[ViewVC] Diff of: cvs/libecb/ecb.h

Comparing libecb/ecb.h (file contents):
Revision 1.168 by root, Sat Nov 21 18:09:36 2015 UTC vs.
Revision 1.204 by root, Fri Mar 25 08:44:14 2022 UTC

…		…
1	/*	1	/*
2	* libecb - http://software.schmorp.de/pkg/libecb	2	* libecb - http://software.schmorp.de/pkg/libecb
3	*	3	*
4	* Copyright (©) 2009-2015 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:
40		40
41	#ifndef ECB_H	41	#ifndef ECB_H
42	#define ECB_H	42	#define ECB_H
43		43
44	/* 16 bits major, 16 bits minor */	44	/* 16 bits major, 16 bits minor */
45	#define ECB_VERSION 0x00010005	45	#define ECB_VERSION 0x0001000c
46		46
47	#ifdef _WIN32	47	#include <string.h> /* for memcpy */
		48
		49	#if defined (_WIN32) && !defined (__MINGW32__)
48	typedef signed char int8_t;	50	typedef signed char int8_t;
49	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;
50	typedef signed short int16_t;	54	typedef signed short int16_t;
51	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;
52	typedef signed int int32_t;	58	typedef signed int int32_t;
53	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;
54	#if __GNUC__	62	#if __GNUC__
55	typedef signed long long int64_t;	63	typedef signed long long int64_t;
56	typedef unsigned long long uint64_t;	64	typedef unsigned long long uint64_t;
57	#else /* _MSC_VER \|\| __BORLANDC__ */	65	#else /* _MSC_VER \|\| __BORLANDC__ */
58	typedef signed __int64 int64_t;	66	typedef signed __int64 int64_t;
59	typedef unsigned __int64 uint64_t;	67	typedef unsigned __int64 uint64_t;
60	#endif	68	#endif
		69	typedef int64_t int_fast64_t;
		70	typedef uint64_t uint_fast64_t;
61	#ifdef _WIN64	71	#ifdef _WIN64
62	#define ECB_PTRSIZE 8	72	#define ECB_PTRSIZE 8
63	typedef uint64_t uintptr_t;	73	typedef uint64_t uintptr_t;
64	typedef int64_t intptr_t;	74	typedef int64_t intptr_t;
65	#else	75	#else
…		…
67	typedef uint32_t uintptr_t;	77	typedef uint32_t uintptr_t;
68	typedef int32_t intptr_t;	78	typedef int32_t intptr_t;
69	#endif	79	#endif
70	#else	80	#else
71	#include <inttypes.h>	81	#include <inttypes.h>
72	#if UINTMAX_MAX > 0xffffffffU	82	#if (defined INTPTR_MAX ? INTPTR_MAX : ULONG_MAX) > 0xffffffffU
73	#define ECB_PTRSIZE 8	83	#define ECB_PTRSIZE 8
74	#else	84	#else
75	#define ECB_PTRSIZE 4	85	#define ECB_PTRSIZE 4
76	#endif	86	#endif
77	#endif	87	#endif
78		88
79	#define ECB_GCC_AMD64 (__amd64 \|\| __amd64__ \|\| __x86_64 \|\| __x86_64__)	89	#define ECB_GCC_AMD64 (__amd64 \|\| __amd64__ \|\| __x86_64 \|\| __x86_64__)
80	#define ECB_MSVC_AMD64 (_M_AMD64 \|\| _M_X64)	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
81		99
82	/* work around x32 idiocy by defining proper macros */	100	/* work around x32 idiocy by defining proper macros */
83	#if ECB_GCC_AMD64 \|\| ECB_MSVC_AMD64	101	#if ECB_GCC_AMD64 \|\| ECB_MSVC_AMD64
84	#if _ILP32	102	#if _ILP32
85	#define ECB_AMD64_X32 1	103	#define ECB_AMD64_X32 1
86	#else	104	#else
87	#define ECB_AMD64 1	105	#define ECB_AMD64 1
88	#endif	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
89	#endif	113	#endif
90		114
91	/* many compilers define _GNUC_ to some versions but then only implement	115	/* many compilers define _GNUC_ to some versions but then only implement
92	* what their idiot authors think are the "more important" extensions,	116	* what their idiot authors think are the "more important" extensions,
93	* causing enormous grief in return for some better fake benchmark numbers.	117	* causing enormous grief in return for some better fake benchmark numbers.
…		…
115	#define ECB_CLANG_EXTENSION(x) 0	139	#define ECB_CLANG_EXTENSION(x) 0
116	#endif	140	#endif
117		141
118	#define ECB_CPP (__cplusplus+0)	142	#define ECB_CPP (__cplusplus+0)
119	#define ECB_CPP11 (__cplusplus >= 201103L)	143	#define ECB_CPP11 (__cplusplus >= 201103L)
		144	#define ECB_CPP14 (__cplusplus >= 201402L)
		145	#define ECB_CPP17 (__cplusplus >= 201703L)
120		146
121	#if ECB_CPP	147	#if ECB_CPP
122	#define ECB_C 0	148	#define ECB_C 0
123	#define ECB_STDC_VERSION 0	149	#define ECB_STDC_VERSION 0
124	#else	150	#else
…		…
126	#define ECB_STDC_VERSION __STDC_VERSION__	152	#define ECB_STDC_VERSION __STDC_VERSION__
127	#endif	153	#endif
128		154
129	#define ECB_C99 (ECB_STDC_VERSION >= 199901L)	155	#define ECB_C99 (ECB_STDC_VERSION >= 199901L)
130	#define ECB_C11 (ECB_STDC_VERSION >= 201112L)	156	#define ECB_C11 (ECB_STDC_VERSION >= 201112L)
		157	#define ECB_C17 (ECB_STDC_VERSION >= 201710L)
131		158
132	#if ECB_CPP	159	#if ECB_CPP
133	#define ECB_EXTERN_C extern "C"	160	#define ECB_EXTERN_C extern "C"
134	#define ECB_EXTERN_C_BEG ECB_EXTERN_C {	161	#define ECB_EXTERN_C_BEG ECB_EXTERN_C {
135	#define ECB_EXTERN_C_END }	162	#define ECB_EXTERN_C_END }
…		…
155	/* http://www-01.ibm.com/support/knowledgecenter/SSGH3R_13.1.0/com.ibm.xlcpp131.aix.doc/compiler_ref/compiler_builtins.html */	182	/* http://www-01.ibm.com/support/knowledgecenter/SSGH3R_13.1.0/com.ibm.xlcpp131.aix.doc/compiler_ref/compiler_builtins.html */
156	#if __xlC__ && ECB_CPP	183	#if __xlC__ && ECB_CPP
157	#include <builtins.h>	184	#include <builtins.h>
158	#endif	185	#endif
159		186
		187	#if 1400 <= _MSC_VER
		188	#include <intrin.h> /* fence functions _ReadBarrier, also bit search functions _BitScanReverse */
		189	#endif
		190
160	#ifndef ECB_MEMORY_FENCE	191	#ifndef ECB_MEMORY_FENCE
161	#if ECB_GCC_VERSION(2,5) \|\| defined __INTEL_COMPILER \|\| (__llvm__ && __GNUC__) \|\| __SUNPRO_C >= 0x5110 \|\| __SUNPRO_CC >= 0x5110	192	#if ECB_GCC_VERSION(2,5) \|\| defined __INTEL_COMPILER \|\| (__llvm__ && __GNUC__) \|\| __SUNPRO_C >= 0x5110 \|\| __SUNPRO_CC >= 0x5110
