… | |
… | |
38 | typedef signed int int32_t; |
38 | typedef signed int int32_t; |
39 | typedef unsigned int uint32_t; |
39 | typedef unsigned int uint32_t; |
40 | #if __GNUC__ |
40 | #if __GNUC__ |
41 | typedef signed long long int64_t; |
41 | typedef signed long long int64_t; |
42 | typedef unsigned long long uint64_t; |
42 | typedef unsigned long long uint64_t; |
43 | #else |
43 | #else /* _MSC_VER || __BORLANDC__ */ |
44 | typedef signed __int64 int64_t; |
44 | typedef signed __int64 int64_t; |
45 | typedef unsigned __int64 uint64_t; |
45 | typedef unsigned __int64 uint64_t; |
46 | #endif |
46 | #endif |
47 | #else |
47 | #else |
48 | #include <inttypes.h> |
48 | #include <inttypes.h> |
49 | #endif |
49 | #endif |
50 | |
50 | |
51 | /* many compilers define _GNUC_ to some versions but then only implement |
51 | /* many compilers define _GNUC_ to some versions but then only implement |
52 | * what their idiot authors think are the "more important" extensions, |
52 | * what their idiot authors think are the "more important" extensions, |
53 | * causing enourmous grief in return for some better fake benchmark numbers. |
53 | * causing enormous grief in return for some better fake benchmark numbers. |
54 | * or so. |
54 | * or so. |
55 | * 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 |
56 | * 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. |
57 | */ |
57 | */ |
58 | #ifndef ECB_GCC_VERSION |
58 | #ifndef ECB_GCC_VERSION |
… | |
… | |
61 | #else |
61 | #else |
62 | #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))) |
63 | #endif |
63 | #endif |
64 | #endif |
64 | #endif |
65 | |
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 | #endif |
|
|
74 | |
|
|
75 | #ifndef ECB_MEMORY_FENCE |
|
|
76 | #if ECB_GCC_VERSION(2,5) || defined(__INTEL_COMPILER) || defined(__clang__) |
|
|
77 | #if __i386__ |
|
|
78 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("lock; orb $0, -1(%%esp)" : : : "memory") |
|
|
79 | #define ECB_MEMORY_FENCE_ACQUIRE ECB_MEMORY_FENCE /* non-lock xchg might be enough */ |
|
|
80 | #define ECB_MEMORY_FENCE_RELEASE do { } while (0) /* unlikely to change in future cpus */ |
|
|
81 | #elif __amd64 |
|
|
82 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("mfence" : : : "memory") |
|
|
83 | #define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("lfence" : : : "memory") |
|
|
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__ ) \ |
|
|
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") |
|
|
90 | #elif defined(__ARM_ARCH_7__ ) || defined(__ARM_ARCH_7A__ ) \ |
|
|
91 | || defined(__ARM_ARCH_7M__) || defined(__ARM_ARCH_7R__ ) |
|
|
92 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("dmb" : : : "memory") |
|
|
93 | #endif |
|
|
94 | #endif |
|
|
95 | #endif |
|
|
96 | |
|
|
97 | #ifndef ECB_MEMORY_FENCE |
|
|
98 | #if ECB_GCC_VERSION(4,4) || defined(__INTEL_COMPILER) || defined(__clang__) |
|
|
99 | #define ECB_MEMORY_FENCE __sync_synchronize () |
|
|
100 | /*#define ECB_MEMORY_FENCE_ACQUIRE ({ char dummy = 0; __sync_lock_test_and_set (&dummy, 1); }) */ |
|
|
101 | /*#define ECB_MEMORY_FENCE_RELEASE ({ char dummy = 1; __sync_lock_release (&dummy ); }) */ |
|
|
102 | #elif _MSC_VER >= 1400 /* VC++ 2005 */ |
|
|
103 | #pragma intrinsic(_ReadBarrier,_WriteBarrier,_ReadWriteBarrier) |
|
|
104 | #define ECB_MEMORY_FENCE _ReadWriteBarrier () |
|
|
105 | #define ECB_MEMORY_FENCE_ACQUIRE _ReadWriteBarrier () /* according to msdn, _ReadBarrier is not a load fence */ |
|
|
106 | #define ECB_MEMORY_FENCE_RELEASE _WriteBarrier () |
|
|
107 | #elif defined(_WIN32) |
|
|
108 | #include <WinNT.h> |
|
|
109 | #define ECB_MEMORY_FENCE MemoryBarrier () /* actually just xchg on x86... scary */ |
|
|
110 | #endif |
|
|
111 | #endif |
|
|
112 | |
|
|
113 | #ifndef ECB_MEMORY_FENCE |
|
|
114 | #if !ECB_AVOID_PTHREADS |
|
|
115 | /* |
|
|
116 | * if you get undefined symbol references to pthread_mutex_lock, |
|
|
117 | * or failure to find pthread.h, then you should implement |
|
|
118 | * the ECB_MEMORY_FENCE operations for your cpu/compiler |
|
|
119 | * OR provide pthread.h and link against the posix thread library |
|
|
120 | * of your system. |
|
|
121 | */ |
|
|
122 | #include <pthread.h> |
|
|
123 | #define ECB_NEEDS_PTHREADS 1 |
|
|
124 | #define ECB_MEMORY_FENCE_NEEDS_PTHREADS 1 |
|
|
125 | |
|
|
126 | static pthread_mutex_t ecb_mf_lock = PTHREAD_MUTEX_INITIALIZER; |
|
|
127 | #define ECB_MEMORY_FENCE do { pthread_mutex_lock (&ecb_mf_lock); pthread_mutex_unlock (&ecb_mf_lock); } while (0) |
|
|
128 | #endif |
|
|
129 | #endif |
|
|
130 | |
|
|
131 | #if !defined(ECB_MEMORY_FENCE_ACQUIRE) && defined(ECB_MEMORY_FENCE) |
|
|
132 | #define ECB_MEMORY_FENCE_ACQUIRE ECB_MEMORY_FENCE |
|
|
133 | #endif |
|
|
134 | |
|
|
135 | #if !defined(ECB_MEMORY_FENCE_RELEASE) && defined(ECB_MEMORY_FENCE) |
|
|
136 | #define ECB_MEMORY_FENCE_RELEASE ECB_MEMORY_FENCE |
|
|
137 | #endif |
|
|
138 | |
|
|
139 | /*****************************************************************************/ |
|
|
140 | |
66 | #define ECB_C99 (__STDC_VERSION__ >= 199901L) |
141 | #define ECB_C99 (__STDC_VERSION__ >= 199901L) |
67 | |
142 | |
68 | #if __cplusplus |
143 | #if __cplusplus |
69 | #define ecb_inline static inline |
144 | #define ecb_inline static inline |
70 | #elif ECB_GCC_VERSION(2,5) |
145 | #elif ECB_GCC_VERSION(2,5) |
… | |
… | |
136 | #define ecb_likely(expr) ecb_expect_true (expr) |
211 | #define ecb_likely(expr) ecb_expect_true (expr) |
137 | #define ecb_unlikely(expr) ecb_expect_false (expr) |
212 | #define ecb_unlikely(expr) ecb_expect_false (expr) |
138 | |
213 | |
139 | /* count trailing zero bits and count # of one bits */ |
214 | /* count trailing zero bits and count # of one bits */ |
140 | #if ECB_GCC_VERSION(3,4) |
215 | #if ECB_GCC_VERSION(3,4) |
|
|
216 | /* we assume int == 32 bit, long == 32 or 64 bit and long long == 64 bit */ |
|
|
217 | #define ecb_ld32(x) (__builtin_clz (x) ^ 31) |
|
|
218 | #define ecb_ld64(x) (__builtin_clzll (x) ^ 63) |
141 | #define ecb_ctz32(x) __builtin_ctz (x) |
219 | #define ecb_ctz32(x) __builtin_ctz (x) |
|
|
220 | #define ecb_ctz64(x) __builtin_ctzll (x) |
142 | #define ecb_popcount32(x) __builtin_popcount (x) |
221 | #define ecb_popcount32(x) __builtin_popcount (x) |
|
|
222 | /* no popcountll */ |
143 | #else |
223 | #else |
144 | ecb_function_ int ecb_ctz32 (uint32_t x) ecb_const; |
224 | ecb_function_ int ecb_ctz32 (uint32_t x) ecb_const; |
145 | ecb_function_ int |
225 | ecb_function_ int |
146 | ecb_ctz32 (uint32_t x) |
226 | ecb_ctz32 (uint32_t x) |
147 | { |
227 | { |
148 | int r = 0; |
228 | int r = 0; |
149 | |
229 | |
150 | x &= (uint32_t)-(int32_t)x; /* this isolates the lowest bit */ |
230 | x &= ~x + 1; /* this isolates the lowest bit */ |
151 | |
231 | |
|
|
232 | #if ECB_branchless_on_i386 |
|
|
233 | r += !!(x & 0xaaaaaaaa) << 0; |
|
|
234 | r += !!(x & 0xcccccccc) << 1; |
|
|
235 | r += !!(x & 0xf0f0f0f0) << 2; |
|
|
236 | r += !!(x & 0xff00ff00) << 3; |
|
|
237 | r += !!(x & 0xffff0000) << 4; |
|
|
238 | #else |
152 | if (x & 0xaaaaaaaa) r += 1; |
239 | if (x & 0xaaaaaaaa) r += 1; |
153 | if (x & 0xcccccccc) r += 2; |
240 | if (x & 0xcccccccc) r += 2; |
154 | if (x & 0xf0f0f0f0) r += 4; |
241 | if (x & 0xf0f0f0f0) r += 4; |
155 | if (x & 0xff00ff00) r += 8; |
242 | if (x & 0xff00ff00) r += 8; |
156 | if (x & 0xffff0000) r += 16; |
243 | if (x & 0xffff0000) r += 16; |
|
|
244 | #endif |
157 | |
245 | |
158 | return r; |
246 | return r; |
|
|
247 | } |
|
|
248 | |
|
|
249 | ecb_function_ int ecb_ctz64 (uint64_t x) ecb_const; |
|
|
250 | ecb_function_ int |
|
|
251 | ecb_ctz64 (uint64_t x) |
|
|
252 | { |
|
|
253 | int shift = x & 0xffffffffU ? 0 : 32; |
|
|
254 | return ecb_ctz32 (x >> shift) + shift; |
159 | } |
255 | } |
160 | |
256 | |
161 | ecb_function_ int ecb_popcount32 (uint32_t x) ecb_const; |
257 | ecb_function_ int ecb_popcount32 (uint32_t x) ecb_const; |
162 | ecb_function_ int |
258 | ecb_function_ int |
163 | ecb_popcount32 (uint32_t x) |
259 | ecb_popcount32 (uint32_t x) |
… | |
… | |
167 | x = ((x >> 4) + x) & 0x0f0f0f0f; |
263 | x = ((x >> 4) + x) & 0x0f0f0f0f; |
168 | x *= 0x01010101; |
264 | x *= 0x01010101; |
169 | |
265 | |
170 | return x >> 24; |
266 | return x >> 24; |
171 | } |
267 | } |
|
|
268 | |
|
|
269 | ecb_function_ int ecb_ld32 (uint32_t x) ecb_const; |
|
|
270 | ecb_function_ int ecb_ld32 (uint32_t x) |
|
|
271 | { |
|
|
272 | int r = 0; |
|
|
273 | |
|
|
274 | if (x >> 16) { x >>= 16; r += 16; } |
|
|
275 | if (x >> 8) { x >>= 8; r += 8; } |
|
|
276 | if (x >> 4) { x >>= 4; r += 4; } |
|
|
277 | if (x >> 2) { x >>= 2; r += 2; } |
|
|
278 | if (x >> 1) { r += 1; } |
|
|
279 | |
|
|
280 | return r; |
|
|
281 | } |
|
|
282 | |
|
|
283 | ecb_function_ int ecb_ld64 (uint64_t x) ecb_const; |
|
|
284 | ecb_function_ int ecb_ld64 (uint64_t x) |
|
|
285 | { |
|
|
286 | int r = 0; |
|
|
287 | |
|
|
288 | if (x >> 32) { x >>= 32; r += 32; } |
|
|
289 | |
|
|
290 | return r + ecb_ld32 (x); |
|
|
291 | } |
172 | #endif |
292 | #endif |
|
|
293 | |
|
|
294 | ecb_function_ uint8_t ecb_bitrev8 (uint8_t x) |
|
|
295 | { |
|
|
296 | return ( (x * 0x0802U & 0x22110U) |
|
|
297 | | (x * 0x8020U & 0x88440U)) * 0x10101U >> 16; |
|
|
298 | } |
|
|
299 | |
|
|
300 | ecb_function_ uint16_t ecb_bitrev16 (uint16_t x) |
|
|
301 | { |
|
|
302 | x = ((x >> 1) & 0x5555) | ((x & 0x5555) << 1); |
|
|
303 | x = ((x >> 2) & 0x3333) | ((x & 0x3333) << 2); |
|
|
304 | x = ((x >> 4) & 0x0f0f) | ((x & 0x0f0f) << 4); |
|
|
305 | x = ( x >> 8 ) | ( x << 8); |
|
|
306 | |
|
|
307 | return x; |
|
|
308 | } |
|
|
309 | |
|
|
310 | ecb_function_ uint32_t ecb_bitrev32 (uint32_t x) |
|
|
311 | { |
|
|
312 | x = ((x >> 1) & 0x55555555) | ((x & 0x55555555) << 1); |
|
|
313 | x = ((x >> 2) & 0x33333333) | ((x & 0x33333333) << 2); |
|
|
314 | x = ((x >> 4) & 0x0f0f0f0f) | ((x & 0x0f0f0f0f) << 4); |
|
|
315 | x = ((x >> 8) & 0x00ff00ff) | ((x & 0x00ff00ff) << 8); |
|
|
316 | x = ( x >> 16 ) | ( x << 16); |
|
|
317 | |
|
|
318 | return x; |
|
|
319 | } |
|
|
320 | |
|
|
321 | /* popcount64 is only available on 64 bit cpus as gcc builtin */ |
|
|
322 | /* so for this version we are lazy */ |
|
|
323 | ecb_function_ int ecb_popcount64 (uint64_t x) ecb_const; |
|
|
324 | ecb_function_ int |
|
|
325 | ecb_popcount64 (uint64_t x) |
|
|
326 | { |
|
|
327 | return ecb_popcount32 (x) + ecb_popcount32 (x >> 32); |
|
|
328 | } |
|
|
329 | |
|
|
330 | ecb_inline uint8_t ecb_rotl8 (uint8_t x, unsigned int count) ecb_const; |
|
|
331 | ecb_inline uint8_t ecb_rotr8 (uint8_t x, unsigned int count) ecb_const; |
|
|
332 | ecb_inline uint16_t ecb_rotl16 (uint16_t x, unsigned int count) ecb_const; |
|
|
333 | ecb_inline uint16_t ecb_rotr16 (uint16_t x, unsigned int count) ecb_const; |
|
|
334 | ecb_inline uint32_t ecb_rotl32 (uint32_t x, unsigned int count) ecb_const; |
|
|
335 | ecb_inline uint32_t ecb_rotr32 (uint32_t x, unsigned int count) ecb_const; |
|
|
336 | ecb_inline uint64_t ecb_rotl64 (uint64_t x, unsigned int count) ecb_const; |
|
|
337 | ecb_inline uint64_t ecb_rotr64 (uint64_t x, unsigned int count) ecb_const; |
|
|
338 | |
|
|
339 | ecb_inline uint8_t ecb_rotl8 (uint8_t x, unsigned int count) { return (x >> ( 8 - count)) | (x << count); } |
|
|
340 | ecb_inline uint8_t ecb_rotr8 (uint8_t x, unsigned int count) { return (x << ( 8 - count)) | (x >> count); } |
|
|
341 | ecb_inline uint16_t ecb_rotl16 (uint16_t x, unsigned int count) { return (x >> (16 - count)) | (x << count); } |
|
|
342 | ecb_inline uint16_t ecb_rotr16 (uint16_t x, unsigned int count) { return (x << (16 - count)) | (x >> count); } |
|
|
343 | ecb_inline uint32_t ecb_rotl32 (uint32_t x, unsigned int count) { return (x >> (32 - count)) | (x << count); } |
|
|
344 | ecb_inline uint32_t ecb_rotr32 (uint32_t x, unsigned int count) { return (x << (32 - count)) | (x >> count); } |
|
|
345 | ecb_inline uint64_t ecb_rotl64 (uint64_t x, unsigned int count) { return (x >> (64 - count)) | (x << count); } |
|
|
346 | ecb_inline uint64_t ecb_rotr64 (uint64_t x, unsigned int count) { return (x << (64 - count)) | (x >> count); } |
173 | |
347 | |
174 | #if ECB_GCC_VERSION(4,3) |
348 | #if ECB_GCC_VERSION(4,3) |
175 | #define ecb_bswap32(x) __builtin_bswap32 (x) |
|
|
176 | #define ecb_bswap16(x) (__builtin_bswap32(x) >> 16) |
349 | #define ecb_bswap16(x) (__builtin_bswap32 (x) >> 16) |
|
|
350 | #define ecb_bswap32(x) __builtin_bswap32 (x) |
|
|
351 | #define ecb_bswap64(x) __builtin_bswap64 (x) |
177 | #else |
352 | #else |
|
|
353 | ecb_function_ uint16_t ecb_bswap16 (uint16_t x) ecb_const; |
|
|
354 | ecb_function_ uint16_t |
|
|
355 | ecb_bswap16 (uint16_t x) |
|
|
356 | { |
|
|
357 | return ecb_rotl16 (x, 8); |
|
|
358 | } |
|
|
359 | |
178 | ecb_function_ uint32_t ecb_bswap32 (uint32_t x) ecb_const; |
360 | ecb_function_ uint32_t ecb_bswap32 (uint32_t x) ecb_const; |
179 | ecb_function_ uint32_t |
361 | ecb_function_ uint32_t |
180 | ecb_bswap32 (uint32_t x) |
362 | ecb_bswap32 (uint32_t x) |
181 | { |
363 | { |
182 | return (x >> 24) |
364 | return (((uint32_t)ecb_bswap16 (x)) << 16) | ecb_bswap16 (x >> 16); |
183 | | ((x >> 8) & 0x0000ff00) |
|
|
184 | | ((x << 8) & 0x00ff0000) |
|
|
185 | | (x << 24); |
|
|
186 | } |
365 | } |
187 | |
366 | |
188 | ecb_function_ uint32_t ecb_bswap16 (uint32_t x) ecb_const; |
367 | ecb_function_ uint64_t ecb_bswap64 (uint64_t x) ecb_const; |
189 | ecb_function_ uint32_t |
368 | ecb_function_ uint64_t |
190 | ecb_bswap16 (uint32_t x) |
369 | ecb_bswap64 (uint64_t x) |
191 | { |
370 | { |
192 | return ((x >> 8) & 0xff) |
371 | return (((uint64_t)ecb_bswap32 (x)) << 32) | ecb_bswap32 (x >> 32); |
