Browse Source

Update wyhash to final4

pull/1232/head
Antonio SJ Musumeci 1 year ago
parent
commit
fdecabfaab
  1. 1
      DEPENDENCIES
  2. 5
      src/fs_mktemp.cpp
  3. 4
      src/gidcache.cpp
  4. 6
      src/rnd.cpp
  5. 195
      src/wyhash.h

1
DEPENDENCIES

@ -3,6 +3,7 @@
### included in repo ### included in repo
* libfuse: https://github.com/libfuse/libfuse (heavily modified fork of v2.x) * libfuse: https://github.com/libfuse/libfuse (heavily modified fork of v2.x)
* wyhash: https://github.com/wangyi-fudan/wyhash
* ghc::filesystem: https://github.com/gulrak/filesystem * ghc::filesystem: https://github.com/gulrak/filesystem
* nonstd::optional: https://github.com/martinmoene/optional-lite * nonstd::optional: https://github.com/martinmoene/optional-lite
* fmt: https://github.com/fmtlib/fmt * fmt: https://github.com/fmtlib/fmt

5
src/fs_mktemp.cpp

@ -19,6 +19,7 @@
#include "errno.hpp" #include "errno.hpp"
#include "fs_open.hpp" #include "fs_open.hpp"
#include "fs_path.hpp" #include "fs_path.hpp"
#include "rnd.hpp"
#include <limits.h> #include <limits.h>
@ -29,6 +30,8 @@
#define MAX_ATTEMPTS 3 #define MAX_ATTEMPTS 3
static char const CHARS[] = "0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz"; static char const CHARS[] = "0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz";
static size_t const CHARS_SIZE = (sizeof(CHARS) - 1);
namespace l namespace l
{ {
@ -41,7 +44,7 @@ namespace l
filename = '.'; filename = '.';
for(int i = 0; i < PAD_LEN; i++) for(int i = 0; i < PAD_LEN; i++)
filename += CHARS[std::rand() % (sizeof(CHARS) - 1)];
filename += CHARS[RND::rand64(CHARS_SIZE)];
path = base_; path = base_;
path /= filename; path /= filename;

4
src/gidcache.cpp

@ -14,6 +14,8 @@
OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
*/ */
#include "rnd.hpp"
#include <grp.h> #include <grp.h>
#include <pwd.h> #include <pwd.h>
#include <stdlib.h> #include <stdlib.h>
@ -81,7 +83,7 @@ GIDRecord *
GIDCache::allocrec(void) GIDCache::allocrec(void)
{ {
if(size == MAXRECS) if(size == MAXRECS)
return &recs[rand() % MAXRECS];
return &recs[RND::rand64(MAXRECS)];
else else
return &recs[size++]; return &recs[size++];
} }

6
src/rnd.cpp

@ -26,9 +26,11 @@
static uint64_t G_SEED; static uint64_t G_SEED;
static RND G_RND;
RND::RND()
__attribute__((constructor))
static
void
_constructor()
{ {
struct timeval tv; struct timeval tv;

