1 | /* |
1 | /* |
2 | * libecb |
2 | * libecb - http://software.schmorp.de/pkg/libecb |
3 | * |
3 | * |
4 | * Copyright (©) 2009-2011 Marc Alexander Lehmann |
4 | * Copyright (©) 2009-2012 Marc Alexander Lehmann <libecb@schmorp.de> |
|
|
5 | * Copyright (©) 2011 Emanuele Giaquinta |
5 | * All rights reserved. |
6 | * All rights reserved. |
6 | * |
7 | * |
7 | * Redistribution and use in source and binary forms, with or without modifica- |
8 | * Redistribution and use in source and binary forms, with or without modifica- |
8 | * tion, are permitted provided that the following conditions are met: |
9 | * tion, are permitted provided that the following conditions are met: |
9 | * |
10 | * |
… | |
… | |
27 | */ |
28 | */ |
28 | |
29 | |
29 | #ifndef ECB_H |
30 | #ifndef ECB_H |
30 | #define ECB_H |
31 | #define ECB_H |
31 | |
32 | |
|
|
33 | /* 16 bits major, 16 bits minor */ |
|
|
34 | #define ECB_VERSION 0x00010001 |
|
|
35 | |
|
|
36 | #ifdef _WIN32 |
|
|
37 | typedef signed char int8_t; |
|
|
38 | typedef unsigned char uint8_t; |
|
|
39 | typedef signed short int16_t; |
|
|
40 | typedef unsigned short uint16_t; |
|
|
41 | typedef signed int int32_t; |
|
|
42 | typedef unsigned int uint32_t; |
|
|
43 | #if __GNUC__ |
|
|
44 | typedef signed long long int64_t; |
|
|
45 | typedef unsigned long long uint64_t; |
|
|
46 | #else /* _MSC_VER || __BORLANDC__ */ |
|
|
47 | typedef signed __int64 int64_t; |
|
|
48 | typedef unsigned __int64 uint64_t; |
|
|
49 | #endif |
|
|
50 | #ifdef _WIN64 |
|
|
51 | #define ECB_PTRSIZE 8 |
|
|
52 | typedef uint64_t uintptr_t; |
|
|
53 | typedef int64_t intptr_t; |
|
|
54 | #else |
|
|
55 | #define ECB_PTRSIZE 4 |
|
|
56 | typedef uint32_t uintptr_t; |
|
|
57 | typedef int32_t intptr_t; |
|
|
58 | #endif |
|
|
59 | typedef intptr_t ptrdiff_t; |
|
|
60 | #else |
|
|
61 | #include <inttypes.h> |
|
|
62 | #if UINTMAX_MAX > 0xffffffffU |
|
|
63 | #define ECB_PTRSIZE 8 |
|
|
64 | #else |
|
|
65 | #define ECB_PTRSIZE 4 |
|
|
66 | #endif |
|
|
67 | #endif |
|
|
68 | |
|
|
69 | /* many compilers define _GNUC_ to some versions but then only implement |
|
|
70 | * what their idiot authors think are the "more important" extensions, |
|
|
71 | * causing enormous grief in return for some better fake benchmark numbers. |
|
|
72 | * or so. |
|
|
73 | * we try to detect these and simply assume they are not gcc - if they have |
|
|
74 | * an issue with that they should have done it right in the first place. |
|
|
75 | */ |
|
|
76 | #ifndef ECB_GCC_VERSION |
|
|
77 | #if !defined __GNUC_MINOR__ || defined __INTEL_COMPILER || defined __SUNPRO_C || defined __SUNPRO_CC || defined __llvm__ || defined __clang__ |
|
|
78 | #define ECB_GCC_VERSION(major,minor) 0 |
|
|
79 | #else |
32 | #define ECB_GCC_VERSION(major,minor) (__GNUC__ > (major) || (__GNUC__ == (major) && __GNUC_MINOR__ >= (minor))) |
80 | #define ECB_GCC_VERSION(major,minor) (__GNUC__ > (major) || (__GNUC__ == (major) && __GNUC_MINOR__ >= (minor))) |
|
|
81 | #endif |
|
|
82 | #endif |
33 | |
83 | |
34 | #define ECB_HEADER_INLINE static inline |
84 | #define ECB_C99 (__STDC_VERSION__ >= 199901L) |
|
|
85 | #define ECB_C11 (__STDC_VERSION__ >= 201112L) |
|
|
