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 0x00010006 |
45 | #define ECB_VERSION 0x00010009 |
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 |
… | |
… | |
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) |
81 | |
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 | |
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. |
… | |
… | |
224 | #if ECB_GCC_VERSION(4,7) |
248 | #if ECB_GCC_VERSION(4,7) |
225 | /* see comment below (stdatomic.h) about the C11 memory model. */ |
249 | /* see comment below (stdatomic.h) about the C11 memory model. */ |
226 | #define ECB_MEMORY_FENCE __atomic_thread_fence (__ATOMIC_SEQ_CST) |
250 | #define ECB_MEMORY_FENCE __atomic_thread_fence (__ATOMIC_SEQ_CST) |
227 | #define ECB_MEMORY_FENCE_ACQUIRE __atomic_thread_fence (__ATOMIC_ACQUIRE) |
251 | #define ECB_MEMORY_FENCE_ACQUIRE __atomic_thread_fence (__ATOMIC_ACQUIRE) |
228 | #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 |
229 | #define ECB_MEMORY_FENCE_RELAXED __atomic_thread_fence (__ATOMIC_RELAXED) |
254 | #define ECB_MEMORY_FENCE_RELAXED __atomic_thread_fence (__ATOMIC_RELAXED) |
230 | |
255 | |
231 | #elif ECB_CLANG_EXTENSION(c_atomic) |
256 | #elif ECB_CLANG_EXTENSION(c_atomic) |
232 | /* see comment below (stdatomic.h) about the C11 memory model. */ |
257 | /* see comment below (stdatomic.h) about the C11 memory model. */ |
233 | #define ECB_MEMORY_FENCE __c11_atomic_thread_fence (__ATOMIC_SEQ_CST) |
258 | #define ECB_MEMORY_FENCE __c11_atomic_thread_fence (__ATOMIC_SEQ_CST) |
234 | #define ECB_MEMORY_FENCE_ACQUIRE __c11_atomic_thread_fence (__ATOMIC_ACQUIRE) |
259 | #define ECB_MEMORY_FENCE_ACQUIRE __c11_atomic_thread_fence (__ATOMIC_ACQUIRE) |
235 | #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 |
236 | #define ECB_MEMORY_FENCE_RELAXED __c11_atomic_thread_fence (__ATOMIC_RELAXED) |
262 | #define ECB_MEMORY_FENCE_RELAXED __c11_atomic_thread_fence (__ATOMIC_RELAXED) |
237 | |
263 | |
238 | #elif ECB_GCC_VERSION(4,4) || defined __INTEL_COMPILER || defined __clang__ |
264 | #elif ECB_GCC_VERSION(4,4) || defined __INTEL_COMPILER || defined __clang__ |
239 | #define ECB_MEMORY_FENCE __sync_synchronize () |
265 | #define ECB_MEMORY_FENCE __sync_synchronize () |
240 | #elif _MSC_VER >= 1500 /* VC++ 2008 */ |
266 | #elif _MSC_VER >= 1500 /* VC++ 2008 */ |
… | |
… | |
592 | 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 >> (32 - count)) | (x << count); } |
593 | 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 << (32 - count)) | (x >> count); } |
594 | 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 >> (64 - count)) | (x << count); } |
595 | 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 << (64 - count)) | (x >> count); } |
596 | |
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 |
|
|
660 | |
597 | #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)) |
598 | #if ECB_GCC_VERSION(4,8) || ECB_CLANG_BUILTIN(__builtin_bswap16) |
662 | #if ECB_GCC_VERSION(4,8) || ECB_CLANG_BUILTIN(__builtin_bswap16) |
599 | #define ecb_bswap16(x) __builtin_bswap16 (x) |
663 | #define ecb_bswap16(x) __builtin_bswap16 (x) |
