src/gmxlib/nonbonded/nb_kernel_avx_256_single/nb_kernel_ElecRFCut_VdwNone_GeomW3W3_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_GeomW3W3_VF_avx_256_single
  38  * Electrostatics interaction: ReactionField
  39  * VdW interaction:            None
  40  * Geometry:                   Water3-Water3
  41  * Calculate force/pot:        PotentialAndForce
  42  */
  43 void
  44 nb_kernel_ElecRFCut_VdwNone_GeomW3W3_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 *           vdwioffsetptr0;
  73     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  74     real *           vdwioffsetptr1;
  75     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  76     real *           vdwioffsetptr2;
  77     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  78     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
  79     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  80     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
  81     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  82     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
  83     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  84     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  85     __m256           dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
  86     __m256           dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
  87     __m256           dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  88     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
  89     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
  90     __m256           dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  91     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
  92     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
  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     iq0              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+0]));
 119     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
 120     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
 121
 122     jq0              = _mm256_set1_ps(charge[inr+0]);
 123     jq1              = _mm256_set1_ps(charge[inr+1]);
 124     jq2              = _mm256_set1_ps(charge[inr+2]);
 125     qq00             = _mm256_mul_ps(iq0,jq0);
 126     qq01             = _mm256_mul_ps(iq0,jq1);
 127     qq02             = _mm256_mul_ps(iq0,jq2);
 128     qq10             = _mm256_mul_ps(iq1,jq0);
 129     qq11             = _mm256_mul_ps(iq1,jq1);
 130     qq12             = _mm256_mul_ps(iq1,jq2);
 131     qq20             = _mm256_mul_ps(iq2,jq0);
 132     qq21             = _mm256_mul_ps(iq2,jq1);
 133     qq22             = _mm256_mul_ps(iq2,jq2);
 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,
 175                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 176
 177         fix0             = _mm256_setzero_ps();
 178         fiy0             = _mm256_setzero_ps();
 179         fiz0             = _mm256_setzero_ps();
 180         fix1             = _mm256_setzero_ps();
 181         fiy1             = _mm256_setzero_ps();
 182         fiz1             = _mm256_setzero_ps();
 183         fix2             = _mm256_setzero_ps();
 184         fiy2             = _mm256_setzero_ps();
 185         fiz2             = _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,x+j_coord_offsetB,
 214                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 215                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 216                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 217                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 218
 219             /* Calculate displacement vector */
 220             dx00             = _mm256_sub_ps(ix0,jx0);
 221             dy00             = _mm256_sub_ps(iy0,jy0);
 222             dz00             = _mm256_sub_ps(iz0,jz0);
 223             dx01             = _mm256_sub_ps(ix0,jx1);
 224             dy01             = _mm256_sub_ps(iy0,jy1);
 225             dz01             = _mm256_sub_ps(iz0,jz1);
 226             dx02             = _mm256_sub_ps(ix0,jx2);
 227             dy02             = _mm256_sub_ps(iy0,jy2);
 228             dz02             = _mm256_sub_ps(iz0,jz2);
 229             dx10             = _mm256_sub_ps(ix1,jx0);
 230             dy10             = _mm256_sub_ps(iy1,jy0);
 231             dz10             = _mm256_sub_ps(iz1,jz0);
 232             dx11             = _mm256_sub_ps(ix1,jx1);
 233             dy11             = _mm256_sub_ps(iy1,jy1);
 234             dz11             = _mm256_sub_ps(iz1,jz1);
 235             dx12             = _mm256_sub_ps(ix1,jx2);
 236             dy12             = _mm256_sub_ps(iy1,jy2);
 237             dz12             = _mm256_sub_ps(iz1,jz2);
 238             dx20             = _mm256_sub_ps(ix2,jx0);
 239             dy20             = _mm256_sub_ps(iy2,jy0);
 240             dz20             = _mm256_sub_ps(iz2,jz0);
 241             dx21             = _mm256_sub_ps(ix2,jx1);
 242             dy21             = _mm256_sub_ps(iy2,jy1);
 243             dz21             = _mm256_sub_ps(iz2,jz1);
 244             dx22             = _mm256_sub_ps(ix2,jx2);
 245             dy22             = _mm256_sub_ps(iy2,jy2);
 246             dz22             = _mm256_sub_ps(iz2,jz2);
 247
 248             /* Calculate squared distance and things based on it */
 249             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 250             rsq01            = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
 251             rsq02            = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
 252             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
 253             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 254             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 255             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
 256             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 257             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 258
 259             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
 260             rinv01           = gmx_mm256_invsqrt_ps(rsq01);
 261             rinv02           = gmx_mm256_invsqrt_ps(rsq02);
 262             rinv10           = gmx_mm256_invsqrt_ps(rsq10);
 263             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 264             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 265             rinv20           = gmx_mm256_invsqrt_ps(rsq20);
 266             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 267             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 268
 269             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
 270             rinvsq01         = _mm256_mul_ps(rinv01,rinv01);
 271             rinvsq02         = _mm256_mul_ps(rinv02,rinv02);
 272             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
 273             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 274             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 275             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
 276             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 277             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 278
 279             fjx0             = _mm256_setzero_ps();
 280             fjy0             = _mm256_setzero_ps();
 281             fjz0             = _mm256_setzero_ps();
 282             fjx1             = _mm256_setzero_ps();
 283             fjy1             = _mm256_setzero_ps();
 284             fjz1             = _mm256_setzero_ps();
 285             fjx2             = _mm256_setzero_ps();
 286             fjy2             = _mm256_setzero_ps();
 287             fjz2             = _mm256_setzero_ps();
 288
 289             /**************************
 290              * CALCULATE INTERACTIONS *
 291              **************************/
 292
 293             if (gmx_mm256_any_lt(rsq00,rcutoff2))
 294             {
 295
 296             /* REACTION-FIELD ELECTROSTATICS */
 297             velec            = _mm256_mul_ps(qq00,_mm256_sub_ps(_mm256_add_ps(rinv00,_mm256_mul_ps(krf,rsq00)),crf));
 298             felec            = _mm256_mul_ps(qq00,_mm256_sub_ps(_mm256_mul_ps(rinv00,rinvsq00),krf2));
 299
 300             cutoff_mask      = _mm256_cmp_ps(rsq00,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,dx00);
 312             ty               = _mm256_mul_ps(fscal,dy00);
 313             tz               = _mm256_mul_ps(fscal,dz00);
 314
 315             /* Update vectorial force */
 316             fix0             = _mm256_add_ps(fix0,tx);
 317             fiy0             = _mm256_add_ps(fiy0,ty);
 318             fiz0             = _mm256_add_ps(fiz0,tz);
 319
 320             fjx0             = _mm256_add_ps(fjx0,tx);
 321             fjy0             = _mm256_add_ps(fjy0,ty);
 322             fjz0             = _mm256_add_ps(fjz0,tz);
 323
 324             }
 325
 326             /**************************
 327              * CALCULATE INTERACTIONS *
 328              **************************/
 329
 330             if (gmx_mm256_any_lt(rsq01,rcutoff2))
 331             {
 332
 333             /* REACTION-FIELD ELECTROSTATICS */
 334             velec            = _mm256_mul_ps(qq01,_mm256_sub_ps(_mm256_add_ps(rinv01,_mm256_mul_ps(krf,rsq01)),crf));
 335             felec            = _mm256_mul_ps(qq01,_mm256_sub_ps(_mm256_mul_ps(rinv01,rinvsq01),krf2));
 336
 337             cutoff_mask      = _mm256_cmp_ps(rsq01,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,dx01);
 349             ty               = _mm256_mul_ps(fscal,dy01);
