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, 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/utility/classhelpers.h"
  59
  60 #include "state_propagator_data_gpu_impl.h"
  61
  62 namespace gmx
  63 {
  64
  65 StatePropagatorDataGpu::Impl::Impl(const void            *pmeStream,
  66                                    const void            *localStream,
  67                                    const void            *nonLocalStream,
  68                                    const void            *deviceContext,
  69                                    GpuApiCallBehavior     transferKind,
  70                                    int                    paddingSize) :
  71     transferKind_(transferKind),
  72     paddingSize_(paddingSize)
  73 {
  74     static_assert(GMX_GPU != GMX_GPU_NONE, "This object should only be constructed on the GPU code-paths.");
  75     GMX_RELEASE_ASSERT(getenv("GMX_USE_GPU_BUFFER_OPS") == nullptr, "GPU buffer ops are not supported in this build.");
  76
  77     // TODO: Refactor when the StreamManager is introduced.
  78     if (GMX_GPU == GMX_GPU_OPENCL)
  79     {
  80         GMX_ASSERT(deviceContext != nullptr, "GPU context should be set in OpenCL builds.");
  81         GMX_ASSERT(pmeStream != nullptr, "GPU PME stream should be set in OpenCL builds.");
  82
  83         // The update stream is set to the PME stream in OpenCL, since PME stream is the only stream created in the PME context.
  84         pmeStream_      = *static_cast<const CommandStream*>(pmeStream);
  85         updateStream_   = *static_cast<const CommandStream*>(pmeStream);
  86         deviceContext_  = *static_cast<const DeviceContext*>(deviceContext);
  87         GMX_UNUSED_VALUE(localStream);
  88         GMX_UNUSED_VALUE(nonLocalStream);
  89     }
  90
  91     if (GMX_GPU == GMX_GPU_CUDA)
  92     {
  93         if (pmeStream != nullptr)
  94         {
  95             pmeStream_ = *static_cast<const CommandStream*>(pmeStream);
  96         }
  97         if (localStream != nullptr)
  98         {
  99             localStream_ = *static_cast<const CommandStream*>(localStream);
 100         }
 101         if (nonLocalStream != nullptr)
 102         {
 103             nonLocalStream_ = *static_cast<const CommandStream*>(nonLocalStream);
 104         }
 105
 106         // TODO: The update stream should be created only when it is needed.
 107 #if (GMX_GPU == GMX_GPU_CUDA)
 108         cudaError_t stat;
 109         stat = cudaStreamCreate(&updateStream_);
 110         CU_RET_ERR(stat, "CUDA stream creation failed in StatePropagatorDataGpu");
 111 #endif
 112         GMX_UNUSED_VALUE(deviceContext);
 113     }
 114
 115     // Map the atom locality to the stream that will be used for coordinates,
 116     // velocities and forces transfers. Same streams are used for H2D and D2H copies.
 117     // Note, that nullptr stream is used here to indicate that the copy is not supported.
 118     xCopyStreams_[AtomLocality::Local]    = updateStream_;
 119     xCopyStreams_[AtomLocality::NonLocal] = nonLocalStream_;
 120     xCopyStreams_[AtomLocality::All]      = updateStream_;
 121
 122     vCopyStreams_[AtomLocality::Local]    = updateStream_;
 123     vCopyStreams_[AtomLocality::NonLocal] = nullptr;
 124     vCopyStreams_[AtomLocality::All]      = updateStream_;
 125
 126     fCopyStreams_[AtomLocality::Local]    = localStream_;
 127     fCopyStreams_[AtomLocality::NonLocal] = nonLocalStream_;
 128     fCopyStreams_[AtomLocality::All]      = nullptr;
 129 }
 130
 131 StatePropagatorDataGpu::Impl::Impl(const void            *pmeStream,
 132                                    const void            *deviceContext,
 133                                    GpuApiCallBehavior     transferKind,
 134                                    int                    paddingSize) :
 135     transferKind_(transferKind),
 136     paddingSize_(paddingSize)
 137 {
 138     static_assert(GMX_GPU != GMX_GPU_NONE, "This object should only be constructed on the GPU code-paths.");
 139     GMX_RELEASE_ASSERT(getenv("GMX_USE_GPU_BUFFER_OPS") == nullptr, "GPU buffer ops are not supported in this build.");
 140
 141     if (GMX_GPU == GMX_GPU_OPENCL)
 142     {
 143         GMX_ASSERT(deviceContext != nullptr, "GPU context should be set in OpenCL builds.");
 144         deviceContext_  = *static_cast<const DeviceContext*>(deviceContext);
 145     }
 146
 147     GMX_ASSERT(pmeStream != nullptr, "GPU PME stream should be set.");
 148     pmeStream_      = *static_cast<const CommandStream*>(pmeStream);
 149
 150     localStream_    = nullptr;
 151     nonLocalStream_ = nullptr;
 152     updateStream_   = nullptr;
 153
 154
 155     // Only local/all coordinates are allowed to be copied in PME-only rank/ PME tests.
 156     // This it temporary measure to make it safe to use this class in those cases.
 157     xCopyStreams_[AtomLocality::Local]    = pmeStream_;
 158     xCopyStreams_[AtomLocality::NonLocal] = nullptr;
 159     xCopyStreams_[AtomLocality::All]      = pmeStream_;
 160
 161     vCopyStreams_[AtomLocality::Local]    = nullptr;
 162     vCopyStreams_[AtomLocality::NonLocal] = nullptr;
 163     vCopyStreams_[AtomLocality::All]      = nullptr;
 164
 165     fCopyStreams_[AtomLocality::Local]    = nullptr;
 166     fCopyStreams_[AtomLocality::NonLocal] = nullptr;
 167     fCopyStreams_[AtomLocality::All]      = nullptr;
 168 }
 169
 170 StatePropagatorDataGpu::Impl::~Impl()
 171 {
 172 }
 173
 174 void StatePropagatorDataGpu::Impl::reinit(int numAtomsLocal, int numAtomsAll)
 175 {
 176     numAtomsLocal_ = numAtomsLocal;
 177     numAtomsAll_   = numAtomsAll;
 178
 179     int numAtomsPadded;
 180     if (paddingSize_ > 0)
 181     {
 182         numAtomsPadded = ((numAtomsAll_ + paddingSize_ - 1 ) / paddingSize_ )*paddingSize_;
 183     }
 184     else
 185     {
 186         numAtomsPadded = numAtomsAll_;
 187     }
 188
 189     reallocateDeviceBuffer(&d_x_, DIM*numAtomsPadded, &d_xSize_, &d_xCapacity_, deviceContext_);
 190
 191     const size_t paddingAllocationSize = numAtomsPadded - numAtomsAll_;
 192     if (paddingAllocationSize > 0)
 193     {
 194         // The PME stream is used here because the padding region of d_x_ is only in the PME task.
 195         clearDeviceBufferAsync(&d_x_, DIM*numAtomsAll_, DIM*paddingAllocationSize, pmeStream_);
 196     }
 197
 198     reallocateDeviceBuffer(&d_v_, DIM*numAtomsAll_, &d_vSize_, &d_vCapacity_, deviceContext_);
 199     reallocateDeviceBuffer(&d_f_, DIM*numAtomsAll_, &d_fSize_, &d_fCapacity_, deviceContext_);
 200
 201 }
 202
 203 std::tuple<int, int> StatePropagatorDataGpu::Impl::getAtomRangesFromAtomLocality(AtomLocality  atomLocality)
 204 {
 205     int atomsStartAt   = 0;
 206     int numAtomsToCopy = 0;
 207     switch (atomLocality)
 208     {
 209         case AtomLocality::All:
 210             atomsStartAt    = 0;
 211             numAtomsToCopy  = numAtomsAll_;
 212             break;
 213         case AtomLocality::Local:
 214             atomsStartAt    = 0;
 215             numAtomsToCopy  = numAtomsLocal_;
 216             break;
 217         case AtomLocality::NonLocal:
 218             atomsStartAt    = numAtomsLocal_;
 219             numAtomsToCopy  = numAtomsAll_ - numAtomsLocal_;
 220             break;
 221         default:
 222             GMX_RELEASE_ASSERT(false, "Wrong range of atoms requested in GPU state data manager. Should be All, Local or NonLocal.");
 223     }
 224     GMX_ASSERT(atomsStartAt   >= 0, "The first elemtnt to copy has negative index. Probably, the GPU propagator state was not initialized.");
 225     GMX_ASSERT(numAtomsToCopy >= 0, "Number of atoms to copy is negative. Probably, the GPU propagator state was not initialized.");
 226     return std::make_tuple(atomsStartAt, numAtomsToCopy);
 227 }
 228
 229 void StatePropagatorDataGpu::Impl::copyToDevice(DeviceBuffer<float>                   d_data,
 230                                                 const gmx::ArrayRef<const gmx::RVec>  h_data,
 231                                                 int                                   dataSize,
 232                                                 AtomLocality                          atomLocality,
 233                                                 CommandStream                         commandStream)
 234 {
 235
 236     GMX_UNUSED_VALUE(dataSize);
 237
 238     GMX_ASSERT(dataSize >= 0, "Trying to copy to device buffer before it was allocated.");
 239
 240     int atomsStartAt, numAtomsToCopy;
 241     std::tie(atomsStartAt, numAtomsToCopy) = getAtomRangesFromAtomLocality(atomLocality);
 242
 243     int elementsStartAt   = atomsStartAt*DIM;
 244     int numElementsToCopy = numAtomsToCopy*DIM;
 245
 246     if (numAtomsToCopy != 0)
 247     {
 248         GMX_ASSERT(elementsStartAt + numElementsToCopy <= dataSize, "The device allocation is smaller than requested copy range.");
 249         GMX_ASSERT(atomsStartAt + numAtomsToCopy <= h_data.ssize(), "The host buffer is smaller than the requested copy range.");
 250
 251         copyToDeviceBuffer(&d_data, reinterpret_cast<const float *>(&h_data.data()[atomsStartAt]),
 252                            elementsStartAt, numElementsToCopy,
 253                            commandStream, transferKind_, nullptr);
 254     }
 255 }
 256
 257 void StatePropagatorDataGpu::Impl::copyFromDevice(gmx::ArrayRef<gmx::RVec>  h_data,
 258                                                   DeviceBuffer<float>       d_data,
 259                                                   int                       dataSize,
 260                                                   AtomLocality              atomLocality,
 261                                                   CommandStream             commandStream)
 262 {
 263
 264     GMX_UNUSED_VALUE(dataSize);
 265
 266     GMX_ASSERT(dataSize >= 0, "Trying to copy from device buffer before it was allocated.");
 267
 268     int atomsStartAt, numAtomsToCopy;
 269     std::tie(atomsStartAt, numAtomsToCopy) = getAtomRangesFromAtomLocality(atomLocality);
 270
 271     int elementsStartAt   = atomsStartAt*DIM;
 272     int numElementsToCopy = numAtomsToCopy*DIM;
 273
 274     if (numAtomsToCopy != 0)
 275     {
 276         GMX_ASSERT(elementsStartAt + numElementsToCopy <= dataSize, "The device allocation is smaller than requested copy range.");
 277         GMX_ASSERT(atomsStartAt + numAtomsToCopy <= h_data.ssize(), "The host buffer is smaller than the requested copy range.");
 278
 279         copyFromDeviceBuffer(reinterpret_cast<float*>(&h_data.data()[atomsStartAt]), &d_data,
 280                              elementsStartAt, numElementsToCopy,
 281                              commandStream, transferKind_, nullptr);
 282     }
 283 }
 284
 285 DeviceBuffer<float> StatePropagatorDataGpu::Impl::getCoordinates()
 286 {
 287     return d_x_;
 288 }
 289
 290 void StatePropagatorDataGpu::Impl::copyCoordinatesToGpu(const gmx::ArrayRef<const gmx::RVec>  h_x,
 291                                                         AtomLocality                          atomLocality)
 292 {
 293     GMX_ASSERT(atomLocality < AtomLocality::Count, "Wrong atom locality.");
 294     CommandStream commandStream = xCopyStreams_[atomLocality];
 295     GMX_ASSERT(commandStream != nullptr, "No stream is valid for copying positions with given atom locality.");
 296
 297     copyToDevice(d_x_, h_x, d_xSize_, atomLocality, commandStream);
 298
 299     // markEvent is skipped in OpenCL as:
 300     //   - it's not needed, copy is done in the same stream as the only consumer task (PME)
 301     //   - we don't consume the events in OpenCL which is not allowed by GpuEventSynchronizer (would leak memory).
