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
  48
  49 #    include "gromacs/gpu_utils/device_stream_manager.h"
  50 #    include "gromacs/gpu_utils/devicebuffer.h"
  51 #    include "gromacs/gpu_utils/gputraits.h"
  52 #    include "gromacs/math/vectypes.h"
  53 #    include "gromacs/mdtypes/state_propagator_data_gpu.h"
  54 #    include "gromacs/timing/wallcycle.h"
  55 #    include "gromacs/utility/classhelpers.h"
  56
  57 #    include "state_propagator_data_gpu_impl.h"
  58
  59
  60 namespace gmx
  61 {
  62
  63 StatePropagatorDataGpu::Impl::Impl(const DeviceStreamManager& deviceStreamManager,
  64                                    GpuApiCallBehavior         transferKind,
  65                                    int                        allocationBlockSizeDivisor,
  66                                    gmx_wallcycle*             wcycle) :
  67     deviceContext_(deviceStreamManager.context()),
  68     transferKind_(transferKind),
  69     allocationBlockSizeDivisor_(allocationBlockSizeDivisor),
  70     wcycle_(wcycle)
  71 {
  72     static_assert(
  73             GMX_GPU,
  74             "GPU state propagator data object should only be constructed on the GPU code-paths.");
  75
  76     // We need to keep local copies for re-initialization.
  77     pmeStream_      = &deviceStreamManager.stream(DeviceStreamType::Pme);
  78     localStream_    = &deviceStreamManager.stream(DeviceStreamType::NonBondedLocal);
  79     nonLocalStream_ = &deviceStreamManager.stream(DeviceStreamType::NonBondedNonLocal);
  80     // PME stream is used in OpenCL for H2D coordinate transfer
  81     updateStream_ = &deviceStreamManager.stream(
  82             GMX_GPU_OPENCL ? DeviceStreamType::Pme : DeviceStreamType::UpdateAndConstraints);
  83
  84     // Map the atom locality to the stream that will be used for coordinates,
  85     // velocities and forces transfers. Same streams are used for H2D and D2H copies.
  86     // Note, that nullptr stream is used here to indicate that the copy is not supported.
  87     xCopyStreams_[AtomLocality::Local]    = updateStream_;
  88     xCopyStreams_[AtomLocality::NonLocal] = nonLocalStream_;
  89     xCopyStreams_[AtomLocality::All]      = updateStream_;
  90
  91     vCopyStreams_[AtomLocality::Local]    = updateStream_;
  92     vCopyStreams_[AtomLocality::NonLocal] = nullptr;
  93     vCopyStreams_[AtomLocality::All]      = updateStream_;
  94
  95     fCopyStreams_[AtomLocality::Local]    = localStream_;
  96     fCopyStreams_[AtomLocality::NonLocal] = nonLocalStream_;
  97     fCopyStreams_[AtomLocality::All]      = updateStream_;
  98 }
  99
 100 StatePropagatorDataGpu::Impl::Impl(const DeviceStream*  pmeStream,
 101                                    const DeviceContext& deviceContext,
 102                                    GpuApiCallBehavior   transferKind,
 103                                    int                  allocationBlockSizeDivisor,
 104                                    gmx_wallcycle*       wcycle) :
 105     deviceContext_(deviceContext),
 106     transferKind_(transferKind),
 107     allocationBlockSizeDivisor_(allocationBlockSizeDivisor),
 108     wcycle_(wcycle)
 109 {
 110     static_assert(
 111             GMX_GPU,
 112             "GPU state propagator data object should only be constructed on the GPU code-paths.");
 113
 114     GMX_ASSERT(pmeStream->isValid(), "GPU PME stream should be valid.");
 115     pmeStream_      = pmeStream;
 116     localStream_    = pmeStream; // For clearing the force buffer
 117     nonLocalStream_ = nullptr;
 118     updateStream_   = nullptr;
 119
 120
 121     // Only local/all coordinates are allowed to be copied in PME-only rank/ PME tests.
 122     // This it temporary measure to make it safe to use this class in those cases.
