src/gromacs/mdtypes/state_propagator_data_gpu_impl_gpu.cpp

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2019,2020, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*! \internal \file
  36  *
  37  * \brief Definitions of interfaces for GPU state data propagator object.
  38  *
  39  * \author Artem Zhmurov <zhmurov@gmail.com>
  40  *
  41  * \ingroup module_mdtypes
  42  */
  43 #include "gmxpre.h"
  44
  45 #include "config.h"
  46
  47 #if GMX_GPU != GMX_GPU_NONE
  48
  49 #    if GMX_GPU == GMX_GPU_CUDA
  50 #        include "gromacs/gpu_utils/cudautils.cuh"
  51 #    endif
  52 #    include "gromacs/gpu_utils/devicebuffer.h"
  53 #    if GMX_GPU == GMX_GPU_OPENCL
  54 #        include "gromacs/gpu_utils/oclutils.h"
  55 #    endif
  56 #    include "gromacs/math/vectypes.h"
  57 #    include "gromacs/mdtypes/state_propagator_data_gpu.h"
  58 #    include "gromacs/timing/wallcycle.h"
  59 #    include "gromacs/utility/classhelpers.h"
  60
  61 #    include "state_propagator_data_gpu_impl.h"
  62
  63
  64 namespace gmx
  65 {
  66
  67 StatePropagatorDataGpu::Impl::Impl(const void*        pmeStream,
  68                                    const void*        localStream,
  69                                    const void*        nonLocalStream,
  70                                    const void*        deviceContext,
  71                                    GpuApiCallBehavior transferKind,
  72                                    int                paddingSize,
  73                                    gmx_wallcycle*     wcycle) :
  74     transferKind_(transferKind),
  75     paddingSize_(paddingSize),
  76     wcycle_(wcycle)
  77 {
  78     static_assert(GMX_GPU != GMX_GPU_NONE,
  79                   "This object should only be constructed on the GPU code-paths.");
  80
  81     // TODO: Refactor when the StreamManager is introduced.
  82     if (GMX_GPU == GMX_GPU_OPENCL)
  83     {
  84         GMX_ASSERT(deviceContext != nullptr, "GPU context should be set in OpenCL builds.");
  85         GMX_ASSERT(pmeStream != nullptr, "GPU PME stream should be set in OpenCL builds.");
  86
  87         // The update stream is set to the PME stream in OpenCL, since PME stream is the only stream created in the PME context.
  88         pmeStream_     = *static_cast<const CommandStream*>(pmeStream);
  89         updateStream_  = *static_cast<const CommandStream*>(pmeStream);
  90         deviceContext_ = *static_cast<const DeviceContext*>(deviceContext);
  91         GMX_UNUSED_VALUE(localStream);
  92         GMX_UNUSED_VALUE(nonLocalStream);
  93     }
  94
  95     if (GMX_GPU == GMX_GPU_CUDA)
  96     {
  97         if (pmeStream != nullptr)
  98         {
  99             pmeStream_ = *static_cast<const CommandStream*>(pmeStream);
 100         }
 101         if (localStream != nullptr)
 102         {
 103             localStream_ = *static_cast<const CommandStream*>(localStream);
 104         }
 105         if (nonLocalStream != nullptr)
 106         {
 107             nonLocalStream_ = *static_cast<const CommandStream*>(nonLocalStream);
 108         }
 109
 110         // TODO: The update stream should be created only when it is needed.
 111 #    if (GMX_GPU == GMX_GPU_CUDA)
 112         cudaError_t stat;
 113         stat = cudaStreamCreate(&updateStream_);
 114         CU_RET_ERR(stat, "CUDA stream creation failed in StatePropagatorDataGpu");
 115 #    endif
 116         GMX_UNUSED_VALUE(deviceContext);
 117     }
 118
 119     // Map the atom locality to the stream that will be used for coordinates,
 120     // velocities and forces transfers. Same streams are used for H2D and D2H copies.
 121     // Note, that nullptr stream is used here to indicate that the copy is not supported.
