2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 1991-2000, University of Groningen, The Netherlands.
5 * Copyright (c) 2001-2004, The GROMACS development team.
6 * Copyright (c) 2011-2019,2020,2021, 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.
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.
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.
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.
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.
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.
39 * \brief Implements the MD runner routine calling all integrators.
41 * \author David van der Spoel <david.vanderspoel@icm.uu.se>
42 * \ingroup module_mdrun
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/makebondedlinks.h"
66 #include "gromacs/domdec/partition.h"
67 #include "gromacs/domdec/reversetopology.h"
68 #include "gromacs/ewald/ewald_utils.h"
69 #include "gromacs/ewald/pme.h"
70 #include "gromacs/ewald/pme_gpu_program.h"
71 #include "gromacs/ewald/pme_only.h"
72 #include "gromacs/ewald/pme_pp_comm_gpu.h"
73 #include "gromacs/fileio/checkpoint.h"
74 #include "gromacs/fileio/gmxfio.h"
75 #include "gromacs/fileio/oenv.h"
76 #include "gromacs/fileio/tpxio.h"
77 #include "gromacs/gmxlib/network.h"
78 #include "gromacs/gmxlib/nrnb.h"
79 #include "gromacs/gpu_utils/device_stream_manager.h"
80 #include "gromacs/hardware/cpuinfo.h"
81 #include "gromacs/hardware/detecthardware.h"
82 #include "gromacs/hardware/device_management.h"
83 #include "gromacs/hardware/hardwaretopology.h"
84 #include "gromacs/hardware/printhardware.h"
85 #include "gromacs/imd/imd.h"
86 #include "gromacs/listed_forces/disre.h"
87 #include "gromacs/listed_forces/listed_forces_gpu.h"
88 #include "gromacs/listed_forces/listed_forces.h"
89 #include "gromacs/listed_forces/orires.h"
90 #include "gromacs/math/functions.h"
91 #include "gromacs/math/utilities.h"
92 #include "gromacs/math/vec.h"
93 #include "gromacs/mdlib/boxdeformation.h"
94 #include "gromacs/mdlib/broadcaststructs.h"
95 #include "gromacs/mdlib/calc_verletbuf.h"
96 #include "gromacs/mdlib/dispersioncorrection.h"
97 #include "gromacs/mdlib/enerdata_utils.h"
98 #include "gromacs/mdlib/force.h"
99 #include "gromacs/mdlib/forcerec.h"
100 #include "gromacs/mdlib/gmx_omp_nthreads.h"
101 #include "gromacs/mdlib/gpuforcereduction.h"
102 #include "gromacs/mdlib/makeconstraints.h"
103 #include "gromacs/mdlib/md_support.h"
104 #include "gromacs/mdlib/mdatoms.h"
105 #include "gromacs/mdlib/sighandler.h"
106 #include "gromacs/mdlib/stophandler.h"
107 #include "gromacs/mdlib/tgroup.h"
108 #include "gromacs/mdlib/updategroups.h"
109 #include "gromacs/mdlib/vsite.h"
110 #include "gromacs/mdrun/mdmodules.h"
111 #include "gromacs/mdrun/simulationcontext.h"
112 #include "gromacs/mdrun/simulationinput.h"
113 #include "gromacs/mdrun/simulationinputhandle.h"
114 #include "gromacs/mdrunutility/handlerestart.h"
115 #include "gromacs/mdrunutility/logging.h"
116 #include "gromacs/mdrunutility/multisim.h"
117 #include "gromacs/mdrunutility/printtime.h"
118 #include "gromacs/mdrunutility/threadaffinity.h"
119 #include "gromacs/mdtypes/checkpointdata.h"
120 #include "gromacs/mdtypes/commrec.h"
121 #include "gromacs/mdtypes/enerdata.h"
122 #include "gromacs/mdtypes/fcdata.h"
123 #include "gromacs/mdtypes/forcerec.h"
124 #include "gromacs/mdtypes/group.h"
125 #include "gromacs/mdtypes/inputrec.h"
126 #include "gromacs/mdtypes/interaction_const.h"
127 #include "gromacs/mdtypes/md_enums.h"
128 #include "gromacs/mdtypes/mdatom.h"
129 #include "gromacs/mdtypes/mdrunoptions.h"
130 #include "gromacs/mdtypes/observableshistory.h"
131 #include "gromacs/mdtypes/observablesreducer.h"
132 #include "gromacs/mdtypes/simulation_workload.h"
133 #include "gromacs/mdtypes/state.h"
134 #include "gromacs/mdtypes/state_propagator_data_gpu.h"
135 #include "gromacs/modularsimulator/modularsimulator.h"
136 #include "gromacs/nbnxm/gpu_data_mgmt.h"
137 #include "gromacs/nbnxm/nbnxm.h"
138 #include "gromacs/nbnxm/pairlist_tuning.h"
139 #include "gromacs/pbcutil/pbc.h"
140 #include "gromacs/pulling/output.h"
141 #include "gromacs/pulling/pull.h"
142 #include "gromacs/pulling/pull_rotation.h"
143 #include "gromacs/restraint/manager.h"
144 #include "gromacs/restraint/restraintmdmodule.h"
145 #include "gromacs/restraint/restraintpotential.h"
146 #include "gromacs/swap/swapcoords.h"
147 #include "gromacs/taskassignment/decidegpuusage.h"
148 #include "gromacs/taskassignment/decidesimulationworkload.h"
149 #include "gromacs/taskassignment/resourcedivision.h"
150 #include "gromacs/taskassignment/taskassignment.h"
151 #include "gromacs/taskassignment/usergpuids.h"
152 #include "gromacs/timing/gpu_timing.h"
153 #include "gromacs/timing/wallcycle.h"
154 #include "gromacs/timing/wallcyclereporting.h"
155 #include "gromacs/topology/mtop_util.h"
156 #include "gromacs/trajectory/trajectoryframe.h"
157 #include "gromacs/utility/basenetwork.h"
158 #include "gromacs/utility/cstringutil.h"
159 #include "gromacs/utility/exceptions.h"
160 #include "gromacs/utility/fatalerror.h"
161 #include "gromacs/utility/filestream.h"
162 #include "gromacs/utility/gmxassert.h"
163 #include "gromacs/utility/gmxmpi.h"
164 #include "gromacs/utility/keyvaluetree.h"
165 #include "gromacs/utility/logger.h"
166 #include "gromacs/utility/loggerbuilder.h"
167 #include "gromacs/utility/mdmodulesnotifiers.h"
168 #include "gromacs/utility/physicalnodecommunicator.h"
169 #include "gromacs/utility/pleasecite.h"
170 #include "gromacs/utility/programcontext.h"
171 #include "gromacs/utility/smalloc.h"
172 #include "gromacs/utility/stringutil.h"
173 #include "gromacs/utility/mpiinfo.h"
175 #include "isimulator.h"
176 #include "membedholder.h"
177 #include "replicaexchange.h"
178 #include "simulatorbuilder.h"
184 /*! \brief Manage any development feature flag variables encountered
186 * The use of dev features indicated by environment variables is
187 * logged in order to ensure that runs with such features enabled can
188 * be identified from their log and standard output. Any cross
189 * dependencies are also checked, and if unsatisfied, a fatal error
192 * Note that some development features overrides are applied already here:
193 * the GPU communication flags are set to false in non-tMPI and non-CUDA builds.
195 * \param[in] mdlog Logger object.
196 * \param[in] useGpuForNonbonded True if the nonbonded task is offloaded in this run.
197 * \param[in] pmeRunMode The PME run mode for this run
198 * \returns The object populated with development feature flags.
200 static DevelopmentFeatureFlags manageDevelopmentFeatures(const gmx::MDLogger& mdlog,
201 const bool useGpuForNonbonded,
202 const PmeRunMode pmeRunMode)
204 DevelopmentFeatureFlags devFlags;
206 devFlags.enableGpuBufferOps = (GMX_GPU_CUDA || GMX_GPU_SYCL) && useGpuForNonbonded
207 && (getenv("GMX_USE_GPU_BUFFER_OPS") != nullptr);
208 devFlags.enableGpuHaloExchange = GMX_MPI && GMX_GPU_CUDA && getenv("GMX_GPU_DD_COMMS") != nullptr;
209 devFlags.forceGpuUpdateDefault = (getenv("GMX_FORCE_UPDATE_DEFAULT_GPU") != nullptr) || GMX_FAHCORE;
210 devFlags.enableGpuPmePPComm = GMX_MPI && GMX_GPU_CUDA && getenv("GMX_GPU_PME_PP_COMMS") != nullptr;
212 // Direct GPU comm path is being used with CUDA_AWARE_MPI
213 // make sure underlying MPI implementation is CUDA-aware
214 if (!GMX_THREAD_MPI && (devFlags.enableGpuPmePPComm || devFlags.enableGpuHaloExchange))
216 const bool haveDetectedCudaAwareMpi =
217 (checkMpiCudaAwareSupport() == CudaAwareMpiStatus::Supported);
218 const bool forceCudaAwareMpi = (getenv("GMX_FORCE_CUDA_AWARE_MPI") != nullptr);
220 if (!haveDetectedCudaAwareMpi && forceCudaAwareMpi)
222 // CUDA-aware support not detected in MPI library but, user has forced it's use
223 GMX_LOG(mdlog.warning)
225 .appendTextFormatted(
226 "This run has forced use of 'CUDA-aware MPI'. "
227 "But, GROMACS cannot determine if underlying MPI "
228 "is CUDA-aware. GROMACS recommends use of latest openMPI version "
229 "for CUDA-aware support. "
230 "If you observe failures at runtime, try unsetting "
231 "GMX_FORCE_CUDA_AWARE_MPI environment variable.");
234 if (haveDetectedCudaAwareMpi || forceCudaAwareMpi)
236 devFlags.usingCudaAwareMpi = true;
237 GMX_LOG(mdlog.warning)
239 .appendTextFormatted(
240 "Using CUDA-aware MPI for 'GPU halo exchange' or 'GPU PME-PP "
241 "communications' feature.");
245 if (devFlags.enableGpuHaloExchange)
247 GMX_LOG(mdlog.warning)
249 .appendTextFormatted(
250 "GMX_GPU_DD_COMMS environment variable detected, but the 'GPU "
251 "halo exchange' feature will not be enabled as GROMACS couldn't "
252 "detect CUDA_aware support in underlying MPI implementation.");
253 devFlags.enableGpuHaloExchange = false;
255 if (devFlags.enableGpuPmePPComm)
257 GMX_LOG(mdlog.warning)
260 "GMX_GPU_PME_PP_COMMS environment variable detected, but the "
261 "'GPU PME-PP communications' feature will not be enabled as "
263 "detect CUDA_aware support in underlying MPI implementation.");
264 devFlags.enableGpuPmePPComm = false;
267 GMX_LOG(mdlog.warning)
269 .appendTextFormatted(
270 "GROMACS recommends use of latest OpenMPI version for CUDA-aware "
272 "If you are certain about CUDA-aware support in your MPI library, "
273 "you can force it's use by setting environment variable "
274 " GMX_FORCE_CUDA_AWARE_MPI.");
278 if (devFlags.enableGpuBufferOps)
280 GMX_LOG(mdlog.warning)
282 .appendTextFormatted(
283 "This run uses the 'GPU buffer ops' feature, enabled by the "
284 "GMX_USE_GPU_BUFFER_OPS environment variable.");
287 if (devFlags.forceGpuUpdateDefault)
289 GMX_LOG(mdlog.warning)
291 .appendTextFormatted(
292 "This run will default to '-update gpu' as requested by the "
293 "GMX_FORCE_UPDATE_DEFAULT_GPU environment variable. GPU update with domain "
294 "decomposition lacks substantial testing and should be used with caution.");
297 if (devFlags.enableGpuHaloExchange)
299 if (useGpuForNonbonded)
301 if (!devFlags.enableGpuBufferOps)
303 GMX_LOG(mdlog.warning)
305 .appendTextFormatted(
306 "Enabling GPU buffer operations required by GMX_GPU_DD_COMMS "
307 "(equivalent with GMX_USE_GPU_BUFFER_OPS=1).");
308 devFlags.enableGpuBufferOps = true;
310 GMX_LOG(mdlog.warning)
312 .appendTextFormatted(
313 "This run has requested the 'GPU halo exchange' feature, enabled by "
315 "GMX_GPU_DD_COMMS environment variable.");
319 GMX_LOG(mdlog.warning)
321 .appendTextFormatted(
322 "GMX_GPU_DD_COMMS environment variable detected, but the 'GPU "
323 "halo exchange' feature will not be enabled as nonbonded interactions "
324 "are not offloaded.");
325 devFlags.enableGpuHaloExchange = false;
329 if (devFlags.enableGpuPmePPComm)
331 if (pmeRunMode == PmeRunMode::GPU)
333 if (!devFlags.enableGpuBufferOps)
335 GMX_LOG(mdlog.warning)
337 .appendTextFormatted(
338 "Enabling GPU buffer operations required by GMX_GPU_PME_PP_COMMS "
339 "(equivalent with GMX_USE_GPU_BUFFER_OPS=1).");
340 devFlags.enableGpuBufferOps = true;
342 GMX_LOG(mdlog.warning)
344 .appendTextFormatted(
345 "This run uses the 'GPU PME-PP communications' feature, enabled "
346 "by the GMX_GPU_PME_PP_COMMS environment variable.");
350 std::string clarification;
351 if (pmeRunMode == PmeRunMode::Mixed)
354 "PME FFT and gather are not offloaded to the GPU (PME is running in mixed "
359 clarification = "PME is not offloaded to the GPU.";
361 GMX_LOG(mdlog.warning)
364 "GMX_GPU_PME_PP_COMMS environment variable detected, but the "
365 "'GPU PME-PP communications' feature was not enabled as "
367 devFlags.enableGpuPmePPComm = false;
374 /*! \brief Barrier for safe simultaneous thread access to mdrunner data
376 * Used to ensure that the master thread does not modify mdrunner during copy
377 * on the spawned threads. */
378 static void threadMpiMdrunnerAccessBarrier()
381 MPI_Barrier(MPI_COMM_WORLD);
385 Mdrunner Mdrunner::cloneOnSpawnedThread() const
387 auto newRunner = Mdrunner(std::make_unique<MDModules>());
389 // All runners in the same process share a restraint manager resource because it is
390 // part of the interface to the client code, which is associated only with the
391 // original thread. Handles to the same resources can be obtained by copy.
