| 1 | /*===---- pmmintrin.h - SSE3 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 __PMMINTRIN_H |
| 25 | #define __PMMINTRIN_H |
| 26 | |
| 27 | #include <emmintrin.h> |
| 28 | |
| 29 | /* Define the default attributes for the functions in this file. */ |
| 30 | #define __DEFAULT_FN_ATTRS \ |
| 31 | __attribute__((__always_inline__, __nodebug__, __target__("sse3"), __min_vector_width__(128))) |
| 32 | |
| 33 | /// Loads data from an unaligned memory location to elements in a 128-bit |
| 34 | /// vector. |
| 35 | /// |
| 36 | /// If the address of the data is not 16-byte aligned, the instruction may |
| 37 | /// read two adjacent aligned blocks of memory to retrieve the requested |
| 38 | /// data. |
| 39 | /// |
| 40 | /// \headerfile <x86intrin.h> |
| 41 | /// |
| 42 | /// This intrinsic corresponds to the <c> VLDDQU </c> instruction. |
| 43 | /// |
| 44 | /// \param __p |
| 45 | /// A pointer to a 128-bit integer vector containing integer values. |
| 46 | /// \returns A 128-bit vector containing the moved values. |
| 47 | static __inline__ __m128i __DEFAULT_FN_ATTRS |
| 48 | _mm_lddqu_si128(__m128i const *__p) |
| 49 | { |
| 50 | return (__m128i)__builtin_ia32_lddqu((char const *)__p); |
| 51 | } |
| 52 | |
| 53 | /// Adds the even-indexed values and subtracts the odd-indexed values of |
| 54 | /// two 128-bit vectors of [4 x float]. |
| 55 | /// |
| 56 | /// \headerfile <x86intrin.h> |
| 57 | /// |
| 58 | /// This intrinsic corresponds to the <c> VADDSUBPS </c> instruction. |
| 59 | /// |
| 60 | /// \param __a |
| 61 | /// A 128-bit vector of [4 x float] containing the left source operand. |
| 62 | /// \param __b |
| 63 | /// A 128-bit vector of [4 x float] containing the right source operand. |
| 64 | /// \returns A 128-bit vector of [4 x float] containing the alternating sums and |
| 65 | /// differences of both operands. |
| 66 | static __inline__ __m128 __DEFAULT_FN_ATTRS |
| 67 | _mm_addsub_ps(__m128 __a, __m128 __b) |
| 68 | { |
| 69 | return __builtin_ia32_addsubps((__v4sf)__a, (__v4sf)__b); |
| 70 | } |
| 71 | |
| 72 | /// Horizontally adds the adjacent pairs of values contained in two |
| 73 | /// 128-bit vectors of [4 x float]. |
| 74 | /// |
| 75 | /// \headerfile <x86intrin.h> |
| 76 | /// |
| 77 | /// This intrinsic corresponds to the <c> VHADDPS </c> instruction. |
| 78 | /// |
| 79 | /// \param __a |
| 80 | /// A 128-bit vector of [4 x float] containing one of the source operands. |
| 81 | /// The horizontal sums of the values are stored in the lower bits of the |
| 82 | /// destination. |
| 83 | /// \param __b |
| 84 | /// A 128-bit vector of [4 x float] containing one of the source operands. |
| 85 | /// The horizontal sums of the values are stored in the upper bits of the |
| 86 | /// destination. |
| 87 | /// \returns A 128-bit vector of [4 x float] containing the horizontal sums of |
| 88 | /// both operands. |
| 89 | static __inline__ __m128 __DEFAULT_FN_ATTRS |
| 90 | _mm_hadd_ps(__m128 __a, __m128 __b) |
| 91 | { |
| 92 | return __builtin_ia32_haddps((__v4sf)__a, (__v4sf)__b); |
| 93 | } |
| 94 | |
| 95 | /// Horizontally subtracts the adjacent pairs of values contained in two |
| 96 | /// 128-bit vectors of [4 x float]. |
| 97 | /// |
| 98 | /// \headerfile <x86intrin.h> |
| 99 | /// |
| 100 | /// This intrinsic corresponds to the <c> VHSUBPS </c> instruction. |
| 101 | /// |
| 102 | /// \param __a |
| 103 | /// A 128-bit vector of [4 x float] containing one of the source operands. |
| 104 | /// The horizontal differences between the values are stored in the lower |
| 105 | /// bits of the destination. |
| 106 | /// \param __b |
| 107 | /// A 128-bit vector of [4 x float] containing one of the source operands. |
| 108 | /// The horizontal differences between the values are stored in the upper |
| 109 | /// bits of the destination. |
| 110 | /// \returns A 128-bit vector of [4 x float] containing the horizontal |
| 111 | /// differences of both operands. |
| 112 | static __inline__ __m128 __DEFAULT_FN_ATTRS |
| 113 | _mm_hsub_ps(__m128 __a, __m128 __b) |
| 114 | { |
| 115 | return __builtin_ia32_hsubps((__v4sf)__a, (__v4sf)__b); |
| 116 | } |
| 117 | |
| 118 | /// Moves and duplicates odd-indexed values from a 128-bit vector |
| 119 | /// of [4 x float] to float values stored in a 128-bit vector of |
| 120 | /// [4 x float]. |
| 121 | /// |
| 122 | /// \headerfile <x86intrin.h> |
| 123 | /// |
| 124 | /// This intrinsic corresponds to the <c> VMOVSHDUP </c> instruction. |
| 125 | /// |
| 126 | /// \param __a |
| 127 | /// A 128-bit vector of [4 x float]. \n |
| 128 | /// Bits [127:96] of the source are written to bits [127:96] and [95:64] of |
| 129 | /// the destination. \n |
| 130 | /// Bits [63:32] of the source are written to bits [63:32] and [31:0] of the |
| 131 | /// destination. |
| 132 | /// \returns A 128-bit vector of [4 x float] containing the moved and duplicated |
| 133 | /// values. |
| 134 | static __inline__ __m128 __DEFAULT_FN_ATTRS |
| 135 | _mm_movehdup_ps(__m128 __a) |
| 136 | { |
| 137 | return __builtin_shufflevector((__v4sf)__a, (__v4sf)__a, 1, 1, 3, 3); |
| 138 | } |
| 139 | |
| 140 | /// Duplicates even-indexed values from a 128-bit vector of |
| 141 | /// [4 x float] to float values stored in a 128-bit vector of [4 x float]. |
| 142 | /// |
| 143 | /// \headerfile <x86intrin.h> |
| 144 | /// |
| 145 | /// This intrinsic corresponds to the <c> VMOVSLDUP </c> instruction. |
| 146 | /// |
| 147 | /// \param __a |
| 148 | /// A 128-bit vector of [4 x float] \n |
| 149 | /// Bits [95:64] of the source are written to bits [127:96] and [95:64] of |
| 150 | /// the destination. \n |
| 151 | /// Bits [31:0] of the source are written to bits [63:32] and [31:0] of the |
| 152 | /// destination. |
| 153 | /// \returns A 128-bit vector of [4 x float] containing the moved and duplicated |
| 154 | /// values. |
| 155 | static __inline__ __m128 __DEFAULT_FN_ATTRS |
| 156 | _mm_moveldup_ps(__m128 __a) |
| 157 | { |
| 158 | return __builtin_shufflevector((__v4sf)__a, (__v4sf)__a, 0, 0, 2, 2); |
| 159 | } |
| 160 | |
| 161 | /// Adds the even-indexed values and subtracts the odd-indexed values of |
| 162 | /// two 128-bit vectors of [2 x double]. |
| 163 | /// |
| 164 | /// \headerfile <x86intrin.h> |
| 165 | /// |
| 166 | /// This intrinsic corresponds to the <c> VADDSUBPD </c> instruction. |
| 167 | /// |
| 168 | /// \param __a |
| 169 | /// A 128-bit vector of [2 x double] containing the left source operand. |
| 170 | /// \param __b |
| 171 | /// A 128-bit vector of [2 x double] containing the right source operand. |
| 172 | /// \returns A 128-bit vector of [2 x double] containing the alternating sums |
| 173 | /// and differences of both operands. |
| 174 | static __inline__ __m128d __DEFAULT_FN_ATTRS |
| 175 | _mm_addsub_pd(__m128d __a, __m128d __b) |
| 176 | { |
| 177 | return __builtin_ia32_addsubpd((__v2df)__a, (__v2df)__b); |
| 178 | } |
| 179 | |
| 180 | /// Horizontally adds the pairs of values contained in two 128-bit |
| 181 | /// vectors of [2 x double]. |
| 182 | /// |
| 183 | /// \headerfile <x86intrin.h> |
| 184 | /// |
| 185 | /// This intrinsic corresponds to the <c> VHADDPD </c> instruction. |
| 186 | /// |
| 187 | /// \param __a |
| 188 | /// A 128-bit vector of [2 x double] containing one of the source operands. |
| 189 | /// The horizontal sum of the values is stored in the lower bits of the |
| 190 | /// destination. |
| 191 | /// \param __b |
| 192 | /// A 128-bit vector of [2 x double] containing one of the source operands. |
| 193 | /// The horizontal sum of the values is stored in the upper bits of the |
| 194 | /// destination. |
| 195 | /// \returns A 128-bit vector of [2 x double] containing the horizontal sums of |
| 196 | /// both operands. |
| 197 | static __inline__ __m128d __DEFAULT_FN_ATTRS |
| 198 | _mm_hadd_pd(__m128d __a, __m128d __b) |
| 199 | { |
| 200 | return __builtin_ia32_haddpd((__v2df)__a, (__v2df)__b); |
| 201 | } |
| 202 | |
| 203 | /// Horizontally subtracts the pairs of values contained in two 128-bit |
| 204 | /// vectors of [2 x double]. |
| 205 | /// |
| 206 | /// \headerfile <x86intrin.h> |
| 207 | /// |
| 208 | /// This intrinsic corresponds to the <c> VHSUBPD </c> instruction. |
| 209 | /// |
| 210 | /// \param __a |
| 211 | /// A 128-bit vector of [2 x double] containing one of the source operands. |
