src/gromacs/mdrun/runner.cpp

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 1991-2000, University of Groningen, The Netherlands.
   5  * Copyright (c) 2001-2004, The GROMACS development team.
   6  * Copyright (c) 2011,2012,2013,2014,2015,2016,2017,2018,2019, by the GROMACS development team, led by
   7  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   8  * and including many others, as listed in the AUTHORS file in the
   9  * top-level source directory and at http://www.gromacs.org.
  10  *
  11  * GROMACS is free software; you can redistribute it and/or
  12  * modify it under the terms of the GNU Lesser General Public License
  13  * as published by the Free Software Foundation; either version 2.1
  14  * of the License, or (at your option) any later version.
  15  *
  16  * GROMACS is distributed in the hope that it will be useful,
  17  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  18  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  19  * Lesser General Public License for more details.
  20  *
  21  * You should have received a copy of the GNU Lesser General Public
  22  * License along with GROMACS; if not, see
  23  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  24  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  25  *
  26  * If you want to redistribute modifications to GROMACS, please
  27  * consider that scientific software is very special. Version
  28  * control is crucial - bugs must be traceable. We will be happy to
  29  * consider code for inclusion in the official distribution, but
  30  * derived work must not be called official GROMACS. Details are found
  31  * in the README & COPYING files - if they are missing, get the
  32  * official version at http://www.gromacs.org.
  33  *
  34  * To help us fund GROMACS development, we humbly ask that you cite
  35  * the research papers on the package. Check out http://www.gromacs.org.
  36  */
  37 /*! \internal \file
  38  *
  39  * \brief Implements the MD runner routine calling all integrators.
  40  *
  41  * \author David van der Spoel <david.vanderspoel@icm.uu.se>
  42  * \ingroup module_mdrun
  43  */
  44 #include "gmxpre.h"
  45
  46 #include "runner.h"
  47
  48 #include "config.h"
  49
  50 #include <cassert>
  51 #include <cinttypes>
  52 #include <csignal>
  53 #include <cstdlib>
  54 #include <cstring>
  55
  56 #include <algorithm>
  57 #include <memory>
  58
  59 #include "gromacs/commandline/filenm.h"
  60 #include "gromacs/domdec/builder.h"
  61 #include "gromacs/domdec/domdec.h"
  62 #include "gromacs/domdec/domdec_struct.h"
  63 #include "gromacs/domdec/gpuhaloexchange.h"
  64 #include "gromacs/domdec/localatomsetmanager.h"
  65 #include "gromacs/domdec/partition.h"
  66 #include "gromacs/ewald/ewald_utils.h"
  67 #include "gromacs/ewald/pme.h"
  68 #include "gromacs/ewald/pme_gpu_program.h"
  69 #include "gromacs/ewald/pme_pp_comm_gpu.h"
  70 #include "gromacs/fileio/checkpoint.h"
  71 #include "gromacs/fileio/gmxfio.h"
  72 #include "gromacs/fileio/oenv.h"
  73 #include "gromacs/fileio/tpxio.h"
  74 #include "gromacs/gmxlib/network.h"
  75 #include "gromacs/gmxlib/nrnb.h"
  76 #include "gromacs/gpu_utils/gpu_utils.h"
  77 #include "gromacs/hardware/cpuinfo.h"
  78 #include "gromacs/hardware/detecthardware.h"
  79 #include "gromacs/hardware/printhardware.h"
  80 #include "gromacs/imd/imd.h"
  81 #include "gromacs/listed_forces/disre.h"
  82 #include "gromacs/listed_forces/gpubonded.h"
  83 #include "gromacs/listed_forces/orires.h"
  84 #include "gromacs/math/functions.h"
  85 #include "gromacs/math/utilities.h"
  86 #include "gromacs/math/vec.h"
  87 #include "gromacs/mdlib/boxdeformation.h"
  88 #include "gromacs/mdlib/broadcaststructs.h"
  89 #include "gromacs/mdlib/calc_verletbuf.h"
  90 #include "gromacs/mdlib/dispersioncorrection.h"
  91 #include "gromacs/mdlib/enerdata_utils.h"
  92 #include "gromacs/mdlib/force.h"
  93 #include "gromacs/mdlib/forcerec.h"
  94 #include "gromacs/mdlib/gmx_omp_nthreads.h"
  95 #include "gromacs/mdlib/makeconstraints.h"
  96 #include "gromacs/mdlib/md_support.h"
  97 #include "gromacs/mdlib/mdatoms.h"
  98 #include "gromacs/mdlib/membed.h"
  99 #include "gromacs/mdlib/qmmm.h"
 100 #include "gromacs/mdlib/sighandler.h"
 101 #include "gromacs/mdlib/stophandler.h"
 102 #include "gromacs/mdrun/mdmodules.h"
 103 #include "gromacs/mdrun/simulationcontext.h"
 104 #include "gromacs/mdrunutility/handlerestart.h"
 105 #include "gromacs/mdrunutility/logging.h"
 106 #include "gromacs/mdrunutility/multisim.h"
 107 #include "gromacs/mdrunutility/printtime.h"
 108 #include "gromacs/mdrunutility/threadaffinity.h"
 109 #include "gromacs/mdtypes/commrec.h"
 110 #include "gromacs/mdtypes/enerdata.h"
 111 #include "gromacs/mdtypes/fcdata.h"
 112 #include "gromacs/mdtypes/group.h"
 113 #include "gromacs/mdtypes/inputrec.h"
 114 #include "gromacs/mdtypes/md_enums.h"
 115 #include "gromacs/mdtypes/mdrunoptions.h"
 116 #include "gromacs/mdtypes/observableshistory.h"
 117 #include "gromacs/mdtypes/simulation_workload.h"
 118 #include "gromacs/mdtypes/state.h"
 119 #include "gromacs/mdtypes/state_propagator_data_gpu.h"
 120 #include "gromacs/nbnxm/gpu_data_mgmt.h"
 121 #include "gromacs/nbnxm/nbnxm.h"
 122 #include "gromacs/nbnxm/pairlist_tuning.h"
 123 #include "gromacs/pbcutil/pbc.h"
 124 #include "gromacs/pulling/output.h"
 125 #include "gromacs/pulling/pull.h"
 126 #include "gromacs/pulling/pull_rotation.h"
 127 #include "gromacs/restraint/manager.h"
 128 #include "gromacs/restraint/restraintmdmodule.h"
 129 #include "gromacs/restraint/restraintpotential.h"
 130 #include "gromacs/swap/swapcoords.h"
 131 #include "gromacs/taskassignment/decidegpuusage.h"
 132 #include "gromacs/taskassignment/decidesimulationworkload.h"
 133 #include "gromacs/taskassignment/resourcedivision.h"
 134 #include "gromacs/taskassignment/taskassignment.h"
 135 #include "gromacs/taskassignment/usergpuids.h"
 136 #include "gromacs/timing/gpu_timing.h"
 137 #include "gromacs/timing/wallcycle.h"
 138 #include "gromacs/timing/wallcyclereporting.h"
 139 #include "gromacs/topology/mtop_util.h"
 140 #include "gromacs/trajectory/trajectoryframe.h"
 141 #include "gromacs/utility/basenetwork.h"
 142 #include "gromacs/utility/cstringutil.h"
 143 #include "gromacs/utility/exceptions.h"
 144 #include "gromacs/utility/fatalerror.h"
 145 #include "gromacs/utility/filestream.h"
 146 #include "gromacs/utility/gmxassert.h"
 147 #include "gromacs/utility/gmxmpi.h"
 148 #include "gromacs/utility/keyvaluetree.h"
 149 #include "gromacs/utility/logger.h"
 150 #include "gromacs/utility/loggerbuilder.h"
 151 #include "gromacs/utility/mdmodulenotification.h"
 152 #include "gromacs/utility/physicalnodecommunicator.h"
 153 #include "gromacs/utility/pleasecite.h"
 154 #include "gromacs/utility/programcontext.h"
 155 #include "gromacs/utility/smalloc.h"
 156 #include "gromacs/utility/stringutil.h"
 157
 158 #include "isimulator.h"
 159 #include "replicaexchange.h"
 160 #include "simulatorbuilder.h"
 161
 162 #if GMX_FAHCORE
 163 #include "corewrap.h"
 164 #endif
 165
 166 namespace gmx
 167 {
 168
 169 /*! \brief Structure that holds boolean flags corresponding to the development
 170  *        features present enabled through environment variables.
 171  *
 172  */
 173 struct DevelopmentFeatureFlags
 174 {
 175     //! True if the Buffer ops development feature is enabled
 176     // TODO: when the trigger of the buffer ops offload is fully automated this should go away
 177     bool enableGpuBufferOps      = false;
 178     //! If true, forces 'mdrun -update auto' default to 'gpu'
 179     bool forceGpuUpdateDefaultOn = false;
 180     //! True if the GPU halo exchange development feature is enabled
 181     bool enableGpuHaloExchange   = false;
 182     //! True if the PME PP direct communication GPU development feature is enabled
 183     bool enableGpuPmePPComm      = false;
 184 };
 185
 186 /*! \brief Manage any development feature flag variables encountered
 187  *
 188  * The use of dev features indicated by environment variables is
 189  * logged in order to ensure that runs with such features enabled can
 190  * be identified from their log and standard output. Any cross
 191  * dependencies are also checked, and if unsatisfied, a fatal error
 192  * issued.
 193  *
 194  * Note that some development features overrides are applied already here:
 195  * the GPU communication flags are set to false in non-tMPI and non-CUDA builds.
 196  *
 197  * \param[in]  mdlog                Logger object.
 198  * \param[in]  useGpuForNonbonded   True if the nonbonded task is offloaded in this run.
 199  * \param[in]  useGpuForPme         True if the PME task is offloaded in this run.
 200  * \returns                         The object populated with development feature flags.
 201  */
 202 static DevelopmentFeatureFlags manageDevelopmentFeatures(const gmx::MDLogger &mdlog,
 203                                                          const bool           useGpuForNonbonded,
 204                                                          const bool           useGpuForPme)
 205 {
 206     DevelopmentFeatureFlags devFlags;
 207
 208     // Some builds of GCC 5 give false positive warnings that these
 209     // getenv results are ignored when clearly they are used.
 210 #pragma GCC diagnostic push
 211 #pragma GCC diagnostic ignored "-Wunused-result"
 212     devFlags.enableGpuBufferOps      = (getenv("GMX_USE_GPU_BUFFER_OPS") != nullptr) && (GMX_GPU == GMX_GPU_CUDA) && useGpuForNonbonded;
 213     devFlags.forceGpuUpdateDefaultOn = (getenv("GMX_FORCE_UPDATE_DEFAULT_GPU") != nullptr);
 214     devFlags.enableGpuHaloExchange   = (getenv("GMX_GPU_DD_COMMS") != nullptr && GMX_THREAD_MPI && (GMX_GPU == GMX_GPU_CUDA));
 215     devFlags.enableGpuPmePPComm      = (getenv("GMX_GPU_PME_PP_COMMS") != nullptr && GMX_THREAD_MPI && (GMX_GPU == GMX_GPU_CUDA));
 216 #pragma GCC diagnostic pop
 217
 218     if (devFlags.enableGpuBufferOps)
 219     {
 220         GMX_LOG(mdlog.warning).asParagraph().appendTextFormatted(
 221                 "NOTE: This run uses the 'GPU buffer ops' feature, enabled by the GMX_USE_GPU_BUFFER_OPS environment variable.");
 222     }
 223
 224     if (devFlags.enableGpuHaloExchange)
 225     {
 226         if (useGpuForNonbonded)
 227         {
 228             if (!devFlags.enableGpuBufferOps)
 229             {
 230                 gmx_fatal(FARGS, "Cannot enable GPU halo exchange without GPU buffer operations, set GMX_USE_GPU_BUFFER_OPS=1\n");
 231             }
 232             GMX_LOG(mdlog.warning).asParagraph().appendTextFormatted(
 233                     "NOTE: This run uses the 'GPU halo exchange' feature, enabled by the GMX_GPU_DD_COMMS environment variable.");
 234         }
 235         else
 236         {
 237             GMX_LOG(mdlog.warning).asParagraph().appendTextFormatted(
 238                     "NOTE: GMX_GPU_DD_COMMS environment variable detected, but the 'GPU halo exchange' feature will not be enabled as nonbonded interactions are not offloaded.");
 239             devFlags.enableGpuHaloExchange = false;
 240         }
 241     }
 242
 243     if (devFlags.forceGpuUpdateDefaultOn)
 244     {
 245         GMX_LOG(mdlog.warning).asParagraph().appendTextFormatted(
 246                 "NOTE: This run will default to '-update gpu' as requested by the GMX_FORCE_UPDATE_DEFAULT_GPU environment variable.");
 247     }
 248
 249     if (devFlags.enableGpuPmePPComm)
 250     {
 251         if (useGpuForPme)
 252         {
 253             GMX_LOG(mdlog.warning).asParagraph().appendTextFormatted(
 254                     "NOTE: This run uses the 'GPU PME-PP communications' feature, enabled by the GMX_GPU_PME_PP_COMMS environment variable.");
 255         }
 256         else
 257         {
 258             GMX_LOG(mdlog.warning).asParagraph().appendTextFormatted(
 259                     "NOTE: GMX_GPU_PME_PP_COMMS environment variable detected, but the 'GPU PME-PP communications' feature was not enabled as PME is not offloaded to the GPU.");
 260             devFlags.enableGpuPmePPComm = false;
 261         }
 262     }
 263
 264     return devFlags;
 265 }
 266
 267 /*! \brief Barrier for safe simultaneous thread access to mdrunner data
 268  *
 269  * Used to ensure that the master thread does not modify mdrunner during copy
 270  * on the spawned threads. */
 271 static void threadMpiMdrunnerAccessBarrier()
 272 {
 273 #if GMX_THREAD_MPI
 274     MPI_Barrier(MPI_COMM_WORLD);
 275 #endif
 276 }
 277
 278 Mdrunner Mdrunner::cloneOnSpawnedThread() const
 279 {
 280     auto newRunner = Mdrunner(std::make_unique<MDModules>());
 281
 282     // All runners in the same process share a restraint manager resource because it is
 283     // part of the interface to the client code, which is associated only with the
 284     // original thread. Handles to the same resources can be obtained by copy.
 285     {
 286         newRunner.restraintManager_ = std::make_unique<RestraintManager>(*restraintManager_);
 287     }
 288
 289     // Copy members of master runner.
 290     // \todo Replace with builder when Simulation context and/or runner phases are better defined.
 291     // Ref https://redmine.gromacs.org/issues/2587 and https://redmine.gromacs.org/issues/2375
 292     newRunner.hw_opt    = hw_opt;
 293     newRunner.filenames = filenames;
 294
 295     newRunner.oenv                = oenv;
 296     newRunner.mdrunOptions        = mdrunOptions;
 297     newRunner.domdecOptions       = domdecOptions;
 298     newRunner.nbpu_opt            = nbpu_opt;
 299     newRunner.pme_opt             = pme_opt;
 300     newRunner.pme_fft_opt         = pme_fft_opt;
 301     newRunner.bonded_opt          = bonded_opt;
 302     newRunner.update_opt          = update_opt;
 303     newRunner.nstlist_cmdline     = nstlist_cmdline;
 304     newRunner.replExParams        = replExParams;
 305     newRunner.pforce              = pforce;
 306     // Give the spawned thread the newly created valid communicator
 307     // for the simulation.
 308     newRunner.communicator        = MPI_COMM_WORLD;
 309     newRunner.ms                  = ms;
 310     newRunner.startingBehavior    = startingBehavior;
 311     newRunner.stopHandlerBuilder_ = std::make_unique<StopHandlerBuilder>(*stopHandlerBuilder_);
 312
 313     threadMpiMdrunnerAccessBarrier();
 314
 315     return newRunner;
 316 }
 317
 318 /*! \brief The callback used for running on spawned threads.
 319  *
 320  * Obtains the pointer to the master mdrunner object from the one
 321  * argument permitted to the thread-launch API call, copies it to make
 322  * a new runner for this thread, reinitializes necessary data, and
 323  * proceeds to the simulation. */
 324 static void mdrunner_start_fn(const void *arg)
 325 {
 326     try
 327     {
 328         auto masterMdrunner = reinterpret_cast<const gmx::Mdrunner *>(arg);
 329         /* copy the arg list to make sure that it's thread-local. This
 330            doesn't copy pointed-to items, of course; fnm, cr and fplog
 331            are reset in the call below, all others should be const. */
 332         gmx::Mdrunner mdrunner = masterMdrunner->cloneOnSpawnedThread();
 333         mdrunner.mdrunner();
 334     }
 335     GMX_CATCH_ALL_AND_EXIT_WITH_FATAL_ERROR;
 336 }
 337
 338
 339 void Mdrunner::spawnThreads(int numThreadsToLaunch)
 340 {
 341 #if GMX_THREAD_MPI
 342     /* now spawn new threads that start mdrunner_start_fn(), while
 343        the main thread returns. Thread affinity is handled later. */
 344     if (tMPI_Init_fn(TRUE, numThreadsToLaunch, TMPI_AFFINITY_NONE,
 345                      mdrunner_start_fn, static_cast<const void*>(this)) != TMPI_SUCCESS)
 346     {
 347         GMX_THROW(gmx::InternalError("Failed to spawn thread-MPI threads"));
 348     }
 349
 350     // Give the master thread the newly created valid communicator for
 351     // the simulation.
 352     communicator = MPI_COMM_WORLD;
 353     threadMpiMdrunnerAccessBarrier();
 354 #else
 355     GMX_UNUSED_VALUE(numThreadsToLaunch);
 356     GMX_UNUSED_VALUE(mdrunner_start_fn);
 357 #endif
 358 }
 359
 360 }  // namespace gmx
 361
 362 /*! \brief Initialize variables for Verlet scheme simulation */
 363 static void prepare_verlet_scheme(FILE                           *fplog,
 364                                   t_commrec                      *cr,
 365                                   t_inputrec                     *ir,
 366                                   int                             nstlist_cmdline,
 367                                   const gmx_mtop_t               *mtop,
 368                                   const matrix                    box,
 369                                   bool                            makeGpuPairList,
 370                                   const gmx::CpuInfo             &cpuinfo)
 371 {
 372     /* For NVE simulations, we will retain the initial list buffer */
 373     if (EI_DYNAMICS(ir->eI) &&
 374         ir->verletbuf_tol > 0 &&
 375         !(EI_MD(ir->eI) && ir->etc == etcNO))
 376     {
 377         /* Update the Verlet buffer size for the current run setup */
 378
 379         /* Here we assume SIMD-enabled kernels are being used. But as currently
 380          * calc_verlet_buffer_size gives the same results for 4x8 and 4x4
 381          * and 4x2 gives a larger buffer than 4x4, this is ok.
 382          */
 383         ListSetupType      listType  = (makeGpuPairList ? ListSetupType::Gpu : ListSetupType::CpuSimdWhenSupported);
 384         VerletbufListSetup listSetup = verletbufGetSafeListSetup(listType);
 385
 386         const real         rlist_new =
 387             calcVerletBufferSize(*mtop, det(box), *ir, ir->nstlist, ir->nstlist - 1, -1, listSetup);
 388
 389         if (rlist_new != ir->rlist)
 390         {
 391             if (fplog != nullptr)
 392             {
 393                 fprintf(fplog, "\nChanging rlist from %g to %g for non-bonded %dx%d atom kernels\n\n",
 394                         ir->rlist, rlist_new,
 395                         listSetup.cluster_size_i, listSetup.cluster_size_j);
 396             }
 397             ir->rlist     = rlist_new;
 398         }
 399     }
 400
 401     if (nstlist_cmdline > 0 && (!EI_DYNAMICS(ir->eI) || ir->verletbuf_tol <= 0))
 402     {
 403         gmx_fatal(FARGS, "Can not set nstlist without %s",
 404                   !EI_DYNAMICS(ir->eI) ? "dynamics" : "verlet-buffer-tolerance");
 405     }
 406
 407     if (EI_DYNAMICS(ir->eI))
 408     {
 409         /* Set or try nstlist values */
 410         increaseNstlist(fplog, cr, ir, nstlist_cmdline, mtop, box, makeGpuPairList, cpuinfo);
 411     }
 412 }
 413
 414 /*! \brief Override the nslist value in inputrec
 415  *
 416  * with value passed on the command line (if any)
 417  */
 418 static void override_nsteps_cmdline(const gmx::MDLogger &mdlog,
 419                                     int64_t              nsteps_cmdline,
 420                                     t_inputrec          *ir)
 421 {
 422     assert(ir);
 423
 424     /* override with anything else than the default -2 */
 425     if (nsteps_cmdline > -2)
 426     {
 427         char sbuf_steps[STEPSTRSIZE];
 428         char sbuf_msg[STRLEN];
 429
 430         ir->nsteps = nsteps_cmdline;
 431         if (EI_DYNAMICS(ir->eI) && nsteps_cmdline != -1)
 432         {
 433             sprintf(sbuf_msg, "Overriding nsteps with value passed on the command line: %s steps, %.3g ps",
 434                     gmx_step_str(nsteps_cmdline, sbuf_steps),
 435                     fabs(nsteps_cmdline*ir->delta_t));
 436         }
 437         else
 438         {
 439             sprintf(sbuf_msg, "Overriding nsteps with value passed on the command line: %s steps",
 440                     gmx_step_str(nsteps_cmdline, sbuf_steps));
 441         }
 442
 443         GMX_LOG(mdlog.warning).asParagraph().appendText(sbuf_msg);
 444     }
 445     else if (nsteps_cmdline < -2)
 446     {
 447         gmx_fatal(FARGS, "Invalid nsteps value passed on the command line: %" PRId64,
 448                   nsteps_cmdline);
 449     }
 450     /* Do nothing if nsteps_cmdline == -2 */
 451 }
 452
 453 namespace gmx
 454 {
 455
 456 /*! \brief Return whether GPU acceleration of nonbondeds is supported with the given settings.
 457  *
 458  * If not, and if a warning may be issued, logs a warning about
 459  * falling back to CPU code. With thread-MPI, only the first
 460  * call to this function should have \c issueWarning true. */
 461 static bool gpuAccelerationOfNonbondedIsUseful(const MDLogger   &mdlog,
 462                                                const t_inputrec &ir,
 463                                                bool              issueWarning)
 464 {
 465     bool        gpuIsUseful = true;
 466     std::string warning;
 467
 468     if (ir.opts.ngener - ir.nwall > 1)
 469     {
 470         /* The GPU code does not support more than one energy group.
 471          * If the user requested GPUs explicitly, a fatal error is given later.
 472          */
 473         gpuIsUseful = false;
 474         warning     =
 475             "Multiple energy groups is not implemented for GPUs, falling back to the CPU. "
 476             "For better performance, run on the GPU without energy groups and then do "
 477             "gmx mdrun -rerun option on the trajectory with an energy group .tpr file.";
 478     }
 479
 480     if (EI_TPI(ir.eI))
 481     {
 482         gpuIsUseful = false;
 483         warning     = "TPI is not implemented for GPUs.";
 484     }
 485
 486     if (!gpuIsUseful && issueWarning)
 487     {
 488         GMX_LOG(mdlog.warning).asParagraph().appendText(warning);
 489     }
 490
 491     return gpuIsUseful;
 492 }
 493
 494 //! Initializes the logger for mdrun.
 495 static gmx::LoggerOwner buildLogger(FILE      *fplog,
 496                                     const bool isSimulationMasterRank)
 497 {
 498     gmx::LoggerBuilder builder;
 499     if (fplog != nullptr)
 500     {
 501         builder.addTargetFile(gmx::MDLogger::LogLevel::Info, fplog);
 502     }
 503     if (isSimulationMasterRank)
 504     {
 505         builder.addTargetStream(gmx::MDLogger::LogLevel::Warning,
 506                                 &gmx::TextOutputFile::standardError());
 507     }
 508     return builder.build();
 509 }
 510
 511 //! Make a TaskTarget from an mdrun argument string.
 512 static TaskTarget findTaskTarget(const char *optionString)
 513 {
 514     TaskTarget returnValue = TaskTarget::Auto;
 515
 516     if (strncmp(optionString, "auto", 3) == 0)
 517     {
 518         returnValue = TaskTarget::Auto;
 519     }
 520     else if (strncmp(optionString, "cpu", 3) == 0)
 521     {
 522         returnValue = TaskTarget::Cpu;
 523     }
 524     else if (strncmp(optionString, "gpu", 3) == 0)
 525     {
 526         returnValue = TaskTarget::Gpu;
 527     }
 528     else
 529     {
 530         GMX_ASSERT(false, "Option string should have been checked for sanity already");
 531     }
 532
 533     return returnValue;
 534 }
 535
 536 //! Finish run, aggregate data to print performance info.
 537 static void finish_run(FILE *fplog,
 538                        const gmx::MDLogger &mdlog,
 539                        const t_commrec *cr,
 540                        const t_inputrec *inputrec,
 541                        t_nrnb nrnb[], gmx_wallcycle_t wcycle,
 542                        gmx_walltime_accounting_t walltime_accounting,
 543                        nonbonded_verlet_t *nbv,
 544                        const gmx_pme_t *pme,
 545                        gmx_bool bWriteStat)
 546 {
 547     double  delta_t  = 0;
 548     double  nbfs     = 0, mflop = 0;
 549     double  elapsed_time,
 550             elapsed_time_over_all_ranks,
 551             elapsed_time_over_all_threads,
 552             elapsed_time_over_all_threads_over_all_ranks;
 553     /* Control whether it is valid to print a report. Only the
 554        simulation master may print, but it should not do so if the run
 555        terminated e.g. before a scheduled reset step. This is
 556        complicated by the fact that PME ranks are unaware of the
 557        reason why they were sent a pmerecvqxFINISH. To avoid
 558        communication deadlocks, we always do the communication for the
 559        report, even if we've decided not to write the report, because
 560        how long it takes to finish the run is not important when we've
 561        decided not to report on the simulation performance.
 562
 563        Further, we only report performance for dynamical integrators,
 564        because those are the only ones for which we plan to
 565        consider doing any optimizations. */
 566     bool printReport = EI_DYNAMICS(inputrec->eI) && SIMMASTER(cr);
 567
 568     if (printReport && !walltime_accounting_get_valid_finish(walltime_accounting))
 569     {
 570         GMX_LOG(mdlog.warning).asParagraph().appendText("Simulation ended prematurely, no performance report will be written.");
 571         printReport = false;
 572     }
 573
 574     t_nrnb                  *nrnb_tot;
 575     std::unique_ptr<t_nrnb>  nrnbTotalStorage;
 576     if (cr->nnodes > 1)
 577     {
 578         nrnbTotalStorage = std::make_unique<t_nrnb>();
 579         nrnb_tot         = nrnbTotalStorage.get();
 580 #if GMX_MPI
 581         MPI_Allreduce(nrnb->n, nrnb_tot->n, eNRNB, MPI_DOUBLE, MPI_SUM,
 582                       cr->mpi_comm_mysim);
 583 #endif
 584     }
 585     else
 586     {
 587         nrnb_tot = nrnb;
 588     }
 589
 590     elapsed_time                  = walltime_accounting_get_time_since_reset(walltime_accounting);
 591     elapsed_time_over_all_threads = walltime_accounting_get_time_since_reset_over_all_threads(walltime_accounting);
 592     if (cr->nnodes > 1)
 593     {
 594 #if GMX_MPI
 595         /* reduce elapsed_time over all MPI ranks in the current simulation */
 596         MPI_Allreduce(&elapsed_time,
 597                       &elapsed_time_over_all_ranks,
 598                       1, MPI_DOUBLE, MPI_SUM,
 599                       cr->mpi_comm_mysim);
 600         elapsed_time_over_all_ranks /= cr->nnodes;
 601         /* Reduce elapsed_time_over_all_threads over all MPI ranks in the
 602          * current simulation. */
 603         MPI_Allreduce(&elapsed_time_over_all_threads,
 604                       &elapsed_time_over_all_threads_over_all_ranks,
 605                       1, MPI_DOUBLE, MPI_SUM,
 606                       cr->mpi_comm_mysim);
 607 #endif
 608     }
 609     else
 610     {
 611         elapsed_time_over_all_ranks                  = elapsed_time;
 612         elapsed_time_over_all_threads_over_all_ranks = elapsed_time_over_all_threads;
 613     }
 614
 615     if (printReport)
 616     {
 617         print_flop(fplog, nrnb_tot, &nbfs, &mflop);
 618     }
 619
 620     if (thisRankHasDuty(cr, DUTY_PP) && DOMAINDECOMP(cr))
 621     {
 622         print_dd_statistics(cr, inputrec, fplog);
 623     }
 624
 625     /* TODO Move the responsibility for any scaling by thread counts
 626      * to the code that handled the thread region, so that there's a
 627      * mechanism to keep cycle counting working during the transition
 628      * to task parallelism. */
 629     int nthreads_pp  = gmx_omp_nthreads_get(emntNonbonded);
 630     int nthreads_pme = gmx_omp_nthreads_get(emntPME);
 631     wallcycle_scale_by_num_threads(wcycle, thisRankHasDuty(cr, DUTY_PME) && !thisRankHasDuty(cr, DUTY_PP), nthreads_pp, nthreads_pme);
 632     auto cycle_sum(wallcycle_sum(cr, wcycle));
 633
 634     if (printReport)
 635     {
 636         auto                    nbnxn_gpu_timings = (nbv != nullptr && nbv->useGpu()) ? Nbnxm::gpu_get_timings(nbv->gpu_nbv) : nullptr;
 637         gmx_wallclock_gpu_pme_t pme_gpu_timings   = {};
 638
 639         if (pme_gpu_task_enabled(pme))
 640         {
 641             pme_gpu_get_timings(pme, &pme_gpu_timings);
 642         }
 643         wallcycle_print(fplog, mdlog, cr->nnodes, cr->npmenodes, nthreads_pp, nthreads_pme,
 644                         elapsed_time_over_all_ranks,
 645                         wcycle, cycle_sum,
 646                         nbnxn_gpu_timings,
 647                         &pme_gpu_timings);
 648
 649         if (EI_DYNAMICS(inputrec->eI))
 650         {
 651             delta_t = inputrec->delta_t;
 652         }
 653
 654         if (fplog)
 655         {
 656             print_perf(fplog, elapsed_time_over_all_threads_over_all_ranks,
 657                        elapsed_time_over_all_ranks,
 658                        walltime_accounting_get_nsteps_done_since_reset(walltime_accounting),
 659                        delta_t, nbfs, mflop);
 660         }
 661         if (bWriteStat)
 662         {
 663             print_perf(stderr, elapsed_time_over_all_threads_over_all_ranks,
 664                        elapsed_time_over_all_ranks,
 665                        walltime_accounting_get_nsteps_done_since_reset(walltime_accounting),
 666                        delta_t, nbfs, mflop);
 667         }
 668     }
 669 }
 670
 671 int Mdrunner::mdrunner()
 672 {
 673     matrix                    box;
 674     t_forcerec               *fr               = nullptr;
 675     t_fcdata                 *fcd              = nullptr;
 676     real                      ewaldcoeff_q     = 0;
 677     real                      ewaldcoeff_lj    = 0;
 678     int                       nChargePerturbed = -1, nTypePerturbed = 0;
 679     gmx_wallcycle_t           wcycle;
 680     gmx_walltime_accounting_t walltime_accounting = nullptr;
 681     gmx_membed_t *            membed              = nullptr;
 682     gmx_hw_info_t            *hwinfo              = nullptr;
 683
 684     /* CAUTION: threads may be started later on in this function, so
 685        cr doesn't reflect the final parallel state right now */
 686     gmx_mtop_t                      mtop;
 687
 688     /* TODO: inputrec should tell us whether we use an algorithm, not a file option */
 689     const bool doEssentialDynamics = opt2bSet("-ei", filenames.size(), filenames.data());
 690     const bool doMembed            = opt2bSet("-membed", filenames.size(), filenames.data());
 691     const bool doRerun             = mdrunOptions.rerun;
 692
 693     // Handle task-assignment related user options.
 694     EmulateGpuNonbonded emulateGpuNonbonded = (getenv("GMX_EMULATE_GPU") != nullptr ?
 695                                                EmulateGpuNonbonded::Yes : EmulateGpuNonbonded::No);
 696
 697     std::vector<int> userGpuTaskAssignment;
 698     try
 699     {
 700         userGpuTaskAssignment = parseUserTaskAssignmentString(hw_opt.userGpuTaskAssignment);
 701     }
 702     GMX_CATCH_ALL_AND_EXIT_WITH_FATAL_ERROR;
 703     auto       nonbondedTarget = findTaskTarget(nbpu_opt);
 704     auto       pmeTarget       = findTaskTarget(pme_opt);
 705     auto       pmeFftTarget    = findTaskTarget(pme_fft_opt);
 706     auto       bondedTarget    = findTaskTarget(bonded_opt);
 707     auto       updateTarget    = findTaskTarget(update_opt);
 708     PmeRunMode pmeRunMode      = PmeRunMode::None;
 709
 710     FILE      *fplog = nullptr;
 711     // If we are appending, we don't write log output because we need
 712     // to check that the old log file matches what the checkpoint file
 713     // expects. Otherwise, we should start to write log output now if
 714     // there is a file ready for it.
 715     if (logFileHandle != nullptr && startingBehavior != StartingBehavior::RestartWithAppending)
 716     {
 717         fplog = gmx_fio_getfp(logFileHandle);
 718     }
 719     const bool       isSimulationMasterRank = findIsSimulationMasterRank(ms, communicator);
 720     gmx::LoggerOwner logOwner(buildLogger(fplog, isSimulationMasterRank));
 721     gmx::MDLogger    mdlog(logOwner.logger());
 722
 723     // TODO The thread-MPI master rank makes a working
 724     // PhysicalNodeCommunicator here, but it gets rebuilt by all ranks
 725     // after the threads have been launched. This works because no use
 726     // is made of that communicator until after the execution paths
 727     // have rejoined. But it is likely that we can improve the way
 728     // this is expressed, e.g. by expressly running detection only the
 729     // master rank for thread-MPI, rather than relying on the mutex
 730     // and reference count.
 731     PhysicalNodeCommunicator physicalNodeComm(communicator, gmx_physicalnode_id_hash());
 732     hwinfo = gmx_detect_hardware(mdlog, physicalNodeComm);
 733
 734     gmx_print_detected_hardware(fplog, isSimulationMasterRank && isMasterSim(ms), mdlog, hwinfo);
 735
 736     std::vector<int> gpuIdsToUse = makeGpuIdsToUse(hwinfo->gpu_info, hw_opt.gpuIdsAvailable);
 737
 738     // Print citation requests after all software/hardware printing
 739     pleaseCiteGromacs(fplog);
 740
 741     // TODO Replace this by unique_ptr once t_inputrec is C++
 742     t_inputrec               inputrecInstance;
 743     t_inputrec              *inputrec = nullptr;
 744     std::unique_ptr<t_state> globalState;
 745
 746     auto                     partialDeserializedTpr = std::make_unique<PartialDeserializedTprFile>();
 747
 748     if (isSimulationMasterRank)
 749     {
 750         /* Only the master rank has the global state */
 751         globalState = std::make_unique<t_state>();
 752
 753         /* Read (nearly) all data required for the simulation
 754          * and keep the partly serialized tpr contents to send to other ranks later
 755          */
 756         *partialDeserializedTpr = read_tpx_state(ftp2fn(efTPR, filenames.size(), filenames.data()), &inputrecInstance, globalState.get(), &mtop);
 757         inputrec                = &inputrecInstance;
 758     }
 759
 760     /* Check and update the hardware options for internal consistency */
 761     checkAndUpdateHardwareOptions(mdlog, &hw_opt, isSimulationMasterRank, domdecOptions.numPmeRanks, inputrec);
 762
 763     if (GMX_THREAD_MPI && isSimulationMasterRank)
 764     {
 765         bool useGpuForNonbonded = false;
 766         bool useGpuForPme       = false;
 767         try
 768         {
 769             GMX_RELEASE_ASSERT(inputrec != nullptr, "Keep the compiler happy");
 770
 771             // If the user specified the number of ranks, then we must
 772             // respect that, but in default mode, we need to allow for
 773             // the number of GPUs to choose the number of ranks.
 774             auto canUseGpuForNonbonded = buildSupportsNonbondedOnGpu(nullptr);
 775             useGpuForNonbonded = decideWhetherToUseGpusForNonbondedWithThreadMpi
 776                     (nonbondedTarget, gpuIdsToUse, userGpuTaskAssignment, emulateGpuNonbonded,
 777                     canUseGpuForNonbonded,
 778                     gpuAccelerationOfNonbondedIsUseful(mdlog, *inputrec, GMX_THREAD_MPI),
 779                     hw_opt.nthreads_tmpi);
 780             useGpuForPme = decideWhetherToUseGpusForPmeWithThreadMpi
 781                     (useGpuForNonbonded, pmeTarget, gpuIdsToUse, userGpuTaskAssignment,
 782                     *hwinfo, *inputrec, mtop, hw_opt.nthreads_tmpi, domdecOptions.numPmeRanks);
 783
 784         }
 785         GMX_CATCH_ALL_AND_EXIT_WITH_FATAL_ERROR;
 786
 787         /* Determine how many thread-MPI ranks to start.
 788          *
 789          * TODO Over-writing the user-supplied value here does
 790          * prevent any possible subsequent checks from working
 791          * correctly. */
 792         hw_opt.nthreads_tmpi = get_nthreads_mpi(hwinfo,
 793                                                 &hw_opt,
 794                                                 gpuIdsToUse,
 795                                                 useGpuForNonbonded,
 796                                                 useGpuForPme,
 797                                                 inputrec, &mtop,
 798                                                 mdlog,
 799                                                 doMembed);
 800
 801         // Now start the threads for thread MPI.
 802         spawnThreads(hw_opt.nthreads_tmpi);
 803         // The spawned threads enter mdrunner() and execution of
 804         // master and spawned threads joins at the end of this block.
 805         physicalNodeComm = PhysicalNodeCommunicator(communicator, gmx_physicalnode_id_hash());
 806     }
 807
 808     GMX_RELEASE_ASSERT(communicator == MPI_COMM_WORLD, "Must have valid world communicator");
 809     CommrecHandle crHandle = init_commrec(communicator, ms);
 810     t_commrec    *cr       = crHandle.get();
 811     GMX_RELEASE_ASSERT(cr != nullptr, "Must have valid commrec");
 812
 813     if (PAR(cr))
 814     {
 815         /* now broadcast everything to the non-master nodes/threads: */
 816         if (!isSimulationMasterRank)
 817         {
 818             inputrec = &inputrecInstance;
 819         }
 820         init_parallel(cr, inputrec, &mtop, partialDeserializedTpr.get());
 821     }
 822     GMX_RELEASE_ASSERT(inputrec != nullptr, "All ranks should have a valid inputrec now");
 823     partialDeserializedTpr.reset(nullptr);
 824
 825     // Now the number of ranks is known to all ranks, and each knows
 826     // the inputrec read by the master rank. The ranks can now all run
 827     // the task-deciding functions and will agree on the result
 828     // without needing to communicate.
 829     //
 830     // TODO Should we do the communication in debug mode to support
 831     // having an assertion?
 832     const bool useDomainDecomposition = (PAR(cr) && !(EI_TPI(inputrec->eI) ||
 833                                                       inputrec->eI == eiNM));
 834
 835     // Note that these variables describe only their own node.
 836     //
 837     // Note that when bonded interactions run on a GPU they always run
 838     // alongside a nonbonded task, so do not influence task assignment
 839     // even though they affect the force calculation workload.
 840     bool useGpuForNonbonded     = false;
 841     bool useGpuForPme           = false;
 842     bool useGpuForBonded        = false;
 843     bool useGpuForUpdate        = false;
 844     bool gpusWereDetected       = hwinfo->ngpu_compatible_tot > 0;
 845     try
 846     {
 847         // It's possible that there are different numbers of GPUs on
 848         // different nodes, which is the user's responsibility to
 849         // handle. If unsuitable, we will notice that during task
 850         // assignment.
 851         auto canUseGpuForNonbonded = buildSupportsNonbondedOnGpu(nullptr);
 852         useGpuForNonbonded = decideWhetherToUseGpusForNonbonded(nonbondedTarget, userGpuTaskAssignment,
 853                                                                 emulateGpuNonbonded,
 854                                                                 canUseGpuForNonbonded,
 855                                                                 gpuAccelerationOfNonbondedIsUseful(mdlog, *inputrec, !GMX_THREAD_MPI),
 856                                                                 gpusWereDetected);
 857         useGpuForPme = decideWhetherToUseGpusForPme(useGpuForNonbonded, pmeTarget, userGpuTaskAssignment,
 858                                                     *hwinfo, *inputrec, mtop,
 859                                                     cr->nnodes, domdecOptions.numPmeRanks,
 860                                                     gpusWereDetected);
 861         auto canUseGpuForBonded = buildSupportsGpuBondeds(nullptr) && inputSupportsGpuBondeds(*inputrec, mtop, nullptr);
 862         useGpuForBonded =
 863             decideWhetherToUseGpusForBonded(useGpuForNonbonded, useGpuForPme,
 864                                             bondedTarget, canUseGpuForBonded,
 865                                             EVDW_PME(inputrec->vdwtype),
 866                                             EEL_PME_EWALD(inputrec->coulombtype),
 867                                             domdecOptions.numPmeRanks, gpusWereDetected);
 868
 869         pmeRunMode   = (useGpuForPme ? PmeRunMode::GPU : PmeRunMode::CPU);
 870         if (pmeRunMode == PmeRunMode::GPU)
 871         {
 872             if (pmeFftTarget == TaskTarget::Cpu)
 873             {
 874                 pmeRunMode = PmeRunMode::Mixed;
 875             }
 876         }
 877         else if (pmeFftTarget == TaskTarget::Gpu)
 878         {
 879             gmx_fatal(FARGS, "Assigning FFTs to GPU requires PME to be assigned to GPU as well. With PME on CPU you should not be using -pmefft.");
 880         }
 881     }
 882     GMX_CATCH_ALL_AND_EXIT_WITH_FATAL_ERROR;
 883
 884     // Initialize development feature flags that enabled by environment variable
 885     // and report those features that are enabled.
 886     const DevelopmentFeatureFlags devFlags = manageDevelopmentFeatures(mdlog, useGpuForNonbonded, useGpuForPme);
 887
 888     // NOTE: The devFlags need decideWhetherToUseGpusForNonbonded(...) and decideWhetherToUseGpusForPme(...) for overrides,
 889     //       decideWhetherToUseGpuForUpdate() needs devFlags for the '-update auto' override, hence the interleaving.
 890     // NOTE: When the simulationWork is constructed, the useGpuForUpdate overrides the devFlags.enableGpuBufferOps.
 891     try
 892     {
 893         useGpuForUpdate = decideWhetherToUseGpuForUpdate(devFlags.forceGpuUpdateDefaultOn,
 894                                                          useDomainDecomposition,
 895                                                          useGpuForPme,
 896                                                          useGpuForNonbonded,
 897                                                          updateTarget,
 898                                                          gpusWereDetected,
 899                                                          *inputrec,
 900                                                          gmx_mtop_interaction_count(mtop, IF_VSITE) > 0,
 901                                                          doEssentialDynamics,
 902                                                          gmx_mtop_ftype_count(mtop, F_ORIRES) > 0,
 903                                                          gmx_mtop_ftype_count(mtop, F_DISRES) > 0,
 904                                                          replExParams.exchangeInterval > 0);
 905     }
 906     GMX_CATCH_ALL_AND_EXIT_WITH_FATAL_ERROR;
 907
 908
 909     // Build restraints.
 910     // TODO: hide restraint implementation details from Mdrunner.
 911     // There is nothing unique about restraints at this point as far as the
 912     // Mdrunner is concerned. The Mdrunner should just be getting a sequence of
 913     // factory functions from the SimulationContext on which to call mdModules_->add().
 914     // TODO: capture all restraints into a single RestraintModule, passed to the runner builder.
 915     for (auto && restraint : restraintManager_->getRestraints())
 916     {
 917         auto module = RestraintMDModule::create(restraint,
 918                                                 restraint->sites());
 919         mdModules_->add(std::move(module));
 920     }
 921
 922     // TODO: Error handling
 923     mdModules_->assignOptionsToModules(*inputrec->params, nullptr);
 924     const auto &mdModulesNotifier = mdModules_->notifier().notifier_;
 925
 926     if (inputrec->internalParameters != nullptr)
 927     {
 928         mdModulesNotifier.notify(*inputrec->internalParameters);
 929     }
 930
 931     if (fplog != nullptr)
 932     {
 933         pr_inputrec(fplog, 0, "Input Parameters", inputrec, FALSE);
 934         fprintf(fplog, "\n");
 935     }
 936
 937     if (SIMMASTER(cr))
 938     {
 939         /* In rerun, set velocities to zero if present */
 940         if (doRerun && ((globalState->flags & (1 << estV)) != 0))
 941         {
 942             // rerun does not use velocities
 943             GMX_LOG(mdlog.info).asParagraph().appendText(
 944                     "Rerun trajectory contains velocities. Rerun does only evaluate "
 945                     "potential energy and forces. The velocities will be ignored.");
 946             for (int i = 0; i < globalState->natoms; i++)
 947             {
 948                 clear_rvec(globalState->v[i]);
 949             }
 950             globalState->flags &= ~(1 << estV);
 951         }
 952
 953         /* now make sure the state is initialized and propagated */
 954         set_state_entries(globalState.get(), inputrec);
 955     }
 956
 957     /* NM and TPI parallelize over force/energy calculations, not atoms,
 958      * so we need to initialize and broadcast the global state.
 959      */
 960     if (inputrec->eI == eiNM || inputrec->eI == eiTPI)
 961     {
 962         if (!MASTER(cr))
 963         {
 964             globalState = std::make_unique<t_state>();
 965         }
 966         broadcastStateWithoutDynamics(cr, globalState.get());
 967     }
 968
 969     /* A parallel command line option consistency check that we can
 970        only do after any threads have started. */
 971     if (!PAR(cr) && (domdecOptions.numCells[XX] > 1 ||
 972                      domdecOptions.numCells[YY] > 1 ||
 973                      domdecOptions.numCells[ZZ] > 1 ||
 974                      domdecOptions.numPmeRanks > 0))
 975     {
 976         gmx_fatal(FARGS,
 977                   "The -dd or -npme option request a parallel simulation, "
 978 #if !GMX_MPI
 979                   "but %s was compiled without threads or MPI enabled", output_env_get_program_display_name(oenv));
 980 #elif GMX_THREAD_MPI
 981                   "but the number of MPI-threads (option -ntmpi) is not set or is 1");
 982 #else
 983                   "but %s was not started through mpirun/mpiexec or only one rank was requested through mpirun/mpiexec", output_env_get_program_display_name(oenv));
 984 #endif
 985     }
 986
 987     if (doRerun &&
 988         (EI_ENERGY_MINIMIZATION(inputrec->eI) || eiNM == inputrec->eI))
 989     {
 990         gmx_fatal(FARGS, "The .mdp file specified an energy mininization or normal mode algorithm, and these are not compatible with mdrun -rerun");
 991     }
 992
 993     if (!(EEL_PME(inputrec->coulombtype) || EVDW_PME(inputrec->vdwtype)))
 994     {
 995         if (domdecOptions.numPmeRanks > 0)
 996         {
 997             gmx_fatal_collective(FARGS, cr->mpi_comm_mysim, MASTER(cr),
 998                                  "PME-only ranks are requested, but the system does not use PME for electrostatics or LJ");
 999         }
1000
1001         domdecOptions.numPmeRanks = 0;
1002     }
1003
1004     if (useGpuForNonbonded && domdecOptions.numPmeRanks < 0)
1005     {
1006         /* With NB GPUs we don't automatically use PME-only CPU ranks. PME ranks can
1007          * improve performance with many threads per GPU, since our OpenMP
1008          * scaling is bad, but it's difficult to automate the setup.
1009          */
1010         domdecOptions.numPmeRanks = 0;
1011     }
1012     if (useGpuForPme)
1013     {
1014         if (domdecOptions.numPmeRanks < 0)
1015         {
1016             domdecOptions.numPmeRanks = 0;
1017             // TODO possibly print a note that one can opt-in for a separate PME GPU rank?