393 newRunner.restraintManager_ = std::make_unique<RestraintManager>(*restraintManager_);
396 // Copy members of master runner.
397 // \todo Replace with builder when Simulation context and/or runner phases are better defined.
398 // Ref https://gitlab.com/gromacs/gromacs/-/issues/2587 and https://gitlab.com/gromacs/gromacs/-/issues/2375
399 newRunner.hw_opt = hw_opt;
400 newRunner.filenames = filenames;
402 newRunner.hwinfo_ = hwinfo_;
403 newRunner.oenv = oenv;
404 newRunner.mdrunOptions = mdrunOptions;
405 newRunner.domdecOptions = domdecOptions;
406 newRunner.nbpu_opt = nbpu_opt;
407 newRunner.pme_opt = pme_opt;
408 newRunner.pme_fft_opt = pme_fft_opt;
409 newRunner.bonded_opt = bonded_opt;
410 newRunner.update_opt = update_opt;
411 newRunner.nstlist_cmdline = nstlist_cmdline;
412 newRunner.replExParams = replExParams;
413 newRunner.pforce = pforce;
414 // Give the spawned thread the newly created valid communicator
415 // for the simulation.
416 newRunner.libraryWorldCommunicator = MPI_COMM_WORLD;
417 newRunner.simulationCommunicator = MPI_COMM_WORLD;
419 newRunner.startingBehavior = startingBehavior;
420 newRunner.stopHandlerBuilder_ = std::make_unique<StopHandlerBuilder>(*stopHandlerBuilder_);
421 newRunner.inputHolder_ = inputHolder_;
423 threadMpiMdrunnerAccessBarrier();
428 /*! \brief The callback used for running on spawned threads.
430 * Obtains the pointer to the master mdrunner object from the one
431 * argument permitted to the thread-launch API call, copies it to make
432 * a new runner for this thread, reinitializes necessary data, and
433 * proceeds to the simulation. */
434 static void mdrunner_start_fn(const void* arg)
438 const auto* masterMdrunner = reinterpret_cast<const gmx::Mdrunner*>(arg);
439 /* copy the arg list to make sure that it's thread-local. This
440 doesn't copy pointed-to items, of course; fnm, cr and fplog
441 are reset in the call below, all others should be const. */
442 gmx::Mdrunner mdrunner = masterMdrunner->cloneOnSpawnedThread();
445 GMX_CATCH_ALL_AND_EXIT_WITH_FATAL_ERROR
449 void Mdrunner::spawnThreads(int numThreadsToLaunch)
452 /* now spawn new threads that start mdrunner_start_fn(), while
453 the main thread returns. Thread affinity is handled later. */
454 if (tMPI_Init_fn(TRUE, numThreadsToLaunch, TMPI_AFFINITY_NONE, mdrunner_start_fn, static_cast<const void*>(this))
457 GMX_THROW(gmx::InternalError("Failed to spawn thread-MPI threads"));
460 // Give the master thread the newly created valid communicator for
462 libraryWorldCommunicator = MPI_COMM_WORLD;
463 simulationCommunicator = MPI_COMM_WORLD;
464 threadMpiMdrunnerAccessBarrier();
466 GMX_UNUSED_VALUE(numThreadsToLaunch);
467 GMX_UNUSED_VALUE(mdrunner_start_fn);
473 /*! \brief Initialize variables for Verlet scheme simulation */
474 static void prepare_verlet_scheme(FILE* fplog,
478 const gmx_mtop_t& mtop,
480 bool makeGpuPairList,
481 const gmx::CpuInfo& cpuinfo)
483 // We checked the cut-offs in grompp, but double-check here.
484 // We have PME+LJcutoff kernels for rcoulomb>rvdw.
485 if (EEL_PME_EWALD(ir->coulombtype) && ir->vdwtype == VanDerWaalsType::Cut)
487 GMX_RELEASE_ASSERT(ir->rcoulomb >= ir->rvdw,
488 "With Verlet lists and PME we should have rcoulomb>=rvdw");
492 GMX_RELEASE_ASSERT(ir->rcoulomb == ir->rvdw,
493 "With Verlet lists and no PME rcoulomb and rvdw should be identical");
495 /* For NVE simulations, we will retain the initial list buffer */
496 if (EI_DYNAMICS(ir->eI) && ir->verletbuf_tol > 0
497 && !(EI_MD(ir->eI) && ir->etc == TemperatureCoupling::No))
499 /* Update the Verlet buffer size for the current run setup */
501 /* Here we assume SIMD-enabled kernels are being used. But as currently
502 * calc_verlet_buffer_size gives the same results for 4x8 and 4x4
503 * and 4x2 gives a larger buffer than 4x4, this is ok.
505 ListSetupType listType =
506 (makeGpuPairList ? ListSetupType::Gpu : ListSetupType::CpuSimdWhenSupported);
507 VerletbufListSetup listSetup = verletbufGetSafeListSetup(listType);
509 const real rlist_new =
510 calcVerletBufferSize(mtop, det(box), *ir, ir->nstlist, ir->nstlist - 1, -1, listSetup);
512 if (rlist_new != ir->rlist)
514 if (fplog != nullptr)
517 "\nChanging rlist from %g to %g for non-bonded %dx%d atom kernels\n\n",
520 listSetup.cluster_size_i,
521 listSetup.cluster_size_j);
523 ir->rlist = rlist_new;
527 if (nstlist_cmdline > 0 && (!EI_DYNAMICS(ir->eI) || ir->verletbuf_tol <= 0))
530 "Can not set nstlist without %s",
531 !EI_DYNAMICS(ir->eI) ? "dynamics" : "verlet-buffer-tolerance");
534 if (EI_DYNAMICS(ir->eI))
536 /* Set or try nstlist values */
537 increaseNstlist(fplog, cr, ir, nstlist_cmdline, &mtop, box, makeGpuPairList, cpuinfo);
541 /*! \brief Override the nslist value in inputrec
543 * with value passed on the command line (if any)
545 static void override_nsteps_cmdline(const gmx::MDLogger& mdlog, int64_t nsteps_cmdline, t_inputrec* ir)
549 /* override with anything else than the default -2 */
550 if (nsteps_cmdline > -2)
552 char sbuf_steps[STEPSTRSIZE];
553 char sbuf_msg[STRLEN];
555 ir->nsteps = nsteps_cmdline;
556 if (EI_DYNAMICS(ir->eI) && nsteps_cmdline != -1)
559 "Overriding nsteps with value passed on the command line: %s steps, %.3g ps",
560 gmx_step_str(nsteps_cmdline, sbuf_steps),
561 fabs(nsteps_cmdline * ir->delta_t));
566 "Overriding nsteps with value passed on the command line: %s steps",
567 gmx_step_str(nsteps_cmdline, sbuf_steps));
570 GMX_LOG(mdlog.warning).asParagraph().appendText(sbuf_msg);
572 else if (nsteps_cmdline < -2)
574 gmx_fatal(FARGS, "Invalid nsteps value passed on the command line: %" PRId64, nsteps_cmdline);
576 /* Do nothing if nsteps_cmdline == -2 */
582 /*! \brief Return whether GPU acceleration of nonbondeds is supported with the given settings.
584 * If not, and if a warning may be issued, logs a warning about
585 * falling back to CPU code. With thread-MPI, only the first
586 * call to this function should have \c issueWarning true. */
587 static bool gpuAccelerationOfNonbondedIsUseful(const MDLogger& mdlog, const t_inputrec& ir, bool issueWarning)
589 bool gpuIsUseful = true;
592 if (ir.opts.ngener - ir.nwall > 1)
594 /* The GPU code does not support more than one energy group.
595 * If the user requested GPUs explicitly, a fatal error is given later.
599 "Multiple energy groups is not implemented for GPUs, falling back to the CPU. "
600 "For better performance, run on the GPU without energy groups and then do "
601 "gmx mdrun -rerun option on the trajectory with an energy group .tpr file.";
607 warning = "TPI is not implemented for GPUs.";
610 if (!gpuIsUseful && issueWarning)
612 GMX_LOG(mdlog.warning).asParagraph().appendText(warning);
618 //! Initializes the logger for mdrun.
619 static gmx::LoggerOwner buildLogger(FILE* fplog, const bool isSimulationMasterRank)
621 gmx::LoggerBuilder builder;
622 if (fplog != nullptr)
624 builder.addTargetFile(gmx::MDLogger::LogLevel::Info, fplog);
626 if (isSimulationMasterRank)
628 builder.addTargetStream(gmx::MDLogger::LogLevel::Warning, &gmx::TextOutputFile::standardError());
630 return builder.build();
633 //! Make a TaskTarget from an mdrun argument string.
634 static TaskTarget findTaskTarget(const char* optionString)
636 TaskTarget returnValue = TaskTarget::Auto;
638 if (strncmp(optionString, "auto", 3) == 0)
640 returnValue = TaskTarget::Auto;
642 else if (strncmp(optionString, "cpu", 3) == 0)
644 returnValue = TaskTarget::Cpu;
646 else if (strncmp(optionString, "gpu", 3) == 0)
648 returnValue = TaskTarget::Gpu;
652 GMX_ASSERT(false, "Option string should have been checked for sanity already");
658 //! Finish run, aggregate data to print performance info.
