arg_def_t timebase = ARG_DEF(NULL, "timebase", 1,
221 "Output timestamp precision (fractional seconds)");
222 static const arg_def_t error_resilient = ARG_DEF(NULL, "error-resilient", 1,
223 "Enable error resiliency features");
224 static const arg_def_t lag_in_frames = ARG_DEF(NULL, "lag-in-frames", 1,
225 "Max number of frames to lag");
227 static const arg_def_t *global_args[] = {
228 &use_yv12, &use_i420, &usage, &threads, &profile,
229 &width, &height, &stereo_mode, &timebase, &framerate,
230 &error_resilient,
231 &lag_in_frames, NULL
234 static const arg_def_t dropframe_thresh = ARG_DEF(NULL, "drop-frame", 1,
235 "Temporal resampling threshold (buf %)");
236 static const arg_def_t resize_allowed = ARG_DEF(NULL, "resize-allowed", 1,
237 "Spatial resampling enabled (bool)");
238 static const arg_def_t resize_width = ARG_DEF(NULL, "resize-width", 1,
239 "Width of encoded frame");
240 static const arg_def_t resize_height = ARG_DEF(NULL, "resize-height", 1,
241 "Height of encoded frame");
242 static const arg_def_t resize_up_thresh = ARG_DEF(NULL, "resize-up", 1,
243 "Upscale threshold (buf %)");
244 static const arg_def_t resize_down_thresh = ARG_DEF(NULL, "resize-down", 1,
245 "Downscale threshold (buf %)");
246 static const struct arg_enum_list end_usage_enum[] = {
247 {"vbr", VPX_VBR},
248 {"cbr", VPX_CBR},
249 {"cq", VPX_CQ},
250 {"q", VPX_Q},
251 {NULL, 0}
253 static const arg_def_t end_usage = ARG_DEF_ENUM(NULL, "end-usage", 1,
254 "Rate control mode", end_usage_enum);
255 static const arg_def_t target_bitrate = ARG_DEF(NULL, "target-bitrate", 1,
256 "Bitrate (kbps)");
257 static const arg_def_t min_quantizer = ARG_DEF(NULL, "min-q", 1,
258 "Minimum (best) quantizer");
259 static const arg_def_t max_quantizer = ARG_DEF(NULL, "max-q", 1,
260 "Maximum (worst) quantizer");
261 static const arg_def_t undershoot_pct = ARG_DEF(NULL, "undershoot-pct", 1,
262 "Datarate undershoot (min) target (%)");
263 static const arg_def_t overshoot_pct = ARG_DEF(NULL, "overshoot-pct", 1,
264 "Datarate overshoot (max) target (%)");
265 static const arg_def_t buf_sz = ARG_DEF(NULL, "buf-sz", 1,
266 "Client buffer size (ms)");
267 static const arg_def_t buf_initial_sz = ARG_DEF(NULL, "buf-initial-sz", 1,
268 "Client initial buffer size (ms)");
269 static const arg_def_t buf_optimal_sz = ARG_DEF(NULL, "buf-optimal-sz", 1,
270 "Client optimal buffer size (ms)");
271 static const arg_def_t *rc_args[] = {
272 &dropframe_thresh, &resize_allowed, &resize_width, &resize_height,
273 &resize_up_thresh, &resize_down_thresh, &end_usage, &target_bitrate,
274 &min_quantizer, &max_quantizer, &undershoot_pct, &overshoot_pct, &buf_sz,
275 &buf_initial_sz, &buf_optimal_sz, NULL
279 static const arg_def_t bias_pct = ARG_DEF(NULL, "bias-pct", 1,
280 "CBR/VBR bias (0=CBR, 100=VBR)");
281 static const arg_def_t minsection_pct = ARG_DEF(NULL, "minsection-pct", 1,
282 "GOP min bitrate (% of target)");
283 static const arg_def_t maxsection_pct = ARG_DEF(NULL, "maxsection-pct", 1,
284 "GOP max bitrate (% of target)");
285 static const arg_def_t *rc_twopass_args[] = {
286 &bias_pct, &minsection_pct, &maxsection_pct, NULL
290 static const arg_def_t kf_min_dist = ARG_DEF(NULL, "kf-min-dist", 1,
291 "Minimum keyframe interval (frames)");
292 static const arg_def_t kf_max_dist = ARG_DEF(NULL, "kf-max-dist", 1,
293 "Maximum keyframe interval (frames)");
294 static const arg_def_t kf_disabled = ARG_DEF(NULL, "disable-kf", 0,
295 "Disable keyframe placement");
296 static const arg_def_t *kf_args[] = {
297 &kf_min_dist, &kf_max_dist, &kf_disabled, NULL
301 static const arg_def_t noise_sens = ARG_DEF(NULL, "noise-sensitivity", 1,
302 "Noise sensitivity (frames to blur)");
303 static const arg_def_t sharpness = ARG_DEF(NULL, "sharpness", 1,
304 "Filter sharpness (0-7)");
305 static const arg_def_t static_thresh = ARG_DEF(NULL, "static-thresh", 1,
306 "Motion detection threshold");
307 static const arg_def_t cpu_used = ARG_DEF(NULL, "cpu-used", 1,
308 "CPU Used (-16..16)");
309 static const arg_def_t auto_altref = ARG_DEF(NULL, "auto-alt-ref", 1,
310 "Enable automatic alt reference frames");
311 static const arg_def_t arnr_maxframes = ARG_DEF(NULL, "arnr-maxframes", 1,
312 "AltRef Max Frames");
313 static const arg_def_t arnr_strength = ARG_DEF(NULL, "arnr-strength", 1,
314 "AltRef Strength");
315 static const arg_def_t arnr_type = ARG_DEF(NULL, "arnr-type", 1,
316 "AltRef Type");
317 static const struct arg_enum_list tuning_enum[] = {
318 {"psnr", VP8_TUNE_PSNR},
319 {"ssim", VP8_TUNE_SSIM},
320 {NULL, 0}
322 static const arg_def_t tune_ssim = ARG_DEF_ENUM(NULL, "tune", 1,
323 "Material to favor", tuning_enum);
324 static const arg_def_t cq_level = ARG_DEF(NULL, "cq-level", 1,
325 "Constant/Constrained Quality level");
326 static const arg_def_t max_intra_rate_pct = ARG_DEF(NULL, "max-intra-rate", 1,
327 "Max I-frame bitrate (pct)");
329 #if CONFIG_VP8_ENCODER
330 static const arg_def_t token_parts =
331 ARG_DEF(NULL, "token-parts", 1, "Number of token partitions to use, log2");
332 static const arg_def_t *vp8_args[] = {
333 &cpu_used, &auto_altref, &noise_sens, &sharpness, &static_thresh,
334 &token_parts, &arnr_maxframes, &arnr_strength, &arnr_type,
335 &tune_ssim, &cq_level, &max_intra_rate_pct,
336 NULL
338 static const int vp8_arg_ctrl_map[] = {
339 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
340 VP8E_SET_NOISE_SENSITIVITY, VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
341 VP8E_SET_TOKEN_PARTITIONS,
342 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
343 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
346 #endif
348 #if CONFIG_VP9_ENCODER
349 static const arg_def_t tile_cols =
350 ARG_DEF(NULL, "tile-columns", 1, "Number of tile columns to use, log2");
351 static const arg_def_t tile_rows =
352 ARG_DEF(NULL, "tile-rows", 1, "Number of tile rows to use, log2");
353 static const arg_def_t lossless = ARG_DEF(NULL, "lossless", 1, "Lossless mode");
354 static const arg_def_t frame_parallel_decoding = ARG_DEF(
355 NULL, "frame-parallel", 1, "Enable frame parallel decodability features");
356 static const arg_def_t aq_mode = ARG_DEF(
