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

Comparing libecb/ecb.h (file contents):
Revision 1.126 by root, Sun Jan 5 02:58:29 2014 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 (©) 2009-2013 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 ()
182 #elif _MSC_VER >= 1500 /* VC++ 2008 */ 214 #elif _MSC_VER >= 1500 /* VC++ 2008 */
183 /* apparently, microsoft broke all the memory barrier stuff in Visual Studio 2008... */ 215 /* apparently, microsoft broke all the memory barrier stuff in Visual Studio 2008... */
212 /* 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. */
213 /* 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 */
214 /* for most usages, or gcc and clang have a bug */ 246 /* for most usages, or gcc and clang have a bug */
215 /* I *currently* lean towards the latter, and inefficiently implement */ 247 /* I *currently* lean towards the latter, and inefficiently implement */
216 /* 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 */
217 #define ECB_MEMORY_FENCE atomic_thread_fence (memory_order_seq_cst) 251 #define ECB_MEMORY_FENCE atomic_thread_fence (memory_order_seq_cst)
218 #endif 252 #endif
219#endif 253#endif
220 254
221#ifndef ECB_MEMORY_FENCE 255#ifndef ECB_MEMORY_FENCE
244 #define ECB_MEMORY_FENCE_RELEASE ECB_MEMORY_FENCE 278 #define ECB_MEMORY_FENCE_RELEASE ECB_MEMORY_FENCE
245#endif 279#endif
246 280
247/*****************************************************************************/ 281/*****************************************************************************/
248 282
249#if __cplusplus 283#if ECB_CPP
250 #define ecb_inline static inline 284 #define ecb_inline static inline
251#elif ECB_GCC_VERSION(2,5) 285#elif ECB_GCC_VERSION(2,5)
252 #define ecb_inline static __inline__ 286 #define ecb_inline static __inline__
253#elif ECB_C99 287#elif ECB_C99
254 #define ecb_inline static inline 288 #define ecb_inline static inline
271#define ECB_STRINGIFY_(a) # a 305#define ECB_STRINGIFY_(a) # a
272#define ECB_STRINGIFY(a) ECB_STRINGIFY_(a) 306#define ECB_STRINGIFY(a) ECB_STRINGIFY_(a)
273 307
274#define ecb_function_ ecb_inline 308#define ecb_function_ ecb_inline
275 309
276#if ECB_GCC_VERSION(3,1) 310#if ECB_GCC_VERSION(3,1) || ECB_CLANG_VERSION(2,8)
277 #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)
278 #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)
279 #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)
280 #define ecb_prefetch(addr,rw,locality) __builtin_prefetch (addr, rw, locality) 333 #define ecb_prefetch(addr,rw,locality) __builtin_prefetch (addr, rw, locality)
281#else 334#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) 335 #define ecb_prefetch(addr,rw,locality)
286#endif 336#endif
287 337
288/* no emulation for ecb_decltype */ 338/* no emulation for ecb_decltype */
289#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; };
290 #define ecb_decltype(x) __decltype(x) 342 #define ecb_decltype(x) ecb_decltype_t<decltype (x)>::type
291#elif ECB_GCC_VERSION(3,0) 343#elif ECB_GCC_VERSION(3,0) || ECB_CLANG_VERSION(2,8)
292 #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__))
293#endif 351#endif
294 352
295#define ecb_noinline ecb_attribute ((__noinline__)) 353#define ecb_noinline ecb_attribute ((__noinline__))
296#define ecb_unused ecb_attribute ((__unused__)) 354#define ecb_unused ecb_attribute ((__unused__))
297#define ecb_const ecb_attribute ((__const__)) 355#define ecb_const ecb_attribute ((__const__))
298#define ecb_pure ecb_attribute ((__pure__)) 356#define ecb_pure ecb_attribute ((__pure__))
299 357
300#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 */
301 #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)
302#else 366#else
303 #define ecb_noreturn ecb_attribute ((__noreturn__)) 367 #define ecb_noreturn ecb_attribute ((__noreturn__))
304#endif 368#endif
305 369
306#if ECB_GCC_VERSION(4,3) 370#if ECB_GCC_VERSION(4,3)
321/* for compatibility to the rest of the world */ 385/* for compatibility to the rest of the world */
322#define ecb_likely(expr) ecb_expect_true (expr) 386#define ecb_likely(expr) ecb_expect_true (expr)
323#define ecb_unlikely(expr) ecb_expect_false (expr) 387#define ecb_unlikely(expr) ecb_expect_false (expr)
324 388
325/* count trailing zero bits and count # of one bits */ 389/* count trailing zero bits and count # of one bits */
326#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))
327 /* 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 */
328 #define ecb_ld32(x) (__builtin_clz (x) ^ 31) 395 #define ecb_ld32(x) (__builtin_clz (x) ^ 31)
329 #define ecb_ld64(x) (__builtin_clzll (x) ^ 63) 396 #define ecb_ld64(x) (__builtin_clzll (x) ^ 63)
330 #define ecb_ctz32(x) __builtin_ctz (x) 397 #define ecb_ctz32(x) __builtin_ctz (x)
331 #define ecb_ctz64(x) __builtin_ctzll (x) 398 #define ecb_ctz64(x) __builtin_ctzll (x)
332 #define ecb_popcount32(x) __builtin_popcount (x) 399 #define ecb_popcount32(x) __builtin_popcount (x)
333 /* no popcountll */ 400 /* no popcountll */
334#else 401#else
335 ecb_function_ int ecb_ctz32 (uint32_t x) ecb_const; 402 ecb_function_ ecb_const int ecb_ctz32 (uint32_t x);
336 ecb_function_ int 403 ecb_function_ ecb_const int
337 ecb_ctz32 (uint32_t x) 404 ecb_ctz32 (uint32_t x)
338 { 405 {
339 int r = 0; 406 int r = 0;
340 407
341 x &= ~x + 1; /* this isolates the lowest bit */ 408 x &= ~x + 1; /* this isolates the lowest bit */
355#endif 422#endif
356 423
357 return r; 424 return r;
358 } 425 }
359 426
360 ecb_function_ int ecb_ctz64 (uint64_t x) ecb_const; 427 ecb_function_ ecb_const int ecb_ctz64 (uint64_t x);
361 ecb_function_ int 428 ecb_function_ ecb_const int
362 ecb_ctz64 (uint64_t x) 429 ecb_ctz64 (uint64_t x)
363 { 430 {
364 int shift = x & 0xffffffffU ? 0 : 32; 431 int shift = x & 0xffffffffU ? 0 : 32;
365 return ecb_ctz32 (x >> shift) + shift; 432 return ecb_ctz32 (x >> shift) + shift;
366 } 433 }
367 434
368 ecb_function_ int ecb_popcount32 (uint32_t x) ecb_const; 435 ecb_function_ ecb_const int ecb_popcount32 (uint32_t x);
369 ecb_function_ int 436 ecb_function_ ecb_const int
370 ecb_popcount32 (uint32_t x) 437 ecb_popcount32 (uint32_t x)
371 { 438 {
372 x -= (x >> 1) & 0x55555555; 439 x -= (x >> 1) & 0x55555555;
373 x = ((x >> 2) & 0x33333333) + (x & 0x33333333); 440 x = ((x >> 2) & 0x33333333) + (x & 0x33333333);
374 x = ((x >> 4) + x) & 0x0f0f0f0f; 441 x = ((x >> 4) + x) & 0x0f0f0f0f;
375 x *= 0x01010101; 442 x *= 0x01010101;
376 443
377 return x >> 24; 444 return x >> 24;
378 } 445 }
379 446
380 ecb_function_ int ecb_ld32 (uint32_t x) ecb_const; 447 ecb_function_ ecb_const int ecb_ld32 (uint32_t x);
381 ecb_function_ int ecb_ld32 (uint32_t x) 448 ecb_function_ ecb_const int ecb_ld32 (uint32_t x)
382 { 449 {
383 int r = 0; 450 int r = 0;
384 451
385 if (x >> 16) { x >>= 16; r += 16; } 452 if (x >> 16) { x >>= 16; r += 16; }
386 if (x >> 8) { x >>= 8; r += 8; } 453 if (x >> 8) { x >>= 8; r += 8; }
389 if (x >> 1) { r += 1; } 456 if (x >> 1) { r += 1; }
390 457
391 return r; 458 return r;
392 } 459 }
393 460
394 ecb_function_ int ecb_ld64 (uint64_t x) ecb_const; 461 ecb_function_ ecb_const int ecb_ld64 (uint64_t x);
395 ecb_function_ int ecb_ld64 (uint64_t x) 462 ecb_function_ ecb_const int ecb_ld64 (uint64_t x)
396 { 463 {
397 int r = 0; 464 int r = 0;
398 465
399 if (x >> 32) { x >>= 32; r += 32; } 466 if (x >> 32) { x >>= 32; r += 32; }
400 467
401 return r + ecb_ld32 (x); 468 return r + ecb_ld32 (x);
402 } 469 }
403#endif 470#endif
404 471
405ecb_function_ ecb_bool ecb_is_pot32 (uint32_t x) ecb_const; 472ecb_function_ ecb_const ecb_bool ecb_is_pot32 (uint32_t x);
406ecb_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)); }
407ecb_function_ ecb_bool ecb_is_pot64 (uint64_t x) ecb_const; 474ecb_function_ ecb_const ecb_bool ecb_is_pot64 (uint64_t x);
408ecb_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)); }
409 476
410ecb_function_ uint8_t ecb_bitrev8 (uint8_t x) ecb_const; 477ecb_function_ ecb_const uint8_t ecb_bitrev8 (uint8_t x);
411ecb_function_ uint8_t ecb_bitrev8 (uint8_t x) 478ecb_function_ ecb_const uint8_t ecb_bitrev8 (uint8_t x)
412{ 479{
413 return ( (x * 0x0802U & 0x22110U) 480 return ( (x * 0x0802U & 0x22110U)
414 | (x * 0x8020U & 0x88440U)) * 0x10101U >> 16; 481 | (x * 0x8020U & 0x88440U)) * 0x10101U >> 16;
415} 482}
416 483
417ecb_function_ uint16_t ecb_bitrev16 (uint16_t x) ecb_const; 484ecb_function_ ecb_const uint16_t ecb_bitrev16 (uint16_t x);
418ecb_function_ uint16_t ecb_bitrev16 (uint16_t x) 485ecb_function_ ecb_const uint16_t ecb_bitrev16 (uint16_t x)
419{ 486{
420 x = ((x >> 1) & 0x5555) | ((x & 0x5555) << 1); 487 x = ((x >> 1) & 0x5555) | ((x & 0x5555) << 1);
421 x = ((x >> 2) & 0x3333) | ((x & 0x3333) << 2); 488 x = ((x >> 2) & 0x3333) | ((x & 0x3333) << 2);
422 x = ((x >> 4) & 0x0f0f) | ((x & 0x0f0f) << 4); 489 x = ((x >> 4) & 0x0f0f) | ((x & 0x0f0f) << 4);
423 x = ( x >> 8 ) | ( x << 8); 490 x = ( x >> 8 ) | ( x << 8);
424 491
425 return x; 492 return x;
426} 493}
427 494
428ecb_function_ uint32_t ecb_bitrev32 (uint32_t x) ecb_const; 495ecb_function_ ecb_const uint32_t ecb_bitrev32 (uint32_t x);
429ecb_function_ uint32_t ecb_bitrev32 (uint32_t x) 496ecb_function_ ecb_const uint32_t ecb_bitrev32 (uint32_t x)
430{ 497{
431 x = ((x >> 1) & 0x55555555) | ((x & 0x55555555) << 1); 498 x = ((x >> 1) & 0x55555555) | ((x & 0x55555555) << 1);
432 x = ((x >> 2) & 0x33333333) | ((x & 0x33333333) << 2); 499 x = ((x >> 2) & 0x33333333) | ((x & 0x33333333) << 2);
433 x = ((x >> 4) & 0x0f0f0f0f) | ((x & 0x0f0f0f0f) << 4); 500 x = ((x >> 4) & 0x0f0f0f0f) | ((x & 0x0f0f0f0f) << 4);
434 x = ((x >> 8) & 0x00ff00ff) | ((x & 0x00ff00ff) << 8); 501 x = ((x >> 8) & 0x00ff00ff) | ((x & 0x00ff00ff) << 8);
437 return x; 504 return x;
438} 505}
439 506
440/* popcount64 is only available on 64 bit cpus as gcc builtin */ 507/* popcount64 is only available on 64 bit cpus as gcc builtin */
441/* so for this version we are lazy */ 508/* so for this version we are lazy */
442ecb_function_ int ecb_popcount64 (uint64_t x) ecb_const; 509ecb_function_ ecb_const int ecb_popcount64 (uint64_t x);
443ecb_function_ int 510ecb_function_ ecb_const int
444ecb_popcount64 (uint64_t x) 511ecb_popcount64 (uint64_t x)
445{ 512{
446 return ecb_popcount32 (x) + ecb_popcount32 (x >> 32); 513 return ecb_popcount32 (x) + ecb_popcount32 (x >> 32);
447} 514}
448 515
449ecb_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);
450ecb_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);
451ecb_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);
452ecb_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);
453ecb_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);
454ecb_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);
455ecb_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);
456ecb_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);
457 524
458ecb_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); }
459ecb_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); }
460ecb_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); }
461ecb_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); }
462ecb_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); }
463ecb_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); }
464ecb_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); }
465ecb_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); }
466 533
467#if ECB_GCC_VERSION(4,3) 534#if ECB_GCC_VERSION(4,3) || (ECB_CLANG_BUILTIN(__builtin_bswap32) && ECB_CLANG_BUILTIN(__builtin_bswap64))
468 #define ecb_bswap16(x) (__builtin_bswap32 (x) >> 16) 535 #define ecb_bswap16(x) (__builtin_bswap32 (x) >> 16)
469 #define ecb_bswap32(x) __builtin_bswap32 (x) 536 #define ecb_bswap32(x) __builtin_bswap32 (x)
470 #define ecb_bswap64(x) __builtin_bswap64 (x) 537 #define ecb_bswap64(x) __builtin_bswap64 (x)
471#else 538#else
472 ecb_function_ uint16_t ecb_bswap16 (uint16_t x) ecb_const; 539 ecb_function_ ecb_const uint16_t ecb_bswap16 (uint16_t x);
473 ecb_function_ uint16_t 540 ecb_function_ ecb_const uint16_t
474 ecb_bswap16 (uint16_t x) 541 ecb_bswap16 (uint16_t x)
475 { 542 {
476 return ecb_rotl16 (x, 8); 543 return ecb_rotl16 (x, 8);
477 } 544 }
478 545
479 ecb_function_ uint32_t ecb_bswap32 (uint32_t x) ecb_const; 546 ecb_function_ ecb_const uint32_t ecb_bswap32 (uint32_t x);
480 ecb_function_ uint32_t 547 ecb_function_ ecb_const uint32_t
481 ecb_bswap32 (uint32_t x) 548 ecb_bswap32 (uint32_t x)
482 { 549 {
483 return (((uint32_t)ecb_bswap16 (x)) << 16) | ecb_bswap16 (x >> 16); 550 return (((uint32_t)ecb_bswap16 (x)) << 16) | ecb_bswap16 (x >> 16);
484 } 551 }
485 552
486 ecb_function_ uint64_t ecb_bswap64 (uint64_t x) ecb_const; 553 ecb_function_ ecb_const uint64_t ecb_bswap64 (uint64_t x);
487 ecb_function_ uint64_t 554 ecb_function_ ecb_const uint64_t
488 ecb_bswap64 (uint64_t x) 555 ecb_bswap64 (uint64_t x)
489 { 556 {
490 return (((uint64_t)ecb_bswap32 (x)) << 32) | ecb_bswap32 (x >> 32); 557 return (((uint64_t)ecb_bswap32 (x)) << 32) | ecb_bswap32 (x >> 32);
491 } 558 }
492#endif 559#endif
493 560
494#if ECB_GCC_VERSION(4,5) 561#if ECB_GCC_VERSION(4,5) || ECB_CLANG_BUILTIN(__builtin_unreachable)
495 #define ecb_unreachable() __builtin_unreachable () 562 #define ecb_unreachable() __builtin_unreachable ()
496#else 563#else
497 /* 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 :/ */
498 ecb_inline void ecb_unreachable (void) ecb_noreturn; 565 ecb_inline ecb_noreturn void ecb_unreachable (void);
499 ecb_inline void ecb_unreachable (void) { } 566 ecb_inline ecb_noreturn void ecb_unreachable (void) { }
500#endif 567#endif
501 568
502/* try to tell the compiler that some condition is definitely true */ 569/* try to tell the compiler that some condition is definitely true */
503#define ecb_assume(cond) if (!(cond)) ecb_unreachable (); else 0 570#define ecb_assume(cond) if (!(cond)) ecb_unreachable (); else 0
504 571
505ecb_inline unsigned char ecb_byteorder_helper (void) ecb_const; 572ecb_inline ecb_const unsigned char ecb_byteorder_helper (void);
506ecb_inline unsigned char 573ecb_inline ecb_const unsigned char
507ecb_byteorder_helper (void) 574ecb_byteorder_helper (void)
508{ 575{
509 /* the union code still generates code under pressure in gcc, */ 576 /* the union code still generates code under pressure in gcc, */
510 /* but less than using pointers, and always seems to */ 577 /* but less than using pointers, and always seems to */
511 /* successfully return a constant. */ 578 /* successfully return a constant. */
526 } u = { 0x11223344 }; 593 } u = { 0x11223344 };
527 return u.c; 594 return u.c;
528#endif 595#endif
529} 596}
530 597
531ecb_inline ecb_bool ecb_big_endian (void) ecb_const; 598ecb_inline ecb_const ecb_bool ecb_big_endian (void);
532ecb_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; }
533ecb_inline ecb_bool ecb_little_endian (void) ecb_const; 600ecb_inline ecb_const ecb_bool ecb_little_endian (void);
534ecb_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; }
535 602
536#if ECB_GCC_VERSION(3,0) || ECB_C99 603#if ECB_GCC_VERSION(3,0) || ECB_C99
537 #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))
538#else 605#else
539 #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)))
540#endif 607#endif
541 608
542#if __cplusplus 609#if ECB_CPP
543 template<typename T> 610 template<typename T>
544 static inline T ecb_div_rd (T val, T div) 611 static inline T ecb_div_rd (T val, T div)
545 { 612 {
546 return val < 0 ? - ((-val + div - 1) / div) : (val ) / div; 613 return val < 0 ? - ((-val + div - 1) / div) : (val ) / div;
547 } 614 }
573/* the only noteworthy exception is ancient armle, which uses order 43218765 */ 640/* the only noteworthy exception is ancient armle, which uses order 43218765 */
574#if 0 \ 641#if 0 \
575 || __i386 || __i386__ \ 642 || __i386 || __i386__ \
576 || __amd64 || __amd64__ || __x86_64 || __x86_64__ \ 643 || __amd64 || __amd64__ || __x86_64 || __x86_64__ \
577 || __powerpc__ || __ppc__ || __powerpc64__ || __ppc64__ \ 644 || __powerpc__ || __ppc__ || __powerpc64__ || __ppc64__ \
578 || defined __arm__ && defined __ARM_EABI__ \
579 || defined __s390__ || defined __s390x__ \ 645 || defined __s390__ || defined __s390x__ \
580 || defined __mips__ \ 646 || defined __mips__ \
581 || defined __alpha__ \ 647 || defined __alpha__ \
582 || defined __hppa__ \ 648 || defined __hppa__ \
583 || defined __ia64__ \ 649 || defined __ia64__ \
584 || defined __m68k__ \ 650 || defined __m68k__ \
585 || defined __m88k__ \ 651 || defined __m88k__ \
586 || defined __sh__ \ 652 || defined __sh__ \
587 || 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__
588 #define ECB_STDFP 1 656 #define ECB_STDFP 1