 350             tz               = _mm256_mul_ps(fscal,dz01);
 351
 352             /* Update vectorial force */
 353             fix0             = _mm256_add_ps(fix0,tx);
 354             fiy0             = _mm256_add_ps(fiy0,ty);
 355             fiz0             = _mm256_add_ps(fiz0,tz);
 356
 357             fjx1             = _mm256_add_ps(fjx1,tx);
 358             fjy1             = _mm256_add_ps(fjy1,ty);
 359             fjz1             = _mm256_add_ps(fjz1,tz);
 360
 361             }
 362
 363             /**************************
 364              * CALCULATE INTERACTIONS *
 365              **************************/
 366
 367             if (gmx_mm256_any_lt(rsq02,rcutoff2))
 368             {
 369
 370             /* REACTION-FIELD ELECTROSTATICS */
 371             velec            = _mm256_mul_ps(qq02,_mm256_sub_ps(_mm256_add_ps(rinv02,_mm256_mul_ps(krf,rsq02)),crf));
 372             felec            = _mm256_mul_ps(qq02,_mm256_sub_ps(_mm256_mul_ps(rinv02,rinvsq02),krf2));
 373
 374             cutoff_mask      = _mm256_cmp_ps(rsq02,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,dx02);
 386             ty               = _mm256_mul_ps(fscal,dy02);
 387             tz               = _mm256_mul_ps(fscal,dz02);
 388
 389             /* Update vectorial force */
 390             fix0             = _mm256_add_ps(fix0,tx);
 391             fiy0             = _mm256_add_ps(fiy0,ty);
 392             fiz0             = _mm256_add_ps(fiz0,tz);
 393
 394             fjx2             = _mm256_add_ps(fjx2,tx);
 395             fjy2             = _mm256_add_ps(fjy2,ty);
 396             fjz2             = _mm256_add_ps(fjz2,tz);
 397
 398             }
 399
 400             /**************************
 401              * CALCULATE INTERACTIONS *
 402              **************************/
 403
 404             if (gmx_mm256_any_lt(rsq10,rcutoff2))
 405             {
 406
 407             /* REACTION-FIELD ELECTROSTATICS */
 408             velec            = _mm256_mul_ps(qq10,_mm256_sub_ps(_mm256_add_ps(rinv10,_mm256_mul_ps(krf,rsq10)),crf));
 409             felec            = _mm256_mul_ps(qq10,_mm256_sub_ps(_mm256_mul_ps(rinv10,rinvsq10),krf2));
 410
 411             cutoff_mask      = _mm256_cmp_ps(rsq10,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,dx10);
 423             ty               = _mm256_mul_ps(fscal,dy10);
 424             tz               = _mm256_mul_ps(fscal,dz10);
 425
 426             /* Update vectorial force */
 427             fix1             = _mm256_add_ps(fix1,tx);
 428             fiy1             = _mm256_add_ps(fiy1,ty);
 429             fiz1             = _mm256_add_ps(fiz1,tz);
 430
 431             fjx0             = _mm256_add_ps(fjx0,tx);
 432             fjy0             = _mm256_add_ps(fjy0,ty);
 433             fjz0             = _mm256_add_ps(fjz0,tz);
 434
 435             }
 436
 437             /**************************
 438              * CALCULATE INTERACTIONS *
 439              **************************/
 440
 441             if (gmx_mm256_any_lt(rsq11,rcutoff2))
 442             {
 443
 444             /* REACTION-FIELD ELECTROSTATICS */
 445             velec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_add_ps(rinv11,_mm256_mul_ps(krf,rsq11)),crf));
 446             felec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_mul_ps(rinv11,rinvsq11),krf2));
 447
 448             cutoff_mask      = _mm256_cmp_ps(rsq11,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,dx11);
 460             ty               = _mm256_mul_ps(fscal,dy11);
 461             tz               = _mm256_mul_ps(fscal,dz11);
 462
 463             /* Update vectorial force */
 464             fix1             = _mm256_add_ps(fix1,tx);
 465             fiy1             = _mm256_add_ps(fiy1,ty);
 466             fiz1             = _mm256_add_ps(fiz1,tz);
 467
 468             fjx1             = _mm256_add_ps(fjx1,tx);
 469             fjy1             = _mm256_add_ps(fjy1,ty);
 470             fjz1             = _mm256_add_ps(fjz1,tz);
 471
 472             }
 473
 474             /**************************
 475              * CALCULATE INTERACTIONS *
 476              **************************/
 477
 478             if (gmx_mm256_any_lt(rsq12,rcutoff2))
 479             {
 480
 481             /* REACTION-FIELD ELECTROSTATICS */
 482             velec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_add_ps(rinv12,_mm256_mul_ps(krf,rsq12)),crf));
 483             felec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_mul_ps(rinv12,rinvsq12),krf2));
 484
 485             cutoff_mask      = _mm256_cmp_ps(rsq12,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,dx12);
 497             ty               = _mm256_mul_ps(fscal,dy12);
 498             tz               = _mm256_mul_ps(fscal,dz12);
 499
 500             /* Update vectorial force */
 501             fix1             = _mm256_add_ps(fix1,tx);
 502             fiy1             = _mm256_add_ps(fiy1,ty);
 503             fiz1             = _mm256_add_ps(fiz1,tz);
 504
 505             fjx2             = _mm256_add_ps(fjx2,tx);
 506             fjy2             = _mm256_add_ps(fjy2,ty);
 507             fjz2             = _mm256_add_ps(fjz2,tz);
 508
 509             }
 510
 511             /**************************
 512              * CALCULATE INTERACTIONS *
 513              **************************/
 514
 515             if (gmx_mm256_any_lt(rsq20,rcutoff2))
 516             {
 517
 518             /* REACTION-FIELD ELECTROSTATICS */
 519             velec            = _mm256_mul_ps(qq20,_mm256_sub_ps(_mm256_add_ps(rinv20,_mm256_mul_ps(krf,rsq20)),crf));
 520             felec            = _mm256_mul_ps(qq20,_mm256_sub_ps(_mm256_mul_ps(rinv20,rinvsq20),krf2));
 521
 522             cutoff_mask      = _mm256_cmp_ps(rsq20,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,dx20);
 534             ty               = _mm256_mul_ps(fscal,dy20);
 535             tz               = _mm256_mul_ps(fscal,dz20);
 536
 537             /* Update vectorial force */
 538             fix2             = _mm256_add_ps(fix2,tx);
 539             fiy2             = _mm256_add_ps(fiy2,ty);
 540             fiz2             = _mm256_add_ps(fiz2,tz);
 541
 542             fjx0             = _mm256_add_ps(fjx0,tx);
 543             fjy0             = _mm256_add_ps(fjy0,ty);
 544             fjz0             = _mm256_add_ps(fjz0,tz);
 545
 546             }
 547
 548             /**************************
 549              * CALCULATE INTERACTIONS *
 550              **************************/
 551
 552             if (gmx_mm256_any_lt(rsq21,rcutoff2))
 553             {
 554
 555             /* REACTION-FIELD ELECTROSTATICS */
 556             velec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_add_ps(rinv21,_mm256_mul_ps(krf,rsq21)),crf));
 557             felec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_mul_ps(rinv21,rinvsq21),krf2));
 558
 559             cutoff_mask      = _mm256_cmp_ps(rsq21,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,dx21);
 571             ty               = _mm256_mul_ps(fscal,dy21);
 572             tz               = _mm256_mul_ps(fscal,dz21);
 573
 574             /* Update vectorial force */
 575             fix2             = _mm256_add_ps(fix2,tx);
 576             fiy2             = _mm256_add_ps(fiy2,ty);
 577             fiz2             = _mm256_add_ps(fiz2,tz);
 578
 579             fjx1             = _mm256_add_ps(fjx1,tx);
 580             fjy1             = _mm256_add_ps(fjy1,ty);
 581             fjz1             = _mm256_add_ps(fjz1,tz);
 582
 583             }
 584
 585             /**************************
 586              * CALCULATE INTERACTIONS *
 587              **************************/
 588
 589             if (gmx_mm256_any_lt(rsq22,rcutoff2))
 590             {
 591
 592             /* REACTION-FIELD ELECTROSTATICS */
 593             velec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_add_ps(rinv22,_mm256_mul_ps(krf,rsq22)),crf));
 594             felec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_mul_ps(rinv22,rinvsq22),krf2));
 595
 596             cutoff_mask      = _mm256_cmp_ps(rsq22,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,dx22);
 608             ty               = _mm256_mul_ps(fscal,dy22);
 609             tz               = _mm256_mul_ps(fscal,dz22);
 610
 611             /* Update vectorial force */
 612             fix2             = _mm256_add_ps(fix2,tx);
 613             fiy2             = _mm256_add_ps(fiy2,ty);
 614             fiz2             = _mm256_add_ps(fiz2,tz);
 615
 616             fjx2             = _mm256_add_ps(fjx2,tx);
 617             fjy2             = _mm256_add_ps(fjy2,ty);
 618             fjz2             = _mm256_add_ps(fjz2,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,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
 632                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 633
 634             /* Inner loop uses 324 flops */
 635         }
 636
 637         if(jidx<j_index_end)
 638         {
 639
 640             /* Get j neighbor index, and coordinate index */
 641             jnrlistA         = jjnr[jidx];
 642             jnrlistB         = jjnr[jidx+1];
 643             jnrlistC         = jjnr[jidx+2];
 644             jnrlistD         = jjnr[jidx+3];
 645             jnrlistE         = jjnr[jidx+4];
 646             jnrlistF         = jjnr[jidx+5];
 647             jnrlistG         = jjnr[jidx+6];
 648             jnrlistH         = jjnr[jidx+7];
 649             /* Sign of each element will be negative for non-real atoms.
