1 /*
2 * libecb - http://software.schmorp.de/pkg/libecb
3 *
4 * Copyright (©) 2009-2011 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 || 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
141 #define ECB_C99 (__STDC_VERSION__ >= 199901L)
142
143 #if __cplusplus
144 #define ecb_inline static inline
145 #elif ECB_GCC_VERSION(2,5)
146 #define ecb_inline static __inline__
147 #elif ECB_C99
148 #define ecb_inline static inline
149 #else
150 #define ecb_inline static
151 #endif
152
153 #if ECB_GCC_VERSION(3,3)
154 #define ecb_restrict __restrict__
155 #elif ECB_C99
156 #define ecb_restrict restrict
157 #else
158 #define ecb_restrict
159 #endif
160
161 typedef int ecb_bool;
162
163 #define ECB_CONCAT_(a, b) a ## b
164 #define ECB_CONCAT(a, b) ECB_CONCAT_(a, b)
165 #define ECB_STRINGIFY_(a) # a
166 #define ECB_STRINGIFY(a) ECB_STRINGIFY_(a)
167
168 #define ecb_function_ ecb_inline
169
170 #if ECB_GCC_VERSION(3,1)
171 #define ecb_attribute(attrlist) __attribute__(attrlist)
172 #define ecb_is_constant(expr) __builtin_constant_p (expr)
173 #define ecb_expect(expr,value) __builtin_expect ((expr),(value))
174 #define ecb_prefetch(addr,rw,locality) __builtin_prefetch (addr, rw, locality)
175 #else
176 #define ecb_attribute(attrlist)
177 #define ecb_is_constant(expr) 0
178 #define ecb_expect(expr,value) (expr)
179 #define ecb_prefetch(addr,rw,locality)
180 #endif
181
182 /* no emulation for ecb_decltype */
183 #if ECB_GCC_VERSION(4,5)
184 #define ecb_decltype(x) __decltype(x)
185 #elif ECB_GCC_VERSION(3,0)
186 #define ecb_decltype(x) __typeof(x)
187 #endif
188
189 #define ecb_noinline ecb_attribute ((__noinline__))
190 #define ecb_noreturn ecb_attribute ((__noreturn__))
191 #define ecb_unused ecb_attribute ((__unused__))
192 #define ecb_const ecb_attribute ((__const__))
193 #define ecb_pure ecb_attribute ((__pure__))
194
195 #if ECB_GCC_VERSION(4,3)
196 #define ecb_artificial ecb_attribute ((__artificial__))
197 #define ecb_hot ecb_attribute ((__hot__))
198 #define ecb_cold ecb_attribute ((__cold__))
199 #else
200 #define ecb_artificial
201 #define ecb_hot
202 #define ecb_cold
203 #endif
204
205 /* put around conditional expressions if you are very sure that the */
206 /* expression is mostly true or mostly false. note that these return */
207 /* booleans, not the expression. */
208 #define ecb_expect_false(expr) ecb_expect (!!(expr), 0)
209 #define ecb_expect_true(expr) ecb_expect (!!(expr), 1)
210 /* for compatibility to the rest of the world */
211 #define ecb_likely(expr) ecb_expect_true (expr)
212 #define ecb_unlikely(expr) ecb_expect_false (expr)
213
214 /* count trailing zero bits and count # of one bits */
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)
219 #define ecb_ctz32(x) __builtin_ctz (x)
220 #define ecb_ctz64(x) __builtin_ctzll (x)
221 #define ecb_popcount32(x) __builtin_popcount (x)
222 /* no popcountll */
223 #else
224 ecb_function_ int ecb_ctz32 (uint32_t x) ecb_const;
225 ecb_function_ int
ecb_ctz32(uint32_t x)226 ecb_ctz32 (uint32_t x)
227 {
228 int r = 0;
229
230 x &= ~x + 1; /* this isolates the lowest bit */
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
239 if (x & 0xaaaaaaaa) r += 1;
240 if (x & 0xcccccccc) r += 2;
241 if (x & 0xf0f0f0f0) r += 4;
242 if (x & 0xff00ff00) r += 8;
243 if (x & 0xffff0000) r += 16;
244 #endif
245
246 return r;
247 }
248
249 ecb_function_ int ecb_ctz64 (uint64_t x) ecb_const;
250 ecb_function_ int
ecb_ctz64(uint64_t x)251 ecb_ctz64 (uint64_t x)
252 {
253 int shift = x & 0xffffffffU ? 0 : 32;
254 return ecb_ctz32 (x >> shift) + shift;
255 }
256
257 ecb_function_ int ecb_popcount32 (uint32_t x) ecb_const;
258 ecb_function_ int
ecb_popcount32(uint32_t x)259 ecb_popcount32 (uint32_t x)
260 {
261 x -= (x >> 1) & 0x55555555;
262 x = ((x >> 2) & 0x33333333) + (x & 0x33333333);
263 x = ((x >> 4) + x) & 0x0f0f0f0f;
264 x *= 0x01010101;
265
266 return x >> 24;
267 }
268
269 ecb_function_ int ecb_ld32 (uint32_t x) ecb_const;
ecb_ld32(uint32_t x)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;
ecb_ld64(uint64_t x)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 }
292 #endif
293
294 /* popcount64 is only available on 64 bit cpus as gcc builtin */
295 /* so for this version we are lazy */
296 ecb_function_ int ecb_popcount64 (uint64_t x) ecb_const;
297 ecb_function_ int
ecb_popcount64(uint64_t x)298 ecb_popcount64 (uint64_t x)
299 {
300 return ecb_popcount32 (x) + ecb_popcount32 (x >> 32);
301 }
302
303 ecb_inline uint8_t ecb_rotl8 (uint8_t x, unsigned int count) ecb_const;
304 ecb_inline uint8_t ecb_rotr8 (uint8_t x, unsigned int count) ecb_const;
305 ecb_inline uint16_t ecb_rotl16 (uint16_t x, unsigned int count) ecb_const;
306 ecb_inline uint16_t ecb_rotr16 (uint16_t x, unsigned int count) ecb_const;
307 ecb_inline uint32_t ecb_rotl32 (uint32_t x, unsigned int count) ecb_const;
308 ecb_inline uint32_t ecb_rotr32 (uint32_t x, unsigned int count) ecb_const;
309 ecb_inline uint64_t ecb_rotl64 (uint64_t x, unsigned int count) ecb_const;
310 ecb_inline uint64_t ecb_rotr64 (uint64_t x, unsigned int count) ecb_const;
311
ecb_rotl8(uint8_t x,unsigned int count)312 ecb_inline uint8_t ecb_rotl8 (uint8_t x, unsigned int count) { return (x >> ( 8 - count)) | (x << count); }
ecb_rotr8(uint8_t x,unsigned int count)313 ecb_inline uint8_t ecb_rotr8 (uint8_t x, unsigned int count) { return (x << ( 8 - count)) | (x >> count); }
ecb_rotl16(uint16_t x,unsigned int count)314 ecb_inline uint16_t ecb_rotl16 (uint16_t x, unsigned int count) { return (x >> (16 - count)) | (x << count); }
ecb_rotr16(uint16_t x,unsigned int count)315 ecb_inline uint16_t ecb_rotr16 (uint16_t x, unsigned int count) { return (x << (16 - count)) | (x >> count); }
ecb_rotl32(uint32_t x,unsigned int count)316 ecb_inline uint32_t ecb_rotl32 (uint32_t x, unsigned int count) { return (x >> (32 - count)) | (x << count); }
ecb_rotr32(uint32_t x,unsigned int count)317 ecb_inline uint32_t ecb_rotr32 (uint32_t x, unsigned int count) { return (x << (32 - count)) | (x >> count); }
ecb_rotl64(uint64_t x,unsigned int count)318 ecb_inline uint64_t ecb_rotl64 (uint64_t x, unsigned int count) { return (x >> (64 - count)) | (x << count); }
ecb_rotr64(uint64_t x,unsigned int count)319 ecb_inline uint64_t ecb_rotr64 (uint64_t x, unsigned int count) { return (x << (64 - count)) | (x >> count); }
320
321 #if ECB_GCC_VERSION(4,3)
322 #define ecb_bswap16(x) (__builtin_bswap32 (x) >> 16)
323 #define ecb_bswap32(x) __builtin_bswap32 (x)
324 #define ecb_bswap64(x) __builtin_bswap64 (x)
325 #else
326 ecb_function_ uint16_t ecb_bswap16 (uint16_t x) ecb_const;
327 ecb_function_ uint16_t
ecb_bswap16(uint16_t x)328 ecb_bswap16 (uint16_t x)
329 {
330 return ecb_rotl16 (x, 8);
331 }
332
333 ecb_function_ uint32_t ecb_bswap32 (uint32_t x) ecb_const;
334 ecb_function_ uint32_t
ecb_bswap32(uint32_t x)335 ecb_bswap32 (uint32_t x)
336 {
337 return (((uint32_t)ecb_bswap16 (x)) << 16) | ecb_bswap16 (x >> 16);
338 }
339
340 ecb_function_ uint64_t ecb_bswap64 (uint64_t x) ecb_const;
341 ecb_function_ uint64_t
ecb_bswap64(uint64_t x)342 ecb_bswap64 (uint64_t x)
343 {
344 return (((uint64_t)ecb_bswap32 (x)) << 32) | ecb_bswap32 (x >> 32);
345 }
346 #endif
347
348 #if ECB_GCC_VERSION(4,5)
349 #define ecb_unreachable() __builtin_unreachable ()
350 #else
351 /* this seems to work fine, but gcc always emits a warning for it :/ */
352 ecb_function_ void ecb_unreachable (void) ecb_noreturn;
ecb_unreachable(void)353 ecb_function_ void ecb_unreachable (void) { }
354 #endif
355
356 /* try to tell the compiler that some condition is definitely true */
357 #define ecb_assume(cond) do { if (!(cond)) ecb_unreachable (); } while (0)
358
359 ecb_function_ unsigned char ecb_byteorder_helper (void) ecb_const;
360 ecb_function_ unsigned char
ecb_byteorder_helper(void)361 ecb_byteorder_helper (void)
362 {
363 const uint32_t u = 0x11223344;
364 return *(unsigned char *)&u;
365 }
366
367 ecb_function_ ecb_bool ecb_big_endian (void) ecb_const;
ecb_big_endian(void)368 ecb_function_ ecb_bool ecb_big_endian (void) { return ecb_byteorder_helper () == 0x11; }
369 ecb_function_ ecb_bool ecb_little_endian (void) ecb_const;
ecb_little_endian(void)370 ecb_function_ ecb_bool ecb_little_endian (void) { return ecb_byteorder_helper () == 0x44; }
371
372 #if ECB_GCC_VERSION(3,0) || ECB_C99
373 #define ecb_mod(m,n) ((m) % (n) + ((m) % (n) < 0 ? (n) : 0))
374 #else
375 #define ecb_mod(m,n) ((m) < 0 ? ((n) - 1 - ((-1 - (m)) % (n))) : ((m) % (n)))
376 #endif
377
378 #if __cplusplus
379 template<typename T>
ecb_div_rd(T val,T div)380 static inline T ecb_div_rd (T val, T div)
381 {
382 return val < 0 ? - ((-val + div - 1) / div) : (val ) / div;
383 }
384 template<typename T>
ecb_div_ru(T val,T div)385 static inline T ecb_div_ru (T val, T div)
386 {
387 return val < 0 ? - ((-val ) / div) : (val + div - 1) / div;
388 }
389 #else
390 #define ecb_div_rd(val,div) ((val) < 0 ? - ((-(val) + (div) - 1) / (div)) : ((val) ) / (div))
391 #define ecb_div_ru(val,div) ((val) < 0 ? - ((-(val) ) / (div)) : ((val) + (div) - 1) / (div))
392 #endif
393
394 #if ecb_cplusplus_does_not_suck
395 /* does not work for local types (http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2657.htm) */
396 template<typename T, int N>
ecb_array_length(const T (& arr)[N])397 static inline int ecb_array_length (const T (&arr)[N])
398 {
399 return N;
400 }
401 #else
402 #define ecb_array_length(name) (sizeof (name) / sizeof (name [0]))
403 #endif
404
405 #endif
406
407