gcc/config/i386/avx512erintrin.h

   1 /* Copyright (C) 2013
   2    Free Software Foundation, Inc.
   3
   4    This file is part of GCC.
   5
   6    GCC is free software; you can redistribute it and/or modify
   7    it under the terms of the GNU General Public License as published by
   8    the Free Software Foundation; either version 3, or (at your option)
   9    any later version.
  10
  11    GCC is distributed in the hope that it will be useful,
  12    but WITHOUT ANY WARRANTY; without even the implied warranty of
  13    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  14    GNU General Public License for more details.
  15
  16    Under Section 7 of GPL version 3, you are granted additional
  17    permissions described in the GCC Runtime Library Exception, version
  18    3.1, as published by the Free Software Foundation.
  19
  20    You should have received a copy of the GNU General Public License and
  21    a copy of the GCC Runtime Library Exception along with this program;
  22    see the files COPYING3 and COPYING.RUNTIME respectively.  If not, see
  23    <http://www.gnu.org/licenses/>.  */
  24
  25 #ifndef _IMMINTRIN_H_INCLUDED
  26 #error "Never use <avx512erintrin.h> directly; include <immintrin.h> instead."
  27 #endif
  28
  29 #ifndef _AVX512ERINTRIN_H_INCLUDED
  30 #define _AVX512ERINTRIN_H_INCLUDED
  31
  32 #ifndef __AVX512ER__
  33 #pragma GCC push_options
  34 #pragma GCC target("avx512er")
  35 #define __DISABLE_AVX512ER__
  36 #endif /* __AVX512ER__ */
  37
  38 /* Internal data types for implementing the intrinsics.  */
  39 typedef double __v8df __attribute__ ((__vector_size__ (64)));
  40 typedef float __v16sf __attribute__ ((__vector_size__ (64)));
  41
  42 /* The Intel API is flexible enough that we must allow aliasing with other
  43    vector types, and their scalar components.  */
  44 typedef float __m512 __attribute__ ((__vector_size__ (64), __may_alias__));
  45 typedef double __m512d __attribute__ ((__vector_size__ (64), __may_alias__));
  46
  47 typedef unsigned char  __mmask8;
  48 typedef unsigned short __mmask16;
  49
  50 #ifdef __OPTIMIZE__
  51 extern __inline __m512d
  52 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
  53 _mm512_exp2a23_round_pd (__m512d __A, int __R)
  54 {
  55   __m512d __W;
  56   return (__m512d) __builtin_ia32_exp2pd_mask ((__v8df) __A,
  57                                                (__v8df) __W,
  58                                                (__mmask8) -1, __R);
  59 }
  60
  61 extern __inline __m512d
  62 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
  63 _mm512_mask_exp2a23_round_pd (__m512d __W, __mmask8 __U, __m512d __A, int __R)
  64 {
  65   return (__m512d) __builtin_ia32_exp2pd_mask ((__v8df) __A,
  66                                                (__v8df) __W,
  67                                                (__mmask8) __U, __R);
  68 }
  69
  70 extern __inline __m512d
  71 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
  72 _mm512_maskz_exp2a23_round_pd (__mmask8 __U, __m512d __A, int __R)
  73 {
  74   return (__m512d) __builtin_ia32_exp2pd_mask ((__v8df) __A,
  75                                                (__v8df) _mm512_setzero_pd (),
  76                                                (__mmask8) __U, __R);
  77 }
  78
  79 extern __inline __m512
  80 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
  81 _mm512_exp2a23_round_ps (__m512 __A, int __R)
  82 {
  83   __m512 __W;
  84   return (__m512) __builtin_ia32_exp2ps_mask ((__v16sf) __A,
  85                                               (__v16sf) __W,
  86                                               (__mmask16) -1, __R);
  87 }
  88
  89 extern __inline __m512
  90 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
  91 _mm512_mask_exp2a23_round_ps (__m512 __W, __mmask16 __U, __m512 __A, int __R)
  92 {
  93   return (__m512) __builtin_ia32_exp2ps_mask ((__v16sf) __A,
  94                                               (__v16sf) __W,
  95                                               (__mmask16) __U, __R);
  96 }
  97
  98 extern __inline __m512
  99 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 100 _mm512_maskz_exp2a23_round_ps (__mmask16 __U, __m512 __A, int __R)
 101 {
 102   return (__m512) __builtin_ia32_exp2ps_mask ((__v16sf) __A,
 103                                               (__v16sf) _mm512_setzero_ps (),
 104                                               (__mmask16) __U, __R);
 105 }
 106
 107 extern __inline __m512d
 108 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 109 _mm512_rcp28_round_pd (__m512d __A, int __R)
 110 {
 111   __m512d __W;
 112   return (__m512d) __builtin_ia32_rcp28pd_mask ((__v8df) __A,
 113                                                 (__v8df) __W,
 114                                                 (__mmask8) -1, __R);
 115 }
 116
 117 extern __inline __m512d
 118 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 119 _mm512_mask_rcp28_round_pd (__m512d __W, __mmask8 __U, __m512d __A, int __R)
 120 {
 121   return (__m512d) __builtin_ia32_rcp28pd_mask ((__v8df) __A,
 122                                                 (__v8df) __W,
 123                                                 (__mmask8) __U, __R);
 124 }
 125
 126 extern __inline __m512d
 127 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 128 _mm512_maskz_rcp28_round_pd (__mmask8 __U, __m512d __A, int __R)
 129 {
 130   return (__m512d) __builtin_ia32_rcp28pd_mask ((__v8df) __A,
 131                                                 (__v8df) _mm512_setzero_pd (),
 132                                                 (__mmask8) __U, __R);
 133 }
 134
 135 extern __inline __m512
 136 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 137 _mm512_rcp28_round_ps (__m512 __A, int __R)
 138 {
 139   __m512 __W;
 140   return (__m512) __builtin_ia32_rcp28ps_mask ((__v16sf) __A,
 141                                                (__v16sf) __W,
 142                                                (__mmask16) -1, __R);
 143 }
 144
 145 extern __inline __m512
 146 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 147 _mm512_mask_rcp28_round_ps (__m512 __W, __mmask16 __U, __m512 __A, int __R)
 148 {
 149   return (__m512) __builtin_ia32_rcp28ps_mask ((__v16sf) __A,
 150                                                (__v16sf) __W,
 151                                                (__mmask16) __U, __R);
 152 }
 153
 154 extern __inline __m512
 155 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 156 _mm512_maskz_rcp28_round_ps (__mmask16 __U, __m512 __A, int __R)
 157 {
 158   return (__m512) __builtin_ia32_rcp28ps_mask ((__v16sf) __A,
 159                                                (__v16sf) _mm512_setzero_ps (),
 160                                                (__mmask16) __U, __R);
 161 }
 162
 163 extern __inline __m512d
 164 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 165 _mm512_rsqrt28_round_pd (__m512d __A, int __R)
 166 {
 167   __m512d __W;
 168   return (__m512d) __builtin_ia32_rsqrt28pd_mask ((__v8df) __A,
 169                                                   (__v8df) __W,
 170                                                   (__mmask8) -1, __R);
 171 }
 172
 173 extern __inline __m512d
 174 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 175 _mm512_mask_rsqrt28_round_pd (__m512d __W, __mmask8 __U, __m512d __A, int __R)
 176 {
 177   return (__m512d) __builtin_ia32_rsqrt28pd_mask ((__v8df) __A,
 178                                                   (__v8df) __W,
 179                                                   (__mmask8) __U, __R);
 180 }
 181
 182 extern __inline __m512d
 183 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 184 _mm512_maskz_rsqrt28_round_pd (__mmask8 __U, __m512d __A, int __R)
 185 {
 186   return (__m512d) __builtin_ia32_rsqrt28pd_mask ((__v8df) __A,
 187                                                   (__v8df) _mm512_setzero_pd (),
 188                                                   (__mmask8) __U, __R);
 189 }
 190
 191 extern __inline __m512
 192 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 193 _mm512_rsqrt28_round_ps (__m512 __A, int __R)
 194 {
 195   __m512 __W;
 196   return (__m512) __builtin_ia32_rsqrt28ps_mask ((__v16sf) __A,
 197                                                  (__v16sf) __W,
 198                                                  (__mmask16) -1, __R);
 199 }
 200
 201 extern __inline __m512
 202 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 203 _mm512_mask_rsqrt28_round_ps (__m512 __W, __mmask16 __U, __m512 __A, int __R)
 204 {
 205   return (__m512) __builtin_ia32_rsqrt28ps_mask ((__v16sf) __A,
 206                                                  (__v16sf) __W,
 207                                                  (__mmask16) __U, __R);
 208 }
 209
 210 extern __inline __m512
 211 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 212 _mm512_maskz_rsqrt28_round_ps (__mmask16 __U, __m512 __A, int __R)
 213 {
 214   return (__m512) __builtin_ia32_rsqrt28ps_mask ((__v16sf) __A,
 215                                                  (__v16sf) _mm512_setzero_ps (),
 216                                                  (__mmask16) __U, __R);
 217 }
 218 #else
 219 #define _mm512_exp2a23_round_pd(A, C)            \
 220     __builtin_ia32_exp2pd_mask(A, (__v8df)_mm512_setzero_pd(), -1, C)
 221
 222 #define _mm512_mask_exp2a23_round_pd(W, U, A, C) \
 223     __builtin_ia32_exp2pd_mask(A, W, U, C)
 224
 225 #define _mm512_maskz_exp2a23_round_pd(U, A, C)   \
 226     __builtin_ia32_exp2pd_mask(A, (__v8df)_mm512_setzero_pd(), U, C)
 227
 228 #define _mm512_exp2a23_round_ps(A, C)            \
 229     __builtin_ia32_exp2ps_mask(A, (__v16sf)_mm512_setzero_ps(), -1, C)
 230
 231 #define _mm512_mask_exp2a23_round_ps(W, U, A, C) \
 232     __builtin_ia32_exp2ps_mask(A, W, U, C)
 233
 234 #define _mm512_maskz_exp2a23_round_ps(U, A, C)   \
 235     __builtin_ia32_exp2ps_mask(A, (__v16sf)_mm512_setzero_ps(), U, C)
 236
 237 #define _mm512_rcp28_round_pd(A, C)            \
 238     __builtin_ia32_rcp28pd_mask(A, (__v8df)_mm512_setzero_pd(), -1, C)
 239
 240 #define _mm512_mask_rcp28_round_pd(W, U, A, C) \
 241     __builtin_ia32_rcp28pd_mask(A, W, U, C)
 242
 243 #define _mm512_maskz_rcp28_round_pd(U, A, C)   \
 244     __builtin_ia32_rcp28pd_mask(A, (__v8df)_mm512_setzero_pd(), U, C)
 245
 246 #define _mm512_rcp28_round_ps(A, C)            \
 247     __builtin_ia32_rcp28ps_mask(A, (__v16sf)_mm512_setzero_ps(), -1, C)
 248
 249 #define _mm512_mask_rcp28_round_ps(W, U, A, C) \
 250     __builtin_ia32_rcp28ps_mask(A, W, U, C)
 251
 252 #define _mm512_maskz_rcp28_round_ps(U, A, C)   \
 253     __builtin_ia32_rcp28ps_mask(A, (__v16sf)_mm512_setzero_ps(), U, C)
 254
 255 #define _mm512_rsqrt28_round_pd(A, C)            \
 256     __builtin_ia32_rsqrt28pd_mask(A, (__v8df)_mm512_setzero_pd(), -1, C)
 257
 258 #define _mm512_mask_rsqrt28_round_pd(W, U, A, C) \
 259     __builtin_ia32_rsqrt28pd_mask(A, W, U, C)
 260
 261 #define _mm512_maskz_rsqrt28_round_pd(U, A, C)   \
 262     __builtin_ia32_rsqrt28pd_mask(A, (__v8df)_mm512_setzero_pd(), U, C)
 263
 264 #define _mm512_rsqrt28_round_ps(A, C)            \
 265     __builtin_ia32_rsqrt28ps_mask(A, (__v16sf)_mm512_setzero_ps(), -1, C)
 266
 267 #define _mm512_mask_rsqrt28_round_ps(W, U, A, C) \
 268     __builtin_ia32_rsqrt28ps_mask(A, W, U, C)
 269
 270 #define _mm512_maskz_rsqrt28_round_ps(U, A, C)   \
 271     __builtin_ia32_rsqrt28ps_mask(A, (__v16sf)_mm512_setzero_ps(), U, C)
 272 #endif
 273
 274 #define _mm512_exp2a23_pd(A)                    \
 275     _mm512_exp2a23_round_pd(A, _MM_FROUND_CUR_DIRECTION)
 276
 277 #define _mm512_mask_exp2a23_pd(W, U, A)   \
 278     _mm512_mask_exp2a23_round_pd(W, U, A, _MM_FROUND_CUR_DIRECTION)
 279
 280 #define _mm512_maskz_exp2a23_pd(U, A)     \
 281     _mm512_maskz_exp2a23_round_pd(U, A, _MM_FROUND_CUR_DIRECTION)
 282
 283 #define _mm512_exp2a23_ps(A)                    \
 284     _mm512_exp2a23_round_ps(A, _MM_FROUND_CUR_DIRECTION)
 285
 286 #define _mm512_mask_exp2a23_ps(W, U, A)   \
 287     _mm512_mask_exp2a23_round_ps(W, U, A, _MM_FROUND_CUR_DIRECTION)
 288
 289 #define _mm512_maskz_exp2a23_ps(U, A)     \
 290     _mm512_maskz_exp2a23_round_ps(U, A, _MM_FROUND_CUR_DIRECTION)
 291
 292 #define _mm512_rcp28_pd(A)                    \
 293     _mm512_rcp28_round_pd(A, _MM_FROUND_CUR_DIRECTION)
 294
 295 #define _mm512_mask_rcp28_pd(W, U, A)   \
 296     _mm512_mask_rcp28_round_pd(W, U, A, _MM_FROUND_CUR_DIRECTION)
 297
 298 #define _mm512_maskz_rcp28_pd(U, A)     \
 299     _mm512_maskz_rcp28_round_pd(U, A, _MM_FROUND_CUR_DIRECTION)
 300
 301 #define _mm512_rcp28_ps(A)                    \
 302     _mm512_rcp28_round_ps(A, _MM_FROUND_CUR_DIRECTION)
 303
 304 #define _mm512_mask_rcp28_ps(W, U, A)   \
 305     _mm512_mask_rcp28_round_ps(W, U, A, _MM_FROUND_CUR_DIRECTION)
 306
 307 #define _mm512_maskz_rcp28_ps(U, A)     \
 308     _mm512_maskz_rcp28_round_ps(U, A, _MM_FROUND_CUR_DIRECTION)
 309
 310 #define _mm512_rsqrt28_pd(A)                    \
 311     _mm512_rsqrt28_round_pd(A, _MM_FROUND_CUR_DIRECTION)
 312
 313 #define _mm512_mask_rsqrt28_pd(W, U, A)   \
 314     _mm512_mask_rsqrt28_round_pd(W, U, A, _MM_FROUND_CUR_DIRECTION)
 315
 316 #define _mm512_maskz_rsqrt28_pd(U, A)     \
 317     _mm512_maskz_rsqrt28_round_pd(U, A, _MM_FROUND_CUR_DIRECTION)
 318
 319 #define _mm512_rsqrt28_ps(A)                    \
 320     _mm512_rsqrt28_round_ps(A, _MM_FROUND_CUR_DIRECTION)
 321
 322 #define _mm512_mask_rsqrt28_ps(W, U, A)   \
 323     _mm512_mask_rsqrt28_round_ps(W, U, A, _MM_FROUND_CUR_DIRECTION)
 324
 325 #define _mm512_maskz_rsqrt28_ps(U, A)     \
 326     _mm512_maskz_rsqrt28_round_ps(U, A, _MM_FROUND_CUR_DIRECTION)
 327
 328 #ifdef __DISABLE_AVX512ER__
 329 #undef __DISABLE_AVX512ER__
 330 #pragma GCC pop_options
 331 #endif /* __DISABLE_AVX512ER__ */
 332
 333 #endif /* _AVX512ERINTRIN_H_INCLUDED */