vpxenc.c 77.3 KB
Newer Older
John Koleszar's avatar
John Koleszar committed
1
/*
2
 *  Copyright (c) 2010 The WebM project authors. All Rights Reserved.
John Koleszar's avatar
John Koleszar committed
3
 *
4
 *  Use of this source code is governed by a BSD-style license
5
6
 *  that can be found in the LICENSE file in the root of the source
 *  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.
John Koleszar's avatar
John Koleszar committed
9
10
11
12
13
14
 */


/* This is a simple program that encodes YV12 files and generates ivf
 * files using the new interface.
 */
15
#if defined(_WIN32) || !CONFIG_OS_SUPPORT
16
17
18
19
#define USE_POSIX_MMAP 0
#else
#define USE_POSIX_MMAP 1
#endif
John Koleszar's avatar
John Koleszar committed
20
21
22
23
24

#include <stdio.h>
#include <stdlib.h>
#include <stdarg.h>
#include <string.h>
John Koleszar's avatar
John Koleszar committed
25
#include <limits.h>
26
#include <assert.h>
27
#include "vpx/vpx_encoder.h"
John Koleszar's avatar
John Koleszar committed
28
29
30
31
32
33
34
#if USE_POSIX_MMAP
#include <sys/types.h>
#include <sys/stat.h>
#include <sys/mman.h>
#include <fcntl.h>
#include <unistd.h>
#endif
35
#include "vpx/vp8cx.h"
John Koleszar's avatar
John Koleszar committed
36
37
#include "vpx_ports/mem_ops.h"
#include "vpx_ports/vpx_timer.h"
John Koleszar's avatar
John Koleszar committed
38
#include "tools_common.h"
39
#include "y4minput.h"
John Koleszar's avatar
John Koleszar committed
40
41
42
43
44
45
46
47
48
49
#include "libmkv/EbmlWriter.h"
#include "libmkv/EbmlIDs.h"

/* Need special handling of these functions on Windows */
#if defined(_MSC_VER)
/* MSVS doesn't define off_t, and uses _f{seek,tell}i64 */
typedef __int64 off_t;
#define fseeko _fseeki64
#define ftello _ftelli64
#elif defined(_WIN32)
James Zern's avatar
James Zern committed
50
51
/* MinGW defines off_t as long
   and uses f{seek,tell}o64/off64_t for large files */
John Koleszar's avatar
John Koleszar committed
52
53
#define fseeko fseeko64
#define ftello ftello64
James Zern's avatar
James Zern committed
54
#define off_t off64_t
John Koleszar's avatar
John Koleszar committed
55
56
#endif

John Koleszar's avatar
John Koleszar committed
57
#define LITERALU64(hi,lo) ((((uint64_t)hi)<<32)|lo)
John Koleszar's avatar
John Koleszar committed
58

59
60
61
62
63
64
65
66
/* We should use 32-bit file operations in WebM file format
 * when building ARM executable file (.axf) with RVCT */
#if !CONFIG_OS_SUPPORT
typedef long off_t;
#define fseeko fseek
#define ftello ftell
#endif

John Koleszar's avatar
John Koleszar committed
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
/* Swallow warnings about unused results of fread/fwrite */
static size_t wrap_fread(void *ptr, size_t size, size_t nmemb,
                         FILE *stream)
{
    return fread(ptr, size, nmemb, stream);
}
#define fread wrap_fread

static size_t wrap_fwrite(const void *ptr, size_t size, size_t nmemb,
                          FILE *stream)
{
    return fwrite(ptr, size, nmemb, stream);
}
#define fwrite wrap_fwrite


John Koleszar's avatar
John Koleszar committed
83
84
85
86
87
static const char *exec_name;

static const struct codec_item
{
    char const              *name;
John Koleszar's avatar
John Koleszar committed
88
    vpx_codec_iface_t       *iface;
John Koleszar's avatar
John Koleszar committed
89
90
91
92
93
94
95
96
97
98
    unsigned int             fourcc;
} codecs[] =
{
#if CONFIG_VP8_ENCODER
    {"vp8",  &vpx_codec_vp8_cx_algo, 0x30385056},
#endif
};

static void usage_exit();

99
100
101
102
103
104
105
106
107
108
109
110
#define LOG_ERROR(label) do \
{\
    const char *l=label;\
    va_list ap;\
    va_start(ap, fmt);\
    if(l)\
        fprintf(stderr, "%s: ", l);\
    vfprintf(stderr, fmt, ap);\
    fprintf(stderr, "\n");\
    va_end(ap);\
} while(0)

John Koleszar's avatar
John Koleszar committed
111
112
void die(const char *fmt, ...)
{
113
    LOG_ERROR(NULL);
John Koleszar's avatar
John Koleszar committed
114
115
116
    usage_exit();
}

117
118
119
120
121
122
123
124
125
126
127
128
129
130

void fatal(const char *fmt, ...)
{
    LOG_ERROR("Fatal");
    exit(EXIT_FAILURE);
}


void warn(const char *fmt, ...)
{
    LOG_ERROR("Warning");
}


131
static void ctx_exit_on_error(vpx_codec_ctx_t *ctx, const char *s, ...)
John Koleszar's avatar
John Koleszar committed
132
{
133
134
135
    va_list ap;

    va_start(ap, s);
John Koleszar's avatar
John Koleszar committed
136
137
138
139
    if (ctx->err)
    {
        const char *detail = vpx_codec_error_detail(ctx);

140
141
        vfprintf(stderr, s, ap);
        fprintf(stderr, ": %s\n", vpx_codec_error(ctx));
John Koleszar's avatar
John Koleszar committed
142
143

        if (detail)
144
            fprintf(stderr, "    %s\n", detail);
John Koleszar's avatar
John Koleszar committed
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194

        exit(EXIT_FAILURE);
    }
}

/* This structure is used to abstract the different ways of handling
 * first pass statistics.
 */
typedef struct
{
    vpx_fixed_buf_t buf;
    int             pass;
    FILE           *file;
    char           *buf_ptr;
    size_t          buf_alloc_sz;
} stats_io_t;

int stats_open_file(stats_io_t *stats, const char *fpf, int pass)
{
    int res;

    stats->pass = pass;

    if (pass == 0)
    {
        stats->file = fopen(fpf, "wb");
        stats->buf.sz = 0;
        stats->buf.buf = NULL,
                   res = (stats->file != NULL);
    }
    else
    {
#if 0
#elif USE_POSIX_MMAP
        struct stat stat_buf;
        int fd;

        fd = open(fpf, O_RDONLY);
        stats->file = fdopen(fd, "rb");
        fstat(fd, &stat_buf);
        stats->buf.sz = stat_buf.st_size;
        stats->buf.buf = mmap(NULL, stats->buf.sz, PROT_READ, MAP_PRIVATE,
                              fd, 0);
        res = (stats->buf.buf != NULL);
#else
        size_t nbytes;

        stats->file = fopen(fpf, "rb");

        if (fseek(stats->file, 0, SEEK_END))
195
            fatal("First-pass stats file must be seekable!");
John Koleszar's avatar
John Koleszar committed
196
197
198
199
200
201
202

        stats->buf.sz = stats->buf_alloc_sz = ftell(stats->file);
        rewind(stats->file);

        stats->buf.buf = malloc(stats->buf_alloc_sz);

        if (!stats->buf.buf)
203
204
            fatal("Failed to allocate first-pass stats buffer (%lu bytes)",
                  (unsigned long)stats->buf_alloc_sz);
John Koleszar's avatar
John Koleszar committed
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231

        nbytes = fread(stats->buf.buf, 1, stats->buf.sz, stats->file);
        res = (nbytes == stats->buf.sz);
#endif
    }

    return res;
}

int stats_open_mem(stats_io_t *stats, int pass)
{
    int res;
    stats->pass = pass;

    if (!pass)
    {
        stats->buf.sz = 0;
        stats->buf_alloc_sz = 64 * 1024;
        stats->buf.buf = malloc(stats->buf_alloc_sz);
    }

    stats->buf_ptr = stats->buf.buf;
    res = (stats->buf.buf != NULL);
    return res;
}


James Berry's avatar
James Berry committed
232
void stats_close(stats_io_t *stats, int last_pass)
John Koleszar's avatar
John Koleszar committed
233
234
235
{
    if (stats->file)
    {
James Berry's avatar
James Berry committed
236
        if (stats->pass == last_pass)
John Koleszar's avatar
John Koleszar committed
237
238
239
240
241
242
243
244
245
246
247
248
249
250
        {
#if 0
#elif USE_POSIX_MMAP
            munmap(stats->buf.buf, stats->buf.sz);
#else
            free(stats->buf.buf);
#endif
        }

        fclose(stats->file);
        stats->file = NULL;
    }
    else
    {
James Berry's avatar
James Berry committed
251
        if (stats->pass == last_pass)
John Koleszar's avatar
John Koleszar committed
252
253
254
255
256
257
258
259
            free(stats->buf.buf);
    }
}

void stats_write(stats_io_t *stats, const void *pkt, size_t len)
{
    if (stats->file)
    {
260
        (void) fwrite(pkt, 1, len, stats->file);
John Koleszar's avatar
John Koleszar committed
261
262
263
264
265
266
267
268
269
270
271
272
273
    }
    else
    {
        if (stats->buf.sz + len > stats->buf_alloc_sz)
        {
            size_t  new_sz = stats->buf_alloc_sz + 64 * 1024;
            char   *new_ptr = realloc(stats->buf.buf, new_sz);

            if (new_ptr)
            {
                stats->buf_ptr = new_ptr + (stats->buf_ptr - (char *)stats->buf.buf);
                stats->buf.buf = new_ptr;
                stats->buf_alloc_sz = new_sz;
274
275
            }
            else
276
                fatal("Failed to realloc firstpass stats buffer.");
John Koleszar's avatar
John Koleszar committed
277
278
279
280
281
282
283
284
285
286
287
288
289
        }

        memcpy(stats->buf_ptr, pkt, len);
        stats->buf.sz += len;
        stats->buf_ptr += len;
    }
}

vpx_fixed_buf_t stats_get(stats_io_t *stats)
{
    return stats->buf;
}

290
291
292
293
294
295
296
297
298
299
/* Stereo 3D packed frame format */
typedef enum stereo_format
{
    STEREO_FORMAT_MONO       = 0,
    STEREO_FORMAT_LEFT_RIGHT = 1,
    STEREO_FORMAT_BOTTOM_TOP = 2,
    STEREO_FORMAT_TOP_BOTTOM = 3,
    STEREO_FORMAT_RIGHT_LEFT = 11
} stereo_format_t;

300
301
302
303
304
305
306
enum video_file_type
{
    FILE_TYPE_RAW,
    FILE_TYPE_IVF,
    FILE_TYPE_Y4M
};

307
308
struct detect_buffer {
    char buf[4];
309
310
    size_t buf_read;
    size_t position;
311
312
313
};


314
315
316
317
318
319
320
321
322
323
324
325
326
327
struct input_state
{
    char                 *fn;
    FILE                 *file;
    y4m_input             y4m;
    struct detect_buffer  detect;
    enum video_file_type  file_type;
    unsigned int          w;
    unsigned int          h;
    struct vpx_rational   framerate;
    int                   use_i420;
};


John Koleszar's avatar
John Koleszar committed
328
#define IVF_FRAME_HDR_SZ (4+8) /* 4 byte size + 8 byte timestamp */
329
static int read_frame(struct input_state *input, vpx_image_t *img)
John Koleszar's avatar
John Koleszar committed
330
{
331
332
333
334
    FILE *f = input->file;
    enum video_file_type file_type = input->file_type;
    y4m_input *y4m = &input->y4m;
    struct detect_buffer *detect = &input->detect;
John Koleszar's avatar
John Koleszar committed
335
    int plane = 0;
336
    int shortread = 0;
John Koleszar's avatar
John Koleszar committed
337

338
    if (file_type == FILE_TYPE_Y4M)
John Koleszar's avatar
John Koleszar committed
339
    {
340
        if (y4m_input_fetch_frame(y4m, f, img) < 1)
341
           return 0;
John Koleszar's avatar
John Koleszar committed
342
    }
343
    else
John Koleszar's avatar
John Koleszar committed
344
    {
345
        if (file_type == FILE_TYPE_IVF)
John Koleszar's avatar
John Koleszar committed
346
        {
347
348
349
350
351
352
            char junk[IVF_FRAME_HDR_SZ];

            /* Skip the frame header. We know how big the frame should be. See
             * write_ivf_frame_header() for documentation on the frame header
             * layout.
             */
353
            (void) fread(junk, 1, IVF_FRAME_HDR_SZ, f);
John Koleszar's avatar
John Koleszar committed
354
355
        }

356
        for (plane = 0; plane < 3; plane++)
John Koleszar's avatar
John Koleszar committed
357
        {
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
            unsigned char *ptr;
            int w = (plane ? (1 + img->d_w) / 2 : img->d_w);
            int h = (plane ? (1 + img->d_h) / 2 : img->d_h);
            int r;

            /* Determine the correct plane based on the image format. The for-loop
             * always counts in Y,U,V order, but this may not match the order of
             * the data on disk.
             */
            switch (plane)
            {
            case 1:
                ptr = img->planes[img->fmt==VPX_IMG_FMT_YV12? VPX_PLANE_V : VPX_PLANE_U];
                break;
            case 2:
                ptr = img->planes[img->fmt==VPX_IMG_FMT_YV12?VPX_PLANE_U : VPX_PLANE_V];
                break;
            default:
                ptr = img->planes[plane];
            }

            for (r = 0; r < h; r++)
            {
381
382
383
384
                size_t needed = w;
                size_t buf_position = 0;
                const size_t left = detect->buf_read - detect->position;
                if (left > 0)
385
                {
386
387
388
389
390
391
392
393
394
                    const size_t more = (left < needed) ? left : needed;
                    memcpy(ptr, detect->buf + detect->position, more);
                    buf_position = more;
                    needed -= more;
                    detect->position += more;
                }
                if (needed > 0)
                {
                    shortread |= (fread(ptr + buf_position, 1, needed, f) < needed);
395
396
                }

397
398
                ptr += img->stride[plane];
            }
John Koleszar's avatar
John Koleszar committed
399
400
401
        }
    }

402
    return !shortread;
John Koleszar's avatar
John Koleszar committed
403
404
405
}


406
unsigned int file_is_y4m(FILE      *infile,
407
408
                         y4m_input *y4m,
                         char       detect[4])
409
{
410
    if(memcmp(detect, "YUV4", 4) == 0)
411
412
413
414
415
416
    {
        return 1;
    }
    return 0;
}

John Koleszar's avatar
John Koleszar committed
417
#define IVF_FILE_HDR_SZ (32)
418
419
unsigned int file_is_ivf(struct input_state *input,
                         unsigned int *fourcc)
John Koleszar's avatar
John Koleszar committed
420
421
422
{
    char raw_hdr[IVF_FILE_HDR_SZ];
    int is_ivf = 0;
423
424
425
426
    FILE *infile = input->file;
    unsigned int *width = &input->w;
    unsigned int *height = &input->h;
    struct detect_buffer *detect = &input->detect;
John Koleszar's avatar
John Koleszar committed
427

428
    if(memcmp(detect->buf, "DKIF", 4) != 0)
429
430
        return 0;

John Koleszar's avatar
John Koleszar committed
431
432
433
    /* See write_ivf_file_header() for more documentation on the file header
     * layout.
     */
434
435
    if (fread(raw_hdr + 4, 1, IVF_FILE_HDR_SZ - 4, infile)
        == IVF_FILE_HDR_SZ - 4)
John Koleszar's avatar
John Koleszar committed
436
437
438
439
440
    {
        {
            is_ivf = 1;

            if (mem_get_le16(raw_hdr + 4) != 0)
441
442
                warn("Unrecognized IVF version! This file may not decode "
                     "properly.");
John Koleszar's avatar
John Koleszar committed
443
444
445
446
447
448
449
450
451

            *fourcc = mem_get_le32(raw_hdr + 8);
        }
    }

    if (is_ivf)
    {
        *width = mem_get_le16(raw_hdr + 12);
        *height = mem_get_le16(raw_hdr + 14);
452
        detect->position = 4;
John Koleszar's avatar
John Koleszar committed
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
    }

    return is_ivf;
}


static void write_ivf_file_header(FILE *outfile,
                                  const vpx_codec_enc_cfg_t *cfg,
                                  unsigned int fourcc,
                                  int frame_cnt)
{
    char header[32];

    if (cfg->g_pass != VPX_RC_ONE_PASS && cfg->g_pass != VPX_RC_LAST_PASS)
        return;

    header[0] = 'D';
    header[1] = 'K';
    header[2] = 'I';
    header[3] = 'F';
    mem_put_le16(header + 4,  0);                 /* version */
    mem_put_le16(header + 6,  32);                /* headersize */
    mem_put_le32(header + 8,  fourcc);            /* headersize */
    mem_put_le16(header + 12, cfg->g_w);          /* width */
    mem_put_le16(header + 14, cfg->g_h);          /* height */
    mem_put_le32(header + 16, cfg->g_timebase.den); /* rate */
    mem_put_le32(header + 20, cfg->g_timebase.num); /* scale */
    mem_put_le32(header + 24, frame_cnt);         /* length */
    mem_put_le32(header + 28, 0);                 /* unused */

483
    (void) fwrite(header, 1, 32, outfile);
John Koleszar's avatar
John Koleszar committed
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
}


static void write_ivf_frame_header(FILE *outfile,
                                   const vpx_codec_cx_pkt_t *pkt)
{
    char             header[12];
    vpx_codec_pts_t  pts;

    if (pkt->kind != VPX_CODEC_CX_FRAME_PKT)
        return;

    pts = pkt->data.frame.pts;
    mem_put_le32(header, pkt->data.frame.sz);
    mem_put_le32(header + 4, pts & 0xFFFFFFFF);
    mem_put_le32(header + 8, pts >> 32);

501
    (void) fwrite(header, 1, 12, outfile);
John Koleszar's avatar
John Koleszar committed
502
503
}

504
505
506
507
static void write_ivf_frame_size(FILE *outfile, size_t size)
{
    char             header[4];
    mem_put_le32(header, size);
508
    (void) fwrite(header, 1, 4, outfile);
509
510
}

John Koleszar's avatar
John Koleszar committed
511
512
513
514
515
516
517
518
519
520
521
522
523

typedef off_t EbmlLoc;


struct cue_entry
{
    unsigned int time;
    uint64_t     loc;
};


struct EbmlGlobal
{
524
525
    int debug;

John Koleszar's avatar
John Koleszar committed
526
    FILE    *stream;
527
    int64_t last_pts_ms;
John Koleszar's avatar
John Koleszar committed
528
529
530
531
532
533
534
535
536
537
    vpx_rational_t  framerate;

    /* These pointers are to the start of an element */
    off_t    position_reference;
    off_t    seek_info_pos;
    off_t    segment_info_pos;
    off_t    track_pos;
    off_t    cue_pos;
    off_t    cluster_pos;

John Koleszar's avatar
John Koleszar committed
538
539
540
    /* This pointer is to a specific element to be serialized */
    off_t    track_id_pos;

John Koleszar's avatar
John Koleszar committed
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
    /* These pointers are to the size field of the element */
    EbmlLoc  startSegment;
    EbmlLoc  startCluster;

    uint32_t cluster_timecode;
    int      cluster_open;

    struct cue_entry *cue_list;
    unsigned int      cues;

};


void Ebml_Write(EbmlGlobal *glob, const void *buffer_in, unsigned long len)
{
556
    (void) fwrite(buffer_in, 1, len, glob->stream);
John Koleszar's avatar
John Koleszar committed
557
558
}

559
560
561
562
563
564
565
#define WRITE_BUFFER(s) \
for(i = len-1; i>=0; i--)\
{ \
    x = *(const s *)buffer_in >> (i * CHAR_BIT); \
    Ebml_Write(glob, &x, 1); \
}
void Ebml_Serialize(EbmlGlobal *glob, const void *buffer_in, int buffer_size, unsigned long len)
John Koleszar's avatar
John Koleszar committed
566
{
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
    char x;
    int i;

    /* buffer_size:
     * 1 - int8_t;
     * 2 - int16_t;
     * 3 - int32_t;
     * 4 - int64_t;
     */
    switch (buffer_size)
    {
        case 1:
            WRITE_BUFFER(int8_t)
            break;
        case 2:
            WRITE_BUFFER(int16_t)
            break;
        case 4:
            WRITE_BUFFER(int32_t)
            break;
        case 8:
            WRITE_BUFFER(int64_t)
            break;
        default:
            break;
    }
John Koleszar's avatar
John Koleszar committed
593
}
594
#undef WRITE_BUFFER
John Koleszar's avatar
John Koleszar committed
595

596
/* Need a fixed size serializer for the track ID. libmkv provides a 64 bit
John Koleszar's avatar
John Koleszar committed
597
598
599
600
601
602
 * one, but not a 32 bit one.
 */
static void Ebml_SerializeUnsigned32(EbmlGlobal *glob, unsigned long class_id, uint64_t ui)
{
    unsigned char sizeSerialized = 4 | 0x80;
    Ebml_WriteID(glob, class_id);
603
604
    Ebml_Serialize(glob, &sizeSerialized, sizeof(sizeSerialized), 1);
    Ebml_Serialize(glob, &ui, sizeof(ui), 4);
John Koleszar's avatar
John Koleszar committed
605
606
607
}


John Koleszar's avatar
John Koleszar committed
608
609
610
611
static void
Ebml_StartSubElement(EbmlGlobal *glob, EbmlLoc *ebmlLoc,
                          unsigned long class_id)
{
John Koleszar's avatar
John Koleszar committed
612
613
    /* todo this is always taking 8 bytes, this may need later optimization */
    /* this is a key that says length unknown */
John Koleszar's avatar
John Koleszar committed
614
    uint64_t unknownLen = LITERALU64(0x01FFFFFF, 0xFFFFFFFF);
John Koleszar's avatar
John Koleszar committed
615
616
617

    Ebml_WriteID(glob, class_id);
    *ebmlLoc = ftello(glob->stream);
618
    Ebml_Serialize(glob, &unknownLen, sizeof(unknownLen), 8);
John Koleszar's avatar
John Koleszar committed
619
620
621
622
623
624
625
626
627
628
629
630
631
}

static void
Ebml_EndSubElement(EbmlGlobal *glob, EbmlLoc *ebmlLoc)
{
    off_t pos;
    uint64_t size;

    /* Save the current stream pointer */
    pos = ftello(glob->stream);

    /* Calculate the size of this element */
    size = pos - *ebmlLoc - 8;
John Koleszar's avatar
John Koleszar committed
632
    size |= LITERALU64(0x01000000,0x00000000);
John Koleszar's avatar
John Koleszar committed
633
634
635

    /* Seek back to the beginning of the element and write the new size */
    fseeko(glob->stream, *ebmlLoc, SEEK_SET);
636
    Ebml_Serialize(glob, &size, sizeof(size), 8);
John Koleszar's avatar
John Koleszar committed
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678

    /* Reset the stream pointer */
    fseeko(glob->stream, pos, SEEK_SET);
}


static void
write_webm_seek_element(EbmlGlobal *ebml, unsigned long id, off_t pos)
{
    uint64_t offset = pos - ebml->position_reference;
    EbmlLoc start;
    Ebml_StartSubElement(ebml, &start, Seek);
    Ebml_SerializeBinary(ebml, SeekID, id);
    Ebml_SerializeUnsigned64(ebml, SeekPosition, offset);
    Ebml_EndSubElement(ebml, &start);
}


static void
write_webm_seek_info(EbmlGlobal *ebml)
{

    off_t pos;

    /* Save the current stream pointer */
    pos = ftello(ebml->stream);

    if(ebml->seek_info_pos)
        fseeko(ebml->stream, ebml->seek_info_pos, SEEK_SET);
    else
        ebml->seek_info_pos = pos;

    {
        EbmlLoc start;

        Ebml_StartSubElement(ebml, &start, SeekHead);
        write_webm_seek_element(ebml, Tracks, ebml->track_pos);
        write_webm_seek_element(ebml, Cues,   ebml->cue_pos);
        write_webm_seek_element(ebml, Info,   ebml->segment_info_pos);
        Ebml_EndSubElement(ebml, &start);
    }
    {
John Koleszar's avatar
John Koleszar committed
679
        /* segment info */
John Koleszar's avatar
John Koleszar committed
680
681
        EbmlLoc startInfo;
        uint64_t frame_time;
682
683
684
685
686
687
688
689
690
691
692
693
        char version_string[64];

        /* Assemble version string */
        if(ebml->debug)
            strcpy(version_string, "vpxenc");
        else
        {
            strcpy(version_string, "vpxenc ");
            strncat(version_string,
                    vpx_codec_version_str(),
                    sizeof(version_string) - 1 - strlen(version_string));
        }
John Koleszar's avatar
John Koleszar committed
694
695
696
697
698
699
700
701

        frame_time = (uint64_t)1000 * ebml->framerate.den
                     / ebml->framerate.num;
        ebml->segment_info_pos = ftello(ebml->stream);
        Ebml_StartSubElement(ebml, &startInfo, Info);
        Ebml_SerializeUnsigned(ebml, TimecodeScale, 1000000);
        Ebml_SerializeFloat(ebml, Segment_Duration,
                            ebml->last_pts_ms + frame_time);
702
703
        Ebml_SerializeString(ebml, 0x4D80, version_string);
        Ebml_SerializeString(ebml, 0x5741, version_string);
John Koleszar's avatar
John Koleszar committed
704
705
706
707
708
709
710
711
        Ebml_EndSubElement(ebml, &startInfo);
    }
}


static void
write_webm_file_header(EbmlGlobal                *glob,
                       const vpx_codec_enc_cfg_t *cfg,
712
713
                       const struct vpx_rational *fps,
                       stereo_format_t            stereo_fmt)
John Koleszar's avatar
John Koleszar committed
714
715
716
717
718
{
    {
        EbmlLoc start;
        Ebml_StartSubElement(glob, &start, EBML);
        Ebml_SerializeUnsigned(glob, EBMLVersion, 1);
John Koleszar's avatar
John Koleszar committed
719
720
721
722
723
724
        Ebml_SerializeUnsigned(glob, EBMLReadVersion, 1);
        Ebml_SerializeUnsigned(glob, EBMLMaxIDLength, 4);
        Ebml_SerializeUnsigned(glob, EBMLMaxSizeLength, 8);
        Ebml_SerializeString(glob, DocType, "webm");
        Ebml_SerializeUnsigned(glob, DocTypeVersion, 2);
        Ebml_SerializeUnsigned(glob, DocTypeReadVersion, 2);
John Koleszar's avatar
John Koleszar committed
725
726
727
        Ebml_EndSubElement(glob, &start);
    }
    {
John Koleszar's avatar
John Koleszar committed
728
        Ebml_StartSubElement(glob, &glob->startSegment, Segment);
John Koleszar's avatar
John Koleszar committed
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
        glob->position_reference = ftello(glob->stream);
        glob->framerate = *fps;
        write_webm_seek_info(glob);

        {
            EbmlLoc trackStart;
            glob->track_pos = ftello(glob->stream);
            Ebml_StartSubElement(glob, &trackStart, Tracks);
            {
                unsigned int trackNumber = 1;
                uint64_t     trackID = 0;

                EbmlLoc start;
                Ebml_StartSubElement(glob, &start, TrackEntry);
                Ebml_SerializeUnsigned(glob, TrackNumber, trackNumber);
John Koleszar's avatar
John Koleszar committed
744
745
                glob->track_id_pos = ftello(glob->stream);
                Ebml_SerializeUnsigned32(glob, TrackUID, trackID);
John Koleszar's avatar
John Koleszar committed
746
                Ebml_SerializeUnsigned(glob, TrackType, 1);
John Koleszar's avatar
John Koleszar committed
747
748
749
750
751
752
753
754
755
756
                Ebml_SerializeString(glob, CodecID, "V_VP8");
                {
                    unsigned int pixelWidth = cfg->g_w;
                    unsigned int pixelHeight = cfg->g_h;
                    float        frameRate   = (float)fps->num/(float)fps->den;

                    EbmlLoc videoStart;
                    Ebml_StartSubElement(glob, &videoStart, Video);
                    Ebml_SerializeUnsigned(glob, PixelWidth, pixelWidth);
                    Ebml_SerializeUnsigned(glob, PixelHeight, pixelHeight);
757
                    Ebml_SerializeUnsigned(glob, StereoMode, stereo_fmt);
John Koleszar's avatar
John Koleszar committed
758
                    Ebml_SerializeFloat(glob, FrameRate, frameRate);
John Koleszar's avatar
John Koleszar committed
759
                    Ebml_EndSubElement(glob, &videoStart);
John Koleszar's avatar
John Koleszar committed
760
                }
John Koleszar's avatar
John Koleszar committed
761
                Ebml_EndSubElement(glob, &start); /* Track Entry */
John Koleszar's avatar
John Koleszar committed
762
763
764
            }
            Ebml_EndSubElement(glob, &trackStart);
        }
John Koleszar's avatar
John Koleszar committed
765
        /* segment element is open */
John Koleszar's avatar
John Koleszar committed
766
767
768
769
770
771
772
773
774
775
776
777
778
    }
}


static void
write_webm_block(EbmlGlobal                *glob,
                 const vpx_codec_enc_cfg_t *cfg,
                 const vpx_codec_cx_pkt_t  *pkt)
{
    unsigned long  block_length;
    unsigned char  track_number;
    unsigned short block_timecode = 0;
    unsigned char  flags;
779
    int64_t        pts_ms;
John Koleszar's avatar
John Koleszar committed
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
    int            start_cluster = 0, is_keyframe;

    /* Calculate the PTS of this frame in milliseconds */
    pts_ms = pkt->data.frame.pts * 1000
             * (uint64_t)cfg->g_timebase.num / (uint64_t)cfg->g_timebase.den;
    if(pts_ms <= glob->last_pts_ms)
        pts_ms = glob->last_pts_ms + 1;
    glob->last_pts_ms = pts_ms;

    /* Calculate the relative time of this block */
    if(pts_ms - glob->cluster_timecode > SHRT_MAX)
        start_cluster = 1;
    else
        block_timecode = pts_ms - glob->cluster_timecode;

    is_keyframe = (pkt->data.frame.flags & VPX_FRAME_IS_KEY);
    if(start_cluster || is_keyframe)
    {
        if(glob->cluster_open)
            Ebml_EndSubElement(glob, &glob->startCluster);

        /* Open the new cluster */
        block_timecode = 0;
        glob->cluster_open = 1;
        glob->cluster_timecode = pts_ms;
        glob->cluster_pos = ftello(glob->stream);
John Koleszar's avatar
John Koleszar committed
806
        Ebml_StartSubElement(glob, &glob->startCluster, Cluster); /* cluster */
John Koleszar's avatar
John Koleszar committed
807
808
809
810
811
        Ebml_SerializeUnsigned(glob, Timecode, glob->cluster_timecode);

        /* Save a cue point if this is a keyframe. */
        if(is_keyframe)
        {
812
813
814
815
816
817
818
            struct cue_entry *cue, *new_cue_list;

            new_cue_list = realloc(glob->cue_list,
                                   (glob->cues+1) * sizeof(struct cue_entry));
            if(new_cue_list)
                glob->cue_list = new_cue_list;
            else
819
                fatal("Failed to realloc cue list.");
John Koleszar's avatar
John Koleszar committed
820
821
822
823
824
825
826
827
828
829
830
831
832

            cue = &glob->cue_list[glob->cues];
            cue->time = glob->cluster_timecode;
            cue->loc = glob->cluster_pos;
            glob->cues++;
        }
    }

    /* Write the Simple Block */
    Ebml_WriteID(glob, SimpleBlock);

    block_length = pkt->data.frame.sz + 4;
    block_length |= 0x10000000;
833
    Ebml_Serialize(glob, &block_length, sizeof(block_length), 4);
John Koleszar's avatar
John Koleszar committed
834
835
836
837
838

    track_number = 1;
    track_number |= 0x80;
    Ebml_Write(glob, &track_number, 1);

839
    Ebml_Serialize(glob, &block_timecode, sizeof(block_timecode), 2);
John Koleszar's avatar
John Koleszar committed
840
841
842
843
844
845
846
847
848
849
850
851
852

    flags = 0;
    if(is_keyframe)
        flags |= 0x80;
    if(pkt->data.frame.flags & VPX_FRAME_IS_INVISIBLE)
        flags |= 0x08;
    Ebml_Write(glob, &flags, 1);

    Ebml_Write(glob, pkt->data.frame.buf, pkt->data.frame.sz);
}


static void
John Koleszar's avatar
John Koleszar committed
853
write_webm_file_footer(EbmlGlobal *glob, long hash)
John Koleszar's avatar
John Koleszar committed
854
855
856
857
858
859
860
{

    if(glob->cluster_open)
        Ebml_EndSubElement(glob, &glob->startCluster);

    {
        EbmlLoc start;
861
        unsigned int i;
John Koleszar's avatar
John Koleszar committed
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890

        glob->cue_pos = ftello(glob->stream);
        Ebml_StartSubElement(glob, &start, Cues);
        for(i=0; i<glob->cues; i++)
        {
            struct cue_entry *cue = &glob->cue_list[i];
            EbmlLoc start;

            Ebml_StartSubElement(glob, &start, CuePoint);
            {
                EbmlLoc start;

                Ebml_SerializeUnsigned(glob, CueTime, cue->time);

                Ebml_StartSubElement(glob, &start, CueTrackPositions);
                Ebml_SerializeUnsigned(glob, CueTrack, 1);
                Ebml_SerializeUnsigned64(glob, CueClusterPosition,
                                         cue->loc - glob->position_reference);
                Ebml_EndSubElement(glob, &start);
            }
            Ebml_EndSubElement(glob, &start);
        }
        Ebml_EndSubElement(glob, &start);
    }

    Ebml_EndSubElement(glob, &glob->startSegment);

    /* Patch up the seek info block */
    write_webm_seek_info(glob);
John Koleszar's avatar
John Koleszar committed
891
892
893
894
895

    /* Patch up the track id */
    fseeko(glob->stream, glob->track_id_pos, SEEK_SET);
    Ebml_SerializeUnsigned32(glob, TrackUID, glob->debug ? 0xDEADBEEF : hash);

John Koleszar's avatar
John Koleszar committed
896
897
898
899
    fseeko(glob->stream, 0, SEEK_END);
}


John Koleszar's avatar
John Koleszar committed
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
/* Murmur hash derived from public domain reference implementation at
 *   http://sites.google.com/site/murmurhash/
 */
static unsigned int murmur ( const void * key, int len, unsigned int seed )
{
    const unsigned int m = 0x5bd1e995;
    const int r = 24;

    unsigned int h = seed ^ len;

    const unsigned char * data = (const unsigned char *)key;

    while(len >= 4)
    {
        unsigned int k;

        k  = data[0];
        k |= data[1] << 8;
        k |= data[2] << 16;
        k |= data[3] << 24;

        k *= m;
        k ^= k >> r;
        k *= m;

        h *= m;
        h ^= k;

        data += 4;
        len -= 4;
    }

    switch(len)
    {
    case 3: h ^= data[2] << 16;
    case 2: h ^= data[1] << 8;
    case 1: h ^= data[0];
            h *= m;
    };

    h ^= h >> 13;
    h *= m;
    h ^= h >> 15;

    return h;
}

947
948
949
950
951
952
953
954
955
#include "math.h"

static double vp8_mse2psnr(double Samples, double Peak, double Mse)
{
    double psnr;

    if ((double)Mse > 0.0)
        psnr = 10.0 * log10(Peak * Peak * Samples / Mse);
    else
John Koleszar's avatar
John Koleszar committed
956
        psnr = 60;      /* Limit to prevent / 0 */
957
958
959
960
961
962
963

    if (psnr > 60)
        psnr = 60;

    return psnr;
}

John Koleszar's avatar
John Koleszar committed
964

John Koleszar's avatar
John Koleszar committed
965
#include "args.h"
966
967
static const arg_def_t debugmode = ARG_DEF("D", "debug", 0,
        "Debug mode (makes output deterministic)");
968
969
static const arg_def_t outputfile = ARG_DEF("o", "output", 1,
        "Output filename");
John Koleszar's avatar
John Koleszar committed
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
static const arg_def_t use_yv12 = ARG_DEF(NULL, "yv12", 0,
                                  "Input file is YV12 ");
static const arg_def_t use_i420 = ARG_DEF(NULL, "i420", 0,
                                  "Input file is I420 (default)");
static const arg_def_t codecarg = ARG_DEF(NULL, "codec", 1,
                                  "Codec to use");
static const arg_def_t passes           = ARG_DEF("p", "passes", 1,
        "Number of passes (1/2)");
static const arg_def_t pass_arg         = ARG_DEF(NULL, "pass", 1,
        "Pass to execute (1/2)");
static const arg_def_t fpf_name         = ARG_DEF(NULL, "fpf", 1,
        "First pass statistics file name");
static const arg_def_t limit = ARG_DEF(NULL, "limit", 1,
                                       "Stop encoding after n input frames");
static const arg_def_t deadline         = ARG_DEF("d", "deadline", 1,
        "Deadline per frame (usec)");
static const arg_def_t best_dl          = ARG_DEF(NULL, "best", 0,
        "Use Best Quality Deadline");
static const arg_def_t good_dl          = ARG_DEF(NULL, "good", 0,
        "Use Good Quality Deadline");
static const arg_def_t rt_dl            = ARG_DEF(NULL, "rt", 0,
        "Use Realtime Quality Deadline");
static const arg_def_t verbosearg       = ARG_DEF("v", "verbose", 0,
        "Show encoder parameters");
static const arg_def_t psnrarg          = ARG_DEF(NULL, "psnr", 0,
        "Show PSNR in status line");
996
static const arg_def_t framerate        = ARG_DEF(NULL, "fps", 1,
997
        "Stream frame rate (rate/scale)");
John Koleszar's avatar
John Koleszar committed
998
999
static const arg_def_t use_ivf          = ARG_DEF(NULL, "ivf", 0,
        "Output IVF (default is WebM)");
1000
1001
static const arg_def_t out_part = ARG_DEF("P", "output-partitions", 0,
        "Makes encoder output partitions. Requires IVF output!");
1002
1003
static const arg_def_t q_hist_n         = ARG_DEF(NULL, "q-hist", 1,
        "Show quantizer histogram (n-buckets)");
1004
1005
static const arg_def_t rate_hist_n         = ARG_DEF(NULL, "rate-hist", 1,
        "Show rate histogram (n-buckets)");
John Koleszar's avatar
John Koleszar committed
1006
1007
static const arg_def_t *main_args[] =
{
1008
    &debugmode,
1009
1010
    &outputfile, &codecarg, &passes, &pass_arg, &fpf_name, &limit, &deadline,
    &best_dl, &good_dl, &rt_dl,
1011
    &verbosearg, &psnrarg, &use_ivf, &out_part, &q_hist_n, &rate_hist_n,
John Koleszar's avatar
John Koleszar committed
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
    NULL
};

static const arg_def_t usage            = ARG_DEF("u", "usage", 1,
        "Usage profile number to use");
static const arg_def_t threads          = ARG_DEF("t", "threads", 1,
        "Max number of threads to use");
static const arg_def_t profile          = ARG_DEF(NULL, "profile", 1,
        "Bitstream profile number to use");
static const arg_def_t width            = ARG_DEF("w", "width", 1,
        "Frame width");
static const arg_def_t height           = ARG_DEF("h", "height", 1,
        "Frame height");
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
static const struct arg_enum_list stereo_mode_enum[] = {
    {"mono"      , STEREO_FORMAT_MONO},
    {"left-right", STEREO_FORMAT_LEFT_RIGHT},
    {"bottom-top", STEREO_FORMAT_BOTTOM_TOP},
    {"top-bottom", STEREO_FORMAT_TOP_BOTTOM},
    {"right-left", STEREO_FORMAT_RIGHT_LEFT},
    {NULL, 0}
};
static const arg_def_t stereo_mode      = ARG_DEF_ENUM(NULL, "stereo-mode", 1,
        "Stereo 3D video format", stereo_mode_enum);
John Koleszar's avatar
John Koleszar committed
1035
static const arg_def_t timebase         = ARG_DEF(NULL, "timebase", 1,
1036
        "Output timestamp precision (fractional seconds)");
John Koleszar's avatar
John Koleszar committed
1037
1038
1039
1040
1041
1042
1043
1044
static const arg_def_t error_resilient  = ARG_DEF(NULL, "error-resilient", 1,
        "Enable error resiliency features");
static const arg_def_t lag_in_frames    = ARG_DEF(NULL, "lag-in-frames", 1,
        "Max number of frames to lag");

static const arg_def_t *global_args[] =
{
    &use_yv12, &use_i420, &usage, &threads, &profile,
1045
    &width, &height, &stereo_mode, &timebase, &framerate, &error_resilient,
John Koleszar's avatar
John Koleszar committed
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
    &lag_in_frames, NULL
};

static const arg_def_t dropframe_thresh   = ARG_DEF(NULL, "drop-frame", 1,
        "Temporal resampling threshold (buf %)");
static const arg_def_t resize_allowed     = ARG_DEF(NULL, "resize-allowed", 1,
        "Spatial resampling enabled (bool)");
static const arg_def_t resize_up_thresh   = ARG_DEF(NULL, "resize-up", 1,
        "Upscale threshold (buf %)");
static const arg_def_t resize_down_thresh = ARG_DEF(NULL, "resize-down", 1,
        "Downscale threshold (buf %)");
1057
1058
1059
1060
1061
1062
1063
1064
static const struct arg_enum_list end_usage_enum[] = {
    {"vbr", VPX_VBR},
    {"cbr", VPX_CBR},
    {"cq",  VPX_CQ},
    {NULL, 0}
};
static const arg_def_t end_usage          = ARG_DEF_ENUM(NULL, "end-usage", 1,
        "Rate control mode", end_usage_enum);
John Koleszar's avatar
John Koleszar committed
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
static const arg_def_t target_bitrate     = ARG_DEF(NULL, "target-bitrate", 1,
        "Bitrate (kbps)");
static const arg_def_t min_quantizer      = ARG_DEF(NULL, "min-q", 1,
        "Minimum (best) quantizer");
static const arg_def_t max_quantizer      = ARG_DEF(NULL, "max-q", 1,
        "Maximum (worst) quantizer");
static const arg_def_t undershoot_pct     = ARG_DEF(NULL, "undershoot-pct", 1,
        "Datarate undershoot (min) target (%)");
static const arg_def_t overshoot_pct      = ARG_DEF(NULL, "overshoot-pct", 1,
        "Datarate overshoot (max) target (%)");
static const arg_def_t buf_sz             = ARG_DEF(NULL, "buf-sz", 1,
        "Client buffer size (ms)");
static const arg_def_t buf_initial_sz     = ARG_DEF(NULL, "buf-initial-sz", 1,
        "Client initial buffer size (ms)");
static const arg_def_t buf_optimal_sz     = ARG_DEF(NULL, "buf-optimal-sz", 1,
        "Client optimal buffer size (ms)");
static const arg_def_t *rc_args[] =
{
    &dropframe_thresh, &resize_allowed, &resize_up_thresh, &resize_down_thresh,
    &end_usage, &target_bitrate, &min_quantizer, &max_quantizer,
    &undershoot_pct, &overshoot_pct, &buf_sz, &buf_initial_sz, &buf_optimal_sz,
    NULL
};


static const arg_def_t bias_pct = ARG_DEF(NULL, "bias-pct", 1,
                                  "CBR/VBR bias (0=CBR, 100=VBR)");
static const arg_def_t minsection_pct = ARG_DEF(NULL, "minsection-pct", 1,
                                        "GOP min bitrate (% of target)");
static const arg_def_t maxsection_pct = ARG_DEF(NULL, "maxsection-pct", 1,
                                        "GOP max bitrate (% of target)");
static const arg_def_t *rc_twopass_args[] =
{
    &bias_pct, &minsection_pct, &maxsection_pct, NULL
};


static const arg_def_t kf_min_dist = ARG_DEF(NULL, "kf-min-dist", 1,
                                     "Minimum keyframe interval (frames)");
static const arg_def_t kf_max_dist = ARG_DEF(NULL, "kf-max-dist", 1,
                                     "Maximum keyframe interval (frames)");
1106
1107
static const arg_def_t kf_disabled = ARG_DEF(NULL, "disable-kf", 0,
                                     "Disable keyframe placement");
John Koleszar's avatar
John Koleszar committed
1108
1109
static const arg_def_t *kf_args[] =
{
1110
    &kf_min_dist, &kf_max_dist, &kf_disabled, NULL
John Koleszar's avatar
John Koleszar committed
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
};


#if CONFIG_VP8_ENCODER
static const arg_def_t noise_sens = ARG_DEF(NULL, "noise-sensitivity", 1,
                                    "Noise sensitivity (frames to blur)");
static const arg_def_t sharpness = ARG_DEF(NULL, "sharpness", 1,
                                   "Filter sharpness (0-7)");
static const arg_def_t static_thresh = ARG_DEF(NULL, "static-thresh", 1,
                                       "Motion detection threshold");
#endif

#if CONFIG_VP8_ENCODER
static const arg_def_t cpu_used = ARG_DEF(NULL, "cpu-used", 1,
                                  "CPU Used (-16..16)");
#endif


#if CONFIG_VP8_ENCODER
static const arg_def_t token_parts = ARG_DEF(NULL, "token-parts", 1,
                                     "Number of token partitions to use, log2");
static const arg_def_t auto_altref = ARG_DEF(NULL, "auto-alt-ref", 1,
                                     "Enable automatic alt reference frames");
static const arg_def_t arnr_maxframes = ARG_DEF(NULL, "arnr-maxframes", 1,
1135
                                        "AltRef Max Frames");
John Koleszar's avatar
John Koleszar committed
1136
static const arg_def_t arnr_strength = ARG_DEF(NULL, "arnr-strength", 1,
1137
                                       "AltRef Strength");
John Koleszar's avatar
John Koleszar committed
1138
static const arg_def_t arnr_type = ARG_DEF(NULL, "arnr-type", 1,
1139
                                   "AltRef Type");
John Koleszar's avatar
John Koleszar committed
1140
1141
1142
1143
1144
1145
1146
static const struct arg_enum_list tuning_enum[] = {
    {"psnr", VP8_TUNE_PSNR},
    {"ssim", VP8_TUNE_SSIM},
    {NULL, 0}
};
static const arg_def_t tune_ssim = ARG_DEF_ENUM(NULL, "tune", 1,
                                   "Material to favor", tuning_enum);
Paul Wilkins's avatar
CQ Mode    
Paul Wilkins committed
1147
1148
static const arg_def_t cq_level = ARG_DEF(NULL, "cq-level", 1,
                                   "Constrained Quality Level");
1149
1150
static const arg_def_t max_intra_rate_pct = ARG_DEF(NULL, "max-intra-rate", 1,
        "Max I-frame bitrate (pct)");
John Koleszar's avatar
John Koleszar committed
1151
1152
1153
1154

static const arg_def_t *vp8_args[] =
{
    &cpu_used, &auto_altref, &noise_sens, &sharpness, &static_thresh,
John Koleszar's avatar
John Koleszar committed
1155
    &token_parts, &arnr_maxframes, &arnr_strength, &arnr_type,
1156
    &tune_ssim, &cq_level, &max_intra_rate_pct, NULL
John Koleszar's avatar
John Koleszar committed
1157
1158
1159
1160
1161
1162
};
static const int vp8_arg_ctrl_map[] =
{
    VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
    VP8E_SET_NOISE_SENSITIVITY, VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
    VP8E_SET_TOKEN_PARTITIONS,
John Koleszar's avatar
John Koleszar committed
1163
    VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH , VP8E_SET_ARNR_TYPE,
1164
    VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT, 0
John Koleszar's avatar
John Koleszar committed
1165
1166
1167
1168
1169
1170
1171
1172
1173
};
#endif

static const arg_def_t *no_args[] = { NULL };

static void usage_exit()
{
    int i;

1174
1175
    fprintf(stderr, "Usage: %s <options> -o dst_filename src_filename \n",
            exec_name);
John Koleszar's avatar
John Koleszar committed
1176

1177
    fprintf(stderr, "\nOptions:\n");
John Koleszar's avatar
John Koleszar committed
1178
    arg_show_usage(stdout, main_args);
1179
    fprintf(stderr, "\nEncoder Global Options:\n");
John Koleszar's avatar
John Koleszar committed
1180
    arg_show_usage(stdout, global_args);
1181
    fprintf(stderr, "\nRate Control Options:\n");
John Koleszar's avatar
John Koleszar committed
1182
    arg_show_usage(stdout, rc_args);
1183
    fprintf(stderr, "\nTwopass Rate Control Options:\n");
John Koleszar's avatar
John Koleszar committed
1184
    arg_show_usage(stdout, rc_twopass_args);
1185
    fprintf(stderr, "\nKeyframe Placement Options:\n");
John Koleszar's avatar
John Koleszar committed
1186
1187
    arg_show_usage(stdout, kf_args);
#if CONFIG_VP8_ENCODER
1188
    fprintf(stderr, "\nVP8 Specific Options:\n");
John Koleszar's avatar
John Koleszar committed
1189
1190
    arg_show_usage(stdout, vp8_args);
#endif
1191
    fprintf(stderr, "\nStream timebase (--timebase):\n"
1192
            "  The desired precision of timestamps in the output, expressed\n"
1193
            "  in fractional seconds. Default is 1/1000.\n");
1194
    fprintf(stderr, "\n"
John Koleszar's avatar
John Koleszar committed
1195
1196
1197
1198
           "Included encoders:\n"
           "\n");

    for (i = 0; i < sizeof(codecs) / sizeof(codecs[0]); i++)
1199
        fprintf(stderr, "    %-6s - %s\n",
John Koleszar's avatar
John Koleszar committed
1200
1201
1202
1203
1204
1205
               codecs[i].name,
               vpx_codec_iface_name(codecs[i].iface));

    exit(EXIT_FAILURE);
}

1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231

#define HIST_BAR_MAX 40
struct hist_bucket
{
    int low, high, count;
};


static int merge_hist_buckets(struct hist_bucket *bucket,
                              int *buckets_,
                              int max_buckets)
{
    int small_bucket = 0, merge_bucket = INT_MAX, big_bucket=0;
    int buckets = *buckets_;
    int i;

    /* Find the extrema for this list of buckets */
    big_bucket = small_bucket = 0;
    for(i=0; i < buckets; i++)
    {
        if(bucket[i].count < bucket[small_bucket].count)
            small_bucket = i;
        if(bucket[i].count > bucket[big_bucket].count)
            big_bucket = i;
    }

1232
    /* If we have too many buckets, merge the smallest with an adjacent
1233
1234
1235
1236
1237
1238
     * bucket.
     */
    while(buckets > max_buckets)
    {
        int last_bucket = buckets - 1;

John Koleszar's avatar
John Koleszar committed
1239
        /* merge the small bucket with an adjacent one. */
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
        if(small_bucket == 0)
            merge_bucket = 1;
        else if(small_bucket == last_bucket)
            merge_bucket = last_bucket - 1;
        else if(bucket[small_bucket - 1].count < bucket[small_bucket + 1].count)
            merge_bucket = small_bucket - 1;
        else
            merge_bucket = small_bucket + 1;

        assert(abs(merge_bucket - small_bucket) <= 1);
        assert(small_bucket < buckets);
        assert(big_bucket < buckets);
        assert(merge_bucket < buckets);

        if(merge_bucket < small_bucket)
        {
            bucket[merge_bucket].high = bucket[small_bucket].high;
            bucket[merge_bucket].count += bucket[small_bucket].count;
        }
        else
        {
            bucket[small_bucket].high = bucket[merge_bucket].high;
            bucket[small_bucket].count += bucket[merge_bucket].count;
            merge_bucket = small_bucket;
        }

        assert(bucket[merge_bucket].low != bucket[merge_bucket].high);

        buckets--;

        /* Remove the merge_bucket from the list, and find the new small
         * and big buckets while we're at it
         */
        big_bucket = small_bucket = 0;
        for(i=0; i < buckets; i++)
        {
            if(i > merge_bucket)
                bucket[i] = bucket[i+1];

            if(bucket[i].count < bucket[small_bucket].count)
                small_bucket = i;
            if(bucket[i].count > bucket[big_bucket].count)
                big_bucket = i;
        }

    }

    *buckets_ = buckets;
    return bucket[big_bucket].count;
}


static void show_histogram(const struct hist_bucket *bucket,
                           int                       buckets,
                           int                       total,
                           int                       scale)
{
1297
    const char *pat1, *pat2;
1298
1299
    int i;

1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
    switch((int)(log(bucket[buckets-1].high)/log(10))+1)
    {
        case 1:
        case 2:
            pat1 = "%4d %2s: ";
            pat2 = "%4d-%2d: ";
            break;
        case 3:
            pat1 = "%5d %3s: ";
            pat2 = "%5d-%3d: ";
            break;
        case 4:
            pat1 = "%6d %4s: ";
            pat2 = "%6d-%4d: ";
            break;
        case 5:
            pat1 = "%7d %5s: ";
            pat2 = "%7d-%5d: ";
            break;
        case 6:
            pat1 = "%8d %6s: ";
            pat2 = "%8d-%6d: ";
            break;
        case 7:
            pat1 = "%9d %7s: ";
            pat2 = "%9d-%7d: ";
            break;
        default:
            pat1 = "%12d %10s: ";
            pat2 = "%12d-%10d: ";
            break;
    }

1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
    for(i=0; i<buckets; i++)
    {
        int len;
        int j;
        float pct;

        pct = 100.0 * (float)bucket[i].count / (float)total;
        len = HIST_BAR_MAX * bucket[i].count / scale;
        if(len < 1)
            len = 1;
        assert(len <= HIST_BAR_MAX);

        if(bucket[i].low == bucket[i].high)
1346
            fprintf(stderr, pat1, bucket[i].low, "");
1347
        else
1348
            fprintf(stderr, pat2, bucket[i].low, bucket[i].high);
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376

        for(j=0; j<HIST_BAR_MAX; j++)
            fprintf(stderr, j<len?"=":" ");
        fprintf(stderr, "\t%5d (%6.2f%%)\n",bucket[i].count,pct);
    }
}


static void show_q_histogram(const int counts[64], int max_buckets)
{
    struct hist_bucket bucket[64];
    int buckets = 0;
    int total = 0;
    int scale;
    int i;


    for(i=0; i<64; i++)
    {
        if(counts[i])
        {
            bucket[buckets].low = bucket[buckets].high = i;
            bucket[buckets].count = counts[i];
            buckets++;
            total += counts[i];
        }
    }

1377
    fprintf(stderr, "\nQuantizer Selection:\n");
1378
1379
1380
1381
1382
    scale = merge_hist_buckets(bucket, &buckets, max_buckets);
    show_histogram(bucket, buckets, total, scale);
}


1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
#define RATE_BINS (100)
struct rate_hist
{
    int64_t            *pts;
    int                *sz;
    int                 samples;
    int                 frames;
    struct hist_bucket  bucket[RATE_BINS];
    int                 total;
};


static void init_rate_histogram(struct rate_hist          *hist,
1396
1397
                                const vpx_codec_enc_cfg_t *cfg,
                                const vpx_rational_t      *fps)
1398
1399
1400
{
    int i;