86 | |
|
|
87 | /*****************************************************************************/ |
|
|
88 | |
|
|
89 | /* ECB_NO_THREADS - ecb is not used by multiple threads, ever */ |
|
|
90 | /* ECB_NO_SMP - ecb might be used in multiple threads, but only on a single cpu */ |
|
|
91 | |
|
|
92 | #if ECB_NO_THREADS |
|
|
93 | # define ECB_NO_SMP 1 |
|
|
94 | #endif |
|
|
95 | |
|
|
96 | #if ECB_NO_THREADS || ECB_NO_SMP |
|
|
97 | #define ECB_MEMORY_FENCE do { } while (0) |
|
|
98 | #endif |
|
|
99 | |
|
|
100 | #ifndef ECB_MEMORY_FENCE |
|
|
101 | #if ECB_C11 && !defined __STDC_NO_ATOMICS__ |
|
|
102 | /* we assume that these memory fences work on all variables/all memory accesses, */ |
|
|
103 | /* not just C11 atomics and atomic accesses */ |
|
|
104 | #include <stdatomic.h> |
|
|
105 | #define ECB_MEMORY_FENCE atomic_thread_fence (memory_order_acq_rel) |
|
|
106 | #define ECB_MEMORY_FENCE_ACQUIRE atomic_thread_fence (memory_order_acquire) |
|
|
107 | #define ECB_MEMORY_FENCE_RELEASE atomic_thread_fence (memory_order_release) |
|
|
108 | #endif |
|
|
109 | #endif |
|
|
110 | |
|
|
111 | #ifndef ECB_MEMORY_FENCE |
|
|
112 | #if ECB_GCC_VERSION(2,5) || defined __INTEL_COMPILER || (__llvm__ && __GNUC__) || __SUNPRO_C >= 0x5110 || __SUNPRO_CC >= 0x5110 |
|
|
113 | #if __i386 || __i386__ |
|
|
114 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("lock; orb $0, -1(%%esp)" : : : "memory") |
|
|
115 | #define ECB_MEMORY_FENCE_ACQUIRE ECB_MEMORY_FENCE /* non-lock xchg might be enough */ |
|
|
116 | #define ECB_MEMORY_FENCE_RELEASE do { } while (0) /* unlikely to change in future cpus */ |
|
|
117 | #elif __amd64 || __amd64__ || __x86_64 || __x86_64__ |
|
|
118 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("mfence" : : : "memory") |
|
|
119 | #define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("lfence" : : : "memory") |
|
|
120 | #define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("sfence") /* play safe - not needed in any current cpu */ |
|
|
121 | #elif __powerpc__ || __ppc__ || __powerpc64__ || __ppc64__ |
|
|
122 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("sync" : : : "memory") |
|
|
123 | #elif defined __ARM_ARCH_6__ || defined __ARM_ARCH_6J__ \ |
|
|
124 | || defined __ARM_ARCH_6K__ || defined __ARM_ARCH_6ZK__ |
|
|
125 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("mcr p15,0,%0,c7,c10,5" : : "r" (0) : "memory") |
|
|
126 | #elif defined __ARM_ARCH_7__ || defined __ARM_ARCH_7A__ \ |
|
|
127 | || defined __ARM_ARCH_7M__ || defined __ARM_ARCH_7R__ |
|
|
128 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("dmb" : : : "memory") |
|
|
129 | #elif __sparc || __sparc__ |
|
|
130 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("membar #LoadStore | #LoadLoad | #StoreStore | #StoreLoad | " : : : "memory") |
|
|
131 | #define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("membar #LoadStore | #LoadLoad" : : : "memory") |
|
|
132 | #define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("membar #LoadStore | #StoreStore") |
|
|
133 | #elif defined __s390__ || defined __s390x__ |
|
|
134 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("bcr 15,0" : : : "memory") |
