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

Comparing libecb/ecb.h (file contents):
Revision 1.121 by root, Sat Oct 26 09:16:15 2013 UTC vs.
Revision 1.153 by root, Wed Feb 11 19:44:46 2015 UTC

1/* 1/*
2 * libecb - http://software.schmorp.de/pkg/libecb 2 * libecb - http://software.schmorp.de/pkg/libecb
3 * 3 *
4 * Copyright (©) 2009-2012 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;
64 #define ECB_PTRSIZE 4 75 #define ECB_PTRSIZE 4
65 #endif 76 #endif
66#endif 77#endif
67 78
68/* work around x32 idiocy by defining proper macros */ 79/* work around x32 idiocy by defining proper macros */
69#if __x86_64 || _M_AMD64 80#if __amd64 || __x86_64 || _M_AMD64 || _M_X64
70 #if _ILP32 81 #if _ILP32
71 #define ECB_AMD64_X32 1 82 #define ECB_AMD64_X32 1
72 #else 83 #else
73 #define ECB_AMD64 1 84 #define ECB_AMD64 1
74 #endif 85 #endif
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__ 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__
144 #define ECB_MEMORY_FENCE __asm__ __volatile__ ("bcr 15,0" : : : "memory") 177 #define ECB_MEMORY_FENCE __asm__ __volatile__ ("bcr 15,0" : : : "memory")
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
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)
267 307
268#define ecb_function_ ecb_inline 308#define ecb_function_ ecb_inline
269 309
270#if ECB_GCC_VERSION(3,1) 310#if ECB_GCC_VERSION(3,1) || ECB_CLANG_VERSION(2,8)
271 #define ecb_attribute(attrlist) __attribute__(attrlist) 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)
272 #define ecb_is_constant(expr) __builtin_constant_p (expr) 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)
273 #define ecb_expect(expr,value) __builtin_expect ((expr),(value)) 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)
274 #define ecb_prefetch(addr,rw,locality) __builtin_prefetch (addr, rw, locality) 333 #define ecb_prefetch(addr,rw,locality) __builtin_prefetch (addr, rw, locality)
275#else 334#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) 335 #define ecb_prefetch(addr,rw,locality)
280#endif 336#endif
281 337
282/* no emulation for ecb_decltype */ 338/* no emulation for ecb_decltype */
283#if ECB_GCC_VERSION(4,5) 339#if ECB_CPP11
340 // older implementations might have problems with decltype(x)::type, work around it
341 template<class T> struct ecb_decltype_t { typedef T type; };
284 #define ecb_decltype(x) __decltype(x) 342 #define ecb_decltype(x) ecb_decltype_t<decltype (x)>::type
285#elif ECB_GCC_VERSION(3,0) 343#elif ECB_GCC_VERSION(3,0) || ECB_CLANG_VERSION(2,8)
286 #define ecb_decltype(x) __typeof(x) 344 #define ecb_decltype(x) __typeof__ (x)
345#endif
346
347#if _MSC_VER >= 1300
348 #define ecb_deprecated __declspec (deprecated)
349#else
350 #define ecb_deprecated ecb_attribute ((__deprecated__))
287#endif 351#endif
288 352
289#define ecb_noinline ecb_attribute ((__noinline__)) 353#define ecb_noinline ecb_attribute ((__noinline__))
290#define ecb_unused ecb_attribute ((__unused__)) 354#define ecb_unused ecb_attribute ((__unused__))
291#define ecb_const ecb_attribute ((__const__)) 355#define ecb_const ecb_attribute ((__const__))
292#define ecb_pure ecb_attribute ((__pure__)) 356#define ecb_pure ecb_attribute ((__pure__))
293 357
294#if ECB_C11 358/* TODO http://msdn.microsoft.com/en-us/library/k6ktzx3s.aspx __declspec(noreturn) */
359#if ECB_C11 || __IBMC_NORETURN
360 /* http://pic.dhe.ibm.com/infocenter/compbg/v121v141/topic/com.ibm.xlcpp121.bg.doc/language_ref/noreturn.html */
295 #define ecb_noreturn _Noreturn 361 #define ecb_noreturn _Noreturn
362#elif ECB_CPP11
363 #define ecb_noreturn [[noreturn]]
364#elif _MSC_VER >= 1200
365 #define ecb_noreturn __declspec (noreturn)
296#else 366#else
297 #define ecb_noreturn ecb_attribute ((__noreturn__)) 367 #define ecb_noreturn ecb_attribute ((__noreturn__))
298#endif 368#endif
299 369
300#if ECB_GCC_VERSION(4,3) 370#if ECB_GCC_VERSION(4,3)
315/* for compatibility to the rest of the world */ 385/* for compatibility to the rest of the world */
316#define ecb_likely(expr) ecb_expect_true (expr) 386#define ecb_likely(expr) ecb_expect_true (expr)
317#define ecb_unlikely(expr) ecb_expect_false (expr) 387#define ecb_unlikely(expr) ecb_expect_false (expr)
318 388
319/* count trailing zero bits and count # of one bits */ 389/* count trailing zero bits and count # of one bits */
320#if ECB_GCC_VERSION(3,4) 390#if ECB_GCC_VERSION(3,4) \
391 || (ECB_CLANG_BUILTIN(__builtin_clz) && ECB_CLANG_BUILTIN(__builtin_clzll) \
392 && ECB_CLANG_BUILTIN(__builtin_ctz) && ECB_CLANG_BUILTIN(__builtin_ctzll) \
393 && ECB_CLANG_BUILTIN(__builtin_popcount))
321 /* we assume int == 32 bit, long == 32 or 64 bit and long long == 64 bit */ 394 /* we assume int == 32 bit, long == 32 or 64 bit and long long == 64 bit */
322 #define ecb_ld32(x) (__builtin_clz (x) ^ 31) 395 #define ecb_ld32(x) (__builtin_clz (x) ^ 31)
323 #define ecb_ld64(x) (__builtin_clzll (x) ^ 63) 396 #define ecb_ld64(x) (__builtin_clzll (x) ^ 63)
324 #define ecb_ctz32(x) __builtin_ctz (x) 397 #define ecb_ctz32(x) __builtin_ctz (x)
325 #define ecb_ctz64(x) __builtin_ctzll (x) 398 #define ecb_ctz64(x) __builtin_ctzll (x)
326 #define ecb_popcount32(x) __builtin_popcount (x) 399 #define ecb_popcount32(x) __builtin_popcount (x)
327 /* no popcountll */ 400 /* no popcountll */
328#else 401#else
329 ecb_function_ int ecb_ctz32 (uint32_t x) ecb_const; 402 ecb_function_ ecb_const int ecb_ctz32 (uint32_t x);
330 ecb_function_ int 403 ecb_function_ ecb_const int
331 ecb_ctz32 (uint32_t x) 404 ecb_ctz32 (uint32_t x)
332 { 405 {
333 int r = 0; 406 int r = 0;
334 407
335 x &= ~x + 1; /* this isolates the lowest bit */ 408 x &= ~x + 1; /* this isolates the lowest bit */
349#endif 422#endif
350 423
351 return r; 424 return r;
352 } 425 }
353 426
354 ecb_function_ int ecb_ctz64 (uint64_t x) ecb_const; 427 ecb_function_ ecb_const int ecb_ctz64 (uint64_t x);
355 ecb_function_ int 428 ecb_function_ ecb_const int
356 ecb_ctz64 (uint64_t x) 429 ecb_ctz64 (uint64_t x)
357 { 430 {
358 int shift = x & 0xffffffffU ? 0 : 32; 431 int shift = x & 0xffffffffU ? 0 : 32;
359 return ecb_ctz32 (x >> shift) + shift; 432 return ecb_ctz32 (x >> shift) + shift;
360 } 433 }
361 434
362 ecb_function_ int ecb_popcount32 (uint32_t x) ecb_const; 435 ecb_function_ ecb_const int ecb_popcount32 (uint32_t x);
363 ecb_function_ int 436 ecb_function_ ecb_const int
364 ecb_popcount32 (uint32_t x) 437 ecb_popcount32 (uint32_t x)
365 { 438 {
366 x -= (x >> 1) & 0x55555555; 439 x -= (x >> 1) & 0x55555555;
367 x = ((x >> 2) & 0x33333333) + (x & 0x33333333); 440 x = ((x >> 2) & 0x33333333) + (x & 0x33333333);
368 x = ((x >> 4) + x) & 0x0f0f0f0f; 441 x = ((x >> 4) + x) & 0x0f0f0f0f;
369 x *= 0x01010101; 442 x *= 0x01010101;
370 443
371 return x >> 24; 444 return x >> 24;
372 } 445 }
373 446
374 ecb_function_ int ecb_ld32 (uint32_t x) ecb_const; 447 ecb_function_ ecb_const int ecb_ld32 (uint32_t x);
375 ecb_function_ int ecb_ld32 (uint32_t x) 448 ecb_function_ ecb_const int ecb_ld32 (uint32_t x)
376 { 449 {
377 int r = 0; 450 int r = 0;
378 451
379 if (x >> 16) { x >>= 16; r += 16; } 452 if (x >> 16) { x >>= 16; r += 16; }
380 if (x >> 8) { x >>= 8; r += 8; } 453 if (x >> 8) { x >>= 8; r += 8; }
383 if (x >> 1) { r += 1; } 456 if (x >> 1) { r += 1; }
384 457
385 return r; 458 return r;
386 } 459 }
387 460
388 ecb_function_ int ecb_ld64 (uint64_t x) ecb_const; 461 ecb_function_ ecb_const int ecb_ld64 (uint64_t x);
389 ecb_function_ int ecb_ld64 (uint64_t x) 462 ecb_function_ ecb_const int ecb_ld64 (uint64_t x)
390 { 463 {
391 int r = 0; 464 int r = 0;
392 465
393 if (x >> 32) { x >>= 32; r += 32; } 466 if (x >> 32) { x >>= 32; r += 32; }
394 467
395 return r + ecb_ld32 (x); 468 return r + ecb_ld32 (x);
396 } 469 }
397#endif 470#endif
398 471
399ecb_function_ ecb_bool ecb_is_pot32 (uint32_t x) ecb_const; 472ecb_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)); } 473ecb_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; 474ecb_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)); } 475ecb_function_ ecb_const ecb_bool ecb_is_pot64 (uint64_t x) { return !(x & (x - 1)); }
403 476
404ecb_function_ uint8_t ecb_bitrev8 (uint8_t x) ecb_const; 477ecb_function_ ecb_const uint8_t ecb_bitrev8 (uint8_t x);
405ecb_function_ uint8_t ecb_bitrev8 (uint8_t x) 478ecb_function_ ecb_const uint8_t ecb_bitrev8 (uint8_t x)
406{ 479{
407 return ( (x * 0x0802U & 0x22110U) 480 return ( (x * 0x0802U & 0x22110U)
408 | (x * 0x8020U & 0x88440U)) * 0x10101U >> 16; 481 | (x * 0x8020U & 0x88440U)) * 0x10101U >> 16;
409} 482}
410 483
411ecb_function_ uint16_t ecb_bitrev16 (uint16_t x) ecb_const; 484ecb_function_ ecb_const uint16_t ecb_bitrev16 (uint16_t x);
412ecb_function_ uint16_t ecb_bitrev16 (uint16_t x) 485ecb_function_ ecb_const uint16_t ecb_bitrev16 (uint16_t x)
413{ 486{
414 x = ((x >> 1) & 0x5555) | ((x & 0x5555) << 1); 487 x = ((x >> 1) & 0x5555) | ((x & 0x5555) << 1);
415 x = ((x >> 2) & 0x3333) | ((x & 0x3333) << 2); 488 x = ((x >> 2) & 0x3333) | ((x & 0x3333) << 2);
416 x = ((x >> 4) & 0x0f0f) | ((x & 0x0f0f) << 4); 489 x = ((x >> 4) & 0x0f0f) | ((x & 0x0f0f) << 4);
417 x = ( x >> 8 ) | ( x << 8); 490 x = ( x >> 8 ) | ( x << 8);
418 491
419 return x; 492 return x;
420} 493}
421 494
422ecb_function_ uint32_t ecb_bitrev32 (uint32_t x) ecb_const; 495ecb_function_ ecb_const uint32_t ecb_bitrev32 (uint32_t x);
423ecb_function_ uint32_t ecb_bitrev32 (uint32_t x) 496ecb_function_ ecb_const uint32_t ecb_bitrev32 (uint32_t x)
424{ 497{
425 x = ((x >> 1) & 0x55555555) | ((x & 0x55555555) << 1); 498 x = ((x >> 1) & 0x55555555) | ((x & 0x55555555) << 1);
426 x = ((x >> 2) & 0x33333333) | ((x & 0x33333333) << 2); 499 x = ((x >> 2) & 0x33333333) | ((x & 0x33333333) << 2);
427 x = ((x >> 4) & 0x0f0f0f0f) | ((x & 0x0f0f0f0f) << 4); 500 x = ((x >> 4) & 0x0f0f0f0f) | ((x & 0x0f0f0f0f) << 4);
428 x = ((x >> 8) & 0x00ff00ff) | ((x & 0x00ff00ff) << 8); 501 x = ((x >> 8) & 0x00ff00ff) | ((x & 0x00ff00ff) << 8);
431 return x; 504 return x;
432} 505}
433 506
434/* popcount64 is only available on 64 bit cpus as gcc builtin */ 507/* popcount64 is only available on 64 bit cpus as gcc builtin */
435/* so for this version we are lazy */ 508/* so for this version we are lazy */
436ecb_function_ int ecb_popcount64 (uint64_t x) ecb_const; 509ecb_function_ ecb_const int ecb_popcount64 (uint64_t x);
437ecb_function_ int 510ecb_function_ ecb_const int
438ecb_popcount64 (uint64_t x) 511ecb_popcount64 (uint64_t x)
439{ 512{
440 return ecb_popcount32 (x) + ecb_popcount32 (x >> 32); 513 return ecb_popcount32 (x) + ecb_popcount32 (x >> 32);
441} 514}
442 515
443ecb_inline uint8_t ecb_rotl8 (uint8_t x, unsigned int count) ecb_const; 516ecb_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; 517ecb_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; 518ecb_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; 519ecb_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; 520ecb_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; 521ecb_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; 522ecb_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; 523ecb_inline ecb_const uint64_t ecb_rotr64 (uint64_t x, unsigned int count);
451 524
452ecb_inline uint8_t ecb_rotl8 (uint8_t x, unsigned int count) { return (x >> ( 8 - count)) | (x << count); } 525ecb_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); } 526ecb_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); } 527ecb_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); } 528ecb_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); } 529ecb_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); } 530ecb_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); } 531ecb_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); } 532ecb_inline ecb_const uint64_t ecb_rotr64 (uint64_t x, unsigned int count) { return (x << (64 - count)) | (x >> count); }
460 533
461#if ECB_GCC_VERSION(4,3) 534#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) 535 #define ecb_bswap16(x) (__builtin_bswap32 (x) >> 16)
463 #define ecb_bswap32(x) __builtin_bswap32 (x) 536 #define ecb_bswap32(x) __builtin_bswap32 (x)
464 #define ecb_bswap64(x) __builtin_bswap64 (x) 537 #define ecb_bswap64(x) __builtin_bswap64 (x)
465#else 538#else
466 ecb_function_ uint16_t ecb_bswap16 (uint16_t x) ecb_const; 539 ecb_function_ ecb_const uint16_t ecb_bswap16 (uint16_t x);
467 ecb_function_ uint16_t 540 ecb_function_ ecb_const uint16_t
468 ecb_bswap16 (uint16_t x) 541 ecb_bswap16 (uint16_t x)
469 { 542 {
470 return ecb_rotl16 (x, 8); 543 return ecb_rotl16 (x, 8);
471 } 544 }
472 545
473 ecb_function_ uint32_t ecb_bswap32 (uint32_t x) ecb_const; 546 ecb_function_ ecb_const uint32_t ecb_bswap32 (uint32_t x);
474 ecb_function_ uint32_t 547 ecb_function_ ecb_const uint32_t
475 ecb_bswap32 (uint32_t x) 548 ecb_bswap32 (uint32_t x)
476 { 549 {
477 return (((uint32_t)ecb_bswap16 (x)) << 16) | ecb_bswap16 (x >> 16); 550 return (((uint32_t)ecb_bswap16 (x)) << 16) | ecb_bswap16 (x >> 16);
478 } 551 }
479 552
480 ecb_function_ uint64_t ecb_bswap64 (uint64_t x) ecb_const; 553 ecb_function_ ecb_const uint64_t ecb_bswap64 (uint64_t x);
481 ecb_function_ uint64_t 554 ecb_function_ ecb_const uint64_t
482 ecb_bswap64 (uint64_t x) 555 ecb_bswap64 (uint64_t x)
483 { 556 {
484 return (((uint64_t)ecb_bswap32 (x)) << 32) | ecb_bswap32 (x >> 32); 557 return (((uint64_t)ecb_bswap32 (x)) << 32) | ecb_bswap32 (x >> 32);
485 } 558 }
486#endif 559#endif
487 560
488#if ECB_GCC_VERSION(4,5) 561#if ECB_GCC_VERSION(4,5) || ECB_CLANG_BUILTIN(__builtin_unreachable)
489 #define ecb_unreachable() __builtin_unreachable () 562 #define ecb_unreachable() __builtin_unreachable ()
490#else 563#else
491 /* this seems to work fine, but gcc always emits a warning for it :/ */ 564 /* this seems to work fine, but gcc always emits a warning for it :/ */
492 ecb_inline void ecb_unreachable (void) ecb_noreturn; 565 ecb_inline ecb_noreturn void ecb_unreachable (void);
493 ecb_inline void ecb_unreachable (void) { } 566 ecb_inline ecb_noreturn void ecb_unreachable (void) { }
494#endif 567#endif
495 568
496/* try to tell the compiler that some condition is definitely true */ 569/* try to tell the compiler that some condition is definitely true */
497#define ecb_assume(cond) if (!(cond)) ecb_unreachable (); else 0 570#define ecb_assume(cond) if (!(cond)) ecb_unreachable (); else 0
498 571
499ecb_inline unsigned char ecb_byteorder_helper (void) ecb_const; 572ecb_inline ecb_const unsigned char ecb_byteorder_helper (void);
500ecb_inline unsigned char 573ecb_inline ecb_const unsigned char
501ecb_byteorder_helper (void) 574ecb_byteorder_helper (void)
502{ 575{
503 /* the union code still generates code under pressure in gcc, */ 576 /* the union code still generates code under pressure in gcc, */
504 /* but less than using pointers, and always seems to */ 577 /* but less than using pointers, and always seems to */
505 /* successfully return a constant. */ 578 /* successfully return a constant. */
520 } u = { 0x11223344 }; 593 } u = { 0x11223344 };
521 return u.c; 594 return u.c;
522#endif 595#endif
523} 596}
524 597
525ecb_inline ecb_bool ecb_big_endian (void) ecb_const; 598ecb_inline ecb_const ecb_bool ecb_big_endian (void);
526ecb_inline ecb_bool ecb_big_endian (void) { return ecb_byteorder_helper () == 0x11; } 599ecb_inline ecb_const ecb_bool ecb_big_endian (void) { return ecb_byteorder_helper () == 0x11; }
527ecb_inline ecb_bool ecb_little_endian (void) ecb_const; 600ecb_inline ecb_const ecb_bool ecb_little_endian (void);
528ecb_inline ecb_bool ecb_little_endian (void) { return ecb_byteorder_helper () == 0x44; } 601ecb_inline ecb_const ecb_bool ecb_little_endian (void) { return ecb_byteorder_helper () == 0x44; }
529 602
530#if ECB_GCC_VERSION(3,0) || ECB_C99 603#if ECB_GCC_VERSION(3,0) || ECB_C99
531 #define ecb_mod(m,n) ((m) % (n) + ((m) % (n) < 0 ? (n) : 0)) 604 #define ecb_mod(m,n) ((m) % (n) + ((m) % (n) < 0 ? (n) : 0))
532#else 605#else
533 #define ecb_mod(m,n) ((m) < 0 ? ((n) - 1 - ((-1 - (m)) % (n))) : ((m) % (n))) 606 #define ecb_mod(m,n) ((m) < 0 ? ((n) - 1 - ((-1 - (m)) % (n))) : ((m) % (n)))
534#endif 607#endif
535 608
536#if __cplusplus 609#if ECB_CPP
537 template<typename T> 610 template<typename T>
538 static inline T ecb_div_rd (T val, T div) 611 static inline T ecb_div_rd (T val, T div)
539 { 612 {
540 return val < 0 ? - ((-val + div - 1) / div) : (val ) / div; 613 return val < 0 ? - ((-val + div - 1) / div) : (val ) / div;
541 } 614 }
567/* the only noteworthy exception is ancient armle, which uses order 43218765 */ 640/* the only noteworthy exception is ancient armle, which uses order 43218765 */
568#if 0 \ 641#if 0 \
569 || __i386 || __i386__ \ 642 || __i386 || __i386__ \
570 || __amd64 || __amd64__ || __x86_64 || __x86_64__ \ 643 || __amd64 || __amd64__ || __x86_64 || __x86_64__ \
571 || __powerpc__ || __ppc__ || __powerpc64__ || __ppc64__ \ 644 || __powerpc__ || __ppc__ || __powerpc64__ || __ppc64__ \
572 || defined __arm__ && defined __ARM_EABI__ \
573 || defined __s390__ || defined __s390x__ \ 645 || defined __s390__ || defined __s390x__ \
574 || defined __mips__ \ 646 || defined __mips__ \
575 || defined __alpha__ \ 647 || defined __alpha__ \
576 || defined __hppa__ \ 648 || defined __hppa__ \
577 || defined __ia64__ \ 649 || defined __ia64__ \
578 || defined __m68k__ \ 650 || defined __m68k__ \
579 || defined __m88k__ \ 651 || defined __m88k__ \
580 || defined __sh__ \ 652 || defined __sh__ \
581 || defined _M_IX86 || defined _M_AMD64 || defined _M_IA64 653 || defined _M_IX86 || defined _M_AMD64 || defined _M_IA64 \
654 || (defined __arm__ && (defined __ARM_EABI__ || defined __EABI__ || defined __VFP_FP__ || defined _WIN32_WCE || defined __ANDROID__)) \
655 || defined __aarch64__
582 #define ECB_STDFP 1 656 #define ECB_STDFP 1
583 #include <string.h> /* for memcpy */ 657 #include <string.h> /* for memcpy */
584#else 658#else
585 #define ECB_STDFP 0 659 #define ECB_STDFP 0
586#endif 660#endif
587 661
588#ifndef ECB_NO_LIBM 662#ifndef ECB_NO_LIBM
589 663
590 #include <math.h> /* for frexp*, ldexp*, INFINITY, NAN */ 664 #include <math.h> /* for frexp*, ldexp*, INFINITY, NAN */
591 665
666 /* only the oldest of old doesn't have this one. solaris. */
667 #ifdef INFINITY
668 #define ECB_INFINITY INFINITY
669 #else
670 #define ECB_INFINITY HUGE_VAL
671 #endif
672
592 #ifdef NEN 673 #ifdef NAN
593 #define ECB_NAN NAN 674 #define ECB_NAN NAN
594 #else 675 #else
595 #define ECB_NAN INFINITY 676 #define ECB_NAN ECB_INFINITY
677 #endif
678
679 #if ECB_C99 || _XOPEN_VERSION >= 600 || _POSIX_VERSION >= 200112L
680 #define ecb_ldexpf(x,e) ldexpf ((x), (e))
681 #else
682 #define ecb_ldexpf(x,e) (float) ldexp ((x), (e))
596 #endif 683 #endif
597 684
598 /* converts an ieee half/binary16 to a float */ 685 /* converts an ieee half/binary16 to a float */
599 ecb_function_ float ecb_binary16_to_float (uint16_t x) ecb_const; 686 ecb_function_ ecb_const float ecb_binary16_to_float (uint16_t x);
600 ecb_function_ float 687 ecb_function_ ecb_const float
601 ecb_binary16_to_float (uint16_t x) 688 ecb_binary16_to_float (uint16_t x)
602 { 689 {
603 int e = (x >> 10) & 0x1f; 690 int e = (x >> 10) & 0x1f;
604 int m = x & 0x3ff; 691 int m = x & 0x3ff;
605 float r; 692 float r;
606 693
607 if (!e ) r = ldexpf (m , -24); 694 if (!e ) r = ecb_ldexpf (m , -24);
608 else if (e != 31) r = ldexpf (m + 0x400, e - 25); 695 else if (e != 31) r = ecb_ldexpf (m + 0x400, e - 25);
609 else if (m ) r = ECB_NAN; 696 else if (m ) r = ECB_NAN;
610 else r = INFINITY; 697 else r = ECB_INFINITY;
611 698
612 return x & 0x8000 ? -r : r; 699 return x & 0x8000 ? -r : r;
613 } 700 }
614 701
615 /* convert a float to ieee single/binary32 */ 702 /* convert a float to ieee single/binary32 */
616 ecb_function_ uint32_t ecb_float_to_binary32 (float x) ecb_const; 703 ecb_function_ ecb_const uint32_t ecb_float_to_binary32 (float x);
617 ecb_function_ uint32_t 704 ecb_function_ ecb_const uint32_t
618 ecb_float_to_binary32 (float x) 705 ecb_float_to_binary32 (float x)
619 { 706 {
620 uint32_t r; 707 uint32_t r;
621 708
622 #if ECB_STDFP 709 #if ECB_STDFP
651 738
652 return r; 739 return r;
653 } 740 }
654 741
655 /* converts an ieee single/binary32 to a float */ 742 /* converts an ieee single/binary32 to a float */
656 ecb_function_ float ecb_binary32_to_float (uint32_t x) ecb_const; 743 ecb_function_ ecb_const float ecb_binary32_to_float (uint32_t x);
657 ecb_function_ float 744 ecb_function_ ecb_const float
658 ecb_binary32_to_float (uint32_t x) 745 ecb_binary32_to_float (uint32_t x)
659 { 746 {
660 float r; 747 float r;
661 748
662 #if ECB_STDFP 749 #if ECB_STDFP
672 x |= 0x800000U; 759 x |= 0x800000U;
673 else 760 else
674 e = 1; 761 e = 1;
675 762
676 /* we distrust ldexpf a bit and do the 2**-24 scaling by an extra multiply */ 763 /* we distrust ldexpf a bit and do the 2**-24 scaling by an extra multiply */
677 r = ldexpf (x * (0.5f / 0x800000U), e - 126); 764 r = ecb_ldexpf (x * (0.5f / 0x800000U), e - 126);
678 765
679 r = neg ? -r : r; 766 r = neg ? -r : r;
680 #endif 767 #endif
681 768
682 return r; 769 return r;
683 } 770 }
684 771
685 /* convert a double to ieee double/binary64 */ 772 /* convert a double to ieee double/binary64 */
686 ecb_function_ uint64_t ecb_double_to_binary64 (double x) ecb_const; 773 ecb_function_ ecb_const uint64_t ecb_double_to_binary64 (double x);
687 ecb_function_ uint64_t 774 ecb_function_ ecb_const uint64_t
688 ecb_double_to_binary64 (double x) 775 ecb_double_to_binary64 (double x)
689 { 776 {
690 uint64_t r; 777 uint64_t r;
691 778
692 #if ECB_STDFP 779 #if ECB_STDFP
721 808
722 return r; 809 return r;
723 } 810 }
724 811
725 /* converts an ieee double/binary64 to a double */ 812 /* converts an ieee double/binary64 to a double */
726 ecb_function_ double ecb_binary64_to_double (uint64_t x) ecb_const; 813 ecb_function_ ecb_const double ecb_binary64_to_double (uint64_t x);
727 ecb_function_ double 814 ecb_function_ ecb_const double
728 ecb_binary64_to_double (uint64_t x) 815 ecb_binary64_to_double (uint64_t x)
729 { 816 {
730 double r; 817 double r;
731 818
732 #if ECB_STDFP 819 #if ECB_STDFP

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines