ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/libecb/ecb.h

Comparing libecb/ecb.h (file contents):
Revision 1.62 by root, Thu Aug 4 14:34:00 2011 UTC vs.
Revision 1.203 by root, Thu Mar 24 00:57:30 2022 UTC

1	/*	1	/*
2	* libecb - http://software.schmorp.de/pkg/libecb	2	* libecb - http://software.schmorp.de/pkg/libecb
3	*	3	*
4	* Copyright (©) 2009-2011 Marc Alexander Lehmann <libecb@schmorp.de>	4	* Copyright (©) 2009-2015,2018-2021 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	#ifdef _WIN32	44	/* 16 bits major, 16 bits minor */
		45	#define ECB_VERSION 0x0001000b
		46
		47	#include <string.h> /* for memcpy */
		48
		49	#if defined (_WIN32) && !defined (__MINGW32__)
34	typedef signed char int8_t;	50	typedef signed char int8_t;
35	typedef unsigned char uint8_t;	51	typedef unsigned char uint8_t;
		52	typedef signed char int_fast8_t;
		53	typedef unsigned char uint_fast8_t;
36	typedef signed short int16_t;	54	typedef signed short int16_t;
37	typedef unsigned short uint16_t;	55	typedef unsigned short uint16_t;
		56	typedef signed int int_fast16_t;
		57	typedef unsigned int uint_fast16_t;
38	typedef signed int int32_t;	58	typedef signed int int32_t;
39	typedef unsigned int uint32_t;	59	typedef unsigned int uint32_t;
		60	typedef signed int int_fast32_t;
		61	typedef unsigned int uint_fast32_t;
40	#if __GNUC__	62	#if __GNUC__
41	typedef signed long long int64_t;	63	typedef signed long long int64_t;
42	typedef unsigned long long uint64_t;	64	typedef unsigned long long uint64_t;
43	#else /* _MSC_VER || __BORLANDC__ */	65	#else /* _MSC_VER || __BORLANDC__ */
44	typedef signed __int64 int64_t;	66	typedef signed __int64 int64_t;
45	typedef unsigned __int64 uint64_t;	67	typedef unsigned __int64 uint64_t;
46	#endif	68	#endif
		69	typedef int64_t int_fast64_t;
		70	typedef uint64_t uint_fast64_t;
		71	#ifdef _WIN64
		72	#define ECB_PTRSIZE 8
		73	typedef uint64_t uintptr_t;
		74	typedef int64_t intptr_t;
		75	#else
		76	#define ECB_PTRSIZE 4
		77	typedef uint32_t uintptr_t;
		78	typedef int32_t intptr_t;
		79	#endif
47	#else	80	#else
48	#include <inttypes.h>	81	#include <inttypes.h>
		82	#if (defined INTPTR_MAX ? INTPTR_MAX : ULONG_MAX) > 0xffffffffU
		83	#define ECB_PTRSIZE 8
		84	#else
		85	#define ECB_PTRSIZE 4
		86	#endif
		87	#endif
		88
		89	#define ECB_GCC_AMD64 (__amd64 || __amd64__ || __x86_64 || __x86_64__)
		90	#define ECB_MSVC_AMD64 (_M_AMD64 || _M_X64)
		91
		92	#ifndef ECB_OPTIMIZE_SIZE
		93	#if __OPTIMIZE_SIZE__
		94	#define ECB_OPTIMIZE_SIZE 1
		95	#else
		96	#define ECB_OPTIMIZE_SIZE 0
		97	#endif
		98	#endif
		99
		100	/* work around x32 idiocy by defining proper macros */
		101	#if ECB_GCC_AMD64 || ECB_MSVC_AMD64
		102	#if _ILP32
		103	#define ECB_AMD64_X32 1
		104	#else
		105	#define ECB_AMD64 1
		106	#endif
		107	#endif
		108
		109	#if ECB_PTRSIZE >= 8 || ECB_AMD64_X32
		110	#define ECB_64BIT_NATIVE 1
		111	#else
		112	#define ECB_64BIT_NATIVE 0
49	#endif	113	#endif
50		114
51	/* many compilers define _GNUC_ to some versions but then only implement	115	/* many compilers define _GNUC_ to some versions but then only implement
52	* what their idiot authors think are the "more important" extensions,	116	* what their idiot authors think are the "more important" extensions,
53	* causing enormous grief in return for some better fake benchmark numbers.	117	* causing enormous grief in return for some better fake benchmark numbers.
54	* or so.	118	* or so.
55	* we try to detect these and simply assume they are not gcc - if they have	119	* we try to detect these and simply assume they are not gcc - if they have
56	* an issue with that they should have done it right in the first place.	120	* an issue with that they should have done it right in the first place.
57	*/	121	*/
58	#ifndef ECB_GCC_VERSION
59	#if !defined(__GNUC_MINOR__) || defined(__INTEL_COMPILER) || defined(__SUNPRO_C) || defined(__SUNPRO_CC) || defined(__llvm__) || defined(__clang__)	122	#if !defined __GNUC_MINOR__ || defined __INTEL_COMPILER || defined __SUNPRO_C || defined __SUNPRO_CC || defined __llvm__ || defined __clang__
60	#define ECB_GCC_VERSION(major,minor) 0	123	#define ECB_GCC_VERSION(major,minor) 0
61	#else	124	#else
62	#define ECB_GCC_VERSION(major,minor) (__GNUC__ > (major) || (__GNUC__ == (major) && __GNUC_MINOR__ >= (minor)))	125	#define ECB_GCC_VERSION(major,minor) (__GNUC__ > (major) || (__GNUC__ == (major) && __GNUC_MINOR__ >= (minor)))
63	#endif	126	#endif
		127
		128	#define ECB_CLANG_VERSION(major,minor) (__clang_major__ > (major) || (__clang_major__ == (major) && __clang_minor__ >= (minor)))
		129
		130	#if __clang__ && defined __has_builtin
		131	#define ECB_CLANG_BUILTIN(x) __has_builtin (x)
		132	#else
		133	#define ECB_CLANG_BUILTIN(x) 0
		134	#endif
		135
		136	#if __clang__ && defined __has_extension
		137	#define ECB_CLANG_EXTENSION(x) __has_extension (x)
		138	#else
		139	#define ECB_CLANG_EXTENSION(x) 0
		140	#endif
		141
		142	#define ECB_CPP (__cplusplus+0)
		143	#define ECB_CPP11 (__cplusplus >= 201103L)
		144	#define ECB_CPP14 (__cplusplus >= 201402L)
		145	#define ECB_CPP17 (__cplusplus >= 201703L)
		146
		147	#if ECB_CPP
		148	#define ECB_C 0
		149	#define ECB_STDC_VERSION 0
		150	#else
		151	#define ECB_C 1
		152	#define ECB_STDC_VERSION __STDC_VERSION__
		153	#endif
		154
		155	#define ECB_C99 (ECB_STDC_VERSION >= 199901L)
		156	#define ECB_C11 (ECB_STDC_VERSION >= 201112L)
		157	#define ECB_C17 (ECB_STDC_VERSION >= 201710L)
		158
		159	#if ECB_CPP
		160	#define ECB_EXTERN_C extern "C"
		161	#define ECB_EXTERN_C_BEG ECB_EXTERN_C {
		162	#define ECB_EXTERN_C_END }
		163	#else
		164	#define ECB_EXTERN_C extern
		165	#define ECB_EXTERN_C_BEG
		166	#define ECB_EXTERN_C_END
64	#endif	167	#endif
65		168
66	/*****************************************************************************/	169	/*****************************************************************************/
67		170
68	/* ECB_NO_THREADS - ecb is not used by multiple threads, ever */	171	/* ECB_NO_THREADS - ecb is not used by multiple threads, ever */
69	/* ECB_NO_SMP - ecb might be used in multiple threads, but only on a single cpu */	172	/* ECB_NO_SMP - ecb might be used in multiple threads, but only on a single cpu */
70		173
71	#if ECB_NO_THREADS || ECB_NO_SMP	174	#if ECB_NO_THREADS
		175	#define ECB_NO_SMP 1
		176	#endif
		177
		178	#if ECB_NO_SMP
72	#define ECB_MEMORY_FENCE do { } while (0)	179	#define ECB_MEMORY_FENCE do { } while (0)
73	#define ECB_MEMORY_FENCE_ACQUIRE do { } while (0)	180	#endif
74	#define ECB_MEMORY_FENCE_RELEASE do { } while (0)	181
		182	/* http://www-01.ibm.com/support/knowledgecenter/SSGH3R_13.1.0/com.ibm.xlcpp131.aix.doc/compiler_ref/compiler_builtins.html */
		183	#if __xlC__ && ECB_CPP
		184	#include <builtins.h>
		185	#endif
		186
		187	#if 1400 <= _MSC_VER
		188	#include <intrin.h> /* fence functions _ReadBarrier, also bit search functions _BitScanReverse */
75	#endif	189	#endif
76		190
77	#ifndef ECB_MEMORY_FENCE	191	#ifndef ECB_MEMORY_FENCE
78	#if ECB_GCC_VERSION(2,5)	192	#if ECB_GCC_VERSION(2,5) || defined __INTEL_COMPILER || (__llvm__ && __GNUC__) || __SUNPRO_C >= 0x5110 || __SUNPRO_CC >= 0x5110
79	#if __x86	193	#define ECB_MEMORY_FENCE_RELAXED __asm__ __volatile__ ("" : : : "memory")
		194	#if __i386 || __i386__
80	#define ECB_MEMORY_FENCE __asm__ __volatile__ ("lock; orb $0, -1(%%esp)" : : : "memory")	195	#define ECB_MEMORY_FENCE __asm__ __volatile__ ("lock; orb $0, -1(%%esp)" : : : "memory")
81	#define ECB_MEMORY_FENCE_ACQUIRE ECB_MEMORY_FENCE /* non-lock xchg might be enough */	196	#define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("" : : : "memory")
82	#define ECB_MEMORY_FENCE_RELEASE do { } while (0) /* unlikely to change in future cpus */	197	#define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("" : : : "memory")
83	#elif __amd64	198	#elif ECB_GCC_AMD64
84	#define ECB_MEMORY_FENCE __asm__ __volatile__ ("mfence" : : : "memory")	199	#define ECB_MEMORY_FENCE __asm__ __volatile__ ("mfence" : : : "memory")
85	#define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("lfence" : : : "memory")	200	#define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("" : : : "memory")
86	#define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("sfence") /* play safe - not needed in any current cpu */	201	#define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("" : : : "memory")
		202	#elif __powerpc__ || __ppc__ || __powerpc64__ || __ppc64__
		203	#define ECB_MEMORY_FENCE __asm__ __volatile__ ("sync" : : : "memory")
		204	#elif defined __ARM_ARCH_2__ \
		205	|| defined __ARM_ARCH_3__ || defined __ARM_ARCH_3M__ \
		206	|| defined __ARM_ARCH_4__ || defined __ARM_ARCH_4T__ \
		207	|| defined __ARM_ARCH_5__ || defined __ARM_ARCH_5E__ \
		208	|| defined __ARM_ARCH_5T__ || defined __ARM_ARCH_5TE__ \
		209	|| defined __ARM_ARCH_5TEJ__
		210	/* should not need any, unless running old code on newer cpu - arm doesn't support that */
87	#elif defined(__ARM_ARCH_6__ ) || defined(__ARM_ARCH_6J__ ) \	211	#elif defined __ARM_ARCH_6__ || defined __ARM_ARCH_6J__ \
88	|| defined(__ARM_ARCH_6K__) || defined(__ARM_ARCH_6ZK__) \	212	|| defined __ARM_ARCH_6K__ || defined __ARM_ARCH_6ZK__ \
		213	|| defined __ARM_ARCH_6T2__
		214	#define ECB_MEMORY_FENCE __asm__ __volatile__ ("mcr p15,0,%0,c7,c10,5" : : "r" (0) : "memory")
		215	#elif defined __ARM_ARCH_7__ || defined __ARM_ARCH_7A__ \
89	|| defined(__ARM_ARCH_7__ ) || defined(__ARM_ARCH_7A__ ) \	216	|| defined __ARM_ARCH_7R__ || defined __ARM_ARCH_7M__
90	|| defined(__ARM_ARCH_7M__) || defined(__ARM_ARCH_7R__ )	217	#define ECB_MEMORY_FENCE __asm__ __volatile__ ("dmb" : : : "memory")
91	#define ECB_MEMORY_FENCE \	218	#elif __aarch64__
92	do { \	219	#define ECB_MEMORY_FENCE __asm__ __volatile__ ("dmb ish" : : : "memory")
93	int null = 0; \	220	#elif (__sparc || __sparc__) && !(__sparc_v8__ || defined __sparcv8)
94	__asm__ __volatile__ ("mcr p15,0,%0,c6,c10,5", : "=&r" (null) : : "memory"); \	221	#define ECB_MEMORY_FENCE __asm__ __volatile__ ("membar #LoadStore | #LoadLoad | #StoreStore | #StoreLoad" : : : "memory")
95	while (0)	222	#define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("membar #LoadStore | #LoadLoad" : : : "memory")
		223	#define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("membar #LoadStore | #StoreStore")
		224	#elif defined __s390__ || defined __s390x__
		225	#define ECB_MEMORY_FENCE __asm__ __volatile__ ("bcr 15,0" : : : "memory")
		226	#elif defined __mips__
		227	/* GNU/Linux emulates sync on mips1 architectures, so we force its use */
		228	/* anybody else who still uses mips1 is supposed to send in their version, with detection code. */
		229	#define ECB_MEMORY_FENCE __asm__ __volatile__ (".set mips2; sync; .set mips0" : : : "memory")
		230	#elif defined __alpha__
		231	#define ECB_MEMORY_FENCE __asm__ __volatile__ ("mb" : : : "memory")
		232	#elif defined __hppa__
		233	#define ECB_MEMORY_FENCE __asm__ __volatile__ ("" : : : "memory")
		234	#define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("")
		235	#elif defined __ia64__
		236	#define ECB_MEMORY_FENCE __asm__ __volatile__ ("mf" : : : "memory")
		237	#elif defined __m68k__
		238	#define ECB_MEMORY_FENCE __asm__ __volatile__ ("" : : : "memory")
		239	#elif defined __m88k__
		240	#define ECB_MEMORY_FENCE __asm__ __volatile__ ("tb1 0,%%r0,128" : : : "memory")
		241	#elif defined __sh__
		242	#define ECB_MEMORY_FENCE __asm__ __volatile__ ("" : : : "memory")
96	#endif	243	#endif
97	#endif	244	#endif
98	#endif	245	#endif
99		246
100	#ifndef ECB_MEMORY_FENCE	247	#ifndef ECB_MEMORY_FENCE
		248	#if ECB_GCC_VERSION(4,7)
		249	/* see comment below (stdatomic.h) about the C11 memory model. */
		250	#define ECB_MEMORY_FENCE __atomic_thread_fence (__ATOMIC_SEQ_CST)
		251	#define ECB_MEMORY_FENCE_ACQUIRE __atomic_thread_fence (__ATOMIC_ACQUIRE)
		252	#define ECB_MEMORY_FENCE_RELEASE __atomic_thread_fence (__ATOMIC_RELEASE)
		253	#undef ECB_MEMORY_FENCE_RELAXED
		254	#define ECB_MEMORY_FENCE_RELAXED __atomic_thread_fence (__ATOMIC_RELAXED)
		255
		256	#elif ECB_CLANG_EXTENSION(c_atomic)
		257	/* see comment below (stdatomic.h) about the C11 memory model. */
		258	#define ECB_MEMORY_FENCE __c11_atomic_thread_fence (__ATOMIC_SEQ_CST)
		259	#define ECB_MEMORY_FENCE_ACQUIRE __c11_atomic_thread_fence (__ATOMIC_ACQUIRE)
		260	#define ECB_MEMORY_FENCE_RELEASE __c11_atomic_thread_fence (__ATOMIC_RELEASE)
		261	#undef ECB_MEMORY_FENCE_RELAXED
		262	#define ECB_MEMORY_FENCE_RELAXED __c11_atomic_thread_fence (__ATOMIC_RELAXED)
		263
101	#if ECB_GCC_VERSION(4,4) || defined(__INTEL_COMPILER)	264	#elif ECB_GCC_VERSION(4,4) || defined __INTEL_COMPILER || defined __clang__
102	#define ECB_MEMORY_FENCE __sync_synchronize ()	265	#define ECB_MEMORY_FENCE __sync_synchronize ()
103	/*#define ECB_MEMORY_FENCE_ACQUIRE ({ char dummy = 0; __sync_lock_test_and_set (&dummy, 1); }) */	266	#elif _MSC_VER >= 1500 /* VC++ 2008 */
104	/*#define ECB_MEMORY_FENCE_RELEASE ({ char dummy = 1; __sync_lock_release (&dummy ); }) */	267	/* apparently, microsoft broke all the memory barrier stuff in Visual Studio 2008... */
		268	#pragma intrinsic(_ReadBarrier,_WriteBarrier,_ReadWriteBarrier)
		269	#define ECB_MEMORY_FENCE _ReadWriteBarrier (); MemoryBarrier()
		270	#define ECB_MEMORY_FENCE_ACQUIRE _ReadWriteBarrier (); MemoryBarrier() /* according to msdn, _ReadBarrier is not a load fence */
		271	#define ECB_MEMORY_FENCE_RELEASE _WriteBarrier (); MemoryBarrier()
105	#elif _MSC_VER >= 1400 /* VC++ 2005 */	272	#elif _MSC_VER >= 1400 /* VC++ 2005 */
106	#pragma intrinsic(_ReadBarrier,_WriteBarrier,_ReadWriteBarrier)	273	#pragma intrinsic(_ReadBarrier,_WriteBarrier,_ReadWriteBarrier)
107	#define ECB_MEMORY_FENCE _ReadWriteBarrier ()	274	#define ECB_MEMORY_FENCE _ReadWriteBarrier ()
108	#define ECB_MEMORY_FENCE_ACQUIRE _ReadWriteBarrier () /* according to msdn, _ReadBarrier is not a load fence */	275	#define ECB_MEMORY_FENCE_ACQUIRE _ReadWriteBarrier () /* according to msdn, _ReadBarrier is not a load fence */
109	#define ECB_MEMORY_FENCE_RELEASE _WriteBarrier ()	276	#define ECB_MEMORY_FENCE_RELEASE _WriteBarrier ()
110	#elif defined(_WIN32)	277	#elif defined _WIN32
111	#include <WinNT.h>	278	#include <WinNT.h>
112	#define ECB_MEMORY_FENCE MemoryBarrier () /* actually just xchg on x86... scary */	279	#define ECB_MEMORY_FENCE MemoryBarrier () /* actually just xchg on x86... scary */
		280	#elif __SUNPRO_C >= 0x5110 || __SUNPRO_CC >= 0x5110
		281	#include <mbarrier.h>
		282	#define ECB_MEMORY_FENCE __machine_rw_barrier ()
		283	#define ECB_MEMORY_FENCE_ACQUIRE __machine_acq_barrier ()
		284	#define ECB_MEMORY_FENCE_RELEASE __machine_rel_barrier ()
		285	#define ECB_MEMORY_FENCE_RELAXED __compiler_barrier ()
		286	#elif __xlC__
		287	#define ECB_MEMORY_FENCE __sync ()
		288	#endif
		289	#endif
		290
		291	#ifndef ECB_MEMORY_FENCE
		292	#if ECB_C11 && !defined __STDC_NO_ATOMICS__
		293	/* we assume that these memory fences work on all variables/all memory accesses, */
		294	/* not just C11 atomics and atomic accesses */
		295	#include <stdatomic.h>
		296	#define ECB_MEMORY_FENCE atomic_thread_fence (memory_order_seq_cst)
		297	#define ECB_MEMORY_FENCE_ACQUIRE atomic_thread_fence (memory_order_acquire)
		298	#define ECB_MEMORY_FENCE_RELEASE atomic_thread_fence (memory_order_release)
113	#endif	299	#endif
114	#endif	300	#endif
115		301
116	#ifndef ECB_MEMORY_FENCE	302	#ifndef ECB_MEMORY_FENCE
117	#if !ECB_AVOID_PTHREADS	303	#if !ECB_AVOID_PTHREADS
…		…
129	static pthread_mutex_t ecb_mf_lock = PTHREAD_MUTEX_INITIALIZER;	315	static pthread_mutex_t ecb_mf_lock = PTHREAD_MUTEX_INITIALIZER;
130	#define ECB_MEMORY_FENCE do { pthread_mutex_lock (&ecb_mf_lock); pthread_mutex_unlock (&ecb_mf_lock); } while (0)	316	#define ECB_MEMORY_FENCE do { pthread_mutex_lock (&ecb_mf_lock); pthread_mutex_unlock (&ecb_mf_lock); } while (0)
131	#endif	317	#endif
132	#endif	318	#endif
133		319
134	#if !defined(ECB_MEMORY_FENCE_ACQUIRE) && defined(ECB_MEMORY_FENCE)	320	#if !defined ECB_MEMORY_FENCE_ACQUIRE && defined ECB_MEMORY_FENCE
135	#define ECB_MEMORY_FENCE_ACQUIRE ECB_MEMORY_FENCE	321	#define ECB_MEMORY_FENCE_ACQUIRE ECB_MEMORY_FENCE
136	#endif	322	#endif
137		323
138	#if !defined(ECB_MEMORY_FENCE_RELEASE) && defined(ECB_MEMORY_FENCE)	324	#if !defined ECB_MEMORY_FENCE_RELEASE && defined ECB_MEMORY_FENCE
139	#define ECB_MEMORY_FENCE_RELEASE ECB_MEMORY_FENCE	325	#define ECB_MEMORY_FENCE_RELEASE ECB_MEMORY_FENCE
140	#endif	326	#endif
141		327
		328	#if !defined ECB_MEMORY_FENCE_RELAXED && defined ECB_MEMORY_FENCE
		329	#define ECB_MEMORY_FENCE_RELAXED ECB_MEMORY_FENCE /* very heavy-handed */
		330	#endif
		331
142	/*****************************************************************************/	332	/*****************************************************************************/
143		333
144	#define ECB_C99 (__STDC_VERSION__ >= 199901L)	334	#if ECB_CPP
145
146	#if __cplusplus
147	#define ecb_inline static inline	335	#define ecb_inline static inline
148	#elif ECB_GCC_VERSION(2,5)	336	#elif ECB_GCC_VERSION(2,5)
149	#define ecb_inline static __inline__	337	#define ecb_inline static __inline__
150	#elif ECB_C99	338	#elif ECB_C99
151	#define ecb_inline static inline	339	#define ecb_inline static inline
…		…
165		353
166	#define ECB_CONCAT_(a, b) a ## b	354	#define ECB_CONCAT_(a, b) a ## b
167	#define ECB_CONCAT(a, b) ECB_CONCAT_(a, b)	355	#define ECB_CONCAT(a, b) ECB_CONCAT_(a, b)
168	#define ECB_STRINGIFY_(a) # a	356	#define ECB_STRINGIFY_(a) # a
169	#define ECB_STRINGIFY(a) ECB_STRINGIFY_(a)	357	#define ECB_STRINGIFY(a) ECB_STRINGIFY_(a)
		358	#define ECB_STRINGIFY_EXPR(expr) ((expr), ECB_STRINGIFY_ (expr))
170		359
171	#define ecb_function_ ecb_inline	360	#define ecb_function_ ecb_inline
172		361
173	#if ECB_GCC_VERSION(3,1)	362	#if ECB_GCC_VERSION(3,1) || ECB_CLANG_VERSION(2,8)
174	#define ecb_attribute(attrlist) __attribute__(attrlist)	363	#define ecb_attribute(attrlist) __attribute__ (attrlist)
		364	#else
		365	#define ecb_attribute(attrlist)
		366	#endif
		367
		368	#if ECB_GCC_VERSION(3,1) || ECB_CLANG_BUILTIN(__builtin_constant_p)
175	#define ecb_is_constant(expr) __builtin_constant_p (expr)	369	#define ecb_is_constant(expr) __builtin_constant_p (expr)
		370	#else
		371	/* possible C11 impl for integral types
		372	typedef struct ecb_is_constant_struct ecb_is_constant_struct;
		373	#define ecb_is_constant(expr) _Generic ((1 ? (struct ecb_is_constant_struct *)0 : (void *)((expr) - (expr)), ecb_is_constant_struct *: 0, default: 1)) */
		374
		375	#define ecb_is_constant(expr) 0
		376	#endif
		377
		378	#if ECB_GCC_VERSION(3,1) || ECB_CLANG_BUILTIN(__builtin_expect)
176	#define ecb_expect(expr,value) __builtin_expect ((expr),(value))	379	#define ecb_expect(expr,value) __builtin_expect ((expr),(value))
		380	#else
		381	#define ecb_expect(expr,value) (expr)
		382	#endif
		383
		384	#if ECB_GCC_VERSION(3,1) || ECB_CLANG_BUILTIN(__builtin_prefetch)
177	#define ecb_prefetch(addr,rw,locality) __builtin_prefetch (addr, rw, locality)	385	#define ecb_prefetch(addr,rw,locality) __builtin_prefetch (addr, rw, locality)
178	#else	386	#else
179	#define ecb_attribute(attrlist)
180	#define ecb_is_constant(expr) 0
181	#define ecb_expect(expr,value) (expr)
182	#define ecb_prefetch(addr,rw,locality)	387	#define ecb_prefetch(addr,rw,locality)
183	#endif	388	#endif
184		389
185	/* no emulation for ecb_decltype */	390	/* no emulation for ecb_decltype */
186	#if ECB_GCC_VERSION(4,5)	391	#if ECB_CPP11
		392	// older implementations might have problems with decltype(x)::type, work around it
		393	template<class T> struct ecb_decltype_t { typedef T type; };
187	#define ecb_decltype(x) __decltype(x)	394	#define ecb_decltype(x) ecb_decltype_t<decltype (x)>::type
188	#elif ECB_GCC_VERSION(3,0)	395	#elif ECB_GCC_VERSION(3,0) || ECB_CLANG_VERSION(2,8)
189	#define ecb_decltype(x) __typeof(x)	396	#define ecb_decltype(x) __typeof__ (x)
190	#endif	397	#endif
191		398
		399	#if _MSC_VER >= 1300
		400	#define ecb_deprecated __declspec (deprecated)
		401	#else
		402	#define ecb_deprecated ecb_attribute ((__deprecated__))
		403	#endif
		404
		405	#if _MSC_VER >= 1500
		406	#define ecb_deprecated_message(msg) __declspec (deprecated (msg))
		407	#elif ECB_GCC_VERSION(4,5)
		408	#define ecb_deprecated_message(msg) ecb_attribute ((__deprecated__ (msg))
		409	#else
		410	#define ecb_deprecated_message(msg) ecb_deprecated
		411	#endif
		412
		413	#if _MSC_VER >= 1400
		414	#define ecb_noinline __declspec (noinline)
		415	#else
192	#define ecb_noinline ecb_attribute ((__noinline__))	416	#define ecb_noinline ecb_attribute ((__noinline__))
193	#define ecb_noreturn ecb_attribute ((__noreturn__))	417	#endif
		418
194	#define ecb_unused ecb_attribute ((__unused__))	419	#define ecb_unused ecb_attribute ((__unused__))
195	#define ecb_const ecb_attribute ((__const__))	420	#define ecb_const ecb_attribute ((__const__))
196	#define ecb_pure ecb_attribute ((__pure__))	421	#define ecb_pure ecb_attribute ((__pure__))
		422
		423	#if ECB_C11 || __IBMC_NORETURN
		424	/* http://www-01.ibm.com/support/knowledgecenter/SSGH3R_13.1.0/com.ibm.xlcpp131.aix.doc/language_ref/noreturn.html */
		425	#define ecb_noreturn _Noreturn
		426	#elif ECB_CPP11
		427	#define ecb_noreturn [[noreturn]]
		428	#elif _MSC_VER >= 1200
		429	/* http://msdn.microsoft.com/en-us/library/k6ktzx3s.aspx */
		430	#define ecb_noreturn __declspec (noreturn)
		431	#else
		432	#define ecb_noreturn ecb_attribute ((__noreturn__))
		433	#endif
197		434
198	#if ECB_GCC_VERSION(4,3)	435	#if ECB_GCC_VERSION(4,3)
199	#define ecb_artificial ecb_attribute ((__artificial__))	436	#define ecb_artificial ecb_attribute ((__artificial__))
200	#define ecb_hot ecb_attribute ((__hot__))	437	#define ecb_hot ecb_attribute ((__hot__))
201	#define ecb_cold ecb_attribute ((__cold__))	438	#define ecb_cold ecb_attribute ((__cold__))
…		…
213	/* for compatibility to the rest of the world */	450	/* for compatibility to the rest of the world */
214	#define ecb_likely(expr) ecb_expect_true (expr)	451	#define ecb_likely(expr) ecb_expect_true (expr)
215	#define ecb_unlikely(expr) ecb_expect_false (expr)	452	#define ecb_unlikely(expr) ecb_expect_false (expr)
216		453
217	/* count trailing zero bits and count # of one bits */	454	/* count trailing zero bits and count # of one bits */
218	#if ECB_GCC_VERSION(3,4)	455	#if ECB_GCC_VERSION(3,4) \
		456	|| (ECB_CLANG_BUILTIN(__builtin_clz) && ECB_CLANG_BUILTIN(__builtin_clzll) \
		457	&& ECB_CLANG_BUILTIN(__builtin_ctz) && ECB_CLANG_BUILTIN(__builtin_ctzll) \
		458	&& ECB_CLANG_BUILTIN(__builtin_popcount))
219	/* we assume int == 32 bit, long == 32 or 64 bit and long long == 64 bit */	459	/* we assume int == 32 bit, long == 32 or 64 bit and long long == 64 bit */
220	#define ecb_ld32(x) (__builtin_clz (x) ^ 31)	460	#define ecb_ld32(x) (__builtin_clz (x) ^ 31)
221	#define ecb_ld64(x) (__builtin_clzll (x) ^ 63)	461	#define ecb_ld64(x) (__builtin_clzll (x) ^ 63)
222	#define ecb_ctz32(x) __builtin_ctz (x)	462	#define ecb_ctz32(x) __builtin_ctz (x)
223	#define ecb_ctz64(x) __builtin_ctzll (x)	463	#define ecb_ctz64(x) __builtin_ctzll (x)
224	#define ecb_popcount32(x) __builtin_popcount (x)	464	#define ecb_popcount32(x) __builtin_popcount (x)
225	/* no popcountll */	465	/* no popcountll */
226	#else	466	#else
227	ecb_function_ int ecb_ctz32 (uint32_t x) ecb_const;	467	ecb_function_ ecb_const int ecb_ctz32 (uint32_t x);
228	ecb_function_ int	468	ecb_function_ ecb_const int
229	ecb_ctz32 (uint32_t x)	469	ecb_ctz32 (uint32_t x)
230	{	470	{
		471	#if 1400 <= _MSC_VER && (_M_IX86 || _M_X64 || _M_IA64 || _M_ARM)
		472	unsigned long r;
		473	_BitScanForward (&r, x);
		474	return (int)r;
		475	#else
231	int r = 0;	476	int r = 0;
232		477
233	x &= ~x + 1; /* this isolates the lowest bit */	478	x &= ~x + 1; /* this isolates the lowest bit */
234		479
235	#if ECB_branchless_on_i386	480	#if ECB_branchless_on_i386
…		…
245	if (x & 0xff00ff00) r += 8;	490	if (x & 0xff00ff00) r += 8;
246	if (x & 0xffff0000) r += 16;	491	if (x & 0xffff0000) r += 16;
247	#endif	492	#endif
248		493
249	return r;	494	return r;
		495	#endif
250	}	496	}
251		497
252	ecb_function_ int ecb_ctz64 (uint64_t x) ecb_const;	498	ecb_function_ ecb_const int ecb_ctz64 (uint64_t x);
253	ecb_function_ int	499	ecb_function_ ecb_const int
254	ecb_ctz64 (uint64_t x)	500	ecb_ctz64 (uint64_t x)
255	{	501	{
		502	#if 1400 <= _MSC_VER && (_M_X64 || _M_IA64 || _M_ARM)
		503	unsigned long r;
		504	_BitScanForward64 (&r, x);
		505	return (int)r;
		506	#else
256	int shift = x & 0xffffffffU ? 0 : 32;	507	int shift = x & 0xffffffff ? 0 : 32;
257	return ecb_ctz32 (x >> shift) + shift;	508	return ecb_ctz32 (x >> shift) + shift;
		509	#endif
258	}	510	}
259		511
260	ecb_function_ int ecb_popcount32 (uint32_t x) ecb_const;	512	ecb_function_ ecb_const int ecb_popcount32 (uint32_t x);
261	ecb_function_ int	513	ecb_function_ ecb_const int
262	ecb_popcount32 (uint32_t x)	514	ecb_popcount32 (uint32_t x)
263	{	515	{
264	x -= (x >> 1) & 0x55555555;	516	x -= (x >> 1) & 0x55555555;
265	x = ((x >> 2) & 0x33333333) + (x & 0x33333333);	517	x = ((x >> 2) & 0x33333333) + (x & 0x33333333);
266	x = ((x >> 4) + x) & 0x0f0f0f0f;	518	x = ((x >> 4) + x) & 0x0f0f0f0f;
267	x *= 0x01010101;	519	x *= 0x01010101;
268		520
269	return x >> 24;	521	return x >> 24;
270	}	522	}
271		523
272	ecb_function_ int ecb_ld32 (uint32_t x) ecb_const;	524	ecb_function_ ecb_const int ecb_ld32 (uint32_t x);
273	ecb_function_ int ecb_ld32 (uint32_t x)	525	ecb_function_ ecb_const int ecb_ld32 (uint32_t x)
274	{	526	{
		527	#if 1400 <= _MSC_VER && (_M_IX86 || _M_X64 || _M_IA64 || _M_ARM)
		528	unsigned long r;
		529	_BitScanReverse (&r, x);
		530	return (int)r;
		531	#else
275	int r = 0;	532	int r = 0;
276		533
277	if (x >> 16) { x >>= 16; r += 16; }	534	if (x >> 16) { x >>= 16; r += 16; }
278	if (x >> 8) { x >>= 8; r += 8; }	535	if (x >> 8) { x >>= 8; r += 8; }
279	if (x >> 4) { x >>= 4; r += 4; }	536	if (x >> 4) { x >>= 4; r += 4; }
280	if (x >> 2) { x >>= 2; r += 2; }	537	if (x >> 2) { x >>= 2; r += 2; }
281	if (x >> 1) { r += 1; }	538	if (x >> 1) { r += 1; }
282		539
283	return r;	540	return r;
		541	#endif
284	}	542	}
285		543
286	ecb_function_ int ecb_ld64 (uint64_t x) ecb_const;	544	ecb_function_ ecb_const int ecb_ld64 (uint64_t x);
287	ecb_function_ int ecb_ld64 (uint64_t x)	545	ecb_function_ ecb_const int ecb_ld64 (uint64_t x)
288	{	546	{
		547	#if 1400 <= _MSC_VER && (_M_X64 || _M_IA64 || _M_ARM)
		548	unsigned long r;
		549	_BitScanReverse64 (&r, x);
		550	return (int)r;
		551	#else
289	int r = 0;	552	int r = 0;
290		553
291	if (x >> 32) { x >>= 32; r += 32; }	554	if (x >> 32) { x >>= 32; r += 32; }
292		555
293	return r + ecb_ld32 (x);	556	return r + ecb_ld32 (x);
294	}
295	#endif	557	#endif
		558	}
		559	#endif
		560
		561	ecb_function_ ecb_const ecb_bool ecb_is_pot32 (uint32_t x);
		562	ecb_function_ ecb_const ecb_bool ecb_is_pot32 (uint32_t x) { return !(x & (x - 1)); }
		563	ecb_function_ ecb_const ecb_bool ecb_is_pot64 (uint64_t x);
		564	ecb_function_ ecb_const ecb_bool ecb_is_pot64 (uint64_t x) { return !(x & (x - 1)); }
		565
		566	ecb_function_ ecb_const uint8_t ecb_bitrev8 (uint8_t x);
		567	ecb_function_ ecb_const uint8_t ecb_bitrev8 (uint8_t x)
		568	{
		569	return ( (x * 0x0802U & 0x22110U)
		570	| (x * 0x8020U & 0x88440U)) * 0x10101U >> 16;
		571	}
		572
		573	ecb_function_ ecb_const uint16_t ecb_bitrev16 (uint16_t x);
		574	ecb_function_ ecb_const uint16_t ecb_bitrev16 (uint16_t x)
		575	{
		576	x = ((x >> 1) & 0x5555) | ((x & 0x5555) << 1);
		577	x = ((x >> 2) & 0x3333) | ((x & 0x3333) << 2);
		578	x = ((x >> 4) & 0x0f0f) | ((x & 0x0f0f) << 4);
		579	x = ( x >> 8 ) | ( x << 8);
		580
		581	return x;
		582	}
		583
		584	ecb_function_ ecb_const uint32_t ecb_bitrev32 (uint32_t x);
		585	ecb_function_ ecb_const uint32_t ecb_bitrev32 (uint32_t x)
		586	{
		587	x = ((x >> 1) & 0x55555555) | ((x & 0x55555555) << 1);
		588	x = ((x >> 2) & 0x33333333) | ((x & 0x33333333) << 2);
		589	x = ((x >> 4) & 0x0f0f0f0f) | ((x & 0x0f0f0f0f) << 4);
		590	x = ((x >> 8) & 0x00ff00ff) | ((x & 0x00ff00ff) << 8);
		591	x = ( x >> 16 ) | ( x << 16);
		592
		593	return x;
		594	}
296		595
297	/* popcount64 is only available on 64 bit cpus as gcc builtin */	596	/* popcount64 is only available on 64 bit cpus as gcc builtin */
298	/* so for this version we are lazy */	597	/* so for this version we are lazy */
299	ecb_function_ int ecb_popcount64 (uint64_t x) ecb_const;	598	ecb_function_ ecb_const int ecb_popcount64 (uint64_t x);
300	ecb_function_ int	599	ecb_function_ ecb_const int
301	ecb_popcount64 (uint64_t x)	600	ecb_popcount64 (uint64_t x)
302	{	601	{
303	return ecb_popcount32 (x) + ecb_popcount32 (x >> 32);	602	return ecb_popcount32 (x) + ecb_popcount32 (x >> 32);
304	}	603	}
305		604
306	ecb_inline uint8_t ecb_rotl8 (uint8_t x, unsigned int count) ecb_const;	605	ecb_inline ecb_const uint8_t ecb_rotl8 (uint8_t x, unsigned int count);
307	ecb_inline uint8_t ecb_rotr8 (uint8_t x, unsigned int count) ecb_const;	606	ecb_inline ecb_const uint8_t ecb_rotr8 (uint8_t x, unsigned int count);
308	ecb_inline uint16_t ecb_rotl16 (uint16_t x, unsigned int count) ecb_const;	607	ecb_inline ecb_const uint16_t ecb_rotl16 (uint16_t x, unsigned int count);
309	ecb_inline uint16_t ecb_rotr16 (uint16_t x, unsigned int count) ecb_const;	608	ecb_inline ecb_const uint16_t ecb_rotr16 (uint16_t x, unsigned int count);
310	ecb_inline uint32_t ecb_rotl32 (uint32_t x, unsigned int count) ecb_const;	609	ecb_inline ecb_const uint32_t ecb_rotl32 (uint32_t x, unsigned int count);
311	ecb_inline uint32_t ecb_rotr32 (uint32_t x, unsigned int count) ecb_const;	610	ecb_inline ecb_const uint32_t ecb_rotr32 (uint32_t x, unsigned int count);
312	ecb_inline uint64_t ecb_rotl64 (uint64_t x, unsigned int count) ecb_const;	611	ecb_inline ecb_const uint64_t ecb_rotl64 (uint64_t x, unsigned int count);
313	ecb_inline uint64_t ecb_rotr64 (uint64_t x, unsigned int count) ecb_const;	612	ecb_inline ecb_const uint64_t ecb_rotr64 (uint64_t x, unsigned int count);
314		613
315	ecb_inline uint8_t ecb_rotl8 (uint8_t x, unsigned int count) { return (x >> ( 8 - count)) | (x << count); }	614	ecb_inline ecb_const uint8_t ecb_rotl8 (uint8_t x, unsigned int count) { return (x >> (-count & 7)) | (x << (count & 7)); }
316	ecb_inline uint8_t ecb_rotr8 (uint8_t x, unsigned int count) { return (x << ( 8 - count)) | (x >> count); }	615	ecb_inline ecb_const uint8_t ecb_rotr8 (uint8_t x, unsigned int count) { return (x << (-count & 7)) | (x >> (count & 7)); }
317	ecb_inline uint16_t ecb_rotl16 (uint16_t x, unsigned int count) { return (x >> (16 - count)) | (x << count); }	616	ecb_inline ecb_const uint16_t ecb_rotl16 (uint16_t x, unsigned int count) { return (x >> (-count & 15)) | (x << (count & 15)); }
318	ecb_inline uint16_t ecb_rotr16 (uint16_t x, unsigned int count) { return (x << (16 - count)) | (x >> count); }	617	ecb_inline ecb_const uint16_t ecb_rotr16 (uint16_t x, unsigned int count) { return (x << (-count & 15)) | (x >> (count & 15)); }
319	ecb_inline uint32_t ecb_rotl32 (uint32_t x, unsigned int count) { return (x >> (32 - count)) | (x << count); }	618	ecb_inline ecb_const uint32_t ecb_rotl32 (uint32_t x, unsigned int count) { return (x >> (-count & 31)) | (x << (count & 31)); }
320	ecb_inline uint32_t ecb_rotr32 (uint32_t x, unsigned int count) { return (x << (32 - count)) | (x >> count); }	619	ecb_inline ecb_const uint32_t ecb_rotr32 (uint32_t x, unsigned int count) { return (x << (-count & 31)) | (x >> (count & 31)); }
321	ecb_inline uint64_t ecb_rotl64 (uint64_t x, unsigned int count) { return (x >> (64 - count)) | (x << count); }	620	ecb_inline ecb_const uint64_t ecb_rotl64 (uint64_t x, unsigned int count) { return (x >> (-count & 63)) | (x << (count & 63)); }
322	ecb_inline uint64_t ecb_rotr64 (uint64_t x, unsigned int count) { return (x << (64 - count)) | (x >> count); }	621	ecb_inline ecb_const uint64_t ecb_rotr64 (uint64_t x, unsigned int count) { return (x << (-count & 63)) | (x >> (count & 63)); }
323		622
324	#if ECB_GCC_VERSION(4,3)	623	#if ECB_CPP
		624
		625	inline uint8_t ecb_ctz (uint8_t v) { return ecb_ctz32 (v); }
		626	inline uint16_t ecb_ctz (uint16_t v) { return ecb_ctz32 (v); }
		627	inline uint32_t ecb_ctz (uint32_t v) { return ecb_ctz32 (v); }
		628	inline uint64_t ecb_ctz (uint64_t v) { return ecb_ctz64 (v); }
		629
		630	inline bool ecb_is_pot (uint8_t v) { return ecb_is_pot32 (v); }
		631	inline bool ecb_is_pot (uint16_t v) { return ecb_is_pot32 (v); }
		632	inline bool ecb_is_pot (uint32_t v) { return ecb_is_pot32 (v); }
		633	inline bool ecb_is_pot (uint64_t v) { return ecb_is_pot64 (v); }
		634
		635	inline int ecb_ld (uint8_t v) { return ecb_ld32 (v); }
		636	inline int ecb_ld (uint16_t v) { return ecb_ld32 (v); }
		637	inline int ecb_ld (uint32_t v) { return ecb_ld32 (v); }
		638	inline int ecb_ld (uint64_t v) { return ecb_ld64 (v); }
		639
		640	inline int ecb_popcount (uint8_t v) { return ecb_popcount32 (v); }
		641	inline int ecb_popcount (uint16_t v) { return ecb_popcount32 (v); }
		642	inline int ecb_popcount (uint32_t v) { return ecb_popcount32 (v); }
		643	inline int ecb_popcount (uint64_t v) { return ecb_popcount64 (v); }
		644
		645	inline uint8_t ecb_bitrev (uint8_t v) { return ecb_bitrev8 (v); }
		646	inline uint16_t ecb_bitrev (uint16_t v) { return ecb_bitrev16 (v); }
		647	inline uint32_t ecb_bitrev (uint32_t v) { return ecb_bitrev32 (v); }
		648
		649	inline uint8_t ecb_rotl (uint8_t v, unsigned int count) { return ecb_rotl8 (v, count); }
		650	inline uint16_t ecb_rotl (uint16_t v, unsigned int count) { return ecb_rotl16 (v, count); }
		651	inline uint32_t ecb_rotl (uint32_t v, unsigned int count) { return ecb_rotl32 (v, count); }
		652	inline uint64_t ecb_rotl (uint64_t v, unsigned int count) { return ecb_rotl64 (v, count); }
		653
		654	inline uint8_t ecb_rotr (uint8_t v, unsigned int count) { return ecb_rotr8 (v, count); }
		655	inline uint16_t ecb_rotr (uint16_t v, unsigned int count) { return ecb_rotr16 (v, count); }
		656	inline uint32_t ecb_rotr (uint32_t v, unsigned int count) { return ecb_rotr32 (v, count); }
		657	inline uint64_t ecb_rotr (uint64_t v, unsigned int count) { return ecb_rotr64 (v, count); }
		658
		659	#endif
		660
		661	#if ECB_GCC_VERSION(4,3) || (ECB_CLANG_BUILTIN(__builtin_bswap32) && ECB_CLANG_BUILTIN(__builtin_bswap64))
		662	#if ECB_GCC_VERSION(4,8) || ECB_CLANG_BUILTIN(__builtin_bswap16)
		663	#define ecb_bswap16(x) __builtin_bswap16 (x)
		664	#else
325	#define ecb_bswap16(x) (__builtin_bswap32 (x) >> 16)	665	#define ecb_bswap16(x) (__builtin_bswap32 (x) >> 16)
		666	#endif
326	#define ecb_bswap32(x) __builtin_bswap32 (x)	667	#define ecb_bswap32(x) __builtin_bswap32 (x)
327	#define ecb_bswap64(x) __builtin_bswap64 (x)	668	#define ecb_bswap64(x) __builtin_bswap64 (x)
		669	#elif _MSC_VER
		670	#include <stdlib.h>
		671	#define ecb_bswap16(x) ((uint16_t)_byteswap_ushort ((uint16_t)(x)))
		672	#define ecb_bswap32(x) ((uint32_t)_byteswap_ulong ((uint32_t)(x)))
		673	#define ecb_bswap64(x) ((uint64_t)_byteswap_uint64 ((uint64_t)(x)))
328	#else	674	#else
329	ecb_function_ uint16_t ecb_bswap16 (uint16_t x) ecb_const;	675	ecb_function_ ecb_const uint16_t ecb_bswap16 (uint16_t x);
330	ecb_function_ uint16_t	676	ecb_function_ ecb_const uint16_t
331	ecb_bswap16 (uint16_t x)	677	ecb_bswap16 (uint16_t x)
332	{	678	{
333	return ecb_rotl16 (x, 8);	679	return ecb_rotl16 (x, 8);
334	}	680	}
335		681
336	ecb_function_ uint32_t ecb_bswap32 (uint32_t x) ecb_const;	682	ecb_function_ ecb_const uint32_t ecb_bswap32 (uint32_t x);
337	ecb_function_ uint32_t	683	ecb_function_ ecb_const uint32_t
338	ecb_bswap32 (uint32_t x)	684	ecb_bswap32 (uint32_t x)
339	{	685	{
340	return (((uint32_t)ecb_bswap16 (x)) << 16) | ecb_bswap16 (x >> 16);	686	return (((uint32_t)ecb_bswap16 (x)) << 16) | ecb_bswap16 (x >> 16);
341	}	687	}
342		688
343	ecb_function_ uint64_t ecb_bswap64 (uint64_t x) ecb_const;	689	ecb_function_ ecb_const uint64_t ecb_bswap64 (uint64_t x);
344	ecb_function_ uint64_t	690	ecb_function_ ecb_const uint64_t
345	ecb_bswap64 (uint64_t x)	691	ecb_bswap64 (uint64_t x)
346	{	692	{
347	return (((uint64_t)ecb_bswap32 (x)) << 32) | ecb_bswap32 (x >> 32);	693	return (((uint64_t)ecb_bswap32 (x)) << 32) | ecb_bswap32 (x >> 32);
348	}	694	}
349	#endif	695	#endif
350		696
351	#if ECB_GCC_VERSION(4,5)	697	#if ECB_GCC_VERSION(4,5) || ECB_CLANG_BUILTIN(__builtin_unreachable)
352	#define ecb_unreachable() __builtin_unreachable ()	698	#define ecb_unreachable() __builtin_unreachable ()
353	#else	699	#else
354	/* this seems to work fine, but gcc always emits a warning for it :/ */	700	/* this seems to work fine, but gcc always emits a warning for it :/ */
355	ecb_function_ void ecb_unreachable (void) ecb_noreturn;	701	ecb_inline ecb_noreturn void ecb_unreachable (void);
356	ecb_function_ void ecb_unreachable (void) { }	702	ecb_inline ecb_noreturn void ecb_unreachable (void) { }
357	#endif	703	#endif
358		704
359	/* try to tell the compiler that some condition is definitely true */	705	/* try to tell the compiler that some condition is definitely true */
360	#define ecb_assume(cond) do { if (!(cond)) ecb_unreachable (); } while (0)	706	#define ecb_assume(cond) if (!(cond)) ecb_unreachable (); else 0
361		707
362	ecb_function_ unsigned char ecb_byteorder_helper (void) ecb_const;	708	ecb_inline ecb_const uint32_t ecb_byteorder_helper (void);
363	ecb_function_ unsigned char	709	ecb_inline ecb_const uint32_t
364	ecb_byteorder_helper (void)	710	ecb_byteorder_helper (void)
365	{	711	{
366	const uint32_t u = 0x11223344;	712	/* the union code still generates code under pressure in gcc, */
367	return *(unsigned char *)&u;	713	/* but less than using pointers, and always seems to */
		714	/* successfully return a constant. */
		715	/* the reason why we have this horrible preprocessor mess */
		716	/* is to avoid it in all cases, at least on common architectures */
		717	/* or when using a recent enough gcc version (>= 4.6) */
		718	#if (defined __BYTE_ORDER__ && __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__) \
		719	|| ((__i386 || __i386__ || _M_IX86 || ECB_GCC_AMD64 || ECB_MSVC_AMD64) && !__VOS__)
		720	#define ECB_LITTLE_ENDIAN 1
		721	return 0x44332211;
		722	#elif (defined __BYTE_ORDER__ && __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__) \
		723	|| ((__AARCH64EB__ || __MIPSEB__ || __ARMEB__) && !__VOS__)
		724	#define ECB_BIG_ENDIAN 1
		725	return 0x11223344;
		726	#else
		727	union
		728	{
		729	uint8_t c[4];
		730	uint32_t u;
		731	} u = { 0x11, 0x22, 0x33, 0x44 };
		732	return u.u;
		733	#endif
368	}	734	}
369		735
370	ecb_function_ ecb_bool ecb_big_endian (void) ecb_const;	736	ecb_inline ecb_const ecb_bool ecb_big_endian (void);
371	ecb_function_ ecb_bool ecb_big_endian (void) { return ecb_byteorder_helper () == 0x11; }	737	ecb_inline ecb_const ecb_bool ecb_big_endian (void) { return ecb_byteorder_helper () == 0x11223344; }
372	ecb_function_ ecb_bool ecb_little_endian (void) ecb_const;	738	ecb_inline ecb_const ecb_bool ecb_little_endian (void);
373	ecb_function_ ecb_bool ecb_little_endian (void) { return ecb_byteorder_helper () == 0x44; }	739	ecb_inline ecb_const ecb_bool ecb_little_endian (void) { return ecb_byteorder_helper () == 0x44332211; }
		740
		741	/*****************************************************************************/
		742	/* unaligned load/store */
		743
		744	ecb_inline uint_fast16_t ecb_be_u16_to_host (uint_fast16_t v) { return ecb_little_endian () ? ecb_bswap16 (v) : v; }
		745	ecb_inline uint_fast32_t ecb_be_u32_to_host (uint_fast32_t v) { return ecb_little_endian () ? ecb_bswap32 (v) : v; }
		746	ecb_inline uint_fast64_t ecb_be_u64_to_host (uint_fast64_t v) { return ecb_little_endian () ? ecb_bswap64 (v) : v; }
		747
		748	ecb_inline uint_fast16_t ecb_le_u16_to_host (uint_fast16_t v) { return ecb_big_endian () ? ecb_bswap16 (v) : v; }
		749	ecb_inline uint_fast32_t ecb_le_u32_to_host (uint_fast32_t v) { return ecb_big_endian () ? ecb_bswap32 (v) : v; }
		750	ecb_inline uint_fast64_t ecb_le_u64_to_host (uint_fast64_t v) { return ecb_big_endian () ? ecb_bswap64 (v) : v; }
		751
		752	ecb_inline uint_fast16_t ecb_peek_u16_u (const void *ptr) { uint16_t v; memcpy (&v, ptr, sizeof (v)); return v; }
		753	ecb_inline uint_fast32_t ecb_peek_u32_u (const void *ptr) { uint32_t v; memcpy (&v, ptr, sizeof (v)); return v; }
		754	ecb_inline uint_fast64_t ecb_peek_u64_u (const void *ptr) { uint64_t v; memcpy (&v, ptr, sizeof (v)); return v; }
		755
		756	ecb_inline uint_fast16_t ecb_peek_be_u16_u (const void *ptr) { return ecb_be_u16_to_host (ecb_peek_u16_u (ptr)); }
		757	ecb_inline uint_fast32_t ecb_peek_be_u32_u (const void *ptr) { return ecb_be_u32_to_host (ecb_peek_u32_u (ptr)); }
		758	ecb_inline uint_fast64_t ecb_peek_be_u64_u (const void *ptr) { return ecb_be_u64_to_host (ecb_peek_u64_u (ptr)); }
		759
		760	ecb_inline uint_fast16_t ecb_peek_le_u16_u (const void *ptr) { return ecb_le_u16_to_host (ecb_peek_u16_u (ptr)); }
		761	ecb_inline uint_fast32_t ecb_peek_le_u32_u (const void *ptr) { return ecb_le_u32_to_host (ecb_peek_u32_u (ptr)); }
		762	ecb_inline uint_fast64_t ecb_peek_le_u64_u (const void *ptr) { return ecb_le_u64_to_host (ecb_peek_u64_u (ptr)); }
		763
		764	ecb_inline uint_fast16_t ecb_host_to_be_u16 (uint_fast16_t v) { return ecb_little_endian () ? ecb_bswap16 (v) : v; }
		765	ecb_inline uint_fast32_t ecb_host_to_be_u32 (uint_fast32_t v) { return ecb_little_endian () ? ecb_bswap32 (v) : v; }
		766	ecb_inline uint_fast64_t ecb_host_to_be_u64 (uint_fast64_t v) { return ecb_little_endian () ? ecb_bswap64 (v) : v; }
		767
		768	ecb_inline uint_fast16_t ecb_host_to_le_u16 (uint_fast16_t v) { return ecb_big_endian () ? ecb_bswap16 (v) : v; }
		769	ecb_inline uint_fast32_t ecb_host_to_le_u32 (uint_fast32_t v) { return ecb_big_endian () ? ecb_bswap32 (v) : v; }
		770	ecb_inline uint_fast64_t ecb_host_to_le_u64 (uint_fast64_t v) { return ecb_big_endian () ? ecb_bswap64 (v) : v; }
		771
		772	ecb_inline void ecb_poke_u16_u (void *ptr, uint16_t v) { memcpy (ptr, &v, sizeof (v)); }
		773	ecb_inline void ecb_poke_u32_u (void *ptr, uint32_t v) { memcpy (ptr, &v, sizeof (v)); }
		774	ecb_inline void ecb_poke_u64_u (void *ptr, uint64_t v) { memcpy (ptr, &v, sizeof (v)); }
		775
		776	ecb_inline void ecb_poke_be_u16_u (void *ptr, uint_fast16_t v) { ecb_poke_u16_u (ptr, ecb_host_to_be_u16 (v)); }
		777	ecb_inline void ecb_poke_be_u32_u (void *ptr, uint_fast32_t v) { ecb_poke_u32_u (ptr, ecb_host_to_be_u32 (v)); }
		778	ecb_inline void ecb_poke_be_u64_u (void *ptr, uint_fast64_t v) { ecb_poke_u64_u (ptr, ecb_host_to_be_u64 (v)); }
		779
		780	ecb_inline void ecb_poke_le_u16_u (void *ptr, uint_fast16_t v) { ecb_poke_u16_u (ptr, ecb_host_to_le_u16 (v)); }
		781	ecb_inline void ecb_poke_le_u32_u (void *ptr, uint_fast32_t v) { ecb_poke_u32_u (ptr, ecb_host_to_le_u32 (v)); }