357 NULL, "aq-mode", 1,
358 "Adaptive quantization mode (0: off (default), 1: variance 2: complexity, "
359 "3: cyclic refresh)");
360 static const arg_def_t frame_periodic_boost = ARG_DEF(
361 NULL, "frame_boost", 1,
362 "Enable frame periodic boost (0: off (default), 1: on)");
364 static const arg_def_t *vp9_args[] = {
365 &cpu_used, &auto_altref, &noise_sens, &sharpness, &static_thresh,
366 &tile_cols, &tile_rows, &arnr_maxframes, &arnr_strength, &arnr_type,
367 &tune_ssim, &cq_level, &max_intra_rate_pct, &lossless,
368 &frame_parallel_decoding, &aq_mode, &frame_periodic_boost,
369 NULL
371 static const int vp9_arg_ctrl_map[] = {
372 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
373 VP8E_SET_NOISE_SENSITIVITY, VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
374 VP9E_SET_TILE_COLUMNS, VP9E_SET_TILE_ROWS,
375 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
376 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
377 VP9E_SET_LOSSLESS, VP9E_SET_FRAME_PARALLEL_DECODING, VP9E_SET_AQ_MODE,
378 VP9E_SET_FRAME_PERIODIC_BOOST,
381 #endif
383 static const arg_def_t *no_args[] = { NULL };
385 void usage_exit() {
386 int i;
388 fprintf(stderr, "Usage: %s <options> -o dst_filename src_filename \n",
389 exec_name);
391 fprintf(stderr, "\nOptions:\n");
392 arg_show_usage(stderr, main_args);
393 fprintf(stderr, "\nEncoder Global Options:\n");
394 arg_show_usage(stderr, global_args);
395 fprintf(stderr, "\nRate Control Options:\n");
396 arg_show_usage(stderr, rc_args);
397 fprintf(stderr, "\nTwopass Rate Control Options:\n");
398 arg_show_usage(stderr, rc_twopass_args);
399 fprintf(stderr, "\nKeyframe Placement Options:\n");
400 arg_show_usage(stderr, kf_args);
401 #if CONFIG_VP8_ENCODER
402 fprintf(stderr, "\nVP8 Specific Options:\n");
403 arg_show_usage(stderr, vp8_args);
404 #endif
405 #if CONFIG_VP9_ENCODER
406 fprintf(stderr, "\nVP9 Specific Options:\n");
407 arg_show_usage(stderr, vp9_args);
408 #endif
409 fprintf(stderr, "\nStream timebase (--timebase):\n"
410 " The desired precision of timestamps in the output, expressed\n"
411 " in fractional seconds. Default is 1/1000.\n");
412 fprintf(stderr, "\nIncluded encoders:\n\n");
414 for (i = 0; i < get_vpx_encoder_count(); ++i) {
415 const VpxInterface *const encoder = get_vpx_encoder_by_index(i);
416 fprintf(stderr, " %-6s - %s\n",
417 encoder->name, vpx_codec_iface_name(encoder->interface()));
420 exit(EXIT_FAILURE);
423 #define mmin(a, b) ((a) < (b) ? (a) : (b))
424 static void find_mismatch(const vpx_image_t *const img1,
425 const vpx_image_t *const img2,
426 int yloc[4], int uloc[4], int vloc[4]) {
427 const uint32_t bsize = 64;
428 const uint32_t bsizey = bsize >> img1->y_chroma_shift;
429 const uint32_t bsizex = bsize >> img1->x_chroma_shift;
430 const uint32_t c_w =
431 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
432 const uint32_t c_h =
433 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
434 int match = 1;
435 uint32_t i, j;
436 yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
437 for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
438 for (j = 0; match && j < img1->d_w; j += bsize) {
439 int k, l;
440 const int si = mmin(i + bsize, img1->d_h) - i;
441 const int sj = mmin(j + bsize, img1->d_w) - j;
442 for (k = 0; match && k < si; ++k) {
443 for (l = 0; match && l < sj; ++l) {
444 if (*(img1->planes[VPX_PLANE_Y] +
445 (i + k) * img1->stride[VPX_PLANE_Y] + j + l) !=
446 *(img2->planes[VPX_PLANE_Y] +
447 (i + k) * img2->stride[VPX_PLANE_Y] + j + l)) {
448 yloc[0] = i + k;
449 yloc[1] = j + l;
450 yloc[2] = *(img1->planes[VPX_PLANE_Y] +
451 (i + k) * img1->stride[VPX_PLANE_Y] + j + l);
452 yloc[3] = *(img2->planes[VPX_PLANE_Y] +
453 (i + k) * img2->stride[VPX_PLANE_Y] + j + l);
454 match = 0;
455 break;
462 uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
463 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
464 for (j = 0; match && j < c_w; j += bsizex) {
465 int k, l;
466 const int si = mmin(i + bsizey, c_h - i);
467 const int sj = mmin(j + bsizex, c_w - j);
468 for (k = 0; match && k < si; ++k) {
469 for (l = 0; match && l < sj; ++l) {
470 if (*(img1->planes[VPX_PLANE_U] +
471 (i + k) * img1->stride[VPX_PLANE_U] + j + l) !=
472 *(img2->planes[VPX_PLANE_U] +
473 (i + k) * img2->stride[VPX_PLANE_U] + j + l)) {
474 uloc[0] = i + k;
475 uloc[1] = j + l;
476 uloc[2] = *(img1->planes[VPX_PLANE_U] +
477 (i + k) * img1->stride[VPX_PLANE_U] + j + l);
478 uloc[3] = *(img2->planes[VPX_PLANE_U] +
479 (i + k) * img2->stride[VPX_PLANE_U] + j + l);
480 match = 0;
481 break;
487 vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
488 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
489 for (j = 0; match && j < c_w; j += bsizex) {
490 int k, l;
491 const int si = mmin(i + bsizey, c_h - i);
492 const int sj = mmin(j + bsizex, c_w - j);
493 for (k = 0; match && k < si; ++k) {
494 for (l = 0; match && l < sj; ++l) {
495 if (*(img1->planes[VPX_PLANE_V] +
496 (i + k) * img1->stride[VPX_PLANE_V] + j + l) !=
497 *(img2->planes[VPX_PLANE_V] +
498 (i + k) * img2->stride[VPX_PLANE_V] + j + l)) {
499 vloc[0] = i + k;
500 vloc[1] = j + l;
501 vloc[2] = *(img1->planes[VPX_PLANE_V] +
502 (i + k) * img1->stride[VPX_PLANE_V] + j + l);
503 vloc[3] = *(img2->planes[VPX_PLANE_V] +
504 (i + k) * img2->stride[VPX_PLANE_V] + j + l);
505 match = 0;
506 break;
514 static int compare_img(const vpx_image_t *const img1,
515 const vpx_image_t *const img2) {
516 const uint32_t c_w =
517 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
518 const uint32_t c_h =
519 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
520 uint32_t i;
521 int match = 1;
523 match &= (img1->fmt == img2->fmt);
524 match &= (img1->d_w == img2->d_w);
525 match &= (img1->d_h == img2->d_h);
527 for (i = 0; i < img1->d_h; ++i)
528 match &= (memcmp(img1->planes[VPX_PLANE_Y] + i * img1->stride[VPX_PLANE_Y],
529 img2->planes[VPX_PLANE_Y] + i * img2->stride[VPX_PLANE_Y],
530 img1->d_w) == 0);
532 for (i = 0; i < c_h; ++i)
533 match &= (memcmp(img1->planes[VPX_PLANE_U] + i * img1->stride[VPX_PLANE_U],
534 img2->planes[VPX_PLANE_U] + i * img2->stride[VPX_PLANE_U],
535 c_w) == 0);
537 for (i = 0; i < c_h; ++i)
538 match &= (memcmp(img1->planes[VPX_PLANE_V] + i * img1->stride[VPX_PLANE_V],
539 img2->planes[VPX_PLANE_V] + i * img2->stride[VPX_PLANE_V],
540 c_w) == 0);
542 return match;
546 #define NELEMENTS(x) (sizeof(x)/sizeof(x[0]))
547 #define MAX(x,y) ((x)>(y)?(x):(y))
548 #if CONFIG_VP8_ENCODER && !CONFIG_VP9_ENCODER
549 #define ARG_CTRL_CNT_MAX NELEMENTS(vp8_arg_ctrl_map)
550 #elif !CONFIG_VP8_ENCODER && CONFIG_VP9_ENCODER
551 #define ARG_CTRL_CNT_MAX NELEMENTS(vp9_arg_ctrl_map)
552 #else
553 #define ARG_CTRL_CNT_MAX MAX(NELEMENTS(vp8_arg_ctrl_map), \
554 NELEMENTS(vp9_arg_ctrl_map))
555 #endif
557 /* Per-stream configuration */
558 struct stream_config {
559 struct vpx_codec_enc_cfg cfg;
560 const char *out_fn;
561 const char *stats_fn;
562 stereo_format_t stereo_fmt;
563 int arg_ctrls[ARG_CTRL_CNT_MAX][2];
564 int arg_ctrl_cnt;
565 int write_webm;
566 int have_kf_max_dist;
570 struct stream_state {
571 int index;
572 struct stream_state *next;
573 struct stream_config config;
574 FILE *file;
575 struct rate_hist *rate_hist;
576 struct EbmlGlobal ebml;
577 uint64_t psnr_sse_total;
578 uint64_t psnr_samples_total;
579 double psnr_totals[4];
580 int psnr_count;
581 int counts[64];
582 vpx_codec_ctx_t encoder;
583 unsigned int frames_out;
584 uint64_t cx_time;
585 size_t nbytes;
586 stats_io_t stats;
587 struct vpx_image *img;
588 vpx_codec_ctx_t decoder;
589 int mismatch_seen;
593 void validate_positive_rational(const char *msg,
594 struct vpx_rational *rat) {
595 if (rat->den < 0) {
596 rat->num *= -1;
597 rat->den *= -1;
600 if (rat->num < 0)
601 die("Error: %s must be positive\n", msg);
603 if (!rat->den)
604 die("Error: %s has zero denominator\n", msg);
608 static void parse_global_config(struct VpxEncoderConfig *global, char **argv) {
609 char **argi, **argj;
610 struct arg arg;
612 /* Initialize default parameters */
613 memset(global, 0, sizeof(*global));
614 global->codec = get_vpx_encoder_by_index(0);
615 global->passes = 0;
616 global->use_i420 = 1;
617 /* Assign default deadline to good quality */
618 global->deadline = VPX_DL_GOOD_QUALITY;
620 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
621 arg.argv_step = 1;
623 if (arg_match(&arg, &codecarg, argi)) {
624 global->codec = get_vpx_encoder_by_name(arg.val);
625 if (!global->codec)
626 die("Error: Unrecognized argument (%s) to --codec\n", arg.val);
627 } else if (arg_match(&arg, &passes, argi)) {
628 global->passes = arg_parse_uint(&arg);
630 if (global->passes < 1 || global->passes > 2)
631 die("Error: Invalid number of passes (%d)\n", global->passes);
632 } else if (arg_match(&arg, &pass_arg, argi)) {
633 global->pass = arg_parse_uint(&arg);
635 if (global->pass < 1 || global->pass > 2)
636 die("Error: Invalid pass selected (%d)\n",
637 global->pass);
638 } else if (arg_match(&arg, &usage, argi))
639 global->usage = arg_parse_uint(&arg);
640 else if (arg_match(&arg, &deadline, argi))
641 global->deadline = arg_parse_uint(&arg);
642 else if (arg_match(&arg, &best_dl, argi))
643 global->deadline = VPX_DL_BEST_QUALITY;
644 else if (arg_match(&arg, &good_dl, argi))
645 global->deadline = VPX_DL_GOOD_QUALITY;
646 else if (arg_match(&arg, &rt_dl, argi))
647 global->deadline = VPX_DL_REALTIME;
648 else if (arg_match(&arg, &use_yv12, argi))
649 global->use_i420 = 0;
650 else if (arg_match(&arg, &use_i420, argi))
651 global->use_i420 = 1;
652 else if (arg_match(&arg, &quietarg, argi))
653 global->quiet = 1;
654 else if (arg_match(&arg, &verbosearg, argi))
655 global->verbose = 1;
656 else if (arg_match(&arg, &limit, argi))
657 global->limit = arg_parse_uint(&arg);
658 else if (arg_match(&arg, &skip, argi))
659 global->skip_frames = arg_parse_uint(&arg);
660 else if (arg_match(&arg, &psnrarg, argi))
661 global->show_psnr = 1;
662 else if (arg_match(&arg, &recontest, argi))
663 global->test_decode = arg_parse_enum_or_int(&arg);
664 else if (arg_match(&arg, &framerate, argi)) {
665 global->framerate = arg_parse_rational(&arg);
666 validate_positive_rational(arg.name, &global->framerate);
667 global->have_framerate = 1;
668 } else if (arg_match(&arg, &out_part, argi))
669 global->out_part = 1;
670 else if (arg_match(&arg, &debugmode, argi))
671 global->debug = 1;
672 else if (arg_match(&arg, &q_hist_n, argi))
673 global->show_q_hist_buckets = arg_parse_uint(&arg);
674 else if (arg_match(&arg, &rate_hist_n, argi))
675 global->show_rate_hist_buckets = arg_parse_uint(&arg);
676 else if (arg_match(&arg, &disable_warnings, argi))
677 global->disable_warnings = 1;
678 else if (arg_match(&arg, &disable_warning_prompt, argi))
679 global->disable_warning_prompt = 1;
680 else if (arg_match(&arg, &experimental_bitstream, argi))
681 global->experimental_bitstream = 1;
682 else
683 argj++;
686 if (global->pass) {
687 /* DWIM: Assume the user meant passes=2 if pass=2 is specified */
688 if (global->pass > global->passes) {
689 warn("Assuming --pass=%d implies --passes=%d\n",
690 global->pass, global->pass);
691 global->passes = global->pass;
694 /* Validate global config */
695 if (global->passes == 0) {
696 #if CONFIG_VP9_ENCODER
697 // Make default VP9 passes = 2 until there is a better quality 1-pass
698 // encoder
699 global->passes = (strcmp(global->codec->name, "vp9") == 0 &&
700 global->deadline != VPX_DL_REALTIME) ? 2 : 1;
701 #else
702 global->passes = 1;
703 #endif
706 if (global->deadline == VPX_DL_REALTIME &&
707 global->passes > 1) {
708 warn("Enforcing one-pass encoding in realtime mode\n");
709 global->passes = 1;
714 void open_input_file(struct VpxInputContext *input) {
715 /* Parse certain options from the input file, if possible */
716 input->file = strcmp(input->filename, "-")
717 ? fopen(input->filename, "rb") : set_binary_mode(stdin);
719 if (!input->file)
720 fatal("Failed to open input file");
722 if (!fseeko(input->file, 0, SEEK_END)) {
723 /* Input file is seekable. Figure out how long it is, so we can get
724 * progress info.
726 input->length = ftello(input->file);
727 rewind(input->file);
730 /* For RAW input sources, these bytes will applied on the first frame
731 * in read_frame().
