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.
12 #include "./vpx_config.h"
23 #include "third_party/libyuv/include/libyuv/scale.h"
26 #include "vpx/vpx_encoder.h"
28 #include "vpx/vpx_decoder.h"
33 #include "./tools_common.h"
35 #if CONFIG_VP8_ENCODER || CONFIG_VP9_ENCODER
36 #include "vpx/vp8cx.h"
38 #if CONFIG_VP8_DECODER || CONFIG_VP9_DECODER
39 #include "vpx/vp8dx.h"
42 #include "vpx/vpx_integer.h"
43 #include "vpx_ports/mem_ops.h"
44 #include "vpx_ports/vpx_timer.h"
45 #include "./rate_hist.h"
46 #include "./vpxstats.h"
47 #include "./warnings.h"
49 #include "./webmenc.h"
51 #include "./y4minput.h"
53 /* Swallow warnings about unused results of fread/fwrite */
54 static size_t wrap_fread(void *ptr
, size_t size
, size_t nmemb
,
56 return fread(ptr
, size
, nmemb
, stream
);
58 #define fread wrap_fread
60 static size_t wrap_fwrite(const void *ptr
, size_t size
, size_t nmemb
,
62 return fwrite(ptr
, size
, nmemb
, stream
);
64 #define fwrite wrap_fwrite
67 static const char *exec_name
;
69 static void warn_or_exit_on_errorv(vpx_codec_ctx_t
*ctx
, int fatal
,
70 const char *s
, va_list ap
) {
72 const char *detail
= vpx_codec_error_detail(ctx
);
74 vfprintf(stderr
, s
, ap
);
75 fprintf(stderr
, ": %s\n", vpx_codec_error(ctx
));
78 fprintf(stderr
, " %s\n", detail
);
85 static void ctx_exit_on_error(vpx_codec_ctx_t
*ctx
, const char *s
, ...) {
89 warn_or_exit_on_errorv(ctx
, 1, s
, ap
);
93 static void warn_or_exit_on_error(vpx_codec_ctx_t
*ctx
, int fatal
,
98 warn_or_exit_on_errorv(ctx
, fatal
, s
, ap
);
102 int read_frame(struct VpxInputContext
*input_ctx
, vpx_image_t
*img
) {
103 FILE *f
= input_ctx
->file
;
104 y4m_input
*y4m
= &input_ctx
->y4m
;
107 if (input_ctx
->file_type
== FILE_TYPE_Y4M
) {
108 if (y4m_input_fetch_frame(y4m
, f
, img
) < 1)
111 shortread
= read_yuv_frame(input_ctx
, img
);
117 int file_is_y4m(const char detect
[4]) {
118 if (memcmp(detect
, "YUV4", 4) == 0) {
124 int fourcc_is_ivf(const char detect
[4]) {
125 if (memcmp(detect
, "DKIF", 4) == 0) {
131 static const arg_def_t debugmode
= ARG_DEF("D", "debug", 0,
132 "Debug mode (makes output deterministic)");
133 static const arg_def_t outputfile
= ARG_DEF("o", "output", 1,
135 static const arg_def_t use_yv12
= ARG_DEF(NULL
, "yv12", 0,
136 "Input file is YV12 ");
137 static const arg_def_t use_i420
= ARG_DEF(NULL
, "i420", 0,
138 "Input file is I420 (default)");
139 static const arg_def_t use_i422
= ARG_DEF(NULL
, "i422", 0,
140 "Input file is I422");
141 static const arg_def_t use_i444
= ARG_DEF(NULL
, "i444", 0,
142 "Input file is I444");
143 static const arg_def_t codecarg
= ARG_DEF(NULL
, "codec", 1,
145 static const arg_def_t passes
= ARG_DEF("p", "passes", 1,
146 "Number of passes (1/2)");
147 static const arg_def_t pass_arg
= ARG_DEF(NULL
, "pass", 1,
148 "Pass to execute (1/2)");
149 static const arg_def_t fpf_name
= ARG_DEF(NULL
, "fpf", 1,
150 "First pass statistics file name");
151 #if CONFIG_FP_MB_STATS
152 static const arg_def_t fpmbf_name
= ARG_DEF(NULL
, "fpmbf", 1,
153 "First pass block statistics file name");
155 static const arg_def_t limit
= ARG_DEF(NULL
, "limit", 1,
156 "Stop encoding after n input frames");
157 static const arg_def_t skip
= ARG_DEF(NULL
, "skip", 1,
158 "Skip the first n input frames");
159 static const arg_def_t deadline
= ARG_DEF("d", "deadline", 1,
160 "Deadline per frame (usec)");
161 static const arg_def_t best_dl
= ARG_DEF(NULL
, "best", 0,
162 "Use Best Quality Deadline");
163 static const arg_def_t good_dl
= ARG_DEF(NULL
, "good", 0,
164 "Use Good Quality Deadline");
165 static const arg_def_t rt_dl
= ARG_DEF(NULL
, "rt", 0,
166 "Use Realtime Quality Deadline");
167 static const arg_def_t quietarg
= ARG_DEF("q", "quiet", 0,
168 "Do not print encode progress");
169 static const arg_def_t verbosearg
= ARG_DEF("v", "verbose", 0,
170 "Show encoder parameters");
171 static const arg_def_t psnrarg
= ARG_DEF(NULL
, "psnr", 0,
172 "Show PSNR in status line");
174 static const struct arg_enum_list test_decode_enum
[] = {
175 {"off", TEST_DECODE_OFF
},
176 {"fatal", TEST_DECODE_FATAL
},
177 {"warn", TEST_DECODE_WARN
},
180 static const arg_def_t recontest
= ARG_DEF_ENUM(NULL
, "test-decode", 1,
181 "Test encode/decode mismatch",
183 static const arg_def_t framerate
= ARG_DEF(NULL
, "fps", 1,
184 "Stream frame rate (rate/scale)");
185 static const arg_def_t use_ivf
= ARG_DEF(NULL
, "ivf", 0,
186 "Output IVF (default is WebM if WebM IO is enabled)");
187 static const arg_def_t out_part
= ARG_DEF("P", "output-partitions", 0,
188 "Makes encoder output partitions. Requires IVF output!");
189 static const arg_def_t q_hist_n
= ARG_DEF(NULL
, "q-hist", 1,
190 "Show quantizer histogram (n-buckets)");
191 static const arg_def_t rate_hist_n
= ARG_DEF(NULL
, "rate-hist", 1,
192 "Show rate histogram (n-buckets)");
193 static const arg_def_t disable_warnings
=
194 ARG_DEF(NULL
, "disable-warnings", 0,
195 "Disable warnings about potentially incorrect encode settings.");
196 static const arg_def_t disable_warning_prompt
=
197 ARG_DEF("y", "disable-warning-prompt", 0,
198 "Display warnings, but do not prompt user to continue.");
199 static const arg_def_t experimental_bitstream
=
200 ARG_DEF(NULL
, "experimental-bitstream", 0,
201 "Allow experimental bitstream features.");
203 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
204 static const arg_def_t test16bitinternalarg
= ARG_DEF(
205 NULL
, "test-16bit-internal", 0, "Force use of 16 bit internal buffer");
208 static const arg_def_t
*main_args
[] = {
210 &outputfile
, &codecarg
, &passes
, &pass_arg
, &fpf_name
, &limit
, &skip
,
211 &deadline
, &best_dl
, &good_dl
, &rt_dl
,
212 &quietarg
, &verbosearg
, &psnrarg
, &use_ivf
, &out_part
, &q_hist_n
,
213 &rate_hist_n
, &disable_warnings
, &disable_warning_prompt
,
217 static const arg_def_t usage
= ARG_DEF("u", "usage", 1,
218 "Usage profile number to use");
219 static const arg_def_t threads
= ARG_DEF("t", "threads", 1,
220 "Max number of threads to use");
221 static const arg_def_t profile
= ARG_DEF(NULL
, "profile", 1,
222 "Bitstream profile number to use");
223 static const arg_def_t width
= ARG_DEF("w", "width", 1,
225 static const arg_def_t height
= ARG_DEF("h", "height", 1,
228 static const struct arg_enum_list stereo_mode_enum
[] = {
229 {"mono", STEREO_FORMAT_MONO
},
230 {"left-right", STEREO_FORMAT_LEFT_RIGHT
},
231 {"bottom-top", STEREO_FORMAT_BOTTOM_TOP
},
232 {"top-bottom", STEREO_FORMAT_TOP_BOTTOM
},
233 {"right-left", STEREO_FORMAT_RIGHT_LEFT
},
236 static const arg_def_t stereo_mode
= ARG_DEF_ENUM(NULL
, "stereo-mode", 1,
237 "Stereo 3D video format", stereo_mode_enum
);
239 static const arg_def_t timebase
= ARG_DEF(NULL
, "timebase", 1,
240 "Output timestamp precision (fractional seconds)");
241 static const arg_def_t error_resilient
= ARG_DEF(NULL
, "error-resilient", 1,
242 "Enable error resiliency features");
243 static const arg_def_t lag_in_frames
= ARG_DEF(NULL
, "lag-in-frames", 1,
244 "Max number of frames to lag");
246 static const arg_def_t
*global_args
[] = {
247 &use_yv12
, &use_i420
, &use_i422
, &use_i444
,
248 &usage
, &threads
, &profile
,
253 &timebase
, &framerate
,
255 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
256 &test16bitinternalarg
,
261 static const arg_def_t dropframe_thresh
= ARG_DEF(NULL
, "drop-frame", 1,
262 "Temporal resampling threshold (buf %)");
263 static const arg_def_t resize_allowed
= ARG_DEF(NULL
, "resize-allowed", 1,
264 "Spatial resampling enabled (bool)");
265 static const arg_def_t resize_width
= ARG_DEF(NULL
, "resize-width", 1,
266 "Width of encoded frame");
267 static const arg_def_t resize_height
= ARG_DEF(NULL
, "resize-height", 1,
268 "Height of encoded frame");
269 static const arg_def_t resize_up_thresh
= ARG_DEF(NULL
, "resize-up", 1,
270 "Upscale threshold (buf %)");
271 static const arg_def_t resize_down_thresh
= ARG_DEF(NULL
, "resize-down", 1,
272 "Downscale threshold (buf %)");
273 static const struct arg_enum_list end_usage_enum
[] = {
280 static const arg_def_t end_usage
= ARG_DEF_ENUM(NULL
, "end-usage", 1,
281 "Rate control mode", end_usage_enum
);
282 static const arg_def_t target_bitrate
= ARG_DEF(NULL
, "target-bitrate", 1,
284 static const arg_def_t min_quantizer
= ARG_DEF(NULL
, "min-q", 1,
285 "Minimum (best) quantizer");
286 static const arg_def_t max_quantizer
= ARG_DEF(NULL
, "max-q", 1,
287 "Maximum (worst) quantizer");
288 static const arg_def_t undershoot_pct
= ARG_DEF(NULL
, "undershoot-pct", 1,
289 "Datarate undershoot (min) target (%)");
290 static const arg_def_t overshoot_pct
= ARG_DEF(NULL
, "overshoot-pct", 1,
291 "Datarate overshoot (max) target (%)");
292 static const arg_def_t buf_sz
= ARG_DEF(NULL
, "buf-sz", 1,
293 "Client buffer size (ms)");
294 static const arg_def_t buf_initial_sz
= ARG_DEF(NULL
, "buf-initial-sz", 1,
295 "Client initial buffer size (ms)");
296 static const arg_def_t buf_optimal_sz
= ARG_DEF(NULL
, "buf-optimal-sz", 1,
297 "Client optimal buffer size (ms)");
298 static const arg_def_t
*rc_args
[] = {
299 &dropframe_thresh
, &resize_allowed
, &resize_width
, &resize_height
,
300 &resize_up_thresh
, &resize_down_thresh
, &end_usage
, &target_bitrate
,
301 &min_quantizer
, &max_quantizer
, &undershoot_pct
, &overshoot_pct
, &buf_sz
,
302 &buf_initial_sz
