src/gromacs/mdlib/nbnxn_pairlist.h

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014,2015,2016,2017,2018, 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
  36 #ifndef _nbnxn_pairlist_h
  37 #define _nbnxn_pairlist_h
  38
  39 #include "config.h"
  40
  41 #include <cstddef>
  42
  43 #include "gromacs/math/vectypes.h"
  44 #include "gromacs/mdtypes/nblist.h"
  45 #include "gromacs/utility/basedefinitions.h"
  46 #include "gromacs/utility/bitmask.h"
  47 #include "gromacs/utility/real.h"
  48
  49 struct tMPI_Atomic;
  50
  51 /*! \cond INTERNAL */
  52
  53 /*! \brief The setup for generating and pruning the nbnxn pair list.
  54  *
  55  * Without dynamic pruning rlistOuter=rlistInner.
  56  */
  57 struct NbnxnListParameters
  58 {
  59     /*! \brief Constructor producing a struct with dynamic pruning disabled
  60      */
  61     NbnxnListParameters(real rlist) :
  62         useDynamicPruning(false),
  63         nstlistPrune(-1),
  64         rlistOuter(rlist),
  65         rlistInner(rlist),
  66         numRollingParts(1)
  67     {
  68     }
  69
  70     bool useDynamicPruning; //!< Are we using dynamic pair-list pruning
  71     int  nstlistPrune;      //!< Pair-list dynamic pruning interval
  72     real rlistOuter;        //!< Cut-off of the larger, outer pair-list
  73     real rlistInner;        //!< Cut-off of the smaller, inner pair-list
  74     int  numRollingParts;   //!< The number parts to divide the pair-list into for rolling pruning, a value of 1 gives no rolling pruning
  75 };
  76
  77 /*! \endcond */
  78
  79 /* With GPU kernels the i and j cluster size is 8 atoms for CUDA and can be set at compile time for OpenCL */
  80 #if GMX_GPU == GMX_GPU_OPENCL
  81 static constexpr int c_nbnxnGpuClusterSize = GMX_OCL_NB_CLUSTER_SIZE;
  82 #else
  83 static constexpr int c_nbnxnGpuClusterSize = 8;
  84 #endif
  85
  86 /* The number of clusters in a super-cluster, used for GPU */
  87 static constexpr int c_nbnxnGpuNumClusterPerSupercluster = 8;
  88
  89 /* With GPU kernels we group cluster pairs in 4 to optimize memory usage
  90  * of integers containing 32 bits.
  91  */
  92 static constexpr int c_nbnxnGpuJgroupSize = 32/c_nbnxnGpuNumClusterPerSupercluster;
  93
  94 /* In CUDA the number of threads in a warp is 32 and we have cluster pairs
  95  * of 8*8=64 atoms, so it's convenient to store data for cluster pair halves.
  96  */
  97 static constexpr int c_nbnxnGpuClusterpairSplit = 2;
  98
  99 /* The fixed size of the exclusion mask array for a half cluster pair */
 100 static constexpr int c_nbnxnGpuExclSize = c_nbnxnGpuClusterSize*c_nbnxnGpuClusterSize/c_nbnxnGpuClusterpairSplit;
 101
 102 /* A buffer data structure of 64 bytes
 103  * to be placed at the beginning and end of structs
 104  * to avoid cache invalidation of the real contents
 105  * of the struct by writes to neighboring memory.
 106  */
 107 typedef struct {
 108     int dummy[16];
 109 } gmx_cache_protect_t;
 110
 111 /* Abstract type for pair searching data */
 112 typedef struct nbnxn_search * nbnxn_search_t;
 113
 114 /* Function that should return a pointer *ptr to memory
 115  * of size nbytes.
 116  * Error handling should be done within this function.
 117  */
 118 typedef void nbnxn_alloc_t (void **ptr, size_t nbytes);
 119
 120 /* Function that should free the memory pointed to by *ptr.
 121  * NULL should not be passed to this function.
 122  */
 123 typedef void nbnxn_free_t (void *ptr);
 124
 125 /* This is the actual cluster-pair list j-entry.
