| 1 |
/*=========================================================================== |
| 2 |
* |
| 3 |
* PUBLIC DOMAIN NOTICE |
| 4 |
* National Center for Biotechnology Information |
| 5 |
* |
| 6 |
* This software/database is a "United States Government Work" under the |
| 7 |
* terms of the United States Copyright Act. It was written as part of |
| 8 |
* the author's official duties as a United States Government employee and |
| 9 |
* thus cannot be copyrighted. This software/database is freely available |
| 10 |
* to the public for use. The National Library of Medicine and the U.S. |
| 11 |
* Government have not placed any restriction on its use or reproduction. |
| 12 |
* |
| 13 |
* Although all reasonable efforts have been taken to ensure the accuracy |
| 14 |
* and reliability of the software and data, the NLM and the U.S. |
| 15 |
* Government do not and cannot warrant the performance or results that |
| 16 |
* may be obtained by using this software or data. The NLM and the U.S. |
| 17 |
* Government disclaim all warranties, express or implied, including |
| 18 |
* warranties of performance, merchantability or fitness for any particular |
| 19 |
* purpose. |
| 20 |
* |
| 21 |
* Please cite the author in any work or product based on this material. |
| 22 |
* |
| 23 |
* =========================================================================== |
| 24 |
* |
| 25 |
*/ |
| 26 |
|
| 27 |
#ifndef _h_atomic32_ |
| 28 |
#define _h_atomic32_ |
| 29 |
|
| 30 |
#ifdef __cplusplus |
| 31 |
extern "C" { |
| 32 |
#endif |
| 33 |
|
| 34 |
/* |
| 35 |
* Make sure gcc doesn't try to be clever and move things around |
| 36 |
* on us. We need to use _exactly_ the address the user gave us, |
| 37 |
* not some alias that contains the same information. |
| 38 |
*/ |
| 39 |
typedef struct atomic32_t atomic32_t; |
| 40 |
struct atomic32_t |
| 41 |
{ |
| 42 |
volatile int counter; |
| 43 |
}; |
| 44 |
|
| 45 |
/* int atomic32_read ( const atomic32_t *v ); */ |
| 46 |
#define atomic32_read( v ) \ |
| 47 |
( ( v ) -> counter ) |
| 48 |
|
| 49 |
/* void atomic32_set ( atomic32_t *v, int i ); */ |
| 50 |
#define atomic32_set( v, i ) \ |
| 51 |
( ( void ) ( ( ( v ) -> counter ) = ( i ) ) ) |
| 52 |
|
| 53 |
/* add to v -> counter and return the prior value */ |
| 54 |
static __inline__ int atomic32_read_and_add ( atomic32_t *v, int i ) |
| 55 |
{ |
| 56 |
int rtn, sum; |
| 57 |
__asm__ __volatile__ |
| 58 |
( |
| 59 |
"mov (%2), %0;" |
| 60 |
"1:" |
| 61 |
"mov %3, %1;" |
| 62 |
"add %0, %1;" |
| 63 |
"lock;" |
| 64 |
"cmpxchg %1, (%2);" |
| 65 |
"jne 1b" |
| 66 |
: "=&a" ( rtn ), "=&r" ( sum ) |
| 67 |
