4 #include "object-store.h"
6 #define DEFAULT_DELTA_CACHE_SIZE (256 * 1024 * 1024)
8 #define OE_DFS_STATE_BITS 2
9 #define OE_DEPTH_BITS 12
10 #define OE_IN_PACK_BITS 10
11 #define OE_Z_DELTA_BITS 20
13 * Note that oe_set_size() becomes expensive when the given size is
14 * above this limit. Don't lower it too much.
16 #define OE_SIZE_BITS 31
17 #define OE_DELTA_SIZE_BITS 20
20 * State flags for depth-first search used for analyzing delta cycles.
22 * The depth is measured in delta-links to the base (so if A is a delta
23 * against B, then A has a depth of 1, and B a depth of 0).
33 * The size of struct nearly determines pack-objects's memory
34 * consumption. This struct is packed tight for that reason. When you
35 * add or reorder something in this struct, think a bit about this.
39 * idx.oid is filled up before delta searching starts. idx.crc32 is
40 * only valid after the object is written out and will be used for
41 * generating the index. idx.offset will be both gradually set and
42 * used in writing phase (base objects get offset first, then deltas
45 * "size" is the uncompressed object size. Compressed size of the raw
46 * data for an object in a pack is not stored anywhere but is computed
47 * and made available when reverse .idx is made. Note that when a
48 * delta is reused, "size" is the uncompressed _delta_ size, not the
49 * canonical one after the delta has been applied.
51 * "hash" contains a path name hash which is used for sorting the
52 * delta list and also during delta searching. Once prepare_pack()
53 * returns it's no longer needed.
57 * The (in_pack, in_pack_offset) tuple contains the location of the
58 * object in the source pack. in_pack_header_size allows quickly
59 * skipping the header and going straight to the zlib stream.
61 * "type" and "in_pack_type" both describe object type. in_pack_type
62 * may contain a delta type, while type is always the canonical type.
66 * Delta links (delta, delta_child and delta_sibling) are created to
67 * reflect that delta graph from the source pack then updated or added
68 * during delta searching phase when we find better deltas.
70 * delta_child and delta_sibling are last needed in
71 * compute_write_order(). "delta" and "delta_size" must remain valid
72 * at object writing phase in case the delta is not cached.
74 * If a delta is cached in memory and is compressed, delta_data points
75 * to the data and z_delta_size contains the compressed size. If it's
76 * uncompressed [1], z_delta_size must be zero. delta_size is always
77 * the uncompressed size and must be valid even if the delta is not
80 * [1] during try_delta phase we don't bother with compressing because
81 * the delta could be quickly replaced with a better one.
84 struct pack_idx_entry idx
;
85 void *delta_data
; /* cached delta (uncompressed) */
87 uint32_t hash
; /* name hint hash */
88 unsigned size_
:OE_SIZE_BITS
;
89 unsigned size_valid
:1;
90 uint32_t delta_idx
; /* delta base object */
91 uint32_t delta_child_idx
; /* deltified objects who bases me */
92 uint32_t delta_sibling_idx
; /* other deltified objects who
93 * uses the same base as me
95 unsigned delta_size_
:OE_DELTA_SIZE_BITS
; /* delta data size (uncompressed) */
96 unsigned delta_size_valid
:1;
97 unsigned in_pack_idx
:OE_IN_PACK_BITS
; /* already in pack */
98 unsigned z_delta_size
:OE_Z_DELTA_BITS
;
99 unsigned type_valid
:1;
100 unsigned type_
:TYPE_BITS
;
101 unsigned no_try_delta
:1;
102 unsigned in_pack_type
:TYPE_BITS
; /* could be delta */
104 unsigned preferred_base
:1; /*
105 * we do not pack this, but is available
106 * to be used as the base object to delta
109 unsigned tagged
:1; /* near the very tip of refs */
110 unsigned filled
:1; /* assigned write-order */
111 unsigned dfs_state
:OE_DFS_STATE_BITS
;
112 unsigned char in_pack_header_size
;
113 unsigned depth
:OE_DEPTH_BITS
;
116 * pahole results on 64-bit linux (gcc and clang)
118 * size: 80, bit_padding: 20 bits, holes: 8 bits
120 * and on 32-bit (gcc)
122 * size: 76, bit_padding: 20 bits, holes: 8 bits
126 struct packing_data
{
127 struct object_entry
*objects
;
128 uint32_t nr_objects
, nr_alloc
;
133 unsigned int *in_pack_pos
;
136 * Only one of these can be non-NULL and they have different
137 * sizes. if in_pack_by_idx is allocated, oe_in_pack() returns
138 * the pack of an object using in_pack_idx field. If not,
139 * in_pack[] array is used the same way as in_pack_pos[]
141 struct packed_git
**in_pack_by_idx
;
142 struct packed_git
**in_pack
;
144 uintmax_t oe_size_limit
;
147 unsigned int *tree_depth
;
148 unsigned char *layer
;
151 void prepare_packing_data(struct packing_data
*pdata
);
152 struct object_entry
*packlist_alloc(struct packing_data
*pdata
,
153 const unsigned char *sha1
,
156 struct object_entry
*packlist_find(struct packing_data
*pdata
,
157 const unsigned char *sha1
,
158 uint32_t *index_pos
);
160 static inline uint32_t pack_name_hash(const char *name
)
162 uint32_t c
, hash
= 0;
168 * This effectively just creates a sortable number from the
169 * last sixteen non-whitespace characters. Last characters
170 * count "most", so things that end in ".c" sort together.
172 while ((c
= *name
++) != 0) {
175 hash
= (hash
>> 2) + (c
<< 24);
180 static inline enum object_type
oe_type(const struct object_entry
*e
)
182 return e
->type_valid
? e
->type_
: OBJ_BAD
;
185 static inline void oe_set_type(struct object_entry
*e
,
186 enum object_type type
)
189 BUG("OBJ_ANY cannot be set in pack-objects code");
191 e
->type_valid
= type
>= OBJ_NONE
;
192 e
->type_
= (unsigned)type
;
195 static inline unsigned int oe_in_pack_pos(const struct packing_data
*pack
,
196 const struct object_entry
*e
)
198 return pack
->in_pack_pos
[e
- pack
->objects
];
201 static inline void oe_set_in_pack_pos(const struct packing_data
*pack
,
202 const struct object_entry
*e
,
205 pack
->in_pack_pos
[e
- pack
->objects
] = pos
;
208 static inline struct packed_git
*oe_in_pack(const struct packing_data
*pack
,
209 const struct object_entry
*e
)
211 if (pack
->in_pack_by_idx
)
212 return pack
->in_pack_by_idx
[e
->in_pack_idx
];
214 return pack
->in_pack
[e
- pack
->objects
];
217 void oe_map_new_pack(struct packing_data
*pack
,
218 struct packed_git
*p
);
219 static inline void oe_set_in_pack(struct packing_data
*pack
,
220 struct object_entry
*e
,
221 struct packed_git
*p
)
224 oe_map_new_pack(pack
, p
);
225 if (pack
->in_pack_by_idx
)
226 e
->in_pack_idx
= p
->index
;
228 pack
->in_pack
[e
- pack
->objects
] = p
;
231 static inline struct object_entry
*oe_delta(
232 const struct packing_data
*pack
,
233 const struct object_entry
*e
)
236 return &pack
->objects
[e
->delta_idx
- 1];
