master
1/*===-------- avx10_2minmaxintrin.h - AVX10_2MINMAX intrinsics -------------===
2 *
3 * Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 * See https://llvm.org/LICENSE.txt for license information.
5 * SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6 *
7 *===-----------------------------------------------------------------------===
8 */
9#ifndef __IMMINTRIN_H
10#error \
11 "Never use <avx10_2minmaxintrin.h> directly; include <immintrin.h> instead."
12#endif // __IMMINTRIN_H
13
14#ifndef __AVX10_2MINMAXINTRIN_H
15#define __AVX10_2MINMAXINTRIN_H
16
17#define _mm_minmax_pbh(A, B, C) \
18 ((__m128bh)__builtin_ia32_vminmaxbf16128((__m128bh)(__v8bf)(A), \
19 (__m128bh)(__v8bf)(B), (int)(C)))
20
21#define _mm_mask_minmax_pbh(W, U, A, B, C) \
22 ((__m128bh)__builtin_ia32_selectpbf_128( \
23 (__mmask8)(U), \
24 (__v8bf)_mm_minmax_pbh((__m128bh)(__v8bf)(A), (__m128bh)(__v8bf)(B), \
25 (int)(C)), \
26 (__v8bf)(W)))
27
28#define _mm_maskz_minmax_pbh(U, A, B, C) \
29 ((__m128bh)__builtin_ia32_selectpbf_128( \
30 (__mmask8)(U), \
31 (__v8bf)_mm_minmax_pbh((__m128bh)(__v8bf)(A), (__m128bh)(__v8bf)(B), \
32 (int)(C)), \
33 (__v8bf) __builtin_bit_cast(__m128bh, _mm_setzero_ps())))
34
35#define _mm256_minmax_pbh(A, B, C) \
36 ((__m256bh)__builtin_ia32_vminmaxbf16256((__m256bh)(__v16bf)(A), \
37 (__m256bh)(__v16bf)(B), (int)(C)))
38
39#define _mm256_mask_minmax_pbh(W, U, A, B, C) \
40 ((__m256bh)__builtin_ia32_selectpbf_256( \
41 (__mmask16)(U), \
42 (__v16bf)_mm256_minmax_pbh((__m256bh)(__v16bf)(A), \
43 (__m256bh)(__v16bf)(B), (int)(C)), \
44 (__v16bf)(W)))
45
46#define _mm256_maskz_minmax_pbh(U, A, B, C) \
47 ((__m256bh)__builtin_ia32_selectpbf_256( \
48 (__mmask16)(U), \
49 (__v16bf)_mm256_minmax_pbh((__m256bh)(__v16bf)(A), \
50 (__m256bh)(__v16bf)(B), (int)(C)), \
51 (__v16bf) __builtin_bit_cast(__m256bh, _mm256_setzero_ps())))
52
53#define _mm_minmax_pd(A, B, C) \
54 ((__m128d)__builtin_ia32_vminmaxpd128_mask( \
55 (__v2df)(__m128d)(A), (__v2df)(__m128d)(B), (int)(C), \
56 (__v2df)_mm_setzero_pd(), (__mmask8)-1))
57
58#define _mm_mask_minmax_pd(W, U, A, B, C) \
59 ((__m128d)__builtin_ia32_vminmaxpd128_mask( \
60 (__v2df)(__m128d)(A), (__v2df)(__m128d)(B), (int)(C), \
61 (__v2df)(__m128d)(W), (__mmask8)(U)))
62
63#define _mm_maskz_minmax_pd(U, A, B, C) \
64 ((__m128d)__builtin_ia32_vminmaxpd128_mask( \
65 (__v2df)(__m128d)(A), (__v2df)(__m128d)(B), (int)(C), \
66 (__v2df)_mm_setzero_pd(), (__mmask8)(U)))
67
68#define _mm256_minmax_pd(A, B, C) \
69 ((__m256d)__builtin_ia32_vminmaxpd256_mask( \
70 (__v4df)(__m256d)(A), (__v4df)(__m256d)(B), (int)(C), \
71 (__v4df)_mm256_setzero_pd(), (__mmask8)-1))
72
73#define _mm256_mask_minmax_pd(W, U, A, B, C) \
74 ((__m256d)__builtin_ia32_vminmaxpd256_mask( \
75 (__v4df)(__m256d)(A), (__v4df)(__m256d)(B), (int)(C), \
76 (__v4df)(__m256d)(W), (__mmask8)(U)))
77
78#define _mm256_maskz_minmax_pd(U, A, B, C) \
79 ((__m256d)__builtin_ia32_vminmaxpd256_mask( \
80 (__v4df)(__m256d)(A), (__v4df)(__m256d)(B), (int)(C), \
81 (__v4df)_mm256_setzero_pd(), (__mmask8)(U)))
82
83#define _mm_minmax_ph(A, B, C) \
84 ((__m128h)__builtin_ia32_vminmaxph128_mask( \
85 (__v8hf)(__m128h)(A), (__v8hf)(__m128h)(B), (int)(C), \
86 (__v8hf)_mm_setzero_ph(), (__mmask8)-1))
87
88#define _mm_mask_minmax_ph(W, U, A, B, C) \
89 ((__m128h)__builtin_ia32_vminmaxph128_mask( \
90 (__v8hf)(__m128h)(A), (__v8hf)(__m128h)(B), (int)(C), \
91 (__v8hf)(__m128h)(W), (__mmask16)-1))
92
93#define _mm_maskz_minmax_ph(U, A, B, C) \
94 ((__m128h)__builtin_ia32_vminmaxph128_mask( \
95 (__v8hf)(__m128h)(A), (__v8hf)(__m128h)(B), (int)(C), \
96 (__v8hf)_mm_setzero_ph(), (__mmask8)(U)))
97
98#define _mm256_minmax_ph(A, B, C) \
99 ((__m256h)__builtin_ia32_vminmaxph256_mask( \
100 (__v16hf)(__m256h)(A), (__v16hf)(__m256h)(B), (int)(C), \
101 (__v16hf)_mm256_setzero_ph(), (__mmask16)-1))
102
103#define _mm256_mask_minmax_ph(W, U, A, B, C) \
104 ((__m256h)__builtin_ia32_vminmaxph256_mask( \
105 (__v16hf)(__m256h)(A), (__v16hf)(__m256h)(B), (int)(C), \
106 (__v16hf)(__m256h)(W), (__mmask16)(U)))
107
108#define _mm256_maskz_minmax_ph(U, A, B, C) \
109 ((__m256h)__builtin_ia32_vminmaxph256_mask( \
110 (__v16hf)(__m256h)(A), (__v16hf)(__m256h)(B), (int)(C), \
111 (__v16hf)_mm256_setzero_ph(), (__mmask16)(U)))
112
113#define _mm_minmax_ps(A, B, C) \
114 ((__m128)__builtin_ia32_vminmaxps128_mask( \
115 (__v4sf)(__m128)(A), (__v4sf)(__m128)(B), (int)(C), \
116 (__v4sf)_mm_setzero_ps(), (__mmask8)-1))
117
118#define _mm_mask_minmax_ps(W, U, A, B, C) \
119 ((__m128)__builtin_ia32_vminmaxps128_mask( \
120 (__v4sf)(__m128)(A), (__v4sf)(__m128)(B), (int)(C), (__v4sf)(__m128)(W), \
121 (__mmask8)(U)))
122
123#define _mm_maskz_minmax_ps(U, A, B, C) \
124 ((__m128)__builtin_ia32_vminmaxps128_mask( \
