6 * Idea here is very simple.
8 * We have total of (sz-N+1) N-byte overlapping sequences in buf whose
9 * size is sz. If the same N-byte sequence appears in both source and
10 * destination, we say the byte that starts that sequence is shared
11 * between them (i.e. copied from source to destination).
13 * For each possible N-byte sequence, if the source buffer has more
14 * instances of it than the destination buffer, that means the
15 * difference are the number of bytes not copied from source to
16 * destination. If the counts are the same, everything was copied
17 * from source to destination. If the destination has more,
18 * everything was copied, and destination added more.
20 * We are doing an approximation so we do not really have to waste
21 * memory by actually storing the sequence. We just hash them into
22 * somewhere around 2^16 hashbuckets and count the occurrences.
24 * The length of the sequence is arbitrarily set to 8 for now.
27 /* Wild guess at the initial hash size */
28 #define INITIAL_HASH_SIZE 9
30 /* We leave more room in smaller hash but do not let it
31 * grow to have unused hole too much.
33 #define INITIAL_FREE(sz_log2) ((1<<(sz_log2))*(sz_log2-3)/(sz_log2))
35 /* A prime rather carefully chosen between 2^16..2^17, so that
36 * HASHBASE < INITIAL_FREE(17). We want to keep the maximum hashtable
37 * size under the current 2<<17 maximum, which can hold this many
38 * different values before overflowing to hashtable of size 2<<18.
40 #define HASHBASE 107927
43 unsigned long hashval;
49 struct spanhash data[FLEX_ARRAY];
52 static struct spanhash *spanhash_find(struct spanhash_top *top,
53 unsigned long hashval)
55 int sz = 1 << top->alloc_log2;
56 int bucket = hashval & (sz - 1);
58 struct spanhash *h = &(top->data[bucket++]);
61 if (h->hashval == hashval)
68 static struct spanhash_top *spanhash_rehash(struct spanhash_top *orig)
70 struct spanhash_top *new;
72 int osz = 1 << orig->alloc_log2;
75 new = xmalloc(sizeof(*orig) + sizeof(struct spanhash) * sz);
76 new->alloc_log2 = orig->alloc_log2 + 1;
77 new->free = INITIAL_FREE(new->alloc_log2);
78 memset(new->data, 0, sizeof(struct spanhash) * sz);
79 for (i = 0; i < osz; i++) {
80 struct spanhash *o = &(orig->data[i]);
84 bucket = o->hashval & (sz - 1);
86 struct spanhash *h = &(new->data[bucket++]);
88 h->hashval = o->hashval;
101 static struct spanhash_top *add_spanhash(struct spanhash_top *top,
102 unsigned long hashval)
107 lim = (1 << top->alloc_log2);
108 bucket = hashval & (lim - 1);
110 h = &(top->data[bucket++]);
112 h->hashval = hashval;
116 return spanhash_rehash(top);
119 if (h->hashval == hashval) {
128 static struct spanhash_top *hash_chars(unsigned char *buf, unsigned long sz)
131 unsigned long accum1, accum2, hashval;
132 struct spanhash_top *hash;
134 i = INITIAL_HASH_SIZE;
135 hash = xmalloc(sizeof(*hash) + sizeof(struct spanhash) * (1<<i));
136 hash->alloc_log2 = i;
137 hash->free = INITIAL_FREE(i);
138 memset(hash->data, 0, sizeof(struct spanhash) * (1<<i));
140 /* an 8-byte shift register made of accum1 and accum2. New
141 * bytes come at LSB of accum2, and shifted up to accum1
143 for (i = accum1 = accum2 = 0; i < 7; i++, sz--) {
144 accum1 = (accum1 << 8) | (accum2 >> 24);
145 accum2 = (accum2 << 8) | *buf++;
148 accum1 = (accum1 << 8) | (accum2 >> 24);
149 accum2 = (accum2 << 8) | *buf++;
150 hashval = (accum1 + accum2 * 0x61) % HASHBASE;
151 hash = add_spanhash(hash, hashval);
157 int diffcore_count_changes(void *src, unsigned long src_size,
158 void *dst, unsigned long dst_size,
161 unsigned long delta_limit,
162 unsigned long *src_copied,
163 unsigned long *literal_added)
166 struct spanhash_top *src_count, *dst_count;
167 unsigned long sc, la;
169 if (src_size < 8 || dst_size < 8)
172 src_count = dst_count = NULL;
174 src_count = *src_count_p;
176 src_count = hash_chars(src, src_size);
178 *src_count_p = src_count;
181 dst_count = *dst_count_p;
183 dst_count = hash_chars(dst, dst_size);
185 *dst_count_p = dst_count;
189 ssz = 1 << src_count->alloc_log2;
190 for (i = 0; i < ssz; i++) {
191 struct spanhash *s = &(src_count->data[i]);
193 unsigned dst_cnt, src_cnt;
197 d = spanhash_find(dst_count, s->hashval);
198 dst_cnt = d ? d->cnt : 0;
199 if (src_cnt < dst_cnt) {
200 la += dst_cnt - src_cnt;