		782	ecb_inline void ecb_poke_le_u64_u (void *ptr, uint_fast64_t v) { ecb_poke_u64_u (ptr, ecb_host_to_le_u64 (v)); }
		783
		784	#if ECB_CPP
		785
		786	inline uint8_t ecb_bswap (uint8_t v) { return v; }
		787	inline uint16_t ecb_bswap (uint16_t v) { return ecb_bswap16 (v); }
		788	inline uint32_t ecb_bswap (uint32_t v) { return ecb_bswap32 (v); }
		789	inline uint64_t ecb_bswap (uint64_t v) { return ecb_bswap64 (v); }
		790
		791	template<typename T> inline T ecb_be_to_host (T v) { return ecb_little_endian () ? ecb_bswap (v) : v; }
		792	template<typename T> inline T ecb_le_to_host (T v) { return ecb_big_endian () ? ecb_bswap (v) : v; }
		793	template<typename T> inline T ecb_peek (const void *ptr) { return *(const T *)ptr; }
		794	template<typename T> inline T ecb_peek_be (const void *ptr) { return ecb_be_to_host (ecb_peek <T> (ptr)); }
		795	template<typename T> inline T ecb_peek_le (const void *ptr) { return ecb_le_to_host (ecb_peek <T> (ptr)); }
		796	template<typename T> inline T ecb_peek_u (const void *ptr) { T v; memcpy (&v, ptr, sizeof (v)); return v; }
		797	template<typename T> inline T ecb_peek_be_u (const void *ptr) { return ecb_be_to_host (ecb_peek_u<T> (ptr)); }
		798	template<typename T> inline T ecb_peek_le_u (const void *ptr) { return ecb_le_to_host (ecb_peek_u<T> (ptr)); }
		799
		800	template<typename T> inline T ecb_host_to_be (T v) { return ecb_little_endian () ? ecb_bswap (v) : v; }
		801	template<typename T> inline T ecb_host_to_le (T v) { return ecb_big_endian () ? ecb_bswap (v) : v; }
		802	template<typename T> inline void ecb_poke (void *ptr, T v) { *(T *)ptr = v; }
		803	template<typename T> inline void ecb_poke_be (void *ptr, T v) { return ecb_poke <T> (ptr, ecb_host_to_be (v)); }
		804	template<typename T> inline void ecb_poke_le (void *ptr, T v) { return ecb_poke <T> (ptr, ecb_host_to_le (v)); }
		805	template<typename T> inline void ecb_poke_u (void *ptr, T v) { memcpy (ptr, &v, sizeof (v)); }
		806	template<typename T> inline void ecb_poke_be_u (void *ptr, T v) { return ecb_poke_u<T> (ptr, ecb_host_to_be (v)); }
		807	template<typename T> inline void ecb_poke_le_u (void *ptr, T v) { return ecb_poke_u<T> (ptr, ecb_host_to_le (v)); }
		808
		809	#endif
		810
		811	/*****************************************************************************/
		812	/* pointer/integer hashing */
		813
		814	/* based on hash by Chris Wellons, https://nullprogram.com/blog/2018/07/31/ */
		815	ecb_function_ uint32_t ecb_mix32 (uint32_t v);
		816	ecb_function_ uint32_t ecb_mix32 (uint32_t v)
		817	{
		818	v ^= v >> 16; v *= 0x7feb352dU;
		819	v ^= v >> 15; v *= 0x846ca68bU;
		820	v ^= v >> 16;
		821	return v;
		822	}
		823
		824	ecb_function_ uint32_t ecb_unmix32 (uint32_t v);
		825	ecb_function_ uint32_t ecb_unmix32 (uint32_t v)
		826	{
		827	v ^= v >> 16 ; v *= 0x43021123U;
		828	v ^= v >> 15 ^ v >> 30; v *= 0x1d69e2a5U;
		829	v ^= v >> 16 ;
		830	return v;
		831	}
		832
		833	/* based on splitmix64, by Sebastiona Vigna, https://prng.di.unimi.it/splitmix64.c */
		834	ecb_function_ uint64_t ecb_mix64 (uint64_t v);
		835	ecb_function_ uint64_t ecb_mix64 (uint64_t v)
		836	{
		837	v ^= v >> 30; v *= 0xbf58476d1ce4e5b9U;
		838	v ^= v >> 27; v *= 0x94d049bb133111ebU;
		839	v ^= v >> 31;
		840	return v;
		841	}
		842
		843	ecb_function_ uint64_t ecb_unmix64 (uint64_t v);
		844	ecb_function_ uint64_t ecb_unmix64 (uint64_t v)
		845	{
		846	v ^= v >> 31 ^ v >> 62; v *= 0x319642b2d24d8ec3U;
		847	v ^= v >> 27 ^ v >> 54; v *= 0x96de1b173f119089U;
		848	v ^= v >> 30 ^ v >> 60;
		849	return v;
		850	}
		851
		852	ecb_function_ uintptr_t ecb_ptrmix (void *p);
		853	ecb_function_ uintptr_t ecb_ptrmix (void *p)
		854	{
		855	#if ECB_PTRSIZE <= 4
		856	return ecb_mix32 ((uint32_t)p);
		857	#else
		858	return ecb_mix64 ((uint64_t)p);
		859	#endif
		860	}
		861
		862	ecb_function_ void *ecb_ptrunmix (uintptr_t v);
		863	ecb_function_ void *ecb_ptrunmix (uintptr_t v)
		864	{
		865	#if ECB_PTRSIZE <= 4
		866	return (void *)ecb_unmix32 (v);
		867	#else
		868	return (void *)ecb_unmix64 (v);
		869	#endif
		870	}
		871
		872	#if ECB_CPP
		873
		874	template<typename T>
		875	inline uintptr_t ecb_ptrmix (T *p)
		876	{
		877	return ecb_ptrmix (static_cast<void *>(p));
		878	}
		879
		880	template<typename T>
		881	inline T *ecb_ptrunmix (uintptr_t v)
		882	{
		883	return static_cast<T *>(ecb_ptrunmix (v));
		884	}
		885
		886	#endif
		887
		888	/*****************************************************************************/
		889	/* gray code */
		890
		891	ecb_function_ uint_fast8_t ecb_gray8_encode (uint_fast8_t b) { return b ^ (b >> 1); }
		892	ecb_function_ uint_fast16_t ecb_gray16_encode (uint_fast16_t b) { return b ^ (b >> 1); }
		893	ecb_function_ uint_fast32_t ecb_gray32_encode (uint_fast32_t b) { return b ^ (b >> 1); }
		894	ecb_function_ uint_fast64_t ecb_gray64_encode (uint_fast64_t b) { return b ^ (b >> 1); }
		895
		896	ecb_function_ uint8_t ecb_gray8_decode (uint8_t g)
		897	{
		898	g ^= g >> 1;
		899	g ^= g >> 2;
		900	g ^= g >> 4;
		901
		902	return g;
		903	}
		904
		905	ecb_function_ uint16_t ecb_gray16_decode (uint16_t g)
		906	{
		907	g ^= g >> 1;
		908	g ^= g >> 2;
		909	g ^= g >> 4;
		910	g ^= g >> 8;
		911
		912	return g;
		913	}
		914
		915	ecb_function_ uint32_t ecb_gray32_decode (uint32_t g)
		916	{
		917	g ^= g >> 1;
		918	g ^= g >> 2;
		919	g ^= g >> 4;
		920	g ^= g >> 8;
		921	g ^= g >> 16;
		922
		923	return g;
		924	}
		925
		926	ecb_function_ uint64_t ecb_gray64_decode (uint64_t g)
		927	{
		928	g ^= g >> 1;
		929	g ^= g >> 2;
		930	g ^= g >> 4;
		931	g ^= g >> 8;
		932	g ^= g >> 16;
		933	g ^= g >> 32;
		934
		935	return g;
		936	}
		937
		938	#if ECB_CPP
		939
		940	ecb_function_ uint8_t ecb_gray_encode (uint8_t b) { return ecb_gray8_encode (b); }
		941	ecb_function_ uint16_t ecb_gray_encode (uint16_t b) { return ecb_gray16_encode (b); }
		942	ecb_function_ uint32_t ecb_gray_encode (uint32_t b) { return ecb_gray32_encode (b); }
		943	ecb_function_ uint64_t ecb_gray_encode (uint64_t b) { return ecb_gray64_encode (b); }
		944
		945	ecb_function_ uint8_t ecb_gray_decode (uint8_t g) { return ecb_gray8_decode (g); }
		946	ecb_function_ uint16_t ecb_gray_decode (uint16_t g) { return ecb_gray16_decode (g); }
		947	ecb_function_ uint32_t ecb_gray_decode (uint32_t g) { return ecb_gray32_decode (g); }
		948	ecb_function_ uint64_t ecb_gray_decode (uint64_t g) { return ecb_gray64_decode (g); }
		949
		950	#endif
		951
		952	/*****************************************************************************/
		953	/* division */
374		954
375	#if ECB_GCC_VERSION(3,0) || ECB_C99	955	#if ECB_GCC_VERSION(3,0) || ECB_C99
		956	/* C99 tightened the definition of %, so we can use a more efficient version */
376	#define ecb_mod(m,n) ((m) % (n) + ((m) % (n) < 0 ? (n) : 0))	957	#define ecb_mod(m,n) ((m) % (n) + ((m) % (n) < 0 ? (n) : 0))
377	#else	958	#else
378	#define ecb_mod(m,n) ((m) < 0 ? ((n) - 1 - ((-1 - (m)) % (n))) : ((m) % (n)))	959	#define ecb_mod(m,n) ((m) < 0 ? ((n) - 1 - ((-1 - (m)) % (n))) : ((m) % (n)))
379	#endif	960	#endif
		961
		962	#if ECB_CPP
		963	template<typename T>
		964	static inline T ecb_div_rd (T val, T div)
		965	{
		966	return val < 0 ? - ((-val + div - 1) / div) : (val ) / div;
		967	}
		968	template<typename T>
		969	static inline T ecb_div_ru (T val, T div)
		970	{
		971	return val < 0 ? - ((-val ) / div) : (val + div - 1) / div;
		972	}
		973	#else
		974	#define ecb_div_rd(val,div) ((val) < 0 ? - ((-(val) + (div) - 1) / (div)) : ((val) ) / (div))
		975	#define ecb_div_ru(val,div) ((val) < 0 ? - ((-(val) ) / (div)) : ((val) + (div) - 1) / (div))
		976	#endif
		977
		978	/*****************************************************************************/
		979	/* array length */
380		980
381	#if ecb_cplusplus_does_not_suck	981	#if ecb_cplusplus_does_not_suck
382	/* does not work for local types (http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2657.htm) */	982	/* does not work for local types (http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2657.htm) */
383	template<typename T, int N>	983	template<typename T, int N>
384	static inline int ecb_array_length (const T (&arr)[N])	984	static inline int ecb_array_length (const T (&arr)[N])
…		…
387	}	987	}
388	#else	988	#else
389	#define ecb_array_length(name) (sizeof (name) / sizeof (name [0]))	989	#define ecb_array_length(name) (sizeof (name) / sizeof (name [0]))
390	#endif	990	#endif
391		991
		992	/*****************************************************************************/
		993	/* IEEE 754-2008 half float conversions */
		994
		995	ecb_function_ ecb_const uint32_t ecb_binary16_to_binary32 (uint32_t x);
		996	ecb_function_ ecb_const uint32_t
		997	ecb_binary16_to_binary32 (uint32_t x)
		998	{
		999	unsigned int s = (x & 0x8000) << (31 - 15);
		1000	int e = (x >> 10) & 0x001f;
		1001	unsigned int m = x & 0x03ff;
		1002
		1003	if (ecb_expect_false (e == 31))
		1004	/* infinity or NaN */
		1005	e = 255 - (127 - 15);
		1006	else if (ecb_expect_false (!e))
		1007	{
		1008	if (ecb_expect_true (!m))
		1009	/* zero, handled by code below by forcing e to 0 */
		1010	e = 0 - (127 - 15);
		1011	else
		1012	{
		1013	/* subnormal, renormalise */
		1014	unsigned int s = 10 - ecb_ld32 (m);
		1015
		1016	m = (m << s) & 0x3ff; /* mask implicit bit */
		1017	e -= s - 1;
		1018	}
		1019	}
		1020
		1021	/* e and m now are normalised, or zero, (or inf or nan) */
		1022	e += 127 - 15;
		1023
		1024	return s | (e << 23) | (m << (23 - 10));
		1025	}
		1026
		1027	ecb_function_ ecb_const uint16_t ecb_binary32_to_binary16 (uint32_t x);
		1028	ecb_function_ ecb_const uint16_t
		1029	ecb_binary32_to_binary16 (uint32_t x)
		1030	{
		1031	unsigned int s = (x >> 16) & 0x00008000; /* sign bit, the easy part */
		1032	int e = ((x >> 23) & 0x000000ff) - (127 - 15); /* the desired exponent */
		1033	unsigned int m = x & 0x007fffff;
		1034
		1035	x &= 0x7fffffff;
		1036
		1037	/* if it's within range of binary16 normals, use fast path */
		1038	if (ecb_expect_true (0x38800000 <= x && x <= 0x477fefff))
		1039	{
		1040	/* mantissa round-to-even */
		1041	m += 0x00000fff + ((m >> (23 - 10)) & 1);
		1042
		1043	/* handle overflow */
		1044	if (ecb_expect_false (m >= 0x00800000))
		1045	{
		1046	m >>= 1;
		1047	e += 1;
		1048	}
		1049
		1050	return s | (e << 10) | (m >> (23 - 10));
		1051	}
		1052
		1053	/* handle large numbers and infinity */
		1054	if (ecb_expect_true (0x477fefff < x && x <= 0x7f800000))
		1055	return s | 0x7c00;
		1056
		1057	/* handle zero, subnormals and small numbers */
		1058	if (ecb_expect_true (x < 0x38800000))
		1059	{
		1060	/* zero */
		1061	if (ecb_expect_true (!x))
		1062	return s;
		1063
		1064	/* handle subnormals */
		1065
		1066	/* too small, will be zero */
		1067	if (e < (14 - 24)) /* might not be sharp, but is good enough */
		1068	return s;
		1069
		1070	m |= 0x00800000; /* make implicit bit explicit */
		1071
		1072	/* very tricky - we need to round to the nearest e (+10) bit value */
		1073	{
		1074	unsigned int bits = 14 - e;
		1075	unsigned int half = (1 << (bits - 1)) - 1;
		1076	unsigned int even = (m >> bits) & 1;
		1077
		1078	/* if this overflows, we will end up with a normalised number */
		1079	m = (m + half + even) >> bits;
		1080	}
		1081
		1082	return s | m;
		1083	}
		1084
		1085	/* handle NaNs, preserve leftmost nan bits, but make sure we don't turn them into infinities */
		1086	m >>= 13;
		1087
		1088	return s | 0x7c00 | m | !m;
		1089	}
		1090
		1091	/*******************************************************************************/
		1092	/* fast integer to ascii */
		1093
		1094	/*
		1095	* This code is pretty complicated because it is general. The idea behind it,
		1096	* however, is pretty simple: first, the number is multiplied with a scaling
		1097	* factor (2**bits / 10**(digits-1)) to convert the integer into a fixed-point
		1098	* number with the first digit in the upper bits.
		1099	* Then this digit is converted to text and masked out. The resulting number
		1100	* is then multiplied by 10, by multiplying the fixed point representation
		1101	* by 5 and shifting the (binary) decimal point one to the right, so a 4.28
		1102	* format becomes 5.27, 6.26 and so on.
		1103	* The rest involves only advancing the pointer if we already generated a
		1104	* non-zero digit, so leading zeroes are overwritten.
		1105	*/
		1106
		1107	/* simply return a mask with "bits" bits set */
		1108	#define ecb_i2a_mask(type,bits) ((((type)1) << (bits)) - 1)
		1109
		1110	/* oputput a single digit. maskvalue is 10**digitidx */
		1111	#define ecb_i2a_digit(type,bits,digitmask,maskvalue,digitidx) \
		1112	if (digitmask >= maskvalue) /* constant, used to decide how many digits to generate */ \
		1113	{ \
		1114	char digit = x >> (bits - digitidx); /* calculate the topmost digit */ \
		1115	*ptr = digit + '0'; /* output it */ \
		1116	nz = (digitmask == maskvalue) || nz || digit; /* first term == always output last digit */ \
		1117	ptr += nz; /* output digit only if non-zero digit seen */ \
		1118	x = (x & ecb_i2a_mask (type, bits - digitidx)) * 5; /* *10, but shift decimal point right */ \
		1119	}
		1120
		1121	/* convert integer to fixed point format and multiply out digits, highest first */
		1122	/* requires magic constants: max. digits and number of bits after the decimal point */
		1123	#define ecb_i2a_def(suffix,ptr,v,type,bits,digitmask,lz) \
		1124	ecb_inline char *ecb_i2a_ ## suffix (char *ptr, uint32_t u) \
		1125	{ \
		1126	char nz = lz; /* non-zero digit seen? */ \
		1127	/* convert to x.bits fixed-point */ \
		1128	type x = u * ((ecb_i2a_mask (type, bits) + digitmask) / digitmask); \
		1129	/* output up to 10 digits */ \
		1130	ecb_i2a_digit (type,bits,digitmask, 1, 0); \
		1131	ecb_i2a_digit (type,bits,digitmask, 10, 1); \
		1132	ecb_i2a_digit (type,bits,digitmask, 100, 2); \
		1133	ecb_i2a_digit (type,bits,digitmask, 1000, 3); \
		1134	ecb_i2a_digit (type,bits,digitmask, 10000, 4); \
		1135	ecb_i2a_digit (type,bits,digitmask, 100000, 5); \
		1136	ecb_i2a_digit (type,bits,digitmask, 1000000, 6); \
		1137	ecb_i2a_digit (type,bits,digitmask, 10000000, 7); \
		1138	ecb_i2a_digit (type,bits,digitmask, 100000000, 8); \
		1139	ecb_i2a_digit (type,bits,digitmask, 1000000000, 9); \
		1140	return ptr; \
		1141	}
		1142
		1143	/* predefined versions of the above, for various digits */
		1144	/* ecb_i2a_xN = almost N digits, limit defined by macro */
		1145	/* ecb_i2a_N = up to N digits, leading zeroes suppressed */
		1146	/* ecb_i2a_0N = exactly N digits, including leading zeroes */
		1147
		1148	/* non-leading-zero versions, limited range */
		1149	#define ECB_I2A_MAX_X5 59074 /* limit for ecb_i2a_x5 */
		1150	#define ECB_I2A_MAX_X10 2932500665 /* limit for ecb_i2a_x10 */
		1151	ecb_i2a_def ( x5, ptr, v, uint32_t, 26, 10000, 0)
		1152	ecb_i2a_def (x10, ptr, v, uint64_t, 60, 1000000000, 0)
		1153
		1154	/* non-leading zero versions, all digits, 4 and 9 are optimal for 32/64 bit */
		1155	ecb_i2a_def ( 2, ptr, v, uint32_t, 10, 10, 0)
		1156	ecb_i2a_def ( 3, ptr, v, uint32_t, 12, 100, 0)
		1157	ecb_i2a_def ( 4, ptr, v, uint32_t, 26, 1000, 0)
		1158	ecb_i2a_def ( 5, ptr, v, uint64_t, 30, 10000, 0)
		1159	ecb_i2a_def ( 6, ptr, v, uint64_t, 36, 100000, 0)
		1160	ecb_i2a_def ( 7, ptr, v, uint64_t, 44, 1000000, 0)
		1161	ecb_i2a_def ( 8, ptr, v, uint64_t, 50, 10000000, 0)
		1162	ecb_i2a_def ( 9, ptr, v, uint64_t, 56, 100000000, 0)
		1163
		1164	/* leading-zero versions, all digits, 04 and 09 are optimal for 32/64 bit */
		1165	ecb_i2a_def (02, ptr, v, uint32_t, 10, 10, 1)
		1166	ecb_i2a_def (03, ptr, v, uint32_t, 12, 100, 1)
		1167	ecb_i2a_def (04, ptr, v, uint32_t, 26, 1000, 1)
		1168	ecb_i2a_def (05, ptr, v, uint64_t, 30, 10000, 1)
		1169	ecb_i2a_def (06, ptr, v, uint64_t, 36, 100000, 1)
		1170	ecb_i2a_def (07, ptr, v, uint64_t, 44, 1000000, 1)
		1171	ecb_i2a_def (08, ptr, v, uint64_t, 50, 10000000, 1)
		1172	ecb_i2a_def (09, ptr, v, uint64_t, 56, 100000000, 1)
		1173
		1174	#define ECB_I2A_I32_DIGITS 11
		1175	#define ECB_I2A_U32_DIGITS 10
		1176	#define ECB_I2A_I64_DIGITS 20
		1177	#define ECB_I2A_U64_DIGITS 21
		1178	#define ECB_I2A_MAX_DIGITS 21
		1179
		1180	ecb_inline char *
		1181	ecb_i2a_u32 (char *ptr, uint32_t u)
		1182	{
		1183	#if ECB_64BIT_NATIVE
		1184	if (ecb_expect_true (u <= ECB_I2A_MAX_X10))
		1185	ptr = ecb_i2a_x10 (ptr, u);
		1186	else /* x10 almost, but not fully, covers 32 bit */
		1187	{
		1188	uint32_t u1 = u % 1000000000;
		1189	uint32_t u2 = u / 1000000000;
		1190
		1191	*ptr++ = u2 + '0';
		1192	ptr = ecb_i2a_09 (ptr, u1);
		1193	}
		1194	#else
		1195	if (ecb_expect_true (u <= ECB_I2A_MAX_X5))
		1196	ecb_i2a_x5 (ptr, u);
		1197	else if (ecb_expect_true (u <= ECB_I2A_MAX_X5 * 10000))
		1198	{
		1199	uint32_t u1 = u % 10000;
		1200	uint32_t u2 = u / 10000;
		1201
		1202	ptr = ecb_i2a_x5 (ptr, u2);
		1203	ptr = ecb_i2a_04 (ptr, u1);
		1204	}
		1205	else
		1206	{
		1207	uint32_t u1 = u % 10000;
		1208	uint32_t ua = u / 10000;
		1209	uint32_t u2 = ua % 10000;
		1210	uint32_t u3 = ua / 10000;
		1211
		1212	ptr = ecb_i2a_2 (ptr, u3);
		1213	ptr = ecb_i2a_04 (ptr, u2);
