?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Removed shadow warnings : postproc.c decodframe.c threading.c
[aom.git] / vpxenc.c
blob5c1819468db57d323dd0079863b6a978ae3864f2
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 (0x00385056)
93 #define VP9_FOURCC (0x00395056)
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 };
330
331
332 #define IVF_FRAME_HDR_SZ (4+8) /* 4 byte size + 8 byte timestamp */
333 static int read_frame(struct input_state *input, vpx_image_t *img) {
334 FILE *f = input->file;
335 enum video_file_type file_type = input->file_type;
336 y4m_input *y4m = &input->y4m;
337 struct detect_buffer *detect = &input->detect;
338 int plane = 0;
339 int shortread = 0;
340
341 if (file_type == FILE_TYPE_Y4M) {
342 if (y4m_input_fetch_frame(y4m, f, img) < 1)
343 return 0;
344 } else {
345 if (file_type == FILE_TYPE_IVF) {
346 char junk[IVF_FRAME_HDR_SZ];
347
348 /* Skip the frame header. We know how big the frame should be. See
349 * write_ivf_frame_header() for documentation on the frame header
350 * layout.
351 */
352 (void) fread(junk, 1, IVF_FRAME_HDR_SZ, f);
353 }
354
355 for (plane = 0; plane < 3; plane++) {
356 unsigned char *ptr;
357 int w = (plane ? (1 + img->d_w) / 2 : img->d_w);
358 int h = (plane ? (1 + img->d_h) / 2 : img->d_h);
359 int r;
360
361 /* Determine the correct plane based on the image format. The for-loop
362 * always counts in Y,U,V order, but this may not match the order of
363 * the data on disk.
364 */
365 switch (plane) {
366 case 1:
367 ptr = img->planes[img->fmt == VPX_IMG_FMT_YV12 ? VPX_PLANE_V : VPX_PLANE_U];
368 break;
369 case 2:
370 ptr = img->planes[img->fmt == VPX_IMG_FMT_YV12 ? VPX_PLANE_U : VPX_PLANE_V];
371 break;
372 default:
373 ptr = img->planes[plane];
374 }
375
376 for (r = 0; r < h; r++) {
377 size_t needed = w;
378 size_t buf_position = 0;
379 const size_t left = detect->buf_read - detect->position;
380 if (left > 0) {
381 const size_t more = (left < needed) ? left : needed;
382 memcpy(ptr, detect->buf + detect->position, more);
383 buf_position = more;
384 needed -= more;
385 detect->position += more;
386 }
387 if (needed > 0) {
388 shortread |= (fread(ptr + buf_position, 1, needed, f) < needed);
389 }
390
391 ptr += img->stride[plane];
392 }
393 }
394 }
395
396 return !shortread;
397 }
398
399
400 unsigned int file_is_y4m(FILE *infile,
401 y4m_input *y4m,
402 char detect[4]) {
403 if (memcmp(detect, "YUV4", 4) == 0) {
404 return 1;
405 }
406 return 0;
407 }
408
409 #define IVF_FILE_HDR_SZ (32)
410 unsigned int file_is_ivf(struct input_state *input,
411 unsigned int *fourcc) {
412 char raw_hdr[IVF_FILE_HDR_SZ];
413 int is_ivf = 0;
414 FILE *infile = input->file;
415 unsigned int *width = &input->w;
416 unsigned int *height = &input->h;
417 struct detect_buffer *detect = &input->detect;
418
419 if (memcmp(detect->buf, "DKIF", 4) != 0)
420 return 0;
421
422 /* See write_ivf_file_header() for more documentation on the file header
423 * layout.
424 */
425 if (fread(raw_hdr + 4, 1, IVF_FILE_HDR_SZ - 4, infile)
426 == IVF_FILE_HDR_SZ - 4) {
427 {
428 is_ivf = 1;
429
430 if (mem_get_le16(raw_hdr + 4) != 0)
431 warn("Unrecognized IVF version! This file may not decode "
432 "properly.");
433
434 *fourcc = mem_get_le32(raw_hdr + 8);
435 }
436 }
437
438 if (is_ivf) {
439 *width = mem_get_le16(raw_hdr + 12);
440 *height = mem_get_le16(raw_hdr + 14);
441 detect->position = 4;
442 }
443
444 return is_ivf;
445 }
446
447
448 static void write_ivf_file_header(FILE *outfile,
449 const vpx_codec_enc_cfg_t *cfg,
450 unsigned int fourcc,
451 int frame_cnt) {
452 char header[32];
453
454 if (cfg->g_pass != VPX_RC_ONE_PASS && cfg->g_pass != VPX_RC_LAST_PASS)
455 return;
456
457 header[0] = 'D';
458 header[1] = 'K';
459 header[2] = 'I';
460 header[3] = 'F';
461 mem_put_le16(header + 4, 0); /* version */
462 mem_put_le16(header + 6, 32); /* headersize */
463 mem_put_le32(header + 8, fourcc); /* headersize */
464 mem_put_le16(header + 12, cfg->g_w); /* width */
465 mem_put_le16(header + 14, cfg->g_h); /* height */
466 mem_put_le32(header + 16, cfg->g_timebase.den); /* rate */
467 mem_put_le32(header + 20, cfg->g_timebase.num); /* scale */
468 mem_put_le32(header + 24, frame_cnt); /* length */
469 mem_put_le32(header + 28, 0); /* unused */
470
471 (void) fwrite(header, 1, 32, outfile);
472 }
473
474
475 static void write_ivf_frame_header(FILE *outfile,
476 const vpx_codec_cx_pkt_t *pkt) {
477 char header[12];
478 vpx_codec_pts_t pts;
479
480 if (pkt->kind != VPX_CODEC_CX_FRAME_PKT)
481 return;
482
483 pts = pkt->data.frame.pts;
484 mem_put_le32(header, (int)pkt->data.frame.sz);
485 mem_put_le32(header + 4, pts & 0xFFFFFFFF);
486 mem_put_le32(header + 8, pts >> 32);
487
488 (void) fwrite(header, 1, 12, outfile);
489 }
490
491 static void write_ivf_frame_size(FILE *outfile, size_t size) {
492 char header[4];
493 mem_put_le32(header, (int)size);
494 (void) fwrite(header, 1, 4, outfile);
495 }
496
497
498 typedef off_t EbmlLoc;
499
500
501 struct cue_entry {
502 unsigned int time;
503 uint64_t loc;
504 };
505
506
507 struct EbmlGlobal {
508 int debug;
509
510 FILE *stream;
511 int64_t last_pts_ms;
512 vpx_rational_t framerate;
513
514 /* These pointers are to the start of an element */
515 off_t position_reference;
516 off_t seek_info_pos;
517 off_t segment_info_pos;
518 off_t track_pos;
519 off_t cue_pos;
520 off_t cluster_pos;
521
522 /* This pointer is to a specific element to be serialized */
523 off_t track_id_pos;
524
525 /* These pointers are to the size field of the element */
526 EbmlLoc startSegment;
527 EbmlLoc startCluster;
528
529 uint32_t cluster_timecode;
530 int cluster_open;
531
532 struct cue_entry *cue_list;
533 unsigned int cues;
534
535 };
536
537
538 void Ebml_Write(EbmlGlobal *glob, const void *buffer_in, unsigned long len) {
539 (void) fwrite(buffer_in, 1, len, glob->stream);
540 }
541
542 #define WRITE_BUFFER(s) \
543 for(i = len-1; i>=0; i--)\
544 { \
545 x = (char)(*(const s *)buffer_in >> (i * CHAR_BIT)); \
546 Ebml_Write(glob, &x, 1); \
547 }
548 void Ebml_Serialize(EbmlGlobal *glob, const void *buffer_in, int buffer_size, unsigned long len) {
549 char x;
550 int i;
551
552 /* buffer_size:
553 * 1 - int8_t;
554 * 2 - int16_t;
555 * 3 - int32_t;
556 * 4 - int64_t;
557 */
558 switch (buffer_size) {
559 case 1:
560 WRITE_BUFFER(int8_t)
561 break;
562 case 2:
563 WRITE_BUFFER(int16_t)
564 break;
565 case 4:
566 WRITE_BUFFER(int32_t)
567 break;
568 case 8:
569 WRITE_BUFFER(int64_t)
570 break;
571 default:
572 break;
573 }
574 }
575 #undef WRITE_BUFFER
576
577 /* Need a fixed size serializer for the track ID. libmkv provides a 64 bit
578 * one, but not a 32 bit one.
579 */
580 static void Ebml_SerializeUnsigned32(EbmlGlobal *glob, unsigned long class_id, uint64_t ui) {
581 unsigned char sizeSerialized = 4 | 0x80;
582 Ebml_WriteID(glob, class_id);
583 Ebml_Serialize(glob, &sizeSerialized, sizeof(sizeSerialized), 1);
584 Ebml_Serialize(glob, &ui, sizeof(ui), 4);
585 }
586
587
588 static void
589 Ebml_StartSubElement(EbmlGlobal *glob, EbmlLoc *ebmlLoc,
590 unsigned long class_id) {
591 /* todo this is always taking 8 bytes, this may need later optimization */
592 /* this is a key that says length unknown */
593 uint64_t unknownLen = LITERALU64(0x01FFFFFF, 0xFFFFFFFF);
594
595 Ebml_WriteID(glob, class_id);
596 *ebmlLoc = ftello(glob->stream);
597 Ebml_Serialize(glob, &unknownLen, sizeof(unknownLen), 8);
598 }
599
600 static void
601 Ebml_EndSubElement(EbmlGlobal *glob, EbmlLoc *ebmlLoc) {
602 off_t pos;
603 uint64_t size;
604
605 /* Save the current stream pointer */
606 pos = ftello(glob->stream);
607
608 /* Calculate the size of this element */
609 size = pos - *ebmlLoc - 8;
610 size |= LITERALU64(0x01000000, 0x00000000);
611
612 /* Seek back to the beginning of the element and write the new size */
613 fseeko(glob->stream, *ebmlLoc, SEEK_SET);
614 Ebml_Serialize(glob, &size, sizeof(size), 8);
615
616 /* Reset the stream pointer */
617 fseeko(glob->stream, pos, SEEK_SET);
618 }
619
620
621 static void
622 write_webm_seek_element(EbmlGlobal *ebml, unsigned long id, off_t pos) {
623 uint64_t offset = pos - ebml->position_reference;
624 EbmlLoc start;
625 Ebml_StartSubElement(ebml, &start, Seek);
626 Ebml_SerializeBinary(ebml, SeekID, id);
627 Ebml_SerializeUnsigned64(ebml, SeekPosition, offset);
628 Ebml_EndSubElement(ebml, &start);
629 }
630
631
632 static void
633 write_webm_seek_info(EbmlGlobal *ebml) {
634
635 off_t pos;
636
637 /* Save the current stream pointer */
638 pos = ftello(ebml->stream);
639
640 if (ebml->seek_info_pos)
641 fseeko(ebml->stream, ebml->seek_info_pos, SEEK_SET);
642 else
643 ebml->seek_info_pos = pos;
644
645 {
646 EbmlLoc start;
647
648 Ebml_StartSubElement(ebml, &start, SeekHead);
649 write_webm_seek_element(ebml, Tracks, ebml->track_pos);
650 write_webm_seek_element(ebml, Cues, ebml->cue_pos);
651 write_webm_seek_element(ebml, Info, ebml->segment_info_pos);
652 Ebml_EndSubElement(ebml, &start);
653 }
654 {
655 /* segment info */
656 EbmlLoc startInfo;
657 uint64_t frame_time;
658 char version_string[64];
659
660 /* Assemble version string */
661 if (ebml->debug)
662 strcpy(version_string, "vpxenc");
663 else {
664 strcpy(version_string, "vpxenc ");
665 strncat(version_string,
666 vpx_codec_version_str(),
667 sizeof(version_string) - 1 - strlen(version_string));
668 }
669
670 frame_time = (uint64_t)1000 * ebml->framerate.den
671 / ebml->framerate.num;
672 ebml->segment_info_pos = ftello(ebml->stream);
673 Ebml_StartSubElement(ebml, &startInfo, Info);
674 Ebml_SerializeUnsigned(ebml, TimecodeScale, 1000000);
675 Ebml_SerializeFloat(ebml, Segment_Duration,
676 (double)(ebml->last_pts_ms + frame_time));
677 Ebml_SerializeString(ebml, 0x4D80, version_string);
678 Ebml_SerializeString(ebml, 0x5741, version_string);
679 Ebml_EndSubElement(ebml, &startInfo);
680 }
681 }
682
683
684 static void
685 write_webm_file_header(EbmlGlobal *glob,
686 const vpx_codec_enc_cfg_t *cfg,
687 const struct vpx_rational *fps,
688 stereo_format_t stereo_fmt,
689 unsigned int fourcc) {
690 {
691 EbmlLoc start;
692 Ebml_StartSubElement(glob, &start, EBML);
693 Ebml_SerializeUnsigned(glob, EBMLVersion, 1);
694 Ebml_SerializeUnsigned(glob, EBMLReadVersion, 1);
695 Ebml_SerializeUnsigned(glob, EBMLMaxIDLength, 4);
696 Ebml_SerializeUnsigned(glob, EBMLMaxSizeLength, 8);
697 Ebml_SerializeString(glob, DocType, "webm");
698 Ebml_SerializeUnsigned(glob, DocTypeVersion, 2);
699 Ebml_SerializeUnsigned(glob, DocTypeReadVersion, 2);
700 Ebml_EndSubElement(glob, &start);
701 }
702 {
703 Ebml_StartSubElement(glob, &glob->startSegment, Segment);
704 glob->position_reference = ftello(glob->stream);
705 glob->framerate = *fps;
706 write_webm_seek_info(glob);
707
708 {
709 EbmlLoc trackStart;
710 glob->track_pos = ftello(glob->stream);
711 Ebml_StartSubElement(glob, &trackStart, Tracks);
712 {
713 unsigned int trackNumber = 1;
714 uint64_t trackID = 0;
715
716 EbmlLoc start;
717 Ebml_StartSubElement(glob, &start, TrackEntry);
718 Ebml_SerializeUnsigned(glob, TrackNumber, trackNumber);
719 glob->track_id_pos = ftello(glob->stream);
720 Ebml_SerializeUnsigned32(glob, TrackUID, trackID);
721 Ebml_SerializeUnsigned(glob, TrackType, 1);
722 Ebml_SerializeString(glob, CodecID,
723 fourcc == VP8_FOURCC ? "V_VP8" : "V_VP9");
724 {
725 unsigned int pixelWidth = cfg->g_w;
726 unsigned int pixelHeight = cfg->g_h;
727 float frameRate = (float)fps->num / (float)fps->den;
728
729 EbmlLoc videoStart;
730 Ebml_StartSubElement(glob, &videoStart, Video);
731 Ebml_SerializeUnsigned(glob, PixelWidth, pixelWidth);
732 Ebml_SerializeUnsigned(glob, PixelHeight, pixelHeight);
733 Ebml_SerializeUnsigned(glob, StereoMode, stereo_fmt);
734 Ebml_SerializeFloat(glob, FrameRate, frameRate);
735 Ebml_EndSubElement(glob, &videoStart);
736 }
737 Ebml_EndSubElement(glob, &start); /* Track Entry */
738 }
739 Ebml_EndSubElement(glob, &trackStart);
740 }
741 /* segment element is open */
742 }
743 }
744
745
746 static void
747 write_webm_block(EbmlGlobal *glob,
748 const vpx_codec_enc_cfg_t *cfg,
749 const vpx_codec_cx_pkt_t *pkt) {
750 unsigned long block_length;
751 unsigned char track_number;
752 unsigned short block_timecode = 0;
753 unsigned char flags;
754 int64_t pts_ms;
755 int start_cluster = 0, is_keyframe;
756
757 /* Calculate the PTS of this frame in milliseconds */
758 pts_ms = pkt->data.frame.pts * 1000
759 * (uint64_t)cfg->g_timebase.num / (uint64_t)cfg->g_timebase.den;
760 if (pts_ms <= glob->last_pts_ms)
761 pts_ms = glob->last_pts_ms + 1;
762 glob->last_pts_ms = pts_ms;
763
764 /* Calculate the relative time of this block */
765 if (pts_ms - glob->cluster_timecode > SHRT_MAX)
766 start_cluster = 1;
767 else
768 block_timecode = (unsigned short)pts_ms - glob->cluster_timecode;
769
770 is_keyframe = (pkt->data.frame.flags & VPX_FRAME_IS_KEY);
771 if (start_cluster || is_keyframe) {
772 if (glob->cluster_open)
773 Ebml_EndSubElement(glob, &glob->startCluster);
774
775 /* Open the new cluster */
776 block_timecode = 0;
777 glob->cluster_open = 1;
778 glob->cluster_timecode = (uint32_t)pts_ms;
779 glob->cluster_pos = ftello(glob->stream);
780 Ebml_StartSubElement(glob, &glob->startCluster, Cluster); /* cluster */
781 Ebml_SerializeUnsigned(glob, Timecode, glob->cluster_timecode);
782
783 /* Save a cue point if this is a keyframe. */
784 if (is_keyframe) {
785 struct cue_entry *cue, *new_cue_list;
786
787 new_cue_list = realloc(glob->cue_list,
788 (glob->cues + 1) * sizeof(struct cue_entry));
789 if (new_cue_list)
790 glob->cue_list = new_cue_list;
791 else
792 fatal("Failed to realloc cue list.");
793
794 cue = &glob->cue_list[glob->cues];
795 cue->time = glob->cluster_timecode;
796 cue->loc = glob->cluster_pos;
797 glob->cues++;
798 }
799 }
800
801 /* Write the Simple Block */
802 Ebml_WriteID(glob, SimpleBlock);
803
804 block_length = (unsigned long)pkt->data.frame.sz + 4;
805 block_length |= 0x10000000;
806 Ebml_Serialize(glob, &block_length, sizeof(block_length), 4);
807
808 track_number = 1;
809 track_number |= 0x80;
810 Ebml_Write(glob, &track_number, 1);
811
812 Ebml_Serialize(glob, &block_timecode, sizeof(block_timecode), 2);
813
814 flags = 0;
815 if (is_keyframe)
816 flags |= 0x80;
817 if (pkt->data.frame.flags & VPX_FRAME_IS_INVISIBLE)
818 flags |= 0x08;
819 Ebml_Write(glob, &flags, 1);
820
821 Ebml_Write(glob, pkt->data.frame.buf, (unsigned long)pkt->data.frame.sz);
822 }
823
824
825 static void
826 write_webm_file_footer(EbmlGlobal *glob, long hash) {
827
828 if (glob->cluster_open)
829 Ebml_EndSubElement(glob, &glob->startCluster);
830
831 {
832 EbmlLoc start;
833 unsigned int i;
834
835 glob->cue_pos = ftello(glob->stream);
836 Ebml_StartSubElement(glob, &start, Cues);
837 for (i = 0; i < glob->cues; i++) {
838 struct cue_entry *cue = &glob->cue_list[i];
839 EbmlLoc start;
840
841 Ebml_StartSubElement(glob, &start, CuePoint);
842 {
843 EbmlLoc start;
844
845 Ebml_SerializeUnsigned(glob, CueTime, cue->time);
846
847 Ebml_StartSubElement(glob, &start, CueTrackPositions);
848 Ebml_SerializeUnsigned(glob, CueTrack, 1);
849 Ebml_SerializeUnsigned64(glob, CueClusterPosition,
850 cue->loc - glob->position_reference);
851 Ebml_EndSubElement(glob, &start);
852 }
853 Ebml_EndSubElement(glob, &start);
854 }
855 Ebml_EndSubElement(glob, &start);
856 }
857
858 Ebml_EndSubElement(glob, &glob->startSegment);
859
860 /* Patch up the seek info block */
861 write_webm_seek_info(glob);
862
863 /* Patch up the track id */
864 fseeko(glob->stream, glob->track_id_pos, SEEK_SET);
865 Ebml_SerializeUnsigned32(glob, TrackUID, glob->debug ? 0xDEADBEEF : hash);
866
867 fseeko(glob->stream, 0, SEEK_END);
868 }
869
870
871 /* Murmur hash derived from public domain reference implementation at
872 * http:// sites.google.com/site/murmurhash/
873 */
874 static unsigned int murmur(const void *key, int len, unsigned int seed) {
875 const unsigned int m = 0x5bd1e995;
876 const int r = 24;
877
878 unsigned int h = seed ^ len;
879
880 const unsigned char *data = (const unsigned char *)key;
881
882 while (len >= 4) {
883 unsigned int k;
884
885 k = data[0];
886 k |= data[1] << 8;
887 k |= data[2] << 16;
888 k |= data[3] << 24;
889
890 k *= m;
891 k ^= k >> r;
892 k *= m;
893
894 h *= m;
895 h ^= k;
896
897 data += 4;
898 len -= 4;
899 }
900
901 switch (len) {
902 case 3:
903 h ^= data[2] << 16;
904 case 2:
905 h ^= data[1] << 8;
906 case 1:
907 h ^= data[0];
908 h *= m;
909 };
910
911 h ^= h >> 13;
912 h *= m;
913 h ^= h >> 15;
914
915 return h;
916 }
917
918 #include "math.h"
919 #define MAX_PSNR 100
920 static double vp8_mse2psnr(double Samples, double Peak, double Mse) {
921 double psnr;
922
923 if ((double)Mse > 0.0)
924 psnr = 10.0 * log10(Peak * Peak * Samples / Mse);
925 else
926 psnr = MAX_PSNR; /* Limit to prevent / 0 */
927
928 if (psnr > MAX_PSNR)
929 psnr = MAX_PSNR;
930
931 return psnr;
932 }
933
934
935 #include "args.h"
936 static const arg_def_t debugmode = ARG_DEF("D", "debug", 0,
937 "Debug mode (makes output deterministic)");
938 static const arg_def_t outputfile = ARG_DEF("o", "output", 1,
939 "Output filename");
940 static const arg_def_t use_yv12 = ARG_DEF(NULL, "yv12", 0,
941 "Input file is YV12 ");
942 static const arg_def_t use_i420 = ARG_DEF(NULL, "i420", 0,
943 "Input file is I420 (default)");
944 static const arg_def_t codecarg = ARG_DEF(NULL, "codec", 1,
945 "Codec to use");
946 static const arg_def_t passes = ARG_DEF("p", "passes", 1,
947 "Number of passes (1/2)");
948 static const arg_def_t pass_arg = ARG_DEF(NULL, "pass", 1,
949 "Pass to execute (1/2)");
950 static const arg_def_t fpf_name = ARG_DEF(NULL, "fpf", 1,
951 "First pass statistics file name");
952 static const arg_def_t limit = ARG_DEF(NULL, "limit", 1,
953 "Stop encoding after n input frames");
954 static const arg_def_t skip = ARG_DEF(NULL, "skip", 1,
955 "Skip the first n input frames");
956 static const arg_def_t deadline = ARG_DEF("d", "deadline", 1,
957 "Deadline per frame (usec)");
958 static const arg_def_t best_dl = ARG_DEF(NULL, "best", 0,
959 "Use Best Quality Deadline");
960 static const arg_def_t good_dl = ARG_DEF(NULL, "good", 0,
961 "Use Good Quality Deadline");
962 static const arg_def_t rt_dl = ARG_DEF(NULL, "rt", 0,
963 "Use Realtime Quality Deadline");
964 static const arg_def_t quietarg = ARG_DEF("q", "quiet", 0,
965 "Do not print encode progress");
966 static const arg_def_t verbosearg = ARG_DEF("v", "verbose", 0,
967 "Show encoder parameters");
968 static const arg_def_t psnrarg = ARG_DEF(NULL, "psnr", 0,
969 "Show PSNR in status line");
970 enum TestDecodeFatality {
971 TEST_DECODE_OFF,
972 TEST_DECODE_FATAL,
973 TEST_DECODE_WARN,
974 };
975 static const struct arg_enum_list test_decode_enum[] = {
976 {"off", TEST_DECODE_OFF},
977 {"fatal", TEST_DECODE_FATAL},
978 {"warn", TEST_DECODE_WARN},
979 {NULL, 0}
980 };
981 static const arg_def_t recontest = ARG_DEF_ENUM(NULL, "test-decode", 1,
982 "Test encode/decode mismatch",
983 test_decode_enum);
984 static const arg_def_t framerate = ARG_DEF(NULL, "fps", 1,
985 "Stream frame rate (rate/scale)");
986 static const arg_def_t use_ivf = ARG_DEF(NULL, "ivf", 0,
987 "Output IVF (default is WebM)");
988 static const arg_def_t out_part = ARG_DEF("P", "output-partitions", 0,
989 "Makes encoder output partitions. Requires IVF output!");
990 static const arg_def_t q_hist_n = ARG_DEF(NULL, "q-hist", 1,
991 "Show quantizer histogram (n-buckets)");
992 static const arg_def_t rate_hist_n = ARG_DEF(NULL, "rate-hist", 1,
993 "Show rate histogram (n-buckets)");