 650              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 651              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 652              */
 653             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
 654                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
 655
 656             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 657             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 658             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 659             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 660             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
 661             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
 662             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
 663             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
 664             j_coord_offsetA  = DIM*jnrA;
 665             j_coord_offsetB  = DIM*jnrB;
 666             j_coord_offsetC  = DIM*jnrC;
 667             j_coord_offsetD  = DIM*jnrD;
 668             j_coord_offsetE  = DIM*jnrE;
 669             j_coord_offsetF  = DIM*jnrF;
 670             j_coord_offsetG  = DIM*jnrG;
 671             j_coord_offsetH  = DIM*jnrH;
 672
 673             /* load j atom coordinates */
 674             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 675                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 676                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 677                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 678                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 679
 680             /* Calculate displacement vector */
 681             dx00             = _mm256_sub_ps(ix0,jx0);
 682             dy00             = _mm256_sub_ps(iy0,jy0);
 683             dz00             = _mm256_sub_ps(iz0,jz0);
 684             dx01             = _mm256_sub_ps(ix0,jx1);
 685             dy01             = _mm256_sub_ps(iy0,jy1);
 686             dz01             = _mm256_sub_ps(iz0,jz1);
 687             dx02             = _mm256_sub_ps(ix0,jx2);
 688             dy02             = _mm256_sub_ps(iy0,jy2);
 689             dz02             = _mm256_sub_ps(iz0,jz2);
 690             dx10             = _mm256_sub_ps(ix1,jx0);
 691             dy10             = _mm256_sub_ps(iy1,jy0);
 692             dz10             = _mm256_sub_ps(iz1,jz0);
 693             dx11             = _mm256_sub_ps(ix1,jx1);
 694             dy11             = _mm256_sub_ps(iy1,jy1);
 695             dz11             = _mm256_sub_ps(iz1,jz1);
 696             dx12             = _mm256_sub_ps(ix1,jx2);
 697             dy12             = _mm256_sub_ps(iy1,jy2);
 698             dz12             = _mm256_sub_ps(iz1,jz2);
 699             dx20             = _mm256_sub_ps(ix2,jx0);
 700             dy20             = _mm256_sub_ps(iy2,jy0);
 701             dz20             = _mm256_sub_ps(iz2,jz0);
 702             dx21             = _mm256_sub_ps(ix2,jx1);
 703             dy21             = _mm256_sub_ps(iy2,jy1);
 704             dz21             = _mm256_sub_ps(iz2,jz1);
 705             dx22             = _mm256_sub_ps(ix2,jx2);
 706             dy22             = _mm256_sub_ps(iy2,jy2);
 707             dz22             = _mm256_sub_ps(iz2,jz2);
 708
 709             /* Calculate squared distance and things based on it */
 710             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 711             rsq01            = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
 712             rsq02            = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
 713             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
 714             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 715             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 716             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
 717             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 718             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 719
 720             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
 721             rinv01           = gmx_mm256_invsqrt_ps(rsq01);
 722             rinv02           = gmx_mm256_invsqrt_ps(rsq02);
 723             rinv10           = gmx_mm256_invsqrt_ps(rsq10);
 724             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 725             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 726             rinv20           = gmx_mm256_invsqrt_ps(rsq20);
 727             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 728             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 729
 730             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
 731             rinvsq01         = _mm256_mul_ps(rinv01,rinv01);
 732             rinvsq02         = _mm256_mul_ps(rinv02,rinv02);
 733             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
 734             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 735             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 736             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
 737             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 738             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 739
 740             fjx0             = _mm256_setzero_ps();
 741             fjy0             = _mm256_setzero_ps();
 742             fjz0             = _mm256_setzero_ps();
 743             fjx1             = _mm256_setzero_ps();
 744             fjy1             = _mm256_setzero_ps();
 745             fjz1             = _mm256_setzero_ps();
 746             fjx2             = _mm256_setzero_ps();
 747             fjy2             = _mm256_setzero_ps();
 748             fjz2             = _mm256_setzero_ps();
 749
 750             /**************************
 751              * CALCULATE INTERACTIONS *
 752              **************************/
 753
 754             if (gmx_mm256_any_lt(rsq00,rcutoff2))
 755             {
 756
 757             /* REACTION-FIELD ELECTROSTATICS */
 758             velec            = _mm256_mul_ps(qq00,_mm256_sub_ps(_mm256_add_ps(rinv00,_mm256_mul_ps(krf,rsq00)),crf));
 759             felec            = _mm256_mul_ps(qq00,_mm256_sub_ps(_mm256_mul_ps(rinv00,rinvsq00),krf2));
 760
 761             cutoff_mask      = _mm256_cmp_ps(rsq00,rcutoff2,_CMP_LT_OQ);
 762
 763             /* Update potential sum for this i atom from the interaction with this j atom. */
 764             velec            = _mm256_and_ps(velec,cutoff_mask);
 765             velec            = _mm256_andnot_ps(dummy_mask,velec);
 766             velecsum         = _mm256_add_ps(velecsum,velec);
 767
 768             fscal            = felec;
 769
 770             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 771
 772             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 773
 774             /* Calculate temporary vectorial force */
 775             tx               = _mm256_mul_ps(fscal,dx00);
 776             ty               = _mm256_mul_ps(fscal,dy00);
 777             tz               = _mm256_mul_ps(fscal,dz00);
 778
 779             /* Update vectorial force */
 780             fix0             = _mm256_add_ps(fix0,tx);
 781             fiy0             = _mm256_add_ps(fiy0,ty);
 782             fiz0             = _mm256_add_ps(fiz0,tz);
 783
 784             fjx0             = _mm256_add_ps(fjx0,tx);
 785             fjy0             = _mm256_add_ps(fjy0,ty);
 786             fjz0             = _mm256_add_ps(fjz0,tz);
 787
 788             }
 789
 790             /**************************
 791              * CALCULATE INTERACTIONS *
 792              **************************/
 793
 794             if (gmx_mm256_any_lt(rsq01,rcutoff2))
 795             {
 796
 797             /* REACTION-FIELD ELECTROSTATICS */
 798             velec            = _mm256_mul_ps(qq01,_mm256_sub_ps(_mm256_add_ps(rinv01,_mm256_mul_ps(krf,rsq01)),crf));
 799             felec            = _mm256_mul_ps(qq01,_mm256_sub_ps(_mm256_mul_ps(rinv01,rinvsq01),krf2));
 800
 801             cutoff_mask      = _mm256_cmp_ps(rsq01,rcutoff2,_CMP_LT_OQ);
 802
 803             /* Update potential sum for this i atom from the interaction with this j atom. */
 804             velec            = _mm256_and_ps(velec,cutoff_mask);
 805             velec            = _mm256_andnot_ps(dummy_mask,velec);
 806             velecsum         = _mm256_add_ps(velecsum,velec);
 807
 808             fscal            = felec;
 809
 810             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 811
 812             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 813
 814             /* Calculate temporary vectorial force */
 815             tx               = _mm256_mul_ps(fscal,dx01);
 816             ty               = _mm256_mul_ps(fscal,dy01);
 817             tz               = _mm256_mul_ps(fscal,dz01);
 818
 819             /* Update vectorial force */
 820             fix0             = _mm256_add_ps(fix0,tx);
 821             fiy0             = _mm256_add_ps(fiy0,ty);
 822             fiz0             = _mm256_add_ps(fiz0,tz);
 823
 824             fjx1             = _mm256_add_ps(fjx1,tx);
 825             fjy1             = _mm256_add_ps(fjy1,ty);
 826             fjz1             = _mm256_add_ps(fjz1,tz);
 827
 828             }
 829
 830             /**************************
 831              * CALCULATE INTERACTIONS *
 832              **************************/
 833
 834             if (gmx_mm256_any_lt(rsq02,rcutoff2))
 835             {
 836
 837             /* REACTION-FIELD ELECTROSTATICS */
 838             velec            = _mm256_mul_ps(qq02,_mm256_sub_ps(_mm256_add_ps(rinv02,_mm256_mul_ps(krf,rsq02)),crf));
 839             felec            = _mm256_mul_ps(qq02,_mm256_sub_ps(_mm256_mul_ps(rinv02,rinvsq02),krf2));
 840
 841             cutoff_mask      = _mm256_cmp_ps(rsq02,rcutoff2,_CMP_LT_OQ);
 842
 843             /* Update potential sum for this i atom from the interaction with this j atom. */
 844             velec            = _mm256_and_ps(velec,cutoff_mask);
 845             velec            = _mm256_andnot_ps(dummy_mask,velec);
 846             velecsum         = _mm256_add_ps(velecsum,velec);
 847
 848             fscal            = felec;
 849
 850             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 851
 852             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 853
 854             /* Calculate temporary vectorial force */
 855             tx               = _mm256_mul_ps(fscal,dx02);
 856             ty               = _mm256_mul_ps(fscal,dy02);
 857             tz               = _mm256_mul_ps(fscal,dz02);
 858
 859             /* Update vectorial force */
 860             fix0             = _mm256_add_ps(fix0,tx);
 861             fiy0             = _mm256_add_ps(fiy0,ty);
 862             fiz0             = _mm256_add_ps(fiz0,tz);
 863
 864             fjx2             = _mm256_add_ps(fjx2,tx);
 865             fjy2             = _mm256_add_ps(fjy2,ty);
 866             fjz2             = _mm256_add_ps(fjz2,tz);
 867
 868             }
 869
 870             /**************************
 871              * CALCULATE INTERACTIONS *
 872              **************************/
 873
 874             if (gmx_mm256_any_lt(rsq10,rcutoff2))
 875             {
 876
 877             /* REACTION-FIELD ELECTROSTATICS */
 878             velec            = _mm256_mul_ps(qq10,_mm256_sub_ps(_mm256_add_ps(rinv10,_mm256_mul_ps(krf,rsq10)),crf));
 879             felec            = _mm256_mul_ps(qq10,_mm256_sub_ps(_mm256_mul_ps(rinv10,rinvsq10),krf2));
 880
 881             cutoff_mask      = _mm256_cmp_ps(rsq10,rcutoff2,_CMP_LT_OQ);
 882
 883             /* Update potential sum for this i atom from the interaction with this j atom. */
 884             velec            = _mm256_and_ps(velec,cutoff_mask);
 885             velec            = _mm256_andnot_ps(dummy_mask,velec);
 886             velecsum         = _mm256_add_ps(velecsum,velec);
 887
 888             fscal            = felec;
 889
 890             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 891
