?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Removed vp9_recon functions
[aom.git] / vpxenc.c
blob8d0bf29299cbd649c03a6f5864872fd92122e48e
1 /*
2 * Copyright (c) 2010 The WebM project authors. All Rights Reserved.
3 *
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 */
10
11 #include "vpx_config.h"
12
13 #if defined(_WIN32) || defined(__OS2__) || !CONFIG_OS_SUPPORT
14 #define USE_POSIX_MMAP 0
15 #else
16 #define USE_POSIX_MMAP 1
17 #endif
18
19 #include <stdio.h>
20 #include <stdlib.h>
21 #include <stdarg.h>
22 #include <string.h>
23 #include <limits.h>
24 #include <assert.h>
25 #include "vpx/vpx_encoder.h"
26 #if CONFIG_DECODERS
27 #include "vpx/vpx_decoder.h"
28 #endif
29 #if USE_POSIX_MMAP
30 #include <sys/types.h>
31 #include <sys/stat.h>
32 #include <sys/mman.h>
33 #include <fcntl.h>
34 #include <unistd.h>
35 #endif
36
37 #if CONFIG_VP8_ENCODER || CONFIG_VP9_ENCODER
38 #include "vpx/vp8cx.h"
39 #endif
40 #if CONFIG_VP8_DECODER || CONFIG_VP9_DECODER
41 #include "vpx/vp8dx.h"
42 #endif
43
44 #include "vpx_ports/mem_ops.h"
45 #include "vpx_ports/vpx_timer.h"
46 #include "tools_common.h"
47 #include "y4minput.h"
48 #include "libmkv/EbmlWriter.h"
49 #include "libmkv/EbmlIDs.h"
50 #include "third_party/libyuv/include/libyuv/scale.h"
51
52 /* Need special handling of these functions on Windows */
53 #if defined(_MSC_VER)
54 /* MSVS doesn't define off_t, and uses _f{seek,tell}i64 */
55 typedef __int64 off_t;
56 #define fseeko _fseeki64
57 #define ftello _ftelli64
58 #elif defined(_WIN32)
59 /* MinGW defines off_t as long
60 and uses f{seek,tell}o64/off64_t for large files */
61 #define fseeko fseeko64
62 #define ftello ftello64
63 #define off_t off64_t
64 #endif
65
66 #define LITERALU64(hi,lo) ((((uint64_t)hi)<<32)|lo)
67
68 /* We should use 32-bit file operations in WebM file format
69 * when building ARM executable file (.axf) with RVCT */
70 #if !CONFIG_OS_SUPPORT
71 typedef long off_t;
72 #define fseeko fseek
73 #define ftello ftell
74 #endif
75
76 /* Swallow warnings about unused results of fread/fwrite */
77 static size_t wrap_fread(void *ptr, size_t size, size_t nmemb,
78 FILE *stream) {
79 return fread(ptr, size, nmemb, stream);
80 }
81 #define fread wrap_fread
82
83 static size_t wrap_fwrite(const void *ptr, size_t size, size_t nmemb,
84 FILE *stream) {
85 return fwrite(ptr, size, nmemb, stream);
86 }
87 #define fwrite wrap_fwrite
88
89
90 static const char *exec_name;
91
92 #define VP8_FOURCC (0x30385056)
93 #define VP9_FOURCC (0x30395056)
94 static const struct codec_item {
95 char const *name;
96 const vpx_codec_iface_t *(*iface)(void);
97 const vpx_codec_iface_t *(*dx_iface)(void);
98 unsigned int fourcc;
99 } codecs[] = {
100 #if CONFIG_VP8_ENCODER && CONFIG_VP8_DECODER
101 {"vp8", &vpx_codec_vp8_cx, &vpx_codec_vp8_dx, VP8_FOURCC},
102 #elif CONFIG_VP8_ENCODER && !CONFIG_VP8_DECODER
103 {"vp8", &vpx_codec_vp8_cx, NULL, VP8_FOURCC},
104 #endif
105 #if CONFIG_VP9_ENCODER && CONFIG_VP9_DECODER
106 {"vp9", &vpx_codec_vp9_cx, &vpx_codec_vp9_dx, VP9_FOURCC},
107 #elif CONFIG_VP9_ENCODER && !CONFIG_VP9_DECODER
108 {"vp9", &vpx_codec_vp9_cx, NULL, VP9_FOURCC},
109 #endif
110 };
111
112 static void usage_exit();
113
114 #define LOG_ERROR(label) do \
115 {\
116 const char *l=label;\
117 va_list ap;\
118 va_start(ap, fmt);\
119 if(l)\
120 fprintf(stderr, "%s: ", l);\
121 vfprintf(stderr, fmt, ap);\
122 fprintf(stderr, "\n");\
123 va_end(ap);\
124 } while(0)
125
126 void die(const char *fmt, ...) {
127 LOG_ERROR(NULL);
128 usage_exit();
129 }
130
131
132 void fatal(const char *fmt, ...) {
133 LOG_ERROR("Fatal");
134 exit(EXIT_FAILURE);
135 }
136
137
138 void warn(const char *fmt, ...) {
139 LOG_ERROR("Warning");
140 }
141
142
143 static void warn_or_exit_on_errorv(vpx_codec_ctx_t *ctx, int fatal,
144 const char *s, va_list ap) {
145 if (ctx->err) {
146 const char *detail = vpx_codec_error_detail(ctx);
147
148 vfprintf(stderr, s, ap);
149 fprintf(stderr, ": %s\n", vpx_codec_error(ctx));
150
151 if (detail)
152 fprintf(stderr, " %s\n", detail);
153
154 if (fatal)
155 exit(EXIT_FAILURE);
156 }
157 }
158
159 static void ctx_exit_on_error(vpx_codec_ctx_t *ctx, const char *s, ...) {
160 va_list ap;
161
162 va_start(ap, s);
163 warn_or_exit_on_errorv(ctx, 1, s, ap);
164 va_end(ap);
165 }
166
167 static void warn_or_exit_on_error(vpx_codec_ctx_t *ctx, int fatal,
168 const char *s, ...) {
169 va_list ap;
170
171 va_start(ap, s);
172 warn_or_exit_on_errorv(ctx, fatal, s, ap);
173 va_end(ap);
174 }
175
176 /* This structure is used to abstract the different ways of handling
177 * first pass statistics.
178 */
179 typedef struct {
180 vpx_fixed_buf_t buf;
181 int pass;
182 FILE *file;
183 char *buf_ptr;
184 size_t buf_alloc_sz;
185 } stats_io_t;
186
187 int stats_open_file(stats_io_t *stats, const char *fpf, int pass) {
188 int res;
189
190 stats->pass = pass;
191
192 if (pass == 0) {
193 stats->file = fopen(fpf, "wb");
194 stats->buf.sz = 0;
195 stats->buf.buf = NULL,
196 res = (stats->file != NULL);
197 } else {
198 #if 0
199 #elif USE_POSIX_MMAP
200 struct stat stat_buf;
201 int fd;
202
203 fd = open(fpf, O_RDONLY);
204 stats->file = fdopen(fd, "rb");
205 fstat(fd, &stat_buf);
206 stats->buf.sz = stat_buf.st_size;
207 stats->buf.buf = mmap(NULL, stats->buf.sz, PROT_READ, MAP_PRIVATE,
208 fd, 0);
209 res = (stats->buf.buf != NULL);
210 #else
211 size_t nbytes;
212
213 stats->file = fopen(fpf, "rb");
214
215 if (fseek(stats->file, 0, SEEK_END))
216 fatal("First-pass stats file must be seekable!");
217
218 stats->buf.sz = stats->buf_alloc_sz = ftell(stats->file);
219 rewind(stats->file);
220
221 stats->buf.buf = malloc(stats->buf_alloc_sz);
222
223 if (!stats->buf.buf)
224 fatal("Failed to allocate first-pass stats buffer (%lu bytes)",
225 (unsigned long)stats->buf_alloc_sz);
226
227 nbytes = fread(stats->buf.buf, 1, stats->buf.sz, stats->file);
228 res = (nbytes == stats->buf.sz);
229 #endif
230 }
231
232 return res;
233 }
234
235 int stats_open_mem(stats_io_t *stats, int pass) {
236 int res;
237 stats->pass = pass;
238
239 if (!pass) {
240 stats->buf.sz = 0;
241 stats->buf_alloc_sz = 64 * 1024;
242 stats->buf.buf = malloc(stats->buf_alloc_sz);
243 }
244
245 stats->buf_ptr = stats->buf.buf;
246 res = (stats->buf.buf != NULL);
247 return res;
248 }
249
250
251 void stats_close(stats_io_t *stats, int last_pass) {
252 if (stats->file) {
253 if (stats->pass == last_pass) {
254 #if 0
255 #elif USE_POSIX_MMAP
256 munmap(stats->buf.buf, stats->buf.sz);
257 #else
258 free(stats->buf.buf);
259 #endif
260 }
261
262 fclose(stats->file);
263 stats->file = NULL;
264 } else {
265 if (stats->pass == last_pass)
266 free(stats->buf.buf);
267 }
268 }
269
270 void stats_write(stats_io_t *stats, const void *pkt, size_t len) {
271 if (stats->file) {
272 (void) fwrite(pkt, 1, len, stats->file);
273 } else {
274 if (stats->buf.sz + len > stats->buf_alloc_sz) {
275 size_t new_sz = stats->buf_alloc_sz + 64 * 1024;
276 char *new_ptr = realloc(stats->buf.buf, new_sz);
277
278 if (new_ptr) {
279 stats->buf_ptr = new_ptr + (stats->buf_ptr - (char *)stats->buf.buf);
280 stats->buf.buf = new_ptr;
281 stats->buf_alloc_sz = new_sz;
282 } else
283 fatal("Failed to realloc firstpass stats buffer.");
284 }
285
286 memcpy(stats->buf_ptr, pkt, len);
287 stats->buf.sz += len;
288 stats->buf_ptr += len;
289 }
290 }
291
292 vpx_fixed_buf_t stats_get(stats_io_t *stats) {
293 return stats->buf;
294 }
295
296 /* Stereo 3D packed frame format */
297 typedef enum stereo_format {
298 STEREO_FORMAT_MONO = 0,
299 STEREO_FORMAT_LEFT_RIGHT = 1,
300 STEREO_FORMAT_BOTTOM_TOP = 2,
301 STEREO_FORMAT_TOP_BOTTOM = 3,
302 STEREO_FORMAT_RIGHT_LEFT = 11
303 } stereo_format_t;
304
305 enum video_file_type {
306 FILE_TYPE_RAW,
307 FILE_TYPE_IVF,
308 FILE_TYPE_Y4M
309 };
310
311 struct detect_buffer {
312 char buf[4];
313 size_t buf_read;
314 size_t position;
315 };
316
317
318 struct input_state {
319 char *fn;
320 FILE *file;
321 off_t length;
322 y4m_input y4m;
323 struct detect_buffer detect;
324 enum video_file_type file_type;
325 unsigned int w;
326 unsigned int h;
327 struct vpx_rational framerate;
328 int use_i420;
329 int only_i420;
330 };
331
332
333 #define IVF_FRAME_HDR_SZ (4+8) /* 4 byte size + 8 byte timestamp */
334 static int read_frame(struct input_state *input, vpx_image_t *img) {
335 FILE *f = input->file;
336 enum video_file_type file_type = input->file_type;
337 y4m_input *y4m = &input->y4m;
338 struct detect_buffer *detect = &input->detect;
339 int plane = 0;
340 int shortread = 0;
341
342 if (file_type == FILE_TYPE_Y4M) {
343 if (y4m_input_fetch_frame(y4m, f, img) < 1)
344 return 0;
345 } else {
346 if (file_type == FILE_TYPE_IVF) {
347 char junk[IVF_FRAME_HDR_SZ];
348
349 /* Skip the frame header. We know how big the frame should be. See
350 * write_ivf_frame_header() for documentation on the frame header
351 * layout.
352 */
353 (void) fread(junk, 1, IVF_FRAME_HDR_SZ, f);
354 }
355
356 for (plane = 0; plane < 3; plane++) {
357 unsigned char *ptr;
358 int w = (plane ? (1 + img->d_w) / 2 : img->d_w);
359 int h = (plane ? (1 + img->d_h) / 2 : img->d_h);
360 int r;
361
362 /* Determine the correct plane based on the image format. The for-loop
363 * always counts in Y,U,V order, but this may not match the order of
364 * the data on disk.
365 */
366 switch (plane) {
367 case 1:
368 ptr = img->planes[img->fmt == VPX_IMG_FMT_YV12 ? VPX_PLANE_V : VPX_PLANE_U];
369 break;
370 case 2:
371 ptr = img->planes[img->fmt == VPX_IMG_FMT_YV12 ? VPX_PLANE_U : VPX_PLANE_V];
372 break;
373 default:
374 ptr = img->planes[plane];
375 }
376
377 for (r = 0; r < h; r++) {
378 size_t needed = w;
379 size_t buf_position = 0;
380 const size_t left = detect->buf_read - detect->position;
381 if (left > 0) {
382 const size_t more = (left < needed) ? left : needed;
383 memcpy(ptr, detect->buf + detect->position, more);
384 buf_position = more;
385 needed -= more;
386 detect->position += more;
387 }
388 if (needed > 0) {
389 shortread |= (fread(ptr + buf_position, 1, needed, f) < needed);
390 }
391
392 ptr += img->stride[plane];
393 }
394 }
395 }
396
397 return !shortread;
398 }
399
400
401 unsigned int file_is_y4m(FILE *infile,
402 y4m_input *y4m,
403 char detect[4]) {
404 if (memcmp(detect, "YUV4", 4) == 0) {
405 return 1;
406 }
407 return 0;
408 }
409
410 #define IVF_FILE_HDR_SZ (32)
411 unsigned int file_is_ivf(struct input_state *input,
412 unsigned int *fourcc) {
413 char raw_hdr[IVF_FILE_HDR_SZ];
414 int is_ivf = 0;
415 FILE *infile = input->file;
416 unsigned int *width = &input->w;
417 unsigned int *height = &input->h;
418 struct detect_buffer *detect = &input->detect;
419
420 if (memcmp(detect->buf, "DKIF", 4) != 0)
421 return 0;
422
423 /* See write_ivf_file_header() for more documentation on the file header
424 * layout.
425 */
426 if (fread(raw_hdr + 4, 1, IVF_FILE_HDR_SZ - 4, infile)
427 == IVF_FILE_HDR_SZ - 4) {
428 {
429 is_ivf = 1;
430
431 if (mem_get_le16(raw_hdr + 4) != 0)
432 warn("Unrecognized IVF version! This file may not decode "
433 "properly.");
434
435 *fourcc = mem_get_le32(raw_hdr + 8);
436 }
437 }
438
439 if (is_ivf) {
440 *width = mem_get_le16(raw_hdr + 12);
441 *height = mem_get_le16(raw_hdr + 14);
442 detect->position = 4;
443 }
444
445 return is_ivf;
446 }
447
448
449 static void write_ivf_file_header(FILE *outfile,
450 const vpx_codec_enc_cfg_t *cfg,
451 unsigned int fourcc,
452 int frame_cnt) {
453 char header[32];
454
455 if (cfg->g_pass != VPX_RC_ONE_PASS && cfg->g_pass != VPX_RC_LAST_PASS)
456 return;
457
458 header[0] = 'D';
459 header[1] = 'K';
460 header[2] = 'I';
461 header[3] = 'F';
462 mem_put_le16(header + 4, 0); /* version */
463 mem_put_le16(header + 6, 32); /* headersize */
464 mem_put_le32(header + 8, fourcc); /* headersize */
465 mem_put_le16(header + 12, cfg->g_w); /* width */
466 mem_put_le16(header + 14, cfg->g_h); /* height */
467 mem_put_le32(header + 16, cfg->g_timebase.den); /* rate */
468 mem_put_le32(header + 20, cfg->g_timebase.num); /* scale */
469 mem_put_le32(header + 24, frame_cnt); /* length */
470 mem_put_le32(header + 28, 0); /* unused */
471
472 (void) fwrite(header, 1, 32, outfile);
473 }
474
475
476 static void write_ivf_frame_header(FILE *outfile,
477 const vpx_codec_cx_pkt_t *pkt) {
478 char header[12];
479 vpx_codec_pts_t pts;
480
481 if (pkt->kind != VPX_CODEC_CX_FRAME_PKT)
482 return;
483
484 pts = pkt->data.frame.pts;
485 mem_put_le32(header, (int)pkt->data.frame.sz);
486 mem_put_le32(header + 4, pts & 0xFFFFFFFF);
487 mem_put_le32(header + 8, pts >> 32);
488
489 (void) fwrite(header, 1, 12, outfile);
490 }
491
492 static void write_ivf_frame_size(FILE *outfile, size_t size) {
493 char header[4];
494 mem_put_le32(header, (int)size);
495 (void) fwrite(header, 1, 4, outfile);
496 }
497
498
499 typedef off_t EbmlLoc;
500
501
502 struct cue_entry {
503 unsigned int time;
504 uint64_t loc;
505 };
506
507
508 struct EbmlGlobal {
509 int debug;
510
511 FILE *stream;
512 int64_t last_pts_ms;
513 vpx_rational_t framerate;
514
515 /* These pointers are to the start of an element */
516 off_t position_reference;
517 off_t seek_info_pos;
518 off_t segment_info_pos;
519 off_t track_pos;
520 off_t cue_pos;
521 off_t cluster_pos;
522
523 /* This pointer is to a specific element to be serialized */
524 off_t track_id_pos;
525
526 /* These pointers are to the size field of the element */
527 EbmlLoc startSegment;
528 EbmlLoc startCluster;
529
530 uint32_t cluster_timecode;
531 int cluster_open;
532
533 struct cue_entry *cue_list;
534 unsigned int cues;
535
536 };
537
538
539 void Ebml_Write(EbmlGlobal *glob, const void *buffer_in, unsigned long len) {
540 (void) fwrite(buffer_in, 1, len, glob->stream);
541 }
542
543 #define WRITE_BUFFER(s) \
544 for(i = len-1; i>=0; i--)\
545 { \
546 x = (char)(*(const s *)buffer_in >> (i * CHAR_BIT)); \
547 Ebml_Write(glob, &x, 1); \
548 }
549 void Ebml_Serialize(EbmlGlobal *glob, const void *buffer_in, int buffer_size, unsigned long len) {
550 char x;
551 int i;
552
553 /* buffer_size:
554 * 1 - int8_t;
555 * 2 - int16_t;
556 * 3 - int32_t;
557 * 4 - int64_t;
558 */
559 switch (buffer_size) {
560 case 1:
561 WRITE_BUFFER(int8_t)
562 break;
563 case 2:
564 WRITE_BUFFER(int16_t)
565 break;
566 case 4:
567 WRITE_BUFFER(int32_t)
568 break;
569 case 8:
570 WRITE_BUFFER(int64_t)
571 break;
572 default:
573 break;
574 }
575 }
576 #undef WRITE_BUFFER
577
578 /* Need a fixed size serializer for the track ID. libmkv provides a 64 bit
579 * one, but not a 32 bit one.
580 */
581 static void Ebml_SerializeUnsigned32(EbmlGlobal *glob, unsigned long class_id, uint64_t ui) {
582 unsigned char sizeSerialized = 4 | 0x80;
583 Ebml_WriteID(glob, class_id);
584 Ebml_Serialize(glob, &sizeSerialized, sizeof(sizeSerialized), 1);
585 Ebml_Serialize(glob, &ui, sizeof(ui), 4);
586 }
587
588
589 static void
590 Ebml_StartSubElement(EbmlGlobal *glob, EbmlLoc *ebmlLoc,
591 unsigned long class_id) {
592 /* todo this is always taking 8 bytes, this may need later optimization */
593 /* this is a key that says length unknown */
594 uint64_t unknownLen = LITERALU64(0x01FFFFFF, 0xFFFFFFFF);
595
596 Ebml_WriteID(glob, class_id);
597 *ebmlLoc = ftello(glob->stream);
598 Ebml_Serialize(glob, &unknownLen, sizeof(unknownLen), 8);
599 }
600
601 static void
602 Ebml_EndSubElement(EbmlGlobal *glob, EbmlLoc *ebmlLoc) {
603 off_t pos;
604 uint64_t size;
605
606 /* Save the current stream pointer */
607 pos = ftello(glob->stream);
608
609 /* Calculate the size of this element */
610 size = pos - *ebmlLoc - 8;
611 size |= LITERALU64(0x01000000, 0x00000000);
612
613 /* Seek back to the beginning of the element and write the new size */
614 fseeko(glob->stream, *ebmlLoc, SEEK_SET);
615 Ebml_Serialize(glob, &size, sizeof(size), 8);
616
617 /* Reset the stream pointer */
618 fseeko(glob->stream, pos, SEEK_SET);
619 }
620
621
622 static void
623 write_webm_seek_element(EbmlGlobal *ebml, unsigned long id, off_t pos) {
624 uint64_t offset = pos - ebml->position_reference;
625 EbmlLoc start;
626 Ebml_StartSubElement(ebml, &start, Seek);
627 Ebml_SerializeBinary(ebml, SeekID, id);
628 Ebml_SerializeUnsigned64(ebml, SeekPosition, offset);
629 Ebml_EndSubElement(ebml, &start);
630 }
631
632
633 static void
634 write_webm_seek_info(EbmlGlobal *ebml) {
635
636 off_t pos;
637
638 /* Save the current stream pointer */
639 pos = ftello(ebml->stream);
640
641 if (ebml->seek_info_pos)
642 fseeko(ebml->stream, ebml->seek_info_pos, SEEK_SET);
643 else
644 ebml->seek_info_pos = pos;
645
646 {
647 EbmlLoc start;
648
649 Ebml_StartSubElement(ebml, &start, SeekHead);
650 write_webm_seek_element(ebml, Tracks, ebml->track_pos);
651 write_webm_seek_element(ebml, Cues, ebml->cue_pos);
652 write_webm_seek_element(ebml, Info, ebml->segment_info_pos);
653 Ebml_EndSubElement(ebml, &start);
654 }
655 {
656 /* segment info */
657 EbmlLoc startInfo;
658 uint64_t frame_time;
659 char version_string[64];
660
661 /* Assemble version string */
662 if (ebml->debug)
663 strcpy(version_string, "vpxenc");
664 else {
665 strcpy(version_string, "vpxenc ");
666 strncat(version_string,
667 vpx_codec_version_str(),
668 sizeof(version_string) - 1 - strlen(version_string));
669 }
670
671 frame_time = (uint64_t)1000 * ebml->framerate.den
672 / ebml->framerate.num;
673 ebml->segment_info_pos = ftello(ebml->stream);
674 Ebml_StartSubElement(ebml, &startInfo, Info);
675 Ebml_SerializeUnsigned(ebml, TimecodeScale, 1000000);
676 Ebml_SerializeFloat(ebml, Segment_Duration,
677 (double)(ebml->last_pts_ms + frame_time));
678 Ebml_SerializeString(ebml, 0x4D80, version_string);
679 Ebml_SerializeString(ebml, 0x5741, version_string);
680 Ebml_EndSubElement(ebml, &startInfo);
681 }
682 }
683
684
685 static void
686 write_webm_file_header(EbmlGlobal *glob,
687 const vpx_codec_enc_cfg_t *cfg,
688 const struct vpx_rational *fps,
689 stereo_format_t stereo_fmt,
690 unsigned int fourcc) {
691 {
692 EbmlLoc start;
693 Ebml_StartSubElement(glob, &start, EBML);
694 Ebml_SerializeUnsigned(glob, EBMLVersion, 1);
695 Ebml_SerializeUnsigned(glob, EBMLReadVersion, 1);
696 Ebml_SerializeUnsigned(glob, EBMLMaxIDLength, 4);
697 Ebml_SerializeUnsigned(glob, EBMLMaxSizeLength, 8);
698 Ebml_SerializeString(glob, DocType, "webm");
699 Ebml_SerializeUnsigned(glob, DocTypeVersion, 2);
700 Ebml_SerializeUnsigned(glob, DocTypeReadVersion, 2);
701 Ebml_EndSubElement(glob, &start);
702 }
703 {
704 Ebml_StartSubElement(glob, &glob->startSegment, Segment);
705 glob->position_reference = ftello(glob->stream);
706 glob->framerate = *fps;
707 write_webm_seek_info(glob);
708
709 {
710 EbmlLoc trackStart;
711 glob->track_pos = ftello(glob->stream);
712 Ebml_StartSubElement(glob, &trackStart, Tracks);
713 {
714 unsigned int trackNumber = 1;
715 uint64_t trackID = 0;
716
717 EbmlLoc start;
718 Ebml_StartSubElement(glob, &start, TrackEntry);
719 Ebml_SerializeUnsigned(glob, TrackNumber, trackNumber);
720 glob->track_id_pos = ftello(glob->stream);
721 Ebml_SerializeUnsigned32(glob, TrackUID, trackID);
722 Ebml_SerializeUnsigned(glob, TrackType, 1);
723 Ebml_SerializeString(glob, CodecID,
724 fourcc == VP8_FOURCC ? "V_VP8" : "V_VP9");
725 {
726 unsigned int pixelWidth = cfg->g_w;
727 unsigned int pixelHeight = cfg->g_h;
728 float frameRate = (float)fps->num / (float)fps->den;
729
730 EbmlLoc videoStart;
731 Ebml_StartSubElement(glob, &videoStart, Video);
732 Ebml_SerializeUnsigned(glob, PixelWidth, pixelWidth);
733 Ebml_SerializeUnsigned(glob, PixelHeight, pixelHeight);
734 Ebml_SerializeUnsigned(glob, StereoMode, stereo_fmt);
735 Ebml_SerializeFloat(glob, FrameRate, frameRate);
736 Ebml_EndSubElement(glob, &videoStart);
737 }
738 Ebml_EndSubElement(glob, &start); /* Track Entry */
739 }
740 Ebml_EndSubElement(glob, &trackStart);
741 }
742 /* segment element is open */
743 }
744 }
745
746
747 static void
748 write_webm_block(EbmlGlobal *glob,
749 const vpx_codec_enc_cfg_t *cfg,
750 const vpx_codec_cx_pkt_t *pkt) {
751 unsigned long block_length;
752 unsigned char track_number;
753 unsigned short block_timecode = 0;
754 unsigned char flags;
755 int64_t pts_ms;
756 int start_cluster = 0, is_keyframe;
757
758 /* Calculate the PTS of this frame in milliseconds */
759 pts_ms = pkt->data.frame.pts * 1000
760 * (uint64_t)cfg->g_timebase.num / (uint64_t)cfg->g_timebase.den;
761 if (pts_ms <= glob->last_pts_ms)
762 pts_ms = glob->last_pts_ms + 1;
763 glob->last_pts_ms = pts_ms;
764
765 /* Calculate the relative time of this block */
766 if (pts_ms - glob->cluster_timecode > SHRT_MAX)
767 start_cluster = 1;
768 else
769 block_timecode = (unsigned short)pts_ms - glob->cluster_timecode;
770
771 is_keyframe = (pkt->data.frame.flags & VPX_FRAME_IS_KEY);
772 if (start_cluster || is_keyframe) {
773 if (glob->cluster_open)
774 Ebml_EndSubElement(glob, &glob->startCluster);
775
776 /* Open the new cluster */
777 block_timecode = 0;
778 glob->cluster_open = 1;
779 glob->cluster_timecode = (uint32_t)pts_ms;
780 glob->cluster_pos = ftello(glob->stream);
781 Ebml_StartSubElement(glob, &glob->startCluster, Cluster); /* cluster */
782 Ebml_SerializeUnsigned(glob, Timecode, glob->cluster_timecode);
783
784 /* Save a cue point if this is a keyframe. */
785 if (is_keyframe) {
786 struct cue_entry *cue, *new_cue_list;
787
788 new_cue_list = realloc(glob->cue_list,
789 (glob->cues + 1) * sizeof(struct cue_entry));
790 if (new_cue_list)
791 glob->cue_list = new_cue_list;
792 else
793 fatal("Failed to realloc cue list.");
794
795 cue = &glob->cue_list[glob->cues];
796 cue->time = glob->cluster_timecode;
797 cue->loc = glob->cluster_pos;
798 glob->cues++;
799 }
800 }
801
802 /* Write the Simple Block */
803 Ebml_WriteID(glob, SimpleBlock);
804
805 block_length = (unsigned long)pkt->data.frame.sz + 4;
806 block_length |= 0x10000000;
807 Ebml_Serialize(glob, &block_length, sizeof(block_length), 4);
808
809 track_number = 1;
810 track_number |= 0x80;
811 Ebml_Write(glob, &track_number, 1);
812
813 Ebml_Serialize(glob, &block_timecode, sizeof(block_timecode), 2);
814
815 flags = 0;
816 if (is_keyframe)
817 flags |= 0x80;
818 if (pkt->data.frame.flags & VPX_FRAME_IS_INVISIBLE)
819 flags |= 0x08;
820 Ebml_Write(glob, &flags, 1);
821
822 Ebml_Write(glob, pkt->data.frame.buf, (unsigned long)pkt->data.frame.sz);
823 }
824
825
826 static void
827 write_webm_file_footer(EbmlGlobal *glob, long hash) {
828
829 if (glob->cluster_open)
830 Ebml_EndSubElement(glob, &glob->startCluster);
831
832 {
833 EbmlLoc start;
834 unsigned int i;
835
836 glob->cue_pos = ftello(glob->stream);
837 Ebml_StartSubElement(glob, &start, Cues);
838 for (i = 0; i < glob->cues; i++) {
839 struct cue_entry *cue = &glob->cue_list[i];
840 EbmlLoc start;
841
842 Ebml_StartSubElement(glob, &start, CuePoint);
843 {
844 EbmlLoc start;
845
846 Ebml_SerializeUnsigned(glob, CueTime, cue->time);
847
848 Ebml_StartSubElement(glob, &start, CueTrackPositions);
849 Ebml_SerializeUnsigned(glob, CueTrack, 1);
850 Ebml_SerializeUnsigned64(glob, CueClusterPosition,
851 cue->loc - glob->position_reference);
852 Ebml_EndSubElement(glob, &start);
853 }
854 Ebml_EndSubElement(glob, &start);
855 }
856 Ebml_EndSubElement(glob, &start);
857 }
858
859 Ebml_EndSubElement(glob, &glob->startSegment);
860
861 /* Patch up the seek info block */
862 write_webm_seek_info(glob);
863
864 /* Patch up the track id */
865 fseeko(glob->stream, glob->track_id_pos, SEEK_SET);
866 Ebml_SerializeUnsigned32(glob, TrackUID, glob->debug ? 0xDEADBEEF : hash);
867
868 fseeko(glob->stream, 0, SEEK_END);
869 }
870
871
872 /* Murmur hash derived from public domain reference implementation at
873 * http:// sites.google.com/site/murmurhash/
874 */
875 static unsigned int murmur(const void *key, int len, unsigned int seed) {
876 const unsigned int m = 0x5bd1e995;
877 const int r = 24;
878
879 unsigned int h = seed ^ len;
880
881 const unsigned char *data = (const unsigned char *)key;
882
883 while (len >= 4) {
884 unsigned int k;
885
886 k = data[0];
887 k |= data[1] << 8;
888 k |= data[2] << 16;
889 k |= data[3] << 24;
890
891 k *= m;
892 k ^= k >> r;
893 k *= m;
894
895 h *= m;
896 h ^= k;
897
898 data += 4;
899 len -= 4;
900 }
901
902 switch (len) {
903 case 3:
904 h ^= data[2] << 16;
905 case 2:
906 h ^= data[1] << 8;
907 case 1:
908 h ^= data[0];
909 h *= m;
910 };
911
912 h ^= h >> 13;
913 h *= m;
914 h ^= h >> 15;
915
916 return h;
917 }
918
919 #include "math.h"
920 #define MAX_PSNR 100
921 static double vp8_mse2psnr(double Samples, double Peak, double Mse) {
922 double psnr;
923
924 if ((double)Mse > 0.0)
925 psnr = 10.0 * log10(Peak * Peak * Samples / Mse);
926 else
927 psnr = MAX_PSNR; /* Limit to prevent / 0 */
928
929 if (psnr > MAX_PSNR)
930 psnr = MAX_PSNR;
931
932 return psnr;
933 }
934
935
936 #include "args.h"
937 static const arg_def_t debugmode = ARG_DEF("D", "debug", 0,
938 "Debug mode (makes output deterministic)");
939 static const arg_def_t outputfile = ARG_DEF("o", "output", 1,
940 "Output filename");
941 static const arg_def_t use_yv12 = ARG_DEF(NULL, "yv12", 0,
942 "Input file is YV12 ");
943 static const arg_def_t use_i420 = ARG_DEF(NULL, "i420", 0,
944 "Input file is I420 (default)");
945 static const arg_def_t codecarg = ARG_DEF(NULL, "codec", 1,
946 "Codec to use");
947 static const arg_def_t passes = ARG_DEF("p", "passes", 1,
948 "Number of passes (1/2)");
949 static const arg_def_t pass_arg = ARG_DEF(NULL, "pass", 1,
950 "Pass to execute (1/2)");
951 static const arg_def_t fpf_name = ARG_DEF(NULL, "fpf", 1,
952 "First pass statistics file name");
953 static const arg_def_t limit = ARG_DEF(NULL, "limit", 1,
954 "Stop encoding after n input frames");
955 static const arg_def_t skip = ARG_DEF(NULL, "skip", 1,
956 "Skip the first n input frames");
957 static const arg_def_t deadline = ARG_DEF("d", "deadline", 1,
958 "Deadline per frame (usec)");
959 static const arg_def_t best_dl = ARG_DEF(NULL, "best", 0,
960 "Use Best Quality Deadline");
961 static const arg_def_t good_dl = ARG_DEF(NULL, "good", 0,
962 "Use Good Quality Deadline");
963 static const arg_def_t rt_dl = ARG_DEF(NULL, "rt", 0,
964 "Use Realtime Quality Deadline");
965 static const arg_def_t quietarg = ARG_DEF("q", "quiet", 0,
966 "Do not print encode progress");
967 static const arg_def_t verbosearg = ARG_DEF("v", "verbose", 0,
968 "Show encoder parameters");
969 static const arg_def_t psnrarg = ARG_DEF(NULL, "psnr", 0,
970 "Show PSNR in status line");
971 enum TestDecodeFatality {
972 TEST_DECODE_OFF,
973 TEST_DECODE_FATAL,
974 TEST_DECODE_WARN,
975 };
976 static const struct arg_enum_list test_decode_enum[] = {
977 {"off", TEST_DECODE_OFF},
978 {"fatal", TEST_DECODE_FATAL},
979 {"warn", TEST_DECODE_WARN},
980 {NULL, 0}
981 };
982 static const arg_def_t recontest = ARG_DEF_ENUM(NULL, "test-decode", 1,
983 "Test encode/decode mismatch",
984 test_decode_enum);
985 static const arg_def_t framerate = ARG_DEF(NULL, "fps", 1,
986 "Stream frame rate (rate/scale)");
987 static const arg_def_t use_ivf = ARG_DEF(NULL, "ivf", 0,
988 "Output IVF (default is WebM)");
989 static const arg_def_t out_part = ARG_DEF("P", "output-partitions", 0,
990 "Makes encoder output partitions. Requires IVF output!");
991 static const arg_def_t q_hist_n = ARG_DEF(NULL, "q-hist", 1,
992 "Show quantizer histogram (n-buckets)");
993 static const arg_def_t rate_hist_n = ARG_DEF(NULL, "rate-hist", 1,
994 "Show rate histogram (n-buckets)");
995 static const arg_def_t *main_args[] = {
996 &debugmode,
997 &outputfile, &codecarg, &passes, &pass_arg, &fpf_name, &limit, &skip,
998 &deadline, &best_dl, &good_dl, &rt_dl,
999 &quietarg, &verbosearg, &psnrarg, &use_ivf, &out_part, &q_hist_n, &rate_hist_n,
1000 NULL
1001 };
1002
1003 static const arg_def_t usage = ARG_DEF("u", "usage", 1,
1004 "Usage profile number to use");
1005 static const arg_def_t threads = ARG_DEF("t", "threads", 1,
1006 "Max number of threads to use");
1007 static const arg_def_t profile = ARG_DEF(NULL, "profile", 1,
1008 "Bitstream profile number to use");
1009 static const arg_def_t width = ARG_DEF("w", "width", 1,
1010 "Frame width");
1011 static const arg_def_t height = ARG_DEF("h", "height", 1,
1012 "Frame height");
1013 static const struct arg_enum_list stereo_mode_enum[] = {
1014 {"mono", STEREO_FORMAT_MONO},
1015 {"left-right", STEREO_FORMAT_LEFT_RIGHT},
1016 {"bottom-top", STEREO_FORMAT_BOTTOM_TOP},
1017 {"top-bottom", STEREO_FORMAT_TOP_BOTTOM},
1018 {"right-left", STEREO_FORMAT_RIGHT_LEFT},
1019 {NULL, 0}
1020 };
1021 static const arg_def_t stereo_mode = ARG_DEF_ENUM(NULL, "stereo-mode", 1,
1022 "Stereo 3D video format", stereo_mode_enum);
1023 static const arg_def_t timebase = ARG_DEF(NULL, "timebase", 1,
1024 "Output timestamp precision (fractional seconds)");
1025 static const arg_def_t error_resilient = ARG_DEF(NULL, "error-resilient", 1,
1026 "Enable error resiliency features");
1027 static const arg_def_t lag_in_frames = ARG_DEF(NULL, "lag-in-frames", 1,
1028 "Max number of frames to lag");
1029
1030 static const arg_def_t *global_args[] = {
1031 &use_yv12, &use_i420, &usage, &threads, &profile,
1032 &width, &height, &stereo_mode, &timebase, &framerate,
1033 &error_resilient,
1034 &lag_in_frames, NULL
1035 };
1036
1037 static const arg_def_t dropframe_thresh = ARG_DEF(NULL, "drop-frame", 1,
1038 "Temporal resampling threshold (buf %)");
1039 static const arg_def_t resize_allowed = ARG_DEF(NULL, "resize-allowed", 1,
1040 "Spatial resampling enabled (bool)");
1041 static const arg_def_t resize_up_thresh = ARG_DEF(NULL, "resize-up", 1,
1042 "Upscale threshold (buf %)");
1043 static const arg_def_t resize_down_thresh = ARG_DEF(NULL, "resize-down", 1,
1044 "Downscale threshold (buf %)");
1045 static const struct arg_enum_list end_usage_enum[] = {
1046 {"vbr", VPX_VBR},
1047 {"cbr", VPX_CBR},
1048 {"cq", VPX_CQ},
1049 {NULL, 0}
1050 };
1051 static const arg_def_t end_usage = ARG_DEF_ENUM(NULL, "end-usage", 1,
1052 "Rate control mode", end_usage_enum);
1053 static const arg_def_t target_bitrate = ARG_DEF(NULL, "target-bitrate", 1,
1054 "Bitrate (kbps)");
1055 static const arg_def_t min_quantizer = ARG_DEF(NULL, "min-q", 1,
1056 "Minimum (best) quantizer");
1057 static const arg_def_t max_quantizer = ARG_DEF(NULL, "max-q", 1,
1058 "Maximum (worst) quantizer");
1059 static const arg_def_t undershoot_pct = ARG_DEF(NULL, "undershoot-pct", 1,
1060 "Datarate undershoot (min) target (%)");
1061 static const arg_def_t overshoot_pct = ARG_DEF(NULL, "overshoot-pct", 1,
1062 "Datarate overshoot (max) target (%)");
1063 static const arg_def_t buf_sz = ARG_DEF(NULL, "buf-sz", 1,
1064 "Client buffer size (ms)");
1065 static const arg_def_t buf_initial_sz = ARG_DEF(NULL, "buf-initial-sz", 1,
1066 "Client initial buffer size (ms)");
1067 static const arg_def_t buf_optimal_sz = ARG_DEF(NULL, "buf-optimal-sz", 1,
1068 "Client optimal buffer size (ms)");
1069 static const arg_def_t *rc_args[] = {
1070 &dropframe_thresh, &resize_allowed, &resize_up_thresh, &resize_down_thresh,
1071 &end_usage, &target_bitrate, &min_quantizer, &max_quantizer,
1072 &undershoot_pct, &overshoot_pct, &buf_sz, &buf_initial_sz, &buf_optimal_sz,
1073 NULL
1074 };
1075
1076
1077 static const arg_def_t bias_pct = ARG_DEF(NULL, "bias-pct", 1,
1078 "CBR/VBR bias (0=CBR, 100=VBR)");
1079 static const arg_def_t minsection_pct = ARG_DEF(NULL, "minsection-pct", 1,
1080 "GOP min bitrate (% of target)");
1081 static const arg_def_t maxsection_pct = ARG_DEF(NULL, "maxsection-pct", 1,
1082 "GOP max bitrate (% of target)");
1083 static const arg_def_t *rc_twopass_args[] = {
1084 &bias_pct, &minsection_pct, &maxsection_pct, NULL
1085 };
1086
1087
1088 static const arg_def_t kf_min_dist = ARG_DEF(NULL, "kf-min-dist", 1,
1089 "Minimum keyframe interval (frames)");
1090 static const arg_def_t kf_max_dist = ARG_DEF(NULL, "kf-max-dist", 1,
1091 "Maximum keyframe interval (frames)");
1092 static const arg_def_t kf_disabled = ARG_DEF(NULL, "disable-kf", 0,
1093 "Disable keyframe placement");
1094 static const arg_def_t *kf_args[] = {
1095 &kf_min_dist, &kf_max_dist, &kf_disabled, NULL
1096 };
1097
1098
1099 static const arg_def_t noise_sens = ARG_DEF(NULL, "noise-sensitivity", 1,
1100 "Noise sensitivity (frames to blur)");
1101 static const arg_def_t sharpness = ARG_DEF(NULL, "sharpness", 1,
1102 "Filter sharpness (0-7)");
1103 static const arg_def_t static_thresh = ARG_DEF(NULL, "static-thresh", 1,
1104 "Motion detection threshold");
1105 static const arg_def_t cpu_used = ARG_DEF(NULL, "cpu-used", 1,
1106 "CPU Used (-16..16)");
1107 static const arg_def_t token_parts = ARG_DEF(NULL, "token-parts", 1,
1108 "Number of token partitions to use, log2");
1109 static const arg_def_t tile_cols = ARG_DEF(NULL, "tile-columns", 1,
1110 "Number of tile columns to use, log2");
1111 static const arg_def_t tile_rows = ARG_DEF(NULL, "tile-rows", 1,
1112 "Number of tile rows to use, log2");
1113 static const arg_def_t auto_altref = ARG_DEF(NULL, "auto-alt-ref", 1,
1114 "Enable automatic alt reference frames");
1115 static const arg_def_t arnr_maxframes = ARG_DEF(NULL, "arnr-maxframes", 1,
1116 "AltRef Max Frames");
1117 static const arg_def_t arnr_strength = ARG_DEF(NULL, "arnr-strength", 1,
1118 "AltRef Strength");
1119 static const arg_def_t arnr_type = ARG_DEF(NULL, "arnr-type", 1,
1120 "AltRef Type");
1121 static const struct arg_enum_list tuning_enum[] = {
1122 {"psnr", VP8_TUNE_PSNR},
1123 {"ssim", VP8_TUNE_SSIM},
1124 {NULL, 0}
1125 };
1126 static const arg_def_t tune_ssim = ARG_DEF_ENUM(NULL, "tune", 1,
1127 "Material to favor", tuning_enum);
1128 static const arg_def_t cq_level = ARG_DEF(NULL, "cq-level", 1,
1129 "Constrained Quality Level");
1130 static const arg_def_t max_intra_rate_pct = ARG_DEF(NULL, "max-intra-rate", 1,
1131 "Max I-frame bitrate (pct)");
1132 static const arg_def_t lossless = ARG_DEF(NULL, "lossless", 1, "Lossless mode");
1133 #if CONFIG_VP9_ENCODER
1134 static const arg_def_t frame_parallel_decoding = ARG_DEF(
1135 NULL, "frame-parallel", 1, "Enable frame parallel decodability features");
1136 #endif
1137
1138 #if CONFIG_VP8_ENCODER
1139 static const arg_def_t *vp8_args[] = {
1140 &cpu_used, &auto_altref, &noise_sens, &sharpness, &static_thresh,
1141 &token_parts, &arnr_maxframes, &arnr_strength, &arnr_type,
1142 &tune_ssim, &cq_level, &max_intra_rate_pct,
1143 NULL
1144 };
1145 static const int vp8_arg_ctrl_map[] = {
1146 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
1147 VP8E_SET_NOISE_SENSITIVITY, VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
1148 VP8E_SET_TOKEN_PARTITIONS,
1149 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
1150 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
1151 0
1152 };
1153 #endif
1154
1155 #if CONFIG_VP9_ENCODER
1156 static const arg_def_t *vp9_args[] = {
1157 &cpu_used, &auto_altref, &noise_sens, &sharpness, &static_thresh,
1158 &tile_cols, &tile_rows, &arnr_maxframes, &arnr_strength, &arnr_type,
1159 &tune_ssim, &cq_level, &max_intra_rate_pct, &lossless,
1160 &frame_parallel_decoding,
1161 NULL
1162 };
1163 static const int vp9_arg_ctrl_map[] = {
1164 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
1165 VP8E_SET_NOISE_SENSITIVITY, VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
1166 VP9E_SET_TILE_COLUMNS, VP9E_SET_TILE_ROWS,
1167 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
1168 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
1169 VP9E_SET_LOSSLESS, VP9E_SET_FRAME_PARALLEL_DECODING,
1170 0
1171 };
1172 #endif
1173
1174 static const arg_def_t *no_args[] = { NULL };
1175
1176 static void usage_exit() {
1177 int i;
1178
1179 fprintf(stderr, "Usage: %s <options> -o dst_filename src_filename \n",
1180 exec_name);
1181
1182 fprintf(stderr, "\nOptions:\n");
1183 arg_show_usage(stdout, main_args);
1184 fprintf(stderr, "\nEncoder Global Options:\n");
1185 arg_show_usage(stdout, global_args);
1186 fprintf(stderr, "\nRate Control Options:\n");
1187 arg_show_usage(stdout, rc_args);
1188 fprintf(stderr, "\nTwopass Rate Control Options:\n");
1189 arg_show_usage(stdout, rc_twopass_args);
1190 fprintf(stderr, "\nKeyframe Placement Options:\n");
1191 arg_show_usage(stdout, kf_args);
1192 #if CONFIG_VP8_ENCODER
1193 fprintf(stderr, "\nVP8 Specific Options:\n");
1194 arg_show_usage(stdout, vp8_args);
1195 #endif
1196 #if CONFIG_VP9_ENCODER
1197 fprintf(stderr, "\nVP9 Specific Options:\n");
1198 arg_show_usage(stdout, vp9_args);
1199 #endif
1200 fprintf(stderr, "\nStream timebase (--timebase):\n"
1201 " The desired precision of timestamps in the output, expressed\n"
1202 " in fractional seconds. Default is 1/1000.\n");
1203 fprintf(stderr, "\n"
1204 "Included encoders:\n"
1205 "\n");
1206
1207 for (i = 0; i < sizeof(codecs) / sizeof(codecs[0]); i++)
1208 fprintf(stderr, " %-6s - %s\n",
1209 codecs[i].name,
1210 vpx_codec_iface_name(codecs[i].iface()));
1211
1212 exit(EXIT_FAILURE);
1213 }
1214
1215
1216 #define HIST_BAR_MAX 40
1217 struct hist_bucket {
1218 int low, high, count;
1219 };
1220
1221
1222 static int merge_hist_buckets(struct hist_bucket *bucket,
1223 int *buckets_,
1224 int max_buckets) {
1225 int small_bucket = 0, merge_bucket = INT_MAX, big_bucket = 0;
1226 int buckets = *buckets_;
1227 int i;
1228
1229 /* Find the extrema for this list of buckets */
1230 big_bucket = small_bucket = 0;
1231 for (i = 0; i < buckets; i++) {
1232 if (bucket[i].count < bucket[small_bucket].count)
1233 small_bucket = i;
1234 if (bucket[i].count > bucket[big_bucket].count)
1235 big_bucket = i;
1236 }
1237
1238 /* If we have too many buckets, merge the smallest with an adjacent
1239 * bucket.
1240 */
1241 while (buckets > max_buckets) {
1242 int last_bucket = buckets - 1;
1243
1244 /* merge the small bucket with an adjacent one. */
1245 if (small_bucket == 0)
1246 merge_bucket = 1;
1247 else if (small_bucket == last_bucket)
1248 merge_bucket = last_bucket - 1;
1249 else if (bucket[small_bucket - 1].count < bucket[small_bucket + 1].count)
1250 merge_bucket = small_bucket - 1;
1251 else
1252 merge_bucket = small_bucket + 1;
1253
1254 assert(abs(merge_bucket - small_bucket) <= 1);
1255 assert(small_bucket < buckets);
1256 assert(big_bucket < buckets);
1257 assert(merge_bucket < buckets);
1258
1259 if (merge_bucket < small_bucket) {
1260 bucket[merge_bucket].high = bucket[small_bucket].high;
1261 bucket[merge_bucket].count += bucket[small_bucket].count;
1262 } else {
1263 bucket[small_bucket].high = bucket[merge_bucket].high;
1264 bucket[small_bucket].count += bucket[merge_bucket].count;
1265 merge_bucket = small_bucket;
1266 }
1267
1268 assert(bucket[merge_bucket].low != bucket[merge_bucket].high);
1269
1270 buckets--;
1271
1272 /* Remove the merge_bucket from the list, and find the new small
1273 * and big buckets while we're at it
1274 */
1275 big_bucket = small_bucket = 0;
1276 for (i = 0; i < buckets; i++) {
1277 if (i > merge_bucket)
1278 bucket[i] = bucket[i + 1];
1279
1280 if (bucket[i].count < bucket[small_bucket].count)
1281 small_bucket = i;
1282 if (bucket[i].count > bucket[big_bucket].count)
1283 big_bucket = i;
1284 }
1285
1286 }
1287
1288 *buckets_ = buckets;
1289 return bucket[big_bucket].count;
1290 }
1291
1292
1293 static void show_histogram(const struct hist_bucket *bucket,
1294 int buckets,
1295 int total,
1296 int scale) {
1297 const char *pat1, *pat2;
1298 int i;
1299
1300 switch ((int)(log(bucket[buckets - 1].high) / log(10)) + 1) {
1301 case 1:
1302 case 2:
1303 pat1 = "%4d %2s: ";
1304 pat2 = "%4d-%2d: ";
1305 break;
1306 case 3:
1307 pat1 = "%5d %3s: ";
1308 pat2 = "%5d-%3d: ";
1309 break;
1310 case 4:
1311 pat1 = "%6d %4s: ";
1312 pat2 = "%6d-%4d: ";
1313 break;
1314 case 5:
1315 pat1 = "%7d %5s: ";
1316 pat2 = "%7d-%5d: ";
1317 break;
1318 case 6:
1319 pat1 = "%8d %6s: ";
1320 pat2 = "%8d-%6d: ";
1321 break;
1322 case 7:
1323 pat1 = "%9d %7s: ";
1324 pat2 = "%9d-%7d: ";
1325 break;
1326 default:
1327 pat1 = "%12d %10s: ";
1328 pat2 = "%12d-%10d: ";
1329 break;
1330 }
1331
1332 for (i = 0; i < buckets; i++) {
1333 int len;
1334 int j;
1335 float pct;
1336
1337 pct = (float)(100.0 * bucket[i].count / total);
1338 len = HIST_BAR_MAX * bucket[i].count / scale;
1339 if (len < 1)
1340 len = 1;
1341 assert(len <= HIST_BAR_MAX);
1342
1343 if (bucket[i].low == bucket[i].high)
1344 fprintf(stderr, pat1, bucket[i].low, "");
1345 else
1346 fprintf(stderr, pat2, bucket[i].low, bucket[i].high);
1347
1348 for (j = 0; j < HIST_BAR_MAX; j++)
1349 fprintf(stderr, j < len ? "=" : " ");
1350 fprintf(stderr, "\t%5d (%6.2f%%)\n", bucket[i].count, pct);
1351 }
1352 }
1353
1354
1355 static void show_q_histogram(const int counts[64], int max_buckets) {
1356 struct hist_bucket bucket[64];
1357 int buckets = 0;
1358 int total = 0;
1359 int scale;
1360 int i;
1361
1362
1363 for (i = 0; i < 64; i++) {
1364 if (counts[i]) {
1365 bucket[buckets].low = bucket[buckets].high = i;
1366 bucket[buckets].count = counts[i];
1367 buckets++;
1368 total += counts[i];
1369 }
1370 }
1371
1372 fprintf(stderr, "\nQuantizer Selection:\n");
1373 scale = merge_hist_buckets(bucket, &buckets, max_buckets);
1374 show_histogram(bucket, buckets, total, scale);
1375 }
1376
1377
1378 #define RATE_BINS (100)
1379 struct rate_hist {
1380 int64_t *pts;
1381 int *sz;
1382 int samples;
1383 int frames;
1384 struct hist_bucket bucket[RATE_BINS];
1385 int total;
1386 };
1387
1388
1389 static void init_rate_histogram(struct rate_hist *hist,
1390 const vpx_codec_enc_cfg_t *cfg,
1391 const vpx_rational_t *fps) {
1392 int i;
1393
1394 /* Determine the number of samples in the buffer. Use the file's framerate
1395 * to determine the number of frames in rc_buf_sz milliseconds, with an
1396 * adjustment (5/4) to account for alt-refs
1397 */
1398 hist->samples = cfg->rc_buf_sz * 5 / 4 * fps->num / fps->den / 1000;
1399
1400 /* prevent division by zero */
1401 if (hist->samples == 0)
1402 hist->samples = 1;
1403
1404 hist->pts = calloc(hist->samples, sizeof(*hist->pts));
1405 hist->sz = calloc(hist->samples, sizeof(*hist->sz));
1406 for (i = 0; i < RATE_BINS; i++) {
1407 hist->bucket[i].low = INT_MAX;
1408 hist->bucket[i].high = 0;
1409 hist->bucket[i].count = 0;
1410 }
1411 }
1412
1413
1414 static void destroy_rate_histogram(struct rate_hist *hist) {
1415 free(hist->pts);
1416 free(hist->sz);
1417 }
1418
1419
1420 static void update_rate_histogram(struct rate_hist *hist,
1421 const vpx_codec_enc_cfg_t *cfg,
1422 const vpx_codec_cx_pkt_t *pkt) {
1423 int i, idx;
1424 int64_t now, then, sum_sz = 0, avg_bitrate;
1425
1426 now = pkt->data.frame.pts * 1000
1427 * (uint64_t)cfg->g_timebase.num / (uint64_t)cfg->g_timebase.den;
1428
1429 idx = hist->frames++ % hist->samples;
1430 hist->pts[idx] = now;
1431 hist->sz[idx] = (int)pkt->data.frame.sz;
1432
1433 if (now < cfg->rc_buf_initial_sz)
1434 return;
1435
1436 then = now;
1437
1438 /* Sum the size over the past rc_buf_sz ms */
1439 for (i = hist->frames; i > 0 && hist->frames - i < hist->samples; i--) {
1440 int i_idx = (i - 1) % hist->samples;
1441
1442 then = hist->pts[i_idx];
1443 if (now - then > cfg->rc_buf_sz)
1444 break;
1445 sum_sz += hist->sz[i_idx];
1446 }
1447
1448 if (now == then)
1449 return;
1450
1451 avg_bitrate = sum_sz * 8 * 1000 / (now - then);
1452 idx = (int)(avg_bitrate * (RATE_BINS / 2) / (cfg->rc_target_bitrate * 1000));
1453 if (idx < 0)
1454 idx = 0;
1455 if (idx > RATE_BINS - 1)
1456 idx = RATE_BINS - 1;
1457 if (hist->bucket[idx].low > avg_bitrate)
1458 hist->bucket[idx].low = (int)avg_bitrate;
1459 if (hist->bucket[idx].high < avg_bitrate)
1460 hist->bucket[idx].high = (int)avg_bitrate;
1461 hist->bucket[idx].count++;
1462 hist->total++;
1463 }
1464
1465
1466 static void show_rate_histogram(struct rate_hist *hist,
1467 const vpx_codec_enc_cfg_t *cfg,
1468 int max_buckets) {
1469 int i, scale;
1470 int buckets = 0;
1471
1472 for (i = 0; i < RATE_BINS; i++) {
1473 if (hist->bucket[i].low == INT_MAX)
1474 continue;
1475 hist->bucket[buckets++] = hist->bucket[i];
1476 }
1477
1478 fprintf(stderr, "\nRate (over %dms window):\n", cfg->rc_buf_sz);
1479 scale = merge_hist_buckets(hist->bucket, &buckets, max_buckets);
1480 show_histogram(hist->bucket, buckets, hist->total, scale);
1481 }
1482
1483 #define mmin(a, b) ((a) < (b) ? (a) : (b))
1484 static void find_mismatch(vpx_image_t *img1, vpx_image_t *img2,
1485 int yloc[4], int uloc[4], int vloc[4]) {
1486 const unsigned int bsize = 64;
1487 const unsigned int bsizey = bsize >> img1->y_chroma_shift;
1488 const unsigned int bsizex = bsize >> img1->x_chroma_shift;
1489 const int c_w = (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
1490 const int c_h = (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
1491 unsigned int match = 1;
1492 unsigned int i, j;
1493 yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
1494 for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
1495 for (j = 0; match && j < img1->d_w; j += bsize) {
1496 int k, l;
1497 int si = mmin(i + bsize, img1->d_h) - i;
1498 int sj = mmin(j + bsize, img1->d_w) - j;
1499 for (k = 0; match && k < si; k++)
1500 for (l = 0; match && l < sj; l++) {
1501 if (*(img1->planes[VPX_PLANE_Y] +
1502 (i + k) * img1->stride[VPX_PLANE_Y] + j + l) !=
1503 *(img2->planes[VPX_PLANE_Y] +
1504 (i + k) * img2->stride[VPX_PLANE_Y] + j + l)) {
1505 yloc[0] = i + k;
1506 yloc[1] = j + l;
1507 yloc[2] = *(img1->planes[VPX_PLANE_Y] +
1508 (i + k) * img1->stride[VPX_PLANE_Y] + j + l);
1509 yloc[3] = *(img2->planes[VPX_PLANE_Y] +
1510 (i + k) * img2->stride[VPX_PLANE_Y] + j + l);
1511 match = 0;
1512 break;
1513 }
1514 }
1515 }
1516 }
1517
1518 uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
1519 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
1520 for (j = 0; j < match && c_w; j += bsizex) {
1521 int k, l;
1522 int si = mmin(i + bsizey, c_h - i);
1523 int sj = mmin(j + bsizex, c_w - j);
1524 for (k = 0; match && k < si; k++)
1525 for (l = 0; match && l < sj; l++) {
1526 if (*(img1->planes[VPX_PLANE_U] +
1527 (i + k) * img1->stride[VPX_PLANE_U] + j + l) !=
1528 *(img2->planes[VPX_PLANE_U] +
1529 (i + k) * img2->stride[VPX_PLANE_U] + j + l)) {
1530 uloc[0] = i + k;
1531 uloc[1] = j + l;
1532 uloc[2] = *(img1->planes[VPX_PLANE_U] +
1533 (i + k) * img1->stride[VPX_PLANE_U] + j + l);
1534 uloc[3] = *(img2->planes[VPX_PLANE_U] +
1535 (i + k) * img2->stride[VPX_PLANE_V] + j + l);
1536 match = 0;
1537 break;
1538 }
1539 }
1540 }
1541 }
1542 vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
1543 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
1544 for (j = 0; j < match && c_w; j += bsizex) {
1545 int k, l;
1546 int si = mmin(i + bsizey, c_h - i);
1547 int sj = mmin(j + bsizex, c_w - j);
1548 for (k = 0; match && k < si; k++)
1549 for (l = 0; match && l < sj; l++) {
1550 if (*(img1->planes[VPX_PLANE_V] +
1551 (i + k) * img1->stride[VPX_PLANE_V] + j + l) !=
1552 *(img2->planes[VPX_PLANE_V] +
1553 (i + k) * img2->stride[VPX_PLANE_V] + j + l)) {
1554 vloc[0] = i + k;
1555 vloc[1] = j + l;
1556 vloc[2] = *(img1->planes[VPX_PLANE_V] +
1557 (i + k) * img1->stride[VPX_PLANE_V] + j + l);
1558 vloc[3] = *(img2->planes[VPX_PLANE_V] +
1559 (i + k) * img2->stride[VPX_PLANE_V] + j + l);
1560 match = 0;
1561 break;
1562 }
1563 }
1564 }
1565 }
1566 }
1567
1568 static int compare_img(vpx_image_t *img1, vpx_image_t *img2)
1569 {
1570 const int c_w = (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
1571 const int c_h = (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
1572 int match = 1;
1573 unsigned int i;
1574
1575 match &= (img1->fmt == img2->fmt);
1576 match &= (img1->w == img2->w);
1577 match &= (img1->h == img2->h);
1578
1579 for (i = 0; i < img1->d_h; i++)
1580 match &= (memcmp(img1->planes[VPX_PLANE_Y]+i*img1->stride[VPX_PLANE_Y],
1581 img2->planes[VPX_PLANE_Y]+i*img2->stride[VPX_PLANE_Y],
1582 img1->d_w) == 0);
1583
1584 for (i = 0; i < c_h; i++)
1585 match &= (memcmp(img1->planes[VPX_PLANE_U]+i*img1->stride[VPX_PLANE_U],
1586 img2->planes[VPX_PLANE_U]+i*img2->stride[VPX_PLANE_U],
1587 c_w) == 0);
1588
1589 for (i = 0; i < c_h; i++)
1590 match &= (memcmp(img1->planes[VPX_PLANE_V]+i*img1->stride[VPX_PLANE_U],
1591 img2->planes[VPX_PLANE_V]+i*img2->stride[VPX_PLANE_U],
1592 c_w) == 0);
1593
1594 return match;
1595 }
1596
1597
1598 #define NELEMENTS(x) (sizeof(x)/sizeof(x[0]))
1599 #define MAX(x,y) ((x)>(y)?(x):(y))
1600 #if CONFIG_VP8_ENCODER && !CONFIG_VP9_ENCODER
1601 #define ARG_CTRL_CNT_MAX NELEMENTS(vp8_arg_ctrl_map)
1602 #elif !CONFIG_VP8_ENCODER && CONFIG_VP9_ENCODER
1603 #define ARG_CTRL_CNT_MAX NELEMENTS(vp9_arg_ctrl_map)
1604 #else
1605 #define ARG_CTRL_CNT_MAX MAX(NELEMENTS(vp8_arg_ctrl_map), \
1606 NELEMENTS(vp9_arg_ctrl_map))
1607 #endif
1608
1609 /* Configuration elements common to all streams */
1610 struct global_config {
1611 const struct codec_item *codec;
1612 int passes;
1613 int pass;
1614 int usage;
1615 int deadline;
1616 int use_i420;
1617 int quiet;
1618 int verbose;
1619 int limit;
1620 int skip_frames;
1621 int show_psnr;
1622 enum TestDecodeFatality test_decode;
1623 int have_framerate;
1624 struct vpx_rational framerate;
1625 int out_part;
1626 int debug;
1627 int show_q_hist_buckets;
1628 int show_rate_hist_buckets;
1629 };
1630
1631
1632 /* Per-stream configuration */
1633 struct stream_config {
1634 struct vpx_codec_enc_cfg cfg;
1635 const char *out_fn;
1636 const char *stats_fn;
1637 stereo_format_t stereo_fmt;
1638 int arg_ctrls[ARG_CTRL_CNT_MAX][2];
1639 int arg_ctrl_cnt;
1640 int write_webm;
1641 int have_kf_max_dist;
1642 };
1643
1644
1645 struct stream_state {
1646 int index;
1647 struct stream_state *next;
1648 struct stream_config config;
1649 FILE *file;
1650 struct rate_hist rate_hist;
1651 EbmlGlobal ebml;
1652 uint32_t hash;
1653 uint64_t psnr_sse_total;
1654 uint64_t psnr_samples_total;
1655 double psnr_totals[4];
1656 int psnr_count;
1657 int counts[64];
1658 vpx_codec_ctx_t encoder;
1659 unsigned int frames_out;
1660 uint64_t cx_time;
1661 size_t nbytes;
1662 stats_io_t stats;
1663 struct vpx_image *img;
1664 vpx_codec_ctx_t decoder;
1665 int mismatch_seen;
1666 };
1667
1668
1669 void validate_positive_rational(const char *msg,
1670 struct vpx_rational *rat) {
1671 if (rat->den < 0) {
1672 rat->num *= -1;
1673 rat->den *= -1;
1674 }
1675
1676 if (rat->num < 0)
1677 die("Error: %s must be positive\n", msg);
1678
1679 if (!rat->den)
1680 die("Error: %s has zero denominator\n", msg);
1681 }
1682
1683
1684 static void parse_global_config(struct global_config *global, char **argv) {
1685 char **argi, **argj;
1686 struct arg arg;
1687
1688 /* Initialize default parameters */
1689 memset(global, 0, sizeof(*global));
1690 global->codec = codecs;
1691 global->passes = 1;
1692 global->use_i420 = 1;
1693
1694 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
1695 arg.argv_step = 1;
1696
1697 if (arg_match(&arg, &codecarg, argi)) {
1698 int j, k = -1;
1699
1700 for (j = 0; j < sizeof(codecs) / sizeof(codecs[0]); j++)
1701 if (!strcmp(codecs[j].name, arg.val))
1702 k = j;
1703
1704 if (k >= 0)
1705 global->codec = codecs + k;
1706 else
1707 die("Error: Unrecognized argument (%s) to --codec\n",
1708 arg.val);
1709
1710 } else if (arg_match(&arg, &passes, argi)) {
1711 global->passes = arg_parse_uint(&arg);
1712
1713 if (global->passes < 1 || global->passes > 2)
1714 die("Error: Invalid number of passes (%d)\n", global->passes);
1715 } else if (arg_match(&arg, &pass_arg, argi)) {
1716 global->pass = arg_parse_uint(&arg);
1717
1718 if (global->pass < 1 || global->pass > 2)
1719 die("Error: Invalid pass selected (%d)\n",
1720 global->pass);
1721 } else if (arg_match(&arg, &usage, argi))
1722 global->usage = arg_parse_uint(&arg);
1723 else if (arg_match(&arg, &deadline, argi))
1724 global->deadline = arg_parse_uint(&arg);
1725 else if (arg_match(&arg, &best_dl, argi))
1726 global->deadline = VPX_DL_BEST_QUALITY;
1727 else if (arg_match(&arg, &good_dl, argi))
1728 global->deadline = VPX_DL_GOOD_QUALITY;
1729 else if (arg_match(&arg, &rt_dl, argi))
1730 global->deadline = VPX_DL_REALTIME;
1731 else if (arg_match(&arg, &use_yv12, argi))
1732 global->use_i420 = 0;
1733 else if (arg_match(&arg, &use_i420, argi))
1734 global->use_i420 = 1;
1735 else if (arg_match(&arg, &quietarg, argi))
1736 global->quiet = 1;
1737 else if (arg_match(&arg, &verbosearg, argi))
1738 global->verbose = 1;
1739 else if (arg_match(&arg, &limit, argi))
1740 global->limit = arg_parse_uint(&arg);
1741 else if (arg_match(&arg, &skip, argi))
1742 global->skip_frames = arg_parse_uint(&arg);
1743 else if (arg_match(&arg, &psnrarg, argi))
1744 global->show_psnr = 1;
1745 else if (arg_match(&arg, &recontest, argi))
1746 global->test_decode = arg_parse_enum_or_int(&arg);
1747 else if (arg_match(&arg, &framerate, argi)) {
1748 global->framerate = arg_parse_rational(&arg);
1749 validate_positive_rational(arg.name, &global->framerate);
1750 global->have_framerate = 1;
1751 } else if (arg_match(&arg, &out_part, argi))
1752 global->out_part = 1;
1753 else if (arg_match(&arg, &debugmode, argi))
1754 global->debug = 1;
1755 else if (arg_match(&arg, &q_hist_n, argi))
1756 global->show_q_hist_buckets = arg_parse_uint(&arg);
1757 else if (arg_match(&arg, &rate_hist_n, argi))
1758 global->show_rate_hist_buckets = arg_parse_uint(&arg);
1759 else
1760 argj++;
1761 }
1762
1763 /* Validate global config */
1764
1765 if (global->pass) {
1766 /* DWIM: Assume the user meant passes=2 if pass=2 is specified */
1767 if (global->pass > global->passes) {
1768 warn("Assuming --pass=%d implies --passes=%d\n",
1769 global->pass, global->pass);
1770 global->passes = global->pass;
1771 }
1772 }
1773 }
1774
1775
1776 void open_input_file(struct input_state *input) {
1777 unsigned int fourcc;
1778
1779 /* Parse certain options from the input file, if possible */
1780 input->file = strcmp(input->fn, "-") ? fopen(input->fn, "rb")
1781 : set_binary_mode(stdin);
1782
1783 if (!input->file)
1784 fatal("Failed to open input file");
1785
1786 if (!fseeko(input->file, 0, SEEK_END)) {
1787 /* Input file is seekable. Figure out how long it is, so we can get
1788 * progress info.
1789 */
1790 input->length = ftello(input->file);
1791 rewind(input->file);
1792 }
1793
1794 /* For RAW input sources, these bytes will applied on the first frame
1795 * in read_frame().
1796 */
1797 input->detect.buf_read = fread(input->detect.buf, 1, 4, input->file);
1798 input->detect.position = 0;
1799
1800 if (input->detect.buf_read == 4
1801 && file_is_y4m(input->file, &input->y4m, input->detect.buf)) {
1802 if (y4m_input_open(&input->y4m, input->file, input->detect.buf, 4,
1803 input->only_i420) >= 0) {
1804 input->file_type = FILE_TYPE_Y4M;
1805 input->w = input->y4m.pic_w;
1806 input->h = input->y4m.pic_h;
1807 input->framerate.num = input->y4m.fps_n;
1808 input->framerate.den = input->y4m.fps_d;
1809 input->use_i420 = 0;
1810 } else
1811 fatal("Unsupported Y4M stream.");
1812 } else if (input->detect.buf_read == 4 && file_is_ivf(input, &fourcc)) {
1813 input->file_type = FILE_TYPE_IVF;
1814 switch (fourcc) {
1815 case 0x32315659:
1816 input->use_i420 = 0;
1817 break;
1818 case 0x30323449:
1819 input->use_i420 = 1;
1820 break;
1821 default:
1822 fatal("Unsupported fourcc (%08x) in IVF", fourcc);
1823 }
1824 } else {
1825 input->file_type = FILE_TYPE_RAW;
1826 }
1827 }
1828
1829
1830 static void close_input_file(struct input_state *input) {
1831 fclose(input->file);
1832 if (input->file_type == FILE_TYPE_Y4M)
1833 y4m_input_close(&input->y4m);
1834 }
1835
1836 static struct stream_state *new_stream(struct global_config *global,
1837 struct stream_state *prev) {
1838 struct stream_state *stream;
1839
1840 stream = calloc(1, sizeof(*stream));
1841 if (!stream)
1842 fatal("Failed to allocate new stream.");
1843 if (prev) {
1844 memcpy(stream, prev, sizeof(*stream));
1845 stream->index++;
1846 prev->next = stream;
1847 } else {
1848 vpx_codec_err_t res;
1849
1850 /* Populate encoder configuration */
1851 res = vpx_codec_enc_config_default(global->codec->iface(),
1852 &stream->config.cfg,
1853 global->usage);
1854 if (res)
1855 fatal("Failed to get config: %s\n", vpx_codec_err_to_string(res));
1856
1857 /* Change the default timebase to a high enough value so that the
1858 * encoder will always create strictly increasing timestamps.
1859 */
1860 stream->config.cfg.g_timebase.den = 1000;
1861
1862 /* Never use the library's default resolution, require it be parsed
1863 * from the file or set on the command line.
1864 */
1865 stream->config.cfg.g_w = 0;
1866 stream->config.cfg.g_h = 0;
1867
1868 /* Initialize remaining stream parameters */
1869 stream->config.stereo_fmt = STEREO_FORMAT_MONO;
1870 stream->config.write_webm = 1;
1871 stream->ebml.last_pts_ms = -1;
1872
1873 /* Allows removal of the application version from the EBML tags */
1874 stream->ebml.debug = global->debug;
1875 }
1876
1877 /* Output files must be specified for each stream */
1878 stream->config.out_fn = NULL;
1879
1880 stream->next = NULL;
1881 return stream;
1882 }
1883
1884
1885 static int parse_stream_params(struct global_config *global,
1886 struct stream_state *stream,
1887 char **argv) {
1888 char **argi, **argj;
1889 struct arg arg;
1890 static const arg_def_t **ctrl_args = no_args;
1891 static const int *ctrl_args_map = NULL;
1892 struct stream_config *config = &stream->config;
1893 int eos_mark_found = 0;
1894
1895 /* Handle codec specific options */
1896 if (0) {
1897 #if CONFIG_VP8_ENCODER
1898 } else if (global->codec->iface == vpx_codec_vp8_cx) {
1899 ctrl_args = vp8_args;
1900 ctrl_args_map = vp8_arg_ctrl_map;
1901 #endif
1902 #if CONFIG_VP9_ENCODER
1903 } else if (global->codec->iface == vpx_codec_vp9_cx) {
1904 ctrl_args = vp9_args;
1905 ctrl_args_map = vp9_arg_ctrl_map;
1906 #endif
1907 }
1908
1909 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
1910 arg.argv_step = 1;
1911
1912 /* Once we've found an end-of-stream marker (--) we want to continue
1913 * shifting arguments but not consuming them.
1914 */
1915 if (eos_mark_found) {
1916 argj++;
1917 continue;
1918 } else if (!strcmp(*argj, "--")) {
1919 eos_mark_found = 1;
1920 continue;
1921 }
1922
1923 if (0);
1924 else if (arg_match(&arg, &outputfile, argi))
1925 config->out_fn = arg.val;
1926 else if (arg_match(&arg, &fpf_name, argi))
1927 config->stats_fn = arg.val;
1928 else if (arg_match(&arg, &use_ivf, argi))
1929 config->write_webm = 0;
1930 else if (arg_match(&arg, &threads, argi))
1931 config->cfg.g_threads = arg_parse_uint(&arg);
1932 else if (arg_match(&arg, &profile, argi))
1933 config->cfg.g_profile = arg_parse_uint(&arg);
1934 else if (arg_match(&arg, &width, argi))
1935 config->cfg.g_w = arg_parse_uint(&arg);
1936 else if (arg_match(&arg, &height, argi))
1937 config->cfg.g_h = arg_parse_uint(&arg);
1938 else if (arg_match(&arg, &stereo_mode, argi))
1939 config->stereo_fmt = arg_parse_enum_or_int(&arg);
1940 else if (arg_match(&arg, &timebase, argi)) {
1941 config->cfg.g_timebase = arg_parse_rational(&arg);
1942 validate_positive_rational(arg.name, &config->cfg.g_timebase);
1943 } else if (arg_match(&arg, &error_resilient, argi))
1944 config->cfg.g_error_resilient = arg_parse_uint(&arg);
1945 else if (arg_match(&arg, &lag_in_frames, argi))
1946 config->cfg.g_lag_in_frames = arg_parse_uint(&arg);
1947 else if (arg_match(&arg, &dropframe_thresh, argi))
1948 config->cfg.rc_dropframe_thresh = arg_parse_uint(&arg);
1949 else if (arg_match(&arg, &resize_allowed, argi))
1950 config->cfg.rc_resize_allowed = arg_parse_uint(&arg);
1951 else if (arg_match(&arg, &resize_up_thresh, argi))
1952 config->cfg.rc_resize_up_thresh = arg_parse_uint(&arg);
1953 else if (arg_match(&arg, &resize_down_thresh, argi))
1954 config->cfg.rc_resize_down_thresh = arg_parse_uint(&arg);
1955 else if (arg_match(&arg, &end_usage, argi))
1956 config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
1957 else if (arg_match(&arg, &target_bitrate, argi))
1958 config->cfg.rc_target_bitrate = arg_parse_uint(&arg);
1959 else if (arg_match(&arg, &min_quantizer, argi))
1960 config->cfg.rc_min_quantizer = arg_parse_uint(&arg);
1961 else if (arg_match(&arg, &max_quantizer, argi))
1962 config->cfg.rc_max_quantizer = arg_parse_uint(&arg);
1963 else if (arg_match(&arg, &undershoot_pct, argi))
1964 config->cfg.rc_undershoot_pct = arg_parse_uint(&arg);
1965 else if (arg_match(&arg, &overshoot_pct, argi))
1966 config->cfg.rc_overshoot_pct = arg_parse_uint(&arg);
1967 else if (arg_match(&arg, &buf_sz, argi))
1968 config->cfg.rc_buf_sz = arg_parse_uint(&arg);
1969 else if (arg_match(&arg, &buf_initial_sz, argi))
1970 config->cfg.rc_buf_initial_sz = arg_parse_uint(&arg);
1971 else if (arg_match(&arg, &buf_optimal_sz, argi))
1972 config->cfg.rc_buf_optimal_sz = arg_parse_uint(&arg);
1973 else if (arg_match(&arg, &bias_pct, argi)) {
1974 config->cfg.rc_2pass_vbr_bias_pct = arg_parse_uint(&arg);
1975
1976 if (global->passes < 2)
1977 warn("option %s ignored in one-pass mode.\n", arg.name);
1978 } else if (arg_match(&arg, &minsection_pct, argi)) {
1979 config->cfg.rc_2pass_vbr_minsection_pct = arg_parse_uint(&arg);
1980
1981 if (global->passes < 2)
1982 warn("option %s ignored in one-pass mode.\n", arg.name);
1983 } else if (arg_match(&arg, &maxsection_pct, argi)) {
1984 config->cfg.rc_2pass_vbr_maxsection_pct = arg_parse_uint(&arg);
1985
1986 if (global->passes < 2)
1987 warn("option %s ignored in one-pass mode.\n", arg.name);
1988 } else if (arg_match(&arg, &kf_min_dist, argi))
1989 config->cfg.kf_min_dist = arg_parse_uint(&arg);
1990 else if (arg_match(&arg, &kf_max_dist, argi)) {
1991 config->cfg.kf_max_dist = arg_parse_uint(&arg);
1992 config->have_kf_max_dist = 1;
1993 } else if (arg_match(&arg, &kf_disabled, argi))
1994 config->cfg.kf_mode = VPX_KF_DISABLED;
1995 else {
1996 int i, match = 0;
1997
1998 for (i = 0; ctrl_args[i]; i++) {
1999 if (arg_match(&arg, ctrl_args[i], argi)) {
2000 int j;
2001 match = 1;
2002
2003 /* Point either to the next free element or the first
2004 * instance of this control.
2005 */
2006 for (j = 0; j < config->arg_ctrl_cnt; j++)
2007 if (config->arg_ctrls[j][0] == ctrl_args_map[i])
2008 break;
2009
2010 /* Update/insert */
2011 assert(j < ARG_CTRL_CNT_MAX);
2012 if (j < ARG_CTRL_CNT_MAX) {
2013 config->arg_ctrls[j][0] = ctrl_args_map[i];
2014 config->arg_ctrls[j][1] = arg_parse_enum_or_int(&arg);
2015 if (j == config->arg_ctrl_cnt)
2016 config->arg_ctrl_cnt++;
2017 }
2018
2019 }
2020 }
2021
2022 if (!match)
2023 argj++;
2024 }
2025 }
2026
2027 return eos_mark_found;
2028 }
2029
2030
2031 #define FOREACH_STREAM(func)\
2032 do\
2033 {\
2034 struct stream_state *stream;\
2035 \
2036 for(stream = streams; stream; stream = stream->next)\
2037 func;\
2038 }while(0)
2039
2040
2041 static void validate_stream_config(struct stream_state *stream) {
2042 struct stream_state *streami;
2043
2044 if (!stream->config.cfg.g_w || !stream->config.cfg.g_h)
2045 fatal("Stream %d: Specify stream dimensions with --width (-w) "
2046 " and --height (-h)", stream->index);
2047
2048 for (streami = stream; streami; streami = streami->next) {
2049 /* All streams require output files */
2050 if (!streami->config.out_fn)
2051 fatal("Stream %d: Output file is required (specify with -o)",
2052 streami->index);
2053
2054 /* Check for two streams outputting to the same file */
2055 if (streami != stream) {
2056 const char *a = stream->config.out_fn;
2057 const char *b = streami->config.out_fn;
2058 if (!strcmp(a, b) && strcmp(a, "/dev/null") && strcmp(a, ":nul"))
2059 fatal("Stream %d: duplicate output file (from stream %d)",
2060 streami->index, stream->index);
2061 }
2062
2063 /* Check for two streams sharing a stats file. */
2064 if (streami != stream) {
2065 const char *a = stream->config.stats_fn;
2066 const char *b = streami->config.stats_fn;
2067 if (a && b && !strcmp(a, b))
2068 fatal("Stream %d: duplicate stats file (from stream %d)",
2069 streami->index, stream->index);
2070 }
2071 }
2072 }
2073
2074
2075 static void set_stream_dimensions(struct stream_state *stream,
2076 unsigned int w,
2077 unsigned int h) {
2078 if (!stream->config.cfg.g_w) {
2079 if (!stream->config.cfg.g_h)
2080 stream->config.cfg.g_w = w;
2081 else
2082 stream->config.cfg.g_w = w * stream->config.cfg.g_h / h;
2083 }
2084 if (!stream->config.cfg.g_h) {
2085 stream->config.cfg.g_h = h * stream->config.cfg.g_w / w;
2086 }
2087 }
2088
2089
2090 static void set_default_kf_interval(struct stream_state *stream,
2091 struct global_config *global) {
2092 /* Use a max keyframe interval of 5 seconds, if none was
2093 * specified on the command line.
2094 */
2095 if (!stream->config.have_kf_max_dist) {
2096 double framerate = (double)global->framerate.num / global->framerate.den;
2097 if (framerate > 0.0)
2098 stream->config.cfg.kf_max_dist = (unsigned int)(5.0 * framerate);
2099 }
2100 }
2101
2102
2103 static void show_stream_config(struct stream_state *stream,
2104 struct global_config *global,
2105 struct input_state *input) {
2106
2107 #define SHOW(field) \
2108 fprintf(stderr, " %-28s = %d\n", #field, stream->config.cfg.field)
2109
2110 if (stream->index == 0) {
2111 fprintf(stderr, "Codec: %s\n",
2112 vpx_codec_iface_name(global->codec->iface()));
2113 fprintf(stderr, "Source file: %s Format: %s\n", input->fn,
2114 input->use_i420 ? "I420" : "YV12");
2115 }
2116 if (stream->next || stream->index)
2117 fprintf(stderr, "\nStream Index: %d\n", stream->index);
2118 fprintf(stderr, "Destination file: %s\n", stream->config.out_fn);
2119 fprintf(stderr, "Encoder parameters:\n");
2120
2121 SHOW(g_usage);
2122 SHOW(g_threads);
2123 SHOW(g_profile);
2124 SHOW(g_w);
2125 SHOW(g_h);
2126 SHOW(g_timebase.num);
2127 SHOW(g_timebase.den);
2128 SHOW(g_error_resilient);
2129 SHOW(g_pass);
2130 SHOW(g_lag_in_frames);
2131 SHOW(rc_dropframe_thresh);
2132 SHOW(rc_resize_allowed);
2133 SHOW(rc_resize_up_thresh);
2134 SHOW(rc_resize_down_thresh);
2135 SHOW(rc_end_usage);
2136 SHOW(rc_target_bitrate);
2137 SHOW(rc_min_quantizer);
2138 SHOW(rc_max_quantizer);
2139 SHOW(rc_undershoot_pct);
2140 SHOW(rc_overshoot_pct);
2141 SHOW(rc_buf_sz);
2142 SHOW(rc_buf_initial_sz);
2143 SHOW(rc_buf_optimal_sz);
2144 SHOW(rc_2pass_vbr_bias_pct);
2145 SHOW(rc_2pass_vbr_minsection_pct);
2146 SHOW(rc_2pass_vbr_maxsection_pct);
2147 SHOW(kf_mode);
2148 SHOW(kf_min_dist);
2149 SHOW(kf_max_dist);
2150 }
2151
2152
2153 static void open_output_file(struct stream_state *stream,
2154 struct global_config *global) {
2155 const char *fn = stream->config.out_fn;
2156
2157 stream->file = strcmp(fn, "-") ? fopen(fn, "wb") : set_binary_mode(stdout);
2158
2159 if (!stream->file)
2160 fatal("Failed to open output file");
2161
2162 if (stream->config.write_webm && fseek(stream->file, 0, SEEK_CUR))
2163 fatal("WebM output to pipes not supported.");
2164
2165 if (stream->config.write_webm) {
2166 stream->ebml.stream = stream->file;
2167 write_webm_file_header(&stream->ebml, &stream->config.cfg,
2168 &global->framerate,
2169 stream->config.stereo_fmt,
2170 global->codec->fourcc);
2171 } else
2172 write_ivf_file_header(stream->file, &stream->config.cfg,
2173 global->codec->fourcc, 0);
2174 }
2175
2176
2177 static void close_output_file(struct stream_state *stream,
2178 unsigned int fourcc) {
2179 if (stream->config.write_webm) {
2180 write_webm_file_footer(&stream->ebml, stream->hash);
2181 free(stream->ebml.cue_list);
2182 stream->ebml.cue_list = NULL;
2183 } else {
2184 if (!fseek(stream->file, 0, SEEK_SET))
2185 write_ivf_file_header(stream->file, &stream->config.cfg,
2186 fourcc,
2187 stream->frames_out);
2188 }
2189
2190 fclose(stream->file);
2191 }
2192
2193
2194 static void setup_pass(struct stream_state *stream,
2195 struct global_config *global,
2196 int pass) {
2197 if (stream->config.stats_fn) {
2198 if (!stats_open_file(&stream->stats, stream->config.stats_fn,
2199 pass))
2200 fatal("Failed to open statistics store");
2201 } else {
2202 if (!stats_open_mem(&stream->stats, pass))
2203 fatal("Failed to open statistics store");
2204 }
2205
2206 stream->config.cfg.g_pass = global->passes == 2
2207 ? pass ? VPX_RC_LAST_PASS : VPX_RC_FIRST_PASS
2208 : VPX_RC_ONE_PASS;
2209 if (pass)
2210 stream->config.cfg.rc_twopass_stats_in = stats_get(&stream->stats);
2211
2212 stream->cx_time = 0;
2213 stream->nbytes = 0;
2214 stream->frames_out = 0;
2215 }
2216
2217
2218 static void initialize_encoder(struct stream_state *stream,
2219 struct global_config *global) {
2220 int i;
2221 int flags = 0;
2222
2223 flags |= global->show_psnr ? VPX_CODEC_USE_PSNR : 0;
2224 flags |= global->out_part ? VPX_CODEC_USE_OUTPUT_PARTITION : 0;
2225
2226 /* Construct Encoder Context */
2227 vpx_codec_enc_init(&stream->encoder, global->codec->iface(),
2228 &stream->config.cfg, flags);
2229 ctx_exit_on_error(&stream->encoder, "Failed to initialize encoder");
2230
2231 /* Note that we bypass the vpx_codec_control wrapper macro because
2232 * we're being clever to store the control IDs in an array. Real
2233 * applications will want to make use of the enumerations directly
2234 */
2235 for (i = 0; i < stream->config.arg_ctrl_cnt; i++) {
2236 int ctrl = stream->config.arg_ctrls[i][0];
2237 int value = stream->config.arg_ctrls[i][1];
2238 if (vpx_codec_control_(&stream->encoder, ctrl, value))
2239 fprintf(stderr, "Error: Tried to set control %d = %d\n",
2240 ctrl, value);
2241
2242 ctx_exit_on_error(&stream->encoder, "Failed to control codec");
2243 }
2244
2245 #if CONFIG_DECODERS
2246 if (global->test_decode != TEST_DECODE_OFF) {
2247 vpx_codec_dec_init(&stream->decoder, global->codec->dx_iface(), NULL, 0);
2248 }
2249 #endif
2250 }
2251
2252
2253 static void encode_frame(struct stream_state *stream,
2254 struct global_config *global,
2255 struct vpx_image *img,
2256 unsigned int frames_in) {
2257 vpx_codec_pts_t frame_start, next_frame_start;
2258 struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
2259 struct vpx_usec_timer timer;
2260
2261 frame_start = (cfg->g_timebase.den * (int64_t)(frames_in - 1)
2262 * global->framerate.den)
2263 / cfg->g_timebase.num / global->framerate.num;
2264 next_frame_start = (cfg->g_timebase.den * (int64_t)(frames_in)
2265 * global->framerate.den)
2266 / cfg->g_timebase.num / global->framerate.num;
2267
2268 /* Scale if necessary */
2269 if (img && (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
2270 if (!stream->img)
2271 stream->img = vpx_img_alloc(NULL, VPX_IMG_FMT_I420,
2272 cfg->g_w, cfg->g_h, 16);
2273 I420Scale(img->planes[VPX_PLANE_Y], img->stride[VPX_PLANE_Y],
2274 img->planes[VPX_PLANE_U], img->stride[VPX_PLANE_U],
2275 img->planes[VPX_PLANE_V], img->stride[VPX_PLANE_V],
2276 img->d_w, img->d_h,
2277 stream->img->planes[VPX_PLANE_Y],
2278 stream->img->stride[VPX_PLANE_Y],
2279 stream->img->planes[VPX_PLANE_U],
2280 stream->img->stride[VPX_PLANE_U],
2281 stream->img->planes[VPX_PLANE_V],
2282 stream->img->stride[VPX_PLANE_V],
2283 stream->img->d_w, stream->img->d_h,
2284 kFilterBox);
2285
2286 img = stream->img;
2287 }
2288
2289 vpx_usec_timer_start(&timer);
2290 vpx_codec_encode(&stream->encoder, img, frame_start,
2291 (unsigned long)(next_frame_start - frame_start),
2292 0, global->deadline);
2293 vpx_usec_timer_mark(&timer);
2294 stream->cx_time += vpx_usec_timer_elapsed(&timer);
2295 ctx_exit_on_error(&stream->encoder, "Stream %d: Failed to encode frame",
2296 stream->index);
2297 }
2298
2299
2300 static void update_quantizer_histogram(struct stream_state *stream) {
2301 if (stream->config.cfg.g_pass != VPX_RC_FIRST_PASS) {
2302 int q;
2303
2304 vpx_codec_control(&stream->encoder, VP8E_GET_LAST_QUANTIZER_64, &q);
2305 ctx_exit_on_error(&stream->encoder, "Failed to read quantizer");
2306 stream->counts[q]++;
2307 }
2308 }
2309
2310
2311 static void get_cx_data(struct stream_state *stream,
2312 struct global_config *global,
2313 int *got_data) {
2314 const vpx_codec_cx_pkt_t *pkt;
2315 const struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
2316 vpx_codec_iter_t iter = NULL;
2317
2318 *got_data = 0;
2319 while ((pkt = vpx_codec_get_cx_data(&stream->encoder, &iter))) {
2320 static size_t fsize = 0;
2321 static off_t ivf_header_pos = 0;
2322
2323 switch (pkt->kind) {
2324 case VPX_CODEC_CX_FRAME_PKT:
2325 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
2326 stream->frames_out++;
2327 }
2328 if (!global->quiet)
2329 fprintf(stderr, " %6luF", (unsigned long)pkt->data.frame.sz);
2330
2331 update_rate_histogram(&stream->rate_hist, cfg, pkt);
2332 if (stream->config.write_webm) {
2333 /* Update the hash */
2334 if (!stream->ebml.debug)
2335 stream->hash = murmur(pkt->data.frame.buf,
2336 (int)pkt->data.frame.sz,
2337 stream->hash);
2338
2339 write_webm_block(&stream->ebml, cfg, pkt);
2340 } else {
2341 if (pkt->data.frame.partition_id <= 0) {
2342 ivf_header_pos = ftello(stream->file);
2343 fsize = pkt->data.frame.sz;
2344
2345 write_ivf_frame_header(stream->file, pkt);
2346 } else {
2347 fsize += pkt->data.frame.sz;
2348
2349 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
2350 off_t currpos = ftello(stream->file);
2351 fseeko(stream->file, ivf_header_pos, SEEK_SET);
2352 write_ivf_frame_size(stream->file, fsize);
2353 fseeko(stream->file, currpos, SEEK_SET);
2354 }
2355 }
2356
2357 (void) fwrite(pkt->data.frame.buf, 1, pkt->data.frame.sz,
2358 stream->file);
2359 }
2360 stream->nbytes += pkt->data.raw.sz;
2361
2362 *got_data = 1;
2363 #if CONFIG_DECODERS
2364 if (global->test_decode != TEST_DECODE_OFF && !stream->mismatch_seen) {
2365 vpx_codec_decode(&stream->decoder, pkt->data.frame.buf,
2366 pkt->data.frame.sz, NULL, 0);
2367 if (stream->decoder.err) {
2368 warn_or_exit_on_error(&stream->decoder,
2369 global->test_decode == TEST_DECODE_FATAL,
2370 "Failed to decode frame %d in stream %d",
2371 stream->frames_out + 1, stream->index);
2372 stream->mismatch_seen = stream->frames_out + 1;
2373 }
2374 }
2375 #endif
2376 break;
2377 case VPX_CODEC_STATS_PKT:
2378 stream->frames_out++;
2379 stats_write(&stream->stats,
2380 pkt->data.twopass_stats.buf,
2381 pkt->data.twopass_stats.sz);
2382 stream->nbytes += pkt->data.raw.sz;
2383 break;
2384 case VPX_CODEC_PSNR_PKT:
2385
2386 if (global->show_psnr) {
2387 int i;
2388
2389 stream->psnr_sse_total += pkt->data.psnr.sse[0];
2390 stream->psnr_samples_total += pkt->data.psnr.samples[0];
2391 for (i = 0; i < 4; i++) {
2392 if (!global->quiet)
2393 fprintf(stderr, "%.3f ", pkt->data.psnr.psnr[i]);
2394 stream->psnr_totals[i] += pkt->data.psnr.psnr[i];
2395 }
2396 stream->psnr_count++;
2397 }
2398
2399 break;
2400 default:
2401 break;
2402 }
2403 }
2404 }
2405
2406
2407 static void show_psnr(struct stream_state *stream) {
2408 int i;
2409 double ovpsnr;
2410
2411 if (!stream->psnr_count)
2412 return;
2413
2414 fprintf(stderr, "Stream %d PSNR (Overall/Avg/Y/U/V)", stream->index);
2415 ovpsnr = vp8_mse2psnr((double)stream->psnr_samples_total, 255.0,
2416 (double)stream->psnr_sse_total);
2417 fprintf(stderr, " %.3f", ovpsnr);
2418
2419 for (i = 0; i < 4; i++) {
2420 fprintf(stderr, " %.3f", stream->psnr_totals[i] / stream->psnr_count);
2421 }
2422 fprintf(stderr, "\n");
2423 }
2424
2425
2426 static float usec_to_fps(uint64_t usec, unsigned int frames) {
2427 return (float)(usec > 0 ? frames * 1000000.0 / (float)usec : 0);
2428 }
2429
2430
2431 static void test_decode(struct stream_state *stream,
2432 enum TestDecodeFatality fatal,
2433 const struct codec_item *codec) {
2434 vpx_image_t enc_img, dec_img;
2435
2436 if (stream->mismatch_seen)
2437 return;
2438
2439 /* Get the internal reference frame */
2440 if (codec->fourcc == VP8_FOURCC) {
2441 struct vpx_ref_frame ref_enc, ref_dec;
2442 int width, height;
2443
2444 width = (stream->config.cfg.g_w + 15) & ~15;
2445 height = (stream->config.cfg.g_h + 15) & ~15;
2446 vpx_img_alloc(&ref_enc.img, VPX_IMG_FMT_I420, width, height, 1);
2447 enc_img = ref_enc.img;
2448 vpx_img_alloc(&ref_dec.img, VPX_IMG_FMT_I420, width, height, 1);
2449 dec_img = ref_dec.img;
2450
2451 ref_enc.frame_type = VP8_LAST_FRAME;
2452 ref_dec.frame_type = VP8_LAST_FRAME;
2453 vpx_codec_control(&stream->encoder, VP8_COPY_REFERENCE, &ref_enc);
2454 vpx_codec_control(&stream->decoder, VP8_COPY_REFERENCE, &ref_dec);
2455 } else {
2456 struct vp9_ref_frame ref;
2457
2458 ref.idx = 0;
2459 vpx_codec_control(&stream->encoder, VP9_GET_REFERENCE, &ref);
2460 enc_img = ref.img;
2461 vpx_codec_control(&stream->decoder, VP9_GET_REFERENCE, &ref);
2462 dec_img = ref.img;
2463 }
2464 ctx_exit_on_error(&stream->encoder, "Failed to get encoder reference frame");
2465 ctx_exit_on_error(&stream->decoder, "Failed to get decoder reference frame");
2466
2467 if (!compare_img(&enc_img, &dec_img)) {
2468 int y[4], u[4], v[4];
2469 find_mismatch(&enc_img, &dec_img, y, u, v);
2470 stream->decoder.err = 1;
2471 warn_or_exit_on_error(&stream->decoder, fatal == TEST_DECODE_FATAL,
2472 "Stream %d: Encode/decode mismatch on frame %d at"
2473 " Y[%d, %d] {%d/%d},"
2474 " U[%d, %d] {%d/%d},"
2475 " V[%d, %d] {%d/%d}",
2476 stream->index, stream->frames_out,
2477 y[0], y[1], y[2], y[3],
2478 u[0], u[1], u[2], u[3],
2479 v[0], v[1], v[2], v[3]);
2480 stream->mismatch_seen = stream->frames_out;
2481 }
2482
2483 vpx_img_free(&enc_img);
2484 vpx_img_free(&dec_img);
2485 }
2486
2487
2488 static void print_time(const char *label, int64_t etl) {
2489 int hours, mins, secs;
2490
2491 if (etl >= 0) {
2492 hours = etl / 3600;
2493 etl -= hours * 3600;
2494 mins = etl / 60;
2495 etl -= mins * 60;
2496 secs = etl;
2497
2498 fprintf(stderr, "[%3s %2d:%02d:%02d] ",
2499 label, hours, mins, secs);
2500 } else {
2501 fprintf(stderr, "[%3s unknown] ", label);
2502 }
2503 }
2504
2505 int main(int argc, const char **argv_) {
2506 int pass;
2507 vpx_image_t raw;
2508 int frame_avail, got_data;
2509
2510 struct input_state input = {0};
2511 struct global_config global;
2512 struct stream_state *streams = NULL;
2513 char **argv, **argi;
2514 uint64_t cx_time = 0;
2515 int stream_cnt = 0;
2516 int res = 0;
2517
2518 exec_name = argv_[0];
2519
2520 if (argc < 3)
2521 usage_exit();
2522
2523 /* Setup default input stream settings */
2524 input.framerate.num = 30;
2525 input.framerate.den = 1;
2526 input.use_i420 = 1;
2527 input.only_i420 = 1;
2528
2529 /* First parse the global configuration values, because we want to apply
2530 * other parameters on top of the default configuration provided by the
2531 * codec.
2532 */
2533 argv = argv_dup(argc - 1, argv_ + 1);
2534 parse_global_config(&global, argv);
2535
2536 {
2537 /* Now parse each stream's parameters. Using a local scope here
2538 * due to the use of 'stream' as loop variable in FOREACH_STREAM
2539 * loops
2540 */
2541 struct stream_state *stream = NULL;
2542
2543 do {
2544 stream = new_stream(&global, stream);
2545 stream_cnt++;
2546 if (!streams)
2547 streams = stream;
2548 } while (parse_stream_params(&global, stream, argv));
2549 }
2550
2551 /* Check for unrecognized options */
2552 for (argi = argv; *argi; argi++)
2553 if (argi[0][0] == '-' && argi[0][1])
2554 die("Error: Unrecognized option %s\n", *argi);
2555
2556 /* Handle non-option arguments */
2557 input.fn = argv[0];
2558
2559 if (!input.fn)
2560 usage_exit();
2561
2562 #if CONFIG_NON420
2563 /* Decide if other chroma subsamplings than 4:2:0 are supported */
2564 if (global.codec->fourcc == VP9_FOURCC)
2565 input.only_i420 = 0;
2566 #endif
2567
2568 for (pass = global.pass ? global.pass - 1 : 0; pass < global.passes; pass++) {
2569 int frames_in = 0, seen_frames = 0;
2570 int64_t estimated_time_left = -1;
2571 int64_t average_rate = -1;
2572 off_t lagged_count = 0;
2573
2574 open_input_file(&input);
2575
2576 /* If the input file doesn't specify its w/h (raw files), try to get
2577 * the data from the first stream's configuration.
2578 */
2579 if (!input.w || !input.h)
2580 FOREACH_STREAM( {
2581 if (stream->config.cfg.g_w && stream->config.cfg.g_h) {
2582 input.w = stream->config.cfg.g_w;
2583 input.h = stream->config.cfg.g_h;
2584 break;
2585 }
2586 });
2587
2588 /* Update stream configurations from the input file's parameters */
2589 if (!input.w || !input.h)
2590 fatal("Specify stream dimensions with --width (-w) "
2591 " and --height (-h)");
2592 FOREACH_STREAM(set_stream_dimensions(stream, input.w, input.h));
2593 FOREACH_STREAM(validate_stream_config(stream));
2594
2595 /* Ensure that --passes and --pass are consistent. If --pass is set and
2596 * --passes=2, ensure --fpf was set.
2597 */
2598 if (global.pass && global.passes == 2)
2599 FOREACH_STREAM( {
2600 if (!stream->config.stats_fn)
2601 die("Stream %d: Must specify --fpf when --pass=%d"
2602 " and --passes=2\n", stream->index, global.pass);
2603 });
2604
2605 /* Use the frame rate from the file only if none was specified
2606 * on the command-line.
2607 */
2608 if (!global.have_framerate)
2609 global.framerate = input.framerate;
2610
2611 FOREACH_STREAM(set_default_kf_interval(stream, &global));
2612
2613 /* Show configuration */
2614 if (global.verbose && pass == 0)
2615 FOREACH_STREAM(show_stream_config(stream, &global, &input));
2616
2617 if (pass == (global.pass ? global.pass - 1 : 0)) {
2618 if (input.file_type == FILE_TYPE_Y4M)
2619 /*The Y4M reader does its own allocation.
2620 Just initialize this here to avoid problems if we never read any
2621 frames.*/
2622 memset(&raw, 0, sizeof(raw));
2623 else
2624 vpx_img_alloc(&raw,
2625 input.use_i420 ? VPX_IMG_FMT_I420
2626 : VPX_IMG_FMT_YV12,
2627 input.w, input.h, 32);
2628
2629 FOREACH_STREAM(init_rate_histogram(&stream->rate_hist,
2630 &stream->config.cfg,
2631 &global.framerate));
2632 }
2633
2634 FOREACH_STREAM(open_output_file(stream, &global));
2635 FOREACH_STREAM(setup_pass(stream, &global, pass));
2636 FOREACH_STREAM(initialize_encoder(stream, &global));
2637
2638 frame_avail = 1;
2639 got_data = 0;
2640
2641 while (frame_avail || got_data) {
2642 struct vpx_usec_timer timer;
2643
2644 if (!global.limit || frames_in < global.limit) {
2645 frame_avail = read_frame(&input, &raw);
2646
2647 if (frame_avail)
2648 frames_in++;
2649 seen_frames = frames_in > global.skip_frames ?
2650 frames_in - global.skip_frames : 0;
2651
2652 if (!global.quiet) {
2653 float fps = usec_to_fps(cx_time, seen_frames);
2654 fprintf(stderr, "\rPass %d/%d ", pass + 1, global.passes);
2655
2656 if (stream_cnt == 1)
2657 fprintf(stderr,
2658 "frame %4d/%-4d %7"PRId64"B ",
2659 frames_in, streams->frames_out, (int64_t)streams->nbytes);
2660 else
2661 fprintf(stderr, "frame %4d ", frames_in);
2662
2663 fprintf(stderr, "%7"PRId64" %s %.2f %s ",
2664 cx_time > 9999999 ? cx_time / 1000 : cx_time,
2665 cx_time > 9999999 ? "ms" : "us",
2666 fps >= 1.0 ? fps : 1000.0 / fps,
2667 fps >= 1.0 ? "fps" : "ms/f");
2668 print_time("ETA", estimated_time_left);
2669 fprintf(stderr, "\033[K");
2670 }
2671
2672 } else
2673 frame_avail = 0;
2674
2675 if (frames_in > global.skip_frames) {
2676 vpx_usec_timer_start(&timer);
2677 FOREACH_STREAM(encode_frame(stream, &global,
2678 frame_avail ? &raw : NULL,
2679 frames_in));
2680 vpx_usec_timer_mark(&timer);
2681 cx_time += vpx_usec_timer_elapsed(&timer);
2682
2683 FOREACH_STREAM(update_quantizer_histogram(stream));
2684
2685 got_data = 0;
2686 FOREACH_STREAM(get_cx_data(stream, &global, &got_data));
2687
2688 if (!got_data && input.length && !streams->frames_out) {
2689 lagged_count = global.limit ? seen_frames : ftello(input.file);
2690 } else if (input.length) {
2691 int64_t remaining;
2692 int64_t rate;
2693
2694 if (global.limit) {
2695 int frame_in_lagged = (seen_frames - lagged_count) * 1000;
2696
2697 rate = cx_time ? frame_in_lagged * (int64_t)1000000 / cx_time : 0;
2698 remaining = 1000 * (global.limit - global.skip_frames
2699 - seen_frames + lagged_count);
2700 } else {
2701 off_t input_pos = ftello(input.file);
2702 off_t input_pos_lagged = input_pos - lagged_count;
2703 int64_t limit = input.length;
2704
2705 rate = cx_time ? input_pos_lagged * (int64_t)1000000 / cx_time : 0;
2706 remaining = limit - input_pos + lagged_count;
2707 }
2708
2709 average_rate = (average_rate <= 0)
2710 ? rate
2711 : (average_rate * 7 + rate) / 8;
2712 estimated_time_left = average_rate ? remaining / average_rate : -1;
2713 }
2714
2715 if (got_data && global.test_decode != TEST_DECODE_OFF)
2716 FOREACH_STREAM(test_decode(stream, global.test_decode, global.codec));
2717 }
2718
2719 fflush(stdout);
2720 }
2721
2722 if (stream_cnt > 1)
2723 fprintf(stderr, "\n");
2724
2725 if (!global.quiet)
2726 FOREACH_STREAM(fprintf(
2727 stderr,
2728 "\rPass %d/%d frame %4d/%-4d %7"PRId64"B %7lub/f %7"PRId64"b/s"
2729 " %7"PRId64" %s (%.2f fps)\033[K\n", pass + 1,
2730 global.passes, frames_in, stream->frames_out, (int64_t)stream->nbytes,
2731 seen_frames ? (unsigned long)(stream->nbytes * 8 / seen_frames) : 0,
2732 seen_frames ? (int64_t)stream->nbytes * 8
2733 * (int64_t)global.framerate.num / global.framerate.den
2734 / seen_frames
2735 : 0,
2736 stream->cx_time > 9999999 ? stream->cx_time / 1000 : stream->cx_time,
2737 stream->cx_time > 9999999 ? "ms" : "us",
2738 usec_to_fps(stream->cx_time, seen_frames));
2739 );
2740
2741 if (global.show_psnr)
2742 FOREACH_STREAM(show_psnr(stream));
2743
2744 FOREACH_STREAM(vpx_codec_destroy(&stream->encoder));
2745
2746 if (global.test_decode != TEST_DECODE_OFF) {
2747 FOREACH_STREAM(vpx_codec_destroy(&stream->decoder));
2748 }
2749
2750 close_input_file(&input);
2751
2752 if (global.test_decode == TEST_DECODE_FATAL) {
2753 FOREACH_STREAM(res |= stream->mismatch_seen);
2754 }
2755 FOREACH_STREAM(close_output_file(stream, global.codec->fourcc));
2756
2757 FOREACH_STREAM(stats_close(&stream->stats, global.passes - 1));
2758
2759 if (global.pass)
2760 break;
2761 }
2762
2763 if (global.show_q_hist_buckets)
2764 FOREACH_STREAM(show_q_histogram(stream->counts,
2765 global.show_q_hist_buckets));
2766
2767 if (global.show_rate_hist_buckets)
2768 FOREACH_STREAM(show_rate_histogram(&stream->rate_hist,
2769 &stream->config.cfg,
2770 global.show_rate_hist_buckets));
2771 FOREACH_STREAM(destroy_rate_histogram(&stream->rate_hist));
2772
2773 #if CONFIG_INTERNAL_STATS
2774 /* TODO(jkoleszar): This doesn't belong in this executable. Do it for now,
2775 * to match some existing utilities.
2776 */
2777 FOREACH_STREAM({
2778 FILE *f = fopen("opsnr.stt", "a");
2779 if (stream->mismatch_seen) {
2780 fprintf(f, "First mismatch occurred in frame %d\n",
2781 stream->mismatch_seen);
2782 } else {
2783 fprintf(f, "No mismatch detected in recon buffers\n");
2784 }
2785 fclose(f);
2786 });
2787 #endif
2788
2789 vpx_img_free(&raw);
2790 free(argv);
2791 free(streams);
2792 return res ? EXIT_FAILURE : EXIT_SUCCESS;
2793 }
AV1 Codec Library