src/gromacs/gmxlib/nonbonded/nb_kernel_avx_256_single/nb_kernel_ElecRFCut_VdwNone_GeomW4W4_avx_256_single.c

   1 /*
   2  * Note: this file was generated by the Gromacs avx_256_single 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_single.h"
  34 #include "kernelutil_x86_avx_256_single.h"
  35
  36 /*
  37  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwNone_GeomW4W4_VF_avx_256_single
  38  * Electrostatics interaction: ReactionField
  39  * VdW interaction:            None
  40  * Geometry:                   Water4-Water4
  41  * Calculate force/pot:        PotentialAndForce
  42  */
  43 void
  44 nb_kernel_ElecRFCut_VdwNone_GeomW4W4_VF_avx_256_single
  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,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight 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              jnrE,jnrF,jnrG,jnrH;
  62     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  63     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
  64     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  65     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
  66     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  67     real             rcutoff_scalar;
  68     real             *shiftvec,*fshift,*x,*f;
  69     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
  70     real             scratch[4*DIM];
  71     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  72     real *           vdwioffsetptr1;
  73     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  74     real *           vdwioffsetptr2;
  75     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  76     real *           vdwioffsetptr3;
  77     __m256           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  78     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
  79     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  80     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
  81     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  82     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D,vdwjidx3E,vdwjidx3F,vdwjidx3G,vdwjidx3H;
  83     __m256           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
  84     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
  85     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
  86     __m256           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
  87     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
  88     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
  89     __m256           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
  90     __m256           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
  91     __m256           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
  92     __m256           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
  93     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
  94     real             *charge;
  95     __m256           dummy_mask,cutoff_mask;
  96     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
  97     __m256           one     = _mm256_set1_ps(1.0);
  98     __m256           two     = _mm256_set1_ps(2.0);
  99     x                = xx[0];
 100     f                = ff[0];
 101
 102     nri              = nlist->nri;
 103     iinr             = nlist->iinr;
 104     jindex           = nlist->jindex;
 105     jjnr             = nlist->jjnr;
 106     shiftidx         = nlist->shift;
 107     gid              = nlist->gid;
 108     shiftvec         = fr->shift_vec[0];
 109     fshift           = fr->fshift[0];
 110     facel            = _mm256_set1_ps(fr->epsfac);
 111     charge           = mdatoms->chargeA;
 112     krf              = _mm256_set1_ps(fr->ic->k_rf);
 113     krf2             = _mm256_set1_ps(fr->ic->k_rf*2.0);
 114     crf              = _mm256_set1_ps(fr->ic->c_rf);
 115
 116     /* Setup water-specific parameters */
 117     inr              = nlist->iinr[0];
 118     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
 119     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
 120     iq3              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+3]));
 121
 122     jq1              = _mm256_set1_ps(charge[inr+1]);
 123     jq2              = _mm256_set1_ps(charge[inr+2]);
 124     jq3              = _mm256_set1_ps(charge[inr+3]);
 125     qq11             = _mm256_mul_ps(iq1,jq1);
 126     qq12             = _mm256_mul_ps(iq1,jq2);
 127     qq13             = _mm256_mul_ps(iq1,jq3);
 128     qq21             = _mm256_mul_ps(iq2,jq1);
 129     qq22             = _mm256_mul_ps(iq2,jq2);
 130     qq23             = _mm256_mul_ps(iq2,jq3);
 131     qq31             = _mm256_mul_ps(iq3,jq1);
 132     qq32             = _mm256_mul_ps(iq3,jq2);
 133     qq33             = _mm256_mul_ps(iq3,jq3);
 134
 135     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 136     rcutoff_scalar   = fr->rcoulomb;
 137     rcutoff          = _mm256_set1_ps(rcutoff_scalar);
 138     rcutoff2         = _mm256_mul_ps(rcutoff,rcutoff);
 139
 140     /* Avoid stupid compiler warnings */
 141     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
 142     j_coord_offsetA = 0;
 143     j_coord_offsetB = 0;
 144     j_coord_offsetC = 0;
 145     j_coord_offsetD = 0;
 146     j_coord_offsetE = 0;
 147     j_coord_offsetF = 0;
 148     j_coord_offsetG = 0;
 149     j_coord_offsetH = 0;
 150
 151     outeriter        = 0;
 152     inneriter        = 0;
 153
 154     for(iidx=0;iidx<4*DIM;iidx++)
 155     {
 156         scratch[iidx] = 0.0;
 157     }
 158
 159     /* Start outer loop over neighborlists */
 160     for(iidx=0; iidx<nri; iidx++)
 161     {
 162         /* Load shift vector for this list */
 163         i_shift_offset   = DIM*shiftidx[iidx];
 164
 165         /* Load limits for loop over neighbors */
 166         j_index_start    = jindex[iidx];
 167         j_index_end      = jindex[iidx+1];
 168
 169         /* Get outer coordinate index */
 170         inr              = iinr[iidx];
 171         i_coord_offset   = DIM*inr;
 172
 173         /* Load i particle coords and add shift vector */
 174         gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
 175                                                     &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 176
 177         fix1             = _mm256_setzero_ps();
 178         fiy1             = _mm256_setzero_ps();
 179         fiz1             = _mm256_setzero_ps();
 180         fix2             = _mm256_setzero_ps();
 181         fiy2             = _mm256_setzero_ps();
 182         fiz2             = _mm256_setzero_ps();
 183         fix3             = _mm256_setzero_ps();
 184         fiy3             = _mm256_setzero_ps();
 185         fiz3             = _mm256_setzero_ps();
 186
 187         /* Reset potential sums */
 188         velecsum         = _mm256_setzero_ps();
 189
 190         /* Start inner kernel loop */
 191         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
 192         {
 193
 194             /* Get j neighbor index, and coordinate index */
 195             jnrA             = jjnr[jidx];
 196             jnrB             = jjnr[jidx+1];
 197             jnrC             = jjnr[jidx+2];
 198             jnrD             = jjnr[jidx+3];
 199             jnrE             = jjnr[jidx+4];
 200             jnrF             = jjnr[jidx+5];
 201             jnrG             = jjnr[jidx+6];
 202             jnrH             = jjnr[jidx+7];
 203             j_coord_offsetA  = DIM*jnrA;
 204             j_coord_offsetB  = DIM*jnrB;
 205             j_coord_offsetC  = DIM*jnrC;
 206             j_coord_offsetD  = DIM*jnrD;
 207             j_coord_offsetE  = DIM*jnrE;
 208             j_coord_offsetF  = DIM*jnrF;
 209             j_coord_offsetG  = DIM*jnrG;
 210             j_coord_offsetH  = DIM*jnrH;
 211
 212             /* load j atom coordinates */
 213             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
 214                                                  x+j_coord_offsetC+DIM,x+j_coord_offsetD+DIM,
 215                                                  x+j_coord_offsetE+DIM,x+j_coord_offsetF+DIM,
 216                                                  x+j_coord_offsetG+DIM,x+j_coord_offsetH+DIM,
 217                                                  &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
 218
 219             /* Calculate displacement vector */
 220             dx11             = _mm256_sub_ps(ix1,jx1);
 221             dy11             = _mm256_sub_ps(iy1,jy1);
 222             dz11             = _mm256_sub_ps(iz1,jz1);
 223             dx12             = _mm256_sub_ps(ix1,jx2);
 224             dy12             = _mm256_sub_ps(iy1,jy2);
 225             dz12             = _mm256_sub_ps(iz1,jz2);
 226             dx13             = _mm256_sub_ps(ix1,jx3);
 227             dy13             = _mm256_sub_ps(iy1,jy3);
 228             dz13             = _mm256_sub_ps(iz1,jz3);
 229             dx21             = _mm256_sub_ps(ix2,jx1);
 230             dy21             = _mm256_sub_ps(iy2,jy1);
 231             dz21             = _mm256_sub_ps(iz2,jz1);
 232             dx22             = _mm256_sub_ps(ix2,jx2);
 233             dy22             = _mm256_sub_ps(iy2,jy2);
 234             dz22             = _mm256_sub_ps(iz2,jz2);
 235             dx23             = _mm256_sub_ps(ix2,jx3);
 236             dy23             = _mm256_sub_ps(iy2,jy3);
 237             dz23             = _mm256_sub_ps(iz2,jz3);
 238             dx31             = _mm256_sub_ps(ix3,jx1);
 239             dy31             = _mm256_sub_ps(iy3,jy1);
 240             dz31             = _mm256_sub_ps(iz3,jz1);
 241             dx32             = _mm256_sub_ps(ix3,jx2);
 242             dy32             = _mm256_sub_ps(iy3,jy2);
 243             dz32             = _mm256_sub_ps(iz3,jz2);
 244             dx33             = _mm256_sub_ps(ix3,jx3);
 245             dy33             = _mm256_sub_ps(iy3,jy3);
 246             dz33             = _mm256_sub_ps(iz3,jz3);
 247
 248             /* Calculate squared distance and things based on it */
 249             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 250             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 251             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
 252             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 253             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 254             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
 255             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
 256             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
 257             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
 258
 259             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 260             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 261             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
 262             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 263             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 264             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
 265             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
 266             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
 267             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
 268
 269             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 270             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 271             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
 272             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 273             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 274             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
 275             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
 276             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
 277             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
 278
 279             fjx1             = _mm256_setzero_ps();
 280             fjy1             = _mm256_setzero_ps();
 281             fjz1             = _mm256_setzero_ps();
 282             fjx2             = _mm256_setzero_ps();
