blob: b6e93c285871319572efd1094b95f6c44eaed413 [file] [log] [blame]
Michael Zuckerman10d6f9a2016-03-09 11:26:45 +00001/*===------------- avx512vbmiintrin.h - VBMI intrinsics ------------------===
Michael Zuckerman9f338482016-03-07 17:04:11 +00002 *
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 <avx512vbmiintrin.h> directly; include <immintrin.h> instead."
26#endif
27
28#ifndef __VBMIINTRIN_H
29#define __VBMIINTRIN_H
30
31/* Define the default attributes for the functions in this file. */
Craig Topper74c10e32018-07-09 19:00:16 +000032#define __DEFAULT_FN_ATTRS __attribute__((__always_inline__, __nodebug__, __target__("avx512vbmi"), __min_vector_width__(512)))
Michael Zuckerman9f338482016-03-07 17:04:11 +000033
34
35static __inline__ __m512i __DEFAULT_FN_ATTRS
Craig Topper68a272d2018-05-29 03:26:38 +000036_mm512_permutex2var_epi8(__m512i __A, __m512i __I, __m512i __B)
Michael Zuckerman9f338482016-03-07 17:04:11 +000037{
Craig Topper68a272d2018-05-29 03:26:38 +000038 return (__m512i)__builtin_ia32_vpermi2varqi512((__v64qi)__A, (__v64qi)__I,
39 (__v64qi) __B);
Michael Zuckerman9f338482016-03-07 17:04:11 +000040}
41
42static __inline__ __m512i __DEFAULT_FN_ATTRS
Craig Topper68a272d2018-05-29 03:26:38 +000043_mm512_mask_permutex2var_epi8(__m512i __A, __mmask64 __U, __m512i __I,
44 __m512i __B)
Michael Zuckerman9f338482016-03-07 17:04:11 +000045{
Craig Topper68a272d2018-05-29 03:26:38 +000046 return (__m512i)__builtin_ia32_selectb_512(__U,
47 (__v64qi)_mm512_permutex2var_epi8(__A, __I, __B),
48 (__v64qi)__A);
Michael Zuckerman9f338482016-03-07 17:04:11 +000049}
50
51static __inline__ __m512i __DEFAULT_FN_ATTRS
Craig Topper68a272d2018-05-29 03:26:38 +000052_mm512_mask2_permutex2var_epi8(__m512i __A, __m512i __I, __mmask64 __U,
53 __m512i __B)
Michael Zuckerman9f338482016-03-07 17:04:11 +000054{
Craig Topper68a272d2018-05-29 03:26:38 +000055 return (__m512i)__builtin_ia32_selectb_512(__U,
56 (__v64qi)_mm512_permutex2var_epi8(__A, __I, __B),
57 (__v64qi)__I);
Michael Zuckerman9f338482016-03-07 17:04:11 +000058}
59
60static __inline__ __m512i __DEFAULT_FN_ATTRS
Craig Topper68a272d2018-05-29 03:26:38 +000061_mm512_maskz_permutex2var_epi8(__mmask64 __U, __m512i __A, __m512i __I,
62 __m512i __B)
Michael Zuckerman9f338482016-03-07 17:04:11 +000063{
Craig Topper68a272d2018-05-29 03:26:38 +000064 return (__m512i)__builtin_ia32_selectb_512(__U,
65 (__v64qi)_mm512_permutex2var_epi8(__A, __I, __B),
66 (__v64qi)_mm512_setzero_si512());
Michael Zuckerman9f338482016-03-07 17:04:11 +000067}
68
Michael Zuckerman8938e832016-04-25 05:32:35 +000069static __inline__ __m512i __DEFAULT_FN_ATTRS
70_mm512_permutexvar_epi8 (__m512i __A, __m512i __B)
71{
Craig Topper55b40672018-05-20 23:34:10 +000072 return (__m512i)__builtin_ia32_permvarqi512((__v64qi) __B, (__v64qi) __A);
Michael Zuckerman8938e832016-04-25 05:32:35 +000073}
74
75static __inline__ __m512i __DEFAULT_FN_ATTRS
76_mm512_maskz_permutexvar_epi8 (__mmask64 __M, __m512i __A,
77 __m512i __B)
78{
Craig Topper55b40672018-05-20 23:34:10 +000079 return (__m512i)__builtin_ia32_selectb_512((__mmask64)__M,
80 (__v64qi)_mm512_permutexvar_epi8(__A, __B),
81 (__v64qi)_mm512_setzero_si512());
Michael Zuckerman8938e832016-04-25 05:32:35 +000082}
83
84static __inline__ __m512i __DEFAULT_FN_ATTRS
85_mm512_mask_permutexvar_epi8 (__m512i __W, __mmask64 __M, __m512i __A,
86 __m512i __B)
87{
Craig Topper55b40672018-05-20 23:34:10 +000088 return (__m512i)__builtin_ia32_selectb_512((__mmask64)__M,
89 (__v64qi)_mm512_permutexvar_epi8(__A, __B),
90 (__v64qi)__W);
Michael Zuckerman8938e832016-04-25 05:32:35 +000091}
92
Michael Zuckermanf86eb712016-05-23 15:04:39 +000093static __inline__ __m512i __DEFAULT_FN_ATTRS
94_mm512_mask_multishift_epi64_epi8 (__m512i __W, __mmask64 __M, __m512i __X, __m512i __Y)
95{
96 return (__m512i) __builtin_ia32_vpmultishiftqb512_mask ((__v64qi) __X,
97 (__v64qi) __Y,
98 (__v64qi) __W,
99 (__mmask64) __M);
100}
101
102static __inline__ __m512i __DEFAULT_FN_ATTRS
103_mm512_maskz_multishift_epi64_epi8 (__mmask64 __M, __m512i __X, __m512i __Y)
104{
105 return (__m512i) __builtin_ia32_vpmultishiftqb512_mask ((__v64qi) __X,
106 (__v64qi) __Y,
Michael Zuckerman33786532016-07-12 12:42:27 +0000107 (__v64qi) _mm512_setzero_si512 (),
Michael Zuckermanf86eb712016-05-23 15:04:39 +0000108 (__mmask64) __M);
109}
110
111static __inline__ __m512i __DEFAULT_FN_ATTRS
112_mm512_multishift_epi64_epi8 (__m512i __X, __m512i __Y)
113{
114 return (__m512i) __builtin_ia32_vpmultishiftqb512_mask ((__v64qi) __X,
115 (__v64qi) __Y,
Michael Zuckerman33786532016-07-12 12:42:27 +0000116 (__v64qi) _mm512_undefined_epi32 (),
Michael Zuckermanf86eb712016-05-23 15:04:39 +0000117 (__mmask64) -1);
118}
119
120
Michael Zuckerman9f338482016-03-07 17:04:11 +0000121#undef __DEFAULT_FN_ATTRS
122
123#endif