bmiintrin.h revision 321369
1/*===---- bmiintrin.h - BMI intrinsics -------------------------------------=== 2 * 3 * Permission is hereby granted, free of charge, to any person obtaining a copy 4 * of this software and associated documentation files (the "Software"), to deal 5 * in the Software without restriction, including without limitation the rights 6 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell 7 * copies of the Software, and to permit persons to whom the Software is 8 * furnished to do so, subject to the following conditions: 9 * 10 * The above copyright notice and this permission notice shall be included in 11 * all copies or substantial portions of the Software. 12 * 13 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 14 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 15 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE 16 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 17 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, 18 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN 19 * THE SOFTWARE. 20 * 21 *===-----------------------------------------------------------------------=== 22 */ 23 24#if !defined __X86INTRIN_H && !defined __IMMINTRIN_H 25#error "Never use <bmiintrin.h> directly; include <x86intrin.h> instead." 26#endif 27 28#ifndef __BMIINTRIN_H 29#define __BMIINTRIN_H 30 31#define _tzcnt_u16(a) (__tzcnt_u16((a))) 32 33#define _andn_u32(a, b) (__andn_u32((a), (b))) 34 35/* _bextr_u32 != __bextr_u32 */ 36#define _blsi_u32(a) (__blsi_u32((a))) 37 38#define _blsmsk_u32(a) (__blsmsk_u32((a))) 39 40#define _blsr_u32(a) (__blsr_u32((a))) 41 42#define _tzcnt_u32(a) (__tzcnt_u32((a))) 43 44/* Define the default attributes for the functions in this file. */ 45#define __DEFAULT_FN_ATTRS __attribute__((__always_inline__, __nodebug__, __target__("bmi"))) 46 47/* Allow using the tzcnt intrinsics even for non-BMI targets. Since the TZCNT 48 instruction behaves as BSF on non-BMI targets, there is code that expects 49 to use it as a potentially faster version of BSF. */ 50#define __RELAXED_FN_ATTRS __attribute__((__always_inline__, __nodebug__)) 51 52/// \brief Counts the number of trailing zero bits in the operand. 53/// 54/// \headerfile <x86intrin.h> 55/// 56/// This intrinsic corresponds to the <c> TZCNT </c> instruction. 57/// 58/// \param __X 59/// An unsigned 16-bit integer whose trailing zeros are to be counted. 60/// \returns An unsigned 16-bit integer containing the number of trailing zero 61/// bits in the operand. 62static __inline__ unsigned short __RELAXED_FN_ATTRS 63__tzcnt_u16(unsigned short __X) 64{ 65 return __X ? __builtin_ctzs(__X) : 16; 66} 67 68/// \brief Performs a bitwise AND of the second operand with the one's 69/// complement of the first operand. 70/// 71/// \headerfile <x86intrin.h> 72/// 73/// This intrinsic corresponds to the <c> ANDN </c> instruction. 74/// 75/// \param __X 76/// An unsigned integer containing one of the operands. 77/// \param __Y 78/// An unsigned integer containing one of the operands. 79/// \returns An unsigned integer containing the bitwise AND of the second 80/// operand with the one's complement of the first operand. 81static __inline__ unsigned int __DEFAULT_FN_ATTRS 82__andn_u32(unsigned int __X, unsigned int __Y) 83{ 84 return ~__X & __Y; 85} 86 87/* AMD-specified, double-leading-underscore version of BEXTR */ 88/// \brief Extracts the specified bits from the first operand and returns them 89/// in the least significant bits of the result. 90/// 91/// \headerfile <x86intrin.h> 92/// 93/// This intrinsic corresponds to the <c> BEXTR </c> instruction. 94/// 95/// \param __X 96/// An unsigned integer whose bits are to be extracted. 97/// \param __Y 98/// An unsigned integer used to specify which bits are extracted. Bits [7:0] 99/// specify the index of the least significant bit. Bits [15:8] specify the 100/// number of bits to be extracted. 101/// \returns An unsigned integer whose least significant bits contain the 102/// extracted bits. 103static __inline__ unsigned int __DEFAULT_FN_ATTRS 104__bextr_u32(unsigned int __X, unsigned int __Y) 105{ 106 return __builtin_ia32_bextr_u32(__X, __Y); 107} 108 109/* Intel-specified, single-leading-underscore version of BEXTR */ 110/// \brief Extracts the specified bits from the first operand and returns them 111/// in the least significant bits of the result. 112/// 113/// \headerfile <x86intrin.h> 114/// 115/// This intrinsic corresponds to the <c> BEXTR </c> instruction. 116/// 117/// \param __X 118/// An unsigned integer whose bits are to be extracted. 119/// \param __Y 120/// An unsigned integer used to specify the index of the least significant 121/// bit for the bits to be extracted. Bits [7:0] specify the index. 122/// \param __Z 123/// An unsigned integer used to specify the number of bits to be extracted. 124/// Bits [7:0] specify the number of bits. 125/// \returns An unsigned integer whose least significant bits contain the 126/// extracted bits. 127static __inline__ unsigned int __DEFAULT_FN_ATTRS 128_bextr_u32(unsigned int __X, unsigned int __Y, unsigned int __Z) 129{ 130 return __builtin_ia32_bextr_u32 (__X, ((__Y & 0xff) | ((__Z & 0xff) << 8))); 131} 132 133/// \brief Clears all bits in the source except for the least significant bit 134/// containing a value of 1 and returns the result. 135/// 136/// \headerfile <x86intrin.h> 137/// 138/// This intrinsic corresponds to the <c> BLSI </c> instruction. 139/// 140/// \param __X 141/// An unsigned integer whose bits are to be cleared. 142/// \returns An unsigned integer containing the result of clearing the bits from 143/// the source operand. 144static __inline__ unsigned int __DEFAULT_FN_ATTRS 145__blsi_u32(unsigned int __X) 146{ 147 return __X & -__X; 148} 149 150/// \brief Creates a mask whose bits are set to 1, using bit 0 up to and 151/// including the least significant bit that is set to 1 in the source 152/// operand and returns the result. 153/// 154/// \headerfile <x86intrin.h> 155/// 156/// This intrinsic corresponds to the <c> BLSMSK </c> instruction. 157/// 158/// \param __X 159/// An unsigned integer used to create the mask. 160/// \returns An unsigned integer containing the newly created mask. 161static __inline__ unsigned int __DEFAULT_FN_ATTRS 162__blsmsk_u32(unsigned int __X) 163{ 164 return __X ^ (__X - 1); 165} 166 167/// \brief Clears the least significant bit that is set to 1 in the source 168/// operand and returns the result. 169/// 170/// \headerfile <x86intrin.h> 171/// 172/// This intrinsic corresponds to the <c> BLSR </c> instruction. 173/// 174/// \param __X 175/// An unsigned integer containing the operand to be cleared. 176/// \returns An unsigned integer containing the result of clearing the source 177/// operand. 178static __inline__ unsigned int __DEFAULT_FN_ATTRS 179__blsr_u32(unsigned int __X) 180{ 181 return __X & (__X - 1); 182} 183 184/// \brief Counts the number of trailing zero bits in the operand. 185/// 186/// \headerfile <x86intrin.h> 187/// 188/// This intrinsic corresponds to the <c> TZCNT </c> instruction. 189/// 190/// \param __X 191/// An unsigned 32-bit integer whose trailing zeros are to be counted. 192/// \returns An unsigned 32-bit integer containing the number of trailing zero 193/// bits in the operand. 194static __inline__ unsigned int __RELAXED_FN_ATTRS 195__tzcnt_u32(unsigned int __X) 196{ 197 return __X ? __builtin_ctz(__X) : 32; 198} 199 200/// \brief Counts the number of trailing zero bits in the operand. 201/// 202/// \headerfile <x86intrin.h> 203/// 204/// This intrinsic corresponds to the <c> TZCNT </c> instruction. 205/// 206/// \param __X 207/// An unsigned 32-bit integer whose trailing zeros are to be counted. 208/// \returns An 32-bit integer containing the number of trailing zero bits in 209/// the operand. 210static __inline__ int __RELAXED_FN_ATTRS 211_mm_tzcnt_32(unsigned int __X) 212{ 213 return __X ? __builtin_ctz(__X) : 32; 214} 215 216#ifdef __x86_64__ 217 218#define _andn_u64(a, b) (__andn_u64((a), (b))) 219 220/* _bextr_u64 != __bextr_u64 */ 221#define _blsi_u64(a) (__blsi_u64((a))) 222 223#define _blsmsk_u64(a) (__blsmsk_u64((a))) 224 225#define _blsr_u64(a) (__blsr_u64((a))) 226 227#define _tzcnt_u64(a) (__tzcnt_u64((a))) 228 229/// \brief Performs a bitwise AND of the second operand with the one's 230/// complement of the first operand. 231/// 232/// \headerfile <x86intrin.h> 233/// 234/// This intrinsic corresponds to the <c> ANDN </c> instruction. 235/// 236/// \param __X 237/// An unsigned 64-bit integer containing one of the operands. 238/// \param __Y 239/// An unsigned 64-bit integer containing one of the operands. 240/// \returns An unsigned 64-bit integer containing the bitwise AND of the second 241/// operand with the one's complement of the first operand. 242static __inline__ unsigned long long __DEFAULT_FN_ATTRS 243__andn_u64 (unsigned long long __X, unsigned long long __Y) 244{ 245 return ~__X & __Y; 246} 247 248/* AMD-specified, double-leading-underscore version of BEXTR */ 249/// \brief Extracts the specified bits from the first operand and returns them 250/// in the least significant bits of the result. 251/// 252/// \headerfile <x86intrin.h> 253/// 254/// This intrinsic corresponds to the <c> BEXTR </c> instruction. 255/// 256/// \param __X 257/// An unsigned 64-bit integer whose bits are to be extracted. 258/// \param __Y 259/// An unsigned 64-bit integer used to specify which bits are extracted. Bits 260/// [7:0] specify the index of the least significant bit. Bits [15:8] specify 261/// the number of bits to be extracted. 262/// \returns An unsigned 64-bit integer whose least significant bits contain the 263/// extracted bits. 264static __inline__ unsigned long long __DEFAULT_FN_ATTRS 265__bextr_u64(unsigned long long __X, unsigned long long __Y) 266{ 267 return __builtin_ia32_bextr_u64(__X, __Y); 268} 269 270/* Intel-specified, single-leading-underscore version of BEXTR */ 271/// \brief Extracts the specified bits from the first operand and returns them 272/// in the least significant bits of the result. 273/// 274/// \headerfile <x86intrin.h> 275/// 276/// This intrinsic corresponds to the <c> BEXTR </c> instruction. 277/// 278/// \param __X 279/// An unsigned 64-bit integer whose bits are to be extracted. 280/// \param __Y 281/// An unsigned integer used to specify the index of the least significant 282/// bit for the bits to be extracted. Bits [7:0] specify the index. 283/// \param __Z 284/// An unsigned integer used to specify the number of bits to be extracted. 285/// Bits [7:0] specify the number of bits. 286/// \returns An unsigned 64-bit integer whose least significant bits contain the 287/// extracted bits. 288static __inline__ unsigned long long __DEFAULT_FN_ATTRS 289_bextr_u64(unsigned long long __X, unsigned int __Y, unsigned int __Z) 290{ 291 return __builtin_ia32_bextr_u64 (__X, ((__Y & 0xff) | ((__Z & 0xff) << 8))); 292} 293 294/// \brief Clears all bits in the source except for the least significant bit 295/// containing a value of 1 and returns the result. 296/// 297/// \headerfile <x86intrin.h> 298/// 299/// This intrinsic corresponds to the <c> BLSI </c> instruction. 300/// 301/// \param __X 302/// An unsigned 64-bit integer whose bits are to be cleared. 303/// \returns An unsigned 64-bit integer containing the result of clearing the 304/// bits from the source operand. 305static __inline__ unsigned long long __DEFAULT_FN_ATTRS 306__blsi_u64(unsigned long long __X) 307{ 308 return __X & -__X; 309} 310 311/// \brief Creates a mask whose bits are set to 1, using bit 0 up to and 312/// including the least significant bit that is set to 1 in the source 313/// operand and returns the result. 314/// 315/// \headerfile <x86intrin.h> 316/// 317/// This intrinsic corresponds to the <c> BLSMSK </c> instruction. 318/// 319/// \param __X 320/// An unsigned 64-bit integer used to create the mask. 321/// \returns An unsigned 64-bit integer containing the newly created mask. 322static __inline__ unsigned long long __DEFAULT_FN_ATTRS 323__blsmsk_u64(unsigned long long __X) 324{ 325 return __X ^ (__X - 1); 326} 327 328/// \brief Clears the least significant bit that is set to 1 in the source 329/// operand and returns the result. 330/// 331/// \headerfile <x86intrin.h> 332/// 333/// This intrinsic corresponds to the <c> BLSR </c> instruction. 334/// 335/// \param __X 336/// An unsigned 64-bit integer containing the operand to be cleared. 337/// \returns An unsigned 64-bit integer containing the result of clearing the 338/// source operand. 339static __inline__ unsigned long long __DEFAULT_FN_ATTRS 340__blsr_u64(unsigned long long __X) 341{ 342 return __X & (__X - 1); 343} 344 345/// \brief Counts the number of trailing zero bits in the operand. 346/// 347/// \headerfile <x86intrin.h> 348/// 349/// This intrinsic corresponds to the <c> TZCNT </c> instruction. 350/// 351/// \param __X 352/// An unsigned 64-bit integer whose trailing zeros are to be counted. 353/// \returns An unsigned 64-bit integer containing the number of trailing zero 354/// bits in the operand. 355static __inline__ unsigned long long __RELAXED_FN_ATTRS 356__tzcnt_u64(unsigned long long __X) 357{ 358 return __X ? __builtin_ctzll(__X) : 64; 359} 360 361/// \brief Counts the number of trailing zero bits in the operand. 362/// 363/// \headerfile <x86intrin.h> 364/// 365/// This intrinsic corresponds to the <c> TZCNT </c> instruction. 366/// 367/// \param __X 368/// An unsigned 64-bit integer whose trailing zeros are to be counted. 369/// \returns An 64-bit integer containing the number of trailing zero bits in 370/// the operand. 371static __inline__ long long __RELAXED_FN_ATTRS 372_mm_tzcnt_64(unsigned long long __X) 373{ 374 return __X ? __builtin_ctzll(__X) : 64; 375} 376 377#endif /* __x86_64__ */ 378 379#undef __DEFAULT_FN_ATTRS 380#undef __RELAXED_FN_ATTRS 381 382#endif /* __BMIINTRIN_H */ 383