125 (__v4sf)(__m128)(A), (__v4sf)(__m128)(B), (int)(C), \
126 (__v4sf)_mm_setzero_ps(), (__mmask8)(U)))
127
128#define _mm256_minmax_ps(A, B, C) \
129 ((__m256)__builtin_ia32_vminmaxps256_mask( \
130 (__v8sf)(__m256)(A), (__v8sf)(__m256)(B), (int)(C), \
131 (__v8sf)_mm256_setzero_ps(), (__mmask8)-1))
132
133#define _mm256_mask_minmax_ps(W, U, A, B, C) \
134 ((__m256)__builtin_ia32_vminmaxps256_mask( \
135 (__v8sf)(__m256)(A), (__v8sf)(__m256)(B), (int)(C), (__v8sf)(__m256)(W), \
136 (__mmask8)(U)))
137
138#define _mm256_maskz_minmax_ps(U, A, B, C) \
139 ((__m256)__builtin_ia32_vminmaxps256_mask( \
140 (__v8sf)(__m256)(A), (__v8sf)(__m256)(B), (int)(C), \
141 (__v8sf)_mm256_setzero_ps(), (__mmask8)(U)))
142
143#define _mm_minmax_sd(A, B, C) \
144 ((__m128d)__builtin_ia32_vminmaxsd_round_mask( \
145 (__v2df)(__m128d)(A), (__v2df)(__m128d)(B), (int)(C), \
146 (__v2df)_mm_undefined_pd(), (__mmask8)-1, _MM_FROUND_CUR_DIRECTION))
147
148#define _mm_mask_minmax_sd(W, U, A, B, C) \
149 ((__m128d)__builtin_ia32_vminmaxsd_round_mask( \
150 (__v2df)(__m128d)(A), (__v2df)(__m128d)(B), (int)(C), \
151 (__v2df)(__m128d)(W), (__mmask8)(U), _MM_FROUND_CUR_DIRECTION))
152
153#define _mm_maskz_minmax_sd(U, A, B, C) \
154 ((__m128d)__builtin_ia32_vminmaxsd_round_mask( \
155 (__v2df)(__m128d)(A), (__v2df)(__m128d)(B), (int)(C), \
156 (__v2df)_mm_setzero_pd(), (__mmask8)(U), _MM_FROUND_CUR_DIRECTION))
157
158#define _mm_minmax_round_sd(A, B, C, R) \
159 ((__m128d)__builtin_ia32_vminmaxsd_round_mask( \
160 (__v2df)(__m128d)(A), (__v2df)(__m128d)(B), (int)(C), \
161 (__v2df)_mm_undefined_pd(), (__mmask8)-1, (int)(R)))
162
163#define _mm_mask_minmax_round_sd(W, U, A, B, C, R) \
164 ((__m128d)__builtin_ia32_vminmaxsd_round_mask( \
165 (__v2df)(__m128d)(A), (__v2df)(__m128d)(B), (int)(C), \
166 (__v2df)(__m128d)(W), (__mmask8)(U), (int)(R)))
167
168#define _mm_maskz_minmax_round_sd(U, A, B, C, R) \
169 ((__m128d)__builtin_ia32_vminmaxsd_round_mask( \
170 (__v2df)(__m128d)(A), (__v2df)(__m128d)(B), (int)(C), \
171 (__v2df)_mm_setzero_pd(), (__mmask8)(U), (int)(R)))
172
173#define _mm_minmax_sh(A, B, C) \
174 ((__m128h)__builtin_ia32_vminmaxsh_round_mask( \
175 (__v8hf)(__m128h)(A), (__v8hf)(__m128h)(B), (int)(C), \
176 (__v8hf)_mm_undefined_ph(), (__mmask8)-1, _MM_FROUND_CUR_DIRECTION))
177
178#define _mm_mask_minmax_sh(W, U, A, B, C) \
179 ((__m128h)__builtin_ia32_vminmaxsh_round_mask( \
180 (__v8hf)(__m128h)(A), (__v8hf)(__m128h)(B), (int)(C), \
181 (__v8hf)(__m128h)(W), (__mmask8)(U), _MM_FROUND_CUR_DIRECTION))
182
183#define _mm_maskz_minmax_sh(U, A, B, C) \
184 ((__m128h)__builtin_ia32_vminmaxsh_round_mask( \
185 (__v8hf)(__m128h)(A), (__v8hf)(__m128h)(B), (int)(C), \
186 (__v8hf)_mm_setzero_ph(), (__mmask8)(U), _MM_FROUND_CUR_DIRECTION))
187
188#define _mm_minmax_round_sh(A, B, C, R) \
189 ((__m128h)__builtin_ia32_vminmaxsh_round_mask( \
190 (__v8hf)(__m128h)(A), (__v8hf)(__m128h)(B), (int)(C), \
191 (__v8hf)_mm_undefined_ph(), (__mmask8)-1, (int)(R)))
192
193#define _mm_mask_minmax_round_sh(W, U, A, B, C, R) \
194 ((__m128h)__builtin_ia32_vminmaxsh_round_mask( \
195 (__v8hf)(__m128h)(A), (__v8hf)(__m128h)(B), (int)(C), \
196 (__v8hf)(__m128h)(W), (__mmask8)(U), (int)(R)))
197
198#define _mm_maskz_minmax_round_sh(U, A, B, C, R) \
199 ((__m128h)__builtin_ia32_vminmaxsh_round_mask( \
200 (__v8hf)(__m128h)(A), (__v8hf)(__m128h)(B), (int)(C), \
201 (__v8hf)_mm_setzero_ph(), (__mmask8)(U), (int)(R)))
202
203#define _mm_minmax_ss(A, B, C) \
204 ((__m128)__builtin_ia32_vminmaxss_round_mask( \
205 (__v4sf)(__m128)(A), (__v4sf)(__m128)(B), (int)(C), \
206 (__v4sf)_mm_undefined_ps(), (__mmask8)-1, _MM_FROUND_CUR_DIRECTION))
207
208#define _mm_mask_minmax_ss(W, U, A, B, C) \
209 ((__m128)__builtin_ia32_vminmaxss_round_mask( \
210 (__v4sf)(__m128)(A), (__v4sf)(__m128)(B), (int)(C), (__v4sf)(W), \
211 (__mmask8)(U), _MM_FROUND_CUR_DIRECTION))
212
213#define _mm_maskz_minmax_ss(U, A, B, C) \
214 ((__m128)__builtin_ia32_vminmaxss_round_mask( \
215 (__v4sf)(__m128)(A), (__v4sf)(__m128)(B), (int)(C), \
216 (__v4sf)_mm_setzero_ps(), (__mmask8)(U), _MM_FROUND_CUR_DIRECTION))
217
218#define _mm_minmax_round_ss(A, B, C, R) \
219 ((__m128)__builtin_ia32_vminmaxss_round_mask( \
220 (__v4sf)(__m128)(A), (__v4sf)(__m128)(B), (int)(C), \
221 (__v4sf)_mm_undefined_ps(), (__mmask8)-1, (int)(R)))
222
223#define _mm_mask_minmax_round_ss(W, U, A, B, C, R) \
224 ((__m128)__builtin_ia32_vminmaxss_round_mask( \
225 (__v4sf)(__m128)(A), (__v4sf)(__m128)(B), (int)(C), (__v4sf)(W), \
226 (__mmask8)(U), (int)(R)))
227
228#define _mm_maskz_minmax_round_ss(U, A, B, C, R) \
229 ((__m128)__builtin_ia32_vminmaxss_round_mask( \
230 (__v4sf)(__m128)(A), (__v4sf)(__m128)(B), (int)(C), \
231 (__v4sf)_mm_setzero_ps(), (__mmask8)(U), (int)(R)))
232#endif // __AVX10_2MINMAXINTRIN_H