195
src/wyhash.h

@ -1,11 +1,29 @@
//Author: Wang Yi <godspeed_china@yeah.net>
#ifndef wyhash_final_version
#define wyhash_final_version
//defines that change behavior
// This is free and unencumbered software released into the public domain under The Unlicense (http://unlicense.org/)
// main repo: https://github.com/wangyi-fudan/wyhash
// author: Wang Yi <godspeed_china@yeah.net>
// contributors: Reini Urban, Dietrich Epp, Joshua Haberman, Tommy Ettinger, Daniel Lemire, Otmar Ertl, cocowalla, leo-yuriev, Diego Barrios Romero, paulie-g, dumblob, Yann Collet, ivte-ms, hyb, James Z.M. Gao, easyaspi314 (Devin), TheOneric
/* quick example:
string s="fjsakfdsjkf";
uint64_t hash=wyhash(s.c_str(), s.size(), 0, _wyp);
*/
#ifndef wyhash_final_version_4
#define wyhash_final_version_4
#ifndef WYHASH_CONDOM #ifndef WYHASH_CONDOM
#define WYHASH_CONDOM 1 //0: read 8 bytes before and after boundaries, dangerous but faster. 1: normal valid behavior 2: extra protection against entropy loss (probability=2^-63), aka. "blind multiplication"
//protections that produce different results:
//1: normal valid behavior
//2: extra protection against entropy loss (probability=2^-63), aka. "blind multiplication"
#define WYHASH_CONDOM 1
#endif #endif
#define WYHASH_32BIT_MUM 0 //faster on 32 bit system
#ifndef WYHASH_32BIT_MUM
//0: normal version, slow on 32 bit systems
//1: faster on 32 bit systems but produces different results, incompatible with wy2u0k function
#define WYHASH_32BIT_MUM 0
#endif
//includes //includes
#include <stdint.h> #include <stdint.h>
#include <string.h> #include <string.h>
@ -13,6 +31,8 @@
#include <intrin.h> #include <intrin.h>
#pragma intrinsic(_umul128) #pragma intrinsic(_umul128)
#endif #endif
//likely and unlikely macros
#if defined(__GNUC__) || defined(__INTEL_COMPILER) || defined(__clang__) #if defined(__GNUC__) || defined(__INTEL_COMPILER) || defined(__clang__)
#define _likely_(x) __builtin_expect(x,1) #define _likely_(x) __builtin_expect(x,1)
#define _unlikely_(x) __builtin_expect(x,0) #define _unlikely_(x) __builtin_expect(x,0)
@ -20,11 +40,12 @@
#define _likely_(x) (x) #define _likely_(x) (x)
#define _unlikely_(x) (x) #define _unlikely_(x) (x)
#endif #endif
//mum function
//128bit multiply function
static inline uint64_t _wyrot(uint64_t x) { return (x>>32)|(x<<32); } static inline uint64_t _wyrot(uint64_t x) { return (x>>32)|(x<<32); }
static inline void _wymum(uint64_t *A, uint64_t *B){ static inline void _wymum(uint64_t *A, uint64_t *B){
#if(WYHASH_32BIT_MUM) #if(WYHASH_32BIT_MUM)
uint64_t hh=(*A>>32)*(*B>>32), hl=(*A>>32)*(unsigned)*B, lh=(unsigned)*A*(*B>>32), ll=(uint64_t)(unsigned)*A*(unsigned)*B;
uint64_t hh=(*A>>32)*(*B>>32), hl=(*A>>32)*(uint32_t)*B, lh=(uint32_t)*A*(*B>>32), ll=(uint64_t)(uint32_t)*A*(uint32_t)*B;
#if(WYHASH_CONDOM>1) #if(WYHASH_CONDOM>1)
*A^=_wyrot(hl)^hh; *B^=_wyrot(lh)^ll; *A^=_wyrot(hl)^hh; *B^=_wyrot(lh)^ll;
#else #else
@ -56,85 +77,155 @@ static inline void _wymum(uint64_t *A, uint64_t *B){
#endif #endif
#endif #endif
} }
//multiply and xor mix function, aka MUM
static inline uint64_t _wymix(uint64_t A, uint64_t B){ _wymum(&A,&B); return A^B; } static inline uint64_t _wymix(uint64_t A, uint64_t B){ _wymum(&A,&B); return A^B; }
//read functions
//endian macros
#ifndef WYHASH_LITTLE_ENDIAN #ifndef WYHASH_LITTLE_ENDIAN
#if defined(_WIN32) || defined(__LITTLE_ENDIAN__) || (defined(__BYTE_ORDER__) && __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__) #if defined(_WIN32) || defined(__LITTLE_ENDIAN__) || (defined(__BYTE_ORDER__) && __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__)
#define WYHASH_LITTLE_ENDIAN 1 #define WYHASH_LITTLE_ENDIAN 1
#elif defined(__BIG_ENDIAN__) || (defined(__BYTE_ORDER__) && __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__) #elif defined(__BIG_ENDIAN__) || (defined(__BYTE_ORDER__) && __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__)
#define WYHASH_LITTLE_ENDIAN 0 #define WYHASH_LITTLE_ENDIAN 0
#else
#warning could not determine endianness! Falling back to little endian.
#define WYHASH_LITTLE_ENDIAN 1
#endif #endif
#endif #endif
//read functions
#if (WYHASH_LITTLE_ENDIAN) #if (WYHASH_LITTLE_ENDIAN)
static inline uint64_t _wyr8(const uint8_t *p) { uint64_t v; memcpy(&v, p, 8); return v;} static inline uint64_t _wyr8(const uint8_t *p) { uint64_t v; memcpy(&v, p, 8); return v;}
static inline uint64_t _wyr4(const uint8_t *p) { unsigned v; memcpy(&v, p, 4); return v;}
static inline uint64_t _wyr4(const uint8_t *p) { uint32_t v; memcpy(&v, p, 4); return v;}
#elif defined(__GNUC__) || defined(__INTEL_COMPILER) || defined(__clang__) #elif defined(__GNUC__) || defined(__INTEL_COMPILER) || defined(__clang__)
static inline uint64_t _wyr8(const uint8_t *p) { uint64_t v; memcpy(&v, p, 8); return __builtin_bswap64(v);} static inline uint64_t _wyr8(const uint8_t *p) { uint64_t v; memcpy(&v, p, 8); return __builtin_bswap64(v);}
static inline uint64_t _wyr4(const uint8_t *p) { unsigned v; memcpy(&v, p, 4); return __builtin_bswap32(v);}
static inline uint64_t _wyr4(const uint8_t *p) { uint32_t v; memcpy(&v, p, 4); return __builtin_bswap32(v);}
#elif defined(_MSC_VER) #elif defined(_MSC_VER)
static inline uint64_t _wyr8(const uint8_t *p) { uint64_t v; memcpy(&v, p, 8); return _byteswap_uint64(v);} static inline uint64_t _wyr8(const uint8_t *p) { uint64_t v; memcpy(&v, p, 8); return _byteswap_uint64(v);}
static inline uint64_t _wyr4(const uint8_t *p) { unsigned v; memcpy(&v, p, 4); return _byteswap_ulong(v);}
#endif
static inline uint64_t _wyr3(const uint8_t *p, unsigned k) { return (((uint64_t)p[0])<<16)|(((uint64_t)p[k>>1])<<8)|p[k-1];}
//wyhash function
static inline uint64_t _wyfinish16(const uint8_t *p, uint64_t len, uint64_t seed, const uint64_t *secret, uint64_t i){
#if(WYHASH_CONDOM>0)
uint64_t a, b;
if(_likely_(i<=8)){
if(_likely_(i>=4)){ a=_wyr4(p); b=_wyr4(p+i-4); }
else if (_likely_(i)){ a=_wyr3(p,i); b=0; }
else a=b=0;
}
else{ a=_wyr8(p); b=_wyr8(p+i-8); }
return _wymix(secret[1]^len,_wymix(a^secret[1], b^seed));
static inline uint64_t _wyr4(const uint8_t *p) { uint32_t v; memcpy(&v, p, 4); return _byteswap_ulong(v);}
#else #else
#define oneshot_shift ((i<8)*((8-i)<<3))
return _wymix(secret[1]^len,_wymix((_wyr8(p)<<oneshot_shift)^secret[1],(_wyr8(p+i-8)>>oneshot_shift)^seed));
#endif
static inline uint64_t _wyr8(const uint8_t *p) {
uint64_t v; memcpy(&v, p, 8);
return (((v >> 56) & 0xff)| ((v >> 40) & 0xff00)| ((v >> 24) & 0xff0000)| ((v >> 8) & 0xff000000)| ((v << 8) & 0xff00000000)| ((v << 24) & 0xff0000000000)| ((v << 40) & 0xff000000000000)| ((v << 56) & 0xff00000000000000));
} }
static inline uint64_t _wyfinish(const uint8_t *p, uint64_t len, uint64_t seed, const uint64_t *secret, uint64_t i){
if(_likely_(i<=16)) return _wyfinish16(p,len,seed,secret,i);
return _wyfinish(p+16,len,_wymix(_wyr8(p)^secret[1],_wyr8(p+8)^seed),secret,i-16);
static inline uint64_t _wyr4(const uint8_t *p) {
uint32_t v; memcpy(&v, p, 4);
return (((v >> 24) & 0xff)| ((v >> 8) & 0xff00)| ((v << 8) & 0xff0000)| ((v << 24) & 0xff000000));
} }
