bitops.h (18510B)
1/* 2 * Bitops Module 3 * 4 * Copyright (C) 2010 Corentin Chary <corentin.chary@gmail.com> 5 * 6 * Mostly inspired by (stolen from) linux/bitmap.h and linux/bitops.h 7 * 8 * This work is licensed under the terms of the GNU LGPL, version 2.1 or later. 9 * See the COPYING.LIB file in the top-level directory. 10 */ 11 12#ifndef BITOPS_H 13#define BITOPS_H 14 15 16#include "host-utils.h" 17#include "atomic.h" 18 19#define BITS_PER_BYTE CHAR_BIT 20#define BITS_PER_LONG (sizeof (unsigned long) * BITS_PER_BYTE) 21 22#define BIT(nr) (1UL << (nr)) 23#define BIT_ULL(nr) (1ULL << (nr)) 24#define BIT_MASK(nr) (1UL << ((nr) % BITS_PER_LONG)) 25#define BIT_WORD(nr) ((nr) / BITS_PER_LONG) 26#define BITS_TO_LONGS(nr) DIV_ROUND_UP(nr, BITS_PER_BYTE * sizeof(long)) 27 28#define MAKE_64BIT_MASK(shift, length) \ 29 (((~0ULL) >> (64 - (length))) << (shift)) 30 31/** 32 * set_bit - Set a bit in memory 33 * @nr: the bit to set 34 * @addr: the address to start counting from 35 */ 36static inline void set_bit(long nr, unsigned long *addr) 37{ 38 unsigned long mask = BIT_MASK(nr); 39 unsigned long *p = addr + BIT_WORD(nr); 40 41 *p |= mask; 42} 43 44/** 45 * set_bit_atomic - Set a bit in memory atomically 46 * @nr: the bit to set 47 * @addr: the address to start counting from 48 */ 49static inline void set_bit_atomic(long nr, unsigned long *addr) 50{ 51 unsigned long mask = BIT_MASK(nr); 52 unsigned long *p = addr + BIT_WORD(nr); 53 54 qatomic_or(p, mask); 55} 56 57/** 58 * clear_bit - Clears a bit in memory 59 * @nr: Bit to clear 60 * @addr: Address to start counting from 61 */ 62static inline void clear_bit(long nr, unsigned long *addr) 63{ 64 unsigned long mask = BIT_MASK(nr); 65 unsigned long *p = addr + BIT_WORD(nr); 66 67 *p &= ~mask; 68} 69 70/** 71 * change_bit - Toggle a bit in memory 72 * @nr: Bit to change 73 * @addr: Address to start counting from 74 */ 75static inline void change_bit(long nr, unsigned long *addr) 76{ 77 unsigned long mask = BIT_MASK(nr); 78 unsigned long *p = addr + BIT_WORD(nr); 79 80 *p ^= mask; 81} 82 83/** 84 * test_and_set_bit - Set a bit and return its old value 85 * @nr: Bit to set 86 * @addr: Address to count from 87 */ 88static inline int test_and_set_bit(long nr, unsigned long *addr) 89{ 90 unsigned long mask = BIT_MASK(nr); 91 unsigned long *p = addr + BIT_WORD(nr); 92 unsigned long old = *p; 93 94 *p = old | mask; 95 return (old & mask) != 0; 96} 97 98/** 99 * test_and_clear_bit - Clear a bit and return its old value 100 * @nr: Bit to clear 101 * @addr: Address to count from 102 */ 103static inline int test_and_clear_bit(long nr, unsigned long *addr) 104{ 105 unsigned long mask = BIT_MASK(nr); 106 unsigned long *p = addr + BIT_WORD(nr); 107 unsigned long old = *p; 108 109 *p = old & ~mask; 110 return (old & mask) != 0; 111} 112 113/** 114 * test_and_change_bit - Change a bit and return its old value 115 * @nr: Bit to change 116 * @addr: Address to count from 117 */ 118static inline int test_and_change_bit(long nr, unsigned long *addr) 119{ 120 unsigned long mask = BIT_MASK(nr); 121 unsigned long *p = addr + BIT_WORD(nr); 122 unsigned long old = *p; 123 124 *p = old ^ mask; 125 return (old & mask) != 0; 126} 127 128/** 129 * test_bit - Determine whether a bit is set 130 * @nr: bit number to test 131 * @addr: Address to start counting from 132 */ 133static inline int test_bit(long nr, const unsigned long *addr) 134{ 135 return 1UL & (addr[BIT_WORD(nr)] >> (nr & (BITS_PER_LONG-1))); 136} 137 138/** 139 * find_last_bit - find the last set bit in a memory region 140 * @addr: The address to start the search at 141 * @size: The maximum size to search 142 * 143 * Returns the bit number of the last set bit, 144 * or @size if there is no set bit in the bitmap. 145 */ 146unsigned long find_last_bit(const unsigned long *addr, 147 unsigned long size); 148 149/** 150 * find_next_bit - find the next set bit in a memory region 151 * @addr: The address to base the search on 152 * @offset: The bitnumber to start searching at 153 * @size: The bitmap size in bits 154 * 155 * Returns the bit number of the next set bit, 156 * or @size if there are no further set bits in the bitmap. 157 */ 158unsigned long find_next_bit(const unsigned long *addr, 159 unsigned long size, 160 unsigned long offset); 161 162/** 163 * find_next_zero_bit - find the next cleared bit in a memory region 164 * @addr: The address to base the search on 165 * @offset: The bitnumber to start searching at 166 * @size: The bitmap size in bits 167 * 168 * Returns the bit number of the next cleared bit, 169 * or @size if there are no further clear bits in the bitmap. 170 */ 171 172unsigned long find_next_zero_bit(const unsigned long *addr, 173 unsigned long size, 174 unsigned long offset); 175 176/** 177 * find_first_bit - find the first set bit in a memory region 178 * @addr: The address to start the search at 179 * @size: The maximum size to search 180 * 181 * Returns the bit number of the first set bit, 182 * or @size if there is no set bit in the bitmap. 183 */ 184static inline unsigned long find_first_bit(const unsigned long *addr, 185 unsigned long size) 186{ 187 unsigned long result, tmp; 188 189 for (result = 0; result < size; result += BITS_PER_LONG) { 190 tmp = *addr++; 191 if (tmp) { 192 result += ctzl(tmp); 193 return result < size ? result : size; 194 } 195 } 196 /* Not found */ 197 return size; 198} 199 200/** 201 * find_first_zero_bit - find the first cleared bit in a memory region 202 * @addr: The address to start the search at 203 * @size: The maximum size to search 204 * 205 * Returns the bit number of the first cleared bit, 206 * or @size if there is no clear bit in the bitmap. 207 */ 208static inline unsigned long find_first_zero_bit(const unsigned long *addr, 209 unsigned long size) 210{ 211 return find_next_zero_bit(addr, size, 0); 212} 213 214/** 215 * rol8 - rotate an 8-bit value left 216 * @word: value to rotate 217 * @shift: bits to roll 218 */ 219static inline uint8_t rol8(uint8_t word, unsigned int shift) 220{ 221 return (word << shift) | (word >> ((8 - shift) & 7)); 222} 223 224/** 225 * ror8 - rotate an 8-bit value right 226 * @word: value to rotate 227 * @shift: bits to roll 228 */ 229static inline uint8_t ror8(uint8_t word, unsigned int shift) 230{ 231 return (word >> shift) | (word << ((8 - shift) & 7)); 232} 233 234/** 235 * rol16 - rotate a 16-bit value left 236 * @word: value to rotate 237 * @shift: bits to roll 238 */ 239static inline uint16_t rol16(uint16_t word, unsigned int shift) 240{ 241 return (word << shift) | (word >> ((16 - shift) & 15)); 242} 243 244/** 245 * ror16 - rotate a 16-bit value right 246 * @word: value to rotate 247 * @shift: bits to roll 248 */ 249static inline uint16_t ror16(uint16_t word, unsigned int shift) 250{ 251 return (word >> shift) | (word << ((16 - shift) & 15)); 252} 253 254/** 255 * rol32 - rotate a 32-bit value left 256 * @word: value to rotate 257 * @shift: bits to roll 258 */ 259static inline uint32_t rol32(uint32_t word, unsigned int shift) 260{ 261 return (word << shift) | (word >> ((32 - shift) & 31)); 262} 263 264/** 265 * ror32 - rotate a 32-bit value right 266 * @word: value to rotate 267 * @shift: bits to roll 268 */ 269static inline uint32_t ror32(uint32_t word, unsigned int shift) 270{ 271 return (word >> shift) | (word << ((32 - shift) & 31)); 272} 273 274/** 275 * rol64 - rotate a 64-bit value left 276 * @word: value to rotate 277 * @shift: bits to roll 278 */ 279static inline uint64_t rol64(uint64_t word, unsigned int shift) 280{ 281 return (word << shift) | (word >> ((64 - shift) & 63)); 282} 283 284/** 285 * ror64 - rotate a 64-bit value right 286 * @word: value to rotate 287 * @shift: bits to roll 288 */ 289static inline uint64_t ror64(uint64_t word, unsigned int shift) 290{ 291 return (word >> shift) | (word << ((64 - shift) & 63)); 292} 293 294/** 295 * hswap32 - swap 16-bit halfwords within a 32-bit value 296 * @h: value to swap 297 */ 298static inline uint32_t hswap32(uint32_t h) 299{ 300 return rol32(h, 16); 301} 302 303/** 304 * hswap64 - swap 16-bit halfwords within a 64-bit value 305 * @h: value to swap 306 */ 307static inline uint64_t hswap64(uint64_t h) 308{ 309 uint64_t m = 0x0000ffff0000ffffull; 310 h = rol64(h, 32); 311 return ((h & m) << 16) | ((h >> 16) & m); 312} 313 314/** 315 * wswap64 - swap 32-bit words within a 64-bit value 316 * @h: value to swap 317 */ 318static inline uint64_t wswap64(uint64_t h) 319{ 320 return rol64(h, 32); 321} 322 323/** 324 * extract32: 325 * @value: the value to extract the bit field from 326 * @start: the lowest bit in the bit field (numbered from 0) 327 * @length: the length of the bit field 328 * 329 * Extract from the 32 bit input @value the bit field specified by the 330 * @start and @length parameters, and return it. The bit field must 331 * lie entirely within the 32 bit word. It is valid to request that 332 * all 32 bits are returned (ie @length 32 and @start 0). 333 * 334 * Returns: the value of the bit field extracted from the input value. 335 */ 336static inline uint32_t extract32(uint32_t value, int start, int length) 337{ 338 assert(start >= 0 && length > 0 && length <= 32 - start); 339 return (value >> start) & (~0U >> (32 - length)); 340} 341 342/** 343 * extract8: 344 * @value: the value to extract the bit field from 345 * @start: the lowest bit in the bit field (numbered from 0) 346 * @length: the length of the bit field 347 * 348 * Extract from the 8 bit input @value the bit field specified by the 349 * @start and @length parameters, and return it. The bit field must 350 * lie entirely within the 8 bit word. It is valid to request that 351 * all 8 bits are returned (ie @length 8 and @start 0). 352 * 353 * Returns: the value of the bit field extracted from the input value. 354 */ 355static inline uint8_t extract8(uint8_t value, int start, int length) 356{ 357 assert(start >= 0 && length > 0 && length <= 8 - start); 358 return extract32(value, start, length); 359} 360 361/** 362 * extract16: 363 * @value: the value to extract the bit field from 364 * @start: the lowest bit in the bit field (numbered from 0) 365 * @length: the length of the bit field 366 * 367 * Extract from the 16 bit input @value the bit field specified by the 368 * @start and @length parameters, and return it. The bit field must 369 * lie entirely within the 16 bit word. It is valid to request that 370 * all 16 bits are returned (ie @length 16 and @start 0). 371 * 372 * Returns: the value of the bit field extracted from the input value. 373 */ 374static inline uint16_t extract16(uint16_t value, int start, int length) 375{ 376 assert(start >= 0 && length > 0 && length <= 16 - start); 377 return extract32(value, start, length); 378} 379 380/** 381 * extract64: 382 * @value: the value to extract the bit field from 383 * @start: the lowest bit in the bit field (numbered from 0) 384 * @length: the length of the bit field 385 * 386 * Extract from the 64 bit input @value the bit field specified by the 387 * @start and @length parameters, and return it. The bit field must 388 * lie entirely within the 64 bit word. It is valid to request that 389 * all 64 bits are returned (ie @length 64 and @start 0). 390 * 391 * Returns: the value of the bit field extracted from the input value. 392 */ 393static inline uint64_t extract64(uint64_t value, int start, int length) 394{ 395 assert(start >= 0 && length > 0 && length <= 64 - start); 396 return (value >> start) & (~0ULL >> (64 - length)); 397} 398 399/** 400 * sextract32: 401 * @value: the value to extract the bit field from 402 * @start: the lowest bit in the bit field (numbered from 0) 403 * @length: the length of the bit field 404 * 405 * Extract from the 32 bit input @value the bit field specified by the 406 * @start and @length parameters, and return it, sign extended to 407 * an int32_t (ie with the most significant bit of the field propagated 408 * to all the upper bits of the return value). The bit field must lie 409 * entirely within the 32 bit word. It is valid to request that 410 * all 32 bits are returned (ie @length 32 and @start 0). 411 * 412 * Returns: the sign extended value of the bit field extracted from the 413 * input value. 414 */ 415static inline int32_t sextract32(uint32_t value, int start, int length) 416{ 417 assert(start >= 0 && length > 0 && length <= 32 - start); 418 /* Note that this implementation relies on right shift of signed 419 * integers being an arithmetic shift. 420 */ 421 return ((int32_t)(value << (32 - length - start))) >> (32 - length); 422} 423 424/** 425 * sextract64: 426 * @value: the value to extract the bit field from 427 * @start: the lowest bit in the bit field (numbered from 0) 428 * @length: the length of the bit field 429 * 430 * Extract from the 64 bit input @value the bit field specified by the 431 * @start and @length parameters, and return it, sign extended to 432 * an int64_t (ie with the most significant bit of the field propagated 433 * to all the upper bits of the return value). The bit field must lie 434 * entirely within the 64 bit word. It is valid to request that 435 * all 64 bits are returned (ie @length 64 and @start 0). 436 * 437 * Returns: the sign extended value of the bit field extracted from the 438 * input value. 439 */ 440static inline int64_t sextract64(uint64_t value, int start, int length) 441{ 442 assert(start >= 0 && length > 0 && length <= 64 - start); 443 /* Note that this implementation relies on right shift of signed 444 * integers being an arithmetic shift. 445 */ 446 return ((int64_t)(value << (64 - length - start))) >> (64 - length); 447} 448 449/** 450 * deposit32: 451 * @value: initial value to insert bit field into 452 * @start: the lowest bit in the bit field (numbered from 0) 453 * @length: the length of the bit field 454 * @fieldval: the value to insert into the bit field 455 * 456 * Deposit @fieldval into the 32 bit @value at the bit field specified 457 * by the @start and @length parameters, and return the modified 458 * @value. Bits of @value outside the bit field are not modified. 459 * Bits of @fieldval above the least significant @length bits are 460 * ignored. The bit field must lie entirely within the 32 bit word. 461 * It is valid to request that all 32 bits are modified (ie @length 462 * 32 and @start 0). 463 * 464 * Returns: the modified @value. 465 */ 466static inline uint32_t deposit32(uint32_t value, int start, int length, 467 uint32_t fieldval) 468{ 469 uint32_t mask; 470 assert(start >= 0 && length > 0 && length <= 32 - start); 471 mask = (~0U >> (32 - length)) << start; 472 return (value & ~mask) | ((fieldval << start) & mask); 473} 474 475/** 476 * deposit64: 477 * @value: initial value to insert bit field into 478 * @start: the lowest bit in the bit field (numbered from 0) 479 * @length: the length of the bit field 480 * @fieldval: the value to insert into the bit field 481 * 482 * Deposit @fieldval into the 64 bit @value at the bit field specified 483 * by the @start and @length parameters, and return the modified 484 * @value. Bits of @value outside the bit field are not modified. 485 * Bits of @fieldval above the least significant @length bits are 486 * ignored. The bit field must lie entirely within the 64 bit word. 487 * It is valid to request that all 64 bits are modified (ie @length 488 * 64 and @start 0). 489 * 490 * Returns: the modified @value. 491 */ 492static inline uint64_t deposit64(uint64_t value, int start, int length, 493 uint64_t fieldval) 494{ 495 uint64_t mask; 496 assert(start >= 0 && length > 0 && length <= 64 - start); 497 mask = (~0ULL >> (64 - length)) << start; 498 return (value & ~mask) | ((fieldval << start) & mask); 499} 500 501/** 502 * half_shuffle32: 503 * @x: 32-bit value (of which only the bottom 16 bits are of interest) 504 * 505 * Given an input value:: 506 * 507 * xxxx xxxx xxxx xxxx ABCD EFGH IJKL MNOP 508 * 509 * return the value where the bottom 16 bits are spread out into 510 * the odd bits in the word, and the even bits are zeroed:: 511 * 512 * 0A0B 0C0D 0E0F 0G0H 0I0J 0K0L 0M0N 0O0P 513 * 514 * Any bits set in the top half of the input are ignored. 515 * 516 * Returns: the shuffled bits. 517 */ 518static inline uint32_t half_shuffle32(uint32_t x) 519{ 520 /* This algorithm is from _Hacker's Delight_ section 7-2 "Shuffling Bits". 521 * It ignores any bits set in the top half of the input. 522 */ 523 x = ((x & 0xFF00) << 8) | (x & 0x00FF); 524 x = ((x << 4) | x) & 0x0F0F0F0F; 525 x = ((x << 2) | x) & 0x33333333; 526 x = ((x << 1) | x) & 0x55555555; 527 return x; 528} 529 530/** 531 * half_shuffle64: 532 * @x: 64-bit value (of which only the bottom 32 bits are of interest) 533 * 534 * Given an input value:: 535 * 536 * xxxx xxxx xxxx .... xxxx xxxx ABCD EFGH IJKL MNOP QRST UVWX YZab cdef 537 * 538 * return the value where the bottom 32 bits are spread out into 539 * the odd bits in the word, and the even bits are zeroed:: 540 * 541 * 0A0B 0C0D 0E0F 0G0H 0I0J 0K0L 0M0N .... 0U0V 0W0X 0Y0Z 0a0b 0c0d 0e0f 542 * 543 * Any bits set in the top half of the input are ignored. 544 * 545 * Returns: the shuffled bits. 546 */ 547static inline uint64_t half_shuffle64(uint64_t x) 548{ 549 /* This algorithm is from _Hacker's Delight_ section 7-2 "Shuffling Bits". 550 * It ignores any bits set in the top half of the input. 551 */ 552 x = ((x & 0xFFFF0000ULL) << 16) | (x & 0xFFFF); 553 x = ((x << 8) | x) & 0x00FF00FF00FF00FFULL; 554 x = ((x << 4) | x) & 0x0F0F0F0F0F0F0F0FULL; 555 x = ((x << 2) | x) & 0x3333333333333333ULL; 556 x = ((x << 1) | x) & 0x5555555555555555ULL; 557 return x; 558} 559 560/** 561 * half_unshuffle32: 562 * @x: 32-bit value (of which only the odd bits are of interest) 563 * 564 * Given an input value:: 565 * 566 * xAxB xCxD xExF xGxH xIxJ xKxL xMxN xOxP 567 * 568 * return the value where all the odd bits are compressed down 569 * into the low half of the word, and the high half is zeroed:: 570 * 571 * 0000 0000 0000 0000 ABCD EFGH IJKL MNOP 572 * 573 * Any even bits set in the input are ignored. 574 * 575 * Returns: the unshuffled bits. 576 */ 577static inline uint32_t half_unshuffle32(uint32_t x) 578{ 579 /* This algorithm is from _Hacker's Delight_ section 7-2 "Shuffling Bits". 580 * where it is called an inverse half shuffle. 581 */ 582 x &= 0x55555555; 583 x = ((x >> 1) | x) & 0x33333333; 584 x = ((x >> 2) | x) & 0x0F0F0F0F; 585 x = ((x >> 4) | x) & 0x00FF00FF; 586 x = ((x >> 8) | x) & 0x0000FFFF; 587 return x; 588} 589 590/** 591 * half_unshuffle64: 592 * @x: 64-bit value (of which only the odd bits are of interest) 593 * 594 * Given an input value:: 595 * 596 * xAxB xCxD xExF xGxH xIxJ xKxL xMxN .... xUxV xWxX xYxZ xaxb xcxd xexf 597 * 598 * return the value where all the odd bits are compressed down 599 * into the low half of the word, and the high half is zeroed:: 600 * 601 * 0000 0000 0000 .... 0000 0000 ABCD EFGH IJKL MNOP QRST UVWX YZab cdef 602 * 603 * Any even bits set in the input are ignored. 604 * 605 * Returns: the unshuffled bits. 606 */ 607static inline uint64_t half_unshuffle64(uint64_t x) 608{ 609 /* This algorithm is from _Hacker's Delight_ section 7-2 "Shuffling Bits". 610 * where it is called an inverse half shuffle. 611 */ 612 x &= 0x5555555555555555ULL; 613 x = ((x >> 1) | x) & 0x3333333333333333ULL; 614 x = ((x >> 2) | x) & 0x0F0F0F0F0F0F0F0FULL; 615 x = ((x >> 4) | x) & 0x00FF00FF00FF00FFULL; 616 x = ((x >> 8) | x) & 0x0000FFFF0000FFFFULL; 617 x = ((x >> 16) | x) & 0x00000000FFFFFFFFULL; 618 return x; 619} 620 621#endif