 283             fjy2             = _mm256_setzero_ps();
 284             fjz2             = _mm256_setzero_ps();
 285             fjx3             = _mm256_setzero_ps();
 286             fjy3             = _mm256_setzero_ps();
 287             fjz3             = _mm256_setzero_ps();
 288
 289             /**************************
 290              * CALCULATE INTERACTIONS *
 291              **************************/
 292
 293             if (gmx_mm256_any_lt(rsq11,rcutoff2))
 294             {
 295
 296             /* REACTION-FIELD ELECTROSTATICS */
 297             velec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_add_ps(rinv11,_mm256_mul_ps(krf,rsq11)),crf));
 298             felec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_mul_ps(rinv11,rinvsq11),krf2));
 299
 300             cutoff_mask      = _mm256_cmp_ps(rsq11,rcutoff2,_CMP_LT_OQ);
 301
 302             /* Update potential sum for this i atom from the interaction with this j atom. */
 303             velec            = _mm256_and_ps(velec,cutoff_mask);
 304             velecsum         = _mm256_add_ps(velecsum,velec);
 305
 306             fscal            = felec;
 307
 308             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 309
 310             /* Calculate temporary vectorial force */
 311             tx               = _mm256_mul_ps(fscal,dx11);
 312             ty               = _mm256_mul_ps(fscal,dy11);
 313             tz               = _mm256_mul_ps(fscal,dz11);
 314
 315             /* Update vectorial force */
 316             fix1             = _mm256_add_ps(fix1,tx);
 317             fiy1             = _mm256_add_ps(fiy1,ty);
 318             fiz1             = _mm256_add_ps(fiz1,tz);
 319
 320             fjx1             = _mm256_add_ps(fjx1,tx);
 321             fjy1             = _mm256_add_ps(fjy1,ty);
 322             fjz1             = _mm256_add_ps(fjz1,tz);
 323
 324             }
 325
 326             /**************************
 327              * CALCULATE INTERACTIONS *
 328              **************************/
 329
 330             if (gmx_mm256_any_lt(rsq12,rcutoff2))
 331             {
 332
 333             /* REACTION-FIELD ELECTROSTATICS */
 334             velec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_add_ps(rinv12,_mm256_mul_ps(krf,rsq12)),crf));
 335             felec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_mul_ps(rinv12,rinvsq12),krf2));
 336
 337             cutoff_mask      = _mm256_cmp_ps(rsq12,rcutoff2,_CMP_LT_OQ);
 338
 339             /* Update potential sum for this i atom from the interaction with this j atom. */
 340             velec            = _mm256_and_ps(velec,cutoff_mask);
 341             velecsum         = _mm256_add_ps(velecsum,velec);
 342
 343             fscal            = felec;
 344
 345             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 346
 347             /* Calculate temporary vectorial force */
 348             tx               = _mm256_mul_ps(fscal,dx12);
 349             ty               = _mm256_mul_ps(fscal,dy12);
 350             tz               = _mm256_mul_ps(fscal,dz12);
 351
 352             /* Update vectorial force */
 353             fix1             = _mm256_add_ps(fix1,tx);
 354             fiy1             = _mm256_add_ps(fiy1,ty);
 355             fiz1             = _mm256_add_ps(fiz1,tz);
 356
 357             fjx2             = _mm256_add_ps(fjx2,tx);
 358             fjy2             = _mm256_add_ps(fjy2,ty);
 359             fjz2             = _mm256_add_ps(fjz2,tz);
 360
 361             }
 362
 363             /**************************
 364              * CALCULATE INTERACTIONS *
 365              **************************/
 366
 367             if (gmx_mm256_any_lt(rsq13,rcutoff2))
 368             {
 369
 370             /* REACTION-FIELD ELECTROSTATICS */
 371             velec            = _mm256_mul_ps(qq13,_mm256_sub_ps(_mm256_add_ps(rinv13,_mm256_mul_ps(krf,rsq13)),crf));
 372             felec            = _mm256_mul_ps(qq13,_mm256_sub_ps(_mm256_mul_ps(rinv13,rinvsq13),krf2));
 373
 374             cutoff_mask      = _mm256_cmp_ps(rsq13,rcutoff2,_CMP_LT_OQ);
 375
 376             /* Update potential sum for this i atom from the interaction with this j atom. */
 377             velec            = _mm256_and_ps(velec,cutoff_mask);
 378             velecsum         = _mm256_add_ps(velecsum,velec);
 379
 380             fscal            = felec;
 381
 382             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 383
 384             /* Calculate temporary vectorial force */
 385             tx               = _mm256_mul_ps(fscal,dx13);
 386             ty               = _mm256_mul_ps(fscal,dy13);
 387             tz               = _mm256_mul_ps(fscal,dz13);
 388
 389             /* Update vectorial force */
 390             fix1             = _mm256_add_ps(fix1,tx);
 391             fiy1             = _mm256_add_ps(fiy1,ty);
 392             fiz1             = _mm256_add_ps(fiz1,tz);
 393
 394             fjx3             = _mm256_add_ps(fjx3,tx);
 395             fjy3             = _mm256_add_ps(fjy3,ty);
 396             fjz3             = _mm256_add_ps(fjz3,tz);
 397
 398             }
 399
 400             /**************************
 401              * CALCULATE INTERACTIONS *
 402              **************************/
 403
 404             if (gmx_mm256_any_lt(rsq21,rcutoff2))
 405             {
 406
 407             /* REACTION-FIELD ELECTROSTATICS */
 408             velec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_add_ps(rinv21,_mm256_mul_ps(krf,rsq21)),crf));
 409             felec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_mul_ps(rinv21,rinvsq21),krf2));
 410
 411             cutoff_mask      = _mm256_cmp_ps(rsq21,rcutoff2,_CMP_LT_OQ);
 412
 413             /* Update potential sum for this i atom from the interaction with this j atom. */
 414             velec            = _mm256_and_ps(velec,cutoff_mask);
 415             velecsum         = _mm256_add_ps(velecsum,velec);
 416
 417             fscal            = felec;
 418
 419             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 420
 421             /* Calculate temporary vectorial force */
 422             tx               = _mm256_mul_ps(fscal,dx21);
 423             ty               = _mm256_mul_ps(fscal,dy21);
 424             tz               = _mm256_mul_ps(fscal,dz21);
 425
 426             /* Update vectorial force */
 427             fix2             = _mm256_add_ps(fix2,tx);
 428             fiy2             = _mm256_add_ps(fiy2,ty);
 429             fiz2             = _mm256_add_ps(fiz2,tz);
 430
 431             fjx1             = _mm256_add_ps(fjx1,tx);
 432             fjy1             = _mm256_add_ps(fjy1,ty);
 433             fjz1             = _mm256_add_ps(fjz1,tz);
 434
 435             }
 436
 437             /**************************
 438              * CALCULATE INTERACTIONS *
 439              **************************/
 440
 441             if (gmx_mm256_any_lt(rsq22,rcutoff2))
 442             {
 443
 444             /* REACTION-FIELD ELECTROSTATICS */
 445             velec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_add_ps(rinv22,_mm256_mul_ps(krf,rsq22)),crf));
 446             felec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_mul_ps(rinv22,rinvsq22),krf2));
 447
 448             cutoff_mask      = _mm256_cmp_ps(rsq22,rcutoff2,_CMP_LT_OQ);
 449
 450             /* Update potential sum for this i atom from the interaction with this j atom. */
 451             velec            = _mm256_and_ps(velec,cutoff_mask);
 452             velecsum         = _mm256_add_ps(velecsum,velec);
 453
 454             fscal            = felec;
 455
 456             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 457
 458             /* Calculate temporary vectorial force */
 459             tx               = _mm256_mul_ps(fscal,dx22);
 460             ty               = _mm256_mul_ps(fscal,dy22);
 461             tz               = _mm256_mul_ps(fscal,dz22);
 462
 463             /* Update vectorial force */
 464             fix2             = _mm256_add_ps(fix2,tx);
 465             fiy2             = _mm256_add_ps(fiy2,ty);
 466             fiz2             = _mm256_add_ps(fiz2,tz);
 467
 468             fjx2             = _mm256_add_ps(fjx2,tx);
 469             fjy2             = _mm256_add_ps(fjy2,ty);
 470             fjz2             = _mm256_add_ps(fjz2,tz);
 471
 472             }
 473
 474             /**************************
 475              * CALCULATE INTERACTIONS *
 476              **************************/
 477
 478             if (gmx_mm256_any_lt(rsq23,rcutoff2))
 479             {
 480
 481             /* REACTION-FIELD ELECTROSTATICS */
 482             velec            = _mm256_mul_ps(qq23,_mm256_sub_ps(_mm256_add_ps(rinv23,_mm256_mul_ps(krf,rsq23)),crf));
 483             felec            = _mm256_mul_ps(qq23,_mm256_sub_ps(_mm256_mul_ps(rinv23,rinvsq23),krf2));
 484
 485             cutoff_mask      = _mm256_cmp_ps(rsq23,rcutoff2,_CMP_LT_OQ);
 486
 487             /* Update potential sum for this i atom from the interaction with this j atom. */
 488             velec            = _mm256_and_ps(velec,cutoff_mask);
 489             velecsum         = _mm256_add_ps(velecsum,velec);
 490
 491             fscal            = felec;
 492
 493             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 494
 495             /* Calculate temporary vectorial force */
 496             tx               = _mm256_mul_ps(fscal,dx23);
 497             ty               = _mm256_mul_ps(fscal,dy23);
 498             tz               = _mm256_mul_ps(fscal,dz23);
 499
 500             /* Update vectorial force */
 501             fix2             = _mm256_add_ps(fix2,tx);
 502             fiy2             = _mm256_add_ps(fiy2,ty);
 503             fiz2             = _mm256_add_ps(fiz2,tz);
 504
 505             fjx3             = _mm256_add_ps(fjx3,tx);
 506             fjy3             = _mm256_add_ps(fjy3,ty);
 507             fjz3             = _mm256_add_ps(fjz3,tz);
 508
 509             }
 510
 511             /**************************
 512              * CALCULATE INTERACTIONS *
 513              **************************/
 514
 515             if (gmx_mm256_any_lt(rsq31,rcutoff2))
 516             {
 517
 518             /* REACTION-FIELD ELECTROSTATICS */
 519             velec            = _mm256_mul_ps(qq31,_mm256_sub_ps(_mm256_add_ps(rinv31,_mm256_mul_ps(krf,rsq31)),crf));
 520             felec            = _mm256_mul_ps(qq31,_mm256_sub_ps(_mm256_mul_ps(rinv31,rinvsq31),krf2));
 521
 522             cutoff_mask      = _mm256_cmp_ps(rsq31,rcutoff2,_CMP_LT_OQ);
 523
 524             /* Update potential sum for this i atom from the interaction with this j atom. */
 525             velec            = _mm256_and_ps(velec,cutoff_mask);
 526             velecsum         = _mm256_add_ps(velecsum,velec);
 527
 528             fscal            = felec;
 529
 530             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 531
 532             /* Calculate temporary vectorial force */
 533             tx               = _mm256_mul_ps(fscal,dx31);
 534             ty               = _mm256_mul_ps(fscal,dy31);
 535             tz               = _mm256_mul_ps(fscal,dz31);
 536
 537             /* Update vectorial force */
 538             fix3             = _mm256_add_ps(fix3,tx);
 539             fiy3             = _mm256_add_ps(fiy3,ty);
 540             fiz3             = _mm256_add_ps(fiz3,tz);
 541
 542             fjx1             = _mm256_add_ps(fjx1,tx);
 543             fjy1             = _mm256_add_ps(fjy1,ty);
 544             fjz1             = _mm256_add_ps(fjz1,tz);
 545
 546             }
 547
 548             /**************************
 549              * CALCULATE INTERACTIONS *
 550              **************************/
 551
 552             if (gmx_mm256_any_lt(rsq32,rcutoff2))
 553             {
 554
 555             /* REACTION-FIELD ELECTROSTATICS */
 556             velec            = _mm256_mul_ps(qq32,_mm256_sub_ps(_mm256_add_ps(rinv32,_mm256_mul_ps(krf,rsq32)),crf));
 557             felec            = _mm256_mul_ps(qq32,_mm256_sub_ps(_mm256_mul_ps(rinv32,rinvsq32),krf2));
 558
 559             cutoff_mask      = _mm256_cmp_ps(rsq32,rcutoff2,_CMP_LT_OQ);
 560
 561             /* Update potential sum for this i atom from the interaction with this j atom. */
 562             velec            = _mm256_and_ps(velec,cutoff_mask);
 563             velecsum         = _mm256_add_ps(velecsum,velec);
 564
 565             fscal            = felec;
 566
 567             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 568
 569             /* Calculate temporary vectorial force */
 570             tx               = _mm256_mul_ps(fscal,dx32);
 571             ty               = _mm256_mul_ps(fscal,dy32);
 572             tz               = _mm256_mul_ps(fscal,dz32);
 573
 574             /* Update vectorial force */
 575             fix3             = _mm256_add_ps(fix3,tx);
 576             fiy3             = _mm256_add_ps(fiy3,ty);
 577             fiz3             = _mm256_add_ps(fiz3,tz);
 578
 579             fjx2             = _mm256_add_ps(fjx2,tx);
 580             fjy2             = _mm256_add_ps(fjy2,ty);
 581             fjz2             = _mm256_add_ps(fjz2,tz);
 582
 583             }
 584
 585             /**************************
 586              * CALCULATE INTERACTIONS *
 587              **************************/
 588
 589             if (gmx_mm256_any_lt(rsq33,rcutoff2))
 590             {
 591
 592             /* REACTION-FIELD ELECTROSTATICS */
 593             velec            = _mm256_mul_ps(qq33,_mm256_sub_ps(_mm256_add_ps(rinv33,_mm256_mul_ps(krf,rsq33)),crf));
 594             felec            = _mm256_mul_ps(qq33,_mm256_sub_ps(_mm256_mul_ps(rinv33,rinvsq33),krf2));
 595
 596             cutoff_mask      = _mm256_cmp_ps(rsq33,rcutoff2,_CMP_LT_OQ);
 597
 598             /* Update potential sum for this i atom from the interaction with this j atom. */
 599             velec            = _mm256_and_ps(velec,cutoff_mask);
 600             velecsum         = _mm256_add_ps(velecsum,velec);
 601
 602             fscal            = felec;
 603
 604             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 605
 606             /* Calculate temporary vectorial force */
 607             tx               = _mm256_mul_ps(fscal,dx33);
 608             ty               = _mm256_mul_ps(fscal,dy33);
 609             tz               = _mm256_mul_ps(fscal,dz33);
 610
 611             /* Update vectorial force */
 612             fix3             = _mm256_add_ps(fix3,tx);
 613             fiy3             = _mm256_add_ps(fiy3,ty);
 614             fiz3             = _mm256_add_ps(fiz3,tz);
 615
 616             fjx3             = _mm256_add_ps(fjx3,tx);
 617             fjy3             = _mm256_add_ps(fjy3,ty);
 618             fjz3             = _mm256_add_ps(fjz3,tz);
 619
 620             }
 621
 622             fjptrA             = f+j_coord_offsetA;
 623             fjptrB             = f+j_coord_offsetB;
 624             fjptrC             = f+j_coord_offsetC;
 625             fjptrD             = f+j_coord_offsetD;
 626             fjptrE             = f+j_coord_offsetE;
 627             fjptrF             = f+j_coord_offsetF;
 628             fjptrG             = f+j_coord_offsetG;
 629             fjptrH             = f+j_coord_offsetH;
 630
 631             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA+DIM,fjptrB+DIM,fjptrC+DIM,fjptrD+DIM,
 632                                                       fjptrE+DIM,fjptrF+DIM,fjptrG+DIM,fjptrH+DIM,
 633                                                       fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 634
 635             /* Inner loop uses 324 flops */
 636         }
 637
 638         if(jidx<j_index_end)
 639         {
 640
 641             /* Get j neighbor index, and coordinate index */
 642             jnrlistA         = jjnr[jidx];
 643             jnrlistB         = jjnr[jidx+1];
 644             jnrlistC         = jjnr[jidx+2];
 645             jnrlistD         = jjnr[jidx+3];
 646             jnrlistE         = jjnr[jidx+4];
 647             jnrlistF         = jjnr[jidx+5];
 648             jnrlistG         = jjnr[jidx+6];
 649             jnrlistH         = jjnr[jidx+7];
 650             /* Sign of each element will be negative for non-real atoms.
 651              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 652              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 653              */
 654             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
 655                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
 656
 657             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 658             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 659             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 660             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 661             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
 662             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
 663             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
 664             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
 665             j_coord_offsetA  = DIM*jnrA;
 666             j_coord_offsetB  = DIM*jnrB;
 667             j_coord_offsetC  = DIM*jnrC;
 668             j_coord_offsetD  = DIM*jnrD;
 669             j_coord_offsetE  = DIM*jnrE;
 670             j_coord_offsetF  = DIM*jnrF;
 671             j_coord_offsetG  = DIM*jnrG;
 672             j_coord_offsetH  = DIM*jnrH;
 673
 674             /* load j atom coordinates */
 675             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
 676                                                  x+j_coord_offsetC+DIM,x+j_coord_offsetD+DIM,
 677                                                  x+j_coord_offsetE+DIM,x+j_coord_offsetF+DIM,
 678                                                  x+j_coord_offsetG+DIM,x+j_coord_offsetH+DIM,
 679                                                  &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
 680
 681             /* Calculate displacement vector */
 682             dx11             = _mm256_sub_ps(ix1,jx1);
 683             dy11             = _mm256_sub_ps(iy1,jy1);
 684             dz11             = _mm256_sub_ps(iz1,jz1);
 685             dx12             = _mm256_sub_ps(ix1,jx2);
 686             dy12             = _mm256_sub_ps(iy1,jy2);
 687             dz12             = _mm256_sub_ps(iz1,jz2);
 688             dx13             = _mm256_sub_ps(ix1,jx3);
 689             dy13             = _mm256_sub_ps(iy1,jy3);
 690             dz13             = _mm256_sub_ps(iz1,jz3);
 691             dx21             = _mm256_sub_ps(ix2,jx1);
 692             dy21             = _mm256_sub_ps(iy2,jy1);
 693             dz21             = _mm256_sub_ps(iz2,jz1);
 694             dx22             = _mm256_sub_ps(ix2,jx2);
 695             dy22             = _mm256_sub_ps(iy2,jy2);
 696             dz22             = _mm256_sub_ps(iz2,jz2);
 697             dx23             = _mm256_sub_ps(ix2,jx3);
 698             dy23             = _mm256_sub_ps(iy2,jy3);
 699             dz23             = _mm256_sub_ps(iz2,jz3);
 700             dx31             = _mm256_sub_ps(ix3,jx1);
 701             dy31             = _mm256_sub_ps(iy3,jy1);
 702             dz31             = _mm256_sub_ps(iz3,jz1);
 703             dx32             = _mm256_sub_ps(ix3,jx2);
 704             dy32             = _mm256_sub_ps(iy3,jy2);
 705             dz32             = _mm256_sub_ps(iz3,jz2);
 706             dx33             = _mm256_sub_ps(ix3,jx3);
 707             dy33             = _mm256_sub_ps(iy3,jy3);
 708             dz33             = _mm256_sub_ps(iz3,jz3);
 709
 710             /* Calculate squared distance and things based on it */
 711             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 712             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 713             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
 714             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 715             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 716             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
 717             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
 718             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
 719             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
 720
 721             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 722             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 723             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
 724             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 725             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 726             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
 727             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
 728             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
 729             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
 730
 731             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 732             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 733             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
 734             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 735             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 736             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
 737             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
 738             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
 739             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
 740
 741             fjx1             = _mm256_setzero_ps();
 742             fjy1             = _mm256_setzero_ps();
 743             fjz1             = _mm256_setzero_ps();
 744             fjx2             = _mm256_setzero_ps();
 745             fjy2             = _mm256_setzero_ps();
 746             fjz2             = _mm256_setzero_ps();
 747             fjx3             = _mm256_setzero_ps();
 748             fjy3             = _mm256_setzero_ps();
 749             fjz3             = _mm256_setzero_ps();
 750
 751             /**************************
 752              * CALCULATE INTERACTIONS *
 753              **************************/
 754
 755             if (gmx_mm256_any_lt(rsq11,rcutoff2))
 756             {
 757
 758             /* REACTION-FIELD ELECTROSTATICS */
 759             velec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_add_ps(rinv11,_mm256_mul_ps(krf,rsq11)),crf));
 760             felec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_mul_ps(rinv11,rinvsq11),krf2));
 761
 762             cutoff_mask      = _mm256_cmp_ps(rsq11,rcutoff2,_CMP_LT_OQ);
 763
 764             /* Update potential sum for this i atom from the interaction with this j atom. */
 765             velec            = _mm256_and_ps(velec,cutoff_mask);
 766             velec            = _mm256_andnot_ps(dummy_mask,velec);
 767             velecsum         = _mm256_add_ps(velecsum,velec);
 768
 769             fscal            = felec;
 770
 771             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 772
 773             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 774
 775             /* Calculate temporary vectorial force */
 776             tx               = _mm256_mul_ps(fscal,dx11);
 777             ty               = _mm256_mul_ps(fscal,dy11);
 778             tz               = _mm256_mul_ps(fscal,dz11);
 779
 780             /* Update vectorial force */
 781             fix1             = _mm256_add_ps(fix1,tx);
 782             fiy1             = _mm256_add_ps(fiy1,ty);
 783             fiz1             = _mm256_add_ps(fiz1,tz);
 784
 785             fjx1             = _mm256_add_ps(fjx1,tx);
 786             fjy1             = _mm256_add_ps(fjy1,ty);
 787             fjz1             = _mm256_add_ps(fjz1,tz);
 788
 789             }
 790
 791             /**************************
 792              * CALCULATE INTERACTIONS *
 793              **************************/
 794
 795             if (gmx_mm256_any_lt(rsq12,rcutoff2))
 796             {
 797
 798             /* REACTION-FIELD ELECTROSTATICS */
 799             velec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_add_ps(rinv12,_mm256_mul_ps(krf,rsq12)),crf));