193 | | ((x << 8) & 0x00ff0000) |
|
|
194 | | (x << 24); |
|
|
195 | } |
372 | } |
196 | #endif |
373 | #endif |
197 | |
374 | |
198 | #if ECB_GCC_VERSION(4,5) |
375 | #if ECB_GCC_VERSION(4,5) |
199 | #define ecb_unreachable() __builtin_unreachable () |
376 | #define ecb_unreachable() __builtin_unreachable () |
… | |
… | |
223 | #define ecb_mod(m,n) ((m) % (n) + ((m) % (n) < 0 ? (n) : 0)) |
400 | #define ecb_mod(m,n) ((m) % (n) + ((m) % (n) < 0 ? (n) : 0)) |
224 | #else |
401 | #else |
225 | #define ecb_mod(m,n) ((m) < 0 ? ((n) - 1 - ((-1 - (m)) % (n))) : ((m) % (n))) |
402 | #define ecb_mod(m,n) ((m) < 0 ? ((n) - 1 - ((-1 - (m)) % (n))) : ((m) % (n))) |
226 | #endif |
403 | #endif |
227 | |
404 | |
|
|
405 | #if __cplusplus |
|
|
406 | template<typename T> |
|
|
407 | static inline T ecb_div_rd (T val, T div) |
|
|
408 | { |
|
|
409 | return val < 0 ? - ((-val + div - 1) / div) : (val ) / div; |
|
|
410 | } |
|
|
411 | template<typename T> |
|
|
412 | static inline T ecb_div_ru (T val, T div) |
|
|
413 | { |
|
|
414 | return val < 0 ? - ((-val ) / div) : (val + div - 1) / div; |
|
|
415 | } |
|
|
416 | #else |
|
|
417 | #define ecb_div_rd(val,div) ((val) < 0 ? - ((-(val) + (div) - 1) / (div)) : ((val) ) / (div)) |
|
|
418 | #define ecb_div_ru(val,div) ((val) < 0 ? - ((-(val) ) / (div)) : ((val) + (div) - 1) / (div)) |
|
|
419 | #endif |
|
|
420 | |
228 | #if ecb_cplusplus_does_not_suck |
421 | #if ecb_cplusplus_does_not_suck |
229 | /* does not work for local types (http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2657.htm) */ |
422 | /* does not work for local types (http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2657.htm) */ |
230 | template<typename T, int N> |
423 | template<typename T, int N> |
231 | static inline int ecb_array_length (const T (&arr)[N]) |
424 | static inline int ecb_array_length (const T (&arr)[N]) |
232 | { |
425 | { |
… | |
… | |
234 | } |
427 | } |
235 | #else |
428 | #else |
236 | #define ecb_array_length(name) (sizeof (name) / sizeof (name [0])) |
429 | #define ecb_array_length(name) (sizeof (name) / sizeof (name [0])) |
237 | #endif |
430 | #endif |
238 | |
431 | |
239 | ecb_inline uint32_t ecb_rotr32 (uint32_t x, unsigned int count) ecb_const; |
|
|
240 | ecb_inline uint32_t |
|
|
241 | ecb_rotr32 (uint32_t x, unsigned int count) |
|
|
242 | { |
|
|
243 | return (x << (32 - count)) | (x >> count); |
|
|
244 | } |
|
|
245 | |
|
|
246 | ecb_inline uint32_t ecb_rotl32 (uint32_t x, unsigned int count) ecb_const; |
|
|
247 | ecb_inline uint32_t |
|
|
248 | ecb_rotl32 (uint32_t x, unsigned int count) |
|
|
249 | { |
|
|
250 | return (x >> (32 - count)) | (x << count); |
|
|
251 | } |
|
|
252 | |
|
|
253 | #endif |
432 | #endif |
254 | |
433 | |