ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/libecb/ecb.h
Revision: 1.126
Committed: Sun Jan 5 02:58:29 2014 UTC (10 years, 5 months ago) by root
Content type: text/plain
Branch: MAIN
Changes since 1.125: +6 -0 lines
Log Message:
*** empty log message ***

File Contents

# Content
1 /*
2 * libecb - http://software.schmorp.de/pkg/libecb
3 *
4 * Copyright (©) 2009-2013 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 /* 16 bits major, 16 bits minor */
34 #define ECB_VERSION 0x00010003
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 #else
60 #include <inttypes.h>
61 #if UINTMAX_MAX > 0xffffffffU
62 #define ECB_PTRSIZE 8
63 #else
64 #define ECB_PTRSIZE 4
65 #endif
66 #endif
67
68 /* work around x32 idiocy by defining proper macros */
69 #if __amd64 || __x86_64 || _M_AMD64 || _M_X64
70 #if _ILP32
71 #define ECB_AMD64_X32 1
72 #else
73 #define ECB_AMD64 1
74 #endif
75 #endif
76
77 /* many compilers define _GNUC_ to some versions but then only implement
78 * what their idiot authors think are the "more important" extensions,
79 * causing enormous grief in return for some better fake benchmark numbers.
80 * or so.
81 * we try to detect these and simply assume they are not gcc - if they have
82 * an issue with that they should have done it right in the first place.
83 */
84 #ifndef ECB_GCC_VERSION
85 #if !defined __GNUC_MINOR__ || defined __INTEL_COMPILER || defined __SUNPRO_C || defined __SUNPRO_CC || defined __llvm__ || defined __clang__
86 #define ECB_GCC_VERSION(major,minor) 0
87 #else
88 #define ECB_GCC_VERSION(major,minor) (__GNUC__ > (major) || (__GNUC__ == (major) && __GNUC_MINOR__ >= (minor)))
89 #endif
90 #endif
91
92 #define ECB_C (__STDC__+0) /* this assumes that __STDC__ is either empty or a number */
93 #define ECB_C99 (__STDC_VERSION__ >= 199901L)
94 #define ECB_C11 (__STDC_VERSION__ >= 201112L)
95 #define ECB_CPP (__cplusplus+0)
96 #define ECB_CPP11 (__cplusplus >= 201103L)
97
98 #if ECB_CPP
99 #define ECB_EXTERN_C extern "C"
100 #define ECB_EXTERN_C_BEG ECB_EXTERN_C {
101 #define ECB_EXTERN_C_END }
102 #else
103 #define ECB_EXTERN_C extern
104 #define ECB_EXTERN_C_BEG
105 #define ECB_EXTERN_C_END
106 #endif
107
108 /*****************************************************************************/
109
110 /* ECB_NO_THREADS - ecb is not used by multiple threads, ever */
111 /* ECB_NO_SMP - ecb might be used in multiple threads, but only on a single cpu */
112
113 #if ECB_NO_THREADS
114 #define ECB_NO_SMP 1
115 #endif
116
117 #if ECB_NO_SMP
118 #define ECB_MEMORY_FENCE do { } while (0)
119 #endif
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 __asm__ __volatile__ ("" : : : "memory")
126 #define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("")
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__ ("" : : : "memory")
130 #define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("")
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__) && !__sparcv8
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 /* GNU/Linux emulates sync on mips1 architectures, so we force its use */
147 /* anybody else who still uses mips1 is supposed to send in their version, with detection code. */
148 #define ECB_MEMORY_FENCE __asm__ __volatile__ (".set mips2; sync; .set mips0" : : : "memory")
149 #elif defined __alpha__
150 #define ECB_MEMORY_FENCE __asm__ __volatile__ ("mb" : : : "memory")
151 #elif defined __hppa__
152 #define ECB_MEMORY_FENCE __asm__ __volatile__ ("" : : : "memory")
153 #define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("")
154 #elif defined __ia64__
155 #define ECB_MEMORY_FENCE __asm__ __volatile__ ("mf" : : : "memory")
156 #elif defined __m68k__
157 #define ECB_MEMORY_FENCE __asm__ __volatile__ ("" : : : "memory")
158 #elif defined __m88k__
159 #define ECB_MEMORY_FENCE __asm__ __volatile__ ("tb1 0,%%r0,128" : : : "memory")
160 #elif defined __sh__
161 #define ECB_MEMORY_FENCE __asm__ __volatile__ ("" : : : "memory")
162 #endif
163 #endif
164 #endif
165
166 #ifndef ECB_MEMORY_FENCE
167 #if ECB_GCC_VERSION(4,7)
168 /* see comment below (stdatomic.h) about the C11 memory model. */
169 #define ECB_MEMORY_FENCE __atomic_thread_fence (__ATOMIC_SEQ_CST)
170
171 /* The __has_feature syntax from clang is so misdesigned that we cannot use it
172 * without risking compile time errors with other compilers. We *could*
173 * define our own ecb_clang_has_feature, but I just can't be bothered to work
174 * around this shit time and again.