659 static void finish_run(FILE* fplog,
660 const gmx::MDLogger& mdlog,
662 const t_inputrec& inputrec,
664 gmx_wallcycle* wcycle,
665 gmx_walltime_accounting_t walltime_accounting,
666 nonbonded_verlet_t* nbv,
667 const gmx_pme_t* pme,
671 double nbfs = 0, mflop = 0;
672 double elapsed_time, elapsed_time_over_all_ranks, elapsed_time_over_all_threads,
673 elapsed_time_over_all_threads_over_all_ranks;
674 /* Control whether it is valid to print a report. Only the
675 simulation master may print, but it should not do so if the run
676 terminated e.g. before a scheduled reset step. This is
677 complicated by the fact that PME ranks are unaware of the
678 reason why they were sent a pmerecvqxFINISH. To avoid
679 communication deadlocks, we always do the communication for the
680 report, even if we've decided not to write the report, because
681 how long it takes to finish the run is not important when we've
682 decided not to report on the simulation performance.
684 Further, we only report performance for dynamical integrators,
685 because those are the only ones for which we plan to
686 consider doing any optimizations. */
687 bool printReport = EI_DYNAMICS(inputrec.eI) && SIMMASTER(cr);
689 if (printReport && !walltime_accounting_get_valid_finish(walltime_accounting))
691 GMX_LOG(mdlog.warning)
693 .appendText("Simulation ended prematurely, no performance report will be written.");
698 std::unique_ptr<t_nrnb> nrnbTotalStorage;
701 nrnbTotalStorage = std::make_unique<t_nrnb>();
702 nrnb_tot = nrnbTotalStorage.get();
704 MPI_Allreduce(nrnb->n.data(), nrnb_tot->n.data(), eNRNB, MPI_DOUBLE, MPI_SUM, cr->mpi_comm_mysim);
712 elapsed_time = walltime_accounting_get_time_since_reset(walltime_accounting);
713 elapsed_time_over_all_threads =
714 walltime_accounting_get_time_since_reset_over_all_threads(walltime_accounting);
718 /* reduce elapsed_time over all MPI ranks in the current simulation */
719 MPI_Allreduce(&elapsed_time, &elapsed_time_over_all_ranks, 1, MPI_DOUBLE, MPI_SUM, cr->mpi_comm_mysim);
720 elapsed_time_over_all_ranks /= cr->nnodes;
721 /* Reduce elapsed_time_over_all_threads over all MPI ranks in the
722 * current simulation. */
723 MPI_Allreduce(&elapsed_time_over_all_threads,
724 &elapsed_time_over_all_threads_over_all_ranks,
733 elapsed_time_over_all_ranks = elapsed_time;
734 elapsed_time_over_all_threads_over_all_ranks = elapsed_time_over_all_threads;
739 print_flop(fplog, nrnb_tot, &nbfs, &mflop);
742 if (thisRankHasDuty(cr, DUTY_PP) && haveDDAtomOrdering(*cr))
744 print_dd_statistics(cr, inputrec, fplog);
747 /* TODO Move the responsibility for any scaling by thread counts
748 * to the code that handled the thread region, so that there's a
749 * mechanism to keep cycle counting working during the transition
750 * to task parallelism. */
751 int nthreads_pp = gmx_omp_nthreads_get(ModuleMultiThread::Nonbonded);
752 int nthreads_pme = gmx_omp_nthreads_get(ModuleMultiThread::Pme);
753 wallcycle_scale_by_num_threads(
754 wcycle, thisRankHasDuty(cr, DUTY_PME) && !thisRankHasDuty(cr, DUTY_PP), nthreads_pp, nthreads_pme);
755 auto cycle_sum(wallcycle_sum(cr, wcycle));
759 auto* nbnxn_gpu_timings =
760 (nbv != nullptr && nbv->useGpu()) ? Nbnxm::gpu_get_timings(nbv->gpu_nbv) : nullptr;
761 gmx_wallclock_gpu_pme_t pme_gpu_timings = {};
763 if (pme_gpu_task_enabled(pme))
765 pme_gpu_get_timings(pme, &pme_gpu_timings);
767 wallcycle_print(fplog,
773 elapsed_time_over_all_ranks,
779 if (EI_DYNAMICS(inputrec.eI))
781 delta_t = inputrec.delta_t;
787 elapsed_time_over_all_threads_over_all_ranks,
788 elapsed_time_over_all_ranks,
789 walltime_accounting_get_nsteps_done_since_reset(walltime_accounting),
797 elapsed_time_over_all_threads_over_all_ranks,
798 elapsed_time_over_all_ranks,
799 walltime_accounting_get_nsteps_done_since_reset(walltime_accounting),
807 int Mdrunner::mdrunner()
810 std::unique_ptr<t_forcerec> fr;
811 real ewaldcoeff_q = 0;
812 real ewaldcoeff_lj = 0;
813 int nChargePerturbed = -1, nTypePerturbed = 0;
814 gmx_walltime_accounting_t walltime_accounting = nullptr;
815 MembedHolder membedHolder(filenames.size(), filenames.data());
817 /* CAUTION: threads may be started later on in this function, so
818 cr doesn't reflect the final parallel state right now */
821 /* TODO: inputrec should tell us whether we use an algorithm, not a file option */
822 const bool doEssentialDynamics = opt2bSet("-ei", filenames.size(), filenames.data());
823 const bool doRerun = mdrunOptions.rerun;
825 // Handle task-assignment related user options.
826 EmulateGpuNonbonded emulateGpuNonbonded =
827 (getenv("GMX_EMULATE_GPU") != nullptr ? EmulateGpuNonbonded::Yes : EmulateGpuNonbonded::No);
829 std::vector<int> userGpuTaskAssignment;
832 userGpuTaskAssignment = parseUserTaskAssignmentString(hw_opt.userGpuTaskAssignment);
834 GMX_CATCH_ALL_AND_EXIT_WITH_FATAL_ERROR
835 auto nonbondedTarget = findTaskTarget(nbpu_opt);
836 auto pmeTarget = findTaskTarget(pme_opt);
837 auto pmeFftTarget = findTaskTarget(pme_fft_opt);
838 auto bondedTarget = findTaskTarget(bonded_opt);
839 auto updateTarget = findTaskTarget(update_opt);
841 FILE* fplog = nullptr;
842 // If we are appending, we don't write log output because we need
843 // to check that the old log file matches what the checkpoint file
844 // expects. Otherwise, we should start to write log output now if
845 // there is a file ready for it.
846 if (logFileHandle != nullptr && startingBehavior != StartingBehavior::RestartWithAppending)
848 fplog = gmx_fio_getfp(logFileHandle);
850 const bool isSimulationMasterRank = findIsSimulationMasterRank(ms, simulationCommunicator);
851 gmx::LoggerOwner logOwner(buildLogger(fplog, isSimulationMasterRank));
852 gmx::MDLogger mdlog(logOwner.logger());
854 gmx_print_detected_hardware(fplog, isSimulationMasterRank && isMasterSim(ms), mdlog, hwinfo_);
856 std::vector<int> availableDevices =
857 makeListOfAvailableDevices(hwinfo_->deviceInfoList, hw_opt.devicesSelectedByUser);
858 const int numAvailableDevices = gmx::ssize(availableDevices);
860 // Print citation requests after all software/hardware printing
861 pleaseCiteGromacs(fplog);
863 // Note: legacy program logic relies on checking whether these pointers are assigned.
864 // Objects may or may not be allocated later.
865 std::unique_ptr<t_inputrec> inputrec;
866 std::unique_ptr<t_state> globalState;
868 auto partialDeserializedTpr = std::make_unique<PartialDeserializedTprFile>();
870 if (isSimulationMasterRank)
872 // Allocate objects to be initialized by later function calls.
873 /* Only the master rank has the global state */
874 globalState = std::make_unique<t_state>();
875 inputrec = std::make_unique<t_inputrec>();
877 /* Read (nearly) all data required for the simulation
878 * and keep the partly serialized tpr contents to send to other ranks later
880 applyGlobalSimulationState(
881 *inputHolder_.get(), partialDeserializedTpr.get(), globalState.get(), inputrec.get(), &mtop);
884 /* Check and update the hardware options for internal consistency */
885 checkAndUpdateHardwareOptions(
886 mdlog, &hw_opt, isSimulationMasterRank, domdecOptions.numPmeRanks, inputrec.get());
888 if (GMX_THREAD_MPI && isSimulationMasterRank)
890 bool useGpuForNonbonded = false;
891 bool useGpuForPme = false;
894 GMX_RELEASE_ASSERT(inputrec != nullptr, "Keep the compiler happy");
896 // If the user specified the number of ranks, then we must
897 // respect that, but in default mode, we need to allow for
898 // the number of GPUs to choose the number of ranks.
899 auto canUseGpuForNonbonded = buildSupportsNonbondedOnGpu(nullptr);
900 useGpuForNonbonded = decideWhetherToUseGpusForNonbondedWithThreadMpi(
902 numAvailableDevices > 0,
903 userGpuTaskAssignment,
905 canUseGpuForNonbonded,
906 gpuAccelerationOfNonbondedIsUseful(mdlog, *inputrec, GMX_THREAD_MPI),
907 hw_opt.nthreads_tmpi);
908 useGpuForPme = decideWhetherToUseGpusForPmeWithThreadMpi(useGpuForNonbonded,
912 userGpuTaskAssignment,
915 hw_opt.nthreads_tmpi,
916 domdecOptions.numPmeRanks);
918 GMX_CATCH_ALL_AND_EXIT_WITH_FATAL_ERROR
920 /* Determine how many thread-MPI ranks to start.
922 * TODO Over-writing the user-supplied value here does
923 * prevent any possible subsequent checks from working
925 hw_opt.nthreads_tmpi = get_nthreads_mpi(hwinfo_,
933 membedHolder.doMembed());
935 // Now start the threads for thread MPI.
936 spawnThreads(hw_opt.nthreads_tmpi);
937 // The spawned threads enter mdrunner() and execution of
938 // master and spawned threads joins at the end of this block.
941 GMX_RELEASE_ASSERT(!GMX_MPI || ms || simulationCommunicator != MPI_COMM_NULL,
942 "Must have valid communicator unless running a multi-simulation");
943 CommrecHandle crHandle = init_commrec(simulationCommunicator);
944 t_commrec* cr = crHandle.get();
945 GMX_RELEASE_ASSERT(cr != nullptr, "Must have valid commrec");
947 PhysicalNodeCommunicator physicalNodeComm(libraryWorldCommunicator, gmx_physicalnode_id_hash());
949 // If we detected the topology on this system, double-check that it makes sense
950 if (hwinfo_->hardwareTopology->isThisSystem())
952 hardwareTopologyDoubleCheckDetection(mdlog, *hwinfo_->hardwareTopology);
957 /* now broadcast everything to the non-master nodes/threads: */
958 if (!isSimulationMasterRank)
960 // Until now, only the master rank has a non-null pointer.
961 // On non-master ranks, allocate the object that will receive data in the following call.
962 inputrec = std::make_unique<t_inputrec>();
964 init_parallel(cr->mpiDefaultCommunicator,
968 partialDeserializedTpr.get());
970 GMX_RELEASE_ASSERT(inputrec != nullptr, "All ranks should have a valid inputrec now");
971 partialDeserializedTpr.reset(nullptr);
974 !inputrec->useConstantAcceleration,
975 "Linear acceleration has been removed in GROMACS 2022, and was broken for many years "
976 "before that. Use GROMACS 4.5 or earlier if you need this feature.");
978 // Now we decide whether to use the domain decomposition machinery.
979 // Note that this does not necessarily imply actually using multiple domains.
980 // Now the number of ranks is known to all ranks, and each knows
981 // the inputrec read by the master rank. The ranks can now all run
982 // the task-deciding functions and will agree on the result
983 // without needing to communicate.
984 // The LBFGS minimizer, test-particle insertion, normal modes and shell dynamics don't support DD
985 const bool useDomainDecomposition =
986 !(inputrec->eI == IntegrationAlgorithm::LBFGS || EI_TPI(inputrec->eI)
987 || inputrec->eI == IntegrationAlgorithm::NM
988 || gmx_mtop_particletype_count(mtop)[ParticleType::Shell] > 0);
990 // Note that these variables describe only their own node.
