1 | /* |
1 | /* |
2 | * libecb - http://software.schmorp.de/pkg/libecb |
2 | * libecb - http://software.schmorp.de/pkg/libecb |
3 | * |
3 | * |
4 | * Copyright (©) 2009-2011 Marc Alexander Lehmann <libecb@schmorp.de> |
4 | * Copyright (©) 2009-2012 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: |
… | |
… | |
28 | */ |
28 | */ |
29 | |
29 | |
30 | #ifndef ECB_H |
30 | #ifndef ECB_H |
31 | #define ECB_H |
31 | #define ECB_H |
32 | |
32 | |
|
|
33 | /* 16 bits major, 16 bits minor */ |
|
|
34 | #define ECB_VERSION 0x00010001 |
|
|
35 | |
33 | #ifdef _WIN32 |
36 | #ifdef _WIN32 |
34 | typedef signed char int8_t; |
37 | typedef signed char int8_t; |
35 | typedef unsigned char uint8_t; |
38 | typedef unsigned char uint8_t; |
36 | typedef signed short int16_t; |
39 | typedef signed short int16_t; |
37 | typedef unsigned short uint16_t; |
40 | typedef unsigned short uint16_t; |
38 | typedef signed int int32_t; |
41 | typedef signed int int32_t; |
39 | typedef unsigned int uint32_t; |
42 | typedef unsigned int uint32_t; |
40 | #if __GNUC__ |
43 | #if __GNUC__ |
41 | typedef signed long long int64_t; |
44 | typedef signed long long int64_t; |
42 | typedef unsigned long long uint64_t; |
45 | typedef unsigned long long uint64_t; |
43 | #else |
46 | #else /* _MSC_VER || __BORLANDC__ */ |
44 | typedef signed __int64 int64_t; |
47 | typedef signed __int64 int64_t; |
45 | typedef unsigned __int64 uint64_t; |
48 | typedef unsigned __int64 uint64_t; |
46 | #endif |
49 | #endif |
|
|
50 | #ifdef _WIN64 |
|
|
51 | #define ECB_PTRSIZE 8 |
|
|
52 | typedef uint64_t uintptr_t; |
|
|
53 | typedef int64_t intptr_t; |
|
|
54 | #else |
|
|
55 | #define ECB_PTRSIZE 4 |
|
|
56 | typedef uint32_t uintptr_t; |
|
|
57 | typedef int32_t intptr_t; |
|
|
58 | #endif |
47 | #else |
59 | #else |
48 | #include <inttypes.h> |
60 | #include <inttypes.h> |
|
|
61 | #if UINTMAX_MAX > 0xffffffffU |
|
|
62 | #define ECB_PTRSIZE 8 |
|
|
63 | #else |
|
|
64 | #define ECB_PTRSIZE 4 |
|
|
65 | #endif |
49 | #endif |
66 | #endif |
50 | |
67 | |
51 | /* many compilers define _GNUC_ to some versions but then only implement |
68 | /* many compilers define _GNUC_ to some versions but then only implement |
52 | * what their idiot authors think are the "more important" extensions, |
69 | * what their idiot authors think are the "more important" extensions, |
53 | * causing enourmous grief in return for some better fake benchmark numbers. |
70 | * causing enormous grief in return for some better fake benchmark numbers. |
54 | * or so. |
71 | * or so. |
55 | * we try to detect these and simply assume they are not gcc - if they have |
72 | * we try to detect these and simply assume they are not gcc - if they have |
56 | * an issue with that they should have done it right in the first place. |
73 | * an issue with that they should have done it right in the first place. |
57 | */ |
74 | */ |
58 | #ifndef ECB_GCC_VERSION |
75 | #ifndef ECB_GCC_VERSION |
59 | #if !defined(__GNUC_MINOR__) || defined(__INTEL_COMPILER) || defined(__SUNPRO_C) || defined(__SUNPRO_CC) || defined(__llvm__) || defined(__clang__) |
76 | #if !defined __GNUC_MINOR__ || defined __INTEL_COMPILER || defined __SUNPRO_C || defined __SUNPRO_CC || defined __llvm__ || defined __clang__ |
60 | #define ECB_GCC_VERSION(major,minor) 0 |
77 | #define ECB_GCC_VERSION(major,minor) 0 |
61 | #else |
78 | #else |
62 | #define ECB_GCC_VERSION(major,minor) (__GNUC__ > (major) || (__GNUC__ == (major) && __GNUC_MINOR__ >= (minor))) |
79 | #define ECB_GCC_VERSION(major,minor) (__GNUC__ > (major) || (__GNUC__ == (major) && __GNUC_MINOR__ >= (minor))) |
63 | #endif |
80 | #endif |
64 | #endif |
81 | #endif |
65 | |
82 | |
|
|
83 | #define ECB_C (__STDC__+0) /* this assumes that __STDC__ is either empty or a number */ |
66 | #define ECB_C99 (__STDC_VERSION__ >= 199901L) |
