1 /* Copyright (C) 2003-2021 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 #include <mwaitintrin.h>
33
34 #ifndef __SSE3__
35 #pragma GCC push_options
36 #pragma GCC target("sse3")
37 #define __DISABLE_SSE3__
38 #endif /* __SSE3__ */
39
40 /* Additional bits in the MXCSR. */
41 #define _MM_DENORMALS_ZERO_MASK 0x0040
42 #define _MM_DENORMALS_ZERO_ON 0x0040
43 #define _MM_DENORMALS_ZERO_OFF 0x0000
44
45 #define _MM_SET_DENORMALS_ZERO_MODE(mode) \
46 _mm_setcsr ((_mm_getcsr () & ~_MM_DENORMALS_ZERO_MASK) | (mode))
47 #define _MM_GET_DENORMALS_ZERO_MODE() \
48 (_mm_getcsr() & _MM_DENORMALS_ZERO_MASK)
49
50 extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_addsub_ps(__m128 __X,__m128 __Y)51 _mm_addsub_ps (__m128 __X, __m128 __Y)
52 {
53 return (__m128) __builtin_ia32_addsubps ((__v4sf)__X, (__v4sf)__Y);
54 }
55
56 extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_hadd_ps(__m128 __X,__m128 __Y)57 _mm_hadd_ps (__m128 __X, __m128 __Y)
58 {
59 return (__m128) __builtin_ia32_haddps ((__v4sf)__X, (__v4sf)__Y);
60 }
61
62 extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_hsub_ps(__m128 __X,__m128 __Y)63 _mm_hsub_ps (__m128 __X, __m128 __Y)
64 {
65 return (__m128) __builtin_ia32_hsubps ((__v4sf)__X, (__v4sf)__Y);
66 }
67
68 extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_movehdup_ps(__m128 __X)69 _mm_movehdup_ps (__m128 __X)
70 {
71 return (__m128) __builtin_ia32_movshdup ((__v4sf)__X);
72 }
73
74 extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_moveldup_ps(__m128 __X)75 _mm_moveldup_ps (__m128 __X)
76 {
77 return (__m128) __builtin_ia32_movsldup ((__v4sf)__X);
78 }
79
80 extern __inline __m128d __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_addsub_pd(__m128d __X,__m128d __Y)81 _mm_addsub_pd (__m128d __X, __m128d __Y)
82 {
83 return (__m128d) __builtin_ia32_addsubpd ((__v2df)__X, (__v2df)__Y);
84 }
85
86 extern __inline __m128d __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_hadd_pd(__m128d __X,__m128d __Y)87 _mm_hadd_pd (__m128d __X, __m128d __Y)
88 {
89 return (__m128d) __builtin_ia32_haddpd ((__v2df)__X, (__v2df)__Y);
90 }
91
92 extern __inline __m128d __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_hsub_pd(__m128d __X,__m128d __Y)93 _mm_hsub_pd (__m128d __X, __m128d __Y)
94 {
95 return (__m128d) __builtin_ia32_hsubpd ((__v2df)__X, (__v2df)__Y);
96 }
97
98 extern __inline __m128d __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_loaddup_pd(double const * __P)99 _mm_loaddup_pd (double const *__P)
100 {
101 return _mm_load1_pd (__P);
102 }
103
104 extern __inline __m128d __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_movedup_pd(__m128d __X)105 _mm_movedup_pd (__m128d __X)
106 {
107 return _mm_shuffle_pd (__X, __X, _MM_SHUFFLE2 (0,0));
108 }
109
110 extern __inline __m128i __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_lddqu_si128(__m128i const * __P)111 _mm_lddqu_si128 (__m128i const *__P)
112 {
113 return (__m128i) __builtin_ia32_lddqu ((char const *)__P);
114 }
115
116 #ifdef __DISABLE_SSE3__
117 #undef __DISABLE_SSE3__
118 #pragma GCC pop_options
119 #endif /* __DISABLE_SSE3__ */
120
121 #endif /* _PMMINTRIN_H_INCLUDED */
122