ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/libecb/ecb.h
(Generate patch)

Comparing libecb/ecb.h (file contents):
Revision 1.125 by root, Wed Dec 4 10:17:49 2013 UTC vs.
Revision 1.157 by root, Fri Feb 20 17:17:26 2015 UTC

1/* 1/*
2 * libecb - http://software.schmorp.de/pkg/libecb 2 * libecb - http://software.schmorp.de/pkg/libecb
3 * 3 *
4 * Copyright (©) 2009-2013 Marc Alexander Lehmann <libecb@schmorp.de> 4 * Copyright (©) 2009-2015 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:
23 * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; 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, 24 * OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
25 * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTH- 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 26 * ERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED
27 * OF THE POSSIBILITY OF SUCH DAMAGE. 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.
28 */ 39 */
29 40
30#ifndef ECB_H 41#ifndef ECB_H
31#define ECB_H 42#define ECB_H
32 43
33/* 16 bits major, 16 bits minor */ 44/* 16 bits major, 16 bits minor */
34#define ECB_VERSION 0x00010003 45#define ECB_VERSION 0x00010004
35 46
36#ifdef _WIN32 47#ifdef _WIN32
37 typedef signed char int8_t; 48 typedef signed char int8_t;
38 typedef unsigned char uint8_t; 49 typedef unsigned char uint8_t;
39 typedef signed short int16_t; 50 typedef signed short int16_t;
79 * causing enormous grief in return for some better fake benchmark numbers. 90 * causing enormous grief in return for some better fake benchmark numbers.
80 * or so. 91 * or so.
81 * we try to detect these and simply assume they are not gcc - if they have 92 * 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. 93 * an issue with that they should have done it right in the first place.
83 */ 94 */
84#ifndef ECB_GCC_VERSION
85 #if !defined __GNUC_MINOR__ || defined __INTEL_COMPILER || defined __SUNPRO_C || defined __SUNPRO_CC || defined __llvm__ || defined __clang__ 95#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 96 #define ECB_GCC_VERSION(major,minor) 0
87 #else 97#else
88 #define ECB_GCC_VERSION(major,minor) (__GNUC__ > (major) || (__GNUC__ == (major) && __GNUC_MINOR__ >= (minor))) 98 #define ECB_GCC_VERSION(major,minor) (__GNUC__ > (major) || (__GNUC__ == (major) && __GNUC_MINOR__ >= (minor)))
89 #endif 99#endif
90#endif
91 100
92#define ECB_C (__STDC__+0) /* this assumes that __STDC__ is either empty or a number */ 101#define ECB_CLANG_VERSION(major,minor) (__clang_major__ > (major) || (__clang_major__ == (major) && __clang_minor__ >= (minor)))
93#define ECB_C99 (__STDC_VERSION__ >= 199901L) 102
94#define ECB_C11 (__STDC_VERSION__ >= 201112L) 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
95#define ECB_CPP (__cplusplus+0) 115#define ECB_CPP (__cplusplus+0)
96#define ECB_CPP11 (__cplusplus >= 201103L) 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)
97 128
98#if ECB_CPP 129#if ECB_CPP
99 #define ECB_EXTERN_C extern "C" 130 #define ECB_EXTERN_C extern "C"
100 #define ECB_EXTERN_C_BEG ECB_EXTERN_C { 131 #define ECB_EXTERN_C_BEG ECB_EXTERN_C {
101 #define ECB_EXTERN_C_END } 132 #define ECB_EXTERN_C_END }
134 || defined __ARM_ARCH_6K__ || defined __ARM_ARCH_6ZK__ 165 || 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") 166 #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__ \ 167 #elif defined __ARM_ARCH_7__ || defined __ARM_ARCH_7A__ \
137 || defined __ARM_ARCH_7M__ || defined __ARM_ARCH_7R__ 168 || defined __ARM_ARCH_7M__ || defined __ARM_ARCH_7R__
138 #define ECB_MEMORY_FENCE __asm__ __volatile__ ("dmb" : : : "memory") 169 #define ECB_MEMORY_FENCE __asm__ __volatile__ ("dmb" : : : "memory")
170 #elif __aarch64__
171 #define ECB_MEMORY_FENCE __asm__ __volatile__ ("dmb ish" : : : "memory")
139 #elif (__sparc || __sparc__) && !__sparcv8 172 #elif (__sparc || __sparc__) && !__sparcv8
140 #define ECB_MEMORY_FENCE __asm__ __volatile__ ("membar #LoadStore | #LoadLoad | #StoreStore | #StoreLoad" : : : "memory") 173 #define ECB_MEMORY_FENCE __asm__ __volatile__ ("membar #LoadStore | #LoadLoad | #StoreStore | #StoreLoad" : : : "memory")
141 #define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("membar #LoadStore | #LoadLoad" : : : "memory") 174 #define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("membar #LoadStore | #LoadLoad" : : : "memory")
142 #define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("membar #LoadStore | #StoreStore") 175 #define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("membar #LoadStore | #StoreStore")
143 #elif defined __s390__ || defined __s390x__ 176 #elif defined __s390__ || defined __s390x__
165 198
166#ifndef ECB_MEMORY_FENCE 199#ifndef ECB_MEMORY_FENCE
167 #if ECB_GCC_VERSION(4,7) 200 #if ECB_GCC_VERSION(4,7)
168 /* see comment below (stdatomic.h) about the C11 memory model. */ 201 /* see comment below (stdatomic.h) about the C11 memory model. */
169 #define ECB_MEMORY_FENCE __atomic_thread_fence (__ATOMIC_SEQ_CST) 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)
170 205
171 /* The __has_feature syntax from clang is so misdesigned that we cannot use it 206 #elif ECB_CLANG_EXTENSION(c_atomic)
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. 207 /* see comment below (stdatomic.h) about the C11 memory model. */
177 * #define ECB_MEMORY_FENCE __c11_atomic_thread_fence (__ATOMIC_SEQ_CST) 208 #define ECB_MEMORY_FENCE __c11_atomic_thread_fence (__ATOMIC_SEQ_CST)
178 */ 209 #define ECB_MEMORY_FENCE_ACQUIRE __c11_atomic_thread_fence (__ATOMIC_ACQUIRE)
210 #define ECB_MEMORY_FENCE_RELEASE __c11_atomic_thread_fence (__ATOMIC_RELEASE)
179 211
180 #elif ECB_GCC_VERSION(4,4) || defined __INTEL_COMPILER || defined __clang__ 212 #elif ECB_GCC_VERSION(4,4) || defined __INTEL_COMPILER || defined __clang__
181 #define ECB_MEMORY_FENCE __sync_synchronize () 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()
182 #elif _MSC_VER >= 1400 /* VC++ 2005 */ 220 #elif _MSC_VER >= 1400 /* VC++ 2005 */
183 #pragma intrinsic(_ReadBarrier,_WriteBarrier,_ReadWriteBarrier) 221 #pragma intrinsic(_ReadBarrier,_WriteBarrier,_ReadWriteBarrier)
184 #define ECB_MEMORY_FENCE _ReadWriteBarrier () 222 #define ECB_MEMORY_FENCE _ReadWriteBarrier ()
185 #define ECB_MEMORY_FENCE_ACQUIRE _ReadWriteBarrier () /* according to msdn, _ReadBarrier is not a load fence */ 223 #define ECB_MEMORY_FENCE_ACQUIRE _ReadWriteBarrier () /* according to msdn, _ReadBarrier is not a load fence */
186 #define ECB_MEMORY_FENCE_RELEASE _WriteBarrier () 224 #define ECB_MEMORY_FENCE_RELEASE _WriteBarrier ()
206 /* any fence other than seq_cst, which isn't very efficient for us. */ 244 /* any fence other than seq_cst, which isn't very efficient for us. */
207 /* Why that is, we don't know - either the C11 memory model is quite useless */ 245 /* Why that is, we don't know - either the C11 memory model is quite useless */
208 /* for most usages, or gcc and clang have a bug */ 246 /* for most usages, or gcc and clang have a bug */
209 /* I *currently* lean towards the latter, and inefficiently implement */ 247 /* I *currently* lean towards the latter, and inefficiently implement */
210 /* all three of ecb's fences as a seq_cst fence */ 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 */
211 #define ECB_MEMORY_FENCE atomic_thread_fence (memory_order_seq_cst) 251 #define ECB_MEMORY_FENCE atomic_thread_fence (memory_order_seq_cst)
212 #endif 252 #endif
213#endif 253#endif
214 254
215#ifndef ECB_MEMORY_FENCE 255#ifndef ECB_MEMORY_FENCE
238 #define ECB_MEMORY_FENCE_RELEASE ECB_MEMORY_FENCE 278 #define ECB_MEMORY_FENCE_RELEASE ECB_MEMORY_FENCE
239#endif 279#endif
240 280
241/*****************************************************************************/ 281/*****************************************************************************/
242 282
243#if __cplusplus 283#if ECB_CPP
244 #define ecb_inline static inline 284 #define ecb_inline static inline
245#elif ECB_GCC_VERSION(2,5) 285#elif ECB_GCC_VERSION(2,5)
246 #define ecb_inline static __inline__ 286 #define ecb_inline static __inline__
247#elif ECB_C99 287#elif ECB_C99
248 #define ecb_inline static inline 288 #define ecb_inline static inline
262 302
263#define ECB_CONCAT_(a, b) a ## b 303#define ECB_CONCAT_(a, b) a ## b
264#define ECB_CONCAT(a, b) ECB_CONCAT_(a, b) 304#define ECB_CONCAT(a, b) ECB_CONCAT_(a, b)
265#define ECB_STRINGIFY_(a) # a 305#define ECB_STRINGIFY_(a) # a
266#define ECB_STRINGIFY(a) ECB_STRINGIFY_(a) 306#define ECB_STRINGIFY(a) ECB_STRINGIFY_(a)
307#define ECB_STRINGIFY_EXPR(expr) ((expr), ECB_STRINGIFY_ (expr))
267 308
268#define ecb_function_ ecb_inline 309#define ecb_function_ ecb_inline
269 310
270#if ECB_GCC_VERSION(3,1) 311#if ECB_GCC_VERSION(3,1) || ECB_CLANG_VERSION(2,8)
271 #define ecb_attribute(attrlist) __attribute__(attrlist) 312 #define ecb_attribute(attrlist) __attribute__ (attrlist)
313#else
314 #define ecb_attribute(attrlist)
315#endif
316
317#if ECB_GCC_VERSION(3,1) || ECB_CLANG_BUILTIN(__builtin_constant_p)
272 #define ecb_is_constant(expr) __builtin_constant_p (expr) 318 #define ecb_is_constant(expr) __builtin_constant_p (expr)
319#else
320 /* possible C11 impl for integral types
321 typedef struct ecb_is_constant_struct ecb_is_constant_struct;
322 #define ecb_is_constant(expr) _Generic ((1 ? (struct ecb_is_constant_struct *)0 : (void *)((expr) - (expr)), ecb_is_constant_struct *: 0, default: 1)) */
323
324 #define ecb_is_constant(expr) 0
325#endif
326
327#if ECB_GCC_VERSION(3,1) || ECB_CLANG_BUILTIN(__builtin_expect)
273 #define ecb_expect(expr,value) __builtin_expect ((expr),(value)) 328 #define ecb_expect(expr,value) __builtin_expect ((expr),(value))
329#else
330 #define ecb_expect(expr,value) (expr)
331#endif
332
333#if ECB_GCC_VERSION(3,1) || ECB_CLANG_BUILTIN(__builtin_prefetch)
274 #define ecb_prefetch(addr,rw,locality) __builtin_prefetch (addr, rw, locality) 334 #define ecb_prefetch(addr,rw,locality) __builtin_prefetch (addr, rw, locality)
275#else 335#else
276 #define ecb_attribute(attrlist)
277 #define ecb_is_constant(expr) 0
278 #define ecb_expect(expr,value) (expr)
279 #define ecb_prefetch(addr,rw,locality) 336 #define ecb_prefetch(addr,rw,locality)
280#endif 337#endif
281 338
282/* no emulation for ecb_decltype */ 339/* no emulation for ecb_decltype */
283#if ECB_GCC_VERSION(4,5) 340#if ECB_CPP11
341 // older implementations might have problems with decltype(x)::type, work around it
342 template<class T> struct ecb_decltype_t { typedef T type; };
284 #define ecb_decltype(x) __decltype(x) 343 #define ecb_decltype(x) ecb_decltype_t<decltype (x)>::type
285#elif ECB_GCC_VERSION(3,0) 344#elif ECB_GCC_VERSION(3,0) || ECB_CLANG_VERSION(2,8)
286 #define ecb_decltype(x) __typeof(x) 345 #define ecb_decltype(x) __typeof__ (x)
287#endif 346#endif
288 347
348#if _MSC_VER >= 1300
349 #define ecb_deprecated __declspec (deprecated)
350#else
351 #define ecb_deprecated ecb_attribute ((__deprecated__))
352#endif
353
354#if __MSC_VER >= 1500
355 #define ecb_deprecated_message(msg) __declspec (deprecated (msg))
356#elif ECB_GCC_VERSION(4,5)
357 #define ecb_deprecated_message(msg) ecb_attribute ((__deprecated__ (msg))
358#else
359 #define ecb_deprecated_message(msg) ecb_deprecated
360#endif
361
362#if _MSC_VER >= 1400
363 #define ecb_noinline __declspec (noinline)
364#else
289#define ecb_noinline ecb_attribute ((__noinline__)) 365 #define ecb_noinline ecb_attribute ((__noinline__))
366#endif
367
290#define ecb_unused ecb_attribute ((__unused__)) 368#define ecb_unused ecb_attribute ((__unused__))
291#define ecb_const ecb_attribute ((__const__)) 369#define ecb_const ecb_attribute ((__const__))
292#define ecb_pure ecb_attribute ((__pure__)) 370#define ecb_pure ecb_attribute ((__pure__))
293 371
294#if ECB_C11 372#if ECB_C11 || __IBMC_NORETURN
373 /* http://pic.dhe.ibm.com/infocenter/compbg/v121v141/topic/com.ibm.xlcpp121.bg.doc/language_ref/noreturn.html */
295 #define ecb_noreturn _Noreturn 374 #define ecb_noreturn _Noreturn
375#elif ECB_CPP11
376 #define ecb_noreturn [[noreturn]]
377#elif _MSC_VER >= 1200
378 /* http://msdn.microsoft.com/en-us/library/k6ktzx3s.aspx */
379 #define ecb_noreturn __declspec (noreturn)
296#else 380#else
297 #define ecb_noreturn ecb_attribute ((__noreturn__)) 381 #define ecb_noreturn ecb_attribute ((__noreturn__))
298#endif 382#endif
299 383
300#if ECB_GCC_VERSION(4,3) 384#if ECB_GCC_VERSION(4,3)
315/* for compatibility to the rest of the world */ 399/* for compatibility to the rest of the world */
316#define ecb_likely(expr) ecb_expect_true (expr) 400#define ecb_likely(expr) ecb_expect_true (expr)
317#define ecb_unlikely(expr) ecb_expect_false (expr) 401#define ecb_unlikely(expr) ecb_expect_false (expr)
318 402
319/* count trailing zero bits and count # of one bits */ 403/* count trailing zero bits and count # of one bits */
320#if ECB_GCC_VERSION(3,4) 404#if ECB_GCC_VERSION(3,4) \
405 || (ECB_CLANG_BUILTIN(__builtin_clz) && ECB_CLANG_BUILTIN(__builtin_clzll) \
406 && ECB_CLANG_BUILTIN(__builtin_ctz) && ECB_CLANG_BUILTIN(__builtin_ctzll) \
407 && ECB_CLANG_BUILTIN(__builtin_popcount))
321 /* we assume int == 32 bit, long == 32 or 64 bit and long long == 64 bit */ 408 /* we assume int == 32 bit, long == 32 or 64 bit and long long == 64 bit */
322 #define ecb_ld32(x) (__builtin_clz (x) ^ 31) 409 #define ecb_ld32(x) (__builtin_clz (x) ^ 31)
323 #define ecb_ld64(x) (__builtin_clzll (x) ^ 63) 410 #define ecb_ld64(x) (__builtin_clzll (x) ^ 63)
324 #define ecb_ctz32(x) __builtin_ctz (x) 411 #define ecb_ctz32(x) __builtin_ctz (x)
325 #define ecb_ctz64(x) __builtin_ctzll (x) 412 #define ecb_ctz64(x) __builtin_ctzll (x)
326 #define ecb_popcount32(x) __builtin_popcount (x) 413 #define ecb_popcount32(x) __builtin_popcount (x)
327 /* no popcountll */ 414 /* no popcountll */
328#else 415#else
329 ecb_function_ int ecb_ctz32 (uint32_t x) ecb_const; 416 ecb_function_ ecb_const int ecb_ctz32 (uint32_t x);
330 ecb_function_ int 417 ecb_function_ ecb_const int
331 ecb_ctz32 (uint32_t x) 418 ecb_ctz32 (uint32_t x)
332 { 419 {
333 int r = 0; 420 int r = 0;
334 421
335 x &= ~x + 1; /* this isolates the lowest bit */ 422 x &= ~x + 1; /* this isolates the lowest bit */
349#endif 436#endif
350 437
351 return r; 438 return r;
352 } 439 }
353 440
354 ecb_function_ int ecb_ctz64 (uint64_t x) ecb_const; 441 ecb_function_ ecb_const int ecb_ctz64 (uint64_t x);
355 ecb_function_ int 442 ecb_function_ ecb_const int
356 ecb_ctz64 (uint64_t x) 443 ecb_ctz64 (uint64_t x)
357 { 444 {
358 int shift = x & 0xffffffffU ? 0 : 32; 445 int shift = x & 0xffffffffU ? 0 : 32;
359 return ecb_ctz32 (x >> shift) + shift; 446 return ecb_ctz32 (x >> shift) + shift;
360 } 447 }
361 448
362 ecb_function_ int ecb_popcount32 (uint32_t x) ecb_const; 449 ecb_function_ ecb_const int ecb_popcount32 (uint32_t x);
363 ecb_function_ int 450 ecb_function_ ecb_const int
364 ecb_popcount32 (uint32_t x) 451 ecb_popcount32 (uint32_t x)
365 { 452 {
366 x -= (x >> 1) & 0x55555555; 453 x -= (x >> 1) & 0x55555555;
367 x = ((x >> 2) & 0x33333333) + (x & 0x33333333); 454 x = ((x >> 2) & 0x33333333) + (x & 0x33333333);
368 x = ((x >> 4) + x) & 0x0f0f0f0f; 455 x = ((x >> 4) + x) & 0x0f0f0f0f;
369 x *= 0x01010101; 456 x *= 0x01010101;
370 457
371 return x >> 24; 458 return x >> 24;
372 } 459 }
373 460
374 ecb_function_ int ecb_ld32 (uint32_t x) ecb_const; 461 ecb_function_ ecb_const int ecb_ld32 (uint32_t x);
375 ecb_function_ int ecb_ld32 (uint32_t x) 462 ecb_function_ ecb_const int ecb_ld32 (uint32_t x)
376 { 463 {
377 int r = 0; 464 int r = 0;
378 465
379 if (x >> 16) { x >>= 16; r += 16; } 466 if (x >> 16) { x >>= 16; r += 16; }
380 if (x >> 8) { x >>= 8; r += 8; } 467 if (x >> 8) { x >>= 8; r += 8; }
383 if (x >> 1) { r += 1; } 470 if (x >> 1) { r += 1; }
384 471
385 return r; 472 return r;
386 } 473 }
387 474
388 ecb_function_ int ecb_ld64 (uint64_t x) ecb_const; 475 ecb_function_ ecb_const int ecb_ld64 (uint64_t x);
389 ecb_function_ int ecb_ld64 (uint64_t x) 476 ecb_function_ ecb_const int ecb_ld64 (uint64_t x)
390 { 477 {
391 int r = 0; 478 int r = 0;
392 479
393 if (x >> 32) { x >>= 32; r += 32; } 480 if (x >> 32) { x >>= 32; r += 32; }
394 481
395 return r + ecb_ld32 (x); 482 return r + ecb_ld32 (x);
396 } 483 }
397#endif 484#endif
398 485
399ecb_function_ ecb_bool ecb_is_pot32 (uint32_t x) ecb_const; 486ecb_function_ ecb_const ecb_bool ecb_is_pot32 (uint32_t x);
400ecb_function_ ecb_bool ecb_is_pot32 (uint32_t x) { return !(x & (x - 1)); } 487ecb_function_ ecb_const ecb_bool ecb_is_pot32 (uint32_t x) { return !(x & (x - 1)); }
401ecb_function_ ecb_bool ecb_is_pot64 (uint64_t x) ecb_const; 488ecb_function_ ecb_const ecb_bool ecb_is_pot64 (uint64_t x);
402ecb_function_ ecb_bool ecb_is_pot64 (uint64_t x) { return !(x & (x - 1)); } 489ecb_function_ ecb_const ecb_bool ecb_is_pot64 (uint64_t x) { return !(x & (x - 1)); }
403 490
404ecb_function_ uint8_t ecb_bitrev8 (uint8_t x) ecb_const; 491ecb_function_ ecb_const uint8_t ecb_bitrev8 (uint8_t x);
405ecb_function_ uint8_t ecb_bitrev8 (uint8_t x) 492ecb_function_ ecb_const uint8_t ecb_bitrev8 (uint8_t x)
406{ 493{
407 return ( (x * 0x0802U & 0x22110U) 494 return ( (x * 0x0802U & 0x22110U)
408 | (x * 0x8020U & 0x88440U)) * 0x10101U >> 16; 495 | (x * 0x8020U & 0x88440U)) * 0x10101U >> 16;
409} 496}
410 497
411ecb_function_ uint16_t ecb_bitrev16 (uint16_t x) ecb_const; 498ecb_function_ ecb_const uint16_t ecb_bitrev16 (uint16_t x);
412ecb_function_ uint16_t ecb_bitrev16 (uint16_t x) 499ecb_function_ ecb_const uint16_t ecb_bitrev16 (uint16_t x)
413{ 500{
414 x = ((x >> 1) & 0x5555) | ((x & 0x5555) << 1); 501 x = ((x >> 1) & 0x5555) | ((x & 0x5555) << 1);
415 x = ((x >> 2) & 0x3333) | ((x & 0x3333) << 2); 502 x = ((x >> 2) & 0x3333) | ((x & 0x3333) << 2);
416 x = ((x >> 4) & 0x0f0f) | ((x & 0x0f0f) << 4); 503 x = ((x >> 4) & 0x0f0f) | ((x & 0x0f0f) << 4);
417 x = ( x >> 8 ) | ( x << 8); 504 x = ( x >> 8 ) | ( x << 8);
418 505
419 return x; 506 return x;
420} 507}
421 508
422ecb_function_ uint32_t ecb_bitrev32 (uint32_t x) ecb_const; 509ecb_function_ ecb_const uint32_t ecb_bitrev32 (uint32_t x);
423ecb_function_ uint32_t ecb_bitrev32 (uint32_t x) 510ecb_function_ ecb_const uint32_t ecb_bitrev32 (uint32_t x)
424{ 511{
425 x = ((x >> 1) & 0x55555555) | ((x & 0x55555555) << 1); 512 x = ((x >> 1) & 0x55555555) | ((x & 0x55555555) << 1);
426 x = ((x >> 2) & 0x33333333) | ((x & 0x33333333) << 2); 513 x = ((x >> 2) & 0x33333333) | ((x & 0x33333333) << 2);
427 x = ((x >> 4) & 0x0f0f0f0f) | ((x & 0x0f0f0f0f) << 4); 514 x = ((x >> 4) & 0x0f0f0f0f) | ((x & 0x0f0f0f0f) << 4);
428 x = ((x >> 8) & 0x00ff00ff) | ((x & 0x00ff00ff) << 8); 515 x = ((x >> 8) & 0x00ff00ff) | ((x & 0x00ff00ff) << 8);
431 return x; 518 return x;
432} 519}
433 520
434/* popcount64 is only available on 64 bit cpus as gcc builtin */ 521/* popcount64 is only available on 64 bit cpus as gcc builtin */
435/* so for this version we are lazy */ 522/* so for this version we are lazy */
436ecb_function_ int ecb_popcount64 (uint64_t x) ecb_const; 523ecb_function_ ecb_const int ecb_popcount64 (uint64_t x);
437ecb_function_ int 524ecb_function_ ecb_const int
438ecb_popcount64 (uint64_t x) 525ecb_popcount64 (uint64_t x)
439{ 526{
440 return ecb_popcount32 (x) + ecb_popcount32 (x >> 32); 527 return ecb_popcount32 (x) + ecb_popcount32 (x >> 32);
441} 528}
442 529
443ecb_inline uint8_t ecb_rotl8 (uint8_t x, unsigned int count) ecb_const; 530ecb_inline ecb_const uint8_t ecb_rotl8 (uint8_t x, unsigned int count);
444ecb_inline uint8_t ecb_rotr8 (uint8_t x, unsigned int count) ecb_const; 531ecb_inline ecb_const uint8_t ecb_rotr8 (uint8_t x, unsigned int count);
445ecb_inline uint16_t ecb_rotl16 (uint16_t x, unsigned int count) ecb_const; 532ecb_inline ecb_const uint16_t ecb_rotl16 (uint16_t x, unsigned int count);
446ecb_inline uint16_t ecb_rotr16 (uint16_t x, unsigned int count) ecb_const; 533ecb_inline ecb_const uint16_t ecb_rotr16 (uint16_t x, unsigned int count);
447ecb_inline uint32_t ecb_rotl32 (uint32_t x, unsigned int count) ecb_const; 534ecb_inline ecb_const uint32_t ecb_rotl32 (uint32_t x, unsigned int count);
448ecb_inline uint32_t ecb_rotr32 (uint32_t x, unsigned int count) ecb_const; 535ecb_inline ecb_const uint32_t ecb_rotr32 (uint32_t x, unsigned int count);
449ecb_inline uint64_t ecb_rotl64 (uint64_t x, unsigned int count) ecb_const; 536ecb_inline ecb_const uint64_t ecb_rotl64 (uint64_t x, unsigned int count);
450ecb_inline uint64_t ecb_rotr64 (uint64_t x, unsigned int count) ecb_const; 537ecb_inline ecb_const uint64_t ecb_rotr64 (uint64_t x, unsigned int count);
451 538
452ecb_inline uint8_t ecb_rotl8 (uint8_t x, unsigned int count) { return (x >> ( 8 - count)) | (x << count); } 539ecb_inline ecb_const uint8_t ecb_rotl8 (uint8_t x, unsigned int count) { return (x >> ( 8 - count)) | (x << count); }
453ecb_inline uint8_t ecb_rotr8 (uint8_t x, unsigned int count) { return (x << ( 8 - count)) | (x >> count); } 540ecb_inline ecb_const uint8_t ecb_rotr8 (uint8_t x, unsigned int count) { return (x << ( 8 - count)) | (x >> count); }
454ecb_inline uint16_t ecb_rotl16 (uint16_t x, unsigned int count) { return (x >> (16 - count)) | (x << count); } 541ecb_inline ecb_const uint16_t ecb_rotl16 (uint16_t x, unsigned int count) { return (x >> (16 - count)) | (x << count); }
455ecb_inline uint16_t ecb_rotr16 (uint16_t x, unsigned int count) { return (x << (16 - count)) | (x >> count); } 542ecb_inline ecb_const uint16_t ecb_rotr16 (uint16_t x, unsigned int count) { return (x << (16 - count)) | (x >> count); }
456ecb_inline uint32_t ecb_rotl32 (uint32_t x, unsigned int count) { return (x >> (32 - count)) | (x << count); } 543ecb_inline ecb_const uint32_t ecb_rotl32 (uint32_t x, unsigned int count) { return (x >> (32 - count)) | (x << count); }
457ecb_inline uint32_t ecb_rotr32 (uint32_t x, unsigned int count) { return (x << (32 - count)) | (x >> count); } 544ecb_inline ecb_const uint32_t ecb_rotr32 (uint32_t x, unsigned int count) { return (x << (32 - count)) | (x >> count); }
458ecb_inline uint64_t ecb_rotl64 (uint64_t x, unsigned int count) { return (x >> (64 - count)) | (x << count); } 545ecb_inline ecb_const uint64_t ecb_rotl64 (uint64_t x, unsigned int count) { return (x >> (64 - count)) | (x << count); }
459ecb_inline uint64_t ecb_rotr64 (uint64_t x, unsigned int count) { return (x << (64 - count)) | (x >> count); } 546ecb_inline ecb_const uint64_t ecb_rotr64 (uint64_t x, unsigned int count) { return (x << (64 - count)) | (x >> count); }
460 547
461#if ECB_GCC_VERSION(4,3) 548#if ECB_GCC_VERSION(4,3) || (ECB_CLANG_BUILTIN(__builtin_bswap32) && ECB_CLANG_BUILTIN(__builtin_bswap64))
462 #define ecb_bswap16(x) (__builtin_bswap32 (x) >> 16) 549 #define ecb_bswap16(x) (__builtin_bswap32 (x) >> 16)
463 #define ecb_bswap32(x) __builtin_bswap32 (x) 550 #define ecb_bswap32(x) __builtin_bswap32 (x)
464 #define ecb_bswap64(x) __builtin_bswap64 (x) 551 #define ecb_bswap64(x) __builtin_bswap64 (x)
465#else 552#else
466 ecb_function_ uint16_t ecb_bswap16 (uint16_t x) ecb_const; 553 ecb_function_ ecb_const uint16_t ecb_bswap16 (uint16_t x);
467 ecb_function_ uint16_t 554 ecb_function_ ecb_const uint16_t
468 ecb_bswap16 (uint16_t x) 555 ecb_bswap16 (uint16_t x)
469 { 556 {
470 return ecb_rotl16 (x, 8); 557 return ecb_rotl16 (x, 8);
471 } 558 }
472 559
473 ecb_function_ uint32_t ecb_bswap32 (uint32_t x) ecb_const; 560 ecb_function_ ecb_const uint32_t ecb_bswap32 (uint32_t x);
474 ecb_function_ uint32_t 561 ecb_function_ ecb_const uint32_t
475 ecb_bswap32 (uint32_t x) 562 ecb_bswap32 (uint32_t x)
476 { 563 {
477 return (((uint32_t)ecb_bswap16 (x)) << 16) | ecb_bswap16 (x >> 16); 564 return (((uint32_t)ecb_bswap16 (x)) << 16) | ecb_bswap16 (x >> 16);
478 } 565 }
479 566
480 ecb_function_ uint64_t ecb_bswap64 (uint64_t x) ecb_const; 567 ecb_function_ ecb_const uint64_t ecb_bswap64 (uint64_t x);
481 ecb_function_ uint64_t 568 ecb_function_ ecb_const uint64_t
482 ecb_bswap64 (uint64_t x) 569 ecb_bswap64 (uint64_t x)
483 { 570 {
484 return (((uint64_t)ecb_bswap32 (x)) << 32) | ecb_bswap32 (x >> 32); 571 return (((uint64_t)ecb_bswap32 (x)) << 32) | ecb_bswap32 (x >> 32);
485 } 572 }
486#endif 573#endif
487 574
488#if ECB_GCC_VERSION(4,5) 575#if ECB_GCC_VERSION(4,5) || ECB_CLANG_BUILTIN(__builtin_unreachable)
489 #define ecb_unreachable() __builtin_unreachable () 576 #define ecb_unreachable() __builtin_unreachable ()
490#else 577#else
491 /* this seems to work fine, but gcc always emits a warning for it :/ */ 578 /* this seems to work fine, but gcc always emits a warning for it :/ */
492 ecb_inline void ecb_unreachable (void) ecb_noreturn; 579 ecb_inline ecb_noreturn void ecb_unreachable (void);
493 ecb_inline void ecb_unreachable (void) { } 580 ecb_inline ecb_noreturn void ecb_unreachable (void) { }
494#endif 581#endif
495 582
496/* try to tell the compiler that some condition is definitely true */ 583/* try to tell the compiler that some condition is definitely true */
497#define ecb_assume(cond) if (!(cond)) ecb_unreachable (); else 0 584#define ecb_assume(cond) if (!(cond)) ecb_unreachable (); else 0
498 585
499ecb_inline unsigned char ecb_byteorder_helper (void) ecb_const; 586ecb_inline ecb_const unsigned char ecb_byteorder_helper (void);
500ecb_inline unsigned char 587ecb_inline ecb_const unsigned char
501ecb_byteorder_helper (void) 588ecb_byteorder_helper (void)
502{ 589{
503 /* the union code still generates code under pressure in gcc, */ 590 /* the union code still generates code under pressure in gcc, */
504 /* but less than using pointers, and always seems to */ 591 /* but less than using pointers, and always seems to */
505 /* successfully return a constant. */ 592 /* successfully return a constant. */
506 /* the reason why we have this horrible preprocessor mess */ 593 /* the reason why we have this horrible preprocessor mess */
507 /* is to avoid it in all cases, at least on common architectures */ 594 /* is to avoid it in all cases, at least on common architectures */
508 /* or when using a recent enough gcc version (>= 4.6) */ 595 /* or when using a recent enough gcc version (>= 4.6) */
509#if __i386 || __i386__ || _M_X86 || __amd64 || __amd64__ || _M_X64 596#if ((__i386 || __i386__) && !__VOS__) || _M_X86 || __amd64 || __amd64__ || _M_X64
510 return 0x44; 597 return 0x44;
511#elif __BYTE_ORDER__ && __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__ 598#elif __BYTE_ORDER__ && __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__
512 return 0x44; 599 return 0x44;
513#elif __BYTE_ORDER__ && __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__ 600#elif __BYTE_ORDER__ && __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__
514 return 0x11; 601 return 0x11;
520 } u = { 0x11223344 }; 607 } u = { 0x11223344 };
521 return u.c; 608 return u.c;
522#endif 609#endif
523} 610}
524 611
525ecb_inline ecb_bool ecb_big_endian (void) ecb_const; 612ecb_inline ecb_const ecb_bool ecb_big_endian (void);
526ecb_inline ecb_bool ecb_big_endian (void) { return ecb_byteorder_helper () == 0x11; } 613ecb_inline ecb_const ecb_bool ecb_big_endian (void) { return ecb_byteorder_helper () == 0x11; }
527ecb_inline ecb_bool ecb_little_endian (void) ecb_const; 614ecb_inline ecb_const ecb_bool ecb_little_endian (void);
528ecb_inline ecb_bool ecb_little_endian (void) { return ecb_byteorder_helper () == 0x44; } 615ecb_inline ecb_const ecb_bool ecb_little_endian (void) { return ecb_byteorder_helper () == 0x44; }
529 616
530#if ECB_GCC_VERSION(3,0) || ECB_C99 617#if ECB_GCC_VERSION(3,0) || ECB_C99
531 #define ecb_mod(m,n) ((m) % (n) + ((m) % (n) < 0 ? (n) : 0)) 618 #define ecb_mod(m,n) ((m) % (n) + ((m) % (n) < 0 ? (n) : 0))
532#else 619#else
533 #define ecb_mod(m,n) ((m) < 0 ? ((n) - 1 - ((-1 - (m)) % (n))) : ((m) % (n))) 620 #define ecb_mod(m,n) ((m) < 0 ? ((n) - 1 - ((-1 - (m)) % (n))) : ((m) % (n)))
534#endif 621#endif
535 622
536#if __cplusplus 623#if ECB_CPP
537 template<typename T> 624 template<typename T>
538 static inline T ecb_div_rd (T val, T div) 625 static inline T ecb_div_rd (T val, T div)
539 { 626 {
540 return val < 0 ? - ((-val + div - 1) / div) : (val ) / div; 627 return val < 0 ? - ((-val + div - 1) / div) : (val ) / div;
541 } 628 }
567/* the only noteworthy exception is ancient armle, which uses order 43218765 */ 654/* the only noteworthy exception is ancient armle, which uses order 43218765 */
568#if 0 \ 655#if 0 \
569 || __i386 || __i386__ \ 656 || __i386 || __i386__ \
570 || __amd64 || __amd64__ || __x86_64 || __x86_64__ \ 657 || __amd64 || __amd64__ || __x86_64 || __x86_64__ \
571 || __powerpc__ || __ppc__ || __powerpc64__ || __ppc64__ \ 658 || __powerpc__ || __ppc__ || __powerpc64__ || __ppc64__ \
572 || defined __arm__ && defined __ARM_EABI__ \
573 || defined __s390__ || defined __s390x__ \ 659 || defined __s390__ || defined __s390x__ \
574 || defined __mips__ \ 660 || defined __mips__ \
575 || defined __alpha__ \ 661 || defined __alpha__ \
576 || defined __hppa__ \ 662 || defined __hppa__ \
577 || defined __ia64__ \ 663 || defined __ia64__ \
578 || defined __m68k__ \ 664 || defined __m68k__ \
579 || defined __m88k__ \ 665 || defined __m88k__ \
580 || defined __sh__ \ 666 || defined __sh__ \
581 || defined _M_IX86 || defined _M_AMD64 || defined _M_IA64 667 || defined _M_IX86 || defined _M_AMD64 || defined _M_IA64 \
