|
150
|
1 /*===------ avx512vlvp2intersectintrin.h - VL VP2INTERSECT 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 <avx512vlvp2intersectintrin.h> directly; include <immintrin.h> instead."
|
|
|
26 #endif
|
|
|
27
|
|
|
28 #ifndef _AVX512VLVP2INTERSECT_H
|
|
|
29 #define _AVX512VLVP2INTERSECT_H
|
|
|
30
|
|
|
31 #define __DEFAULT_FN_ATTRS128 \
|
|
|
32 __attribute__((__always_inline__, __nodebug__, __target__("avx512vl,avx512vp2intersect"), \
|
|
|
33 __min_vector_width__(128)))
|
|
|
34
|
|
|
35 #define __DEFAULT_FN_ATTRS256 \
|
|
|
36 __attribute__((__always_inline__, __nodebug__, __target__("avx512vl,avx512vp2intersect"), \
|
|
|
37 __min_vector_width__(256)))
|
|
|
38 /// Store, in an even/odd pair of mask registers, the indicators of the
|
|
|
39 /// locations of value matches between dwords in operands __a and __b.
|
|
|
40 ///
|
|
|
41 /// \headerfile <x86intrin.h>
|
|
|
42 ///
|
|
|
43 /// This intrinsic corresponds to the <c> VP2INTERSECTD </c> instruction.
|
|
|
44 ///
|
|
|
45 /// \param __a
|
|
|
46 /// A 256-bit vector of [8 x i32].
|
|
|
47 /// \param __b
|
|
|
48 /// A 256-bit vector of [8 x i32]
|
|
|
49 /// \param __m0
|
|
|
50 /// A pointer point to 8-bit mask
|
|
|
51 /// \param __m1
|
|
|
52 /// A pointer point to 8-bit mask
|
|
|
53 static __inline__ void __DEFAULT_FN_ATTRS256
|
|
|
54 _mm256_2intersect_epi32(__m256i __a, __m256i __b, __mmask8 *__m0, __mmask8 *__m1) {
|
|
|
55 __builtin_ia32_vp2intersect_d_256((__v8si)__a, (__v8si)__b, __m0, __m1);
|
|
|
56 }
|
|
|
57
|
|
|
58 /// Store, in an even/odd pair of mask registers, the indicators of the
|
|
|
59 /// locations of value matches between quadwords in operands __a and __b.
|
|
|
60 ///
|
|
|
61 /// \headerfile <x86intrin.h>
|
|
|
62 ///
|
|
|
63 /// This intrinsic corresponds to the <c> VP2INTERSECTQ </c> instruction.
|
|
|
64 ///
|
|
|
65 /// \param __a
|
|
|
66 /// A 256-bit vector of [4 x i64].
|
|
|
67 /// \param __b
|
|
|
68 /// A 256-bit vector of [4 x i64]
|
|
|
69 /// \param __m0
|
|
|
70 /// A pointer point to 8-bit mask
|
|
|
71 /// \param __m1
|
|
|
72 /// A pointer point to 8-bit mask
|
|
|
73 static __inline__ void __DEFAULT_FN_ATTRS256
|
|
|
74 _mm256_2intersect_epi64(__m256i __a, __m256i __b, __mmask8 *__m0, __mmask8 *__m1) {
|
|
|
75 __builtin_ia32_vp2intersect_q_256((__v4di)__a, (__v4di)__b, __m0, __m1);
|
|
|
76 }
|
|
|
77
|
|
|
78 /// Store, in an even/odd pair of mask registers, the indicators of the
|
|
|
79 /// locations of value matches between dwords in operands __a and __b.
|
|
|
80 ///
|
|
|
81 /// \headerfile <x86intrin.h>
|
|
|
82 ///
|
|
|
83 /// This intrinsic corresponds to the <c> VP2INTERSECTD </c> instruction.
|
|
|
84 ///
|
|
|
85 /// \param __a
|
|
|
86 /// A 128-bit vector of [4 x i32].
|
|
|
87 /// \param __b
|
|
|
88 /// A 128-bit vector of [4 x i32]
|
|
|
89 /// \param __m0
|
|
|
90 /// A pointer point to 8-bit mask
|
|
|
91 /// \param __m1
|
|
|
92 /// A pointer point to 8-bit mask
|
|
|
93 static __inline__ void __DEFAULT_FN_ATTRS128
|
|
|
94 _mm_2intersect_epi32(__m128i __a, __m128i __b, __mmask8 *__m0, __mmask8 *__m1) {
|
|
|
95 __builtin_ia32_vp2intersect_d_128((__v4si)__a, (__v4si)__b, __m0, __m1);
|
|
|
96 }
|
|
|
97
|
|
|
98 /// Store, in an even/odd pair of mask registers, the indicators of the
|
|
|
99 /// locations of value matches between quadwords in operands __a and __b.
|
|
|
100 ///
|
|
|
101 /// \headerfile <x86intrin.h>
|
|
|
102 ///
|
|
|
103 /// This intrinsic corresponds to the <c> VP2INTERSECTQ </c> instruction.
|
|
|
104 ///
|
|
|
105 /// \param __a
|
|
|
106 /// A 128-bit vector of [2 x i64].
|
|
|
107 /// \param __b
|
|
|
108 /// A 128-bit vector of [2 x i64]
|
|
|
109 /// \param __m0
|
|
|
110 /// A pointer point to 8-bit mask
|
|
|
111 /// \param __m1
|
|
|
112 /// A pointer point to 8-bit mask
|
|
|
113 static __inline__ void __DEFAULT_FN_ATTRS128
|
|
|
114 _mm_2intersect_epi64(__m128i __a, __m128i __b, __mmask8 *__m0, __mmask8 *__m1) {
|
|
|
115 __builtin_ia32_vp2intersect_q_128((__v2di)__a, (__v2di)__b, __m0, __m1);
|
|
|
116 }
|
|
|
117
|
|
|
118 #undef __DEFAULT_FN_ATTRS128
|
|
|
119 #undef __DEFAULT_FN_ATTRS256
|
|
|
120
|
|
|
121 #endif
|
