Ying Wang | b335bb0 | 2011-11-29 10:23:55 -0800 | [diff] [blame] | 1 | /*===---- mm3dnow.h - 3DNow! intrinsics ------------------------------------=== |
| 2 | * |
| 3 | * Permission is hereby granted, free of charge, to any person obtaining a copy |
| 4 | * of this software and associated documentation files (the "Software"), to deal |
| 5 | * in the Software without restriction, including without limitation the rights |
| 6 | * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell |
| 7 | * copies of the Software, and to permit persons to whom the Software is |
| 8 | * furnished to do so, subject to the following conditions: |
| 9 | * |
| 10 | * The above copyright notice and this permission notice shall be included in |
| 11 | * all copies or substantial portions of the Software. |
| 12 | * |
| 13 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
| 14 | * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
| 15 | * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE |
| 16 | * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER |
| 17 | * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, |
| 18 | * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN |
| 19 | * THE SOFTWARE. |
| 20 | * |
| 21 | *===-----------------------------------------------------------------------=== |
| 22 | */ |
| 23 | |
| 24 | #ifndef _MM3DNOW_H_INCLUDED |
| 25 | #define _MM3DNOW_H_INCLUDED |
| 26 | |
| 27 | #include <mmintrin.h> |
Stephen Hines | 996e4dc | 2013-08-13 01:04:14 -0700 | [diff] [blame] | 28 | #include <prfchwintrin.h> |
Ying Wang | b335bb0 | 2011-11-29 10:23:55 -0800 | [diff] [blame] | 29 | |
| 30 | typedef float __v2sf __attribute__((__vector_size__(8))); |
| 31 | |
| 32 | static __inline__ void __attribute__((__always_inline__, __nodebug__)) |
| 33 | _m_femms() { |
| 34 | __builtin_ia32_femms(); |
| 35 | } |
| 36 | |
| 37 | static __inline__ __m64 __attribute__((__always_inline__, __nodebug__)) |
| 38 | _m_pavgusb(__m64 __m1, __m64 __m2) { |
| 39 | return (__m64)__builtin_ia32_pavgusb((__v8qi)__m1, (__v8qi)__m2); |
| 40 | } |
| 41 | |
| 42 | static __inline__ __m64 __attribute__((__always_inline__, __nodebug__)) |
| 43 | _m_pf2id(__m64 __m) { |
| 44 | return (__m64)__builtin_ia32_pf2id((__v2sf)__m); |
| 45 | } |
| 46 | |
| 47 | static __inline__ __m64 __attribute__((__always_inline__, __nodebug__)) |
| 48 | _m_pfacc(__m64 __m1, __m64 __m2) { |
| 49 | return (__m64)__builtin_ia32_pfacc((__v2sf)__m1, (__v2sf)__m2); |
| 50 | } |
| 51 | |
| 52 | static __inline__ __m64 __attribute__((__always_inline__, __nodebug__)) |
| 53 | _m_pfadd(__m64 __m1, __m64 __m2) { |
| 54 | return (__m64)__builtin_ia32_pfadd((__v2sf)__m1, (__v2sf)__m2); |
| 55 | } |
| 56 | |
| 57 | static __inline__ __m64 __attribute__((__always_inline__, __nodebug__)) |
| 58 | _m_pfcmpeq(__m64 __m1, __m64 __m2) { |
| 59 | return (__m64)__builtin_ia32_pfcmpeq((__v2sf)__m1, (__v2sf)__m2); |
| 60 | } |
| 61 | |
| 62 | static __inline__ __m64 __attribute__((__always_inline__, __nodebug__)) |
| 63 | _m_pfcmpge(__m64 __m1, __m64 __m2) { |
| 64 | return (__m64)__builtin_ia32_pfcmpge((__v2sf)__m1, (__v2sf)__m2); |
| 65 | } |
| 66 | |
| 67 | static __inline__ __m64 __attribute__((__always_inline__, __nodebug__)) |
| 68 | _m_pfcmpgt(__m64 __m1, __m64 __m2) { |
| 69 | return (__m64)__builtin_ia32_pfcmpgt((__v2sf)__m1, (__v2sf)__m2); |
| 70 | } |
| 71 | |
| 72 | static __inline__ __m64 __attribute__((__always_inline__, __nodebug__)) |
| 73 | _m_pfmax(__m64 __m1, __m64 __m2) { |
| 74 | return (__m64)__builtin_ia32_pfmax((__v2sf)__m1, (__v2sf)__m2); |
| 75 | } |
| 76 | |
| 77 | static __inline__ __m64 __attribute__((__always_inline__, __nodebug__)) |