994 static const arg_def_t *main_args[] = {
995 &debugmode,
996 &outputfile, &codecarg, &passes, &pass_arg, &fpf_name, &limit, &skip,
997 &deadline, &best_dl, &good_dl, &rt_dl,
998 &quietarg, &verbosearg, &psnrarg, &use_ivf, &out_part, &q_hist_n, &rate_hist_n,
999 NULL
1000 };
1001
1002 static const arg_def_t usage = ARG_DEF("u", "usage", 1,
1003 "Usage profile number to use");
1004 static const arg_def_t threads = ARG_DEF("t", "threads", 1,
1005 "Max number of threads to use");
1006 static const arg_def_t profile = ARG_DEF(NULL, "profile", 1,
1007 "Bitstream profile number to use");
1008 static const arg_def_t width = ARG_DEF("w", "width", 1,
1009 "Frame width");
1010 static const arg_def_t height = ARG_DEF("h", "height", 1,
1011 "Frame height");
1012 static const struct arg_enum_list stereo_mode_enum[] = {
1013 {"mono", STEREO_FORMAT_MONO},
1014 {"left-right", STEREO_FORMAT_LEFT_RIGHT},
1015 {"bottom-top", STEREO_FORMAT_BOTTOM_TOP},
1016 {"top-bottom", STEREO_FORMAT_TOP_BOTTOM},
1017 {"right-left", STEREO_FORMAT_RIGHT_LEFT},
1018 {NULL, 0}
1019 };
1020 static const arg_def_t stereo_mode = ARG_DEF_ENUM(NULL, "stereo-mode", 1,
1021 "Stereo 3D video format", stereo_mode_enum);
1022 static const arg_def_t timebase = ARG_DEF(NULL, "timebase", 1,
1023 "Output timestamp precision (fractional seconds)");
1024 static const arg_def_t error_resilient = ARG_DEF(NULL, "error-resilient", 1,
1025 "Enable error resiliency features");
1026 static const arg_def_t lag_in_frames = ARG_DEF(NULL, "lag-in-frames", 1,
1027 "Max number of frames to lag");
1028
1029 static const arg_def_t *global_args[] = {
1030 &use_yv12, &use_i420, &usage, &threads, &profile,
1031 &width, &height, &stereo_mode, &timebase, &framerate, &error_resilient,
1032 &lag_in_frames, NULL
1033 };
1034
1035 static const arg_def_t dropframe_thresh = ARG_DEF(NULL, "drop-frame", 1,
1036 "Temporal resampling threshold (buf %)");
1037 static const arg_def_t resize_allowed = ARG_DEF(NULL, "resize-allowed", 1,
1038 "Spatial resampling enabled (bool)");
1039 static const arg_def_t resize_up_thresh = ARG_DEF(NULL, "resize-up", 1,
1040 "Upscale threshold (buf %)");
1041 static const arg_def_t resize_down_thresh = ARG_DEF(NULL, "resize-down", 1,
1042 "Downscale threshold (buf %)");
1043 static const struct arg_enum_list end_usage_enum[] = {
1044 {"vbr", VPX_VBR},
1045 {"cbr", VPX_CBR},
1046 {"cq", VPX_CQ},
1047 {NULL, 0}
1048 };
1049 static const arg_def_t end_usage = ARG_DEF_ENUM(NULL, "end-usage", 1,
1050 "Rate control mode", end_usage_enum);
1051 static const arg_def_t target_bitrate = ARG_DEF(NULL, "target-bitrate", 1,
1052 "Bitrate (kbps)");
1053 static const arg_def_t min_quantizer = ARG_DEF(NULL, "min-q", 1,
1054 "Minimum (best) quantizer");
1055 static const arg_def_t max_quantizer = ARG_DEF(NULL, "max-q", 1,
1056 "Maximum (worst) quantizer");
1057 static const arg_def_t undershoot_pct = ARG_DEF(NULL, "undershoot-pct", 1,
1058 "Datarate undershoot (min) target (%)");
1059 static const arg_def_t overshoot_pct = ARG_DEF(NULL, "overshoot-pct", 1,
1060 "Datarate overshoot (max) target (%)");
1061 static const arg_def_t buf_sz = ARG_DEF(NULL, "buf-sz", 1,
1062 "Client buffer size (ms)");
1063 static const arg_def_t buf_initial_sz = ARG_DEF(NULL, "buf-initial-sz", 1,
1064 "Client initial buffer size (ms)");
1065 static const arg_def_t buf_optimal_sz = ARG_DEF(NULL, "buf-optimal-sz", 1,
1066 "Client optimal buffer size (ms)");
1067 static const arg_def_t *rc_args[] = {
1068 &dropframe_thresh, &resize_allowed, &resize_up_thresh, &resize_down_thresh,
1069 &end_usage, &target_bitrate, &min_quantizer, &max_quantizer,
1070 &undershoot_pct, &overshoot_pct, &buf_sz, &buf_initial_sz, &buf_optimal_sz,
1071 NULL
1072 };
1073
1074
1075 static const arg_def_t bias_pct = ARG_DEF(NULL, "bias-pct", 1,
1076 "CBR/VBR bias (0=CBR, 100=VBR)");
1077 static const arg_def_t minsection_pct = ARG_DEF(NULL, "minsection-pct", 1,
1078 "GOP min bitrate (% of target)");
1079 static const arg_def_t maxsection_pct = ARG_DEF(NULL, "maxsection-pct", 1,
1080 "GOP max bitrate (% of target)");
1081 static const arg_def_t *rc_twopass_args[] = {
1082 &bias_pct, &minsection_pct, &maxsection_pct, NULL
1083 };
1084
1085
1086 static const arg_def_t kf_min_dist = ARG_DEF(NULL, "kf-min-dist", 1,
1087 "Minimum keyframe interval (frames)");
1088 static const arg_def_t kf_max_dist = ARG_DEF(NULL, "kf-max-dist", 1,
1089 "Maximum keyframe interval (frames)");
1090 static const arg_def_t kf_disabled = ARG_DEF(NULL, "disable-kf", 0,
1091 "Disable keyframe placement");
1092 static const arg_def_t *kf_args[] = {
1093 &kf_min_dist, &kf_max_dist, &kf_disabled, NULL
1094 };
1095
1096
1097 static const arg_def_t noise_sens = ARG_DEF(NULL, "noise-sensitivity", 1,
1098 "Noise sensitivity (frames to blur)");
1099 static const arg_def_t sharpness = ARG_DEF(NULL, "sharpness", 1,
1100 "Filter sharpness (0-7)");
1101 static const arg_def_t static_thresh = ARG_DEF(NULL, "static-thresh", 1,
1102 "Motion detection threshold");
1103 static const arg_def_t cpu_used = ARG_DEF(NULL, "cpu-used", 1,
1104 "CPU Used (-16..16)");
1105 static const arg_def_t token_parts = ARG_DEF(NULL, "token-parts", 1,
1106 "Number of token partitions to use, log2");
1107 static const arg_def_t auto_altref = ARG_DEF(NULL, "auto-alt-ref", 1,
1108 "Enable automatic alt reference frames");
1109 static const arg_def_t arnr_maxframes = ARG_DEF(NULL, "arnr-maxframes", 1,
1110 "AltRef Max Frames");
1111 static const arg_def_t arnr_strength = ARG_DEF(NULL, "arnr-strength", 1,
1112 "AltRef Strength");
1113 static const arg_def_t arnr_type = ARG_DEF(NULL, "arnr-type", 1,
1114 "AltRef Type");
1115 static const struct arg_enum_list tuning_enum[] = {
1116 {"psnr", VP8_TUNE_PSNR},
1117 {"ssim", VP8_TUNE_SSIM},
1118 {NULL, 0}
1119 };
1120 static const arg_def_t tune_ssim = ARG_DEF_ENUM(NULL, "tune", 1,
1121 "Material to favor", tuning_enum);
1122 static const arg_def_t cq_level = ARG_DEF(NULL, "cq-level", 1,
1123 "Constrained Quality Level");
1124 static const arg_def_t max_intra_rate_pct = ARG_DEF(NULL, "max-intra-rate", 1,
1125 "Max I-frame bitrate (pct)");
1126 #if CONFIG_LOSSLESS
1127 static const arg_def_t lossless = ARG_DEF(NULL, "lossless", 1, "Lossless mode");
1128 #endif
1129
1130 #if CONFIG_VP8_ENCODER
1131 static const arg_def_t *vp8_args[] = {
1132 &cpu_used, &auto_altref, &noise_sens, &sharpness, &static_thresh,
1133 &token_parts, &arnr_maxframes, &arnr_strength, &arnr_type,
1134 &tune_ssim, &cq_level, &max_intra_rate_pct,
1135 NULL
1136 };
1137 static const int vp8_arg_ctrl_map[] = {
1138 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
1139 VP8E_SET_NOISE_SENSITIVITY, VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
1140 VP8E_SET_TOKEN_PARTITIONS,
1141 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
1142 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
1143 0
1144 };
1145 #endif
1146
1147 #if CONFIG_VP9_ENCODER
1148 static const arg_def_t *vp9_args[] = {
1149 &cpu_used, &auto_altref, &noise_sens, &sharpness, &static_thresh,
1150 &token_parts, &arnr_maxframes, &arnr_strength, &arnr_type,
1151 &tune_ssim, &cq_level, &max_intra_rate_pct,
1152 #if CONFIG_LOSSLESS
1153 &lossless,
1154 #endif
1155 NULL
1156 };
1157 static const int vp9_arg_ctrl_map[] = {
1158 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
1159 VP8E_SET_NOISE_SENSITIVITY, VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
1160 VP8E_SET_TOKEN_PARTITIONS,
1161 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
1162 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
1163 #if CONFIG_LOSSLESS
1164 VP9E_SET_LOSSLESS,
1165 #endif
1166 0
1167 };
1168 #endif
1169
1170 static const arg_def_t *no_args[] = { NULL };
1171
1172 static void usage_exit() {
1173 int i;
1174
1175 fprintf(stderr, "Usage: %s <options> -o dst_filename src_filename \n",
1176 exec_name);
1177
1178 fprintf(stderr, "\nOptions:\n");
1179 arg_show_usage(stdout, main_args);
1180 fprintf(stderr, "\nEncoder Global Options:\n");
1181 arg_show_usage(stdout, global_args);
1182 fprintf(stderr, "\nRate Control Options:\n");
1183 arg_show_usage(stdout, rc_args);
1184 fprintf(stderr, "\nTwopass Rate Control Options:\n");
1185 arg_show_usage(stdout, rc_twopass_args);
1186 fprintf(stderr, "\nKeyframe Placement Options:\n");
1187 arg_show_usage(stdout, kf_args);
1188 #if CONFIG_VP8_ENCODER
1189 fprintf(stderr, "\nVP8 Specific Options:\n");
1190 arg_show_usage(stdout, vp8_args);
1191 #endif
1192 #if CONFIG_VP9_ENCODER
1193 fprintf(stderr, "\nVP9 Specific Options:\n");
1194 arg_show_usage(stdout, vp9_args);
1195 #endif
1196 fprintf(stderr, "\nStream timebase (--timebase):\n"
1197 " The desired precision of timestamps in the output, expressed\n"
1198 " in fractional seconds. Default is 1/1000.\n");
1199 fprintf(stderr, "\n"
1200 "Included encoders:\n"
1201 "\n");
1202
1203 for (i = 0; i < sizeof(codecs) / sizeof(codecs[0]); i++)
1204 fprintf(stderr, " %-6s - %s\n",
1205 codecs[i].name,
1206 vpx_codec_iface_name(codecs[i].iface()));
1207
1208 exit(EXIT_FAILURE);
1209 }
1210
1211
1212 #define HIST_BAR_MAX 40
1213 struct hist_bucket {
1214 int low, high, count;
1215 };
1216
1217
1218 static int merge_hist_buckets(struct hist_bucket *bucket,
1219 int *buckets_,
1220 int max_buckets) {
1221 int small_bucket = 0, merge_bucket = INT_MAX, big_bucket = 0;
1222 int buckets = *buckets_;
1223 int i;
1224
1225 /* Find the extrema for this list of buckets */
1226 big_bucket = small_bucket = 0;
1227 for (i = 0; i < buckets; i++) {
1228 if (bucket[i].count < bucket[small_bucket].count)
1229 small_bucket = i;
1230 if (bucket[i].count > bucket[big_bucket].count)
1231 big_bucket = i;
1232 }
1233
1234 /* If we have too many buckets, merge the smallest with an adjacent
1235 * bucket.
1236 */
1237 while (buckets > max_buckets) {
1238 int last_bucket = buckets - 1;
1239
1240 /* merge the small bucket with an adjacent one. */
1241 if (small_bucket == 0)
1242 merge_bucket = 1;
1243 else if (small_bucket == last_bucket)
1244 merge_bucket = last_bucket - 1;
1245 else if (bucket[small_bucket - 1].count < bucket[small_bucket + 1].count)
1246 merge_bucket = small_bucket - 1;
1247 else
1248 merge_bucket = small_bucket + 1;
1249
1250 assert(abs(merge_bucket - small_bucket) <= 1);
1251 assert(small_bucket < buckets);
1252 assert(big_bucket < buckets);
1253 assert(merge_bucket < buckets);
1254
1255 if (merge_bucket < small_bucket) {
1256 bucket[merge_bucket].high = bucket[small_bucket].high;
1257 bucket[merge_bucket].count += bucket[small_bucket].count;
1258 } else {
1259 bucket[small_bucket].high = bucket[merge_bucket].high;
1260 bucket[small_bucket].count += bucket[merge_bucket].count;
1261 merge_bucket = small_bucket;
1262 }
1263
1264 assert(bucket[merge_bucket].low != bucket[merge_bucket].high);
1265
1266 buckets--;
1267
1268 /* Remove the merge_bucket from the list, and find the new small
1269 * and big buckets while we're at it
1270 */
1271 big_bucket = small_bucket = 0;
1272 for (i = 0; i < buckets; i++) {
1273 if (i > merge_bucket)
1274 bucket[i] = bucket[i + 1];
1275
1276 if (bucket[i].count < bucket[small_bucket].count)
1277 small_bucket = i;
1278 if (bucket[i].count > bucket[big_bucket].count)
1279 big_bucket = i;
1280 }
1281
1282 }
1283
1284 *buckets_ = buckets;
1285 return bucket[big_bucket].count;
1286 }
1287
1288
1289 static void show_histogram(const struct hist_bucket *bucket,
1290 int buckets,
1291 int total,
1292 int scale) {
1293 const char *pat1, *pat2;
1294 int i;
1295
1296 switch ((int)(log(bucket[buckets - 1].high) / log(10)) + 1) {
1297 case 1:
1298 case 2:
1299 pat1 = "%4d %2s: ";
1300 pat2 = "%4d-%2d: ";
1301 break;
1302 case 3:
1303 pat1 = "%5d %3s: ";
1304 pat2 = "%5d-%3d: ";
1305 break;
1306 case 4:
1307 pat1 = "%6d %4s: ";
1308 pat2 = "%6d-%4d: ";
1309 break;
1310 case 5:
1311 pat1 = "%7d %5s: ";
1312 pat2 = "%7d-%5d: ";
1313 break;
1314 case 6:
1315 pat1 = "%8d %6s: ";
1316 pat2 = "%8d-%6d: ";
1317 break;
1318 case 7:
1319 pat1 = "%9d %7s: ";
1320 pat2 = "%9d-%7d: ";
1321 break;
1322 default:
1323 pat1 = "%12d %10s: ";
1324 pat2 = "%12d-%10d: ";
1325 break;
1326 }
1327
1328 for (i = 0; i < buckets; i++) {
1329 int len;
1330 int j;
1331 float pct;
1332
1333 pct = (float)(100.0 * bucket[i].count / total);
1334 len = HIST_BAR_MAX * bucket[i].count / scale;
1335 if (len < 1)
1336 len = 1;
1337 assert(len <= HIST_BAR_MAX);
1338
1339 if (bucket[i].low == bucket[i].high)
1340 fprintf(stderr, pat1, bucket[i].low, "");
1341 else
1342 fprintf(stderr, pat2, bucket[i].low, bucket[i].high);
1343
1344 for (j = 0; j < HIST_BAR_MAX; j++)
1345 fprintf(stderr, j < len ? "=" : " ");
1346 fprintf(stderr, "\t%5d (%6.2f%%)\n", bucket[i].count, pct);
1347 }
1348 }
1349
1350
1351 static void show_q_histogram(const int counts[64], int max_buckets) {
1352 struct hist_bucket bucket[64];
1353 int buckets = 0;
1354 int total = 0;
1355 int scale;
1356 int i;
1357
1358
1359 for (i = 0; i < 64; i++) {
1360 if (counts[i]) {
1361 bucket[buckets].low = bucket[buckets].high = i;
1362 bucket[buckets].count = counts[i];
1363 buckets++;
1364 total += counts[i];
1365 }
1366 }
1367
1368 fprintf(stderr, "\nQuantizer Selection:\n");
1369 scale = merge_hist_buckets(bucket, &buckets, max_buckets);
1370 show_histogram(bucket, buckets, total, scale);
1371 }
1372
1373
1374 #define RATE_BINS (100)
1375 struct rate_hist {
1376 int64_t *pts;
1377 int *sz;
1378 int samples;
1379 int frames;
1380 struct hist_bucket bucket[RATE_BINS];
1381 int total;
1382 };
1383
1384
1385 static void init_rate_histogram(struct rate_hist *hist,
1386 const vpx_codec_enc_cfg_t *cfg,
1387 const vpx_rational_t *fps) {
1388 int i;
1389
1390 /* Determine the number of samples in the buffer. Use the file's framerate
1391 * to determine the number of frames in rc_buf_sz milliseconds, with an
1392 * adjustment (5/4) to account for alt-refs
1393 */
1394 hist->samples = cfg->rc_buf_sz * 5 / 4 * fps->num / fps->den / 1000;
1395
1396 /* prevent division by zero */
1397 if (hist->samples == 0)
1398 hist->samples = 1;
1399
1400 hist->pts = calloc(hist->samples, sizeof(*hist->pts));
1401 hist->sz = calloc(hist->samples, sizeof(*hist->sz));
1402 for (i = 0; i < RATE_BINS; i++) {
1403 hist->bucket[i].low = INT_MAX;
1404 hist->bucket[i].high = 0;
1405 hist->bucket[i].count = 0;
1406 }
1407 }
1408
1409
1410 static void destroy_rate_histogram(struct rate_hist *hist) {
1411 free(hist->pts);
1412 free(hist->sz);
1413 }
1414
1415
1416 static void update_rate_histogram(struct rate_hist *hist,
1417 const vpx_codec_enc_cfg_t *cfg,
1418 const vpx_codec_cx_pkt_t *pkt) {
1419 int i, idx;
1420 int64_t now, then, sum_sz = 0, avg_bitrate;
1421
1422 now = pkt->data.frame.pts * 1000
1423 * (uint64_t)cfg->g_timebase.num / (uint64_t)cfg->g_timebase.den;
1424
1425 idx = hist->frames++ % hist->samples;
1426 hist->pts[idx] = now;
1427 hist->sz[idx] = (int)pkt->data.frame.sz;
1428
1429 if (now < cfg->rc_buf_initial_sz)
1430 return;
1431
1432 then = now;
1433
1434 /* Sum the size over the past rc_buf_sz ms */
1435 for (i = hist->frames; i > 0 && hist->frames - i < hist->samples; i--) {
1436 int i_idx = (i - 1) % hist->samples;
1437
1438 then = hist->pts[i_idx];
1439 if (now - then > cfg->rc_buf_sz)
1440 break;
1441 sum_sz += hist->sz[i_idx];
1442 }
1443
1444 if (now == then)
1445 return;
1446
1447 avg_bitrate = sum_sz * 8 * 1000 / (now - then);
1448 idx = (int)(avg_bitrate * (RATE_BINS / 2) / (cfg->rc_target_bitrate * 1000));
1449 if (idx < 0)
1450 idx = 0;
1451 if (idx > RATE_BINS - 1)
1452 idx = RATE_BINS - 1;
1453 if (hist->bucket[idx].low > avg_bitrate)
1454 hist->bucket[idx].low = (int)avg_bitrate;
1455 if (hist->bucket[idx].high < avg_bitrate)
1456 hist->bucket[idx].high = (int)avg_bitrate;
1457 hist->bucket[idx].count++;
1458 hist->total++;
1459 }
1460
1461
1462 static void show_rate_histogram(struct rate_hist *hist,
1463 const vpx_codec_enc_cfg_t *cfg,
1464 int max_buckets) {
1465 int i, scale;
1466 int buckets = 0;
1467
1468 for (i = 0; i < RATE_BINS; i++) {
1469 if (hist->bucket[i].low == INT_MAX)
1470 continue;
1471 hist->bucket[buckets++] = hist->bucket[i];
1472 }
1473
1474 fprintf(stderr, "\nRate (over %dms window):\n", cfg->rc_buf_sz);
1475 scale = merge_hist_buckets(hist->bucket, &buckets, max_buckets);
1476 show_histogram(hist->bucket, buckets, hist->total, scale);
1477 }
1478
1479 #define mmin(a, b) ((a) < (b) ? (a) : (b))
1480 static void find_mismatch(vpx_image_t *img1, vpx_image_t *img2,
1481 int yloc[2], int uloc[2], int vloc[2]) {
1482 int match = 1;
1483 int i, j;
1484 yloc[0] = yloc[1] = -1;
1485 for (i = 0, match = 1; match && i < img1->d_h; i+=32) {
1486 for (j = 0; match && j < img1->d_w; j+=32) {
1487 int k, l;
1488 int si = mmin(i + 32, img1->d_h) - i;
1489 int sj = mmin(j + 32, img1->d_w) - j;
1490 for (k = 0; match && k < si; k++)
1491 for (l = 0; match && l < sj; l++) {
1492 if (*(img1->planes[VPX_PLANE_Y] +
1493 (i + k) * img1->stride[VPX_PLANE_Y] + j + l) !=
1494 *(img2->planes[VPX_PLANE_Y] +
1495 (i + k) * img2->stride[VPX_PLANE_Y] + j + l)) {
1496 yloc[0] = i + k;
1497 yloc[1] = j + l;
1498 match = 0;
1499 break;
1500 }
1501 }
1502 }
1503 }
1504 uloc[0] = uloc[1] = -1;
1505 for (i = 0, match = 1; match && i < (img1->d_h + 1) / 2; i+=16) {
1506 for (j = 0; j < match && (img1->d_w + 1) / 2; j+=16) {
1507 int k, l;
1508 int si = mmin(i + 16, (img1->d_h + 1) / 2) - i;
1509 int sj = mmin(j + 16, (img1->d_w + 1) / 2) - j;
1510 for (k = 0; match && k < si; k++)
1511 for (l = 0; match && l < sj; l++) {
1512 if (*(img1->planes[VPX_PLANE_U] +
1513 (i + k) * img1->stride[VPX_PLANE_U] + j + l) !=
1514 *(img2->planes[VPX_PLANE_U] +
1515 (i + k) * img2->stride[VPX_PLANE_U] + j + l)) {
1516 uloc[0] = i + k;
1517 uloc[1] = j + l;
1518 match = 0;
1519 break;
1520 }
1521 }
1522 }
1523 }
1524 vloc[0] = vloc[1] = -1;
1525 for (i = 0, match = 1; match && i < (img1->d_h + 1) / 2; i+=16) {
1526 for (j = 0; j < match && (img1->d_w + 1) / 2; j+=16) {
1527 int k, l;
1528 int si = mmin(i + 16, (img1->d_h + 1) / 2) - i;
1529 int sj = mmin(j + 16, (img1->d_w + 1) / 2) - j;
1530 for (k = 0; match && k < si; k++)
1531 for (l = 0; match && l < sj; l++) {
1532 if (*(img1->planes[VPX_PLANE_V] +
1533 (i + k) * img1->stride[VPX_PLANE_V] + j + l) !=
1534 *(img2->planes[VPX_PLANE_V] +
1535 (i + k) * img2->stride[VPX_PLANE_V] + j + l)) {
1536 vloc[0] = i + k;
1537 vloc[1] = j + l;
1538 match = 0;
1539 break;
1540 }
1541 }
1542 }
1543 }
1544 }
1545
1546 static int compare_img(vpx_image_t *img1, vpx_image_t *img2)
1547 {
1548 int match = 1;
1549 int i;
1550
1551 match &= (img1->fmt == img2->fmt);
1552 match &= (img1->w == img2->w);
1553 match &= (img1->h == img2->h);
1554
1555 for (i = 0; i < img1->d_h; i++)
1556 match &= (memcmp(img1->planes[VPX_PLANE_Y]+i*img1->stride[VPX_PLANE_Y],
1557 img2->planes[VPX_PLANE_Y]+i*img2->stride[VPX_PLANE_Y],
1558 img1->d_w) == 0);
1559
1560 for (i = 0; i < img1->d_h/2; i++)
1561 match &= (memcmp(img1->planes[VPX_PLANE_U]+i*img1->stride[VPX_PLANE_U],
1562 img2->planes[VPX_PLANE_U]+i*img2->stride[VPX_PLANE_U],
1563 (img1->d_w + 1) / 2) == 0);
1564
1565 for (i = 0; i < img1->d_h/2; i++)
1566 match &= (memcmp(img1->planes[VPX_PLANE_V]+i*img1->stride[VPX_PLANE_U],
1567 img2->planes[VPX_PLANE_V]+i*img2->stride[VPX_PLANE_U],
1568 (img1->d_w + 1) / 2) == 0);
1569
1570 return match;
1571 }
1572
1573
1574 #define NELEMENTS(x) (sizeof(x)/sizeof(x[0]))
1575 #define MAX(x,y) ((x)>(y)?(x):(y))
1576 #if CONFIG_VP8_ENCODER && !CONFIG_VP9_ENCODER
1577 #define ARG_CTRL_CNT_MAX NELEMENTS(vp8_arg_ctrl_map)
1578 #elif !CONFIG_VP8_ENCODER && CONFIG_VP9_ENCODER
1579 #define ARG_CTRL_CNT_MAX NELEMENTS(vp9_arg_ctrl_map)
1580 #else
1581 #define ARG_CTRL_CNT_MAX MAX(NELEMENTS(vp8_arg_ctrl_map), \
1582 NELEMENTS(vp9_arg_ctrl_map))
1583 #endif
1584
1585 /* Configuration elements common to all streams */
1586 struct global_config {
1587 const struct codec_item *codec;
1588 int passes;
1589 int pass;
1590 int usage;
1591 int deadline;
1592 int use_i420;
1593 int quiet;
1594 int verbose;
1595 int limit;
1596 int skip_frames;
1597 int show_psnr;
1598 enum TestDecodeFatality test_decode;
1599 int have_framerate;
1600 struct vpx_rational framerate;
1601 int out_part;
1602 int debug;
1603 int show_q_hist_buckets;
1604 int show_rate_hist_buckets;
1605 };
1606
1607
1608 /* Per-stream configuration */
1609 struct stream_config {
1610 struct vpx_codec_enc_cfg cfg;
1611 const char *out_fn;
1612 const char *stats_fn;
1613 stereo_format_t stereo_fmt;
1614 int arg_ctrls[ARG_CTRL_CNT_MAX][2];
1615 int arg_ctrl_cnt;
1616 int write_webm;
1617 int have_kf_max_dist;
1618 };
1619
1620
1621 struct stream_state {
1622 int index;
1623 struct stream_state *next;
1624 struct stream_config config;
1625 FILE *file;
1626 struct rate_hist rate_hist;
1627 EbmlGlobal ebml;
1628 uint32_t hash;
1629 uint64_t psnr_sse_total;
1630 uint64_t psnr_samples_total;
1631 double psnr_totals[4];
1632 int psnr_count;
1633 int counts[64];
1634 vpx_codec_ctx_t encoder;
1635 unsigned int frames_out;
1636 uint64_t cx_time;
1637 size_t nbytes;
1638 stats_io_t stats;
1639 struct vpx_image *img;
1640 vpx_codec_ctx_t decoder;
1641 vpx_ref_frame_t ref_enc;
1642 vpx_ref_frame_t ref_dec;
1643 int mismatch_seen;
1644 };
1645
1646
1647 void validate_positive_rational(const char *msg,
1648 struct vpx_rational *rat) {
1649 if (rat->den < 0) {
1650 rat->num *= -1;
1651 rat->den *= -1;
1652 }
1653
1654 if (rat->num < 0)
1655 die("Error: %s must be positive\n", msg);
1656
1657 if (!rat->den)
1658 die("Error: %s has zero denominator\n", msg);
1659 }
1660
1661
1662 static void parse_global_config(struct global_config *global, char **argv) {
1663 char **argi, **argj;
1664 struct arg arg;
1665
1666 /* Initialize default parameters */
1667 memset(global, 0, sizeof(*global));
1668 global->codec = codecs;
1669 global->passes = 1;
1670 global->use_i420 = 1;
1671
1672 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
1673 arg.argv_step = 1;
1674
1675 if (arg_match(&arg, &codecarg, argi)) {
1676 int j, k = -1;
1677
1678 for (j = 0; j < sizeof(codecs) / sizeof(codecs[0]); j++)
1679 if (!strcmp(codecs[j].name, arg.val))
1680 k = j;
1681
1682 if (k >= 0)
1683 global->codec = codecs + k;
1684 else
1685 die("Error: Unrecognized argument (%s) to --codec\n",
1686 arg.val);
1687
1688 } else if (arg_match(&arg, &passes, argi)) {
1689 global->passes = arg_parse_uint(&arg);
1690
1691 if (global->passes < 1 || global->passes > 2)
1692 die("Error: Invalid number of passes (%d)\n", global->passes);
1693 } else if (arg_match(&arg, &pass_arg, argi)) {
1694 global->pass = arg_parse_uint(&arg);
1695
1696 if (global->pass < 1 || global->pass > 2)
1697 die("Error: Invalid pass selected (%d)\n",
1698 global->pass);
1699 } else if (arg_match(&arg, &usage, argi))
1700 global->usage = arg_parse_uint(&arg);
1701 else if (arg_match(&arg, &deadline, argi))
1702 global->deadline = arg_parse_uint(&arg);
1703 else if (arg_match(&arg, &best_dl, argi))
1704 global->deadline = VPX_DL_BEST_QUALITY;
1705 else if (arg_match(&arg, &good_dl, argi))
1706 global->deadline = VPX_DL_GOOD_QUALITY;
1707 else if (arg_match(&arg, &rt_dl, argi))
1708 global->deadline = VPX_DL_REALTIME;
1709 else if (arg_match(&arg, &use_yv12, argi))
1710 global->use_i420 = 0;
1711 else if (arg_match(&arg, &use_i420, argi))
1712 global->use_i420 = 1;
1713 else if (arg_match(&arg, &quietarg, argi))
1714 global->quiet = 1;
1715 else if (arg_match(&arg, &verbosearg, argi))
1716 global->verbose = 1;
1717 else if (arg_match(&arg, &limit, argi))
1718 global->limit = arg_parse_uint(&arg);
1719 else if (arg_match(&arg, &skip, argi))
1720 global->skip_frames = arg_parse_uint(&arg);
1721 else if (arg_match(&arg, &psnrarg, argi))
1722 global->show_psnr = 1;
1723 else if (arg_match(&arg, &recontest, argi))
1724 global->test_decode = arg_parse_enum_or_int(&arg);
1725 else if (arg_match(&arg, &framerate, argi)) {
1726 global->framerate = arg_parse_rational(&arg);
1727 validate_positive_rational(arg.name, &global->framerate);
1728 global->have_framerate = 1;
1729 } else if (arg_match(&arg, &out_part, argi))
1730 global->out_part = 1;
1731 else if (arg_match(&arg, &debugmode, argi))
1732 global->debug = 1;
1733 else if (arg_match(&arg, &q_hist_n, argi))
1734 global->show_q_hist_buckets = arg_parse_uint(&arg);
1735 else if (arg_match(&arg, &rate_hist_n, argi))
1736 global->show_rate_hist_buckets = arg_parse_uint(&arg);
1737 else
1738 argj++;
1739 }
1740
1741 /* Validate global config */
1742
1743 if (global->pass) {
1744 /* DWIM: Assume the user meant passes=2 if pass=2 is specified */
1745 if (global->pass > global->passes) {
1746 warn("Assuming --pass=%d implies --passes=%d\n",
1747 global->pass, global->pass);
1748 global->passes = global->pass;
1749 }
1750 }
1751 }
1752
1753
1754 void open_input_file(struct input_state *input) {
1755 unsigned int fourcc;
1756
1757 /* Parse certain options from the input file, if possible */
1758 input->file = strcmp(input->fn, "-") ? fopen(input->fn, "rb")
1759 : set_binary_mode(stdin);
1760
1761 if (!input->file)
1762 fatal("Failed to open input file");
1763
1764 if (!fseeko(input->file, 0, SEEK_END)) {
1765 /* Input file is seekable. Figure out how long it is, so we can get
1766 * progress info.
1767 */
1768 input->length = ftello(input->file);
1769 rewind(input->file);
1770 }
1771
1772 /* For RAW input sources, these bytes will applied on the first frame
1773 * in read_frame().
1774 */
1775 input->detect.buf_read = fread(input->detect.buf, 1, 4, input->file);
1776 input->detect.position = 0;
1777
1778 if (input->detect.buf_read == 4
1779 && file_is_y4m(input->file, &input->y4m, input->detect.buf)) {
1780 if (y4m_input_open(&input->y4m, input->file, input->detect.buf, 4) >= 0) {
1781 input->file_type = FILE_TYPE_Y4M;
1782 input->w = input->y4m.pic_w;
1783 input->h = input->y4m.pic_h;
1784 input->framerate.num = input->y4m.fps_n;
1785 input->framerate.den = input->y4m.fps_d;
1786 input->use_i420 = 0;
1787 } else
1788 fatal("Unsupported Y4M stream.");
1789 } else if (input->detect.buf_read == 4 && file_is_ivf(input, &fourcc)) {
1790 input->file_type = FILE_TYPE_IVF;
1791 switch (fourcc) {
1792 case 0x32315659:
1793 input->use_i420 = 0;
1794 break;
1795 case 0x30323449:
1796 input->use_i420 = 1;
1797 break;
1798 default:
1799 fatal("Unsupported fourcc (%08x) in IVF", fourcc);
1800 }
1801 } else {
1802 input->file_type = FILE_TYPE_RAW;
1803 }
1804 }
1805
1806
1807 static void close_input_file(struct input_state *input) {
1808 fclose(input->file);
1809 if (input->file_type == FILE_TYPE_Y4M)
1810 y4m_input_close(&input->y4m);
1811 }
1812
1813 static struct stream_state *new_stream(struct global_config *global,
1814 struct stream_state *prev) {
1815 struct stream_state *stream;
1816
1817 stream = calloc(1, sizeof(*stream));
1818 if (!stream)
1819 fatal("Failed to allocate new stream.");
1820 if (prev) {
1821 memcpy(stream, prev, sizeof(*stream));
1822 stream->index++;
1823 prev->next = stream;
1824 } else {
1825 vpx_codec_err_t res;
1826
1827 /* Populate encoder configuration */
1828 res = vpx_codec_enc_config_default(global->codec->iface(),
1829 &stream->config.cfg,
1830 global->usage);
1831 if (res)
1832 fatal("Failed to get config: %s\n", vpx_codec_err_to_string(res));
1833
1834 /* Change the default timebase to a high enough value so that the
1835 * encoder will always create strictly increasing timestamps.
1836 */
1837 stream->config.cfg.g_timebase.den = 1000;
1838
1839 /* Never use the library's default resolution, require it be parsed
1840 * from the file or set on the command line.
1841 */
1842 stream->config.cfg.g_w = 0;
1843 stream->config.cfg.g_h = 0;
1844
1845 /* Initialize remaining stream parameters */
1846 stream->config.stereo_fmt = STEREO_FORMAT_MONO;
1847 stream->config.write_webm = 1;
1848 stream->ebml.last_pts_ms = -1;
1849
1850 /* Allows removal of the application version from the EBML tags */
1851 stream->ebml.debug = global->debug;
1852 }
1853
1854 /* Output files must be specified for each stream */
1855 stream->config.out_fn = NULL;
1856
1857 stream->next = NULL;
1858 return stream;
1859 }
1860
1861
1862 static int parse_stream_params(struct global_config *global,
1863 struct stream_state *stream,
1864 char **argv) {
1865 char **argi, **argj;
1866 struct arg arg;
1867 static const arg_def_t **ctrl_args = no_args;
1868 static const int *ctrl_args_map = NULL;
1869 struct stream_config *config = &stream->config;
1870 int eos_mark_found = 0;
1871
1872 /* Handle codec specific options */
1873 if (0) {
1874 #if CONFIG_VP8_ENCODER
1875 } else if (global->codec->iface == vpx_codec_vp8_cx) {
1876 ctrl_args = vp8_args;
1877 ctrl_args_map = vp8_arg_ctrl_map;
1878 #endif
1879 #if CONFIG_VP9_ENCODER
1880 } else if (global->codec->iface == vpx_codec_vp9_cx) {
1881 ctrl_args = vp9_args;
1882 ctrl_args_map = vp9_arg_ctrl_map;
1883 #endif
1884 }
1885
1886 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
1887 arg.argv_step = 1;
1888
1889 /* Once we've found an end-of-stream marker (--) we want to continue
1890 * shifting arguments but not consuming them.
1891 */
1892 if (eos_mark_found) {
1893 argj++;
1894 continue;
1895 } else if (!strcmp(*argj, "--")) {
1896 eos_mark_found = 1;
1897 continue;
1898 }
1899
1900 if (0);
1901 else if (arg_match(&arg, &outputfile, argi))
1902 config->out_fn = arg.val;
1903 else if (arg_match(&arg, &fpf_name, argi))
1904 config->stats_fn = arg.val;
1905 else if (arg_match(&arg, &use_ivf, argi))
1906 config->write_webm = 0;
1907 else if (arg_match(&arg, &threads, argi))
1908 config->cfg.g_threads = arg_parse_uint(&arg);
1909 else if (arg_match(&arg, &profile, argi))
1910 config->cfg.g_profile = arg_parse_uint(&arg);
1911 else if (arg_match(&arg, &width, argi))
1912 config->cfg.g_w = arg_parse_uint(&arg);
1913 else if (arg_match(&arg, &height, argi))
1914 config->cfg.g_h = arg_parse_uint(&arg);
1915 else if (arg_match(&arg, &stereo_mode, argi))
1916 config->stereo_fmt = arg_parse_enum_or_int(&arg);
1917 else if (arg_match(&arg, &timebase, argi)) {
1918 config->cfg.g_timebase = arg_parse_rational(&arg);
1919 validate_positive_rational(arg.name, &config->cfg.g_timebase);
1920 } else if (arg_match(&arg, &error_resilient, argi))
1921 config->cfg.g_error_resilient = arg_parse_uint(&arg);
1922 else if (arg_match(&arg, &lag_in_frames, argi))
1923 config->cfg.g_lag_in_frames = arg_parse_uint(&arg);
1924 else if (arg_match(&arg, &dropframe_thresh, argi))
1925 config->cfg.rc_dropframe_thresh = arg_parse_uint(&arg);
1926 else if (arg_match(&arg, &resize_allowed, argi))
1927 config->cfg.rc_resize_allowed = arg_parse_uint(&arg);
1928 else if (arg_match(&arg, &resize_up_thresh, argi))
1929 config->cfg.rc_resize_up_thresh = arg_parse_uint(&arg);
1930 else if (arg_match(&arg, &resize_down_thresh, argi))
1931 config->cfg.rc_resize_down_thresh = arg_parse_uint(&arg);
1932 else if (arg_match(&arg, &end_usage, argi))
1933 config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
1934 else if (arg_match(&arg, &target_bitrate, argi))
1935 config->cfg.rc_target_bitrate = arg_parse_uint(&arg);
1936 else if (arg_match(&arg, &min_quantizer, argi))
1937 config->cfg.rc_min_quantizer = arg_parse_uint(&arg);
1938 else if (arg_match(&arg, &max_quantizer, argi))
1939 config->cfg.rc_max_quantizer = arg_parse_uint(&arg);
1940 else if (arg_match(&arg, &undershoot_pct, argi))
1941 config->cfg.rc_undershoot_pct = arg_parse_uint(&arg);
1942 else if (arg_match(&arg, &overshoot_pct, argi))
1943 config->cfg.rc_overshoot_pct = arg_parse_uint(&arg);
1944 else if (arg_match(&arg, &buf_sz, argi))
1945 config->cfg.rc_buf_sz = arg_parse_uint(&arg);
1946 else if (arg_match(&arg, &buf_initial_sz, argi))
1947 config->cfg.rc_buf_initial_sz = arg_parse_uint(&arg);
1948 else if (arg_match(&arg, &buf_optimal_sz, argi))
1949 config->cfg.rc_buf_optimal_sz = arg_parse_uint(&arg);
1950 else if (arg_match(&arg, &bias_pct, argi)) {
1951 config->cfg.rc_2pass_vbr_bias_pct = arg_parse_uint(&arg);
1952
1953 if (global->passes < 2)
1954 warn("option %s ignored in one-pass mode.\n", arg.name);
1955 } else if (arg_match(&arg, &minsection_pct, argi)) {
1956 config->cfg.rc_2pass_vbr_minsection_pct = arg_parse_uint(&arg);
1957
1958 if (global->passes < 2)
1959 warn("option %s ignored in one-pass mode.\n", arg.name);
1960 } else if (arg_match(&arg, &maxsection_pct, argi)) {
1961 config->cfg.rc_2pass_vbr_maxsection_pct = arg_parse_uint(&arg);
1962
1963 if (global->passes < 2)
1964 warn("option %s ignored in one-pass mode.\n", arg.name);
1965 } else if (arg_match(&arg, &kf_min_dist, argi))
1966 config->cfg.kf_min_dist = arg_parse_uint(&arg);
1967 else if (arg_match(&arg, &kf_max_dist, argi)) {
1968 config->cfg.kf_max_dist = arg_parse_uint(&arg);
1969 config->have_kf_max_dist = 1;
1970 } else if (arg_match(&arg, &kf_disabled, argi))
1971 config->cfg.kf_mode = VPX_KF_DISABLED;
1972 else {
1973 int i, match = 0;
1974
1975 for (i = 0; ctrl_args[i]; i++) {
1976 if (arg_match(&arg, ctrl_args[i], argi)) {
1977 int j;
1978 match = 1;
1979
1980 /* Point either to the next free element or the first
1981 * instance of this control.
1982 */
1983 for (j = 0; j < config->arg_ctrl_cnt; j++)
1984 if (config->arg_ctrls[j][0] == ctrl_args_map[i])
1985 break;
1986
1987 /* Update/insert */
1988 assert(j < ARG_CTRL_CNT_MAX);
1989 if (j < ARG_CTRL_CNT_MAX) {
1990 config->arg_ctrls[j][0] = ctrl_args_map[i];
1991 config->arg_ctrls[j][1] = arg_parse_enum_or_int(&arg);
1992 if (j == config->arg_ctrl_cnt)
1993 config->arg_ctrl_cnt++;
1994 }
1995
1996 }
1997 }
1998
1999 if (!match)
2000 argj++;
2001 }
2002 }
2003
2004 return eos_mark_found;
2005 }
2006
2007
2008 #define FOREACH_STREAM(func)\
2009 do\
2010 {\
2011 struct stream_state *stream;\
2012 \
2013 for(stream = streams; stream; stream = stream->next)\
2014 func;\
2015 }while(0)
2016
2017
2018 static void validate_stream_config(struct stream_state *stream) {
2019 struct stream_state *streami;
2020
2021 if (!stream->config.cfg.g_w || !stream->config.cfg.g_h)
2022 fatal("Stream %d: Specify stream dimensions with --width (-w) "
2023 " and --height (-h)", stream->index);
2024
2025 for (streami = stream; streami; streami = streami->next) {
2026 /* All streams require output files */
2027 if (!streami->config.out_fn)
2028 fatal("Stream %d: Output file is required (specify with -o)",
2029 streami->index);
2030
2031 /* Check for two streams outputting to the same file */
2032 if (streami != stream) {
2033 const char *a = stream->config.out_fn;
2034 const char *b = streami->config.out_fn;
2035 if (!strcmp(a, b) && strcmp(a, "/dev/null") && strcmp(a, ":nul"))
2036 fatal("Stream %d: duplicate output file (from stream %d)",
2037 streami->index, stream->index);
2038 }
2039
2040 /* Check for two streams sharing a stats file. */
2041 if (streami != stream) {
2042 const char *a = stream->config.stats_fn;
2043 const char *b = streami->config.stats_fn;
2044 if (a && b && !strcmp(a, b))
2045 fatal("Stream %d: duplicate stats file (from stream %d)",
2046 streami->index, stream->index);
2047 }
2048 }
2049 }
2050
2051
2052 static void set_stream_dimensions(struct stream_state *stream,
2053 unsigned int w,
2054 unsigned int h) {
2055 if (!stream->config.cfg.g_w) {
2056 if (!stream->config.cfg.g_h)
2057 stream->config.cfg.g_w = w;
2058 else
2059 stream->config.cfg.g_w = w * stream->config.cfg.g_h / h;
2060 }
2061 if (!stream->config.cfg.g_h) {
2062 stream->config.cfg.g_h = h * stream->config.cfg.g_w / w;
2063 }
2064 }
2065
2066
2067 static void set_default_kf_interval(struct stream_state *stream,
2068 struct global_config *global) {
2069 /* Use a max keyframe interval of 5 seconds, if none was
2070 * specified on the command line.
