1 |
/* |
2 |
* libecb - http://software.schmorp.de/pkg/libecb |
3 |
* |
4 |
* Copyright (©) 2009-2012 Marc Alexander Lehmann <libecb@schmorp.de> |
5 |
* Copyright (©) 2011 Emanuele Giaquinta |
6 |
* All rights reserved. |
7 |
* |
8 |
* Redistribution and use in source and binary forms, with or without modifica- |
9 |
* tion, are permitted provided that the following conditions are met: |
10 |
* |
11 |
* 1. Redistributions of source code must retain the above copyright notice, |
12 |
* this list of conditions and the following disclaimer. |
13 |
* |
14 |
* 2. Redistributions in binary form must reproduce the above copyright |
15 |
* notice, this list of conditions and the following disclaimer in the |
16 |
* documentation and/or other materials provided with the distribution. |
17 |
* |
18 |
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR IMPLIED |
19 |
* WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MER- |
20 |
* CHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO |
21 |
* EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPE- |
22 |
* CIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, |
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, |
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 |
27 |
* OF THE POSSIBILITY OF SUCH DAMAGE. |
28 |
*/ |
29 |
|
30 |
#ifndef ECB_H |
31 |
#define ECB_H |
32 |
|
33 |
/* 16 bits major, 16 bits minor */ |
34 |
#define ECB_VERSION 0x00010001 |
35 |
|
36 |
#ifdef _WIN32 |
37 |
typedef signed char int8_t; |
38 |
typedef unsigned char uint8_t; |
39 |
typedef signed short int16_t; |
40 |
typedef unsigned short uint16_t; |
41 |
typedef signed int int32_t; |
42 |
typedef unsigned int uint32_t; |
43 |
#if __GNUC__ |
44 |
typedef signed long long int64_t; |
45 |
typedef unsigned long long uint64_t; |
46 |
#else /* _MSC_VER || __BORLANDC__ */ |
47 |
typedef signed __int64 int64_t; |
48 |
typedef unsigned __int64 uint64_t; |
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 |
59 |
#else |
60 |
#include <inttypes.h> |
61 |
#if UINTMAX_MAX > 0xffffffffU |
62 |
#define ECB_PTRSIZE 8 |
63 |
#else |
64 |
#define ECB_PTRSIZE 4 |
65 |
#endif |
66 |
#endif |
67 |
|
68 |
/* many compilers define _GNUC_ to some versions but then only implement |
69 |
* what their idiot authors think are the "more important" extensions, |
70 |
* causing enormous grief in return for some better fake benchmark numbers. |
71 |
* or so. |
72 |
* we try to detect these and simply assume they are not gcc - if they have |
73 |
* an issue with that they should have done it right in the first place. |
74 |
*/ |
75 |
#ifndef ECB_GCC_VERSION |
76 |
#if !defined __GNUC_MINOR__ || defined __INTEL_COMPILER || defined __SUNPRO_C || defined __SUNPRO_CC || defined __llvm__ || defined __clang__ |
77 |
#define ECB_GCC_VERSION(major,minor) 0 |
78 |
#else |
79 |
#define ECB_GCC_VERSION(major,minor) (__GNUC__ > (major) || (__GNUC__ == (major) && __GNUC_MINOR__ >= (minor))) |
80 |
#endif |
81 |
#endif |
82 |
|
83 |
#define ECB_C (__STDC__+0) /* this assumes that __STDC__ is either empty or a number */ |
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 |
/*****************************************************************************/ |
218 |
|
219 |
#if __cplusplus |
220 |
#define ecb_inline static inline |
221 |
#elif ECB_GCC_VERSION(2,5) |
222 |
#define ecb_inline static __inline__ |
223 |
#elif ECB_C99 |
224 |
#define ecb_inline static inline |
225 |
#else |
226 |
#define ecb_inline static |
227 |
#endif |
228 |
|
229 |
#if ECB_GCC_VERSION(3,3) |
230 |
#define ecb_restrict __restrict__ |
231 |
#elif ECB_C99 |
232 |
#define ecb_restrict restrict |
233 |
#else |
234 |
#define ecb_restrict |
235 |
#endif |
236 |
|
237 |
typedef int ecb_bool; |
238 |
|
239 |
#define ECB_CONCAT_(a, b) a ## b |
240 |
#define ECB_CONCAT(a, b) ECB_CONCAT_(a, b) |
241 |
#define ECB_STRINGIFY_(a) # a |
242 |
#define ECB_STRINGIFY(a) ECB_STRINGIFY_(a) |
243 |
|
244 |
#define ecb_function_ ecb_inline |
245 |
|
246 |
#if ECB_GCC_VERSION(3,1) |
247 |
#define ecb_attribute(attrlist) __attribute__(attrlist) |
248 |
#define ecb_is_constant(expr) __builtin_constant_p (expr) |
249 |
#define ecb_expect(expr,value) __builtin_expect ((expr),(value)) |
250 |
#define ecb_prefetch(addr,rw,locality) __builtin_prefetch (addr, rw, locality) |
251 |
#else |
252 |
#define ecb_attribute(attrlist) |
253 |
#define ecb_is_constant(expr) 0 |
254 |
#define ecb_expect(expr,value) (expr) |
255 |
#define ecb_prefetch(addr,rw,locality) |
256 |
#endif |
257 |
|
258 |
/* no emulation for ecb_decltype */ |
259 |
#if ECB_GCC_VERSION(4,5) |
260 |
#define ecb_decltype(x) __decltype(x) |
261 |
#elif ECB_GCC_VERSION(3,0) |
262 |
#define ecb_decltype(x) __typeof(x) |
263 |
#endif |
264 |
|
265 |
#define ecb_noinline ecb_attribute ((__noinline__)) |
266 |
#define ecb_unused ecb_attribute ((__unused__)) |
267 |
#define ecb_const ecb_attribute ((__const__)) |
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 |
275 |
|
276 |
#if ECB_GCC_VERSION(4,3) |
277 |
#define ecb_artificial ecb_attribute ((__artificial__)) |
278 |
#define ecb_hot ecb_attribute ((__hot__)) |
279 |
#define ecb_cold ecb_attribute ((__cold__)) |
280 |
#else |
281 |
#define ecb_artificial |
282 |
#define ecb_hot |
283 |
#define ecb_cold |
284 |
#endif |
285 |
|
286 |
/* put around conditional expressions if you are very sure that the */ |
287 |
/* expression is mostly true or mostly false. note that these return */ |
288 |
/* booleans, not the expression. */ |
289 |
#define ecb_expect_false(expr) ecb_expect (!!(expr), 0) |
290 |
#define ecb_expect_true(expr) ecb_expect (!!(expr), 1) |
291 |
/* for compatibility to the rest of the world */ |
292 |
#define ecb_likely(expr) ecb_expect_true (expr) |
293 |
#define ecb_unlikely(expr) ecb_expect_false (expr) |
294 |
|
295 |
/* count trailing zero bits and count # of one bits */ |
296 |
#if ECB_GCC_VERSION(3,4) |
297 |
/* we assume int == 32 bit, long == 32 or 64 bit and long long == 64 bit */ |
298 |
#define ecb_ld32(x) (__builtin_clz (x) ^ 31) |
299 |
#define ecb_ld64(x) (__builtin_clzll (x) ^ 63) |
300 |
#define ecb_ctz32(x) __builtin_ctz (x) |
301 |
#define ecb_ctz64(x) __builtin_ctzll (x) |
302 |
#define ecb_popcount32(x) __builtin_popcount (x) |
303 |
/* no popcountll */ |
304 |
#else |
305 |
ecb_function_ int ecb_ctz32 (uint32_t x) ecb_const; |
306 |
ecb_function_ int |
307 |
ecb_ctz32 (uint32_t x) |
308 |
{ |
309 |
int r = 0; |
310 |
|
311 |
x &= ~x + 1; /* this isolates the lowest bit */ |
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 |
320 |
if (x & 0xaaaaaaaa) r += 1; |
321 |
if (x & 0xcccccccc) r += 2; |
322 |
if (x & 0xf0f0f0f0) r += 4; |
323 |
if (x & 0xff00ff00) r += 8; |
324 |
if (x & 0xffff0000) r += 16; |
325 |
#endif |
326 |
|
327 |
return r; |
328 |
} |
329 |
|
330 |
ecb_function_ int ecb_ctz64 (uint64_t x) ecb_const; |
331 |
ecb_function_ int |
332 |
ecb_ctz64 (uint64_t x) |
333 |
{ |
334 |
int shift = x & 0xffffffffU ? 0 : 32; |
335 |
return ecb_ctz32 (x >> shift) + shift; |
336 |
} |
337 |
|
338 |
ecb_function_ int ecb_popcount32 (uint32_t x) ecb_const; |
339 |
ecb_function_ int |
340 |
ecb_popcount32 (uint32_t x) |
341 |
{ |
342 |
x -= (x >> 1) & 0x55555555; |
343 |
x = ((x >> 2) & 0x33333333) + (x & 0x33333333); |
344 |
x = ((x >> 4) + x) & 0x0f0f0f0f; |
345 |
x *= 0x01010101; |
346 |
|
347 |
return x >> 24; |
348 |
} |
349 |
|
350 |
ecb_function_ int ecb_ld32 (uint32_t x) ecb_const; |
351 |
ecb_function_ int ecb_ld32 (uint32_t x) |
352 |
{ |
353 |
int r = 0; |
354 |
|
355 |
if (x >> 16) { x >>= 16; r += 16; } |
356 |
if (x >> 8) { x >>= 8; r += 8; } |
357 |
if (x >> 4) { x >>= 4; r += 4; } |
358 |
if (x >> 2) { x >>= 2; r += 2; } |
359 |
if (x >> 1) { r += 1; } |
360 |
|
361 |
return r; |
362 |
} |
363 |
|
364 |
ecb_function_ int ecb_ld64 (uint64_t x) ecb_const; |
365 |
ecb_function_ int ecb_ld64 (uint64_t x) |
366 |
{ |
367 |
int r = 0; |
368 |
|
369 |
if (x >> 32) { x >>= 32; r += 32; } |
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 |
|
395 |
return x; |
396 |
} |
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 |
} |
409 |
|
410 |
/* popcount64 is only available on 64 bit cpus as gcc builtin */ |
411 |
/* so for this version we are lazy */ |
412 |
ecb_function_ int ecb_popcount64 (uint64_t x) ecb_const; |
413 |
ecb_function_ int |
414 |
ecb_popcount64 (uint64_t x) |
415 |
{ |
416 |
return ecb_popcount32 (x) + ecb_popcount32 (x >> 32); |
417 |
} |
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 |
|
437 |
#if ECB_GCC_VERSION(4,3) |
438 |
#define ecb_bswap16(x) (__builtin_bswap32 (x) >> 16) |
439 |
#define ecb_bswap32(x) __builtin_bswap32 (x) |
440 |
#define ecb_bswap64(x) __builtin_bswap64 (x) |
441 |
#else |
442 |
ecb_function_ uint16_t ecb_bswap16 (uint16_t x) ecb_const; |
443 |
ecb_function_ uint16_t |
444 |
ecb_bswap16 (uint16_t x) |
445 |
{ |
446 |
return ecb_rotl16 (x, 8); |
447 |
} |
448 |
|
449 |
ecb_function_ uint32_t ecb_bswap32 (uint32_t x) ecb_const; |
450 |
ecb_function_ uint32_t |
451 |
ecb_bswap32 (uint32_t x) |
452 |
{ |
453 |
return (((uint32_t)ecb_bswap16 (x)) << 16) | ecb_bswap16 (x >> 16); |
454 |
} |
455 |
|
456 |
ecb_function_ uint64_t ecb_bswap64 (uint64_t x) ecb_const; |
457 |
ecb_function_ uint64_t |
458 |
ecb_bswap64 (uint64_t x) |
459 |
{ |
460 |
return (((uint64_t)ecb_bswap32 (x)) << 32) | ecb_bswap32 (x >> 32); |
461 |
} |
462 |
#endif |
463 |
|
464 |
#if ECB_GCC_VERSION(4,5) |
465 |
#define ecb_unreachable() __builtin_unreachable () |
466 |
#else |
467 |
/* this seems to work fine, but gcc always emits a warning for it :/ */ |
468 |
ecb_inline void ecb_unreachable (void) ecb_noreturn; |
469 |
ecb_inline void ecb_unreachable (void) { } |
470 |
#endif |
471 |
|
472 |
/* try to tell the compiler that some condition is definitely true */ |
473 |
#define ecb_assume(cond) if (!(cond)) ecb_unreachable (); else 0 |
474 |
|
475 |
ecb_inline unsigned char ecb_byteorder_helper (void) ecb_const; |
476 |
ecb_inline unsigned char |
477 |
ecb_byteorder_helper (void) |
478 |
{ |
479 |
/* the union code still generates code under pressure in gcc, */ |
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 |
499 |
} |
500 |
|
501 |
ecb_inline ecb_bool ecb_big_endian (void) ecb_const; |
502 |
ecb_inline ecb_bool ecb_big_endian (void) { return ecb_byteorder_helper () == 0x11; } |
503 |
ecb_inline ecb_bool ecb_little_endian (void) ecb_const; |
504 |
ecb_inline ecb_bool ecb_little_endian (void) { return ecb_byteorder_helper () == 0x44; } |
505 |
|
506 |
#if ECB_GCC_VERSION(3,0) || ECB_C99 |
507 |
#define ecb_mod(m,n) ((m) % (n) + ((m) % (n) < 0 ? (n) : 0)) |
508 |
#else |
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)) |
526 |
#endif |
527 |
|
528 |
#if ecb_cplusplus_does_not_suck |
529 |
/* does not work for local types (http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2657.htm) */ |
530 |
template<typename T, int N> |
531 |
static inline int ecb_array_length (const T (&arr)[N]) |
532 |
{ |
533 |
return N; |
534 |
} |
535 |
#else |
536 |
#define ecb_array_length(name) (sizeof (name) / sizeof (name [0])) |
537 |
#endif |
538 |
|
539 |
/*******************************************************************************/ |
540 |
/* floating point stuff, can be disabled by defining ECB_NO_LIBM */ |
541 |
|
542 |
/* basically, everything uses "ieee pure-endian" floating point numbers */ |
543 |
/* the only noteworthy exception is ancient armle, which uses order 43218765 */ |
544 |
#if 0 \ |
545 |
|| __i386 || __i386__ \ |
546 |
|| __amd64 || __amd64__ || __x86_64 || __x86_64__ \ |
547 |
|| __powerpc__ || __ppc__ || __powerpc64__ || __ppc64__ \ |
548 |
|| defined __arm__ && defined __ARM_EABI__ \ |
549 |
|| defined __s390__ || defined __s390x__ \ |
550 |
|| defined __mips__ \ |
551 |
|| defined __alpha__ \ |
552 |
|| defined __hppa__ \ |
553 |
|| defined __ia64__ \ |
554 |
|| defined _M_IX86 || defined _M_AMD64 || defined _M_IA64 |
555 |
#define ECB_STDFP 1 |
556 |
#include <string.h> /* for memcpy */ |
557 |
#else |
558 |
#define ECB_STDFP 0 |
559 |
#endif |
560 |
|
561 |
#ifndef ECB_NO_LIBM |
562 |
|
563 |
/* convert a float to ieee single/binary32 */ |
564 |
ecb_function_ uint32_t ecb_float_to_binary32 (float x) ecb_const; |
565 |
ecb_function_ uint32_t |
566 |
ecb_float_to_binary32 (float x) |
567 |
{ |
568 |
uint32_t r; |
569 |
|
570 |
#if ECB_STDFP |
571 |
memcpy (&r, &x, 4); |
572 |
#else |
573 |
/* slow emulation, works for anything but -0 */ |
574 |
ECB_EXTERN_C float frexpf (float v, int *e); |
575 |
uint32_t m; |
576 |
int e; |
577 |
|
578 |
if (x == 0e0f ) return 0; |
579 |
if (x > +3.40282346638528860e+38f) return 0x7f800000U; |
580 |
if (x < -3.40282346638528860e+38f) return 0xff800000U; |
581 |
if (x != x ) return 0x7fbfffffU; |
582 |
|
583 |
m = frexpf (x, &e) * 0x1000000U; |
584 |
|
585 |
r = m & 0x80000000U; |
586 |
|
587 |
if (r) |
588 |
m = -m; |
589 |
|
590 |
if (e < -125) |
591 |
{ |
592 |
m &= 0xffffffU; |
593 |
m >>= (-125 - e); |
594 |
e = -126; |
595 |
} |
596 |
|
597 |
r |= (e + 126) << 23; |
598 |
r |= m & 0x7fffffU; |
599 |
#endif |
600 |
|
601 |
return r; |
602 |
} |
603 |
|
604 |
/* converts an ieee single/binary32 to a float */ |
605 |
ecb_function_ float ecb_binary32_to_float (uint32_t x) ecb_const; |
606 |
ecb_function_ float |
607 |
ecb_binary32_to_float (uint32_t x) |
608 |
{ |
609 |
float r; |
610 |
|
611 |
#if ECB_STDFP |
612 |
memcpy (&r, &x, 4); |
613 |
#else |
614 |
/* emulation, only works for normals and subnormals and +0 */ |
615 |
ECB_EXTERN_C float ldexpf (float x, int e); |
616 |
|
617 |
int neg = x >> 31; |
618 |
int e = (x >> 23) & 0xffU; |
619 |
|
620 |
x &= 0x7fffffU; |
621 |
|
622 |
if (e) |
623 |
x |= 0x800000U; |
624 |
else |
625 |
e = 1; |
626 |
|
627 |
/* we distrust ldexpf a bit and do the 2**-24 scaling by an extra multiply */ |
628 |
r = ldexpf (x * (1.f / 0x1000000U), e - 126); |
629 |
|
630 |
r = neg ? -r : r; |
631 |
#endif |
632 |
|
633 |
return r; |
634 |
} |
635 |
|
636 |
/* convert a double to ieee double/binary64 */ |
637 |
ecb_function_ uint64_t ecb_double_to_binary64 (double x) ecb_const; |
638 |
ecb_function_ uint64_t |
639 |
ecb_double_to_binary64 (double x) |
640 |
{ |
641 |
uint64_t r; |
642 |
|
643 |
#if ECB_STDFP |
644 |
memcpy (&r, &x, 8); |
645 |
#else |
646 |
/* slow emulation, works for anything but -0 */ |
647 |
ECB_EXTERN_C double frexp (double v, int *e); |
648 |
uint64_t m; |
649 |
int e; |
650 |
|
651 |
if (x == 0e0 ) return 0; |
652 |
if (x > +1.79769313486231470e+308) return 0x7ff0000000000000U; |
653 |
if (x < -1.79769313486231470e+308) return 0xfff0000000000000U; |
654 |
if (x != x ) return 0X7ff7ffffffffffffU; |
655 |
|
656 |
m = frexp (x, &e) * 0x20000000000000U; |
657 |
|
658 |
r = m & 0x8000000000000000;; |
659 |
|
660 |
if (r) |
661 |
m = -m; |
662 |
|
663 |
if (e < -1021) |
664 |
{ |
665 |
m &= 0x1fffffffffffffU; |
666 |
m >>= (-1021 - e); |
667 |
e = -1022; |
668 |
} |
669 |
|
670 |
r |= ((uint64_t)(e + 1022)) << 52; |
671 |
r |= m & 0xfffffffffffffU; |
672 |
#endif |
673 |
|
674 |
return r; |
675 |
} |
676 |
|
677 |
/* converts an ieee double/binary64 to a double */ |
678 |
ecb_function_ double ecb_binary64_to_double (uint64_t x) ecb_const; |
679 |
ecb_function_ double |
680 |
ecb_binary64_to_double (uint64_t x) |
681 |
{ |
682 |
double r; |
683 |
|
684 |
#if ECB_STDFP |
685 |
memcpy (&r, &x, 8); |
686 |
#else |
687 |
/* emulation, only works for normals and subnormals and +0 */ |
688 |
ECB_EXTERN_C double ldexp (double x, int e); |
689 |
|
690 |
int neg = x >> 63; |
691 |
int e = (x >> 52) & 0x7ffU; |
692 |
|
693 |
x &= 0xfffffffffffffU; |
694 |
|
695 |
if (e) |
696 |
x |= 0x10000000000000U; |
697 |
else |
698 |
e = 1; |
699 |
|
700 |
/* we distrust ldexp a bit and do the 2**-53 scaling by an extra multiply */ |
701 |
r = ldexp (x * (1. / 0x20000000000000U), e - 1022); |
702 |
|
703 |
r = neg ? -r : r; |
704 |
#endif |
705 |
|
706 |
return r; |
707 |
} |
708 |
|
709 |
#endif |
710 |
|
711 |
#endif |
712 |
|