src/gromacs/mdlib/update_constrain_gpu_impl.cpp

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2019,2020,2021, 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 Implements update and constraints class.
  38  *
  39  * The class combines Leap-Frog integrator with LINCS and SETTLE constraints.
  40  *
  41  * \todo The computational procedures in members should be integrated to improve
  42  *       computational performance.
  43  *
  44  * \author Artem Zhmurov <zhmurov@gmail.com>
  45  *
  46  * \ingroup module_mdlib
  47  */
  48 #include "gmxpre.h"
  49
  50 #include "update_constrain_gpu_impl.h"
  51
  52 #include <assert.h>
  53 #include <stdio.h>
  54
  55 #include <cmath>
  56
  57 #include <algorithm>
  58
  59 #include "gromacs/gpu_utils/device_context.h"
  60 #include "gromacs/gpu_utils/device_stream.h"
  61 #include "gromacs/gpu_utils/devicebuffer.h"
  62 #if GMX_GPU_CUDA
  63 #    include "gromacs/gpu_utils/gpueventsynchronizer.cuh"
  64 #elif GMX_GPU_SYCL
  65 #    include "gromacs/gpu_utils/gpueventsynchronizer_sycl.h"
  66 #endif
  67 #include "gromacs/gpu_utils/gputraits.h"
  68 #include "gromacs/mdlib/leapfrog_gpu.h"
  69 #include "gromacs/mdlib/update_constrain_gpu.h"
  70 #include "gromacs/mdlib/update_constrain_gpu_internal.h"
  71 #include "gromacs/mdtypes/mdatom.h"
  72 #include "gromacs/timing/wallcycle.h"
  73
  74 namespace gmx
  75 {
  76
  77 void UpdateConstrainGpu::Impl::integrate(GpuEventSynchronizer*             fReadyOnDevice,
  78                                          const real                        dt,
  79                                          const bool                        updateVelocities,
  80                                          const bool                        computeVirial,
  81                                          tensor                            virial,
  82                                          const bool                        doTemperatureScaling,
  83                                          gmx::ArrayRef<const t_grp_tcstat> tcstat,
  84                                          const bool                        doParrinelloRahman,
  85                                          const float                       dtPressureCouple,
  86                                          const matrix                      prVelocityScalingMatrix)
  87 {
  88     wallcycle_start_nocount(wcycle_, WallCycleCounter::LaunchGpu);
  89     wallcycle_sub_start(wcycle_, WallCycleSubCounter::LaunchGpuUpdateConstrain);
  90
  91     // Clearing virial matrix
  92     // TODO There is no point in having separate virial matrix for constraints
  93     clear_mat(virial);
  94
  95     // Make sure that the forces are ready on device before proceeding with the update.
  96     fReadyOnDevice->enqueueWaitEvent(deviceStream_);
  97
  98     // The integrate should save a copy of the current coordinates in d_xp_ and write updated
  99     // once into d_x_. The d_xp_ is only needed by constraints.
 100     integrator_->integrate(
 101             d_x_, d_xp_, d_v_, d_f_, dt, doTemperatureScaling, tcstat, doParrinelloRahman, dtPressureCouple, prVelocityScalingMatrix);
 102     // Constraints need both coordinates before (d_x_) and after (d_xp_) update. However, after constraints
 103     // are applied, the d_x_ can be discarded. So we intentionally swap the d_x_ and d_xp_ here to avoid the
 104     // d_xp_ -> d_x_ copy after constraints. Note that the integrate saves them in the wrong order as well.
 105     lincsGpu_->apply(d_xp_, d_x_, updateVelocities, d_v_, 1.0 / dt, computeVirial, virial, pbcAiuc_);
 106     settleGpu_->apply(d_xp_, d_x_, updateVelocities, d_v_, 1.0 / dt, computeVirial, virial, pbcAiuc_);
 107
 108     // scaledVirial -> virial (methods above returns scaled values)
 109     float scaleFactor = 0.5F / (dt * dt);
 110     for (int i = 0; i < DIM; i++)
 111     {
 112         for (int j = 0; j < DIM; j++)
 113         {
 114             virial[i][j] = scaleFactor * virial[i][j];
 115         }
 116     }
 117
 118     coordinatesReady_->markEvent(deviceStream_);
 119
 120     wallcycle_sub_stop(wcycle_, WallCycleSubCounter::LaunchGpuUpdateConstrain);
 121     wallcycle_stop(wcycle_, WallCycleCounter::LaunchGpu);
 122
 123     return;
 124 }
 125
 126 void UpdateConstrainGpu::Impl::scaleCoordinates(const matrix scalingMatrix)
 127 {
 128     wallcycle_start_nocount(wcycle_, WallCycleCounter::LaunchGpu);
 129     wallcycle_sub_start(wcycle_, WallCycleSubCounter::LaunchGpuUpdateConstrain);
 130
 131     ScalingMatrix mu(scalingMatrix);
 132
 133     launchScaleCoordinatesKernel(numAtoms_, d_x_, mu, deviceStream_);
 134
 135     wallcycle_sub_stop(wcycle_, WallCycleSubCounter::LaunchGpuUpdateConstrain);
 136     wallcycle_stop(wcycle_, WallCycleCounter::LaunchGpu);
 137 }
 138
 139 void UpdateConstrainGpu::Impl::scaleVelocities(const matrix scalingMatrix)
 140 {
 141     wallcycle_start_nocount(wcycle_, WallCycleCounter::LaunchGpu);
 142     wallcycle_sub_start(wcycle_, WallCycleSubCounter::LaunchGpuUpdateConstrain);
 143
 144     ScalingMatrix mu(scalingMatrix);
 145
 146     launchScaleCoordinatesKernel(numAtoms_, d_v_, mu, deviceStream_);
 147
 148     wallcycle_sub_stop(wcycle_, WallCycleSubCounter::LaunchGpuUpdateConstrain);
 149     wallcycle_stop(wcycle_, WallCycleCounter::LaunchGpu);
 150 }
 151
 152 UpdateConstrainGpu::Impl::Impl(const t_inputrec&     ir,
 153                                const gmx_mtop_t&     mtop,
 154                                const int             numTempScaleValues,
 155                                const DeviceContext&  deviceContext,
 156                                const DeviceStream&   deviceStream,
 157                                GpuEventSynchronizer* xUpdatedOnDevice,
 158                                gmx_wallcycle*        wcycle) :
 159     deviceContext_(deviceContext),
 160     deviceStream_(deviceStream),
 161     coordinatesReady_(xUpdatedOnDevice),
 162     wcycle_(wcycle)
 163 {
 164     GMX_ASSERT(xUpdatedOnDevice != nullptr, "The event synchronizer can not be nullptr.");
 165
 166     integrator_ = std::make_unique<LeapFrogGpu>(deviceContext_, deviceStream_, numTempScaleValues);
 167     lincsGpu_ = std::make_unique<LincsGpu>(ir.nLincsIter, ir.nProjOrder, deviceContext_, deviceStream_);
 168     settleGpu_ = std::make_unique<SettleGpu>(mtop, deviceContext_, deviceStream_);
 169 }
 170
 171 UpdateConstrainGpu::Impl::~Impl() {}
 172
 173 void UpdateConstrainGpu::Impl::set(DeviceBuffer<Float3>          d_x,
 174                                    DeviceBuffer<Float3>          d_v,
 175                                    const DeviceBuffer<Float3>    d_f,
 176                                    const InteractionDefinitions& idef,
 177                                    const t_mdatoms&              md)
 178 {
 179     wallcycle_start_nocount(wcycle_, WallCycleCounter::LaunchGpu);
 180     wallcycle_sub_start(wcycle_, WallCycleSubCounter::LaunchGpuUpdateConstrain);
 181
 182     GMX_ASSERT(d_x, "Coordinates device buffer should not be null.");
 183     GMX_ASSERT(d_v, "Velocities device buffer should not be null.");
 184     GMX_ASSERT(d_f, "Forces device buffer should not be null.");
 185
 186     d_x_ = d_x;
 187     d_v_ = d_v;
 188     d_f_ = d_f;
 189
 190     numAtoms_ = md.nr;
 191
 192     reallocateDeviceBuffer(&d_xp_, numAtoms_, &numXp_, &numXpAlloc_, deviceContext_);
 193
 194     reallocateDeviceBuffer(
 195             &d_inverseMasses_, numAtoms_, &numInverseMasses_, &numInverseMassesAlloc_, deviceContext_);
 196
 197     // Integrator should also update something, but it does not even have a method yet
 198     integrator_->set(numAtoms_, md.invmass, md.cTC);
 199     lincsGpu_->set(idef, numAtoms_, md.invmass);
 200     settleGpu_->set(idef);
 201
 202     wallcycle_sub_stop(wcycle_, WallCycleSubCounter::LaunchGpuUpdateConstrain);
 203     wallcycle_stop(wcycle_, WallCycleCounter::LaunchGpu);
 204 }
 205
 206 void UpdateConstrainGpu::Impl::setPbc(const PbcType pbcType, const matrix box)
 207 {
 208     // TODO wallcycle
 209     setPbcAiuc(numPbcDimensions(pbcType), box, &pbcAiuc_);
 210 }
 211
 212 GpuEventSynchronizer* UpdateConstrainGpu::Impl::getCoordinatesReadySync()
 213 {
 214     return coordinatesReady_;
 215 }
 216
 217 UpdateConstrainGpu::UpdateConstrainGpu(const t_inputrec&     ir,
 218                                        const gmx_mtop_t&     mtop,
 219                                        const int             numTempScaleValues,
 220                                        const DeviceContext&  deviceContext,
 221                                        const DeviceStream&   deviceStream,
 222                                        GpuEventSynchronizer* xUpdatedOnDevice,
 223                                        gmx_wallcycle*        wcycle) :
 224     impl_(new Impl(ir, mtop, numTempScaleValues, deviceContext, deviceStream, xUpdatedOnDevice, wcycle))
 225 {
 226 }
 227
 228 UpdateConstrainGpu::~UpdateConstrainGpu() = default;
 229
 230 void UpdateConstrainGpu::integrate(GpuEventSynchronizer*             fReadyOnDevice,
 231                                    const real                        dt,
 232                                    const bool                        updateVelocities,
 233                                    const bool                        computeVirial,
 234                                    tensor                            virialScaled,
 235                                    const bool                        doTemperatureScaling,
 236                                    gmx::ArrayRef<const t_grp_tcstat> tcstat,
 237                                    const bool                        doParrinelloRahman,
 238                                    const float                       dtPressureCouple,
 239                                    const matrix                      prVelocityScalingMatrix)
 240 {
 241     impl_->integrate(fReadyOnDevice,
 242                      dt,
 243                      updateVelocities,
 244                      computeVirial,
 245                      virialScaled,
 246                      doTemperatureScaling,
 247                      tcstat,
 248                      doParrinelloRahman,
 249                      dtPressureCouple,
 250                      prVelocityScalingMatrix);
 251 }
 252
 253 void UpdateConstrainGpu::scaleCoordinates(const matrix scalingMatrix)
 254 {
 255     impl_->scaleCoordinates(scalingMatrix);
 256 }
 257
 258 void UpdateConstrainGpu::scaleVelocities(const matrix scalingMatrix)
 259 {
 260     impl_->scaleVelocities(scalingMatrix);
 261 }
 262
 263 void UpdateConstrainGpu::set(DeviceBuffer<Float3>          d_x,
 264                              DeviceBuffer<Float3>          d_v,
 265                              const DeviceBuffer<Float3>    d_f,
 266                              const InteractionDefinitions& idef,
 267                              const t_mdatoms&              md)
 268 {
 269     impl_->set(d_x, d_v, d_f, idef, md);
 270 }
 271
 272 void UpdateConstrainGpu::setPbc(const PbcType pbcType, const matrix box)
 273 {
 274     impl_->setPbc(pbcType, box);
 275 }
 276
 277 GpuEventSynchronizer* UpdateConstrainGpu::getCoordinatesReadySync()
 278 {
 279     return impl_->getCoordinatesReadySync();
 280 }
 281
 282 bool UpdateConstrainGpu::isNumCoupledConstraintsSupported(const gmx_mtop_t& mtop)
 283 {
 284     return LincsGpu::isNumCoupledConstraintsSupported(mtop);
 285 }
 286
 287 } // namespace gmx