992 // Note that when bonded interactions run on a GPU they always run
993 // alongside a nonbonded task, so do not influence task assignment
994 // even though they affect the force calculation workload.
995 bool useGpuForNonbonded = false;
996 bool useGpuForPme = false;
997 bool useGpuForBonded = false;
998 bool useGpuForUpdate = false;
999 bool gpusWereDetected = hwinfo_->ngpu_compatible_tot > 0;
1002 // It's possible that there are different numbers of GPUs on
1003 // different nodes, which is the user's responsibility to
1004 // handle. If unsuitable, we will notice that during task
1006 auto canUseGpuForNonbonded = buildSupportsNonbondedOnGpu(nullptr);
1007 useGpuForNonbonded = decideWhetherToUseGpusForNonbonded(
1009 userGpuTaskAssignment,
1010 emulateGpuNonbonded,
1011 canUseGpuForNonbonded,
1012 gpuAccelerationOfNonbondedIsUseful(mdlog, *inputrec, !GMX_THREAD_MPI),
1014 useGpuForPme = decideWhetherToUseGpusForPme(useGpuForNonbonded,
1017 userGpuTaskAssignment,
1020 cr->sizeOfDefaultCommunicator,
1021 domdecOptions.numPmeRanks,
1023 useGpuForBonded = decideWhetherToUseGpusForBonded(
1024 useGpuForNonbonded, useGpuForPme, bondedTarget, *inputrec, mtop, domdecOptions.numPmeRanks, gpusWereDetected);
1026 GMX_CATCH_ALL_AND_EXIT_WITH_FATAL_ERROR
1028 const PmeRunMode pmeRunMode = determinePmeRunMode(useGpuForPme, pmeFftTarget, *inputrec);
1030 // Initialize development feature flags that enabled by environment variable
1031 // and report those features that are enabled.
1032 const DevelopmentFeatureFlags devFlags =
1033 manageDevelopmentFeatures(mdlog, useGpuForNonbonded, pmeRunMode);
1035 const bool useModularSimulator = checkUseModularSimulator(false,
1042 doEssentialDynamics,
1043 membedHolder.doMembed());
1045 ObservablesReducerBuilder observablesReducerBuilder;
1047 // Build restraints.
1048 // TODO: hide restraint implementation details from Mdrunner.
1049 // There is nothing unique about restraints at this point as far as the
1050 // Mdrunner is concerned. The Mdrunner should just be getting a sequence of
1051 // factory functions from the SimulationContext on which to call mdModules_->add().
1052 // TODO: capture all restraints into a single RestraintModule, passed to the runner builder.
1053 for (auto&& restraint : restraintManager_->getRestraints())
1055 auto module = RestraintMDModule::create(restraint, restraint->sites());
1056 mdModules_->add(std::move(module));
1059 // TODO: Error handling
1060 mdModules_->assignOptionsToModules(*inputrec->params, nullptr);
1061 // now that the MDModules know their options, they know which callbacks to sign up to
1062 mdModules_->subscribeToSimulationSetupNotifications();
1063 const auto& setupNotifier = mdModules_->notifiers().simulationSetupNotifier_;
1065 // Notify MdModules of existing logger
1066 setupNotifier.notify(mdlog);
1068 // Notify MdModules of internal parameters, saved into KVT
1069 if (inputrec->internalParameters != nullptr)
1071 setupNotifier.notify(*inputrec->internalParameters);
1074 // Let MdModules know the .tpr filename
1076 gmx::MdRunInputFilename mdRunInputFilename = { ftp2fn(efTPR, filenames.size(), filenames.data()) };
1077 setupNotifier.notify(mdRunInputFilename);
1080 if (fplog != nullptr)
1082 pr_inputrec(fplog, 0, "Input Parameters", inputrec.get(), FALSE);
1083 fprintf(fplog, "\n");
1088 /* In rerun, set velocities to zero if present */
1089 if (doRerun && ((globalState->flags & enumValueToBitMask(StateEntry::V)) != 0))
1091 // rerun does not use velocities
1095 "Rerun trajectory contains velocities. Rerun does only evaluate "
1096 "potential energy and forces. The velocities will be ignored.");
1097 for (int i = 0; i < globalState->natoms; i++)
1099 clear_rvec(globalState->v[i]);
1101 globalState->flags &= ~enumValueToBitMask(StateEntry::V);
1104 /* now make sure the state is initialized and propagated */
1105 set_state_entries(globalState.get(), inputrec.get(), useModularSimulator);
1108 /* NM and TPI parallelize over force/energy calculations, not atoms,
1109 * so we need to initialize and broadcast the global state.
1111 if (inputrec->eI == IntegrationAlgorithm::NM || inputrec->eI == IntegrationAlgorithm::TPI)
1115 globalState = std::make_unique<t_state>();
1117 broadcastStateWithoutDynamics(
1118 cr->mpiDefaultCommunicator, haveDDAtomOrdering(*cr), PAR(cr), globalState.get());
1121 /* A parallel command line option consistency check that we can
1122 only do after any threads have started. */
1124 && (domdecOptions.numCells[XX] > 1 || domdecOptions.numCells[YY] > 1
1125 || domdecOptions.numCells[ZZ] > 1 || domdecOptions.numPmeRanks > 0))
1128 "The -dd or -npme option request a parallel simulation, "
1130 "but %s was compiled without threads or MPI enabled",
1131 output_env_get_program_display_name(oenv));
1132 #elif GMX_THREAD_MPI
1133 "but the number of MPI-threads (option -ntmpi) is not set or is 1");
1135 "but %s was not started through mpirun/mpiexec or only one rank was requested "
1136 "through mpirun/mpiexec",
1137 output_env_get_program_display_name(oenv));
1141 if (doRerun && (EI_ENERGY_MINIMIZATION(inputrec->eI) || IntegrationAlgorithm::NM == inputrec->eI))
1144 "The .mdp file specified an energy mininization or normal mode algorithm, and "
1145 "these are not compatible with mdrun -rerun");
1148 /* NMR restraints must be initialized before load_checkpoint,
1149 * since with time averaging the history is added to t_state.
1150 * For proper consistency check we therefore need to extend
1152 * So the PME-only nodes (if present) will also initialize
1153 * the distance restraints.
1156 /* This needs to be called before read_checkpoint to extend the state */
1157 t_disresdata* disresdata;
1158 snew(disresdata, 1);
1162 DisResRunMode::MDRun,
1163 MASTER(cr) ? DDRole::Master : DDRole::Agent,
1164 PAR(cr) ? NumRanks::Multiple : NumRanks::Single,
1169 replExParams.exchangeInterval > 0);
1171 if (gmx_mtop_ftype_count(mtop, F_ORIRES) > 0 && isSimulationMasterRank)
1173 extendStateWithOriresHistory(mtop, *inputrec, globalState.get());
1176 auto deform = prepareBoxDeformation(globalState != nullptr ? globalState->box : box,
1177 MASTER(cr) ? DDRole::Master : DDRole::Agent,
1178 PAR(cr) ? NumRanks::Multiple : NumRanks::Single,
1179 cr->mpi_comm_mygroup,
1183 /* We have to remember the generation's first step before reading checkpoint.
1184 This way, we can report to the F@H core both the generation's first step
1185 and the restored first step, thus making it able to distinguish between
1186 an interruption/resume and start of the n-th generation simulation.
1187 Having this information, the F@H core can correctly calculate and report
1190 int gen_first_step = 0;
1193 gen_first_step = inputrec->init_step;
1197 ObservablesHistory observablesHistory = {};
1199 auto modularSimulatorCheckpointData = std::make_unique<ReadCheckpointDataHolder>();
1200 if (startingBehavior != StartingBehavior::NewSimulation)
1202 /* Check if checkpoint file exists before doing continuation.
1203 * This way we can use identical input options for the first and subsequent runs...
1205 if (mdrunOptions.numStepsCommandline > -2)
1207 /* Temporarily set the number of steps to unlimited to avoid
1208 * triggering the nsteps check in load_checkpoint().
1209 * This hack will go away soon when the -nsteps option is removed.
1211 inputrec->nsteps = -1;
1214 // Finish applying initial simulation state information from external sources on all ranks.
1215 // Reconcile checkpoint file data with Mdrunner state established up to this point.
1216 applyLocalState(*inputHolder_.get(),
1219 domdecOptions.numCells,
1222 &observablesHistory,
1223 mdrunOptions.reproducible,
1224 mdModules_->notifiers(),
1225 modularSimulatorCheckpointData.get(),
1226 useModularSimulator);
1227 // TODO: (#3652) Synchronize filesystem state, SimulationInput contents, and program
1229 // on all code paths.
1230 // Write checkpoint or provide hook to update SimulationInput.
1231 // If there was a checkpoint file, SimulationInput contains more information
1232 // than if there wasn't. At this point, we have synchronized the in-memory
1233 // state with the filesystem state only for restarted simulations. We should
1234 // be calling applyLocalState unconditionally and expect that the completeness
1235 // of SimulationInput is not dependent on its creation method.
1237 if (startingBehavior == StartingBehavior::RestartWithAppending && logFileHandle)
1239 // Now we can start normal logging to the truncated log file.
1240 fplog = gmx_fio_getfp(logFileHandle);
1241 prepareLogAppending(fplog);
1242 logOwner = buildLogger(fplog, MASTER(cr));
1243 mdlog = logOwner.logger();
1250 fcRegisterSteps(inputrec->nsteps + inputrec->init_step, gen_first_step);
1254 if (mdrunOptions.numStepsCommandline > -2)
1259 "The -nsteps functionality is deprecated, and may be removed in a future "
1261 "Consider using gmx convert-tpr -nsteps or changing the appropriate .mdp "
1264 /* override nsteps with value set on the commandline */
1265 override_nsteps_cmdline(mdlog, mdrunOptions.numStepsCommandline, inputrec.get());
1267 if (isSimulationMasterRank)
1269 copy_mat(globalState->box, box);
1274 gmx_bcast(sizeof(box), box, cr->mpiDefaultCommunicator);
1277 if (inputrec->cutoff_scheme != CutoffScheme::Verlet)
1280 "This group-scheme .tpr file can no longer be run by mdrun. Please update to the "
1281 "Verlet scheme, or use an earlier version of GROMACS if necessary.");
1283 /* Update rlist and nstlist. */
1284 /* Note: prepare_verlet_scheme is calling increaseNstlist(...), which (while attempting to
1285 * increase rlist) tries to check if the newly chosen value fits with the DD scheme. As this is
1286 * run before any DD scheme is set up, this check is never executed. See #3334 for more details.
1288 prepare_verlet_scheme(fplog,
1294 useGpuForNonbonded || (emulateGpuNonbonded == EmulateGpuNonbonded::Yes),
1297 // We need to decide on update groups early, as this affects
1298 // inter-domain communication distances.
1299 auto updateGroupingsPerMoleculeType = makeUpdateGroupingsPerMoleculeType(mtop);
1300 const real maxUpdateGroupRadius = computeMaxUpdateGroupRadius(
1301 mtop, updateGroupingsPerMoleculeType, maxReferenceTemperature(*inputrec));
1302 const real cutoffMargin = std::sqrt(max_cutoff2(inputrec->pbcType, box)) - inputrec->rlist;
1303 UpdateGroups updateGroups = makeUpdateGroups(mdlog,
1304 std::move(updateGroupingsPerMoleculeType),
1305 maxUpdateGroupRadius,
1306 useDomainDecomposition,
1307 systemHasConstraintsOrVsites(mtop),
1312 const bool haveFrozenAtoms = inputrecFrozenAtoms(inputrec.get());
1314 useGpuForUpdate = decideWhetherToUseGpuForUpdate(useDomainDecomposition,
1315 updateGroups.useUpdateGroups(),
1317 domdecOptions.numPmeRanks > 0,
1323 doEssentialDynamics,
1324 gmx_mtop_ftype_count(mtop, F_ORIRES) > 0,
1330 GMX_CATCH_ALL_AND_EXIT_WITH_FATAL_ERROR
1332 bool useGpuDirectHalo = false;
1334 if (useGpuForNonbonded)
1336 // cr->npmenodes is not yet initialized.
1337 // domdecOptions.numPmeRanks == -1 results in 0 separate PME ranks when useGpuForNonbonded is true.
1338 // Todo: remove this assumption later once auto mode has support for separate PME rank
1339 const int numPmeRanks = domdecOptions.numPmeRanks > 0 ? domdecOptions.numPmeRanks : 0;
1340 bool havePPDomainDecomposition = (cr->sizeOfDefaultCommunicator - numPmeRanks) > 1;
1341 useGpuDirectHalo = decideWhetherToUseGpuForHalo(devFlags,
1342 havePPDomainDecomposition,
1344 useModularSimulator,
1346 EI_ENERGY_MINIMIZATION(inputrec->eI));
1349 // This builder is necessary while we have multi-part construction
1350 // of DD. Before DD is constructed, we use the existence of
1351 // the builder object to indicate that further construction of DD
1353 std::unique_ptr<DomainDecompositionBuilder> ddBuilder;
1354 if (useDomainDecomposition)
1356 // P2P GPU comm + GPU update leads to case in which we enqueue async work for multiple
1357 // timesteps. DLB needs to be disabled in that case
1358 const bool directGpuCommUsedWithGpuUpdate = GMX_THREAD_MPI && useGpuDirectHalo && useGpuForUpdate;
1359 ddBuilder = std::make_unique<DomainDecompositionBuilder>(
1366 mdModules_->notifiers(),
1368 updateGroups.updateGroupingPerMoleculeType(),
1369 updateGroups.useUpdateGroups(),
1370 updateGroups.maxUpdateGroupRadius(),
1371 positionsFromStatePointer(globalState.get()),
1374 directGpuCommUsedWithGpuUpdate);
1378 /* PME, if used, is done on all nodes with 1D decomposition */
1379 cr->nnodes = cr->sizeOfDefaultCommunicator;
1380 cr->sim_nodeid = cr->rankInDefaultCommunicator;
1381 cr->nodeid = cr->rankInDefaultCommunicator;
1383 cr->duty = (DUTY_PP | DUTY_PME);
1385 if (inputrec->pbcType == PbcType::Screw)
1387 gmx_fatal(FARGS, "pbc=screw is only implemented with domain decomposition");
1391 // Produce the task assignment for this rank - done after DD is constructed
1392 GpuTaskAssignments gpuTaskAssignments = GpuTaskAssignmentsBuilder::build(
1394 userGpuTaskAssignment,
1396 simulationCommunicator,
1404 thisRankHasDuty(cr, DUTY_PP),
1405 // TODO cr->duty & DUTY_PME should imply that a PME
1406 // algorithm is active, but currently does not.
1407 EEL_PME(inputrec->coulombtype) && thisRankHasDuty(cr, DUTY_PME));
1409 // Get the device handles for the modules, nullptr when no task is assigned.
1411 DeviceInformation* deviceInfo = gpuTaskAssignments.initDevice(&deviceId);
1413 // timing enabling - TODO put this in gpu_utils (even though generally this is just option handling?)
1414 bool useTiming = true;
1418 /* WARNING: CUDA timings are incorrect with multiple streams.
1419 * This is the main reason why they are disabled by default.
1421 // TODO: Consider turning on by default when we can detect nr of streams.
1422 useTiming = (getenv("GMX_ENABLE_GPU_TIMING") != nullptr);
1424 else if (GMX_GPU_OPENCL)
1426 useTiming = (getenv("GMX_DISABLE_GPU_TIMING") == nullptr);
1429 // TODO Currently this is always built, yet DD partition code
1430 // checks if it is built before using it. Probably it should
1431 // become an MDModule that is made only when another module
1432 // requires it (e.g. pull, CompEl, density fitting), so that we
1433 // don't update the local atom sets unilaterally every step.
1434 LocalAtomSetManager atomSets;
1436 // Local state and topology are declared (and perhaps constructed)
1437 // now, because DD needs them for the LocalTopologyChecker, but
1438 // they do not contain valid data until after the first DD
1440 std::unique_ptr<t_state> localStateInstance;
1441 t_state* localState;
1442 gmx_localtop_t localTopology(mtop.ffparams);
1446 localStateInstance = std::make_unique<t_state>();
1447 localState = localStateInstance.get();
1448 // TODO Pass the GPU streams to ddBuilder to use in buffer
1449 // transfers (e.g. halo exchange)
1450 cr->dd = ddBuilder->build(&atomSets, localTopology, *localState, &observablesReducerBuilder);
1451 // The builder's job is done, so destruct it
1452 ddBuilder.reset(nullptr);
1453 // Note that local state still does not exist yet.
1457 // Without DD, the local state is merely an alias to the global state,
1458 // so we don't need to allocate anything.
1459 localState = globalState.get();
1462 // Ensure that all atoms within the same update group are in the
1463 // same periodic image. Otherwise, a simulation that did not use
1464 // update groups (e.g. a single-rank simulation) cannot always be
1465 // correctly restarted in a way that does use update groups
1466 // (e.g. a multi-rank simulation).
1467 if (isSimulationMasterRank)
1469 const bool useUpdateGroups = cr->dd ? ddUsesUpdateGroups(*cr->dd) : false;
1470 if (useUpdateGroups)
1472 putUpdateGroupAtomsInSamePeriodicImage(*cr->dd, mtop, globalState->box, globalState->x);
1476 const bool printHostName = (cr->nnodes > 1);
1477 gpuTaskAssignments.reportGpuUsage(mdlog, printHostName, useGpuForBonded, pmeRunMode, useGpuForUpdate);
1479 const bool disableNonbondedCalculation = (getenv("GMX_NO_NONBONDED") != nullptr);
1480 if (disableNonbondedCalculation)
1482 /* turn off non-bonded calculations */
1483 GMX_LOG(mdlog.warning)
1486 "Found environment variable GMX_NO_NONBONDED.\n"
1487 "Disabling nonbonded calculations.");
1490 MdrunScheduleWorkload runScheduleWork;
1492 // Also populates the simulation constant workload description.
1493 // Note: currently the default duty is DUTY_PP | DUTY_PME for all simulations, including those without PME,
1494 // so this boolean is sufficient on all ranks to determine whether separate PME ranks are used,
1495 // but this will no longer be the case if cr->duty is changed for !EEL_PME(fr->ic->eeltype).
1496 const bool haveSeparatePmeRank = (!thisRankHasDuty(cr, DUTY_PP) || !thisRankHasDuty(cr, DUTY_PME));
1497 runScheduleWork.simulationWork = createSimulationWorkload(*inputrec,
1498 disableNonbondedCalculation,
1500 havePPDomainDecomposition(cr),
1501 haveSeparatePmeRank,
1508 std::unique_ptr<DeviceStreamManager> deviceStreamManager = nullptr;
1510 if (deviceInfo != nullptr)
1512 if (runScheduleWork.simulationWork.havePpDomainDecomposition && thisRankHasDuty(cr, DUTY_PP))
1514 dd_setup_dlb_resource_sharing(cr, deviceId);
1516 deviceStreamManager = std::make_unique<DeviceStreamManager>(
1517 *deviceInfo, havePPDomainDecomposition(cr), runScheduleWork.simulationWork, useTiming);
1520 // If the user chose a task assignment, give them some hints
1521 // where appropriate.
1522 if (!userGpuTaskAssignment.empty())
1524 gpuTaskAssignments.logPerformanceHints(mdlog, numAvailableDevices);
1529 /* After possible communicator splitting in make_dd_communicators.
1530 * we can set up the intra/inter node communication.
1532 gmx_setup_nodecomm(fplog, cr);
1538 GMX_LOG(mdlog.warning)
1540 .appendTextFormatted(
1541 "This is simulation %d out of %d running as a composite GROMACS\n"
1542 "multi-simulation job. Setup for this simulation:\n",
1543 ms->simulationIndex_,
1544 ms->numSimulations_);
1546 GMX_LOG(mdlog.warning)
1547 .appendTextFormatted("Using %d MPI %s\n",
1550 cr->nnodes == 1 ? "thread" : "threads"
1552 cr->nnodes == 1 ? "process" : "processes"
1558 // If mdrun -pin auto honors any affinity setting that already
1559 // exists. If so, it is nice to provide feedback about whether
1560 // that existing affinity setting was from OpenMP or something
1561 // else, so we run this code both before and after we initialize
1562 // the OpenMP support.
1563 gmx_check_thread_affinity_set(mdlog, &hw_opt, hwinfo_->nthreads_hw_avail, FALSE);
1564 /* Check and update the number of OpenMP threads requested */
1565 checkAndUpdateRequestedNumOpenmpThreads(
1566 &hw_opt, *hwinfo_, cr, ms, physicalNodeComm.size_, pmeRunMode, mtop, *inputrec);
1568 gmx_omp_nthreads_init(mdlog,
1570 hwinfo_->nthreads_hw_avail,
1571 physicalNodeComm.size_,
1572 hw_opt.nthreads_omp,
1573 hw_opt.nthreads_omp_pme,
1574 !thisRankHasDuty(cr, DUTY_PP));
1576 const bool bEnableFPE = gmxShouldEnableFPExceptions();
1577 // FIXME - reconcile with gmx_feenableexcept() call from CommandLineModuleManager::run()
1580 gmx_feenableexcept();
1583 /* Now that we know the setup is consistent, check for efficiency */
1584 check_resource_division_efficiency(
1585 hwinfo_, gpuTaskAssignments.thisRankHasAnyGpuTask(), mdrunOptions.ntompOptionIsSet, cr, mdlog);
1587 /* getting number of PP/PME threads on this MPI / tMPI rank.
1588 PME: env variable should be read only on one node to make sure it is
1589 identical everywhere;
1591 const int numThreadsOnThisRank = thisRankHasDuty(cr, DUTY_PP)
1592 ? gmx_omp_nthreads_get(ModuleMultiThread::Nonbonded)
1593 : gmx_omp_nthreads_get(ModuleMultiThread::Pme);
1594 checkHardwareOversubscription(
1595 numThreadsOnThisRank, cr->nodeid, *hwinfo_->hardwareTopology, physicalNodeComm, mdlog);
1597 // Enable Peer access between GPUs where available
1598 // Only for DD, only master PP rank needs to perform setup, and only if thread MPI plus
1599 // any of the GPU communication features are active.
1600 if (haveDDAtomOrdering(*cr) && MASTER(cr) && thisRankHasDuty(cr, DUTY_PP) && GMX_THREAD_MPI
1601 && (runScheduleWork.simulationWork.useGpuHaloExchange
1602 || runScheduleWork.simulationWork.useGpuPmePpCommunication))
1604 setupGpuDevicePeerAccess(gpuTaskAssignments.deviceIdsAssigned(), mdlog);
1607 if (hw_opt.threadAffinity != ThreadAffinity::Off)
1609 /* Before setting affinity, check whether the affinity has changed
1610 * - which indicates that probably the OpenMP library has changed it
1611 * since we first checked).
1613 gmx_check_thread_affinity_set(mdlog, &hw_opt, hwinfo_->nthreads_hw_avail, TRUE);
1615 int numThreadsOnThisNode, intraNodeThreadOffset;
1616 analyzeThreadsOnThisNode(
1617 physicalNodeComm, numThreadsOnThisRank, &numThreadsOnThisNode, &intraNodeThreadOffset);
1619 /* Set the CPU affinity */
1620 gmx_set_thread_affinity(mdlog,
1623 *hwinfo_->hardwareTopology,
1624 numThreadsOnThisRank,
1625 numThreadsOnThisNode,
1626 intraNodeThreadOffset,
1630 if (mdrunOptions.timingOptions.resetStep > -1)
1635 "The -resetstep functionality is deprecated, and may be removed in a "
1638 std::unique_ptr<gmx_wallcycle> wcycle =
1639 wallcycle_init(fplog, mdrunOptions.timingOptions.resetStep, cr);
1643 /* Master synchronizes its value of reset_counters with all nodes
1644 * including PME only nodes */
1645 int64_t reset_counters = wcycle_get_reset_counters(wcycle.get());
1646 gmx_bcast(sizeof(reset_counters), &reset_counters, cr->mpi_comm_mysim);
1647 wcycle_set_reset_counters(wcycle.get(), reset_counters);
1650 // Membrane embedding must be initialized before we call init_forcerec()
1651 membedHolder.initializeMembed(fplog,
1658 &mdrunOptions.checkpointOptions.period);
1660 const bool thisRankHasPmeGpuTask = gpuTaskAssignments.thisRankHasPmeGpuTask();
1661 std::unique_ptr<MDAtoms> mdAtoms;
1662 std::unique_ptr<VirtualSitesHandler> vsite;
1665 if (thisRankHasDuty(cr, DUTY_PP))
1667 setupNotifier.notify(*cr);
1668 setupNotifier.notify(&atomSets);
1669 setupNotifier.notify(mtop);
1670 setupNotifier.notify(inputrec->pbcType);
1671 setupNotifier.notify(SimulationTimeStep{ inputrec->delta_t });
1672 /* Initiate forcerecord */
1673 fr = std::make_unique<t_forcerec>();
1674 fr->forceProviders = mdModules_->initForceProviders();
1675 init_forcerec(fplog,
1677 runScheduleWork.simulationWork,
1683 opt2fn("-table", filenames.size(), filenames.data()),
1684 opt2fn("-tablep", filenames.size(), filenames.data()),
1685 opt2fns("-tableb", filenames.size(), filenames.data()),
1687 // Dirty hack, for fixing disres and orires should be made mdmodules
1688 fr->fcdata->disres = disresdata;
1689 if (gmx_mtop_ftype_count(mtop, F_ORIRES) > 0)
1691 fr->fcdata->orires = std::make_unique<t_oriresdata>(
1692 fplog, mtop, *inputrec, ms, globalState.get(), &atomSets);
1695 // Save a handle to device stream manager to use elsewhere in the code
1696 // TODO: Forcerec is not a correct place to store it.
1697 fr->deviceStreamManager = deviceStreamManager.get();
1699 if (runScheduleWork.simulationWork.useGpuPmePpCommunication && !thisRankHasDuty(cr, DUTY_PME))
1702 deviceStreamManager != nullptr,
1703 "GPU device stream manager should be valid in order to use PME-PP direct "
1706 deviceStreamManager->streamIsValid(DeviceStreamType::PmePpTransfer),
1707 "GPU PP-PME stream should be valid in order to use GPU PME-PP direct "
1709 fr->pmePpCommGpu = std::make_unique<gmx::PmePpCommGpu>(
1712 &cr->dd->pmeForceReceiveBuffer,
1713 deviceStreamManager->context(),
1714 deviceStreamManager->stream(DeviceStreamType::PmePpTransfer));
1717 fr->nbv = Nbnxm::init_nb_verlet(mdlog,
1722 runScheduleWork.simulationWork.useGpuNonbonded,
1723 deviceStreamManager.get(),
1727 // TODO: Move the logic below to a GPU bonded builder
1728 if (runScheduleWork.simulationWork.useGpuBonded)
1730 GMX_RELEASE_ASSERT(deviceStreamManager != nullptr,
1731 "GPU device stream manager should be valid in order to use GPU "
1732 "version of bonded forces.");
1733 fr->listedForcesGpu = std::make_unique<ListedForcesGpu>(
1735 fr->ic->epsfac * fr->fudgeQQ,
1736 deviceStreamManager->context(),
1737 deviceStreamManager->bondedStream(havePPDomainDecomposition(cr)),
1740 fr->longRangeNonbondeds = std::make_unique<CpuPpLongRangeNonbondeds>(fr->n_tpi,
1741 fr->ic->ewaldcoeff_q,
1751 /* Initialize the mdAtoms structure.
1752 * mdAtoms is not filled with atom data,
1753 * as this can not be done now with domain decomposition.
1755 mdAtoms = makeMDAtoms(fplog, mtop, *inputrec, thisRankHasPmeGpuTask);
1756 if (globalState && thisRankHasPmeGpuTask)
1758 // The pinning of coordinates in the global state object works, because we only use
1759 // PME on GPU without DD or on a separate PME rank, and because the local state pointer
1760 // points to the global state object without DD.
1761 // FIXME: MD and EM separately set up the local state - this should happen in the same
1762 // function, which should also perform the pinning.
1763 changePinningPolicy(&globalState->x, pme_get_pinning_policy());
1766 /* Initialize the virtual site communication */
1767 vsite = makeVirtualSitesHandler(
1768 mtop, cr, fr->pbcType, updateGroups.updateGroupingPerMoleculeType());
1770 calc_shifts(box, fr->shift_vec);
1772 /* With periodic molecules the charge groups should be whole at start up
1773 * and the virtual sites should not be far from their proper positions.
1775 if (!inputrec->bContinuation && MASTER(cr)
1776 && !(inputrec->pbcType != PbcType::No && inputrec->bPeriodicMols))
1778 /* Make molecules whole at start of run */
1779 if (fr->pbcType != PbcType::No)
1781 do_pbc_first_mtop(fplog, inputrec->pbcType, box, &mtop, globalState->x.rvec_array());
1785 /* Correct initial vsite positions are required
1786 * for the initial distribution in the domain decomposition
1787 * and for the initial shell prediction.
1789 constructVirtualSitesGlobal(mtop, globalState->x);
1792 // Make the DD reverse topology, now that any vsites that are present are available
1793 if (haveDDAtomOrdering(*cr))
1795 dd_make_reverse_top(fplog, cr->dd, mtop, vsite.get(), *inputrec, domdecOptions.ddBondedChecking);
1798 if (EEL_PME(fr->ic->eeltype) || EVDW_PME(fr->ic->vdwtype))
1800 ewaldcoeff_q = fr->ic->ewaldcoeff_q;
1801 ewaldcoeff_lj = fr->ic->ewaldcoeff_lj;
1806 /* This is a PME only node */
1808 GMX_ASSERT(globalState == nullptr,
1809 "We don't need the state on a PME only rank and expect it to be unitialized");
1811 ewaldcoeff_q = calc_ewaldcoeff_q(inputrec->rcoulomb, inputrec->ewald_rtol);
1812 ewaldcoeff_lj = calc_ewaldcoeff_lj(inputrec->rvdw, inputrec->ewald_rtol_lj);
1815 gmx_pme_t* sepPmeData = nullptr;
1816 // This reference hides the fact that PME data is owned by runner on PME-only ranks and by forcerec on other ranks
1817 GMX_ASSERT(thisRankHasDuty(cr, DUTY_PP) == (fr != nullptr),
1818 "Double-checking that only PME-only ranks have no forcerec");
1819 gmx_pme_t*& pmedata = fr ? fr->pmedata : sepPmeData;
1821 // TODO should live in ewald module once its testing is improved
1823 // Later, this program could contain kernels that might be later
1824 // re-used as auto-tuning progresses, or subsequent simulations
1826 PmeGpuProgramStorage pmeGpuProgram;
1827 if (thisRankHasPmeGpuTask)
1830 (deviceStreamManager != nullptr),
1831 "GPU device stream manager should be initialized in order to use GPU for PME.");
1832 GMX_RELEASE_ASSERT((deviceInfo != nullptr),
1833 "GPU device should be initialized in order to use GPU for PME.");
1834 pmeGpuProgram = buildPmeGpuProgram(deviceStreamManager->context());
1837 /* Initiate PME if necessary,
1838 * either on all nodes or on dedicated PME nodes only. */
1839 if (EEL_PME(inputrec->coulombtype) || EVDW_PME(inputrec->vdwtype))
1841 if (mdAtoms && mdAtoms->mdatoms())
1843 nChargePerturbed = mdAtoms->mdatoms()->nChargePerturbed;
1844 if (EVDW_PME(inputrec->vdwtype))
1846 nTypePerturbed = mdAtoms->mdatoms()->nTypePerturbed;
1849 if (cr->npmenodes > 0)
1851 /* The PME only nodes need to know nChargePerturbed(FEP on Q) and nTypePerturbed(FEP on LJ)*/
1852 gmx_bcast(sizeof(nChargePerturbed), &nChargePerturbed, cr->mpi_comm_mysim);
1853 gmx_bcast(sizeof(nTypePerturbed), &nTypePerturbed, cr->mpi_comm_mysim);
1856 if (thisRankHasDuty(cr, DUTY_PME))
1860 // TODO: This should be in the builder.
1861 GMX_RELEASE_ASSERT(!runScheduleWork.simulationWork.useGpuPme
1862 || (deviceStreamManager != nullptr),
1863 "Device stream manager should be valid in order to use GPU "
1866 !runScheduleWork.simulationWork.useGpuPme
1867 || deviceStreamManager->streamIsValid(DeviceStreamType::Pme),
1868 "GPU PME stream should be valid in order to use GPU version of PME.");
1870 const DeviceContext* deviceContext = runScheduleWork.simulationWork.useGpuPme
1871 ? &deviceStreamManager->context()
1873 const DeviceStream* pmeStream =
1874 runScheduleWork.simulationWork.useGpuPme
1875 ? &deviceStreamManager->stream(DeviceStreamType::Pme)
1878 pmedata = gmx_pme_init(cr,
1879 getNumPmeDomains(cr->dd),
1881 nChargePerturbed != 0,
1882 nTypePerturbed != 0,
1883 mdrunOptions.reproducible,
1886 gmx_omp_nthreads_get(ModuleMultiThread::Pme),
1891 pmeGpuProgram.get(),
1894 GMX_CATCH_ALL_AND_EXIT_WITH_FATAL_ERROR
1899 if (EI_DYNAMICS(inputrec->eI))
1901 /* Turn on signal handling on all nodes */
1903 * (A user signal from the PME nodes (if any)
1904 * is communicated to the PP nodes.
1906 signal_handler_install();
1909 pull_t* pull_work = nullptr;
1910 if (thisRankHasDuty(cr, DUTY_PP))
1912 /* Assumes uniform use of the number of OpenMP threads */
1913 walltime_accounting = walltime_accounting_init(gmx_omp_nthreads_get(ModuleMultiThread::Default));
1915 if (inputrec->bPull)
1917 /* Initialize pull code */
1918 pull_work = init_pull(fplog,
1919 inputrec->pull.get(),
1924 inputrec->fepvals->init_lambda);
1925 if (inputrec->pull->bXOutAverage || inputrec->pull->bFOutAverage)
1927 initPullHistory(pull_work, &observablesHistory);
1929 if (EI_DYNAMICS(inputrec->eI) && MASTER(cr))
1931 init_pull_output_files(pull_work, filenames.size(), filenames.data(), oenv, startingBehavior);
1935 std::unique_ptr<EnforcedRotation> enforcedRotation;
1938 /* Initialize enforced rotation code */
1939 enforcedRotation = init_rot(fplog,
1952 t_swap* swap = nullptr;
1953 if (inputrec->eSwapCoords != SwapType::No)
1955 /* Initialize ion swapping code */
1956 swap = init_swapcoords(fplog,
1958 opt2fn_master("-swap", filenames.size(), filenames.data(), cr),
1961 &observablesHistory,
1969 /* Let makeConstraints know whether we have essential dynamics constraints. */
1970 auto constr = makeConstraints(mtop,
1973 doEssentialDynamics,
1976 updateGroups.useUpdateGroups(),
1981 &observablesReducerBuilder);
1983 /* Energy terms and groups */
1984 gmx_enerdata_t enerd(mtop.groups.groups[SimulationAtomGroupType::EnergyOutput].size(),
1985 inputrec->fepvals->n_lambda);
1987 // cos acceleration is only supported by md, but older tpr
1988 // files might still combine it with other integrators
1989 GMX_RELEASE_ASSERT(inputrec->cos_accel == 0.0 || inputrec->eI == IntegrationAlgorithm::MD,
1990 "cos_acceleration is only supported by integrator=md");
1992 /* Kinetic energy data */
1993 gmx_ekindata_t ekind(inputrec->opts.ngtc,
1994 inputrec->cos_accel,
1995 gmx_omp_nthreads_get(ModuleMultiThread::Update));
1997 /* Set up interactive MD (IMD) */
1998 auto imdSession = makeImdSession(inputrec.get(),
2005 MASTER(cr) ? globalState->x : gmx::ArrayRef<gmx::RVec>(),
2009 mdrunOptions.imdOptions,
2012 if (haveDDAtomOrdering(*cr))
2014 GMX_RELEASE_ASSERT(fr, "fr was NULL while cr->duty was DUTY_PP");
2015 /* This call is not included in init_domain_decomposition
2016 * because fr->atomInfoForEachMoleculeBlock is set later.
2018 makeBondedLinks(cr->dd, mtop, fr->atomInfoForEachMoleculeBlock);
2021 if (runScheduleWork.simulationWork.useGpuBufferOps)
2023 fr->gpuForceReduction[gmx::AtomLocality::Local] = std::make_unique<gmx::GpuForceReduction>(
2024 deviceStreamManager->context(),
2025 deviceStreamManager->stream(gmx::DeviceStreamType::NonBondedLocal),
2027 fr->gpuForceReduction[gmx::AtomLocality::NonLocal] = std::make_unique<gmx::GpuForceReduction>(
2028 deviceStreamManager->context(),
2029 deviceStreamManager->stream(gmx::DeviceStreamType::NonBondedNonLocal),
2033 std::unique_ptr<gmx::StatePropagatorDataGpu> stateGpu;
2034 if (gpusWereDetected
2035 && ((runScheduleWork.simulationWork.useGpuPme && thisRankHasDuty(cr, DUTY_PME))
2036 || runScheduleWork.simulationWork.useGpuBufferOps))
2038 GpuApiCallBehavior transferKind =
2039 (inputrec->eI == IntegrationAlgorithm::MD && !doRerun && !useModularSimulator)
2040 ? GpuApiCallBehavior::Async
2041 : GpuApiCallBehavior::Sync;
2042 GMX_RELEASE_ASSERT(deviceStreamManager != nullptr,
2043 "GPU device stream manager should be initialized to use GPU.");
2044 stateGpu = std::make_unique<gmx::StatePropagatorDataGpu>(
2045 *deviceStreamManager, transferKind, pme_gpu_get_block_size(fr->pmedata), wcycle.get());
2046 fr->stateGpu = stateGpu.get();
2049 GMX_ASSERT(stopHandlerBuilder_, "Runner must provide StopHandlerBuilder to simulator.");
2050 SimulatorBuilder simulatorBuilder;
2052 simulatorBuilder.add(SimulatorStateData(
2053 globalState.get(), localState, &observablesHistory, &enerd, &ekind));
2054 simulatorBuilder.add(std::move(membedHolder));
2055 simulatorBuilder.add(std::move(stopHandlerBuilder_));
2056 simulatorBuilder.add(SimulatorConfig(mdrunOptions, startingBehavior, &runScheduleWork));
2059 simulatorBuilder.add(SimulatorEnv(fplog, cr, ms, mdlog, oenv, &observablesReducerBuilder));
2060 simulatorBuilder.add(Profiling(&nrnb, walltime_accounting, wcycle.get()));
2061 simulatorBuilder.add(ConstraintsParam(
2062 constr.get(), enforcedRotation ? enforcedRotation->getLegacyEnfrot() : nullptr, vsite.get()));
2063 // TODO: Separate `fr` to a separate add, and make the `build` handle the coupling sensibly.
2064 simulatorBuilder.add(LegacyInput(
2065 static_cast<int>(filenames.size()), filenames.data(), inputrec.get(), fr.get()));
2066 simulatorBuilder.add(ReplicaExchangeParameters(replExParams));
2067 simulatorBuilder.add(InteractiveMD(imdSession.get()));
2068 simulatorBuilder.add(SimulatorModules(mdModules_->outputProvider(), mdModules_->notifiers()));
2069 simulatorBuilder.add(CenterOfMassPulling(pull_work));
2070 // Todo move to an MDModule
2071 simulatorBuilder.add(IonSwapping(swap));
2072 simulatorBuilder.add(TopologyData(mtop, &localTopology, mdAtoms.get()));
2073 simulatorBuilder.add(BoxDeformationHandle(deform.get()));
2074 simulatorBuilder.add(std::move(modularSimulatorCheckpointData));
2076 // build and run simulator object based on user-input
2077 auto simulator = simulatorBuilder.build(useModularSimulator);
2080 if (fr->pmePpCommGpu)
2082 // destroy object since it is no longer required. (This needs to be done while the GPU context still exists.)
2083 fr->pmePpCommGpu.reset();
2086 if (inputrec->bPull)
2088 finish_pull(pull_work);
2090 finish_swapcoords(swap);
2094 GMX_RELEASE_ASSERT(pmedata, "pmedata was NULL while cr->duty was not DUTY_PP");
2096 walltime_accounting = walltime_accounting_init(gmx_omp_nthreads_get(ModuleMultiThread::Pme));
2097 gmx_pmeonly(pmedata,
2101 walltime_accounting,
2104 runScheduleWork.simulationWork.useGpuPmePpCommunication,
2105 deviceStreamManager.get());
2108 wallcycle_stop(wcycle.get(), WallCycleCounter::Run);
2110 /* Finish up, write some stuff
2111 * if rerunMD, don't write last frame again
2119 walltime_accounting,
2120 fr ? fr->nbv.get() : nullptr,
2122 EI_DYNAMICS(inputrec->eI) && !isMultiSim(ms));
2125 deviceStreamManager.reset(nullptr);
2129 gmx_pme_destroy(pmedata);
2133 // FIXME: this is only here to manually unpin mdAtoms->chargeA_ and state->x,
2134 // before we destroy the GPU context(s)
2135 // Pinned buffers are associated with contexts in CUDA.
2136 // As soon as we destroy GPU contexts after mdrunner() exits, these lines should go.
2137 mdAtoms.reset(nullptr);
2138 globalState.reset(nullptr);
2139 localStateInstance.reset(nullptr);
2140 mdModules_.reset(nullptr); // destruct force providers here as they might also use the GPU
2141 fr.reset(nullptr); // destruct forcerec before gpu
2142 // TODO convert to C++ so we can get rid of these frees
2145 if (!hwinfo_->deviceInfoList.empty())
2147 /* stop the GPU profiler (only CUDA) */
2151 /* With tMPI we need to wait for all ranks to finish deallocation before
2152 * destroying the CUDA context as some tMPI ranks may be sharing
2155 * This is not a concern in OpenCL where we use one context per rank.
2157 * Note: it is safe to not call the barrier on the ranks which do not use GPU,
2158 * but it is easier and more futureproof to call it on the whole node.
2160 * Note that this function needs to be called even if GPUs are not used
2161 * in this run because the PME ranks have no knowledge of whether GPUs
2162 * are used or not, but all ranks need to enter the barrier below.
2163 * \todo Remove this physical node barrier after making sure
2164 * that it's not needed anymore (with a shared GPU run).
2168 physicalNodeComm.barrier();
2171 if (!devFlags.usingCudaAwareMpi)
2173 // Don't reset GPU in case of CUDA-AWARE MPI
2174 // UCX creates CUDA buffers which are cleaned-up as part of MPI_Finalize()
2175 // resetting the device before MPI_Finalize() results in crashes inside UCX
2176 releaseDevice(deviceInfo);
2179 /* Does what it says */
2180 print_date_and_time(fplog, cr->nodeid, "Finished mdrun", gmx_gettime());
2181 walltime_accounting_destroy(walltime_accounting);
2183 // Ensure log file content is written
2186 gmx_fio_flush(logFileHandle);
2189 /* Reset FPEs (important for unit tests) by disabling them. Assumes no
2190 * exceptions were enabled before function was called. */
2193 gmx_fedisableexcept();
2196 auto rc = static_cast<int>(gmx_get_stop_condition());
2199 /* we need to join all threads. The sub-threads join when they
2200 exit this function, but the master thread needs to be told to
2210 Mdrunner::~Mdrunner()
2212 // Clean up of the Manager.
2213 // This will end up getting called on every thread-MPI rank, which is unnecessary,
2214 // but okay as long as threads synchronize some time before adding or accessing
2215 // a new set of restraints.
2216 if (restraintManager_)
2218 restraintManager_->clear();
2219 GMX_ASSERT(restraintManager_->countRestraints() == 0,
2220 "restraints added during runner life time should be cleared at runner "
2225 void Mdrunner::addPotential(std::shared_ptr<gmx::IRestraintPotential> puller, const std::string& name)
2227 GMX_ASSERT(restraintManager_, "Mdrunner must have a restraint manager.");
2228 // Not sure if this should be logged through the md logger or something else,
2229 // but it is helpful to have some sort of INFO level message sent somewhere.
2230 // std::cout << "Registering restraint named " << name << std::endl;
2232 // When multiple restraints are used, it may be wasteful to register them separately.
2233 // Maybe instead register an entire Restraint Manager as a force provider.
2234 restraintManager_->addToSpec(std::move(puller), name);
2237 Mdrunner::Mdrunner(std::unique_ptr<MDModules> mdModules) : mdModules_(std::move(mdModules)) {}
2239 Mdrunner::Mdrunner(Mdrunner&&) noexcept = default;
2241 //NOLINTNEXTLINE(performance-noexcept-move-constructor) working around GCC bug 58265 in CentOS 7
2242 Mdrunner& Mdrunner::operator=(Mdrunner&& /*handle*/) noexcept(BUGFREE_NOEXCEPT_STRING) = default;
2244 class Mdrunner::BuilderImplementation
2247 BuilderImplementation() = delete;
2248 BuilderImplementation(std::unique_ptr<MDModules> mdModules, compat::not_null<SimulationContext*> context);
2249 ~BuilderImplementation();
2251 BuilderImplementation& setExtraMdrunOptions(const MdrunOptions& options,
2252 real forceWarningThreshold,
2253 StartingBehavior startingBehavior);
2255 void addHardwareDetectionResult(const gmx_hw_info_t* hwinfo);
2257 void addDomdec(const DomdecOptions& options);
2259 void addInput(SimulationInputHandle inputHolder);
2261 void addVerletList(int nstlist);
2263 void addReplicaExchange(const ReplicaExchangeParameters& params);
2265 void addNonBonded(const char* nbpu_opt);
2267 void addPME(const char* pme_opt_, const char* pme_fft_opt_);
2269 void addBondedTaskAssignment(const char* bonded_opt);
2271 void addUpdateTaskAssignment(const char* update_opt);
2273 void addHardwareOptions(const gmx_hw_opt_t& hardwareOptions);
2275 void addFilenames(ArrayRef<const t_filenm> filenames);
2277 void addOutputEnvironment(gmx_output_env_t* outputEnvironment);
2279 void addLogFile(t_fileio* logFileHandle);
2281 void addStopHandlerBuilder(std::unique_ptr<StopHandlerBuilder> builder);
2286 // Default parameters copied from runner.h
2287 // \todo Clarify source(s) of default parameters.
2289 const char* nbpu_opt_ = nullptr;
2290 const char* pme_opt_ = nullptr;
2291 const char* pme_fft_opt_ = nullptr;
2292 const char* bonded_opt_ = nullptr;
2293 const char* update_opt_ = nullptr;
2295 MdrunOptions mdrunOptions_;
2297 DomdecOptions domdecOptions_;
2299 ReplicaExchangeParameters replicaExchangeParameters_;
2301 //! Command-line override for the duration of a neighbor list with the Verlet scheme.
2304 //! World communicator, used for hardware detection and task assignment
2305 MPI_Comm libraryWorldCommunicator_ = MPI_COMM_NULL;
2307 //! Multisim communicator handle.
2308 gmx_multisim_t* multiSimulation_;
2310 //! mdrun communicator
2311 MPI_Comm simulationCommunicator_ = MPI_COMM_NULL;
2313 //! Print a warning if any force is larger than this (in kJ/mol nm).
2314 real forceWarningThreshold_ = -1;
2316 //! Whether the simulation will start afresh, or restart with/without appending.
2317 StartingBehavior startingBehavior_ = StartingBehavior::NewSimulation;
2319 //! The modules that comprise the functionality of mdrun.
2320 std::unique_ptr<MDModules> mdModules_;
2322 //! Detected hardware.
2323 const gmx_hw_info_t* hwinfo_ = nullptr;
2325 //! \brief Parallelism information.
2326 gmx_hw_opt_t hardwareOptions_;
2328 //! filename options for simulation.
2329 ArrayRef<const t_filenm> filenames_;
2331 /*! \brief Handle to output environment.
2333 * \todo gmx_output_env_t needs lifetime management.
2335 gmx_output_env_t* outputEnvironment_ = nullptr;
2337 /*! \brief Non-owning handle to MD log file.
2339 * \todo Context should own output facilities for client.
2340 * \todo Improve log file handle management.
2342 * Code managing the FILE* relies on the ability to set it to
2343 * nullptr to check whether the filehandle is valid.
2345 t_fileio* logFileHandle_ = nullptr;
2348 * \brief Builder for simulation stop signal handler.
2350 std::unique_ptr<StopHandlerBuilder> stopHandlerBuilder_ = nullptr;
2353 * \brief Sources for initial simulation state.
2355 * See issue #3652 for near-term refinements to the SimulationInput interface.
2357 * See issue #3379 for broader discussion on API aspects of simulation inputs and outputs.
2359 SimulationInputHandle inputHolder_;
2362 Mdrunner::BuilderImplementation::BuilderImplementation(std::unique_ptr<MDModules> mdModules,
2363 compat::not_null<SimulationContext*> context) :
2364 mdModules_(std::move(mdModules))
2366 libraryWorldCommunicator_ = context->libraryWorldCommunicator_;
2367 simulationCommunicator_ = context->simulationCommunicator_;
2368 multiSimulation_ = context->multiSimulation_.get();
2371 Mdrunner::BuilderImplementation::~BuilderImplementation() = default;
2373 Mdrunner::BuilderImplementation&
2374 Mdrunner::BuilderImplementation::setExtraMdrunOptions(const MdrunOptions& options,
2375 const real forceWarningThreshold,
2376 const StartingBehavior startingBehavior)
2378 mdrunOptions_ = options;
2379 forceWarningThreshold_ = forceWarningThreshold;
2380 startingBehavior_ = startingBehavior;
2384 void Mdrunner::BuilderImplementation::addDomdec(const DomdecOptions& options)
2386 domdecOptions_ = options;
2389 void Mdrunner::BuilderImplementation::addVerletList(int nstlist)
2394 void Mdrunner::BuilderImplementation::addReplicaExchange(const ReplicaExchangeParameters& params)
2396 replicaExchangeParameters_ = params;
2399 Mdrunner Mdrunner::BuilderImplementation::build()
2401 auto newRunner = Mdrunner(std::move(mdModules_));
2403 newRunner.mdrunOptions = mdrunOptions_;
2404 newRunner.pforce = forceWarningThreshold_;
2405 newRunner.startingBehavior = startingBehavior_;
2406 newRunner.domdecOptions = domdecOptions_;
2408 // \todo determine an invariant to check or confirm that all gmx_hw_opt_t objects are valid
2409 newRunner.hw_opt = hardwareOptions_;
2411 // No invariant to check. This parameter exists to optionally override other behavior.
2412 newRunner.nstlist_cmdline = nstlist_;
2414 newRunner.replExParams = replicaExchangeParameters_;
2416 newRunner.filenames = filenames_;
2418 newRunner.libraryWorldCommunicator = libraryWorldCommunicator_;
2420 newRunner.simulationCommunicator = simulationCommunicator_;
2422 // nullptr is a valid value for the multisim handle
2423 newRunner.ms = multiSimulation_;
2427 newRunner.hwinfo_ = hwinfo_;
2431 GMX_THROW(gmx::APIError(
2432 "MdrunnerBuilder::addHardwareDetectionResult() is required before build()"));
2437 newRunner.inputHolder_ = std::move(inputHolder_);
2441 GMX_THROW(gmx::APIError("MdrunnerBuilder::addInput() is required before build()."));
2444 // \todo Clarify ownership and lifetime management for gmx_output_env_t
2445 // \todo Update sanity checking when output environment has clearly specified invariants.
2446 // Initialization and default values for oenv are not well specified in the current version.
2447 if (outputEnvironment_)
2449 newRunner.oenv = outputEnvironment_;
2453 GMX_THROW(gmx::APIError(
2454 "MdrunnerBuilder::addOutputEnvironment() is required before build()"));
2457 newRunner.logFileHandle = logFileHandle_;
2461 newRunner.nbpu_opt = nbpu_opt_;
2465 GMX_THROW(gmx::APIError("MdrunnerBuilder::addNonBonded() is required before build()"));
2468 if (pme_opt_ && pme_fft_opt_)
2470 newRunner.pme_opt = pme_opt_;
2471 newRunner.pme_fft_opt = pme_fft_opt_;
2475 GMX_THROW(gmx::APIError("MdrunnerBuilder::addElectrostatics() is required before build()"));
2480 newRunner.bonded_opt = bonded_opt_;
2484 GMX_THROW(gmx::APIError(
2485 "MdrunnerBuilder::addBondedTaskAssignment() is required before build()"));
2490 newRunner.update_opt = update_opt_;
2494 GMX_THROW(gmx::APIError(
2495 "MdrunnerBuilder::addUpdateTaskAssignment() is required before build() "));
2499 newRunner.restraintManager_ = std::make_unique<gmx::RestraintManager>();
2501 if (stopHandlerBuilder_)
2503 newRunner.stopHandlerBuilder_ = std::move(stopHandlerBuilder_);
2507 newRunner.stopHandlerBuilder_ = std::make_unique<StopHandlerBuilder>();
2513 void Mdrunner::BuilderImplementation::addHardwareDetectionResult(const gmx_hw_info_t* hwinfo)
2518 void Mdrunner::BuilderImplementation::addNonBonded(const char* nbpu_opt)
2520 nbpu_opt_ = nbpu_opt;
2523 void Mdrunner::BuilderImplementation::addPME(const char* pme_opt, const char* pme_fft_opt)
2526 pme_fft_opt_ = pme_fft_opt;
2529 void Mdrunner::BuilderImplementation::addBondedTaskAssignment(const char* bonded_opt)
2531 bonded_opt_ = bonded_opt;
2534 void Mdrunner::BuilderImplementation::addUpdateTaskAssignment(const char* update_opt)
2536 update_opt_ = update_opt;
2539 void Mdrunner::BuilderImplementation::addHardwareOptions(const gmx_hw_opt_t& hardwareOptions)
2541 hardwareOptions_ = hardwareOptions;
2544 void Mdrunner::BuilderImplementation::addFilenames(ArrayRef<const t_filenm> filenames)
2546 filenames_ = filenames;
2549 void Mdrunner::BuilderImplementation::addOutputEnvironment(gmx_output_env_t* outputEnvironment)
2551 outputEnvironment_ = outputEnvironment;
2554 void Mdrunner::BuilderImplementation::addLogFile(t_fileio* logFileHandle)
2556 logFileHandle_ = logFileHandle;
2559 void Mdrunner::BuilderImplementation::addStopHandlerBuilder(std::unique_ptr<StopHandlerBuilder> builder)
2561 stopHandlerBuilder_ = std::move(builder);
2564 void Mdrunner::BuilderImplementation::addInput(SimulationInputHandle inputHolder)
2566 inputHolder_ = std::move(inputHolder);
2569 MdrunnerBuilder::MdrunnerBuilder(std::unique_ptr<MDModules> mdModules,
2570 compat::not_null<SimulationContext*> context) :
2571 impl_{ std::make_unique<Mdrunner::BuilderImplementation>(std::move(mdModules), context) }
2575 MdrunnerBuilder::~MdrunnerBuilder() = default;
2577 MdrunnerBuilder& MdrunnerBuilder::addHardwareDetectionResult(const gmx_hw_info_t* hwinfo)
2579 impl_->addHardwareDetectionResult(hwinfo);
2583 MdrunnerBuilder& MdrunnerBuilder::addSimulationMethod(const MdrunOptions& options,
2584 real forceWarningThreshold,
2585 const StartingBehavior startingBehavior)
2587 impl_->setExtraMdrunOptions(options, forceWarningThreshold, startingBehavior);
2591 MdrunnerBuilder& MdrunnerBuilder::addDomainDecomposition(const DomdecOptions& options)
2593 impl_->addDomdec(options);
2597 MdrunnerBuilder& MdrunnerBuilder::addNeighborList(int nstlist)
2599 impl_->addVerletList(nstlist);
2603 MdrunnerBuilder& MdrunnerBuilder::addReplicaExchange(const ReplicaExchangeParameters& params)
2605 impl_->addReplicaExchange(params);
2609 MdrunnerBuilder& MdrunnerBuilder::addNonBonded(const char* nbpu_opt)
2611 impl_->addNonBonded(nbpu_opt);
2615 MdrunnerBuilder& MdrunnerBuilder::addElectrostatics(const char* pme_opt, const char* pme_fft_opt)
2617 // The builder method may become more general in the future, but in this version,
2618 // parameters for PME electrostatics are both required and the only parameters
2620 if (pme_opt && pme_fft_opt)
2622 impl_->addPME(pme_opt, pme_fft_opt);
2627 gmx::InvalidInputError("addElectrostatics() arguments must be non-null pointers."));
2632 MdrunnerBuilder& MdrunnerBuilder::addBondedTaskAssignment(const char* bonded_opt)
2634 impl_->addBondedTaskAssignment(bonded_opt);
2638 MdrunnerBuilder& MdrunnerBuilder::addUpdateTaskAssignment(const char* update_opt)
2640 impl_->addUpdateTaskAssignment(update_opt);
2644 Mdrunner MdrunnerBuilder::build()
2646 return impl_->build();
2649 MdrunnerBuilder& MdrunnerBuilder::addHardwareOptions(const gmx_hw_opt_t& hardwareOptions)
2651 impl_->addHardwareOptions(hardwareOptions);
2655 MdrunnerBuilder& MdrunnerBuilder::addFilenames(ArrayRef<const t_filenm> filenames)
2657 impl_->addFilenames(filenames);
2661 MdrunnerBuilder& MdrunnerBuilder::addOutputEnvironment(gmx_output_env_t* outputEnvironment)
2663 impl_->addOutputEnvironment(outputEnvironment);
2667 MdrunnerBuilder& MdrunnerBuilder::addLogFile(t_fileio* logFileHandle)
2669 impl_->addLogFile(logFileHandle);
2673 MdrunnerBuilder& MdrunnerBuilder::addStopHandlerBuilder(std::unique_ptr<StopHandlerBuilder> builder)
2675 impl_->addStopHandlerBuilder(std::move(builder));
2679 MdrunnerBuilder& MdrunnerBuilder::addInput(SimulationInputHandle input)
2681 impl_->addInput(std::move(input));
2685 MdrunnerBuilder::MdrunnerBuilder(MdrunnerBuilder&&) noexcept = default;
2687 MdrunnerBuilder& MdrunnerBuilder::operator=(MdrunnerBuilder&&) noexcept = default;