 122     xCopyStreams_[AtomLocality::Local]    = updateStream_;
 123     xCopyStreams_[AtomLocality::NonLocal] = nonLocalStream_;
 124     xCopyStreams_[AtomLocality::All]      = updateStream_;
 125
 126     vCopyStreams_[AtomLocality::Local]    = updateStream_;
 127     vCopyStreams_[AtomLocality::NonLocal] = nullptr;
 128     vCopyStreams_[AtomLocality::All]      = updateStream_;
 129
 130     fCopyStreams_[AtomLocality::Local]    = localStream_;
 131     fCopyStreams_[AtomLocality::NonLocal] = nonLocalStream_;
 132     fCopyStreams_[AtomLocality::All]      = updateStream_;
 133 }
 134
 135 StatePropagatorDataGpu::Impl::Impl(const void*        pmeStream,
 136                                    const void*        deviceContext,
 137                                    GpuApiCallBehavior transferKind,
 138                                    int                paddingSize,
 139                                    gmx_wallcycle*     wcycle) :
 140     transferKind_(transferKind),
 141     paddingSize_(paddingSize),
 142     wcycle_(wcycle)
 143 {
 144     static_assert(GMX_GPU != GMX_GPU_NONE,
 145                   "This object should only be constructed on the GPU code-paths.");
 146
 147     if (GMX_GPU == GMX_GPU_OPENCL)
 148     {
 149         GMX_ASSERT(deviceContext != nullptr, "GPU context should be set in OpenCL builds.");
 150         deviceContext_ = *static_cast<const DeviceContext*>(deviceContext);
 151     }
 152
 153     GMX_ASSERT(pmeStream != nullptr, "GPU PME stream should be set.");
 154     pmeStream_ = *static_cast<const CommandStream*>(pmeStream);
 155
 156     localStream_    = nullptr;
 157     nonLocalStream_ = nullptr;
 158     updateStream_   = nullptr;
 159
 160
 161     // Only local/all coordinates are allowed to be copied in PME-only rank/ PME tests.
 162     // This it temporary measure to make it safe to use this class in those cases.
 163     xCopyStreams_[AtomLocality::Local]    = pmeStream_;
 164     xCopyStreams_[AtomLocality::NonLocal] = nullptr;
 165     xCopyStreams_[AtomLocality::All]      = pmeStream_;
 166
 167     vCopyStreams_[AtomLocality::Local]    = nullptr;
 168     vCopyStreams_[AtomLocality::NonLocal] = nullptr;
 169     vCopyStreams_[AtomLocality::All]      = nullptr;
 170
 171     fCopyStreams_[AtomLocality::Local]    = nullptr;
 172     fCopyStreams_[AtomLocality::NonLocal] = nullptr;
 173     fCopyStreams_[AtomLocality::All]      = nullptr;
 174 }
 175
 176 StatePropagatorDataGpu::Impl::~Impl() {}
 177
 178 void StatePropagatorDataGpu::Impl::reinit(int numAtomsLocal, int numAtomsAll)
 179 {
 180     wallcycle_start_nocount(wcycle_, ewcLAUNCH_GPU);
 181     wallcycle_sub_start_nocount(wcycle_, ewcsLAUNCH_STATE_PROPAGATOR_DATA);
 182
 183     numAtomsLocal_ = numAtomsLocal;
 184     numAtomsAll_   = numAtomsAll;
 185
 186     int numAtomsPadded;
 187     if (paddingSize_ > 0)
 188     {
 189         numAtomsPadded = ((numAtomsAll_ + paddingSize_ - 1) / paddingSize_) * paddingSize_;
 190     }
 191     else
 192     {
 193         numAtomsPadded = numAtomsAll_;
 194     }
 195
 196     reallocateDeviceBuffer(&d_x_, numAtomsPadded, &d_xSize_, &d_xCapacity_, deviceContext_);
 197
 198     const size_t paddingAllocationSize = numAtomsPadded - numAtomsAll_;
 199     if (paddingAllocationSize > 0)
 200     {
 201         // The PME stream is used here because the padding region of d_x_ is only in the PME task.
 202         clearDeviceBufferAsync(&d_x_, numAtomsAll_, paddingAllocationSize, pmeStream_);
 203     }
 204
 205     reallocateDeviceBuffer(&d_v_, numAtomsAll_, &d_vSize_, &d_vCapacity_, deviceContext_);
 206     const int d_fOldCapacity = d_fCapacity_;
 207     reallocateDeviceBuffer(&d_f_, numAtomsAll_, &d_fSize_, &d_fCapacity_, deviceContext_);
 208     // Clearing of the forces can be done in local stream since the nonlocal stream cannot reach
 209     // the force accumulation stage before syncing with the local stream. Only done in CUDA,
 210     // since the force buffer ops are not implemented in OpenCL.
