1 /////////////////////////////////////////////////////////////////////////////// 2 // 3 /// \file tuklib_integer.h 4 /// \brief Various integer and bit operations 5 /// 6 /// This file provides macros or functions to do some basic integer and bit 7 /// operations. 8 /// 9 /// Native endian inline functions (XX = 16, 32, or 64): 10 /// - Unaligned native endian reads: readXXne(ptr) 11 /// - Unaligned native endian writes: writeXXne(ptr, num) 12 /// - Aligned native endian reads: aligned_readXXne(ptr) 13 /// - Aligned native endian writes: aligned_writeXXne(ptr, num) 14 /// 15 /// Endianness-converting integer operations (these can be macros!) 16 /// (XX = 16, 32, or 64; Y = b or l): 17 /// - Byte swapping: bswapXX(num) 18 /// - Byte order conversions to/from native (byteswaps if Y isn't 19 /// the native endianness): convXXYe(num) 20 /// - Unaligned reads (16/32-bit only): readXXYe(ptr) 21 /// - Unaligned writes (16/32-bit only): writeXXYe(ptr, num) 22 /// - Aligned reads: aligned_readXXYe(ptr) 23 /// - Aligned writes: aligned_writeXXYe(ptr, num) 24 /// 25 /// Since the above can macros, the arguments should have no side effects 26 /// because they may be evaluated more than once. 27 /// 28 /// Bit scan operations for non-zero 32-bit integers (inline functions): 29 /// - Bit scan reverse (find highest non-zero bit): bsr32(num) 30 /// - Count leading zeros: clz32(num) 31 /// - Count trailing zeros: ctz32(num) 32 /// - Bit scan forward (simply an alias for ctz32()): bsf32(num) 33 /// 34 /// The above bit scan operations return 0-31. If num is zero, 35 /// the result is undefined. 36 // 37 // Authors: Lasse Collin 38 // Joachim Henke 39 // 40 // This file has been put into the public domain. 41 // You can do whatever you want with this file. 42 // 43 /////////////////////////////////////////////////////////////////////////////// 44 45 #ifndef TUKLIB_INTEGER_H 46 #define TUKLIB_INTEGER_H 47 48 #include "tuklib_common.h" 49 #include <string.h> 50 51 // Newer Intel C compilers require immintrin.h for _bit_scan_reverse() 52 // and such functions. 53 #if defined(__INTEL_COMPILER) && (__INTEL_COMPILER >= 1500) 54 # include <immintrin.h> 55 #endif 56 57 58 /////////////////// 59 // Byte swapping // 60 /////////////////// 61 62 #if defined(HAVE___BUILTIN_BSWAPXX) 63 // GCC >= 4.8 and Clang 64 # define bswap16(n) __builtin_bswap16(n) 65 # define bswap32(n) __builtin_bswap32(n) 66 # define bswap64(n) __builtin_bswap64(n) 67 68 #elif defined(HAVE_BYTESWAP_H) 69 // glibc, uClibc, dietlibc 70 # include <byteswap.h> 71 # ifdef HAVE_BSWAP_16 72 # define bswap16(num) bswap_16(num) 73 # endif 74 # ifdef HAVE_BSWAP_32 75 # define bswap32(num) bswap_32(num) 76 # endif 77 # ifdef HAVE_BSWAP_64 78 # define bswap64(num) bswap_64(num) 79 # endif 80 81 #elif defined(HAVE_SYS_ENDIAN_H) 82 // *BSDs and Darwin 83 # include <sys/endian.h> 84 85 #elif defined(HAVE_SYS_BYTEORDER_H) 86 // Solaris 87 # include <sys/byteorder.h> 88 # ifdef BSWAP_16 89 # define bswap16(num) BSWAP_16(num) 90 # endif 91 # ifdef BSWAP_32 92 # define bswap32(num) BSWAP_32(num) 93 # endif 94 # ifdef BSWAP_64 95 # define bswap64(num) BSWAP_64(num) 96 # endif 97 # ifdef BE_16 98 # define conv16be(num) BE_16(num) 99 # endif 100 # ifdef BE_32 101 # define conv32be(num) BE_32(num) 102 # endif 103 # ifdef BE_64 104 # define conv64be(num) BE_64(num) 105 # endif 106 # ifdef LE_16 107 # define conv16le(num) LE_16(num) 108 # endif 109 # ifdef LE_32 110 # define conv32le(num) LE_32(num) 111 # endif 112 # ifdef LE_64 113 # define conv64le(num) LE_64(num) 114 # endif 115 #endif 116 117 #ifndef bswap16 118 # define bswap16(n) (uint16_t)( \ 119 (((n) & 0x00FFU) << 8) \ 120 | (((n) & 0xFF00U) >> 8) \ 121 ) 122 #endif 123 124 #ifndef bswap32 125 # define bswap32(n) (uint32_t)( \ 126 (((n) & UINT32_C(0x000000FF)) << 24) \ 127 | (((n) & UINT32_C(0x0000FF00)) << 8) \ 128 | (((n) & UINT32_C(0x00FF0000)) >> 8) \ 129 | (((n) & UINT32_C(0xFF000000)) >> 24) \ 130 ) 131 #endif 132 133 #ifndef bswap64 134 # define bswap64(n) (uint64_t)( \ 135 (((n) & UINT64_C(0x00000000000000FF)) << 56) \ 136 | (((n) & UINT64_C(0x000000000000FF00)) << 40) \ 137 | (((n) & UINT64_C(0x0000000000FF0000)) << 24) \ 138 | (((n) & UINT64_C(0x00000000FF000000)) << 8) \ 139 | (((n) & UINT64_C(0x000000FF00000000)) >> 8) \ 140 | (((n) & UINT64_C(0x0000FF0000000000)) >> 24) \ 141 | (((n) & UINT64_C(0x00FF000000000000)) >> 40) \ 142 | (((n) & UINT64_C(0xFF00000000000000)) >> 56) \ 143 ) 144 #endif 145 146 // Define conversion macros using the basic byte swapping macros. 147 #ifdef WORDS_BIGENDIAN 148 # ifndef conv16be 149 # define conv16be(num) ((uint16_t)(num)) 150 # endif 151 # ifndef conv32be 152 # define conv32be(num) ((uint32_t)(num)) 153 # endif 154 # ifndef conv64be 155 # define conv64be(num) ((uint64_t)(num)) 156 # endif 157 # ifndef conv16le 158 # define conv16le(num) bswap16(num) 159 # endif 160 # ifndef conv32le 161 # define conv32le(num) bswap32(num) 162 # endif 163 # ifndef conv64le 164 # define conv64le(num) bswap64(num) 165 # endif 166 #else 167 # ifndef conv16be 168 # define conv16be(num) bswap16(num) 169 # endif 170 # ifndef conv32be 171 # define conv32be(num) bswap32(num) 172 # endif 173 # ifndef conv64be 174 # define conv64be(num) bswap64(num) 175 # endif 176 # ifndef conv16le 177 # define conv16le(num) ((uint16_t)(num)) 178 # endif 179 # ifndef conv32le 180 # define conv32le(num) ((uint32_t)(num)) 181 # endif 182 # ifndef conv64le 183 # define conv64le(num) ((uint64_t)(num)) 184 # endif 185 #endif 186 187 188 //////////////////////////////// 189 // Unaligned reads and writes // 190 //////////////////////////////// 191 192 // The traditional way of casting e.g. *(const uint16_t *)uint8_pointer 193 // is bad even if the uint8_pointer is properly aligned because this kind 194 // of casts break strict aliasing rules and result in undefined behavior. 195 // With unaligned pointers it's even worse: compilers may emit vector 196 // instructions that require aligned pointers even if non-vector 197 // instructions work with unaligned pointers. 198 // 199 // Using memcpy() is the standard compliant way to do unaligned access. 200 // Many modern compilers inline it so there is no function call overhead. 201 // For those compilers that don't handle the memcpy() method well, the 202 // old casting method (that violates strict aliasing) can be requested at 203 // build time. A third method, casting to a packed struct, would also be 204 // an option but isn't provided to keep things simpler (it's already a mess). 205 // Hopefully this is flexible enough in practice. 206 207 static inline uint16_t 208 read16ne(const uint8_t *buf) 209 { 210 #if defined(TUKLIB_FAST_UNALIGNED_ACCESS) \ 211 && defined(TUKLIB_USE_UNSAFE_TYPE_PUNNING) 212 return *(const uint16_t *)buf; 213 #else 214 uint16_t num; 215 memcpy(&num, buf, sizeof(num)); 216 return num; 217 #endif 218 } 219 220 221 static inline uint32_t 222 read32ne(const uint8_t *buf) 223 { 224 #if defined(TUKLIB_FAST_UNALIGNED_ACCESS) \ 225 && defined(TUKLIB_USE_UNSAFE_TYPE_PUNNING) 226 return *(const uint32_t *)buf; 227 #else 228 uint32_t num; 229 memcpy(&num, buf, sizeof(num)); 230 return num; 231 #endif 232 } 233 234 235 static inline uint64_t 236 read64ne(const uint8_t *buf) 237 { 238 #if defined(TUKLIB_FAST_UNALIGNED_ACCESS) \ 239 && defined(TUKLIB_USE_UNSAFE_TYPE_PUNNING) 240 return *(const uint64_t *)buf; 241 #else 242 uint64_t num; 243 memcpy(&num, buf, sizeof(num)); 244 return num; 245 #endif 246 } 247 248 249 static inline void 250 write16ne(uint8_t *buf, uint16_t num) 251 { 252 #if defined(TUKLIB_FAST_UNALIGNED_ACCESS) \ 253 && defined(TUKLIB_USE_UNSAFE_TYPE_PUNNING) 254 *(uint16_t *)buf = num; 255 #else 256 memcpy(buf, &num, sizeof(num)); 257 #endif 258 return; 259 } 260 261 262 static inline void 263 write32ne(uint8_t *buf, uint32_t num) 264 { 265 #if defined(TUKLIB_FAST_UNALIGNED_ACCESS) \ 266 && defined(TUKLIB_USE_UNSAFE_TYPE_PUNNING) 267 *(uint32_t *)buf = num; 268 #else 269 memcpy(buf, &num, sizeof(num)); 270 #endif 271 return; 272 } 273 274 275 static inline void 276 write64ne(uint8_t *buf, uint64_t num) 277 { 278 #if defined(TUKLIB_FAST_UNALIGNED_ACCESS) \ 279 && defined(TUKLIB_USE_UNSAFE_TYPE_PUNNING) 280 *(uint64_t *)buf = num; 281 #else 282 memcpy(buf, &num, sizeof(num)); 283 #endif 284 return; 285 } 286 287 288 static inline uint16_t 289 read16be(const uint8_t *buf) 290 { 291 #if defined(WORDS_BIGENDIAN) || defined(TUKLIB_FAST_UNALIGNED_ACCESS) 292 uint16_t num = read16ne(buf); 293 return conv16be(num); 294 #else 295 uint16_t num = ((uint16_t)buf[0] << 8) | (uint16_t)buf[1]; 296 return num; 297 #endif 298 } 299 300 301 static inline uint16_t 302 read16le(const uint8_t *buf) 303 { 304 #if !defined(WORDS_BIGENDIAN) || defined(TUKLIB_FAST_UNALIGNED_ACCESS) 305 uint16_t num = read16ne(buf); 306 return conv16le(num); 307 #else 308 uint16_t num = ((uint16_t)buf[0]) | ((uint16_t)buf[1] << 8); 309 return num; 310 #endif 311 } 312 313 314 static inline uint32_t 315 read32be(const uint8_t *buf) 316 { 317 #if defined(WORDS_BIGENDIAN) || defined(TUKLIB_FAST_UNALIGNED_ACCESS) 318 uint32_t num = read32ne(buf); 319 return conv32be(num); 320 #else 321 uint32_t num = (uint32_t)buf[0] << 24; 322 num |= (uint32_t)buf[1] << 16; 323 num |= (uint32_t)buf[2] << 8; 324 num |= (uint32_t)buf[3]; 325 return num; 326 #endif 327 } 328 329 330 static inline uint32_t 331 read32le(const uint8_t *buf) 332 { 333 #if !defined(WORDS_BIGENDIAN) || defined(TUKLIB_FAST_UNALIGNED_ACCESS) 334 uint32_t num = read32ne(buf); 335 return conv32le(num); 336 #else 337 uint32_t num = (uint32_t)buf[0]; 338 num |= (uint32_t)buf[1] << 8; 339 num |= (uint32_t)buf[2] << 16; 340 num |= (uint32_t)buf[3] << 24; 341 return num; 342 #endif 343 } 344 345 346 // NOTE: Possible byte swapping must be done in a macro to allow the compiler 347 // to optimize byte swapping of constants when using glibc's or *BSD's 348 // byte swapping macros. The actual write is done in an inline function 349 // to make type checking of the buf pointer possible. 350 #if defined(WORDS_BIGENDIAN) || defined(TUKLIB_FAST_UNALIGNED_ACCESS) 351 # define write16be(buf, num) write16ne(buf, conv16be(num)) 352 # define write32be(buf, num) write32ne(buf, conv32be(num)) 353 #endif 354 355 #if !defined(WORDS_BIGENDIAN) || defined(TUKLIB_FAST_UNALIGNED_ACCESS) 356 # define write16le(buf, num) write16ne(buf, conv16le(num)) 357 # define write32le(buf, num) write32ne(buf, conv32le(num)) 358 #endif 359 360 361 #ifndef write16be 362 static inline void 363 write16be(uint8_t *buf, uint16_t num) 364 { 365 buf[0] = (uint8_t)(num >> 8); 366 buf[1] = (uint8_t)num; 367 return; 368 } 369 #endif 370 371 372 #ifndef write16le 373 static inline void 374 write16le(uint8_t *buf, uint16_t num) 375 { 376 buf[0] = (uint8_t)num; 377 buf[1] = (uint8_t)(num >> 8); 378 return; 379 } 380 #endif 381 382 383 #ifndef write32be 384 static inline void 385 write32be(uint8_t *buf, uint32_t num) 386 { 387 buf[0] = (uint8_t)(num >> 24); 388 buf[1] = (uint8_t)(num >> 16); 389 buf[2] = (uint8_t)(num >> 8); 390 buf[3] = (uint8_t)num; 391 return; 392 } 393 #endif 394 395 396 #ifndef write32le 397 static inline void 398 write32le(uint8_t *buf, uint32_t num) 399 { 400 buf[0] = (uint8_t)num; 401 buf[1] = (uint8_t)(num >> 8); 402 buf[2] = (uint8_t)(num >> 16); 403 buf[3] = (uint8_t)(num >> 24); 404 return; 405 } 406 #endif 407 408 409 ////////////////////////////// 410 // Aligned reads and writes // 411 ////////////////////////////// 412 413 // Separate functions for aligned reads and writes are provided since on 414 // strict-align archs aligned access is much faster than unaligned access. 415 // 416 // Just like in the unaligned case, memcpy() is needed to avoid 417 // strict aliasing violations. However, on archs that don't support 418 // unaligned access the compiler cannot know that the pointers given 419 // to memcpy() are aligned which results in slow code. As of C11 there is 420 // no standard way to tell the compiler that we know that the address is 421 // aligned but some compilers have language extensions to do that. With 422 // such language extensions the memcpy() method gives excellent results. 423 // 424 // What to do on a strict-align system when no known language extentensions 425 // are available? Falling back to byte-by-byte access would be safe but ruin 426 // optimizations that have been made specifically with aligned access in mind. 427 // As a compromise, aligned reads will fall back to non-compliant type punning 428 // but aligned writes will be byte-by-byte, that is, fast reads are preferred 429 // over fast writes. This obviously isn't great but hopefully it's a working 430 // compromise for now. 431 // 432 // __builtin_assume_aligned is support by GCC >= 4.7 and clang >= 3.6. 433 #ifdef HAVE___BUILTIN_ASSUME_ALIGNED 434 # define tuklib_memcpy_aligned(dest, src, size) \ 435 memcpy(dest, __builtin_assume_aligned(src, size), size) 436 #else 437 # define tuklib_memcpy_aligned(dest, src, size) \ 438 memcpy(dest, src, size) 439 # ifndef TUKLIB_FAST_UNALIGNED_ACCESS 440 # define TUKLIB_USE_UNSAFE_ALIGNED_READS 1 441 # endif 442 #endif 443 444 445 static inline uint16_t 446 aligned_read16ne(const uint8_t *buf) 447 { 448 #if defined(TUKLIB_USE_UNSAFE_TYPE_PUNNING) \ 449 || defined(TUKLIB_USE_UNSAFE_ALIGNED_READS) 450 return *(const uint16_t *)buf; 451 #else 452 uint16_t num; 453 tuklib_memcpy_aligned(&num, buf, sizeof(num)); 454 return num; 455 #endif 456 } 457 458 459 static inline uint32_t 460 aligned_read32ne(const uint8_t *buf) 461 { 462 #if defined(TUKLIB_USE_UNSAFE_TYPE_PUNNING) \ 463 || defined(TUKLIB_USE_UNSAFE_ALIGNED_READS) 464 return *(const uint32_t *)buf; 465 #else 466 uint32_t num; 467 tuklib_memcpy_aligned(&num, buf, sizeof(num)); 468 return num; 469 #endif 470 } 471 472 473 static inline uint64_t 474 aligned_read64ne(const uint8_t *buf) 475 { 476 #if defined(TUKLIB_USE_UNSAFE_TYPE_PUNNING) \ 477 || defined(TUKLIB_USE_UNSAFE_ALIGNED_READS) 478 return *(const uint64_t *)buf; 479 #else 480 uint64_t num; 481 tuklib_memcpy_aligned(&num, buf, sizeof(num)); 482 return num; 483 #endif 484 } 485 486 487 static inline void 488 aligned_write16ne(uint8_t *buf, uint16_t num) 489 { 490 #ifdef TUKLIB_USE_UNSAFE_TYPE_PUNNING 491 *(uint16_t *)buf = num; 492 #else 493 tuklib_memcpy_aligned(buf, &num, sizeof(num)); 494 #endif 495 return; 496 } 497 498 499 static inline void 500 aligned_write32ne(uint8_t *buf, uint32_t num) 501 { 502 #ifdef TUKLIB_USE_UNSAFE_TYPE_PUNNING 503 *(uint32_t *)buf = num; 504 #else 505 tuklib_memcpy_aligned(buf, &num, sizeof(num)); 506 #endif 507 return; 508 } 509 510 511 static inline void 512 aligned_write64ne(uint8_t *buf, uint64_t num) 513 { 514 #ifdef TUKLIB_USE_UNSAFE_TYPE_PUNNING 515 *(uint64_t *)buf = num; 516 #else 517 tuklib_memcpy_aligned(buf, &num, sizeof(num)); 518 #endif 519 return; 520 } 521 522 523 static inline uint16_t 524 aligned_read16be(const uint8_t *buf) 525 { 526 uint16_t num = aligned_read16ne(buf); 527 return conv16be(num); 528 } 529 530 531 static inline uint16_t 532 aligned_read16le(const uint8_t *buf) 533 { 534 uint16_t num = aligned_read16ne(buf); 535 return conv16le(num); 536 } 537 538 539 static inline uint32_t 540 aligned_read32be(const uint8_t *buf) 541 { 542 uint32_t num = aligned_read32ne(buf); 543 return conv32be(num); 544 } 545 546 547 static inline uint32_t 548 aligned_read32le(const uint8_t *buf) 549 { 550 uint32_t num = aligned_read32ne(buf); 551 return conv32le(num); 552 } 553 554 555 static inline uint64_t 556 aligned_read64be(const uint8_t *buf) 557 { 558 uint64_t num = aligned_read64ne(buf); 559 return conv64be(num); 560 } 561 562 563 static inline uint64_t 564 aligned_read64le(const uint8_t *buf) 565 { 566 uint64_t num = aligned_read64ne(buf); 567 return conv64le(num); 568 } 569 570 571 // These need to be macros like in the unaligned case. 572 #define aligned_write16be(buf, num) aligned_write16ne((buf), conv16be(num)) 573 #define aligned_write16le(buf, num) aligned_write16ne((buf), conv16le(num)) 574 #define aligned_write32be(buf, num) aligned_write32ne((buf), conv32be(num)) 575 #define aligned_write32le(buf, num) aligned_write32ne((buf), conv32le(num)) 576 #define aligned_write64be(buf, num) aligned_write64ne((buf), conv64be(num)) 577 #define aligned_write64le(buf, num) aligned_write64ne((buf), conv64le(num)) 578 579 580 //////////////////// 581 // Bit operations // 582 //////////////////// 583 584 static inline uint32_t 585 bsr32(uint32_t n) 586 { 587 // Check for ICC first, since it tends to define __GNUC__ too. 588 #if defined(__INTEL_COMPILER) 589 return _bit_scan_reverse(n); 590 591 #elif TUKLIB_GNUC_REQ(3, 4) && UINT_MAX == UINT32_MAX 592 // GCC >= 3.4 has __builtin_clz(), which gives good results on 593 // multiple architectures. On x86, __builtin_clz() ^ 31U becomes 594 // either plain BSR (so the XOR gets optimized away) or LZCNT and 595 // XOR (if -march indicates that SSE4a instructions are supported). 596 return (uint32_t)__builtin_clz(n) ^ 31U; 597 598 #elif defined(__GNUC__) && (defined(__i386__) || defined(__x86_64__)) 599 uint32_t i; 600 __asm__("bsrl %1, %0" : "=r" (i) : "rm" (n)); 601 return i; 602 603 #elif defined(_MSC_VER) 604 unsigned long i; 605 _BitScanReverse(&i, n); 606 return i; 607 608 #else 609 uint32_t i = 31; 610 611 if ((n & 0xFFFF0000) == 0) { 612 n <<= 16; 613 i = 15; 614 } 615 616 if ((n & 0xFF000000) == 0) { 617 n <<= 8; 618 i -= 8; 619 } 620 621 if ((n & 0xF0000000) == 0) { 622 n <<= 4; 623 i -= 4; 624 } 625 626 if ((n & 0xC0000000) == 0) { 627 n <<= 2; 628 i -= 2; 629 } 630 631 if ((n & 0x80000000) == 0) 632 --i; 633 634 return i; 635 #endif 636 } 637 638 639 static inline uint32_t 640 clz32(uint32_t n) 641 { 642 #if defined(__INTEL_COMPILER) 643 return _bit_scan_reverse(n) ^ 31U; 644 645 #elif TUKLIB_GNUC_REQ(3, 4) && UINT_MAX == UINT32_MAX 646 return (uint32_t)__builtin_clz(n); 647 648 #elif defined(__GNUC__) && (defined(__i386__) || defined(__x86_64__)) 649 uint32_t i; 650 __asm__("bsrl %1, %0\n\t" 651 "xorl $31, %0" 652 : "=r" (i) : "rm" (n)); 653 return i; 654 655 #elif defined(_MSC_VER) 656 unsigned long i; 657 _BitScanReverse(&i, n); 658 return i ^ 31U; 659 660 #else 661 uint32_t i = 0; 662 663 if ((n & 0xFFFF0000) == 0) { 664 n <<= 16; 665 i = 16; 666 } 667 668 if ((n & 0xFF000000) == 0) { 669 n <<= 8; 670 i += 8; 671 } 672 673 if ((n & 0xF0000000) == 0) { 674 n <<= 4; 675 i += 4; 676 } 677 678 if ((n & 0xC0000000) == 0) { 679 n <<= 2; 680 i += 2; 681 } 682 683 if ((n & 0x80000000) == 0) 684 ++i; 685 686 return i; 687 #endif 688 } 689 690 691 static inline uint32_t 692 ctz32(uint32_t n) 693 { 694 #if defined(__INTEL_COMPILER) 695 return _bit_scan_forward(n); 696 697 #elif TUKLIB_GNUC_REQ(3, 4) && UINT_MAX >= UINT32_MAX 698 return (uint32_t)__builtin_ctz(n); 699 700 #elif defined(__GNUC__) && (defined(__i386__) || defined(__x86_64__)) 701 uint32_t i; 702 __asm__("bsfl %1, %0" : "=r" (i) : "rm" (n)); 703 return i; 704 705 #elif defined(_MSC_VER) 706 unsigned long i; 707 _BitScanForward(&i, n); 708 return i; 709 710 #else 711 uint32_t i = 0; 712 713 if ((n & 0x0000FFFF) == 0) { 714 n >>= 16; 715 i = 16; 716 } 717 718 if ((n & 0x000000FF) == 0) { 719 n >>= 8; 720 i += 8; 721 } 722 723 if ((n & 0x0000000F) == 0) { 724 n >>= 4; 725 i += 4; 726 } 727 728 if ((n & 0x00000003) == 0) { 729 n >>= 2; 730 i += 2; 731 } 732 733 if ((n & 0x00000001) == 0) 734 ++i; 735 736 return i; 737 #endif 738 } 739 740 #define bsf32 ctz32 741 742 #endif 743