 123     xCopyStreams_[AtomLocality::Local]    = pmeStream_;
 124     xCopyStreams_[AtomLocality::NonLocal] = nullptr;
 125     xCopyStreams_[AtomLocality::All]      = pmeStream_;
 126
 127     vCopyStreams_[AtomLocality::Local]    = nullptr;
 128     vCopyStreams_[AtomLocality::NonLocal] = nullptr;
 129     vCopyStreams_[AtomLocality::All]      = nullptr;
 130
 131     fCopyStreams_[AtomLocality::Local]    = nullptr;
 132     fCopyStreams_[AtomLocality::NonLocal] = nullptr;
 133     fCopyStreams_[AtomLocality::All]      = nullptr;
 134 }
 135
 136 StatePropagatorDataGpu::Impl::~Impl() {}
 137
 138 void StatePropagatorDataGpu::Impl::reinit(int numAtomsLocal, int numAtomsAll)
 139 {
 140     wallcycle_start_nocount(wcycle_, ewcLAUNCH_GPU);
 141     wallcycle_sub_start_nocount(wcycle_, ewcsLAUNCH_STATE_PROPAGATOR_DATA);
 142
 143     numAtomsLocal_ = numAtomsLocal;
 144     numAtomsAll_   = numAtomsAll;
 145
 146     int numAtomsPadded;
 147     if (allocationBlockSizeDivisor_ > 0)
 148     {
 149         numAtomsPadded = ((numAtomsAll_ + allocationBlockSizeDivisor_ - 1) / allocationBlockSizeDivisor_)
 150                          * allocationBlockSizeDivisor_;
 151     }
 152     else
 153     {
 154         numAtomsPadded = numAtomsAll_;
 155     }
 156
 157     reallocateDeviceBuffer(&d_x_, numAtomsPadded, &d_xSize_, &d_xCapacity_, deviceContext_);
 158
 159     const size_t paddingAllocationSize = numAtomsPadded - numAtomsAll_;
 160     if (paddingAllocationSize > 0)
 161     {
 162         // The PME stream is used here because the padding region of d_x_ is only in the PME task.
 163         clearDeviceBufferAsync(&d_x_, numAtomsAll_, paddingAllocationSize, *pmeStream_);
 164     }
 165
 166     reallocateDeviceBuffer(&d_v_, numAtomsAll_, &d_vSize_, &d_vCapacity_, deviceContext_);
 167     const int d_fOldCapacity = d_fCapacity_;
 168     reallocateDeviceBuffer(&d_f_, numAtomsAll_, &d_fSize_, &d_fCapacity_, deviceContext_);
 169
 170     // Clearing of the forces can be done in local stream since the nonlocal stream cannot reach
 171     // the force accumulation stage before syncing with the local stream. Only done in CUDA,
 172     // since the force buffer ops are not implemented in OpenCL.
 173     if (GMX_GPU_CUDA && d_fCapacity_ != d_fOldCapacity)
 174     {
 175         clearDeviceBufferAsync(&d_f_, 0, d_fCapacity_, *localStream_);
 176     }
 177
 178     wallcycle_sub_stop(wcycle_, ewcsLAUNCH_STATE_PROPAGATOR_DATA);
 179     wallcycle_stop(wcycle_, ewcLAUNCH_GPU);
 180 }
 181
 182 std::tuple<int, int> StatePropagatorDataGpu::Impl::getAtomRangesFromAtomLocality(AtomLocality atomLocality)
 183 {
 184     int atomsStartAt   = 0;
 185     int numAtomsToCopy = 0;
 186     switch (atomLocality)
 187     {
 188         case AtomLocality::All:
 189             atomsStartAt   = 0;
 190             numAtomsToCopy = numAtomsAll_;
 191             break;
 192         case AtomLocality::Local:
 193             atomsStartAt   = 0;
 194             numAtomsToCopy = numAtomsLocal_;
 195             break;
 196         case AtomLocality::NonLocal:
 197             atomsStartAt   = numAtomsLocal_;
 198             numAtomsToCopy = numAtomsAll_ - numAtomsLocal_;
 199             break;
 200         default:
 201             GMX_RELEASE_ASSERT(false,
 202                                "Wrong range of atoms requested in GPU state data manager. Should "
 203                                "be All, Local or NonLocal.");
 204     }
 205     GMX_ASSERT(atomsStartAt >= 0,
 206                "The first elemtnt to copy has negative index. Probably, the GPU propagator state "
 207                "was not initialized.");
 208     GMX_ASSERT(numAtomsToCopy >= 0,
 209                "Number of atoms to copy is negative. Probably, the GPU propagator state was not "
 210                "initialized.");
 211     return std::make_tuple(atomsStartAt, numAtomsToCopy);
 212 }
 213
 214 void StatePropagatorDataGpu::Impl::copyToDevice(DeviceBuffer<RVec>                   d_data,
 215                                                 const gmx::ArrayRef<const gmx::RVec> h_data,
 216                                                 int                                  dataSize,
 217                                                 AtomLocality                         atomLocality,
 218                                                 const DeviceStream&                  deviceStream)
 219 {
 220     GMX_UNUSED_VALUE(dataSize);
 221
 222     GMX_ASSERT(atomLocality < AtomLocality::Count, "Wrong atom locality.");
 223
 224     GMX_ASSERT(dataSize >= 0, "Trying to copy to device buffer before it was allocated.");
 225
 226     GMX_ASSERT(deviceStream.isValid(), "No stream is valid for copying with given atom locality.");
 227     wallcycle_start_nocount(wcycle_, ewcLAUNCH_GPU);
 228     wallcycle_sub_start(wcycle_, ewcsLAUNCH_STATE_PROPAGATOR_DATA);
 229
 230     int atomsStartAt, numAtomsToCopy;
 231     std::tie(atomsStartAt, numAtomsToCopy) = getAtomRangesFromAtomLocality(atomLocality);
 232
 233     if (numAtomsToCopy != 0)
 234     {
 235         GMX_ASSERT(atomsStartAt + numAtomsToCopy <= dataSize,
 236                    "The device allocation is smaller than requested copy range.");
 237         GMX_ASSERT(atomsStartAt + numAtomsToCopy <= h_data.ssize(),
 238                    "The host buffer is smaller than the requested copy range.");
 239
 240         copyToDeviceBuffer(&d_data, reinterpret_cast<const RVec*>(&h_data.data()[atomsStartAt]),
 241                            atomsStartAt, numAtomsToCopy, deviceStream, transferKind_, nullptr);
 242     }
 243
 244     wallcycle_sub_stop(wcycle_, ewcsLAUNCH_STATE_PROPAGATOR_DATA);
 245     wallcycle_stop(wcycle_, ewcLAUNCH_GPU);
 246 }
 247
 248 void StatePropagatorDataGpu::Impl::copyFromDevice(gmx::ArrayRef<gmx::RVec> h_data,
 249                                                   DeviceBuffer<RVec>       d_data,
 250                                                   int                      dataSize,
 251                                                   AtomLocality             atomLocality,
 252                                                   const DeviceStream&      deviceStream)
 253 {
 254     GMX_UNUSED_VALUE(dataSize);
 255
 256     GMX_ASSERT(atomLocality < AtomLocality::Count, "Wrong atom locality.");
 257
 258     GMX_ASSERT(dataSize >= 0, "Trying to copy from device buffer before it was allocated.");
 259
 260     GMX_ASSERT(deviceStream.isValid(), "No stream is valid for copying with given atom locality.");
 261     wallcycle_start_nocount(wcycle_, ewcLAUNCH_GPU);
 262     wallcycle_sub_start(wcycle_, ewcsLAUNCH_STATE_PROPAGATOR_DATA);
 263
 264     int atomsStartAt, numAtomsToCopy;
 265     std::tie(atomsStartAt, numAtomsToCopy) = getAtomRangesFromAtomLocality(atomLocality);
 266
 267     if (numAtomsToCopy != 0)
 268     {
 269         GMX_ASSERT(atomsStartAt + numAtomsToCopy <= dataSize,
 270                    "The device allocation is smaller than requested copy range.");
 271         GMX_ASSERT(atomsStartAt + numAtomsToCopy <= h_data.ssize(),
 272                    "The host buffer is smaller than the requested copy range.");
 273
 274         copyFromDeviceBuffer(reinterpret_cast<RVec*>(&h_data.data()[atomsStartAt]), &d_data,
 275                              atomsStartAt, numAtomsToCopy, deviceStream, transferKind_, nullptr);
 276     }
 277
 278     wallcycle_sub_stop(wcycle_, ewcsLAUNCH_STATE_PROPAGATOR_DATA);
 279     wallcycle_stop(wcycle_, ewcLAUNCH_GPU);
 280 }
 281
 282 DeviceBuffer<RVec> StatePropagatorDataGpu::Impl::getCoordinates()
 283 {
 284     return d_x_;
 285 }
 286
 287 void StatePropagatorDataGpu::Impl::copyCoordinatesToGpu(const gmx::ArrayRef<const gmx::RVec> h_x,
 288                                                         AtomLocality atomLocality)
 289 {
 290     GMX_ASSERT(atomLocality < AtomLocality::Count, "Wrong atom locality.");
 291     const DeviceStream* deviceStream = xCopyStreams_[atomLocality];
 292     GMX_ASSERT(deviceStream != nullptr,
 293                "No stream is valid for copying positions with given atom locality.");
 294
 295     wallcycle_start_nocount(wcycle_, ewcLAUNCH_GPU);
 296     wallcycle_sub_start(wcycle_, ewcsLAUNCH_STATE_PROPAGATOR_DATA);
 297
 298     copyToDevice(d_x_, h_x, d_xSize_, atomLocality, *deviceStream);
 299
 300     // markEvent is skipped in OpenCL as:
 301     //   - it's not needed, copy is done in the same stream as the only consumer task (PME)
 302     //   - we don't consume the events in OpenCL which is not allowed by GpuEventSynchronizer (would leak memory).
 303     // TODO: remove this by adding an event-mark free flavor of this function
 304     if (GMX_GPU_CUDA)
 305     {
 306         xReadyOnDevice_[atomLocality].markEvent(*deviceStream);
 307     }
 308
 309     wallcycle_sub_stop(wcycle_, ewcsLAUNCH_STATE_PROPAGATOR_DATA);
 310     wallcycle_stop(wcycle_, ewcLAUNCH_GPU);
 311 }
 312
 313 GpuEventSynchronizer*
 314 StatePropagatorDataGpu::Impl::getCoordinatesReadyOnDeviceEvent(AtomLocality atomLocality,
 315                                                                const SimulationWorkload& simulationWork,
 316                                                                const StepWorkload&       stepWork)
 317 {
 318     // The provider of the coordinates may be different for local atoms. If the update is offloaded
 319     // and this is not a neighbor search step, then the consumer needs to wait for the update
 320     // to complete. Otherwise, the coordinates are copied from the host and we need to wait for
 321     // the copy event. Non-local coordinates are always provided by the H2D copy.
 322     //
 323     // TODO: This should be reconsidered to support the halo exchange.
 324     //
 325     // In OpenCL no events are used as coordinate sync is not necessary
 326     if (GMX_GPU_OPENCL)
 327     {
 328         return nullptr;
 329     }
 330     if (atomLocality == AtomLocality::Local && simulationWork.useGpuUpdate && !stepWork.doNeighborSearch)
 331     {
 332         return &xUpdatedOnDevice_;
 333     }
 334     else
 335     {
 336         return &xReadyOnDevice_[atomLocality];
 337     }
 338 }
 339
 340 void StatePropagatorDataGpu::Impl::waitCoordinatesCopiedToDevice(AtomLocality atomLocality)
 341 {
 342     wallcycle_start(wcycle_, ewcWAIT_GPU_STATE_PROPAGATOR_DATA);
 343     GMX_ASSERT(atomLocality < AtomLocality::Count, "Wrong atom locality.");
 344     xReadyOnDevice_[atomLocality].waitForEvent();
 345     wallcycle_stop(wcycle_, ewcWAIT_GPU_STATE_PROPAGATOR_DATA);
 346 }
 347
 348 GpuEventSynchronizer* StatePropagatorDataGpu::Impl::xUpdatedOnDevice()
 349 {
 350     return &xUpdatedOnDevice_;
 351 }
 352
 353 void StatePropagatorDataGpu::Impl::copyCoordinatesFromGpu(gmx::ArrayRef<gmx::RVec> h_x, AtomLocality atomLocality)
 354 {
 355     GMX_ASSERT(atomLocality < AtomLocality::Count, "Wrong atom locality.");
 356     const DeviceStream* deviceStream = xCopyStreams_[atomLocality];
 357     GMX_ASSERT(deviceStream != nullptr,
 358                "No stream is valid for copying positions with given atom locality.");
 359
 360     wallcycle_start_nocount(wcycle_, ewcLAUNCH_GPU);
 361     wallcycle_sub_start(wcycle_, ewcsLAUNCH_STATE_PROPAGATOR_DATA);
 362
 363     copyFromDevice(h_x, d_x_, d_xSize_, atomLocality, *deviceStream);
 364     // Note: unlike copyCoordinatesToGpu this is not used in OpenCL, and the conditional is not needed.
 365     xReadyOnHost_[atomLocality].markEvent(*deviceStream);
 366
 367     wallcycle_sub_stop(wcycle_, ewcsLAUNCH_STATE_PROPAGATOR_DATA);
 368     wallcycle_stop(wcycle_, ewcLAUNCH_GPU);
 369 }
 370
 371 void StatePropagatorDataGpu::Impl::waitCoordinatesReadyOnHost(AtomLocality atomLocality)
 372 {
 373     wallcycle_start(wcycle_, ewcWAIT_GPU_STATE_PROPAGATOR_DATA);
 374     xReadyOnHost_[atomLocality].waitForEvent();
 375     wallcycle_stop(wcycle_, ewcWAIT_GPU_STATE_PROPAGATOR_DATA);
 376 }
 377
 378
 379 DeviceBuffer<RVec> 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     GMX_ASSERT(atomLocality < AtomLocality::Count, "Wrong atom locality.");
 388     const DeviceStream* deviceStream = vCopyStreams_[atomLocality];
 389     GMX_ASSERT(deviceStream != nullptr,
 390                "No stream is valid for copying velocities with given atom locality.");
 391
 392     wallcycle_start_nocount(wcycle_, ewcLAUNCH_GPU);
 393     wallcycle_sub_start(wcycle_, ewcsLAUNCH_STATE_PROPAGATOR_DATA);
 394
 395     copyToDevice(d_v_, h_v, d_vSize_, atomLocality, *deviceStream);
 396     vReadyOnDevice_[atomLocality].markEvent(*deviceStream);
 397
 398     wallcycle_sub_stop(wcycle_, ewcsLAUNCH_STATE_PROPAGATOR_DATA);
 399     wallcycle_stop(wcycle_, ewcLAUNCH_GPU);
 400 }
 401
 402 GpuEventSynchronizer* StatePropagatorDataGpu::Impl::getVelocitiesReadyOnDeviceEvent(AtomLocality atomLocality)
 403 {
 404     return &vReadyOnDevice_[atomLocality];
 405 }
 406
 407
 408 void StatePropagatorDataGpu::Impl::copyVelocitiesFromGpu(gmx::ArrayRef<gmx::RVec> h_v, AtomLocality atomLocality)
 409 {
 410     GMX_ASSERT(atomLocality < AtomLocality::Count, "Wrong atom locality.");
 411     const DeviceStream* deviceStream = vCopyStreams_[atomLocality];
 412     GMX_ASSERT(deviceStream != nullptr,
 413                "No stream is valid for copying velocities with given atom locality.");
 414
 415     wallcycle_start_nocount(wcycle_, ewcLAUNCH_GPU);
 416     wallcycle_sub_start(wcycle_, ewcsLAUNCH_STATE_PROPAGATOR_DATA);
 417
 418     copyFromDevice(h_v, d_v_, d_vSize_, atomLocality, *deviceStream);
 419     vReadyOnHost_[atomLocality].markEvent(*deviceStream);
 420
 421     wallcycle_sub_stop(wcycle_, ewcsLAUNCH_STATE_PROPAGATOR_DATA);
 422     wallcycle_stop(wcycle_, ewcLAUNCH_GPU);
 423 }
 424
 425 void StatePropagatorDataGpu::Impl::waitVelocitiesReadyOnHost(AtomLocality atomLocality)
 426 {
 427     wallcycle_start(wcycle_, ewcWAIT_GPU_STATE_PROPAGATOR_DATA);
 428     vReadyOnHost_[atomLocality].waitForEvent();
 429     wallcycle_stop(wcycle_, ewcWAIT_GPU_STATE_PROPAGATOR_DATA);
 430 }
 431
 432
 433 DeviceBuffer<RVec> StatePropagatorDataGpu::Impl::getForces()
 434 {
 435     return d_f_;
 436 }
 437
 438 void StatePropagatorDataGpu::Impl::copyForcesToGpu(const gmx::ArrayRef<const gmx::RVec> h_f,
 439                                                    AtomLocality atomLocality)
 440 {
 441     GMX_ASSERT(atomLocality < AtomLocality::Count, "Wrong atom locality.");
 442     const DeviceStream* deviceStream = fCopyStreams_[atomLocality];
 443     GMX_ASSERT(deviceStream != nullptr,
 444                "No stream is valid for copying forces with given atom locality.");
 445
 446     wallcycle_start_nocount(wcycle_, ewcLAUNCH_GPU);
 447     wallcycle_sub_start(wcycle_, ewcsLAUNCH_STATE_PROPAGATOR_DATA);
 448
 449     copyToDevice(d_f_, h_f, d_fSize_, atomLocality, *deviceStream);
 450     fReadyOnDevice_[atomLocality].markEvent(*deviceStream);
 451
 452     wallcycle_sub_stop(wcycle_, ewcsLAUNCH_STATE_PROPAGATOR_DATA);
 453     wallcycle_stop(wcycle_, ewcLAUNCH_GPU);
 454 }
 455
 456 GpuEventSynchronizer* StatePropagatorDataGpu::Impl::getForcesReadyOnDeviceEvent(AtomLocality atomLocality,
 457                                                                                 bool useGpuFBufferOps)
 458 {
 459     if ((atomLocality == AtomLocality::Local || atomLocality == AtomLocality::NonLocal) && useGpuFBufferOps)
 460     {
 461         return &fReducedOnDevice_;
 462     }
 463     else
 464     {
 465         return &fReadyOnDevice_[atomLocality];
 466     }
 467 }
 468
 469 GpuEventSynchronizer* StatePropagatorDataGpu::Impl::fReducedOnDevice()
 470 {
 471     return &fReducedOnDevice_;
 472 }
 473
 474 void StatePropagatorDataGpu::Impl::copyForcesFromGpu(gmx::ArrayRef<gmx::RVec> h_f, AtomLocality atomLocality)
 475 {
 476     GMX_ASSERT(atomLocality < AtomLocality::Count, "Wrong atom locality.");
 477     const DeviceStream* deviceStream = fCopyStreams_[atomLocality];
 478     GMX_ASSERT(deviceStream != nullptr,
 479                "No stream is valid for copying forces with given atom locality.");
 480
 481     wallcycle_start_nocount(wcycle_, ewcLAUNCH_GPU);
 482     wallcycle_sub_start(wcycle_, ewcsLAUNCH_STATE_PROPAGATOR_DATA);
 483
 484     copyFromDevice(h_f, d_f_, d_fSize_, atomLocality, *deviceStream);
 485     fReadyOnHost_[atomLocality].markEvent(*deviceStream);
 486
 487     wallcycle_sub_stop(wcycle_, ewcsLAUNCH_STATE_PROPAGATOR_DATA);
 488     wallcycle_stop(wcycle_, ewcLAUNCH_GPU);
 489 }
 490
 491 void StatePropagatorDataGpu::Impl::waitForcesReadyOnHost(AtomLocality atomLocality)
 492 {
 493     wallcycle_start(wcycle_, ewcWAIT_GPU_STATE_PROPAGATOR_DATA);
 494     fReadyOnHost_[atomLocality].waitForEvent();
 495     wallcycle_stop(wcycle_, ewcWAIT_GPU_STATE_PROPAGATOR_DATA);
 496 }
 497
 498 const DeviceStream* StatePropagatorDataGpu::Impl::getUpdateStream()
 499 {
 500     return updateStream_;
 501 }
 502
 503 int StatePropagatorDataGpu::Impl::numAtomsLocal()
 504 {
 505     return numAtomsLocal_;
 506 }
 507
 508 int StatePropagatorDataGpu::Impl::numAtomsAll()
 509 {
 510     return numAtomsAll_;
 511 }
 512
 513
 514 StatePropagatorDataGpu::StatePropagatorDataGpu(const DeviceStreamManager& deviceStreamManager,
 515                                                GpuApiCallBehavior         transferKind,
 516                                                int            allocationBlockSizeDivisor,
 517                                                gmx_wallcycle* wcycle) :
 518     impl_(new Impl(deviceStreamManager, transferKind, allocationBlockSizeDivisor, wcycle))
 519 {
 520 }
 521
 522 StatePropagatorDataGpu::StatePropagatorDataGpu(const DeviceStream*  pmeStream,
 523                                                const DeviceContext& deviceContext,
 524                                                GpuApiCallBehavior   transferKind,
 525                                                int                  allocationBlockSizeDivisor,
 526                                                gmx_wallcycle*       wcycle) :
 527     impl_(new Impl(pmeStream, deviceContext, transferKind, allocationBlockSizeDivisor, wcycle))
 528 {
 529 }
 530
 531 StatePropagatorDataGpu::StatePropagatorDataGpu(StatePropagatorDataGpu&& /* other */) noexcept = default;
 532
 533 StatePropagatorDataGpu& StatePropagatorDataGpu::operator=(StatePropagatorDataGpu&& /* other */) noexcept = default;
 534
 535 StatePropagatorDataGpu::~StatePropagatorDataGpu() = default;
 536
 537
 538 void StatePropagatorDataGpu::reinit(int numAtomsLocal, int numAtomsAll)
 539 {
 540     return impl_->reinit(numAtomsLocal, numAtomsAll);
 541 }
 542
 543 std::tuple<int, int> StatePropagatorDataGpu::getAtomRangesFromAtomLocality(AtomLocality atomLocality)
 544 {
 545     return impl_->getAtomRangesFromAtomLocality(atomLocality);
 546 }
 547
 548
 549 DeviceBuffer<RVec> StatePropagatorDataGpu::getCoordinates()
 550 {
 551     return impl_->getCoordinates();
 552 }
 553
 554 void StatePropagatorDataGpu::copyCoordinatesToGpu(const gmx::ArrayRef<const gmx::RVec> h_x,
 555                                                   AtomLocality                         atomLocality)
 556 {
 557     return impl_->copyCoordinatesToGpu(h_x, atomLocality);
 558 }
 559
 560 GpuEventSynchronizer*
 561 StatePropagatorDataGpu::getCoordinatesReadyOnDeviceEvent(AtomLocality              atomLocality,
 562                                                          const SimulationWorkload& simulationWork,
 563                                                          const StepWorkload&       stepWork)
 564 {
 565     return impl_->getCoordinatesReadyOnDeviceEvent(atomLocality, simulationWork, stepWork);
 566 }
 567
 568 void StatePropagatorDataGpu::waitCoordinatesCopiedToDevice(AtomLocality atomLocality)
 569 {
 570     return impl_->waitCoordinatesCopiedToDevice(atomLocality);
 571 }
 572
 573 GpuEventSynchronizer* StatePropagatorDataGpu::xUpdatedOnDevice()
 574 {
 575     return impl_->xUpdatedOnDevice();
 576 }
 577
 578 void StatePropagatorDataGpu::copyCoordinatesFromGpu(gmx::ArrayRef<RVec> h_x, AtomLocality atomLocality)
 579 {
 580     return impl_->copyCoordinatesFromGpu(h_x, atomLocality);
 581 }
 582
 583 void StatePropagatorDataGpu::waitCoordinatesReadyOnHost(AtomLocality atomLocality)
 584 {
 585     return impl_->waitCoordinatesReadyOnHost(atomLocality);
 586 }
 587
 588
 589 DeviceBuffer<RVec> StatePropagatorDataGpu::getVelocities()
 590 {
 591     return impl_->getVelocities();
 592 }
 593
 594 void StatePropagatorDataGpu::copyVelocitiesToGpu(const gmx::ArrayRef<const gmx::RVec> h_v,
 595                                                  AtomLocality                         atomLocality)
 596 {
 597     return impl_->copyVelocitiesToGpu(h_v, atomLocality);
 598 }
 599
 600 GpuEventSynchronizer* StatePropagatorDataGpu::getVelocitiesReadyOnDeviceEvent(AtomLocality atomLocality)
 601 {
 602     return impl_->getVelocitiesReadyOnDeviceEvent(atomLocality);
 603 }
 604
 605 void StatePropagatorDataGpu::copyVelocitiesFromGpu(gmx::ArrayRef<RVec> h_v, AtomLocality atomLocality)
 606 {
 607     return impl_->copyVelocitiesFromGpu(h_v, atomLocality);
 608 }
 609
 610 void StatePropagatorDataGpu::waitVelocitiesReadyOnHost(AtomLocality atomLocality)
 611 {
 612     return impl_->waitVelocitiesReadyOnHost(atomLocality);
 613 }
 614
 615
 616 DeviceBuffer<RVec> StatePropagatorDataGpu::getForces()
 617 {
 618     return impl_->getForces();
 619 }
 620
 621 void StatePropagatorDataGpu::copyForcesToGpu(const gmx::ArrayRef<const gmx::RVec> h_f, AtomLocality atomLocality)
 622 {
 623     return impl_->copyForcesToGpu(h_f, atomLocality);
 624 }
 625
 626 GpuEventSynchronizer* StatePropagatorDataGpu::getForcesReadyOnDeviceEvent(AtomLocality atomLocality,
 627                                                                           bool useGpuFBufferOps)
 628 {
 629     return impl_->getForcesReadyOnDeviceEvent(atomLocality, useGpuFBufferOps);
 630 }
 631
 632 GpuEventSynchronizer* StatePropagatorDataGpu::fReducedOnDevice()
 633 {
 634     return impl_->fReducedOnDevice();
 635 }
 636
 637 void StatePropagatorDataGpu::copyForcesFromGpu(gmx::ArrayRef<RVec> h_f, AtomLocality atomLocality)
 638 {
 639     return impl_->copyForcesFromGpu(h_f, atomLocality);
 640 }
 641
 642 void StatePropagatorDataGpu::waitForcesReadyOnHost(AtomLocality atomLocality)
 643 {
 644     return impl_->waitForcesReadyOnHost(atomLocality);
 645 }
 646
 647
 648 const DeviceStream* StatePropagatorDataGpu::getUpdateStream()
 649 {
 650     return impl_->getUpdateStream();
 651 }
 652
 653 int StatePropagatorDataGpu::numAtomsLocal()
 654 {
 655     return impl_->numAtomsLocal();
 656 }
 657
 658 int StatePropagatorDataGpu::numAtomsAll()
 659 {
 660     return impl_->numAtomsAll();
 661 }
 662
 663 } // namespace gmx
 664
 665 #endif // GMX_GPU