ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/libecb/ecb.h
Revision: 1.140
Committed: Tue Oct 14 22:14:32 2014 UTC (9 years, 8 months ago) by sf-exg
Content type: text/plain
Branch: MAIN
Changes since 1.139: +11 -10 lines
Log Message:
Add support for C11 atomic on clang.

File Contents

# Content
1 /*
2 * libecb - http://software.schmorp.de/pkg/libecb
3 *
4 * Copyright (©) 2009-2014 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 * Alternatively, the contents of this file may be used under the terms of
30 * the GNU General Public License ("GPL") version 2 or any later version,
31 * in which case the provisions of the GPL are applicable instead of
32 * the above. If you wish to allow the use of your version of this file
33 * only under the terms of the GPL and not to allow others to use your
34 * version of this file under the BSD license, indicate your decision
35 * by deleting the provisions above and replace them with the notice
36 * and other provisions required by the GPL. If you do not delete the
37 * provisions above, a recipient may use your version of this file under
38 * either the BSD or the GPL.
39 */
40
41 #ifndef ECB_H
42 #define ECB_H
43
44 /* 16 bits major, 16 bits minor */
45 #define ECB_VERSION 0x00010003
46
47 #ifdef _WIN32
48 typedef signed char int8_t;
49 typedef unsigned char uint8_t;
50 typedef signed short int16_t;
51 typedef unsigned short uint16_t;
52 typedef signed int int32_t;
53 typedef unsigned int uint32_t;
54 #if __GNUC__
55 typedef signed long long int64_t;
56 typedef unsigned long long uint64_t;
57 #else /* _MSC_VER || __BORLANDC__ */
58 typedef signed __int64 int64_t;
59 typedef unsigned __int64 uint64_t;
60 #endif
61 #ifdef _WIN64
62 #define ECB_PTRSIZE 8
63 typedef uint64_t uintptr_t;
64 typedef int64_t intptr_t;
65 #else
66 #define ECB_PTRSIZE 4
67 typedef uint32_t uintptr_t;
68 typedef int32_t intptr_t;
69 #endif
70 #else
71 #include <inttypes.h>
72 #if UINTMAX_MAX > 0xffffffffU
73 #define ECB_PTRSIZE 8
74 #else
75 #define ECB_PTRSIZE 4
76 #endif
77 #endif
78
79 /* work around x32 idiocy by defining proper macros */
80 #if __amd64 || __x86_64 || _M_AMD64 || _M_X64
81 #if _ILP32
82 #define ECB_AMD64_X32 1
83 #else
84 #define ECB_AMD64 1
85 #endif
86 #endif
87
88 /* many compilers define _GNUC_ to some versions but then only implement
89 * what their idiot authors think are the "more important" extensions,
90 * causing enormous grief in return for some better fake benchmark numbers.
91 * or so.
92 * we try to detect these and simply assume they are not gcc - if they have
93 * an issue with that they should have done it right in the first place.
94 */
95 #if !defined __GNUC_MINOR__ || defined __INTEL_COMPILER || defined __SUNPRO_C || defined __SUNPRO_CC || defined __llvm__ || defined __clang__
96 #define ECB_GCC_VERSION(major,minor) 0
97 #else
98 #define ECB_GCC_VERSION(major,minor) (__GNUC__ > (major) || (__GNUC__ == (major) && __GNUC_MINOR__ >= (minor)))
99 #endif
100
101 #define ECB_CLANG_VERSION(major,minor) (__clang_major__ > (major) || (__clang_major__ == (major) && __clang_minor__ >= (minor)))
102
103 #if __clang__ && defined(__has_builtin)
104 #define ECB_CLANG_BUILTIN(x) __has_builtin(x)
105 #else
106 #define ECB_CLANG_BUILTIN(x) 0
107 #endif
108
109 #if __clang__ && defined(__has_extension)
110 #define ECB_CLANG_EXTENSION(x) __has_extension(x)
111 #else
112 #define ECB_CLANG_EXTENSION(x) 0
113 #endif
114
115 #define ECB_CPP (__cplusplus+0)
116 #define ECB_CPP11 (__cplusplus >= 201103L)
117
118 #if ECB_CPP
119 #define ECB_C 0
