… | |
… | |
67 | |
67 | |
68 | /* ECB_NO_THREADS - ecb is not used by multiple threads, ever */ |
68 | /* ECB_NO_THREADS - ecb is not used by multiple threads, ever */ |
69 | /* ECB_NO_SMP - ecb might be used in multiple threads, but only on a single cpu */ |
69 | /* ECB_NO_SMP - ecb might be used in multiple threads, but only on a single cpu */ |
70 | |
70 | |
71 | #if ECB_NO_THREADS || ECB_NO_SMP |
71 | #if ECB_NO_THREADS || ECB_NO_SMP |
72 | #define ECB_MEMORY_FENCE do { } while (0) |
72 | #define ECB_MEMORY_FENCE do { } while (0) |
73 | #define ECB_MEMORY_FENCE_ACQUIRE do { } while (0) |
|
|
74 | #define ECB_MEMORY_FENCE_RELEASE do { } while (0) |
|
|
75 | #endif |
73 | #endif |
76 | |
74 | |
77 | #ifndef ECB_MEMORY_FENCE |
75 | #ifndef ECB_MEMORY_FENCE |
78 | #if ECB_GCC_VERSION(2,5) |
76 | #if ECB_GCC_VERSION(2,5) || defined(__INTEL_COMPILER) || defined(__clang__) |
79 | #if __x86 |
77 | #if __i386__ |
80 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("lock; orb $0, -1(%%esp)" : : : "memory") |
78 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("lock; orb $0, -1(%%esp)" : : : "memory") |
81 | #define ECB_MEMORY_FENCE_ACQUIRE ECB_MEMORY_FENCE /* non-lock xchg might be enough */ |
79 | #define ECB_MEMORY_FENCE_ACQUIRE ECB_MEMORY_FENCE /* non-lock xchg might be enough */ |
82 | #define ECB_MEMORY_FENCE_RELEASE do { } while (0) /* unlikely to change in future cpus */ |
80 | #define ECB_MEMORY_FENCE_RELEASE do { } while (0) /* unlikely to change in future cpus */ |
83 | #elif __amd64 |
81 | #elif __amd64 |
84 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("mfence" : : : "memory") |
82 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("mfence" : : : "memory") |
85 | #define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("lfence" : : : "memory") |
83 | #define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("lfence" : : : "memory") |
86 | #define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("sfence") /* play safe - not needed in any current cpu */ |
84 | #define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("sfence") /* play safe - not needed in any current cpu */ |
|
|
85 | #elif __powerpc__ || __ppc__ || __powerpc64__ || __ppc64__ |
|
|
86 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("sync" : : : "memory") |
87 | #elif defined(__ARM_ARCH_6__ ) || defined(__ARM_ARCH_6J__ ) \ |
87 | #elif defined(__ARM_ARCH_6__ ) || defined(__ARM_ARCH_6J__ ) \ |
88 | || defined(__ARM_ARCH_6K__) || defined(__ARM_ARCH_6ZK__) \ |
88 | || defined(__ARM_ARCH_6K__) || defined(__ARM_ARCH_6ZK__) |
|
|
89 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("mcr p15,0,%0,c7,c10,5" : : "r" (0) : "memory") |
89 | || defined(__ARM_ARCH_7__ ) || defined(__ARM_ARCH_7A__ ) \ |
90 | #elif defined(__ARM_ARCH_7__ ) || defined(__ARM_ARCH_7A__ ) \ |
90 | || defined(__ARM_ARCH_7M__) || defined(__ARM_ARCH_7R__ ) |
91 | || defined(__ARM_ARCH_7M__) || defined(__ARM_ARCH_7R__ ) |
91 | #define ECB_MEMORY_FENCE \ |
92 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("dmb" : : : "memory") |
92 | do { \ |
|
|
93 | int null = 0; \ |
|
|
94 | __asm__ __volatile__ ("mcr p15,0,%0,c6,c10,5", : "=&r" (null) : : "memory"); \ |
|
|
95 | while (0) |
|
|
96 | #endif |
93 | #endif |
97 | #endif |
94 | #endif |
98 | #endif |
95 | #endif |
99 | |
96 | |
100 | #ifndef ECB_MEMORY_FENCE |
97 | #ifndef ECB_MEMORY_FENCE |
101 | #if ECB_GCC_VERSION(4,4) || defined(__INTEL_COMPILER) |
98 | #if ECB_GCC_VERSION(4,4) || defined(__INTEL_COMPILER) || defined(__clang__) |
102 | #define ECB_MEMORY_FENCE __sync_synchronize () |
99 | #define ECB_MEMORY_FENCE __sync_synchronize () |
103 | /*#define ECB_MEMORY_FENCE_ACQUIRE ({ char dummy = 0; __sync_lock_test_and_set (&dummy, 1); }) */ |
100 | /*#define ECB_MEMORY_FENCE_ACQUIRE ({ char dummy = 0; __sync_lock_test_and_set (&dummy, 1); }) */ |
104 | /*#define ECB_MEMORY_FENCE_RELEASE ({ char dummy = 1; __sync_lock_release (&dummy ); }) */ |
101 | /*#define ECB_MEMORY_FENCE_RELEASE ({ char dummy = 1; __sync_lock_release (&dummy ); }) */ |
105 | #elif _MSC_VER >= 1400 /* VC++ 2005 */ |
102 | #elif _MSC_VER >= 1400 /* VC++ 2005 */ |
106 | #pragma intrinsic(_ReadBarrier,_WriteBarrier,_ReadWriteBarrier) |
103 | #pragma intrinsic(_ReadBarrier,_WriteBarrier,_ReadWriteBarrier) |
… | |
… | |
292 | |
289 | |
293 | return r + ecb_ld32 (x); |
290 | return r + ecb_ld32 (x); |
294 | } |
291 | } |
295 | #endif |
292 | #endif |
296 | |
293 | |
|
|
294 | ecb_function_ uint8_t ecb_bitrev8 (uint8_t x) ecb_const; |
|
|
295 | ecb_function_ uint8_t ecb_bitrev8 (uint8_t x) |
|
|
296 | { |
|
|
297 | return ( (x * 0x0802U & 0x22110U) |
|
|
298 | | (x * 0x8020U & 0x88440U)) * 0x10101U >> 16; |
|
|
299 | } |
|
|
300 | |
|
|
301 | ecb_function_ uint16_t ecb_bitrev16 (uint16_t x) ecb_const; |
|
|
302 | ecb_function_ uint16_t ecb_bitrev16 (uint16_t x) |
|
|
303 | { |
|
|
304 | x = ((x >> 1) & 0x5555) | ((x & 0x5555) << 1); |
|
|
305 | x = ((x >> 2) & 0x3333) | ((x & 0x3333) << 2); |
|
|
306 | x = ((x >> 4) & 0x0f0f) | ((x & 0x0f0f) << 4); |
|
|
307 | x = ( x >> 8 ) | ( x << 8); |
|
|
308 | |
|
|
309 | return x; |
|
|
310 | } |
|
|
311 | |
|
|
312 | ecb_function_ uint32_t ecb_bitrev32 (uint32_t x) ecb_const; |
|
|
313 | ecb_function_ uint32_t ecb_bitrev32 (uint32_t x) |
|
|
314 | { |
|
|
315 | x = ((x >> 1) & 0x55555555) | ((x & 0x55555555) << 1); |
|
|
316 | x = ((x >> 2) & 0x33333333) | ((x & 0x33333333) << 2); |
|
|
317 | x = ((x >> 4) & 0x0f0f0f0f) | ((x & 0x0f0f0f0f) << 4); |
|
|
318 | x = ((x >> 8) & 0x00ff00ff) | ((x & 0x00ff00ff) << 8); |
|
|
319 | x = ( x >> 16 ) | ( x << 16); |
|
|
320 | |
|
|
321 | return x; |
|
|
322 | } |
|
|
323 | |
297 | /* popcount64 is only available on 64 bit cpus as gcc builtin */ |
324 | /* popcount64 is only available on 64 bit cpus as gcc builtin */ |
298 | /* so for this version we are lazy */ |
325 | /* so for this version we are lazy */ |
299 | ecb_function_ int ecb_popcount64 (uint64_t x) ecb_const; |
326 | ecb_function_ int ecb_popcount64 (uint64_t x) ecb_const; |
300 | ecb_function_ int |
327 | ecb_function_ int |
301 | ecb_popcount64 (uint64_t x) |
328 | ecb_popcount64 (uint64_t x) |
… | |
… | |
376 | #define ecb_mod(m,n) ((m) % (n) + ((m) % (n) < 0 ? (n) : 0)) |
403 | #define ecb_mod(m,n) ((m) % (n) + ((m) % (n) < 0 ? (n) : 0)) |
377 | #else |
404 | #else |
378 | #define ecb_mod(m,n) ((m) < 0 ? ((n) - 1 - ((-1 - (m)) % (n))) : ((m) % (n))) |
405 | #define ecb_mod(m,n) ((m) < 0 ? ((n) - 1 - ((-1 - (m)) % (n))) : ((m) % (n))) |
379 | #endif |
406 | #endif |
380 | |
407 | |
|
|
408 | #if __cplusplus |
|
|
409 | template<typename T> |
|
|
410 | static inline T ecb_div_rd (T val, T div) |
|
|
411 | { |
|
|
412 | return val < 0 ? - ((-val + div - 1) / div) : (val ) / div; |
|
|
413 | } |
|
|
414 | template<typename T> |
|
|
415 | static inline T ecb_div_ru (T val, T div) |
|
|
416 | { |
|
|
417 | return val < 0 ? - ((-val ) / div) : (val + div - 1) / div; |
|
|
418 | } |
|
|
419 | #else |
|
|
420 | #define ecb_div_rd(val,div) ((val) < 0 ? - ((-(val) + (div) - 1) / (div)) : ((val) ) / (div)) |
|
|
421 | #define ecb_div_ru(val,div) ((val) < 0 ? - ((-(val) ) / (div)) : ((val) + (div) - 1) / (div)) |
|
|
422 | #endif |
|
|
423 | |
381 | #if ecb_cplusplus_does_not_suck |
424 | #if ecb_cplusplus_does_not_suck |
382 | /* does not work for local types (http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2657.htm) */ |
425 | /* does not work for local types (http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2657.htm) */ |
383 | template<typename T, int N> |
426 | template<typename T, int N> |
384 | static inline int ecb_array_length (const T (&arr)[N]) |
427 | static inline int ecb_array_length (const T (&arr)[N]) |
385 | { |
428 | { |