1 /**************************************************************************** 2 * Copyright (C) 2017 Intel Corporation. All Rights Reserved. 3 * 4 * Permission is hereby granted, free of charge, to any person obtaining a 5 * copy of this software and associated documentation files (the "Software"), 6 * to deal in the Software without restriction, including without limitation 7 * the rights to use, copy, modify, merge, publish, distribute, sublicense, 8 * and/or sell copies of the Software, and to permit persons to whom the 9 * Software is furnished to do so, subject to the following conditions: 10 * 11 * The above copyright notice and this permission notice (including the next 12 * paragraph) shall be included in all copies or substantial portions of the 13 * Software. 14 * 15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 16 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 17 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL 18 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 19 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING 20 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS 21 * IN THE SOFTWARE. 22 ****************************************************************************/ 23 #if !defined(__SIMD_LIB_AVX512_HPP__) 24 #error Do not include this file directly, use "simdlib.hpp" instead. 25 #endif 26 27 //============================================================================ 28 // SIMD16 AVX512 (F) implementation for Knights Family Processors 29 // 30 //============================================================================ 31 32 #define SIMD_WRAPPER_1_(op, intrin) \ 33 static SIMDINLINE Float SIMDCALL op(Float a) \ 34 {\ 35 return intrin(a);\ 36 } 37 38 #define SIMD_WRAPPER_1(op) \ 39 SIMD_WRAPPER_1_(op, _mm512_##op) 40 41 #define SIMD_WRAPPER_2_(op, intrin) \ 42 static SIMDINLINE Float SIMDCALL op(Float a, Float b) \ 43 {\ 44 return _mm512_##intrin(a, b);\ 45 } 46 #define SIMD_WRAPPER_2(op) SIMD_WRAPPER_2_(op, op) 47 48 #define SIMD_WRAPPERI_2_(op, intrin) \ 49 static SIMDINLINE Float SIMDCALL op(Float a, Float b) \ 50 {\ 51 return _mm512_castsi512_ps(_mm512_##intrin(\ 52 _mm512_castps_si512(a), _mm512_castps_si512(b)));\ 53 } 54 55 #define SIMD_DWRAPPER_2(op) \ 56 static SIMDINLINE Double SIMDCALL op(Double a, Double b) \ 57 {\ 58 return _mm512_##op(a, b);\ 59 } 60 61 #define SIMD_WRAPPER_2I_(op, intrin) \ 62 template<int ImmT>\ 63 static SIMDINLINE Float SIMDCALL op(Float a, Float b) \ 64 {\ 65 return _mm512_##intrin(a, b, ImmT);\ 66 } 67 #define SIMD_WRAPPER_2I(op) SIMD_WRAPPER_2I_(op, op) 68 69 #define SIMD_DWRAPPER_2I_(op, intrin) \ 70 template<int ImmT>\ 71 static SIMDINLINE Double SIMDCALL op(Double a, Double b) \ 72 {\ 73 return _mm512_##intrin(a, b, ImmT);\ 74 } 75 #define SIMD_DWRAPPER_2I(op) SIMD_DWRAPPER_2I_(op, op) 76 77 #define SIMD_WRAPPER_3(op) \ 78 static SIMDINLINE Float SIMDCALL op(Float a, Float b, Float c) \ 79 {\ 80 return _mm512_##op(a, b, c);\ 81 } 82 83 #define SIMD_IWRAPPER_1(op) \ 84 static SIMDINLINE Integer SIMDCALL op(Integer a) \ 85 {\ 86 return _mm512_##op(a);\ 87 } 88 #define SIMD_IWRAPPER_1_8(op) \ 89 static SIMDINLINE Integer SIMDCALL op(SIMD256Impl::Integer a) \ 90 {\ 91 return _mm512_##op(a);\ 92 } 93 94 #define SIMD_IWRAPPER_1_4(op) \ 95 static SIMDINLINE Integer SIMDCALL op(SIMD128Impl::Integer a) \ 96 {\ 97 return _mm512_##op(a);\ 98 } 99 100 #define SIMD_IWRAPPER_1I_(op, intrin) \ 101 template<int ImmT> \ 102 static SIMDINLINE Integer SIMDCALL op(Integer a) \ 103 {\ 104 return intrin(a, ImmT);\ 105 } 106 #define SIMD_IWRAPPER_1I(op) SIMD_IWRAPPER_1I_(op, _mm512_##op) 107 108 #define SIMD_IWRAPPER_2_(op, intrin) \ 109 static SIMDINLINE Integer SIMDCALL op(Integer a, Integer b) \ 110 {\ 111 return _mm512_##intrin(a, b);\ 112 } 113 #define SIMD_IWRAPPER_2(op) SIMD_IWRAPPER_2_(op, op) 114 115 #define SIMD_IWRAPPER_2_CMP(op, cmp) \ 116 static SIMDINLINE Integer SIMDCALL op(Integer a, Integer b) \ 117 {\ 118 return cmp(a, b);\ 119 } 120 121 #define SIMD_IFWRAPPER_2(op, intrin) \ 122 static SIMDINLINE Integer SIMDCALL op(Integer a, Integer b) \ 123 {\ 124 return castps_si(_mm512_##intrin(castsi_ps(a), castsi_ps(b)) );\ 125 } 126 127 #define SIMD_IWRAPPER_2I_(op, intrin) \ 128 template<int ImmT>\ 129 static SIMDINLINE Integer SIMDCALL op(Integer a, Integer b) \ 130 {\ 131 return _mm512_##intrin(a, b, ImmT);\ 132 } 133 #define SIMD_IWRAPPER_2I(op) SIMD_IWRAPPER_2I_(op, op) 134 135 SIMD_WRAPPERI_2_(and_ps, and_epi32); // return a & b (float treated as int) 136 SIMD_WRAPPERI_2_(andnot_ps, andnot_epi32); // return (~a) & b (float treated as int) 137 SIMD_WRAPPERI_2_(or_ps, or_epi32); // return a | b (float treated as int) 138 SIMD_WRAPPERI_2_(xor_ps, xor_epi32); // return a ^ b (float treated as int) 139 140 #undef SIMD_WRAPPER_1_ 141 #undef SIMD_WRAPPER_1 142 #undef SIMD_WRAPPER_2 143 #undef SIMD_WRAPPER_2_ 144 #undef SIMD_WRAPPERI_2_ 145 #undef SIMD_DWRAPPER_2 146 #undef SIMD_DWRAPPER_2I 147 #undef SIMD_WRAPPER_2I_ 148 #undef SIMD_WRAPPER_3_ 149 #undef SIMD_WRAPPER_2I 150 #undef SIMD_WRAPPER_3 151 #undef SIMD_IWRAPPER_1 152 #undef SIMD_IWRAPPER_2 153 #undef SIMD_IFWRAPPER_2 154 #undef SIMD_IWRAPPER_2I 155 #undef SIMD_IWRAPPER_1 156 #undef SIMD_IWRAPPER_1I 157 #undef SIMD_IWRAPPER_1I_ 158 #undef SIMD_IWRAPPER_2 159 #undef SIMD_IWRAPPER_2_ 160 #undef SIMD_IWRAPPER_2I 161 162