Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1 | /*===---- mmintrin.h - MMX intrinsics --------------------------------------=== |
| 2 | * |
Logan Chien | df4f766 | 2019-09-04 16:45:23 -0700 | [diff] [blame] | 3 | * Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. |
| 4 | * See https://llvm.org/LICENSE.txt for license information. |
| 5 | * SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 6 | * |
| 7 | *===-----------------------------------------------------------------------=== |
| 8 | */ |
| 9 | |
| 10 | #ifndef __MMINTRIN_H |
| 11 | #define __MMINTRIN_H |
| 12 | |
Pirama Arumuga Nainar | ec8c89d | 2022-02-23 09:26:16 -0800 | [diff] [blame] | 13 | #if !defined(__i386__) && !defined(__x86_64__) |
| 14 | #error "This header is only meant to be used on x86 and x64 architecture" |
| 15 | #endif |
| 16 | |
Logan Chien | dbcf412 | 2019-03-21 10:50:25 +0800 | [diff] [blame] | 17 | typedef long long __m64 __attribute__((__vector_size__(8), __aligned__(8))); |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 18 | |
| 19 | typedef long long __v1di __attribute__((__vector_size__(8))); |
| 20 | typedef int __v2si __attribute__((__vector_size__(8))); |
| 21 | typedef short __v4hi __attribute__((__vector_size__(8))); |
| 22 | typedef char __v8qi __attribute__((__vector_size__(8))); |
| 23 | |
| 24 | /* Define the default attributes for the functions in this file. */ |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 25 | #define __DEFAULT_FN_ATTRS __attribute__((__always_inline__, __nodebug__, __target__("mmx"), __min_vector_width__(64))) |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 26 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 27 | /// Clears the MMX state by setting the state of the x87 stack registers |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 28 | /// to empty. |
| 29 | /// |
| 30 | /// \headerfile <x86intrin.h> |
| 31 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 32 | /// This intrinsic corresponds to the <c> EMMS </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 33 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 34 | static __inline__ void __attribute__((__always_inline__, __nodebug__, __target__("mmx"))) |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 35 | _mm_empty(void) |
| 36 | { |
| 37 | __builtin_ia32_emms(); |
| 38 | } |
| 39 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 40 | /// Constructs a 64-bit integer vector, setting the lower 32 bits to the |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 41 | /// value of the 32-bit integer parameter and setting the upper 32 bits to 0. |
| 42 | /// |
| 43 | /// \headerfile <x86intrin.h> |
| 44 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 45 | /// This intrinsic corresponds to the <c> MOVD </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 46 | /// |
| 47 | /// \param __i |
| 48 | /// A 32-bit integer value. |
| 49 | /// \returns A 64-bit integer vector. The lower 32 bits contain the value of the |
| 50 | /// parameter. The upper 32 bits are set to 0. |
| 51 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 52 | _mm_cvtsi32_si64(int __i) |
| 53 | { |
| 54 | return (__m64)__builtin_ia32_vec_init_v2si(__i, 0); |
| 55 | } |
| 56 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 57 | /// Returns the lower 32 bits of a 64-bit integer vector as a 32-bit |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 58 | /// signed integer. |
| 59 | /// |
| 60 | /// \headerfile <x86intrin.h> |
| 61 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 62 | /// This intrinsic corresponds to the <c> MOVD </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 63 | /// |
| 64 | /// \param __m |
| 65 | /// A 64-bit integer vector. |
| 66 | /// \returns A 32-bit signed integer value containing the lower 32 bits of the |
| 67 | /// parameter. |
| 68 | static __inline__ int __DEFAULT_FN_ATTRS |
| 69 | _mm_cvtsi64_si32(__m64 __m) |
| 70 | { |
| 71 | return __builtin_ia32_vec_ext_v2si((__v2si)__m, 0); |
| 72 | } |
| 73 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 74 | /// Casts a 64-bit signed integer value into a 64-bit integer vector. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 75 | /// |
| 76 | /// \headerfile <x86intrin.h> |
| 77 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 78 | /// This intrinsic corresponds to the <c> MOVQ </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 79 | /// |
| 80 | /// \param __i |
| 81 | /// A 64-bit signed integer. |
| 82 | /// \returns A 64-bit integer vector containing the same bitwise pattern as the |
| 83 | /// parameter. |
| 84 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 85 | _mm_cvtsi64_m64(long long __i) |
| 86 | { |
| 87 | return (__m64)__i; |
| 88 | } |
| 89 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 90 | /// Casts a 64-bit integer vector into a 64-bit signed integer value. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 91 | /// |
| 92 | /// \headerfile <x86intrin.h> |
| 93 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 94 | /// This intrinsic corresponds to the <c> MOVQ </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 95 | /// |
| 96 | /// \param __m |
| 97 | /// A 64-bit integer vector. |
| 98 | /// \returns A 64-bit signed integer containing the same bitwise pattern as the |
| 99 | /// parameter. |
| 100 | static __inline__ long long __DEFAULT_FN_ATTRS |
| 101 | _mm_cvtm64_si64(__m64 __m) |
| 102 | { |
| 103 | return (long long)__m; |
| 104 | } |
| 105 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 106 | /// Converts 16-bit signed integers from both 64-bit integer vector |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 107 | /// parameters of [4 x i16] into 8-bit signed integer values, and constructs |
| 108 | /// a 64-bit integer vector of [8 x i8] as the result. Positive values |
| 109 | /// greater than 0x7F are saturated to 0x7F. Negative values less than 0x80 |
| 110 | /// are saturated to 0x80. |
| 111 | /// |
| 112 | /// \headerfile <x86intrin.h> |
| 113 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 114 | /// This intrinsic corresponds to the <c> PACKSSWB </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 115 | /// |
| 116 | /// \param __m1 |
| 117 | /// A 64-bit integer vector of [4 x i16]. Each 16-bit element is treated as a |
| 118 | /// 16-bit signed integer and is converted to an 8-bit signed integer with |
| 119 | /// saturation. Positive values greater than 0x7F are saturated to 0x7F. |
| 120 | /// Negative values less than 0x80 are saturated to 0x80. The converted |
| 121 | /// [4 x i8] values are written to the lower 32 bits of the result. |
| 122 | /// \param __m2 |
| 123 | /// A 64-bit integer vector of [4 x i16]. Each 16-bit element is treated as a |
| 124 | /// 16-bit signed integer and is converted to an 8-bit signed integer with |
| 125 | /// saturation. Positive values greater than 0x7F are saturated to 0x7F. |
| 126 | /// Negative values less than 0x80 are saturated to 0x80. The converted |
| 127 | /// [4 x i8] values are written to the upper 32 bits of the result. |
| 128 | /// \returns A 64-bit integer vector of [8 x i8] containing the converted |
| 129 | /// values. |
| 130 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 131 | _mm_packs_pi16(__m64 __m1, __m64 __m2) |
| 132 | { |
| 133 | return (__m64)__builtin_ia32_packsswb((__v4hi)__m1, (__v4hi)__m2); |
| 134 | } |
| 135 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 136 | /// Converts 32-bit signed integers from both 64-bit integer vector |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 137 | /// parameters of [2 x i32] into 16-bit signed integer values, and constructs |
| 138 | /// a 64-bit integer vector of [4 x i16] as the result. Positive values |
| 139 | /// greater than 0x7FFF are saturated to 0x7FFF. Negative values less than |
| 140 | /// 0x8000 are saturated to 0x8000. |
| 141 | /// |
| 142 | /// \headerfile <x86intrin.h> |
| 143 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 144 | /// This intrinsic corresponds to the <c> PACKSSDW </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 145 | /// |
| 146 | /// \param __m1 |
| 147 | /// A 64-bit integer vector of [2 x i32]. Each 32-bit element is treated as a |
| 148 | /// 32-bit signed integer and is converted to a 16-bit signed integer with |
| 149 | /// saturation. Positive values greater than 0x7FFF are saturated to 0x7FFF. |
| 150 | /// Negative values less than 0x8000 are saturated to 0x8000. The converted |
| 151 | /// [2 x i16] values are written to the lower 32 bits of the result. |
| 152 | /// \param __m2 |
| 153 | /// A 64-bit integer vector of [2 x i32]. Each 32-bit element is treated as a |
| 154 | /// 32-bit signed integer and is converted to a 16-bit signed integer with |
| 155 | /// saturation. Positive values greater than 0x7FFF are saturated to 0x7FFF. |
| 156 | /// Negative values less than 0x8000 are saturated to 0x8000. The converted |
| 157 | /// [2 x i16] values are written to the upper 32 bits of the result. |
| 158 | /// \returns A 64-bit integer vector of [4 x i16] containing the converted |
| 159 | /// values. |
| 160 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 161 | _mm_packs_pi32(__m64 __m1, __m64 __m2) |
| 162 | { |
| 163 | return (__m64)__builtin_ia32_packssdw((__v2si)__m1, (__v2si)__m2); |
| 164 | } |
| 165 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 166 | /// Converts 16-bit signed integers from both 64-bit integer vector |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 167 | /// parameters of [4 x i16] into 8-bit unsigned integer values, and |
| 168 | /// constructs a 64-bit integer vector of [8 x i8] as the result. Values |
| 169 | /// greater than 0xFF are saturated to 0xFF. Values less than 0 are saturated |
| 170 | /// to 0. |
| 171 | /// |
| 172 | /// \headerfile <x86intrin.h> |
| 173 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 174 | /// This intrinsic corresponds to the <c> PACKUSWB </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 175 | /// |
| 176 | /// \param __m1 |
| 177 | /// A 64-bit integer vector of [4 x i16]. Each 16-bit element is treated as a |
| 178 | /// 16-bit signed integer and is converted to an 8-bit unsigned integer with |
| 179 | /// saturation. Values greater than 0xFF are saturated to 0xFF. Values less |
| 180 | /// than 0 are saturated to 0. The converted [4 x i8] values are written to |
| 181 | /// the lower 32 bits of the result. |
| 182 | /// \param __m2 |
| 183 | /// A 64-bit integer vector of [4 x i16]. Each 16-bit element is treated as a |
| 184 | /// 16-bit signed integer and is converted to an 8-bit unsigned integer with |
| 185 | /// saturation. Values greater than 0xFF are saturated to 0xFF. Values less |
| 186 | /// than 0 are saturated to 0. The converted [4 x i8] values are written to |
| 187 | /// the upper 32 bits of the result. |
| 188 | /// \returns A 64-bit integer vector of [8 x i8] containing the converted |
| 189 | /// values. |
| 190 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 191 | _mm_packs_pu16(__m64 __m1, __m64 __m2) |
| 192 | { |
| 193 | return (__m64)__builtin_ia32_packuswb((__v4hi)__m1, (__v4hi)__m2); |
| 194 | } |
| 195 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 196 | /// Unpacks the upper 32 bits from two 64-bit integer vectors of [8 x i8] |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 197 | /// and interleaves them into a 64-bit integer vector of [8 x i8]. |
| 198 | /// |
| 199 | /// \headerfile <x86intrin.h> |
| 200 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 201 | /// This intrinsic corresponds to the <c> PUNPCKHBW </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 202 | /// |
| 203 | /// \param __m1 |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 204 | /// A 64-bit integer vector of [8 x i8]. \n |
| 205 | /// Bits [39:32] are written to bits [7:0] of the result. \n |
| 206 | /// Bits [47:40] are written to bits [23:16] of the result. \n |
| 207 | /// Bits [55:48] are written to bits [39:32] of the result. \n |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 208 | /// Bits [63:56] are written to bits [55:48] of the result. |
| 209 | /// \param __m2 |
| 210 | /// A 64-bit integer vector of [8 x i8]. |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 211 | /// Bits [39:32] are written to bits [15:8] of the result. \n |
| 212 | /// Bits [47:40] are written to bits [31:24] of the result. \n |
| 213 | /// Bits [55:48] are written to bits [47:40] of the result. \n |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 214 | /// Bits [63:56] are written to bits [63:56] of the result. |
| 215 | /// \returns A 64-bit integer vector of [8 x i8] containing the interleaved |
| 216 | /// values. |
| 217 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 218 | _mm_unpackhi_pi8(__m64 __m1, __m64 __m2) |
| 219 | { |
| 220 | return (__m64)__builtin_ia32_punpckhbw((__v8qi)__m1, (__v8qi)__m2); |
| 221 | } |
| 222 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 223 | /// Unpacks the upper 32 bits from two 64-bit integer vectors of |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 224 | /// [4 x i16] and interleaves them into a 64-bit integer vector of [4 x i16]. |
| 225 | /// |
| 226 | /// \headerfile <x86intrin.h> |
| 227 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 228 | /// This intrinsic corresponds to the <c> PUNPCKHWD </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 229 | /// |
| 230 | /// \param __m1 |
| 231 | /// A 64-bit integer vector of [4 x i16]. |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 232 | /// Bits [47:32] are written to bits [15:0] of the result. \n |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 233 | /// Bits [63:48] are written to bits [47:32] of the result. |
| 234 | /// \param __m2 |
| 235 | /// A 64-bit integer vector of [4 x i16]. |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 236 | /// Bits [47:32] are written to bits [31:16] of the result. \n |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 237 | /// Bits [63:48] are written to bits [63:48] of the result. |
| 238 | /// \returns A 64-bit integer vector of [4 x i16] containing the interleaved |
| 239 | /// values. |
| 240 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 241 | _mm_unpackhi_pi16(__m64 __m1, __m64 __m2) |
| 242 | { |
| 243 | return (__m64)__builtin_ia32_punpckhwd((__v4hi)__m1, (__v4hi)__m2); |
| 244 | } |
| 245 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 246 | /// Unpacks the upper 32 bits from two 64-bit integer vectors of |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 247 | /// [2 x i32] and interleaves them into a 64-bit integer vector of [2 x i32]. |
| 248 | /// |
| 249 | /// \headerfile <x86intrin.h> |
| 250 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 251 | /// This intrinsic corresponds to the <c> PUNPCKHDQ </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 252 | /// |
| 253 | /// \param __m1 |
| 254 | /// A 64-bit integer vector of [2 x i32]. The upper 32 bits are written to |
| 255 | /// the lower 32 bits of the result. |
| 256 | /// \param __m2 |
| 257 | /// A 64-bit integer vector of [2 x i32]. The upper 32 bits are written to |
| 258 | /// the upper 32 bits of the result. |
| 259 | /// \returns A 64-bit integer vector of [2 x i32] containing the interleaved |
| 260 | /// values. |
| 261 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 262 | _mm_unpackhi_pi32(__m64 __m1, __m64 __m2) |
| 263 | { |
| 264 | return (__m64)__builtin_ia32_punpckhdq((__v2si)__m1, (__v2si)__m2); |
| 265 | } |
| 266 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 267 | /// Unpacks the lower 32 bits from two 64-bit integer vectors of [8 x i8] |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 268 | /// and interleaves them into a 64-bit integer vector of [8 x i8]. |
| 269 | /// |
| 270 | /// \headerfile <x86intrin.h> |
| 271 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 272 | /// This intrinsic corresponds to the <c> PUNPCKLBW </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 273 | /// |
| 274 | /// \param __m1 |
| 275 | /// A 64-bit integer vector of [8 x i8]. |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 276 | /// Bits [7:0] are written to bits [7:0] of the result. \n |
| 277 | /// Bits [15:8] are written to bits [23:16] of the result. \n |
| 278 | /// Bits [23:16] are written to bits [39:32] of the result. \n |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 279 | /// Bits [31:24] are written to bits [55:48] of the result. |
| 280 | /// \param __m2 |
| 281 | /// A 64-bit integer vector of [8 x i8]. |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 282 | /// Bits [7:0] are written to bits [15:8] of the result. \n |
| 283 | /// Bits [15:8] are written to bits [31:24] of the result. \n |
| 284 | /// Bits [23:16] are written to bits [47:40] of the result. \n |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 285 | /// Bits [31:24] are written to bits [63:56] of the result. |
| 286 | /// \returns A 64-bit integer vector of [8 x i8] containing the interleaved |
| 287 | /// values. |
| 288 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 289 | _mm_unpacklo_pi8(__m64 __m1, __m64 __m2) |
| 290 | { |
| 291 | return (__m64)__builtin_ia32_punpcklbw((__v8qi)__m1, (__v8qi)__m2); |
| 292 | } |
| 293 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 294 | /// Unpacks the lower 32 bits from two 64-bit integer vectors of |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 295 | /// [4 x i16] and interleaves them into a 64-bit integer vector of [4 x i16]. |
| 296 | /// |
| 297 | /// \headerfile <x86intrin.h> |
| 298 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 299 | /// This intrinsic corresponds to the <c> PUNPCKLWD </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 300 | /// |
| 301 | /// \param __m1 |
| 302 | /// A 64-bit integer vector of [4 x i16]. |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 303 | /// Bits [15:0] are written to bits [15:0] of the result. \n |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 304 | /// Bits [31:16] are written to bits [47:32] of the result. |
| 305 | /// \param __m2 |
| 306 | /// A 64-bit integer vector of [4 x i16]. |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 307 | /// Bits [15:0] are written to bits [31:16] of the result. \n |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 308 | /// Bits [31:16] are written to bits [63:48] of the result. |
| 309 | /// \returns A 64-bit integer vector of [4 x i16] containing the interleaved |
| 310 | /// values. |
| 311 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 312 | _mm_unpacklo_pi16(__m64 __m1, __m64 __m2) |
| 313 | { |
| 314 | return (__m64)__builtin_ia32_punpcklwd((__v4hi)__m1, (__v4hi)__m2); |
| 315 | } |
| 316 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 317 | /// Unpacks the lower 32 bits from two 64-bit integer vectors of |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 318 | /// [2 x i32] and interleaves them into a 64-bit integer vector of [2 x i32]. |
| 319 | /// |
| 320 | /// \headerfile <x86intrin.h> |
| 321 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 322 | /// This intrinsic corresponds to the <c> PUNPCKLDQ </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 323 | /// |
| 324 | /// \param __m1 |
| 325 | /// A 64-bit integer vector of [2 x i32]. The lower 32 bits are written to |
| 326 | /// the lower 32 bits of the result. |
| 327 | /// \param __m2 |
| 328 | /// A 64-bit integer vector of [2 x i32]. The lower 32 bits are written to |
| 329 | /// the upper 32 bits of the result. |
| 330 | /// \returns A 64-bit integer vector of [2 x i32] containing the interleaved |
| 331 | /// values. |
| 332 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 333 | _mm_unpacklo_pi32(__m64 __m1, __m64 __m2) |
| 334 | { |
| 335 | return (__m64)__builtin_ia32_punpckldq((__v2si)__m1, (__v2si)__m2); |
| 336 | } |
| 337 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 338 | /// Adds each 8-bit integer element of the first 64-bit integer vector |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 339 | /// of [8 x i8] to the corresponding 8-bit integer element of the second |
| 340 | /// 64-bit integer vector of [8 x i8]. The lower 8 bits of the results are |
| 341 | /// packed into a 64-bit integer vector of [8 x i8]. |
| 342 | /// |
| 343 | /// \headerfile <x86intrin.h> |
| 344 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 345 | /// This intrinsic corresponds to the <c> PADDB </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 346 | /// |
| 347 | /// \param __m1 |
| 348 | /// A 64-bit integer vector of [8 x i8]. |
| 349 | /// \param __m2 |
| 350 | /// A 64-bit integer vector of [8 x i8]. |
| 351 | /// \returns A 64-bit integer vector of [8 x i8] containing the sums of both |
| 352 | /// parameters. |
| 353 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 354 | _mm_add_pi8(__m64 __m1, __m64 __m2) |
| 355 | { |
| 356 | return (__m64)__builtin_ia32_paddb((__v8qi)__m1, (__v8qi)__m2); |
| 357 | } |
| 358 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 359 | /// Adds each 16-bit integer element of the first 64-bit integer vector |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 360 | /// of [4 x i16] to the corresponding 16-bit integer element of the second |
| 361 | /// 64-bit integer vector of [4 x i16]. The lower 16 bits of the results are |
| 362 | /// packed into a 64-bit integer vector of [4 x i16]. |
| 363 | /// |
| 364 | /// \headerfile <x86intrin.h> |
| 365 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 366 | /// This intrinsic corresponds to the <c> PADDW </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 367 | /// |