 892             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 893
 894             /* Calculate temporary vectorial force */
 895             tx               = _mm256_mul_ps(fscal,dx10);
 896             ty               = _mm256_mul_ps(fscal,dy10);
 897             tz               = _mm256_mul_ps(fscal,dz10);
 898
 899             /* Update vectorial force */
 900             fix1             = _mm256_add_ps(fix1,tx);
 901             fiy1             = _mm256_add_ps(fiy1,ty);
 902             fiz1             = _mm256_add_ps(fiz1,tz);
 903
 904             fjx0             = _mm256_add_ps(fjx0,tx);
 905             fjy0             = _mm256_add_ps(fjy0,ty);
 906             fjz0             = _mm256_add_ps(fjz0,tz);
 907
 908             }
 909
 910             /**************************
 911              * CALCULATE INTERACTIONS *
 912              **************************/
 913
 914             if (gmx_mm256_any_lt(rsq11,rcutoff2))
 915             {
 916
 917             /* REACTION-FIELD ELECTROSTATICS */
 918             velec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_add_ps(rinv11,_mm256_mul_ps(krf,rsq11)),crf));
 919             felec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_mul_ps(rinv11,rinvsq11),krf2));
 920
 921             cutoff_mask      = _mm256_cmp_ps(rsq11,rcutoff2,_CMP_LT_OQ);
 922
 923             /* Update potential sum for this i atom from the interaction with this j atom. */
 924             velec            = _mm256_and_ps(velec,cutoff_mask);
 925             velec            = _mm256_andnot_ps(dummy_mask,velec);
 926             velecsum         = _mm256_add_ps(velecsum,velec);
 927
 928             fscal            = felec;
 929
 930             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 931
 932             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 933
 934             /* Calculate temporary vectorial force */
 935             tx               = _mm256_mul_ps(fscal,dx11);
 936             ty               = _mm256_mul_ps(fscal,dy11);
 937             tz               = _mm256_mul_ps(fscal,dz11);
 938
 939             /* Update vectorial force */
 940             fix1             = _mm256_add_ps(fix1,tx);
 941             fiy1             = _mm256_add_ps(fiy1,ty);
 942             fiz1             = _mm256_add_ps(fiz1,tz);
 943
 944             fjx1             = _mm256_add_ps(fjx1,tx);
 945             fjy1             = _mm256_add_ps(fjy1,ty);
 946             fjz1             = _mm256_add_ps(fjz1,tz);
 947
 948             }
 949
 950             /**************************
 951              * CALCULATE INTERACTIONS *
 952              **************************/
 953
 954             if (gmx_mm256_any_lt(rsq12,rcutoff2))
 955             {
 956
 957             /* REACTION-FIELD ELECTROSTATICS */
 958             velec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_add_ps(rinv12,_mm256_mul_ps(krf,rsq12)),crf));
 959             felec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_mul_ps(rinv12,rinvsq12),krf2));
 960
 961             cutoff_mask      = _mm256_cmp_ps(rsq12,rcutoff2,_CMP_LT_OQ);
 962
 963             /* Update potential sum for this i atom from the interaction with this j atom. */
 964             velec            = _mm256_and_ps(velec,cutoff_mask);
 965             velec            = _mm256_andnot_ps(dummy_mask,velec);
 966             velecsum         = _mm256_add_ps(velecsum,velec);
 967
 968             fscal            = felec;
 969
 970             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 971
 972             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 973
 974             /* Calculate temporary vectorial force */
 975             tx               = _mm256_mul_ps(fscal,dx12);
 976             ty               = _mm256_mul_ps(fscal,dy12);
 977             tz               = _mm256_mul_ps(fscal,dz12);
 978
 979             /* Update vectorial force */
 980             fix1             = _mm256_add_ps(fix1,tx);
 981             fiy1             = _mm256_add_ps(fiy1,ty);
 982             fiz1             = _mm256_add_ps(fiz1,tz);
 983
 984             fjx2             = _mm256_add_ps(fjx2,tx);
 985             fjy2             = _mm256_add_ps(fjy2,ty);
 986             fjz2             = _mm256_add_ps(fjz2,tz);
 987
 988             }
 989
 990             /**************************
 991              * CALCULATE INTERACTIONS *
 992              **************************/
 993
 994             if (gmx_mm256_any_lt(rsq20,rcutoff2))
 995             {
 996
 997             /* REACTION-FIELD ELECTROSTATICS */
 998             velec            = _mm256_mul_ps(qq20,_mm256_sub_ps(_mm256_add_ps(rinv20,_mm256_mul_ps(krf,rsq20)),crf));
 999             felec            = _mm256_mul_ps(qq20,_mm256_sub_ps(_mm256_mul_ps(rinv20,rinvsq20),krf2));
1000
1001             cutoff_mask      = _mm256_cmp_ps(rsq20,rcutoff2,_CMP_LT_OQ);
1002
1003             /* Update potential sum for this i atom from the interaction with this j atom. */
1004             velec            = _mm256_and_ps(velec,cutoff_mask);
1005             velec            = _mm256_andnot_ps(dummy_mask,velec);
1006             velecsum         = _mm256_add_ps(velecsum,velec);
1007
1008             fscal            = felec;
1009
1010             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1011
1012             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1013
1014             /* Calculate temporary vectorial force */
1015             tx               = _mm256_mul_ps(fscal,dx20);
1016             ty               = _mm256_mul_ps(fscal,dy20);
1017             tz               = _mm256_mul_ps(fscal,dz20);
1018
1019             /* Update vectorial force */
1020             fix2             = _mm256_add_ps(fix2,tx);
1021             fiy2             = _mm256_add_ps(fiy2,ty);
1022             fiz2             = _mm256_add_ps(fiz2,tz);
1023
1024             fjx0             = _mm256_add_ps(fjx0,tx);
1025             fjy0             = _mm256_add_ps(fjy0,ty);
1026             fjz0             = _mm256_add_ps(fjz0,tz);
1027
1028             }
1029
1030             /**************************
1031              * CALCULATE INTERACTIONS *
1032              **************************/
1033
1034             if (gmx_mm256_any_lt(rsq21,rcutoff2))
1035             {
1036
1037             /* REACTION-FIELD ELECTROSTATICS */
1038             velec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_add_ps(rinv21,_mm256_mul_ps(krf,rsq21)),crf));
1039             felec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_mul_ps(rinv21,rinvsq21),krf2));
1040
1041             cutoff_mask      = _mm256_cmp_ps(rsq21,rcutoff2,_CMP_LT_OQ);
1042
1043             /* Update potential sum for this i atom from the interaction with this j atom. */
1044             velec            = _mm256_and_ps(velec,cutoff_mask);
1045             velec            = _mm256_andnot_ps(dummy_mask,velec);
1046             velecsum         = _mm256_add_ps(velecsum,velec);
1047
1048             fscal            = felec;
1049
1050             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1051
1052             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1053
1054             /* Calculate temporary vectorial force */
1055             tx               = _mm256_mul_ps(fscal,dx21);
1056             ty               = _mm256_mul_ps(fscal,dy21);
1057             tz               = _mm256_mul_ps(fscal,dz21);
1058
1059             /* Update vectorial force */
1060             fix2             = _mm256_add_ps(fix2,tx);
1061             fiy2             = _mm256_add_ps(fiy2,ty);
1062             fiz2             = _mm256_add_ps(fiz2,tz);
1063
1064             fjx1             = _mm256_add_ps(fjx1,tx);
1065             fjy1             = _mm256_add_ps(fjy1,ty);
1066             fjz1             = _mm256_add_ps(fjz1,tz);
1067
1068             }
1069
1070             /**************************
1071              * CALCULATE INTERACTIONS *
1072              **************************/
1073
1074             if (gmx_mm256_any_lt(rsq22,rcutoff2))
1075             {
1076
1077             /* REACTION-FIELD ELECTROSTATICS */
1078             velec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_add_ps(rinv22,_mm256_mul_ps(krf,rsq22)),crf));
1079             felec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_mul_ps(rinv22,rinvsq22),krf2));
1080
1081             cutoff_mask      = _mm256_cmp_ps(rsq22,rcutoff2,_CMP_LT_OQ);
1082
1083             /* Update potential sum for this i atom from the interaction with this j atom. */
1084             velec            = _mm256_and_ps(velec,cutoff_mask);
1085             velec            = _mm256_andnot_ps(dummy_mask,velec);
1086             velecsum         = _mm256_add_ps(velecsum,velec);
1087
1088             fscal            = felec;
1089
1090             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1091
1092             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1093
1094             /* Calculate temporary vectorial force */
1095             tx               = _mm256_mul_ps(fscal,dx22);
1096             ty               = _mm256_mul_ps(fscal,dy22);
1097             tz               = _mm256_mul_ps(fscal,dz22);
1098
1099             /* Update vectorial force */
1100             fix2             = _mm256_add_ps(fix2,tx);
1101             fiy2             = _mm256_add_ps(fiy2,ty);
1102             fiz2             = _mm256_add_ps(fiz2,tz);
1103
1104             fjx2             = _mm256_add_ps(fjx2,tx);
1105             fjy2             = _mm256_add_ps(fjy2,ty);
1106             fjz2             = _mm256_add_ps(fjz2,tz);
1107
1108             }
1109
1110             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1111             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1112             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1113             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1114             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1115             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1116             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
1117             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
1118
1119             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1120                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1121
1122             /* Inner loop uses 324 flops */
1123         }
1124
1125         /* End of innermost loop */
1126
1127         gmx_mm256_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1128                                                  f+i_coord_offset,fshift+i_shift_offset);
1129
1130         ggid                        = gid[iidx];
1131         /* Update potential energies */
1132         gmx_mm256_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
1133
1134         /* Increment number of inner iterations */
1135         inneriter                  += j_index_end - j_index_start;
1136
1137         /* Outer loop uses 19 flops */
1138     }
1139
1140     /* Increment number of outer iterations */
1141     outeriter        += nri;
1142
1143     /* Update outer/inner flops */
1144
1145     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3W3_VF,outeriter*19 + inneriter*324);
1146 }
1147 /*
1148  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwNone_GeomW3W3_F_avx_256_single
1149  * Electrostatics interaction: ReactionField
1150  * VdW interaction:            None
1151  * Geometry:                   Water3-Water3
1152  * Calculate force/pot:        Force
1153  */
1154 void
1155 nb_kernel_ElecRFCut_VdwNone_GeomW3W3_F_avx_256_single
1156                     (t_nblist * gmx_restrict                nlist,
1157                      rvec * gmx_restrict                    xx,
1158                      rvec * gmx_restrict                    ff,
1159                      t_forcerec * gmx_restrict              fr,
1160                      t_mdatoms * gmx_restrict               mdatoms,
1161                      nb_kernel_data_t * gmx_restrict        kernel_data,
1162                      t_nrnb * gmx_restrict                  nrnb)
1163 {
1164     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1165      * just 0 for non-waters.
1166      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
1167      * jnr indices corresponding to data put in the four positions in the SIMD register.
1168      */
1169     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1170     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1171     int              jnrA,jnrB,jnrC,jnrD;
1172     int              jnrE,jnrF,jnrG,jnrH;
1173     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1174     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1175     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1176     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
1177     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1178     real             rcutoff_scalar;
1179     real             *shiftvec,*fshift,*x,*f;
1180     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
1181     real             scratch[4*DIM];
1182     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1183     real *           vdwioffsetptr0;
1184     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
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     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
1190     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
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     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1196     __m256           dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1197     __m256           dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1198     __m256           dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1199     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1200     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1201     __m256           dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1202     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1203     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1204     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
1205     real             *charge;
1206     __m256           dummy_mask,cutoff_mask;
1207     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
1208     __m256           one     = _mm256_set1_ps(1.0);
1209     __m256           two     = _mm256_set1_ps(2.0);
1210     x                = xx[0];
1211     f                = ff[0];
1212
1213     nri              = nlist->nri;
1214     iinr             = nlist->iinr;
1215     jindex           = nlist->jindex;
1216     jjnr             = nlist->jjnr;
1217     shiftidx         = nlist->shift;
1218     gid              = nlist->gid;
1219     shiftvec         = fr->shift_vec[0];
1220     fshift           = fr->fshift[0];
1221     facel            = _mm256_set1_ps(fr->epsfac);
1222     charge           = mdatoms->chargeA;
1223     krf              = _mm256_set1_ps(fr->ic->k_rf);
1224     krf2             = _mm256_set1_ps(fr->ic->k_rf*2.0);
1225     crf              = _mm256_set1_ps(fr->ic->c_rf);
1226
1227     /* Setup water-specific parameters */
1228     inr              = nlist->iinr[0];
1229     iq0              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+0]));
1230     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
1231     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
1232
1233     jq0              = _mm256_set1_ps(charge[inr+0]);
1234     jq1              = _mm256_set1_ps(charge[inr+1]);
1235     jq2              = _mm256_set1_ps(charge[inr+2]);
1236     qq00             = _mm256_mul_ps(iq0,jq0);
1237     qq01             = _mm256_mul_ps(iq0,jq1);
1238     qq02             = _mm256_mul_ps(iq0,jq2);
1239     qq10             = _mm256_mul_ps(iq1,jq0);
1240     qq11             = _mm256_mul_ps(iq1,jq1);
1241     qq12             = _mm256_mul_ps(iq1,jq2);
1242     qq20             = _mm256_mul_ps(iq2,jq0);
1243     qq21             = _mm256_mul_ps(iq2,jq1);
1244     qq22             = _mm256_mul_ps(iq2,jq2);
1245
1246     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
1247     rcutoff_scalar   = fr->rcoulomb;
1248     rcutoff          = _mm256_set1_ps(rcutoff_scalar);
1249     rcutoff2         = _mm256_mul_ps(rcutoff,rcutoff);
1250
1251     /* Avoid stupid compiler warnings */
1252     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
1253     j_coord_offsetA = 0;
1254     j_coord_offsetB = 0;
1255     j_coord_offsetC = 0;
1256     j_coord_offsetD = 0;
1257     j_coord_offsetE = 0;
1258     j_coord_offsetF = 0;
1259     j_coord_offsetG = 0;
1260     j_coord_offsetH = 0;
1261
1262     outeriter        = 0;
1263     inneriter        = 0;
1264
1265     for(iidx=0;iidx<4*DIM;iidx++)
1266     {
1267         scratch[iidx] = 0.0;
1268     }
1269
1270     /* Start outer loop over neighborlists */
1271     for(iidx=0; iidx<nri; iidx++)
1272     {
1273         /* Load shift vector for this list */
1274         i_shift_offset   = DIM*shiftidx[iidx];
1275
1276         /* Load limits for loop over neighbors */
1277         j_index_start    = jindex[iidx];
1278         j_index_end      = jindex[iidx+1];
1279
1280         /* Get outer coordinate index */
1281         inr              = iinr[iidx];
1282         i_coord_offset   = DIM*inr;
1283
1284         /* Load i particle coords and add shift vector */
1285         gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1286                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1287
1288         fix0             = _mm256_setzero_ps();
1289         fiy0             = _mm256_setzero_ps();
1290         fiz0             = _mm256_setzero_ps();
1291         fix1             = _mm256_setzero_ps();
1292         fiy1             = _mm256_setzero_ps();
1293         fiz1             = _mm256_setzero_ps();
1294         fix2             = _mm256_setzero_ps();
1295         fiy2             = _mm256_setzero_ps();
1296         fiz2             = _mm256_setzero_ps();
1297
1298         /* Start inner kernel loop */
1299         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
1300         {
1301
1302             /* Get j neighbor index, and coordinate index */
1303             jnrA             = jjnr[jidx];
1304             jnrB             = jjnr[jidx+1];
1305             jnrC             = jjnr[jidx+2];
1306             jnrD             = jjnr[jidx+3];
1307             jnrE             = jjnr[jidx+4];
1308             jnrF             = jjnr[jidx+5];
1309             jnrG             = jjnr[jidx+6];
1310             jnrH             = jjnr[jidx+7];
1311             j_coord_offsetA  = DIM*jnrA;
1312             j_coord_offsetB  = DIM*jnrB;
1313             j_coord_offsetC  = DIM*jnrC;
1314             j_coord_offsetD  = DIM*jnrD;
1315             j_coord_offsetE  = DIM*jnrE;
1316             j_coord_offsetF  = DIM*jnrF;
1317             j_coord_offsetG  = DIM*jnrG;
1318             j_coord_offsetH  = DIM*jnrH;
1319
1320             /* load j atom coordinates */
1321             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1322                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1323                                                  x+j_coord_offsetE,x+j_coord_offsetF,
1324                                                  x+j_coord_offsetG,x+j_coord_offsetH,
1325                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1326
1327             /* Calculate displacement vector */
1328             dx00             = _mm256_sub_ps(ix0,jx0);
1329             dy00             = _mm256_sub_ps(iy0,jy0);
1330             dz00             = _mm256_sub_ps(iz0,jz0);
1331             dx01             = _mm256_sub_ps(ix0,jx1);
1332             dy01             = _mm256_sub_ps(iy0,jy1);
1333             dz01             = _mm256_sub_ps(iz0,jz1);
1334             dx02             = _mm256_sub_ps(ix0,jx2);
1335             dy02             = _mm256_sub_ps(iy0,jy2);
1336             dz02             = _mm256_sub_ps(iz0,jz2);
1337             dx10             = _mm256_sub_ps(ix1,jx0);
1338             dy10             = _mm256_sub_ps(iy1,jy0);
1339             dz10             = _mm256_sub_ps(iz1,jz0);
1340             dx11             = _mm256_sub_ps(ix1,jx1);
1341             dy11             = _mm256_sub_ps(iy1,jy1);
1342             dz11             = _mm256_sub_ps(iz1,jz1);
1343             dx12             = _mm256_sub_ps(ix1,jx2);
1344             dy12             = _mm256_sub_ps(iy1,jy2);
1345             dz12             = _mm256_sub_ps(iz1,jz2);
1346             dx20             = _mm256_sub_ps(ix2,jx0);
1347             dy20             = _mm256_sub_ps(iy2,jy0);
1348             dz20             = _mm256_sub_ps(iz2,jz0);
1349             dx21             = _mm256_sub_ps(ix2,jx1);
1350             dy21             = _mm256_sub_ps(iy2,jy1);
1351             dz21             = _mm256_sub_ps(iz2,jz1);
1352             dx22             = _mm256_sub_ps(ix2,jx2);
1353             dy22             = _mm256_sub_ps(iy2,jy2);
1354             dz22             = _mm256_sub_ps(iz2,jz2);
1355
1356             /* Calculate squared distance and things based on it */
1357             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1358             rsq01            = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
1359             rsq02            = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
1360             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
1361             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1362             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1363             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
1364             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1365             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1366
1367             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
1368             rinv01           = gmx_mm256_invsqrt_ps(rsq01);
1369             rinv02           = gmx_mm256_invsqrt_ps(rsq02);
1370             rinv10           = gmx_mm256_invsqrt_ps(rsq10);
1371             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1372             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1373             rinv20           = gmx_mm256_invsqrt_ps(rsq20);
1374             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1375             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1376
1377             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
1378             rinvsq01         = _mm256_mul_ps(rinv01,rinv01);
1379             rinvsq02         = _mm256_mul_ps(rinv02,rinv02);
1380             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
1381             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1382             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1383             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
1384             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1385             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1386
1387             fjx0             = _mm256_setzero_ps();
1388             fjy0             = _mm256_setzero_ps();
1389             fjz0             = _mm256_setzero_ps();
1390             fjx1             = _mm256_setzero_ps();
1391             fjy1             = _mm256_setzero_ps();
1392             fjz1             = _mm256_setzero_ps();
1393             fjx2             = _mm256_setzero_ps();
1394             fjy2             = _mm256_setzero_ps();
1395             fjz2             = _mm256_setzero_ps();
1396
1397             /**************************
1398              * CALCULATE INTERACTIONS *
1399              **************************/
1400
1401             if (gmx_mm256_any_lt(rsq00,rcutoff2))
1402             {
1403
1404             /* REACTION-FIELD ELECTROSTATICS */
1405             felec            = _mm256_mul_ps(qq00,_mm256_sub_ps(_mm256_mul_ps(rinv00,rinvsq00),krf2));
1406
1407             cutoff_mask      = _mm256_cmp_ps(rsq00,rcutoff2,_CMP_LT_OQ);
1408
1409             fscal            = felec;
1410
1411             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1412
1413             /* Calculate temporary vectorial force */
1414             tx               = _mm256_mul_ps(fscal,dx00);
