| 1 |
/* |
| 2 |
* libecb - http://software.schmorp.de/pkg/libecb |
| 3 |
* |
| 4 |
* Copyright (©) 2009-2012 Marc Alexander Lehmann <libecb@schmorp.de> |
| 5 |
* Copyright (©) 2011 Emanuele Giaquinta |
| 6 |
* All rights reserved. |
| 7 |
* |
| 8 |
* Redistribution and use in source and binary forms, with or without modifica- |
| 9 |
* tion, are permitted provided that the following conditions are met: |
| 10 |
* |
| 11 |
* 1. Redistributions of source code must retain the above copyright notice, |
| 12 |
* this list of conditions and the following disclaimer. |
| 13 |
* |
| 14 |
* 2. Redistributions in binary form must reproduce the above copyright |
| 15 |
* notice, this list of conditions and the following disclaimer in the |
| 16 |
* documentation and/or other materials provided with the distribution. |
| 17 |
* |
| 18 |
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR IMPLIED |
| 19 |
* WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MER- |
| 20 |
* CHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO |
| 21 |
* EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPE- |
| 22 |
* CIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, |
| 23 |
* PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; |
| 24 |
* OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, |
| 25 |
* WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTH- |
| 26 |
* ERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED |
| 27 |
* OF THE POSSIBILITY OF SUCH DAMAGE. |
| 28 |
*/ |
| 29 |
|
| 30 |
#ifndef ECB_H |
| 31 |
#define ECB_H |
| 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 |
| 48 |
#include <inttypes.h> |
| 49 |
#endif |
| 50 |
|
| 51 |
/* many compilers define _GNUC_ to some versions but then only implement |
| 52 |
* what their idiot authors think are the "more important" extensions, |
| 53 |
* causing enormous grief in return for some better fake benchmark numbers. |
| 54 |
* or so. |
| 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. |
| 57 |
*/ |
| 58 |
#ifndef ECB_GCC_VERSION |
| 59 |
#if !defined(__GNUC_MINOR__) || defined(__INTEL_COMPILER) || defined(__SUNPRO_C) || defined(__SUNPRO_CC) || defined(__llvm__) || defined(__clang__) |
| 60 |
#define ECB_GCC_VERSION(major,minor) 0 |
| 61 |
#else |
| 62 |
#define ECB_GCC_VERSION(major,minor) (__GNUC__ > (major) || (__GNUC__ == (major) && __GNUC_MINOR__ >= (minor))) |
| 63 |
#endif |
| 64 |
#endif |
| 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 |
| 72 |
# define ECB_NO_SMP 1 |
| 73 |
#endif |
| 74 |
|
| 75 |
#if ECB_NO_THREADS || ECB_NO_SMP |
| 76 |
#define ECB_MEMORY_FENCE do { } while (0) |
| 77 |
#endif |
| 78 |
|
| 79 |
#ifndef ECB_MEMORY_FENCE |
| 80 |
#if ECB_GCC_VERSION(2,5) || defined(__INTEL_COMPILER) || (__llvm__ && __GNUC__) || __SUNPRO_C >= 0x5110 || __SUNPRO_CC >= 0x5110 |
| 81 |
#if __i386 || __i386__ |
| 82 |
#define ECB_MEMORY_FENCE __asm__ __volatile__ ("lock; orb $0, -1(%%esp)" : : : "memory") |
| 83 |
#define ECB_MEMORY_FENCE_ACQUIRE ECB_MEMORY_FENCE /* non-lock xchg might be enough */ |
| 84 |
#define ECB_MEMORY_FENCE_RELEASE do { } while (0) /* unlikely to change in future cpus */ |
| 85 |
#elif __amd64 || __amd64__ || __x86_64 || __x86_64__ |
| 86 |
#define ECB_MEMORY_FENCE __asm__ __volatile__ ("mfence" : : : "memory") |
| 87 |
#define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("lfence" : : : "memory") |
| 88 |
#define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("sfence") /* play safe - not needed in any current cpu */ |
| 89 |
#elif __powerpc__ || __ppc__ || __powerpc64__ || __ppc64__ |
| 90 |
#define ECB_MEMORY_FENCE __asm__ __volatile__ ("sync" : : : "memory") |
| 91 |
#elif defined(__ARM_ARCH_6__ ) || defined(__ARM_ARCH_6J__ ) \ |
| 92 |
|| defined(__ARM_ARCH_6K__) || defined(__ARM_ARCH_6ZK__) |
| 93 |
#define ECB_MEMORY_FENCE __asm__ __volatile__ ("mcr p15,0,%0,c7,c10,5" : : "r" (0) : "memory") |
| 94 |
#elif defined(__ARM_ARCH_7__ ) || defined(__ARM_ARCH_7A__ ) \ |
| 95 |
|| defined(__ARM_ARCH_7M__) || defined(__ARM_ARCH_7R__ ) |
| 96 |
#define ECB_MEMORY_FENCE __asm__ __volatile__ ("dmb" : : : "memory") |
| 97 |
#elif __sparc || __sparc__ |
| 98 |
#define ECB_MEMORY_FENCE __asm__ __volatile__ ("membar #LoadStore | #LoadLoad | #StoreStore | #StoreLoad | " : : : "memory") |
| 99 |
#define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("membar #LoadStore | #LoadLoad" : : : "memory") |
| 100 |
#define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("membar #LoadStore | #StoreStore") |
| 101 |
#elif defined(__s390__) || defined(__s390x__) |
| 102 |
#define ECB_MEMORY_FENCE __asm__ __volatile__ ("bcr 15,0" : : : "memory") |
| 103 |
#elif defined(__mips__) |
| 104 |
#define ECB_MEMORY_FENCE __asm__ __volatile__ ("sync" : : : "memory") |
| 105 |
#elif __xlC__ |
| 106 |
#define ECB_MEMORY_FENCE __lwsync () |
| 107 |
#endif |
| 108 |
#endif |
| 109 |
#endif |
| 110 |
|
| 111 |
#ifndef ECB_MEMORY_FENCE |
| 112 |
#if ECB_GCC_VERSION(4,4) || defined(__INTEL_COMPILER) || defined(__clang__) |
| 113 |
#define ECB_MEMORY_FENCE __sync_synchronize () |
| 114 |
/*#define ECB_MEMORY_FENCE_ACQUIRE ({ char dummy = 0; __sync_lock_test_and_set (&dummy, 1); }) */ |
| 115 |
/*#define ECB_MEMORY_FENCE_RELEASE ({ char dummy = 1; __sync_lock_release (&dummy ); }) */ |
| 116 |
#elif _MSC_VER >= 1400 /* VC++ 2005 */ |
| 117 |
#pragma intrinsic(_ReadBarrier,_WriteBarrier,_ReadWriteBarrier) |
| 118 |
#define ECB_MEMORY_FENCE _ReadWriteBarrier () |
| 119 |
#define ECB_MEMORY_FENCE_ACQUIRE _ReadWriteBarrier () /* according to msdn, _ReadBarrier is not a load fence */ |
| 120 |
#define ECB_MEMORY_FENCE_RELEASE _WriteBarrier () |
| 121 |
#elif defined(_WIN32) |
| 122 |
#include <WinNT.h> |
| 123 |
#define ECB_MEMORY_FENCE MemoryBarrier () /* actually just xchg on x86... scary */ |
| 124 |
#elif __SUNPRO_C >= 0x5110 || __SUNPRO_CC >= 0x5110 |
| 125 |
#include <mbarrier.h> |
| 126 |
#define ECB_MEMORY_FENCE __machine_rw_barrier () |
| 127 |
#define ECB_MEMORY_FENCE_ACQUIRE __machine_r_barrier () |
| 128 |
#define ECB_MEMORY_FENCE_RELEASE __machine_w_barrier () |
| 129 |
#endif |
| 130 |
#endif |
| 131 |
|
| 132 |
#ifndef ECB_MEMORY_FENCE |
| 133 |
#if !ECB_AVOID_PTHREADS |
| 134 |
/* |
| 135 |
* if you get undefined symbol references to pthread_mutex_lock, |
| 136 |
* or failure to find pthread.h, then you should implement |
| 137 |
* the ECB_MEMORY_FENCE operations for your cpu/compiler |
| 138 |
* OR provide pthread.h and link against the posix thread library |
| 139 |
* of your system. |
| 140 |
*/ |
| 141 |
#include <pthread.h> |
| 142 |
#define ECB_NEEDS_PTHREADS 1 |
| 143 |
#define ECB_MEMORY_FENCE_NEEDS_PTHREADS 1 |
| 144 |
|
| 145 |
static pthread_mutex_t ecb_mf_lock = PTHREAD_MUTEX_INITIALIZER; |
| 146 |
#define ECB_MEMORY_FENCE do { pthread_mutex_lock (&ecb_mf_lock); pthread_mutex_unlock (&ecb_mf_lock); } while (0) |
| 147 |
#endif |
| 148 |
#endif |
| 149 |
|
| 150 |
#if !defined(ECB_MEMORY_FENCE_ACQUIRE) && defined(ECB_MEMORY_FENCE) |
| 151 |
#define ECB_MEMORY_FENCE_ACQUIRE ECB_MEMORY_FENCE |
| 152 |
#endif |
| 153 |
|
| 154 |
#if !defined(ECB_MEMORY_FENCE_RELEASE) && defined(ECB_MEMORY_FENCE) |
| 155 |
#define ECB_MEMORY_FENCE_RELEASE ECB_MEMORY_FENCE |
| 156 |
#endif |
| 157 |
|
| 158 |
/*****************************************************************************/ |
| 159 |
|
| 160 |
#define ECB_C99 (__STDC_VERSION__ >= 199901L) |
| 161 |
|
| 162 |
#if __cplusplus |
| 163 |
#define ecb_inline static inline |
| 164 |
#elif ECB_GCC_VERSION(2,5) |
| 165 |
#define ecb_inline static __inline__ |
| 166 |
#elif ECB_C99 |
| 167 |
#define ecb_inline static inline |
| 168 |
#else |
| 169 |
#define ecb_inline static |
| 170 |
#endif |
| 171 |
|
| 172 |
#if ECB_GCC_VERSION(3,3) |
| 173 |
#define ecb_restrict __restrict__ |
| 174 |
#elif ECB_C99 |
| 175 |
#define ecb_restrict restrict |
| 176 |
#else |
| 177 |
#define ecb_restrict |
| 178 |
#endif |
| 179 |
|
| 180 |
typedef int ecb_bool; |
| 181 |
|
| 182 |
#define ECB_CONCAT_(a, b) a ## b |
| 183 |
#define ECB_CONCAT(a, b) ECB_CONCAT_(a, b) |
| 184 |
#define ECB_STRINGIFY_(a) # a |
| 185 |
#define ECB_STRINGIFY(a) ECB_STRINGIFY_(a) |
| 186 |
|
| 187 |
#define ecb_function_ ecb_inline |
| 188 |
|
| 189 |
#if ECB_GCC_VERSION(3,1) |
| 190 |
#define ecb_attribute(attrlist) __attribute__(attrlist) |
| 191 |
#define ecb_is_constant(expr) __builtin_constant_p (expr) |
| 192 |
#define ecb_expect(expr,value) __builtin_expect ((expr),(value)) |
| 193 |
#define ecb_prefetch(addr,rw,locality) __builtin_prefetch (addr, rw, locality) |
| 194 |
#else |
| 195 |
#define ecb_attribute(attrlist) |
| 196 |
#define ecb_is_constant(expr) 0 |
| 197 |
#define ecb_expect(expr,value) (expr) |
| 198 |
#define ecb_prefetch(addr,rw,locality) |
| 199 |
#endif |
| 200 |
|
| 201 |
/* no emulation for ecb_decltype */ |
| 202 |
#if ECB_GCC_VERSION(4,5) |
| 203 |
#define ecb_decltype(x) __decltype(x) |
| 204 |
#elif ECB_GCC_VERSION(3,0) |
| 205 |
#define ecb_decltype(x) __typeof(x) |
| 206 |
#endif |
| 207 |
|
| 208 |
#define ecb_noinline ecb_attribute ((__noinline__)) |
| 209 |
#define ecb_noreturn ecb_attribute ((__noreturn__)) |
| 210 |
#define ecb_unused ecb_attribute ((__unused__)) |
| 211 |
#define ecb_const ecb_attribute ((__const__)) |
| 212 |
#define ecb_pure ecb_attribute ((__pure__)) |
| 213 |
|
| 214 |
#if ECB_GCC_VERSION(4,3) |
| 215 |
#define ecb_artificial ecb_attribute ((__artificial__)) |
| 216 |
#define ecb_hot ecb_attribute ((__hot__)) |
| 217 |
#define ecb_cold ecb_attribute ((__cold__)) |
| 218 |
#else |
| 219 |
#define ecb_artificial |
| 220 |
#define ecb_hot |
| 221 |
#define ecb_cold |
| 222 |
#endif |
| 223 |
|
| 224 |
/* put around conditional expressions if you are very sure that the */ |
| 225 |
/* expression is mostly true or mostly false. note that these return */ |
| 226 |
/* booleans, not the expression. */ |
| 227 |
#define ecb_expect_false(expr) ecb_expect (!!(expr), 0) |
| 228 |
#define ecb_expect_true(expr) ecb_expect (!!(expr), 1) |
| 229 |
/* for compatibility to the rest of the world */ |
| 230 |
#define ecb_likely(expr) ecb_expect_true (expr) |
| 231 |
#define ecb_unlikely(expr) ecb_expect_false (expr) |
| 232 |
|
| 233 |
/* count trailing zero bits and count # of one bits */ |
| 234 |
#if ECB_GCC_VERSION(3,4) |
| 235 |
/* we assume int == 32 bit, long == 32 or 64 bit and long long == 64 bit */ |
| 236 |
#define ecb_ld32(x) (__builtin_clz (x) ^ 31) |
| 237 |
#define ecb_ld64(x) (__builtin_clzll (x) ^ 63) |
| 238 |
#define ecb_ctz32(x) __builtin_ctz (x) |
| 239 |
#define ecb_ctz64(x) __builtin_ctzll (x) |
| 240 |
#define ecb_popcount32(x) __builtin_popcount (x) |
| 241 |
/* no popcountll */ |
| 242 |
#else |
| 243 |
ecb_function_ int ecb_ctz32 (uint32_t x) ecb_const; |
| 244 |
ecb_function_ int |
| 245 |
ecb_ctz32 (uint32_t x) |
| 246 |
{ |
| 247 |
int r = 0; |
| 248 |
|
| 249 |
x &= ~x + 1; /* this isolates the lowest bit */ |
| 250 |
|
| 251 |
#if ECB_branchless_on_i386 |
| 252 |
r += !!(x & 0xaaaaaaaa) << 0; |
| 253 |
r += !!(x & 0xcccccccc) << 1; |
| 254 |
r += !!(x & 0xf0f0f0f0) << 2; |
| 255 |
r += !!(x & 0xff00ff00) << 3; |
| 256 |
r += !!(x & 0xffff0000) << 4; |
| 257 |
#else |
| 258 |
if (x & 0xaaaaaaaa) r += 1; |
| 259 |
if (x & 0xcccccccc) r += 2; |
| 260 |
if (x & 0xf0f0f0f0) r += 4; |
| 261 |
if (x & 0xff00ff00) r += 8; |
| 262 |
if (x & 0xffff0000) r += 16; |
| 263 |
#endif |
| 264 |
|
| 265 |
return r; |
| 266 |
} |
| 267 |
|
| 268 |
ecb_function_ int ecb_ctz64 (uint64_t x) ecb_const; |
| 269 |
ecb_function_ int |
| 270 |
ecb_ctz64 (uint64_t x) |
| 271 |
{ |
| 272 |
int shift = x & 0xffffffffU ? 0 : 32; |
| 273 |
return ecb_ctz32 (x >> shift) + shift; |
| 274 |
} |
| 275 |
|
| 276 |
ecb_function_ int ecb_popcount32 (uint32_t x) ecb_const; |
| 277 |
ecb_function_ int |
| 278 |
ecb_popcount32 (uint32_t x) |
| 279 |
{ |
| 280 |
x -= (x >> 1) & 0x55555555; |
| 281 |
x = ((x >> 2) & 0x33333333) + (x & 0x33333333); |
| 282 |
x = ((x >> 4) + x) & 0x0f0f0f0f; |
| 283 |
x *= 0x01010101; |
| 284 |
|
| 285 |
return x >> 24; |
| 286 |
} |
| 287 |
|
| 288 |
ecb_function_ int ecb_ld32 (uint32_t x) ecb_const; |
| 289 |
ecb_function_ int ecb_ld32 (uint32_t x) |
| 290 |
{ |
| 291 |
int r = 0; |
| 292 |
|
| 293 |
if (x >> 16) { x >>= 16; r += 16; } |
| 294 |
if (x >> 8) { x >>= 8; r += 8; } |
| 295 |
if (x >> 4) { x >>= 4; r += 4; } |
| 296 |
if (x >> 2) { x >>= 2; r += 2; } |
| 297 |
if (x >> 1) { r += 1; } |
| 298 |
|
| 299 |
return r; |
| 300 |
} |
| 301 |
|
| 302 |
ecb_function_ int ecb_ld64 (uint64_t x) ecb_const; |
| 303 |
ecb_function_ int ecb_ld64 (uint64_t x) |
| 304 |
{ |
| 305 |
int r = 0; |
| 306 |
|
| 307 |
if (x >> 32) { x >>= 32; r += 32; } |
| 308 |
|
| 309 |
return r + ecb_ld32 (x); |
| 310 |
} |
| 311 |
#endif |
| 312 |
|
| 313 |
ecb_function_ uint8_t ecb_bitrev8 (uint8_t x) ecb_const; |
| 314 |
ecb_function_ uint8_t ecb_bitrev8 (uint8_t x) |
| 315 |
{ |
| 316 |
return ( (x * 0x0802U & 0x22110U) |
| 317 |
| (x * 0x8020U & 0x88440U)) * 0x10101U >> 16; |
| 318 |
} |
| 319 |
|
| 320 |
ecb_function_ uint16_t ecb_bitrev16 (uint16_t x) ecb_const; |
| 321 |
ecb_function_ uint16_t ecb_bitrev16 (uint16_t x) |
| 322 |
{ |
| 323 |
x = ((x >> 1) & 0x5555) | ((x & 0x5555) << 1); |
| 324 |
x = ((x >> 2) & 0x3333) | ((x & 0x3333) << 2); |
| 325 |
x = ((x >> 4) & 0x0f0f) | ((x & 0x0f0f) << 4); |
| 326 |
x = ( x >> 8 ) | ( x << 8); |
| 327 |
|
| 328 |
return x; |
| 329 |
} |
| 330 |
|
| 331 |
ecb_function_ uint32_t ecb_bitrev32 (uint32_t x) ecb_const; |
| 332 |
ecb_function_ uint32_t ecb_bitrev32 (uint32_t x) |
| 333 |
{ |
| 334 |
x = ((x >> 1) & 0x55555555) | ((x & 0x55555555) << 1); |
| 335 |
x = ((x >> 2) & 0x33333333) | ((x & 0x33333333) << 2); |
| 336 |
x = ((x >> 4) & 0x0f0f0f0f) | ((x & 0x0f0f0f0f) << 4); |
| 337 |
x = ((x >> 8) & 0x00ff00ff) | ((x & 0x00ff00ff) << 8); |
| 338 |
x = ( x >> 16 ) | ( x << 16); |
| 339 |
|
| 340 |
return x; |
| 341 |
} |
| 342 |
|
| 343 |
/* popcount64 is only available on 64 bit cpus as gcc builtin */ |
| 344 |
/* so for this version we are lazy */ |
| 345 |
ecb_function_ int ecb_popcount64 (uint64_t x) ecb_const; |
| 346 |
ecb_function_ int |
| 347 |
ecb_popcount64 (uint64_t x) |
| 348 |
{ |
| 349 |
return ecb_popcount32 (x) + ecb_popcount32 (x >> 32); |
| 350 |
} |
| 351 |
|
| 352 |
ecb_inline uint8_t ecb_rotl8 (uint8_t x, unsigned int count) ecb_const; |
| 353 |
ecb_inline uint8_t ecb_rotr8 (uint8_t x, unsigned int count) ecb_const; |
| 354 |
ecb_inline uint16_t ecb_rotl16 (uint16_t x, unsigned int count) ecb_const; |
| 355 |
ecb_inline uint16_t ecb_rotr16 (uint16_t x, unsigned int count) ecb_const; |
| 356 |
ecb_inline uint32_t ecb_rotl32 (uint32_t x, unsigned int count) ecb_const; |
| 357 |
ecb_inline uint32_t ecb_rotr32 (uint32_t x, unsigned int count) ecb_const; |
| 358 |
ecb_inline uint64_t ecb_rotl64 (uint64_t x, unsigned int count) ecb_const; |
| 359 |
ecb_inline uint64_t ecb_rotr64 (uint64_t x, unsigned int count) ecb_const; |
| 360 |
|
| 361 |
ecb_inline uint8_t ecb_rotl8 (uint8_t x, unsigned int count) { return (x >> ( 8 - count)) | (x << count); } |
| 362 |
ecb_inline uint8_t ecb_rotr8 (uint8_t x, unsigned int count) { return (x << ( 8 - count)) | (x >> count); } |
| 363 |
ecb_inline uint16_t ecb_rotl16 (uint16_t x, unsigned int count) { return (x >> (16 - count)) | (x << count); } |
| 364 |
ecb_inline uint16_t ecb_rotr16 (uint16_t x, unsigned int count) { return (x << (16 - count)) | (x >> count); } |
| 365 |
ecb_inline uint32_t ecb_rotl32 (uint32_t x, unsigned int count) { return (x >> (32 - count)) | (x << count); } |
| 366 |
ecb_inline uint32_t ecb_rotr32 (uint32_t x, unsigned int count) { return (x << (32 - count)) | (x >> count); } |
| 367 |
ecb_inline uint64_t ecb_rotl64 (uint64_t x, unsigned int count) { return (x >> (64 - count)) | (x << count); } |
| 368 |
ecb_inline uint64_t ecb_rotr64 (uint64_t x, unsigned int count) { return (x << (64 - count)) | (x >> count); } |
| 369 |
|
| 370 |
#if ECB_GCC_VERSION(4,3) |
| 371 |
#define ecb_bswap16(x) (__builtin_bswap32 (x) >> 16) |
| 372 |
#define ecb_bswap32(x) __builtin_bswap32 (x) |
| 373 |
#define ecb_bswap64(x) __builtin_bswap64 (x) |
| 374 |
#else |
| 375 |
ecb_function_ uint16_t ecb_bswap16 (uint16_t x) ecb_const; |
| 376 |
ecb_function_ uint16_t |
| 377 |
ecb_bswap16 (uint16_t x) |
| 378 |
{ |
| 379 |
return ecb_rotl16 (x, 8); |
| 380 |
} |
| 381 |
|
| 382 |
ecb_function_ uint32_t ecb_bswap32 (uint32_t x) ecb_const; |
| 383 |
ecb_function_ uint32_t |
| 384 |
ecb_bswap32 (uint32_t x) |
| 385 |
{ |
| 386 |
return (((uint32_t)ecb_bswap16 (x)) << 16) | ecb_bswap16 (x >> 16); |
| 387 |
} |
| 388 |
|
| 389 |
ecb_function_ uint64_t ecb_bswap64 (uint64_t x) ecb_const; |
| 390 |
ecb_function_ uint64_t |
| 391 |
ecb_bswap64 (uint64_t x) |
| 392 |
{ |
| 393 |
return (((uint64_t)ecb_bswap32 (x)) << 32) | ecb_bswap32 (x >> 32); |
| 394 |
} |
| 395 |
#endif |
| 396 |
|
| 397 |
#if ECB_GCC_VERSION(4,5) |
| 398 |
#define ecb_unreachable() __builtin_unreachable () |
| 399 |
#else |
| 400 |
/* this seems to work fine, but gcc always emits a warning for it :/ */ |
| 401 |
ecb_inline void ecb_unreachable (void) ecb_noreturn; |
| 402 |
ecb_inline void ecb_unreachable (void) { } |
| 403 |
#endif |
| 404 |
|
| 405 |
/* try to tell the compiler that some condition is definitely true */ |
| 406 |
#define ecb_assume(cond) do { if (!(cond)) ecb_unreachable (); } while (0) |
| 407 |
|
| 408 |
ecb_inline unsigned char ecb_byteorder_helper (void) ecb_const; |
| 409 |
ecb_inline unsigned char |
| 410 |
ecb_byteorder_helper (void) |
| 411 |
{ |
| 412 |
const uint32_t u = 0x11223344; |
| 413 |
return *(unsigned char *)&u; |
| 414 |
} |
| 415 |
|
| 416 |
ecb_inline ecb_bool ecb_big_endian (void) ecb_const; |
| 417 |
ecb_inline ecb_bool ecb_big_endian (void) { return ecb_byteorder_helper () == 0x11; } |
| 418 |
ecb_inline ecb_bool ecb_little_endian (void) ecb_const; |
| 419 |
ecb_inline ecb_bool ecb_little_endian (void) { return ecb_byteorder_helper () == 0x44; } |
| 420 |
|
| 421 |
#if ECB_GCC_VERSION(3,0) || ECB_C99 |
| 422 |
#define ecb_mod(m,n) ((m) % (n) + ((m) % (n) < 0 ? (n) : 0)) |
| 423 |
#else |
| 424 |
#define ecb_mod(m,n) ((m) < 0 ? ((n) - 1 - ((-1 - (m)) % (n))) : ((m) % (n))) |
| 425 |
#endif |
| 426 |
|
| 427 |
#if __cplusplus |
| 428 |
template<typename T> |
| 429 |
static inline T ecb_div_rd (T val, T div) |
| 430 |
{ |
| 431 |
return val < 0 ? - ((-val + div - 1) / div) : (val ) / div; |
| 432 |
} |
| 433 |
template<typename T> |
| 434 |
static inline T ecb_div_ru (T val, T div) |
| 435 |
{ |
| 436 |
return val < 0 ? - ((-val ) / div) : (val + div - 1) / div; |
| 437 |
} |
| 438 |
#else |
| 439 |
#define ecb_div_rd(val,div) ((val) < 0 ? - ((-(val) + (div) - 1) / (div)) : ((val) ) / (div)) |
| 440 |
#define ecb_div_ru(val,div) ((val) < 0 ? - ((-(val) ) / (div)) : ((val) + (div) - 1) / (div)) |
| 441 |
#endif |
| 442 |
|
| 443 |
#if ecb_cplusplus_does_not_suck |
| 444 |
/* does not work for local types (http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2657.htm) */ |
| 445 |
template<typename T, int N> |
| 446 |
static inline int ecb_array_length (const T (&arr)[N]) |
| 447 |
{ |
| 448 |
return N; |
| 449 |
} |
| 450 |
#else |
| 451 |
#define ecb_array_length(name) (sizeof (name) / sizeof (name [0])) |
| 452 |
#endif |
| 453 |
|
| 454 |
#endif |
| 455 |
|