733 input->detect.buf_read = fread(input->detect.buf, 1, 4, input->file);
734 input->detect.position = 0;
736 if (input->detect.buf_read == 4
737 && file_is_y4m(input->detect.buf)) {
738 if (y4m_input_open(&input->y4m, input->file, input->detect.buf, 4,
739 input->only_i420) >= 0) {
740 input->file_type = FILE_TYPE_Y4M;
741 input->width = input->y4m.pic_w;
742 input->height = input->y4m.pic_h;
743 input->framerate.numerator = input->y4m.fps_n;
744 input->framerate.denominator = input->y4m.fps_d;
745 input->use_i420 = 0;
746 } else
747 fatal("Unsupported Y4M stream.");
748 } else if (input->detect.buf_read == 4 && fourcc_is_ivf(input->detect.buf)) {
749 fatal("IVF is not supported as input.");
750 } else {
751 input->file_type = FILE_TYPE_RAW;
756 static void close_input_file(struct VpxInputContext *input) {
757 fclose(input->file);
758 if (input->file_type == FILE_TYPE_Y4M)
759 y4m_input_close(&input->y4m);
762 static struct stream_state *new_stream(struct VpxEncoderConfig *global,
763 struct stream_state *prev) {
764 struct stream_state *stream;
766 stream = calloc(1, sizeof(*stream));
767 if (!stream)
768 fatal("Failed to allocate new stream.");
769 if (prev) {
770 memcpy(stream, prev, sizeof(*stream));
771 stream->index++;
772 prev->next = stream;
773 } else {
774 vpx_codec_err_t res;
776 /* Populate encoder configuration */
777 res = vpx_codec_enc_config_default(global->codec->interface(),
778 &stream->config.cfg,
779 global->usage);
780 if (res)
781 fatal("Failed to get config: %s\n", vpx_codec_err_to_string(res));
783 /* Change the default timebase to a high enough value so that the
784 * encoder will always create strictly increasing timestamps.
786 stream->config.cfg.g_timebase.den = 1000;
788 /* Never use the library's default resolution, require it be parsed
789 * from the file or set on the command line.
791 stream->config.cfg.g_w = 0;
792 stream->config.cfg.g_h = 0;
794 /* Initialize remaining stream parameters */
795 stream->config.stereo_fmt = STEREO_FORMAT_MONO;
796 stream->config.write_webm = 1;
797 #if CONFIG_WEBM_IO
798 stream->ebml.last_pts_ns = -1;
799 stream->ebml.writer = NULL;
800 stream->ebml.segment = NULL;
801 #endif
803 /* Allows removal of the application version from the EBML tags */
804 stream->ebml.debug = global->debug;
806 /* Default lag_in_frames is 0 in realtime mode */
807 if (global->deadline == VPX_DL_REALTIME)
808 stream->config.cfg.g_lag_in_frames = 0;
811 /* Output files must be specified for each stream */
812 stream->config.out_fn = NULL;
814 stream->next = NULL;
815 return stream;
819 static int parse_stream_params(struct VpxEncoderConfig *global,
820 struct stream_state *stream,
821 char **argv) {
822 char **argi, **argj;
823 struct arg arg;
824 static const arg_def_t **ctrl_args = no_args;
825 static const int *ctrl_args_map = NULL;
826 struct stream_config *config = &stream->config;
827 int eos_mark_found = 0;
829 // Handle codec specific options
830 if (0) {
831 #if CONFIG_VP8_ENCODER
832 } else if (strcmp(global->codec->name, "vp8") == 0) {
833 ctrl_args = vp8_args;
834 ctrl_args_map = vp8_arg_ctrl_map;
835 #endif
836 #if CONFIG_VP9_ENCODER
837 } else if (strcmp(global->codec->name, "vp9") == 0) {
838 ctrl_args = vp9_args;
839 ctrl_args_map = vp9_arg_ctrl_map;
840 #endif
843 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
844 arg.argv_step = 1;
846 /* Once we've found an end-of-stream marker (--) we want to continue
847 * shifting arguments but not consuming them.
849 if (eos_mark_found) {
850 argj++;
851 continue;
852 } else if (!strcmp(*argj, "--")) {
853 eos_mark_found = 1;
854 continue;
857 if (0) {
858 } else if (arg_match(&arg, &outputfile, argi)) {
859 config->out_fn = arg.val;
860 } else if (arg_match(&arg, &fpf_name, argi)) {
861 config->stats_fn = arg.val;
862 } else if (arg_match(&arg, &use_ivf, argi)) {
863 config->write_webm = 0;
864 } else if (arg_match(&arg, &threads, argi)) {
865 config->cfg.g_threads = arg_parse_uint(&arg);
866 } else if (arg_match(&arg, &profile, argi)) {
867 config->cfg.g_profile = arg_parse_uint(&arg);
868 } else if (arg_match(&arg, &width, argi)) {
869 config->cfg.g_w = arg_parse_uint(&arg);
870 } else if (arg_match(&arg, &height, argi)) {
871 config->cfg.g_h = arg_parse_uint(&arg);
872 } else if (arg_match(&arg, &stereo_mode, argi)) {
873 config->stereo_fmt = arg_parse_enum_or_int(&arg);
874 } else if (arg_match(&arg, &timebase, argi)) {
875 config->cfg.g_timebase = arg_parse_rational(&arg);
876 validate_positive_rational(arg.name, &config->cfg.g_timebase);
877 } else if (arg_match(&arg, &error_resilient, argi)) {
878 config->cfg.g_error_resilient = arg_parse_uint(&arg);
879 } else if (arg_match(&arg, &lag_in_frames, argi)) {
880 config->cfg.g_lag_in_frames = arg_parse_uint(&arg);
881 if (global->deadline == VPX_DL_REALTIME &&
882 config->cfg.g_lag_in_frames != 0) {
883 warn("non-zero %s option ignored in realtime mode.\n", arg.name);
884 config->cfg.g_lag_in_frames = 0;
886 } else if (arg_match(&arg, &dropframe_thresh, argi)) {
887 config->cfg.rc_dropframe_thresh = arg_parse_uint(&arg);
888 } else if (arg_match(&arg, &resize_allowed, argi)) {
889 config->cfg.rc_resize_allowed = arg_parse_uint(&arg);
890 } else if (arg_match(&arg, &resize_width, argi)) {
891 config->cfg.rc_scaled_width = arg_parse_uint(&arg);
892 } else if (arg_match(&arg, &resize_height, argi)) {
893 config->cfg.rc_scaled_height = arg_parse_uint(&arg);
894 } else if (arg_match(&arg, &resize_up_thresh, argi)) {
895 config->cfg.rc_resize_up_thresh = arg_parse_uint(&arg);
896 } else if (arg_match(&arg, &resize_down_thresh, argi)) {
897 config->cfg.rc_resize_down_thresh = arg_parse_uint(&arg);
898 } else if (arg_match(&arg, &end_usage, argi)) {
899 config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
900 } else if (arg_match(&arg, &target_bitrate, argi)) {
901 config->cfg.rc_target_bitrate = arg_parse_uint(&arg);
902 } else if (arg_match(&arg, &min_quantizer, argi)) {
903 config->cfg.rc_min_quantizer = arg_parse_uint(&arg);
904 } else if (arg_match(&arg, &max_quantizer, argi)) {
905 config->cfg.rc_max_quantizer = arg_parse_uint(&arg);
906 } else if (arg_match(&arg, &undershoot_pct, argi)) {
907 config->cfg.rc_undershoot_pct = arg_parse_uint(&arg);
908 } else if (arg_match(&arg, &overshoot_pct, argi)) {
909 config->cfg.rc_overshoot_pct = arg_parse_uint(&arg);
910 } else if (arg_match(&arg, &buf_sz, argi)) {
911 config->cfg.rc_buf_sz = arg_parse_uint(&arg);
912 } else if (arg_match(&arg, &buf_initial_sz, argi)) {
913 config->cfg.rc_buf_initial_sz = arg_parse_uint(&arg);
914 } else if (arg_match(&arg, &buf_optimal_sz, argi)) {
915 config->cfg.rc_buf_optimal_sz = arg_parse_uint(&arg);
916 } else if (arg_match(&arg, &bias_pct, argi)) {
917 config->cfg.rc_2pass_vbr_bias_pct = arg_parse_uint(&arg);
918 if (global->passes < 2)
919 warn("option %s ignored in one-pass mode.\n", arg.name);
920 } else if (arg_match(&arg, &minsection_pct, argi)) {
921 config->cfg.rc_2pass_vbr_minsection_pct = arg_parse_uint(&arg);
923 if (global->passes < 2)
924 warn("option %s ignored in one-pass mode.\n", arg.name);
925 } else if (arg_match(&arg, &maxsection_pct, argi)) {
926 config->cfg.rc_2pass_vbr_maxsection_pct = arg_parse_uint(&arg);
928 if (global->passes < 2)
929 warn("option %s ignored in one-pass mode.\n", arg.name);
930 } else if (arg_match(&arg, &kf_min_dist, argi)) {
931 config->cfg.kf_min_dist = arg_parse_uint(&arg);
932 } else if (arg_match(&arg, &kf_max_dist, argi)) {
933 config->cfg.kf_max_dist = arg_parse_uint(&arg);
934 config->have_kf_max_dist = 1;
935 } else if (arg_match(&arg, &kf_disabled, argi)) {
936 config->cfg.kf_mode = VPX_KF_DISABLED;
937 } else {
938 int i, match = 0;
939 for (i = 0; ctrl_args[i]; i++) {
940 if (arg_match(&arg, ctrl_args[i], argi)) {
941 int j;
942 match = 1;
944 /* Point either to the next free element or the first
945 * instance of this control.
947 for (j = 0; j < config->arg_ctrl_cnt; j++)
948 if (config->arg_ctrls[j][0] == ctrl_args_map[i])
949 break;
951 /* Update/insert */
952 assert(j < ARG_CTRL_CNT_MAX);
953 if (j < ARG_CTRL_CNT_MAX) {
954 config->arg_ctrls[j][0] = ctrl_args_map[i];
955 config->arg_ctrls[j][1] = arg_parse_enum_or_int(&arg);
956 if (j == config->arg_ctrl_cnt)
957 config->arg_ctrl_cnt++;
962 if (!match)
963 argj++;
966 return eos_mark_found;
970 #define FOREACH_STREAM(func) \
971 do { \
972 struct stream_state *stream; \
973 for (stream = streams; stream; stream = stream->next) { \
974 func; \
976 } while (0)
979 static void validate_stream_config(const struct stream_state *stream,
980 const struct VpxEncoderConfig *global) {
981 const struct stream_state *streami;
983 if (!stream->config.cfg.g_w || !stream->config.cfg.g_h)
984 fatal("Stream %d: Specify stream dimensions with --width (-w) "
985 " and --height (-h)", stream->index);
987 if (stream->config.cfg.g_profile != 0 && !global->experimental_bitstream) {
988 fatal("Stream %d: profile %d is experimental and requires the --%s flag",
989 stream->index, stream->config.cfg.g_profile,
990 experimental_bitstream.long_name);
993 for (streami = stream; streami; streami = streami->next) {
994 /* All streams require output files */
995 if (!streami->config.out_fn)
996 fatal("Stream %d: Output file is required (specify with -o)",
997 streami->index);
999 /* Check for two streams outputting to the same file */
1000 if (streami != stream) {
1001 const char *a = stream->config.out_fn;
1002 const char *b = streami->config.out_fn;
1003 if (!strcmp(a, b) && strcmp(a, "/dev/null") && strcmp(a, ":nul"))
1004 fatal("Stream %d: duplicate output file (from stream %d)",
1005 streami->index, stream->index);
1008 /* Check for two streams sharing a stats file. */
1009 if (streami != stream) {
1010 const char *a = stream->config.stats_fn;
1011 const char *b = streami->config.stats_fn;
1012 if (a && b && !strcmp(a, b))
1013 fatal("Stream %d: duplicate stats file (from stream %d)",
1014 streami->index, stream->index);
1020 static void set_stream_dimensions(struct stream_state *stream,
1021 unsigned int w,
1022 unsigned int h) {
1023 if (!stream->config.cfg.g_w) {
1024 if (!stream->config.cfg.g_h)
1025 stream->config.cfg.g_w = w;
1026 else
1027 stream->config.cfg.g_w = w * stream->config.cfg.g_h / h;
1029 if (!stream->config.cfg.g_h) {
1030 stream->config.cfg.g_h = h * stream->config.cfg.g_w / w;
1035 static void set_default_kf_interval(struct stream_state *stream,
1036 struct VpxEncoderConfig *global) {
1037 /* Use a max keyframe interval of 5 seconds, if none was
1038 * specified on the command line.
1040 if (!stream->config.have_kf_max_dist) {
1041 double framerate = (double)global->framerate.num / global->framerate.den;
1042 if (framerate > 0.0)
1043 stream->config.cfg.kf_max_dist = (unsigned int)(5.0 * framerate);
1048 static void show_stream_config(struct stream_state *stream,
1049 struct VpxEncoderConfig *global,
1050 struct VpxInputContext *input) {
1052 #define SHOW(field) \
1053 fprintf(stderr, " %-28s = %d\n", #field, stream->config.cfg.field)
1055 if (stream->index == 0) {
1056 fprintf(stderr, "Codec: %s\n",
1057 vpx_codec_iface_name(global->codec->interface()));
1058 fprintf(stderr, "Source file: %s Format: %s\n", input->filename,
1059 input->use_i420 ? "I420" : "YV12");
1061 if (stream->next || stream->index)
1062 fprintf(stderr, "\nStream Index: %d\n", stream->index);
1063 fprintf(stderr, "Destination file: %s\n", stream->config.out_fn);
1064 fprintf(stderr, "Encoder parameters:\n");
1066 SHOW(g_usage);
1067 SHOW(g_threads);
1068 SHOW(g_profile);
1069 SHOW(g_w);
1070 SHOW(g_h);
1071 SHOW(g_timebase.num);
1072 SHOW(g_timebase.den);
1073 SHOW(g_error_resilient);
1074 SHOW(g_pass);
1075 SHOW(g_lag_in_frames);
1076 SHOW(rc_dropframe_thresh);
1077 SHOW(rc_resize_allowed);
1078 SHOW(rc_scaled_width);
1079 SHOW(rc_scaled_height);
1080 SHOW(rc_resize_up_thresh);
1081 SHOW(rc_resize_down_thresh);
1082 SHOW(rc_end_usage);
1083 SHOW(rc_target_bitrate);
1084 SHOW(rc_min_quantizer);
1085 SHOW(rc_max_quantizer);
1086 SHOW(rc_undershoot_pct);
1087 SHOW(rc_overshoot_pct);
1088 SHOW(rc_buf_sz);
1089 SHOW(rc_buf_initial_sz);
1090 SHOW(rc_buf_optimal_sz);
1091 SHOW(rc_2pass_vbr_bias_pct);
1092 SHOW(rc_2pass_vbr_minsection_pct);
1093 SHOW(rc_2pass_vbr_maxsection_pct);
1094 SHOW(kf_mode);
1095 SHOW(kf_min_dist);
1096 SHOW(kf_max_dist);
1100 static void open_output_file(struct stream_state *stream,
1101 struct VpxEncoderConfig *global) {
1102 const char *fn = stream->config.out_fn;
1103 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1105 if (cfg->g_pass == VPX_RC_FIRST_PASS)
1106 return;
1108 stream->file = strcmp(fn, "-") ? fopen(fn, "wb") : set_binary_mode(stdout);
1110 if (!stream->file)
1111 fatal("Failed to open output file");
1113 if (stream->config.write_webm && fseek(stream->file, 0, SEEK_CUR))
1114 fatal("WebM output to pipes not supported.");
1116 #if CONFIG_WEBM_IO
1117 if (stream->config.write_webm) {
1118 stream->ebml.stream = stream->file;
1119 write_webm_file_header(&stream->ebml, cfg,
1120 &global->framerate,
1121 stream->config.stereo_fmt,
1122 global->codec->fourcc);
1124 #endif
1126 if (!stream->config.write_webm) {
1127 ivf_write_file_header(stream->file, cfg, global->codec->fourcc, 0);
1132 static void close_output_file(struct stream_state *stream,
1133 unsigned int fourcc) {
1134 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1136 if (cfg->g_pass == VPX_RC_FIRST_PASS)
1137 return;
1139 #if CONFIG_WEBM_IO
1140 if (stream->config.write_webm) {
1141 write_webm_file_footer(&stream->ebml);
1143 #endif
1145 if (!stream->config.write_webm) {
1146 if (!fseek(stream->file, 0, SEEK_SET))
1147 ivf_write_file_header(stream->file, &stream->config.cfg,
1148 fourcc,
1149 stream->frames_out);
1152 fclose(stream->file);
1156 static void setup_pass(struct stream_state *stream,
1157 struct VpxEncoderConfig *global,
1158 int pass) {
1159 if (stream->config.stats_fn) {
1160 if (!stats_open_file(&stream->stats, stream->config.stats_fn,
1161 pass))
1162 fatal("Failed to open statistics store");
1163 } else {
1164 if (!stats_open_mem(&stream->stats, pass))
1165 fatal("Failed to open statistics store");
1168 stream->config.cfg.g_pass = global->passes == 2
1169 ? pass ? VPX_RC_LAST_PASS : VPX_RC_FIRST_PASS
1170 : VPX_RC_ONE_PASS;
1171 if (pass)
1172 stream->config.cfg.rc_twopass_stats_in = stats_get(&stream->stats);
1174 stream->cx_time = 0;
1175 stream->nbytes = 0;
1176 stream->frames_out = 0;
1180 static void initialize_encoder(struct stream_state *stream,
1181 struct VpxEncoderConfig *global) {
1182 int i;
1183 int flags = 0;
1185 flags |= global->show_psnr ? VPX_CODEC_USE_PSNR : 0;
1186 flags |= global->out_part ? VPX_CODEC_USE_OUTPUT_PARTITION : 0;
1188 /* Construct Encoder Context */
1189 vpx_codec_enc_init(&stream->encoder, global->codec->interface(),
1190 &stream->config.cfg, flags);
1191 ctx_exit_on_error(&stream->encoder, "Failed to initialize encoder");
1193 /* Note that we bypass the vpx_codec_control wrapper macro because
1194 * we're being clever to store the control IDs in an array. Real
1195 * applications will want to make use of the enumerations directly
1197 for (i = 0; i < stream->config.arg_ctrl_cnt; i++) {
1198 int ctrl = stream->config.arg_ctrls[i][0];
1199 int value = stream->config.arg_ctrls[i][1];
1200 if (vpx_codec_control_(&stream->encoder, ctrl, value))
1201 fprintf(stderr, "Error: Tried to set control %d = %d\n",
1202 ctrl, value);
1204 ctx_exit_on_error(&stream->encoder, "Failed to control codec");
1207 #if CONFIG_DECODERS
1208 if (global->test_decode != TEST_DECODE_OFF) {
1209 const VpxInterface *decoder = get_vpx_decoder_by_name(global->codec->name);
1210 vpx_codec_dec_init(&stream->decoder, decoder->interface(), NULL, 0);
1212 #endif
1216 static void encode_frame(struct stream_state *stream,
1217 struct VpxEncoderConfig *global,
1218 struct vpx_image *img,
1219 unsigned int frames_in) {
1220 vpx_codec_pts_t frame_start, next_frame_start;
1221 struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1222 struct vpx_usec_timer timer;
1224 frame_start = (cfg->g_timebase.den * (int64_t)(frames_in - 1)
1225 * global->framerate.den)
1226 / cfg->g_timebase.num / global->framerate.num;
1227 next_frame_start = (cfg->g_timebase.den * (int64_t)(frames_in)
1228 * global->framerate.den)
1229 / cfg->g_timebase.num / global->framerate.num;
1231 /* Scale if necessary */
1232 if (img && (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1233 if (!stream->img)
1234 stream->img = vpx_img_alloc(NULL, VPX_IMG_FMT_I420,
1235 cfg->g_w, cfg->g_h, 16);
1236 I420Scale(img->planes[VPX_PLANE_Y], img->stride[VPX_PLANE_Y],
1237 img->planes[VPX_PLANE_U], img->stride[VPX_PLANE_U],
1238 img->planes[VPX_PLANE_V], img->stride[VPX_PLANE_V],
1239 img->d_w, img->d_h,
1240 stream->img->planes[VPX_PLANE_Y],
1241 stream->img->stride[VPX_PLANE_Y],
1242 stream->img->planes[VPX_PLANE_U],
1243 stream->img->stride[VPX_PLANE_U],
1244 stream->img->planes[VPX_PLANE_V],
1245 stream->img->stride[VPX_PLANE_V],
1246 stream->img->d_w, stream->img->d_h,
1247 kFilterBox);
1249 img = stream->img;
1252 vpx_usec_timer_start(&timer);
1253 vpx_codec_encode(&stream->encoder, img, frame_start,
1254 (unsigned long)(next_frame_start - frame_start),
1255 0, global->deadline);
1256 vpx_usec_timer_mark(&timer);
1257 stream->cx_time += vpx_usec_timer_elapsed(&timer);
1258 ctx_exit_on_error(&stream->encoder, "Stream %d: Failed to encode frame",
1259 stream->index);
1263 static void update_quantizer_histogram(struct stream_state *stream) {
1264 if (stream->config.cfg.g_pass != VPX_RC_FIRST_PASS) {
1265 int q;
1267 vpx_codec_control(&stream->encoder, VP8E_GET_LAST_QUANTIZER_64, &q);
1268 ctx_exit_on_error(&stream->encoder, "Failed to read quantizer");
1269 stream->counts[q]++;
1274 static void get_cx_data(struct stream_state *stream,
1275 struct VpxEncoderConfig *global,
1276 int *got_data) {
1277 const vpx_codec_cx_pkt_t *pkt;
1278 const struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1279 vpx_codec_iter_t iter = NULL;
1281 *got_data = 0;
1282 while ((pkt = vpx_codec_get_cx_data(&stream->encoder, &iter))) {
1283 static size_t fsize = 0;
1284 static int64_t ivf_header_pos = 0;
1286 switch (pkt->kind) {
1287 case VPX_CODEC_CX_FRAME_PKT:
1288 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1289 stream->frames_out++;
1291 if (!global->quiet)
1292 fprintf(stderr, " %6luF", (unsigned long)pkt->data.frame.sz);
1294 update_rate_histogram(stream->rate_hist, cfg, pkt);
1295 #if CONFIG_WEBM_IO
1296 if (stream->config.write_webm) {
1297 write_webm_block(&stream->ebml, cfg, pkt);
1299 #endif
1300 if (!stream->config.write_webm) {
1301 if (pkt->data.frame.partition_id <= 0) {
1302 ivf_header_pos = ftello(stream->file);
1303 fsize = pkt->data.frame.sz;
1305 ivf_write_frame_header(stream->file, pkt->data.frame.pts, fsize);
1306 } else {
1307 fsize += pkt->data.frame.sz;
1309 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1310 const int64_t currpos = ftello(stream->file);
1311 fseeko(stream->file, ivf_header_pos, SEEK_SET);
1312 ivf_write_frame_size(stream->file, fsize);
1313 fseeko(stream->file, currpos, SEEK_SET);
1317 (void) fwrite(pkt->data.frame.buf, 1, pkt->data.frame.sz,
1318 stream->file);
1320 stream->nbytes += pkt->data.raw.sz;
1322 *got_data = 1;
1323 #if CONFIG_DECODERS
1324 if (global->test_decode != TEST_DECODE_OFF && !stream->mismatch_seen) {
1325 vpx_codec_decode(&stream->decoder, pkt->data.frame.buf,
1326 (unsigned int)pkt->data.frame.sz, NULL, 0);
1327 if (stream->decoder.err) {
1328 warn_or_exit_on_error(&stream->decoder,
1329 global->test_decode == TEST_DECODE_FATAL,
1330 "Failed to decode frame %d in stream %d",
1331 stream->frames_out + 1, stream->index);
1332 stream->mismatch_seen = stream->frames_out + 1;
1335 #endif
1336 break;
1337 case VPX_CODEC_STATS_PKT:
1338 stream->frames_out++;
1339 stats_write(&stream->stats,
1340 pkt->data.twopass_stats.buf,
1341 pkt->data.twopass_stats.sz);
1342 stream->nbytes += pkt->data.raw.sz;
1343 break;
1344 case VPX_CODEC_PSNR_PKT:
1346 if (global->show_psnr) {
1347 int i;
1349 stream->psnr_sse_total += pkt->data.psnr.sse[0];
1350 stream->psnr_samples_total += pkt->data.psnr.samples[0];
1351 for (i = 0; i < 4; i++) {
1352 if (!global->quiet)
1353 fprintf(stderr, "%.3f ", pkt->data.psnr.psnr[i]);
1354 stream->psnr_totals[i] += pkt->data.psnr.psnr[i];
1356 stream->psnr_count++;
1359 break;
1360 default:
1361 break;
1367 static void show_psnr(struct stream_state *stream) {
1368 int i;
1369 double ovpsnr;
1371 if (!stream->psnr_count)
1372 return;
1374 fprintf(stderr, "Stream %d PSNR (Overall/Avg/Y/U/V)", stream->index);
1375 ovpsnr = sse_to_psnr((double)stream->psnr_samples_total, 255.0,
1376 (double)stream->psnr_sse_total);
1377 fprintf(stderr, " %.3f", ovpsnr);
1379 for (i = 0; i < 4; i++) {
1380 fprintf(stderr, " %.3f", stream->psnr_totals[i] / stream->psnr_count);
1382 fprintf(stderr, "\n");
1386 static float usec_to_fps(uint64_t usec, unsigned int frames) {
1387 return (float)(usec > 0 ? frames * 1000000.0 / (float)usec : 0);
1391 static void test_decode(struct stream_state *stream,
1392 enum TestDecodeFatality fatal,
1393 const VpxInterface *codec) {
1394 vpx_image_t enc_img, dec_img;
1396 if (stream->mismatch_seen)
1397 return;
1399 /* Get the internal reference frame */
1400 if (strcmp(codec->name, "vp8") == 0) {
1401 struct vpx_ref_frame ref_enc, ref_dec;
1402 int width, height;
1404 width = (stream->config.cfg.g_w + 15) & ~15;
1405 height = (stream->config.cfg.g_h + 15) & ~15;
1406 vpx_img_alloc(&ref_enc.img, VPX_IMG_FMT_I420, width, height, 1);
1407 enc_img = ref_enc.img;
1408 vpx_img_alloc(&ref_dec.img, VPX_IMG_FMT_I420, width, height, 1);
1409 dec_img = ref_dec.img;
1411 ref_enc.frame_type = VP8_LAST_FRAME;
1412 ref_dec.frame_type = VP8_LAST_FRAME;
1413 vpx_codec_control(&stream->encoder, VP8_COPY_REFERENCE, &ref_enc);
1414 vpx_codec_control(&stream->decoder, VP8_COPY_REFERENCE, &ref_dec);
1415 } else {
1416 struct vp9_ref_frame ref;
1418 ref.idx = 0;
1419 vpx_codec_control(&stream->encoder, VP9_GET_REFERENCE, &ref);
1420 enc_img = ref.img;
1421 vpx_codec_control(&stream->decoder, VP9_GET_REFERENCE, &ref);
1422 dec_img = ref.img;
1424 ctx_exit_on_error(&stream->encoder, "Failed to get encoder reference frame");
1425 ctx_exit_on_error(&stream->decoder, "Failed to get decoder reference frame");
1427 if (!compare_img(&enc_img, &dec_img)) {
1428 int y[4], u[4], v[4];
1429 find_mismatch(&enc_img, &dec_img, y, u, v);
1430 stream->decoder.err = 1;
1431 warn_or_exit_on_error(&stream->decoder, fatal == TEST_DECODE_FATAL,
1432 "Stream %d: Encode/decode mismatch on frame %d at"
1433 " Y[%d, %d] {%d/%d},"
1434 " U[%d, %d] {%d/%d},"
1435 " V[%d, %d] {%d/%d}",
1436 stream->index, stream->frames_out,
1437 y[0], y[1], y[2], y[3],
1438 u[0], u[1], u[2], u[3],
1439 v[0], v[1], v[2], v[3]);
1440 stream->mismatch_seen = stream->frames_out;
1443 vpx_img_free(&enc_img);
1444 vpx_img_free(&dec_img);
1448 static void print_time(const char *label, int64_t etl) {
1449 int64_t hours;
1450 int64_t mins;
1451 int64_t secs;
1453 if (etl >= 0) {
1454 hours = etl / 3600;
1455 etl -= hours * 3600;
1456 mins = etl / 60;
1457 etl -= mins * 60;
1458 secs = etl;
1460 fprintf(stderr, "[%3s %2"PRId64":%02"PRId64":%02"PRId64"] ",
1461 label, hours, mins, secs);
1462 } else {
1463 fprintf(stderr, "[%3s unknown] ", label);
1468 int main(int argc, const char **argv_) {
1469 int pass;
1470 vpx_image_t raw;
1471 int frame_avail, got_data;
1473 struct VpxInputContext input = {0};
1474 struct VpxEncoderConfig global;
1475 struct stream_state *streams = NULL;
1476 char **argv, **argi;
1477 uint64_t cx_time = 0;
1478 int stream_cnt = 0;
1479 int res = 0;
1481 exec_name = argv_[0];
1483 if (argc < 3)
1484 usage_exit();
1486 /* Setup default input stream settings */
1487 input.framerate.numerator = 30;
1488 input.framerate.denominator = 1;
1489 input.use_i420 = 1;
1490 input.only_i420 = 1;
1492 /* First parse the global configuration values, because we want to apply
1493 * other parameters on top of the default configuration provided by the
1494 * codec.
1496 argv = argv_dup(argc - 1, argv_ + 1);
1497 parse_global_config(&global, argv);
1501 /* Now parse each stream's parameters. Using a local scope here
1502 * due to the use of 'stream' as loop variable in FOREACH_STREAM
1503 * loops
1505 struct stream_state *stream = NULL;
1507 do {
1508 stream = new_stream(&global, stream);
1509 stream_cnt++;
1510 if (!streams)
1511 streams = stream;
1512 } while (parse_stream_params(&global, stream, argv));
1515 /* Check for unrecognized options */
1516 for (argi = argv; *argi; argi++)
1517 if (argi[0][0] == '-' && argi[0][1])
1518 die("Error: Unrecognized option %s\n", *argi);
1520 FOREACH_STREAM(check_encoder_config(global.disable_warning_prompt,
1521 &global, &stream->config.cfg););
1523 /* Handle non-option arguments */
1524 input.filename = argv[0];
1526 if (!input.filename)
1527 usage_exit();
1529 /* Decide if other chroma subsamplings than 4:2:0 are supported */
1530 if (global.codec->fourcc == VP9_FOURCC)
1531 input.only_i420 = 0;
1533 for (pass = global.pass ? global.pass - 1 : 0; pass < global.passes; pass++) {
1534 int frames_in = 0, seen_frames = 0;
1535 int64_t estimated_time_left = -1;
1536 int64_t average_rate = -1;
1537 int64_t lagged_count = 0;
1539 open_input_file(&input);
1541 /* If the input file doesn't specify its w/h (raw files), try to get
1542 * the data from the first stream's configuration.
1544 if (!input.width || !input.height)
1545 FOREACH_STREAM( {
1546 if (stream->config.cfg.g_w && stream->config.cfg.g_h) {
1547 input.width = stream->config.cfg.g_w;
1548 input.height = stream->config.cfg.g_h;
1549 break;
1553 /* Update stream configurations from the input file's parameters */
1554 if (!input.width || !input.height)
1555 fatal("Specify stream dimensions with --width (-w) "
1556 " and --height (-h)");
1557 FOREACH_STREAM(set_stream_dimensions(stream, input.width, input.height));
1558 FOREACH_STREAM(validate_stream_config(stream, &global));
1560 /* Ensure that --passes and --pass are consistent. If --pass is set and
1561 * --passes=2, ensure --fpf was set.
