diff options
Diffstat (limited to 'dviware/dvisvgm/libs/xxHash/xxhash.c')
-rw-r--r-- | dviware/dvisvgm/libs/xxHash/xxhash.c | 688 |
1 files changed, 394 insertions, 294 deletions
diff --git a/dviware/dvisvgm/libs/xxHash/xxhash.c b/dviware/dvisvgm/libs/xxHash/xxhash.c index eb881ff3b1..3f49b7d1d8 100644 --- a/dviware/dvisvgm/libs/xxHash/xxhash.c +++ b/dviware/dvisvgm/libs/xxHash/xxhash.c @@ -33,6 +33,12 @@ */ +/* since xxhash.c can be included (via XXH_INLINE_ALL), + * it's good practice to protect it with guard + * in case of multiples inclusions */ +#ifndef XXHASH_C_01393879 +#define XXHASH_C_01393879 + /* ************************************* * Tuning parameters ***************************************/ @@ -50,14 +56,10 @@ * Prefer these methods in priority order (0 > 1 > 2) */ #ifndef XXH_FORCE_MEMORY_ACCESS /* can be defined externally, on command line for example */ -# if defined(__GNUC__) && ( defined(__ARM_ARCH_6__) || defined(__ARM_ARCH_6J__) \ - || defined(__ARM_ARCH_6K__) || defined(__ARM_ARCH_6Z__) \ - || defined(__ARM_ARCH_6ZK__) || defined(__ARM_ARCH_6T2__) ) +# if !defined(__clang__) && defined(__GNUC__) && defined(__ARM_FEATURE_UNALIGNED) && defined(__ARM_ARCH) && (__ARM_ARCH == 6) # define XXH_FORCE_MEMORY_ACCESS 2 -# elif (defined(__INTEL_COMPILER) && !defined(_WIN32)) || \ - (defined(__GNUC__) && ( defined(__ARM_ARCH_7__) || defined(__ARM_ARCH_7A__) \ - || defined(__ARM_ARCH_7R__) || defined(__ARM_ARCH_7M__) \ - || defined(__ARM_ARCH_7S__) )) +# elif !defined(__clang__) && ((defined(__INTEL_COMPILER) && !defined(_WIN32)) || \ + (defined(__GNUC__) && (defined(__ARM_ARCH) && __ARM_ARCH >= 7))) # define XXH_FORCE_MEMORY_ACCESS 1 # endif #endif @@ -86,6 +88,18 @@ # endif #endif +/*!XXH_REROLL: + * Whether to reroll XXH32_finalize, and XXH64_finalize, + * instead of using an unrolled jump table/if statement loop. + * + * This is automatically defined on -Os/-Oz on GCC and Clang. */ +#ifndef XXH_REROLL +# if defined(__OPTIMIZE_SIZE__) +# define XXH_REROLL 1 +# else +# define XXH_REROLL 0 +# endif +#endif /* ************************************* * Includes & Memory related functions @@ -99,7 +113,7 @@ static void XXH_free (void* p) { free(p); } #include <string.h> static void* XXH_memcpy(void* dest, const void* src, size_t size) { return memcpy(dest,src,size); } -#include <assert.h> /* assert */ +#include <limits.h> /* ULLONG_MAX */ #define XXH_STATIC_LINKING_ONLY #include "xxhash.h" @@ -111,36 +125,57 @@ static void* XXH_memcpy(void* dest, const void* src, size_t size) { return memcp #ifdef _MSC_VER /* Visual Studio */ # pragma warning(disable : 4127) /* disable: C4127: conditional expression is constant */ # define XXH_FORCE_INLINE static __forceinline +# define XXH_NO_INLINE static __declspec(noinline) #else # if defined (__cplusplus) || defined (__STDC_VERSION__) && __STDC_VERSION__ >= 199901L /* C99 */ # ifdef __GNUC__ # define XXH_FORCE_INLINE static inline __attribute__((always_inline)) +# define XXH_NO_INLINE static __attribute__((noinline)) # else # define XXH_FORCE_INLINE static inline +# define XXH_NO_INLINE static # endif # else # define XXH_FORCE_INLINE static +# define XXH_NO_INLINE static # endif /* __STDC_VERSION__ */ #endif + +/* ************************************* +* Debug +***************************************/ +/* DEBUGLEVEL is expected to be defined externally, + * typically through compiler command line. + * Value must be a number. */ +#ifndef DEBUGLEVEL +# define DEBUGLEVEL 0 +#endif + +#if (DEBUGLEVEL>=1) +# include <assert.h> /* note : can still be disabled with NDEBUG */ +# define XXH_ASSERT(c) assert(c) +#else +# define XXH_ASSERT(c) ((void)0) +#endif + +/* note : use after variable declarations */ +#define XXH_STATIC_ASSERT(c) { enum { XXH_sa = 1/(int)(!!