|
|
135 | #elif defined __mips__ |
|
|
136 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("sync" : : : "memory") |
|
|
137 | #elif defined __alpha__ |
|
|
138 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("mb" : : : "memory") |
|
|
139 | #endif |
|
|
140 | #endif |
|
|
141 | #endif |
|
|
142 | |
|
|
143 | #ifndef ECB_MEMORY_FENCE |
|
|
144 | #if ECB_GCC_VERSION(4,4) || defined __INTEL_COMPILER || defined __clang__ |
|
|
145 | #define ECB_MEMORY_FENCE __sync_synchronize () |
|
|
146 | /*#define ECB_MEMORY_FENCE_ACQUIRE ({ char dummy = 0; __sync_lock_test_and_set (&dummy, 1); }) */ |
|
|
147 | /*#define ECB_MEMORY_FENCE_RELEASE ({ char dummy = 1; __sync_lock_release (&dummy ); }) */ |
|
|
148 | #elif _MSC_VER >= 1400 /* VC++ 2005 */ |
|
|
149 | #pragma intrinsic(_ReadBarrier,_WriteBarrier,_ReadWriteBarrier) |
|
|
150 | #define ECB_MEMORY_FENCE _ReadWriteBarrier () |
|
|
151 | #define ECB_MEMORY_FENCE_ACQUIRE _ReadWriteBarrier () /* according to msdn, _ReadBarrier is not a load fence */ |
|
|
152 | #define ECB_MEMORY_FENCE_RELEASE _WriteBarrier () |
|
|
153 | #elif defined _WIN32 |
|
|
154 | #include <WinNT.h> |
|
|
155 | #define ECB_MEMORY_FENCE MemoryBarrier () /* actually just xchg on x86... scary */ |
|
|
156 | #elif __SUNPRO_C >= 0x5110 || __SUNPRO_CC >= 0x5110 |
|
|
157 | #include <mbarrier.h> |
|
|
158 | #define ECB_MEMORY_FENCE __machine_rw_barrier () |
|
|
159 | #define ECB_MEMORY_FENCE_ACQUIRE __machine_r_barrier () |
|
|
160 | #define ECB_MEMORY_FENCE_RELEASE __machine_w_barrier () |
|
|
161 | #elif __xlC__ |
|
|
162 | #define ECB_MEMORY_FENCE __sync () |
|
|
163 | #endif |
|
|
164 | #endif |
|
|
165 | |
|
|
166 | #ifndef ECB_MEMORY_FENCE |
|
|
167 | #if !ECB_AVOID_PTHREADS |
|
|
168 | /* |
|
|
169 | * if you get undefined symbol references to pthread_mutex_lock, |
|
|
170 | * or failure to find pthread.h, then you should implement |
|
|
171 | * the ECB_MEMORY_FENCE operations for your cpu/compiler |
|
|
172 | * OR provide pthread.h and link against the posix thread library |
|
|
173 | * of your system. |
|
|
174 | */ |
|
|
175 | #include <pthread.h> |
|
|
176 | #define ECB_NEEDS_PTHREADS 1 |
|
|
177 | #define ECB_MEMORY_FENCE_NEEDS_PTHREADS 1 |
|
|
178 | |
|
|
179 | static pthread_mutex_t ecb_mf_lock = PTHREAD_MUTEX_INITIALIZER; |
|
|
180 | #define ECB_MEMORY_FENCE do { pthread_mutex_lock (&ecb_mf_lock); pthread_mutex_unlock (&ecb_mf_lock); } while (0) |
|
|
181 | #endif |
|
|
182 | #endif |
|
|
183 | |
|
|
184 | #if !defined ECB_MEMORY_FENCE_ACQUIRE && defined ECB_MEMORY_FENCE |
|
|
185 | #define ECB_MEMORY_FENCE_ACQUIRE ECB_MEMORY_FENCE |
|
|
186 | #endif |
|
|
187 | |
|
|
188 | #if !defined ECB_MEMORY_FENCE_RELEASE && defined ECB_MEMORY_FENCE |
|
|
189 | #define ECB_MEMORY_FENCE_RELEASE ECB_MEMORY_FENCE |
|
|
190 | #endif |
|
|
191 | |
|
|
192 | /*****************************************************************************/ |
|
|
193 | |
|
|
194 | #if __cplusplus |
|
|
195 | #define ecb_inline static inline |
|
|
196 | #elif ECB_GCC_VERSION(2,5) |
|
|
197 | #define ecb_inline static __inline__ |
|
|
198 | #elif ECB_C99 |
|
|
199 | #define ecb_inline static inline |
|
|
200 | #else |
|
|
