| 1 |
/*===---- avx512erintrin.h - AVX512ER 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 "Never use <avx512erintrin.h> directly; include <immintrin.h> instead." |
| 11 |
#endif |
| 12 |
|
| 13 |
#ifndef __AVX512ERINTRIN_H |
| 14 |
#define __AVX512ERINTRIN_H |
| 15 |
|
| 16 |
/* exp2a23 */ |
| 17 |
#define _mm512_exp2a23_round_pd(A, R) \ |
| 18 |
((__m512d)__builtin_ia32_exp2pd_mask((__v8df)(__m512d)(A), \ |
| 19 |
(__v8df)_mm512_setzero_pd(), \ |
| 20 |
(__mmask8)-1, (int)(R))) |
| 21 |
|
| 22 |
#define _mm512_mask_exp2a23_round_pd(S, M, A, R) \ |
| 23 |
((__m512d)__builtin_ia32_exp2pd_mask((__v8df)(__m512d)(A), \ |
| 24 |
(__v8df)(__m512d)(S), (__mmask8)(M), \ |
| 25 |
(int)(R))) |
| 26 |
|
| 27 |
#define _mm512_maskz_exp2a23_round_pd(M, A, R) \ |
| 28 |
((__m512d)__builtin_ia32_exp2pd_mask((__v8df)(__m512d)(A), \ |
| 29 |
(__v8df)_mm512_setzero_pd(), \ |
| 30 |
(__mmask8)(M), (int)(R))) |
| 31 |
|
| 32 |
#define _mm512_exp2a23_pd(A) \ |
| 33 |
_mm512_exp2a23_round_pd((A), _MM_FROUND_CUR_DIRECTION) |
| 34 |
|
| 35 |
#define _mm512_mask_exp2a23_pd(S, M, A) \ |
| 36 |
_mm512_mask_exp2a23_round_pd((S), (M), (A), _MM_FROUND_CUR_DIRECTION) |
| 37 |
|
| 38 |
#define _mm512_maskz_exp2a23_pd(M, A) \ |
| 39 |
_mm512_maskz_exp2a23_round_pd((M), (A), _MM_FROUND_CUR_DIRECTION) |
| 40 |
|
| 41 |
#define _mm512_exp2a23_round_ps(A, R) \ |
| 42 |
((__m512)__builtin_ia32_exp2ps_mask((__v16sf)(__m512)(A), \ |
| 43 |
(__v16sf)_mm512_setzero_ps(), \ |
| 44 |
(__mmask16)-1, (int)(R))) |
| 45 |
|
| 46 |
#define _mm512_mask_exp2a23_round_ps(S, M, A, R) \ |
| 47 |
((__m512)__builtin_ia32_exp2ps_mask((__v16sf)(__m512)(A), \ |
| 48 |
(__v16sf)(__m512)(S), (__mmask16)(M), \ |
| 49 |
(int)(R))) |
| 50 |
|
| 51 |
#define _mm512_maskz_exp2a23_round_ps(M, A, R) \ |
| 52 |
((__m512)__builtin_ia32_exp2ps_mask((__v16sf)(__m512)(A), \ |
| 53 |
(__v16sf)_mm512_setzero_ps(), \ |
| 54 |
(__mmask16)(M), (int)(R))) |
| 55 |
|
| 56 |
#define _mm512_exp2a23_ps(A) \ |
| 57 |
_mm512_exp2a23_round_ps((A), _MM_FROUND_CUR_DIRECTION) |
| 58 |
|
| 59 |
#define _mm512_mask_exp2a23_ps(S, M, A) \ |
| 60 |
_mm512_mask_exp2a23_round_ps((S), (M), (A), _MM_FROUND_CUR_DIRECTION) |
| 61 |
|
| 62 |