		193	#define ECB_MEMORY_FENCE_RELAXED __asm__ __volatile__ ("" : : : "memory")
162	#if __i386 \|\| __i386__	194	#if __i386 \|\| __i386__
163	#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")
164	#define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("" : : : "memory")	196	#define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("" : : : "memory")
165	#define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("")	197	#define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("" : : : "memory")
166	#elif ECB_GCC_AMD64	198	#elif ECB_GCC_AMD64
167	#define ECB_MEMORY_FENCE __asm__ __volatile__ ("mfence" : : : "memory")	199	#define ECB_MEMORY_FENCE __asm__ __volatile__ ("mfence" : : : "memory")
168	#define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("" : : : "memory")	200	#define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("" : : : "memory")
169	#define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("")	201	#define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("" : : : "memory")
170	#elif __powerpc__ \|\| __ppc__ \|\| __powerpc64__ \|\| __ppc64__	202	#elif __powerpc__ \|\| __ppc__ \|\| __powerpc64__ \|\| __ppc64__
171	#define ECB_MEMORY_FENCE __asm__ __volatile__ ("sync" : : : "memory")	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 */
172	#elif defined __ARM_ARCH_6__ \|\| defined __ARM_ARCH_6J__ \	211	#elif defined __ARM_ARCH_6__ \|\| defined __ARM_ARCH_6J__ \
173	\|\| defined __ARM_ARCH_6K__ \|\| defined __ARM_ARCH_6ZK__	212	\|\| defined __ARM_ARCH_6K__ \|\| defined __ARM_ARCH_6ZK__ \
		213	\|\| defined __ARM_ARCH_6T2__
174	#define ECB_MEMORY_FENCE __asm__ __volatile__ ("mcr p15,0,%0,c7,c10,5" : : "r" (0) : "memory")	214	#define ECB_MEMORY_FENCE __asm__ __volatile__ ("mcr p15,0,%0,c7,c10,5" : : "r" (0) : "memory")
175	#elif defined __ARM_ARCH_7__ \|\| defined __ARM_ARCH_7A__ \	215	#elif defined __ARM_ARCH_7__ \|\| defined __ARM_ARCH_7A__ \
176	\|\| defined __ARM_ARCH_7M__ \|\| defined __ARM_ARCH_7R__	216	\|\| defined __ARM_ARCH_7R__ \|\| defined __ARM_ARCH_7M__
177	#define ECB_MEMORY_FENCE __asm__ __volatile__ ("dmb" : : : "memory")	217	#define ECB_MEMORY_FENCE __asm__ __volatile__ ("dmb" : : : "memory")
178	#elif __aarch64__	218	#elif __aarch64__
179	#define ECB_MEMORY_FENCE __asm__ __volatile__ ("dmb ish" : : : "memory")	219	#define ECB_MEMORY_FENCE __asm__ __volatile__ ("dmb ish" : : : "memory")
180	#elif (__sparc \|\| __sparc__) && !(__sparc_v8__ \|\| defined __sparcv8)	220	#elif (__sparc \|\| __sparc__) && !(__sparc_v8__ \|\| defined __sparcv8)
181	#define ECB_MEMORY_FENCE __asm__ __volatile__ ("membar #LoadStore \| #LoadLoad \| #StoreStore \| #StoreLoad" : : : "memory")	221	#define ECB_MEMORY_FENCE __asm__ __volatile__ ("membar #LoadStore \| #LoadLoad \| #StoreStore \| #StoreLoad" : : : "memory")
…		…
208	#if ECB_GCC_VERSION(4,7)	248	#if ECB_GCC_VERSION(4,7)
209	/* see comment below (stdatomic.h) about the C11 memory model. */	249	/* see comment below (stdatomic.h) about the C11 memory model. */
210	#define ECB_MEMORY_FENCE __atomic_thread_fence (__ATOMIC_SEQ_CST)	250	#define ECB_MEMORY_FENCE __atomic_thread_fence (__ATOMIC_SEQ_CST)
211	#define ECB_MEMORY_FENCE_ACQUIRE __atomic_thread_fence (__ATOMIC_ACQUIRE)	251	#define ECB_MEMORY_FENCE_ACQUIRE __atomic_thread_fence (__ATOMIC_ACQUIRE)
212	#define ECB_MEMORY_FENCE_RELEASE __atomic_thread_fence (__ATOMIC_RELEASE)	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)
213		255
214	#elif ECB_CLANG_EXTENSION(c_atomic)	256	#elif ECB_CLANG_EXTENSION(c_atomic)
215	/* see comment below (stdatomic.h) about the C11 memory model. */	257	/* see comment below (stdatomic.h) about the C11 memory model. */
216	#define ECB_MEMORY_FENCE __c11_atomic_thread_fence (__ATOMIC_SEQ_CST)	258	#define ECB_MEMORY_FENCE __c11_atomic_thread_fence (__ATOMIC_SEQ_CST)
217	#define ECB_MEMORY_FENCE_ACQUIRE __c11_atomic_thread_fence (__ATOMIC_ACQUIRE)	259	#define ECB_MEMORY_FENCE_ACQUIRE __c11_atomic_thread_fence (__ATOMIC_ACQUIRE)
218	#define ECB_MEMORY_FENCE_RELEASE __c11_atomic_thread_fence (__ATOMIC_RELEASE)	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)
219		263
220	#elif ECB_GCC_VERSION(4,4) \|\| defined __INTEL_COMPILER \|\| defined __clang__	264	#elif ECB_GCC_VERSION(4,4) \|\| defined __INTEL_COMPILER \|\| defined __clang__
221	#define ECB_MEMORY_FENCE __sync_synchronize ()	265	#define ECB_MEMORY_FENCE __sync_synchronize ()
222	#elif _MSC_VER >= 1500 /* VC++ 2008 */	266	#elif _MSC_VER >= 1500 /* VC++ 2008 */
223	/* apparently, microsoft broke all the memory barrier stuff in Visual Studio 2008... */	267	/* apparently, microsoft broke all the memory barrier stuff in Visual Studio 2008... */
…		…
233	#elif defined _WIN32	277	#elif defined _WIN32
234	#include <WinNT.h>	278	#include <WinNT.h>
235	#define ECB_MEMORY_FENCE MemoryBarrier () /* actually just xchg on x86... scary */	279	#define ECB_MEMORY_FENCE MemoryBarrier () /* actually just xchg on x86... scary */
236	#elif __SUNPRO_C >= 0x5110 \|\| __SUNPRO_CC >= 0x5110	280	#elif __SUNPRO_C >= 0x5110 \|\| __SUNPRO_CC >= 0x5110
237	#include <mbarrier.h>	281	#include <mbarrier.h>
238	#define ECB_MEMORY_FENCE __machine_rw_barrier ()	282	#define ECB_MEMORY_FENCE __machine_rw_barrier ()
239	#define ECB_MEMORY_FENCE_ACQUIRE __machine_r_barrier ()	283	#define ECB_MEMORY_FENCE_ACQUIRE __machine_acq_barrier ()
240	#define ECB_MEMORY_FENCE_RELEASE __machine_w_barrier ()	284	#define ECB_MEMORY_FENCE_RELEASE __machine_rel_barrier ()
		285	#define ECB_MEMORY_FENCE_RELAXED __compiler_barrier ()
241	#elif __xlC__	286	#elif __xlC__
242	#define ECB_MEMORY_FENCE __sync ()	287	#define ECB_MEMORY_FENCE __sync ()
243	#endif	288	#endif
244	#endif	289	#endif
245		290
246	#ifndef ECB_MEMORY_FENCE	291	#ifndef ECB_MEMORY_FENCE
247	#if ECB_C11 && !defined __STDC_NO_ATOMICS__	292	#if ECB_C11 && !defined __STDC_NO_ATOMICS__
248	/* we assume that these memory fences work on all variables/all memory accesses, */	293	/* we assume that these memory fences work on all variables/all memory accesses, */
249	/* not just C11 atomics and atomic accesses */	294	/* not just C11 atomics and atomic accesses */
250	#include <stdatomic.h>	295	#include <stdatomic.h>
251	/* Unfortunately, neither gcc 4.7 nor clang 3.1 generate any instructions for */
252	/* any fence other than seq_cst, which isn't very efficient for us. */
253	/* Why that is, we don't know - either the C11 memory model is quite useless */
254	/* for most usages, or gcc and clang have a bug */
255	/* I currently lean towards the latter, and inefficiently implement */
256	/* all three of ecb's fences as a seq_cst fence */
257	/* Update, gcc-4.8 generates mfence for all c++ fences, but nothing */
258	/* for all __atomic_thread_fence's except seq_cst */
259	#define ECB_MEMORY_FENCE atomic_thread_fence (memory_order_seq_cst)	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)
260	#endif	299	#endif
261	#endif	300	#endif
262		301
263	#ifndef ECB_MEMORY_FENCE	302	#ifndef ECB_MEMORY_FENCE
264	#if !ECB_AVOID_PTHREADS	303	#if !ECB_AVOID_PTHREADS
…		…
284		323
285	#if !defined ECB_MEMORY_FENCE_RELEASE && defined ECB_MEMORY_FENCE	324	#if !defined ECB_MEMORY_FENCE_RELEASE && defined ECB_MEMORY_FENCE
286	#define ECB_MEMORY_FENCE_RELEASE ECB_MEMORY_FENCE	325	#define ECB_MEMORY_FENCE_RELEASE ECB_MEMORY_FENCE
287	#endif	326	#endif
288		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
289	/*****************************************************************************/	332	/*****************************************************************************/
290		333
291	#if ECB_CPP	334	#if ECB_CPP
292	#define ecb_inline static inline	335	#define ecb_inline static inline
293	#elif ECB_GCC_VERSION(2,5)	336	#elif ECB_GCC_VERSION(2,5)
…		…
423	#else	466	#else
424	ecb_function_ ecb_const int ecb_ctz32 (uint32_t x);	467	ecb_function_ ecb_const int ecb_ctz32 (uint32_t x);
425	ecb_function_ ecb_const int	468	ecb_function_ ecb_const int
426	ecb_ctz32 (uint32_t x)	469	ecb_ctz32 (uint32_t x)
427	{	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
428	int r = 0;	476	int r = 0;
429		477
430	x &= ~x + 1; /* this isolates the lowest bit */	478	x &= ~x + 1; /* this isolates the lowest bit */
431		479
432	#if ECB_branchless_on_i386	480	#if ECB_branchless_on_i386
…		…
442	if (x & 0xff00ff00) r += 8;	490	if (x & 0xff00ff00) r += 8;
443	if (x & 0xffff0000) r += 16;	491	if (x & 0xffff0000) r += 16;
444	#endif	492	#endif
445		493
446	return r;	494	return r;
		495	#endif
447	}	496	}
448		497
449	ecb_function_ ecb_const int ecb_ctz64 (uint64_t x);	498	ecb_function_ ecb_const int ecb_ctz64 (uint64_t x);
450	ecb_function_ ecb_const int	499	ecb_function_ ecb_const int
451	ecb_ctz64 (uint64_t x)	500	ecb_ctz64 (uint64_t x)
452	{	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
453	int shift = x & 0xffffffff ? 0 : 32;	507	int shift = x & 0xffffffff ? 0 : 32;
454	return ecb_ctz32 (x >> shift) + shift;	508	return ecb_ctz32 (x >> shift) + shift;
		509	#endif
455	}	510	}
456		511
457	ecb_function_ ecb_const int ecb_popcount32 (uint32_t x);	512	ecb_function_ ecb_const int ecb_popcount32 (uint32_t x);
458	ecb_function_ ecb_const int	513	ecb_function_ ecb_const int
459	ecb_popcount32 (uint32_t x)	514	ecb_popcount32 (uint32_t x)
…		…
467	}	522	}
468		523
469	ecb_function_ ecb_const int ecb_ld32 (uint32_t x);	524	ecb_function_ ecb_const int ecb_ld32 (uint32_t x);
470	ecb_function_ ecb_const int ecb_ld32 (uint32_t x)	525	ecb_function_ ecb_const int ecb_ld32 (uint32_t x)
471	{	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
472	int r = 0;	532	int r = 0;
473		533
474	if (x >> 16) { x >>= 16; r += 16; }	534	if (x >> 16) { x >>= 16; r += 16; }
475	if (x >> 8) { x >>= 8; r += 8; }	535	if (x >> 8) { x >>= 8; r += 8; }
476	if (x >> 4) { x >>= 4; r += 4; }	536	if (x >> 4) { x >>= 4; r += 4; }
477	if (x >> 2) { x >>= 2; r += 2; }	537	if (x >> 2) { x >>= 2; r += 2; }
478	if (x >> 1) { r += 1; }	538	if (x >> 1) { r += 1; }
479		539
480	return r;	540	return r;
		541	#endif
481	}	542	}
482		543
483	ecb_function_ ecb_const int ecb_ld64 (uint64_t x);	544	ecb_function_ ecb_const int ecb_ld64 (uint64_t x);
484	ecb_function_ ecb_const int ecb_ld64 (uint64_t x)	545	ecb_function_ ecb_const int ecb_ld64 (uint64_t x)
485	{	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
486	int r = 0;	552	int r = 0;
487		553
488	if (x >> 32) { x >>= 32; r += 32; }	554	if (x >> 32) { x >>= 32; r += 32; }
489		555
490	return r + ecb_ld32 (x);	556	return r + ecb_ld32 (x);
		557	#endif
491	}	558	}
492	#endif	559	#endif
493		560
494	ecb_function_ ecb_const ecb_bool ecb_is_pot32 (uint32_t x);	561	ecb_function_ ecb_const ecb_bool ecb_is_pot32 (uint32_t x);
495	ecb_function_ ecb_const ecb_bool ecb_is_pot32 (uint32_t x) { return !(x & (x - 1)); }	562	ecb_function_ ecb_const ecb_bool ecb_is_pot32 (uint32_t x) { return !(x & (x - 1)); }
