Chromium Code Reviews
chromiumcodereview-hr@appspot.gserviceaccount.com (chromiumcodereview-hr) | Please choose your nickname with Settings | Help | Chromium Project | Gerrit Changes | Sign out
(435)

Side by Side Diff: source/libvpx/vpxenc.c

Issue 11555023: libvpx: Add VP9 decoder. (Closed) Base URL: svn://chrome-svn/chrome/trunk/deps/third_party/libvpx/
Patch Set: Created 8 years ago
Use n/p to move between diff chunks; N/P to move between comments. Draft comments are only viewable by you.
Jump to:
View unified diff | Download patch | Annotate | Revision Log
OLDNEW
1 /* 1 /*
2 * Copyright (c) 2010 The WebM project authors. All Rights Reserved. 2 * Copyright (c) 2010 The WebM project authors. All Rights Reserved.
3 * 3 *
4 * Use of this source code is governed by a BSD-style license 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 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 6 * tree. An additional intellectual property rights grant can be found
7 * in the file PATENTS. All contributing project authors may 7 * in the file PATENTS. All contributing project authors may
8 * be found in the AUTHORS file in the root of the source tree. 8 * be found in the AUTHORS file in the root of the source tree.
9 */ 9 */
10 10
11 #include "vpx_config.h"
11 12
12 /* This is a simple program that encodes YV12 files and generates ivf
13 * files using the new interface.
14 */
15 #if defined(_WIN32) || !CONFIG_OS_SUPPORT 13 #if defined(_WIN32) || !CONFIG_OS_SUPPORT
16 #define USE_POSIX_MMAP 0 14 #define USE_POSIX_MMAP 0
17 #else 15 #else
18 #define USE_POSIX_MMAP 1 16 #define USE_POSIX_MMAP 1
19 #endif 17 #endif
20 18
21 #include <stdio.h> 19 #include <stdio.h>
22 #include <stdlib.h> 20 #include <stdlib.h>
23 #include <stdarg.h> 21 #include <stdarg.h>
24 #include <string.h> 22 #include <string.h>
25 #include <limits.h> 23 #include <limits.h>
26 #include <assert.h> 24 #include <assert.h>
27 #include "vpx/vpx_encoder.h" 25 #include "vpx/vpx_encoder.h"
26 #include "vpx/vpx_decoder.h"
28 #if USE_POSIX_MMAP 27 #if USE_POSIX_MMAP
29 #include <sys/types.h> 28 #include <sys/types.h>
30 #include <sys/stat.h> 29 #include <sys/stat.h>
31 #include <sys/mman.h> 30 #include <sys/mman.h>
32 #include <fcntl.h> 31 #include <fcntl.h>
33 #include <unistd.h> 32 #include <unistd.h>
34 #endif 33 #endif
34
35 #if CONFIG_VP8_ENCODER || CONFIG_VP9_ENCODER
35 #include "vpx/vp8cx.h" 36 #include "vpx/vp8cx.h"
37 #endif
38 #if CONFIG_VP8_DECODER || CONFIG_VP9_DECODER
39 #include "vpx/vp8dx.h"
40 #endif
41
36 #include "vpx_ports/mem_ops.h" 42 #include "vpx_ports/mem_ops.h"
37 #include "vpx_ports/vpx_timer.h" 43 #include "vpx_ports/vpx_timer.h"
38 #include "tools_common.h" 44 #include "tools_common.h"
39 #include "y4minput.h" 45 #include "y4minput.h"
40 #include "libmkv/EbmlWriter.h" 46 #include "libmkv/EbmlWriter.h"
41 #include "libmkv/EbmlIDs.h" 47 #include "libmkv/EbmlIDs.h"
42 48
43 /* Need special handling of these functions on Windows */ 49 /* Need special handling of these functions on Windows */
44 #if defined(_MSC_VER) 50 #if defined(_MSC_VER)
45 /* MSVS doesn't define off_t, and uses _f{seek,tell}i64 */ 51 /* MSVS doesn't define off_t, and uses _f{seek,tell}i64 */
(...skipping 13 matching lines...) Expand all
59 /* We should use 32-bit file operations in WebM file format 65 /* We should use 32-bit file operations in WebM file format
60 * when building ARM executable file (.axf) with RVCT */ 66 * when building ARM executable file (.axf) with RVCT */
61 #if !CONFIG_OS_SUPPORT 67 #if !CONFIG_OS_SUPPORT
62 typedef long off_t; 68 typedef long off_t;
63 #define fseeko fseek 69 #define fseeko fseek
64 #define ftello ftell 70 #define ftello ftell
65 #endif 71 #endif
66 72
67 /* Swallow warnings about unused results of fread/fwrite */ 73 /* Swallow warnings about unused results of fread/fwrite */
68 static size_t wrap_fread(void *ptr, size_t size, size_t nmemb, 74 static size_t wrap_fread(void *ptr, size_t size, size_t nmemb,
69 FILE *stream) 75 FILE *stream) {
70 { 76 return fread(ptr, size, nmemb, stream);
71 return fread(ptr, size, nmemb, stream);
72 } 77 }
73 #define fread wrap_fread 78 #define fread wrap_fread
74 79
75 static size_t wrap_fwrite(const void *ptr, size_t size, size_t nmemb, 80 static size_t wrap_fwrite(const void *ptr, size_t size, size_t nmemb,
76 FILE *stream) 81 FILE *stream) {
77 { 82 return fwrite(ptr, size, nmemb, stream);
78 return fwrite(ptr, size, nmemb, stream);
79 } 83 }
80 #define fwrite wrap_fwrite 84 #define fwrite wrap_fwrite
81 85
82 86
83 static const char *exec_name; 87 static const char *exec_name;
84 88
85 static const struct codec_item 89 #define VP8_FOURCC (0x00385056)
86 { 90 #define VP9_FOURCC (0x00395056)
87 char const *name; 91 static const struct codec_item {
88 vpx_codec_iface_t *iface; 92 char const *name;
89 unsigned int fourcc; 93 const vpx_codec_iface_t *(*iface)(void);
90 } codecs[] = 94 const vpx_codec_iface_t *(*dx_iface)(void);
91 { 95 unsigned int fourcc;
92 #if CONFIG_VP8_ENCODER 96 } codecs[] = {
93 {"vp8", &vpx_codec_vp8_cx_algo, 0x30385056}, 97 #if CONFIG_VP8_ENCODER && CONFIG_VP8_DECODER
98 {"vp8", &vpx_codec_vp8_cx, &vpx_codec_vp8_dx, VP8_FOURCC},
99 #elif CONFIG_VP8_ENCODER && !CONFIG_VP8_DECODER
100 {"vp8", &vpx_codec_vp8_cx, NULL, VP8_FOURCC},
101 #endif
102 #if CONFIG_VP9_ENCODER && CONFIG_VP9_DECODER
103 {"vp9", &vpx_codec_vp9_cx, &vpx_codec_vp9_dx, VP9_FOURCC},
104 #elif CONFIG_VP9_ENCODER && !CONFIG_VP9_DECODER
105 {"vp9", &vpx_codec_vp9_cx, NULL, VP9_FOURCC},
94 #endif 106 #endif
95 }; 107 };
96 108
97 static void usage_exit(); 109 static void usage_exit();
98 110
99 #define LOG_ERROR(label) do \ 111 #define LOG_ERROR(label) do \
100 {\ 112 {\
101 const char *l=label;\ 113 const char *l=label;\
102 va_list ap;\ 114 va_list ap;\
103 va_start(ap, fmt);\ 115 va_start(ap, fmt);\
104 if(l)\ 116 if(l)\
105 fprintf(stderr, "%s: ", l);\ 117 fprintf(stderr, "%s: ", l);\
106 vfprintf(stderr, fmt, ap);\ 118 vfprintf(stderr, fmt, ap);\
107 fprintf(stderr, "\n");\ 119 fprintf(stderr, "\n");\
108 va_end(ap);\ 120 va_end(ap);\
109 } while(0) 121 } while(0)
110 122
111 void die(const char *fmt, ...) 123 void die(const char *fmt, ...) {
112 { 124 LOG_ERROR(NULL);
113 LOG_ERROR(NULL); 125 usage_exit();
114 usage_exit(); 126 }
115 } 127
116 128
117 129 void fatal(const char *fmt, ...) {
118 void fatal(const char *fmt, ...) 130 LOG_ERROR("Fatal");
119 { 131 exit(EXIT_FAILURE);
120 LOG_ERROR("Fatal"); 132 }
133
134
135 void warn(const char *fmt, ...) {
136 LOG_ERROR("Warning");
137 }
138
139
140 static void ctx_exit_on_error(vpx_codec_ctx_t *ctx, const char *s, ...) {
141 va_list ap;
142
143 va_start(ap, s);
144 if (ctx->err) {
145 const char *detail = vpx_codec_error_detail(ctx);
146
147 vfprintf(stderr, s, ap);
148 fprintf(stderr, ": %s\n", vpx_codec_error(ctx));
149
150 if (detail)
151 fprintf(stderr, " %s\n", detail);
152
121 exit(EXIT_FAILURE); 153 exit(EXIT_FAILURE);
122 } 154 }
123
124
125 void warn(const char *fmt, ...)
126 {
127 LOG_ERROR("Warning");
128 }
129
130
131 static void ctx_exit_on_error(vpx_codec_ctx_t *ctx, const char *s, ...)
132 {
133 va_list ap;
134
135 va_start(ap, s);
136 if (ctx->err)
137 {
138 const char *detail = vpx_codec_error_detail(ctx);
139
140 vfprintf(stderr, s, ap);
141 fprintf(stderr, ": %s\n", vpx_codec_error(ctx));
142
143 if (detail)
144 fprintf(stderr, " %s\n", detail);
145
146 exit(EXIT_FAILURE);
147 }
148 } 155 }
149 156
150 /* This structure is used to abstract the different ways of handling 157 /* This structure is used to abstract the different ways of handling
151 * first pass statistics. 158 * first pass statistics.
152 */ 159 */
153 typedef struct 160 typedef struct {
154 { 161 vpx_fixed_buf_t buf;
155 vpx_fixed_buf_t buf; 162 int pass;
156 int pass; 163 FILE *file;
157 FILE *file; 164 char *buf_ptr;
158 char *buf_ptr; 165 size_t buf_alloc_sz;
159 size_t buf_alloc_sz;
160 } stats_io_t; 166 } stats_io_t;
161 167
162 int stats_open_file(stats_io_t *stats, const char *fpf, int pass) 168 int stats_open_file(stats_io_t *stats, const char *fpf, int pass) {
163 { 169 int res;
164 int res; 170
165 171 stats->pass = pass;
166 stats->pass = pass; 172
167 173 if (pass == 0) {
168 if (pass == 0) 174 stats->file = fopen(fpf, "wb");
169 { 175 stats->buf.sz = 0;
170 stats->file = fopen(fpf, "wb"); 176 stats->buf.buf = NULL,
171 stats->buf.sz = 0; 177 res = (stats->file != NULL);
172 stats->buf.buf = NULL, 178 } else {
173 res = (stats->file != NULL);
174 }
175 else
176 {
177 #if 0 179 #if 0
178 #elif USE_POSIX_MMAP 180 #elif USE_POSIX_MMAP
179 struct stat stat_buf; 181 struct stat stat_buf;
180 int fd; 182 int fd;
181 183
182 fd = open(fpf, O_RDONLY); 184 fd = open(fpf, O_RDONLY);
183 stats->file = fdopen(fd, "rb"); 185 stats->file = fdopen(fd, "rb");
184 fstat(fd, &stat_buf); 186 fstat(fd, &stat_buf);
185 stats->buf.sz = stat_buf.st_size; 187 stats->buf.sz = stat_buf.st_size;
186 stats->buf.buf = mmap(NULL, stats->buf.sz, PROT_READ, MAP_PRIVATE, 188 stats->buf.buf = mmap(NULL, stats->buf.sz, PROT_READ, MAP_PRIVATE,
187 fd, 0); 189 fd, 0);
188 res = (stats->buf.buf != NULL); 190 res = (stats->buf.buf != NULL);
189 #else 191 #else
190 size_t nbytes; 192 size_t nbytes;
191 193
192 stats->file = fopen(fpf, "rb"); 194 stats->file = fopen(fpf, "rb");
193 195
194 if (fseek(stats->file, 0, SEEK_END)) 196 if (fseek(stats->file, 0, SEEK_END))
195 fatal("First-pass stats file must be seekable!"); 197 fatal("First-pass stats file must be seekable!");
196 198
197 stats->buf.sz = stats->buf_alloc_sz = ftell(stats->file); 199 stats->buf.sz = stats->buf_alloc_sz = ftell(stats->file);
198 rewind(stats->file); 200 rewind(stats->file);
199 201
200 stats->buf.buf = malloc(stats->buf_alloc_sz); 202 stats->buf.buf = malloc(stats->buf_alloc_sz);
201 203
202 if (!stats->buf.buf) 204 if (!stats->buf.buf)
203 fatal("Failed to allocate first-pass stats buffer (%lu bytes)", 205 fatal("Failed to allocate first-pass stats buffer (%lu bytes)",
204 (unsigned long)stats->buf_alloc_sz); 206 (unsigned long)stats->buf_alloc_sz);
205 207
206 nbytes = fread(stats->buf.buf, 1, stats->buf.sz, stats->file); 208 nbytes = fread(stats->buf.buf, 1, stats->buf.sz, stats->file);
207 res = (nbytes == stats->buf.sz); 209 res = (nbytes == stats->buf.sz);
210 #endif
211 }
212
213 return res;
214 }
215
216 int stats_open_mem(stats_io_t *stats, int pass) {
217 int res;
218 stats->pass = pass;
219
220 if (!pass) {
221 stats->buf.sz = 0;
222 stats->buf_alloc_sz = 64 * 1024;
223 stats->buf.buf = malloc(stats->buf_alloc_sz);
224 }
225
226 stats->buf_ptr = stats->buf.buf;
227 res = (stats->buf.buf != NULL);
228 return res;
229 }
230
231
232 void stats_close(stats_io_t *stats, int last_pass) {
233 if (stats->file) {
234 if (stats->pass == last_pass) {
235 #if 0
236 #elif USE_POSIX_MMAP
237 munmap(stats->buf.buf, stats->buf.sz);
238 #else
239 free(stats->buf.buf);
208 #endif 240 #endif
209 } 241 }
210 242
211 return res; 243 fclose(stats->file);
212 } 244 stats->file = NULL;
213 245 } else {
214 int stats_open_mem(stats_io_t *stats, int pass) 246 if (stats->pass == last_pass)
215 { 247 free(stats->buf.buf);
216 int res; 248 }
217 stats->pass = pass; 249 }
218 250
219 if (!pass) 251 void stats_write(stats_io_t *stats, const void *pkt, size_t len) {
220 { 252 if (stats->file) {
221 stats->buf.sz = 0; 253 (void) fwrite(pkt, 1, len, stats->file);
222 stats->buf_alloc_sz = 64 * 1024; 254 } else {
223 stats->buf.buf = malloc(stats->buf_alloc_sz); 255 if (stats->buf.sz + len > stats->buf_alloc_sz) {
256 size_t new_sz = stats->buf_alloc_sz + 64 * 1024;
257 char *new_ptr = realloc(stats->buf.buf, new_sz);
258
259 if (new_ptr) {
260 stats->buf_ptr = new_ptr + (stats->buf_ptr - (char *)stats->buf.buf);
261 stats->buf.buf = new_ptr;
262 stats->buf_alloc_sz = new_sz;
263 } else
264 fatal("Failed to realloc firstpass stats buffer.");
224 } 265 }
225 266
226 stats->buf_ptr = stats->buf.buf; 267 memcpy(stats->buf_ptr, pkt, len);
227 res = (stats->buf.buf != NULL); 268 stats->buf.sz += len;
228 return res; 269 stats->buf_ptr += len;
229 } 270 }
230 271 }
231 272
232 void stats_close(stats_io_t *stats, int last_pass) 273 vpx_fixed_buf_t stats_get(stats_io_t *stats) {
233 { 274 return stats->buf;
234 if (stats->file) 275 }
235 { 276
236 if (stats->pass == last_pass) 277 /* Stereo 3D packed frame format */
237 { 278 typedef enum stereo_format {
238 #if 0 279 STEREO_FORMAT_MONO = 0,
239 #elif USE_POSIX_MMAP 280 STEREO_FORMAT_LEFT_RIGHT = 1,
240 munmap(stats->buf.buf, stats->buf.sz); 281 STEREO_FORMAT_BOTTOM_TOP = 2,
241 #else 282 STEREO_FORMAT_TOP_BOTTOM = 3,
242 free(stats->buf.buf); 283 STEREO_FORMAT_RIGHT_LEFT = 11
243 #endif 284 } stereo_format_t;
285
286 enum video_file_type {
287 FILE_TYPE_RAW,
288 FILE_TYPE_IVF,
289 FILE_TYPE_Y4M
290 };
291
292 struct detect_buffer {
293 char buf[4];
294 size_t buf_read;
295 size_t position;
296 };
297
298
299 struct input_state {
300 char *fn;
301 FILE *file;
302 y4m_input y4m;
303 struct detect_buffer detect;
304 enum video_file_type file_type;
305 unsigned int w;
306 unsigned int h;
307 struct vpx_rational framerate;
308 int use_i420;
309 };
310
311
312 #define IVF_FRAME_HDR_SZ (4+8) /* 4 byte size + 8 byte timestamp */
313 static int read_frame(struct input_state *input, vpx_image_t *img) {
314 FILE *f = input->file;
315 enum video_file_type file_type = input->file_type;
316 y4m_input *y4m = &input->y4m;
317 struct detect_buffer *detect = &input->detect;
318 int plane = 0;
319 int shortread = 0;
320
321 if (file_type == FILE_TYPE_Y4M) {
322 if (y4m_input_fetch_frame(y4m, f, img) < 1)
323 return 0;
324 } else {
325 if (file_type == FILE_TYPE_IVF) {
326 char junk[IVF_FRAME_HDR_SZ];
327
328 /* Skip the frame header. We know how big the frame should be. See
329 * write_ivf_frame_header() for documentation on the frame header
330 * layout.
331 */
332 (void) fread(junk, 1, IVF_FRAME_HDR_SZ, f);
333 }
334
335 for (plane = 0; plane < 3; plane++) {
336 unsigned char *ptr;
337 int w = (plane ? (1 + img->d_w) / 2 : img->d_w);
338 int h = (plane ? (1 + img->d_h) / 2 : img->d_h);
339 int r;
340
341 /* Determine the correct plane based on the image format. The for-loop
342 * always counts in Y,U,V order, but this may not match the order of
343 * the data on disk.
344 */
345 switch (plane) {
346 case 1:
347 ptr = img->planes[img->fmt == VPX_IMG_FMT_YV12 ? VPX_PLANE_V : VPX_PLA NE_U];
348 break;
349 case 2:
350 ptr = img->planes[img->fmt == VPX_IMG_FMT_YV12 ? VPX_PLANE_U : VPX_PLA NE_V];
351 break;
352 default:
353 ptr = img->planes[plane];
354 }
355
356 for (r = 0; r < h; r++) {
357 size_t needed = w;
358 size_t buf_position = 0;
359 const size_t left = detect->buf_read - detect->position;
360 if (left > 0) {
361 const size_t more = (left < needed) ? left : needed;
362 memcpy(ptr, detect->buf + detect->position, more);
363 buf_position = more;
364 needed -= more;
365 detect->position += more;
244 } 366 }
245 367 if (needed > 0) {
246 fclose(stats->file); 368 shortread |= (fread(ptr + buf_position, 1, needed, f) < needed);
247 stats->file = NULL; 369 }
370
371 ptr += img->stride[plane];
372 }
248 } 373 }
249 else 374 }
250 { 375
251 if (stats->pass == last_pass) 376 return !shortread;
252 free(stats->buf.buf);
253 }
254 }
255
256 void stats_write(stats_io_t *stats, const void *pkt, size_t len)
257 {
258 if (stats->file)
259 {
260 (void) fwrite(pkt, 1, len, stats->file);
261 }
262 else
263 {
264 if (stats->buf.sz + len > stats->buf_alloc_sz)
265 {
266 size_t new_sz = stats->buf_alloc_sz + 64 * 1024;
267 char *new_ptr = realloc(stats->buf.buf, new_sz);
268
269 if (new_ptr)
270 {
271 stats->buf_ptr = new_ptr + (stats->buf_ptr - (char *)stats->buf. buf);
272 stats->buf.buf = new_ptr;
273 stats->buf_alloc_sz = new_sz;
274 }
275 else
276 fatal("Failed to realloc firstpass stats buffer.");
277 }
278
279 memcpy(stats->buf_ptr, pkt, len);
280 stats->buf.sz += len;
281 stats->buf_ptr += len;
282 }
283 }
284
285 vpx_fixed_buf_t stats_get(stats_io_t *stats)
286 {
287 return stats->buf;
288 }
289
290 /* Stereo 3D packed frame format */
291 typedef enum stereo_format
292 {
293 STEREO_FORMAT_MONO = 0,
294 STEREO_FORMAT_LEFT_RIGHT = 1,
295 STEREO_FORMAT_BOTTOM_TOP = 2,
296 STEREO_FORMAT_TOP_BOTTOM = 3,
297 STEREO_FORMAT_RIGHT_LEFT = 11
298 } stereo_format_t;
299
300 enum video_file_type
301 {
302 FILE_TYPE_RAW,
303 FILE_TYPE_IVF,
304 FILE_TYPE_Y4M
305 };
306
307 struct detect_buffer {
308 char buf[4];
309 size_t buf_read;
310 size_t position;
311 };
312
313
314 struct input_state
315 {
316 char *fn;
317 FILE *file;
318 y4m_input y4m;
319 struct detect_buffer detect;
320 enum video_file_type file_type;
321 unsigned int w;
322 unsigned int h;
323 struct vpx_rational framerate;
324 int use_i420;
325 };
326
327
328 #define IVF_FRAME_HDR_SZ (4+8) /* 4 byte size + 8 byte timestamp */
329 static int read_frame(struct input_state *input, vpx_image_t *img)
330 {
331 FILE *f = input->file;
332 enum video_file_type file_type = input->file_type;
333 y4m_input *y4m = &input->y4m;
334 struct detect_buffer *detect = &input->detect;
335 int plane = 0;
336 int shortread = 0;
337
338 if (file_type == FILE_TYPE_Y4M)
339 {
340 if (y4m_input_fetch_frame(y4m, f, img) < 1)
341 return 0;
342 }
343 else
344 {
345 if (file_type == FILE_TYPE_IVF)
346 {
347 char junk[IVF_FRAME_HDR_SZ];
348
349 /* Skip the frame header. We know how big the frame should be. See
350 * write_ivf_frame_header() for documentation on the frame header
351 * layout.
352 */
353 (void) fread(junk, 1, IVF_FRAME_HDR_SZ, f);
354 }
355
356 for (plane = 0; plane < 3; plane++)
357 {
358 unsigned char *ptr;
359 int w = (plane ? (1 + img->d_w) / 2 : img->d_w);
360 int h = (plane ? (1 + img->d_h) / 2 : img->d_h);
361 int r;
362
363 /* Determine the correct plane based on the image format. The for-lo op
364 * always counts in Y,U,V order, but this may not match the order of
365 * the data on disk.
