Merge "Fix issue 262, vp8cx_pack_tokens_into_partitions_armv5"
[libvpx.git] / vpxenc.c
blob2c133e718e41213d6ce40a2ba3045148d131e796
1 /*
2 * Copyright (c) 2010 The WebM project authors. All Rights Reserved.
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 */
12 /* This is a simple program that encodes YV12 files and generates ivf
13 * files using the new interface.
15 #if defined(_WIN32)
16 #define USE_POSIX_MMAP 0
17 #else
18 #define USE_POSIX_MMAP 1
19 #endif
21 #include <stdio.h>
22 #include <stdlib.h>
23 #include <stdarg.h>
24 #include <string.h>
25 #include <limits.h>
26 #include "vpx/vpx_encoder.h"
27 #if USE_POSIX_MMAP
28 #include <sys/types.h>
29 #include <sys/stat.h>
30 #include <sys/mman.h>
31 #include <fcntl.h>
32 #include <unistd.h>
33 #endif
34 #include "vpx_version.h"
35 #include "vpx/vp8cx.h"
36 #include "vpx_ports/mem_ops.h"
37 #include "vpx_ports/vpx_timer.h"
38 #include "tools_common.h"
39 #include "y4minput.h"
40 #include "libmkv/EbmlWriter.h"
41 #include "libmkv/EbmlIDs.h"
43 /* Need special handling of these functions on Windows */
44 #if defined(_MSC_VER)
45 /* MSVS doesn't define off_t, and uses _f{seek,tell}i64 */
46 typedef __int64 off_t;
47 #define fseeko _fseeki64
48 #define ftello _ftelli64
49 #elif defined(_WIN32)
50 /* MinGW defines off_t, and uses f{seek,tell}o64 */
51 #define fseeko fseeko64
52 #define ftello ftello64
53 #endif
55 #if defined(_MSC_VER)
56 #define LITERALU64(n) n
57 #else
58 #define LITERALU64(n) n##LLU
59 #endif
61 static const char *exec_name;
63 static const struct codec_item
65 char const *name;
66 const vpx_codec_iface_t *iface;
67 unsigned int fourcc;
68 } codecs[] =
70 #if CONFIG_VP8_ENCODER
71 {"vp8", &vpx_codec_vp8_cx_algo, 0x30385056},
72 #endif
75 static void usage_exit();
77 void die(const char *fmt, ...)
79 va_list ap;
80 va_start(ap, fmt);
81 vfprintf(stderr, fmt, ap);
82 fprintf(stderr, "\n");
83 usage_exit();
86 static void ctx_exit_on_error(vpx_codec_ctx_t *ctx, const char *s)
88 if (ctx->err)
90 const char *detail = vpx_codec_error_detail(ctx);
92 fprintf(stderr, "%s: %s\n", s, vpx_codec_error(ctx));
94 if (detail)
95 fprintf(stderr, " %s\n", detail);
97 exit(EXIT_FAILURE);
101 /* This structure is used to abstract the different ways of handling
102 * first pass statistics.
104 typedef struct
106 vpx_fixed_buf_t buf;
107 int pass;
108 FILE *file;
109 char *buf_ptr;
110 size_t buf_alloc_sz;
111 } stats_io_t;
113 int stats_open_file(stats_io_t *stats, const char *fpf, int pass)
115 int res;
117 stats->pass = pass;
119 if (pass == 0)
121 stats->file = fopen(fpf, "wb");
122 stats->buf.sz = 0;
123 stats->buf.buf = NULL,
124 res = (stats->file != NULL);
126 else
128 #if 0
129 #elif USE_POSIX_MMAP
130 struct stat stat_buf;
131 int fd;
133 fd = open(fpf, O_RDONLY);
134 stats->file = fdopen(fd, "rb");
135 fstat(fd, &stat_buf);
136 stats->buf.sz = stat_buf.st_size;
137 stats->buf.buf = mmap(NULL, stats->buf.sz, PROT_READ, MAP_PRIVATE,
138 fd, 0);
139 res = (stats->buf.buf != NULL);
140 #else
141 size_t nbytes;
143 stats->file = fopen(fpf, "rb");
145 if (fseek(stats->file, 0, SEEK_END))
147 fprintf(stderr, "First-pass stats file must be seekable!\n");
148 exit(EXIT_FAILURE);
151 stats->buf.sz = stats->buf_alloc_sz = ftell(stats->file);
152 rewind(stats->file);
154 stats->buf.buf = malloc(stats->buf_alloc_sz);
156 if (!stats->buf.buf)
158 fprintf(stderr, "Failed to allocate first-pass stats buffer (%d bytes)\n",
159 stats->buf_alloc_sz);
160 exit(EXIT_FAILURE);
163 nbytes = fread(stats->buf.buf, 1, stats->buf.sz, stats->file);
164 res = (nbytes == stats->buf.sz);
165 #endif
168 return res;
171 int stats_open_mem(stats_io_t *stats, int pass)
173 int res;
174 stats->pass = pass;
176 if (!pass)
178 stats->buf.sz = 0;
179 stats->buf_alloc_sz = 64 * 1024;
180 stats->buf.buf = malloc(stats->buf_alloc_sz);
183 stats->buf_ptr = stats->buf.buf;
184 res = (stats->buf.buf != NULL);
185 return res;
189 void stats_close(stats_io_t *stats, int last_pass)
191 if (stats->file)
193 if (stats->pass == last_pass)
195 #if 0
196 #elif USE_POSIX_MMAP
197 munmap(stats->buf.buf, stats->buf.sz);
198 #else
199 free(stats->buf.buf);
200 #endif
203 fclose(stats->file);
204 stats->file = NULL;
206 else
208 if (stats->pass == last_pass)
209 free(stats->buf.buf);
213 void stats_write(stats_io_t *stats, const void *pkt, size_t len)
215 if (stats->file)
217 if(fwrite(pkt, 1, len, stats->file));
219 else
221 if (stats->buf.sz + len > stats->buf_alloc_sz)
223 size_t new_sz = stats->buf_alloc_sz + 64 * 1024;
224 char *new_ptr = realloc(stats->buf.buf, new_sz);
226 if (new_ptr)
228 stats->buf_ptr = new_ptr + (stats->buf_ptr - (char *)stats->buf.buf);
229 stats->buf.buf = new_ptr;
230 stats->buf_alloc_sz = new_sz;
231 } /* else ... */
234 memcpy(stats->buf_ptr, pkt, len);
235 stats->buf.sz += len;
236 stats->buf_ptr += len;
240 vpx_fixed_buf_t stats_get(stats_io_t *stats)
242 return stats->buf;
245 enum video_file_type
247 FILE_TYPE_RAW,
248 FILE_TYPE_IVF,
249 FILE_TYPE_Y4M
252 struct detect_buffer {
253 char buf[4];
254 size_t buf_read;
255 size_t position;
259 #define IVF_FRAME_HDR_SZ (4+8) /* 4 byte size + 8 byte timestamp */
260 static int read_frame(FILE *f, vpx_image_t *img, unsigned int file_type,
261 y4m_input *y4m, struct detect_buffer *detect)
263 int plane = 0;
264 int shortread = 0;
266 if (file_type == FILE_TYPE_Y4M)
268 if (y4m_input_fetch_frame(y4m, f, img) < 1)
269 return 0;
271 else
273 if (file_type == FILE_TYPE_IVF)
275 char junk[IVF_FRAME_HDR_SZ];
277 /* Skip the frame header. We know how big the frame should be. See
278 * write_ivf_frame_header() for documentation on the frame header
279 * layout.
