1 /*===------------- avx512vlvnniintrin.h - VNNI 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 <avx512vlvnniintrin.h> directly; include <immintrin.h> instead."
28 #ifndef __AVX512VLVNNIINTRIN_H
29 #define __AVX512VLVNNIINTRIN_H
31 /* Define the default attributes for the functions in this file. */
32 #define __DEFAULT_FN_ATTRS __attribute__((__always_inline__, __nodebug__, __target__("avx512vl,avx512vnni")))
35 static __inline__ __m256i __DEFAULT_FN_ATTRS
36 _mm256_mask_dpbusd_epi32(__m256i __S, __mmask8 __U, __m256i __A, __m256i __B)
38 return (__m256i) __builtin_ia32_vpdpbusd256_mask ((__v8si) __S,
44 static __inline__ __m256i __DEFAULT_FN_ATTRS
45 _mm256_maskz_dpbusd_epi32(__mmask8 __U, __m256i __S, __m256i __A, __m256i __B)
47 return (__m256i) __builtin_ia32_vpdpbusd256_maskz ((__v8si) __S,
53 static __inline__ __m256i __DEFAULT_FN_ATTRS
54 _mm256_dpbusd_epi32(__m256i __S, __m256i __A, __m256i __B)
56 return (__m256i) __builtin_ia32_vpdpbusd256_mask ((__v8si) __S,
62 static __inline__ __m256i __DEFAULT_FN_ATTRS
63 _mm256_mask_dpbusds_epi32(__m256i __S, __mmask8 __U, __m256i __A, __m256i __B)
65 return (__m256i) __builtin_ia32_vpdpbusds256_mask ((__v8si) __S,
71 static __inline__ __m256i __DEFAULT_FN_ATTRS
72 _mm256_maskz_dpbusds_epi32(__mmask8 __U, __m256i __S, __m256i __A, __m256i __B)
74 return (__m256i) __builtin_ia32_vpdpbusds256_maskz ((__v8si) __S,
80 static __inline__ __m256i __DEFAULT_FN_ATTRS
81 _mm256_dpbusds_epi32(__m256i __S, __m256i __A, __m256i __B)
83 return (__m256i) __builtin_ia32_vpdpbusds256_mask ((__v8si) __S,
89 static __inline__ __m256i __DEFAULT_FN_ATTRS
90 _mm256_mask_dpwssd_epi32(__m256i __S, __mmask8 __U, __m256i __A, __m256i __B)
92 return (__m256i) __builtin_ia32_vpdpwssd256_mask ((__v8si) __S,
98 static __inline__ __m256i __DEFAULT_FN_ATTRS
99 _mm256_maskz_dpwssd_epi32(__mmask8 __U, __m256i __S, __m256i __A, __m256i __B)
101 return (__m256i) __builtin_ia32_vpdpwssd256_maskz ((__v8si) __S,
107 static __inline__ __m256i __DEFAULT_FN_ATTRS
108 _mm256_dpwssd_epi32(__m256i __S, __m256i __A, __m256i __B)
110 return (__m256i) __builtin_ia32_vpdpwssd256_mask ((__v8si) __S,
116 static __inline__ __m256i __DEFAULT_FN_ATTRS
117 _mm256_mask_dpwssds_epi32(__m256i __S, __mmask8 __U, __m256i __A, __m256i __B)
119 return (__m256i) __builtin_ia32_vpdpwssds256_mask ((__v8si) __S,
125 static __inline__ __m256i __DEFAULT_FN_ATTRS
126 _mm256_maskz_dpwssds_epi32(__mmask8 __U, __m256i __S, __m256i __A, __m256i __B)
128 return (__m256i) __builtin_ia32_vpdpwssds256_maskz ((__v8si) __S,
134 static __inline__ __m256i __DEFAULT_FN_ATTRS
135 _mm256_dpwssds_epi32(__m256i __S, __m256i __A, __m256i __B)
137 return (__m256i) __builtin_ia32_vpdpwssds256_mask ((__v8si) __S,
143 static __inline__ __m128i __DEFAULT_FN_ATTRS
144 _mm128_mask_dpbusd_epi32(__m128i __S, __mmask8 __U, __m128i __A, __m128i __B)
146 return (__m128i) __builtin_ia32_vpdpbusd128_mask ((__v4si) __S,
152 static __inline__ __m128i __DEFAULT_FN_ATTRS
153 _mm128_maskz_dpbusd_epi32(__mmask8 __U, __m128i __S, __m128i __A, __m128i __B)
155 return (__m128i) __builtin_ia32_vpdpbusd128_maskz ((__v4si) __S,
161 static __inline__ __m128i __DEFAULT_FN_ATTRS
162 _mm128_dpbusd_epi32(__m128i __S, __m128i __A, __m128i __B)
164 return (__m128i) __builtin_ia32_vpdpbusd128_mask ((__v4si) __S,
170 static __inline__ __m128i __DEFAULT_FN_ATTRS
171 _mm128_mask_dpbusds_epi32(__m128i __S, __mmask8 __U, __m128i __A, __m128i __B)
173 return (__m128i) __builtin_ia32_vpdpbusds128_mask ((__v4si) __S,
179 static __inline__ __m128i __DEFAULT_FN_ATTRS
180 _mm128_maskz_dpbusds_epi32(__mmask8 __U, __m128i __S, __m128i __A, __m128i __B)
182 return (__m128i) __builtin_ia32_vpdpbusds128_maskz ((__v4si) __S,
188 static __inline__ __m128i __DEFAULT_FN_ATTRS
189 _mm128_dpbusds_epi32(__m128i __S, __m128i __A, __m128i __B)
191 return (__m128i) __builtin_ia32_vpdpbusds128_mask ((__v4si) __S,
197 static __inline__ __m128i __DEFAULT_FN_ATTRS
198 _mm128_mask_dpwssd_epi32(__m128i __S, __mmask8 __U, __m128i __A, __m128i __B)
200 return (__m128i) __builtin_ia32_vpdpwssd128_mask ((__v4si) __S,
206 static __inline__ __m128i __DEFAULT_FN_ATTRS
207 _mm128_maskz_dpwssd_epi32(__mmask8 __U, __m128i __S, __m128i __A, __m128i __B)
209 return (__m128i) __builtin_ia32_vpdpwssd128_maskz ((__v4si) __S,
215 static __inline__ __m128i __DEFAULT_FN_ATTRS
216 _mm128_dpwssd_epi32(__m128i __S, __m128i __A, __m128i __B)
218 return (__m128i) __builtin_ia32_vpdpwssd128_mask ((__v4si) __S,
224 static __inline__ __m128i __DEFAULT_FN_ATTRS
225 _mm128_mask_dpwssds_epi32(__m128i __S, __mmask8 __U, __m128i __A, __m128i __B)
227 return (__m128i) __builtin_ia32_vpdpwssds128_mask ((__v4si) __S,
233 static __inline__ __m128i __DEFAULT_FN_ATTRS
234 _mm128_maskz_dpwssds_epi32(__mmask8 __U, __m128i __S, __m128i __A, __m128i __B)
236 return (__m128i) __builtin_ia32_vpdpwssds128_maskz ((__v4si) __S,
242 static __inline__ __m128i __DEFAULT_FN_ATTRS
243 _mm128_dpwssds_epi32(__m128i __S, __m128i __A, __m128i __B)
245 return (__m128i) __builtin_ia32_vpdpwssds128_mask ((__v4si) __S,
252 #undef __DEFAULT_FN_ATTRS