| 368 | /// \param __m1 |
| 369 | /// A 64-bit integer vector of [4 x i16]. |
| 370 | /// \param __m2 |
| 371 | /// A 64-bit integer vector of [4 x i16]. |
| 372 | /// \returns A 64-bit integer vector of [4 x i16] containing the sums of both |
| 373 | /// parameters. |
| 374 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 375 | _mm_add_pi16(__m64 __m1, __m64 __m2) |
| 376 | { |
| 377 | return (__m64)__builtin_ia32_paddw((__v4hi)__m1, (__v4hi)__m2); |
| 378 | } |
| 379 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 380 | /// Adds each 32-bit integer element of the first 64-bit integer vector |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 381 | /// of [2 x i32] to the corresponding 32-bit integer element of the second |
| 382 | /// 64-bit integer vector of [2 x i32]. The lower 32 bits of the results are |
| 383 | /// packed into a 64-bit integer vector of [2 x i32]. |
| 384 | /// |
| 385 | /// \headerfile <x86intrin.h> |
| 386 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 387 | /// This intrinsic corresponds to the <c> PADDD </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 388 | /// |
| 389 | /// \param __m1 |
| 390 | /// A 64-bit integer vector of [2 x i32]. |
| 391 | /// \param __m2 |
| 392 | /// A 64-bit integer vector of [2 x i32]. |
| 393 | /// \returns A 64-bit integer vector of [2 x i32] containing the sums of both |
| 394 | /// parameters. |
| 395 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 396 | _mm_add_pi32(__m64 __m1, __m64 __m2) |
| 397 | { |
| 398 | return (__m64)__builtin_ia32_paddd((__v2si)__m1, (__v2si)__m2); |
| 399 | } |
| 400 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 401 | /// Adds each 8-bit signed integer element of the first 64-bit integer |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 402 | /// vector of [8 x i8] to the corresponding 8-bit signed integer element of |
| 403 | /// the second 64-bit integer vector of [8 x i8]. Positive sums greater than |
| 404 | /// 0x7F are saturated to 0x7F. Negative sums less than 0x80 are saturated to |
| 405 | /// 0x80. The results are packed into a 64-bit integer vector of [8 x i8]. |
| 406 | /// |
| 407 | /// \headerfile <x86intrin.h> |
| 408 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 409 | /// This intrinsic corresponds to the <c> PADDSB </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 410 | /// |
| 411 | /// \param __m1 |
| 412 | /// A 64-bit integer vector of [8 x i8]. |
| 413 | /// \param __m2 |
| 414 | /// A 64-bit integer vector of [8 x i8]. |
| 415 | /// \returns A 64-bit integer vector of [8 x i8] containing the saturated sums |
| 416 | /// of both parameters. |
| 417 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 418 | _mm_adds_pi8(__m64 __m1, __m64 __m2) |
| 419 | { |
| 420 | return (__m64)__builtin_ia32_paddsb((__v8qi)__m1, (__v8qi)__m2); |
| 421 | } |
| 422 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 423 | /// Adds each 16-bit signed integer element of the first 64-bit integer |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 424 | /// vector of [4 x i16] to the corresponding 16-bit signed integer element of |
| 425 | /// the second 64-bit integer vector of [4 x i16]. Positive sums greater than |
| 426 | /// 0x7FFF are saturated to 0x7FFF. Negative sums less than 0x8000 are |
| 427 | /// saturated to 0x8000. The results are packed into a 64-bit integer vector |
| 428 | /// of [4 x i16]. |
| 429 | /// |
| 430 | /// \headerfile <x86intrin.h> |
| 431 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 432 | /// This intrinsic corresponds to the <c> PADDSW </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 433 | /// |
| 434 | /// \param __m1 |
| 435 | /// A 64-bit integer vector of [4 x i16]. |
| 436 | /// \param __m2 |
| 437 | /// A 64-bit integer vector of [4 x i16]. |
| 438 | /// \returns A 64-bit integer vector of [4 x i16] containing the saturated sums |
| 439 | /// of both parameters. |
| 440 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 441 | _mm_adds_pi16(__m64 __m1, __m64 __m2) |
| 442 | { |
| 443 | return (__m64)__builtin_ia32_paddsw((__v4hi)__m1, (__v4hi)__m2); |
| 444 | } |
| 445 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 446 | /// Adds each 8-bit unsigned integer element of the first 64-bit integer |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 447 | /// vector of [8 x i8] to the corresponding 8-bit unsigned integer element of |
| 448 | /// the second 64-bit integer vector of [8 x i8]. Sums greater than 0xFF are |
| 449 | /// saturated to 0xFF. The results are packed into a 64-bit integer vector of |
| 450 | /// [8 x i8]. |
| 451 | /// |
| 452 | /// \headerfile <x86intrin.h> |
| 453 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 454 | /// This intrinsic corresponds to the <c> PADDUSB </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 455 | /// |
| 456 | /// \param __m1 |
| 457 | /// A 64-bit integer vector of [8 x i8]. |
| 458 | /// \param __m2 |
| 459 | /// A 64-bit integer vector of [8 x i8]. |
| 460 | /// \returns A 64-bit integer vector of [8 x i8] containing the saturated |
| 461 | /// unsigned sums of both parameters. |
| 462 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 463 | _mm_adds_pu8(__m64 __m1, __m64 __m2) |
| 464 | { |
| 465 | return (__m64)__builtin_ia32_paddusb((__v8qi)__m1, (__v8qi)__m2); |
| 466 | } |
| 467 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 468 | /// Adds each 16-bit unsigned integer element of the first 64-bit integer |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 469 | /// vector of [4 x i16] to the corresponding 16-bit unsigned integer element |
| 470 | /// of the second 64-bit integer vector of [4 x i16]. Sums greater than |
| 471 | /// 0xFFFF are saturated to 0xFFFF. The results are packed into a 64-bit |
| 472 | /// integer vector of [4 x i16]. |
| 473 | /// |
| 474 | /// \headerfile <x86intrin.h> |
| 475 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 476 | /// This intrinsic corresponds to the <c> PADDUSW </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 477 | /// |
| 478 | /// \param __m1 |
| 479 | /// A 64-bit integer vector of [4 x i16]. |
| 480 | /// \param __m2 |
| 481 | /// A 64-bit integer vector of [4 x i16]. |
| 482 | /// \returns A 64-bit integer vector of [4 x i16] containing the saturated |
| 483 | /// unsigned sums of both parameters. |
| 484 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 485 | _mm_adds_pu16(__m64 __m1, __m64 __m2) |
| 486 | { |
| 487 | return (__m64)__builtin_ia32_paddusw((__v4hi)__m1, (__v4hi)__m2); |
| 488 | } |
| 489 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 490 | /// Subtracts each 8-bit integer element of the second 64-bit integer |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 491 | /// vector of [8 x i8] from the corresponding 8-bit integer element of the |
| 492 | /// first 64-bit integer vector of [8 x i8]. The lower 8 bits of the results |
| 493 | /// are packed into a 64-bit integer vector of [8 x i8]. |
| 494 | /// |
| 495 | /// \headerfile <x86intrin.h> |
| 496 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 497 | /// This intrinsic corresponds to the <c> PSUBB </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 498 | /// |
| 499 | /// \param __m1 |
| 500 | /// A 64-bit integer vector of [8 x i8] containing the minuends. |
| 501 | /// \param __m2 |
| 502 | /// A 64-bit integer vector of [8 x i8] containing the subtrahends. |
| 503 | /// \returns A 64-bit integer vector of [8 x i8] containing the differences of |
| 504 | /// both parameters. |
| 505 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 506 | _mm_sub_pi8(__m64 __m1, __m64 __m2) |
| 507 | { |
| 508 | return (__m64)__builtin_ia32_psubb((__v8qi)__m1, (__v8qi)__m2); |
| 509 | } |
| 510 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 511 | /// Subtracts each 16-bit integer element of the second 64-bit integer |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 512 | /// vector of [4 x i16] from the corresponding 16-bit integer element of the |
| 513 | /// first 64-bit integer vector of [4 x i16]. The lower 16 bits of the |
| 514 | /// results are packed into a 64-bit integer vector of [4 x i16]. |
| 515 | /// |
| 516 | /// \headerfile <x86intrin.h> |
| 517 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 518 | /// This intrinsic corresponds to the <c> PSUBW </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 519 | /// |
| 520 | /// \param __m1 |
| 521 | /// A 64-bit integer vector of [4 x i16] containing the minuends. |
| 522 | /// \param __m2 |
| 523 | /// A 64-bit integer vector of [4 x i16] containing the subtrahends. |
| 524 | /// \returns A 64-bit integer vector of [4 x i16] containing the differences of |
| 525 | /// both parameters. |
| 526 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 527 | _mm_sub_pi16(__m64 __m1, __m64 __m2) |
| 528 | { |
| 529 | return (__m64)__builtin_ia32_psubw((__v4hi)__m1, (__v4hi)__m2); |
| 530 | } |
| 531 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 532 | /// Subtracts each 32-bit integer element of the second 64-bit integer |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 533 | /// vector of [2 x i32] from the corresponding 32-bit integer element of the |
| 534 | /// first 64-bit integer vector of [2 x i32]. The lower 32 bits of the |
| 535 | /// results are packed into a 64-bit integer vector of [2 x i32]. |
| 536 | /// |
| 537 | /// \headerfile <x86intrin.h> |
| 538 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 539 | /// This intrinsic corresponds to the <c> PSUBD </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 540 | /// |
| 541 | /// \param __m1 |
| 542 | /// A 64-bit integer vector of [2 x i32] containing the minuends. |
| 543 | /// \param __m2 |
| 544 | /// A 64-bit integer vector of [2 x i32] containing the subtrahends. |
| 545 | /// \returns A 64-bit integer vector of [2 x i32] containing the differences of |
| 546 | /// both parameters. |
| 547 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 548 | _mm_sub_pi32(__m64 __m1, __m64 __m2) |
| 549 | { |
| 550 | return (__m64)__builtin_ia32_psubd((__v2si)__m1, (__v2si)__m2); |
| 551 | } |
| 552 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 553 | /// Subtracts each 8-bit signed integer element of the second 64-bit |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 554 | /// integer vector of [8 x i8] from the corresponding 8-bit signed integer |
| 555 | /// element of the first 64-bit integer vector of [8 x i8]. Positive results |
| 556 | /// greater than 0x7F are saturated to 0x7F. Negative results less than 0x80 |
| 557 | /// are saturated to 0x80. The results are packed into a 64-bit integer |
| 558 | /// vector of [8 x i8]. |
| 559 | /// |
| 560 | /// \headerfile <x86intrin.h> |
| 561 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 562 | /// This intrinsic corresponds to the <c> PSUBSB </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 563 | /// |
| 564 | /// \param __m1 |
| 565 | /// A 64-bit integer vector of [8 x i8] containing the minuends. |
| 566 | /// \param __m2 |
| 567 | /// A 64-bit integer vector of [8 x i8] containing the subtrahends. |
| 568 | /// \returns A 64-bit integer vector of [8 x i8] containing the saturated |
| 569 | /// differences of both parameters. |
| 570 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 571 | _mm_subs_pi8(__m64 __m1, __m64 __m2) |
| 572 | { |
| 573 | return (__m64)__builtin_ia32_psubsb((__v8qi)__m1, (__v8qi)__m2); |
| 574 | } |
| 575 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 576 | /// Subtracts each 16-bit signed integer element of the second 64-bit |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 577 | /// integer vector of [4 x i16] from the corresponding 16-bit signed integer |
| 578 | /// element of the first 64-bit integer vector of [4 x i16]. Positive results |
| 579 | /// greater than 0x7FFF are saturated to 0x7FFF. Negative results less than |
| 580 | /// 0x8000 are saturated to 0x8000. The results are packed into a 64-bit |
| 581 | /// integer vector of [4 x i16]. |
| 582 | /// |
| 583 | /// \headerfile <x86intrin.h> |
| 584 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 585 | /// This intrinsic corresponds to the <c> PSUBSW </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 586 | /// |
| 587 | /// \param __m1 |
| 588 | /// A 64-bit integer vector of [4 x i16] containing the minuends. |
| 589 | /// \param __m2 |
| 590 | /// A 64-bit integer vector of [4 x i16] containing the subtrahends. |
| 591 | /// \returns A 64-bit integer vector of [4 x i16] containing the saturated |
| 592 | /// differences of both parameters. |
| 593 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 594 | _mm_subs_pi16(__m64 __m1, __m64 __m2) |
| 595 | { |
| 596 | return (__m64)__builtin_ia32_psubsw((__v4hi)__m1, (__v4hi)__m2); |
| 597 | } |
| 598 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 599 | /// Subtracts each 8-bit unsigned integer element of the second 64-bit |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 600 | /// integer vector of [8 x i8] from the corresponding 8-bit unsigned integer |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 601 | /// element of the first 64-bit integer vector of [8 x i8]. |
| 602 | /// |
| 603 | /// If an element of the first vector is less than the corresponding element |
| 604 | /// of the second vector, the result is saturated to 0. The results are |
| 605 | /// packed into a 64-bit integer vector of [8 x i8]. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 606 | /// |
| 607 | /// \headerfile <x86intrin.h> |
| 608 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 609 | /// This intrinsic corresponds to the <c> PSUBUSB </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 610 | /// |
| 611 | /// \param __m1 |
| 612 | /// A 64-bit integer vector of [8 x i8] containing the minuends. |
| 613 | /// \param __m2 |
| 614 | /// A 64-bit integer vector of [8 x i8] containing the subtrahends. |
| 615 | /// \returns A 64-bit integer vector of [8 x i8] containing the saturated |
| 616 | /// differences of both parameters. |
| 617 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 618 | _mm_subs_pu8(__m64 __m1, __m64 __m2) |
| 619 | { |
| 620 | return (__m64)__builtin_ia32_psubusb((__v8qi)__m1, (__v8qi)__m2); |
| 621 | } |
| 622 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 623 | /// Subtracts each 16-bit unsigned integer element of the second 64-bit |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 624 | /// integer vector of [4 x i16] from the corresponding 16-bit unsigned |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 625 | /// integer element of the first 64-bit integer vector of [4 x i16]. |
| 626 | /// |
| 627 | /// If an element of the first vector is less than the corresponding element |
| 628 | /// of the second vector, the result is saturated to 0. The results are |
| 629 | /// packed into a 64-bit integer vector of [4 x i16]. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 630 | /// |
| 631 | /// \headerfile <x86intrin.h> |
| 632 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 633 | /// This intrinsic corresponds to the <c> PSUBUSW </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 634 | /// |
| 635 | /// \param __m1 |
| 636 | /// A 64-bit integer vector of [4 x i16] containing the minuends. |
| 637 | /// \param __m2 |
| 638 | /// A 64-bit integer vector of [4 x i16] containing the subtrahends. |
| 639 | /// \returns A 64-bit integer vector of [4 x i16] containing the saturated |
| 640 | /// differences of both parameters. |
| 641 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 642 | _mm_subs_pu16(__m64 __m1, __m64 __m2) |
| 643 | { |
| 644 | return (__m64)__builtin_ia32_psubusw((__v4hi)__m1, (__v4hi)__m2); |
| 645 | } |
| 646 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 647 | /// Multiplies each 16-bit signed integer element of the first 64-bit |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 648 | /// integer vector of [4 x i16] by the corresponding 16-bit signed integer |
| 649 | /// element of the second 64-bit integer vector of [4 x i16] and get four |
| 650 | /// 32-bit products. Adds adjacent pairs of products to get two 32-bit sums. |
| 651 | /// The lower 32 bits of these two sums are packed into a 64-bit integer |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 652 | /// vector of [2 x i32]. |
| 653 | /// |
| 654 | /// For example, bits [15:0] of both parameters are multiplied, bits [31:16] |
| 655 | /// of both parameters are multiplied, and the sum of both results is written |
| 656 | /// to bits [31:0] of the result. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 657 | /// |
| 658 | /// \headerfile <x86intrin.h> |
| 659 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 660 | /// This intrinsic corresponds to the <c> PMADDWD </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 661 | /// |
| 662 | /// \param __m1 |
| 663 | /// A 64-bit integer vector of [4 x i16]. |
| 664 | /// \param __m2 |
| 665 | /// A 64-bit integer vector of [4 x i16]. |
| 666 | /// \returns A 64-bit integer vector of [2 x i32] containing the sums of |
| 667 | /// products of both parameters. |
| 668 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 669 | _mm_madd_pi16(__m64 __m1, __m64 __m2) |
| 670 | { |
| 671 | return (__m64)__builtin_ia32_pmaddwd((__v4hi)__m1, (__v4hi)__m2); |
| 672 | } |
| 673 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 674 | /// Multiplies each 16-bit signed integer element of the first 64-bit |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 675 | /// integer vector of [4 x i16] by the corresponding 16-bit signed integer |
| 676 | /// element of the second 64-bit integer vector of [4 x i16]. Packs the upper |
| 677 | /// 16 bits of the 32-bit products into a 64-bit integer vector of [4 x i16]. |
| 678 | /// |
| 679 | /// \headerfile <x86intrin.h> |
| 680 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 681 | /// This intrinsic corresponds to the <c> PMULHW </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 682 | /// |
| 683 | /// \param __m1 |
| 684 | /// A 64-bit integer vector of [4 x i16]. |
| 685 | /// \param __m2 |
| 686 | /// A 64-bit integer vector of [4 x i16]. |
| 687 | /// \returns A 64-bit integer vector of [4 x i16] containing the upper 16 bits |
| 688 | /// of the products of both parameters. |
| 689 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 690 | _mm_mulhi_pi16(__m64 __m1, __m64 __m2) |
| 691 | { |
| 692 | return (__m64)__builtin_ia32_pmulhw((__v4hi)__m1, (__v4hi)__m2); |
| 693 | } |
| 694 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 695 | /// Multiplies each 16-bit signed integer element of the first 64-bit |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 696 | /// integer vector of [4 x i16] by the corresponding 16-bit signed integer |
| 697 | /// element of the second 64-bit integer vector of [4 x i16]. Packs the lower |
| 698 | /// 16 bits of the 32-bit products into a 64-bit integer vector of [4 x i16]. |
| 699 | /// |
| 700 | /// \headerfile <x86intrin.h> |
| 701 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 702 | /// This intrinsic corresponds to the <c> PMULLW </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 703 | /// |
| 704 | /// \param __m1 |
| 705 | /// A 64-bit integer vector of [4 x i16]. |
| 706 | /// \param __m2 |
| 707 | /// A 64-bit integer vector of [4 x i16]. |
| 708 | /// \returns A 64-bit integer vector of [4 x i16] containing the lower 16 bits |
| 709 | /// of the products of both parameters. |
| 710 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 711 | _mm_mullo_pi16(__m64 __m1, __m64 __m2) |
| 712 | { |
| 713 | return (__m64)__builtin_ia32_pmullw((__v4hi)__m1, (__v4hi)__m2); |
| 714 | } |
| 715 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 716 | /// Left-shifts each 16-bit signed integer element of the first |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 717 | /// parameter, which is a 64-bit integer vector of [4 x i16], by the number |
| 718 | /// of bits specified by the second parameter, which is a 64-bit integer. The |
| 719 | /// lower 16 bits of the results are packed into a 64-bit integer vector of |
| 720 | /// [4 x i16]. |
| 721 | /// |
| 722 | /// \headerfile <x86intrin.h> |
| 723 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 724 | /// This intrinsic corresponds to the <c> PSLLW </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 725 | /// |
| 726 | /// \param __m |
| 727 | /// A 64-bit integer vector of [4 x i16]. |
| 728 | /// \param __count |
| 729 | /// A 64-bit integer vector interpreted as a single 64-bit integer. |
| 730 | /// \returns A 64-bit integer vector of [4 x i16] containing the left-shifted |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 731 | /// values. If \a __count is greater or equal to 16, the result is set to all |
| 732 | /// 0. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 733 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 734 | _mm_sll_pi16(__m64 __m, __m64 __count) |
| 735 | { |
| 736 | return (__m64)__builtin_ia32_psllw((__v4hi)__m, __count); |
| 737 | } |
| 738 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 739 | /// Left-shifts each 16-bit signed integer element of a 64-bit integer |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 740 | /// vector of [4 x i16] by the number of bits specified by a 32-bit integer. |
| 741 | /// The lower 16 bits of the results are packed into a 64-bit integer vector |
| 742 | /// of [4 x i16]. |
| 743 | /// |
| 744 | /// \headerfile <x86intrin.h> |
| 745 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 746 | /// This intrinsic corresponds to the <c> PSLLW </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 747 | /// |
| 748 | /// \param __m |
| 749 | /// A 64-bit integer vector of [4 x i16]. |
| 750 | /// \param __count |
| 751 | /// A 32-bit integer value. |
| 752 | /// \returns A 64-bit integer vector of [4 x i16] containing the left-shifted |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 753 | /// values. If \a __count is greater or equal to 16, the result is set to all |
| 754 | /// 0. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 755 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 756 | _mm_slli_pi16(__m64 __m, int __count) |
| 757 | { |
| 758 | return (__m64)__builtin_ia32_psllwi((__v4hi)__m, __count); |
| 759 | } |
| 760 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 761 | /// Left-shifts each 32-bit signed integer element of the first |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 762 | /// parameter, which is a 64-bit integer vector of [2 x i32], by the number |
| 763 | /// of bits specified by the second parameter, which is a 64-bit integer. The |
| 764 | /// lower 32 bits of the results are packed into a 64-bit integer vector of |
| 765 | /// [2 x i32]. |
| 766 | /// |
| 767 | /// \headerfile <x86intrin.h> |
| 768 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 769 | /// This intrinsic corresponds to the <c> PSLLD </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 770 | /// |
| 771 | /// \param __m |
| 772 | /// A 64-bit integer vector of [2 x i32]. |
| 773 | /// \param __count |
| 774 | /// A 64-bit integer vector interpreted as a single 64-bit integer. |
| 775 | /// \returns A 64-bit integer vector of [2 x i32] containing the left-shifted |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 776 | /// values. If \a __count is greater or equal to 32, the result is set to all |
| 777 | /// 0. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 778 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 779 | _mm_sll_pi32(__m64 __m, __m64 __count) |
| 780 | { |
| 781 | return (__m64)__builtin_ia32_pslld((__v2si)__m, __count); |
| 782 | } |
| 783 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 784 | /// Left-shifts each 32-bit signed integer element of a 64-bit integer |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 785 | /// vector of [2 x i32] by the number of bits specified by a 32-bit integer. |
| 786 | /// The lower 32 bits of the results are packed into a 64-bit integer vector |
| 787 | /// of [2 x i32]. |
| 788 | /// |
| 789 | /// \headerfile <x86intrin.h> |
| 790 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 791 | /// This intrinsic corresponds to the <c> PSLLD </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 792 | /// |
| 793 | /// \param __m |
| 794 | /// A 64-bit integer vector of [2 x i32]. |
| 795 | /// \param __count |
| 796 | /// A 32-bit integer value. |
| 797 | /// \returns A 64-bit integer vector of [2 x i32] containing the left-shifted |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 798 | /// values. If \a __count is greater or equal to 32, the result is set to all |
| 799 | /// 0. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 800 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 801 | _mm_slli_pi32(__m64 __m, int __count) |
| 802 | { |
| 803 | return (__m64)__builtin_ia32_pslldi((__v2si)__m, __count); |
| 804 | } |
| 805 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 806 | /// Left-shifts the first 64-bit integer parameter by the number of bits |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 807 | /// specified by the second 64-bit integer parameter. The lower 64 bits of |
| 808 | /// result are returned. |
| 809 | /// |
| 810 | /// \headerfile <x86intrin.h> |
| 811 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 812 | /// This intrinsic corresponds to the <c> PSLLQ </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 813 | /// |
| 814 | /// \param __m |
| 815 | /// A 64-bit integer vector interpreted as a single 64-bit integer. |
| 816 | /// \param __count |
| 817 | /// A 64-bit integer vector interpreted as a single 64-bit integer. |
| 818 | /// \returns A 64-bit integer vector containing the left-shifted value. If |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 819 | /// \a __count is greater or equal to 64, the result is set to 0. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 820 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 821 | _mm_sll_si64(__m64 __m, __m64 __count) |
| 822 | { |
| 823 | return (__m64)__builtin_ia32_psllq((__v1di)__m, __count); |
| 824 | } |
| 825 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 826 | /// Left-shifts the first parameter, which is a 64-bit integer, by the |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 827 | /// number of bits specified by the second parameter, which is a 32-bit |
| 828 | /// integer. The lower 64 bits of result are returned. |
| 829 | /// |
| 830 | /// \headerfile <x86intrin.h> |
| 831 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 832 | /// This intrinsic corresponds to the <c> PSLLQ </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 833 | /// |
| 834 | /// \param __m |
| 835 | /// A 64-bit integer vector interpreted as a single 64-bit integer. |
| 836 | /// \param __count |
| 837 | /// A 32-bit integer value. |
| 838 | /// \returns A 64-bit integer vector containing the left-shifted value. If |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 839 | /// \a __count is greater or equal to 64, the result is set to 0. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 840 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 841 | _mm_slli_si64(__m64 __m, int __count) |
| 842 | { |
| 843 | return (__m64)__builtin_ia32_psllqi((__v1di)__m, __count); |
| 844 | } |
| 845 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 846 | /// Right-shifts each 16-bit integer element of the first parameter, |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 847 | /// which is a 64-bit integer vector of [4 x i16], by the number of bits |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 848 | /// specified by the second parameter, which is a 64-bit integer. |
| 849 | /// |
| 850 | /// High-order bits are filled with the sign bit of the initial value of each |
| 851 | /// 16-bit element. The 16-bit results are packed into a 64-bit integer |
| 852 | /// vector of [4 x i16]. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 853 | /// |
| 854 | /// \headerfile <x86intrin.h> |
| 855 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 856 | /// This intrinsic corresponds to the <c> PSRAW </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 857 | /// |
| 858 | /// \param __m |
| 859 | /// A 64-bit integer vector of [4 x i16]. |
| 860 | /// \param __count |
| 861 | /// A 64-bit integer vector interpreted as a single 64-bit integer. |
| 862 | /// \returns A 64-bit integer vector of [4 x i16] containing the right-shifted |
| 863 | /// values. |
| 864 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 865 | _mm_sra_pi16(__m64 __m, __m64 __count) |
| 866 | { |
| 867 | return (__m64)__builtin_ia32_psraw((__v4hi)__m, __count); |
| 868 | } |
| 869 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 870 | /// Right-shifts each 16-bit integer element of a 64-bit integer vector |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 871 | /// of [4 x i16] by the number of bits specified by a 32-bit integer. |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 872 | /// |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 873 | /// High-order bits are filled with the sign bit of the initial value of each |
| 874 | /// 16-bit element. The 16-bit results are packed into a 64-bit integer |
| 875 | /// vector of [4 x i16]. |
| 876 | /// |
| 877 | /// \headerfile <x86intrin.h> |
| 878 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 879 | /// This intrinsic corresponds to the <c> PSRAW </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 880 | /// |
| 881 | /// \param __m |
| 882 | /// A 64-bit integer vector of [4 x i16]. |
| 883 | /// \param __count |
| 884 | /// A 32-bit integer value. |
| 885 | /// \returns A 64-bit integer vector of [4 x i16] containing the right-shifted |
| 886 | /// values. |
| 887 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 888 | _mm_srai_pi16(__m64 __m, int __count) |
| 889 | { |
| 890 | return (__m64)__builtin_ia32_psrawi((__v4hi)__m, __count); |
| 891 | } |
| 892 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 893 | /// Right-shifts each 32-bit integer element of the first parameter, |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 894 | /// which is a 64-bit integer vector of [2 x i32], by the number of bits |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 895 | /// specified by the second parameter, which is a 64-bit integer. |
| 896 | /// |
| 897 | /// High-order bits are filled with the sign bit of the initial value of each |
| 898 | /// 32-bit element. The 32-bit results are packed into a 64-bit integer |
| 899 | /// vector of [2 x i32]. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 900 | /// |
| 901 | /// \headerfile <x86intrin.h> |
| 902 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 903 | /// This intrinsic corresponds to the <c> PSRAD </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 904 | /// |
| 905 | /// \param __m |
| 906 | /// A 64-bit integer vector of [2 x i32]. |
| 907 | /// \param __count |
| 908 | /// A 64-bit integer vector interpreted as a single 64-bit integer. |
| 909 | /// \returns A 64-bit integer vector of [2 x i32] containing the right-shifted |
| 910 | /// values. |
| 911 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 912 | _mm_sra_pi32(__m64 __m, __m64 __count) |
| 913 | { |
| 914 | return (__m64)__builtin_ia32_psrad((__v2si)__m, __count); |
| 915 | } |
| 916 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 917 | /// Right-shifts each 32-bit integer element of a 64-bit integer vector |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 918 | /// of [2 x i32] by the number of bits specified by a 32-bit integer. |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 919 | /// |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 920 | /// High-order bits are filled with the sign bit of the initial value of each |
| 921 | /// 32-bit element. The 32-bit results are packed into a 64-bit integer |
| 922 | /// vector of [2 x i32]. |
| 923 | /// |
| 924 | /// \headerfile <x86intrin.h> |
| 925 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 926 | /// This intrinsic corresponds to the <c> PSRAD </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 927 | /// |
| 928 | /// \param __m |
| 929 | /// A 64-bit integer vector of [2 x i32]. |
| 930 | /// \param __count |
| 931 | /// A 32-bit integer value. |
| 932 | /// \returns A 64-bit integer vector of [2 x i32] containing the right-shifted |
| 933 | /// values. |
| 934 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 935 | _mm_srai_pi32(__m64 __m, int __count) |
| 936 | { |
| 937 | return (__m64)__builtin_ia32_psradi((__v2si)__m, __count); |
| 938 | } |
| 939 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 940 | /// Right-shifts each 16-bit integer element of the first parameter, |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 941 | /// which is a 64-bit integer vector of [4 x i16], by the number of bits |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 942 | /// specified by the second parameter, which is a 64-bit integer. |
| 943 | /// |
| 944 | /// High-order bits are cleared. The 16-bit results are packed into a 64-bit |
| 945 | /// integer vector of [4 x i16]. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 946 | /// |
| 947 | /// \headerfile <x86intrin.h> |
| 948 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 949 | /// This intrinsic corresponds to the <c> PSRLW </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 950 | /// |
| 951 | /// \param __m |
| 952 | /// A 64-bit integer vector of [4 x i16]. |
| 953 | /// \param __count |
| 954 | /// A 64-bit integer vector interpreted as a single 64-bit integer. |
| 955 | /// \returns A 64-bit integer vector of [4 x i16] containing the right-shifted |
| 956 | /// values. |
| 957 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 958 | _mm_srl_pi16(__m64 __m, __m64 __count) |
| 959 | { |
| 960 | return (__m64)__builtin_ia32_psrlw((__v4hi)__m, __count); |
| 961 | } |
| 962 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 963 | /// Right-shifts each 16-bit integer element of a 64-bit integer vector |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 964 | /// of [4 x i16] by the number of bits specified by a 32-bit integer. |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 965 | /// |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 966 | /// High-order bits are cleared. The 16-bit results are packed into a 64-bit |
| 967 | /// integer vector of [4 x i16]. |
| 968 | /// |
| 969 | /// \headerfile <x86intrin.h> |
| 970 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 971 | /// This intrinsic corresponds to the <c> PSRLW </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 972 | /// |
| 973 | /// \param __m |
| 974 | /// A 64-bit integer vector of [4 x i16]. |
| 975 | /// \param __count |
| 976 | /// A 32-bit integer value. |
| 977 | /// \returns A 64-bit integer vector of [4 x i16] containing the right-shifted |
| 978 | /// values. |
| 979 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 980 | _mm_srli_pi16(__m64 __m, int __count) |
| 981 | { |
| 982 | return (__m64)__builtin_ia32_psrlwi((__v4hi)__m, __count); |
| 983 | } |
| 984 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 985 | /// Right-shifts each 32-bit integer element of the first parameter, |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 986 | /// which is a 64-bit integer vector of [2 x i32], by the number of bits |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 987 | /// specified by the second parameter, which is a 64-bit integer. |
| 988 | /// |
| 989 | /// High-order bits are cleared. The 32-bit results are packed into a 64-bit |
| 990 | /// integer vector of [2 x i32]. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 991 | /// |
| 992 | /// \headerfile <x86intrin.h> |
| 993 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 994 | /// This intrinsic corresponds to the <c> PSRLD </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 995 | /// |
| 996 | /// \param __m |
| 997 | /// A 64-bit integer vector of [2 x i32]. |
| 998 | /// \param __count |
| 999 | /// A 64-bit integer vector interpreted as a single 64-bit integer. |
| 1000 | /// \returns A 64-bit integer vector of [2 x i32] containing the right-shifted |
| 1001 | /// values. |
| 1002 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 1003 | _mm_srl_pi32(__m64 __m, __m64 __count) |
| 1004 | { |
| 1005 | return (__m64)__builtin_ia32_psrld((__v2si)__m, __count); |
| 1006 | } |
| 1007 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1008 | /// Right-shifts each 32-bit integer element of a 64-bit integer vector |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1009 | /// of [2 x i32] by the number of bits specified by a 32-bit integer. |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1010 | /// |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1011 | /// High-order bits are cleared. The 32-bit results are packed into a 64-bit |
| 1012 | /// integer vector of [2 x i32]. |
| 1013 | /// |
| 1014 | /// \headerfile <x86intrin.h> |
| 1015 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1016 | /// This intrinsic corresponds to the <c> PSRLD </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1017 | /// |
| 1018 | /// \param __m |
| 1019 | /// A 64-bit integer vector of [2 x i32]. |
| 1020 | /// \param __count |
| 1021 | /// A 32-bit integer value. |
| 1022 | /// \returns A 64-bit integer vector of [2 x i32] containing the right-shifted |
| 1023 | /// values. |
| 1024 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 1025 | _mm_srli_pi32(__m64 __m, int __count) |
| 1026 | { |
| 1027 | return (__m64)__builtin_ia32_psrldi((__v2si)__m, __count); |
| 1028 | } |
| 1029 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1030 | /// Right-shifts the first 64-bit integer parameter by the number of bits |
| 1031 | /// specified by the second 64-bit integer parameter. |
| 1032 | /// |
| 1033 | /// High-order bits are cleared. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1034 | /// |
| 1035 | /// \headerfile <x86intrin.h> |
| 1036 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1037 | /// This intrinsic corresponds to the <c> PSRLQ </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1038 | /// |
| 1039 | /// \param __m |
| 1040 | /// A 64-bit integer vector interpreted as a single 64-bit integer. |
| 1041 | /// \param __count |
| 1042 | /// A 64-bit integer vector interpreted as a single 64-bit integer. |
| 1043 | /// \returns A 64-bit integer vector containing the right-shifted value. |
| 1044 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 1045 | _mm_srl_si64(__m64 __m, __m64 __count) |
| 1046 | { |
| 1047 | return (__m64)__builtin_ia32_psrlq((__v1di)__m, __count); |
| 1048 | } |
| 1049 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1050 | /// Right-shifts the first parameter, which is a 64-bit integer, by the |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1051 | /// number of bits specified by the second parameter, which is a 32-bit |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1052 | /// integer. |
| 1053 | /// |
| 1054 | /// High-order bits are cleared. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1055 | /// |
| 1056 | /// \headerfile <x86intrin.h> |
| 1057 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1058 | /// This intrinsic corresponds to the <c> PSRLQ </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1059 | /// |
| 1060 | /// \param __m |
| 1061 | /// A 64-bit integer vector interpreted as a single 64-bit integer. |
| 1062 | /// \param __count |
| 1063 | /// A 32-bit integer value. |
| 1064 | /// \returns A 64-bit integer vector containing the right-shifted value. |
| 1065 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 1066 | _mm_srli_si64(__m64 __m, int __count) |
| 1067 | { |
| 1068 | return (__m64)__builtin_ia32_psrlqi((__v1di)__m, __count); |
| 1069 | } |
| 1070 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1071 | /// Performs a bitwise AND of two 64-bit integer vectors. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1072 | /// |
| 1073 | /// \headerfile <x86intrin.h> |
| 1074 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1075 | /// This intrinsic corresponds to the <c> PAND </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1076 | /// |
| 1077 | /// \param __m1 |
| 1078 | /// A 64-bit integer vector. |
| 1079 | /// \param __m2 |
| 1080 | /// A 64-bit integer vector. |
| 1081 | /// \returns A 64-bit integer vector containing the bitwise AND of both |
| 1082 | /// parameters. |
| 1083 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 1084 | _mm_and_si64(__m64 __m1, __m64 __m2) |
| 1085 | { |
| 1086 | return __builtin_ia32_pand((__v1di)__m1, (__v1di)__m2); |
| 1087 | } |
| 1088 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1089 | /// Performs a bitwise NOT of the first 64-bit integer vector, and then |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1090 | /// performs a bitwise AND of the intermediate result and the second 64-bit |
| 1091 | /// integer vector. |
| 1092 | /// |
| 1093 | /// \headerfile <x86intrin.h> |
| 1094 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1095 | /// This intrinsic corresponds to the <c> PANDN </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1096 | /// |
| 1097 | /// \param __m1 |
| 1098 | /// A 64-bit integer vector. The one's complement of this parameter is used |
| 1099 | /// in the bitwise AND. |
| 1100 | /// \param __m2 |
| 1101 | /// A 64-bit integer vector. |
| 1102 | /// \returns A 64-bit integer vector containing the bitwise AND of the second |
| 1103 | /// parameter and the one's complement of the first parameter. |
| 1104 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 1105 | _mm_andnot_si64(__m64 __m1, __m64 __m2) |
| 1106 | { |
| 1107 | return __builtin_ia32_pandn((__v1di)__m1, (__v1di)__m2); |
| 1108 | } |
| 1109 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1110 | /// Performs a bitwise OR of two 64-bit integer vectors. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1111 | /// |
| 1112 | /// \headerfile <x86intrin.h> |
| 1113 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1114 | /// This intrinsic corresponds to the <c> POR </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1115 | /// |
| 1116 | /// \param __m1 |
| 1117 | /// A 64-bit integer vector. |
| 1118 | /// \param __m2 |
| 1119 | /// A 64-bit integer vector. |
| 1120 | /// \returns A 64-bit integer vector containing the bitwise OR of both |
| 1121 | /// parameters. |
| 1122 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 1123 | _mm_or_si64(__m64 __m1, __m64 __m2) |
| 1124 | { |
| 1125 | return __builtin_ia32_por((__v1di)__m1, (__v1di)__m2); |
| 1126 | } |
| 1127 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1128 | /// Performs a bitwise exclusive OR of two 64-bit integer vectors. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1129 | /// |
| 1130 | /// \headerfile <x86intrin.h> |
| 1131 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1132 | /// This intrinsic corresponds to the <c> PXOR </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1133 | /// |
| 1134 | /// \param __m1 |
| 1135 | /// A 64-bit integer vector. |
| 1136 | /// \param __m2 |
| 1137 | /// A 64-bit integer vector. |
| 1138 | /// \returns A 64-bit integer vector containing the bitwise exclusive OR of both |
| 1139 | /// parameters. |
| 1140 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 1141 | _mm_xor_si64(__m64 __m1, __m64 __m2) |
| 1142 | { |
| 1143 | return __builtin_ia32_pxor((__v1di)__m1, (__v1di)__m2); |
| 1144 | } |
| 1145 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1146 | /// Compares the 8-bit integer elements of two 64-bit integer vectors of |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1147 | /// [8 x i8] to determine if the element of the first vector is equal to the |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1148 | /// corresponding element of the second vector. |
| 1149 | /// |
| 1150 | /// The comparison yields 0 for false, 0xFF for true. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1151 | /// |
| 1152 | /// \headerfile <x86intrin.h> |
| 1153 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1154 | /// This intrinsic corresponds to the <c> PCMPEQB </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1155 | /// |
| 1156 | /// \param __m1 |
| 1157 | /// A 64-bit integer vector of [8 x i8]. |
| 1158 | /// \param __m2 |
| 1159 | /// A 64-bit integer vector of [8 x i8]. |
| 1160 | /// \returns A 64-bit integer vector of [8 x i8] containing the comparison |
| 1161 | /// results. |
| 1162 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 1163 | _mm_cmpeq_pi8(__m64 __m1, __m64 __m2) |
| 1164 | { |
| 1165 | return (__m64)__builtin_ia32_pcmpeqb((__v8qi)__m1, (__v8qi)__m2); |
| 1166 | } |
| 1167 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1168 | /// Compares the 16-bit integer elements of two 64-bit integer vectors of |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1169 | /// [4 x i16] to determine if the element of the first vector is equal to the |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1170 | /// corresponding element of the second vector. |
| 1171 | /// |
| 1172 | /// The comparison yields 0 for false, 0xFFFF for true. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1173 | /// |
| 1174 | /// \headerfile <x86intrin.h> |
| 1175 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1176 | /// This intrinsic corresponds to the <c> PCMPEQW </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1177 | /// |
| 1178 | /// \param __m1 |
| 1179 | /// A 64-bit integer vector of [4 x i16]. |
| 1180 | /// \param __m2 |
| 1181 | /// A 64-bit integer vector of [4 x i16]. |
| 1182 | /// \returns A 64-bit integer vector of [4 x i16] containing the comparison |
| 1183 | /// results. |
| 1184 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 1185 | _mm_cmpeq_pi16(__m64 __m1, __m64 __m2) |
| 1186 | { |
| 1187 | return (__m64)__builtin_ia32_pcmpeqw((__v4hi)__m1, (__v4hi)__m2); |
| 1188 | } |
| 1189 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1190 | /// Compares the 32-bit integer elements of two 64-bit integer vectors of |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1191 | /// [2 x i32] to determine if the element of the first vector is equal to the |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1192 | /// corresponding element of the second vector. |
| 1193 | /// |
| 1194 | /// The comparison yields 0 for false, 0xFFFFFFFF for true. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1195 | /// |
| 1196 | /// \headerfile <x86intrin.h> |
| 1197 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1198 | /// This intrinsic corresponds to the <c> PCMPEQD </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1199 | /// |
| 1200 | /// \param __m1 |
| 1201 | /// A 64-bit integer vector of [2 x i32]. |
| 1202 | /// \param __m2 |
| 1203 | /// A 64-bit integer vector of [2 x i32]. |
| 1204 | /// \returns A 64-bit integer vector of [2 x i32] containing the comparison |
| 1205 | /// results. |
| 1206 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 1207 | _mm_cmpeq_pi32(__m64 __m1, __m64 __m2) |
| 1208 | { |
| 1209 | return (__m64)__builtin_ia32_pcmpeqd((__v2si)__m1, (__v2si)__m2); |
| 1210 | } |
| 1211 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1212 | /// Compares the 8-bit integer elements of two 64-bit integer vectors of |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1213 | /// [8 x i8] to determine if the element of the first vector is greater than |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1214 | /// the corresponding element of the second vector. |
| 1215 | /// |
| 1216 | /// The comparison yields 0 for false, 0xFF for true. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1217 | /// |
| 1218 | /// \headerfile <x86intrin.h> |
| 1219 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1220 | /// This intrinsic corresponds to the <c> PCMPGTB </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1221 | /// |
| 1222 | /// \param __m1 |
| 1223 | /// A 64-bit integer vector of [8 x i8]. |
| 1224 | /// \param __m2 |
| 1225 | /// A 64-bit integer vector of [8 x i8]. |
| 1226 | /// \returns A 64-bit integer vector of [8 x i8] containing the comparison |
| 1227 | /// results. |
| 1228 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 1229 | _mm_cmpgt_pi8(__m64 __m1, __m64 __m2) |
| 1230 | { |
| 1231 | return (__m64)__builtin_ia32_pcmpgtb((__v8qi)__m1, (__v8qi)__m2); |
| 1232 | } |
| 1233 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1234 | /// Compares the 16-bit integer elements of two 64-bit integer vectors of |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1235 | /// [4 x i16] to determine if the element of the first vector is greater than |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1236 | /// the corresponding element of the second vector. |
| 1237 | /// |
| 1238 | /// The comparison yields 0 for false, 0xFFFF for true. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1239 | /// |
| 1240 | /// \headerfile <x86intrin.h> |
| 1241 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1242 | /// This intrinsic corresponds to the <c> PCMPGTW </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1243 | /// |
| 1244 | /// \param __m1 |
| 1245 | /// A 64-bit integer vector of [4 x i16]. |
| 1246 | /// \param __m2 |
| 1247 | /// A 64-bit integer vector of [4 x i16]. |
| 1248 | /// \returns A 64-bit integer vector of [4 x i16] containing the comparison |
| 1249 | /// results. |
| 1250 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 1251 | _mm_cmpgt_pi16(__m64 __m1, __m64 __m2) |
| 1252 | { |
| 1253 | return (__m64)__builtin_ia32_pcmpgtw((__v4hi)__m1, (__v4hi)__m2); |
| 1254 | } |
| 1255 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1256 | /// Compares the 32-bit integer elements of two 64-bit integer vectors of |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1257 | /// [2 x i32] to determine if the element of the first vector is greater than |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1258 | /// the corresponding element of the second vector. |
| 1259 | /// |
| 1260 | /// The comparison yields 0 for false, 0xFFFFFFFF for true. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1261 | /// |
| 1262 | /// \headerfile <x86intrin.h> |
| 1263 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1264 | /// This intrinsic corresponds to the <c> PCMPGTD </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1265 | /// |
| 1266 | /// \param __m1 |
| 1267 | /// A 64-bit integer vector of [2 x i32]. |
| 1268 | /// \param __m2 |
| 1269 | /// A 64-bit integer vector of [2 x i32]. |
| 1270 | /// \returns A 64-bit integer vector of [2 x i32] containing the comparison |
| 1271 | /// results. |
| 1272 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 1273 | _mm_cmpgt_pi32(__m64 __m1, __m64 __m2) |
| 1274 | { |
| 1275 | return (__m64)__builtin_ia32_pcmpgtd((__v2si)__m1, (__v2si)__m2); |
| 1276 | } |
| 1277 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1278 | /// Constructs a 64-bit integer vector initialized to zero. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1279 | /// |
| 1280 | /// \headerfile <x86intrin.h> |
| 1281 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1282 | /// This intrinsic corresponds to the <c> PXOR </c> instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1283 | /// |
| 1284 | /// \returns An initialized 64-bit integer vector with all elements set to zero. |
| 1285 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 1286 | _mm_setzero_si64(void) |
| 1287 | { |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1288 | return __extension__ (__m64){ 0LL }; |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1289 | } |
| 1290 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1291 | /// Constructs a 64-bit integer vector initialized with the specified |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1292 | /// 32-bit integer values. |
| 1293 | /// |
| 1294 | /// \headerfile <x86intrin.h> |
| 1295 | /// |
| 1296 | /// This intrinsic is a utility function and does not correspond to a specific |
| 1297 | /// instruction. |
| 1298 | /// |
| 1299 | /// \param __i1 |
| 1300 | /// A 32-bit integer value used to initialize the upper 32 bits of the |
| 1301 | /// result. |
| 1302 | /// \param __i0 |
| 1303 | /// A 32-bit integer value used to initialize the lower 32 bits of the |
| 1304 | /// result. |
| 1305 | /// \returns An initialized 64-bit integer vector. |
| 1306 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 1307 | _mm_set_pi32(int __i1, int __i0) |
| 1308 | { |
| 1309 | return (__m64)__builtin_ia32_vec_init_v2si(__i0, __i1); |
| 1310 | } |
| 1311 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1312 | /// Constructs a 64-bit integer vector initialized with the specified |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1313 | /// 16-bit integer values. |
| 1314 | /// |
| 1315 | /// \headerfile <x86intrin.h> |
| 1316 | /// |
| 1317 | /// This intrinsic is a utility function and does not correspond to a specific |
| 1318 | /// instruction. |
| 1319 | /// |
| 1320 | /// \param __s3 |
| 1321 | /// A 16-bit integer value used to initialize bits [63:48] of the result. |
| 1322 | /// \param __s2 |
| 1323 | /// A 16-bit integer value used to initialize bits [47:32] of the result. |
| 1324 | /// \param __s1 |
| 1325 | /// A 16-bit integer value used to initialize bits [31:16] of the result. |
| 1326 | /// \param __s0 |
| 1327 | /// A 16-bit integer value used to initialize bits [15:0] of the result. |
| 1328 | /// \returns An initialized 64-bit integer vector. |
| 1329 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 1330 | _mm_set_pi16(short __s3, short __s2, short __s1, short __s0) |
| 1331 | { |
| 1332 | return (__m64)__builtin_ia32_vec_init_v4hi(__s0, __s1, __s2, __s3); |
| 1333 | } |
| 1334 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1335 | /// Constructs a 64-bit integer vector initialized with the specified |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1336 | /// 8-bit integer values. |
