src/gmxlib/nonbonded/nb_kernel_avx_256_double/nb_kernel_ElecEwSw_VdwLJSw_GeomW3P1_avx_256_double.c

   1 /*
   2  * Note: this file was generated by the Gromacs avx_256_double kernel generator.
   3  *
   4  *                This source code is part of
   5  *
   6  *                 G   R   O   M   A   C   S
   7  *
   8  * Copyright (c) 2001-2012, The GROMACS Development Team
   9  *
  10  * Gromacs is a library for molecular simulation and trajectory analysis,
  11  * written by Erik Lindahl, David van der Spoel, Berk Hess, and others - for
  12  * a full list of developers and information, check out http://www.gromacs.org
  13  *
  14  * This program is free software; you can redistribute it and/or modify it under
  15  * the terms of the GNU Lesser General Public License as published by the Free
  16  * Software Foundation; either version 2 of the License, or (at your option) any
  17  * later version.
  18  *
  19  * To help fund GROMACS development, we humbly ask that you cite
  20  * the papers people have written on it - you can find them on the website.
  21  */
  22 #ifdef HAVE_CONFIG_H
  23 #include <config.h>
  24 #endif
  25
  26 #include <math.h>
  27
  28 #include "../nb_kernel.h"
  29 #include "types/simple.h"
  30 #include "vec.h"
  31 #include "nrnb.h"
  32
  33 #include "gmx_math_x86_avx_256_double.h"
  34 #include "kernelutil_x86_avx_256_double.h"
  35
  36 /*
  37  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSw_VdwLJSw_GeomW3P1_VF_avx_256_double
  38  * Electrostatics interaction: Ewald
  39  * VdW interaction:            LennardJones
  40  * Geometry:                   Water3-Particle
  41  * Calculate force/pot:        PotentialAndForce
  42  */
  43 void
  44 nb_kernel_ElecEwSw_VdwLJSw_GeomW3P1_VF_avx_256_double
  45                     (t_nblist * gmx_restrict                nlist,
  46                      rvec * gmx_restrict                    xx,
  47                      rvec * gmx_restrict                    ff,
  48                      t_forcerec * gmx_restrict              fr,
  49                      t_mdatoms * gmx_restrict               mdatoms,
  50                      nb_kernel_data_t * gmx_restrict        kernel_data,
  51                      t_nrnb * gmx_restrict                  nrnb)
  52 {
  53     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  54      * just 0 for non-waters.
  55      * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
  56      * jnr indices corresponding to data put in the four positions in the SIMD register.
  57      */
  58     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  59     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  60     int              jnrA,jnrB,jnrC,jnrD;
  61     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  62     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
  63     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  64     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  65     real             rcutoff_scalar;
  66     real             *shiftvec,*fshift,*x,*f;
  67     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
  68     real             scratch[4*DIM];
  69     __m256d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  70     real *           vdwioffsetptr0;
  71     __m256d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  72     real *           vdwioffsetptr1;
  73     __m256d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  74     real *           vdwioffsetptr2;
  75     __m256d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  76     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
  77     __m256d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  78     __m256d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  79     __m256d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  80     __m256d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  81     __m256d          velec,felec,velecsum,facel,crf,krf,krf2;
  82     real             *charge;
  83     int              nvdwtype;
  84     __m256d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  85     int              *vdwtype;
  86     real             *vdwparam;
  87     __m256d          one_sixth   = _mm256_set1_pd(1.0/6.0);
  88     __m256d          one_twelfth = _mm256_set1_pd(1.0/12.0);
  89     __m128i          ewitab;
  90     __m256d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
  91     __m256d          beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
  92     real             *ewtab;
  93     __m256d          rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
  94     real             rswitch_scalar,d_scalar;
  95     __m256d          dummy_mask,cutoff_mask;
  96     __m128           tmpmask0,tmpmask1;
  97     __m256d          signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
  98     __m256d          one     = _mm256_set1_pd(1.0);
  99     __m256d          two     = _mm256_set1_pd(2.0);
 100     x                = xx[0];
 101     f                = ff[0];
 102
 103     nri              = nlist->nri;
 104     iinr             = nlist->iinr;
 105     jindex           = nlist->jindex;
 106     jjnr             = nlist->jjnr;
 107     shiftidx         = nlist->shift;
 108     gid              = nlist->gid;
 109     shiftvec         = fr->shift_vec[0];
 110     fshift           = fr->fshift[0];
 111     facel            = _mm256_set1_pd(fr->epsfac);
 112     charge           = mdatoms->chargeA;
 113     nvdwtype         = fr->ntype;
 114     vdwparam         = fr->nbfp;
 115     vdwtype          = mdatoms->typeA;
 116
 117     sh_ewald         = _mm256_set1_pd(fr->ic->sh_ewald);
 118     beta             = _mm256_set1_pd(fr->ic->ewaldcoeff);
 119     beta2            = _mm256_mul_pd(beta,beta);
 120     beta3            = _mm256_mul_pd(beta,beta2);
 121
 122     ewtab            = fr->ic->tabq_coul_FDV0;
 123     ewtabscale       = _mm256_set1_pd(fr->ic->tabq_scale);
 124     ewtabhalfspace   = _mm256_set1_pd(0.5/fr->ic->tabq_scale);
 125
 126     /* Setup water-specific parameters */
 127     inr              = nlist->iinr[0];
 128     iq0              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+0]));
 129     iq1              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
 130     iq2              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
 131     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
 132
 133     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 134     rcutoff_scalar   = fr->rcoulomb;
 135     rcutoff          = _mm256_set1_pd(rcutoff_scalar);
 136     rcutoff2         = _mm256_mul_pd(rcutoff,rcutoff);
 137
 138     rswitch_scalar   = fr->rcoulomb_switch;
 139     rswitch          = _mm256_set1_pd(rswitch_scalar);
 140     /* Setup switch parameters */
 141     d_scalar         = rcutoff_scalar-rswitch_scalar;
 142     d                = _mm256_set1_pd(d_scalar);
 143     swV3             = _mm256_set1_pd(-10.0/(d_scalar*d_scalar*d_scalar));
 144     swV4             = _mm256_set1_pd( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 145     swV5             = _mm256_set1_pd( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 146     swF2             = _mm256_set1_pd(-30.0/(d_scalar*d_scalar*d_scalar));
 147     swF3             = _mm256_set1_pd( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 148     swF4             = _mm256_set1_pd(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 149
 150     /* Avoid stupid compiler warnings */
 151     jnrA = jnrB = jnrC = jnrD = 0;
 152     j_coord_offsetA = 0;
 153     j_coord_offsetB = 0;
 154     j_coord_offsetC = 0;
 155     j_coord_offsetD = 0;
 156
 157     outeriter        = 0;
 158     inneriter        = 0;
 159
 160     for(iidx=0;iidx<4*DIM;iidx++)
 161     {
 162         scratch[iidx] = 0.0;
 163     }
 164
 165     /* Start outer loop over neighborlists */
 166     for(iidx=0; iidx<nri; iidx++)
 167     {
 168         /* Load shift vector for this list */
 169         i_shift_offset   = DIM*shiftidx[iidx];
 170
 171         /* Load limits for loop over neighbors */
 172         j_index_start    = jindex[iidx];
 173         j_index_end      = jindex[iidx+1];
 174
 175         /* Get outer coordinate index */
 176         inr              = iinr[iidx];
 177         i_coord_offset   = DIM*inr;
 178
 179         /* Load i particle coords and add shift vector */
 180         gmx_mm256_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 181                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 182
 183         fix0             = _mm256_setzero_pd();
 184         fiy0             = _mm256_setzero_pd();
 185         fiz0             = _mm256_setzero_pd();
 186         fix1             = _mm256_setzero_pd();
 187         fiy1             = _mm256_setzero_pd();
 188         fiz1             = _mm256_setzero_pd();
 189         fix2             = _mm256_setzero_pd();
 190         fiy2             = _mm256_setzero_pd();
 191         fiz2             = _mm256_setzero_pd();
 192
 193         /* Reset potential sums */
 194         velecsum         = _mm256_setzero_pd();
 195         vvdwsum          = _mm256_setzero_pd();
 196
 197         /* Start inner kernel loop */
 198         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 199         {
 200
 201             /* Get j neighbor index, and coordinate index */
 202             jnrA             = jjnr[jidx];
 203             jnrB             = jjnr[jidx+1];
 204             jnrC             = jjnr[jidx+2];
 205             jnrD             = jjnr[jidx+3];
 206             j_coord_offsetA  = DIM*jnrA;
 207             j_coord_offsetB  = DIM*jnrB;
 208             j_coord_offsetC  = DIM*jnrC;
 209             j_coord_offsetD  = DIM*jnrD;
 210
 211             /* load j atom coordinates */
 212             gmx_mm256_load_1rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 213                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 214                                                  &jx0,&jy0,&jz0);
 215
 216             /* Calculate displacement vector */
 217             dx00             = _mm256_sub_pd(ix0,jx0);
 218             dy00             = _mm256_sub_pd(iy0,jy0);
 219             dz00             = _mm256_sub_pd(iz0,jz0);
 220             dx10             = _mm256_sub_pd(ix1,jx0);
 221             dy10             = _mm256_sub_pd(iy1,jy0);
 222             dz10             = _mm256_sub_pd(iz1,jz0);
 223             dx20             = _mm256_sub_pd(ix2,jx0);
 224             dy20             = _mm256_sub_pd(iy2,jy0);
 225             dz20             = _mm256_sub_pd(iz2,jz0);
 226
 227             /* Calculate squared distance and things based on it */
 228             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 229             rsq10            = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
 230             rsq20            = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
 231
 232             rinv00           = gmx_mm256_invsqrt_pd(rsq00);
 233             rinv10           = gmx_mm256_invsqrt_pd(rsq10);
 234             rinv20           = gmx_mm256_invsqrt_pd(rsq20);
 235
 236             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
 237             rinvsq10         = _mm256_mul_pd(rinv10,rinv10);
 238             rinvsq20         = _mm256_mul_pd(rinv20,rinv20);
 239
 240             /* Load parameters for j particles */
 241             jq0              = gmx_mm256_load_4real_swizzle_pd(charge+jnrA+0,charge+jnrB+0,
 242                                                                  charge+jnrC+0,charge+jnrD+0);
 243             vdwjidx0A        = 2*vdwtype[jnrA+0];
 244             vdwjidx0B        = 2*vdwtype[jnrB+0];
 245             vdwjidx0C        = 2*vdwtype[jnrC+0];
 246             vdwjidx0D        = 2*vdwtype[jnrD+0];
 247
 248             fjx0             = _mm256_setzero_pd();
 249             fjy0             = _mm256_setzero_pd();
 250             fjz0             = _mm256_setzero_pd();
 251
 252             /**************************
 253              * CALCULATE INTERACTIONS *
 254              **************************/
 255
 256             if (gmx_mm256_any_lt(rsq00,rcutoff2))
 257             {
 258
 259             r00              = _mm256_mul_pd(rsq00,rinv00);
 260
 261             /* Compute parameters for interactions between i and j atoms */
 262             qq00             = _mm256_mul_pd(iq0,jq0);
 263             gmx_mm256_load_4pair_swizzle_pd(vdwioffsetptr0+vdwjidx0A,
 264                                             vdwioffsetptr0+vdwjidx0B,
 265                                             vdwioffsetptr0+vdwjidx0C,
 266                                             vdwioffsetptr0+vdwjidx0D,
 267                                             &c6_00,&c12_00);
 268
 269             /* EWALD ELECTROSTATICS */
 270
 271             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 272             ewrt             = _mm256_mul_pd(r00,ewtabscale);
 273             ewitab           = _mm256_cvttpd_epi32(ewrt);
 274             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 275             ewitab           = _mm_slli_epi32(ewitab,2);
 276             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 277             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 278             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 279             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 280             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 281             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 282             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 283             velec            = _mm256_mul_pd(qq00,_mm256_sub_pd(rinv00,velec));
 284             felec            = _mm256_mul_pd(_mm256_mul_pd(qq00,rinv00),_mm256_sub_pd(rinvsq00,felec));
 285
 286             /* LENNARD-JONES DISPERSION/REPULSION */
 287
 288             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 289             vvdw6            = _mm256_mul_pd(c6_00,rinvsix);
 290             vvdw12           = _mm256_mul_pd(c12_00,_mm256_mul_pd(rinvsix,rinvsix));
 291             vvdw             = _mm256_sub_pd( _mm256_mul_pd(vvdw12,one_twelfth) , _mm256_mul_pd(vvdw6,one_sixth) );
 292             fvdw             = _mm256_mul_pd(_mm256_sub_pd(vvdw12,vvdw6),rinvsq00);
 293
 294             d                = _mm256_sub_pd(r00,rswitch);
 295             d                = _mm256_max_pd(d,_mm256_setzero_pd());
 296             d2               = _mm256_mul_pd(d,d);
 297             sw               = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
 298
 299             dsw              = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
 300
 301             /* Evaluate switch function */
 302             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 303             felec            = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv00,_mm256_mul_pd(velec,dsw)) );
 304             fvdw             = _mm256_sub_pd( _mm256_mul_pd(fvdw,sw) , _mm256_mul_pd(rinv00,_mm256_mul_pd(vvdw,dsw)) );
 305             velec            = _mm256_mul_pd(velec,sw);
 306             vvdw             = _mm256_mul_pd(vvdw,sw);
 307             cutoff_mask      = _mm256_cmp_pd(rsq00,rcutoff2,_CMP_LT_OQ);
 308
 309             /* Update potential sum for this i atom from the interaction with this j atom. */
 310             velec            = _mm256_and_pd(velec,cutoff_mask);
 311             velecsum         = _mm256_add_pd(velecsum,velec);
 312             vvdw             = _mm256_and_pd(vvdw,cutoff_mask);
 313             vvdwsum          = _mm256_add_pd(vvdwsum,vvdw);
 314
 315             fscal            = _mm256_add_pd(felec,fvdw);
 316
 317             fscal            = _mm256_and_pd(fscal,cutoff_mask);
 318
 319             /* Calculate temporary vectorial force */
 320             tx               = _mm256_mul_pd(fscal,dx00);
 321             ty               = _mm256_mul_pd(fscal,dy00);
 322             tz               = _mm256_mul_pd(fscal,dz00);
 323
 324             /* Update vectorial force */
 325             fix0             = _mm256_add_pd(fix0,tx);
 326             fiy0             = _mm256_add_pd(fiy0,ty);
 327             fiz0             = _mm256_add_pd(fiz0,tz);
 328
 329             fjx0             = _mm256_add_pd(fjx0,tx);
 330             fjy0             = _mm256_add_pd(fjy0,ty);
 331             fjz0             = _mm256_add_pd(fjz0,tz);
 332
 333             }
 334
 335             /**************************
 336              * CALCULATE INTERACTIONS *
 337              **************************/
 338
 339             if (gmx_mm256_any_lt(rsq10,rcutoff2))
 340             {
 341
 342             r10              = _mm256_mul_pd(rsq10,rinv10);
 343
 344             /* Compute parameters for interactions between i and j atoms */
 345             qq10             = _mm256_mul_pd(iq1,jq0);
 346
 347             /* EWALD ELECTROSTATICS */
 348
 349             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 350             ewrt             = _mm256_mul_pd(r10,ewtabscale);
 351             ewitab           = _mm256_cvttpd_epi32(ewrt);
 352             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 353             ewitab           = _mm_slli_epi32(ewitab,2);
 354             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 355             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 356             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 357             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 358             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 359             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 360             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 361             velec            = _mm256_mul_pd(qq10,_mm256_sub_pd(rinv10,velec));
 362             felec            = _mm256_mul_pd(_mm256_mul_pd(qq10,rinv10),_mm256_sub_pd(rinvsq10,felec));
 363
 364             d                = _mm256_sub_pd(r10,rswitch);
 365             d                = _mm256_max_pd(d,_mm256_setzero_pd());
 366             d2               = _mm256_mul_pd(d,d);
 367             sw               = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
 368
 369             dsw              = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
 370
 371             /* Evaluate switch function */
 372             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 373             felec            = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv10,_mm256_mul_pd(velec,dsw)) );
 374             velec            = _mm256_mul_pd(velec,sw);
 375             cutoff_mask      = _mm256_cmp_pd(rsq10,rcutoff2,_CMP_LT_OQ);
 376
 377             /* Update potential sum for this i atom from the interaction with this j atom. */
 378             velec            = _mm256_and_pd(velec,cutoff_mask);
 379             velecsum         = _mm256_add_pd(velecsum,velec);
 380
 381             fscal            = felec;
 382
 383             fscal            = _mm256_and_pd(fscal,cutoff_mask);
 384
 385             /* Calculate temporary vectorial force */
 386             tx               = _mm256_mul_pd(fscal,dx10);
 387             ty               = _mm256_mul_pd(fscal,dy10);
 388             tz               = _mm256_mul_pd(fscal,dz10);
 389
 390             /* Update vectorial force */
 391             fix1             = _mm256_add_pd(fix1,tx);
 392             fiy1             = _mm256_add_pd(fiy1,ty);
 393             fiz1             = _mm256_add_pd(fiz1,tz);
 394
 395             fjx0             = _mm256_add_pd(fjx0,tx);
 396             fjy0             = _mm256_add_pd(fjy0,ty);
 397             fjz0             = _mm256_add_pd(fjz0,tz);
 398
 399             }
 400
 401             /**************************
 402              * CALCULATE INTERACTIONS *
 403              **************************/
 404
 405             if (gmx_mm256_any_lt(rsq20,rcutoff2))
 406             {
 407
 408             r20              = _mm256_mul_pd(rsq20,rinv20);
 409
 410             /* Compute parameters for interactions between i and j atoms */
 411             qq20             = _mm256_mul_pd(iq2,jq0);
 412
 413             /* EWALD ELECTROSTATICS */
 414
 415             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 416             ewrt             = _mm256_mul_pd(r20,ewtabscale);
 417             ewitab           = _mm256_cvttpd_epi32(ewrt);
 418             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 419             ewitab           = _mm_slli_epi32(ewitab,2);
 420             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 421             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 422             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 423             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 424             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 425             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 426             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 427             velec            = _mm256_mul_pd(qq20,_mm256_sub_pd(rinv20,velec));
 428             felec            = _mm256_mul_pd(_mm256_mul_pd(qq20,rinv20),_mm256_sub_pd(rinvsq20,felec));
 429
 430             d                = _mm256_sub_pd(r20,rswitch);
 431             d                = _mm256_max_pd(d,_mm256_setzero_pd());
 432             d2               = _mm256_mul_pd(d,d);
 433             sw               = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
 434
 435             dsw              = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
 436
 437             /* Evaluate switch function */
 438             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 439             felec            = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv20,_mm256_mul_pd(velec,dsw)) );
 440             velec            = _mm256_mul_pd(velec,sw);
 441             cutoff_mask      = _mm256_cmp_pd(rsq20,rcutoff2,_CMP_LT_OQ);
 442
 443             /* Update potential sum for this i atom from the interaction with this j atom. */
 444             velec            = _mm256_and_pd(velec,cutoff_mask);
 445             velecsum         = _mm256_add_pd(velecsum,velec);
 446
 447             fscal            = felec;
 448
 449             fscal            = _mm256_and_pd(fscal,cutoff_mask);
 450
 451             /* Calculate temporary vectorial force */
 452             tx               = _mm256_mul_pd(fscal,dx20);
 453             ty               = _mm256_mul_pd(fscal,dy20);
 454             tz               = _mm256_mul_pd(fscal,dz20);
 455
 456             /* Update vectorial force */
 457             fix2             = _mm256_add_pd(fix2,tx);
 458             fiy2             = _mm256_add_pd(fiy2,ty);
 459             fiz2             = _mm256_add_pd(fiz2,tz);
 460
 461             fjx0             = _mm256_add_pd(fjx0,tx);
 462             fjy0             = _mm256_add_pd(fjy0,ty);
 463             fjz0             = _mm256_add_pd(fjz0,tz);
 464
 465             }
 466
 467             fjptrA             = f+j_coord_offsetA;
 468             fjptrB             = f+j_coord_offsetB;
 469             fjptrC             = f+j_coord_offsetC;
 470             fjptrD             = f+j_coord_offsetD;
 471
 472             gmx_mm256_decrement_1rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,fjx0,fjy0,fjz0);
 473
 474             /* Inner loop uses 216 flops */
 475         }
 476
 477         if(jidx<j_index_end)
 478         {
 479
 480             /* Get j neighbor index, and coordinate index */
 481             jnrlistA         = jjnr[jidx];
 482             jnrlistB         = jjnr[jidx+1];
 483             jnrlistC         = jjnr[jidx+2];
 484             jnrlistD         = jjnr[jidx+3];
 485             /* Sign of each element will be negative for non-real atoms.
 486              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 487              * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
 488              */
 489             tmpmask0 = gmx_mm_castsi128_pd(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 490
 491             tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
 492             tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
 493             dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
 494
 495             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 496             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 497             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 498             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 499             j_coord_offsetA  = DIM*jnrA;
 500             j_coord_offsetB  = DIM*jnrB;
 501             j_coord_offsetC  = DIM*jnrC;
 502             j_coord_offsetD  = DIM*jnrD;
 503
 504             /* load j atom coordinates */
 505             gmx_mm256_load_1rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 506                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 507                                                  &jx0,&jy0,&jz0);
 508
 509             /* Calculate displacement vector */
 510             dx00             = _mm256_sub_pd(ix0,jx0);
 511             dy00             = _mm256_sub_pd(iy0,jy0);
 512             dz00             = _mm256_sub_pd(iz0,jz0);
 513             dx10             = _mm256_sub_pd(ix1,jx0);
 514             dy10             = _mm256_sub_pd(iy1,jy0);
 515             dz10             = _mm256_sub_pd(iz1,jz0);
 516             dx20             = _mm256_sub_pd(ix2,jx0);
 517             dy20             = _mm256_sub_pd(iy2,jy0);
 518             dz20             = _mm256_sub_pd(iz2,jz0);
 519
 520             /* Calculate squared distance and things based on it */
 521             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 522             rsq10            = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
 523             rsq20            = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
 524
 525             rinv00           = gmx_mm256_invsqrt_pd(rsq00);
 526             rinv10           = gmx_mm256_invsqrt_pd(rsq10);
 527             rinv20           = gmx_mm256_invsqrt_pd(rsq20);
 528
 529             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
 530             rinvsq10         = _mm256_mul_pd(rinv10,rinv10);
 531             rinvsq20         = _mm256_mul_pd(rinv20,rinv20);
 532
 533             /* Load parameters for j particles */
 534             jq0              = gmx_mm256_load_4real_swizzle_pd(charge+jnrA+0,charge+jnrB+0,
 535                                                                  charge+jnrC+0,charge+jnrD+0);
 536             vdwjidx0A        = 2*vdwtype[jnrA+0];
 537             vdwjidx0B        = 2*vdwtype[jnrB+0];
 538             vdwjidx0C        = 2*vdwtype[jnrC+0];
 539             vdwjidx0D        = 2*vdwtype[jnrD+0];
 540
 541             fjx0             = _mm256_setzero_pd();
 542             fjy0             = _mm256_setzero_pd();
 543             fjz0             = _mm256_setzero_pd();
 544
 545             /**************************
 546              * CALCULATE INTERACTIONS *
 547              **************************/
 548
 549             if (gmx_mm256_any_lt(rsq00,rcutoff2))
 550             {
 551
 552             r00              = _mm256_mul_pd(rsq00,rinv00);
 553             r00              = _mm256_andnot_pd(dummy_mask,r00);
 554
 555             /* Compute parameters for interactions between i and j atoms */
 556             qq00             = _mm256_mul_pd(iq0,jq0);
 557             gmx_mm256_load_4pair_swizzle_pd(vdwioffsetptr0+vdwjidx0A,
 558                                             vdwioffsetptr0+vdwjidx0B,
 559                                             vdwioffsetptr0+vdwjidx0C,
 560                                             vdwioffsetptr0+vdwjidx0D,
 561                                             &c6_00,&c12_00);
 562
 563             /* EWALD ELECTROSTATICS */
 564
 565             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 566             ewrt             = _mm256_mul_pd(r00,ewtabscale);
 567             ewitab           = _mm256_cvttpd_epi32(ewrt);
 568             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 569             ewitab           = _mm_slli_epi32(ewitab,2);
 570             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 571             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 572             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 573             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 574             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 575             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 576             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 577             velec            = _mm256_mul_pd(qq00,_mm256_sub_pd(rinv00,velec));
 578             felec            = _mm256_mul_pd(_mm256_mul_pd(qq00,rinv00),_mm256_sub_pd(rinvsq00,felec));
 579
 580             /* LENNARD-JONES DISPERSION/REPULSION */
 581
 582             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 583             vvdw6            = _mm256_mul_pd(c6_00,rinvsix);
 584             vvdw12           = _mm256_mul_pd(c12_00,_mm256_mul_pd(rinvsix,rinvsix));
 585             vvdw             = _mm256_sub_pd( _mm256_mul_pd(vvdw12,one_twelfth) , _mm256_mul_pd(vvdw6,one_sixth) );
 586             fvdw             = _mm256_mul_pd(_mm256_sub_pd(vvdw12,vvdw6),rinvsq00);
 587
 588             d                = _mm256_sub_pd(r00,rswitch);
 589             d                = _mm256_max_pd(d,_mm256_setzero_pd());
 590             d2               = _mm256_mul_pd(d,d);
 591             sw               = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
 592
 593             dsw              = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
 594
 595             /* Evaluate switch function */
 596             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 597             felec            = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv00,_mm256_mul_pd(velec,dsw)) );
 598             fvdw             = _mm256_sub_pd( _mm256_mul_pd(fvdw,sw) , _mm256_mul_pd(rinv00,_mm256_mul_pd(vvdw,dsw)) );
 599             velec            = _mm256_mul_pd(velec,sw);
 600             vvdw             = _mm256_mul_pd(vvdw,sw);
 601             cutoff_mask      = _mm256_cmp_pd(rsq00,rcutoff2,_CMP_LT_OQ);
 602
 603             /* Update potential sum for this i atom from the interaction with this j atom. */
 604             velec            = _mm256_and_pd(velec,cutoff_mask);
 605             velec            = _mm256_andnot_pd(dummy_mask,velec);
 606             velecsum         = _mm256_add_pd(velecsum,velec);
 607             vvdw             = _mm256_and_pd(vvdw,cutoff_mask);
 608             vvdw             = _mm256_andnot_pd(dummy_mask,vvdw);
 609             vvdwsum          = _mm256_add_pd(vvdwsum,vvdw);
 610
 611             fscal            = _mm256_add_pd(felec,fvdw);
 612
 613             fscal            = _mm256_and_pd(fscal,cutoff_mask);
 614
 615             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 616
 617             /* Calculate temporary vectorial force */
 618             tx               = _mm256_mul_pd(fscal,dx00);
 619             ty               = _mm256_mul_pd(fscal,dy00);
 620             tz               = _mm256_mul_pd(fscal,dz00);
 621
 622             /* Update vectorial force */
 623             fix0             = _mm256_add_pd(fix0,tx);
 624             fiy0             = _mm256_add_pd(fiy0,ty);
 625             fiz0             = _mm256_add_pd(fiz0,tz);
 626
 627             fjx0             = _mm256_add_pd(fjx0,tx);
 628             fjy0             = _mm256_add_pd(fjy0,ty);
 629             fjz0             = _mm256_add_pd(fjz0,tz);
 630
 631             }
 632
 633             /**************************
 634              * CALCULATE INTERACTIONS *
 635              **************************/
 636
 637             if (gmx_mm256_any_lt(rsq10,rcutoff2))
 638             {
 639
 640             r10              = _mm256_mul_pd(rsq10,rinv10);
 641             r10              = _mm256_andnot_pd(dummy_mask,r10);
 642
 643             /* Compute parameters for interactions between i and j atoms */
 644             qq10             = _mm256_mul_pd(iq1,jq0);
 645
 646             /* EWALD ELECTROSTATICS */
 647
 648             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 649             ewrt             = _mm256_mul_pd(r10,ewtabscale);
 650             ewitab           = _mm256_cvttpd_epi32(ewrt);
 651             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 652             ewitab           = _mm_slli_epi32(ewitab,2);
 653             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 654             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 655             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 656             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 657             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 658             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 659             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 660             velec            = _mm256_mul_pd(qq10,_mm256_sub_pd(rinv10,velec));
 661             felec            = _mm256_mul_pd(_mm256_mul_pd(qq10,rinv10),_mm256_sub_pd(rinvsq10,felec));
 662
 663             d                = _mm256_sub_pd(r10,rswitch);
 664             d                = _mm256_max_pd(d,_mm256_setzero_pd());
 665             d2               = _mm256_mul_pd(d,d);
 666             sw               = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
 667
 668             dsw              = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
 669
 670             /* Evaluate switch function */
 671             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 672             felec            = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv10,_mm256_mul_pd(velec,dsw)) );
 673             velec            = _mm256_mul_pd(velec,sw);
 674             cutoff_mask      = _mm256_cmp_pd(rsq10,rcutoff2,_CMP_LT_OQ);
 675
 676             /* Update potential sum for this i atom from the interaction with this j atom. */
 677             velec            = _mm256_and_pd(velec,cutoff_mask);
 678             velec            = _mm256_andnot_pd(dummy_mask,velec);
 679             velecsum         = _mm256_add_pd(velecsum,velec);
 680
 681             fscal            = felec;
 682
 683             fscal            = _mm256_and_pd(fscal,cutoff_mask);
 684
 685             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 686
 687             /* Calculate temporary vectorial force */
 688             tx               = _mm256_mul_pd(fscal,dx10);
 689             ty               = _mm256_mul_pd(fscal,dy10);
 690             tz               = _mm256_mul_pd(fscal,dz10);
 691
 692             /* Update vectorial force */
 693             fix1             = _mm256_add_pd(fix1,tx);
 694             fiy1             = _mm256_add_pd(fiy1,ty);
 695             fiz1             = _mm256_add_pd(fiz1,tz);
 696
 697             fjx0             = _mm256_add_pd(fjx0,tx);
 698             fjy0             = _mm256_add_pd(fjy0,ty);
 699             fjz0             = _mm256_add_pd(fjz0,tz);
 700
 701             }
 702
 703             /**************************
 704              * CALCULATE INTERACTIONS *
 705              **************************/
 706
 707             if (gmx_mm256_any_lt(rsq20,rcutoff2))
 708             {
 709
 710             r20              = _mm256_mul_pd(rsq20,rinv20);
 711             r20              = _mm256_andnot_pd(dummy_mask,r20);
 712
 713             /* Compute parameters for interactions between i and j atoms */
 714             qq20             = _mm256_mul_pd(iq2,jq0);
 715
 716             /* EWALD ELECTROSTATICS */
 717
 718             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 719             ewrt             = _mm256_mul_pd(r20,ewtabscale);
 720             ewitab           = _mm256_cvttpd_epi32(ewrt);
 721             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 722             ewitab           = _mm_slli_epi32(ewitab,2);
 723             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 724             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 725             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 726             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 727             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 728             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 729             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 730             velec            = _mm256_mul_pd(qq20,_mm256_sub_pd(rinv20,velec));
 731             felec            = _mm256_mul_pd(_mm256_mul_pd(qq20,rinv20),_mm256_sub_pd(rinvsq20,felec));
 732
 733             d                = _mm256_sub_pd(r20,rswitch);
 734             d                = _mm256_max_pd(d,_mm256_setzero_pd());
 735             d2               = _mm256_mul_pd(d,d);
 736             sw               = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
 737
 738             dsw              = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
 739
 740             /* Evaluate switch function */
 741             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 742             felec            = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv20,_mm256_mul_pd(velec,dsw)) );
 743             velec            = _mm256_mul_pd(velec,sw);
 744             cutoff_mask      = _mm256_cmp_pd(rsq20,rcutoff2,_CMP_LT_OQ);
 745
 746             /* Update potential sum for this i atom from the interaction with this j atom. */
 747             velec            = _mm256_and_pd(velec,cutoff_mask);
 748             velec            = _mm256_andnot_pd(dummy_mask,velec);
 749             velecsum         = _mm256_add_pd(velecsum,velec);
 750
 751             fscal            = felec;
 752
 753             fscal            = _mm256_and_pd(fscal,cutoff_mask);
 754
 755             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 756
 757             /* Calculate temporary vectorial force */
 758             tx               = _mm256_mul_pd(fscal,dx20);
 759             ty               = _mm256_mul_pd(fscal,dy20);
 760             tz               = _mm256_mul_pd(fscal,dz20);
 761
 762             /* Update vectorial force */
 763             fix2             = _mm256_add_pd(fix2,tx);
 764             fiy2             = _mm256_add_pd(fiy2,ty);
 765             fiz2             = _mm256_add_pd(fiz2,tz);
 766
 767             fjx0             = _mm256_add_pd(fjx0,tx);
 768             fjy0             = _mm256_add_pd(fjy0,ty);
 769             fjz0             = _mm256_add_pd(fjz0,tz);
 770
 771             }
 772
 773             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
 774             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
 775             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
 776             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
 777
 778             gmx_mm256_decrement_1rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,fjx0,fjy0,fjz0);
 779
 780             /* Inner loop uses 219 flops */
 781         }
 782
 783         /* End of innermost loop */
 784
 785         gmx_mm256_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 786                                                  f+i_coord_offset,fshift+i_shift_offset);
 787
 788         ggid                        = gid[iidx];
 789         /* Update potential energies */
 790         gmx_mm256_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
 791         gmx_mm256_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
 792
 793         /* Increment number of inner iterations */
 794         inneriter                  += j_index_end - j_index_start;
 795
 796         /* Outer loop uses 20 flops */
 797     }
 798
 799     /* Increment number of outer iterations */
 800     outeriter        += nri;
 801
 802     /* Update outer/inner flops */
 803
 804     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*219);
 805 }
 806 /*
 807  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSw_VdwLJSw_GeomW3P1_F_avx_256_double
 808  * Electrostatics interaction: Ewald
 809  * VdW interaction:            LennardJones
 810  * Geometry:                   Water3-Particle
 811  * Calculate force/pot:        Force
 812  */
 813 void
 814 nb_kernel_ElecEwSw_VdwLJSw_GeomW3P1_F_avx_256_double
 815                     (t_nblist * gmx_restrict                nlist,
 816                      rvec * gmx_restrict                    xx,
 817                      rvec * gmx_restrict                    ff,
 818                      t_forcerec * gmx_restrict              fr,
 819                      t_mdatoms * gmx_restrict               mdatoms,
 820                      nb_kernel_data_t * gmx_restrict        kernel_data,
 821                      t_nrnb * gmx_restrict                  nrnb)
 822 {
 823     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 824      * just 0 for non-waters.
 825      * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
 826      * jnr indices corresponding to data put in the four positions in the SIMD register.
 827      */
 828     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 829     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 830     int              jnrA,jnrB,jnrC,jnrD;
 831     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
 832     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
 833     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
 834     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 835     real             rcutoff_scalar;
 836     real             *shiftvec,*fshift,*x,*f;
 837     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
 838     real             scratch[4*DIM];
 839     __m256d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 840     real *           vdwioffsetptr0;
 841     __m256d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 842     real *           vdwioffsetptr1;
 843     __m256d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 844     real *           vdwioffsetptr2;
 845     __m256d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 846     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
 847     __m256d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 848     __m256d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 849     __m256d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 850     __m256d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 851     __m256d          velec,felec,velecsum,facel,crf,krf,krf2;
 852     real             *charge;
 853     int              nvdwtype;
 854     __m256d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 855     int              *vdwtype;
 856     real             *vdwparam;
 857     __m256d          one_sixth   = _mm256_set1_pd(1.0/6.0);
 858     __m256d          one_twelfth = _mm256_set1_pd(1.0/12.0);
 859     __m128i          ewitab;
 860     __m256d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 861     __m256d          beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
 862     real             *ewtab;
 863     __m256d          rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
 864     real             rswitch_scalar,d_scalar;
 865     __m256d          dummy_mask,cutoff_mask;
 866     __m128           tmpmask0,tmpmask1;
 867     __m256d          signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
 868     __m256d          one     = _mm256_set1_pd(1.0);
 869     __m256d          two     = _mm256_set1_pd(2.0);
 870     x                = xx[0];
 871     f                = ff[0];
 872
 873     nri              = nlist->nri;
 874     iinr             = nlist->iinr;
 875     jindex           = nlist->jindex;
 876     jjnr             = nlist->jjnr;
 877     shiftidx         = nlist->shift;
 878     gid              = nlist->gid;
 879     shiftvec         = fr->shift_vec[0];
 880     fshift           = fr->fshift[0];
 881     facel            = _mm256_set1_pd(fr->epsfac);
 882     charge           = mdatoms->chargeA;
 883     nvdwtype         = fr->ntype;
 884     vdwparam         = fr->nbfp;
 885     vdwtype          = mdatoms->typeA;
 886
 887     sh_ewald         = _mm256_set1_pd(fr->ic->sh_ewald);
 888     beta             = _mm256_set1_pd(fr->ic->ewaldcoeff);
 889     beta2            = _mm256_mul_pd(beta,beta);
 890     beta3            = _mm256_mul_pd(beta,beta2);
 891
 892     ewtab            = fr->ic->tabq_coul_FDV0;
 893     ewtabscale       = _mm256_set1_pd(fr->ic->tabq_scale);
 894     ewtabhalfspace   = _mm256_set1_pd(0.5/fr->ic->tabq_scale);
 895
 896     /* Setup water-specific parameters */
 897     inr              = nlist->iinr[0];
 898     iq0              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+0]));
 899     iq1              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
 900     iq2              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
 901     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
 902
 903     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 904     rcutoff_scalar   = fr->rcoulomb;
 905     rcutoff          = _mm256_set1_pd(rcutoff_scalar);
 906     rcutoff2         = _mm256_mul_pd(rcutoff,rcutoff);
 907
 908     rswitch_scalar   = fr->rcoulomb_switch;
 909     rswitch          = _mm256_set1_pd(rswitch_scalar);
 910     /* Setup switch parameters */
 911     d_scalar         = rcutoff_scalar-rswitch_scalar;
 912     d                = _mm256_set1_pd(d_scalar);
 913     swV3             = _mm256_set1_pd(-10.0/(d_scalar*d_scalar*d_scalar));
 914     swV4             = _mm256_set1_pd( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 915     swV5             = _mm256_set1_pd( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 916     swF2             = _mm256_set1_pd(-30.0/(d_scalar*d_scalar*d_scalar));
 917     swF3             = _mm256_set1_pd( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 918     swF4             = _mm256_set1_pd(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 919
 920     /* Avoid stupid compiler warnings */
 921     jnrA = jnrB = jnrC = jnrD = 0;
 922     j_coord_offsetA = 0;
 923     j_coord_offsetB = 0;
 924     j_coord_offsetC = 0;
 925     j_coord_offsetD = 0;
 926
 927     outeriter        = 0;
 928     inneriter        = 0;
 929
 930     for(iidx=0;iidx<4*DIM;iidx++)
 931     {
 932         scratch[iidx] = 0.0;
 933     }
 934
 935     /* Start outer loop over neighborlists */
 936     for(iidx=0; iidx<nri; iidx++)
 937     {
 938         /* Load shift vector for this list */
 939         i_shift_offset   = DIM*shiftidx[iidx];
 940
 941         /* Load limits for loop over neighbors */
 942         j_index_start    = jindex[iidx];
 943         j_index_end      = jindex[iidx+1];
 944
 945         /* Get outer coordinate index */
 946         inr              = iinr[iidx];
 947         i_coord_offset   = DIM*inr;
 948
 949         /* Load i particle coords and add shift vector */
 950         gmx_mm256_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 951                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 952
 953         fix0             = _mm256_setzero_pd();
 954         fiy0             = _mm256_setzero_pd();
 955         fiz0             = _mm256_setzero_pd();
 956         fix1             = _mm256_setzero_pd();
 957         fiy1             = _mm256_setzero_pd();
 958         fiz1             = _mm256_setzero_pd();
 959         fix2             = _mm256_setzero_pd();
 960         fiy2             = _mm256_setzero_pd();
 961         fiz2             = _mm256_setzero_pd();
 962
 963         /* Start inner kernel loop */
 964         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 965         {
 966
 967             /* Get j neighbor index, and coordinate index */
 968             jnrA             = jjnr[jidx];
 969             jnrB             = jjnr[jidx+1];
 970             jnrC             = jjnr[jidx+2];
 971             jnrD             = jjnr[jidx+3];
 972             j_coord_offsetA  = DIM*jnrA;
 973             j_coord_offsetB  = DIM*jnrB;
 974             j_coord_offsetC  = DIM*jnrC;
 975             j_coord_offsetD  = DIM*jnrD;
 976
 977             /* load j atom coordinates */
 978             gmx_mm256_load_1rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 979                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 980                                                  &jx0,&jy0,&jz0);
 981
 982             /* Calculate displacement vector */
 983             dx00             = _mm256_sub_pd(ix0,jx0);
 984             dy00             = _mm256_sub_pd(iy0,jy0);
 985             dz00             = _mm256_sub_pd(iz0,jz0);
 986             dx10             = _mm256_sub_pd(ix1,jx0);
 987             dy10             = _mm256_sub_pd(iy1,jy0);
 988             dz10             = _mm256_sub_pd(iz1,jz0);
 989             dx20             = _mm256_sub_pd(ix2,jx0);
 990             dy20             = _mm256_sub_pd(iy2,jy0);
 991             dz20             = _mm256_sub_pd(iz2,jz0);
 992
 993             /* Calculate squared distance and things based on it */
 994             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 995             rsq10            = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
 996             rsq20            = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
 997
 998             rinv00           = gmx_mm256_invsqrt_pd(rsq00);
 999             rinv10           = gmx_mm256_invsqrt_pd(rsq10);
1000             rinv20           = gmx_mm256_invsqrt_pd(rsq20);
1001
1002             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
1003             rinvsq10         = _mm256_mul_pd(rinv10,rinv10);
1004             rinvsq20         = _mm256_mul_pd(rinv20,rinv20);
1005
1006             /* Load parameters for j particles */
1007             jq0              = gmx_mm256_load_4real_swizzle_pd(charge+jnrA+0,charge+jnrB+0,
1008                                                                  charge+jnrC+0,charge+jnrD+0);
1009             vdwjidx0A        = 2*vdwtype[jnrA+0];
1010             vdwjidx0B        = 2*vdwtype[jnrB+0];
1011             vdwjidx0C        = 2*vdwtype[jnrC+0];
1012             vdwjidx0D        = 2*vdwtype[jnrD+0];
1013
1014             fjx0             = _mm256_setzero_pd();
1015             fjy0             = _mm256_setzero_pd();
1016             fjz0             = _mm256_setzero_pd();
1017
1018             /**************************
1019              * CALCULATE INTERACTIONS *
1020              **************************/
1021
1022             if (gmx_mm256_any_lt(rsq00,rcutoff2))
1023             {
1024
1025             r00              = _mm256_mul_pd(rsq00,rinv00);
1026
1027             /* Compute parameters for interactions between i and j atoms */
1028             qq00             = _mm256_mul_pd(iq0,jq0);
1029             gmx_mm256_load_4pair_swizzle_pd(vdwioffsetptr0+vdwjidx0A,
1030                                             vdwioffsetptr0+vdwjidx0B,
1031                                             vdwioffsetptr0+vdwjidx0C,
1032                                             vdwioffsetptr0+vdwjidx0D,
1033                                             &c6_00,&c12_00);
1034
1035             /* EWALD ELECTROSTATICS */
1036
1037             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1038             ewrt             = _mm256_mul_pd(r00,ewtabscale);
1039             ewitab           = _mm256_cvttpd_epi32(ewrt);
1040             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1041             ewitab           = _mm_slli_epi32(ewitab,2);
1042             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1043             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1044             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1045             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1046             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1047             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1048             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1049             velec            = _mm256_mul_pd(qq00,_mm256_sub_pd(rinv00,velec));
1050             felec            = _mm256_mul_pd(_mm256_mul_pd(qq00,rinv00),_mm256_sub_pd(rinvsq00,felec));
1051
1052             /* LENNARD-JONES DISPERSION/REPULSION */
1053
1054             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1055             vvdw6            = _mm256_mul_pd(c6_00,rinvsix);
1056             vvdw12           = _mm256_mul_pd(c12_00,_mm256_mul_pd(rinvsix,rinvsix));
1057             vvdw             = _mm256_sub_pd( _mm256_mul_pd(vvdw12,one_twelfth) , _mm256_mul_pd(vvdw6,one_sixth) );
1058             fvdw             = _mm256_mul_pd(_mm256_sub_pd(vvdw12,vvdw6),rinvsq00);
1059
1060             d                = _mm256_sub_pd(r00,rswitch);
1061             d                = _mm256_max_pd(d,_mm256_setzero_pd());
1062             d2               = _mm256_mul_pd(d,d);
1063             sw               = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
1064
1065             dsw              = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
1066
1067             /* Evaluate switch function */
1068             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1069             felec            = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv00,_mm256_mul_pd(velec,dsw)) );
1070             fvdw             = _mm256_sub_pd( _mm256_mul_pd(fvdw,sw) , _mm256_mul_pd(rinv00,_mm256_mul_pd(vvdw,dsw)) );
1071             cutoff_mask      = _mm256_cmp_pd(rsq00,rcutoff2,_CMP_LT_OQ);
1072
1073             fscal            = _mm256_add_pd(felec,fvdw);
1074
1075             fscal            = _mm256_and_pd(fscal,cutoff_mask);
1076
1077             /* Calculate temporary vectorial force */
1078             tx               = _mm256_mul_pd(fscal,dx00);
1079             ty               = _mm256_mul_pd(fscal,dy00);
1080             tz               = _mm256_mul_pd(fscal,dz00);
1081
1082             /* Update vectorial force */
1083             fix0             = _mm256_add_pd(fix0,tx);
1084             fiy0             = _mm256_add_pd(fiy0,ty);
1085             fiz0             = _mm256_add_pd(fiz0,tz);
1086
1087             fjx0             = _mm256_add_pd(fjx0,tx);
1088             fjy0             = _mm256_add_pd(fjy0,ty);
1089             fjz0             = _mm256_add_pd(fjz0,tz);
1090
1091             }
1092
1093             /**************************
1094              * CALCULATE INTERACTIONS *
1095              **************************/
1096
1097             if (gmx_mm256_any_lt(rsq10,rcutoff2))
1098             {
1099
1100             r10              = _mm256_mul_pd(rsq10,rinv10);
1101
1102             /* Compute parameters for interactions between i and j atoms */
1103             qq10             = _mm256_mul_pd(iq1,jq0);
1104
1105             /* EWALD ELECTROSTATICS */
1106
1107             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1108             ewrt             = _mm256_mul_pd(r10,ewtabscale);
1109             ewitab           = _mm256_cvttpd_epi32(ewrt);
1110             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1111             ewitab           = _mm_slli_epi32(ewitab,2);
1112             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1113             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1114             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1115             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1116             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1117             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1118             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1119             velec            = _mm256_mul_pd(qq10,_mm256_sub_pd(rinv10,velec));
1120             felec            = _mm256_mul_pd(_mm256_mul_pd(qq10,rinv10),_mm256_sub_pd(rinvsq10,felec));
1121
1122             d                = _mm256_sub_pd(r10,rswitch);
1123             d                = _mm256_max_pd(d,_mm256_setzero_pd());
1124             d2               = _mm256_mul_pd(d,d);
1125             sw               = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
1126
1127             dsw              = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
1128
1129             /* Evaluate switch function */
1130             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1131             felec            = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv10,_mm256_mul_pd(velec,dsw)) );
1132             cutoff_mask      = _mm256_cmp_pd(rsq10,rcutoff2,_CMP_LT_OQ);
1133
1134             fscal            = felec;
1135
1136             fscal            = _mm256_and_pd(fscal,cutoff_mask);
1137
1138             /* Calculate temporary vectorial force */
1139             tx               = _mm256_mul_pd(fscal,dx10);
1140             ty               = _mm256_mul_pd(fscal,dy10);
1141             tz               = _mm256_mul_pd(fscal,dz10);
1142
1143             /* Update vectorial force */
1144             fix1             = _mm256_add_pd(fix1,tx);
1145             fiy1             = _mm256_add_pd(fiy1,ty);
1146             fiz1             = _mm256_add_pd(fiz1,tz);
1147
1148             fjx0             = _mm256_add_pd(fjx0,tx);
1149             fjy0             = _mm256_add_pd(fjy0,ty);
1150             fjz0             = _mm256_add_pd(fjz0,tz);
1151
1152             }
1153
1154             /**************************
1155              * CALCULATE INTERACTIONS *
1156              **************************/
1157
1158             if (gmx_mm256_any_lt(rsq20,rcutoff2))
1159             {
1160
1161             r20              = _mm256_mul_pd(rsq20,rinv20);
1162
1163             /* Compute parameters for interactions between i and j atoms */
1164             qq20             = _mm256_mul_pd(iq2,jq0);
1165
1166             /* EWALD ELECTROSTATICS */
1167
1168             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1169             ewrt             = _mm256_mul_pd(r20,ewtabscale);
1170             ewitab           = _mm256_cvttpd_epi32(ewrt);
1171             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1172             ewitab           = _mm_slli_epi32(ewitab,2);
1173             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1174             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1175             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1176             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1177             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1178             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1179             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1180             velec            = _mm256_mul_pd(qq20,_mm256_sub_pd(rinv20,velec));
1181             felec            = _mm256_mul_pd(_mm256_mul_pd(qq20,rinv20),_mm256_sub_pd(rinvsq20,felec));
1182
1183             d                = _mm256_sub_pd(r20,rswitch);
1184             d                = _mm256_max_pd(d,_mm256_setzero_pd());
1185             d2               = _mm256_mul_pd(d,d);
1186             sw               = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
1187
1188             dsw              = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
1189
1190             /* Evaluate switch function */
1191             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1192             felec            = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv20,_mm256_mul_pd(velec,dsw)) );
1193             cutoff_mask      = _mm256_cmp_pd(rsq20,rcutoff2,_CMP_LT_OQ);
1194
1195             fscal            = felec;
1196
1197             fscal            = _mm256_and_pd(fscal,cutoff_mask);
1198
1199             /* Calculate temporary vectorial force */
1200             tx               = _mm256_mul_pd(fscal,dx20);
1201             ty               = _mm256_mul_pd(fscal,dy20);
1202             tz               = _mm256_mul_pd(fscal,dz20);
1203
1204             /* Update vectorial force */
1205             fix2             = _mm256_add_pd(fix2,tx);
1206             fiy2             = _mm256_add_pd(fiy2,ty);
1207             fiz2             = _mm256_add_pd(fiz2,tz);
1208
1209             fjx0             = _mm256_add_pd(fjx0,tx);
1210             fjy0             = _mm256_add_pd(fjy0,ty);
1211             fjz0             = _mm256_add_pd(fjz0,tz);
1212
1213             }
1214
1215             fjptrA             = f+j_coord_offsetA;
1216             fjptrB             = f+j_coord_offsetB;
1217             fjptrC             = f+j_coord_offsetC;
1218             fjptrD             = f+j_coord_offsetD;
1219
1220             gmx_mm256_decrement_1rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,fjx0,fjy0,fjz0);
1221
1222             /* Inner loop uses 204 flops */
1223         }
1224
1225         if(jidx<j_index_end)
1226         {
1227
1228             /* Get j neighbor index, and coordinate index */
1229             jnrlistA         = jjnr[jidx];
1230             jnrlistB         = jjnr[jidx+1];
1231             jnrlistC         = jjnr[jidx+2];
1232             jnrlistD         = jjnr[jidx+3];
1233             /* Sign of each element will be negative for non-real atoms.
1234              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1235              * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
1236              */
1237             tmpmask0 = gmx_mm_castsi128_pd(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
1238
1239             tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
1240             tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
1241             dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
1242
1243             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1244             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1245             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1246             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1247             j_coord_offsetA  = DIM*jnrA;
1248             j_coord_offsetB  = DIM*jnrB;
1249             j_coord_offsetC  = DIM*jnrC;
1250             j_coord_offsetD  = DIM*jnrD;
1251
1252             /* load j atom coordinates */
1253             gmx_mm256_load_1rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1254                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1255                                                  &jx0,&jy0,&jz0);
1256
1257             /* Calculate displacement vector */
1258             dx00             = _mm256_sub_pd(ix0,jx0);
1259             dy00             = _mm256_sub_pd(iy0,jy0);
1260             dz00             = _mm256_sub_pd(iz0,jz0);
1261             dx10             = _mm256_sub_pd(ix1,jx0);
1262             dy10             = _mm256_sub_pd(iy1,jy0);
1263             dz10             = _mm256_sub_pd(iz1,jz0);
1264             dx20             = _mm256_sub_pd(ix2,jx0);
1265             dy20             = _mm256_sub_pd(iy2,jy0);
1266             dz20             = _mm256_sub_pd(iz2,jz0);
1267
1268             /* Calculate squared distance and things based on it */
1269             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
1270             rsq10            = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
1271             rsq20            = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
1272
1273             rinv00           = gmx_mm256_invsqrt_pd(rsq00);
1274             rinv10           = gmx_mm256_invsqrt_pd(rsq10);
1275             rinv20           = gmx_mm256_invsqrt_pd(rsq20);
1276
1277             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
1278             rinvsq10         = _mm256_mul_pd(rinv10,rinv10);
1279             rinvsq20         = _mm256_mul_pd(rinv20,rinv20);
1280
1281             /* Load parameters for j particles */
1282             jq0              = gmx_mm256_load_4real_swizzle_pd(charge+jnrA+0,charge+jnrB+0,
1283                                                                  charge+jnrC+0,charge+jnrD+0);
1284             vdwjidx0A        = 2*vdwtype[jnrA+0];
1285             vdwjidx0B        = 2*vdwtype[jnrB+0];
1286             vdwjidx0C        = 2*vdwtype[jnrC+0];
1287             vdwjidx0D        = 2*vdwtype[jnrD+0];
1288
1289             fjx0             = _mm256_setzero_pd();
1290             fjy0             = _mm256_setzero_pd();
1291             fjz0             = _mm256_setzero_pd();
1292
1293             /**************************
1294              * CALCULATE INTERACTIONS *
1295              **************************/
1296
1297             if (gmx_mm256_any_lt(rsq00,rcutoff2))
1298             {
1299
1300             r00              = _mm256_mul_pd(rsq00,rinv00);
1301             r00              = _mm256_andnot_pd(dummy_mask,r00);
1302
1303             /* Compute parameters for interactions between i and j atoms */
1304             qq00             = _mm256_mul_pd(iq0,jq0);
1305             gmx_mm256_load_4pair_swizzle_pd(vdwioffsetptr0+vdwjidx0A,
1306                                             vdwioffsetptr0+vdwjidx0B,
1307                                             vdwioffsetptr0+vdwjidx0C,
1308                                             vdwioffsetptr0+vdwjidx0D,
1309                                             &c6_00,&c12_00);
1310
1311             /* EWALD ELECTROSTATICS */
1312
1313             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1314             ewrt             = _mm256_mul_pd(r00,ewtabscale);
1315             ewitab           = _mm256_cvttpd_epi32(ewrt);
1316             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1317             ewitab           = _mm_slli_epi32(ewitab,2);
1318             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1319             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1320             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1321             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1322             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1323             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1324             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1325             velec            = _mm256_mul_pd(qq00,_mm256_sub_pd(rinv00,velec));
1326             felec            = _mm256_mul_pd(_mm256_mul_pd(qq00,rinv00),_mm256_sub_pd(rinvsq00,felec));
1327
1328             /* LENNARD-JONES DISPERSION/REPULSION */
1329
1330             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1331             vvdw6            = _mm256_mul_pd(c6_00,rinvsix);
1332             vvdw12           = _mm256_mul_pd(c12_00,_mm256_mul_pd(rinvsix,rinvsix));
1333             vvdw             = _mm256_sub_pd( _mm256_mul_pd(vvdw12,one_twelfth) , _mm256_mul_pd(vvdw6,one_sixth) );
1334             fvdw             = _mm256_mul_pd(_mm256_sub_pd(vvdw12,vvdw6),rinvsq00);
1335
1336             d                = _mm256_sub_pd(r00,rswitch);
1337             d                = _mm256_max_pd(d,_mm256_setzero_pd());
1338             d2               = _mm256_mul_pd(d,d);
1339             sw               = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
1340
1341             dsw              = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
1342
1343             /* Evaluate switch function */
1344             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1345             felec            = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv00,_mm256_mul_pd(velec,dsw)) );
1346             fvdw             = _mm256_sub_pd( _mm256_mul_pd(fvdw,sw) , _mm256_mul_pd(rinv00,_mm256_mul_pd(vvdw,dsw)) );
1347             cutoff_mask      = _mm256_cmp_pd(rsq00,rcutoff2,_CMP_LT_OQ);
1348
1349             fscal            = _mm256_add_pd(felec,fvdw);
1350
1351             fscal            = _mm256_and_pd(fscal,cutoff_mask);
1352
1353             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1354
1355             /* Calculate temporary vectorial force */
1356             tx               = _mm256_mul_pd(fscal,dx00);
1357             ty               = _mm256_mul_pd(fscal,dy00);
1358             tz               = _mm256_mul_pd(fscal,dz00);
1359
1360             /* Update vectorial force */
1361             fix0             = _mm256_add_pd(fix0,tx);
1362             fiy0             = _mm256_add_pd(fiy0,ty);
1363             fiz0             = _mm256_add_pd(fiz0,tz);
1364
1365             fjx0             = _mm256_add_pd(fjx0,tx);
1366             fjy0             = _mm256_add_pd(fjy0,ty);
1367             fjz0             = _mm256_add_pd(fjz0,tz);
1368
1369             }
1370
1371             /**************************
1372              * CALCULATE INTERACTIONS *
1373              **************************/
1374
1375             if (gmx_mm256_any_lt(rsq10,rcutoff2))
1376             {
1377
1378             r10              = _mm256_mul_pd(rsq10,rinv10);
1379             r10              = _mm256_andnot_pd(dummy_mask,r10);
1380
1381             /* Compute parameters for interactions between i and j atoms */
1382             qq10             = _mm256_mul_pd(iq1,jq0);
1383
1384             /* EWALD ELECTROSTATICS */
1385
1386             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1387             ewrt             = _mm256_mul_pd(r10,ewtabscale);
1388             ewitab           = _mm256_cvttpd_epi32(ewrt);
1389             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1390             ewitab           = _mm_slli_epi32(ewitab,2);
1391             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1392             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1393             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1394             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1395             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1396             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1397             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1398             velec            = _mm256_mul_pd(qq10,_mm256_sub_pd(rinv10,velec));
1399             felec            = _mm256_mul_pd(_mm256_mul_pd(qq10,rinv10),_mm256_sub_pd(rinvsq10,felec));
1400
1401             d                = _mm256_sub_pd(r10,rswitch);
1402             d                = _mm256_max_pd(d,_mm256_setzero_pd());
1403             d2               = _mm256_mul_pd(d,d);
1404             sw               = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
1405
1406             dsw              = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
1407
1408             /* Evaluate switch function */
1409             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1410             felec            = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv10,_mm256_mul_pd(velec,dsw)) );
1411             cutoff_mask      = _mm256_cmp_pd(rsq10,rcutoff2,_CMP_LT_OQ);
1412
1413             fscal            = felec;
1414
1415             fscal            = _mm256_and_pd(fscal,cutoff_mask);
1416
1417             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1418
1419             /* Calculate temporary vectorial force */
1420             tx               = _mm256_mul_pd(fscal,dx10);
1421             ty               = _mm256_mul_pd(fscal,dy10);
1422             tz               = _mm256_mul_pd(fscal,dz10);
1423
1424             /* Update vectorial force */
1425             fix1             = _mm256_add_pd(fix1,tx);
1426             fiy1             = _mm256_add_pd(fiy1,ty);
1427             fiz1             = _mm256_add_pd(fiz1,tz);
1428
1429             fjx0             = _mm256_add_pd(fjx0,tx);
1430             fjy0             = _mm256_add_pd(fjy0,ty);
1431             fjz0             = _mm256_add_pd(fjz0,tz);
1432
1433             }
1434
1435             /**************************
1436              * CALCULATE INTERACTIONS *
1437              **************************/
1438
1439             if (gmx_mm256_any_lt(rsq20,rcutoff2))
1440             {
1441
1442             r20              = _mm256_mul_pd(rsq20,rinv20);
1443             r20              = _mm256_andnot_pd(dummy_mask,r20);
1444
1445             /* Compute parameters for interactions between i and j atoms */
1446             qq20             = _mm256_mul_pd(iq2,jq0);
1447
1448             /* EWALD ELECTROSTATICS */
1449
1450             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1451             ewrt             = _mm256_mul_pd(r20,ewtabscale);
1452             ewitab           = _mm256_cvttpd_epi32(ewrt);
1453             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1454             ewitab           = _mm_slli_epi32(ewitab,2);
1455             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1456             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1457             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1458             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1459             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1460             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1461             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1462             velec            = _mm256_mul_pd(qq20,_mm256_sub_pd(rinv20,velec));
1463             felec            = _mm256_mul_pd(_mm256_mul_pd(qq20,rinv20),_mm256_sub_pd(rinvsq20,felec));
1464
1465             d                = _mm256_sub_pd(r20,rswitch);
1466             d                = _mm256_max_pd(d,_mm256_setzero_pd());
1467             d2               = _mm256_mul_pd(d,d);
1468             sw               = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
1469
1470             dsw              = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
1471
1472             /* Evaluate switch function */
1473             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1474             felec            = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv20,_mm256_mul_pd(velec,dsw)) );
1475             cutoff_mask      = _mm256_cmp_pd(rsq20,rcutoff2,_CMP_LT_OQ);
1476
1477             fscal            = felec;
1478
1479             fscal            = _mm256_and_pd(fscal,cutoff_mask);
1480
1481             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1482
1483             /* Calculate temporary vectorial force */
1484             tx               = _mm256_mul_pd(fscal,dx20);
1485             ty               = _mm256_mul_pd(fscal,dy20);
1486             tz               = _mm256_mul_pd(fscal,dz20);
1487
1488             /* Update vectorial force */
1489             fix2             = _mm256_add_pd(fix2,tx);
1490             fiy2             = _mm256_add_pd(fiy2,ty);
1491             fiz2             = _mm256_add_pd(fiz2,tz);
1492
1493             fjx0             = _mm256_add_pd(fjx0,tx);
1494             fjy0             = _mm256_add_pd(fjy0,ty);
1495             fjz0             = _mm256_add_pd(fjz0,tz);
1496
1497             }
1498
1499             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1500             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1501             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1502             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1503
1504             gmx_mm256_decrement_1rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,fjx0,fjy0,fjz0);
1505
1506             /* Inner loop uses 207 flops */
1507         }
1508
1509         /* End of innermost loop */
1510
1511         gmx_mm256_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1512                                                  f+i_coord_offset,fshift+i_shift_offset);
1513
1514         /* Increment number of inner iterations */
1515         inneriter                  += j_index_end - j_index_start;
1516
1517         /* Outer loop uses 18 flops */
1518     }
1519
1520     /* Increment number of outer iterations */
1521     outeriter        += nri;
1522
1523     /* Update outer/inner flops */
1524
1525     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_F,outeriter*18 + inneriter*207);
1526 }