static inline uint64_t wyhash(const void *key, uint64_t len, uint64_t seed, const uint64_t *secret){
const uint8_t *p=(const uint8_t *)key;
uint64_t i=len; seed^=*secret;
if(_unlikely_(i>64)){
uint64_t see1=seed;
#endif
static inline uint64_t _wyr3(const uint8_t *p, size_t k) { return (((uint64_t)p[0])<<16)|(((uint64_t)p[k>>1])<<8)|p[k-1];}
//wyhash main function
static inline uint64_t wyhash(const void *key, size_t len, uint64_t seed, const uint64_t *secret){
const uint8_t *p=(const uint8_t *)key; seed^=_wymix(seed^secret[0],secret[1]); uint64_t a, b;
if(_likely_(len<=16)){
if(_likely_(len>=4)){ a=(_wyr4(p)<<32)|_wyr4(p+((len>>3)<<2)); b=(_wyr4(p+len-4)<<32)|_wyr4(p+len-4-((len>>3)<<2)); }
else if(_likely_(len>0)){ a=_wyr3(p,len); b=0;}
else a=b=0;
}
else{
size_t i=len;
if(_unlikely_(i>48)){
uint64_t see1=seed, see2=seed;
do{ do{
seed=_wymix(_wyr8(p)^secret[1],_wyr8(p+8)^seed)^_wymix(_wyr8(p+16)^secret[2],_wyr8(p+24)^seed);
see1=_wymix(_wyr8(p+32)^secret[3],_wyr8(p+40)^see1)^_wymix(_wyr8(p+48)^secret[4],_wyr8(p+56)^see1);
p+=64; i-=64;
}while(i>64);
seed^=see1;
seed=_wymix(_wyr8(p)^secret[1],_wyr8(p+8)^seed);
see1=_wymix(_wyr8(p+16)^secret[2],_wyr8(p+24)^see1);
see2=_wymix(_wyr8(p+32)^secret[3],_wyr8(p+40)^see2);
p+=48; i-=48;
}while(_likely_(i>48));
seed^=see1^see2;
}
while(_unlikely_(i>16)){ seed=_wymix(_wyr8(p)^secret[1],_wyr8(p+8)^seed); i-=16; p+=16; }
a=_wyr8(p+i-16); b=_wyr8(p+i-8);
} }
return _wyfinish(p,len,seed,secret,i);
a^=secret[1]; b^=seed; _wymum(&a,&b);
return _wymix(a^secret[0]^len,b^secret[1]);
} }
//utility functions
const uint64_t _wyp[5] = {0xa0761d6478bd642full, 0xe7037ed1a0b428dbull, 0x8ebc6af09c88c6e3ull, 0x589965cc75374cc3ull, 0x1d8e4e27c47d124full};
static inline uint64_t wyhash64(uint64_t A, uint64_t B){ A^=_wyp[0]; B^=_wyp[1]; _wymum(&A,&B); return _wymix(A^_wyp[0],B^_wyp[1]);}
static inline uint64_t wyrand(uint64_t *seed){ *seed+=_wyp[0]; return _wymix(*seed,*seed^_wyp[1]);}
//the default secret parameters
static const uint64_t _wyp[4] = {0xa0761d6478bd642full, 0xe7037ed1a0b428dbull, 0x8ebc6af09c88c6e3ull, 0x589965cc75374cc3ull};
//a useful 64bit-64bit mix function to produce deterministic pseudo random numbers that can pass BigCrush and PractRand
static inline uint64_t wyhash64(uint64_t A, uint64_t B){ A^=0xa0761d6478bd642full; B^=0xe7037ed1a0b428dbull; _wymum(&A,&B); return _wymix(A^0xa0761d6478bd642full,B^0xe7037ed1a0b428dbull);}
//The wyrand PRNG that pass BigCrush and PractRand
static inline uint64_t wyrand(uint64_t *seed){ *seed+=0xa0761d6478bd642full; return _wymix(*seed,*seed^0xe7037ed1a0b428dbull);}
//convert any 64 bit pseudo random numbers to uniform distribution [0,1). It can be combined with wyrand, wyhash64 or wyhash.
static inline double wy2u01(uint64_t r){ const double _wynorm=1.0/(1ull<<52); return (r>>12)*_wynorm;} static inline double wy2u01(uint64_t r){ const double _wynorm=1.0/(1ull<<52); return (r>>12)*_wynorm;}
//convert any 64 bit pseudo random numbers to APPROXIMATE Gaussian distribution. It can be combined with wyrand, wyhash64 or wyhash.
static inline double wy2gau(uint64_t r){ const double _wynorm=1.0/(1ull<<20); return ((r&0x1fffff)+((r>>21)&0x1fffff)+((r>>42)&0x1fffff))*_wynorm-3.0;} static inline double wy2gau(uint64_t r){ const double _wynorm=1.0/(1ull<<20); return ((r&0x1fffff)+((r>>21)&0x1fffff)+((r>>42)&0x1fffff))*_wynorm-3.0;}
#ifdef WYTRNG
#include <sys/time.h>
//The wytrand true random number generator, passed BigCrush.
static inline uint64_t wytrand(uint64_t *seed){
struct timeval t; gettimeofday(&t,0);
