All Downloads are FREE. Search and download functionalities are using the official Maven repository.
Many resources are needed to download a project. Please understand that we have to compensate our server costs. Thank you in advance.
Project price only 1 $
You can buy this project and download/modify it how often you want.
org.infinispan.commons.hash.MurmurHash3 Maven / Gradle / Ivy
package org.infinispan.commons.hash;
import java.io.ObjectInput;
import java.util.Collections;
import java.util.Set;
import org.infinispan.commons.marshall.Ids;
import org.infinispan.commons.marshall.WrappedBytes;
import org.infinispan.commons.marshall.exts.NoStateExternalizer;
import net.jcip.annotations.Immutable;
import net.jcip.annotations.ThreadSafe;
/**
* MurmurHash3 implementation in Java, based on Austin Appleby's original in C
*
* Only implementing x64 version, because this should always be faster on 64 bit
* native processors, even 64 bit being ran with a 32 bit OS; this should also
* be as fast or faster than the x86 version on some modern 32 bit processors.
*
* @author Patrick McFarland
* @see MurmurHash website
* @see MurmurHash entry on Wikipedia
* @since 5.0
*/
@ThreadSafe
@Immutable
public class MurmurHash3 implements Hash {
private final static MurmurHash3 instance = new MurmurHash3();
public static final byte INVALID_CHAR = (byte) '?';
public static MurmurHash3 getInstance() {
return instance;
}
private MurmurHash3() {
}
static class State {
long h1;
long h2;
long k1;
long k2;
long c1;
long c2;
}
static long getblock(byte[] key, int i) {
return
((key[i + 0] & 0x00000000000000FFL))
| ((key[i + 1] & 0x00000000000000FFL) << 8)
| ((key[i + 2] & 0x00000000000000FFL) << 16)
| ((key[i + 3] & 0x00000000000000FFL) << 24)
| ((key[i + 4] & 0x00000000000000FFL) << 32)
| ((key[i + 5] & 0x00000000000000FFL) << 40)
| ((key[i + 6] & 0x00000000000000FFL) << 48)
| ((key[i + 7] & 0x00000000000000FFL) << 56);
}
static void bmix(State state) {
state.k1 *= state.c1;
state.k1 = (state.k1 << 23) | (state.k1 >>> 64 - 23);
state.k1 *= state.c2;
state.h1 ^= state.k1;
state.h1 += state.h2;
state.h2 = (state.h2 << 41) | (state.h2 >>> 64 - 41);
state.k2 *= state.c2;
state.k2 = (state.k2 << 23) | (state.k2 >>> 64 - 23);
state.k2 *= state.c1;
state.h2 ^= state.k2;
state.h2 += state.h1;
state.h1 = state.h1 * 3 + 0x52dce729;
state.h2 = state.h2 * 3 + 0x38495ab5;
state.c1 = state.c1 * 5 + 0x7b7d159c;
state.c2 = state.c2 * 5 + 0x6bce6396;
}
static long fmix(long k) {
k ^= k >>> 33;
k *= 0xff51afd7ed558ccdL;
k ^= k >>> 33;
k *= 0xc4ceb9fe1a85ec53L;
k ^= k >>> 33;
return k;
}
/**
* Hash a value using the x64 128 bit variant of MurmurHash3
*
* @param key value to hash
* @param seed random value
* @return 128 bit hashed key, in an array containing two longs
*/
public static long[] MurmurHash3_x64_128(final byte[] key, final int seed) {
State state = new State();
state.h1 = 0x9368e53c2f6af274L ^ seed;
state.h2 = 0x586dcd208f7cd3fdL ^ seed;
state.c1 = 0x87c37b91114253d5L;
state.c2 = 0x4cf5ad432745937fL;
for (int i = 0; i < key.length / 16; i++) {
state.k1 = getblock(key, i * 2 * 8);
state.k2 = getblock(key, (i * 2 + 1) * 8);
bmix(state);
}
state.k1 = 0;
state.k2 = 0;
int tail = (key.length >>> 4) << 4;
switch (key.length & 15) {
case 15: state.k2 ^= (long) key[tail + 14] << 48;
