gcc/config/i386/avx512pfintrin.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 <avx512pfintrin.h> directly; include <immintrin.h> instead."
  27 #endif
  28
  29 #ifndef _AVX512PFINTRIN_H_INCLUDED
  30 #define _AVX512PFINTRIN_H_INCLUDED
  31
  32 #ifndef __AVX512PF__
  33 #pragma GCC push_options
  34 #pragma GCC target("avx512pf")
  35 #define __DISABLE_AVX512PF__
  36 #endif /* __AVX512PF__ */
  37
  38 /* Internal data types for implementing the intrinsics.  */
  39 typedef long long __v8di __attribute__ ((__vector_size__ (64)));
  40 typedef int __v16si __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 long long __m512i __attribute__ ((__vector_size__ (64), __may_alias__));
  45
  46 typedef unsigned char  __mmask8;
  47 typedef unsigned short __mmask16;
  48
  49 #ifdef __OPTIMIZE__
  50 extern __inline void
  51 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
  52 _mm512_mask_prefetch_i32gather_ps (__m512i index, __mmask16 mask,
  53                                    int const *addr, int scale, int hint)
  54 {
  55   __builtin_ia32_gatherpfdps (mask, (__v16si) index, addr, scale, hint);
  56 }
  57
  58 extern __inline void
  59 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
  60 _mm512_mask_prefetch_i64gather_ps (__m512i index, __mmask8 mask,
  61                                    int const *addr, int scale, int hint)
  62 {
  63   __builtin_ia32_gatherpfqps (mask, (__v8di) index, addr, scale, hint);
  64 }
  65
  66 extern __inline void
  67 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
  68 _mm512_prefetch_i32scatter_ps (int const *addr, __m512i index, int scale,
  69                                int hint)
  70 {
  71   __builtin_ia32_scatterpfdps ((__mmask16) 0xFFFF, (__v16si) index, addr, scale,
  72                                hint);
  73 }
  74
  75 extern __inline void
  76 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
  77 _mm512_mask_prefetch_i32scatter_ps (int const *addr, __mmask16 mask,
  78                                     __m512i index, int scale, int hint)
  79 {
  80   __builtin_ia32_scatterpfdps (mask, (__v16si) index, addr, scale, hint);
  81 }
  82
  83 extern __inline void
  84 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
  85 _mm512_prefetch_i64scatter_ps (int const *addr, __m512i index, int scale,
  86                                int hint)
  87 {
  88   __builtin_ia32_scatterpfqps ((__mmask8) 0xFF, (__v8di) index, addr, scale,
  89                                hint);
  90 }
  91
  92 extern __inline void
  93 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
  94 _mm512_mask_prefetch_i64scatter_ps (int const *addr, __mmask16 mask,
  95                                     __m512i index, int scale, int hint)
  96 {
  97   __builtin_ia32_scatterpfqps (mask, (__v8di) index, addr, scale, hint);
  98 }
  99 #else
 100 #define _mm512_mask_prefetch_i32gather_ps(INDEX, MASK, ADDR, SCALE, HINT)    \
 101   __builtin_ia32_gatherpfdps ((__mmask16)MASK, (__v16si)(__m512i)INDEX,      \
 102                               (int const *)ADDR, (int)SCALE, (int)HINT)
 103
 104 #define _mm512_mask_prefetch_i64gather_ps(INDEX, MASK, ADDR, SCALE, HINT)    \
 105   __builtin_ia32_gatherpfqps ((__mmask8)MASK, (__v8di)(__m512i)INDEX,        \
 106                               (int const *)ADDR, (int)SCALE, (int)HINT)
 107
 108 #define _mm512_prefetch_i32scatter_ps(ADDR, INDEX, SCALE, HINT)              \
 109   __builtin_ia32_scatterpfdps ((__mmask16)0xFFFF, (__v16si)(__m512i)INDEX,   \
 110                                (int const *)ADDR, (int)SCALE, (int)HINT)
 111
 112 #define _mm512_mask_prefetch_i32scatter_ps(ADDR, MASK, INDEX, SCALE, HINT)   \
 113   __builtin_ia32_scatterpfdps ((__mmask16)MASK, (__v16si)(__m512i)INDEX,     \
 114                                (int const *)ADDR, (int)SCALE, (int)HINT)
 115
 116 #define _mm512_prefetch_i64scatter_ps(ADDR, INDEX, SCALE, HINT)              \
 117   __builtin_ia32_scatterpfqps ((__mmask8)0xFF, (__v8di)(__m512i)INDEX,       \
 118                                (int const *)ADDR, (int)SCALE, (int)HINT)
 119
 120 #define _mm512_mask_prefetch_i64scatter_ps(ADDR, MASK, INDEX, SCALE, HINT)   \
 121   __builtin_ia32_scatterpfqps ((__mmask8)MASK, (__v8di)(__m512i)INDEX,       \
 122                                (int const *)ADDR, (int)SCALE, (int)HINT)
 123 #endif
 124
 125 #ifdef __DISABLE_AVX512PF__
 126 #undef __DISABLE_AVX512PF__
 127 #pragma GCC pop_options
 128 #endif /* __DISABLE_AVX512PF__ */
 129
 130 #endif /* _AVX512PFINTRIN_H_INCLUDED */