uint64_t teed=(((uint64_t)t.tv_sec)<<32)|t.tv_usec;
teed=_wymix(teed^_wyp[0],*seed^_wyp[1]);
*seed=_wymix(teed^_wyp[0],_wyp[2]);
return _wymix(*seed,*seed^_wyp[3]);
}
#endif
#if(!WYHASH_32BIT_MUM)
//fast range integer random number generation on [0,k) credit to Daniel Lemire. May not work when WYHASH_32BIT_MUM=1. It can be combined with wyrand, wyhash64 or wyhash.
static inline uint64_t wy2u0k(uint64_t r, uint64_t k){ _wymum(&r,&k); return k; }
#endif
//make your own secret
static inline void make_secret(uint64_t seed, uint64_t *secret){ static inline void make_secret(uint64_t seed, uint64_t *secret){
uint8_t c[] = {15, 23, 27, 29, 30, 39, 43, 45, 46, 51, 53, 54, 57, 58, 60, 71, 75, 77, 78, 83, 85, 86, 89, 90, 92, 99, 101, 102, 105, 106, 108, 113, 114, 116, 120, 135, 139, 141, 142, 147, 149, 150, 153, 154, 156, 163, 165, 166, 169, 170, 172, 177, 178, 180, 184, 195, 197, 198, 201, 202, 204, 209, 210, 212, 216, 225, 226, 228, 232, 240 }; uint8_t c[] = {15, 23, 27, 29, 30, 39, 43, 45, 46, 51, 53, 54, 57, 58, 60, 71, 75, 77, 78, 83, 85, 86, 89, 90, 92, 99, 101, 102, 105, 106, 108, 113, 114, 116, 120, 135, 139, 141, 142, 147, 149, 150, 153, 154, 156, 163, 165, 166, 169, 170, 172, 177, 178, 180, 184, 195, 197, 198, 201, 202, 204, 209, 210, 212, 216, 225, 226, 228, 232, 240 };
for(size_t i=0;i<5;i++){
for(size_t i=0;i<4;i++){
uint8_t ok; uint8_t ok;
do{ do{
ok=1; secret[i]=0; ok=1; secret[i]=0;
for(size_t j=0;j<64;j+=8) secret[i]|=((uint64_t)c[wyrand(&seed)%sizeof(c)])<<j; for(size_t j=0;j<64;j+=8) secret[i]|=((uint64_t)c[wyrand(&seed)%sizeof(c)])<<j;
if(secret[i]%2==0){ ok=0; continue; } if(secret[i]%2==0){ ok=0; continue; }
for(size_t j=0;j<i;j++)
for(size_t j=0;j<i;j++) {
#if defined(__GNUC__) || defined(__INTEL_COMPILER) || defined(__clang__) #if defined(__GNUC__) || defined(__INTEL_COMPILER) || defined(__clang__)
if(__builtin_popcountll(secret[j]^secret[i])!=32){ ok=0; break; } if(__builtin_popcountll(secret[j]^secret[i])!=32){ ok=0; break; }
#elif defined(_MSC_VER) && defined(_M_X64) #elif defined(_MSC_VER) && defined(_M_X64)
if(_mm_popcnt_u64(secret[j]^secret[i])!=32){ ok=0; break; } if(_mm_popcnt_u64(secret[j]^secret[i])!=32){ ok=0; break; }
#else
//manual popcount
uint64_t x = secret[j]^secret[i];
x -= (x >> 1) & 0x5555555555555555;
x = (x & 0x3333333333333333) + ((x >> 2) & 0x3333333333333333);
x = (x + (x >> 4)) & 0x0f0f0f0f0f0f0f0f;
x = (x * 0x0101010101010101) >> 56;
if(x!=32){ ok=0; break; }
#endif #endif
if(!ok)continue;
for(uint64_t j=3;j<0x100000000ull;j+=2) if(secret[i]%j==0){ ok=0; break; }
}
}while(!ok); }while(!ok);
} }
} }
#endif #endif
/* The Unlicense
This is free and unencumbered software released into the public domain.
Anyone is free to copy, modify, publish, use, compile, sell, or
distribute this software, either in source code form or as a compiled
binary, for any purpose, commercial or non-commercial, and by any
means.
In jurisdictions that recognize copyright laws, the author or authors
of this software dedicate any and all copyright interest in the
software to the public domain. We make this dedication for the benefit
of the public at large and to the detriment of our heirs and
successors. We intend this dedication to be an overt act of
relinquishment in perpetuity of all present and future rights to this
software under copyright law.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR
OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
OTHER DEALINGS IN THE SOFTWARE.
For more information, please refer to <http://unlicense.org/>
*/
Loading…
Cancel
Save