1 | /* |
1 | /* |
2 | * libecb - http://software.schmorp.de/pkg/libecb |
2 | * libecb - http://software.schmorp.de/pkg/libecb |
3 | * |
3 | * |
4 | * Copyright (©) 2009-2013 Marc Alexander Lehmann <libecb@schmorp.de> |
4 | * Copyright (©) 2009-2015 Marc Alexander Lehmann <libecb@schmorp.de> |
5 | * Copyright (©) 2011 Emanuele Giaquinta |
5 | * Copyright (©) 2011 Emanuele Giaquinta |
6 | * All rights reserved. |
6 | * All rights reserved. |
7 | * |
7 | * |
8 | * Redistribution and use in source and binary forms, with or without modifica- |
8 | * Redistribution and use in source and binary forms, with or without modifica- |
9 | * tion, are permitted provided that the following conditions are met: |
9 | * tion, are permitted provided that the following conditions are met: |
… | |
… | |
23 | * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; |
23 | * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; |
24 | * OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, |
24 | * OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, |
25 | * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTH- |
25 | * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTH- |
26 | * ERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED |
26 | * ERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED |
27 | * OF THE POSSIBILITY OF SUCH DAMAGE. |
27 | * OF THE POSSIBILITY OF SUCH DAMAGE. |
|
|
28 | * |
|
|
29 | * Alternatively, the contents of this file may be used under the terms of |
|
|
30 | * the GNU General Public License ("GPL") version 2 or any later version, |
|
|
31 | * in which case the provisions of the GPL are applicable instead of |
|
|
32 | * the above. If you wish to allow the use of your version of this file |
|
|
33 | * only under the terms of the GPL and not to allow others to use your |
|
|
34 | * version of this file under the BSD license, indicate your decision |
|
|
35 | * by deleting the provisions above and replace them with the notice |
|
|
36 | * and other provisions required by the GPL. If you do not delete the |
|
|
37 | * provisions above, a recipient may use your version of this file under |
|
|
38 | * either the BSD or the GPL. |
28 | */ |
39 | */ |
29 | |
40 | |
30 | #ifndef ECB_H |
41 | #ifndef ECB_H |
31 | #define ECB_H |
42 | #define ECB_H |
32 | |
43 | |
33 | /* 16 bits major, 16 bits minor */ |
44 | /* 16 bits major, 16 bits minor */ |
34 | #define ECB_VERSION 0x00010003 |
45 | #define ECB_VERSION 0x00010008 |
35 | |
46 | |
36 | #ifdef _WIN32 |
47 | #ifdef _WIN32 |
37 | typedef signed char int8_t; |
48 | typedef signed char int8_t; |
38 | typedef unsigned char uint8_t; |
49 | typedef unsigned char uint8_t; |
|
|
50 | typedef signed char int_fast8_t; |
|
|
51 | typedef unsigned char uint_fast8_t; |
39 | typedef signed short int16_t; |
52 | typedef signed short int16_t; |
40 | typedef unsigned short uint16_t; |
53 | typedef unsigned short uint16_t; |
|
|
54 | typedef signed int int_fast16_t; |
|
|
55 | typedef unsigned int uint_fast16_t; |
41 | typedef signed int int32_t; |
56 | typedef signed int int32_t; |
42 | typedef unsigned int uint32_t; |
57 | typedef unsigned int uint32_t; |
|
|
58 | typedef signed int int_fast32_t; |
|
|
59 | typedef unsigned int uint_fast32_t; |
43 | #if __GNUC__ |
60 | #if __GNUC__ |
44 | typedef signed long long int64_t; |
61 | typedef signed long long int64_t; |
45 | typedef unsigned long long uint64_t; |
62 | typedef unsigned long long uint64_t; |
46 | #else /* _MSC_VER || __BORLANDC__ */ |
63 | #else /* _MSC_VER || __BORLANDC__ */ |
47 | typedef signed __int64 int64_t; |
64 | typedef signed __int64 int64_t; |
48 | typedef unsigned __int64 uint64_t; |
65 | typedef unsigned __int64 uint64_t; |
49 | #endif |
66 | #endif |
|
|
67 | typedef int64_t int_fast64_t; |
|
|
68 | typedef uint64_t uint_fast64_t; |
50 | #ifdef _WIN64 |
69 | #ifdef _WIN64 |
51 | #define ECB_PTRSIZE 8 |
70 | #define ECB_PTRSIZE 8 |
52 | typedef uint64_t uintptr_t; |
71 | typedef uint64_t uintptr_t; |
53 | typedef int64_t intptr_t; |
72 | typedef int64_t intptr_t; |
54 | #else |
73 | #else |
… | |
… | |
56 | typedef uint32_t uintptr_t; |
75 | typedef uint32_t uintptr_t; |
57 | typedef int32_t intptr_t; |
76 | typedef int32_t intptr_t; |
58 | #endif |
77 | #endif |
59 | #else |
78 | #else |
60 | #include <inttypes.h> |
79 | #include <inttypes.h> |
61 | #if UINTMAX_MAX > 0xffffffffU |
80 | #if (defined INTPTR_MAX ? INTPTR_MAX : ULONG_MAX) > 0xffffffffU |
62 | #define ECB_PTRSIZE 8 |
81 | #define ECB_PTRSIZE 8 |
63 | #else |
82 | #else |
64 | #define ECB_PTRSIZE 4 |
83 | #define ECB_PTRSIZE 4 |
65 | #endif |
84 | #endif |
66 | #endif |
85 | #endif |
67 | |
86 | |
|
|
87 | #define ECB_GCC_AMD64 (__amd64 || __amd64__ || __x86_64 || __x86_64__) |
|
|
88 | #define ECB_MSVC_AMD64 (_M_AMD64 || _M_X64) |
|
|
89 | |
|
|
90 | #ifndef ECB_OPTIMIZE_SIZE |
|
|
91 | #if __OPTIMIZE_SIZE__ |
|
|
92 | #define ECB_OPTIMIZE_SIZE 1 |
|
|
93 | #else |
|
|
94 | #define ECB_OPTIMIZE_SIZE 0 |
|
|
95 | #endif |
|
|
96 | #endif |
|
|
97 | |
68 | /* work around x32 idiocy by defining proper macros */ |
98 | /* work around x32 idiocy by defining proper macros */ |
69 | #if __x86_64 || _M_AMD64 |
99 | #if ECB_GCC_AMD64 || ECB_MSVC_AMD64 |
70 | #if _ILP32 |
100 | #if _ILP32 |
71 | #define ECB_AMD64_X32 1 |
101 | #define ECB_AMD64_X32 1 |
72 | #else |
102 | #else |
73 | #define ECB_AMD64 1 |
103 | #define ECB_AMD64 1 |
74 | #endif |
104 | #endif |
… | |
… | |
79 | * causing enormous grief in return for some better fake benchmark numbers. |
109 | * causing enormous grief in return for some better fake benchmark numbers. |
80 | * or so. |
110 | * or so. |
81 | * we try to detect these and simply assume they are not gcc - if they have |
111 | * we try to detect these and simply assume they are not gcc - if they have |
82 | * an issue with that they should have done it right in the first place. |
112 | * an issue with that they should have done it right in the first place. |
83 | */ |
113 | */ |
84 | #ifndef ECB_GCC_VERSION |
|
|
85 | #if !defined __GNUC_MINOR__ || defined __INTEL_COMPILER || defined __SUNPRO_C || defined __SUNPRO_CC || defined __llvm__ || defined __clang__ |
114 | #if !defined __GNUC_MINOR__ || defined __INTEL_COMPILER || defined __SUNPRO_C || defined __SUNPRO_CC || defined __llvm__ || defined __clang__ |
86 | #define ECB_GCC_VERSION(major,minor) 0 |
115 | #define ECB_GCC_VERSION(major,minor) 0 |
87 | #else |
116 | #else |
88 | #define ECB_GCC_VERSION(major,minor) (__GNUC__ > (major) || (__GNUC__ == (major) && __GNUC_MINOR__ >= (minor))) |
117 | #define ECB_GCC_VERSION(major,minor) (__GNUC__ > (major) || (__GNUC__ == (major) && __GNUC_MINOR__ >= (minor))) |
89 | #endif |
118 | #endif |
90 | #endif |
|
|
91 | |
119 | |
92 | #define ECB_C (__STDC__+0) /* this assumes that __STDC__ is either empty or a number */ |
120 | #define ECB_CLANG_VERSION(major,minor) (__clang_major__ > (major) || (__clang_major__ == (major) && __clang_minor__ >= (minor))) |
93 | #define ECB_C99 (__STDC_VERSION__ >= 199901L) |
121 | |
94 | #define ECB_C11 (__STDC_VERSION__ >= 201112L) |
122 | #if __clang__ && defined __has_builtin |
|
|
123 | #define ECB_CLANG_BUILTIN(x) __has_builtin (x) |
|
|
124 | #else |
|
|
125 | #define ECB_CLANG_BUILTIN(x) 0 |
|
|
126 | #endif |
|
|
127 | |
|
|
128 | #if __clang__ && defined __has_extension |
|
|
129 | #define ECB_CLANG_EXTENSION(x) __has_extension (x) |
|
|
130 | #else |
|
|
131 | #define ECB_CLANG_EXTENSION(x) 0 |
|
|
132 | #endif |
|
|
133 | |
95 | #define ECB_CPP (__cplusplus+0) |
134 | #define ECB_CPP (__cplusplus+0) |
96 | #define ECB_CPP11 (__cplusplus >= 201103L) |
135 | #define ECB_CPP11 (__cplusplus >= 201103L) |
|
|
136 | #define ECB_CPP14 (__cplusplus >= 201402L) |
|
|
137 | #define ECB_CPP17 (__cplusplus >= 201703L) |
|
|
138 | |
|
|
139 | #if ECB_CPP |
|
|
140 | #define ECB_C 0 |
|
|