1415             ty               = _mm256_mul_ps(fscal,dy00);
1416             tz               = _mm256_mul_ps(fscal,dz00);
1417
1418             /* Update vectorial force */
1419             fix0             = _mm256_add_ps(fix0,tx);
1420             fiy0             = _mm256_add_ps(fiy0,ty);
1421             fiz0             = _mm256_add_ps(fiz0,tz);
1422
1423             fjx0             = _mm256_add_ps(fjx0,tx);
1424             fjy0             = _mm256_add_ps(fjy0,ty);
1425             fjz0             = _mm256_add_ps(fjz0,tz);
1426
1427             }
1428
1429             /**************************
1430              * CALCULATE INTERACTIONS *
1431              **************************/
1432
1433             if (gmx_mm256_any_lt(rsq01,rcutoff2))
1434             {
1435
1436             /* REACTION-FIELD ELECTROSTATICS */
1437             felec            = _mm256_mul_ps(qq01,_mm256_sub_ps(_mm256_mul_ps(rinv01,rinvsq01),krf2));
1438
1439             cutoff_mask      = _mm256_cmp_ps(rsq01,rcutoff2,_CMP_LT_OQ);
1440
1441             fscal            = felec;
1442
1443             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1444
1445             /* Calculate temporary vectorial force */
1446             tx               = _mm256_mul_ps(fscal,dx01);
1447             ty               = _mm256_mul_ps(fscal,dy01);
1448             tz               = _mm256_mul_ps(fscal,dz01);
1449
1450             /* Update vectorial force */
1451             fix0             = _mm256_add_ps(fix0,tx);
1452             fiy0             = _mm256_add_ps(fiy0,ty);
1453             fiz0             = _mm256_add_ps(fiz0,tz);
1454
1455             fjx1             = _mm256_add_ps(fjx1,tx);
1456             fjy1             = _mm256_add_ps(fjy1,ty);
1457             fjz1             = _mm256_add_ps(fjz1,tz);
1458
1459             }
1460
1461             /**************************
1462              * CALCULATE INTERACTIONS *
1463              **************************/
1464
1465             if (gmx_mm256_any_lt(rsq02,rcutoff2))
1466             {
1467
1468             /* REACTION-FIELD ELECTROSTATICS */
1469             felec            = _mm256_mul_ps(qq02,_mm256_sub_ps(_mm256_mul_ps(rinv02,rinvsq02),krf2));
1470
1471             cutoff_mask      = _mm256_cmp_ps(rsq02,rcutoff2,_CMP_LT_OQ);
1472
1473             fscal            = felec;
1474
1475             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1476
1477             /* Calculate temporary vectorial force */
1478             tx               = _mm256_mul_ps(fscal,dx02);
1479             ty               = _mm256_mul_ps(fscal,dy02);
1480             tz               = _mm256_mul_ps(fscal,dz02);
1481
1482             /* Update vectorial force */
1483             fix0             = _mm256_add_ps(fix0,tx);
1484             fiy0             = _mm256_add_ps(fiy0,ty);
1485             fiz0             = _mm256_add_ps(fiz0,tz);
1486
1487             fjx2             = _mm256_add_ps(fjx2,tx);
1488             fjy2             = _mm256_add_ps(fjy2,ty);
1489             fjz2             = _mm256_add_ps(fjz2,tz);
1490
1491             }
1492
1493             /**************************
1494              * CALCULATE INTERACTIONS *
1495              **************************/
1496
1497             if (gmx_mm256_any_lt(rsq10,rcutoff2))
1498             {
1499
1500             /* REACTION-FIELD ELECTROSTATICS */
1501             felec            = _mm256_mul_ps(qq10,_mm256_sub_ps(_mm256_mul_ps(rinv10,rinvsq10),krf2));
1502
1503             cutoff_mask      = _mm256_cmp_ps(rsq10,rcutoff2,_CMP_LT_OQ);
1504
1505             fscal            = felec;
1506
1507             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1508
1509             /* Calculate temporary vectorial force */
1510             tx               = _mm256_mul_ps(fscal,dx10);
1511             ty               = _mm256_mul_ps(fscal,dy10);
1512             tz               = _mm256_mul_ps(fscal,dz10);
1513
1514             /* Update vectorial force */
1515             fix1             = _mm256_add_ps(fix1,tx);
1516             fiy1             = _mm256_add_ps(fiy1,ty);
1517             fiz1             = _mm256_add_ps(fiz1,tz);
1518
1519             fjx0             = _mm256_add_ps(fjx0,tx);
1520             fjy0             = _mm256_add_ps(fjy0,ty);
1521             fjz0             = _mm256_add_ps(fjz0,tz);
1522
1523             }
1524
1525             /**************************
1526              * CALCULATE INTERACTIONS *
1527              **************************/
1528
1529             if (gmx_mm256_any_lt(rsq11,rcutoff2))
1530             {
1531
1532             /* REACTION-FIELD ELECTROSTATICS */
1533             felec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_mul_ps(rinv11,rinvsq11),krf2));
1534
1535             cutoff_mask      = _mm256_cmp_ps(rsq11,rcutoff2,_CMP_LT_OQ);
1536
1537             fscal            = felec;
1538
1539             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1540
1541             /* Calculate temporary vectorial force */
1542             tx               = _mm256_mul_ps(fscal,dx11);
1543             ty               = _mm256_mul_ps(fscal,dy11);
1544             tz               = _mm256_mul_ps(fscal,dz11);
1545
1546             /* Update vectorial force */
1547             fix1             = _mm256_add_ps(fix1,tx);
1548             fiy1             = _mm256_add_ps(fiy1,ty);
1549             fiz1             = _mm256_add_ps(fiz1,tz);
1550
1551             fjx1             = _mm256_add_ps(fjx1,tx);
1552             fjy1             = _mm256_add_ps(fjy1,ty);
1553             fjz1             = _mm256_add_ps(fjz1,tz);
1554
1555             }
1556
1557             /**************************
1558              * CALCULATE INTERACTIONS *
1559              **************************/
1560
1561             if (gmx_mm256_any_lt(rsq12,rcutoff2))
1562             {
1563
1564             /* REACTION-FIELD ELECTROSTATICS */
1565             felec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_mul_ps(rinv12,rinvsq12),krf2));
1566
1567             cutoff_mask      = _mm256_cmp_ps(rsq12,rcutoff2,_CMP_LT_OQ);
1568
1569             fscal            = felec;
1570
1571             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1572
1573             /* Calculate temporary vectorial force */
1574             tx               = _mm256_mul_ps(fscal,dx12);
1575             ty               = _mm256_mul_ps(fscal,dy12);
1576             tz               = _mm256_mul_ps(fscal,dz12);
1577
1578             /* Update vectorial force */
1579             fix1             = _mm256_add_ps(fix1,tx);
1580             fiy1             = _mm256_add_ps(fiy1,ty);
1581             fiz1             = _mm256_add_ps(fiz1,tz);
1582
1583             fjx2             = _mm256_add_ps(fjx2,tx);
1584             fjy2             = _mm256_add_ps(fjy2,ty);
1585             fjz2             = _mm256_add_ps(fjz2,tz);
1586
1587             }
1588
1589             /**************************
1590              * CALCULATE INTERACTIONS *
1591              **************************/
1592
1593             if (gmx_mm256_any_lt(rsq20,rcutoff2))
1594             {
1595
1596             /* REACTION-FIELD ELECTROSTATICS */
1597             felec            = _mm256_mul_ps(qq20,_mm256_sub_ps(_mm256_mul_ps(rinv20,rinvsq20),krf2));
1598
1599             cutoff_mask      = _mm256_cmp_ps(rsq20,rcutoff2,_CMP_LT_OQ);
1600
1601             fscal            = felec;
1602
1603             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1604
1605             /* Calculate temporary vectorial force */
1606             tx               = _mm256_mul_ps(fscal,dx20);
1607             ty               = _mm256_mul_ps(fscal,dy20);
1608             tz               = _mm256_mul_ps(fscal,dz20);
1609
1610             /* Update vectorial force */
1611             fix2             = _mm256_add_ps(fix2,tx);
1612             fiy2             = _mm256_add_ps(fiy2,ty);
1613             fiz2             = _mm256_add_ps(fiz2,tz);
1614
1615             fjx0             = _mm256_add_ps(fjx0,tx);
1616             fjy0             = _mm256_add_ps(fjy0,ty);
1617             fjz0             = _mm256_add_ps(fjz0,tz);
1618
1619             }
1620
1621             /**************************
1622              * CALCULATE INTERACTIONS *
1623              **************************/
1624
1625             if (gmx_mm256_any_lt(rsq21,rcutoff2))
1626             {
1627
1628             /* REACTION-FIELD ELECTROSTATICS */
1629             felec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_mul_ps(rinv21,rinvsq21),krf2));
1630
1631             cutoff_mask      = _mm256_cmp_ps(rsq21,rcutoff2,_CMP_LT_OQ);
1632
1633             fscal            = felec;
1634
1635             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1636
1637             /* Calculate temporary vectorial force */
1638             tx               = _mm256_mul_ps(fscal,dx21);
1639             ty               = _mm256_mul_ps(fscal,dy21);
1640             tz               = _mm256_mul_ps(fscal,dz21);
1641
1642             /* Update vectorial force */
1643             fix2             = _mm256_add_ps(fix2,tx);
1644             fiy2             = _mm256_add_ps(fiy2,ty);
1645             fiz2             = _mm256_add_ps(fiz2,tz);
1646
1647             fjx1             = _mm256_add_ps(fjx1,tx);
1648             fjy1             = _mm256_add_ps(fjy1,ty);
1649             fjz1             = _mm256_add_ps(fjz1,tz);
1650
1651             }
1652
1653             /**************************
1654              * CALCULATE INTERACTIONS *
1655              **************************/
1656
1657             if (gmx_mm256_any_lt(rsq22,rcutoff2))
1658             {
1659
1660             /* REACTION-FIELD ELECTROSTATICS */
1661             felec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_mul_ps(rinv22,rinvsq22),krf2));
1662
1663             cutoff_mask      = _mm256_cmp_ps(rsq22,rcutoff2,_CMP_LT_OQ);
1664
1665             fscal            = felec;
1666
1667             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1668
1669             /* Calculate temporary vectorial force */
1670             tx               = _mm256_mul_ps(fscal,dx22);
1671             ty               = _mm256_mul_ps(fscal,dy22);
1672             tz               = _mm256_mul_ps(fscal,dz22);
1673
1674             /* Update vectorial force */
1675             fix2             = _mm256_add_ps(fix2,tx);
1676             fiy2             = _mm256_add_ps(fiy2,ty);
1677             fiz2             = _mm256_add_ps(fiz2,tz);
1678
1679             fjx2             = _mm256_add_ps(fjx2,tx);
1680             fjy2             = _mm256_add_ps(fjy2,ty);
1681             fjz2             = _mm256_add_ps(fjz2,tz);
1682
1683             }
1684
1685             fjptrA             = f+j_coord_offsetA;
1686             fjptrB             = f+j_coord_offsetB;
1687             fjptrC             = f+j_coord_offsetC;
1688             fjptrD             = f+j_coord_offsetD;
1689             fjptrE             = f+j_coord_offsetE;
1690             fjptrF             = f+j_coord_offsetF;
1691             fjptrG             = f+j_coord_offsetG;
1692             fjptrH             = f+j_coord_offsetH;
1693
1694             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1695                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1696
1697             /* Inner loop uses 270 flops */
1698         }
1699
1700         if(jidx<j_index_end)
1701         {
1702
1703             /* Get j neighbor index, and coordinate index */
1704             jnrlistA         = jjnr[jidx];
1705             jnrlistB         = jjnr[jidx+1];
1706             jnrlistC         = jjnr[jidx+2];
1707             jnrlistD         = jjnr[jidx+3];
1708             jnrlistE         = jjnr[jidx+4];
1709             jnrlistF         = jjnr[jidx+5];
1710             jnrlistG         = jjnr[jidx+6];
1711             jnrlistH         = jjnr[jidx+7];
1712             /* Sign of each element will be negative for non-real atoms.
1713              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1714              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1715              */
1716             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
1717                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
1718
1719             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1720             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1721             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1722             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1723             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
1724             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
1725             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
1726             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
1727             j_coord_offsetA  = DIM*jnrA;
1728             j_coord_offsetB  = DIM*jnrB;
1729             j_coord_offsetC  = DIM*jnrC;
1730             j_coord_offsetD  = DIM*jnrD;
1731             j_coord_offsetE  = DIM*jnrE;
1732             j_coord_offsetF  = DIM*jnrF;
1733             j_coord_offsetG  = DIM*jnrG;
1734             j_coord_offsetH  = DIM*jnrH;
1735
1736             /* load j atom coordinates */
1737             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1738                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1739                                                  x+j_coord_offsetE,x+j_coord_offsetF,
1740                                                  x+j_coord_offsetG,x+j_coord_offsetH,
1741                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1742
1743             /* Calculate displacement vector */
1744             dx00             = _mm256_sub_ps(ix0,jx0);
1745             dy00             = _mm256_sub_ps(iy0,jy0);
1746             dz00             = _mm256_sub_ps(iz0,jz0);
1747             dx01             = _mm256_sub_ps(ix0,jx1);
1748             dy01             = _mm256_sub_ps(iy0,jy1);
1749             dz01             = _mm256_sub_ps(iz0,jz1);
1750             dx02             = _mm256_sub_ps(ix0,jx2);
1751             dy02             = _mm256_sub_ps(iy0,jy2);
1752             dz02             = _mm256_sub_ps(iz0,jz2);
1753             dx10             = _mm256_sub_ps(ix1,jx0);
1754             dy10             = _mm256_sub_ps(iy1,jy0);
1755             dz10             = _mm256_sub_ps(iz1,jz0);
1756             dx11             = _mm256_sub_ps(ix1,jx1);
1757             dy11             = _mm256_sub_ps(iy1,jy1);
1758             dz11             = _mm256_sub_ps(iz1,jz1);
1759             dx12             = _mm256_sub_ps(ix1,jx2);
1760             dy12             = _mm256_sub_ps(iy1,jy2);
1761             dz12             = _mm256_sub_ps(iz1,jz2);
1762             dx20             = _mm256_sub_ps(ix2,jx0);
1763             dy20             = _mm256_sub_ps(iy2,jy0);
1764             dz20             = _mm256_sub_ps(iz2,jz0);
1765             dx21             = _mm256_sub_ps(ix2,jx1);
1766             dy21             = _mm256_sub_ps(iy2,jy1);
1767             dz21             = _mm256_sub_ps(iz2,jz1);
1768             dx22             = _mm256_sub_ps(ix2,jx2);
1769             dy22             = _mm256_sub_ps(iy2,jy2);
1770             dz22             = _mm256_sub_ps(iz2,jz2);
1771
1772             /* Calculate squared distance and things based on it */
1773             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1774             rsq01            = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
1775             rsq02            = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
1776             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
1777             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1778             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1779             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
1780             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1781             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1782
1783             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
1784             rinv01           = gmx_mm256_invsqrt_ps(rsq01);
1785             rinv02           = gmx_mm256_invsqrt_ps(rsq02);
1786             rinv10           = gmx_mm256_invsqrt_ps(rsq10);
1787             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1788             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1789             rinv20           = gmx_mm256_invsqrt_ps(rsq20);
1790             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1791             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1792
1793             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
1794             rinvsq01         = _mm256_mul_ps(rinv01,rinv01);
1795             rinvsq02         = _mm256_mul_ps(rinv02,rinv02);
1796             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
1797             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1798             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1799             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
1800             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1801             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1802
1803             fjx0             = _mm256_setzero_ps();
1804             fjy0             = _mm256_setzero_ps();
1805             fjz0             = _mm256_setzero_ps();
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
1813             /**************************
1814              * CALCULATE INTERACTIONS *
1815              **************************/
1816
1817             if (gmx_mm256_any_lt(rsq00,rcutoff2))
1818             {
1819
1820             /* REACTION-FIELD ELECTROSTATICS */
1821             felec            = _mm256_mul_ps(qq00,_mm256_sub_ps(_mm256_mul_ps(rinv00,rinvsq00),krf2));
1822
1823             cutoff_mask      = _mm256_cmp_ps(rsq00,rcutoff2,_CMP_LT_OQ);
1824
1825             fscal            = felec;
1826
1827             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1828
1829             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1830
1831             /* Calculate temporary vectorial force */
1832             tx               = _mm256_mul_ps(fscal,dx00);
1833             ty               = _mm256_mul_ps(fscal,dy00);
1834             tz               = _mm256_mul_ps(fscal,dz00);
1835
1836             /* Update vectorial force */
1837             fix0             = _mm256_add_ps(fix0,tx);
1838             fiy0             = _mm256_add_ps(fiy0,ty);
1839             fiz0             = _mm256_add_ps(fiz0,tz);
1840
1841             fjx0             = _mm256_add_ps(fjx0,tx);
1842             fjy0             = _mm256_add_ps(fjy0,ty);
1843             fjz0             = _mm256_add_ps(fjz0,tz);
1844
1845             }
1846
1847             /**************************
1848              * CALCULATE INTERACTIONS *
1849              **************************/
1850
1851             if (gmx_mm256_any_lt(rsq01,rcutoff2))
1852             {
1853
1854             /* REACTION-FIELD ELECTROSTATICS */
1855             felec            = _mm256_mul_ps(qq01,_mm256_sub_ps(_mm256_mul_ps(rinv01,rinvsq01),krf2));
1856
1857             cutoff_mask      = _mm256_cmp_ps(rsq01,rcutoff2,_CMP_LT_OQ);
1858
1859             fscal            = felec;
1860
1861             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1862
1863             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1864
1865             /* Calculate temporary vectorial force */
1866             tx               = _mm256_mul_ps(fscal,dx01);
1867             ty               = _mm256_mul_ps(fscal,dy01);
1868             tz               = _mm256_mul_ps(fscal,dz01);
1869
1870             /* Update vectorial force */
1871             fix0             = _mm256_add_ps(fix0,tx);
1872             fiy0             = _mm256_add_ps(fiy0,ty);
1873             fiz0             = _mm256_add_ps(fiz0,tz);
1874
1875             fjx1             = _mm256_add_ps(fjx1,tx);
1876             fjy1             = _mm256_add_ps(fjy1,ty);
1877             fjz1             = _mm256_add_ps(fjz1,tz);
1878
1879             }
1880
1881             /**************************
1882              * CALCULATE INTERACTIONS *
1883              **************************/
1884
1885             if (gmx_mm256_any_lt(rsq02,rcutoff2))
1886             {
1887
1888             /* REACTION-FIELD ELECTROSTATICS */
1889             felec            = _mm256_mul_ps(qq02,_mm256_sub_ps(_mm256_mul_ps(rinv02,rinvsq02),krf2));
1890
1891             cutoff_mask      = _mm256_cmp_ps(rsq02,rcutoff2,_CMP_LT_OQ);
1892
1893             fscal            = felec;
1894
1895             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1896
1897             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1898
1899             /* Calculate temporary vectorial force */
1900             tx               = _mm256_mul_ps(fscal,dx02);
1901             ty               = _mm256_mul_ps(fscal,dy02);
1902             tz               = _mm256_mul_ps(fscal,dz02);
1903
1904             /* Update vectorial force */
1905             fix0             = _mm256_add_ps(fix0,tx);
1906             fiy0             = _mm256_add_ps(fiy0,ty);
1907             fiz0             = _mm256_add_ps(fiz0,tz);
1908
1909             fjx2             = _mm256_add_ps(fjx2,tx);
1910             fjy2             = _mm256_add_ps(fjy2,ty);
1911             fjz2             = _mm256_add_ps(fjz2,tz);
1912
1913             }
1914
1915             /**************************
1916              * CALCULATE INTERACTIONS *
1917              **************************/
1918
1919             if (gmx_mm256_any_lt(rsq10,rcutoff2))
1920             {
1921
1922             /* REACTION-FIELD ELECTROSTATICS */
1923             felec            = _mm256_mul_ps(qq10,_mm256_sub_ps(_mm256_mul_ps(rinv10,rinvsq10),krf2));
1924
1925             cutoff_mask      = _mm256_cmp_ps(rsq10,rcutoff2,_CMP_LT_OQ);
1926
1927             fscal            = felec;
1928
1929             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1930
1931             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1932
1933             /* Calculate temporary vectorial force */
1934             tx               = _mm256_mul_ps(fscal,dx10);
1935             ty               = _mm256_mul_ps(fscal,dy10);
1936             tz               = _mm256_mul_ps(fscal,dz10);
1937
1938             /* Update vectorial force */
1939             fix1             = _mm256_add_ps(fix1,tx);
1940             fiy1             = _mm256_add_ps(fiy1,ty);
1941             fiz1             = _mm256_add_ps(fiz1,tz);
1942
1943             fjx0             = _mm256_add_ps(fjx0,tx);
1944             fjy0             = _mm256_add_ps(fjy0,ty);
1945             fjz0             = _mm256_add_ps(fjz0,tz);
1946
1947             }
1948
1949             /**************************
1950              * CALCULATE INTERACTIONS *
1951              **************************/
1952
1953             if (gmx_mm256_any_lt(rsq11,rcutoff2))
1954             {
1955
1956             /* REACTION-FIELD ELECTROSTATICS */
1957             felec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_mul_ps(rinv11,rinvsq11),krf2));
1958
1959             cutoff_mask      = _mm256_cmp_ps(rsq11,rcutoff2,_CMP_LT_OQ);
1960
1961             fscal            = felec;
1962
1963             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1964
1965             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1966
1967             /* Calculate temporary vectorial force */
1968             tx               = _mm256_mul_ps(fscal,dx11);
1969             ty               = _mm256_mul_ps(fscal,dy11);
1970             tz               = _mm256_mul_ps(fscal,dz11);
1971
1972             /* Update vectorial force */
1973             fix1             = _mm256_add_ps(fix1,tx);
1974             fiy1             = _mm256_add_ps(fiy1,ty);
1975             fiz1             = _mm256_add_ps(fiz1,tz);
1976
1977             fjx1             = _mm256_add_ps(fjx1,tx);
1978             fjy1             = _mm256_add_ps(fjy1,ty);
1979             fjz1             = _mm256_add_ps(fjz1,tz);
1980
1981             }
1982
1983             /**************************
1984              * CALCULATE INTERACTIONS *
1985              **************************/
1986
1987             if (gmx_mm256_any_lt(rsq12,rcutoff2))
1988             {
1989
1990             /* REACTION-FIELD ELECTROSTATICS */
1991             felec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_mul_ps(rinv12,rinvsq12),krf2));
1992
1993             cutoff_mask      = _mm256_cmp_ps(rsq12,rcutoff2,_CMP_LT_OQ);
1994
1995             fscal            = felec;
1996
1997             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1998
1999             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2000
2001             /* Calculate temporary vectorial force */
2002             tx               = _mm256_mul_ps(fscal,dx12);
2003             ty               = _mm256_mul_ps(fscal,dy12);
2004             tz               = _mm256_mul_ps(fscal,dz12);
2005
2006             /* Update vectorial force */
2007             fix1             = _mm256_add_ps(fix1,tx);
2008             fiy1             = _mm256_add_ps(fiy1,ty);
2009             fiz1             = _mm256_add_ps(fiz1,tz);
2010
2011             fjx2             = _mm256_add_ps(fjx2,tx);
2012             fjy2             = _mm256_add_ps(fjy2,ty);
2013             fjz2             = _mm256_add_ps(fjz2,tz);
2014
2015             }
2016
2017             /**************************
2018              * CALCULATE INTERACTIONS *
2019              **************************/
2020
2021             if (gmx_mm256_any_lt(rsq20,rcutoff2))
2022             {
2023
2024             /* REACTION-FIELD ELECTROSTATICS */
2025             felec            = _mm256_mul_ps(qq20,_mm256_sub_ps(_mm256_mul_ps(rinv20,rinvsq20),krf2));
2026
2027             cutoff_mask      = _mm256_cmp_ps(rsq20,rcutoff2,_CMP_LT_OQ);
2028
2029             fscal            = felec;
2030
2031             fscal            = _mm256_and_ps(fscal,cutoff_mask);
2032
2033             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2034
2035             /* Calculate temporary vectorial force */
2036             tx               = _mm256_mul_ps(fscal,dx20);
2037             ty               = _mm256_mul_ps(fscal,dy20);
2038             tz               = _mm256_mul_ps(fscal,dz20);
2039
2040             /* Update vectorial force */
2041             fix2             = _mm256_add_ps(fix2,tx);
2042             fiy2             = _mm256_add_ps(fiy2,ty);
2043             fiz2             = _mm256_add_ps(fiz2,tz);
2044
2045             fjx0             = _mm256_add_ps(fjx0,tx);
2046             fjy0             = _mm256_add_ps(fjy0,ty);
2047             fjz0             = _mm256_add_ps(fjz0,tz);
2048
2049             }
2050
2051             /**************************
2052              * CALCULATE INTERACTIONS *
2053              **************************/
2054
2055             if (gmx_mm256_any_lt(rsq21,rcutoff2))
2056             {
2057
2058             /* REACTION-FIELD ELECTROSTATICS */
2059             felec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_mul_ps(rinv21,rinvsq21),krf2));
2060
2061             cutoff_mask      = _mm256_cmp_ps(rsq21,rcutoff2,_CMP_LT_OQ);
2062
2063             fscal            = felec;
2064
2065             fscal            = _mm256_and_ps(fscal,cutoff_mask);
2066
2067             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2068
2069             /* Calculate temporary vectorial force */
2070             tx               = _mm256_mul_ps(fscal,dx21);
2071             ty               = _mm256_mul_ps(fscal,dy21);
2072             tz               = _mm256_mul_ps(fscal,dz21);
2073
2074             /* Update vectorial force */
2075             fix2             = _mm256_add_ps(fix2,tx);
2076             fiy2             = _mm256_add_ps(fiy2,ty);
2077             fiz2             = _mm256_add_ps(fiz2,tz);
2078
2079             fjx1             = _mm256_add_ps(fjx1,tx);
2080             fjy1             = _mm256_add_ps(fjy1,ty);
2081             fjz1             = _mm256_add_ps(fjz1,tz);
2082
2083             }
2084
2085             /**************************
2086              * CALCULATE INTERACTIONS *
2087              **************************/
2088
2089             if (gmx_mm256_any_lt(rsq22,rcutoff2))
2090             {
2091
2092             /* REACTION-FIELD ELECTROSTATICS */
2093             felec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_mul_ps(rinv22,rinvsq22),krf2));
2094
2095             cutoff_mask      = _mm256_cmp_ps(rsq22,rcutoff2,_CMP_LT_OQ);
2096
2097             fscal            = felec;
2098
2099             fscal            = _mm256_and_ps(fscal,cutoff_mask);
2100
2101             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2102
2103             /* Calculate temporary vectorial force */
2104             tx               = _mm256_mul_ps(fscal,dx22);
2105             ty               = _mm256_mul_ps(fscal,dy22);
2106             tz               = _mm256_mul_ps(fscal,dz22);
2107
2108             /* Update vectorial force */
2109             fix2             = _mm256_add_ps(fix2,tx);
2110             fiy2             = _mm256_add_ps(fiy2,ty);
2111             fiz2             = _mm256_add_ps(fiz2,tz);
2112
2113             fjx2             = _mm256_add_ps(fjx2,tx);
2114             fjy2             = _mm256_add_ps(fjy2,ty);
2115             fjz2             = _mm256_add_ps(fjz2,tz);
2116
2117             }
2118
2119             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
2120             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
2121             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
2122             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
2123             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
2124             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
2125             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
2126             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
2127
2128             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
2129                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
2130
2131             /* Inner loop uses 270 flops */
2132         }
2133
2134         /* End of innermost loop */
2135
2136         gmx_mm256_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
2137                                                  f+i_coord_offset,fshift+i_shift_offset);
2138
2139         /* Increment number of inner iterations */
2140         inneriter                  += j_index_end - j_index_start;
2141
2142         /* Outer loop uses 18 flops */
2143     }
2144
2145     /* Increment number of outer iterations */
2146     outeriter        += nri;
2147
2148     /* Update outer/inner flops */
2149
2150     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3W3_F,outeriter*18 + inneriter*270);
2151 }