… | |
… | |
28 | */ |
28 | */ |
29 | |
29 | |
30 | #ifndef ECB_H |
30 | #ifndef ECB_H |
31 | #define ECB_H |
31 | #define ECB_H |
32 | |
32 | |
|
|
33 | #ifdef _WIN32 |
|
|
34 | typedef signed char int8_t; |
|
|
35 | typedef unsigned char uint8_t; |
|
|
36 | typedef signed short int16_t; |
|
|
37 | typedef unsigned short uint16_t; |
|
|
38 | typedef signed int int32_t; |
|
|
39 | typedef unsigned int uint32_t; |
|
|
40 | #if __GNUC__ |
|
|
41 | typedef signed long long int64_t; |
|
|
42 | typedef unsigned long long uint64_t; |
|
|
43 | #else /* _MSC_VER || __BORLANDC__ */ |
|
|
44 | typedef signed __int64 int64_t; |
|
|
45 | typedef unsigned __int64 uint64_t; |
|
|
46 | #endif |
|
|
47 | #else |
33 | #include <inttypes.h> |
48 | #include <inttypes.h> |
|
|
49 | #endif |
34 | |
50 | |
35 | /* many compilers define _GNUC_ to some versions but then only implement |
51 | /* many compilers define _GNUC_ to some versions but then only implement |
36 | * what their idiot authors think are the "more important" extensions, |
52 | * what their idiot authors think are the "more important" extensions, |
37 | * causing enourmous grief in return for some better fake benchmark numbers. |
53 | * causing enormous grief in return for some better fake benchmark numbers. |
38 | * or so. |
54 | * or so. |
39 | * we try to detect these and simply assume they are not gcc - if they have |
55 | * we try to detect these and simply assume they are not gcc - if they have |
40 | * an issue with that they should have done it right in the first place. |
56 | * an issue with that they should have done it right in the first place. |
41 | */ |
57 | */ |
42 | #ifndef ECB_GCC_VERSION |
58 | #ifndef ECB_GCC_VERSION |
43 | #if defined(__INTEL_COMPILER) || defined(__SUNPRO_C) || defined(__llvm__) |
59 | #if !defined(__GNUC_MINOR__) || defined(__INTEL_COMPILER) || defined(__SUNPRO_C) || defined(__SUNPRO_CC) || defined(__llvm__) || defined(__clang__) |
44 | #define ECB_GCC_VERSION(major,minor) 0 |
60 | #define ECB_GCC_VERSION(major,minor) 0 |
45 | #else |
61 | #else |
46 | #define ECB_GCC_VERSION(major,minor) (__GNUC__ > (major) || (__GNUC__ == (major) && __GNUC_MINOR__ >= (minor))) |
62 | #define ECB_GCC_VERSION(major,minor) (__GNUC__ > (major) || (__GNUC__ == (major) && __GNUC_MINOR__ >= (minor))) |
47 | #endif |
63 | #endif |
48 | #endif |
64 | #endif |
49 | |
65 | |
|
|
66 | /*****************************************************************************/ |
|
|
67 | |
|
|
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 */ |
|
|
70 | |
|
|
71 | #if ECB_NO_THREADS || ECB_NO_SMP |
|
|
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 |
|
|
76 | |
|
|
77 | #ifndef ECB_MEMORY_FENCE |
|
|
78 | #if ECB_GCC_VERSION(2,5) |
|
|
79 | #if __x86 |
|
|
80 | #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 */ |
|
|
82 | #define ECB_MEMORY_FENCE_RELEASE do { } while (0) /* unlikely to change in future cpus */ |
|
|
83 | #elif __amd64 |
|
|
84 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("mfence" : : : "memory") |
|
|
85 | #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 */ |
|
|
87 | #elif defined(__ARM_ARCH_6__ ) || defined(__ARM_ARCH_6J__ ) \ |
|
|
88 | || defined(__ARM_ARCH_6K__) || defined(__ARM_ARCH_6ZK__) \ |
|
|
89 | || defined(__ARM_ARCH_7__ ) || defined(__ARM_ARCH_7A__ ) \ |
|
|
90 | || defined(__ARM_ARCH_7M__) || defined(__ARM_ARCH_7R__ ) |
|
|
91 | #define ECB_MEMORY_FENCE \ |
|
|
92 | do { \ |
|
|
93 | int null = 0; \ |
|
|
94 | __asm__ __volatile__ ("mcr p15,0,%0,c6,c10,5", : "=&r" (null) : : "memory"); \ |
|
|
95 | while (0) |
|
|
96 | #endif |
|
|
97 | #endif |
|
|
98 | #endif |
|
|
99 | |
|
|
100 | #ifndef ECB_MEMORY_FENCE |
|
|
101 | #if ECB_GCC_VERSION(4,4) || defined(__INTEL_COMPILER) |
|
|
102 | #define ECB_MEMORY_FENCE __sync_synchronize () |
|
|
103 | /*#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 ); }) */ |
|
|
105 | #elif _MSC_VER >= 1400 /* VC++ 2005 */ |
|
|
106 | #pragma intrinsic(_ReadBarrier,_WriteBarrier,_ReadWriteBarrier) |
|
|
107 | #define ECB_MEMORY_FENCE _ReadWriteBarrier () |
|
|
108 | #define ECB_MEMORY_FENCE_ACQUIRE _ReadWriteBarrier () /* according to msdn, _ReadBarrier is not a load fence */ |
|
|
109 | #define ECB_MEMORY_FENCE_RELEASE _WriteBarrier () |
|
|
110 | #elif defined(_WIN32) |
|
|
111 | #include <WinNT.h> |
|
|
112 | #define ECB_MEMORY_FENCE MemoryBarrier () /* actually just xchg on x86... scary */ |
|
|
113 | #endif |
|
|
114 | #endif |
|
|
115 | |
|
|
116 | #ifndef ECB_MEMORY_FENCE |
|
|
117 | #if !ECB_AVOID_PTHREADS |
|
|
118 | /* |
|
|
119 | * if you get undefined symbol references to pthread_mutex_lock, |
|
|
120 | * or failure to find pthread.h, then you should implement |
|
|
121 | * the ECB_MEMORY_FENCE operations for your cpu/compiler |
|
|
122 | * OR provide pthread.h and link against the posix thread library |
|
|
123 | * of your system. |
|
|
124 | */ |
|
|
125 | #include <pthread.h> |
|
|
126 | #define ECB_NEEDS_PTHREADS 1 |
|
|
127 | #define ECB_MEMORY_FENCE_NEEDS_PTHREADS 1 |
|
|
128 | |
|
|
129 | static pthread_mutex_t ecb_mf_lock = PTHREAD_MUTEX_INITIALIZER; |
|
|
130 | #define ECB_MEMORY_FENCE do { pthread_mutex_lock (&ecb_mf_lock); pthread_mutex_unlock (&ecb_mf_lock); } while (0) |
|
|
131 | #endif |
|
|
132 | #endif |
|
|
133 | |
|
|
134 | #if !defined(ECB_MEMORY_FENCE_ACQUIRE) && defined(ECB_MEMORY_FENCE) |
|
|
135 | #define ECB_MEMORY_FENCE_ACQUIRE ECB_MEMORY_FENCE |
|
|
136 | #endif |
|
|
137 | |
|
|
138 | #if !defined(ECB_MEMORY_FENCE_RELEASE) && defined(ECB_MEMORY_FENCE) |
|
|
139 | #define ECB_MEMORY_FENCE_RELEASE ECB_MEMORY_FENCE |
|
|
140 | #endif |
|
|
141 | |
|
|
142 | /*****************************************************************************/ |
|
|
143 | |
50 | #define ECB_C99 (__STDC_VERSION__ >= 199901L) |
144 | #define ECB_C99 (__STDC_VERSION__ >= 199901L) |
51 | |
145 | |
52 | #if __cplusplus |
146 | #if __cplusplus |
53 | #define ECB_INLINE static inline |
147 | #define ecb_inline static inline |
54 | #elif ECB_GCC_VERSION(2,5) |
148 | #elif ECB_GCC_VERSION(2,5) |
55 | #define ECB_INLINE static __inline__ |
149 | #define ecb_inline static __inline__ |
56 | #elif ECB_C99 |
150 | #elif ECB_C99 |
57 | #define ECB_INLINE static inline |
151 | #define ecb_inline static inline |
58 | #else |
152 | #else |
59 | #define ECB_INLINE static |
153 | #define ecb_inline static |
60 | #endif |
154 | #endif |
61 | |
155 | |
62 | #if ECB_GCC_VERSION(3,3) |
156 | #if ECB_GCC_VERSION(3,3) |
63 | #define ecb_restrict __restrict__ |
157 | #define ecb_restrict __restrict__ |
64 | #elif ECB_C99 |
158 | #elif ECB_C99 |
… | |
… | |
72 | #define ECB_CONCAT_(a, b) a ## b |
166 | #define ECB_CONCAT_(a, b) a ## b |
73 | #define ECB_CONCAT(a, b) ECB_CONCAT_(a, b) |
167 | #define ECB_CONCAT(a, b) ECB_CONCAT_(a, b) |
74 | #define ECB_STRINGIFY_(a) # a |
168 | #define ECB_STRINGIFY_(a) # a |
75 | #define ECB_STRINGIFY(a) ECB_STRINGIFY_(a) |
169 | #define ECB_STRINGIFY(a) ECB_STRINGIFY_(a) |
76 | |
170 | |
77 | #define ecb_function_ ECB_INLINE |
171 | #define ecb_function_ ecb_inline |
78 | |
172 | |
79 | #if ECB_GCC_VERSION(3,1) |
173 | #if ECB_GCC_VERSION(3,1) |
80 | #define ecb_attribute(attrlist) __attribute__(attrlist) |
174 | #define ecb_attribute(attrlist) __attribute__(attrlist) |
81 | #define ecb_is_constant(expr) __builtin_constant_p (expr) |
175 | #define ecb_is_constant(expr) __builtin_constant_p (expr) |
82 | #define ecb_expect(expr,value) __builtin_expect ((expr),(value)) |
176 | #define ecb_expect(expr,value) __builtin_expect ((expr),(value)) |
… | |
… | |
120 | #define ecb_likely(expr) ecb_expect_true (expr) |
214 | #define ecb_likely(expr) ecb_expect_true (expr) |
121 | #define ecb_unlikely(expr) ecb_expect_false (expr) |
215 | #define ecb_unlikely(expr) ecb_expect_false (expr) |
122 | |
216 | |
123 | /* count trailing zero bits and count # of one bits */ |
217 | /* count trailing zero bits and count # of one bits */ |
124 | #if ECB_GCC_VERSION(3,4) |
218 | #if ECB_GCC_VERSION(3,4) |
|
|
219 | /* we assume int == 32 bit, long == 32 or 64 bit and long long == 64 bit */ |
|
|
220 | #define ecb_ld32(x) (__builtin_clz (x) ^ 31) |
|
|
221 | #define ecb_ld64(x) (__builtin_clzll (x) ^ 63) |
125 | #define ecb_ctz32(x) __builtin_ctz (x) |
222 | #define ecb_ctz32(x) __builtin_ctz (x) |
|
|
223 | #define ecb_ctz64(x) __builtin_ctzll (x) |
126 | #define ecb_popcount32(x) __builtin_popcount (x) |
224 | #define ecb_popcount32(x) __builtin_popcount (x) |
|
|
225 | /* no popcountll */ |
127 | #else |
226 | #else |
128 | ecb_function_ int ecb_ctz32 (uint32_t x) ecb_const; |
227 | ecb_function_ int ecb_ctz32 (uint32_t x) ecb_const; |
129 | ecb_function_ int |
228 | ecb_function_ int |
130 | ecb_ctz32 (uint32_t x) |
229 | ecb_ctz32 (uint32_t x) |
131 | { |
230 | { |
132 | int r = 0; |
231 | int r = 0; |
133 | |
232 | |
134 | x &= -x; /* this isolates the lowest bit */ |
233 | x &= ~x + 1; /* this isolates the lowest bit */ |
135 | |
234 | |
|
|
235 | #if ECB_branchless_on_i386 |
|
|
236 | r += !!(x & 0xaaaaaaaa) << 0; |
|
|
237 | r += !!(x & 0xcccccccc) << 1; |
|
|
238 | r += !!(x & 0xf0f0f0f0) << 2; |
|
|
239 | r += !!(x & 0xff00ff00) << 3; |
|
|
240 | r += !!(x & 0xffff0000) << 4; |
|
|
241 | #else |
136 | if (x & 0xaaaaaaaa) r += 1; |
242 | if (x & 0xaaaaaaaa) r += 1; |
137 | if (x & 0xcccccccc) r += 2; |
243 | if (x & 0xcccccccc) r += 2; |
138 | if (x & 0xf0f0f0f0) r += 4; |
244 | if (x & 0xf0f0f0f0) r += 4; |
139 | if (x & 0xff00ff00) r += 8; |
245 | if (x & 0xff00ff00) r += 8; |
140 | if (x & 0xffff0000) r += 16; |
246 | if (x & 0xffff0000) r += 16; |