120 #define ECB_STDC_VERSION 0
121 #else
122 #define ECB_C 1
123 #define ECB_STDC_VERSION __STDC_VERSION__
124 #endif
125
126 #define ECB_C99 (ECB_STDC_VERSION >= 199901L)
127 #define ECB_C11 (ECB_STDC_VERSION >= 201112L)
128
129 #if ECB_CPP
130 #define ECB_EXTERN_C extern "C"
131 #define ECB_EXTERN_C_BEG ECB_EXTERN_C {
132 #define ECB_EXTERN_C_END }
133 #else
134 #define ECB_EXTERN_C extern
135 #define ECB_EXTERN_C_BEG
136 #define ECB_EXTERN_C_END
137 #endif
138
139 /*****************************************************************************/
140
141 /* ECB_NO_THREADS - ecb is not used by multiple threads, ever */
142 /* ECB_NO_SMP - ecb might be used in multiple threads, but only on a single cpu */
143
144 #if ECB_NO_THREADS
145 #define ECB_NO_SMP 1
146 #endif
147
148 #if ECB_NO_SMP
149 #define ECB_MEMORY_FENCE do { } while (0)
150 #endif
151
152 #ifndef ECB_MEMORY_FENCE
153 #if ECB_GCC_VERSION(2,5) || defined __INTEL_COMPILER || (__llvm__ && __GNUC__) || __SUNPRO_C >= 0x5110 || __SUNPRO_CC >= 0x5110
154 #if __i386 || __i386__
155 #define ECB_MEMORY_FENCE __asm__ __volatile__ ("lock; orb $0, -1(%%esp)" : : : "memory")
156 #define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("" : : : "memory")
157 #define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("")
158 #elif __amd64 || __amd64__ || __x86_64 || __x86_64__
159 #define ECB_MEMORY_FENCE __asm__ __volatile__ ("mfence" : : : "memory")
160 #define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("" : : : "memory")
161 #define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("")
162 #elif __powerpc__ || __ppc__ || __powerpc64__ || __ppc64__
163 #define ECB_MEMORY_FENCE __asm__ __volatile__ ("sync" : : : "memory")
164 #elif defined __ARM_ARCH_6__ || defined __ARM_ARCH_6J__ \
165 || defined __ARM_ARCH_6K__ || defined __ARM_ARCH_6ZK__
166 #define ECB_MEMORY_FENCE __asm__ __volatile__ ("mcr p15,0,%0,c7,c10,5" : : "r" (0) : "memory")
167 #elif defined __ARM_ARCH_7__ || defined __ARM_ARCH_7A__ \
168 || defined __ARM_ARCH_7M__ || defined __ARM_ARCH_7R__
169 #define ECB_MEMORY_FENCE __asm__ __volatile__ ("dmb" : : : "memory")
170 #elif __aarch64__
171 #define ECB_MEMORY_FENCE __asm__ __volatile__ ("dmb ish" : : : "memory")
172 #elif (__sparc || __sparc__) && !__sparcv8
173 #define ECB_MEMORY_FENCE __asm__ __volatile__ ("membar #LoadStore | #LoadLoad | #StoreStore | #StoreLoad" : : : "memory")
174 #define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("membar #LoadStore | #LoadLoad" : : : "memory")
175 #define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("membar #LoadStore | #StoreStore")
176 #elif defined __s390__ || defined __s390x__
177 #define ECB_MEMORY_FENCE __asm__ __volatile__ ("bcr 15,0" : : : "memory")
178 #elif defined __mips__
179 /* GNU/Linux emulates sync on mips1 architectures, so we force its use */
180 /* anybody else who still uses mips1 is supposed to send in their version, with detection code. */
181 #define ECB_MEMORY_FENCE __asm__ __volatile__ (".set mips2; sync; .set mips0" : : : "memory")
182 #elif defined __alpha__
183 #define ECB_MEMORY_FENCE __asm__ __volatile__ ("mb" : : : "memory")
184 #elif defined __hppa__
185 #define ECB_MEMORY_FENCE __asm__ __volatile__ ("" : : : "memory")
186 #define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("")
187 #elif defined __ia64__
188 #define ECB_MEMORY_FENCE __asm__ __volatile__ ("mf" : : : "memory")
189 #elif defined __m68k__
190 #define ECB_MEMORY_FENCE __asm__ __volatile__ ("" : : : "memory")
191 #elif defined __m88k__
192 #define ECB_MEMORY_FENCE __asm__ __volatile__ ("tb1 0,%%r0,128" : : : "memory")
193 #elif defined __sh__
194 #define ECB_MEMORY_FENCE __asm__ __volatile__ ("" : : : "memory")
195 #endif
196 #endif
197 #endif
198
199 #ifndef ECB_MEMORY_FENCE
200 #if ECB_GCC_VERSION(4,7)
201 /* see comment below (stdatomic.h) about the C11 memory model. */
202 #define ECB_MEMORY_FENCE __atomic_thread_fence (__ATOMIC_SEQ_CST)
203 #define ECB_MEMORY_FENCE_ACQUIRE __atomic_thread_fence (__ATOMIC_ACQUIRE)
204 #define ECB_MEMORY_FENCE_RELEASE __atomic_thread_fence (__ATOMIC_RELEASE)
205
206 #elif ECB_CLANG_EXTENSION(c_atomic)
207 /* see comment below (stdatomic.h) about the C11 memory model. */
208 #define ECB_MEMORY_FENCE __c11_atomic_thread_fence (__ATOMIC_SEQ_CST)
209 #define ECB_MEMORY_FENCE_ACQUIRE __c11_atomic_thread_fence (__ATOMIC_ACQUIRE)
210 #define ECB_MEMORY_FENCE_RELEASE __c11_atomic_thread_fence (__ATOMIC_RELEASE)
211
212 #elif ECB_GCC_VERSION(4,4) || defined __INTEL_COMPILER || defined __clang__
213 #define ECB_MEMORY_FENCE __sync_synchronize ()
214 #elif _MSC_VER >= 1500 /* VC++ 2008 */
215 /* apparently, microsoft broke all the memory barrier stuff in Visual Studio 2008... */
216 #pragma intrinsic(_ReadBarrier,_WriteBarrier,_ReadWriteBarrier)
217 #define ECB_MEMORY_FENCE _ReadWriteBarrier (); MemoryBarrier()
218 #define ECB_MEMORY_FENCE_ACQUIRE _ReadWriteBarrier (); MemoryBarrier() /* according to msdn, _ReadBarrier is not a load fence */
219 #define ECB_MEMORY_FENCE_RELEASE _WriteBarrier (); MemoryBarrier()
220 #elif _MSC_VER >= 1400 /* VC++ 2005 */
221 #pragma intrinsic(_ReadBarrier,_WriteBarrier,_ReadWriteBarrier)
222 #define ECB_MEMORY_FENCE _ReadWriteBarrier ()
223 #define ECB_MEMORY_FENCE_ACQUIRE _ReadWriteBarrier () /* according to msdn, _ReadBarrier is not a load fence */
224 #define ECB_MEMORY_FENCE_RELEASE _WriteBarrier ()
225 #elif defined _WIN32
226 #include <WinNT.h>
227 #define ECB_MEMORY_FENCE MemoryBarrier () /* actually just xchg on x86... scary */
228 #elif __SUNPRO_C >= 0x5110 || __SUNPRO_CC >= 0x5110
229 #include <mbarrier.h>
230 #define ECB_MEMORY_FENCE __machine_rw_barrier ()
231 #define ECB_MEMORY_FENCE_ACQUIRE __machine_r_barrier ()
232 #define ECB_MEMORY_FENCE_RELEASE __machine_w_barrier ()
233 #elif __xlC__
234 #define ECB_MEMORY_FENCE __sync ()
235 #endif
236 #endif
237
238 #ifndef ECB_MEMORY_FENCE
239 #if ECB_C11 && !defined __STDC_NO_ATOMICS__
240 /* we assume that these memory fences work on all variables/all memory accesses, */
241 /* not just C11 atomics and atomic accesses */
242 #include <stdatomic.h>
243 /* Unfortunately, neither gcc 4.7 nor clang 3.1 generate any instructions for */
244 /* any fence other than seq_cst, which isn't very efficient for us. */
245 /* Why that is, we don't know - either the C11 memory model is quite useless */
246 /* for most usages, or gcc and clang have a bug */
247 /* I *currently* lean towards the latter, and inefficiently implement */
248 /* all three of ecb's fences as a seq_cst fence */
249 /* Update, gcc-4.8 generates mfence for all c++ fences, but nothing */
250 /* for all __atomic_thread_fence's except seq_cst */
251 #define ECB_MEMORY_FENCE atomic_thread_fence (memory_order_seq_cst)
252 #endif
253 #endif
254
255 #ifndef ECB_MEMORY_FENCE
256 #if !ECB_AVOID_PTHREADS
257 /*
258 * if you get undefined symbol references to pthread_mutex_lock,
259 * or failure to find pthread.h, then you should implement
260 * the ECB_MEMORY_FENCE operations for your cpu/compiler
261 * OR provide pthread.h and link against the posix thread library
262 * of your system.
