src/gromacs/hardware/hardwareassign.cpp

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2016,2017, 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 #include "gmxpre.h"
  36
  37 #include "hardwareassign.h"
  38
  39 #include "config.h"
  40
  41 #include <cstring>
  42
  43 #include <algorithm>
  44 #include <string>
  45
  46 #include "gromacs/gmxlib/network.h"
  47 #include "gromacs/gpu_utils/gpu_utils.h"
  48 #include "gromacs/hardware/detecthardware.h"
  49 #include "gromacs/hardware/gpu_hw_info.h"
  50 #include "gromacs/mdtypes/commrec.h"
  51 #include "gromacs/utility/basenetwork.h"
  52 #include "gromacs/utility/exceptions.h"
  53 #include "gromacs/utility/fatalerror.h"
  54 #include "gromacs/utility/gmxassert.h"
  55 #include "gromacs/utility/logger.h"
  56 #include "gromacs/utility/smalloc.h"
  57 #include "gromacs/utility/stringutil.h"
  58 #include "gromacs/utility/sysinfo.h"
  59
  60 #define HOSTNAMELEN 80
  61
  62 /*! \internal \brief
  63  * This function is responsible for mapping the GPUs to the processes on a single node
  64  * (filling the gpu_opt->dev_use array).
  65  *
  66  * \param[in]        compatibleGpus       Vector of GPUs that are compatible
  67  * \param[in,out]    gpu_opt              Input/output GPU assignment data.
  68  * \param[in]        nrank                Number of PP GPU ranks on the node.
  69  * \param[in]        rank                 Index of PP GPU rank on the node.
  70  *
  71  * This selects the GPUs we will use. This is an operation local to each physical node.
  72  * If we have less MPI ranks than GPUs, we will waste some GPUs.
  73  */
  74 static void assign_rank_gpu_ids(const std::vector<int> &compatibleGpus,
  75                                 gmx_gpu_opt_t *gpu_opt, int nrank, int rank)
  76 {
  77     int numCompatibleGpus = static_cast<int>(compatibleGpus.size());
  78     GMX_RELEASE_ASSERT(gpu_opt, "Invalid gpu_opt pointer passed");
  79     GMX_RELEASE_ASSERT(nrank >= 1,
  80                        gmx::formatString("Invalid limit (%d) for the number of GPUs (detected %d compatible GPUs)",
  81                                          rank, numCompatibleGpus).c_str());
  82
  83     if (numCompatibleGpus == 0)
  84     {
  85         char host[HOSTNAMELEN];
  86
  87         gmx_gethostname(host, HOSTNAMELEN);
  88         gmx_fatal(FARGS, "A GPU was requested on host %s, but no compatible GPUs were detected. All nodes with PP ranks need to have GPUs. If you intended to use GPU acceleration in a parallel run, you can either avoid using the nodes that don't have GPUs or place PME ranks on these nodes.", host);
  89     }
  90
  91     int nshare;
  92
  93     nshare = 1;
  94     if (nrank > numCompatibleGpus)
  95     {
  96         if (nrank % numCompatibleGpus == 0)
  97         {
  98             nshare = gmx_gpu_sharing_supported() ? nrank/numCompatibleGpus : 1;
  99         }
 100         else
 101         {
 102             if (rank == 0)
 103             {
 104                 gmx_fatal(FARGS, "The number of MPI ranks (%d) in a physical node is not a multiple of the number of GPUs (%d). Select a different number of MPI ranks or use the -gpu_id option to manually specify the GPU to be used.",
 105                           nrank, numCompatibleGpus);
 106             }
 107
 108 #if GMX_MPI
 109             /* We use a global barrier to prevent ranks from continuing with
 110              * an invalid setup.
 111              */
 112             MPI_Barrier(MPI_COMM_WORLD);
 113 #endif
 114         }
 115     }
 116
 117     /* Here we will waste GPUs when nrank < numCompatibleGpus */
 118     gpu_opt->n_dev_use = std::min(numCompatibleGpus*nshare, nrank);
 119     if (!gmx_multiple_gpu_per_node_supported())
 120     {
 121         gpu_opt->n_dev_use = std::min(gpu_opt->n_dev_use, 1);
 122     }
 123     snew(gpu_opt->dev_use, gpu_opt->n_dev_use);
 124     for (int i = 0; i != gpu_opt->n_dev_use; ++i)
 125     {
 126         /* TODO: improve this implementation: either sort GPUs or remove the weakest here */
 127         gpu_opt->dev_use[i] = compatibleGpus[i/nshare];
 128     }
 129 }
 130
 131 /*! \brief Return whether all selected GPUs are compatible.
 132  *
 133  * Given the list of selected GPU device IDs in \c gpu_opt and
 134  * detected GPUs in \c gpu_info, return whether all selected GPUs are
 135  * compatible. If not, place a suitable string in \c errorMessage.
 136  *
 137  * \param[in]   gpu_info      Information about detected GPUs
 138  * \param[in]   gpu_opt       pointer to structure holding GPU options
 139  * \param[out]  errorMessage  pointer to string to hold a possible error message (is not updated when returning true)
 140  * \returns                   true if every requested GPU is compatible
 141  */
 142 static bool checkGpuSelection(const gmx_gpu_info_t &gpu_info,
 143                               const gmx_gpu_opt_t  *gpu_opt,
 144                               std::string          *errorMessage)
 145 {
 146     bool        allOK   = true;
 147     std::string message = "Some of the requested GPUs do not exist, behave strangely, or are not compatible:\n";
 148     for (int i = 0; i < gpu_opt->n_dev_use; i++)
 149     {
 150         GMX_ASSERT(gpu_opt, "Invalid gpu_opt");
 151         GMX_ASSERT(gpu_opt->dev_use, "Invalid gpu_opt->dev_use");
 152
 153         int id     = gpu_opt->dev_use[i];
 154         if (!isGpuCompatible(gpu_info, id))
 155         {
 156             allOK    = false;
 157             message += gmx::formatString("    GPU #%d: %s\n",
 158                                          id,
 159                                          getGpuCompatibilityDescription(gpu_info, id));
 160         }
 161     }
 162     if (!allOK && errorMessage)
 163     {
 164         *errorMessage = message;
 165     }
 166     return allOK;
 167 }
 168
 169 std::vector<int> getCompatibleGpus(const gmx_gpu_info_t &gpu_info)
 170 {
 171     // Possible minor over-allocation here, but not important for anything
 172     std::vector<int> compatibleGpus;
 173     compatibleGpus.reserve(gpu_info.n_dev);
 174     for (int i = 0; i < gpu_info.n_dev; i++)
 175     {
 176         GMX_ASSERT(gpu_info.gpu_dev, "Invalid gpu_info.gpu_dev");
 177         if (isGpuCompatible(gpu_info, i))
 178         {
 179             compatibleGpus.push_back(i);
 180         }
 181     }
 182     return compatibleGpus;
 183 }
 184
 185 void gmx_select_rank_gpu_ids(const t_commrec      *cr,
 186                              const gmx_gpu_info_t &gpu_info,
 187                              bool                  userSetGpuIds,
 188                              gmx_gpu_opt_t        *gpu_opt)
 189 {
 190     if (!(cr->duty & DUTY_PP))
 191     {
 192         /* Our rank is not doing PP, we don't use a GPU */
 193         return;
 194     }
 195
 196     if (userSetGpuIds)
 197     {
 198         /* Check the GPU IDs passed by the user.
 199          * (GPU IDs have been parsed by gmx_parse_gpu_ids before)
 200          */
 201         std::string errorMessage;
 202         if (!checkGpuSelection(gpu_info, gpu_opt, &errorMessage))
 203         {
 204             gmx_fatal(FARGS, errorMessage.c_str());
 205         }
 206     }
 207     else
 208     {
 209         auto compatibleGpus = getCompatibleGpus(gpu_info);
 210         assign_rank_gpu_ids(compatibleGpus, gpu_opt, cr->nrank_pp_intranode, cr->rank_pp_intranode);
 211     }
 212 }