 800             felec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_mul_ps(rinv12,rinvsq12),krf2));
 801
 802             cutoff_mask      = _mm256_cmp_ps(rsq12,rcutoff2,_CMP_LT_OQ);
 803
 804             /* Update potential sum for this i atom from the interaction with this j atom. */
 805             velec            = _mm256_and_ps(velec,cutoff_mask);
 806             velec            = _mm256_andnot_ps(dummy_mask,velec);
 807             velecsum         = _mm256_add_ps(velecsum,velec);
 808
 809             fscal            = felec;
 810
 811             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 812
 813             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 814
 815             /* Calculate temporary vectorial force */
 816             tx               = _mm256_mul_ps(fscal,dx12);
 817             ty               = _mm256_mul_ps(fscal,dy12);
 818             tz               = _mm256_mul_ps(fscal,dz12);
 819
 820             /* Update vectorial force */
 821             fix1             = _mm256_add_ps(fix1,tx);
 822             fiy1             = _mm256_add_ps(fiy1,ty);
 823             fiz1             = _mm256_add_ps(fiz1,tz);
 824
 825             fjx2             = _mm256_add_ps(fjx2,tx);
 826             fjy2             = _mm256_add_ps(fjy2,ty);
 827             fjz2             = _mm256_add_ps(fjz2,tz);
 828
 829             }
 830
 831             /**************************
 832              * CALCULATE INTERACTIONS *
 833              **************************/
 834
 835             if (gmx_mm256_any_lt(rsq13,rcutoff2))
 836             {
 837
 838             /* REACTION-FIELD ELECTROSTATICS */
 839             velec            = _mm256_mul_ps(qq13,_mm256_sub_ps(_mm256_add_ps(rinv13,_mm256_mul_ps(krf,rsq13)),crf));
 840             felec            = _mm256_mul_ps(qq13,_mm256_sub_ps(_mm256_mul_ps(rinv13,rinvsq13),krf2));
 841
 842             cutoff_mask      = _mm256_cmp_ps(rsq13,rcutoff2,_CMP_LT_OQ);
 843
 844             /* Update potential sum for this i atom from the interaction with this j atom. */
 845             velec            = _mm256_and_ps(velec,cutoff_mask);
 846             velec            = _mm256_andnot_ps(dummy_mask,velec);
 847             velecsum         = _mm256_add_ps(velecsum,velec);
 848
 849             fscal            = felec;
 850
 851             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 852
 853             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 854
 855             /* Calculate temporary vectorial force */
 856             tx               = _mm256_mul_ps(fscal,dx13);
 857             ty               = _mm256_mul_ps(fscal,dy13);
 858             tz               = _mm256_mul_ps(fscal,dz13);
 859
 860             /* Update vectorial force */
 861             fix1             = _mm256_add_ps(fix1,tx);
 862             fiy1             = _mm256_add_ps(fiy1,ty);
 863             fiz1             = _mm256_add_ps(fiz1,tz);
 864
 865             fjx3             = _mm256_add_ps(fjx3,tx);
 866             fjy3             = _mm256_add_ps(fjy3,ty);
 867             fjz3             = _mm256_add_ps(fjz3,tz);
 868
 869             }
 870
 871             /**************************
 872              * CALCULATE INTERACTIONS *
 873              **************************/
 874
 875             if (gmx_mm256_any_lt(rsq21,rcutoff2))
 876             {
 877
 878             /* REACTION-FIELD ELECTROSTATICS */
 879             velec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_add_ps(rinv21,_mm256_mul_ps(krf,rsq21)),crf));
 880             felec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_mul_ps(rinv21,rinvsq21),krf2));
 881
 882             cutoff_mask      = _mm256_cmp_ps(rsq21,rcutoff2,_CMP_LT_OQ);
 883
 884             /* Update potential sum for this i atom from the interaction with this j atom. */
 885             velec            = _mm256_and_ps(velec,cutoff_mask);
 886             velec            = _mm256_andnot_ps(dummy_mask,velec);
 887             velecsum         = _mm256_add_ps(velecsum,velec);
 888
 889             fscal            = felec;
 890
 891             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 892
 893             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 894
 895             /* Calculate temporary vectorial force */
 896             tx               = _mm256_mul_ps(fscal,dx21);
 897             ty               = _mm256_mul_ps(fscal,dy21);
 898             tz               = _mm256_mul_ps(fscal,dz21);
 899
 900             /* Update vectorial force */
 901             fix2             = _mm256_add_ps(fix2,tx);
 902             fiy2             = _mm256_add_ps(fiy2,ty);
 903             fiz2             = _mm256_add_ps(fiz2,tz);
 904
 905             fjx1             = _mm256_add_ps(fjx1,tx);
 906             fjy1             = _mm256_add_ps(fjy1,ty);
 907             fjz1             = _mm256_add_ps(fjz1,tz);
 908
 909             }
 910
 911             /**************************
 912              * CALCULATE INTERACTIONS *
 913              **************************/
 914
 915             if (gmx_mm256_any_lt(rsq22,rcutoff2))
 916             {
 917
 918             /* REACTION-FIELD ELECTROSTATICS */
 919             velec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_add_ps(rinv22,_mm256_mul_ps(krf,rsq22)),crf));
 920             felec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_mul_ps(rinv22,rinvsq22),krf2));
 921
 922             cutoff_mask      = _mm256_cmp_ps(rsq22,rcutoff2,_CMP_LT_OQ);
 923
 924             /* Update potential sum for this i atom from the interaction with this j atom. */
 925             velec            = _mm256_and_ps(velec,cutoff_mask);
 926             velec            = _mm256_andnot_ps(dummy_mask,velec);
 927             velecsum         = _mm256_add_ps(velecsum,velec);
 928
 929             fscal            = felec;
 930
 931             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 932
 933             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 934
 935             /* Calculate temporary vectorial force */
 936             tx               = _mm256_mul_ps(fscal,dx22);
 937             ty               = _mm256_mul_ps(fscal,dy22);
 938             tz               = _mm256_mul_ps(fscal,dz22);
 939
 940             /* Update vectorial force */
 941             fix2             = _mm256_add_ps(fix2,tx);
 942             fiy2             = _mm256_add_ps(fiy2,ty);
 943             fiz2             = _mm256_add_ps(fiz2,tz);
 944
 945             fjx2             = _mm256_add_ps(fjx2,tx);
 946             fjy2             = _mm256_add_ps(fjy2,ty);
 947             fjz2             = _mm256_add_ps(fjz2,tz);
 948
 949             }
 950
 951             /**************************
 952              * CALCULATE INTERACTIONS *
 953              **************************/
 954
 955             if (gmx_mm256_any_lt(rsq23,rcutoff2))
 956             {
 957
 958             /* REACTION-FIELD ELECTROSTATICS */
 959             velec            = _mm256_mul_ps(qq23,_mm256_sub_ps(_mm256_add_ps(rinv23,_mm256_mul_ps(krf,rsq23)),crf));
 960             felec            = _mm256_mul_ps(qq23,_mm256_sub_ps(_mm256_mul_ps(rinv23,rinvsq23),krf2));
 961
 962             cutoff_mask      = _mm256_cmp_ps(rsq23,rcutoff2,_CMP_LT_OQ);
 963
 964             /* Update potential sum for this i atom from the interaction with this j atom. */
 965             velec            = _mm256_and_ps(velec,cutoff_mask);
 966             velec            = _mm256_andnot_ps(dummy_mask,velec);
 967             velecsum         = _mm256_add_ps(velecsum,velec);
 968
 969             fscal            = felec;
 970
 971             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 972
 973             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 974
 975             /* Calculate temporary vectorial force */
 976             tx               = _mm256_mul_ps(fscal,dx23);
 977             ty               = _mm256_mul_ps(fscal,dy23);
 978             tz               = _mm256_mul_ps(fscal,dz23);
 979
 980             /* Update vectorial force */
 981             fix2             = _mm256_add_ps(fix2,tx);
 982             fiy2             = _mm256_add_ps(fiy2,ty);
 983             fiz2             = _mm256_add_ps(fiz2,tz);
 984
 985             fjx3             = _mm256_add_ps(fjx3,tx);
 986             fjy3             = _mm256_add_ps(fjy3,ty);
 987             fjz3             = _mm256_add_ps(fjz3,tz);
 988
 989             }
 990
 991             /**************************
 992              * CALCULATE INTERACTIONS *
 993              **************************/
 994
 995             if (gmx_mm256_any_lt(rsq31,rcutoff2))
 996             {
 997
 998             /* REACTION-FIELD ELECTROSTATICS */
 999             velec            = _mm256_mul_ps(qq31,_mm256_sub_ps(_mm256_add_ps(rinv31,_mm256_mul_ps(krf,rsq31)),crf));
1000             felec            = _mm256_mul_ps(qq31,_mm256_sub_ps(_mm256_mul_ps(rinv31,rinvsq31),krf2));
1001
1002             cutoff_mask      = _mm256_cmp_ps(rsq31,rcutoff2,_CMP_LT_OQ);
1003
1004             /* Update potential sum for this i atom from the interaction with this j atom. */
1005             velec            = _mm256_and_ps(velec,cutoff_mask);
1006             velec            = _mm256_andnot_ps(dummy_mask,velec);
1007             velecsum         = _mm256_add_ps(velecsum,velec);
1008
1009             fscal            = felec;
1010
1011             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1012
1013             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1014
1015             /* Calculate temporary vectorial force */
1016             tx               = _mm256_mul_ps(fscal,dx31);
1017             ty               = _mm256_mul_ps(fscal,dy31);
1018             tz               = _mm256_mul_ps(fscal,dz31);
1019
1020             /* Update vectorial force */
1021             fix3             = _mm256_add_ps(fix3,tx);
1022             fiy3             = _mm256_add_ps(fiy3,ty);
1023             fiz3             = _mm256_add_ps(fiz3,tz);
1024
1025             fjx1             = _mm256_add_ps(fjx1,tx);
1026             fjy1             = _mm256_add_ps(fjy1,ty);
1027             fjz1             = _mm256_add_ps(fjz1,tz);
1028
1029             }
1030
1031             /**************************
1032              * CALCULATE INTERACTIONS *
1033              **************************/
1034
1035             if (gmx_mm256_any_lt(rsq32,rcutoff2))
1036             {
1037
1038             /* REACTION-FIELD ELECTROSTATICS */
1039             velec            = _mm256_mul_ps(qq32,_mm256_sub_ps(_mm256_add_ps(rinv32,_mm256_mul_ps(krf,rsq32)),crf));
1040             felec            = _mm256_mul_ps(qq32,_mm256_sub_ps(_mm256_mul_ps(rinv32,rinvsq32),krf2));
1041
1042             cutoff_mask      = _mm256_cmp_ps(rsq32,rcutoff2,_CMP_LT_OQ);
1043
1044             /* Update potential sum for this i atom from the interaction with this j atom. */
1045             velec            = _mm256_and_ps(velec,cutoff_mask);
1046             velec            = _mm256_andnot_ps(dummy_mask,velec);
1047             velecsum         = _mm256_add_ps(velecsum,velec);
1048
1049             fscal            = felec;
1050
1051             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1052
1053             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1054
1055             /* Calculate temporary vectorial force */
1056             tx               = _mm256_mul_ps(fscal,dx32);
1057             ty               = _mm256_mul_ps(fscal,dy32);
1058             tz               = _mm256_mul_ps(fscal,dz32);
1059
1060             /* Update vectorial force */
1061             fix3             = _mm256_add_ps(fix3,tx);
1062             fiy3             = _mm256_add_ps(fiy3,ty);
1063             fiz3             = _mm256_add_ps(fiz3,tz);
1064
1065             fjx2             = _mm256_add_ps(fjx2,tx);
1066             fjy2             = _mm256_add_ps(fjy2,ty);
1067             fjz2             = _mm256_add_ps(fjz2,tz);
1068
1069             }
1070
1071             /**************************
1072              * CALCULATE INTERACTIONS *
1073              **************************/
1074
1075             if (gmx_mm256_any_lt(rsq33,rcutoff2))
1076             {
1077
1078             /* REACTION-FIELD ELECTROSTATICS */
1079             velec            = _mm256_mul_ps(qq33,_mm256_sub_ps(_mm256_add_ps(rinv33,_mm256_mul_ps(krf,rsq33)),crf));
1080             felec            = _mm256_mul_ps(qq33,_mm256_sub_ps(_mm256_mul_ps(rinv33,rinvsq33),krf2));
1081
1082             cutoff_mask      = _mm256_cmp_ps(rsq33,rcutoff2,_CMP_LT_OQ);
1083
1084             /* Update potential sum for this i atom from the interaction with this j atom. */
1085             velec            = _mm256_and_ps(velec,cutoff_mask);
1086             velec            = _mm256_andnot_ps(dummy_mask,velec);
1087             velecsum         = _mm256_add_ps(velecsum,velec);
1088
1089             fscal            = felec;
1090
1091             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1092
1093             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1094
1095             /* Calculate temporary vectorial force */
1096             tx               = _mm256_mul_ps(fscal,dx33);
1097             ty               = _mm256_mul_ps(fscal,dy33);
1098             tz               = _mm256_mul_ps(fscal,dz33);
1099
1100             /* Update vectorial force */
1101             fix3             = _mm256_add_ps(fix3,tx);
1102             fiy3             = _mm256_add_ps(fiy3,ty);
1103             fiz3             = _mm256_add_ps(fiz3,tz);
1104
1105             fjx3             = _mm256_add_ps(fjx3,tx);
1106             fjy3             = _mm256_add_ps(fjy3,ty);
1107             fjz3             = _mm256_add_ps(fjz3,tz);
1108
1109             }
1110
1111             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1112             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1113             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1114             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1115             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1116             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1117             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
1118             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
1119
1120             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA+DIM,fjptrB+DIM,fjptrC+DIM,fjptrD+DIM,
1121                                                       fjptrE+DIM,fjptrF+DIM,fjptrG+DIM,fjptrH+DIM,
1122                                                       fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1123
1124             /* Inner loop uses 324 flops */
1125         }
1126
1127         /* End of innermost loop */
1128
1129         gmx_mm256_update_iforce_3atom_swizzle_ps(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1130                                                  f+i_coord_offset+DIM,fshift+i_shift_offset);
1131
1132         ggid                        = gid[iidx];
1133         /* Update potential energies */
1134         gmx_mm256_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
1135
1136         /* Increment number of inner iterations */
1137         inneriter                  += j_index_end - j_index_start;
1138
1139         /* Outer loop uses 19 flops */
1140     }
1141
1142     /* Increment number of outer iterations */
1143     outeriter        += nri;
1144
1145     /* Update outer/inner flops */
1146
1147     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4W4_VF,outeriter*19 + inneriter*324);
1148 }
1149 /*
1150  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwNone_GeomW4W4_F_avx_256_single
1151  * Electrostatics interaction: ReactionField
1152  * VdW interaction:            None
1153  * Geometry:                   Water4-Water4
1154  * Calculate force/pot:        Force
1155  */
1156 void
1157 nb_kernel_ElecRFCut_VdwNone_GeomW4W4_F_avx_256_single
1158                     (t_nblist * gmx_restrict                nlist,
1159                      rvec * gmx_restrict                    xx,
1160                      rvec * gmx_restrict                    ff,
1161                      t_forcerec * gmx_restrict              fr,
1162                      t_mdatoms * gmx_restrict               mdatoms,
1163                      nb_kernel_data_t * gmx_restrict        kernel_data,
1164                      t_nrnb * gmx_restrict                  nrnb)
1165 {
1166     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1167      * just 0 for non-waters.
1168      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
1169      * jnr indices corresponding to data put in the four positions in the SIMD register.
1170      */
1171     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1172     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1173     int              jnrA,jnrB,jnrC,jnrD;
1174     int              jnrE,jnrF,jnrG,jnrH;
1175     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1176     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1177     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1178     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
1179     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1180     real             rcutoff_scalar;
1181     real             *shiftvec,*fshift,*x,*f;
1182     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
1183     real             scratch[4*DIM];
1184     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1185     real *           vdwioffsetptr1;
1186     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1187     real *           vdwioffsetptr2;
1188     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1189     real *           vdwioffsetptr3;
1190     __m256           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1191     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
1192     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1193     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
1194     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1195     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D,vdwjidx3E,vdwjidx3F,vdwjidx3G,vdwjidx3H;
1196     __m256           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1197     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1198     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1199     __m256           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1200     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1201     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1202     __m256           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1203     __m256           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1204     __m256           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1205     __m256           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1206     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
1207     real             *charge;
1208     __m256           dummy_mask,cutoff_mask;
1209     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
1210     __m256           one     = _mm256_set1_ps(1.0);
1211     __m256           two     = _mm256_set1_ps(2.0);
1212     x                = xx[0];
1213     f                = ff[0];
1214
1215     nri              = nlist->nri;
1216     iinr             = nlist->iinr;
1217     jindex           = nlist->jindex;
1218     jjnr             = nlist->jjnr;
1219     shiftidx         = nlist->shift;
1220     gid              = nlist->gid;
1221     shiftvec         = fr->shift_vec[0];
1222     fshift           = fr->fshift[0];
1223     facel            = _mm256_set1_ps(fr->epsfac);
1224     charge           = mdatoms->chargeA;
1225     krf              = _mm256_set1_ps(fr->ic->k_rf);
1226     krf2             = _mm256_set1_ps(fr->ic->k_rf*2.0);
1227     crf              = _mm256_set1_ps(fr->ic->c_rf);
1228
1229     /* Setup water-specific parameters */
1230     inr              = nlist->iinr[0];
1231     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
1232     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
1233     iq3              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+3]));
1234
1235     jq1              = _mm256_set1_ps(charge[inr+1]);
1236     jq2              = _mm256_set1_ps(charge[inr+2]);
1237     jq3              = _mm256_set1_ps(charge[inr+3]);
1238     qq11             = _mm256_mul_ps(iq1,jq1);
1239     qq12             = _mm256_mul_ps(iq1,jq2);
1240     qq13             = _mm256_mul_ps(iq1,jq3);
1241     qq21             = _mm256_mul_ps(iq2,jq1);
1242     qq22             = _mm256_mul_ps(iq2,jq2);
1243     qq23             = _mm256_mul_ps(iq2,jq3);
1244     qq31             = _mm256_mul_ps(iq3,jq1);
1245     qq32             = _mm256_mul_ps(iq3,jq2);
1246     qq33             = _mm256_mul_ps(iq3,jq3);
1247
1248     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
1249     rcutoff_scalar   = fr->rcoulomb;
1250     rcutoff          = _mm256_set1_ps(rcutoff_scalar);
1251     rcutoff2         = _mm256_mul_ps(rcutoff,rcutoff);
1252
1253     /* Avoid stupid compiler warnings */
1254     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
1255     j_coord_offsetA = 0;
1256     j_coord_offsetB = 0;
1257     j_coord_offsetC = 0;
1258     j_coord_offsetD = 0;
1259     j_coord_offsetE = 0;
1260     j_coord_offsetF = 0;
1261     j_coord_offsetG = 0;
1262     j_coord_offsetH = 0;
1263
1264     outeriter        = 0;
1265     inneriter        = 0;
1266
1267     for(iidx=0;iidx<4*DIM;iidx++)
1268     {
1269         scratch[iidx] = 0.0;
1270     }
1271
1272     /* Start outer loop over neighborlists */
1273     for(iidx=0; iidx<nri; iidx++)
1274     {
1275         /* Load shift vector for this list */
1276         i_shift_offset   = DIM*shiftidx[iidx];
1277
1278         /* Load limits for loop over neighbors */
1279         j_index_start    = jindex[iidx];
1280         j_index_end      = jindex[iidx+1];
1281
1282         /* Get outer coordinate index */
1283         inr              = iinr[iidx];
1284         i_coord_offset   = DIM*inr;
1285
1286         /* Load i particle coords and add shift vector */
1287         gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
1288                                                     &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1289
1290         fix1             = _mm256_setzero_ps();
1291         fiy1             = _mm256_setzero_ps();
1292         fiz1             = _mm256_setzero_ps();
1293         fix2             = _mm256_setzero_ps();
1294         fiy2             = _mm256_setzero_ps();
1295         fiz2             = _mm256_setzero_ps();
1296         fix3             = _mm256_setzero_ps();
1297         fiy3             = _mm256_setzero_ps();
1298         fiz3             = _mm256_setzero_ps();
1299
1300         /* Start inner kernel loop */
1301         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
1302         {
1303
1304             /* Get j neighbor index, and coordinate index */
1305             jnrA             = jjnr[jidx];
1306             jnrB             = jjnr[jidx+1];
1307             jnrC             = jjnr[jidx+2];
1308             jnrD             = jjnr[jidx+3];
1309             jnrE             = jjnr[jidx+4];
1310             jnrF             = jjnr[jidx+5];
1311             jnrG             = jjnr[jidx+6];
1312             jnrH             = jjnr[jidx+7];
1313             j_coord_offsetA  = DIM*jnrA;
1314             j_coord_offsetB  = DIM*jnrB;
1315             j_coord_offsetC  = DIM*jnrC;
1316             j_coord_offsetD  = DIM*jnrD;
1317             j_coord_offsetE  = DIM*jnrE;
1318             j_coord_offsetF  = DIM*jnrF;
1319             j_coord_offsetG  = DIM*jnrG;
1320             j_coord_offsetH  = DIM*jnrH;
1321
1322             /* load j atom coordinates */
1323             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
1324                                                  x+j_coord_offsetC+DIM,x+j_coord_offsetD+DIM,
1325                                                  x+j_coord_offsetE+DIM,x+j_coord_offsetF+DIM,
1326                                                  x+j_coord_offsetG+DIM,x+j_coord_offsetH+DIM,
1327                                                  &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
1328
1329             /* Calculate displacement vector */
