include/gmx_x86_sse4_1.h

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012, by the GROMACS development team, led by
   5  * David van der Spoel, Berk Hess, Erik Lindahl, and including many
   6  * others, as listed in the AUTHORS file in the top-level source
   7  * directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 #ifndef _gmx_x86_sse4_1_h_
  36 #define _gmx_x86_sse4_1_h_
  37
  38 #include <smmintrin.h>
  39
  40 #include <stdio.h>
  41
  42 #include "types/simple.h"
  43
  44 /* Create some basic definitions that are not 100% SSE2 standard and thus not
  45  * available on all compilers. These should be fairly self-evident by comparing
  46  * with an arbitrary emmintrin.h.
  47  */
  48
  49
  50 #define gmx_mm_extract_epi32(x, imm) _mm_extract_epi32((x),(imm))
  51
  52 #define GMX_MM_TRANSPOSE2_PD(row0, row1) {           \
  53     __m128d __gmx_t1 = row0;                         \
  54     row0           = _mm_unpacklo_pd(row0,row1);     \
  55     row1           = _mm_unpackhi_pd(__gmx_t1,row1); \
  56 }
  57
  58 #define _GMX_MM_BLEND(b3,b2,b1,b0) (((b3) << 3) | ((b2) << 2) | ((b1) << 1) | ((b0)))
  59
  60 #if (defined (_MSC_VER) || defined(__INTEL_COMPILER))
  61 #  define gmx_mm_castsi128_ps(a) _mm_castsi128_ps(a)
  62 #  define gmx_mm_castps_si128(a) _mm_castps_si128(a)
  63 #  define gmx_mm_castps_ps128(a) (a)
  64 #  define gmx_mm_castsi128_pd(a) _mm_castsi128_pd(a)
  65 #  define gmx_mm_castpd_si128(a) _mm_castpd_si128(a)
  66 #elif defined(__GNUC__)
  67 #  define gmx_mm_castsi128_ps(a) ((__m128)(a))
  68 #  define gmx_mm_castps_si128(a) ((__m128i)(a))
  69 #  define gmx_mm_castps_ps128(a) ((__m128)(a))
  70 #  define gmx_mm_castsi128_pd(a) ((__m128d)(a))
  71 #  define gmx_mm_castpd_si128(a) ((__m128i)(a))
  72 #else
  73 static __m128  gmx_mm_castsi128_ps(__m128i a)
  74 {
  75     return *(__m128 *) &a;
  76 }
  77 static __m128i gmx_mm_castps_si128(__m128 a)
  78 {
  79     return *(__m128i *) &a;
  80 }
  81 static __m128  gmx_mm_castps_ps128(__m128 a)
  82 {
  83     return *(__m128 *) &a;
  84 }
  85 static __m128d gmx_mm_castsi128_pd(__m128i a)
  86 {
  87     return *(__m128d *) &a;
  88 }
  89 static __m128i gmx_mm_castpd_si128(__m128d a)
  90 {
  91     return *(__m128i *) &a;
  92 }
  93 #endif
  94
  95
  96 static void
  97 gmx_mm_printxmm_ps(const char *s,__m128 xmm)
  98 {
  99     float f[4];
 100
 101     _mm_storeu_ps(f,xmm);
 102     printf("%s: %15.10e %15.10e %15.10e %15.10e\n",s,f[0],f[1],f[2],f[3]);
 103 }
 104
 105
 106 static void
 107 gmx_mm_printxmmsum_ps(const char *s,__m128 xmm)
 108 {
 109     float f[4];
 110
 111     _mm_storeu_ps(f,xmm);
 112     printf("%s (sum): %15.10g\n",s,f[0]+f[1]+f[2]+f[3]);
 113 }
 114
 115
 116 static void
 117 gmx_mm_printxmm_pd(const char *s,__m128d xmm)
 118 {
 119     double f[2];
 120
 121     _mm_storeu_pd(f,xmm);
 122     printf("%s: %30.20e %30.20e\n",s,f[0],f[1]);
 123 }
 124
 125 static void
 126 gmx_mm_printxmmsum_pd(const char *s,__m128d xmm)
 127 {
 128     double f[2];
 129
 130     _mm_storeu_pd(f,xmm);
 131     printf("%s (sum): %15.10g\n",s,f[0]+f[1]);
 132 }
 133
 134
 135 static void
 136 gmx_mm_printxmm_epi32(const char *s,__m128i xmmi)
 137 {
 138     int i[4];
 139
 140     _mm_storeu_si128((__m128i *)i,xmmi);
 141     printf("%10s: %2d %2d %2d %2d\n",s,i[0],i[1],i[2],i[3]);
 142 }
 143
 144
 145
 146 static int gmx_mm_check_and_reset_overflow(void)
 147 {
 148     int MXCSR;
 149     int sse_overflow;
 150
 151     MXCSR = _mm_getcsr();
 152     /* The overflow flag is bit 3 in the register */
 153     if (MXCSR & 0x0008)
 154     {
 155         sse_overflow = 1;
 156         /* Set the overflow flag to zero */
 157         MXCSR = MXCSR & 0xFFF7;
 158         _mm_setcsr(MXCSR);
 159     }
 160     else
 161     {
 162         sse_overflow = 0;
 163     }
 164
 165     return sse_overflow;
 166 }
 167
 168
 169 #endif /* _gmx_x86_sse4_1_h_ */