(c)) }; } + + /* ************************************* * Basic Types ***************************************/ -#ifndef MEM_MODULE -# if !defined (__VMS) \ - && (defined (__cplusplus) \ - || (defined (__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L) /* C99 */) ) -# include <stdint.h> - typedef uint8_t BYTE; - typedef uint16_t U16; - typedef uint32_t U32; -# else - typedef unsigned char BYTE; - typedef unsigned short U16; - typedef unsigned int U32; -# endif +#if !defined (__VMS) \ + && (defined (__cplusplus) \ + || (defined (__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L) /* C99 */) ) +# include <stdint.h> + typedef uint8_t xxh_u8; +#else + typedef unsigned char xxh_u8; #endif +typedef XXH32_hash_t xxh_u32; /* === Memory access === */ @@ -148,23 +183,23 @@ static void* XXH_memcpy(void* dest, const void* src, size_t size) { return memcp #if (defined(XXH_FORCE_MEMORY_ACCESS) && (XXH_FORCE_MEMORY_ACCESS==2)) /* Force direct memory access. Only works on CPU which support unaligned memory access in hardware */ -static U32 XXH_read32(const void* memPtr) { return *(const U32*) memPtr; } +static xxh_u32 XXH_read32(const void* memPtr) { return *(const xxh_u32*) memPtr; } #elif (defined(XXH_FORCE_MEMORY_ACCESS) && (XXH_FORCE_MEMORY_ACCESS==1)) /* __pack instructions are safer, but compiler specific, hence potentially problematic for some compilers */ /* currently only defined for gcc and icc */ -typedef union { U32 u32; } __attribute__((packed)) unalign; -static U32 XXH_read32(const void* ptr) { return ((const unalign*)ptr)->u32; } +typedef union { xxh_u32 u32; } __attribute__((packed)) unalign; +static xxh_u32 XXH_read32(const void* ptr) { return ((const unalign*)ptr)->u32; } #else /* portable and safe solution. Generally efficient. * see : http://stackoverflow.com/a/32095106/646947 */ -static U32 XXH_read32(const void* memPtr) +static xxh_u32 XXH_read32(const void* memPtr) { - U32 val; + xxh_u32 val; memcpy(&val, memPtr, sizeof(val)); return val; } @@ -177,12 +212,21 @@ typedef enum { XXH_bigEndian=0, XXH_littleEndian=1 } XXH_endianess; /* XXH_CPU_LITTLE_ENDIAN can be defined externally, for example on the compiler command line */ #ifndef XXH_CPU_LITTLE_ENDIAN +# if defined(_WIN32) /* Windows is always little endian */ \ + || defined(__LITTLE_ENDIAN__) \ + || (defined(__BYTE_ORDER__) && __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__) +# define XXH_CPU_LITTLE_ENDIAN 1 +# elif defined(__BIG_ENDIAN__) \ + || (defined(__BYTE_ORDER__) && __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__) +# define XXH_CPU_LITTLE_ENDIAN 0 +# else static int XXH_isLittleEndian(void) { - const union { U32 u; BYTE c[4]; } one = { 1 }; /* don't use static : performance detrimental */ + const union { xxh_u32 u; xxh_u8 c[4]; } one = { 1 }; /* don't use static : performance detrimental */ return one.c[0]; } # define XXH_CPU_LITTLE_ENDIAN XXH_isLittleEndian() +# endif #endif @@ -193,8 +237,15 @@ static int XXH_isLittleEndian(void) ******************************************/ #define XXH_GCC_VERSION (__GNUC__ * 100 + __GNUC_MINOR__) +#ifndef __has_builtin +# define __has_builtin(x) 0 +#endif + +#if !defined(NO_CLANG_BUILTIN) && __has_builtin(__builtin_rotateleft32) && __has_builtin(__builtin_rotateleft64) +# define XXH_rotl32 __builtin_rotateleft32 +# define XXH_rotl64 __builtin_rotateleft64 /* Note : although _rotl exists for minGW (GCC under windows), performance seems poor */ -#if defined(_MSC_VER) +#elif defined(_MSC_VER) # define XXH_rotl32(x,r) _rotl(x,r) # define XXH_rotl64(x,r) _rotl64(x,r) #else @@ -207,7 +258,7 @@ static int XXH_isLittleEndian(void) #elif XXH_GCC_VERSION >= 403 # define XXH_swap32 __builtin_bswap32 #else -static U32 XXH_swap32 (U32 x) +static xxh_u32 XXH_swap32 (xxh_u32 x) { return ((x << 24) & 0xff000000 ) | ((x << 8) & 0x00ff0000 ) | @@ -222,44 +273,43 @@ static U32 XXH_swap32 (U32 x) *****************************/ typedef enum { XXH_aligned, XXH_unaligned } XXH_alignment; -XXH_FORCE_INLINE U32 XXH_readLE32(const void* ptr) +XXH_FORCE_INLINE xxh_u32 XXH_readLE32(const void* ptr) { return XXH_CPU_LITTLE_ENDIAN ? XXH_read32(ptr) : XXH_swap32(XXH_read32(ptr)); } -static U32 XXH_readBE32(const void* ptr) +static xxh_u32 XXH_readBE32(const void* ptr) { return XXH_CPU_LITTLE_ENDIAN ? XXH_swap32(XXH_read32(ptr)) : XXH_read32(ptr); } -XXH_FORCE_INLINE U32 +XXH_FORCE_INLINE xxh_u32 XXH_readLE32_align(const void* ptr, XXH_alignment align) { if (align==XXH_unaligned) { return XXH_readLE32(ptr); } else { - return XXH_CPU_LITTLE_ENDIAN ? *(const U32*)ptr : XXH_swap32(*(const U32*)ptr); + return XXH_CPU_LITTLE_ENDIAN ? *(const xxh_u32*)ptr : XXH_swap32(*(const xxh_u32*)ptr); } } /* ************************************* -* Macros +* Misc ***************************************/ -#define XXH_STATIC_ASSERT(c) { enum { XXH_sa = 1/(int)(!!