: "r" ( & v -> counter ), "r" ( i ) |
| 68 |
); |
| 69 |
return rtn; |
| 70 |
} |
| 71 |
|
| 72 |
/* if no read is needed, define the least expensive atomic add */ |
| 73 |
#define atomic32_add( v, i ) \ |
| 74 |
atomic32_read_and_add ( v, i ) |
| 75 |
|
| 76 |
/* add to v -> counter and return the result */ |
| 77 |
static __inline__ int atomic32_add_and_read ( atomic32_t *v, int i ) |
| 78 |
{ |
| 79 |
int rtn, cmp; |
| 80 |
__asm__ __volatile__ |
| 81 |
( |
| 82 |
"mov (%2), %0;" |
| 83 |
"1:" |
| 84 |
"mov %3, %1;" |
| 85 |
"add %0, %1;" |
| 86 |
"lock;" |
| 87 |
"cmpxchg %1,(%2);" |
| 88 |
"jne 1b;" |
| 89 |
: "=&a" ( cmp ), "=&r" ( rtn ) |
| 90 |
: "r" ( & v -> counter ), "r" ( i ) |
| 91 |
); |
| 92 |
return rtn; |
| 93 |
} |
| 94 |
|
| 95 |
/* just don't try to find out what the result was */ |
| 96 |
static __inline__ void atomic32_inc ( atomic32_t *v ) |
| 97 |
{ |
| 98 |
__asm__ __volatile__ |
| 99 |
( |
| 100 |
"lock;" |
| 101 |
"incl %0" |
| 102 |
: "=m" ( v -> counter ) |
| 103 |
: "m" ( v -> counter ) |
| 104 |
); |
| 105 |
} |
| 106 |
|
| 107 |
static __inline__ void atomic32_dec ( atomic32_t *v ) |
| 108 |
{ |
| 109 |
__asm__ __volatile__ |
| 110 |
( |
| 111 |
"lock;" |
| 112 |
"decl %0" |
| 113 |
: "=m" ( v -> counter ) |
| 114 |
: "m" ( v -> counter ) |
| 115 |
); |
| 116 |
} |
| 117 |
|
| 118 |
/* decrement by one and test result for 0 */ |
| 119 |
static __inline__ int atomic32_dec_and_test ( atomic32_t *v ) |
| 120 |
{ |
| 121 |
unsigned char c; |
| 122 |
__asm__ __volatile__ |
| 123 |
( |
| 124 |
"lock;" |
| 125 |
"decl %1;" |
| 126 |
"sete %0" |
| 127 |
: "=r" ( c ), "=m" ( v -> counter ) |
| 128 |
: "m" ( v -> counter ) |
| 129 |
); |
| 130 |
return c; |
| 131 |
} |
| 132 |
|
| 133 |
/* when atomic32_dec_and_test uses predecrement, you want |
| 134 |
postincrement to this function. so it isn't very useful */ |
| 135 |
static __inline__ int atomic32_inc_and_test ( atomic32_t *v ) |
| 136 |
{ |
| 137 |
unsigned char c; |
| 138 |
__asm__ __volatile__ |
| 139 |
( |
| 140 |
"lock;" |
| 141 |
"incl %1;" |
| 142 |
"sete %0" |
| 143 |
: "=r" ( c ), "=m" ( v -> counter ) |
| 144 |
: "m" ( v -> counter ) |
| 145 |
); |
| 146 |
return c; |
| 147 |
} |
| 148 |
|
| 149 |
/* HERE's useful */ |
| 150 |
#define atomic32_test_and_inc( v ) \ |
| 151 |
( atomic32_read_and_add ( v, 1 ) == 0 ) |
| 152 |
|
| 153 |
static __inline__ int atomic32_test_and_set ( atomic32_t *v, int s, int t ) |
| 154 |
{ |
| 155 |
int rtn; |
| 156 |
__asm__ __volatile__ |
| 157 |
( |
| 158 |
"lock;" |
| 159 |
"cmpxchg %2, (%1)" |
| 160 |
: "=a" ( rtn ) |
| 161 |
: "r" ( & v -> counter ), "r" ( s ), "a" ( t ) |
| 162 |
); |
| 163 |
return rtn; |
| 164 |
} |
| 165 |
|
| 166 |
/* conditional modifications */ |
| 167 |
static __inline__ |
| 168 |
int atomic32_read_and_add_lt ( atomic32_t *v, int i, int t ) |
| 169 |
{ |
| 170 |
int rtn, sum; |
| 171 |
__asm__ __volatile__ |
| 172 |
( |
| 173 |
"mov (%2), %0;" |
| 174 |
"1:" |
| 175 |
"cmp %4, %0;" |
| 176 |
"mov %3, %1;" |
| 177 |
"jge 2f;" |
| 178 |
"add %0, %1;" |
| 179 |
"lock;" |
| 180 |
"cmpxchg %1, (%2);" |
| 181 |
"jne 1b;" |
| 182 |
"2:" |
| 183 |
: "=&a" ( rtn ), "=&r" ( sum ) |
| 184 |
: "r" ( & v -> counter ), "r" ( i ), "r" ( t ) |
| 185 |
); |
| 186 |
return rtn; |
| 187 |
} |
| 188 |
|
| 189 |
#define atomic32_add_if_lt( v, i, t ) \ |
| 190 |
( atomic32_read_and_add_lt ( v, i, t ) < ( t ) ) |
| 191 |
|
| 192 |
static __inline__ |
| 193 |
int atomic32_read_and_add_le ( atomic32_t *v, int i, int t ) |
| 194 |
{ |
| 195 |
int rtn, sum; |
| 196 |
__asm__ __volatile__ |
| 197 |
( |
| 198 |
"mov (%2), %0;" |
| 199 |
"1:" |
| 200 |
"cmp %4, %0;" |
| 201 |
"mov %3, %1;" |
| 202 |
"jg 2f;" |
| 203 |
"add %0, %1;" |
| 204 |
"lock;" |
| 205 |
"cmpxchg %1, (%2);" |
| 206 |
"jne 1b;" |
| 207 |
"2:" |
| 208 |
: "=&a" ( rtn ), "=&r" ( sum ) |
| 209 |
: "r" ( & v -> counter ), "r" ( i ), "r" ( t ) |
| 210 |
); |
| 211 |
return rtn; |
| 212 |
} |
| 213 |
|
| 214 |
#define atomic32_add_if_le( v, i, t ) \ |
| 215 |
( atomic32_read_and_add_le ( v, i, t ) <= ( t ) ) |
| 216 |
|
| 217 |
static __inline__ |
| 218 |
int atomic32_read_and_add_eq ( atomic32_t *v, int i, int t ) |
| 219 |
{ |
| 220 |
int rtn, sum; |
| 221 |
__asm__ __volatile__ |
| 222 |
( |
| 223 |
"mov (%2), %0;" |
| 224 |
"1:" |
| 225 |
"cmp %4, %0;" |
| 226 |
"mov %3, %1;" |
| 227 |
"jne 2f;" |
| 228 |
"add %0, %1;" |
| 229 |
"lock;" |
| 230 |
"cmpxchg %1, (%2);" |
| 231 |
"jne 1b;" |
| 232 |
"2:" |
| 233 |
: "=&a" ( rtn ), "=&r" ( sum ) |
| 234 |
: "r" ( & v -> counter ), "r" ( i ), "r" ( t ) |
| 235 |
); |
| 236 |
return rtn; |
| 237 |
} |
| 238 |
|
| 239 |
#define atomic32_add_if_eq( v, i, t ) \ |
| 240 |
( atomic32_read_and_add_eq ( v, i, t ) == ( t ) ) |
| 241 |
|
| 242 |
static __inline__ |
| 243 |
int atomic32_read_and_add_ne ( atomic32_t *v, int i, int t ) |
| 244 |
{ |
| 245 |
int rtn, sum; |
| 246 |
__asm__ __volatile__ |
| 247 |
( |
| 248 |
"mov (%2), %0;" |
| 249 |
"1:" |
| 250 |
"cmp %4, %0;" |
| 251 |
"mov %3, %1;" |
| 252 |
"je 2f;" |
| 253 |
"add %0, %1;" |
| 254 |
"lock;" |
| 255 |
"cmpxchg %1, (%2);" |
| 256 |
"jne 1b;" |
| 257 |
"2:" |
| 258 |
: "=&a" ( rtn ), "=&r" ( sum ) |
| 259 |
: "r" ( & v -> counter ), "r" ( i ), "r" ( t ) |
| 260 |
); |
| 261 |
return rtn; |
| 262 |
} |
| 263 |
|
| 264 |
#define atomic32_add_if_ne( v, i, t ) \ |
| 265 |
( atomic32_read_and_add_ne ( v, i, t ) != ( t ) ) |
| 266 |
|
| 267 |
static __inline__ |
| 268 |
int atomic32_read_and_add_ge ( atomic32_t *v, int i, int t ) |
| 269 |
{ |
| 270 |
int rtn, sum; |
| 271 |
__asm__ __volatile__ |
| 272 |
( |
| 273 |
"mov (%2), %0;" |
| 274 |
"1:" |
| 275 |
"cmp %4, %0;" |
| 276 |
"mov %3, %1;" |
| 277 |
"jl 2f;" |
| 278 |
"add %0, %1;" |
| 279 |
"lock;" |
| 280 |
"cmpxchg %1, (%2);" |
| 281 |
"jne 1b;" |
| 282 |
"2:" |
| 283 |
: "=&a" ( rtn ), "=&r" ( sum ) |
| 284 |
: "r" ( & v -> counter ), "r" ( i ), "r" ( t ) |
| 285 |
); |
| 286 |
return rtn; |
| 287 |
} |
| 288 |
|
| 289 |
#define atomic32_add_if_ge( v, i, t ) \ |
| 290 |
( atomic32_read_and_add_ge ( v, i, t ) >= ( t ) ) |
| 291 |
|
| 292 |
static __inline__ |
| 293 |
int atomic32_read_and_add_gt ( atomic32_t *v, int i, int t ) |
| 294 |
{ |
| 295 |
int rtn, sum; |
| 296 |
__asm__ __volatile__ |
| 297 |
( |
| 298 |
"mov (%2), %0;" |
| 299 |
"1:" |
| 300 |
"cmp %4, %0;" |
| 301 |
"mov %3, %1;" |
| 302 |
"jle 2f;" |
| 303 |
"add %0, %1;" |
| 304 |
"lock;" |
| 305 |
"cmpxchg %1, (%2);" |
| 306 |
"jne 1b;" |
| 307 |
"2:" |
| 308 |
: "=&a" ( rtn ), "=&r" ( sum ) |
| 309 |
: "r" ( & v -> counter ), "r" ( i ), "r" ( t ) |
| 310 |
); |
| 311 |
return rtn; |
| 312 |
} |
| 313 |
|
| 314 |
#define atomic32_add_if_gt( v, i, t ) \ |
| 315 |
( atomic32_read_and_add_gt ( v, i, t ) > ( t ) ) |
| 316 |
|
| 317 |
static __inline__ |
| 318 |
int atomic32_read_and_add_odd ( atomic32_t *v, int i ) |
| 319 |
{ |
| 320 |
int rtn, sum; |
| 321 |
__asm__ __volatile__ |
| 322 |
( |
| 323 |
"mov (%2), %0;" |
| 324 |
"1:" |
| 325 |
"bt $0, %0;" |
| 326 |
"mov %3, %1;" |
| 327 |
"jnc 2f;" |
| 328 |
"add %0, %1;" |
| 329 |
"lock;" |
| 330 |
"cmpxchg %1, (%2);" |
| 331 |
"jne 1b;" |
| 332 |
"2:" |
| 333 |
: "=&a" ( rtn ), "=&r" ( sum ) |
| 334 |
: "r" ( & v -> counter ), "r" ( i ) |
| 335 |
); |
| 336 |
return rtn; |
| 337 |
} |
| 338 |
|
| 339 |
static __inline__ |
| 340 |
int atomic32_read_and_add_even ( atomic32_t *v, int i ) |
| 341 |
{ |
| 342 |
int rtn, sum; |
| 343 |
__asm__ __volatile__ |
| 344 |
( |
| 345 |
"mov (%2), %0;" |
| 346 |
"1:" |
| 347 |
"bt $0, %0;" |
| 348 |
"mov %3, %1;" |
| 349 |
"jc 2f;" |
| 350 |
"add %0, %1;" |
| 351 |
"lock;" |
| 352 |
"cmpxchg %1, (%2);" |
| 353 |
"jne 1b;" |
| 354 |
"2:" |
| 355 |
: "=&a" ( rtn ), "=&r" ( sum ) |
| 356 |
: "r" ( & v -> counter ), "r" ( i ) |
| 357 |
); |
| 358 |
return rtn; |
| 359 |
} |
| 360 |
|
| 361 |
|
| 362 |
#ifdef __cplusplus |
| 363 |
} |
| 364 |
#endif |
| 365 |
|
| 366 |
#endif /* _h_atomic32_ */ |