2071 */
2072 if (!stream->config.have_kf_max_dist) {
2073 double framerate = (double)global->framerate.num / global->framerate.den;
2074 if (framerate > 0.0)
2075 stream->config.cfg.kf_max_dist = (unsigned int)(5.0 * framerate);
2076 }
2077 }
2078
2079
2080 static void show_stream_config(struct stream_state *stream,
2081 struct global_config *global,
2082 struct input_state *input) {
2083
2084 #define SHOW(field) \
2085 fprintf(stderr, " %-28s = %d\n", #field, stream->config.cfg.field)
2086
2087 if (stream->index == 0) {
2088 fprintf(stderr, "Codec: %s\n",
2089 vpx_codec_iface_name(global->codec->iface()));
2090 fprintf(stderr, "Source file: %s Format: %s\n", input->fn,
2091 input->use_i420 ? "I420" : "YV12");
2092 }
2093 if (stream->next || stream->index)
2094 fprintf(stderr, "\nStream Index: %d\n", stream->index);
2095 fprintf(stderr, "Destination file: %s\n", stream->config.out_fn);
2096 fprintf(stderr, "Encoder parameters:\n");
2097
2098 SHOW(g_usage);
2099 SHOW(g_threads);
2100 SHOW(g_profile);
2101 SHOW(g_w);
2102 SHOW(g_h);
2103 SHOW(g_timebase.num);
2104 SHOW(g_timebase.den);
2105 SHOW(g_error_resilient);
2106 SHOW(g_pass);
2107 SHOW(g_lag_in_frames);
2108 SHOW(rc_dropframe_thresh);
2109 SHOW(rc_resize_allowed);
2110 SHOW(rc_resize_up_thresh);
2111 SHOW(rc_resize_down_thresh);
2112 SHOW(rc_end_usage);
2113 SHOW(rc_target_bitrate);
2114 SHOW(rc_min_quantizer);
2115 SHOW(rc_max_quantizer);
2116 SHOW(rc_undershoot_pct);
2117 SHOW(rc_overshoot_pct);
2118 SHOW(rc_buf_sz);
2119 SHOW(rc_buf_initial_sz);
2120 SHOW(rc_buf_optimal_sz);
2121 SHOW(rc_2pass_vbr_bias_pct);
2122 SHOW(rc_2pass_vbr_minsection_pct);
2123 SHOW(rc_2pass_vbr_maxsection_pct);
2124 SHOW(kf_mode);
2125 SHOW(kf_min_dist);
2126 SHOW(kf_max_dist);
2127 }
2128
2129
2130 static void open_output_file(struct stream_state *stream,
2131 struct global_config *global) {
2132 const char *fn = stream->config.out_fn;
2133
2134 stream->file = strcmp(fn, "-") ? fopen(fn, "wb") : set_binary_mode(stdout);
2135
2136 if (!stream->file)
2137 fatal("Failed to open output file");
2138
2139 if (stream->config.write_webm && fseek(stream->file, 0, SEEK_CUR))
2140 fatal("WebM output to pipes not supported.");
2141
2142 if (stream->config.write_webm) {
2143 stream->ebml.stream = stream->file;
2144 write_webm_file_header(&stream->ebml, &stream->config.cfg,
2145 &global->framerate,
2146 stream->config.stereo_fmt,
2147 global->codec->fourcc);
2148 } else
2149 write_ivf_file_header(stream->file, &stream->config.cfg,
2150 global->codec->fourcc, 0);
2151 }
2152
2153
2154 static void close_output_file(struct stream_state *stream,
2155 unsigned int fourcc) {
2156 if (stream->config.write_webm) {
2157 write_webm_file_footer(&stream->ebml, stream->hash);
2158 free(stream->ebml.cue_list);
2159 stream->ebml.cue_list = NULL;
2160 } else {
2161 if (!fseek(stream->file, 0, SEEK_SET))
2162 write_ivf_file_header(stream->file, &stream->config.cfg,
2163 fourcc,
2164 stream->frames_out);
2165 }
2166
2167 fclose(stream->file);
2168 }
2169
2170
2171 static void setup_pass(struct stream_state *stream,
2172 struct global_config *global,
2173 int pass) {
2174 if (stream->config.stats_fn) {
2175 if (!stats_open_file(&stream->stats, stream->config.stats_fn,
2176 pass))
2177 fatal("Failed to open statistics store");
2178 } else {
2179 if (!stats_open_mem(&stream->stats, pass))
2180 fatal("Failed to open statistics store");
2181 }
2182
2183 stream->config.cfg.g_pass = global->passes == 2
2184 ? pass ? VPX_RC_LAST_PASS : VPX_RC_FIRST_PASS
2185 : VPX_RC_ONE_PASS;
2186 if (pass)
2187 stream->config.cfg.rc_twopass_stats_in = stats_get(&stream->stats);
2188
2189 stream->cx_time = 0;
2190 stream->nbytes = 0;
2191 stream->frames_out = 0;
2192 }
2193
2194
2195 static void initialize_encoder(struct stream_state *stream,
2196 struct global_config *global) {
2197 int i;
2198 int flags = 0;
2199
2200 flags |= global->show_psnr ? VPX_CODEC_USE_PSNR : 0;
2201 flags |= global->out_part ? VPX_CODEC_USE_OUTPUT_PARTITION : 0;
2202
2203 /* Construct Encoder Context */
2204 vpx_codec_enc_init(&stream->encoder, global->codec->iface(),
2205 &stream->config.cfg, flags);
2206 ctx_exit_on_error(&stream->encoder, "Failed to initialize encoder");
2207
2208 /* Note that we bypass the vpx_codec_control wrapper macro because
2209 * we're being clever to store the control IDs in an array. Real
2210 * applications will want to make use of the enumerations directly
2211 */
2212 for (i = 0; i < stream->config.arg_ctrl_cnt; i++) {
2213 int ctrl = stream->config.arg_ctrls[i][0];
2214 int value = stream->config.arg_ctrls[i][1];
2215 if (vpx_codec_control_(&stream->encoder, ctrl, value))
2216 fprintf(stderr, "Error: Tried to set control %d = %d\n",
2217 ctrl, value);
2218
2219 ctx_exit_on_error(&stream->encoder, "Failed to control codec");
2220 }
2221
2222 #if CONFIG_DECODERS
2223 if (global->test_decode != TEST_DECODE_OFF) {
2224 int width, height;
2225
2226 vpx_codec_dec_init(&stream->decoder, global->codec->dx_iface(), NULL, 0);
2227
2228 width = (stream->config.cfg.g_w + 15) & ~15;
2229 height = (stream->config.cfg.g_h + 15) & ~15;
2230 vpx_img_alloc(&stream->ref_enc.img, VPX_IMG_FMT_I420, width, height, 1);
2231 vpx_img_alloc(&stream->ref_dec.img, VPX_IMG_FMT_I420, width, height, 1);
2232 stream->ref_enc.frame_type = VP8_LAST_FRAME;
2233 stream->ref_dec.frame_type = VP8_LAST_FRAME;
2234 }
2235 #endif
2236 }
2237
2238
2239 static void encode_frame(struct stream_state *stream,
2240 struct global_config *global,
2241 struct vpx_image *img,
2242 unsigned int frames_in) {
2243 vpx_codec_pts_t frame_start, next_frame_start;
2244 struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
2245 struct vpx_usec_timer timer;
2246
2247 frame_start = (cfg->g_timebase.den * (int64_t)(frames_in - 1)
2248 * global->framerate.den)
2249 / cfg->g_timebase.num / global->framerate.num;
2250 next_frame_start = (cfg->g_timebase.den * (int64_t)(frames_in)
2251 * global->framerate.den)
2252 / cfg->g_timebase.num / global->framerate.num;
2253
2254 /* Scale if necessary */
2255 if (img && (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
2256 if (!stream->img)
2257 stream->img = vpx_img_alloc(NULL, VPX_IMG_FMT_I420,
2258 cfg->g_w, cfg->g_h, 16);
2259 I420Scale(img->planes[VPX_PLANE_Y], img->stride[VPX_PLANE_Y],
2260 img->planes[VPX_PLANE_U], img->stride[VPX_PLANE_U],
2261 img->planes[VPX_PLANE_V], img->stride[VPX_PLANE_V],
2262 img->d_w, img->d_h,
2263 stream->img->planes[VPX_PLANE_Y],
2264 stream->img->stride[VPX_PLANE_Y],
2265 stream->img->planes[VPX_PLANE_U],
2266 stream->img->stride[VPX_PLANE_U],
2267 stream->img->planes[VPX_PLANE_V],
2268 stream->img->stride[VPX_PLANE_V],
2269 stream->img->d_w, stream->img->d_h,
2270 kFilterBox);
2271
2272 img = stream->img;
2273 }
2274
2275 vpx_usec_timer_start(&timer);
2276 vpx_codec_encode(&stream->encoder, img, frame_start,
2277 (unsigned long)(next_frame_start - frame_start),
2278 0, global->deadline);
2279 vpx_usec_timer_mark(&timer);
2280 stream->cx_time += vpx_usec_timer_elapsed(&timer);
2281 ctx_exit_on_error(&stream->encoder, "Stream %d: Failed to encode frame",
2282 stream->index);
2283 }
2284
2285
2286 static void update_quantizer_histogram(struct stream_state *stream) {
2287 if (stream->config.cfg.g_pass != VPX_RC_FIRST_PASS) {
2288 int q;
2289
2290 vpx_codec_control(&stream->encoder, VP8E_GET_LAST_QUANTIZER_64, &q);
2291 ctx_exit_on_error(&stream->encoder, "Failed to read quantizer");
2292 stream->counts[q]++;
2293 }
2294 }
2295
2296
2297 static void get_cx_data(struct stream_state *stream,
2298 struct global_config *global,
2299 int *got_data) {
2300 const vpx_codec_cx_pkt_t *pkt;
2301 const struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
2302 vpx_codec_iter_t iter = NULL;
2303
2304 *got_data = 0;
2305 while ((pkt = vpx_codec_get_cx_data(&stream->encoder, &iter))) {
2306 static size_t fsize = 0;
2307 static off_t ivf_header_pos = 0;
2308
2309 switch (pkt->kind) {
2310 case VPX_CODEC_CX_FRAME_PKT:
2311 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
2312 stream->frames_out++;
2313 }
2314
2315 update_rate_histogram(&stream->rate_hist, cfg, pkt);
2316 if (stream->config.write_webm) {
2317 /* Update the hash */
2318 if (!stream->ebml.debug)
2319 stream->hash = murmur(pkt->data.frame.buf,
2320 (int)pkt->data.frame.sz,
2321 stream->hash);
2322
2323 write_webm_block(&stream->ebml, cfg, pkt);
2324 } else {
2325 if (pkt->data.frame.partition_id <= 0) {
2326 ivf_header_pos = ftello(stream->file);
2327 fsize = pkt->data.frame.sz;
2328
2329 write_ivf_frame_header(stream->file, pkt);
2330 } else {
2331 fsize += pkt->data.frame.sz;
2332
2333 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
2334 off_t currpos = ftello(stream->file);
2335 fseeko(stream->file, ivf_header_pos, SEEK_SET);
2336 write_ivf_frame_size(stream->file, fsize);
2337 fseeko(stream->file, currpos, SEEK_SET);
2338 }
2339 }
2340
2341 (void) fwrite(pkt->data.frame.buf, 1, pkt->data.frame.sz,
2342 stream->file);
2343 }
2344 stream->nbytes += pkt->data.raw.sz;
2345
2346 *got_data = 1;
2347 #if CONFIG_DECODERS
2348 if (global->test_decode != TEST_DECODE_OFF && !stream->mismatch_seen) {
2349 vpx_codec_decode(&stream->decoder, pkt->data.frame.buf,
2350 pkt->data.frame.sz, NULL, 0);
2351 if (stream->decoder.err) {
2352 warn_or_exit_on_error(&stream->decoder,
2353 global->test_decode == TEST_DECODE_FATAL,
2354 "Failed to decode frame %d in stream %d",
2355 stream->frames_out + 1, stream->index);
2356 stream->mismatch_seen = stream->frames_out + 1;
2357 }
2358 }
2359 #endif
2360 break;
2361 case VPX_CODEC_STATS_PKT:
2362 stream->frames_out++;
2363 stats_write(&stream->stats,
2364 pkt->data.twopass_stats.buf,
2365 pkt->data.twopass_stats.sz);
2366 stream->nbytes += pkt->data.raw.sz;
2367 break;
2368 case VPX_CODEC_PSNR_PKT:
2369
2370 if (global->show_psnr) {
2371 int i;
2372
2373 stream->psnr_sse_total += pkt->data.psnr.sse[0];
2374 stream->psnr_samples_total += pkt->data.psnr.samples[0];
2375 for (i = 0; i < 4; i++) {
2376 stream->psnr_totals[i] += pkt->data.psnr.psnr[i];
2377 }
2378 stream->psnr_count++;
2379 }
2380
2381 break;
2382 default:
2383 break;
2384 }
2385 }
2386 }
2387
2388
2389 static void show_psnr(struct stream_state *stream) {
2390 int i;
2391 double ovpsnr;
2392
2393 if (!stream->psnr_count)
2394 return;
2395
2396 fprintf(stderr, "Stream %d PSNR (Overall/Avg/Y/U/V)", stream->index);
2397 ovpsnr = vp8_mse2psnr((double)stream->psnr_samples_total, 255.0,
2398 (double)stream->psnr_sse_total);
2399 fprintf(stderr, " %.3f", ovpsnr);
2400
2401 for (i = 0; i < 4; i++) {
2402 fprintf(stderr, " %.3f", stream->psnr_totals[i] / stream->psnr_count);
2403 }
2404 fprintf(stderr, "\n");
2405 }
2406
2407
2408 static float usec_to_fps(uint64_t usec, unsigned int frames) {
2409 return (float)(usec > 0 ? frames * 1000000.0 / (float)usec : 0);
2410 }
2411
2412
2413 static void test_decode(struct stream_state *stream,
2414 enum TestDecodeFatality fatal) {
2415 if (stream->mismatch_seen)
2416 return;
2417
2418 vpx_codec_control(&stream->encoder, VP8_COPY_REFERENCE, &stream->ref_enc);
2419 ctx_exit_on_error(&stream->encoder, "Failed to get encoder reference frame");
2420 vpx_codec_control(&stream->decoder, VP8_COPY_REFERENCE, &stream->ref_dec);
2421 ctx_exit_on_error(&stream->decoder, "Failed to get decoder reference frame");
2422
2423 if (!compare_img(&stream->ref_enc.img, &stream->ref_dec.img)) {
2424 int y[2], u[2], v[2];
2425 find_mismatch(&stream->ref_enc.img, &stream->ref_dec.img,
2426 y, u, v);
2427 warn_or_exit_on_error(&stream->decoder, fatal == TEST_DECODE_FATAL,
2428 "Stream %d: Encode/decode mismatch on frame %d"
2429 " at Y[%d, %d], U[%d, %d], V[%d, %d]",
2430 stream->index, stream->frames_out,
2431 y[0], y[1], u[0], u[1], v[0], v[1]);
2432 stream->mismatch_seen = stream->frames_out;
2433 }
2434 }
2435
2436
2437 static void print_time(const char *label, int64_t etl) {
2438 int hours, mins, secs;
2439
2440 if (etl >= 0) {
2441 hours = etl / 3600;
2442 etl -= hours * 3600;
2443 mins = etl / 60;
2444 etl -= mins * 60;
2445 secs = etl;
2446
2447 fprintf(stderr, "[%3s %2d:%02d:%02d] ",
2448 label, hours, mins, secs);
2449 } else {
2450 fprintf(stderr, "[%3s unknown] ", label);
2451 }
2452 }
2453
2454 int main(int argc, const char **argv_) {
2455 int pass;
2456 vpx_image_t raw;
2457 int frame_avail, got_data;
2458
2459 struct input_state input = {0};
2460 struct global_config global;
2461 struct stream_state *streams = NULL;
2462 char **argv, **argi;
2463 uint64_t cx_time = 0;
2464 int stream_cnt = 0;
2465 int res = 0;
2466
2467 exec_name = argv_[0];
2468
2469 if (argc < 3)
2470 usage_exit();
2471
2472 /* Setup default input stream settings */
2473 input.framerate.num = 30;
2474 input.framerate.den = 1;
2475 input.use_i420 = 1;
2476
2477 /* First parse the global configuration values, because we want to apply
2478 * other parameters on top of the default configuration provided by the
2479 * codec.
