-
Notifications
You must be signed in to change notification settings - Fork 0
/
allocator.go
372 lines (338 loc) · 10.5 KB
/
allocator.go
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
package radix
import (
"bufio"
"bytes"
"encoding/binary"
"fmt"
"io"
"strconv"
"sync/atomic"
)
// An Allocator is responsible for allocating new uint64 slots for storing tree nodes and manages the recycling of removed nodes.
// Since lookups are lock free an node might be recycled through an update operation while a reader is accessing it. To prevent the node
// from being used again until the read operation is done readers reserve the current epoch using the allocator.
// It's normally not needed to work directly with an allocator, but for bulk lookup/delete/update operations you will get
// better performance by acquiring an allocator and using it directly.
//
// a := trie.GetAllocator()
// for {
// v := a.Lookup(key)
// ...
// }
// trie.ReleaseAllocator(a)
//
// An allocator is not safe for concurrent use.
type Allocator struct {
idx *Tree
id int
empty bool
nextFree uint64
block int
blockSize uint64
offset uint64
recycler [][]recycledItem
newBlocks chan int
manager *epochManager
globalEpoch *uint64
reserveSlot *uint64
counter uint64
// the local item counter, could be negative
itemCounter int64
// cache padding
_ [8]uint64
}
type recycledItem struct {
n uint64
e uint64
}
func newAllocator(idx *Tree, id int, blocksize uint64, newBlocks chan int) *Allocator {
if id > idx.manager.n {
panic(fmt.Sprintf("ID %d overflows number of allocators in manager", id))
}
return &Allocator{
idx: idx,
manager: idx.manager,
id: id,
reserveSlot: idx.manager.reservations[id],
globalEpoch: idx.manager.globalEpoch,
empty: true,
nextFree: blocksize,
blockSize: blocksize,
recycler: make([][]recycledItem, 258),
newBlocks: newBlocks,
}
}
// FlushStats updates tree global stats with the allocators private counters.
func (a *Allocator) FlushStats() {
atomic.AddInt64(&a.idx.liveObjects, a.itemCounter)
a.itemCounter = 0
}
// unused reports the number of bytes unused in the current block
func (a *Allocator) unused() (bytes int) {
return int(a.blockSize - a.nextFree)
}
// Lookup searches the tree for a key and if a candidate leaf is found returns the value stored with it.
// The caller needs to verify that the candidate is equal to the lookup key, since if the key didn't exist, the candidate
// will be another key sharing a prefix with the lookup key.
func (a *Allocator) Lookup(key []byte) (value uint64, found bool) {
a.startOp()
value, found = a.idx.search(key)
a.endOp()
return
}
// PrepareUpdate prepares an update operation for the given key. It returns if it did find a key candidate
// and an UpdateOperation used to finalize or abort the operation.
// If not found it's a simple insert operation:
//
// found, op := a.PrepareUpdate([]byte("hello"))
// if !found {
// // this is an insert
// if inserted := op.Finalize(NEW_VALUE); !inserted {
// // write conflict, we need to restart with a new PrepareUpdate
// }
// }
//
// If a key candidate is found the caller must fetch the full key for op.Value into the op.FetchedKey field. The caller should compare this value
// to the update key (op.Key) and set op.Match accordingly before finalizing.
// Since operation structs are pooled it's best to reuse the slice at op.FetchedKey.
//
// if found {
// op.FetchedKey = fetchKeyFromSomeStorage(op.FetchedKey[:0], op.Value)
// if bytes.Equal(op.Key, op.FetchedKey) {
// // this is an update
// // do some update logic
// op.Match = true
// if updated := op.Finalize(NEW_VALUE); !updated {
// // write conflict, we need to restart with a new PrepareUpdate
// }
// } else {
// // this is an insert
// // do some insert logic
// if inserted := op.Finalize(NEW_VALUE); !inserted {
// // write conflict, we need to restart with a new PrepareUpdate
// }
// }
// }
//
func (a *Allocator) PrepareUpdate(key []byte) (found bool, op *UpdateOperation) {
op = newUpdateOperation(a.idx, a, false)
return op.prepareUpdate(key), op
}
// PrepareDelete prepares an delete operation on the given key. It returns if it did find a candidate key
// and an DeleteOperation used to finalize or abort the operation.
// If returns a nil op if and only if found is false. In this case there is nothing to delete.
//
// found, op := a.PreparDelete([]byte("hello"))
// if !found {
// // key didn't exist
// }
//
// If a candidate is found the caller must fetch the full key for op.Value into the op.FetchedKey field. The caller should compare this value
// to the update key (op.Key) to verify we are deleting the right key. If they don't match, call op.Abort.
// Since operation structs are pooled it's best to reuse the slice at op.FetchedKey.
//
// if found {
// op.FetchedKey = fetchKeyFromSomeStorage(op.FetchedKey[:0], op.Value)
// if bytes.Equal(op.Key, op.FetchedKey) {
// // it was a match, go ahead with delete
// if deleted := op.Finalize(); !deleted {
// // write conflict, we need to restart with a new PrepareDelete
// }
// } else {
// // wrong key, abort
// op.Abort()
// }
// }
//
func (a *Allocator) PrepareDelete(key []byte) (found bool, op *DeleteOperation) {
op = newDeleteOperation(a.idx, a, false)
if op.prepare(key) {
return true, op
}
op.Abort()
return false, nil
}
// startOp fetches the current global epoch and makes an allocator specific reservation with this epoch, meaning that only memory blocks recycled before this epoch are safe to reuse.
