[lttng-tools.git] / src / common / hashtable / utils.c

/*
 * Copyright (C) - Bob Jenkins, May 2006
 * Copyright (C) 2011 - David Goulet <david.goulet@polymtl.ca>
 * Copyright (C) 2011 - Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License, version 2 only,
 * as published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it will be useful, but WITHOUT
 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
 * more details.
 *
 * You should have received a copy of the GNU General Public License along
 * with this program; if not, write to the Free Software Foundation, Inc.,
 * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 */

/*
 * These are functions for producing 32-bit hashes for hash table lookup.
 * hashword(), hashlittle(), hashlittle2(), hashbig(), mix(), and final() are
 * externally useful functions.  Routines to test the hash are included if
 * SELF_TEST is defined.  You can use this free for any purpose.  It's in the
 * public domain.  It has no warranty.
 *
 * You probably want to use hashlittle().  hashlittle() and hashbig() hash byte
 * arrays.  hashlittle() is is faster than hashbig() on little-endian machines.
 * Intel and AMD are little-endian machines.  On second thought, you probably
 * want hashlittle2(), which is identical to hashlittle() except it returns two
 * 32-bit hashes for the price of one.  You could implement hashbig2() if you
 * wanted but I haven't bothered here.
 *
 * If you want to find a hash of, say, exactly 7 integers, do
 *   a = i1;  b = i2;  c = i3;
 *   mix(a,b,c);
 *   a += i4; b += i5; c += i6;
 *   mix(a,b,c);
 *   a += i7;
 *   final(a,b,c);
 * then use c as the hash value.  If you have a variable length array of
 * 4-byte integers to hash, use hashword().  If you have a byte array (like
 * a character string), use hashlittle().  If you have several byte arrays, or
 * a mix of things, see the comments above hashlittle().
 *
 * Why is this so big?  I read 12 bytes at a time into 3 4-byte integers, then
 * mix those integers.  This is fast (you can do a lot more thorough mixing
 * with 12*3 instructions on 3 integers than you can with 3 instructions on 1
 * byte), but shoehorning those bytes into integers efficiently is messy.
 */

#define _LGPL_SOURCE
#include <assert.h>
#include <stdint.h>     /* defines uint32_t etc */
#include <stdio.h>      /* defines printf for tests */
#include <string.h>
#include <sys/param.h>  /* attempt to define endianness */
#include <time.h>       /* defines time_t for timings in the test */
#include <urcu/compiler.h>

#include "utils.h"
#include <common/compat/endian.h>    /* attempt to define endianness */
#include <common/common.h>
#include <common/hashtable/hashtable.h>

/*
 * My best guess at if you are big-endian or little-endian.  This may
 * need adjustment.
 */
#if (defined(__BYTE_ORDER) && defined(__LITTLE_ENDIAN) && \
     __BYTE_ORDER == __LITTLE_ENDIAN) || \
    (defined(i386) || defined(__i386__) || defined(__i486__) || \
     defined(__i586__) || defined(__i686__) || defined(vax) || defined(MIPSEL))
# define HASH_LITTLE_ENDIAN 1
# define HASH_BIG_ENDIAN 0
#elif (defined(__BYTE_ORDER) && defined(__BIG_ENDIAN) && \
       __BYTE_ORDER == __BIG_ENDIAN) || \
      (defined(sparc) || defined(POWERPC) || defined(mc68000) || defined(sel))
# define HASH_LITTLE_ENDIAN 0
# define HASH_BIG_ENDIAN 1
#else
# define HASH_LITTLE_ENDIAN 0
# define HASH_BIG_ENDIAN 0
#endif

#define hashsize(n) ((uint32_t)1<<(n))
#define hashmask(n) (hashsize(n)-1)
#define rot(x,k) (((x)<<(k)) | ((x)>>(32-(k))))

/*
 * mix -- mix 3 32-bit values reversibly.
 *
 * This is reversible, so any information in (a,b,c) before mix() is
 * still in (a,b,c) after mix().
 *
 * If four pairs of (a,b,c) inputs are run through mix(), or through
 * mix() in reverse, there are at least 32 bits of the output that
 * are sometimes the same for one pair and different for another pair.
 * This was tested for:
 * * pairs that differed by one bit, by two bits, in any combination
 *   of top bits of (a,b,c), or in any combination of bottom bits of
 *   (a,b,c).
 * * "differ" is defined as +, -, ^, or ~^.  For + and -, I transformed
 *   the output delta to a Gray code (a^(a>>1)) so a string of 1's (as
 *   is commonly produced by subtraction) look like a single 1-bit
 *   difference.
 * * the base values were pseudorandom, all zero but one bit set, or
 *   all zero plus a counter that starts at zero.
 *
 * Some k values for my "a-=c; a^=rot(c,k); c+=b;" arrangement that
 * satisfy this are
 *     4  6  8 16 19  4
 *     9 15  3 18 27 15
 *    14  9  3  7 17  3
 * Well, "9 15 3 18 27 15" didn't quite get 32 bits diffing
 * for "differ" defined as + with a one-bit base and a two-bit delta.  I
 * used http://burtleburtle.net/bob/hash/avalanche.html to choose
 * the operations, constants, and arrangements of the variables.
 *
 * This does not achieve avalanche.  There are input bits of (a,b,c)
 * that fail to affect some output bits of (a,b,c), especially of a.  The
 * most thoroughly mixed value is c, but it doesn't really even achieve
 * avalanche in c.
 *
 * This allows some parallelism.  Read-after-writes are good at doubling
 * the number of bits affected, so the goal of mixing pulls in the opposite
 * direction as the goal of parallelism.  I did what I could.  Rotates
 * seem to cost as much as shifts on every machine I could lay my hands
 * on, and rotates are much kinder to the top and bottom bits, so I used
 * rotates.
 */
#define mix(a,b,c) \
{ \
  a -= c;  a ^= rot(c, 4);  c += b; \
  b -= a;  b ^= rot(a, 6);  a += c; \
  c -= b;  c ^= rot(b, 8);  b += a; \
  a -= c;  a ^= rot(c,16);  c += b; \
  b -= a;  b ^= rot(a,19);  a += c; \
  c -= b;  c ^= rot(b, 4);  b += a; \
}

/*
 * final -- final mixing of 3 32-bit values (a,b,c) into c
 *
 * Pairs of (a,b,c) values differing in only a few bits will usually
 * produce values of c that look totally different.  This was tested for
 * * pairs that differed by one bit, by two bits, in any combination
 *   of top bits of (a,b,c), or in any combination of bottom bits of
 *   (a,b,c).
 * * "differ" is defined as +, -, ^, or ~^.  For + and -, I transformed
 *   the output delta to a Gray code (a^(a>>1)) so a string of 1's (as
 *   is commonly produced by subtraction) look like a single 1-bit
 *   difference.
 * * the base values were pseudorandom, all zero but one bit set, or
 *   all zero plus a counter that starts at zero.
 *
 * These constants passed:
 *  14 11 25 16 4 14 24
 *  12 14 25 16 4 14 24
 * and these came close:
 *   4  8 15 26 3 22 24
 *  10  8 15 26 3 22 24
 *  11  8 15 26 3 22 24
 */
#define final(a,b,c) \
{ \
  c ^= b; c -= rot(b,14); \
  a ^= c; a -= rot(c,11); \
  b ^= a; b -= rot(a,25); \
  c ^= b; c -= rot(b,16); \
  a ^= c; a -= rot(c,4);  \
  b ^= a; b -= rot(a,14); \
  c ^= b; c -= rot(b,24); \
}

/*
 * k - the key, an array of uint32_t values
 * length - the length of the key, in uint32_ts
 * initval - the previous hash, or an arbitrary value
 */
static uint32_t __attribute__((unused)) hashword(const uint32_t *k,
		size_t length, uint32_t initval)
{
	uint32_t a, b, c;

	/* Set up the internal state */
	a = b = c = 0xdeadbeef + (((uint32_t) length) << 2) + initval;

	/*----------------------------------------- handle most of the key */
	while (length > 3) {
		a += k[0];
		b += k[1];
		c += k[2];
		mix(a, b, c);
		length -= 3;
		k += 3;
	}

	/*----------------------------------- handle the last 3 uint32_t's */
	switch (length) {	/* all the case statements fall through */
	case 3: c += k[2];
	case 2: b += k[1];
	case 1: a += k[0];
		final(a, b, c);
	case 0:			/* case 0: nothing left to add */
		break;
	}
	/*---------------------------------------------- report the result */
	return c;
}


/*
 * hashword2() -- same as hashword(), but take two seeds and return two 32-bit
 * values.  pc and pb must both be nonnull, and *pc and *pb must both be
 * initialized with seeds.  If you pass in (*pb)==0, the output (*pc) will be
 * the same as the return value from hashword().
 */
static void __attribute__((unused)) hashword2(const uint32_t *k, size_t length,
		uint32_t *pc, uint32_t *pb)
{
	uint32_t a, b, c;

	/* Set up the internal state */
	a = b = c = 0xdeadbeef + ((uint32_t) (length << 2)) + *pc;
	c += *pb;

	while (length > 3) {
		a += k[0];
		b += k[1];
		c += k[2];
		mix(a, b, c);
		length -= 3;
		k += 3;
	}

	switch (length) {
	case 3 :
		c += k[2];
	case 2 :
		b += k[1];
	case 1 :
		a += k[0];
		final(a, b, c);
	case 0:     /* case 0: nothing left to add */
		break;
	}

	*pc = c;
	*pb = b;
}

/*
 * hashlittle() -- hash a variable-length key into a 32-bit value
 *   k       : the key (the unaligned variable-length array of bytes)
 *   length  : the length of the key, counting by bytes
 *   initval : can be any 4-byte value
 * Returns a 32-bit value.  Every bit of the key affects every bit of
 * the return value.  Two keys differing by one or two bits will have
 * totally different hash values.
 *
 * The best hash table sizes are powers of 2.  There is no need to do
 * mod a prime (mod is sooo slow!).  If you need less than 32 bits,
 * use a bitmask.  For example, if you need only 10 bits, do
 *   h = (h & hashmask(10));
 * In which case, the hash table should have hashsize(10) elements.
 *
 * If you are hashing n strings (uint8_t **)k, do it like this:
 *   for (i=0, h=0; i<n; ++i) h = hashlittle( k[i], len[i], h);
 *
 * By Bob Jenkins, 2006.  bob_jenkins@burtleburtle.net.  You may use this
 * code any way you wish, private, educational, or commercial.  It's free.
 *
 * Use for hash table lookup, or anything where one collision in 2^^32 is
 * acceptable.  Do NOT use for cryptographic purposes.
 */
static uint32_t __attribute__((unused)) hashlittle(const void *key,
		size_t length, uint32_t initval)
{
	uint32_t a,b,c;
	union {
		const void *ptr;
		size_t i;
	} u;     /* needed for Mac Powerbook G4 */

	/* Set up the internal state */
	a = b = c = 0xdeadbeef + ((uint32_t)length) + initval;

	u.ptr = key;
	if (HASH_LITTLE_ENDIAN && ((u.i & 0x3) == 0)) {
		const uint32_t *k = (const uint32_t *)key;         /* read 32-bit chunks */

		/*------ all but last block: aligned reads and affect 32 bits of (a,b,c) */
		while (length > 12) {
			a += k[0];
			b += k[1];
			c += k[2];
			mix(a,b,c);
			length -= 12;
			k += 3;
		}

		/*
		 * "k[2]&0xffffff" actually reads beyond the end of the string, but
		 * then masks off the part it's not allowed to read.  Because the
		 * string is aligned, the masked-off tail is in the same word as the
		 * rest of the string.  Every machine with memory protection I've seen
		 * does it on word boundaries, so is OK with this.  But VALGRIND will
		 * still catch it and complain.  The masking trick does make the hash
		 * noticably faster for short strings (like English words).
		 */
#ifndef VALGRIND

		switch (length) {
		case 12: c+=k[2]; b+=k[1]; a+=k[0]; break;
		case 11: c+=k[2]&0xffffff; b+=k[1]; a+=k[0]; break;
		case 10: c+=k[2]&0xffff; b+=k[1]; a+=k[0]; break;
		case 9 : c+=k[2]&0xff; b+=k[1]; a+=k[0]; break;
		case 8 : b+=k[1]; a+=k[0]; break;
		case 7 : b+=k[1]&0xffffff; a+=k[0]; break;
		case 6 : b+=k[1]&0xffff; a+=k[0]; break;
		case 5 : b+=k[1]&0xff; a+=k[0]; break;
		case 4 : a+=k[0]; break;
		case 3 : a+=k[0]&0xffffff; break;
		case 2 : a+=k[0]&0xffff; break;
		case 1 : a+=k[0]&0xff; break;
		case 0 : return c;              /* zero length strings require no mixing */
		}
#else /* make valgrind happy */
		const uint8_t *k8;

		k8 = (const uint8_t *)k;
		switch (length) {
		case 12: c+=k[2]; b+=k[1]; a+=k[0]; break;
		case 11: c+=((uint32_t)k8[10])<<16;  /* fall through */
		case 10: c+=((uint32_t)k8[9])<<8;    /* fall through */
		case 9 : c+=k8[8];                   /* fall through */
		case 8 : b+=k[1]; a+=k[0]; break;
		case 7 : b+=((uint32_t)k8[6])<<16;   /* fall through */
		case 6 : b+=((uint32_t)k8[5])<<8;    /* fall through */
		case 5 : b+=k8[4];                   /* fall through */
		case 4 : a+=k[0]; break;
		case 3 : a+=((uint32_t)k8[2])<<16;   /* fall through */
		case 2 : a+=((uint32_t)k8[1])<<8;    /* fall through */
		case 1 : a+=k8[0]; break;
		case 0 : return c;
		}
#endif /* !valgrind */
	} else if (HASH_LITTLE_ENDIAN && ((u.i & 0x1) == 0)) {
		const uint16_t *k = (const uint16_t *)key;         /* read 16-bit chunks */
		const uint8_t *k8;

		/*--------------- all but last block: aligned reads and different mixing */
		while (length > 12) {
			a += k[0] + (((uint32_t)k[1])<<16);
			b += k[2] + (((uint32_t)k[3])<<16);
			c += k[4] + (((uint32_t)k[5])<<16);
			mix(a,b,c);
			length -= 12;
			k += 6;
		}

		k8 = (const uint8_t *)k;
		switch (length) {
		case 12:
			c+=k[4]+(((uint32_t)k[5])<<16);
			b+=k[2]+(((uint32_t)k[3])<<16);
			a+=k[0]+(((uint32_t)k[1])<<16);
			break;
		case 11:
			c+=((uint32_t)k8[10])<<16;     /* fall through */
		case 10:
			c+=k[4];
			b+=k[2]+(((uint32_t)k[3])<<16);
			a+=k[0]+(((uint32_t)k[1])<<16);
			break;
		case 9:
			c+=k8[8];                      /* fall through */
		case 8:
			b+=k[2]+(((uint32_t)k[3])<<16);
			a+=k[0]+(((uint32_t)k[1])<<16);
			break;
		case 7:
			b+=((uint32_t)k8[6])<<16;      /* fall through */
		case 6:
			b+=k[2];
			a+=k[0]+(((uint32_t)k[1])<<16);
			break;
		case 5:
			b+=k8[4];                      /* fall through */
		case 4:
			a+=k[0]+(((uint32_t)k[1])<<16);
			break;
		case 3:
			a+=((uint32_t)k8[2])<<16;      /* fall through */
		case 2:
			a+=k[0];
			break;
		case 1:
			a+=k8[0];
			break;
		case 0:
			return c;   /* zero length requires no mixing */
		}

	} else {    /* need to read the key one byte at a time */
		const uint8_t *k = (const uint8_t *)key;

		while (length > 12) {
			a += k[0];
			a += ((uint32_t)k[1])<<8;
			a += ((uint32_t)k[2])<<16;
			a += ((uint32_t)k[3])<<24;
			b += k[4];
			b += ((uint32_t)k[5])<<8;
			b += ((uint32_t)k[6])<<16;
			b += ((uint32_t)k[7])<<24;
			c += k[8];
			c += ((uint32_t)k[9])<<8;
			c += ((uint32_t)k[10])<<16;
			c += ((uint32_t)k[11])<<24;
			mix(a,b,c);
			length -= 12;
			k += 12;
		}

		switch(length) {                  /* all the case statements fall through */
		case 12: c+=((uint32_t)k[11])<<24;
		case 11: c+=((uint32_t)k[10])<<16;
		case 10: c+=((uint32_t)k[9])<<8;
		case 9: c+=k[8];
		case 8: b+=((uint32_t)k[7])<<24;
		case 7: b+=((uint32_t)k[6])<<16;
		case 6: b+=((uint32_t)k[5])<<8;
		case 5: b+=k[4];
		case 4: a+=((uint32_t)k[3])<<24;
		case 3: a+=((uint32_t)k[2])<<16;
		case 2: a+=((uint32_t)k[1])<<8;
		case 1:
			a+=k[0];
			break;
		case 0:
			return c;
		}
	}

	final(a,b,c);
	return c;
}

LTTNG_HIDDEN
unsigned long hash_key_u64(const void *_key, unsigned long seed)
{
	union {
		uint64_t v64;
		uint32_t v32[2];
	} v;
	union {
		uint64_t v64;
		uint32_t v32[2];
	} key;

	v.v64 = (uint64_t) seed;
	key.v64 = *(const uint64_t *) _key;
	hashword2(key.v32, 2, &v.v32[0], &v.v32[1]);
	return v.v64;
}

#if (CAA_BITS_PER_LONG == 64)
/*
 * Hash function for number value.
 * Pass the value itself as the key, not its address.
 */
LTTNG_HIDDEN
unsigned long hash_key_ulong(const void *_key, unsigned long seed)
{
	uint64_t __key = (uint64_t) _key;
	return (unsigned long) hash_key_u64(&__key, seed);
}
#else
/*
 * Hash function for number value.
 * Pass the value itself as the key, not its address.
 */
LTTNG_HIDDEN
unsigned long hash_key_ulong(const void *_key, unsigned long seed)
{
	uint32_t key = (uint32_t) _key;

	return hashword(&key, 1, seed);
}
#endif /* CAA_BITS_PER_LONG */

/*
 * Hash function for string.
 */
LTTNG_HIDDEN
unsigned long hash_key_str(const void *key, unsigned long seed)
{
	return hashlittle(key, strlen((const char *) key), seed);
}

/*
 * Hash function for two uint64_t.
 */
LTTNG_HIDDEN
unsigned long hash_key_two_u64(const void *key, unsigned long seed)
{
	const struct lttng_ht_two_u64 *k =
		(const struct lttng_ht_two_u64 *) key;

	return hash_key_u64(&k->key1, seed) ^ hash_key_u64(&k->key2, seed);
}

/*
 * Hash function compare for number value.
 */
LTTNG_HIDDEN
int hash_match_key_ulong(const void *key1, const void *key2)
{
	if (key1 == key2) {
		return 1;
	}

	return 0;
}

/*
 * Hash function compare for number value.
 */
LTTNG_HIDDEN
int hash_match_key_u64(const void *key1, const void *key2)
{
	if (*(const uint64_t *) key1 == *(const uint64_t *) key2) {
		return 1;
	}

	return 0;
}

/*
 * Hash compare function for string.
 */
LTTNG_HIDDEN
int hash_match_key_str(const void *key1, const void *key2)
{
	if (strcmp(key1, key2) == 0) {
		return 1;
	}

	return 0;
}

/*
 * Hash function compare two uint64_t.
 */
LTTNG_HIDDEN
int hash_match_key_two_u64(const void *key1, const void *key2)
{
	const struct lttng_ht_two_u64 *k1 =
		(const struct lttng_ht_two_u64 *) key1;
	const struct lttng_ht_two_u64 *k2 =
		(const struct lttng_ht_two_u64 *) key2;

	if (hash_match_key_u64(&k1->key1, &k2->key1) &&
			hash_match_key_u64(&k1->key2, &k2->key2)) {
		return 1;
	}

