--- libecb/ecb.h 2011/06/12 08:14:03 1.42 +++ libecb/ecb.h 2012/05/29 21:06:01 1.95 @@ -1,7 +1,7 @@ /* * libecb - http://software.schmorp.de/pkg/libecb * - * Copyright (©) 2009-2011 Marc Alexander Lehmann + * Copyright (©) 2009-2012 Marc Alexander Lehmann * Copyright (©) 2011 Emanuele Giaquinta * All rights reserved. * @@ -30,33 +30,188 @@ #ifndef ECB_H #define ECB_H -#include +/* 16 bits major, 16 bits minor */ +#define ECB_VERSION 0x00010001 + +#ifdef _WIN32 + typedef signed char int8_t; + typedef unsigned char uint8_t; + typedef signed short int16_t; + typedef unsigned short uint16_t; + typedef signed int int32_t; + typedef unsigned int uint32_t; + #if __GNUC__ + typedef signed long long int64_t; + typedef unsigned long long uint64_t; + #else /* _MSC_VER || __BORLANDC__ */ + typedef signed __int64 int64_t; + typedef unsigned __int64 uint64_t; + #endif + #ifdef _WIN64 + #define ECB_PTRSIZE 8 + typedef uint64_t uintptr_t; + typedef int64_t intptr_t; + #else + #define ECB_PTRSIZE 4 + typedef uint32_t uintptr_t; + typedef int32_t intptr_t; + #endif + typedef intptr_t ptrdiff_t; +#else + #include + #if UINTMAX_MAX > 0xffffffffU + #define ECB_PTRSIZE 8 + #else + #define ECB_PTRSIZE 4 + #endif +#endif /* many compilers define _GNUC_ to some versions but then only implement * what their idiot authors think are the "more important" extensions, - * causing enourmous grief in return for some better fake benchmark numbers. + * causing enormous grief in return for some better fake benchmark numbers. * or so. * we try to detect these and simply assume they are not gcc - if they have * an issue with that they should have done it right in the first place. */ #ifndef ECB_GCC_VERSION - #if defined(__INTEL_COMPILER) || defined(__SUNPRO_C) || defined(__llvm__) + #if !defined __GNUC_MINOR__ || defined __INTEL_COMPILER || defined __SUNPRO_C || defined __SUNPRO_CC || defined __llvm__ || defined __clang__ #define ECB_GCC_VERSION(major,minor) 0 #else #define ECB_GCC_VERSION(major,minor) (__GNUC__ > (major) || (__GNUC__ == (major) && __GNUC_MINOR__ >= (minor))) #endif #endif -#define ECB_C99 (__STDC_VERSION__ >= 199901L) +#define ECB_C (__STDC__+0) /* this assumes that __STDC__ is either empty or a number */ +#define ECB_C99 (__STDC_VERSION__ >= 199901L) +#define ECB_C11 (__STDC_VERSION__ >= 201112L) +#define ECB_CPP (__cplusplus+0) +#define ECB_CPP98 (__cplusplus >= 199711L) +#define ECB_CPP11 (__cplusplus >= 201103L) + +/*****************************************************************************/ + +/* ECB_NO_THREADS - ecb is not used by multiple threads, ever */ +/* ECB_NO_SMP - ecb might be used in multiple threads, but only on a single cpu */ + +#if ECB_NO_THREADS + #define ECB_NO_SMP 1 +#endif + +#if ECB_NO_SMP + #define ECB_MEMORY_FENCE do { } while (0) +#endif + +#ifndef ECB_MEMORY_FENCE + #if ECB_GCC_VERSION(2,5) || defined __INTEL_COMPILER || (__llvm__ && __GNUC__) || __SUNPRO_C >= 0x5110 || __SUNPRO_CC >= 0x5110 + #if __i386 || __i386__ + #define ECB_MEMORY_FENCE __asm__ __volatile__ ("lock; orb $0, -1(%%esp)" : : : "memory") + #define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("" : : : "memory") + #define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("") + #elif __amd64 || __amd64__ || __x86_64 || __x86_64__ + #define ECB_MEMORY_FENCE __asm__ __volatile__ ("mfence" : : : "memory") + #define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("" : : : "memory") + #define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("") + #elif __powerpc__ || __ppc__ || __powerpc64__ || __ppc64__ + #define ECB_MEMORY_FENCE __asm__ __volatile__ ("sync" : : : "memory") + #elif defined __ARM_ARCH_6__ || defined __ARM_ARCH_6J__ \ + || defined __ARM_ARCH_6K__ || defined __ARM_ARCH_6ZK__ + #define ECB_MEMORY_FENCE __asm__ __volatile__ ("mcr p15,0,%0,c7,c10,5" : : "r" (0) : "memory") + #elif defined __ARM_ARCH_7__ || defined __ARM_ARCH_7A__ \ + || defined __ARM_ARCH_7M__ || defined __ARM_ARCH_7R__ + #define ECB_MEMORY_FENCE __asm__ __volatile__ ("dmb" : : : "memory") + #elif __sparc || __sparc__ + #define ECB_MEMORY_FENCE __asm__ __volatile__ ("membar #LoadStore | #LoadLoad | #StoreStore | #StoreLoad" : : : "memory") + #define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("membar #LoadStore | #LoadLoad" : : : "memory") + #define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("membar #LoadStore | #StoreStore") + #elif defined __s390__ || defined __s390x__ + #define ECB_MEMORY_FENCE __asm__ __volatile__ ("bcr 15,0" : : : "memory") + #elif defined __mips__ + #define ECB_MEMORY_FENCE __asm__ __volatile__ ("sync" : : : "memory") + #elif defined __alpha__ + #define ECB_MEMORY_FENCE __asm__ __volatile__ ("mb" : : : "memory") + #elif defined __hppa__ + #define ECB_MEMORY_FENCE __asm__ __volatile__ ("" : : : "memory") + #define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("") + #elif defined __ia64__ + #define ECB_MEMORY_FENCE __asm__ __volatile__ ("mf" : : : "memory") + #endif + #endif +#endif + +#ifndef ECB_MEMORY_FENCE + #if ECB_GCC_VERSION(4,7) + /* see comment below about the C11 memory model. in short - avoid */ + #define ECB_MEMORY_FENCE __atomic_thread_fence (__ATOMIC_SEQ_CST) + #elif defined __clang && __has_feature (cxx_atomic) + /* see above */ + #define ECB_MEMORY_FENCE __c11_atomic_thread_fence (__ATOMIC_SEQ_CST) + #elif ECB_GCC_VERSION(4,4) || defined __INTEL_COMPILER || defined __clang__ + #define ECB_MEMORY_FENCE __sync_synchronize () + #elif _MSC_VER >= 1400 /* VC++ 2005 */ + #pragma intrinsic(_ReadBarrier,_WriteBarrier,_ReadWriteBarrier) + #define ECB_MEMORY_FENCE _ReadWriteBarrier () + #define ECB_MEMORY_FENCE_ACQUIRE _ReadWriteBarrier () /* according to msdn, _ReadBarrier is not a load fence */ + #define ECB_MEMORY_FENCE_RELEASE _WriteBarrier () + #elif defined _WIN32 + #include + #define ECB_MEMORY_FENCE MemoryBarrier () /* actually just xchg on x86... scary */ + #elif __SUNPRO_C >= 0x5110 || __SUNPRO_CC >= 0x5110 + #include + #define ECB_MEMORY_FENCE __machine_rw_barrier () + #define ECB_MEMORY_FENCE_ACQUIRE __machine_r_barrier () + #define ECB_MEMORY_FENCE_RELEASE __machine_w_barrier () + #elif __xlC__ + #define ECB_MEMORY_FENCE __sync () + #endif +#endif + +#ifndef ECB_MEMORY_FENCE + #if ECB_C11 && !defined __STDC_NO_ATOMICS__ + /* we assume that these memory fences work on all variables/all memory accesses, */ + /* not just C11 atomics and atomic accesses */ + #include + /* unfortunately, the C11 memory model seems to be very limited, and unable to express */ + /* simple barrier semantics. That means we need to take out thor's hammer. */ + #define ECB_MEMORY_FENCE atomic_thread_fence (memory_order_seq_cst) + #endif +#endif + +#ifndef ECB_MEMORY_FENCE + #if !ECB_AVOID_PTHREADS + /* + * if you get undefined symbol references to pthread_mutex_lock, + * or failure to find pthread.h, then you should implement + * the ECB_MEMORY_FENCE operations for your cpu/compiler + * OR provide pthread.h and link against the posix thread library + * of your system. + */ + #include + #define ECB_NEEDS_PTHREADS 1 + #define ECB_MEMORY_FENCE_NEEDS_PTHREADS 1 + + static pthread_mutex_t ecb_mf_lock = PTHREAD_MUTEX_INITIALIZER; + #define ECB_MEMORY_FENCE do { pthread_mutex_lock (&ecb_mf_lock); pthread_mutex_unlock (&ecb_mf_lock); } while (0) + #endif +#endif + +#if !defined ECB_MEMORY_FENCE_ACQUIRE && defined ECB_MEMORY_FENCE + #define ECB_MEMORY_FENCE_ACQUIRE ECB_MEMORY_FENCE +#endif + +#if !defined ECB_MEMORY_FENCE_RELEASE && defined ECB_MEMORY_FENCE + #define ECB_MEMORY_FENCE_RELEASE ECB_MEMORY_FENCE +#endif + +/*****************************************************************************/ #if __cplusplus - #define ECB_INLINE static inline + #define ecb_inline static inline #elif ECB_GCC_VERSION(2,5) - #define ECB_INLINE static __inline__ + #define ecb_inline static __inline__ #elif ECB_C99 - #define ECB_INLINE static inline + #define ecb_inline static inline #else - #define ECB_INLINE static + #define ecb_inline static #endif #if ECB_GCC_VERSION(3,3) @@ -74,7 +229,7 @@ #define ECB_STRINGIFY_(a) # a #define ECB_STRINGIFY(a) ECB_STRINGIFY_(a) -#define ecb_function_ ECB_INLINE +#define ecb_function_ ecb_inline #if ECB_GCC_VERSION(3,1) #define ecb_attribute(attrlist) __attribute__(attrlist) @@ -96,11 +251,16 @@ #endif #define ecb_noinline ecb_attribute ((__noinline__)) -#define ecb_noreturn ecb_attribute ((__noreturn__)) #define ecb_unused ecb_attribute ((__unused__)) #define ecb_const ecb_attribute ((__const__)) #define ecb_pure ecb_attribute ((__pure__)) +#if ECB_C11 + #define ecb_noreturn _Noreturn +#else + #define ecb_noreturn ecb_attribute ((__noreturn__)) +#endif + #if ECB_GCC_VERSION(4,3) #define ecb_artificial ecb_attribute ((__artificial__)) #define ecb_hot ecb_attribute ((__hot__)) @@ -122,8 +282,13 @@ /* count trailing zero bits and count # of one bits */ #if ECB_GCC_VERSION(3,4) + /* we assume int == 32 bit, long == 32 or 64 bit and long long == 64 bit */ + #define ecb_ld32(x) (__builtin_clz (x) ^ 31) + #define ecb_ld64(x) (__builtin_clzll (x) ^ 63) #define ecb_ctz32(x) __builtin_ctz (x) + #define ecb_ctz64(x) __builtin_ctzll (x) #define ecb_popcount32(x) __builtin_popcount (x) + /* no popcountll */ #else ecb_function_ int ecb_ctz32 (uint32_t x) ecb_const; ecb_function_ int @@ -131,17 +296,33 @@ { int r = 0; - x &= -x; /* this isolates the lowest bit */ + x &= ~x + 1; /* this isolates the lowest bit */ +#if ECB_branchless_on_i386 + r += !!(x & 0xaaaaaaaa) << 0; + r += !!(x & 0xcccccccc) << 1; + r += !!(x & 0xf0f0f0f0) << 2; + r += !!(x & 0xff00ff00) << 3; + r += !!(x & 0xffff0000) << 4; +#else if (x & 0xaaaaaaaa) r += 1; if (x & 0xcccccccc) r += 2; if (x & 0xf0f0f0f0) r += 4; if (x & 0xff00ff00) r += 8; if (x & 0xffff0000) r += 16; +#endif return r; } + ecb_function_ int ecb_ctz64 (uint64_t x) ecb_const; + ecb_function_ int + ecb_ctz64 (uint64_t x) + { + int shift = x & 0xffffffffU ? 0 : 32; + return ecb_ctz32 (x >> shift) + shift; + } + ecb_function_ int ecb_popcount32 (uint32_t x) ecb_const; ecb_function_ int ecb_popcount32 (uint32_t x) @@ -153,29 +334,118 @@ return x >> 24; } + + ecb_function_ int ecb_ld32 (uint32_t x) ecb_const; + ecb_function_ int ecb_ld32 (uint32_t x) + { + int r = 0; + + if (x >> 16) { x >>= 16; r += 16; } + if (x >> 8) { x >>= 8; r += 8; } + if (x >> 4) { x >>= 4; r += 4; } + if (x >> 2) { x >>= 2; r += 2; } + if (x >> 1) { r += 1; } + + return r; + } + + ecb_function_ int ecb_ld64 (uint64_t x) ecb_const; + ecb_function_ int ecb_ld64 (uint64_t x) + { + int r = 0; + + if (x >> 32) { x >>= 32; r += 32; } + + return r + ecb_ld32 (x); + } #endif +ecb_function_ ecb_bool ecb_is_pot32 (uint32_t x) ecb_const; +ecb_function_ ecb_bool ecb_is_pot32 (uint32_t x) { return !(x & (x - 1)); } +ecb_function_ ecb_bool ecb_is_pot64 (uint64_t x) ecb_const; +ecb_function_ ecb_bool ecb_is_pot64 (uint64_t x) { return !(x & (x - 1)); } + +ecb_function_ uint8_t ecb_bitrev8 (uint8_t x) ecb_const; +ecb_function_ uint8_t ecb_bitrev8 (uint8_t x) +{ + return ( (x * 0x0802U & 0x22110U) + | (x * 0x8020U & 0x88440U)) * 0x10101U >> 16; +} + +ecb_function_ uint16_t ecb_bitrev16 (uint16_t x) ecb_const; +ecb_function_ uint16_t ecb_bitrev16 (uint16_t x) +{ + x = ((x >> 1) & 0x5555) | ((x & 0x5555) << 1); + x = ((x >> 2) & 0x3333) | ((x & 0x3333) << 2); + x = ((x >> 4) & 0x0f0f) | ((x & 0x0f0f) << 4); + x = ( x >> 8 ) | ( x << 8); + + return x; +} + +ecb_function_ uint32_t ecb_bitrev32 (uint32_t x) ecb_const; +ecb_function_ uint32_t ecb_bitrev32 (uint32_t x) +{ + x = ((x >> 1) & 0x55555555) | ((x & 0x55555555) << 1); + x = ((x >> 2) & 0x33333333) | ((x & 0x33333333) << 2); + x = ((x >> 4) & 0x0f0f0f0f) | ((x & 0x0f0f0f0f) << 4); + x = ((x >> 8) & 0x00ff00ff) | ((x & 0x00ff00ff) << 8); + x = ( x >> 16 ) | ( x << 16); + + return x; +} + +/* popcount64 is only available on 64 bit cpus as gcc builtin */ +/* so for this version we are lazy */ +ecb_function_ int ecb_popcount64 (uint64_t x) ecb_const; +ecb_function_ int +ecb_popcount64 (uint64_t x) +{ + return ecb_popcount32 (x) + ecb_popcount32 (x >> 32); +} + +ecb_inline uint8_t ecb_rotl8 (uint8_t x, unsigned int count) ecb_const; +ecb_inline uint8_t ecb_rotr8 (uint8_t x, unsigned int count) ecb_const; +ecb_inline uint16_t ecb_rotl16 (uint16_t x, unsigned int count) ecb_const; +ecb_inline uint16_t ecb_rotr16 (uint16_t x, unsigned int count) ecb_const; +ecb_inline uint32_t ecb_rotl32 (uint32_t x, unsigned int count) ecb_const; +ecb_inline uint32_t ecb_rotr32 (uint32_t x, unsigned int count) ecb_const; +ecb_inline uint64_t ecb_rotl64 (uint64_t x, unsigned int count) ecb_const; +ecb_inline uint64_t ecb_rotr64 (uint64_t x, unsigned int count) ecb_const; + +ecb_inline uint8_t ecb_rotl8 (uint8_t x, unsigned int count) { return (x >> ( 8 - count)) | (x << count); } +ecb_inline uint8_t ecb_rotr8 (uint8_t x, unsigned int count) { return (x << ( 8 - count)) | (x >> count); } +ecb_inline uint16_t ecb_rotl16 (uint16_t x, unsigned int count) { return (x >> (16 - count)) | (x << count); } +ecb_inline uint16_t ecb_rotr16 (uint16_t x, unsigned int count) { return (x << (16 - count)) | (x >> count); } +ecb_inline uint32_t ecb_rotl32 (uint32_t x, unsigned int count) { return (x >> (32 - count)) | (x << count); } +ecb_inline uint32_t ecb_rotr32 (uint32_t x, unsigned int count) { return (x << (32 - count)) | (x >> count); } +ecb_inline uint64_t ecb_rotl64 (uint64_t x, unsigned int count) { return (x >> (64 - count)) | (x << count); } +ecb_inline uint64_t ecb_rotr64 (uint64_t x, unsigned int count) { return (x << (64 - count)) | (x >> count); } + #if ECB_GCC_VERSION(4,3) - #define ecb_bswap32(x) __builtin_bswap32 (x) - #define ecb_bswap16(x) (__builtin_bswap32(x) >> 16) + #define ecb_bswap16(x) (__builtin_bswap32 (x) >> 16) + #define ecb_bswap32(x) __builtin_bswap32 (x) + #define ecb_bswap64(x) __builtin_bswap64 (x) #else + ecb_function_ uint16_t