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 #include "gromacs/topology/mtop_util.h"
  74
  75 static constexpr bool sc_haveGpuConstraintSupport = GMX_GPU_CUDA;
  76
  77 namespace gmx
  78 {
  79
  80 void UpdateConstrainGpu::Impl::integrate(GpuEventSynchronizer*             fReadyOnDevice,
  81                                          const real                        dt,
  82                                          const bool                        updateVelocities,
  83                                          const bool                        computeVirial,
  84                                          tensor                            virial,
  85                                          const bool                        doTemperatureScaling,
  86                                          gmx::ArrayRef<const t_grp_tcstat> tcstat,
  87                                          const bool                        doParrinelloRahman,
  88                                          const float                       dtPressureCouple,
  89                                          const matrix                      prVelocityScalingMatrix)
  90 {
  91     wallcycle_start_nocount(wcycle_, WallCycleCounter::LaunchGpu);
  92     wallcycle_sub_start(wcycle_, WallCycleSubCounter::LaunchGpuUpdateConstrain);
  93
  94     // Clearing virial matrix
  95     // TODO There is no point in having separate virial matrix for constraints
  96     clear_mat(virial);
  97
  98     // Make sure that the forces are ready on device before proceeding with the update.
  99     fReadyOnDevice->enqueueWaitEvent(deviceStream_);
 100
 101     // The integrate should save a copy of the current coordinates in d_xp_ and write updated
 102     // once into d_x_. The d_xp_ is only needed by constraints.
 103     integrator_->integrate(
 104             d_x_, d_xp_, d_v_, d_f_, dt, doTemperatureScaling, tcstat, doParrinelloRahman, dtPressureCouple, prVelocityScalingMatrix);
 105     // Constraints need both coordinates before (d_x_) and after (d_xp_) update. However, after constraints
 106     // are applied, the d_x_ can be discarded. So we intentionally swap the d_x_ and d_xp_ here to avoid the
 107     // d_xp_ -> d_x_ copy after constraints. Note that the integrate saves them in the wrong order as well.
 108     if (sc_haveGpuConstraintSupport)
 109     {
 110         lincsGpu_->apply(d_xp_, d_x_, updateVelocities, d_v_, 1.0 / dt, computeVirial, virial, pbcAiuc_);
 111         settleGpu_->apply(d_xp_, d_x_, updateVelocities, d_v_, 1.0 / dt, computeVirial, virial, pbcAiuc_);
 112     }
 113
 114     // scaledVirial -> virial (methods above returns scaled values)
 115     float scaleFactor = 0.5F / (dt * dt);
 116     for (int i = 0; i < DIM; i++)
 117     {
 118         for (int j = 0; j < DIM; j++)
 119         {
 120             virial[i][j] = scaleFactor * virial[i][j];
 121         }
 122     }
 123
 124     coordinatesReady_->markEvent(deviceStream_);
 125
 126     wallcycle_sub_stop(wcycle_, WallCycleSubCounter::LaunchGpuUpdateConstrain);
 127     wallcycle_stop(wcycle_, WallCycleCounter::LaunchGpu);
 128 }
 129
 130 void UpdateConstrainGpu::Impl::scaleCoordinates(const matrix scalingMatrix)
 131 {
 132     wallcycle_start_nocount(wcycle_, WallCycleCounter::LaunchGpu);
 133     wallcycle_sub_start(wcycle_, WallCycleSubCounter::LaunchGpuUpdateConstrain);
 134
 135     ScalingMatrix mu(scalingMatrix);
 136
 137     launchScaleCoordinatesKernel(numAtoms_, d_x_, mu, deviceStream_);
 138
 139     wallcycle_sub_stop(wcycle_, WallCycleSubCounter::LaunchGpuUpdateConstrain);
 140     wallcycle_stop(wcycle_, WallCycleCounter::LaunchGpu);
 141 }
 142
 143 void UpdateConstrainGpu::Impl::scaleVelocities(const matrix scalingMatrix)
 144 {
 145     wallcycle_start_nocount(wcycle_, WallCycleCounter::LaunchGpu);
 146     wallcycle_sub_start(wcycle_, WallCycleSubCounter::LaunchGpuUpdateConstrain);
 147
 148     ScalingMatrix mu(scalingMatrix);
 149
 150     launchScaleCoordinatesKernel(numAtoms_, d_v_, mu, deviceStream_);
 151
 152     wallcycle_sub_stop(wcycle_, WallCycleSubCounter::LaunchGpuUpdateConstrain);
 153     wallcycle_stop(wcycle_, WallCycleCounter::LaunchGpu);
 154 }
 155
 156 UpdateConstrainGpu::Impl::Impl(const t_inputrec&     ir,
 157                                const gmx_mtop_t&     mtop,
 158                                const int             numTempScaleValues,
 159                                const DeviceContext&  deviceContext,
 160                                const DeviceStream&   deviceStream,
 161                                GpuEventSynchronizer* xUpdatedOnDevice,
 162                                gmx_wallcycle*        wcycle) :
 163     deviceContext_(deviceContext),
 164     deviceStream_(deviceStream),
 165     coordinatesReady_(xUpdatedOnDevice),
 166     wcycle_(wcycle)
 167 {
 168     GMX_ASSERT(xUpdatedOnDevice != nullptr, "The event synchronizer can not be nullptr.");
 169
 170     integrator_ = std::make_unique<LeapFrogGpu>(deviceContext_, deviceStream_, numTempScaleValues);
 171     if (sc_haveGpuConstraintSupport)
 172     {
 173         lincsGpu_ = std::make_unique<LincsGpu>(ir.nLincsIter, ir.nProjOrder, deviceContext_, deviceStream_);
 174         settleGpu_ = std::make_unique<SettleGpu>(mtop, deviceContext_, deviceStream_);
 175     }
 176 }
 177
 178 UpdateConstrainGpu::Impl::~Impl() {}
 179
 180 void UpdateConstrainGpu::Impl::set(DeviceBuffer<Float3>          d_x,
 181                                    DeviceBuffer<Float3>          d_v,
 182                                    const DeviceBuffer<Float3>    d_f,
 183                                    const InteractionDefinitions& idef,
 184                                    const t_mdatoms&              md)
 185 {
 186     wallcycle_start_nocount(wcycle_, WallCycleCounter::LaunchGpu);
 187     wallcycle_sub_start(wcycle_, WallCycleSubCounter::LaunchGpuUpdateConstrain);
 188
 189     GMX_ASSERT(d_x, "Coordinates device buffer should not be null.");
 190     GMX_ASSERT(d_v, "Velocities device buffer should not be null.");
 191     GMX_ASSERT(d_f, "Forces device buffer should not be null.");
 192
 193     d_x_ = d_x;
 194     d_v_ = d_v;
 195     d_f_ = d_f;
 196
 197     numAtoms_ = md.nr;
 198
 199     reallocateDeviceBuffer(&d_xp_, numAtoms_, &numXp_, &numXpAlloc_, deviceContext_);
 200
 201     reallocateDeviceBuffer(
 202             &d_inverseMasses_, numAtoms_, &numInverseMasses_, &numInverseMassesAlloc_, deviceContext_);
 203
 204     // Integrator should also update something, but it does not even have a method yet
 205     integrator_->set(numAtoms_, md.invmass, md.cTC);
 206     if (sc_haveGpuConstraintSupport)
 207     {
 208         lincsGpu_->set(idef, numAtoms_, md.invmass);
 209         settleGpu_->set(idef);
 210     }
 211     else
 212     {
 213         GMX_ASSERT(idef.il[F_SETTLE].empty(), "SETTLE not supported");
 214         GMX_ASSERT(idef.il[F_CONSTR].empty(), "LINCS not supported");
 215     }
 216
 217     wallcycle_sub_stop(wcycle_, WallCycleSubCounter::LaunchGpuUpdateConstrain);
 218     wallcycle_stop(wcycle_, WallCycleCounter::LaunchGpu);
 219 }
 220
 221 void UpdateConstrainGpu::Impl::setPbc(const PbcType pbcType, const matrix box)
 222 {
 223     // TODO wallcycle
 224     setPbcAiuc(numPbcDimensions(pbcType), box, &pbcAiuc_);
 225 }
 226
 227 GpuEventSynchronizer* UpdateConstrainGpu::Impl::getCoordinatesReadySync()
 228 {
 229     return coordinatesReady_;
 230 }
 231
 232 UpdateConstrainGpu::UpdateConstrainGpu(const t_inputrec&     ir,
 233                                        const gmx_mtop_t&     mtop,
 234                                        const int             numTempScaleValues,
 235                                        const DeviceContext&  deviceContext,
 236                                        const DeviceStream&   deviceStream,
 237                                        GpuEventSynchronizer* xUpdatedOnDevice,
 238                                        gmx_wallcycle*        wcycle) :
 239     impl_(new Impl(ir, mtop, numTempScaleValues, deviceContext, deviceStream, xUpdatedOnDevice, wcycle))
 240 {
 241 }
 242
 243 UpdateConstrainGpu::~UpdateConstrainGpu() = default;
 244
 245 void UpdateConstrainGpu::integrate(GpuEventSynchronizer*             fReadyOnDevice,
 246                                    const real                        dt,
 247                                    const bool                        updateVelocities,
 248                                    const bool                        computeVirial,
 249                                    tensor                            virialScaled,
 250                                    const bool                        doTemperatureScaling,
 251                                    gmx::ArrayRef<const t_grp_tcstat> tcstat,
 252                                    const bool                        doParrinelloRahman,
 253                                    const float                       dtPressureCouple,
 254                                    const matrix                      prVelocityScalingMatrix)
 255 {
 256     impl_->integrate(fReadyOnDevice,
 257                      dt,
 258                      updateVelocities,
 259                      computeVirial,
 260                      virialScaled,
 261                      doTemperatureScaling,
 262                      tcstat,
 263                      doParrinelloRahman,
 264                      dtPressureCouple,
 265                      prVelocityScalingMatrix);
 266 }
 267
 268 void UpdateConstrainGpu::scaleCoordinates(const matrix scalingMatrix)
 269 {
 270     impl_->scaleCoordinates(scalingMatrix);
 271 }
 272
 273 void UpdateConstrainGpu::scaleVelocities(const matrix scalingMatrix)
 274 {
 275     impl_->scaleVelocities(scalingMatrix);
 276 }
 277
 278 void UpdateConstrainGpu::set(DeviceBuffer<Float3>          d_x,
 279                              DeviceBuffer<Float3>          d_v,
 280                              const DeviceBuffer<Float3>    d_f,
 281                              const InteractionDefinitions& idef,
 282                              const t_mdatoms&              md)
 283 {
 284     impl_->set(d_x, d_v, d_f, idef, md);
 285 }
 286
 287 void UpdateConstrainGpu::setPbc(const PbcType pbcType, const matrix box)
 288 {
 289     impl_->setPbc(pbcType, box);
 290 }
 291
 292 GpuEventSynchronizer* UpdateConstrainGpu::getCoordinatesReadySync()
 293 {
 294     return impl_->getCoordinatesReadySync();
 295 }
 296
 297 bool UpdateConstrainGpu::isNumCoupledConstraintsSupported(const gmx_mtop_t& mtop)
 298 {
 299     return LincsGpu::isNumCoupledConstraintsSupported(mtop);
 300 }
 301
 302 bool UpdateConstrainGpu::areConstraintsSupported()
 303 {
 304     return sc_haveGpuConstraintSupport;
 305 }
 306
 307 } // namespace gmx