2480 */
2481 argv = argv_dup(argc - 1, argv_ + 1);
2482 parse_global_config(&global, argv);
2483
2484 {
2485 /* Now parse each stream's parameters. Using a local scope here
2486 * due to the use of 'stream' as loop variable in FOREACH_STREAM
2487 * loops
2488 */
2489 struct stream_state *stream = NULL;
2490
2491 do {
2492 stream = new_stream(&global, stream);
2493 stream_cnt++;
2494 if (!streams)
2495 streams = stream;
2496 } while (parse_stream_params(&global, stream, argv));
2497 }
2498
2499 /* Check for unrecognized options */
2500 for (argi = argv; *argi; argi++)
2501 if (argi[0][0] == '-' && argi[0][1])
2502 die("Error: Unrecognized option %s\n", *argi);
2503
2504 /* Handle non-option arguments */
2505 input.fn = argv[0];
2506
2507 if (!input.fn)
2508 usage_exit();
2509
2510 for (pass = global.pass ? global.pass - 1 : 0; pass < global.passes; pass++) {
2511 int frames_in = 0, seen_frames = 0;
2512 int64_t estimated_time_left = -1;
2513 int64_t average_rate = -1;
2514 off_t lagged_count = 0;
2515
2516 open_input_file(&input);
2517
2518 /* If the input file doesn't specify its w/h (raw files), try to get
2519 * the data from the first stream's configuration.
2520 */
2521 if (!input.w || !input.h)
2522 FOREACH_STREAM( {
2523 if (stream->config.cfg.g_w && stream->config.cfg.g_h) {
2524 input.w = stream->config.cfg.g_w;
2525 input.h = stream->config.cfg.g_h;
2526 break;
2527 }
2528 });
2529
2530 /* Update stream configurations from the input file's parameters */
2531 if (!input.w || !input.h)
2532 fatal("Specify stream dimensions with --width (-w) "
2533 " and --height (-h)");
2534 FOREACH_STREAM(set_stream_dimensions(stream, input.w, input.h));
2535 FOREACH_STREAM(validate_stream_config(stream));
2536
2537 /* Ensure that --passes and --pass are consistent. If --pass is set and
2538 * --passes=2, ensure --fpf was set.
2539 */
2540 if (global.pass && global.passes == 2)
2541 FOREACH_STREAM( {
2542 if (!stream->config.stats_fn)
2543 die("Stream %d: Must specify --fpf when --pass=%d"
2544 " and --passes=2\n", stream->index, global.pass);
2545 });
2546
2547
2548 /* Use the frame rate from the file only if none was specified
2549 * on the command-line.
2550 */
2551 if (!global.have_framerate)
2552 global.framerate = input.framerate;
2553
2554 FOREACH_STREAM(set_default_kf_interval(stream, &global));
2555
2556 /* Show configuration */
2557 if (global.verbose && pass == 0)
2558 FOREACH_STREAM(show_stream_config(stream, &global, &input));
2559
2560 if (pass == (global.pass ? global.pass - 1 : 0)) {
2561 if (input.file_type == FILE_TYPE_Y4M)
2562 /*The Y4M reader does its own allocation.
2563 Just initialize this here to avoid problems if we never read any
2564 frames.*/
2565 memset(&raw, 0, sizeof(raw));
2566 else
2567 vpx_img_alloc(&raw,
2568 input.use_i420 ? VPX_IMG_FMT_I420
2569 : VPX_IMG_FMT_YV12,
2570 input.w, input.h, 32);
2571
2572 FOREACH_STREAM(init_rate_histogram(&stream->rate_hist,
2573 &stream->config.cfg,
2574 &global.framerate));
2575 }
2576
2577 FOREACH_STREAM(open_output_file(stream, &global));
2578 FOREACH_STREAM(setup_pass(stream, &global, pass));
2579 FOREACH_STREAM(initialize_encoder(stream, &global));
2580
2581 frame_avail = 1;
2582 got_data = 0;
2583
2584 while (frame_avail || got_data) {
2585 struct vpx_usec_timer timer;
2586
2587 if (!global.limit || frames_in < global.limit) {
2588 frame_avail = read_frame(&input, &raw);
2589
2590 if (frame_avail)
2591 frames_in++;
2592 seen_frames = frames_in > global.skip_frames ?
2593 frames_in - global.skip_frames : 0;
2594
2595 if (!global.quiet) {
2596 float fps = usec_to_fps(cx_time, seen_frames);
2597 fprintf(stderr, "\rPass %d/%d ", pass + 1, global.passes);
2598
2599 if (stream_cnt == 1)
2600 fprintf(stderr,
2601 "frame %4d/%-4d %7"PRId64"B ",
2602 frames_in, streams->frames_out, (int64_t)streams->nbytes);
2603 else
2604 fprintf(stderr, "frame %4d ", frames_in);
2605
2606 fprintf(stderr, "%7"PRId64" %s %.2f %s ",
2607 cx_time > 9999999 ? cx_time / 1000 : cx_time,
2608 cx_time > 9999999 ? "ms" : "us",
2609 fps >= 1.0 ? fps : 1000.0 / fps,
2610 fps >= 1.0 ? "fps" : "ms/f");
2611 print_time("ETA", estimated_time_left);
2612 fprintf(stderr, "\033[K");
2613 }
2614
2615 } else
2616 frame_avail = 0;
2617
2618 if (frames_in > global.skip_frames) {
2619 vpx_usec_timer_start(&timer);
2620 FOREACH_STREAM(encode_frame(stream, &global,
2621 frame_avail ? &raw : NULL,
2622 frames_in));
2623 vpx_usec_timer_mark(&timer);
2624 cx_time += vpx_usec_timer_elapsed(&timer);
2625
2626 FOREACH_STREAM(update_quantizer_histogram(stream));
2627
2628 got_data = 0;
2629 FOREACH_STREAM(get_cx_data(stream, &global, &got_data));
2630
2631 if (!got_data && input.length && !streams->frames_out) {
2632 lagged_count = global.limit ? seen_frames : ftello(input.file);
2633 } else if (input.length) {
2634 int64_t remaining;
2635 int64_t rate;
2636
2637 if (global.limit) {
2638 int frame_in_lagged = (seen_frames - lagged_count) * 1000;
2639
2640 rate = cx_time ? frame_in_lagged * (int64_t)1000000 / cx_time : 0;
2641 remaining = 1000 * (global.limit - global.skip_frames
2642 - seen_frames + lagged_count);
2643 } else {
2644 off_t input_pos = ftello(input.file);
2645 off_t input_pos_lagged = input_pos - lagged_count;
2646 int64_t limit = input.length;
2647
2648 rate = cx_time ? input_pos_lagged * (int64_t)1000000 / cx_time : 0;
2649 remaining = limit - input_pos + lagged_count;
2650 }
2651
2652 average_rate = (average_rate <= 0)
2653 ? rate
2654 : (average_rate * 7 + rate) / 8;
2655 estimated_time_left = average_rate ? remaining / average_rate : -1;
2656 }
2657
2658 if (got_data && global.test_decode != TEST_DECODE_OFF)
2659 FOREACH_STREAM(test_decode(stream, global.test_decode));
2660 }
2661
2662 fflush(stdout);
2663 }
2664
2665 if (stream_cnt > 1)
2666 fprintf(stderr, "\n");
2667
2668 if (!global.quiet)
2669 FOREACH_STREAM(fprintf(
2670 stderr,
2671 "\rPass %d/%d frame %4d/%-4d %7"PRId64"B %7lub/f %7"PRId64"b/s"
2672 " %7"PRId64" %s (%.2f fps)\033[K\n", pass + 1,
2673 global.passes, frames_in, stream->frames_out, (int64_t)stream->nbytes,
2674 seen_frames ? (unsigned long)(stream->nbytes * 8 / seen_frames) : 0,
2675 seen_frames ? (int64_t)stream->nbytes * 8
2676 * (int64_t)global.framerate.num / global.framerate.den
2677 / seen_frames
2678 : 0,
2679 stream->cx_time > 9999999 ? stream->cx_time / 1000 : stream->cx_time,
2680 stream->cx_time > 9999999 ? "ms" : "us",
2681 usec_to_fps(stream->cx_time, seen_frames));
2682 );
2683
2684 if (global.show_psnr)
2685 FOREACH_STREAM(show_psnr(stream));
2686
2687 FOREACH_STREAM(vpx_codec_destroy(&stream->encoder));
2688
2689 if (global.test_decode != TEST_DECODE_OFF) {
2690 FOREACH_STREAM(vpx_codec_destroy(&stream->decoder));
2691 FOREACH_STREAM(vpx_img_free(&stream->ref_enc.img));
2692 FOREACH_STREAM(vpx_img_free(&stream->ref_dec.img));
2693 }
2694
2695 close_input_file(&input);
2696
2697 if (global.test_decode == TEST_DECODE_FATAL) {
2698 FOREACH_STREAM(res |= stream->mismatch_seen);
2699 }
2700 FOREACH_STREAM(close_output_file(stream, global.codec->fourcc));
2701
2702 FOREACH_STREAM(stats_close(&stream->stats, global.passes - 1));
2703
2704 if (global.pass)
2705 break;
2706 }
2707
2708 if (global.show_q_hist_buckets)
2709 FOREACH_STREAM(show_q_histogram(stream->counts,
2710 global.show_q_hist_buckets));
2711
2712 if (global.show_rate_hist_buckets)
2713 FOREACH_STREAM(show_rate_histogram(&stream->rate_hist,
2714 &stream->config.cfg,
2715 global.show_rate_hist_buckets));
2716 FOREACH_STREAM(destroy_rate_histogram(&stream->rate_hist));
2717
2718 #if CONFIG_INTERNAL_STATS
2719 /* TODO(jkoleszar): This doesn't belong in this executable. Do it for now,
2720 * to match some existing utilities.
2721 */
2722 FOREACH_STREAM({
2723 FILE *f = fopen("opsnr.stt", "a");
2724 if (stream->mismatch_seen) {
2725 fprintf(f, "First mismatch occurred in frame %d\n",
2726 stream->mismatch_seen);
2727 } else {
2728 fprintf(f, "No mismatch detected in recon buffers\n");
2729 }
2730 fclose(f);
2731 });
2732 #endif
2733
2734 vpx_img_free(&raw);
2735 free(argv);
2736 free(streams);
2737 return res ? EXIT_FAILURE : EXIT_SUCCESS;
2738 }
AV1 Codec Library