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,
  75                   "This object should only be constructed on the GPU code-paths.");
  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]      = updateStream_;
 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,
 139                   "This object should only be constructed on the GPU code-paths.");
 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 void StatePropagatorDataGpu::Impl::reinit(int numAtomsLocal, int numAtomsAll)
 173 {
 174     numAtomsLocal_ = numAtomsLocal;
 175     numAtomsAll_   = numAtomsAll;
 176
 177     int numAtomsPadded;
 178     if (paddingSize_ > 0)
 179     {
 180         numAtomsPadded = ((numAtomsAll_ + paddingSize_ - 1) / paddingSize_) * paddingSize_;
 181     }
 182     else
 183     {
 184         numAtomsPadded = numAtomsAll_;
 185     }
 186
 187     reallocateDeviceBuffer(&d_x_, DIM * numAtomsPadded, &d_xSize_, &d_xCapacity_, deviceContext_);
 188
 189     const size_t paddingAllocationSize = numAtomsPadded - numAtomsAll_;
 190     if (paddingAllocationSize > 0)
 191     {
 192         // The PME stream is used here because the padding region of d_x_ is only in the PME task.
 193         clearDeviceBufferAsync(&d_x_, DIM * numAtomsAll_, DIM * paddingAllocationSize, pmeStream_);
 194     }
 195
 196     reallocateDeviceBuffer(&d_v_, DIM * numAtomsAll_, &d_vSize_, &d_vCapacity_, deviceContext_);
 197     const int d_fOldCapacity = d_fCapacity_;
 198     reallocateDeviceBuffer(&d_f_, DIM * numAtomsAll_, &d_fSize_, &d_fCapacity_, deviceContext_);
 199     // Clearing of the forces can be done in local stream since the nonlocal stream cannot reach
 200     // the force accumulation stage before syncing with the local stream. Only done in CUDA,
 201     // since the force buffer ops are not implemented in OpenCL.
 202     if (GMX_GPU == GMX_GPU_CUDA && d_fCapacity_ != d_fOldCapacity)
 203     {
 204         clearDeviceBufferAsync(&d_f_, 0, d_fCapacity_, localStream_);
 205     }
 206 }
 207
 208 std::tuple<int, int> StatePropagatorDataGpu::Impl::getAtomRangesFromAtomLocality(AtomLocality atomLocality)
 209 {
 210     int atomsStartAt   = 0;
 211     int numAtomsToCopy = 0;
 212     switch (atomLocality)
 213     {
 214         case AtomLocality::All:
 215             atomsStartAt   = 0;
 216             numAtomsToCopy = numAtomsAll_;
 217             break;
 218         case AtomLocality::Local:
 219             atomsStartAt   = 0;
 220             numAtomsToCopy = numAtomsLocal_;
 221             break;
 222         case AtomLocality::NonLocal:
 223             atomsStartAt   = numAtomsLocal_;
 224             numAtomsToCopy = numAtomsAll_ - numAtomsLocal_;
 225             break;
 226         default:
 227             GMX_RELEASE_ASSERT(false,
 228                                "Wrong range of atoms requested in GPU state data manager. Should "
 229                                "be All, Local or NonLocal.");
 230     }
 231     GMX_ASSERT(atomsStartAt >= 0,
 232                "The first elemtnt to copy has negative index. Probably, the GPU propagator state "
 233                "was not initialized.");
 234     GMX_ASSERT(numAtomsToCopy >= 0,
 235                "Number of atoms to copy is negative. Probably, the GPU propagator state was not "
 236                "initialized.");
 237     return std::make_tuple(atomsStartAt, numAtomsToCopy);
 238 }
 239
 240 void StatePropagatorDataGpu::Impl::copyToDevice(DeviceBuffer<float>                  d_data,
 241                                                 const gmx::ArrayRef<const gmx::RVec> h_data,
 242                                                 int                                  dataSize,
 243                                                 AtomLocality                         atomLocality,
 244                                                 CommandStream                        commandStream)
 245 {
 246
 247     GMX_UNUSED_VALUE(dataSize);
 248
 249     GMX_ASSERT(atomLocality < AtomLocality::Count, "Wrong atom locality.");
 250
 251     GMX_ASSERT(dataSize >= 0, "Trying to copy to device buffer before it was allocated.");
 252
 253     GMX_ASSERT(commandStream != nullptr,
 254                "No stream is valid for copying with given atom locality.");
 255
 256     int atomsStartAt, numAtomsToCopy;
 257     std::tie(atomsStartAt, numAtomsToCopy) = getAtomRangesFromAtomLocality(atomLocality);
 258
 259     int elementsStartAt   = atomsStartAt * DIM;
 260     int numElementsToCopy = numAtomsToCopy * DIM;
 261
 262     if (numAtomsToCopy != 0)
 263     {
 264         GMX_ASSERT(elementsStartAt + numElementsToCopy <= dataSize,
 265                    "The device allocation is smaller than requested copy range.");
 266         GMX_ASSERT(atomsStartAt + numAtomsToCopy <= h_data.ssize(),
 267                    "The host buffer is smaller than the requested copy range.");
 268
 269         copyToDeviceBuffer(&d_data, reinterpret_cast<const float*>(&h_data.data()[atomsStartAt]),
 270                            elementsStartAt, numElementsToCopy, commandStream, transferKind_, nullptr);
 271     }
 272 }
 273
 274 void StatePropagatorDataGpu::Impl::copyFromDevice(gmx::ArrayRef<gmx::RVec> h_data,
 275                                                   DeviceBuffer<float>      d_data,
 276                                                   int                      dataSize,
 277                                                   AtomLocality             atomLocality,
 278                                                   CommandStream            commandStream)
 279 {
 280
 281     GMX_UNUSED_VALUE(dataSize);
 282
 283     GMX_ASSERT(atomLocality < AtomLocality::Count, "Wrong atom locality.");
 284
 285     GMX_ASSERT(dataSize >= 0, "Trying to copy from device buffer before it was allocated.");
 286
 287     GMX_ASSERT(commandStream != nullptr,
 288                "No stream is valid for copying with given atom locality.");
 289
 290     int atomsStartAt, numAtomsToCopy;
 291     std::tie(atomsStartAt, numAtomsToCopy) = getAtomRangesFromAtomLocality(atomLocality);
 292
 293     int elementsStartAt   = atomsStartAt * DIM;
 294     int numElementsToCopy = numAtomsToCopy * DIM;
 295
 296     if (numAtomsToCopy != 0)
 297     {
 298         GMX_ASSERT(elementsStartAt + numElementsToCopy <= dataSize,
 299                    "The device allocation is smaller than requested copy range.");
 300         GMX_ASSERT(atomsStartAt + numAtomsToCopy <= h_data.ssize(),
 301                    "The host buffer is smaller than the requested copy range.");
 302
 303         copyFromDeviceBuffer(reinterpret_cast<float*>(&h_data.data()[atomsStartAt]), &d_data,
 304                              elementsStartAt, numElementsToCopy, commandStream, transferKind_, nullptr);
 305     }
 306 }
 307
 308 DeviceBuffer<float> StatePropagatorDataGpu::Impl::getCoordinates()
 309 {
 310     return d_x_;
 311 }
 312
 313 void StatePropagatorDataGpu::Impl::copyCoordinatesToGpu(const gmx::ArrayRef<const gmx::RVec> h_x,
 314                                                         AtomLocality atomLocality)
 315 {
 316     copyToDevice(d_x_, h_x, d_xSize_, atomLocality, xCopyStreams_[atomLocality]);
 317
 318     // markEvent is skipped in OpenCL as:
 319     //   - it's not needed, copy is done in the same stream as the only consumer task (PME)
 320     //   - we don't consume the events in OpenCL which is not allowed by GpuEventSynchronizer (would leak memory).
