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(
132 "D", "debug", 0, "Debug mode (makes output deterministic)");
133 static const arg_def_t outputfile
= ARG_DEF(
134 "o", "output", 1, "Output filename");
135 static const arg_def_t use_yv12
= ARG_DEF(
136 NULL
, "yv12", 0, "Input file is YV12 ");
137 static const arg_def_t use_i420
= ARG_DEF(
138 NULL
, "i420", 0, "Input file is I420 (default)");
139 static const arg_def_t use_i422
= ARG_DEF(
140 NULL
, "i422", 0, "Input file is I422");
141 static const arg_def_t use_i444
= ARG_DEF(
142 NULL
, "i444", 0, "Input file is I444");
143 static const arg_def_t use_i440
= ARG_DEF(
144 NULL
, "i440", 0, "Input file is I440");
145 static const arg_def_t codecarg
= ARG_DEF(
146 NULL
, "codec", 1, "Codec to use");
147 static const arg_def_t passes
= ARG_DEF(
148 "p", "passes", 1, "Number of passes (1/2)");
149 static const arg_def_t pass_arg
= ARG_DEF(
150 NULL
, "pass", 1, "Pass to execute (1/2)");
151 static const arg_def_t fpf_name
= ARG_DEF(
152 NULL
, "fpf", 1, "First pass statistics file name");
153 #if CONFIG_FP_MB_STATS
154 static const arg_def_t fpmbf_name
= ARG_DEF(
155 NULL
, "fpmbf", 1, "First pass block statistics file name");
157 static const arg_def_t limit
= ARG_DEF(
158 NULL
, "limit", 1, "Stop encoding after n input frames");
159 static const arg_def_t skip
= ARG_DEF(
160 NULL
, "skip", 1, "Skip the first n input frames");
161 static const arg_def_t deadline
= ARG_DEF(
162 "d", "deadline", 1, "Deadline per frame (usec)");
163 static const arg_def_t best_dl
= ARG_DEF(
164 NULL
, "best", 0, "Use Best Quality Deadline");
165 static const arg_def_t good_dl
= ARG_DEF(
166 NULL
, "good", 0, "Use Good Quality Deadline");
167 static const arg_def_t rt_dl
= ARG_DEF(
168 NULL
, "rt", 0, "Use Realtime Quality Deadline");
169 static const arg_def_t quietarg
= ARG_DEF(
170 "q", "quiet", 0, "Do not print encode progress");
171 static const arg_def_t verbosearg
= ARG_DEF(
172 "v", "verbose", 0, "Show encoder parameters");
173 static const arg_def_t psnrarg
= ARG_DEF(
174 NULL
, "psnr", 0, "Show PSNR in status line");
176 static const struct arg_enum_list test_decode_enum
[] = {
177 {"off", TEST_DECODE_OFF
},
178 {"fatal", TEST_DECODE_FATAL
},
179 {"warn", TEST_DECODE_WARN
},
182 static const arg_def_t recontest
= ARG_DEF_ENUM(
183 NULL
, "test-decode", 1, "Test encode/decode mismatch", test_decode_enum
);
184 static const arg_def_t framerate
= ARG_DEF(
185 NULL
, "fps", 1, "Stream frame rate (rate/scale)");
186 static const arg_def_t use_webm
= ARG_DEF(
187 NULL
, "webm", 0, "Output WebM (default when WebM IO is enabled)");
188 static const arg_def_t use_ivf
= ARG_DEF(
189 NULL
, "ivf", 0, "Output IVF");
190 static const arg_def_t out_part
= ARG_DEF(
191 "P", "output-partitions", 0,
192 "Makes encoder output partitions. Requires IVF output!");
193 static const arg_def_t q_hist_n
= ARG_DEF(
194 NULL
, "q-hist", 1, "Show quantizer histogram (n-buckets)");
195 static const arg_def_t rate_hist_n
= ARG_DEF(
196 NULL
, "rate-hist", 1, "Show rate histogram (n-buckets)");
197 static const arg_def_t disable_warnings
= ARG_DEF(
198 NULL
, "disable-warnings", 0,
199 "Disable warnings about potentially incorrect encode settings.");
200 static const arg_def_t disable_warning_prompt
= ARG_DEF(
201 "y", "disable-warning-prompt", 0,
202 "Display warnings, but do not prompt user to continue.");
204 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
205 static const arg_def_t test16bitinternalarg
= ARG_DEF(
206 NULL
, "test-16bit-internal", 0, "Force use of 16 bit internal buffer");
209 static const arg_def_t
*main_args
[] = {
211 &outputfile
, &codecarg
, &passes
, &pass_arg
, &fpf_name
, &limit
, &skip
,
212 &deadline
, &best_dl
, &good_dl
, &rt_dl
,
213 &quietarg
, &verbosearg
, &psnrarg
, &use_webm
, &use_ivf
, &out_part
, &q_hist_n
,
214 &rate_hist_n
, &disable_warnings
, &disable_warning_prompt
,
218 static const arg_def_t usage
= ARG_DEF(
219 "u", "usage", 1, "Usage profile number to use");
220 static const arg_def_t threads
= ARG_DEF(
221 "t", "threads", 1, "Max number of threads to use");
222 static const arg_def_t profile
= ARG_DEF(
223 NULL
, "profile", 1, "Bitstream profile number to use");
224 static const arg_def_t width
= ARG_DEF("w", "width", 1, "Frame width");
225 static const arg_def_t height
= ARG_DEF("h", "height", 1, "Frame height");
227 static const struct arg_enum_list stereo_mode_enum
[] = {
228 {"mono", STEREO_FORMAT_MONO
},
229 {"left-right", STEREO_FORMAT_LEFT_RIGHT
},
230 {"bottom-top", STEREO_FORMAT_BOTTOM_TOP
},
231 {"top-bottom", STEREO_FORMAT_TOP_BOTTOM
},
232 {"right-left", STEREO_FORMAT_RIGHT_LEFT
},
235 static const arg_def_t stereo_mode
= ARG_DEF_ENUM(
236 NULL
, "stereo-mode", 1, "Stereo 3D video format", stereo_mode_enum
);
238 static const arg_def_t timebase
= ARG_DEF(
239 NULL
, "timebase", 1, "Output timestamp precision (fractional seconds)");
240 static const arg_def_t error_resilient
= ARG_DEF(
241 NULL
, "error-resilient", 1, "Enable error resiliency features");
242 static const arg_def_t lag_in_frames
= ARG_DEF(
243 NULL
, "lag-in-frames", 1, "Max number of frames to lag");
245 static const arg_def_t
*global_args
[] = {
246 &use_yv12
, &use_i420
, &use_i422
, &use_i444
, &use_i440
,
247 &usage
, &threads
, &profile
,
252 &timebase
, &framerate
,
254 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
255 &test16bitinternalarg
,
260 static const arg_def_t dropframe_thresh
= ARG_DEF(
261 NULL
, "drop-frame", 1, "Temporal resampling threshold (buf %)");
262 static const arg_def_t resize_allowed
= ARG_DEF(
263 NULL
, "resize-allowed", 1, "Spatial resampling enabled (bool)");
264 static const arg_def_t resize_width
= ARG_DEF(
265 NULL
, "resize-width", 1, "Width of encoded frame");
266 static const arg_def_t resize_height
= ARG_DEF(
267 NULL
, "resize-height", 1, "Height of encoded frame");
268 static const arg_def_t resize_up_thresh
= ARG_DEF(
269 NULL
, "resize-up", 1, "Upscale threshold (buf %)");
270 static const arg_def_t resize_down_thresh
= ARG_DEF(
271 NULL
, "resize-down", 1, "Downscale threshold (buf %)");
272 static const struct arg_enum_list end_usage_enum
[] = {
279 static const arg_def_t end_usage
= ARG_DEF_ENUM(
280 NULL
, "end-usage", 1, "Rate control mode", end_usage_enum
);
281 static const arg_def_t target_bitrate
= ARG_DEF(
282 NULL
, "target-bitrate", 1, "Bitrate (kbps)");
283 static const arg_def_t min_quantizer
= ARG_DEF(
284 NULL
, "min-q", 1, "Minimum (best) quantizer");
285 static const arg_def_t max_quantizer
= ARG_DEF(
286 NULL
, "max-q", 1, "Maximum (worst) quantizer");
287 static const arg_def_t undershoot_pct
= ARG_DEF(
288 NULL
, "undershoot-pct", 1, "Datarate undershoot (min) target (%)");
289 static const arg_def_t overshoot_pct
= ARG_DEF(
290 NULL
, "overshoot-pct", 1, "Datarate overshoot (max) target (%)");
291 static const arg_def_t buf_sz
= ARG_DEF(
292 NULL
, "buf-sz", 1, "Client buffer size (ms)");
293 static const arg_def_t buf_initial_sz
= ARG_DEF(
294 NULL
, "buf-initial-sz", 1, "Client initial buffer size (ms)");
295 static const arg_def_t buf_optimal_sz
= ARG_DEF(
296 NULL
, "buf-optimal-sz", 1, "Client optimal buffer size (ms)");
297 static const arg_def_t
*rc_args
[] = {
298 &dropframe_thresh
, &resize_allowed
, &resize_width
, &resize_height
,
299 &resize_up_thresh
, &resize_down_thresh
, &end_usage
, &target_bitrate
,
300 &min_quantizer
, &max_quantizer
, &undershoot_pct
, &overshoot_pct
, &buf_sz
,
301 &buf_initial_sz
, &buf_optimal_sz
, NULL
305 static const arg_def_t bias_pct
= ARG_DEF(
306 NULL
, "bias-pct", 1, "CBR/VBR bias (0=CBR, 100=VBR)");
307 static const arg_def_t minsection_pct
= ARG_DEF(
308 NULL
, "minsection-pct", 1, "GOP min bitrate (% of target)");
309 static const arg_def_t maxsection_pct
= ARG_DEF(
310 NULL
, "maxsection-pct", 1, "GOP max bitrate (% of target)");
311 static const arg_def_t
*rc_twopass_args
[] = {
312 &bias_pct
, &minsection_pct
, &maxsection_pct
, NULL
316 static const arg_def_t kf_min_dist
= ARG_DEF(
317 NULL
, "kf-min-dist", 1, "Minimum keyframe interval (frames)");
318 static const arg_def_t kf_max_dist
= ARG_DEF(
319 NULL
, "kf-max-dist", 1, "Maximum keyframe interval (frames)");
320 static const arg_def_t kf_disabled
= ARG_DEF(
321 NULL
, "disable-kf", 0, "Disable keyframe placement");
322 static const arg_def_t
*kf_args
[] = {
323 &kf_min_dist
, &kf_max_dist
, &kf_disabled
, NULL
327 static const arg_def_t noise_sens
= ARG_DEF(
328 NULL
, "noise-sensitivity", 1, "Noise sensitivity (frames to blur)");
329 static const arg_def_t sharpness
= ARG_DEF(
330 NULL
, "sharpness", 1, "Loop filter sharpness (0..7)");
331 static const arg_def_t static_thresh
= ARG_DEF(
332 NULL
, "static-thresh", 1, "Motion detection threshold");
333 static const arg_def_t cpu_used_vp8
= ARG_DEF(
334 NULL
, "cpu-used", 1, "CPU Used (-16..16)");
335 static const arg_def_t cpu_used_vp9
= ARG_DEF(
336 NULL
, "cpu-used", 1, "CPU Used (-8..8)");
337 static const arg_def_t auto_altref
= ARG_DEF(
338 NULL
, "auto-alt-ref", 1, "Enable automatic alt reference frames");
339 static const arg_def_t arnr_maxframes
= ARG_DEF(
340 NULL
, "arnr-maxframes", 1, "AltRef max frames (0..15)");
341 static const arg_def_t arnr_strength
= ARG_DEF(
342 NULL
, "arnr-strength", 1, "AltRef filter strength (0..6)");
343 static const arg_def_t arnr_type
= ARG_DEF(
344 NULL
, "arnr-type", 1, "AltRef type");
345 static const struct arg_enum_list tuning_enum
[] = {
346 {"psnr", VP8_TUNE_PSNR
},
347 {"ssim", VP8_TUNE_SSIM
},
350 static const arg_def_t tune_ssim
= ARG_DEF_ENUM(
351 NULL
, "tune", 1, "Material to favor", tuning_enum
);
352 static const arg_def_t cq_level
= ARG_DEF(
353 NULL
, "cq-level", 1, "Constant/Constrained Quality level");
354 static const arg_def_t max_intra_rate_pct
= ARG_DEF(
355 NULL
, "max-intra-rate", 1, "Max I-frame bitrate (pct)");
356 static const arg_def_t max_inter_rate_pct
= ARG_DEF(
357 NULL
, "max-inter-rate", 1, "Max P-frame bitrate (pct)");
358 static const arg_def_t gf_cbr_boost_pct
= ARG_DEF(
359 NULL
, "gf-cbr-boost", 1, "Boost for Golden Frame in CBR mode (pct)");
361 static const arg_def_t screen_content_mode
= ARG_DEF(NULL
, "screen-content-mode", 1,
362 "Screen content mode");
364 #if CONFIG_VP8_ENCODER
365 static const arg_def_t token_parts
= ARG_DEF(
366 NULL
, "token-parts", 1, "Number of token partitions to use, log2");
367 static const arg_def_t
*vp8_args
[] = {
368 &cpu_used_vp8
, &auto_altref
, &noise_sens
, &sharpness
, &static_thresh
,
369 &token_parts
, &arnr_maxframes
, &arnr_strength
, &arnr_type
,
370 &tune_ssim
, &cq_level
, &max_intra_rate_pct
, &screen_content_mode
,
373 static const int vp8_arg_ctrl_map
[] = {
374 VP8E_SET_CPUUSED
, VP8E_SET_ENABLEAUTOALTREF
,
375 VP8E_SET_NOISE_SENSITIVITY
, VP8E_SET_SHARPNESS
, VP8E_SET_STATIC_THRESHOLD
,
376 VP8E_SET_TOKEN_PARTITIONS
,
377 VP8E_SET_ARNR_MAXFRAMES
, VP8E_SET_ARNR_STRENGTH
, VP8E_SET_ARNR_TYPE
,
378 VP8E_SET_TUNING
, VP8E_SET_CQ_LEVEL
, VP8E_SET_MAX_INTRA_BITRATE_PCT
,
379 VP8E_SET_SCREEN_CONTENT_MODE
,
384 #if CONFIG_VP9_ENCODER
385 static const arg_def_t tile_cols
= ARG_DEF(
386 NULL
, "tile-columns", 1, "Number of tile columns to use, log2");
387 static const arg_def_t tile_rows
= ARG_DEF(
388 NULL
, "tile-rows", 1, "Number of tile rows to use, log2");
389 static const arg_def_t lossless
= ARG_DEF(
390 NULL
, "lossless", 1, "Lossless mode");
391 static const arg_def_t frame_parallel_decoding
= ARG_DEF(
392 NULL
, "frame-parallel", 1, "Enable frame parallel decodability features");
393 static const arg_def_t aq_mode
= ARG_DEF(
395 "Adaptive quantization mode (0: off (default), 1: variance 2: complexity, "
396 "3: cyclic refresh)");
397 static const arg_def_t frame_periodic_boost
= ARG_DEF(
398 NULL
, "frame-boost", 1,
399 "Enable frame periodic boost (0: off (default), 1: on)");
401 static const struct arg_enum_list color_space_enum
[] = {
402 { "unknown", VPX_CS_UNKNOWN
},
403 { "bt601", VPX_CS_BT_601
},
404 { "bt709", VPX_CS_BT_709
},
405 { "smpte170", VPX_CS_SMPTE_170
},
406 { "smpte240", VPX_CS_SMPTE_240
},
407 { "bt2020", VPX_CS_BT_2020
},
408 { "reserved", VPX_CS_RESERVED
},
409 { "sRGB", VPX_CS_SRGB
},
413 static const arg_def_t input_color_space
= ARG_DEF_ENUM(
414 NULL
, "color-space", 1,
415 "The color space of input content:", color_space_enum
);
417 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
418 static const struct arg_enum_list bitdepth_enum
[] = {
425 static const arg_def_t bitdeptharg
= ARG_DEF_ENUM(
427 "Bit depth for codec (8 for version <=1, 10 or 12 for version 2)",
429 static const arg_def_t inbitdeptharg
= ARG_DEF(
430 NULL
, "input-bit-depth", 1, "Bit depth of input");
433 static const struct arg_enum_list tune_content_enum
[] = {
434 {"default", VP9E_CONTENT_DEFAULT
},
435 {"screen", VP9E_CONTENT_SCREEN
},
439 static const arg_def_t tune_content
= ARG_DEF_ENUM(
440 NULL
, "tune-content", 1, "Tune content type", tune_content_enum
);
442 static const arg_def_t
*vp9_args
[] = {
443 &cpu_used_vp9
, &auto_altref
, &sharpness
, &static_thresh
,
444 &tile_cols
, &tile_rows
, &arnr_maxframes
, &arnr_strength
, &arnr_type
,
445 &tune_ssim
, &cq_level
, &max_intra_rate_pct
, &max_inter_rate_pct
,
446 &gf_cbr_boost_pct
, &lossless
,
447 &frame_parallel_decoding
, &aq_mode
, &frame_periodic_boost
,
448 &noise_sens
, &tune_content
, &input_color_space
,
449 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
450 &bitdeptharg
, &inbitdeptharg
,
454 static const int vp9_arg_ctrl_map
[] = {
455 VP8E_SET_CPUUSED
, VP8E_SET_ENABLEAUTOALTREF
,
456 VP8E_SET_SHARPNESS
, VP8E_SET_STATIC_THRESHOLD
,
457 VP9E_SET_TILE_COLUMNS
, VP9E_SET_TILE_ROWS
,
458 VP8E_SET_ARNR_MAXFRAMES
, VP8E_SET_ARNR_STRENGTH
, VP8E_SET_ARNR_TYPE
,
459 VP8E_SET_TUNING
, VP8E_SET_CQ_LEVEL
, VP8E_SET_MAX_INTRA_BITRATE_PCT
,
460 VP9E_SET_MAX_INTER_BITRATE_PCT
, VP9E_SET_GF_CBR_BOOST_PCT
,
461 VP9E_SET_LOSSLESS
, VP9E_SET_FRAME_PARALLEL_DECODING
, VP9E_SET_AQ_MODE
,
462 VP9E_SET_FRAME_PERIODIC_BOOST
, VP9E_SET_NOISE_SENSITIVITY
,
463 VP9E_SET_TUNE_CONTENT
, VP9E_SET_COLOR_SPACE
,
468 static const arg_def_t
*no_args
[] = { NULL
};
473 fprintf(stderr
, "Usage: %s <options> -o dst_filename src_filename \n",
476 fprintf(stderr
, "\nOptions:\n");
477 arg_show_usage(stderr
, main_args
);
478 fprintf(stderr
, "\nEncoder Global Options:\n");
479 arg_show_usage(stderr
, global_args
);
480 fprintf(stderr
, "\nRate Control Options:\n");
481 arg_show_usage(stderr
, rc_args
);
482 fprintf(stderr
, "\nTwopass Rate Control Options:\n");
483 arg_show_usage(stderr
, rc_twopass_args
);
484 fprintf(stderr
, "\nKeyframe Placement Options:\n");
485 arg_show_usage(stderr
, kf_args
);
486 #if CONFIG_VP8_ENCODER
487 fprintf(stderr
, "\nVP8 Specific Options:\n");
488 arg_show_usage(stderr
, vp8_args
);
490 #if CONFIG_VP9_ENCODER
491 fprintf(stderr
, "\nVP9 Specific Options:\n");
492 arg_show_usage(stderr
, vp9_args
);
494 fprintf(stderr
, "\nStream timebase (--timebase):\n"
495 " The desired precision of timestamps in the output, expressed\n"
496 " in fractional seconds. Default is 1/1000.\n");
497 fprintf(stderr
, "\nIncluded encoders:\n\n");
499 for (i
= 0; i
< get_vpx_encoder_count(); ++i
) {
500 const VpxInterface
*const encoder
= get_vpx_encoder_by_index(i
);
501 fprintf(stderr
, " %-6s - %s\n",
502 encoder
->name
, vpx_codec_iface_name(encoder
->codec_interface()));
508 #define mmin(a, b) ((a) < (b) ? (a) : (b))
510 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
511 static void find_mismatch_high(const vpx_image_t
*const img1
,
512 const vpx_image_t
*const img2
,
513 int yloc
[4], int uloc
[4], int vloc
[4]) {
514 uint16_t *plane1
, *plane2
;
515 uint32_t stride1
, stride2
;
516 const uint32_t bsize
= 64;
517 const uint32_t bsizey
= bsize
>> img1
->y_chroma_shift
;
518 const uint32_t bsizex
= bsize
>> img1
->x_chroma_shift
;
520 (img1
->d_w
+ img1
->x_chroma_shift
) >> img1
->x_chroma_shift
;
522 (img1
->d_h
+ img1
->y_chroma_shift
) >> img1
->y_chroma_shift
;
525 yloc
[0] = yloc
[1] = yloc
[2] = yloc
[3] = -1;
526 plane1
= (uint16_t*)img1
->planes
[VPX_PLANE_Y
];
527 plane2
= (uint16_t*)img2
->planes
[VPX_PLANE_Y
];
528 stride1
= img1
->stride
[VPX_PLANE_Y
]/2;
529 stride2
= img2
->stride
[VPX_PLANE_Y
]/2;
530 for (i
= 0, match
= 1; match
&& i
< img1
->d_h
; i
+= bsize
) {
531 for (j
= 0; match
&& j
< img1
->d_w
; j
+= bsize
) {
533 const int si
= mmin(i
+ bsize
, img1
->d_h
) - i
;
534 const int sj
= mmin(j
+ bsize
, img1
->d_w
) - j
;
535 for (k
= 0; match
&& k
< si
; ++k
) {
536 for (l
= 0; match
&& l
< sj
; ++l
) {
537 if (*(plane1
+ (i
+ k
) * stride1
+ j
+ l
) !=
538 *(plane2
+ (i
+ k
) * stride2
+ j
+ l
)) {
541 yloc
[2] = *(plane1
+ (i
+ k
) * stride1
+ j
+ l
);
542 yloc
[3] = *(plane2
+ (i
+ k
) * stride2
+ j
+ l
);
551 uloc
[0] = uloc
[1] = uloc
[2] = uloc
[3] = -1;
552 plane1
= (uint16_t*)img1
->planes
[VPX_PLANE_U
];
553 plane2
= (uint16_t*)img2
->planes
[VPX_PLANE_U
];
554 stride1
= img1
->stride
[VPX_PLANE_U
]/2;
555 stride2
= img2
->stride
[VPX_PLANE_U
]/2;
556 for (i
= 0, match
= 1; match
&& i
< c_h
; i
+= bsizey
) {
557 for (j
= 0; match
&& j
< c_w
; j
+= bsizex
) {
559 const int si
= mmin(i
+ bsizey
, c_h
- i
);
560 const int sj
= mmin(j
+ bsizex
, c_w
- j
);
561 for (k
= 0; match
&& k
< si
; ++k
) {
562 for (l
= 0; match
&& l
< sj
; ++l
) {
563 if (*(plane1
+ (i
+ k
) * stride1
+ j
+ l
) !=
564 *(plane2
+ (i
+ k
) * stride2
+ j
+ l
)) {
567 uloc
[2] = *(plane1
+ (i
+ k
) * stride1
+ j
+ l
);
568 uloc
[3] = *(plane2
+ (i
+ k
) * stride2
+ j
+ l
);
577 vloc
[0] = vloc
[1] = vloc
[2] = vloc
[3] = -1;
578 plane1
= (uint16_t*)img1
->planes
[VPX_PLANE_V
];
579 plane2
= (uint16_t*)img2
->planes
[VPX_PLANE_V
];
580 stride1
= img1
->stride
[VPX_PLANE_V
]/2;
581 stride2
= img2
->stride
[VPX_PLANE_V
]/2;
582 for (i
= 0, match
= 1; match
&& i
< c_h
; i
+= bsizey
) {
583 for (j
= 0; match
&& j
< c_w
; j
+= bsizex
) {
585 const int si
= mmin(i
+ bsizey
, c_h
- i
);
586 const int sj
= mmin(j
+ bsizex
, c_w
- j
);
587 for (k
= 0; match
&& k
< si
; ++k
) {
588 for (l
= 0; match
&& l
< sj
; ++l
) {
589 if (*(plane1
+ (i
+ k
) * stride1
+ j
+ l
) !=
590 *(plane2
+ (i
+ k
) * stride2
+ j
+ l
)) {
593 vloc
[2] = *(plane1
+ (i
+ k
) * stride1
+ j
+ l
);
594 vloc
[3] = *(plane2
+ (i
+ k
) * stride2
+ j
+ l
);
605 static void find_mismatch(const vpx_image_t
*const img1
,
606 const vpx_image_t
*const img2
,
607 int yloc
[4], int uloc
[4], int vloc
[4]) {
608 const uint32_t bsize
= 64;
609 const uint32_t bsizey
= bsize
>> img1
->y_chroma_shift
;
610 const uint32_t bsizex
= bsize
>> img1
->x_chroma_shift
;
612 (img1
->d_w
+ img1
->x_chroma_shift
) >> img1
->x_chroma_shift
;
614 (img1
->d_h
+ img1
->y_chroma_shift
) >> img1
->y_chroma_shift
;
617 yloc
[0] = yloc
[1] = yloc
[2] = yloc
[3] = -1;
618 for (i
= 0, match
= 1; match
&& i
< img1
->d_h
; i
+= bsize
) {
619 for (j
= 0; match
&& j
< img1
->d_w
; j
+= bsize
) {
621 const int si
= mmin(i
+ bsize
, img1
->d_h
) - i
;
622 const int sj
= mmin(j
+ bsize
, img1
->d_w
) - j
;
623 for (k
= 0; match
&& k
< si
; ++k
) {
624 for (l
= 0; match
&& l
< sj
; ++l
) {
625 if (*(img1
->planes
[VPX_PLANE_Y
] +
626 (i
+ k
) * img1
->stride
[VPX_PLANE_Y
] + j
+ l
) !=
627 *(img2
->planes
[VPX_PLANE_Y
] +
628 (i
+ k
) * img2
->stride
[VPX_PLANE_Y
] + j
+ l
)) {
631 yloc
[2] = *(img1
->planes
[VPX_PLANE_Y
] +
632 (i
+ k
) * img1
->stride
[VPX_PLANE_Y
] + j
+ l
);
633 yloc
[3] = *(img2
->planes
[VPX_PLANE_Y
] +
634 (i
+ k
) * img2
->stride
[VPX_PLANE_Y
] + j
+ l
);
643 uloc
[0] = uloc
[1] = uloc
[2] = uloc
[3] = -1;
644 for (i
= 0, match
= 1; match
&& i
< c_h
; i
+= bsizey
) {
645 for (j
= 0; match
&& j
< c_w
; j
+= bsizex
) {
647 const int si
= mmin(i
+ bsizey
, c_h
- i
);
648 const int sj
= mmin(j
+ bsizex
, c_w
- j
);
649 for (k
= 0; match
&& k
< si
; ++k
) {
650 for (l
= 0; match
&& l
< sj
; ++l
) {
651 if (*(img1
->planes
[VPX_PLANE_U
] +
652 (i
+ k
) * img1
->stride
[VPX_PLANE_U
] + j
+ l
) !=
653 *(img2
->planes
[VPX_PLANE_U
] +
654 (i
+ k
) * img2
->stride
[VPX_PLANE_U
] + j
+ l
)) {
657 uloc
[2] = *(img1
->planes
[VPX_PLANE_U
] +
658 (i
+ k
) * img1
->stride
[VPX_PLANE_U
] + j
+ l
);
659 uloc
[3] = *(img2
->planes
[VPX_PLANE_U
] +
660 (i
+ k
) * img2
->stride
[VPX_PLANE_U
] + j
+ l
);
668 vloc
[0] = vloc
[1] = vloc
[2] = vloc
[3] = -1;
669 for (i
= 0, match
= 1; match
&& i
< c_h
; i
+= bsizey
) {
670 for (j
= 0; match
&& j
< c_w
; j
+= bsizex
) {
672 const int si
= mmin(i
+ bsizey
, c_h
- i
);
673 const int sj
= mmin(j
+ bsizex
, c_w
- j
);
674 for (k
= 0; match
&& k
< si
; ++k
) {
675 for (l
= 0; match
&& l
< sj
; ++l
) {
676 if (*(img1
->planes
[VPX_PLANE_V
] +
677 (i
+ k
) * img1
->stride
[VPX_PLANE_V
] + j
+ l
) !=
678 *(img2
->planes
[VPX_PLANE_V
] +
679 (i
+ k
) * img2
->stride
[VPX_PLANE_V
] + j
+ l
)) {
682 vloc
[2] = *(img1
->planes
[VPX_PLANE_V
] +
683 (i
+ k
) * img1
->stride
[VPX_PLANE_V
] + j
+ l
);
684 vloc
[3] = *(img2
->planes
[VPX_PLANE_V
] +
685 (i
+ k
) * img2
->stride
[VPX_PLANE_V
] + j
+ l
);
695 static int compare_img(const vpx_image_t
*const img1
,
696 const vpx_image_t
*const img2
) {
697 uint32_t l_w
= img1
->d_w
;
699 (img1
->d_w
+ img1
->x_chroma_shift
) >> img1
->x_chroma_shift
;
701 (img1
->d_h
+ img1
->y_chroma_shift
) >> img1
->y_chroma_shift
;
705 match
&= (img1
->fmt
== img2
->fmt
);
706 match
&= (img1
->d_w
== img2
->d_w
);
707 match
&= (img1
->d_h
== img2
->d_h
);
708 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
709 if (img1
->fmt
& VPX_IMG_FMT_HIGHBITDEPTH
) {
715 for (i
= 0; i
< img1
->d_h
; ++i
)
716 match
&= (memcmp(img1
->planes
[VPX_PLANE_Y
] + i
* img1
->stride
[VPX_PLANE_Y
],
717 img2
->planes
[VPX_PLANE_Y
] + i
* img2
->stride
[VPX_PLANE_Y
],
720 for (i
= 0; i
< c_h
; ++i
)
721 match
&= (memcmp(img1
->planes
[VPX_PLANE_U
] + i
* img1
->stride
[VPX_PLANE_U
],
722 img2
->planes
[VPX_PLANE_U
] + i
* img2
->stride
[VPX_PLANE_U
],
725 for (i
= 0; i
< c_h
; ++i
)
726 match
&= (memcmp(img1
->planes
[VPX_PLANE_V
] + i
* img1
->stride
[VPX_PLANE_V
],
727 img2
->planes
[VPX_PLANE_V
] + i
* img2
->stride
[VPX_PLANE_V
],
734 #define NELEMENTS(x) (sizeof(x)/sizeof(x[0]))
735 #define MAX(x,y) ((x)>(y)?(x):(y))
736 #if CONFIG_VP8_ENCODER && !CONFIG_VP9_ENCODER
737 #define ARG_CTRL_CNT_MAX NELEMENTS(vp8_arg_ctrl_map)
738 #elif !CONFIG_VP8_ENCODER && CONFIG_VP9_ENCODER
739 #define ARG_CTRL_CNT_MAX NELEMENTS(vp9_arg_ctrl_map)
741 #define ARG_CTRL_CNT_MAX MAX(NELEMENTS(vp8_arg_ctrl_map), \
742 NELEMENTS(vp9_arg_ctrl_map))
746 typedef int stereo_format_t
;
747 struct EbmlGlobal
{ int debug
; };
750 /* Per-stream configuration */
751 struct stream_config
{
752 struct vpx_codec_enc_cfg cfg
;
754 const char *stats_fn
;
755 #if CONFIG_FP_MB_STATS
756 const char *fpmb_stats_fn
;
758 stereo_format_t stereo_fmt
;
759 int arg_ctrls
[ARG_CTRL_CNT_MAX
][2];
762 int have_kf_max_dist
;
763 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
764 // whether to use 16bit internal buffers
765 int use_16bit_internal
;
770 struct stream_state
{
772 struct stream_state
*next
;
773 struct stream_config config
;
775 struct rate_hist
*rate_hist
;
776 struct EbmlGlobal ebml
;
777 uint64_t psnr_sse_total
;
778 uint64_t psnr_samples_total
;
779 double psnr_totals
[4];
782 vpx_codec_ctx_t encoder
;
783 unsigned int frames_out
;
787 #if CONFIG_FP_MB_STATS
788 stats_io_t fpmb_stats
;
790 struct vpx_image
*img
;
791 vpx_codec_ctx_t decoder
;
796 void validate_positive_rational(const char *msg
,
797 struct vpx_rational
*rat
) {
804 die("Error: %s must be positive\n", msg
);
807 die("Error: %s has zero denominator\n", msg
);
811 static void parse_global_config(struct VpxEncoderConfig
*global
, char **argv
) {
815 /* Initialize default parameters */
816 memset(global
, 0, sizeof(*global
));
817 global
->codec
= get_vpx_encoder_by_index(0);
819 global
->color_type
= I420
;
820 /* Assign default deadline to good quality */
821 global
->deadline
= VPX_DL_GOOD_QUALITY
;
823 for (argi
= argj
= argv
; (*argj
= *argi
); argi
+= arg
.argv_step
) {
826 if (arg_match(&arg
, &codecarg
, argi
)) {
827 global
->codec
= get_vpx_encoder_by_name(arg
.val
);
829 die("Error: Unrecognized argument (%s) to --codec\n", arg
.val
);
830 } else if (arg_match(&arg
, &passes
, argi
)) {
831 global
->passes
= arg_parse_uint(&arg
);
833 if (global
->passes
< 1 || global
->passes
> 2)
834 die("Error: Invalid number of passes (%d)\n", global
->passes
);
835 } else if (arg_match(&arg
, &pass_arg
, argi
)) {
836 global
->pass
= arg_parse_uint(&arg
);
838 if (global
->pass
< 1 || global
->pass
> 2)
839 die("Error: Invalid pass selected (%d)\n",
841 } else if (arg_match(&arg
, &usage
, argi
))
842 global
->usage
= arg_parse_uint(&arg
);
843 else if (arg_match(&arg
, &deadline
, argi
))
844 global
->deadline
= arg_parse_uint(&arg
);
845 else if (arg_match(&arg
, &best_dl
, argi
))
846 global
->deadline
= VPX_DL_BEST_QUALITY
;
847 else if (arg_match(&arg
, &good_dl
, argi
))
848 global
->deadline
= VPX_DL_GOOD_QUALITY
;
849 else if (arg_match(&arg
, &rt_dl
, argi
))
850 global
->deadline
= VPX_DL_REALTIME
;
851 else if (arg_match(&arg
, &use_yv12
, argi
))
852 global
->color_type
= YV12
;
853 else if (arg_match(&arg
, &use_i420
, argi
))
854 global
->color_type
= I420
;
855 else if (arg_match(&arg
, &use_i422
, argi
))
856 global
->color_type
= I422
;
857 else if (arg_match(&arg
, &use_i444
, argi
))
858 global
->color_type
= I444
;
859 else if (arg_match(&arg
, &use_i440
, argi
))
860 global
->color_type
= I440
;
861 else if (arg_match(&arg
, &quietarg
, argi
))
863 else if (arg_match(&arg
, &verbosearg
, argi
))
865 else if (arg_match(&arg
, &limit
, argi
))
866 global
->limit
= arg_parse_uint(&arg
);
867 else if (arg_match(&arg
, &skip
, argi
))
868 global
->skip_frames
= arg_parse_uint(&arg
);
869 else if (arg_match(&arg
, &psnrarg
, argi
))
870 global
->show_psnr
= 1;
871 else if (arg_match(&arg
, &recontest
, argi
))
872 global
->test_decode
= arg_parse_enum_or_int(&arg
);
873 else if (arg_match(&arg
, &framerate
, argi
)) {
874 global
->framerate
= arg_parse_rational(&arg
);
875 validate_positive_rational(arg
.name
, &global
->framerate
);
876 global
->have_framerate
= 1;
877 } else if (arg_match(&arg
, &out_part
, argi
))
878 global
->out_part
= 1;
879 else if (arg_match(&arg
, &debugmode
, argi
))
881 else if (arg_match(&arg
, &q_hist_n
, argi
))
882 global
->show_q_hist_buckets
= arg_parse_uint(&arg
);
883 else if (arg_match(&arg
, &rate_hist_n
, argi
))
884 global
->show_rate_hist_buckets
= arg_parse_uint(&arg
);
885 else if (arg_match(&arg
, &disable_warnings
, argi
))
886 global
->disable_warnings
= 1;
887 else if (arg_match(&arg
, &disable_warning_prompt
, argi
))
888 global
->disable_warning_prompt
= 1;
894 /* DWIM: Assume the user meant passes=2 if pass=2 is specified */
895 if (global
->pass
> global
->passes
) {
896 warn("Assuming --pass=%d implies --passes=%d\n",
897 global
->pass
, global
->pass
);
898 global
->passes
= global
->pass
;
901 /* Validate global config */
902 if (global
->passes
== 0) {
903 #if CONFIG_VP9_ENCODER
904 // Make default VP9 passes = 2 until there is a better quality 1-pass
906 if (global
->codec
!= NULL
&& global
->codec
->name
!= NULL
)
907 global
->passes
= (strcmp(global
->codec
->name
, "vp9") == 0 &&
908 global
->deadline
!= VPX_DL_REALTIME
) ? 2 : 1;
914 if (global
->deadline
== VPX_DL_REALTIME
&&
915 global
->passes
> 1) {
916 warn("Enforcing one-pass encoding in realtime mode\n");
922 void open_input_file(struct VpxInputContext
*input
) {
923 /* Parse certain options from the input file, if possible */
924 input
->file
= strcmp(input
->filename
, "-")
925 ? fopen(input
->filename
, "rb") : set_binary_mode(stdin
);
928 fatal("Failed to open input file");
930 if (!fseeko(input
->file
, 0, SEEK_END
)) {
931 /* Input file is seekable. Figure out how long it is, so we can get
934 input
->length
= ftello(input
->file
);
938 /* For RAW input sources, these bytes will applied on the first frame
941 input
->detect
.buf_read
= fread(input
->detect
.buf
, 1, 4, input
->file
);
942 input
->detect
.position
= 0;
944 if (input
->detect
.buf_read
== 4
945 && file_is_y4m(input
->detect
.buf
)) {
946 if (y4m_input_open(&input
->y4m
, input
->file
, input
->detect
.buf
, 4,
947 input
->only_i420
) >= 0) {
948 input
->file_type
= FILE_TYPE_Y4M
;
949 input
->width
= input
->y4m
.pic_w
;
950 input
->height
= input
->y4m
.pic_h
;
951 input
->framerate
.numerator
= input
->y4m
.fps_n
;
952 input
->framerate
.denominator
= input
->y4m
.fps_d
;
953 input
->fmt
= input
->y4m
.vpx_fmt
;
954 input
->bit_depth
= input
->y4m
.bit_depth
;
956 fatal("Unsupported Y4M stream.");
957 } else if (input
->detect
.buf_read
== 4 && fourcc_is_ivf(input
->detect
.buf
)) {
958 fatal("IVF is not supported as input.");
960 input
->file_type
= FILE_TYPE_RAW
;
965 static void close_input_file(struct VpxInputContext
*input
) {
967 if (input
->file_type
== FILE_TYPE_Y4M
)
968 y4m_input_close(&input
->y4m
);
971 static struct stream_state
*new_stream(struct VpxEncoderConfig
*global
,
972 struct stream_state
*prev
) {
973 struct stream_state
*stream
;
975 stream
= calloc(1, sizeof(*stream
));
976 if (stream
== NULL
) {
977 fatal("Failed to allocate new stream.");
981 memcpy(stream
, prev
, sizeof(*stream
));
987 /* Populate encoder configuration */
988 res
= vpx_codec_enc_config_default(global
->codec
->codec_interface(),
992 fatal("Failed to get config: %s\n", vpx_codec_err_to_string(res
));
994 /* Change the default timebase to a high enough value so that the
995 * encoder will always create strictly increasing timestamps.
997 stream
->config
.cfg
.g_timebase
.den
= 1000;
999 /* Never use the library's default resolution, require it be parsed
1000 * from the file or set on the command line.
1002 stream
->config
.cfg
.g_w
= 0;
1003 stream
->config
.cfg
.g_h
= 0;
1005 /* Initialize remaining stream parameters */
1006 stream
->config
.write_webm
= 1;
1008 stream
->config
.stereo_fmt
= STEREO_FORMAT_MONO
;
1009 stream
->ebml
.last_pts_ns
= -1;
1010 stream
->ebml
.writer
= NULL
;
1011 stream
->ebml
.segment
= NULL
;
1014 /* Allows removal of the application version from the EBML tags */
1015 stream
->ebml
.debug
= global
->debug
;
1017 /* Default lag_in_frames is 0 in realtime mode */
1018 if (global
->deadline
== VPX_DL_REALTIME
)
1019 stream
->config
.cfg
.g_lag_in_frames
= 0;
1022 /* Output files must be specified for each stream */
1023 stream
->config
.out_fn
= NULL
;
1025 stream
->next
= NULL
;
1030 static int parse_stream_params(struct VpxEncoderConfig
*global
,
1031 struct stream_state
*stream
,
1033 char **argi
, **argj
;
1035 static const arg_def_t
**ctrl_args
= no_args
;
1036 static const int *ctrl_args_map
= NULL
;
1037 struct stream_config
*config
= &stream
->config
;
1038 int eos_mark_found
= 0;
1039 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1040 int test_16bit_internal
= 0;
1043 // Handle codec specific options
1045 #if CONFIG_VP8_ENCODER
1046 } else if (strcmp(global
->codec
->name
, "vp8") == 0) {
1047 ctrl_args
= vp8_args
;
1048 ctrl_args_map
= vp8_arg_ctrl_map
;
1050 #if CONFIG_VP9_ENCODER
1051 } else if (strcmp(global
->codec
->name
, "vp9") == 0) {
1052 ctrl_args
= vp9_args
;
1053 ctrl_args_map
= vp9_arg_ctrl_map
;
1057 for (argi
= argj
= argv
; (*argj
= *argi
); argi
+= arg
.argv_step
) {
1060 /* Once we've found an end-of-stream marker (--) we want to continue
1061 * shifting arguments but not consuming them.
1063 if (eos_mark_found
) {
1066 } else if (!strcmp(*argj
, "--")) {
1071 if (arg_match(&arg
, &outputfile
, argi
)) {
1072 config
->out_fn
= arg
.val
;
1073 } else if (arg_match(&arg
, &fpf_name
, argi
)) {
1074 config
->stats_fn
= arg
.val
;
1075 #if CONFIG_FP_MB_STATS
1076 } else if (arg_match(&arg
, &fpmbf_name
, argi
)) {
1077 config
->fpmb_stats_fn
= arg
.val
;
1079 } else if (arg_match(&arg
, &use_webm
, argi
)) {
1081 config
->write_webm
= 1;
1083 die("Error: --webm specified but webm is disabled.");
1085 } else if (arg_match(&arg
, &use_ivf
, argi
)) {
1086 config
->write_webm
= 0;
1087 } else if (arg_match(&arg
, &threads
, argi
)) {
1088 config
->cfg
.g_threads
= arg_parse_uint(&arg
);
1089 } else if (arg_match(&arg
, &profile
, argi
)) {
1090 config
->cfg
.g_profile
= arg_parse_uint(&arg
);
1091 } else if (arg_match(&arg
, &width
, argi
)) {
1092 config
->cfg
.g_w
= arg_parse_uint(&arg
);
1093 } else if (arg_match(&arg
, &height
, argi
)) {
1094 config
->cfg
.g_h
= arg_parse_uint(&arg
);
1095 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1096 } else if (arg_match(&arg
, &bitdeptharg
, argi
)) {
1097 config
->cfg
.g_bit_depth
= arg_parse_enum_or_int(&arg
);
1098 } else if (arg_match(&arg
, &inbitdeptharg
, argi
)) {
1099 config
->cfg
.g_input_bit_depth
= arg_parse_uint(&arg
);
1102 } else if (arg_match(&arg
, &stereo_mode
, argi
)) {
1103 config
->stereo_fmt
= arg_parse_enum_or_int(&arg
);
1105 } else if (arg_match(&arg
, &timebase
, argi
)) {
1106 config
->cfg
.g_timebase
= arg_parse_rational(&arg
);
1107 validate_positive_rational(arg
.name
, &config
->cfg
.g_timebase
);
1108 } else if (arg_match(&arg
, &error_resilient
, argi
)) {
1109 config
->cfg
.g_error_resilient
= arg_parse_uint(&arg
);
1110 } else if (arg_match(&arg
, &lag_in_frames
, argi
)) {
1111 config
->cfg
.g_lag_in_frames
= arg_parse_uint(&arg
);
1112 if (global
->deadline
== VPX_DL_REALTIME
&&
1113 config
->cfg
.g_lag_in_frames
!= 0) {
1114 warn("non-zero %s option ignored in realtime mode.\n", arg
.name
);
1115 config
->cfg
.g_lag_in_frames
= 0;
1117 } else if (arg_match(&arg
, &dropframe_thresh
, argi
)) {
1118 config
->cfg
.rc_dropframe_thresh
= arg_parse_uint(&arg
);
1119 } else if (arg_match(&arg
, &resize_allowed
, argi
)) {
1120 config
->cfg
.rc_resize_allowed
= arg_parse_uint(&arg
);
1121 } else if (arg_match(&arg
, &resize_width
, argi
)) {
1122 config
->cfg
.rc_scaled_width
= arg_parse_uint(&arg
);
1123 } else if (arg_match(&arg
, &resize_height
, argi
)) {
1124 config
->cfg
.rc_scaled_height
= arg_parse_uint(&arg
);
1125 } else if (arg_match(&arg
, &resize_up_thresh
, argi
)) {
1126 config
->cfg
.rc_resize_up_thresh
= arg_parse_uint(&arg
);
1127 } else if (arg_match(&arg
, &resize_down_thresh
, argi
)) {
1128 config
->cfg
.rc_resize_down_thresh
= arg_parse_uint(&arg
);
1129 } else if (arg_match(&arg
, &end_usage
, argi
)) {
1130 config
->cfg
.rc_end_usage
= arg_parse_enum_or_int(&arg
);
1131 } else if (arg_match(&arg
, &target_bitrate
, argi
)) {
1132 config
->cfg
.rc_target_bitrate
= arg_parse_uint(&arg
);
1133 } else if (arg_match(&arg
, &min_quantizer
, argi
)) {
1134 config
->cfg
.rc_min_quantizer
= arg_parse_uint(&arg
);
1135 } else if (arg_match(&arg
, &max_quantizer
, argi
)) {
1136 config
->cfg
.rc_max_quantizer
= arg_parse_uint(&arg
);
1137 } else if (arg_match(&arg
, &undershoot_pct
, argi
)) {
1138 config
->cfg
.rc_undershoot_pct
= arg_parse_uint(&arg
);
1139 } else if (arg_match(&arg
, &overshoot_pct
, argi
)) {
1140 config
->cfg
.rc_overshoot_pct
= arg_parse_uint(&arg
);
1141 } else if (arg_match(&arg
, &buf_sz
, argi
)) {
1142 config
->cfg
.rc_buf_sz
= arg_parse_uint(&arg
);
1143 } else if (arg_match(&arg
, &buf_initial_sz
, argi
)) {
1144 config
->cfg
.rc_buf_initial_sz
= arg_parse_uint(&arg
);
1145 } else if (arg_match(&arg
, &buf_optimal_sz
, argi
)) {
1146 config
->cfg
.rc_buf_optimal_sz
= arg_parse_uint(&arg
);
1147 } else if (arg_match(&arg
, &bias_pct
, argi
)) {
1148 config
->cfg
.rc_2pass_vbr_bias_pct
= arg_parse_uint(&arg
);
1149 if (global
->passes
< 2)
1150 warn("option %s ignored in one-pass mode.\n", arg
.name
);
1151 } else if (arg_match(&arg
, &minsection_pct
, argi
)) {
1152 config
->cfg
.rc_2pass_vbr_minsection_pct
= arg_parse_uint(&arg
);
1154 if (global
->passes
< 2)
1155 warn("option %s ignored in one-pass mode.\n", arg
.name
);
1156 } else if (arg_match(&arg
, &maxsection_pct
, argi
)) {
1157 config
->cfg
.rc_2pass_vbr_maxsection_pct
= arg_parse_uint(&arg
);
1159 if (global
->passes
< 2)
1160 warn("option %s ignored in one-pass mode.\n", arg
.name
);
1161 } else if (arg_match(&arg
, &kf_min_dist
, argi
)) {
1162 config
->cfg
.kf_min_dist
= arg_parse_uint(&arg
);
1163 } else if (arg_match(&arg
, &kf_max_dist
, argi
)) {
1164 config
->cfg
.kf_max_dist
= arg_parse_uint(&arg
);
1165 config
->have_kf_max_dist
= 1;
1166 } else if (arg_match(&arg
, &kf_disabled
, argi
)) {
1167 config
->cfg
.kf_mode
= VPX_KF_DISABLED
;
1168 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1169 } else if (arg_match(&arg
, &test16bitinternalarg
, argi
)) {
1170 if (strcmp(global
->codec
->name
, "vp9") == 0) {
1171 test_16bit_internal
= 1;
1176 for (i
= 0; ctrl_args
[i
]; i
++) {
1177 if (arg_match(&arg
, ctrl_args
[i
], argi
)) {
1181 /* Point either to the next free element or the first
1182 * instance of this control.
1184 for (j
= 0; j
< config
->arg_ctrl_cnt
; j
++)
1185 if (ctrl_args_map
!= NULL
&&
1186 config
->arg_ctrls
[j
][0] == ctrl_args_map
[i
])
1190 assert(j
< (int)ARG_CTRL_CNT_MAX
);
1191 if (ctrl_args_map
!= NULL
&& j
< (int)ARG_CTRL_CNT_MAX
) {
1192 config
->arg_ctrls
[j
][0] = ctrl_args_map
[i
];
1193 config
->arg_ctrls
[j
][1] = arg_parse_enum_or_int(&arg
);
1194 if (j
== config
->arg_ctrl_cnt
)
1195 config
->arg_ctrl_cnt
++;
1203 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1204 if (strcmp(global
->codec
->name
, "vp9") == 0) {
1205 config
->use_16bit_internal
= test_16bit_internal
|
1206 (config
->cfg
.g_profile
> 1);
1209 return eos_mark_found
;
1213 #define FOREACH_STREAM(func) \
1215 struct stream_state *stream; \
1216 for (stream = streams; stream; stream = stream->next) { \
1222 static void validate_stream_config(const struct stream_state
*stream
,
1223 const struct VpxEncoderConfig
*global
) {
1224 const struct stream_state
*streami
;
1227 if (!stream
->config
.cfg
.g_w
|| !stream
->config
.cfg
.g_h
)
1228 fatal("Stream %d: Specify stream dimensions with --width (-w) "
1229 " and --height (-h)", stream
->index
);
1231 // Check that the codec bit depth is greater than the input bit depth.
1232 if (stream
->config
.cfg
.g_input_bit_depth
>
1233 (unsigned int)stream
->config
.cfg
.g_bit_depth
) {
1234 fatal("Stream %d: codec bit depth (%d) less than input bit depth (%d)",
1235 stream
->index
, (int)stream
->config
.cfg
.g_bit_depth
,
1236 stream
->config
.cfg
.g_input_bit_depth
);
1239 for (streami
= stream
; streami
; streami
= streami
->next
) {
1240 /* All streams require output files */
1241 if (!streami
->config
.out_fn
)
1242 fatal("Stream %d: Output file is required (specify with -o)",
1245 /* Check for two streams outputting to the same file */
1246 if (streami
!= stream
) {
1247 const char *a
= stream
->config
.out_fn
;
1248 const char *b
= streami
->config
.out_fn
;
1249 if (!strcmp(a
, b
) && strcmp(a
, "/dev/null") && strcmp(a
, ":nul"))
1250 fatal("Stream %d: duplicate output file (from stream %d)",
1251 streami
->index
, stream
->index
);
1254 /* Check for two streams sharing a stats file. */
1255 if (streami
!= stream
) {
1256 const char *a
= stream
->config
.stats_fn
;
1257 const char *b
= streami
->config
.stats_fn
;
1258 if (a
&& b
&& !strcmp(a
, b
))
1259 fatal("Stream %d: duplicate stats file (from stream %d)",
1260 streami
->index
, stream
->index
);
1263 #if CONFIG_FP_MB_STATS
1264 /* Check for two streams sharing a mb stats file. */
1265 if (streami
!= stream
) {
1266 const char *a
= stream
->config
.fpmb_stats_fn
;
1267 const char *b
= streami
->config
.fpmb_stats_fn
;
1268 if (a
&& b
&& !strcmp(a
, b
))
1269 fatal("Stream %d: duplicate mb stats file (from stream %d)",
1270 streami
->index
, stream
->index
);
1277 static void set_stream_dimensions(struct stream_state
*stream
,
1280 if (!stream
->config
.cfg
.g_w
) {
1281 if (!stream
->config
.cfg
.g_h
)
1282 stream
->config
.cfg
.g_w
= w
;
1284 stream
->config
.cfg
.g_w
= w
* stream
->config
.cfg
.g_h
/ h
;
1286 if (!stream
->config
.cfg
.g_h
) {
1287 stream
->config
.cfg
.g_h
= h
* stream
->config
.cfg
.g_w
/ w
;
1292 static void set_default_kf_interval(struct stream_state
*stream
,
1293 struct VpxEncoderConfig
*global
) {
1294 /* Use a max keyframe interval of 5 seconds, if none was
1295 * specified on the command line.
1297 if (!stream
->config
.have_kf_max_dist
) {
1298 double framerate
= (double)global
->framerate
.num
/ global
->framerate
.den
;
1299 if (framerate
> 0.0)
1300 stream
->config
.cfg
.kf_max_dist
= (unsigned int)(5.0 * framerate
);
1304 static const char* file_type_to_string(enum VideoFileType t
) {
1306 case FILE_TYPE_RAW
: return "RAW";
1307 case FILE_TYPE_Y4M
: return "Y4M";
1308 default: return "Other";
1312 static const char* image_format_to_string(vpx_img_fmt_t f
) {
1314 case VPX_IMG_FMT_I420
: return "I420";
1315 case VPX_IMG_FMT_I422
: return "I422";
1316 case VPX_IMG_FMT_I444
: return "I444";
1317 case VPX_IMG_FMT_I440
: return "I440";
1318 case VPX_IMG_FMT_YV12
: return "YV12";
1319 case VPX_IMG_FMT_I42016
: return "I42016";
1320 case VPX_IMG_FMT_I42216
: return "I42216";
1321 case VPX_IMG_FMT_I44416
: return "I44416";
1322 case VPX_IMG_FMT_I44016
: return "I44016";
1323 default: return "Other";
1327 static void show_stream_config(struct stream_state
*stream
,
1328 struct VpxEncoderConfig
*global
,
1329 struct VpxInputContext
*input
) {
1331 #define SHOW(field) \
1332 fprintf(stderr, " %-28s = %d\n", #field, stream->config.cfg.field)
1334 if (stream
->index
== 0) {
1335 fprintf(stderr
, "Codec: %s\n",
1336 vpx_codec_iface_name(global
->codec
->codec_interface()));
1337 fprintf(stderr
, "Source file: %s File Type: %s Format: %s\n",
1339 file_type_to_string(input
->file_type
),
1340 image_format_to_string(input
->fmt
));
1342 if (stream
->next
|| stream
->index
)
1343 fprintf(stderr
, "\nStream Index: %d\n", stream
->index
);
1344 fprintf(stderr
, "Destination file: %s\n", stream
->config
.out_fn
);
1345 fprintf(stderr
, "Encoder parameters:\n");
1353 SHOW(g_input_bit_depth
);
1354 SHOW(g_timebase
.num
);
1355 SHOW(g_timebase
.den
);
1356 SHOW(g_error_resilient
);
1358 SHOW(g_lag_in_frames
);
1359 SHOW(rc_dropframe_thresh
);
1360 SHOW(rc_resize_allowed
);
1361 SHOW(rc_scaled_width
);
1362 SHOW(rc_scaled_height
);
1363 SHOW(rc_resize_up_thresh
);
1364 SHOW(rc_resize_down_thresh
);
1366 SHOW(rc_target_bitrate
);
1367 SHOW(rc_min_quantizer
);
1368 SHOW(rc_max_quantizer
);
1369 SHOW(rc_undershoot_pct
);
1370 SHOW(rc_overshoot_pct
);
1372 SHOW(rc_buf_initial_sz
);
1373 SHOW(rc_buf_optimal_sz
);
1374 SHOW(rc_2pass_vbr_bias_pct
);
1375 SHOW(rc_2pass_vbr_minsection_pct
);
1376 SHOW(rc_2pass_vbr_maxsection_pct
);
1383 static void open_output_file(struct stream_state
*stream
,
1384 struct VpxEncoderConfig
*global
) {
1385 const char *fn
= stream
->config
.out_fn
;
1386 const struct vpx_codec_enc_cfg
*const cfg
= &stream
->config
.cfg
;
1388 if (cfg
->g_pass
== VPX_RC_FIRST_PASS
)
1391 stream
->file
= strcmp(fn
, "-") ? fopen(fn
, "wb") : set_binary_mode(stdout
);
1394 fatal("Failed to open output file");
1396 if (stream
->config
.write_webm
&& fseek(stream
->file
, 0, SEEK_CUR
))
1397 fatal("WebM output to pipes not supported.");
1400 if (stream
->config
.write_webm
) {
1401 stream
->ebml
.stream
= stream
->file
;
1402 write_webm_file_header(&stream
->ebml
, cfg
,
1404 stream
->config
.stereo_fmt
,
1405 global
->codec
->fourcc
);
1409 if (!stream
->config
.write_webm
) {
1410 ivf_write_file_header(stream
->file
, cfg
, global
->codec
->fourcc
, 0);
1415 static void close_output_file(struct stream_state
*stream
,
1416 unsigned int fourcc
) {
1417 const struct vpx_codec_enc_cfg
*const cfg
= &stream
->config
.cfg
;
1419 if (cfg
->g_pass
== VPX_RC_FIRST_PASS
)
1423 if (stream
->config
.write_webm
) {
1424 write_webm_file_footer(&stream
->ebml
);
1428 if (!stream
->config
.write_webm
) {
1429 if (!fseek(stream
->file
, 0, SEEK_SET
))
1430 ivf_write_file_header(stream
->file
, &stream
->config
.cfg
,
1432 stream
->frames_out
);
1435 fclose(stream
->file
);
1439 static void setup_pass(struct stream_state
*stream
,
1440 struct VpxEncoderConfig
*global
,
1442 if (stream
->config
.stats_fn
) {
1443 if (!stats_open_file(&stream
->stats
, stream
->config
.stats_fn
,
1445 fatal("Failed to open statistics store");
1447 if (!stats_open_mem(&stream
->stats
, pass
))
1448 fatal("Failed to open statistics store");
1451 #if CONFIG_FP_MB_STATS
1452 if (stream
->config
.fpmb_stats_fn
) {
1453 if (!stats_open_file(&stream
->fpmb_stats
,
1454 stream
->config
.fpmb_stats_fn
, pass
))
1455 fatal("Failed to open mb statistics store");
1457 if (!stats_open_mem(&stream
->fpmb_stats
, pass
))
1458 fatal("Failed to open mb statistics store");
1462 stream
->config
.cfg
.g_pass
= global
->passes
== 2
1463 ? pass
? VPX_RC_LAST_PASS
: VPX_RC_FIRST_PASS
1466 stream
->config
.cfg
.rc_twopass_stats_in
= stats_get(&stream
->stats
);
1467 #if CONFIG_FP_MB_STATS
1468 stream
->config
.cfg
.rc_firstpass_mb_stats_in
=
1469 stats_get(&stream
->fpmb_stats
);
1473 stream
->cx_time
= 0;
1475 stream
->frames_out
= 0;
1479 static void initialize_encoder(struct stream_state
*stream
,
1480 struct VpxEncoderConfig
*global
) {
1484 flags
|= global
->show_psnr
? VPX_CODEC_USE_PSNR
: 0;
1485 flags
|= global
->out_part
? VPX_CODEC_USE_OUTPUT_PARTITION
: 0;
1486 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1487 flags
|= stream
->config
.use_16bit_internal
? VPX_CODEC_USE_HIGHBITDEPTH
: 0;
1490 /* Construct Encoder Context */
1491 vpx_codec_enc_init(&stream
->encoder
, global
->codec
->codec_interface(),
1492 &stream
->config
.cfg
, flags
);
1493 ctx_exit_on_error(&stream
->encoder
, "Failed to initialize encoder");
1495 /* Note that we bypass the vpx_codec_control wrapper macro because
1496 * we're being clever to store the control IDs in an array. Real
1497 * applications will want to make use of the enumerations directly
1499 for (i
= 0; i
< stream
->config
.arg_ctrl_cnt
; i
++) {
1500 int ctrl
= stream
->config
.arg_ctrls
[i
][0];
1501 int value
= stream
->config
.arg_ctrls
[i
][1];
1502 if (vpx_codec_control_(&stream
->encoder
, ctrl
, value
))
1503 fprintf(stderr
, "Error: Tried to set control %d = %d\n",
1506 ctx_exit_on_error(&stream
->encoder
, "Failed to control codec");
1510 if (global
->test_decode
!= TEST_DECODE_OFF
) {
1511 const VpxInterface
*decoder
= get_vpx_decoder_by_name(global
->codec
->name
);
1512 vpx_codec_dec_init(&stream
->decoder
, decoder
->codec_interface(), NULL
, 0);
1518 static void encode_frame(struct stream_state
*stream
,
1519 struct VpxEncoderConfig
*global
,
1520 struct vpx_image
*img
,
1521 unsigned int frames_in
) {
1522 vpx_codec_pts_t frame_start
, next_frame_start
;
1523 struct vpx_codec_enc_cfg
*cfg
= &stream
->config
.cfg
;
1524 struct vpx_usec_timer timer
;
1526 frame_start
= (cfg
->g_timebase
.den
* (int64_t)(frames_in
- 1)
1527 * global
->framerate
.den
)
1528 / cfg
->g_timebase
.num
/ global
->framerate
.num
;
1529 next_frame_start
= (cfg
->g_timebase
.den
* (int64_t)(frames_in
)
1530 * global
->framerate
.den
)
1531 / cfg
->g_timebase
.num
/ global
->framerate
.num
;
1533 /* Scale if necessary */
1534 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1536 if ((img
->fmt
& VPX_IMG_FMT_HIGHBITDEPTH
) &&
1537 (img
->d_w
!= cfg
->g_w
|| img
->d_h
!= cfg
->g_h
)) {
1538 if (img
->fmt
!= VPX_IMG_FMT_I42016
) {
1539 fprintf(stderr
, "%s can only scale 4:2:0 inputs\n", exec_name
);
1544 stream
->img
= vpx_img_alloc(NULL
, VPX_IMG_FMT_I42016
,
1545 cfg
->g_w
, cfg
->g_h
, 16);
1547 I420Scale_16((uint16
*)img
->planes
[VPX_PLANE_Y
],
1548 img
->stride
[VPX_PLANE_Y
]/2,
1549 (uint16
*)img
->planes
[VPX_PLANE_U
],
1550 img
->stride
[VPX_PLANE_U
]/2,
1551 (uint16
*)img
->planes
[VPX_PLANE_V
],
1552 img
->stride
[VPX_PLANE_V
]/2,
1554 (uint16
*)stream
->img
->planes
[VPX_PLANE_Y
],
1555 stream
->img
->stride
[VPX_PLANE_Y
]/2,
1556 (uint16
*)stream
->img
->planes
[VPX_PLANE_U
],
1557 stream
->img
->stride
[VPX_PLANE_U
]/2,
1558 (uint16
*)stream
->img
->planes
[VPX_PLANE_V
],
1559 stream
->img
->stride
[VPX_PLANE_V
]/2,
1560 stream
->img
->d_w
, stream
->img
->d_h
,
1564 stream
->encoder
.err
= 1;
1565 ctx_exit_on_error(&stream
->encoder
,
1566 "Stream %d: Failed to encode frame.\n"
1567 "Scaling disabled in this configuration. \n"
1568 "To enable, configure with --enable-libyuv\n",
1574 if (img
&& (img
->d_w
!= cfg
->g_w
|| img
->d_h
!= cfg
->g_h
)) {
1575 if (img
->fmt
!= VPX_IMG_FMT_I420
&& img
->fmt
!= VPX_IMG_FMT_YV12
) {
1576 fprintf(stderr
, "%s can only scale 4:2:0 8bpp inputs\n", exec_name
);
1581 stream
->img
= vpx_img_alloc(NULL
, VPX_IMG_FMT_I420
,
1582 cfg
->g_w
, cfg
->g_h
, 16);
1583 I420Scale(img
->planes
[VPX_PLANE_Y
], img
->stride
[VPX_PLANE_Y
],
1584 img
->planes
[VPX_PLANE_U
], img
->stride
[VPX_PLANE_U
],
1585 img
->planes
[VPX_PLANE_V
], img
->stride
[VPX_PLANE_V
],
1587 stream
->img
->planes
[VPX_PLANE_Y
],
1588 stream
->img
->stride
[VPX_PLANE_Y
],
1589 stream
->img
->planes
[VPX_PLANE_U
],
1590 stream
->img
->stride
[VPX_PLANE_U
],
1591 stream
->img
->planes
[VPX_PLANE_V
],
1592 stream
->img
->stride
[VPX_PLANE_V
],
1593 stream
->img
->d_w
, stream
->img
->d_h
,
1597 stream
->encoder
.err
= 1;
1598 ctx_exit_on_error(&stream
->encoder
,
1599 "Stream %d: Failed to encode frame.\n"
1600 "Scaling disabled in this configuration. \n"
1601 "To enable, configure with --enable-libyuv\n",
1606 vpx_usec_timer_start(&timer
);
1607 vpx_codec_encode(&stream
->encoder
, img
, frame_start
,
1608 (unsigned long)(next_frame_start
- frame_start
),
1609 0, global
->deadline
);
1610 vpx_usec_timer_mark(&timer
);
1611 stream
->cx_time
+= vpx_usec_timer_elapsed(&timer
);
1612 ctx_exit_on_error(&stream
->encoder
, "Stream %d: Failed to encode frame",
1617 static void update_quantizer_histogram(struct stream_state
*stream
) {
1618 if (stream
->config
.cfg
.g_pass
!= VPX_RC_FIRST_PASS
) {
1621 vpx_codec_control(&stream
->encoder
, VP8E_GET_LAST_QUANTIZER_64
, &q
);
1622 ctx_exit_on_error(&stream
->encoder
, "Failed to read quantizer");
1623 stream
->counts
[q
]++;
1628 static void get_cx_data(struct stream_state
*stream
,
1629 struct VpxEncoderConfig
*global
,
1631 const vpx_codec_cx_pkt_t
*pkt
;
1632 const struct vpx_codec_enc_cfg
*cfg
= &stream
->config
.cfg
;
1633 vpx_codec_iter_t iter
= NULL
;
1636 while ((pkt
= vpx_codec_get_cx_data(&stream
->encoder
, &iter
))) {
1637 static size_t fsize
= 0;
1638 static int64_t ivf_header_pos
= 0;
1640 switch (pkt
->kind
) {
1641 case VPX_CODEC_CX_FRAME_PKT
:
1642 if (!(pkt
->data
.frame
.flags
& VPX_FRAME_IS_FRAGMENT
)) {
1643 stream
->frames_out
++;
1646 fprintf(stderr
, " %6luF", (unsigned long)pkt
->data
.frame
.sz
);
1648 update_rate_histogram(stream
->rate_hist
, cfg
, pkt
);
1650 if (stream
->config
.write_webm
) {
1651 write_webm_block(&stream
->ebml
, cfg
, pkt
);
1654 if (!stream
->config
.write_webm
) {
1655 if (pkt
->data
.frame
.partition_id
<= 0) {
1656 ivf_header_pos
= ftello(stream
->file
);
1657 fsize
= pkt
->data
.frame
.sz
;
1659 ivf_write_frame_header(stream
->file
, pkt
->data
.frame
.pts
, fsize
);
1661 fsize
+= pkt
->data
.frame
.sz
;
1663 if (!(pkt
->data
.frame
.flags
& VPX_FRAME_IS_FRAGMENT
)) {
1664 const int64_t currpos
= ftello(stream
->file
);
1665 fseeko(stream
->file
, ivf_header_pos
, SEEK_SET
);
1666 ivf_write_frame_size(stream
->file
, fsize
);
1667 fseeko(stream
->file
, currpos
, SEEK_SET
);
1671 (void) fwrite(pkt
->data
.frame
.buf
, 1, pkt
->data
.frame
.sz
,
1674 stream
->nbytes
+= pkt
->data
.raw
.sz
;
1678 if (global
->test_decode
!= TEST_DECODE_OFF
&& !stream
->mismatch_seen
) {
1679 vpx_codec_decode(&stream
->decoder
, pkt
->data
.frame
.buf
,
1680 (unsigned int)pkt
->data
.frame
.sz
, NULL
, 0);
1681 if (stream
->decoder
.err
) {
1682 warn_or_exit_on_error(&stream
->decoder
,
1683 global
->test_decode
== TEST_DECODE_FATAL
,
1684 "Failed to decode frame %d in stream %d",
1685 stream
->frames_out
+ 1, stream
->index
);
1686 stream
->mismatch_seen
= stream
->frames_out
+ 1;
1691 case VPX_CODEC_STATS_PKT
:
1692 stream
->frames_out
++;
1693 stats_write(&stream
->stats
,
1694 pkt
->data
.twopass_stats
.buf
,
1695 pkt
->data
.twopass_stats
.sz
);
1696 stream
->nbytes
+= pkt
->data
.raw
.sz
;
1698 #if CONFIG_FP_MB_STATS
1699 case VPX_CODEC_FPMB_STATS_PKT
:
1700 stats_write(&stream
->fpmb_stats
,
1701 pkt
->data
.firstpass_mb_stats
.buf
,
1702 pkt
->data
.firstpass_mb_stats
.sz
);
1703 stream
->nbytes
+= pkt
->data
.raw
.sz
;
1706 case VPX_CODEC_PSNR_PKT
:
1708 if (global
->show_psnr
) {
1711 stream
->psnr_sse_total
+= pkt
->data
.psnr
.sse
[0];
1712 stream
->psnr_samples_total
+= pkt
->data
.psnr
.samples
[0];
1713 for (i
= 0; i
< 4; i
++) {
1715 fprintf(stderr
, "%.3f ", pkt
->data
.psnr
.psnr
[i
]);
1716 stream
->psnr_totals
[i
] += pkt
->data
.psnr
.psnr
[i
];
1718 stream
->psnr_count
++;
1729 static void show_psnr(struct stream_state
*stream
, double peak
) {
1733 if (!stream
->psnr_count
)
1736 fprintf(stderr
, "Stream %d PSNR (Overall/Avg/Y/U/V)", stream
->index
);
1737 ovpsnr
= sse_to_psnr((double)stream
->psnr_samples_total
, peak
,
1738 (double)stream
->psnr_sse_total
);
1739 fprintf(stderr
, " %.3f", ovpsnr
);
1741 for (i
= 0; i
< 4; i
++) {
1742 fprintf(stderr
, " %.3f", stream
->psnr_totals
[i
] / stream
->psnr_count
);
1744 fprintf(stderr
, "\n");
1748 static float usec_to_fps(uint64_t usec
, unsigned int frames
) {
1749 return (float)(usec
> 0 ? frames
* 1000000.0 / (float)usec
: 0);
1752 static void test_decode(struct stream_state
*stream
,
1753 enum TestDecodeFatality fatal
,
1754 const VpxInterface
*codec
) {
1755 vpx_image_t enc_img
, dec_img
;
1757 if (stream
->mismatch_seen
)
1760 /* Get the internal reference frame */
1761 if (strcmp(codec
->name
, "vp8") == 0) {
1762 struct vpx_ref_frame ref_enc
, ref_dec
;
1765 width
= (stream
->config
.cfg
.g_w
+ 15) & ~15;
1766 height
= (stream
->config
.cfg
.g_h
+ 15) & ~15;
1767 vpx_img_alloc(&ref_enc
.img
, VPX_IMG_FMT_I420
, width
, height
, 1);
1768 enc_img
= ref_enc
.img
;
1769 vpx_img_alloc(&ref_dec
.img
, VPX_IMG_FMT_I420
, width
, height
, 1);
1770 dec_img
= ref_dec
.img
;
1772 ref_enc
.frame_type
= VP8_LAST_FRAME
;
1773 ref_dec
.frame_type
= VP8_LAST_FRAME
;
1774 vpx_codec_control(&stream
->encoder
, VP8_COPY_REFERENCE
, &ref_enc
);
1775 vpx_codec_control(&stream
->decoder
, VP8_COPY_REFERENCE
, &ref_dec
);
1777 struct vp9_ref_frame ref_enc
, ref_dec
;
1781 vpx_codec_control(&stream
->encoder
, VP9_GET_REFERENCE
, &ref_enc
);
1782 enc_img
= ref_enc
.img
;
1783 vpx_codec_control(&stream
->decoder
, VP9_GET_REFERENCE
, &ref_dec
);
1784 dec_img
= ref_dec
.img
;
1785 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1786 if ((enc_img
.fmt
& VPX_IMG_FMT_HIGHBITDEPTH
) !=
1787 (dec_img
.fmt
& VPX_IMG_FMT_HIGHBITDEPTH
)) {
1788 if (enc_img
.fmt
& VPX_IMG_FMT_HIGHBITDEPTH
) {
1789 vpx_img_alloc(&enc_img
, enc_img
.fmt
- VPX_IMG_FMT_HIGHBITDEPTH
,
1790 enc_img
.d_w
, enc_img
.d_h
, 16);
1791 vpx_img_truncate_16_to_8(&enc_img
, &ref_enc
.img
);
1793 if (dec_img
.fmt
& VPX_IMG_FMT_HIGHBITDEPTH
) {
1794 vpx_img_alloc(&dec_img
, dec_img
.fmt
- VPX_IMG_FMT_HIGHBITDEPTH
,
1795 dec_img
.d_w
, dec_img
.d_h
, 16);
1796 vpx_img_truncate_16_to_8(&dec_img
, &ref_dec
.img
);
1801 ctx_exit_on_error(&stream
->encoder
, "Failed to get encoder reference frame");
1802 ctx_exit_on_error(&stream
->decoder
, "Failed to get decoder reference frame");
1804 if (!compare_img(&enc_img
, &dec_img
)) {
1805 int y
[4], u
[4], v
[4];
1806 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1807 if (enc_img
.fmt
& VPX_IMG_FMT_HIGHBITDEPTH
) {
1808 find_mismatch_high(&enc_img
, &dec_img
, y
, u
, v
);
1810 find_mismatch(&enc_img
, &dec_img
, y
, u
, v
);
1813 find_mismatch(&enc_img
, &dec_img
, y
, u
, v
);
1815 stream
->decoder
.err
= 1;
1816 warn_or_exit_on_error(&stream
->decoder
, fatal
== TEST_DECODE_FATAL
,
1817 "Stream %d: Encode/decode mismatch on frame %d at"
1818 " Y[%d, %d] {%d/%d},"
1819 " U[%d, %d] {%d/%d},"
1820 " V[%d, %d] {%d/%d}",
1821 stream
->index
, stream
->frames_out
,
1822 y
[0], y
[1], y
[2], y
[3],
1823 u
[0], u
[1], u
[2], u
[3],
1824 v
[0], v
[1], v
[2], v
[3]);
1825 stream
->mismatch_seen
= stream
->frames_out
;
1828 vpx_img_free(&enc_img
);
1829 vpx_img_free(&dec_img
);
1833 static void print_time(const char *label
, int64_t etl
) {
1840 etl
-= hours
* 3600;
1845 fprintf(stderr
, "[%3s %2"PRId64
":%02"PRId64
":%02"PRId64
"] ",
1846 label
, hours
, mins
, secs
);
1848 fprintf(stderr
, "[%3s unknown] ", label
);
1853 int main(int argc
, const char **argv_
) {
1856 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1857 vpx_image_t raw_shift
;
1858 int allocated_raw_shift
= 0;
1859 int use_16bit_internal
= 0;
1860 int input_shift
= 0;
1862 int frame_avail
, got_data
;
1864 struct VpxInputContext input
;
1865 struct VpxEncoderConfig global
;
1866 struct stream_state
*streams
= NULL
;
1867 char **argv
, **argi
;
1868 uint64_t cx_time
= 0;
1872 memset(&input
, 0, sizeof(input
));
1873 exec_name
= argv_
[0];
1878 /* Setup default input stream settings */
1879 input
.framerate
.numerator
= 30;
1880 input
.framerate
.denominator
= 1;
1881 input
.only_i420
= 1;
1882 input
.bit_depth
= 0;
1884 /* First parse the global configuration values, because we want to apply
1885 * other parameters on top of the default configuration provided by the
1888 argv
= argv_dup(argc
- 1, argv_
+ 1);
1889 parse_global_config(&global
, argv
);
1891 switch (global
.color_type
) {
1893 input
.fmt
= VPX_IMG_FMT_I420
;
1896 input
.fmt
= VPX_IMG_FMT_I422
;
1899 input
.fmt
= VPX_IMG_FMT_I444
;
1902 input
.fmt
= VPX_IMG_FMT_I440
;
1905 input
.fmt
= VPX_IMG_FMT_YV12
;
1910 /* Now parse each stream's parameters. Using a local scope here
1911 * due to the use of 'stream' as loop variable in FOREACH_STREAM
1914 struct stream_state
*stream
= NULL
;
1917 stream
= new_stream(&global
, stream
);
1921 } while (parse_stream_params(&global
, stream
, argv
));
1924 /* Check for unrecognized options */
1925 for (argi
= argv
; *argi
; argi
++)
1926 if (argi
[0][0] == '-' && argi
[0][1])
1927 die("Error: Unrecognized option %s\n", *argi
);
1929 FOREACH_STREAM(check_encoder_config(global
.disable_warning_prompt
,
1930 &global
, &stream
->config
.cfg
););
1932 /* Handle non-option arguments */
1933 input
.filename
= argv
[0];
1935 if (!input
.filename
)
1938 /* Decide if other chroma subsamplings than 4:2:0 are supported */
1939 if (global
.codec
->fourcc
== VP9_FOURCC
)
1940 input
.only_i420
= 0;
1942 for (pass
= global
.pass
? global
.pass
- 1 : 0; pass
< global
.passes
; pass
++) {
1943 int frames_in
= 0, seen_frames
= 0;
1944 int64_t estimated_time_left
= -1;
1945 int64_t average_rate
= -1;
1946 int64_t lagged_count
= 0;
1948 open_input_file(&input
);
1950 /* If the input file doesn't specify its w/h (raw files), try to get
1951 * the data from the first stream's configuration.
1953 if (!input
.width
|| !input
.height
) {
1955 if (stream
->config
.cfg
.g_w
&& stream
->config
.cfg
.g_h
) {
1956 input
.width
= stream
->config
.cfg
.g_w
;
1957 input
.height
= stream
->config
.cfg
.g_h
;
1963 /* Update stream configurations from the input file's parameters */
1964 if (!input
.width
|| !input
.height
)
1965 fatal("Specify stream dimensions with --width (-w) "
1966 " and --height (-h)");
1968 /* If input file does not specify bit-depth but input-bit-depth parameter
1969 * exists, assume that to be the input bit-depth. However, if the
1970 * input-bit-depth paramter does not exist, assume the input bit-depth
1971 * to be the same as the codec bit-depth.
1973 if (!input
.bit_depth
) {
1975 if (stream
->config
.cfg
.g_input_bit_depth
)
1976 input
.bit_depth
= stream
->config
.cfg
.g_input_bit_depth
;
1978 input
.bit_depth
= stream
->config
.cfg
.g_input_bit_depth
=
1979 (int)stream
->config
.cfg
.g_bit_depth
;
1981 if (input
.bit_depth
> 8) input
.fmt
|= VPX_IMG_FMT_HIGHBITDEPTH
;
1984 stream
->config
.cfg
.g_input_bit_depth
= input
.bit_depth
;
1988 FOREACH_STREAM(set_stream_dimensions(stream
, input
.width
, input
.height
));
1989 FOREACH_STREAM(validate_stream_config(stream
, &global
));
1991 /* Ensure that --passes and --pass are consistent. If --pass is set and
1992 * --passes=2, ensure --fpf was set.
1994 if (global
.pass
&& global
.passes
== 2)
1996 if (!stream
->config
.stats_fn
)
1997 die("Stream %d: Must specify --fpf when --pass=%d"
1998 " and --passes=2\n", stream
->index
, global
.pass
);
2003 stream
->config
.write_webm
= 0;
2004 warn("vpxenc was compiled without WebM container support."
2005 "Producing IVF output");
2009 /* Use the frame rate from the file only if none was specified
2010 * on the command-line.
2012 if (!global
.have_framerate
) {
2013 global
.framerate
.num
= input
.framerate
.numerator
;
2014 global
.framerate
.den
= input
.framerate
.denominator
;
2017 FOREACH_STREAM(set_default_kf_interval(stream
, &global
));
2019 /* Show configuration */
2020 if (global
.verbose
&& pass
== 0)
2021 FOREACH_STREAM(show_stream_config(stream
, &global
, &input
));
2023 if (pass
== (global
.pass
? global
.pass
- 1 : 0)) {
2024 if (input
.file_type
== FILE_TYPE_Y4M
)
2025 /*The Y4M reader does its own allocation.
2026 Just initialize this here to avoid problems if we never read any
2028 memset(&raw
, 0, sizeof(raw
));
2030 vpx_img_alloc(&raw
, input
.fmt
, input
.width
, input
.height
, 32);
2032 FOREACH_STREAM(stream
->rate_hist
=
2033 init_rate_histogram(&stream
->config
.cfg
,
2034 &global
.framerate
));
2037 FOREACH_STREAM(setup_pass(stream
, &global
, pass
));
2038 FOREACH_STREAM(open_output_file(stream
, &global
));
2039 FOREACH_STREAM(initialize_encoder(stream
, &global
));
2041 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
2042 if (strcmp(global
.codec
->name
, "vp9") == 0) {
2043 // Check to see if at least one stream uses 16 bit internal.
2044 // Currently assume that the bit_depths for all streams using
2045 // highbitdepth are the same.
2047 if (stream
->config
.use_16bit_internal
) {
2048 use_16bit_internal
= 1;
2050 if (stream
->config
.cfg
.g_profile
== 0) {
2053 input_shift
= (int)stream
->config
.cfg
.g_bit_depth
-
2054 stream
->config
.cfg
.g_input_bit_depth
;
2063 while (frame_avail
|| got_data
) {
2064 struct vpx_usec_timer timer
;
2066 if (!global
.limit
|| frames_in
< global
.limit
) {
2067 frame_avail
= read_frame(&input
, &raw
);
2071 seen_frames
= frames_in
> global
.skip_frames
?
2072 frames_in
- global
.skip_frames
: 0;
2074 if (!global
.quiet
) {
2075 float fps
= usec_to_fps(cx_time
, seen_frames
);
2076 fprintf(stderr
, "\rPass %d/%d ", pass
+ 1, global
.passes
);
2078 if (stream_cnt
== 1)
2080 "frame %4d/%-4d %7"PRId64
"B ",
2081 frames_in
, streams
->frames_out
, (int64_t)streams
->nbytes
);
2083 fprintf(stderr
, "frame %4d ", frames_in
);
2085 fprintf(stderr
, "%7"PRId64
" %s %.2f %s ",
2086 cx_time
> 9999999 ? cx_time
/ 1000 : cx_time
,
2087 cx_time
> 9999999 ? "ms" : "us",
2088 fps
>= 1.0 ? fps
: fps
* 60,
2089 fps
>= 1.0 ? "fps" : "fpm");
2090 print_time("ETA", estimated_time_left
);
2096 if (frames_in
> global
.skip_frames
) {
2097 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
2098 vpx_image_t
*frame_to_encode
;
2099 if (input_shift
|| (use_16bit_internal
&& input
.bit_depth
== 8)) {
2100 assert(use_16bit_internal
);
2101 // Input bit depth and stream bit depth do not match, so up
2102 // shift frame to stream bit depth
2103 if (!allocated_raw_shift
) {
2104 vpx_img_alloc(&raw_shift
, raw
.fmt
| VPX_IMG_FMT_HIGHBITDEPTH
,
2105 input
.width
, input
.height
, 32);
2106 allocated_raw_shift
= 1;
2108 vpx_img_upshift(&raw_shift
, &raw
, input_shift
);
2109 frame_to_encode
= &raw_shift
;
2111 frame_to_encode
= &raw
;
2113 vpx_usec_timer_start(&timer
);
2114 if (use_16bit_internal
) {
2115 assert(frame_to_encode
->fmt
& VPX_IMG_FMT_HIGHBITDEPTH
);
2117 if (stream
->config
.use_16bit_internal
)
2118 encode_frame(stream
, &global
,
2119 frame_avail
? frame_to_encode
: NULL
,
2125 assert((frame_to_encode
->fmt
& VPX_IMG_FMT_HIGHBITDEPTH
) == 0);
2126 FOREACH_STREAM(encode_frame(stream
, &global
,
2127 frame_avail
? frame_to_encode
: NULL
,
2131 vpx_usec_timer_start(&timer
);
2132 FOREACH_STREAM(encode_frame(stream
, &global
,
2133 frame_avail
? &raw
: NULL
,
2136 vpx_usec_timer_mark(&timer
);
2137 cx_time
+= vpx_usec_timer_elapsed(&timer
);
2139 FOREACH_STREAM(update_quantizer_histogram(stream
));
2142 FOREACH_STREAM(get_cx_data(stream
, &global
, &got_data
));
2144 if (!got_data
&& input
.length
&& streams
!= NULL
&&
2145 !streams
->frames_out
) {
2146 lagged_count
= global
.limit
? seen_frames
: ftello(input
.file
);
2147 } else if (input
.length
) {
2152 const int64_t frame_in_lagged
= (seen_frames
- lagged_count
) * 1000;
2154 rate
= cx_time
? frame_in_lagged
* (int64_t)1000000 / cx_time
: 0;
2155 remaining
= 1000 * (global
.limit
- global
.skip_frames
2156 - seen_frames
+ lagged_count
);
2158 const int64_t input_pos
= ftello(input
.file
);
2159 const int64_t input_pos_lagged
= input_pos
- lagged_count
;
2160 const int64_t limit
= input
.length
;
2162 rate
= cx_time
? input_pos_lagged
* (int64_t)1000000 / cx_time
: 0;
2163 remaining
= limit
- input_pos
+ lagged_count
;
2166 average_rate
= (average_rate
<= 0)
2168 : (average_rate
* 7 + rate
) / 8;
2169 estimated_time_left
= average_rate
? remaining
/ average_rate
: -1;
2172 if (got_data
&& global
.test_decode
!= TEST_DECODE_OFF
)
2173 FOREACH_STREAM(test_decode(stream
, global
.test_decode
, global
.codec
));
2178 fprintf(stderr
, "\033[K");
2182 fprintf(stderr
, "\n");
2184 if (!global
.quiet
) {
2185 FOREACH_STREAM(fprintf(stderr
,
2186 "\rPass %d/%d frame %4d/%-4d %7"PRId64
"B %7"PRId64
"b/f %7"PRId64
"b/s"
2187 " %7"PRId64
" %s (%.2f fps)\033[K\n",
2189 global
.passes
, frames_in
, stream
->frames_out
, (int64_t)stream
->nbytes
,
2190 seen_frames
? (int64_t)(stream
->nbytes
* 8 / seen_frames
) : 0,
2191 seen_frames
? (int64_t)stream
->nbytes
* 8 *
2192 (int64_t)global
.framerate
.num
/ global
.framerate
.den
/
2194 stream
->cx_time
> 9999999 ? stream
->cx_time
/ 1000 : stream
->cx_time
,
2195 stream
->cx_time
> 9999999 ? "ms" : "us",
2196 usec_to_fps(stream
->cx_time
, seen_frames
)));
2199 if (global
.show_psnr
) {
2200 if (global
.codec
->fourcc
== VP9_FOURCC
) {
2202 show_psnr(stream
, (1 << stream
->config
.cfg
.g_input_bit_depth
) - 1));
2204 FOREACH_STREAM(show_psnr(stream
, 255.0));
2208 FOREACH_STREAM(vpx_codec_destroy(&stream
->encoder
));
2210 if (global
.test_decode
!= TEST_DECODE_OFF
) {
2211 FOREACH_STREAM(vpx_codec_destroy(&stream
->decoder
));
2214 close_input_file(&input
);
2216 if (global
.test_decode
== TEST_DECODE_FATAL
) {
2217 FOREACH_STREAM(res
|= stream
->mismatch_seen
);
2219 FOREACH_STREAM(close_output_file(stream
, global
.codec
->fourcc
));
2221 FOREACH_STREAM(stats_close(&stream
->stats
, global
.passes
- 1));
2223 #if CONFIG_FP_MB_STATS
2224 FOREACH_STREAM(stats_close(&stream
->fpmb_stats
, global
.passes
- 1));
2231 if (global
.show_q_hist_buckets
)
2232 FOREACH_STREAM(show_q_histogram(stream
->counts
,
2233 global
.show_q_hist_buckets
));
2235 if (global
.show_rate_hist_buckets
)
2236 FOREACH_STREAM(show_rate_histogram(stream
->rate_hist
,
2237 &stream
->config
.cfg
,
2238 global
.show_rate_hist_buckets
));
2239 FOREACH_STREAM(destroy_rate_histogram(stream
->rate_hist
));
2241 #if CONFIG_INTERNAL_STATS
2242 /* TODO(jkoleszar): This doesn't belong in this executable. Do it for now,
2243 * to match some existing utilities.
2245 if (!(global
.pass
== 1 && global
.passes
== 2))
2247 FILE *f
= fopen("opsnr.stt", "a");
2248 if (stream
->mismatch_seen
) {
2249 fprintf(f
, "First mismatch occurred in frame %d\n",
2250 stream
->mismatch_seen
);
2252 fprintf(f
, "No mismatch detected in recon buffers\n");
2258 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
2259 if (allocated_raw_shift
)
2260 vpx_img_free(&raw_shift
);
2265 return res
? EXIT_FAILURE
: EXIT_SUCCESS
;