201 | #define ecb_inline static |
|
|
202 | #endif |
|
|
203 | |
|
|
204 | #if ECB_GCC_VERSION(3,3) |
|
|
205 | #define ecb_restrict __restrict__ |
|
|
206 | #elif ECB_C99 |
|
|
207 | #define ecb_restrict restrict |
|
|
208 | #else |
|
|
209 | #define ecb_restrict |
|
|
210 | #endif |
|
|
211 | |
|
|
212 | typedef int ecb_bool; |
|
|
213 | |
|
|
214 | #define ECB_CONCAT_(a, b) a ## b |
|
|
215 | #define ECB_CONCAT(a, b) ECB_CONCAT_(a, b) |
|
|
216 | #define ECB_STRINGIFY_(a) # a |
|
|
217 | #define ECB_STRINGIFY(a) ECB_STRINGIFY_(a) |
|
|
218 | |
|
|
219 | #define ecb_function_ ecb_inline |
35 | |
220 | |
36 | #if ECB_GCC_VERSION(3,1) |
221 | #if ECB_GCC_VERSION(3,1) |
37 | # define ecb_attribute(attrlist) __attribute__(attrlist) |
222 | #define ecb_attribute(attrlist) __attribute__(attrlist) |
38 | # define ecb_is_constant(c) __builtin_constant_p (c) |
223 | #define ecb_is_constant(expr) __builtin_constant_p (expr) |
39 | # define ecb_expect(expr,value) __builtin_expect ((expr),(value)) |
224 | #define ecb_expect(expr,value) __builtin_expect ((expr),(value)) |
40 | # define ecb_prefetch(addr,rw,locality) __builtin_prefetch (addr, rw, locality) |
225 | #define ecb_prefetch(addr,rw,locality) __builtin_prefetch (addr, rw, locality) |
41 | #else |
226 | #else |
42 | # define ecb_attribute(attrlist) |
227 | #define ecb_attribute(attrlist) |
43 | # define ecb_is_constant(c) 0 |
228 | #define ecb_is_constant(expr) 0 |
44 | # define ecb_expect(expr,value) (expr) |
229 | #define ecb_expect(expr,value) (expr) |
45 | # define ecb_prefetch(addr,rw,locality) |
230 | #define ecb_prefetch(addr,rw,locality) |
46 | #endif |
231 | #endif |
47 | |
232 | |
48 | /* no emulation for ecb_decltype */ |
233 | /* no emulation for ecb_decltype */ |
49 | #if ECB_GCC_VERSION(4,5) |
234 | #if ECB_GCC_VERSION(4,5) |
50 | # define ecb_decltype(x) __decltype(x) |
235 | #define ecb_decltype(x) __decltype(x) |
51 | #elif ECB_GCC_VERSION(3,0) |
236 | #elif ECB_GCC_VERSION(3,0) |
52 | # define ecb_decltype(x) typeof(x) |
237 | #define ecb_decltype(x) __typeof(x) |
53 | #endif |
238 | #endif |
54 | |
239 | |
55 | #define ecb_noinline ecb_attribute ((noinline)) |
240 | #define ecb_noinline ecb_attribute ((__noinline__)) |
56 | #define ecb_noreturn ecb_attribute ((noreturn)) |
|
|
57 | #define ecb_unused ecb_attribute ((unused)) |
241 | #define ecb_unused ecb_attribute ((__unused__)) |
58 | #define ecb_const ecb_attribute ((const)) |
242 | #define ecb_const ecb_attribute ((__const__)) |
59 | #define ecb_pure ecb_attribute ((pure)) |
243 | #define ecb_pure ecb_attribute ((__pure__)) |
60 | #define ecb_hot ecb_attribute ((hot)) // 4.3 |
|
|
61 | #define ecb_cold ecb_attribute ((cold)) // 4.3 |
|
|
62 | |
244 | |
|
|
245 | #if ECB_C11 |
|
|
246 | #define ecb_noreturn _Noreturn |
|
|
247 | #else |
|
|
248 | #define ecb_noreturn ecb_attribute ((__noreturn__)) |
|
|
249 | #endif |
|
|
250 | |
63 | #if ECB_GCC_VERSION(4,5) |
251 | #if ECB_GCC_VERSION(4,3) |
64 | # define ecb_unreachable() __builtin_unreachable () |
252 | #define ecb_artificial ecb_attribute ((__artificial__)) |
|
|
253 | #define ecb_hot ecb_attribute ((__hot__)) |
|
|
254 | #define ecb_cold ecb_attribute ((__cold__)) |
65 | #else |
255 | #else |
66 | /* this seems to work fine, but gcc always emits a warning for it :/ */ |
256 | #define ecb_artificial |
67 | ECB_HEADER_INLINE void ecb_unreachable () ecb_attribute ((noreturn)); |
257 | #define ecb_hot |
68 | ECB_HEADER_INLINE void ecb_unreachable () { } |
258 | #define ecb_cold |
69 | #endif |
259 | #endif |
70 | |
260 | |
71 | /* put into ifs if you are very sure that the expression */ |
261 | /* put around conditional expressions if you are very sure that the */ |
72 | /* is mostly true or mosty false. note that these return */ |
262 | /* expression is mostly true or mostly false. note that these return */ |
73 | /* booleans, not the expression. */ |
263 | /* booleans, not the expression. */ |
74 | #define ecb_expect_false(expr) ecb_expect ((expr) ? 1 : 0, 0) |
264 | #define ecb_expect_false(expr) ecb_expect (!!(expr), 0) |
75 | #define ecb_expect_true(expr) ecb_expect ((expr) ? 1 : 0, 1) |
265 | #define ecb_expect_true(expr) ecb_expect (!!(expr), 1) |
76 | |
266 | /* for compatibility to the rest of the world */ |
77 | /* try to tell the compiler that some condition is definitely true */ |
267 | #define ecb_likely(expr) ecb_expect_true (expr) |
78 | #define ecb_assume(cond) do { if (!(cond)) unreachable (); } while (0) |
268 | #define ecb_unlikely(expr) ecb_expect_false (expr) |
79 | |
269 | |
80 | /* count trailing zero bits and count # of one bits */ |
270 | /* count trailing zero bits and count # of one bits */ |
81 | #if ECB_GCC_VERSION(3,4) |
271 | #if ECB_GCC_VERSION(3,4) |
82 | ECB_HEADER_INLINE int ecb_ctz (unsigned int x) { return __builtin_ctz (x); } |
272 | /* we assume int == 32 bit, long == 32 or 64 bit and long long == 64 bit */ |
83 | ECB_HEADER_INLINE int ecb_popcount (unsigned int x) { return __builtin_popcount (x); } |
273 | #define ecb_ld32(x) (__builtin_clz (x) ^ 31) |
|
|
274 | #define ecb_ld64(x) (__builtin_clzll (x) ^ 63) |
|
|
275 | #define ecb_ctz32(x) __builtin_ctz (x) |
|
|
276 | #define ecb_ctz64(x) __builtin_ctzll (x) |
|
|
277 | #define ecb_popcount32(x) __builtin_popcount (x) |
|
|
278 | /* no popcountll */ |
84 | #else |
279 | #else |
85 | ECB_HEADER_INLINE |
280 | ecb_function_ int ecb_ctz32 (uint32_t x) ecb_const; |
86 | ecb_ctz (unsigned int x) ecb_const |
281 | ecb_function_ int |
|
|
282 | ecb_ctz32 (uint32_t x) |
|
|
283 | { |
|
|
284 | int r = 0; |
|
|
285 | |
|
|
286 | x &= ~x + 1; /* this isolates the lowest bit */ |
|
|
287 | |
|
|
288 | #if ECB_branchless_on_i386 |
|
|
289 | r += !!(x & 0xaaaaaaaa) << 0; |
|
|
290 | r += !!(x & 0xcccccccc) << 1; |
|
|
291 | r += !!(x & 0xf0f0f0f0) << 2; |
|
|
292 | r += !!(x & 0xff00ff00) << 3; |
|
|
293 | r += !!(x & 0xffff0000) << 4; |
|
|
294 | #else |
|
|
295 | if (x & 0xaaaaaaaa) r += 1; |
|
|
296 | if (x & 0xcccccccc) r += 2; |
|
|
297 | if (x & 0xf0f0f0f0) r += 4; |
|
|
298 | if (x & 0xff00ff00) r += 8; |
|
|
299 | if (x & 0xffff0000) r += 16; |
|
|
300 | #endif |
|
|
301 | |
|
|
302 | return r; |
|
|
303 | } |
|
|
304 | |
|
|
305 | ecb_function_ int ecb_ctz64 (uint64_t x) ecb_const; |
|
|
306 | ecb_function_ int |
|
|
307 | ecb_ctz64 (uint64_t x) |
|
|
308 | { |
|
|
309 | int shift = x & 0xffffffffU ? 0 : 32; |
|
|
310 | return ecb_ctz32 (x >> shift) + shift; |
|
|
311 | } |
|
|
312 | |
|
|
313 | ecb_function_ int ecb_popcount32 (uint32_t x) ecb_const; |
|
|
314 | ecb_function_ int |
|
|
315 | ecb_popcount32 (uint32_t x) |
|
|
316 | { |
|
|
317 | x -= (x >> 1) & 0x55555555; |
|
|
318 | x = ((x >> 2) & 0x33333333) + (x & 0x33333333); |
|
|
319 | x = ((x >> 4) + x) & 0x0f0f0f0f; |
|
|
320 | x *= 0x01010101; |
|
|
321 | |
|
|
322 | return x >> 24; |
|
|
323 | } |
|
|
324 | |
|
|
325 | ecb_function_ int ecb_ld32 (uint32_t x) ecb_const; |
|
|
326 | ecb_function_ int ecb_ld32 (uint32_t x) |
|
|
327 | { |
|
|
328 | int r = 0; |
|
|
329 | |
|
|
330 | if (x >> 16) { x >>= 16; r += 16; } |
|
|
331 | if (x >> 8) { x >>= 8; r += 8; } |
|
|
332 | if (x >> 4) { x >>= 4; r += 4; } |
|
|
333 | if (x >> 2) { x >>= 2; r += 2; } |
|
|
334 | if (x >> 1) { r += 1; } |
|
|
335 | |
|
|
336 | return r; |
|
|
337 | } |
|
|
338 | |
|
|
339 | ecb_function_ int ecb_ld64 (uint64_t x) ecb_const; |
|
|
340 | ecb_function_ int ecb_ld64 (uint64_t x) |
|
|
341 | { |
|
|
342 | int r = 0; |
|
|
343 | |
|
|
344 | if (x >> 32) { x >>= 32; r += 32; } |
|
|
345 | |
|
|
346 | return r + ecb_ld32 (x); |
|
|
347 | } |
|
|
348 | #endif |
|
|
349 | |
|
|
350 | ecb_function_ ecb_bool ecb_is_pot32 (uint32_t x) ecb_const; |
|
|
351 | ecb_function_ ecb_bool ecb_is_pot32 (uint32_t x) { return !(x & (x - 1)); } |
|
|
352 | ecb_function_ ecb_bool ecb_is_pot64 (uint64_t x) ecb_const; |
|
|
353 | ecb_function_ ecb_bool ecb_is_pot64 (uint64_t x) { return !(x & (x - 1)); } |
|
|
354 | |
|
|
355 | ecb_function_ uint8_t ecb_bitrev8 (uint8_t x) ecb_const; |
|
|
356 | ecb_function_ uint8_t ecb_bitrev8 (uint8_t x) |
87 | { |
357 | { |
88 | int r = 0; |
358 | return ( (x * 0x0802U & 0x22110U) |
89 | |
359 | | (x * 0x8020U & 0x88440U)) * 0x10101U >> 16; |
90 | x &= -x; // this isolates the lowest bit |
|
|
91 | |
|
|
92 | if (x & 0xaaaaaaaa) r += 1; |
|
|
93 | if (x & 0xcccccccc) r += 2; |
|
|
94 | if (x & 0xf0f0f0f0) r += 4; |
|
|
95 | if (x & 0xff00ff00) r += 8; |
|
|
96 | if (x & 0xffff0000) r += 16; |
|
|
97 | |
|
|
98 | return r; |
|
|
99 | } |
360 | } |
100 | |
361 | |
101 | ECB_HEADER_INLINE |
362 | ecb_function_ uint16_t ecb_bitrev16 (uint16_t x) ecb_const; |
102 | ecb_popcount (unsigned int x) ecb_const |
363 | ecb_function_ uint16_t ecb_bitrev16 (uint16_t x) |
103 | { |
364 | { |
104 | x -= (x >> 1) & 0x55555555; |
365 | x = ((x >> 1) & 0x5555) | ((x & 0x5555) << 1); |
105 | x = ((x >> 2) & 0x33333333) + (x & 0x33333333); |
366 | x = ((x >> 2) & 0x3333) | ((x & 0x3333) << 2); |
106 | x = ((x >> 4) + x) & 0x0f0f0f0f; |
367 | x = ((x >> 4) & 0x0f0f) | ((x & 0x0f0f) << 4); |
107 | x *= 0x01010101; |
368 | x = ( x >> 8 ) | ( x << 8); |
108 | |
369 | |
109 | return x >> 24; |
370 | return x; |
110 | } |
371 | } |
111 | #endif |
|
|
112 | |
372 | |
113 | ECB_HEADER_INLINE unsigned char |
373 | ecb_function_ uint32_t ecb_bitrev32 (uint32_t x) ecb_const; |
|
|
374 | ecb_function_ uint32_t ecb_bitrev32 (uint32_t x) |
|
|
375 | { |
|
|
376 | x = ((x >> 1) & 0x55555555) | ((x & 0x55555555) << 1); |
|
|
377 | x = ((x >> 2) & 0x33333333) | ((x & 0x33333333) << 2); |
|
|
378 | x = ((x >> 4) & 0x0f0f0f0f) | ((x & 0x0f0f0f0f) << 4); |
|
|
379 | x = ((x >> 8) & 0x00ff00ff) | ((x & 0x00ff00ff) << 8); |
|
|
380 | x = ( x >> 16 ) | ( x << 16); |
|
|
381 | |
|
|
382 | return x; |
|
|
383 | } |
|
|
384 | |
|
|
385 | /* popcount64 is only available on 64 bit cpus as gcc builtin */ |
|
|
386 | /* so for this version we are lazy */ |
|
|
387 | ecb_function_ int ecb_popcount64 (uint64_t x) ecb_const; |
|
|
388 | ecb_function_ int |
|
|
389 | ecb_popcount64 (uint64_t x) |
|
|
390 | { |
|
|
391 | return ecb_popcount32 (x) + ecb_popcount32 (x >> 32); |
|
|
392 | } |
|
|
393 | |
|
|
394 | ecb_inline uint8_t ecb_rotl8 (uint8_t x, unsigned int count) ecb_const; |
|
|
395 | ecb_inline uint8_t ecb_rotr8 (uint8_t x, unsigned int count) ecb_const; |
|
|
396 | ecb_inline uint16_t ecb_rotl16 (uint16_t x, unsigned int count) ecb_const; |
|
|
397 | ecb_inline uint16_t ecb_rotr16 (uint16_t x, unsigned int count) ecb_const; |
|
|
398 | ecb_inline uint32_t ecb_rotl32 (uint32_t x, unsigned int count) ecb_const; |
|
|
399 | ecb_inline uint32_t ecb_rotr32 (uint32_t x, unsigned int count) ecb_const; |
|
|
400 | ecb_inline uint64_t ecb_rotl64 (uint64_t x, unsigned int count) ecb_const; |
|
|
401 | ecb_inline uint64_t ecb_rotr64 (uint64_t x, unsigned int count) ecb_const; |
|
|
402 | |
|
|
403 | ecb_inline uint8_t ecb_rotl8 (uint8_t x, unsigned int count) { return (x >> ( 8 - count)) | (x << count); } |
|
|
404 | ecb_inline uint8_t ecb_rotr8 (uint8_t x, unsigned int count) { return (x << ( 8 - count)) | (x >> count); } |
|
|
405 | ecb_inline uint16_t ecb_rotl16 (uint16_t x, unsigned int count) { return (x >> (16 - count)) | (x << count); } |
|
|
406 | ecb_inline uint16_t ecb_rotr16 (uint16_t x, unsigned int count) { return (x << (16 - count)) | (x >> count); } |
|
|
407 | ecb_inline uint32_t ecb_rotl32 (uint32_t x, unsigned int count) { return (x >> (32 - count)) | (x << count); } |
|
|
408 | ecb_inline uint32_t ecb_rotr32 (uint32_t x, unsigned int count) { return (x << (32 - count)) | (x >> count); } |
|
|
409 | ecb_inline uint64_t ecb_rotl64 (uint64_t x, unsigned int count) { return (x >> (64 - count)) | (x << count); } |
|
|
410 | ecb_inline uint64_t ecb_rotr64 (uint64_t x, unsigned int count) { return (x << (64 - count)) | (x >> count); } |
|
|
411 | |
|
|
412 | #if ECB_GCC_VERSION(4,3) |
|
|
413 | #define ecb_bswap16(x) (__builtin_bswap32 (x) >> 16) |
|
|
414 | #define ecb_bswap32(x) __builtin_bswap32 (x) |
|
|
415 | #define ecb_bswap64(x) __builtin_bswap64 (x) |
|
|
416 | #else |
|
|
417 | ecb_function_ uint16_t ecb_bswap16 (uint16_t x) ecb_const; |
|
|
418 | ecb_function_ uint16_t |
|
|
419 | ecb_bswap16 (uint16_t x) |
|
|
420 | { |
|
|
421 | return ecb_rotl16 (x, 8); |
|
|
422 | } |
|
|
423 | |
|
|
424 | ecb_function_ uint32_t ecb_bswap32 (uint32_t x) ecb_const; |
|
|
425 | ecb_function_ uint32_t |
|
|
426 | ecb_bswap32 (uint32_t x) |
|
|
427 | { |
|
|
428 | return (((uint32_t)ecb_bswap16 (x)) << 16) | ecb_bswap16 (x >> 16); |
|
|
429 | } |
|
|
430 | |
|
|
431 | ecb_function_ uint64_t ecb_bswap64 (uint64_t x) ecb_const; |
|
|
432 | ecb_function_ uint64_t |
|
|
433 | ecb_bswap64 (uint64_t x) |
|
|
434 | { |
|
|
435 | return (((uint64_t)ecb_bswap32 (x)) << 32) | ecb_bswap32 (x >> 32); |
|
|
436 | } |
|
|
437 | #endif |
|
|
438 | |
|
|
439 | #if ECB_GCC_VERSION(4,5) |
|
|
440 | #define ecb_unreachable() __builtin_unreachable () |
|
|
441 | #else |
|
|
442 | /* this seems to work fine, but gcc always emits a warning for it :/ */ |
|
|
443 | ecb_inline void ecb_unreachable (void) ecb_noreturn; |
|
|
444 | ecb_inline void ecb_unreachable (void) { } |
|
|
445 | #endif |
|
|
446 | |
|
|
447 | /* try to tell the compiler that some condition is definitely true */ |
|
|
448 | #define ecb_assume(cond) do { if (!(cond)) ecb_unreachable (); } while (0) |
|
|
449 | |
|
|
450 | ecb_inline unsigned char ecb_byteorder_helper (void) ecb_const; |
|
|
451 | ecb_inline unsigned char |
114 | ecb_byteorder_helper () ecb_const |
452 | ecb_byteorder_helper (void) |
115 | { |
453 | { |
116 | const uint32_t u = 0x11223344; |
454 | const uint32_t u = 0x11223344; |
117 | return *(unsigned char *)&u; |
455 | return *(unsigned char *)&u; |
118 | } |
456 | } |
119 | |
457 | |
|
|
458 | ecb_inline ecb_bool ecb_big_endian (void) ecb_const; |
120 | ECB_HEADER_INLINE bool ecb_big_endian () ecb_const { return ecb_byteorder_helper () == 0x11; }; |
459 | ecb_inline ecb_bool ecb_big_endian (void) { return ecb_byteorder_helper () == 0x11; } |
121 | ECB_HEADER_INLINE bool ecb_network () ecb_const { return big_endian (); }; |
460 | ecb_inline ecb_bool ecb_little_endian (void) ecb_const; |
122 | ECB_HEADER_INLINE bool ecb_little_endian () ecb_const { return ecb_byteorder_helper () == 0x44; }; |
461 | ecb_inline ecb_bool ecb_little_endian (void) { return ecb_byteorder_helper () == 0x44; } |
123 | ECB_HEADER_INLINE bool ecb_vax () ecb_const { return little_endian (); }; |
|
|
124 | |
462 | |
125 | |
463 | #if ECB_GCC_VERSION(3,0) || ECB_C99 |
|
|
464 | #define ecb_mod(m,n) ((m) % (n) + ((m) % (n) < 0 ? (n) : 0)) |
|
|
465 | #else |
|
|
466 | #define ecb_mod(m,n) ((m) < 0 ? ((n) - 1 - ((-1 - (m)) % (n))) : ((m) % (n))) |
126 | #endif |
467 | #endif |
127 | |
468 | |
|
|
469 | #if __cplusplus |
|
|
470 | template<typename T> |
|
|
471 | static inline T ecb_div_rd (T val, T div) |
|
|
472 | { |
|
|
473 | return val < 0 ? - ((-val + div - 1) / div) : (val ) / div; |
|
|
474 | } |
|
|
475 | template<typename T> |
|
|
476 | static inline T ecb_div_ru (T val, T div) |
|
|
477 | { |
|
|
478 | return val < 0 ? - ((-val ) / div) : (val + div - 1) / div; |
|
|
479 | } |
|
|
480 | #else |
|
|
481 | #define ecb_div_rd(val,div) ((val) < 0 ? - ((-(val) + (div) - 1) / (div)) : ((val) ) / (div)) |
|
|
482 | #define ecb_div_ru(val,div) ((val) < 0 ? - ((-(val) ) / (div)) : ((val) + (div) - 1) / (div)) |
|
|
483 | #endif |
|
|
484 | |
|
|
485 | #if ecb_cplusplus_does_not_suck |
|
|
486 | /* does not work for local types (http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2657.htm) */ |
|
|
487 | template<typename T, int N> |
|
|
488 | static inline int ecb_array_length (const T (&arr)[N]) |
|
|
489 | { |
|
|
490 | return N; |
|
|
491 | } |
|
|
492 | #else |
|
|
493 | #define ecb_array_length(name) (sizeof (name) / sizeof (name [0])) |
|
|
494 | #endif |
|
|
495 | |
|
|
496 | #endif |
|
|
497 | |