589 #include <string.h> /* for memcpy */ 657 #include <string.h> /* for memcpy */
590#else 658#else
591 #define ECB_STDFP 0 659 #define ECB_STDFP 0
592#endif 660#endif
606 #define ECB_NAN NAN 674 #define ECB_NAN NAN
607 #else 675 #else
608 #define ECB_NAN ECB_INFINITY 676 #define ECB_NAN ECB_INFINITY
609 #endif 677 #endif
610 678
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))
683 #endif
684
611 /* converts an ieee half/binary16 to a float */ 685 /* converts an ieee half/binary16 to a float */
612 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);
613 ecb_function_ float 687 ecb_function_ ecb_const float
614 ecb_binary16_to_float (uint16_t x) 688 ecb_binary16_to_float (uint16_t x)
615 { 689 {
616 int e = (x >> 10) & 0x1f; 690 int e = (x >> 10) & 0x1f;
617 int m = x & 0x3ff; 691 int m = x & 0x3ff;
618 float r; 692 float r;
619 693
620 if (!e ) r = ldexpf (m , -24); 694 if (!e ) r = ecb_ldexpf (m , -24);
621 else if (e != 31) r = ldexpf (m + 0x400, e - 25); 695 else if (e != 31) r = ecb_ldexpf (m + 0x400, e - 25);
622 else if (m ) r = ECB_NAN; 696 else if (m ) r = ECB_NAN;
623 else r = ECB_INFINITY; 697 else r = ECB_INFINITY;
624 698
625 return x & 0x8000 ? -r : r; 699 return x & 0x8000 ? -r : r;
626 } 700 }
627 701
628 /* convert a float to ieee single/binary32 */ 702 /* convert a float to ieee single/binary32 */
629 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);
630 ecb_function_ uint32_t 704 ecb_function_ ecb_const uint32_t
631 ecb_float_to_binary32 (float x) 705 ecb_float_to_binary32 (float x)
632 { 706 {
633 uint32_t r; 707 uint32_t r;
634 708
635 #if ECB_STDFP 709 #if ECB_STDFP
664 738
665 return r; 739 return r;
666 } 740 }
667 741
668 /* converts an ieee single/binary32 to a float */ 742 /* converts an ieee single/binary32 to a float */
669 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);
670 ecb_function_ float 744 ecb_function_ ecb_const float
671 ecb_binary32_to_float (uint32_t x) 745 ecb_binary32_to_float (uint32_t x)
672 { 746 {
673 float r; 747 float r;
674 748
675 #if ECB_STDFP 749 #if ECB_STDFP
685 x |= 0x800000U; 759 x |= 0x800000U;
686 else 760 else
687 e = 1; 761 e = 1;
688 762
689 /* 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 */
690 r = ldexpf (x * (0.5f / 0x800000U), e - 126); 764 r = ecb_ldexpf (x * (0.5f / 0x800000U), e - 126);
691 765
692 r = neg ? -r : r; 766 r = neg ? -r : r;
693 #endif 767 #endif
694 768
695 return r; 769 return r;
696 } 770 }
697 771
698 /* convert a double to ieee double/binary64 */ 772 /* convert a double to ieee double/binary64 */
699 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);
700 ecb_function_ uint64_t 774 ecb_function_ ecb_const uint64_t
701 ecb_double_to_binary64 (double x) 775 ecb_double_to_binary64 (double x)
702 { 776 {
703 uint64_t r; 777 uint64_t r;
704 778
705 #if ECB_STDFP 779 #if ECB_STDFP
734 808
735 return r; 809 return r;
736 } 810 }
737 811
738 /* converts an ieee double/binary64 to a double */ 812 /* converts an ieee double/binary64 to a double */
739 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);
740 ecb_function_ double 814 ecb_function_ ecb_const double
741 ecb_binary64_to_double (uint64_t x) 815 ecb_binary64_to_double (uint64_t x)
742 { 816 {
743 double r; 817 double r;
744 818
745 #if ECB_STDFP 819 #if ECB_STDFP

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines