7 * The author dedicates this file to the public domain.
8 * You can do whatever you want with this file.
10 * ----------------------------------------------------------------------------
12 * This is a Binary Trees (bt) based matchfinder.
14 * The data structure is a hash table where each hash bucket contains a binary
15 * tree of sequences whose first 3 bytes share the same hash code. Each
16 * sequence is identified by its starting position in the input buffer. Each
17 * binary tree is always sorted such that each left child represents a sequence
18 * lexicographically lesser than its parent and each right child represents a
19 * sequence lexicographically greater than its parent.
21 * The algorithm processes the input buffer sequentially. At each byte
22 * position, the hash code of the first 3 bytes of the sequence beginning at
23 * that position (the sequence being matched against) is computed. This
24 * identifies the hash bucket to use for that position. Then, a new binary tree
25 * node is created to represent the current sequence. Then, in a single tree
26 * traversal, the hash bucket's binary tree is searched for matches and is
27 * re-rooted at the new node.
29 * Compared to the simpler algorithm that uses linked lists instead of binary
30 * trees (see hc_matchfinder.h), the binary tree version gains more information
31 * at each node visitation. Ideally, the binary tree version will examine only
32 * 'log(n)' nodes to find the same matches that the linked list version will
33 * find by examining 'n' nodes. In addition, the binary tree version can
34 * examine fewer bytes at each node by taking advantage of the common prefixes
35 * that result from the sort order, whereas the linked list version may have to
36 * examine up to the full length of the match at each node.
38 * However, it is not always best to use the binary tree version. It requires
39 * nearly twice as much memory as the linked list version, and it takes time to
40 * keep the binary trees sorted, even at positions where the compressor does not
41 * need matches. Generally, when doing fast compression on small buffers,
42 * binary trees are the wrong approach. They are best suited for thorough
43 * compression and/or large buffers.
45 * ----------------------------------------------------------------------------
48 #ifndef _BT_MATCHFINDER_H
49 #define _BT_MATCHFINDER_H
51 #ifndef MATCHFINDER_MAX_WINDOW_ORDER
52 # error "MATCHFINDER_MAX_WINDOW_ORDER must be defined!"
57 #include "wimlib/lz_extend.h"
58 #include "wimlib/lz_hash.h"
60 #if MATCHFINDER_MAX_WINDOW_ORDER < 13
61 # define BT_MATCHFINDER_HASH_ORDER 14
62 #elif MATCHFINDER_MAX_WINDOW_ORDER < 15
63 # define BT_MATCHFINDER_HASH_ORDER 15
65 # define BT_MATCHFINDER_HASH_ORDER 16
68 #if MATCHFINDER_MAX_WINDOW_ORDER <= 16
74 /* Representation of a match found by the bt_matchfinder */
77 /* The number of bytes matched. */
80 /* The offset back from the current position that was matched. */
84 struct bt_matchfinder {
85 pos_t hash_tab[1UL << BT_MATCHFINDER_HASH_ORDER];
89 /* Return the number of bytes that must be allocated for a 'bt_matchfinder' that
90 * can work with buffers up to the specified size. */
92 bt_matchfinder_size(size_t max_bufsize)
94 return sizeof(struct bt_matchfinder) + (2 * max_bufsize * sizeof(pos_t));
97 /* Prepare the matchfinder for a new input buffer. */
99 bt_matchfinder_init(struct bt_matchfinder *mf)
101 memset(mf, 0, sizeof(*mf));
105 bt_matchfinder_hash_3_bytes(const u8 *in_next)
107 return lz_hash_3_bytes(in_next, BT_MATCHFINDER_HASH_ORDER);
110 static inline pos_t *
111 bt_child(struct bt_matchfinder *mf, pos_t node, int offset)
113 if (MATCHFINDER_MAX_WINDOW_ORDER < sizeof(pos_t) * 8) {
115 return &mf->child_tab[(node << 1) + offset];
117 return &mf->child_tab[((size_t)node << 1) + offset];
121 static inline pos_t *
122 bt_left_child(struct bt_matchfinder *mf, pos_t node)
124 return bt_child(mf, node, 0);
127 static inline pos_t *
128 bt_right_child(struct bt_matchfinder *mf, pos_t node)
130 return bt_child(mf, node, 1);
134 * Retrieve a list of matches with the current position.
137 * The matchfinder structure.
139 * Pointer to the beginning of the input buffer.
141 * Pointer to the next byte in the input buffer to process. This is the
142 * pointer to the sequence being matched against.
144 * Only record matches that are at least this long.
146 * The maximum permissible match length at this position.
148 * Stop searching if a match of at least this length is found.
149 * Must be <= @max_len.
151 * Limit on the number of potential matches to consider. Must be >= 1.
153 * Pointer to the hash code for the current sequence, which was computed
154 * one position in advance so that the binary tree root could be
155 * prefetched. This is an input/output parameter.
157 * The length of the longest match found is written here. (This is
158 * actually redundant with the 'struct lz_match' array, but this is easier
159 * for the compiler to optimize when inlined and the caller immediately
160 * does a check against 'best_len'.)
162 * An array in which this function will record the matches. The recorded
163 * matches will be sorted by strictly increasing length and strictly
164 * increasing offset. The maximum number of matches that may be found is
165 * 'min(nice_len, max_len) - 3 + 1'.
167 * The return value is a pointer to the next available slot in the @lz_matchptr
168 * array. (If no matches were found, this will be the same as @lz_matchptr.)
170 static inline struct lz_match *
171 bt_matchfinder_get_matches(struct bt_matchfinder * const restrict mf,
172 const u8 * const in_begin,
173 const u8 * const in_next,
174 const unsigned min_len,
175 const unsigned max_len,
176 const unsigned nice_len,
177 const unsigned max_search_depth,
178 u32 * restrict next_hash,
179 unsigned * restrict best_len_ret,
180 struct lz_match * restrict lz_matchptr)
182 unsigned depth_remaining = max_search_depth;
186 pos_t *pending_lt_ptr, *pending_gt_ptr;
187 unsigned best_lt_len, best_gt_len;
189 unsigned best_len = min_len - 1;
191 if (unlikely(max_len < LZ_HASH3_REQUIRED_NBYTES + 1)) {
192 *best_len_ret = best_len;
197 *next_hash = bt_matchfinder_hash_3_bytes(in_next + 1);
198 cur_node = mf->hash_tab[hash];
199 mf->hash_tab[hash] = in_next - in_begin;
200 prefetchw(&mf->hash_tab[*next_hash]);
202 pending_lt_ptr = bt_left_child(mf, in_next - in_begin);
203 pending_gt_ptr = bt_right_child(mf, in_next - in_begin);
211 *best_len_ret = best_len;
216 matchptr = &in_begin[cur_node];
218 if (matchptr[len] == in_next[len]) {
219 len = lz_extend(in_next, matchptr, len + 1, max_len);
220 if (len > best_len) {
222 lz_matchptr->length = len;
223 lz_matchptr->offset = in_next - matchptr;
225 if (len >= nice_len) {
226 *pending_lt_ptr = *bt_left_child(mf, cur_node);
227 *pending_gt_ptr = *bt_right_child(mf, cur_node);
228 *best_len_ret = best_len;
234 if (matchptr[len] < in_next[len]) {
235 *pending_lt_ptr = cur_node;
236 pending_lt_ptr = bt_right_child(mf, cur_node);
237 cur_node = *pending_lt_ptr;
239 if (best_gt_len < len)
242 *pending_gt_ptr = cur_node;
243 pending_gt_ptr = bt_left_child(mf, cur_node);
244 cur_node = *pending_gt_ptr;
246 if (best_lt_len < len)
250 if (!cur_node || !--depth_remaining) {
253 *best_len_ret = best_len;
260 * Advance the matchfinder, but don't record any matches.
263 * The matchfinder structure.
265 * Pointer to the beginning of the input buffer.
267 * Pointer to the next byte in the input buffer to process.
269 * Pointer to the end of the input buffer.
271 * Stop searching if a match of at least this length is found.
273 * Limit on the number of potential matches to consider.
275 * Pointer to the hash code for the current sequence, which was computed
276 * one position in advance so that the binary tree root could be
277 * prefetched. This is an input/output parameter.
279 * Note: this is very similar to bt_matchfinder_get_matches() because both
280 * functions must do hashing and tree re-rooting. This version just doesn't
281 * actually record any matches.
284 bt_matchfinder_skip_position(struct bt_matchfinder * const restrict mf,
285 const u8 * const in_begin,
286 const u8 * const in_next,
287 const u8 * const in_end,
288 const unsigned nice_len,
289 const unsigned max_search_depth,
290 u32 * restrict next_hash)
292 unsigned depth_remaining = max_search_depth;
296 pos_t *pending_lt_ptr, *pending_gt_ptr;
297 unsigned best_lt_len, best_gt_len;
300 if (unlikely(in_end - in_next < LZ_HASH3_REQUIRED_NBYTES + 1))
304 *next_hash = bt_matchfinder_hash_3_bytes(in_next + 1);
305 cur_node = mf->hash_tab[hash];
306 mf->hash_tab[hash] = in_next - in_begin;
307 prefetchw(&mf->hash_tab[*next_hash]);
309 depth_remaining = max_search_depth;
310 pending_lt_ptr = bt_left_child(mf, in_next - in_begin);
311 pending_gt_ptr = bt_right_child(mf, in_next - in_begin);
323 matchptr = &in_begin[cur_node];
325 if (matchptr[len] == in_next[len]) {
326 len = lz_extend(in_next, matchptr, len + 1, nice_len);
327 if (len == nice_len) {
328 *pending_lt_ptr = *bt_left_child(mf, cur_node);
329 *pending_gt_ptr = *bt_right_child(mf, cur_node);
334 if (matchptr[len] < in_next[len]) {
335 *pending_lt_ptr = cur_node;
336 pending_lt_ptr = bt_right_child(mf, cur_node);
337 cur_node = *pending_lt_ptr;
339 if (best_gt_len < len)
342 *pending_gt_ptr = cur_node;
343 pending_gt_ptr = bt_left_child(mf, cur_node);
344 cur_node = *pending_gt_ptr;
346 if (best_lt_len < len)
350 if (!cur_node || !--depth_remaining) {
358 #endif /* _BT_MATCHFINDER_H */