summaryrefslogtreecommitdiff
path: root/vendor/github.com/klauspost/compress/flate/level6.go
blob: 00a3119776d4d0d0ff2a21d7e22e1e1f0369a5e3 (plain)
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
package flate

import "fmt"

type fastEncL6 struct {
	fastGen
	table  [tableSize]tableEntry
	bTable [tableSize]tableEntryPrev
}

func (e *fastEncL6) Encode(dst *tokens, src []byte) {
	const (
		inputMargin            = 12 - 1
		minNonLiteralBlockSize = 1 + 1 + inputMargin
	)
	if debugDecode && e.cur < 0 {
		panic(fmt.Sprint("e.cur < 0: ", e.cur))
	}

	// Protect against e.cur wraparound.
	for e.cur >= bufferReset {
		if len(e.hist) == 0 {
			for i := range e.table[:] {
				e.table[i] = tableEntry{}
			}
			for i := range e.bTable[:] {
				e.bTable[i] = tableEntryPrev{}
			}
			e.cur = maxMatchOffset
			break
		}
		// Shift down everything in the table that isn't already too far away.
		minOff := e.cur + int32(len(e.hist)) - maxMatchOffset
		for i := range e.table[:] {
			v := e.table[i].offset
			if v <= minOff {
				v = 0
			} else {
				v = v - e.cur + maxMatchOffset
			}
			e.table[i].offset = v
		}
		for i := range e.bTable[:] {
			v := e.bTable[i]
			if v.Cur.offset <= minOff {
				v.Cur.offset = 0
				v.Prev.offset = 0
			} else {
				v.Cur.offset = v.Cur.offset - e.cur + maxMatchOffset
				if v.Prev.offset <= minOff {
					v.Prev.offset = 0
				} else {
					v.Prev.offset = v.Prev.offset - e.cur + maxMatchOffset
				}
			}
			e.bTable[i] = v
		}
		e.cur = maxMatchOffset
	}

	s := e.addBlock(src)

	// This check isn't in the Snappy implementation, but there, the caller
	// instead of the callee handles this case.
	if len(src) < minNonLiteralBlockSize {
		// We do not fill the token table.
		// This will be picked up by caller.
		dst.n = uint16(len(src))
		return
	}

	// Override src
	src = e.hist
	nextEmit := s

	// sLimit is when to stop looking for offset/length copies. The inputMargin
	// lets us use a fast path for emitLiteral in the main loop, while we are
	// looking for copies.
	sLimit := int32(len(src) - inputMargin)

	// nextEmit is where in src the next emitLiteral should start from.
	cv := load6432(src, s)
	// Repeat MUST be > 1 and within range
	repeat := int32(1)
	for {
		const skipLog = 7
		const doEvery = 1

		nextS := s
		var l int32
		var t int32
		for {
			nextHashS := hash4x64(cv, tableBits)
			nextHashL := hash7(cv, tableBits)
			s = nextS
			nextS = s + doEvery + (s-nextEmit)>>skipLog
			if nextS > sLimit {
				goto emitRemainder
			}
			// Fetch a short+long candidate
			sCandidate := e.table[nextHashS]
			lCandidate := e.bTable[nextHashL]
			next := load6432(src, nextS)
			entry := tableEntry{offset: s + e.cur, val: uint32(cv)}
			e.table[nextHashS] = entry
			eLong := &e.bTable[nextHashL]
			eLong.Cur, eLong.Prev = entry, eLong.Cur

			// Calculate hashes of 'next'
			nextHashS = hash4x64(next, tableBits)
			nextHashL = hash7(next, tableBits)

			t = lCandidate.Cur.offset - e.cur
			if s-t < maxMatchOffset {
				if uint32(cv) == lCandidate.Cur.val {
					// Long candidate matches at least 4 bytes.

					// Store the next match
					e.table[nextHashS] = tableEntry{offset: nextS + e.cur, val: uint32(next)}
					eLong := &e.bTable[nextHashL]
					eLong.Cur, eLong.Prev = tableEntry{offset: nextS + e.cur, val: uint32(next)}, eLong.Cur

					// Check the previous long candidate as well.
					t2 := lCandidate.Prev.offset - e.cur
					if s-t2 < maxMatchOffset && uint32(cv) == lCandidate.Prev.val {
						l = e.matchlen(s+4, t+4, src) + 4
						ml1 := e.matchlen(s+4, t2+4, src) + 4
						if ml1 > l {
							t = t2
							l = ml1
							break
						}
					}
					break
				}
				// Current value did not match, but check if previous long value does.
				t = lCandidate.Prev.offset - e.cur
				if s-t < maxMatchOffset && uint32(cv) == lCandidate.Prev.val {
					// Store the next match
					e.table[nextHashS] = tableEntry{offset: nextS + e.cur, val: uint32(next)}
					eLong := &e.bTable[nextHashL]
					eLong.Cur, eLong.Prev = tableEntry{offset: nextS + e.cur, val: uint32(next)}, eLong.Cur
					break
				}
			}

			t = sCandidate.offset - e.cur
			if s-t < maxMatchOffset && uint32(cv) == sCandidate.val {
				// Found a 4 match...
				l = e.matchlen(s+4, t+4, src) + 4

				// Look up next long candidate (at nextS)
				lCandidate = e.bTable[nextHashL]

				// Store the next match
				e.table[nextHashS] = tableEntry{offset: nextS + e.cur, val: uint32(next)}
				eLong := &e.bTable[nextHashL]
				eLong.Cur, eLong.Prev = tableEntry{offset: nextS + e.cur, val: uint32(next)}, eLong.Cur

				// Check repeat at s + repOff
				const repOff = 1
				t2 := s - repeat + repOff
				if load3232(src, t2) == uint32(cv>>(8*repOff)) {
					ml := e.matchlen(s+4+repOff, t2+4, src) + 4
					if ml > l {
						t = t2
						l = ml
						s += repOff
						// Not worth checking more.
						break
					}
				}

				// If the next long is a candidate, use that...
				t2 = lCandidate.Cur.offset - e.cur
				if nextS-t2 < maxMatchOffset {
					if lCandidate.Cur.val == uint32(next) {
						ml := e.matchlen(nextS+4, t2+4, src) + 4
						if ml > l {
							t = t2
							s = nextS
							l = ml
							// This is ok, but check previous as well.
						}
					}
					// If the previous long is a candidate, use that...
					t2 = lCandidate.Prev.offset - e.cur
					if nextS-t2 < maxMatchOffset && lCandidate.Prev.val == uint32(next) {
						ml := e.matchlen(nextS+4, t2+4, src) + 4
						if ml > l {
							t = t2
							s = nextS
							l = ml
							break
						}
					}
				}
				break
			}
			cv = next
		}

		// A 4-byte match has been found. We'll later see if more than 4 bytes
		// match. But, prior to the match, src[nextEmit:s] are unmatched. Emit
		// them as literal bytes.

		// Extend the 4-byte match as long as possible.
		if l == 0 {
			l = e.matchlenLong(s+4, t+4, src) + 4
		} else if l == maxMatchLength {
			l += e.matchlenLong(s+l, t+l, src)
		}

		// Extend backwards
		for t > 0 && s > nextEmit && src[t-1] == src[s-1] {
			s--
			t--
			l++
		}
		if nextEmit < s {
			emitLiteral(dst, src[nextEmit:s])
		}
		if false {
			if t >= s {
				panic(fmt.Sprintln("s-t", s, t))
			}
			if (s - t) > maxMatchOffset {
				panic(fmt.Sprintln("mmo", s-t))
			}
			if l < baseMatchLength {
				panic("bml")
			}
		}

		dst.AddMatchLong(l, uint32(s-t-baseMatchOffset))
		repeat = s - t
		s += l
		nextEmit = s
		if nextS >= s {
			s = nextS + 1
		}

		if s >= sLimit {
			// Index after match end.
			for i := nextS + 1; i < int32(len(src))-8; i += 2 {
				cv := load6432(src, i)
				e.table[hash4x64(cv, tableBits)] = tableEntry{offset: i + e.cur, val: uint32(cv)}
				eLong := &e.bTable[hash7(cv, tableBits)]
				eLong.Cur, eLong.Prev = tableEntry{offset: i + e.cur, val: uint32(cv)}, eLong.Cur
			}
			goto emitRemainder
		}

		// Store every long hash in-between and every second short.
		if true {
			for i := nextS + 1; i < s-1; i += 2 {
				cv := load6432(src, i)
				t := tableEntry{offset: i + e.cur, val: uint32(cv)}
				t2 := tableEntry{offset: t.offset + 1, val: uint32(cv >> 8)}
				eLong := &e.bTable[hash7(cv, tableBits)]
				eLong2 := &e.bTable[hash7(cv>>8, tableBits)]
				e.table[hash4x64(cv, tableBits)] = t
				eLong.Cur, eLong.Prev = t, eLong.Cur
				eLong2.Cur, eLong2.Prev = t2, eLong2.Cur
			}
		}

		// We could immediately start working at s now, but to improve
		// compression we first update the hash table at s-1 and at s.
		cv = load6432(src, s)
	}

emitRemainder:
	if int(nextEmit) < len(src) {
		// If nothing was added, don't encode literals.
		if dst.n == 0 {
			return
		}

		emitLiteral(dst, src[nextEmit:])
	}
}