141 | #define ECB_STDC_VERSION 0 |
|
|
142 | #else |
|
|
143 | #define ECB_C 1 |
|
|
144 | #define ECB_STDC_VERSION __STDC_VERSION__ |
|
|
145 | #endif |
|
|
146 | |
|
|
147 | #define ECB_C99 (ECB_STDC_VERSION >= 199901L) |
|
|
148 | #define ECB_C11 (ECB_STDC_VERSION >= 201112L) |
|
|
149 | #define ECB_C17 (ECB_STDC_VERSION >= 201710L) |
97 | |
150 | |
98 | #if ECB_CPP |
151 | #if ECB_CPP |
99 | #define ECB_EXTERN_C extern "C" |
152 | #define ECB_EXTERN_C extern "C" |
100 | #define ECB_EXTERN_C_BEG ECB_EXTERN_C { |
153 | #define ECB_EXTERN_C_BEG ECB_EXTERN_C { |
101 | #define ECB_EXTERN_C_END } |
154 | #define ECB_EXTERN_C_END } |
… | |
… | |
116 | |
169 | |
117 | #if ECB_NO_SMP |
170 | #if ECB_NO_SMP |
118 | #define ECB_MEMORY_FENCE do { } while (0) |
171 | #define ECB_MEMORY_FENCE do { } while (0) |
119 | #endif |
172 | #endif |
120 | |
173 | |
|
|
174 | /* http://www-01.ibm.com/support/knowledgecenter/SSGH3R_13.1.0/com.ibm.xlcpp131.aix.doc/compiler_ref/compiler_builtins.html */ |
|
|
175 | #if __xlC__ && ECB_CPP |
|
|
176 | #include <builtins.h> |
|
|
177 | #endif |
|
|
178 | |
|
|
179 | #if 1400 <= _MSC_VER |
|
|
180 | #include <intrin.h> /* fence functions _ReadBarrier, also bit search functions _BitScanReverse */ |
|
|
181 | #endif |
|
|
182 | |
121 | #ifndef ECB_MEMORY_FENCE |
183 | #ifndef ECB_MEMORY_FENCE |
122 | #if ECB_GCC_VERSION(2,5) || defined __INTEL_COMPILER || (__llvm__ && __GNUC__) || __SUNPRO_C >= 0x5110 || __SUNPRO_CC >= 0x5110 |
184 | #if ECB_GCC_VERSION(2,5) || defined __INTEL_COMPILER || (__llvm__ && __GNUC__) || __SUNPRO_C >= 0x5110 || __SUNPRO_CC >= 0x5110 |
|
|
185 | #define ECB_MEMORY_FENCE_RELAXED __asm__ __volatile__ ("" : : : "memory") |
123 | #if __i386 || __i386__ |
186 | #if __i386 || __i386__ |
124 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("lock; orb $0, -1(%%esp)" : : : "memory") |
187 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("lock; orb $0, -1(%%esp)" : : : "memory") |
125 | #define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("" : : : "memory") |
188 | #define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("" : : : "memory") |
126 | #define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("") |
189 | #define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("" : : : "memory") |
127 | #elif __amd64 || __amd64__ || __x86_64 || __x86_64__ |
190 | #elif ECB_GCC_AMD64 |
128 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("mfence" : : : "memory") |
191 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("mfence" : : : "memory") |
129 | #define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("" : : : "memory") |
192 | #define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("" : : : "memory") |
130 | #define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("") |
193 | #define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("" : : : "memory") |
131 | #elif __powerpc__ || __ppc__ || __powerpc64__ || __ppc64__ |
194 | #elif __powerpc__ || __ppc__ || __powerpc64__ || __ppc64__ |
132 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("sync" : : : "memory") |
195 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("sync" : : : "memory") |
|
|
196 | #elif defined __ARM_ARCH_2__ \ |
|
|
197 | || defined __ARM_ARCH_3__ || defined __ARM_ARCH_3M__ \ |
|
|
198 | || defined __ARM_ARCH_4__ || defined __ARM_ARCH_4T__ \ |
|
|
199 | || defined __ARM_ARCH_5__ || defined __ARM_ARCH_5E__ \ |
|
|
200 | || defined __ARM_ARCH_5T__ || defined __ARM_ARCH_5TE__ \ |
|
|
201 | || defined __ARM_ARCH_5TEJ__ |
|
|
202 | /* should not need any, unless running old code on newer cpu - arm doesn't support that */ |
133 | #elif defined __ARM_ARCH_6__ || defined __ARM_ARCH_6J__ \ |
203 | #elif defined __ARM_ARCH_6__ || defined __ARM_ARCH_6J__ \ |
134 | || defined __ARM_ARCH_6K__ || defined __ARM_ARCH_6ZK__ |
204 | || defined __ARM_ARCH_6K__ || defined __ARM_ARCH_6ZK__ \ |
|
|
205 | || defined __ARM_ARCH_6T2__ |
135 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("mcr p15,0,%0,c7,c10,5" : : "r" (0) : "memory") |
206 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("mcr p15,0,%0,c7,c10,5" : : "r" (0) : "memory") |
136 | #elif defined __ARM_ARCH_7__ || defined __ARM_ARCH_7A__ \ |
207 | #elif defined __ARM_ARCH_7__ || defined __ARM_ARCH_7A__ \ |
137 | || defined __ARM_ARCH_7M__ || defined __ARM_ARCH_7R__ |
208 | || defined __ARM_ARCH_7R__ || defined __ARM_ARCH_7M__ |
138 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("dmb" : : : "memory") |
209 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("dmb" : : : "memory") |
139 | #elif __sparc || __sparc__ |
210 | #elif __aarch64__ |
|
|
211 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("dmb ish" : : : "memory") |
|
|
212 | #elif (__sparc || __sparc__) && !(__sparc_v8__ || defined __sparcv8) |
140 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("membar #LoadStore | #LoadLoad | #StoreStore | #StoreLoad" : : : "memory") |
213 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("membar #LoadStore | #LoadLoad | #StoreStore | #StoreLoad" : : : "memory") |
141 | #define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("membar #LoadStore | #LoadLoad" : : : "memory") |
214 | #define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("membar #LoadStore | #LoadLoad" : : : "memory") |
142 | #define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("membar #LoadStore | #StoreStore") |
215 | #define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("membar #LoadStore | #StoreStore") |
143 | #elif defined __s390__ || defined __s390x__ |
216 | #elif defined __s390__ || defined __s390x__ |
144 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("bcr 15,0" : : : "memory") |
217 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("bcr 15,0" : : : "memory") |
… | |
… | |
165 | |
238 | |
166 | #ifndef ECB_MEMORY_FENCE |
239 | #ifndef ECB_MEMORY_FENCE |
167 | #if ECB_GCC_VERSION(4,7) |
240 | #if ECB_GCC_VERSION(4,7) |
168 | /* see comment below (stdatomic.h) about the C11 memory model. */ |
241 | /* see comment below (stdatomic.h) about the C11 memory model. */ |
169 | #define ECB_MEMORY_FENCE __atomic_thread_fence (__ATOMIC_SEQ_CST) |
242 | #define ECB_MEMORY_FENCE __atomic_thread_fence (__ATOMIC_SEQ_CST) |
|
|
243 | #define ECB_MEMORY_FENCE_ACQUIRE __atomic_thread_fence (__ATOMIC_ACQUIRE) |
|
|
244 | #define ECB_MEMORY_FENCE_RELEASE __atomic_thread_fence (__ATOMIC_RELEASE) |
|
|
245 | #define ECB_MEMORY_FENCE_RELAXED __atomic_thread_fence (__ATOMIC_RELAXED) |
170 | |
246 | |
171 | /* The __has_feature syntax from clang is so misdesigned that we cannot use it |
247 | #elif ECB_CLANG_EXTENSION(c_atomic) |
172 | * without risking compile time errors with other compilers. We *could* |
|
|
173 | * define our own ecb_clang_has_feature, but I just can't be bothered to work |
|
|
174 | * around this shit time and again. |
|
|
175 | * #elif defined __clang && __has_feature (cxx_atomic) |
|
|
176 | * // see comment below (stdatomic.h) about the C11 memory model. |
248 | /* see comment below (stdatomic.h) about the C11 memory model. */ |
177 | * #define ECB_MEMORY_FENCE __c11_atomic_thread_fence (__ATOMIC_SEQ_CST) |
249 | #define ECB_MEMORY_FENCE __c11_atomic_thread_fence (__ATOMIC_SEQ_CST) |
178 | */ |
250 | #define ECB_MEMORY_FENCE_ACQUIRE __c11_atomic_thread_fence (__ATOMIC_ACQUIRE) |
|
|
251 | #define ECB_MEMORY_FENCE_RELEASE __c11_atomic_thread_fence (__ATOMIC_RELEASE) |
|
|
252 | #define ECB_MEMORY_FENCE_RELAXED __c11_atomic_thread_fence (__ATOMIC_RELAXED) |
179 | |
253 | |
180 | #elif ECB_GCC_VERSION(4,4) || defined __INTEL_COMPILER || defined __clang__ |
254 | #elif ECB_GCC_VERSION(4,4) || defined __INTEL_COMPILER || defined __clang__ |
181 | #define ECB_MEMORY_FENCE __sync_synchronize () |
255 | #define ECB_MEMORY_FENCE __sync_synchronize () |
|
|
256 | #elif _MSC_VER >= 1500 /* VC++ 2008 */ |
|
|
257 | /* apparently, microsoft broke all the memory barrier stuff in Visual Studio 2008... */ |
|
|
258 | #pragma intrinsic(_ReadBarrier,_WriteBarrier,_ReadWriteBarrier) |
|
|
259 | #define ECB_MEMORY_FENCE _ReadWriteBarrier (); MemoryBarrier() |
|
|
260 | #define ECB_MEMORY_FENCE_ACQUIRE _ReadWriteBarrier (); MemoryBarrier() /* according to msdn, _ReadBarrier is not a load fence */ |
|
|
261 | #define ECB_MEMORY_FENCE_RELEASE _WriteBarrier (); MemoryBarrier() |
182 | #elif _MSC_VER >= 1400 /* VC++ 2005 */ |
262 | #elif _MSC_VER >= 1400 /* VC++ 2005 */ |
183 | #pragma intrinsic(_ReadBarrier,_WriteBarrier,_ReadWriteBarrier) |
263 | #pragma intrinsic(_ReadBarrier,_WriteBarrier,_ReadWriteBarrier) |
184 | #define ECB_MEMORY_FENCE _ReadWriteBarrier () |
264 | #define ECB_MEMORY_FENCE _ReadWriteBarrier () |
185 | #define ECB_MEMORY_FENCE_ACQUIRE _ReadWriteBarrier () /* according to msdn, _ReadBarrier is not a load fence */ |
265 | #define ECB_MEMORY_FENCE_ACQUIRE _ReadWriteBarrier () /* according to msdn, _ReadBarrier is not a load fence */ |
186 | #define ECB_MEMORY_FENCE_RELEASE _WriteBarrier () |
266 | #define ECB_MEMORY_FENCE_RELEASE _WriteBarrier () |
187 | #elif defined _WIN32 |
267 | #elif defined _WIN32 |
188 | #include <WinNT.h> |
268 | #include <WinNT.h> |
189 | #define ECB_MEMORY_FENCE MemoryBarrier () /* actually just xchg on x86... scary */ |
269 | #define ECB_MEMORY_FENCE MemoryBarrier () /* actually just xchg on x86... scary */ |
190 | #elif __SUNPRO_C >= 0x5110 || __SUNPRO_CC >= 0x5110 |
270 | #elif __SUNPRO_C >= 0x5110 || __SUNPRO_CC >= 0x5110 |
191 | #include <mbarrier.h> |
271 | #include <mbarrier.h> |
192 | #define ECB_MEMORY_FENCE __machine_rw_barrier () |
272 | #define ECB_MEMORY_FENCE __machine_rw_barrier () |
193 | #define ECB_MEMORY_FENCE_ACQUIRE __machine_r_barrier () |
273 | #define ECB_MEMORY_FENCE_ACQUIRE __machine_acq_barrier () |
194 | #define ECB_MEMORY_FENCE_RELEASE __machine_w_barrier () |
274 | #define ECB_MEMORY_FENCE_RELEASE __machine_rel_barrier () |
|
|
275 | #define ECB_MEMORY_FENCE_RELAXED __compiler_barrier () |
195 | #elif __xlC__ |
276 | #elif __xlC__ |
196 | #define ECB_MEMORY_FENCE __sync () |
277 | #define ECB_MEMORY_FENCE __sync () |
197 | #endif |
278 | #endif |
198 | #endif |
279 | #endif |
199 | |
280 | |
200 | #ifndef ECB_MEMORY_FENCE |
281 | #ifndef ECB_MEMORY_FENCE |
201 | #if ECB_C11 && !defined __STDC_NO_ATOMICS__ |
282 | #if ECB_C11 && !defined __STDC_NO_ATOMICS__ |
202 | /* we assume that these memory fences work on all variables/all memory accesses, */ |
283 | /* we assume that these memory fences work on all variables/all memory accesses, */ |
203 | /* not just C11 atomics and atomic accesses */ |
284 | /* not just C11 atomics and atomic accesses */ |
204 | #include <stdatomic.h> |
285 | #include <stdatomic.h> |
205 | /* Unfortunately, neither gcc 4.7 nor clang 3.1 generate any instructions for */ |
|
|
206 | /* any fence other than seq_cst, which isn't very efficient for us. */ |
|
|
207 | /* Why that is, we don't know - either the C11 memory model is quite useless */ |
|
|
208 | /* for most usages, or gcc and clang have a bug */ |
|
|
209 | /* I *currently* lean towards the latter, and inefficiently implement */ |
|
|
210 | /* all three of ecb's fences as a seq_cst fence */ |
|
|
211 | #define ECB_MEMORY_FENCE atomic_thread_fence (memory_order_seq_cst) |
286 | #define ECB_MEMORY_FENCE atomic_thread_fence (memory_order_seq_cst) |
|
|
287 | #define ECB_MEMORY_FENCE_ACQUIRE atomic_thread_fence (memory_order_acquire) |
|
|
288 | #define ECB_MEMORY_FENCE_RELEASE atomic_thread_fence (memory_order_release) |
212 | #endif |
289 | #endif |
213 | #endif |
290 | #endif |
214 | |
291 | |
215 | #ifndef ECB_MEMORY_FENCE |
292 | #ifndef ECB_MEMORY_FENCE |
216 | #if !ECB_AVOID_PTHREADS |
293 | #if !ECB_AVOID_PTHREADS |
… | |
… | |
236 | |
313 | |
237 | #if !defined ECB_MEMORY_FENCE_RELEASE && defined ECB_MEMORY_FENCE |
314 | #if !defined ECB_MEMORY_FENCE_RELEASE && defined ECB_MEMORY_FENCE |
238 | #define ECB_MEMORY_FENCE_RELEASE ECB_MEMORY_FENCE |
315 | #define ECB_MEMORY_FENCE_RELEASE ECB_MEMORY_FENCE |
239 | #endif |
316 | #endif |
240 | |
317 | |
|
|
318 | #if !defined ECB_MEMORY_FENCE_RELAXED && defined ECB_MEMORY_FENCE |
|
|
319 | #define ECB_MEMORY_FENCE_RELAXED ECB_MEMORY_FENCE /* very heavy-handed */ |
|
|
320 | #endif |
|
|
321 | |
241 | /*****************************************************************************/ |
322 | /*****************************************************************************/ |
242 | |
323 | |
243 | #if __cplusplus |
324 | #if ECB_CPP |
244 | #define ecb_inline static inline |
325 | #define ecb_inline static inline |
245 | #elif ECB_GCC_VERSION(2,5) |
326 | #elif ECB_GCC_VERSION(2,5) |
246 | #define ecb_inline static __inline__ |
327 | #define ecb_inline static __inline__ |
247 | #elif ECB_C99 |
328 | #elif ECB_C99 |
248 | #define ecb_inline static inline |
329 | #define ecb_inline static inline |
… | |
… | |
262 | |
343 | |
263 | #define ECB_CONCAT_(a, b) a ## b |
344 | #define ECB_CONCAT_(a, b) a ## b |
264 | #define ECB_CONCAT(a, b) ECB_CONCAT_(a, b) |
345 | #define ECB_CONCAT(a, b) ECB_CONCAT_(a, b) |
265 | #define ECB_STRINGIFY_(a) # a |
346 | #define ECB_STRINGIFY_(a) # a |
266 | #define ECB_STRINGIFY(a) ECB_STRINGIFY_(a) |
347 | #define ECB_STRINGIFY(a) ECB_STRINGIFY_(a) |
|
|
348 | #define ECB_STRINGIFY_EXPR(expr) ((expr), ECB_STRINGIFY_ (expr)) |
267 | |
349 | |
268 | #define ecb_function_ ecb_inline |
350 | #define ecb_function_ ecb_inline |
269 | |
351 | |
270 | #if ECB_GCC_VERSION(3,1) |
352 | #if ECB_GCC_VERSION(3,1) || ECB_CLANG_VERSION(2,8) |
271 | #define ecb_attribute(attrlist) __attribute__(attrlist) |
353 | #define ecb_attribute(attrlist) __attribute__ (attrlist) |
|
|
354 | #else |
|
|
355 | #define ecb_attribute(attrlist) |
|
|
356 | #endif |
|
|
357 | |
|
|
358 | #if ECB_GCC_VERSION(3,1) || ECB_CLANG_BUILTIN(__builtin_constant_p) |
272 | #define ecb_is_constant(expr) __builtin_constant_p (expr) |
359 | #define ecb_is_constant(expr) __builtin_constant_p (expr) |
|
|
360 | #else |
|
|
361 | /* possible C11 impl for integral types |
|
|
362 | typedef struct ecb_is_constant_struct ecb_is_constant_struct; |
|
|
363 | #define ecb_is_constant(expr) _Generic ((1 ? (struct ecb_is_constant_struct *)0 : (void *)((expr) - (expr)), ecb_is_constant_struct *: 0, default: 1)) */ |
|
|
364 | |
|
|
365 | #define ecb_is_constant(expr) 0 |
|
|
366 | #endif |
|
|
367 | |
|
|
368 | #if ECB_GCC_VERSION(3,1) || ECB_CLANG_BUILTIN(__builtin_expect) |
273 | #define ecb_expect(expr,value) __builtin_expect ((expr),(value)) |
369 | #define ecb_expect(expr,value) __builtin_expect ((expr),(value)) |
|
|
370 | #else |
|
|
371 | #define ecb_expect(expr,value) (expr) |
|
|
372 | #endif |
|
|
373 | |
|
|
374 | #if ECB_GCC_VERSION(3,1) || ECB_CLANG_BUILTIN(__builtin_prefetch) |
274 | #define ecb_prefetch(addr,rw,locality) __builtin_prefetch (addr, rw, locality) |
375 | #define ecb_prefetch(addr,rw,locality) __builtin_prefetch (addr, rw, locality) |
275 | #else |
376 | #else |
276 | #define ecb_attribute(attrlist) |
|
|
277 | #define ecb_is_constant(expr) 0 |
|
|
278 | #define ecb_expect(expr,value) (expr) |
|
|
279 | #define ecb_prefetch(addr,rw,locality) |
377 | #define ecb_prefetch(addr,rw,locality) |
280 | #endif |
378 | #endif |
281 | |
379 | |
282 | /* no emulation for ecb_decltype */ |
380 | /* no emulation for ecb_decltype */ |
283 | #if ECB_GCC_VERSION(4,5) |
381 | #if ECB_CPP11 |
|
|
382 | // older implementations might have problems with decltype(x)::type, work around it |
|
|
383 | template<class T> struct ecb_decltype_t { typedef T type; }; |
284 | #define ecb_decltype(x) __decltype(x) |
384 | #define ecb_decltype(x) ecb_decltype_t<decltype (x)>::type |
285 | #elif ECB_GCC_VERSION(3,0) |
385 | #elif ECB_GCC_VERSION(3,0) || ECB_CLANG_VERSION(2,8) |
286 | #define ecb_decltype(x) __typeof(x) |
386 | #define ecb_decltype(x) __typeof__ (x) |
287 | #endif |
387 | #endif |
288 | |
388 | |
|
|
389 | #if _MSC_VER >= 1300 |
|
|
390 | #define ecb_deprecated __declspec (deprecated) |
|
|
391 | #else |
|
|
392 | #define ecb_deprecated ecb_attribute ((__deprecated__)) |
|
|
393 | #endif |
|
|
394 | |
|
|
395 | #if _MSC_VER >= 1500 |
|
|
396 | #define ecb_deprecated_message(msg) __declspec (deprecated (msg)) |
|
|
397 | #elif ECB_GCC_VERSION(4,5) |
|
|
398 | #define ecb_deprecated_message(msg) ecb_attribute ((__deprecated__ (msg)) |
|
|
399 | #else |
|
|
400 | #define ecb_deprecated_message(msg) ecb_deprecated |
|
|
401 | #endif |
|
|
402 | |
|
|
403 | #if _MSC_VER >= 1400 |
|
|
404 | #define ecb_noinline __declspec (noinline) |
|
|
405 | #else |
289 | #define ecb_noinline ecb_attribute ((__noinline__)) |
406 | #define ecb_noinline ecb_attribute ((__noinline__)) |
|
|
407 | #endif |
|
|
408 | |
290 | #define ecb_unused ecb_attribute ((__unused__)) |
409 | #define ecb_unused ecb_attribute ((__unused__)) |
291 | #define ecb_const ecb_attribute ((__const__)) |
410 | #define ecb_const ecb_attribute ((__const__)) |
292 | #define ecb_pure ecb_attribute ((__pure__)) |
411 | #define ecb_pure ecb_attribute ((__pure__)) |
293 | |
412 | |
294 | #if ECB_C11 |
413 | #if ECB_C11 || __IBMC_NORETURN |
|
|
414 | /* http://www-01.ibm.com/support/knowledgecenter/SSGH3R_13.1.0/com.ibm.xlcpp131.aix.doc/language_ref/noreturn.html */ |
295 | #define ecb_noreturn _Noreturn |
415 | #define ecb_noreturn _Noreturn |
|
|
416 | #elif ECB_CPP11 |
|
|
417 | #define ecb_noreturn [[noreturn]] |
|
|
418 | #elif _MSC_VER >= 1200 |
|
|
419 | /* http://msdn.microsoft.com/en-us/library/k6ktzx3s.aspx */ |
|
|
420 | #define ecb_noreturn __declspec (noreturn) |
296 | #else |
421 | #else |
297 | #define ecb_noreturn ecb_attribute ((__noreturn__)) |
422 | #define ecb_noreturn ecb_attribute ((__noreturn__)) |
298 | #endif |
423 | #endif |
299 | |
424 | |
300 | #if ECB_GCC_VERSION(4,3) |
425 | #if ECB_GCC_VERSION(4,3) |
… | |
… | |
315 | /* for compatibility to the rest of the world */ |
440 | /* for compatibility to the rest of the world */ |
316 | #define ecb_likely(expr) ecb_expect_true (expr) |
441 | #define ecb_likely(expr) ecb_expect_true (expr) |
317 | #define ecb_unlikely(expr) ecb_expect_false (expr) |
442 | #define ecb_unlikely(expr) ecb_expect_false (expr) |
318 | |
443 | |
319 | /* count trailing zero bits and count # of one bits */ |
444 | /* count trailing zero bits and count # of one bits */ |
320 | #if ECB_GCC_VERSION(3,4) |
445 | #if ECB_GCC_VERSION(3,4) \ |
|
|
446 | || (ECB_CLANG_BUILTIN(__builtin_clz) && ECB_CLANG_BUILTIN(__builtin_clzll) \ |
|
|
447 | && ECB_CLANG_BUILTIN(__builtin_ctz) && ECB_CLANG_BUILTIN(__builtin_ctzll) \ |
|
|
448 | && ECB_CLANG_BUILTIN(__builtin_popcount)) |
321 | /* we assume int == 32 bit, long == 32 or 64 bit and long long == 64 bit */ |
449 | /* we assume int == 32 bit, long == 32 or 64 bit and long long == 64 bit */ |
322 | #define ecb_ld32(x) (__builtin_clz (x) ^ 31) |
450 | #define ecb_ld32(x) (__builtin_clz (x) ^ 31) |
323 | #define ecb_ld64(x) (__builtin_clzll (x) ^ 63) |
451 | #define ecb_ld64(x) (__builtin_clzll (x) ^ 63) |
324 | #define ecb_ctz32(x) __builtin_ctz (x) |
452 | #define ecb_ctz32(x) __builtin_ctz (x) |
325 | #define ecb_ctz64(x) __builtin_ctzll (x) |
453 | #define ecb_ctz64(x) __builtin_ctzll (x) |
326 | #define ecb_popcount32(x) __builtin_popcount (x) |
454 | #define ecb_popcount32(x) __builtin_popcount (x) |
327 | /* no popcountll */ |
455 | /* no popcountll */ |
328 | #else |
456 | #else |
329 | ecb_function_ int ecb_ctz32 (uint32_t x) ecb_const; |
457 | ecb_function_ ecb_const int ecb_ctz32 (uint32_t x); |
330 | ecb_function_ int |
458 | ecb_function_ ecb_const int |
331 | ecb_ctz32 (uint32_t x) |
459 | ecb_ctz32 (uint32_t x) |
332 | { |
460 | { |
|
|
461 | #if 1400 <= _MSC_VER && (_M_IX86 || _M_X64 || _M_IA64 || _M_ARM) |
|
|
462 | unsigned long r; |
|
|
463 | _BitScanForward (&r, x); |
|
|
464 | return (int)r; |
|
|
465 | #else |
333 | int r = 0; |
466 | int r = 0; |
334 | |
467 | |
335 | x &= ~x + 1; /* this isolates the lowest bit */ |
468 | x &= ~x + 1; /* this isolates the lowest bit */ |
336 | |
469 | |
337 | #if ECB_branchless_on_i386 |
470 | #if ECB_branchless_on_i386 |
… | |
… | |
347 | if (x & 0xff00ff00) r += 8; |
480 | if (x & 0xff00ff00) r += 8; |
348 | if (x & 0xffff0000) r += 16; |
481 | if (x & 0xffff0000) r += 16; |
349 | #endif |
482 | #endif |
350 | |
483 | |
351 | return r; |
484 | return r; |
|
|
485 | #endif |
352 | } |
486 | } |
353 | |
487 | |
354 | ecb_function_ int ecb_ctz64 (uint64_t x) ecb_const; |
488 | ecb_function_ ecb_const int ecb_ctz64 (uint64_t x); |
355 | ecb_function_ int |
489 | ecb_function_ ecb_const int |
356 | ecb_ctz64 (uint64_t x) |
490 | ecb_ctz64 (uint64_t x) |
357 | { |
491 | { |
|
|
492 | #if 1400 <= _MSC_VER && (_M_X64 || _M_IA64 || _M_ARM) |
|
|
493 | unsigned long r; |
|
|
494 | _BitScanForward64 (&r, x); |
|
|
495 | return (int)r; |
|
|
496 | #else |
358 | int shift = x & 0xffffffffU ? 0 : 32; |
497 | int shift = x & 0xffffffff ? 0 : 32; |
359 | return ecb_ctz32 (x >> shift) + shift; |
498 | return ecb_ctz32 (x >> shift) + shift; |
|
|
499 | #endif |
360 | } |
500 | } |
361 | |
501 | |
362 | ecb_function_ int ecb_popcount32 (uint32_t x) ecb_const; |
502 | ecb_function_ ecb_const int ecb_popcount32 (uint32_t x); |
363 | ecb_function_ int |
503 | ecb_function_ ecb_const int |
364 | ecb_popcount32 (uint32_t x) |
504 | ecb_popcount32 (uint32_t x) |
365 | { |
505 | { |
366 | x -= (x >> 1) & 0x55555555; |
506 | x -= (x >> 1) & 0x55555555; |
367 | x = ((x >> 2) & 0x33333333) + (x & 0x33333333); |
507 | x = ((x >> 2) & 0x33333333) + (x & 0x33333333); |
368 | x = ((x >> 4) + x) & 0x0f0f0f0f; |
508 | x = ((x >> 4) + x) & 0x0f0f0f0f; |
369 | x *= 0x01010101; |
509 | x *= 0x01010101; |
370 | |
510 | |
371 | return x >> 24; |
511 | return x >> 24; |
372 | } |
512 | } |
373 | |
513 | |
374 | ecb_function_ int ecb_ld32 (uint32_t x) ecb_const; |
514 | ecb_function_ ecb_const int ecb_ld32 (uint32_t x); |
375 | ecb_function_ int ecb_ld32 (uint32_t x) |
515 | ecb_function_ ecb_const int ecb_ld32 (uint32_t x) |
376 | { |
516 | { |
|
|
517 | #if 1400 <= _MSC_VER && (_M_IX86 || _M_X64 || _M_IA64 || _M_ARM) |
|
|
518 | unsigned long r; |
|
|
519 | _BitScanReverse (&r, x); |
|
|
520 | return (int)r; |
|
|
521 | #else |
377 | int r = 0; |
522 | int r = 0; |
378 | |
523 | |
379 | if (x >> 16) { x >>= 16; r += 16; } |
524 | if (x >> 16) { x >>= 16; r += 16; } |
380 | if (x >> 8) { x >>= 8; r += 8; } |
525 | if (x >> 8) { x >>= 8; r += 8; } |
381 | if (x >> 4) { x >>= 4; r += 4; } |
526 | if (x >> 4) { x >>= 4; r += 4; } |
382 | if (x >> 2) { x >>= 2; r += 2; } |
527 | if (x >> 2) { x >>= 2; r += 2; } |
383 | if (x >> 1) { r += 1; } |
528 | if (x >> 1) { r += 1; } |
384 | |
529 | |
385 | return r; |
530 | return r; |
|
|
531 | #endif |
386 | } |
532 | } |
387 | |
533 | |
388 | ecb_function_ int ecb_ld64 (uint64_t x) ecb_const; |
534 | ecb_function_ ecb_const int ecb_ld64 (uint64_t x); |
389 | ecb_function_ int ecb_ld64 (uint64_t x) |
535 | ecb_function_ ecb_const int ecb_ld64 (uint64_t x) |
390 | { |
536 | { |
|
|
537 | #if 1400 <= _MSC_VER && (_M_X64 || _M_IA64 || _M_ARM) |
|
|
538 | unsigned long r; |
|
|
539 | _BitScanReverse64 (&r, x); |
|
|
540 | return (int)r; |
|
|
541 | #else |
391 | int r = 0; |
542 | int r = 0; |
392 | |
543 | |
393 | if (x >> 32) { x >>= 32; r += 32; } |
544 | if (x >> 32) { x >>= 32; r += 32; } |
394 | |
545 | |
395 | return r + ecb_ld32 (x); |
546 | return r + ecb_ld32 (x); |
396 | } |
|
|
397 | #endif |
547 | #endif |
|
|
548 | } |
|
|
549 | #endif |
398 | |
550 | |
399 | ecb_function_ ecb_bool ecb_is_pot32 (uint32_t x) ecb_const; |
551 | ecb_function_ ecb_const ecb_bool ecb_is_pot32 (uint32_t x); |
400 | ecb_function_ ecb_bool ecb_is_pot32 (uint32_t x) { return !(x & (x - 1)); } |
552 | ecb_function_ ecb_const ecb_bool ecb_is_pot32 (uint32_t x) { return !(x & (x - 1)); } |
401 | ecb_function_ ecb_bool ecb_is_pot64 (uint64_t x) ecb_const; |
553 | ecb_function_ ecb_const ecb_bool ecb_is_pot64 (uint64_t x); |
402 | ecb_function_ ecb_bool ecb_is_pot64 (uint64_t x) { return !(x & (x - 1)); } |
554 | ecb_function_ ecb_const ecb_bool ecb_is_pot64 (uint64_t x) { return !(x & (x - 1)); } |
403 | |
555 | |
404 | ecb_function_ uint8_t ecb_bitrev8 (uint8_t x) ecb_const; |
556 | ecb_function_ ecb_const uint8_t ecb_bitrev8 (uint8_t x); |
405 | ecb_function_ uint8_t ecb_bitrev8 (uint8_t x) |
557 | ecb_function_ ecb_const uint8_t ecb_bitrev8 (uint8_t x) |
406 | { |
558 | { |
407 | return ( (x * 0x0802U & 0x22110U) |
559 | return ( (x * 0x0802U & 0x22110U) |
408 | | (x * 0x8020U & 0x88440U)) * 0x10101U >> 16; |
560 | | (x * 0x8020U & 0x88440U)) * 0x10101U >> 16; |
409 | } |
561 | } |
410 | |
562 | |
411 | ecb_function_ uint16_t ecb_bitrev16 (uint16_t x) ecb_const; |
563 | ecb_function_ ecb_const uint16_t ecb_bitrev16 (uint16_t x); |
412 | ecb_function_ uint16_t ecb_bitrev16 (uint16_t x) |
564 | ecb_function_ ecb_const uint16_t ecb_bitrev16 (uint16_t x) |
413 | { |
565 | { |
414 | x = ((x >> 1) & 0x5555) | ((x & 0x5555) << 1); |
566 | x = ((x >> 1) & 0x5555) | ((x & 0x5555) << 1); |
415 | x = ((x >> 2) & 0x3333) | ((x & 0x3333) << 2); |
567 | x = ((x >> 2) & 0x3333) | ((x & 0x3333) << 2); |
416 | x = ((x >> 4) & 0x0f0f) | ((x & 0x0f0f) << 4); |
568 | x = ((x >> 4) & 0x0f0f) | ((x & 0x0f0f) << 4); |
417 | x = ( x >> 8 ) | ( x << 8); |
569 | x = ( x >> 8 ) | ( x << 8); |
418 | |
570 | |
419 | return x; |
571 | return x; |
420 | } |
572 | } |
421 | |
573 | |
422 | ecb_function_ uint32_t ecb_bitrev32 (uint32_t x) ecb_const; |
574 | ecb_function_ ecb_const uint32_t ecb_bitrev32 (uint32_t x); |
423 | ecb_function_ uint32_t ecb_bitrev32 (uint32_t x) |
575 | ecb_function_ ecb_const uint32_t ecb_bitrev32 (uint32_t x) |
424 | { |
576 | { |
425 | x = ((x >> 1) & 0x55555555) | ((x & 0x55555555) << 1); |
577 | x = ((x >> 1) & 0x55555555) | ((x & 0x55555555) << 1); |
426 | x = ((x >> 2) & 0x33333333) | ((x & 0x33333333) << 2); |
578 | x = ((x >> 2) & 0x33333333) | ((x & 0x33333333) << 2); |
427 | x = ((x >> 4) & 0x0f0f0f0f) | ((x & 0x0f0f0f0f) << 4); |
579 | x = ((x >> 4) & 0x0f0f0f0f) | ((x & 0x0f0f0f0f) << 4); |
428 | x = ((x >> 8) & 0x00ff00ff) | ((x & 0x00ff00ff) << 8); |
580 | x = ((x >> 8) & 0x00ff00ff) | ((x & 0x00ff00ff) << 8); |
… | |
… | |
431 | return x; |
583 | return x; |
432 | } |
584 | } |
433 | |
585 | |
434 | /* popcount64 is only available on 64 bit cpus as gcc builtin */ |
586 | /* popcount64 is only available on 64 bit cpus as gcc builtin */ |
435 | /* so for this version we are lazy */ |
587 | /* so for this version we are lazy */ |
436 | ecb_function_ int ecb_popcount64 (uint64_t x) ecb_const; |
588 | ecb_function_ ecb_const int ecb_popcount64 (uint64_t x); |
437 | ecb_function_ int |
589 | ecb_function_ ecb_const int |
438 | ecb_popcount64 (uint64_t x) |
590 | ecb_popcount64 (uint64_t x) |
439 | { |
591 | { |
440 | return ecb_popcount32 (x) + ecb_popcount32 (x >> 32); |
592 | return ecb_popcount32 (x) + ecb_popcount32 (x >> 32); |
441 | } |
593 | } |
442 | |
594 | |
443 | ecb_inline uint8_t ecb_rotl8 (uint8_t x, unsigned int count) ecb_const; |
595 | ecb_inline ecb_const uint8_t ecb_rotl8 (uint8_t x, unsigned int count); |
444 | ecb_inline uint8_t ecb_rotr8 (uint8_t x, unsigned int count) ecb_const; |
596 | ecb_inline ecb_const uint8_t ecb_rotr8 (uint8_t x, unsigned int count); |
445 | ecb_inline uint16_t ecb_rotl16 (uint16_t x, unsigned int count) ecb_const; |
597 | ecb_inline ecb_const uint16_t ecb_rotl16 (uint16_t x, unsigned int count); |
446 | ecb_inline uint16_t ecb_rotr16 (uint16_t x, unsigned int count) ecb_const; |
598 | ecb_inline ecb_const uint16_t ecb_rotr16 (uint16_t x, unsigned int count); |
447 | ecb_inline uint32_t ecb_rotl32 (uint32_t x, unsigned int count) ecb_const; |
599 | ecb_inline ecb_const uint32_t ecb_rotl32 (uint32_t x, unsigned int count); |
448 | ecb_inline uint32_t ecb_rotr32 (uint32_t x, unsigned int count) ecb_const; |
600 | ecb_inline ecb_const uint32_t ecb_rotr32 (uint32_t x, unsigned int count); |
449 | ecb_inline uint64_t ecb_rotl64 (uint64_t x, unsigned int count) ecb_const; |
601 | ecb_inline ecb_const uint64_t ecb_rotl64 (uint64_t x, unsigned int count); |
450 | ecb_inline uint64_t ecb_rotr64 (uint64_t x, unsigned int count) ecb_const; |
602 | ecb_inline ecb_const uint64_t ecb_rotr64 (uint64_t x, unsigned int count); |
451 | |
603 | |
452 | ecb_inline uint8_t ecb_rotl8 (uint8_t x, unsigned int count) { return (x >> ( 8 - count)) | (x << count); } |
604 | ecb_inline ecb_const uint8_t ecb_rotl8 (uint8_t x, unsigned int count) { return (x >> ( 8 - count)) | (x << count); } |
453 | ecb_inline uint8_t ecb_rotr8 (uint8_t x, unsigned int count) { return (x << ( 8 - count)) | (x >> count); } |
605 | ecb_inline ecb_const uint8_t ecb_rotr8 (uint8_t x, unsigned int count) { return (x << ( 8 - count)) | (x >> count); } |
454 | ecb_inline uint16_t ecb_rotl16 (uint16_t x, unsigned int count) { return (x >> (16 - count)) | (x << count); } |
606 | ecb_inline ecb_const uint16_t ecb_rotl16 (uint16_t x, unsigned int count) { return (x >> (16 - count)) | (x << count); } |
455 | ecb_inline uint16_t ecb_rotr16 (uint16_t x, unsigned int count) { return (x << (16 - count)) | (x >> count); } |
607 | ecb_inline ecb_const uint16_t ecb_rotr16 (uint16_t x, unsigned int count) { return (x << (16 - count)) | (x >> count); } |
456 | ecb_inline uint32_t ecb_rotl32 (uint32_t x, unsigned int count) { return (x >> (32 - count)) | (x << count); } |
608 | ecb_inline ecb_const uint32_t ecb_rotl32 (uint32_t x, unsigned int count) { return (x >> (32 - count)) | (x << count); } |
457 | ecb_inline uint32_t ecb_rotr32 (uint32_t x, unsigned int count) { return (x << (32 - count)) | (x >> count); } |
609 | ecb_inline ecb_const uint32_t ecb_rotr32 (uint32_t x, unsigned int count) { return (x << (32 - count)) | (x >> count); } |
458 | ecb_inline uint64_t ecb_rotl64 (uint64_t x, unsigned int count) { return (x >> (64 - count)) | (x << count); } |
610 | ecb_inline ecb_const uint64_t ecb_rotl64 (uint64_t x, unsigned int count) { return (x >> (64 - count)) | (x << count); } |
459 | ecb_inline uint64_t ecb_rotr64 (uint64_t x, unsigned int count) { return (x << (64 - count)) | (x >> count); } |
611 | ecb_inline ecb_const uint64_t ecb_rotr64 (uint64_t x, unsigned int count) { return (x << (64 - count)) | (x >> count); } |
460 | |
612 | |
461 | #if ECB_GCC_VERSION(4,3) |
613 | #if ECB_GCC_VERSION(4,3) || (ECB_CLANG_BUILTIN(__builtin_bswap32) && ECB_CLANG_BUILTIN(__builtin_bswap64)) |
|
|
614 | #if ECB_GCC_VERSION(4,8) || ECB_CLANG_BUILTIN(__builtin_bswap16) |
|
|
615 | #define ecb_bswap16(x) __builtin_bswap16 (x) |
|
|
616 | #else |
462 | #define ecb_bswap16(x) (__builtin_bswap32 (x) >> 16) |
617 | #define ecb_bswap16(x) (__builtin_bswap32 (x) >> 16) |
|
|
618 | #endif |
463 | #define ecb_bswap32(x) __builtin_bswap32 (x) |
619 | #define ecb_bswap32(x) __builtin_bswap32 (x) |
464 | #define ecb_bswap64(x) __builtin_bswap64 (x) |
620 | #define ecb_bswap64(x) __builtin_bswap64 (x) |
|
|
621 | #elif _MSC_VER |
|
|
622 | #include <stdlib.h> |
|
|
623 | #define ecb_bswap16(x) ((uint16_t)_byteswap_ushort ((uint16_t)(x))) |
|
|
624 | #define ecb_bswap32(x) ((uint32_t)_byteswap_ulong ((uint32_t)(x))) |
|
|
625 | #define ecb_bswap64(x) ((uint64_t)_byteswap_uint64 ((uint64_t)(x))) |
465 | #else |
626 | #else |
466 | ecb_function_ uint16_t ecb_bswap16 (uint16_t x) ecb_const; |
627 | ecb_function_ ecb_const uint16_t ecb_bswap16 (uint16_t x); |
467 | ecb_function_ uint16_t |
628 | ecb_function_ ecb_const uint16_t |
468 | ecb_bswap16 (uint16_t x) |
629 | ecb_bswap16 (uint16_t x) |
469 | { |
630 | { |
470 | return ecb_rotl16 (x, 8); |
631 | return ecb_rotl16 (x, 8); |
471 | } |
632 | } |
472 | |
633 | |
473 | ecb_function_ uint32_t ecb_bswap32 (uint32_t x) ecb_const; |
634 | ecb_function_ ecb_const uint32_t ecb_bswap32 (uint32_t x); |
474 | ecb_function_ uint32_t |
635 | ecb_function_ ecb_const uint32_t |
475 | ecb_bswap32 (uint32_t x) |
636 | ecb_bswap32 (uint32_t x) |
476 | { |
637 | { |
477 | return (((uint32_t)ecb_bswap16 (x)) << 16) | ecb_bswap16 (x >> 16); |
638 | return (((uint32_t)ecb_bswap16 (x)) << 16) | ecb_bswap16 (x >> 16); |
478 | } |
639 | } |
479 | |
640 | |
480 | ecb_function_ uint64_t ecb_bswap64 (uint64_t x) ecb_const; |
641 | ecb_function_ ecb_const uint64_t ecb_bswap64 (uint64_t x); |
481 | ecb_function_ uint64_t |
642 | ecb_function_ ecb_const uint64_t |
482 | ecb_bswap64 (uint64_t x) |
643 | ecb_bswap64 (uint64_t x) |
483 | { |
644 | { |
484 | return (((uint64_t)ecb_bswap32 (x)) << 32) | ecb_bswap32 (x >> 32); |
645 | return (((uint64_t)ecb_bswap32 (x)) << 32) | ecb_bswap32 (x >> 32); |
485 | } |
646 | } |
486 | #endif |
647 | #endif |
487 | |
648 | |
488 | #if ECB_GCC_VERSION(4,5) |
649 | #if ECB_GCC_VERSION(4,5) || ECB_CLANG_BUILTIN(__builtin_unreachable) |
489 | #define ecb_unreachable() __builtin_unreachable () |
650 | #define ecb_unreachable() __builtin_unreachable () |
490 | #else |
651 | #else |
491 | /* this seems to work fine, but gcc always emits a warning for it :/ */ |
652 | /* this seems to work fine, but gcc always emits a warning for it :/ */ |
492 | ecb_inline void ecb_unreachable (void) ecb_noreturn; |
653 | ecb_inline ecb_noreturn void ecb_unreachable (void); |
493 | ecb_inline void ecb_unreachable (void) { } |
654 | ecb_inline ecb_noreturn void ecb_unreachable (void) { } |
494 | #endif |
655 | #endif |
495 | |
656 | |
496 | /* try to tell the compiler that some condition is definitely true */ |
657 | /* try to tell the compiler that some condition is definitely true */ |
497 | #define ecb_assume(cond) if (!(cond)) ecb_unreachable (); else 0 |
658 | #define ecb_assume(cond) if (!(cond)) ecb_unreachable (); else 0 |
498 | |
659 | |
499 | ecb_inline unsigned char ecb_byteorder_helper (void) ecb_const; |
660 | ecb_inline ecb_const uint32_t ecb_byteorder_helper (void); |
500 | ecb_inline unsigned char |
661 | ecb_inline ecb_const uint32_t |
501 | ecb_byteorder_helper (void) |
662 | ecb_byteorder_helper (void) |
502 | { |
663 | { |
503 | /* the union code still generates code under pressure in gcc, */ |
664 | /* the union code still generates code under pressure in gcc, */ |
504 | /* but less than using pointers, and always seems to */ |
665 | /* but less than using pointers, and always seems to */ |
505 | /* successfully return a constant. */ |
666 | /* successfully return a constant. */ |
506 | /* the reason why we have this horrible preprocessor mess */ |
667 | /* the reason why we have this horrible preprocessor mess */ |
507 | /* is to avoid it in all cases, at least on common architectures */ |
668 | /* is to avoid it in all cases, at least on common architectures */ |
508 | /* or when using a recent enough gcc version (>= 4.6) */ |
669 | /* or when using a recent enough gcc version (>= 4.6) */ |
509 | #if __i386 || __i386__ || _M_X86 || __amd64 || __amd64__ || _M_X64 |
|
|
510 | return 0x44; |
|
|
511 | #elif __BYTE_ORDER__ && __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__ |
670 | #if (defined __BYTE_ORDER__ && __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__) \ |
|
|
671 | || ((__i386 || __i386__ || _M_IX86 || ECB_GCC_AMD64 || ECB_MSVC_AMD64) && !__VOS__) |
|
|
672 | #define ECB_LITTLE_ENDIAN 1 |
512 | return 0x44; |
673 | return 0x44332211; |
513 | #elif __BYTE_ORDER__ && __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__ |
674 | #elif (defined __BYTE_ORDER__ && __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__) \ |
|
|
675 | || ((__AARCH64EB__ || __MIPSEB__ || __ARMEB__) && !__VOS__) |
|
|
676 | #define ECB_BIG_ENDIAN 1 |
514 | return 0x11; |
677 | return 0x11223344; |
515 | #else |
678 | #else |
516 | union |
679 | union |
517 | { |
680 | { |
|
|
681 | uint8_t c[4]; |
518 | uint32_t i; |
682 | uint32_t u; |
519 | uint8_t c; |
|
|
520 | } u = { 0x11223344 }; |
683 | } u = { 0x11, 0x22, 0x33, 0x44 }; |
521 | return u.c; |
684 | return u.u; |
522 | #endif |
685 | #endif |
523 | } |
686 | } |
524 | |
687 | |
525 | ecb_inline ecb_bool ecb_big_endian (void) ecb_const; |
688 | ecb_inline ecb_const ecb_bool ecb_big_endian (void); |
526 | ecb_inline ecb_bool ecb_big_endian (void) { return ecb_byteorder_helper () == 0x11; } |
689 | ecb_inline ecb_const ecb_bool ecb_big_endian (void) { return ecb_byteorder_helper () == 0x11223344; } |
527 | ecb_inline ecb_bool ecb_little_endian (void) ecb_const; |
690 | ecb_inline ecb_const ecb_bool ecb_little_endian (void); |
528 | ecb_inline ecb_bool ecb_little_endian (void) { return ecb_byteorder_helper () == 0x44; } |
691 | ecb_inline ecb_const ecb_bool ecb_little_endian (void) { return ecb_byteorder_helper () == 0x44332211; } |
|
|
692 | |
|
|
693 | /*****************************************************************************/ |
|
|
694 | /* unaligned load/store */ |
|
|
695 | |
|
|
696 | ecb_inline uint_fast16_t ecb_be_u16_to_host (uint_fast16_t v) { return ecb_little_endian () ? ecb_bswap16 (v) : v; } |
|
|
697 | ecb_inline uint_fast32_t ecb_be_u32_to_host (uint_fast32_t v) { return ecb_little_endian () ? ecb_bswap32 (v) : v; } |
|
|
698 | ecb_inline uint_fast64_t ecb_be_u64_to_host (uint_fast64_t v) { return ecb_little_endian () ? ecb_bswap64 (v) : v; } |
|
|
699 | |
|
|
700 | ecb_inline uint_fast16_t ecb_le_u16_to_host (uint_fast16_t v) { return ecb_big_endian () ? ecb_bswap16 (v) : v; } |
|
|
701 | ecb_inline uint_fast32_t ecb_le_u32_to_host (uint_fast32_t v) { return ecb_big_endian () ? ecb_bswap32 (v) : v; } |
|
|
702 | ecb_inline uint_fast64_t ecb_le_u64_to_host (uint_fast64_t v) { return ecb_big_endian () ? ecb_bswap64 (v) : v; } |
|
|
703 | |
|
|
704 | ecb_inline uint_fast16_t ecb_peek_u16_u (const void *ptr) { uint16_t v; memcpy (&v, ptr, sizeof (v)); return v; } |
|
|
705 | ecb_inline uint_fast32_t ecb_peek_u32_u (const void *ptr) { uint32_t v; memcpy (&v, ptr, sizeof (v)); return v; } |
|
|
706 | ecb_inline uint_fast64_t ecb_peek_u64_u (const void *ptr) { uint64_t v; memcpy (&v, ptr, sizeof (v)); return v; } |
|
|
707 | |
|
|
708 | ecb_inline uint_fast16_t ecb_peek_be_u16_u (const void *ptr) { return ecb_be_u16_to_host (ecb_peek_u16_u (ptr)); } |
|
|
709 | ecb_inline uint_fast32_t ecb_peek_be_u32_u (const void *ptr) { return ecb_be_u32_to_host (ecb_peek_u32_u (ptr)); } |
|
|
710 | ecb_inline uint_fast64_t ecb_peek_be_u64_u (const void *ptr) { return ecb_be_u64_to_host (ecb_peek_u64_u (ptr)); } |
|
|
711 | |
|
|
712 | ecb_inline uint_fast16_t ecb_peek_le_u16_u (const void *ptr) { return ecb_le_u16_to_host (ecb_peek_u16_u (ptr)); } |
|
|
713 | ecb_inline uint_fast32_t ecb_peek_le_u32_u (const void *ptr) { return ecb_le_u32_to_host (ecb_peek_u32_u (ptr)); } |
|
|
714 | ecb_inline uint_fast64_t ecb_peek_le_u64_u (const void *ptr) { return ecb_le_u64_to_host (ecb_peek_u64_u (ptr)); } |
|
|
715 | |
|
|
716 | ecb_inline uint_fast16_t ecb_host_to_be_u16 (uint_fast16_t v) { return ecb_little_endian () ? ecb_bswap16 (v) : v; } |
|
|
717 | ecb_inline uint_fast32_t ecb_host_to_be_u32 (uint_fast32_t v) { return ecb_little_endian () ? ecb_bswap32 (v) : v; } |
|
|
718 | ecb_inline uint_fast64_t ecb_host_to_be_u64 (uint_fast64_t v) { return ecb_little_endian () ? ecb_bswap64 (v) : v; } |
|
|
719 | |
|
|
720 | ecb_inline uint_fast16_t ecb_host_to_le_u16 (uint_fast16_t v) { return ecb_big_endian () ? ecb_bswap16 (v) : v; } |
|
|
721 | ecb_inline uint_fast32_t ecb_host_to_le_u32 (uint_fast32_t v) { return ecb_big_endian () ? ecb_bswap32 (v) : v; } |
|
|
722 | ecb_inline uint_fast64_t ecb_host_to_le_u64 (uint_fast64_t v) { return ecb_big_endian () ? ecb_bswap64 (v) : v; } |
|
|
723 | |
|
|
724 | ecb_inline void ecb_poke_u16_u (void *ptr, uint16_t v) { memcpy (ptr, &v, sizeof (v)); } |
|
|
725 | ecb_inline void ecb_poke_u32_u (void *ptr, uint32_t v) { memcpy (ptr, &v, sizeof (v)); } |
|
|
726 | ecb_inline void ecb_poke_u64_u (void *ptr, uint64_t v) { memcpy (ptr, &v, sizeof (v)); } |
|
|
727 | |
|
|
728 | 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)); } |
|
|
729 | 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)); } |
|
|
730 | 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)); } |
|
|
731 | |
|
|
732 | 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)); } |
|
|
733 | 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)); } |
|
|
734 | 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)); } |
|
|
735 | |
|
|
736 | #ifdef __cplusplus |
|
|
737 | |
|
|
738 | inline uint8_t ecb_bswap (uint8_t v) { return v; } |
|
|
739 | inline uint16_t ecb_bswap (uint16_t v) { return ecb_bswap16 (v); } |
|
|
740 | inline uint32_t ecb_bswap (uint32_t v) { return ecb_bswap32 (v); } |
|
|
741 | inline uint64_t ecb_bswap (uint64_t v) { return ecb_bswap64 (v); } |
|
|
742 | |
|
|
743 | template<typename T> inline T ecb_be_to_host (T v) { return ecb_little_endian () ? ecb_bswap (v) : v; } |
|
|
744 | template<typename T> inline T ecb_le_to_host (T v) { return ecb_big_endian () ? ecb_bswap (v) : v; } |
|
|
745 | template<typename T> inline T ecb_peek (const void *ptr) { return *(const T *)ptr; } |
|
|
746 | template<typename T> inline T ecb_peek_be (const void *ptr) { return ecb_be_to_host (ecb_peek <T> (ptr)); } |
|
|
747 | template<typename T> inline T ecb_peek_le (const void *ptr) { return ecb_le_to_host (ecb_peek <T> (ptr)); } |
|
|
748 | template<typename T> inline T ecb_peek_u (const void *ptr) { T v; std::memcpy (&v, ptr, sizeof (v)); return v; } |
|
|
749 | template<typename T> inline T ecb_peek_be_u (const void *ptr) { return ecb_be_to_host (ecb_peek_u<T> (ptr)); } |
|
|
750 | template<typename T> inline T ecb_peek_le_u (const void *ptr) { return ecb_le_to_host (ecb_peek_u<T> (ptr)); } |
|
|
751 | |
|
|
752 | template<typename T> inline T ecb_host_to_be (T v) { return ecb_little_endian () ? ecb_bswap (v) : v; } |
|
|
753 | template<typename T> inline T ecb_host_to_le (T v) { return ecb_big_endian () ? ecb_bswap (v) : v; } |
|
|
754 | template<typename T> inline void ecb_poke (void *ptr, T v) { *(T *)ptr = v; } |
|
|
755 | template<typename T> inline void ecb_poke_be (void *ptr, T v) { return ecb_poke <T> (ptr, ecb_host_to_be (v)); } |
|
|
756 | template<typename T> inline void ecb_poke_le (void *ptr, T v) { return ecb_poke <T> (ptr, ecb_host_to_le (v)); } |
|
|
757 | template<typename T> inline void ecb_poke_u (void *ptr, T v) { std::memcpy (ptr, &v, sizeof (v)); } |
|
|
758 | template<typename T> inline void ecb_poke_be_u (void *ptr, T v) { return ecb_poke_u<T> (ptr, ecb_host_to_be (v)); } |
|
|
759 | template<typename T> inline void ecb_poke_le_u (void *ptr, T v) { return ecb_poke_u<T> (ptr, ecb_host_to_le (v)); } |
|
|
760 | |
|
|
761 | #endif |
|
|
762 | |
|
|
763 | /*****************************************************************************/ |
529 | |
764 | |
530 | #if ECB_GCC_VERSION(3,0) || ECB_C99 |
765 | #if ECB_GCC_VERSION(3,0) || ECB_C99 |
531 | #define ecb_mod(m,n) ((m) % (n) + ((m) % (n) < 0 ? (n) : 0)) |
766 | #define ecb_mod(m,n) ((m) % (n) + ((m) % (n) < 0 ? (n) : 0)) |
532 | #else |
767 | #else |
533 | #define ecb_mod(m,n) ((m) < 0 ? ((n) - 1 - ((-1 - (m)) % (n))) : ((m) % (n))) |
768 | #define ecb_mod(m,n) ((m) < 0 ? ((n) - 1 - ((-1 - (m)) % (n))) : ((m) % (n))) |
534 | #endif |
769 | #endif |
535 | |
770 | |
536 | #if __cplusplus |
771 | #if ECB_CPP |
537 | template<typename T> |
772 | template<typename T> |
538 | static inline T ecb_div_rd (T val, T div) |
773 | static inline T ecb_div_rd (T val, T div) |
539 | { |
774 | { |
540 | return val < 0 ? - ((-val + div - 1) / div) : (val ) / div; |
775 | return val < 0 ? - ((-val + div - 1) / div) : (val ) / div; |
541 | } |
776 | } |
… | |
… | |
558 | } |
793 | } |
559 | #else |
794 | #else |
560 | #define ecb_array_length(name) (sizeof (name) / sizeof (name [0])) |
795 | #define ecb_array_length(name) (sizeof (name) / sizeof (name [0])) |
561 | #endif |
796 | #endif |
562 | |
797 | |
|
|
798 | /*****************************************************************************/ |
|
|
799 | |
|
|
800 | ecb_function_ ecb_const uint32_t ecb_binary16_to_binary32 (uint32_t x); |
|
|
801 | ecb_function_ ecb_const uint32_t |
|
|
802 | ecb_binary16_to_binary32 (uint32_t x) |
|
|
803 | { |
|
|
804 | unsigned int s = (x & 0x8000) << (31 - 15); |
|
|
805 | int e = (x >> 10) & 0x001f; |
|
|
806 | unsigned int m = x & 0x03ff; |
|
|
807 | |
|
|
808 | if (ecb_expect_false (e == 31)) |
|
|
809 | /* infinity or NaN */ |
|
|
810 | e = 255 - (127 - 15); |
|
|
811 | else if (ecb_expect_false (!e)) |
|
|
812 | { |
|
|
813 | if (ecb_expect_true (!m)) |
|
|
814 | /* zero, handled by code below by forcing e to 0 */ |
|
|
815 | e = 0 - (127 - 15); |
|
|
816 | else |
|
|
817 | { |
|
|
818 | /* subnormal, renormalise */ |
|
|
819 | unsigned int s = 10 - ecb_ld32 (m); |
|
|
820 | |
|
|
821 | m = (m << s) & 0x3ff; /* mask implicit bit */ |
|
|
822 | e -= s - 1; |
|
|
823 | } |
|
|
824 | } |
|
|
825 | |
|
|
826 | /* e and m now are normalised, or zero, (or inf or nan) */ |
|
|
827 | e += 127 - 15; |
|
|
828 | |
|
|
829 | return s | (e << 23) | (m << (23 - 10)); |
|
|
830 | } |
|
|
831 | |
|
|
832 | ecb_function_ ecb_const uint16_t ecb_binary32_to_binary16 (uint32_t x); |
|
|
833 | ecb_function_ ecb_const uint16_t |
|
|
834 | ecb_binary32_to_binary16 (uint32_t x) |
|
|
835 | { |
|
|
836 | unsigned int s = (x >> 16) & 0x00008000; /* sign bit, the easy part */ |
|
|
837 | unsigned int e = ((x >> 23) & 0x000000ff) - (127 - 15); /* the desired exponent */ |
|
|
838 | unsigned int m = x & 0x007fffff; |
|
|
839 | |
|
|
840 | x &= 0x7fffffff; |
|
|
841 | |
|
|
842 | /* if it's within range of binary16 normals, use fast path */ |
|
|
843 | if (ecb_expect_true (0x38800000 <= x && x <= 0x477fefff)) |
|
|
844 | { |
|
|
845 | /* mantissa round-to-even */ |
|
|
846 | m += 0x00000fff + ((m >> (23 - 10)) & 1); |
|
|
847 | |
|
|
848 | /* handle overflow */ |
|
|
849 | if (ecb_expect_false (m >= 0x00800000)) |
|
|
850 | { |
|
|
851 | m >>= 1; |
|
|
852 | e += 1; |
|
|
853 | } |
|
|
854 | |
|
|
855 | return s | (e << 10) | (m >> (23 - 10)); |
|
|
856 | } |
|
|
857 | |
|
|
858 | /* handle large numbers and infinity */ |
|
|
859 | if (ecb_expect_true (0x477fefff < x && x <= 0x7f800000)) |
|
|
860 | return s | 0x7c00; |
|
|
861 | |
|
|
862 | /* handle zero, subnormals and small numbers */ |
|
|
863 | if (ecb_expect_true (x < 0x38800000)) |
|
|
864 | { |
|
|
865 | /* zero */ |
|
|
866 | if (ecb_expect_true (!x)) |
|
|
867 | return s; |
|
|
868 | |
|
|
869 | /* handle subnormals */ |
|
|
870 | |
|
|
871 | /* too small, will be zero */ |
|
|
872 | if (e < (14 - 24)) /* might not be sharp, but is good enough */ |
|
|
873 | return s; |
|
|
874 | |
|
|
875 | m |= 0x00800000; /* make implicit bit explicit */ |
|
|
876 | |
|
|
877 | /* very tricky - we need to round to the nearest e (+10) bit value */ |
|
|
878 | { |
|
|
879 | unsigned int bits = 14 - e; |
|
|
880 | unsigned int half = (1 << (bits - 1)) - 1; |
|
|
881 | unsigned int even = (m >> bits) & 1; |
|
|
882 | |
|
|
883 | /* if this overflows, we will end up with a normalised number */ |
|
|
884 | m = (m + half + even) >> bits; |
|
|
885 | } |
|
|
886 | |
|
|
887 | return s | m; |
|
|
888 | } |
|
|
889 | |
|
|
890 | /* handle NaNs, preserve leftmost nan bits, but make sure we don't turn them into infinities */ |
|
|
891 | m >>= 13; |
|
|
892 | |
|
|
893 | return s | 0x7c00 | m | !m; |
|
|
894 | } |
|