281 if(fread(junk, 1, IVF_FRAME_HDR_SZ, f));
284 for (plane = 0; plane < 3; plane++)
286 unsigned char *ptr;
287 int w = (plane ? (1 + img->d_w) / 2 : img->d_w);
288 int h = (plane ? (1 + img->d_h) / 2 : img->d_h);
289 int r;
291 /* Determine the correct plane based on the image format. The for-loop
292 * always counts in Y,U,V order, but this may not match the order of
293 * the data on disk.
295 switch (plane)
297 case 1:
298 ptr = img->planes[img->fmt==VPX_IMG_FMT_YV12? VPX_PLANE_V : VPX_PLANE_U];
299 break;
300 case 2:
301 ptr = img->planes[img->fmt==VPX_IMG_FMT_YV12?VPX_PLANE_U : VPX_PLANE_V];
302 break;
303 default:
304 ptr = img->planes[plane];
307 for (r = 0; r < h; r++)
309 size_t needed = w;
310 size_t buf_position = 0;
311 const size_t left = detect->buf_read - detect->position;
312 if (left > 0)
314 const size_t more = (left < needed) ? left : needed;
315 memcpy(ptr, detect->buf + detect->position, more);
316 buf_position = more;
317 needed -= more;
318 detect->position += more;
320 if (needed > 0)
322 shortread |= (fread(ptr + buf_position, 1, needed, f) < needed);
325 ptr += img->stride[plane];
330 return !shortread;
334 unsigned int file_is_y4m(FILE *infile,
335 y4m_input *y4m,
336 char detect[4])
338 if(memcmp(detect, "YUV4", 4) == 0)
340 return 1;
342 return 0;
345 #define IVF_FILE_HDR_SZ (32)
346 unsigned int file_is_ivf(FILE *infile,
347 unsigned int *fourcc,
348 unsigned int *width,
349 unsigned int *height,
350 struct detect_buffer *detect)
352 char raw_hdr[IVF_FILE_HDR_SZ];
353 int is_ivf = 0;
355 if(memcmp(detect->buf, "DKIF", 4) != 0)
356 return 0;
358 /* See write_ivf_file_header() for more documentation on the file header
359 * layout.
361 if (fread(raw_hdr + 4, 1, IVF_FILE_HDR_SZ - 4, infile)
362 == IVF_FILE_HDR_SZ - 4)
365 is_ivf = 1;
367 if (mem_get_le16(raw_hdr + 4) != 0)
368 fprintf(stderr, "Error: Unrecognized IVF version! This file may not"
369 " decode properly.");
371 *fourcc = mem_get_le32(raw_hdr + 8);
375 if (is_ivf)
377 *width = mem_get_le16(raw_hdr + 12);
378 *height = mem_get_le16(raw_hdr + 14);
379 detect->position = 4;
382 return is_ivf;
386 static void write_ivf_file_header(FILE *outfile,
387 const vpx_codec_enc_cfg_t *cfg,
388 unsigned int fourcc,
389 int frame_cnt)
391 char header[32];
393 if (cfg->g_pass != VPX_RC_ONE_PASS && cfg->g_pass != VPX_RC_LAST_PASS)
394 return;
396 header[0] = 'D';
397 header[1] = 'K';
398 header[2] = 'I';
399 header[3] = 'F';
400 mem_put_le16(header + 4, 0); /* version */
401 mem_put_le16(header + 6, 32); /* headersize */
402 mem_put_le32(header + 8, fourcc); /* headersize */
403 mem_put_le16(header + 12, cfg->g_w); /* width */
404 mem_put_le16(header + 14, cfg->g_h); /* height */
405 mem_put_le32(header + 16, cfg->g_timebase.den); /* rate */
406 mem_put_le32(header + 20, cfg->g_timebase.num); /* scale */
407 mem_put_le32(header + 24, frame_cnt); /* length */
408 mem_put_le32(header + 28, 0); /* unused */
410 if(fwrite(header, 1, 32, outfile));
414 static void write_ivf_frame_header(FILE *outfile,
415 const vpx_codec_cx_pkt_t *pkt)
417 char header[12];
418 vpx_codec_pts_t pts;
420 if (pkt->kind != VPX_CODEC_CX_FRAME_PKT)
421 return;
423 pts = pkt->data.frame.pts;
424 mem_put_le32(header, pkt->data.frame.sz);
425 mem_put_le32(header + 4, pts & 0xFFFFFFFF);
426 mem_put_le32(header + 8, pts >> 32);
428 if(fwrite(header, 1, 12, outfile));
432 typedef off_t EbmlLoc;
435 struct cue_entry
437 unsigned int time;
438 uint64_t loc;
442 struct EbmlGlobal
444 int debug;
446 FILE *stream;
447 int64_t last_pts_ms;
448 vpx_rational_t framerate;
450 /* These pointers are to the start of an element */
451 off_t position_reference;
452 off_t seek_info_pos;
453 off_t segment_info_pos;
454 off_t track_pos;
455 off_t cue_pos;
456 off_t cluster_pos;
458 /* This pointer is to a specific element to be serialized */
459 off_t track_id_pos;
461 /* These pointers are to the size field of the element */
462 EbmlLoc startSegment;
463 EbmlLoc startCluster;
465 uint32_t cluster_timecode;
466 int cluster_open;
468 struct cue_entry *cue_list;
469 unsigned int cues;
474 void Ebml_Write(EbmlGlobal *glob, const void *buffer_in, unsigned long len)
476 if(fwrite(buffer_in, 1, len, glob->stream));
480 void Ebml_Serialize(EbmlGlobal *glob, const void *buffer_in, unsigned long len)
482 const unsigned char *q = (const unsigned char *)buffer_in + len - 1;
484 for(; len; len--)
485 Ebml_Write(glob, q--, 1);
489 /* Need a fixed size serializer for the track ID. libmkv provdes a 64 bit
490 * one, but not a 32 bit one.
492 static void Ebml_SerializeUnsigned32(EbmlGlobal *glob, unsigned long class_id, uint64_t ui)
494 unsigned char sizeSerialized = 4 | 0x80;
495 Ebml_WriteID(glob, class_id);
496 Ebml_Serialize(glob, &sizeSerialized, 1);
497 Ebml_Serialize(glob, &ui, 4);
501 static void
502 Ebml_StartSubElement(EbmlGlobal *glob, EbmlLoc *ebmlLoc,
503 unsigned long class_id)
505 //todo this is always taking 8 bytes, this may need later optimization
506 //this is a key that says lenght unknown
507 unsigned long long unknownLen = LITERALU64(0x01FFFFFFFFFFFFFF);
509 Ebml_WriteID(glob, class_id);
510 *ebmlLoc = ftello(glob->stream);
511 Ebml_Serialize(glob, &unknownLen, 8);
514 static void
515 Ebml_EndSubElement(EbmlGlobal *glob, EbmlLoc *ebmlLoc)
517 off_t pos;
518 uint64_t size;
520 /* Save the current stream pointer */
521 pos = ftello(glob->stream);
523 /* Calculate the size of this element */
524 size = pos - *ebmlLoc - 8;
525 size |= LITERALU64(0x0100000000000000);
527 /* Seek back to the beginning of the element and write the new size */
528 fseeko(glob->stream, *ebmlLoc, SEEK_SET);
529 Ebml_Serialize(glob, &size, 8);
531 /* Reset the stream pointer */
532 fseeko(glob->stream, pos, SEEK_SET);
536 static void
537 write_webm_seek_element(EbmlGlobal *ebml, unsigned long id, off_t pos)
539 uint64_t offset = pos - ebml->position_reference;
540 EbmlLoc start;
541 Ebml_StartSubElement(ebml, &start, Seek);
542 Ebml_SerializeBinary(ebml, SeekID, id);
543 Ebml_SerializeUnsigned64(ebml, SeekPosition, offset);
544 Ebml_EndSubElement(ebml, &start);
548 static void
549 write_webm_seek_info(EbmlGlobal *ebml)
552 off_t pos;
554 /* Save the current stream pointer */
555 pos = ftello(ebml->stream);
557 if(ebml->seek_info_pos)
558 fseeko(ebml->stream, ebml->seek_info_pos, SEEK_SET);
559 else
560 ebml->seek_info_pos = pos;
563 EbmlLoc start;
565 Ebml_StartSubElement(ebml, &start, SeekHead);
566 write_webm_seek_element(ebml, Tracks, ebml->track_pos);
567 write_webm_seek_element(ebml, Cues, ebml->cue_pos);
568 write_webm_seek_element(ebml, Info, ebml->segment_info_pos);
569 Ebml_EndSubElement(ebml, &start);
572 //segment info
573 EbmlLoc startInfo;
574 uint64_t frame_time;
576 frame_time = (uint64_t)1000 * ebml->framerate.den
577 / ebml->framerate.num;
578 ebml->segment_info_pos = ftello(ebml->stream);
579 Ebml_StartSubElement(ebml, &startInfo, Info);
580 Ebml_SerializeUnsigned(ebml, TimecodeScale, 1000000);
581 Ebml_SerializeFloat(ebml, Segment_Duration,
582 ebml->last_pts_ms + frame_time);
583 Ebml_SerializeString(ebml, 0x4D80,
584 ebml->debug ? "vpxenc" : "vpxenc" VERSION_STRING);
585 Ebml_SerializeString(ebml, 0x5741,
586 ebml->debug ? "vpxenc" : "vpxenc" VERSION_STRING);
587 Ebml_EndSubElement(ebml, &startInfo);
592 static void
593 write_webm_file_header(EbmlGlobal *glob,
594 const vpx_codec_enc_cfg_t *cfg,
595 const struct vpx_rational *fps)
598 EbmlLoc start;
599 Ebml_StartSubElement(glob, &start, EBML);
600 Ebml_SerializeUnsigned(glob, EBMLVersion, 1);
601 Ebml_SerializeUnsigned(glob, EBMLReadVersion, 1); //EBML Read Version
602 Ebml_SerializeUnsigned(glob, EBMLMaxIDLength, 4); //EBML Max ID Length
603 Ebml_SerializeUnsigned(glob, EBMLMaxSizeLength, 8); //EBML Max Size Length
604 Ebml_SerializeString(glob, DocType, "webm"); //Doc Type
605 Ebml_SerializeUnsigned(glob, DocTypeVersion, 2); //Doc Type Version
606 Ebml_SerializeUnsigned(glob, DocTypeReadVersion, 2); //Doc Type Read Version
607 Ebml_EndSubElement(glob, &start);
610 Ebml_StartSubElement(glob, &glob->startSegment, Segment); //segment
611 glob->position_reference = ftello(glob->stream);
612 glob->framerate = *fps;
613 write_webm_seek_info(glob);
616 EbmlLoc trackStart;
617 glob->track_pos = ftello(glob->stream);
618 Ebml_StartSubElement(glob, &trackStart, Tracks);
620 unsigned int trackNumber = 1;
621 uint64_t trackID = 0;
623 EbmlLoc start;
624 Ebml_StartSubElement(glob, &start, TrackEntry);
625 Ebml_SerializeUnsigned(glob, TrackNumber, trackNumber);
626 glob->track_id_pos = ftello(glob->stream);
627 Ebml_SerializeUnsigned32(glob, TrackUID, trackID);
628 Ebml_SerializeUnsigned(glob, TrackType, 1); //video is always 1
629 Ebml_SerializeString(glob, CodecID, "V_VP8");
631 unsigned int pixelWidth = cfg->g_w;
632 unsigned int pixelHeight = cfg->g_h;
633 float frameRate = (float)fps->num/(float)fps->den;
635 EbmlLoc videoStart;
636 Ebml_StartSubElement(glob, &videoStart, Video);
637 Ebml_SerializeUnsigned(glob, PixelWidth, pixelWidth);
638 Ebml_SerializeUnsigned(glob, PixelHeight, pixelHeight);
639 Ebml_SerializeFloat(glob, FrameRate, frameRate);
640 Ebml_EndSubElement(glob, &videoStart); //Video
642 Ebml_EndSubElement(glob, &start); //Track Entry
644 Ebml_EndSubElement(glob, &trackStart);
646 // segment element is open
651 static void
652 write_webm_block(EbmlGlobal *glob,
653 const vpx_codec_enc_cfg_t *cfg,
654 const vpx_codec_cx_pkt_t *pkt)
656 unsigned long block_length;
657 unsigned char track_number;
658 unsigned short block_timecode = 0;
659 unsigned char flags;
660 int64_t pts_ms;
661 int start_cluster = 0, is_keyframe;
663 /* Calculate the PTS of this frame in milliseconds */
664 pts_ms = pkt->data.frame.pts * 1000
665 * (uint64_t)cfg->g_timebase.num / (uint64_t)cfg->g_timebase.den;
666 if(pts_ms <= glob->last_pts_ms)
667 pts_ms = glob->last_pts_ms + 1;
668 glob->last_pts_ms = pts_ms;
670 /* Calculate the relative time of this block */
671 if(pts_ms - glob->cluster_timecode > SHRT_MAX)
672 start_cluster = 1;
673 else
674 block_timecode = pts_ms - glob->cluster_timecode;
676 is_keyframe = (pkt->data.frame.flags & VPX_FRAME_IS_KEY);
677 if(start_cluster || is_keyframe)
679 if(glob->cluster_open)
680 Ebml_EndSubElement(glob, &glob->startCluster);
682 /* Open the new cluster */
683 block_timecode = 0;
684 glob->cluster_open = 1;
685 glob->cluster_timecode = pts_ms;
686 glob->cluster_pos = ftello(glob->stream);
687 Ebml_StartSubElement(glob, &glob->startCluster, Cluster); //cluster
688 Ebml_SerializeUnsigned(glob, Timecode, glob->cluster_timecode);
690 /* Save a cue point if this is a keyframe. */
691 if(is_keyframe)
693 struct cue_entry *cue;
695 glob->cue_list = realloc(glob->cue_list,
696 (glob->cues+1) * sizeof(struct cue_entry));
697 cue = &glob->cue_list[glob->cues];
698 cue->time = glob->cluster_timecode;
699 cue->loc = glob->cluster_pos;
700 glob->cues++;
704 /* Write the Simple Block */
705 Ebml_WriteID(glob, SimpleBlock);
707 block_length = pkt->data.frame.sz + 4;
708 block_length |= 0x10000000;
709 Ebml_Serialize(glob, &block_length, 4);
711 track_number = 1;
712 track_number |= 0x80;
713 Ebml_Write(glob, &track_number, 1);
715 Ebml_Serialize(glob, &block_timecode, 2);
717 flags = 0;
718 if(is_keyframe)
719 flags |= 0x80;
720 if(pkt->data.frame.flags & VPX_FRAME_IS_INVISIBLE)
721 flags |= 0x08;
722 Ebml_Write(glob, &flags, 1);
724 Ebml_Write(glob, pkt->data.frame.buf, pkt->data.frame.sz);
728 static void
729 write_webm_file_footer(EbmlGlobal *glob, long hash)
732 if(glob->cluster_open)
733 Ebml_EndSubElement(glob, &glob->startCluster);
736 EbmlLoc start;
737 int i;
739 glob->cue_pos = ftello(glob->stream);
740 Ebml_StartSubElement(glob, &start, Cues);
741 for(i=0; i<glob->cues; i++)
743 struct cue_entry *cue = &glob->cue_list[i];
744 EbmlLoc start;
746 Ebml_StartSubElement(glob, &start, CuePoint);
748 EbmlLoc start;
750 Ebml_SerializeUnsigned(glob, CueTime, cue->time);
752 Ebml_StartSubElement(glob, &start, CueTrackPositions);
753 Ebml_SerializeUnsigned(glob, CueTrack, 1);
754 Ebml_SerializeUnsigned64(glob, CueClusterPosition,
755 cue->loc - glob->position_reference);
756 //Ebml_SerializeUnsigned(glob, CueBlockNumber, cue->blockNumber);
757 Ebml_EndSubElement(glob, &start);
759 Ebml_EndSubElement(glob, &start);
761 Ebml_EndSubElement(glob, &start);
764 Ebml_EndSubElement(glob, &glob->startSegment);
766 /* Patch up the seek info block */
767 write_webm_seek_info(glob);
769 /* Patch up the track id */
770 fseeko(glob->stream, glob->track_id_pos, SEEK_SET);
771 Ebml_SerializeUnsigned32(glob, TrackUID, glob->debug ? 0xDEADBEEF : hash);
773 fseeko(glob->stream, 0, SEEK_END);
777 /* Murmur hash derived from public domain reference implementation at
778 * http://sites.google.com/site/murmurhash/
780 static unsigned int murmur ( const void * key, int len, unsigned int seed )
782 const unsigned int m = 0x5bd1e995;
783 const int r = 24;
785 unsigned int h = seed ^ len;
787 const unsigned char * data = (const unsigned char *)key;
789 while(len >= 4)
791 unsigned int k;
793 k = data[0];
794 k |= data[1] << 8;
795 k |= data[2] << 16;
796 k |= data[3] << 24;
798 k *= m;
799 k ^= k >> r;
800 k *= m;
802 h *= m;
803 h ^= k;
805 data += 4;
806 len -= 4;
809 switch(len)
811 case 3: h ^= data[2] << 16;
812 case 2: h ^= data[1] << 8;
813 case 1: h ^= data[0];
814 h *= m;
817 h ^= h >> 13;
818 h *= m;
819 h ^= h >> 15;
821 return h;
824 #include "math.h"
826 static double vp8_mse2psnr(double Samples, double Peak, double Mse)
828 double psnr;
830 if ((double)Mse > 0.0)
831 psnr = 10.0 * log10(Peak * Peak * Samples / Mse);
832 else
833 psnr = 60; // Limit to prevent / 0
835 if (psnr > 60)
836 psnr = 60;
838 return psnr;
842 #include "args.h"
844 static const arg_def_t debugmode = ARG_DEF("D", "debug", 0,
845 "Debug mode (makes output deterministic)");
846 static const arg_def_t outputfile = ARG_DEF("o", "output", 1,
847 "Output filename");
848 static const arg_def_t use_yv12 = ARG_DEF(NULL, "yv12", 0,
849 "Input file is YV12 ");
850 static const arg_def_t use_i420 = ARG_DEF(NULL, "i420", 0,
851 "Input file is I420 (default)");
852 static const arg_def_t codecarg = ARG_DEF(NULL, "codec", 1,
853 "Codec to use");
854 static const arg_def_t passes = ARG_DEF("p", "passes", 1,
855 "Number of passes (1/2)");
856 static const arg_def_t pass_arg = ARG_DEF(NULL, "pass", 1,
857 "Pass to execute (1/2)");
858 static const arg_def_t fpf_name = ARG_DEF(NULL, "fpf", 1,
859 "First pass statistics file name");
860 static const arg_def_t limit = ARG_DEF(NULL, "limit", 1,
861 "Stop encoding after n input frames");
862 static const arg_def_t deadline = ARG_DEF("d", "deadline", 1,
863 "Deadline per frame (usec)");
864 static const arg_def_t best_dl = ARG_DEF(NULL, "best", 0,
865 "Use Best Quality Deadline");
866 static const arg_def_t good_dl = ARG_DEF(NULL, "good", 0,
867 "Use Good Quality Deadline");
868 static const arg_def_t rt_dl = ARG_DEF(NULL, "rt", 0,
869 "Use Realtime Quality Deadline");
870 static const arg_def_t verbosearg = ARG_DEF("v", "verbose", 0,
871 "Show encoder parameters");
872 static const arg_def_t psnrarg = ARG_DEF(NULL, "psnr", 0,
873 "Show PSNR in status line");
874 static const arg_def_t framerate = ARG_DEF(NULL, "fps", 1,
875 "Stream frame rate (rate/scale)");
876 static const arg_def_t use_ivf = ARG_DEF(NULL, "ivf", 0,
877 "Output IVF (default is WebM)");
878 static const arg_def_t *main_args[] =
880 &debugmode,
881 &outputfile, &codecarg, &passes, &pass_arg, &fpf_name, &limit, &deadline,
882 &best_dl, &good_dl, &rt_dl,
883 &verbosearg, &psnrarg, &use_ivf, &framerate,
884 NULL
887 static const arg_def_t usage = ARG_DEF("u", "usage", 1,
888 "Usage profile number to use");
889 static const arg_def_t threads = ARG_DEF("t", "threads", 1,
890 "Max number of threads to use");
891 static const arg_def_t profile = ARG_DEF(NULL, "profile", 1,
892 "Bitstream profile number to use");
893 static const arg_def_t width = ARG_DEF("w", "width", 1,
894 "Frame width");
895 static const arg_def_t height = ARG_DEF("h", "height", 1,
896 "Frame height");
897 static const arg_def_t timebase = ARG_DEF(NULL, "timebase", 1,
898 "Stream timebase (frame duration)");
899 static const arg_def_t error_resilient = ARG_DEF(NULL, "error-resilient", 1,
900 "Enable error resiliency features");
901 static const arg_def_t lag_in_frames = ARG_DEF(NULL, "lag-in-frames", 1,
902 "Max number of frames to lag");
904 static const arg_def_t *global_args[] =
906 &use_yv12, &use_i420, &usage, &threads, &profile,
907 &width, &height, &timebase, &framerate, &error_resilient,
908 &lag_in_frames, NULL
911 static const arg_def_t dropframe_thresh = ARG_DEF(NULL, "drop-frame", 1,
912 "Temporal resampling threshold (buf %)");
913 static const arg_def_t resize_allowed = ARG_DEF(NULL, "resize-allowed", 1,
914 "Spatial resampling enabled (bool)");
915 static const arg_def_t resize_up_thresh = ARG_DEF(NULL, "resize-up", 1,
916 "Upscale threshold (buf %)");
917 static const arg_def_t resize_down_thresh = ARG_DEF(NULL, "resize-down", 1,
918 "Downscale threshold (buf %)");
919 static const arg_def_t end_usage = ARG_DEF(NULL, "end-usage", 1,
920 "VBR=0 | CBR=1 | CQ=2");
921 static const arg_def_t target_bitrate = ARG_DEF(NULL, "target-bitrate", 1,
922 "Bitrate (kbps)");
923 static const arg_def_t min_quantizer = ARG_DEF(NULL, "min-q", 1,
924 "Minimum (best) quantizer");
925 static const arg_def_t max_quantizer = ARG_DEF(NULL, "max-q", 1,
926 "Maximum (worst) quantizer");
927 static const arg_def_t undershoot_pct = ARG_DEF(NULL, "undershoot-pct", 1,
928 "Datarate undershoot (min) target (%)");
929 static const arg_def_t overshoot_pct = ARG_DEF(NULL, "overshoot-pct", 1,
930 "Datarate overshoot (max) target (%)");
931 static const arg_def_t buf_sz = ARG_DEF(NULL, "buf-sz", 1,
932 "Client buffer size (ms)");
933 static const arg_def_t buf_initial_sz = ARG_DEF(NULL, "buf-initial-sz", 1,
934 "Client initial buffer size (ms)");
935 static const arg_def_t buf_optimal_sz = ARG_DEF(NULL, "buf-optimal-sz", 1,
936 "Client optimal buffer size (ms)");
937 static const arg_def_t *rc_args[] =
939 &dropframe_thresh, &resize_allowed, &resize_up_thresh, &resize_down_thresh,
940 &end_usage, &target_bitrate, &min_quantizer, &max_quantizer,
941 &undershoot_pct, &overshoot_pct, &buf_sz, &buf_initial_sz, &buf_optimal_sz,
942 NULL
946 static const arg_def_t bias_pct = ARG_DEF(NULL, "bias-pct", 1,
947 "CBR/VBR bias (0=CBR, 100=VBR)");
948 static const arg_def_t minsection_pct = ARG_DEF(NULL, "minsection-pct", 1,
949 "GOP min bitrate (% of target)");
950 static const arg_def_t maxsection_pct = ARG_DEF(NULL, "maxsection-pct", 1,
951 "GOP max bitrate (% of target)");
952 static const arg_def_t *rc_twopass_args[] =
954 &bias_pct, &minsection_pct, &maxsection_pct, NULL
958 static const arg_def_t kf_min_dist = ARG_DEF(NULL, "kf-min-dist", 1,
959 "Minimum keyframe interval (frames)");
960 static const arg_def_t kf_max_dist = ARG_DEF(NULL, "kf-max-dist", 1,
961 "Maximum keyframe interval (frames)");
962 static const arg_def_t kf_disabled = ARG_DEF(NULL, "disable-kf", 0,
963 "Disable keyframe placement");
964 static const arg_def_t *kf_args[] =
966 &kf_min_dist, &kf_max_dist, &kf_disabled, NULL
970 #if CONFIG_VP8_ENCODER
971 static const arg_def_t noise_sens = ARG_DEF(NULL, "noise-sensitivity", 1,
972 "Noise sensitivity (frames to blur)");
973 static const arg_def_t sharpness = ARG_DEF(NULL, "sharpness", 1,
974 "Filter sharpness (0-7)");
975 static const arg_def_t static_thresh = ARG_DEF(NULL, "static-thresh", 1,
976 "Motion detection threshold");
977 #endif
979 #if CONFIG_VP8_ENCODER
980 static const arg_def_t cpu_used = ARG_DEF(NULL, "cpu-used", 1,
981 "CPU Used (-16..16)");
982 #endif
985 #if CONFIG_VP8_ENCODER
986 static const arg_def_t token_parts = ARG_DEF(NULL, "token-parts", 1,
987 "Number of token partitions to use, log2");
988 static const arg_def_t auto_altref = ARG_DEF(NULL, "auto-alt-ref", 1,
989 "Enable automatic alt reference frames");
990 static const arg_def_t arnr_maxframes = ARG_DEF(NULL, "arnr-maxframes", 1,
991 "AltRef Max Frames");
992 static const arg_def_t arnr_strength = ARG_DEF(NULL, "arnr-strength", 1,
993 "AltRef Strength");
994 static const arg_def_t arnr_type = ARG_DEF(NULL, "arnr-type", 1,
995 "AltRef Type");
996 static const struct arg_enum_list tuning_enum[] = {
997 {"psnr", VP8_TUNE_PSNR},
998 {"ssim", VP8_TUNE_SSIM},
999 {NULL, 0}
1001 static const arg_def_t tune_ssim = ARG_DEF_ENUM(NULL, "tune", 1,
1002 "Material to favor", tuning_enum);
1003 static const arg_def_t cq_level = ARG_DEF(NULL, "cq-level", 1,
1004 "Constrained Quality Level");
1006 static const arg_def_t *vp8_args[] =
1008 &cpu_used, &auto_altref, &noise_sens, &sharpness, &static_thresh,
1009 &token_parts, &arnr_maxframes, &arnr_strength, &arnr_type,
1010 &tune_ssim, &cq_level, NULL
1012 static const int vp8_arg_ctrl_map[] =
1014 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
1015 VP8E_SET_NOISE_SENSITIVITY, VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
1016 VP8E_SET_TOKEN_PARTITIONS,
1017 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH , VP8E_SET_ARNR_TYPE,
1018 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, 0
1020 #endif
1022 static const arg_def_t *no_args[] = { NULL };
1024 static void usage_exit()
1026 int i;
1028 fprintf(stderr, "Usage: %s <options> -o dst_filename src_filename \n",
1029 exec_name);
1031 fprintf(stderr, "\nOptions:\n");
1032 arg_show_usage(stdout, main_args);
1033 fprintf(stderr, "\nEncoder Global Options:\n");
1034 arg_show_usage(stdout, global_args);
1035 fprintf(stderr, "\nRate Control Options:\n");
1036 arg_show_usage(stdout, rc_args);
1037 fprintf(stderr, "\nTwopass Rate Control Options:\n");
1038 arg_show_usage(stdout, rc_twopass_args);
1039 fprintf(stderr, "\nKeyframe Placement Options:\n");
1040 arg_show_usage(stdout, kf_args);
1041 #if CONFIG_VP8_ENCODER
1042 fprintf(stderr, "\nVP8 Specific Options:\n");
1043 arg_show_usage(stdout, vp8_args);
1044 #endif
1045 fprintf(stderr, "\n"
1046 "Included encoders:\n"
1047 "\n");
1049 for (i = 0; i < sizeof(codecs) / sizeof(codecs[0]); i++)
1050 fprintf(stderr, " %-6s - %s\n",
1051 codecs[i].name,
1052 vpx_codec_iface_name(codecs[i].iface));
1054 exit(EXIT_FAILURE);
1057 #define ARG_CTRL_CNT_MAX 10
1060 int main(int argc, const char **argv_)
1062 vpx_codec_ctx_t encoder;
1063 const char *in_fn = NULL, *out_fn = NULL, *stats_fn = NULL;
1064 int i;
1065 FILE *infile, *outfile;
1066 vpx_codec_enc_cfg_t cfg;
1067 vpx_codec_err_t res;
1068 int pass, one_pass_only = 0;
1069 stats_io_t stats;
1070 vpx_image_t raw;
1071 const struct codec_item *codec = codecs;
1072 int frame_avail, got_data;
1074 struct arg arg;
1075 char **argv, **argi, **argj;
1076 int arg_usage = 0, arg_passes = 1, arg_deadline = 0;
1077 int arg_ctrls[ARG_CTRL_CNT_MAX][2], arg_ctrl_cnt = 0;
1078 int arg_limit = 0;
1079 static const arg_def_t **ctrl_args = no_args;
1080 static const int *ctrl_args_map = NULL;
1081 int verbose = 0, show_psnr = 0;
1082 int arg_use_i420 = 1;
1083 unsigned long cx_time = 0;
1084 unsigned int file_type, fourcc;
1085 y4m_input y4m;
1086 struct vpx_rational arg_framerate = {30, 1};
1087 int arg_have_framerate = 0;
1088 int write_webm = 1;
1089 EbmlGlobal ebml = {0};
1090 uint32_t hash = 0;
1091 uint64_t psnr_sse_total = 0;
1092 uint64_t psnr_samples_total = 0;
1093 double psnr_totals[4] = {0, 0, 0, 0};
1094 int psnr_count = 0;
1096 exec_name = argv_[0];
1097 ebml.last_pts_ms = -1;
1099 if (argc < 3)
1100 usage_exit();
1103 /* First parse the codec and usage values, because we want to apply other
1104 * parameters on top of the default configuration provided by the codec.
1106 argv = argv_dup(argc - 1, argv_ + 1);
1108 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step)
1110 arg.argv_step = 1;
1112 if (arg_match(&arg, &codecarg, argi))
1114 int j, k = -1;
1116 for (j = 0; j < sizeof(codecs) / sizeof(codecs[0]); j++)
1117 if (!strcmp(codecs[j].name, arg.val))
1118 k = j;
1120 if (k >= 0)
1121 codec = codecs + k;
1122 else
1123 die("Error: Unrecognized argument (%s) to --codec\n",
1124 arg.val);
1127 else if (arg_match(&arg, &passes, argi))
1129 arg_passes = arg_parse_uint(&arg);
1131 if (arg_passes < 1 || arg_passes > 2)
1132 die("Error: Invalid number of passes (%d)\n", arg_passes);
1134 else if (arg_match(&arg, &pass_arg, argi))
1136 one_pass_only = arg_parse_uint(&arg);
1138 if (one_pass_only < 1 || one_pass_only > 2)
1139 die("Error: Invalid pass selected (%d)\n", one_pass_only);
1141 else if (arg_match(&arg, &fpf_name, argi))
1142 stats_fn = arg.val;
1143 else if (arg_match(&arg, &usage, argi))
1144 arg_usage = arg_parse_uint(&arg);
1145 else if (arg_match(&arg, &deadline, argi))
1146 arg_deadline = arg_parse_uint(&arg);
1147 else if (arg_match(&arg, &best_dl, argi))
1148 arg_deadline = VPX_DL_BEST_QUALITY;
1149 else if (arg_match(&arg, &good_dl, argi))
1150 arg_deadline = VPX_DL_GOOD_QUALITY;
1151 else if (arg_match(&arg, &rt_dl, argi))
1152 arg_deadline = VPX_DL_REALTIME;
1153 else if (arg_match(&arg, &use_yv12, argi))
1155 arg_use_i420 = 0;
1157 else if (arg_match(&arg, &use_i420, argi))
1159 arg_use_i420 = 1;
1161 else if (arg_match(&arg, &verbosearg, argi))
1162 verbose = 1;
1163 else if (arg_match(&arg, &limit, argi))
1164 arg_limit = arg_parse_uint(&arg);
1165 else if (arg_match(&arg, &psnrarg, argi))
1166 show_psnr = 1;
1167 else if (arg_match(&arg, &framerate, argi))
1169 arg_framerate = arg_parse_rational(&arg);
1170 arg_have_framerate = 1;
1172 else if (arg_match(&arg, &use_ivf, argi))
1173 write_webm = 0;
1174 else if (arg_match(&arg, &outputfile, argi))
1175 out_fn = arg.val;
1176 else if (arg_match(&arg, &debugmode, argi))
1177 ebml.debug = 1;
1178 else
1179 argj++;
1182 /* Ensure that --passes and --pass are consistent. If --pass is set and --passes=2,
1183 * ensure --fpf was set.
1185 if (one_pass_only)
1187 /* DWIM: Assume the user meant passes=2 if pass=2 is specified */
1188 if (one_pass_only > arg_passes)
1190 fprintf(stderr, "Warning: Assuming --pass=%d implies --passes=%d\n",
1191 one_pass_only, one_pass_only);
1192 arg_passes = one_pass_only;
1195 if (arg_passes == 2 && !stats_fn)
1196 die("Must specify --fpf when --pass=%d and --passes=2\n", one_pass_only);
1199 /* Populate encoder configuration */
1200 res = vpx_codec_enc_config_default(codec->iface, &cfg, arg_usage);
1202 if (res)
1204 fprintf(stderr, "Failed to get config: %s\n",
1205 vpx_codec_err_to_string(res));
1206 return EXIT_FAILURE;
1209 /* Change the default timebase to a high enough value so that the encoder
1210 * will always create strictly increasing timestamps.
1212 cfg.g_timebase.den = 1000;
1214 /* Never use the library's default resolution, require it be parsed
1215 * from the file or set on the command line.
1217 cfg.g_w = 0;
1218 cfg.g_h = 0;
1220 /* Now parse the remainder of the parameters. */
1221 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step)
1223 arg.argv_step = 1;
1225 if (0);
1226 else if (arg_match(&arg, &threads, argi))
1227 cfg.g_threads = arg_parse_uint(&arg);
1228 else if (arg_match(&arg, &profile, argi))
1229 cfg.g_profile = arg_parse_uint(&arg);
1230 else if (arg_match(&arg, &width, argi))
1231 cfg.g_w = arg_parse_uint(&arg);
1232 else if (arg_match(&arg, &height, argi))
1233 cfg.g_h = arg_parse_uint(&arg);
1234 else if (arg_match(&arg, &timebase, argi))
1235 cfg.g_timebase = arg_parse_rational(&arg);
1236 else if (arg_match(&arg, &error_resilient, argi))
1237 cfg.g_error_resilient = arg_parse_uint(&arg);
1238 else if (arg_match(&arg, &lag_in_frames, argi))
1239 cfg.g_lag_in_frames = arg_parse_uint(&arg);
1240 else if (arg_match(&arg, &dropframe_thresh, argi))
1241 cfg.rc_dropframe_thresh = arg_parse_uint(&arg);
1242 else if (arg_match(&arg, &resize_allowed, argi))
1243 cfg.rc_resize_allowed = arg_parse_uint(&arg);
1244 else if (arg_match(&arg, &resize_up_thresh, argi))
1245 cfg.rc_resize_up_thresh = arg_parse_uint(&arg);
1246 else if (arg_match(&arg, &resize_down_thresh, argi))
1247 cfg.rc_resize_down_thresh = arg_parse_uint(&arg);
1248 else if (arg_match(&arg, &resize_down_thresh, argi))
1249 cfg.rc_resize_down_thresh = arg_parse_uint(&arg);
1250 else if (arg_match(&arg, &end_usage, argi))
1251 cfg.rc_end_usage = arg_parse_uint(&arg);
1252 else if (arg_match(&arg, &target_bitrate, argi))
1253 cfg.rc_target_bitrate = arg_parse_uint(&arg);
1254 else if (arg_match(&arg, &min_quantizer, argi))
1255 cfg.rc_min_quantizer = arg_parse_uint(&arg);
1256 else if (arg_match(&arg, &max_quantizer, argi))
1257 cfg.rc_max_quantizer = arg_parse_uint(&arg);
1258 else if (arg_match(&arg, &undershoot_pct, argi))
1259 cfg.rc_undershoot_pct = arg_parse_uint(&arg);
1260 else if (arg_match(&arg, &overshoot_pct, argi))
1261 cfg.rc_overshoot_pct = arg_parse_uint(&arg);
1262 else if (arg_match(&arg, &buf_sz, argi))
1263 cfg.rc_buf_sz = arg_parse_uint(&arg);
1264 else if (arg_match(&arg, &buf_initial_sz, argi))
1265 cfg.rc_buf_initial_sz = arg_parse_uint(&arg);
1266 else if (arg_match(&arg, &buf_optimal_sz, argi))
1267 cfg.rc_buf_optimal_sz = arg_parse_uint(&arg);
1268 else if (arg_match(&arg, &bias_pct, argi))
1270 cfg.rc_2pass_vbr_bias_pct = arg_parse_uint(&arg);
1272 if (arg_passes < 2)
1273 fprintf(stderr,
1274 "Warning: option %s ignored in one-pass mode.\n",
1275 arg.name);
1277 else if (arg_match(&arg, &minsection_pct, argi))
1279 cfg.rc_2pass_vbr_minsection_pct = arg_parse_uint(&arg);
1281 if (arg_passes < 2)
1282 fprintf(stderr,
1283 "Warning: option %s ignored in one-pass mode.\n",
1284 arg.name);
1286 else if (arg_match(&arg, &maxsection_pct, argi))
1288 cfg.rc_2pass_vbr_maxsection_pct = arg_parse_uint(&arg);
1290 if (arg_passes < 2)
1291 fprintf(stderr,
1292 "Warning: option %s ignored in one-pass mode.\n",
1293 arg.name);
1295 else if (arg_match(&arg, &kf_min_dist, argi))
1296 cfg.kf_min_dist = arg_parse_uint(&arg);
1297 else if (arg_match(&arg, &kf_max_dist, argi))
1298 cfg.kf_max_dist = arg_parse_uint(&arg);
1299 else if (arg_match(&arg, &kf_disabled, argi))
1300 cfg.kf_mode = VPX_KF_DISABLED;
1301 else
1302 argj++;
1305 /* Handle codec specific options */
1306 #if CONFIG_VP8_ENCODER
1308 if (codec->iface == &vpx_codec_vp8_cx_algo)
1310 ctrl_args = vp8_args;
1311 ctrl_args_map = vp8_arg_ctrl_map;
1314 #endif
1316 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step)
1318 int match = 0;
1320 arg.argv_step = 1;
1322 for (i = 0; ctrl_args[i]; i++)
1324 if (arg_match(&arg, ctrl_args[i], argi))
1326 match = 1;
1328 if (arg_ctrl_cnt < ARG_CTRL_CNT_MAX)
1330 arg_ctrls[arg_ctrl_cnt][0] = ctrl_args_map[i];
1331 arg_ctrls[arg_ctrl_cnt][1] = arg_parse_enum_or_int(&arg);
1332 arg_ctrl_cnt++;
1337 if (!match)
1338 argj++;
1341 /* Check for unrecognized options */
1342 for (argi = argv; *argi; argi++)
1343 if (argi[0][0] == '-' && argi[0][1])
1344 die("Error: Unrecognized option %s\n", *argi);
1346 /* Handle non-option arguments */
1347 in_fn = argv[0];
1349 if (!in_fn)
1350 usage_exit();
1352 if(!out_fn)
1353 die("Error: Output file is required (specify with -o)\n");
1355 memset(&stats, 0, sizeof(stats));
1357 for (pass = one_pass_only ? one_pass_only - 1 : 0; pass < arg_passes; pass++)
1359 int frames_in = 0, frames_out = 0;
1360 unsigned long nbytes = 0;
1361 struct detect_buffer detect;
1363 /* Parse certain options from the input file, if possible */
1364 infile = strcmp(in_fn, "-") ? fopen(in_fn, "rb")
1365 : set_binary_mode(stdin);
1367 if (!infile)
1369 fprintf(stderr, "Failed to open input file\n");
1370 return EXIT_FAILURE;
1373 /* For RAW input sources, these bytes will applied on the first frame
1374 * in read_frame().
1376 detect.buf_read = fread(detect.buf, 1, 4, infile);
1377 detect.position = 0;
1379 if (detect.buf_read == 4 && file_is_y4m(infile, &y4m, detect.buf))
1381 if (y4m_input_open(&y4m, infile, detect.buf, 4) >= 0)
1383 file_type = FILE_TYPE_Y4M;
1384 cfg.g_w = y4m.pic_w;
1385 cfg.g_h = y4m.pic_h;
1387 /* Use the frame rate from the file only if none was specified
1388 * on the command-line.
1390 if (!arg_have_framerate)
1392 arg_framerate.num = y4m.fps_n;
1393 arg_framerate.den = y4m.fps_d;
1396 arg_use_i420 = 0;
1398 else
1400 fprintf(stderr, "Unsupported Y4M stream.\n");
1401 return EXIT_FAILURE;
1404 else if (detect.buf_read == 4 &&
1405 file_is_ivf(infile, &fourcc, &cfg.g_w, &cfg.g_h, &detect))
1407 file_type = FILE_TYPE_IVF;
1408 switch (fourcc)
1410 case 0x32315659:
1411 arg_use_i420 = 0;
1412 break;
1413 case 0x30323449:
1414 arg_use_i420 = 1;
1415 break;
1416 default:
1417 fprintf(stderr, "Unsupported fourcc (%08x) in IVF\n", fourcc);
1418 return EXIT_FAILURE;
1421 else
1423 file_type = FILE_TYPE_RAW;
1426 if(!cfg.g_w || !cfg.g_h)
1428 fprintf(stderr, "Specify stream dimensions with --width (-w) "
1429 " and --height (-h).\n");
1430 return EXIT_FAILURE;
1433 #define SHOW(field) fprintf(stderr, " %-28s = %d\n", #field, cfg.field)
1435 if (verbose && pass == 0)
1437 fprintf(stderr, "Codec: %s\n", vpx_codec_iface_name(codec->iface));
1438 fprintf(stderr, "Source file: %s Format: %s\n", in_fn,
1439 arg_use_i420 ? "I420" : "YV12");
1440 fprintf(stderr, "Destination file: %s\n", out_fn);
1441 fprintf(stderr, "Encoder parameters:\n");
1443 SHOW(g_usage);
1444 SHOW(g_threads);
1445 SHOW(g_profile);
1446 SHOW(g_w);
1447 SHOW(g_h);
1448 SHOW(g_timebase.num);
1449 SHOW(g_timebase.den);
1450 SHOW(g_error_resilient);
1451 SHOW(g_pass);
1452 SHOW(g_lag_in_frames);
1453 SHOW(rc_dropframe_thresh);
1454 SHOW(rc_resize_allowed);
1455 SHOW(rc_resize_up_thresh);
1456 SHOW(rc_resize_down_thresh);
1457 SHOW(rc_end_usage);
1458 SHOW(rc_target_bitrate);
1459 SHOW(rc_min_quantizer);
1460 SHOW(rc_max_quantizer);
1461 SHOW(rc_undershoot_pct);
1462 SHOW(rc_overshoot_pct);
1463 SHOW(rc_buf_sz);
1464 SHOW(rc_buf_initial_sz);
1465 SHOW(rc_buf_optimal_sz);
1466 SHOW(rc_2pass_vbr_bias_pct);
1467 SHOW(rc_2pass_vbr_minsection_pct);
1468 SHOW(rc_2pass_vbr_maxsection_pct);
1469 SHOW(kf_mode);
1470 SHOW(kf_min_dist);
1471 SHOW(kf_max_dist);
1474 if(pass == (one_pass_only ? one_pass_only - 1 : 0)) {
1475 if (file_type == FILE_TYPE_Y4M)
1476 /*The Y4M reader does its own allocation.
1477 Just initialize this here to avoid problems if we never read any
1478 frames.*/
1479 memset(&raw, 0, sizeof(raw));
1480 else
1481 vpx_img_alloc(&raw, arg_use_i420 ? VPX_IMG_FMT_I420 : VPX_IMG_FMT_YV12,
1482 cfg.g_w, cfg.g_h, 1);
1485 outfile = strcmp(out_fn, "-") ? fopen(out_fn, "wb")
1486 : set_binary_mode(stdout);
1488 if (!outfile)
1490 fprintf(stderr, "Failed to open output file\n");
1491 return EXIT_FAILURE;
1494 if(write_webm && fseek(outfile, 0, SEEK_CUR))
1496 fprintf(stderr, "WebM output to pipes not supported.\n");
1497 return EXIT_FAILURE;
1500 if (stats_fn)
1502 if (!stats_open_file(&stats, stats_fn, pass))
1504 fprintf(stderr, "Failed to open statistics store\n");
1505 return EXIT_FAILURE;
1508 else
1510 if (!stats_open_mem(&stats, pass))
1512 fprintf(stderr, "Failed to open statistics store\n");
1513 return EXIT_FAILURE;
1517 cfg.g_pass = arg_passes == 2
1518 ? pass ? VPX_RC_LAST_PASS : VPX_RC_FIRST_PASS
1519 : VPX_RC_ONE_PASS;
1520 #if VPX_ENCODER_ABI_VERSION > (1 + VPX_CODEC_ABI_VERSION)
1522 if (pass)
1524 cfg.rc_twopass_stats_in = stats_get(&stats);
1527 #endif
1529 if(write_webm)
1531 ebml.stream = outfile;
1532 write_webm_file_header(&ebml, &cfg, &arg_framerate);
1534 else
1535 write_ivf_file_header(outfile, &cfg, codec->fourcc, 0);
1538 /* Construct Encoder Context */
1539 vpx_codec_enc_init(&encoder, codec->iface, &cfg,
1540 show_psnr ? VPX_CODEC_USE_PSNR : 0);
1541 ctx_exit_on_error(&encoder, "Failed to initialize encoder");
1543 /* Note that we bypass the vpx_codec_control wrapper macro because
1544 * we're being clever to store the control IDs in an array. Real
1545 * applications will want to make use of the enumerations directly
1547 for (i = 0; i < arg_ctrl_cnt; i++)
1549 if (vpx_codec_control_(&encoder, arg_ctrls[i][0], arg_ctrls[i][1]))
1550 fprintf(stderr, "Error: Tried to set control %d = %d\n",
1551 arg_ctrls[i][0], arg_ctrls[i][1]);
1553 ctx_exit_on_error(&encoder, "Failed to control codec");
1556 frame_avail = 1;
1557 got_data = 0;
1559 while (frame_avail || got_data)
1561 vpx_codec_iter_t iter = NULL;
1562 const vpx_codec_cx_pkt_t *pkt;
1563 struct vpx_usec_timer timer;
1564 int64_t frame_start, next_frame_start;
1566 if (!arg_limit || frames_in < arg_limit)
1568 frame_avail = read_frame(infile, &raw, file_type, &y4m,
1569 &detect);
1571 if (frame_avail)
1572 frames_in++;
1574 fprintf(stderr,
1575 "\rPass %d/%d frame %4d/%-4d %7ldB \033[K", pass + 1,
1576 arg_passes, frames_in, frames_out, nbytes);
1578 else
1579 frame_avail = 0;
1581 vpx_usec_timer_start(&timer);
1583 frame_start = (cfg.g_timebase.den * (int64_t)(frames_in - 1)
1584 * arg_framerate.den) / cfg.g_timebase.num / arg_framerate.num;
1585 next_frame_start = (cfg.g_timebase.den * (int64_t)(frames_in)
1586 * arg_framerate.den)
1587 / cfg.g_timebase.num / arg_framerate.num;
1588 vpx_codec_encode(&encoder, frame_avail ? &raw : NULL, frame_start,
1589 next_frame_start - frame_start,
1590 0, arg_deadline);
1591 vpx_usec_timer_mark(&timer);
1592 cx_time += vpx_usec_timer_elapsed(&timer);
1593 ctx_exit_on_error(&encoder, "Failed to encode frame");
1594 got_data = 0;
1596 while ((pkt = vpx_codec_get_cx_data(&encoder, &iter)))
1598 got_data = 1;
1600 switch (pkt->kind)
1602 case VPX_CODEC_CX_FRAME_PKT:
1603 frames_out++;
1604 fprintf(stderr, " %6luF",
1605 (unsigned long)pkt->data.frame.sz);
1607 if(write_webm)
1609 /* Update the hash */
1610 if(!ebml.debug)
1611 hash = murmur(pkt->data.frame.buf,
1612 pkt->data.frame.sz, hash);
1614 write_webm_block(&ebml, &cfg, pkt);
1616 else
1618 write_ivf_frame_header(outfile, pkt);
1619 if(fwrite(pkt->data.frame.buf, 1,
1620 pkt->data.frame.sz, outfile));
1622 nbytes += pkt->data.raw.sz;
1623 break;
1624 case VPX_CODEC_STATS_PKT:
1625 frames_out++;
1626 fprintf(stderr, " %6luS",
1627 (unsigned long)pkt->data.twopass_stats.sz);
1628 stats_write(&stats,
1629 pkt->data.twopass_stats.buf,
1630 pkt->data.twopass_stats.sz);
1631 nbytes += pkt->data.raw.sz;
1632 break;
1633 case VPX_CODEC_PSNR_PKT:
1635 if (show_psnr)
1637 int i;
1639 psnr_sse_total += pkt->data.psnr.sse[0];
1640 psnr_samples_total += pkt->data.psnr.samples[0];
1641 for (i = 0; i < 4; i++)
1643 fprintf(stderr, "%.3lf ", pkt->data.psnr.psnr[i]);
1644 psnr_totals[i] += pkt->data.psnr.psnr[i];
1646 psnr_count++;
1649 break;
1650 default:
1651 break;
1655 fflush(stdout);
1658 fprintf(stderr,
1659 "\rPass %d/%d frame %4d/%-4d %7ldB %7ldb/f %7"PRId64"b/s"
1660 " %7lu %s (%.2f fps)\033[K", pass + 1,
1661 arg_passes, frames_in, frames_out, nbytes, nbytes * 8 / frames_in,
1662 nbytes * 8 *(int64_t)arg_framerate.num / arg_framerate.den / frames_in,
1663 cx_time > 9999999 ? cx_time / 1000 : cx_time,
1664 cx_time > 9999999 ? "ms" : "us",
1665 (float)frames_in * 1000000.0 / (float)cx_time);
1667 if ( (show_psnr) && (psnr_count>0) )
1669 int i;
1670 double ovpsnr = vp8_mse2psnr(psnr_samples_total, 255.0,
1671 psnr_sse_total);
1673 fprintf(stderr, "\nPSNR (Overall/Avg/Y/U/V)");
1675 fprintf(stderr, " %.3lf", ovpsnr);
1676 for (i = 0; i < 4; i++)
1678 fprintf(stderr, " %.3lf", psnr_totals[i]/psnr_count);
1682 vpx_codec_destroy(&encoder);
1684 fclose(infile);
1686 if(write_webm)
1688 write_webm_file_footer(&ebml, hash);
1690 else
1692 if (!fseek(outfile, 0, SEEK_SET))
1693 write_ivf_file_header(outfile, &cfg, codec->fourcc, frames_out);
1696 fclose(outfile);
1697 stats_close(&stats, arg_passes-1);
1698 fprintf(stderr, "\n");
1700 if (one_pass_only)
1701 break;
1704 vpx_img_free(&raw);
1705 free(argv);
1706 return EXIT_SUCCESS;