1563 if (global.pass && global.passes == 2)
1564 FOREACH_STREAM( {
1565 if (!stream->config.stats_fn)
1566 die("Stream %d: Must specify --fpf when --pass=%d"
1567 " and --passes=2\n", stream->index, global.pass);
1570 #if !CONFIG_WEBM_IO
1571 FOREACH_STREAM({
1572 stream->config.write_webm = 0;
1573 warn("vpxenc was compiled without WebM container support."
1574 "Producing IVF output");
1576 #endif
1578 /* Use the frame rate from the file only if none was specified
1579 * on the command-line.
1581 if (!global.have_framerate) {
1582 global.framerate.num = input.framerate.numerator;
1583 global.framerate.den = input.framerate.denominator;
1586 FOREACH_STREAM(set_default_kf_interval(stream, &global));
1588 /* Show configuration */
1589 if (global.verbose && pass == 0)
1590 FOREACH_STREAM(show_stream_config(stream, &global, &input));
1592 if (pass == (global.pass ? global.pass - 1 : 0)) {
1593 if (input.file_type == FILE_TYPE_Y4M)
1594 /*The Y4M reader does its own allocation.
1595 Just initialize this here to avoid problems if we never read any
1596 frames.*/
1597 memset(&raw, 0, sizeof(raw));
1598 else
1599 vpx_img_alloc(&raw,
1600 input.use_i420 ? VPX_IMG_FMT_I420
1601 : VPX_IMG_FMT_YV12,
1602 input.width, input.height, 32);
1604 FOREACH_STREAM(stream->rate_hist =
1605 init_rate_histogram(&stream->config.cfg,
1606 &global.framerate));
1609 FOREACH_STREAM(setup_pass(stream, &global, pass));
1610 FOREACH_STREAM(open_output_file(stream, &global));
1611 FOREACH_STREAM(initialize_encoder(stream, &global));
1613 frame_avail = 1;
1614 got_data = 0;
1616 while (frame_avail || got_data) {
1617 struct vpx_usec_timer timer;
1619 if (!global.limit || frames_in < global.limit) {
1620 frame_avail = read_frame(&input, &raw);
1622 if (frame_avail)
1623 frames_in++;
1624 seen_frames = frames_in > global.skip_frames ?
1625 frames_in - global.skip_frames : 0;
1627 if (!global.quiet) {
1628 float fps = usec_to_fps(cx_time, seen_frames);
1629 fprintf(stderr, "\rPass %d/%d ", pass + 1, global.passes);
1631 if (stream_cnt == 1)
1632 fprintf(stderr,
1633 "frame %4d/%-4d %7"PRId64"B ",
1634 frames_in, streams->frames_out, (int64_t)streams->nbytes);
1635 else
1636 fprintf(stderr, "frame %4d ", frames_in);
1638 fprintf(stderr, "%7"PRId64" %s %.2f %s ",
1639 cx_time > 9999999 ? cx_time / 1000 : cx_time,
1640 cx_time > 9999999 ? "ms" : "us",
1641 fps >= 1.0 ? fps : fps * 60,
1642 fps >= 1.0 ? "fps" : "fpm");
1643 print_time("ETA", estimated_time_left);
1644 fprintf(stderr, "\033[K");
1647 } else
1648 frame_avail = 0;
1650 if (frames_in > global.skip_frames) {
1651 vpx_usec_timer_start(&timer);
1652 FOREACH_STREAM(encode_frame(stream, &global,
1653 frame_avail ? &raw : NULL,
1654 frames_in));
1655 vpx_usec_timer_mark(&timer);
1656 cx_time += vpx_usec_timer_elapsed(&timer);
1658 FOREACH_STREAM(update_quantizer_histogram(stream));
1660 got_data = 0;
1661 FOREACH_STREAM(get_cx_data(stream, &global, &got_data));
1663 if (!got_data && input.length && !streams->frames_out) {
1664 lagged_count = global.limit ? seen_frames : ftello(input.file);
1665 } else if (input.length) {
1666 int64_t remaining;
1667 int64_t rate;
1669 if (global.limit) {
1670 const int64_t frame_in_lagged = (seen_frames - lagged_count) * 1000;
1672 rate = cx_time ? frame_in_lagged * (int64_t)1000000 / cx_time : 0;
1673 remaining = 1000 * (global.limit - global.skip_frames
1674 - seen_frames + lagged_count);
1675 } else {
1676 const int64_t input_pos = ftello(input.file);
1677 const int64_t input_pos_lagged = input_pos - lagged_count;
1678 const int64_t limit = input.length;
1680 rate = cx_time ? input_pos_lagged * (int64_t)1000000 / cx_time : 0;
1681 remaining = limit - input_pos + lagged_count;
1684 average_rate = (average_rate <= 0)
1685 ? rate
1686 : (average_rate * 7 + rate) / 8;
1687 estimated_time_left = average_rate ? remaining / average_rate : -1;
1690 if (got_data && global.test_decode != TEST_DECODE_OFF)
1691 FOREACH_STREAM(test_decode(stream, global.test_decode, global.codec));
1694 fflush(stdout);
1697 if (stream_cnt > 1)
1698 fprintf(stderr, "\n");
1700 if (!global.quiet)
1701 FOREACH_STREAM(fprintf(
1702 stderr,
1703 "\rPass %d/%d frame %4d/%-4d %7"PRId64"B %7lub/f %7"PRId64"b/s"
1704 " %7"PRId64" %s (%.2f fps)\033[K\n", pass + 1,
1705 global.passes, frames_in, stream->frames_out, (int64_t)stream->nbytes,
1706 seen_frames ? (unsigned long)(stream->nbytes * 8 / seen_frames) : 0,
1707 seen_frames ? (int64_t)stream->nbytes * 8
1708 * (int64_t)global.framerate.num / global.framerate.den
1709 / seen_frames
1710 : 0,
1711 stream->cx_time > 9999999 ? stream->cx_time / 1000 : stream->cx_time,
1712 stream->cx_time > 9999999 ? "ms" : "us",
1713 usec_to_fps(stream->cx_time, seen_frames));
1716 if (global.show_psnr)
1717 FOREACH_STREAM(show_psnr(stream));
1719 FOREACH_STREAM(vpx_codec_destroy(&stream->encoder));
1721 if (global.test_decode != TEST_DECODE_OFF) {
1722 FOREACH_STREAM(vpx_codec_destroy(&stream->decoder));
1725 close_input_file(&input);
1727 if (global.test_decode == TEST_DECODE_FATAL) {
1728 FOREACH_STREAM(res |= stream->mismatch_seen);
1730 FOREACH_STREAM(close_output_file(stream, global.codec->fourcc));
1732 FOREACH_STREAM(stats_close(&stream->stats, global.passes - 1));
1734 if (global.pass)
1735 break;
1738 if (global.show_q_hist_buckets)
1739 FOREACH_STREAM(show_q_histogram(stream->counts,
1740 global.show_q_hist_buckets));
1742 if (global.show_rate_hist_buckets)
1743 FOREACH_STREAM(show_rate_histogram(stream->rate_hist,
1744 &stream->config.cfg,
1745 global.show_rate_hist_buckets));
1746 FOREACH_STREAM(destroy_rate_histogram(stream->rate_hist));
1748 #if CONFIG_INTERNAL_STATS
1749 /* TODO(jkoleszar): This doesn't belong in this executable. Do it for now,
1750 * to match some existing utilities.
1752 if (!(global.pass == 1 && global.passes == 2))
1753 FOREACH_STREAM({
1754 FILE *f = fopen("opsnr.stt", "a");
1755 if (stream->mismatch_seen) {
1756 fprintf(f, "First mismatch occurred in frame %d\n",
1757 stream->mismatch_seen);
1758 } else {
1759 fprintf(f, "No mismatch detected in recon buffers\n");
1761 fclose(f);
1763 #endif
1765 vpx_img_free(&raw);
1766 free(argv);
1767 free(streams);
1768 return res ? EXIT_FAILURE : EXIT_SUCCESS;