AOMedia AV1 Codec
inspect
1 /*
2  * Copyright (c) 2016, Alliance for Open Media. All rights reserved.
3  *
4  * This source code is subject to the terms of the BSD 2 Clause License and
5  * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
6  * was not distributed with this source code in the LICENSE file, you can
7  * obtain it at www.aomedia.org/license/software. If the Alliance for Open
8  * Media Patent License 1.0 was not distributed with this source code in the
9  * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
10  */
11 
12 // Inspect Decoder
13 // ================
14 //
15 // This is a simple decoder loop that writes JSON stats to stdout. This tool
16 // can also be compiled with Emscripten and used as a library.
17 
18 #include <stdio.h>
19 #include <stdlib.h>
20 #include <string.h>
21 
22 #ifdef __EMSCRIPTEN__
23 #include <emscripten.h>
24 #else
25 #define EMSCRIPTEN_KEEPALIVE
26 #endif
27 
28 #include "config/aom_config.h"
29 
30 #include "aom/aom_decoder.h"
31 #include "aom/aomdx.h"
32 #include "av1/common/av1_common_int.h"
33 
34 #if CONFIG_ACCOUNTING
35 #include "av1/decoder/accounting.h"
36 #endif
37 
38 #include "av1/decoder/inspection.h"
39 #include "common/args.h"
40 #include "common/tools_common.h"
41 #include "common/video_common.h"
42 #include "common/video_reader.h"
43 
44 // Max JSON buffer size.
45 const int MAX_BUFFER = 1024 * 1024 * 256;
46 
47 typedef enum {
48  ACCOUNTING_LAYER = 1,
49  BLOCK_SIZE_LAYER = 1 << 1,
50  TRANSFORM_SIZE_LAYER = 1 << 2,
51  TRANSFORM_TYPE_LAYER = 1 << 3,
52  MODE_LAYER = 1 << 4,
53  SKIP_LAYER = 1 << 5,
54  FILTER_LAYER = 1 << 6,
55  CDEF_LAYER = 1 << 7,
56  REFERENCE_FRAME_LAYER = 1 << 8,
57  MOTION_VECTORS_LAYER = 1 << 9,
58  UV_MODE_LAYER = 1 << 10,
59  CFL_LAYER = 1 << 11,
60  DUAL_FILTER_LAYER = 1 << 12,
61  Q_INDEX_LAYER = 1 << 13,
62  SEGMENT_ID_LAYER = 1 << 14,
63  MOTION_MODE_LAYER = 1 << 15,
64  COMPOUND_TYPE_LAYER = 1 << 16,
65  INTRABC_LAYER = 1 << 17,
66  PALETTE_LAYER = 1 << 18,
67  UV_PALETTE_LAYER = 1 << 19,
68  ALL_LAYERS = (1 << 20) - 1
69 } LayerType;
70 
71 static LayerType layers = 0;
72 
73 static int stop_after = 0;
74 static int compress = 0;
75 
76 static const arg_def_t limit_arg =
77  ARG_DEF(NULL, "limit", 1, "Stop decoding after n frames");
78 static const arg_def_t dump_all_arg = ARG_DEF("A", "all", 0, "Dump All");
79 static const arg_def_t compress_arg =
80  ARG_DEF("x", "compress", 0, "Compress JSON using RLE");
81 static const arg_def_t dump_accounting_arg =
82  ARG_DEF("a", "accounting", 0, "Dump Accounting");
83 static const arg_def_t dump_block_size_arg =
84  ARG_DEF("bs", "blockSize", 0, "Dump Block Size");
85 static const arg_def_t dump_motion_vectors_arg =
86  ARG_DEF("mv", "motionVectors", 0, "Dump Motion Vectors");
87 static const arg_def_t dump_transform_size_arg =
88  ARG_DEF("ts", "transformSize", 0, "Dump Transform Size");
89 static const arg_def_t dump_transform_type_arg =
90  ARG_DEF("tt", "transformType", 0, "Dump Transform Type");
91 static const arg_def_t dump_mode_arg = ARG_DEF("m", "mode", 0, "Dump Mode");
92 static const arg_def_t dump_motion_mode_arg =
93  ARG_DEF("mm", "motion_mode", 0, "Dump Motion Modes");
94 static const arg_def_t dump_compound_type_arg =
95  ARG_DEF("ct", "compound_type", 0, "Dump Compound Types");
96 static const arg_def_t dump_uv_mode_arg =
97  ARG_DEF("uvm", "uv_mode", 0, "Dump UV Intra Prediction Modes");
98 static const arg_def_t dump_skip_arg = ARG_DEF("s", "skip", 0, "Dump Skip");
99 static const arg_def_t dump_filter_arg =
100  ARG_DEF("f", "filter", 0, "Dump Filter");
101 static const arg_def_t dump_cdef_arg = ARG_DEF("c", "cdef", 0, "Dump CDEF");
102 static const arg_def_t dump_cfl_arg =
103  ARG_DEF("cfl", "chroma_from_luma", 0, "Dump Chroma from Luma Alphas");
104 static const arg_def_t dump_dual_filter_type_arg =
