ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/libecb/ecb.h
(Generate patch)

Comparing libecb/ecb.h (file contents):
Revision 1.37 by root, Sat Jun 11 13:47:47 2011 UTC vs.
Revision 1.57 by root, Wed Aug 3 15:15:00 2011 UTC

28 */ 28 */
29 29
30#ifndef ECB_H 30#ifndef ECB_H
31#define ECB_H 31#define ECB_H
32 32
33#ifdef _WIN32
34 typedef signed char int8_t;
35 typedef unsigned char uint8_t;
36 typedef signed short int16_t;
37 typedef unsigned short uint16_t;
38 typedef signed int int32_t;
39 typedef unsigned int uint32_t;
40 #if __GNUC__
41 typedef signed long long int64_t;
42 typedef unsigned long long uint64_t;
43 #else /* _MSC_VER || __BORLANDC__ */
44 typedef signed __int64 int64_t;
45 typedef unsigned __int64 uint64_t;
46 #endif
47#else
33#include <inttypes.h> 48 #include <inttypes.h>
49#endif
34 50
35/* many compilers define _GNUC_ to some versions but then only implement 51/* many compilers define _GNUC_ to some versions but then only implement
36 * what their idiot authors think are the "more important" extensions, 52 * what their idiot authors think are the "more important" extensions,
37 * causing enourmous grief in return for some better fake benchmark numbers. 53 * causing enourmous grief in return for some better fake benchmark numbers.
38 * or so. 54 * or so.
39 * we try to detect these and simply assume they are not gcc - if they have 55 * we try to detect these and simply assume they are not gcc - if they have
40 * an issue with that they should have done it right in the first place. 56 * an issue with that they should have done it right in the first place.
41 */ 57 */
42#ifndef ECB_GCC_VERSION 58#ifndef ECB_GCC_VERSION
43 #if defined(__INTEL_COMPILER) || defined(__SUNPRO_C) || defined(__llvm__) 59 #if !defined(__GNUC_MINOR__) || defined(__INTEL_COMPILER) || defined(__SUNPRO_C) || defined(__SUNPRO_CC) || defined(__llvm__) || defined(__clang__)
44 #define ECB_GCC_VERSION(major,minor) 0 60 #define ECB_GCC_VERSION(major,minor) 0
45 #else 61 #else
46 #define ECB_GCC_VERSION(major,minor) (__GNUC__ > (major) || (__GNUC__ == (major) && __GNUC_MINOR__ >= (minor))) 62 #define ECB_GCC_VERSION(major,minor) (__GNUC__ > (major) || (__GNUC__ == (major) && __GNUC_MINOR__ >= (minor)))
47 #endif 63 #endif
48#endif 64#endif
49 65
50#ifndef __cplusplus 66/*****************************************************************************/
51 #if __STDC_VERSION__ >= 199901L 67
52 #define ECB_INLINE static inline 68#ifndef ECB_MEMORY_FENCE
53 typedef _Bool ecb_bool; 69 #if ECB_GCC_VERSION(2,5)
54 #else 70 #if __x86
55 #define ECB_INLINE static inline /* we assume the extension is ubiquituous, please tell us when we are wrong, or upgrade to the GCC */ 71 #define ECB_MEMORY_FENCE __asm__ __volatile__ ("lock; orb $0, -1(%%esp)" : : : "memory")
56 typedef int ecb_bool; 72 #define ECB_MEMORY_FENCE_ACQUIRE ECB_MEMORY_FENCE /* non-lock xchg might be enough */
73 #define ECB_MEMORY_FENCE_RELEASE do { } while (0) /* unlikely to change in future cpus */
74 #elif __amd64
75 #define ECB_MEMORY_FENCE __asm__ __volatile__ ("mfence" : : : "memory")
76 #define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("lfence" : : : "memory")
77 #define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("sfence") /* play safe - not needed in any current cpu */
78 #endif
57 #endif 79 #endif
58#else 80#endif
59 #define ECB_INLINE static inline 81
60 typedef bool ecb_bool; 82#ifndef ECB_MEMORY_FENCE
83 #if ECB_GCC_VERSION(4,4)
84 #define ECB_MEMORY_FENCE __sync_synchronize ()
85 #define ECB_MEMORY_FENCE_ACQUIRE ({ char dummy = 0; __sync_lock_test_and_set (&dummy, 1); })
86 #define ECB_MEMORY_FENCE_RELEASE ({ char dummy = 1; __sync_lock_release (&dummy ); })
87 #elif _MSC_VER >= 1400 /* VC++ 2005 */
88 #pragma intrinsic(_ReadBarrier,_WriteBarrier,_ReadWriteBarrier)
89 #define ECB_MEMORY_FENCE _ReadWriteBarrier ()
90 #define ECB_MEMORY_FENCE_ACQUIRE _ReadWriteBarrier () /* according to msdn, _ReadBarrier is not a load fence */
91 #define ECB_MEMORY_FENCE_RELEASE _WriteBarrier ()
92 #elif defined(_WIN32)
93 #include <WinNT.h>
94 #define ECB_MEMORY_FENCE MemoryBarrier () /* actually just xchg on x86... scary */
95 #define ECB_MEMORY_FENCE_ACQUIRE ECB_MEMORY_FENCE
96 #define ECB_MEMORY_FENCE_RELEASE ECB_MEMORY_FENCE
61#endif 97 #endif
98#endif
99
100#ifndef ECB_MEMORY_FENCE
101 /*
102 * if you get undefined symbol references to pthread_mutex_lock,
103 * or failure to find pthread.h, then you should implement
104 * the ECB_MEMORY_FENCE operations for your cpu/compiler
105 * OR proide pthread.h and link against the posix thread library
106 * of your system.
107 */
108 #include <pthread.h>
109
110 static pthread_mutex_t ecb_mf_lock = PTHREAD_MUTEX_INITIALIZER;
111 #define ECB_MEMORY_FENCE do { pthread_mutex_lock (&ecb_mf_lock); pthread_mutex_unlock (&ecb_mf_lock); } while (0)
112 #define ECB_MEMORY_FENCE_ACQUIRE ECB_MEMORY_FENCE
113 #define ECB_MEMORY_FENCE_RELEASE ECB_MEMORY_FENCE
114#endif
115
116/*****************************************************************************/
117
118#define ECB_C99 (__STDC_VERSION__ >= 199901L)
119
120#if __cplusplus
121 #define ecb_inline static inline
122#elif ECB_GCC_VERSION(2,5)
123 #define ecb_inline static __inline__
124#elif ECB_C99
125 #define ecb_inline static inline
126#else
127 #define ecb_inline static
128#endif
129
130#if ECB_GCC_VERSION(3,3)
131 #define ecb_restrict __restrict__
132#elif ECB_C99
133 #define ecb_restrict restrict
134#else
135 #define ecb_restrict
136#endif
137
138typedef int ecb_bool;
62 139
63#define ECB_CONCAT_(a, b) a ## b 140#define ECB_CONCAT_(a, b) a ## b
64#define ECB_CONCAT(a, b) ECB_CONCAT_(a, b) 141#define ECB_CONCAT(a, b) ECB_CONCAT_(a, b)
65#define ECB_STRINGIFY_(a) # a 142#define ECB_STRINGIFY_(a) # a
66#define ECB_STRINGIFY(a) ECB_STRINGIFY_(a) 143#define ECB_STRINGIFY(a) ECB_STRINGIFY_(a)
67 144
68#define ECB_HEADER_INLINE ECB_INLINE 145#define ecb_function_ ecb_inline
69 146
70#if ECB_GCC_VERSION(3,1) 147#if ECB_GCC_VERSION(3,1)
71 #define ecb_attribute(attrlist) __attribute__(attrlist) 148 #define ecb_attribute(attrlist) __attribute__(attrlist)
72 #define ecb_is_constant(expr) __builtin_constant_p (expr) 149 #define ecb_is_constant(expr) __builtin_constant_p (expr)
73 #define ecb_expect(expr,value) __builtin_expect ((expr),(value)) 150 #define ecb_expect(expr,value) __builtin_expect ((expr),(value))
81 158
82/* no emulation for ecb_decltype */ 159/* no emulation for ecb_decltype */
83#if ECB_GCC_VERSION(4,5) 160#if ECB_GCC_VERSION(4,5)
84 #define ecb_decltype(x) __decltype(x) 161 #define ecb_decltype(x) __decltype(x)
85#elif ECB_GCC_VERSION(3,0) 162#elif ECB_GCC_VERSION(3,0)
86 #define ecb_decltype(x) typeof(x) 163 #define ecb_decltype(x) __typeof(x)
87#endif 164#endif
88 165
89#define ecb_noinline ecb_attribute ((__noinline__)) 166#define ecb_noinline ecb_attribute ((__noinline__))
90#define ecb_noreturn ecb_attribute ((__noreturn__)) 167#define ecb_noreturn ecb_attribute ((__noreturn__))
91#define ecb_unused ecb_attribute ((__unused__)) 168#define ecb_unused ecb_attribute ((__unused__))
92#define ecb_const ecb_attribute ((__const__)) 169#define ecb_const ecb_attribute ((__const__))
93#define ecb_pure ecb_attribute ((__pure__)) 170#define ecb_pure ecb_attribute ((__pure__))
94 171
95#if ECB_GCC_VERSION(4,3) 172#if ECB_GCC_VERSION(4,3)
96 #define ecb_artificial ecb_attribute ((__artificial__)) /* 4.3 */ 173 #define ecb_artificial ecb_attribute ((__artificial__))
97 #define ecb_hot ecb_attribute ((__hot__)) /* 4.3 */ 174 #define ecb_hot ecb_attribute ((__hot__))
98 #define ecb_cold ecb_attribute ((__cold__)) /* 4.3 */ 175 #define ecb_cold ecb_attribute ((__cold__))
99#else 176#else
100 #define ecb_artificial 177 #define ecb_artificial
101 #define ecb_hot 178 #define ecb_hot
102 #define ecb_cold 179 #define ecb_cold
103#endif 180#endif
104 181
105/* put into if's if you are very sure that the expression */ 182/* put around conditional expressions if you are very sure that the */
106/* is mostly true or mostly false. note that these return */ 183/* expression is mostly true or mostly false. note that these return */
107/* booleans, not the expression. */ 184/* booleans, not the expression. */
108#define ecb_expect_false(expr) ecb_expect (!!(expr), 0) 185#define ecb_expect_false(expr) ecb_expect (!!(expr), 0)
109#define ecb_expect_true(expr) ecb_expect (!!(expr), 1) 186#define ecb_expect_true(expr) ecb_expect (!!(expr), 1)
110/* for compatibility to the rest of the world */ 187/* for compatibility to the rest of the world */
111#define ecb_likely(expr) ecb_expect_true (expr) 188#define ecb_likely(expr) ecb_expect_true (expr)
112#define ecb_unlikely(expr) ecb_expect_false (expr) 189#define ecb_unlikely(expr) ecb_expect_false (expr)
113 190
114/* try to tell the compiler that some condition is definitely true */
115#define ecb_assume(cond) do { if (!(cond)) ecb_unreachable (); } while (0)
116
117/* count trailing zero bits and count # of one bits */ 191/* count trailing zero bits and count # of one bits */
118#if ECB_GCC_VERSION(3,4) 192#if ECB_GCC_VERSION(3,4)
193 /* we assume int == 32 bit, long == 32 or 64 bit and long long == 64 bit */
194 #define ecb_ld32(x) (__builtin_clz (x) ^ 31)
195 #define ecb_ld64(x) (__builtin_clzll (x) ^ 63)
119 #define ecb_ctz32(x) __builtin_ctz (x) 196 #define ecb_ctz32(x) __builtin_ctz (x)
197 #define ecb_ctz64(x) __builtin_ctzll (x)
120 #define ecb_popcount32(x) __builtin_popcount (x) 198 #define ecb_popcount32(x) __builtin_popcount (x)
199 /* no popcountll */
121#else 200#else
122 ECB_HEADER_INLINE int ecb_ctz32 (uint32_t x) ecb_const; 201 ecb_function_ int ecb_ctz32 (uint32_t x) ecb_const;
123 ECB_HEADER_INLINE int 202 ecb_function_ int
124 ecb_ctz32 (uint32_t x) 203 ecb_ctz32 (uint32_t x)
125 { 204 {
126 int r = 0; 205 int r = 0;
127 206
128 x &= -x; /* this isolates the lowest bit */ 207 x &= ~x + 1; /* this isolates the lowest bit */
129 208
209#if ECB_branchless_on_i386
210 r += !!(x & 0xaaaaaaaa) << 0;
211 r += !!(x & 0xcccccccc) << 1;
212 r += !!(x & 0xf0f0f0f0) << 2;
213 r += !!(x & 0xff00ff00) << 3;
214 r += !!(x & 0xffff0000) << 4;
215#else
130 if (x & 0xaaaaaaaa) r += 1; 216 if (x & 0xaaaaaaaa) r += 1;
131 if (x & 0xcccccccc) r += 2; 217 if (x & 0xcccccccc) r += 2;
132 if (x & 0xf0f0f0f0) r += 4; 218 if (x & 0xf0f0f0f0) r += 4;
133 if (x & 0xff00ff00) r += 8; 219 if (x & 0xff00ff00) r += 8;
134 if (x & 0xffff0000) r += 16; 220 if (x & 0xffff0000) r += 16;
221#endif
135 222
136 return r; 223 return r;
137 } 224 }
138 225
226 ecb_function_ int ecb_ctz64 (uint64_t x) ecb_const;
227 ecb_function_ int
228 ecb_ctz64 (uint64_t x)
229 {
230 int shift = x & 0xffffffffU ? 0 : 32;
231 return ecb_ctz32 (x >> shift) + shift;
232 }
233
139 ECB_HEADER_INLINE int ecb_popcount32 (uint32_t x) ecb_const; 234 ecb_function_ int ecb_popcount32 (uint32_t x) ecb_const;
140 ECB_HEADER_INLINE int 235 ecb_function_ int
141 ecb_popcount32 (uint32_t x) 236 ecb_popcount32 (uint32_t x)
142 { 237 {
143 x -= (x >> 1) & 0x55555555; 238 x -= (x >> 1) & 0x55555555;
144 x = ((x >> 2) & 0x33333333) + (x & 0x33333333); 239 x = ((x >> 2) & 0x33333333) + (x & 0x33333333);
145 x = ((x >> 4) + x) & 0x0f0f0f0f; 240 x = ((x >> 4) + x) & 0x0f0f0f0f;
146 x *= 0x01010101; 241 x *= 0x01010101;
147 242
148 return x >> 24; 243 return x >> 24;
149 } 244 }
245
246 ecb_function_ int ecb_ld32 (uint32_t x) ecb_const;
247 ecb_function_ int ecb_ld32 (uint32_t x)
248 {
249 int r = 0;
250
251 if (x >> 16) { x >>= 16; r += 16; }
252 if (x >> 8) { x >>= 8; r += 8; }
253 if (x >> 4) { x >>= 4; r += 4; }
254 if (x >> 2) { x >>= 2; r += 2; }
255 if (x >> 1) { r += 1; }
256
257 return r;
258 }
259
260 ecb_function_ int ecb_ld64 (uint64_t x) ecb_const;
261 ecb_function_ int ecb_ld64 (uint64_t x)
262 {
263 int r = 0;
264
265 if (x >> 32) { x >>= 32; r += 32; }
266
267 return r + ecb_ld32 (x);
268 }
150#endif 269#endif
270
271/* popcount64 is only available on 64 bit cpus as gcc builtin */
272/* so for this version we are lazy */
273ecb_function_ int ecb_popcount64 (uint64_t x) ecb_const;
274ecb_function_ int
275ecb_popcount64 (uint64_t x)
276{
277 return ecb_popcount32 (x) + ecb_popcount32 (x >> 32);
278}
279
280ecb_inline uint8_t ecb_rotl8 (uint8_t x, unsigned int count) ecb_const;
281ecb_inline uint8_t ecb_rotr8 (uint8_t x, unsigned int count) ecb_const;
282ecb_inline uint16_t ecb_rotl16 (uint16_t x, unsigned int count) ecb_const;
283ecb_inline uint16_t ecb_rotr16 (uint16_t x, unsigned int count) ecb_const;
284ecb_inline uint32_t ecb_rotl32 (uint32_t x, unsigned int count) ecb_const;
285ecb_inline uint32_t ecb_rotr32 (uint32_t x, unsigned int count) ecb_const;
286ecb_inline uint64_t ecb_rotl64 (uint64_t x, unsigned int count) ecb_const;
287ecb_inline uint64_t ecb_rotr64 (uint64_t x, unsigned int count) ecb_const;
288
289ecb_inline uint8_t ecb_rotl8 (uint8_t x, unsigned int count) { return (x >> ( 8 - count)) | (x << count); }
290ecb_inline uint8_t ecb_rotr8 (uint8_t x, unsigned int count) { return (x << ( 8 - count)) | (x >> count); }
291ecb_inline uint16_t ecb_rotl16 (uint16_t x, unsigned int count) { return (x >> (16 - count)) | (x << count); }
292ecb_inline uint16_t ecb_rotr16 (uint16_t x, unsigned int count) { return (x << (16 - count)) | (x >> count); }
293ecb_inline uint32_t ecb_rotl32 (uint32_t x, unsigned int count) { return (x >> (32 - count)) | (x << count); }
294ecb_inline uint32_t ecb_rotr32 (uint32_t x, unsigned int count) { return (x << (32 - count)) | (x >> count); }
295ecb_inline uint64_t ecb_rotl64 (uint64_t x, unsigned int count) { return (x >> (64 - count)) | (x << count); }
296ecb_inline uint64_t ecb_rotr64 (uint64_t x, unsigned int count) { return (x << (64 - count)) | (x >> count); }
151 297
152#if ECB_GCC_VERSION(4,3) 298#if ECB_GCC_VERSION(4,3)
153 #define ecb_bswap32(x) __builtin_bswap32 (x)
154 #define ecb_bswap16(x) (__builtin_bswap32(x) >> 16) 299 #define ecb_bswap16(x) (__builtin_bswap32 (x) >> 16)
300 #define ecb_bswap32(x) __builtin_bswap32 (x)
301 #define ecb_bswap64(x) __builtin_bswap64 (x)
155#else 302#else
303 ecb_function_ uint16_t ecb_bswap16 (uint16_t x) ecb_const;
304 ecb_function_ uint16_t
305 ecb_bswap16 (uint16_t x)
306 {
307 return ecb_rotl16 (x, 8);
308 }
309
156 ECB_HEADER_INLINE uint32_t ecb_bswap32 (uint32_t x) ecb_const; 310 ecb_function_ uint32_t ecb_bswap32 (uint32_t x) ecb_const;
157 ECB_HEADER_INLINE uint32_t 311 ecb_function_ uint32_t
158 ecb_bswap32 (uint32_t x) 312 ecb_bswap32 (uint32_t x)
159 { 313 {
160 return (x >> 24) 314 return (((uint32_t)ecb_bswap16 (x)) << 16) | ecb_bswap16 (x >> 16);
161 | ((x >> 8) & 0x0000ff00)
162 | ((x << 8) & 0x00ff0000)
163 | (x << 24);
164 } 315 }
165 316
166 ECB_HEADER_INLINE uint32_t ecb_bswap16 (uint32_t x) ecb_const; 317 ecb_function_ uint64_t ecb_bswap64 (uint64_t x) ecb_const;
167 ECB_HEADER_INLINE uint32_t 318 ecb_function_ uint64_t
168 ecb_bswap16 (uint32_t x) 319 ecb_bswap64 (uint64_t x)
169 { 320 {
170 return ((x >> 8) & 0xff) 321 return (((uint64_t)ecb_bswap32 (x)) << 32) | ecb_bswap32 (x >> 32);
171 | ((x << 8) & 0x00ff0000)
172 | (x << 24);
173 } 322 }
174#endif 323#endif
175 324
176#if ECB_GCC_VERSION(4,5) 325#if ECB_GCC_VERSION(4,5)
177 #define ecb_unreachable() __builtin_unreachable () 326 #define ecb_unreachable() __builtin_unreachable ()
178#else 327#else
179 /* this seems to work fine, but gcc always emits a warning for it :/ */ 328 /* this seems to work fine, but gcc always emits a warning for it :/ */
180 ECB_HEADER_INLINE void ecb_unreachable (void) ecb_noreturn; 329 ecb_function_ void ecb_unreachable (void) ecb_noreturn;
181 ECB_HEADER_INLINE void ecb_unreachable (void) { } 330 ecb_function_ void ecb_unreachable (void) { }
182#endif 331#endif
183 332
333/* try to tell the compiler that some condition is definitely true */
334#define ecb_assume(cond) do { if (!(cond)) ecb_unreachable (); } while (0)
335
184ECB_HEADER_INLINE unsigned char ecb_byteorder_helper (void) ecb_const; 336ecb_function_ unsigned char ecb_byteorder_helper (void) ecb_const;
185ECB_HEADER_INLINE unsigned char 337ecb_function_ unsigned char
186ecb_byteorder_helper (void) 338ecb_byteorder_helper (void)
187{ 339{
188 const uint32_t u = 0x11223344; 340 const uint32_t u = 0x11223344;
189 return *(unsigned char *)&u; 341 return *(unsigned char *)&u;
190} 342}
191 343
192ECB_HEADER_INLINE ecb_bool ecb_big_endian (void) ecb_const; 344ecb_function_ ecb_bool ecb_big_endian (void) ecb_const;
193ECB_HEADER_INLINE ecb_bool ecb_big_endian (void) { return ecb_byteorder_helper () == 0x11; }; 345ecb_function_ ecb_bool ecb_big_endian (void) { return ecb_byteorder_helper () == 0x11; }
194ECB_HEADER_INLINE ecb_bool ecb_little_endian (void) ecb_const; 346ecb_function_ ecb_bool ecb_little_endian (void) ecb_const;
195ECB_HEADER_INLINE ecb_bool ecb_little_endian (void) { return ecb_byteorder_helper () == 0x44; }; 347ecb_function_ ecb_bool ecb_little_endian (void) { return ecb_byteorder_helper () == 0x44; }
196 348
197#if ECB_GCC_VERSION(3,0) 349#if ECB_GCC_VERSION(3,0) || ECB_C99
198 #define ecb_mod(m,n) ((m) % (n) + ((m) % (n) < 0 ? (n) : 0)) 350 #define ecb_mod(m,n) ((m) % (n) + ((m) % (n) < 0 ? (n) : 0))
199#else 351#else
200 #define ecb_mod(m,n) ((m) < 0 ? ((n) - 1 - ((-1 - (m)) % (n))) : ((m) % (n))) 352 #define ecb_mod(m,n) ((m) < 0 ? ((n) - 1 - ((-1 - (m)) % (n))) : ((m) % (n)))
201#endif 353#endif
202 354
203#if ecb_cplusplus_does_not_suck 355#if ecb_cplusplus_does_not_suck
204 // does not work for local types (http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2657.htm) 356 /* does not work for local types (http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2657.htm) */
205 template<typename T, int N> 357 template<typename T, int N>
206 static inline int ecb_array_length (const T (&arr)[N]) 358 static inline int ecb_array_length (const T (&arr)[N])
207 { 359 {
208 return N; 360 return N;
209 } 361 }
210#else 362#else
211 #define ecb_array_length(name) (sizeof (name) / sizeof (name [0])) 363 #define ecb_array_length(name) (sizeof (name) / sizeof (name [0]))
212#endif 364#endif
213 365
214ECB_INLINE uint32_t ecb_rotr32 (uint32_t x, unsigned int count) ecb_const;
215ECB_INLINE uint32_t
216ecb_rotr32 (uint32_t x, unsigned int count)
217{
218 return (x << (32 - count)) | (x >> count);
219}
220
221ECB_INLINE uint32_t ecb_rotl32 (uint32_t x, unsigned int count) ecb_const;
222ECB_INLINE uint32_t
223ecb_rotl32 (uint32_t x, unsigned int count)
224{
225 return (x >> (32 - count)) | (x << count);
226}
227
228#endif 366#endif
229 367

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines