src/gromacs/simd/support.cpp

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2015, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source 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
  36 /*! \internal \file
  37  *
  38  * \brief Implements SIMD architecture support query routines
  39  *
  40  * \author Erik Lindahl <erik.lindahl@scilifelab.se>
  41  *
  42  * \ingroup module_simd
  43  */
  44
  45 #include "gmxpre.h"
  46
  47 #include "support.h"
  48
  49 #include "config.h"
  50
  51 #include <cstdio>
  52 #include <cstdlib>
  53
  54 #include <map>
  55 #include <string>
  56
  57 #include "gromacs/hardware/cpuinfo.h"
  58
  59 namespace gmx
  60 {
  61
  62 /*! \cond libapi */
  63
  64 const std::string &
  65 simdString(SimdType s)
  66 {
  67     static const std::map<SimdType, std::string> name =
  68     {
  69         { SimdType::None,           "None"            },
  70         { SimdType::Reference,      "Reference"       },
  71         { SimdType::Generic,        "Generic"         },
  72         { SimdType::X86_Sse2,       "SSE2"            },
  73         { SimdType::X86_Sse4_1,     "SSE4.1"          },
  74         { SimdType::X86_Avx128Fma,  "AVX_128_FMA"     },
  75         { SimdType::X86_Avx,        "AVX_256"         },
  76         { SimdType::X86_Avx2,       "AVX2_256"        },
  77         { SimdType::X86_Avx512F,    "AVX_512F"        },
  78         { SimdType::X86_Avx512ER,   "AVX_512ER"       },
  79         { SimdType::X86_Mic,        "X86_MIC"         },
  80         { SimdType::Arm_Neon,       "ARM_NEON"        },
  81         { SimdType::Arm_NeonAsimd,  "ARM_NEON_ASIMD"  },
  82         { SimdType::Ibm_Qpx,        "IBM_QPX"         },
  83         { SimdType::Ibm_Vmx,        "IBM_VMX"         },
  84         { SimdType::Ibm_Vsx,        "IBM_VSX"         },
  85         { SimdType::Fujitsu_HpcAce, "Fujitsu HPC-ACE" }
  86     };
  87
  88     return name.at(s);
  89 }
  90
  91 SimdType
  92 simdSuggested(const CpuInfo &c)
  93 {
  94     SimdType suggested = SimdType::None;
  95
  96     if (c.supportLevel() >= CpuInfo::SupportLevel::Features)
  97     {
  98         switch (c.vendor())
  99         {
 100             case CpuInfo::Vendor::Intel:
 101                 if (c.feature(CpuInfo::Feature::X86_Avx2))
 102                 {
 103                     suggested = SimdType::X86_Avx2;
 104                 }
 105                 else if (c.feature(CpuInfo::Feature::X86_Avx))
 106                 {
 107                     suggested = SimdType::X86_Avx;
 108                 }
 109                 else if (c.feature(CpuInfo::Feature::X86_Sse4_1))
 110                 {
 111                     suggested = SimdType::X86_Sse4_1;
 112                 }
 113                 else if (c.feature(CpuInfo::Feature::X86_Sse2))
 114                 {
 115                     suggested = SimdType::X86_Sse2;
 116                 }
 117                 break;
 118             case CpuInfo::Vendor::Amd:
 119                 if (c.feature(CpuInfo::Feature::X86_Avx2))
 120                 {
 121                     // When Amd starts supporting Avx2 we assume it will be 256 bits
 122                     suggested = SimdType::X86_Avx2;
 123                 }
 124                 else if (c.feature(CpuInfo::Feature::X86_Avx))
 125                 {
 126                     // For vanilla Avx, we should use the 128-bit FMA flavor in Amd
 127                     suggested = SimdType::X86_Avx128Fma;
 128                 }
 129                 else if (c.feature(CpuInfo::Feature::X86_Sse4_1))
 130                 {
 131                     suggested = SimdType::X86_Sse4_1;
 132                 }
 133                 else if (c.feature(CpuInfo::Feature::X86_Sse2))
 134                 {
 135                     suggested = SimdType::X86_Sse2;
 136                 }
 137
 138                 break;
 139             case CpuInfo::Vendor::Arm:
 140                 if (c.feature(CpuInfo::Feature::Arm_NeonAsimd))
 141                 {
 142                     suggested = SimdType::Arm_NeonAsimd;
 143                 }
 144                 else if (c.feature(CpuInfo::Feature::Arm_Neon))
 145                 {
 146                     suggested = SimdType::Arm_Neon;
 147                 }
 148                 break;
 149             case CpuInfo::Vendor::Ibm:
 150                 if (c.feature(CpuInfo::Feature::Ibm_Vsx))
 151                 {
 152                     suggested = SimdType::Ibm_Vsx;
 153                 }
 154                 else if (c.feature(CpuInfo::Feature::Ibm_Vsx))
 155                 {
 156                     suggested = SimdType::Ibm_Vsx;
 157                 }
 158                 else if (c.feature(CpuInfo::Feature::Ibm_Qpx))
 159                 {
 160                     suggested = SimdType::Ibm_Qpx;
 161                 }
 162                 break;
 163             case CpuInfo::Vendor::Fujitsu:
 164                 if (c.feature(CpuInfo::Feature::Fujitsu_HpcAce))
 165                 {
 166                     suggested = SimdType::Fujitsu_HpcAce;
 167                 }
 168                 break;
 169             default:
 170                 break;
 171         }
 172     }
 173     return suggested;
 174 }
 175
 176 SimdType
 177 simdCompiled()
 178 {
 179 #if GMX_SIMD_X86_AVX_512ER
 180     return SimdType::X86_Avx512ER;
 181 #elif GMX_SIMD_X86_AVX_512F
 182     return SimdType::X86_Avx512F;
 183 #elif GMX_SIMD_X86_MIC
 184     return SimdType::X86_Mic;
 185 #elif GMX_SIMD_X86_AVX2_256
 186     return SimdType::X86_Avx2;
 187 #elif GMX_SIMD_X86_AVX_256
 188     return SimdType::X86_Avx;
 189 #elif GMX_SIMD_X86_AVX_128_FMA
 190     return SimdType::X86_Avx128Fma;
 191 #elif GMX_SIMD_X86_SSE4_1
 192     return SimdType::X86_Sse4_1;
 193 #elif GMX_SIMD_X86_SSE2
 194     return SimdType::X86_Sse2;
 195 #elif GMX_SIMD_ARM_NEON
 196     return SimdType::Arm_Neon;
 197 #elif GMX_SIMD_ARM_NEON_ASIMD
 198     return SimdType::Arm_NeonAsimd;
 199 #elif GMX_SIMD_IBM_QPX
 200     return SimdType::Ibm_Qpx;
 201 #elif GMX_SIMD_IBM_VMX
 202     return SimdType::Ibm_Vmx;
 203 #elif GMX_SIMD_IBM_VSX
 204     return SimdType::Ibm_Vsx;
 205 #elif GMX_SIMD_SPARC64_HPC_ACE
 206     return SimdType::Fujitsu_HpcAce;
 207 #elif GMX_SIMD_REFERENCE
 208     return SimdType::Reference;
 209 #else
 210     return SimdType::None;
 211 #endif
 212 }
 213
 214 bool
 215 simdCheck(gmx::SimdType    wanted,
 216           FILE *           log,
 217           bool             warnToStdErr)
 218 {
 219     SimdType compiled = simdCompiled();
 220
 221     // Normally it is close to catastrophic if the compiled SIMD type is larger than
 222     // the supported one, but AVX128Fma is an exception: AMD CPUs will (strongly) prefer
 223     // AVX128Fma, but they will work fine with AVX too. Thus, make an exception for this.
 224     if (compiled > wanted && !(compiled == SimdType::X86_Avx && wanted == SimdType::X86_Avx128Fma))
 225     {
 226         fprintf(stderr, "Warning: SIMD instructions newer than hardware. Program will likely crash.\n"
 227                 "SIMD instructions most likely to fit this hardware: %s\n"
 228                 "SIMD instructions selected at GROMACS compile time: %s\n\n",
 229                 simdString(wanted).c_str(),
 230                 simdString(compiled).c_str());
 231     }
 232     else
 233     {
 234         // This warning will also occur if compiled is X86_Avx and wanted is X86_Avx128Fma
 235
 236         if (log != nullptr)
 237         {
 238             fprintf(log, "\nBinary not matching hardware - you might be losing performance.\n"
 239                     "SIMD instructions most likely to fit this hardware: %s\n"
 240                     "SIMD instructions selected at GROMACS compile time: %s\n\n",
 241                     simdString(wanted).c_str(),
 242                     simdString(compiled).c_str());
 243         }
 244         if (warnToStdErr)
 245         {
 246             fprintf(stderr, "Compiled SIMD instructions: %s, GROMACS could use %s on this machine, which is better.\n\n",
 247                     simdString(compiled).c_str(),
 248                     simdString(wanted).c_str());
 249         }
 250     }
 251     return (wanted == compiled);
 252 }
 253
 254 /*! \endcond */
 255
 256 }