105  ARG_DEF("df", "dualFilterType", 0, "Dump Dual Filter Type");
106 static const arg_def_t dump_reference_frame_arg =
107  ARG_DEF("r", "referenceFrame", 0, "Dump Reference Frame");
108 static const arg_def_t dump_delta_q_arg =
109  ARG_DEF("dq", "delta_q", 0, "Dump QIndex");
110 static const arg_def_t dump_seg_id_arg =
111  ARG_DEF("si", "seg_id", 0, "Dump Segment ID");
112 static const arg_def_t dump_intrabc_arg =
113  ARG_DEF("ibc", "intrabc", 0, "Dump If IntraBC Is Used");
114 static const arg_def_t dump_palette_arg =
115  ARG_DEF("plt", "palette", 0, "Dump Palette Size");
116 static const arg_def_t dump_uv_palette_arg =
117  ARG_DEF("uvp", "uv_palette", 0, "Dump UV Palette Size");
118 static const arg_def_t usage_arg = ARG_DEF("h", "help", 0, "Help");
119 static const arg_def_t skip_non_transform_arg = ARG_DEF(
120  "snt", "skip_non_transform", 1, "Skip is counted as a non transform.");
121 static const arg_def_t combined_arg =
122  ARG_DEF("comb", "combined", 1, "combinining parameters into one output.");
123 
124 int combined_parm_list[15];
125 int combined_parm_count = 0;
126 
127 static const arg_def_t *main_args[] = { &limit_arg,
128  &dump_all_arg,
129  &compress_arg,
130 #if CONFIG_ACCOUNTING
131  &dump_accounting_arg,
132 #endif
133  &dump_block_size_arg,
134  &dump_transform_size_arg,
135  &dump_transform_type_arg,
136  &dump_mode_arg,
137  &dump_uv_mode_arg,
138  &dump_motion_mode_arg,
139  &dump_compound_type_arg,
140  &dump_skip_arg,
141  &dump_filter_arg,
142  &dump_cdef_arg,
143  &dump_dual_filter_type_arg,
144  &dump_cfl_arg,
145  &dump_reference_frame_arg,
146  &dump_motion_vectors_arg,
147  &dump_delta_q_arg,
148  &dump_seg_id_arg,
149  &dump_intrabc_arg,
150  &dump_palette_arg,
151  &dump_uv_palette_arg,
152  &usage_arg,
153  &skip_non_transform_arg,
154  &combined_arg,
155  NULL };
156 #define ENUM(name) \
157  { #name, name }
158 #define LAST_ENUM \
159  { NULL, 0 }
160 typedef struct map_entry {
161  const char *name;
162  int value;
163 } map_entry;
164 
165 const map_entry refs_map[] = {
166  ENUM(INTRA_FRAME), ENUM(LAST_FRAME), ENUM(LAST2_FRAME),
167  ENUM(LAST3_FRAME), ENUM(GOLDEN_FRAME), ENUM(BWDREF_FRAME),
168  ENUM(ALTREF2_FRAME), ENUM(ALTREF_FRAME), LAST_ENUM
169 };
170 
171 const map_entry block_size_map[] = {
172  ENUM(BLOCK_4X4), ENUM(BLOCK_4X8), ENUM(BLOCK_8X4),
173  ENUM(BLOCK_8X8), ENUM(BLOCK_8X16), ENUM(BLOCK_16X8),
174  ENUM(BLOCK_16X16), ENUM(BLOCK_16X32), ENUM(BLOCK_32X16),
175  ENUM(BLOCK_32X32), ENUM(BLOCK_32X64), ENUM(BLOCK_64X32),
176  ENUM(BLOCK_64X64), ENUM(BLOCK_64X128), ENUM(BLOCK_128X64),
177  ENUM(BLOCK_128X128), ENUM(BLOCK_4X16), ENUM(BLOCK_16X4),
178  ENUM(BLOCK_8X32), ENUM(BLOCK_32X8), ENUM(BLOCK_16X64),
179  ENUM(BLOCK_64X16), LAST_ENUM
180 };
181 
182 #define TX_SKIP -1
183 
184 const map_entry tx_size_map[] = {
185  ENUM(TX_4X4), ENUM(TX_8X8), ENUM(TX_16X16), ENUM(TX_32X32),
186  ENUM(TX_64X64), ENUM(TX_4X8), ENUM(TX_8X4), ENUM(TX_8X16),
187  ENUM(TX_16X8), ENUM(TX_16X32), ENUM(TX_32X16), ENUM(TX_32X64),
188  ENUM(TX_64X32), ENUM(TX_4X16), ENUM(TX_16X4), ENUM(TX_8X32),
189  ENUM(TX_32X8), ENUM(TX_16X64), ENUM(TX_64X16), LAST_ENUM
190 };
191 
192 const map_entry tx_type_map[] = { ENUM(DCT_DCT),
193  ENUM(ADST_DCT),
194  ENUM(DCT_ADST),
195  ENUM(ADST_ADST),
196  ENUM(FLIPADST_DCT),
197  ENUM(DCT_FLIPADST),
198  ENUM(FLIPADST_FLIPADST),
199  ENUM(ADST_FLIPADST),
200  ENUM(FLIPADST_ADST),
201  ENUM(IDTX),
202  ENUM(V_DCT),
203  ENUM(H_DCT),
204  ENUM(V_ADST),
205  ENUM(H_ADST),
206  ENUM(V_FLIPADST),
207  ENUM(H_FLIPADST),
208  LAST_ENUM };
209 const map_entry dual_filter_map[] = { ENUM(REG_REG), ENUM(REG_SMOOTH),
210  ENUM(REG_SHARP), ENUM(SMOOTH_REG),
211  ENUM(SMOOTH_SMOOTH), ENUM(SMOOTH_SHARP),
212  ENUM(SHARP_REG), ENUM(SHARP_SMOOTH),
213  ENUM(SHARP_SHARP), LAST_ENUM };
214 
215 const map_entry prediction_mode_map[] = {
216  ENUM(DC_PRED), ENUM(V_PRED), ENUM(H_PRED),
217  ENUM(D45_PRED), ENUM(D135_PRED), ENUM(D113_PRED),
218  ENUM(D157_PRED), ENUM(D203_PRED), ENUM(D67_PRED),
219  ENUM(SMOOTH_PRED), ENUM(SMOOTH_V_PRED), ENUM(SMOOTH_H_PRED),
220  ENUM(PAETH_PRED), ENUM(NEARESTMV), ENUM(NEARMV),
221  ENUM(GLOBALMV), ENUM(NEWMV), ENUM(NEAREST_NEARESTMV),
222  ENUM(NEAR_NEARMV), ENUM(NEAREST_NEWMV), ENUM(NEW_NEARESTMV),
223  ENUM(NEAR_NEWMV), ENUM(NEW_NEARMV), ENUM(GLOBAL_GLOBALMV),
224  ENUM(NEW_NEWMV), ENUM(INTRA_INVALID), LAST_ENUM
225 };
226 
227 const map_entry motion_mode_map[] = { ENUM(SIMPLE_TRANSLATION),
228  ENUM(OBMC_CAUSAL), // 2-sided OBMC
229  ENUM(WARPED_CAUSAL), // 2-sided WARPED
230  LAST_ENUM };
231 
232 const map_entry compound_type_map[] = { ENUM(COMPOUND_AVERAGE),
233  ENUM(COMPOUND_WEDGE),
234  ENUM(COMPOUND_DIFFWTD), LAST_ENUM };
235 
236 const map_entry uv_prediction_mode_map[] = {
237  ENUM(UV_DC_PRED), ENUM(UV_V_PRED),
238  ENUM(UV_H_PRED), ENUM(UV_D45_PRED),
239  ENUM(UV_D135_PRED), ENUM(UV_D113_PRED),
240  ENUM(UV_D157_PRED), ENUM(UV_D203_PRED),
241  ENUM(UV_D67_PRED), ENUM(UV_SMOOTH_PRED),
242  ENUM(UV_SMOOTH_V_PRED), ENUM(UV_SMOOTH_H_PRED),
243  ENUM(UV_PAETH_PRED), ENUM(UV_CFL_PRED),
244  ENUM(UV_MODE_INVALID), LAST_ENUM
245 };
246 #define NO_SKIP 0
247 #define SKIP 1
248 
249 const map_entry skip_map[] = { ENUM(SKIP), ENUM(NO_SKIP), LAST_ENUM };
250 
251 const map_entry intrabc_map[] = { { "INTRABC", 1 },
252  { "NO_INTRABC", 0 },
253  LAST_ENUM };
254 
255 const map_entry palette_map[] = {
256  { "ZERO_COLORS", 0 }, { "TWO_COLORS", 2 }, { "THREE_COLORS", 3 },
257  { "FOUR_COLORS", 4 }, { "FIVE_COLORS", 5 }, { "SIX_COLORS", 6 },
258  { "SEVEN_COLORS", 7 }, { "EIGHT_COLORS", 8 }, LAST_ENUM
259 };
260 
261 const map_entry config_map[] = { ENUM(MI_SIZE), LAST_ENUM };
262 
263 static const char *exec_name;
264 
265 struct parm_offset {
266  char parm[60];
267  char offset;
268 };
269 struct parm_offset parm_offsets[] = {
270  { "blockSize", offsetof(insp_mi_data, bsize) },
271  { "transformSize", offsetof(insp_mi_data, tx_size) },
272  { "transformType", offsetof(insp_mi_data, tx_type) },
273  { "dualFilterType", offsetof(insp_mi_data, dual_filter_type) },
274  { "mode", offsetof(insp_mi_data, mode) },
275  { "uv_mode", offsetof(insp_mi_data, uv_mode) },
276  { "motion_mode", offsetof(insp_mi_data, motion_mode) },
277  { "compound_type", offsetof(insp_mi_data, compound_type) },
278  { "referenceFrame", offsetof(insp_mi_data, ref_frame) },
279  { "skip", offsetof(insp_mi_data, skip) },
280 };
281 int parm_count = sizeof(parm_offsets) / sizeof(parm_offsets[0]);
282 
283 static int convert_to_indices(char *str, int *indices, int maxCount,
284  int *count) {
285  *count = 0;
286  do {
287  char *comma = strchr(str, ',');
288  int length = (comma ? (int)(comma - str) : (int)strlen(str));
289  int i;
290  for (i = 0; i < parm_count; ++i) {
291  if (!strncmp(str, parm_offsets[i].parm, length)) {
292  break;
293  }
294  }
295  if (i == parm_count) return 0;
296  indices[(*count)++] = i;
297  if (*count > maxCount) return 0;
298  str += length + 1;
299  } while (strlen(str) > 0);
300  return 1;
301 }
302 
303 insp_frame_data frame_data;
304 int frame_count = 0;
305 int decoded_frame_count = 0;
306 aom_codec_ctx_t codec;
307 AvxVideoReader *reader = NULL;
308 const AvxVideoInfo *info = NULL;
309 aom_image_t *img = NULL;
310 
311 static void on_frame_decoded_dump(char *json) {
312 #ifdef __EMSCRIPTEN__
313  EM_ASM_({ Module.on_frame_decoded_json($0); }, json);
314 #else
315  printf("%s", json);
316 #endif
317 }
318 
319 // Writing out the JSON buffer using snprintf is very slow, especially when
320 // compiled with emscripten, these functions speed things up quite a bit.
321 static int put_str(char *buffer, const char *str) {
322  int i;
323  for (i = 0; str[i] != '\0'; i++) {
324  buffer[i] = str[i];
325  }
326  return i;
327 }
328 
329 static int put_str_with_escape(char *buffer, const char *str) {
330  int i;
331  int j = 0;
332  for (i = 0; str[i] != '\0'; i++) {
333  if (str[i] < ' ') {
334  continue;
335  } else if (str[i] == '"' || str[i] == '\\') {
336  buffer[j++] = '\\';
337  }
338  buffer[j++] = str[i];
339  }
340  return j;
341 }
342 
343 static int put_num(char *buffer, char prefix, int num, char suffix) {
344  int i = 0;
345  char *buf = buffer;
346  int is_neg = 0;
347  if (prefix) {
348  buf[i++] = prefix;
349  }
350  if (num == 0) {
351  buf[i++] = '0';
352  } else {
353  if (num < 0) {
354  num = -num;
355  is_neg = 1;
356  }
357  int s = i;
358  while (num != 0) {
359  buf[i++] = '0' + (num % 10);
360  num = num / 10;
361  }
362  if (is_neg) {
363  buf[i++] = '-';
364  }
365  int e = i - 1;
366  while (s < e) {
367  int t = buf[s];
368  buf[s] = buf[e];
369  buf[e] = t;
370  s++;
371  e--;
372  }
373  }
374  if (suffix) {
375  buf[i++] = suffix;
376  }
377  return i;
378 }
379 
380 static int put_map(char *buffer, const map_entry *map) {
381  char *buf = buffer;
382  const map_entry *entry = map;
383  while (entry->name != NULL) {
384  *(buf++) = '"';
385  buf += put_str(buf, entry->name);
386  *(buf++) = '"';
387  buf += put_num(buf, ':', entry->value, 0);
388  entry++;
389  if (entry->name != NULL) {
390  *(buf++) = ',';
391  }
392  }
393  return (int)(buf - buffer);
394 }
395 
396 #if 0
397 static int put_reference_frame(char *buffer) {
398  const int mi_rows = frame_data.mi_rows;
399  const int mi_cols = frame_data.mi_cols;
400  char *buf = buffer;
401  int r, c, t;
402  buf += put_str(buf, " \"referenceFrameMap\": {");
403  buf += put_map(buf, refs_map);
404  buf += put_str(buf, "},\n");
405  buf += put_str(buf, " \"referenceFrame\": [");
406  for (r = 0; r < mi_rows; ++r) {
407  *(buf++) = '[';
408  for (c = 0; c < mi_cols; ++c) {
409  insp_mi_data *mi = &frame_data.mi_grid[r * mi_cols + c];
410  buf += put_num(buf, '[', mi->ref_frame[0], 0);
411  buf += put_num(buf, ',', mi->ref_frame[1], ']');
412  if (compress) { // RLE
413  for (t = c + 1; t < mi_cols; ++t) {
414  insp_mi_data *next_mi = &frame_data.mi_grid[r * mi_cols + t];
415  if (mi->ref_frame[0] != next_mi->ref_frame[0] ||
416  mi->ref_frame[1] != next_mi->ref_frame[1]) {
417  break;
418  }
419  }
420  if (t - c > 1) {
421  *(buf++) = ',';
422  buf += put_num(buf, '[', t - c - 1, ']');
423  c = t - 1;
424  }
425  }
426  if (c < mi_cols - 1) *(buf++) = ',';
427  }
428  *(buf++) = ']';
429  if (r < mi_rows - 1) *(buf++) = ',';
430  }
431  buf += put_str(buf, "],\n");
432  return (int)(buf - buffer);
433 }
434 #endif
435 
436 static int put_motion_vectors(char *buffer) {
437  const int mi_rows = frame_data.mi_rows;
438  const int mi_cols = frame_data.mi_cols;
439  char *buf = buffer;
440  int r, c, t;
441  buf += put_str(buf, " \"motionVectors\": [");
442  for (r = 0; r < mi_rows; ++r) {
443  *(buf++) = '[';
444  for (c = 0; c < mi_cols; ++c) {
445  insp_mi_data *mi = &frame_data.mi_grid[r * mi_cols + c];
446  buf += put_num(buf, '[', mi->mv[0].col, 0);
447  buf += put_num(buf, ',', mi->mv[0].row, 0);
448  buf += put_num(buf, ',', mi->mv[1].col, 0);
449  buf += put_num(buf, ',', mi->mv[1].row, ']');
450  if (compress) { // RLE
451  for (t = c + 1; t < mi_cols; ++t) {
452  insp_mi_data *next_mi = &frame_data.mi_grid[r * mi_cols + t];
453  if (mi->mv[0].col != next_mi->mv[0].col ||
454  mi->mv[0].row != next_mi->mv[0].row ||
455  mi->mv[1].col != next_mi->mv[1].col ||
456  mi->mv[1].row != next_mi->mv[1].row) {
457  break;
458  }
459  }
460  if (t - c > 1) {
461  *(buf++) = ',';
462  buf += put_num(buf, '[', t - c - 1, ']');
463  c = t - 1;
464  }
465  }
466  if (c < mi_cols - 1) *(buf++) = ',';
467  }
468  *(buf++) = ']';
469  if (r < mi_rows - 1) *(buf++) = ',';
470  }
471  buf += put_str(buf, "],\n");
472  return (int)(buf - buffer);
473 }
474 
475 static int put_combined(char *buffer) {
476  const int mi_rows = frame_data.mi_rows;
477  const int mi_cols = frame_data.mi_cols;
478  char *buf = buffer;
479  int r, c, p;
480  buf += put_str(buf, " \"");
481  for (p = 0; p < combined_parm_count; ++p) {
482  if (p) buf += put_str(buf, "&");
483  buf += put_str(buf, parm_offsets[combined_parm_list[p]].parm);
