src/gromacs/nbnxm/kernels_simd_4xm/kernel_prune.cpp

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2016,2017,2018,2019, 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 #include "gmxpre.h"
  37
  38 #include "kernel_prune.h"
  39
  40 #include "gromacs/nbnxm/atomdata.h"
  41 #include "gromacs/nbnxm/nbnxm_simd.h"
  42 #include "gromacs/nbnxm/pairlist.h"
  43 #include "gromacs/utility/gmxassert.h"
  44
  45 #ifdef GMX_NBNXN_SIMD_4XN
  46 #    define GMX_SIMD_J_UNROLL_SIZE 1
  47 #    include "kernel_common.h"
  48 #endif
  49
  50 /* Prune a single nbnxn_pairtlist_t entry with distance rlistInner */
  51 void nbnxn_kernel_prune_4xn(NbnxnPairlistCpu*       nbl,
  52                             const nbnxn_atomdata_t* nbat,
  53                             const rvec* gmx_restrict shift_vec,
  54                             real                     rlistInner)
  55 {
  56 #ifdef GMX_NBNXN_SIMD_4XN
  57     using namespace gmx;
  58
  59     /* We avoid push_back() for efficiency reasons and resize after filling */
  60     nbl->ci.resize(nbl->ciOuter.size());
  61     nbl->cj.resize(nbl->cjOuter.size());
  62
  63     const nbnxn_ci_t* gmx_restrict ciOuter = nbl->ciOuter.data();
  64     nbnxn_ci_t* gmx_restrict ciInner       = nbl->ci.data();
  65
  66     const nbnxn_cj_t* gmx_restrict cjOuter = nbl->cjOuter.data();
  67     nbnxn_cj_t* gmx_restrict cjInner       = nbl->cj.data();
  68
  69     const real* gmx_restrict shiftvec = shift_vec[0];
  70     const real* gmx_restrict x        = nbat->x().data();
  71
  72     const SimdReal rlist2_S(rlistInner * rlistInner);
  73
  74     /* Initialize the new list count as empty and add pairs that are in range */
  75     int       nciInner = 0;
  76     int       ncjInner = 0;
  77     const int nciOuter = nbl->ciOuter.size();
  78     for (int i = 0; i < nciOuter; i++)
  79     {
  80         const nbnxn_ci_t* gmx_restrict ciEntry = &ciOuter[i];
  81
  82         /* Copy the original list entry to the pruned entry */
  83         ciInner[nciInner].ci           = ciEntry->ci;
  84         ciInner[nciInner].shift        = ciEntry->shift;
  85         ciInner[nciInner].cj_ind_start = ncjInner;
  86
  87         /* Extract shift data */
  88         int ish  = (ciEntry->shift & NBNXN_CI_SHIFT);
  89         int ish3 = ish * 3;
  90         int ci   = ciEntry->ci;
  91
  92         SimdReal shX_S = SimdReal(shiftvec[ish3]);
  93         SimdReal shY_S = SimdReal(shiftvec[ish3 + 1]);
  94         SimdReal shZ_S = SimdReal(shiftvec[ish3 + 2]);
  95
  96 #    if UNROLLJ <= 4
  97         int scix = ci * STRIDE * DIM;
  98 #    else
  99         int scix = (ci >> 1) * STRIDE * DIM + (ci & 1) * (STRIDE >> 1);
 100 #    endif
 101
 102         /* Load i atom data */
 103         int      sciy  = scix + STRIDE;
 104         int      sciz  = sciy + STRIDE;
 105         SimdReal ix_S0 = SimdReal(x[scix]) + shX_S;
 106         SimdReal ix_S1 = SimdReal(x[scix + 1]) + shX_S;
 107         SimdReal ix_S2 = SimdReal(x[scix + 2]) + shX_S;
 108         SimdReal ix_S3 = SimdReal(x[scix + 3]) + shX_S;
 109         SimdReal iy_S0 = SimdReal(x[sciy]) + shY_S;
 110         SimdReal iy_S1 = SimdReal(x[sciy + 1]) + shY_S;
 111         SimdReal iy_S2 = SimdReal(x[sciy + 2]) + shY_S;
 112         SimdReal iy_S3 = SimdReal(x[sciy + 3]) + shY_S;
 113         SimdReal iz_S0 = SimdReal(x[sciz]) + shZ_S;
 114         SimdReal iz_S1 = SimdReal(x[sciz + 1]) + shZ_S;
 115         SimdReal iz_S2 = SimdReal(x[sciz + 2]) + shZ_S;
 116         SimdReal iz_S3 = SimdReal(x[sciz + 3]) + shZ_S;
 117
 118         for (int cjind = ciEntry->cj_ind_start; cjind < ciEntry->cj_ind_end; cjind++)
 119         {
 120             /* j-cluster index */
 121             int cj = cjOuter[cjind].cj;
 122
 123             /* Atom indices (of the first atom in the cluster) */
 124 #    if UNROLLJ == STRIDE
 125             int aj  = cj * UNROLLJ;
 126             int ajx = aj * DIM;
 127 #    else
 128             int ajx = (cj >> 1) * DIM * STRIDE + (cj & 1) * UNROLLJ;
 129 #    endif
 130             int ajy = ajx + STRIDE;
 131             int ajz = ajy + STRIDE;
 132
 133             /* load j atom coordinates */
 134             SimdReal jx_S = load<SimdReal>(x + ajx);
 135             SimdReal jy_S = load<SimdReal>(x + ajy);
 136             SimdReal jz_S = load<SimdReal>(x + ajz);
 137
 138             /* Calculate distance */
 139             SimdReal dx_S0 = ix_S0 - jx_S;
 140             SimdReal dy_S0 = iy_S0 - jy_S;
 141             SimdReal dz_S0 = iz_S0 - jz_S;
 142             SimdReal dx_S1 = ix_S1 - jx_S;
 143             SimdReal dy_S1 = iy_S1 - jy_S;
 144             SimdReal dz_S1 = iz_S1 - jz_S;
 145             SimdReal dx_S2 = ix_S2 - jx_S;
 146             SimdReal dy_S2 = iy_S2 - jy_S;
 147             SimdReal dz_S2 = iz_S2 - jz_S;
 148             SimdReal dx_S3 = ix_S3 - jx_S;
 149             SimdReal dy_S3 = iy_S3 - jy_S;
 150             SimdReal dz_S3 = iz_S3 - jz_S;
 151
 152             /* rsq = dx*dx+dy*dy+dz*dz */
 153             SimdReal rsq_S0 = norm2(dx_S0, dy_S0, dz_S0);
 154             SimdReal rsq_S1 = norm2(dx_S1, dy_S1, dz_S1);
 155             SimdReal rsq_S2 = norm2(dx_S2, dy_S2, dz_S2);
 156             SimdReal rsq_S3 = norm2(dx_S3, dy_S3, dz_S3);
 157
 158             /* Do the cut-off check */
 159             SimdBool wco_S0 = (rsq_S0 < rlist2_S);
 160             SimdBool wco_S1 = (rsq_S1 < rlist2_S);
 161             SimdBool wco_S2 = (rsq_S2 < rlist2_S);
 162             SimdBool wco_S3 = (rsq_S3 < rlist2_S);
 163
 164             wco_S0 = wco_S0 || wco_S1;
 165             wco_S2 = wco_S2 || wco_S3;
 166             wco_S0 = wco_S0 || wco_S2;
 167
 168             /* Putting the assignment inside the conditional is slower */
 169             cjInner[ncjInner] = cjOuter[cjind];
 170             if (anyTrue(wco_S0))
 171             {
 172                 ncjInner++;
 173             }
 174         }
 175
 176         if (ncjInner > ciInner[nciInner].cj_ind_start)
 177         {
 178             ciInner[nciInner].cj_ind_end = ncjInner;
 179             nciInner++;
 180         }
 181     }
 182
 183     nbl->ci.resize(nciInner);
 184     nbl->cj.resize(ncjInner);
 185
 186 #else /* GMX_NBNXN_SIMD_4XN */
 187
 188     GMX_RELEASE_ASSERT(false, "4xN kernel called without 4xN support");
 189
 190     GMX_UNUSED_VALUE(nbl);
 191     GMX_UNUSED_VALUE(nbat);
 192     GMX_UNUSED_VALUE(shift_vec);
 193     GMX_UNUSED_VALUE(rlistInner);
 194
 195 #endif /* GMX_NBNXN_SIMD_4XN */
 196 }