 321     // TODO: remove this by adding an event-mark free flavor of this function
 322     if (GMX_GPU == GMX_GPU_CUDA)
 323     {
 324         xReadyOnDevice_[atomLocality].markEvent(xCopyStreams_[atomLocality]);
 325     }
 326 }
 327
 328 GpuEventSynchronizer*
 329 StatePropagatorDataGpu::Impl::getCoordinatesReadyOnDeviceEvent(AtomLocality atomLocality,
 330                                                                const SimulationWorkload& simulationWork,
 331                                                                const StepWorkload&       stepWork)
 332 {
 333     // The provider of the coordinates may be different for local atoms. If the update is offloaded
 334     // and this is not a neighbor search step, then the consumer needs to wait for the update
 335     // to complete. Otherwise, the coordinates are copied from the host and we need to wait for
 336     // the copy event. Non-local coordinates are always provided by the H2D copy.
 337     //
 338     // TODO: This should be reconsidered to support the halo exchange.
 339     //
 340     // In OpenCL no events are used as coordinate sync is not necessary
 341     if (GMX_GPU == GMX_GPU_OPENCL)
 342     {
 343         return nullptr;
 344     }
 345     if (atomLocality == AtomLocality::Local && simulationWork.useGpuUpdate && !stepWork.doNeighborSearch)
 346     {
 347         return &xUpdatedOnDevice_;
 348     }
 349     else
 350     {
 351         return &xReadyOnDevice_[atomLocality];
 352     }
 353 }
 354
 355 void StatePropagatorDataGpu::Impl::waitCoordinatesCopiedToDevice(AtomLocality atomLocality)
 356 {
 357     GMX_ASSERT(atomLocality < AtomLocality::Count, "Wrong atom locality.");
 358     xReadyOnDevice_[atomLocality].waitForEvent();
 359 }
 360
 361 GpuEventSynchronizer* StatePropagatorDataGpu::Impl::xUpdatedOnDevice()
 362 {
 363     return &xUpdatedOnDevice_;
 364 }
 365
 366 void StatePropagatorDataGpu::Impl::copyCoordinatesFromGpu(gmx::ArrayRef<gmx::RVec> h_x, AtomLocality atomLocality)
 367 {
 368     copyFromDevice(h_x, d_x_, d_xSize_, atomLocality, xCopyStreams_[atomLocality]);
 369     // Note: unlike copyCoordinatesToGpu this is not used in OpenCL, and the conditional is not needed.
 370     xReadyOnHost_[atomLocality].markEvent(xCopyStreams_[atomLocality]);
 371 }
 372
 373 void StatePropagatorDataGpu::Impl::waitCoordinatesReadyOnHost(AtomLocality atomLocality)
 374 {
 375     xReadyOnHost_[atomLocality].waitForEvent();
 376 }
 377
 378
 379 DeviceBuffer<float> StatePropagatorDataGpu::Impl::getVelocities()
 380 {
 381     return d_v_;
 382 }
 383
 384 void StatePropagatorDataGpu::Impl::copyVelocitiesToGpu(const gmx::ArrayRef<const gmx::RVec> h_v,
 385                                                        AtomLocality atomLocality)
 386 {
 387     copyToDevice(d_v_, h_v, d_vSize_, atomLocality, vCopyStreams_[atomLocality]);
 388     vReadyOnDevice_[atomLocality].markEvent(vCopyStreams_[atomLocality]);
 389 }
 390
 391 GpuEventSynchronizer* StatePropagatorDataGpu::Impl::getVelocitiesReadyOnDeviceEvent(AtomLocality atomLocality)
 392 {
 393     return &vReadyOnDevice_[atomLocality];
 394 }
 395
 396
 397 void StatePropagatorDataGpu::Impl::copyVelocitiesFromGpu(gmx::ArrayRef<gmx::RVec> h_v, AtomLocality atomLocality)
 398 {
 399     copyFromDevice(h_v, d_v_, d_vSize_, atomLocality, vCopyStreams_[atomLocality]);
 400     vReadyOnHost_[atomLocality].markEvent(vCopyStreams_[atomLocality]);
 401 }
 402
 403 void StatePropagatorDataGpu::Impl::waitVelocitiesReadyOnHost(AtomLocality atomLocality)
 404 {
 405     vReadyOnHost_[atomLocality].waitForEvent();
 406 }
 407
 408
 409 DeviceBuffer<float> StatePropagatorDataGpu::Impl::getForces()
 410 {
 411     return d_f_;
 412 }
 413
 414 void StatePropagatorDataGpu::Impl::copyForcesToGpu(const gmx::ArrayRef<const gmx::RVec> h_f,
 415                                                    AtomLocality atomLocality)
 416 {
 417     copyToDevice(d_f_, h_f, d_fSize_, atomLocality, fCopyStreams_[atomLocality]);
 418     fReadyOnDevice_[atomLocality].markEvent(fCopyStreams_[atomLocality]);
 419 }
 420
 421 GpuEventSynchronizer* StatePropagatorDataGpu::Impl::getForcesReadyOnDeviceEvent(AtomLocality atomLocality,
 422                                                                                 bool useGpuFBufferOps)
 423 {
 424     if ((atomLocality == AtomLocality::Local || atomLocality == AtomLocality::NonLocal) && useGpuFBufferOps)
 425     {
 426         return &fReducedOnDevice_;
 427     }
 428     else
 429     {
 430         return &fReadyOnDevice_[atomLocality];
 431     }
 432 }
 433
 434 GpuEventSynchronizer* StatePropagatorDataGpu::Impl::fReducedOnDevice()
 435 {
 436     return &fReducedOnDevice_;
 437 }
 438
 439 void StatePropagatorDataGpu::Impl::copyForcesFromGpu(gmx::ArrayRef<gmx::RVec> h_f, AtomLocality atomLocality)
 440 {
 441     copyFromDevice(h_f, d_f_, d_fSize_, atomLocality, fCopyStreams_[atomLocality]);
 442     fReadyOnHost_[atomLocality].markEvent(fCopyStreams_[atomLocality]);
 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 StatePropagatorDataGpu::StatePropagatorDataGpu(const void*        pmeStream,
 467                                                const void*        localStream,
 468                                                const void*        nonLocalStream,
 469                                                const void*        deviceContext,
 470                                                GpuApiCallBehavior transferKind,
 471                                                int                paddingSize) :
 472     impl_(new Impl(pmeStream, localStream, nonLocalStream, deviceContext, transferKind, paddingSize))
 473 {
 474 }
 475
 476 StatePropagatorDataGpu::StatePropagatorDataGpu(const void*        pmeStream,
 477                                                const void*        deviceContext,
 478                                                GpuApiCallBehavior transferKind,
 479                                                int                paddingSize) :
 480     impl_(new Impl(pmeStream, deviceContext, transferKind, paddingSize))
 481 {
 482 }
 483
 484 StatePropagatorDataGpu::StatePropagatorDataGpu(StatePropagatorDataGpu&& /* other */) noexcept = default;
 485
 486 StatePropagatorDataGpu& StatePropagatorDataGpu::operator=(StatePropagatorDataGpu&& /* other */) noexcept = default;
 487
 488 StatePropagatorDataGpu::~StatePropagatorDataGpu() = default;
 489
 490
 491 void StatePropagatorDataGpu::reinit(int numAtomsLocal, int numAtomsAll)
 492 {
 493     return impl_->reinit(numAtomsLocal, numAtomsAll);
 494 }
 495
 496 std::tuple<int, int> StatePropagatorDataGpu::getAtomRangesFromAtomLocality(AtomLocality atomLocality)
 497 {
 498     return impl_->getAtomRangesFromAtomLocality(atomLocality);
 499 }
 500
 501
 502 DeviceBuffer<float> StatePropagatorDataGpu::getCoordinates()
 503 {
 504     return impl_->getCoordinates();
 505 }
 506
 507 void StatePropagatorDataGpu::copyCoordinatesToGpu(const gmx::ArrayRef<const gmx::RVec> h_x,
 508                                                   AtomLocality                         atomLocality)
 509 {
 510     return impl_->copyCoordinatesToGpu(h_x, atomLocality);
 511 }
 512
 513 GpuEventSynchronizer*
 514 StatePropagatorDataGpu::getCoordinatesReadyOnDeviceEvent(AtomLocality              atomLocality,
 515                                                          const SimulationWorkload& simulationWork,
 516                                                          const StepWorkload&       stepWork)
 517 {
 518     return impl_->getCoordinatesReadyOnDeviceEvent(atomLocality, simulationWork, stepWork);
 519 }
 520
 521 void StatePropagatorDataGpu::waitCoordinatesCopiedToDevice(AtomLocality atomLocality)
 522 {
 523     return impl_->waitCoordinatesCopiedToDevice(atomLocality);
 524 }
 525
 526 GpuEventSynchronizer* StatePropagatorDataGpu::xUpdatedOnDevice()
 527 {
 528     return impl_->xUpdatedOnDevice();
 529 }
 530
 531 void StatePropagatorDataGpu::copyCoordinatesFromGpu(gmx::ArrayRef<RVec> h_x, AtomLocality atomLocality)
 532 {
 533     return impl_->copyCoordinatesFromGpu(h_x, atomLocality);
 534 }
 535
 536 void StatePropagatorDataGpu::waitCoordinatesReadyOnHost(AtomLocality atomLocality)
 537 {
 538     return impl_->waitCoordinatesReadyOnHost(atomLocality);
 539 }
 540
 541
 542 DeviceBuffer<float> StatePropagatorDataGpu::getVelocities()
 543 {
 544     return impl_->getVelocities();
 545 }
 546
 547 void StatePropagatorDataGpu::copyVelocitiesToGpu(const gmx::ArrayRef<const gmx::RVec> h_v,
 548                                                  AtomLocality                         atomLocality)
 549 {
 550     return impl_->copyVelocitiesToGpu(h_v, atomLocality);
 551 }
 552
 553 GpuEventSynchronizer* StatePropagatorDataGpu::getVelocitiesReadyOnDeviceEvent(AtomLocality atomLocality)
 554 {
 555     return impl_->getVelocitiesReadyOnDeviceEvent(atomLocality);
 556 }
 557
 558 void StatePropagatorDataGpu::copyVelocitiesFromGpu(gmx::ArrayRef<RVec> h_v, AtomLocality atomLocality)
 559 {
 560     return impl_->copyVelocitiesFromGpu(h_v, atomLocality);
 561 }
 562
 563 void StatePropagatorDataGpu::waitVelocitiesReadyOnHost(AtomLocality atomLocality)
 564 {
 565     return impl_->waitVelocitiesReadyOnHost(atomLocality);
 566 }
 567
 568
 569 DeviceBuffer<float> StatePropagatorDataGpu::getForces()
 570 {
 571     return impl_->getForces();
 572 }
 573
 574 void StatePropagatorDataGpu::copyForcesToGpu(const gmx::ArrayRef<const gmx::RVec> h_f, AtomLocality atomLocality)
 575 {
 576     return impl_->copyForcesToGpu(h_f, atomLocality);
 577 }
 578
 579 GpuEventSynchronizer* StatePropagatorDataGpu::getForcesReadyOnDeviceEvent(AtomLocality atomLocality,
 580                                                                           bool useGpuFBufferOps)
 581 {
 582     return impl_->getForcesReadyOnDeviceEvent(atomLocality, useGpuFBufferOps);
 583 }
 584
 585 GpuEventSynchronizer* StatePropagatorDataGpu::fReducedOnDevice()
 586 {
 587     return impl_->fReducedOnDevice();
 588 }
 589
 590 void StatePropagatorDataGpu::copyForcesFromGpu(gmx::ArrayRef<RVec> h_f, AtomLocality atomLocality)
 591 {
 592     return impl_->copyForcesFromGpu(h_f, atomLocality);
 593 }
 594
 595 void StatePropagatorDataGpu::waitForcesReadyOnHost(AtomLocality atomLocality)
 596 {
 597     return impl_->waitForcesReadyOnHost(atomLocality);
 598 }
 599
 600
 601 void* StatePropagatorDataGpu::getUpdateStream()
 602 {
 603     return impl_->getUpdateStream();
 604 }
 605
 606 int StatePropagatorDataGpu::numAtomsLocal()
 607 {
 608     return impl_->numAtomsLocal();
 609 }
 610
 611 int StatePropagatorDataGpu::numAtomsAll()
 612 {
 613     return impl_->numAtomsAll();
 614 }
 615
 616 } // namespace gmx
 617
 618 #endif // GMX_GPU == GMX_GPU_NONE