484  }
485  buf += put_str(buf, "\": [");
486  for (r = 0; r < mi_rows; ++r) {
487  *(buf++) = '[';
488  for (c = 0; c < mi_cols; ++c) {
489  insp_mi_data *mi = &frame_data.mi_grid[r * mi_cols + c];
490  *(buf++) = '[';
491  for (p = 0; p < combined_parm_count; ++p) {
492  if (p) *(buf++) = ',';
493  int16_t *v = (int16_t *)(((int8_t *)mi) +
494  parm_offsets[combined_parm_list[p]].offset);
495  buf += put_num(buf, 0, v[0], 0);
496  }
497  *(buf++) = ']';
498  if (c < mi_cols - 1) *(buf++) = ',';
499  }
500  *(buf++) = ']';
501  if (r < mi_rows - 1) *(buf++) = ',';
502  }
503  buf += put_str(buf, "],\n");
504  return (int)(buf - buffer);
505 }
506 
507 static int put_block_info(char *buffer, const map_entry *map, const char *name,
508  size_t offset, int len) {
509  const int mi_rows = frame_data.mi_rows;
510  const int mi_cols = frame_data.mi_cols;
511  char *buf = buffer;
512  int r, c, t, i;
513  if (compress && len == 1) {
514  die("Can't encode scalars as arrays when RLE compression is enabled.");
515  }
516  if (map) {
517  buf += snprintf(buf, MAX_BUFFER, " \"%sMap\": {", name);
518  buf += put_map(buf, map);
519  buf += put_str(buf, "},\n");
520  }
521  buf += snprintf(buf, MAX_BUFFER, " \"%s\": [", name);
522  for (r = 0; r < mi_rows; ++r) {
523  *(buf++) = '[';
524  for (c = 0; c < mi_cols; ++c) {
525  insp_mi_data *mi = &frame_data.mi_grid[r * mi_cols + c];
526  int16_t *v = (int16_t *)(((int8_t *)mi) + offset);
527  if (len == 0) {
528  buf += put_num(buf, 0, v[0], 0);
529  } else {
530  buf += put_str(buf, "[");
531  for (i = 0; i < len; i++) {
532  buf += put_num(buf, 0, v[i], 0);
533  if (i < len - 1) {
534  buf += put_str(buf, ",");
535  }
536  }
537  buf += put_str(buf, "]");
538  }
539  if (compress) { // RLE
540  for (t = c + 1; t < mi_cols; ++t) {
541  insp_mi_data *next_mi = &frame_data.mi_grid[r * mi_cols + t];
542  int16_t *nv = (int16_t *)(((int8_t *)next_mi) + offset);
543  int same = 0;
544  if (len == 0) {
545  same = v[0] == nv[0];
546  } else {
547  for (i = 0; i < len; i++) {
548  same = v[i] == nv[i];
549  if (!same) {
550  break;
551  }
552  }
553  }
554  if (!same) {
555  break;
556  }
557  }
558  if (t - c > 1) {
559  *(buf++) = ',';
560  buf += put_num(buf, '[', t - c - 1, ']');
561  c = t - 1;
562  }
563  }
564  if (c < mi_cols - 1) *(buf++) = ',';
565  }
566  *(buf++) = ']';
567  if (r < mi_rows - 1) *(buf++) = ',';
568  }
569  buf += put_str(buf, "],\n");
570  return (int)(buf - buffer);
571 }
572 
573 #if CONFIG_ACCOUNTING
574 static int put_accounting(char *buffer) {
575  char *buf = buffer;
576  int i;
577  const Accounting *accounting = frame_data.accounting;
578  if (accounting == NULL) {
579  printf("XXX\n");
580  return 0;
581  }
582  const int num_syms = accounting->syms.num_syms;
583  const int num_strs = accounting->syms.dictionary.num_strs;
584  buf += put_str(buf, " \"symbolsMap\": [");
585  for (i = 0; i < num_strs; i++) {
586  buf += snprintf(buf, MAX_BUFFER, "\"%s\"",
587  accounting->syms.dictionary.strs[i]);
588  if (i < num_strs - 1) *(buf++) = ',';
589  }
590  buf += put_str(buf, "],\n");
591  buf += put_str(buf, " \"symbols\": [\n ");
592  AccountingSymbolContext context;
593  context.x = -2;
594  context.y = -2;
595  AccountingSymbol *sym;
596  for (i = 0; i < num_syms; i++) {
597  sym = &accounting->syms.syms[i];
598  if (memcmp(&context, &sym->context, sizeof(AccountingSymbolContext)) != 0) {
599  buf += put_num(buf, '[', sym->context.x, 0);
600  buf += put_num(buf, ',', sym->context.y, ']');
601  } else {
602  buf += put_num(buf, '[', sym->id, 0);
603  buf += put_num(buf, ',', sym->bits, 0);
604  buf += put_num(buf, ',', sym->samples, ']');
605  }
606  context = sym->context;
607  if (i < num_syms - 1) *(buf++) = ',';
608  }
609  buf += put_str(buf, "],\n");
610  return (int)(buf - buffer);
611 }
612 #endif
613 
614 int skip_non_transform = 0;
615 
616 static void inspect(void *pbi, void *data) {
617  /* Fetch frame data. */
618  ifd_inspect(&frame_data, pbi, skip_non_transform);
619 
620  // Show existing frames just show a reference buffer we've already decoded.
621  // There's no information to show.
622  if (frame_data.show_existing_frame) return;
623 
624  (void)data;
625  // We allocate enough space and hope we don't write out of bounds. Totally
626  // unsafe but this speeds things up, especially when compiled to Javascript.
627  char *buffer = malloc(MAX_BUFFER);
628  if (!buffer) {
629  fprintf(stderr, "Error allocating inspect info buffer\n");
630  abort();
631  }
632  char *buf = buffer;
633  buf += put_str(buf, "{\n");
634  if (layers & BLOCK_SIZE_LAYER) {
635  buf += put_block_info(buf, block_size_map, "blockSize",
636  offsetof(insp_mi_data, bsize), 0);
637  }
638  if (layers & TRANSFORM_SIZE_LAYER) {
639  buf += put_block_info(buf, tx_size_map, "transformSize",
640  offsetof(insp_mi_data, tx_size), 0);
641  }
642  if (layers & TRANSFORM_TYPE_LAYER) {
643  buf += put_block_info(buf, tx_type_map, "transformType",
644  offsetof(insp_mi_data, tx_type), 0);
645  }
646  if (layers & DUAL_FILTER_LAYER) {
647  buf += put_block_info(buf, dual_filter_map, "dualFilterType",
648  offsetof(insp_mi_data, dual_filter_type), 0);
649  }
650  if (layers & MODE_LAYER) {
651  buf += put_block_info(buf, prediction_mode_map, "mode",
652  offsetof(insp_mi_data, mode), 0);
653  }
654  if (layers & UV_MODE_LAYER) {
655  buf += put_block_info(buf, uv_prediction_mode_map, "uv_mode",
656  offsetof(insp_mi_data, uv_mode), 0);
657  }
658  if (layers & MOTION_MODE_LAYER) {
659  buf += put_block_info(buf, motion_mode_map, "motion_mode",
660  offsetof(insp_mi_data, motion_mode), 0);
661  }
662  if (layers & COMPOUND_TYPE_LAYER) {
663  buf += put_block_info(buf, compound_type_map, "compound_type",
664  offsetof(insp_mi_data, compound_type), 0);
665  }
666  if (layers & SKIP_LAYER) {
667  buf +=
668  put_block_info(buf, skip_map, "skip", offsetof(insp_mi_data, skip), 0);
669  }
670  if (layers & FILTER_LAYER) {
671  buf +=
672  put_block_info(buf, NULL, "filter", offsetof(insp_mi_data, filter), 2);
673  }
674  if (layers & CDEF_LAYER) {
675  buf += put_block_info(buf, NULL, "cdef_level",
676  offsetof(insp_mi_data, cdef_level), 0);
677  buf += put_block_info(buf, NULL, "cdef_strength",
678  offsetof(insp_mi_data, cdef_strength), 0);
679  }
680  if (layers & CFL_LAYER) {
681  buf += put_block_info(buf, NULL, "cfl_alpha_idx",
682  offsetof(insp_mi_data, cfl_alpha_idx), 0);
683  buf += put_block_info(buf, NULL, "cfl_alpha_sign",
684  offsetof(insp_mi_data, cfl_alpha_sign), 0);
685  }
686  if (layers & Q_INDEX_LAYER) {
687  buf += put_block_info(buf, NULL, "delta_q",
688  offsetof(insp_mi_data, current_qindex), 0);
689  }
690  if (layers & SEGMENT_ID_LAYER) {
691  buf += put_block_info(buf, NULL, "seg_id",
692  offsetof(insp_mi_data, segment_id), 0);
693  }
694  if (layers & MOTION_VECTORS_LAYER) {
695  buf += put_motion_vectors(buf);
696  }
697  if (layers & INTRABC_LAYER) {
698  buf += put_block_info(buf, intrabc_map, "intrabc",
699  offsetof(insp_mi_data, intrabc), 0);
700  }
701  if (layers & PALETTE_LAYER) {
702  buf += put_block_info(buf, palette_map, "palette",
703  offsetof(insp_mi_data, palette), 0);
704  }
705  if (layers & UV_PALETTE_LAYER) {
706  buf += put_block_info(buf, palette_map, "uv_palette",
707  offsetof(insp_mi_data, uv_palette), 0);
708  }
709  if (combined_parm_count > 0) buf += put_combined(buf);
710  if (layers & REFERENCE_FRAME_LAYER) {
711  buf += put_block_info(buf, refs_map, "referenceFrame",
712  offsetof(insp_mi_data, ref_frame), 2);
713  }
714 #if CONFIG_ACCOUNTING
715  if (layers & ACCOUNTING_LAYER) {
716  buf += put_accounting(buf);
717  }
718 #endif
719  buf +=
720  snprintf(buf, MAX_BUFFER, " \"frame\": %d,\n", frame_data.frame_number);
721  buf += snprintf(buf, MAX_BUFFER, " \"showFrame\": %d,\n",
722  frame_data.show_frame);
723  buf += snprintf(buf, MAX_BUFFER, " \"frameType\": %d,\n",
724  frame_data.frame_type);
725  buf += snprintf(buf, MAX_BUFFER, " \"baseQIndex\": %d,\n",
726  frame_data.base_qindex);
727  buf += snprintf(buf, MAX_BUFFER, " \"tileCols\": %d,\n",
728  frame_data.tile_mi_cols);
729  buf += snprintf(buf, MAX_BUFFER, " \"tileRows\": %d,\n",
730  frame_data.tile_mi_rows);
731  buf += snprintf(buf, MAX_BUFFER, " \"deltaQPresentFlag\": %d,\n",
732  frame_data.delta_q_present_flag);
733  buf += snprintf(buf, MAX_BUFFER, " \"deltaQRes\": %d,\n",
734  frame_data.delta_q_res);
735  buf += put_str(buf, " \"config\": {");
736  buf += put_map(buf, config_map);
737  buf += put_str(buf, "},\n");
738  buf += put_str(buf, " \"configString\": \"");
739  buf += put_str_with_escape(buf, aom_codec_build_config());
740  buf += put_str(buf, "\"\n");
741  decoded_frame_count++;
742  buf += put_str(buf, "},\n");
743  *(buf++) = 0;
744  on_frame_decoded_dump(buffer);
745  free(buffer);
746 }
747 
748 static void ifd_init_cb(void) {
749  aom_inspect_init ii;
750  ii.inspect_cb = inspect;
751  ii.inspect_ctx = NULL;
753 }
754 
755 EMSCRIPTEN_KEEPALIVE int open_file(char *file);
756 
757 EMSCRIPTEN_KEEPALIVE
758 int open_file(char *file) {
759  if (file == NULL) {
760  // The JS analyzer puts the .ivf file at this location.
761  file = "/tmp/input.ivf";
762  }
763  reader = aom_video_reader_open(file);
764  if (!reader) die("Failed to open %s for reading.", file);
765  info = aom_video_reader_get_info(reader);
766  aom_codec_iface_t *decoder = get_aom_decoder_by_fourcc(info->codec_fourcc);
767  if (!decoder) die("Unknown input codec.");
768  fprintf(stderr, "Using %s\n", aom_codec_iface_name(decoder));
769  if (aom_codec_dec_init(&codec, decoder, NULL, 0))
770  die("Failed to initialize decoder.");
771  ifd_init(&frame_data, info->frame_width, info->frame_height);
772  ifd_init_cb();
773  return EXIT_SUCCESS;
774 }
775 
776 Av1DecodeReturn adr;
777 int have_frame = 0;
778 const unsigned char *frame;
779 const unsigned char *end_frame;
780 size_t frame_size = 0;
781 struct av1_ref_frame ref_dec;
782 
783 EMSCRIPTEN_KEEPALIVE int read_frame(void);
784 
785 EMSCRIPTEN_KEEPALIVE
786 int read_frame(void) {
787  img = NULL;
788 
789  // This loop skips over any frames that are show_existing_frames, as
790  // there is nothing to analyze.