(c)) }; } /* use after variable declarations */ XXH_PUBLIC_API unsigned XXH_versionNumber (void) { return XXH_VERSION_NUMBER; } /* ******************************************************************* * 32-bit hash functions *********************************************************************/ -static const U32 PRIME32_1 = 2654435761U; /* 0b10011110001101110111100110110001 */ -static const U32 PRIME32_2 = 2246822519U; /* 0b10000101111010111100101001110111 */ -static const U32 PRIME32_3 = 3266489917U; /* 0b11000010101100101010111000111101 */ -static const U32 PRIME32_4 = 668265263U; /* 0b00100111110101001110101100101111 */ -static const U32 PRIME32_5 = 374761393U; /* 0b00010110010101100110011110110001 */ +static const xxh_u32 PRIME32_1 = 0x9E3779B1U; /* 0b10011110001101110111100110110001 */ +static const xxh_u32 PRIME32_2 = 0x85EBCA77U; /* 0b10000101111010111100101001110111 */ +static const xxh_u32 PRIME32_3 = 0xC2B2AE3DU; /* 0b11000010101100101010111000111101 */ +static const xxh_u32 PRIME32_4 = 0x27D4EB2FU; /* 0b00100111110101001110101100101111 */ +static const xxh_u32 PRIME32_5 = 0x165667B1U; /* 0b00010110010101100110011110110001 */ -static U32 XXH32_round(U32 acc, U32 input) +static xxh_u32 XXH32_round(xxh_u32 acc, xxh_u32 input) { acc += input * PRIME32_2; acc = XXH_rotl32(acc, 13); @@ -312,7 +362,7 @@ static U32 XXH32_round(U32 acc, U32 input) } /* mix all bits */ -static U32 XXH32_avalanche(U32 h32) +static xxh_u32 XXH32_avalanche(xxh_u32 h32) { h32 ^= h32 >> 15; h32 *= PRIME32_2; @@ -324,92 +374,101 @@ static U32 XXH32_avalanche(U32 h32) #define XXH_get32bits(p) XXH_readLE32_align(p, align) -static U32 -XXH32_finalize(U32 h32, const void* ptr, size_t len, XXH_alignment align) - +static xxh_u32 +XXH32_finalize(xxh_u32 h32, const xxh_u8* ptr, size_t len, XXH_alignment align) { - const BYTE* p = (const BYTE*)ptr; - #define PROCESS1 \ - h32 += (*p++) * PRIME32_5; \ + h32 += (*ptr++) * PRIME32_5; \ h32 = XXH_rotl32(h32, 11) * PRIME32_1 ; #define PROCESS4 \ - h32 += XXH_get32bits(p) * PRIME32_3; \ - p+=4; \ + h32 += XXH_get32bits(ptr) * PRIME32_3; \ + ptr+=4; \ h32 = XXH_rotl32(h32, 17) * PRIME32_4 ; - switch(len&15) /* or switch(bEnd - p) */ - { - case 12: PROCESS4; - /* fallthrough */ - case 8: PROCESS4; - /* fallthrough */ - case 4: PROCESS4; - return XXH32_avalanche(h32); - - case 13: PROCESS4; - /* fallthrough */ - case 9: PROCESS4; - /* fallthrough */ - case 5: PROCESS4; - PROCESS1; - return XXH32_avalanche(h32); - - case 14: PROCESS4; - /* fallthrough */ - case 10: PROCESS4; - /* fallthrough */ - case 6: PROCESS4; - PROCESS1; - PROCESS1; - return XXH32_avalanche(h32); - - case 15: PROCESS4; - /* fallthrough */ - case 11: PROCESS4; - /* fallthrough */ - case 7: PROCESS4; - /* fallthrough */ - case 3: PROCESS1; - /* fallthrough */ - case 2: PROCESS1; - /* fallthrough */ - case 1: PROCESS1; - /* fallthrough */ - case 0: return XXH32_avalanche(h32); + /* Compact rerolled version */ + if (XXH_REROLL) { + len &= 15; + while (len >= 4) { + PROCESS4; + len -= 4; + } + while (len > 0) { + PROCESS1; + --len; + } + return XXH32_avalanche(h32); + } else { + switch(len&15) /* or switch(bEnd - p) */ { + case 12: PROCESS4; + /* fallthrough */ + case 8: PROCESS4; + /* fallthrough */ + case 4: PROCESS4; + return XXH32_avalanche(h32); + + case 13: PROCESS4; + /* fallthrough */ + case 9: PROCESS4; + /* fallthrough */ + case 5: PROCESS4; + PROCESS1; + return XXH32_avalanche(h32); + + case 14: PROCESS4; + /* fallthrough */ + case 10: PROCESS4; + /* fallthrough */ + case 6: PROCESS4; + PROCESS1; + PROCESS1; + return XXH32_avalanche(h32); + + case 15: PROCESS4; + /* fallthrough */ + case 11: PROCESS4; + /* fallthrough */ + case 7: PROCESS4; + /* fallthrough */ + case 3: PROCESS1; + /* fallthrough */ + case 2: PROCESS1; + /* fallthrough */ + case 1: PROCESS1; + /* fallthrough */ + case 0: return XXH32_avalanche(h32); + } + XXH_ASSERT(0); + return h32; /* reaching this point is deemed impossible */ } - assert(0); - return h32; /* reaching this point is deemed impossible */ } -XXH_FORCE_INLINE U32 -XXH32_endian_align(const void* input, size_t len, U32 seed, XXH_alignment align) +XXH_FORCE_INLINE xxh_u32 +XXH32_endian_align(const xxh_u8* input, size_t len, xxh_u32 seed, XXH_alignment align) { - const BYTE* p = (const BYTE*)input; - const BYTE* bEnd = p + len; - U32 h32; + const xxh_u8* bEnd = input + len; + xxh_u32 h32; #if defined(XXH_ACCEPT_NULL_INPUT_POINTER) && (XXH_ACCEPT_NULL_INPUT_POINTER>=1) - if (p==NULL) { + if (input==NULL) { len=0; - bEnd=p=(const