24 #include "../tools_common.h"
25 #include "../video_writer.h"
27 #include "../vpx_ports/vpx_timer.h"
28 #include "./svc_context.h"
31 #include "../vpxstats.h"
32 #include "vp9/encoder/vp9_encoder.h"
33 #include "./y4minput.h"
35 #define OUTPUT_RC_STATS 1
37 #define SIMULCAST_MODE 0
39 static const arg_def_t outputfile =
40 ARG_DEF(
"o",
"output", 1,
"Output filename");
41 static const arg_def_t skip_frames_arg =
42 ARG_DEF(
"s",
"skip-frames", 1,
"input frames to skip");
43 static const arg_def_t frames_arg =
44 ARG_DEF(
"f",
"frames", 1,
"number of frames to encode");
45 static const arg_def_t threads_arg =
46 ARG_DEF(
"th",
"threads", 1,
"number of threads to use");
48 static const arg_def_t output_rc_stats_arg =
49 ARG_DEF(
"rcstat",
"output_rc_stats", 1,
"output rc stats");
51 static const arg_def_t width_arg = ARG_DEF(
"w",
"width", 1,
"source width");
52 static const arg_def_t height_arg = ARG_DEF(
"h",
"height", 1,
"source height");
53 static const arg_def_t timebase_arg =
54 ARG_DEF(
"t",
"timebase", 1,
"timebase (num/den)");
55 static const arg_def_t bitrate_arg = ARG_DEF(
56 "b",
"target-bitrate", 1,
"encoding bitrate, in kilobits per second");
57 static const arg_def_t spatial_layers_arg =
58 ARG_DEF(
"sl",
"spatial-layers", 1,
"number of spatial SVC layers");
59 static const arg_def_t temporal_layers_arg =
60 ARG_DEF(
"tl",
"temporal-layers", 1,
"number of temporal SVC layers");
61 static const arg_def_t temporal_layering_mode_arg =
62 ARG_DEF(
"tlm",
"temporal-layering-mode", 1,
63 "temporal layering scheme."
64 "VP9E_TEMPORAL_LAYERING_MODE");
65 static const arg_def_t kf_dist_arg =
66 ARG_DEF(
"k",
"kf-dist", 1,
"number of frames between keyframes");
67 static const arg_def_t scale_factors_arg =
68 ARG_DEF(
"r",
"scale-factors", 1,
"scale factors (lowest to highest layer)");
69 static const arg_def_t passes_arg =
70 ARG_DEF(
"p",
"passes", 1,
"Number of passes (1/2)");
71 static const arg_def_t pass_arg =
72 ARG_DEF(NULL,
"pass", 1,
"Pass to execute (1/2)");
73 static const arg_def_t fpf_name_arg =
74 ARG_DEF(NULL,
"fpf", 1,
"First pass statistics file name");
75 static const arg_def_t min_q_arg =
76 ARG_DEF(NULL,
"min-q", 1,
"Minimum quantizer");
77 static const arg_def_t max_q_arg =
78 ARG_DEF(NULL,
"max-q", 1,
"Maximum quantizer");
79 static const arg_def_t min_bitrate_arg =
80 ARG_DEF(NULL,
"min-bitrate", 1,
"Minimum bitrate");
81 static const arg_def_t max_bitrate_arg =
82 ARG_DEF(NULL,
"max-bitrate", 1,
"Maximum bitrate");
83 static const arg_def_t lag_in_frame_arg =
84 ARG_DEF(NULL,
"lag-in-frames", 1,
85 "Number of frame to input before "
86 "generating any outputs");
87 static const arg_def_t rc_end_usage_arg =
88 ARG_DEF(NULL,
"rc-end-usage", 1,
"0 - 3: VBR, CBR, CQ, Q");
89 static const arg_def_t speed_arg =
90 ARG_DEF(
"sp",
"speed", 1,
"speed configuration");
91 static const arg_def_t aqmode_arg =
92 ARG_DEF(
"aq",
"aqmode", 1,
"aq-mode off/on");
93 static const arg_def_t bitrates_arg =
94 ARG_DEF(
"bl",
"bitrates", 1,
"bitrates[sl * num_tl + tl]");
95 static const arg_def_t dropframe_thresh_arg =
96 ARG_DEF(NULL,
"drop-frame", 1,
"Temporal resampling threshold (buf %)");
97 static const struct arg_enum_list tune_content_enum[] = {
98 {
"default", VP9E_CONTENT_DEFAULT },
99 {
"screen", VP9E_CONTENT_SCREEN },
100 {
"film", VP9E_CONTENT_FILM },
104 static const arg_def_t tune_content_arg = ARG_DEF_ENUM(
105 NULL,
"tune-content", 1,
"Tune content type", tune_content_enum);
106 static const arg_def_t inter_layer_pred_arg = ARG_DEF(
107 NULL,
"inter-layer-pred", 1,
"0 - 3: On, Off, Key-frames, Constrained");
109 #if CONFIG_VP9_HIGHBITDEPTH
110 static const struct arg_enum_list bitdepth_enum[] = {
114 static const arg_def_t bitdepth_arg = ARG_DEF_ENUM(
115 "d",
"bit-depth", 1,
"Bit depth for codec 8, 10 or 12. ", bitdepth_enum);
118 static const arg_def_t *svc_args[] = { &frames_arg,
135 &temporal_layers_arg,
136 &temporal_layering_mode_arg,
141 &output_rc_stats_arg,
144 #if CONFIG_VP9_HIGHBITDEPTH
150 &dropframe_thresh_arg,
152 &inter_layer_pred_arg,
155 static const uint32_t default_frames_to_skip = 0;
156 static const uint32_t default_frames_to_code = 60 * 60;
157 static const uint32_t default_width = 1920;
158 static const uint32_t default_height = 1080;
159 static const uint32_t default_timebase_num = 1;
160 static const uint32_t default_timebase_den = 60;
161 static const uint32_t default_bitrate = 1000;
162 static const uint32_t default_spatial_layers = 5;
163 static const uint32_t default_temporal_layers = 1;
164 static const uint32_t default_kf_dist = 100;
165 static const uint32_t default_temporal_layering_mode = 0;
166 static const uint32_t default_output_rc_stats = 0;
167 static const int32_t default_speed = -1;
168 static const uint32_t default_threads = 0;
171 const char *output_filename;
172 uint32_t frames_to_code;
173 uint32_t frames_to_skip;
174 struct VpxInputContext input_ctx;
179 int inter_layer_pred;
182 static const char *exec_name;
184 void usage_exit(
void) {
185 fprintf(stderr,
"Usage: %s <options> input_filename -o output_filename\n",
187 fprintf(stderr,
"Options:\n");
188 arg_show_usage(stderr, svc_args);
192 static void parse_command_line(
int argc,
const char **argv_,
193 AppInput *app_input, SvcContext *svc_ctx,
202 const char *fpf_file_name = NULL;
203 unsigned int min_bitrate = 0;
204 unsigned int max_bitrate = 0;
205 char string_options[1024] = { 0 };
208 svc_ctx->log_level = SVC_LOG_DEBUG;
209 svc_ctx->spatial_layers = default_spatial_layers;
210 svc_ctx->temporal_layers = default_temporal_layers;
211 svc_ctx->temporal_layering_mode = default_temporal_layering_mode;
213 svc_ctx->output_rc_stat = default_output_rc_stats;
215 svc_ctx->speed = default_speed;
216 svc_ctx->threads = default_threads;
224 enc_cfg->
g_w = default_width;
225 enc_cfg->
g_h = default_height;
234 app_input->frames_to_code = default_frames_to_code;
235 app_input->frames_to_skip = default_frames_to_skip;
238 argv = argv_dup(argc - 1, argv_ + 1);
239 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
242 if (arg_match(&arg, &frames_arg, argi)) {
243 app_input->frames_to_code = arg_parse_uint(&arg);
244 }
else if (arg_match(&arg, &outputfile, argi)) {
245 app_input->output_filename = arg.val;
246 }
else if (arg_match(&arg, &width_arg, argi)) {
247 enc_cfg->
g_w = arg_parse_uint(&arg);
248 }
else if (arg_match(&arg, &height_arg, argi)) {
249 enc_cfg->
g_h = arg_parse_uint(&arg);
250 }
else if (arg_match(&arg, &timebase_arg, argi)) {
251 enc_cfg->
g_timebase = arg_parse_rational(&arg);
252 }
else if (arg_match(&arg, &bitrate_arg, argi)) {
254 }
else if (arg_match(&arg, &skip_frames_arg, argi)) {
255 app_input->frames_to_skip = arg_parse_uint(&arg);
256 }
else if (arg_match(&arg, &spatial_layers_arg, argi)) {
257 svc_ctx->spatial_layers = arg_parse_uint(&arg);
258 }
else if (arg_match(&arg, &temporal_layers_arg, argi)) {
259 svc_ctx->temporal_layers = arg_parse_uint(&arg);
261 }
else if (arg_match(&arg, &output_rc_stats_arg, argi)) {
262 svc_ctx->output_rc_stat = arg_parse_uint(&arg);
264 }
else if (arg_match(&arg, &speed_arg, argi)) {
265 svc_ctx->speed = arg_parse_uint(&arg);
266 if (svc_ctx->speed > 9) {
267 warn(
"Mapping speed %d to speed 9.\n", svc_ctx->speed);
269 }
else if (arg_match(&arg, &aqmode_arg, argi)) {
270 svc_ctx->aqmode = arg_parse_uint(&arg);
271 }
else if (arg_match(&arg, &threads_arg, argi)) {
272 svc_ctx->threads = arg_parse_uint(&arg);
273 }
else if (arg_match(&arg, &temporal_layering_mode_arg, argi)) {
276 if (svc_ctx->temporal_layering_mode) {
279 }
else if (arg_match(&arg, &kf_dist_arg, argi)) {
282 }
else if (arg_match(&arg, &scale_factors_arg, argi)) {
283 strncat(string_options,
" scale-factors=",
284 sizeof(string_options) - strlen(string_options) - 1);
285 strncat(string_options, arg.val,
286 sizeof(string_options) - strlen(string_options) - 1);
287 }
else if (arg_match(&arg, &bitrates_arg, argi)) {
288 strncat(string_options,
" bitrates=",
289 sizeof(string_options) - strlen(string_options) - 1);
290 strncat(string_options, arg.val,
291 sizeof(string_options) - strlen(string_options) - 1);
292 }
else if (arg_match(&arg, &passes_arg, argi)) {
293 passes = arg_parse_uint(&arg);
294 if (passes < 1 || passes > 2) {
295 die(
"Error: Invalid number of passes (%d)\n", passes);
297 }
else if (arg_match(&arg, &pass_arg, argi)) {
298 pass = arg_parse_uint(&arg);
299 if (pass < 1 || pass > 2) {
300 die(
"Error: Invalid pass selected (%d)\n", pass);
302 }
else if (arg_match(&arg, &fpf_name_arg, argi)) {
303 fpf_file_name = arg.val;
304 }
else if (arg_match(&arg, &min_q_arg, argi)) {
305 strncat(string_options,
" min-quantizers=",
306 sizeof(string_options) - strlen(string_options) - 1);
307 strncat(string_options, arg.val,
308 sizeof(string_options) - strlen(string_options) - 1);
309 }
else if (arg_match(&arg, &max_q_arg, argi)) {
310 strncat(string_options,
" max-quantizers=",
311 sizeof(string_options) - strlen(string_options) - 1);
312 strncat(string_options, arg.val,
313 sizeof(string_options) - strlen(string_options) - 1);
314 }
else if (arg_match(&arg, &min_bitrate_arg, argi)) {
315 min_bitrate = arg_parse_uint(&arg);
316 }
else if (arg_match(&arg, &max_bitrate_arg, argi)) {
317 max_bitrate = arg_parse_uint(&arg);
318 }
else if (arg_match(&arg, &lag_in_frame_arg, argi)) {
320 }
else if (arg_match(&arg, &rc_end_usage_arg, argi)) {
322 #if CONFIG_VP9_HIGHBITDEPTH
323 }
else if (arg_match(&arg, &bitdepth_arg, argi)) {
324 enc_cfg->
g_bit_depth = arg_parse_enum_or_int(&arg);
339 die(
"Error: Invalid bit depth selected (%d)\n", enc_cfg->
g_bit_depth);
343 }
else if (arg_match(&arg, &dropframe_thresh_arg, argi)) {
345 }
else if (arg_match(&arg, &tune_content_arg, argi)) {
346 app_input->tune_content = arg_parse_uint(&arg);
347 }
else if (arg_match(&arg, &inter_layer_pred_arg, argi)) {
348 app_input->inter_layer_pred = arg_parse_uint(&arg);
355 if (strlen(string_options) > 0)
356 vpx_svc_set_options(svc_ctx, string_options + 1);
358 if (passes == 0 || passes == 1) {
360 fprintf(stderr,
"pass is ignored since there's only one pass\n");
365 die(
"pass must be specified when passes is 2\n");
368 if (fpf_file_name == NULL) {
369 die(
"fpf must be specified when passes is 2\n");
374 if (!stats_open_file(&app_input->rc_stats, fpf_file_name, 0)) {
375 fatal(
"Failed to open statistics store");
379 if (!stats_open_file(&app_input->rc_stats, fpf_file_name, 1)) {
380 fatal(
"Failed to open statistics store");
384 app_input->passes = passes;
385 app_input->pass = pass;
389 if (min_bitrate > 0) {
393 if (max_bitrate > 0) {
400 for (argi = argv; *argi; ++argi)
401 if (argi[0][0] ==
'-' && strlen(argi[0]) > 1)
402 die(
"Error: Unrecognized option %s\n", *argi);
404 if (argv[0] == NULL) {
407 app_input->input_ctx.filename = argv[0];
410 open_input_file(&app_input->input_ctx);
411 if (app_input->input_ctx.file_type == FILE_TYPE_Y4M) {
412 enc_cfg->
g_w = app_input->input_ctx.width;
413 enc_cfg->
g_h = app_input->input_ctx.height;
416 if (enc_cfg->
g_w < 16 || enc_cfg->
g_w % 2 || enc_cfg->
g_h < 16 ||
418 die(
"Invalid resolution: %d x %d\n", enc_cfg->
g_w, enc_cfg->
g_h);
421 "Codec %s\nframes: %d, skip: %d\n"
423 "width %d, height: %d,\n"
424 "num: %d, den: %d, bitrate: %d,\n"
427 app_input->frames_to_skip, svc_ctx->spatial_layers, enc_cfg->
g_w,
434 struct RateControlStats {
453 double avg_st_encoding_bitrate;
455 double variance_st_encoding_bitrate;
464 static void set_rate_control_stats(
struct RateControlStats *rc,
475 rc->layer_framerate[layer] = framerate;
479 rc->layer_pfb[layer] =
483 (rc->layer_framerate[layer] - rc->layer_framerate[layer - 1]);
486 rc->layer_framerate[layer];
488 rc->layer_input_frames[layer] = 0;
489 rc->layer_enc_frames[layer] = 0;
490 rc->layer_tot_enc_frames[layer] = 0;
491 rc->layer_encoding_bitrate[layer] = 0.0;
492 rc->layer_avg_frame_size[layer] = 0.0;
493 rc->layer_avg_rate_mismatch[layer] = 0.0;
496 rc->window_count = 0;
497 rc->window_size = 15;
498 rc->avg_st_encoding_bitrate = 0.0;
499 rc->variance_st_encoding_bitrate = 0.0;
502 static void printout_rate_control_summary(
struct RateControlStats *rc,
506 double perc_fluctuation = 0.0;
507 int tot_num_frames = 0;
508 printf(
"Total number of processed frames: %d\n\n", frame_cnt - 1);
509 printf(
"Rate control layer stats for sl%d tl%d layer(s):\n\n",
515 const int num_dropped =
517 ? (rc->layer_input_frames[layer] - rc->layer_enc_frames[layer])
518 : (rc->layer_input_frames[layer] - rc->layer_enc_frames[layer] -
520 tot_num_frames += rc->layer_input_frames[layer];
521 rc->layer_encoding_bitrate[layer] = 0.001 * rc->layer_framerate[layer] *
522 rc->layer_encoding_bitrate[layer] /
524 rc->layer_avg_frame_size[layer] =
525 rc->layer_avg_frame_size[layer] / rc->layer_enc_frames[layer];
526 rc->layer_avg_rate_mismatch[layer] = 100.0 *
527 rc->layer_avg_rate_mismatch[layer] /
528 rc->layer_enc_frames[layer];
529 printf(
"For layer#: sl%d tl%d \n", sl, tl);
530 printf(
"Bitrate (target vs actual): %d %f.0 kbps\n",
532 rc->layer_encoding_bitrate[layer]);
533 printf(
"Average frame size (target vs actual): %f %f bits\n",
534 rc->layer_pfb[layer], rc->layer_avg_frame_size[layer]);
535 printf(
"Average rate_mismatch: %f\n", rc->layer_avg_rate_mismatch[layer]);
537 "Number of input frames, encoded (non-key) frames, "
538 "and percent dropped frames: %d %d %f.0 \n",
539 rc->layer_input_frames[layer], rc->layer_enc_frames[layer],
540 100.0 * num_dropped / rc->layer_input_frames[layer]);
544 rc->avg_st_encoding_bitrate = rc->avg_st_encoding_bitrate / rc->window_count;
545 rc->variance_st_encoding_bitrate =
546 rc->variance_st_encoding_bitrate / rc->window_count -
547 (rc->avg_st_encoding_bitrate * rc->avg_st_encoding_bitrate);
548 perc_fluctuation = 100.0 * sqrt(rc->variance_st_encoding_bitrate) /
549 rc->avg_st_encoding_bitrate;
550 printf(
"Short-time stats, for window of %d frames: \n", rc->window_size);
551 printf(
"Average, rms-variance, and percent-fluct: %f %f %f \n",
552 rc->avg_st_encoding_bitrate, sqrt(rc->variance_st_encoding_bitrate),
554 printf(
"Num of input, num of encoded (super) frames: %d %d \n", frame_cnt,
559 size_t data_sz, uint64_t sizes[8],
569 marker = *(data + data_sz - 1);
572 if ((marker & 0xe0) == 0xc0) {
573 const uint32_t frames = (marker & 0x7) + 1;
574 const uint32_t mag = ((marker >> 3) & 0x3) + 1;
575 const size_t index_sz = 2 + mag * frames;
582 const uint8_t marker2 = *(data + data_sz - index_sz);
593 const uint8_t *x = &data[data_sz - index_sz + 1];
595 for (i = 0; i < frames; ++i) {
596 uint32_t this_sz = 0;
598 for (j = 0; j < mag; ++j) this_sz |= (*x++) << (j * 8);
612 static void set_frame_flags_bypass_mode_ex0(
613 int tl,
int num_spatial_layers,
int is_key_frame,
616 for (sl = 0; sl < num_spatial_layers; ++sl)
619 for (sl = 0; sl < num_spatial_layers; ++sl) {
634 }
else if (tl == 1) {
636 ref_frame_config->
gld_fb_idx[sl] = num_spatial_layers + sl - 1;
637 ref_frame_config->
alt_fb_idx[sl] = num_spatial_layers + sl;
664 }
else if (tl == 1) {
674 if (sl < num_spatial_layers - 1) {
680 }
else if (sl == num_spatial_layers - 1) {
694 static void set_frame_flags_bypass_mode_ex1(
695 int tl,
int num_spatial_layers,
int is_key_frame,
698 for (sl = 0; sl < num_spatial_layers; ++sl)
757 #if CONFIG_VP9_DECODER && !SIMULCAST_MODE
759 const int frames_out,
int *mismatch_seen) {
762 if (*mismatch_seen)
return;
767 enc_img = ref_enc.img;
769 dec_img = ref_dec.img;
770 #if CONFIG_VP9_HIGHBITDEPTH
775 enc_img.
d_w, enc_img.
d_h, 16);
776 vpx_img_truncate_16_to_8(&enc_img, &ref_enc.img);
780 dec_img.
d_w, dec_img.
d_h, 16);
781 vpx_img_truncate_16_to_8(&dec_img, &ref_dec.img);
786 if (!compare_img(&enc_img, &dec_img)) {
787 int y[4], u[4], v[4];
788 #if CONFIG_VP9_HIGHBITDEPTH
790 find_mismatch_high(&enc_img, &dec_img, y, u, v);
792 find_mismatch(&enc_img, &dec_img, y, u, v);
795 find_mismatch(&enc_img, &dec_img, y, u, v);
799 "Encode/decode mismatch on frame %d at"
800 " Y[%d, %d] {%d/%d},"
801 " U[%d, %d] {%d/%d},"
802 " V[%d, %d] {%d/%d}\n",
803 frames_out, y[0], y[1], y[2], y[3], u[0], u[1], u[2], u[3], v[0], v[1],
805 *mismatch_seen = frames_out;
814 static void svc_output_rc_stats(
817 struct RateControlStats *rc, VpxVideoWriter **outfile,
818 const uint32_t frame_cnt,
const double framerate) {
819 int num_layers_encoded = 0;
822 uint64_t sizes_parsed[8];
824 double sum_bitrate = 0.0;
825 double sum_bitrate2 = 0.0;
827 vp9_zero(sizes_parsed);
830 sizes_parsed, &count);
834 if (cx_pkt->
data.
frame.spatial_layer_encoded[sl]) {
835 sizes[sl] = sizes_parsed[num_layers_encoded];
836 num_layers_encoded++;
841 uint64_t tot_size = 0;
843 for (sl2 = 0; sl2 < sl; ++sl2) {
844 if (cx_pkt->
data.
frame.spatial_layer_encoded[sl2]) tot_size += sizes[sl2];
846 vpx_video_writer_write_frame(outfile[sl],
847 (uint8_t *)(cx_pkt->
data.
frame.buf) + tot_size,
848 (
size_t)(sizes[sl]), cx_pkt->
data.
frame.pts);
850 for (sl2 = 0; sl2 <= sl; ++sl2) {
851 if (cx_pkt->
data.
frame.spatial_layer_encoded[sl2]) tot_size += sizes[sl2];
854 vpx_video_writer_write_frame(outfile[sl], cx_pkt->
data.
frame.buf,
855 (
size_t)(tot_size), cx_pkt->
data.
frame.pts);
859 if (cx_pkt->
data.
frame.spatial_layer_encoded[sl]) {
863 ++rc->layer_tot_enc_frames[layer];
864 rc->layer_encoding_bitrate[layer] += 8.0 * sizes[sl];
869 rc->layer_avg_frame_size[layer] += 8.0 * sizes[sl];
870 rc->layer_avg_rate_mismatch[layer] +=
871 fabs(8.0 * sizes[sl] - rc->layer_pfb[layer]) /
872 rc->layer_pfb[layer];
873 ++rc->layer_enc_frames[layer];
882 if (frame_cnt > (
unsigned int)rc->window_size) {
884 if (cx_pkt->
data.
frame.spatial_layer_encoded[sl])
885 sum_bitrate += 0.001 * 8.0 * sizes[sl] * framerate;
887 if (frame_cnt % rc->window_size == 0) {
888 rc->window_count += 1;
889 rc->avg_st_encoding_bitrate += sum_bitrate / rc->window_size;
890 rc->variance_st_encoding_bitrate +=
891 (sum_bitrate / rc->window_size) * (sum_bitrate / rc->window_size);
896 if (frame_cnt > (
unsigned int)(rc->window_size + rc->window_size / 2)) {
898 sum_bitrate2 += 0.001 * 8.0 * sizes[sl] * framerate;
901 if (frame_cnt > (
unsigned int)(2 * rc->window_size) &&
902 frame_cnt % rc->window_size == 0) {
903 rc->window_count += 1;
904 rc->avg_st_encoding_bitrate += sum_bitrate2 / rc->window_size;
905 rc->variance_st_encoding_bitrate +=
906 (sum_bitrate2 / rc->window_size) * (sum_bitrate2 / rc->window_size);
912 int main(
int argc,
const char **argv) {
914 VpxVideoWriter *writer = NULL;
921 uint32_t frame_cnt = 0;
925 int frame_duration = 1;
926 int end_of_stream = 0;
927 int frames_received = 0;
930 struct RateControlStats rc;
934 double framerate = 30.0;
936 struct vpx_usec_timer timer;
938 #if CONFIG_INTERNAL_STATS
939 FILE *f = fopen(
"opsnr.stt",
"a");
941 #if CONFIG_VP9_DECODER && !SIMULCAST_MODE
942 int mismatch_seen = 0;
945 memset(&svc_ctx, 0,
sizeof(svc_ctx));
946 memset(&app_input, 0,
sizeof(AppInput));
947 memset(&info, 0,
sizeof(VpxVideoInfo));
949 memset(&rc, 0,
sizeof(
struct RateControlStats));
953 app_input.input_ctx.framerate.numerator = 30;
954 app_input.input_ctx.framerate.denominator = 1;
955 app_input.input_ctx.only_i420 = 1;
956 app_input.input_ctx.bit_depth = 0;
958 parse_command_line(argc, argv, &app_input, &svc_ctx, &enc_cfg);
961 if (app_input.input_ctx.file_type != FILE_TYPE_Y4M) {
963 #if CONFIG_VP9_HIGHBITDEPTH
967 enc_cfg.
g_w, enc_cfg.
g_h, 32)) {
968 die(
"Failed to allocate image %dx%d\n", enc_cfg.
g_w, enc_cfg.
g_h);
972 die(
"Failed to allocate image %dx%d\n", enc_cfg.
g_w, enc_cfg.
g_h);
978 if (vpx_svc_init(&svc_ctx, &encoder, vpx_codec_vp9_cx(), &enc_cfg) !=
980 die(
"Failed to initialize encoder\n");
981 #if CONFIG_VP9_DECODER && !SIMULCAST_MODE
983 &decoder, get_vpx_decoder_by_name(
"vp9")->codec_interface(), NULL, 0))
984 die(
"Failed to initialize decoder\n");
990 rc.avg_st_encoding_bitrate = 0.0;
991 rc.variance_st_encoding_bitrate = 0.0;
992 if (svc_ctx.output_rc_stat) {
993 set_rate_control_stats(&rc, &enc_cfg);
998 info.codec_fourcc = VP9_FOURCC;
999 info.frame_width = enc_cfg.
g_w;
1000 info.frame_height = enc_cfg.
g_h;
1004 if (!(app_input.passes == 2 && app_input.pass == 1)) {
1007 vpx_video_writer_open(app_input.output_filename, kContainerIVF, &info);
1009 die(
"Failed to open %s for writing\n", app_input.output_filename);
1014 if (svc_ctx.output_rc_stat) {
1016 char file_name[PATH_MAX];
1018 snprintf(file_name,
sizeof(file_name),
"%s_s%d.ivf",
1019 app_input.output_filename, sl);
1020 outfile[sl] = vpx_video_writer_open(file_name, kContainerIVF, &info);
1021 if (!outfile[sl]) die(
"Failed to open %s for writing", file_name);
1027 for (i = 0; i < app_input.frames_to_skip; ++i)
1028 read_frame(&app_input.input_ctx, &raw);
1030 if (svc_ctx.speed != -1)
1032 if (svc_ctx.threads) {
1034 get_msb(svc_ctx.threads));
1035 if (svc_ctx.threads > 1)
1040 if (svc_ctx.speed >= 5 && svc_ctx.aqmode == 1)
1042 if (svc_ctx.speed >= 5)
1047 app_input.inter_layer_pred);
1054 for (sl = 0; sl < (
unsigned int)svc_ctx.spatial_layers; ++sl)
1060 while (!end_of_stream) {
1068 int example_pattern = 0;
1069 if (frame_cnt >= app_input.frames_to_code ||
1070 !read_frame(&app_input.input_ctx, &raw)) {
1083 layer_id.spatial_layer_id = 0;
1085 if (frame_cnt % 2 == 0) {
1086 layer_id.temporal_layer_id = 0;
1088 layer_id.temporal_layer_id_per_spatial[i] = 0;
1090 layer_id.temporal_layer_id = 1;
1092 layer_id.temporal_layer_id_per_spatial[i] = 1;
1094 if (example_pattern == 1) {
1096 assert(svc_ctx.spatial_layers == 2);
1097 assert(svc_ctx.temporal_layers == 2);
1098 if (frame_cnt % 2 == 0) {
1100 layer_id.temporal_layer_id_per_spatial[0] = 0;
1101 layer_id.temporal_layer_id_per_spatial[1] = 0;
1102 layer_id.spatial_layer_id = 0;
1105 layer_id.temporal_layer_id_per_spatial[1] = 1;
1106 layer_id.spatial_layer_id = 1;
1112 if (example_pattern == 0) {
1113 set_frame_flags_bypass_mode_ex0(layer_id.temporal_layer_id,
1114 svc_ctx.spatial_layers, frame_cnt == 0,
1116 }
else if (example_pattern == 1) {
1117 set_frame_flags_bypass_mode_ex1(layer_id.temporal_layer_id,
1118 svc_ctx.spatial_layers, frame_cnt == 0,
1121 ref_frame_config.
duration[0] = frame_duration * 1;
1122 ref_frame_config.
duration[1] = frame_duration * 1;
1130 layer_id.temporal_layer_id];
1134 unsigned int tl = 0;
1136 tl = (frame_cnt % 2 != 0);
1138 if (frame_cnt % 2 != 0) tl = 2;
1139 if ((frame_cnt > 1) && ((frame_cnt - 2) % 4 == 0)) tl = 1;
1145 vpx_usec_timer_start(&timer);
1146 res = vpx_svc_encode(
1147 &svc_ctx, &encoder, (end_of_stream ? NULL : &raw), pts, frame_duration,
1149 vpx_usec_timer_mark(&timer);
1150 cx_time += vpx_usec_timer_elapsed(&timer);
1154 die_codec(&encoder,
"Failed to encode frame");
1158 switch (cx_pkt->
kind) {
1160 SvcInternal_t *
const si = (SvcInternal_t *)svc_ctx.internal;
1162 vpx_video_writer_write_frame(writer, cx_pkt->
data.
frame.buf,
1166 if (svc_ctx.output_rc_stat) {
1167 svc_output_rc_stats(&encoder, &enc_cfg, &layer_id, cx_pkt, &rc,
1168 outfile, frame_cnt, framerate);
1178 si->bytes_sum[0] += (int)cx_pkt->
data.
frame.sz;
1180 #
if CONFIG_VP9_DECODER && !SIMULCAST_MODE
1182 (
unsigned int)cx_pkt->
data.
frame.sz, NULL, 0))
1183 die_codec(&decoder,
"Failed to decode frame.");
1195 #if CONFIG_VP9_DECODER && !SIMULCAST_MODE
1200 !(layer_id.temporal_layer_id > 0 &&
1204 test_decode(&encoder, &decoder, frame_cnt, &mismatch_seen);
1209 if (!end_of_stream) {
1211 pts += frame_duration;
1215 printf(
"Processed %d frames\n", frame_cnt);
1217 close_input_file(&app_input.input_ctx);
1220 if (svc_ctx.output_rc_stat) {
1221 printout_rate_control_summary(&rc, &enc_cfg, frame_cnt);
1226 die_codec(&encoder,
"Failed to destroy codec");
1227 if (app_input.passes == 2) stats_close(&app_input.rc_stats, 1);
1229 vpx_video_writer_close(writer);
1232 if (svc_ctx.output_rc_stat) {
1234 vpx_video_writer_close(outfile[sl]);
1238 #if CONFIG_INTERNAL_STATS
1239 if (mismatch_seen) {
1240 fprintf(f,
"First mismatch occurred in frame %d\n", mismatch_seen);
1242 fprintf(f,
"No mismatch detected in recon buffers\n");
1246 printf(
"Frame cnt and encoding time/FPS stats for encoding: %d %f %f \n",
1247 frame_cnt, 1000 * (
float)cx_time / (
double)(frame_cnt * 1000000),
1248 1000000 * (
double)frame_cnt / (
double)cx_time);
1249 if (app_input.input_ctx.file_type != FILE_TYPE_Y4M) {
1253 vpx_svc_dump_statistics(&svc_ctx);
1254 vpx_svc_release(&svc_ctx);
1255 return EXIT_SUCCESS;
vpx_codec_err_t vpx_codec_destroy(vpx_codec_ctx_t *ctx)
Destroy a codec instance.
const void * vpx_codec_iter_t
Iterator.
Definition: vpx_codec.h:190
const char * vpx_codec_iface_name(vpx_codec_iface_t *iface)
Return the name for a given interface.
const char * vpx_codec_err_to_string(vpx_codec_err_t err)
Convert error number to printable string.
#define vpx_codec_control(ctx, id, data)
vpx_codec_control wrapper macro
Definition: vpx_codec.h:407
vpx_codec_err_t
Algorithm return codes.
Definition: vpx_codec.h:93
@ VPX_CODEC_CORRUPT_FRAME
The coded data for this stream is corrupt or incomplete.
Definition: vpx_codec.h:133
@ VPX_CODEC_OK
Operation completed without error.
Definition: vpx_codec.h:95
@ VPX_BITS_8
Definition: vpx_codec.h:221
@ VPX_BITS_12
Definition: vpx_codec.h:223
@ VPX_BITS_10
Definition: vpx_codec.h:222
vpx_codec_err_t vpx_codec_decode(vpx_codec_ctx_t *ctx, const uint8_t *data, unsigned int data_sz, void *user_priv, long deadline)
Decode data.
#define vpx_codec_dec_init(ctx, iface, cfg, flags)
Convenience macro for vpx_codec_dec_init_ver()
Definition: vpx_decoder.h:143
#define VPX_DL_REALTIME
deadline parameter analogous to VPx REALTIME mode.
Definition: vpx_encoder.h:830
#define VPX_DL_GOOD_QUALITY
deadline parameter analogous to VPx GOOD QUALITY mode.
Definition: vpx_encoder.h:832
#define VPX_MAX_LAYERS
Definition: vpx_encoder.h:43
#define VPX_FRAME_IS_KEY
Definition: vpx_encoder.h:116
#define VPX_SS_MAX_LAYERS
Definition: vpx_encoder.h:46
vpx_codec_err_t vpx_codec_enc_config_default(vpx_codec_iface_t *iface, vpx_codec_enc_cfg_t *cfg, unsigned int usage)
Get a default configuration.
const vpx_codec_cx_pkt_t * vpx_codec_get_cx_data(vpx_codec_ctx_t *ctx, vpx_codec_iter_t *iter)
Encoded data iterator.
@ VPX_CODEC_CX_FRAME_PKT
Definition: vpx_encoder.h:147
@ VPX_CODEC_STATS_PKT
Definition: vpx_encoder.h:148
@ VPX_RC_LAST_PASS
Definition: vpx_encoder.h:227
@ VPX_RC_ONE_PASS
Definition: vpx_encoder.h:225
@ VPX_RC_FIRST_PASS
Definition: vpx_encoder.h:226
@ VPX_CQ
Definition: vpx_encoder.h:234
@ FULL_SUPERFRAME_DROP
Definition: vp8cx.h:849
@ VP9E_SET_SVC_LAYER_ID
Codec control function to set svc layer for spatial and temporal.
Definition: vp8cx.h:453
@ VP8E_SET_MAX_INTRA_BITRATE_PCT
Codec control function to set Max data rate for Intra frames.
Definition: vp8cx.h:257
@ VP9E_SET_SVC_INTER_LAYER_PRED
Codec control function to constrain the inter-layer prediction (prediction of lower spatial resolutio...
Definition: vp8cx.h:619
@ VP9E_SET_AQ_MODE
Codec control function to set adaptive quantization mode.
Definition: vp8cx.h:398
@ VP9E_SET_TUNE_CONTENT
Codec control function to set content type.
Definition: vp8cx.h:463
@ VP9E_SET_ROW_MT
Codec control function to set row level multi-threading.
Definition: vp8cx.h:570
@ VP8E_SET_CPUUSED
Codec control function to set encoder internal speed settings.
Definition: vp8cx.h:155
@ VP9E_SET_TILE_COLUMNS
Codec control function to set number of tile columns.
Definition: vp8cx.h:351
@ VP9E_SET_SVC_FRAME_DROP_LAYER
Codec control function to set mode and thresholds for frame dropping in SVC. Drop frame thresholds ar...
Definition: vp8cx.h:628
@ VP9E_SET_SVC_REF_FRAME_CONFIG
Codec control function to set the frame flags and buffer indices for spatial layers....
Definition: vp8cx.h:545
@ VP8E_SET_STATIC_THRESHOLD
Codec control function to set the threshold for MBs treated static.
Definition: vp8cx.h:188
@ VP9E_SET_NOISE_SENSITIVITY
Codec control function to set noise sensitivity.
Definition: vp8cx.h:421
@ VP9E_GET_SVC_LAYER_ID
Codec control function to get svc layer ID.
Definition: vp8cx.h:471
@ VP9E_TEMPORAL_LAYERING_MODE_BYPASS
Bypass mode. Used when application needs to control temporal layering. This will only work when the n...
Definition: vp8cx.h:716
@ VP9_GET_REFERENCE
Definition: vp8.h:55
VP9 specific reference frame data struct.
Definition: vp8.h:110
int idx
Definition: vp8.h:111
Codec context structure.
Definition: vpx_codec.h:200
vpx_codec_err_t err
Definition: vpx_codec.h:203
Encoder output packet.
Definition: vpx_encoder.h:159
vpx_fixed_buf_t twopass_stats
Definition: vpx_encoder.h:182
enum vpx_codec_cx_pkt_kind kind
Definition: vpx_encoder.h:160
struct vpx_codec_cx_pkt::@1::@2 frame
union vpx_codec_cx_pkt::@1 data
Encoder configuration structure.
Definition: vpx_encoder.h:268
int temporal_layering_mode
Temporal layering mode indicating which temporal layering scheme to use.
Definition: vpx_encoder.h:693
unsigned int kf_min_dist
Keyframe minimum interval.
Definition: vpx_encoder.h:605
unsigned int ts_number_layers
Number of temporal coding layers.
Definition: vpx_encoder.h:644
vpx_fixed_buf_t rc_twopass_stats_in
Two-pass stats buffer.
Definition: vpx_encoder.h:447
unsigned int ss_number_layers
Number of spatial coding layers.
Definition: vpx_encoder.h:624
unsigned int rc_2pass_vbr_minsection_pct
Two-pass mode per-GOP minimum bitrate.
Definition: vpx_encoder.h:570
unsigned int g_profile
Bitstream profile to use.
Definition: vpx_encoder.h:295
unsigned int layer_target_bitrate[12]
Target bitrate for each spatial/temporal layer.
Definition: vpx_encoder.h:684
unsigned int g_h
Height of the frame.
Definition: vpx_encoder.h:313
vpx_codec_er_flags_t g_error_resilient
Enable error resilient modes.
Definition: vpx_encoder.h:351
unsigned int g_w
Width of the frame.
Definition: vpx_encoder.h:304
unsigned int rc_dropframe_thresh
Temporal resampling configuration, if supported by the codec.
Definition: vpx_encoder.h:391
struct vpx_rational g_timebase
Stream timebase units.
Definition: vpx_encoder.h:343
enum vpx_enc_pass g_pass
Multi-pass Encoding Mode.
Definition: vpx_encoder.h:358
unsigned int g_lag_in_frames
Allow lagged encoding.
Definition: vpx_encoder.h:372
enum vpx_rc_mode rc_end_usage
Rate control algorithm to use.
Definition: vpx_encoder.h:440
vpx_bit_depth_t g_bit_depth
Bit-depth of the codec.
Definition: vpx_encoder.h:321
unsigned int rc_2pass_vbr_maxsection_pct
Two-pass mode per-GOP maximum bitrate.
Definition: vpx_encoder.h:577
unsigned int rc_target_bitrate
Target data rate.
Definition: vpx_encoder.h:460
unsigned int g_input_bit_depth
Bit-depth of the input frames.
Definition: vpx_encoder.h:329
unsigned int ts_rate_decimator[5]
Frame rate decimation factor for each temporal layer.
Definition: vpx_encoder.h:658
unsigned int kf_max_dist
Keyframe maximum interval.
Definition: vpx_encoder.h:614
size_t sz
Definition: vpx_encoder.h:98
void * buf
Definition: vpx_encoder.h:97
Image Descriptor.
Definition: vpx_image.h:72
vpx_img_fmt_t fmt
Definition: vpx_image.h:73
unsigned int d_h
Definition: vpx_image.h:84
unsigned int d_w
Definition: vpx_image.h:83
int den
Definition: vpx_encoder.h:220
int num
Definition: vpx_encoder.h:219
vp9 svc frame dropping parameters.
Definition: vp8cx.h:861
int framedrop_thresh[5]
Definition: vp8cx.h:862
SVC_LAYER_DROP_MODE framedrop_mode
Definition: vp8cx.h:864
int max_consec_drop
Definition: vp8cx.h:865
vp9 svc layer parameters
Definition: vp8cx.h:810
int temporal_layer_id
Definition: vp8cx.h:813
vp9 svc frame flag parameters.
Definition: vp8cx.h:825
int lst_fb_idx[5]
Definition: vp8cx.h:826
int update_buffer_slot[5]
Definition: vp8cx.h:829
int gld_fb_idx[5]
Definition: vp8cx.h:827
int reference_last[5]
Definition: vp8cx.h:834
int reference_golden[5]
Definition: vp8cx.h:835
int reference_alt_ref[5]
Definition: vp8cx.h:836
int64_t duration[5]
Definition: vp8cx.h:837
int alt_fb_idx[5]
Definition: vp8cx.h:828
Provides definitions for using VP8 or VP9 encoder algorithm within the vpx Codec Interface.
Describes the encoder algorithm interface to applications.
#define VPX_IMG_FMT_HIGHBITDEPTH
Definition: vpx_image.h:35
@ VPX_IMG_FMT_I42016
Definition: vpx_image.h:47
@ VPX_IMG_FMT_I420
Definition: vpx_image.h:42
vpx_image_t * vpx_img_alloc(vpx_image_t *img, vpx_img_fmt_t fmt, unsigned int d_w, unsigned int d_h, unsigned int align)
Open a descriptor, allocating storage for the underlying image.
void vpx_img_free(vpx_image_t *img)
Close an image descriptor.