1 /* license:BSD-3-Clause
2 * copyright-holders:Aaron Giles
3 ****************************************************************************
7 Static Huffman compression and decompression helpers.
9 ****************************************************************************
11 Maximum codelength is officially (alphabetsize - 1). This would be 255 bits
12 (since we use 1 byte values). However, it is also dependent upon the number
13 of samples used, as follows:
15 2 bits -> 3..4 samples
16 3 bits -> 5..7 samples
17 4 bits -> 8..12 samples
18 5 bits -> 13..20 samples
19 6 bits -> 21..33 samples
20 7 bits -> 34..54 samples
21 8 bits -> 55..88 samples
22 9 bits -> 89..143 samples
23 10 bits -> 144..232 samples
24 11 bits -> 233..376 samples
25 12 bits -> 377..609 samples
26 13 bits -> 610..986 samples
27 14 bits -> 987..1596 samples
28 15 bits -> 1597..2583 samples
29 16 bits -> 2584..4180 samples -> note that a 4k data size guarantees codelength <= 16 bits
30 17 bits -> 4181..6764 samples
31 18 bits -> 6765..10945 samples
32 19 bits -> 10946..17710 samples
33 20 bits -> 17711..28656 samples
34 21 bits -> 28657..46367 samples
35 22 bits -> 46368..75024 samples
36 23 bits -> 75025..121392 samples
37 24 bits -> 121393..196417 samples
38 25 bits -> 196418..317810 samples
39 26 bits -> 317811..514228 samples
40 27 bits -> 514229..832039 samples
41 28 bits -> 832040..1346268 samples
42 29 bits -> 1346269..2178308 samples
43 30 bits -> 2178309..3524577 samples
44 31 bits -> 3524578..5702886 samples
45 32 bits -> 5702887..9227464 samples
47 Looking at it differently, here is where powers of 2 fall into these buckets:
49 256 samples -> 11 bits max
50 512 samples -> 12 bits max
51 1k samples -> 14 bits max
52 2k samples -> 15 bits max
53 4k samples -> 16 bits max
54 8k samples -> 18 bits max
55 16k samples -> 19 bits max
56 32k samples -> 21 bits max
57 64k samples -> 22 bits max
58 128k samples -> 24 bits max
59 256k samples -> 25 bits max
60 512k samples -> 27 bits max
61 1M samples -> 28 bits max
62 2M samples -> 29 bits max
63 4M samples -> 31 bits max
64 8M samples -> 32 bits max
66 ****************************************************************************
68 Delta-RLE encoding works as follows:
70 Starting value is assumed to be 0. All data is encoded as a delta
71 from the previous value, such that final[i] = final[i - 1] + delta.
72 Long runs of 0s are RLE-encoded as follows:
74 0x100 = repeat count of 8
75 0x101 = repeat count of 9
76 0x102 = repeat count of 10
77 0x103 = repeat count of 11
78 0x104 = repeat count of 12
79 0x105 = repeat count of 13
80 0x106 = repeat count of 14
81 0x107 = repeat count of 15
82 0x108 = repeat count of 16
83 0x109 = repeat count of 32
84 0x10a = repeat count of 64
85 0x10b = repeat count of 128
86 0x10c = repeat count of 256
87 0x10d = repeat count of 512
88 0x10e = repeat count of 1024
89 0x10f = repeat count of 2048
91 Note that repeat counts are reset at the end of a row, so if a 0 run
92 extends to the end of a row, a large repeat count may be used.
94 The reason for starting the run counts at 8 is that 0 is expected to
95 be the most common symbol, and is typically encoded in 1 or 2 bits.
97 ***************************************************************************/
103 #include <libchdr/huffman.h>
105 #define MAX(x,y) ((x) > (y) ? (x) : (y))
107 /***************************************************************************
109 ***************************************************************************
112 #define MAKE_LOOKUP(code,bits) (((code) << 5) | ((bits) & 0x1f))
114 /***************************************************************************
116 ***************************************************************************
119 /*-------------------------------------------------
120 * huffman_context_base - create an encoding/
122 *-------------------------------------------------
125 struct huffman_decoder* create_huffman_decoder(int numcodes, int maxbits)
127 struct huffman_decoder* decoder = NULL;
129 /* limit to 24 bits */
133 decoder = (struct huffman_decoder*)malloc(sizeof(struct huffman_decoder));
134 decoder->numcodes = numcodes;
135 decoder->maxbits = maxbits;
136 decoder->lookup = (lookup_value*)malloc(sizeof(lookup_value) * (1 << maxbits));
137 decoder->huffnode = (struct node_t*)malloc(sizeof(struct node_t) * numcodes);
138 decoder->datahisto = NULL;
139 decoder->prevdata = 0;
140 decoder->rleremaining = 0;
144 void delete_huffman_decoder(struct huffman_decoder* decoder)
148 if (decoder->lookup != NULL)
149 free(decoder->lookup);
150 if (decoder->huffnode != NULL)
151 free(decoder->huffnode);
156 /*-------------------------------------------------
157 * decode_one - decode a single code from the
159 *-------------------------------------------------
162 uint32_t huffman_decode_one(struct huffman_decoder* decoder, struct bitstream* bitbuf)
164 /* peek ahead to get maxbits worth of data */
165 uint32_t bits = bitstream_peek(bitbuf, decoder->maxbits);
167 /* look it up, then remove the actual number of bits for this code */
168 lookup_value lookup = decoder->lookup[bits];
169 bitstream_remove(bitbuf, lookup & 0x1f);
171 /* return the value */
175 /*-------------------------------------------------
176 * import_tree_rle - import an RLE-encoded
177 * huffman tree from a source data stream
178 *-------------------------------------------------
181 enum huffman_error huffman_import_tree_rle(struct huffman_decoder* decoder, struct bitstream* bitbuf)
185 enum huffman_error error;
187 /* bits per entry depends on the maxbits */
188 if (decoder->maxbits >= 16)
190 else if (decoder->maxbits >= 8)
195 /* loop until we read all the nodes */
196 for (curnode = 0; curnode < decoder->numcodes; )
198 /* a non-one value is just raw */
199 int nodebits = bitstream_read(bitbuf, numbits);
201 decoder->huffnode[curnode++].numbits = nodebits;
203 /* a one value is an escape code */
206 /* a double 1 is just a single 1 */
207 nodebits = bitstream_read(bitbuf, numbits);
209 decoder->huffnode[curnode++].numbits = nodebits;
211 /* otherwise, we need one for value for the repeat count */
214 int repcount = bitstream_read(bitbuf, numbits) + 3;
215 if (repcount + curnode > decoder->numcodes)
216 return HUFFERR_INVALID_DATA;
218 decoder->huffnode[curnode++].numbits = nodebits;
223 /* make sure we ended up with the right number */
224 if (curnode != decoder->numcodes)
225 return HUFFERR_INVALID_DATA;
227 /* assign canonical codes for all nodes based on their code lengths */
228 error = huffman_assign_canonical_codes(decoder);
229 if (error != HUFFERR_NONE)
232 /* build the lookup table */
233 huffman_build_lookup_table(decoder);
235 /* determine final input length and report errors */
236 return bitstream_overflow(bitbuf) ? HUFFERR_INPUT_BUFFER_TOO_SMALL : HUFFERR_NONE;
240 /*-------------------------------------------------
241 * import_tree_huffman - import a huffman-encoded
242 * huffman tree from a source data stream
243 *-------------------------------------------------
246 enum huffman_error huffman_import_tree_huffman(struct huffman_decoder* decoder, struct bitstream* bitbuf)
253 uint8_t rlefullbits = 0;
255 enum huffman_error error;
256 /* start by parsing the lengths for the small tree */
257 struct huffman_decoder* smallhuff = create_huffman_decoder(24, 6);
258 smallhuff->huffnode[0].numbits = bitstream_read(bitbuf, 3);