84 | #define ECB_C99 (__STDC_VERSION__ >= 199901L) |
|
|
85 | #define ECB_C11 (__STDC_VERSION__ >= 201112L) |
|
|
86 | #define ECB_CPP (__cplusplus+0) |
|
|
87 | #define ECB_CPP11 (__cplusplus >= 201103L) |
|
|
88 | |
|
|
89 | #if ECB_CPP |
|
|
90 | #define ECB_EXTERN_C extern "C" |
|
|
91 | #define ECB_EXTERN_C_BEG ECB_EXTERN_C { |
|
|
92 | #define ECB_EXTERN_C_END } |
|
|
93 | #else |
|
|
94 | #define ECB_EXTERN_C extern |
|
|
95 | #define ECB_EXTERN_C_BEG |
|
|
96 | #define ECB_EXTERN_C_END |
|
|
97 | #endif |
|
|
98 | |
|
|
99 | /*****************************************************************************/ |
|
|
100 | |
|
|
101 | /* ECB_NO_THREADS - ecb is not used by multiple threads, ever */ |
|
|
102 | /* ECB_NO_SMP - ecb might be used in multiple threads, but only on a single cpu */ |
|
|
103 | |
|
|
104 | #if ECB_NO_THREADS |
|
|
105 | #define ECB_NO_SMP 1 |
|
|
106 | #endif |
|
|
107 | |
|
|
108 | #if ECB_NO_SMP |
|
|
109 | #define ECB_MEMORY_FENCE do { } while (0) |
|
|
110 | #endif |
|
|
111 | |
|
|
112 | #ifndef ECB_MEMORY_FENCE |
|
|
113 | #if ECB_GCC_VERSION(2,5) || defined __INTEL_COMPILER || (__llvm__ && __GNUC__) || __SUNPRO_C >= 0x5110 || __SUNPRO_CC >= 0x5110 |
|
|
114 | #if __i386 || __i386__ |
|
|
115 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("lock; orb $0, -1(%%esp)" : : : "memory") |
|
|
116 | #define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("" : : : "memory") |
|
|
117 | #define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("") |
|
|
118 | #elif __amd64 || __amd64__ || __x86_64 || __x86_64__ |
|
|
119 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("mfence" : : : "memory") |
|
|
120 | #define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("" : : : "memory") |
|
|
121 | #define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("") |
|
|
122 | #elif __powerpc__ || __ppc__ || __powerpc64__ || __ppc64__ |
|
|
123 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("sync" : : : "memory") |
|
|
124 | #elif defined __ARM_ARCH_6__ || defined __ARM_ARCH_6J__ \ |
|
|
125 | || defined __ARM_ARCH_6K__ || defined __ARM_ARCH_6ZK__ |
|
|
126 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("mcr p15,0,%0,c7,c10,5" : : "r" (0) : "memory") |
|
|
127 | #elif defined __ARM_ARCH_7__ || defined __ARM_ARCH_7A__ \ |
|
|
128 | || defined __ARM_ARCH_7M__ || defined __ARM_ARCH_7R__ |
|
|
129 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("dmb" : : : "memory") |
|
|
130 | #elif __sparc || __sparc__ |
|
|
131 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("membar #LoadStore | #LoadLoad | #StoreStore | #StoreLoad" : : : "memory") |
|
|
132 | #define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("membar #LoadStore | #LoadLoad" : : : "memory") |
|
|
133 | #define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("membar #LoadStore | #StoreStore") |
|
|
134 | #elif defined __s390__ || defined __s390x__ |
|
|
135 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("bcr 15,0" : : : "memory") |
|
|
136 | #elif defined __mips__ |
|
|
137 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("sync" : : : "memory") |
|
|
138 | #elif defined __alpha__ |
|
|
139 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("mb" : : : "memory") |
|
|
140 | #elif defined __hppa__ |
|
|
141 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("" : : : "memory") |
|
|
142 | #define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("") |
|
|
143 | #elif defined __ia64__ |
|
|
144 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("mf" : : : "memory") |
|
|
145 | #endif |
|
|
146 | #endif |
|
|
147 | #endif |
|
|
148 | |
|
|
149 | #ifndef ECB_MEMORY_FENCE |
|
|
150 | #if ECB_GCC_VERSION(4,7) |
|
|
151 | /* see comment below (stdatomic.h) about the C11 memory model. */ |
|
|
152 | #define ECB_MEMORY_FENCE __atomic_thread_fence (__ATOMIC_SEQ_CST) |
|
|
153 | /*#elif defined __clang && __has_feature (cxx_atomic)*/ |
|
|
154 | /* see comment below (stdatomic.h) about the C11 memory model. */ |
|
|
155 | /*#define ECB_MEMORY_FENCE __c11_atomic_thread_fence (__ATOMIC_SEQ_CST)*/ |
|
|
156 | #elif ECB_GCC_VERSION(4,4) || defined __INTEL_COMPILER || defined __clang__ |
|
|
157 | #define ECB_MEMORY_FENCE __sync_synchronize () |
|
|
158 | #elif _MSC_VER >= 1400 /* VC++ 2005 */ |
|
|
159 | #pragma intrinsic(_ReadBarrier,_WriteBarrier,_ReadWriteBarrier) |
|
|
160 | #define ECB_MEMORY_FENCE _ReadWriteBarrier () |
|
|
161 | #define ECB_MEMORY_FENCE_ACQUIRE _ReadWriteBarrier () /* according to msdn, _ReadBarrier is not a load fence */ |
|
|
162 | #define ECB_MEMORY_FENCE_RELEASE _WriteBarrier () |
|
|
163 | #elif defined _WIN32 |
|
|
164 | #include <WinNT.h> |
|
|
165 | #define ECB_MEMORY_FENCE MemoryBarrier () /* actually just xchg on x86... scary */ |
|
|
166 | #elif __SUNPRO_C >= 0x5110 || __SUNPRO_CC >= 0x5110 |
|
|
167 | #include <mbarrier.h> |
|
|
168 | #define ECB_MEMORY_FENCE __machine_rw_barrier () |
|
|
169 | #define ECB_MEMORY_FENCE_ACQUIRE __machine_r_barrier () |
|
|
170 | #define ECB_MEMORY_FENCE_RELEASE __machine_w_barrier () |
|
|
171 | #elif __xlC__ |
|
|
172 | #define ECB_MEMORY_FENCE __sync () |
|
|
173 | #endif |
|
|
174 | #endif |
|
|
175 | |
|
|
176 | #ifndef ECB_MEMORY_FENCE |
|
|
177 | #if ECB_C11 && !defined __STDC_NO_ATOMICS__ |
|
|
178 | /* we assume that these memory fences work on all variables/all memory accesses, */ |
|
|
179 | /* not just C11 atomics and atomic accesses */ |
|
|
180 | #include <stdatomic.h> |
|
|
181 | /* Unfortunately, neither gcc 4.7 nor clang 3.1 generate any instructions for */ |
|
|
182 | /* any fence other than seq_cst, which isn't very efficient for us. */ |
|
|
183 | /* Why that is, we don't know - either the C11 memory model is quite useless */ |
|
|
184 | /* for most usages, or gcc and clang have a bug */ |
|
|
185 | /* I *currently* lean towards the latter, and inefficiently implement */ |
|
|
186 | /* all three of ecb's fences as a seq_cst fence */ |
|
|
187 | #define ECB_MEMORY_FENCE atomic_thread_fence (memory_order_seq_cst) |
|
|
188 | #endif |
|
|
189 | #endif |
|
|
190 | |
|
|
191 | #ifndef ECB_MEMORY_FENCE |
|
|
192 | #if !ECB_AVOID_PTHREADS |
|
|
193 | /* |
|
|
194 | * if you get undefined symbol references to pthread_mutex_lock, |
|
|
195 | * or failure to find pthread.h, then you should implement |
|
|
196 | * the ECB_MEMORY_FENCE operations for your cpu/compiler |
|
|
197 | * OR provide pthread.h and link against the posix thread library |
|
|
198 | * of your system. |
|
|
199 | */ |
|
|
200 | #include <pthread.h> |
|
|
201 | #define ECB_NEEDS_PTHREADS 1 |
|
|
202 | #define ECB_MEMORY_FENCE_NEEDS_PTHREADS 1 |
|
|
203 | |
|
|
204 | static pthread_mutex_t ecb_mf_lock = PTHREAD_MUTEX_INITIALIZER; |
|
|
205 | #define ECB_MEMORY_FENCE do { pthread_mutex_lock (&ecb_mf_lock); pthread_mutex_unlock (&ecb_mf_lock); } while (0) |
|
|
206 | #endif |
|
|
207 | #endif |
|
|
208 | |
|
|
209 | #if !defined ECB_MEMORY_FENCE_ACQUIRE && defined ECB_MEMORY_FENCE |
|
|
210 | #define ECB_MEMORY_FENCE_ACQUIRE ECB_MEMORY_FENCE |
|
|
211 | #endif |
|
|
212 | |
|
|
213 | #if !defined ECB_MEMORY_FENCE_RELEASE && defined ECB_MEMORY_FENCE |
|
|
214 | #define ECB_MEMORY_FENCE_RELEASE ECB_MEMORY_FENCE |
|
|
215 | #endif |
|
|
216 | |
|
|
217 | /*****************************************************************************/ |
67 | |
218 | |
68 | #if __cplusplus |
219 | #if __cplusplus |
69 | #define ecb_inline static inline |
220 | #define ecb_inline static inline |
70 | #elif ECB_GCC_VERSION(2,5) |
221 | #elif ECB_GCC_VERSION(2,5) |
71 | #define ecb_inline static __inline__ |
222 | #define ecb_inline static __inline__ |
… | |
… | |
110 | #elif ECB_GCC_VERSION(3,0) |
261 | #elif ECB_GCC_VERSION(3,0) |
111 | #define ecb_decltype(x) __typeof(x) |
262 | #define ecb_decltype(x) __typeof(x) |
112 | #endif |
263 | #endif |
113 | |
264 | |
114 | #define ecb_noinline ecb_attribute ((__noinline__)) |
265 | #define ecb_noinline ecb_attribute ((__noinline__)) |
115 | #define ecb_noreturn ecb_attribute ((__noreturn__)) |
|
|
116 | #define ecb_unused ecb_attribute ((__unused__)) |
266 | #define ecb_unused ecb_attribute ((__unused__)) |
117 | #define ecb_const ecb_attribute ((__const__)) |
267 | #define ecb_const ecb_attribute ((__const__)) |
118 | #define ecb_pure ecb_attribute ((__pure__)) |
268 | #define ecb_pure ecb_attribute ((__pure__)) |
|
|
269 | |
|
|
270 | #if ECB_C11 |
|
|
271 | #define ecb_noreturn _Noreturn |
|
|
272 | #else |
|
|
273 | #define ecb_noreturn ecb_attribute ((__noreturn__)) |
|
|
274 | #endif |
119 | |
275 | |
120 | #if ECB_GCC_VERSION(4,3) |
276 | #if ECB_GCC_VERSION(4,3) |
121 | #define ecb_artificial ecb_attribute ((__artificial__)) |
277 | #define ecb_artificial ecb_attribute ((__artificial__)) |
122 | #define ecb_hot ecb_attribute ((__hot__)) |
278 | #define ecb_hot ecb_attribute ((__hot__)) |
123 | #define ecb_cold ecb_attribute ((__cold__)) |
279 | #define ecb_cold ecb_attribute ((__cold__)) |
… | |
… | |
152 | { |
308 | { |
153 | int r = 0; |
309 | int r = 0; |
154 | |
310 | |
155 | x &= ~x + 1; /* this isolates the lowest bit */ |
311 | x &= ~x + 1; /* this isolates the lowest bit */ |
156 | |
312 | |
|
|
313 | #if ECB_branchless_on_i386 |
|
|
314 | r += !!(x & 0xaaaaaaaa) << 0; |
|
|
315 | r += !!(x & 0xcccccccc) << 1; |
|
|
316 | r += !!(x & 0xf0f0f0f0) << 2; |
|
|
317 | r += !!(x & 0xff00ff00) << 3; |
|
|
318 | r += !!(x & 0xffff0000) << 4; |
|
|
319 | #else |
157 | if (x & 0xaaaaaaaa) r += 1; |
320 | if (x & 0xaaaaaaaa) r += 1; |
158 | if (x & 0xcccccccc) r += 2; |
321 | if (x & 0xcccccccc) r += 2; |
159 | if (x & 0xf0f0f0f0) r += 4; |
322 | if (x & 0xf0f0f0f0) r += 4; |
160 | if (x & 0xff00ff00) r += 8; |
323 | if (x & 0xff00ff00) r += 8; |
161 | if (x & 0xffff0000) r += 16; |
324 | if (x & 0xffff0000) r += 16; |
|
|
325 | #endif |
162 | |
326 | |
163 | return r; |
327 | return r; |
164 | } |
328 | } |
165 | |
329 | |
166 | ecb_function_ int ecb_ctz64 (uint64_t x) ecb_const; |
330 | ecb_function_ int ecb_ctz64 (uint64_t x) ecb_const; |
167 | ecb_function_ int |
331 | ecb_function_ int |
168 | ecb_ctz64 (uint64_t x) |
332 | ecb_ctz64 (uint64_t x) |
169 | { |
333 | { |
170 | int shift = x & 0xffffffffU ? 0 : 32; |
334 | int shift = x & 0xffffffffU ? 0 : 32; |
171 | return ecb_ctz (x >> shift) + shift; |
335 | return ecb_ctz32 (x >> shift) + shift; |
172 | } |
336 | } |
173 | |
337 | |
174 | ecb_function_ int ecb_popcount32 (uint32_t x) ecb_const; |
338 | ecb_function_ int ecb_popcount32 (uint32_t x) ecb_const; |
175 | ecb_function_ int |
339 | ecb_function_ int |
176 | ecb_popcount32 (uint32_t x) |
340 | ecb_popcount32 (uint32_t x) |
… | |
… | |
181 | x *= 0x01010101; |
345 | x *= 0x01010101; |
182 | |
346 | |
183 | return x >> 24; |
347 | return x >> 24; |
184 | } |
348 | } |
185 | |
349 | |
186 | /* you have the choice beetween something with a table lookup, */ |
|
|
187 | /* something using lots of bit arithmetic and a simple loop */ |
|
|
188 | /* we went for the loop */ |
|
|
189 | ecb_function_ int ecb_ld32 (uint32_t x) ecb_const; |
350 | ecb_function_ int ecb_ld32 (uint32_t x) ecb_const; |
190 | ecb_function_ int ecb_ld32 (uint32_t x) |
351 | ecb_function_ int ecb_ld32 (uint32_t x) |
191 | { |
352 | { |
192 | int r = -1; |
353 | int r = 0; |
193 | |
354 | |
194 | do |
355 | if (x >> 16) { x >>= 16; r += 16; } |
195 | { |
356 | if (x >> 8) { x >>= 8; r += 8; } |
196 | x >>= 1; |
357 | if (x >> 4) { x >>= 4; r += 4; } |
197 | ++r; |
358 | if (x >> 2) { x >>= 2; r += 2; } |
198 | } |
359 | if (x >> 1) { r += 1; } |
199 | while (x); |
|
|
200 | |
360 | |
201 | return r; |
361 | return r; |
202 | } |
362 | } |
203 | |
363 | |
204 | ecb_function_ int ecb_ld64 (uint64_t x) ecb_const; |
364 | ecb_function_ int ecb_ld64 (uint64_t x) ecb_const; |
205 | ecb_function_ int ecb_ld64 (uint64_t x) |
365 | ecb_function_ int ecb_ld64 (uint64_t x) |
206 | { |
366 | { |
207 | int r = -1; |
367 | int r = 0; |
208 | |
368 | |
209 | do |
369 | if (x >> 32) { x >>= 32; r += 32; } |
210 | { |
|
|
211 | x >>= 1; |
|
|
212 | ++r; |
|
|
213 | } |
|
|
214 | while (x); |
|
|
215 | |
370 | |
|
|
371 | return r + ecb_ld32 (x); |
|
|
372 | } |
|
|
373 | #endif |
|
|
374 | |
|
|
375 | ecb_function_ ecb_bool ecb_is_pot32 (uint32_t x) ecb_const; |
|
|
376 | ecb_function_ ecb_bool ecb_is_pot32 (uint32_t x) { return !(x & (x - 1)); } |
|
|
377 | ecb_function_ ecb_bool ecb_is_pot64 (uint64_t x) ecb_const; |
|
|
378 | ecb_function_ ecb_bool ecb_is_pot64 (uint64_t x) { return !(x & (x - 1)); } |
|
|
379 | |
|
|
380 | ecb_function_ uint8_t ecb_bitrev8 (uint8_t x) ecb_const; |
|
|
381 | ecb_function_ uint8_t ecb_bitrev8 (uint8_t x) |
|
|
382 | { |
|
|
383 | return ( (x * 0x0802U & 0x22110U) |
|
|
384 | | (x * 0x8020U & 0x88440U)) * 0x10101U >> 16; |
|
|
385 | } |
|
|
386 | |
|
|
387 | ecb_function_ uint16_t ecb_bitrev16 (uint16_t x) ecb_const; |
|
|
388 | ecb_function_ uint16_t ecb_bitrev16 (uint16_t x) |
|
|
389 | { |
|
|
390 | x = ((x >> 1) & 0x5555) | ((x & 0x5555) << 1); |
|
|
391 | x = ((x >> 2) & 0x3333) | ((x & 0x3333) << 2); |
|
|
392 | x = ((x >> 4) & 0x0f0f) | ((x & 0x0f0f) << 4); |
|
|
393 | x = ( x >> 8 ) | ( x << 8); |
|
|
394 | |
216 | return r; |
395 | return x; |
217 | } |
396 | } |
218 | #endif |
397 | |
|
|
398 | ecb_function_ uint32_t ecb_bitrev32 (uint32_t x) ecb_const; |
|
|
399 | ecb_function_ uint32_t ecb_bitrev32 (uint32_t x) |
|
|
400 | { |
|
|
401 | x = ((x >> 1) & 0x55555555) | ((x & 0x55555555) << 1); |
|
|
402 | x = ((x >> 2) & 0x33333333) | ((x & 0x33333333) << 2); |
|
|
403 | x = ((x >> 4) & 0x0f0f0f0f) | ((x & 0x0f0f0f0f) << 4); |
|
|
404 | x = ((x >> 8) & 0x00ff00ff) | ((x & 0x00ff00ff) << 8); |
|
|
405 | x = ( x >> 16 ) | ( x << 16); |
|
|
406 | |
|
|
407 | return x; |
|
|
408 | } |
219 | |
409 | |
220 | /* popcount64 is only available on 64 bit cpus as gcc builtin */ |
410 | /* popcount64 is only available on 64 bit cpus as gcc builtin */ |
221 | /* so for this version we are lazy */ |
411 | /* so for this version we are lazy */ |
222 | ecb_function_ int ecb_popcount64 (uint64_t x) ecb_const; |
412 | ecb_function_ int ecb_popcount64 (uint64_t x) ecb_const; |
223 | ecb_function_ int |
413 | ecb_function_ int |
224 | ecb_popcount64 (uint64_t x) |
414 | ecb_popcount64 (uint64_t x) |
225 | { |
415 | { |
226 | return ecb_popcount32 (x) + ecb_popcount32 (x >> 32); |
416 | return ecb_popcount32 (x) + ecb_popcount32 (x >> 32); |
227 | } |
417 | } |
228 | |
418 | |
|
|
419 | ecb_inline uint8_t ecb_rotl8 (uint8_t x, unsigned int count) ecb_const; |
|
|
420 | ecb_inline uint8_t ecb_rotr8 (uint8_t x, unsigned int count) ecb_const; |
|
|
421 | ecb_inline uint16_t ecb_rotl16 (uint16_t x, unsigned int count) ecb_const; |
|
|
422 | ecb_inline uint16_t ecb_rotr16 (uint16_t x, unsigned int count) ecb_const; |
|
|
423 | ecb_inline uint32_t ecb_rotl32 (uint32_t x, unsigned int count) ecb_const; |
|
|
424 | ecb_inline uint32_t ecb_rotr32 (uint32_t x, unsigned int count) ecb_const; |
|
|
425 | ecb_inline uint64_t ecb_rotl64 (uint64_t x, unsigned int count) ecb_const; |
|
|
426 | ecb_inline uint64_t ecb_rotr64 (uint64_t x, unsigned int count) ecb_const; |
|
|
427 | |
|
|
428 | ecb_inline uint8_t ecb_rotl8 (uint8_t x, unsigned int count) { return (x >> ( 8 - count)) | (x << count); } |
|
|
429 | ecb_inline uint8_t ecb_rotr8 (uint8_t x, unsigned int count) { return (x << ( 8 - count)) | (x >> count); } |
|
|
430 | ecb_inline uint16_t ecb_rotl16 (uint16_t x, unsigned int count) { return (x >> (16 - count)) | (x << count); } |
|
|
431 | ecb_inline uint16_t ecb_rotr16 (uint16_t x, unsigned int count) { return (x << (16 - count)) | (x >> count); } |
|
|
432 | ecb_inline uint32_t ecb_rotl32 (uint32_t x, unsigned int count) { return (x >> (32 - count)) | (x << count); } |
|
|
433 | ecb_inline uint32_t ecb_rotr32 (uint32_t x, unsigned int count) { return (x << (32 - count)) | (x >> count); } |
|
|
434 | ecb_inline uint64_t ecb_rotl64 (uint64_t x, unsigned int count) { return (x >> (64 - count)) | (x << count); } |
|
|
435 | ecb_inline uint64_t ecb_rotr64 (uint64_t x, unsigned int count) { return (x << (64 - count)) | (x >> count); } |
|
|
436 | |
229 | #if ECB_GCC_VERSION(4,3) |
437 | #if ECB_GCC_VERSION(4,3) |
230 | #define ecb_bswap16(x) (__builtin_bswap32 (x) >> 16) |
438 | #define ecb_bswap16(x) (__builtin_bswap32 (x) >> 16) |
231 | #define ecb_bswap32(x) __builtin_bswap32 (x) |
439 | #define ecb_bswap32(x) __builtin_bswap32 (x) |
232 | #define ecb_bswap64(x) __builtin_bswap64 (x) |
440 | #define ecb_bswap64(x) __builtin_bswap64 (x) |
233 | #else |
441 | #else |
234 | ecb_function_ uint32_t ecb_bswap16 (uint32_t x) ecb_const; |
442 | ecb_function_ uint16_t ecb_bswap16 (uint16_t x) ecb_const; |
235 | ecb_function_ uint32_t |
443 | ecb_function_ uint16_t |
236 | ecb_bswap16 (uint32_t x) |
444 | ecb_bswap16 (uint16_t x) |
237 | { |
445 | { |
238 | return ((x >> 8) & 0xff) |
446 | return ecb_rotl16 (x, 8); |
239 | | ((x << 8) & 0x00ff0000) |
|
|
240 | | (x << 24); |
|
|
241 | } |
447 | } |
242 | |
448 | |
243 | ecb_function_ uint32_t ecb_bswap32 (uint32_t x) ecb_const; |
449 | ecb_function_ uint32_t ecb_bswap32 (uint32_t x) ecb_const; |
244 | ecb_function_ uint32_t |
450 | ecb_function_ uint32_t |
245 | ecb_bswap32 (uint32_t x) |
451 | ecb_bswap32 (uint32_t x) |
246 | { |
452 | { |
247 | return (x >> 24) |
453 | return (((uint32_t)ecb_bswap16 (x)) << 16) | ecb_bswap16 (x >> 16); |
248 | | ((x >> 8) & 0x0000ff00) |
|
|
249 | | ((x << 8) & 0x00ff0000) |
|
|
250 | | (x << 24); |
|
|
251 | } |
454 | } |
252 | |
455 | |
253 | ecb_function_ uint64_t ecb_bswap64 (uint64_t x) ecb_const; |
456 | ecb_function_ uint64_t ecb_bswap64 (uint64_t x) ecb_const; |
254 | ecb_function_ uint64_t |
457 | ecb_function_ uint64_t |
255 | ecb_bswap64 (uint64_t x) |
458 | ecb_bswap64 (uint64_t x) |
256 | { |
459 | { |
257 | return (((uint64_t)ecb_bswap32 (x)) << 32) |
460 | return (((uint64_t)ecb_bswap32 (x)) << 32) | ecb_bswap32 (x >> 32); |
258 | | ecb_bswap32 (x >> 32); |
|
|
259 | } |
461 | } |
260 | #endif |
462 | #endif |
261 | |
463 | |
262 | #if ECB_GCC_VERSION(4,5) |
464 | #if ECB_GCC_VERSION(4,5) |
263 | #define ecb_unreachable() __builtin_unreachable () |
465 | #define ecb_unreachable() __builtin_unreachable () |
264 | #else |
466 | #else |
265 | /* this seems to work fine, but gcc always emits a warning for it :/ */ |
467 | /* this seems to work fine, but gcc always emits a warning for it :/ */ |
266 | ecb_function_ void ecb_unreachable (void) ecb_noreturn; |
468 | ecb_inline void ecb_unreachable (void) ecb_noreturn; |
267 | ecb_function_ void ecb_unreachable (void) { } |
469 | ecb_inline void ecb_unreachable (void) { } |
268 | #endif |
470 | #endif |
269 | |
471 | |
270 | /* try to tell the compiler that some condition is definitely true */ |
472 | /* try to tell the compiler that some condition is definitely true */ |
271 | #define ecb_assume(cond) do { if (!(cond)) ecb_unreachable (); } while (0) |
473 | #define ecb_assume(cond) if (!(cond)) ecb_unreachable (); else 0 |
272 | |
474 | |
273 | ecb_function_ unsigned char ecb_byteorder_helper (void) ecb_const; |
475 | ecb_inline unsigned char ecb_byteorder_helper (void) ecb_const; |
274 | ecb_function_ unsigned char |
476 | ecb_inline unsigned char |
275 | ecb_byteorder_helper (void) |
477 | ecb_byteorder_helper (void) |
276 | { |
478 | { |
277 | const uint32_t u = 0x11223344; |
479 | /* the union code still generates code under pressure in gcc, */ |
278 | return *(unsigned char *)&u; |
480 | /* but less than using pointers, and always seem to */ |
|
|
481 | /* successfully return a constant. */ |
|
|
482 | /* the reason why we have this horrible preprocessor mess */ |
|
|
483 | /* is to avoid it in all cases, at least on common architectures */ |
|
|
484 | /* and yes, gcc defines __BYTE_ORDER__, g++ does not */ |
|
|
485 | #if __i386 || __i386__ || _M_X86 || __amd64 || __amd64__ || _M_X64 |
|
|
486 | return 0x44; |
|
|
487 | #elif __BYTE_ORDER__ && __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__ |
|
|
488 | return 0x44; |
|
|
489 | #elif __BYTE_ORDER__ && __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__ |
|
|
490 | retrurn 0x11; |
|
|
491 | #else |
|
|
492 | union |
|
|
493 | { |
|
|
494 | uint32_t i; |
|
|
495 | uint8_t c; |
|
|
496 | } u = { 0x11223344 }; |
|
|
497 | return u.c; |
|
|
498 | #endif |
279 | } |
499 | } |
280 | |
500 | |
281 | ecb_function_ ecb_bool ecb_big_endian (void) ecb_const; |
501 | ecb_inline ecb_bool ecb_big_endian (void) ecb_const; |
282 | ecb_function_ ecb_bool ecb_big_endian (void) { return ecb_byteorder_helper () == 0x11; } |
502 | ecb_inline ecb_bool ecb_big_endian (void) { return ecb_byteorder_helper () == 0x11; } |
283 | ecb_function_ ecb_bool ecb_little_endian (void) ecb_const; |
503 | ecb_inline ecb_bool ecb_little_endian (void) ecb_const; |
284 | ecb_function_ ecb_bool ecb_little_endian (void) { return ecb_byteorder_helper () == 0x44; } |
504 | ecb_inline ecb_bool ecb_little_endian (void) { return ecb_byteorder_helper () == 0x44; } |
285 | |
505 | |
286 | #if ECB_GCC_VERSION(3,0) || ECB_C99 |
506 | #if ECB_GCC_VERSION(3,0) || ECB_C99 |
287 | #define ecb_mod(m,n) ((m) % (n) + ((m) % (n) < 0 ? (n) : 0)) |
507 | #define ecb_mod(m,n) ((m) % (n) + ((m) % (n) < 0 ? (n) : 0)) |
288 | #else |
508 | #else |
289 | #define ecb_mod(m,n) ((m) < 0 ? ((n) - 1 - ((-1 - (m)) % (n))) : ((m) % (n))) |
509 | #define ecb_mod(m,n) ((m) < 0 ? ((n) - 1 - ((-1 - (m)) % (n))) : ((m) % (n))) |
|
|
510 | #endif |
|
|
511 | |
|
|
512 | #if __cplusplus |
|
|
513 | template<typename T> |
|
|
514 | static inline T ecb_div_rd (T val, T div) |
|
|
515 | { |
|
|
516 | return val < 0 ? - ((-val + div - 1) / div) : (val ) / div; |
|
|
517 | } |
|
|
518 | template<typename T> |
|
|
519 | static inline T ecb_div_ru (T val, T div) |
|
|
520 | { |
|
|
521 | return val < 0 ? - ((-val ) / div) : (val + div - 1) / div; |
|
|
522 | } |
|
|
523 | #else |
|
|
524 | #define ecb_div_rd(val,div) ((val) < 0 ? - ((-(val) + (div) - 1) / (div)) : ((val) ) / (div)) |
|
|
525 | #define ecb_div_ru(val,div) ((val) < 0 ? - ((-(val) ) / (div)) : ((val) + (div) - 1) / (div)) |
290 | #endif |
526 | #endif |
291 | |
527 | |
292 | #if ecb_cplusplus_does_not_suck |
528 | #if ecb_cplusplus_does_not_suck |
293 | /* does not work for local types (http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2657.htm) */ |
529 | /* does not work for local types (http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2657.htm) */ |
294 | template<typename T, int N> |
530 | template<typename T, int N> |
… | |
… | |
298 | } |
534 | } |
299 | #else |
535 | #else |
300 | #define ecb_array_length(name) (sizeof (name) / sizeof (name [0])) |
536 | #define ecb_array_length(name) (sizeof (name) / sizeof (name [0])) |
301 | #endif |
537 | #endif |
302 | |
538 | |
303 | ecb_inline uint32_t ecb_rotr32 (uint32_t x, unsigned int count) ecb_const; |
539 | #if __STDC_IEC_559__ |
304 | ecb_inline uint32_t |
540 | // we assume this is defined for most C and many C++ compilers |
305 | ecb_rotr32 (uint32_t x, unsigned int count) |
541 | ecb_inline ecb_bool ecb_float_ieee (void) ecb_const; |
306 | { |
542 | ecb_inline ecb_bool ecb_float_ieee (void) { return 1; } |
307 | return (x << (32 - count)) | (x >> count); |
543 | ecb_inline ecb_bool ecb_double_ieee (void) ecb_const; |
308 | } |
544 | ecb_inline ecb_bool ecb_double_ieee (void) { return 1; } |
309 | |
545 | #elif ECB_CPP |
310 | ecb_inline uint32_t ecb_rotl32 (uint32_t x, unsigned int count) ecb_const; |
546 | #include <limits> |
311 | ecb_inline uint32_t |
547 | ecb_inline ecb_bool ecb_float_ieee (void) ecb_const; |
312 | ecb_rotl32 (uint32_t x, unsigned int count) |
548 | ecb_inline ecb_bool ecb_float_ieee (void) { return std::numeric_limits<float >::is_iec559; } |
313 | { |
549 | ecb_inline ecb_bool ecb_double_ieee (void) ecb_const; |
314 | return (x >> (32 - count)) | (x << count); |
550 | ecb_inline ecb_bool ecb_double_ieee (void) { return std::numeric_limits<double>::is_iec559; } |
315 | } |
551 | #else |
316 | |
552 | ecb_inline ecb_bool ecb_float_ieee (void) ecb_const; |
317 | ecb_inline uint64_t ecb_rotr64 (uint64_t x, unsigned int count) ecb_const; |
553 | ecb_inline ecb_bool ecb_float_ieee (void) { return 0; } |
318 | ecb_inline uint64_t |
554 | ecb_inline ecb_bool ecb_double_ieee (void) ecb_const; |
319 | ecb_rotr64 (uint64_t x, unsigned int count) |
555 | ecb_inline ecb_bool ecb_double_ieee (void) { return 0; } |
320 | { |
|
|
321 | return (x << (64 - count)) | (x >> count); |
|
|
322 | } |
|
|
323 | |
|
|
324 | ecb_inline uint64_t ecb_rotl64 (uint64_t x, unsigned int count) ecb_const; |
|
|
325 | ecb_inline uint64_t |
|
|
326 | ecb_rotl64 (uint64_t x, unsigned int count) |
|
|
327 | { |
|
|
328 | return (x >> (64 - count)) | (x << count); |
|
|
329 | } |
|
|
330 | |
|
|
331 | #endif |
556 | #endif |
332 | |
557 | |
|
|
558 | // convert a float to ieee single/binary32 |
|
|
559 | ecb_function_ uint32_t ecb_float_to_binary32 (float x) ecb_const; |
|
|
560 | ecb_function_ uint32_t |
|
|
561 | ecb_float_to_binary32 (float x) |
|
|
562 | { |
|
|
563 | /* slow emulation, works for anything but nan's and -0 */ |
|
|
564 | ECB_EXTERN_C float frexpf (float v, int *e); |
|
|
565 | uint32_t r, m; |
|
|
566 | int e; |
|
|
567 | |
|
|
568 | if (x == 0e0f ) return 0; |
|
|
569 | if (x > +3.40282346638528860e+38f) return 0x7f800000U; |
|
|
570 | if (x < -3.40282346638528860e+38f) return 0xff800000U; |
|
|
571 | |
|
|
572 | m = frexpf (x, &e) * 0x1000000U; |
|
|
573 | |
|
|
574 | r = m & 0x80000000U; |
|
|
575 | |
|
|
576 | if (r) |
|
|
577 | m = -m; |
|
|
578 | |
|
|
579 | if (e < -125) |
|
|
580 | { |
|
|
581 | m &= 0xffffffU; |
|
|
582 | m >>= (-125 - e); |
|
|
583 | e = -126; |
|
|
584 | } |
|
|
585 | |
|
|
586 | r |= (e + 126) << 23; |
|
|
587 | r |= m & 0x7fffffU; |
|
|
588 | |
|
|
589 | return r; |
|
|
590 | } |
|
|
591 | |
|
|
592 | // converts a ieee single/binary32 to a float |
|
|
593 | ecb_function_ float ecb_binary32_to_float (uint32_t x) ecb_const; |
|
|
594 | ecb_function_ float |
|
|
595 | ecb_binary32_to_float (uint32_t x) |
|
|
596 | { |
|
|
597 | /* emulation, only works for normals and subnormals and +0 */ |
|
|
598 | ECB_EXTERN_C float ldexpf (float x, int e); |
|
|
599 | |
|
|
600 | int neg = x >> 31; |
|
|
601 | int e = (x >> 23) & 0xffU; |
|
|
602 | float r; |
|
|
603 | |
|
|
604 | x &= 0x7fffffU; |
|
|
605 | |
|
|
606 | if (e) |
|
|
607 | x |= 0x800000U; |
|
|
608 | |
|
|
609 | /* we distrust ldexpf a bit and do the 2**-24 scaling by an extra multiply */ |
|
|
610 | r = ldexpf (x * (1.f / 0x1000000U), e - 126); |
|
|
611 | |
|
|
612 | return neg ? -r : r; |
|
|
613 | } |
|
|
614 | |
|
|
615 | ecb_function_ uint64_t ecb_double_to_binary64 (double x) ecb_const; |
|
|
616 | ecb_function_ uint64_t |
|
|
617 | ecb_double_to_binary64 (double x) |
|
|
618 | { |
|
|
619 | } |
|
|
620 | |
|
|
621 | #endif |
|
|
622 | |