 211     if (GMX_GPU == GMX_GPU_CUDA && d_fCapacity_ != d_fOldCapacity)
 212     {
 213         clearDeviceBufferAsync(&d_f_, 0, d_fCapacity_, localStream_);
 214     }
 215
 216     wallcycle_sub_stop(wcycle_, ewcsLAUNCH_STATE_PROPAGATOR_DATA);
 217     wallcycle_stop(wcycle_, ewcLAUNCH_GPU);
 218 }
 219
 220 std::tuple<int, int> StatePropagatorDataGpu::Impl::getAtomRangesFromAtomLocality(AtomLocality atomLocality)
 221 {
 222     int atomsStartAt   = 0;
 223     int numAtomsToCopy = 0;
 224     switch (atomLocality)
 225     {
 226         case AtomLocality::All:
 227             atomsStartAt   = 0;
 228             numAtomsToCopy = numAtomsAll_;
 229             break;
 230         case AtomLocality::Local:
 231             atomsStartAt   = 0;
 232             numAtomsToCopy = numAtomsLocal_;
 233             break;
 234         case AtomLocality::NonLocal:
 235             atomsStartAt   = numAtomsLocal_;
 236             numAtomsToCopy = numAtomsAll_ - numAtomsLocal_;
 237             break;
 238         default:
 239             GMX_RELEASE_ASSERT(false,
 240                                "Wrong range of atoms requested in GPU state data manager. Should "
 241                                "be All, Local or NonLocal.");
 242     }
 243     GMX_ASSERT(atomsStartAt >= 0,
 244                "The first elemtnt to copy has negative index. Probably, the GPU propagator state "
 245                "was not initialized.");
 246     GMX_ASSERT(numAtomsToCopy >= 0,
 247                "Number of atoms to copy is negative. Probably, the GPU propagator state was not "
 248                "initialized.");
 249     return std::make_tuple(atomsStartAt, numAtomsToCopy);
 250 }
 251
 252 void StatePropagatorDataGpu::Impl::copyToDevice(DeviceBuffer<RVec>                   d_data,
 253                                                 const gmx::ArrayRef<const gmx::RVec> h_data,
 254                                                 int                                  dataSize,
 255                                                 AtomLocality                         atomLocality,
 256                                                 CommandStream                        commandStream)
 257 {
 258     GMX_UNUSED_VALUE(dataSize);
 259
 260     GMX_ASSERT(atomLocality < AtomLocality::Count, "Wrong atom locality.");
 261
 262     GMX_ASSERT(dataSize >= 0, "Trying to copy to device buffer before it was allocated.");
 263
 264     GMX_ASSERT(commandStream != nullptr,
 265                "No stream is valid for copying with given atom locality.");
 266     wallcycle_start_nocount(wcycle_, ewcLAUNCH_GPU);
 267     wallcycle_sub_start(wcycle_, ewcsLAUNCH_STATE_PROPAGATOR_DATA);
 268
 269     int atomsStartAt, numAtomsToCopy;
 270     std::tie(atomsStartAt, numAtomsToCopy) = getAtomRangesFromAtomLocality(atomLocality);
 271
 272     if (numAtomsToCopy != 0)
 273     {
 274         GMX_ASSERT(atomsStartAt + numAtomsToCopy <= dataSize,
 275                    "The device allocation is smaller than requested copy range.");
 276         GMX_ASSERT(atomsStartAt + numAtomsToCopy <= h_data.ssize(),
 277                    "The host buffer is smaller than the requested copy range.");
 278
 279         copyToDeviceBuffer(&d_data, reinterpret_cast<const RVec*>(&h_data.data()[atomsStartAt]),
 280                            atomsStartAt, numAtomsToCopy, commandStream, transferKind_, nullptr);
 281     }
 282
 283     wallcycle_sub_stop(wcycle_, ewcsLAUNCH_STATE_PROPAGATOR_DATA);
 284     wallcycle_stop(wcycle_, ewcLAUNCH_GPU);
 285 }
 286
 287 void StatePropagatorDataGpu::Impl::copyFromDevice(gmx::ArrayRef<gmx::RVec> h_data,
 288                                                   DeviceBuffer<RVec>       d_data,