175 * #elif defined __clang && __has_feature (cxx_atomic)
176 * // see comment below (stdatomic.h) about the C11 memory model.
177 * #define ECB_MEMORY_FENCE __c11_atomic_thread_fence (__ATOMIC_SEQ_CST)
178 */
179
180 #elif ECB_GCC_VERSION(4,4) || defined __INTEL_COMPILER || defined __clang__
181 #define ECB_MEMORY_FENCE __sync_synchronize ()
182 #elif _MSC_VER >= 1500 /* VC++ 2008 */
183 /* apparently, microsoft broke all the memory barrier stuff in Visual Studio 2008... */
184 #pragma intrinsic(_ReadBarrier,_WriteBarrier,_ReadWriteBarrier)
185 #define ECB_MEMORY_FENCE _ReadWriteBarrier (); MemoryBarrier()
186 #define ECB_MEMORY_FENCE_ACQUIRE _ReadWriteBarrier (); MemoryBarrier() /* according to msdn, _ReadBarrier is not a load fence */
187 #define ECB_MEMORY_FENCE_RELEASE _WriteBarrier (); MemoryBarrier()
188 #elif _MSC_VER >= 1400 /* VC++ 2005 */
189 #pragma intrinsic(_ReadBarrier,_WriteBarrier,_ReadWriteBarrier)
190 #define ECB_MEMORY_FENCE _ReadWriteBarrier ()
191 #define ECB_MEMORY_FENCE_ACQUIRE _ReadWriteBarrier () /* according to msdn, _ReadBarrier is not a load fence */
192 #define ECB_MEMORY_FENCE_RELEASE _WriteBarrier ()
193 #elif defined _WIN32
194 #include <WinNT.h>
195 #define ECB_MEMORY_FENCE MemoryBarrier () /* actually just xchg on x86... scary */
196 #elif __SUNPRO_C >= 0x5110 || __SUNPRO_CC >= 0x5110
197 #include <mbarrier.h>
198 #define ECB_MEMORY_FENCE __machine_rw_barrier ()
199 #define ECB_MEMORY_FENCE_ACQUIRE __machine_r_barrier ()
200 #define ECB_MEMORY_FENCE_RELEASE __machine_w_barrier ()
201 #elif __xlC__
202 #define ECB_MEMORY_FENCE __sync ()
203 #endif
204 #endif
205
206 #ifndef ECB_MEMORY_FENCE
207 #if ECB_C11 && !defined __STDC_NO_ATOMICS__
208 /* we assume that these memory fences work on all variables/all memory accesses, */
209 /* not just C11 atomics and atomic accesses */
210 #include <stdatomic.h>
211 /* Unfortunately, neither gcc 4.7 nor clang 3.1 generate any instructions for */
212 /* any fence other than seq_cst, which isn't very efficient for us. */
213 /* Why that is, we don't know - either the C11 memory model is quite useless */
214 /* for most usages, or gcc and clang have a bug */
215 /* I *currently* lean towards the latter, and inefficiently implement */
216 /* all three of ecb's fences as a seq_cst fence */
217 #define ECB_MEMORY_FENCE atomic_thread_fence (memory_order_seq_cst)
218 #endif
219 #endif
220
221 #ifndef ECB_MEMORY_FENCE
222 #if !ECB_AVOID_PTHREADS
223 /*
224 * if you get undefined symbol references to pthread_mutex_lock,
225 * or failure to find pthread.h, then you should implement
226 * the ECB_MEMORY_FENCE operations for your cpu/compiler
227 * OR provide pthread.h and link against the posix thread library
228 * of your system.
229 */
230 #include <pthread.h>
231 #define ECB_NEEDS_PTHREADS 1
232 #define ECB_MEMORY_FENCE_NEEDS_PTHREADS 1
233
234 static pthread_mutex_t ecb_mf_lock = PTHREAD_MUTEX_INITIALIZER;
235 #define ECB_MEMORY_FENCE do { pthread_mutex_lock (&ecb_mf_lock); pthread_mutex_unlock (&ecb_mf_lock); } while (0)
236 #endif
237 #endif
238
239 #if !defined ECB_MEMORY_FENCE_ACQUIRE && defined ECB_MEMORY_FENCE
240 #define ECB_MEMORY_FENCE_ACQUIRE ECB_MEMORY_FENCE
241 #endif
242
243 #if !defined ECB_MEMORY_FENCE_RELEASE && defined ECB_MEMORY_FENCE
244 #define ECB_MEMORY_FENCE_RELEASE ECB_MEMORY_FENCE
245 #endif
246
247 /*****************************************************************************/
248
249 #if __cplusplus
250 #define ecb_inline static inline
251 #elif ECB_GCC_VERSION(2,5)
252 #define ecb_inline static __inline__
253 #elif ECB_C99
254 #define ecb_inline static inline
255 #else
256 #define ecb_inline static
257 #endif
258
259 #if ECB_GCC_VERSION(3,3)
260 #define ecb_restrict __restrict__
261 #elif ECB_C99
262 #define ecb_restrict restrict
263 #else
264 #define ecb_restrict
265 #endif
266
267 typedef int ecb_bool;
268
269 #define ECB_CONCAT_(a, b) a ## b
270 #define ECB_CONCAT(a, b) ECB_CONCAT_(a, b)
271 #define ECB_STRINGIFY_(a) # a
272 #define ECB_STRINGIFY(a) ECB_STRINGIFY_(a)
273
274 #define ecb_function_ ecb_inline
275
276 #if ECB_GCC_VERSION(3,1)
277 #define ecb_attribute(attrlist) __attribute__(attrlist)
278 #define ecb_is_constant(expr) __builtin_constant_p (expr)
279 #define ecb_expect(expr,value) __builtin_expect ((expr),(value))
280 #define ecb_prefetch(addr,rw,locality) __builtin_prefetch (addr, rw, locality)
281 #else
282 #define ecb_attribute(attrlist)
283 #define ecb_is_constant(expr) 0
284 #define ecb_expect(expr,value) (expr)
285 #define ecb_prefetch(addr,rw,locality)
286 #endif
287
288 /* no emulation for ecb_decltype */
289 #if ECB_GCC_VERSION(4,5)
290 #define ecb_decltype(x) __decltype(x)
291 #elif ECB_GCC_VERSION(3,0)
292 #define ecb_decltype(x) __typeof(x)
293 #endif
294
295 #define ecb_noinline ecb_attribute ((__noinline__))
296 #define ecb_unused ecb_attribute ((__unused__))
297 #define ecb_const ecb_attribute ((__const__))
298 #define ecb_pure ecb_attribute ((__pure__))
299
300 #if ECB_C11
301 #define ecb_noreturn _Noreturn
302 #else
303 #define ecb_noreturn ecb_attribute ((__noreturn__))
304 #endif
305
306 #if ECB_GCC_VERSION(4,3)
307 #define ecb_artificial ecb_attribute ((__artificial__))
308 #define ecb_hot ecb_attribute ((__hot__))
309 #define ecb_cold ecb_attribute ((__cold__))
310 #else
311 #define ecb_artificial
312 #define ecb_hot
313 #define ecb_cold
314 #endif
315
316 /* put around conditional expressions if you are very sure that the */
317 /* expression is mostly true or mostly false. note that these return */
318 /* booleans, not the expression. */
319 #define ecb_expect_false(expr) ecb_expect (!!(expr), 0)
320 #define ecb_expect_true(expr) ecb_expect (!!(expr), 1)
321 /* for compatibility to the rest of the world */
322 #define ecb_likely(expr) ecb_expect_true (expr)
323 #define ecb_unlikely(expr) ecb_expect_false (expr)
324
325 /* count trailing zero bits and count # of one bits */
326 #if ECB_GCC_VERSION(3,4)
327 /* we assume int == 32 bit, long == 32 or 64 bit and long long == 64 bit */
328 #define ecb_ld32(x) (__builtin_clz (x) ^ 31)
329 #define ecb_ld64(x) (__builtin_clzll (x) ^ 63)
330 #define ecb_ctz32(x) __builtin_ctz (x)
331 #define ecb_ctz64(x) __builtin_ctzll (x)
332 #define ecb_popcount32(x) __builtin_popcount (x)
333 /* no popcountll */
334 #else
335 ecb_function_ int ecb_ctz32 (uint32_t x) ecb_const;
336 ecb_function_ int
337 ecb_ctz32 (uint32_t x)
338 {
339 int r = 0;
340
341 x &= ~x + 1; /* this isolates the lowest bit */
342
343 #if ECB_branchless_on_i386
344 r += !!(x & 0xaaaaaaaa) << 0;
345 r += !!(x & 0xcccccccc) << 1;
346 r += !!(x & 0xf0f0f0f0) << 2;
347 r += !!(x & 0xff00ff00) << 3;
348 r += !!(x & 0xffff0000) << 4;
349 #else
350 if (x & 0xaaaaaaaa) r += 1;
351 if (x & 0xcccccccc) r += 2;
352 if (x & 0xf0f0f0f0) r += 4;
353 if (x & 0xff00ff00) r += 8;
354 if (x & 0xffff0000) r += 16;
355 #endif
356
357 return r;
358 }
359
360 ecb_function_ int ecb_ctz64 (uint64_t x) ecb_const;
361 ecb_function_ int
362 ecb_ctz64 (uint64_t x)
363 {
364 int shift = x & 0xffffffffU ? 0 : 32;
365 return ecb_ctz32 (x >> shift) + shift;
366 }
367
368 ecb_function_ int ecb_popcount32 (uint32_t x) ecb_const;
369 ecb_function_ int
370 ecb_popcount32 (uint32_t x)
371 {
372 x -= (x >> 1) & 0x55555555;
373 x = ((x >> 2) & 0x33333333) + (x & 0x33333333);
374 x = ((x >> 4) + x) & 0x0f0f0f0f;
375 x *= 0x01010101;
376
377 return x >> 24;
378 }
379
380 ecb_function_ int ecb_ld32 (uint32_t x) ecb_const;
381 ecb_function_ int ecb_ld32 (uint32_t x)
382 {
383 int r = 0;
384
385 if (x >> 16) { x >>= 16; r += 16; }
386 if (x >> 8) { x >>= 8; r += 8; }
387 if (x >> 4) { x >>= 4; r += 4; }
388 if (x >> 2) { x >>= 2; r += 2; }
389 if (x >> 1) { r += 1; }
390
391 return r;
392 }
393
394 ecb_function_ int ecb_ld64 (uint64_t x) ecb_const;
395 ecb_function_ int ecb_ld64 (uint64_t x)
396 {
397 int r = 0;
398
399 if (x >> 32) { x >>= 32; r += 32; }
400
401 return r + ecb_ld32 (x);
402 }
403 #endif
404
405 ecb_function_ ecb_bool ecb_is_pot32 (uint32_t x) ecb_const;
406 ecb_function_ ecb_bool ecb_is_pot32 (uint32_t x) { return !(x & (x - 1)); }
407 ecb_function_ ecb_bool ecb_is_pot64 (uint64_t x) ecb_const;
408 ecb_function_ ecb_bool ecb_is_pot64 (uint64_t x) { return !(x & (x - 1)); }
409
410 ecb_function_ uint8_t ecb_bitrev8 (uint8_t x) ecb_const;
411 ecb_function_ uint8_t ecb_bitrev8 (uint8_t x)
412 {
413 return ( (x * 0x0802U & 0x22110U)
414 | (x * 0x8020U & 0x88440U)) * 0x10101U >> 16;
415 }
416
417 ecb_function_ uint16_t ecb_bitrev16 (uint16_t x) ecb_const;
418 ecb_function_ uint16_t ecb_bitrev16 (uint16_t x)
419 {
420 x = ((x >> 1) & 0x5555) | ((x & 0x5555) << 1);
421 x = ((x >> 2) & 0x3333) | ((x & 0x3333) << 2);
422 x = ((x >> 4) & 0x0f0f) | ((x & 0x0f0f) << 4);
423 x = ( x >> 8 ) | ( x << 8);
424
425 return x;
426 }
427
428 ecb_function_ uint32_t ecb_bitrev32 (uint32_t x) ecb_const;
429 ecb_function_ uint32_t ecb_bitrev32 (uint32_t x)
430 {
431 x = ((x >> 1) & 0x55555555) | ((x & 0x55555555) << 1);
432 x = ((x >> 2) & 0x33333333) | ((x & 0x33333333) << 2);
433 x = ((x >> 4) & 0x0f0f0f0f) | ((x & 0x0f0f0f0f) << 4);
434 x = ((x >> 8) & 0x00ff00ff) | ((x & 0x00ff00ff) << 8);
435 x = ( x >> 16 ) | ( x << 16);
436
437 return x;
438 }
439
440 /* popcount64 is only available on 64 bit cpus as gcc builtin */
441 /* so for this version we are lazy */
442 ecb_function_ int ecb_popcount64 (uint64_t x) ecb_const;
443 ecb_function_ int
444 ecb_popcount64 (uint64_t x)
445 {
446 return ecb_popcount32 (x) + ecb_popcount32 (x >> 32);
447 }
448
449 ecb_inline uint8_t ecb_rotl8 (uint8_t x, unsigned int count) ecb_const;
450 ecb_inline uint8_t ecb_rotr8 (uint8_t x, unsigned int count) ecb_const;
451 ecb_inline uint16_t ecb_rotl16 (uint16_t x, unsigned int count) ecb_const;
452 ecb_inline uint16_t ecb_rotr16 (uint16_t x, unsigned int count) ecb_const;
453 ecb_inline uint32_t ecb_rotl32 (uint32_t x, unsigned int count) ecb_const;
454 ecb_inline uint32_t ecb_rotr32 (uint32_t x, unsigned int count) ecb_const;
455 ecb_inline uint64_t ecb_rotl64 (uint64_t x, unsigned int count) ecb_const;
456 ecb_inline uint64_t ecb_rotr64 (uint64_t x, unsigned int count) ecb_const;
457
458 ecb_inline uint8_t ecb_rotl8 (uint8_t x, unsigned int count) { return (x >> ( 8 - count)) | (x << count); }
459 ecb_inline uint8_t ecb_rotr8 (uint8_t x, unsigned int count) { return (x << ( 8 - count)) | (x >> count); }
460 ecb_inline uint16_t ecb_rotl16 (uint16_t x, unsigned int count) { return (x >> (16 - count)) | (x << count); }
461 ecb_inline uint16_t ecb_rotr16 (uint16_t x, unsigned int count) { return (x << (16 - count)) | (x >> count); }
462 ecb_inline uint32_t ecb_rotl32 (uint32_t x, unsigned int count) { return (x >> (32 - count)) | (x << count); }
463 ecb_inline uint32_t ecb_rotr32 (uint32_t x, unsigned int count) { return (x << (32 - count)) | (x >> count); }
464 ecb_inline uint64_t ecb_rotl64 (uint64_t x, unsigned int count) { return (x >> (64 - count)) | (x << count); }
465 ecb_inline uint64_t ecb_rotr64 (uint64_t x, unsigned int count) { return (x << (64 - count)) | (x >> count); }
466
467 #if ECB_GCC_VERSION(4,3)
468 #define ecb_bswap16(x) (__builtin_bswap32 (x) >> 16)
469 #define ecb_bswap32(x) __builtin_bswap32 (x)
470 #define ecb_bswap64(x) __builtin_bswap64 (x)
471 #else
472 ecb_function_ uint16_t ecb_bswap16 (uint16_t x) ecb_const;
473 ecb_function_ uint16_t
474 ecb_bswap16 (uint16_t x)
475 {
476 return ecb_rotl16 (x, 8);
477 }
478
479 ecb_function_ uint32_t ecb_bswap32 (uint32_t x) ecb_const;
480 ecb_function_ uint32_t
481 ecb_bswap32 (uint32_t x)
482 {
483 return (((uint32_t)ecb_bswap16 (x)) << 16) | ecb_bswap16 (x >> 16);
484 }
485
486 ecb_function_ uint64_t ecb_bswap64 (uint64_t x) ecb_const;
487 ecb_function_ uint64_t
488 ecb_bswap64 (uint64_t x)
489 {
490 return (((uint64_t)ecb_bswap32 (x)) << 32) | ecb_bswap32 (x >> 32);
491 }
492 #endif
493
494 #if ECB_GCC_VERSION(4,5)
495 #define ecb_unreachable() __builtin_unreachable ()
496 #else
497 /* this seems to work fine, but gcc always emits a warning for it :/ */
498 ecb_inline void ecb_unreachable (void) ecb_noreturn;
499 ecb_inline void ecb_unreachable (void) { }
500 #endif
501
502 /* try to tell the compiler that some condition is definitely true */
503 #define ecb_assume(cond) if (!(cond)) ecb_unreachable (); else 0
504
505 ecb_inline unsigned char ecb_byteorder_helper (void) ecb_const;
506 ecb_inline unsigned char
507 ecb_byteorder_helper (void)
508 {
509 /* the union code still generates code under pressure in gcc, */
510 /* but less than using pointers, and always seems to */
511 /* successfully return a constant. */
512 /* the reason why we have this horrible preprocessor mess */
513 /* is to avoid it in all cases, at least on common architectures */
514 /* or when using a recent enough gcc version (>= 4.6) */
515 #if __i386 || __i386__ || _M_X86 || __amd64 || __amd64__ || _M_X64
516 return 0x44;
517 #elif __BYTE_ORDER__ && __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__
518 return 0x44;
519 #elif __BYTE_ORDER__ && __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__
520 return 0x11;
521 #else
522 union
523 {
524 uint32_t i;
525 uint8_t c;
526 } u = { 0x11223344 };
527 return u.c;
528 #endif
529 }
530
531 ecb_inline ecb_bool ecb_big_endian (void) ecb_const;
532 ecb_inline ecb_bool ecb_big_endian (void) { return ecb_byteorder_helper () == 0x11; }
533 ecb_inline ecb_bool ecb_little_endian (void) ecb_const;
534 ecb_inline ecb_bool ecb_little_endian (void) { return ecb_byteorder_helper () == 0x44; }
535
536 #if ECB_GCC_VERSION(3,0) || ECB_C99
537 #define ecb_mod(m,n) ((m) % (n) + ((m) % (n) < 0 ? (n) : 0))
538 #else
539 #define ecb_mod(m,n) ((m) < 0 ? ((n) - 1 - ((-1 - (m)) % (n))) : ((m) % (n)))
540 #endif
541
542 #if __cplusplus
543 template<typename T>
544 static inline T ecb_div_rd (T val, T div)
545 {
546 return val < 0 ? - ((-val + div - 1) / div) : (val ) / div;
547 }
548 template<typename T>
549 static inline T ecb_div_ru (T val, T div)
550 {
551 return val < 0 ? - ((-val ) / div) : (val + div - 1) / div;
552 }
553 #else
554 #define ecb_div_rd(val,div) ((val) < 0 ? - ((-(val) + (div) - 1) / (div)) : ((val) ) / (div))
555 #define ecb_div_ru(val,div) ((val) < 0 ? - ((-(val) ) / (div)) : ((val) + (div) - 1) / (div))
556 #endif
557
558 #if ecb_cplusplus_does_not_suck
559 /* does not work for local types (http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2657.htm) */
560 template<typename T, int N>
561 static inline int ecb_array_length (const T (&arr)[N])
562 {
563 return N;
564 }
565 #else
566 #define ecb_array_length(name) (sizeof (name) / sizeof (name [0]))
567 #endif
568
569 /*******************************************************************************/
570 /* floating point stuff, can be disabled by defining ECB_NO_LIBM */
571
572 /* basically, everything uses "ieee pure-endian" floating point numbers */
573 /* the only noteworthy exception is ancient armle, which uses order 43218765 */
574 #if 0 \
575 || __i386 || __i386__ \
576 || __amd64 || __amd64__ || __x86_64 || __x86_64__ \
577 || __powerpc__ || __ppc__ || __powerpc64__ || __ppc64__ \
578 || defined __arm__ && defined __ARM_EABI__ \
579 || defined __s390__ || defined __s390x__ \
580 || defined __mips__ \
581 || defined __alpha__ \
582 || defined __hppa__ \
583 || defined __ia64__ \
584 || defined __m68k__ \
585 || defined __m88k__ \
586 || defined __sh__ \
587 || defined _M_IX86 || defined _M_AMD64 || defined _M_IA64
588 #define ECB_STDFP 1
589 #include <string.h> /* for memcpy */
590 #else
591 #define ECB_STDFP 0
592 #endif
593
594 #ifndef ECB_NO_LIBM
595
596 #include <math.h> /* for frexp*, ldexp*, INFINITY, NAN */
597
598 /* only the oldest of old doesn't have this one. solaris. */
599 #ifdef INFINITY
600 #define ECB_INFINITY INFINITY
601 #else
602 #define ECB_INFINITY HUGE_VAL
603 #endif
604
605 #ifdef NAN
606 #define ECB_NAN NAN
607 #else
608 #define ECB_NAN ECB_INFINITY
609 #endif
610
611 /* converts an ieee half/binary16 to a float */
612 ecb_function_ float ecb_binary16_to_float (uint16_t x) ecb_const;
613 ecb_function_ float
614 ecb_binary16_to_float (uint16_t x)
615 {
616 int e = (x >> 10) & 0x1f;
617 int m = x & 0x3ff;
618 float r;
619
620 if (!e ) r = ldexpf (m , -24);
621 else if (e != 31) r = ldexpf (m + 0x400, e - 25);
622 else if (m ) r = ECB_NAN;
623 else r = ECB_INFINITY;
624
625 return x & 0x8000 ? -r : r;
626 }
627
628 /* convert a float to ieee single/binary32 */
629 ecb_function_ uint32_t ecb_float_to_binary32 (float x) ecb_const;
630 ecb_function_ uint32_t
631 ecb_float_to_binary32 (float x)
632 {
633 uint32_t r;
634
635 #if ECB_STDFP
636 memcpy (&r, &x, 4);
637 #else
638 /* slow emulation, works for anything but -0 */
639 uint32_t m;
640 int e;
641
642 if (x == 0e0f ) return 0x00000000U;
643 if (x > +3.40282346638528860e+38f) return 0x7f800000U;
644 if (x < -3.40282346638528860e+38f) return 0xff800000U;
645 if (x != x ) return 0x7fbfffffU;
646
647 m = frexpf (x, &e) * 0x1000000U;
648
649 r = m & 0x80000000U;
650
651 if (r)
652 m = -m;
653
654 if (e <= -126)
655 {
656 m &= 0xffffffU;
657 m >>= (-125 - e);
658 e = -126;
659 }
660
661 r |= (e + 126) << 23;
662 r |= m & 0x7fffffU;
663 #endif
664
665 return r;
666 }
667
668 /* converts an ieee single/binary32 to a float */
669 ecb_function_ float ecb_binary32_to_float (uint32_t x) ecb_const;
670 ecb_function_ float
671 ecb_binary32_to_float (uint32_t x)
672 {
673 float r;
674
675 #if ECB_STDFP
676 memcpy (&r, &x, 4);
677 #else
678 /* emulation, only works for normals and subnormals and +0 */
679 int neg = x >> 31;
680 int e = (x >> 23) & 0xffU;
681
682 x &= 0x7fffffU;
683
684 if (e)
685 x |= 0x800000U;
686 else
687 e = 1;
688
689 /* we distrust ldexpf a bit and do the 2**-24 scaling by an extra multiply */
690 r = ldexpf (x * (0.5f / 0x800000U), e - 126);
691
692 r = neg ? -r : r;
693 #endif
694
695 return r;
696 }
697
698 /* convert a double to ieee double/binary64 */
699 ecb_function_ uint64_t ecb_double_to_binary64 (double x) ecb_const;
700 ecb_function_ uint64_t
701 ecb_double_to_binary64 (double x)
702 {
703 uint64_t r;
704
705 #if ECB_STDFP
706 memcpy (&r, &x, 8);
707 #else
708 /* slow emulation, works for anything but -0 */
709 uint64_t m;
710 int e;
711
712 if (x == 0e0 ) return 0x0000000000000000U;
713 if (x > +1.79769313486231470e+308) return 0x7ff0000000000000U;
714 if (x < -1.79769313486231470e+308) return 0xfff0000000000000U;
715 if (x != x ) return 0X7ff7ffffffffffffU;
716
717 m = frexp (x, &e) * 0x20000000000000U;
718
719 r = m & 0x8000000000000000;;
720
721 if (r)
722 m = -m;
723
724 if (e <= -1022)
725 {
726 m &= 0x1fffffffffffffU;
727 m >>= (-1021 - e);
728 e = -1022;
729 }
730
731 r |= ((uint64_t)(e + 1022)) << 52;
732 r |= m & 0xfffffffffffffU;
733 #endif
734
735 return r;
736 }
737
738 /* converts an ieee double/binary64 to a double */
739 ecb_function_ double ecb_binary64_to_double (uint64_t x) ecb_const;
740 ecb_function_ double
741 ecb_binary64_to_double (uint64_t x)
742 {
743 double r;
744
745 #if ECB_STDFP
746 memcpy (&r, &x, 8);
747 #else
748 /* emulation, only works for normals and subnormals and +0 */
749 int neg = x >> 63;
750 int e = (x >> 52) & 0x7ffU;
751
752 x &= 0xfffffffffffffU;
753
754 if (e)
755 x |= 0x10000000000000U;
756 else
757 e = 1;
758
759 /* we distrust ldexp a bit and do the 2**-53 scaling by an extra multiply */
760 r = ldexp (x * (0.5 / 0x10000000000000U), e - 1022);
761
762 r = neg ? -r : r;
763 #endif
764
765 return r;
766 }
767
768 #endif
769
770 #endif
771