|
|
247 | #endif |
141 | |
248 | |
142 | return r; |
249 | return r; |
|
|
250 | } |
|
|
251 | |
|
|
252 | ecb_function_ int ecb_ctz64 (uint64_t x) ecb_const; |
|
|
253 | ecb_function_ int |
|
|
254 | ecb_ctz64 (uint64_t x) |
|
|
255 | { |
|
|
256 | int shift = x & 0xffffffffU ? 0 : 32; |
|
|
257 | return ecb_ctz32 (x >> shift) + shift; |
143 | } |
258 | } |
144 | |
259 | |
145 | ecb_function_ int ecb_popcount32 (uint32_t x) ecb_const; |
260 | ecb_function_ int ecb_popcount32 (uint32_t x) ecb_const; |
146 | ecb_function_ int |
261 | ecb_function_ int |
147 | ecb_popcount32 (uint32_t x) |
262 | ecb_popcount32 (uint32_t x) |
… | |
… | |
151 | x = ((x >> 4) + x) & 0x0f0f0f0f; |
266 | x = ((x >> 4) + x) & 0x0f0f0f0f; |
152 | x *= 0x01010101; |
267 | x *= 0x01010101; |
153 | |
268 | |
154 | return x >> 24; |
269 | return x >> 24; |
155 | } |
270 | } |
|
|
271 | |
|
|
272 | ecb_function_ int ecb_ld32 (uint32_t x) ecb_const; |
|
|
273 | ecb_function_ int ecb_ld32 (uint32_t x) |
|
|
274 | { |
|
|
275 | int r = 0; |
|
|
276 | |
|
|
277 | if (x >> 16) { x >>= 16; r += 16; } |
|
|
278 | if (x >> 8) { x >>= 8; r += 8; } |
|
|
279 | if (x >> 4) { x >>= 4; r += 4; } |
|
|
280 | if (x >> 2) { x >>= 2; r += 2; } |
|
|
281 | if (x >> 1) { r += 1; } |
|
|
282 | |
|
|
283 | return r; |
|
|
284 | } |
|
|
285 | |
|
|
286 | ecb_function_ int ecb_ld64 (uint64_t x) ecb_const; |
|
|
287 | ecb_function_ int ecb_ld64 (uint64_t x) |
|
|
288 | { |
|
|
289 | int r = 0; |
|
|
290 | |
|
|
291 | if (x >> 32) { x >>= 32; r += 32; } |
|
|
292 | |
|
|
293 | return r + ecb_ld32 (x); |
|
|
294 | } |
156 | #endif |
295 | #endif |
|
|
296 | |
|
|
297 | /* popcount64 is only available on 64 bit cpus as gcc builtin */ |
|
|
298 | /* so for this version we are lazy */ |
|
|
299 | ecb_function_ int ecb_popcount64 (uint64_t x) ecb_const; |
|
|
300 | ecb_function_ int |
|
|
301 | ecb_popcount64 (uint64_t x) |
|
|
302 | { |
|
|
303 | return ecb_popcount32 (x) + ecb_popcount32 (x >> 32); |
|
|
304 | } |
|
|
305 | |
|
|
306 | ecb_inline uint8_t ecb_rotl8 (uint8_t x, unsigned int count) ecb_const; |
|
|
307 | ecb_inline uint8_t ecb_rotr8 (uint8_t x, unsigned int count) ecb_const; |
|
|
308 | ecb_inline uint16_t ecb_rotl16 (uint16_t x, unsigned int count) ecb_const; |
|
|
309 | ecb_inline uint16_t ecb_rotr16 (uint16_t x, unsigned int count) ecb_const; |
|
|
310 | ecb_inline uint32_t ecb_rotl32 (uint32_t x, unsigned int count) ecb_const; |
|
|
311 | ecb_inline uint32_t ecb_rotr32 (uint32_t x, unsigned int count) ecb_const; |
|
|
312 | ecb_inline uint64_t ecb_rotl64 (uint64_t x, unsigned int count) ecb_const; |
|
|
313 | ecb_inline uint64_t ecb_rotr64 (uint64_t x, unsigned int count) ecb_const; |
|
|
314 | |
|
|
315 | ecb_inline uint8_t ecb_rotl8 (uint8_t x, unsigned int count) { return (x >> ( 8 - count)) | (x << count); } |
|
|
316 | ecb_inline uint8_t ecb_rotr8 (uint8_t x, unsigned int count) { return (x << ( 8 - count)) | (x >> count); } |
|
|
317 | ecb_inline uint16_t ecb_rotl16 (uint16_t x, unsigned int count) { return (x >> (16 - count)) | (x << count); } |
|
|
318 | ecb_inline uint16_t ecb_rotr16 (uint16_t x, unsigned int count) { return (x << (16 - count)) | (x >> count); } |
|
|
319 | ecb_inline uint32_t ecb_rotl32 (uint32_t x, unsigned int count) { return (x >> (32 - count)) | (x << count); } |
|
|
320 | ecb_inline uint32_t ecb_rotr32 (uint32_t x, unsigned int count) { return (x << (32 - count)) | (x >> count); } |
|
|
321 | ecb_inline uint64_t ecb_rotl64 (uint64_t x, unsigned int count) { return (x >> (64 - count)) | (x << count); } |
|
|
322 | ecb_inline uint64_t ecb_rotr64 (uint64_t x, unsigned int count) { return (x << (64 - count)) | (x >> count); } |
157 | |
323 | |
158 | #if ECB_GCC_VERSION(4,3) |
324 | #if ECB_GCC_VERSION(4,3) |
159 | #define ecb_bswap32(x) __builtin_bswap32 (x) |
|
|
160 | #define ecb_bswap16(x) (__builtin_bswap32(x) >> 16) |
325 | #define ecb_bswap16(x) (__builtin_bswap32 (x) >> 16) |
|
|
326 | #define ecb_bswap32(x) __builtin_bswap32 (x) |
|
|
327 | #define ecb_bswap64(x) __builtin_bswap64 (x) |
161 | #else |
328 | #else |
|
|
329 | ecb_function_ uint16_t ecb_bswap16 (uint16_t x) ecb_const; |
|
|
330 | ecb_function_ uint16_t |
|
|
331 | ecb_bswap16 (uint16_t x) |
|
|
332 | { |
|
|
333 | return ecb_rotl16 (x, 8); |
|
|
334 | } |
|
|
335 | |
162 | ecb_function_ uint32_t ecb_bswap32 (uint32_t x) ecb_const; |
336 | ecb_function_ uint32_t ecb_bswap32 (uint32_t x) ecb_const; |
163 | ecb_function_ uint32_t |
337 | ecb_function_ uint32_t |
164 | ecb_bswap32 (uint32_t x) |
338 | ecb_bswap32 (uint32_t x) |
165 | { |
339 | { |
166 | return (x >> 24) |
340 | return (((uint32_t)ecb_bswap16 (x)) << 16) | ecb_bswap16 (x >> 16); |
167 | | ((x >> 8) & 0x0000ff00) |
|
|
168 | | ((x << 8) & 0x00ff0000) |
|
|
169 | | (x << 24); |
|
|
170 | } |
341 | } |
171 | |
342 | |
172 | ecb_function_ uint32_t ecb_bswap16 (uint32_t x) ecb_const; |
343 | ecb_function_ uint64_t ecb_bswap64 (uint64_t x) ecb_const; |
173 | ecb_function_ uint32_t |
344 | ecb_function_ uint64_t |
174 | ecb_bswap16 (uint32_t x) |
345 | ecb_bswap64 (uint64_t x) |
175 | { |
346 | { |
176 | return ((x >> 8) & 0xff) |
347 | return (((uint64_t)ecb_bswap32 (x)) << 32) | ecb_bswap32 (x >> 32); |
177 | | ((x << 8) & 0x00ff0000) |
|
|
178 | | (x << 24); |
|
|
179 | } |
348 | } |
180 | #endif |
349 | #endif |
181 | |
350 | |
182 | #if ECB_GCC_VERSION(4,5) |
351 | #if ECB_GCC_VERSION(4,5) |
183 | #define ecb_unreachable() __builtin_unreachable () |
352 | #define ecb_unreachable() __builtin_unreachable () |
… | |
… | |
218 | } |
387 | } |
219 | #else |
388 | #else |
220 | #define ecb_array_length(name) (sizeof (name) / sizeof (name [0])) |
389 | #define ecb_array_length(name) (sizeof (name) / sizeof (name [0])) |
221 | #endif |
390 | #endif |
222 | |
391 | |
223 | ECB_INLINE uint32_t ecb_rotr32 (uint32_t x, unsigned int count) ecb_const; |
|
|
224 | ECB_INLINE uint32_t |
|
|
225 | ecb_rotr32 (uint32_t x, unsigned int count) |
|
|
226 | { |
|
|
227 | return (x << (32 - count)) | (x >> count); |
|
|
228 | } |
|
|
229 | |
|
|
230 | ECB_INLINE uint32_t ecb_rotl32 (uint32_t x, unsigned int count) ecb_const; |
|
|
231 | ECB_INLINE uint32_t |
|
|
232 | ecb_rotl32 (uint32_t x, unsigned int count) |
|
|
233 | { |
|
|
234 | return (x >> (32 - count)) | (x << count); |
|
|
235 | } |
|
|
236 | |
|
|
237 | #endif |
392 | #endif |
238 | |
393 | |