366 */
367 switch (plane)
368 {
369 case 1:
370 ptr = img->planes[img->fmt==VPX_IMG_FMT_YV12? VPX_PLANE_V : VPX_ PLANE_U];
371 break;
372 case 2:
373 ptr = img->planes[img->fmt==VPX_IMG_FMT_YV12?VPX_PLANE_U : VPX_P LANE_V];
374 break;
375 default:
376 ptr = img->planes[plane];
377 }
378
379 for (r = 0; r < h; r++)
380 {
381 size_t needed = w;
382 size_t buf_position = 0;
383 const size_t left = detect->buf_read - detect->position;
384 if (left > 0)
385 {
386 const size_t more = (left < needed) ? left : needed;
387 memcpy(ptr, detect->buf + detect->position, more);
388 buf_position = more;
389 needed -= more;
390 detect->position += more;
391 }
392 if (needed > 0)
393 {
394 shortread |= (fread(ptr + buf_position, 1, needed, f) < need ed);
395 }
396
397 ptr += img->stride[plane];
398 }
399 }
400 }
401
402 return !shortread;
403 } 377 }
404 378
405 379
406 unsigned int file_is_y4m(FILE *infile, 380 unsigned int file_is_y4m(FILE *infile,
407 y4m_input *y4m, 381 y4m_input *y4m,
408 char detect[4]) 382 char detect[4]) {
409 { 383 if (memcmp(detect, "YUV4", 4) == 0) {
410 if(memcmp(detect, "YUV4", 4) == 0) 384 return 1;
411 { 385 }
412 return 1; 386 return 0;
413 }
414 return 0;
415 } 387 }
416 388
417 #define IVF_FILE_HDR_SZ (32) 389 #define IVF_FILE_HDR_SZ (32)
418 unsigned int file_is_ivf(struct input_state *input, 390 unsigned int file_is_ivf(struct input_state *input,
419 unsigned int *fourcc) 391 unsigned int *fourcc) {
420 { 392 char raw_hdr[IVF_FILE_HDR_SZ];
421 char raw_hdr[IVF_FILE_HDR_SZ]; 393 int is_ivf = 0;
422 int is_ivf = 0; 394 FILE *infile = input->file;
423 FILE *infile = input->file; 395 unsigned int *width = &input->w;
424 unsigned int *width = &input->w; 396 unsigned int *height = &input->h;
425 unsigned int *height = &input->h; 397 struct detect_buffer *detect = &input->detect;
426 struct detect_buffer *detect = &input->detect; 398
427 399 if (memcmp(detect->buf, "DKIF", 4) != 0)
428 if(memcmp(detect->buf, "DKIF", 4) != 0) 400 return 0;
429 return 0; 401
430 402 /* See write_ivf_file_header() for more documentation on the file header
431 /* See write_ivf_file_header() for more documentation on the file header 403 * layout.
432 * layout. 404 */
433 */ 405 if (fread(raw_hdr + 4, 1, IVF_FILE_HDR_SZ - 4, infile)
434 if (fread(raw_hdr + 4, 1, IVF_FILE_HDR_SZ - 4, infile) 406 == IVF_FILE_HDR_SZ - 4) {
435 == IVF_FILE_HDR_SZ - 4)
436 { 407 {
437 { 408 is_ivf = 1;
438 is_ivf = 1; 409
439 410 if (mem_get_le16(raw_hdr + 4) != 0)
440 if (mem_get_le16(raw_hdr + 4) != 0) 411 warn("Unrecognized IVF version! This file may not decode "
441 warn("Unrecognized IVF version! This file may not decode " 412 "properly.");
442 "properly."); 413
443 414 *fourcc = mem_get_le32(raw_hdr + 8);
444 *fourcc = mem_get_le32(raw_hdr + 8);
445 }
446 } 415 }
447 416 }
448 if (is_ivf) 417
449 { 418 if (is_ivf) {
450 *width = mem_get_le16(raw_hdr + 12); 419 *width = mem_get_le16(raw_hdr + 12);
451 *height = mem_get_le16(raw_hdr + 14); 420 *height = mem_get_le16(raw_hdr + 14);
452 detect->position = 4; 421 detect->position = 4;
453 } 422 }
454 423
455 return is_ivf; 424 return is_ivf;
456 } 425 }
457 426
458 427
459 static void write_ivf_file_header(FILE *outfile, 428 static void write_ivf_file_header(FILE *outfile,
460 const vpx_codec_enc_cfg_t *cfg, 429 const vpx_codec_enc_cfg_t *cfg,
461 unsigned int fourcc, 430 unsigned int fourcc,
462 int frame_cnt) 431 int frame_cnt) {
463 { 432 char header[32];
464 char header[32]; 433
465 434 if (cfg->g_pass != VPX_RC_ONE_PASS && cfg->g_pass != VPX_RC_LAST_PASS)
466 if (cfg->g_pass != VPX_RC_ONE_PASS && cfg->g_pass != VPX_RC_LAST_PASS) 435 return;
467 return; 436
468 437 header[0] = 'D';
469 header[0] = 'D'; 438 header[1] = 'K';
470 header[1] = 'K'; 439 header[2] = 'I';
471 header[2] = 'I'; 440 header[3] = 'F';
472 header[3] = 'F'; 441 mem_put_le16(header + 4, 0); /* version */
473 mem_put_le16(header + 4, 0); /* version */ 442 mem_put_le16(header + 6, 32); /* headersize */
474 mem_put_le16(header + 6, 32); /* headersize */ 443 mem_put_le32(header + 8, fourcc); /* headersize */
475 mem_put_le32(header + 8, fourcc); /* headersize */ 444 mem_put_le16(header + 12, cfg->g_w); /* width */
476 mem_put_le16(header + 12, cfg->g_w); /* width */ 445 mem_put_le16(header + 14, cfg->g_h); /* height */
477 mem_put_le16(header + 14, cfg->g_h); /* height */ 446 mem_put_le32(header + 16, cfg->g_timebase.den); /* rate */
478 mem_put_le32(header + 16, cfg->g_timebase.den); /* rate */ 447 mem_put_le32(header + 20, cfg->g_timebase.num); /* scale */
479 mem_put_le32(header + 20, cfg->g_timebase.num); /* scale */ 448 mem_put_le32(header + 24, frame_cnt); /* length */
480 mem_put_le32(header + 24, frame_cnt); /* length */ 449 mem_put_le32(header + 28, 0); /* unused */
481 mem_put_le32(header + 28, 0); /* unused */ 450
482 451 (void) fwrite(header, 1, 32, outfile);
483 (void) fwrite(header, 1, 32, outfile);
484 } 452 }
485 453
486 454
487 static void write_ivf_frame_header(FILE *outfile, 455 static void write_ivf_frame_header(FILE *outfile,
488 const vpx_codec_cx_pkt_t *pkt) 456 const vpx_codec_cx_pkt_t *pkt) {
489 { 457 char header[12];
490 char header[12]; 458 vpx_codec_pts_t pts;
491 vpx_codec_pts_t pts; 459
492 460 if (pkt->kind != VPX_CODEC_CX_FRAME_PKT)
493 if (pkt->kind != VPX_CODEC_CX_FRAME_PKT) 461 return;
494 return; 462
495 463 pts = pkt->data.frame.pts;
496 pts = pkt->data.frame.pts; 464 mem_put_le32(header, (int)pkt->data.frame.sz);
497 mem_put_le32(header, (int)pkt->data.frame.sz); 465 mem_put_le32(header + 4, pts & 0xFFFFFFFF);
498 mem_put_le32(header + 4, pts & 0xFFFFFFFF); 466 mem_put_le32(header + 8, pts >> 32);
499 mem_put_le32(header + 8, pts >> 32); 467
500 468 (void) fwrite(header, 1, 12, outfile);
501 (void) fwrite(header, 1, 12, outfile); 469 }
502 } 470
503 471 static void write_ivf_frame_size(FILE *outfile, size_t size) {
504 static void write_ivf_frame_size(FILE *outfile, size_t size) 472 char header[4];
505 { 473 mem_put_le32(header, (int)size);
506 char header[4]; 474 (void) fwrite(header, 1, 4, outfile);
507 mem_put_le32(header, (int)size);
508 (void) fwrite(header, 1, 4, outfile);
509 } 475 }
510 476
511 477
512 typedef off_t EbmlLoc; 478 typedef off_t EbmlLoc;
513 479
514 480
515 struct cue_entry 481 struct cue_entry {
516 { 482 unsigned int time;
517 unsigned int time; 483 uint64_t loc;
518 uint64_t loc;
519 }; 484 };
520 485
521 486
522 struct EbmlGlobal 487 struct EbmlGlobal {
523 { 488 int debug;
524 int debug; 489
525 490 FILE *stream;
526 FILE *stream; 491 int64_t last_pts_ms;
527 int64_t last_pts_ms; 492 vpx_rational_t framerate;
528 vpx_rational_t framerate; 493
529 494 /* These pointers are to the start of an element */
530 /* These pointers are to the start of an element */ 495 off_t position_reference;
531 off_t position_reference; 496 off_t seek_info_pos;
532 off_t seek_info_pos; 497 off_t segment_info_pos;
533 off_t segment_info_pos; 498 off_t track_pos;
534 off_t track_pos; 499 off_t cue_pos;
535 off_t cue_pos; 500 off_t cluster_pos;
536 off_t cluster_pos; 501
537 502 /* This pointer is to a specific element to be serialized */
538 /* This pointer is to a specific element to be serialized */ 503 off_t track_id_pos;
539 off_t track_id_pos; 504
540 505 /* These pointers are to the size field of the element */
541 /* These pointers are to the size field of the element */ 506 EbmlLoc startSegment;
542 EbmlLoc startSegment; 507 EbmlLoc startCluster;
543 EbmlLoc startCluster; 508
544 509 uint32_t cluster_timecode;
545 uint32_t cluster_timecode; 510 int cluster_open;
546 int cluster_open; 511
547 512 struct cue_entry *cue_list;
548 struct cue_entry *cue_list; 513 unsigned int cues;
549 unsigned int cues;
550 514
551 }; 515 };
552 516
553 517
554 void Ebml_Write(EbmlGlobal *glob, const void *buffer_in, unsigned long len) 518 void Ebml_Write(EbmlGlobal *glob, const void *buffer_in, unsigned long len) {
555 { 519 (void) fwrite(buffer_in, 1, len, glob->stream);
556 (void) fwrite(buffer_in, 1, len, glob->stream);
557 } 520 }
558 521
559 #define WRITE_BUFFER(s) \ 522 #define WRITE_BUFFER(s) \
560 for(i = len-1; i>=0; i--)\ 523 for(i = len-1; i>=0; i--)\
561 { \ 524 { \
562 x = (char)(*(const s *)buffer_in >> (i * CHAR_BIT)); \ 525 x = (char)(*(const s *)buffer_in >> (i * CHAR_BIT)); \
563 Ebml_Write(glob, &x, 1); \ 526 Ebml_Write(glob, &x, 1); \
564 } 527 }
565 void Ebml_Serialize(EbmlGlobal *glob, const void *buffer_in, int buffer_size, un signed long len) 528 void Ebml_Serialize(EbmlGlobal *glob, const void *buffer_in, int buffer_size, un signed long len) {
566 { 529 char x;
567 char x; 530 int i;
568 int i; 531
569 532 /* buffer_size:
570 /* buffer_size: 533 * 1 - int8_t;
571 * 1 - int8_t; 534 * 2 - int16_t;
572 * 2 - int16_t; 535 * 3 - int32_t;
573 * 3 - int32_t; 536 * 4 - int64_t;
574 * 4 - int64_t; 537 */
575 */ 538 switch (buffer_size) {
576 switch (buffer_size) 539 case 1:
577 { 540 WRITE_BUFFER(int8_t)
578 case 1: 541 break;
579 WRITE_BUFFER(int8_t) 542 case 2:
580 break; 543 WRITE_BUFFER(int16_t)
581 case 2: 544 break;
582 WRITE_BUFFER(int16_t) 545 case 4:
583 break; 546 WRITE_BUFFER(int32_t)
584 case 4: 547 break;
585 WRITE_BUFFER(int32_t) 548 case 8:
586 break; 549 WRITE_BUFFER(int64_t)
587 case 8: 550 break;
588 WRITE_BUFFER(int64_t) 551 default:
589 break; 552 break;
590 default: 553 }
591 break;
592 }
593 } 554 }
594 #undef WRITE_BUFFER 555 #undef WRITE_BUFFER
595 556
596 /* Need a fixed size serializer for the track ID. libmkv provides a 64 bit 557 /* Need a fixed size serializer for the track ID. libmkv provides a 64 bit
597 * one, but not a 32 bit one. 558 * one, but not a 32 bit one.
598 */ 559 */
599 static void Ebml_SerializeUnsigned32(EbmlGlobal *glob, unsigned long class_id, u int64_t ui) 560 static void Ebml_SerializeUnsigned32(EbmlGlobal *glob, unsigned long class_id, u int64_t ui) {
600 { 561 unsigned char sizeSerialized = 4 | 0x80;
601 unsigned char sizeSerialized = 4 | 0x80; 562 Ebml_WriteID(glob, class_id);
602 Ebml_WriteID(glob, class_id); 563 Ebml_Serialize(glob, &sizeSerialized, sizeof(sizeSerialized), 1);
603 Ebml_Serialize(glob, &sizeSerialized, sizeof(sizeSerialized), 1); 564 Ebml_Serialize(glob, &ui, sizeof(ui), 4);
604 Ebml_Serialize(glob, &ui, sizeof(ui), 4);
605 } 565 }
606 566
607 567
608 static void 568 static void
609 Ebml_StartSubElement(EbmlGlobal *glob, EbmlLoc *ebmlLoc, 569 Ebml_StartSubElement(EbmlGlobal *glob, EbmlLoc *ebmlLoc,
610 unsigned long class_id) 570 unsigned long class_id) {
611 { 571 /* todo this is always taking 8 bytes, this may need later optimization */
612 /* todo this is always taking 8 bytes, this may need later optimization */ 572 /* this is a key that says length unknown */
613 /* this is a key that says length unknown */ 573 uint64_t unknownLen = LITERALU64(0x01FFFFFF, 0xFFFFFFFF);
614 uint64_t unknownLen = LITERALU64(0x01FFFFFF, 0xFFFFFFFF); 574
615 575 Ebml_WriteID(glob, class_id);
616 Ebml_WriteID(glob, class_id); 576 *ebmlLoc = ftello(glob->stream);
617 *ebmlLoc = ftello(glob->stream); 577 Ebml_Serialize(glob, &unknownLen, sizeof(unknownLen), 8);
618 Ebml_Serialize(glob, &unknownLen, sizeof(unknownLen), 8);
619 } 578 }
620 579
621 static void 580 static void
622 Ebml_EndSubElement(EbmlGlobal *glob, EbmlLoc *ebmlLoc) 581 Ebml_EndSubElement(EbmlGlobal *glob, EbmlLoc *ebmlLoc) {
623 { 582 off_t pos;
624 off_t pos; 583 uint64_t size;
625 uint64_t size; 584
626 585 /* Save the current stream pointer */
627 /* Save the current stream pointer */ 586 pos = ftello(glob->stream);
628 pos = ftello(glob->stream); 587
629 588 /* Calculate the size of this element */
630 /* Calculate the size of this element */ 589 size = pos - *ebmlLoc - 8;
631 size = pos - *ebmlLoc - 8; 590 size |= LITERALU64(0x01000000, 0x00000000);
632 size |= LITERALU64(0x01000000,0x00000000); 591
633 592 /* Seek back to the beginning of the element and write the new size */
634 /* Seek back to the beginning of the element and write the new size */ 593 fseeko(glob->stream, *ebmlLoc, SEEK_SET);
635 fseeko(glob->stream, *ebmlLoc, SEEK_SET); 594 Ebml_Serialize(glob, &size, sizeof(size), 8);
636 Ebml_Serialize(glob, &size, sizeof(size), 8); 595
637 596 /* Reset the stream pointer */
638 /* Reset the stream pointer */ 597 fseeko(glob->stream, pos, SEEK_SET);
639 fseeko(glob->stream, pos, SEEK_SET);
640 } 598 }
641 599
642 600
643 static void 601 static void
644 write_webm_seek_element(EbmlGlobal *ebml, unsigned long id, off_t pos) 602 write_webm_seek_element(EbmlGlobal *ebml, unsigned long id, off_t pos) {
645 { 603 uint64_t offset = pos - ebml->position_reference;
646 uint64_t offset = pos - ebml->position_reference; 604 EbmlLoc start;
605 Ebml_StartSubElement(ebml, &start, Seek);
606 Ebml_SerializeBinary(ebml, SeekID, id);
607 Ebml_SerializeUnsigned64(ebml, SeekPosition, offset);
608 Ebml_EndSubElement(ebml, &start);
609 }
610
611
612 static void
613 write_webm_seek_info(EbmlGlobal *ebml) {
614
615 off_t pos;
616
617 /* Save the current stream pointer */
618 pos = ftello(ebml->stream);
619
620 if (ebml->seek_info_pos)
621 fseeko(ebml->stream, ebml->seek_info_pos, SEEK_SET);
622 else
623 ebml->seek_info_pos = pos;
624
625 {
647 EbmlLoc start; 626 EbmlLoc start;
648 Ebml_StartSubElement(ebml, &start, Seek); 627
649 Ebml_SerializeBinary(ebml, SeekID, id); 628 Ebml_StartSubElement(ebml, &start, SeekHead);
650 Ebml_SerializeUnsigned64(ebml, SeekPosition, offset); 629 write_webm_seek_element(ebml, Tracks, ebml->track_pos);
630 write_webm_seek_element(ebml, Cues, ebml->cue_pos);
631 write_webm_seek_element(ebml, Info, ebml->segment_info_pos);
651 Ebml_EndSubElement(ebml, &start); 632 Ebml_EndSubElement(ebml, &start);
652 } 633 }
653 634 {
654 635 /* segment info */
655 static void 636 EbmlLoc startInfo;
656 write_webm_seek_info(EbmlGlobal *ebml) 637 uint64_t frame_time;
657 { 638 char version_string[64];
658 639
659 off_t pos; 640 /* Assemble version string */
660 641 if (ebml->debug)
661 /* Save the current stream pointer */ 642 strcpy(version_string, "vpxenc");
662 pos = ftello(ebml->stream); 643 else {
663 644 strcpy(version_string, "vpxenc ");
664 if(ebml->seek_info_pos) 645 strncat(version_string,
665 fseeko(ebml->stream, ebml->seek_info_pos, SEEK_SET); 646 vpx_codec_version_str(),
666 else 647 sizeof(version_string) - 1 - strlen(version_string));
667 ebml->seek_info_pos = pos;
668
669 {
670 EbmlLoc start;
671
672 Ebml_StartSubElement(ebml, &start, SeekHead);
673 write_webm_seek_element(ebml, Tracks, ebml->track_pos);
674 write_webm_seek_element(ebml, Cues, ebml->cue_pos);
675 write_webm_seek_element(ebml, Info, ebml->segment_info_pos);
676 Ebml_EndSubElement(ebml, &start);
677 } 648 }
678 { 649
679 /* segment info */ 650 frame_time = (uint64_t)1000 * ebml->framerate.den
680 EbmlLoc startInfo; 651 / ebml->framerate.num;
681 uint64_t frame_time; 652 ebml->segment_info_pos = ftello(ebml->stream);
682 char version_string[64]; 653 Ebml_StartSubElement(ebml, &startInfo, Info);
683 654 Ebml_SerializeUnsigned(ebml, TimecodeScale, 1000000);
684 /* Assemble version string */ 655 Ebml_SerializeFloat(ebml, Segment_Duration,
685 if(ebml->debug) 656 (double)(ebml->last_pts_ms + frame_time));
686 strcpy(version_string, "vpxenc"); 657 Ebml_SerializeString(ebml, 0x4D80, version_string);
687 else 658 Ebml_SerializeString(ebml, 0x5741, version_string);
688 { 659 Ebml_EndSubElement(ebml, &startInfo);
689 strcpy(version_string, "vpxenc "); 660 }
690 strncat(version_string, 661 }
691 vpx_codec_version_str(), 662
692 sizeof(version_string) - 1 - strlen(version_string)); 663
693 }
694
695 frame_time = (uint64_t)1000 * ebml->framerate.den
696 / ebml->framerate.num;
697 ebml->segment_info_pos = ftello(ebml->stream);
698 Ebml_StartSubElement(ebml, &startInfo, Info);
699 Ebml_SerializeUnsigned(ebml, TimecodeScale, 1000000);
700 Ebml_SerializeFloat(ebml, Segment_Duration,
701 (double)(ebml->last_pts_ms + frame_time));
702 Ebml_SerializeString(ebml, 0x4D80, version_string);
703 Ebml_SerializeString(ebml, 0x5741, version_string);
704 Ebml_EndSubElement(ebml, &startInfo);
705 }
706 }
707
708
709 static void 664 static void
710 write_webm_file_header(EbmlGlobal *glob, 665 write_webm_file_header(EbmlGlobal *glob,
711 const vpx_codec_enc_cfg_t *cfg, 666 const vpx_codec_enc_cfg_t *cfg,
712 const struct vpx_rational *fps, 667 const struct vpx_rational *fps,
713 stereo_format_t stereo_fmt) 668 stereo_format_t stereo_fmt,
714 { 669 unsigned int fourcc) {
670 {
671 EbmlLoc start;
672 Ebml_StartSubElement(glob, &start, EBML);
673 Ebml_SerializeUnsigned(glob, EBMLVersion, 1);
674 Ebml_SerializeUnsigned(glob, EBMLReadVersion, 1);
675 Ebml_SerializeUnsigned(glob, EBMLMaxIDLength, 4);
676 Ebml_SerializeUnsigned(glob, EBMLMaxSizeLength, 8);
677 Ebml_SerializeString(glob, DocType, "webm");
678 Ebml_SerializeUnsigned(glob, DocTypeVersion, 2);
679 Ebml_SerializeUnsigned(glob, DocTypeReadVersion, 2);
680 Ebml_EndSubElement(glob, &start);
681 }
682 {
683 Ebml_StartSubElement(glob, &glob->startSegment, Segment);
684 glob->position_reference = ftello(glob->stream);
685 glob->framerate = *fps;
686 write_webm_seek_info(glob);
687
715 { 688 {
689 EbmlLoc trackStart;
690 glob->track_pos = ftello(glob->stream);
691 Ebml_StartSubElement(glob, &trackStart, Tracks);
692 {
693 unsigned int trackNumber = 1;
694 uint64_t trackID = 0;
695
716 EbmlLoc start; 696 EbmlLoc start;
717 Ebml_StartSubElement(glob, &start, EBML); 697 Ebml_StartSubElement(glob, &start, TrackEntry);
718 Ebml_SerializeUnsigned(glob, EBMLVersion, 1); 698 Ebml_SerializeUnsigned(glob, TrackNumber, trackNumber);
719 Ebml_SerializeUnsigned(glob, EBMLReadVersion, 1); 699 glob->track_id_pos = ftello(glob->stream);
720 Ebml_SerializeUnsigned(glob, EBMLMaxIDLength, 4); 700 Ebml_SerializeUnsigned32(glob, TrackUID, trackID);
721 Ebml_SerializeUnsigned(glob, EBMLMaxSizeLength, 8); 701 Ebml_SerializeUnsigned(glob, TrackType, 1);
722 Ebml_SerializeString(glob, DocType, "webm"); 702 Ebml_SerializeString(glob, CodecID,
723 Ebml_SerializeUnsigned(glob, DocTypeVersion, 2); 703 fourcc == VP8_FOURCC ? "V_VP8" : "V_VP9");
724 Ebml_SerializeUnsigned(glob, DocTypeReadVersion, 2); 704 {
725 Ebml_EndSubElement(glob, &start); 705 unsigned int pixelWidth = cfg->g_w;
706 unsigned int pixelHeight = cfg->g_h;
707 float frameRate = (float)fps->num / (float)fps->den;
708
709 EbmlLoc videoStart;
710 Ebml_StartSubElement(glob, &videoStart, Video);
711 Ebml_SerializeUnsigned(glob, PixelWidth, pixelWidth);
712 Ebml_SerializeUnsigned(glob, PixelHeight, pixelHeight);
713 Ebml_SerializeUnsigned(glob, StereoMode, stereo_fmt);
714 Ebml_SerializeFloat(glob, FrameRate, frameRate);
715 Ebml_EndSubElement(glob, &videoStart);
716 }
717 Ebml_EndSubElement(glob, &start); /* Track Entry */
718 }
719 Ebml_EndSubElement(glob, &trackStart);
726 } 720 }
727 { 721 /* segment element is open */
728 Ebml_StartSubElement(glob, &glob->startSegment, Segment); 722 }
729 glob->position_reference = ftello(glob->stream);
730 glob->framerate = *fps;
731 write_webm_seek_info(glob);
732
733 {
734 EbmlLoc trackStart;
735 glob->track_pos = ftello(glob->stream);
736 Ebml_StartSubElement(glob, &trackStart, Tracks);
737 {
738 unsigned int trackNumber = 1;
739 uint64_t trackID = 0;
740
741 EbmlLoc start;
742 Ebml_StartSubElement(glob, &start, TrackEntry);
743 Ebml_SerializeUnsigned(glob, TrackNumber, trackNumber);
744 glob->track_id_pos = ftello(glob->stream);
745 Ebml_SerializeUnsigned32(glob, TrackUID, trackID);
746 Ebml_SerializeUnsigned(glob, TrackType, 1);
747 Ebml_SerializeString(glob, CodecID, "V_VP8");
748 {
749 unsigned int pixelWidth = cfg->g_w;
750 unsigned int pixelHeight = cfg->g_h;
751 float frameRate = (float)fps->num/(float)fps->den;
752
753 EbmlLoc videoStart;
754 Ebml_StartSubElement(glob, &videoStart, Video);
755 Ebml_SerializeUnsigned(glob, PixelWidth, pixelWidth);
756 Ebml_SerializeUnsigned(glob, PixelHeight, pixelHeight);
757 Ebml_SerializeUnsigned(glob, StereoMode, stereo_fmt);
758 Ebml_SerializeFloat(glob, FrameRate, frameRate);
759 Ebml_EndSubElement(glob, &videoStart);
760 }
761 Ebml_EndSubElement(glob, &start); /* Track Entry */
762 }
763 Ebml_EndSubElement(glob, &trackStart);
764 }
765 /* segment element is open */
766 }
767 } 723 }
768 724
769 725
770 static void 726 static void
771 write_webm_block(EbmlGlobal *glob, 727 write_webm_block(EbmlGlobal *glob,
772 const vpx_codec_enc_cfg_t *cfg, 728 const vpx_codec_enc_cfg_t *cfg,
773 const vpx_codec_cx_pkt_t *pkt) 729 const vpx_codec_cx_pkt_t *pkt) {
774 { 730 unsigned long block_length;
775 unsigned long block_length; 731 unsigned char track_number;
776 unsigned char track_number; 732 unsigned short block_timecode = 0;
777 unsigned short block_timecode = 0; 733 unsigned char flags;
778 unsigned char flags; 734 int64_t pts_ms;
779 int64_t pts_ms; 735 int start_cluster = 0, is_keyframe;
780 int start_cluster = 0, is_keyframe; 736
781 737 /* Calculate the PTS of this frame in milliseconds */
782 /* Calculate the PTS of this frame in milliseconds */ 738 pts_ms = pkt->data.frame.pts * 1000
783 pts_ms = pkt->data.frame.pts * 1000 739 * (uint64_t)cfg->g_timebase.num / (uint64_t)cfg->g_timebase.den;
784 * (uint64_t)cfg->g_timebase.num / (uint64_t)cfg->g_timebase.den; 740 if (pts_ms <= glob->last_pts_ms)
785 if(pts_ms <= glob->last_pts_ms) 741 pts_ms = glob->last_pts_ms + 1;
786 pts_ms = glob->last_pts_ms + 1; 742 glob->last_pts_ms = pts_ms;
787 glob->last_pts_ms = pts_ms; 743
788 744 /* Calculate the relative time of this block */
789 /* Calculate the relative time of this block */ 745 if (pts_ms - glob->cluster_timecode > SHRT_MAX)
790 if(pts_ms - glob->cluster_timecode > SHRT_MAX) 746 start_cluster = 1;
791 start_cluster = 1; 747 else
792 else 748 block_timecode = (unsigned short)pts_ms - glob->cluster_timecode;
793 block_timecode = (unsigned short)pts_ms - glob->cluster_timecode; 749
794 750 is_keyframe = (pkt->data.frame.flags & VPX_FRAME_IS_KEY);
795 is_keyframe = (pkt->data.frame.flags & VPX_FRAME_IS_KEY); 751 if (start_cluster || is_keyframe) {
796 if(start_cluster || is_keyframe) 752 if (glob->cluster_open)
797 { 753 Ebml_EndSubElement(glob, &glob->startCluster);
798 if(glob->cluster_open) 754
799 Ebml_EndSubElement(glob, &glob->startCluster); 755 /* Open the new cluster */
800 756 block_timecode = 0;
801 /* Open the new cluster */ 757 glob->cluster_open = 1;
802 block_timecode = 0; 758 glob->cluster_timecode = (uint32_t)pts_ms;
803 glob->cluster_open = 1; 759 glob->cluster_pos = ftello(glob->stream);
804 glob->cluster_timecode = (uint32_t)pts_ms; 760 Ebml_StartSubElement(glob, &glob->startCluster, Cluster); /* cluster */
805 glob->cluster_pos = ftello(glob->stream); 761 Ebml_SerializeUnsigned(glob, Timecode, glob->cluster_timecode);
806 Ebml_StartSubElement(glob, &glob->startCluster, Cluster); /* cluster */ 762
807 Ebml_SerializeUnsigned(glob, Timecode, glob->cluster_timecode); 763 /* Save a cue point if this is a keyframe. */
808 764 if (is_keyframe) {
809 /* Save a cue point if this is a keyframe. */ 765 struct cue_entry *cue, *new_cue_list;
810 if(is_keyframe) 766
811 { 767 new_cue_list = realloc(glob->cue_list,
812 struct cue_entry *cue, *new_cue_list; 768 (glob->cues + 1) * sizeof(struct cue_entry));
813 769 if (new_cue_list)
814 new_cue_list = realloc(glob->cue_list, 770 glob->cue_list = new_cue_list;
815 (glob->cues+1) * sizeof(struct cue_entry)); 771 else
816 if(new_cue_list) 772 fatal("Failed to realloc cue list.");
817 glob->cue_list = new_cue_list; 773
818 else 774 cue = &glob->cue_list[glob->cues];
819 fatal("Failed to realloc cue list."); 775 cue->time = glob->cluster_timecode;
820 776 cue->loc = glob->cluster_pos;
821 cue = &glob->cue_list[glob->cues]; 777 glob->cues++;
822 cue->time = glob->cluster_timecode;
823 cue->loc = glob->cluster_pos;
824 glob->cues++;
825 }
826 } 778 }
827 779 }
828 /* Write the Simple Block */ 780
829 Ebml_WriteID(glob, SimpleBlock); 781 /* Write the Simple Block */
830 782 Ebml_WriteID(glob, SimpleBlock);
831 block_length = (unsigned long)pkt->data.frame.sz + 4; 783
832 block_length |= 0x10000000; 784 block_length = (unsigned long)pkt->data.frame.sz + 4;
833 Ebml_Serialize(glob, &block_length, sizeof(block_length), 4); 785 block_length |= 0x10000000;
834 786 Ebml_Serialize(glob, &block_length, sizeof(block_length), 4);
835 track_number = 1; 787
836 track_number |= 0x80; 788 track_number = 1;
837 Ebml_Write(glob, &track_number, 1); 789 track_number |= 0x80;
838 790 Ebml_Write(glob, &track_number, 1);
839 Ebml_Serialize(glob, &block_timecode, sizeof(block_timecode), 2); 791
840 792 Ebml_Serialize(glob, &block_timecode, sizeof(block_timecode), 2);
841 flags = 0; 793
842 if(is_keyframe) 794 flags = 0;
843 flags |= 0x80; 795 if (is_keyframe)
844 if(pkt->data.frame.flags & VPX_FRAME_IS_INVISIBLE) 796 flags |= 0x80;
845 flags |= 0x08; 797 if (pkt->data.frame.flags & VPX_FRAME_IS_INVISIBLE)
846 Ebml_Write(glob, &flags, 1); 798 flags |= 0x08;
847 799 Ebml_Write(glob, &flags, 1);
848 Ebml_Write(glob, pkt->data.frame.buf, (unsigned long)pkt->data.frame.sz); 800
801 Ebml_Write(glob, pkt->data.frame.buf, (unsigned long)pkt->data.frame.sz);
849 } 802 }
850 803
851 804
852 static void 805 static void
853 write_webm_file_footer(EbmlGlobal *glob, long hash) 806 write_webm_file_footer(EbmlGlobal *glob, long hash) {
854 { 807
855 808 if (glob->cluster_open)
856 if(glob->cluster_open) 809 Ebml_EndSubElement(glob, &glob->startCluster);
857 Ebml_EndSubElement(glob, &glob->startCluster); 810
858 811 {
859 { 812 EbmlLoc start;
813 unsigned int i;
814
815 glob->cue_pos = ftello(glob->stream);
816 Ebml_StartSubElement(glob, &start, Cues);
817 for (i = 0; i < glob->cues; i++) {
818 struct cue_entry *cue = &glob->cue_list[i];
819 EbmlLoc start;
820
821 Ebml_StartSubElement(glob, &start, CuePoint);
822 {
860 EbmlLoc start; 823 EbmlLoc start;
861 unsigned int i; 824
862 825 Ebml_SerializeUnsigned(glob, CueTime, cue->time);
863 glob->cue_pos = ftello(glob->stream); 826
864 Ebml_StartSubElement(glob, &start, Cues); 827 Ebml_StartSubElement(glob, &start, CueTrackPositions);
865 for(i=0; i<glob->cues; i++) 828 Ebml_SerializeUnsigned(glob, CueTrack, 1);
866 { 829 Ebml_SerializeUnsigned64(glob, CueClusterPosition,
867 struct cue_entry *cue = &glob->cue_list[i]; 830 cue->loc - glob->position_reference);
868 EbmlLoc start;
869
870 Ebml_StartSubElement(glob, &start, CuePoint);
871 {
872 EbmlLoc start;
873
874 Ebml_SerializeUnsigned(glob, CueTime, cue->time);
875
876 Ebml_StartSubElement(glob, &start, CueTrackPositions);
877 Ebml_SerializeUnsigned(glob, CueTrack, 1);
878 Ebml_SerializeUnsigned64(glob, CueClusterPosition,
879 cue->loc - glob->position_reference);
880 Ebml_EndSubElement(glob, &start);
881 }
882 Ebml_EndSubElement(glob, &start);
883 }
884 Ebml_EndSubElement(glob, &start); 831 Ebml_EndSubElement(glob, &start);
832 }
833 Ebml_EndSubElement(glob, &start);
885 } 834 }
886 835 Ebml_EndSubElement(glob, &start);
887 Ebml_EndSubElement(glob, &glob->startSegment); 836 }
888 837
889 /* Patch up the seek info block */ 838 Ebml_EndSubElement(glob, &glob->startSegment);
890 write_webm_seek_info(glob); 839
891 840 /* Patch up the seek info block */
892 /* Patch up the track id */ 841 write_webm_seek_info(glob);
893 fseeko(glob->stream, glob->track_id_pos, SEEK_SET); 842
894 Ebml_SerializeUnsigned32(glob, TrackUID, glob->debug ? 0xDEADBEEF : hash); 843 /* Patch up the track id */
895 844 fseeko(glob->stream, glob->track_id_pos, SEEK_SET);
896 fseeko(glob->stream, 0, SEEK_END); 845 Ebml_SerializeUnsigned32(glob, TrackUID, glob->debug ? 0xDEADBEEF : hash);
846
847 fseeko(glob->stream, 0, SEEK_END);
897 } 848 }
898 849
899 850
900 /* Murmur hash derived from public domain reference implementation at 851 /* Murmur hash derived from public domain reference implementation at
901 * http://sites.google.com/site/murmurhash/ 852 * http:// sites.google.com/site/murmurhash/
902 */ 853 */
903 static unsigned int murmur ( const void * key, int len, unsigned int seed ) 854 static unsigned int murmur(const void *key, int len, unsigned int seed) {
904 { 855 const unsigned int m = 0x5bd1e995;
905 const unsigned int m = 0x5bd1e995; 856 const int r = 24;
906 const int r = 24; 857
907 858 unsigned int h = seed ^ len;
908 unsigned int h = seed ^ len; 859
909 860 const unsigned char *data = (const unsigned char *)key;
910 const unsigned char * data = (const unsigned char *)key; 861
911 862 while (len >= 4) {
912 while(len >= 4) 863 unsigned int k;
913 { 864
914 unsigned int k; 865 k = data[0];
915 866 k |= data[1] << 8;
916 k = data[0]; 867 k |= data[2] << 16;
917 k |= data[1] << 8; 868 k |= data[3] << 24;
918 k |= data[2] << 16; 869
919 k |= data[3] << 24; 870 k *= m;
920 871 k ^= k >> r;
921 k *= m; 872 k *= m;
922 k ^= k >> r; 873
923 k *= m;
924
925 h *= m;
926 h ^= k;
927
928 data += 4;
929 len -= 4;
930 }
931
932 switch(len)
933 {
934 case 3: h ^= data[2] << 16;
935 case 2: h ^= data[1] << 8;
936 case 1: h ^= data[0];
937 h *= m;
938 };
939
940 h ^= h >> 13;
941 h *= m; 874 h *= m;
942 h ^= h >> 15; 875 h ^= k;
943 876
944 return h; 877 data += 4;
878 len -= 4;
879 }
880
881 switch (len) {
882 case 3:
883 h ^= data[2] << 16;
884 case 2:
885 h ^= data[1] << 8;
886 case 1:
887 h ^= data[0];
888 h *= m;
889 };
890
891 h ^= h >> 13;
892 h *= m;
893 h ^= h >> 15;
894
895 return h;
945 } 896 }
946 897
947 #include "math.h" 898 #include "math.h"
948 899 #define MAX_PSNR 100
949 static double vp8_mse2psnr(double Samples, double Peak, double Mse) 900 static double vp8_mse2psnr(double Samples, double Peak, double Mse) {
950 { 901 double psnr;
951 double psnr; 902
952 903 if ((double)Mse > 0.0)
953 if ((double)Mse > 0.0) 904 psnr = 10.0 * log10(Peak * Peak * Samples / Mse);
954 psnr = 10.0 * log10(Peak * Peak * Samples / Mse); 905 else
955 else 906 psnr = MAX_PSNR; /* Limit to prevent / 0 */
956 psnr = 60; /* Limit to prevent / 0 */ 907
957 908 if (psnr > MAX_PSNR)
958 if (psnr > 60) 909 psnr = MAX_PSNR;
959 psnr = 60; 910
960 911 return psnr;
961 return psnr;
962 } 912 }
963 913
964 914
965 #include "args.h" 915 #include "args.h"
966 static const arg_def_t debugmode = ARG_DEF("D", "debug", 0, 916 static const arg_def_t debugmode = ARG_DEF("D", "debug", 0,
967 "Debug mode (makes output deterministic)"); 917 "Debug mode (makes output determinist ic)");
968 static const arg_def_t outputfile = ARG_DEF("o", "output", 1, 918 static const arg_def_t outputfile = ARG_DEF("o", "output", 1,
969 "Output filename"); 919 "Output filename");
970 static const arg_def_t use_yv12 = ARG_DEF(NULL, "yv12", 0, 920 static const arg_def_t use_yv12 = ARG_DEF(NULL, "yv12", 0,
971 "Input file is YV12 "); 921 "Input file is YV12 ");
972 static const arg_def_t use_i420 = ARG_DEF(NULL, "i420", 0, 922 static const arg_def_t use_i420 = ARG_DEF(NULL, "i420", 0,
973 "Input file is I420 (default)"); 923 "Input file is I420 (default)");
974 static const arg_def_t codecarg = ARG_DEF(NULL, "codec", 1, 924 static const arg_def_t codecarg = ARG_DEF(NULL, "codec", 1,
975 "Codec to use"); 925 "Codec to use");
976 static const arg_def_t passes = ARG_DEF("p", "passes", 1, 926 static const arg_def_t passes = ARG_DEF("p", "passes", 1,
977 "Number of passes (1/2)"); 927 "Number of passes (1/2)");
978 static const arg_def_t pass_arg = ARG_DEF(NULL, "pass", 1, 928 static const arg_def_t pass_arg = ARG_DEF(NULL, "pass", 1,
979 "Pass to execute (1/2)"); 929 "Pass to execute (1/2)");
980 static const arg_def_t fpf_name = ARG_DEF(NULL, "fpf", 1, 930 static const arg_def_t fpf_name = ARG_DEF(NULL, "fpf", 1,
981 "First pass statistics file name"); 931 "First pass statistics file na me");
982 static const arg_def_t limit = ARG_DEF(NULL, "limit", 1, 932 static const arg_def_t limit = ARG_DEF(NULL, "limit", 1,
983 "Stop encoding after n input frames"); 933 "Stop encoding after n input frames");
934 static const arg_def_t skip = ARG_DEF(NULL, "skip", 1,
935 "Skip the first n input frames");
984 static const arg_def_t deadline = ARG_DEF("d", "deadline", 1, 936 static const arg_def_t deadline = ARG_DEF("d", "deadline", 1,
985 "Deadline per frame (usec)"); 937 "Deadline per frame (usec)");
986 static const arg_def_t best_dl = ARG_DEF(NULL, "best", 0, 938 static const arg_def_t best_dl = ARG_DEF(NULL, "best", 0,
987 "Use Best Quality Deadline"); 939 "Use Best Quality Deadline");
988 static const arg_def_t good_dl = ARG_DEF(NULL, "good", 0, 940 static const arg_def_t good_dl = ARG_DEF(NULL, "good", 0,
989 "Use Good Quality Deadline"); 941 "Use Good Quality Deadline");
990 static const arg_def_t rt_dl = ARG_DEF(NULL, "rt", 0, 942 static const arg_def_t rt_dl = ARG_DEF(NULL, "rt", 0,
991 "Use Realtime Quality Deadline"); 943 "Use Realtime Quality Deadline ");
992 static const arg_def_t quietarg = ARG_DEF("q", "quiet", 0, 944 static const arg_def_t quietarg = ARG_DEF("q", "quiet", 0,
993 "Do not print encode progress"); 945 "Do not print encode progress" );
994 static const arg_def_t verbosearg = ARG_DEF("v", "verbose", 0, 946 static const arg_def_t verbosearg = ARG_DEF("v", "verbose", 0,
995 "Show encoder parameters"); 947 "Show encoder parameters");
996 static const arg_def_t psnrarg = ARG_DEF(NULL, "psnr", 0, 948 static const arg_def_t psnrarg = ARG_DEF(NULL, "psnr", 0,
997 "Show PSNR in status line"); 949 "Show PSNR in status line");
950 static const arg_def_t recontest = ARG_DEF(NULL, "test-decode", 0,
951 "Test encode/decode mismatch") ;
998 static const arg_def_t framerate = ARG_DEF(NULL, "fps", 1, 952 static const arg_def_t framerate = ARG_DEF(NULL, "fps", 1,
999 "Stream frame rate (rate/scale)"); 953 "Stream frame rate (rate/scale )");
1000 static const arg_def_t use_ivf = ARG_DEF(NULL, "ivf", 0, 954 static const arg_def_t use_ivf = ARG_DEF(NULL, "ivf", 0,
1001 "Output IVF (default is WebM)"); 955 "Output IVF (default is WebM)" );
1002 static const arg_def_t out_part = ARG_DEF("P", "output-partitions", 0, 956 static const arg_def_t out_part = ARG_DEF("P", "output-partitions", 0,
1003 "Makes encoder output partitions. Requires IVF output!"); 957 "Makes encoder output partitions. Requ ires IVF output!");
1004 static const arg_def_t q_hist_n = ARG_DEF(NULL, "q-hist", 1, 958 static const arg_def_t q_hist_n = ARG_DEF(NULL, "q-hist", 1,
1005 "Show quantizer histogram (n-buckets)"); 959 "Show quantizer histogram (n-b uckets)");
1006 static const arg_def_t rate_hist_n = ARG_DEF(NULL, "rate-hist", 1, 960 static const arg_def_t rate_hist_n = ARG_DEF(NULL, "rate-hist", 1,
1007 "Show rate histogram (n-buckets)"); 961 "Show rate histogram (n-buc kets)");
1008 static const arg_def_t *main_args[] = 962 static const arg_def_t *main_args[] = {
1009 { 963 &debugmode,
1010 &debugmode, 964 &outputfile, &codecarg, &passes, &pass_arg, &fpf_name, &limit, &skip,
1011 &outputfile, &codecarg, &passes, &pass_arg, &fpf_name, &limit, &deadline, 965 &deadline, &best_dl, &good_dl, &rt_dl,
1012 &best_dl, &good_dl, &rt_dl, 966 &quietarg, &verbosearg, &psnrarg, &use_ivf, &out_part, &q_hist_n, &rate_hist_n ,
1013 &quietarg, &verbosearg, &psnrarg, &use_ivf, &out_part, &q_hist_n, &rate_hist _n, 967 NULL
1014 NULL
1015 }; 968 };
1016 969
1017 static const arg_def_t usage = ARG_DEF("u", "usage", 1, 970 static const arg_def_t usage = ARG_DEF("u", "usage", 1,
1018 "Usage profile number to use"); 971 "Usage profile number to use") ;
1019 static const arg_def_t threads = ARG_DEF("t", "threads", 1, 972 static const arg_def_t threads = ARG_DEF("t", "threads", 1,
1020 "Max number of threads to use"); 973 "Max number of threads to use" );
1021 static const arg_def_t profile = ARG_DEF(NULL, "profile", 1, 974 static const arg_def_t profile = ARG_DEF(NULL, "profile", 1,
1022 "Bitstream profile number to use"); 975 "Bitstream profile number to u se");
1023 static const arg_def_t width = ARG_DEF("w", "width", 1, 976 static const arg_def_t width = ARG_DEF("w", "width", 1,
1024 "Frame width"); 977 "Frame width");
1025 static const arg_def_t height = ARG_DEF("h", "height", 1, 978 static const arg_def_t height = ARG_DEF("h", "height", 1,
1026 "Frame height"); 979 "Frame height");
1027 static const struct arg_enum_list stereo_mode_enum[] = { 980 static const struct arg_enum_list stereo_mode_enum[] = {
1028 {"mono" , STEREO_FORMAT_MONO}, 981 {"mono", STEREO_FORMAT_MONO},
1029 {"left-right", STEREO_FORMAT_LEFT_RIGHT}, 982 {"left-right", STEREO_FORMAT_LEFT_RIGHT},
1030 {"bottom-top", STEREO_FORMAT_BOTTOM_TOP}, 983 {"bottom-top", STEREO_FORMAT_BOTTOM_TOP},
1031 {"top-bottom", STEREO_FORMAT_TOP_BOTTOM}, 984 {"top-bottom", STEREO_FORMAT_TOP_BOTTOM},
1032 {"right-left", STEREO_FORMAT_RIGHT_LEFT}, 985 {"right-left", STEREO_FORMAT_RIGHT_LEFT},
1033 {NULL, 0} 986 {NULL, 0}
1034 }; 987 };
1035 static const arg_def_t stereo_mode = ARG_DEF_ENUM(NULL, "stereo-mode", 1, 988 static const arg_def_t stereo_mode = ARG_DEF_ENUM(NULL, "stereo-mode", 1,
1036 "Stereo 3D video format", stereo_mode_enum); 989 "Stereo 3D video format", stereo_mode_enum);
1037 static const arg_def_t timebase = ARG_DEF(NULL, "timebase", 1, 990 static const arg_def_t timebase = ARG_DEF(NULL, "timebase", 1,
1038 "Output timestamp precision (fractional seconds)"); 991 "Output timestamp precision (f ractional seconds)");
1039 static const arg_def_t error_resilient = ARG_DEF(NULL, "error-resilient", 1, 992 static const arg_def_t error_resilient = ARG_DEF(NULL, "error-resilient", 1,
1040 "Enable error resiliency features"); 993 "Enable error resiliency featu res");
1041 static const arg_def_t lag_in_frames = ARG_DEF(NULL, "lag-in-frames", 1, 994 static const arg_def_t lag_in_frames = ARG_DEF(NULL, "lag-in-frames", 1,
1042 "Max number of frames to lag"); 995 "Max number of frames to lag") ;
1043 996
1044 static const arg_def_t *global_args[] = 997 static const arg_def_t *global_args[] = {
1045 { 998 &use_yv12, &use_i420, &usage, &threads, &profile,
1046 &use_yv12, &use_i420, &usage, &threads, &profile, 999 &width, &height, &stereo_mode, &timebase, &framerate, &error_resilient,
1047 &width, &height, &stereo_mode, &timebase, &framerate, &error_resilient, 1000 &lag_in_frames, NULL
1048 &lag_in_frames, NULL
1049 }; 1001 };
1050 1002
1051 static const arg_def_t dropframe_thresh = ARG_DEF(NULL, "drop-frame", 1, 1003 static const arg_def_t dropframe_thresh = ARG_DEF(NULL, "drop-frame", 1,
1052 "Temporal resampling threshold (buf %)"); 1004 "Temporal resampling thresho ld (buf %)");
1053 static const arg_def_t resize_allowed = ARG_DEF(NULL, "resize-allowed", 1, 1005 static const arg_def_t resize_allowed = ARG_DEF(NULL, "resize-allowed", 1,
1054 "Spatial resampling enabled (bool)"); 1006 "Spatial resampling enabled (bool)");
1055 static const arg_def_t resize_up_thresh = ARG_DEF(NULL, "resize-up", 1, 1007 static const arg_def_t resize_up_thresh = ARG_DEF(NULL, "resize-up", 1,
1056 "Upscale threshold (buf %)"); 1008 "Upscale threshold (buf %)") ;
1057 static const arg_def_t resize_down_thresh = ARG_DEF(NULL, "resize-down", 1, 1009 static const arg_def_t resize_down_thresh = ARG_DEF(NULL, "resize-down", 1,
1058 "Downscale threshold (buf %)"); 1010 "Downscale threshold (buf %) ");
1059 static const struct arg_enum_list end_usage_enum[] = { 1011 static const struct arg_enum_list end_usage_enum[] = {
1060 {"vbr", VPX_VBR}, 1012 {"vbr", VPX_VBR},
1061 {"cbr", VPX_CBR}, 1013 {"cbr", VPX_CBR},
1062 {"cq", VPX_CQ}, 1014 {"cq", VPX_CQ},
1063 {NULL, 0} 1015 {NULL, 0}
1064 }; 1016 };
1065 static const arg_def_t end_usage = ARG_DEF_ENUM(NULL, "end-usage", 1, 1017 static const arg_def_t end_usage = ARG_DEF_ENUM(NULL, "end-usage", 1,
1066 "Rate control mode", end_usage_enum); 1018 "Rate control mode", en d_usage_enum);
1067 static const arg_def_t target_bitrate = ARG_DEF(NULL, "target-bitrate", 1, 1019 static const arg_def_t target_bitrate = ARG_DEF(NULL, "target-bitrate", 1,
1068 "Bitrate (kbps)"); 1020 "Bitrate (kbps)");
1069 static const arg_def_t min_quantizer = ARG_DEF(NULL, "min-q", 1, 1021 static const arg_def_t min_quantizer = ARG_DEF(NULL, "min-q", 1,
1070 "Minimum (best) quantizer"); 1022 "Minimum (best) quantizer");
1071 static const arg_def_t max_quantizer = ARG_DEF(NULL, "max-q", 1, 1023 static const arg_def_t max_quantizer = ARG_DEF(NULL, "max-q", 1,
1072 "Maximum (worst) quantizer"); 1024 "Maximum (worst) quantizer") ;
1073 static const arg_def_t undershoot_pct = ARG_DEF(NULL, "undershoot-pct", 1, 1025 static const arg_def_t undershoot_pct = ARG_DEF(NULL, "undershoot-pct", 1,
1074 "Datarate undershoot (min) target (%)"); 1026 "Datarate undershoot (min) t arget (%)");
1075 static const arg_def_t overshoot_pct = ARG_DEF(NULL, "overshoot-pct", 1, 1027 static const arg_def_t overshoot_pct = ARG_DEF(NULL, "overshoot-pct", 1,
1076 "Datarate overshoot (max) target (%)"); 1028 "Datarate overshoot (max) ta rget (%)");
1077 static const arg_def_t buf_sz = ARG_DEF(NULL, "buf-sz", 1, 1029 static const arg_def_t buf_sz = ARG_DEF(NULL, "buf-sz", 1,
1078 "Client buffer size (ms)"); 1030 "Client buffer size (ms)");
1079 static const arg_def_t buf_initial_sz = ARG_DEF(NULL, "buf-initial-sz", 1, 1031 static const arg_def_t buf_initial_sz = ARG_DEF(NULL, "buf-initial-sz", 1,
1080 "Client initial buffer size (ms)"); 1032 "Client initial buffer size (ms)");
1081 static const arg_def_t buf_optimal_sz = ARG_DEF(NULL, "buf-optimal-sz", 1, 1033 static const arg_def_t buf_optimal_sz = ARG_DEF(NULL, "buf-optimal-sz", 1,
1082 "Client optimal buffer size (ms)"); 1034 "Client optimal buffer size (ms)");
1083 static const arg_def_t *rc_args[] = 1035 static const arg_def_t *rc_args[] = {
1084 { 1036 &dropframe_thresh, &resize_allowed, &resize_up_thresh, &resize_down_thresh,
1085 &dropframe_thresh, &resize_allowed, &resize_up_thresh, &resize_down_thresh, 1037 &end_usage, &target_bitrate, &min_quantizer, &max_quantizer,
1086 &end_usage, &target_bitrate, &min_quantizer, &max_quantizer, 1038 &undershoot_pct, &overshoot_pct, &buf_sz, &buf_initial_sz, &buf_optimal_sz,
1087 &undershoot_pct, &overshoot_pct, &buf_sz, &buf_initial_sz, &buf_optimal_sz, 1039 NULL
1088 NULL
1089 }; 1040 };
1090 1041
1091 1042
1092 static const arg_def_t bias_pct = ARG_DEF(NULL, "bias-pct", 1, 1043 static const arg_def_t bias_pct = ARG_DEF(NULL, "bias-pct", 1,
1093 "CBR/VBR bias (0=CBR, 100=VBR)"); 1044 "CBR/VBR bias (0=CBR, 100=VBR)");
1094 static const arg_def_t minsection_pct = ARG_DEF(NULL, "minsection-pct", 1, 1045 static const arg_def_t minsection_pct = ARG_DEF(NULL, "minsection-pct", 1,
1095 "GOP min bitrate (% of target)"); 1046 "GOP min bitrate (% of target)") ;
1096 static const arg_def_t maxsection_pct = ARG_DEF(NULL, "maxsection-pct", 1, 1047 static const arg_def_t maxsection_pct = ARG_DEF(NULL, "maxsection-pct", 1,
1097 "GOP max bitrate (% of target)"); 1048 "GOP max bitrate (% of target)") ;
1098 static const arg_def_t *rc_twopass_args[] = 1049 static const arg_def_t *rc_twopass_args[] = {
1099 { 1050 &bias_pct, &minsection_pct, &maxsection_pct, NULL
1100 &bias_pct, &minsection_pct, &maxsection_pct, NULL
1101 }; 1051 };
1102 1052
1103 1053
1104 static const arg_def_t kf_min_dist = ARG_DEF(NULL, "kf-min-dist", 1, 1054 static const arg_def_t kf_min_dist = ARG_DEF(NULL, "kf-min-dist", 1,
1105 "Minimum keyframe interval (frames)"); 1055 "Minimum keyframe interval (frames) ");
1106 static const arg_def_t kf_max_dist = ARG_DEF(NULL, "kf-max-dist", 1, 1056 static const arg_def_t kf_max_dist = ARG_DEF(NULL, "kf-max-dist", 1,
1107 "Maximum keyframe interval (frames)"); 1057 "Maximum keyframe interval (frames) ");
1108 static const arg_def_t kf_disabled = ARG_DEF(NULL, "disable-kf", 0, 1058 static const arg_def_t kf_disabled = ARG_DEF(NULL, "disable-kf", 0,
1109 "Disable keyframe placement"); 1059 "Disable keyframe placement");
1110 static const arg_def_t *kf_args[] = 1060 static const arg_def_t *kf_args[] = {
1111 { 1061 &kf_min_dist, &kf_max_dist, &kf_disabled, NULL
1112 &kf_min_dist, &kf_max_dist, &kf_disabled, NULL
1113 }; 1062 };
1114 1063
1115 1064
1116 #if CONFIG_VP8_ENCODER
1117 static const arg_def_t noise_sens = ARG_DEF(NULL, "noise-sensitivity", 1, 1065 static const arg_def_t noise_sens = ARG_DEF(NULL, "noise-sensitivity", 1,
1118 "Noise sensitivity (frames to blur)"); 1066 "Noise sensitivity (frames to blur)" );
1119 static const arg_def_t sharpness = ARG_DEF(NULL, "sharpness", 1, 1067 static const arg_def_t sharpness = ARG_DEF(NULL, "sharpness", 1,
1120 "Filter sharpness (0-7)"); 1068 "Filter sharpness (0-7)");
1121 static const arg_def_t static_thresh = ARG_DEF(NULL, "static-thresh", 1, 1069 static const arg_def_t static_thresh = ARG_DEF(NULL, "static-thresh", 1,
1122 "Motion detection threshold"); 1070 "Motion detection threshold");
1123 #endif
1124
1125 #if CONFIG_VP8_ENCODER
1126 static const arg_def_t cpu_used = ARG_DEF(NULL, "cpu-used", 1, 1071 static const arg_def_t cpu_used = ARG_DEF(NULL, "cpu-used", 1,
1127 "CPU Used (-16..16)"); 1072 "CPU Used (-16..16)");
1128 #endif
1129
1130
1131 #if CONFIG_VP8_ENCODER
1132 static const arg_def_t token_parts = ARG_DEF(NULL, "token-parts", 1, 1073 static const arg_def_t token_parts = ARG_DEF(NULL, "token-parts", 1,
1133 "Number of token partitions to use, log2"); 1074 "Number of token partitions to use, log2");
1134 static const arg_def_t auto_altref = ARG_DEF(NULL, "auto-alt-ref", 1, 1075 static const arg_def_t auto_altref = ARG_DEF(NULL, "auto-alt-ref", 1,
1135 "Enable automatic alt reference frames"); 1076 "Enable automatic alt reference fra mes");
1136 static const arg_def_t arnr_maxframes = ARG_DEF(NULL, "arnr-maxframes", 1, 1077 static const arg_def_t arnr_maxframes = ARG_DEF(NULL, "arnr-maxframes", 1,
1137 "AltRef Max Frames"); 1078 "AltRef Max Frames");
1138 static const arg_def_t arnr_strength = ARG_DEF(NULL, "arnr-strength", 1, 1079 static const arg_def_t arnr_strength = ARG_DEF(NULL, "arnr-strength", 1,
1139 "AltRef Strength"); 1080 "AltRef Strength");
1140 static const arg_def_t arnr_type = ARG_DEF(NULL, "arnr-type", 1, 1081 static const arg_def_t arnr_type = ARG_DEF(NULL, "arnr-type", 1,
1141 "AltRef Type"); 1082 "AltRef Type");
1142 static const struct arg_enum_list tuning_enum[] = { 1083 static const struct arg_enum_list tuning_enum[] = {
1143 {"psnr", VP8_TUNE_PSNR}, 1084 {"psnr", VP8_TUNE_PSNR},
1144 {"ssim", VP8_TUNE_SSIM}, 1085 {"ssim", VP8_TUNE_SSIM},
1145 {NULL, 0} 1086 {NULL, 0}
1146 }; 1087 };
1147 static const arg_def_t tune_ssim = ARG_DEF_ENUM(NULL, "tune", 1, 1088 static const arg_def_t tune_ssim = ARG_DEF_ENUM(NULL, "tune", 1,
1148 "Material to favor", tuning_enum); 1089 "Material to favor", tuning_enum );
1149 static const arg_def_t cq_level = ARG_DEF(NULL, "cq-level", 1, 1090 static const arg_def_t cq_level = ARG_DEF(NULL, "cq-level", 1,
1150 "Constrained Quality Level"); 1091 "Constrained Quality Level");
1151 static const arg_def_t max_intra_rate_pct = ARG_DEF(NULL, "max-intra-rate", 1, 1092 static const arg_def_t max_intra_rate_pct = ARG_DEF(NULL, "max-intra-rate", 1,
1152 "Max I-frame bitrate (pct)"); 1093 "Max I-frame bitrate (pct)") ;
1153 1094 #if CONFIG_LOSSLESS
1154 static const arg_def_t *vp8_args[] = 1095 static const arg_def_t lossless = ARG_DEF(NULL, "lossless", 1, "Lossless mode");
1155 { 1096 #endif
1156 &cpu_used, &auto_altref, &noise_sens, &sharpness, &static_thresh, 1097
1157 &token_parts, &arnr_maxframes, &arnr_strength, &arnr_type, 1098 #if CONFIG_VP8_ENCODER
1158 &tune_ssim, &cq_level, &max_intra_rate_pct, NULL 1099 static const arg_def_t *vp8_args[] = {
1100 &cpu_used, &auto_altref, &noise_sens, &sharpness, &static_thresh,
1101 &token_parts, &arnr_maxframes, &arnr_strength, &arnr_type,
1102 &tune_ssim, &cq_level, &max_intra_rate_pct,
1103 NULL
1159 }; 1104 };
1160 static const int vp8_arg_ctrl_map[] = 1105 static const int vp8_arg_ctrl_map[] = {
1161 { 1106 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
1162 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF, 1107 VP8E_SET_NOISE_SENSITIVITY, VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
1163 VP8E_SET_NOISE_SENSITIVITY, VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD, 1108 VP8E_SET_TOKEN_PARTITIONS,
1164 VP8E_SET_TOKEN_PARTITIONS, 1109 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
1165 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH , VP8E_SET_ARNR_TYPE, 1110 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
1166 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT, 0 1111 0
1167 }; 1112 };
1168 #endif 1113 #endif
1169 1114
1115 #if CONFIG_VP9_ENCODER
1116 static const arg_def_t *vp9_args[] = {
1117 &cpu_used, &auto_altref, &noise_sens, &sharpness, &static_thresh,
1118 &token_parts, &arnr_maxframes, &arnr_strength, &arnr_type,
1119 &tune_ssim, &cq_level, &max_intra_rate_pct,
1120 #if CONFIG_LOSSLESS
1121 &lossless,
1122 #endif
1123 NULL
1124 };
1125 static const int vp9_arg_ctrl_map[] = {
1126 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
1127 VP8E_SET_NOISE_SENSITIVITY, VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
1128 VP8E_SET_TOKEN_PARTITIONS,
1129 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
1130 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
1131 #if CONFIG_LOSSLESS
1132 VP9E_SET_LOSSLESS,
1133 #endif
1134 0
1135 };
1136 #endif
1137
1170 static const arg_def_t *no_args[] = { NULL }; 1138 static const arg_def_t *no_args[] = { NULL };
1171 1139
1172 static void usage_exit() 1140 static void usage_exit() {
1173 { 1141 int i;
1174 int i; 1142
1175 1143 fprintf(stderr, "Usage: %s <options> -o dst_filename src_filename \n",
1176 fprintf(stderr, "Usage: %s <options> -o dst_filename src_filename \n", 1144 exec_name);
1177 exec_name); 1145
1178 1146 fprintf(stderr, "\nOptions:\n");
1179 fprintf(stderr, "\nOptions:\n"); 1147 arg_show_usage(stdout, main_args);
1180 arg_show_usage(stdout, main_args); 1148 fprintf(stderr, "\nEncoder Global Options:\n");
1181 fprintf(stderr, "\nEncoder Global Options:\n"); 1149 arg_show_usage(stdout, global_args);
1182 arg_show_usage(stdout, global_args); 1150 fprintf(stderr, "\nRate Control Options:\n");
1183 fprintf(stderr, "\nRate Control Options:\n"); 1151 arg_show_usage(stdout, rc_args);
1184 arg_show_usage(stdout, rc_args); 1152 fprintf(stderr, "\nTwopass Rate Control Options:\n");
1185 fprintf(stderr, "\nTwopass Rate Control Options:\n"); 1153 arg_show_usage(stdout, rc_twopass_args);
1186 arg_show_usage(stdout, rc_twopass_args); 1154 fprintf(stderr, "\nKeyframe Placement Options:\n");
1187 fprintf(stderr, "\nKeyframe Placement Options:\n"); 1155 arg_show_usage(stdout, kf_args);
1188 arg_show_usage(stdout, kf_args);
1189 #if CONFIG_VP8_ENCODER 1156 #if CONFIG_VP8_ENCODER
1190 fprintf(stderr, "\nVP8 Specific Options:\n"); 1157 fprintf(stderr, "\nVP8 Specific Options:\n");
1191 arg_show_usage(stdout, vp8_args); 1158 arg_show_usage(stdout, vp8_args);
1192 #endif 1159 #endif
1193 fprintf(stderr, "\nStream timebase (--timebase):\n" 1160 #if CONFIG_VP9_ENCODER
1194 " The desired precision of timestamps in the output, expressed\n" 1161 fprintf(stderr, "\nVP9 Specific Options:\n");
1195 " in fractional seconds. Default is 1/1000.\n"); 1162 arg_show_usage(stdout, vp9_args);
1196 fprintf(stderr, "\n" 1163 #endif
1197 "Included encoders:\n" 1164 fprintf(stderr, "\nStream timebase (--timebase):\n"
1198 "\n"); 1165 " The desired precision of timestamps in the output, expressed\n"
1199 1166 " in fractional seconds. Default is 1/1000.\n");
1200 for (i = 0; i < sizeof(codecs) / sizeof(codecs[0]); i++) 1167 fprintf(stderr, "\n"
1201 fprintf(stderr, " %-6s - %s\n", 1168 "Included encoders:\n"
1202 codecs[i].name, 1169 "\n");
1203 vpx_codec_iface_name(codecs[i].iface)); 1170
1204 1171 for (i = 0; i < sizeof(codecs) / sizeof(codecs[0]); i++)
1205 exit(EXIT_FAILURE); 1172 fprintf(stderr, " %-6s - %s\n",
1173 codecs[i].name,
1174 vpx_codec_iface_name(codecs[i].iface()));
1175
1176 exit(EXIT_FAILURE);
1206 } 1177 }
1207 1178
1208 1179
1209 #define HIST_BAR_MAX 40 1180 #define HIST_BAR_MAX 40
1210 struct hist_bucket 1181 struct hist_bucket {
1211 { 1182 int low, high, count;
1212 int low, high, count;
1213 }; 1183 };
1214 1184
1215 1185
1216 static int merge_hist_buckets(struct hist_bucket *bucket, 1186 static int merge_hist_buckets(struct hist_bucket *bucket,
1217 int *buckets_, 1187 int *buckets_,
1218 int max_buckets) 1188 int max_buckets) {
1219 { 1189 int small_bucket = 0, merge_bucket = INT_MAX, big_bucket = 0;
1220 int small_bucket = 0, merge_bucket = INT_MAX, big_bucket=0; 1190 int buckets = *buckets_;
1221 int buckets = *buckets_; 1191 int i;
1222 int i; 1192
1223 1193 /* Find the extrema for this list of buckets */
1224 /* Find the extrema for this list of buckets */ 1194 big_bucket = small_bucket = 0;
1195 for (i = 0; i < buckets; i++) {
1196 if (bucket[i].count < bucket[small_bucket].count)
1197 small_bucket = i;
1198 if (bucket[i].count > bucket[big_bucket].count)
1199 big_bucket = i;
1200 }
1201
1202 /* If we have too many buckets, merge the smallest with an adjacent
1203 * bucket.
1204 */
1205 while (buckets > max_buckets) {
1206 int last_bucket = buckets - 1;
1207
1208 /* merge the small bucket with an adjacent one. */
1209 if (small_bucket == 0)
1210 merge_bucket = 1;
1211 else if (small_bucket == last_bucket)
1212 merge_bucket = last_bucket - 1;
1213 else if (bucket[small_bucket - 1].count < bucket[small_bucket + 1].count)
1214 merge_bucket = small_bucket - 1;
1215 else
1216 merge_bucket = small_bucket + 1;
1217
1218 assert(abs(merge_bucket - small_bucket) <= 1);
1219 assert(small_bucket < buckets);
1220 assert(big_bucket < buckets);
1221 assert(merge_bucket < buckets);
1222
1223 if (merge_bucket < small_bucket) {
1224 bucket[merge_bucket].high = bucket[small_bucket].high;
1225 bucket[merge_bucket].count += bucket[small_bucket].count;
1226 } else {
1227 bucket[small_bucket].high = bucket[merge_bucket].high;
1228 bucket[small_bucket].count += bucket[merge_bucket].count;
1229 merge_bucket = small_bucket;
1230 }
1231
1232 assert(bucket[merge_bucket].low != bucket[merge_bucket].high);
1233
1234 buckets--;
1235
1236 /* Remove the merge_bucket from the list, and find the new small
1237 * and big buckets while we're at it
1238 */
1225 big_bucket = small_bucket = 0; 1239 big_bucket = small_bucket = 0;
1226 for(i=0; i < buckets; i++) 1240 for (i = 0; i < buckets; i++) {
1227 { 1241 if (i > merge_bucket)
1228 if(bucket[i].count < bucket[small_bucket].count) 1242 bucket[i] = bucket[i + 1];
1229 small_bucket = i; 1243
1230 if(bucket[i].count > bucket[big_bucket].count) 1244 if (bucket[i].count < bucket[small_bucket].count)
1231 big_bucket = i; 1245 small_bucket = i;
1246 if (bucket[i].count > bucket[big_bucket].count)
1247 big_bucket = i;
1232 } 1248 }
1233 1249
1234 /* If we have too many buckets, merge the smallest with an adjacent 1250 }
1235 * bucket. 1251
1236 */ 1252 *buckets_ = buckets;
1237 while(buckets > max_buckets) 1253 return bucket[big_bucket].count;
1238 {
1239 int last_bucket = buckets - 1;
1240
1241 /* merge the small bucket with an adjacent one. */
1242 if(small_bucket == 0)
1243 merge_bucket = 1;
1244 else if(small_bucket == last_bucket)
1245 merge_bucket = last_bucket - 1;
1246 else if(bucket[small_bucket - 1].count < bucket[small_bucket + 1].count)
1247 merge_bucket = small_bucket - 1;
1248 else
1249 merge_bucket = small_bucket + 1;
1250
1251 assert(abs(merge_bucket - small_bucket) <= 1);
1252 assert(small_bucket < buckets);
1253 assert(big_bucket < buckets);
1254 assert(merge_bucket < buckets);
1255
1256 if(merge_bucket < small_bucket)
1257 {
1258 bucket[merge_bucket].high = bucket[small_bucket].high;
1259 bucket[merge_bucket].count += bucket[small_bucket].count;
1260 }
1261 else
1262 {
1263 bucket[small_bucket].high = bucket[merge_bucket].high;
1264 bucket[small_bucket].count += bucket[merge_bucket].count;
1265 merge_bucket = small_bucket;
1266 }
1267
1268 assert(bucket[merge_bucket].low != bucket[merge_bucket].high);
1269
1270 buckets--;
1271
1272 /* Remove the merge_bucket from the list, and find the new small
1273 * and big buckets while we're at it
1274 */
1275 big_bucket = small_bucket = 0;
1276 for(i=0; i < buckets; i++)
1277 {
1278 if(i > merge_bucket)
1279 bucket[i] = bucket[i+1];
1280
1281 if(bucket[i].count < bucket[small_bucket].count)
1282 small_bucket = i;
1283 if(bucket[i].count > bucket[big_bucket].count)
1284 big_bucket = i;
1285 }
1286
1287 }
1288
1289 *buckets_ = buckets;
1290 return bucket[big_bucket].count;
1291 } 1254 }
1292 1255
1293 1256
1294 static void show_histogram(const struct hist_bucket *bucket, 1257 static void show_histogram(const struct hist_bucket *bucket,
1295 int buckets, 1258 int buckets,
1296 int total, 1259 int total,
1297 int scale) 1260 int scale) {
1298 { 1261 const char *pat1, *pat2;
1299 const char *pat1, *pat2; 1262 int i;
1300 int i; 1263
1301 1264 switch ((int)(log(bucket[buckets - 1].high) / log(10)) + 1) {
1302 switch((int)(log(bucket[buckets-1].high)/log(10))+1) 1265 case 1:
1303 { 1266 case 2:
1304 case 1: 1267 pat1 = "%4d %2s: ";
1305 case 2: 1268 pat2 = "%4d-%2d: ";
1306 pat1 = "%4d %2s: "; 1269 break;
1307 pat2 = "%4d-%2d: "; 1270 case 3:
1271 pat1 = "%5d %3s: ";
1272 pat2 = "%5d-%3d: ";
1273 break;
1274 case 4:
1275 pat1 = "%6d %4s: ";
1276 pat2 = "%6d-%4d: ";
1277 break;
1278 case 5:
1279 pat1 = "%7d %5s: ";
1280 pat2 = "%7d-%5d: ";
1281 break;
1282 case 6:
1283 pat1 = "%8d %6s: ";
1284 pat2 = "%8d-%6d: ";
1285 break;
1286 case 7:
1287 pat1 = "%9d %7s: ";
1288 pat2 = "%9d-%7d: ";
1289 break;
1290 default:
1291 pat1 = "%12d %10s: ";
1292 pat2 = "%12d-%10d: ";
1293 break;
1294 }
1295
1296 for (i = 0; i < buckets; i++) {
1297 int len;
1298 int j;
1299 float pct;
1300
1301 pct = (float)(100.0 * bucket[i].count / total);
1302 len = HIST_BAR_MAX * bucket[i].count / scale;
1303 if (len < 1)
1304 len = 1;
1305 assert(len <= HIST_BAR_MAX);
1306
1307 if (bucket[i].low == bucket[i].high)
1308 fprintf(stderr, pat1, bucket[i].low, "");
1309 else
1310 fprintf(stderr, pat2, bucket[i].low, bucket[i].high);
1311
1312 for (j = 0; j < HIST_BAR_MAX; j++)
1313 fprintf(stderr, j < len ? "=" : " ");
1314 fprintf(stderr, "\t%5d (%6.2f%%)\n", bucket[i].count, pct);
1315 }
1316 }
1317
1318
1319 static void show_q_histogram(const int counts[64], int max_buckets) {
1320 struct hist_bucket bucket[64];
1321 int buckets = 0;
1322 int total = 0;
1323 int scale;
1324 int i;
1325
1326
1327 for (i = 0; i < 64; i++) {
1328 if (counts[i]) {
1329 bucket[buckets].low = bucket[buckets].high = i;
1330 bucket[buckets].count = counts[i];
1331 buckets++;
1332 total += counts[i];
1333 }
1334 }
1335
1336 fprintf(stderr, "\nQuantizer Selection:\n");
1337 scale = merge_hist_buckets(bucket, &buckets, max_buckets);
1338 show_histogram(bucket, buckets, total, scale);
1339 }
1340
1341
1342 #define RATE_BINS (100)
1343 struct rate_hist {
1344 int64_t *pts;
1345 int *sz;
1346 int samples;
1347 int frames;
1348 struct hist_bucket bucket[RATE_BINS];
1349 int total;
1350 };
1351
1352
1353 static void init_rate_histogram(struct rate_hist *hist,
1354 const vpx_codec_enc_cfg_t *cfg,
1355 const vpx_rational_t *fps) {
1356 int i;
1357
1358 /* Determine the number of samples in the buffer. Use the file's framerate
1359 * to determine the number of frames in rc_buf_sz milliseconds, with an
1360 * adjustment (5/4) to account for alt-refs
1361 */
1362 hist->samples = cfg->rc_buf_sz * 5 / 4 * fps->num / fps->den / 1000;
1363
1364 /* prevent division by zero */
1365 if (hist->samples == 0)
1366 hist->samples = 1;
1367
1368 hist->pts = calloc(hist->samples, sizeof(*hist->pts));
1369 hist->sz = calloc(hist->samples, sizeof(*hist->sz));
1370 for (i = 0; i < RATE_BINS; i++) {
1371 hist->bucket[i].low = INT_MAX;
1372 hist->bucket[i].high = 0;
1373 hist->bucket[i].count = 0;
1374 }
1375 }
1376
1377
1378 static void destroy_rate_histogram(struct rate_hist *hist) {
1379 free(hist->pts);
1380 free(hist->sz);
1381 }
1382
1383
1384 static void update_rate_histogram(struct rate_hist *hist,
1385 const vpx_codec_enc_cfg_t *cfg,
1386 const vpx_codec_cx_pkt_t *pkt) {
1387 int i, idx;
1388 int64_t now, then, sum_sz = 0, avg_bitrate;
1389
1390 now = pkt->data.frame.pts * 1000
1391 * (uint64_t)cfg->g_timebase.num / (uint64_t)cfg->g_timebase.den;
1392
1393 idx = hist->frames++ % hist->samples;
1394 hist->pts[idx] = now;
1395 hist->sz[idx] = (int)pkt->data.frame.sz;
1396
1397 if (now < cfg->rc_buf_initial_sz)
1398 return;
1399
1400 then = now;
1401
1402 /* Sum the size over the past rc_buf_sz ms */
1403 for (i = hist->frames; i > 0 && hist->frames - i < hist->samples; i--) {
1404 int i_idx = (i - 1) % hist->samples;
1405
1406 then = hist->pts[i_idx];
1407 if (now - then > cfg->rc_buf_sz)
1408 break;
1409 sum_sz += hist->sz[i_idx];
1410 }
1411
1412 if (now == then)
1413 return;
1414
1415 avg_bitrate = sum_sz * 8 * 1000 / (now - then);
1416 idx = (int)(avg_bitrate * (RATE_BINS / 2) / (cfg->rc_target_bitrate * 1000));
1417 if (idx < 0)
1418 idx = 0;
1419 if (idx > RATE_BINS - 1)
1420 idx = RATE_BINS - 1;
1421 if (hist->bucket[idx].low > avg_bitrate)
1422 hist->bucket[idx].low = (int)avg_bitrate;
1423 if (hist->bucket[idx].high < avg_bitrate)
1424 hist->bucket[idx].high = (int)avg_bitrate;
1425 hist->bucket[idx].count++;
1426 hist->total++;
1427 }
1428
1429
1430 static void show_rate_histogram(struct rate_hist *hist,
1431 const vpx_codec_enc_cfg_t *cfg,
1432 int max_buckets) {
1433 int i, scale;
1434 int buckets = 0;
1435
1436 for (i = 0; i < RATE_BINS; i++) {
1437 if (hist->bucket[i].low == INT_MAX)
1438 continue;
1439 hist->bucket[buckets++] = hist->bucket[i];
1440 }
1441
1442 fprintf(stderr, "\nRate (over %dms window):\n", cfg->rc_buf_sz);
1443 scale = merge_hist_buckets(hist->bucket, &buckets, max_buckets);
1444 show_histogram(hist->bucket, buckets, hist->total, scale);
1445 }
1446
1447 #define mmin(a, b) ((a) < (b) ? (a) : (b))
1448 static void find_mismatch(vpx_image_t *img1, vpx_image_t *img2,
1449 int yloc[2], int uloc[2], int vloc[2]) {
1450 int match = 1;
1451 int i, j;
1452 yloc[0] = yloc[1] = -1;
1453 for (i = 0, match = 1; match && i < img1->d_h; i+=32) {
1454 for (j = 0; match && j < img1->d_w; j+=32) {
1455 int k, l;
1456 int si = mmin(i + 32, img1->d_h) - i;
1457 int sj = mmin(j + 32, img1->d_w) - j;
1458 for (k = 0; match && k < si; k++)
1459 for (l = 0; match && l < sj; l++) {
1460 if (*(img1->planes[VPX_PLANE_Y] +
1461 (i + k) * img1->stride[VPX_PLANE_Y] + j + l) !=
1462 *(img2->planes[VPX_PLANE_Y] +
1463 (i + k) * img2->stride[VPX_PLANE_Y] + j + l)) {
1464 yloc[0] = i + k;
1465 yloc[1] = j + l;
1466 match = 0;
1308 break; 1467 break;
1309 case 3: 1468 }
1310 pat1 = "%5d %3s: ";
1311 pat2 = "%5d-%3d: ";
1312 break;
1313 case 4:
1314 pat1 = "%6d %4s: ";
1315 pat2 = "%6d-%4d: ";
1316 break;
1317 case 5:
1318 pat1 = "%7d %5s: ";
1319 pat2 = "%7d-%5d: ";
1320 break;
1321 case 6:
1322 pat1 = "%8d %6s: ";
1323 pat2 = "%8d-%6d: ";
1324 break;
1325 case 7:
1326 pat1 = "%9d %7s: ";
1327 pat2 = "%9d-%7d: ";
1328 break;
1329 default:
1330 pat1 = "%12d %10s: ";
1331 pat2 = "%12d-%10d: ";
1332 break;
1333 }
1334
1335 for(i=0; i<buckets; i++)
1336 {
1337 int len;
1338 int j;
1339 float pct;
1340
1341 pct = (float)(100.0 * bucket[i].count / total);
1342 len = HIST_BAR_MAX * bucket[i].count / scale;
1343 if(len < 1)
1344 len = 1;
1345 assert(len <= HIST_BAR_MAX);
1346
1347 if(bucket[i].low == bucket[i].high)
1348 fprintf(stderr, pat1, bucket[i].low, "");
1349 else
1350 fprintf(stderr, pat2, bucket[i].low, bucket[i].high);
1351
1352 for(j=0; j<HIST_BAR_MAX; j++)
1353 fprintf(stderr, j<len?"=":" ");
1354 fprintf(stderr, "\t%5d (%6.2f%%)\n",bucket[i].count,pct);
1355 }
1356 }
1357
1358
1359 static void show_q_histogram(const int counts[64], int max_buckets)
1360 {
1361 struct hist_bucket bucket[64];
1362 int buckets = 0;
1363 int total = 0;
1364 int scale;
1365 int i;
1366
1367
1368 for(i=0; i<64; i++)
1369 {
1370 if(counts[i])
1371 {
1372 bucket[buckets].low = bucket[buckets].high = i;
1373 bucket[buckets].count = counts[i];
1374 buckets++;
1375 total += counts[i];
1376 } 1469 }
1377 } 1470 }
1378 1471 }
1379 fprintf(stderr, "\nQuantizer Selection:\n"); 1472 uloc[0] = uloc[1] = -1;
1380 scale = merge_hist_buckets(bucket, &buckets, max_buckets); 1473 for (i = 0, match = 1; match && i < (img1->d_h + 1) / 2; i+=16) {
1381 show_histogram(bucket, buckets, total, scale); 1474 for (j = 0; j < match && (img1->d_w + 1) / 2; j+=16) {
1382 } 1475 int k, l;
1383 1476 int si = mmin(i + 16, (img1->d_h + 1) / 2) - i;
1384 1477 int sj = mmin(j + 16, (img1->d_w + 1) / 2) - j;
1385 #define RATE_BINS (100) 1478 for (k = 0; match && k < si; k++)
1386 struct rate_hist 1479 for (l = 0; match && l < sj; l++) {
1387 { 1480 if (*(img1->planes[VPX_PLANE_U] +
1388 int64_t *pts; 1481 (i + k) * img1->stride[VPX_PLANE_U] + j + l) !=
1389 int *sz; 1482 *(img2->planes[VPX_PLANE_U] +
1390 int samples; 1483 (i + k) * img2->stride[VPX_PLANE_U] + j + l)) {
1391 int frames; 1484 uloc[0] = i + k;
1392 struct hist_bucket bucket[RATE_BINS]; 1485 uloc[1] = j + l;
1393 int total; 1486 match = 0;
1394 };
1395
1396
1397 static void init_rate_histogram(struct rate_hist *hist,
1398 const vpx_codec_enc_cfg_t *cfg,
1399 const vpx_rational_t *fps)
1400 {
1401 int i;
1402
1403 /* Determine the number of samples in the buffer. Use the file's framerate
1404 * to determine the number of frames in rc_buf_sz milliseconds, with an
1405 * adjustment (5/4) to account for alt-refs
1406 */
1407 hist->samples = cfg->rc_buf_sz * 5 / 4 * fps->num / fps->den / 1000;
1408
1409 /* prevent division by zero */
1410 if (hist->samples == 0)
1411 hist->samples=1;
1412
1413 hist->pts = calloc(hist->samples, sizeof(*hist->pts));
1414 hist->sz = calloc(hist->samples, sizeof(*hist->sz));
1415 for(i=0; i<RATE_BINS; i++)
1416 {
1417 hist->bucket[i].low = INT_MAX;
1418 hist->bucket[i].high = 0;
1419 hist->bucket[i].count = 0;
1420 }
1421 }
1422
1423
1424 static void destroy_rate_histogram(struct rate_hist *hist)
1425 {
1426 free(hist->pts);
1427 free(hist->sz);
1428 }
1429
1430
1431 static void update_rate_histogram(struct rate_hist *hist,
1432 const vpx_codec_enc_cfg_t *cfg,
1433 const vpx_codec_cx_pkt_t *pkt)
1434 {
1435 int i, idx;
1436 int64_t now, then, sum_sz = 0, avg_bitrate;
1437
1438 now = pkt->data.frame.pts * 1000
1439 * (uint64_t)cfg->g_timebase.num / (uint64_t)cfg->g_timebase.den;
1440
1441 idx = hist->frames++ % hist->samples;
1442 hist->pts[idx] = now;
1443 hist->sz[idx] = (int)pkt->data.frame.sz;
1444
1445 if(now < cfg->rc_buf_initial_sz)
1446 return;
1447
1448 then = now;
1449
1450 /* Sum the size over the past rc_buf_sz ms */
1451 for(i = hist->frames; i > 0 && hist->frames - i < hist->samples; i--)
1452 {
1453 int i_idx = (i-1) % hist->samples;
1454
1455 then = hist->pts[i_idx];
1456 if(now - then > cfg->rc_buf_sz)
1457 break; 1487 break;
1458 sum_sz += hist->sz[i_idx]; 1488 }
1459 }
1460
1461 if (now == then)
1462 return;
1463
1464 avg_bitrate = sum_sz * 8 * 1000 / (now - then);
1465 idx = (int)(avg_bitrate * (RATE_BINS/2) / (cfg->rc_target_bitrate * 1000));
1466 if(idx < 0)
1467 idx = 0;
1468 if(idx > RATE_BINS-1)
1469 idx = RATE_BINS-1;
1470 if(hist->bucket[idx].low > avg_bitrate)
1471 hist->bucket[idx].low = (int)avg_bitrate;
1472 if(hist->bucket[idx].high < avg_bitrate)
1473 hist->bucket[idx].high = (int)avg_bitrate;
1474 hist->bucket[idx].count++;
1475 hist->total++;
1476 }
1477
1478
1479 static void show_rate_histogram(struct rate_hist *hist,
1480 const vpx_codec_enc_cfg_t *cfg,
1481 int max_buckets)
1482 {
1483 int i, scale;
1484 int buckets = 0;
1485
1486 for(i = 0; i < RATE_BINS; i++)
1487 {
1488 if(hist->bucket[i].low == INT_MAX)
1489 continue;
1490 hist->bucket[buckets++] = hist->bucket[i];
1491 }
1492
1493 fprintf(stderr, "\nRate (over %dms window):\n", cfg->rc_buf_sz);
1494 scale = merge_hist_buckets(hist->bucket, &buckets, max_buckets);
1495 show_histogram(hist->bucket, buckets, hist->total, scale);
1496 }
1497
1498 #define NELEMENTS(x) (sizeof(x)/sizeof(x[0]))
1499 #define ARG_CTRL_CNT_MAX NELEMENTS(vp8_arg_ctrl_map)
1500
1501
1502 /* Configuration elements common to all streams */
1503 struct global_config
1504 {
1505 const struct codec_item *codec;
1506 int passes;
1507 int pass;
1508 int usage;
1509 int deadline;
1510 int use_i420;
1511 int quiet;
1512 int verbose;
1513 int limit;
1514 int show_psnr;
1515 int have_framerate;
1516 struct vpx_rational framerate;
1517 int out_part;
1518 int debug;
1519 int show_q_hist_buckets;
1520 int show_rate_hist_buckets;
1521 };
1522
1523
1524 /* Per-stream configuration */
1525 struct stream_config
1526 {
1527 struct vpx_codec_enc_cfg cfg;
1528 const char *out_fn;
1529 const char *stats_fn;
1530 stereo_format_t stereo_fmt;
1531 int arg_ctrls[ARG_CTRL_CNT_MAX][2];
1532 int arg_ctrl_cnt;
1533 int write_webm;
1534 int have_kf_max_dist;
1535 };
1536
1537
1538 struct stream_state
1539 {
1540 int index;
1541 struct stream_state *next;
1542 struct stream_config config;
1543 FILE *file;
1544 struct rate_hist rate_hist;
1545 EbmlGlobal ebml;
1546 uint32_t hash;
1547 uint64_t psnr_sse_total;
1548 uint64_t psnr_samples_total;
1549 double psnr_totals[4];
1550 int psnr_count;
1551 int counts[64];
1552 vpx_codec_ctx_t encoder;
1553 unsigned int frames_out;
1554 uint64_t cx_time;
1555 size_t nbytes;
1556 stats_io_t stats;
1557 };
1558
1559
1560 void validate_positive_rational(const char *msg,
1561 struct vpx_rational *rat)
1562 {
1563 if (rat->den < 0)
1564 {
1565 rat->num *= -1;
1566 rat->den *= -1;
1567 }
1568
1569 if (rat->num < 0)
1570 die("Error: %s must be positive\n", msg);
1571
1572 if (!rat->den)
1573 die("Error: %s has zero denominator\n", msg);
1574 }
1575
1576
1577 static void parse_global_config(struct global_config *global, char **argv)
1578 {
1579 char **argi, **argj;
1580 struct arg arg;
1581
1582 /* Initialize default parameters */
1583 memset(global, 0, sizeof(*global));
1584 global->codec = codecs;
1585 global->passes = 1;
1586 global->use_i420 = 1;
1587
1588 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step)
1589 {
1590 arg.argv_step = 1;
1591
1592 if (arg_match(&arg, &codecarg, argi))
1593 {
1594 int j, k = -1;
1595
1596 for (j = 0; j < sizeof(codecs) / sizeof(codecs[0]); j++)
1597 if (!strcmp(codecs[j].name, arg.val))
1598 k = j;
1599
1600 if (k >= 0)
1601 global->codec = codecs + k;
1602 else
1603 die("Error: Unrecognized argument (%s) to --codec\n",
1604 arg.val);
1605
1606 }
1607 else if (arg_match(&arg, &passes, argi))
1608 {
1609 global->passes = arg_parse_uint(&arg);
1610
1611 if (global->passes < 1 || global->passes > 2)
1612 die("Error: Invalid number of passes (%d)\n", global->passes);
1613 }
1614 else if (arg_match(&arg, &pass_arg, argi))
1615 {
1616 global->pass = arg_parse_uint(&arg);
1617
1618 if (global->pass < 1 || global->pass > 2)
1619 die("Error: Invalid pass selected (%d)\n",
1620 global->pass);
1621 }
1622 else if (arg_match(&arg, &usage, argi))
1623 global->usage = arg_parse_uint(&arg);
1624 else if (arg_match(&arg, &deadline, argi))
1625 global->deadline = arg_parse_uint(&arg);
1626 else if (arg_match(&arg, &best_dl, argi))
1627 global->deadline = VPX_DL_BEST_QUALITY;
1628 else if (arg_match(&arg, &good_dl, argi))
1629 global->deadline = VPX_DL_GOOD_QUALITY;
1630 else if (arg_match(&arg, &rt_dl, argi))
1631 global->deadline = VPX_DL_REALTIME;
1632 else if (arg_match(&arg, &use_yv12, argi))
1633 global->use_i420 = 0;
1634 else if (arg_match(&arg, &use_i420, argi))
1635 global->use_i420 = 1;
1636 else if (arg_match(&arg, &quietarg, argi))
1637 global->quiet = 1;
1638 else if (arg_match(&arg, &verbosearg, argi))
1639 global->verbose = 1;
1640 else if (arg_match(&arg, &limit, argi))
1641 global->limit = arg_parse_uint(&arg);
1642 else if (arg_match(&arg, &psnrarg, argi))
1643 global->show_psnr = 1;
1644 else if (arg_match(&arg, &framerate, argi))
1645 {
1646 global->framerate = arg_parse_rational(&arg);
1647 validate_positive_rational(arg.name, &global->framerate);
1648 global->have_framerate = 1;
1649 }
1650 else if (arg_match(&arg,&out_part, argi))
1651 global->out_part = 1;
1652 else if (arg_match(&arg, &debugmode, argi))
1653 global->debug = 1;
1654 else if (arg_match(&arg, &q_hist_n, argi))
1655 global->show_q_hist_buckets = arg_parse_uint(&arg);
1656 else if (arg_match(&arg, &rate_hist_n, argi))
1657 global->show_rate_hist_buckets = arg_parse_uint(&arg);
1658 else
1659 argj++;
1660 }
1661
1662 /* Validate global config */
1663
1664 if (global->pass)
1665 {
1666 /* DWIM: Assume the user meant passes=2 if pass=2 is specified */
1667 if (global->pass > global->passes)
1668 {
1669 warn("Assuming --pass=%d implies --passes=%d\n",
1670 global->pass, global->pass);
1671 global->passes = global->pass;
1672 } 1489 }
1673 } 1490 }
1674 } 1491 }
1675 1492 vloc[0] = vloc[1] = -1;
1676 1493 for (i = 0, match = 1; match && i < (img1->d_h + 1) / 2; i+=16) {
1677 void open_input_file(struct input_state *input) 1494 for (j = 0; j < match && (img1->d_w + 1) / 2; j+=16) {
1678 { 1495 int k, l;
1679 unsigned int fourcc; 1496 int si = mmin(i + 16, (img1->d_h + 1) / 2) - i;
1680 1497 int sj = mmin(j + 16, (img1->d_w + 1) / 2) - j;
1681 /* Parse certain options from the input file, if possible */ 1498 for (k = 0; match && k < si; k++)
1682 input->file = strcmp(input->fn, "-") ? fopen(input->fn, "rb") 1499 for (l = 0; match && l < sj; l++) {
1683 : set_binary_mode(stdin); 1500 if (*(img1->planes[VPX_PLANE_V] +
1684 1501 (i + k) * img1->stride[VPX_PLANE_V] + j + l) !=
1685 if (!input->file) 1502 *(img2->planes[VPX_PLANE_V] +
1686 fatal("Failed to open input file"); 1503 (i + k) * img2->stride[VPX_PLANE_V] + j + l)) {
1687 1504 vloc[0] = i + k;
1688 /* For RAW input sources, these bytes will applied on the first frame 1505 vloc[1] = j + l;
1689 * in read_frame(). 1506 match = 0;
1690 */
1691 input->detect.buf_read = fread(input->detect.buf, 1, 4, input->file);
1692 input->detect.position = 0;
1693
1694 if (input->detect.buf_read == 4
1695 && file_is_y4m(input->file, &input->y4m, input->detect.buf))
1696 {
1697 if (y4m_input_open(&input->y4m, input->file, input->detect.buf, 4) >= 0)
1698 {
1699 input->file_type = FILE_TYPE_Y4M;
1700 input->w = input->y4m.pic_w;
1701 input->h = input->y4m.pic_h;
1702 input->framerate.num = input->y4m.fps_n;
1703 input->framerate.den = input->y4m.fps_d;
1704 input->use_i420 = 0;
1705 }
1706 else
1707 fatal("Unsupported Y4M stream.");
1708 }
1709 else if (input->detect.buf_read == 4 && file_is_ivf(input, &fourcc))
1710 {
1711 input->file_type = FILE_TYPE_IVF;
1712 switch (fourcc)
1713 {
1714 case 0x32315659:
1715 input->use_i420 = 0;
1716 break; 1507 break;
1717 case 0x30323449: 1508 }
1718 input->use_i420 = 1;
1719 break;
1720 default:
1721 fatal("Unsupported fourcc (%08x) in IVF", fourcc);
1722 } 1509 }
1723 } 1510 }
1511 }
1512 }
1513
1514 static int compare_img(vpx_image_t *img1, vpx_image_t *img2)
1515 {
1516 int match = 1;
1517 int i;
1518
1519 match &= (img1->fmt == img2->fmt);
1520 match &= (img1->w == img2->w);
1521 match &= (img1->h == img2->h);
1522
1523 for (i = 0; i < img1->d_h; i++)
1524 match &= (memcmp(img1->planes[VPX_PLANE_Y]+i*img1->stride[VPX_PLANE_Y],
1525 img2->planes[VPX_PLANE_Y]+i*img2->stride[VPX_PLANE_Y],
1526 img1->d_w) == 0);
1527
1528 for (i = 0; i < img1->d_h/2; i++)
1529 match &= (memcmp(img1->planes[VPX_PLANE_U]+i*img1->stride[VPX_PLANE_U],
1530 img2->planes[VPX_PLANE_U]+i*img2->stride[VPX_PLANE_U],
1531 (img1->d_w + 1) / 2) == 0);
1532
1533 for (i = 0; i < img1->d_h/2; i++)
1534 match &= (memcmp(img1->planes[VPX_PLANE_V]+i*img1->stride[VPX_PLANE_U],
1535 img2->planes[VPX_PLANE_V]+i*img2->stride[VPX_PLANE_U],
1536 (img1->d_w + 1) / 2) == 0);
1537
1538 return match;
1539 }
1540
1541
1542 #define NELEMENTS(x) (sizeof(x)/sizeof(x[0]))
1543 #define MAX(x,y) ((x)>(y)?(x):(y))
1544 #if CONFIG_VP8_ENCODER && !CONFIG_VP9_ENCODER
1545 #define ARG_CTRL_CNT_MAX NELEMENTS(vp8_arg_ctrl_map)
1546 #elif !CONFIG_VP8_ENCODER && CONFIG_VP9_ENCODER
1547 #define ARG_CTRL_CNT_MAX NELEMENTS(vp9_arg_ctrl_map)
1548 #else
1549 #define ARG_CTRL_CNT_MAX MAX(NELEMENTS(vp8_arg_ctrl_map), \
1550 NELEMENTS(vp9_arg_ctrl_map))
1551 #endif
1552
1553 /* Configuration elements common to all streams */
1554 struct global_config {
1555 const struct codec_item *codec;
1556 int passes;
1557 int pass;
1558 int usage;
1559 int deadline;
1560 int use_i420;
1561 int quiet;
1562 int verbose;
1563 int limit;
1564 int skip_frames;
1565 int show_psnr;
1566 int test_decode;
1567 int have_framerate;
1568 struct vpx_rational framerate;
1569 int out_part;
1570 int debug;
1571 int show_q_hist_buckets;
1572 int show_rate_hist_buckets;
1573 };
1574
1575
1576 /* Per-stream configuration */
1577 struct stream_config {
1578 struct vpx_codec_enc_cfg cfg;
1579 const char *out_fn;
1580 const char *stats_fn;
1581 stereo_format_t stereo_fmt;
1582 int arg_ctrls[ARG_CTRL_CNT_MAX][2];
1583 int arg_ctrl_cnt;
1584 int write_webm;
1585 int have_kf_max_dist;
1586 };
1587
1588
1589 struct stream_state {
1590 int index;
1591 struct stream_state *next;
1592 struct stream_config config;
1593 FILE *file;
1594 struct rate_hist rate_hist;
1595 EbmlGlobal ebml;
1596 uint32_t hash;
1597 uint64_t psnr_sse_total;
1598 uint64_t psnr_samples_total;
1599 double psnr_totals[4];
1600 int psnr_count;
1601 int counts[64];
1602 vpx_codec_ctx_t encoder;
1603 unsigned int frames_out;
1604 uint64_t cx_time;
1605 size_t nbytes;
1606 stats_io_t stats;
1607 vpx_codec_ctx_t decoder;
1608 vpx_ref_frame_t ref_enc;
1609 vpx_ref_frame_t ref_dec;
1610 int mismatch_seen;
1611 };
1612
1613
1614 void validate_positive_rational(const char *msg,
1615 struct vpx_rational *rat) {
1616 if (rat->den < 0) {
1617 rat->num *= -1;
1618 rat->den *= -1;
1619 }
1620
1621 if (rat->num < 0)
1622 die("Error: %s must be positive\n", msg);
1623
1624 if (!rat->den)
1625 die("Error: %s has zero denominator\n", msg);
1626 }
1627
1628
1629 static void parse_global_config(struct global_config *global, char **argv) {
1630 char **argi, **argj;
1631 struct arg arg;
1632
1633 /* Initialize default parameters */
1634 memset(global, 0, sizeof(*global));
1635 global->codec = codecs;
1636 global->passes = 1;
1637 global->use_i420 = 1;
1638
1639 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
1640 arg.argv_step = 1;
1641
1642 if (arg_match(&arg, &codecarg, argi)) {
1643 int j, k = -1;
1644
1645 for (j = 0; j < sizeof(codecs) / sizeof(codecs[0]); j++)
1646 if (!strcmp(codecs[j].name, arg.val))
1647 k = j;
1648
1649 if (k >= 0)
1650 global->codec = codecs + k;
1651 else
1652 die("Error: Unrecognized argument (%s) to --codec\n",
1653 arg.val);
1654
1655 } else if (arg_match(&arg, &passes, argi)) {
1656 global->passes = arg_parse_uint(&arg);
1657
1658 if (global->passes < 1 || global->passes > 2)
1659 die("Error: Invalid number of passes (%d)\n", global->passes);
1660 } else if (arg_match(&arg, &pass_arg, argi)) {
1661 global->pass = arg_parse_uint(&arg);
1662
1663 if (global->pass < 1 || global->pass > 2)
1664 die("Error: Invalid pass selected (%d)\n",
1665 global->pass);
1666 } else if (arg_match(&arg, &usage, argi))
1667 global->usage = arg_parse_uint(&arg);
1668 else if (arg_match(&arg, &deadline, argi))
1669 global->deadline = arg_parse_uint(&arg);
1670 else if (arg_match(&arg, &best_dl, argi))
1671 global->deadline = VPX_DL_BEST_QUALITY;
1672 else if (arg_match(&arg, &good_dl, argi))
1673 global->deadline = VPX_DL_GOOD_QUALITY;
1674 else if (arg_match(&arg, &rt_dl, argi))
1675 global->deadline = VPX_DL_REALTIME;
1676 else if (arg_match(&arg, &use_yv12, argi))
1677 global->use_i420 = 0;
1678 else if (arg_match(&arg, &use_i420, argi))
1679 global->use_i420 = 1;
1680 else if (arg_match(&arg, &quietarg, argi))
1681 global->quiet = 1;
1682 else if (arg_match(&arg, &verbosearg, argi))
1683 global->verbose = 1;
1684 else if (arg_match(&arg, &limit, argi))
1685 global->limit = arg_parse_uint(&arg);
1686 else if (arg_match(&arg, &skip, argi))
1687 global->skip_frames = arg_parse_uint(&arg);
1688 else if (arg_match(&arg, &psnrarg, argi))
1689 global->show_psnr = 1;
1690 else if (arg_match(&arg, &recontest, argi))
1691 global->test_decode = 1;
1692 else if (arg_match(&arg, &framerate, argi)) {
1693 global->framerate = arg_parse_rational(&arg);
1694 validate_positive_rational(arg.name, &global->framerate);
1695 global->have_framerate = 1;
1696 } else if (arg_match(&arg, &out_part, argi))
1697 global->out_part = 1;
1698 else if (arg_match(&arg, &debugmode, argi))
1699 global->debug = 1;
1700 else if (arg_match(&arg, &q_hist_n, argi))
1701 global->show_q_hist_buckets = arg_parse_uint(&arg);
1702 else if (arg_match(&arg, &rate_hist_n, argi))
1703 global->show_rate_hist_buckets = arg_parse_uint(&arg);
1724 else 1704 else
1725 { 1705 argj++;
1726 input->file_type = FILE_TYPE_RAW; 1706 }
1707
1708 /* Validate global config */
1709
1710 if (global->pass) {
1711 /* DWIM: Assume the user meant passes=2 if pass=2 is specified */
1712 if (global->pass > global->passes) {
1713 warn("Assuming --pass=%d implies --passes=%d\n",
1714 global->pass, global->pass);
1715 global->passes = global->pass;
1727 } 1716 }
1728 } 1717 }
1729 1718 }
1730 1719
1731 static void close_input_file(struct input_state *input) 1720
1732 { 1721 void open_input_file(struct input_state *input) {
1733 fclose(input->file); 1722 unsigned int fourcc;
1734 if (input->file_type == FILE_TYPE_Y4M) 1723
1735 y4m_input_close(&input->y4m); 1724 /* Parse certain options from the input file, if possible */
1725 input->file = strcmp(input->fn, "-") ? fopen(input->fn, "rb")
1726 : set_binary_mode(stdin);
1727
1728 if (!input->file)
1729 fatal("Failed to open input file");
1730
1731 /* For RAW input sources, these bytes will applied on the first frame
1732 * in read_frame().
1733 */
1734 input->detect.buf_read = fread(input->detect.buf, 1, 4, input->file);
1735 input->detect.position = 0;
1736
1737 if (input->detect.buf_read == 4
1738 && file_is_y4m(input->file, &input->y4m, input->detect.buf)) {
1739 if (y4m_input_open(&input->y4m, input->file, input->detect.buf, 4) >= 0) {
1740 input->file_type = FILE_TYPE_Y4M;
1741 input->w = input->y4m.pic_w;
1742 input->h = input->y4m.pic_h;
1743 input->framerate.num = input->y4m.fps_n;
1744 input->framerate.den = input->y4m.fps_d;
1745 input->use_i420 = 0;
1746 } else
1747 fatal("Unsupported Y4M stream.");
1748 } else if (input->detect.buf_read == 4 && file_is_ivf(input, &fourcc)) {
1749 input->file_type = FILE_TYPE_IVF;
1750 switch (fourcc) {
1751 case 0x32315659:
1752 input->use_i420 = 0;
1753 break;
1754 case 0x30323449:
1755 input->use_i420 = 1;
1756 break;
1757 default:
1758 fatal("Unsupported fourcc (%08x) in IVF", fourcc);
1759 }
1760 } else {
1761 input->file_type = FILE_TYPE_RAW;
1762 }
1763 }
1764
1765
1766 static void close_input_file(struct input_state *input) {
1767 fclose(input->file);
1768 if (input->file_type == FILE_TYPE_Y4M)
1769 y4m_input_close(&input->y4m);
1736 } 1770 }
1737 1771
1738 static struct stream_state *new_stream(struct global_config *global, 1772 static struct stream_state *new_stream(struct global_config *global,
1739 struct stream_state *prev) 1773 struct stream_state *prev) {
1740 { 1774 struct stream_state *stream;
1741 struct stream_state *stream; 1775
1742 1776 stream = calloc(1, sizeof(*stream));
1743 stream = calloc(1, sizeof(*stream)); 1777 if (!stream)
1744 if(!stream) 1778 fatal("Failed to allocate new stream.");
1745 fatal("Failed to allocate new stream."); 1779 if (prev) {
1746 if(prev) 1780 memcpy(stream, prev, sizeof(*stream));
1747 { 1781 stream->index++;
1748 memcpy(stream, prev, sizeof(*stream)); 1782 prev->next = stream;
1749 stream->index++; 1783 } else {
1750 prev->next = stream; 1784 vpx_codec_err_t res;
1751 } 1785
1752 else 1786 /* Populate encoder configuration */
1753 { 1787 res = vpx_codec_enc_config_default(global->codec->iface(),
1754 vpx_codec_err_t res; 1788 &stream->config.cfg,
1755 1789 global->usage);
1756 /* Populate encoder configuration */ 1790 if (res)
1757 res = vpx_codec_enc_config_default(global->codec->iface, 1791 fatal("Failed to get config: %s\n", vpx_codec_err_to_string(res));
1758 &stream->config.cfg, 1792
1759 global->usage); 1793 /* Change the default timebase to a high enough value so that the
1760 if (res) 1794 * encoder will always create strictly increasing timestamps.
1761 fatal("Failed to get config: %s\n", vpx_codec_err_to_string(res)); 1795 */
1762 1796 stream->config.cfg.g_timebase.den = 1000;
1763 /* Change the default timebase to a high enough value so that the 1797
1764 * encoder will always create strictly increasing timestamps. 1798 /* Never use the library's default resolution, require it be parsed
1765 */ 1799 * from the file or set on the command line.
1766 stream->config.cfg.g_timebase.den = 1000; 1800 */
1767 1801 stream->config.cfg.g_w = 0;
1768 /* Never use the library's default resolution, require it be parsed 1802 stream->config.cfg.g_h = 0;
1769 * from the file or set on the command line. 1803
1770 */ 1804 /* Initialize remaining stream parameters */
1771 stream->config.cfg.g_w = 0; 1805 stream->config.stereo_fmt = STEREO_FORMAT_MONO;
1772 stream->config.cfg.g_h = 0; 1806 stream->config.write_webm = 1;
1773 1807 stream->ebml.last_pts_ms = -1;
1774 /* Initialize remaining stream parameters */ 1808
1775 stream->config.stereo_fmt = STEREO_FORMAT_MONO; 1809 /* Allows removal of the application version from the EBML tags */
1776 stream->config.write_webm = 1; 1810 stream->ebml.debug = global->debug;
1777 stream->ebml.last_pts_ms = -1; 1811 }
1778 1812
1779 /* Allows removal of the application version from the EBML tags */ 1813 /* Output files must be specified for each stream */
1780 stream->ebml.debug = global->debug; 1814 stream->config.out_fn = NULL;
1781 } 1815
1782 1816 stream->next = NULL;
1783 /* Output files must be specified for each stream */ 1817 return stream;
1784 stream->config.out_fn = NULL;
1785
1786 stream->next = NULL;
1787 return stream;
1788 } 1818 }
1789 1819
1790 1820
1791 static int parse_stream_params(struct global_config *global, 1821 static int parse_stream_params(struct global_config *global,
1792 struct stream_state *stream, 1822 struct stream_state *stream,
1793 char **argv) 1823 char **argv) {
1794 { 1824 char **argi, **argj;
1795 char **argi, **argj; 1825 struct arg arg;
1796 struct arg arg; 1826 static const arg_def_t **ctrl_args = no_args;
1797 static const arg_def_t **ctrl_args = no_args; 1827 static const int *ctrl_args_map = NULL;
1798 static const int *ctrl_args_map = NULL; 1828 struct stream_config *config = &stream->config;
1799 struct stream_config *config = &stream->config; 1829 int eos_mark_found = 0;
1800 int eos_mark_found = 0; 1830
1801 1831 /* Handle codec specific options */
1802 /* Handle codec specific options */ 1832 if (0) {
1803 if (global->codec->iface == &vpx_codec_vp8_cx_algo) 1833 #if CONFIG_VP8_ENCODER
1804 { 1834 } else if (global->codec->iface == vpx_codec_vp8_cx) {
1805 ctrl_args = vp8_args; 1835 ctrl_args = vp8_args;
1806 ctrl_args_map = vp8_arg_ctrl_map; 1836 ctrl_args_map = vp8_arg_ctrl_map;
1837 #endif
1838 #if CONFIG_VP9_ENCODER
1839 } else if (global->codec->iface == vpx_codec_vp9_cx) {
1840 ctrl_args = vp9_args;
1841 ctrl_args_map = vp9_arg_ctrl_map;
1842 #endif
1843 }
1844
1845 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
1846 arg.argv_step = 1;
1847
1848 /* Once we've found an end-of-stream marker (--) we want to continue
1849 * shifting arguments but not consuming them.
1850 */
1851 if (eos_mark_found) {
1852 argj++;
1853 continue;
1854 } else if (!strcmp(*argj, "--")) {
1855 eos_mark_found = 1;
1856 continue;
1807 } 1857 }
1808 1858
1809 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) 1859 if (0);
1810 { 1860 else if (arg_match(&arg, &outputfile, argi))
1811 arg.argv_step = 1; 1861 config->out_fn = arg.val;
1812 1862 else if (arg_match(&arg, &fpf_name, argi))
1813 /* Once we've found an end-of-stream marker (--) we want to continue 1863 config->stats_fn = arg.val;
1814 * shifting arguments but not consuming them. 1864 else if (arg_match(&arg, &use_ivf, argi))
1815 */ 1865 config->write_webm = 0;
1816 if (eos_mark_found) 1866 else if (arg_match(&arg, &threads, argi))
1817 { 1867 config->cfg.g_threads = arg_parse_uint(&arg);
1818 argj++; 1868 else if (arg_match(&arg, &profile, argi))
1819 continue; 1869 config->cfg.g_profile = arg_parse_uint(&arg);
1870 else if (arg_match(&arg, &width, argi))
1871 config->cfg.g_w = arg_parse_uint(&arg);
1872 else if (arg_match(&arg, &height, argi))
1873 config->cfg.g_h = arg_parse_uint(&arg);
1874 else if (arg_match(&arg, &stereo_mode, argi))
1875 config->stereo_fmt = arg_parse_enum_or_int(&arg);
1876 else if (arg_match(&arg, &timebase, argi)) {
1877 config->cfg.g_timebase = arg_parse_rational(&arg);
1878 validate_positive_rational(arg.name, &config->cfg.g_timebase);
1879 } else if (arg_match(&arg, &error_resilient, argi))
1880 config->cfg.g_error_resilient = arg_parse_uint(&arg);
1881 else if (arg_match(&arg, &lag_in_frames, argi))
1882 config->cfg.g_lag_in_frames = arg_parse_uint(&arg);
1883 else if (arg_match(&arg, &dropframe_thresh, argi))
1884 config->cfg.rc_dropframe_thresh = arg_parse_uint(&arg);
1885 else if (arg_match(&arg, &resize_allowed, argi))
1886 config->cfg.rc_resize_allowed = arg_parse_uint(&arg);
1887 else if (arg_match(&arg, &resize_up_thresh, argi))
1888 config->cfg.rc_resize_up_thresh = arg_parse_uint(&arg);
1889 else if (arg_match(&arg, &resize_down_thresh, argi))
1890 config->cfg.rc_resize_down_thresh = arg_parse_uint(&arg);
1891 else if (arg_match(&arg, &end_usage, argi))
1892 config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
1893 else if (arg_match(&arg, &target_bitrate, argi))
1894 config->cfg.rc_target_bitrate = arg_parse_uint(&arg);
1895 else if (arg_match(&arg, &min_quantizer, argi))
1896 config->cfg.rc_min_quantizer = arg_parse_uint(&arg);
1897 else if (arg_match(&arg, &max_quantizer, argi))
1898 config->cfg.rc_max_quantizer = arg_parse_uint(&arg);
1899 else if (arg_match(&arg, &undershoot_pct, argi))
1900 config->cfg.rc_undershoot_pct = arg_parse_uint(&arg);
1901 else if (arg_match(&arg, &overshoot_pct, argi))
1902 config->cfg.rc_overshoot_pct = arg_parse_uint(&arg);
1903 else if (arg_match(&arg, &buf_sz, argi))
1904 config->cfg.rc_buf_sz = arg_parse_uint(&arg);
1905 else if (arg_match(&arg, &buf_initial_sz, argi))
1906 config->cfg.rc_buf_initial_sz = arg_parse_uint(&arg);
1907 else if (arg_match(&arg, &buf_optimal_sz, argi))
1908 config->cfg.rc_buf_optimal_sz = arg_parse_uint(&arg);
1909 else if (arg_match(&arg, &bias_pct, argi)) {
1910 config->cfg.rc_2pass_vbr_bias_pct = arg_parse_uint(&arg);
1911
1912 if (global->passes < 2)
1913 warn("option %s ignored in one-pass mode.\n", arg.name);
1914 } else if (arg_match(&arg, &minsection_pct, argi)) {
1915 config->cfg.rc_2pass_vbr_minsection_pct = arg_parse_uint(&arg);
1916
1917 if (global->passes < 2)
1918 warn("option %s ignored in one-pass mode.\n", arg.name);
1919 } else if (arg_match(&arg, &maxsection_pct, argi)) {
1920 config->cfg.rc_2pass_vbr_maxsection_pct = arg_parse_uint(&arg);
1921
1922 if (global->passes < 2)
1923 warn("option %s ignored in one-pass mode.\n", arg.name);
1924 } else if (arg_match(&arg, &kf_min_dist, argi))
1925 config->cfg.kf_min_dist = arg_parse_uint(&arg);
1926 else if (arg_match(&arg, &kf_max_dist, argi)) {
1927 config->cfg.kf_max_dist = arg_parse_uint(&arg);
1928 config->have_kf_max_dist = 1;
1929 } else if (arg_match(&arg, &kf_disabled, argi))
1930 config->cfg.kf_mode = VPX_KF_DISABLED;
1931 else {
1932 int i, match = 0;
1933
1934 for (i = 0; ctrl_args[i]; i++) {
1935 if (arg_match(&arg, ctrl_args[i], argi)) {
1936 int j;
1937 match = 1;
1938
1939 /* Point either to the next free element or the first
1940 * instance of this control.
1941 */
1942 for (j = 0; j < config->arg_ctrl_cnt; j++)
1943 if (config->arg_ctrls[j][0] == ctrl_args_map[i])
1944 break;
1945
1946 /* Update/insert */
1947 assert(j < ARG_CTRL_CNT_MAX);
1948 if (j < ARG_CTRL_CNT_MAX) {
1949 config->arg_ctrls[j][0] = ctrl_args_map[i];
1950 config->arg_ctrls[j][1] = arg_parse_enum_or_int(&arg);
1951 if (j == config->arg_ctrl_cnt)
1952 config->arg_ctrl_cnt++;
1953 }
1954
1820 } 1955 }
1821 else if (!strcmp(*argj, "--")) 1956 }
1822 { 1957
1823 eos_mark_found = 1; 1958 if (!match)
1824 continue; 1959 argj++;
1825 }
1826
1827 if (0);
1828 else if (arg_match(&arg, &outputfile, argi))
1829 config->out_fn = arg.val;
1830 else if (arg_match(&arg, &fpf_name, argi))
1831 config->stats_fn = arg.val;
1832 else if (arg_match(&arg, &use_ivf, argi))
1833 config->write_webm = 0;
1834 else if (arg_match(&arg, &threads, argi))
1835 config->cfg.g_threads = arg_parse_uint(&arg);
1836 else if (arg_match(&arg, &profile, argi))
1837 config->cfg.g_profile = arg_parse_uint(&arg);
1838 else if (arg_match(&arg, &width, argi))
1839 config->cfg.g_w = arg_parse_uint(&arg);
1840 else if (arg_match(&arg, &height, argi))
1841 config->cfg.g_h = arg_parse_uint(&arg);
1842 else if (arg_match(&arg, &stereo_mode, argi))
1843 config->stereo_fmt = arg_parse_enum_or_int(&arg);
1844 else if (arg_match(&arg, &timebase, argi))
1845 {
1846 config->cfg.g_timebase = arg_parse_rational(&arg);
1847 validate_positive_rational(arg.name, &config->cfg.g_timebase);
1848 }
1849 else if (arg_match(&arg, &error_resilient, argi))
1850 config->cfg.g_error_resilient = arg_parse_uint(&arg);
1851 else if (arg_match(&arg, &lag_in_frames, argi))
1852 config->cfg.g_lag_in_frames = arg_parse_uint(&arg);
1853 else if (arg_match(&arg, &dropframe_thresh, argi))
1854 config->cfg.rc_dropframe_thresh = arg_parse_uint(&arg);
1855 else if (arg_match(&arg, &resize_allowed, argi))
1856 config->cfg.rc_resize_allowed = arg_parse_uint(&arg);
1857 else if (arg_match(&arg, &resize_up_thresh, argi))
1858 config->cfg.rc_resize_up_thresh = arg_parse_uint(&arg);
1859 else if (arg_match(&arg, &resize_down_thresh, argi))
1860 config->cfg.rc_resize_down_thresh = arg_parse_uint(&arg);
1861 else if (arg_match(&arg, &end_usage, argi))
1862 config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
1863 else if (arg_match(&arg, &target_bitrate, argi))
1864 config->cfg.rc_target_bitrate = arg_parse_uint(&arg);
1865 else if (arg_match(&arg, &min_quantizer, argi))
1866 config->cfg.rc_min_quantizer = arg_parse_uint(&arg);
1867 else if (arg_match(&arg, &max_quantizer, argi))
1868 config->cfg.rc_max_quantizer = arg_parse_uint(&arg);
1869 else if (arg_match(&arg, &undershoot_pct, argi))
1870 config->cfg.rc_undershoot_pct = arg_parse_uint(&arg);
1871 else if (arg_match(&arg, &overshoot_pct, argi))
1872 config->cfg.rc_overshoot_pct = arg_parse_uint(&arg);
1873 else if (arg_match(&arg, &buf_sz, argi))
1874 config->cfg.rc_buf_sz = arg_parse_uint(&arg);
1875 else if (arg_match(&arg, &buf_initial_sz, argi))
1876 config->cfg.rc_buf_initial_sz = arg_parse_uint(&arg);
1877 else if (arg_match(&arg, &buf_optimal_sz, argi))
1878 config->cfg.rc_buf_optimal_sz = arg_parse_uint(&arg);
1879 else if (arg_match(&arg, &bias_pct, argi))
1880 {
1881 config->cfg.rc_2pass_vbr_bias_pct = arg_parse_uint(&arg);
1882
1883 if (global->passes < 2)
1884 warn("option %s ignored in one-pass mode.\n", arg.name);
1885 }
1886 else if (arg_match(&arg, &minsection_pct, argi))
1887 {
1888 config->cfg.rc_2pass_vbr_minsection_pct = arg_parse_uint(&arg);
1889
1890 if (global->passes < 2)
1891 warn("option %s ignored in one-pass mode.\n", arg.name);
1892 }
1893 else if (arg_match(&arg, &maxsection_pct, argi))
1894 {
1895 config->cfg.rc_2pass_vbr_maxsection_pct = arg_parse_uint(&arg);
1896
1897 if (global->passes < 2)
1898 warn("option %s ignored in one-pass mode.\n", arg.name);
1899 }
1900 else if (arg_match(&arg, &kf_min_dist, argi))
1901 config->cfg.kf_min_dist = arg_parse_uint(&arg);
1902 else if (arg_match(&arg, &kf_max_dist, argi))
1903 {
1904 config->cfg.kf_max_dist = arg_parse_uint(&arg);
1905 config->have_kf_max_dist = 1;
1906 }
1907 else if (arg_match(&arg, &kf_disabled, argi))
1908 config->cfg.kf_mode = VPX_KF_DISABLED;
1909 else
1910 {
1911 int i, match = 0;
1912
1913 for (i = 0; ctrl_args[i]; i++)
1914 {
1915 if (arg_match(&arg, ctrl_args[i], argi))
1916 {
1917 int j;
1918 match = 1;
1919
1920 /* Point either to the next free element or the first
1921 * instance of this control.
1922 */
1923 for(j=0; j<config->arg_ctrl_cnt; j++)
1924 if(config->arg_ctrls[j][0] == ctrl_args_map[i])
1925 break;
1926
1927 /* Update/insert */
1928 assert(j < ARG_CTRL_CNT_MAX);
1929 if (j < ARG_CTRL_CNT_MAX)
1930 {
1931 config->arg_ctrls[j][0] = ctrl_args_map[i];
1932 config->arg_ctrls[j][1] = arg_parse_enum_or_int(&arg);
1933 if(j == config->arg_ctrl_cnt)
1934 config->arg_ctrl_cnt++;
1935 }
1936
1937 }
1938 }
1939
1940 if (!match)
1941 argj++;
1942 }
1943 } 1960 }
1944 1961 }
1945 return eos_mark_found; 1962
1963 return eos_mark_found;
1946 } 1964 }
1947 1965
1948 1966
1949 #define FOREACH_STREAM(func)\ 1967 #define FOREACH_STREAM(func)\
1950 do\ 1968 do\
1951 {\ 1969 {\
1952 struct stream_state *stream;\ 1970 struct stream_state *stream;\
1953 \ 1971 \
1954 for(stream = streams; stream; stream = stream->next)\ 1972 for(stream = streams; stream; stream = stream->next)\
1955 func;\ 1973 func;\
1956 }while(0) 1974 }while(0)
1957 1975
1958 1976
1959 static void validate_stream_config(struct stream_state *stream) 1977 static void validate_stream_config(struct stream_state *stream) {
1960 { 1978 struct stream_state *streami;
1961 struct stream_state *streami; 1979
1962 1980 if (!stream->config.cfg.g_w || !stream->config.cfg.g_h)
1963 if(!stream->config.cfg.g_w || !stream->config.cfg.g_h) 1981 fatal("Stream %d: Specify stream dimensions with --width (-w) "
1964 fatal("Stream %d: Specify stream dimensions with --width (-w) " 1982 " and --height (-h)", stream->index);
1965 " and --height (-h)", stream->index); 1983
1966 1984 for (streami = stream; streami; streami = streami->next) {
1967 for(streami = stream; streami; streami = streami->next) 1985 /* All streams require output files */
1968 { 1986 if (!streami->config.out_fn)
1969 /* All streams require output files */ 1987 fatal("Stream %d: Output file is required (specify with -o)",
1970 if(!streami->config.out_fn) 1988 streami->index);
1971 fatal("Stream %d: Output file is required (specify with -o)", 1989
1972 streami->index); 1990 /* Check for two streams outputting to the same file */
1973 1991 if (streami != stream) {
1974 /* Check for two streams outputting to the same file */ 1992 const char *a = stream->config.out_fn;
1975 if(streami != stream) 1993 const char *b = streami->config.out_fn;
1976 { 1994 if (!strcmp(a, b) && strcmp(a, "/dev/null") && strcmp(a, ":nul"))
1977 const char *a = stream->config.out_fn; 1995 fatal("Stream %d: duplicate output file (from stream %d)",
1978 const char *b = streami->config.out_fn; 1996 streami->index, stream->index);
1979 if(!strcmp(a,b) && strcmp(a, "/dev/null") && strcmp(a, ":nul"))
1980 fatal("Stream %d: duplicate output file (from stream %d)",
1981 streami->index, stream->index);
1982 }
1983
1984 /* Check for two streams sharing a stats file. */
1985 if(streami != stream)
1986 {
1987 const char *a = stream->config.stats_fn;
1988 const char *b = streami->config.stats_fn;
1989 if(a && b && !strcmp(a,b))
1990 fatal("Stream %d: duplicate stats file (from stream %d)",
1991 streami->index, stream->index);
1992 }
1993 } 1997 }
1998
1999 /* Check for two streams sharing a stats file. */
2000 if (streami != stream) {
2001 const char *a = stream->config.stats_fn;
2002 const char *b = streami->config.stats_fn;
2003 if (a && b && !strcmp(a, b))
2004 fatal("Stream %d: duplicate stats file (from stream %d)",
2005 streami->index, stream->index);
2006 }
2007 }
1994 } 2008 }
1995 2009
1996 2010
1997 static void set_stream_dimensions(struct stream_state *stream, 2011 static void set_stream_dimensions(struct stream_state *stream,
1998 unsigned int w, 2012 unsigned int w,
1999 unsigned int h) 2013 unsigned int h) {
2000 { 2014 if ((stream->config.cfg.g_w && stream->config.cfg.g_w != w)
2001 if ((stream->config.cfg.g_w && stream->config.cfg.g_w != w) 2015 || (stream->config.cfg.g_h && stream->config.cfg.g_h != h))
2002 ||(stream->config.cfg.g_h && stream->config.cfg.g_h != h)) 2016 fatal("Stream %d: Resizing not yet supported", stream->index);
2003 fatal("Stream %d: Resizing not yet supported", stream->index); 2017 stream->config.cfg.g_w = w;
2004 stream->config.cfg.g_w = w; 2018 stream->config.cfg.g_h = h;
2005 stream->config.cfg.g_h = h;
2006 } 2019 }
2007 2020
2008 2021
2009 static void set_default_kf_interval(struct stream_state *stream, 2022 static void set_default_kf_interval(struct stream_state *stream,
2010 struct global_config *global) 2023 struct global_config *global) {
2011 { 2024 /* Use a max keyframe interval of 5 seconds, if none was
2012 /* Use a max keyframe interval of 5 seconds, if none was 2025 * specified on the command line.
2013 * specified on the command line. 2026 */
2014 */ 2027 if (!stream->config.have_kf_max_dist) {
2015 if (!stream->config.have_kf_max_dist) 2028 double framerate = (double)global->framerate.num / global->framerate.den;
2016 { 2029 if (framerate > 0.0)
2017 double framerate = (double)global->framerate.num/global->framerate.den; 2030 stream->config.cfg.kf_max_dist = (unsigned int)(5.0 * framerate);
2018 if (framerate > 0.0) 2031 }
2019 stream->config.cfg.kf_max_dist = (unsigned int)(5.0*framerate);
2020 }
2021 } 2032 }
2022 2033
2023 2034
2024 static void show_stream_config(struct stream_state *stream, 2035 static void show_stream_config(struct stream_state *stream,
2025 struct global_config *global, 2036 struct global_config *global,
2026 struct input_state *input) 2037 struct input_state *input) {
2027 {
2028 2038
2029 #define SHOW(field) \ 2039 #define SHOW(field) \
2030 fprintf(stderr, " %-28s = %d\n", #field, stream->config.cfg.field) 2040 fprintf(stderr, " %-28s = %d\n", #field, stream->config.cfg.field)
2031 2041
2032 if(stream->index == 0) 2042 if (stream->index == 0) {
2033 { 2043 fprintf(stderr, "Codec: %s\n",
2034 fprintf(stderr, "Codec: %s\n", 2044 vpx_codec_iface_name(global->codec->iface()));
2035 vpx_codec_iface_name(global->codec->iface)); 2045 fprintf(stderr, "Source file: %s Format: %s\n", input->fn,
2036 fprintf(stderr, "Source file: %s Format: %s\n", input->fn, 2046 input->use_i420 ? "I420" : "YV12");
2037 input->use_i420 ? "I420" : "YV12"); 2047 }
2038 } 2048 if (stream->next || stream->index)
2039 if(stream->next || stream->index) 2049 fprintf(stderr, "\nStream Index: %d\n", stream->index);
2040 fprintf(stderr, "\nStream Index: %d\n", stream->index); 2050 fprintf(stderr, "Destination file: %s\n", stream->config.out_fn);
2041 fprintf(stderr, "Destination file: %s\n", stream->config.out_fn); 2051 fprintf(stderr, "Encoder parameters:\n");
2042 fprintf(stderr, "Encoder parameters:\n"); 2052
2043 2053 SHOW(g_usage);
2044 SHOW(g_usage); 2054 SHOW(g_threads);
2045 SHOW(g_threads); 2055 SHOW(g_profile);
2046 SHOW(g_profile); 2056 SHOW(g_w);
2047 SHOW(g_w); 2057 SHOW(g_h);
2048 SHOW(g_h); 2058 SHOW(g_timebase.num);
2049 SHOW(g_timebase.num); 2059 SHOW(g_timebase.den);
2050 SHOW(g_timebase.den); 2060 SHOW(g_error_resilient);
2051 SHOW(g_error_resilient); 2061 SHOW(g_pass);
2052 SHOW(g_pass); 2062 SHOW(g_lag_in_frames);
2053 SHOW(g_lag_in_frames); 2063 SHOW(rc_dropframe_thresh);
2054 SHOW(rc_dropframe_thresh); 2064 SHOW(rc_resize_allowed);
2055 SHOW(rc_resize_allowed); 2065 SHOW(rc_resize_up_thresh);
2056 SHOW(rc_resize_up_thresh); 2066 SHOW(rc_resize_down_thresh);
2057 SHOW(rc_resize_down_thresh); 2067 SHOW(rc_end_usage);
2058 SHOW(rc_end_usage); 2068 SHOW(rc_target_bitrate);
2059 SHOW(rc_target_bitrate); 2069 SHOW(rc_min_quantizer);
2060 SHOW(rc_min_quantizer); 2070 SHOW(rc_max_quantizer);
2061 SHOW(rc_max_quantizer); 2071 SHOW(rc_undershoot_pct);
2062 SHOW(rc_undershoot_pct); 2072 SHOW(rc_overshoot_pct);
2063 SHOW(rc_overshoot_pct); 2073 SHOW(rc_buf_sz);
2064 SHOW(rc_buf_sz); 2074 SHOW(rc_buf_initial_sz);
2065 SHOW(rc_buf_initial_sz); 2075 SHOW(rc_buf_optimal_sz);
2066 SHOW(rc_buf_optimal_sz); 2076 SHOW(rc_2pass_vbr_bias_pct);
2067 SHOW(rc_2pass_vbr_bias_pct); 2077 SHOW(rc_2pass_vbr_minsection_pct);
2068 SHOW(rc_2pass_vbr_minsection_pct); 2078 SHOW(rc_2pass_vbr_maxsection_pct);
2069 SHOW(rc_2pass_vbr_maxsection_pct); 2079 SHOW(kf_mode);
2070 SHOW(kf_mode); 2080 SHOW(kf_min_dist);
2071 SHOW(kf_min_dist); 2081 SHOW(kf_max_dist);
2072 SHOW(kf_max_dist);
2073 } 2082 }
2074 2083
2075 2084
2076 static void open_output_file(struct stream_state *stream, 2085 static void open_output_file(struct stream_state *stream,
2077 struct global_config *global) 2086 struct global_config *global) {
2078 { 2087 const char *fn = stream->config.out_fn;
2079 const char *fn = stream->config.out_fn; 2088
2080 2089 stream->file = strcmp(fn, "-") ? fopen(fn, "wb") : set_binary_mode(stdout);
2081 stream->file = strcmp(fn, "-") ? fopen(fn, "wb") : set_binary_mode(stdout); 2090
2082 2091 if (!stream->file)
2083 if (!stream->file) 2092 fatal("Failed to open output file");
2084 fatal("Failed to open output file"); 2093
2085 2094 if (stream->config.write_webm && fseek(stream->file, 0, SEEK_CUR))
2086 if(stream->config.write_webm && fseek(stream->file, 0, SEEK_CUR)) 2095 fatal("WebM output to pipes not supported.");
2087 fatal("WebM output to pipes not supported."); 2096
2088 2097 if (stream->config.write_webm) {
2089 if(stream->config.write_webm) 2098 stream->ebml.stream = stream->file;
2090 { 2099 write_webm_file_header(&stream->ebml, &stream->config.cfg,
2091 stream->ebml.stream = stream->file; 2100 &global->framerate,
2092 write_webm_file_header(&stream->ebml, &stream->config.cfg, 2101 stream->config.stereo_fmt,
2093 &global->framerate, 2102 global->codec->fourcc);
2094 stream->config.stereo_fmt); 2103 } else
2095 } 2104 write_ivf_file_header(stream->file, &stream->config.cfg,
2096 else 2105 global->codec->fourcc, 0);
2097 write_ivf_file_header(stream->file, &stream->config.cfg,
2098 global->codec->fourcc, 0);
2099 } 2106 }
2100 2107
2101 2108
2102 static void close_output_file(struct stream_state *stream, 2109 static void close_output_file(struct stream_state *stream,
2103 unsigned int fourcc) 2110 unsigned int fourcc) {
2104 { 2111 if (stream->config.write_webm) {
2105 if(stream->config.write_webm) 2112 write_webm_file_footer(&stream->ebml, stream->hash);
2106 { 2113 free(stream->ebml.cue_list);
2107 write_webm_file_footer(&stream->ebml, stream->hash); 2114 stream->ebml.cue_list = NULL;
2108 free(stream->ebml.cue_list); 2115 } else {
2109 stream->ebml.cue_list = NULL; 2116 if (!fseek(stream->file, 0, SEEK_SET))
2110 } 2117 write_ivf_file_header(stream->file, &stream->config.cfg,
2111 else 2118 fourcc,
2112 { 2119 stream->frames_out);
2113 if (!fseek(stream->file, 0, SEEK_SET)) 2120 }
2114 write_ivf_file_header(stream->file, &stream->config.cfg, 2121
2115 fourcc, 2122 fclose(stream->file);
2116 stream->frames_out);
2117 }
2118
2119 fclose(stream->file);
2120 } 2123 }
2121 2124
2122 2125
2123 static void setup_pass(struct stream_state *stream, 2126 static void setup_pass(struct stream_state *stream,
2124 struct global_config *global, 2127 struct global_config *global,
2125 int pass) 2128 int pass) {
2126 { 2129 if (stream->config.stats_fn) {
2127 if (stream->config.stats_fn) 2130 if (!stats_open_file(&stream->stats, stream->config.stats_fn,
2128 { 2131 pass))
2129 if (!stats_open_file(&stream->stats, stream->config.stats_fn, 2132 fatal("Failed to open statistics store");
2130 pass)) 2133 } else {
2131 fatal("Failed to open statistics store"); 2134 if (!stats_open_mem(&stream->stats, pass))
2132 } 2135 fatal("Failed to open statistics store");
2133 else 2136 }
2134 { 2137
2135 if (!stats_open_mem(&stream->stats, pass)) 2138 stream->config.cfg.g_pass = global->passes == 2
2136 fatal("Failed to open statistics store"); 2139 ? pass ? VPX_RC_LAST_PASS : VPX_RC_FIRST_PASS
2137 } 2140 : VPX_RC_ONE_PASS;
2138 2141 if (pass)
2139 stream->config.cfg.g_pass = global->passes == 2 2142 stream->config.cfg.rc_twopass_stats_in = stats_get(&stream->stats);
2140 ? pass ? VPX_RC_LAST_PASS : VPX_RC_FIRST_PASS 2143
2141 : VPX_RC_ONE_PASS; 2144 stream->cx_time = 0;
2142 if (pass) 2145 stream->nbytes = 0;
2143 stream->config.cfg.rc_twopass_stats_in = stats_get(&stream->stats); 2146 stream->frames_out = 0;
2144
2145 stream->cx_time = 0;
2146 stream->nbytes = 0;
2147 stream->frames_out = 0;
2148 } 2147 }
2149 2148
2150 2149
2151 static void initialize_encoder(struct stream_state *stream, 2150 static void initialize_encoder(struct stream_state *stream,
2152 struct global_config *global) 2151 struct global_config *global) {
2153 { 2152 int i;
2154 int i; 2153 int flags = 0;
2155 int flags = 0; 2154
2156 2155 flags |= global->show_psnr ? VPX_CODEC_USE_PSNR : 0;
2157 flags |= global->show_psnr ? VPX_CODEC_USE_PSNR : 0; 2156 flags |= global->out_part ? VPX_CODEC_USE_OUTPUT_PARTITION : 0;
2158 flags |= global->out_part ? VPX_CODEC_USE_OUTPUT_PARTITION : 0; 2157
2159 2158 /* Construct Encoder Context */
2160 /* Construct Encoder Context */ 2159 vpx_codec_enc_init(&stream->encoder, global->codec->iface(),
2161 vpx_codec_enc_init(&stream->encoder, global->codec->iface, 2160 &stream->config.cfg, flags);
2162 &stream->config.cfg, flags); 2161 ctx_exit_on_error(&stream->encoder, "Failed to initialize encoder");
2163 ctx_exit_on_error(&stream->encoder, "Failed to initialize encoder"); 2162
2164 2163 /* Note that we bypass the vpx_codec_control wrapper macro because
2165 /* Note that we bypass the vpx_codec_control wrapper macro because 2164 * we're being clever to store the control IDs in an array. Real
2166 * we're being clever to store the control IDs in an array. Real 2165 * applications will want to make use of the enumerations directly
2167 * applications will want to make use of the enumerations directly 2166 */
2168 */ 2167 for (i = 0; i < stream->config.arg_ctrl_cnt; i++) {
2169 for (i = 0; i < stream->config.arg_ctrl_cnt; i++) 2168 int ctrl = stream->config.arg_ctrls[i][0];
2170 { 2169 int value = stream->config.arg_ctrls[i][1];
2171 int ctrl = stream->config.arg_ctrls[i][0]; 2170 if (vpx_codec_control_(&stream->encoder, ctrl, value))
2172 int value = stream->config.arg_ctrls[i][1]; 2171 fprintf(stderr, "Error: Tried to set control %d = %d\n",
2173 if (vpx_codec_control_(&stream->encoder, ctrl, value)) 2172 ctrl, value);
2174 fprintf(stderr, "Error: Tried to set control %d = %d\n", 2173
2175 ctrl, value); 2174 ctx_exit_on_error(&stream->encoder, "Failed to control codec");
2176 2175 }
2177 ctx_exit_on_error(&stream->encoder, "Failed to control codec"); 2176
2178 } 2177 if (global->test_decode) {
2178 int width, height;
2179
2180 vpx_codec_dec_init(&stream->decoder, global->codec->dx_iface(), NULL, 0);
2181
2182 width = (stream->config.cfg.g_w + 15) & ~15;
2183 height = (stream->config.cfg.g_h + 15) & ~15;
2184 vpx_img_alloc(&stream->ref_enc.img, VPX_IMG_FMT_I420, width, height, 1);
2185 vpx_img_alloc(&stream->ref_dec.img, VPX_IMG_FMT_I420, width, height, 1);
2186 stream->ref_enc.frame_type = VP8_LAST_FRAME;
2187 stream->ref_dec.frame_type = VP8_LAST_FRAME;
2188 }
2179 } 2189 }
2180 2190
2181 2191
2182 static void encode_frame(struct stream_state *stream, 2192 static void encode_frame(struct stream_state *stream,
2183 struct global_config *global, 2193 struct global_config *global,
2184 struct vpx_image *img, 2194 struct vpx_image *img,
2185 unsigned int frames_in) 2195 unsigned int frames_in) {
2186 { 2196 vpx_codec_pts_t frame_start, next_frame_start;
2187 vpx_codec_pts_t frame_start, next_frame_start; 2197 struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
2188 struct vpx_codec_enc_cfg *cfg = &stream->config.cfg; 2198 struct vpx_usec_timer timer;
2189 struct vpx_usec_timer timer; 2199
2190 2200 frame_start = (cfg->g_timebase.den * (int64_t)(frames_in - 1)
2191 frame_start = (cfg->g_timebase.den * (int64_t)(frames_in - 1) 2201 * global->framerate.den)
2192 * global->framerate.den) 2202 / cfg->g_timebase.num / global->framerate.num;
2193 / cfg->g_timebase.num / global->framerate.num; 2203 next_frame_start = (cfg->g_timebase.den * (int64_t)(frames_in)
2194 next_frame_start = (cfg->g_timebase.den * (int64_t)(frames_in) 2204 * global->framerate.den)
2195 * global->framerate.den) 2205 / cfg->g_timebase.num / global->framerate.num;
2196 / cfg->g_timebase.num / global->framerate.num; 2206 vpx_usec_timer_start(&timer);
2197 vpx_usec_timer_start(&timer); 2207 vpx_codec_encode(&stream->encoder, img, frame_start,
2198 vpx_codec_encode(&stream->encoder, img, frame_start, 2208 (unsigned long)(next_frame_start - frame_start),
2199 (unsigned long)(next_frame_start - frame_start), 2209 0, global->deadline);
2200 0, global->deadline); 2210 vpx_usec_timer_mark(&timer);
2201 vpx_usec_timer_mark(&timer); 2211 stream->cx_time += vpx_usec_timer_elapsed(&timer);
2202 stream->cx_time += vpx_usec_timer_elapsed(&timer); 2212 ctx_exit_on_error(&stream->encoder, "Stream %d: Failed to encode frame",
2203 ctx_exit_on_error(&stream->encoder, "Stream %d: Failed to encode frame", 2213 stream->index);
2204 stream->index); 2214 }
2205 } 2215
2206 2216
2207 2217 static void update_quantizer_histogram(struct stream_state *stream) {
2208 static void update_quantizer_histogram(struct stream_state *stream) 2218 if (stream->config.cfg.g_pass != VPX_RC_FIRST_PASS) {
2209 { 2219 int q;
2210 if(stream->config.cfg.g_pass != VPX_RC_FIRST_PASS) 2220
2211 { 2221 vpx_codec_control(&stream->encoder, VP8E_GET_LAST_QUANTIZER_64, &q);
2212 int q; 2222 ctx_exit_on_error(&stream->encoder, "Failed to read quantizer");
2213 2223 stream->counts[q]++;
2214 vpx_codec_control(&stream->encoder, VP8E_GET_LAST_QUANTIZER_64, &q); 2224 }
2215 ctx_exit_on_error(&stream->encoder, "Failed to read quantizer");
2216 stream->counts[q]++;
2217 }
2218 } 2225 }
2219 2226
2220 2227
2221 static void get_cx_data(struct stream_state *stream, 2228 static void get_cx_data(struct stream_state *stream,
2222 struct global_config *global, 2229 struct global_config *global,
2223 int *got_data) 2230 int *got_data) {
2224 { 2231 const vpx_codec_cx_pkt_t *pkt;
2225 const vpx_codec_cx_pkt_t *pkt; 2232 const struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
2226 const struct vpx_codec_enc_cfg *cfg = &stream->config.cfg; 2233 vpx_codec_iter_t iter = NULL;
2227 vpx_codec_iter_t iter = NULL; 2234
2228 2235 *got_data = 0;
2229 while ((pkt = vpx_codec_get_cx_data(&stream->encoder, &iter))) 2236 while ((pkt = vpx_codec_get_cx_data(&stream->encoder, &iter))) {
2230 { 2237 static size_t fsize = 0;
2231 static size_t fsize = 0; 2238 static off_t ivf_header_pos = 0;
2232 static off_t ivf_header_pos = 0; 2239
2240 switch (pkt->kind) {
2241 case VPX_CODEC_CX_FRAME_PKT:
2242 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
2243 stream->frames_out++;
2244 }
2245 if (!global->quiet)
2246 fprintf(stderr, " %6luF",
2247 (unsigned long)pkt->data.frame.sz);
2248
2249 update_rate_histogram(&stream->rate_hist, cfg, pkt);
2250 if (stream->config.write_webm) {
2251 /* Update the hash */
2252 if (!stream->ebml.debug)
2253 stream->hash = murmur(pkt->data.frame.buf,
2254 (int)pkt->data.frame.sz,
2255 stream->hash);
2256
2257 write_webm_block(&stream->ebml, cfg, pkt);
2258 } else {
2259 if (pkt->data.frame.partition_id <= 0) {
2260 ivf_header_pos = ftello(stream->file);
2261 fsize = pkt->data.frame.sz;
2262
2263 write_ivf_frame_header(stream->file, pkt);
2264 } else {
2265 fsize += pkt->data.frame.sz;
2266
2267 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
2268 off_t currpos = ftello(stream->file);
2269 fseeko(stream->file, ivf_header_pos, SEEK_SET);
2270 write_ivf_frame_size(stream->file, fsize);
2271 fseeko(stream->file, currpos, SEEK_SET);
2272 }
2273 }
2274
2275 (void) fwrite(pkt->data.frame.buf, 1, pkt->data.frame.sz,
2276 stream->file);
2277 }
2278 stream->nbytes += pkt->data.raw.sz;
2233 2279
2234 *got_data = 1; 2280 *got_data = 1;
2235 2281 if (global->test_decode) {
2236 switch (pkt->kind) 2282 vpx_codec_decode(&stream->decoder, pkt->data.frame.buf,
2237 { 2283 pkt->data.frame.sz, NULL, 0);
2238 case VPX_CODEC_CX_FRAME_PKT: 2284 ctx_exit_on_error(&stream->decoder, "Failed to decode frame");
2239 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) 2285 }
2240 { 2286 break;
2241 stream->frames_out++; 2287 case VPX_CODEC_STATS_PKT:
2242 } 2288 stream->frames_out++;
2289 fprintf(stderr, " %6luS",
2290 (unsigned long)pkt->data.twopass_stats.sz);
2291 stats_write(&stream->stats,
2292 pkt->data.twopass_stats.buf,
2293 pkt->data.twopass_stats.sz);
2294 stream->nbytes += pkt->data.raw.sz;
2295 break;
2296 case VPX_CODEC_PSNR_PKT:
2297
2298 if (global->show_psnr) {
2299 int i;
2300
2301 stream->psnr_sse_total += pkt->data.psnr.sse[0];
2302 stream->psnr_samples_total += pkt->data.psnr.samples[0];
2303 for (i = 0; i < 4; i++) {
2243 if (!global->quiet) 2304 if (!global->quiet)
2244 fprintf(stderr, " %6luF", 2305 fprintf(stderr, "%.3f ", pkt->data.psnr.psnr[i]);
2245 (unsigned long)pkt->data.frame.sz); 2306 stream->psnr_totals[i] += pkt->data.psnr.psnr[i];
2246 2307 }
2247 update_rate_histogram(&stream->rate_hist, cfg, pkt); 2308 stream->psnr_count++;
2248 if(stream->config.write_webm)
2249 {
2250 /* Update the hash */
2251 if(!stream->ebml.debug)
2252 stream->hash = murmur(pkt->data.frame.buf,
2253 (int)pkt->data.frame.sz,
2254 stream->hash);
2255
2256 write_webm_block(&stream->ebml, cfg, pkt);
2257 }
2258 else
2259 {
2260 if (pkt->data.frame.partition_id <= 0)
2261 {
2262 ivf_header_pos = ftello(stream->file);
2263 fsize = pkt->data.frame.sz;
2264
2265 write_ivf_frame_header(stream->file, pkt);
2266 }
2267 else
2268 {
2269 fsize += pkt->data.frame.sz;
2270
2271 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT))
2272 {
2273 off_t currpos = ftello(stream->file);
2274 fseeko(stream->file, ivf_header_pos, SEEK_SET);
2275 write_ivf_frame_size(stream->file, fsize);
2276 fseeko(stream->file, currpos, SEEK_SET);
2277 }
2278 }
2279
2280 (void) fwrite(pkt->data.frame.buf, 1, pkt->data.frame.sz,
2281 stream->file);
2282 }
2283 stream->nbytes += pkt->data.raw.sz;
2284 break;
2285 case VPX_CODEC_STATS_PKT:
2286 stream->frames_out++;
2287 fprintf(stderr, " %6luS",
2288 (unsigned long)pkt->data.twopass_stats.sz);
2289 stats_write(&stream->stats,
2290 pkt->data.twopass_stats.buf,
2291 pkt->data.twopass_stats.sz);
2292 stream->nbytes += pkt->data.raw.sz;
2293 break;
2294 case VPX_CODEC_PSNR_PKT:
2295
2296 if (global->show_psnr)
2297 {
2298 int i;
2299
2300 stream->psnr_sse_total += pkt->data.psnr.sse[0];
2301 stream->psnr_samples_total += pkt->data.psnr.samples[0];
2302 for (i = 0; i < 4; i++)
2303 {
2304 if (!global->quiet)
2305 fprintf(stderr, "%.3f ", pkt->data.psnr.psnr[i]);
2306 stream->psnr_totals[i] += pkt->data.psnr.psnr[i];
2307 }
2308 stream->psnr_count++;
2309 }
2310
2311 break;
2312 default:
2313 break;
2314 } 2309 }
2310
2311 break;
2312 default:
2313 break;
2315 } 2314 }
2316 } 2315 }
2317 2316 }
2318 2317
2319 static void show_psnr(struct stream_state *stream) 2318
2320 { 2319 static void show_psnr(struct stream_state *stream) {
2321 int i; 2320 int i;
2322 double ovpsnr; 2321 double ovpsnr;
2323 2322
2324 if (!stream->psnr_count) 2323 if (!stream->psnr_count)
2325 return; 2324 return;
2326 2325
2327 fprintf(stderr, "Stream %d PSNR (Overall/Avg/Y/U/V)", stream->index); 2326 fprintf(stderr, "Stream %d PSNR (Overall/Avg/Y/U/V)", stream->index);
2328 ovpsnr = vp8_mse2psnr((double)stream->psnr_samples_total, 255.0, 2327 ovpsnr = vp8_mse2psnr((double)stream->psnr_samples_total, 255.0,
2329 (double)stream->psnr_sse_total); 2328 (double)stream->psnr_sse_total);
2330 fprintf(stderr, " %.3f", ovpsnr); 2329 fprintf(stderr, " %.3f", ovpsnr);
2331 2330
2332 for (i = 0; i < 4; i++) 2331 for (i = 0; i < 4; i++) {
2333 { 2332 fprintf(stderr, " %.3f", stream->psnr_totals[i] / stream->psnr_count);
2334 fprintf(stderr, " %.3f", stream->psnr_totals[i]/stream->psnr_count); 2333 }
2334 fprintf(stderr, "\n");
2335 }
2336
2337
2338 float usec_to_fps(uint64_t usec, unsigned int frames) {
2339 return (float)(usec > 0 ? frames * 1000000.0 / (float)usec : 0);
2340 }
2341
2342
2343 static void test_decode(struct stream_state *stream) {
2344 vpx_codec_control(&stream->encoder, VP8_COPY_REFERENCE, &stream->ref_enc);
2345 ctx_exit_on_error(&stream->encoder, "Failed to get encoder reference frame");
2346 vpx_codec_control(&stream->decoder, VP8_COPY_REFERENCE, &stream->ref_dec);
2347 ctx_exit_on_error(&stream->decoder, "Failed to get decoder reference frame");
2348
2349 if (!stream->mismatch_seen
2350 && !compare_img(&stream->ref_enc.img, &stream->ref_dec.img)) {
2351 /* TODO(jkoleszar): make fatal. */
2352 int y[2], u[2], v[2];
2353 find_mismatch(&stream->ref_enc.img, &stream->ref_dec.img,
2354 y, u, v);
2355 warn("Stream %d: Encode/decode mismatch on frame %d"
2356 " at Y[%d, %d], U[%d, %d], V[%d, %d]",
2357 stream->index, stream->frames_out,
2358 y[0], y[1], u[0], u[1], v[0], v[1]);
2359 stream->mismatch_seen = stream->frames_out;
2360 }
2361 }
2362
2363 int main(int argc, const char **argv_) {
2364 int pass;
2365 vpx_image_t raw;
2366 int frame_avail, got_data;
2367
2368 struct input_state input = {0};
2369 struct global_config global;
2370 struct stream_state *streams = NULL;
2371 char **argv, **argi;
2372 unsigned long cx_time = 0;
2373 int stream_cnt = 0;
2374
2375 exec_name = argv_[0];
2376
2377 if (argc < 3)
2378 usage_exit();
2379
2380 /* Setup default input stream settings */
2381 input.framerate.num = 30;
2382 input.framerate.den = 1;
2383 input.use_i420 = 1;
2384
2385 /* First parse the global configuration values, because we want to apply
2386 * other parameters on top of the default configuration provided by the
2387 * codec.
2388 */
2389 argv = argv_dup(argc - 1, argv_ + 1);
2390 parse_global_config(&global, argv);
2391
2392 {
2393 /* Now parse each stream's parameters. Using a local scope here
2394 * due to the use of 'stream' as loop variable in FOREACH_STREAM
2395 * loops
2396 */
2397 struct stream_state *stream = NULL;
2398
2399 do {
2400 stream = new_stream(&global, stream);
2401 stream_cnt++;
2402 if (!streams)
2403 streams = stream;
2404 } while (parse_stream_params(&global, stream, argv));
2405 }
2406
2407 /* Check for unrecognized options */
2408 for (argi = argv; *argi; argi++)
2409 if (argi[0][0] == '-' && argi[0][1])
2410 die("Error: Unrecognized option %s\n", *argi);
2411
2412 /* Handle non-option arguments */
2413 input.fn = argv[0];
2414
2415 if (!input.fn)
2416 usage_exit();
2417
2418 for (pass = global.pass ? global.pass - 1 : 0; pass < global.passes; pass++) {
2419 int frames_in = 0;
2420
2421 open_input_file(&input);
2422
2423 /* If the input file doesn't specify its w/h (raw files), try to get
2424 * the data from the first stream's configuration.
2425 */
2426 if (!input.w || !input.h)
2427 FOREACH_STREAM( {
2428 if (stream->config.cfg.g_w && stream->config.cfg.g_h) {
2429 input.w = stream->config.cfg.g_w;
2430 input.h = stream->config.cfg.g_h;
2431 break;
2432 }
2433 });
2434
2435 /* Update stream configurations from the input file's parameters */
2436 FOREACH_STREAM(set_stream_dimensions(stream, input.w, input.h));
2437 FOREACH_STREAM(validate_stream_config(stream));
2438
2439 /* Ensure that --passes and --pass are consistent. If --pass is set and
2440 * --passes=2, ensure --fpf was set.
2441 */
2442 if (global.pass && global.passes == 2)
2443 FOREACH_STREAM( {
2444 if (!stream->config.stats_fn)
2445 die("Stream %d: Must specify --fpf when --pass=%d"
2446 " and --passes=2\n", stream->index, global.pass);
2447 });
2448
2449
2450 /* Use the frame rate from the file only if none was specified
2451 * on the command-line.
2452 */
2453 if (!global.have_framerate)
2454 global.framerate = input.framerate;
2455
2456 FOREACH_STREAM(set_default_kf_interval(stream, &global));
2457
2458 /* Show configuration */
2459 if (global.verbose && pass == 0)
2460 FOREACH_STREAM(show_stream_config(stream, &global, &input));
2461
2462 if (pass == (global.pass ? global.pass - 1 : 0)) {
2463 if (input.file_type == FILE_TYPE_Y4M)
2464 /*The Y4M reader does its own allocation.
2465 Just initialize this here to avoid problems if we never read any
2466 frames.*/
2467 memset(&raw, 0, sizeof(raw));
2468 else
2469 vpx_img_alloc(&raw,
2470 input.use_i420 ? VPX_IMG_FMT_I420
2471 : VPX_IMG_FMT_YV12,
2472 input.w, input.h, 32);
2473
2474 FOREACH_STREAM(init_rate_histogram(&stream->rate_hist,
2475 &stream->config.cfg,
2476 &global.framerate));
2335 } 2477 }
2336 fprintf(stderr, "\n"); 2478
2337 } 2479 FOREACH_STREAM(open_output_file(stream, &global));
2338 2480 FOREACH_STREAM(setup_pass(stream, &global, pass));
2339 2481 FOREACH_STREAM(initialize_encoder(stream, &global));
2340 float usec_to_fps(uint64_t usec, unsigned int frames) 2482
2341 { 2483 frame_avail = 1;
2342 return (float)(usec > 0 ? frames * 1000000.0 / (float)usec : 0); 2484 got_data = 0;
2343 } 2485
2344 2486 while (frame_avail || got_data) {
2345 2487 struct vpx_usec_timer timer;
2346 int main(int argc, const char **argv_) 2488
2347 { 2489 if (!global.limit || frames_in < global.limit) {
2348 int pass; 2490 frame_avail = read_frame(&input, &raw);
2349 vpx_image_t raw; 2491
2350 int frame_avail, got_data; 2492 if (frame_avail)
2351 2493 frames_in++;
2352 struct input_state input = {0}; 2494
2353 struct global_config global; 2495 if (!global.quiet) {
2354 struct stream_state *streams = NULL; 2496 if (stream_cnt == 1)
2355 char **argv, **argi; 2497 fprintf(stderr,
2356 unsigned long cx_time = 0; 2498 "\rPass %d/%d frame %4d/%-4d %7"PRId64"B \033[K",
2357 int stream_cnt = 0; 2499 pass + 1, global.passes, frames_in,
2358 2500 streams->frames_out, (int64_t)streams->nbytes);
2359 exec_name = argv_[0]; 2501 else
2360 2502 fprintf(stderr,
2361 if (argc < 3) 2503 "\rPass %d/%d frame %4d %7lu %s (%.2f fps)\033[K",
2362 usage_exit(); 2504 pass + 1, global.passes, frames_in,
2363 2505 cx_time > 9999999 ? cx_time / 1000 : cx_time,
2364 /* Setup default input stream settings */ 2506 cx_time > 9999999 ? "ms" : "us",
2365 input.framerate.num = 30; 2507 usec_to_fps(cx_time, frames_in));
2366 input.framerate.den = 1; 2508 }
2367 input.use_i420 = 1; 2509
2368 2510 } else
2369 /* First parse the global configuration values, because we want to apply 2511 frame_avail = 0;
2370 * other parameters on top of the default configuration provided by the 2512
2371 * codec. 2513 if (frames_in > global.skip_frames) {
2372 */ 2514 vpx_usec_timer_start(&timer);
2373 argv = argv_dup(argc - 1, argv_ + 1); 2515 FOREACH_STREAM(encode_frame(stream, &global,
2374 parse_global_config(&global, argv); 2516 frame_avail ? &raw : NULL,
2375 2517 frames_in));
2376 { 2518 vpx_usec_timer_mark(&timer);
2377 /* Now parse each stream's parameters. Using a local scope here 2519 cx_time += (unsigned long)vpx_usec_timer_elapsed(&timer);
2378 * due to the use of 'stream' as loop variable in FOREACH_STREAM 2520
2379 * loops 2521 FOREACH_STREAM(update_quantizer_histogram(stream));
2380 */ 2522
2381 struct stream_state *stream = NULL; 2523 got_data = 0;
2382 2524 FOREACH_STREAM(get_cx_data(stream, &global, &got_data));
2383 do 2525
2384 { 2526 if (got_data && global.test_decode)
2385 stream = new_stream(&global, stream); 2527 FOREACH_STREAM(test_decode(stream));
2386 stream_cnt++; 2528 }
2387 if(!streams) 2529
2388 streams = stream; 2530 fflush(stdout);
2389 } while(parse_stream_params(&global, stream, argv));
2390 } 2531 }
2391 2532
2392 /* Check for unrecognized options */ 2533 if (stream_cnt > 1)
2393 for (argi = argv; *argi; argi++) 2534 fprintf(stderr, "\n");
2394 if (argi[0][0] == '-' && argi[0][1]) 2535
2395 die("Error: Unrecognized option %s\n", *argi); 2536 if (!global.quiet)
2396 2537 FOREACH_STREAM(fprintf(
2397 /* Handle non-option arguments */ 2538 stderr,
2398 input.fn = argv[0]; 2539 "\rPass %d/%d frame %4d/%-4d %7"PRId64"B %7lub/f %7"PRId6 4"b/s"
2399 2540 " %7"PRId64" %s (%.2f fps)\033[K\n", pass + 1,
2400 if (!input.fn) 2541 global.passes, frames_in, stream->frames_out, (int64_t)st ream->nbytes,
2401 usage_exit(); 2542 frames_in ? (unsigned long)(stream->nbytes * 8 / frames_i n) : 0,
2402 2543 frames_in ? (int64_t)stream->nbytes * 8
2403 for (pass = global.pass ? global.pass - 1 : 0; pass < global.passes; pass++) 2544 * (int64_t)global.framerate.num / global.framerate.den
2404 { 2545 / frames_in
2405 int frames_in = 0; 2546 : 0,
2406 2547 stream->cx_time > 9999999 ? stream->cx_time / 1000 : stre am->cx_time,
2407 open_input_file(&input); 2548 stream->cx_time > 9999999 ? "ms" : "us",
2408 2549 usec_to_fps(stream->cx_time, frames_in));
2409 /* If the input file doesn't specify its w/h (raw files), try to get 2550 );
2410 * the data from the first stream's configuration. 2551
2411 */ 2552 if (global.show_psnr)
2412 if(!input.w || !input.h) 2553 FOREACH_STREAM(show_psnr(stream));
2413 FOREACH_STREAM({ 2554
2414 if(stream->config.cfg.g_w && stream->config.cfg.g_h) 2555 FOREACH_STREAM(vpx_codec_destroy(&stream->encoder));
2415 { 2556
2416 input.w = stream->config.cfg.g_w; 2557 if (global.test_decode) {
2417 input.h = stream->config.cfg.g_h; 2558 FOREACH_STREAM(vpx_codec_destroy(&stream->decoder));
2418 break; 2559 FOREACH_STREAM(vpx_img_free(&stream->ref_enc.img));
2419 } 2560 FOREACH_STREAM(vpx_img_free(&stream->ref_dec.img));
2420 });
2421
2422 /* Update stream configurations from the input file's parameters */
2423 FOREACH_STREAM(set_stream_dimensions(stream, input.w, input.h));
2424 FOREACH_STREAM(validate_stream_config(stream));
2425
2426 /* Ensure that --passes and --pass are consistent. If --pass is set and
2427 * --passes=2, ensure --fpf was set.
2428 */
2429 if (global.pass && global.passes == 2)
2430 FOREACH_STREAM({
2431 if(!stream->config.stats_fn)
2432 die("Stream %d: Must specify --fpf when --pass=%d"
2433 " and --passes=2\n", stream->index, global.pass);
2434 });
2435
2436
2437 /* Use the frame rate from the file only if none was specified
2438 * on the command-line.
2439 */
2440 if (!global.have_framerate)
2441 global.framerate = input.framerate;
2442
2443 FOREACH_STREAM(set_default_kf_interval(stream, &global));
2444
2445 /* Show configuration */
2446 if (global.verbose && pass == 0)
2447 FOREACH_STREAM(show_stream_config(stream, &global, &input));
2448
2449 if(pass == (global.pass ? global.pass - 1 : 0)) {
2450 if (input.file_type == FILE_TYPE_Y4M)
2451 /*The Y4M reader does its own allocation.
2452 Just initialize this here to avoid problems if we never read a ny
2453 frames.*/
2454 memset(&raw, 0, sizeof(raw));
2455 else
2456 vpx_img_alloc(&raw,
2457 input.use_i420 ? VPX_IMG_FMT_I420
2458 : VPX_IMG_FMT_YV12,
2459 input.w, input.h, 32);
2460
2461 FOREACH_STREAM(init_rate_histogram(&stream->rate_hist,
2462 &stream->config.cfg,
2463 &global.framerate));
2464 }
2465
2466 FOREACH_STREAM(open_output_file(stream, &global));
2467 FOREACH_STREAM(setup_pass(stream, &global, pass));
2468 FOREACH_STREAM(initialize_encoder(stream, &global));
2469
2470 frame_avail = 1;
2471 got_data = 0;
2472
2473 while (frame_avail || got_data)
2474 {
2475 struct vpx_usec_timer timer;
2476
2477 if (!global.limit || frames_in < global.limit)
2478 {
2479 frame_avail = read_frame(&input, &raw);
2480
2481 if (frame_avail)
2482 frames_in++;
2483
2484 if (!global.quiet)
2485 {
2486 if(stream_cnt == 1)
2487 fprintf(stderr,
2488 "\rPass %d/%d frame %4d/%-4d %7"PRId64"B \033[K" ,
2489 pass + 1, global.passes, frames_in,
2490 streams->frames_out, (int64_t)streams->nbytes);
2491 else
2492 fprintf(stderr,
2493 "\rPass %d/%d frame %4d %7lu %s (%.2f fps)\033[K ",
2494 pass + 1, global.passes, frames_in,
2495 cx_time > 9999999 ? cx_time / 1000 : cx_time,
2496 cx_time > 9999999 ? "ms" : "us",
2497 usec_to_fps(cx_time, frames_in));
2498 }
2499
2500 }
2501 else
2502 frame_avail = 0;
2503
2504 vpx_usec_timer_start(&timer);
2505 FOREACH_STREAM(encode_frame(stream, &global,
2506 frame_avail ? &raw : NULL,
2507 frames_in));
2508 vpx_usec_timer_mark(&timer);
2509 cx_time += (unsigned long)vpx_usec_timer_elapsed(&timer);
2510
2511 FOREACH_STREAM(update_quantizer_histogram(stream));
2512
2513 got_data = 0;
2514 FOREACH_STREAM(get_cx_data(stream, &global, &got_data));
2515
2516 fflush(stdout);
2517 }
2518
2519 if(stream_cnt > 1)
2520 fprintf(stderr, "\n");
2521
2522 if (!global.quiet)
2523 FOREACH_STREAM(fprintf(
2524 stderr,
2525 "\rPass %d/%d frame %4d/%-4d %7"PRId64"B %7lub/f %7"PRId64"b/s"
2526 " %7"PRId64" %s (%.2f fps)\033[K\n", pass + 1,
2527 global.passes, frames_in, stream->frames_out, (int64_t)stream->n bytes,
2528 frames_in ? (unsigned long)(stream->nbytes * 8 / frames_in) : 0,
2529 frames_in ? (int64_t)stream->nbytes * 8
2530 * (int64_t)global.framerate.num / global.framerate.d en
2531 / frames_in
2532 : 0,
2533 stream->cx_time > 9999999 ? stream->cx_time / 1000 : stream->cx_ time,
2534 stream->cx_time > 9999999 ? "ms" : "us",
2535 usec_to_fps(stream->cx_time, frames_in));
2536 );
2537
2538 if (global.show_psnr)
2539 FOREACH_STREAM(show_psnr(stream));
2540
2541 FOREACH_STREAM(vpx_codec_destroy(&stream->encoder));
2542
2543 close_input_file(&input);
2544
2545 FOREACH_STREAM(close_output_file(stream, global.codec->fourcc));
2546
2547 FOREACH_STREAM(stats_close(&stream->stats, global.passes-1));
2548
2549 if (global.pass)
2550 break;
2551 } 2561 }
2552 2562
2553 if (global.show_q_hist_buckets) 2563 close_input_file(&input);
2554 FOREACH_STREAM(show_q_histogram(stream->counts, 2564
2555 global.show_q_hist_buckets)); 2565 FOREACH_STREAM(close_output_file(stream, global.codec->fourcc));
2556 2566
2557 if (global.show_rate_hist_buckets) 2567 FOREACH_STREAM(stats_close(&stream->stats, global.passes - 1));
2558 FOREACH_STREAM(show_rate_histogram(&stream->rate_hist, 2568
2559 &stream->config.cfg, 2569 if (global.pass)
2560 global.show_rate_hist_buckets)); 2570 break;
2561 FOREACH_STREAM(destroy_rate_histogram(&stream->rate_hist)); 2571 }
2562 2572
2563 vpx_img_free(&raw); 2573 if (global.show_q_hist_buckets)
2564 free(argv); 2574 FOREACH_STREAM(show_q_histogram(stream->counts,
2565 free(streams); 2575 global.show_q_hist_buckets));
2566 return EXIT_SUCCESS; 2576
2567 } 2577 if (global.show_rate_hist_buckets)
2578 FOREACH_STREAM(show_rate_histogram(&stream->rate_hist,
2579 &stream->config.cfg,
2580 global.show_rate_hist_buckets));
2581 FOREACH_STREAM(destroy_rate_histogram(&stream->rate_hist));
2582
2583 #if CONFIG_INTERNAL_STATS
2584 /* TODO(jkoleszar): This doesn't belong in this executable. Do it for now,
2585 * to match some existing utilities.
2586 */
2587 FOREACH_STREAM({
2588 FILE *f = fopen("opsnr.stt", "a");
2589 if (stream->mismatch_seen) {
2590 fprintf(f, "First mismatch occurred in frame %d\n",
2591 stream->mismatch_seen);
2592 } else {
2593 fprintf(f, "No mismatch detected in recon buffers\n");
2594 }
2595 fclose(f);
2596 });
2597 #endif
2598
2599 vpx_img_free(&raw);
2600 free(argv);
2601 free(streams);
2602 return EXIT_SUCCESS;
2603 }
OLDNEW
« libvpx.gyp ('K') | « source/libvpx/vpxdec.c ('k') | source/libvpx/y4minput.h » ('j') | no next file with comments »

Powered by Google App Engine
This is Rietveld 408576698