600 | #else |
664 | #else |
601 | #define ecb_bswap16(x) (__builtin_bswap32 (x) >> 16) |
665 | #define ecb_bswap16(x) (__builtin_bswap32 (x) >> 16) |
… | |
… | |
672 | ecb_inline ecb_const ecb_bool ecb_big_endian (void); |
736 | ecb_inline ecb_const ecb_bool ecb_big_endian (void); |
673 | ecb_inline ecb_const ecb_bool ecb_big_endian (void) { return ecb_byteorder_helper () == 0x11223344; } |
737 | ecb_inline ecb_const ecb_bool ecb_big_endian (void) { return ecb_byteorder_helper () == 0x11223344; } |
674 | ecb_inline ecb_const ecb_bool ecb_little_endian (void); |
738 | ecb_inline ecb_const ecb_bool ecb_little_endian (void); |
675 | ecb_inline ecb_const ecb_bool ecb_little_endian (void) { return ecb_byteorder_helper () == 0x44332211; } |
739 | ecb_inline ecb_const ecb_bool ecb_little_endian (void) { return ecb_byteorder_helper () == 0x44332211; } |
676 | |
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 | /* division */ |
|
|
813 | |
677 | #if ECB_GCC_VERSION(3,0) || ECB_C99 |
814 | #if ECB_GCC_VERSION(3,0) || ECB_C99 |
|
|
815 | /* C99 tightened the definition of %, so we can use a more efficient version */ |
678 | #define ecb_mod(m,n) ((m) % (n) + ((m) % (n) < 0 ? (n) : 0)) |
816 | #define ecb_mod(m,n) ((m) % (n) + ((m) % (n) < 0 ? (n) : 0)) |
679 | #else |
817 | #else |
680 | #define ecb_mod(m,n) ((m) < 0 ? ((n) - 1 - ((-1 - (m)) % (n))) : ((m) % (n))) |
818 | #define ecb_mod(m,n) ((m) < 0 ? ((n) - 1 - ((-1 - (m)) % (n))) : ((m) % (n))) |
681 | #endif |
819 | #endif |
682 | |
820 | |
… | |
… | |
693 | } |
831 | } |
694 | #else |
832 | #else |
695 | #define ecb_div_rd(val,div) ((val) < 0 ? - ((-(val) + (div) - 1) / (div)) : ((val) ) / (div)) |
833 | #define ecb_div_rd(val,div) ((val) < 0 ? - ((-(val) + (div) - 1) / (div)) : ((val) ) / (div)) |
696 | #define ecb_div_ru(val,div) ((val) < 0 ? - ((-(val) ) / (div)) : ((val) + (div) - 1) / (div)) |
834 | #define ecb_div_ru(val,div) ((val) < 0 ? - ((-(val) ) / (div)) : ((val) + (div) - 1) / (div)) |
697 | #endif |
835 | #endif |
|
|
836 | |
|
|
837 | /*****************************************************************************/ |
|
|
838 | /* array length */ |
698 | |
839 | |
699 | #if ecb_cplusplus_does_not_suck |
840 | #if ecb_cplusplus_does_not_suck |
700 | /* does not work for local types (http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2657.htm) */ |
841 | /* does not work for local types (http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2657.htm) */ |
701 | template<typename T, int N> |
842 | template<typename T, int N> |
702 | static inline int ecb_array_length (const T (&arr)[N]) |
843 | static inline int ecb_array_length (const T (&arr)[N]) |
… | |
… | |
704 | return N; |
845 | return N; |
705 | } |
846 | } |
706 | #else |
847 | #else |
707 | #define ecb_array_length(name) (sizeof (name) / sizeof (name [0])) |
848 | #define ecb_array_length(name) (sizeof (name) / sizeof (name [0])) |
708 | #endif |
849 | #endif |
|
|
850 | |
|
|
851 | /*****************************************************************************/ |
|
|
852 | /* IEEE 754-2008 half float conversions */ |
709 | |
853 | |
710 | ecb_function_ ecb_const uint32_t ecb_binary16_to_binary32 (uint32_t x); |
854 | ecb_function_ ecb_const uint32_t ecb_binary16_to_binary32 (uint32_t x); |
711 | ecb_function_ ecb_const uint32_t |
855 | ecb_function_ ecb_const uint32_t |
712 | ecb_binary16_to_binary32 (uint32_t x) |
856 | ecb_binary16_to_binary32 (uint32_t x) |
713 | { |
857 | { |
… | |
… | |
742 | ecb_function_ ecb_const uint16_t ecb_binary32_to_binary16 (uint32_t x); |
886 | ecb_function_ ecb_const uint16_t ecb_binary32_to_binary16 (uint32_t x); |
743 | ecb_function_ ecb_const uint16_t |
887 | ecb_function_ ecb_const uint16_t |
744 | ecb_binary32_to_binary16 (uint32_t x) |
888 | ecb_binary32_to_binary16 (uint32_t x) |
745 | { |
889 | { |
746 | unsigned int s = (x >> 16) & 0x00008000; /* sign bit, the easy part */ |
890 | unsigned int s = (x >> 16) & 0x00008000; /* sign bit, the easy part */ |
747 | unsigned int e = ((x >> 23) & 0x000000ff) - (127 - 15); /* the desired exponent */ |
891 | int e = ((x >> 23) & 0x000000ff) - (127 - 15); /* the desired exponent */ |
748 | unsigned int m = x & 0x007fffff; |
892 | unsigned int m = x & 0x007fffff; |
749 | |
893 | |
750 | x &= 0x7fffffff; |
894 | x &= 0x7fffffff; |
751 | |
895 | |
752 | /* if it's within range of binary16 normals, use fast path */ |
896 | /* if it's within range of binary16 normals, use fast path */ |
… | |
… | |
799 | |
943 | |
800 | /* handle NaNs, preserve leftmost nan bits, but make sure we don't turn them into infinities */ |
944 | /* handle NaNs, preserve leftmost nan bits, but make sure we don't turn them into infinities */ |
801 | m >>= 13; |
945 | m >>= 13; |
802 | |
946 | |
803 | return s | 0x7c00 | m | !m; |
947 | return s | 0x7c00 | m | !m; |
|
|
948 | } |
|
|
949 | |
|
|
950 | /*******************************************************************************/ |
|
|
951 | /* fast integer to ascii */ |
|
|
952 | |
|
|
953 | /* |
|
|
954 | * This code is pretty complicated because it is general. The idea behind it, |
|
|
955 | * however, is pretty simple: first, the number is multiplied with a scaling |
|
|
956 | * factor (2**bits / 10**(digits-1)) to convert the integer into a fixed-point |
|
|
957 | * number with the first digit in the upper bits. |
|
|
958 | * Then this digit is converted to text and masked out. The resulting number |
|
|
959 | * is then multiplied by 10, by multiplying the fixed point representation |
|
|
960 | * by 5 and shifting the (binary) decimal point one to the right, so a 4.28 |
|
|
961 | * format becomes 5.27, 6.26 and so on. |
|
|
962 | * The rest involves only advancing the pointer if we already generated a |
|
|
963 | * non-zero digit, so leading zeroes are overwritten. |
|
|
964 | */ |
|
|
965 | |
|
|
966 | // simply return a mask with "bits" bits set |
|
|
967 | #define ecb_i2a_mask(type,bits) ((((type)1) << (bits)) - 1) |
|
|
968 | |
|
|
969 | // oputput a single digit. maskvalue is 10**digitidx |
|
|
970 | #define ecb_i2a_digit(type,bits,digitmask,maskvalue,digitidx) \ |
|
|
971 | if (digitmask >= maskvalue) /* constant, used to decide how many digits to generate */ \ |
|
|
972 | { \ |
|
|
973 | char digit = x >> (bits - digitidx); /* calculate the topmost digit */ \ |
|
|
974 | *ptr = digit + '0'; /* output it */ \ |
|
|
975 | nz = (digitmask == maskvalue) || nz || digit; /* first term == always output last digit */ \ |
|
|
976 | ptr += nz; /* output digit only if non-zero digit seen */ \ |
|
|
977 | x = (x & ecb_i2a_mask (type, bits - digitidx)) * 5; /* *10, but shift decimal point right */ \ |
|
|
978 | } |
|
|
979 | |
|
|
980 | // convert integer to fixed point format and multiply out digits, highest first |
|
|
981 | // requires magic constants: max. digits and number of bits after the decimal point |
|
|
982 | #define ecb_i2a_def(suffix,ptr,v,type,bits,digitmask,lz) \ |
|
|
983 | ecb_inline char *ecb_i2a_ ## suffix (char *ptr, uint32_t u) \ |
|
|
984 | { \ |
|
|
985 | char nz = lz; /* non-zero digit seen? */ \ |
|
|
986 | /* convert to x.bits fixed-point */ \ |
|
|
987 | type x = u * ((ecb_i2a_mask (type, bits) + digitmask) / digitmask); \ |
|
|
988 | /* output up to 10 digits */ \ |
|
|
989 | ecb_i2a_digit (type,bits,digitmask, 1, 0); \ |
|
|
990 | ecb_i2a_digit (type,bits,digitmask, 10, 1); \ |
|
|
991 | ecb_i2a_digit (type,bits,digitmask, 100, 2); \ |
|
|
992 | ecb_i2a_digit (type,bits,digitmask, 1000, 3); \ |
|
|
993 | ecb_i2a_digit (type,bits,digitmask, 10000, 4); \ |
|
|
994 | ecb_i2a_digit (type,bits,digitmask, 100000, 5); \ |
|
|
995 | ecb_i2a_digit (type,bits,digitmask, 1000000, 6); \ |
|
|
996 | ecb_i2a_digit (type,bits,digitmask, 10000000, 7); \ |
|
|
997 | ecb_i2a_digit (type,bits,digitmask, 100000000, 8); \ |
|
|
998 | ecb_i2a_digit (type,bits,digitmask, 1000000000, 9); \ |
|
|
999 | return ptr; \ |
|
|
1000 | } |
|
|
1001 | |
|
|
1002 | // predefined versions of the above, for various digits |
|
|
1003 | // ecb_i2a_xN = almost N digits, limit defined by macro |
|
|
1004 | // ecb_i2a_N = up to N digits, leading zeroes suppressed |
|
|
1005 | // ecb_i2a_0N = exactly N digits, including leading zeroes |
|
|
1006 | |
|
|
1007 | // non-leading-zero versions, limited range |
|
|
1008 | #define ECB_I2A_MAX_X5 59074 // limit for ecb_i2a_x5 |
|
|
1009 | #define ECB_I2A_MAX_X10 2932500665 // limit for ecb_i2a_x10 |
|
|
1010 | ecb_i2a_def ( x5, ptr, v, uint32_t, 26, 10000, 0) |
|
|
1011 | ecb_i2a_def (x10, ptr, v, uint64_t, 60, 1000000000, 0) |
|
|
1012 | |
|
|
1013 | // non-leading zero versions, all digits, 4 and 9 are optimal for 32/64 bit |
|
|
1014 | ecb_i2a_def ( 2, ptr, v, uint32_t, 10, 10, 0) |
|
|
1015 | ecb_i2a_def ( 3, ptr, v, uint32_t, 12, 100, 0) |
|
|
1016 | ecb_i2a_def ( 4, ptr, v, uint32_t, 26, 1000, 0) |
|
|
1017 | ecb_i2a_def ( 5, ptr, v, uint64_t, 30, 10000, 0) |
|
|
1018 | ecb_i2a_def ( 6, ptr, v, uint64_t, 36, 100000, 0) |
|
|
1019 | ecb_i2a_def ( 7, ptr, v, uint64_t, 44, 1000000, 0) |
|
|
1020 | ecb_i2a_def ( 8, ptr, v, uint64_t, 50, 10000000, 0) |
|
|
1021 | ecb_i2a_def ( 9, ptr, v, uint64_t, 56, 100000000, 0) |
|
|
1022 | |
|
|
1023 | // leading-zero versions, all digits, 04 and 09 are optimal for 32/64 bit |
|
|
1024 | ecb_i2a_def (02, ptr, v, uint32_t, 10, 10, 1) |
|
|
1025 | ecb_i2a_def (03, ptr, v, uint32_t, 12, 100, 1) |
|
|
1026 | ecb_i2a_def (04, ptr, v, uint32_t, 26, 1000, 1) |
|
|
1027 | ecb_i2a_def (05, ptr, v, uint64_t, 30, 10000, 1) |
|
|
1028 | ecb_i2a_def (06, ptr, v, uint64_t, 36, 100000, 1) |
|
|
1029 | ecb_i2a_def (07, ptr, v, uint64_t, 44, 1000000, 1) |
|
|
1030 | ecb_i2a_def (08, ptr, v, uint64_t, 50, 10000000, 1) |
|
|
1031 | ecb_i2a_def (09, ptr, v, uint64_t, 56, 100000000, 1) |
|
|
1032 | |
|
|
1033 | #define ECB_I2A_I32_DIGITS 11 |
|
|
1034 | #define ECB_I2A_U32_DIGITS 10 |
|
|
1035 | #define ECB_I2A_I64_DIGITS 20 |
|
|
1036 | #define ECB_I2A_U64_DIGITS 21 |
|
|
1037 | #define ECB_I2A_MAX_DIGITS 21 |
|
|
1038 | |
|
|
1039 | ecb_inline char * |
|
|
1040 | ecb_i2a_u32 (char *ptr, uint32_t u) |
|
|
1041 | { |
|
|
1042 | #if ECB_64BIT_NATIVE |
|
|
1043 | if (ecb_expect_true (u <= ECB_I2A_MAX_X10)) |
|
|
1044 | ptr = ecb_i2a_x10 (ptr, u); |
|
|
1045 | else // x10 almost, but not fully, covers 32 bit |
|
|
1046 | { |
|
|
1047 | uint32_t u1 = u % 1000000000; |
|
|
1048 | uint32_t u2 = u / 1000000000; |
|
|
1049 | |
|
|
1050 | *ptr++ = u2 + '0'; |
|
|
1051 | ptr = ecb_i2a_09 (ptr, u1); |
|
|
1052 | } |
|
|
1053 | #else |
|
|
1054 | if (ecb_expect_true (u <= ECB_I2A_MAX_X5)) |
|
|
1055 | ecb_i2a_x5 (ptr, u); |
|
|
1056 | else if (ecb_expect_true (u <= ECB_I2A_MAX_X5 * 10000)) |
|
|
1057 | { |
|
|
1058 | uint32_t u1 = u % 10000; |
|
|
1059 | uint32_t u2 = u / 10000; |
|
|
1060 | |
|
|
1061 | ptr = ecb_i2a_x5 (ptr, u2); |
|
|
1062 | ptr = ecb_i2a_04 (ptr, u1); |
|
|
1063 | } |
|
|
1064 | else |
|
|
1065 | { |
|
|
1066 | uint32_t u1 = u % 10000; |
|
|
1067 | uint32_t ua = u / 10000; |
|
|
1068 | uint32_t u2 = ua % 10000; |
|
|
1069 | uint32_t u3 = ua / 10000; |
|
|
1070 | |
|
|
1071 | ptr = ecb_i2a_2 (ptr, u3); |
|
|
1072 | ptr = ecb_i2a_04 (ptr, u2); |
|
|
1073 | ptr = ecb_i2a_04 (ptr, u1); |
|
|
1074 | } |
|
|
1075 | #endif |
|
|
1076 | |
|
|
1077 | return ptr; |
|
|
1078 | } |
|
|
1079 | |
|
|
1080 | ecb_inline char * |
|
|
1081 | ecb_i2a_i32 (char *ptr, int32_t v) |
|
|
1082 | { |
|
|
1083 | *ptr = '-'; ptr += v < 0; |
|
|
1084 | uint32_t u = v < 0 ? -(uint32_t)v : v; |
|
|
1085 | |
|
|
1086 | #if ECB_64BIT_NATIVE |
|
|
1087 | ptr = ecb_i2a_x10 (ptr, u); // x10 fully covers 31 bit |
|
|
1088 | #else |
|
|
1089 | ptr = ecb_i2a_u32 (ptr, u); |
|
|
1090 | #endif |
|
|
1091 | |
|
|
1092 | return ptr; |
|
|
1093 | } |
|
|
1094 | |
|
|
1095 | ecb_inline char * |
|
|
1096 | ecb_i2a_u64 (char *ptr, uint64_t u) |
|
|
1097 | { |
|
|
1098 | #if ECB_64BIT_NATIVE |
|
|
1099 | if (ecb_expect_true (u <= ECB_I2A_MAX_X10)) |
|
|
1100 | ptr = ecb_i2a_x10 (ptr, u); |
|
|
1101 | else if (ecb_expect_false (u <= ECB_I2A_MAX_X10 * 1000000000)) |
|
|
1102 | { |
|
|
1103 | uint64_t u1 = u % 1000000000; |
|
|
1104 | uint64_t u2 = u / 1000000000; |
|
|
1105 | |
|
|
1106 | ptr = ecb_i2a_x10 (ptr, u2); |
|
|
1107 | ptr = ecb_i2a_09 (ptr, u1); |
|
|
1108 | } |
|
|
1109 | else |
|
|
1110 | { |
|
|
1111 | uint64_t u1 = u % 1000000000; |
|
|
