bmiintrin.h revision 341825
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/// 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/// 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/// 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. 103/// \see _bextr_u32 104static __inline__ unsigned int __DEFAULT_FN_ATTRS 105__bextr_u32(unsigned int __X, unsigned int __Y) 106{ 107 return __builtin_ia32_bextr_u32(__X, __Y); 108} 109 110/* Intel-specified, single-leading-underscore version of BEXTR */ 111/// Extracts the specified bits from the first operand and returns them 112/// in the least significant bits of the result. 113/// 114/// \headerfile <x86intrin.h> 115/// 116/// This intrinsic corresponds to the <c> BEXTR </c> instruction. 117/// 118/// \param __X 119/// An unsigned integer whose bits are to be extracted. 120/// \param __Y 121/// An unsigned integer used to specify the index of the least significant 122/// bit for the bits to be extracted. Bits [7:0] specify the index. 123/// \param __Z 124/// An unsigned integer used to specify the number of bits to be extracted. 125/// Bits [7:0] specify the number of bits. 126/// \returns An unsigned integer whose least significant bits contain the 127/// extracted bits. 128/// \see __bextr_u32 129static __inline__ unsigned int __DEFAULT_FN_ATTRS 130_bextr_u32(unsigned int __X, unsigned int __Y, unsigned int __Z) 131{ 132 return __builtin_ia32_bextr_u32 (__X, ((__Y & 0xff) | ((__Z & 0xff) << 8))); 133} 134 135/// Clears all bits in the source except for the least significant bit 136/// containing a value of 1 and returns the result. 137/// 138/// \headerfile <x86intrin.h> 139/// 140/// This intrinsic corresponds to the <c> BLSI </c> instruction. 141/// 142/// \param __X 143/// An unsigned integer whose bits are to be cleared. 144/// \returns An unsigned integer containing the result of clearing the bits from 145/// the source operand. 146static __inline__ unsigned int __DEFAULT_FN_ATTRS 147__blsi_u32(unsigned int __X) 148{ 149 return __X & -__X; 150} 151 152/// Creates a mask whose bits are set to 1, using bit 0 up to and 153/// including the least significant bit that is set to 1 in the source 154/// operand and returns the result. 155/// 156/// \headerfile <x86intrin.h> 157/// 158/// This intrinsic corresponds to the <c> BLSMSK </c> instruction. 159/// 160/// \param __X 161/// An unsigned integer used to create the mask. 162/// \returns An unsigned integer containing the newly created mask. 163static __inline__ unsigned int __DEFAULT_FN_ATTRS 164__blsmsk_u32(unsigned int __X) 165{ 166 return __X ^ (__X - 1); 167} 168 169/// Clears the least significant bit that is set to 1 in the source 170/// operand and returns the result. 171/// 172/// \headerfile <x86intrin.h> 173/// 174/// This intrinsic corresponds to the <c> BLSR </c> instruction. 175/// 176/// \param __X 177/// An unsigned integer containing the operand to be cleared. 178/// \returns An unsigned integer containing the result of clearing the source 179/// operand. 180static __inline__ unsigned int __DEFAULT_FN_ATTRS 181__blsr_u32(unsigned int __X) 182{ 183 return __X & (__X - 1); 184} 185 186/// Counts the number of trailing zero bits in the operand. 187/// 188/// \headerfile <x86intrin.h> 189/// 190/// This intrinsic corresponds to the <c> TZCNT </c> instruction. 191/// 192/// \param __X 193/// An unsigned 32-bit integer whose trailing zeros are to be counted. 194/// \returns An unsigned 32-bit integer containing the number of trailing zero 195/// bits in the operand. 196static __inline__ unsigned int __RELAXED_FN_ATTRS 197__tzcnt_u32(unsigned int __X) 198{ 199 return __X ? __builtin_ctz(__X) : 32; 200} 201 202/// Counts the number of trailing zero bits in the operand. 203/// 204/// \headerfile <x86intrin.h> 205/// 206/// This intrinsic corresponds to the <c> TZCNT </c> instruction. 207/// 208/// \param __X 209/// An unsigned 32-bit integer whose trailing zeros are to be counted. 210/// \returns An 32-bit integer containing the number of trailing zero bits in 211/// the operand. 212static __inline__ int __RELAXED_FN_ATTRS 213_mm_tzcnt_32(unsigned int __X) 214{ 215 return __X ? __builtin_ctz(__X) : 32; 216} 217 218#ifdef __x86_64__ 219 220#define _andn_u64(a, b) (__andn_u64((a), (b))) 221 222/* _bextr_u64 != __bextr_u64 */ 223#define _blsi_u64(a) (__blsi_u64((a))) 224 225#define _blsmsk_u64(a) (__blsmsk_u64((a))) 226 227#define _blsr_u64(a) (__blsr_u64((a))) 228 229#define _tzcnt_u64(a) (__tzcnt_u64((a))) 230 231/// Performs a bitwise AND of the second operand with the one's 232/// complement of the first operand. 233/// 234/// \headerfile <x86intrin.h> 235/// 236/// This intrinsic corresponds to the <c> ANDN </c> instruction. 237/// 238/// \param __X 239/// An unsigned 64-bit integer containing one of the operands. 240/// \param __Y 241/// An unsigned 64-bit integer containing one of the operands. 242/// \returns An unsigned 64-bit integer containing the bitwise AND of the second 243/// operand with the one's complement of the first operand. 