259 start = bitstream_read(bitbuf, 3) + 1;
260 for (index = 1; index < 24; index++)
262 if (index < start || count == 7)
263 smallhuff->huffnode[index].numbits = 0;
266 count = bitstream_read(bitbuf, 3);
267 smallhuff->huffnode[index].numbits = (count == 7) ? 0 : count;
271 /* then regenerate the tree */
272 error = huffman_assign_canonical_codes(smallhuff);
273 if (error != HUFFERR_NONE)
275 huffman_build_lookup_table(smallhuff);
277 /* determine the maximum length of an RLE count */
278 temp = decoder->numcodes - 9;
280 temp >>= 1, rlefullbits++;
282 /* now process the rest of the data */
283 for (curcode = 0; curcode < decoder->numcodes; )
285 int value = huffman_decode_one(smallhuff, bitbuf);
287 decoder->huffnode[curcode++].numbits = last = value - 1;
290 int count = bitstream_read(bitbuf, 3) + 2;
292 count += bitstream_read(bitbuf, rlefullbits);
293 for ( ; count != 0 && curcode < decoder->numcodes; count--)
294 decoder->huffnode[curcode++].numbits = last;
298 /* make sure we free the local huffman decoder */
299 delete_huffman_decoder(smallhuff);
301 /* make sure we ended up with the right number */
302 if (curcode != decoder->numcodes)
303 return HUFFERR_INVALID_DATA;
305 /* assign canonical codes for all nodes based on their code lengths */
306 error = huffman_assign_canonical_codes(decoder);
307 if (error != HUFFERR_NONE)
310 /* build the lookup table */
311 huffman_build_lookup_table(decoder);
313 /* determine final input length and report errors */
314 return bitstream_overflow(bitbuf) ? HUFFERR_INPUT_BUFFER_TOO_SMALL : HUFFERR_NONE;
317 /*-------------------------------------------------
318 * compute_tree_from_histo - common backend for
319 * computing a tree based on the data histogram
320 *-------------------------------------------------
323 enum huffman_error huffman_compute_tree_from_histo(struct huffman_decoder* decoder)
326 uint32_t lowerweight;
327 uint32_t upperweight;
328 /* compute the number of data items in the histogram */
329 uint32_t sdatacount = 0;
330 for (i = 0; i < decoder->numcodes; i++)
331 sdatacount += decoder->datahisto[i];
333 /* binary search to achieve the optimum encoding */
335 upperweight = sdatacount * 2;
338 /* build a tree using the current weight */
339 uint32_t curweight = (upperweight + lowerweight) / 2;
340 int curmaxbits = huffman_build_tree(decoder, sdatacount, curweight);
342 /* apply binary search here */
343 if (curmaxbits <= decoder->maxbits)
345 lowerweight = curweight;
347 /* early out if it worked with the raw weights, or if we're done searching */
348 if (curweight == sdatacount || (upperweight - lowerweight) <= 1)
352 upperweight = curweight;
355 /* assign canonical codes for all nodes based on their code lengths */
356 return huffman_assign_canonical_codes(decoder);
359 /***************************************************************************
361 ***************************************************************************
364 /*-------------------------------------------------
365 * tree_node_compare - compare two tree nodes
367 *-------------------------------------------------
370 static int huffman_tree_node_compare(const void *item1, const void *item2)
372 const struct node_t *node1 = *(const struct node_t **)item1;
373 const struct node_t *node2 = *(const struct node_t **)item2;
374 if (node2->weight != node1->weight)
375 return node2->weight - node1->weight;
376 if (node2->bits - node1->bits == 0)
377 fprintf(stderr, "identical node sort keys, should not happen!\n");
378 return (int)node1->bits - (int)node2->bits;
381 /*-------------------------------------------------
382 * build_tree - build a huffman tree based on the
384 *-------------------------------------------------
387 int huffman_build_tree(struct huffman_decoder* decoder, uint32_t totaldata, uint32_t totalweight)
393 /* make a list of all non-zero nodes */
394 struct node_t** list = (struct node_t**)malloc(sizeof(struct node_t*) * decoder->numcodes * 2);
395 memset(decoder->huffnode, 0, decoder->numcodes * sizeof(decoder->huffnode[0]));
396 for (curcode = 0; curcode < decoder->numcodes; curcode++)
397 if (decoder->datahisto[curcode] != 0)
399 list[listitems++] = &decoder->huffnode[curcode];
400 decoder->huffnode[curcode].count = decoder->datahisto[curcode];
401 decoder->huffnode[curcode].bits = curcode;
403 /* scale the weight by the current effective length, ensuring we don't go to 0 */
404 decoder->huffnode[curcode].weight = ((uint64_t)decoder->datahisto[curcode]) * ((uint64_t)totalweight) / ((uint64_t)totaldata);
405 if (decoder->huffnode[curcode].weight == 0)
406 decoder->huffnode[curcode].weight = 1;
410 fprintf(stderr, "Pre-sort:\n");
411 for (int i = 0; i < listitems; i++) {
412 fprintf(stderr, "weight: %d code: %d\n", list[i]->m_weight, list[i]->m_bits);
416 /* sort the list by weight, largest weight first */
417 qsort(&list[0], listitems, sizeof(list[0]), huffman_tree_node_compare);
420 fprintf(stderr, "Post-sort:\n");
421 for (int i = 0; i < listitems; i++) {
422 fprintf(stderr, "weight: %d code: %d\n", list[i]->m_weight, list[i]->m_bits);
424 fprintf(stderr, "===================\n");
427 /* now build the tree */
428 nextalloc = decoder->numcodes;
429 while (listitems > 1)
432 /* remove lowest two items */
433 struct node_t* node1 = &(*list[--listitems]);
434 struct node_t* node0 = &(*list[--listitems]);
436 /* create new node */
437 struct node_t* newnode = &decoder->huffnode[nextalloc++];
438 newnode->parent = NULL;
439 node0->parent = node1->parent = newnode;
440 newnode->weight = node0->weight + node1->weight;
442 /* insert into list at appropriate location */
443 for (curitem = 0; curitem < listitems; curitem++)
444 if (newnode->weight > list[curitem]->weight)
446 memmove(&list[curitem+1], &list[curitem], (listitems - curitem) * sizeof(list[0]));
449 list[curitem] = newnode;
453 /* compute the number of bits in each code, and fill in another histogram */
454 for (curcode = 0; curcode < decoder->numcodes; curcode++)
456 struct node_t *curnode;
457 struct node_t* node = &decoder->huffnode[curcode];
461 /* if we have a non-zero weight, compute the number of bits */
462 if (node->weight > 0)
464 /* determine the number of bits for this node */
465 for (curnode = node; curnode->parent != NULL; curnode = curnode->parent)
467 if (node->numbits == 0)
470 /* keep track of the max */
471 maxbits = MAX(maxbits, ((int)node->numbits));
477 /*-------------------------------------------------
478 * assign_canonical_codes - assign canonical codes
479 * to all the nodes based on the number of bits
481 *-------------------------------------------------
484 enum huffman_error huffman_assign_canonical_codes(struct huffman_decoder* decoder)
488 uint32_t curstart = 0;
489 /* build up a histogram of bit lengths */
490 uint32_t bithisto[33] = { 0 };
491 for (curcode = 0; curcode < decoder->numcodes; curcode++)
493 struct node_t* node = &decoder->huffnode[curcode];
494 if (node->numbits > decoder->maxbits)
495 return HUFFERR_INTERNAL_INCONSISTENCY;
496 if (node->numbits <= 32)
497 bithisto[node->numbits]++;
500 /* for each code length, determine the starting code number */
501 for (codelen = 32; codelen > 0; codelen--)
503 uint32_t nextstart = (curstart + bithisto[codelen]) >> 1;
504 if (codelen != 1 && nextstart * 2 != (curstart + bithisto[codelen]))
505 return HUFFERR_INTERNAL_INCONSISTENCY;
506 bithisto[codelen] = curstart;
507 curstart = nextstart;
510 /* now assign canonical codes */
511 for (curcode = 0; curcode < decoder->numcodes; curcode++)
513 struct node_t* node = &decoder->huffnode[curcode];
514 if (node->numbits > 0)
515 node->bits = bithisto[node->numbits]++;
520 /*-------------------------------------------------
521 * build_lookup_table - build a lookup table for
523 *-------------------------------------------------
526 void huffman_build_lookup_table(struct huffman_decoder* decoder)
529 /* iterate over all codes */
530 for (curcode = 0; curcode < decoder->numcodes; curcode++)
532 /* process all nodes which have non-zero bits */
533 struct node_t* node = &decoder->huffnode[curcode];
534 if (node->numbits > 0)
538 lookup_value *destend;
539 /* set up the entry */
540 lookup_value value = MAKE_LOOKUP(curcode, node->numbits);
542 /* fill all matching entries */
543 shift = decoder->maxbits - node->numbits;
544 dest = &decoder->lookup[node->bits << shift];
545 destend = &decoder->lookup[((node->bits + 1) << shift) - 1];
546 while (dest <= destend)