, &buf_optimal_sz
, NULL
306 static const arg_def_t bias_pct
= ARG_DEF(NULL
, "bias-pct", 1,
307 "CBR/VBR bias (0=CBR, 100=VBR)");
308 static const arg_def_t minsection_pct
= ARG_DEF(NULL
, "minsection-pct", 1,
309 "GOP min bitrate (% of target)");
310 static const arg_def_t maxsection_pct
= ARG_DEF(NULL
, "maxsection-pct", 1,
311 "GOP max bitrate (% of target)");
312 static const arg_def_t
*rc_twopass_args
[] = {
313 &bias_pct
, &minsection_pct
, &maxsection_pct
, NULL
317 static const arg_def_t kf_min_dist
= ARG_DEF(NULL
, "kf-min-dist", 1,
318 "Minimum keyframe interval (frames)");
319 static const arg_def_t kf_max_dist
= ARG_DEF(NULL
, "kf-max-dist", 1,
320 "Maximum keyframe interval (frames)");
321 static const arg_def_t kf_disabled
= ARG_DEF(NULL
, "disable-kf", 0,
322 "Disable keyframe placement");
323 static const arg_def_t
*kf_args
[] = {
324 &kf_min_dist
, &kf_max_dist
, &kf_disabled
, NULL
328 static const arg_def_t noise_sens
= ARG_DEF(NULL
, "noise-sensitivity", 1,
329 "Noise sensitivity (frames to blur)");
330 static const arg_def_t sharpness
= ARG_DEF(NULL
, "sharpness", 1,
331 "Loop filter sharpness (0..7)");
332 static const arg_def_t static_thresh
= ARG_DEF(NULL
, "static-thresh", 1,
333 "Motion detection threshold");
334 static const arg_def_t cpu_used
= ARG_DEF(NULL
, "cpu-used", 1,
335 "CPU Used (-16..16)");
336 static const arg_def_t auto_altref
= ARG_DEF(NULL
, "auto-alt-ref", 1,
337 "Enable automatic alt reference frames");
338 static const arg_def_t arnr_maxframes
= ARG_DEF(NULL
, "arnr-maxframes", 1,
339 "AltRef max frames (0..15)");
340 static const arg_def_t arnr_strength
= ARG_DEF(NULL
, "arnr-strength", 1,
341 "AltRef filter strength (0..6)");
342 static const arg_def_t arnr_type
= ARG_DEF(NULL
, "arnr-type", 1,
344 static const struct arg_enum_list tuning_enum
[] = {
345 {"psnr", VP8_TUNE_PSNR
},
346 {"ssim", VP8_TUNE_SSIM
},
349 static const arg_def_t tune_ssim
= ARG_DEF_ENUM(NULL
, "tune", 1,
350 "Material to favor", tuning_enum
);
351 static const arg_def_t cq_level
= ARG_DEF(NULL
, "cq-level", 1,
352 "Constant/Constrained Quality level");
353 static const arg_def_t max_intra_rate_pct
= ARG_DEF(NULL
, "max-intra-rate", 1,
354 "Max I-frame bitrate (pct)");
356 #if CONFIG_VP8_ENCODER
357 static const arg_def_t token_parts
=
358 ARG_DEF(NULL
, "token-parts", 1, "Number of token partitions to use, log2");
359 static const arg_def_t
*vp8_args
[] = {
360 &cpu_used
, &auto_altref
, &noise_sens
, &sharpness
, &static_thresh
,
361 &token_parts
, &arnr_maxframes
, &arnr_strength
, &arnr_type
,
362 &tune_ssim
, &cq_level
, &max_intra_rate_pct
,
365 static const int vp8_arg_ctrl_map
[] = {
366 VP8E_SET_CPUUSED
, VP8E_SET_ENABLEAUTOALTREF
,
367 VP8E_SET_NOISE_SENSITIVITY
, VP8E_SET_SHARPNESS
, VP8E_SET_STATIC_THRESHOLD
,
368 VP8E_SET_TOKEN_PARTITIONS
,
369 VP8E_SET_ARNR_MAXFRAMES
, VP8E_SET_ARNR_STRENGTH
, VP8E_SET_ARNR_TYPE
,
370 VP8E_SET_TUNING
, VP8E_SET_CQ_LEVEL
, VP8E_SET_MAX_INTRA_BITRATE_PCT
,
375 #if CONFIG_VP9_ENCODER
376 static const arg_def_t tile_cols
=
377 ARG_DEF(NULL
, "tile-columns", 1, "Number of tile columns to use, log2");
378 static const arg_def_t tile_rows
=
379 ARG_DEF(NULL
, "tile-rows", 1, "Number of tile rows to use, log2");
380 static const arg_def_t lossless
= ARG_DEF(NULL
, "lossless", 1, "Lossless mode");
381 static const arg_def_t frame_parallel_decoding
= ARG_DEF(
382 NULL
, "frame-parallel", 1, "Enable frame parallel decodability features");
383 static const arg_def_t aq_mode
= ARG_DEF(
385 "Adaptive quantization mode (0: off (default), 1: variance 2: complexity, "
386 "3: cyclic refresh)");
387 static const arg_def_t frame_periodic_boost
= ARG_DEF(
388 NULL
, "frame-boost", 1,
389 "Enable frame periodic boost (0: off (default), 1: on)");
391 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
392 static const struct arg_enum_list bitdepth_enum
[] = {
399 static const arg_def_t bitdeptharg
= ARG_DEF_ENUM("b", "bit-depth", 1,
400 "Bit depth for codec "
401 "(8 for version <=1, "
402 "10 or 12 for version 2)",
404 static const arg_def_t inbitdeptharg
= ARG_DEF(NULL
, "input-bit-depth", 1,
405 "Bit depth of input");
408 static const struct arg_enum_list tune_content_enum
[] = {
409 {"default", VP9E_CONTENT_DEFAULT
},
410 {"screen", VP9E_CONTENT_SCREEN
},
414 static const arg_def_t tune_content
= ARG_DEF_ENUM(
415 NULL
, "tune-content", 1, "Tune content type", tune_content_enum
);
417 static const arg_def_t
*vp9_args
[] = {
418 &cpu_used
, &auto_altref
, &sharpness
, &static_thresh
,
419 &tile_cols
, &tile_rows
, &arnr_maxframes
, &arnr_strength
, &arnr_type
,
420 &tune_ssim
, &cq_level
, &max_intra_rate_pct
, &lossless
,
421 &frame_parallel_decoding
, &aq_mode
, &frame_periodic_boost
,
422 &noise_sens
, &tune_content
,
423 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
424 &bitdeptharg
, &inbitdeptharg
,
428 static const int vp9_arg_ctrl_map
[] = {
429 VP8E_SET_CPUUSED
, VP8E_SET_ENABLEAUTOALTREF
,
430 VP8E_SET_SHARPNESS
, VP8E_SET_STATIC_THRESHOLD
,
431 VP9E_SET_TILE_COLUMNS
, VP9E_SET_TILE_ROWS
,
432 VP8E_SET_ARNR_MAXFRAMES
, VP8E_SET_ARNR_STRENGTH
, VP8E_SET_ARNR_TYPE
,
433 VP8E_SET_TUNING
, VP8E_SET_CQ_LEVEL
, VP8E_SET_MAX_INTRA_BITRATE_PCT
,
434 VP9E_SET_LOSSLESS
, VP9E_SET_FRAME_PARALLEL_DECODING
, VP9E_SET_AQ_MODE
,
435 VP9E_SET_FRAME_PERIODIC_BOOST
, VP9E_SET_NOISE_SENSITIVITY
,
436 VP9E_SET_TUNE_CONTENT
,
441 static const arg_def_t
*no_args
[] = { NULL
};
446 fprintf(stderr
, "Usage: %s <options> -o dst_filename src_filename \n",
449 fprintf(stderr
, "\nOptions:\n");
450 arg_show_usage(stderr
, main_args
);
451 fprintf(stderr
, "\nEncoder Global Options:\n");
452 arg_show_usage(stderr
, global_args
);
453 fprintf(stderr
, "\nRate Control Options:\n");
454 arg_show_usage(stderr
, rc_args
);
455 fprintf(stderr
, "\nTwopass Rate Control Options:\n");
456 arg_show_usage(stderr
, rc_twopass_args
);
457 fprintf(stderr
, "\nKeyframe Placement Options:\n");
458 arg_show_usage(stderr
, kf_args
);
459 #if CONFIG_VP8_ENCODER
460 fprintf(stderr
, "\nVP8 Specific Options:\n");
461 arg_show_usage(stderr
, vp8_args
);
463 #if CONFIG_VP9_ENCODER
464 fprintf(stderr
, "\nVP9 Specific Options:\n");
465 arg_show_usage(stderr
, vp9_args
);
467 fprintf(stderr
, "\nStream timebase (--timebase):\n"
468 " The desired precision of timestamps in the output, expressed\n"
469 " in fractional seconds. Default is 1/1000.\n");
470 fprintf(stderr
, "\nIncluded encoders:\n\n");
472 for (i
= 0; i
< get_vpx_encoder_count(); ++i
) {
473 const VpxInterface
*const encoder
= get_vpx_encoder_by_index(i
);
474 fprintf(stderr
, " %-6s - %s\n",
475 encoder
->name
, vpx_codec_iface_name(encoder
->codec_interface()));
481 #define mmin(a, b) ((a) < (b) ? (a) : (b))
483 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
484 static void find_mismatch_high(const vpx_image_t
*const img1
,
485 const vpx_image_t
*const img2
,
486 int yloc
[4], int uloc
[4], int vloc
[4]) {
487 uint16_t *plane1
, *plane2
;
488 uint32_t stride1
, stride2
;
489 const uint32_t bsize
= 64;
490 const uint32_t bsizey
= bsize
>> img1
->y_chroma_shift
;
491 const uint32_t bsizex
= bsize
>> img1
->x_chroma_shift
;
493 (img1
->d_w
+ img1
->x_chroma_shift
) >> img1
->x_chroma_shift
;
495 (img1
->d_h
+ img1
->y_chroma_shift
) >> img1
->y_chroma_shift
;
498 yloc
[0] = yloc
[1] = yloc
[2] = yloc
[3] = -1;
499 plane1
= (uint16_t*)img1
->planes
[VPX_PLANE_Y
];
500 plane2
= (uint16_t*)img2
->planes
[VPX_PLANE_Y
];
501 stride1
= img1
->stride
[VPX_PLANE_Y
]/2;
502 stride2
= img2
->stride
[VPX_PLANE_Y
]/2;
503 for (i
= 0, match
= 1; match
&& i
< img1
->d_h
; i
+= bsize
) {
504 for (j
= 0; match
&& j
< img1
->d_w
; j
+= bsize
) {
506 const int si
= mmin(i
+ bsize
, img1
->d_h
) - i
;
507 const int sj
= mmin(j
+ bsize
, img1
->d_w
) - j
;
508 for (k
= 0; match
&& k
< si
; ++k
) {
509 for (l
= 0; match
&& l
< sj
; ++l
) {
510 if (*(plane1
+ (i
+ k
) * stride1
+ j
+ l
) !=
511 *(plane2
+ (i
+ k
) * stride2
+ j
+ l
)) {
514 yloc
[2] = *(plane1
+ (i
+ k
) * stride1
+ j
+ l
);
515 yloc
[3] = *(plane2
+ (i
+ k
) * stride2
+ j
+ l
);
524 uloc
[0] = uloc
[1] = uloc
[2] = uloc
[3] = -1;
525 plane1
= (uint16_t*)img1
->planes
[VPX_PLANE_U
];
526 plane2
= (uint16_t*)img2
->planes
[VPX_PLANE_U
];
527 stride1
= img1
->stride
[VPX_PLANE_U
]/2;
528 stride2
= img2
->stride
[VPX_PLANE_U
]/2;
529 for (i
= 0, match
= 1; match
&& i
< c_h
; i
+= bsizey
) {
530 for (j
= 0; match
&& j
< c_w
; j
+= bsizex
) {
532 const int si
= mmin(i
+ bsizey
, c_h
- i
);
533 const int sj
= mmin(j
+ bsizex
, c_w
- j
);
534 for (k
= 0; match
&& k
< si
; ++k
) {
535 for (l
= 0; match
&& l
< sj
; ++l
) {
536 if (*(plane1
+ (i
+ k
) * stride1
+ j
+ l
) !=
537 *(plane2
+ (i
+ k
) * stride2
+ j
+ l
)) {
540 uloc
[2] = *(plane1
+ (i
+ k
) * stride1
+ j
+ l
);
541 uloc
[3] = *(plane2
+ (i
+ k
) * stride2
+ j
+ l
);
550 vloc
[0] = vloc
[1] = vloc
[2] = vloc
[3] = -1;
551 plane1
= (uint16_t*)img1
->planes
[VPX_PLANE_V
];
552 plane2
= (uint16_t*)img2
->planes
[VPX_PLANE_V
];
553 stride1
= img1
->stride
[VPX_PLANE_V
]/2;
554 stride2
= img2
->stride
[VPX_PLANE_V
]/2;
555 for (i
= 0, match
= 1; match
&& i
< c_h
; i
+= bsizey
) {
556 for (j
= 0; match
&& j
< c_w
; j
+= bsizex
) {
558 const int si
= mmin(i
+ bsizey
, c_h
- i
);
559 const int sj
= mmin(j
+ bsizex
, c_w
- j
);
560 for (k
= 0; match
&& k
< si
; ++k
) {
561 for (l
= 0; match
&& l
< sj
; ++l
) {
562 if (*(plane1
+ (i
+ k
) * stride1
+ j
+ l
) !=
563 *(plane2
+ (i
+ k
) * stride2
+ j
+ l
)) {
566 vloc
[2] = *(plane1
+ (i
+ k
) * stride1
+ j
+ l
);
567 vloc
[3] = *(plane2
+ (i
+ k
) * stride2
+ j
+ l
);
578 static void find_mismatch(const vpx_image_t
*const img1
,
579 const vpx_image_t
*const img2
,
580 int yloc
[4], int uloc
[4], int vloc
[4]) {
581 const uint32_t bsize
= 64;
582 const uint32_t bsizey
= bsize
>> img1
->y_chroma_shift
;
583 const uint32_t bsizex
= bsize
>> img1
->x_chroma_shift
;
585 (img1
->d_w
+ img1
->x_chroma_shift
) >> img1
->x_chroma_shift
;
587 (img1
->d_h
+ img1
->y_chroma_shift
) >> img1
->y_chroma_shift
;
590 yloc
[0] = yloc
[1] = yloc
[2] = yloc
[3] = -1;
591 for (i
= 0, match
= 1; match
&& i
< img1
->d_h
; i
+= bsize
) {
592 for (j
= 0; match
&& j
< img1
->d_w
; j
+= bsize
) {
594 const int si
= mmin(i
+ bsize
, img1
->d_h
) - i
;
595 const int sj
= mmin(j
+ bsize
, img1
->d_w
) - j
;
596 for (k
= 0; match
&& k
< si
; ++k
) {
597 for (l
= 0; match
&& l
< sj
; ++l
) {
598 if (*(img1
->planes
[VPX_PLANE_Y
] +
599 (i
+ k
) * img1
->stride
[VPX_PLANE_Y
] + j
+ l
) !=
600 *(img2
->planes
[VPX_PLANE_Y
] +
601 (i
+ k
) * img2
->stride
[VPX_PLANE_Y
] + j
+ l
)) {
604 yloc
[2] = *(img1
->planes
[VPX_PLANE_Y
] +
605 (i
+ k
) * img1
->stride
[VPX_PLANE_Y
] + j
+ l
);
606 yloc
[3] = *(img2
->planes
[VPX_PLANE_Y
] +
607 (i
+ k
) * img2
->stride
[VPX_PLANE_Y
] + j
+ l
);
616 uloc
[0] = uloc
[1] = uloc
[2] = uloc
[3] = -1;
617 for (i
= 0, match
= 1; match
&& i
< c_h
; i
+= bsizey
) {
618 for (j
= 0; match
&& j
< c_w
; j
+= bsizex
) {
620 const int si
= mmin(i
+ bsizey
, c_h
- i
);
621 const int sj
= mmin(j
+ bsizex
, c_w
- j
);
622 for (k
= 0; match
&& k
< si
; ++k
) {
623 for (l
= 0; match
&& l
< sj
; ++l
) {
624 if (*(img1
->planes
[VPX_PLANE_U
] +
625 (i
+ k
) * img1
->stride
[VPX_PLANE_U
] + j
+ l
) !=
626 *(img2
->planes
[VPX_PLANE_U
] +
627 (i
+ k
) * img2
->stride
[VPX_PLANE_U
] + j
+ l
)) {
630 uloc
[2] = *(img1
->planes
[VPX_PLANE_U
] +
631 (i
+ k
) * img1
->stride
[VPX_PLANE_U
] + j
+ l
);
632 uloc
[3] = *(img2
->planes
[VPX_PLANE_U
] +
633 (i
+ k
) * img2
->stride
[VPX_PLANE_U
] + j
+ l
);
641 vloc
[0] = vloc
[1] = vloc
[2] = vloc
[3] = -1;
642 for (i
= 0, match
= 1; match
&& i
< c_h
; i
+= bsizey
) {
643 for (j
= 0; match
&& j
< c_w
; j
+= bsizex
) {
645 const int si
= mmin(i
+ bsizey
, c_h
- i
);
646 const int sj
= mmin(j
+ bsizex
, c_w
- j
);
647 for (k
= 0; match
&& k
< si
; ++k
) {
648 for (l
= 0; match
&& l
< sj
; ++l
) {
649 if (*(img1
->planes
[VPX_PLANE_V
] +
650 (i
+ k
) * img1
->stride
[VPX_PLANE_V
] + j
+ l
) !=
651 *(img2
->planes
[VPX_PLANE_V
] +
652 (i
+ k
) * img2
->stride
[VPX_PLANE_V
] + j
+ l
)) {
655 vloc
[2] = *(img1
->planes
[VPX_PLANE_V
] +
656 (i
+ k
) * img1
->stride
[VPX_PLANE_V
] + j
+ l
);
657 vloc
[3] = *(img2
->planes
[VPX_PLANE_V
] +
658 (i
+ k
) * img2
->stride
[VPX_PLANE_V
] + j
+ l
);
668 static int compare_img(const vpx_image_t
*const img1
,
669 const vpx_image_t
*const img2
) {
670 uint32_t l_w
= img1
->d_w
;
672 (img1
->d_w
+ img1
->x_chroma_shift
) >> img1
->x_chroma_shift
;
674 (img1
->d_h
+ img1
->y_chroma_shift
) >> img1
->y_chroma_shift
;
678 match
&= (img1
->fmt
== img2
->fmt
);
679 match
&= (img1
->d_w
== img2
->d_w
);
680 match
&= (img1
->d_h
== img2
->d_h
);
681 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
682 if (img1
->fmt
& VPX_IMG_FMT_HIGHBITDEPTH
) {
688 for (i
= 0; i
< img1
->d_h
; ++i
)
689 match
&= (memcmp(img1
->planes
[VPX_PLANE_Y
] + i
* img1
->stride
[VPX_PLANE_Y
],
690 img2
->planes
[VPX_PLANE_Y
] + i
* img2
->stride
[VPX_PLANE_Y
],
693 for (i
= 0; i
< c_h
; ++i
)
694 match
&= (memcmp(img1
->planes
[VPX_PLANE_U
] + i
* img1
->stride
[VPX_PLANE_U
],
695 img2
->planes
[VPX_PLANE_U
] + i
* img2
->stride
[VPX_PLANE_U
],
698 for (i
= 0; i
< c_h
; ++i
)
699 match
&= (memcmp(img1
->planes
[VPX_PLANE_V
] + i
* img1
->stride
[VPX_PLANE_V
],
700 img2
->planes
[VPX_PLANE_V
] + i
* img2
->stride
[VPX_PLANE_V
],
707 #define NELEMENTS(x) (sizeof(x)/sizeof(x[0]))
708 #define MAX(x,y) ((x)>(y)?(x):(y))
709 #if CONFIG_VP8_ENCODER && !CONFIG_VP9_ENCODER
710 #define ARG_CTRL_CNT_MAX NELEMENTS(vp8_arg_ctrl_map)
711 #elif !CONFIG_VP8_ENCODER && CONFIG_VP9_ENCODER
712 #define ARG_CTRL_CNT_MAX NELEMENTS(vp9_arg_ctrl_map)
714 #define ARG_CTRL_CNT_MAX MAX(NELEMENTS(vp8_arg_ctrl_map), \
715 NELEMENTS(vp9_arg_ctrl_map))
719 typedef int stereo_format_t
;
720 struct EbmlGlobal
{ int debug
; };
723 /* Per-stream configuration */
724 struct stream_config
{
725 struct vpx_codec_enc_cfg cfg
;
727 const char *stats_fn
;
728 #if CONFIG_FP_MB_STATS
729 const char *fpmb_stats_fn
;
731 stereo_format_t stereo_fmt
;
732 int arg_ctrls
[ARG_CTRL_CNT_MAX
][2];
735 int have_kf_max_dist
;
736 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
737 // whether to use 16bit internal buffers
738 int use_16bit_internal
;
743 struct stream_state
{
745 struct stream_state
*next
;
746 struct stream_config config
;
748 struct rate_hist
*rate_hist
;
749 struct EbmlGlobal ebml
;
750 uint64_t psnr_sse_total
;
751 uint64_t psnr_samples_total
;
752 double psnr_totals
[4];
755 vpx_codec_ctx_t encoder
;
756 unsigned int frames_out
;
760 #if CONFIG_FP_MB_STATS
761 stats_io_t fpmb_stats
;
763 struct vpx_image
*img
;
764 vpx_codec_ctx_t decoder
;
769 void validate_positive_rational(const char *msg
,
770 struct vpx_rational
*rat
) {
777 die("Error: %s must be positive\n", msg
);
780 die("Error: %s has zero denominator\n", msg
);
784 static void parse_global_config(struct VpxEncoderConfig
*global
, char **argv
) {
788 /* Initialize default parameters */
789 memset(global
, 0, sizeof(*global
));
790 global
->codec
= get_vpx_encoder_by_index(0);
792 global
->color_type
= I420
;
793 /* Assign default deadline to good quality */
794 global
->deadline
= VPX_DL_GOOD_QUALITY
;
796 for (argi
= argj
= argv
; (*argj
= *argi
); argi
+= arg
.argv_step
) {
799 if (arg_match(&arg
, &codecarg
, argi
)) {
800 global
->codec
= get_vpx_encoder_by_name(arg
.val
);
802 die("Error: Unrecognized argument (%s) to --codec\n", arg
.val
);
803 } else if (arg_match(&arg
, &passes
, argi
)) {
804 global
->passes
= arg_parse_uint(&arg
);
806 if (global
->passes
< 1 || global
->passes
> 2)
807 die("Error: Invalid number of passes (%d)\n", global
->passes
);
808 } else if (arg_match(&arg
, &pass_arg
, argi
)) {
809 global
->pass
= arg_parse_uint(&arg
);
811 if (global
->pass
< 1 || global
->pass
> 2)
812 die("Error: Invalid pass selected (%d)\n",
814 } else if (arg_match(&arg
, &usage
, argi
))
815 global
->usage
= arg_parse_uint(&arg
);
816 else if (arg_match(&arg
, &deadline
, argi
))
817 global
->deadline
= arg_parse_uint(&arg
);
818 else if (arg_match(&arg
, &best_dl
, argi
))
819 global
->deadline
= VPX_DL_BEST_QUALITY
;
820 else if (arg_match(&arg
, &good_dl
, argi
))
821 global
->deadline
= VPX_DL_GOOD_QUALITY
;
822 else if (arg_match(&arg
, &rt_dl
, argi
))
823 global
->deadline
= VPX_DL_REALTIME
;
824 else if (arg_match(&arg
, &use_yv12
, argi
))
825 global
->color_type
= YV12
;
826 else if (arg_match(&arg
, &use_i420
, argi
))
827 global
->color_type
= I420
;
828 else if (arg_match(&arg
, &use_i422
, argi
))
829 global
->color_type
= I422
;
830 else if (arg_match(&arg
, &use_i444
, argi
))
831 global
->color_type
= I444
;
832 else if (arg_match(&arg
, &quietarg
, argi
))
834 else if (arg_match(&arg
, &verbosearg
, argi
))
836 else if (arg_match(&arg
, &limit
, argi
))
837 global
->limit
= arg_parse_uint(&arg
);
838 else if (arg_match(&arg
, &skip
, argi
))
839 global
->skip_frames
= arg_parse_uint(&arg
);
840 else if (arg_match(&arg
, &psnrarg
, argi
))
841 global
->show_psnr
= 1;
842 else if (arg_match(&arg
, &recontest
, argi
))
843 global
->test_decode
= arg_parse_enum_or_int(&arg
);
844 else if (arg_match(&arg
, &framerate
, argi
)) {
845 global
->framerate
= arg_parse_rational(&arg
);
846 validate_positive_rational(arg
.name
, &global
->framerate
);
847 global
->have_framerate
= 1;
848 } else if (arg_match(&arg
, &out_part
, argi
))
849 global
->out_part
= 1;
850 else if (arg_match(&arg
, &debugmode
, argi
))
852 else if (arg_match(&arg
, &q_hist_n
, argi
))
853 global
->show_q_hist_buckets
= arg_parse_uint(&arg
);
854 else if (arg_match(&arg
, &rate_hist_n
, argi
))
855 global
->show_rate_hist_buckets
= arg_parse_uint(&arg
);
856 else if (arg_match(&arg
, &disable_warnings
, argi
))
857 global
->disable_warnings
= 1;
858 else if (arg_match(&arg
, &disable_warning_prompt
, argi
))
859 global
->disable_warning_prompt
= 1;
860 else if (arg_match(&arg
, &experimental_bitstream
, argi
))
861 global
->experimental_bitstream
= 1;
867 /* DWIM: Assume the user meant passes=2 if pass=2 is specified */
868 if (global
->pass
> global
->passes
) {
869 warn("Assuming --pass=%d implies --passes=%d\n",
870 global
->pass
, global
->pass
);
871 global
->passes
= global
->pass
;
874 /* Validate global config */
875 if (global
->passes
== 0) {
876 #if CONFIG_VP9_ENCODER
877 // Make default VP9 passes = 2 until there is a better quality 1-pass
879 if (global
->codec
!= NULL
&& global
->codec
->name
!= NULL
)
880 global
->passes
= (strcmp(global
->codec
->name
, "vp9") == 0 &&
881 global
->deadline
!= VPX_DL_REALTIME
) ? 2 : 1;
887 if (global
->deadline
== VPX_DL_REALTIME
&&
888 global
->passes
> 1) {
889 warn("Enforcing one-pass encoding in realtime mode\n");
895 void open_input_file(struct VpxInputContext
*input
) {
896 /* Parse certain options from the input file, if possible */
897 input
->file
= strcmp(input
->filename
, "-")
898 ? fopen(input
->filename
, "rb") : set_binary_mode(stdin
);
901 fatal("Failed to open input file");
903 if (!fseeko(input
->file
, 0, SEEK_END
)) {
904 /* Input file is seekable. Figure out how long it is, so we can get
907 input
->length
= ftello(input
->file
);
911 /* For RAW input sources, these bytes will applied on the first frame
914 input
->detect
.buf_read
= fread(input
->detect
.buf
, 1, 4, input
->file
);
915 input
->detect
.position
= 0;
917 if (input
->detect
.buf_read
== 4
918 && file_is_y4m(input
->detect
.buf
)) {
919 if (y4m_input_open(&input
->y4m
, input
->file
, input
->detect
.buf
, 4,
920 input
->only_i420
) >= 0) {
921 input
->file_type
= FILE_TYPE_Y4M
;
922 input
->width
= input
->y4m
.pic_w
;
923 input
->height
= input
->y4m
.pic_h
;
924 input
->framerate
.numerator
= input
->y4m
.fps_n
;
925 input
->framerate
.denominator
= input
->y4m
.fps_d
;
926 input
->fmt
= input
->y4m
.vpx_fmt
;
927 input
->bit_depth
= input
->y4m
.bit_depth
;
929 fatal("Unsupported Y4M stream.");
930 } else if (input
->detect
.buf_read
== 4 && fourcc_is_ivf(input
->detect
.buf
)) {
931 fatal("IVF is not supported as input.");
933 input
->file_type
= FILE_TYPE_RAW
;
938 static void close_input_file(struct VpxInputContext
*input
) {
940 if (input
->file_type
== FILE_TYPE_Y4M
)
941 y4m_input_close(&input
->y4m
);
944 static struct stream_state
*new_stream(struct VpxEncoderConfig
*global
,
945 struct stream_state
*prev
) {
946 struct stream_state
*stream
;
948 stream
= calloc(1, sizeof(*stream
));
949 if (stream
== NULL
) {
950 fatal("Failed to allocate new stream.");
954 memcpy(stream
, prev
, sizeof(*stream
));
960 /* Populate encoder configuration */
961 res
= vpx_codec_enc_config_default(global
->codec
->codec_interface(),
965 fatal("Failed to get config: %s\n", vpx_codec_err_to_string(res
));
967 /* Change the default timebase to a high enough value so that the
968 * encoder will always create strictly increasing timestamps.
970 stream
->config
.cfg
.g_timebase
.den
= 1000;
972 /* Never use the library's default resolution, require it be parsed
973 * from the file or set on the command line.
975 stream
->config
.cfg
.g_w
= 0;
976 stream
->config
.cfg
.g_h
= 0;
978 /* Initialize remaining stream parameters */
979 stream
->config
.write_webm
= 1;
981 stream
->config
.stereo_fmt
= STEREO_FORMAT_MONO
;
982 stream
->ebml
.last_pts_ns
= -1;
983 stream
->ebml
.writer
= NULL
;
984 stream
->ebml
.segment
= NULL
;
987 /* Allows removal of the application version from the EBML tags */
988 stream
->ebml
.debug
= global
->debug
;
990 /* Default lag_in_frames is 0 in realtime mode */
991 if (global
->deadline
== VPX_DL_REALTIME
)
992 stream
->config
.cfg
.g_lag_in_frames
= 0;
995 /* Output files must be specified for each stream */
996 stream
->config
.out_fn
= NULL
;
1003 static int parse_stream_params(struct VpxEncoderConfig
*global
,
1004 struct stream_state
*stream
,
1006 char **argi
, **argj
;
1008 static const arg_def_t
**ctrl_args
= no_args
;
1009 static const int *ctrl_args_map
= NULL
;
1010 struct stream_config
*config
= &stream
->config
;
1011 int eos_mark_found
= 0;
1012 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1013 int test_16bit_internal
= 0;
1016 // Handle codec specific options
1018 #if CONFIG_VP8_ENCODER
1019 } else if (strcmp(global
->codec
->name
, "vp8") == 0) {
1020 ctrl_args
= vp8_args
;
1021 ctrl_args_map
= vp8_arg_ctrl_map
;
1023 #if CONFIG_VP9_ENCODER
1024 } else if (strcmp(global
->codec
->name
, "vp9") == 0) {
1025 ctrl_args
= vp9_args
;
1026 ctrl_args_map
= vp9_arg_ctrl_map
;
1030 for (argi
= argj
= argv
; (*argj
= *argi
); argi
+= arg
.argv_step
) {
1033 /* Once we've found an end-of-stream marker (--) we want to continue
1034 * shifting arguments but not consuming them.
1036 if (eos_mark_found
) {
1039 } else if (!strcmp(*argj
, "--")) {
1045 } else if (arg_match(&arg
, &outputfile
, argi
)) {
1046 config
->out_fn
= arg
.val
;
1047 } else if (arg_match(&arg
, &fpf_name
, argi
)) {
1048 config
->stats_fn
= arg
.val
;
1049 #if CONFIG_FP_MB_STATS
1050 } else if (arg_match(&arg
, &fpmbf_name
, argi
)) {
1051 config
->fpmb_stats_fn
= arg
.val
;
1053 } else if (arg_match(&arg
, &use_ivf
, argi
)) {
1054 config
->write_webm
= 0;
1055 } else if (arg_match(&arg
, &threads
, argi
)) {
1056 config
->cfg
.g_threads
= arg_parse_uint(&arg
);
1057 } else if (arg_match(&arg
, &profile
, argi
)) {
1058 config
->cfg
.g_profile
= arg_parse_uint(&arg
);
1059 } else if (arg_match(&arg
, &width
, argi
)) {
1060 config
->cfg
.g_w
= arg_parse_uint(&arg
);
1061 } else if (arg_match(&arg
, &height
, argi
)) {
1062 config
->cfg
.g_h
= arg_parse_uint(&arg
);
1063 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1064 } else if (arg_match(&arg
, &bitdeptharg
, argi
)) {
1065 config
->cfg
.g_bit_depth
= arg_parse_enum_or_int(&arg
);
1066 } else if (arg_match(&arg
, &inbitdeptharg
, argi
)) {
1067 config
->cfg
.g_input_bit_depth
= arg_parse_uint(&arg
);
1070 } else if (arg_match(&arg
, &stereo_mode
, argi
)) {
1071 config
->stereo_fmt
= arg_parse_enum_or_int(&arg
);
1073 } else if (arg_match(&arg
, &timebase
, argi
)) {
1074 config
->cfg
.g_timebase
= arg_parse_rational(&arg
);
1075 validate_positive_rational(arg
.name
, &config
->cfg
.g_timebase
);
1076 } else if (arg_match(&arg
, &error_resilient
, argi
)) {
1077 config
->cfg
.g_error_resilient
= arg_parse_uint(&arg
);
1078 } else if (arg_match(&arg
, &lag_in_frames
, argi
)) {
1079 config
->cfg
.g_lag_in_frames
= arg_parse_uint(&arg
);
1080 if (global
->deadline
== VPX_DL_REALTIME
&&
1081 config
->cfg
.g_lag_in_frames
!= 0) {
1082 warn("non-zero %s option ignored in realtime mode.\n", arg
.name
);
1083 config
->cfg
.g_lag_in_frames
= 0;
1085 } else if (arg_match(&arg
, &dropframe_thresh
, argi
)) {
1086 config
->cfg
.rc_dropframe_thresh
= arg_parse_uint(&arg
);
1087 } else if (arg_match(&arg
, &resize_allowed
, argi
)) {
1088 config
->cfg
.rc_resize_allowed
= arg_parse_uint(&arg
);
1089 } else if (arg_match(&arg
, &resize_width
, argi
)) {
1090 config
->cfg
.rc_scaled_width
= arg_parse_uint(&arg
);
1091 } else if (arg_match(&arg
, &resize_height
, argi
)) {
1092 config
->cfg
.rc_scaled_height
= arg_parse_uint(&arg
);
1093 } else if (arg_match(&arg
, &resize_up_thresh
, argi
)) {
1094 config
->cfg
.rc_resize_up_thresh
= arg_parse_uint(&arg
);
1095 } else if (arg_match(&arg
, &resize_down_thresh
, argi
)) {
1096 config
->cfg
.rc_resize_down_thresh
= arg_parse_uint(&arg
);
1097 } else if (arg_match(&arg
, &end_usage
, argi
)) {
1098 config
->cfg
.rc_end_usage
= arg_parse_enum_or_int(&arg
);
1099 } else if (arg_match(&arg
, &target_bitrate
, argi
)) {
1100 config
->cfg
.rc_target_bitrate
= arg_parse_uint(&arg
);
1101 } else if (arg_match(&arg
, &min_quantizer
, argi
)) {
1102 config
->cfg
.rc_min_quantizer
= arg_parse_uint(&arg
);
1103 } else if (arg_match(&arg
, &max_quantizer
, argi
)) {
1104 config
->cfg
.rc_max_quantizer
= arg_parse_uint(&arg
);
1105 } else if (arg_match(&arg
, &undershoot_pct
, argi
)) {
1106 config
->cfg
.rc_undershoot_pct
= arg_parse_uint(&arg
);
1107 } else if (arg_match(&arg
, &overshoot_pct
, argi
)) {
1108 config
->cfg
.rc_overshoot_pct
= arg_parse_uint(&arg
);
1109 } else if (arg_match(&arg
, &buf_sz
, argi
)) {
1110 config
->cfg
.rc_buf_sz
= arg_parse_uint(&arg
);
1111 } else if (arg_match(&arg
, &buf_initial_sz
, argi
)) {
1112 config
->cfg
.rc_buf_initial_sz
= arg_parse_uint(&arg
);
1113 } else if (arg_match(&arg
, &buf_optimal_sz
, argi
)) {
1114 config
->cfg
.rc_buf_optimal_sz
= arg_parse_uint(&arg
);
1115 } else if (arg_match(&arg
, &bias_pct
, argi
)) {
1116 config
->cfg
.rc_2pass_vbr_bias_pct
= arg_parse_uint(&arg
);
1117 if (global
->passes
< 2)
1118 warn("option %s ignored in one-pass mode.\n", arg
.name
);
1119 } else if (arg_match(&arg
, &minsection_pct
, argi
)) {
1120 config
->cfg
.rc_2pass_vbr_minsection_pct
= arg_parse_uint(&arg
);
1122 if (global
->passes
< 2)
1123 warn("option %s ignored in one-pass mode.\n", arg
.name
);
1124 } else if (arg_match(&arg
, &maxsection_pct
, argi
)) {
1125 config
->cfg
.rc_2pass_vbr_maxsection_pct
= arg_parse_uint(&arg
);
1127 if (global
->passes
< 2)
1128 warn("option %s ignored in one-pass mode.\n", arg
.name
);
1129 } else if (arg_match(&arg
, &kf_min_dist
, argi
)) {
1130 config
->cfg
.kf_min_dist
= arg_parse_uint(&arg
);
1131 } else if (arg_match(&arg
, &kf_max_dist
, argi
)) {
1132 config
->cfg
.kf_max_dist
= arg_parse_uint(&arg
);
1133 config
->have_kf_max_dist
= 1;
1134 } else if (arg_match(&arg
, &kf_disabled
, argi
)) {
1135 config
->cfg
.kf_mode
= VPX_KF_DISABLED
;
1136 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1137 } else if (arg_match(&arg
, &test16bitinternalarg
, argi
)) {
1138 if (strcmp(global
->codec
->name
, "vp9") == 0) {
1139 test_16bit_internal
= 1;
1144 for (i
= 0; ctrl_args
[i
]; i
++) {
1145 if (arg_match(&arg
, ctrl_args
[i
], argi
)) {
1149 /* Point either to the next free element or the first
1150 * instance of this control.