|
895 | |
563 | /*******************************************************************************/ |
896 | /*******************************************************************************/ |
564 | /* floating point stuff, can be disabled by defining ECB_NO_LIBM */ |
897 | /* floating point stuff, can be disabled by defining ECB_NO_LIBM */ |
565 | |
898 | |
566 | /* basically, everything uses "ieee pure-endian" floating point numbers */ |
899 | /* basically, everything uses "ieee pure-endian" floating point numbers */ |
567 | /* the only noteworthy exception is ancient armle, which uses order 43218765 */ |
900 | /* the only noteworthy exception is ancient armle, which uses order 43218765 */ |
568 | #if 0 \ |
901 | #if 0 \ |
569 | || __i386 || __i386__ \ |
902 | || __i386 || __i386__ \ |
570 | || __amd64 || __amd64__ || __x86_64 || __x86_64__ \ |
903 | || ECB_GCC_AMD64 \ |
571 | || __powerpc__ || __ppc__ || __powerpc64__ || __ppc64__ \ |
904 | || __powerpc__ || __ppc__ || __powerpc64__ || __ppc64__ \ |
572 | || defined __arm__ && defined __ARM_EABI__ \ |
|
|
573 | || defined __s390__ || defined __s390x__ \ |
905 | || defined __s390__ || defined __s390x__ \ |
574 | || defined __mips__ \ |
906 | || defined __mips__ \ |
575 | || defined __alpha__ \ |
907 | || defined __alpha__ \ |
576 | || defined __hppa__ \ |
908 | || defined __hppa__ \ |
577 | || defined __ia64__ \ |
909 | || defined __ia64__ \ |
578 | || defined __m68k__ \ |
910 | || defined __m68k__ \ |
579 | || defined __m88k__ \ |
911 | || defined __m88k__ \ |
580 | || defined __sh__ \ |
912 | || defined __sh__ \ |
581 | || defined _M_IX86 || defined _M_AMD64 || defined _M_IA64 |
913 | || defined _M_IX86 || defined ECB_MSVC_AMD64 || defined _M_IA64 \ |
|
|
914 | || (defined __arm__ && (defined __ARM_EABI__ || defined __EABI__ || defined __VFP_FP__ || defined _WIN32_WCE || defined __ANDROID__)) \ |
|
|
915 | || defined __aarch64__ |
582 | #define ECB_STDFP 1 |
916 | #define ECB_STDFP 1 |
583 | #include <string.h> /* for memcpy */ |
917 | #include <string.h> /* for memcpy */ |
584 | #else |
918 | #else |
585 | #define ECB_STDFP 0 |
919 | #define ECB_STDFP 0 |
586 | #endif |
920 | #endif |
… | |
… | |
600 | #define ECB_NAN NAN |
934 | #define ECB_NAN NAN |
601 | #else |
935 | #else |
602 | #define ECB_NAN ECB_INFINITY |
936 | #define ECB_NAN ECB_INFINITY |
603 | #endif |
937 | #endif |
604 | |
938 | |
605 | /* converts an ieee half/binary16 to a float */ |
939 | #if ECB_C99 || _XOPEN_VERSION >= 600 || _POSIX_VERSION >= 200112L |
606 | ecb_function_ float ecb_binary16_to_float (uint16_t x) ecb_const; |
940 | #define ecb_ldexpf(x,e) ldexpf ((x), (e)) |
607 | ecb_function_ float |
941 | #define ecb_frexpf(x,e) frexpf ((x), (e)) |
608 | ecb_binary16_to_float (uint16_t x) |
942 | #else |
609 | { |
943 | #define ecb_ldexpf(x,e) (float) ldexp ((double) (x), (e)) |
610 | int e = (x >> 10) & 0x1f; |
944 | #define ecb_frexpf(x,e) (float) frexp ((double) (x), (e)) |
611 | int m = x & 0x3ff; |
945 | #endif |
612 | float r; |
|
|
613 | |
|
|
614 | if (!e ) r = ldexpf (m , -24); |
|
|
615 | else if (e != 31) r = ldexpf (m + 0x400, e - 25); |
|
|
616 | else if (m ) r = ECB_NAN; |
|
|
617 | else r = ECB_INFINITY; |
|
|
618 | |
|
|
619 | return x & 0x8000 ? -r : r; |
|
|
620 | } |
|
|
621 | |
946 | |
622 | /* convert a float to ieee single/binary32 */ |
947 | /* convert a float to ieee single/binary32 */ |
623 | ecb_function_ uint32_t ecb_float_to_binary32 (float x) ecb_const; |
948 | ecb_function_ ecb_const uint32_t ecb_float_to_binary32 (float x); |
624 | ecb_function_ uint32_t |
949 | ecb_function_ ecb_const uint32_t |
625 | ecb_float_to_binary32 (float x) |
950 | ecb_float_to_binary32 (float x) |
626 | { |
951 | { |
627 | uint32_t r; |
952 | uint32_t r; |
628 | |
953 | |
629 | #if ECB_STDFP |
954 | #if ECB_STDFP |
… | |
… | |
636 | if (x == 0e0f ) return 0x00000000U; |
961 | if (x == 0e0f ) return 0x00000000U; |
637 | if (x > +3.40282346638528860e+38f) return 0x7f800000U; |
962 | if (x > +3.40282346638528860e+38f) return 0x7f800000U; |
638 | if (x < -3.40282346638528860e+38f) return 0xff800000U; |
963 | if (x < -3.40282346638528860e+38f) return 0xff800000U; |
639 | if (x != x ) return 0x7fbfffffU; |
964 | if (x != x ) return 0x7fbfffffU; |
640 | |
965 | |
641 | m = frexpf (x, &e) * 0x1000000U; |
966 | m = ecb_frexpf (x, &e) * 0x1000000U; |
642 | |
967 | |
643 | r = m & 0x80000000U; |
968 | r = m & 0x80000000U; |
644 | |
969 | |
645 | if (r) |
970 | if (r) |
646 | m = -m; |
971 | m = -m; |
… | |
… | |
658 | |
983 | |
659 | return r; |
984 | return r; |
660 | } |
985 | } |
661 | |
986 | |
662 | /* converts an ieee single/binary32 to a float */ |
987 | /* converts an ieee single/binary32 to a float */ |
663 | ecb_function_ float ecb_binary32_to_float (uint32_t x) ecb_const; |
988 | ecb_function_ ecb_const float ecb_binary32_to_float (uint32_t x); |
664 | ecb_function_ float |
989 | ecb_function_ ecb_const float |
665 | ecb_binary32_to_float (uint32_t x) |
990 | ecb_binary32_to_float (uint32_t x) |
666 | { |
991 | { |
667 | float r; |
992 | float r; |
668 | |
993 | |
669 | #if ECB_STDFP |
994 | #if ECB_STDFP |
… | |
… | |
679 | x |= 0x800000U; |
1004 | x |= 0x800000U; |
680 | else |
1005 | else |
681 | e = 1; |
1006 | e = 1; |
682 | |
1007 | |
683 | /* we distrust ldexpf a bit and do the 2**-24 scaling by an extra multiply */ |
1008 | /* we distrust ldexpf a bit and do the 2**-24 scaling by an extra multiply */ |
684 | r = ldexpf (x * (0.5f / 0x800000U), e - 126); |
1009 | r = ecb_ldexpf (x * (0.5f / 0x800000U), e - 126); |
685 | |
1010 | |
686 | r = neg ? -r : r; |
1011 | r = neg ? -r : r; |
687 | #endif |
1012 | #endif |
688 | |
1013 | |
689 | return r; |
1014 | return r; |
690 | } |
1015 | } |
691 | |
1016 | |
692 | /* convert a double to ieee double/binary64 */ |
1017 | /* convert a double to ieee double/binary64 */ |
693 | ecb_function_ uint64_t ecb_double_to_binary64 (double x) ecb_const; |
1018 | ecb_function_ ecb_const uint64_t ecb_double_to_binary64 (double x); |
694 | ecb_function_ uint64_t |
1019 | ecb_function_ ecb_const uint64_t |
695 | ecb_double_to_binary64 (double x) |
1020 | ecb_double_to_binary64 (double x) |
696 | { |
1021 | { |
697 | uint64_t r; |
1022 | uint64_t r; |
698 | |
1023 | |
699 | #if ECB_STDFP |
1024 | #if ECB_STDFP |
… | |
… | |
728 | |
1053 | |
729 | return r; |
1054 | return r; |
730 | } |
1055 | } |
731 | |
1056 | |
732 | /* converts an ieee double/binary64 to a double */ |
1057 | /* converts an ieee double/binary64 to a double */ |
733 | ecb_function_ double ecb_binary64_to_double (uint64_t x) ecb_const; |
1058 | ecb_function_ ecb_const double ecb_binary64_to_double (uint64_t x); |
734 | ecb_function_ double |
1059 | ecb_function_ ecb_const double |
735 | ecb_binary64_to_double (uint64_t x) |
1060 | ecb_binary64_to_double (uint64_t x) |
736 | { |
1061 | { |
737 | double r; |
1062 | double r; |
738 | |
1063 | |
739 | #if ECB_STDFP |
1064 | #if ECB_STDFP |
… | |
… | |
757 | #endif |
1082 | #endif |
758 | |
1083 | |
759 | return r; |
1084 | return r; |
760 | } |
1085 | } |
761 | |
1086 | |
762 | #endif |
1087 | /* convert a float to ieee half/binary16 */ |
|
|
1088 | ecb_function_ ecb_const uint16_t ecb_float_to_binary16 (float x); |
|
|
1089 | ecb_function_ ecb_const uint16_t |
|
|
1090 | ecb_float_to_binary16 (float x) |
|
|
1091 | { |
|
|
1092 | return ecb_binary32_to_binary16 (ecb_float_to_binary32 (x)); |
|
|
1093 | } |
763 | |
1094 | |
764 | #endif |
1095 | /* convert an ieee half/binary16 to float */ |
|
|
1096 | ecb_function_ ecb_const float ecb_binary16_to_float (uint16_t x); |
|
|
1097 | ecb_function_ ecb_const float |
|
|
1098 | ecb_binary16_to_float (uint16_t x) |
|
|
1099 | { |
|
|
1100 | return ecb_binary32_to_float (ecb_binary16_to_binary32 (x)); |
|
|
1101 | } |
765 | |
1102 | |
|
|
1103 | #endif |
|
|
1104 | |
|
|
1105 | #endif |
|
|
1106 | |