blob 4d0e770f (8692B) - Raw
1 /*===---- ammintrin.h - SSE4a 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 #ifndef __AMMINTRIN_H 25 #define __AMMINTRIN_H 26 27 #ifndef __SSE4A__ 28 #error "SSE4A instruction set not enabled" 29 #else 30 31 #include <pmmintrin.h> 32 33 /* Define the default attributes for the functions in this file. */ 34 #define __DEFAULT_FN_ATTRS __attribute__((__always_inline__, __nodebug__)) 35 36 /// \brief Extracts the specified bits from the lower 64 bits of the 128-bit 37 /// integer vector operand at the index idx and of the length len. 38 /// 39 /// \headerfile <x86intrin.h> 40 /// 41 /// \code 42 /// __m128i _mm_extracti_si64(__m128i x, const int len, const int idx); 43 /// \endcode 44 /// 45 /// \code 46 /// This intrinsic corresponds to the \c EXTRQ instruction. 47 /// \endcode 48 /// 49 /// \param x 50 /// The value from which bits are extracted. 51 /// \param len 52 /// Bits [5:0] specify the length; the other bits are ignored. If bits [5:0] 53 /// are zero, the length is interpreted as 64. 54 /// \param idx 55 /// Bits [5:0] specify the index of the least significant bit; the other 56 /// bits are ignored. If the sum of the index and length is greater than 57 /// 64, the result is undefined. If the length and index are both zero, 58 /// bits [63:0] of parameter x are extracted. If the length is zero 59 /// but the index is non-zero, the result is undefined. 60 /// \returns A 128-bit integer vector whose lower 64 bits contain the bits 61 /// extracted from the source operand. 62 #define _mm_extracti_si64(x, len, idx) \ 63 ((__m128i)__builtin_ia32_extrqi((__v2di)(__m128i)(x), \ 64 (char)(len), (char)(idx))) 65 66 /// \brief Extracts the specified bits from the lower 64 bits of the 128-bit 67 /// integer vector operand at the index and of the length specified by __y. 68 /// 69 /// \headerfile <x86intrin.h> 70 /// 71 /// \code 72 /// This intrinsic corresponds to the \c EXTRQ instruction. 73 /// \endcode 74 /// 75 /// \param __x 76 /// The value from which bits are extracted. 77 /// \param __y 78 /// Specifies the index of the least significant bit at [13:8] 79 /// and the length at [5:0]; all other bits are ignored. 80 /// If bits [5:0] are zero, the length is interpreted as 64. 81 /// If the sum of the index and length is greater than 64, the result is 82 /// undefined. If the length and index are both zero, bits [63:0] of 83 /// parameter __x are extracted. If the length is zero but the index is 84 /// non-zero, the result is undefined. 85 /// \returns A 128-bit vector whose lower 64 bits contain the bits extracted 86 /// from the source operand. 87 static __inline__ __m128i __DEFAULT_FN_ATTRS 88 _mm_extract_si64(__m128i __x, __m128i __y) 89 { 90 return (__m128i)__builtin_ia32_extrq((__v2di)__x, (__v16qi)__y); 91 } 92 93 /// \brief Inserts bits of a specified length from the source integer vector 94 /// y into the lower 64 bits of the destination integer vector x at the 95 /// index idx and of the length len. 96 /// 97 /// \headerfile <x86intrin.h> 98 /// 99 /// \code 100 /// __m128i _mm_inserti_si64(__m128i x, __m128i y, const int len, 101 /// const int idx); 102 /// \endcode 103 /// 104 /// \code 105 /// This intrinsic corresponds to the \c INSERTQ instruction. 106 /// \endcode 107 /// 108 /// \param x 109 /// The destination operand where bits will be inserted. The inserted bits 110 /// are defined by the length len and by the index idx specifying the least 111 /// significant bit. 112 /// \param y 113 /// The source operand containing the bits to be extracted. The extracted 114 /// bits are the least significant bits of operand y of length len. 115 /// \param len 116 /// Bits [5:0] specify the length; the other bits are ignored. If bits [5:0] 117 /// are zero, the length is interpreted as 64. 118 /// \param idx 119 /// Bits [5:0] specify the index of the least significant bit; the other 120 /// bits are ignored. If the sum of the index and length is greater than 121 /// 64, the result is undefined. If the length and index are both zero, 122 /// bits [63:0] of parameter y are inserted into parameter x. If the 123 /// length is zero but the index is non-zero, the result is undefined. 124 /// \returns A 128-bit integer vector containing the original lower 64-bits 125 /// of destination operand x with the specified bitfields replaced by the 126 /// lower bits of source operand y. The upper 64 bits of the return value 127 /// are undefined. 128 129 #define _mm_inserti_si64(x, y, len, idx) \ 130 ((__m128i)__builtin_ia32_insertqi((__v2di)(__m128i)(x), \ 131 (__v2di)(__m128i)(y), \ 132 (char)(len), (char)(idx))) 133 134 /// \brief Inserts bits of a specified length from the source integer vector 135 /// __y into the lower 64 bits of the destination integer vector __x at 136 /// the index and of the length specified by __y. 137 /// 138 /// \headerfile <x86intrin.h> 139 /// 140 /// \code 141 /// This intrinsic corresponds to the \c INSERTQ instruction. 142 /// \endcode 143 /// 144 /// \param __x 145 /// The destination operand where bits will be inserted. The inserted bits 146 /// are defined by the length and by the index of the least significant bit 147 /// specified by operand __y. 148 /// \param __y 149 /// The source operand containing the bits to be extracted. The extracted 150 /// bits are the least significant bits of operand __y with length specified 151 /// by bits [69:64]. These are inserted into the destination at the index 152 /// specified by bits [77:72]; all other bits are ignored. 153 /// If bits [69:64] are zero, the length is interpreted as 64. 154 /// If the sum of the index and length is greater than 64, the result is 155 /// undefined. If the length and index are both zero, bits [63:0] of 156 /// parameter __y are inserted into parameter __x. If the length 157 /// is zero but the index is non-zero, the result is undefined. 158 /// \returns A 128-bit integer vector containing the original lower 64-bits 159 /// of destination operand __x with the specified bitfields replaced by the 160 /// lower bits of source operand __y. The upper 64 bits of the return value 161 /// are undefined. 162 163 static __inline__ __m128i __DEFAULT_FN_ATTRS 164 _mm_insert_si64(__m128i __x, __m128i __y) 165 { 166 return (__m128i)__builtin_ia32_insertq((__v2di)__x, (__v2di)__y); 167 } 168 169 /// \brief Stores a 64-bit double-precision value in a 64-bit memory location. 170 /// To minimize caching, the data is flagged as non-temporal (unlikely to be 171 /// used again soon). 172 /// 173 /// \headerfile <x86intrin.h> 174 /// 175 /// \code 176 /// This intrinsic corresponds to the \c MOVNTSD instruction. 177 /// \endcode 178 /// 179 /// \param __p 180 /// The 64-bit memory location used to store the register value. 181 /// \param __a 182 /// The 64-bit double-precision floating-point register value to 183 /// be stored. 184 static __inline__ void __DEFAULT_FN_ATTRS 185 _mm_stream_sd(double *__p, __m128d __a) 186 { 187 __builtin_ia32_movntsd(__p, (__v2df)__a); 188 } 189 190 /// \brief Stores a 32-bit single-precision floating-point value in a 32-bit 191 /// memory location. To minimize caching, the data is flagged as 192 /// non-temporal (unlikely to be used again soon). 193 /// 194 /// \headerfile <x86intrin.h> 195 /// 196 /// \code 197 /// This intrinsic corresponds to the \c MOVNTSS instruction. 198 /// \endcode 199 /// 200 /// \param __p 201 /// The 32-bit memory location used to store the register value. 202 /// \param __a 203 /// The 32-bit single-precision floating-point register value to 204 /// be stored. 205 static __inline__ void __DEFAULT_FN_ATTRS 206 _mm_stream_ss(float *__p, __m128 __a) 207 { 208 __builtin_ia32_movntss(__p, (__v4sf)__a); 209 } 210 211 #undef __DEFAULT_FN_ATTRS 212 213 #endif /* __SSE4A__ */ 214 215 #endif /* __AMMINTRIN_H */