1152 for (j
= 0; j
< config
->arg_ctrl_cnt
; j
++)
1153 if (ctrl_args_map
!= NULL
&&
1154 config
->arg_ctrls
[j
][0] == ctrl_args_map
[i
])
1158 assert(j
< (int)ARG_CTRL_CNT_MAX
);
1159 if (ctrl_args_map
!= NULL
&& j
< (int)ARG_CTRL_CNT_MAX
) {
1160 config
->arg_ctrls
[j
][0] = ctrl_args_map
[i
];
1161 config
->arg_ctrls
[j
][1] = arg_parse_enum_or_int(&arg
);
1162 if (j
== config
->arg_ctrl_cnt
)
1163 config
->arg_ctrl_cnt
++;
1172 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1173 if (strcmp(global
->codec
->name
, "vp9") == 0) {
1174 config
->use_16bit_internal
= test_16bit_internal
|
1175 (config
->cfg
.g_profile
> 1);
1178 return eos_mark_found
;
1182 #define FOREACH_STREAM(func) \
1184 struct stream_state *stream; \
1185 for (stream = streams; stream; stream = stream->next) { \
1191 static void validate_stream_config(const struct stream_state
*stream
,
1192 const struct VpxEncoderConfig
*global
) {
1193 const struct stream_state
*streami
;
1195 if (!stream
->config
.cfg
.g_w
|| !stream
->config
.cfg
.g_h
)
1196 fatal("Stream %d: Specify stream dimensions with --width (-w) "
1197 " and --height (-h)", stream
->index
);
1199 if (stream
->config
.cfg
.g_profile
!= 0 && !global
->experimental_bitstream
) {
1200 fatal("Stream %d: profile %d is experimental and requires the --%s flag",
1201 stream
->index
, stream
->config
.cfg
.g_profile
,
1202 experimental_bitstream
.long_name
);
1205 // Check that the codec bit depth is greater than the input bit depth.
1206 if (stream
->config
.cfg
.g_input_bit_depth
>
1207 (unsigned int)stream
->config
.cfg
.g_bit_depth
) {
1208 fatal("Stream %d: codec bit depth (%d) less than input bit depth (%d)",
1209 stream
->index
, (int)stream
->config
.cfg
.g_bit_depth
,
1210 stream
->config
.cfg
.g_input_bit_depth
);
1213 for (streami
= stream
; streami
; streami
= streami
->next
) {
1214 /* All streams require output files */
1215 if (!streami
->config
.out_fn
)
1216 fatal("Stream %d: Output file is required (specify with -o)",
1219 /* Check for two streams outputting to the same file */
1220 if (streami
!= stream
) {
1221 const char *a
= stream
->config
.out_fn
;
1222 const char *b
= streami
->config
.out_fn
;
1223 if (!strcmp(a
, b
) && strcmp(a
, "/dev/null") && strcmp(a
, ":nul"))
1224 fatal("Stream %d: duplicate output file (from stream %d)",
1225 streami
->index
, stream
->index
);
1228 /* Check for two streams sharing a stats file. */
1229 if (streami
!= stream
) {
1230 const char *a
= stream
->config
.stats_fn
;
1231 const char *b
= streami
->config
.stats_fn
;
1232 if (a
&& b
&& !strcmp(a
, b
))
1233 fatal("Stream %d: duplicate stats file (from stream %d)",
1234 streami
->index
, stream
->index
);
1237 #if CONFIG_FP_MB_STATS
1238 /* Check for two streams sharing a mb stats file. */
1239 if (streami
!= stream
) {
1240 const char *a
= stream
->config
.fpmb_stats_fn
;
1241 const char *b
= streami
->config
.fpmb_stats_fn
;
1242 if (a
&& b
&& !strcmp(a
, b
))
1243 fatal("Stream %d: duplicate mb stats file (from stream %d)",
1244 streami
->index
, stream
->index
);
1251 static void set_stream_dimensions(struct stream_state
*stream
,
1254 if (!stream
->config
.cfg
.g_w
) {
1255 if (!stream
->config
.cfg
.g_h
)
1256 stream
->config
.cfg
.g_w
= w
;
1258 stream
->config
.cfg
.g_w
= w
* stream
->config
.cfg
.g_h
/ h
;
1260 if (!stream
->config
.cfg
.g_h
) {
1261 stream
->config
.cfg
.g_h
= h
* stream
->config
.cfg
.g_w
/ w
;
1266 static void set_default_kf_interval(struct stream_state
*stream
,
1267 struct VpxEncoderConfig
*global
) {
1268 /* Use a max keyframe interval of 5 seconds, if none was
1269 * specified on the command line.
1271 if (!stream
->config
.have_kf_max_dist
) {
1272 double framerate
= (double)global
->framerate
.num
/ global
->framerate
.den
;
1273 if (framerate
> 0.0)
1274 stream
->config
.cfg
.kf_max_dist
= (unsigned int)(5.0 * framerate
);
1278 static const char* file_type_to_string(enum VideoFileType t
) {
1280 case FILE_TYPE_RAW
: return "RAW";
1281 case FILE_TYPE_Y4M
: return "Y4M";
1282 default: return "Other";
1286 static const char* image_format_to_string(vpx_img_fmt_t f
) {
1288 case VPX_IMG_FMT_I420
: return "I420";
1289 case VPX_IMG_FMT_I422
: return "I422";
1290 case VPX_IMG_FMT_I444
: return "I444";
1291 case VPX_IMG_FMT_YV12
: return "YV12";
1292 default: return "Other";
1296 static void show_stream_config(struct stream_state
*stream
,
1297 struct VpxEncoderConfig
*global
,
1298 struct VpxInputContext
*input
) {
1300 #define SHOW(field) \
1301 fprintf(stderr, " %-28s = %d\n", #field, stream->config.cfg.field)
1303 if (stream
->index
== 0) {
1304 fprintf(stderr
, "Codec: %s\n",
1305 vpx_codec_iface_name(global
->codec
->codec_interface()));
1306 fprintf(stderr
, "Source file: %s File Type: %s Format: %s\n",
1308 file_type_to_string(input
->file_type
),
1309 image_format_to_string(input
->fmt
));
1311 if (stream
->next
|| stream
->index
)
1312 fprintf(stderr
, "\nStream Index: %d\n", stream
->index
);
1313 fprintf(stderr
, "Destination file: %s\n", stream
->config
.out_fn
);
1314 fprintf(stderr
, "Encoder parameters:\n");
1322 SHOW(g_input_bit_depth
);
1323 SHOW(g_timebase
.num
);
1324 SHOW(g_timebase
.den
);
1325 SHOW(g_error_resilient
);
1327 SHOW(g_lag_in_frames
);
1328 SHOW(rc_dropframe_thresh
);
1329 SHOW(rc_resize_allowed
);
1330 SHOW(rc_scaled_width
);
1331 SHOW(rc_scaled_height
);
1332 SHOW(rc_resize_up_thresh
);
1333 SHOW(rc_resize_down_thresh
);
1335 SHOW(rc_target_bitrate
);
1336 SHOW(rc_min_quantizer
);
1337 SHOW(rc_max_quantizer
);
1338 SHOW(rc_undershoot_pct
);
1339 SHOW(rc_overshoot_pct
);
1341 SHOW(rc_buf_initial_sz
);
1342 SHOW(rc_buf_optimal_sz
);
1343 SHOW(rc_2pass_vbr_bias_pct
);
1344 SHOW(rc_2pass_vbr_minsection_pct
);
1345 SHOW(rc_2pass_vbr_maxsection_pct
);
1352 static void open_output_file(struct stream_state
*stream
,
1353 struct VpxEncoderConfig
*global
) {
1354 const char *fn
= stream
->config
.out_fn
;
1355 const struct vpx_codec_enc_cfg
*const cfg
= &stream
->config
.cfg
;
1357 if (cfg
->g_pass
== VPX_RC_FIRST_PASS
)
1360 stream
->file
= strcmp(fn
, "-") ? fopen(fn
, "wb") : set_binary_mode(stdout
);
1363 fatal("Failed to open output file");
1365 if (stream
->config
.write_webm
&& fseek(stream
->file
, 0, SEEK_CUR
))
1366 fatal("WebM output to pipes not supported.");
1369 if (stream
->config
.write_webm
) {
1370 stream
->ebml
.stream
= stream
->file
;
1371 write_webm_file_header(&stream
->ebml
, cfg
,
1373 stream
->config
.stereo_fmt
,
1374 global
->codec
->fourcc
);
1378 if (!stream
->config
.write_webm
) {
1379 ivf_write_file_header(stream
->file
, cfg
, global
->codec
->fourcc
, 0);
1384 static void close_output_file(struct stream_state
*stream
,
1385 unsigned int fourcc
) {
1386 const struct vpx_codec_enc_cfg
*const cfg
= &stream
->config
.cfg
;
1388 if (cfg
->g_pass
== VPX_RC_FIRST_PASS
)
1392 if (stream
->config
.write_webm
) {
1393 write_webm_file_footer(&stream
->ebml
);
1397 if (!stream
->config
.write_webm
) {
1398 if (!fseek(stream
->file
, 0, SEEK_SET
))
1399 ivf_write_file_header(stream
->file
, &stream
->config
.cfg
,
1401 stream
->frames_out
);
1404 fclose(stream
->file
);
1408 static void setup_pass(struct stream_state
*stream
,
1409 struct VpxEncoderConfig
*global
,
1411 if (stream
->config
.stats_fn
) {
1412 if (!stats_open_file(&stream
->stats
, stream
->config
.stats_fn
,
1414 fatal("Failed to open statistics store");
1416 if (!stats_open_mem(&stream
->stats
, pass
))
1417 fatal("Failed to open statistics store");
1420 #if CONFIG_FP_MB_STATS
1421 if (stream
->config
.fpmb_stats_fn
) {
1422 if (!stats_open_file(&stream
->fpmb_stats
,
1423 stream
->config
.fpmb_stats_fn
, pass
))
1424 fatal("Failed to open mb statistics store");
1426 if (!stats_open_mem(&stream
->fpmb_stats
, pass
))
1427 fatal("Failed to open mb statistics store");
1431 stream
->config
.cfg
.g_pass
= global
->passes
== 2
1432 ? pass
? VPX_RC_LAST_PASS
: VPX_RC_FIRST_PASS
1435 stream
->config
.cfg
.rc_twopass_stats_in
= stats_get(&stream
->stats
);
1436 #if CONFIG_FP_MB_STATS
1437 stream
->config
.cfg
.rc_firstpass_mb_stats_in
=
1438 stats_get(&stream
->fpmb_stats
);
1442 stream
->cx_time
= 0;
1444 stream
->frames_out
= 0;
1448 static void initialize_encoder(struct stream_state
*stream
,
1449 struct VpxEncoderConfig
*global
) {
1453 flags
|= global
->show_psnr
? VPX_CODEC_USE_PSNR
: 0;
1454 flags
|= global
->out_part
? VPX_CODEC_USE_OUTPUT_PARTITION
: 0;
1455 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1456 flags
|= stream
->config
.use_16bit_internal
? VPX_CODEC_USE_HIGHBITDEPTH
: 0;
1459 /* Construct Encoder Context */
1460 vpx_codec_enc_init(&stream
->encoder
, global
->codec
->codec_interface(),
1461 &stream
->config
.cfg
, flags
);
1462 ctx_exit_on_error(&stream
->encoder
, "Failed to initialize encoder");
1464 /* Note that we bypass the vpx_codec_control wrapper macro because
1465 * we're being clever to store the control IDs in an array. Real
1466 * applications will want to make use of the enumerations directly
1468 for (i
= 0; i
< stream
->config
.arg_ctrl_cnt
; i
++) {
1469 int ctrl
= stream
->config
.arg_ctrls
[i
][0];
1470 int value
= stream
->config
.arg_ctrls
[i
][1];
1471 if (vpx_codec_control_(&stream
->encoder
, ctrl
, value
))
1472 fprintf(stderr
, "Error: Tried to set control %d = %d\n",
1475 ctx_exit_on_error(&stream
->encoder
, "Failed to control codec");
1479 if (global
->test_decode
!= TEST_DECODE_OFF
) {
1480 const VpxInterface
*decoder
= get_vpx_decoder_by_name(global
->codec
->name
);
1481 vpx_codec_dec_init(&stream
->decoder
, decoder
->codec_interface(), NULL
, 0);
1487 static void encode_frame(struct stream_state
*stream
,
1488 struct VpxEncoderConfig
*global
,
1489 struct vpx_image
*img
,
1490 unsigned int frames_in
) {
1491 vpx_codec_pts_t frame_start
, next_frame_start
;
1492 struct vpx_codec_enc_cfg
*cfg
= &stream
->config
.cfg
;
1493 struct vpx_usec_timer timer
;
1495 frame_start
= (cfg
->g_timebase
.den
* (int64_t)(frames_in
- 1)
1496 * global
->framerate
.den
)
1497 / cfg
->g_timebase
.num
/ global
->framerate
.num
;
1498 next_frame_start
= (cfg
->g_timebase
.den
* (int64_t)(frames_in
)
1499 * global
->framerate
.den
)
1500 / cfg
->g_timebase
.num
/ global
->framerate
.num
;
1502 /* Scale if necessary */
1503 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1505 if ((img
->fmt
& VPX_IMG_FMT_HIGHBITDEPTH
) &&
1506 (img
->d_w
!= cfg
->g_w
|| img
->d_h
!= cfg
->g_h
)) {
1507 if (img
->fmt
!= VPX_IMG_FMT_I42016
) {
1508 fprintf(stderr
, "%s can only scale 4:2:0 inputs\n", exec_name
);
1513 stream
->img
= vpx_img_alloc(NULL
, VPX_IMG_FMT_I42016
,
1514 cfg
->g_w
, cfg
->g_h
, 16);
1516 I420Scale_16((uint16
*)img
->planes
[VPX_PLANE_Y
],
1517 img
->stride
[VPX_PLANE_Y
]/2,
1518 (uint16
*)img
->planes
[VPX_PLANE_U
],
1519 img
->stride
[VPX_PLANE_U
]/2,
1520 (uint16
*)img
->planes
[VPX_PLANE_V
],
1521 img
->stride
[VPX_PLANE_V
]/2,
1523 (uint16
*)stream
->img
->planes
[VPX_PLANE_Y
],
1524 stream
->img
->stride
[VPX_PLANE_Y
]/2,
1525 (uint16
*)stream
->img
->planes
[VPX_PLANE_U
],
1526 stream
->img
->stride
[VPX_PLANE_U
]/2,
1527 (uint16
*)stream
->img
->planes
[VPX_PLANE_V
],
1528 stream
->img
->stride
[VPX_PLANE_V
]/2,
1529 stream
->img
->d_w
, stream
->img
->d_h
,
1533 stream
->encoder
.err
= 1;
1534 ctx_exit_on_error(&stream
->encoder
,
1535 "Stream %d: Failed to encode frame.\n"
1536 "Scaling disabled in this configuration. \n"
1537 "To enable, configure with --enable-libyuv\n",
1543 if (img
&& (img
->d_w
!= cfg
->g_w
|| img
->d_h
!= cfg
->g_h
)) {
1544 if (img
->fmt
!= VPX_IMG_FMT_I420
&& img
->fmt
!= VPX_IMG_FMT_YV12
) {
1545 fprintf(stderr
, "%s can only scale 4:2:0 8bpp inputs\n", exec_name
);
1550 stream
->img
= vpx_img_alloc(NULL
, VPX_IMG_FMT_I420
,
1551 cfg
->g_w
, cfg
->g_h
, 16);
1552 I420Scale(img
->planes
[VPX_PLANE_Y
], img
->stride
[VPX_PLANE_Y
],
1553 img
->planes
[VPX_PLANE_U
], img
->stride
[VPX_PLANE_U
],
1554 img
->planes
[VPX_PLANE_V
], img
->stride
[VPX_PLANE_V
],
1556 stream
->img
->planes
[VPX_PLANE_Y
],
1557 stream
->img
->stride
[VPX_PLANE_Y
],
1558 stream
->img
->planes
[VPX_PLANE_U
],
1559 stream
->img
->stride
[VPX_PLANE_U
],
1560 stream
->img
->planes
[VPX_PLANE_V
],
1561 stream
->img
->stride
[VPX_PLANE_V
],
1562 stream
->img
->d_w
, stream
->img
->d_h
,
1566 stream
->encoder
.err
= 1;
1567 ctx_exit_on_error(&stream
->encoder
,
1568 "Stream %d: Failed to encode frame.\n"
1569 "Scaling disabled in this configuration. \n"
1570 "To enable, configure with --enable-libyuv\n",
1575 vpx_usec_timer_start(&timer
);
1576 vpx_codec_encode(&stream
->encoder
, img
, frame_start
,
1577 (unsigned long)(next_frame_start
- frame_start
),
1578 0, global
->deadline
);
1579 vpx_usec_timer_mark(&timer
);
1580 stream
->cx_time
+= vpx_usec_timer_elapsed(&timer
);
1581 ctx_exit_on_error(&stream
->encoder
, "Stream %d: Failed to encode frame",
1586 static void update_quantizer_histogram(struct stream_state
*stream
) {
1587 if (stream
->config
.cfg
.g_pass
!= VPX_RC_FIRST_PASS
) {
1590 vpx_codec_control(&stream
->encoder
, VP8E_GET_LAST_QUANTIZER_64
, &q
);
1591 ctx_exit_on_error(&stream
->encoder
, "Failed to read quantizer");
1592 stream
->counts
[q
]++;
1597 static void get_cx_data(struct stream_state
*stream
,
1598 struct VpxEncoderConfig
*global
,
1600 const vpx_codec_cx_pkt_t
*pkt
;
1601 const struct vpx_codec_enc_cfg
*cfg
= &stream
->config
.cfg
;
1602 vpx_codec_iter_t iter
= NULL
;
1605 while ((pkt
= vpx_codec_get_cx_data(&stream
->encoder
, &iter
))) {
1606 static size_t fsize
= 0;
1607 static int64_t ivf_header_pos
= 0;
1609 switch (pkt
->kind
) {
1610 case VPX_CODEC_CX_FRAME_PKT
:
1611 if (!(pkt
->data
.frame
.flags
& VPX_FRAME_IS_FRAGMENT
)) {
1612 stream
->frames_out
++;
1615 fprintf(stderr
, " %6luF", (unsigned long)pkt
->data
.frame
.sz
);
1617 update_rate_histogram(stream
->rate_hist
, cfg
, pkt
);
1619 if (stream
->config
.write_webm
) {
1620 write_webm_block(&stream
->ebml
, cfg
, pkt
);
1623 if (!stream
->config
.write_webm
) {
1624 if (pkt
->data
.frame
.partition_id
<= 0) {
1625 ivf_header_pos
= ftello(stream
->file
);
1626 fsize
= pkt
->data
.frame
.sz
;
1628 ivf_write_frame_header(stream
->file
, pkt
->data
.frame
.pts
, fsize
);
1630 fsize
+= pkt
->data
.frame
.sz
;
1632 if (!(pkt
->data
.frame
.flags
& VPX_FRAME_IS_FRAGMENT
)) {
1633 const int64_t currpos
= ftello(stream
->file
);
1634 fseeko(stream
->file
, ivf_header_pos
, SEEK_SET
);
1635 ivf_write_frame_size(stream
->file
, fsize
);
1636 fseeko(stream
->file
, currpos
, SEEK_SET
);
1640 (void) fwrite(pkt
->data
.frame
.buf
, 1, pkt
->data
.frame
.sz
,
1643 stream
->nbytes
+= pkt
->data
.raw
.sz
;
1647 if (global
->test_decode
!= TEST_DECODE_OFF
&& !stream
->mismatch_seen
) {
1648 vpx_codec_decode(&stream
->decoder
, pkt
->data
.frame
.buf
,
1649 (unsigned int)pkt
->data
.frame
.sz
, NULL
, 0);
1650 if (stream
->decoder
.err
) {
1651 warn_or_exit_on_error(&stream
->decoder
,
1652 global
->test_decode
== TEST_DECODE_FATAL
,
1653 "Failed to decode frame %d in stream %d",
1654 stream
->frames_out
+ 1, stream
->index
);
1655 stream
->mismatch_seen
= stream
->frames_out
+ 1;
1660 case VPX_CODEC_STATS_PKT
:
1661 stream
->frames_out
++;
1662 stats_write(&stream
->stats
,
1663 pkt
->data
.twopass_stats
.buf
,
1664 pkt
->data
.twopass_stats
.sz
);
1665 stream
->nbytes
+= pkt
->data
.raw
.sz
;
1667 #if CONFIG_FP_MB_STATS
1668 case VPX_CODEC_FPMB_STATS_PKT
:
1669 stats_write(&stream
->fpmb_stats
,
1670 pkt
->data
.firstpass_mb_stats
.buf
,
1671 pkt
->data
.firstpass_mb_stats
.sz
);
1672 stream
->nbytes
+= pkt
->data
.raw
.sz
;
1675 case VPX_CODEC_PSNR_PKT
:
1677 if (global
->show_psnr
) {
1680 stream
->psnr_sse_total
+= pkt
->data
.psnr
.sse
[0];
1681 stream
->psnr_samples_total
+= pkt
->data
.psnr
.samples
[0];
1682 for (i
= 0; i
< 4; i
++) {
1684 fprintf(stderr
, "%.3f ", pkt
->data
.psnr
.psnr
[i
]);
1685 stream
->psnr_totals
[i
] += pkt
->data
.psnr
.psnr
[i
];
1687 stream
->psnr_count
++;
1698 static void show_psnr(struct stream_state
*stream
) {
1702 if (!stream
->psnr_count
)
1705 fprintf(stderr
, "Stream %d PSNR (Overall/Avg/Y/U/V)", stream
->index
);
1706 ovpsnr
= sse_to_psnr((double)stream
->psnr_samples_total
, 255.0,
1707 (double)stream
->psnr_sse_total
);
1708 fprintf(stderr
, " %.3f", ovpsnr
);
1710 for (i
= 0; i
< 4; i
++) {
1711 fprintf(stderr
, " %.3f", stream
->psnr_totals
[i
] / stream
->psnr_count
);
1713 fprintf(stderr
, "\n");
1717 static float usec_to_fps(uint64_t usec
, unsigned int frames
) {
1718 return (float)(usec
> 0 ? frames
* 1000000.0 / (float)usec
: 0);
1721 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1722 static void high_img_upshift(vpx_image_t
*dst
, vpx_image_t
*src
,
1724 // Note the offset is 1 less than half
1725 const int offset
= input_shift
> 0 ? (1 << (input_shift
- 1)) - 1 : 0;
1727 if (dst
->w
!= src
->w
|| dst
->h
!= src
->h
||
1728 dst
->x_chroma_shift
!= src
->x_chroma_shift
||
1729 dst
->y_chroma_shift
!= src
->y_chroma_shift
||
1730 dst
->fmt
!= src
->fmt
|| input_shift
< 0) {
1731 fatal("Unsupported image conversion");
1734 case VPX_IMG_FMT_I42016
:
1735 case VPX_IMG_FMT_I42216
:
1736 case VPX_IMG_FMT_I44416
:
1739 fatal("Unsupported image conversion");
1742 for (plane
= 0; plane
< 3; plane
++) {
1747 w
>>= src
->x_chroma_shift
;
1748 h
>>= src
->y_chroma_shift
;
1750 for (y
= 0; y
< h
; y
++) {
1751 uint16_t *p_src
= (uint16_t *)(src
->planes
[plane
] +
1752 y
* src
->stride
[plane
]);
1753 uint16_t *p_dst
= (uint16_t *)(dst
->planes
[plane
] +
1754 y
* dst
->stride
[plane
]);
1755 for (x
= 0; x
< w
; x
++)
1756 *p_dst
++ = (*p_src
++ << input_shift
) + offset
;
1761 static void low_img_upshift(vpx_image_t
*dst
, vpx_image_t
*src
,
1763 // Note the offset is 1 less than half
1764 const int offset
= input_shift
> 0 ? (1 << (input_shift
- 1)) - 1 : 0;
1766 if (dst
->w
!= src
->w
|| dst
->h
!= src
->h
||
1767 dst
->x_chroma_shift
!= src
->x_chroma_shift
||
1768 dst
->y_chroma_shift
!= src
->y_chroma_shift
||
1769 dst
->fmt
!= src
->fmt
+ VPX_IMG_FMT_HIGHBITDEPTH
||
1771 fatal("Unsupported image conversion");
1774 case VPX_IMG_FMT_I420
:
1775 case VPX_IMG_FMT_I422
:
1776 case VPX_IMG_FMT_I444
:
1779 fatal("Unsupported image conversion");
1782 for (plane
= 0; plane
< 3; plane
++) {
1787 w
>>= src
->x_chroma_shift
;
1788 h
>>= src
->y_chroma_shift
;
1790 for (y
= 0; y
< h
; y
++) {
1791 uint8_t *p_src
= src
->planes
[plane
] + y
* src
->stride
[plane
];
1792 uint16_t *p_dst
= (uint16_t *)(dst
->planes
[plane
] +
1793 y
* dst
->stride
[plane
]);
1794 for (x
= 0; x
< w
; x
++) {
1795 *p_dst
++ = (*p_src
++ << input_shift
) + offset
;
1801 static void img_upshift(vpx_image_t
*dst
, vpx_image_t
*src
,
1803 if (src
->fmt
& VPX_IMG_FMT_HIGHBITDEPTH
) {
1804 high_img_upshift(dst
, src
, input_shift
);
1806 low_img_upshift(dst
, src
, input_shift
);
1810 static void img_cast_16_to_8(vpx_image_t
*dst
, vpx_image_t
*src
) {
1812 if (dst
->fmt
+ VPX_IMG_FMT_HIGHBITDEPTH
!= src
->fmt
||
1813 dst
->d_w
!= src
->d_w
|| dst
->d_h
!= src
->d_h
||
1814 dst
->x_chroma_shift
!= src
->x_chroma_shift
||
1815 dst
->y_chroma_shift
!= src
->y_chroma_shift
) {
1816 fatal("Unsupported image conversion");
1819 case VPX_IMG_FMT_I420
:
1820 case VPX_IMG_FMT_I422
:
1821 case VPX_IMG_FMT_I444
:
1824 fatal("Unsupported image conversion");
1827 for (plane
= 0; plane
< 3; plane
++) {
1832 w
>>= src
->x_chroma_shift
;
1833 h
>>= src
->y_chroma_shift
;
1835 for (y
= 0; y
< h
; y
++) {
1836 uint16_t *p_src
= (uint16_t *)(src
->planes
[plane
] +
1837 y
* src
->stride
[plane
]);
1838 uint8_t *p_dst
= dst
->planes
[plane
] + y
* dst
->stride
[plane
];
1839 for (x
= 0; x
< w
; x
++) {
1840 *p_dst
++ = *p_src
++;
1847 static void test_decode(struct stream_state
*stream
,
1848 enum TestDecodeFatality fatal
,
1849 const VpxInterface
*codec
) {
1850 vpx_image_t enc_img
, dec_img
;
1852 if (stream
->mismatch_seen
)
1855 /* Get the internal reference frame */
1856 if (strcmp(codec
->name
, "vp8") == 0) {
1857 struct vpx_ref_frame ref_enc
, ref_dec
;
1860 width
= (stream
->config
.cfg
.g_w
+ 15) & ~15;
1861 height
= (stream
->config
.cfg
.g_h
+ 15) & ~15;
1862 vpx_img_alloc(&ref_enc
.img
, VPX_IMG_FMT_I420
, width
, height
, 1);
1863 enc_img
= ref_enc
.img
;
1864 vpx_img_alloc(&ref_dec
.img
, VPX_IMG_FMT_I420
, width
, height
, 1);
1865 dec_img
= ref_dec
.img
;
1867 ref_enc
.frame_type
= VP8_LAST_FRAME
;
1868 ref_dec
.frame_type
= VP8_LAST_FRAME
;
1869 vpx_codec_control(&stream
->encoder
, VP8_COPY_REFERENCE
, &ref_enc
);
1870 vpx_codec_control(&stream
->decoder
, VP8_COPY_REFERENCE
, &ref_dec
);
1872 struct vp9_ref_frame ref_enc
, ref_dec
;
1876 vpx_codec_control(&stream
->encoder
, VP9_GET_REFERENCE
, &ref_enc
);
1877 enc_img
= ref_enc
.img
;
1878 vpx_codec_control(&stream
->decoder
, VP9_GET_REFERENCE
, &ref_dec
);
1879 dec_img
= ref_dec
.img
;
1880 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1881 if ((enc_img
.fmt
& VPX_IMG_FMT_HIGHBITDEPTH
) !=
1882 (dec_img
.fmt
& VPX_IMG_FMT_HIGHBITDEPTH
)) {
1883 if (enc_img
.fmt
& VPX_IMG_FMT_HIGHBITDEPTH
) {
1884 vpx_img_alloc(&enc_img
, enc_img
.fmt
- VPX_IMG_FMT_HIGHBITDEPTH
,
1885 enc_img
.d_w
, enc_img
.d_h
, 16);
1886 img_cast_16_to_8(&enc_img
, &ref_enc
.img
);
1888 if (dec_img
.fmt
& VPX_IMG_FMT_HIGHBITDEPTH
) {
1889 vpx_img_alloc(&dec_img
, dec_img
.fmt
- VPX_IMG_FMT_HIGHBITDEPTH
,
1890 dec_img
.d_w
, dec_img
.d_h
, 16);
1891 img_cast_16_to_8(&dec_img
, &ref_dec
.img
);
1896 ctx_exit_on_error(&stream
->encoder
, "Failed to get encoder reference frame");
1897 ctx_exit_on_error(&stream
->decoder
, "Failed to get decoder reference frame");
1899 if (!compare_img(&enc_img
, &dec_img
)) {
1900 int y
[4], u
[4], v
[4];
1901 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1902 if (enc_img
.fmt
& VPX_IMG_FMT_HIGHBITDEPTH
) {
1903 find_mismatch_high(&enc_img
, &dec_img
, y
, u
, v
);
1905 find_mismatch(&enc_img
, &dec_img
, y
, u
, v
);
1908 find_mismatch(&enc_img
, &dec_img
, y
, u
, v
);
1910 stream
->decoder
.err
= 1;
1911 warn_or_exit_on_error(&stream
->decoder
, fatal
== TEST_DECODE_FATAL
,
1912 "Stream %d: Encode/decode mismatch on frame %d at"
1913 " Y[%d, %d] {%d/%d},"
1914 " U[%d, %d] {%d/%d},"
1915 " V[%d, %d] {%d/%d}",
1916 stream
->index
, stream
->frames_out
,
1917 y
[0], y
[1], y
[2], y
[3],
1918 u
[0], u
[1], u
[2], u
[3],
1919 v
[0], v
[1], v
[2], v
[3]);
1920 stream
->mismatch_seen
= stream
->frames_out
;
1923 vpx_img_free(&enc_img
);
1924 vpx_img_free(&dec_img
);
1928 static void print_time(const char *label
, int64_t etl
) {
1935 etl
-= hours
* 3600;
1940 fprintf(stderr
, "[%3s %2"PRId64
":%02"PRId64
":%02"PRId64
"] ",
1941 label
, hours
, mins
, secs
);
1943 fprintf(stderr
, "[%3s unknown] ", label
);
1948 int main(int argc
, const char **argv_
) {
1951 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1952 vpx_image_t raw_shift
;
1953 int allocated_raw_shift
= 0;
1954 int use_16bit_internal
= 0;
1955 int input_shift
= 0;
1957 int frame_avail
, got_data
;
1959 struct VpxInputContext input
;
1960 struct VpxEncoderConfig global
;
1961 struct stream_state
*streams
= NULL
;
1962 char **argv
, **argi
;
1963 uint64_t cx_time
= 0;
1967 memset(&input
, 0, sizeof(input
));
1968 exec_name
= argv_
[0];
1973 /* Setup default input stream settings */
1974 input
.framerate
.numerator
= 30;
1975 input
.framerate
.denominator
= 1;
1976 input
.only_i420
= 1;
1977 input
.bit_depth
= 0;
1979 /* First parse the global configuration values, because we want to apply
1980 * other parameters on top of the default configuration provided by the
1983 argv
= argv_dup(argc
- 1, argv_
+ 1);
1984 parse_global_config(&global
, argv
);
1986 switch (global
.color_type
) {
1988 input
.fmt
= VPX_IMG_FMT_I420
;
1991 input
.fmt
= VPX_IMG_FMT_I422
;
1994 input
.fmt
= VPX_IMG_FMT_I444
;
1997 input
.fmt
= VPX_IMG_FMT_YV12
;
2002 /* Now parse each stream's parameters. Using a local scope here
2003 * due to the use of 'stream' as loop variable in FOREACH_STREAM
2006 struct stream_state
*stream
= NULL
;
2009 stream
= new_stream(&global
, stream
);
2013 } while (parse_stream_params(&global
, stream
, argv
));
2016 /* Check for unrecognized options */
2017 for (argi
= argv
; *argi
; argi
++)
2018 if (argi
[0][0] == '-' && argi
[0][1])
2019 die("Error: Unrecognized option %s\n", *argi
);
2021 FOREACH_STREAM(check_encoder_config(global
.disable_warning_prompt
,
2022 &global
, &stream
->config
.cfg
););
2024 /* Handle non-option arguments */
2025 input
.filename
= argv
[0];
2027 if (!input
.filename
)
2030 /* Decide if other chroma subsamplings than 4:2:0 are supported */
2031 if (global
.codec
->fourcc
== VP9_FOURCC
)
2032 input
.only_i420
= 0;
2034 for (pass
= global
.pass
? global
.pass
- 1 : 0; pass
< global
.passes
; pass
++) {
2035 int frames_in
= 0, seen_frames
= 0;
2036 int64_t estimated_time_left
= -1;
2037 int64_t average_rate
= -1;
2038 int64_t lagged_count
= 0;
2040 open_input_file(&input
);
2042 /* If the input file doesn't specify its w/h (raw files), try to get
2043 * the data from the first stream's configuration.
2045 if (!input
.width
|| !input
.height
)
2047 if (stream
->config
.cfg
.g_w
&& stream
->config
.cfg
.g_h
) {
2048 input
.width
= stream
->config
.cfg
.g_w
;
2049 input
.height
= stream
->config
.cfg
.g_h
;
2054 /* Update stream configurations from the input file's parameters */
2055 if (!input
.width
|| !input
.height
)
2056 fatal("Specify stream dimensions with --width (-w) "
2057 " and --height (-h)");
2059 /* If input file does not specify bit-depth but input-bit-depth parameter
2060 * exists, assume that to be the input bit-depth. However, if the
2061 * input-bit-depth paramter does not exist, assume the input bit-depth
2062 * to be the same as the codec bit-depth.
2064 if (!input
.bit_depth
) {
2066 if (stream
->config
.cfg
.g_input_bit_depth
)
2067 input
.bit_depth
= stream
->config
.cfg
.g_input_bit_depth
;
2069 input
.bit_depth
= stream
->config
.cfg
.g_input_bit_depth
=
2070 (int)stream
->config
.cfg
.g_bit_depth
;
2072 if (input
.bit_depth
> 8) input
.fmt
|= VPX_IMG_FMT_HIGHBITDEPTH
;
2075 stream
->config
.cfg
.g_input_bit_depth
= input
.bit_depth
;
2079 FOREACH_STREAM(set_stream_dimensions(stream
, input
.width
, input
.height
));
2080 FOREACH_STREAM(validate_stream_config(stream
, &global
));
2082 /* Ensure that --passes and --pass are consistent. If --pass is set and
2083 * --passes=2, ensure --fpf was set.
2085 if (global
.pass
&& global
.passes
== 2)
2087 if (!stream
->config
.stats_fn
)
2088 die("Stream %d: Must specify --fpf when --pass=%d"
2089 " and --passes=2\n", stream
->index
, global
.pass
);
2094 stream
->config
.write_webm
= 0;
2095 warn("vpxenc was compiled without WebM container support."
2096 "Producing IVF output");
2100 /* Use the frame rate from the file only if none was specified
2101 * on the command-line.
2103 if (!global
.have_framerate
) {
2104 global
.framerate
.num
= input
.framerate
.numerator
;
2105 global
.framerate
.den
= input
.framerate
.denominator
;
2108 FOREACH_STREAM(set_default_kf_interval(stream
, &global
));
2110 /* Show configuration */
2111 if (global
.verbose
&& pass
== 0)
2112 FOREACH_STREAM(show_stream_config(stream
, &global
, &input
));
2114 if (pass
== (global
.pass
? global
.pass
- 1 : 0)) {
2115 if (input
.file_type
== FILE_TYPE_Y4M
)
2116 /*The Y4M reader does its own allocation.
2117 Just initialize this here to avoid problems if we never read any
2119 memset(&raw
, 0, sizeof(raw
));
2121 vpx_img_alloc(&raw
, input
.fmt
, input
.width
, input
.height
, 32);
2123 FOREACH_STREAM(stream
->rate_hist
=
2124 init_rate_histogram(&stream
->config
.cfg
,
2125 &global
.framerate
));
2128 FOREACH_STREAM(setup_pass(stream
, &global
, pass
));
2129 FOREACH_STREAM(open_output_file(stream
, &global
));
2130 FOREACH_STREAM(initialize_encoder(stream
, &global
));
2132 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
2133 if (strcmp(global
.codec
->name
, "vp9") == 0) {
2134 // Check to see if at least one stream uses 16 bit internal.
2135 // Currently assume that the bit_depths for all streams using
2136 // highbitdepth are the same.
2138 if (stream
->config
.use_16bit_internal
) {
2139 use_16bit_internal
= 1;
2141 if (stream
->config
.cfg
.g_profile
== 0) {
2144 input_shift
= (int)stream
->config
.cfg
.g_bit_depth
-
2145 stream
->config
.cfg
.g_input_bit_depth
;
2154 while (frame_avail
|| got_data
) {
2155 struct vpx_usec_timer timer
;
2157 if (!global
.limit
|| frames_in
< global
.limit
) {
2158 frame_avail
= read_frame(&input
, &raw
);
2162 seen_frames
= frames_in
> global
.skip_frames
?
2163 frames_in
- global
.skip_frames
: 0;
2165 if (!global
.quiet
) {
2166 float fps
= usec_to_fps(cx_time
, seen_frames
);
2167 fprintf(stderr
, "\rPass %d/%d ", pass
+ 1, global
.passes
);
2169 if (stream_cnt
== 1)
2171 "frame %4d/%-4d %7"PRId64
"B ",
2172 frames_in
, streams
->frames_out
, (int64_t)streams
->nbytes
);
2174 fprintf(stderr
, "frame %4d ", frames_in
);
2176 fprintf(stderr
, "%7"PRId64
" %s %.2f %s ",
2177 cx_time
> 9999999 ? cx_time
/ 1000 : cx_time
,
2178 cx_time
> 9999999 ? "ms" : "us",
2179 fps
>= 1.0 ? fps
: fps
* 60,
2180 fps
>= 1.0 ? "fps" : "fpm");
2181 print_time("ETA", estimated_time_left
);
2187 if (frames_in
> global
.skip_frames
) {
2188 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
2189 vpx_image_t
*frame_to_encode
;
2190 if (input_shift
|| (use_16bit_internal
&& input
.bit_depth
== 8)) {
2191 assert(use_16bit_internal
);
2192 // Input bit depth and stream bit depth do not match, so up
2193 // shift frame to stream bit depth
2194 if (!allocated_raw_shift
) {
2195 vpx_img_alloc(&raw_shift
, raw
.fmt
| VPX_IMG_FMT_HIGHBITDEPTH
,
2196 input
.width
, input
.height
, 32);
2197 allocated_raw_shift
= 1;
2199 img_upshift(&raw_shift
, &raw
, input_shift
);
2200 frame_to_encode
= &raw_shift
;
2202 frame_to_encode
= &raw
;
2204 vpx_usec_timer_start(&timer
);
2205 if (use_16bit_internal
) {
2206 assert(frame_to_encode
->fmt
& VPX_IMG_FMT_HIGHBITDEPTH
);
2208 if (stream
->config
.use_16bit_internal
)
2209 encode_frame(stream
, &global
,
2210 frame_avail
? frame_to_encode
: NULL
,
2216 assert((frame_to_encode
->fmt
& VPX_IMG_FMT_HIGHBITDEPTH
) == 0);
2217 FOREACH_STREAM(encode_frame(stream
, &global
,
2218 frame_avail
? frame_to_encode
: NULL
,
2222 vpx_usec_timer_start(&timer
);
2223 FOREACH_STREAM(encode_frame(stream
, &global
,
2224 frame_avail
? &raw
: NULL
,
2227 vpx_usec_timer_mark(&timer
);
2228 cx_time
+= vpx_usec_timer_elapsed(&timer
);
2230 FOREACH_STREAM(update_quantizer_histogram(stream
));
2233 FOREACH_STREAM(get_cx_data(stream
, &global
, &got_data
));
2235 if (!got_data
&& input
.length
&& streams
!= NULL
&&
2236 !streams
->frames_out
) {
2237 lagged_count
= global
.limit
? seen_frames
: ftello(input
.file
);
2238 } else if (input
.length
) {
2243 const int64_t frame_in_lagged
= (seen_frames
- lagged_count
) * 1000;
2245 rate
= cx_time
? frame_in_lagged
* (int64_t)1000000 / cx_time
: 0;
2246 remaining
= 1000 * (global
.limit
- global
.skip_frames
2247 - seen_frames
+ lagged_count
);
2249 const int64_t input_pos
= ftello(input
.file
);
2250 const int64_t input_pos_lagged
= input_pos
- lagged_count
;
2251 const int64_t limit
= input
.length
;
2253 rate
= cx_time
? input_pos_lagged
* (int64_t)1000000 / cx_time
: 0;
2254 remaining
= limit
- input_pos
+ lagged_count
;
2257 average_rate
= (average_rate
<= 0)
2259 : (average_rate
* 7 + rate
) / 8;
2260 estimated_time_left
= average_rate
? remaining
/ average_rate
: -1;
2263 if (got_data
&& global
.test_decode
!= TEST_DECODE_OFF
)
2264 FOREACH_STREAM(test_decode(stream
, global
.test_decode
, global
.codec
));
2269 fprintf(stderr
, "\033[K");
2273 fprintf(stderr
, "\n");
2276 FOREACH_STREAM(fprintf(
2278 "\rPass %d/%d frame %4d/%-4d %7"PRId64
"B %7lub/f %7"PRId64
"b/s"
2279 " %7"PRId64
" %s (%.2f fps)\033[K\n", pass
+ 1,
2280 global
.passes
, frames_in
, stream
->frames_out
, (int64_t)stream
->nbytes
,
2281 seen_frames
? (unsigned long)(stream
->nbytes
* 8 / seen_frames
) : 0,
2282 seen_frames
? (int64_t)stream
->nbytes
* 8
2283 * (int64_t)global
.framerate
.num
/ global
.framerate
.den
2286 stream
->cx_time
> 9999999 ? stream
->cx_time
/ 1000 : stream
->cx_time
,
2287 stream
->cx_time
> 9999999 ? "ms" : "us",
2288 usec_to_fps(stream
->cx_time
, seen_frames
));
2291 if (global
.show_psnr
)
2292 FOREACH_STREAM(show_psnr(stream
));
2294 FOREACH_STREAM(vpx_codec_destroy(&stream
->encoder
));
2296 if (global
.test_decode
!= TEST_DECODE_OFF
) {
2297 FOREACH_STREAM(vpx_codec_destroy(&stream
->decoder
));
2300 close_input_file(&input
);
2302 if (global
.test_decode
== TEST_DECODE_FATAL
) {
2303 FOREACH_STREAM(res
|= stream
->mismatch_seen
);
2305 FOREACH_STREAM(close_output_file(stream
, global
.codec
->fourcc
));
2307 FOREACH_STREAM(stats_close(&stream
->stats
, global
.passes
- 1));
2309 #if CONFIG_FP_MB_STATS
2310 FOREACH_STREAM(stats_close(&stream
->fpmb_stats
, global
.passes
- 1));
2317 if (global
.show_q_hist_buckets
)
2318 FOREACH_STREAM(show_q_histogram(stream
->counts
,
2319 global
.show_q_hist_buckets
));
2321 if (global
.show_rate_hist_buckets
)
2322 FOREACH_STREAM(show_rate_histogram(stream
->rate_hist
,
2323 &stream
->config
.cfg
,
2324 global
.show_rate_hist_buckets
));
2325 FOREACH_STREAM(destroy_rate_histogram(stream
->rate_hist
));
2327 #if CONFIG_INTERNAL_STATS
2328 /* TODO(jkoleszar): This doesn't belong in this executable. Do it for now,
2329 * to match some existing utilities.
2331 if (!(global
.pass
== 1 && global
.passes
== 2))
2333 FILE *f
= fopen("opsnr.stt", "a");
2334 if (stream
->mismatch_seen
) {
2335 fprintf(f
, "First mismatch occurred in frame %d\n",
2336 stream
->mismatch_seen
);
2338 fprintf(f
, "No mismatch detected in recon buffers\n");
2344 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
2345 if (allocated_raw_shift
)
2346 vpx_img_free(&raw_shift
);
2351 return res
? EXIT_FAILURE
: EXIT_SUCCESS
;