case 14: state.k2 ^= (long) key[tail + 13] << 40;
case 13: state.k2 ^= (long) key[tail + 12] << 32;
case 12: state.k2 ^= (long) key[tail + 11] << 24;
case 11: state.k2 ^= (long) key[tail + 10] << 16;
case 10: state.k2 ^= (long) key[tail + 9] << 8;
case 9: state.k2 ^= key[tail + 8];
case 8: state.k1 ^= (long) key[tail + 7] << 56;
case 7: state.k1 ^= (long) key[tail + 6] << 48;
case 6: state.k1 ^= (long) key[tail + 5] << 40;
case 5: state.k1 ^= (long) key[tail + 4] << 32;
case 4: state.k1 ^= (long) key[tail + 3] << 24;
case 3: state.k1 ^= (long) key[tail + 2] << 16;
case 2: state.k1 ^= (long) key[tail + 1] << 8;
case 1: state.k1 ^= key[tail + 0];
bmix(state);
}
state.h2 ^= key.length;
state.h1 += state.h2;
state.h2 += state.h1;
state.h1 = fmix(state.h1);
state.h2 = fmix(state.h2);
state.h1 += state.h2;
state.h2 += state.h1;
return new long[] { state.h1, state.h2 };
}
/**
* Hash a value using the x64 64 bit variant of MurmurHash3
*
* @param key value to hash
* @param seed random value
* @return 64 bit hashed key
*/
public static long MurmurHash3_x64_64(final byte[] key, final int seed) {
// Exactly the same as MurmurHash3_x64_128, except it only returns state.h1
State state = new State();
state.h1 = 0x9368e53c2f6af274L ^ seed;
state.h2 = 0x586dcd208f7cd3fdL ^ seed;
state.c1 = 0x87c37b91114253d5L;
state.c2 = 0x4cf5ad432745937fL;
for (int i = 0; i < key.length / 16; i++) {
state.k1 = getblock(key, i * 2 * 8);
state.k2 = getblock(key, (i * 2 + 1) * 8);
bmix(state);
}
state.k1 = 0;
state.k2 = 0;
int tail = (key.length >>> 4) << 4;
switch (key.length & 15) {
case 15: state.k2 ^= (long) key[tail + 14] << 48;
case 14: state.k2 ^= (long) key[tail + 13] << 40;
case 13: state.k2 ^= (long) key[tail + 12] << 32;
case 12: state.k2 ^= (long) key[tail + 11] << 24;
case 11: state.k2 ^= (long) key[tail + 10] << 16;
case 10: state.k2 ^= (long) key[tail + 9] << 8;
case 9: state.k2 ^= key[tail + 8];
case 8: state.k1 ^= (long) key[tail + 7] << 56;
case 7: state.k1 ^= (long) key[tail + 6] << 48;
case 6: state.k1 ^= (long) key[tail + 5] << 40;
case 5: state.k1 ^= (long) key[tail + 4] << 32;
case 4: state.k1 ^= (long) key[tail + 3] << 24;
case 3: state.k1 ^= (long) key[tail + 2] << 16;
case 2: state.k1 ^= (long) key[tail + 1] << 8;
case 1: state.k1 ^= key[tail + 0];
bmix(state);
}
state.h2 ^= key.length;
state.h1 += state.h2;
state.h2 += state.h1;
state.h1 = fmix(state.h1);
state.h2 = fmix(state.h2);
state.h1 += state.h2;
state.h2 += state.h1;
return state.h1;
}
/**
* Hash a value using the x64 32 bit variant of MurmurHash3
*
* @param key value to hash
* @param seed random value
* @return 32 bit hashed key
*/
public static int MurmurHash3_x64_32(final byte[] key, final int seed) {
return (int) (MurmurHash3_x64_64(key, seed) >>> 32);
}
/**
* Hash a value using the x64 128 bit variant of MurmurHash3
*
* @param key value to hash
* @param seed random value
* @return 128 bit hashed key, in an array containing two longs
*/
public static long[] MurmurHash3_x64_128(final long[] key, final int seed) {
State state = new State();
state.h1 = 0x9368e53c2f6af274L ^ seed;
state.h2 = 0x586dcd208f7cd3fdL ^ seed;
state.c1 = 0x87c37b91114253d5L;
state.c2 = 0x4cf5ad432745937fL;
for (int i = 0; i < key.length / 2; i++) {
state.k1 = key[i * 2];
state.k2 = key[i * 2 + 1];
bmix(state);
}
long tail = key[key.length - 1];
// Key length is odd
if ((key.length & 1) == 1) {
state.k1 ^= tail;
bmix(state);
}
state.h2 ^= key.length * 8;
state.h1 += state.h2;
state.h2 += state.h1;
state.h1 = fmix(state.h1);
state.h2 = fmix(state.h2);
state.h1 += state.h2;
state.h2 += state.h1;
return new long[] { state.h1, state.h2 };
}
/**
* Hash a value using the x64 64 bit variant of MurmurHash3
*
* @param key value to hash
* @param seed random value
* @return 64 bit hashed key
*/
public static long MurmurHash3_x64_64(final long[] key, final int seed) {
// Exactly the same as MurmurHash3_x64_128, except it only returns state.h1
State state = new State();
state.h1 = 0x9368e53c2f6af274L ^ seed;
state.h2 = 0x586dcd208f7cd3fdL ^ seed;
state.c1 = 0x87c37b91114253d5L;
state.c2 = 0x4cf5ad432745937fL;
for (int i = 0; i < key.length / 2; i++) {
state.k1 = key[i * 2];
state.k2 = key[i * 2 + 1];
bmix(state);
}
long tail = key[key.length - 1];
if (key.length % 2 != 0) {
state.k1 ^= tail;
bmix(state);
}
state.h2 ^= key.length * 8;
state.h1 += state.h2;
state.h2 += state.h1;
state.h1 = fmix(state.h1);
state.h2 = fmix(state.h2);
state.h1 += state.h2;
state.h2 += state.h1;
return state.h1;
}
/**
* Hash a value using the x64 32 bit variant of MurmurHash3
*
* @param key value to hash
* @param seed random value
* @return 32 bit hashed key
*/
public static int MurmurHash3_x64_32(final long[] key, final int seed) {
return (int) (MurmurHash3_x64_64(key, seed) >>> 32);
}
@Override
public int hash(byte[] payload) {
return MurmurHash3_x64_32(payload, 9001);
}
/**
* Hashes a byte array efficiently.
*
* @param payload a byte array to hash
* @return a hash code for the byte array
*/
public static int hash(long[] payload) {
return MurmurHash3_x64_32(payload, 9001);
}
@Override
public int hash(int hashcode) {
// Obtained by inlining MurmurHash3_x64_32(byte[], 9001) and removing all the unused code
// (since we know the input is always 4 bytes and we only need 4 bytes of output)
byte b0 = (byte) hashcode;
byte b1 = (byte) (hashcode >>> 8);
byte b2 = (byte) (hashcode >>> 16);
byte b3 = (byte) (hashcode >>> 24);
State state = new State();
state.h1 = 0x9368e53c2f6af274L ^ 9001;
state.h2 = 0x586dcd208f7cd3fdL ^ 9001;
state.c1 = 0x87c37b91114253d5L;
state.c2 = 0x4cf5ad432745937fL;
state.k1 = 0;
state.k2 = 0;
state.k1 ^= (long) b3 << 24;
state.k1 ^= (long) b2 << 16;
state.k1 ^= (long) b1 << 8;
state.k1 ^= b0;
bmix(state);
state.h2 ^= 4;
state.h1 += state.h2;
state.h2 += state.h1;
state.h1 = fmix(state.h1);
state.h2 = fmix(state.h2);
state.h1 += state.h2;
state.h2 += state.h1;
return (int) (state.h1 >>> 32);
}
@Override
public int hash(Object o) {
if (o instanceof byte[])
return hash((byte[]) o);
else if (o instanceof WrappedBytes) {
return hash(((WrappedBytes) o).getBytes());
}
else if (o instanceof long[])
return hash((long[]) o);
else if (o instanceof String)
return hashString((String) o);
else
return hash(o.hashCode());
}
private int hashString(String s) {
return (int) (MurmurHash3_x64_64_String(s, 9001) >> 32);
}
private long MurmurHash3_x64_64_String(String s, long seed) {
// Exactly the same as MurmurHash3_x64_64, except it works directly on a String's chars
MurmurHash3.State state = new MurmurHash3.State();
state.h1 = 0x9368e53c2f6af274L ^ seed;
state.h2 = 0x586dcd208f7cd3fdL ^ seed;
state.c1 = 0x87c37b91114253d5L;
state.c2 = 0x4cf5ad432745937fL;
int byteLen = 0;
int stringLen = s.length();
for (int i = 0; i < stringLen; i++) {
char c1 = s.charAt(i);
int cp;
if (!Character.isSurrogate(c1)) {
cp = c1;
} else if (Character.isHighSurrogate(c1)){
if (i + 1 < stringLen) {
char c2 = s.charAt(i + 1);
if (Character.isLowSurrogate(c2)) {
i++;
cp = Character.toCodePoint(c1, c2);
} else {
cp = INVALID_CHAR;
}
} else {
cp = INVALID_CHAR;
}
} else {
cp = INVALID_CHAR;
}
if (cp <= 0x7f) {
addByte(state, (byte) cp, byteLen++);
} else if (cp <= 0x07ff) {
byte b1 = (byte) (0xc0 | (0x1f & (cp >> 6)));
byte b2 = (byte) (0x80 | (0x3f & cp));
addByte(state, b1, byteLen++);
addByte(state, b2, byteLen++);
} else if (cp <= 0xffff) {
byte b1 = (byte) (0xe0 | (0x0f & (cp >> 12)));
byte b2 = (byte) (0x80 | (0x3f & (cp >> 6)));
byte b3 = (byte) (0x80 | (0x3f & cp));
addByte(state, b1, byteLen++);
addByte(state, b2, byteLen++);
addByte(state, b3, byteLen++);
} else {
byte b1 = (byte) (0xf0 | (0x07 & (cp >> 18)));
byte b2 = (byte) (0x80 | (0x3f & (cp >> 12)));
byte b3 = (byte) (0x80 | (0x3f & (cp >> 6)));
byte b4 = (byte) (0x80 | (0x3f & cp));
addByte(state, b1, byteLen++);
addByte(state, b2, byteLen++);
addByte(state, b3, byteLen++);
addByte(state, b4, byteLen++);
}
}
long savedK1 = state.k1;
long savedK2 = state.k2;
state.k1 = 0;
state.k2 = 0;
switch (byteLen & 15) {
case 15:
state.k2 ^= (long) ((byte)(savedK2 >> 48)) << 48;
case 14:
state.k2 ^= (long) ((byte) (savedK2 >> 40)) << 40;
case 13:
state.k2 ^= (long) ((byte) (savedK2 >> 32)) << 32;
case 12:
state.k2 ^= (long) ((byte) (savedK2 >> 24)) << 24;
case 11:
state.k2 ^= (long) ((byte) (savedK2 >> 16)) << 16;
case 10:
state.k2 ^= (long) ((byte) (savedK2 >> 8)) << 8;
case 9:
state.k2 ^= ((byte) savedK2);
case 8:
state.k1 ^= (long) ((byte) (savedK1 >> 56)) << 56;
case 7:
state.k1 ^= (long) ((byte) (savedK1 >> 48)) << 48;
case 6:
state.k1 ^= (long) ((byte) (savedK1 >> 40)) << 40;
case 5:
state.k1 ^= (long) ((byte) (savedK1 >> 32)) << 32;
case 4:
state.k1 ^= (long) ((byte) (savedK1 >> 24)) << 24;
case 3:
state.k1 ^= (long) ((byte) (savedK1 >> 16)) << 16;
case 2:
state.k1 ^= (long) ((byte) (savedK1 >> 8)) << 8;
case 1:
state.k1 ^= ((byte) savedK1);
bmix(state);
}
state.h2 ^= byteLen;
state.h1 += state.h2;
state.h2 += state.h1;
state.h1 = fmix(state.h1);
state.h2 = fmix(state.h2);
state.h1 += state.h2;
state.h2 += state.h1;
return state.h1;
}
private void addByte(State state, byte b, int len) {
int shift = (len & 0x7) * 8;
long bb = (b & 0xffL) << shift;
if ((len & 0x8) == 0) {
state.k1 |= bb;
} else {
state.k2 |= bb;
if ((len & 0xf) == 0xf) {
bmix(state);
state.k1 = 0;
state.k2 = 0;
}
}
}
@Override
public boolean equals(Object other) {
return other != null && other.getClass() == getClass();
}
@Override
public int hashCode() {
return 0;
}
@Override
public String toString() {
return "MurmurHash3";
}
public static class Externalizer extends NoStateExternalizer {
@Override
public Set> getTypeClasses() {
return Collections.singleton(MurmurHash3.class);
}
@Override
public MurmurHash3 readObject(ObjectInput input) {
return instance;
}
@Override
public Integer getId() {
return Ids.MURMURHASH_3;
}
}
}