1330             dx11             = _mm256_sub_ps(ix1,jx1);
1331             dy11             = _mm256_sub_ps(iy1,jy1);
1332             dz11             = _mm256_sub_ps(iz1,jz1);
1333             dx12             = _mm256_sub_ps(ix1,jx2);
1334             dy12             = _mm256_sub_ps(iy1,jy2);
1335             dz12             = _mm256_sub_ps(iz1,jz2);
1336             dx13             = _mm256_sub_ps(ix1,jx3);
1337             dy13             = _mm256_sub_ps(iy1,jy3);
1338             dz13             = _mm256_sub_ps(iz1,jz3);
1339             dx21             = _mm256_sub_ps(ix2,jx1);
1340             dy21             = _mm256_sub_ps(iy2,jy1);
1341             dz21             = _mm256_sub_ps(iz2,jz1);
1342             dx22             = _mm256_sub_ps(ix2,jx2);
1343             dy22             = _mm256_sub_ps(iy2,jy2);
1344             dz22             = _mm256_sub_ps(iz2,jz2);
1345             dx23             = _mm256_sub_ps(ix2,jx3);
1346             dy23             = _mm256_sub_ps(iy2,jy3);
1347             dz23             = _mm256_sub_ps(iz2,jz3);
1348             dx31             = _mm256_sub_ps(ix3,jx1);
1349             dy31             = _mm256_sub_ps(iy3,jy1);
1350             dz31             = _mm256_sub_ps(iz3,jz1);
1351             dx32             = _mm256_sub_ps(ix3,jx2);
1352             dy32             = _mm256_sub_ps(iy3,jy2);
1353             dz32             = _mm256_sub_ps(iz3,jz2);
1354             dx33             = _mm256_sub_ps(ix3,jx3);
1355             dy33             = _mm256_sub_ps(iy3,jy3);
1356             dz33             = _mm256_sub_ps(iz3,jz3);
1357
1358             /* Calculate squared distance and things based on it */
1359             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1360             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1361             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
1362             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1363             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1364             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
1365             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
1366             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
1367             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
1368
1369             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1370             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1371             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
1372             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1373             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1374             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
1375             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
1376             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
1377             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
1378
1379             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1380             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1381             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
1382             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1383             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1384             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
1385             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
1386             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
1387             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
1388
1389             fjx1             = _mm256_setzero_ps();
1390             fjy1             = _mm256_setzero_ps();
1391             fjz1             = _mm256_setzero_ps();
1392             fjx2             = _mm256_setzero_ps();
1393             fjy2             = _mm256_setzero_ps();
1394             fjz2             = _mm256_setzero_ps();
1395             fjx3             = _mm256_setzero_ps();
1396             fjy3             = _mm256_setzero_ps();
1397             fjz3             = _mm256_setzero_ps();
1398
1399             /**************************
1400              * CALCULATE INTERACTIONS *
1401              **************************/
1402
1403             if (gmx_mm256_any_lt(rsq11,rcutoff2))
1404             {
1405
1406             /* REACTION-FIELD ELECTROSTATICS */
1407             felec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_mul_ps(rinv11,rinvsq11),krf2));
1408
1409             cutoff_mask      = _mm256_cmp_ps(rsq11,rcutoff2,_CMP_LT_OQ);
1410
1411             fscal            = felec;
1412
1413             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1414
1415             /* Calculate temporary vectorial force */
1416             tx               = _mm256_mul_ps(fscal,dx11);
1417             ty               = _mm256_mul_ps(fscal,dy11);
1418             tz               = _mm256_mul_ps(fscal,dz11);
1419
1420             /* Update vectorial force */
1421             fix1             = _mm256_add_ps(fix1,tx);
1422             fiy1             = _mm256_add_ps(fiy1,ty);
1423             fiz1             = _mm256_add_ps(fiz1,tz);
1424
1425             fjx1             = _mm256_add_ps(fjx1,tx);
1426             fjy1             = _mm256_add_ps(fjy1,ty);
1427             fjz1             = _mm256_add_ps(fjz1,tz);
1428
1429             }
1430
1431             /**************************
1432              * CALCULATE INTERACTIONS *
1433              **************************/
1434
1435             if (gmx_mm256_any_lt(rsq12,rcutoff2))
1436             {
1437
1438             /* REACTION-FIELD ELECTROSTATICS */
1439             felec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_mul_ps(rinv12,rinvsq12),krf2));
1440
1441             cutoff_mask      = _mm256_cmp_ps(rsq12,rcutoff2,_CMP_LT_OQ);
1442
1443             fscal            = felec;
1444
1445             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1446
1447             /* Calculate temporary vectorial force */
1448             tx               = _mm256_mul_ps(fscal,dx12);
1449             ty               = _mm256_mul_ps(fscal,dy12);
1450             tz               = _mm256_mul_ps(fscal,dz12);
1451
1452             /* Update vectorial force */
1453             fix1             = _mm256_add_ps(fix1,tx);
1454             fiy1             = _mm256_add_ps(fiy1,ty);
1455             fiz1             = _mm256_add_ps(fiz1,tz);
1456
1457             fjx2             = _mm256_add_ps(fjx2,tx);
1458             fjy2             = _mm256_add_ps(fjy2,ty);
1459             fjz2             = _mm256_add_ps(fjz2,tz);
1460
1461             }
1462
1463             /**************************
1464              * CALCULATE INTERACTIONS *
1465              **************************/
1466
1467             if (gmx_mm256_any_lt(rsq13,rcutoff2))
1468             {
1469
1470             /* REACTION-FIELD ELECTROSTATICS */
1471             felec            = _mm256_mul_ps(qq13,_mm256_sub_ps(_mm256_mul_ps(rinv13,rinvsq13),krf2));
1472
1473             cutoff_mask      = _mm256_cmp_ps(rsq13,rcutoff2,_CMP_LT_OQ);
1474
1475             fscal            = felec;
1476
1477             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1478
1479             /* Calculate temporary vectorial force */
1480             tx               = _mm256_mul_ps(fscal,dx13);
1481             ty               = _mm256_mul_ps(fscal,dy13);
1482             tz               = _mm256_mul_ps(fscal,dz13);
1483
1484             /* Update vectorial force */
1485             fix1             = _mm256_add_ps(fix1,tx);
1486             fiy1             = _mm256_add_ps(fiy1,ty);
1487             fiz1             = _mm256_add_ps(fiz1,tz);
1488
1489             fjx3             = _mm256_add_ps(fjx3,tx);
1490             fjy3             = _mm256_add_ps(fjy3,ty);
1491             fjz3             = _mm256_add_ps(fjz3,tz);
1492
1493             }
1494
1495             /**************************
1496              * CALCULATE INTERACTIONS *
1497              **************************/
1498
1499             if (gmx_mm256_any_lt(rsq21,rcutoff2))
1500             {
1501
1502             /* REACTION-FIELD ELECTROSTATICS */
1503             felec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_mul_ps(rinv21,rinvsq21),krf2));
1504
1505             cutoff_mask      = _mm256_cmp_ps(rsq21,rcutoff2,_CMP_LT_OQ);
1506
1507             fscal            = felec;
1508
1509             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1510
1511             /* Calculate temporary vectorial force */
1512             tx               = _mm256_mul_ps(fscal,dx21);
1513             ty               = _mm256_mul_ps(fscal,dy21);
1514             tz               = _mm256_mul_ps(fscal,dz21);
1515
1516             /* Update vectorial force */
1517             fix2             = _mm256_add_ps(fix2,tx);
1518             fiy2             = _mm256_add_ps(fiy2,ty);
1519             fiz2             = _mm256_add_ps(fiz2,tz);
1520
1521             fjx1             = _mm256_add_ps(fjx1,tx);
1522             fjy1             = _mm256_add_ps(fjy1,ty);
1523             fjz1             = _mm256_add_ps(fjz1,tz);
1524
1525             }
1526
1527             /**************************
1528              * CALCULATE INTERACTIONS *
1529              **************************/
1530
1531             if (gmx_mm256_any_lt(rsq22,rcutoff2))
1532             {
1533
1534             /* REACTION-FIELD ELECTROSTATICS */
1535             felec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_mul_ps(rinv22,rinvsq22),krf2));
1536
1537             cutoff_mask      = _mm256_cmp_ps(rsq22,rcutoff2,_CMP_LT_OQ);
1538
1539             fscal            = felec;
1540
1541             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1542
1543             /* Calculate temporary vectorial force */
1544             tx               = _mm256_mul_ps(fscal,dx22);
1545             ty               = _mm256_mul_ps(fscal,dy22);
1546             tz               = _mm256_mul_ps(fscal,dz22);
1547
1548             /* Update vectorial force */
1549             fix2             = _mm256_add_ps(fix2,tx);
1550             fiy2             = _mm256_add_ps(fiy2,ty);
1551             fiz2             = _mm256_add_ps(fiz2,tz);
1552
1553             fjx2             = _mm256_add_ps(fjx2,tx);
1554             fjy2             = _mm256_add_ps(fjy2,ty);
1555             fjz2             = _mm256_add_ps(fjz2,tz);
1556
1557             }
1558
1559             /**************************
1560              * CALCULATE INTERACTIONS *
1561              **************************/
1562
1563             if (gmx_mm256_any_lt(rsq23,rcutoff2))
1564             {
1565
1566             /* REACTION-FIELD ELECTROSTATICS */
1567             felec            = _mm256_mul_ps(qq23,_mm256_sub_ps(_mm256_mul_ps(rinv23,rinvsq23),krf2));
1568
1569             cutoff_mask      = _mm256_cmp_ps(rsq23,rcutoff2,_CMP_LT_OQ);
1570
1571             fscal            = felec;
1572
1573             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1574
1575             /* Calculate temporary vectorial force */
1576             tx               = _mm256_mul_ps(fscal,dx23);
1577             ty               = _mm256_mul_ps(fscal,dy23);
1578             tz               = _mm256_mul_ps(fscal,dz23);
1579
1580             /* Update vectorial force */
1581             fix2             = _mm256_add_ps(fix2,tx);
1582             fiy2             = _mm256_add_ps(fiy2,ty);
1583             fiz2             = _mm256_add_ps(fiz2,tz);
1584
1585             fjx3             = _mm256_add_ps(fjx3,tx);
1586             fjy3             = _mm256_add_ps(fjy3,ty);
1587             fjz3             = _mm256_add_ps(fjz3,tz);
1588
1589             }
1590
1591             /**************************
1592              * CALCULATE INTERACTIONS *
1593              **************************/
1594
1595             if (gmx_mm256_any_lt(rsq31,rcutoff2))
1596             {
1597
1598             /* REACTION-FIELD ELECTROSTATICS */
1599             felec            = _mm256_mul_ps(qq31,_mm256_sub_ps(_mm256_mul_ps(rinv31,rinvsq31),krf2));
1600
1601             cutoff_mask      = _mm256_cmp_ps(rsq31,rcutoff2,_CMP_LT_OQ);
1602
1603             fscal            = felec;
1604
1605             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1606
1607             /* Calculate temporary vectorial force */
1608             tx               = _mm256_mul_ps(fscal,dx31);
1609             ty               = _mm256_mul_ps(fscal,dy31);
1610             tz               = _mm256_mul_ps(fscal,dz31);
1611
1612             /* Update vectorial force */
1613             fix3             = _mm256_add_ps(fix3,tx);
1614             fiy3             = _mm256_add_ps(fiy3,ty);
1615             fiz3             = _mm256_add_ps(fiz3,tz);
1616
1617             fjx1             = _mm256_add_ps(fjx1,tx);
1618             fjy1             = _mm256_add_ps(fjy1,ty);
1619             fjz1             = _mm256_add_ps(fjz1,tz);
1620
1621             }
1622
1623             /**************************
1624              * CALCULATE INTERACTIONS *
1625              **************************/
1626
1627             if (gmx_mm256_any_lt(rsq32,rcutoff2))
1628             {
1629
1630             /* REACTION-FIELD ELECTROSTATICS */
1631             felec            = _mm256_mul_ps(qq32,_mm256_sub_ps(_mm256_mul_ps(rinv32,rinvsq32),krf2));
1632
1633             cutoff_mask      = _mm256_cmp_ps(rsq32,rcutoff2,_CMP_LT_OQ);
1634
1635             fscal            = felec;
1636
1637             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1638
1639             /* Calculate temporary vectorial force */
1640             tx               = _mm256_mul_ps(fscal,dx32);
1641             ty               = _mm256_mul_ps(fscal,dy32);
1642             tz               = _mm256_mul_ps(fscal,dz32);
1643
1644             /* Update vectorial force */
1645             fix3             = _mm256_add_ps(fix3,tx);
1646             fiy3             = _mm256_add_ps(fiy3,ty);
1647             fiz3             = _mm256_add_ps(fiz3,tz);
1648
1649             fjx2             = _mm256_add_ps(fjx2,tx);
1650             fjy2             = _mm256_add_ps(fjy2,ty);
1651             fjz2             = _mm256_add_ps(fjz2,tz);
1652
1653             }
1654
1655             /**************************
1656              * CALCULATE INTERACTIONS *
1657              **************************/
1658
1659             if (gmx_mm256_any_lt(rsq33,rcutoff2))
1660             {
1661
1662             /* REACTION-FIELD ELECTROSTATICS */
1663             felec            = _mm256_mul_ps(qq33,_mm256_sub_ps(_mm256_mul_ps(rinv33,rinvsq33),krf2));
1664
1665             cutoff_mask      = _mm256_cmp_ps(rsq33,rcutoff2,_CMP_LT_OQ);
1666
1667             fscal            = felec;
1668
1669             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1670
1671             /* Calculate temporary vectorial force */
1672             tx               = _mm256_mul_ps(fscal,dx33);
1673             ty               = _mm256_mul_ps(fscal,dy33);
1674             tz               = _mm256_mul_ps(fscal,dz33);
1675
1676             /* Update vectorial force */
1677             fix3             = _mm256_add_ps(fix3,tx);
1678             fiy3             = _mm256_add_ps(fiy3,ty);
1679             fiz3             = _mm256_add_ps(fiz3,tz);
1680
1681             fjx3             = _mm256_add_ps(fjx3,tx);
1682             fjy3             = _mm256_add_ps(fjy3,ty);
1683             fjz3             = _mm256_add_ps(fjz3,tz);
1684
1685             }
1686
1687             fjptrA             = f+j_coord_offsetA;
1688             fjptrB             = f+j_coord_offsetB;
1689             fjptrC             = f+j_coord_offsetC;
1690             fjptrD             = f+j_coord_offsetD;
1691             fjptrE             = f+j_coord_offsetE;
1692             fjptrF             = f+j_coord_offsetF;
1693             fjptrG             = f+j_coord_offsetG;
1694             fjptrH             = f+j_coord_offsetH;
1695
1696             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA+DIM,fjptrB+DIM,fjptrC+DIM,fjptrD+DIM,
1697                                                       fjptrE+DIM,fjptrF+DIM,fjptrG+DIM,fjptrH+DIM,
1698                                                       fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1699
1700             /* Inner loop uses 270 flops */
1701         }
1702
1703         if(jidx<j_index_end)
1704         {
1705
1706             /* Get j neighbor index, and coordinate index */
1707             jnrlistA         = jjnr[jidx];
1708             jnrlistB         = jjnr[jidx+1];
1709             jnrlistC         = jjnr[jidx+2];
1710             jnrlistD         = jjnr[jidx+3];
1711             jnrlistE         = jjnr[jidx+4];
1712             jnrlistF         = jjnr[jidx+5];
1713             jnrlistG         = jjnr[jidx+6];
1714             jnrlistH         = jjnr[jidx+7];
1715             /* Sign of each element will be negative for non-real atoms.
1716              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1717              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1718              */
1719             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
1720                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
1721
1722             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1723             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1724             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1725             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1726             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
1727             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
1728             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
1729             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
1730             j_coord_offsetA  = DIM*jnrA;
1731             j_coord_offsetB  = DIM*jnrB;
1732             j_coord_offsetC  = DIM*jnrC;
1733             j_coord_offsetD  = DIM*jnrD;
1734             j_coord_offsetE  = DIM*jnrE;
1735             j_coord_offsetF  = DIM*jnrF;
1736             j_coord_offsetG  = DIM*jnrG;
1737             j_coord_offsetH  = DIM*jnrH;
1738
1739             /* load j atom coordinates */
1740             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
1741                                                  x+j_coord_offsetC+DIM,x+j_coord_offsetD+DIM,
1742                                                  x+j_coord_offsetE+DIM,x+j_coord_offsetF+DIM,
1743                                                  x+j_coord_offsetG+DIM,x+j_coord_offsetH+DIM,
1744                                                  &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
1745
1746             /* Calculate displacement vector */
1747             dx11             = _mm256_sub_ps(ix1,jx1);
1748             dy11             = _mm256_sub_ps(iy1,jy1);
1749             dz11             = _mm256_sub_ps(iz1,jz1);
1750             dx12             = _mm256_sub_ps(ix1,jx2);
1751             dy12             = _mm256_sub_ps(iy1,jy2);
1752             dz12             = _mm256_sub_ps(iz1,jz2);
1753             dx13             = _mm256_sub_ps(ix1,jx3);
1754             dy13             = _mm256_sub_ps(iy1,jy3);
1755             dz13             = _mm256_sub_ps(iz1,jz3);
1756             dx21             = _mm256_sub_ps(ix2,jx1);
1757             dy21             = _mm256_sub_ps(iy2,jy1);
1758             dz21             = _mm256_sub_ps(iz2,jz1);
1759             dx22             = _mm256_sub_ps(ix2,jx2);
1760             dy22             = _mm256_sub_ps(iy2,jy2);
1761             dz22             = _mm256_sub_ps(iz2,jz2);
1762             dx23             = _mm256_sub_ps(ix2,jx3);
1763             dy23             = _mm256_sub_ps(iy2,jy3);
1764             dz23             = _mm256_sub_ps(iz2,jz3);
1765             dx31             = _mm256_sub_ps(ix3,jx1);
1766             dy31             = _mm256_sub_ps(iy3,jy1);
1767             dz31             = _mm256_sub_ps(iz3,jz1);
1768             dx32             = _mm256_sub_ps(ix3,jx2);
1769             dy32             = _mm256_sub_ps(iy3,jy2);
1770             dz32             = _mm256_sub_ps(iz3,jz2);
1771             dx33             = _mm256_sub_ps(ix3,jx3);
1772             dy33             = _mm256_sub_ps(iy3,jy3);
1773             dz33             = _mm256_sub_ps(iz3,jz3);
1774
1775             /* Calculate squared distance and things based on it */
1776             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1777             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1778             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
1779             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1780             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1781             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
1782             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
1783             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
1784             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
1785
1786             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1787             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1788             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
1789             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1790             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1791             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
1792             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
1793             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
1794             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
1795
1796             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1797             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1798             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
1799             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1800             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1801             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
1802             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
1803             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
1804             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
1805
1806             fjx1             = _mm256_setzero_ps();
1807             fjy1             = _mm256_setzero_ps();
1808             fjz1             = _mm256_setzero_ps();
1809             fjx2             = _mm256_setzero_ps();
1810             fjy2             = _mm256_setzero_ps();
1811             fjz2             = _mm256_setzero_ps();
1812             fjx3             = _mm256_setzero_ps();
1813             fjy3             = _mm256_setzero_ps();
1814             fjz3             = _mm256_setzero_ps();
1815
1816             /**************************
1817              * CALCULATE INTERACTIONS *
1818              **************************/
1819
1820             if (gmx_mm256_any_lt(rsq11,rcutoff2))
1821             {
1822
1823             /* REACTION-FIELD ELECTROSTATICS */
1824             felec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_mul_ps(rinv11,rinvsq11),krf2));
1825
1826             cutoff_mask      = _mm256_cmp_ps(rsq11,rcutoff2,_CMP_LT_OQ);
1827
1828             fscal            = felec;
1829
1830             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1831
1832             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1833
1834             /* Calculate temporary vectorial force */
1835             tx               = _mm256_mul_ps(fscal,dx11);
1836             ty               = _mm256_mul_ps(fscal,dy11);
1837             tz               = _mm256_mul_ps(fscal,dz11);
1838
1839             /* Update vectorial force */
1840             fix1             = _mm256_add_ps(fix1,tx);
1841             fiy1             = _mm256_add_ps(fiy1,ty);
1842             fiz1             = _mm256_add_ps(fiz1,tz);
1843
1844             fjx1             = _mm256_add_ps(fjx1,tx);
1845             fjy1             = _mm256_add_ps(fjy1,ty);
1846             fjz1             = _mm256_add_ps(fjz1,tz);
1847
1848             }
1849
1850             /**************************
1851              * CALCULATE INTERACTIONS *
1852              **************************/
1853
1854             if (gmx_mm256_any_lt(rsq12,rcutoff2))
1855             {
1856
1857             /* REACTION-FIELD ELECTROSTATICS */
1858             felec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_mul_ps(rinv12,rinvsq12),krf2));
1859
1860             cutoff_mask      = _mm256_cmp_ps(rsq12,rcutoff2,_CMP_LT_OQ);
1861
1862             fscal            = felec;
1863
1864             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1865
1866             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1867
1868             /* Calculate temporary vectorial force */
1869             tx               = _mm256_mul_ps(fscal,dx12);
1870             ty               = _mm256_mul_ps(fscal,dy12);
1871             tz               = _mm256_mul_ps(fscal,dz12);
1872
1873             /* Update vectorial force */
1874             fix1             = _mm256_add_ps(fix1,tx);
1875             fiy1             = _mm256_add_ps(fiy1,ty);
1876             fiz1             = _mm256_add_ps(fiz1,tz);
1877
1878             fjx2             = _mm256_add_ps(fjx2,tx);
1879             fjy2             = _mm256_add_ps(fjy2,ty);
1880             fjz2             = _mm256_add_ps(fjz2,tz);
1881
1882             }
1883
1884             /**************************
1885              * CALCULATE INTERACTIONS *
1886              **************************/
1887
1888             if (gmx_mm256_any_lt(rsq13,rcutoff2))
1889             {
1890
1891             /* REACTION-FIELD ELECTROSTATICS */
1892             felec            = _mm256_mul_ps(qq13,_mm256_sub_ps(_mm256_mul_ps(rinv13,rinvsq13),krf2));
1893
1894             cutoff_mask      = _mm256_cmp_ps(rsq13,rcutoff2,_CMP_LT_OQ);
1895
1896             fscal            = felec;
1897
1898             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1899
1900             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1901
1902             /* Calculate temporary vectorial force */
1903             tx               = _mm256_mul_ps(fscal,dx13);
1904             ty               = _mm256_mul_ps(fscal,dy13);
1905             tz               = _mm256_mul_ps(fscal,dz13);
1906
1907             /* Update vectorial force */
1908             fix1             = _mm256_add_ps(fix1,tx);
1909             fiy1             = _mm256_add_ps(fiy1,ty);
1910             fiz1             = _mm256_add_ps(fiz1,tz);
1911
1912             fjx3             = _mm256_add_ps(fjx3,tx);
1913             fjy3             = _mm256_add_ps(fjy3,ty);
1914             fjz3             = _mm256_add_ps(fjz3,tz);
1915
1916             }
1917
1918             /**************************
1919              * CALCULATE INTERACTIONS *
1920              **************************/
1921
1922             if (gmx_mm256_any_lt(rsq21,rcutoff2))
1923             {
1924
1925             /* REACTION-FIELD ELECTROSTATICS */
1926             felec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_mul_ps(rinv21,rinvsq21),krf2));
1927
1928             cutoff_mask      = _mm256_cmp_ps(rsq21,rcutoff2,_CMP_LT_OQ);
1929
1930             fscal            = felec;
1931
1932             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1933
1934             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1935
1936             /* Calculate temporary vectorial force */
1937             tx               = _mm256_mul_ps(fscal,dx21);
1938             ty               = _mm256_mul_ps(fscal,dy21);
1939             tz               = _mm256_mul_ps(fscal,dz21);
1940
1941             /* Update vectorial force */
1942             fix2             = _mm256_add_ps(fix2,tx);
1943             fiy2             = _mm256_add_ps(fiy2,ty);
1944             fiz2             = _mm256_add_ps(fiz2,tz);
1945
1946             fjx1             = _mm256_add_ps(fjx1,tx);
1947             fjy1             = _mm256_add_ps(fjy1,ty);
1948             fjz1             = _mm256_add_ps(fjz1,tz);
1949
1950             }
1951
1952             /**************************
1953              * CALCULATE INTERACTIONS *
1954              **************************/
1955
1956             if (gmx_mm256_any_lt(rsq22,rcutoff2))
1957             {
1958
1959             /* REACTION-FIELD ELECTROSTATICS */
1960             felec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_mul_ps(rinv22,rinvsq22),krf2));
1961
1962             cutoff_mask      = _mm256_cmp_ps(rsq22,rcutoff2,_CMP_LT_OQ);
1963
1964             fscal            = felec;
1965
1966             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1967
1968             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1969
1970             /* Calculate temporary vectorial force */
1971             tx               = _mm256_mul_ps(fscal,dx22);
1972             ty               = _mm256_mul_ps(fscal,dy22);
1973             tz               = _mm256_mul_ps(fscal,dz22);
1974
1975             /* Update vectorial force */
1976             fix2             = _mm256_add_ps(fix2,tx);
1977             fiy2             = _mm256_add_ps(fiy2,ty);
1978             fiz2             = _mm256_add_ps(fiz2,tz);
1979
1980             fjx2             = _mm256_add_ps(fjx2,tx);
1981             fjy2             = _mm256_add_ps(fjy2,ty);
1982             fjz2             = _mm256_add_ps(fjz2,tz);
1983
1984             }
1985
1986             /**************************
1987              * CALCULATE INTERACTIONS *
1988              **************************/
1989
1990             if (gmx_mm256_any_lt(rsq23,rcutoff2))
1991             {
1992
1993             /* REACTION-FIELD ELECTROSTATICS */
1994             felec            = _mm256_mul_ps(qq23,_mm256_sub_ps(_mm256_mul_ps(rinv23,rinvsq23),krf2));
1995
1996             cutoff_mask      = _mm256_cmp_ps(rsq23,rcutoff2,_CMP_LT_OQ);
1997
1998             fscal            = felec;
1999
2000             fscal            = _mm256_and_ps(fscal,cutoff_mask);
2001
2002             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2003
2004             /* Calculate temporary vectorial force */
2005             tx               = _mm256_mul_ps(fscal,dx23);
2006             ty               = _mm256_mul_ps(fscal,dy23);
2007             tz               = _mm256_mul_ps(fscal,dz23);
2008
2009             /* Update vectorial force */
2010             fix2             = _mm256_add_ps(fix2,tx);
2011             fiy2             = _mm256_add_ps(fiy2,ty);
2012             fiz2             = _mm256_add_ps(fiz2,tz);
2013
2014             fjx3             = _mm256_add_ps(fjx3,tx);
2015             fjy3             = _mm256_add_ps(fjy3,ty);
2016             fjz3             = _mm256_add_ps(fjz3,tz);
2017
2018             }
2019
2020             /**************************
2021              * CALCULATE INTERACTIONS *
2022              **************************/
2023
2024             if (gmx_mm256_any_lt(rsq31,rcutoff2))
2025             {
2026
2027             /* REACTION-FIELD ELECTROSTATICS */
2028             felec            = _mm256_mul_ps(qq31,_mm256_sub_ps(_mm256_mul_ps(rinv31,rinvsq31),krf2));
2029
2030             cutoff_mask      = _mm256_cmp_ps(rsq31,rcutoff2,_CMP_LT_OQ);
2031
2032             fscal            = felec;
2033
2034             fscal            = _mm256_and_ps(fscal,cutoff_mask);
2035
2036             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2037
2038             /* Calculate temporary vectorial force */
2039             tx               = _mm256_mul_ps(fscal,dx31);
2040             ty               = _mm256_mul_ps(fscal,dy31);
2041             tz               = _mm256_mul_ps(fscal,dz31);
2042
2043             /* Update vectorial force */
2044             fix3             = _mm256_add_ps(fix3,tx);
2045             fiy3             = _mm256_add_ps(fiy3,ty);
2046             fiz3             = _mm256_add_ps(fiz3,tz);
2047
2048             fjx1             = _mm256_add_ps(fjx1,tx);
2049             fjy1             = _mm256_add_ps(fjy1,ty);
2050             fjz1             = _mm256_add_ps(fjz1,tz);
2051
2052             }
2053
2054             /**************************
2055              * CALCULATE INTERACTIONS *
2056              **************************/
2057
2058             if (gmx_mm256_any_lt(rsq32,rcutoff2))
2059             {
2060
2061             /* REACTION-FIELD ELECTROSTATICS */
2062             felec            = _mm256_mul_ps(qq32,_mm256_sub_ps(_mm256_mul_ps(rinv32,rinvsq32),krf2));
2063
2064             cutoff_mask      = _mm256_cmp_ps(rsq32,rcutoff2,_CMP_LT_OQ);
2065
2066             fscal            = felec;
2067
2068             fscal            = _mm256_and_ps(fscal,cutoff_mask);
2069
2070             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2071
2072             /* Calculate temporary vectorial force */
2073             tx               = _mm256_mul_ps(fscal,dx32);
2074             ty               = _mm256_mul_ps(fscal,dy32);
2075             tz               = _mm256_mul_ps(fscal,dz32);
2076
2077             /* Update vectorial force */
2078             fix3             = _mm256_add_ps(fix3,tx);
2079             fiy3             = _mm256_add_ps(fiy3,ty);
2080             fiz3             = _mm256_add_ps(fiz3,tz);
2081
2082             fjx2             = _mm256_add_ps(fjx2,tx);
2083             fjy2             = _mm256_add_ps(fjy2,ty);
2084             fjz2             = _mm256_add_ps(fjz2,tz);
2085
2086             }
2087
2088             /**************************
2089              * CALCULATE INTERACTIONS *
2090              **************************/
2091
2092             if (gmx_mm256_any_lt(rsq33,rcutoff2))
2093             {
2094
2095             /* REACTION-FIELD ELECTROSTATICS */
2096             felec            = _mm256_mul_ps(qq33,_mm256_sub_ps(_mm256_mul_ps(rinv33,rinvsq33),krf2));
2097
2098             cutoff_mask      = _mm256_cmp_ps(rsq33,rcutoff2,_CMP_LT_OQ);
2099
2100             fscal            = felec;
2101
2102             fscal            = _mm256_and_ps(fscal,cutoff_mask);
2103
2104             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2105
2106             /* Calculate temporary vectorial force */
2107             tx               = _mm256_mul_ps(fscal,dx33);
2108             ty               = _mm256_mul_ps(fscal,dy33);
2109             tz               = _mm256_mul_ps(fscal,dz33);
2110
2111             /* Update vectorial force */
2112             fix3             = _mm256_add_ps(fix3,tx);
2113             fiy3             = _mm256_add_ps(fiy3,ty);
2114             fiz3             = _mm256_add_ps(fiz3,tz);
2115
2116             fjx3             = _mm256_add_ps(fjx3,tx);
2117             fjy3             = _mm256_add_ps(fjy3,ty);
2118             fjz3             = _mm256_add_ps(fjz3,tz);
2119
2120             }
2121
2122             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
2123             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
2124             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
2125             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
2126             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
2127             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
2128             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
2129             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
2130
2131             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA+DIM,fjptrB+DIM,fjptrC+DIM,fjptrD+DIM,
2132                                                       fjptrE+DIM,fjptrF+DIM,fjptrG+DIM,fjptrH+DIM,
2133                                                       fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2134
2135             /* Inner loop uses 270 flops */
2136         }
2137
2138         /* End of innermost loop */
2139
2140         gmx_mm256_update_iforce_3atom_swizzle_ps(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2141                                                  f+i_coord_offset+DIM,fshift+i_shift_offset);
2142
2143         /* Increment number of inner iterations */
2144         inneriter                  += j_index_end - j_index_start;
2145
2146         /* Outer loop uses 18 flops */
2147     }
2148
2149     /* Increment number of outer iterations */
2150     outeriter        += nri;
2151
2152     /* Update outer/inner flops */
2153
2154     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4W4_F,outeriter*18 + inneriter*270);
2155 }