| 78 | _m_pfmin(__m64 __m1, __m64 __m2) { |
| 79 | return (__m64)__builtin_ia32_pfmin((__v2sf)__m1, (__v2sf)__m2); |
| 80 | } |
| 81 | |
| 82 | static __inline__ __m64 __attribute__((__always_inline__, __nodebug__)) |
| 83 | _m_pfmul(__m64 __m1, __m64 __m2) { |
| 84 | return (__m64)__builtin_ia32_pfmul((__v2sf)__m1, (__v2sf)__m2); |
| 85 | } |
| 86 | |
| 87 | static __inline__ __m64 __attribute__((__always_inline__, __nodebug__)) |
| 88 | _m_pfrcp(__m64 __m) { |
| 89 | return (__m64)__builtin_ia32_pfrcp((__v2sf)__m); |
| 90 | } |
| 91 | |
| 92 | static __inline__ __m64 __attribute__((__always_inline__, __nodebug__)) |
| 93 | _m_pfrcpit1(__m64 __m1, __m64 __m2) { |
| 94 | return (__m64)__builtin_ia32_pfrcpit1((__v2sf)__m1, (__v2sf)__m2); |
| 95 | } |
| 96 | |
| 97 | static __inline__ __m64 __attribute__((__always_inline__, __nodebug__)) |
| 98 | _m_pfrcpit2(__m64 __m1, __m64 __m2) { |
| 99 | return (__m64)__builtin_ia32_pfrcpit2((__v2sf)__m1, (__v2sf)__m2); |
| 100 | } |
| 101 | |
| 102 | static __inline__ __m64 __attribute__((__always_inline__, __nodebug__)) |
| 103 | _m_pfrsqrt(__m64 __m) { |
| 104 | return (__m64)__builtin_ia32_pfrsqrt((__v2sf)__m); |
| 105 | } |
| 106 | |
| 107 | static __inline__ __m64 __attribute__((__always_inline__, __nodebug__)) |
| 108 | _m_pfrsqrtit1(__m64 __m1, __m64 __m2) { |
Ying Wang | 6099914 | 2013-01-07 13:59:36 -0800 | [diff] [blame] | 109 | return (__m64)__builtin_ia32_pfrsqit1((__v2sf)__m1, (__v2sf)__m2); |
Ying Wang | b335bb0 | 2011-11-29 10:23:55 -0800 | [diff] [blame] | 110 | } |
| 111 | |
| 112 | static __inline__ __m64 __attribute__((__always_inline__, __nodebug__)) |
| 113 | _m_pfsub(__m64 __m1, __m64 __m2) { |
| 114 | return (__m64)__builtin_ia32_pfsub((__v2sf)__m1, (__v2sf)__m2); |
| 115 | } |
| 116 | |
| 117 | static __inline__ __m64 __attribute__((__always_inline__, __nodebug__)) |
| 118 | _m_pfsubr(__m64 __m1, __m64 __m2) { |
| 119 | return (__m64)__builtin_ia32_pfsubr((__v2sf)__m1, (__v2sf)__m2); |
| 120 | } |
| 121 | |
| 122 | static __inline__ __m64 __attribute__((__always_inline__, __nodebug__)) |
| 123 | _m_pi2fd(__m64 __m) { |
| 124 | return (__m64)__builtin_ia32_pi2fd((__v2si)__m); |
| 125 | } |
| 126 | |
| 127 | static __inline__ __m64 __attribute__((__always_inline__, __nodebug__)) |
| 128 | _m_pmulhrw(__m64 __m1, __m64 __m2) { |
| 129 | return (__m64)__builtin_ia32_pmulhrw((__v4hi)__m1, (__v4hi)__m2); |
| 130 | } |
| 131 | |
| 132 | static __inline__ __m64 __attribute__((__always_inline__, __nodebug__)) |
| 133 | _m_pf2iw(__m64 __m) { |
| 134 | return (__m64)__builtin_ia32_pf2iw((__v2sf)__m); |
| 135 | } |
| 136 | |
| 137 | static __inline__ __m64 __attribute__((__always_inline__, __nodebug__)) |
| 138 | _m_pfnacc(__m64 __m1, __m64 __m2) { |
| 139 | return (__m64)__builtin_ia32_pfnacc((__v2sf)__m1, (__v2sf)__m2); |
| 140 | } |
| 141 | |
| 142 | static __inline__ __m64 __attribute__((__always_inline__, __nodebug__)) |
| 143 | _m_pfpnacc(__m64 __m1, __m64 __m2) { |
| 144 | return (__m64)__builtin_ia32_pfpnacc((__v2sf)__m1, (__v2sf)__m2); |
| 145 | } |
| 146 | |
| 147 | static __inline__ __m64 __attribute__((__always_inline__, __nodebug__)) |
| 148 | _m_pi2fw(__m64 __m) { |
| 149 | return (__m64)__builtin_ia32_pi2fw((__v2si)__m); |
| 150 | } |
| 151 | |
| 152 | static __inline__ __m64 __attribute__((__always_inline__, __nodebug__)) |
| 153 | _m_pswapdsf(__m64 __m) { |
| 154 | return (__m64)__builtin_ia32_pswapdsf((__v2sf)__m); |
| 155 | } |
| 156 | |
| 157 | static __inline__ __m64 __attribute__((__always_inline__, __nodebug__)) |
| 158 | _m_pswapdsi(__m64 __m) { |
| 159 | return (__m64)__builtin_ia32_pswapdsi((__v2si)__m); |
| 160 | } |
| 161 | |
| 162 | #endif |