1 |
709 |
jeremybenn |
/* Copyright (C) 2011 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 |
|
|
#ifndef _IMMINTRIN_H_INCLUDED
|
25 |
|
|
# error "Never use <fmaintrin.h> directly; include <immintrin.h> instead."
|
26 |
|
|
#endif
|
27 |
|
|
|
28 |
|
|
#ifndef _FMAINTRIN_H_INCLUDED
|
29 |
|
|
#define _FMAINTRIN_H_INCLUDED
|
30 |
|
|
|
31 |
|
|
#ifndef __FMA__
|
32 |
|
|
# error "FMA instruction set not enabled"
|
33 |
|
|
#else
|
34 |
|
|
|
35 |
|
|
extern __inline __m128d
|
36 |
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
37 |
|
|
_mm_fmadd_pd (__m128d __A, __m128d __B, __m128d __C)
|
38 |
|
|
{
|
39 |
|
|
return (__m128d)__builtin_ia32_vfmaddpd ((__v2df)__A, (__v2df)__B,
|
40 |
|
|
(__v2df)__C);
|
41 |
|
|
}
|
42 |
|
|
|
43 |
|
|
extern __inline __m256d
|
44 |
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
45 |
|
|
_mm256_fmadd_pd (__m256d __A, __m256d __B, __m256d __C)
|
46 |
|
|
{
|
47 |
|
|
return (__m256d)__builtin_ia32_vfmaddpd256 ((__v4df)__A, (__v4df)__B,
|
48 |
|
|
(__v4df)__C);
|
49 |
|
|
}
|
50 |
|
|
|
51 |
|
|
extern __inline __m128
|
52 |
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
53 |
|
|
_mm_fmadd_ps (__m128 __A, __m128 __B, __m128 __C)
|
54 |
|
|
{
|
55 |
|
|
return (__m128)__builtin_ia32_vfmaddps ((__v4sf)__A, (__v4sf)__B,
|
56 |
|
|
(__v4sf)__C);
|
57 |
|
|
}
|
58 |
|
|
|
59 |
|
|
extern __inline __m256
|
60 |
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
61 |
|
|
_mm256_fmadd_ps (__m256 __A, __m256 __B, __m256 __C)
|
62 |
|
|
{
|
63 |
|
|
return (__m256)__builtin_ia32_vfmaddps256 ((__v8sf)__A, (__v8sf)__B,
|
64 |
|
|
(__v8sf)__C);
|
65 |
|
|
}
|
66 |
|
|
|
67 |
|
|
extern __inline __m128d
|
68 |
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
69 |
|
|
_mm_fmadd_sd (__m128d __A, __m128d __B, __m128d __C)
|
70 |
|
|
{
|
71 |
|
|
return (__m128d) __builtin_ia32_vfmaddsd3 ((__v2df)__A, (__v2df)__B,
|
72 |
|
|
(__v2df)__C);
|
73 |
|
|
}
|
74 |
|
|
|
75 |
|
|
extern __inline __m128
|
76 |
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
77 |
|
|
_mm_fmadd_ss (__m128 __A, __m128 __B, __m128 __C)
|
78 |
|
|
{
|
79 |
|
|
return (__m128) __builtin_ia32_vfmaddss3 ((__v4sf)__A, (__v4sf)__B,
|
80 |
|
|
(__v4sf)__C);
|
81 |
|
|
}
|
82 |
|
|
|
83 |
|
|
extern __inline __m128d
|
84 |
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
85 |
|
|
_mm_fmsub_pd (__m128d __A, __m128d __B, __m128d __C)
|
86 |
|
|
{
|
87 |
|
|
return (__m128d)__builtin_ia32_vfmaddpd ((__v2df)__A, (__v2df)__B,
|
88 |
|
|
-(__v2df)__C);
|
89 |
|
|
}
|
90 |
|
|
|
91 |
|
|
extern __inline __m256d
|
92 |
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
93 |
|
|
_mm256_fmsub_pd (__m256d __A, __m256d __B, __m256d __C)
|
94 |
|
|
{
|
95 |
|
|
return (__m256d)__builtin_ia32_vfmaddpd256 ((__v4df)__A, (__v4df)__B,
|
96 |
|
|
-(__v4df)__C);
|
97 |
|
|
}
|
98 |
|
|
|
99 |
|
|
extern __inline __m128
|
100 |
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
101 |
|
|
_mm_fmsub_ps (__m128 __A, __m128 __B, __m128 __C)
|
102 |
|
|
{
|
103 |
|
|
return (__m128)__builtin_ia32_vfmaddps ((__v4sf)__A, (__v4sf)__B,
|
104 |
|
|
-(__v4sf)__C);
|
105 |
|
|
}
|
106 |
|
|
|
107 |
|
|
extern __inline __m256
|
108 |
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
109 |
|
|
_mm256_fmsub_ps (__m256 __A, __m256 __B, __m256 __C)
|
110 |
|
|
{
|
111 |
|
|
return (__m256)__builtin_ia32_vfmaddps256 ((__v8sf)__A, (__v8sf)__B,
|
112 |
|
|
-(__v8sf)__C);
|
113 |
|
|
}
|
114 |
|
|
|
115 |
|
|
extern __inline __m128d
|
116 |
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
117 |
|
|
_mm_fmsub_sd (__m128d __A, __m128d __B, __m128d __C)
|
118 |
|
|
{
|
119 |
|
|
return (__m128d)__builtin_ia32_vfmaddsd3 ((__v2df)__A, (__v2df)__B,
|
120 |
|
|
-(__v2df)__C);
|
121 |
|
|
}
|
122 |
|
|
|
123 |
|
|
extern __inline __m128
|
124 |
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
125 |
|
|
_mm_fmsub_ss (__m128 __A, __m128 __B, __m128 __C)
|
126 |
|
|
{
|
127 |
|
|
return (__m128)__builtin_ia32_vfmaddss3 ((__v4sf)__A, (__v4sf)__B,
|
128 |
|
|
-(__v4sf)__C);
|
129 |
|
|
}
|
130 |
|
|
|
131 |
|
|
extern __inline __m128d
|
132 |
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
133 |
|
|
_mm_fnmadd_pd (__m128d __A, __m128d __B, __m128d __C)
|
134 |
|
|
{
|
135 |
|
|
return (__m128d)__builtin_ia32_vfmaddpd (-(__v2df)__A, (__v2df)__B,
|
136 |
|
|
(__v2df)__C);
|
137 |
|
|
}
|
138 |
|
|
|
139 |
|
|
extern __inline __m256d
|
140 |
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
141 |
|
|
_mm256_fnmadd_pd (__m256d __A, __m256d __B, __m256d __C)
|
142 |
|
|
{
|
143 |
|
|
return (__m256d)__builtin_ia32_vfmaddpd256 (-(__v4df)__A, (__v4df)__B,
|
144 |
|
|
(__v4df)__C);
|
145 |
|
|
}
|
146 |
|
|
|
147 |
|
|
extern __inline __m128
|
148 |
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
149 |
|
|
_mm_fnmadd_ps (__m128 __A, __m128 __B, __m128 __C)
|
150 |
|
|
{
|
151 |
|
|
return (__m128)__builtin_ia32_vfmaddps (-(__v4sf)__A, (__v4sf)__B,
|
152 |
|
|
(__v4sf)__C);
|
153 |
|
|
}
|
154 |
|
|
|
155 |
|
|
extern __inline __m256
|
156 |
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
157 |
|
|
_mm256_fnmadd_ps (__m256 __A, __m256 __B, __m256 __C)
|
158 |
|
|
{
|
159 |
|
|
return (__m256)__builtin_ia32_vfmaddps256 (-(__v8sf)__A, (__v8sf)__B,
|
160 |
|
|
(__v8sf)__C);
|
161 |
|
|
}
|
162 |
|
|
|
163 |
|
|
extern __inline __m128d
|
164 |
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
165 |
|
|
_mm_fnmadd_sd (__m128d __A, __m128d __B, __m128d __C)
|
166 |
|
|
{
|
167 |
|
|
return (__m128d)__builtin_ia32_vfmaddsd3 (-(__v2df)__A, (__v2df)__B,
|
168 |
|
|
(__v2df)__C);
|
169 |
|
|
}
|
170 |
|
|
|
171 |
|
|
extern __inline __m128
|
172 |
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
173 |
|
|
_mm_fnmadd_ss (__m128 __A, __m128 __B, __m128 __C)
|
174 |
|
|
{
|
175 |
|
|
return (__m128)__builtin_ia32_vfmaddss3 (-(__v4sf)__A, (__v4sf)__B,
|
176 |
|
|
(__v4sf)__C);
|
177 |
|
|
}
|
178 |
|
|
|
179 |
|
|
extern __inline __m128d
|
180 |
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
181 |
|
|
_mm_fnmsub_pd (__m128d __A, __m128d __B, __m128d __C)
|
182 |
|
|
{
|
183 |
|
|
return (__m128d)__builtin_ia32_vfmaddpd (-(__v2df)__A, (__v2df)__B,
|
184 |
|
|
-(__v2df)__C);
|
185 |
|
|
}
|
186 |
|
|
|
187 |
|
|
extern __inline __m256d
|
188 |
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
189 |
|
|
_mm256_fnmsub_pd (__m256d __A, __m256d __B, __m256d __C)
|
190 |
|
|
{
|
191 |
|
|
return (__m256d)__builtin_ia32_vfmaddpd256 (-(__v4df)__A, (__v4df)__B,
|
192 |
|
|
-(__v4df)__C);
|
193 |
|
|
}
|
194 |
|
|
|
195 |
|
|
extern __inline __m128
|
196 |
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
197 |
|
|
_mm_fnmsub_ps (__m128 __A, __m128 __B, __m128 __C)
|
198 |
|
|
{
|
199 |
|
|
return (__m128)__builtin_ia32_vfmaddps (-(__v4sf)__A, (__v4sf)__B,
|
200 |
|
|
-(__v4sf)__C);
|
201 |
|
|
}
|
202 |
|
|
|
203 |
|
|
extern __inline __m256
|
204 |
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
205 |
|
|
_mm256_fnmsub_ps (__m256 __A, __m256 __B, __m256 __C)
|
206 |
|
|
{
|
207 |
|
|
return (__m256)__builtin_ia32_vfmaddps256 (-(__v8sf)__A, (__v8sf)__B,
|
208 |
|
|
-(__v8sf)__C);
|
209 |
|
|
}
|
210 |
|
|
|
211 |
|
|
extern __inline __m128d
|
212 |
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
213 |
|
|
_mm_fnmsub_sd (__m128d __A, __m128d __B, __m128d __C)
|
214 |
|
|
{
|
215 |
|
|
return (__m128d)__builtin_ia32_vfmaddsd3 (-(__v2df)__A, (__v2df)__B,
|
216 |
|
|
-(__v2df)__C);
|
217 |
|
|
}
|
218 |
|
|
|
219 |
|
|
extern __inline __m128
|
220 |
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
221 |
|
|
_mm_fnmsub_ss (__m128 __A, __m128 __B, __m128 __C)
|
222 |
|
|
{
|
223 |
|
|
return (__m128)__builtin_ia32_vfmaddss3 (-(__v4sf)__A, (__v4sf)__B,
|
224 |
|
|
-(__v4sf)__C);
|
225 |
|
|
}
|
226 |
|
|
|
227 |
|
|
extern __inline __m128d
|
228 |
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
229 |
|
|
_mm_fmaddsub_pd (__m128d __A, __m128d __B, __m128d __C)
|
230 |
|
|
{
|
231 |
|
|
return (__m128d)__builtin_ia32_vfmaddsubpd ((__v2df)__A, (__v2df)__B,
|
232 |
|
|
(__v2df)__C);
|
233 |
|
|
}
|
234 |
|
|
|
235 |
|
|
extern __inline __m256d
|
236 |
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
237 |
|
|
_mm256_fmaddsub_pd (__m256d __A, __m256d __B, __m256d __C)
|
238 |
|
|
{
|
239 |
|
|
return (__m256d)__builtin_ia32_vfmaddsubpd256 ((__v4df)__A,
|
240 |
|
|
(__v4df)__B,
|
241 |
|
|
(__v4df)__C);
|
242 |
|
|
}
|
243 |
|
|
|
244 |
|
|
extern __inline __m128
|
245 |
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
246 |
|
|
_mm_fmaddsub_ps (__m128 __A, __m128 __B, __m128 __C)
|
247 |
|
|
{
|
248 |
|
|
return (__m128)__builtin_ia32_vfmaddsubps ((__v4sf)__A, (__v4sf)__B,
|
249 |
|
|
(__v4sf)__C);
|
250 |
|
|
}
|
251 |
|
|
|
252 |
|
|
extern __inline __m256
|
253 |
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
254 |
|
|
_mm256_fmaddsub_ps (__m256 __A, __m256 __B, __m256 __C)
|
255 |
|
|
{
|
256 |
|
|
return (__m256)__builtin_ia32_vfmaddsubps256 ((__v8sf)__A,
|
257 |
|
|
(__v8sf)__B,
|
258 |
|
|
(__v8sf)__C);
|
259 |
|
|
}
|
260 |
|
|
|
261 |
|
|
extern __inline __m128d
|
262 |
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
263 |
|
|
_mm_fmsubadd_pd (__m128d __A, __m128d __B, __m128d __C)
|
264 |
|
|
{
|
265 |
|
|
return (__m128d)__builtin_ia32_vfmaddsubpd ((__v2df)__A, (__v2df)__B,
|
266 |
|
|
-(__v2df)__C);
|
267 |
|
|
}
|
268 |
|
|
|
269 |
|
|
extern __inline __m256d
|
270 |
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
271 |
|
|
_mm256_fmsubadd_pd (__m256d __A, __m256d __B, __m256d __C)
|
272 |
|
|
{
|
273 |
|
|
return (__m256d)__builtin_ia32_vfmaddsubpd256 ((__v4df)__A,
|
274 |
|
|
(__v4df)__B,
|
275 |
|
|
-(__v4df)__C);
|
276 |
|
|
}
|
277 |
|
|
|
278 |
|
|
extern __inline __m128
|
279 |
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
280 |
|
|
_mm_fmsubadd_ps (__m128 __A, __m128 __B, __m128 __C)
|
281 |
|
|
{
|
282 |
|
|
return (__m128)__builtin_ia32_vfmaddsubps ((__v4sf)__A, (__v4sf)__B,
|
283 |
|
|
-(__v4sf)__C);
|
284 |
|
|
}
|
285 |
|
|
|
286 |
|
|
extern __inline __m256
|
287 |
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
288 |
|
|
_mm256_fmsubadd_ps (__m256 __A, __m256 __B, __m256 __C)
|
289 |
|
|
{
|
290 |
|
|
return (__m256)__builtin_ia32_vfmaddsubps256 ((__v8sf)__A,
|
291 |
|
|
(__v8sf)__B,
|
292 |
|
|
-(__v8sf)__C);
|
293 |
|
|
}
|
294 |
|
|
|
295 |
|
|
#endif
|
296 |
|
|
|
297 |
|
|
#endif
|