		1214	ptr = ecb_i2a_04 (ptr, u1);
		1215	}
392	#endif	1216	#endif
393		1217
		1218	return ptr;
		1219	}
		1220
		1221	ecb_inline char *
		1222	ecb_i2a_i32 (char *ptr, int32_t v)
		1223	{
		1224	*ptr = '-'; ptr += v < 0;
		1225	uint32_t u = v < 0 ? -(uint32_t)v : v;
		1226
		1227	#if ECB_64BIT_NATIVE
		1228	ptr = ecb_i2a_x10 (ptr, u); /* x10 fully covers 31 bit */
		1229	#else
		1230	ptr = ecb_i2a_u32 (ptr, u);
		1231	#endif
		1232
		1233	return ptr;
		1234	}
		1235
		1236	ecb_inline char *
		1237	ecb_i2a_u64 (char *ptr, uint64_t u)
		1238	{
		1239	#if ECB_64BIT_NATIVE
		1240	if (ecb_expect_true (u <= ECB_I2A_MAX_X10))
		1241	ptr = ecb_i2a_x10 (ptr, u);
		1242	else if (ecb_expect_false (u <= ECB_I2A_MAX_X10 * 1000000000))
		1243	{
		1244	uint64_t u1 = u % 1000000000;
		1245	uint64_t u2 = u / 1000000000;
		1246
		1247	ptr = ecb_i2a_x10 (ptr, u2);
		1248	ptr = ecb_i2a_09 (ptr, u1);
		1249	}
		1250	else
		1251	{
		1252	uint64_t u1 = u % 1000000000;
		1253	uint64_t ua = u / 1000000000;
		1254	uint64_t u2 = ua % 1000000000;
		1255	uint64_t u3 = ua / 1000000000;
		1256
		1257	ptr = ecb_i2a_2 (ptr, u3);
		1258	ptr = ecb_i2a_09 (ptr, u2);
		1259	ptr = ecb_i2a_09 (ptr, u1);
		1260	}
		1261	#else
		1262	if (ecb_expect_true (u <= ECB_I2A_MAX_X5))
		1263	ptr = ecb_i2a_x5 (ptr, u);
		1264	else
		1265	{
		1266	uint64_t u1 = u % 10000;
		1267	uint64_t u2 = u / 10000;
		1268
		1269	ptr = ecb_i2a_u64 (ptr, u2);
		1270	ptr = ecb_i2a_04 (ptr, u1);
		1271	}
		1272	#endif
		1273
		1274	return ptr;
		1275	}
		1276
		1277	ecb_inline char *
		1278	ecb_i2a_i64 (char *ptr, int64_t v)
		1279	{
		1280	*ptr = '-'; ptr += v < 0;
		1281	uint64_t u = v < 0 ? -(uint64_t)v : v;
		1282
		1283	#if ECB_64BIT_NATIVE
		1284	if (ecb_expect_true (u <= ECB_I2A_MAX_X10))
		1285	ptr = ecb_i2a_x10 (ptr, u);
		1286	else if (ecb_expect_false (u <= ECB_I2A_MAX_X10 * 1000000000))
		1287	{
		1288	uint64_t u1 = u % 1000000000;
		1289	uint64_t u2 = u / 1000000000;
		1290
		1291	ptr = ecb_i2a_x10 (ptr, u2);
		1292	ptr = ecb_i2a_09 (ptr, u1);
		1293	}
		1294	else
		1295	{
		1296	uint64_t u1 = u % 1000000000;
		1297	uint64_t ua = u / 1000000000;
		1298	uint64_t u2 = ua % 1000000000;
		1299	uint64_t u3 = ua / 1000000000;
		1300
		1301	/* 2**31 is 19 digits, so the top is exactly one digit */
		1302	*ptr++ = u3 + '0';
		1303	ptr = ecb_i2a_09 (ptr, u2);
		1304	ptr = ecb_i2a_09 (ptr, u1);
		1305	}
		1306	#else
		1307	ptr = ecb_i2a_u64 (ptr, u);
		1308	#endif
		1309
		1310	return ptr;
		1311	}
		1312
		1313	/*******************************************************************************/
		1314	/* floating point stuff, can be disabled by defining ECB_NO_LIBM */
		1315
		1316	/* basically, everything uses "ieee pure-endian" floating point numbers */
		1317	/* the only noteworthy exception is ancient armle, which uses order 43218765 */
		1318	#if 0 \
		1319	|| __i386 || __i386__ \
		1320	|| ECB_GCC_AMD64 \
		1321	|| __powerpc__ || __ppc__ || __powerpc64__ || __ppc64__ \
		1322	|| defined __s390__ || defined __s390x__ \
		1323	|| defined __mips__ \
		1324	|| defined __alpha__ \
		1325	|| defined __hppa__ \
		1326	|| defined __ia64__ \
		1327	|| defined __m68k__ \
		1328	|| defined __m88k__ \
		1329	|| defined __sh__ \
		1330	|| defined _M_IX86 || defined ECB_MSVC_AMD64 || defined _M_IA64 \
		1331	|| (defined __arm__ && (defined __ARM_EABI__ || defined __EABI__ || defined __VFP_FP__ || defined _WIN32_WCE || defined __ANDROID__)) \
		1332	|| defined __aarch64__
		1333	#define ECB_STDFP 1
		1334	#else
		1335	#define ECB_STDFP 0
		1336	#endif
		1337
		1338	#ifndef ECB_NO_LIBM
		1339
		1340	#include <math.h> /* for frexp*, ldexp*, INFINITY, NAN */
		1341
		1342	/* only the oldest of old doesn't have this one. solaris. */
		1343	#ifdef INFINITY
		1344	#define ECB_INFINITY INFINITY
		1345	#else
		1346	#define ECB_INFINITY HUGE_VAL
		1347	#endif
		1348
		1349	#ifdef NAN
		1350	#define ECB_NAN NAN
		1351	#else
		1352	#define ECB_NAN ECB_INFINITY
		1353	#endif
		1354
		1355	#if ECB_C99 || _XOPEN_VERSION >= 600 || _POSIX_VERSION >= 200112L
		1356	#define ecb_ldexpf(x,e) ldexpf ((x), (e))
		1357	#define ecb_frexpf(x,e) frexpf ((x), (e))
		1358	#else
		1359	#define ecb_ldexpf(x,e) (float) ldexp ((double) (x), (e))
		1360	#define ecb_frexpf(x,e) (float) frexp ((double) (x), (e))
		1361	#endif
		1362
		1363	/* convert a float to ieee single/binary32 */
		1364	ecb_function_ ecb_const uint32_t ecb_float_to_binary32 (float x);
		1365	ecb_function_ ecb_const uint32_t
		1366	ecb_float_to_binary32 (float x)
		1367	{
		1368	uint32_t r;
		1369
		1370	#if ECB_STDFP
		1371	memcpy (&r, &x, 4);
		1372	#else
		1373	/* slow emulation, works for anything but -0 */
		1374	uint32_t m;
		1375	int e;
		1376
		1377	if (x == 0e0f ) return 0x00000000U;
		1378	if (x > +3.40282346638528860e+38f) return 0x7f800000U;
		1379	if (x < -3.40282346638528860e+38f) return 0xff800000U;
		1380	if (x != x ) return 0x7fbfffffU;
		1381
		1382	m = ecb_frexpf (x, &e) * 0x1000000U;
		1383
		1384	r = m & 0x80000000U;
		1385
		1386	if (r)
		1387	m = -m;
		1388
		1389	if (e <= -126)
		1390	{
		1391	m &= 0xffffffU;
		1392	m >>= (-125 - e);
		1393	e = -126;
		1394	}
		1395
		1396	r |= (e + 126) << 23;
		1397	r |= m & 0x7fffffU;
		1398	#endif
		1399
		1400	return r;
		1401	}
		1402
		1403	/* converts an ieee single/binary32 to a float */
		1404	ecb_function_ ecb_const float ecb_binary32_to_float (uint32_t x);
		1405	ecb_function_ ecb_const float
		1406	ecb_binary32_to_float (uint32_t x)
		1407	{
		1408	float r;
		1409
		1410	#if ECB_STDFP
		1411	memcpy (&r, &x, 4);
		1412	#else
		1413	/* emulation, only works for normals and subnormals and +0 */
		1414	int neg = x >> 31;
		1415	int e = (x >> 23) & 0xffU;
		1416
		1417	x &= 0x7fffffU;
		1418
		1419	if (e)
		1420	x |= 0x800000U;
		1421	else
		1422	e = 1;
		1423
		1424	/* we distrust ldexpf a bit and do the 2**-24 scaling by an extra multiply */
		1425	r = ecb_ldexpf (x * (0.5f / 0x800000U), e - 126);
		1426
		1427	r = neg ? -r : r;
		1428	#endif
		1429
		1430	return r;
		1431	}
		1432
		1433	/* convert a double to ieee double/binary64 */
		1434	ecb_function_ ecb_const uint64_t ecb_double_to_binary64 (double x);
		1435	ecb_function_ ecb_const uint64_t
		1436	ecb_double_to_binary64 (double x)
		1437	{
		1438	uint64_t r;
		1439
		1440	#if ECB_STDFP
		1441	memcpy (&r, &x, 8);
		1442	#else
		1443	/* slow emulation, works for anything but -0 */
		1444	uint64_t m;
		1445	int e;
		1446
		1447	if (x == 0e0 ) return 0x0000000000000000U;
		1448	if (x > +1.79769313486231470e+308) return 0x7ff0000000000000U;
		1449	if (x < -1.79769313486231470e+308) return 0xfff0000000000000U;
		1450	if (x != x ) return 0X7ff7ffffffffffffU;
		1451
		1452	m = frexp (x, &e) * 0x20000000000000U;
		1453
		1454	r = m & 0x8000000000000000;;
		1455
		1456	if (r)
		1457	m = -m;
		1458
		1459	if (e <= -1022)
		1460	{
		1461	m &= 0x1fffffffffffffU;
		1462	m >>= (-1021 - e);
		1463	e = -1022;
		1464	}
		1465
		1466	r |= ((uint64_t)(e + 1022)) << 52;
		1467	r |= m & 0xfffffffffffffU;
		1468	#endif
		1469
		1470	return r;
		1471	}
		1472
		1473	/* converts an ieee double/binary64 to a double */
		1474	ecb_function_ ecb_const double ecb_binary64_to_double (uint64_t x);
		1475	ecb_function_ ecb_const double
		1476	ecb_binary64_to_double (uint64_t x)
		1477	{
		1478	double r;
		1479
		1480	#if ECB_STDFP
		1481	memcpy (&r, &x, 8);
		1482	#else
		1483	/* emulation, only works for normals and subnormals and +0 */
		1484	int neg = x >> 63;
		1485	int e = (x >> 52) & 0x7ffU;
		1486
		1487	x &= 0xfffffffffffffU;
		1488
		1489	if (e)
		1490	x |= 0x10000000000000U;
		1491	else
		1492	e = 1;
		1493
		1494	/* we distrust ldexp a bit and do the 2**-53 scaling by an extra multiply */
		1495	r = ldexp (x * (0.5 / 0x10000000000000U), e - 1022);
		1496
		1497	r = neg ? -r : r;
		1498	#endif
		1499
		1500	return r;
		1501	}
		1502
		1503	/* convert a float to ieee half/binary16 */
		1504	ecb_function_ ecb_const uint16_t ecb_float_to_binary16 (float x);
		1505	ecb_function_ ecb_const uint16_t
		1506	ecb_float_to_binary16 (float x)
		1507	{
		1508	return ecb_binary32_to_binary16 (ecb_float_to_binary32 (x));
		1509	}
		1510
		1511	/* convert an ieee half/binary16 to float */
		1512	ecb_function_ ecb_const float ecb_binary16_to_float (uint16_t x);
		1513	ecb_function_ ecb_const float
		1514	ecb_binary16_to_float (uint16_t x)
		1515	{
		1516	return ecb_binary32_to_float (ecb_binary16_to_binary32 (x));
		1517	}
		1518
		1519	#endif
		1520
		1521	#endif
		1522

Diff Legend

–	Removed lines
+	Added lines
<	Changed lines
>	Changed lines