func (a *Allocator) startOp() {
atomic.StoreUint64(a.reserveSlot, atomic.LoadUint64(a.globalEpoch))
}
// endOp clears the reservation made by a previous call to startOp
func (a *Allocator) endOp() {
atomic.StoreUint64(a.reserveSlot, maxUint64)
}
func (a *Allocator) recycle(node uint64, slots int) {
if slots > len(a.recycler) {
return
}
e := atomic.LoadUint64(a.globalEpoch)
a.recycler[slots] = append(a.recycler[slots], recycledItem{node, e})
a.counter++
if a.counter >= 1000 {
atomic.AddUint64(a.globalEpoch, 1)
a.counter = 0
a.FlushStats()
}
//a.slotsReleased += uint64(slots)
}
func (a *Allocator) printStats(out io.Writer, i int) {
if a.empty {
fmt.Fprintf(out, "Allocator %d - EMPTY (blocksize %d)\n", i, a.blockSize)
} else {
fmt.Fprintf(out, "Allocator %d - current block %d (offset %d, blocksize %d), next free position at %d\n", i, a.block, a.offset, a.blockSize, a.nextFree)
}
for j, k := range a.recycler {
if len(k) > 0 || cap(k) > 0 {
fmt.Fprintf(out, " recycled size %d - %d entries (cap %d)\n", j, len(k), cap(k))
}
}
}
// newNode returns a node with room for the given number of slots, node is a global offset in 8 byte multiples
func (a *Allocator) newNode(slots int) (n uint64) {
minReserved := a.manager.minReservation()
for t := slots; t < len(a.recycler); t++ {
l := len(a.recycler[t])
if l > 0 {
for i, item := range a.recycler[t] {
if item.e < minReserved {
// pick this item
n = item.n
a.recycler[t][i] = a.recycler[t][l-1]
a.recycler[t] = a.recycler[t][:l-1]
unusedSlots := t - slots
if unusedSlots >= 2 {
n2 := n + uint64(slots)
a.recycler[unusedSlots] = append(a.recycler[unusedSlots], recycledItem{n2, item.e})
//a.slotsReleased += uint64(unusedSlots)
}
//a.slotsReused += uint64(slots)
return n
}
}
}
// skip the slots+1 size, since this would leave us with 1 lost unused slot
if t == slots {
t++
}
}
return a.allocateNode(slots)
}
// allocateNode allocates a new block of memory and returns it's global offset counted in 8 byte multiples
func (a *Allocator) allocateNode(slots int) (n uint64) {
// a.offset and a.nextFree are still in bytes
n = (a.offset + a.nextFree) >> 3
a.nextFree += uint64(slots) * 8
if a.nextFree >= blockSize {
a.block = <-a.newBlocks
a.offset = uint64(a.block) * blockSize
// avoid the 0 slot in block 0 !
if a.offset == 0 {
a.offset = 8
}
n = a.offset >> 3
a.nextFree = uint64(slots) * 8
a.empty = false
}
//a.slotsAllocated += uint64(slots)
return
}
func (a *Allocator) loadState(reader *bufio.Reader) error {
// reset recyclers
for i := range a.recycler {
a.recycler[i] = a.recycler[i][:0]
}
for {
line, err := reader.ReadBytes('\n')
if err != nil {
return err
}
// an single \n means allocator section is over
if len(line) == 1 {
return nil
}
space := bytes.IndexByte(line, ' ')
if space < 0 {
return fmt.Errorf("Corrupted state file, ignoring bad line %v", line)
}
// trim \n !
values := line[space+1 : len(line)-1]
v, _ := strconv.ParseUint(string(values), 10, 64)
switch string(line[:space]) {
case "empty":
a.empty = bytes.Equal(line[space+1:], []byte("true"))
case "next_free":
a.nextFree = v
case "block":
a.block = int(v)
case "block_size":
a.blockSize = v
case "offset":
a.offset = v
case "recycler":
// recycler <size> <count>
space = bytes.IndexByte(values, ' ')
if space < 0 {
return fmt.Errorf("Corrupted state file, ignoring bad line %v", line)
}
size, _ := strconv.Atoi(string(values[:space]))
count, _ := strconv.Atoi(string(values[space+1:]))
if count <= cap(a.recycler[size]) {
a.recycler[size] = a.recycler[size][0:count]
} else {
a.recycler[size] = make([]recycledItem, count)
}
var buf [8]byte
for i := 0; i < count; i++ {
_, err := io.ReadFull(reader, buf[:])
if err != nil {
return err
}
a.recycler[size][i] = recycledItem{binary.LittleEndian.Uint64(buf[:]), 0}
}
}
}
}
func (a *Allocator) writeState(out io.Writer) (int, error) {
written := 0
// write metadata: blocksize, number of blocks, root node location
n, err := out.Write([]byte(fmt.Sprintf("empty %v\nnext_free %d\nblock %d\nblock_size %d\noffset %d\n", a.empty, a.nextFree, a.block, a.blockSize, a.offset)))
if err != nil {
return n, err
}
written += n
var buf []byte
// write recyclers
for i := range a.recycler {
// skip empty lists
if len(a.recycler[i]) == 0 {
continue
}
n, err := out.Write([]byte(fmt.Sprintf("recycler %d %d\n", i, len(a.recycler[i]))))
if err != nil {
return n, err
}
written += n
sz := 8 * len(a.recycler[i])
if cap(buf) < sz {
buf = make([]byte, sz)
} else {
buf = buf[:sz]
}
o := 0
for _, item := range a.recycler[i] {
binary.LittleEndian.PutUint64(buf[o:o+8], item.n)
o += 8
}
n, err = out.Write(buf)
if err != nil {
return n, err
}
written += n
}
// a single \n to denote end of allocator section
n, err = out.Write([]byte("\n"))
if err != nil {
return n, err
}
written += n
return written, nil
}