old-autovect-branch/gcc/config/i386/mm3dnow.h

   1 /* Copyright (C) 2004 Free Software Foundation, Inc.
   2
   3    This file is part of GCC.
   4
   5    GCC is free software; you can redistribute it and/or modify
   6    it under the terms of the GNU General Public License as published by
   7    the Free Software Foundation; either version 2, or (at your option)
   8    any later version.
   9
  10    GCC is distributed in the hope that it will be useful,
  11    but WITHOUT ANY WARRANTY; without even the implied warranty of
  12    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  13    GNU General Public License for more details.
  14
  15    You should have received a copy of the GNU General Public License
  16    along with GCC; see the file COPYING.  If not, write to
  17    the Free Software Foundation, 51 Franklin Street, Fifth Floor,
  18    Boston, MA 02110-1301, USA.  */
  19
  20 /* As a special exception, if you include this header file into source
  21    files compiled by GCC, this header file does not by itself cause
  22    the resulting executable to be covered by the GNU General Public
  23    License.  This exception does not however invalidate any other
  24    reasons why the executable file might be covered by the GNU General
  25    Public License.  */
  26
  27 /* Implemented from the mm3dnow.h (of supposedly AMD origin) included with
  28    MSVC 7.1.  */
  29
  30 #ifndef _MM3DNOW_H_INCLUDED
  31 #define _MM3DNOW_H_INCLUDED
  32
  33 #ifdef __3dNOW__
  34
  35 #include <mmintrin.h>
  36
  37 /* Internal data types for implementing the intrinsics.  */
  38 typedef float __v2sf __attribute__ ((__vector_size__ (8)));
  39
  40 static __inline void
  41 _m_femms (void)
  42 {
  43   __builtin_ia32_femms();
  44 }
  45
  46 static __inline __m64
  47 _m_pavgusb (__m64 __A, __m64 __B)
  48 {
  49   return (__m64)__builtin_ia32_pavgusb ((__v8qi)__A, (__v8qi)__B);
  50 }
  51
  52 static __inline __m64
  53 _m_pf2id (__m64 __A)
  54 {
  55   return (__m64)__builtin_ia32_pf2id ((__v2sf)__A);
  56 }
  57
  58 static __inline __m64
  59 _m_pfacc (__m64 __A, __m64 __B)
  60 {
  61   return (__m64)__builtin_ia32_pfacc ((__v2sf)__A, (__v2sf)__B);
  62 }
  63
  64 static __inline __m64
  65 _m_pfadd (__m64 __A, __m64 __B)
  66 {
  67   return (__m64)__builtin_ia32_pfadd ((__v2sf)__A, (__v2sf)__B);
  68 }
  69
  70 static __inline __m64
  71 _m_pfcmpeq (__m64 __A, __m64 __B)
  72 {
  73   return (__m64)__builtin_ia32_pfcmpeq ((__v2sf)__A, (__v2sf)__B);
  74 }
  75
  76 static __inline __m64
  77 _m_pfcmpge (__m64 __A, __m64 __B)
  78 {
  79   return (__m64)__builtin_ia32_pfcmpge ((__v2sf)__A, (__v2sf)__B);
  80 }
  81
  82 static __inline __m64
  83 _m_pfcmpgt (__m64 __A, __m64 __B)
  84 {
  85   return (__m64)__builtin_ia32_pfcmpgt ((__v2sf)__A, (__v2sf)__B);
  86 }
  87
  88 static __inline __m64
  89 _m_pfmax (__m64 __A, __m64 __B)
  90 {
  91   return (__m64)__builtin_ia32_pfmax ((__v2sf)__A, (__v2sf)__B);
  92 }
  93
  94 static __inline __m64
  95 _m_pfmin (__m64 __A, __m64 __B)
  96 {
  97   return (__m64)__builtin_ia32_pfmin ((__v2sf)__A, (__v2sf)__B);
  98 }
  99
 100 static __inline __m64
 101 _m_pfmul (__m64 __A, __m64 __B)
 102 {
 103   return (__m64)__builtin_ia32_pfmul ((__v2sf)__A, (__v2sf)__B);
 104 }
 105
 106 static __inline __m64
 107 _m_pfrcp (__m64 __A)
 108 {
 109   return (__m64)__builtin_ia32_pfrcp ((__v2sf)__A);
 110 }
 111
 112 static __inline __m64
 113 _m_pfrcpit1 (__m64 __A, __m64 __B)
 114 {
 115   return (__m64)__builtin_ia32_pfrcpit1 ((__v2sf)__A, (__v2sf)__B);
 116 }
 117
 118 static __inline __m64
 119 _m_pfrcpit2 (__m64 __A, __m64 __B)
 120 {
 121   return (__m64)__builtin_ia32_pfrcpit2 ((__v2sf)__A, (__v2sf)__B);
 122 }
 123
 124 static __inline __m64
 125 _m_pfrsqrt (__m64 __A)
 126 {
 127   return (__m64)__builtin_ia32_pfrsqrt ((__v2sf)__A);
 128 }
 129
 130 static __inline __m64
 131 _m_pfrsqit1 (__m64 __A, __m64 __B)
 132 {
 133   return (__m64)__builtin_ia32_pfrsqit1 ((__v2sf)__A, (__v2sf)__B);
 134 }
 135
 136 static __inline __m64
 137 _m_pfsub (__m64 __A, __m64 __B)
 138 {
 139   return (__m64)__builtin_ia32_pfsub ((__v2sf)__A, (__v2sf)__B);
 140 }
 141
 142 static __inline __m64
 143 _m_pfsubr (__m64 __A, __m64 __B)
 144 {
 145   return (__m64)__builtin_ia32_pfsubr ((__v2sf)__A, (__v2sf)__B);
 146 }
 147
 148 static __inline __m64
 149 _m_pi2fd (__m64 __A)
 150 {
 151   return (__m64)__builtin_ia32_pi2fd ((__v2si)__A);
 152 }
 153
 154 static __inline __m64
 155 _m_pmulhrw (__m64 __A, __m64 __B)
 156 {
 157   return (__m64)__builtin_ia32_pmulhrw ((__v4hi)__A, (__v4hi)__B);
 158 }
 159
 160 static __inline void
 161 _m_prefetch (void *__P)
 162 {
 163   __builtin_prefetch (__P, 0, 3 /* _MM_HINT_T0 */);
 164 }
 165
 166 static __inline void
 167 _m_prefetchw (void *__P)
 168 {
 169   __builtin_prefetch (__P, 1, 3 /* _MM_HINT_T0 */);
 170 }
 171
 172 static __inline __m64
 173 _m_from_float (float __A)
 174 {
 175   return (__m64)(__v2sf){ __A, 0 };
 176 }
 177
 178 static __inline float
 179 _m_to_float (__m64 __A)
 180 {
 181   union { __v2sf v; float a[2]; } __tmp = { (__v2sf)__A };
 182   return __tmp.a[0];
 183 }
 184
 185 #ifdef __3dNOW_A__
 186
 187 static __inline __m64
 188 _m_pf2iw (__m64 __A)
 189 {
 190   return (__m64)__builtin_ia32_pf2iw ((__v2sf)__A);
 191 }
 192
 193 static __inline __m64
 194 _m_pfnacc (__m64 __A, __m64 __B)
 195 {
 196   return (__m64)__builtin_ia32_pfnacc ((__v2sf)__A, (__v2sf)__B);
 197 }
 198
 199 static __inline __m64
 200 _m_pfpnacc (__m64 __A, __m64 __B)
 201 {
 202   return (__m64)__builtin_ia32_pfpnacc ((__v2sf)__A, (__v2sf)__B);
 203 }
 204
 205 static __inline __m64
 206 _m_pi2fw (__m64 __A)
 207 {
 208   return (__m64)__builtin_ia32_pi2fw ((__v2si)__A);
 209 }
 210
 211 static __inline __m64
 212 _m_pswapd (__m64 __A)
 213 {
 214   return (__m64)__builtin_ia32_pswapdsf ((__v2sf)__A);
 215 }
 216
 217 #endif /* __3dNOW_A__ */
 218 #endif /* __3dNOW__ */
 219
 220 #endif /* _MM3DNOW_H_INCLUDED */