| 212 | /// The horizontal difference of the values is stored in the lower bits of |
| 213 | /// the destination. |
| 214 | /// \param __b |
| 215 | /// A 128-bit vector of [2 x double] containing one of the source operands. |
| 216 | /// The horizontal difference of the values is stored in the upper bits of |
| 217 | /// the destination. |
| 218 | /// \returns A 128-bit vector of [2 x double] containing the horizontal |
| 219 | /// differences of both operands. |
| 220 | static __inline__ __m128d __DEFAULT_FN_ATTRS |
| 221 | _mm_hsub_pd(__m128d __a, __m128d __b) |
| 222 | { |
| 223 | return __builtin_ia32_hsubpd((__v2df)__a, (__v2df)__b); |
| 224 | } |
| 225 | |
| 226 | /// Moves and duplicates one double-precision value to double-precision |
| 227 | /// values stored in a 128-bit vector of [2 x double]. |
| 228 | /// |
| 229 | /// \headerfile <x86intrin.h> |
| 230 | /// |
| 231 | /// \code |
| 232 | /// __m128d _mm_loaddup_pd(double const *dp); |
| 233 | /// \endcode |
| 234 | /// |
| 235 | /// This intrinsic corresponds to the <c> VMOVDDUP </c> instruction. |
| 236 | /// |
| 237 | /// \param dp |
| 238 | /// A pointer to a double-precision value to be moved and duplicated. |
| 239 | /// \returns A 128-bit vector of [2 x double] containing the moved and |
| 240 | /// duplicated values. |
| 241 | #define _mm_loaddup_pd(dp) _mm_load1_pd(dp) |
| 242 | |
| 243 | /// Moves and duplicates the double-precision value in the lower bits of |
| 244 | /// a 128-bit vector of [2 x double] to double-precision values stored in a |
| 245 | /// 128-bit vector of [2 x double]. |
| 246 | /// |
| 247 | /// \headerfile <x86intrin.h> |
| 248 | /// |
| 249 | /// This intrinsic corresponds to the <c> VMOVDDUP </c> instruction. |
| 250 | /// |
| 251 | /// \param __a |
| 252 | /// A 128-bit vector of [2 x double]. Bits [63:0] are written to bits |
| 253 | /// [127:64] and [63:0] of the destination. |
| 254 | /// \returns A 128-bit vector of [2 x double] containing the moved and |
| 255 | /// duplicated values. |
| 256 | static __inline__ __m128d __DEFAULT_FN_ATTRS |
| 257 | _mm_movedup_pd(__m128d __a) |
| 258 | { |
| 259 | return __builtin_shufflevector((__v2df)__a, (__v2df)__a, 0, 0); |
| 260 | } |
| 261 | |
| 262 | /// Establishes a linear address memory range to be monitored and puts |
| 263 | /// the processor in the monitor event pending state. Data stored in the |
| 264 | /// monitored address range causes the processor to exit the pending state. |
| 265 | /// |
| 266 | /// \headerfile <x86intrin.h> |
| 267 | /// |
| 268 | /// This intrinsic corresponds to the <c> MONITOR </c> instruction. |
| 269 | /// |
| 270 | /// \param __p |
| 271 | /// The memory range to be monitored. The size of the range is determined by |
| 272 | /// CPUID function 0000_0005h. |
| 273 | /// \param __extensions |
| 274 | /// Optional extensions for the monitoring state. |
| 275 | /// \param __hints |
| 276 | /// Optional hints for the monitoring state. |
| 277 | static __inline__ void __DEFAULT_FN_ATTRS |
| 278 | _mm_monitor(void const *__p, unsigned __extensions, unsigned __hints) |
| 279 | { |
| 280 | __builtin_ia32_monitor((void *)__p, __extensions, __hints); |
| 281 | } |
| 282 | |
| 283 | /// Used with the MONITOR instruction to wait while the processor is in |
| 284 | /// the monitor event pending state. Data stored in the monitored address |
| 285 | /// range causes the processor to exit the pending state. |
| 286 | /// |
| 287 | /// \headerfile <x86intrin.h> |
| 288 | /// |
| 289 | /// This intrinsic corresponds to the <c> MWAIT </c> instruction. |
| 290 | /// |
| 291 | /// \param __extensions |
| 292 | /// Optional extensions for the monitoring state, which may vary by |
| 293 | /// processor. |
| 294 | /// \param __hints |
| 295 | /// Optional hints for the monitoring state, which may vary by processor. |
| 296 | static __inline__ void __DEFAULT_FN_ATTRS |
| 297 | _mm_mwait(unsigned __extensions, unsigned __hints) |
| 298 | { |
| 299 | __builtin_ia32_mwait(__extensions, __hints); |
| 300 | } |
| 301 | |
| 302 | #undef __DEFAULT_FN_ATTRS |
| 303 | |
| 304 | #endif /* __PMMINTRIN_H */ |
| 305 |