…		…
542	ecb_inline ecb_const uint32_t ecb_rotl32 (uint32_t x, unsigned int count);	609	ecb_inline ecb_const uint32_t ecb_rotl32 (uint32_t x, unsigned int count);
543	ecb_inline ecb_const uint32_t ecb_rotr32 (uint32_t x, unsigned int count);	610	ecb_inline ecb_const uint32_t ecb_rotr32 (uint32_t x, unsigned int count);
544	ecb_inline ecb_const uint64_t ecb_rotl64 (uint64_t x, unsigned int count);	611	ecb_inline ecb_const uint64_t ecb_rotl64 (uint64_t x, unsigned int count);
545	ecb_inline ecb_const uint64_t ecb_rotr64 (uint64_t x, unsigned int count);	612	ecb_inline ecb_const uint64_t ecb_rotr64 (uint64_t x, unsigned int count);
546		613
547	ecb_inline ecb_const 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)); }
548	ecb_inline ecb_const 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)); }
549	ecb_inline ecb_const 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)); }
550	ecb_inline ecb_const 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)); }
551	ecb_inline ecb_const 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)); }
552	ecb_inline ecb_const 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)); }
553	ecb_inline ecb_const 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)); }
554	ecb_inline ecb_const 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)); }
		622
		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
555		660
556	#if ECB_GCC_VERSION(4,3) \|\| (ECB_CLANG_BUILTIN(__builtin_bswap32) && ECB_CLANG_BUILTIN(__builtin_bswap64))	661	#if ECB_GCC_VERSION(4,3) \|\| (ECB_CLANG_BUILTIN(__builtin_bswap32) && ECB_CLANG_BUILTIN(__builtin_bswap64))
557	#if ECB_GCC_VERSION(4,8) \|\| ECB_CLANG_BUILTIN(__builtin_bswap16)	662	#if ECB_GCC_VERSION(4,8) \|\| ECB_CLANG_BUILTIN(__builtin_bswap16)
558	#define ecb_bswap16(x) __builtin_bswap16 (x)	663	#define ecb_bswap16(x) __builtin_bswap16 (x)
559	#else	664	#else
…		…
598	#endif	703	#endif
599		704
600	/* try to tell the compiler that some condition is definitely true */	705	/* try to tell the compiler that some condition is definitely true */
601	#define ecb_assume(cond) if (!(cond)) ecb_unreachable (); else 0	706	#define ecb_assume(cond) if (!(cond)) ecb_unreachable (); else 0
602		707
603	ecb_inline ecb_const unsigned char ecb_byteorder_helper (void);	708	ecb_inline ecb_const uint32_t ecb_byteorder_helper (void);
604	ecb_inline ecb_const unsigned char	709	ecb_inline ecb_const uint32_t
605	ecb_byteorder_helper (void)	710	ecb_byteorder_helper (void)
606	{	711	{
607	/* the union code still generates code under pressure in gcc, */	712	/* the union code still generates code under pressure in gcc, */
608	/* but less than using pointers, and always seems to */	713	/* but less than using pointers, and always seems to */
609	/* successfully return a constant. */	714	/* successfully return a constant. */
610	/* the reason why we have this horrible preprocessor mess */	715	/* the reason why we have this horrible preprocessor mess */
611	/* is to avoid it in all cases, at least on common architectures */	716	/* is to avoid it in all cases, at least on common architectures */
612	/* or when using a recent enough gcc version (>= 4.6) */	717	/* or when using a recent enough gcc version (>= 4.6) */
613	#if ((__i386 \|\| __i386__) && !__VOS__) \|\| _M_IX86 \|\| ECB_GCC_AMD64 \|\| ECB_MSVC_AMD64
614	return 0x44;
615	#elif __BYTE_ORDER__ && __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__	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
616	return 0x44;	721	return 0x44332211;
617	#elif __BYTE_ORDER__ && __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__	722	#elif (defined __BYTE_ORDER__ && __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__) \
		723	\|\| ((__AARCH64EB__ \|\| __MIPSEB__ \|\| __ARMEB__) && !__VOS__)
		724	#define ECB_BIG_ENDIAN 1
618	return 0x11;	725	return 0x11223344;
619	#else	726	#else
620	union	727	union
621	{	728	{
		729	uint8_t c[4];
622	uint32_t i;	730	uint32_t u;
623	uint8_t c;
624	} u = { 0x11223344 };	731	} u = { 0x11, 0x22, 0x33, 0x44 };
625	return u.c;	732	return u.u;
626	#endif	733	#endif
627	}	734	}
628		735
629	ecb_inline ecb_const ecb_bool ecb_big_endian (void);	736	ecb_inline ecb_const ecb_bool ecb_big_endian (void);
630	ecb_inline ecb_const 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; }
631	ecb_inline ecb_const ecb_bool ecb_little_endian (void);	738	ecb_inline ecb_const ecb_bool ecb_little_endian (void);
632	ecb_inline ecb_const 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	/* 2d hilbert curves */
		954
		955	/* algorithm from the book Hacker's Delight, modified to not */
		956	/* run into undefined behaviour for n==16 */
		957	static uint32_t
		958	ecb_hilbert2d_index_to_coord32 (int n, uint32_t s)
		959	{
		960	uint32_t comp, swap, cs, t, sr;
		961
		962	/* pad s on the left (unused) bits with 01 (no change groups) */
		963	s \|= 0x55555555U << n << n;
		964	/* "s shift right" */
		965	sr = (s >> 1) & 0x55555555U;
		966	/* compute complement and swap info in two-bit groups */
		967	cs = ((s & 0x55555555U) + sr) ^ 0x55555555U;
		968
		969	/* parallel prefix xor op to propagate both complement
		970	* and swap info together from left to right (there is
		971	* no step "cs ^= cs >> 1", so in effect it computes
		972	* two independent parallel prefix operations on two
		973	* interleaved sets of sixteen bits).
		974	*/
		975	cs ^= cs >> 2;
		976	cs ^= cs >> 4;
		977	cs ^= cs >> 8;
		978	cs ^= cs >> 16;
		979
		980	/* separate swap and complement bits */
		981	swap = cs & 0x55555555U;
		982	comp = (cs >> 1) & 0x55555555U;
		983
		984	/* calculate coordinates in odd and even bit positions */
		985	t = (s & swap) ^ comp;
		986	s = s ^ sr ^ t ^ (t << 1);
		987
		988	/* unpad/clear out any junk on the left */
		989	s = s & ((1 << n << n) - 1);
		990
		991	/* Now "unshuffle" to separate the x and y bits. */
		992	t = (s ^ (s >> 1)) & 0x22222222U; s ^= t ^ (t << 1);
		993	t = (s ^ (s >> 2)) & 0x0c0c0c0cU; s ^= t ^ (t << 2);
		994	t = (s ^ (s >> 4)) & 0x00f000f0U; s ^= t ^ (t << 4);
		995	t = (s ^ (s >> 8)) & 0x0000ff00U; s ^= t ^ (t << 8);
		996
		997	/* now s contains two 16-bit coordinates */
		998	return s;
		999	}
		1000
		1001	/* 64 bit, a straightforward extension to the 32 bit case */
		1002	static uint64_t
		1003	ecb_hilbert2d_index_to_coord64 (int n, uint64_t s)
		1004	{
		1005	uint64_t comp, swap, cs, t, sr;
		1006
		1007	/* pad s on the left (unused) bits with 01 (no change groups) */
		1008	s \|= 0x5555555555555555U << n << n;
		1009	/* "s shift right" */
		1010	sr = (s >> 1) & 0x5555555555555555U;
		1011	/* compute complement and swap info in two-bit groups */
		1012	cs = ((s & 0x5555555555555555U) + sr) ^ 0x5555555555555555U;
		1013
		1014	/* parallel prefix xor op to propagate both complement
		1015	* and swap info together from left to right (there is
		1016	* no step "cs ^= cs >> 1", so in effect it computes
		1017	* two independent parallel prefix operations on two
		1018	* interleaved sets of thirty-two bits).
		1019	*/
		1020	cs ^= cs >> 2;
		1021	cs ^= cs >> 4;
		1022	cs ^= cs >> 8;
		1023	cs ^= cs >> 16;
		1024	cs ^= cs >> 32;
		1025
		1026	/* separate swap and complement bits */
		1027	swap = cs & 0x5555555555555555U;
		1028	comp = (cs >> 1) & 0x5555555555555555U;
		1029
		1030	/* calculate coordinates in odd and even bit positions */
		1031	t = (s & swap) ^ comp;
		1032	s = s ^ sr ^ t ^ (t << 1);
		1033
		1034	/* unpad/clear out any junk on the left */
		1035	s = s & ((1 << n << n) - 1);
		1036
		1037	/* Now "unshuffle" to separate the x and y bits. */
		1038	t = (s ^ (s >> 1)) & 0x2222222222222222U; s ^= t ^ (t << 1);
		1039	t = (s ^ (s >> 2)) & 0x0c0c0c0c0c0c0c0cU; s ^= t ^ (t << 2);
		1040	t = (s ^ (s >> 4)) & 0x00f000f000f000f0U; s ^= t ^ (t << 4);
		1041	t = (s ^ (s >> 8)) & 0x0000ff000000ff00U; s ^= t ^ (t << 8);
		1042	t = (s ^ (s >> 16)) & 0x00000000ffff0000U; s ^= t ^ (t << 16);
		1043
		1044	/* now s contains two 32-bit coordinates */
		1045	return s;
		1046	}
		1047
		1048	/* algorithm from the book Hacker's Delight, but a similar algorithm*/
		1049	/* is given in https://doi.org/10.1002/spe.4380160103 */
		1050	/* this has been slightly improved over the original version */
		1051	ecb_function_ uint32_t
		1052	ecb_hilbert2d_coord_to_index32 (int n, uint32_t xy)
		1053	{
		1054	uint32_t row;
		1055	uint32_t state = 0;
		1056	uint32_t s = 0;
		1057
		1058	do
		1059	{
		1060	--n;
		1061
		1062	row = 4 * state
		1063	\| (2 & (xy >> n >> 15))
		1064	\| (1 & (xy >> n ));
		1065
		1066	/* these funky constants are lookup tables for two-bit values */
		1067	s = (s << 2) \| (0x361e9cb4U >> 2 * row) & 3;
		1068	state = (0x8fe65831U >> 2 * row) & 3;
		1069	}
		1070	while (n > 0);
		1071
		1072	return s;
		1073	}
		1074
		1075	/* 64 bit, essentially the same as 32 bit */
		1076	ecb_function_ uint64_t
		1077	ecb_hilbert2d_coord_to_index64 (int n, uint64_t xy)
		1078	{
		1079	uint32_t row;
		1080	uint32_t state = 0;
		1081	uint64_t s = 0;
		1082
		1083	do
		1084	{
		1085	--n;
		1086
		1087	row = 4 * state
		1088	\| (2 & (xy >> n >> 31))
		1089	\| (1 & (xy >> n ));
		1090
		1091	/* these funky constants are lookup tables for two-bit values */
		1092	s = (s << 2) \| (0x361e9cb4U >> 2 * row) & 3;
		1093	state = (0x8fe65831U >> 2 * row) & 3;
		1094	}
		1095	while (n > 0);
		1096
		1097	return s;
		1098	}
		1099
		1100	/*****************************************************************************/
		1101	/* division */
633		1102
634	#if ECB_GCC_VERSION(3,0) \|\| ECB_C99	1103	#if ECB_GCC_VERSION(3,0) \|\| ECB_C99
		1104	/* C99 tightened the definition of %, so we can use a more efficient version */
635	#define ecb_mod(m,n) ((m) % (n) + ((m) % (n) < 0 ? (n) : 0))	1105	#define ecb_mod(m,n) ((m) % (n) + ((m) % (n) < 0 ? (n) : 0))
636	#else	1106	#else
637	#define ecb_mod(m,n) ((m) < 0 ? ((n) - 1 - ((-1 - (m)) % (n))) : ((m) % (n)))	1107	#define ecb_mod(m,n) ((m) < 0 ? ((n) - 1 - ((-1 - (m)) % (n))) : ((m) % (n)))
638	#endif	1108	#endif
639		1109
…		…
651	#else	1121	#else
652	#define ecb_div_rd(val,div) ((val) < 0 ? - ((-(val) + (div) - 1) / (div)) : ((val) ) / (div))	1122	#define ecb_div_rd(val,div) ((val) < 0 ? - ((-(val) + (div) - 1) / (div)) : ((val) ) / (div))
653	#define ecb_div_ru(val,div) ((val) < 0 ? - ((-(val) ) / (div)) : ((val) + (div) - 1) / (div))	1123	#define ecb_div_ru(val,div) ((val) < 0 ? - ((-(val) ) / (div)) : ((val) + (div) - 1) / (div))
654	#endif	1124	#endif
655		1125
		1126	/*****************************************************************************/
		1127	/* array length */
		1128
656	#if ecb_cplusplus_does_not_suck	1129	#if ecb_cplusplus_does_not_suck
657	/* does not work for local types (http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2657.htm) */	1130	/* does not work for local types (http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2657.htm) */
658	template<typename T, int N>	1131	template<typename T, int N>
659	static inline int ecb_array_length (const T (&arr)[N])	1132	static inline int ecb_array_length (const T (&arr)[N])
660	{	1133	{
…		…
662	}	1135	}
663	#else	1136	#else
664	#define ecb_array_length(name) (sizeof (name) / sizeof (name [0]))	1137	#define ecb_array_length(name) (sizeof (name) / sizeof (name [0]))
665	#endif	1138	#endif
666		1139
		1140	/*****************************************************************************/
		1141	/* IEEE 754-2008 half float conversions */
		1142
667	ecb_function_ ecb_const uint32_t ecb_binary16_to_binary32 (uint16_t x);	1143	ecb_function_ ecb_const uint32_t ecb_binary16_to_binary32 (uint32_t x);
668	ecb_function_ ecb_const uint32_t	1144	ecb_function_ ecb_const uint32_t
669	ecb_binary16_to_binary32 (uint16_t x)	1145	ecb_binary16_to_binary32 (uint32_t x)
670	{	1146	{
671	unsigned int s = (x & 0x8000) << (31 - 15);	1147	unsigned int s = (x & 0x8000) << (31 - 15);
672	int e = (x >> 10) & 0x001f;	1148	int e = (x >> 10) & 0x001f;
673	unsigned int m = x & 0x03ff;	1149	unsigned int m = x & 0x03ff;
674		1150
…		…
699	ecb_function_ ecb_const uint16_t ecb_binary32_to_binary16 (uint32_t x);	1175	ecb_function_ ecb_const uint16_t ecb_binary32_to_binary16 (uint32_t x);
700	ecb_function_ ecb_const uint16_t	1176	ecb_function_ ecb_const uint16_t
701	ecb_binary32_to_binary16 (uint32_t x)	1177	ecb_binary32_to_binary16 (uint32_t x)
702	{	1178	{
703	unsigned int s = (x >> 16) & 0x00008000; /* sign bit, the easy part */	1179	unsigned int s = (x >> 16) & 0x00008000; /* sign bit, the easy part */
704	unsigned int e = ((x >> 23) & 0x000000ff) - (127 - 15); /* the desired exponent */	1180	int e = ((x >> 23) & 0x000000ff) - (127 - 15); /* the desired exponent */
705	unsigned int m = x & 0x007fffff;	1181	unsigned int m = x & 0x007fffff;
706		1182
707	x &= 0x7fffffff;	1183	x &= 0x7fffffff;
708		1184
709	/* if it's within range of binary16 normals, use fast path */	1185	/* if it's within range of binary16 normals, use fast path */
…		…
724		1200
725	/* handle large numbers and infinity */	1201	/* handle large numbers and infinity */
726	if (ecb_expect_true (0x477fefff < x && x <= 0x7f800000))	1202	if (ecb_expect_true (0x477fefff < x && x <= 0x7f800000))
727	return s \| 0x7c00;	1203	return s \| 0x7c00;
728		1204
729	/* handle zero and subnormals */	1205	/* handle zero, subnormals and small numbers */
730	if (ecb_expect_true (x < 0x38800000))	1206	if (ecb_expect_true (x < 0x38800000))
731	{	1207	{
732	/* zero */	1208	/* zero */
733	if (ecb_expect_true (!x))	1209	if (ecb_expect_true (!x))
734	return s;	1210	return s;
735		1211
736	/* handle subnormals */	1212	/* handle subnormals */
737		1213
		1214	/* too small, will be zero */
		1215	if (e < (14 - 24)) /* might not be sharp, but is good enough */
		1216	return s;
		1217
738	m \|= 0x00800000; /* make implicit bit explicit */	1218	m \|= 0x00800000; /* make implicit bit explicit */
739		1219
740	/* very tricky - we need to round to the nearest e (+10) bit value */	1220	/* very tricky - we need to round to the nearest e (+10) bit value */
741	{	1221	{
742	unsigned int bits = 14 - e;	1222	unsigned int bits = 14 - e;
…		…
752		1232
753	/* handle NaNs, preserve leftmost nan bits, but make sure we don't turn them into infinities */	1233	/* handle NaNs, preserve leftmost nan bits, but make sure we don't turn them into infinities */
754	m >>= 13;	1234	m >>= 13;
755		1235
756	return s \| 0x7c00 \| m \| !m;	1236	return s \| 0x7c00 \| m \| !m;
		1237	}
		1238
		1239	/*******************************************************************************/
		1240	/* fast integer to ascii */
		1241
		1242	/*
		1243	* This code is pretty complicated because it is general. The idea behind it,
		1244	* however, is pretty simple: first, the number is multiplied with a scaling
		1245	* factor (2bits / 10(digits-1)) to convert the integer into a fixed-point
		1246	* number with the first digit in the upper bits.
		1247	* Then this digit is converted to text and masked out. The resulting number
		1248	* is then multiplied by 10, by multiplying the fixed point representation
		1249	* by 5 and shifting the (binary) decimal point one to the right, so a 4.28
		1250	* format becomes 5.27, 6.26 and so on.
		1251	* The rest involves only advancing the pointer if we already generated a
		1252	* non-zero digit, so leading zeroes are overwritten.
		1253	*/
		1254
		1255	/* simply return a mask with "bits" bits set */
		1256	#define ecb_i2a_mask(type,bits) ((((type)1) << (bits)) - 1)
		1257
		1258	/* oputput a single digit. maskvalue is 10*digitidx /
		1259	#define ecb_i2a_digit(type,bits,digitmask,maskvalue,digitidx) \
		1260	if (digitmask >= maskvalue) /* constant, used to decide how many digits to generate */ \
		1261	{ \
		1262	char digit = x >> (bits - digitidx); /* calculate the topmost digit */ \
		1263	ptr = digit + '0'; / output it */ \
		1264	nz = (digitmask == maskvalue) \|\| nz \|\| digit; /* first term == always output last digit */ \
		1265	ptr += nz; /* output digit only if non-zero digit seen */ \
		1266	x = (x & ecb_i2a_mask (type, bits - digitidx)) * 5; /* 10, but shift decimal point right / \
		1267	}
		1268
		1269	/* convert integer to fixed point format and multiply out digits, highest first */
		1270	/* requires magic constants: max. digits and number of bits after the decimal point */
		1271	#define ecb_i2a_def(suffix,ptr,v,type,bits,digitmask,lz) \
		1272	ecb_inline char ecb_i2a_ ## suffix (char ptr, uint32_t u) \
		1273	{ \
		1274	char nz = lz; /* non-zero digit seen? */ \
		1275	/* convert to x.bits fixed-point */ \
		1276	type x = u * ((ecb_i2a_mask (type, bits) + digitmask) / digitmask); \
		1277	/* output up to 10 digits */ \
		1278	ecb_i2a_digit (type,bits,digitmask, 1, 0); \
		1279	ecb_i2a_digit (type,bits,digitmask, 10, 1); \
		1280	ecb_i2a_digit (type,bits,digitmask, 100, 2); \
		1281	ecb_i2a_digit (type,bits,digitmask, 1000, 3); \
		1282	ecb_i2a_digit (type,bits,digitmask, 10000, 4); \
		1283	ecb_i2a_digit (type,bits,digitmask, 100000, 5); \
		1284	ecb_i2a_digit (type,bits,digitmask, 1000000, 6); \
		1285	ecb_i2a_digit (type,bits,digitmask, 10000000, 7); \
		1286	ecb_i2a_digit (type,bits,digitmask, 100000000, 8); \
		1287	ecb_i2a_digit (type,bits,digitmask, 1000000000, 9); \
		1288	return ptr; \
		1289	}
		1290
		1291	/* predefined versions of the above, for various digits */
		1292	/* ecb_i2a_xN = almost N digits, limit defined by macro */
		1293	/* ecb_i2a_N = up to N digits, leading zeroes suppressed */
		1294	/* ecb_i2a_0N = exactly N digits, including leading zeroes */
		1295
		1296	/* non-leading-zero versions, limited range */
		1297	#define ECB_I2A_MAX_X5 59074 /* limit for ecb_i2a_x5 */
		1298	#define ECB_I2A_MAX_X10 2932500665 /* limit for ecb_i2a_x10 */
		1299	ecb_i2a_def ( x5, ptr, v, uint32_t, 26, 10000, 0)
		1300	ecb_i2a_def (x10, ptr, v, uint64_t, 60, 1000000000, 0)
		1301
		1302	/* non-leading zero versions, all digits, 4 and 9 are optimal for 32/64 bit */
		1303	ecb_i2a_def ( 2, ptr, v, uint32_t, 10, 10, 0)
		1304	ecb_i2a_def ( 3, ptr, v, uint32_t, 12, 100, 0)
		1305	ecb_i2a_def ( 4, ptr, v, uint32_t, 26, 1000, 0)
		1306	ecb_i2a_def ( 5, ptr, v, uint64_t, 30, 10000, 0)
		1307	ecb_i2a_def ( 6, ptr, v, uint64_t, 36, 100000, 0)
		1308	ecb_i2a_def ( 7, ptr, v, uint64_t, 44, 1000000, 0)
		1309	ecb_i2a_def ( 8, ptr, v, uint64_t, 50, 10000000, 0)
		1310	ecb_i2a_def ( 9, ptr, v, uint64_t, 56, 100000000, 0)
		1311
		1312	/* leading-zero versions, all digits, 04 and 09 are optimal for 32/64 bit */
		1313	ecb_i2a_def (02, ptr, v, uint32_t, 10, 10, 1)
		1314	ecb_i2a_def (03, ptr, v, uint32_t, 12, 100, 1)
		1315	ecb_i2a_def (04, ptr, v, uint32_t, 26, 1000, 1)
		1316	ecb_i2a_def (05, ptr, v, uint64_t, 30, 10000, 1)
		1317	ecb_i2a_def (06, ptr, v, uint64_t, 36, 100000, 1)
		1318	ecb_i2a_def (07, ptr, v, uint64_t, 44, 1000000, 1)
		1319	ecb_i2a_def (08, ptr, v, uint64_t, 50, 10000000, 1)
		1320	ecb_i2a_def (09, ptr, v, uint64_t, 56, 100000000, 1)
		1321
		1322	#define ECB_I2A_I32_DIGITS 11
		1323	#define ECB_I2A_U32_DIGITS 10
		1324	#define ECB_I2A_I64_DIGITS 20
		1325	#define ECB_I2A_U64_DIGITS 21
		1326	#define ECB_I2A_MAX_DIGITS 21
		1327
		1328	ecb_inline char *
		1329	ecb_i2a_u32 (char *ptr, uint32_t u)
		1330	{
		1331	#if ECB_64BIT_NATIVE
		1332	if (ecb_expect_true (u <= ECB_I2A_MAX_X10))
		1333	ptr = ecb_i2a_x10 (ptr, u);
		1334	else /* x10 almost, but not fully, covers 32 bit */
		1335	{
		1336	uint32_t u1 = u % 1000000000;
		1337	uint32_t u2 = u / 1000000000;
		1338
		1339	*ptr++ = u2 + '0';
		1340	ptr = ecb_i2a_09 (ptr, u1);
		1341	}
		1342	#else
		1343	if (ecb_expect_true (u <= ECB_I2A_MAX_X5))
		1344	ecb_i2a_x5 (ptr, u);
		1345	else if (ecb_expect_true (u <= ECB_I2A_MAX_X5 * 10000))
		1346	{
		1347	uint32_t u1 = u % 10000;
		1348	uint32_t u2 = u / 10000;
		1349
		1350	ptr = ecb_i2a_x5 (ptr, u2);
		1351	ptr = ecb_i2a_04 (ptr, u1);
		1352	}
		1353	else
		1354	{
		1355	uint32_t u1 = u % 10000;
		1356	uint32_t ua = u / 10000;
		1357	uint32_t u2 = ua % 10000;
		1358	uint32_t u3 = ua / 10000;
		1359
		1360	ptr = ecb_i2a_2 (ptr, u3);
		1361	ptr = ecb_i2a_04 (ptr, u2);
		1362	ptr = ecb_i2a_04 (ptr, u1);
		1363	}
		1364	#endif
		1365
		1366	return ptr;
		1367	}
		1368
		1369	ecb_inline char *
		1370	ecb_i2a_i32 (char *ptr, int32_t v)
		1371	{
		1372	*ptr = '-'; ptr += v < 0;
		1373	uint32_t u = v < 0 ? -(uint32_t)v : v;
		1374
		1375	#if ECB_64BIT_NATIVE
		1376	ptr = ecb_i2a_x10 (ptr, u); /* x10 fully covers 31 bit */
		1377	#else
		1378	ptr = ecb_i2a_u32 (ptr, u);
		1379	#endif
		1380
		1381	return ptr;
		1382	}
		1383
		1384	ecb_inline char *
		1385	ecb_i2a_u64 (char *ptr, uint64_t u)
		1386	{
		1387	#if ECB_64BIT_NATIVE
		1388	if (ecb_expect_true (u <= ECB_I2A_MAX_X10))
		1389	ptr = ecb_i2a_x10 (ptr, u);
		1390	else if (ecb_expect_false (u <= ECB_I2A_MAX_X10 * 1000000000))
		1391	{
		1392	uint64_t u1 = u % 1000000000;
		1393	uint64_t u2 = u / 1000000000;
		1394
		1395	ptr = ecb_i2a_x10 (ptr, u2);
		1396	ptr = ecb_i2a_09 (ptr, u1);
		1397	}
		1398	else
		1399	{
		1400	uint64_t u1 = u % 1000000000;
		1401	uint64_t ua = u / 1000000000;
		1402	uint64_t u2 = ua % 1000000000;
		1403	uint64_t u3 = ua / 1000000000;
		1404
		1405	ptr = ecb_i2a_2 (ptr, u3);
		1406	ptr = ecb_i2a_09 (ptr, u2);
		1407	ptr = ecb_i2a_09 (ptr, u1);
		1408	}
		1409	#else
		1410	if (ecb_expect_true (u <= ECB_I2A_MAX_X5))
		1411	ptr = ecb_i2a_x5 (ptr, u);
		1412	else
		1413	{
		1414	uint64_t u1 = u % 10000;
		1415	uint64_t u2 = u / 10000;
		1416
		1417	ptr = ecb_i2a_u64 (ptr, u2);
		1418	ptr = ecb_i2a_04 (ptr, u1);
		1419	}
		1420	#endif
		1421
		1422	return ptr;
		1423	}
		1424
		1425	ecb_inline char *
		1426	ecb_i2a_i64 (char *ptr, int64_t v)
		1427	{
		1428	*ptr = '-'; ptr += v < 0;
		1429	uint64_t u = v < 0 ? -(uint64_t)v : v;
		1430
		1431	#if ECB_64BIT_NATIVE
		1432	if (ecb_expect_true (u <= ECB_I2A_MAX_X10))
		1433	ptr = ecb_i2a_x10 (ptr, u);
		1434	else if (ecb_expect_false (u <= ECB_I2A_MAX_X10 * 1000000000))
		1435	{
		1436	uint64_t u1 = u % 1000000000;
		1437	uint64_t u2 = u / 1000000000;
		1438
		1439	ptr = ecb_i2a_x10 (ptr, u2);
		1440	ptr = ecb_i2a_09 (ptr, u1);
		1441	}
		1442	else
		1443	{
		1444	uint64_t u1 = u % 1000000000;
		1445	uint64_t ua = u / 1000000000;
		1446	uint64_t u2 = ua % 1000000000;
		1447	uint64_t u3 = ua / 1000000000;
		1448
		1449	/* 2*31 is 19 digits, so the top is exactly one digit /
		1450	*ptr++ = u3 + '0';
		1451	ptr = ecb_i2a_09 (ptr, u2);
		1452	ptr = ecb_i2a_09 (ptr, u1);
		1453	}
		1454	#else
		1455	ptr = ecb_i2a_u64 (ptr, u);
		1456	#endif
		1457
		1458	return ptr;
757	}	1459	}
758		1460
759	/*******************************************************************************/	1461	/*******************************************************************************/
760	/* floating point stuff, can be disabled by defining ECB_NO_LIBM */	1462	/* floating point stuff, can be disabled by defining ECB_NO_LIBM */
761		1463
…		…
775	\|\| defined __sh__ \	1477	\|\| defined __sh__ \
776	\|\| defined _M_IX86 \|\| defined ECB_MSVC_AMD64 \|\| defined _M_IA64 \	1478	\|\| defined _M_IX86 \|\| defined ECB_MSVC_AMD64 \|\| defined _M_IA64 \
777	\|\| (defined __arm__ && (defined __ARM_EABI__ \|\| defined __EABI__ \|\| defined __VFP_FP__ \|\| defined _WIN32_WCE \|\| defined __ANDROID__)) \	1479	\|\| (defined __arm__ && (defined __ARM_EABI__ \|\| defined __EABI__ \|\| defined __VFP_FP__ \|\| defined _WIN32_WCE \|\| defined __ANDROID__)) \
778	\|\| defined __aarch64__	1480	\|\| defined __aarch64__
779	#define ECB_STDFP 1	1481	#define ECB_STDFP 1
780	#include <string.h> /* for memcpy */
781	#else	1482	#else
782	#define ECB_STDFP 0	1483	#define ECB_STDFP 0
783	#endif	1484	#endif
784		1485
785	#ifndef ECB_NO_LIBM	1486	#ifndef ECB_NO_LIBM

Diff Legend

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

Comparing libecb/ecb.h (file contents): Revision 1.168 by root, Sat Nov 21 18:09:36 2015 UTC vs. Revision 1.204 by root, Fri Mar 25 08:44:14 2022 UTC

Diff Legend

Comparing libecb/ecb.h (file contents):
Revision 1.168 by root, Sat Nov 21 18:09:36 2015 UTC vs.
Revision 1.204 by root, Fri Mar 25 08:44:14 2022 UTC