791  do {
792  if (!have_frame) {
793  if (!aom_video_reader_read_frame(reader)) return EXIT_FAILURE;
794  frame = aom_video_reader_get_frame(reader, &frame_size);
795 
796  have_frame = 1;
797  end_frame = frame + frame_size;
798  }
799 
800  if (aom_codec_decode(&codec, frame, (unsigned int)frame_size, &adr) !=
801  AOM_CODEC_OK) {
802  die_codec(&codec, "Failed to decode frame.");
803  }
804 
805  frame = adr.buf;
806  frame_size = end_frame - frame;
807  if (frame == end_frame) have_frame = 0;
808  } while (adr.show_existing);
809 
810  int got_any_frames = 0;
811  aom_image_t *frame_img;
812  ref_dec.idx = adr.idx;
813 
814  // ref_dec.idx is the index to the reference buffer idx to AV1_GET_REFERENCE
815  // if its -1 the decoder didn't update any reference buffer and the only
816  // way to see the frame is aom_codec_get_frame.
817  if (ref_dec.idx == -1) {
818  aom_codec_iter_t iter = NULL;
819  img = frame_img = aom_codec_get_frame(&codec, &iter);
820  ++frame_count;
821  got_any_frames = 1;
822  } else if (!aom_codec_control(&codec, AV1_GET_REFERENCE, &ref_dec)) {
823  img = frame_img = &ref_dec.img;
824  ++frame_count;
825  got_any_frames = 1;
826  }
827  if (!got_any_frames) {
828  return EXIT_FAILURE;
829  }
830  return EXIT_SUCCESS;
831 }
832 
833 EMSCRIPTEN_KEEPALIVE const char *get_aom_codec_build_config(void);
834 
835 EMSCRIPTEN_KEEPALIVE
836 const char *get_aom_codec_build_config(void) {
837  return aom_codec_build_config();
838 }
839 
840 EMSCRIPTEN_KEEPALIVE int get_bit_depth(void);
841 
842 EMSCRIPTEN_KEEPALIVE
843 int get_bit_depth(void) { return img->bit_depth; }
844 
845 EMSCRIPTEN_KEEPALIVE int get_bits_per_sample(void);
846 
847 EMSCRIPTEN_KEEPALIVE
848 int get_bits_per_sample(void) { return img->bps; }
849 
850 EMSCRIPTEN_KEEPALIVE int get_image_format(void);
851 
852 EMSCRIPTEN_KEEPALIVE
853 int get_image_format(void) { return img->fmt; }
854 
855 EMSCRIPTEN_KEEPALIVE unsigned char *get_plane(int plane);
856 
857 EMSCRIPTEN_KEEPALIVE
858 unsigned char *get_plane(int plane) { return img->planes[plane]; }
859 
860 EMSCRIPTEN_KEEPALIVE int get_plane_stride(int plane);
861 
862 EMSCRIPTEN_KEEPALIVE
863 int get_plane_stride(int plane) { return img->stride[plane]; }
864 
865 EMSCRIPTEN_KEEPALIVE int get_plane_width(int plane);
866 
867 EMSCRIPTEN_KEEPALIVE
868 int get_plane_width(int plane) { return aom_img_plane_width(img, plane); }
869 
870 EMSCRIPTEN_KEEPALIVE int get_plane_height(int plane);
871 
872 EMSCRIPTEN_KEEPALIVE
873 int get_plane_height(int plane) { return aom_img_plane_height(img, plane); }
874 
875 EMSCRIPTEN_KEEPALIVE int get_frame_width(void);
876 
877 EMSCRIPTEN_KEEPALIVE
878 int get_frame_width(void) { return info->frame_width; }
879 
880 EMSCRIPTEN_KEEPALIVE int get_frame_height(void);
881 
882 EMSCRIPTEN_KEEPALIVE
883 int get_frame_height(void) { return info->frame_height; }
884 
885 static void parse_args(char **argv) {
886  char **argi, **argj;
887  struct arg arg;
888  (void)dump_accounting_arg;
889  (void)dump_cdef_arg;
890  for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
891  arg.argv_step = 1;
892  if (arg_match(&arg, &dump_block_size_arg, argi)) layers |= BLOCK_SIZE_LAYER;
893 #if CONFIG_ACCOUNTING
894  else if (arg_match(&arg, &dump_accounting_arg, argi))
895  layers |= ACCOUNTING_LAYER;
896 #endif
897  else if (arg_match(&arg, &dump_transform_size_arg, argi))
898  layers |= TRANSFORM_SIZE_LAYER;
899  else if (arg_match(&arg, &dump_transform_type_arg, argi))
900  layers |= TRANSFORM_TYPE_LAYER;
901  else if (arg_match(&arg, &dump_mode_arg, argi))
902  layers |= MODE_LAYER;
903  else if (arg_match(&arg, &dump_uv_mode_arg, argi))
904  layers |= UV_MODE_LAYER;
905  else if (arg_match(&arg, &dump_motion_mode_arg, argi))
906  layers |= MOTION_MODE_LAYER;
907  else if (arg_match(&arg, &dump_compound_type_arg, argi))
908  layers |= COMPOUND_TYPE_LAYER;
909  else if (arg_match(&arg, &dump_skip_arg, argi))
910  layers |= SKIP_LAYER;
911  else if (arg_match(&arg, &dump_filter_arg, argi))
912  layers |= FILTER_LAYER;
913  else if (arg_match(&arg, &dump_cdef_arg, argi))
914  layers |= CDEF_LAYER;
915  else if (arg_match(&arg, &dump_cfl_arg, argi))
916  layers |= CFL_LAYER;
917  else if (arg_match(&arg, &dump_reference_frame_arg, argi))
918  layers |= REFERENCE_FRAME_LAYER;
919  else if (arg_match(&arg, &dump_motion_vectors_arg, argi))
920  layers |= MOTION_VECTORS_LAYER;
921  else if (arg_match(&arg, &dump_dual_filter_type_arg, argi))
922  layers |= DUAL_FILTER_LAYER;