 126  * cj is the j-cluster.
 127  * The interaction bits in excl are indexed i-major, j-minor.
 128  * The cj entries are sorted such that ones with exclusions come first.
 129  * This means that once a full mask (=NBNXN_INTERACTION_MASK_ALL)
 130  * is found, all subsequent j-entries in the i-entry also have full masks.
 131  */
 132 typedef struct {
 133     int          cj;    /* The j-cluster                    */
 134     unsigned int excl;  /* The exclusion (interaction) bits */
 135 } nbnxn_cj_t;
 136
 137 /* In nbnxn_ci_t the integer shift contains the shift in the lower 7 bits.
 138  * The upper bits contain information for non-bonded kernel optimization.
 139  * Simply calculating LJ and Coulomb for all pairs in a cluster pair is fine.
 140  * But three flags can be used to skip interactions, currently only for subc=0
 141  * !(shift & NBNXN_CI_DO_LJ(subc))   => we can skip LJ for all pairs
 142  * shift & NBNXN_CI_HALF_LJ(subc)    => we can skip LJ for the second half of i
 143  * !(shift & NBNXN_CI_DO_COUL(subc)) => we can skip Coulomb for all pairs
 144  */
 145 #define NBNXN_CI_SHIFT          127
 146 #define NBNXN_CI_DO_LJ(subc)    (1<<(7+3*(subc)))
 147 #define NBNXN_CI_HALF_LJ(subc)  (1<<(8+3*(subc)))
 148 #define NBNXN_CI_DO_COUL(subc)  (1<<(9+3*(subc)))
 149
 150 /* Simple pair-list i-unit */
 151 typedef struct {
 152     int ci;             /* i-cluster             */
 153     int shift;          /* Shift vector index plus possible flags, see above */
 154     int cj_ind_start;   /* Start index into cj   */
 155     int cj_ind_end;     /* End index into cj     */
 156 } nbnxn_ci_t;
 157
 158 /* Grouped pair-list i-unit */
 159 typedef struct {
 160     int sci;            /* i-super-cluster       */
 161     int shift;          /* Shift vector index plus possible flags */
 162     int cj4_ind_start;  /* Start index into cj4  */
 163     int cj4_ind_end;    /* End index into cj4    */
 164 } nbnxn_sci_t;
 165
 166 typedef struct {
 167     unsigned int imask;    /* The i-cluster interactions mask for 1 warp  */
 168     int          excl_ind; /* Index into the exclusion array for 1 warp   */
 169 } nbnxn_im_ei_t;
 170
 171 typedef struct {
 172     int           cj[c_nbnxnGpuJgroupSize];         /* The 4 j-clusters */
 173     nbnxn_im_ei_t imei[c_nbnxnGpuClusterpairSplit]; /* The i-cluster mask data       for 2 warps   */
 174 } nbnxn_cj4_t;
 175
 176 typedef struct {
 177     unsigned int pair[c_nbnxnGpuExclSize]; /* Topology exclusion interaction bits for one warp,
 178                                             * each unsigned has bitS for 4*8 i clusters
 179                                             */
 180 } nbnxn_excl_t;
 181
 182 typedef struct nbnxn_pairlist_t {
 183     gmx_cache_protect_t cp0;
 184
 185     nbnxn_alloc_t      *alloc;
 186     nbnxn_free_t       *free;
 187
 188     gmx_bool            bSimple;         /* Simple list has na_sc=na_s and uses cj   *
 189                                           * Complex list uses cj4                    */
 190
 191     int                     na_ci;       /* The number of atoms per i-cluster        */
 192     int                     na_cj;       /* The number of atoms per j-cluster        */
 193     int                     na_sc;       /* The number of atoms per super cluster    */
 194     real                    rlist;       /* The radius for constructing the list     */
 195     int                     nci;         /* The number of i-clusters in the list     */
 196     int                     nciOuter;    /* The number of i-clusters in the outer, unpruned list, -1 when invalid */
 197     nbnxn_ci_t             *ci;          /* The i-cluster list, size nci             */
 198     nbnxn_ci_t             *ciOuter;     /* The outer, unpruned i-cluster list, size nciOuter(=-1 when invalid) */
 199     int                     ci_nalloc;   /* The allocation size of ci/ciOuter        */
 200     int                     nsci;        /* The number of i-super-clusters in the list */
 201     nbnxn_sci_t            *sci;         /* The i-super-cluster list                 */
 202     int                     sci_nalloc;  /* The allocation size of sci               */
 203
 204     int                     ncj;         /* The number of j-clusters in the list     */
 205     nbnxn_cj_t             *cj;          /* The j-cluster list, size ncj             */
 206     nbnxn_cj_t             *cjOuter;     /* The outer, unpruned j-cluster list, size ncj    */
 207     int                     cj_nalloc;   /* The allocation size of cj/cj0            */
 208     int                     ncjInUse;    /* The number of j-clusters that are used by ci entries in this list, will be <= ncj */
 209
 210     int                     ncj4;        /* The total number of 4*j clusters         */
 211     nbnxn_cj4_t            *cj4;         /* The 4*j cluster list, size ncj4          */
 212     int                     cj4_nalloc;  /* The allocation size of cj4               */
 213     int                     nexcl;       /* The count for excl                       */
 214     nbnxn_excl_t           *excl;        /* Atom interaction bits (non-exclusions)   */
 215     int                     excl_nalloc; /* The allocation size for excl             */
 216     int                     nci_tot;     /* The total number of i clusters           */
 217
 218     struct nbnxn_list_work *work;
 219
 220     gmx_cache_protect_t     cp1;
 221 } nbnxn_pairlist_t;
 222
 223 typedef struct {
 224     int                nnbl;                  /* number of lists */
 225     nbnxn_pairlist_t **nbl;                   /* lists */
 226     nbnxn_pairlist_t **nbl_work;              /* work space for rebalancing lists */
 227     gmx_bool           bCombined;             /* TRUE if lists get combined into one (the 1st) */
 228     gmx_bool           bSimple;               /* TRUE if the list of of type "simple"
 229                                                  (na_sc=na_s, no super-clusters used) */
 230     int                natpair_ljq;           /* Total number of atom pairs for LJ+Q kernel */
 231     int                natpair_lj;            /* Total number of atom pairs for LJ kernel   */
 232     int                natpair_q;             /* Total number of atom pairs for Q kernel    */
 233     t_nblist         **nbl_fep;               /* List of free-energy atom pair interactions */
 234     int64_t            outerListCreationStep; /* Step at which the outer list was created */
 235 } nbnxn_pairlist_set_t;
 236
 237 enum {
 238     nbatXYZ, nbatXYZQ, nbatX4, nbatX8
 239 };
 240
 241 typedef struct {
 242     real *f;      /* f, size natoms*fstride                             */
 243     real *fshift; /* Shift force array, size SHIFTS*DIM                 */
 244     int   nV;     /* The size of *Vvdw and *Vc                          */
 245     real *Vvdw;   /* Temporary Van der Waals group energy storage       */
 246     real *Vc;     /* Temporary Coulomb group energy storage             */
 247     int   nVS;    /* The size of *VSvdw and *VSc                        */
 248     real *VSvdw;  /* Temporary SIMD Van der Waals group energy storage  */
 249     real *VSc;    /* Temporary SIMD Coulomb group energy storage        */
 250 } nbnxn_atomdata_output_t;
 251
 252 /* Block size in atoms for the non-bonded thread force-buffer reduction,
 253  * should be a multiple of all cell and x86 SIMD sizes (i.e. 2, 4 and 8).
 254  * Should be small to reduce the reduction and zeroing cost,
 255  * but too small will result in overhead.
 256  * Currently the block size is NBNXN_BUFFERFLAG_SIZE*3*sizeof(real)=192 bytes.
 257  */
 258 #if GMX_DOUBLE
 259 #define NBNXN_BUFFERFLAG_SIZE   8
 260 #else
 261 #define NBNXN_BUFFERFLAG_SIZE  16
 262 #endif
 263
 264 /* We store the reduction flags as gmx_bitmask_t.
 265  * This limits the number of flags to BITMASK_SIZE.
 266  */
 267 #define NBNXN_BUFFERFLAG_MAX_THREADS  (BITMASK_SIZE)
 268
 269 /* Flags for telling if threads write to force output buffers */
 270 typedef struct {
 271     int               nflag;       /* The number of flag blocks                         */
 272     gmx_bitmask_t    *flag;        /* Bit i is set when thread i writes to a cell-block */
 273     int               flag_nalloc; /* Allocation size of cxy_flag                       */
 274 } nbnxn_buffer_flags_t;
 275
 276 /* LJ combination rules: geometric, Lorentz-Berthelot, none */
 277 enum {
 278     ljcrGEOM, ljcrLB, ljcrNONE, ljcrNR
 279 };
 280
 281 typedef struct nbnxn_atomdata_t { //NOLINT(clang-analyzer-optin.performance.Padding)
 282     nbnxn_alloc_t           *alloc;
 283     nbnxn_free_t            *free;
 284     int                      ntype;           /* The number of different atom types                 */
 285     real                    *nbfp;            /* Lennard-Jones 6*C6 and 12*C12 params, size ntype^2*2 */
 286     int                      comb_rule;       /* Combination rule, see enum above                   */
 287     real                    *nbfp_comb;       /* LJ parameter per atom type, size ntype*2           */
 288     real                    *nbfp_aligned;    /* As nbfp, but with an alignment (stride) suitable
 289                                                * for the present SIMD architectures
 290                                                */
 291     int                      natoms;          /* Number of atoms                                    */
 292     int                      natoms_local;    /* Number of local atoms                           */
 293     int                     *type;            /* Atom types                                         */
 294     real                    *lj_comb;         /* LJ parameters per atom for combining for pairs     */
 295     int                      XFormat;         /* The format of x (and q), enum                      */
 296     int                      FFormat;         /* The format of f, enum                              */
 297     real                    *q;               /* Charges, can be NULL if incorporated in x          */
 298     int                      na_c;            /* The number of atoms per cluster                    */
 299     int                      nenergrp;        /* The number of energy groups                        */
 300     int                      neg_2log;        /* Log2 of nenergrp                                   */
 301     int                     *energrp;         /* The energy groups per cluster, can be NULL         */
 302     gmx_bool                 bDynamicBox;     /* Do we need to update shift_vec every step?    */
 303     rvec                    *shift_vec;       /* Shift vectors, copied from t_forcerec              */
 304     int                      xstride;         /* stride for a coordinate in x (usually 3 or 4)      */
 305     int                      fstride;         /* stride for a coordinate in f (usually 3 or 4)      */
 306     real                    *x;               /* x and possibly q, size natoms*xstride              */
 307
 308     /* j-atom minus i-atom index for generating self and Newton exclusions
 309      * cluster-cluster pairs of the diagonal, for 4xn and 2xnn kernels.
 310      */
 311     real                    *simd_4xn_diagonal_j_minus_i;
 312     real                    *simd_2xnn_diagonal_j_minus_i;
 313     /* Filters for topology exclusion masks for the SIMD kernels. */
 314     uint32_t                *simd_exclusion_filter;
 315     uint64_t                *simd_exclusion_filter64; //!< Used for double w/o SIMD int32 logical support
 316     real                    *simd_interaction_array;  /* Array of masks needed for exclusions */
 317     int                      nout;                    /* The number of force arrays                         */
 318     nbnxn_atomdata_output_t *out;                     /* Output data structures               */
 319     int                      nalloc;                  /* Allocation size of all arrays (for x/f *x/fstride) */
 320     gmx_bool                 bUseBufferFlags;         /* Use the flags or operate on all atoms     */
 321     nbnxn_buffer_flags_t     buffer_flags;            /* Flags for buffer zeroing+reduc.  */
 322     gmx_bool                 bUseTreeReduce;          /* Use tree for force reduction */
 323     tMPI_Atomic             *syncStep;                /* Synchronization step for tree reduce */
 324 } nbnxn_atomdata_t;
 325
 326 #endif