f8a751837e3e67fef57e3de9670e5a2defc6652a
1#include "cache.h"
2#include "diff.h"
3#include "diffcore.h"
4/*
6 * Idea here is very simple.
7 *
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).
12 *
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.
19 *
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.
23 *
24 * The length of the sequence is arbitrarily set to 8 for now.
25 */
26/* Wild guess at the initial hash size */
28#define INITIAL_HASH_SIZE 9
29#define HASHBASE 65537 /* next_prime(2^16) */
30/* We leave more room in smaller hash but do not let it
31 * grow to have unused hole too much.
32 */
33#define INITIAL_FREE(sz_log2) ((1<<(sz_log2))*(sz_log2-3)/(sz_log2))
34struct spanhash {
36 unsigned long hashval;
37 unsigned long cnt;
38};
39struct spanhash_top {
40 int alloc_log2;
41 int free;
42 struct spanhash data[FLEX_ARRAY];
43};
44static struct spanhash *spanhash_find(struct spanhash_top *top,
46 unsigned long hashval)
47{
48 int sz = 1 << top->alloc_log2;
49 int bucket = hashval & (sz - 1);
50 while (1) {
51 struct spanhash *h = &(top->data[bucket++]);
52 if (!h->cnt)
53 return NULL;
54 if (h->hashval == hashval)
55 return h;
56 if (sz <= bucket)
57 bucket = 0;
58 }
59}
60static struct spanhash_top *spanhash_rehash(struct spanhash_top *orig)
62{
63 struct spanhash_top *new;
64 int i;
65 int osz = 1 << orig->alloc_log2;
66 int sz = osz << 1;
67 new = xmalloc(sizeof(*orig) + sizeof(struct spanhash) * sz);
69 new->alloc_log2 = orig->alloc_log2 + 1;
70 new->free = INITIAL_FREE(new->alloc_log2);
71 memset(new->data, 0, sizeof(struct spanhash) * sz);
72 for (i = 0; i < osz; i++) {
73 struct spanhash *o = &(orig->data[i]);
74 int bucket;
75 if (!o->cnt)
76 continue;
77 bucket = o->hashval & (sz - 1);
78 while (1) {
79 struct spanhash *h = &(new->data[bucket++]);
80 if (!h->cnt) {
81 h->hashval = o->hashval;
82 h->cnt = o->cnt;
83 new->free--;
84 break;
85 }
86 if (sz <= bucket)
87 bucket = 0;
88 }
89 }
90 free(orig);
91 return new;
92}
93static struct spanhash_top *add_spanhash(struct spanhash_top *top,
95 unsigned long hashval)
96{
97 int bucket, lim;
98 struct spanhash *h;
99 lim = (1 << top->alloc_log2);
101 bucket = hashval & (lim - 1);
102 while (1) {
103 h = &(top->data[bucket++]);
104 if (!h->cnt) {
105 h->hashval = hashval;
106 h->cnt = 1;
107 top->free--;
108 if (top->free < 0)
109 return spanhash_rehash(top);
110 return top;
111 }
112 if (h->hashval == hashval) {
113 h->cnt++;
114 return top;
115 }
116 if (lim <= bucket)
117 bucket = 0;
118 }
119}
120static struct spanhash_top *hash_chars(unsigned char *buf, unsigned long sz)
122{
123 int i;
124 unsigned long accum1, accum2, hashval;
125 struct spanhash_top *hash;
126 i = INITIAL_HASH_SIZE;
128 hash = xmalloc(sizeof(*hash) + sizeof(struct spanhash) * (1<<i));
129 hash->alloc_log2 = i;
130 hash->free = INITIAL_FREE(i);
131 memset(hash->data, 0, sizeof(struct spanhash) * (1<<i));
132 /* an 8-byte shift register made of accum1 and accum2. New
134 * bytes come at LSB of accum2, and shifted up to accum1
135 */
136 for (i = accum1 = accum2 = 0; i < 7; i++, sz--) {
137 accum1 = (accum1 << 8) | (accum2 >> 24);
138 accum2 = (accum2 << 8) | *buf++;
139 }
140 while (sz) {
141 accum1 = (accum1 << 8) | (accum2 >> 24);
142 accum2 = (accum2 << 8) | *buf++;
143 hashval = (accum1 + accum2 * 0x61) % HASHBASE;
144 hash = add_spanhash(hash, hashval);
145 sz--;
146 }
147 return hash;
148}
149int diffcore_count_changes(void *src, unsigned long src_size,
151 void *dst, unsigned long dst_size,
152 void **src_count_p,
153 void **dst_count_p,
154 unsigned long delta_limit,
155 unsigned long *src_copied,
156 unsigned long *literal_added)
157{
158 int i, ssz;
159 struct spanhash_top *src_count, *dst_count;
160 unsigned long sc, la;
161 if (src_size < 8 || dst_size < 8)
163 return -1;
164 src_count = dst_count = NULL;
166 if (src_count_p)
167 src_count = *src_count_p;
168 if (!src_count) {
169 src_count = hash_chars(src, src_size);
170 if (src_count_p)
171 *src_count_p = src_count;
172 }
173 if (dst_count_p)
174 dst_count = *dst_count_p;
175 if (!dst_count) {
176 dst_count = hash_chars(dst, dst_size);
177 if (dst_count_p)
178 *dst_count_p = dst_count;
179 }
180 sc = la = 0;
181 ssz = 1 << src_count->alloc_log2;
183 for (i = 0; i < ssz; i++) {
184 struct spanhash *s = &(src_count->data[i]);
185 struct spanhash *d;
186 unsigned dst_cnt, src_cnt;
187 if (!s->cnt)
188 continue;
189 src_cnt = s->cnt;
190 d = spanhash_find(dst_count, s->hashval);
191 dst_cnt = d ? d->cnt : 0;
192 if (src_cnt < dst_cnt) {
193 la += dst_cnt - src_cnt;
194 sc += src_cnt;
195 }
196 else
197 sc += dst_cnt;
198 }
199 if (!src_count_p)
201 free(src_count);
202 if (!dst_count_p)
203 free(dst_count);
204 *src_copied = sc;
205 *literal_added = la;
206 return 0;
207}