1018         }
1019         else
1020         {
1021             GMX_RELEASE_ASSERT(domdecOptions.numPmeRanks <= 1, "PME GPU decomposition is not supported");
1022         }
1023     }
1024
1025 #if GMX_FAHCORE
1026     if (MASTER(cr))
1027     {
1028         fcRegisterSteps(inputrec->nsteps, inputrec->init_step);
1029     }
1030 #endif
1031
1032     /* NMR restraints must be initialized before load_checkpoint,
1033      * since with time averaging the history is added to t_state.
1034      * For proper consistency check we therefore need to extend
1035      * t_state here.
1036      * So the PME-only nodes (if present) will also initialize
1037      * the distance restraints.
1038      */
1039     snew(fcd, 1);
1040
1041     /* This needs to be called before read_checkpoint to extend the state */
1042     init_disres(fplog, &mtop, inputrec, cr, ms, fcd, globalState.get(), replExParams.exchangeInterval > 0);
1043
1044     init_orires(fplog, &mtop, inputrec, cr, ms, globalState.get(), &(fcd->orires));
1045
1046     auto                 deform = prepareBoxDeformation(globalState->box, cr, *inputrec);
1047
1048     ObservablesHistory   observablesHistory = {};
1049
1050     if (startingBehavior != StartingBehavior::NewSimulation)
1051     {
1052         /* Check if checkpoint file exists before doing continuation.
1053          * This way we can use identical input options for the first and subsequent runs...
1054          */
1055         if (mdrunOptions.numStepsCommandline > -2)
1056         {
1057             /* Temporarily set the number of steps to unlimited to avoid
1058              * triggering the nsteps check in load_checkpoint().
1059              * This hack will go away soon when the -nsteps option is removed.
1060              */
1061             inputrec->nsteps = -1;
1062         }
1063
1064         load_checkpoint(opt2fn_master("-cpi", filenames.size(), filenames.data(), cr),
1065                         logFileHandle,
1066                         cr, domdecOptions.numCells,
1067                         inputrec, globalState.get(),
1068                         &observablesHistory,
1069                         mdrunOptions.reproducible, mdModules_->notifier());
1070
1071         if (startingBehavior == StartingBehavior::RestartWithAppending && logFileHandle)
1072         {
1073             // Now we can start normal logging to the truncated log file.
1074             fplog    = gmx_fio_getfp(logFileHandle);
1075             prepareLogAppending(fplog);
1076             logOwner = buildLogger(fplog, MASTER(cr));
1077             mdlog    = logOwner.logger();
1078         }
1079     }
1080
1081     if (mdrunOptions.numStepsCommandline > -2)
1082     {
1083         GMX_LOG(mdlog.info).asParagraph().
1084             appendText("The -nsteps functionality is deprecated, and may be removed in a future version. "
1085                        "Consider using gmx convert-tpr -nsteps or changing the appropriate .mdp file field.");
1086     }
1087     /* override nsteps with value set on the commandline */
1088     override_nsteps_cmdline(mdlog, mdrunOptions.numStepsCommandline, inputrec);
1089
1090     if (SIMMASTER(cr))
1091     {
1092         copy_mat(globalState->box, box);
1093     }
1094
1095     if (PAR(cr))
1096     {
1097         gmx_bcast(sizeof(box), box, cr);
1098     }
1099
1100     if (inputrec->cutoff_scheme != ecutsVERLET)
1101     {
1102         gmx_fatal(FARGS, "This group-scheme .tpr file can no longer be run by mdrun. Please update to the Verlet scheme, or use an earlier version of GROMACS if necessary.");
1103     }
1104     /* Update rlist and nstlist. */
1105     prepare_verlet_scheme(fplog, cr, inputrec, nstlist_cmdline, &mtop, box,
1106                           useGpuForNonbonded || (emulateGpuNonbonded == EmulateGpuNonbonded::Yes), *hwinfo->cpuInfo);
1107
1108     const bool prefer1DAnd1PulseDD = (devFlags.enableGpuHaloExchange && useGpuForNonbonded);
1109     // This builder is necessary while we have multi-part construction
1110     // of DD. Before DD is constructed, we use the existence of
1111     // the builder object to indicate that further construction of DD
1112     // is needed.
1113     std::unique_ptr<DomainDecompositionBuilder> ddBuilder;
1114     if (useDomainDecomposition)
1115     {
1116         ddBuilder = std::make_unique<DomainDecompositionBuilder>
1117                 (mdlog, cr, domdecOptions, mdrunOptions,
1118                 prefer1DAnd1PulseDD, mtop, *inputrec,
1119                 box, positionsFromStatePointer(globalState.get()));
1120     }
1121     else
1122     {
1123         /* PME, if used, is done on all nodes with 1D decomposition */
1124         cr->npmenodes = 0;
1125         cr->duty      = (DUTY_PP | DUTY_PME);
1126
1127         if (inputrec->ePBC == epbcSCREW)
1128         {
1129             gmx_fatal(FARGS,
1130                       "pbc=screw is only implemented with domain decomposition");
1131         }
1132     }
1133
1134     // Produce the task assignment for this rank.
1135     GpuTaskAssignmentsBuilder gpuTaskAssignmentsBuilder;
1136     GpuTaskAssignments        gpuTaskAssignments =
1137         gpuTaskAssignmentsBuilder.build(gpuIdsToUse,
1138                                         userGpuTaskAssignment,
1139                                         *hwinfo,
1140                                         cr,
1141                                         ms,
1142                                         physicalNodeComm,
1143                                         nonbondedTarget,
1144                                         pmeTarget,
1145                                         bondedTarget,
1146                                         updateTarget,
1147                                         useGpuForNonbonded,
1148                                         useGpuForPme,
1149                                         thisRankHasDuty(cr, DUTY_PP),
1150                                         // TODO cr->duty & DUTY_PME should imply that a PME
1151                                         // algorithm is active, but currently does not.
1152                                         EEL_PME(inputrec->coulombtype) &&
1153                                         thisRankHasDuty(cr, DUTY_PME));
1154
1155     const bool printHostName = (cr->nnodes > 1);
1156     gpuTaskAssignments.reportGpuUsage(mdlog, printHostName, useGpuForBonded, pmeRunMode);
1157
1158     // If the user chose a task assignment, give them some hints
1159     // where appropriate.
1160     if (!userGpuTaskAssignment.empty())
1161     {
1162         gpuTaskAssignments.logPerformanceHints(mdlog,
1163                                                ssize(gpuIdsToUse));
1164     }
1165
1166     // Get the device handles for the modules, nullptr when no task is assigned.
1167     gmx_device_info_t *nonbondedDeviceInfo   = gpuTaskAssignments.initNonbondedDevice(cr);
1168     gmx_device_info_t *pmeDeviceInfo         = gpuTaskAssignments.initPmeDevice();
1169
1170     // TODO Initialize GPU streams here.
1171
1172     // TODO Currently this is always built, yet DD partition code
1173     // checks if it is built before using it. Probably it should
1174     // become an MDModule that is made only when another module
1175     // requires it (e.g. pull, CompEl, density fitting), so that we
1176     // don't update the local atom sets unilaterally every step.
1177     LocalAtomSetManager atomSets;
1178     if (ddBuilder)
1179     {
1180         // TODO Pass the GPU streams to ddBuilder to use in buffer
1181         // transfers (e.g. halo exchange)
1182         cr->dd = ddBuilder->build(&atomSets);
1183         // The builder's job is done, so destruct it
1184         ddBuilder.reset(nullptr);
1185         // Note that local state still does not exist yet.
1186     }
1187
1188     if (PAR(cr))
1189     {
1190         /* After possible communicator splitting in make_dd_communicators.
1191          * we can set up the intra/inter node communication.
1192          */
1193         gmx_setup_nodecomm(fplog, cr);
1194     }
1195
1196 #if GMX_MPI
1197     if (isMultiSim(ms))
1198     {
1199         GMX_LOG(mdlog.warning).asParagraph().appendTextFormatted(
1200                 "This is simulation %d out of %d running as a composite GROMACS\n"
1201                 "multi-simulation job. Setup for this simulation:\n",
1202                 ms->sim, ms->nsim);
1203     }
1204     GMX_LOG(mdlog.warning).appendTextFormatted(
1205             "Using %d MPI %s\n",
1206             cr->nnodes,
1207 #if GMX_THREAD_MPI
1208             cr->nnodes == 1 ? "thread" : "threads"
1209 #else
1210             cr->nnodes == 1 ? "process" : "processes"
1211 #endif
1212             );
1213     fflush(stderr);
1214 #endif
1215
1216     // If mdrun -pin auto honors any affinity setting that already
1217     // exists. If so, it is nice to provide feedback about whether
1218     // that existing affinity setting was from OpenMP or something
1219     // else, so we run this code both before and after we initialize
1220     // the OpenMP support.
1221     gmx_check_thread_affinity_set(mdlog,
1222                                   &hw_opt, hwinfo->nthreads_hw_avail, FALSE);
1223     /* Check and update the number of OpenMP threads requested */
1224     checkAndUpdateRequestedNumOpenmpThreads(&hw_opt, *hwinfo, cr, ms, physicalNodeComm.size_,
1225                                             pmeRunMode, mtop, *inputrec);
1226
1227     gmx_omp_nthreads_init(mdlog, cr,
1228                           hwinfo->nthreads_hw_avail,
1229                           physicalNodeComm.size_,
1230                           hw_opt.nthreads_omp,
1231                           hw_opt.nthreads_omp_pme,
1232                           !thisRankHasDuty(cr, DUTY_PP));
1233
1234     // Enable FP exception detection, but not in
1235     // Release mode and not for compilers with known buggy FP
1236     // exception support (clang with any optimization) or suspected
1237     // buggy FP exception support (gcc 7.* with optimization).
1238 #if !defined NDEBUG && \
1239     !((defined __clang__ || (defined(__GNUC__) && !defined(__ICC) && __GNUC__ == 7)) \
1240     && defined __OPTIMIZE__)
1241     const bool bEnableFPE = true;
1242 #else
1243     const bool bEnableFPE = false;
1244 #endif
1245     //FIXME - reconcile with gmx_feenableexcept() call from CommandLineModuleManager::run()
1246     if (bEnableFPE)
1247     {
1248         gmx_feenableexcept();
1249     }
1250
1251     /* Now that we know the setup is consistent, check for efficiency */
1252     check_resource_division_efficiency(hwinfo,
1253                                        gpuTaskAssignments.thisRankHasAnyGpuTask(),
1254                                        mdrunOptions.ntompOptionIsSet,
1255                                        cr,
1256                                        mdlog);
1257
1258     /* getting number of PP/PME threads on this MPI / tMPI rank.
1259        PME: env variable should be read only on one node to make sure it is
1260        identical everywhere;
1261      */
1262     const int numThreadsOnThisRank =
1263         thisRankHasDuty(cr, DUTY_PP) ? gmx_omp_nthreads_get(emntNonbonded) : gmx_omp_nthreads_get(emntPME);
1264     checkHardwareOversubscription(numThreadsOnThisRank, cr->nodeid,
1265                                   *hwinfo->hardwareTopology,
1266                                   physicalNodeComm, mdlog);
1267
1268     // Enable Peer access between GPUs where available
1269     // Only for DD, only master PP rank needs to perform setup, and only if thread MPI plus
1270     // any of the GPU communication features are active.
1271     if (DOMAINDECOMP(cr) && MASTER(cr) && thisRankHasDuty(cr, DUTY_PP) && GMX_THREAD_MPI &&
1272         (devFlags.enableGpuHaloExchange || devFlags.enableGpuPmePPComm))
1273     {
1274         setupGpuDevicePeerAccess(gpuIdsToUse, mdlog);
1275     }
1276
1277     if (hw_opt.threadAffinity != ThreadAffinity::Off)
1278     {
1279         /* Before setting affinity, check whether the affinity has changed
1280          * - which indicates that probably the OpenMP library has changed it
1281          * since we first checked).
1282          */
1283         gmx_check_thread_affinity_set(mdlog,
1284                                       &hw_opt, hwinfo->nthreads_hw_avail, TRUE);
1285
1286         int numThreadsOnThisNode, intraNodeThreadOffset;
1287         analyzeThreadsOnThisNode(physicalNodeComm, numThreadsOnThisRank, &numThreadsOnThisNode,
1288                                  &intraNodeThreadOffset);
1289
1290         /* Set the CPU affinity */
1291         gmx_set_thread_affinity(mdlog, cr, &hw_opt, *hwinfo->hardwareTopology,
1292                                 numThreadsOnThisRank, numThreadsOnThisNode,
1293                                 intraNodeThreadOffset, nullptr);
1294     }
1295
1296     if (mdrunOptions.timingOptions.resetStep > -1)
1297     {
1298         GMX_LOG(mdlog.info).asParagraph().
1299             appendText("The -resetstep functionality is deprecated, and may be removed in a future version.");
1300     }
1301     wcycle = wallcycle_init(fplog, mdrunOptions.timingOptions.resetStep, cr);
1302
1303     if (PAR(cr))
1304     {
1305         /* Master synchronizes its value of reset_counters with all nodes
1306          * including PME only nodes */
1307         int64_t reset_counters = wcycle_get_reset_counters(wcycle);
1308         gmx_bcast_sim(sizeof(reset_counters), &reset_counters, cr);
1309         wcycle_set_reset_counters(wcycle, reset_counters);
1310     }
1311
1312     // Membrane embedding must be initialized before we call init_forcerec()
1313     if (doMembed)
1314     {
1315         if (MASTER(cr))
1316         {
1317             fprintf(stderr, "Initializing membed");
1318         }
1319         /* Note that membed cannot work in parallel because mtop is
1320          * changed here. Fix this if we ever want to make it run with
1321          * multiple ranks. */
1322         membed = init_membed(fplog, filenames.size(), filenames.data(), &mtop, inputrec, globalState.get(), cr,
1323                              &mdrunOptions
1324                                  .checkpointOptions.period);
1325     }
1326
1327     const bool                   thisRankHasPmeGpuTask = gpuTaskAssignments.thisRankHasPmeGpuTask();
1328     std::unique_ptr<MDAtoms>     mdAtoms;
1329     std::unique_ptr<gmx_vsite_t> vsite;
1330
1331     t_nrnb nrnb;
1332     if (thisRankHasDuty(cr, DUTY_PP))
1333     {
1334         mdModulesNotifier.notify(*cr);
1335         mdModulesNotifier.notify(&atomSets);
1336         mdModulesNotifier.notify(PeriodicBoundaryConditionType {inputrec->ePBC});
1337         mdModulesNotifier.notify(SimulationTimeStep { inputrec->delta_t });
1338         /* Initiate forcerecord */
1339         fr                 = new t_forcerec;
1340         fr->forceProviders = mdModules_->initForceProviders();
1341         init_forcerec(fplog, mdlog, fr, fcd,
1342                       inputrec, &mtop, cr, box,
1343                       opt2fn("-table", filenames.size(), filenames.data()),
1344                       opt2fn("-tablep", filenames.size(), filenames.data()),
1345                       opt2fns("-tableb", filenames.size(), filenames.data()),
1346                       *hwinfo, nonbondedDeviceInfo,
1347                       useGpuForBonded,
1348                       pmeRunMode == PmeRunMode::GPU && !thisRankHasDuty(cr, DUTY_PME),
1349                       pforce,
1350                       wcycle);
1351
1352         // TODO Move this to happen during domain decomposition setup,
1353         // once stream and event handling works well with that.
1354         // TODO remove need to pass local stream into GPU halo exchange - Redmine #3093
1355         if (havePPDomainDecomposition(cr) && prefer1DAnd1PulseDD && is1DAnd1PulseDD(*cr->dd))
1356         {
1357             GMX_RELEASE_ASSERT(devFlags.enableGpuBufferOps, "Must use GMX_USE_GPU_BUFFER_OPS=1 to use GMX_GPU_DD_COMMS=1");
1358             void *streamLocal              = Nbnxm::gpu_get_command_stream(fr->nbv->gpu_nbv, InteractionLocality::Local);
1359             void *streamNonLocal           = Nbnxm::gpu_get_command_stream(fr->nbv->gpu_nbv, InteractionLocality::NonLocal);
1360             void *coordinatesOnDeviceEvent = fr->nbv->get_x_on_device_event();
1361             GMX_LOG(mdlog.warning).asParagraph().appendTextFormatted(
1362                     "NOTE: This run uses the 'GPU halo exchange' feature, enabled by the GMX_GPU_DD_COMMS environment variable.");
1363             cr->dd->gpuHaloExchange = std::make_unique<GpuHaloExchange>(cr->dd, cr->mpi_comm_mysim, streamLocal,
1364                                                                         streamNonLocal, coordinatesOnDeviceEvent);
1365         }
1366
1367         /* Initialize the mdAtoms structure.
1368          * mdAtoms is not filled with atom data,
1369          * as this can not be done now with domain decomposition.
1370          */
1371         mdAtoms = makeMDAtoms(fplog, mtop, *inputrec, thisRankHasPmeGpuTask);
1372         if (globalState && thisRankHasPmeGpuTask)
1373         {
1374             // The pinning of coordinates in the global state object works, because we only use
1375             // PME on GPU without DD or on a separate PME rank, and because the local state pointer
1376             // points to the global state object without DD.
1377             // FIXME: MD and EM separately set up the local state - this should happen in the same function,
1378             // which should also perform the pinning.
1379             changePinningPolicy(&globalState->x, pme_get_pinning_policy());
1380         }
1381
1382         /* Initialize the virtual site communication */
1383         vsite = initVsite(mtop, cr);
1384
1385         calc_shifts(box, fr->shift_vec);
1386
1387         /* With periodic molecules the charge groups should be whole at start up
1388          * and the virtual sites should not be far from their proper positions.
1389          */
1390         if (!inputrec->bContinuation && MASTER(cr) &&
1391             !(inputrec->ePBC != epbcNONE && inputrec->bPeriodicMols))
1392         {
1393             /* Make molecules whole at start of run */
1394             if (fr->ePBC != epbcNONE)
1395             {
1396                 do_pbc_first_mtop(fplog, inputrec->ePBC, box, &mtop, globalState->x.rvec_array());
1397             }
1398             if (vsite)
1399             {
1400                 /* Correct initial vsite positions are required
1401                  * for the initial distribution in the domain decomposition
1402                  * and for the initial shell prediction.
1403                  */
1404                 constructVsitesGlobal(mtop, globalState->x);
1405             }
1406         }
1407
1408         if (EEL_PME(fr->ic->eeltype) || EVDW_PME(fr->ic->vdwtype))
1409         {
1410             ewaldcoeff_q  = fr->ic->ewaldcoeff_q;
1411             ewaldcoeff_lj = fr->ic->ewaldcoeff_lj;
1412         }
1413     }
1414     else
1415     {
1416         /* This is a PME only node */
1417
1418         GMX_ASSERT(globalState == nullptr, "We don't need the state on a PME only rank and expect it to be unitialized");
1419
1420         ewaldcoeff_q  = calc_ewaldcoeff_q(inputrec->rcoulomb, inputrec->ewald_rtol);
1421         ewaldcoeff_lj = calc_ewaldcoeff_lj(inputrec->rvdw, inputrec->ewald_rtol_lj);
1422     }
1423
1424     gmx_pme_t *sepPmeData = nullptr;
1425     // This reference hides the fact that PME data is owned by runner on PME-only ranks and by forcerec on other ranks
1426     GMX_ASSERT(thisRankHasDuty(cr, DUTY_PP) == (fr != nullptr), "Double-checking that only PME-only ranks have no forcerec");
1427     gmx_pme_t * &pmedata = fr ? fr->pmedata : sepPmeData;
1428
1429     // TODO should live in ewald module once its testing is improved
1430     //
1431     // Later, this program could contain kernels that might be later
1432     // re-used as auto-tuning progresses, or subsequent simulations
1433     // are invoked.
1434     PmeGpuProgramStorage pmeGpuProgram;
1435     if (thisRankHasPmeGpuTask)
1436     {
1437         pmeGpuProgram = buildPmeGpuProgram(pmeDeviceInfo);
1438     }
1439
1440     /* Initiate PME if necessary,
1441      * either on all nodes or on dedicated PME nodes only. */
1442     if (EEL_PME(inputrec->coulombtype) || EVDW_PME(inputrec->vdwtype))
1443     {
1444         if (mdAtoms && mdAtoms->mdatoms())
1445         {
1446             nChargePerturbed = mdAtoms->mdatoms()->nChargePerturbed;
1447             if (EVDW_PME(inputrec->vdwtype))
1448             {
1449                 nTypePerturbed   = mdAtoms->mdatoms()->nTypePerturbed;
1450             }
1451         }
1452         if (cr->npmenodes > 0)
1453         {
1454             /* The PME only nodes need to know nChargePerturbed(FEP on Q) and nTypePerturbed(FEP on LJ)*/
1455             gmx_bcast_sim(sizeof(nChargePerturbed), &nChargePerturbed, cr);
1456             gmx_bcast_sim(sizeof(nTypePerturbed), &nTypePerturbed, cr);
1457         }
1458
1459         if (thisRankHasDuty(cr, DUTY_PME))
1460         {
1461             try
1462             {
1463                 pmedata = gmx_pme_init(cr,
1464                                        getNumPmeDomains(cr->dd),
1465                                        inputrec,
1466                                        nChargePerturbed != 0, nTypePerturbed != 0,
1467                                        mdrunOptions.reproducible,
1468                                        ewaldcoeff_q, ewaldcoeff_lj,
1469                                        gmx_omp_nthreads_get(emntPME),
1470                                        pmeRunMode, nullptr,
1471                                        pmeDeviceInfo, pmeGpuProgram.get(), mdlog);
1472             }
1473             GMX_CATCH_ALL_AND_EXIT_WITH_FATAL_ERROR;
1474         }
1475     }
1476
1477
1478     if (EI_DYNAMICS(inputrec->eI))
1479     {
1480         /* Turn on signal handling on all nodes */
1481         /*
1482          * (A user signal from the PME nodes (if any)
1483          * is communicated to the PP nodes.
1484          */
1485         signal_handler_install();
1486     }
1487
1488     pull_t *pull_work = nullptr;
1489     if (thisRankHasDuty(cr, DUTY_PP))
1490     {
1491         /* Assumes uniform use of the number of OpenMP threads */
1492         walltime_accounting = walltime_accounting_init(gmx_omp_nthreads_get(emntDefault));
1493
1494         if (inputrec->bPull)
1495         {
1496             /* Initialize pull code */
1497             pull_work =
1498                 init_pull(fplog, inputrec->pull, inputrec,
1499                           &mtop, cr, &atomSets, inputrec->fepvals->init_lambda);
1500             if (inputrec->pull->bXOutAverage || inputrec->pull->bFOutAverage)
1501             {
1502                 initPullHistory(pull_work, &observablesHistory);
1503             }
1504             if (EI_DYNAMICS(inputrec->eI) && MASTER(cr))
1505             {
1506                 init_pull_output_files(pull_work,
1507                                        filenames.size(), filenames.data(), oenv,
1508                                        startingBehavior);
1509             }
1510         }
1511
1512         std::unique_ptr<EnforcedRotation> enforcedRotation;
1513         if (inputrec->bRot)
1514         {
1515             /* Initialize enforced rotation code */
1516             enforcedRotation = init_rot(fplog,
1517                                         inputrec,
1518                                         filenames.size(),
1519                                         filenames.data(),
1520                                         cr,
1521                                         &atomSets,
1522                                         globalState.get(),
1523                                         &mtop,
1524                                         oenv,
1525                                         mdrunOptions,
1526                                         startingBehavior);
1527         }
1528
1529         t_swap *swap = nullptr;
1530         if (inputrec->eSwapCoords != eswapNO)
1531         {
1532             /* Initialize ion swapping code */
1533             swap = init_swapcoords(fplog, inputrec,
1534                                    opt2fn_master("-swap", filenames.size(), filenames.data(), cr),
1535                                    &mtop, globalState.get(), &observablesHistory,
1536                                    cr, &atomSets, oenv, mdrunOptions,
1537                                    startingBehavior);
1538         }
1539
1540         /* Let makeConstraints know whether we have essential dynamics constraints. */
1541         auto constr = makeConstraints(mtop, *inputrec, pull_work, doEssentialDynamics,
1542                                       fplog, *mdAtoms->mdatoms(),
1543                                       cr, ms, &nrnb, wcycle, fr->bMolPBC);
1544
1545         /* Energy terms and groups */
1546         gmx_enerdata_t enerd(mtop.groups.groups[SimulationAtomGroupType::EnergyOutput].size(), inputrec->fepvals->n_lambda);
1547
1548         /* Kinetic energy data */
1549         gmx_ekindata_t ekind;
1550         init_ekindata(fplog, &mtop, &(inputrec->opts), &ekind, inputrec->cos_accel);
1551
1552         /* Set up interactive MD (IMD) */
1553         auto imdSession = makeImdSession(inputrec, cr, wcycle, &enerd, ms, &mtop, mdlog,
1554                                          MASTER(cr) ? globalState->x.rvec_array() : nullptr,
1555                                          filenames.size(), filenames.data(), oenv, mdrunOptions.imdOptions,
1556                                          startingBehavior);
1557
1558         if (DOMAINDECOMP(cr))
1559         {
1560             GMX_RELEASE_ASSERT(fr, "fr was NULL while cr->duty was DUTY_PP");
1561             /* This call is not included in init_domain_decomposition mainly
1562              * because fr->cginfo_mb is set later.
1563              */
1564             dd_init_bondeds(fplog, cr->dd, &mtop, vsite.get(), inputrec,
1565                             domdecOptions.checkBondedInteractions,
1566                             fr->cginfo_mb);
1567         }
1568
1569         const bool inputIsCompatibleWithModularSimulator = ModularSimulator::isInputCompatible(
1570                     false,
1571                     inputrec, doRerun, vsite.get(), ms, replExParams,
1572                     fcd, static_cast<int>(filenames.size()), filenames.data(),
1573                     &observablesHistory, membed);
1574
1575         const bool useModularSimulator = inputIsCompatibleWithModularSimulator && !(getenv("GMX_DISABLE_MODULAR_SIMULATOR") != nullptr);
1576
1577         // TODO This is not the right place to manage the lifetime of
1578         // this data structure, but currently it's the easiest way to
1579         // make it work.
1580         MdrunScheduleWorkload runScheduleWork;
1581         // Also populates the simulation constant workload description.
1582         runScheduleWork.simulationWork = createSimulationWorkload(useGpuForNonbonded,
1583                                                                   pmeRunMode,
1584                                                                   useGpuForBonded,
1585                                                                   useGpuForUpdate,
1586                                                                   devFlags.enableGpuBufferOps,
1587                                                                   devFlags.enableGpuHaloExchange,
1588                                                                   devFlags.enableGpuPmePPComm);
1589
1590         std::unique_ptr<gmx::StatePropagatorDataGpu> stateGpu;
1591         if (gpusWereDetected && ((useGpuForPme && thisRankHasDuty(cr, DUTY_PME)) || runScheduleWork.simulationWork.useGpuBufferOps))
1592         {
1593             const void         *pmeStream      = pme_gpu_get_device_stream(fr->pmedata);
1594             const void         *localStream    = fr->nbv->gpu_nbv != nullptr ? Nbnxm::gpu_get_command_stream(fr->nbv->gpu_nbv, InteractionLocality::Local) : nullptr;
1595             const void         *nonLocalStream = fr->nbv->gpu_nbv != nullptr ? Nbnxm::gpu_get_command_stream(fr->nbv->gpu_nbv, InteractionLocality::NonLocal) : nullptr;
1596             const void         *deviceContext  = pme_gpu_get_device_context(fr->pmedata);
1597             const int           paddingSize    = pme_gpu_get_padding_size(fr->pmedata);
1598             GpuApiCallBehavior  transferKind   = (inputrec->eI == eiMD && !doRerun && !useModularSimulator) ? GpuApiCallBehavior::Async : GpuApiCallBehavior::Sync;
1599
1600             stateGpu = std::make_unique<gmx::StatePropagatorDataGpu>(pmeStream,
1601                                                                      localStream,
1602                                                                      nonLocalStream,
1603                                                                      deviceContext,
1604                                                                      transferKind,
1605                                                                      paddingSize);
1606             fr->stateGpu = stateGpu.get();
1607         }
1608
1609         GMX_ASSERT(stopHandlerBuilder_, "Runner must provide StopHandlerBuilder to simulator.");
1610         SimulatorBuilder simulatorBuilder;
1611
1612         // build and run simulator object based on user-input
1613         auto simulator = simulatorBuilder.build(
1614                     inputIsCompatibleWithModularSimulator,
1615                     fplog, cr, ms, mdlog, static_cast<int>(filenames.size()), filenames.data(),
1616                     oenv,
1617                     mdrunOptions,
1618                     startingBehavior,
1619                     vsite.get(), constr.get(),
1620                     enforcedRotation ? enforcedRotation->getLegacyEnfrot() : nullptr,
1621                     deform.get(),
1622                     mdModules_->outputProvider(),
1623                     mdModules_->notifier(),
1624                     inputrec, imdSession.get(), pull_work, swap, &mtop,
1625                     fcd,
1626                     globalState.get(),
1627                     &observablesHistory,
1628                     mdAtoms.get(), &nrnb, wcycle, fr,
1629                     &enerd,
1630                     &ekind,
1631                     &runScheduleWork,
1632                     replExParams,
1633                     membed,
1634                     walltime_accounting,
1635                     std::move(stopHandlerBuilder_),
1636                     doRerun);
1637         simulator->run();
1638
1639         if (inputrec->bPull)
1640         {
1641             finish_pull(pull_work);
1642         }
1643         finish_swapcoords(swap);
1644     }
1645     else
1646     {
1647         GMX_RELEASE_ASSERT(pmedata, "pmedata was NULL while cr->duty was not DUTY_PP");
1648         /* do PME only */
1649         walltime_accounting = walltime_accounting_init(gmx_omp_nthreads_get(emntPME));
1650         gmx_pmeonly(pmedata, cr, &nrnb, wcycle, walltime_accounting, inputrec, pmeRunMode);
1651     }
1652
1653     wallcycle_stop(wcycle, ewcRUN);
1654
1655     /* Finish up, write some stuff
1656      * if rerunMD, don't write last frame again
1657      */
1658     finish_run(fplog, mdlog, cr,
1659                inputrec, &nrnb, wcycle, walltime_accounting,
1660                fr ? fr->nbv.get() : nullptr,
1661                pmedata,
1662                EI_DYNAMICS(inputrec->eI) && !isMultiSim(ms));
1663
1664     // clean up cycle counter
1665     wallcycle_destroy(wcycle);
1666
1667 // Free PME data
1668     if (pmedata)
1669     {
1670         gmx_pme_destroy(pmedata);
1671         pmedata = nullptr;
1672     }
1673
1674     // FIXME: this is only here to manually unpin mdAtoms->chargeA_ and state->x,
1675     // before we destroy the GPU context(s) in free_gpu_resources().
1676     // Pinned buffers are associated with contexts in CUDA.
1677     // As soon as we destroy GPU contexts after mdrunner() exits, these lines should go.
1678     mdAtoms.reset(nullptr);
1679     globalState.reset(nullptr);
1680     mdModules_.reset(nullptr);   // destruct force providers here as they might also use the GPU
1681
1682     /* Free GPU memory and set a physical node tMPI barrier (which should eventually go away) */
1683     free_gpu_resources(fr, physicalNodeComm, hwinfo->gpu_info);
1684     free_gpu(nonbondedDeviceInfo);
1685     free_gpu(pmeDeviceInfo);
1686     done_forcerec(fr, mtop.molblock.size());
1687     sfree(fcd);
1688
1689     if (doMembed)
1690     {
1691         free_membed(membed);
1692     }
1693
1694     /* Does what it says */
1695     print_date_and_time(fplog, cr->nodeid, "Finished mdrun", gmx_gettime());
1696     walltime_accounting_destroy(walltime_accounting);
1697
1698     // Ensure log file content is written
1699     if (logFileHandle)
1700     {
1701         gmx_fio_flush(logFileHandle);
1702     }
1703
1704     /* Reset FPEs (important for unit tests) by disabling them. Assumes no
1705      * exceptions were enabled before function was called. */
1706     if (bEnableFPE)
1707     {
1708         gmx_fedisableexcept();
1709     }
1710
1711     auto rc = static_cast<int>(gmx_get_stop_condition());
1712
1713 #if GMX_THREAD_MPI
1714     /* we need to join all threads. The sub-threads join when they
1715        exit this function, but the master thread needs to be told to
1716        wait for that. */
1717     if (PAR(cr) && MASTER(cr))
1718     {
1719         tMPI_Finalize();
1720     }
1721 #endif
1722     return rc;
1723 }
1724
1725 Mdrunner::~Mdrunner()
1726 {
1727     // Clean up of the Manager.
1728     // This will end up getting called on every thread-MPI rank, which is unnecessary,
1729     // but okay as long as threads synchronize some time before adding or accessing
1730     // a new set of restraints.
1731     if (restraintManager_)
1732     {
1733         restraintManager_->clear();
1734         GMX_ASSERT(restraintManager_->countRestraints() == 0,
1735                    "restraints added during runner life time should be cleared at runner destruction.");
1736     }
1737 };
1738
1739 void Mdrunner::addPotential(std::shared_ptr<gmx::IRestraintPotential> puller,
1740                             const std::string                        &name)
1741 {
1742     GMX_ASSERT(restraintManager_, "Mdrunner must have a restraint manager.");
1743     // Not sure if this should be logged through the md logger or something else,
1744     // but it is helpful to have some sort of INFO level message sent somewhere.
1745     //    std::cout << "Registering restraint named " << name << std::endl;
1746
1747     // When multiple restraints are used, it may be wasteful to register them separately.
1748     // Maybe instead register an entire Restraint Manager as a force provider.
1749     restraintManager_->addToSpec(std::move(puller),
1750                                  name);
1751 }
1752
1753 Mdrunner::Mdrunner(std::unique_ptr<MDModules> mdModules)
1754     : mdModules_(std::move(mdModules))
1755 {
1756 }
1757
1758 Mdrunner::Mdrunner(Mdrunner &&) noexcept = default;
1759
1760 //NOLINTNEXTLINE(performance-noexcept-move-constructor) working around GCC bug 58265
1761 Mdrunner &Mdrunner::operator=(Mdrunner && /*handle*/) noexcept(BUGFREE_NOEXCEPT_STRING) = default;
1762
1763 class Mdrunner::BuilderImplementation
1764 {
1765     public:
1766         BuilderImplementation() = delete;
1767         BuilderImplementation(std::unique_ptr<MDModules>           mdModules,
1768                               compat::not_null<SimulationContext*> context);
1769         ~BuilderImplementation();
1770
1771         BuilderImplementation &setExtraMdrunOptions(const MdrunOptions &options,
1772                                                     real                forceWarningThreshold,
1773                                                     StartingBehavior    startingBehavior);
1774
1775         void addDomdec(const DomdecOptions &options);
1776
1777         void addVerletList(int nstlist);
1778
1779         void addReplicaExchange(const ReplicaExchangeParameters &params);
1780
1781         void addNonBonded(const char* nbpu_opt);
1782
1783         void addPME(const char* pme_opt_, const char* pme_fft_opt_);
1784
1785         void addBondedTaskAssignment(const char* bonded_opt);
1786
1787         void addUpdateTaskAssignment(const char* update_opt);
1788
1789         void addHardwareOptions(const gmx_hw_opt_t &hardwareOptions);
1790
1791         void addFilenames(ArrayRef <const t_filenm> filenames);
1792
1793         void addOutputEnvironment(gmx_output_env_t* outputEnvironment);
1794
1795         void addLogFile(t_fileio *logFileHandle);
1796
1797         void addStopHandlerBuilder(std::unique_ptr<StopHandlerBuilder> builder);
1798
1799         Mdrunner build();
1800
1801     private:
1802
1803         // Default parameters copied from runner.h
1804         // \todo Clarify source(s) of default parameters.
1805
1806         const char* nbpu_opt_          = nullptr;
1807         const char* pme_opt_           = nullptr;
1808         const char* pme_fft_opt_       = nullptr;
1809         const char *bonded_opt_        = nullptr;
1810         const char *update_opt_        = nullptr;
1811
1812         MdrunOptions                          mdrunOptions_;
1813
1814         DomdecOptions                         domdecOptions_;
1815
1816         ReplicaExchangeParameters             replicaExchangeParameters_;
1817
1818         //! Command-line override for the duration of a neighbor list with the Verlet scheme.
1819         int         nstlist_ = 0;
1820
1821         //! Multisim communicator handle.
1822         gmx_multisim_t *multiSimulation_;
1823
1824         //! mdrun communicator
1825         MPI_Comm communicator_ = MPI_COMM_NULL;
1826
1827         //! Print a warning if any force is larger than this (in kJ/mol nm).
1828         real forceWarningThreshold_ = -1;
1829
1830         //! Whether the simulation will start afresh, or restart with/without appending.
1831         StartingBehavior startingBehavior_ = StartingBehavior::NewSimulation;
1832
1833         //! The modules that comprise the functionality of mdrun.
1834         std::unique_ptr<MDModules> mdModules_;
1835
1836         //! \brief Parallelism information.
1837         gmx_hw_opt_t hardwareOptions_;
1838
1839         //! filename options for simulation.
1840         ArrayRef<const t_filenm> filenames_;
1841
1842         /*! \brief Handle to output environment.
1843          *
1844          * \todo gmx_output_env_t needs lifetime management.
1845          */
1846         gmx_output_env_t*    outputEnvironment_ = nullptr;
1847
1848         /*! \brief Non-owning handle to MD log file.
1849          *
1850          * \todo Context should own output facilities for client.
1851          * \todo Improve log file handle management.
1852          * \internal
1853          * Code managing the FILE* relies on the ability to set it to
1854          * nullptr to check whether the filehandle is valid.
1855          */
1856         t_fileio* logFileHandle_ = nullptr;
1857
1858         /*!
1859          * \brief Builder for simulation stop signal handler.
1860          */
1861         std::unique_ptr<StopHandlerBuilder> stopHandlerBuilder_ = nullptr;
1862 };
1863
1864 Mdrunner::BuilderImplementation::BuilderImplementation(std::unique_ptr<MDModules>           mdModules,
1865                                                        compat::not_null<SimulationContext*> context) :
1866     mdModules_(std::move(mdModules))
1867 {
1868     communicator_    = context->communicator_;
1869     multiSimulation_ = context->multiSimulation_.get();
1870 }
1871
1872 Mdrunner::BuilderImplementation::~BuilderImplementation() = default;
1873
1874 Mdrunner::BuilderImplementation &
1875 Mdrunner::BuilderImplementation::setExtraMdrunOptions(const MdrunOptions    &options,
1876                                                       const real             forceWarningThreshold,
1877                                                       const StartingBehavior startingBehavior)
1878 {
1879     mdrunOptions_          = options;
1880     forceWarningThreshold_ = forceWarningThreshold;
1881     startingBehavior_      = startingBehavior;
1882     return *this;
1883 }
1884
1885 void Mdrunner::BuilderImplementation::addDomdec(const DomdecOptions &options)
1886 {
1887     domdecOptions_ = options;
1888 }
1889
1890 void Mdrunner::BuilderImplementation::addVerletList(int nstlist)
1891 {
1892     nstlist_ = nstlist;
1893 }
1894
1895 void Mdrunner::BuilderImplementation::addReplicaExchange(const ReplicaExchangeParameters &params)
1896 {
1897     replicaExchangeParameters_ = params;
1898 }
1899
1900 Mdrunner Mdrunner::BuilderImplementation::build()
1901 {
1902     auto newRunner = Mdrunner(std::move(mdModules_));
1903
1904     newRunner.mdrunOptions          = mdrunOptions_;
1905     newRunner.pforce                = forceWarningThreshold_;
1906     newRunner.startingBehavior      = startingBehavior_;
1907     newRunner.domdecOptions         = domdecOptions_;
1908
1909     // \todo determine an invariant to check or confirm that all gmx_hw_opt_t objects are valid
1910     newRunner.hw_opt          = hardwareOptions_;
1911
1912     // No invariant to check. This parameter exists to optionally override other behavior.
1913     newRunner.nstlist_cmdline = nstlist_;
1914
1915     newRunner.replExParams    = replicaExchangeParameters_;
1916
1917     newRunner.filenames = filenames_;
1918
1919     newRunner.communicator = communicator_;
1920
1921     // nullptr is a valid value for the multisim handle
1922     newRunner.ms = multiSimulation_;
1923
1924     // \todo Clarify ownership and lifetime management for gmx_output_env_t
1925     // \todo Update sanity checking when output environment has clearly specified invariants.
1926     // Initialization and default values for oenv are not well specified in the current version.
1927     if (outputEnvironment_)
1928     {
1929         newRunner.oenv  = outputEnvironment_;
1930     }
1931     else
1932     {
1933         GMX_THROW(gmx::APIError("MdrunnerBuilder::addOutputEnvironment() is required before build()"));
1934     }
1935
1936     newRunner.logFileHandle = logFileHandle_;
1937
1938     if (nbpu_opt_)
1939     {
1940         newRunner.nbpu_opt    = nbpu_opt_;
1941     }
1942     else
1943     {
1944         GMX_THROW(gmx::APIError("MdrunnerBuilder::addNonBonded() is required before build()"));
1945     }
1946
1947     if (pme_opt_ && pme_fft_opt_)
1948     {
1949         newRunner.pme_opt     = pme_opt_;
1950         newRunner.pme_fft_opt = pme_fft_opt_;
1951     }
1952     else
1953     {
1954         GMX_THROW(gmx::APIError("MdrunnerBuilder::addElectrostatics() is required before build()"));
1955     }
1956
1957     if (bonded_opt_)
1958     {
1959         newRunner.bonded_opt = bonded_opt_;
1960     }
1961     else
1962     {
1963         GMX_THROW(gmx::APIError("MdrunnerBuilder::addBondedTaskAssignment() is required before build()"));
1964     }
1965
1966     if (update_opt_)
1967     {
1968         newRunner.update_opt = update_opt_;
1969     }
1970     else
1971     {
1972         GMX_THROW(gmx::APIError("MdrunnerBuilder::addUpdateTaskAssignment() is required before build()  "));
1973     }
1974
1975
1976     newRunner.restraintManager_ = std::make_unique<gmx::RestraintManager>();
1977
1978     if (stopHandlerBuilder_)
1979     {
1980         newRunner.stopHandlerBuilder_ = std::move(stopHandlerBuilder_);
1981     }
1982     else
1983     {
1984         newRunner.stopHandlerBuilder_ = std::make_unique<StopHandlerBuilder>();
1985     }
1986
1987     return newRunner;
1988 }
1989
1990 void Mdrunner::BuilderImplementation::addNonBonded(const char* nbpu_opt)
1991 {
1992     nbpu_opt_ = nbpu_opt;
1993 }
1994
1995 void Mdrunner::BuilderImplementation::addPME(const char* pme_opt,
1996                                              const char* pme_fft_opt)
1997 {
1998     pme_opt_     = pme_opt;
1999     pme_fft_opt_ = pme_fft_opt;
2000 }
2001
2002 void Mdrunner::BuilderImplementation::addBondedTaskAssignment(const char* bonded_opt)
2003 {
2004     bonded_opt_ = bonded_opt;
2005 }
2006
2007 void Mdrunner::BuilderImplementation::addUpdateTaskAssignment(const char* update_opt)
2008 {
2009     update_opt_ = update_opt;
2010 }
2011
2012 void Mdrunner::BuilderImplementation::addHardwareOptions(const gmx_hw_opt_t &hardwareOptions)
2013 {
2014     hardwareOptions_ = hardwareOptions;
2015 }
2016
2017 void Mdrunner::BuilderImplementation::addFilenames(ArrayRef<const t_filenm> filenames)
2018 {
2019     filenames_ = filenames;
2020 }
2021
2022 void Mdrunner::BuilderImplementation::addOutputEnvironment(gmx_output_env_t* outputEnvironment)
2023 {
2024     outputEnvironment_ = outputEnvironment;
2025 }
2026
2027 void Mdrunner::BuilderImplementation::addLogFile(t_fileio *logFileHandle)
2028 {
2029     logFileHandle_ = logFileHandle;
2030 }
2031
2032 void Mdrunner::BuilderImplementation::addStopHandlerBuilder(std::unique_ptr<StopHandlerBuilder> builder)
2033 {
2034     stopHandlerBuilder_ = std::move(builder);
2035 }
2036
2037 MdrunnerBuilder::MdrunnerBuilder(std::unique_ptr<MDModules>           mdModules,
2038                                  compat::not_null<SimulationContext*> context) :
2039     impl_ {std::make_unique<Mdrunner::BuilderImplementation>(std::move(mdModules), context)}
2040 {
2041 }
2042
2043 MdrunnerBuilder::~MdrunnerBuilder() = default;
2044
2045 MdrunnerBuilder &MdrunnerBuilder::addSimulationMethod(const MdrunOptions    &options,
2046                                                       real                   forceWarningThreshold,
2047                                                       const StartingBehavior startingBehavior)
2048 {
2049     impl_->setExtraMdrunOptions(options, forceWarningThreshold, startingBehavior);
2050     return *this;
2051 }
2052
2053 MdrunnerBuilder &MdrunnerBuilder::addDomainDecomposition(const DomdecOptions &options)
2054 {
2055     impl_->addDomdec(options);
2056     return *this;
2057 }
2058
2059 MdrunnerBuilder &MdrunnerBuilder::addNeighborList(int nstlist)
2060 {
2061     impl_->addVerletList(nstlist);
2062     return *this;
2063 }
2064
2065 MdrunnerBuilder &MdrunnerBuilder::addReplicaExchange(const ReplicaExchangeParameters &params)
2066 {
2067     impl_->addReplicaExchange(params);
2068     return *this;
2069 }
2070
2071 MdrunnerBuilder &MdrunnerBuilder::addNonBonded(const char* nbpu_opt)
2072 {
2073     impl_->addNonBonded(nbpu_opt);
2074     return *this;
2075 }
2076
2077 MdrunnerBuilder &MdrunnerBuilder::addElectrostatics(const char* pme_opt,
2078                                                     const char* pme_fft_opt)
2079 {
2080     // The builder method may become more general in the future, but in this version,
2081     // parameters for PME electrostatics are both required and the only parameters
2082     // available.
2083     if (pme_opt && pme_fft_opt)
2084     {
2085         impl_->addPME(pme_opt, pme_fft_opt);
2086     }
2087     else
2088     {
2089         GMX_THROW(gmx::InvalidInputError("addElectrostatics() arguments must be non-null pointers."));
2090     }
2091     return *this;
2092 }
2093
2094 MdrunnerBuilder &MdrunnerBuilder::addBondedTaskAssignment(const char* bonded_opt)
2095 {
2096     impl_->addBondedTaskAssignment(bonded_opt);
2097     return *this;
2098 }
2099
2100 MdrunnerBuilder &MdrunnerBuilder::addUpdateTaskAssignment(const char* update_opt)
2101 {
2102     impl_->addUpdateTaskAssignment(update_opt);
2103     return *this;
2104 }
2105
2106 Mdrunner MdrunnerBuilder::build()
2107 {
2108     return impl_->build();
2109 }
2110
2111 MdrunnerBuilder &MdrunnerBuilder::addHardwareOptions(const gmx_hw_opt_t &hardwareOptions)
2112 {
2113     impl_->addHardwareOptions(hardwareOptions);
2114     return *this;
2115 }
2116
2117 MdrunnerBuilder &MdrunnerBuilder::addFilenames(ArrayRef<const t_filenm> filenames)
2118 {
2119     impl_->addFilenames(filenames);
2120     return *this;
2121 }
2122
2123 MdrunnerBuilder &MdrunnerBuilder::addOutputEnvironment(gmx_output_env_t* outputEnvironment)
2124 {
2125     impl_->addOutputEnvironment(outputEnvironment);
2126     return *this;
2127 }
2128
2129 MdrunnerBuilder &MdrunnerBuilder::addLogFile(t_fileio *logFileHandle)
2130 {
2131     impl_->addLogFile(logFileHandle);
2132     return *this;
2133 }
2134
2135 MdrunnerBuilder &MdrunnerBuilder::addStopHandlerBuilder(std::unique_ptr<StopHandlerBuilder> builder)
2136 {
2137     impl_->addStopHandlerBuilder(std::move(builder));
2138     return *this;
2139 }
2140
2141 MdrunnerBuilder::MdrunnerBuilder(MdrunnerBuilder &&) noexcept = default;
2142
2143 MdrunnerBuilder &MdrunnerBuilder::operator=(MdrunnerBuilder &&) noexcept = default;
2144
2145 } // namespace gmx