| 1337 | /// |
| 1338 | /// \headerfile <x86intrin.h> |
| 1339 | /// |
| 1340 | /// This intrinsic is a utility function and does not correspond to a specific |
| 1341 | /// instruction. |
| 1342 | /// |
| 1343 | /// \param __b7 |
| 1344 | /// An 8-bit integer value used to initialize bits [63:56] of the result. |
| 1345 | /// \param __b6 |
| 1346 | /// An 8-bit integer value used to initialize bits [55:48] of the result. |
| 1347 | /// \param __b5 |
| 1348 | /// An 8-bit integer value used to initialize bits [47:40] of the result. |
| 1349 | /// \param __b4 |
| 1350 | /// An 8-bit integer value used to initialize bits [39:32] of the result. |
| 1351 | /// \param __b3 |
| 1352 | /// An 8-bit integer value used to initialize bits [31:24] of the result. |
| 1353 | /// \param __b2 |
| 1354 | /// An 8-bit integer value used to initialize bits [23:16] of the result. |
| 1355 | /// \param __b1 |
| 1356 | /// An 8-bit integer value used to initialize bits [15:8] of the result. |
| 1357 | /// \param __b0 |
| 1358 | /// An 8-bit integer value used to initialize bits [7:0] of the result. |
| 1359 | /// \returns An initialized 64-bit integer vector. |
| 1360 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 1361 | _mm_set_pi8(char __b7, char __b6, char __b5, char __b4, char __b3, char __b2, |
| 1362 | char __b1, char __b0) |
| 1363 | { |
| 1364 | return (__m64)__builtin_ia32_vec_init_v8qi(__b0, __b1, __b2, __b3, |
| 1365 | __b4, __b5, __b6, __b7); |
| 1366 | } |
| 1367 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1368 | /// Constructs a 64-bit integer vector of [2 x i32], with each of the |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1369 | /// 32-bit integer vector elements set to the specified 32-bit integer |
| 1370 | /// value. |
| 1371 | /// |
| 1372 | /// \headerfile <x86intrin.h> |
| 1373 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1374 | /// This intrinsic is a utility function and does not correspond to a specific |
| 1375 | /// instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1376 | /// |
| 1377 | /// \param __i |
| 1378 | /// A 32-bit integer value used to initialize each vector element of the |
| 1379 | /// result. |
| 1380 | /// \returns An initialized 64-bit integer vector of [2 x i32]. |
| 1381 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 1382 | _mm_set1_pi32(int __i) |
| 1383 | { |
| 1384 | return _mm_set_pi32(__i, __i); |
| 1385 | } |
| 1386 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1387 | /// Constructs a 64-bit integer vector of [4 x i16], with each of the |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1388 | /// 16-bit integer vector elements set to the specified 16-bit integer |
| 1389 | /// value. |
| 1390 | /// |
| 1391 | /// \headerfile <x86intrin.h> |
| 1392 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1393 | /// This intrinsic is a utility function and does not correspond to a specific |
| 1394 | /// instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1395 | /// |
| 1396 | /// \param __w |
| 1397 | /// A 16-bit integer value used to initialize each vector element of the |
| 1398 | /// result. |
| 1399 | /// \returns An initialized 64-bit integer vector of [4 x i16]. |
| 1400 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 1401 | _mm_set1_pi16(short __w) |
| 1402 | { |
| 1403 | return _mm_set_pi16(__w, __w, __w, __w); |
| 1404 | } |
| 1405 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1406 | /// Constructs a 64-bit integer vector of [8 x i8], with each of the |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1407 | /// 8-bit integer vector elements set to the specified 8-bit integer value. |
| 1408 | /// |
| 1409 | /// \headerfile <x86intrin.h> |
| 1410 | /// |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1411 | /// This intrinsic is a utility function and does not correspond to a specific |
| 1412 | /// instruction. |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1413 | /// |
| 1414 | /// \param __b |
| 1415 | /// An 8-bit integer value used to initialize each vector element of the |
| 1416 | /// result. |
| 1417 | /// \returns An initialized 64-bit integer vector of [8 x i8]. |
| 1418 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 1419 | _mm_set1_pi8(char __b) |
| 1420 | { |
| 1421 | return _mm_set_pi8(__b, __b, __b, __b, __b, __b, __b, __b); |
| 1422 | } |
| 1423 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1424 | /// Constructs a 64-bit integer vector, initialized in reverse order with |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1425 | /// the specified 32-bit integer values. |
| 1426 | /// |
| 1427 | /// \headerfile <x86intrin.h> |
| 1428 | /// |
| 1429 | /// This intrinsic is a utility function and does not correspond to a specific |
| 1430 | /// instruction. |
| 1431 | /// |
| 1432 | /// \param __i0 |
| 1433 | /// A 32-bit integer value used to initialize the lower 32 bits of the |
| 1434 | /// result. |
| 1435 | /// \param __i1 |
| 1436 | /// A 32-bit integer value used to initialize the upper 32 bits of the |
| 1437 | /// result. |
| 1438 | /// \returns An initialized 64-bit integer vector. |
| 1439 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 1440 | _mm_setr_pi32(int __i0, int __i1) |
| 1441 | { |
| 1442 | return _mm_set_pi32(__i1, __i0); |
| 1443 | } |
| 1444 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1445 | /// Constructs a 64-bit integer vector, initialized in reverse order with |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1446 | /// the specified 16-bit integer values. |
| 1447 | /// |
| 1448 | /// \headerfile <x86intrin.h> |
| 1449 | /// |
| 1450 | /// This intrinsic is a utility function and does not correspond to a specific |
| 1451 | /// instruction. |
| 1452 | /// |
| 1453 | /// \param __w0 |
| 1454 | /// A 16-bit integer value used to initialize bits [15:0] of the result. |
| 1455 | /// \param __w1 |
| 1456 | /// A 16-bit integer value used to initialize bits [31:16] of the result. |
| 1457 | /// \param __w2 |
| 1458 | /// A 16-bit integer value used to initialize bits [47:32] of the result. |
| 1459 | /// \param __w3 |
| 1460 | /// A 16-bit integer value used to initialize bits [63:48] of the result. |
| 1461 | /// \returns An initialized 64-bit integer vector. |
| 1462 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 1463 | _mm_setr_pi16(short __w0, short __w1, short __w2, short __w3) |
| 1464 | { |
| 1465 | return _mm_set_pi16(__w3, __w2, __w1, __w0); |
| 1466 | } |
| 1467 | |
Logan Chien | 55afb0a | 2018-10-15 10:42:14 +0800 | [diff] [blame] | 1468 | /// Constructs a 64-bit integer vector, initialized in reverse order with |
Logan Chien | 2833ffb | 2018-10-09 10:03:24 +0800 | [diff] [blame] | 1469 | /// the specified 8-bit integer values. |
| 1470 | /// |
| 1471 | /// \headerfile <x86intrin.h> |
| 1472 | /// |
| 1473 | /// This intrinsic is a utility function and does not correspond to a specific |
| 1474 | /// instruction. |
| 1475 | /// |
| 1476 | /// \param __b0 |
| 1477 | /// An 8-bit integer value used to initialize bits [7:0] of the result. |
| 1478 | /// \param __b1 |
| 1479 | /// An 8-bit integer value used to initialize bits [15:8] of the result. |
| 1480 | /// \param __b2 |
| 1481 | /// An 8-bit integer value used to initialize bits [23:16] of the result. |
| 1482 | /// \param __b3 |
| 1483 | /// An 8-bit integer value used to initialize bits [31:24] of the result. |
| 1484 | /// \param __b4 |
| 1485 | /// An 8-bit integer value used to initialize bits [39:32] of the result. |
| 1486 | /// \param __b5 |
| 1487 | /// An 8-bit integer value used to initialize bits [47:40] of the result. |
| 1488 | /// \param __b6 |
| 1489 | /// An 8-bit integer value used to initialize bits [55:48] of the result. |
| 1490 | /// \param __b7 |
| 1491 | /// An 8-bit integer value used to initialize bits [63:56] of the result. |
| 1492 | /// \returns An initialized 64-bit integer vector. |
| 1493 | static __inline__ __m64 __DEFAULT_FN_ATTRS |
| 1494 | _mm_setr_pi8(char __b0, char __b1, char __b2, char __b3, char __b4, char __b5, |
| 1495 | char __b6, char __b7) |
| 1496 | { |
| 1497 | return _mm_set_pi8(__b7, __b6, __b5, __b4, __b3, __b2, __b1, __b0); |
| 1498 | } |
| 1499 | |
| 1500 | #undef __DEFAULT_FN_ATTRS |
| 1501 | |
| 1502 | /* Aliases for compatibility. */ |
| 1503 | #define _m_empty _mm_empty |
| 1504 | #define _m_from_int _mm_cvtsi32_si64 |
| 1505 | #define _m_from_int64 _mm_cvtsi64_m64 |
| 1506 | #define _m_to_int _mm_cvtsi64_si32 |
| 1507 | #define _m_to_int64 _mm_cvtm64_si64 |
| 1508 | #define _m_packsswb _mm_packs_pi16 |
| 1509 | #define _m_packssdw _mm_packs_pi32 |
| 1510 | #define _m_packuswb _mm_packs_pu16 |
| 1511 | #define _m_punpckhbw _mm_unpackhi_pi8 |
| 1512 | #define _m_punpckhwd _mm_unpackhi_pi16 |
| 1513 | #define _m_punpckhdq _mm_unpackhi_pi32 |
| 1514 | #define _m_punpcklbw _mm_unpacklo_pi8 |
| 1515 | #define _m_punpcklwd _mm_unpacklo_pi16 |
| 1516 | #define _m_punpckldq _mm_unpacklo_pi32 |
| 1517 | #define _m_paddb _mm_add_pi8 |
| 1518 | #define _m_paddw _mm_add_pi16 |
| 1519 | #define _m_paddd _mm_add_pi32 |
| 1520 | #define _m_paddsb _mm_adds_pi8 |
| 1521 | #define _m_paddsw _mm_adds_pi16 |
| 1522 | #define _m_paddusb _mm_adds_pu8 |
| 1523 | #define _m_paddusw _mm_adds_pu16 |
| 1524 | #define _m_psubb _mm_sub_pi8 |
| 1525 | #define _m_psubw _mm_sub_pi16 |
| 1526 | #define _m_psubd _mm_sub_pi32 |
| 1527 | #define _m_psubsb _mm_subs_pi8 |
| 1528 | #define _m_psubsw _mm_subs_pi16 |
| 1529 | #define _m_psubusb _mm_subs_pu8 |
| 1530 | #define _m_psubusw _mm_subs_pu16 |
| 1531 | #define _m_pmaddwd _mm_madd_pi16 |
| 1532 | #define _m_pmulhw _mm_mulhi_pi16 |
| 1533 | #define _m_pmullw _mm_mullo_pi16 |
| 1534 | #define _m_psllw _mm_sll_pi16 |
| 1535 | #define _m_psllwi _mm_slli_pi16 |
| 1536 | #define _m_pslld _mm_sll_pi32 |
| 1537 | #define _m_pslldi _mm_slli_pi32 |
| 1538 | #define _m_psllq _mm_sll_si64 |
| 1539 | #define _m_psllqi _mm_slli_si64 |
| 1540 | #define _m_psraw _mm_sra_pi16 |
| 1541 | #define _m_psrawi _mm_srai_pi16 |
| 1542 | #define _m_psrad _mm_sra_pi32 |
| 1543 | #define _m_psradi _mm_srai_pi32 |
| 1544 | #define _m_psrlw _mm_srl_pi16 |
| 1545 | #define _m_psrlwi _mm_srli_pi16 |
| 1546 | #define _m_psrld _mm_srl_pi32 |
| 1547 | #define _m_psrldi _mm_srli_pi32 |
| 1548 | #define _m_psrlq _mm_srl_si64 |
| 1549 | #define _m_psrlqi _mm_srli_si64 |
| 1550 | #define _m_pand _mm_and_si64 |
| 1551 | #define _m_pandn _mm_andnot_si64 |
| 1552 | #define _m_por _mm_or_si64 |
| 1553 | #define _m_pxor _mm_xor_si64 |
| 1554 | #define _m_pcmpeqb _mm_cmpeq_pi8 |
| 1555 | #define _m_pcmpeqw _mm_cmpeq_pi16 |
| 1556 | #define _m_pcmpeqd _mm_cmpeq_pi32 |
| 1557 | #define _m_pcmpgtb _mm_cmpgt_pi8 |
| 1558 | #define _m_pcmpgtw _mm_cmpgt_pi16 |
| 1559 | #define _m_pcmpgtd _mm_cmpgt_pi32 |
| 1560 | |
| 1561 | #endif /* __MMINTRIN_H */ |
| 1562 | |