2 * This file is part of the GROMACS molecular simulation package.
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.
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.
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.
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.
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.
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.
35 /*! \libinternal \file
37 * \brief Declares the class for GPU implementation of LINCS.
39 * \author Artem Zhmurov <zhmurov@gmail.com>
41 * \ingroup module_mdlib
44 #ifndef GMX_MDLIB_LINCS_GPU_CUH
45 #define GMX_MDLIB_LINCS_GPU_CUH
47 #include "gromacs/gpu_utils/device_context.h"
48 #include "gromacs/gpu_utils/device_stream.h"
49 #include "gromacs/gpu_utils/gputraits.cuh"
50 #include "gromacs/mdlib/constr.h"
51 #include "gromacs/pbcutil/pbc_aiuc.h"
52 #include "gromacs/utility/classhelpers.h"
54 class InteractionDefinitions;
59 /* \brief LINCS parameters and GPU pointers
61 * This is used to accumulate all the parameters and pointers so they can be passed
62 * to the GPU as a single structure.
65 struct LincsGpuKernelParameters
67 //! Periodic boundary data
69 //! Order of expansion when inverting the matrix
71 //! Number of iterations used to correct the projection
73 //! 1/mass for all atoms (GPU)
74 float* d_inverseMasses;
75 //! Scaled virial tensor (6 floats: [XX, XY, XZ, YY, YZ, ZZ], GPU)
76 float* d_virialScaled;
77 /*! \brief Total number of threads.
79 * This covers all constraints and gaps in the ends of the thread blocks
80 * that are necessary to avoid inter-block synchronizations.
81 * Should be a multiple of block size (the last block is filled with dummy to the end).
83 int numConstraintsThreads;
84 //! List of constrained atoms (GPU memory)
86 //! Equilibrium distances for the constraints (GPU)
87 float* d_constraintsTargetLengths;
88 //! Number of constraints, coupled with the current one (GPU)
89 int* d_coupledConstraintsCounts;
90 //! List of coupled with the current one (GPU)
91 int* d_coupledConstraintsIndices;
92 //! Elements of the coupling matrix.
94 //! Mass factors (GPU)
98 /*! \internal \brief Class with interfaces and data for GPU version of LINCS. */
103 /*! \brief Constructor.
105 * \param[in] numIterations Number of iteration for the correction of the projection.
106 * \param[in] expansionOrder Order of the matrix inversion algorithm.
107 * \param[in] deviceContext Device context (dummy in CUDA).
108 * \param[in] deviceStream Device command stream.
110 LincsGpu(int numIterations,
112 const DeviceContext& deviceContext,
113 const DeviceStream& deviceStream);
114 /*! \brief Destructor.*/
117 /*! \brief Apply LINCS.
119 * Applies LINCS to coordinates and velocities, stored on GPU.
120 * The results are not automatically copied back to the CPU memory.
121 * Method uses this class data structures which should be updated
122 * when needed using set() method.
124 * \param[in] d_x Coordinates before timestep (in GPU memory)
125 * \param[in,out] d_xp Coordinates after timestep (in GPU memory). The
126 * resulting constrained coordinates will be saved here.
127 * \param[in] updateVelocities If the velocities should be updated.
128 * \param[in,out] d_v Velocities to update (in GPU memory, can be nullptr
130 * \param[in] invdt Reciprocal timestep (to scale Lagrange
131 * multipliers when velocities are updated)
132 * \param[in] computeVirial If virial should be updated.
133 * \param[in,out] virialScaled Scaled virial tensor to be updated.
134 * \param[in] pbcAiuc PBC data.
136 void apply(const float3* d_x,
138 const bool updateVelocities,
141 const bool computeVirial,
143 const PbcAiuc pbcAiuc);
146 * Update data-structures (e.g. after NB search step).
148 * Updates the constraints data and copies it to the GPU. Should be
149 * called if the particles were sorted, redistributed between domains, etc.
150 * This version uses common data formats so it can be called from anywhere
151 * in the code. Does not recycle the data preparation routines from the CPU
152 * version. Works only with simple case when all the constraints in idef are
153 * are handled by a single GPU. Triangles are not handled as special case.
155 * Information about constraints is taken from:
156 * idef.il[F_CONSTR].iatoms --- type (T) of constraint and two atom indexes (i1, i2)
157 * idef.iparams[T].constr.dA --- target length for constraint of type T
159 * \param[in] idef Local topology data to get information on constraints from.
160 * \param[in] numAtoms Number of atoms.
161 * \param[in] invmass Inverse masses of atoms.
163 void set(const InteractionDefinitions& idef, int numAtoms, const real* invmass);
166 * Returns whether the maximum number of coupled constraints is supported
167 * by the GPU LINCS code.
169 * \param[in] mtop The molecular topology
171 static bool isNumCoupledConstraintsSupported(const gmx_mtop_t& mtop);
174 //! GPU context object
175 const DeviceContext& deviceContext_;
177 const DeviceStream& deviceStream_;
179 //! Parameters and pointers, passed to the GPU kernel
180 LincsGpuKernelParameters kernelParams_;
182 //! Scaled virial tensor (6 floats: [XX, XY, XZ, YY, YZ, ZZ])
183 std::vector<float> h_virialScaled_;
185 /*! \brief Maximum total number of constraints so far.
187 * If the new number of constraints is larger then previous maximum, the GPU data arrays are
190 int numConstraintsThreadsAlloc_;
192 /*! \brief Maximum total number of atoms so far.
194 * If the new number of atoms is larger then previous maximum, the GPU array with masses is
202 #endif // GMX_MDLIB_LINCS_GPU_CUH