 289                                                   int                      dataSize,
 290                                                   AtomLocality             atomLocality,
 291                                                   CommandStream            commandStream)
 292 {
 293     GMX_UNUSED_VALUE(dataSize);
 294
 295     GMX_ASSERT(atomLocality < AtomLocality::Count, "Wrong atom locality.");
 296
 297     GMX_ASSERT(dataSize >= 0, "Trying to copy from device buffer before it was allocated.");
 298
 299     GMX_ASSERT(commandStream != nullptr,
 300                "No stream is valid for copying with given atom locality.");
 301     wallcycle_start_nocount(wcycle_, ewcLAUNCH_GPU);
 302     wallcycle_sub_start(wcycle_, ewcsLAUNCH_STATE_PROPAGATOR_DATA);
 303
 304     int atomsStartAt, numAtomsToCopy;
 305     std::tie(atomsStartAt, numAtomsToCopy) = getAtomRangesFromAtomLocality(atomLocality);
 306
 307     if (numAtomsToCopy != 0)
 308     {
 309         GMX_ASSERT(atomsStartAt + numAtomsToCopy <= dataSize,
 310                    "The device allocation is smaller than requested copy range.");
 311         GMX_ASSERT(atomsStartAt + numAtomsToCopy <= h_data.ssize(),
 312                    "The host buffer is smaller than the requested copy range.");
 313
 314         copyFromDeviceBuffer(reinterpret_cast<RVec*>(&h_data.data()[atomsStartAt]), &d_data,
 315                              atomsStartAt, numAtomsToCopy, commandStream, transferKind_, nullptr);
 316     }
 317
 318     wallcycle_sub_stop(wcycle_, ewcsLAUNCH_STATE_PROPAGATOR_DATA);
 319     wallcycle_stop(wcycle_, ewcLAUNCH_GPU);
 320 }
 321
 322 DeviceBuffer<RVec> StatePropagatorDataGpu::Impl::getCoordinates()
 323 {
 324     return d_x_;
 325 }
 326
 327 void StatePropagatorDataGpu::Impl::copyCoordinatesToGpu(const gmx::ArrayRef<const gmx::RVec> h_x,
 328                                                         AtomLocality atomLocality)
 329 {
 330     GMX_ASSERT(atomLocality < AtomLocality::Count, "Wrong atom locality.");
 331     CommandStream commandStream = xCopyStreams_[atomLocality];
 332     GMX_ASSERT(commandStream != nullptr,
 333                "No stream is valid for copying positions with given atom locality.");
 334
 335     wallcycle_start_nocount(wcycle_, ewcLAUNCH_GPU);
 336     wallcycle_sub_start(wcycle_, ewcsLAUNCH_STATE_PROPAGATOR_DATA);
 337
 338     copyToDevice(d_x_, h_x, d_xSize_, atomLocality, commandStream);
 339
 340     // markEvent is skipped in OpenCL as:
 341     //   - it's not needed, copy is done in the same stream as the only consumer task (PME)
 342     //   - we don't consume the events in OpenCL which is not allowed by GpuEventSynchronizer (would leak memory).
 343     // TODO: remove this by adding an event-mark free flavor of this function
 344     if (GMX_GPU == GMX_GPU_CUDA)
 345     {
 346         xReadyOnDevice_[atomLocality].markEvent(xCopyStreams_[atomLocality]);
 347     }
 348
 349     wallcycle_sub_stop(wcycle_, ewcsLAUNCH_STATE_PROPAGATOR_DATA);
 350     wallcycle_stop(wcycle_, ewcLAUNCH_GPU);
 351 }
 352
 353 GpuEventSynchronizer*
 354 StatePropagatorDataGpu::Impl::getCoordinatesReadyOnDeviceEvent(AtomLocality atomLocality,
 355                                                                const SimulationWorkload& simulationWork,
 356                                                                const StepWorkload&       stepWork)
 357 {
 358     // The provider of the coordinates may be different for local atoms. If the update is offloaded
 359     // and this is not a neighbor search step, then the consumer needs to wait for the update
 360     // to complete. Otherwise, the coordinates are copied from the host and we need to wait for
 361     // the copy event. Non-local coordinates are always provided by the H2D copy.
 362     //
 363     // TODO: This should be reconsidered to support the halo exchange.
 364     //
 365     // In OpenCL no events are used as coordinate sync is not necessary
 366     if (GMX_GPU == GMX_GPU_OPENCL)
 367     {
 368         return nullptr;
 369     }
 370     if (atomLocality == AtomLocality::Local && simulationWork.useGpuUpdate && !stepWork.doNeighborSearch)
 371     {
 372         return &xUpdatedOnDevice_;
 373     }
 374     else
 375     {
 376         return &xReadyOnDevice_[atomLocality];
 377     }
 378 }
 379
 380 void StatePropagatorDataGpu::Impl::waitCoordinatesCopiedToDevice(AtomLocality atomLocality)
 381 {
 382     wallcycle_start(wcycle_, ewcWAIT_GPU_STATE_PROPAGATOR_DATA);
 383     GMX_ASSERT(atomLocality < AtomLocality::Count, "Wrong atom locality.");
 384     xReadyOnDevice_[atomLocality].waitForEvent();
 385     wallcycle_stop(wcycle_, ewcWAIT_GPU_STATE_PROPAGATOR_DATA);
 386 }
 387
 388 GpuEventSynchronizer* StatePropagatorDataGpu::Impl::xUpdatedOnDevice()
 389 {
 390     return &xUpdatedOnDevice_;
 391 }
 392
 393 void StatePropagatorDataGpu::Impl::copyCoordinatesFromGpu(gmx::ArrayRef<gmx::RVec> h_x, AtomLocality atomLocality)
 394 {
 395     GMX_ASSERT(atomLocality < AtomLocality::Count, "Wrong atom locality.");
 396     CommandStream commandStream = xCopyStreams_[atomLocality];
 397     GMX_ASSERT(commandStream != nullptr,
 398                "No stream is valid for copying positions with given atom locality.");
 399
 400     wallcycle_start_nocount(wcycle_, ewcLAUNCH_GPU);
 401     wallcycle_sub_start(wcycle_, ewcsLAUNCH_STATE_PROPAGATOR_DATA);
 402
 403     copyFromDevice(h_x, d_x_, d_xSize_, atomLocality, commandStream);
 404     // Note: unlike copyCoordinatesToGpu this is not used in OpenCL, and the conditional is not needed.
 405     xReadyOnHost_[atomLocality].markEvent(commandStream);
 406
 407     wallcycle_sub_stop(wcycle_, ewcsLAUNCH_STATE_PROPAGATOR_DATA);
 408     wallcycle_stop(wcycle_, ewcLAUNCH_GPU);
 409 }
 410
 411 void StatePropagatorDataGpu::Impl::waitCoordinatesReadyOnHost(AtomLocality atomLocality)
 412 {
 413     wallcycle_start(wcycle_, ewcWAIT_GPU_STATE_PROPAGATOR_DATA);
 414     xReadyOnHost_[atomLocality].waitForEvent();
 415     wallcycle_stop(wcycle_, ewcWAIT_GPU_STATE_PROPAGATOR_DATA);
 416 }
 417
 418
 419 DeviceBuffer<RVec> StatePropagatorDataGpu::Impl::getVelocities()
 420 {
 421     return d_v_;
 422 }
 423
 424 void StatePropagatorDataGpu::Impl::copyVelocitiesToGpu(const gmx::ArrayRef<const gmx::RVec> h_v,
 425                                                        AtomLocality atomLocality)
 426 {
 427     GMX_ASSERT(atomLocality < AtomLocality::Count, "Wrong atom locality.");
 428     CommandStream commandStream = vCopyStreams_[atomLocality];
 429     GMX_ASSERT(commandStream != nullptr,
 430                "No stream is valid for copying velocities with given atom locality.");
 431
 432     wallcycle_start_nocount(wcycle_, ewcLAUNCH_GPU);
 433     wallcycle_sub_start(wcycle_, ewcsLAUNCH_STATE_PROPAGATOR_DATA);
 434
 435     copyToDevice(d_v_, h_v, d_vSize_, atomLocality, commandStream);
 436     vReadyOnDevice_[atomLocality].markEvent(commandStream);
 437
 438     wallcycle_sub_stop(wcycle_, ewcsLAUNCH_STATE_PROPAGATOR_DATA);
 439     wallcycle_stop(wcycle_, ewcLAUNCH_GPU);
 440 }
 441
 442 GpuEventSynchronizer* StatePropagatorDataGpu::Impl::getVelocitiesReadyOnDeviceEvent(AtomLocality atomLocality)
 443 {
 444     return &vReadyOnDevice_[atomLocality];
 445 }
 446
 447
 448 void StatePropagatorDataGpu::Impl::copyVelocitiesFromGpu(gmx::ArrayRef<gmx::RVec> h_v, AtomLocality atomLocality)
 449 {
 450     GMX_ASSERT(atomLocality < AtomLocality::Count, "Wrong atom locality.");
 451     CommandStream commandStream = vCopyStreams_[atomLocality];
 452     GMX_ASSERT(commandStream != nullptr,
 453                "No stream is valid for copying velocities with given atom locality.");
 454
 455     wallcycle_start_nocount(wcycle_, ewcLAUNCH_GPU);
 456     wallcycle_sub_start(wcycle_, ewcsLAUNCH_STATE_PROPAGATOR_DATA);
 457
 458     copyFromDevice(h_v, d_v_, d_vSize_, atomLocality, commandStream);
 459     vReadyOnHost_[atomLocality].markEvent(commandStream);
 460
 461     wallcycle_sub_stop(wcycle_, ewcsLAUNCH_STATE_PROPAGATOR_DATA);
 462     wallcycle_stop(wcycle_, ewcLAUNCH_GPU);
 463 }
 464
 465 void StatePropagatorDataGpu::Impl::waitVelocitiesReadyOnHost(AtomLocality atomLocality)
 466 {
 467     wallcycle_start(wcycle_, ewcWAIT_GPU_STATE_PROPAGATOR_DATA);
 468     vReadyOnHost_[atomLocality].waitForEvent();
 469     wallcycle_stop(wcycle_, ewcWAIT_GPU_STATE_PROPAGATOR_DATA);
 470 }
 471
 472
 473 DeviceBuffer<RVec> StatePropagatorDataGpu::Impl::getForces()
 474 {
 475     return d_f_;
 476 }
 477
 478 void StatePropagatorDataGpu::Impl::copyForcesToGpu(const gmx::ArrayRef<const gmx::RVec> h_f,
 479                                                    AtomLocality atomLocality)
 480 {
 481     GMX_ASSERT(atomLocality < AtomLocality::Count, "Wrong atom locality.");
 482     CommandStream commandStream = fCopyStreams_[atomLocality];
 483     GMX_ASSERT(commandStream != nullptr,
 484                "No stream is valid for copying forces with given atom locality.");
 485
 486     wallcycle_start_nocount(wcycle_, ewcLAUNCH_GPU);
 487     wallcycle_sub_start(wcycle_, ewcsLAUNCH_STATE_PROPAGATOR_DATA);
 488
 489     copyToDevice(d_f_, h_f, d_fSize_, atomLocality, commandStream);
 490     fReadyOnDevice_[atomLocality].markEvent(commandStream);
 491
 492     wallcycle_sub_stop(wcycle_, ewcsLAUNCH_STATE_PROPAGATOR_DATA);
 493     wallcycle_stop(wcycle_, ewcLAUNCH_GPU);
 494 }
 495
 496 GpuEventSynchronizer* StatePropagatorDataGpu::Impl::getForcesReadyOnDeviceEvent(AtomLocality atomLocality,
 497                                                                                 bool useGpuFBufferOps)
 498 {
 499     if ((atomLocality == AtomLocality::Local || atomLocality == AtomLocality::NonLocal) && useGpuFBufferOps)
 500     {
 501         return &fReducedOnDevice_;
 502     }
 503     else
 504     {
 505         return &fReadyOnDevice_[atomLocality];
 506     }
 507 }
 508
 509 GpuEventSynchronizer* StatePropagatorDataGpu::Impl::fReducedOnDevice()
 510 {
 511     return &fReducedOnDevice_;
 512 }
 513
 514 void StatePropagatorDataGpu::Impl::copyForcesFromGpu(gmx::ArrayRef<gmx::RVec> h_f, AtomLocality atomLocality)
 515 {
 516     GMX_ASSERT(atomLocality < AtomLocality::Count, "Wrong atom locality.");
 517     CommandStream commandStream = fCopyStreams_[atomLocality];
 518     GMX_ASSERT(commandStream != nullptr,
 519                "No stream is valid for copying forces with given atom locality.");
 520
 521     wallcycle_start_nocount(wcycle_, ewcLAUNCH_GPU);
 522     wallcycle_sub_start(wcycle_, ewcsLAUNCH_STATE_PROPAGATOR_DATA);
 523
 524     copyFromDevice(h_f, d_f_, d_fSize_, atomLocality, commandStream);
 525     fReadyOnHost_[atomLocality].markEvent(commandStream);
 526
 527     wallcycle_sub_stop(wcycle_, ewcsLAUNCH_STATE_PROPAGATOR_DATA);
 528     wallcycle_stop(wcycle_, ewcLAUNCH_GPU);
 529 }
 530
 531 void StatePropagatorDataGpu::Impl::waitForcesReadyOnHost(AtomLocality atomLocality)
 532 {
 533     wallcycle_start(wcycle_, ewcWAIT_GPU_STATE_PROPAGATOR_DATA);
 534     fReadyOnHost_[atomLocality].waitForEvent();
 535     wallcycle_stop(wcycle_, ewcWAIT_GPU_STATE_PROPAGATOR_DATA);
 536 }
 537
 538 void* StatePropagatorDataGpu::Impl::getUpdateStream()
 539 {
 540     return &updateStream_;
 541 }
 542
 543 int StatePropagatorDataGpu::Impl::numAtomsLocal()
 544 {
 545     return numAtomsLocal_;
 546 }
 547
 548 int StatePropagatorDataGpu::Impl::numAtomsAll()
 549 {
 550     return numAtomsAll_;
 551 }
 552
 553
 554 StatePropagatorDataGpu::StatePropagatorDataGpu(const void*        pmeStream,
 555                                                const void*        localStream,
 556                                                const void*        nonLocalStream,
 557                                                const void*        deviceContext,
 558                                                GpuApiCallBehavior transferKind,
 559                                                int                paddingSize,
 560                                                gmx_wallcycle*     wcycle) :
 561     impl_(new Impl(pmeStream, localStream, nonLocalStream, deviceContext, transferKind, paddingSize, wcycle))
 562 {
 563 }
 564
 565 StatePropagatorDataGpu::StatePropagatorDataGpu(const void*        pmeStream,
 566                                                const void*        deviceContext,
 567                                                GpuApiCallBehavior transferKind,
 568                                                int                paddingSize,
 569                                                gmx_wallcycle*     wcycle) :
 570     impl_(new Impl(pmeStream, deviceContext, transferKind, paddingSize, wcycle))
 571 {
 572 }
 573
 574 StatePropagatorDataGpu::StatePropagatorDataGpu(StatePropagatorDataGpu&& /* other */) noexcept = default;
 575
 576 StatePropagatorDataGpu& StatePropagatorDataGpu::operator=(StatePropagatorDataGpu&& /* other */) noexcept = default;
 577
 578 StatePropagatorDataGpu::~StatePropagatorDataGpu() = default;
 579
 580
 581 void StatePropagatorDataGpu::reinit(int numAtomsLocal, int numAtomsAll)
 582 {
 583     return impl_->reinit(numAtomsLocal, numAtomsAll);
 584 }
 585
 586 std::tuple<int, int> StatePropagatorDataGpu::getAtomRangesFromAtomLocality(AtomLocality atomLocality)
 587 {
 588     return impl_->getAtomRangesFromAtomLocality(atomLocality);
 589 }
 590
 591
 592 DeviceBuffer<RVec> StatePropagatorDataGpu::getCoordinates()
 593 {
 594     return impl_->getCoordinates();
 595 }
 596
 597 void StatePropagatorDataGpu::copyCoordinatesToGpu(const gmx::ArrayRef<const gmx::RVec> h_x,
 598                                                   AtomLocality                         atomLocality)
 599 {
 600     return impl_->copyCoordinatesToGpu(h_x, atomLocality);
 601 }
 602
 603 GpuEventSynchronizer*
 604 StatePropagatorDataGpu::getCoordinatesReadyOnDeviceEvent(AtomLocality              atomLocality,
 605                                                          const SimulationWorkload& simulationWork,
 606                                                          const StepWorkload&       stepWork)
 607 {
 608     return impl_->getCoordinatesReadyOnDeviceEvent(atomLocality, simulationWork, stepWork);
 609 }
 610
 611 void StatePropagatorDataGpu::waitCoordinatesCopiedToDevice(AtomLocality atomLocality)
 612 {
 613     return impl_->waitCoordinatesCopiedToDevice(atomLocality);
 614 }
 615
 616 GpuEventSynchronizer* StatePropagatorDataGpu::xUpdatedOnDevice()
 617 {
 618     return impl_->xUpdatedOnDevice();
 619 }
 620
 621 void StatePropagatorDataGpu::copyCoordinatesFromGpu(gmx::ArrayRef<RVec> h_x, AtomLocality atomLocality)
 622 {
 623     return impl_->copyCoordinatesFromGpu(h_x, atomLocality);
 624 }
 625
 626 void StatePropagatorDataGpu::waitCoordinatesReadyOnHost(AtomLocality atomLocality)
 627 {
 628     return impl_->waitCoordinatesReadyOnHost(atomLocality);
 629 }
 630
 631
 632 DeviceBuffer<RVec> StatePropagatorDataGpu::getVelocities()
 633 {
 634     return impl_->getVelocities();
 635 }
 636
 637 void StatePropagatorDataGpu::copyVelocitiesToGpu(const gmx::ArrayRef<const gmx::RVec> h_v,
 638                                                  AtomLocality                         atomLocality)
 639 {
 640     return impl_->copyVelocitiesToGpu(h_v, atomLocality);
 641 }
 642
 643 GpuEventSynchronizer* StatePropagatorDataGpu::getVelocitiesReadyOnDeviceEvent(AtomLocality atomLocality)
 644 {
 645     return impl_->getVelocitiesReadyOnDeviceEvent(atomLocality);
 646 }
 647
 648 void StatePropagatorDataGpu::copyVelocitiesFromGpu(gmx::ArrayRef<RVec> h_v, AtomLocality atomLocality)
 649 {
 650     return impl_->copyVelocitiesFromGpu(h_v, atomLocality);
 651 }
 652
 653 void StatePropagatorDataGpu::waitVelocitiesReadyOnHost(AtomLocality atomLocality)
 654 {
 655     return impl_->waitVelocitiesReadyOnHost(atomLocality);
 656 }
 657
 658
 659 DeviceBuffer<RVec> StatePropagatorDataGpu::getForces()
 660 {
 661     return impl_->getForces();
 662 }
 663
 664 void StatePropagatorDataGpu::copyForcesToGpu(const gmx::ArrayRef<const gmx::RVec> h_f, AtomLocality atomLocality)
 665 {
 666     return impl_->copyForcesToGpu(h_f, atomLocality);
 667 }
 668
 669 GpuEventSynchronizer* StatePropagatorDataGpu::getForcesReadyOnDeviceEvent(AtomLocality atomLocality,
 670                                                                           bool useGpuFBufferOps)
 671 {
 672     return impl_->getForcesReadyOnDeviceEvent(atomLocality, useGpuFBufferOps);
 673 }
 674
 675 GpuEventSynchronizer* StatePropagatorDataGpu::fReducedOnDevice()
 676 {
 677     return impl_->fReducedOnDevice();
 678 }
 679
 680 void StatePropagatorDataGpu::copyForcesFromGpu(gmx::ArrayRef<RVec> h_f, AtomLocality atomLocality)
 681 {
 682     return impl_->copyForcesFromGpu(h_f, atomLocality);
 683 }
 684
 685 void StatePropagatorDataGpu::waitForcesReadyOnHost(AtomLocality atomLocality)
 686 {
 687     return impl_->waitForcesReadyOnHost(atomLocality);
 688 }
 689
 690
 691 void* StatePropagatorDataGpu::getUpdateStream()
 692 {
 693     return impl_->getUpdateStream();
 694 }
 695
 696 int StatePropagatorDataGpu::numAtomsLocal()
 697 {
 698     return impl_->numAtomsLocal();
 699 }
 700
 701 int StatePropagatorDataGpu::numAtomsAll()
 702 {
 703     return impl_->numAtomsAll();
 704 }
 705
 706 } // namespace gmx
 707
 708 #endif // GMX_GPU == GMX_GPU_NONE