1401
1402
1403
1404
1405
    /* Determine the number of samples in the buffer. Use the file's framerate
     * to determine the number of frames in rc_buf_sz milliseconds, with an
     * adjustment (5/4) to account for alt-refs
     */
    hist->samples = cfg->rc_buf_sz * 5 / 4 * fps->num / fps->den / 1000;
1406

John Koleszar's avatar
John Koleszar committed
1407
    /* prevent division by zero */
1408
1409
1410
    if (hist->samples == 0)
      hist->samples=1;

1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
    hist->pts = calloc(hist->samples, sizeof(*hist->pts));
    hist->sz = calloc(hist->samples, sizeof(*hist->sz));
    for(i=0; i<RATE_BINS; i++)
    {
        hist->bucket[i].low = INT_MAX;
        hist->bucket[i].high = 0;
        hist->bucket[i].count = 0;
    }
}


static void destroy_rate_histogram(struct rate_hist *hist)
{
    free(hist->pts);
    free(hist->sz);
}


static void update_rate_histogram(struct rate_hist          *hist,
                                  const vpx_codec_enc_cfg_t *cfg,
                                  const vpx_codec_cx_pkt_t  *pkt)
{
    int i, idx;
    int64_t now, then, sum_sz = 0, avg_bitrate;

    now = pkt->data.frame.pts * 1000
          * (uint64_t)cfg->g_timebase.num / (uint64_t)cfg->g_timebase.den;

    idx = hist->frames++ % hist->samples;
    hist->pts[idx] = now;
    hist->sz[idx] = pkt->data.frame.sz;

    if(now < cfg->rc_buf_initial_sz)
        return;

Johann's avatar
Johann committed
1446
1447
    then = now;

1448
    /* Sum the size over the past rc_buf_sz ms */
1449
    for(i = hist->frames; i > 0 && hist->frames - i < hist->samples; i--)
1450
1451
1452
1453
1454
1455
1456
1457
1458
    {
        int i_idx = (i-1) % hist->samples;

        then = hist->pts[i_idx];
        if(now - then > cfg->rc_buf_sz)
            break;
        sum_sz += hist->sz[i_idx];
    }

Johann's avatar
Johann committed
1459
1460
1461
    if (now == then)
        return;

1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
    avg_bitrate = sum_sz * 8 * 1000 / (now - then);
    idx = avg_bitrate * (RATE_BINS/2) / (cfg->rc_target_bitrate * 1000);
    if(idx < 0)
        idx = 0;
    if(idx > RATE_BINS-1)
        idx = RATE_BINS-1;
    if(hist->bucket[idx].low > avg_bitrate)
        hist->bucket[idx].low = avg_bitrate;
    if(hist->bucket[idx].high < avg_bitrate)
        hist->bucket[idx].high = avg_bitrate;
    hist->bucket[idx].count++;
    hist->total++;
}


static void show_rate_histogram(struct rate_hist          *hist,
                                const vpx_codec_enc_cfg_t *cfg,
                                int                        max_buckets)
{
    int i, scale;
    int buckets = 0;

    for(i = 0; i < RATE_BINS; i++)
    {
        if(hist->bucket[i].low == INT_MAX)
            continue;
        hist->bucket[buckets++] = hist->bucket[i];
    }

    fprintf(stderr, "\nRate (over %dms window):\n", cfg->rc_buf_sz);
    scale = merge_hist_buckets(hist->bucket, &buckets, max_buckets);
    show_histogram(hist->bucket, buckets, hist->total, scale);
}

1496
1497
#define NELEMENTS(x) (sizeof(x)/sizeof(x[0]))
#define ARG_CTRL_CNT_MAX NELEMENTS(vp8_arg_ctrl_map)
John Koleszar's avatar
John Koleszar committed
1498
1499


1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
/* Configuration elements common to all streams */
struct global_config
{
    const struct codec_item  *codec;
    int                       passes;
    int                       pass;
    int                       usage;
    int                       deadline;
    int                       use_i420;
    int                       verbose;
    int                       limit;
    int                       show_psnr;
    int                       have_framerate;
    struct vpx_rational       framerate;
1514
    int                       out_part;
1515
1516
1517
1518
    int                       debug;
    int                       show_q_hist_buckets;
    int                       show_rate_hist_buckets;
};
John Koleszar's avatar
John Koleszar committed
1519
1520


1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
/* Per-stream configuration */
struct stream_config
{
    struct vpx_codec_enc_cfg  cfg;
    const char               *out_fn;
    const char               *stats_fn;
    stereo_format_t           stereo_fmt;
    int                       arg_ctrls[ARG_CTRL_CNT_MAX][2];
    int                       arg_ctrl_cnt;
    int                       write_webm;
1531
    int                       have_kf_max_dist;
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
};


struct stream_state
{
    int                       index;
    struct stream_state      *next;
    struct stream_config      config;
    FILE                     *file;
    struct rate_hist          rate_hist;
    EbmlGlobal                ebml;
    uint32_t                  hash;
    uint64_t                  psnr_sse_total;
    uint64_t                  psnr_samples_total;
    double                    psnr_totals[4];
    int                       psnr_count;
    int                       counts[64];
    vpx_codec_ctx_t           encoder;
    unsigned int              frames_out;
    uint64_t                  cx_time;
    size_t                    nbytes;
    stats_io_t                stats;
};


1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
void validate_positive_rational(const char          *msg,
                                struct vpx_rational *rat)
{
    if (rat->den < 0)
    {
        rat->num *= -1;
        rat->den *= -1;
    }

    if (rat->num < 0)
        die("Error: %s must be positive\n", msg);

    if (!rat->den)
        die("Error: %s has zero denominator\n", msg);
}


1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
static void parse_global_config(struct global_config *global, char **argv)
{
    char       **argi, **argj;
    struct arg   arg;

    /* Initialize default parameters */
    memset(global, 0, sizeof(*global));
    global->codec = codecs;
    global->passes = 1;
    global->use_i420 = 1;
John Koleszar's avatar
John Koleszar committed
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597

    for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step)
    {
        arg.argv_step = 1;

        if (arg_match(&arg, &codecarg, argi))
        {
            int j, k = -1;

            for (j = 0; j < sizeof(codecs) / sizeof(codecs[0]); j++)
                if (!strcmp(codecs[j].name, arg.val))
                    k = j;

            if (k >= 0)
1598
                global->codec = codecs + k;
John Koleszar's avatar
John Koleszar committed
1599
1600
1601
1602
1603
1604
1605
            else
                die("Error: Unrecognized argument (%s) to --codec\n",
                    arg.val);

        }
        else if (arg_match(&arg, &passes, argi))
        {
1606
            global->passes = arg_parse_uint(&arg);
John Koleszar's avatar
John Koleszar committed
1607

1608
1609
            if (global->passes < 1 || global->passes > 2)
                die("Error: Invalid number of passes (%d)\n", global->passes);
John Koleszar's avatar
John Koleszar committed
1610
1611
1612
        }
        else if (arg_match(&arg, &pass_arg, argi))
        {
1613
            global->pass = arg_parse_uint(&arg);
John Koleszar's avatar
John Koleszar committed
1614

1615
1616
1617
            if (global->pass < 1 || global->pass > 2)
                die("Error: Invalid pass selected (%d)\n",
                    global->pass);
John Koleszar's avatar
John Koleszar committed
1618
1619
        }
        else if (arg_match(&arg, &usage, argi))
1620
            global->usage = arg_parse_uint(&arg);
John Koleszar's avatar
John Koleszar committed
1621
        else if (arg_match(&arg, &deadline, argi))
1622
            global->deadline = arg_parse_uint(&arg);
John Koleszar's avatar
John Koleszar committed
1623
        else if (arg_match(&arg, &best_dl, argi))
1624
            global->deadline = VPX_DL_BEST_QUALITY;
John Koleszar's avatar
John Koleszar committed
1625
        else if (arg_match(&arg, &good_dl, argi))
1626
            global->deadline = VPX_DL_GOOD_QUALITY;
John Koleszar's avatar
John Koleszar committed
1627
        else if (arg_match(&arg, &rt_dl, argi))
1628
            global->deadline = VPX_DL_REALTIME;