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 | * Copyright (©) 2011 Emanuele Giaquinta |
6 | * All rights reserved. |
6 | * All rights reserved. |
7 | * |
7 | * |
8 | * 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- |
9 | * tion, are permitted provided that the following conditions are met: |
9 | * tion, are permitted provided that the following conditions are met: |
… | |
… | |
28 | */ |
28 | */ |
29 | |
29 | |
30 | #ifndef ECB_H |
30 | #ifndef ECB_H |
31 | #define ECB_H |
31 | #define ECB_H |
32 | |
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 |
33 | #include <inttypes.h> |
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 |
34 | |
68 | |
35 | /* many compilers define _GNUC_ to some versions but then only implement |
69 | /* many compilers define _GNUC_ to some versions but then only implement |
36 | * what their idiot authors think are the "more important" extensions, |
70 | * what their idiot authors think are the "more important" extensions, |
37 | * causing enourmous grief for some better fake benchmark numbers or so. |
71 | * causing enormous grief in return for some better fake benchmark numbers. |
|
|
72 | * or so. |
38 | * we try to detect these and simply assume they are not gcc - if they have |
73 | * we try to detect these and simply assume they are not gcc - if they have |
39 | * an issue with that they should have done it right in the first place. |
74 | * an issue with that they should have done it right in the first place. |
40 | */ |
75 | */ |
41 | #ifndef ECB_GCC_VERSION |
76 | #ifndef ECB_GCC_VERSION |
42 | # if defined(__INTEL_COMPILER) || defined(__SUNPRO_C) || defined(__llvm__) |
77 | #if !defined __GNUC_MINOR__ || defined __INTEL_COMPILER || defined __SUNPRO_C || defined __SUNPRO_CC || defined __llvm__ || defined __clang__ |
43 | # define ECB_GCC_VERSION(major,minor) 0 |
78 | #define ECB_GCC_VERSION(major,minor) 0 |
44 | # else |
79 | #else |
45 | # 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 |
46 | # endif |
82 | #endif |
|
|
83 | |
|
|
84 | #define ECB_C (__STDC__+0) /* this assumes that __STDC__ is either empty or a number */ |
|
|
85 | #define ECB_C99 (__STDC_VERSION__ >= 199901L) |
|
|
86 | #define ECB_C11 (__STDC_VERSION__ >= 201112L) |
|
|
87 | #define ECB_CPP (__cplusplus+0) |
|
|
88 | #define ECB_CPP98 (__cplusplus >= 199711L) |
|
|
89 | #define ECB_CPP11 (__cplusplus >= 201103L) |
|
|
90 | |
|
|
91 | /*****************************************************************************/ |
|
|
92 | |
|
|
93 | /* ECB_NO_THREADS - ecb is not used by multiple threads, ever */ |
|
|
94 | /* ECB_NO_SMP - ecb might be used in multiple threads, but only on a single cpu */ |
|
|
95 | |
|
|
96 | #if ECB_NO_THREADS |
|
|
97 | # define ECB_NO_SMP 1 |
|
|
98 | #endif |
|
|
99 | |
|
|
100 | #if ECB_NO_SMP |
|
|
101 | #define ECB_MEMORY_FENCE do { } while (0) |
|
|
102 | #endif |
|
|
103 | |
|
|
104 | #ifndef ECB_MEMORY_FENCE |
|
|
105 | #if ECB_C11 && !defined __STDC_NO_ATOMICS__ |
|
|
106 | /* we assume that these memory fences work on all variables/all memory accesses, */ |
|
|
107 | /* not just C11 atomics and atomic accesses */ |
|
|
108 | #include <stdatomic.h> |
|
|
109 | #if 0 |
|
|
110 | #define ECB_MEMORY_FENCE atomic_thread_fence (memory_order_acq_rel) |
|
|
111 | #define ECB_MEMORY_FENCE_ACQUIRE atomic_thread_fence (memory_order_acquire) |
|
|
112 | #define ECB_MEMORY_FENCE_RELEASE atomic_thread_fence (memory_order_release) |
|
|
113 | #else |
|
|
114 | /* the above *should* be enough in my book, but after experiences with gcc-4.7 */ |
|
|
115 | /* and clang, better play safe */ |
|
|
116 | #define ECB_MEMORY_FENCE atomic_thread_fence (memory_order_seq_cst) |
|
|
117 | #endif |
47 | #endif |
118 | #endif |
|
|
119 | #endif |
48 | |
120 | |
|
|
121 | #ifndef ECB_MEMORY_FENCE |
|
|
122 | #if ECB_GCC_VERSION(2,5) || defined __INTEL_COMPILER || (__llvm__ && __GNUC__) || __SUNPRO_C >= 0x5110 || __SUNPRO_CC >= 0x5110 |
|
|
123 | #if __i386 || __i386__ |
|
|
124 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("lock; orb $0, -1(%%esp)" : : : "memory") |
|
|
125 | #define ECB_MEMORY_FENCE_ACQUIRE ECB_MEMORY_FENCE /* non-lock xchg might be enough */ |
|
|
126 | #define ECB_MEMORY_FENCE_RELEASE do { } while (0) /* unlikely to change in future cpus */ |
|
|
127 | #elif __amd64 || __amd64__ || __x86_64 || __x86_64__ |
|
|
128 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("mfence" : : : "memory") |
|
|
129 | #define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("lfence" : : : "memory") |
|
|
130 | #define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("sfence") /* play safe - not needed in any current cpu */ |
|
|
131 | #elif __powerpc__ || __ppc__ || __powerpc64__ || __ppc64__ |
|
|
132 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("sync" : : : "memory") |
|
|
133 | #elif defined __ARM_ARCH_6__ || defined __ARM_ARCH_6J__ \ |
|
|
134 | || defined __ARM_ARCH_6K__ || defined __ARM_ARCH_6ZK__ |
|
|
135 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("mcr p15,0,%0,c7,c10,5" : : "r" (0) : "memory") |
|
|
136 | #elif defined __ARM_ARCH_7__ || defined __ARM_ARCH_7A__ \ |
|
|
137 | || defined __ARM_ARCH_7M__ || defined __ARM_ARCH_7R__ |
|
|
138 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("dmb" : : : "memory") |
|
|
139 | #elif __sparc || __sparc__ |
|
|
140 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("membar #LoadStore | #LoadLoad | #StoreStore | #StoreLoad | " : : : "memory") |
|
|
141 | #define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("membar #LoadStore | #LoadLoad" : : : "memory") |
|
|
142 | #define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("membar #LoadStore | #StoreStore") |
|
|
143 | #elif defined __s390__ || defined __s390x__ |
|
|
144 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("bcr 15,0" : : : "memory") |
|
|
145 | #elif defined __mips__ |
|
|
146 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("sync" : : : "memory") |
|
|
147 | #elif defined __alpha__ |
|
|
148 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("mb" : : : "memory") |
|
|
149 | #endif |
|
|
150 | #endif |
|
|
151 | #endif |
|
|
152 | |
|
|
153 | #ifndef ECB_MEMORY_FENCE |
|
|
154 | #if ECB_GCC_VERSION(4,7) |
|
|
155 | /* unsolved mystery: ACQ_REL should be enough, but doesn't generate any code */ |
|
|
156 | /* which in turn actually breaks libev */ |
|
|
157 | #define ECB_MEMORY_FENCE __atomic_thread_fence (__ATOMIC_SEQ_CST) |
|
|
158 | #elif defined __clang && __has_feature (cxx_atomic) |
|
|
159 | /* see above */ |
|
|
160 | #define ECB_MEMORY_FENCE __c11_atomic_thread_fence (__ATOMIC_SEQ_CST) |
|
|
161 | #elif ECB_GCC_VERSION(4,4) || defined __INTEL_COMPILER || defined __clang__ |
|
|
162 | #define ECB_MEMORY_FENCE __sync_synchronize () |
|
|
163 | /*#define ECB_MEMORY_FENCE_ACQUIRE ({ char dummy = 0; __sync_lock_test_and_set (&dummy, 1); }) */ |
|
|
164 | /*#define ECB_MEMORY_FENCE_RELEASE ({ char dummy = 1; __sync_lock_release (&dummy ); }) */ |
|
|
165 | #elif _MSC_VER >= 1400 /* VC++ 2005 */ |
|
|
166 | #pragma intrinsic(_ReadBarrier,_WriteBarrier,_ReadWriteBarrier) |
|
|
167 | #define ECB_MEMORY_FENCE _ReadWriteBarrier () |
|
|
168 | #define ECB_MEMORY_FENCE_ACQUIRE _ReadWriteBarrier () /* according to msdn, _ReadBarrier is not a load fence */ |
|
|
169 | #define ECB_MEMORY_FENCE_RELEASE _WriteBarrier () |
|
|
170 | #elif defined _WIN32 |
|
|
171 | #include <WinNT.h> |
|
|
172 | #define ECB_MEMORY_FENCE MemoryBarrier () /* actually just xchg on x86... scary */ |
|
|
173 | #elif __SUNPRO_C >= 0x5110 || __SUNPRO_CC >= 0x5110 |
|
|
174 | #include <mbarrier.h> |
|
|
175 | #define ECB_MEMORY_FENCE __machine_rw_barrier () |
|
|
176 | #define ECB_MEMORY_FENCE_ACQUIRE __machine_r_barrier () |
|
|
177 | #define ECB_MEMORY_FENCE_RELEASE __machine_w_barrier () |
|
|
178 | #elif __xlC__ |
|
|
179 | #define ECB_MEMORY_FENCE __sync () |
|
|
180 | #endif |
|
|
181 | #endif |
|
|
182 | |
|
|
183 | #ifndef ECB_MEMORY_FENCE |
|
|
184 | #if !ECB_AVOID_PTHREADS |
|
|
185 | /* |
|
|
186 | * if you get undefined symbol references to pthread_mutex_lock, |
|
|
187 | * or failure to find pthread.h, then you should implement |
|
|
188 | * the ECB_MEMORY_FENCE operations for your cpu/compiler |
|
|
189 | * OR provide pthread.h and link against the posix thread library |
|
|
190 | * of your system. |
|
|
191 | */ |
|
|
192 | #include <pthread.h> |
|
|
193 | #define ECB_NEEDS_PTHREADS 1 |
|
|
194 | #define ECB_MEMORY_FENCE_NEEDS_PTHREADS 1 |
|
|
195 | |
|
|
196 | static pthread_mutex_t ecb_mf_lock = PTHREAD_MUTEX_INITIALIZER; |
|
|
197 | #define ECB_MEMORY_FENCE do { pthread_mutex_lock (&ecb_mf_lock); pthread_mutex_unlock (&ecb_mf_lock); } while (0) |
|
|
198 | #endif |
|
|
199 | #endif |
|
|
200 | |
|
|
201 | #if !defined ECB_MEMORY_FENCE_ACQUIRE && defined ECB_MEMORY_FENCE |
|
|
202 | #define ECB_MEMORY_FENCE_ACQUIRE ECB_MEMORY_FENCE |
|
|
203 | #endif |
|
|
204 | |
|
|
205 | #if !defined ECB_MEMORY_FENCE_RELEASE && defined ECB_MEMORY_FENCE |
|
|
206 | #define ECB_MEMORY_FENCE_RELEASE ECB_MEMORY_FENCE |
|
|
207 | #endif |
|
|
208 | |
|
|
209 | /*****************************************************************************/ |
|
|
210 | |
49 | #ifndef __cplusplus |
211 | #if __cplusplus |
50 | # if __STDC_VERSION__ >= 199901L |
212 | #define ecb_inline static inline |
51 | # define ECB_INLINE inline |
213 | #elif ECB_GCC_VERSION(2,5) |
52 | typedef _Bool ecb_bool; |
214 | #define ecb_inline static __inline__ |
|
|
215 | #elif ECB_C99 |
|
|
216 | #define ecb_inline static inline |
53 | # else |
217 | #else |
54 | # define ECB_INLINE inline /* yeah! */ |
218 | #define ecb_inline static |
|
|
219 | #endif |
|
|
220 | |
|
|
221 | #if ECB_GCC_VERSION(3,3) |
|
|
222 | #define ecb_restrict __restrict__ |
|
|
223 | #elif ECB_C99 |
|
|
224 | #define ecb_restrict restrict |
|
|
225 | #else |
|
|
226 | #define ecb_restrict |
|
|
227 | #endif |
|
|
228 | |
55 | typedef int ecb_bool; |
229 | typedef int ecb_bool; |
56 | # endif |
|
|
57 | #endif |
|
|
58 | |
230 | |
59 | #define ECB_CONCAT_(a, b) a ## b |
231 | #define ECB_CONCAT_(a, b) a ## b |
60 | #define ECB_CONCAT(a, b) ECB_CONCAT_(a, b) |