ecb_bswap16 (uint16_t x) ecb_const; + ecb_function_ uint16_t + ecb_bswap16 (uint16_t x) + { + return ecb_rotl16 (x, 8); + } + ecb_function_ uint32_t ecb_bswap32 (uint32_t x) ecb_const; ecb_function_ uint32_t ecb_bswap32 (uint32_t x) { - return (x >> 24) - | ((x >> 8) & 0x0000ff00) - | ((x << 8) & 0x00ff0000) - | (x << 24); + return (((uint32_t)ecb_bswap16 (x)) << 16) | ecb_bswap16 (x >> 16); } - ecb_function_ uint32_t ecb_bswap16 (uint32_t x) ecb_const; - ecb_function_ uint32_t - ecb_bswap16 (uint32_t x) + ecb_function_ uint64_t ecb_bswap64 (uint64_t x) ecb_const; + ecb_function_ uint64_t + ecb_bswap64 (uint64_t x) { - return ((x >> 8) & 0xff) - | ((x << 8) & 0x00ff0000) - | (x << 24); + return (((uint64_t)ecb_bswap32 (x)) << 32) | ecb_bswap32 (x >> 32); } #endif @@ -183,25 +453,25 @@ #define ecb_unreachable() __builtin_unreachable () #else /* this seems to work fine, but gcc always emits a warning for it :/ */ - ecb_function_ void ecb_unreachable (void) ecb_noreturn; - ecb_function_ void ecb_unreachable (void) { } + ecb_inline void ecb_unreachable (void) ecb_noreturn; + ecb_inline void ecb_unreachable (void) { } #endif /* try to tell the compiler that some condition is definitely true */ #define ecb_assume(cond) do { if (!(cond)) ecb_unreachable (); } while (0) -ecb_function_ unsigned char ecb_byteorder_helper (void) ecb_const; -ecb_function_ unsigned char +ecb_inline unsigned char ecb_byteorder_helper (void) ecb_const; +ecb_inline unsigned char ecb_byteorder_helper (void) { const uint32_t u = 0x11223344; return *(unsigned char *)&u; } -ecb_function_ ecb_bool ecb_big_endian (void) ecb_const; -ecb_function_ ecb_bool ecb_big_endian (void) { return ecb_byteorder_helper () == 0x11; } -ecb_function_ ecb_bool ecb_little_endian (void) ecb_const; -ecb_function_ ecb_bool ecb_little_endian (void) { return ecb_byteorder_helper () == 0x44; } +ecb_inline ecb_bool ecb_big_endian (void) ecb_const; +ecb_inline ecb_bool ecb_big_endian (void) { return ecb_byteorder_helper () == 0x11; } +ecb_inline ecb_bool ecb_little_endian (void) ecb_const; +ecb_inline ecb_bool ecb_little_endian (void) { return ecb_byteorder_helper () == 0x44; } #if ECB_GCC_VERSION(3,0) || ECB_C99 #define ecb_mod(m,n) ((m) % (n) + ((m) % (n) < 0 ? (n) : 0)) @@ -209,6 +479,22 @@ #define ecb_mod(m,n) ((m) < 0 ? ((n) - 1 - ((-1 - (m)) % (n))) : ((m) % (n))) #endif +#if __cplusplus + template + static inline T ecb_div_rd (T val, T div) + { + return val < 0 ? - ((-val + div - 1) / div) : (val ) / div; + } + template + static inline T ecb_div_ru (T val, T div) + { + return val < 0 ? - ((-val ) / div) : (val + div - 1) / div; + } +#else + #define ecb_div_rd(val,div) ((val) < 0 ? - ((-(val) + (div) - 1) / (div)) : ((val) ) / (div)) + #define ecb_div_ru(val,div) ((val) < 0 ? - ((-(val) ) / (div)) : ((val) + (div) - 1) / (div)) +#endif + #if ecb_cplusplus_does_not_suck /* does not work for local types (http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2657.htm) */ template @@ -220,19 +506,5 @@ #define ecb_array_length(name) (sizeof (name) / sizeof (name [0])) #endif -ECB_INLINE uint32_t ecb_rotr32 (uint32_t x, unsigned int count) ecb_const; -ECB_INLINE uint32_t -ecb_rotr32 (uint32_t x, unsigned int count) -{ - return (x << (32 - count)) | (x >> count); -} - -ECB_INLINE uint32_t ecb_rotl32 (uint32_t x, unsigned int count) ecb_const; -ECB_INLINE uint32_t -ecb_rotl32 (uint32_t x, unsigned int count) -{ - return (x >> (32 - count)) | (x << count); -} - #endif