923  else if (arg_match(&arg, &dump_delta_q_arg, argi))
924  layers |= Q_INDEX_LAYER;
925  else if (arg_match(&arg, &dump_seg_id_arg, argi))
926  layers |= SEGMENT_ID_LAYER;
927  else if (arg_match(&arg, &dump_intrabc_arg, argi))
928  layers |= INTRABC_LAYER;
929  else if (arg_match(&arg, &dump_palette_arg, argi))
930  layers |= PALETTE_LAYER;
931  else if (arg_match(&arg, &dump_uv_palette_arg, argi))
932  layers |= UV_PALETTE_LAYER;
933  else if (arg_match(&arg, &dump_all_arg, argi))
934  layers |= ALL_LAYERS;
935  else if (arg_match(&arg, &compress_arg, argi))
936  compress = 1;
937  else if (arg_match(&arg, &usage_arg, argi))
938  usage_exit();
939  else if (arg_match(&arg, &limit_arg, argi))
940  stop_after = arg_parse_uint(&arg);
941  else if (arg_match(&arg, &skip_non_transform_arg, argi))
942  skip_non_transform = arg_parse_uint(&arg);
943  else if (arg_match(&arg, &combined_arg, argi))
944  convert_to_indices(
945  (char *)arg.val, combined_parm_list,
946  sizeof(combined_parm_list) / sizeof(combined_parm_list[0]),
947  &combined_parm_count);
948  else
949  argj++;
950  }
951 }
952 
953 static const char *exec_name;
954 
955 void usage_exit(void) {
956  fprintf(stderr, "Usage: %s src_filename <options>\n", exec_name);
957  fprintf(stderr, "\nOptions:\n");
958  arg_show_usage(stderr, main_args);
959  exit(EXIT_FAILURE);
960 }
961 
962 EMSCRIPTEN_KEEPALIVE
963 int main(int argc, char **argv) {
964  exec_name = argv[0];
965  parse_args(argv);
966  if (argc >= 2) {
967  open_file(argv[1]);
968  printf("[\n");
969  while (1) {
970  if (stop_after && (decoded_frame_count >= stop_after)) break;
971  if (read_frame()) break;
972  }
973  printf("null\n");
974  printf("]");
975  } else {
976  usage_exit();
977  }
978 }
979 
980 EMSCRIPTEN_KEEPALIVE void quit(void);
981 
982 EMSCRIPTEN_KEEPALIVE
983 void quit(void) {
984  if (aom_codec_destroy(&codec)) die_codec(&codec, "Failed to destroy codec");
985  aom_video_reader_close(reader);
986 }
987 
988 EMSCRIPTEN_KEEPALIVE void set_layers(LayerType v);
989 
990 EMSCRIPTEN_KEEPALIVE
991 void set_layers(LayerType v) { layers = v; }
992 
993 EMSCRIPTEN_KEEPALIVE void set_compress(int v);
994 
995 EMSCRIPTEN_KEEPALIVE
996 void set_compress(int v) { compress = v; }
Codec control function to get a pointer to a reference frame.
Definition: aom.h:51
Operation completed without error.
Definition: aom_codec.h:157
int show_existing
Definition: aomdx.h:84
aom_inspect_cb inspect_cb
Definition: aomdx.h:66
aom_image_t * aom_codec_get_frame(aom_codec_ctx_t *ctx, aom_codec_iter_t *iter)
Decoded frames iterator.
unsigned char * planes[3]
Definition: aom_image.h:215
int idx
Definition: aomdx.h:82
Codec context structure.
Definition: aom_codec.h:315
const unsigned char * buf
Definition: aomdx.h:80
Describes the decoder algorithm interface to applications.
Image Descriptor.
Definition: aom_image.h:182
aom_codec_err_t aom_codec_decode(aom_codec_ctx_t *ctx, const uint8_t *data, size_t data_sz, void *user_priv)
Decode data.
const struct aom_codec_iface aom_codec_iface_t
Codec interface structure.
Definition: aom_codec.h:271
#define aom_codec_dec_init(ctx, iface, cfg, flags)
Convenience macro for aom_codec_dec_init_ver()
Definition: aom_decoder.h:129
const char * aom_codec_iface_name(aom_codec_iface_t *iface)
Return the name for a given interface.
aom_codec_err_t aom_codec_destroy(aom_codec_ctx_t *ctx)
Destroy a codec instance.
Codec control function to set an aom_inspect_cb callback that is invoked each time a frame is decoded...
Definition: aomdx.h:382
int aom_img_plane_width(const aom_image_t *img, int plane)
Get the width of a plane.
void * inspect_ctx
Definition: aomdx.h:69
const char * aom_codec_build_config(void)
Return the build configuration.
AV1 specific reference frame data struct.
Definition: aom.h:89
int aom_img_plane_height(const aom_image_t *img, int plane)
Get the height of a plane.
Structure to hold inspection callback and context.
Definition: aomdx.h:64
const void * aom_codec_iter_t
Iterator.
Definition: aom_codec.h:305
int bps
Definition: aom_image.h:219
Provides definitions for using AOM or AV1 within the aom Decoder interface.
int stride[3]
Definition: aom_image.h:216
Structure to collect a buffer index when inspecting.
Definition: aomdx.h:78
unsigned int bit_depth
Definition: aom_image.h:194
aom_codec_err_t aom_codec_control(aom_codec_ctx_t *ctx, int ctrl_id,...)
Algorithm Control.
aom_img_fmt_t fmt
Definition: aom_image.h:183
struct Accounting Accounting
Definition: aomdx.h:50