1 /*-----------------------------------------------------------------------------
2 * MurmurHash3 was written by Austin Appleby, and is placed in the public
3 * domain.
4 *
5 * This implementation was written by Shane Day, and is also public domain.
6 *
7 * This is a portable ANSI C implementation of MurmurHash3_x86_32 (Murmur3A)
8 * with support for progressive processing.
9 */
10
11 /*-----------------------------------------------------------------------------
12
13 If you want to understand the MurmurHash algorithm you would be much better
14 off reading the original source. Just point your browser at:
15 http://code.google.com/p/smhasher/source/browse/trunk/MurmurHash3.cpp
16
17
18 What this version provides?
19
20 1. Progressive data feeding. Useful when the entire payload to be hashed
21 does not fit in memory or when the data is streamed through the application.
22 Also useful when hashing a number of strings with a common prefix. A partial
23 hash of a prefix string can be generated and reused for each suffix string.
24
25 How does it work?
26
27 We can only process entire 32 bit chunks of input, except for the very end
28 that may be shorter. So along with the partial hash we need to give back to
29 the caller a carry containing up to 3 bytes that we were unable to process.
30 This carry also needs to record the number of bytes the carry holds. I use
31 the low 2 bits as a count (0..3) and the carry bytes are shifted into the
32 high byte in stream order.
33
34 To handle endianess I simply use a macro that reads a uint32_t and define
35 that macro to be a direct read on little endian machines, a read and swap
36 on big endian machines, or a byte-by-byte read if the endianess is unknown.
37
38 -----------------------------------------------------------------------------*/
39
40
41 #include "PMurHash.h"
42
43 // /* MSVC warnings we choose to ignore */
44 // #if defined(_MSC_VER)
45 // #pragma warning(disable: 4127) /* conditional expression is constant */
46 // #endif
47
48 /*-----------------------------------------------------------------------------
49 * Endianess, misalignment capabilities and util macros
50 *
51 * The following 3 macros are defined in this section. The other macros defined
52 * are only needed to help derive these 3.
53 *
54 * READ_UINT32(x) Read a little endian unsigned 32-bit int
55 * UNALIGNED_SAFE Defined if READ_UINT32 works on non-word boundaries
56 * ROTL32(x,r) Rotate x left by r bits
57 */
58
59 /* I386 or AMD64 */
60 #if defined(_M_I86) || defined(_M_IX86) || defined(_X86_) || defined(__i386__) || defined(__i386) || defined(i386) \
61 || defined(_M_X64) || defined(__x86_64__) || defined(__x86_64) || defined(__amd64__) || defined(__amd64)
62 #define UNALIGNED_SAFE
63 #endif
64 /* I386 or AMD64 */
65 #if defined(_M_I86) || defined(_M_IX86) || defined(_X86_) || defined(__i386__) || defined(__i386) || defined(i386) \
66 || defined(_M_X64) || defined(__x86_64__) || defined(__x86_64) || defined(__amd64__) || defined(__amd64)
67 #define UNALIGNED_SAFE
68 #endif
69
70 /* Find best way to ROTL */
71 #if defined(_MSC_VER)
72 #define FORCE_INLINE static __forceinline
73 #include <stdlib.h> /* Microsoft put _rotl declaration in here */
74 #define ROTL32(x,y) _rotl(x,y)
75 #else
76 #define FORCE_INLINE static inline __attribute__((always_inline))
77 /* gcc recognises this code and generates a rotate instruction for CPUs with one */
78 #define ROTL32(x,r) (((uint32_t)x << r) | ((uint32_t)x >> (32 - r)))
79 #endif
80
81 #include "endianness.h"
82
83 #define READ_UINT32(ptr) getblock32((uint32_t *)ptr, 0)
84
85 /*-----------------------------------------------------------------------------
86 * Core murmurhash algorithm macros */
87
88 static const uint32_t kC1 = 0xcc9e2d51;
89 static const uint32_t kC2 = 0x1b873593;
90
91 /* This is the main processing body of the algorithm. It operates
92 * on each full 32-bits of input. */
93 #define doblock(h1, k1) \
94 do {\
95 k1 *= kC1;\
96 k1 = ROTL32(k1,15);\
97 k1 *= kC2;\
98 \
99 h1 ^= k1;\
100 h1 = ROTL32(h1,13);\
101 h1 = h1*5+0xe6546b64;\
102 } while(0)
103
104 /* Append unaligned bytes to carry, forcing hash churn if we have 4 bytes */
105 /* cnt=bytes to process, h1=name of h1 var, c=carry, n=bytes in c, ptr/len=payload */
106 #define dobytes(cnt, h1, c, n, ptr, len) \
107 do {\
108 unsigned __cnt = cnt;\
109 while(__cnt--) {\
110 c = c>>8 | (uint32_t)*ptr++<<24;\
111 n++; len--;\
112 if(n==4) {\
113 doblock(h1, c);\
114 n = 0;\
115 }\
116 }\
117 } while(0)
118
119 /*---------------------------------------------------------------------------*/
120
121 /* Main hashing function. Initialise carry to 0 and h1 to 0 or an initial seed
122 * if wanted. Both ph1 and pcarry are required arguments. */
PMurHash32_Process(uint32_t * ph1,uint32_t * pcarry,const void * key,int len)123 void PMurHash32_Process(uint32_t *ph1, uint32_t *pcarry, const void *key, int len)
124 {
125 uint32_t h1 = *ph1;
126 uint32_t c = *pcarry;
127
128 const uint8_t *ptr = (uint8_t*)key;
129 const uint8_t *end;
130
131 /* Extract carry count from low 2 bits of c value */
132 int n = c & 3;
133
134 #if defined(UNALIGNED_SAFE)
135 /* This CPU handles unaligned word access */
136 // #pragma message ( "UNALIGNED_SAFE" )
137 /* Consume any carry bytes */
138 int i = (4-n) & 3;
139 if(i && i <= len) {
140 dobytes(i, h1, c, n, ptr, len);
141 }
142
143 /* Process 32-bit chunks */
144 end = ptr + (len & ~3);
145 for( ; ptr < end ; ptr+=4) {
146 uint32_t k1 = READ_UINT32(ptr);
147 doblock(h1, k1);
148 }
149
150 #else /*UNALIGNED_SAFE*/
151 /* This CPU does not handle unaligned word access */
152 // #pragma message ( "ALIGNED" )
153 /* Consume enough so that the next data byte is word aligned */
154 int i = -(intptr_t)(void *)ptr & 3;
155 if(i && i <= len) {
156 dobytes(i, h1, c, n, ptr, len);
157 }
158
159 /* We're now aligned. Process in aligned blocks. Specialise for each possible carry count */
160 end = ptr + (len & ~3);
161 switch(n) { /* how many bytes in c */
162 case 0: /* c=[----] w=[3210] b=[3210]=w c'=[----] */
163 for( ; ptr < end ; ptr+=4) {
164 uint32_t k1 = READ_UINT32(ptr);
165 doblock(h1, k1);
166 }
167 break;
168 case 1: /* c=[0---] w=[4321] b=[3210]=c>>24|w<<8 c'=[4---] */
169 for( ; ptr < end ; ptr+=4) {
170 uint32_t k1 = c>>24;
171 c = READ_UINT32(ptr);
172 k1 |= c<<8;
173 doblock(h1, k1);
174 }
175 break;
176 case 2: /* c=[10--] w=[5432] b=[3210]=c>>16|w<<16 c'=[54--] */
177 for( ; ptr < end ; ptr+=4) {
178 uint32_t k1 = c>>16;
179 c = READ_UINT32(ptr);
180 k1 |= c<<16;
181 doblock(h1, k1);
182 }
183 break;
184 case 3: /* c=[210-] w=[6543] b=[3210]=c>>8|w<<24 c'=[654-] */
185 for( ; ptr < end ; ptr+=4) {
186 uint32_t k1 = c>>8;
187 c = READ_UINT32(ptr);
188 k1 |= c<<24;
189 doblock(h1, k1);
190 }
191 }
192 #endif /*UNALIGNED_SAFE*/
193
194 /* Advance over whole 32-bit chunks, possibly leaving 1..3 bytes */
195 len -= len & ~3;
196
197 /* Append any remaining bytes into carry */
198 dobytes(len, h1, c, n, ptr, len);
199
200 /* Copy out new running hash and carry */
201 *ph1 = h1;
202 *pcarry = (c & ~0xff) | n;
203 }
204
205 /*---------------------------------------------------------------------------*/
206
207 /* Finalize a hash. To match the original Murmur3A the total_length must be provided */
PMurHash32_Result(uint32_t h,uint32_t carry,uint32_t total_length)208 uint32_t PMurHash32_Result(uint32_t h, uint32_t carry, uint32_t total_length)
209 {
210 uint32_t k1;
211 int n = carry & 3;
212 if(n) {
213 k1 = carry >> (4-n)*8;
214 k1 *= kC1; k1 = ROTL32(k1,15); k1 *= kC2; h ^= k1;
215 }
216 h ^= total_length;
217
218 /* fmix */
219 h ^= h >> 16;
220 h *= 0x85ebca6b;
221 h ^= h >> 13;
222 h *= 0xc2b2ae35;
223 h ^= h >> 16;
224
225 return h;
226 }
227
