1 /*===------------- avx512vbmivlintrin.h - VBMI intrinsics ------------------===
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:
11 * The above copyright notice and this permission notice shall be included in
12 * all copies or substantial portions of the Software.
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
22 *===-----------------------------------------------------------------------===
25 #error "Never use <avx512vbmivlintrin.h> directly; include <immintrin.h> instead."
28 #ifndef __VBMIVLINTRIN_H
29 #define __VBMIVLINTRIN_H
31 /* Define the default attributes for the functions in this file. */
32 #define __DEFAULT_FN_ATTRS128 __attribute__((__always_inline__, __nodebug__, __target__("avx512vbmi,avx512vl"), __min_vector_width__(128)))
33 #define __DEFAULT_FN_ATTRS256 __attribute__((__always_inline__, __nodebug__, __target__("avx512vbmi,avx512vl"), __min_vector_width__(256)))
36 static __inline__ __m128i __DEFAULT_FN_ATTRS128
37 _mm_permutex2var_epi8(__m128i __A, __m128i __I, __m128i __B)
39 return (__m128i)__builtin_ia32_vpermi2varqi128((__v16qi)__A,
44 static __inline__ __m128i __DEFAULT_FN_ATTRS128
45 _mm_mask_permutex2var_epi8(__m128i __A, __mmask16 __U, __m128i __I,
48 return (__m128i)__builtin_ia32_selectb_128(__U,
49 (__v16qi)_mm_permutex2var_epi8(__A, __I, __B),
53 static __inline__ __m128i __DEFAULT_FN_ATTRS128
54 _mm_mask2_permutex2var_epi8(__m128i __A, __m128i __I, __mmask16 __U,
57 return (__m128i)__builtin_ia32_selectb_128(__U,
58 (__v16qi)_mm_permutex2var_epi8(__A, __I, __B),
62 static __inline__ __m128i __DEFAULT_FN_ATTRS128
63 _mm_maskz_permutex2var_epi8(__mmask16 __U, __m128i __A, __m128i __I,
66 return (__m128i)__builtin_ia32_selectb_128(__U,
67 (__v16qi)_mm_permutex2var_epi8(__A, __I, __B),
68 (__v16qi)_mm_setzero_si128());
71 static __inline__ __m256i __DEFAULT_FN_ATTRS256
72 _mm256_permutex2var_epi8(__m256i __A, __m256i __I, __m256i __B)
74 return (__m256i)__builtin_ia32_vpermi2varqi256((__v32qi)__A, (__v32qi)__I,
78 static __inline__ __m256i __DEFAULT_FN_ATTRS256
79 _mm256_mask_permutex2var_epi8(__m256i __A, __mmask32 __U, __m256i __I,
82 return (__m256i)__builtin_ia32_selectb_256(__U,
83 (__v32qi)_mm256_permutex2var_epi8(__A, __I, __B),
87 static __inline__ __m256i __DEFAULT_FN_ATTRS256
88 _mm256_mask2_permutex2var_epi8(__m256i __A, __m256i __I, __mmask32 __U,
91 return (__m256i)__builtin_ia32_selectb_256(__U,
92 (__v32qi)_mm256_permutex2var_epi8(__A, __I, __B),
96 static __inline__ __m256i __DEFAULT_FN_ATTRS256
97 _mm256_maskz_permutex2var_epi8(__mmask32 __U, __m256i __A, __m256i __I,
100 return (__m256i)__builtin_ia32_selectb_256(__U,
101 (__v32qi)_mm256_permutex2var_epi8(__A, __I, __B),
102 (__v32qi)_mm256_setzero_si256());
105 static __inline__ __m128i __DEFAULT_FN_ATTRS128
106 _mm_permutexvar_epi8 (__m128i __A, __m128i __B)
108 return (__m128i)__builtin_ia32_permvarqi128((__v16qi)__B, (__v16qi)__A);
111 static __inline__ __m128i __DEFAULT_FN_ATTRS128
112 _mm_maskz_permutexvar_epi8 (__mmask16 __M, __m128i __A, __m128i __B)
114 return (__m128i)__builtin_ia32_selectb_128((__mmask16)__M,
115 (__v16qi)_mm_permutexvar_epi8(__A, __B),
116 (__v16qi)_mm_setzero_si128());
119 static __inline__ __m128i __DEFAULT_FN_ATTRS128
120 _mm_mask_permutexvar_epi8 (__m128i __W, __mmask16 __M, __m128i __A,
123 return (__m128i)__builtin_ia32_selectb_128((__mmask16)__M,
124 (__v16qi)_mm_permutexvar_epi8(__A, __B),
128 static __inline__ __m256i __DEFAULT_FN_ATTRS256
129 _mm256_permutexvar_epi8 (__m256i __A, __m256i __B)
131 return (__m256i)__builtin_ia32_permvarqi256((__v32qi) __B, (__v32qi) __A);
134 static __inline__ __m256i __DEFAULT_FN_ATTRS256
135 _mm256_maskz_permutexvar_epi8 (__mmask32 __M, __m256i __A,
138 return (__m256i)__builtin_ia32_selectb_256((__mmask32)__M,
139 (__v32qi)_mm256_permutexvar_epi8(__A, __B),
140 (__v32qi)_mm256_setzero_si256());
143 static __inline__ __m256i __DEFAULT_FN_ATTRS256
144 _mm256_mask_permutexvar_epi8 (__m256i __W, __mmask32 __M, __m256i __A,
147 return (__m256i)__builtin_ia32_selectb_256((__mmask32)__M,
148 (__v32qi)_mm256_permutexvar_epi8(__A, __B),
152 static __inline__ __m128i __DEFAULT_FN_ATTRS128
153 _mm_multishift_epi64_epi8(__m128i __X, __m128i __Y)
155 return (__m128i)__builtin_ia32_vpmultishiftqb128((__v16qi)__X, (__v16qi)__Y);
158 static __inline__ __m128i __DEFAULT_FN_ATTRS128
159 _mm_mask_multishift_epi64_epi8(__m128i __W, __mmask16 __M, __m128i __X,
162 return (__m128i)__builtin_ia32_selectb_128((__mmask16)__M,
163 (__v16qi)_mm_multishift_epi64_epi8(__X, __Y),
167 static __inline__ __m128i __DEFAULT_FN_ATTRS128
168 _mm_maskz_multishift_epi64_epi8(__mmask16 __M, __m128i __X, __m128i __Y)
170 return (__m128i)__builtin_ia32_selectb_128((__mmask16)__M,
171 (__v16qi)_mm_multishift_epi64_epi8(__X, __Y),
172 (__v16qi)_mm_setzero_si128());
175 static __inline__ __m256i __DEFAULT_FN_ATTRS256
176 _mm256_multishift_epi64_epi8(__m256i __X, __m256i __Y)
178 return (__m256i)__builtin_ia32_vpmultishiftqb256((__v32qi)__X, (__v32qi)__Y);
181 static __inline__ __m256i __DEFAULT_FN_ATTRS256
182 _mm256_mask_multishift_epi64_epi8(__m256i __W, __mmask32 __M, __m256i __X,
185 return (__m256i)__builtin_ia32_selectb_256((__mmask32)__M,
186 (__v32qi)_mm256_multishift_epi64_epi8(__X, __Y),
190 static __inline__ __m256i __DEFAULT_FN_ATTRS256
191 _mm256_maskz_multishift_epi64_epi8(__mmask32 __M, __m256i __X, __m256i __Y)
193 return (__m256i)__builtin_ia32_selectb_256((__mmask32)__M,
194 (__v32qi)_mm256_multishift_epi64_epi8(__X, __Y),
195 (__v32qi)_mm256_setzero_si256());
199 #undef __DEFAULT_FN_ATTRS128
200 #undef __DEFAULT_FN_ATTRS256