2 * Copyright (c) 2010 The VP8 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.
12 /* This is a simple program that encodes YV12 files and generates ivf
13 * files using the new interface.
16 #define USE_POSIX_MMAP 0
18 #define USE_POSIX_MMAP 1
25 #include "vpx/vpx_encoder.h"
27 #include <sys/types.h>
33 #include "vpx/vp8cx.h"
34 #include "vpx_ports/mem_ops.h"
35 #include "vpx_ports/vpx_timer.h"
38 static const char *exec_name
;
40 static const struct codec_item
43 const vpx_codec_iface_t
*iface
;
47 #if CONFIG_VP8_ENCODER
48 {"vp8", &vpx_codec_vp8_cx_algo
, 0x30385056},
52 static void usage_exit();
54 void die(const char *fmt
, ...)
63 static void ctx_exit_on_error(vpx_codec_ctx_t
*ctx
, const char *s
)
67 const char *detail
= vpx_codec_error_detail(ctx
);
69 printf("%s: %s\n", s
, vpx_codec_error(ctx
));
72 printf(" %s\n", detail
);
78 /* This structure is used to abstract the different ways of handling
79 * first pass statistics.
90 int stats_open_file(stats_io_t
*stats
, const char *fpf
, int pass
)
98 stats
->file
= fopen(fpf
, "wb");
100 stats
->buf
.buf
= NULL
,
101 res
= (stats
->file
!= NULL
);
107 struct stat stat_buf
;
110 fd
= open(fpf
, O_RDONLY
);
111 stats
->file
= fdopen(fd
, "rb");
112 fstat(fd
, &stat_buf
);
113 stats
->buf
.sz
= stat_buf
.st_size
;
114 stats
->buf
.buf
= mmap(NULL
, stats
->buf
.sz
, PROT_READ
, MAP_PRIVATE
,
116 res
= (stats
->buf
.buf
!= NULL
);
120 stats
->file
= fopen(fpf
, "rb");
122 if (fseek(stats
->file
, 0, SEEK_END
))
124 fprintf(stderr
, "First-pass stats file must be seekable!\n");
128 stats
->buf
.sz
= stats
->buf_alloc_sz
= ftell(stats
->file
);
131 stats
->buf
.buf
= malloc(stats
->buf_alloc_sz
);
135 fprintf(stderr
, "Failed to allocate first-pass stats buffer (%d bytes)\n",
136 stats
->buf_alloc_sz
);
140 nbytes
= fread(stats
->buf
.buf
, 1, stats
->buf
.sz
, stats
->file
);
141 res
= (nbytes
== stats
->buf
.sz
);
148 int stats_open_mem(stats_io_t
*stats
, int pass
)
156 stats
->buf_alloc_sz
= 64 * 1024;
157 stats
->buf
.buf
= malloc(stats
->buf_alloc_sz
);
160 stats
->buf_ptr
= stats
->buf
.buf
;
161 res
= (stats
->buf
.buf
!= NULL
);
166 void stats_close(stats_io_t
*stats
)
170 if (stats
->pass
== 1)
174 munmap(stats
->buf
.buf
, stats
->buf
.sz
);
176 free(stats
->buf
.buf
);
185 if (stats
->pass
== 1)
186 free(stats
->buf
.buf
);
190 void stats_write(stats_io_t
*stats
, const void *pkt
, size_t len
)
194 fwrite(pkt
, 1, len
, stats
->file
);
198 if (stats
->buf
.sz
+ len
> stats
->buf_alloc_sz
)
200 size_t new_sz
= stats
->buf_alloc_sz
+ 64 * 1024;
201 char *new_ptr
= realloc(stats
->buf
.buf
, new_sz
);
205 stats
->buf_ptr
= new_ptr
+ (stats
->buf_ptr
- (char *)stats
->buf
.buf
);
206 stats
->buf
.buf
= new_ptr
;
207 stats
->buf_alloc_sz
= new_sz
;
211 memcpy(stats
->buf_ptr
, pkt
, len
);
212 stats
->buf
.sz
+= len
;
213 stats
->buf_ptr
+= len
;
217 vpx_fixed_buf_t
stats_get(stats_io_t
*stats
)
229 #define IVF_FRAME_HDR_SZ (4+8) /* 4 byte size + 8 byte timestamp */
230 static int read_frame(FILE *f
, vpx_image_t
*img
, unsigned int file_type
,
235 if (file_type
== FILE_TYPE_Y4M
)
237 if (y4m_input_fetch_frame(y4m
, f
, img
) < 0)
242 if (file_type
== FILE_TYPE_IVF
)
244 char junk
[IVF_FRAME_HDR_SZ
];
246 /* Skip the frame header. We know how big the frame should be. See
247 * write_ivf_frame_header() for documentation on the frame header
250 fread(junk
, 1, IVF_FRAME_HDR_SZ
, f
);
253 for (plane
= 0; plane
< 3; plane
++)
256 int w
= (plane
? (1 + img
->d_w
) / 2 : img
->d_w
);
257 int h
= (plane
? (1 + img
->d_h
) / 2 : img
->d_h
);
260 /* Determine the correct plane based on the image format. The for-loop
261 * always counts in Y,U,V order, but this may not match the order of
267 ptr
= img
->planes
[img
->fmt
==VPX_IMG_FMT_YV12
? VPX_PLANE_V
: VPX_PLANE_U
];
270 ptr
= img
->planes
[img
->fmt
==VPX_IMG_FMT_YV12
?VPX_PLANE_U
: VPX_PLANE_V
];
273 ptr
= img
->planes
[plane
];
276 for (r
= 0; r
< h
; r
++)
279 ptr
+= img
->stride
[plane
];
288 unsigned int file_is_y4m(FILE *infile
,
292 if (fread(raw_hdr
, 1, 4, infile
) == 4 &&
293 memcmp(raw_hdr
, "YUV4", 4) == 0 &&
294 y4m_input_open(y4m
, infile
, raw_hdr
, 4) >= 0)
302 #define IVF_FILE_HDR_SZ (32)
303 unsigned int file_is_ivf(FILE *infile
,
304 unsigned int *fourcc
,
306 unsigned int *height
)
308 char raw_hdr
[IVF_FILE_HDR_SZ
];
311 /* See write_ivf_file_header() for more documentation on the file header
314 if (fread(raw_hdr
, 1, IVF_FILE_HDR_SZ
, infile
) == IVF_FILE_HDR_SZ
)
316 if (raw_hdr
[0] == 'D' && raw_hdr
[1] == 'K'
317 && raw_hdr
[2] == 'I' && raw_hdr
[3] == 'F')
321 if (mem_get_le16(raw_hdr
+ 4) != 0)
322 fprintf(stderr
, "Error: Unrecognized IVF version! This file may not"
323 " decode properly.");
325 *fourcc
= mem_get_le32(raw_hdr
+ 8);
331 *width
= mem_get_le16(raw_hdr
+ 12);
332 *height
= mem_get_le16(raw_hdr
+ 14);
341 static void write_ivf_file_header(FILE *outfile
,
342 const vpx_codec_enc_cfg_t
*cfg
,
348 if (cfg
->g_pass
!= VPX_RC_ONE_PASS
&& cfg
->g_pass
!= VPX_RC_LAST_PASS
)
355 mem_put_le16(header
+ 4, 0); /* version */
356 mem_put_le16(header
+ 6, 32); /* headersize */
357 mem_put_le32(header
+ 8, fourcc
); /* headersize */
358 mem_put_le16(header
+ 12, cfg
->g_w
); /* width */
359 mem_put_le16(header
+ 14, cfg
->g_h
); /* height */
360 mem_put_le32(header
+ 16, cfg
->g_timebase
.den
); /* rate */
361 mem_put_le32(header
+ 20, cfg
->g_timebase
.num
); /* scale */
362 mem_put_le32(header
+ 24, frame_cnt
); /* length */
363 mem_put_le32(header
+ 28, 0); /* unused */
365 fwrite(header
, 1, 32, outfile
);
369 static void write_ivf_frame_header(FILE *outfile
,
370 const vpx_codec_cx_pkt_t
*pkt
)
375 if (pkt
->kind
!= VPX_CODEC_CX_FRAME_PKT
)
378 pts
= pkt
->data
.frame
.pts
;
379 mem_put_le32(header
, pkt
->data
.frame
.sz
);
380 mem_put_le32(header
+ 4, pts
& 0xFFFFFFFF);
381 mem_put_le32(header
+ 8, pts
>> 32);
383 fwrite(header
, 1, 12, outfile
);
388 static const arg_def_t use_yv12
= ARG_DEF(NULL
, "yv12", 0,
389 "Input file is YV12 ");
390 static const arg_def_t use_i420
= ARG_DEF(NULL
, "i420", 0,
391 "Input file is I420 (default)");
392 static const arg_def_t codecarg
= ARG_DEF(NULL
, "codec", 1,
394 static const arg_def_t passes
= ARG_DEF("p", "passes", 1,
395 "Number of passes (1/2)");
396 static const arg_def_t pass_arg
= ARG_DEF(NULL
, "pass", 1,
397 "Pass to execute (1/2)");
398 static const arg_def_t fpf_name
= ARG_DEF(NULL
, "fpf", 1,
399 "First pass statistics file name");
400 static const arg_def_t limit
= ARG_DEF(NULL
, "limit", 1,
401 "Stop encoding after n input frames");
402 static const arg_def_t deadline
= ARG_DEF("d", "deadline", 1,
403 "Deadline per frame (usec)");
404 static const arg_def_t best_dl
= ARG_DEF(NULL
, "best", 0,
405 "Use Best Quality Deadline");
406 static const arg_def_t good_dl
= ARG_DEF(NULL
, "good", 0,
407 "Use Good Quality Deadline");
408 static const arg_def_t rt_dl
= ARG_DEF(NULL
, "rt", 0,
409 "Use Realtime Quality Deadline");
410 static const arg_def_t verbosearg
= ARG_DEF("v", "verbose", 0,
411 "Show encoder parameters");
412 static const arg_def_t psnrarg
= ARG_DEF(NULL
, "psnr", 0,
413 "Show PSNR in status line");
414 static const arg_def_t
*main_args
[] =
416 &codecarg
, &passes
, &pass_arg
, &fpf_name
, &limit
, &deadline
, &best_dl
, &good_dl
, &rt_dl
,
417 &verbosearg
, &psnrarg
,
421 static const arg_def_t usage
= ARG_DEF("u", "usage", 1,
422 "Usage profile number to use");
423 static const arg_def_t threads
= ARG_DEF("t", "threads", 1,
424 "Max number of threads to use");
425 static const arg_def_t profile
= ARG_DEF(NULL
, "profile", 1,
426 "Bitstream profile number to use");
427 static const arg_def_t width
= ARG_DEF("w", "width", 1,
429 static const arg_def_t height
= ARG_DEF("h", "height", 1,
431 static const arg_def_t timebase
= ARG_DEF(NULL
, "timebase", 1,
432 "Stream timebase (frame duration)");
433 static const arg_def_t error_resilient
= ARG_DEF(NULL
, "error-resilient", 1,
434 "Enable error resiliency features");
435 static const arg_def_t lag_in_frames
= ARG_DEF(NULL
, "lag-in-frames", 1,
436 "Max number of frames to lag");
438 static const arg_def_t
*global_args
[] =
440 &use_yv12
, &use_i420
, &usage
, &threads
, &profile
,
441 &width
, &height
, &timebase
, &error_resilient
,
445 static const arg_def_t dropframe_thresh
= ARG_DEF(NULL
, "drop-frame", 1,
446 "Temporal resampling threshold (buf %)");
447 static const arg_def_t resize_allowed
= ARG_DEF(NULL
, "resize-allowed", 1,
448 "Spatial resampling enabled (bool)");
449 static const arg_def_t resize_up_thresh
= ARG_DEF(NULL
, "resize-up", 1,
450 "Upscale threshold (buf %)");
451 static const arg_def_t resize_down_thresh
= ARG_DEF(NULL
, "resize-down", 1,
452 "Downscale threshold (buf %)");
453 static const arg_def_t end_usage
= ARG_DEF(NULL
, "end-usage", 1,
455 static const arg_def_t target_bitrate
= ARG_DEF(NULL
, "target-bitrate", 1,
457 static const arg_def_t min_quantizer
= ARG_DEF(NULL
, "min-q", 1,
458 "Minimum (best) quantizer");
459 static const arg_def_t max_quantizer
= ARG_DEF(NULL
, "max-q", 1,
460 "Maximum (worst) quantizer");
461 static const arg_def_t undershoot_pct
= ARG_DEF(NULL
, "undershoot-pct", 1,
462 "Datarate undershoot (min) target (%)");
463 static const arg_def_t overshoot_pct
= ARG_DEF(NULL
, "overshoot-pct", 1,
464 "Datarate overshoot (max) target (%)");
465 static const arg_def_t buf_sz
= ARG_DEF(NULL
, "buf-sz", 1,
466 "Client buffer size (ms)");
467 static const arg_def_t buf_initial_sz
= ARG_DEF(NULL
, "buf-initial-sz", 1,
468 "Client initial buffer size (ms)");
469 static const arg_def_t buf_optimal_sz
= ARG_DEF(NULL
, "buf-optimal-sz", 1,
470 "Client optimal buffer size (ms)");
471 static const arg_def_t
*rc_args
[] =
473 &dropframe_thresh
, &resize_allowed
, &resize_up_thresh
, &resize_down_thresh
,
474 &end_usage
, &target_bitrate
, &min_quantizer
, &max_quantizer
,
475 &undershoot_pct
, &overshoot_pct
, &buf_sz
, &buf_initial_sz
, &buf_optimal_sz
,
480 static const arg_def_t bias_pct
= ARG_DEF(NULL
, "bias-pct", 1,
481 "CBR/VBR bias (0=CBR, 100=VBR)");
482 static const arg_def_t minsection_pct
= ARG_DEF(NULL
, "minsection-pct", 1,
483 "GOP min bitrate (% of target)");
484 static const arg_def_t maxsection_pct
= ARG_DEF(NULL
, "maxsection-pct", 1,
485 "GOP max bitrate (% of target)");
486 static const arg_def_t
*rc_twopass_args
[] =
488 &bias_pct
, &minsection_pct
, &maxsection_pct
, NULL
492 static const arg_def_t kf_min_dist
= ARG_DEF(NULL
, "kf-min-dist", 1,
493 "Minimum keyframe interval (frames)");
494 static const arg_def_t kf_max_dist
= ARG_DEF(NULL
, "kf-max-dist", 1,
495 "Maximum keyframe interval (frames)");
496 static const arg_def_t
*kf_args
[] =
498 &kf_min_dist
, &kf_max_dist
, NULL
502 #if CONFIG_VP8_ENCODER
503 static const arg_def_t noise_sens
= ARG_DEF(NULL
, "noise-sensitivity", 1,
504 "Noise sensitivity (frames to blur)");
505 static const arg_def_t sharpness
= ARG_DEF(NULL
, "sharpness", 1,
506 "Filter sharpness (0-7)");
507 static const arg_def_t static_thresh
= ARG_DEF(NULL
, "static-thresh", 1,
508 "Motion detection threshold");
511 #if CONFIG_VP8_ENCODER
512 static const arg_def_t cpu_used
= ARG_DEF(NULL
, "cpu-used", 1,
513 "CPU Used (-16..16)");
517 #if CONFIG_VP8_ENCODER
518 static const arg_def_t token_parts
= ARG_DEF(NULL
, "token-parts", 1,
519 "Number of token partitions to use, log2");
520 static const arg_def_t auto_altref
= ARG_DEF(NULL
, "auto-alt-ref", 1,
521 "Enable automatic alt reference frames");
522 static const arg_def_t arnr_maxframes
= ARG_DEF(NULL
, "arnr-maxframes", 1,
523 "alt_ref Max Frames");
524 static const arg_def_t arnr_strength
= ARG_DEF(NULL
, "arnr-strength", 1,
526 static const arg_def_t arnr_type
= ARG_DEF(NULL
, "arnr-type", 1,
529 static const arg_def_t
*vp8_args
[] =
531 &cpu_used
, &auto_altref
, &noise_sens
, &sharpness
, &static_thresh
,
532 &token_parts
, &arnr_maxframes
, &arnr_strength
, &arnr_type
, NULL
534 static const int vp8_arg_ctrl_map
[] =
536 VP8E_SET_CPUUSED
, VP8E_SET_ENABLEAUTOALTREF
,
537 VP8E_SET_NOISE_SENSITIVITY
, VP8E_SET_SHARPNESS
, VP8E_SET_STATIC_THRESHOLD
,
538 VP8E_SET_TOKEN_PARTITIONS
,
539 VP8E_SET_ARNR_MAXFRAMES
, VP8E_SET_ARNR_STRENGTH
, VP8E_SET_ARNR_TYPE
, 0
543 static const arg_def_t
*no_args
[] = { NULL
};
545 static void usage_exit()
549 printf("Usage: %s <options> src_filename dst_filename\n", exec_name
);
551 printf("\n_options:\n");
552 arg_show_usage(stdout
, main_args
);
553 printf("\n_encoder Global Options:\n");
554 arg_show_usage(stdout
, global_args
);
555 printf("\n_rate Control Options:\n");
556 arg_show_usage(stdout
, rc_args
);
557 printf("\n_twopass Rate Control Options:\n");
558 arg_show_usage(stdout
, rc_twopass_args
);
559 printf("\n_keyframe Placement Options:\n");
560 arg_show_usage(stdout
, kf_args
);
561 #if CONFIG_VP8_ENCODER
562 printf("\n_vp8 Specific Options:\n");
563 arg_show_usage(stdout
, vp8_args
);
566 "Included encoders:\n"
569 for (i
= 0; i
< sizeof(codecs
) / sizeof(codecs
[0]); i
++)
570 printf(" %-6s - %s\n",
572 vpx_codec_iface_name(codecs
[i
].iface
));
577 #define ARG_CTRL_CNT_MAX 10
580 int main(int argc
, const char **argv_
)
582 vpx_codec_ctx_t encoder
;
583 const char *in_fn
= NULL
, *out_fn
= NULL
, *stats_fn
= NULL
;
585 FILE *infile
, *outfile
;
586 vpx_codec_enc_cfg_t cfg
;
588 int pass
, one_pass_only
= 0;
591 const struct codec_item
*codec
= codecs
;
592 int frame_avail
, got_data
;
595 char **argv
, **argi
, **argj
;
596 int arg_usage
= 0, arg_passes
= 1, arg_deadline
= 0;
597 int arg_ctrls
[ARG_CTRL_CNT_MAX
][2], arg_ctrl_cnt
= 0;
599 static const arg_def_t
**ctrl_args
= no_args
;
600 static const int *ctrl_args_map
= NULL
;
601 int verbose
= 0, show_psnr
= 0;
602 int arg_use_i420
= 1;
603 int arg_have_timebase
= 0;
604 unsigned long cx_time
= 0;
605 unsigned int file_type
, fourcc
;
608 exec_name
= argv_
[0];
614 /* First parse the codec and usage values, because we want to apply other
615 * parameters on top of the default configuration provided by the codec.
617 argv
= argv_dup(argc
- 1, argv_
+ 1);
619 for (argi
= argj
= argv
; (*argj
= *argi
); argi
+= arg
.argv_step
)
623 if (arg_match(&arg
, &codecarg
, argi
))
627 for (j
= 0; j
< sizeof(codecs
) / sizeof(codecs
[0]); j
++)
628 if (!strcmp(codecs
[j
].name
, arg
.val
))
634 die("Error: Unrecognized argument (%s) to --codec\n",
638 else if (arg_match(&arg
, &passes
, argi
))
640 arg_passes
= arg_parse_uint(&arg
);
642 if (arg_passes
< 1 || arg_passes
> 2)
643 die("Error: Invalid number of passes (%d)\n", arg_passes
);
645 else if (arg_match(&arg
, &pass_arg
, argi
))
647 one_pass_only
= arg_parse_uint(&arg
);
649 if (one_pass_only
< 1 || one_pass_only
> 2)
650 die("Error: Invalid pass selected (%d)\n", one_pass_only
);
652 else if (arg_match(&arg
, &fpf_name
, argi
))
654 else if (arg_match(&arg
, &usage
, argi
))
655 arg_usage
= arg_parse_uint(&arg
);
656 else if (arg_match(&arg
, &deadline
, argi
))
657 arg_deadline
= arg_parse_uint(&arg
);
658 else if (arg_match(&arg
, &best_dl
, argi
))
659 arg_deadline
= VPX_DL_BEST_QUALITY
;
660 else if (arg_match(&arg
, &good_dl
, argi
))
661 arg_deadline
= VPX_DL_GOOD_QUALITY
;
662 else if (arg_match(&arg
, &rt_dl
, argi
))
663 arg_deadline
= VPX_DL_REALTIME
;
664 else if (arg_match(&arg
, &use_yv12
, argi
))
668 else if (arg_match(&arg
, &use_i420
, argi
))
672 else if (arg_match(&arg
, &verbosearg
, argi
))
674 else if (arg_match(&arg
, &limit
, argi
))
675 arg_limit
= arg_parse_uint(&arg
);
676 else if (arg_match(&arg
, &psnrarg
, argi
))
682 /* Ensure that --passes and --pass are consistent. If --pass is set and --passes=2,
683 * ensure --fpf was set.
687 /* DWIM: Assume the user meant passes=2 if pass=2 is specified */
688 if (one_pass_only
> arg_passes
)
690 printf("Warning: Assuming --pass=%d implies --passes=%d\n",
691 one_pass_only
, one_pass_only
);
692 arg_passes
= one_pass_only
;
695 if (arg_passes
== 2 && !stats_fn
)
696 die("Must specify --fpf when --pass=%d and --passes=2\n", one_pass_only
);
699 /* Populate encoder configuration */
700 res
= vpx_codec_enc_config_default(codec
->iface
, &cfg
, arg_usage
);
704 printf("Failed to get config: %s\n", vpx_codec_err_to_string(res
));
708 /* Now parse the remainder of the parameters. */
709 for (argi
= argj
= argv
; (*argj
= *argi
); argi
+= arg
.argv_step
)
714 else if (arg_match(&arg
, &threads
, argi
))
715 cfg
.g_threads
= arg_parse_uint(&arg
);
716 else if (arg_match(&arg
, &profile
, argi
))
717 cfg
.g_profile
= arg_parse_uint(&arg
);
718 else if (arg_match(&arg
, &width
, argi
))
719 cfg
.g_w
= arg_parse_uint(&arg
);
720 else if (arg_match(&arg
, &height
, argi
))
721 cfg
.g_h
= arg_parse_uint(&arg
);
722 else if (arg_match(&arg
, &timebase
, argi
))
724 cfg
.g_timebase
= arg_parse_rational(&arg
);
725 arg_have_timebase
= 1;
727 else if (arg_match(&arg
, &error_resilient
, argi
))
728 cfg
.g_error_resilient
= arg_parse_uint(&arg
);
729 else if (arg_match(&arg
, &lag_in_frames
, argi
))
730 cfg
.g_lag_in_frames
= arg_parse_uint(&arg
);
731 else if (arg_match(&arg
, &dropframe_thresh
, argi
))
732 cfg
.rc_dropframe_thresh
= arg_parse_uint(&arg
);
733 else if (arg_match(&arg
, &resize_allowed
, argi
))
734 cfg
.rc_resize_allowed
= arg_parse_uint(&arg
);
735 else if (arg_match(&arg
, &resize_up_thresh
, argi
))
736 cfg
.rc_resize_up_thresh
= arg_parse_uint(&arg
);
737 else if (arg_match(&arg
, &resize_down_thresh
, argi
))
738 cfg
.rc_resize_down_thresh
= arg_parse_uint(&arg
);
739 else if (arg_match(&arg
, &resize_down_thresh
, argi
))
740 cfg
.rc_resize_down_thresh
= arg_parse_uint(&arg
);
741 else if (arg_match(&arg
, &end_usage
, argi
))
742 cfg
.rc_end_usage
= arg_parse_uint(&arg
);
743 else if (arg_match(&arg
, &target_bitrate
, argi
))
744 cfg
.rc_target_bitrate
= arg_parse_uint(&arg
);
745 else if (arg_match(&arg
, &min_quantizer
, argi
))
746 cfg
.rc_min_quantizer
= arg_parse_uint(&arg
);
747 else if (arg_match(&arg
, &max_quantizer
, argi
))
748 cfg
.rc_max_quantizer
= arg_parse_uint(&arg
);
749 else if (arg_match(&arg
, &undershoot_pct
, argi
))
750 cfg
.rc_undershoot_pct
= arg_parse_uint(&arg
);
751 else if (arg_match(&arg
, &overshoot_pct
, argi
))
752 cfg
.rc_overshoot_pct
= arg_parse_uint(&arg
);
753 else if (arg_match(&arg
, &buf_sz
, argi
))
754 cfg
.rc_buf_sz
= arg_parse_uint(&arg
);
755 else if (arg_match(&arg
, &buf_initial_sz
, argi
))
756 cfg
.rc_buf_initial_sz
= arg_parse_uint(&arg
);
757 else if (arg_match(&arg
, &buf_optimal_sz
, argi
))
758 cfg
.rc_buf_optimal_sz
= arg_parse_uint(&arg
);
759 else if (arg_match(&arg
, &bias_pct
, argi
))
761 cfg
.rc_2pass_vbr_bias_pct
= arg_parse_uint(&arg
);
764 printf("Warning: option %s ignored in one-pass mode.\n",
767 else if (arg_match(&arg
, &minsection_pct
, argi
))
769 cfg
.rc_2pass_vbr_minsection_pct
= arg_parse_uint(&arg
);
772 printf("Warning: option %s ignored in one-pass mode.\n",
775 else if (arg_match(&arg
, &maxsection_pct
, argi
))
777 cfg
.rc_2pass_vbr_maxsection_pct
= arg_parse_uint(&arg
);
780 printf("Warning: option %s ignored in one-pass mode.\n",
783 else if (arg_match(&arg
, &kf_min_dist
, argi
))
784 cfg
.kf_min_dist
= arg_parse_uint(&arg
);
785 else if (arg_match(&arg
, &kf_max_dist
, argi
))
786 cfg
.kf_max_dist
= arg_parse_uint(&arg
);
791 /* Handle codec specific options */
792 #if CONFIG_VP8_ENCODER
794 if (codec
->iface
== &vpx_codec_vp8_cx_algo
)
796 ctrl_args
= vp8_args
;
797 ctrl_args_map
= vp8_arg_ctrl_map
;
802 for (argi
= argj
= argv
; (*argj
= *argi
); argi
+= arg
.argv_step
)
808 for (i
= 0; ctrl_args
[i
]; i
++)
810 if (arg_match(&arg
, ctrl_args
[i
], argi
))
814 if (arg_ctrl_cnt
< ARG_CTRL_CNT_MAX
)
816 arg_ctrls
[arg_ctrl_cnt
][0] = ctrl_args_map
[i
];
817 arg_ctrls
[arg_ctrl_cnt
][1] = arg_parse_int(&arg
);
827 /* Check for unrecognized options */
828 for (argi
= argv
; *argi
; argi
++)
829 if (argi
[0][0] == '-')
830 die("Error: Unrecognized option %s\n", *argi
);
832 /* Handle non-option arguments */
836 if (!in_fn
|| !out_fn
)
839 /* Parse certain options from the input file, if possible */
840 infile
= fopen(in_fn
, "rb");
844 printf("Failed to open input file");
848 if (file_is_y4m(infile
, &y4m
))
850 file_type
= FILE_TYPE_Y4M
;
853 /* Use the frame rate from the file only if none was specified on the
856 if (!arg_have_timebase
)
858 cfg
.g_timebase
.num
= y4m
.fps_d
;
859 cfg
.g_timebase
.den
= y4m
.fps_n
;
863 else if (file_is_ivf(infile
, &fourcc
, &cfg
.g_w
, &cfg
.g_h
))
865 file_type
= FILE_TYPE_IVF
;
875 printf("Unsupported fourcc (%08x) in IVF\n", fourcc
);
880 file_type
= FILE_TYPE_RAW
;
885 #define SHOW(field) printf(" %-28s = %d\n", #field, cfg.field)
889 printf("Codec: %s\n", vpx_codec_iface_name(codec
->iface
));
890 printf("Source file: %s Format: %s\n", in_fn
, arg_use_i420
? "I420" : "YV12");
891 printf("Destination file: %s\n", out_fn
);
892 printf("Encoder parameters:\n");
899 SHOW(g_timebase
.num
);
900 SHOW(g_timebase
.den
);
901 SHOW(g_error_resilient
);
903 SHOW(g_lag_in_frames
);
904 SHOW(rc_dropframe_thresh
);
905 SHOW(rc_resize_allowed
);
906 SHOW(rc_resize_up_thresh
);
907 SHOW(rc_resize_down_thresh
);
909 SHOW(rc_target_bitrate
);
910 SHOW(rc_min_quantizer
);
911 SHOW(rc_max_quantizer
);
912 SHOW(rc_undershoot_pct
);
913 SHOW(rc_overshoot_pct
);
915 SHOW(rc_buf_initial_sz
);
916 SHOW(rc_buf_optimal_sz
);
917 SHOW(rc_2pass_vbr_bias_pct
);
918 SHOW(rc_2pass_vbr_minsection_pct
);
919 SHOW(rc_2pass_vbr_maxsection_pct
);
925 if (file_type
== FILE_TYPE_Y4M
)
926 /*The Y4M reader does its own allocation.
927 Just initialize this here to avoid problems if we never read any
929 memset(&raw
, 0, sizeof(raw
));
931 vpx_img_alloc(&raw
, arg_use_i420
? VPX_IMG_FMT_I420
: VPX_IMG_FMT_YV12
,
932 cfg
.g_w
, cfg
.g_h
, 1);
934 // This was added so that ivfenc will create monotically increasing
935 // timestamps. Since we create new timestamps for alt-reference frames
936 // we need to make room in the series of timestamps. Since there can
937 // only be 1 alt-ref frame ( current bitstream) multiplying by 2
938 // gives us enough room.
939 cfg
.g_timebase
.den
*= 2;
941 memset(&stats
, 0, sizeof(stats
));
943 for (pass
= one_pass_only
? one_pass_only
- 1 : 0; pass
< arg_passes
; pass
++)
945 int frames_in
= 0, frames_out
= 0;
946 unsigned long nbytes
= 0;
948 infile
= fopen(in_fn
, "rb");
952 printf("Failed to open input file");
956 /*Skip the file header.*/
957 if (file_type
== FILE_TYPE_IVF
)
959 char raw_hdr
[IVF_FILE_HDR_SZ
];
960 (void)fread(raw_hdr
, 1, IVF_FILE_HDR_SZ
, infile
);
962 else if(file_type
== FILE_TYPE_Y4M
)
965 (void)fgets(buffer
, sizeof(buffer
)/sizeof(*buffer
) - 1, infile
);
968 outfile
= fopen(out_fn
, "wb");
972 printf("Failed to open output file");
978 if (!stats_open_file(&stats
, stats_fn
, pass
))
980 printf("Failed to open statistics store\n");
986 if (!stats_open_mem(&stats
, pass
))
988 printf("Failed to open statistics store\n");
993 cfg
.g_pass
= arg_passes
== 2
994 ? pass
? VPX_RC_LAST_PASS
: VPX_RC_FIRST_PASS
996 #if VPX_ENCODER_ABI_VERSION > (1 + VPX_CODEC_ABI_VERSION)
1000 cfg
.rc_twopass_stats_in
= stats_get(&stats
);
1005 write_ivf_file_header(outfile
, &cfg
, codec
->fourcc
, 0);
1008 /* Construct Encoder Context */
1009 if (cfg
.kf_min_dist
== cfg
.kf_max_dist
)
1010 cfg
.kf_mode
= VPX_KF_FIXED
;
1012 vpx_codec_enc_init(&encoder
, codec
->iface
, &cfg
,
1013 show_psnr
? VPX_CODEC_USE_PSNR
: 0);
1014 ctx_exit_on_error(&encoder
, "Failed to initialize encoder");
1016 /* Note that we bypass the vpx_codec_control wrapper macro because
1017 * we're being clever to store the control IDs in an array. Real
1018 * applications will want to make use of the enumerations directly
1020 for (i
= 0; i
< arg_ctrl_cnt
; i
++)
1022 if (vpx_codec_control_(&encoder
, arg_ctrls
[i
][0], arg_ctrls
[i
][1]))
1023 printf("Error: Tried to set control %d = %d\n",
1024 arg_ctrls
[i
][0], arg_ctrls
[i
][1]);
1026 ctx_exit_on_error(&encoder
, "Failed to control codec");
1032 while (frame_avail
|| got_data
)
1034 vpx_codec_iter_t iter
= NULL
;
1035 const vpx_codec_cx_pkt_t
*pkt
;
1036 struct vpx_usec_timer timer
;
1038 if (!arg_limit
|| frames_in
< arg_limit
)
1040 frame_avail
= read_frame(infile
, &raw
, file_type
, &y4m
);
1045 printf("\rPass %d/%d frame %4d/%-4d %7ldB \033[K", pass
+ 1,
1046 arg_passes
, frames_in
, frames_out
, nbytes
);
1051 vpx_usec_timer_start(&timer
);
1053 // since we halved our timebase we need to double the timestamps
1054 // and duration we pass in.
1055 vpx_codec_encode(&encoder
, frame_avail
? &raw
: NULL
, (frames_in
- 1) * 2,
1056 2, 0, arg_deadline
);
1057 vpx_usec_timer_mark(&timer
);
1058 cx_time
+= vpx_usec_timer_elapsed(&timer
);
1059 ctx_exit_on_error(&encoder
, "Failed to encode frame");
1062 while ((pkt
= vpx_codec_get_cx_data(&encoder
, &iter
)))
1068 case VPX_CODEC_CX_FRAME_PKT
:
1071 (unsigned long)pkt
->data
.frame
.sz
);
1072 write_ivf_frame_header(outfile
, pkt
);
1073 fwrite(pkt
->data
.frame
.buf
, 1, pkt
->data
.frame
.sz
, outfile
);
1074 nbytes
+= pkt
->data
.raw
.sz
;
1076 case VPX_CODEC_STATS_PKT
:
1079 (unsigned long)pkt
->data
.twopass_stats
.sz
);
1081 pkt
->data
.twopass_stats
.buf
,
1082 pkt
->data
.twopass_stats
.sz
);
1083 nbytes
+= pkt
->data
.raw
.sz
;
1085 case VPX_CODEC_PSNR_PKT
:
1091 for (i
= 0; i
< 4; i
++)
1092 printf("%.3lf ", pkt
->data
.psnr
.psnr
[i
]);
1104 /* this bitrate calc is simplified and relies on the fact that this
1105 * application uses 1/timebase for framerate.
1107 printf("\rPass %d/%d frame %4d/%-4d %7ldB %7ldb/f %7"PRId64
"b/s"
1108 " %7lu %s (%.2f fps)\033[K", pass
+ 1,
1109 arg_passes
, frames_in
, frames_out
, nbytes
, nbytes
* 8 / frames_in
,
1110 nbytes
* 8 *(int64_t)cfg
.g_timebase
.den
/2/ cfg
.g_timebase
.num
/ frames_in
,
1111 cx_time
> 9999999 ? cx_time
/ 1000 : cx_time
,
1112 cx_time
> 9999999 ? "ms" : "us",
1113 (float)frames_in
* 1000000.0 / (float)cx_time
);
1115 vpx_codec_destroy(&encoder
);
1119 if (!fseek(outfile
, 0, SEEK_SET
))
1120 write_ivf_file_header(outfile
, &cfg
, codec
->fourcc
, frames_out
);
1123 stats_close(&stats
);
1132 return EXIT_SUCCESS
;