263 */
264 #include <pthread.h>
265 #define ECB_NEEDS_PTHREADS 1
266 #define ECB_MEMORY_FENCE_NEEDS_PTHREADS 1
267
268 static pthread_mutex_t ecb_mf_lock = PTHREAD_MUTEX_INITIALIZER;
269 #define ECB_MEMORY_FENCE do { pthread_mutex_lock (&ecb_mf_lock); pthread_mutex_unlock (&ecb_mf_lock); } while (0)
270 #endif
271 #endif
272
273 #if !defined ECB_MEMORY_FENCE_ACQUIRE && defined ECB_MEMORY_FENCE
274 #define ECB_MEMORY_FENCE_ACQUIRE ECB_MEMORY_FENCE
275 #endif
276
277 #if !defined ECB_MEMORY_FENCE_RELEASE && defined ECB_MEMORY_FENCE
278 #define ECB_MEMORY_FENCE_RELEASE ECB_MEMORY_FENCE
279 #endif
280
281 /*****************************************************************************/
282
283 #if __cplusplus
284 #define ecb_inline static inline
285 #elif ECB_GCC_VERSION(2,5)
286 #define ecb_inline static __inline__
287 #elif ECB_C99
288 #define ecb_inline static inline
289 #else
290 #define ecb_inline static
291 #endif
292
293 #if ECB_GCC_VERSION(3,3)
294 #define ecb_restrict __restrict__
295 #elif ECB_C99
296 #define ecb_restrict restrict
297 #else
298 #define ecb_restrict
299 #endif
300
301 typedef int ecb_bool;
302
303 #define ECB_CONCAT_(a, b) a ## b
304 #define ECB_CONCAT(a, b) ECB_CONCAT_(a, b)
305 #define ECB_STRINGIFY_(a) # a
306 #define ECB_STRINGIFY(a) ECB_STRINGIFY_(a)
307
308 #define ecb_function_ ecb_inline
309
310 #if ECB_GCC_VERSION(3,1) || ECB_CLANG_VERSION(2,8)
311 #define ecb_attribute(attrlist) __attribute__(attrlist)
312 #else
313 #define ecb_attribute(attrlist)
314 #endif
315
316 #if ECB_GCC_VERSION(3,1) || ECB_CLANG_BUILTIN(__builtin_constant_p)
317 #define ecb_is_constant(expr) __builtin_constant_p (expr)
318 #else
319 /* possible C11 impl for integral types
320 typedef struct ecb_is_constant_struct ecb_is_constant_struct;
321 #define ecb_is_constant(expr) _Generic ((1 ? (struct ecb_is_constant_struct *)0 : (void *)((expr) - (expr)), ecb_is_constant_struct *: 0, default: 1)) */
322
323 #define ecb_is_constant(expr) 0
324 #endif
325
326 #if ECB_GCC_VERSION(3,1) || ECB_CLANG_BUILTIN(__builtin_expect)
327 #define ecb_expect(expr,value) __builtin_expect ((expr),(value))
328 #else
329 #define ecb_expect(expr,value) (expr)
330 #endif
331
332 #if ECB_GCC_VERSION(3,1) || ECB_CLANG_BUILTIN(__builtin_prefetch)
333 #define ecb_prefetch(addr,rw,locality) __builtin_prefetch (addr, rw, locality)
334 #else
335 #define ecb_prefetch(addr,rw,locality)
336 #endif
337
338 /* no emulation for ecb_decltype */
339 #if ECB_GCC_VERSION(4,5)
340 #define ecb_decltype(x) __decltype(x)
341 #elif ECB_GCC_VERSION(3,0)
342 #define ecb_decltype(x) __typeof(x)
343 #endif
344
345 #if _MSC_VER >= 1300
346 #define ecb_deprecated __declspec(deprecated)
347 #else
348 #define ecb_deprecated ecb_attribute ((__deprecated__))
349 #endif
350
351 #define ecb_noinline ecb_attribute ((__noinline__))
352 #define ecb_unused ecb_attribute ((__unused__))
353 #define ecb_const ecb_attribute ((__const__))
354 #define ecb_pure ecb_attribute ((__pure__))
355
356 /* http://msdn.microsoft.com/en-us/library/k6ktzx3s.aspx __declspec(noreturn) */
357 #if ECB_C11
358 #define ecb_noreturn _Noreturn
359 #else
360 #define ecb_noreturn ecb_attribute ((__noreturn__))
361 #endif
362
363 #if ECB_GCC_VERSION(4,3)
364 #define ecb_artificial ecb_attribute ((__artificial__))
365 #define ecb_hot ecb_attribute ((__hot__))
366 #define ecb_cold ecb_attribute ((__cold__))
367 #else
368 #define ecb_artificial
369 #define ecb_hot
370 #define ecb_cold
371 #endif
372
373 /* put around conditional expressions if you are very sure that the */
374 /* expression is mostly true or mostly false. note that these return */
375 /* booleans, not the expression. */
376 #define ecb_expect_false(expr) ecb_expect (!!(expr), 0)
377 #define ecb_expect_true(expr) ecb_expect (!!(expr), 1)
378 /* for compatibility to the rest of the world */
379 #define ecb_likely(expr) ecb_expect_true (expr)
380 #define ecb_unlikely(expr) ecb_expect_false (expr)
381
382 /* count trailing zero bits and count # of one bits */
383 #if ECB_GCC_VERSION(3,4) \
384 || (ECB_CLANG_BUILTIN(__builtin_clz) && ECB_CLANG_BUILTIN(__builtin_clzll) \
385 && ECB_CLANG_BUILTIN(__builtin_ctz) && ECB_CLANG_BUILTIN(__builtin_ctzll) \
386 && ECB_CLANG_BUILTIN(__builtin_popcount))
387 /* we assume int == 32 bit, long == 32 or 64 bit and long long == 64 bit */
388 #define ecb_ld32(x) (__builtin_clz (x) ^ 31)
389 #define ecb_ld64(x) (__builtin_clzll (x) ^ 63)
390 #define ecb_ctz32(x) __builtin_ctz (x)
391 #define ecb_ctz64(x) __builtin_ctzll (x)
392 #define ecb_popcount32(x) __builtin_popcount (x)
393 /* no popcountll */
394 #else
395 ecb_function_ int ecb_ctz32 (uint32_t x) ecb_const;
396 ecb_function_ int
397 ecb_ctz32 (uint32_t x)
398 {
399 int r = 0;
400
401 x &= ~x + 1; /* this isolates the lowest bit */
402
403 #if ECB_branchless_on_i386
404 r += !!(x & 0xaaaaaaaa) << 0;
405 r += !!(x & 0xcccccccc) << 1;
406 r += !!(x & 0xf0f0f0f0) << 2;
407 r += !!(x & 0xff00ff00) << 3;
408 r += !!(x & 0xffff0000) << 4;
409 #else
410 if (x & 0xaaaaaaaa) r += 1;
411 if (x & 0xcccccccc) r += 2;
412 if (x & 0xf0f0f0f0) r += 4;
413 if (x & 0xff00ff00) r += 8;
414 if (x & 0xffff0000) r += 16;
415 #endif
416
417 return r;
418 }
419
420 ecb_function_ int ecb_ctz64 (uint64_t x) ecb_const;
421 ecb_function_ int
422 ecb_ctz64 (uint64_t x)
423 {
424 int shift = x & 0xffffffffU ? 0 : 32;
425 return ecb_ctz32 (x >> shift) + shift;
426 }
427
428 ecb_function_ int ecb_popcount32 (uint32_t x) ecb_const;
429 ecb_function_ int
430 ecb_popcount32 (uint32_t x)
431 {
432 x -= (x >> 1) & 0x55555555;
433 x = ((x >> 2) & 0x33333333) + (x & 0x33333333);
434 x = ((x >> 4) + x) & 0x0f0f0f0f;
435 x *= 0x01010101;
436
437 return x >> 24;
438 }
439
440 ecb_function_ int ecb_ld32 (uint32_t x) ecb_const;
441 ecb_function_ int ecb_ld32 (uint32_t x)
442 {
443 int r = 0;
444
445 if (x >> 16) { x >>= 16; r += 16; }
446 if (x >> 8) { x >>= 8; r += 8; }
447 if (x >> 4) { x >>= 4; r += 4; }
448 if (x >> 2) { x >>= 2; r += 2; }
449 if (x >> 1) { r += 1; }
450
451 return r;
452 }
453
454 ecb_function_ int ecb_ld64 (uint64_t x) ecb_const;
455 ecb_function_ int ecb_ld64 (uint64_t x)
456 {
457 int r = 0;
458
459 if (x >> 32) { x >>= 32; r += 32; }
460
461 return r + ecb_ld32 (x);
462 }
463 #endif
464
465 ecb_function_ ecb_bool ecb_is_pot32 (uint32_t x) ecb_const;
466 ecb_function_ ecb_bool ecb_is_pot32 (uint32_t x) { return !(x & (x - 1)); }
467 ecb_function_ ecb_bool ecb_is_pot64 (uint64_t x) ecb_const;
468 ecb_function_ ecb_bool ecb_is_pot64 (uint64_t x) { return !(x & (x - 1)); }
469
470 ecb_function_ uint8_t ecb_bitrev8 (uint8_t x) ecb_const;
471 ecb_function_ uint8_t ecb_bitrev8 (uint8_t x)
472 {
473 return ( (x * 0x0802U & 0x22110U)
474 | (x * 0x8020U & 0x88440U)) * 0x10101U >> 16;
475 }
476
477 ecb_function_ uint16_t ecb_bitrev16 (uint16_t x) ecb_const;
478 ecb_function_ uint16_t ecb_bitrev16 (uint16_t x)
479 {
480 x = ((x >> 1) & 0x5555) | ((x & 0x5555) << 1);
481 x = ((x >> 2) & 0x3333) | ((x & 0x3333) << 2);
482 x = ((x >> 4) & 0x0f0f) | ((x & 0x0f0f) << 4);
483 x = ( x >> 8 ) | ( x << 8);
484
485 return x;
486 }
487
488 ecb_function_ uint32_t ecb_bitrev32 (uint32_t x) ecb_const;
489 ecb_function_ uint32_t ecb_bitrev32 (uint32_t x)
490 {
491 x = ((x >> 1) & 0x55555555) | ((x & 0x55555555) << 1);
492 x = ((x >> 2) & 0x33333333) | ((x & 0x33333333) << 2);
493 x = ((x >> 4) & 0x0f0f0f0f) | ((x & 0x0f0f0f0f) << 4);
494 x = ((x >> 8) & 0x00ff00ff) | ((x & 0x00ff00ff) << 8);
495 x = ( x >> 16 ) | ( x << 16);
496
497 return x;
498 }
499
500 /* popcount64 is only available on 64 bit cpus as gcc builtin */
501 /* so for this version we are lazy */
502 ecb_function_ int ecb_popcount64 (uint64_t x) ecb_const;
503 ecb_function_ int
504 ecb_popcount64 (uint64_t x)
505 {
506 return ecb_popcount32 (x) + ecb_popcount32 (x >> 32);
507 }
508
509 ecb_inline uint8_t ecb_rotl8 (uint8_t x, unsigned int count) ecb_const;
510 ecb_inline uint8_t ecb_rotr8 (uint8_t x, unsigned int count) ecb_const;
511 ecb_inline uint16_t ecb_rotl16 (uint16_t x, unsigned int count) ecb_const;
512 ecb_inline uint16_t ecb_rotr16 (uint16_t x, unsigned int count) ecb_const;
513 ecb_inline uint32_t ecb_rotl32 (uint32_t x, unsigned int count) ecb_const;
514 ecb_inline uint32_t ecb_rotr32 (uint32_t x, unsigned int count) ecb_const;
515 ecb_inline uint64_t ecb_rotl64 (uint64_t x, unsigned int count) ecb_const;
516 ecb_inline uint64_t ecb_rotr64 (uint64_t x, unsigned int count) ecb_const;
517
518 ecb_inline uint8_t ecb_rotl8 (uint8_t x, unsigned int count) { return (x >> ( 8 - count)) | (x << count); }
519 ecb_inline uint8_t ecb_rotr8 (uint8_t x, unsigned int count) { return (x << ( 8 - count)) | (x >> count); }
520 ecb_inline uint16_t ecb_rotl16 (uint16_t x, unsigned int count) { return (x >> (16 - count)) | (x << count); }
521 ecb_inline uint16_t ecb_rotr16 (uint16_t x, unsigned int count) { return (x << (16 - count)) | (x >> count); }
522 ecb_inline uint32_t ecb_rotl32 (uint32_t x, unsigned int count) { return (x >> (32 - count)) | (x << count); }
523 ecb_inline uint32_t ecb_rotr32 (uint32_t x, unsigned int count) { return (x << (32 - count)) | (x >> count); }
524 ecb_inline uint64_t ecb_rotl64 (uint64_t x, unsigned int count) { return (x >> (64 - count)) | (x << count); }
525 ecb_inline uint64_t ecb_rotr64 (uint64_t x, unsigned int count) { return (x << (64 - count)) | (x >> count); }
526
527 #if ECB_GCC_VERSION(4,3) || (ECB_CLANG_BUILTIN(__builtin_bswap32) && ECB_CLANG_BUILTIN(__builtin_bswap64))
528 #define ecb_bswap16(x) (__builtin_bswap32 (x) >> 16)
529 #define ecb_bswap32(x) __builtin_bswap32 (x)
530 #define ecb_bswap64(x) __builtin_bswap64 (x)
531 #else
532 ecb_function_ uint16_t ecb_bswap16 (uint16_t x) ecb_const;
533 ecb_function_ uint16_t
534 ecb_bswap16 (uint16_t x)
535 {
536 return ecb_rotl16 (x, 8);
537 }
538
539 ecb_function_ uint32_t ecb_bswap32 (uint32_t x) ecb_const;
540 ecb_function_ uint32_t
541 ecb_bswap32 (uint32_t x)
542 {
543 return (((uint32_t)ecb_bswap16 (x)) << 16) | ecb_bswap16 (x >> 16);
544 }
545
546 ecb_function_ uint64_t ecb_bswap64 (uint64_t x) ecb_const;
547 ecb_function_ uint64_t
548 ecb_bswap64 (uint64_t x)
549 {
550 return (((uint64_t)ecb_bswap32 (x)) << 32) | ecb_bswap32 (x >> 32);
551 }
552 #endif
553
554 #if ECB_GCC_VERSION(4,5) || ECB_CLANG_BUILTIN(__builtin_unreachable)
555 #define ecb_unreachable() __builtin_unreachable ()
556 #else
557 /* this seems to work fine, but gcc always emits a warning for it :/ */
558 ecb_inline void ecb_unreachable (void) ecb_noreturn;
559 ecb_inline void ecb_unreachable (void) { }
560 #endif
561
562 /* try to tell the compiler that some condition is definitely true */
563 #define ecb_assume(cond) if (!(cond)) ecb_unreachable (); else 0
564
565 ecb_inline unsigned char ecb_byteorder_helper (void) ecb_const;
566 ecb_inline unsigned char