232 | #define ECB_CONCAT(a, b) ECB_CONCAT_(a, b) |
61 | #define ECB_STRINGIFY_(a) # a |
233 | #define ECB_STRINGIFY_(a) # a |
62 | #define ECB_STRINGIFY(a) ECB_STRINGIFY_(a) |
234 | #define ECB_STRINGIFY(a) ECB_STRINGIFY_(a) |
63 | |
235 | |
64 | #define ECB_HEADER_INLINE static ECB_INLINE |
236 | #define ecb_function_ ecb_inline |
65 | |
237 | |
66 | #if ECB_GCC_VERSION(3,1) |
238 | #if ECB_GCC_VERSION(3,1) |
67 | # define ecb_attribute(attrlist) __attribute__(attrlist) |
239 | #define ecb_attribute(attrlist) __attribute__(attrlist) |
68 | # define ecb_is_constant(c) __builtin_constant_p (c) |
240 | #define ecb_is_constant(expr) __builtin_constant_p (expr) |
69 | # define ecb_expect(expr,value) __builtin_expect ((expr),(value)) |
241 | #define ecb_expect(expr,value) __builtin_expect ((expr),(value)) |
70 | # define ecb_prefetch(addr,rw,locality) __builtin_prefetch (addr, rw, locality) |
242 | #define ecb_prefetch(addr,rw,locality) __builtin_prefetch (addr, rw, locality) |
71 | #else |
243 | #else |
72 | # define ecb_attribute(attrlist) |
244 | #define ecb_attribute(attrlist) |
73 | # define ecb_is_constant(c) 0 |
245 | #define ecb_is_constant(expr) 0 |
74 | # define ecb_expect(expr,value) (expr) |
246 | #define ecb_expect(expr,value) (expr) |
75 | # define ecb_prefetch(addr,rw,locality) |
247 | #define ecb_prefetch(addr,rw,locality) |
76 | #endif |
248 | #endif |
77 | |
249 | |
78 | /* no emulation for ecb_decltype */ |
250 | /* no emulation for ecb_decltype */ |
79 | #if ECB_GCC_VERSION(4,5) |
251 | #if ECB_GCC_VERSION(4,5) |
80 | # define ecb_decltype(x) __decltype(x) |
252 | #define ecb_decltype(x) __decltype(x) |
81 | #elif ECB_GCC_VERSION(3,0) |
253 | #elif ECB_GCC_VERSION(3,0) |
82 | # define ecb_decltype(x) typeof(x) |
254 | #define ecb_decltype(x) __typeof(x) |
83 | #endif |
255 | #endif |
84 | |
256 | |
85 | #define ecb_noinline ecb_attribute ((noinline)) |
257 | #define ecb_noinline ecb_attribute ((__noinline__)) |
86 | #define ecb_noreturn ecb_attribute ((noreturn)) |
|
|
87 | #define ecb_unused ecb_attribute ((unused)) |
258 | #define ecb_unused ecb_attribute ((__unused__)) |
88 | #define ecb_const ecb_attribute ((const)) |
259 | #define ecb_const ecb_attribute ((__const__)) |
89 | #define ecb_pure ecb_attribute ((pure)) |
260 | #define ecb_pure ecb_attribute ((__pure__)) |
|
|
261 | |
|
|
262 | #if ECB_C11 |
|
|
263 | #define ecb_noreturn _Noreturn |
|
|
264 | #else |
|
|
265 | #define ecb_noreturn ecb_attribute ((__noreturn__)) |
|
|
266 | #endif |
|
|
267 | |
|
|
268 | #if ECB_GCC_VERSION(4,3) |
|
|
269 | #define ecb_artificial ecb_attribute ((__artificial__)) |
90 | #define ecb_hot ecb_attribute ((hot)) /* 4.3 */ |
270 | #define ecb_hot ecb_attribute ((__hot__)) |
91 | #define ecb_cold ecb_attribute ((cold)) /* 4.3 */ |
271 | #define ecb_cold ecb_attribute ((__cold__)) |
|
|
272 | #else |
|
|
273 | #define ecb_artificial |
|
|
274 | #define ecb_hot |
|
|
275 | #define ecb_cold |
|
|
276 | #endif |
92 | |
277 | |
93 | /* put into ifs if you are very sure that the expression */ |
278 | /* put around conditional expressions if you are very sure that the */ |
94 | /* is mostly true or mosty false. note that these return */ |
279 | /* expression is mostly true or mostly false. note that these return */ |
95 | /* booleans, not the expression. */ |
280 | /* booleans, not the expression. */ |
96 | #define ecb_unlikely(expr) ecb_expect (!!(expr), 0) |
281 | #define ecb_expect_false(expr) ecb_expect (!!(expr), 0) |
|
|
282 | #define ecb_expect_true(expr) ecb_expect (!!(expr), 1) |
|
|
283 | /* for compatibility to the rest of the world */ |
97 | #define ecb_likely(expr) ecb_expect (!!(expr), 1) |
284 | #define ecb_likely(expr) ecb_expect_true (expr) |
98 | |
285 | #define ecb_unlikely(expr) ecb_expect_false (expr) |
99 | /* try to tell the compiler that some condition is definitely true */ |
|
|
100 | #define ecb_assume(cond) do { if (!(cond)) unreachable (); } while (0) |
|
|
101 | |
286 | |
102 | /* count trailing zero bits and count # of one bits */ |
287 | /* count trailing zero bits and count # of one bits */ |
103 | #if ECB_GCC_VERSION(3,4) |
288 | #if ECB_GCC_VERSION(3,4) |
|
|
289 | /* we assume int == 32 bit, long == 32 or 64 bit and long long == 64 bit */ |
|
|
290 | #define ecb_ld32(x) (__builtin_clz (x) ^ 31) |
|
|
291 | #define ecb_ld64(x) (__builtin_clzll (x) ^ 63) |
104 | #define ecb_ctz32 (x) __builtin_ctz (x) |
292 | #define ecb_ctz32(x) __builtin_ctz (x) |
|
|
293 | #define ecb_ctz64(x) __builtin_ctzll (x) |
105 | #define ecb_popcount32 (x) __builtin_popcount (x) |
294 | #define ecb_popcount32(x) __builtin_popcount (x) |
|
|
295 | /* no popcountll */ |
106 | #else |
296 | #else |
107 | ECB_HEADER_INLINE int ecb_ctz32 (uint32_t x) ecb_const; |
297 | ecb_function_ int ecb_ctz32 (uint32_t x) ecb_const; |
108 | ECB_HEADER_INLINE |
298 | ecb_function_ int |
109 | ecb_ctz32 (uint32_t x) |
299 | ecb_ctz32 (uint32_t x) |
|
|
300 | { |
|
|
301 | int r = 0; |
|
|
302 | |
|
|
303 | x &= ~x + 1; /* this isolates the lowest bit */ |
|
|
304 | |
|
|
305 | #if ECB_branchless_on_i386 |
|
|
306 | r += !!(x & 0xaaaaaaaa) << 0; |
|
|
307 | r += !!(x & 0xcccccccc) << 1; |
|
|
308 | r += !!(x & 0xf0f0f0f0) << 2; |
|
|
309 | r += !!(x & 0xff00ff00) << 3; |
|
|
310 | r += !!(x & 0xffff0000) << 4; |
|
|
311 | #else |
|
|
312 | if (x & 0xaaaaaaaa) r += 1; |
|
|
313 | if (x & 0xcccccccc) r += 2; |
|
|
314 | if (x & 0xf0f0f0f0) r += 4; |
|
|
315 | if (x & 0xff00ff00) r += 8; |
|
|
316 | if (x & 0xffff0000) r += 16; |
|
|
317 | #endif |
|
|
318 | |
|
|
319 | return r; |
|
|
320 | } |
|
|
321 | |
|
|
322 | ecb_function_ int ecb_ctz64 (uint64_t x) ecb_const; |
|
|
323 | ecb_function_ int |
|
|
324 | ecb_ctz64 (uint64_t x) |
|
|
325 | { |
|
|
326 | int shift = x & 0xffffffffU ? 0 : 32; |
|
|
327 | return ecb_ctz32 (x >> shift) + shift; |
|
|
328 | } |
|
|
329 | |
|
|
330 | ecb_function_ int ecb_popcount32 (uint32_t x) ecb_const; |
|
|
331 | ecb_function_ int |
|
|
332 | ecb_popcount32 (uint32_t x) |
|
|
333 | { |
|
|
334 | x -= (x >> 1) & 0x55555555; |
|
|
335 | x = ((x >> 2) & 0x33333333) + (x & 0x33333333); |
|
|
336 | x = ((x >> 4) + x) & 0x0f0f0f0f; |
|
|
337 | x *= 0x01010101; |
|
|
338 | |
|
|
339 | return x >> 24; |
|
|
340 | } |
|
|
341 | |
|
|
342 | ecb_function_ int ecb_ld32 (uint32_t x) ecb_const; |
|
|
343 | ecb_function_ int ecb_ld32 (uint32_t x) |
|
|
344 | { |
|
|
345 | int r = 0; |
|
|
346 | |
|
|
347 | if (x >> 16) { x >>= 16; r += 16; } |
|
|
348 | if (x >> 8) { x >>= 8; r += 8; } |
|
|
349 | if (x >> 4) { x >>= 4; r += 4; } |
|
|
350 | if (x >> 2) { x >>= 2; r += 2; } |
|
|
351 | if (x >> 1) { r += 1; } |
|
|
352 | |
|
|
353 | return r; |
|
|
354 | } |
|
|
355 | |
|
|
356 | ecb_function_ int ecb_ld64 (uint64_t x) ecb_const; |
|
|
357 | ecb_function_ int ecb_ld64 (uint64_t x) |
|
|
358 | { |
|
|
359 | int r = 0; |
|
|
360 | |
|
|
361 | if (x >> 32) { x >>= 32; r += 32; } |
|
|
362 | |
|
|
363 | return r + ecb_ld32 (x); |
|
|
364 | } |
|
|
365 | #endif |
|
|
366 | |
|
|
367 | ecb_function_ ecb_bool ecb_is_pot32 (uint32_t x) ecb_const; |
|
|
368 | ecb_function_ ecb_bool ecb_is_pot32 (uint32_t x) { return !(x & (x - 1)); } |
|
|
369 | ecb_function_ ecb_bool ecb_is_pot64 (uint64_t x) ecb_const; |
|
|
370 | ecb_function_ ecb_bool ecb_is_pot64 (uint64_t x) { return !(x & (x - 1)); } |
|
|
371 | |
|
|
372 | ecb_function_ uint8_t ecb_bitrev8 (uint8_t x) ecb_const; |
|
|
373 | ecb_function_ uint8_t ecb_bitrev8 (uint8_t x) |
110 | { |
374 | { |
111 | int r = 0; |
375 | return ( (x * 0x0802U & 0x22110U) |
112 | |
376 | | (x * 0x8020U & 0x88440U)) * 0x10101U >> 16; |
113 | x &= -x; /* this isolates the lowest bit */ |
|
|
114 | |
|
|
115 | if (x & 0xaaaaaaaa) r += 1; |
|
|
116 | if (x & 0xcccccccc) r += 2; |
|
|
117 | if (x & 0xf0f0f0f0) r += 4; |
|
|
118 | if (x & 0xff00ff00) r += 8; |
|
|
119 | if (x & 0xffff0000) r += 16; |
|
|
120 | |
|
|
121 | return r; |
|
|
122 | } |
377 | } |
123 | |
378 | |
124 | ECB_HEADER_INLINE int ecb_popcount32 (uint32_t x) ecb_const; |
379 | ecb_function_ uint16_t ecb_bitrev16 (uint16_t x) ecb_const; |
125 | ECB_HEADER_INLINE |
380 | ecb_function_ uint16_t ecb_bitrev16 (uint16_t x) |
126 | ecb_popcount32 (uint32_t x) |
|
|
127 | { |
381 | { |
128 | x -= (x >> 1) & 0x55555555; |
382 | x = ((x >> 1) & 0x5555) | ((x & 0x5555) << 1); |
129 | x = ((x >> 2) & 0x33333333) + (x & 0x33333333); |
383 | x = ((x >> 2) & 0x3333) | ((x & 0x3333) << 2); |
130 | x = ((x >> 4) + x) & 0x0f0f0f0f; |
384 | x = ((x >> 4) & 0x0f0f) | ((x & 0x0f0f) << 4); |
131 | x *= 0x01010101; |
385 | x = ( x >> 8 ) | ( x << 8); |
132 | |
386 | |
133 | return x >> 24; |
387 | return x; |
134 | } |
388 | } |
135 | #endif |
389 | |
|
|
390 | ecb_function_ uint32_t ecb_bitrev32 (uint32_t x) ecb_const; |
|
|
391 | ecb_function_ uint32_t ecb_bitrev32 (uint32_t x) |
|
|
392 | { |
|
|
393 | x = ((x >> 1) & 0x55555555) | ((x & 0x55555555) << 1); |
|
|
394 | x = ((x >> 2) & 0x33333333) | ((x & 0x33333333) << 2); |
|
|
395 | x = ((x >> 4) & 0x0f0f0f0f) | ((x & 0x0f0f0f0f) << 4); |
|
|
396 | x = ((x >> 8) & 0x00ff00ff) | ((x & 0x00ff00ff) << 8); |
|
|
397 | x = ( x >> 16 ) | ( x << 16); |
|
|
398 | |
|
|
399 | return x; |
|
|
400 | } |
|
|
401 | |
|
|
402 | /* popcount64 is only available on 64 bit cpus as gcc builtin */ |
|
|
403 | /* so for this version we are lazy */ |
|
|
404 | ecb_function_ int ecb_popcount64 (uint64_t x) ecb_const; |
|
|
405 | ecb_function_ int |
|
|
406 | ecb_popcount64 (uint64_t x) |
|
|
407 | { |
|
|
408 | return ecb_popcount32 (x) + ecb_popcount32 (x >> 32); |
|
|
409 | } |
|
|
410 | |
|
|
411 | ecb_inline uint8_t ecb_rotl8 (uint8_t x, unsigned int count) ecb_const; |
|
|
412 | ecb_inline uint8_t ecb_rotr8 (uint8_t x, unsigned int count) ecb_const; |
|
|
413 | ecb_inline uint16_t ecb_rotl16 (uint16_t x, unsigned int count) ecb_const; |
|
|
414 | ecb_inline uint16_t ecb_rotr16 (uint16_t x, unsigned int count) ecb_const; |
|
|
415 | ecb_inline uint32_t ecb_rotl32 (uint32_t x, unsigned int count) ecb_const; |
|
|
416 | ecb_inline uint32_t ecb_rotr32 (uint32_t x, unsigned int count) ecb_const; |
|
|
417 | ecb_inline uint64_t ecb_rotl64 (uint64_t x, unsigned int count) ecb_const; |
|
|
418 | ecb_inline uint64_t ecb_rotr64 (uint64_t x, unsigned int count) ecb_const; |
|
|
419 | |
|
|
420 | ecb_inline uint8_t ecb_rotl8 (uint8_t x, unsigned int count) { return (x >> ( 8 - count)) | (x << count); } |
|
|
421 | ecb_inline uint8_t ecb_rotr8 (uint8_t x, unsigned int count) { return (x << ( 8 - count)) | (x >> count); } |
|
|
422 | ecb_inline uint16_t ecb_rotl16 (uint16_t x, unsigned int count) { return (x >> (16 - count)) | (x << count); } |
|
|
423 | ecb_inline uint16_t ecb_rotr16 (uint16_t x, unsigned int count) { return (x << (16 - count)) | (x >> count); } |
|
|
424 | ecb_inline uint32_t ecb_rotl32 (uint32_t x, unsigned int count) { return (x >> (32 - count)) | (x << count); } |
|
|
425 | ecb_inline uint32_t ecb_rotr32 (uint32_t x, unsigned int count) { return (x << (32 - count)) | (x >> count); } |
|
|
426 | ecb_inline uint64_t ecb_rotl64 (uint64_t x, unsigned int count) { return (x >> (64 - count)) | (x << count); } |
|
|
427 | ecb_inline uint64_t ecb_rotr64 (uint64_t x, unsigned int count) { return (x << (64 - count)) | (x >> count); } |
136 | |
428 | |
137 | #if ECB_GCC_VERSION(4,3) |
429 | #if ECB_GCC_VERSION(4,3) |
|
|
430 | #define ecb_bswap16(x) (__builtin_bswap32 (x) >> 16) |
138 | # define ecb_bswap32(x) __builtin_bswap32 (x) |
431 | #define ecb_bswap32(x) __builtin_bswap32 (x) |
|
|
432 | #define ecb_bswap64(x) __builtin_bswap64 (x) |
139 | #else |
433 | #else |
|
|
434 | ecb_function_ uint16_t ecb_bswap16 (uint16_t x) ecb_const; |
|
|
435 | ecb_function_ uint16_t |
|
|
436 | ecb_bswap16 (uint16_t x) |
|
|
437 | { |
|
|
438 | return ecb_rotl16 (x, 8); |
|
|
439 | } |
|
|
440 | |
140 | ECB_HEADER_INLINE uint32_t ecb_bswap32 (uint32_t x) ecb_const; |
441 | ecb_function_ uint32_t ecb_bswap32 (uint32_t x) ecb_const; |
141 | ECB_HEADER_INLINE uint32_t |
442 | ecb_function_ uint32_t |
142 | ecb_bswap32 (uint32_t x) |
443 | ecb_bswap32 (uint32_t x) |
143 | { |
444 | { |
144 | return (x >> 24) |
445 | return (((uint32_t)ecb_bswap16 (x)) << 16) | ecb_bswap16 (x >> 16); |
145 | | ((x >> 8) & 0x0000ff00) |
446 | } |
146 | | ((x << 8) & 0x00ff0000) |
447 | |
147 | | (x << 24); |
448 | ecb_function_ uint64_t ecb_bswap64 (uint64_t x) ecb_const; |
148 | } |
449 | ecb_function_ uint64_t |
|
|
450 | ecb_bswap64 (uint64_t x) |
|
|
451 | { |
|
|
452 | return (((uint64_t)ecb_bswap32 (x)) << 32) | ecb_bswap32 (x >> 32); |
|
|
453 | } |
149 | #endif |
454 | #endif |
150 | |
455 | |
151 | #if ECB_GCC_VERSION(4,5) |
456 | #if ECB_GCC_VERSION(4,5) |
152 | # define ecb_unreachable() __builtin_unreachable () |
457 | #define ecb_unreachable() __builtin_unreachable () |
153 | #else |
458 | #else |
154 | /* this seems to work fine, but gcc always emits a warning for it :/ */ |
459 | /* this seems to work fine, but gcc always emits a warning for it :/ */ |
155 | ECB_HEADER_INLINE void ecb_unreachable (void) ecb_attribute ((noreturn)); |
460 | ecb_inline void ecb_unreachable (void) ecb_noreturn; |
156 | ECB_HEADER_INLINE void ecb_unreachable (void) { } |
461 | ecb_inline void ecb_unreachable (void) { } |
157 | #endif |
462 | #endif |
158 | |
463 | |
|
|
464 | /* try to tell the compiler that some condition is definitely true */ |
|
|
465 | #define ecb_assume(cond) do { if (!(cond)) ecb_unreachable (); } while (0) |
|
|
466 | |
159 | ECB_HEADER_INLINE unsigned char ecb_byteorder_helper () ecb_const; |
467 | ecb_inline unsigned char ecb_byteorder_helper (void) ecb_const; |
160 | ECB_HEADER_INLINE unsigned char |
468 | ecb_inline unsigned char |
161 | ecb_byteorder_helper () |
469 | ecb_byteorder_helper (void) |
162 | { |
470 | { |
163 | const uint32_t u = 0x11223344; |
471 | const uint32_t u = 0x11223344; |
164 | return *(unsigned char *)&u; |
472 | return *(unsigned char *)&u; |
165 | } |
473 | } |
166 | |
474 | |
167 | ECB_HEADER_INLINE ecb_bool ecb_big_endian () ecb_const; |
475 | ecb_inline ecb_bool ecb_big_endian (void) ecb_const; |
168 | ECB_HEADER_INLINE ecb_bool ecb_big_endian () { return ecb_byteorder_helper () == 0x11; }; |
476 | ecb_inline ecb_bool ecb_big_endian (void) { return ecb_byteorder_helper () == 0x11; } |
169 | ECB_HEADER_INLINE ecb_bool ecb_little_endian () ecb_const; |
477 | ecb_inline ecb_bool ecb_little_endian (void) ecb_const; |
170 | ECB_HEADER_INLINE ecb_bool ecb_little_endian () { return ecb_byteorder_helper () == 0x44; }; |
478 | ecb_inline ecb_bool ecb_little_endian (void) { return ecb_byteorder_helper () == 0x44; } |
|
|
479 | |
|
|
480 | #if ECB_GCC_VERSION(3,0) || ECB_C99 |
|
|
481 | #define ecb_mod(m,n) ((m) % (n) + ((m) % (n) < 0 ? (n) : 0)) |
|
|
482 | #else |
|
|
483 | #define ecb_mod(m,n) ((m) < 0 ? ((n) - 1 - ((-1 - (m)) % (n))) : ((m) % (n))) |
|
|
484 | #endif |
|
|
485 | |
|
|
486 | #if __cplusplus |
|
|
487 | template<typename T> |
|
|
488 | static inline T ecb_div_rd (T val, T div) |
|
|
489 | { |
|
|
490 | return val < 0 ? - ((-val + div - 1) / div) : (val ) / div; |
|
|
491 | } |
|
|
492 | template<typename T> |
|
|
493 | static inline T ecb_div_ru (T val, T div) |
|
|
494 | { |
|
|
495 | return val < 0 ? - ((-val ) / div) : (val + div - 1) / div; |
|
|
496 | } |
|
|
497 | #else |
|
|
498 | #define ecb_div_rd(val,div) ((val) < 0 ? - ((-(val) + (div) - 1) / (div)) : ((val) ) / (div)) |
|
|
499 | #define ecb_div_ru(val,div) ((val) < 0 ? - ((-(val) ) / (div)) : ((val) + (div) - 1) / (div)) |
|
|
500 | #endif |
171 | |
501 | |
172 | #if ecb_cplusplus_does_not_suck |
502 | #if ecb_cplusplus_does_not_suck |
173 | // does not work for local types (http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2657.htm) |
503 | /* does not work for local types (http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2657.htm) */ |
174 | template<typename T, int N> |
504 | template<typename T, int N> |
175 | static inline int ecb_array_length (const T (&arr)[N]) |
505 | static inline int ecb_array_length (const T (&arr)[N]) |
176 | { |
506 | { |
177 | return N; |
507 | return N; |
178 | } |
508 | } |
179 | #else |
509 | #else |
180 | #define ecb_array_length(name) (sizeof (name) / sizeof (name [0])) |
510 | #define ecb_array_length(name) (sizeof (name) / sizeof (name [0])) |
181 | #endif |
511 | #endif |
182 | |
512 | |
183 | ECB_INLINE uint32_t ecb_rotr32 (uint32_t x, unsigned int count) ecb_const; |
|
|
184 | ECB_INLINE uint32_t |
|
|
185 | ecb_rotr32 (uint32_t x, unsigned int count) |
|
|
186 | { |
|
|
187 | return (x << (32 - count)) | (x >> count); |
|
|
188 | } |
|
|
189 | |
|
|
190 | ECB_INLINE uint32_t ecb_rotl32 (uint32_t x, unsigned int count) ecb_const; |
|
|
191 | ECB_INLINE uint32_t |
|
|
192 | ecb_rotl32 (uint32_t x, unsigned int count) |
|
|
193 | { |
|
|
194 | return (x >> (32 - count)) | (x << count); |
|
|
195 | } |
|
|
196 | |
|
|
197 | #endif |
513 | #endif |
198 | |
514 | |