668 || (defined __arm__ && (defined __ARM_EABI__ || defined __EABI__ || defined __VFP_FP__ || defined _WIN32_WCE || defined __ANDROID__)) \
669 || defined __aarch64__
582 #define ECB_STDFP 1 670 #define ECB_STDFP 1
583 #include <string.h> /* for memcpy */ 671 #include <string.h> /* for memcpy */
584#else 672#else
585 #define ECB_STDFP 0 673 #define ECB_STDFP 0
586#endif 674#endif
600 #define ECB_NAN NAN 688 #define ECB_NAN NAN
601 #else 689 #else
602 #define ECB_NAN ECB_INFINITY 690 #define ECB_NAN ECB_INFINITY
603 #endif 691 #endif
604 692
693 #if ECB_C99 || _XOPEN_VERSION >= 600 || _POSIX_VERSION >= 200112L
694 #define ecb_ldexpf(x,e) ldexpf ((x), (e))
695 #else
696 #define ecb_ldexpf(x,e) (float) ldexp ((x), (e))
697 #endif
698
605 /* converts an ieee half/binary16 to a float */ 699 /* converts an ieee half/binary16 to a float */
606 ecb_function_ float ecb_binary16_to_float (uint16_t x) ecb_const; 700 ecb_function_ ecb_const float ecb_binary16_to_float (uint16_t x);
607 ecb_function_ float 701 ecb_function_ ecb_const float
608 ecb_binary16_to_float (uint16_t x) 702 ecb_binary16_to_float (uint16_t x)
609 { 703 {
610 int e = (x >> 10) & 0x1f; 704 int e = (x >> 10) & 0x1f;
611 int m = x & 0x3ff; 705 int m = x & 0x3ff;
612 float r; 706 float r;
613 707
614 if (!e ) r = ldexpf (m , -24); 708 if (!e ) r = ecb_ldexpf (m , -24);
615 else if (e != 31) r = ldexpf (m + 0x400, e - 25); 709 else if (e != 31) r = ecb_ldexpf (m + 0x400, e - 25);
616 else if (m ) r = ECB_NAN; 710 else if (m ) r = ECB_NAN;
617 else r = ECB_INFINITY; 711 else r = ECB_INFINITY;
618 712
619 return x & 0x8000 ? -r : r; 713 return x & 0x8000 ? -r : r;
620 } 714 }
621 715
622 /* convert a float to ieee single/binary32 */ 716 /* convert a float to ieee single/binary32 */
623 ecb_function_ uint32_t ecb_float_to_binary32 (float x) ecb_const; 717 ecb_function_ ecb_const uint32_t ecb_float_to_binary32 (float x);
624 ecb_function_ uint32_t 718 ecb_function_ ecb_const uint32_t
625 ecb_float_to_binary32 (float x) 719 ecb_float_to_binary32 (float x)
626 { 720 {
627 uint32_t r; 721 uint32_t r;
628 722
629 #if ECB_STDFP 723 #if ECB_STDFP
658 752
659 return r; 753 return r;
660 } 754 }
661 755
662 /* converts an ieee single/binary32 to a float */ 756 /* converts an ieee single/binary32 to a float */
663 ecb_function_ float ecb_binary32_to_float (uint32_t x) ecb_const; 757 ecb_function_ ecb_const float ecb_binary32_to_float (uint32_t x);
664 ecb_function_ float 758 ecb_function_ ecb_const float
665 ecb_binary32_to_float (uint32_t x) 759 ecb_binary32_to_float (uint32_t x)
666 { 760 {
667 float r; 761 float r;
668 762
669 #if ECB_STDFP 763 #if ECB_STDFP
679 x |= 0x800000U; 773 x |= 0x800000U;
680 else 774 else
681 e = 1; 775 e = 1;
682 776
683 /* we distrust ldexpf a bit and do the 2**-24 scaling by an extra multiply */ 777 /* we distrust ldexpf a bit and do the 2**-24 scaling by an extra multiply */
684 r = ldexpf (x * (0.5f / 0x800000U), e - 126); 778 r = ecb_ldexpf (x * (0.5f / 0x800000U), e - 126);
685 779
686 r = neg ? -r : r; 780 r = neg ? -r : r;
687 #endif 781 #endif
688 782
689 return r; 783 return r;
690 } 784 }
691 785
692 /* convert a double to ieee double/binary64 */ 786 /* convert a double to ieee double/binary64 */
693 ecb_function_ uint64_t ecb_double_to_binary64 (double x) ecb_const; 787 ecb_function_ ecb_const uint64_t ecb_double_to_binary64 (double x);
694 ecb_function_ uint64_t 788 ecb_function_ ecb_const uint64_t
695 ecb_double_to_binary64 (double x) 789 ecb_double_to_binary64 (double x)
696 { 790 {
697 uint64_t r; 791 uint64_t r;
698 792
699 #if ECB_STDFP 793 #if ECB_STDFP
728 822
729 return r; 823 return r;
730 } 824 }
731 825
732 /* converts an ieee double/binary64 to a double */ 826 /* converts an ieee double/binary64 to a double */
733 ecb_function_ double ecb_binary64_to_double (uint64_t x) ecb_const; 827 ecb_function_ ecb_const double ecb_binary64_to_double (uint64_t x);
734 ecb_function_ double 828 ecb_function_ ecb_const double
735 ecb_binary64_to_double (uint64_t x) 829 ecb_binary64_to_double (uint64_t x)
736 { 830 {
737 double r; 831 double r;
738 832
739 #if ECB_STDFP 833 #if ECB_STDFP

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines