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

Comparing libecb/ecb.h (file contents):
Revision 1.48 by root, Fri Jun 17 16:17:26 2011 UTC vs.
Revision 1.61 by root, Thu Aug 4 14:27:40 2011 UTC

38 typedef signed int int32_t; 38 typedef signed int int32_t;
39 typedef unsigned int uint32_t; 39 typedef unsigned int uint32_t;
40 #if __GNUC__ 40 #if __GNUC__
41 typedef signed long long int64_t; 41 typedef signed long long int64_t;
42 typedef unsigned long long uint64_t; 42 typedef unsigned long long uint64_t;
43 #else 43 #else /* _MSC_VER || __BORLANDC__ */
44 typedef signed __int64 int64_t; 44 typedef signed __int64 int64_t;
45 typedef unsigned __int64 uint64_t; 45 typedef unsigned __int64 uint64_t;
46 #endif 46 #endif
47#else 47#else
48 #include <inttypes.h> 48 #include <inttypes.h>
49#endif 49#endif
50 50
51/* many compilers define _GNUC_ to some versions but then only implement 51/* many compilers define _GNUC_ to some versions but then only implement
52 * what their idiot authors think are the "more important" extensions, 52 * what their idiot authors think are the "more important" extensions,
53 * causing enourmous grief in return for some better fake benchmark numbers. 53 * causing enormous grief in return for some better fake benchmark numbers.
54 * or so. 54 * or so.
55 * 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
56 * 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.
57 */ 57 */
58#ifndef ECB_GCC_VERSION 58#ifndef ECB_GCC_VERSION
61 #else 61 #else
62 #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)))
63 #endif 63 #endif
64#endif 64#endif
65 65
66/*****************************************************************************/
67
68/* ECB_NO_THREADS - ecb is not used by multiple threads, ever */
69/* ECB_NO_SMP - ecb might be used in multiple threads, but only on a single cpu */
70
71#if ECB_NO_THREADS || ECB_NO_SMP
72 #define ECB_MEMORY_FENCE do { } while (0)
73 #define ECB_MEMORY_FENCE_ACQUIRE do { } while (0)
74 #define ECB_MEMORY_FENCE_RELEASE do { } while (0)
75#endif
76
77#ifndef ECB_MEMORY_FENCE
78 #if ECB_GCC_VERSION(2,5)
79 #if __x86
80 #define ECB_MEMORY_FENCE __asm__ __volatile__ ("lock; orb $0, -1(%%esp)" : : : "memory")
81 #define ECB_MEMORY_FENCE_ACQUIRE ECB_MEMORY_FENCE /* non-lock xchg might be enough */
82 #define ECB_MEMORY_FENCE_RELEASE do { } while (0) /* unlikely to change in future cpus */
83 #elif __amd64
84 #define ECB_MEMORY_FENCE __asm__ __volatile__ ("mfence" : : : "memory")
85 #define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("lfence" : : : "memory")
86 #define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("sfence") /* play safe - not needed in any current cpu */
87 #endif
88 #endif
89#endif
90
91#ifndef ECB_MEMORY_FENCE
92 #if ECB_GCC_VERSION(4,4) || defined(__INTEL_COMPILER)
93 #define ECB_MEMORY_FENCE __sync_synchronize ()
94 #define ECB_MEMORY_FENCE_ACQUIRE __sync_synchronize () /* ({ char dummy = 0; __sync_lock_test_and_set (&dummy, 1); }) */
95 #define ECB_MEMORY_FENCE_RELEASE __sync_synchronize () /* ({ char dummy = 1; __sync_lock_release (&dummy ); }) */
96 #elif _MSC_VER >= 1400 /* VC++ 2005 */
97 #pragma intrinsic(_ReadBarrier,_WriteBarrier,_ReadWriteBarrier)
98 #define ECB_MEMORY_FENCE _ReadWriteBarrier ()
99 #define ECB_MEMORY_FENCE_ACQUIRE _ReadWriteBarrier () /* according to msdn, _ReadBarrier is not a load fence */
100 #define ECB_MEMORY_FENCE_RELEASE _WriteBarrier ()
101 #elif defined(_WIN32)
102 #include <WinNT.h>
103 #define ECB_MEMORY_FENCE MemoryBarrier () /* actually just xchg on x86... scary */
104 #define ECB_MEMORY_FENCE_ACQUIRE ECB_MEMORY_FENCE
105 #define ECB_MEMORY_FENCE_RELEASE ECB_MEMORY_FENCE
106 #endif
107#endif
108
109#ifndef ECB_MEMORY_FENCE
110 /*
111 * if you get undefined symbol references to pthread_mutex_lock,
112 * or failure to find pthread.h, then you should implement
113 * the ECB_MEMORY_FENCE operations for your cpu/compiler
114 * OR provide pthread.h and link against the posix thread library
115 * of your system.
116 */
117 #include <pthread.h>
118 #define ECB_NEEDS_PTHREADS 1
119 #define ECB_MEMORY_FENCE_NEEDS_PTHREADS 1
120
121 static pthread_mutex_t ecb_mf_lock = PTHREAD_MUTEX_INITIALIZER;
122 #define ECB_MEMORY_FENCE do { pthread_mutex_lock (&ecb_mf_lock); pthread_mutex_unlock (&ecb_mf_lock); } while (0)
123 #define ECB_MEMORY_FENCE_ACQUIRE ECB_MEMORY_FENCE
124 #define ECB_MEMORY_FENCE_RELEASE ECB_MEMORY_FENCE
125#endif
126
127/*****************************************************************************/
128
66#define ECB_C99 (__STDC_VERSION__ >= 199901L) 129#define ECB_C99 (__STDC_VERSION__ >= 199901L)
67 130
68#if __cplusplus 131#if __cplusplus
69 #define ecb_inline static inline 132 #define ecb_inline static inline
70#elif ECB_GCC_VERSION(2,5) 133#elif ECB_GCC_VERSION(2,5)
136#define ecb_likely(expr) ecb_expect_true (expr) 199#define ecb_likely(expr) ecb_expect_true (expr)
137#define ecb_unlikely(expr) ecb_expect_false (expr) 200#define ecb_unlikely(expr) ecb_expect_false (expr)
138 201
139/* count trailing zero bits and count # of one bits */ 202/* count trailing zero bits and count # of one bits */
140#if ECB_GCC_VERSION(3,4) 203#if ECB_GCC_VERSION(3,4)
204 /* we assume int == 32 bit, long == 32 or 64 bit and long long == 64 bit */
205 #define ecb_ld32(x) (__builtin_clz (x) ^ 31)
206 #define ecb_ld64(x) (__builtin_clzll (x) ^ 63)
141 #define ecb_ctz32(x) __builtin_ctz (x) 207 #define ecb_ctz32(x) __builtin_ctz (x)
208 #define ecb_ctz64(x) __builtin_ctzll (x)
142 #define ecb_popcount32(x) __builtin_popcount (x) 209 #define ecb_popcount32(x) __builtin_popcount (x)
210 /* no popcountll */
143#else 211#else
144 ecb_function_ int ecb_ctz32 (uint32_t x) ecb_const; 212 ecb_function_ int ecb_ctz32 (uint32_t x) ecb_const;
145 ecb_function_ int 213 ecb_function_ int
146 ecb_ctz32 (uint32_t x) 214 ecb_ctz32 (uint32_t x)
147 { 215 {
148 int r = 0; 216 int r = 0;
149 217
150 x &= ~x + 1; /* this isolates the lowest bit */ 218 x &= ~x + 1; /* this isolates the lowest bit */
151 219
220#if ECB_branchless_on_i386
221 r += !!(x & 0xaaaaaaaa) << 0;
222 r += !!(x & 0xcccccccc) << 1;
223 r += !!(x & 0xf0f0f0f0) << 2;
224 r += !!(x & 0xff00ff00) << 3;
225 r += !!(x & 0xffff0000) << 4;
226#else
152 if (x & 0xaaaaaaaa) r += 1; 227 if (x & 0xaaaaaaaa) r += 1;
153 if (x & 0xcccccccc) r += 2; 228 if (x & 0xcccccccc) r += 2;
154 if (x & 0xf0f0f0f0) r += 4; 229 if (x & 0xf0f0f0f0) r += 4;
155 if (x & 0xff00ff00) r += 8; 230 if (x & 0xff00ff00) r += 8;
156 if (x & 0xffff0000) r += 16; 231 if (x & 0xffff0000) r += 16;
232#endif
157 233
158 return r; 234 return r;
235 }
236
237 ecb_function_ int ecb_ctz64 (uint64_t x) ecb_const;
238 ecb_function_ int
239 ecb_ctz64 (uint64_t x)
240 {
241 int shift = x & 0xffffffffU ? 0 : 32;
242 return ecb_ctz32 (x >> shift) + shift;
159 } 243 }
160 244
161 ecb_function_ int ecb_popcount32 (uint32_t x) ecb_const; 245 ecb_function_ int ecb_popcount32 (uint32_t x) ecb_const;
162 ecb_function_ int 246 ecb_function_ int
163 ecb_popcount32 (uint32_t x) 247 ecb_popcount32 (uint32_t x)
167 x = ((x >> 4) + x) & 0x0f0f0f0f; 251 x = ((x >> 4) + x) & 0x0f0f0f0f;
168 x *= 0x01010101; 252 x *= 0x01010101;
169 253
170 return x >> 24; 254 return x >> 24;
171 } 255 }
256
257 ecb_function_ int ecb_ld32 (uint32_t x) ecb_const;
258 ecb_function_ int ecb_ld32 (uint32_t x)
259 {
260 int r = 0;
261
262 if (x >> 16) { x >>= 16; r += 16; }
263 if (x >> 8) { x >>= 8; r += 8; }
264 if (x >> 4) { x >>= 4; r += 4; }
265 if (x >> 2) { x >>= 2; r += 2; }
266 if (x >> 1) { r += 1; }
267
268 return r;
269 }
270
271 ecb_function_ int ecb_ld64 (uint64_t x) ecb_const;
272 ecb_function_ int ecb_ld64 (uint64_t x)
273 {
274 int r = 0;
275
276 if (x >> 32) { x >>= 32; r += 32; }
277
278 return r + ecb_ld32 (x);
279 }
172#endif 280#endif
281
282/* popcount64 is only available on 64 bit cpus as gcc builtin */
283/* so for this version we are lazy */
284ecb_function_ int ecb_popcount64 (uint64_t x) ecb_const;
285ecb_function_ int
286ecb_popcount64 (uint64_t x)
287{
288 return ecb_popcount32 (x) + ecb_popcount32 (x >> 32);
289}
290
291ecb_inline uint8_t ecb_rotl8 (uint8_t x, unsigned int count) ecb_const;
292ecb_inline uint8_t ecb_rotr8 (uint8_t x, unsigned int count) ecb_const;
293ecb_inline uint16_t ecb_rotl16 (uint16_t x, unsigned int count) ecb_const;
294ecb_inline uint16_t ecb_rotr16 (uint16_t x, unsigned int count) ecb_const;
295ecb_inline uint32_t ecb_rotl32 (uint32_t x, unsigned int count) ecb_const;
296ecb_inline uint32_t ecb_rotr32 (uint32_t x, unsigned int count) ecb_const;
297ecb_inline uint64_t ecb_rotl64 (uint64_t x, unsigned int count) ecb_const;
298ecb_inline uint64_t ecb_rotr64 (uint64_t x, unsigned int count) ecb_const;
299
300ecb_inline uint8_t ecb_rotl8 (uint8_t x, unsigned int count) { return (x >> ( 8 - count)) | (x << count); }
301ecb_inline uint8_t ecb_rotr8 (uint8_t x, unsigned int count) { return (x << ( 8 - count)) | (x >> count); }
302ecb_inline uint16_t ecb_rotl16 (uint16_t x, unsigned int count) { return (x >> (16 - count)) | (x << count); }
303ecb_inline uint16_t ecb_rotr16 (uint16_t x, unsigned int count) { return (x << (16 - count)) | (x >> count); }
304ecb_inline uint32_t ecb_rotl32 (uint32_t x, unsigned int count) { return (x >> (32 - count)) | (x << count); }
305ecb_inline uint32_t ecb_rotr32 (uint32_t x, unsigned int count) { return (x << (32 - count)) | (x >> count); }
306ecb_inline uint64_t ecb_rotl64 (uint64_t x, unsigned int count) { return (x >> (64 - count)) | (x << count); }
307ecb_inline uint64_t ecb_rotr64 (uint64_t x, unsigned int count) { return (x << (64 - count)) | (x >> count); }
173 308
174#if ECB_GCC_VERSION(4,3) 309#if ECB_GCC_VERSION(4,3)
175 #define ecb_bswap32(x) __builtin_bswap32 (x)
176 #define ecb_bswap16(x) (__builtin_bswap32(x) >> 16) 310 #define ecb_bswap16(x) (__builtin_bswap32 (x) >> 16)
311 #define ecb_bswap32(x) __builtin_bswap32 (x)
312 #define ecb_bswap64(x) __builtin_bswap64 (x)
177#else 313#else
314 ecb_function_ uint16_t ecb_bswap16 (uint16_t x) ecb_const;
315 ecb_function_ uint16_t
316 ecb_bswap16 (uint16_t x)
317 {
318 return ecb_rotl16 (x, 8);
319 }
320
178 ecb_function_ uint32_t ecb_bswap32 (uint32_t x) ecb_const; 321 ecb_function_ uint32_t ecb_bswap32 (uint32_t x) ecb_const;
179 ecb_function_ uint32_t 322 ecb_function_ uint32_t
180 ecb_bswap32 (uint32_t x) 323 ecb_bswap32 (uint32_t x)
181 { 324 {
182 return (x >> 24) 325 return (((uint32_t)ecb_bswap16 (x)) << 16) | ecb_bswap16 (x >> 16);
183 | ((x >> 8) & 0x0000ff00)
184 | ((x << 8) & 0x00ff0000)
185 | (x << 24);
186 } 326 }
187 327
188 ecb_function_ uint32_t ecb_bswap16 (uint32_t x) ecb_const; 328 ecb_function_ uint64_t ecb_bswap64 (uint64_t x) ecb_const;
189 ecb_function_ uint32_t 329 ecb_function_ uint64_t
190 ecb_bswap16 (uint32_t x) 330 ecb_bswap64 (uint64_t x)
191 { 331 {
192 return ((x >> 8) & 0xff) 332 return (((uint64_t)ecb_bswap32 (x)) << 32) | ecb_bswap32 (x >> 32);
193 | ((x << 8) & 0x00ff0000)
194 | (x << 24);
195 } 333 }
196#endif 334#endif
197 335
198#if ECB_GCC_VERSION(4,5) 336#if ECB_GCC_VERSION(4,5)
199 #define ecb_unreachable() __builtin_unreachable () 337 #define ecb_unreachable() __builtin_unreachable ()
234 } 372 }
235#else 373#else
236 #define ecb_array_length(name) (sizeof (name) / sizeof (name [0])) 374 #define ecb_array_length(name) (sizeof (name) / sizeof (name [0]))
237#endif 375#endif
238 376
239ecb_inline uint32_t ecb_rotr32 (uint32_t x, unsigned int count) ecb_const;
240ecb_inline uint32_t
241ecb_rotr32 (uint32_t x, unsigned int count)
242{
243 return (x << (32 - count)) | (x >> count);
244}
245
246ecb_inline uint32_t ecb_rotl32 (uint32_t x, unsigned int count) ecb_const;
247ecb_inline uint32_t
248ecb_rotl32 (uint32_t x, unsigned int count)
249{
250 return (x >> (32 - count)) | (x << count);
251}
252
253#endif 377#endif
254 378

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines