23#include "./vpx_config.h"
32#include "vpx_ports/vpx_timer.h"
35#include "vpx_ports/mem_ops.h"
36#include "../tools_common.h"
37#define interface (vpx_codec_vp8_cx())
38#define fourcc 0x30385056
40void usage_exit(
void) { exit(EXIT_FAILURE); }
55#define MAX_NUM_TEMPORAL_LAYERS 3
58#include "third_party/libyuv/include/libyuv/basic_types.h"
59#include "third_party/libyuv/include/libyuv/scale.h"
60#include "third_party/libyuv/include/libyuv/cpu_id.h"
64static int mulres_read_frame(FILE *f,
vpx_image_t *img) {
65 size_t nbytes, to_read;
68 to_read = img->
w * img->
h * 3 / 2;
69 nbytes = fread(img->
planes[0], 1, to_read, f);
70 if (nbytes != to_read) {
73 printf(
"Warning: Read partial frame. Check your width & height!\n");
78static int mulres_read_frame_by_row(FILE *f,
vpx_image_t *img) {
79 size_t nbytes, to_read;
83 for (plane = 0; plane < 3; plane++) {
85 int w = (plane ? (1 + img->
d_w) / 2 : img->
d_w);
86 int h = (plane ? (1 + img->
d_h) / 2 : img->
d_h);
102 default: ptr = img->
planes[plane];
105 for (r = 0; r < h; r++) {
108 nbytes = fread(ptr, 1, to_read, f);
109 if (nbytes != to_read) {
112 printf(
"Warning: Read partial frame. Check your width & height!\n");
116 ptr += img->
stride[plane];
133 mem_put_le16(header + 4, 0);
134 mem_put_le16(header + 6, 32);
135 mem_put_le32(header + 8, fourcc);
136 mem_put_le16(header + 12, cfg->
g_w);
137 mem_put_le16(header + 14, cfg->
g_h);
140 mem_put_le32(header + 24, frame_cnt);
141 mem_put_le32(header + 28, 0);
143 (void)fwrite(header, 1, 32, outfile);
146static void write_ivf_frame_header(FILE *outfile,
155 mem_put_le32(header + 4, pts & 0xFFFFFFFF);
156 mem_put_le32(header + 8, pts >> 32);
158 (void)fwrite(header, 1, 12, outfile);
166static void set_temporal_layer_pattern(
int num_temporal_layers,
169 assert(num_temporal_layers <= MAX_NUM_TEMPORAL_LAYERS);
170 switch (num_temporal_layers) {
216 layer_flags[4] = layer_flags[2];
219 layer_flags[5] = layer_flags[3];
222 layer_flags[6] = layer_flags[4];
225 layer_flags[7] = layer_flags[5];
271 layer_flags[5] = layer_flags[3];
277 layer_flags[7] = layer_flags[3];
284static int periodicity_to_num_layers[MAX_NUM_TEMPORAL_LAYERS] = { 1, 8, 8 };
286int main(
int argc,
char **argv) {
287 FILE *infile, *outfile[NUM_ENCODERS];
288 FILE *downsampled_input[NUM_ENCODERS - 1];
306 int flag_periodicity;
315 int key_frame_insert = 0;
316 uint64_t psnr_sse_total[NUM_ENCODERS] = { 0 };
317 uint64_t psnr_samples_total[NUM_ENCODERS] = { 0 };
318 double psnr_totals[NUM_ENCODERS][4] = { { 0, 0 } };
319 int psnr_count[NUM_ENCODERS] = { 0 };
327 unsigned int target_bitrate[NUM_ENCODERS] = { 1000, 500, 100 };
336 vpx_rational_t dsf[NUM_ENCODERS] = { { 2, 1 }, { 2, 1 }, { 1, 1 } };
340 unsigned int num_temporal_layers[NUM_ENCODERS] = { 3, 3, 3 };
342 if (argc != (7 + 3 * NUM_ENCODERS))
343 die(
"Usage: %s <width> <height> <frame_rate> <infile> <outfile(s)> "
344 "<rate_encoder(s)> <temporal_layer(s)> <key_frame_insert> <output "
350 width = (int)strtol(argv[1], NULL, 0);
351 height = (int)strtol(argv[2], NULL, 0);
352 framerate = (int)strtol(argv[3], NULL, 0);
354 if (width < 16 || width % 2 || height < 16 || height % 2)
355 die(
"Invalid resolution: %dx%d", width, height);
358 if (!(infile = fopen(argv[4],
"rb")))
359 die(
"Failed to open %s for reading", argv[4]);
362 for (i = 0; i < NUM_ENCODERS; i++) {
363 if (!target_bitrate[i]) {
368 if (!(outfile[i] = fopen(argv[i + 5],
"wb")))
369 die(
"Failed to open %s for writing", argv[i + 4]);
373 for (i = 0; i < NUM_ENCODERS; i++) {
374 target_bitrate[i] = (int)strtol(argv[NUM_ENCODERS + 5 + i], NULL, 0);
378 for (i = 0; i < NUM_ENCODERS; i++) {
379 num_temporal_layers[i] =
380 (int)strtol(argv[2 * NUM_ENCODERS + 5 + i], NULL, 0);
381 if (num_temporal_layers[i] < 1 || num_temporal_layers[i] > 3)
382 die(
"Invalid temporal layers: %d, Must be 1, 2, or 3. \n",
383 num_temporal_layers[i]);
387 for (i = 0; i < NUM_ENCODERS - 1; i++) {
389 if (sprintf(filename,
"ds%d.yuv", NUM_ENCODERS - i) < 0) {
392 downsampled_input[i] = fopen(filename,
"wb");
395 key_frame_insert = (int)strtol(argv[3 * NUM_ENCODERS + 5], NULL, 0);
397 show_psnr = (int)strtol(argv[3 * NUM_ENCODERS + 6], NULL, 0);
400 for (i = 0; i < NUM_ENCODERS; i++) {
440 for (i = 1; i < NUM_ENCODERS; i++) {
450 unsigned int iw = cfg[i - 1].
g_w * dsf[i - 1].
den + dsf[i - 1].
num - 1;
451 unsigned int ih = cfg[i - 1].
g_h * dsf[i - 1].
den + dsf[i - 1].
num - 1;
452 cfg[i].
g_w = iw / dsf[i - 1].
num;
453 cfg[i].
g_h = ih / dsf[i - 1].
num;
458 if ((cfg[i].g_w) % 2) cfg[i].
g_w++;
459 if ((cfg[i].g_h) % 2) cfg[i].
g_h++;
469 for (i = 0; i < NUM_ENCODERS; i++)
471 die(
"Failed to allocate image (%dx%d)", cfg[i].g_w, cfg[i].g_h);
474 read_frame_p = mulres_read_frame;
476 read_frame_p = mulres_read_frame_by_row;
478 for (i = 0; i < NUM_ENCODERS; i++)
479 if (outfile[i]) write_ivf_file_header(outfile[i], &cfg[i], 0);
482 for (i = 0; i < NUM_ENCODERS; i++) {
483 set_temporal_layer_pattern(num_temporal_layers[i], &cfg[i],
484 cfg[i].rc_target_bitrate,
491 die_codec(&codec[0],
"Failed to initialize encoder");
495 for (i = 0; i < NUM_ENCODERS; i++) {
498 if (i == NUM_ENCODERS - 1) speed = -4;
500 die_codec(&codec[i],
"Failed to set cpu_used");
504 for (i = 0; i < NUM_ENCODERS; i++) {
506 die_codec(&codec[i],
"Failed to set static threshold");
512 die_codec(&codec[0],
"Failed to set noise_sensitivity");
514 die_codec(&codec[1],
"Failed to set noise_sensitivity");
515 for (i = 2; i < NUM_ENCODERS; i++) {
517 die_codec(&codec[i],
"Failed to set noise_sensitivity");
521 for (i = 0; i < NUM_ENCODERS; i++) {
523 die_codec(&codec[i],
"Failed to set static threshold");
527 for (i = 0; i < NUM_ENCODERS; i++) {
528 unsigned int max_intra_size_pct =
529 (int)(((
double)cfg[0].rc_buf_optimal_sz * 0.5) * framerate / 10);
532 die_codec(&codec[i],
"Failed to set static threshold");
539 while (frame_avail || got_data) {
540 struct vpx_usec_timer timer;
545 frame_avail = read_frame_p(infile, &raw[0]);
548 for (i = 1; i < NUM_ENCODERS; i++) {
555 raw[i - 1].d_w, raw[i - 1].d_h, raw[i].planes[
VPX_PLANE_Y],
558 raw[i].stride[
VPX_PLANE_V], raw[i].d_w, raw[i].d_h, 1);
560 length_frame = cfg[i].
g_w * cfg[i].
g_h * 3 / 2;
561 if (fwrite(raw[i].planes[0], 1, length_frame,
562 downsampled_input[NUM_ENCODERS - i - 1]) !=
563 (
unsigned int)length_frame) {
570 for (i = 0; i < NUM_ENCODERS; i++) {
573 flag_periodicity = periodicity_to_num_layers[num_temporal_layers[i] - 1];
575 frame_cnt % flag_periodicity];
577 if (frame_cnt == 0) {
580 if (frame_cnt > 0 && frame_cnt == key_frame_insert) {
591 vpx_usec_timer_start(&timer);
594 die_codec(&codec[0],
"Failed to encode frame");
596 vpx_usec_timer_mark(&timer);
597 cx_time += vpx_usec_timer_elapsed(&timer);
599 for (i = NUM_ENCODERS - 1; i >= 0; i--) {
603 switch (pkt[i]->kind) {
605 write_ivf_frame_header(outfile[i], pkt[i]);
606 (void)fwrite(pkt[i]->data.frame.buf, 1, pkt[i]->data.frame.sz,
613 psnr_sse_total[i] += pkt[i]->
data.
psnr.sse[0];
614 psnr_samples_total[i] += pkt[i]->
data.
psnr.samples[0];
615 for (j = 0; j < 4; j++) {
616 psnr_totals[i][j] += pkt[i]->
data.
psnr.psnr[j];
630 printf(
"Frame cnt and encoding time/FPS stats for encoding: %d %f %f \n",
631 frame_cnt, 1000 * (
float)cx_time / (
double)(frame_cnt * 1000000),
632 1000000 * (
double)frame_cnt / (
double)cx_time);
636 printf(
"Processed %ld frames.\n", (
long int)frame_cnt - 1);
637 for (i = 0; i < NUM_ENCODERS; i++) {
639 if ((show_psnr) && (psnr_count[i] > 0)) {
642 sse_to_psnr(psnr_samples_total[i], 255.0, psnr_sse_total[i]);
644 fprintf(stderr,
"\n ENC%d PSNR (Overall/Avg/Y/U/V)", i);
646 fprintf(stderr,
" %.3lf", ovpsnr);
647 for (j = 0; j < 4; j++) {
648 fprintf(stderr,
" %.3lf", psnr_totals[i][j] / psnr_count[i]);
653 die_codec(&codec[i],
"Failed to destroy codec");
657 if (!outfile[i])
continue;
660 if (!fseek(outfile[i], 0, SEEK_SET))
661 write_ivf_file_header(outfile[i], &cfg[i], frame_cnt - 1);
const char * vpx_codec_err_to_string(vpx_codec_err_t err)
Convert error number to printable string.
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.
#define vpx_codec_control(ctx, id, data)
vpx_codec_control wrapper macro
Definition vpx_codec.h:408
vpx_codec_err_t
Algorithm return codes.
Definition vpx_codec.h:93
#define VPX_DL_REALTIME
deadline parameter analogous to VPx REALTIME mode.
Definition vpx_encoder.h:1004
#define VPX_EFLAG_FORCE_KF
Definition vpx_encoder.h:268
#define VPX_CODEC_USE_PSNR
Initialization-time Feature Enabling.
Definition vpx_encoder.h:94
const vpx_codec_cx_pkt_t * vpx_codec_get_cx_data(vpx_codec_ctx_t *ctx, vpx_codec_iter_t *iter)
Encoded data iterator.
#define VPX_TS_MAX_PERIODICITY
Definition vpx_encoder.h:38
int64_t vpx_codec_pts_t
Time Stamp Type.
Definition vpx_encoder.h:113
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.
#define vpx_codec_enc_init_multi(ctx, iface, cfg, num_enc, flags, dsf)
Convenience macro for vpx_codec_enc_init_multi_ver()
Definition vpx_encoder.h:931
vpx_codec_err_t vpx_codec_encode(vpx_codec_ctx_t *ctx, const vpx_image_t *img, vpx_codec_pts_t pts, unsigned long duration, vpx_enc_frame_flags_t flags, vpx_enc_deadline_t deadline)
Encode a frame.
@ VPX_CODEC_PSNR_PKT
Definition vpx_encoder.h:158
@ VPX_CODEC_CX_FRAME_PKT
Definition vpx_encoder.h:155
@ VPX_RC_LAST_PASS
Definition vpx_encoder.h:235
@ VPX_RC_ONE_PASS
Definition vpx_encoder.h:233
@ VPX_KF_AUTO
Definition vpx_encoder.h:256
@ VPX_CBR
Definition vpx_encoder.h:241
#define VP8_EFLAG_NO_UPD_ARF
Don't update the alternate reference frame.
Definition vp8cx.h:112
#define VP8_EFLAG_NO_UPD_ENTROPY
Disable entropy update.
Definition vp8cx.h:133
#define VP8_EFLAG_NO_UPD_LAST
Don't update the last frame.
Definition vp8cx.h:98
#define VP8_EFLAG_NO_UPD_GF
Don't update the golden frame.
Definition vp8cx.h:105
#define VP8_EFLAG_NO_REF_GF
Don't reference the golden frame.
Definition vp8cx.h:83
@ VP8E_SET_MAX_INTRA_BITRATE_PCT
Codec control function to set Max data rate for Intra frames.
Definition vp8cx.h:275
@ VP8E_SET_NOISE_SENSITIVITY
control function to set noise sensitivity
Definition vp8cx.h:191
@ VP8E_SET_TOKEN_PARTITIONS
Codec control function to set the number of token partitions.
Definition vp8cx.h:212
@ VP8E_SET_CPUUSED
Codec control function to set encoder internal speed settings.
Definition vp8cx.h:173
@ VP8E_SET_TEMPORAL_LAYER_ID
Codec control function to set the temporal layer id.
Definition vp8cx.h:322
@ VP8E_SET_STATIC_THRESHOLD
Codec control function to set the threshold for MBs treated static.
Definition vp8cx.h:206
@ VP8E_SET_FRAME_FLAGS
Codec control function to set reference and update frame flags.
Definition vp8cx.h:281
Codec context structure.
Definition vpx_codec.h:200
Encoder output packet.
Definition vpx_encoder.h:167
enum vpx_codec_cx_pkt_kind kind
Definition vpx_encoder.h:168
double psnr[4]
Definition vpx_encoder.h:195
struct vpx_codec_cx_pkt::@1::@2 frame
size_t sz
Definition vpx_encoder.h:172
vpx_codec_pts_t pts
time stamp to show frame (in timebase units)
Definition vpx_encoder.h:174
union vpx_codec_cx_pkt::@1 data
Encoder configuration structure.
Definition vpx_encoder.h:276
unsigned int rc_resize_allowed
Enable/disable spatial resampling, if supported by the codec.
Definition vpx_encoder.h:408
unsigned int kf_min_dist
Keyframe minimum interval.
Definition vpx_encoder.h:615
unsigned int rc_min_quantizer
Minimum (Best Quality) Quantizer.
Definition vpx_encoder.h:483
unsigned int ts_number_layers
Number of temporal coding layers.
Definition vpx_encoder.h:654
unsigned int g_h
Height of the frame.
Definition vpx_encoder.h:321
enum vpx_kf_mode kf_mode
Keyframe placement mode.
Definition vpx_encoder.h:606
unsigned int ts_layer_id[16]
Template defining the membership of frames to temporal layers.
Definition vpx_encoder.h:686
vpx_codec_er_flags_t g_error_resilient
Enable error resilient modes.
Definition vpx_encoder.h:359
unsigned int ts_periodicity
Length of the sequence defining frame temporal layer membership.
Definition vpx_encoder.h:677
unsigned int rc_overshoot_pct
Rate control adaptation overshoot control.
Definition vpx_encoder.h:526
unsigned int g_w
Width of the frame.
Definition vpx_encoder.h:312
unsigned int rc_buf_sz
Decoder Buffer Size.
Definition vpx_encoder.h:541
unsigned int rc_dropframe_thresh
Temporal resampling configuration, if supported by the codec.
Definition vpx_encoder.h:399
struct vpx_rational g_timebase
Stream timebase units.
Definition vpx_encoder.h:351
enum vpx_enc_pass g_pass
Multi-pass Encoding Mode.
Definition vpx_encoder.h:366
unsigned int rc_max_quantizer
Maximum (Worst Quality) Quantizer.
Definition vpx_encoder.h:492
unsigned int g_lag_in_frames
Allow lagged encoding.
Definition vpx_encoder.h:380
enum vpx_rc_mode rc_end_usage
Rate control algorithm to use.
Definition vpx_encoder.h:448
unsigned int rc_buf_initial_sz
Decoder Buffer Initial Size.
Definition vpx_encoder.h:550
unsigned int rc_buf_optimal_sz
Decoder Buffer Optimal Size.
Definition vpx_encoder.h:559
unsigned int rc_target_bitrate
Target data rate.
Definition vpx_encoder.h:470
unsigned int ts_target_bitrate[5]
Target bitrate for each temporal layer.
Definition vpx_encoder.h:661
unsigned int rc_undershoot_pct
Rate control adaptation undershoot control.
Definition vpx_encoder.h:511
unsigned int ts_rate_decimator[5]
Frame rate decimation factor for each temporal layer.
Definition vpx_encoder.h:668
unsigned int kf_max_dist
Keyframe maximum interval.
Definition vpx_encoder.h:624
unsigned int g_threads
Maximum number of threads to use.
Definition vpx_encoder.h:293
Image Descriptor.
Definition vpx_image.h:76
vpx_img_fmt_t fmt
Definition vpx_image.h:77
unsigned int d_h
Definition vpx_image.h:88
unsigned int d_w
Definition vpx_image.h:87
unsigned int h
Definition vpx_image.h:83
unsigned char * planes[4]
Definition vpx_image.h:104
unsigned int w
Definition vpx_image.h:82
int stride[4]
Definition vpx_image.h:105
Rational Number.
Definition vpx_encoder.h:226
int den
Definition vpx_encoder.h:228
int num
Definition vpx_encoder.h:227
Provides definitions for using VP8 or VP9 encoder algorithm within the vpx Codec Interface.
Describes the encoder algorithm interface to applications.
#define VPX_PLANE_Y
Definition vpx_image.h:100
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.
#define VPX_PLANE_U
Definition vpx_image.h:101
@ VPX_IMG_FMT_YV12
Definition vpx_image.h:40
@ VPX_IMG_FMT_I420
Definition vpx_image.h:42
#define VPX_PLANE_V
Definition vpx_image.h:102
void vpx_img_free(vpx_image_t *img)
Close an image descriptor.