BYTE*)(size_t)16; + bEnd=input=(const xxh_u8*)(size_t)16; } #endif if (len>=16) { - const BYTE* const limit = bEnd - 15; - U32 v1 = seed + PRIME32_1 + PRIME32_2; - U32 v2 = seed + PRIME32_2; - U32 v3 = seed + 0; - U32 v4 = seed - PRIME32_1; + const xxh_u8* const limit = bEnd - 15; + xxh_u32 v1 = seed + PRIME32_1 + PRIME32_2; + xxh_u32 v2 = seed + PRIME32_2; + xxh_u32 v3 = seed + 0; + xxh_u32 v4 = seed - PRIME32_1; do { - v1 = XXH32_round(v1, XXH_get32bits(p)); p+=4; - v2 = XXH32_round(v2, XXH_get32bits(p)); p+=4; - v3 = XXH32_round(v3, XXH_get32bits(p)); p+=4; - v4 = XXH32_round(v4, XXH_get32bits(p)); p+=4; - } while (p < limit); + v1 = XXH32_round(v1, XXH_get32bits(input)); input += 4; + v2 = XXH32_round(v2, XXH_get32bits(input)); input += 4; + v3 = XXH32_round(v3, XXH_get32bits(input)); input += 4; + v4 = XXH32_round(v4, XXH_get32bits(input)); input += 4; + } while (input < limit); h32 = XXH_rotl32(v1, 1) + XXH_rotl32(v2, 7) + XXH_rotl32(v3, 12) + XXH_rotl32(v4, 18); @@ -417,29 +476,29 @@ XXH32_endian_align(const void* input, size_t len, U32 seed, XXH_alignment align) h32 = seed + PRIME32_5; } - h32 += (U32)len; + h32 += (xxh_u32)len; - return XXH32_finalize(h32, p, len&15, align); + return XXH32_finalize(h32, input, len&15, align); } -XXH_PUBLIC_API unsigned int XXH32 (const void* input, size_t len, unsigned int seed) +XXH_PUBLIC_API XXH32_hash_t XXH32 (const void* input, size_t len, XXH32_hash_t seed) { #if 0 /* Simple version, good for code maintenance, but unfortunately slow for small inputs */ XXH32_state_t state; XXH32_reset(&state, seed); - XXH32_update(&state, input, len); + XXH32_update(&state, (const xxh_u8*)input, len); return XXH32_digest(&state); #else if (XXH_FORCE_ALIGN_CHECK) { if ((((size_t)input) & 3) == 0) { /* Input is 4-bytes aligned, leverage the speed benefit */ - return XXH32_endian_align(input, len, seed, XXH_aligned); + return XXH32_endian_align((const xxh_u8*)input, len, seed, XXH_aligned); } } - return XXH32_endian_align(input, len, seed, XXH_unaligned); + return XXH32_endian_align((const xxh_u8*)input, len, seed, XXH_unaligned); #endif } @@ -462,7 +521,7 @@ XXH_PUBLIC_API void XXH32_copyState(XXH32_state_t* dstState, const XXH32_state_t memcpy(dstState, srcState, sizeof(*dstState)); } -XXH_PUBLIC_API XXH_errorcode XXH32_reset(XXH32_state_t* statePtr, unsigned int seed) +XXH_PUBLIC_API XXH_errorcode XXH32_reset(XXH32_state_t* statePtr, XXH32_hash_t seed) { XXH32_state_t state; /* using a local state to memcpy() in order to avoid strict-aliasing warnings */ memset(&state, 0, sizeof(state)); @@ -486,21 +545,21 @@ XXH32_update(XXH32_state_t* state, const void* input, size_t len) return XXH_ERROR; #endif - { const BYTE* p = (const BYTE*)input; - const BYTE* const bEnd = p + len; + { const xxh_u8* p = (const xxh_u8*)input; + const xxh_u8* const bEnd = p + len; state->total_len_32 += (XXH32_hash_t)len; state->large_len |= (XXH32_hash_t)((len>=16) | (state->total_len_32>=16)); if (state->memsize + len < 16) { /* fill in tmp buffer */ - XXH_memcpy((BYTE*)(state->mem32) + state->memsize, input, len); + XXH_memcpy((xxh_u8*)(state->mem32) + state->memsize, input, len); state->memsize += (XXH32_hash_t)len; return XXH_OK; } if (state->memsize) { /* some data left from previous update */ - XXH_memcpy((BYTE*)(state->mem32) + state->memsize, input, 16-state->memsize); - { const U32* p32 = state->mem32; + XXH_memcpy((xxh_u8*)(state->mem32) + state->memsize, input, 16-state->memsize); + { const xxh_u32* p32 = state->mem32; state->v1 = XXH32_round(state->v1, XXH_readLE32(p32)); p32++; state->v2 = XXH32_round(state->v2, XXH_readLE32(p32)); p32++; state->v3 = XXH32_round(state->v3, XXH_readLE32(p32)); p32++; @@ -511,11 +570,11 @@ XXH32_update(XXH32_state_t* state, const void* input, size_t len) } if (p <= bEnd-16) { - const BYTE* const limit = bEnd - 16; - U32 v1 = state->v1; - U32 v2 = state->v2; - U32 v3 = state->v3; - U32 v4 = state->v4; + const xxh_u8* const limit = bEnd - 16; + xxh_u32 v1 = state->v1; + xxh_u32 v2 = state->v2; + xxh_u32 v3 = state->v3; + xxh_u32 v4 = state->v4; do { v1 = XXH32_round(v1, XXH_readLE32(p)); p+=4; @@ -540,9 +599,9 @@ XXH32_update(XXH32_state_t* state, const void* input, size_t len) } -XXH_PUBLIC_API unsigned int XXH32_digest (const XXH32_state_t* state) +XXH_PUBLIC_API XXH32_hash_t XXH32_digest (const XXH32_state_t* state) { - U32 h32; + xxh_u32 h32; if (state->large_len) { h32 = XXH_rotl32(state->v1, 1) @@ -555,7 +614,7 @@ XXH_PUBLIC_API unsigned int XXH32_digest (const XXH32_state_t* state) h32 += state->total_len_32; - return XXH32_finalize(h32, state->mem32, state->memsize, XXH_aligned); + return XXH32_finalize(h32, (const xxh_u8*)state->mem32, state->memsize, XXH_aligned); } @@ -588,31 +647,46 @@ XXH_PUBLIC_API XXH32_hash_t XXH32_hashFromCanonical(const XXH32_canonical_t* src /*====== Memory access ======*/ -#ifndef MEM_MODULE -# define MEM_MODULE -# if !defined (__VMS) \ - && (defined (__cplusplus) \ - || (defined (__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L) /* C99 */) ) -# include <stdint.h> - typedef uint64_t U64; -# else - /* if compiler doesn't support unsigned long long, replace by another 64-bit type */ - typedef unsigned long long U64; -# endif -#endif - +typedef XXH64_hash_t xxh_u64; + + +/*! XXH_REROLL_XXH64: + * Whether to reroll the XXH64_finalize() loop. + * + * Just like XXH32, we can unroll the XXH64_finalize() loop. This can be a performance gain + * on 64-bit hosts, as only one jump is required. + * + * However, on 32-bit hosts, because arithmetic needs to be done with two 32-bit registers, + * and 64-bit arithmetic needs to be simulated, it isn't beneficial to unroll. The code becomes + * ridiculously large (the largest function in the binary on i386!), and rerolling it saves + * anywhere from 3kB to 20kB. It is also slightly faster because it fits into cache better + * and is more likely to be inlined by the compiler. + * + * If XXH_REROLL is defined, this is ignored and the loop is always rerolled. */ +#ifndef XXH_REROLL_XXH64 +# if (defined(__ILP32__) || defined(_ILP32)) /* ILP32 is often defined on 32-bit GCC family */ \ + || !(defined(__x86_64__) || defined(_M_X64) || defined(_M_AMD64) /* x86-64 */ \ + || defined(_M_ARM64) || defined(__aarch64__) || defined(__arm64__) /* aarch64 */ \ + || defined(__PPC64__) || defined(__PPC64LE__) || defined(__ppc64__) || defined(__powerpc64__) /* ppc64 */ \ + || defined(__mips64__) || defined(__mips64)) /* mips64 */ \ + || (!defined(SIZE_MAX) || SIZE_MAX < ULLONG_MAX) /* check limits */ +# define XXH_REROLL_XXH64 1 +# else +# define XXH_REROLL_XXH64 0 +# endif +#endif /* !defined(XXH_REROLL_XXH64) */ #if (defined(XXH_FORCE_MEMORY_ACCESS) && (XXH_FORCE_MEMORY_ACCESS==2)) /* Force direct memory access. Only works on CPU which support unaligned memory access in hardware */ -static U64 XXH_read64(const void* memPtr) { return *(const U64*) memPtr; } +static xxh_u64 XXH_read64(const void* memPtr) { return *(const xxh_u64*) memPtr; } #elif (defined(XXH_FORCE_MEMORY_ACCESS) && (XXH_FORCE_MEMORY_ACCESS==1)) /* __pack instructions are safer, but compiler specific, hence potentially problematic for some compilers */ /* currently only defined for gcc and icc */ -typedef union { U32 u32; U64 u64; } __attribute__((packed)) unalign64; -static U64 XXH_read64(const void* ptr) { return ((const unalign64*)ptr)->u64; } +typedef union { xxh_u32 u32; xxh_u64 u64; } __attribute__((packed)) unalign64; +static xxh_u64 XXH_read64(const void* ptr) { return ((const unalign64*)ptr)->u64; } #else @@ -620,9 +694,9 @@ static U64 XXH_read64(const void* ptr) { return ((const unalign64*)ptr)->u64; } * see : http://stackoverflow.com/a/32095106/646947 */ -static U64 XXH_read64(const void* memPtr) +static xxh_u64 XXH_read64(const void* memPtr) { - U64 val; + xxh_u64 val; memcpy(&val, memPtr, sizeof(val)); return val; } @@ -634,7 +708,7 @@ static U64 XXH_read64(const void* memPtr) #elif XXH_GCC_VERSION >= 403 # define XXH_swap64 __builtin_bswap64 #else -static U64 XXH_swap64 (U64 x) +static xxh_u64 XXH_swap64 (xxh_u64 x) { return ((x << 56) & 0xff00000000000000ULL) | ((x << 40) & 0x00ff000000000000ULL) | @@ -647,35 +721,35 @@ static U64 XXH_swap64 (U64 x) } #endif -XXH_FORCE_INLINE U64 XXH_readLE64(const void* ptr) +XXH_FORCE_INLINE xxh_u64 XXH_readLE64(const void* ptr) { return XXH_CPU_LITTLE_ENDIAN ? XXH_read64(ptr) : XXH_swap64(XXH_read64(ptr)); } -static U64 XXH_readBE64(const void* ptr) +static xxh_u64 XXH_readBE64(const void* ptr) { return XXH_CPU_LITTLE_ENDIAN ? XXH_swap64(XXH_read64(ptr)) : XXH_read64(ptr); } -XXH_FORCE_INLINE U64 +XXH_FORCE_INLINE xxh_u64 XXH_readLE64_align(const void* ptr, XXH_alignment align) { if (align==XXH_unaligned) return XXH_readLE64(ptr); else - return XXH_CPU_LITTLE_ENDIAN ? *(const