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

Comparing libeio/ecb.h (file contents):
Revision 1.2 by root, Fri Jun 17 18:55:42 2011 UTC vs.
Revision 1.7 by root, Wed Aug 3 15:25:38 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>
… …
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#ifndef ECB_MEMORY_FENCE
69 #if ECB_GCC_VERSION(2,5)
70 #if __x86
71 #define ECB_MEMORY_FENCE __asm__ __volatile__ ("lock; orb $0, -1(%%esp)" : : : "memory")
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
79 #endif
80#endif
81
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
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
66#define ECB_C99 (__STDC_VERSION__ >= 199901L) 118#define ECB_C99 (__STDC_VERSION__ >= 199901L)
67 119
68#if __cplusplus 120#if __cplusplus
69 #define ecb_inline static inline 121 #define ecb_inline static inline
70#elif ECB_GCC_VERSION(2,5) 122#elif ECB_GCC_VERSION(2,5)
… …
152 { 204 {
153 int r = 0; 205 int r = 0;
154 206
155 x &= ~x + 1; /* this isolates the lowest bit */ 207 x &= ~x + 1; /* this isolates the lowest bit */
156 208
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
157 if (x & 0xaaaaaaaa) r += 1; 216 if (x & 0xaaaaaaaa) r += 1;
158 if (x & 0xcccccccc) r += 2; 217 if (x & 0xcccccccc) r += 2;
159 if (x & 0xf0f0f0f0) r += 4; 218 if (x & 0xf0f0f0f0) r += 4;
160 if (x & 0xff00ff00) r += 8; 219 if (x & 0xff00ff00) r += 8;
161 if (x & 0xffff0000) r += 16; 220 if (x & 0xffff0000) r += 16;
221#endif
162 222
163 return r; 223 return r;
164 } 224 }
165 225
166 ecb_function_ int ecb_ctz64 (uint64_t x) ecb_const; 226 ecb_function_ int ecb_ctz64 (uint64_t x) ecb_const;
167 ecb_function_ int 227 ecb_function_ int
168 ecb_ctz64 (uint64_t x) 228 ecb_ctz64 (uint64_t x)
169 { 229 {
170 int shift = x & 0xffffffffU ? 0 : 32; 230 int shift = x & 0xffffffffU ? 0 : 32;
171 return ecb_ctz (x >> shift) + shift; 231 return ecb_ctz32 (x >> shift) + shift;
172 } 232 }
173 233
174 ecb_function_ int ecb_popcount32 (uint32_t x) ecb_const; 234 ecb_function_ int ecb_popcount32 (uint32_t x) ecb_const;
175 ecb_function_ int 235 ecb_function_ int
176 ecb_popcount32 (uint32_t x) 236 ecb_popcount32 (uint32_t x)
… …
187 /* something using lots of bit arithmetic and a simple loop */ 247 /* something using lots of bit arithmetic and a simple loop */
188 /* we went for the loop */ 248 /* we went for the loop */
189 ecb_function_ int ecb_ld32 (uint32_t x) ecb_const; 249 ecb_function_ int ecb_ld32 (uint32_t x) ecb_const;
190 ecb_function_ int ecb_ld32 (uint32_t x) 250 ecb_function_ int ecb_ld32 (uint32_t x)
191 { 251 {
192 int r = -1; 252 int r = 0;
193 253
194 do 254 if (x >> 16) { x >>= 16; r += 16; }
195 { 255 if (x >> 8) { x >>= 8; r += 8; }
196 x >>= 1; 256 if (x >> 4) { x >>= 4; r += 4; }
197 ++r; 257 if (x >> 2) { x >>= 2; r += 2; }
198 } 258 if (x >> 1) { r += 1; }
199 while (x);
200 259
201 return r; 260 return r;
202 } 261 }
203 262
204 ecb_function_ int ecb_ld64 (uint64_t x) ecb_const; 263 ecb_function_ int ecb_ld64 (uint64_t x) ecb_const;
205 ecb_function_ int ecb_ld64 (uint64_t x) 264 ecb_function_ int ecb_ld64 (uint64_t x)
206 { 265 {
207 int r = -1; 266 int r = 0;
208 267
209 do 268 if (x >> 32) { x >>= 32; r += 32; }
210 {
211 x >>= 1;
212 ++r;
213 }
214 while (x);
215 269
216 return r; 270 return r + ecb_ld32 (x);
217 } 271 }
218#endif 272#endif
219 273
220/* popcount64 is only available on 64 bit cpus as gcc builtin */ 274/* popcount64 is only available on 64 bit cpus as gcc builtin */
221/* so for this version we are lazy */ 275/* so for this version we are lazy */
… …
224ecb_popcount64 (uint64_t x) 278ecb_popcount64 (uint64_t x)
225{ 279{
226 return ecb_popcount32 (x) + ecb_popcount32 (x >> 32); 280 return ecb_popcount32 (x) + ecb_popcount32 (x >> 32);
227} 281}
228 282
283ecb_inline uint8_t ecb_rotl8 (uint8_t x, unsigned int count) ecb_const;
284ecb_inline uint8_t ecb_rotr8 (uint8_t x, unsigned int count) ecb_const;
285ecb_inline uint16_t ecb_rotl16 (uint16_t x, unsigned int count) ecb_const;
286ecb_inline uint16_t ecb_rotr16 (uint16_t x, unsigned int count) ecb_const;
287ecb_inline uint32_t ecb_rotl32 (uint32_t x, unsigned int count) ecb_const;
288ecb_inline uint32_t ecb_rotr32 (uint32_t x, unsigned int count) ecb_const;
289ecb_inline uint64_t ecb_rotl64 (uint64_t x, unsigned int count) ecb_const;
290ecb_inline uint64_t ecb_rotr64 (uint64_t x, unsigned int count) ecb_const;
291
292ecb_inline uint8_t ecb_rotl8 (uint8_t x, unsigned int count) { return (x >> ( 8 - count)) | (x << count); }
293ecb_inline uint8_t ecb_rotr8 (uint8_t x, unsigned int count) { return (x << ( 8 - count)) | (x >> count); }
294ecb_inline uint16_t ecb_rotl16 (uint16_t x, unsigned int count) { return (x >> (16 - count)) | (x << count); }
295ecb_inline uint16_t ecb_rotr16 (uint16_t x, unsigned int count) { return (x << (16 - count)) | (x >> count); }
296ecb_inline uint32_t ecb_rotl32 (uint32_t x, unsigned int count) { return (x >> (32 - count)) | (x << count); }
297ecb_inline uint32_t ecb_rotr32 (uint32_t x, unsigned int count) { return (x << (32 - count)) | (x >> count); }
298ecb_inline uint64_t ecb_rotl64 (uint64_t x, unsigned int count) { return (x >> (64 - count)) | (x << count); }
299ecb_inline uint64_t ecb_rotr64 (uint64_t x, unsigned int count) { return (x << (64 - count)) | (x >> count); }
300
229#if ECB_GCC_VERSION(4,3) 301#if ECB_GCC_VERSION(4,3)
230 #define ecb_bswap16(x) (__builtin_bswap32 (x) >> 16) 302 #define ecb_bswap16(x) (__builtin_bswap32 (x) >> 16)
231 #define ecb_bswap32(x) __builtin_bswap32 (x) 303 #define ecb_bswap32(x) __builtin_bswap32 (x)
232 #define ecb_bswap64(x) __builtin_bswap64 (x) 304 #define ecb_bswap64(x) __builtin_bswap64 (x)
233#else 305#else
234 ecb_function_ uint32_t ecb_bswap16 (uint32_t x) ecb_const; 306 ecb_function_ uint16_t ecb_bswap16 (uint16_t x) ecb_const;
235 ecb_function_ uint32_t 307 ecb_function_ uint16_t
236 ecb_bswap16 (uint32_t x) 308 ecb_bswap16 (uint16_t x)
237 { 309 {
238 return ((x >> 8) & 0xff) 310 return ecb_rotl16 (x, 8);
239 | ((x << 8) & 0x00ff0000)
240 | (x << 24);
241 } 311 }
242 312
243 ecb_function_ uint32_t ecb_bswap32 (uint32_t x) ecb_const; 313 ecb_function_ uint32_t ecb_bswap32 (uint32_t x) ecb_const;
244 ecb_function_ uint32_t 314 ecb_function_ uint32_t
245 ecb_bswap32 (uint32_t x) 315 ecb_bswap32 (uint32_t x)
246 { 316 {
247 return (x >> 24) 317 return (((uint32_t)ecb_bswap16 (x)) << 16) | ecb_bswap16 (x >> 16);
248 | ((x >> 8) & 0x0000ff00)
249 | ((x << 8) & 0x00ff0000)
250 | (x << 24);
251 } 318 }
252 319
253 ecb_function_ uint64_t ecb_bswap64 (uint64_t x) ecb_const; 320 ecb_function_ uint64_t ecb_bswap64 (uint64_t x) ecb_const;
254 ecb_function_ uint64_t 321 ecb_function_ uint64_t
255 ecb_bswap64 (uint64_t x) 322 ecb_bswap64 (uint64_t x)
256 { 323 {
257 return (((uint64_t)ecb_bswap32 (x)) << 32) 324 return (((uint64_t)ecb_bswap32 (x)) << 32) | ecb_bswap32 (x >> 32);
258 | ecb_bswap32 (x >> 32);
259 } 325 }
260#endif 326#endif
261 327
262#if ECB_GCC_VERSION(4,5) 328#if ECB_GCC_VERSION(4,5)
263 #define ecb_unreachable() __builtin_unreachable () 329 #define ecb_unreachable() __builtin_unreachable ()
… …
298 } 364 }
299#else 365#else
300 #define ecb_array_length(name) (sizeof (name) / sizeof (name [0])) 366 #define ecb_array_length(name) (sizeof (name) / sizeof (name [0]))
301#endif 367#endif
302 368
303ecb_inline uint32_t ecb_rotr32 (uint32_t x, unsigned int count) ecb_const;
304ecb_inline uint32_t
305ecb_rotr32 (uint32_t x, unsigned int count)
306{
307 return (x << (32 - count)) | (x >> count);
308}
309
310ecb_inline uint32_t ecb_rotl32 (uint32_t x, unsigned int count) ecb_const;
311ecb_inline uint32_t
312ecb_rotl32 (uint32_t x, unsigned int count)
313{
314 return (x >> (32 - count)) | (x << count);
315}
316
317ecb_inline uint64_t ecb_rotr64 (uint64_t x, unsigned int count) ecb_const;
318ecb_inline uint64_t
319ecb_rotr64 (uint64_t x, unsigned int count)
320{
321 return (x << (64 - count)) | (x >> count);
322}
323
324ecb_inline uint64_t ecb_rotl64 (uint64_t x, unsigned int count) ecb_const;
325ecb_inline uint64_t
326ecb_rotl64 (uint64_t x, unsigned int count)
327{
328 return (x >> (64 - count)) | (x << count);
329}
330
331#endif 369#endif
332 370

Diff Legend

– Removed lines
+ Added lines
< Changed lines
> Changed lines