blob: c7fa3cf313e327b9894b73bb48f4b1af5cc291d3 [file] [log] [blame]
Louis Pullen-Freilich2d897072020-01-30 14:59:03 +00001/*===------------- avx512pfintrin.h - PF intrinsics ------------------===
2 *
3 *
4 * Permission is hereby granted, free of charge, to any person obtaining a copy
5 * of this software and associated documentation files (the "Software"), to deal
6 * in the Software without restriction, including without limitation the rights
7 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
8 * copies of the Software, and to permit persons to whom the Software is
9 * furnished to do so, subject to the following conditions:
10 *
11 * The above copyright notice and this permission notice shall be included in
12 * all copies or substantial portions of the Software.
13 *
14 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
15 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
16 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
17 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
18 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
19 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
20 * THE SOFTWARE.
21 *
22 *===-----------------------------------------------------------------------===
23 */
24#ifndef __IMMINTRIN_H
25#error "Never use <avx512pfintrin.h> directly; include <immintrin.h> instead."
26#endif
27
28#ifndef __AVX512PFINTRIN_H
29#define __AVX512PFINTRIN_H
30
31/* Define the default attributes for the functions in this file. */
32#define __DEFAULT_FN_ATTRS __attribute__((__always_inline__, __nodebug__, __target__("avx512pf")))
33
34#define _mm512_mask_prefetch_i32gather_pd(index, mask, addr, scale, hint) __extension__ ({\
35 __builtin_ia32_gatherpfdpd((__mmask8)(mask), (__v8si)(__m256i)(index), \
36 (long long const *)(addr), (int)(scale), \
37 (int)(hint)); })
38
39#define _mm512_prefetch_i32gather_pd(index, addr, scale, hint) __extension__ ({\
40 __builtin_ia32_gatherpfdpd((__mmask8) -1, (__v8si)(__m256i)(index), \
41 (long long const *)(addr), (int)(scale), \
42 (int)(hint)); })
43
44#define _mm512_mask_prefetch_i32gather_ps(index, mask, addr, scale, hint) ({\
45 __builtin_ia32_gatherpfdps((__mmask16)(mask), \
46 (__v16si)(__m512i)(index), (int const *)(addr), \
47 (int)(scale), (int)(hint)); })
48
49#define _mm512_prefetch_i32gather_ps(index, addr, scale, hint) ({\
50 __builtin_ia32_gatherpfdps((__mmask16) -1, \
51 (__v16si)(__m512i)(index), (int const *)(addr), \
52 (int)(scale), (int)(hint)); })
53
54#define _mm512_mask_prefetch_i64gather_pd(index, mask, addr, scale, hint) __extension__ ({\
55 __builtin_ia32_gatherpfqpd((__mmask8)(mask), (__v8di)(__m512i)(index), \
56 (long long const *)(addr), (int)(scale), \
57 (int)(hint)); })
58
59#define _mm512_prefetch_i64gather_pd(index, addr, scale, hint) __extension__ ({\
60 __builtin_ia32_gatherpfqpd((__mmask8) -1, (__v8di)(__m512i)(index), \
61 (long long const *)(addr), (int)(scale), \
62 (int)(hint)); })
63
64#define _mm512_mask_prefetch_i64gather_ps(index, mask, addr, scale, hint) ({\
65 __builtin_ia32_gatherpfqps((__mmask8)(mask), (__v8di)(__m512i)(index), \
66 (int const *)(addr), (int)(scale), (int)(hint)); })
67
68#define _mm512_prefetch_i64gather_ps(index, addr, scale, hint) ({\
69 __builtin_ia32_gatherpfqps((__mmask8) -1, (__v8di)(__m512i)(index), \
70 (int const *)(addr), (int)(scale), (int)(hint)); })
71
72#define _mm512_prefetch_i32scatter_pd(addr, index, scale, hint) __extension__ ({\
73 __builtin_ia32_scatterpfdpd((__mmask8)-1, (__v8si)(__m256i)(index), \
74 (long long *)(addr), (int)(scale), \
75 (int)(hint)); })
76
77#define _mm512_mask_prefetch_i32scatter_pd(addr, mask, index, scale, hint) __extension__ ({\
78 __builtin_ia32_scatterpfdpd((__mmask8)(mask), (__v8si)(__m256i)(index), \
79 (long long *)(addr), (int)(scale), \
80 (int)(hint)); })
81
82#define _mm512_prefetch_i32scatter_ps(addr, index, scale, hint) __extension__ ({\
83 __builtin_ia32_scatterpfdps((__mmask16)-1, (__v16si)(__m512i)(index), \
84 (int *)(addr), (int)(scale), (int)(hint)); })
85
86#define _mm512_mask_prefetch_i32scatter_ps(addr, mask, index, scale, hint) __extension__ ({\
87 __builtin_ia32_scatterpfdps((__mmask16)(mask), \
88 (__v16si)(__m512i)(index), (int *)(addr), \
89 (int)(scale), (int)(hint)); })
90
91#define _mm512_prefetch_i64scatter_pd(addr, index, scale, hint) __extension__ ({\
92 __builtin_ia32_scatterpfqpd((__mmask8)-1, (__v8di)(__m512i)(index), \
93 (long long *)(addr), (int)(scale), \
94 (int)(hint)); })
95
96#define _mm512_mask_prefetch_i64scatter_pd(addr, mask, index, scale, hint) __extension__ ({\
97 __builtin_ia32_scatterpfqpd((__mmask8)(mask), (__v8di)(__m512i)(index), \
98 (long long *)(addr), (int)(scale), \
99 (int)(hint)); })
100
101#define _mm512_prefetch_i64scatter_ps(addr, index, scale, hint) __extension__ ({\
102 __builtin_ia32_scatterpfqps((__mmask8)-1, (__v8di)(__m512i)(index), \
103 (int *)(addr), (int)(scale), (int)(hint)); })
104
105#define _mm512_mask_prefetch_i64scatter_ps(addr, mask, index, scale, hint) __extension__ ({\
106 __builtin_ia32_scatterpfqps((__mmask8)(mask), (__v8di)(__m512i)(index), \
107 (int *)(addr), (int)(scale), (int)(hint)); })
108
109#undef __DEFAULT_FN_ATTRS
110
111#endif