	return 0;
}
Commit	Line	Data
819dc7d4	1	/*
66c60361	2	* Copyright (C) - Bob Jenkins, May 2006
819dc7d4	3	* Copyright (C) 2011 - David Goulet <david.goulet@polymtl.ca>
bec39940	4	* Copyright (C) 2011 - Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
819dc7d4	5	*
d14d33bf AM	6	* This program is free software; you can redistribute it and/or modify
	7	* it under the terms of the GNU General Public License, version 2 only,
	8	* as published by the Free Software Foundation.
66c60361 DG	9	*
	10	* This program is distributed in the hope that it will be useful, but WITHOUT
	11	* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
	12	* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
	13	* more details.
	14	*
d14d33bf AM	15	* You should have received a copy of the GNU General Public License along
	16	* with this program; if not, write to the Free Software Foundation, Inc.,
	17	* 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
66c60361 DG	18	*/
	19
	20	/*
819dc7d4 DG	21	* These are functions for producing 32-bit hashes for hash table lookup.
	22	* hashword(), hashlittle(), hashlittle2(), hashbig(), mix(), and final() are
	23	* externally useful functions. Routines to test the hash are included if
	24	* SELF_TEST is defined. You can use this free for any purpose. It's in the
	25	* public domain. It has no warranty.
	26	*
	27	* You probably want to use hashlittle(). hashlittle() and hashbig() hash byte
	28	* arrays. hashlittle() is is faster than hashbig() on little-endian machines.
	29	* Intel and AMD are little-endian machines. On second thought, you probably
	30	* want hashlittle2(), which is identical to hashlittle() except it returns two
	31	* 32-bit hashes for the price of one. You could implement hashbig2() if you
	32	* wanted but I haven't bothered here.
	33	*
	34	* If you want to find a hash of, say, exactly 7 integers, do
	35	* a = i1; b = i2; c = i3;
	36	* mix(a,b,c);
	37	* a += i4; b += i5; c += i6;
	38	* mix(a,b,c);
	39	* a += i7;
	40	* final(a,b,c);
	41	* then use c as the hash value. If you have a variable length array of
	42	* 4-byte integers to hash, use hashword(). If you have a byte array (like
	43	* a character string), use hashlittle(). If you have several byte arrays, or
	44	* a mix of things, see the comments above hashlittle().
	45	*
	46	* Why is this so big? I read 12 bytes at a time into 3 4-byte integers, then
	47	* mix those integers. This is fast (you can do a lot more thorough mixing
	48	* with 12*3 instructions on 3 integers than you can with 3 instructions on 1
	49	* byte), but shoehorning those bytes into integers efficiently is messy.
	50	*/
890d8fe4	51
6c1c0768	52	#define _LGPL_SOURCE
bec39940	53	#include <assert.h>
bec39940 DG	54	#include <stdint.h> /* defines uint32_t etc */
bec39940 DG	55	#include <stdio.h> /* defines printf for tests */
819dc7d4	56	#include <string.h>
bec39940 DG	57	#include <sys/param.h> /* attempt to define endianness */
bec39940 DG	58	#include <time.h> /* defines time_t for timings in the test */
0df502fd	59	#include <urcu/compiler.h>
819dc7d4	60
bec39940	61	#include "utils.h"
eb71a0aa	62	#include <common/compat/endian.h> /* attempt to define endianness */
90e535ef	63	#include <common/common.h>
3c4599b9	64	#include <common/hashtable/hashtable.h>
bec39940	65
819dc7d4 DG	66	/*
	67	* My best guess at if you are big-endian or little-endian. This may
	68	* need adjustment.
	69	*/
	70	#if (defined(__BYTE_ORDER) && defined(__LITTLE_ENDIAN) && \
	71	__BYTE_ORDER == __LITTLE_ENDIAN) \|\| \
	72	(defined(i386) \|\| defined(__i386__) \|\| defined(__i486__) \|\| \
	73	defined(__i586__) \|\| defined(__i686__) \|\| defined(vax) \|\| defined(MIPSEL))
	74	# define HASH_LITTLE_ENDIAN 1
	75	# define HASH_BIG_ENDIAN 0
	76	#elif (defined(__BYTE_ORDER) && defined(__BIG_ENDIAN) && \
	77	__BYTE_ORDER == __BIG_ENDIAN) \|\| \
	78	(defined(sparc) \|\| defined(POWERPC) \|\| defined(mc68000) \|\| defined(sel))
	79	# define HASH_LITTLE_ENDIAN 0
	80	# define HASH_BIG_ENDIAN 1
	81	#else
	82	# define HASH_LITTLE_ENDIAN 0
	83	# define HASH_BIG_ENDIAN 0
	84	#endif
	85
	86	#define hashsize(n) ((uint32_t)1<<(n))
	87	#define hashmask(n) (hashsize(n)-1)
	88	#define rot(x,k) (((x)<<(k)) \| ((x)>>(32-(k))))
	89
	90	/*
	91	* mix -- mix 3 32-bit values reversibly.
	92	*
	93	* This is reversible, so any information in (a,b,c) before mix() is
	94	* still in (a,b,c) after mix().
	95	*
	96	* If four pairs of (a,b,c) inputs are run through mix(), or through
	97	* mix() in reverse, there are at least 32 bits of the output that
	98	* are sometimes the same for one pair and different for another pair.
	99	* This was tested for:
	100	* * pairs that differed by one bit, by two bits, in any combination
	101	* of top bits of (a,b,c), or in any combination of bottom bits of
	102	* (a,b,c).
	103	* * "differ" is defined as +, -, ^, or ~^. For + and -, I transformed
	104	* the output delta to a Gray code (a^(a>>1)) so a string of 1's (as
	105	* is commonly produced by subtraction) look like a single 1-bit
	106	* difference.
	107	* * the base values were pseudorandom, all zero but one bit set, or
	108	* all zero plus a counter that starts at zero.
	109	*
	110	* Some k values for my "a-=c; a^=rot(c,k); c+=b;" arrangement that
	111	* satisfy this are
	112	* 4 6 8 16 19 4
	113	* 9 15 3 18 27 15
	114	* 14 9 3 7 17 3
	115	* Well, "9 15 3 18 27 15" didn't quite get 32 bits diffing
	116	* for "differ" defined as + with a one-bit base and a two-bit delta. I
	117	* used http://burtleburtle.net/bob/hash/avalanche.html to choose
	118	* the operations, constants, and arrangements of the variables.
	119	*
	120	* This does not achieve avalanche. There are input bits of (a,b,c)
	121	* that fail to affect some output bits of (a,b,c), especially of a. The
	122	* most thoroughly mixed value is c, but it doesn't really even achieve
	123	* avalanche in c.
	124	*
	125	* This allows some parallelism. Read-after-writes are good at doubling
	126	* the number of bits affected, so the goal of mixing pulls in the opposite
	127	* direction as the goal of parallelism. I did what I could. Rotates
	128	* seem to cost as much as shifts on every machine I could lay my hands
	129	* on, and rotates are much kinder to the top and bottom bits, so I used
130	* rotates.
131	*/
132	#define mix(a,b,c) \
133	{ \
134	a -= c; a ^= rot(c, 4); c += b; \
135	b -= a; b ^= rot(a, 6); a += c; \
136	c -= b; c ^= rot(b, 8); b += a; \
137	a -= c; a ^= rot(c,16); c += b; \
138	b -= a; b ^= rot(a,19); a += c; \
139	c -= b; c ^= rot(b, 4); b += a; \
140	}
141
142	/*
143	* final -- final mixing of 3 32-bit values (a,b,c) into c
144	*
145	* Pairs of (a,b,c) values differing in only a few bits will usually
146	* produce values of c that look totally different. This was tested for
147	* * pairs that differed by one bit, by two bits, in any combination
148	* of top bits of (a,b,c), or in any combination of bottom bits of
149	* (a,b,c).
150	* * "differ" is defined as +, -, ^, or ~^. For + and -, I transformed
151	* the output delta to a Gray code (a^(a>>1)) so a string of 1's (as
152	* is commonly produced by subtraction) look like a single 1-bit
153	* difference.
154	* * the base values were pseudorandom, all zero but one bit set, or
155	* all zero plus a counter that starts at zero.
156	*
157	* These constants passed:
158	* 14 11 25 16 4 14 24
159	* 12 14 25 16 4 14 24
160	* and these came close:
161	* 4 8 15 26 3 22 24
162	* 10 8 15 26 3 22 24
163	* 11 8 15 26 3 22 24
164	*/
165	#define final(a,b,c) \
166	{ \
167	c ^= b; c -= rot(b,14); \
168	a ^= c; a -= rot(c,11); \
169	b ^= a; b -= rot(a,25); \
170	c ^= b; c -= rot(b,16); \
171	a ^= c; a -= rot(c,4); \
172	b ^= a; b -= rot(a,14); \
173	c ^= b; c -= rot(b,24); \
174	}
175
bec39940 DG	176	/*
	177	* k - the key, an array of uint32_t values
	178	* length - the length of the key, in uint32_ts
	179	* initval - the previous hash, or an arbitrary value
	180	*/
	181	static uint32_t __attribute__((unused)) hashword(const uint32_t *k,
	182	size_t length, uint32_t initval)
0df502fd MD	183	{
	184	uint32_t a, b, c;
	185
	186	/* Set up the internal state */
	187	a = b = c = 0xdeadbeef + (((uint32_t) length) << 2) + initval;
	188
	189	/----------------------------------------- handle most of the key /
	190	while (length > 3) {
	191	a += k[0];
	192	b += k[1];
	193	c += k[2];
	194	mix(a, b, c);
	195	length -= 3;
	196	k += 3;
	197	}
	198
	199	/----------------------------------- handle the last 3 uint32_t's /
	200	switch (length) { /* all the case statements fall through */
	201	case 3: c += k[2];
	202	case 2: b += k[1];
	203	case 1: a += k[0];
	204	final(a, b, c);
	205	case 0: /* case 0: nothing left to add */
	206	break;
	207	}
	208	/---------------------------------------------- report the result /
	209	return c;
	210	}
	211
	212
819dc7d4 DG	213	/*
	214	* hashword2() -- same as hashword(), but take two seeds and return two 32-bit
	215	* values. pc and pb must both be nonnull, and pc and pb must both be
	216	* initialized with seeds. If you pass in (pb)==0, the output (pc) will be
	217	* the same as the return value from hashword().
	218	*/
bec39940	219	static void __attribute__((unused)) hashword2(const uint32_t *k, size_t length,
819dc7d4 DG	220	uint32_t pc, uint32_t pb)
	221	{
	222	uint32_t a, b, c;
	223
	224	/* Set up the internal state */
	225	a = b = c = 0xdeadbeef + ((uint32_t) (length << 2)) + *pc;
	226	c += *pb;
	227
	228	while (length > 3) {
	229	a += k[0];
	230	b += k[1];
	231	c += k[2];
	232	mix(a, b, c);
	233	length -= 3;
	234	k += 3;
	235	}
	236
	237	switch (length) {
	238	case 3 :
	239	c += k[2];
	240	case 2 :
	241	b += k[1];
	242	case 1 :
	243	a += k[0];
	244	final(a, b, c);
	245	case 0: /* case 0: nothing left to add */
	246	break;
	247	}
	248
	249	*pc = c;
	250	*pb = b;
	251	}
	252
	253	/*
	254	* hashlittle() -- hash a variable-length key into a 32-bit value
	255	* k : the key (the unaligned variable-length array of bytes)
	256	* length : the length of the key, counting by bytes
	257	* initval : can be any 4-byte value
	258	* Returns a 32-bit value. Every bit of the key affects every bit of
	259	* the return value. Two keys differing by one or two bits will have
	260	* totally different hash values.
	261	*
	262	* The best hash table sizes are powers of 2. There is no need to do
	263	* mod a prime (mod is sooo slow!). If you need less than 32 bits,
	264	* use a bitmask. For example, if you need only 10 bits, do
	265	* h = (h & hashmask(10));
	266	* In which case, the hash table should have hashsize(10) elements.
	267	*
	268	* If you are hashing n strings (uint8_t **)k, do it like this:
	269	* for (i=0, h=0; i<n; ++i) h = hashlittle( k[i], len[i], h);
	270	*
	271	* By Bob Jenkins, 2006. bob_jenkins@burtleburtle.net. You may use this
	272	* code any way you wish, private, educational, or commercial. It's free.
	273	*
	274	* Use for hash table lookup, or anything where one collision in 2^^32 is
	275	* acceptable. Do NOT use for cryptographic purposes.
	276	*/
bec39940 DG	277	static uint32_t __attribute__((unused)) hashlittle(const void *key,
bec39940 DG	278	size_t length, uint32_t initval)
819dc7d4 DG	279	{
	280	uint32_t a,b,c;
	281	union {
	282	const void *ptr;
	283	size_t i;
	284	} u; /* needed for Mac Powerbook G4 */
	285
	286	/* Set up the internal state */
	287	a = b = c = 0xdeadbeef + ((uint32_t)length) + initval;
	288
	289	u.ptr = key;
	290	if (HASH_LITTLE_ENDIAN && ((u.i & 0x3) == 0)) {
	291	const uint32_t k = (const uint32_t )key; /* read 32-bit chunks */
	292
	293	/------ all but last block: aligned reads and affect 32 bits of (a,b,c) /
	294	while (length > 12) {
	295	a += k[0];
	296	b += k[1];
	297	c += k[2];
	298	mix(a,b,c);
	299	length -= 12;
	300	k += 3;
	301	}
	302
	303	/*
	304	* "k[2]&0xffffff" actually reads beyond the end of the string, but
	305	* then masks off the part it's not allowed to read. Because the
	306	* string is aligned, the masked-off tail is in the same word as the
	307	* rest of the string. Every machine with memory protection I've seen
	308	* does it on word boundaries, so is OK with this. But VALGRIND will
	309	* still catch it and complain. The masking trick does make the hash
	310	* noticably faster for short strings (like English words).
	311	*/
	312	#ifndef VALGRIND
	313
	314	switch (length) {
	315	case 12: c+=k[2]; b+=k[1]; a+=k[0]; break;
	316	case 11: c+=k[2]&0xffffff; b+=k[1]; a+=k[0]; break;
	317	case 10: c+=k[2]&0xffff; b+=k[1]; a+=k[0]; break;
	318	case 9 : c+=k[2]&0xff; b+=k[1]; a+=k[0]; break;
	319	case 8 : b+=k[1]; a+=k[0]; break;
	320	case 7 : b+=k[1]&0xffffff; a+=k[0]; break;
	321	case 6 : b+=k[1]&0xffff; a+=k[0]; break;
	322	case 5 : b+=k[1]&0xff; a+=k[0]; break;
	323	case 4 : a+=k[0]; break;
	324	case 3 : a+=k[0]&0xffffff; break;
	325	case 2 : a+=k[0]&0xffff; break;
	326	case 1 : a+=k[0]&0xff; break;
	327	case 0 : return c; /* zero length strings require no mixing */
	328	}
	329	#else /* make valgrind happy */
	330	const uint8_t *k8;
	331
	332	k8 = (const uint8_t *)k;
	333	switch (length) {
	334	case 12: c+=k[2]; b+=k[1]; a+=k[0]; break;
	335	case 11: c+=((uint32_t)k8[10])<<16; /* fall through */
	336	case 10: c+=((uint32_t)k8[9])<<8; /* fall through */
	337	case 9 : c+=k8[8]; /* fall through */
	338	case 8 : b+=k[1]; a+=k[0]; break;
	339	case 7 : b+=((uint32_t)k8[6])<<16; /* fall through */
	340	case 6 : b+=((uint32_t)k8[5])<<8; /* fall through */
	341	case 5 : b+=k8[4]; /* fall through */
	342	case 4 : a+=k[0]; break;
343	case 3 : a+=((uint32_t)k8[2])<<16; /* fall through */
344	case 2 : a+=((uint32_t)k8[1])<<8; /* fall through */
345	case 1 : a+=k8[0]; break;
346	case 0 : return c;
347	}
348	#endif /* !valgrind */
349	} else if (HASH_LITTLE_ENDIAN && ((u.i & 0x1) == 0)) {
350	const uint16_t k = (const uint16_t )key; /* read 16-bit chunks */
351	const uint8_t *k8;
352
353	/--------------- all but last block: aligned reads and different mixing /
354	while (length > 12) {
355	a += k[0] + (((uint32_t)k[1])<<16);
356	b += k[2] + (((uint32_t)k[3])<<16);
357	c += k[4] + (((uint32_t)k[5])<<16);
358	mix(a,b,c);
359	length -= 12;
360	k += 6;
361	}
362
363	k8 = (const uint8_t *)k;
364	switch (length) {
365	case 12:
366	c+=k[4]+(((uint32_t)k[5])<<16);
367	b+=k[2]+(((uint32_t)k[3])<<16);
368	a+=k[0]+(((uint32_t)k[1])<<16);
369	break;
370	case 11:
371	c+=((uint32_t)k8[10])<<16; /* fall through */
372	case 10:
373	c+=k[4];
374	b+=k[2]+(((uint32_t)k[3])<<16);
375	a+=k[0]+(((uint32_t)k[1])<<16);
376	break;
377	case 9:
378	c+=k8[8]; /* fall through */
379	case 8:
380	b+=k[2]+(((uint32_t)k[3])<<16);
381	a+=k[0]+(((uint32_t)k[1])<<16);
382	break;
383	case 7:
384	b+=((uint32_t)k8[6])<<16; /* fall through */
385	case 6:
386	b+=k[2];
