blob: e63902743c0606911da316386ef953b37c30db02 [file] [log] [blame]
Logan Chien2833ffb2018-10-09 10:03:24 +08001/*===------------- avx512cdintrin.h - AVX512CD 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 <avx512cdintrin.h> directly; include <immintrin.h> instead."
26#endif
27
28#ifndef __AVX512CDINTRIN_H
29#define __AVX512CDINTRIN_H
30
31/* Define the default attributes for the functions in this file. */
Logan Chien55afb0a2018-10-15 10:42:14 +080032#define __DEFAULT_FN_ATTRS __attribute__((__always_inline__, __nodebug__, __target__("avx512cd"), __min_vector_width__(512)))
Logan Chien2833ffb2018-10-09 10:03:24 +080033
34static __inline__ __m512i __DEFAULT_FN_ATTRS
35_mm512_conflict_epi64 (__m512i __A)
36{
37 return (__m512i) __builtin_ia32_vpconflictdi_512_mask ((__v8di) __A,
38 (__v8di) _mm512_setzero_si512 (),
39 (__mmask8) -1);
40}
41
42static __inline__ __m512i __DEFAULT_FN_ATTRS
43_mm512_mask_conflict_epi64 (__m512i __W, __mmask8 __U, __m512i __A)
44{
45 return (__m512i) __builtin_ia32_vpconflictdi_512_mask ((__v8di) __A,
46 (__v8di) __W,
47 (__mmask8) __U);
48}
49
50static __inline__ __m512i __DEFAULT_FN_ATTRS
51_mm512_maskz_conflict_epi64 (__mmask8 __U, __m512i __A)
52{
53 return (__m512i) __builtin_ia32_vpconflictdi_512_mask ((__v8di) __A,
54 (__v8di) _mm512_setzero_si512 (),
55 (__mmask8) __U);
56}
57
58static __inline__ __m512i __DEFAULT_FN_ATTRS
59_mm512_conflict_epi32 (__m512i __A)
60{
61 return (__m512i) __builtin_ia32_vpconflictsi_512_mask ((__v16si) __A,
62 (__v16si) _mm512_setzero_si512 (),
63 (__mmask16) -1);
64}
65
66static __inline__ __m512i __DEFAULT_FN_ATTRS
67_mm512_mask_conflict_epi32 (__m512i __W, __mmask16 __U, __m512i __A)
68{
69 return (__m512i) __builtin_ia32_vpconflictsi_512_mask ((__v16si) __A,
70 (__v16si) __W,
71 (__mmask16) __U);
72}
73
74static __inline__ __m512i __DEFAULT_FN_ATTRS
75_mm512_maskz_conflict_epi32 (__mmask16 __U, __m512i __A)
76{
77 return (__m512i) __builtin_ia32_vpconflictsi_512_mask ((__v16si) __A,
78 (__v16si) _mm512_setzero_si512 (),
79 (__mmask16) __U);
80}
81
82static __inline__ __m512i __DEFAULT_FN_ATTRS
83_mm512_lzcnt_epi32 (__m512i __A)
84{
Logan Chien55afb0a2018-10-15 10:42:14 +080085 return (__m512i) __builtin_ia32_vplzcntd_512 ((__v16si) __A);
Logan Chien2833ffb2018-10-09 10:03:24 +080086}
87
88static __inline__ __m512i __DEFAULT_FN_ATTRS
89_mm512_mask_lzcnt_epi32 (__m512i __W, __mmask16 __U, __m512i __A)
90{
Logan Chien55afb0a2018-10-15 10:42:14 +080091 return (__m512i)__builtin_ia32_selectd_512((__mmask16)__U,
92 (__v16si)_mm512_lzcnt_epi32(__A),
93 (__v16si)__W);
Logan Chien2833ffb2018-10-09 10:03:24 +080094}
95
96static __inline__ __m512i __DEFAULT_FN_ATTRS
97_mm512_maskz_lzcnt_epi32 (__mmask16 __U, __m512i __A)
98{
Logan Chien55afb0a2018-10-15 10:42:14 +080099 return (__m512i)__builtin_ia32_selectd_512((__mmask16)__U,
100 (__v16si)_mm512_lzcnt_epi32(__A),
101 (__v16si)_mm512_setzero_si512());
Logan Chien2833ffb2018-10-09 10:03:24 +0800102}
103
104static __inline__ __m512i __DEFAULT_FN_ATTRS
105_mm512_lzcnt_epi64 (__m512i __A)
106{
Logan Chien55afb0a2018-10-15 10:42:14 +0800107 return (__m512i) __builtin_ia32_vplzcntq_512 ((__v8di) __A);
Logan Chien2833ffb2018-10-09 10:03:24 +0800108}
109
110static __inline__ __m512i __DEFAULT_FN_ATTRS
111_mm512_mask_lzcnt_epi64 (__m512i __W, __mmask8 __U, __m512i __A)
112{
Logan Chien55afb0a2018-10-15 10:42:14 +0800113 return (__m512i)__builtin_ia32_selectq_512((__mmask8)__U,
114 (__v8di)_mm512_lzcnt_epi64(__A),
115 (__v8di)__W);
Logan Chien2833ffb2018-10-09 10:03:24 +0800116}
117
118static __inline__ __m512i __DEFAULT_FN_ATTRS
119_mm512_maskz_lzcnt_epi64 (__mmask8 __U, __m512i __A)
120{
Logan Chien55afb0a2018-10-15 10:42:14 +0800121 return (__m512i)__builtin_ia32_selectq_512((__mmask8)__U,
122 (__v8di)_mm512_lzcnt_epi64(__A),
123 (__v8di)_mm512_setzero_si512());
Logan Chien2833ffb2018-10-09 10:03:24 +0800124}
125
126static __inline__ __m512i __DEFAULT_FN_ATTRS
127_mm512_broadcastmb_epi64 (__mmask8 __A)
128{
Logan Chien55afb0a2018-10-15 10:42:14 +0800129 return (__m512i) _mm512_set1_epi64((long long) __A);
Logan Chien2833ffb2018-10-09 10:03:24 +0800130}
131
132static __inline__ __m512i __DEFAULT_FN_ATTRS
133_mm512_broadcastmw_epi32 (__mmask16 __A)
134{
Logan Chien55afb0a2018-10-15 10:42:14 +0800135 return (__m512i) _mm512_set1_epi32((int) __A);
136
Logan Chien2833ffb2018-10-09 10:03:24 +0800137}
138
139#undef __DEFAULT_FN_ATTRS
140
141#endif