240 static inline void oe_set_delta(struct packing_data
*pack
,
241 struct object_entry
*e
,
242 struct object_entry
*delta
)
245 e
->delta_idx
= (delta
- pack
->objects
) + 1;
250 static inline struct object_entry
*oe_delta_child(
251 const struct packing_data
*pack
,
252 const struct object_entry
*e
)
254 if (e
->delta_child_idx
)
255 return &pack
->objects
[e
->delta_child_idx
- 1];
259 static inline void oe_set_delta_child(struct packing_data
*pack
,
260 struct object_entry
*e
,
261 struct object_entry
*delta
)
264 e
->delta_child_idx
= (delta
- pack
->objects
) + 1;
266 e
->delta_child_idx
= 0;
269 static inline struct object_entry
*oe_delta_sibling(
270 const struct packing_data
*pack
,
271 const struct object_entry
*e
)
273 if (e
->delta_sibling_idx
)
274 return &pack
->objects
[e
->delta_sibling_idx
- 1];
278 static inline void oe_set_delta_sibling(struct packing_data
*pack
,
279 struct object_entry
*e
,
280 struct object_entry
*delta
)
283 e
->delta_sibling_idx
= (delta
- pack
->objects
) + 1;
285 e
->delta_sibling_idx
= 0;
288 unsigned long oe_get_size_slow(struct packing_data
*pack
,
289 const struct object_entry
*e
);
290 static inline unsigned long oe_size(struct packing_data
*pack
,
291 const struct object_entry
*e
)
296 return oe_get_size_slow(pack
, e
);
299 static inline int oe_size_less_than(struct packing_data
*pack
,
300 const struct object_entry
*lhs
,
304 return lhs
->size_
< rhs
;
305 if (rhs
< pack
->oe_size_limit
) /* rhs < 2^x <= lhs ? */
307 return oe_get_size_slow(pack
, lhs
) < rhs
;
310 static inline int oe_size_greater_than(struct packing_data
*pack
,
311 const struct object_entry
*lhs
,
315 return lhs
->size_
> rhs
;
316 if (rhs
< pack
->oe_size_limit
) /* rhs < 2^x <= lhs ? */
318 return oe_get_size_slow(pack
, lhs
) > rhs
;
321 static inline void oe_set_size(struct packing_data
*pack
,
322 struct object_entry
*e
,
325 if (size
< pack
->oe_size_limit
) {
330 if (oe_get_size_slow(pack
, e
) != size
)
331 BUG("'size' is supposed to be the object size!");
335 static inline unsigned long oe_delta_size(struct packing_data
*pack
,
336 const struct object_entry
*e
)
338 if (e
->delta_size_valid
)
339 return e
->delta_size_
;
340 return oe_size(pack
, e
);
343 static inline void oe_set_delta_size(struct packing_data
*pack
,
344 struct object_entry
*e
,
347 e
->delta_size_
= size
;
348 e
->delta_size_valid
= e
->delta_size_
== size
;
349 if (!e
->delta_size_valid
&& size
!= oe_size(pack
, e
))
350 BUG("this can only happen in check_object() "
351 "where delta size is the same as entry size");
354 static inline unsigned int oe_tree_depth(struct packing_data
*pack
,
355 struct object_entry
*e
)
357 if (!pack
->tree_depth
)
359 return pack
->tree_depth
[e
- pack
->objects
];
362 static inline void oe_set_tree_depth(struct packing_data
*pack
,
363 struct object_entry
*e
,
364 unsigned int tree_depth
)
366 if (!pack
->tree_depth
)
367 CALLOC_ARRAY(pack
->tree_depth
, pack
->nr_alloc
);
368 pack
->tree_depth
[e
- pack
->objects
] = tree_depth
;
371 static inline unsigned char oe_layer(struct packing_data
*pack
,
372 struct object_entry
*e
)
376 return pack
->layer
[e
- pack
->objects
];
379 static inline void oe_set_layer(struct packing_data
*pack
,
380 struct object_entry
*e
,
384 CALLOC_ARRAY(pack
->layer
, pack
->nr_alloc
);
385 pack
->layer
[e
- pack
->objects
] = layer
;