src/gromacs/nbnxm/gpu_types_common.h

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2017,2018,2019,2020, 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  * \brief Implements common internal types for different NBNXN GPU implementations
  37  *
  38  * \author Szilárd Páll <pall.szilard@gmail.com>
  39  * \ingroup module_nbnxm
  40  */
  41
  42 #ifndef GMX_MDLIB_NBNXN_GPU_COMMON_TYPES_H
  43 #define GMX_MDLIB_NBNXN_GPU_COMMON_TYPES_H
  44
  45 #include "config.h"
  46
  47 #include "gromacs/mdtypes/locality.h"
  48 #include "gromacs/utility/enumerationhelpers.h"
  49
  50 #include "pairlist.h"
  51
  52 #if GMX_GPU == GMX_GPU_OPENCL
  53 #    include "gromacs/gpu_utils/gpuregiontimer_ocl.h"
  54 #endif
  55
  56 #if GMX_GPU == GMX_GPU_CUDA
  57 #    include "gromacs/gpu_utils/gpuregiontimer.cuh"
  58 #endif
  59
  60 /** \internal
  61  * \brief Parameters required for the GPU nonbonded calculations.
  62  */
  63 struct NBParamGpu
  64 {
  65
  66     //! type of electrostatics, takes values from #eelType
  67     int eeltype;
  68     //! type of VdW impl., takes values from #evdwType
  69     int vdwtype;
  70
  71     //! charge multiplication factor
  72     float epsfac;
  73     //! Reaction-field/plain cutoff electrostatics const.
  74     float c_rf;
  75     //! Reaction-field electrostatics constant
  76     float two_k_rf;
  77     //! Ewald/PME parameter
  78     float ewald_beta;
  79     //! Ewald/PME correction term substracted from the direct-space potential
  80     float sh_ewald;
  81     //! LJ-Ewald/PME correction term added to the correction potential
  82     float sh_lj_ewald;
  83     //! LJ-Ewald/PME coefficient
  84     float ewaldcoeff_lj;
  85
  86     //! Coulomb cut-off squared
  87     float rcoulomb_sq;
  88
  89     //! VdW cut-off squared
  90     float rvdw_sq;
  91     //! VdW switched cut-off
  92     float rvdw_switch;
  93     //! Full, outer pair-list cut-off squared
  94     float rlistOuter_sq;
  95     //! Inner, dynamic pruned pair-list cut-off squared
  96     float rlistInner_sq;
  97     //! True if we use dynamic pair-list pruning
  98     bool useDynamicPruning;
  99
 100     //! VdW shift dispersion constants
 101     shift_consts_t dispersion_shift;
 102     //! VdW shift repulsion constants
 103     shift_consts_t repulsion_shift;
 104     //! VdW switch constants
 105     switch_consts_t vdw_switch;
 106
 107     /* LJ non-bonded parameters - accessed through texture memory */
 108     //! nonbonded parameter table with C6/C12 pairs per atom type-pair, 2*ntype^2 elements
 109     DeviceBuffer<float> nbfp;
 110     //! texture object bound to nbfp
 111     DeviceTexture nbfp_texobj;
 112     //! nonbonded parameter table per atom type, 2*ntype elements
 113     DeviceBuffer<float> nbfp_comb;
 114     //! texture object bound to nbfp_comb
 115     DeviceTexture nbfp_comb_texobj;
 116
 117     /* Ewald Coulomb force table data - accessed through texture memory */
 118     //! table scale/spacing
 119     float coulomb_tab_scale;
 120     //! pointer to the table in the device memory
 121     DeviceBuffer<float> coulomb_tab;
 122     //! texture object bound to coulomb_tab
 123     DeviceTexture coulomb_tab_texobj;
 124 };
 125
 126 namespace Nbnxm
 127 {
 128
 129 using gmx::AtomLocality;
 130 using gmx::InteractionLocality;
 131
 132 /*! \internal
 133  * \brief GPU region timers used for timing GPU kernels and H2D/D2H transfers.
 134  *
 135  * The two-sized arrays hold the local and non-local values and should always
 136  * be indexed with eintLocal/eintNonlocal.
 137  */
 138 struct gpu_timers_t
 139 {
 140     /*! \internal
 141      * \brief Timers for local or non-local coordinate/force transfers
 142      */
 143     struct XFTransfers
 144     {
 145         //! timer for x/q H2D transfers (l/nl, every step)
 146         GpuRegionTimer nb_h2d;
 147         //! timer for f D2H transfer (l/nl, every step)
 148         GpuRegionTimer nb_d2h;
 149     };
 150
 151     /*! \internal
 152      * \brief Timers for local or non-local interaction related operations
 153      */
 154     struct Interaction
 155     {
 156         //! timer for pair-list H2D transfers (l/nl, every PS step)
 157         GpuRegionTimer pl_h2d;
 158         //! true when a pair-list transfer has been done at this step
 159         bool didPairlistH2D = false;
 160         //! timer for non-bonded kernels (l/nl, every step)
 161         GpuRegionTimer nb_k;
 162         //! timer for the 1st pass list pruning kernel (l/nl, every PS step)
 163         GpuRegionTimer prune_k;
 164         //! true when we timed pruning and the timings need to be accounted for
 165         bool didPrune = false;
 166         //! timer for rolling pruning kernels (l/nl, frequency depends on chunk size)
 167         GpuRegionTimer rollingPrune_k;
 168         //! true when we timed rolling pruning (at the previous step) and the timings need to be accounted for
 169         bool didRollingPrune = false;
 170     };
 171
 172     //! timer for atom data transfer (every PS step)
 173     GpuRegionTimer atdat;
 174     //! timers for coordinate/force transfers (every step)
 175     gmx::EnumerationArray<AtomLocality, XFTransfers> xf;
 176     //! timers for interaction related transfers
 177     gmx::EnumerationArray<InteractionLocality, Nbnxm::gpu_timers_t::Interaction> interaction;
 178 };
 179
 180 /*! \internal
 181  * \brief GPU pair list structure */
 182 struct gpu_plist
 183 {
 184     //! number of atoms per cluster
 185     int na_c;
 186
 187     //! size of sci, # of i clusters in the list
 188     int nsci;
 189     //! allocation size of sci
 190     int sci_nalloc;
 191     //! list of i-cluster ("super-clusters")
 192     DeviceBuffer<nbnxn_sci_t> sci;
 193
 194     //! total # of 4*j clusters
 195     int ncj4;
 196     //! allocation size of cj4
 197     int cj4_nalloc;
 198     //! 4*j cluster list, contains j cluster number and index into the i cluster list
 199     DeviceBuffer<nbnxn_cj4_t> cj4;
 200     //! # of 4*j clusters * # of warps
 201     int nimask;
 202     //! allocation size of imask
 203     int imask_nalloc;
 204     //! imask for 2 warps for each 4*j cluster group
 205     DeviceBuffer<unsigned int> imask;
 206     //! atom interaction bits
 207     DeviceBuffer<nbnxn_excl_t> excl;
 208     //! count for excl
 209     int nexcl;
 210     //! allocation size of excl
 211     int excl_nalloc;
 212
 213     /* parameter+variables for normal and rolling pruning */
 214     //! true after search, indictes that initial pruning with outer prunning is needed
 215     bool haveFreshList;
 216     //! the number of parts/steps over which one cyle of roling pruning takes places
 217     int rollingPruningNumParts;
 218     //! the next part to which the roling pruning needs to be applied
 219     int rollingPruningPart;
 220 };
 221
 222 } // namespace Nbnxm
 223
 224 #endif