244static __inline__ unsigned long long __DEFAULT_FN_ATTRS 245__andn_u64 (unsigned long long __X, unsigned long long __Y) 246{ 247 return ~__X & __Y; 248} 249 250/* AMD-specified, double-leading-underscore version of BEXTR */ 251/// Extracts the specified bits from the first operand and returns them 252/// in the least significant bits of the result. 253/// 254/// \headerfile <x86intrin.h> 255/// 256/// This intrinsic corresponds to the <c> BEXTR </c> instruction. 257/// 258/// \param __X 259/// An unsigned 64-bit integer whose bits are to be extracted. 260/// \param __Y 261/// An unsigned 64-bit integer used to specify which bits are extracted. Bits 262/// [7:0] specify the index of the least significant bit. Bits [15:8] specify 263/// the number of bits to be extracted. 264/// \returns An unsigned 64-bit integer whose least significant bits contain the 265/// extracted bits. 266/// \see _bextr_u64 267static __inline__ unsigned long long __DEFAULT_FN_ATTRS 268__bextr_u64(unsigned long long __X, unsigned long long __Y) 269{ 270 return __builtin_ia32_bextr_u64(__X, __Y); 271} 272 273/* Intel-specified, single-leading-underscore version of BEXTR */ 274/// Extracts the specified bits from the first operand and returns them 275/// in the least significant bits of the result. 276/// 277/// \headerfile <x86intrin.h> 278/// 279/// This intrinsic corresponds to the <c> BEXTR </c> instruction. 280/// 281/// \param __X 282/// An unsigned 64-bit integer whose bits are to be extracted. 283/// \param __Y 284/// An unsigned integer used to specify the index of the least significant 285/// bit for the bits to be extracted. Bits [7:0] specify the index. 286/// \param __Z 287/// An unsigned integer used to specify the number of bits to be extracted. 288/// Bits [7:0] specify the number of bits. 289/// \returns An unsigned 64-bit integer whose least significant bits contain the 290/// extracted bits. 291/// \see __bextr_u64 292static __inline__ unsigned long long __DEFAULT_FN_ATTRS 293_bextr_u64(unsigned long long __X, unsigned int __Y, unsigned int __Z) 294{ 295 return __builtin_ia32_bextr_u64 (__X, ((__Y & 0xff) | ((__Z & 0xff) << 8))); 296} 297 298/// Clears all bits in the source except for the least significant bit 299/// containing a value of 1 and returns the result. 300/// 301/// \headerfile <x86intrin.h> 302/// 303/// This intrinsic corresponds to the <c> BLSI </c> instruction. 304/// 305/// \param __X 306/// An unsigned 64-bit integer whose bits are to be cleared. 307/// \returns An unsigned 64-bit integer containing the result of clearing the 308/// bits from the source operand. 309static __inline__ unsigned long long __DEFAULT_FN_ATTRS 310__blsi_u64(unsigned long long __X) 311{ 312 return __X & -__X; 313} 314 315/// Creates a mask whose bits are set to 1, using bit 0 up to and 316/// including the least significant bit that is set to 1 in the source 317/// operand and returns the result. 318/// 319/// \headerfile <x86intrin.h> 320/// 321/// This intrinsic corresponds to the <c> BLSMSK </c> instruction. 322/// 323/// \param __X 324/// An unsigned 64-bit integer used to create the mask. 325/// \returns An unsigned 64-bit integer containing the newly created mask. 326static __inline__ unsigned long long __DEFAULT_FN_ATTRS 327__blsmsk_u64(unsigned long long __X) 328{ 329 return __X ^ (__X - 1); 330} 331 332/// Clears the least significant bit that is set to 1 in the source 333/// operand and returns the result. 334/// 335/// \headerfile <x86intrin.h> 336/// 337/// This intrinsic corresponds to the <c> BLSR </c> instruction. 338/// 339/// \param __X 340/// An unsigned 64-bit integer containing the operand to be cleared. 341/// \returns An unsigned 64-bit integer containing the result of clearing the 342/// source operand. 343static __inline__ unsigned long long __DEFAULT_FN_ATTRS 344__blsr_u64(unsigned long long __X) 345{ 346 return __X & (__X - 1); 347} 348 349/// Counts the number of trailing zero bits in the operand. 350/// 351/// \headerfile <x86intrin.h> 352/// 353/// This intrinsic corresponds to the <c> TZCNT </c> instruction. 354/// 355/// \param __X 356/// An unsigned 64-bit integer whose trailing zeros are to be counted. 357/// \returns An unsigned 64-bit integer containing the number of trailing zero 358/// bits in the operand. 359static __inline__ unsigned long long __RELAXED_FN_ATTRS 360__tzcnt_u64(unsigned long long __X) 361{ 362 return __X ? __builtin_ctzll(__X) : 64; 363} 364 365/// Counts the number of trailing zero bits in the operand. 366/// 367/// \headerfile <x86intrin.h> 368/// 369/// This intrinsic corresponds to the <c> TZCNT </c> instruction. 370/// 371/// \param __X 372/// An unsigned 64-bit integer whose trailing zeros are to be counted. 373/// \returns An 64-bit integer containing the number of trailing zero bits in 374/// the operand. 375static __inline__ long long __RELAXED_FN_ATTRS 376_mm_tzcnt_64(unsigned long long __X) 377{ 378 return __X ? __builtin_ctzll(__X) : 64; 379} 380 381#endif /* __x86_64__ */ 382 383#undef __DEFAULT_FN_ATTRS 384#undef __RELAXED_FN_ATTRS 385 386#endif /* __BMIINTRIN_H */ 387