--- libecb/ecb.h 2012/01/18 12:13:00 1.72 +++ libecb/ecb.h 2012/05/29 21:06:01 1.95 @@ -30,6 +30,9 @@ #ifndef ECB_H #define ECB_H +/* 16 bits major, 16 bits minor */ +#define ECB_VERSION 0x00010001 + #ifdef _WIN32 typedef signed char int8_t; typedef unsigned char uint8_t; @@ -44,8 +47,23 @@ 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 @@ -56,59 +74,85 @@ * an issue with that they should have done it right in the first place. */ #ifndef ECB_GCC_VERSION - #if !defined(__GNUC_MINOR__) || defined(__INTEL_COMPILER) || defined(__SUNPRO_C) || defined(__SUNPRO_CC) || defined(__llvm__) || defined(__clang__) + #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_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 || ECB_NO_SMP +#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) || defined(__clang__) || __SUNPRO_C >= 0x5110 || __SUNPRO_CC >= 0x5110 - #if __i386__ + #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 ECB_MEMORY_FENCE /* non-lock xchg might be enough */ - #define ECB_MEMORY_FENCE_RELEASE do { } while (0) /* unlikely to change in future cpus */ - #elif __amd64 - #define ECB_MEMORY_FENCE __asm__ __volatile__ ("mfence" : : : "memory") - #define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("lfence" : : : "memory") - #define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("sfence") /* play safe - not needed in any current cpu */ + #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 defined(__sparc) - #define ECB_MEMORY_FENCE __asm__ __volatile__ ("membar #LoadStore | #StoreLoad | #LoadLoad | #StoreStore" : : : "memory") - #define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("membar #LoadLoad" : : : "memory") - #define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("membar #StoreStore") + #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(4,4) || defined(__INTEL_COMPILER) || defined(__clang__) + #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 () - /*#define ECB_MEMORY_FENCE_ACQUIRE ({ char dummy = 0; __sync_lock_test_and_set (&dummy, 1); }) */ - /*#define ECB_MEMORY_FENCE_RELEASE ({ char dummy = 1; __sync_lock_release (&dummy ); }) */ #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) + #elif defined _WIN32 #include #define ECB_MEMORY_FENCE MemoryBarrier () /* actually just xchg on x86... scary */ #elif __SUNPRO_C >= 0x5110 || __SUNPRO_CC >= 0x5110 @@ -116,6 +160,19 @@ #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 @@ -137,18 +194,16 @@ #endif #endif -#if !defined(ECB_MEMORY_FENCE_ACQUIRE) && defined(ECB_MEMORY_FENCE) +#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) +#if !defined ECB_MEMORY_FENCE_RELEASE && defined ECB_MEMORY_FENCE #define ECB_MEMORY_FENCE_RELEASE ECB_MEMORY_FENCE #endif /*****************************************************************************/ -#define ECB_C99 (__STDC_VERSION__ >= 199901L) - #if __cplusplus #define ecb_inline static inline #elif ECB_GCC_VERSION(2,5) @@ -196,11 +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__)) @@ -300,6 +360,11 @@ } #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) { @@ -388,25 +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))