387	a+=k[0]+(((uint32_t)k[1])<<16);
388	break;
389	case 5:
390	b+=k8[4]; /* fall through */
391	case 4:
392	a+=k[0]+(((uint32_t)k[1])<<16);
393	break;
394	case 3:
395	a+=((uint32_t)k8[2])<<16; /* fall through */
396	case 2:
397	a+=k[0];
398	break;
399	case 1:
400	a+=k8[0];
401	break;
402	case 0:
403	return c; /* zero length requires no mixing */
404	}
405
406	} else { /* need to read the key one byte at a time */
407	const uint8_t k = (const uint8_t )key;
408
409	while (length > 12) {
410	a += k[0];
411	a += ((uint32_t)k[1])<<8;
412	a += ((uint32_t)k[2])<<16;
413	a += ((uint32_t)k[3])<<24;
414	b += k[4];
415	b += ((uint32_t)k[5])<<8;
416	b += ((uint32_t)k[6])<<16;
417	b += ((uint32_t)k[7])<<24;
418	c += k[8];
419	c += ((uint32_t)k[9])<<8;
420	c += ((uint32_t)k[10])<<16;
421	c += ((uint32_t)k[11])<<24;
422	mix(a,b,c);
423	length -= 12;
424	k += 12;
425	}
426
427	switch(length) { /* all the case statements fall through */
428	case 12: c+=((uint32_t)k[11])<<24;
429	case 11: c+=((uint32_t)k[10])<<16;
430	case 10: c+=((uint32_t)k[9])<<8;
431	case 9: c+=k[8];
432	case 8: b+=((uint32_t)k[7])<<24;
433	case 7: b+=((uint32_t)k[6])<<16;
434	case 6: b+=((uint32_t)k[5])<<8;
435	case 5: b+=k[4];
436	case 4: a+=((uint32_t)k[3])<<24;
437	case 3: a+=((uint32_t)k[2])<<16;
438	case 2: a+=((uint32_t)k[1])<<8;
439	case 1:
440	a+=k[0];
441	break;
442	case 0:
443	return c;
444	}
445	}
446
447	final(a,b,c);
448	return c;
449	}
450
90e535ef	451	LTTNG_HIDDEN
bcd52dd9	452	unsigned long hash_key_u64(const void *_key, unsigned long seed)
0df502fd MD	453	{
	454	union {
	455	uint64_t v64;
	456	uint32_t v32[2];
	457	} v;
	458	union {
	459	uint64_t v64;
	460	uint32_t v32[2];
	461	} key;
	462
0df502fd	463	v.v64 = (uint64_t) seed;
bcd52dd9	464	key.v64 = (const uint64_t ) _key;
0df502fd MD	465	hashword2(key.v32, 2, &v.v32[0], &v.v32[1]);
	466	return v.v64;
	467	}
d88aee68 DG	468
	469	#if (CAA_BITS_PER_LONG == 64)
	470	/*
	471	* Hash function for number value.
9a2746aa	472	* Pass the value itself as the key, not its address.
d88aee68 DG	473	*/
d88aee68 DG	474	LTTNG_HIDDEN
bcd52dd9	475	unsigned long hash_key_ulong(const void *_key, unsigned long seed)
d88aee68 DG	476	{
	477	uint64_t __key = (uint64_t) _key;
	478	return (unsigned long) hash_key_u64(&__key, seed);
	479	}
0df502fd	480	#else
819dc7d4 DG	481	/*
819dc7d4 DG	482	* Hash function for number value.
9a2746aa	483	* Pass the value itself as the key, not its address.
819dc7d4	484	*/
90e535ef	485	LTTNG_HIDDEN
bcd52dd9	486	unsigned long hash_key_ulong(const void *_key, unsigned long seed)
819dc7d4	487	{
8da9ba32	488	uint32_t key = (uint32_t) _key;
0df502fd	489
0df502fd	490	return hashword(&key, 1, seed);
819dc7d4	491	}
bec39940	492	#endif /* CAA_BITS_PER_LONG */
819dc7d4 DG	493
	494	/*
	495	* Hash function for string.
	496	*/
90e535ef	497	LTTNG_HIDDEN
bcd52dd9	498	unsigned long hash_key_str(const void *key, unsigned long seed)
819dc7d4	499	{
bcd52dd9	500	return hashlittle(key, strlen((const char *) key), seed);
819dc7d4 DG	501	}
819dc7d4 DG	502
3c4599b9 JD	503	/*
	504	* Hash function for two uint64_t.
	505	*/
	506	LTTNG_HIDDEN
bcd52dd9	507	unsigned long hash_key_two_u64(const void *key, unsigned long seed)
3c4599b9	508	{
bcd52dd9 JG	509	const struct lttng_ht_two_u64 *k =
bcd52dd9 JG	510	(const struct lttng_ht_two_u64 *) key;
3c4599b9 JD	511
	512	return hash_key_u64(&k->key1, seed) ^ hash_key_u64(&k->key2, seed);
	513	}
	514
819dc7d4 DG	515	/*
	516	* Hash function compare for number value.
	517	*/
90e535ef	518	LTTNG_HIDDEN
bcd52dd9	519	int hash_match_key_ulong(const void key1, const void key2)
819dc7d4	520	{
819dc7d4	521	if (key1 == key2) {
bec39940	522	return 1;
819dc7d4 DG	523	}
819dc7d4 DG	524
bec39940	525	return 0;
819dc7d4 DG	526	}
819dc7d4 DG	527
d88aee68 DG	528	/*
	529	* Hash function compare for number value.
	530	*/
	531	LTTNG_HIDDEN
bcd52dd9	532	int hash_match_key_u64(const void key1, const void key2)
d88aee68	533	{
bcd52dd9	534	if ((const uint64_t ) key1 == (const uint64_t ) key2) {
d88aee68 DG	535	return 1;
	536	}
	537
	538	return 0;
	539	}
	540
819dc7d4 DG	541	/*
	542	* Hash compare function for string.
	543	*/
90e535ef	544	LTTNG_HIDDEN
bcd52dd9	545	int hash_match_key_str(const void key1, const void key2)
819dc7d4	546	{
bec39940 DG	547	if (strcmp(key1, key2) == 0) {
bec39940 DG	548	return 1;
819dc7d4 DG	549	}
819dc7d4 DG	550
bec39940	551	return 0;
819dc7d4	552	}
3c4599b9 JD	553
	554	/*
	555	* Hash function compare two uint64_t.
	556	*/
	557	LTTNG_HIDDEN
bcd52dd9	558	int hash_match_key_two_u64(const void key1, const void key2)
3c4599b9	559	{
bcd52dd9 JG	560	const struct lttng_ht_two_u64 *k1 =
	561	(const struct lttng_ht_two_u64 *) key1;
	562	const struct lttng_ht_two_u64 *k2 =
	563	(const struct lttng_ht_two_u64 *) key2;
3c4599b9 JD	564
	565	if (hash_match_key_u64(&k1->key1, &k2->key1) &&
	566	hash_match_key_u64(&k1->key2, &k2->key2)) {
	567	return 1;
	568	}
	569
	570	return 0;
	571	}