X-Git-Url: https://perl5.git.perl.org/perl5.git/blobdiff_plain/54e07e2b21cb1f58c04d67bca2a311715ba8815e..74fe8880c4667610a3b0aa9e8167fe1b407dc5c5:/hv_func.h diff --git a/hv_func.h b/hv_func.h index 473ec46..708c6c7 100644 --- a/hv_func.h +++ b/hv_func.h @@ -21,6 +21,8 @@ || defined(PERL_HASH_FUNC_ONE_AT_A_TIME) \ || defined(PERL_HASH_FUNC_ONE_AT_A_TIME_HARD) \ || defined(PERL_HASH_FUNC_ONE_AT_A_TIME_OLD) \ + || defined(PERL_HASH_FUNC_MURMUR_HASH_64A) \ + || defined(PERL_HASH_FUNC_MURMUR_HASH_64B) \ ) #define PERL_HASH_FUNC_ONE_AT_A_TIME_HARD #endif @@ -28,38 +30,46 @@ #if defined(PERL_HASH_FUNC_SIPHASH) # define PERL_HASH_FUNC "SIPHASH_2_4" # define PERL_HASH_SEED_BYTES 16 -# define PERL_HASH(hash,str,len) (hash)= S_perl_hash_siphash_2_4(PERL_HASH_SEED,(U8*)(str),(len)) +# define PERL_HASH_WITH_SEED(seed,hash,str,len) (hash)= S_perl_hash_siphash_2_4((seed),(U8*)(str),(len)) #elif defined(PERL_HASH_FUNC_SUPERFAST) # define PERL_HASH_FUNC "SUPERFAST" # define PERL_HASH_SEED_BYTES 4 -# define PERL_HASH(hash,str,len) (hash)= S_perl_hash_superfast(PERL_HASH_SEED,(U8*)(str),(len)) +# define PERL_HASH_WITH_SEED(seed,hash,str,len) (hash)= S_perl_hash_superfast((seed),(U8*)(str),(len)) #elif defined(PERL_HASH_FUNC_MURMUR3) # define PERL_HASH_FUNC "MURMUR3" # define PERL_HASH_SEED_BYTES 4 -# define PERL_HASH(hash,str,len) (hash)= S_perl_hash_murmur3(PERL_HASH_SEED,(U8*)(str),(len)) +# define PERL_HASH_WITH_SEED(seed,hash,str,len) (hash)= S_perl_hash_murmur3((seed),(U8*)(str),(len)) #elif defined(PERL_HASH_FUNC_DJB2) # define PERL_HASH_FUNC "DJB2" # define PERL_HASH_SEED_BYTES 4 -# define PERL_HASH(hash,str,len) (hash)= S_perl_hash_djb2(PERL_HASH_SEED,(U8*)(str),(len)) +# define PERL_HASH_WITH_SEED(seed,hash,str,len) (hash)= S_perl_hash_djb2((seed),(U8*)(str),(len)) #elif defined(PERL_HASH_FUNC_SDBM) # define PERL_HASH_FUNC "SDBM" # define PERL_HASH_SEED_BYTES 4 -# define PERL_HASH(hash,str,len) (hash)= S_perl_hash_sdbm(PERL_HASH_SEED,(U8*)(str),(len)) +# define PERL_HASH_WITH_SEED(seed,hash,str,len) (hash)= S_perl_hash_sdbm((seed),(U8*)(str),(len)) #elif defined(PERL_HASH_FUNC_ONE_AT_A_TIME_HARD) # define PERL_HASH_FUNC "ONE_AT_A_TIME_HARD" # define PERL_HASH_SEED_BYTES 8 -# define PERL_HASH(hash,str,len) (hash)= S_perl_hash_one_at_a_time_hard(PERL_HASH_SEED,(U8*)(str),(len)) +# define PERL_HASH_WITH_SEED(seed,hash,str,len) (hash)= S_perl_hash_one_at_a_time_hard((seed),(U8*)(str),(len)) #elif defined(PERL_HASH_FUNC_ONE_AT_A_TIME) # define PERL_HASH_FUNC "ONE_AT_A_TIME" # define PERL_HASH_SEED_BYTES 4 -# define PERL_HASH(hash,str,len) (hash)= S_perl_hash_one_at_a_time(PERL_HASH_SEED,(U8*)(str),(len)) +# define PERL_HASH_WITH_SEED(seed,hash,str,len) (hash)= S_perl_hash_one_at_a_time((seed),(U8*)(str),(len)) #elif defined(PERL_HASH_FUNC_ONE_AT_A_TIME_OLD) # define PERL_HASH_FUNC "ONE_AT_A_TIME_OLD" # define PERL_HASH_SEED_BYTES 4 -# define PERL_HASH(hash,str,len) (hash)= S_perl_hash_old_one_at_a_time(PERL_HASH_SEED,(U8*)(str),(len)) +# define PERL_HASH_WITH_SEED(seed,hash,str,len) (hash)= S_perl_hash_old_one_at_a_time((seed),(U8*)(str),(len)) +#elif defined(PERL_HASH_FUNC_MURMUR_HASH_64A) +# define PERL_HASH_FUNC "MURMUR_HASH_64A" +# define PERL_HASH_SEED_BYTES 8 +# define PERL_HASH_WITH_SEED(seed,hash,str,len) (hash)= S_perl_hash_murmur_hash_64a((seed),(U8*)(str),(len)) +#elif defined(PERL_HASH_FUNC_MURMUR_HASH_64B) +# define PERL_HASH_FUNC "MURMUR_HASH_64B" +# define PERL_HASH_SEED_BYTES 8 +# define PERL_HASH_WITH_SEED(seed,hash,str,len) (hash)= S_perl_hash_murmur_hash_64b((seed),(U8*)(str),(len)) #endif -#ifndef PERL_HASH +#ifndef PERL_HASH_WITH_SEED #error "No hash function defined!" #endif #ifndef PERL_HASH_SEED_BYTES @@ -81,6 +91,8 @@ # endif #endif +#define PERL_HASH(hash,str,len) PERL_HASH_WITH_SEED(PERL_HASH_SEED,hash,str,len) + /*----------------------------------------------------------------------------- * Endianess, misalignment capabilities and util macros * @@ -257,7 +269,7 @@ S_perl_hash_siphash_2_4(const unsigned char * const seed, const unsigned char *i PERL_STATIC_INLINE U32 S_perl_hash_superfast(const unsigned char * const seed, const unsigned char *str, STRLEN len) { - U32 hash = *((U32*)seed) + len; + U32 hash = *((U32*)seed) + (U32)len; U32 tmp; int rem= len & 3; len >>= 2; @@ -367,7 +379,7 @@ S_perl_hash_murmur3(const unsigned char * const seed, const unsigned char *ptr, const unsigned char *end; int bytes_in_carry = 0; /* bytes in carry */ - I32 total_length= len; + I32 total_length= (I32)len; #if defined(UNALIGNED_SAFE) /* Handle carry: commented out as its only used in incremental mode - it never fires for us @@ -390,7 +402,7 @@ S_perl_hash_murmur3(const unsigned char * const seed, const unsigned char *ptr, /* Consume enough so that the next data byte is word aligned */ STRLEN i = -PTR2IV(ptr) & 3; if(i && i <= len) { - MURMUR_DOBYTES(i, h1, carry, bytes_in_carry, ptr, len); + MURMUR_DOBYTES((int)i, h1, carry, bytes_in_carry, ptr, len); } /* We're now aligned. Process in aligned blocks. Specialise for each possible carry count */ @@ -431,7 +443,7 @@ S_perl_hash_murmur3(const unsigned char * const seed, const unsigned char *ptr, len -= len/4*4; /* Append any remaining bytes into carry */ - MURMUR_DOBYTES(len, h1, carry, bytes_in_carry, ptr, len); + MURMUR_DOBYTES((int)len, h1, carry, bytes_in_carry, ptr, len); if (bytes_in_carry) { k1 = carry >> ( 4 - bytes_in_carry ) * 8; @@ -455,7 +467,7 @@ S_perl_hash_murmur3(const unsigned char * const seed, const unsigned char *ptr, PERL_STATIC_INLINE U32 S_perl_hash_djb2(const unsigned char * const seed, const unsigned char *str, const STRLEN len) { const unsigned char * const end = (const unsigned char *)str + len; - U32 hash = *((U32*)seed) + len; + U32 hash = *((U32*)seed) + (U32)len; while (str < end) { hash = ((hash << 5) + hash) + *str++; } @@ -465,7 +477,7 @@ S_perl_hash_djb2(const unsigned char * const seed, const unsigned char *str, con PERL_STATIC_INLINE U32 S_perl_hash_sdbm(const unsigned char * const seed, const unsigned char *str, const STRLEN len) { const unsigned char * const end = (const unsigned char *)str + len; - U32 hash = *((U32*)seed) + len; + U32 hash = *((U32*)seed) + (U32)len; while (str < end) { hash = (hash << 6) + (hash << 16) - hash + *str++; } @@ -491,7 +503,7 @@ S_perl_hash_sdbm(const unsigned char * const seed, const unsigned char *str, con PERL_STATIC_INLINE U32 S_perl_hash_one_at_a_time(const unsigned char * const seed, const unsigned char *str, const STRLEN len) { const unsigned char * const end = (const unsigned char *)str + len; - U32 hash = *((U32*)seed) + len; + U32 hash = *((U32*)seed) + (U32)len; while (str < end) { hash += *str++; hash += (hash << 10); @@ -506,7 +518,7 @@ S_perl_hash_one_at_a_time(const unsigned char * const seed, const unsigned char PERL_STATIC_INLINE U32 S_perl_hash_one_at_a_time_hard(const unsigned char * const seed, const unsigned char *str, const STRLEN len) { const unsigned char * const end = (const unsigned char *)str + len; - U32 hash = *((U32*)seed) + len; + U32 hash = *((U32*)seed) + (U32)len; while (str < end) { hash += (hash << 10); @@ -552,6 +564,134 @@ S_perl_hash_old_one_at_a_time(const unsigned char * const seed, const unsigned c return (hash + (hash << 15)); } +#ifdef PERL_HASH_FUNC_MURMUR_HASH_64A +/* This code is from Austin Appleby and is in the public domain. + Altered by Yves Orton to match Perl's hash interface, and to + return a 32 bit hash. + + Note uses unaligned 64 bit loads - will NOT work on machines with + strict alginment requirements. + + Also this code may not be suitable for big-endian machines. +*/ + +/* a 64 bit hash where we only use the low 32 bits */ +PERL_STATIC_INLINE U32 +S_perl_hash_murmur_hash_64a (const unsigned char * const seed, const unsigned char *str, const STRLEN len) +{ + const U64TYPE m = 0xc6a4a7935bd1e995; + const int r = 47; + U64TYPE h = *((U64TYPE*)seed) ^ len; + const U64TYPE * data = (const U64TYPE *)str; + const U64TYPE * end = data + (len/8); + const unsigned char * data2; + + while(data != end) + { + U64TYPE k = *data++; + + k *= m; + k ^= k >> r; + k *= m; + + h ^= k; + h *= m; + } + + data2 = (const unsigned char *)data; + + switch(len & 7) + { + case 7: h ^= (U64TYPE)(data2[6]) << 48; /* fallthrough */ + case 6: h ^= (U64TYPE)(data2[5]) << 40; /* fallthrough */ + case 5: h ^= (U64TYPE)(data2[4]) << 32; /* fallthrough */ + case 4: h ^= (U64TYPE)(data2[3]) << 24; /* fallthrough */ + case 3: h ^= (U64TYPE)(data2[2]) << 16; /* fallthrough */ + case 2: h ^= (U64TYPE)(data2[1]) << 8; /* fallthrough */ + case 1: h ^= (U64TYPE)(data2[0]); /* fallthrough */ + h *= m; + }; + + h ^= h >> r; + h *= m; + h ^= h >> r; + + /* was: return h; */ + return h & 0xFFFFFFFF; +} + +#endif + +#ifdef PERL_HASH_FUNC_MURMUR_HASH_64B +/* This code is from Austin Appleby and is in the public domain. + Altered by Yves Orton to match Perl's hash interface and return + a 32 bit value + + Note uses unaligned 32 bit loads - will NOT work on machines with + strict alginment requirements. + + Also this code may not be suitable for big-endian machines. +*/ + +/* a 64-bit hash for 32-bit platforms where we only use the low 32 bits */ +PERL_STATIC_INLINE U32 +S_perl_hash_murmur_hash_64b (const unsigned char * const seed, const unsigned char *str, STRLEN len) +{ + const U32 m = 0x5bd1e995; + const int r = 24; + + U32 h1 = ((U32 *)seed)[0] ^ len; + U32 h2 = ((U32 *)seed)[1]; + + const U32 * data = (const U32 *)str; + + while(len >= 8) + { + U32 k1, k2; + k1 = *data++; + k1 *= m; k1 ^= k1 >> r; k1 *= m; + h1 *= m; h1 ^= k1; + len -= 4; + + k2 = *data++; + k2 *= m; k2 ^= k2 >> r; k2 *= m; + h2 *= m; h2 ^= k2; + len -= 4; + } + + if(len >= 4) + { + U32 k1 = *data++; + k1 *= m; k1 ^= k1 >> r; k1 *= m; + h1 *= m; h1 ^= k1; + len -= 4; + } + + switch(len) + { + case 3: h2 ^= ((unsigned char*)data)[2] << 16; /* fallthrough */ + case 2: h2 ^= ((unsigned char*)data)[1] << 8; /* fallthrough */ + case 1: h2 ^= ((unsigned char*)data)[0]; /* fallthrough */ + h2 *= m; + }; + + h1 ^= h2 >> 18; h1 *= m; + h2 ^= h1 >> 22; h2 *= m; + /* + The following code has been removed as it is unused + when only the low 32 bits are used. -- Yves + + h1 ^= h2 >> 17; h1 *= m; + + U64TYPE h = h1; + + h = (h << 32) | h2; + */ + + return h2; +} +#endif + /* legacy - only mod_perl should be doing this. */ #ifdef PERL_HASH_INTERNAL_ACCESS #define PERL_HASH_INTERNAL(hash,str,len) PERL_HASH(hash,str,len) @@ -560,11 +700,5 @@ S_perl_hash_old_one_at_a_time(const unsigned char * const seed, const unsigned c #endif /*compile once*/ /* - * Local variables: - * c-indentation-style: bsd - * c-basic-offset: 4 - * indent-tabs-mode: nil - * End: - * * ex: set ts=8 sts=4 sw=4 et: */