Rafael Espindola | eaeb6d9 | 2016-09-27 15:45:57 +0000 | [diff] [blame] | 1 | /* |
| 2 | * xxHash - Fast Hash algorithm |
| 3 | * Copyright (C) 2012-2016, Yann Collet |
| 4 | * |
| 5 | * BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php) |
| 6 | * |
| 7 | * Redistribution and use in source and binary forms, with or without |
| 8 | * modification, are permitted provided that the following conditions are |
| 9 | * met: |
| 10 | * |
| 11 | * * Redistributions of source code must retain the above copyright |
| 12 | * notice, this list of conditions and the following disclaimer. |
| 13 | * * Redistributions in binary form must reproduce the above |
| 14 | * copyright notice, this list of conditions and the following disclaimer |
| 15 | * in the documentation and/or other materials provided with the |
| 16 | * distribution. |
| 17 | * |
| 18 | * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS |
| 19 | * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT |
| 20 | * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR |
| 21 | * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT |
| 22 | * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, |
| 23 | * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT |
| 24 | * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, |
| 25 | * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY |
| 26 | * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT |
| 27 | * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE |
| 28 | * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. |
| 29 | * |
| 30 | * You can contact the author at : |
| 31 | * - xxHash homepage: http://www.xxhash.com |
| 32 | * - xxHash source repository : https://github.com/Cyan4973/xxHash |
| 33 | */ |
| 34 | |
| 35 | /* based on revision d2df04efcbef7d7f6886d345861e5dfda4edacc1 Removed |
| 36 | * everything but a simple interface for computing XXh64. */ |
| 37 | |
| 38 | #include "llvm/Support/xxhash.h" |
| 39 | #include "llvm/Support/Endian.h" |
| 40 | |
| 41 | #include <stdlib.h> |
| 42 | #include <string.h> |
| 43 | |
| 44 | using namespace llvm; |
| 45 | using namespace support; |
| 46 | |
| 47 | static uint64_t rotl64(uint64_t X, size_t R) { |
| 48 | return (X << R) | (X >> (64 - R)); |
| 49 | } |
| 50 | |
| 51 | static const uint64_t PRIME64_1 = 11400714785074694791ULL; |
| 52 | static const uint64_t PRIME64_2 = 14029467366897019727ULL; |
| 53 | static const uint64_t PRIME64_3 = 1609587929392839161ULL; |
| 54 | static const uint64_t PRIME64_4 = 9650029242287828579ULL; |
| 55 | static const uint64_t PRIME64_5 = 2870177450012600261ULL; |
| 56 | |
| 57 | static uint64_t round(uint64_t Acc, uint64_t Input) { |
| 58 | Acc += Input * PRIME64_2; |
| 59 | Acc = rotl64(Acc, 31); |
| 60 | Acc *= PRIME64_1; |
| 61 | return Acc; |
| 62 | } |
| 63 | |
| 64 | static uint64_t mergeRound(uint64_t Acc, uint64_t Val) { |
| 65 | Val = round(0, Val); |
| 66 | Acc ^= Val; |
| 67 | Acc = Acc * PRIME64_1 + PRIME64_4; |
| 68 | return Acc; |
| 69 | } |
| 70 | |
| 71 | uint64_t llvm::xxHash64(StringRef Data) { |
| 72 | size_t Len = Data.size(); |
| 73 | uint64_t Seed = 0; |
| 74 | const char *P = Data.data(); |
| 75 | const char *const BEnd = P + Len; |
| 76 | uint64_t H64; |
| 77 | |
| 78 | if (Len >= 32) { |
| 79 | const char *const Limit = BEnd - 32; |
| 80 | uint64_t V1 = Seed + PRIME64_1 + PRIME64_2; |
| 81 | uint64_t V2 = Seed + PRIME64_2; |
| 82 | uint64_t V3 = Seed + 0; |
| 83 | uint64_t V4 = Seed - PRIME64_1; |
| 84 | |
| 85 | do { |
| 86 | V1 = round(V1, endian::read64le(P)); |
| 87 | P += 8; |
| 88 | V2 = round(V2, endian::read64le(P)); |
| 89 | P += 8; |
| 90 | V3 = round(V3, endian::read64le(P)); |
| 91 | P += 8; |
| 92 | V4 = round(V4, endian::read64le(P)); |
| 93 | P += 8; |
| 94 | } while (P <= Limit); |
| 95 | |
| 96 | H64 = rotl64(V1, 1) + rotl64(V2, 7) + rotl64(V3, 12) + rotl64(V4, 18); |
| 97 | H64 = mergeRound(H64, V1); |
| 98 | H64 = mergeRound(H64, V2); |
| 99 | H64 = mergeRound(H64, V3); |
| 100 | H64 = mergeRound(H64, V4); |
| 101 | |
| 102 | } else { |
| 103 | H64 = Seed + PRIME64_5; |
| 104 | } |
| 105 | |
| 106 | H64 += (uint64_t)Len; |
| 107 | |
| 108 | while (P + 8 <= BEnd) { |
| 109 | uint64_t const K1 = round(0, endian::read64le(P)); |
| 110 | H64 ^= K1; |
| 111 | H64 = rotl64(H64, 27) * PRIME64_1 + PRIME64_4; |
| 112 | P += 8; |
| 113 | } |
| 114 | |
| 115 | if (P + 4 <= BEnd) { |
| 116 | H64 ^= (uint64_t)(endian::read32le(P)) * PRIME64_1; |
| 117 | H64 = rotl64(H64, 23) * PRIME64_2 + PRIME64_3; |
| 118 | P += 4; |
| 119 | } |
| 120 | |
| 121 | while (P < BEnd) { |
| 122 | H64 ^= (*P) * PRIME64_5; |
| 123 | H64 = rotl64(H64, 11) * PRIME64_1; |
| 124 | P++; |
| 125 | } |
| 126 | |
| 127 | H64 ^= H64 >> 33; |
| 128 | H64 *= PRIME64_2; |
| 129 | H64 ^= H64 >> 29; |
| 130 | H64 *= PRIME64_3; |
| 131 | H64 ^= H64 >> 32; |
| 132 | |
| 133 | return H64; |
| 134 | } |