U64*)ptr : XXH_swap64(*(const U64*)ptr); + return XXH_CPU_LITTLE_ENDIAN ? *(const xxh_u64*)ptr : XXH_swap64(*(const xxh_u64*)ptr); } /*====== xxh64 ======*/ -static const U64 PRIME64_1 = 11400714785074694791ULL; /* 0b1001111000110111011110011011000110000101111010111100101010000111 */ -static const U64 PRIME64_2 = 14029467366897019727ULL; /* 0b1100001010110010101011100011110100100111110101001110101101001111 */ -static const U64 PRIME64_3 = 1609587929392839161ULL; /* 0b0001011001010110011001111011000110011110001101110111100111111001 */ -static const U64 PRIME64_4 = 9650029242287828579ULL; /* 0b1000010111101011110010100111011111000010101100101010111001100011 */ -static const U64 PRIME64_5 = 2870177450012600261ULL; /* 0b0010011111010100111010110010111100010110010101100110011111000101 */ +static const xxh_u64 PRIME64_1 = 0x9E3779B185EBCA87ULL; /* 0b1001111000110111011110011011000110000101111010111100101010000111 */ +static const xxh_u64 PRIME64_2 = 0xC2B2AE3D27D4EB4FULL; /* 0b1100001010110010101011100011110100100111110101001110101101001111 */ +static const xxh_u64 PRIME64_3 = 0x165667B19E3779F9ULL; /* 0b0001011001010110011001111011000110011110001101110111100111111001 */ +static const xxh_u64 PRIME64_4 = 0x85EBCA77C2B2AE63ULL; /* 0b1000010111101011110010100111011111000010101100101010111001100011 */ +static const xxh_u64 PRIME64_5 = 0x27D4EB2F165667C5ULL; /* 0b0010011111010100111010110010111100010110010101100110011111000101 */ -static U64 XXH64_round(U64 acc, U64 input) +static xxh_u64 XXH64_round(xxh_u64 acc, xxh_u64 input) { acc += input * PRIME64_2; acc = XXH_rotl64(acc, 31); @@ -683,7 +757,7 @@ static U64 XXH64_round(U64 acc, U64 input) return acc; } -static U64 XXH64_mergeRound(U64 acc, U64 val) +static xxh_u64 XXH64_mergeRound(xxh_u64 acc, xxh_u64 val) { val = XXH64_round(0, val); acc ^= val; @@ -691,7 +765,7 @@ static U64 XXH64_mergeRound(U64 acc, U64 val) return acc; } -static U64 XXH64_avalanche(U64 h64) +static xxh_u64 XXH64_avalanche(xxh_u64 h64) { h64 ^= h64 >> 33; h64 *= PRIME64_2; @@ -704,141 +778,155 @@ static U64 XXH64_avalanche(U64 h64) #define XXH_get64bits(p) XXH_readLE64_align(p, align) -static U64 -XXH64_finalize(U64 h64, const void* ptr, size_t len, XXH_alignment align) +static xxh_u64 +XXH64_finalize(xxh_u64 h64, const xxh_u8* ptr, size_t len, XXH_alignment align) { - const BYTE* p = (const BYTE*)ptr; - #define PROCESS1_64 \ - h64 ^= (*p++) * PRIME64_5; \ + h64 ^= (*ptr++) * PRIME64_5; \ h64 = XXH_rotl64(h64, 11) * PRIME64_1; #define PROCESS4_64 \ - h64 ^= (U64)(XXH_get32bits(p)) * PRIME64_1; \ - p+=4; \ + h64 ^= (xxh_u64)(XXH_get32bits(ptr)) * PRIME64_1; \ + ptr+=4; \ h64 = XXH_rotl64(h64, 23) * PRIME64_2 + PRIME64_3; #define PROCESS8_64 { \ - U64 const k1 = XXH64_round(0, XXH_get64bits(p)); \ - p+=8; \ + xxh_u64 const k1 = XXH64_round(0, XXH_get64bits(ptr)); \ + ptr+=8; \ h64 ^= k1; \ h64 = XXH_rotl64(h64,27) * PRIME64_1 + PRIME64_4; \ } - switch(len&31) { - case 24: PROCESS8_64; - /* fallthrough */ - case 16: PROCESS8_64; - /* fallthrough */ - case 8: PROCESS8_64; - return XXH64_avalanche(h64); - - case 28: PROCESS8_64; - /* fallthrough */ - case 20: PROCESS8_64; - /* fallthrough */ - case 12: PROCESS8_64; - /* fallthrough */ - case 4: PROCESS4_64; - return XXH64_avalanche(h64); - - case 25: PROCESS8_64; - /* fallthrough */ - case 17: PROCESS8_64; - /* fallthrough */ - case 9: PROCESS8_64; - PROCESS1_64; - return XXH64_avalanche(h64); - - case 29: PROCESS8_64; - /* fallthrough */ - case 21: PROCESS8_64; - /* fallthrough */ - case 13: PROCESS8_64; - /* fallthrough */ - case 5: PROCESS4_64; - PROCESS1_64; - return XXH64_avalanche(h64); - - case 26: PROCESS8_64; - /* fallthrough */ - case 18: PROCESS8_64; - /* fallthrough */ - case 10: PROCESS8_64; - PROCESS1_64; - PROCESS1_64; - return XXH64_avalanche(h64); - - case 30: PROCESS8_64; - /* fallthrough */ - case 22: PROCESS8_64; - /* fallthrough */ - case 14: PROCESS8_64; - /* fallthrough */ - case 6: PROCESS4_64; - PROCESS1_64; - PROCESS1_64; - return XXH64_avalanche(h64); - - case 27: PROCESS8_64; - /* fallthrough */ - case 19: PROCESS8_64; - /* fallthrough */ - case 11: PROCESS8_64; - PROCESS1_64; - PROCESS1_64; - PROCESS1_64; - return XXH64_avalanche(h64); - - case 31: PROCESS8_64; - /* fallthrough */ - case 23: PROCESS8_64; - /* fallthrough */ - case 15: PROCESS8_64; - /* fallthrough */ - case 7: PROCESS4_64; - /* fallthrough */ - case 3: PROCESS1_64; - /* fallthrough */ - case 2: PROCESS1_64; - /* fallthrough */ - case 1: PROCESS1_64; - /* fallthrough */ - case 0: return XXH64_avalanche(h64); + /* Rerolled version for 32-bit targets is faster and much smaller. */ + if (XXH_REROLL || XXH_REROLL_XXH64) { + len &= 31; + while (len >= 8) { + PROCESS8_64; + len -= 8; + } + if (len >= 4) { + PROCESS4_64; + len -= 4; + } + while (len > 0) { + PROCESS1_64; + --len; + } + return XXH64_avalanche(h64); + } else { + switch(len & 31) { + case 24: PROCESS8_64; + /* fallthrough */ + case 16: PROCESS8_64; + /* fallthrough */ + case 8: PROCESS8_64; + return XXH64_avalanche(h64); + + case 28: PROCESS8_64; + /* fallthrough */ + case 20: PROCESS8_64; + /* fallthrough */ + case 12: PROCESS8_64; + /* fallthrough */ + case 4: PROCESS4_64; + return XXH64_avalanche(h64); + + case 25: PROCESS8_64; + /* fallthrough */ + case 17: PROCESS8_64; + /* fallthrough */ + case 9: PROCESS8_64; + PROCESS1_64; + return XXH64_avalanche(h64); + + case 29: PROCESS8_64; + /* fallthrough */ + case 21: PROCESS8_64; + /* fallthrough */ + case 13: PROCESS8_64; + /* fallthrough */ + case 5: PROCESS4_64; + PROCESS1_64; + return XXH64_avalanche(h64); + + case 26: PROCESS8_64; + /* fallthrough */ + case 18: PROCESS8_64; + /* fallthrough */ + case 10: PROCESS8_64; + PROCESS1_64; + PROCESS1_64; + return XXH64_avalanche(h64); + + case 30: PROCESS8_64; + /* fallthrough */ + case 22: PROCESS8_64; + /* fallthrough */ + case 14: PROCESS8_64; + /* fallthrough */ + case 6: PROCESS4_64; + PROCESS1_64; + PROCESS1_64; + return XXH64_avalanche(h64); + + case 27: PROCESS8_64; + /* fallthrough */ + case 19: PROCESS8_64; + /* fallthrough */ + case 11: PROCESS8_64; + PROCESS1_64; + PROCESS1_64; + PROCESS1_64; + return XXH64_avalanche(h64); + + case 31: PROCESS8_64; + /* fallthrough */ + case 23: PROCESS8_64; + /* fallthrough */ + case 15: PROCESS8_64; + /* fallthrough */ + case 7: PROCESS4_64; + /* fallthrough */ + case 3: PROCESS1_64; + /* fallthrough */ + case 2: PROCESS1_64; + /* fallthrough */ + case 1: PROCESS1_64; + /* fallthrough */ + case 0: return XXH64_avalanche(h64); + } } - /* impossible to reach */ - assert(0); + XXH_ASSERT(0); return 0; /* unreachable, but some compilers complain without it */ } -XXH_FORCE_INLINE U64 -XXH64_endian_align(const void* input, size_t len, U64 seed, XXH_alignment align) +XXH_FORCE_INLINE xxh_u64 +XXH64_endian_align(const xxh_u8* input, size_t len, xxh_u64 seed, XXH_alignment align) { - const BYTE* p = (const BYTE*)input; - const BYTE* bEnd = p + len; - U64 h64; + const xxh_u8* bEnd = input + len; + xxh_u64 h64; #if defined(XXH_ACCEPT_NULL_INPUT_POINTER) && (XXH_ACCEPT_NULL_INPUT_POINTER>=1) - if (p==NULL) { + if (input==NULL) { len=0; - bEnd=p=(const BYTE*)(size_t)32; + bEnd=input=(const xxh_u8*)(size_t)32; } #endif if (len>=32) { - const BYTE* const limit = bEnd - 32; - U64 v1 = seed + PRIME64_1 + PRIME64_2; - U64 v2 = seed + PRIME64_2; - U64 v3 = seed + 0; - U64 v4 = seed - PRIME64_1; + const xxh_u8* const limit = bEnd - 32; + xxh_u64 v1 = seed + PRIME64_1 + PRIME64_2; + xxh_u64 v2 = seed + PRIME64_2; + xxh_u64 v3 = seed + 0; + xxh_u64 v4 = seed - PRIME64_1; do { - v1 = XXH64_round(v1, XXH_get64bits(p)); p+=8; - v2 = XXH64_round(v2, XXH_get64bits(p)); p+=8; - v3 = XXH64_round(v3, XXH_get64bits(p)); p+=8; - v4 = XXH64_round(v4, XXH_get64bits(p)); p+=8; - } while (p<=limit); + v1 = XXH64_round(v1, XXH_get64bits(input)); input+=8; + v2 = XXH64_round(v2, XXH_get64bits(input)); input+=8; + v3 = XXH64_round(v3, XXH_get64bits(input)); input+=8; + v4 = XXH64_round(v4, XXH_get64bits(input)); input+=8; + } while (input<=limit); h64 = XXH_rotl64(v1, 1) + XXH_rotl64(v2, 7) + XXH_rotl64(v3, 12) + XXH_rotl64(v4, 18); h64 = XXH64_mergeRound(h64, v1); @@ -850,29 +938,29 @@ XXH64_endian_align(const void* input, size_t len, U64 seed, XXH_alignment align) h64 = seed + PRIME64_5; } - h64 += (U64) len; + h64 += (xxh_u64) len; - return XXH64_finalize(h64, p, len, align); + return XXH64_finalize(h64, input, len, align); } -XXH_PUBLIC_API unsigned long long XXH64 (const void* input, size_t len, unsigned long long seed) +XXH_PUBLIC_API XXH64_hash_t XXH64 (const void* input, size_t len, XXH64_hash_t seed) { #if 0 /* Simple version, good for code maintenance, but unfortunately slow for small inputs */ XXH64_state_t state; XXH64_reset(&state, seed); - XXH64_update(&state, input, len); + XXH64_update(&state, (const xxh_u8*)input, len); return XXH64_digest(&state); #else if (XXH_FORCE_ALIGN_CHECK) { if ((((size_t)input) & 7)==0) { /* Input is aligned, let's leverage the speed advantage */ - return XXH64_endian_align(input, len, seed, XXH_aligned); + return XXH64_endian_align((const xxh_u8*)input, len, seed, XXH_aligned); } } - return XXH64_endian_align(input, len, seed, XXH_unaligned); + return XXH64_endian_align((const xxh_u8*)input, len, seed, XXH_unaligned); #endif } @@ -894,7 +982,7 @@ XXH_PUBLIC_API void XXH64_copyState(XXH64_state_t* dstState, const XXH64_state_t memcpy(dstState, srcState, sizeof(*dstState)); } -XXH_PUBLIC_API XXH_errorcode XXH64_reset(XXH64_state_t* statePtr, unsigned long long seed) +XXH_PUBLIC_API XXH_errorcode XXH64_reset(XXH64_state_t* statePtr, XXH64_hash_t seed) { XXH64_state_t state; /* using a local state to memcpy() in order to avoid strict-aliasing warnings */ memset(&state, 0, sizeof(state)); @@ -902,8 +990,8 @@ XXH_PUBLIC_API XXH_errorcode XXH64_reset(XXH64_state_t* statePtr, unsigned long state.v2 = seed + PRIME64_2; state.v3 = seed + 0; state.v4 = seed - PRIME64_1; - /* do not write into reserved, planned to be removed in a future version */ - memcpy(statePtr, &state, sizeof(state) - sizeof(state.reserved)); + /* do not write into reserved64, might be removed in a future version */ + memcpy(statePtr, &state, sizeof(state) - sizeof(state.reserved64)); return XXH_OK; } @@ -917,19 +1005,19 @@ XXH64_update (XXH64_state_t* state, const void* input, size_t len) return XXH_ERROR; #endif - { const BYTE* p = (const BYTE*)input; - const BYTE* const bEnd = p + len; + { const xxh_u8* p = (const xxh_u8*)input; + const xxh_u8* const bEnd = p + len; state->total_len += len; if (state->memsize + len < 32) { /* fill in tmp buffer */ - XXH_memcpy(((BYTE*)state->mem64) + state->memsize, input, len); - state->memsize += (U32)len; + XXH_memcpy(((xxh_u8*)state->mem64) + state->memsize, input, len); + state->memsize += (xxh_u32)len; return XXH_OK; } if (state->memsize) { /* tmp buffer is full */ - XXH_memcpy(((BYTE*)state->mem64) + state->memsize, input, 32-state->memsize); + XXH_memcpy(((xxh_u8*)state->mem64) + state->memsize, input, 32-state->memsize); state->v1 = XXH64_round(state->v1, XXH_readLE64(state->mem64+0)); state->v2 = XXH64_round(state->v2, XXH_readLE64(state->mem64+1)); state->v3 = XXH64_round(state->v3, XXH_readLE64(state->mem64+2)); @@ -939,11 +1027,11 @@ XXH64_update (XXH64_state_t* state, const void* input, size_t len) } if (p+32 <= bEnd) { - const BYTE* const limit = bEnd - 32; - U64 v1 = state->v1; - U64 v2 = state->v2; - U64 v3 = state->v3; - U64 v4 = state->v4; + const xxh_u8* const limit = bEnd - 32; + xxh_u64 v1 = state->v1; + xxh_u64 v2 = state->v2; + xxh_u64 v3 = state->v3; + xxh_u64 v4 = state->v4; do { v1 = XXH64_round(v1, XXH_readLE64(p)); p+=8; @@ -968,15 +1056,15 @@ XXH64_update (XXH64_state_t* state, const void* input, size_t len) } -XXH_PUBLIC_API unsigned long long XXH64_digest (const XXH64_state_t* state) +XXH_PUBLIC_API XXH64_hash_t XXH64_digest (const XXH64_state_t* state) { - U64 h64; + xxh_u64 h64; if (state->total_len >= 32) { - U64 const v1 = state->v1; - U64 const v2 = state->v2; - U64 const v3 = state->v3; - U64 const v4 = state->v4; + xxh_u64 const v1 = state->v1; + xxh_u64 const v2 = state->v2; + xxh_u64 const v3 = state->v3; + xxh_u64 const v4 = state->v4; h64 = XXH_rotl64(v1, 1) + XXH_rotl64(v2, 7) + XXH_rotl64(v3, 12) + XXH_rotl64(v4, 18); h64 = XXH64_mergeRound(h64, v1); @@ -987,9 +1075,9 @@ XXH_PUBLIC_API unsigned long long XXH64_digest (const XXH64_state_t* state) h64 = state->v3 /*seed*/ + PRIME64_5; } - h64 += (U64) state->total_len; + h64 += (xxh_u64) state->total_len; - return XXH64_finalize(h64, state->mem64, (size_t)state->total_len, XXH_aligned); + return XXH64_finalize(h64, (const xxh_u8*)state->mem64, (size_t)state->total_len, XXH_aligned); } @@ -1007,4 +1095,16 @@ XXH_PUBLIC_API XXH64_hash_t XXH64_hashFromCanonical(const XXH64_canonical_t* src return XXH_readBE64(src); } + + +/* ********************************************************************* +* XXH3 +* New generation hash designed for speed on small keys and vectorization +************************************************************************ */ + +#include "xxh3.h" + + #endif /* XXH_NO_LONG_LONG */ + +#endif /* XXHASH_C_01393879 */ |