src/gromacs/gmxlib/nonbonded/nb_kernel_avx_256_double/nb_kernel_ElecRF_VdwCSTab_GeomW3W3_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_ElecRF_VdwCSTab_GeomW3W3_VF_avx_256_double
  38  * Electrostatics interaction: ReactionField
  39  * VdW interaction:            CubicSplineTable
  40  * Geometry:                   Water3-Water3
  41  * Calculate force/pot:        PotentialAndForce
  42  */
  43 void
  44 nb_kernel_ElecRF_VdwCSTab_GeomW3W3_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     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
  79     __m256d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  80     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
  81     __m256d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  82     __m256d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  83     __m256d          dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
  84     __m256d          dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
  85     __m256d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  86     __m256d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
  87     __m256d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
  88     __m256d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  89     __m256d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
  90     __m256d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
  91     __m256d          velec,felec,velecsum,facel,crf,krf,krf2;
  92     real             *charge;
  93     int              nvdwtype;
  94     __m256d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  95     int              *vdwtype;
  96     real             *vdwparam;
  97     __m256d          one_sixth   = _mm256_set1_pd(1.0/6.0);
  98     __m256d          one_twelfth = _mm256_set1_pd(1.0/12.0);
  99     __m128i          vfitab;
 100     __m128i          ifour       = _mm_set1_epi32(4);
 101     __m256d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
 102     real             *vftab;
 103     __m256d          dummy_mask,cutoff_mask;
 104     __m128           tmpmask0,tmpmask1;
 105     __m256d          signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
 106     __m256d          one     = _mm256_set1_pd(1.0);
 107     __m256d          two     = _mm256_set1_pd(2.0);
 108     x                = xx[0];
 109     f                = ff[0];
 110
 111     nri              = nlist->nri;
 112     iinr             = nlist->iinr;
 113     jindex           = nlist->jindex;
 114     jjnr             = nlist->jjnr;
 115     shiftidx         = nlist->shift;
 116     gid              = nlist->gid;
 117     shiftvec         = fr->shift_vec[0];
 118     fshift           = fr->fshift[0];
 119     facel            = _mm256_set1_pd(fr->epsfac);
 120     charge           = mdatoms->chargeA;
 121     krf              = _mm256_set1_pd(fr->ic->k_rf);
 122     krf2             = _mm256_set1_pd(fr->ic->k_rf*2.0);
 123     crf              = _mm256_set1_pd(fr->ic->c_rf);
 124     nvdwtype         = fr->ntype;
 125     vdwparam         = fr->nbfp;
 126     vdwtype          = mdatoms->typeA;
 127
 128     vftab            = kernel_data->table_vdw->data;
 129     vftabscale       = _mm256_set1_pd(kernel_data->table_vdw->scale);
 130
 131     /* Setup water-specific parameters */
 132     inr              = nlist->iinr[0];
 133     iq0              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+0]));
 134     iq1              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
 135     iq2              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
 136     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
 137
 138     jq0              = _mm256_set1_pd(charge[inr+0]);
 139     jq1              = _mm256_set1_pd(charge[inr+1]);
 140     jq2              = _mm256_set1_pd(charge[inr+2]);
 141     vdwjidx0A        = 2*vdwtype[inr+0];
 142     qq00             = _mm256_mul_pd(iq0,jq0);
 143     c6_00            = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A]);
 144     c12_00           = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A+1]);
 145     qq01             = _mm256_mul_pd(iq0,jq1);
 146     qq02             = _mm256_mul_pd(iq0,jq2);
 147     qq10             = _mm256_mul_pd(iq1,jq0);
 148     qq11             = _mm256_mul_pd(iq1,jq1);
 149     qq12             = _mm256_mul_pd(iq1,jq2);
 150     qq20             = _mm256_mul_pd(iq2,jq0);
 151     qq21             = _mm256_mul_pd(iq2,jq1);
 152     qq22             = _mm256_mul_pd(iq2,jq2);
 153
 154     /* Avoid stupid compiler warnings */
 155     jnrA = jnrB = jnrC = jnrD = 0;
 156     j_coord_offsetA = 0;
 157     j_coord_offsetB = 0;
 158     j_coord_offsetC = 0;
 159     j_coord_offsetD = 0;
 160
 161     outeriter        = 0;
 162     inneriter        = 0;
 163
 164     for(iidx=0;iidx<4*DIM;iidx++)
 165     {
 166         scratch[iidx] = 0.0;
 167     }
 168
 169     /* Start outer loop over neighborlists */
 170     for(iidx=0; iidx<nri; iidx++)
 171     {
 172         /* Load shift vector for this list */
 173         i_shift_offset   = DIM*shiftidx[iidx];
 174
 175         /* Load limits for loop over neighbors */
 176         j_index_start    = jindex[iidx];
 177         j_index_end      = jindex[iidx+1];
 178
 179         /* Get outer coordinate index */
 180         inr              = iinr[iidx];
 181         i_coord_offset   = DIM*inr;
 182
 183         /* Load i particle coords and add shift vector */
 184         gmx_mm256_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 185                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 186
 187         fix0             = _mm256_setzero_pd();
 188         fiy0             = _mm256_setzero_pd();
 189         fiz0             = _mm256_setzero_pd();
 190         fix1             = _mm256_setzero_pd();
 191         fiy1             = _mm256_setzero_pd();
 192         fiz1             = _mm256_setzero_pd();
 193         fix2             = _mm256_setzero_pd();
 194         fiy2             = _mm256_setzero_pd();
 195         fiz2             = _mm256_setzero_pd();
 196
 197         /* Reset potential sums */
 198         velecsum         = _mm256_setzero_pd();
 199         vvdwsum          = _mm256_setzero_pd();
 200
 201         /* Start inner kernel loop */
 202         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 203         {
 204
 205             /* Get j neighbor index, and coordinate index */
 206             jnrA             = jjnr[jidx];
 207             jnrB             = jjnr[jidx+1];
 208             jnrC             = jjnr[jidx+2];
 209             jnrD             = jjnr[jidx+3];
 210             j_coord_offsetA  = DIM*jnrA;
 211             j_coord_offsetB  = DIM*jnrB;
 212             j_coord_offsetC  = DIM*jnrC;
 213             j_coord_offsetD  = DIM*jnrD;
 214
 215             /* load j atom coordinates */
 216             gmx_mm256_load_3rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 217                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 218                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 219
 220             /* Calculate displacement vector */
 221             dx00             = _mm256_sub_pd(ix0,jx0);
 222             dy00             = _mm256_sub_pd(iy0,jy0);
 223             dz00             = _mm256_sub_pd(iz0,jz0);
 224             dx01             = _mm256_sub_pd(ix0,jx1);
 225             dy01             = _mm256_sub_pd(iy0,jy1);
 226             dz01             = _mm256_sub_pd(iz0,jz1);
 227             dx02             = _mm256_sub_pd(ix0,jx2);
 228             dy02             = _mm256_sub_pd(iy0,jy2);
 229             dz02             = _mm256_sub_pd(iz0,jz2);
 230             dx10             = _mm256_sub_pd(ix1,jx0);
 231             dy10             = _mm256_sub_pd(iy1,jy0);
 232             dz10             = _mm256_sub_pd(iz1,jz0);
 233             dx11             = _mm256_sub_pd(ix1,jx1);
 234             dy11             = _mm256_sub_pd(iy1,jy1);
 235             dz11             = _mm256_sub_pd(iz1,jz1);
 236             dx12             = _mm256_sub_pd(ix1,jx2);
 237             dy12             = _mm256_sub_pd(iy1,jy2);
 238             dz12             = _mm256_sub_pd(iz1,jz2);
 239             dx20             = _mm256_sub_pd(ix2,jx0);
 240             dy20             = _mm256_sub_pd(iy2,jy0);
 241             dz20             = _mm256_sub_pd(iz2,jz0);
 242             dx21             = _mm256_sub_pd(ix2,jx1);
 243             dy21             = _mm256_sub_pd(iy2,jy1);
 244             dz21             = _mm256_sub_pd(iz2,jz1);
 245             dx22             = _mm256_sub_pd(ix2,jx2);
 246             dy22             = _mm256_sub_pd(iy2,jy2);
 247             dz22             = _mm256_sub_pd(iz2,jz2);
 248
 249             /* Calculate squared distance and things based on it */
 250             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 251             rsq01            = gmx_mm256_calc_rsq_pd(dx01,dy01,dz01);
 252             rsq02            = gmx_mm256_calc_rsq_pd(dx02,dy02,dz02);
 253             rsq10            = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
 254             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
 255             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
 256             rsq20            = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
 257             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
 258             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
 259
 260             rinv00           = gmx_mm256_invsqrt_pd(rsq00);
 261             rinv01           = gmx_mm256_invsqrt_pd(rsq01);
 262             rinv02           = gmx_mm256_invsqrt_pd(rsq02);
 263             rinv10           = gmx_mm256_invsqrt_pd(rsq10);
 264             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
 265             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
 266             rinv20           = gmx_mm256_invsqrt_pd(rsq20);
 267             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
 268             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
 269
 270             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
 271             rinvsq01         = _mm256_mul_pd(rinv01,rinv01);
 272             rinvsq02         = _mm256_mul_pd(rinv02,rinv02);
 273             rinvsq10         = _mm256_mul_pd(rinv10,rinv10);
 274             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
 275             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
 276             rinvsq20         = _mm256_mul_pd(rinv20,rinv20);
 277             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
 278             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
 279
 280             fjx0             = _mm256_setzero_pd();
 281             fjy0             = _mm256_setzero_pd();
 282             fjz0             = _mm256_setzero_pd();
 283             fjx1             = _mm256_setzero_pd();
 284             fjy1             = _mm256_setzero_pd();
 285             fjz1             = _mm256_setzero_pd();
 286             fjx2             = _mm256_setzero_pd();
 287             fjy2             = _mm256_setzero_pd();
 288             fjz2             = _mm256_setzero_pd();
 289
 290             /**************************
 291              * CALCULATE INTERACTIONS *
 292              **************************/
 293
 294             r00              = _mm256_mul_pd(rsq00,rinv00);
 295
 296             /* Calculate table index by multiplying r with table scale and truncate to integer */
 297             rt               = _mm256_mul_pd(r00,vftabscale);
 298             vfitab           = _mm256_cvttpd_epi32(rt);
 299             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
 300             vfitab           = _mm_slli_epi32(vfitab,3);
 301
 302             /* REACTION-FIELD ELECTROSTATICS */
 303             velec            = _mm256_mul_pd(qq00,_mm256_sub_pd(_mm256_add_pd(rinv00,_mm256_mul_pd(krf,rsq00)),crf));
 304             felec            = _mm256_mul_pd(qq00,_mm256_sub_pd(_mm256_mul_pd(rinv00,rinvsq00),krf2));
 305
 306             /* CUBIC SPLINE TABLE DISPERSION */
 307             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 308             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 309             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 310             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 311             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 312             Heps             = _mm256_mul_pd(vfeps,H);
 313             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 314             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 315             vvdw6            = _mm256_mul_pd(c6_00,VV);
 316             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 317             fvdw6            = _mm256_mul_pd(c6_00,FF);
 318
 319             /* CUBIC SPLINE TABLE REPULSION */
 320             vfitab           = _mm_add_epi32(vfitab,ifour);
 321             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 322             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 323             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 324             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 325             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 326             Heps             = _mm256_mul_pd(vfeps,H);
 327             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 328             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 329             vvdw12           = _mm256_mul_pd(c12_00,VV);
 330             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 331             fvdw12           = _mm256_mul_pd(c12_00,FF);
 332             vvdw             = _mm256_add_pd(vvdw12,vvdw6);
 333             fvdw             = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_add_pd(fvdw6,fvdw12),_mm256_mul_pd(vftabscale,rinv00)));
 334
 335             /* Update potential sum for this i atom from the interaction with this j atom. */
 336             velecsum         = _mm256_add_pd(velecsum,velec);
 337             vvdwsum          = _mm256_add_pd(vvdwsum,vvdw);
 338
 339             fscal            = _mm256_add_pd(felec,fvdw);
 340
 341             /* Calculate temporary vectorial force */
 342             tx               = _mm256_mul_pd(fscal,dx00);
 343             ty               = _mm256_mul_pd(fscal,dy00);
 344             tz               = _mm256_mul_pd(fscal,dz00);
 345
 346             /* Update vectorial force */
 347             fix0             = _mm256_add_pd(fix0,tx);
 348             fiy0             = _mm256_add_pd(fiy0,ty);
 349             fiz0             = _mm256_add_pd(fiz0,tz);
 350
 351             fjx0             = _mm256_add_pd(fjx0,tx);
 352             fjy0             = _mm256_add_pd(fjy0,ty);
 353             fjz0             = _mm256_add_pd(fjz0,tz);
 354
 355             /**************************
 356              * CALCULATE INTERACTIONS *
 357              **************************/
 358
 359             /* REACTION-FIELD ELECTROSTATICS */
 360             velec            = _mm256_mul_pd(qq01,_mm256_sub_pd(_mm256_add_pd(rinv01,_mm256_mul_pd(krf,rsq01)),crf));
 361             felec            = _mm256_mul_pd(qq01,_mm256_sub_pd(_mm256_mul_pd(rinv01,rinvsq01),krf2));
 362
 363             /* Update potential sum for this i atom from the interaction with this j atom. */
 364             velecsum         = _mm256_add_pd(velecsum,velec);
 365
 366             fscal            = felec;
 367
 368             /* Calculate temporary vectorial force */
 369             tx               = _mm256_mul_pd(fscal,dx01);
 370             ty               = _mm256_mul_pd(fscal,dy01);
 371             tz               = _mm256_mul_pd(fscal,dz01);
 372
 373             /* Update vectorial force */
 374             fix0             = _mm256_add_pd(fix0,tx);
 375             fiy0             = _mm256_add_pd(fiy0,ty);
 376             fiz0             = _mm256_add_pd(fiz0,tz);
 377
 378             fjx1             = _mm256_add_pd(fjx1,tx);
 379             fjy1             = _mm256_add_pd(fjy1,ty);
 380             fjz1             = _mm256_add_pd(fjz1,tz);
 381
 382             /**************************
 383              * CALCULATE INTERACTIONS *
 384              **************************/
 385
 386             /* REACTION-FIELD ELECTROSTATICS */
 387             velec            = _mm256_mul_pd(qq02,_mm256_sub_pd(_mm256_add_pd(rinv02,_mm256_mul_pd(krf,rsq02)),crf));
 388             felec            = _mm256_mul_pd(qq02,_mm256_sub_pd(_mm256_mul_pd(rinv02,rinvsq02),krf2));
 389
 390             /* Update potential sum for this i atom from the interaction with this j atom. */
 391             velecsum         = _mm256_add_pd(velecsum,velec);
 392
 393             fscal            = felec;
 394
 395             /* Calculate temporary vectorial force */
 396             tx               = _mm256_mul_pd(fscal,dx02);
 397             ty               = _mm256_mul_pd(fscal,dy02);
 398             tz               = _mm256_mul_pd(fscal,dz02);
 399
 400             /* Update vectorial force */
 401             fix0             = _mm256_add_pd(fix0,tx);
 402             fiy0             = _mm256_add_pd(fiy0,ty);
 403             fiz0             = _mm256_add_pd(fiz0,tz);
 404
 405             fjx2             = _mm256_add_pd(fjx2,tx);
 406             fjy2             = _mm256_add_pd(fjy2,ty);
 407             fjz2             = _mm256_add_pd(fjz2,tz);
 408
 409             /**************************
 410              * CALCULATE INTERACTIONS *
 411              **************************/
 412
 413             /* REACTION-FIELD ELECTROSTATICS */
 414             velec            = _mm256_mul_pd(qq10,_mm256_sub_pd(_mm256_add_pd(rinv10,_mm256_mul_pd(krf,rsq10)),crf));
 415             felec            = _mm256_mul_pd(qq10,_mm256_sub_pd(_mm256_mul_pd(rinv10,rinvsq10),krf2));
 416
 417             /* Update potential sum for this i atom from the interaction with this j atom. */
 418             velecsum         = _mm256_add_pd(velecsum,velec);
 419
 420             fscal            = felec;
 421
 422             /* Calculate temporary vectorial force */
 423             tx               = _mm256_mul_pd(fscal,dx10);
 424             ty               = _mm256_mul_pd(fscal,dy10);
 425             tz               = _mm256_mul_pd(fscal,dz10);
 426
 427             /* Update vectorial force */
 428             fix1             = _mm256_add_pd(fix1,tx);
 429             fiy1             = _mm256_add_pd(fiy1,ty);
 430             fiz1             = _mm256_add_pd(fiz1,tz);
 431
 432             fjx0             = _mm256_add_pd(fjx0,tx);
 433             fjy0             = _mm256_add_pd(fjy0,ty);
 434             fjz0             = _mm256_add_pd(fjz0,tz);
 435
 436             /**************************
 437              * CALCULATE INTERACTIONS *
 438              **************************/
 439
 440             /* REACTION-FIELD ELECTROSTATICS */
 441             velec            = _mm256_mul_pd(qq11,_mm256_sub_pd(_mm256_add_pd(rinv11,_mm256_mul_pd(krf,rsq11)),crf));
 442             felec            = _mm256_mul_pd(qq11,_mm256_sub_pd(_mm256_mul_pd(rinv11,rinvsq11),krf2));
 443
 444             /* Update potential sum for this i atom from the interaction with this j atom. */
 445             velecsum         = _mm256_add_pd(velecsum,velec);
 446
 447             fscal            = felec;
 448
 449             /* Calculate temporary vectorial force */
 450             tx               = _mm256_mul_pd(fscal,dx11);
 451             ty               = _mm256_mul_pd(fscal,dy11);
 452             tz               = _mm256_mul_pd(fscal,dz11);
 453
 454             /* Update vectorial force */
 455             fix1             = _mm256_add_pd(fix1,tx);
 456             fiy1             = _mm256_add_pd(fiy1,ty);
 457             fiz1             = _mm256_add_pd(fiz1,tz);
 458
 459             fjx1             = _mm256_add_pd(fjx1,tx);
 460             fjy1             = _mm256_add_pd(fjy1,ty);
 461             fjz1             = _mm256_add_pd(fjz1,tz);
 462
 463             /**************************
 464              * CALCULATE INTERACTIONS *
 465              **************************/
 466
 467             /* REACTION-FIELD ELECTROSTATICS */
 468             velec            = _mm256_mul_pd(qq12,_mm256_sub_pd(_mm256_add_pd(rinv12,_mm256_mul_pd(krf,rsq12)),crf));
 469             felec            = _mm256_mul_pd(qq12,_mm256_sub_pd(_mm256_mul_pd(rinv12,rinvsq12),krf2));
 470
 471             /* Update potential sum for this i atom from the interaction with this j atom. */
 472             velecsum         = _mm256_add_pd(velecsum,velec);
 473
 474             fscal            = felec;
 475
 476             /* Calculate temporary vectorial force */
 477             tx               = _mm256_mul_pd(fscal,dx12);
 478             ty               = _mm256_mul_pd(fscal,dy12);
 479             tz               = _mm256_mul_pd(fscal,dz12);
 480
 481             /* Update vectorial force */
 482             fix1             = _mm256_add_pd(fix1,tx);
 483             fiy1             = _mm256_add_pd(fiy1,ty);
 484             fiz1             = _mm256_add_pd(fiz1,tz);
 485
 486             fjx2             = _mm256_add_pd(fjx2,tx);
 487             fjy2             = _mm256_add_pd(fjy2,ty);
 488             fjz2             = _mm256_add_pd(fjz2,tz);
 489
 490             /**************************
 491              * CALCULATE INTERACTIONS *
 492              **************************/
 493
 494             /* REACTION-FIELD ELECTROSTATICS */
 495             velec            = _mm256_mul_pd(qq20,_mm256_sub_pd(_mm256_add_pd(rinv20,_mm256_mul_pd(krf,rsq20)),crf));
 496             felec            = _mm256_mul_pd(qq20,_mm256_sub_pd(_mm256_mul_pd(rinv20,rinvsq20),krf2));
 497
 498             /* Update potential sum for this i atom from the interaction with this j atom. */
 499             velecsum         = _mm256_add_pd(velecsum,velec);
 500
 501             fscal            = felec;
 502
 503             /* Calculate temporary vectorial force */
 504             tx               = _mm256_mul_pd(fscal,dx20);
 505             ty               = _mm256_mul_pd(fscal,dy20);
 506             tz               = _mm256_mul_pd(fscal,dz20);
 507
 508             /* Update vectorial force */
 509             fix2             = _mm256_add_pd(fix2,tx);
 510             fiy2             = _mm256_add_pd(fiy2,ty);
 511             fiz2             = _mm256_add_pd(fiz2,tz);
 512
 513             fjx0             = _mm256_add_pd(fjx0,tx);
 514             fjy0             = _mm256_add_pd(fjy0,ty);
 515             fjz0             = _mm256_add_pd(fjz0,tz);
 516
 517             /**************************
 518              * CALCULATE INTERACTIONS *
 519              **************************/
 520
 521             /* REACTION-FIELD ELECTROSTATICS */
 522             velec            = _mm256_mul_pd(qq21,_mm256_sub_pd(_mm256_add_pd(rinv21,_mm256_mul_pd(krf,rsq21)),crf));
 523             felec            = _mm256_mul_pd(qq21,_mm256_sub_pd(_mm256_mul_pd(rinv21,rinvsq21),krf2));
 524
 525             /* Update potential sum for this i atom from the interaction with this j atom. */
 526             velecsum         = _mm256_add_pd(velecsum,velec);
 527
 528             fscal            = felec;
 529
 530             /* Calculate temporary vectorial force */
 531             tx               = _mm256_mul_pd(fscal,dx21);
 532             ty               = _mm256_mul_pd(fscal,dy21);
 533             tz               = _mm256_mul_pd(fscal,dz21);
 534
 535             /* Update vectorial force */
 536             fix2             = _mm256_add_pd(fix2,tx);
 537             fiy2             = _mm256_add_pd(fiy2,ty);
 538             fiz2             = _mm256_add_pd(fiz2,tz);
 539
 540             fjx1             = _mm256_add_pd(fjx1,tx);
 541             fjy1             = _mm256_add_pd(fjy1,ty);
 542             fjz1             = _mm256_add_pd(fjz1,tz);
 543
 544             /**************************
 545              * CALCULATE INTERACTIONS *
 546              **************************/
 547
 548             /* REACTION-FIELD ELECTROSTATICS */
 549             velec            = _mm256_mul_pd(qq22,_mm256_sub_pd(_mm256_add_pd(rinv22,_mm256_mul_pd(krf,rsq22)),crf));
 550             felec            = _mm256_mul_pd(qq22,_mm256_sub_pd(_mm256_mul_pd(rinv22,rinvsq22),krf2));
 551
 552             /* Update potential sum for this i atom from the interaction with this j atom. */
 553             velecsum         = _mm256_add_pd(velecsum,velec);
 554
 555             fscal            = felec;
 556
 557             /* Calculate temporary vectorial force */
 558             tx               = _mm256_mul_pd(fscal,dx22);
 559             ty               = _mm256_mul_pd(fscal,dy22);
 560             tz               = _mm256_mul_pd(fscal,dz22);
 561
 562             /* Update vectorial force */
 563             fix2             = _mm256_add_pd(fix2,tx);
 564             fiy2             = _mm256_add_pd(fiy2,ty);
 565             fiz2             = _mm256_add_pd(fiz2,tz);
 566
 567             fjx2             = _mm256_add_pd(fjx2,tx);
 568             fjy2             = _mm256_add_pd(fjy2,ty);
 569             fjz2             = _mm256_add_pd(fjz2,tz);
 570
 571             fjptrA             = f+j_coord_offsetA;
 572             fjptrB             = f+j_coord_offsetB;
 573             fjptrC             = f+j_coord_offsetC;
 574             fjptrD             = f+j_coord_offsetD;
 575
 576             gmx_mm256_decrement_3rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
 577                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 578
 579             /* Inner loop uses 323 flops */
 580         }
 581
 582         if(jidx<j_index_end)
 583         {
 584
 585             /* Get j neighbor index, and coordinate index */
 586             jnrlistA         = jjnr[jidx];
 587             jnrlistB         = jjnr[jidx+1];
 588             jnrlistC         = jjnr[jidx+2];
 589             jnrlistD         = jjnr[jidx+3];
 590             /* Sign of each element will be negative for non-real atoms.
 591              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 592              * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
 593              */
 594             tmpmask0 = gmx_mm_castsi128_pd(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 595
 596             tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
 597             tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
 598             dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
 599
 600             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 601             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 602             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 603             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 604             j_coord_offsetA  = DIM*jnrA;
 605             j_coord_offsetB  = DIM*jnrB;
 606             j_coord_offsetC  = DIM*jnrC;
 607             j_coord_offsetD  = DIM*jnrD;
 608
 609             /* load j atom coordinates */
 610             gmx_mm256_load_3rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 611                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 612                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 613
 614             /* Calculate displacement vector */
 615             dx00             = _mm256_sub_pd(ix0,jx0);
 616             dy00             = _mm256_sub_pd(iy0,jy0);
 617             dz00             = _mm256_sub_pd(iz0,jz0);
 618             dx01             = _mm256_sub_pd(ix0,jx1);
 619             dy01             = _mm256_sub_pd(iy0,jy1);
 620             dz01             = _mm256_sub_pd(iz0,jz1);
 621             dx02             = _mm256_sub_pd(ix0,jx2);
 622             dy02             = _mm256_sub_pd(iy0,jy2);
 623             dz02             = _mm256_sub_pd(iz0,jz2);
 624             dx10             = _mm256_sub_pd(ix1,jx0);
 625             dy10             = _mm256_sub_pd(iy1,jy0);
 626             dz10             = _mm256_sub_pd(iz1,jz0);
 627             dx11             = _mm256_sub_pd(ix1,jx1);
 628             dy11             = _mm256_sub_pd(iy1,jy1);
 629             dz11             = _mm256_sub_pd(iz1,jz1);
 630             dx12             = _mm256_sub_pd(ix1,jx2);
 631             dy12             = _mm256_sub_pd(iy1,jy2);
 632             dz12             = _mm256_sub_pd(iz1,jz2);
 633             dx20             = _mm256_sub_pd(ix2,jx0);
 634             dy20             = _mm256_sub_pd(iy2,jy0);
 635             dz20             = _mm256_sub_pd(iz2,jz0);
 636             dx21             = _mm256_sub_pd(ix2,jx1);
 637             dy21             = _mm256_sub_pd(iy2,jy1);
 638             dz21             = _mm256_sub_pd(iz2,jz1);
 639             dx22             = _mm256_sub_pd(ix2,jx2);
 640             dy22             = _mm256_sub_pd(iy2,jy2);
 641             dz22             = _mm256_sub_pd(iz2,jz2);
 642
 643             /* Calculate squared distance and things based on it */
 644             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 645             rsq01            = gmx_mm256_calc_rsq_pd(dx01,dy01,dz01);
 646             rsq02            = gmx_mm256_calc_rsq_pd(dx02,dy02,dz02);
 647             rsq10            = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
 648             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
 649             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
 650             rsq20            = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
 651             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
 652             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
 653
 654             rinv00           = gmx_mm256_invsqrt_pd(rsq00);
 655             rinv01           = gmx_mm256_invsqrt_pd(rsq01);
 656             rinv02           = gmx_mm256_invsqrt_pd(rsq02);
 657             rinv10           = gmx_mm256_invsqrt_pd(rsq10);
 658             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
 659             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
 660             rinv20           = gmx_mm256_invsqrt_pd(rsq20);
 661             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
 662             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
 663
 664             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
 665             rinvsq01         = _mm256_mul_pd(rinv01,rinv01);
 666             rinvsq02         = _mm256_mul_pd(rinv02,rinv02);
 667             rinvsq10         = _mm256_mul_pd(rinv10,rinv10);
 668             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
 669             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
 670             rinvsq20         = _mm256_mul_pd(rinv20,rinv20);
 671             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
 672             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
 673
 674             fjx0             = _mm256_setzero_pd();
 675             fjy0             = _mm256_setzero_pd();
 676             fjz0             = _mm256_setzero_pd();
 677             fjx1             = _mm256_setzero_pd();
 678             fjy1             = _mm256_setzero_pd();
 679             fjz1             = _mm256_setzero_pd();
 680             fjx2             = _mm256_setzero_pd();
 681             fjy2             = _mm256_setzero_pd();
 682             fjz2             = _mm256_setzero_pd();
 683
 684             /**************************
 685              * CALCULATE INTERACTIONS *
 686              **************************/
 687
 688             r00              = _mm256_mul_pd(rsq00,rinv00);
 689             r00              = _mm256_andnot_pd(dummy_mask,r00);
 690
 691             /* Calculate table index by multiplying r with table scale and truncate to integer */
 692             rt               = _mm256_mul_pd(r00,vftabscale);
 693             vfitab           = _mm256_cvttpd_epi32(rt);
 694             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
 695             vfitab           = _mm_slli_epi32(vfitab,3);
 696
 697             /* REACTION-FIELD ELECTROSTATICS */
 698             velec            = _mm256_mul_pd(qq00,_mm256_sub_pd(_mm256_add_pd(rinv00,_mm256_mul_pd(krf,rsq00)),crf));
 699             felec            = _mm256_mul_pd(qq00,_mm256_sub_pd(_mm256_mul_pd(rinv00,rinvsq00),krf2));
 700
 701             /* CUBIC SPLINE TABLE DISPERSION */
 702             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 703             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 704             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 705             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 706             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 707             Heps             = _mm256_mul_pd(vfeps,H);
 708             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 709             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 710             vvdw6            = _mm256_mul_pd(c6_00,VV);
 711             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 712             fvdw6            = _mm256_mul_pd(c6_00,FF);
 713
 714             /* CUBIC SPLINE TABLE REPULSION */
 715             vfitab           = _mm_add_epi32(vfitab,ifour);
 716             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 717             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 718             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 719             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 720             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 721             Heps             = _mm256_mul_pd(vfeps,H);
 722             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 723             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 724             vvdw12           = _mm256_mul_pd(c12_00,VV);
 725             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 726             fvdw12           = _mm256_mul_pd(c12_00,FF);
 727             vvdw             = _mm256_add_pd(vvdw12,vvdw6);
 728             fvdw             = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_add_pd(fvdw6,fvdw12),_mm256_mul_pd(vftabscale,rinv00)));
 729
 730             /* Update potential sum for this i atom from the interaction with this j atom. */
 731             velec            = _mm256_andnot_pd(dummy_mask,velec);
 732             velecsum         = _mm256_add_pd(velecsum,velec);
 733             vvdw             = _mm256_andnot_pd(dummy_mask,vvdw);
 734             vvdwsum          = _mm256_add_pd(vvdwsum,vvdw);
 735
 736             fscal            = _mm256_add_pd(felec,fvdw);
 737
 738             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 739
 740             /* Calculate temporary vectorial force */
 741             tx               = _mm256_mul_pd(fscal,dx00);
 742             ty               = _mm256_mul_pd(fscal,dy00);
 743             tz               = _mm256_mul_pd(fscal,dz00);
 744
 745             /* Update vectorial force */
 746             fix0             = _mm256_add_pd(fix0,tx);
 747             fiy0             = _mm256_add_pd(fiy0,ty);
 748             fiz0             = _mm256_add_pd(fiz0,tz);
 749
 750             fjx0             = _mm256_add_pd(fjx0,tx);
 751             fjy0             = _mm256_add_pd(fjy0,ty);
 752             fjz0             = _mm256_add_pd(fjz0,tz);
 753
 754             /**************************
 755              * CALCULATE INTERACTIONS *
 756              **************************/
 757
 758             /* REACTION-FIELD ELECTROSTATICS */
 759             velec            = _mm256_mul_pd(qq01,_mm256_sub_pd(_mm256_add_pd(rinv01,_mm256_mul_pd(krf,rsq01)),crf));
 760             felec            = _mm256_mul_pd(qq01,_mm256_sub_pd(_mm256_mul_pd(rinv01,rinvsq01),krf2));
 761
 762             /* Update potential sum for this i atom from the interaction with this j atom. */
 763             velec            = _mm256_andnot_pd(dummy_mask,velec);
 764             velecsum         = _mm256_add_pd(velecsum,velec);
 765
 766             fscal            = felec;
 767
 768             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 769
 770             /* Calculate temporary vectorial force */
 771             tx               = _mm256_mul_pd(fscal,dx01);
 772             ty               = _mm256_mul_pd(fscal,dy01);
 773             tz               = _mm256_mul_pd(fscal,dz01);
 774
 775             /* Update vectorial force */
 776             fix0             = _mm256_add_pd(fix0,tx);
 777             fiy0             = _mm256_add_pd(fiy0,ty);
 778             fiz0             = _mm256_add_pd(fiz0,tz);
 779
 780             fjx1             = _mm256_add_pd(fjx1,tx);
 781             fjy1             = _mm256_add_pd(fjy1,ty);
 782             fjz1             = _mm256_add_pd(fjz1,tz);
 783
 784             /**************************
 785              * CALCULATE INTERACTIONS *
 786              **************************/
 787
 788             /* REACTION-FIELD ELECTROSTATICS */
 789             velec            = _mm256_mul_pd(qq02,_mm256_sub_pd(_mm256_add_pd(rinv02,_mm256_mul_pd(krf,rsq02)),crf));
 790             felec            = _mm256_mul_pd(qq02,_mm256_sub_pd(_mm256_mul_pd(rinv02,rinvsq02),krf2));
 791
 792             /* Update potential sum for this i atom from the interaction with this j atom. */
 793             velec            = _mm256_andnot_pd(dummy_mask,velec);
 794             velecsum         = _mm256_add_pd(velecsum,velec);
 795
 796             fscal            = felec;
 797
 798             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 799
 800             /* Calculate temporary vectorial force */
 801             tx               = _mm256_mul_pd(fscal,dx02);
 802             ty               = _mm256_mul_pd(fscal,dy02);
 803             tz               = _mm256_mul_pd(fscal,dz02);
 804
 805             /* Update vectorial force */
 806             fix0             = _mm256_add_pd(fix0,tx);
 807             fiy0             = _mm256_add_pd(fiy0,ty);
 808             fiz0             = _mm256_add_pd(fiz0,tz);
 809
 810             fjx2             = _mm256_add_pd(fjx2,tx);
 811             fjy2             = _mm256_add_pd(fjy2,ty);
 812             fjz2             = _mm256_add_pd(fjz2,tz);
 813
 814             /**************************
 815              * CALCULATE INTERACTIONS *
 816              **************************/
 817
 818             /* REACTION-FIELD ELECTROSTATICS */
 819             velec            = _mm256_mul_pd(qq10,_mm256_sub_pd(_mm256_add_pd(rinv10,_mm256_mul_pd(krf,rsq10)),crf));
 820             felec            = _mm256_mul_pd(qq10,_mm256_sub_pd(_mm256_mul_pd(rinv10,rinvsq10),krf2));
 821
 822             /* Update potential sum for this i atom from the interaction with this j atom. */
 823             velec            = _mm256_andnot_pd(dummy_mask,velec);
 824             velecsum         = _mm256_add_pd(velecsum,velec);
 825
 826             fscal            = felec;
 827
 828             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 829
 830             /* Calculate temporary vectorial force */
 831             tx               = _mm256_mul_pd(fscal,dx10);
 832             ty               = _mm256_mul_pd(fscal,dy10);
 833             tz               = _mm256_mul_pd(fscal,dz10);
 834
 835             /* Update vectorial force */
 836             fix1             = _mm256_add_pd(fix1,tx);
 837             fiy1             = _mm256_add_pd(fiy1,ty);
 838             fiz1             = _mm256_add_pd(fiz1,tz);
 839
 840             fjx0             = _mm256_add_pd(fjx0,tx);
 841             fjy0             = _mm256_add_pd(fjy0,ty);
 842             fjz0             = _mm256_add_pd(fjz0,tz);
 843
 844             /**************************
 845              * CALCULATE INTERACTIONS *
 846              **************************/
 847
 848             /* REACTION-FIELD ELECTROSTATICS */
 849             velec            = _mm256_mul_pd(qq11,_mm256_sub_pd(_mm256_add_pd(rinv11,_mm256_mul_pd(krf,rsq11)),crf));
 850             felec            = _mm256_mul_pd(qq11,_mm256_sub_pd(_mm256_mul_pd(rinv11,rinvsq11),krf2));
 851
 852             /* Update potential sum for this i atom from the interaction with this j atom. */
 853             velec            = _mm256_andnot_pd(dummy_mask,velec);
 854             velecsum         = _mm256_add_pd(velecsum,velec);
 855
 856             fscal            = felec;
 857
 858             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 859
 860             /* Calculate temporary vectorial force */
 861             tx               = _mm256_mul_pd(fscal,dx11);
 862             ty               = _mm256_mul_pd(fscal,dy11);
 863             tz               = _mm256_mul_pd(fscal,dz11);
 864
 865             /* Update vectorial force */
 866             fix1             = _mm256_add_pd(fix1,tx);
 867             fiy1             = _mm256_add_pd(fiy1,ty);
 868             fiz1             = _mm256_add_pd(fiz1,tz);
 869
 870             fjx1             = _mm256_add_pd(fjx1,tx);
 871             fjy1             = _mm256_add_pd(fjy1,ty);
 872             fjz1             = _mm256_add_pd(fjz1,tz);
 873
 874             /**************************
 875              * CALCULATE INTERACTIONS *
 876              **************************/
 877
 878             /* REACTION-FIELD ELECTROSTATICS */
 879             velec            = _mm256_mul_pd(qq12,_mm256_sub_pd(_mm256_add_pd(rinv12,_mm256_mul_pd(krf,rsq12)),crf));
 880             felec            = _mm256_mul_pd(qq12,_mm256_sub_pd(_mm256_mul_pd(rinv12,rinvsq12),krf2));
 881
 882             /* Update potential sum for this i atom from the interaction with this j atom. */
 883             velec            = _mm256_andnot_pd(dummy_mask,velec);
 884             velecsum         = _mm256_add_pd(velecsum,velec);
 885
 886             fscal            = felec;
 887
 888             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 889
 890             /* Calculate temporary vectorial force */
 891             tx               = _mm256_mul_pd(fscal,dx12);
 892             ty               = _mm256_mul_pd(fscal,dy12);
 893             tz               = _mm256_mul_pd(fscal,dz12);
 894
 895             /* Update vectorial force */
 896             fix1             = _mm256_add_pd(fix1,tx);
 897             fiy1             = _mm256_add_pd(fiy1,ty);
 898             fiz1             = _mm256_add_pd(fiz1,tz);
 899
 900             fjx2             = _mm256_add_pd(fjx2,tx);
 901             fjy2             = _mm256_add_pd(fjy2,ty);
 902             fjz2             = _mm256_add_pd(fjz2,tz);
 903
 904             /**************************
 905              * CALCULATE INTERACTIONS *
 906              **************************/
 907
 908             /* REACTION-FIELD ELECTROSTATICS */
 909             velec            = _mm256_mul_pd(qq20,_mm256_sub_pd(_mm256_add_pd(rinv20,_mm256_mul_pd(krf,rsq20)),crf));
 910             felec            = _mm256_mul_pd(qq20,_mm256_sub_pd(_mm256_mul_pd(rinv20,rinvsq20),krf2));
 911
 912             /* Update potential sum for this i atom from the interaction with this j atom. */
 913             velec            = _mm256_andnot_pd(dummy_mask,velec);
 914             velecsum         = _mm256_add_pd(velecsum,velec);
 915
 916             fscal            = felec;
 917
 918             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 919
 920             /* Calculate temporary vectorial force */
 921             tx               = _mm256_mul_pd(fscal,dx20);
 922             ty               = _mm256_mul_pd(fscal,dy20);
 923             tz               = _mm256_mul_pd(fscal,dz20);
 924
 925             /* Update vectorial force */
 926             fix2             = _mm256_add_pd(fix2,tx);
 927             fiy2             = _mm256_add_pd(fiy2,ty);
 928             fiz2             = _mm256_add_pd(fiz2,tz);
 929
 930             fjx0             = _mm256_add_pd(fjx0,tx);
 931             fjy0             = _mm256_add_pd(fjy0,ty);
 932             fjz0             = _mm256_add_pd(fjz0,tz);
 933
 934             /**************************
 935              * CALCULATE INTERACTIONS *
 936              **************************/
 937
 938             /* REACTION-FIELD ELECTROSTATICS */
 939             velec            = _mm256_mul_pd(qq21,_mm256_sub_pd(_mm256_add_pd(rinv21,_mm256_mul_pd(krf,rsq21)),crf));
 940             felec            = _mm256_mul_pd(qq21,_mm256_sub_pd(_mm256_mul_pd(rinv21,rinvsq21),krf2));
 941
 942             /* Update potential sum for this i atom from the interaction with this j atom. */
 943             velec            = _mm256_andnot_pd(dummy_mask,velec);
 944             velecsum         = _mm256_add_pd(velecsum,velec);
 945
 946             fscal            = felec;
 947
 948             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 949
 950             /* Calculate temporary vectorial force */
 951             tx               = _mm256_mul_pd(fscal,dx21);
 952             ty               = _mm256_mul_pd(fscal,dy21);
 953             tz               = _mm256_mul_pd(fscal,dz21);
 954
 955             /* Update vectorial force */
 956             fix2             = _mm256_add_pd(fix2,tx);
 957             fiy2             = _mm256_add_pd(fiy2,ty);
 958             fiz2             = _mm256_add_pd(fiz2,tz);
 959
 960             fjx1             = _mm256_add_pd(fjx1,tx);
 961             fjy1             = _mm256_add_pd(fjy1,ty);
 962             fjz1             = _mm256_add_pd(fjz1,tz);
 963
 964             /**************************
 965              * CALCULATE INTERACTIONS *
 966              **************************/
 967
 968             /* REACTION-FIELD ELECTROSTATICS */
 969             velec            = _mm256_mul_pd(qq22,_mm256_sub_pd(_mm256_add_pd(rinv22,_mm256_mul_pd(krf,rsq22)),crf));
 970             felec            = _mm256_mul_pd(qq22,_mm256_sub_pd(_mm256_mul_pd(rinv22,rinvsq22),krf2));
 971
 972             /* Update potential sum for this i atom from the interaction with this j atom. */
 973             velec            = _mm256_andnot_pd(dummy_mask,velec);
 974             velecsum         = _mm256_add_pd(velecsum,velec);
 975
 976             fscal            = felec;
 977
 978             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 979
 980             /* Calculate temporary vectorial force */
 981             tx               = _mm256_mul_pd(fscal,dx22);
 982             ty               = _mm256_mul_pd(fscal,dy22);
 983             tz               = _mm256_mul_pd(fscal,dz22);
 984
 985             /* Update vectorial force */
 986             fix2             = _mm256_add_pd(fix2,tx);
 987             fiy2             = _mm256_add_pd(fiy2,ty);
 988             fiz2             = _mm256_add_pd(fiz2,tz);
 989
 990             fjx2             = _mm256_add_pd(fjx2,tx);
 991             fjy2             = _mm256_add_pd(fjy2,ty);
 992             fjz2             = _mm256_add_pd(fjz2,tz);
 993
 994             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
 995             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
 996             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
 997             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
 998
 999             gmx_mm256_decrement_3rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1000                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1001
1002             /* Inner loop uses 324 flops */
1003         }
1004
1005         /* End of innermost loop */
1006
1007         gmx_mm256_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1008                                                  f+i_coord_offset,fshift+i_shift_offset);
1009
1010         ggid                        = gid[iidx];
1011         /* Update potential energies */
1012         gmx_mm256_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1013         gmx_mm256_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1014
1015         /* Increment number of inner iterations */
1016         inneriter                  += j_index_end - j_index_start;
1017
1018         /* Outer loop uses 20 flops */
1019     }
1020
1021     /* Increment number of outer iterations */
1022     outeriter        += nri;
1023
1024     /* Update outer/inner flops */
1025
1026     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_VF,outeriter*20 + inneriter*324);
1027 }
1028 /*
1029  * Gromacs nonbonded kernel:   nb_kernel_ElecRF_VdwCSTab_GeomW3W3_F_avx_256_double
1030  * Electrostatics interaction: ReactionField
1031  * VdW interaction:            CubicSplineTable
1032  * Geometry:                   Water3-Water3
1033  * Calculate force/pot:        Force
1034  */
1035 void
1036 nb_kernel_ElecRF_VdwCSTab_GeomW3W3_F_avx_256_double
1037                     (t_nblist * gmx_restrict                nlist,
1038                      rvec * gmx_restrict                    xx,
1039                      rvec * gmx_restrict                    ff,
1040                      t_forcerec * gmx_restrict              fr,
1041                      t_mdatoms * gmx_restrict               mdatoms,
1042                      nb_kernel_data_t * gmx_restrict        kernel_data,
1043                      t_nrnb * gmx_restrict                  nrnb)
1044 {
1045     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1046      * just 0 for non-waters.
1047      * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
1048      * jnr indices corresponding to data put in the four positions in the SIMD register.
1049      */
1050     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1051     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1052     int              jnrA,jnrB,jnrC,jnrD;
1053     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1054     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1055     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1056     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1057     real             rcutoff_scalar;
1058     real             *shiftvec,*fshift,*x,*f;
1059     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
1060     real             scratch[4*DIM];
1061     __m256d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1062     real *           vdwioffsetptr0;
1063     __m256d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1064     real *           vdwioffsetptr1;
1065     __m256d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1066     real *           vdwioffsetptr2;
1067     __m256d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1068     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
1069     __m256d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1070     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
1071     __m256d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1072     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
1073     __m256d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1074     __m256d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1075     __m256d          dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1076     __m256d          dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1077     __m256d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1078     __m256d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1079     __m256d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1080     __m256d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1081     __m256d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1082     __m256d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1083     __m256d          velec,felec,velecsum,facel,crf,krf,krf2;
1084     real             *charge;
1085     int              nvdwtype;
1086     __m256d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1087     int              *vdwtype;
1088     real             *vdwparam;
1089     __m256d          one_sixth   = _mm256_set1_pd(1.0/6.0);
1090     __m256d          one_twelfth = _mm256_set1_pd(1.0/12.0);
1091     __m128i          vfitab;
1092     __m128i          ifour       = _mm_set1_epi32(4);
1093     __m256d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
1094     real             *vftab;
1095     __m256d          dummy_mask,cutoff_mask;
1096     __m128           tmpmask0,tmpmask1;
1097     __m256d          signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
1098     __m256d          one     = _mm256_set1_pd(1.0);
1099     __m256d          two     = _mm256_set1_pd(2.0);
1100     x                = xx[0];
1101     f                = ff[0];
1102
1103     nri              = nlist->nri;
1104     iinr             = nlist->iinr;
1105     jindex           = nlist->jindex;
1106     jjnr             = nlist->jjnr;
1107     shiftidx         = nlist->shift;
1108     gid              = nlist->gid;
1109     shiftvec         = fr->shift_vec[0];
1110     fshift           = fr->fshift[0];
1111     facel            = _mm256_set1_pd(fr->epsfac);
1112     charge           = mdatoms->chargeA;
1113     krf              = _mm256_set1_pd(fr->ic->k_rf);
1114     krf2             = _mm256_set1_pd(fr->ic->k_rf*2.0);
1115     crf              = _mm256_set1_pd(fr->ic->c_rf);
1116     nvdwtype         = fr->ntype;
1117     vdwparam         = fr->nbfp;
1118     vdwtype          = mdatoms->typeA;
1119
1120     vftab            = kernel_data->table_vdw->data;
1121     vftabscale       = _mm256_set1_pd(kernel_data->table_vdw->scale);
1122
1123     /* Setup water-specific parameters */
1124     inr              = nlist->iinr[0];
1125     iq0              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+0]));
1126     iq1              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
1127     iq2              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
1128     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
1129
1130     jq0              = _mm256_set1_pd(charge[inr+0]);
1131     jq1              = _mm256_set1_pd(charge[inr+1]);
1132     jq2              = _mm256_set1_pd(charge[inr+2]);
1133     vdwjidx0A        = 2*vdwtype[inr+0];
1134     qq00             = _mm256_mul_pd(iq0,jq0);
1135     c6_00            = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A]);
1136     c12_00           = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A+1]);
1137     qq01             = _mm256_mul_pd(iq0,jq1);
1138     qq02             = _mm256_mul_pd(iq0,jq2);
1139     qq10             = _mm256_mul_pd(iq1,jq0);
1140     qq11             = _mm256_mul_pd(iq1,jq1);
1141     qq12             = _mm256_mul_pd(iq1,jq2);
1142     qq20             = _mm256_mul_pd(iq2,jq0);
1143     qq21             = _mm256_mul_pd(iq2,jq1);
1144     qq22             = _mm256_mul_pd(iq2,jq2);
1145
1146     /* Avoid stupid compiler warnings */
1147     jnrA = jnrB = jnrC = jnrD = 0;
1148     j_coord_offsetA = 0;
1149     j_coord_offsetB = 0;
1150     j_coord_offsetC = 0;
1151     j_coord_offsetD = 0;
1152
1153     outeriter        = 0;
1154     inneriter        = 0;
1155
1156     for(iidx=0;iidx<4*DIM;iidx++)
1157     {
1158         scratch[iidx] = 0.0;
1159     }
1160
1161     /* Start outer loop over neighborlists */
1162     for(iidx=0; iidx<nri; iidx++)
1163     {
1164         /* Load shift vector for this list */
1165         i_shift_offset   = DIM*shiftidx[iidx];
1166
1167         /* Load limits for loop over neighbors */
1168         j_index_start    = jindex[iidx];
1169         j_index_end      = jindex[iidx+1];
1170
1171         /* Get outer coordinate index */
1172         inr              = iinr[iidx];
1173         i_coord_offset   = DIM*inr;
1174
1175         /* Load i particle coords and add shift vector */
1176         gmx_mm256_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1177                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1178
1179         fix0             = _mm256_setzero_pd();
1180         fiy0             = _mm256_setzero_pd();
1181         fiz0             = _mm256_setzero_pd();
1182         fix1             = _mm256_setzero_pd();
1183         fiy1             = _mm256_setzero_pd();
1184         fiz1             = _mm256_setzero_pd();
1185         fix2             = _mm256_setzero_pd();
1186         fiy2             = _mm256_setzero_pd();
1187         fiz2             = _mm256_setzero_pd();
1188
1189         /* Start inner kernel loop */
1190         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
1191         {
1192
1193             /* Get j neighbor index, and coordinate index */
1194             jnrA             = jjnr[jidx];
1195             jnrB             = jjnr[jidx+1];
1196             jnrC             = jjnr[jidx+2];
1197             jnrD             = jjnr[jidx+3];
1198             j_coord_offsetA  = DIM*jnrA;
1199             j_coord_offsetB  = DIM*jnrB;
1200             j_coord_offsetC  = DIM*jnrC;
1201             j_coord_offsetD  = DIM*jnrD;
1202
1203             /* load j atom coordinates */
1204             gmx_mm256_load_3rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1205                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1206                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1207
1208             /* Calculate displacement vector */
1209             dx00             = _mm256_sub_pd(ix0,jx0);
1210             dy00             = _mm256_sub_pd(iy0,jy0);
1211             dz00             = _mm256_sub_pd(iz0,jz0);
1212             dx01             = _mm256_sub_pd(ix0,jx1);
1213             dy01             = _mm256_sub_pd(iy0,jy1);
1214             dz01             = _mm256_sub_pd(iz0,jz1);
1215             dx02             = _mm256_sub_pd(ix0,jx2);
1216             dy02             = _mm256_sub_pd(iy0,jy2);
1217             dz02             = _mm256_sub_pd(iz0,jz2);
1218             dx10             = _mm256_sub_pd(ix1,jx0);
1219             dy10             = _mm256_sub_pd(iy1,jy0);
1220             dz10             = _mm256_sub_pd(iz1,jz0);
1221             dx11             = _mm256_sub_pd(ix1,jx1);
1222             dy11             = _mm256_sub_pd(iy1,jy1);
1223             dz11             = _mm256_sub_pd(iz1,jz1);
1224             dx12             = _mm256_sub_pd(ix1,jx2);
1225             dy12             = _mm256_sub_pd(iy1,jy2);
1226             dz12             = _mm256_sub_pd(iz1,jz2);
1227             dx20             = _mm256_sub_pd(ix2,jx0);
1228             dy20             = _mm256_sub_pd(iy2,jy0);
1229             dz20             = _mm256_sub_pd(iz2,jz0);
1230             dx21             = _mm256_sub_pd(ix2,jx1);
1231             dy21             = _mm256_sub_pd(iy2,jy1);
1232             dz21             = _mm256_sub_pd(iz2,jz1);
1233             dx22             = _mm256_sub_pd(ix2,jx2);
1234             dy22             = _mm256_sub_pd(iy2,jy2);
1235             dz22             = _mm256_sub_pd(iz2,jz2);
1236
1237             /* Calculate squared distance and things based on it */
1238             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
1239             rsq01            = gmx_mm256_calc_rsq_pd(dx01,dy01,dz01);
1240             rsq02            = gmx_mm256_calc_rsq_pd(dx02,dy02,dz02);
1241             rsq10            = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
1242             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
1243             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
1244             rsq20            = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
1245             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
1246             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
1247
1248             rinv00           = gmx_mm256_invsqrt_pd(rsq00);
1249             rinv01           = gmx_mm256_invsqrt_pd(rsq01);
1250             rinv02           = gmx_mm256_invsqrt_pd(rsq02);
1251             rinv10           = gmx_mm256_invsqrt_pd(rsq10);
1252             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
1253             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
1254             rinv20           = gmx_mm256_invsqrt_pd(rsq20);
1255             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
1256             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
1257
1258             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
1259             rinvsq01         = _mm256_mul_pd(rinv01,rinv01);
1260             rinvsq02         = _mm256_mul_pd(rinv02,rinv02);
1261             rinvsq10         = _mm256_mul_pd(rinv10,rinv10);
1262             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
1263             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
1264             rinvsq20         = _mm256_mul_pd(rinv20,rinv20);
1265             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
1266             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
1267
1268             fjx0             = _mm256_setzero_pd();
1269             fjy0             = _mm256_setzero_pd();
1270             fjz0             = _mm256_setzero_pd();
1271             fjx1             = _mm256_setzero_pd();
1272             fjy1             = _mm256_setzero_pd();
1273             fjz1             = _mm256_setzero_pd();
1274             fjx2             = _mm256_setzero_pd();
1275             fjy2             = _mm256_setzero_pd();
1276             fjz2             = _mm256_setzero_pd();
1277
1278             /**************************
1279              * CALCULATE INTERACTIONS *
1280              **************************/
1281
1282             r00              = _mm256_mul_pd(rsq00,rinv00);
1283
1284             /* Calculate table index by multiplying r with table scale and truncate to integer */
1285             rt               = _mm256_mul_pd(r00,vftabscale);
1286             vfitab           = _mm256_cvttpd_epi32(rt);
1287             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
1288             vfitab           = _mm_slli_epi32(vfitab,3);
1289
1290             /* REACTION-FIELD ELECTROSTATICS */
1291             felec            = _mm256_mul_pd(qq00,_mm256_sub_pd(_mm256_mul_pd(rinv00,rinvsq00),krf2));
1292
1293             /* CUBIC SPLINE TABLE DISPERSION */
1294             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1295             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1296             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1297             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1298             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1299             Heps             = _mm256_mul_pd(vfeps,H);
1300             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1301             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1302             fvdw6            = _mm256_mul_pd(c6_00,FF);
1303
1304             /* CUBIC SPLINE TABLE REPULSION */
1305             vfitab           = _mm_add_epi32(vfitab,ifour);
1306             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1307             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1308             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1309             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1310             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1311             Heps             = _mm256_mul_pd(vfeps,H);
1312             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1313             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1314             fvdw12           = _mm256_mul_pd(c12_00,FF);
1315             fvdw             = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_add_pd(fvdw6,fvdw12),_mm256_mul_pd(vftabscale,rinv00)));
1316
1317             fscal            = _mm256_add_pd(felec,fvdw);
1318
1319             /* Calculate temporary vectorial force */
1320             tx               = _mm256_mul_pd(fscal,dx00);
1321             ty               = _mm256_mul_pd(fscal,dy00);
1322             tz               = _mm256_mul_pd(fscal,dz00);
1323
1324             /* Update vectorial force */
1325             fix0             = _mm256_add_pd(fix0,tx);
1326             fiy0             = _mm256_add_pd(fiy0,ty);
1327             fiz0             = _mm256_add_pd(fiz0,tz);
1328
1329             fjx0             = _mm256_add_pd(fjx0,tx);
1330             fjy0             = _mm256_add_pd(fjy0,ty);
1331             fjz0             = _mm256_add_pd(fjz0,tz);
1332
1333             /**************************
1334              * CALCULATE INTERACTIONS *
1335              **************************/
1336
1337             /* REACTION-FIELD ELECTROSTATICS */
1338             felec            = _mm256_mul_pd(qq01,_mm256_sub_pd(_mm256_mul_pd(rinv01,rinvsq01),krf2));
1339
1340             fscal            = felec;
1341
1342             /* Calculate temporary vectorial force */
1343             tx               = _mm256_mul_pd(fscal,dx01);
1344             ty               = _mm256_mul_pd(fscal,dy01);
1345             tz               = _mm256_mul_pd(fscal,dz01);
1346
1347             /* Update vectorial force */
1348             fix0             = _mm256_add_pd(fix0,tx);
1349             fiy0             = _mm256_add_pd(fiy0,ty);
1350             fiz0             = _mm256_add_pd(fiz0,tz);
1351
1352             fjx1             = _mm256_add_pd(fjx1,tx);
1353             fjy1             = _mm256_add_pd(fjy1,ty);
1354             fjz1             = _mm256_add_pd(fjz1,tz);
1355
1356             /**************************
1357              * CALCULATE INTERACTIONS *
1358              **************************/
1359
1360             /* REACTION-FIELD ELECTROSTATICS */
1361             felec            = _mm256_mul_pd(qq02,_mm256_sub_pd(_mm256_mul_pd(rinv02,rinvsq02),krf2));
1362
1363             fscal            = felec;
1364
1365             /* Calculate temporary vectorial force */
1366             tx               = _mm256_mul_pd(fscal,dx02);
1367             ty               = _mm256_mul_pd(fscal,dy02);
1368             tz               = _mm256_mul_pd(fscal,dz02);
1369
1370             /* Update vectorial force */
1371             fix0             = _mm256_add_pd(fix0,tx);
1372             fiy0             = _mm256_add_pd(fiy0,ty);
1373             fiz0             = _mm256_add_pd(fiz0,tz);
1374
1375             fjx2             = _mm256_add_pd(fjx2,tx);
1376             fjy2             = _mm256_add_pd(fjy2,ty);
1377             fjz2             = _mm256_add_pd(fjz2,tz);
1378
1379             /**************************
1380              * CALCULATE INTERACTIONS *
1381              **************************/
1382
1383             /* REACTION-FIELD ELECTROSTATICS */
1384             felec            = _mm256_mul_pd(qq10,_mm256_sub_pd(_mm256_mul_pd(rinv10,rinvsq10),krf2));
1385
1386             fscal            = felec;
1387
1388             /* Calculate temporary vectorial force */
1389             tx               = _mm256_mul_pd(fscal,dx10);
1390             ty               = _mm256_mul_pd(fscal,dy10);
1391             tz               = _mm256_mul_pd(fscal,dz10);
1392
1393             /* Update vectorial force */
1394             fix1             = _mm256_add_pd(fix1,tx);
1395             fiy1             = _mm256_add_pd(fiy1,ty);
1396             fiz1             = _mm256_add_pd(fiz1,tz);
1397
1398             fjx0             = _mm256_add_pd(fjx0,tx);
1399             fjy0             = _mm256_add_pd(fjy0,ty);
1400             fjz0             = _mm256_add_pd(fjz0,tz);
1401
1402             /**************************
1403              * CALCULATE INTERACTIONS *
1404              **************************/
1405
1406             /* REACTION-FIELD ELECTROSTATICS */
1407             felec            = _mm256_mul_pd(qq11,_mm256_sub_pd(_mm256_mul_pd(rinv11,rinvsq11),krf2));
1408
1409             fscal            = felec;
1410
1411             /* Calculate temporary vectorial force */
1412             tx               = _mm256_mul_pd(fscal,dx11);
1413             ty               = _mm256_mul_pd(fscal,dy11);
1414             tz               = _mm256_mul_pd(fscal,dz11);
1415
1416             /* Update vectorial force */
1417             fix1             = _mm256_add_pd(fix1,tx);
1418             fiy1             = _mm256_add_pd(fiy1,ty);
1419             fiz1             = _mm256_add_pd(fiz1,tz);
1420
1421             fjx1             = _mm256_add_pd(fjx1,tx);
1422             fjy1             = _mm256_add_pd(fjy1,ty);
1423             fjz1             = _mm256_add_pd(fjz1,tz);
1424
1425             /**************************
1426              * CALCULATE INTERACTIONS *
1427              **************************/
1428
1429             /* REACTION-FIELD ELECTROSTATICS */
1430             felec            = _mm256_mul_pd(qq12,_mm256_sub_pd(_mm256_mul_pd(rinv12,rinvsq12),krf2));
1431
1432             fscal            = felec;
1433
1434             /* Calculate temporary vectorial force */
1435             tx               = _mm256_mul_pd(fscal,dx12);
1436             ty               = _mm256_mul_pd(fscal,dy12);
1437             tz               = _mm256_mul_pd(fscal,dz12);
1438
1439             /* Update vectorial force */
1440             fix1             = _mm256_add_pd(fix1,tx);
1441             fiy1             = _mm256_add_pd(fiy1,ty);
1442             fiz1             = _mm256_add_pd(fiz1,tz);
1443
1444             fjx2             = _mm256_add_pd(fjx2,tx);
1445             fjy2             = _mm256_add_pd(fjy2,ty);
1446             fjz2             = _mm256_add_pd(fjz2,tz);
1447
1448             /**************************
1449              * CALCULATE INTERACTIONS *
1450              **************************/
1451
1452             /* REACTION-FIELD ELECTROSTATICS */
1453             felec            = _mm256_mul_pd(qq20,_mm256_sub_pd(_mm256_mul_pd(rinv20,rinvsq20),krf2));
1454
1455             fscal            = felec;
1456
1457             /* Calculate temporary vectorial force */
1458             tx               = _mm256_mul_pd(fscal,dx20);
1459             ty               = _mm256_mul_pd(fscal,dy20);
1460             tz               = _mm256_mul_pd(fscal,dz20);
1461
1462             /* Update vectorial force */
1463             fix2             = _mm256_add_pd(fix2,tx);
1464             fiy2             = _mm256_add_pd(fiy2,ty);
1465             fiz2             = _mm256_add_pd(fiz2,tz);
1466
1467             fjx0             = _mm256_add_pd(fjx0,tx);
1468             fjy0             = _mm256_add_pd(fjy0,ty);
1469             fjz0             = _mm256_add_pd(fjz0,tz);
1470
1471             /**************************
1472              * CALCULATE INTERACTIONS *
1473              **************************/
1474
1475             /* REACTION-FIELD ELECTROSTATICS */
1476             felec            = _mm256_mul_pd(qq21,_mm256_sub_pd(_mm256_mul_pd(rinv21,rinvsq21),krf2));
1477
1478             fscal            = felec;
1479
1480             /* Calculate temporary vectorial force */
1481             tx               = _mm256_mul_pd(fscal,dx21);
1482             ty               = _mm256_mul_pd(fscal,dy21);
1483             tz               = _mm256_mul_pd(fscal,dz21);
1484
1485             /* Update vectorial force */
1486             fix2             = _mm256_add_pd(fix2,tx);
1487             fiy2             = _mm256_add_pd(fiy2,ty);
1488             fiz2             = _mm256_add_pd(fiz2,tz);
1489
1490             fjx1             = _mm256_add_pd(fjx1,tx);
1491             fjy1             = _mm256_add_pd(fjy1,ty);
1492             fjz1             = _mm256_add_pd(fjz1,tz);
1493
1494             /**************************
1495              * CALCULATE INTERACTIONS *
1496              **************************/
1497
1498             /* REACTION-FIELD ELECTROSTATICS */
1499             felec            = _mm256_mul_pd(qq22,_mm256_sub_pd(_mm256_mul_pd(rinv22,rinvsq22),krf2));
1500
1501             fscal            = felec;
1502
1503             /* Calculate temporary vectorial force */
1504             tx               = _mm256_mul_pd(fscal,dx22);
1505             ty               = _mm256_mul_pd(fscal,dy22);
1506             tz               = _mm256_mul_pd(fscal,dz22);
1507
1508             /* Update vectorial force */
1509             fix2             = _mm256_add_pd(fix2,tx);
1510             fiy2             = _mm256_add_pd(fiy2,ty);
1511             fiz2             = _mm256_add_pd(fiz2,tz);
1512
1513             fjx2             = _mm256_add_pd(fjx2,tx);
1514             fjy2             = _mm256_add_pd(fjy2,ty);
1515             fjz2             = _mm256_add_pd(fjz2,tz);
1516
1517             fjptrA             = f+j_coord_offsetA;
1518             fjptrB             = f+j_coord_offsetB;
1519             fjptrC             = f+j_coord_offsetC;
1520             fjptrD             = f+j_coord_offsetD;
1521
1522             gmx_mm256_decrement_3rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1523                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1524
1525             /* Inner loop uses 270 flops */
1526         }
1527
1528         if(jidx<j_index_end)
1529         {
1530
1531             /* Get j neighbor index, and coordinate index */
1532             jnrlistA         = jjnr[jidx];
1533             jnrlistB         = jjnr[jidx+1];
1534             jnrlistC         = jjnr[jidx+2];
1535             jnrlistD         = jjnr[jidx+3];
1536             /* Sign of each element will be negative for non-real atoms.
1537              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1538              * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
1539              */
1540             tmpmask0 = gmx_mm_castsi128_pd(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
1541
1542             tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
1543             tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
1544             dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
1545
1546             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1547             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1548             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1549             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1550             j_coord_offsetA  = DIM*jnrA;
1551             j_coord_offsetB  = DIM*jnrB;
1552             j_coord_offsetC  = DIM*jnrC;
1553             j_coord_offsetD  = DIM*jnrD;
1554
1555             /* load j atom coordinates */
1556             gmx_mm256_load_3rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1557                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1558                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1559
1560             /* Calculate displacement vector */
1561             dx00             = _mm256_sub_pd(ix0,jx0);
1562             dy00             = _mm256_sub_pd(iy0,jy0);
1563             dz00             = _mm256_sub_pd(iz0,jz0);
1564             dx01             = _mm256_sub_pd(ix0,jx1);
1565             dy01             = _mm256_sub_pd(iy0,jy1);
1566             dz01             = _mm256_sub_pd(iz0,jz1);
1567             dx02             = _mm256_sub_pd(ix0,jx2);
1568             dy02             = _mm256_sub_pd(iy0,jy2);
1569             dz02             = _mm256_sub_pd(iz0,jz2);
1570             dx10             = _mm256_sub_pd(ix1,jx0);
1571             dy10             = _mm256_sub_pd(iy1,jy0);
1572             dz10             = _mm256_sub_pd(iz1,jz0);
1573             dx11             = _mm256_sub_pd(ix1,jx1);
1574             dy11             = _mm256_sub_pd(iy1,jy1);
1575             dz11             = _mm256_sub_pd(iz1,jz1);
1576             dx12             = _mm256_sub_pd(ix1,jx2);
1577             dy12             = _mm256_sub_pd(iy1,jy2);
1578             dz12             = _mm256_sub_pd(iz1,jz2);
1579             dx20             = _mm256_sub_pd(ix2,jx0);
1580             dy20             = _mm256_sub_pd(iy2,jy0);
1581             dz20             = _mm256_sub_pd(iz2,jz0);
1582             dx21             = _mm256_sub_pd(ix2,jx1);
1583             dy21             = _mm256_sub_pd(iy2,jy1);
1584             dz21             = _mm256_sub_pd(iz2,jz1);
1585             dx22             = _mm256_sub_pd(ix2,jx2);
1586             dy22             = _mm256_sub_pd(iy2,jy2);
1587             dz22             = _mm256_sub_pd(iz2,jz2);
1588
1589             /* Calculate squared distance and things based on it */
1590             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
1591             rsq01            = gmx_mm256_calc_rsq_pd(dx01,dy01,dz01);
1592             rsq02            = gmx_mm256_calc_rsq_pd(dx02,dy02,dz02);
1593             rsq10            = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
1594             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
1595             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
1596             rsq20            = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
1597             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
1598             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
1599
1600             rinv00           = gmx_mm256_invsqrt_pd(rsq00);
1601             rinv01           = gmx_mm256_invsqrt_pd(rsq01);
1602             rinv02           = gmx_mm256_invsqrt_pd(rsq02);
1603             rinv10           = gmx_mm256_invsqrt_pd(rsq10);
1604             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
1605             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
1606             rinv20           = gmx_mm256_invsqrt_pd(rsq20);
1607             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
1608             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
1609
1610             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
1611             rinvsq01         = _mm256_mul_pd(rinv01,rinv01);
1612             rinvsq02         = _mm256_mul_pd(rinv02,rinv02);
1613             rinvsq10         = _mm256_mul_pd(rinv10,rinv10);
1614             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
1615             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
1616             rinvsq20         = _mm256_mul_pd(rinv20,rinv20);
1617             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
1618             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
1619
1620             fjx0             = _mm256_setzero_pd();
1621             fjy0             = _mm256_setzero_pd();
1622             fjz0             = _mm256_setzero_pd();
1623             fjx1             = _mm256_setzero_pd();
1624             fjy1             = _mm256_setzero_pd();
1625             fjz1             = _mm256_setzero_pd();
1626             fjx2             = _mm256_setzero_pd();
1627             fjy2             = _mm256_setzero_pd();
1628             fjz2             = _mm256_setzero_pd();
1629
1630             /**************************
1631              * CALCULATE INTERACTIONS *
1632              **************************/
1633
1634             r00              = _mm256_mul_pd(rsq00,rinv00);
1635             r00              = _mm256_andnot_pd(dummy_mask,r00);
1636
1637             /* Calculate table index by multiplying r with table scale and truncate to integer */
1638             rt               = _mm256_mul_pd(r00,vftabscale);
1639             vfitab           = _mm256_cvttpd_epi32(rt);
1640             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
1641             vfitab           = _mm_slli_epi32(vfitab,3);
1642
1643             /* REACTION-FIELD ELECTROSTATICS */
1644             felec            = _mm256_mul_pd(qq00,_mm256_sub_pd(_mm256_mul_pd(rinv00,rinvsq00),krf2));
1645
1646             /* CUBIC SPLINE TABLE DISPERSION */
1647             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1648             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1649             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1650             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1651             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1652             Heps             = _mm256_mul_pd(vfeps,H);
1653             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1654             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1655             fvdw6            = _mm256_mul_pd(c6_00,FF);
1656
1657             /* CUBIC SPLINE TABLE REPULSION */
1658             vfitab           = _mm_add_epi32(vfitab,ifour);
1659             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1660             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1661             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1662             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1663             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1664             Heps             = _mm256_mul_pd(vfeps,H);
1665             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1666             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1667             fvdw12           = _mm256_mul_pd(c12_00,FF);
1668             fvdw             = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_add_pd(fvdw6,fvdw12),_mm256_mul_pd(vftabscale,rinv00)));
1669
1670             fscal            = _mm256_add_pd(felec,fvdw);
1671
1672             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1673
1674             /* Calculate temporary vectorial force */
1675             tx               = _mm256_mul_pd(fscal,dx00);
1676             ty               = _mm256_mul_pd(fscal,dy00);
1677             tz               = _mm256_mul_pd(fscal,dz00);
1678
1679             /* Update vectorial force */
1680             fix0             = _mm256_add_pd(fix0,tx);
1681             fiy0             = _mm256_add_pd(fiy0,ty);
1682             fiz0             = _mm256_add_pd(fiz0,tz);
1683
1684             fjx0             = _mm256_add_pd(fjx0,tx);
1685             fjy0             = _mm256_add_pd(fjy0,ty);
1686             fjz0             = _mm256_add_pd(fjz0,tz);
1687
1688             /**************************
1689              * CALCULATE INTERACTIONS *
1690              **************************/
1691
1692             /* REACTION-FIELD ELECTROSTATICS */
1693             felec            = _mm256_mul_pd(qq01,_mm256_sub_pd(_mm256_mul_pd(rinv01,rinvsq01),krf2));
1694
1695             fscal            = felec;
1696
1697             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1698
1699             /* Calculate temporary vectorial force */
1700             tx               = _mm256_mul_pd(fscal,dx01);
1701             ty               = _mm256_mul_pd(fscal,dy01);
1702             tz               = _mm256_mul_pd(fscal,dz01);
1703
1704             /* Update vectorial force */
1705             fix0             = _mm256_add_pd(fix0,tx);
1706             fiy0             = _mm256_add_pd(fiy0,ty);
1707             fiz0             = _mm256_add_pd(fiz0,tz);
1708
1709             fjx1             = _mm256_add_pd(fjx1,tx);
1710             fjy1             = _mm256_add_pd(fjy1,ty);
1711             fjz1             = _mm256_add_pd(fjz1,tz);
1712
1713             /**************************
1714              * CALCULATE INTERACTIONS *
1715              **************************/
1716
1717             /* REACTION-FIELD ELECTROSTATICS */
1718             felec            = _mm256_mul_pd(qq02,_mm256_sub_pd(_mm256_mul_pd(rinv02,rinvsq02),krf2));
1719
1720             fscal            = felec;
1721
1722             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1723
1724             /* Calculate temporary vectorial force */
1725             tx               = _mm256_mul_pd(fscal,dx02);
1726             ty               = _mm256_mul_pd(fscal,dy02);
1727             tz               = _mm256_mul_pd(fscal,dz02);
1728
1729             /* Update vectorial force */
1730             fix0             = _mm256_add_pd(fix0,tx);
1731             fiy0             = _mm256_add_pd(fiy0,ty);
1732             fiz0             = _mm256_add_pd(fiz0,tz);
1733
1734             fjx2             = _mm256_add_pd(fjx2,tx);
1735             fjy2             = _mm256_add_pd(fjy2,ty);
1736             fjz2             = _mm256_add_pd(fjz2,tz);
1737
1738             /**************************
1739              * CALCULATE INTERACTIONS *
1740              **************************/
1741
1742             /* REACTION-FIELD ELECTROSTATICS */
1743             felec            = _mm256_mul_pd(qq10,_mm256_sub_pd(_mm256_mul_pd(rinv10,rinvsq10),krf2));
1744
1745             fscal            = felec;
1746
1747             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1748
1749             /* Calculate temporary vectorial force */
1750             tx               = _mm256_mul_pd(fscal,dx10);
1751             ty               = _mm256_mul_pd(fscal,dy10);
1752             tz               = _mm256_mul_pd(fscal,dz10);
1753
1754             /* Update vectorial force */
1755             fix1             = _mm256_add_pd(fix1,tx);
1756             fiy1             = _mm256_add_pd(fiy1,ty);
1757             fiz1             = _mm256_add_pd(fiz1,tz);
1758
1759             fjx0             = _mm256_add_pd(fjx0,tx);
1760             fjy0             = _mm256_add_pd(fjy0,ty);
1761             fjz0             = _mm256_add_pd(fjz0,tz);
1762
1763             /**************************
1764              * CALCULATE INTERACTIONS *
1765              **************************/
1766
1767             /* REACTION-FIELD ELECTROSTATICS */
1768             felec            = _mm256_mul_pd(qq11,_mm256_sub_pd(_mm256_mul_pd(rinv11,rinvsq11),krf2));
1769
1770             fscal            = felec;
1771
1772             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1773
1774             /* Calculate temporary vectorial force */
1775             tx               = _mm256_mul_pd(fscal,dx11);
1776             ty               = _mm256_mul_pd(fscal,dy11);
1777             tz               = _mm256_mul_pd(fscal,dz11);
1778
1779             /* Update vectorial force */
1780             fix1             = _mm256_add_pd(fix1,tx);
1781             fiy1             = _mm256_add_pd(fiy1,ty);
1782             fiz1             = _mm256_add_pd(fiz1,tz);
1783
1784             fjx1             = _mm256_add_pd(fjx1,tx);
1785             fjy1             = _mm256_add_pd(fjy1,ty);
1786             fjz1             = _mm256_add_pd(fjz1,tz);
1787
1788             /**************************
1789              * CALCULATE INTERACTIONS *
1790              **************************/
1791
1792             /* REACTION-FIELD ELECTROSTATICS */
1793             felec            = _mm256_mul_pd(qq12,_mm256_sub_pd(_mm256_mul_pd(rinv12,rinvsq12),krf2));
1794
1795             fscal            = felec;
1796
1797             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1798
1799             /* Calculate temporary vectorial force */
1800             tx               = _mm256_mul_pd(fscal,dx12);
1801             ty               = _mm256_mul_pd(fscal,dy12);
1802             tz               = _mm256_mul_pd(fscal,dz12);
1803
1804             /* Update vectorial force */
1805             fix1             = _mm256_add_pd(fix1,tx);
1806             fiy1             = _mm256_add_pd(fiy1,ty);
1807             fiz1             = _mm256_add_pd(fiz1,tz);
1808
1809             fjx2             = _mm256_add_pd(fjx2,tx);
1810             fjy2             = _mm256_add_pd(fjy2,ty);
1811             fjz2             = _mm256_add_pd(fjz2,tz);
1812
1813             /**************************
1814              * CALCULATE INTERACTIONS *
1815              **************************/
1816
1817             /* REACTION-FIELD ELECTROSTATICS */
1818             felec            = _mm256_mul_pd(qq20,_mm256_sub_pd(_mm256_mul_pd(rinv20,rinvsq20),krf2));
1819
1820             fscal            = felec;
1821
1822             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1823
1824             /* Calculate temporary vectorial force */
1825             tx               = _mm256_mul_pd(fscal,dx20);
1826             ty               = _mm256_mul_pd(fscal,dy20);
1827             tz               = _mm256_mul_pd(fscal,dz20);
1828
1829             /* Update vectorial force */
1830             fix2             = _mm256_add_pd(fix2,tx);
1831             fiy2             = _mm256_add_pd(fiy2,ty);
1832             fiz2             = _mm256_add_pd(fiz2,tz);
1833
1834             fjx0             = _mm256_add_pd(fjx0,tx);
1835             fjy0             = _mm256_add_pd(fjy0,ty);
1836             fjz0             = _mm256_add_pd(fjz0,tz);
1837
1838             /**************************
1839              * CALCULATE INTERACTIONS *
1840              **************************/
1841
1842             /* REACTION-FIELD ELECTROSTATICS */
1843             felec            = _mm256_mul_pd(qq21,_mm256_sub_pd(_mm256_mul_pd(rinv21,rinvsq21),krf2));
1844
1845             fscal            = felec;
1846
1847             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1848
1849             /* Calculate temporary vectorial force */
1850             tx               = _mm256_mul_pd(fscal,dx21);
1851             ty               = _mm256_mul_pd(fscal,dy21);
1852             tz               = _mm256_mul_pd(fscal,dz21);
1853
1854             /* Update vectorial force */
1855             fix2             = _mm256_add_pd(fix2,tx);
1856             fiy2             = _mm256_add_pd(fiy2,ty);
1857             fiz2             = _mm256_add_pd(fiz2,tz);
1858
1859             fjx1             = _mm256_add_pd(fjx1,tx);
1860             fjy1             = _mm256_add_pd(fjy1,ty);
1861             fjz1             = _mm256_add_pd(fjz1,tz);
1862
1863             /**************************
1864              * CALCULATE INTERACTIONS *
1865              **************************/
1866
1867             /* REACTION-FIELD ELECTROSTATICS */
1868             felec            = _mm256_mul_pd(qq22,_mm256_sub_pd(_mm256_mul_pd(rinv22,rinvsq22),krf2));
1869
1870             fscal            = felec;
1871
1872             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1873
1874             /* Calculate temporary vectorial force */
1875             tx               = _mm256_mul_pd(fscal,dx22);
1876             ty               = _mm256_mul_pd(fscal,dy22);
1877             tz               = _mm256_mul_pd(fscal,dz22);
1878
1879             /* Update vectorial force */
1880             fix2             = _mm256_add_pd(fix2,tx);
1881             fiy2             = _mm256_add_pd(fiy2,ty);
1882             fiz2             = _mm256_add_pd(fiz2,tz);
1883
1884             fjx2             = _mm256_add_pd(fjx2,tx);
1885             fjy2             = _mm256_add_pd(fjy2,ty);
1886             fjz2             = _mm256_add_pd(fjz2,tz);
1887
1888             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1889             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1890             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1891             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1892
1893             gmx_mm256_decrement_3rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1894                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1895
1896             /* Inner loop uses 271 flops */
1897         }
1898
1899         /* End of innermost loop */
1900
1901         gmx_mm256_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1902                                                  f+i_coord_offset,fshift+i_shift_offset);
1903
1904         /* Increment number of inner iterations */
1905         inneriter                  += j_index_end - j_index_start;
1906
1907         /* Outer loop uses 18 flops */
1908     }
1909
1910     /* Increment number of outer iterations */
1911     outeriter        += nri;
1912
1913     /* Update outer/inner flops */
1914
1915     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_F,outeriter*18 + inneriter*271);
1916 }