#define _mm512_maskz_exp2a23_ps(M, A) \ |
| 63 |
_mm512_maskz_exp2a23_round_ps((M), (A), _MM_FROUND_CUR_DIRECTION) |
| 64 |
|
| 65 |
/* rsqrt28 */ |
| 66 |
#define _mm512_rsqrt28_round_pd(A, R) \ |
| 67 |
((__m512d)__builtin_ia32_rsqrt28pd_mask((__v8df)(__m512d)(A), \ |
| 68 |
(__v8df)_mm512_setzero_pd(), \ |
| 69 |
(__mmask8)-1, (int)(R))) |
| 70 |
|
| 71 |
#define _mm512_mask_rsqrt28_round_pd(S, M, A, R) \ |
| 72 |
((__m512d)__builtin_ia32_rsqrt28pd_mask((__v8df)(__m512d)(A), \ |
| 73 |
(__v8df)(__m512d)(S), (__mmask8)(M), \ |
| 74 |
(int)(R))) |
| 75 |
|
| 76 |
#define _mm512_maskz_rsqrt28_round_pd(M, A, R) \ |
| 77 |
((__m512d)__builtin_ia32_rsqrt28pd_mask((__v8df)(__m512d)(A), \ |
| 78 |
(__v8df)_mm512_setzero_pd(), \ |
| 79 |
(__mmask8)(M), (int)(R))) |
| 80 |
|
| 81 |
#define _mm512_rsqrt28_pd(A) \ |
| 82 |
_mm512_rsqrt28_round_pd((A), _MM_FROUND_CUR_DIRECTION) |
| 83 |
|
| 84 |
#define _mm512_mask_rsqrt28_pd(S, M, A) \ |
| 85 |
_mm512_mask_rsqrt28_round_pd((S), (M), (A), _MM_FROUND_CUR_DIRECTION) |
| 86 |
|
| 87 |
#define _mm512_maskz_rsqrt28_pd(M, A) \ |
| 88 |
_mm512_maskz_rsqrt28_round_pd((M), (A), _MM_FROUND_CUR_DIRECTION) |
| 89 |
|
| 90 |
#define _mm512_rsqrt28_round_ps(A, R) \ |
| 91 |
((__m512)__builtin_ia32_rsqrt28ps_mask((__v16sf)(__m512)(A), \ |
| 92 |
(__v16sf)_mm512_setzero_ps(), \ |
| 93 |
(__mmask16)-1, (int)(R))) |
| 94 |
|
| 95 |
#define _mm512_mask_rsqrt28_round_ps(S, M, A, R) \ |
| 96 |
((__m512)__builtin_ia32_rsqrt28ps_mask((__v16sf)(__m512)(A), \ |
| 97 |
(__v16sf)(__m512)(S), (__mmask16)(M), \ |
| 98 |
(int)(R))) |
| 99 |
|
| 100 |
#define _mm512_maskz_rsqrt28_round_ps(M, A, R) \ |
| 101 |
((__m512)__builtin_ia32_rsqrt28ps_mask((__v16sf)(__m512)(A), \ |
| 102 |
(__v16sf)_mm512_setzero_ps(), \ |
| 103 |
(__mmask16)(M), (int)(R))) |
| 104 |
|
| 105 |
#define _mm512_rsqrt28_ps(A) \ |
| 106 |
_mm512_rsqrt28_round_ps((A), _MM_FROUND_CUR_DIRECTION) |
| 107 |
|
| 108 |
#define _mm512_mask_rsqrt28_ps(S, M, A) \ |
| 109 |
_mm512_mask_rsqrt28_round_ps((S), (M), A, _MM_FROUND_CUR_DIRECTION) |
| 110 |
|
| 111 |
#define _mm512_maskz_rsqrt28_ps(M, A) \ |
| 112 |
_mm512_maskz_rsqrt28_round_ps((M), (A), _MM_FROUND_CUR_DIRECTION) |
| 113 |
|
| 114 |
#define _mm_rsqrt28_round_ss(A, B, R) \ |
| 115 |
((__m128)__builtin_ia32_rsqrt28ss_round_mask((__v4sf)(__m128)(A), \ |
| 116 |
(__v4sf)(__m128)(B), \ |
| 117 |
(__v4sf)_mm_setzero_ps(), \ |
| 118 |
(__mmask8)-1, (int)(R))) |
| 119 |
|
| 120 |
#define _mm_mask_rsqrt28_round_ss(S, M, A, B, R) \ |
| 121 |
((__m128)__builtin_ia32_rsqrt28ss_round_mask((__v4sf)(__m128)(A), \ |
| 122 |
(__v4sf)(__m128)(B), \ |
| 123 |
(__v4sf)(__m128)(S), \ |
| 124 |
(__mmask8)(M), (int)(R))) |
| 125 |
|
| 126 |
#define _mm_maskz_rsqrt28_round_ss(M, A, B, R) \ |
| 127 |
((__m128)__builtin_ia32_rsqrt28ss_round_mask((__v4sf)(__m128)(A), \ |
| 128 |
(__v4sf)(__m128)(B), \ |
| 129 |
(__v4sf)_mm_setzero_ps(), \ |
| 130 |
(__mmask8)(M), (int)(R))) |
| 131 |
|
| 132 |
#define _mm_rsqrt28_ss(A, B) \ |
| 133 |
_mm_rsqrt28_round_ss((A), (B), _MM_FROUND_CUR_DIRECTION) |
| 134 |
|
| 135 |
#define _mm_mask_rsqrt28_ss(S, M, A, B) \ |
| 136 |
_mm_mask_rsqrt28_round_ss((S), (M), (A), (B), _MM_FROUND_CUR_DIRECTION) |
| 137 |
|
| 138 |
#define _mm_maskz_rsqrt28_ss(M, A, B) \ |
| 139 |
_mm_maskz_rsqrt28_round_ss((M), (A), (B), _MM_FROUND_CUR_DIRECTION) |
| 140 |
|
| 141 |
#define _mm_rsqrt28_round_sd(A, B, R) \ |
| 142 |
((__m128d)__builtin_ia32_rsqrt28sd_round_mask((__v2df)(__m128d)(A), \ |
| 143 |
(__v2df)(__m128d)(B), \ |
| 144 |
(__v2df)_mm_setzero_pd(), \ |
| 145 |
(__mmask8)-1, (int)(R))) |
| 146 |
|
| 147 |
#define _mm_mask_rsqrt28_round_sd(S, M, A, B, R) \ |
| 148 |
((__m128d)__builtin_ia32_rsqrt28sd_round_mask((__v2df)(__m128d)(A), \ |
| 149 |
(__v2df)(__m128d)(B), \ |
| 150 |
(__v2df)(__m128d)(S), \ |
| 151 |
(__mmask8)(M), (int)(R))) |
| 152 |
|
| 153 |
#define _mm_maskz_rsqrt28_round_sd(M, A, B, R) \ |
| 154 |
((__m128d)__builtin_ia32_rsqrt28sd_round_mask((__v2df)(__m128d)(A), \ |
| 155 |
(__v2df)(__m128d)(B), \ |
| 156 |
(__v2df)_mm_setzero_pd(), \ |
| 157 |
(__mmask8)(M), (int)(R))) |
| 158 |
|
| 159 |
#define _mm_rsqrt28_sd(A, B) \ |
| 160 |
_mm_rsqrt28_round_sd((A), (B), _MM_FROUND_CUR_DIRECTION) |
| 161 |
|
| 162 |
#define _mm_mask_rsqrt28_sd(S, M, A, B) \ |
| 163 |
_mm_mask_rsqrt28_round_sd((S), (M), (A), (B), _MM_FROUND_CUR_DIRECTION) |
| 164 |
|
| 165 |
#define _mm_maskz_rsqrt28_sd(M, A, B) \ |
| 166 |
_mm_maskz_rsqrt28_round_sd((M), (A), (B), _MM_FROUND_CUR_DIRECTION) |
| 167 |
|
| 168 |
/* rcp28 */ |
| 169 |
#define _mm512_rcp28_round_pd(A, R) \ |
| 170 |
((__m512d)__builtin_ia32_rcp28pd_mask((__v8df)(__m512d)(A), \ |
| 171 |
(__v8df)_mm512_setzero_pd(), \ |
| 172 |
(__mmask8)-1, (int)(R))) |
| 173 |
|
| 174 |
#define _mm512_mask_rcp28_round_pd(S, M, A, R) \ |
| 175 |
((__m512d)__builtin_ia32_rcp28pd_mask((__v8df)(__m512d)(A), \ |
| 176 |
(__v8df)(__m512d)(S), (__mmask8)(M), \ |
| 177 |
(int)(R))) |
| 178 |
|
| 179 |
#define _mm512_maskz_rcp28_round_pd(M, A, R) \ |
| 180 |
((__m512d)__builtin_ia32_rcp28pd_mask((__v8df)(__m512d)(A), \ |
| 181 |
(__v8df)_mm512_setzero_pd(), \ |
| 182 |
(__mmask8)(M), (int)(R))) |
| 183 |
|
| 184 |
#define _mm512_rcp28_pd(A) \ |
| 185 |
_mm512_rcp28_round_pd((A), _MM_FROUND_CUR_DIRECTION) |
| 186 |
|
| 187 |
#define _mm512_mask_rcp28_pd(S, M, A) \ |
| 188 |
_mm512_mask_rcp28_round_pd((S), (M), (A), _MM_FROUND_CUR_DIRECTION) |
| 189 |
|
| 190 |
#define _mm512_maskz_rcp28_pd(M, A) \ |
| 191 |
_mm512_maskz_rcp28_round_pd((M), (A), _MM_FROUND_CUR_DIRECTION) |
| 192 |
|
| 193 |
#define _mm512_rcp28_round_ps(A, R) \ |
| 194 |
((__m512)__builtin_ia32_rcp28ps_mask((__v16sf)(__m512)(A), \ |
| 195 |
(__v16sf)_mm512_setzero_ps(), \ |
| 196 |
(__mmask16)-1, (int)(R))) |
| 197 |
|
| 198 |
#define _mm512_mask_rcp28_round_ps(S, M, A, R) \ |
| 199 |
((__m512)__builtin_ia32_rcp28ps_mask((__v16sf)(__m512)(A), \ |
| 200 |
(__v16sf)(__m512)(S), (__mmask16)(M), \ |
| 201 |
(int)(R))) |
| 202 |
|
| 203 |
#define _mm512_maskz_rcp28_round_ps(M, A, R) \ |
| 204 |
((__m512)__builtin_ia32_rcp28ps_mask((__v16sf)(__m512)(A), \ |
| 205 |
(__v16sf)_mm512_setzero_ps(), \ |
| 206 |
(__mmask16)(M), (int)(R))) |
| 207 |
|
| 208 |
#define _mm512_rcp28_ps(A) \ |
| 209 |
_mm512_rcp28_round_ps((A), _MM_FROUND_CUR_DIRECTION) |
| 210 |
|
| 211 |
#define _mm512_mask_rcp28_ps(S, M, A) \ |
| 212 |
_mm512_mask_rcp28_round_ps((S), (M), (A), _MM_FROUND_CUR_DIRECTION) |
| 213 |
|
| 214 |
#define _mm512_maskz_rcp28_ps(M, A) \ |
| 215 |
_mm512_maskz_rcp28_round_ps((M), (A), _MM_FROUND_CUR_DIRECTION) |
| 216 |
|
| 217 |
#define _mm_rcp28_round_ss(A, B, R) \ |
| 218 |
((__m128)__builtin_ia32_rcp28ss_round_mask((__v4sf)(__m128)(A), \ |
| 219 |
(__v4sf)(__m128)(B), \ |
| 220 |
(__v4sf)_mm_setzero_ps(), \ |
| 221 |
(__mmask8)-1, (int)(R))) |
| 222 |
|
| 223 |
#define _mm_mask_rcp28_round_ss(S, M, A, B, R) \ |
| 224 |
((__m128)__builtin_ia32_rcp28ss_round_mask((__v4sf)(__m128)(A), \ |
| 225 |
(__v4sf)(__m128)(B), \ |
| 226 |
(__v4sf)(__m128)(S), \ |
| 227 |
(__mmask8)(M), (int)(R))) |
| 228 |
|
| 229 |
#define _mm_maskz_rcp28_round_ss(M, A, B, R) \ |
| 230 |
((__m128)__builtin_ia32_rcp28ss_round_mask((__v4sf)(__m128)(A), \ |
| 231 |
(__v4sf)(__m128)(B), \ |
| 232 |
(__v4sf)_mm_setzero_ps(), \ |
| 233 |
(__mmask8)(M), (int)(R))) |
| 234 |
|
| 235 |
#define _mm_rcp28_ss(A, B) \ |
| 236 |
_mm_rcp28_round_ss((A), (B), _MM_FROUND_CUR_DIRECTION) |
| 237 |
|
| 238 |
#define _mm_mask_rcp28_ss(S, M, A, B) \ |
| 239 |
_mm_mask_rcp28_round_ss((S), (M), (A), (B), _MM_FROUND_CUR_DIRECTION) |
| 240 |
|
| 241 |
#define _mm_maskz_rcp28_ss(M, A, B) \ |
| 242 |
_mm_maskz_rcp28_round_ss((M), (A), (B), _MM_FROUND_CUR_DIRECTION) |
| 243 |
|
| 244 |
#define _mm_rcp28_round_sd(A, B, R) \ |
| 245 |
((__m128d)__builtin_ia32_rcp28sd_round_mask((__v2df)(__m128d)(A), \ |
| 246 |
(__v2df)(__m128d)(B), \ |
| 247 |
(__v2df)_mm_setzero_pd(), \ |
| 248 |
(__mmask8)-1, (int)(R))) |
| 249 |
|
| 250 |
#define _mm_mask_rcp28_round_sd(S, M, A, B, R) \ |
| 251 |
((__m128d)__builtin_ia32_rcp28sd_round_mask((__v2df)(__m128d)(A), \ |
| 252 |
(__v2df)(__m128d)(B), \ |
| 253 |
(__v2df)(__m128d)(S), \ |
| 254 |
(__mmask8)(M), (int)(R))) |
| 255 |
|
| 256 |
#define _mm_maskz_rcp28_round_sd(M, A, B, R) \ |
| 257 |
((__m128d)__builtin_ia32_rcp28sd_round_mask((__v2df)(__m128d)(A), \ |
| 258 |
(__v2df)(__m128d)(B), \ |
| 259 |
(__v2df)_mm_setzero_pd(), \ |
| 260 |
(__mmask8)(M), (int)(R))) |
| 261 |
|
| 262 |
#define _mm_rcp28_sd(A, B) \ |
| 263 |
_mm_rcp28_round_sd((A), (B), _MM_FROUND_CUR_DIRECTION) |
| 264 |
|
| 265 |
#define _mm_mask_rcp28_sd(S, M, A, B) \ |
| 266 |
_mm_mask_rcp28_round_sd((S), (M), (A), (B), _MM_FROUND_CUR_DIRECTION) |
| 267 |
|
| 268 |
#define _mm_maskz_rcp28_sd(M, A, B) \ |
| 269 |
_mm_maskz_rcp28_round_sd((M), (A), (B), _MM_FROUND_CUR_DIRECTION) |
| 270 |
|
| 271 |
#endif /* __AVX512ERINTRIN_H */ |