Joey Armstrong | e8c091f | 2023-01-17 16:56:26 -0500 | [diff] [blame] | 1 | // Package xxhash implements the 64-bit variant of xxHash (XXH64) as described |
| 2 | // at http://cyan4973.github.io/xxHash/. |
| 3 | package xxhash |
| 4 | |
| 5 | import ( |
| 6 | "encoding/binary" |
| 7 | "errors" |
| 8 | "math/bits" |
| 9 | ) |
| 10 | |
| 11 | const ( |
| 12 | prime1 uint64 = 11400714785074694791 |
| 13 | prime2 uint64 = 14029467366897019727 |
| 14 | prime3 uint64 = 1609587929392839161 |
| 15 | prime4 uint64 = 9650029242287828579 |
| 16 | prime5 uint64 = 2870177450012600261 |
| 17 | ) |
| 18 | |
| 19 | // NOTE(caleb): I'm using both consts and vars of the primes. Using consts where |
| 20 | // possible in the Go code is worth a small (but measurable) performance boost |
| 21 | // by avoiding some MOVQs. Vars are needed for the asm and also are useful for |
| 22 | // convenience in the Go code in a few places where we need to intentionally |
| 23 | // avoid constant arithmetic (e.g., v1 := prime1 + prime2 fails because the |
| 24 | // result overflows a uint64). |
| 25 | var ( |
| 26 | prime1v = prime1 |
| 27 | prime2v = prime2 |
| 28 | prime3v = prime3 |
| 29 | prime4v = prime4 |
| 30 | prime5v = prime5 |
| 31 | ) |
| 32 | |
| 33 | // Digest implements hash.Hash64. |
| 34 | type Digest struct { |
| 35 | v1 uint64 |
| 36 | v2 uint64 |
| 37 | v3 uint64 |
| 38 | v4 uint64 |
| 39 | total uint64 |
| 40 | mem [32]byte |
| 41 | n int // how much of mem is used |
| 42 | } |
| 43 | |
| 44 | // New creates a new Digest that computes the 64-bit xxHash algorithm. |
| 45 | func New() *Digest { |
| 46 | var d Digest |
| 47 | d.Reset() |
| 48 | return &d |
| 49 | } |
| 50 | |
| 51 | // Reset clears the Digest's state so that it can be reused. |
| 52 | func (d *Digest) Reset() { |
| 53 | d.v1 = prime1v + prime2 |
| 54 | d.v2 = prime2 |
| 55 | d.v3 = 0 |
| 56 | d.v4 = -prime1v |
| 57 | d.total = 0 |
| 58 | d.n = 0 |
| 59 | } |
| 60 | |
| 61 | // Size always returns 8 bytes. |
| 62 | func (d *Digest) Size() int { return 8 } |
| 63 | |
| 64 | // BlockSize always returns 32 bytes. |
| 65 | func (d *Digest) BlockSize() int { return 32 } |
| 66 | |
| 67 | // Write adds more data to d. It always returns len(b), nil. |
| 68 | func (d *Digest) Write(b []byte) (n int, err error) { |
| 69 | n = len(b) |
| 70 | d.total += uint64(n) |
| 71 | |
| 72 | if d.n+n < 32 { |
| 73 | // This new data doesn't even fill the current block. |
| 74 | copy(d.mem[d.n:], b) |
| 75 | d.n += n |
| 76 | return |
| 77 | } |
| 78 | |
| 79 | if d.n > 0 { |
| 80 | // Finish off the partial block. |
| 81 | copy(d.mem[d.n:], b) |
| 82 | d.v1 = round(d.v1, u64(d.mem[0:8])) |
| 83 | d.v2 = round(d.v2, u64(d.mem[8:16])) |
| 84 | d.v3 = round(d.v3, u64(d.mem[16:24])) |
| 85 | d.v4 = round(d.v4, u64(d.mem[24:32])) |
| 86 | b = b[32-d.n:] |
| 87 | d.n = 0 |
| 88 | } |
| 89 | |
| 90 | if len(b) >= 32 { |
| 91 | // One or more full blocks left. |
| 92 | nw := writeBlocks(d, b) |
| 93 | b = b[nw:] |
| 94 | } |
| 95 | |
| 96 | // Store any remaining partial block. |
| 97 | copy(d.mem[:], b) |
| 98 | d.n = len(b) |
| 99 | |
| 100 | return |
| 101 | } |
| 102 | |
| 103 | // Sum appends the current hash to b and returns the resulting slice. |
| 104 | func (d *Digest) Sum(b []byte) []byte { |
| 105 | s := d.Sum64() |
| 106 | return append( |
| 107 | b, |
| 108 | byte(s>>56), |
| 109 | byte(s>>48), |
| 110 | byte(s>>40), |
| 111 | byte(s>>32), |
| 112 | byte(s>>24), |
| 113 | byte(s>>16), |
| 114 | byte(s>>8), |
| 115 | byte(s), |
| 116 | ) |
| 117 | } |
| 118 | |
| 119 | // Sum64 returns the current hash. |
| 120 | func (d *Digest) Sum64() uint64 { |
| 121 | var h uint64 |
| 122 | |
| 123 | if d.total >= 32 { |
| 124 | v1, v2, v3, v4 := d.v1, d.v2, d.v3, d.v4 |
| 125 | h = rol1(v1) + rol7(v2) + rol12(v3) + rol18(v4) |
| 126 | h = mergeRound(h, v1) |
| 127 | h = mergeRound(h, v2) |
| 128 | h = mergeRound(h, v3) |
| 129 | h = mergeRound(h, v4) |
| 130 | } else { |
| 131 | h = d.v3 + prime5 |
| 132 | } |
| 133 | |
| 134 | h += d.total |
| 135 | |
| 136 | i, end := 0, d.n |
| 137 | for ; i+8 <= end; i += 8 { |
| 138 | k1 := round(0, u64(d.mem[i:i+8])) |
| 139 | h ^= k1 |
| 140 | h = rol27(h)*prime1 + prime4 |
| 141 | } |
| 142 | if i+4 <= end { |
| 143 | h ^= uint64(u32(d.mem[i:i+4])) * prime1 |
| 144 | h = rol23(h)*prime2 + prime3 |
| 145 | i += 4 |
| 146 | } |
| 147 | for i < end { |
| 148 | h ^= uint64(d.mem[i]) * prime5 |
| 149 | h = rol11(h) * prime1 |
| 150 | i++ |
| 151 | } |
| 152 | |
| 153 | h ^= h >> 33 |
| 154 | h *= prime2 |
| 155 | h ^= h >> 29 |
| 156 | h *= prime3 |
| 157 | h ^= h >> 32 |
| 158 | |
| 159 | return h |
| 160 | } |
| 161 | |
| 162 | const ( |
| 163 | magic = "xxh\x06" |
| 164 | marshaledSize = len(magic) + 8*5 + 32 |
| 165 | ) |
| 166 | |
| 167 | // MarshalBinary implements the encoding.BinaryMarshaler interface. |
| 168 | func (d *Digest) MarshalBinary() ([]byte, error) { |
| 169 | b := make([]byte, 0, marshaledSize) |
| 170 | b = append(b, magic...) |
| 171 | b = appendUint64(b, d.v1) |
| 172 | b = appendUint64(b, d.v2) |
| 173 | b = appendUint64(b, d.v3) |
| 174 | b = appendUint64(b, d.v4) |
| 175 | b = appendUint64(b, d.total) |
| 176 | b = append(b, d.mem[:d.n]...) |
| 177 | b = b[:len(b)+len(d.mem)-d.n] |
| 178 | return b, nil |
| 179 | } |
| 180 | |
| 181 | // UnmarshalBinary implements the encoding.BinaryUnmarshaler interface. |
| 182 | func (d *Digest) UnmarshalBinary(b []byte) error { |
| 183 | if len(b) < len(magic) || string(b[:len(magic)]) != magic { |
| 184 | return errors.New("xxhash: invalid hash state identifier") |
| 185 | } |
| 186 | if len(b) != marshaledSize { |
| 187 | return errors.New("xxhash: invalid hash state size") |
| 188 | } |
| 189 | b = b[len(magic):] |
| 190 | b, d.v1 = consumeUint64(b) |
| 191 | b, d.v2 = consumeUint64(b) |
| 192 | b, d.v3 = consumeUint64(b) |
| 193 | b, d.v4 = consumeUint64(b) |
| 194 | b, d.total = consumeUint64(b) |
| 195 | copy(d.mem[:], b) |
| 196 | b = b[len(d.mem):] |
| 197 | d.n = int(d.total % uint64(len(d.mem))) |
| 198 | return nil |
| 199 | } |
| 200 | |
| 201 | func appendUint64(b []byte, x uint64) []byte { |
| 202 | var a [8]byte |
| 203 | binary.LittleEndian.PutUint64(a[:], x) |
| 204 | return append(b, a[:]...) |
| 205 | } |
| 206 | |
| 207 | func consumeUint64(b []byte) ([]byte, uint64) { |
| 208 | x := u64(b) |
| 209 | return b[8:], x |
| 210 | } |
| 211 | |
| 212 | func u64(b []byte) uint64 { return binary.LittleEndian.Uint64(b) } |
| 213 | func u32(b []byte) uint32 { return binary.LittleEndian.Uint32(b) } |
| 214 | |
| 215 | func round(acc, input uint64) uint64 { |
| 216 | acc += input * prime2 |
| 217 | acc = rol31(acc) |
| 218 | acc *= prime1 |
| 219 | return acc |
| 220 | } |
| 221 | |
| 222 | func mergeRound(acc, val uint64) uint64 { |
| 223 | val = round(0, val) |
| 224 | acc ^= val |
| 225 | acc = acc*prime1 + prime4 |
| 226 | return acc |
| 227 | } |
| 228 | |
| 229 | func rol1(x uint64) uint64 { return bits.RotateLeft64(x, 1) } |
| 230 | func rol7(x uint64) uint64 { return bits.RotateLeft64(x, 7) } |
| 231 | func rol11(x uint64) uint64 { return bits.RotateLeft64(x, 11) } |
| 232 | func rol12(x uint64) uint64 { return bits.RotateLeft64(x, 12) } |
| 233 | func rol18(x uint64) uint64 { return bits.RotateLeft64(x, 18) } |
| 234 | func rol23(x uint64) uint64 { return bits.RotateLeft64(x, 23) } |
| 235 | func rol27(x uint64) uint64 { return bits.RotateLeft64(x, 27) } |
| 236 | func rol31(x uint64) uint64 { return bits.RotateLeft64(x, 31) } |