567 ecb_byteorder_helper (void)
568 {
569 /* the union code still generates code under pressure in gcc, */
570 /* but less than using pointers, and always seems to */
571 /* successfully return a constant. */
572 /* the reason why we have this horrible preprocessor mess */
573 /* is to avoid it in all cases, at least on common architectures */
574 /* or when using a recent enough gcc version (>= 4.6) */
575 #if __i386 || __i386__ || _M_X86 || __amd64 || __amd64__ || _M_X64
576 return 0x44;
577 #elif __BYTE_ORDER__ && __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__
578 return 0x44;
579 #elif __BYTE_ORDER__ && __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__
580 return 0x11;
581 #else
582 union
583 {
584 uint32_t i;
585 uint8_t c;
586 } u = { 0x11223344 };
587 return u.c;
588 #endif
589 }
590
591 ecb_inline ecb_bool ecb_big_endian (void) ecb_const;
592 ecb_inline ecb_bool ecb_big_endian (void) { return ecb_byteorder_helper () == 0x11; }
593 ecb_inline ecb_bool ecb_little_endian (void) ecb_const;
594 ecb_inline ecb_bool ecb_little_endian (void) { return ecb_byteorder_helper () == 0x44; }
595
596 #if ECB_GCC_VERSION(3,0) || ECB_C99
597 #define ecb_mod(m,n) ((m) % (n) + ((m) % (n) < 0 ? (n) : 0))
598 #else
599 #define ecb_mod(m,n) ((m) < 0 ? ((n) - 1 - ((-1 - (m)) % (n))) : ((m) % (n)))
600 #endif
601
602 #if __cplusplus
603 template<typename T>
604 static inline T ecb_div_rd (T val, T div)
605 {
606 return val < 0 ? - ((-val + div - 1) / div) : (val ) / div;
607 }
608 template<typename T>
609 static inline T ecb_div_ru (T val, T div)
610 {
611 return val < 0 ? - ((-val ) / div) : (val + div - 1) / div;
612 }
613 #else
614 #define ecb_div_rd(val,div) ((val) < 0 ? - ((-(val) + (div) - 1) / (div)) : ((val) ) / (div))
615 #define ecb_div_ru(val,div) ((val) < 0 ? - ((-(val) ) / (div)) : ((val) + (div) - 1) / (div))
616 #endif
617
618 #if ecb_cplusplus_does_not_suck
619 /* does not work for local types (http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2657.htm) */
620 template<typename T, int N>
621 static inline int ecb_array_length (const T (&arr)[N])
622 {
623 return N;
624 }
625 #else
626 #define ecb_array_length(name) (sizeof (name) / sizeof (name [0]))
627 #endif
628
629 /*******************************************************************************/
630 /* floating point stuff, can be disabled by defining ECB_NO_LIBM */
631
632 /* basically, everything uses "ieee pure-endian" floating point numbers */
633 /* the only noteworthy exception is ancient armle, which uses order 43218765 */
634 #if 0 \
635 || __i386 || __i386__ \
636 || __amd64 || __amd64__ || __x86_64 || __x86_64__ \
637 || __powerpc__ || __ppc__ || __powerpc64__ || __ppc64__ \
638 || defined __s390__ || defined __s390x__ \
639 || defined __mips__ \
640 || defined __alpha__ \
641 || defined __hppa__ \
642 || defined __ia64__ \
643 || defined __m68k__ \
644 || defined __m88k__ \
645 || defined __sh__ \
646 || defined _M_IX86 || defined _M_AMD64 || defined _M_IA64 \
647 || (defined __arm__ && (defined __ARM_EABI__ || defined __EABI__ || defined __VFP_FP__ || defined _WIN32_WCE || defined __ANDROID__)) \
648 || defined __aarch64__
649 #define ECB_STDFP 1
650 #include <string.h> /* for memcpy */
651 #else
652 #define ECB_STDFP 0
653 #endif
654
655 #ifndef ECB_NO_LIBM
656
657 #include <math.h> /* for frexp*, ldexp*, INFINITY, NAN */
658
659 /* only the oldest of old doesn't have this one. solaris. */
660 #ifdef INFINITY
661 #define ECB_INFINITY INFINITY
662 #else
663 #define ECB_INFINITY HUGE_VAL
664 #endif
665
666 #ifdef NAN
667 #define ECB_NAN NAN
668 #else
669 #define ECB_NAN ECB_INFINITY
670 #endif
671
672 /* converts an ieee half/binary16 to a float */
673 ecb_function_ float ecb_binary16_to_float (uint16_t x) ecb_const;
674 ecb_function_ float
675 ecb_binary16_to_float (uint16_t x)
676 {
677 int e = (x >> 10) & 0x1f;
678 int m = x & 0x3ff;
679 float r;
680
681 if (!e ) r = ldexpf (m , -24);
682 else if (e != 31) r = ldexpf (m + 0x400, e - 25);
683 else if (m ) r = ECB_NAN;
684 else r = ECB_INFINITY;
685
686 return x & 0x8000 ? -r : r;
687 }
688
689 /* convert a float to ieee single/binary32 */
690 ecb_function_ uint32_t ecb_float_to_binary32 (float x) ecb_const;
691 ecb_function_ uint32_t
692 ecb_float_to_binary32 (float x)
693 {
694 uint32_t r;
695
696 #if ECB_STDFP
697 memcpy (&r, &x, 4);
698 #else
699 /* slow emulation, works for anything but -0 */
700 uint32_t m;
701 int e;
702
703 if (x == 0e0f ) return 0x00000000U;
704 if (x > +3.40282346638528860e+38f) return 0x7f800000U;
705 if (x < -3.40282346638528860e+38f) return 0xff800000U;
706 if (x != x ) return 0x7fbfffffU;
707
708 m = frexpf (x, &e) * 0x1000000U;
709
710 r = m & 0x80000000U;
711
712 if (r)
713 m = -m;
714
715 if (e <= -126)
716 {
717 m &= 0xffffffU;
718 m >>= (-125 - e);
719 e = -126;
720 }
721
722 r |= (e + 126) << 23;
723 r |= m & 0x7fffffU;
724 #endif
725
726 return r;
727 }
728
729 /* converts an ieee single/binary32 to a float */
730 ecb_function_ float ecb_binary32_to_float (uint32_t x) ecb_const;
731 ecb_function_ float
732 ecb_binary32_to_float (uint32_t x)
733 {
734 float r;
735
736 #if ECB_STDFP
737 memcpy (&r, &x, 4);
738 #else
739 /* emulation, only works for normals and subnormals and +0 */
740 int neg = x >> 31;
741 int e = (x >> 23) & 0xffU;
742
743 x &= 0x7fffffU;
744
745 if (e)
746 x |= 0x800000U;
747 else
748 e = 1;
749
750 /* we distrust ldexpf a bit and do the 2**-24 scaling by an extra multiply */
751 r = ldexpf (x * (0.5f / 0x800000U), e - 126);
752
753 r = neg ? -r : r;
754 #endif
755
756 return r;
757 }
758
759 /* convert a double to ieee double/binary64 */
760 ecb_function_ uint64_t ecb_double_to_binary64 (double x) ecb_const;
761 ecb_function_ uint64_t
762 ecb_double_to_binary64 (double x)
763 {
764 uint64_t r;
765
766 #if ECB_STDFP
767 memcpy (&r, &x, 8);
768 #else
769 /* slow emulation, works for anything but -0 */
770 uint64_t m;
771 int e;
772
773 if (x == 0e0 ) return 0x0000000000000000U;
774 if (x > +1.79769313486231470e+308) return 0x7ff0000000000000U;
775 if (x < -1.79769313486231470e+308) return 0xfff0000000000000U;
776 if (x != x ) return 0X7ff7ffffffffffffU;
777
778 m = frexp (x, &e) * 0x20000000000000U;
779
780 r = m & 0x8000000000000000;;
781
782 if (r)
783 m = -m;
784
785 if (e <= -1022)
786 {
787 m &= 0x1fffffffffffffU;
788 m >>= (-1021 - e);
789 e = -1022;
790 }
791
792 r |= ((uint64_t)(e + 1022)) << 52;
793 r |= m & 0xfffffffffffffU;
794 #endif
795
796 return r;
797 }
798
799 /* converts an ieee double/binary64 to a double */
800 ecb_function_ double ecb_binary64_to_double (uint64_t x) ecb_const;
801 ecb_function_ double
802 ecb_binary64_to_double (uint64_t x)
803 {
804 double r;
805
806 #if ECB_STDFP
807 memcpy (&r, &x, 8);
808 #else
809 /* emulation, only works for normals and subnormals and +0 */
810 int neg = x >> 63;
811 int e = (x >> 52) & 0x7ffU;
812
813 x &= 0xfffffffffffffU;
814
815 if (e)
816 x |= 0x10000000000000U;
817 else
818 e = 1;
819
820 /* we distrust ldexp a bit and do the 2**-53 scaling by an extra multiply */
821 r = ldexp (x * (0.5 / 0x10000000000000U), e - 1022);
822
823 r = neg ? -r : r;
824 #endif
825
826 return r;
827 }
828
829 #endif
830
831 #endif
832