1112 | uint64_t ua = u / 1000000000; |
|
|
1113 | uint64_t u2 = ua % 1000000000; |
|
|
1114 | uint64_t u3 = ua / 1000000000; |
|
|
1115 | |
|
|
1116 | ptr = ecb_i2a_2 (ptr, u3); |
|
|
1117 | ptr = ecb_i2a_09 (ptr, u2); |
|
|
1118 | ptr = ecb_i2a_09 (ptr, u1); |
|
|
1119 | } |
|
|
1120 | #else |
|
|
1121 | if (ecb_expect_true (u <= ECB_I2A_MAX_X5)) |
|
|
1122 | ptr = ecb_i2a_x5 (ptr, u); |
|
|
1123 | else |
|
|
1124 | { |
|
|
1125 | uint64_t u1 = u % 10000; |
|
|
1126 | uint64_t u2 = u / 10000; |
|
|
1127 | |
|
|
1128 | ptr = ecb_i2a_u64 (ptr, u2); |
|
|
1129 | ptr = ecb_i2a_04 (ptr, u1); |
|
|
1130 | } |
|
|
1131 | #endif |
|
|
1132 | |
|
|
1133 | return ptr; |
|
|
1134 | } |
|
|
1135 | |
|
|
1136 | ecb_inline char * |
|
|
1137 | ecb_i2a_i64 (char *ptr, int64_t v) |
|
|
1138 | { |
|
|
1139 | *ptr = '-'; ptr += v < 0; |
|
|
1140 | uint64_t u = v < 0 ? -(uint64_t)v : v; |
|
|
1141 | |
|
|
1142 | #if ECB_64BIT_NATIVE |
|
|
1143 | if (ecb_expect_true (u <= ECB_I2A_MAX_X10)) |
|
|
1144 | ptr = ecb_i2a_x10 (ptr, u); |
|
|
1145 | else if (ecb_expect_false (u <= ECB_I2A_MAX_X10 * 1000000000)) |
|
|
1146 | { |
|
|
1147 | uint64_t u1 = u % 1000000000; |
|
|
1148 | uint64_t u2 = u / 1000000000; |
|
|
1149 | |
|
|
1150 | ptr = ecb_i2a_x10 (ptr, u2); |
|
|
1151 | ptr = ecb_i2a_09 (ptr, u1); |
|
|
1152 | } |
|
|
1153 | else |
|
|
1154 | { |
|
|
1155 | uint64_t u1 = u % 1000000000; |
|
|
1156 | uint64_t ua = u / 1000000000; |
|
|
1157 | uint64_t u2 = ua % 1000000000; |
|
|
1158 | uint64_t u3 = ua / 1000000000; |
|
|
1159 | |
|
|
1160 | // 2**31 is 19 digits, so the top is exactly one digit |
|
|
1161 | *ptr++ = u3 + '0'; |
|
|
1162 | ptr = ecb_i2a_09 (ptr, u2); |
|
|
1163 | ptr = ecb_i2a_09 (ptr, u1); |
|
|
1164 | } |
|
|
1165 | #else |
|
|
1166 | ptr = ecb_i2a_u64 (ptr, u); |
|
|
1167 | #endif |
|
|
1168 | |
|
|
1169 | return ptr; |
804 | } |
1170 | } |
805 | |
1171 | |
806 | /*******************************************************************************/ |
1172 | /*******************************************************************************/ |
807 | /* floating point stuff, can be disabled by defining ECB_NO_LIBM */ |
1173 | /* floating point stuff, can be disabled by defining ECB_NO_LIBM */ |
808 | |
1174 | |
… | |
… | |
822 | || defined __sh__ \ |
1188 | || defined __sh__ \ |
823 | || defined _M_IX86 || defined ECB_MSVC_AMD64 || defined _M_IA64 \ |
1189 | || defined _M_IX86 || defined ECB_MSVC_AMD64 || defined _M_IA64 \ |
824 | || (defined __arm__ && (defined __ARM_EABI__ || defined __EABI__ || defined __VFP_FP__ || defined _WIN32_WCE || defined __ANDROID__)) \ |
1190 | || (defined __arm__ && (defined __ARM_EABI__ || defined __EABI__ || defined __VFP_FP__ || defined _WIN32_WCE || defined __ANDROID__)) \ |
825 | || defined __aarch64__ |
1191 | || defined __aarch64__ |
826 | #define ECB_STDFP 1 |
1192 | #define ECB_STDFP 1 |
827 | #include <string.h> /* for memcpy */ |
|
|
828 | #else |
1193 | #else |
829 | #define ECB_STDFP 0 |
1194 | #define ECB_STDFP 0 |
830 | #endif |
1195 | #endif |
831 | |
1196 | |
832 | #ifndef ECB_NO_LIBM |
1197 | #ifndef ECB_NO_LIBM |