 302     // TODO: remove this by adding an event-mark free flavor of this function
 303     if (GMX_GPU == GMX_GPU_CUDA)
 304     {
 305         xReadyOnDevice_[atomLocality].markEvent(commandStream);
 306         // TODO: Remove When event-based synchronization is introduced
 307         gpuStreamSynchronize(commandStream);
 308     }
 309 }
 310
 311 GpuEventSynchronizer* StatePropagatorDataGpu::Impl::getCoordinatesReadyOnDeviceEvent(AtomLocality              atomLocality,
 312                                                                                      const SimulationWorkload &simulationWork,
 313                                                                                      const StepWorkload       &stepWork)
 314 {
 315     // The provider of the coordinates may be different for local atoms. If the update is offloaded
 316     // and this is not a neighbor search step, then the consumer needs to wait for the update
 317     // to complete. Otherwise, the coordinates are copied from the host and we need to wait for
 318     // the copy event. Non-local coordinates are always provided by the H2D copy.
 319     //
 320     // TODO: This should be reconsidered to support the halo exchange.
 321     //
 322     if (atomLocality == AtomLocality::Local && simulationWork.useGpuUpdate && !stepWork.doNeighborSearch)
 323     {
 324         return &xUpdatedOnDevice_;
 325     }
 326     else
 327     {
 328         return &xReadyOnDevice_[atomLocality];
 329     }
 330 }
 331
 332 GpuEventSynchronizer* StatePropagatorDataGpu::Impl::xUpdatedOnDevice()
 333 {
 334     return &xUpdatedOnDevice_;
 335 }
 336
 337 void StatePropagatorDataGpu::Impl::copyCoordinatesFromGpu(gmx::ArrayRef<gmx::RVec>  h_x,
 338                                                           AtomLocality              atomLocality)
 339 {
 340     GMX_ASSERT(atomLocality < AtomLocality::Count, "Wrong atom locality.");
 341     CommandStream commandStream = xCopyStreams_[atomLocality];
 342     GMX_ASSERT(commandStream != nullptr, "No stream is valid for copying positions with given atom locality.");
 343
 344     copyFromDevice(h_x, d_x_, d_xSize_, atomLocality, commandStream);
 345     // TODO: Remove When event-based synchronization is introduced
 346     gpuStreamSynchronize(commandStream);
 347     // Note: unlike copyCoordinatesToGpu this is not used in OpenCL, and the conditional is not needed.
 348     xReadyOnHost_[atomLocality].markEvent(commandStream);
 349 }
 350
 351 void StatePropagatorDataGpu::Impl::waitCoordinatesReadyOnHost(AtomLocality  atomLocality)
 352 {
 353     xReadyOnHost_[atomLocality].waitForEvent();
 354 }
 355
 356
 357 DeviceBuffer<float> StatePropagatorDataGpu::Impl::getVelocities()
 358 {
 359     return d_v_;
 360 }
 361
 362 void StatePropagatorDataGpu::Impl::copyVelocitiesToGpu(const gmx::ArrayRef<const gmx::RVec>  h_v,
 363                                                        AtomLocality                          atomLocality)
 364 {
 365     GMX_ASSERT(atomLocality < AtomLocality::Count, "Wrong atom locality.");
 366     CommandStream commandStream = vCopyStreams_[atomLocality];
 367     GMX_ASSERT(commandStream != nullptr, "No stream is valid for copying velocities with given atom locality.");
 368
 369     copyToDevice(d_v_, h_v, d_vSize_, atomLocality, commandStream);
 370     // TODO: Remove When event-based synchronization is introduced
 371     gpuStreamSynchronize(commandStream);
 372     vReadyOnDevice_[atomLocality].markEvent(commandStream);
 373 }
 374
 375 GpuEventSynchronizer* StatePropagatorDataGpu::Impl::getVelocitiesReadyOnDeviceEvent(AtomLocality  atomLocality)
 376 {
 377     return &vReadyOnDevice_[atomLocality];
 378 }
 379
 380
 381 void StatePropagatorDataGpu::Impl::copyVelocitiesFromGpu(gmx::ArrayRef<gmx::RVec>  h_v,
 382                                                          AtomLocality              atomLocality)
 383 {
 384     GMX_ASSERT(atomLocality < AtomLocality::Count, "Wrong atom locality.");
 385     CommandStream commandStream = vCopyStreams_[atomLocality];
 386     GMX_ASSERT(commandStream != nullptr, "No stream is valid for copying velocities with given atom locality.");
 387
 388     copyFromDevice(h_v, d_v_, d_vSize_, atomLocality, commandStream);
 389     // TODO: Remove When event-based synchronization is introduced
 390     gpuStreamSynchronize(commandStream);
 391     vReadyOnHost_[atomLocality].markEvent(commandStream);
 392 }
 393
 394 void StatePropagatorDataGpu::Impl::waitVelocitiesReadyOnHost(AtomLocality  atomLocality)
 395 {
 396     vReadyOnHost_[atomLocality].waitForEvent();
 397 }
 398
 399
 400 DeviceBuffer<float> StatePropagatorDataGpu::Impl::getForces()
 401 {
 402     return d_f_;
 403 }
 404
 405 void StatePropagatorDataGpu::Impl::copyForcesToGpu(const gmx::ArrayRef<const gmx::RVec>  h_f,
 406                                                    AtomLocality                          atomLocality)
 407 {
 408     GMX_ASSERT(atomLocality < AtomLocality::Count, "Wrong atom locality.");
 409     CommandStream commandStream = fCopyStreams_[atomLocality];
 410     GMX_ASSERT(commandStream != nullptr, "No stream is valid for copying forces with given atom locality.");
 411
 412     copyToDevice(d_f_, h_f, d_fSize_, atomLocality, commandStream);
 413     fReadyOnDevice_[atomLocality].markEvent(commandStream);
 414 }
 415
 416 GpuEventSynchronizer* StatePropagatorDataGpu::Impl::getForcesReadyOnDeviceEvent(AtomLocality  atomLocality,
 417                                                                                 bool          useGpuFBufferOps)
 418 {
 419     if ((atomLocality == AtomLocality::Local || atomLocality == AtomLocality::NonLocal) && useGpuFBufferOps)
 420     {
 421         return &fReducedOnDevice_;
 422     }
 423     else
 424     {
 425         return &fReadyOnDevice_[atomLocality];
 426     }
 427 }
 428
 429 GpuEventSynchronizer* StatePropagatorDataGpu::Impl::fReducedOnDevice()
 430 {
 431     return &fReducedOnDevice_;
 432 }
 433
 434 void StatePropagatorDataGpu::Impl::copyForcesFromGpu(gmx::ArrayRef<gmx::RVec>  h_f,
 435                                                      AtomLocality              atomLocality)
 436 {
 437     GMX_ASSERT(atomLocality < AtomLocality::Count, "Wrong atom locality.");
 438     CommandStream commandStream = fCopyStreams_[atomLocality];
 439     GMX_ASSERT(commandStream != nullptr, "No stream is valid for copying forces with given atom locality.");
 440
 441     copyFromDevice(h_f, d_f_, d_fSize_, atomLocality, commandStream);
 442     fReadyOnHost_[atomLocality].markEvent(commandStream);
 443 }
 444
 445 void StatePropagatorDataGpu::Impl::waitForcesReadyOnHost(AtomLocality  atomLocality)
 446 {
 447     fReadyOnHost_[atomLocality].waitForEvent();
 448 }
 449
 450 void* StatePropagatorDataGpu::Impl::getUpdateStream()
 451 {
 452     return updateStream_;
 453 }
 454
 455 int StatePropagatorDataGpu::Impl::numAtomsLocal()
 456 {
 457     return numAtomsLocal_;
 458 }
 459
 460 int StatePropagatorDataGpu::Impl::numAtomsAll()
 461 {
 462     return numAtomsAll_;
 463 }
 464
 465
 466
 467 StatePropagatorDataGpu::StatePropagatorDataGpu(const void        *pmeStream,
 468                                                const void        *localStream,
 469                                                const void        *nonLocalStream,
 470                                                const void        *deviceContext,
 471                                                GpuApiCallBehavior transferKind,
 472                                                int                paddingSize)
 473     : impl_(new Impl(pmeStream,
 474                      localStream,
 475                      nonLocalStream,
 476                      deviceContext,
 477                      transferKind,
 478                      paddingSize))
 479 {
 480 }
 481
 482 StatePropagatorDataGpu::StatePropagatorDataGpu(const void        *pmeStream,
 483                                                const void        *deviceContext,
 484                                                GpuApiCallBehavior transferKind,
 485                                                int                paddingSize)
 486     : impl_(new Impl(pmeStream,
 487                      deviceContext,
 488                      transferKind,
 489                      paddingSize))
 490 {
 491 }
 492
 493 StatePropagatorDataGpu::StatePropagatorDataGpu(StatePropagatorDataGpu && /* other */) noexcept = default;
 494
 495 StatePropagatorDataGpu &StatePropagatorDataGpu::operator=(StatePropagatorDataGpu && /* other */) noexcept = default;
 496
 497 StatePropagatorDataGpu::~StatePropagatorDataGpu() = default;
 498
 499
 500 void StatePropagatorDataGpu::reinit(int numAtomsLocal, int numAtomsAll)
 501 {
 502     return impl_->reinit(numAtomsLocal, numAtomsAll);
 503 }
 504
 505 std::tuple<int, int> StatePropagatorDataGpu::getAtomRangesFromAtomLocality(AtomLocality  atomLocality)
 506 {
 507     return impl_->getAtomRangesFromAtomLocality(atomLocality);
 508 }
 509
 510
 511 DeviceBuffer<float> StatePropagatorDataGpu::getCoordinates()
 512 {
 513     return impl_->getCoordinates();
 514 }
 515
 516 void StatePropagatorDataGpu::copyCoordinatesToGpu(const gmx::ArrayRef<const gmx::RVec>  h_x,
 517                                                   AtomLocality                          atomLocality)
 518 {
 519     return impl_->copyCoordinatesToGpu(h_x, atomLocality);
 520 }
 521
 522 GpuEventSynchronizer* StatePropagatorDataGpu::getCoordinatesReadyOnDeviceEvent(AtomLocality              atomLocality,
 523                                                                                const SimulationWorkload &simulationWork,
 524                                                                                const StepWorkload       &stepWork)
 525 {
 526     return impl_->getCoordinatesReadyOnDeviceEvent(atomLocality, simulationWork, stepWork);
 527 }
 528
 529 GpuEventSynchronizer* StatePropagatorDataGpu::xUpdatedOnDevice()
 530 {
 531     return impl_->xUpdatedOnDevice();
 532 }
 533
 534 void StatePropagatorDataGpu::copyCoordinatesFromGpu(gmx::ArrayRef<RVec>  h_x,
 535                                                     AtomLocality         atomLocality)
 536 {
 537     return impl_->copyCoordinatesFromGpu(h_x, atomLocality);
 538 }
 539
 540 void StatePropagatorDataGpu::waitCoordinatesReadyOnHost(AtomLocality  atomLocality)
 541 {
 542     return impl_->waitCoordinatesReadyOnHost(atomLocality);
 543 }
 544
 545
 546 DeviceBuffer<float> StatePropagatorDataGpu::getVelocities()
 547 {
 548     return impl_->getVelocities();
 549 }
 550
 551 void StatePropagatorDataGpu::copyVelocitiesToGpu(const gmx::ArrayRef<const gmx::RVec>  h_v,
 552                                                  AtomLocality                          atomLocality)
 553 {
 554     return impl_->copyVelocitiesToGpu(h_v, atomLocality);
 555 }
 556
 557 GpuEventSynchronizer* StatePropagatorDataGpu::getVelocitiesReadyOnDeviceEvent(AtomLocality  atomLocality)
 558 {
 559     return impl_->getVelocitiesReadyOnDeviceEvent(atomLocality);
 560 }
 561
 562 void StatePropagatorDataGpu::copyVelocitiesFromGpu(gmx::ArrayRef<RVec>  h_v,
 563                                                    AtomLocality         atomLocality)
 564 {
 565     return impl_->copyVelocitiesFromGpu(h_v, atomLocality);
 566 }
 567
 568 void StatePropagatorDataGpu::waitVelocitiesReadyOnHost(AtomLocality  atomLocality)
 569 {
 570     return impl_->waitVelocitiesReadyOnHost(atomLocality);
 571 }
 572
 573
 574 DeviceBuffer<float> StatePropagatorDataGpu::getForces()
 575 {
 576     return impl_->getForces();
 577 }
 578
 579 void StatePropagatorDataGpu::copyForcesToGpu(const gmx::ArrayRef<const gmx::RVec>  h_f,
 580                                              AtomLocality                          atomLocality)
 581 {
 582     return impl_->copyForcesToGpu(h_f, atomLocality);
 583 }
 584
 585 GpuEventSynchronizer* StatePropagatorDataGpu::getForcesReadyOnDeviceEvent(AtomLocality  atomLocality,
 586                                                                           bool          useGpuFBufferOps)
 587 {
 588     return impl_->getForcesReadyOnDeviceEvent(atomLocality, useGpuFBufferOps);
 589 }
 590
 591 GpuEventSynchronizer* StatePropagatorDataGpu::fReducedOnDevice()
 592 {
 593     return impl_->fReducedOnDevice();
 594 }
 595
 596 void StatePropagatorDataGpu::copyForcesFromGpu(gmx::ArrayRef<RVec>  h_f,
 597                                                AtomLocality         atomLocality)
 598 {
 599     return impl_->copyForcesFromGpu(h_f, atomLocality);
 600 }
 601
 602 void StatePropagatorDataGpu::waitForcesReadyOnHost(AtomLocality  atomLocality)
 603 {
 604     return impl_->waitForcesReadyOnHost(atomLocality);
 605 }
 606
 607
 608 void* StatePropagatorDataGpu::getUpdateStream()
 609 {
 610     return impl_->getUpdateStream();
 611 }
 612
 613 int StatePropagatorDataGpu::numAtomsLocal()
 614 {
 615     return impl_->numAtomsLocal();
 616 }
 617
 618 int StatePropagatorDataGpu::numAtomsAll()
 619 {
 620     return impl_->numAtomsAll();
 621 }
 622
 623 }      // namespace gmx
 624
 625 #endif // GMX_GPU == GMX_GPU_NONE