src/gromacs/simd/support.cpp

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2015,2016, 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_Avx512,     "AVX_512"         },
  78         { SimdType::X86_Avx512Knl,  "AVX_512_KNL"     },
  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_Avx512ER))
 102                 {
 103                     suggested = SimdType::X86_Avx512Knl;
 104                 }
 105                 else if (c.feature(CpuInfo::Feature::X86_Avx512F))
 106                 {
 107                     suggested = SimdType::X86_Avx512;
 108                 }
 109                 else if (c.feature(CpuInfo::Feature::X86_Avx2))
 110                 {
 111                     suggested = SimdType::X86_Avx2;
 112                 }
 113                 else if (c.feature(CpuInfo::Feature::X86_Avx))
 114                 {
 115                     suggested = SimdType::X86_Avx;
 116                 }
 117                 else if (c.feature(CpuInfo::Feature::X86_Sse4_1))
 118                 {
 119                     suggested = SimdType::X86_Sse4_1;
 120                 }
 121                 else if (c.feature(CpuInfo::Feature::X86_Sse2))
 122                 {
 123                     suggested = SimdType::X86_Sse2;
 124                 }
 125                 break;
 126             case CpuInfo::Vendor::Amd:
 127                 if (c.feature(CpuInfo::Feature::X86_Avx2))
 128                 {
 129                     // When Amd starts supporting Avx2 we assume it will be 256 bits
 130                     suggested = SimdType::X86_Avx2;
 131                 }
 132                 else if (c.feature(CpuInfo::Feature::X86_Avx))
 133                 {
 134                     // Use 128-bit FMA SIMD if Fma4 flag is set, otherwise plain 256-bit AVX
 135                     if (c.feature(CpuInfo::Feature::X86_Fma4))
 136                     {
 137                         suggested = SimdType::X86_Avx128Fma;
 138                     }
 139                     else
 140                     {
 141                         suggested = SimdType::X86_Avx;
 142                     }
 143                 }
 144                 else if (c.feature(CpuInfo::Feature::X86_Sse4_1))
 145                 {
 146                     suggested = SimdType::X86_Sse4_1;
 147                 }
 148                 else if (c.feature(CpuInfo::Feature::X86_Sse2))
 149                 {
 150                     suggested = SimdType::X86_Sse2;
 151                 }
 152
 153                 break;
 154             case CpuInfo::Vendor::Arm:
 155                 if (c.feature(CpuInfo::Feature::Arm_NeonAsimd))
 156                 {
 157                     suggested = SimdType::Arm_NeonAsimd;
 158                 }
 159                 else if (c.feature(CpuInfo::Feature::Arm_Neon))
 160                 {
 161                     suggested = SimdType::Arm_Neon;
 162                 }
 163                 break;
 164             case CpuInfo::Vendor::Ibm:
 165                 if (c.feature(CpuInfo::Feature::Ibm_Vsx))
 166                 {
 167                     suggested = SimdType::Ibm_Vsx;
 168                 }
 169                 else if (c.feature(CpuInfo::Feature::Ibm_Vsx))
 170                 {
 171                     suggested = SimdType::Ibm_Vsx;
 172                 }
 173                 else if (c.feature(CpuInfo::Feature::Ibm_Qpx))
 174                 {
 175                     suggested = SimdType::Ibm_Qpx;
 176                 }
 177                 break;
 178             case CpuInfo::Vendor::Fujitsu:
 179                 if (c.feature(CpuInfo::Feature::Fujitsu_HpcAce))
 180                 {
 181                     suggested = SimdType::Fujitsu_HpcAce;
 182                 }
 183                 break;
 184             default:
 185                 break;
 186         }
 187     }
 188     return suggested;
 189 }
 190
 191 SimdType
 192 simdCompiled()
 193 {
 194 #if GMX_SIMD_X86_AVX_512_KNL
 195     return SimdType::X86_Avx512Knl;
 196 #elif GMX_SIMD_X86_AVX_512
 197     return SimdType::X86_Avx512;
 198 #elif GMX_SIMD_X86_MIC
 199     return SimdType::X86_Mic;
 200 #elif GMX_SIMD_X86_AVX2_256
 201     return SimdType::X86_Avx2;
 202 #elif GMX_SIMD_X86_AVX_256
 203     return SimdType::X86_Avx;
 204 #elif GMX_SIMD_X86_AVX_128_FMA
 205     return SimdType::X86_Avx128Fma;
 206 #elif GMX_SIMD_X86_SSE4_1
 207     return SimdType::X86_Sse4_1;
 208 #elif GMX_SIMD_X86_SSE2
 209     return SimdType::X86_Sse2;
 210 #elif GMX_SIMD_ARM_NEON
 211     return SimdType::Arm_Neon;
 212 #elif GMX_SIMD_ARM_NEON_ASIMD
 213     return SimdType::Arm_NeonAsimd;
 214 #elif GMX_SIMD_IBM_QPX
 215     return SimdType::Ibm_Qpx;
 216 #elif GMX_SIMD_IBM_VMX
 217     return SimdType::Ibm_Vmx;
 218 #elif GMX_SIMD_IBM_VSX
 219     return SimdType::Ibm_Vsx;
 220 #elif GMX_SIMD_SPARC64_HPC_ACE
 221     return SimdType::Fujitsu_HpcAce;
 222 #elif GMX_SIMD_REFERENCE
 223     return SimdType::Reference;
 224 #else
 225     return SimdType::None;
 226 #endif
 227 }
 228
 229 bool
 230 simdCheck(gmx::SimdType    wanted,
 231           FILE *           log,
 232           bool             warnToStdErr)
 233 {
 234     SimdType compiled = simdCompiled();
 235
 236     // Normally it is close to catastrophic if the compiled SIMD type is larger than
 237     // the supported one, but AVX128Fma is an exception: AMD CPUs will (strongly) prefer
 238     // AVX128Fma, but they will work fine with AVX too. Thus, make an exception for this.
 239     if (compiled > wanted && !(compiled == SimdType::X86_Avx && wanted == SimdType::X86_Avx128Fma))
 240     {
 241         fprintf(stderr, "Warning: SIMD instructions newer than hardware. Program will likely crash.\n"
 242                 "SIMD instructions most likely to fit this hardware: %s\n"
 243                 "SIMD instructions selected at GROMACS compile time: %s\n\n",
 244                 simdString(wanted).c_str(),
 245                 simdString(compiled).c_str());
 246     }
 247     else if (wanted != compiled)
 248     {
 249         // This warning will also occur if compiled is X86_Avx and wanted is X86_Avx128Fma
 250
 251         if (log != nullptr)
 252         {
 253             fprintf(log, "\nBinary not matching hardware - you might be losing performance.\n"
 254                     "SIMD instructions most likely to fit this hardware: %s\n"
 255                     "SIMD instructions selected at GROMACS compile time: %s\n\n",
 256                     simdString(wanted).c_str(),
 257                     simdString(compiled).c_str());
 258         }
 259         if (warnToStdErr)
 260         {
 261             fprintf(stderr, "Compiled SIMD instructions: %s, GROMACS could use %s on this machine, which is better.\n\n",
 262                     simdString(compiled).c_str(),
 263                     simdString(wanted).c_str());
 264         }
 265     }
 266     return (wanted == compiled);
 267 }
 268
 269 /*! \endcond */
 270
 271 }