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 |
|
|
#if !defined _X86INTRIN_H_INCLUDED && !defined _IMMINTRIN_H_INCLUDED
|
25 |
|
|
# error "Never use <f16intrin.h> directly; include <x86intrin.h> or <immintrin.h> instead."
|
26 |
|
|
#endif
|
27 |
|
|
|
28 |
|
|
#ifndef __F16C__
|
29 |
|
|
# error "F16C instruction set not enabled"
|
30 |
|
|
#else
|
31 |
|
|
|
32 |
|
|
#ifndef _F16CINTRIN_H_INCLUDED
|
33 |
|
|
#define _F16CINTRIN_H_INCLUDED
|
34 |
|
|
|
35 |
|
|
extern __inline float __attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
36 |
|
|
_cvtsh_ss (unsigned short __S)
|
37 |
|
|
{
|
38 |
|
|
__v8hi __H = __extension__ (__v8hi){ __S, 0, 0, 0, 0, 0, 0, 0 };
|
39 |
|
|
__v4sf __A = __builtin_ia32_vcvtph2ps (__H);
|
40 |
|
|
return __builtin_ia32_vec_ext_v4sf (__A, 0);
|
41 |
|
|
}
|
42 |
|
|
|
43 |
|
|
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
44 |
|
|
_mm_cvtph_ps (__m128i __A)
|
45 |
|
|
{
|
46 |
|
|
return (__m128) __builtin_ia32_vcvtph2ps ((__v8hi) __A);
|
47 |
|
|
}
|
48 |
|
|
|
49 |
|
|
extern __inline __m256 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
50 |
|
|
_mm256_cvtph_ps (__m128i __A)
|
51 |
|
|
{
|
52 |
|
|
return (__m256) __builtin_ia32_vcvtph2ps256 ((__v8hi) __A);
|
53 |
|
|
}
|
54 |
|
|
|
55 |
|
|
#ifdef __OPTIMIZE__
|
56 |
|
|
extern __inline unsigned short __attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
57 |
|
|
_cvtss_sh (float __F, const int __I)
|
58 |
|
|
{
|
59 |
|
|
__v4sf __A = __extension__ (__v4sf){ __F, 0, 0, 0 };
|
60 |
|
|
__v8hi __H = __builtin_ia32_vcvtps2ph (__A, __I);
|
61 |
|
|
return (unsigned short) __builtin_ia32_vec_ext_v8hi (__H, 0);
|
62 |
|
|
}
|
63 |
|
|
|
64 |
|
|
extern __inline __m128i __attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
65 |
|
|
_mm_cvtps_ph (__m128 __A, const int __I)
|
66 |
|
|
{
|
67 |
|
|
return (__m128i) __builtin_ia32_vcvtps2ph ((__v4sf) __A, __I);
|
68 |
|
|
}
|
69 |
|
|
|
70 |
|
|
extern __inline __m128i __attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
71 |
|
|
_mm256_cvtps_ph (__m256 __A, const int __I)
|
72 |
|
|
{
|
73 |
|
|
return (__m128i) __builtin_ia32_vcvtps2ph256 ((__v8sf) __A, __I);
|
74 |
|
|
}
|
75 |
|
|
#else
|
76 |
|
|
#define _cvtss_sh(__F, __I) \
|
77 |
|
|
(__extension__ \
|
78 |
|
|
({ \
|
79 |
|
|
__v4sf __A = __extension__ (__v4sf){ __F, 0, 0, 0 }; \
|
80 |
|
|
__v8hi __H = __builtin_ia32_vcvtps2ph (__A, __I); \
|
81 |
|
|
(unsigned short) __builtin_ia32_vec_ext_v8hi (__H, 0); \
|
82 |
|
|
}))
|
83 |
|
|
|
84 |
|
|
#define _mm_cvtps_ph(A, I) \
|
85 |
|
|
((__m128i) __builtin_ia32_vcvtps2ph ((__v4sf)(__m128) A, (int) (I)))
|
86 |
|
|
|
87 |
|
|
#define _mm256_cvtps_ph(A, I) \
|
88 |
|
|
((__m128i) __builtin_ia32_vcvtps2ph256 ((__v8sf)(__m256) A, (int) (I)))
|
89 |
|
|
#endif /* __OPTIMIZE */
|
90 |
|
|
|
91 |
|
|
#endif /* _F16CINTRIN_H_INCLUDED */
|
92 |
|
|
#endif /* __F16C__ */
|