?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
d153 intra prediction (16x16) ssse3 using bytes
[aom.git] / vpxenc.c
blob71cf01feffdebc80f76f4de77b2992e8feb1e762
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
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_EndSubElement(glob, &videoStart);
735 }
736 Ebml_EndSubElement(glob, &start); /* Track Entry */
737 }
738 Ebml_EndSubElement(glob, &trackStart);
739 }
740 /* segment element is open */
741 }
742 }
743
744
745 static void
746 write_webm_block(EbmlGlobal *glob,
747 const vpx_codec_enc_cfg_t *cfg,
748 const vpx_codec_cx_pkt_t *pkt) {
749 unsigned long block_length;
750 unsigned char track_number;
751 unsigned short block_timecode = 0;
752 unsigned char flags;
753 int64_t pts_ms;
754 int start_cluster = 0, is_keyframe;
755
756 /* Calculate the PTS of this frame in milliseconds */
757 pts_ms = pkt->data.frame.pts * 1000
758 * (uint64_t)cfg->g_timebase.num / (uint64_t)cfg->g_timebase.den;
759 if (pts_ms <= glob->last_pts_ms)
760 pts_ms = glob->last_pts_ms + 1;
761 glob->last_pts_ms = pts_ms;
762
763 /* Calculate the relative time of this block */
764 if (pts_ms - glob->cluster_timecode > SHRT_MAX)
765 start_cluster = 1;
766 else
767 block_timecode = (unsigned short)pts_ms - glob->cluster_timecode;
768
769 is_keyframe = (pkt->data.frame.flags & VPX_FRAME_IS_KEY);
770 if (start_cluster || is_keyframe) {
771 if (glob->cluster_open)
772 Ebml_EndSubElement(glob, &glob->startCluster);
773
774 /* Open the new cluster */
775 block_timecode = 0;
776 glob->cluster_open = 1;
777 glob->cluster_timecode = (uint32_t)pts_ms;
778 glob->cluster_pos = ftello(glob->stream);
779 Ebml_StartSubElement(glob, &glob->startCluster, Cluster); /* cluster */
780 Ebml_SerializeUnsigned(glob, Timecode, glob->cluster_timecode);
781
782 /* Save a cue point if this is a keyframe. */
783 if (is_keyframe) {
784 struct cue_entry *cue, *new_cue_list;
785
786 new_cue_list = realloc(glob->cue_list,
787 (glob->cues + 1) * sizeof(struct cue_entry));
788 if (new_cue_list)
789 glob->cue_list = new_cue_list;
790 else
791 fatal("Failed to realloc cue list.");
792
793 cue = &glob->cue_list[glob->cues];
794 cue->time = glob->cluster_timecode;
795 cue->loc = glob->cluster_pos;
796 glob->cues++;
797 }
798 }
799
800 /* Write the Simple Block */
801 Ebml_WriteID(glob, SimpleBlock);
802
803 block_length = (unsigned long)pkt->data.frame.sz + 4;
804 block_length |= 0x10000000;
805 Ebml_Serialize(glob, &block_length, sizeof(block_length), 4);
806
807 track_number = 1;
808 track_number |= 0x80;
809 Ebml_Write(glob, &track_number, 1);
810
811 Ebml_Serialize(glob, &block_timecode, sizeof(block_timecode), 2);
812
813 flags = 0;
814 if (is_keyframe)
815 flags |= 0x80;
816 if (pkt->data.frame.flags & VPX_FRAME_IS_INVISIBLE)
817 flags |= 0x08;
818 Ebml_Write(glob, &flags, 1);
819
820 Ebml_Write(glob, pkt->data.frame.buf, (unsigned long)pkt->data.frame.sz);
821 }
822
823
824 static void
825 write_webm_file_footer(EbmlGlobal *glob, long hash) {
826
827 if (glob->cluster_open)
828 Ebml_EndSubElement(glob, &glob->startCluster);
829
830 {
831 EbmlLoc start;
832 unsigned int i;
833
834 glob->cue_pos = ftello(glob->stream);
835 Ebml_StartSubElement(glob, &start, Cues);
836 for (i = 0; i < glob->cues; i++) {
837 struct cue_entry *cue = &glob->cue_list[i];
838 EbmlLoc start;
839
840 Ebml_StartSubElement(glob, &start, CuePoint);
841 {
842 EbmlLoc start;
843
844 Ebml_SerializeUnsigned(glob, CueTime, cue->time);
845
846 Ebml_StartSubElement(glob, &start, CueTrackPositions);
847 Ebml_SerializeUnsigned(glob, CueTrack, 1);
848 Ebml_SerializeUnsigned64(glob, CueClusterPosition,
849 cue->loc - glob->position_reference);
850 Ebml_EndSubElement(glob, &start);
851 }
852 Ebml_EndSubElement(glob, &start);
853 }
854 Ebml_EndSubElement(glob, &start);
855 }
856
857 Ebml_EndSubElement(glob, &glob->startSegment);
858
859 /* Patch up the seek info block */
860 write_webm_seek_info(glob);
861
862 /* Patch up the track id */
863 fseeko(glob->stream, glob->track_id_pos, SEEK_SET);
864 Ebml_SerializeUnsigned32(glob, TrackUID, glob->debug ? 0xDEADBEEF : hash);
865
866 fseeko(glob->stream, 0, SEEK_END);
867 }
868
869
870 /* Murmur hash derived from public domain reference implementation at
871 * http:// sites.google.com/site/murmurhash/
872 */
873 static unsigned int murmur(const void *key, int len, unsigned int seed) {
874 const unsigned int m = 0x5bd1e995;
875 const int r = 24;
876
877 unsigned int h = seed ^ len;
878
879 const unsigned char *data = (const unsigned char *)key;
880
881 while (len >= 4) {
882 unsigned int k;
883
884 k = (unsigned int)data[0];
885 k |= (unsigned int)data[1] << 8;
886 k |= (unsigned int)data[2] << 16;
887 k |= (unsigned int)data[3] << 24;
888
889 k *= m;
890 k ^= k >> r;
891 k *= m;
892
893 h *= m;
894 h ^= k;
895
896 data += 4;
897 len -= 4;
898 }
899
900 switch (len) {
901 case 3:
902 h ^= data[2] << 16;
903 case 2:
904 h ^= data[1] << 8;
905 case 1:
906 h ^= data[0];
907 h *= m;
908 };
909
910 h ^= h >> 13;
911 h *= m;
912 h ^= h >> 15;
913
914 return h;
915 }
916
917 #include "math.h"
918 #define MAX_PSNR 100
919 static double vp8_mse2psnr(double Samples, double Peak, double Mse) {
920 double psnr;
921
922 if ((double)Mse > 0.0)
923 psnr = 10.0 * log10(Peak * Peak * Samples / Mse);
924 else
925 psnr = MAX_PSNR; /* Limit to prevent / 0 */
926
927 if (psnr > MAX_PSNR)
928 psnr = MAX_PSNR;
929
930 return psnr;
931 }
932
933
934 #include "args.h"
935 static const arg_def_t debugmode = ARG_DEF("D", "debug", 0,
936 "Debug mode (makes output deterministic)");
937 static const arg_def_t outputfile = ARG_DEF("o", "output", 1,
938 "Output filename");
939 static const arg_def_t use_yv12 = ARG_DEF(NULL, "yv12", 0,
940 "Input file is YV12 ");
941 static const arg_def_t use_i420 = ARG_DEF(NULL, "i420", 0,
942 "Input file is I420 (default)");
943 static const arg_def_t codecarg = ARG_DEF(NULL, "codec", 1,
944 "Codec to use");
945 static const arg_def_t passes = ARG_DEF("p", "passes", 1,
946 "Number of passes (1/2)");
947 static const arg_def_t pass_arg = ARG_DEF(NULL, "pass", 1,
948 "Pass to execute (1/2)");
949 static const arg_def_t fpf_name = ARG_DEF(NULL, "fpf", 1,
950 "First pass statistics file name");
951 static const arg_def_t limit = ARG_DEF(NULL, "limit", 1,
952 "Stop encoding after n input frames");
953 static const arg_def_t skip = ARG_DEF(NULL, "skip", 1,
954 "Skip the first n input frames");
955 static const arg_def_t deadline = ARG_DEF("d", "deadline", 1,
956 "Deadline per frame (usec)");
957 static const arg_def_t best_dl = ARG_DEF(NULL, "best", 0,
958 "Use Best Quality Deadline");
959 static const arg_def_t good_dl = ARG_DEF(NULL, "good", 0,
960 "Use Good Quality Deadline");
961 static const arg_def_t rt_dl = ARG_DEF(NULL, "rt", 0,
962 "Use Realtime Quality Deadline");
963 static const arg_def_t quietarg = ARG_DEF("q", "quiet", 0,
964 "Do not print encode progress");
965 static const arg_def_t verbosearg = ARG_DEF("v", "verbose", 0,
966 "Show encoder parameters");
967 static const arg_def_t psnrarg = ARG_DEF(NULL, "psnr", 0,
968 "Show PSNR in status line");
969 enum TestDecodeFatality {
970 TEST_DECODE_OFF,
971 TEST_DECODE_FATAL,
972 TEST_DECODE_WARN,
973 };
974 static const struct arg_enum_list test_decode_enum[] = {
975 {"off", TEST_DECODE_OFF},
976 {"fatal", TEST_DECODE_FATAL},
977 {"warn", TEST_DECODE_WARN},
978 {NULL, 0}
979 };
980 static const arg_def_t recontest = ARG_DEF_ENUM(NULL, "test-decode", 1,
981 "Test encode/decode mismatch",
982 test_decode_enum);
983 static const arg_def_t framerate = ARG_DEF(NULL, "fps", 1,
984 "Stream frame rate (rate/scale)");
985 static const arg_def_t use_ivf = ARG_DEF(NULL, "ivf", 0,
986 "Output IVF (default is WebM)");
987 static const arg_def_t out_part = ARG_DEF("P", "output-partitions", 0,
988 "Makes encoder output partitions. Requires IVF output!");
989 static const arg_def_t q_hist_n = ARG_DEF(NULL, "q-hist", 1,
990 "Show quantizer histogram (n-buckets)");
991 static const arg_def_t rate_hist_n = ARG_DEF(NULL, "rate-hist", 1,
992 "Show rate histogram (n-buckets)");
993 static const arg_def_t *main_args[] = {
994 &debugmode,
995 &outputfile, &codecarg, &passes, &pass_arg, &fpf_name, &limit, &skip,
996 &deadline, &best_dl, &good_dl, &rt_dl,
997 &quietarg, &verbosearg, &psnrarg, &use_ivf, &out_part, &q_hist_n, &rate_hist_n,
998 NULL
999 };
1000
1001 static const arg_def_t usage = ARG_DEF("u", "usage", 1,
1002 "Usage profile number to use");
1003 static const arg_def_t threads = ARG_DEF("t", "threads", 1,
1004 "Max number of threads to use");
1005 static const arg_def_t profile = ARG_DEF(NULL, "profile", 1,
1006 "Bitstream profile number to use");
1007 static const arg_def_t width = ARG_DEF("w", "width", 1,
1008 "Frame width");
1009 static const arg_def_t height = ARG_DEF("h", "height", 1,
1010 "Frame height");
1011 static const struct arg_enum_list stereo_mode_enum[] = {
1012 {"mono", STEREO_FORMAT_MONO},
1013 {"left-right", STEREO_FORMAT_LEFT_RIGHT},
1014 {"bottom-top", STEREO_FORMAT_BOTTOM_TOP},
1015 {"top-bottom", STEREO_FORMAT_TOP_BOTTOM},
1016 {"right-left", STEREO_FORMAT_RIGHT_LEFT},
1017 {NULL, 0}
1018 };
1019 static const arg_def_t stereo_mode = ARG_DEF_ENUM(NULL, "stereo-mode", 1,
1020 "Stereo 3D video format", stereo_mode_enum);
1021 static const arg_def_t timebase = ARG_DEF(NULL, "timebase", 1,
1022 "Output timestamp precision (fractional seconds)");
1023 static const arg_def_t error_resilient = ARG_DEF(NULL, "error-resilient", 1,
1024 "Enable error resiliency features");
1025 static const arg_def_t lag_in_frames = ARG_DEF(NULL, "lag-in-frames", 1,
1026 "Max number of frames to lag");
1027
1028 static const arg_def_t *global_args[] = {
1029 &use_yv12, &use_i420, &usage, &threads, &profile,
1030 &width, &height, &stereo_mode, &timebase, &framerate,
1031 &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 {"q", VPX_Q},
1048 {NULL, 0}
1049 };
1050 static const arg_def_t end_usage = ARG_DEF_ENUM(NULL, "end-usage", 1,
1051 "Rate control mode", end_usage_enum);
1052 static const arg_def_t target_bitrate = ARG_DEF(NULL, "target-bitrate", 1,
1053 "Bitrate (kbps)");
1054 static const arg_def_t min_quantizer = ARG_DEF(NULL, "min-q", 1,
1055 "Minimum (best) quantizer");
1056 static const arg_def_t max_quantizer = ARG_DEF(NULL, "max-q", 1,
1057 "Maximum (worst) quantizer");
1058 static const arg_def_t undershoot_pct = ARG_DEF(NULL, "undershoot-pct", 1,
1059 "Datarate undershoot (min) target (%)");
1060 static const arg_def_t overshoot_pct = ARG_DEF(NULL, "overshoot-pct", 1,
1061 "Datarate overshoot (max) target (%)");
1062 static const arg_def_t buf_sz = ARG_DEF(NULL, "buf-sz", 1,
1063 "Client buffer size (ms)");
1064 static const arg_def_t buf_initial_sz = ARG_DEF(NULL, "buf-initial-sz", 1,
1065 "Client initial buffer size (ms)");
1066 static const arg_def_t buf_optimal_sz = ARG_DEF(NULL, "buf-optimal-sz", 1,
1067 "Client optimal buffer size (ms)");
1068 static const arg_def_t *rc_args[] = {
1069 &dropframe_thresh, &resize_allowed, &resize_up_thresh, &resize_down_thresh,
1070 &end_usage, &target_bitrate, &min_quantizer, &max_quantizer,
1071 &undershoot_pct, &overshoot_pct, &buf_sz, &buf_initial_sz, &buf_optimal_sz,
1072 NULL
1073 };
1074
1075
1076 static const arg_def_t bias_pct = ARG_DEF(NULL, "bias-pct", 1,
1077 "CBR/VBR bias (0=CBR, 100=VBR)");
1078 static const arg_def_t minsection_pct = ARG_DEF(NULL, "minsection-pct", 1,
1079 "GOP min bitrate (% of target)");
1080 static const arg_def_t maxsection_pct = ARG_DEF(NULL, "maxsection-pct", 1,
1081 "GOP max bitrate (% of target)");
1082 static const arg_def_t *rc_twopass_args[] = {
1083 &bias_pct, &minsection_pct, &maxsection_pct, NULL
1084 };
1085
1086
1087 static const arg_def_t kf_min_dist = ARG_DEF(NULL, "kf-min-dist", 1,
1088 "Minimum keyframe interval (frames)");
1089 static const arg_def_t kf_max_dist = ARG_DEF(NULL, "kf-max-dist", 1,
1090 "Maximum keyframe interval (frames)");
1091 static const arg_def_t kf_disabled = ARG_DEF(NULL, "disable-kf", 0,
1092 "Disable keyframe placement");
1093 static const arg_def_t *kf_args[] = {
1094 &kf_min_dist, &kf_max_dist, &kf_disabled, NULL
1095 };
1096
1097
1098 static const arg_def_t noise_sens = ARG_DEF(NULL, "noise-sensitivity", 1,
1099 "Noise sensitivity (frames to blur)");
1100 static const arg_def_t sharpness = ARG_DEF(NULL, "sharpness", 1,
1101 "Filter sharpness (0-7)");
1102 static const arg_def_t static_thresh = ARG_DEF(NULL, "static-thresh", 1,
1103 "Motion detection threshold");
1104 static const arg_def_t cpu_used = ARG_DEF(NULL, "cpu-used", 1,
1105 "CPU Used (-16..16)");
1106 static const arg_def_t token_parts = ARG_DEF(NULL, "token-parts", 1,
1107 "Number of token partitions to use, log2");
1108 static const arg_def_t tile_cols = ARG_DEF(NULL, "tile-columns", 1,
1109 "Number of tile columns to use, log2");
1110 static const arg_def_t tile_rows = ARG_DEF(NULL, "tile-rows", 1,
1111 "Number of tile rows to use, log2");
1112 static const arg_def_t auto_altref = ARG_DEF(NULL, "auto-alt-ref", 1,
1113 "Enable automatic alt reference frames");
1114 static const arg_def_t arnr_maxframes = ARG_DEF(NULL, "arnr-maxframes", 1,
1115 "AltRef Max Frames");
1116 static const arg_def_t arnr_strength = ARG_DEF(NULL, "arnr-strength", 1,
1117 "AltRef Strength");
1118 static const arg_def_t arnr_type = ARG_DEF(NULL, "arnr-type", 1,
1119 "AltRef Type");
1120 static const struct arg_enum_list tuning_enum[] = {
1121 {"psnr", VP8_TUNE_PSNR},
1122 {"ssim", VP8_TUNE_SSIM},
1123 {NULL, 0}
1124 };
1125 static const arg_def_t tune_ssim = ARG_DEF_ENUM(NULL, "tune", 1,
1126 "Material to favor", tuning_enum);
1127 static const arg_def_t cq_level = ARG_DEF(NULL, "cq-level", 1,
1128 "Constant/Constrained Quality level");
1129 static const arg_def_t max_intra_rate_pct = ARG_DEF(NULL, "max-intra-rate", 1,
1130 "Max I-frame bitrate (pct)");
1131 static const arg_def_t lossless = ARG_DEF(NULL, "lossless", 1, "Lossless mode");
1132 #if CONFIG_VP9_ENCODER
1133 static const arg_def_t frame_parallel_decoding = ARG_DEF(
1134 NULL, "frame-parallel", 1, "Enable frame parallel decodability features");
1135 #endif
1136
1137 #if CONFIG_VP8_ENCODER
1138 static const arg_def_t *vp8_args[] = {
1139 &cpu_used, &auto_altref, &noise_sens, &sharpness, &static_thresh,
1140 &token_parts, &arnr_maxframes, &arnr_strength, &arnr_type,
1141 &tune_ssim, &cq_level, &max_intra_rate_pct,
1142 NULL
1143 };
1144 static const int vp8_arg_ctrl_map[] = {
1145 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
1146 VP8E_SET_NOISE_SENSITIVITY, VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
1147 VP8E_SET_TOKEN_PARTITIONS,
1148 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
1149 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
1150 0
1151 };
1152 #endif
1153
1154 #if CONFIG_VP9_ENCODER
1155 static const arg_def_t *vp9_args[] = {
1156 &cpu_used, &auto_altref, &noise_sens, &sharpness, &static_thresh,
1157 &tile_cols, &tile_rows, &arnr_maxframes, &arnr_strength, &arnr_type,
1158 &tune_ssim, &cq_level, &max_intra_rate_pct, &lossless,
1159 &frame_parallel_decoding,
1160 NULL
1161 };
1162 static const int vp9_arg_ctrl_map[] = {
1163 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
1164 VP8E_SET_NOISE_SENSITIVITY, VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
1165 VP9E_SET_TILE_COLUMNS, VP9E_SET_TILE_ROWS,
1166 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
1167 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
1168 VP9E_SET_LOSSLESS, VP9E_SET_FRAME_PARALLEL_DECODING,
1169 0
1170 };
1171 #endif
1172
1173 static const arg_def_t *no_args[] = { NULL };
1174
1175 static void usage_exit() {
1176 int i;
1177
1178 fprintf(stderr, "Usage: %s <options> -o dst_filename src_filename \n",
1179 exec_name);
1180
1181 fprintf(stderr, "\nOptions:\n");
1182 arg_show_usage(stderr, main_args);
1183 fprintf(stderr, "\nEncoder Global Options:\n");
1184 arg_show_usage(stderr, global_args);
1185 fprintf(stderr, "\nRate Control Options:\n");
1186 arg_show_usage(stderr, rc_args);
1187 fprintf(stderr, "\nTwopass Rate Control Options:\n");
1188 arg_show_usage(stderr, rc_twopass_args);
1189 fprintf(stderr, "\nKeyframe Placement Options:\n");
1190 arg_show_usage(stderr, kf_args);
1191 #if CONFIG_VP8_ENCODER
1192 fprintf(stderr, "\nVP8 Specific Options:\n");
1193 arg_show_usage(stderr, vp8_args);
1194 #endif
1195 #if CONFIG_VP9_ENCODER
1196 fprintf(stderr, "\nVP9 Specific Options:\n");
1197 arg_show_usage(stderr, vp9_args);
1198 #endif
1199 fprintf(stderr, "\nStream timebase (--timebase):\n"
1200 " The desired precision of timestamps in the output, expressed\n"
1201 " in fractional seconds. Default is 1/1000.\n");
1202 fprintf(stderr, "\n"
1203 "Included encoders:\n"
1204 "\n");
1205
1206 for (i = 0; i < sizeof(codecs) / sizeof(codecs[0]); i++)
1207 fprintf(stderr, " %-6s - %s\n",
1208 codecs[i].name,
1209 vpx_codec_iface_name(codecs[i].iface()));
1210
1211 exit(EXIT_FAILURE);
1212 }
1213
1214
1215 #define HIST_BAR_MAX 40
1216 struct hist_bucket {
1217 int low, high, count;
1218 };
1219
1220
1221 static int merge_hist_buckets(struct hist_bucket *bucket,
1222 int *buckets_,
1223 int max_buckets) {
1224 int small_bucket = 0, merge_bucket = INT_MAX, big_bucket = 0;
1225 int buckets = *buckets_;
1226 int i;
1227
1228 /* Find the extrema for this list of buckets */
1229 big_bucket = small_bucket = 0;
1230 for (i = 0; i < buckets; i++) {
1231 if (bucket[i].count < bucket[small_bucket].count)
1232 small_bucket = i;
1233 if (bucket[i].count > bucket[big_bucket].count)
1234 big_bucket = i;
1235 }
1236
1237 /* If we have too many buckets, merge the smallest with an adjacent
1238 * bucket.
1239 */
1240 while (buckets > max_buckets) {
1241 int last_bucket = buckets - 1;
1242
1243 /* merge the small bucket with an adjacent one. */
1244 if (small_bucket == 0)
1245 merge_bucket = 1;
1246 else if (small_bucket == last_bucket)
1247 merge_bucket = last_bucket - 1;
1248 else if (bucket[small_bucket - 1].count < bucket[small_bucket + 1].count)
1249 merge_bucket = small_bucket - 1;
1250 else
1251 merge_bucket = small_bucket + 1;
1252
1253 assert(abs(merge_bucket - small_bucket) <= 1);
1254 assert(small_bucket < buckets);
1255 assert(big_bucket < buckets);
1256 assert(merge_bucket < buckets);
1257
1258 if (merge_bucket < small_bucket) {
1259 bucket[merge_bucket].high = bucket[small_bucket].high;
1260 bucket[merge_bucket].count += bucket[small_bucket].count;
1261 } else {
1262 bucket[small_bucket].high = bucket[merge_bucket].high;
1263 bucket[small_bucket].count += bucket[merge_bucket].count;
1264 merge_bucket = small_bucket;
1265 }
1266
1267 assert(bucket[merge_bucket].low != bucket[merge_bucket].high);
1268
1269 buckets--;
1270
1271 /* Remove the merge_bucket from the list, and find the new small
1272 * and big buckets while we're at it
1273 */
1274 big_bucket = small_bucket = 0;
1275 for (i = 0; i < buckets; i++) {
1276 if (i > merge_bucket)
1277 bucket[i] = bucket[i + 1];
1278
1279 if (bucket[i].count < bucket[small_bucket].count)
1280 small_bucket = i;
1281 if (bucket[i].count > bucket[big_bucket].count)
1282 big_bucket = i;
1283 }
1284
1285 }
1286
1287 *buckets_ = buckets;
1288 return bucket[big_bucket].count;
1289 }
1290
1291
1292 static void show_histogram(const struct hist_bucket *bucket,
1293 int buckets,
1294 int total,
1295 int scale) {
1296 const char *pat1, *pat2;
1297 int i;
1298
1299 switch ((int)(log(bucket[buckets - 1].high) / log(10)) + 1) {
1300 case 1:
1301 case 2:
1302 pat1 = "%4d %2s: ";
1303 pat2 = "%4d-%2d: ";
1304 break;
1305 case 3:
1306 pat1 = "%5d %3s: ";
1307 pat2 = "%5d-%3d: ";
1308 break;
1309 case 4:
1310 pat1 = "%6d %4s: ";
1311 pat2 = "%6d-%4d: ";
1312 break;
1313 case 5:
1314 pat1 = "%7d %5s: ";
1315 pat2 = "%7d-%5d: ";
1316 break;
1317 case 6:
1318 pat1 = "%8d %6s: ";
1319 pat2 = "%8d-%6d: ";
1320 break;
1321 case 7:
1322 pat1 = "%9d %7s: ";
1323 pat2 = "%9d-%7d: ";
1324 break;
1325 default:
1326 pat1 = "%12d %10s: ";
1327 pat2 = "%12d-%10d: ";
1328 break;
1329 }
1330
1331 for (i = 0; i < buckets; i++) {
1332 int len;
1333 int j;
1334 float pct;
1335
1336 pct = (float)(100.0 * bucket[i].count / total);
1337 len = HIST_BAR_MAX * bucket[i].count / scale;
1338 if (len < 1)
1339 len = 1;
1340 assert(len <= HIST_BAR_MAX);
1341
1342 if (bucket[i].low == bucket[i].high)
1343 fprintf(stderr, pat1, bucket[i].low, "");
1344 else
1345 fprintf(stderr, pat2, bucket[i].low, bucket[i].high);
1346
1347 for (j = 0; j < HIST_BAR_MAX; j++)
1348 fprintf(stderr, j < len ? "=" : " ");
1349 fprintf(stderr, "\t%5d (%6.2f%%)\n", bucket[i].count, pct);
1350 }
1351 }
1352
1353
1354 static void show_q_histogram(const int counts[64], int max_buckets) {
1355 struct hist_bucket bucket[64];
1356 int buckets = 0;
1357 int total = 0;
1358 int scale;
1359 int i;
1360
1361
1362 for (i = 0; i < 64; i++) {
1363 if (counts[i]) {
1364 bucket[buckets].low = bucket[buckets].high = i;
1365 bucket[buckets].count = counts[i];
1366 buckets++;
1367 total += counts[i];
1368 }
1369 }
1370
1371 fprintf(stderr, "\nQuantizer Selection:\n");
1372 scale = merge_hist_buckets(bucket, &buckets, max_buckets);
1373 show_histogram(bucket, buckets, total, scale);
1374 }
1375
1376
1377 #define RATE_BINS (100)
1378 struct rate_hist {
1379 int64_t *pts;
1380 int *sz;
1381 int samples;
1382 int frames;
1383 struct hist_bucket bucket[RATE_BINS];
1384 int total;
1385 };
1386
1387
1388 static void init_rate_histogram(struct rate_hist *hist,
1389 const vpx_codec_enc_cfg_t *cfg,
1390 const vpx_rational_t *fps) {
1391 int i;
1392
1393 /* Determine the number of samples in the buffer. Use the file's framerate
1394 * to determine the number of frames in rc_buf_sz milliseconds, with an
1395 * adjustment (5/4) to account for alt-refs
1396 */
1397 hist->samples = cfg->rc_buf_sz * 5 / 4 * fps->num / fps->den / 1000;
1398
1399 /* prevent division by zero */
1400 if (hist->samples == 0)
1401 hist->samples = 1;
1402
1403 hist->pts = calloc(hist->samples, sizeof(*hist->pts));
1404 hist->sz = calloc(hist->samples, sizeof(*hist->sz));
1405 for (i = 0; i < RATE_BINS; i++) {
1406 hist->bucket[i].low = INT_MAX;
1407 hist->bucket[i].high = 0;
1408 hist->bucket[i].count = 0;
1409 }
1410 }
1411
1412
1413 static void destroy_rate_histogram(struct rate_hist *hist) {
1414 free(hist->pts);
1415 free(hist->sz);
1416 }
1417
1418
1419 static void update_rate_histogram(struct rate_hist *hist,
1420 const vpx_codec_enc_cfg_t *cfg,
1421 const vpx_codec_cx_pkt_t *pkt) {
1422 int i, idx;
1423 int64_t now, then, sum_sz = 0, avg_bitrate;
1424
1425 now = pkt->data.frame.pts * 1000
1426 * (uint64_t)cfg->g_timebase.num / (uint64_t)cfg->g_timebase.den;
1427
1428 idx = hist->frames++ % hist->samples;
1429 hist->pts[idx] = now;
1430 hist->sz[idx] = (int)pkt->data.frame.sz;
1431
1432 if (now < cfg->rc_buf_initial_sz)
1433 return;
1434
1435 then = now;
1436
1437 /* Sum the size over the past rc_buf_sz ms */
1438 for (i = hist->frames; i > 0 && hist->frames - i < hist->samples; i--) {
1439 int i_idx = (i - 1) % hist->samples;
1440
1441 then = hist->pts[i_idx];
1442 if (now - then > cfg->rc_buf_sz)
1443 break;
1444 sum_sz += hist->sz[i_idx];
1445 }
1446
1447 if (now == then)
1448 return;
1449
1450 avg_bitrate = sum_sz * 8 * 1000 / (now - then);
1451 idx = (int)(avg_bitrate * (RATE_BINS / 2) / (cfg->rc_target_bitrate * 1000));
1452 if (idx < 0)
1453 idx = 0;
1454 if (idx > RATE_BINS - 1)
1455 idx = RATE_BINS - 1;
1456 if (hist->bucket[idx].low > avg_bitrate)
1457 hist->bucket[idx].low = (int)avg_bitrate;
1458 if (hist->bucket[idx].high < avg_bitrate)
1459 hist->bucket[idx].high = (int)avg_bitrate;
1460 hist->bucket[idx].count++;
1461 hist->total++;
1462 }
1463
1464
1465 static void show_rate_histogram(struct rate_hist *hist,
1466 const vpx_codec_enc_cfg_t *cfg,
1467 int max_buckets) {
1468 int i, scale;
1469 int buckets = 0;
1470
1471 for (i = 0; i < RATE_BINS; i++) {
1472 if (hist->bucket[i].low == INT_MAX)
1473 continue;
1474 hist->bucket[buckets++] = hist->bucket[i];
1475 }
1476
1477 fprintf(stderr, "\nRate (over %dms window):\n", cfg->rc_buf_sz);
1478 scale = merge_hist_buckets(hist->bucket, &buckets, max_buckets);
1479 show_histogram(hist->bucket, buckets, hist->total, scale);
1480 }
1481
1482 #define mmin(a, b) ((a) < (b) ? (a) : (b))
1483 static void find_mismatch(vpx_image_t *img1, vpx_image_t *img2,
1484 int yloc[4], int uloc[4], int vloc[4]) {
1485 const unsigned int bsize = 64;
1486 const unsigned int bsizey = bsize >> img1->y_chroma_shift;
1487 const unsigned int bsizex = bsize >> img1->x_chroma_shift;
1488 const int c_w = (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
1489 const int c_h = (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
1490 unsigned int match = 1;
1491 unsigned int i, j;
1492 yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
1493 for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
1494 for (j = 0; match && j < img1->d_w; j += bsize) {
1495 int k, l;
1496 int si = mmin(i + bsize, img1->d_h) - i;
1497 int sj = mmin(j + bsize, img1->d_w) - j;
1498 for (k = 0; match && k < si; k++)
1499 for (l = 0; match && l < sj; l++) {
1500 if (*(img1->planes[VPX_PLANE_Y] +
1501 (i + k) * img1->stride[VPX_PLANE_Y] + j + l) !=
1502 *(img2->planes[VPX_PLANE_Y] +
1503 (i + k) * img2->stride[VPX_PLANE_Y] + j + l)) {
1504 yloc[0] = i + k;
1505 yloc[1] = j + l;
1506 yloc[2] = *(img1->planes[VPX_PLANE_Y] +
1507 (i + k) * img1->stride[VPX_PLANE_Y] + j + l);
1508 yloc[3] = *(img2->planes[VPX_PLANE_Y] +
1509 (i + k) * img2->stride[VPX_PLANE_Y] + j + l);
1510 match = 0;
1511 break;
1512 }
1513 }
1514 }
1515 }
1516
1517 uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
1518 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
1519 for (j = 0; match && j < c_w; j += bsizex) {
1520 int k, l;
1521 int si = mmin(i + bsizey, c_h - i);
1522 int sj = mmin(j + bsizex, c_w - j);
1523 for (k = 0; match && k < si; k++)
1524 for (l = 0; match && l < sj; l++) {
1525 if (*(img1->planes[VPX_PLANE_U] +
1526 (i + k) * img1->stride[VPX_PLANE_U] + j + l) !=
1527 *(img2->planes[VPX_PLANE_U] +
1528 (i + k) * img2->stride[VPX_PLANE_U] + j + l)) {
1529 uloc[0] = i + k;
1530 uloc[1] = j + l;
1531 uloc[2] = *(img1->planes[VPX_PLANE_U] +
1532 (i + k) * img1->stride[VPX_PLANE_U] + j + l);
1533 uloc[3] = *(img2->planes[VPX_PLANE_U] +
1534 (i + k) * img2->stride[VPX_PLANE_V] + j + l);
1535 match = 0;
1536 break;
1537 }
1538 }
1539 }
1540 }
1541 vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
1542 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
1543 for (j = 0; match && j < c_w; j += bsizex) {
1544 int k, l;
1545 int si = mmin(i + bsizey, c_h - i);
1546 int sj = mmin(j + bsizex, c_w - j);
1547 for (k = 0; match && k < si; k++)
1548 for (l = 0; match && l < sj; l++) {
1549 if (*(img1->planes[VPX_PLANE_V] +
1550 (i + k) * img1->stride[VPX_PLANE_V] + j + l) !=
1551 *(img2->planes[VPX_PLANE_V] +
1552 (i + k) * img2->stride[VPX_PLANE_V] + j + l)) {
1553 vloc[0] = i + k;
1554 vloc[1] = j + l;
1555 vloc[2] = *(img1->planes[VPX_PLANE_V] +
1556 (i + k) * img1->stride[VPX_PLANE_V] + j + l);
1557 vloc[3] = *(img2->planes[VPX_PLANE_V] +
1558 (i + k) * img2->stride[VPX_PLANE_V] + j + l);
1559 match = 0;
1560 break;
1561 }
1562 }
1563 }
1564 }
1565 }
1566
1567 static int compare_img(vpx_image_t *img1, vpx_image_t *img2)
1568 {
1569 const int c_w = (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
1570 const int c_h = (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
1571 int match = 1;
1572 unsigned int i;
1573
1574 match &= (img1->fmt == img2->fmt);
1575 match &= (img1->w == img2->w);
1576 match &= (img1->h == img2->h);
1577
1578 for (i = 0; i < img1->d_h; i++)
1579 match &= (memcmp(img1->planes[VPX_PLANE_Y]+i*img1->stride[VPX_PLANE_Y],
1580 img2->planes[VPX_PLANE_Y]+i*img2->stride[VPX_PLANE_Y],
1581 img1->d_w) == 0);
1582
1583 for (i = 0; i < c_h; i++)
1584 match &= (memcmp(img1->planes[VPX_PLANE_U]+i*img1->stride[VPX_PLANE_U],
1585 img2->planes[VPX_PLANE_U]+i*img2->stride[VPX_PLANE_U],
1586 c_w) == 0);
1587
1588 for (i = 0; i < c_h; i++)
1589 match &= (memcmp(img1->planes[VPX_PLANE_V]+i*img1->stride[VPX_PLANE_U],
1590 img2->planes[VPX_PLANE_V]+i*img2->stride[VPX_PLANE_U],
1591 c_w) == 0);
1592
1593 return match;
1594 }
1595
1596
1597 #define NELEMENTS(x) (sizeof(x)/sizeof(x[0]))
1598 #define MAX(x,y) ((x)>(y)?(x):(y))
1599 #if CONFIG_VP8_ENCODER && !CONFIG_VP9_ENCODER
1600 #define ARG_CTRL_CNT_MAX NELEMENTS(vp8_arg_ctrl_map)
1601 #elif !CONFIG_VP8_ENCODER && CONFIG_VP9_ENCODER
1602 #define ARG_CTRL_CNT_MAX NELEMENTS(vp9_arg_ctrl_map)
1603 #else
1604 #define ARG_CTRL_CNT_MAX MAX(NELEMENTS(vp8_arg_ctrl_map), \
1605 NELEMENTS(vp9_arg_ctrl_map))
1606 #endif
1607
1608 /* Configuration elements common to all streams */
1609 struct global_config {
1610 const struct codec_item *codec;
1611 int passes;
1612 int pass;
1613 int usage;
1614 int deadline;
1615 int use_i420;
1616 int quiet;
1617 int verbose;
1618 int limit;
1619 int skip_frames;
1620 int show_psnr;
1621 enum TestDecodeFatality test_decode;
1622 int have_framerate;
1623 struct vpx_rational framerate;
1624 int out_part;
1625 int debug;
1626 int show_q_hist_buckets;
1627 int show_rate_hist_buckets;
1628 };
1629
1630
1631 /* Per-stream configuration */
1632 struct stream_config {
1633 struct vpx_codec_enc_cfg cfg;
1634 const char *out_fn;
1635 const char *stats_fn;
1636 stereo_format_t stereo_fmt;
1637 int arg_ctrls[ARG_CTRL_CNT_MAX][2];
1638 int arg_ctrl_cnt;
1639 int write_webm;
1640 int have_kf_max_dist;
1641 };
1642
1643
1644 struct stream_state {
1645 int index;
1646 struct stream_state *next;
1647 struct stream_config config;
1648 FILE *file;
1649 struct rate_hist rate_hist;
1650 EbmlGlobal ebml;
1651 uint32_t hash;
1652 uint64_t psnr_sse_total;
1653 uint64_t psnr_samples_total;
1654 double psnr_totals[4];
1655 int psnr_count;
1656 int counts[64];
1657 vpx_codec_ctx_t encoder;
1658 unsigned int frames_out;
1659 uint64_t cx_time;
1660 size_t nbytes;
1661 stats_io_t stats;
1662 struct vpx_image *img;
1663 vpx_codec_ctx_t decoder;
1664 int mismatch_seen;
1665 };
1666
1667
1668 void validate_positive_rational(const char *msg,
1669 struct vpx_rational *rat) {
1670 if (rat->den < 0) {
1671 rat->num *= -1;
1672 rat->den *= -1;
1673 }
1674
1675 if (rat->num < 0)
1676 die("Error: %s must be positive\n", msg);
1677
1678 if (!rat->den)
1679 die("Error: %s has zero denominator\n", msg);
1680 }
1681
1682
1683 static void parse_global_config(struct global_config *global, char **argv) {
1684 char **argi, **argj;
1685 struct arg arg;
1686
1687 /* Initialize default parameters */
1688 memset(global, 0, sizeof(*global));
1689 global->codec = codecs;
1690 global->passes = 0;
1691 global->use_i420 = 1;
1692 /* Assign default deadline to good quality */
1693 global->deadline = VPX_DL_GOOD_QUALITY;
1694
1695 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
1696 arg.argv_step = 1;
1697
1698 if (arg_match(&arg, &codecarg, argi)) {
1699 int j, k = -1;
1700
1701 for (j = 0; j < sizeof(codecs) / sizeof(codecs[0]); j++)
1702 if (!strcmp(codecs[j].name, arg.val))
1703 k = j;
1704
1705 if (k >= 0)
1706 global->codec = codecs + k;
1707 else
1708 die("Error: Unrecognized argument (%s) to --codec\n",
1709 arg.val);
1710
1711 } else if (arg_match(&arg, &passes, argi)) {
1712 global->passes = arg_parse_uint(&arg);
1713
1714 if (global->passes < 1 || global->passes > 2)
1715 die("Error: Invalid number of passes (%d)\n", global->passes);
1716 } else if (arg_match(&arg, &pass_arg, argi)) {
1717 global->pass = arg_parse_uint(&arg);
1718
1719 if (global->pass < 1 || global->pass > 2)
1720 die("Error: Invalid pass selected (%d)\n",
1721 global->pass);
1722 } else if (arg_match(&arg, &usage, argi))
1723 global->usage = arg_parse_uint(&arg);
1724 else if (arg_match(&arg, &deadline, argi))
1725 global->deadline = arg_parse_uint(&arg);
1726 else if (arg_match(&arg, &best_dl, argi))
1727 global->deadline = VPX_DL_BEST_QUALITY;
1728 else if (arg_match(&arg, &good_dl, argi))
1729 global->deadline = VPX_DL_GOOD_QUALITY;
1730 else if (arg_match(&arg, &rt_dl, argi))
1731 global->deadline = VPX_DL_REALTIME;
1732 else if (arg_match(&arg, &use_yv12, argi))
1733 global->use_i420 = 0;
1734 else if (arg_match(&arg, &use_i420, argi))
1735 global->use_i420 = 1;
1736 else if (arg_match(&arg, &quietarg, argi))
1737 global->quiet = 1;
1738 else if (arg_match(&arg, &verbosearg, argi))
1739 global->verbose = 1;
1740 else if (arg_match(&arg, &limit, argi))
1741 global->limit = arg_parse_uint(&arg);
1742 else if (arg_match(&arg, &skip, argi))
1743 global->skip_frames = arg_parse_uint(&arg);
1744 else if (arg_match(&arg, &psnrarg, argi))
1745 global->show_psnr = 1;
1746 else if (arg_match(&arg, &recontest, argi))
1747 global->test_decode = arg_parse_enum_or_int(&arg);
1748 else if (arg_match(&arg, &framerate, argi)) {
1749 global->framerate = arg_parse_rational(&arg);
1750 validate_positive_rational(arg.name, &global->framerate);
1751 global->have_framerate = 1;
1752 } else if (arg_match(&arg, &out_part, argi))
1753 global->out_part = 1;
1754 else if (arg_match(&arg, &debugmode, argi))
1755 global->debug = 1;
1756 else if (arg_match(&arg, &q_hist_n, argi))
1757 global->show_q_hist_buckets = arg_parse_uint(&arg);
1758 else if (arg_match(&arg, &rate_hist_n, argi))
1759 global->show_rate_hist_buckets = arg_parse_uint(&arg);
1760 else
1761 argj++;
1762 }
1763
1764 /* Validate global config */
1765 if (global->passes == 0) {
1766 #if CONFIG_VP9_ENCODER
1767 // Make default VP9 passes = 2 until there is a better quality 1-pass
1768 // encoder
1769 global->passes = (global->codec->iface == vpx_codec_vp9_cx ? 2 : 1);
1770 #else
1771 global->passes = 1;
1772 #endif
1773 }
1774
1775 if (global->pass) {
1776 /* DWIM: Assume the user meant passes=2 if pass=2 is specified */
1777 if (global->pass > global->passes) {
1778 warn("Assuming --pass=%d implies --passes=%d\n",
1779 global->pass, global->pass);
1780 global->passes = global->pass;
1781 }
1782 }
1783 }
1784
1785
1786 void open_input_file(struct input_state *input) {
1787 unsigned int fourcc;
1788
1789 /* Parse certain options from the input file, if possible */
1790 input->file = strcmp(input->fn, "-") ? fopen(input->fn, "rb")
1791 : set_binary_mode(stdin);
1792
1793 if (!input->file)
1794 fatal("Failed to open input file");
1795
1796 if (!fseeko(input->file, 0, SEEK_END)) {
1797 /* Input file is seekable. Figure out how long it is, so we can get
1798 * progress info.
1799 */
1800 input->length = ftello(input->file);
1801 rewind(input->file);
1802 }
1803
1804 /* For RAW input sources, these bytes will applied on the first frame
1805 * in read_frame().
1806 */
1807 input->detect.buf_read = fread(input->detect.buf, 1, 4, input->file);
1808 input->detect.position = 0;
1809
1810 if (input->detect.buf_read == 4
1811 && file_is_y4m(input->file, &input->y4m, input->detect.buf)) {
1812 if (y4m_input_open(&input->y4m, input->file, input->detect.buf, 4,
1813 input->only_i420) >= 0) {
1814 input->file_type = FILE_TYPE_Y4M;
1815 input->w = input->y4m.pic_w;
1816 input->h = input->y4m.pic_h;
1817 input->framerate.num = input->y4m.fps_n;
1818 input->framerate.den = input->y4m.fps_d;
1819 input->use_i420 = 0;
1820 } else
1821 fatal("Unsupported Y4M stream.");
1822 } else if (input->detect.buf_read == 4 && file_is_ivf(input, &fourcc)) {
1823 input->file_type = FILE_TYPE_IVF;
1824 switch (fourcc) {
1825 case 0x32315659:
1826 input->use_i420 = 0;
1827 break;
1828 case 0x30323449:
1829 input->use_i420 = 1;
1830 break;
1831 default:
1832 fatal("Unsupported fourcc (%08x) in IVF", fourcc);
1833 }
1834 } else {
1835 input->file_type = FILE_TYPE_RAW;
1836 }
1837 }
1838
1839
1840 static void close_input_file(struct input_state *input) {
1841 fclose(input->file);
1842 if (input->file_type == FILE_TYPE_Y4M)
1843 y4m_input_close(&input->y4m);
1844 }
1845
1846 static struct stream_state *new_stream(struct global_config *global,
1847 struct stream_state *prev) {
1848 struct stream_state *stream;
1849
1850 stream = calloc(1, sizeof(*stream));
1851 if (!stream)
1852 fatal("Failed to allocate new stream.");
1853 if (prev) {
1854 memcpy(stream, prev, sizeof(*stream));
1855 stream->index++;
1856 prev->next = stream;
1857 } else {
1858 vpx_codec_err_t res;
1859
1860 /* Populate encoder configuration */
1861 res = vpx_codec_enc_config_default(global->codec->iface(),
1862 &stream->config.cfg,
1863 global->usage);
1864 if (res)
1865 fatal("Failed to get config: %s\n", vpx_codec_err_to_string(res));
1866
1867 /* Change the default timebase to a high enough value so that the
1868 * encoder will always create strictly increasing timestamps.
1869 */
1870 stream->config.cfg.g_timebase.den = 1000;
1871
1872 /* Never use the library's default resolution, require it be parsed
1873 * from the file or set on the command line.
1874 */
1875 stream->config.cfg.g_w = 0;
1876 stream->config.cfg.g_h = 0;
1877
1878 /* Initialize remaining stream parameters */
1879 stream->config.stereo_fmt = STEREO_FORMAT_MONO;
1880 stream->config.write_webm = 1;
1881 stream->ebml.last_pts_ms = -1;
1882
1883 /* Allows removal of the application version from the EBML tags */
1884 stream->ebml.debug = global->debug;
1885 }
1886
1887 /* Output files must be specified for each stream */
1888 stream->config.out_fn = NULL;
1889
1890 stream->next = NULL;
1891 return stream;
1892 }
1893
1894
1895 static int parse_stream_params(struct global_config *global,
1896 struct stream_state *stream,
1897 char **argv) {
1898 char **argi, **argj;
1899 struct arg arg;
1900 static const arg_def_t **ctrl_args = no_args;
1901 static const int *ctrl_args_map = NULL;
1902 struct stream_config *config = &stream->config;
1903 int eos_mark_found = 0;
1904
1905 /* Handle codec specific options */
1906 if (0) {
1907 #if CONFIG_VP8_ENCODER
1908 } else if (global->codec->iface == vpx_codec_vp8_cx) {
1909 ctrl_args = vp8_args;
1910 ctrl_args_map = vp8_arg_ctrl_map;
1911 #endif
1912 #if CONFIG_VP9_ENCODER
1913 } else if (global->codec->iface == vpx_codec_vp9_cx) {
1914 ctrl_args = vp9_args;
1915 ctrl_args_map = vp9_arg_ctrl_map;
1916 #endif
1917 }
1918
1919 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
1920 arg.argv_step = 1;
1921
1922 /* Once we've found an end-of-stream marker (--) we want to continue
1923 * shifting arguments but not consuming them.
1924 */
1925 if (eos_mark_found) {
1926 argj++;
1927 continue;
1928 } else if (!strcmp(*argj, "--")) {
1929 eos_mark_found = 1;
1930 continue;
1931 }
1932
1933 if (0);
1934 else if (arg_match(&arg, &outputfile, argi))
1935 config->out_fn = arg.val;
1936 else if (arg_match(&arg, &fpf_name, argi))
1937 config->stats_fn = arg.val;
1938 else if (arg_match(&arg, &use_ivf, argi))
1939 config->write_webm = 0;
1940 else if (arg_match(&arg, &threads, argi))
1941 config->cfg.g_threads = arg_parse_uint(&arg);
1942 else if (arg_match(&arg, &profile, argi))
1943 config->cfg.g_profile = arg_parse_uint(&arg);
1944 else if (arg_match(&arg, &width, argi))
1945 config->cfg.g_w = arg_parse_uint(&arg);
1946 else if (arg_match(&arg, &height, argi))
1947 config->cfg.g_h = arg_parse_uint(&arg);
1948 else if (arg_match(&arg, &stereo_mode, argi))
1949 config->stereo_fmt = arg_parse_enum_or_int(&arg);
1950 else if (arg_match(&arg, &timebase, argi)) {
1951 config->cfg.g_timebase = arg_parse_rational(&arg);
1952 validate_positive_rational(arg.name, &config->cfg.g_timebase);
1953 } else if (arg_match(&arg, &error_resilient, argi))
1954 config->cfg.g_error_resilient = arg_parse_uint(&arg);
1955 else if (arg_match(&arg, &lag_in_frames, argi))
1956 config->cfg.g_lag_in_frames = arg_parse_uint(&arg);
1957 else if (arg_match(&arg, &dropframe_thresh, argi))
1958 config->cfg.rc_dropframe_thresh = arg_parse_uint(&arg);
1959 else if (arg_match(&arg, &resize_allowed, argi))
1960 config->cfg.rc_resize_allowed = arg_parse_uint(&arg);
1961 else if (arg_match(&arg, &resize_up_thresh, argi))
1962 config->cfg.rc_resize_up_thresh = arg_parse_uint(&arg);
1963 else if (arg_match(&arg, &resize_down_thresh, argi))
1964 config->cfg.rc_resize_down_thresh = arg_parse_uint(&arg);
1965 else if (arg_match(&arg, &end_usage, argi))
1966 config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
1967 else if (arg_match(&arg, &target_bitrate, argi))
1968 config->cfg.rc_target_bitrate = arg_parse_uint(&arg);
1969 else if (arg_match(&arg, &min_quantizer, argi))
1970 config->cfg.rc_min_quantizer = arg_parse_uint(&arg);
1971 else if (arg_match(&arg, &max_quantizer, argi))
1972 config->cfg.rc_max_quantizer = arg_parse_uint(&arg);
1973 else if (arg_match(&arg, &undershoot_pct, argi))
1974 config->cfg.rc_undershoot_pct = arg_parse_uint(&arg);
1975 else if (arg_match(&arg, &overshoot_pct, argi))
1976 config->cfg.rc_overshoot_pct = arg_parse_uint(&arg);
1977 else if (arg_match(&arg, &buf_sz, argi))
1978 config->cfg.rc_buf_sz = arg_parse_uint(&arg);
1979 else if (arg_match(&arg, &buf_initial_sz, argi))
1980 config->cfg.rc_buf_initial_sz = arg_parse_uint(&arg);
1981 else if (arg_match(&arg, &buf_optimal_sz, argi))
1982 config->cfg.rc_buf_optimal_sz = arg_parse_uint(&arg);
1983 else if (arg_match(&arg, &bias_pct, argi)) {
1984 config->cfg.rc_2pass_vbr_bias_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, &minsection_pct, argi)) {
1989 config->cfg.rc_2pass_vbr_minsection_pct = arg_parse_uint(&arg);
1990
1991 if (global->passes < 2)
1992 warn("option %s ignored in one-pass mode.\n", arg.name);
1993 } else if (arg_match(&arg, &maxsection_pct, argi)) {
1994 config->cfg.rc_2pass_vbr_maxsection_pct = arg_parse_uint(&arg);
1995
1996 if (global->passes < 2)
1997 warn("option %s ignored in one-pass mode.\n", arg.name);
1998 } else if (arg_match(&arg, &kf_min_dist, argi))
1999 config->cfg.kf_min_dist = arg_parse_uint(&arg);
2000 else if (arg_match(&arg, &kf_max_dist, argi)) {
2001 config->cfg.kf_max_dist = arg_parse_uint(&arg);
2002 config->have_kf_max_dist = 1;
2003 } else if (arg_match(&arg, &kf_disabled, argi))
2004 config->cfg.kf_mode = VPX_KF_DISABLED;
2005 else {
2006 int i, match = 0;
2007
2008 for (i = 0; ctrl_args[i]; i++) {
2009 if (arg_match(&arg, ctrl_args[i], argi)) {
2010 int j;
2011 match = 1;
2012
2013 /* Point either to the next free element or the first
2014 * instance of this control.
2015 */
2016 for (j = 0; j < config->arg_ctrl_cnt; j++)
2017 if (config->arg_ctrls[j][0] == ctrl_args_map[i])
2018 break;
2019
2020 /* Update/insert */
2021 assert(j < ARG_CTRL_CNT_MAX);
2022 if (j < ARG_CTRL_CNT_MAX) {
2023 config->arg_ctrls[j][0] = ctrl_args_map[i];
2024 config->arg_ctrls[j][1] = arg_parse_enum_or_int(&arg);
2025 if (j == config->arg_ctrl_cnt)
2026 config->arg_ctrl_cnt++;
2027 }
2028
2029 }
2030 }
2031
2032 if (!match)
2033 argj++;
2034 }
2035 }
2036
2037 return eos_mark_found;
2038 }
2039
2040
2041 #define FOREACH_STREAM(func)\
2042 do\
2043 {\
2044 struct stream_state *stream;\
2045 \
2046 for(stream = streams; stream; stream = stream->next)\
2047 func;\
2048 }while(0)
2049
2050
2051 static void validate_stream_config(struct stream_state *stream) {
2052 struct stream_state *streami;
2053
2054 if (!stream->config.cfg.g_w || !stream->config.cfg.g_h)
2055 fatal("Stream %d: Specify stream dimensions with --width (-w) "
2056 " and --height (-h)", stream->index);
2057
2058 for (streami = stream; streami; streami = streami->next) {
2059 /* All streams require output files */
2060 if (!streami->config.out_fn)
2061 fatal("Stream %d: Output file is required (specify with -o)",
2062 streami->index);
2063
2064 /* Check for two streams outputting to the same file */
2065 if (streami != stream) {
2066 const char *a = stream->config.out_fn;
2067 const char *b = streami->config.out_fn;
2068 if (!strcmp(a, b) && strcmp(a, "/dev/null") && strcmp(a, ":nul"))
2069 fatal("Stream %d: duplicate output file (from stream %d)",
2070 streami->index, stream->index);
2071 }
2072
2073 /* Check for two streams sharing a stats file. */
2074 if (streami != stream) {
2075 const char *a = stream->config.stats_fn;
2076 const char *b = streami->config.stats_fn;
2077 if (a && b && !strcmp(a, b))
2078 fatal("Stream %d: duplicate stats file (from stream %d)",
2079 streami->index, stream->index);
2080 }
2081 }
2082 }
2083
2084
2085 static void set_stream_dimensions(struct stream_state *stream,
2086 unsigned int w,
2087 unsigned int h) {
2088 if (!stream->config.cfg.g_w) {
2089 if (!stream->config.cfg.g_h)
2090 stream->config.cfg.g_w = w;
2091 else
2092 stream->config.cfg.g_w = w * stream->config.cfg.g_h / h;
2093 }
2094 if (!stream->config.cfg.g_h) {
2095 stream->config.cfg.g_h = h * stream->config.cfg.g_w / w;
2096 }
2097 }
2098
2099
2100 static void set_default_kf_interval(struct stream_state *stream,
2101 struct global_config *global) {
2102 /* Use a max keyframe interval of 5 seconds, if none was
2103 * specified on the command line.
2104 */
2105 if (!stream->config.have_kf_max_dist) {
2106 double framerate = (double)global->framerate.num / global->framerate.den;
2107 if (framerate > 0.0)
2108 stream->config.cfg.kf_max_dist = (unsigned int)(5.0 * framerate);
2109 }
2110 }
2111
2112
2113 static void show_stream_config(struct stream_state *stream,
2114 struct global_config *global,
2115 struct input_state *input) {
2116
2117 #define SHOW(field) \
2118 fprintf(stderr, " %-28s = %d\n", #field, stream->config.cfg.field)
2119
2120 if (stream->index == 0) {
2121 fprintf(stderr, "Codec: %s\n",
2122 vpx_codec_iface_name(global->codec->iface()));
2123 fprintf(stderr, "Source file: %s Format: %s\n", input->fn,
2124 input->use_i420 ? "I420" : "YV12");
2125 }
2126 if (stream->next || stream->index)
2127 fprintf(stderr, "\nStream Index: %d\n", stream->index);
2128 fprintf(stderr, "Destination file: %s\n", stream->config.out_fn);
2129 fprintf(stderr, "Encoder parameters:\n");
2130
2131 SHOW(g_usage);
2132 SHOW(g_threads);
2133 SHOW(g_profile);
2134 SHOW(g_w);
2135 SHOW(g_h);
2136 SHOW(g_timebase.num);
2137 SHOW(g_timebase.den);
2138 SHOW(g_error_resilient);
2139 SHOW(g_pass);
2140 SHOW(g_lag_in_frames);
2141 SHOW(rc_dropframe_thresh);
2142 SHOW(rc_resize_allowed);
2143 SHOW(rc_resize_up_thresh);
2144 SHOW(rc_resize_down_thresh);
2145 SHOW(rc_end_usage);
2146 SHOW(rc_target_bitrate);
2147 SHOW(rc_min_quantizer);
2148 SHOW(rc_max_quantizer);
2149 SHOW(rc_undershoot_pct);
2150 SHOW(rc_overshoot_pct);
2151 SHOW(rc_buf_sz);
2152 SHOW(rc_buf_initial_sz);
2153 SHOW(rc_buf_optimal_sz);
2154 SHOW(rc_2pass_vbr_bias_pct);
2155 SHOW(rc_2pass_vbr_minsection_pct);
2156 SHOW(rc_2pass_vbr_maxsection_pct);
2157 SHOW(kf_mode);
2158 SHOW(kf_min_dist);
2159 SHOW(kf_max_dist);
2160 }
2161
2162
2163 static void open_output_file(struct stream_state *stream,
2164 struct global_config *global) {
2165 const char *fn = stream->config.out_fn;
2166
2167 stream->file = strcmp(fn, "-") ? fopen(fn, "wb") : set_binary_mode(stdout);
2168
2169 if (!stream->file)
2170 fatal("Failed to open output file");
2171
2172 if (stream->config.write_webm && fseek(stream->file, 0, SEEK_CUR))
2173 fatal("WebM output to pipes not supported.");
2174
2175 if (stream->config.write_webm) {
2176 stream->ebml.stream = stream->file;
2177 write_webm_file_header(&stream->ebml, &stream->config.cfg,
2178 &global->framerate,
2179 stream->config.stereo_fmt,
2180 global->codec->fourcc);
2181 } else
2182 write_ivf_file_header(stream->file, &stream->config.cfg,
2183 global->codec->fourcc, 0);
2184 }
2185
2186
2187 static void close_output_file(struct stream_state *stream,
2188 unsigned int fourcc) {
2189 if (stream->config.write_webm) {
2190 write_webm_file_footer(&stream->ebml, stream->hash);
2191 free(stream->ebml.cue_list);
2192 stream->ebml.cue_list = NULL;
2193 } else {
2194 if (!fseek(stream->file, 0, SEEK_SET))
2195 write_ivf_file_header(stream->file, &stream->config.cfg,
2196 fourcc,
2197 stream->frames_out);
2198 }
2199
2200 fclose(stream->file);
2201 }
2202
2203
2204 static void setup_pass(struct stream_state *stream,
2205 struct global_config *global,
2206 int pass) {
2207 if (stream->config.stats_fn) {
2208 if (!stats_open_file(&stream->stats, stream->config.stats_fn,
2209 pass))
2210 fatal("Failed to open statistics store");
2211 } else {
2212 if (!stats_open_mem(&stream->stats, pass))
2213 fatal("Failed to open statistics store");
2214 }
2215
2216 stream->config.cfg.g_pass = global->passes == 2
2217 ? pass ? VPX_RC_LAST_PASS : VPX_RC_FIRST_PASS
2218 : VPX_RC_ONE_PASS;
2219 if (pass)
2220 stream->config.cfg.rc_twopass_stats_in = stats_get(&stream->stats);
2221
2222 stream->cx_time = 0;
2223 stream->nbytes = 0;
2224 stream->frames_out = 0;
2225 }
2226
2227
2228 static void initialize_encoder(struct stream_state *stream,
2229 struct global_config *global) {
2230 int i;
2231 int flags = 0;
2232
2233 flags |= global->show_psnr ? VPX_CODEC_USE_PSNR : 0;
2234 flags |= global->out_part ? VPX_CODEC_USE_OUTPUT_PARTITION : 0;
2235
2236 /* Construct Encoder Context */
2237 vpx_codec_enc_init(&stream->encoder, global->codec->iface(),
2238 &stream->config.cfg, flags);
2239 ctx_exit_on_error(&stream->encoder, "Failed to initialize encoder");
2240
2241 /* Note that we bypass the vpx_codec_control wrapper macro because
2242 * we're being clever to store the control IDs in an array. Real
2243 * applications will want to make use of the enumerations directly
2244 */
2245 for (i = 0; i < stream->config.arg_ctrl_cnt; i++) {
2246 int ctrl = stream->config.arg_ctrls[i][0];
2247 int value = stream->config.arg_ctrls[i][1];
2248 if (vpx_codec_control_(&stream->encoder, ctrl, value))
2249 fprintf(stderr, "Error: Tried to set control %d = %d\n",
2250 ctrl, value);
2251
2252 ctx_exit_on_error(&stream->encoder, "Failed to control codec");
2253 }
2254
2255 #if CONFIG_DECODERS
2256 if (global->test_decode != TEST_DECODE_OFF) {
2257 vpx_codec_dec_init(&stream->decoder, global->codec->dx_iface(), NULL, 0);
2258 }
2259 #endif
2260 }
2261
2262
2263 static void encode_frame(struct stream_state *stream,
2264 struct global_config *global,
2265 struct vpx_image *img,
2266 unsigned int frames_in) {
2267 vpx_codec_pts_t frame_start, next_frame_start;
2268 struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
2269 struct vpx_usec_timer timer;
2270
2271 frame_start = (cfg->g_timebase.den * (int64_t)(frames_in - 1)
2272 * global->framerate.den)
2273 / cfg->g_timebase.num / global->framerate.num;
2274 next_frame_start = (cfg->g_timebase.den * (int64_t)(frames_in)
2275 * global->framerate.den)
2276 / cfg->g_timebase.num / global->framerate.num;
2277
2278 /* Scale if necessary */
2279 if (img && (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
2280 if (!stream->img)
2281 stream->img = vpx_img_alloc(NULL, VPX_IMG_FMT_I420,
2282 cfg->g_w, cfg->g_h, 16);
2283 I420Scale(img->planes[VPX_PLANE_Y], img->stride[VPX_PLANE_Y],
2284 img->planes[VPX_PLANE_U], img->stride[VPX_PLANE_U],
2285 img->planes[VPX_PLANE_V], img->stride[VPX_PLANE_V],
2286 img->d_w, img->d_h,
2287 stream->img->planes[VPX_PLANE_Y],
2288 stream->img->stride[VPX_PLANE_Y],
2289 stream->img->planes[VPX_PLANE_U],
2290 stream->img->stride[VPX_PLANE_U],
2291 stream->img->planes[VPX_PLANE_V],
2292 stream->img->stride[VPX_PLANE_V],
2293 stream->img->d_w, stream->img->d_h,
2294 kFilterBox);
2295
2296 img = stream->img;
2297 }
2298
2299 vpx_usec_timer_start(&timer);
2300 vpx_codec_encode(&stream->encoder, img, frame_start,
2301 (unsigned long)(next_frame_start - frame_start),
2302 0, global->deadline);
2303 vpx_usec_timer_mark(&timer);
2304 stream->cx_time += vpx_usec_timer_elapsed(&timer);
2305 ctx_exit_on_error(&stream->encoder, "Stream %d: Failed to encode frame",
2306 stream->index);
2307 }
2308
2309
2310 static void update_quantizer_histogram(struct stream_state *stream) {
2311 if (stream->config.cfg.g_pass != VPX_RC_FIRST_PASS) {
2312 int q;
2313
2314 vpx_codec_control(&stream->encoder, VP8E_GET_LAST_QUANTIZER_64, &q);
2315 ctx_exit_on_error(&stream->encoder, "Failed to read quantizer");
2316 stream->counts[q]++;
2317 }
2318 }
2319
2320
2321 static void get_cx_data(struct stream_state *stream,
2322 struct global_config *global,
2323 int *got_data) {
2324 const vpx_codec_cx_pkt_t *pkt;
2325 const struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
2326 vpx_codec_iter_t iter = NULL;
2327
2328 *got_data = 0;
2329 while ((pkt = vpx_codec_get_cx_data(&stream->encoder, &iter))) {
2330 static size_t fsize = 0;
2331 static off_t ivf_header_pos = 0;
2332
2333 switch (pkt->kind) {
2334 case VPX_CODEC_CX_FRAME_PKT:
2335 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
2336 stream->frames_out++;
2337 }
2338 if (!global->quiet)
2339 fprintf(stderr, " %6luF", (unsigned long)pkt->data.frame.sz);
2340
2341 update_rate_histogram(&stream->rate_hist, cfg, pkt);
2342 if (stream->config.write_webm) {
2343 /* Update the hash */
2344 if (!stream->ebml.debug)
2345 stream->hash = murmur(pkt->data.frame.buf,
2346 (int)pkt->data.frame.sz,
2347 stream->hash);
2348
2349 write_webm_block(&stream->ebml, cfg, pkt);
2350 } else {
2351 if (pkt->data.frame.partition_id <= 0) {
2352 ivf_header_pos = ftello(stream->file);
2353 fsize = pkt->data.frame.sz;
2354
2355 write_ivf_frame_header(stream->file, pkt);
2356 } else {
2357 fsize += pkt->data.frame.sz;
2358
2359 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
2360 off_t currpos = ftello(stream->file);
2361 fseeko(stream->file, ivf_header_pos, SEEK_SET);
2362 write_ivf_frame_size(stream->file, fsize);
2363 fseeko(stream->file, currpos, SEEK_SET);
2364 }
2365 }
2366
2367 (void) fwrite(pkt->data.frame.buf, 1, pkt->data.frame.sz,
2368 stream->file);
2369 }
2370 stream->nbytes += pkt->data.raw.sz;
2371
2372 *got_data = 1;
2373 #if CONFIG_DECODERS
2374 if (global->test_decode != TEST_DECODE_OFF && !stream->mismatch_seen) {
2375 vpx_codec_decode(&stream->decoder, pkt->data.frame.buf,
2376 pkt->data.frame.sz, NULL, 0);
2377 if (stream->decoder.err) {
2378 warn_or_exit_on_error(&stream->decoder,
2379 global->test_decode == TEST_DECODE_FATAL,
2380 "Failed to decode frame %d in stream %d",
2381 stream->frames_out + 1, stream->index);
2382 stream->mismatch_seen = stream->frames_out + 1;
2383 }
2384 }
2385 #endif
2386 break;
2387 case VPX_CODEC_STATS_PKT:
2388 stream->frames_out++;
2389 stats_write(&stream->stats,
2390 pkt->data.twopass_stats.buf,
2391 pkt->data.twopass_stats.sz);
2392 stream->nbytes += pkt->data.raw.sz;
2393 break;
2394 case VPX_CODEC_PSNR_PKT:
2395
2396 if (global->show_psnr) {
2397 int i;
2398
2399 stream->psnr_sse_total += pkt->data.psnr.sse[0];
2400 stream->psnr_samples_total += pkt->data.psnr.samples[0];
2401 for (i = 0; i < 4; i++) {
2402 if (!global->quiet)
2403 fprintf(stderr, "%.3f ", pkt->data.psnr.psnr[i]);
2404 stream->psnr_totals[i] += pkt->data.psnr.psnr[i];
2405 }
2406 stream->psnr_count++;
2407 }
2408
2409 break;
2410 default:
2411 break;
2412 }
2413 }
2414 }
2415
2416
2417 static void show_psnr(struct stream_state *stream) {
2418 int i;
2419 double ovpsnr;
2420
2421 if (!stream->psnr_count)
2422 return;
2423
2424 fprintf(stderr, "Stream %d PSNR (Overall/Avg/Y/U/V)", stream->index);
2425 ovpsnr = vp8_mse2psnr((double)stream->psnr_samples_total, 255.0,
2426 (double)stream->psnr_sse_total);
2427 fprintf(stderr, " %.3f", ovpsnr);
2428
2429 for (i = 0; i < 4; i++) {
2430 fprintf(stderr, " %.3f", stream->psnr_totals[i] / stream->psnr_count);
2431 }
2432 fprintf(stderr, "\n");
2433 }
2434
2435
2436 static float usec_to_fps(uint64_t usec, unsigned int frames) {
2437 return (float)(usec > 0 ? frames * 1000000.0 / (float)usec : 0);
2438 }
2439
2440
2441 static void test_decode(struct stream_state *stream,
2442 enum TestDecodeFatality fatal,
2443 const struct codec_item *codec) {
2444 vpx_image_t enc_img, dec_img;
2445
2446 if (stream->mismatch_seen)
2447 return;
2448
2449 /* Get the internal reference frame */
2450 if (codec->fourcc == VP8_FOURCC) {
2451 struct vpx_ref_frame ref_enc, ref_dec;
2452 int width, height;
2453
2454 width = (stream->config.cfg.g_w + 15) & ~15;
2455 height = (stream->config.cfg.g_h + 15) & ~15;
2456 vpx_img_alloc(&ref_enc.img, VPX_IMG_FMT_I420, width, height, 1);
2457 enc_img = ref_enc.img;
2458 vpx_img_alloc(&ref_dec.img, VPX_IMG_FMT_I420, width, height, 1);
2459 dec_img = ref_dec.img;
2460
2461 ref_enc.frame_type = VP8_LAST_FRAME;
2462 ref_dec.frame_type = VP8_LAST_FRAME;
2463 vpx_codec_control(&stream->encoder, VP8_COPY_REFERENCE, &ref_enc);
2464 vpx_codec_control(&stream->decoder, VP8_COPY_REFERENCE, &ref_dec);
2465 } else {
2466 struct vp9_ref_frame ref;
2467
2468 ref.idx = 0;
2469 vpx_codec_control(&stream->encoder, VP9_GET_REFERENCE, &ref);
2470 enc_img = ref.img;
2471 vpx_codec_control(&stream->decoder, VP9_GET_REFERENCE, &ref);
2472 dec_img = ref.img;
2473 }
2474 ctx_exit_on_error(&stream->encoder, "Failed to get encoder reference frame");
2475 ctx_exit_on_error(&stream->decoder, "Failed to get decoder reference frame");
2476
2477 if (!compare_img(&enc_img, &dec_img)) {
2478 int y[4], u[4], v[4];
2479 find_mismatch(&enc_img, &dec_img, y, u, v);
2480 stream->decoder.err = 1;
2481 warn_or_exit_on_error(&stream->decoder, fatal == TEST_DECODE_FATAL,
2482 "Stream %d: Encode/decode mismatch on frame %d at"
2483 " Y[%d, %d] {%d/%d},"
2484 " U[%d, %d] {%d/%d},"
2485 " V[%d, %d] {%d/%d}",
2486 stream->index, stream->frames_out,
2487 y[0], y[1], y[2], y[3],
2488 u[0], u[1], u[2], u[3],
2489 v[0], v[1], v[2], v[3]);
2490 stream->mismatch_seen = stream->frames_out;
2491 }
2492
2493 vpx_img_free(&enc_img);
2494 vpx_img_free(&dec_img);
2495 }
2496
2497
2498 static void print_time(const char *label, int64_t etl) {
2499 int hours, mins, secs;
2500
2501 if (etl >= 0) {
2502 hours = etl / 3600;
2503 etl -= hours * 3600;
2504 mins = etl / 60;
2505 etl -= mins * 60;
2506 secs = etl;
2507
2508 fprintf(stderr, "[%3s %2d:%02d:%02d] ",
2509 label, hours, mins, secs);
2510 } else {
2511 fprintf(stderr, "[%3s unknown] ", label);
2512 }
2513 }
2514
2515 int main(int argc, const char **argv_) {
2516 int pass;
2517 vpx_image_t raw;
2518 int frame_avail, got_data;
2519
2520 struct input_state input = {0};
2521 struct global_config global;
2522 struct stream_state *streams = NULL;
2523 char **argv, **argi;
2524 uint64_t cx_time = 0;
2525 int stream_cnt = 0;
2526 int res = 0;
2527
2528 exec_name = argv_[0];
2529
2530 if (argc < 3)
2531 usage_exit();
2532
2533 /* Setup default input stream settings */
2534 input.framerate.num = 30;
2535 input.framerate.den = 1;
2536 input.use_i420 = 1;
2537 input.only_i420 = 1;
2538
2539 /* First parse the global configuration values, because we want to apply
2540 * other parameters on top of the default configuration provided by the
2541 * codec.
2542 */
2543 argv = argv_dup(argc - 1, argv_ + 1);
2544 parse_global_config(&global, argv);
2545
2546 {
2547 /* Now parse each stream's parameters. Using a local scope here
2548 * due to the use of 'stream' as loop variable in FOREACH_STREAM
2549 * loops
2550 */
2551 struct stream_state *stream = NULL;
2552
2553 do {
2554 stream = new_stream(&global, stream);
2555 stream_cnt++;
2556 if (!streams)
2557 streams = stream;
2558 } while (parse_stream_params(&global, stream, argv));
2559 }
2560
2561 /* Check for unrecognized options */
2562 for (argi = argv; *argi; argi++)
2563 if (argi[0][0] == '-' && argi[0][1])
2564 die("Error: Unrecognized option %s\n", *argi);
2565
2566 /* Handle non-option arguments */
2567 input.fn = argv[0];
2568
2569 if (!input.fn)
2570 usage_exit();
2571
2572 #if CONFIG_NON420
2573 /* Decide if other chroma subsamplings than 4:2:0 are supported */
2574 if (global.codec->fourcc == VP9_FOURCC)
2575 input.only_i420 = 0;
2576 #endif
2577
2578 for (pass = global.pass ? global.pass - 1 : 0; pass < global.passes; pass++) {
2579 int frames_in = 0, seen_frames = 0;
2580 int64_t estimated_time_left = -1;
2581 int64_t average_rate = -1;
2582 off_t lagged_count = 0;
2583
2584 open_input_file(&input);
2585
2586 /* If the input file doesn't specify its w/h (raw files), try to get
2587 * the data from the first stream's configuration.
2588 */
2589 if (!input.w || !input.h)
2590 FOREACH_STREAM( {
2591 if (stream->config.cfg.g_w && stream->config.cfg.g_h) {
2592 input.w = stream->config.cfg.g_w;
2593 input.h = stream->config.cfg.g_h;
2594 break;
2595 }
2596 });
2597
2598 /* Update stream configurations from the input file's parameters */
2599 if (!input.w || !input.h)
2600 fatal("Specify stream dimensions with --width (-w) "
2601 " and --height (-h)");
2602 FOREACH_STREAM(set_stream_dimensions(stream, input.w, input.h));
2603 FOREACH_STREAM(validate_stream_config(stream));
2604
2605 /* Ensure that --passes and --pass are consistent. If --pass is set and
2606 * --passes=2, ensure --fpf was set.
2607 */
2608 if (global.pass && global.passes == 2)
2609 FOREACH_STREAM( {
2610 if (!stream->config.stats_fn)
2611 die("Stream %d: Must specify --fpf when --pass=%d"
2612 " and --passes=2\n", stream->index, global.pass);
2613 });
2614
2615 /* Use the frame rate from the file only if none was specified
2616 * on the command-line.
2617 */
2618 if (!global.have_framerate)
2619 global.framerate = input.framerate;
2620
2621 FOREACH_STREAM(set_default_kf_interval(stream, &global));
2622
2623 /* Show configuration */
2624 if (global.verbose && pass == 0)
2625 FOREACH_STREAM(show_stream_config(stream, &global, &input));
2626
2627 if (pass == (global.pass ? global.pass - 1 : 0)) {
2628 if (input.file_type == FILE_TYPE_Y4M)
2629 /*The Y4M reader does its own allocation.
2630 Just initialize this here to avoid problems if we never read any
2631 frames.*/
2632 memset(&raw, 0, sizeof(raw));
2633 else
2634 vpx_img_alloc(&raw,
2635 input.use_i420 ? VPX_IMG_FMT_I420
2636 : VPX_IMG_FMT_YV12,
2637 input.w, input.h, 32);
2638
2639 FOREACH_STREAM(init_rate_histogram(&stream->rate_hist,
2640 &stream->config.cfg,
2641 &global.framerate));
2642 }
2643
2644 FOREACH_STREAM(setup_pass(stream, &global, pass));
2645 FOREACH_STREAM(open_output_file(stream, &global));
2646 FOREACH_STREAM(initialize_encoder(stream, &global));
2647
2648 frame_avail = 1;
2649 got_data = 0;
2650
2651 while (frame_avail || got_data) {
2652 struct vpx_usec_timer timer;
2653
2654 if (!global.limit || frames_in < global.limit) {
2655 frame_avail = read_frame(&input, &raw);
2656
2657 if (frame_avail)
2658 frames_in++;
2659 seen_frames = frames_in > global.skip_frames ?
2660 frames_in - global.skip_frames : 0;
2661
2662 if (!global.quiet) {
2663 float fps = usec_to_fps(cx_time, seen_frames);
2664 fprintf(stderr, "\rPass %d/%d ", pass + 1, global.passes);
2665
2666 if (stream_cnt == 1)
2667 fprintf(stderr,
2668 "frame %4d/%-4d %7"PRId64"B ",
2669 frames_in, streams->frames_out, (int64_t)streams->nbytes);
2670 else
2671 fprintf(stderr, "frame %4d ", frames_in);
2672
2673 fprintf(stderr, "%7"PRId64" %s %.2f %s ",
2674 cx_time > 9999999 ? cx_time / 1000 : cx_time,
2675 cx_time > 9999999 ? "ms" : "us",
2676 fps >= 1.0 ? fps : fps * 60,
2677 fps >= 1.0 ? "fps" : "fpm");
2678 print_time("ETA", estimated_time_left);
2679 fprintf(stderr, "\033[K");
2680 }
2681
2682 } else
2683 frame_avail = 0;
2684
2685 if (frames_in > global.skip_frames) {
2686 vpx_usec_timer_start(&timer);
2687 FOREACH_STREAM(encode_frame(stream, &global,
2688 frame_avail ? &raw : NULL,
2689 frames_in));
2690 vpx_usec_timer_mark(&timer);
2691 cx_time += vpx_usec_timer_elapsed(&timer);
2692
2693 FOREACH_STREAM(update_quantizer_histogram(stream));
2694
2695 got_data = 0;
2696 FOREACH_STREAM(get_cx_data(stream, &global, &got_data));
2697
2698 if (!got_data && input.length && !streams->frames_out) {
2699 lagged_count = global.limit ? seen_frames : ftello(input.file);
2700 } else if (input.length) {
2701 int64_t remaining;
2702 int64_t rate;
2703
2704 if (global.limit) {
2705 int frame_in_lagged = (seen_frames - lagged_count) * 1000;
2706
2707 rate = cx_time ? frame_in_lagged * (int64_t)1000000 / cx_time : 0;
2708 remaining = 1000 * (global.limit - global.skip_frames
2709 - seen_frames + lagged_count);
2710 } else {
2711 off_t input_pos = ftello(input.file);
2712 off_t input_pos_lagged = input_pos - lagged_count;
2713 int64_t limit = input.length;
2714
2715 rate = cx_time ? input_pos_lagged * (int64_t)1000000 / cx_time : 0;
2716 remaining = limit - input_pos + lagged_count;
2717 }
2718
2719 average_rate = (average_rate <= 0)
2720 ? rate
2721 : (average_rate * 7 + rate) / 8;
2722 estimated_time_left = average_rate ? remaining / average_rate : -1;
2723 }
2724
2725 if (got_data && global.test_decode != TEST_DECODE_OFF)
2726 FOREACH_STREAM(test_decode(stream, global.test_decode, global.codec));
2727 }
2728
2729 fflush(stdout);
2730 }
2731
2732 if (stream_cnt > 1)
2733 fprintf(stderr, "\n");
2734
2735 if (!global.quiet)
2736 FOREACH_STREAM(fprintf(
2737 stderr,
2738 "\rPass %d/%d frame %4d/%-4d %7"PRId64"B %7lub/f %7"PRId64"b/s"
2739 " %7"PRId64" %s (%.2f fps)\033[K\n", pass + 1,
2740 global.passes, frames_in, stream->frames_out, (int64_t)stream->nbytes,
2741 seen_frames ? (unsigned long)(stream->nbytes * 8 / seen_frames) : 0,
2742 seen_frames ? (int64_t)stream->nbytes * 8
2743 * (int64_t)global.framerate.num / global.framerate.den
2744 / seen_frames
2745 : 0,
2746 stream->cx_time > 9999999 ? stream->cx_time / 1000 : stream->cx_time,
2747 stream->cx_time > 9999999 ? "ms" : "us",
2748 usec_to_fps(stream->cx_time, seen_frames));
2749 );
2750
2751 if (global.show_psnr)
2752 FOREACH_STREAM(show_psnr(stream));
2753
2754 FOREACH_STREAM(vpx_codec_destroy(&stream->encoder));
2755
2756 if (global.test_decode != TEST_DECODE_OFF) {
2757 FOREACH_STREAM(vpx_codec_destroy(&stream->decoder));
2758 }
2759
2760 close_input_file(&input);
2761
2762 if (global.test_decode == TEST_DECODE_FATAL) {
2763 FOREACH_STREAM(res |= stream->mismatch_seen);
2764 }
2765 FOREACH_STREAM(close_output_file(stream, global.codec->fourcc));
2766
2767 FOREACH_STREAM(stats_close(&stream->stats, global.passes - 1));
2768
2769 if (global.pass)
2770 break;
2771 }
2772
2773 if (global.show_q_hist_buckets)
2774 FOREACH_STREAM(show_q_histogram(stream->counts,
2775 global.show_q_hist_buckets));
2776
2777 if (global.show_rate_hist_buckets)
2778 FOREACH_STREAM(show_rate_histogram(&stream->rate_hist,
2779 &stream->config.cfg,
2780 global.show_rate_hist_buckets));
2781 FOREACH_STREAM(destroy_rate_histogram(&stream->rate_hist));
2782
2783 #if CONFIG_INTERNAL_STATS
2784 /* TODO(jkoleszar): This doesn't belong in this executable. Do it for now,
2785 * to match some existing utilities.
2786 */
2787 FOREACH_STREAM({
2788 FILE *f = fopen("opsnr.stt", "a");
2789 if (stream->mismatch_seen) {
2790 fprintf(f, "First mismatch occurred in frame %d\n",
2791 stream->mismatch_seen);
2792 } else {
2793 fprintf(f, "No mismatch detected in recon buffers\n");
2794 }
2795 fclose(f);
2796 });
2797 #endif
2798
2799 vpx_img_free(&raw);
2800 free(argv);
2801 free(streams);
2802 return res ? EXIT_FAILURE : EXIT_SUCCESS;
2803 }
AV1 Codec Library