| 1 |
/* Copyright (C) 2003-2015 Free Software Foundation, Inc. |
| 2 |
|
| 3 |
This file is part of GCC. |
| 4 |
|
| 5 |
GCC is free software; you can redistribute it and/or modify |
| 6 |
it under the terms of the GNU General Public License as published by |
| 7 |
the Free Software Foundation; either version 3, or (at your option) |
| 8 |
any later version. |
| 9 |
|
| 10 |
GCC is distributed in the hope that it will be useful, |
| 11 |
but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 12 |
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 13 |
GNU General Public License for more details. |
| 14 |
|
| 15 |
Under Section 7 of GPL version 3, you are granted additional |
| 16 |
permissions described in the GCC Runtime Library Exception, version |
| 17 |
3.1, as published by the Free Software Foundation. |
| 18 |
|
| 19 |
You should have received a copy of the GNU General Public License and |
| 20 |
a copy of the GCC Runtime Library Exception along with this program; |
| 21 |
see the files COPYING3 and COPYING.RUNTIME respectively. If not, see |
| 22 |
<http://www.gnu.org/licenses/>. */ |
| 23 |
|
| 24 |
/* Implemented from the specification included in the Intel C++ Compiler |
| 25 |
User Guide and Reference, version 9.0. */ |
| 26 |
|
| 27 |
#ifndef _PMMINTRIN_H_INCLUDED |
| 28 |
#define _PMMINTRIN_H_INCLUDED |
| 29 |
|
| 30 |
/* We need definitions from the SSE2 and SSE header files*/ |
| 31 |
#include <emmintrin.h> |
| 32 |
|
| 33 |
#ifndef __SSE3__ |
| 34 |
#pragma GCC push_options |
| 35 |
#pragma GCC target("sse3") |
| 36 |
#define __DISABLE_SSE3__ |
| 37 |
#endif /* __SSE3__ */ |
| 38 |
|
| 39 |
/* Additional bits in the MXCSR. */ |
| 40 |
#define _MM_DENORMALS_ZERO_MASK 0x0040 |
| 41 |
#define _MM_DENORMALS_ZERO_ON 0x0040 |
| 42 |
#define _MM_DENORMALS_ZERO_OFF 0x0000 |
| 43 |
|
| 44 |
#define _MM_SET_DENORMALS_ZERO_MODE(mode) \ |
| 45 |
_mm_setcsr ((_mm_getcsr () & ~_MM_DENORMALS_ZERO_MASK) | (mode)) |
| 46 |
#define _MM_GET_DENORMALS_ZERO_MODE() \ |
| 47 |
(_mm_getcsr() & _MM_DENORMALS_ZERO_MASK) |
| 48 |
|
| 49 |
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__)) |
| 50 |
_mm_addsub_ps (__m128 __X, __m128 __Y) |
| 51 |
{ |
| 52 |
return (__m128) __builtin_ia32_addsubps ((__v4sf)__X, (__v4sf)__Y); |
| 53 |
} |
| 54 |
|
| 55 |
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__)) |
| 56 |
_mm_hadd_ps (__m128 __X, __m128 __Y) |
| 57 |
{ |
| 58 |
return (__m128) __builtin_ia32_haddps ((__v4sf)__X, (__v4sf)__Y); |
| 59 |
} |
| 60 |
|
| 61 |
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__)) |
| 62 |
_mm_hsub_ps (__m128 __X, __m128 __Y) |
| 63 |
{ |
| 64 |
return (__m128) __builtin_ia32_hsubps ((__v4sf)__X, (__v4sf)__Y); |
| 65 |
} |
| 66 |
|
| 67 |
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__)) |
| 68 |
_mm_movehdup_ps (__m128 __X) |
| 69 |
{ |
| 70 |
return (__m128) __builtin_ia32_movshdup ((__v4sf)__X); |
| 71 |
} |
| 72 |
|
| 73 |
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__)) |
| 74 |
_mm_moveldup_ps (__m128 __X) |
| 75 |
{ |
| 76 |
return (__m128) __builtin_ia32_movsldup ((__v4sf)__X); |
| 77 |
} |
| 78 |
|
| 79 |
extern __inline __m128d __attribute__((__gnu_inline__, __always_inline__, __artificial__)) |
| 80 |
_mm_addsub_pd (__m128d __X, __m128d __Y) |
| 81 |
{ |
| 82 |
return (__m128d) __builtin_ia32_addsubpd ((__v2df)__X, (__v2df)__Y); |
| 83 |
} |
| 84 |
|
| 85 |
extern __inline __m128d __attribute__((__gnu_inline__, __always_inline__, __artificial__)) |
| 86 |
_mm_hadd_pd (__m128d __X, __m128d __Y) |
| 87 |
{ |
| 88 |
return (__m128d) __builtin_ia32_haddpd ((__v2df)__X, (__v2df)__Y); |
| 89 |
} |
| 90 |
|
| 91 |
extern __inline __m128d __attribute__((__gnu_inline__, __always_inline__, __artificial__)) |
| 92 |
_mm_hsub_pd (__m128d __X, __m128d __Y) |
| 93 |
{ |
| 94 |
return (__m128d) __builtin_ia32_hsubpd ((__v2df)__X, (__v2df)__Y); |
| 95 |
} |
| 96 |
|
| 97 |
extern __inline __m128d __attribute__((__gnu_inline__, __always_inline__, __artificial__)) |
| 98 |
_mm_loaddup_pd (double const *__P) |
| 99 |
{ |
| 100 |
return _mm_load1_pd (__P); |
| 101 |
} |
| 102 |
|
| 103 |
extern __inline __m128d __attribute__((__gnu_inline__, __always_inline__, __artificial__)) |
| 104 |
_mm_movedup_pd (__m128d __X) |
| 105 |
{ |
| 106 |
return _mm_shuffle_pd (__X, __X, _MM_SHUFFLE2 (0,0)); |
| 107 |
} |
| 108 |
|
| 109 |
extern __inline __m128i __attribute__((__gnu_inline__, __always_inline__, __artificial__)) |
| 110 |
_mm_lddqu_si128 (__m128i const *__P) |
| 111 |
{ |
| 112 |
return (__m128i) __builtin_ia32_lddqu ((char const *)__P); |
| 113 |
} |
| 114 |
|
| 115 |
extern __inline void __attribute__((__gnu_inline__, __always_inline__, __artificial__)) |
| 116 |
_mm_monitor (void const * __P, unsigned int __E, unsigned int __H) |
| 117 |
{ |
| 118 |
__builtin_ia32_monitor (__P, __E, __H); |
| 119 |
} |
| 120 |
|
| 121 |
extern __inline void __attribute__((__gnu_inline__, __always_inline__, __artificial__)) |
| 122 |
_mm_mwait (unsigned int __E, unsigned int __H) |
| 123 |
{ |
| 124 |
__builtin_ia32_mwait (__E, __H); |
| 125 |
} |
| 126 |
|
| 127 |
#ifdef __DISABLE_SSE3__ |
| 128 |
#undef __DISABLE_SSE3__ |
| 129 |
#pragma GCC pop_options |
| 130 |
#endif /* __DISABLE_SSE3__ */ |
| 131 |
|
| 132 |
#endif /* _PMMINTRIN_H_INCLUDED */ |
