src/gmxlib/nonbonded/nb_kernel_avx_256_single/nb_kernel_ElecCoul_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_ElecCoul_VdwNone_GeomW3W3_VF_avx_256_single
  38  * Electrostatics interaction: Coulomb
  39  * VdW interaction:            None
  40  * Geometry:                   Water3-Water3
  41  * Calculate force/pot:        PotentialAndForce
  42  */
  43 void
  44 nb_kernel_ElecCoul_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
 113     /* Setup water-specific parameters */
 114     inr              = nlist->iinr[0];
 115     iq0              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+0]));
 116     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
 117     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
 118
 119     jq0              = _mm256_set1_ps(charge[inr+0]);
 120     jq1              = _mm256_set1_ps(charge[inr+1]);
 121     jq2              = _mm256_set1_ps(charge[inr+2]);
 122     qq00             = _mm256_mul_ps(iq0,jq0);
 123     qq01             = _mm256_mul_ps(iq0,jq1);
 124     qq02             = _mm256_mul_ps(iq0,jq2);
 125     qq10             = _mm256_mul_ps(iq1,jq0);
 126     qq11             = _mm256_mul_ps(iq1,jq1);
 127     qq12             = _mm256_mul_ps(iq1,jq2);
 128     qq20             = _mm256_mul_ps(iq2,jq0);
 129     qq21             = _mm256_mul_ps(iq2,jq1);
 130     qq22             = _mm256_mul_ps(iq2,jq2);
 131
 132     /* Avoid stupid compiler warnings */
 133     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
 134     j_coord_offsetA = 0;
 135     j_coord_offsetB = 0;
 136     j_coord_offsetC = 0;
 137     j_coord_offsetD = 0;
 138     j_coord_offsetE = 0;
 139     j_coord_offsetF = 0;
 140     j_coord_offsetG = 0;
 141     j_coord_offsetH = 0;
 142
 143     outeriter        = 0;
 144     inneriter        = 0;
 145
 146     for(iidx=0;iidx<4*DIM;iidx++)
 147     {
 148         scratch[iidx] = 0.0;
 149     }
 150
 151     /* Start outer loop over neighborlists */
 152     for(iidx=0; iidx<nri; iidx++)
 153     {
 154         /* Load shift vector for this list */
 155         i_shift_offset   = DIM*shiftidx[iidx];
 156
 157         /* Load limits for loop over neighbors */
 158         j_index_start    = jindex[iidx];
 159         j_index_end      = jindex[iidx+1];
 160
 161         /* Get outer coordinate index */
 162         inr              = iinr[iidx];
 163         i_coord_offset   = DIM*inr;
 164
 165         /* Load i particle coords and add shift vector */
 166         gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
 167                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 168
 169         fix0             = _mm256_setzero_ps();
 170         fiy0             = _mm256_setzero_ps();
 171         fiz0             = _mm256_setzero_ps();
 172         fix1             = _mm256_setzero_ps();
 173         fiy1             = _mm256_setzero_ps();
 174         fiz1             = _mm256_setzero_ps();
 175         fix2             = _mm256_setzero_ps();
 176         fiy2             = _mm256_setzero_ps();
 177         fiz2             = _mm256_setzero_ps();
 178
 179         /* Reset potential sums */
 180         velecsum         = _mm256_setzero_ps();
 181
 182         /* Start inner kernel loop */
 183         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
 184         {
 185
 186             /* Get j neighbor index, and coordinate index */
 187             jnrA             = jjnr[jidx];
 188             jnrB             = jjnr[jidx+1];
 189             jnrC             = jjnr[jidx+2];
 190             jnrD             = jjnr[jidx+3];
 191             jnrE             = jjnr[jidx+4];
 192             jnrF             = jjnr[jidx+5];
 193             jnrG             = jjnr[jidx+6];
 194             jnrH             = jjnr[jidx+7];
 195             j_coord_offsetA  = DIM*jnrA;
 196             j_coord_offsetB  = DIM*jnrB;
 197             j_coord_offsetC  = DIM*jnrC;
 198             j_coord_offsetD  = DIM*jnrD;
 199             j_coord_offsetE  = DIM*jnrE;
 200             j_coord_offsetF  = DIM*jnrF;
 201             j_coord_offsetG  = DIM*jnrG;
 202             j_coord_offsetH  = DIM*jnrH;
 203
 204             /* load j atom coordinates */
 205             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 206                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 207                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 208                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 209                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 210
 211             /* Calculate displacement vector */
 212             dx00             = _mm256_sub_ps(ix0,jx0);
 213             dy00             = _mm256_sub_ps(iy0,jy0);
 214             dz00             = _mm256_sub_ps(iz0,jz0);
 215             dx01             = _mm256_sub_ps(ix0,jx1);
 216             dy01             = _mm256_sub_ps(iy0,jy1);
 217             dz01             = _mm256_sub_ps(iz0,jz1);
 218             dx02             = _mm256_sub_ps(ix0,jx2);
 219             dy02             = _mm256_sub_ps(iy0,jy2);
 220             dz02             = _mm256_sub_ps(iz0,jz2);
 221             dx10             = _mm256_sub_ps(ix1,jx0);
 222             dy10             = _mm256_sub_ps(iy1,jy0);
 223             dz10             = _mm256_sub_ps(iz1,jz0);
 224             dx11             = _mm256_sub_ps(ix1,jx1);
 225             dy11             = _mm256_sub_ps(iy1,jy1);
 226             dz11             = _mm256_sub_ps(iz1,jz1);
 227             dx12             = _mm256_sub_ps(ix1,jx2);
 228             dy12             = _mm256_sub_ps(iy1,jy2);
 229             dz12             = _mm256_sub_ps(iz1,jz2);
 230             dx20             = _mm256_sub_ps(ix2,jx0);
 231             dy20             = _mm256_sub_ps(iy2,jy0);
 232             dz20             = _mm256_sub_ps(iz2,jz0);
 233             dx21             = _mm256_sub_ps(ix2,jx1);
 234             dy21             = _mm256_sub_ps(iy2,jy1);
 235             dz21             = _mm256_sub_ps(iz2,jz1);
 236             dx22             = _mm256_sub_ps(ix2,jx2);
 237             dy22             = _mm256_sub_ps(iy2,jy2);
 238             dz22             = _mm256_sub_ps(iz2,jz2);
 239
 240             /* Calculate squared distance and things based on it */
 241             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 242             rsq01            = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
 243             rsq02            = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
 244             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
 245             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 246             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 247             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
 248             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 249             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 250
 251             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
 252             rinv01           = gmx_mm256_invsqrt_ps(rsq01);
 253             rinv02           = gmx_mm256_invsqrt_ps(rsq02);
 254             rinv10           = gmx_mm256_invsqrt_ps(rsq10);
 255             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 256             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 257             rinv20           = gmx_mm256_invsqrt_ps(rsq20);
 258             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 259             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 260
 261             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
 262             rinvsq01         = _mm256_mul_ps(rinv01,rinv01);
 263             rinvsq02         = _mm256_mul_ps(rinv02,rinv02);
 264             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
 265             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 266             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 267             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
 268             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 269             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 270
 271             fjx0             = _mm256_setzero_ps();
 272             fjy0             = _mm256_setzero_ps();
 273             fjz0             = _mm256_setzero_ps();
 274             fjx1             = _mm256_setzero_ps();
 275             fjy1             = _mm256_setzero_ps();
 276             fjz1             = _mm256_setzero_ps();
 277             fjx2             = _mm256_setzero_ps();
 278             fjy2             = _mm256_setzero_ps();
 279             fjz2             = _mm256_setzero_ps();
 280
 281             /**************************
 282              * CALCULATE INTERACTIONS *
 283              **************************/
 284
 285             /* COULOMB ELECTROSTATICS */
 286             velec            = _mm256_mul_ps(qq00,rinv00);
 287             felec            = _mm256_mul_ps(velec,rinvsq00);
 288
 289             /* Update potential sum for this i atom from the interaction with this j atom. */
 290             velecsum         = _mm256_add_ps(velecsum,velec);
 291
 292             fscal            = felec;
 293
 294             /* Calculate temporary vectorial force */
 295             tx               = _mm256_mul_ps(fscal,dx00);
 296             ty               = _mm256_mul_ps(fscal,dy00);
 297             tz               = _mm256_mul_ps(fscal,dz00);
 298
 299             /* Update vectorial force */
 300             fix0             = _mm256_add_ps(fix0,tx);
 301             fiy0             = _mm256_add_ps(fiy0,ty);
 302             fiz0             = _mm256_add_ps(fiz0,tz);
 303
 304             fjx0             = _mm256_add_ps(fjx0,tx);
 305             fjy0             = _mm256_add_ps(fjy0,ty);
 306             fjz0             = _mm256_add_ps(fjz0,tz);
 307
 308             /**************************
 309              * CALCULATE INTERACTIONS *
 310              **************************/
 311
 312             /* COULOMB ELECTROSTATICS */
 313             velec            = _mm256_mul_ps(qq01,rinv01);
 314             felec            = _mm256_mul_ps(velec,rinvsq01);
 315
 316             /* Update potential sum for this i atom from the interaction with this j atom. */
 317             velecsum         = _mm256_add_ps(velecsum,velec);
 318
 319             fscal            = felec;
 320
 321             /* Calculate temporary vectorial force */
 322             tx               = _mm256_mul_ps(fscal,dx01);
 323             ty               = _mm256_mul_ps(fscal,dy01);
 324             tz               = _mm256_mul_ps(fscal,dz01);
 325
 326             /* Update vectorial force */
 327             fix0             = _mm256_add_ps(fix0,tx);
 328             fiy0             = _mm256_add_ps(fiy0,ty);
 329             fiz0             = _mm256_add_ps(fiz0,tz);
 330
 331             fjx1             = _mm256_add_ps(fjx1,tx);
 332             fjy1             = _mm256_add_ps(fjy1,ty);
 333             fjz1             = _mm256_add_ps(fjz1,tz);
 334
 335             /**************************
 336              * CALCULATE INTERACTIONS *
 337              **************************/
 338
 339             /* COULOMB ELECTROSTATICS */
 340             velec            = _mm256_mul_ps(qq02,rinv02);
 341             felec            = _mm256_mul_ps(velec,rinvsq02);
 342
 343             /* Update potential sum for this i atom from the interaction with this j atom. */
 344             velecsum         = _mm256_add_ps(velecsum,velec);
 345
 346             fscal            = felec;
 347
 348             /* Calculate temporary vectorial force */
 349             tx               = _mm256_mul_ps(fscal,dx02);
 350             ty               = _mm256_mul_ps(fscal,dy02);
 351             tz               = _mm256_mul_ps(fscal,dz02);
 352
 353             /* Update vectorial force */
 354             fix0             = _mm256_add_ps(fix0,tx);
 355             fiy0             = _mm256_add_ps(fiy0,ty);
 356             fiz0             = _mm256_add_ps(fiz0,tz);
 357
 358             fjx2             = _mm256_add_ps(fjx2,tx);
 359             fjy2             = _mm256_add_ps(fjy2,ty);
 360             fjz2             = _mm256_add_ps(fjz2,tz);
 361
 362             /**************************
 363              * CALCULATE INTERACTIONS *
 364              **************************/
 365
 366             /* COULOMB ELECTROSTATICS */
 367             velec            = _mm256_mul_ps(qq10,rinv10);
 368             felec            = _mm256_mul_ps(velec,rinvsq10);
 369
 370             /* Update potential sum for this i atom from the interaction with this j atom. */
 371             velecsum         = _mm256_add_ps(velecsum,velec);
 372
 373             fscal            = felec;
 374
 375             /* Calculate temporary vectorial force */
 376             tx               = _mm256_mul_ps(fscal,dx10);
 377             ty               = _mm256_mul_ps(fscal,dy10);
 378             tz               = _mm256_mul_ps(fscal,dz10);
 379
 380             /* Update vectorial force */
 381             fix1             = _mm256_add_ps(fix1,tx);
 382             fiy1             = _mm256_add_ps(fiy1,ty);
 383             fiz1             = _mm256_add_ps(fiz1,tz);
 384
 385             fjx0             = _mm256_add_ps(fjx0,tx);
 386             fjy0             = _mm256_add_ps(fjy0,ty);
 387             fjz0             = _mm256_add_ps(fjz0,tz);
 388
 389             /**************************
 390              * CALCULATE INTERACTIONS *
 391              **************************/
 392
 393             /* COULOMB ELECTROSTATICS */
 394             velec            = _mm256_mul_ps(qq11,rinv11);
 395             felec            = _mm256_mul_ps(velec,rinvsq11);
 396
 397             /* Update potential sum for this i atom from the interaction with this j atom. */
 398             velecsum         = _mm256_add_ps(velecsum,velec);
 399
 400             fscal            = felec;
 401
 402             /* Calculate temporary vectorial force */
 403             tx               = _mm256_mul_ps(fscal,dx11);
 404             ty               = _mm256_mul_ps(fscal,dy11);
 405             tz               = _mm256_mul_ps(fscal,dz11);
 406
 407             /* Update vectorial force */
 408             fix1             = _mm256_add_ps(fix1,tx);
 409             fiy1             = _mm256_add_ps(fiy1,ty);
 410             fiz1             = _mm256_add_ps(fiz1,tz);
 411
 412             fjx1             = _mm256_add_ps(fjx1,tx);
 413             fjy1             = _mm256_add_ps(fjy1,ty);
 414             fjz1             = _mm256_add_ps(fjz1,tz);
 415
 416             /**************************
 417              * CALCULATE INTERACTIONS *
 418              **************************/
 419
 420             /* COULOMB ELECTROSTATICS */
 421             velec            = _mm256_mul_ps(qq12,rinv12);
 422             felec            = _mm256_mul_ps(velec,rinvsq12);
 423
 424             /* Update potential sum for this i atom from the interaction with this j atom. */
 425             velecsum         = _mm256_add_ps(velecsum,velec);
 426
 427             fscal            = felec;
 428
 429             /* Calculate temporary vectorial force */
 430             tx               = _mm256_mul_ps(fscal,dx12);
 431             ty               = _mm256_mul_ps(fscal,dy12);
 432             tz               = _mm256_mul_ps(fscal,dz12);
 433
 434             /* Update vectorial force */
 435             fix1             = _mm256_add_ps(fix1,tx);
 436             fiy1             = _mm256_add_ps(fiy1,ty);
 437             fiz1             = _mm256_add_ps(fiz1,tz);
 438
 439             fjx2             = _mm256_add_ps(fjx2,tx);
 440             fjy2             = _mm256_add_ps(fjy2,ty);
 441             fjz2             = _mm256_add_ps(fjz2,tz);
 442
 443             /**************************
 444              * CALCULATE INTERACTIONS *
 445              **************************/
 446
 447             /* COULOMB ELECTROSTATICS */
 448             velec            = _mm256_mul_ps(qq20,rinv20);
 449             felec            = _mm256_mul_ps(velec,rinvsq20);
 450
 451             /* Update potential sum for this i atom from the interaction with this j atom. */
 452             velecsum         = _mm256_add_ps(velecsum,velec);
 453
 454             fscal            = felec;
 455
 456             /* Calculate temporary vectorial force */
 457             tx               = _mm256_mul_ps(fscal,dx20);
 458             ty               = _mm256_mul_ps(fscal,dy20);
 459             tz               = _mm256_mul_ps(fscal,dz20);
 460
 461             /* Update vectorial force */
 462             fix2             = _mm256_add_ps(fix2,tx);
 463             fiy2             = _mm256_add_ps(fiy2,ty);
 464             fiz2             = _mm256_add_ps(fiz2,tz);
 465
 466             fjx0             = _mm256_add_ps(fjx0,tx);
 467             fjy0             = _mm256_add_ps(fjy0,ty);
 468             fjz0             = _mm256_add_ps(fjz0,tz);
 469
 470             /**************************
 471              * CALCULATE INTERACTIONS *
 472              **************************/
 473
 474             /* COULOMB ELECTROSTATICS */
 475             velec            = _mm256_mul_ps(qq21,rinv21);
 476             felec            = _mm256_mul_ps(velec,rinvsq21);
 477
 478             /* Update potential sum for this i atom from the interaction with this j atom. */
 479             velecsum         = _mm256_add_ps(velecsum,velec);
 480
 481             fscal            = felec;
 482
 483             /* Calculate temporary vectorial force */
 484             tx               = _mm256_mul_ps(fscal,dx21);
 485             ty               = _mm256_mul_ps(fscal,dy21);
 486             tz               = _mm256_mul_ps(fscal,dz21);
 487
 488             /* Update vectorial force */
 489             fix2             = _mm256_add_ps(fix2,tx);
 490             fiy2             = _mm256_add_ps(fiy2,ty);
 491             fiz2             = _mm256_add_ps(fiz2,tz);
 492
 493             fjx1             = _mm256_add_ps(fjx1,tx);
 494             fjy1             = _mm256_add_ps(fjy1,ty);
 495             fjz1             = _mm256_add_ps(fjz1,tz);
 496
 497             /**************************
 498              * CALCULATE INTERACTIONS *
 499              **************************/
 500
 501             /* COULOMB ELECTROSTATICS */
 502             velec            = _mm256_mul_ps(qq22,rinv22);
 503             felec            = _mm256_mul_ps(velec,rinvsq22);
 504
 505             /* Update potential sum for this i atom from the interaction with this j atom. */
 506             velecsum         = _mm256_add_ps(velecsum,velec);
 507
 508             fscal            = felec;
 509
 510             /* Calculate temporary vectorial force */
 511             tx               = _mm256_mul_ps(fscal,dx22);
 512             ty               = _mm256_mul_ps(fscal,dy22);
 513             tz               = _mm256_mul_ps(fscal,dz22);
 514
 515             /* Update vectorial force */
 516             fix2             = _mm256_add_ps(fix2,tx);
 517             fiy2             = _mm256_add_ps(fiy2,ty);
 518             fiz2             = _mm256_add_ps(fiz2,tz);
 519
 520             fjx2             = _mm256_add_ps(fjx2,tx);
 521             fjy2             = _mm256_add_ps(fjy2,ty);
 522             fjz2             = _mm256_add_ps(fjz2,tz);
 523
 524             fjptrA             = f+j_coord_offsetA;
 525             fjptrB             = f+j_coord_offsetB;
 526             fjptrC             = f+j_coord_offsetC;
 527             fjptrD             = f+j_coord_offsetD;
 528             fjptrE             = f+j_coord_offsetE;
 529             fjptrF             = f+j_coord_offsetF;
 530             fjptrG             = f+j_coord_offsetG;
 531             fjptrH             = f+j_coord_offsetH;
 532
 533             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
 534                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 535
 536             /* Inner loop uses 243 flops */
 537         }
 538
 539         if(jidx<j_index_end)
 540         {
 541
 542             /* Get j neighbor index, and coordinate index */
 543             jnrlistA         = jjnr[jidx];
 544             jnrlistB         = jjnr[jidx+1];
 545             jnrlistC         = jjnr[jidx+2];
 546             jnrlistD         = jjnr[jidx+3];
 547             jnrlistE         = jjnr[jidx+4];
 548             jnrlistF         = jjnr[jidx+5];
 549             jnrlistG         = jjnr[jidx+6];
 550             jnrlistH         = jjnr[jidx+7];
 551             /* Sign of each element will be negative for non-real atoms.
 552              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 553              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 554              */
 555             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
 556                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
 557
 558             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 559             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 560             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 561             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 562             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
 563             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
 564             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
 565             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
 566             j_coord_offsetA  = DIM*jnrA;
 567             j_coord_offsetB  = DIM*jnrB;
 568             j_coord_offsetC  = DIM*jnrC;
 569             j_coord_offsetD  = DIM*jnrD;
 570             j_coord_offsetE  = DIM*jnrE;
 571             j_coord_offsetF  = DIM*jnrF;
 572             j_coord_offsetG  = DIM*jnrG;
 573             j_coord_offsetH  = DIM*jnrH;
 574
 575             /* load j atom coordinates */
 576             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 577                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 578                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 579                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 580                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 581
 582             /* Calculate displacement vector */
 583             dx00             = _mm256_sub_ps(ix0,jx0);
 584             dy00             = _mm256_sub_ps(iy0,jy0);
 585             dz00             = _mm256_sub_ps(iz0,jz0);
 586             dx01             = _mm256_sub_ps(ix0,jx1);
 587             dy01             = _mm256_sub_ps(iy0,jy1);
 588             dz01             = _mm256_sub_ps(iz0,jz1);
 589             dx02             = _mm256_sub_ps(ix0,jx2);
 590             dy02             = _mm256_sub_ps(iy0,jy2);
 591             dz02             = _mm256_sub_ps(iz0,jz2);
 592             dx10             = _mm256_sub_ps(ix1,jx0);
 593             dy10             = _mm256_sub_ps(iy1,jy0);
 594             dz10             = _mm256_sub_ps(iz1,jz0);
 595             dx11             = _mm256_sub_ps(ix1,jx1);
 596             dy11             = _mm256_sub_ps(iy1,jy1);
 597             dz11             = _mm256_sub_ps(iz1,jz1);
 598             dx12             = _mm256_sub_ps(ix1,jx2);
 599             dy12             = _mm256_sub_ps(iy1,jy2);
 600             dz12             = _mm256_sub_ps(iz1,jz2);
 601             dx20             = _mm256_sub_ps(ix2,jx0);
 602             dy20             = _mm256_sub_ps(iy2,jy0);
 603             dz20             = _mm256_sub_ps(iz2,jz0);
 604             dx21             = _mm256_sub_ps(ix2,jx1);
 605             dy21             = _mm256_sub_ps(iy2,jy1);
 606             dz21             = _mm256_sub_ps(iz2,jz1);
 607             dx22             = _mm256_sub_ps(ix2,jx2);
 608             dy22             = _mm256_sub_ps(iy2,jy2);
 609             dz22             = _mm256_sub_ps(iz2,jz2);
 610
 611             /* Calculate squared distance and things based on it */
 612             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 613             rsq01            = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
 614             rsq02            = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
 615             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
 616             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 617             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 618             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
 619             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 620             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 621
 622             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
 623             rinv01           = gmx_mm256_invsqrt_ps(rsq01);
 624             rinv02           = gmx_mm256_invsqrt_ps(rsq02);
 625             rinv10           = gmx_mm256_invsqrt_ps(rsq10);
 626             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 627             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 628             rinv20           = gmx_mm256_invsqrt_ps(rsq20);
 629             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 630             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 631
 632             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
 633             rinvsq01         = _mm256_mul_ps(rinv01,rinv01);
 634             rinvsq02         = _mm256_mul_ps(rinv02,rinv02);
 635             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
 636             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 637             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 638             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
 639             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 640             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 641
 642             fjx0             = _mm256_setzero_ps();
 643             fjy0             = _mm256_setzero_ps();
 644             fjz0             = _mm256_setzero_ps();
 645             fjx1             = _mm256_setzero_ps();
 646             fjy1             = _mm256_setzero_ps();
 647             fjz1             = _mm256_setzero_ps();
 648             fjx2             = _mm256_setzero_ps();
 649             fjy2             = _mm256_setzero_ps();
 650             fjz2             = _mm256_setzero_ps();
 651
 652             /**************************
 653              * CALCULATE INTERACTIONS *
 654              **************************/
 655
 656             /* COULOMB ELECTROSTATICS */
 657             velec            = _mm256_mul_ps(qq00,rinv00);
 658             felec            = _mm256_mul_ps(velec,rinvsq00);
 659
 660             /* Update potential sum for this i atom from the interaction with this j atom. */
 661             velec            = _mm256_andnot_ps(dummy_mask,velec);
 662             velecsum         = _mm256_add_ps(velecsum,velec);
 663
 664             fscal            = felec;
 665
 666             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 667
 668             /* Calculate temporary vectorial force */
 669             tx               = _mm256_mul_ps(fscal,dx00);
 670             ty               = _mm256_mul_ps(fscal,dy00);
 671             tz               = _mm256_mul_ps(fscal,dz00);
 672
 673             /* Update vectorial force */
 674             fix0             = _mm256_add_ps(fix0,tx);
 675             fiy0             = _mm256_add_ps(fiy0,ty);
 676             fiz0             = _mm256_add_ps(fiz0,tz);
 677
 678             fjx0             = _mm256_add_ps(fjx0,tx);
 679             fjy0             = _mm256_add_ps(fjy0,ty);
 680             fjz0             = _mm256_add_ps(fjz0,tz);
 681
 682             /**************************
 683              * CALCULATE INTERACTIONS *
 684              **************************/
 685
 686             /* COULOMB ELECTROSTATICS */
 687             velec            = _mm256_mul_ps(qq01,rinv01);
 688             felec            = _mm256_mul_ps(velec,rinvsq01);
 689
 690             /* Update potential sum for this i atom from the interaction with this j atom. */
 691             velec            = _mm256_andnot_ps(dummy_mask,velec);
 692             velecsum         = _mm256_add_ps(velecsum,velec);
 693
 694             fscal            = felec;
 695
 696             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 697
 698             /* Calculate temporary vectorial force */
 699             tx               = _mm256_mul_ps(fscal,dx01);
 700             ty               = _mm256_mul_ps(fscal,dy01);
 701             tz               = _mm256_mul_ps(fscal,dz01);
 702
 703             /* Update vectorial force */
 704             fix0             = _mm256_add_ps(fix0,tx);
 705             fiy0             = _mm256_add_ps(fiy0,ty);
 706             fiz0             = _mm256_add_ps(fiz0,tz);
 707
 708             fjx1             = _mm256_add_ps(fjx1,tx);
 709             fjy1             = _mm256_add_ps(fjy1,ty);
 710             fjz1             = _mm256_add_ps(fjz1,tz);
 711
 712             /**************************
 713              * CALCULATE INTERACTIONS *
 714              **************************/
 715
 716             /* COULOMB ELECTROSTATICS */
 717             velec            = _mm256_mul_ps(qq02,rinv02);
 718             felec            = _mm256_mul_ps(velec,rinvsq02);
 719
 720             /* Update potential sum for this i atom from the interaction with this j atom. */
 721             velec            = _mm256_andnot_ps(dummy_mask,velec);
 722             velecsum         = _mm256_add_ps(velecsum,velec);
 723
 724             fscal            = felec;
 725
 726             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 727
 728             /* Calculate temporary vectorial force */
 729             tx               = _mm256_mul_ps(fscal,dx02);
 730             ty               = _mm256_mul_ps(fscal,dy02);
 731             tz               = _mm256_mul_ps(fscal,dz02);
 732
 733             /* Update vectorial force */
 734             fix0             = _mm256_add_ps(fix0,tx);
 735             fiy0             = _mm256_add_ps(fiy0,ty);
 736             fiz0             = _mm256_add_ps(fiz0,tz);
 737
 738             fjx2             = _mm256_add_ps(fjx2,tx);
 739             fjy2             = _mm256_add_ps(fjy2,ty);
 740             fjz2             = _mm256_add_ps(fjz2,tz);
 741
 742             /**************************
 743              * CALCULATE INTERACTIONS *
 744              **************************/
 745
 746             /* COULOMB ELECTROSTATICS */
 747             velec            = _mm256_mul_ps(qq10,rinv10);
 748             felec            = _mm256_mul_ps(velec,rinvsq10);
 749
 750             /* Update potential sum for this i atom from the interaction with this j atom. */
 751             velec            = _mm256_andnot_ps(dummy_mask,velec);
 752             velecsum         = _mm256_add_ps(velecsum,velec);
 753
 754             fscal            = felec;
 755
 756             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 757
 758             /* Calculate temporary vectorial force */
 759             tx               = _mm256_mul_ps(fscal,dx10);
 760             ty               = _mm256_mul_ps(fscal,dy10);
 761             tz               = _mm256_mul_ps(fscal,dz10);
 762
 763             /* Update vectorial force */
 764             fix1             = _mm256_add_ps(fix1,tx);
 765             fiy1             = _mm256_add_ps(fiy1,ty);
 766             fiz1             = _mm256_add_ps(fiz1,tz);
 767
 768             fjx0             = _mm256_add_ps(fjx0,tx);
 769             fjy0             = _mm256_add_ps(fjy0,ty);
 770             fjz0             = _mm256_add_ps(fjz0,tz);
 771
 772             /**************************
 773              * CALCULATE INTERACTIONS *
 774              **************************/
 775
 776             /* COULOMB ELECTROSTATICS */
 777             velec            = _mm256_mul_ps(qq11,rinv11);
 778             felec            = _mm256_mul_ps(velec,rinvsq11);
 779
 780             /* Update potential sum for this i atom from the interaction with this j atom. */
 781             velec            = _mm256_andnot_ps(dummy_mask,velec);
 782             velecsum         = _mm256_add_ps(velecsum,velec);
 783
 784             fscal            = felec;
 785
 786             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 787
 788             /* Calculate temporary vectorial force */
 789             tx               = _mm256_mul_ps(fscal,dx11);
 790             ty               = _mm256_mul_ps(fscal,dy11);
 791             tz               = _mm256_mul_ps(fscal,dz11);
 792
 793             /* Update vectorial force */
 794             fix1             = _mm256_add_ps(fix1,tx);
 795             fiy1             = _mm256_add_ps(fiy1,ty);
 796             fiz1             = _mm256_add_ps(fiz1,tz);
 797
 798             fjx1             = _mm256_add_ps(fjx1,tx);
 799             fjy1             = _mm256_add_ps(fjy1,ty);
 800             fjz1             = _mm256_add_ps(fjz1,tz);
 801
 802             /**************************
 803              * CALCULATE INTERACTIONS *
 804              **************************/
 805
 806             /* COULOMB ELECTROSTATICS */
 807             velec            = _mm256_mul_ps(qq12,rinv12);
 808             felec            = _mm256_mul_ps(velec,rinvsq12);
 809
 810             /* Update potential sum for this i atom from the interaction with this j atom. */
 811             velec            = _mm256_andnot_ps(dummy_mask,velec);
 812             velecsum         = _mm256_add_ps(velecsum,velec);
 813
 814             fscal            = felec;
 815
 816             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 817
 818             /* Calculate temporary vectorial force */
 819             tx               = _mm256_mul_ps(fscal,dx12);
 820             ty               = _mm256_mul_ps(fscal,dy12);
 821             tz               = _mm256_mul_ps(fscal,dz12);
 822
 823             /* Update vectorial force */
 824             fix1             = _mm256_add_ps(fix1,tx);
 825             fiy1             = _mm256_add_ps(fiy1,ty);
 826             fiz1             = _mm256_add_ps(fiz1,tz);
 827
 828             fjx2             = _mm256_add_ps(fjx2,tx);
 829             fjy2             = _mm256_add_ps(fjy2,ty);
 830             fjz2             = _mm256_add_ps(fjz2,tz);
 831
 832             /**************************
 833              * CALCULATE INTERACTIONS *
 834              **************************/
 835
 836             /* COULOMB ELECTROSTATICS */
 837             velec            = _mm256_mul_ps(qq20,rinv20);
 838             felec            = _mm256_mul_ps(velec,rinvsq20);
 839
 840             /* Update potential sum for this i atom from the interaction with this j atom. */
 841             velec            = _mm256_andnot_ps(dummy_mask,velec);
 842             velecsum         = _mm256_add_ps(velecsum,velec);
 843
 844             fscal            = felec;
 845
 846             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 847
 848             /* Calculate temporary vectorial force */
 849             tx               = _mm256_mul_ps(fscal,dx20);
 850             ty               = _mm256_mul_ps(fscal,dy20);
 851             tz               = _mm256_mul_ps(fscal,dz20);
 852
 853             /* Update vectorial force */
 854             fix2             = _mm256_add_ps(fix2,tx);
 855             fiy2             = _mm256_add_ps(fiy2,ty);
 856             fiz2             = _mm256_add_ps(fiz2,tz);
 857
 858             fjx0             = _mm256_add_ps(fjx0,tx);
 859             fjy0             = _mm256_add_ps(fjy0,ty);
 860             fjz0             = _mm256_add_ps(fjz0,tz);
 861
 862             /**************************
 863              * CALCULATE INTERACTIONS *
 864              **************************/
 865
 866             /* COULOMB ELECTROSTATICS */
 867             velec            = _mm256_mul_ps(qq21,rinv21);
 868             felec            = _mm256_mul_ps(velec,rinvsq21);
 869
 870             /* Update potential sum for this i atom from the interaction with this j atom. */
 871             velec            = _mm256_andnot_ps(dummy_mask,velec);
 872             velecsum         = _mm256_add_ps(velecsum,velec);
 873
 874             fscal            = felec;
 875
 876             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 877
 878             /* Calculate temporary vectorial force */
 879             tx               = _mm256_mul_ps(fscal,dx21);
 880             ty               = _mm256_mul_ps(fscal,dy21);
 881             tz               = _mm256_mul_ps(fscal,dz21);
 882
 883             /* Update vectorial force */
 884             fix2             = _mm256_add_ps(fix2,tx);
 885             fiy2             = _mm256_add_ps(fiy2,ty);
 886             fiz2             = _mm256_add_ps(fiz2,tz);
 887
 888             fjx1             = _mm256_add_ps(fjx1,tx);
 889             fjy1             = _mm256_add_ps(fjy1,ty);
 890             fjz1             = _mm256_add_ps(fjz1,tz);
 891
 892             /**************************
 893              * CALCULATE INTERACTIONS *
 894              **************************/
 895
 896             /* COULOMB ELECTROSTATICS */
 897             velec            = _mm256_mul_ps(qq22,rinv22);
 898             felec            = _mm256_mul_ps(velec,rinvsq22);
 899
 900             /* Update potential sum for this i atom from the interaction with this j atom. */
 901             velec            = _mm256_andnot_ps(dummy_mask,velec);
 902             velecsum         = _mm256_add_ps(velecsum,velec);
 903
 904             fscal            = felec;
 905
 906             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 907
 908             /* Calculate temporary vectorial force */
 909             tx               = _mm256_mul_ps(fscal,dx22);
 910             ty               = _mm256_mul_ps(fscal,dy22);
 911             tz               = _mm256_mul_ps(fscal,dz22);
 912
 913             /* Update vectorial force */
 914             fix2             = _mm256_add_ps(fix2,tx);
 915             fiy2             = _mm256_add_ps(fiy2,ty);
 916             fiz2             = _mm256_add_ps(fiz2,tz);
 917
 918             fjx2             = _mm256_add_ps(fjx2,tx);
 919             fjy2             = _mm256_add_ps(fjy2,ty);
 920             fjz2             = _mm256_add_ps(fjz2,tz);
 921
 922             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
 923             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
 924             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
 925             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
 926             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
 927             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
 928             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
 929             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
 930
 931             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
 932                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 933
 934             /* Inner loop uses 243 flops */
 935         }
 936
 937         /* End of innermost loop */
 938
 939         gmx_mm256_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 940                                                  f+i_coord_offset,fshift+i_shift_offset);
 941
 942         ggid                        = gid[iidx];
 943         /* Update potential energies */
 944         gmx_mm256_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
 945
 946         /* Increment number of inner iterations */
 947         inneriter                  += j_index_end - j_index_start;
 948
 949         /* Outer loop uses 19 flops */
 950     }
 951
 952     /* Increment number of outer iterations */
 953     outeriter        += nri;
 954
 955     /* Update outer/inner flops */
 956
 957     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3W3_VF,outeriter*19 + inneriter*243);
 958 }
 959 /*
 960  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwNone_GeomW3W3_F_avx_256_single
 961  * Electrostatics interaction: Coulomb
 962  * VdW interaction:            None
 963  * Geometry:                   Water3-Water3
 964  * Calculate force/pot:        Force
 965  */
 966 void
 967 nb_kernel_ElecCoul_VdwNone_GeomW3W3_F_avx_256_single
 968                     (t_nblist * gmx_restrict                nlist,
 969                      rvec * gmx_restrict                    xx,
 970                      rvec * gmx_restrict                    ff,
 971                      t_forcerec * gmx_restrict              fr,
 972                      t_mdatoms * gmx_restrict               mdatoms,
 973                      nb_kernel_data_t * gmx_restrict        kernel_data,
 974                      t_nrnb * gmx_restrict                  nrnb)
 975 {
 976     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 977      * just 0 for non-waters.
 978      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
 979      * jnr indices corresponding to data put in the four positions in the SIMD register.
 980      */
 981     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 982     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 983     int              jnrA,jnrB,jnrC,jnrD;
 984     int              jnrE,jnrF,jnrG,jnrH;
 985     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
 986     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
 987     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
 988     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
 989     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 990     real             rcutoff_scalar;
 991     real             *shiftvec,*fshift,*x,*f;
 992     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
 993     real             scratch[4*DIM];
 994     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 995     real *           vdwioffsetptr0;
 996     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 997     real *           vdwioffsetptr1;
 998     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 999     real *           vdwioffsetptr2;
1000     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1001     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
1002     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1003     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
1004     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1005     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
1006     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1007     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1008     __m256           dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1009     __m256           dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1010     __m256           dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1011     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1012     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1013     __m256           dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1014     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1015     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1016     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
1017     real             *charge;
1018     __m256           dummy_mask,cutoff_mask;
1019     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
1020     __m256           one     = _mm256_set1_ps(1.0);
1021     __m256           two     = _mm256_set1_ps(2.0);
1022     x                = xx[0];
1023     f                = ff[0];
1024
1025     nri              = nlist->nri;
1026     iinr             = nlist->iinr;
1027     jindex           = nlist->jindex;
1028     jjnr             = nlist->jjnr;
1029     shiftidx         = nlist->shift;
1030     gid              = nlist->gid;
1031     shiftvec         = fr->shift_vec[0];
1032     fshift           = fr->fshift[0];
1033     facel            = _mm256_set1_ps(fr->epsfac);
1034     charge           = mdatoms->chargeA;
1035
1036     /* Setup water-specific parameters */
1037     inr              = nlist->iinr[0];
1038     iq0              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+0]));
1039     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
1040     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
1041
1042     jq0              = _mm256_set1_ps(charge[inr+0]);
1043     jq1              = _mm256_set1_ps(charge[inr+1]);
1044     jq2              = _mm256_set1_ps(charge[inr+2]);
1045     qq00             = _mm256_mul_ps(iq0,jq0);
1046     qq01             = _mm256_mul_ps(iq0,jq1);
1047     qq02             = _mm256_mul_ps(iq0,jq2);
1048     qq10             = _mm256_mul_ps(iq1,jq0);
1049     qq11             = _mm256_mul_ps(iq1,jq1);
1050     qq12             = _mm256_mul_ps(iq1,jq2);
1051     qq20             = _mm256_mul_ps(iq2,jq0);
1052     qq21             = _mm256_mul_ps(iq2,jq1);
1053     qq22             = _mm256_mul_ps(iq2,jq2);
1054
1055     /* Avoid stupid compiler warnings */
1056     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
1057     j_coord_offsetA = 0;
1058     j_coord_offsetB = 0;
1059     j_coord_offsetC = 0;
1060     j_coord_offsetD = 0;
1061     j_coord_offsetE = 0;
1062     j_coord_offsetF = 0;
1063     j_coord_offsetG = 0;
1064     j_coord_offsetH = 0;
1065
1066     outeriter        = 0;
1067     inneriter        = 0;
1068
1069     for(iidx=0;iidx<4*DIM;iidx++)
1070     {
1071         scratch[iidx] = 0.0;
1072     }
1073
1074     /* Start outer loop over neighborlists */
1075     for(iidx=0; iidx<nri; iidx++)
1076     {
1077         /* Load shift vector for this list */
1078         i_shift_offset   = DIM*shiftidx[iidx];
1079
1080         /* Load limits for loop over neighbors */
1081         j_index_start    = jindex[iidx];
1082         j_index_end      = jindex[iidx+1];
1083
1084         /* Get outer coordinate index */
1085         inr              = iinr[iidx];
1086         i_coord_offset   = DIM*inr;
1087
1088         /* Load i particle coords and add shift vector */
1089         gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1090                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1091
1092         fix0             = _mm256_setzero_ps();
1093         fiy0             = _mm256_setzero_ps();
1094         fiz0             = _mm256_setzero_ps();
1095         fix1             = _mm256_setzero_ps();
1096         fiy1             = _mm256_setzero_ps();
1097         fiz1             = _mm256_setzero_ps();
1098         fix2             = _mm256_setzero_ps();
1099         fiy2             = _mm256_setzero_ps();
1100         fiz2             = _mm256_setzero_ps();
1101
1102         /* Start inner kernel loop */
1103         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
1104         {
1105
1106             /* Get j neighbor index, and coordinate index */
1107             jnrA             = jjnr[jidx];
1108             jnrB             = jjnr[jidx+1];
1109             jnrC             = jjnr[jidx+2];
1110             jnrD             = jjnr[jidx+3];
1111             jnrE             = jjnr[jidx+4];
1112             jnrF             = jjnr[jidx+5];
1113             jnrG             = jjnr[jidx+6];
1114             jnrH             = jjnr[jidx+7];
1115             j_coord_offsetA  = DIM*jnrA;
1116             j_coord_offsetB  = DIM*jnrB;
1117             j_coord_offsetC  = DIM*jnrC;
1118             j_coord_offsetD  = DIM*jnrD;
1119             j_coord_offsetE  = DIM*jnrE;
1120             j_coord_offsetF  = DIM*jnrF;
1121             j_coord_offsetG  = DIM*jnrG;
1122             j_coord_offsetH  = DIM*jnrH;
1123
1124             /* load j atom coordinates */
1125             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1126                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1127                                                  x+j_coord_offsetE,x+j_coord_offsetF,
1128                                                  x+j_coord_offsetG,x+j_coord_offsetH,
1129                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1130
1131             /* Calculate displacement vector */
1132             dx00             = _mm256_sub_ps(ix0,jx0);
1133             dy00             = _mm256_sub_ps(iy0,jy0);
1134             dz00             = _mm256_sub_ps(iz0,jz0);
1135             dx01             = _mm256_sub_ps(ix0,jx1);
1136             dy01             = _mm256_sub_ps(iy0,jy1);
1137             dz01             = _mm256_sub_ps(iz0,jz1);
1138             dx02             = _mm256_sub_ps(ix0,jx2);
1139             dy02             = _mm256_sub_ps(iy0,jy2);
1140             dz02             = _mm256_sub_ps(iz0,jz2);
1141             dx10             = _mm256_sub_ps(ix1,jx0);
1142             dy10             = _mm256_sub_ps(iy1,jy0);
1143             dz10             = _mm256_sub_ps(iz1,jz0);
1144             dx11             = _mm256_sub_ps(ix1,jx1);
1145             dy11             = _mm256_sub_ps(iy1,jy1);
1146             dz11             = _mm256_sub_ps(iz1,jz1);
1147             dx12             = _mm256_sub_ps(ix1,jx2);
1148             dy12             = _mm256_sub_ps(iy1,jy2);
1149             dz12             = _mm256_sub_ps(iz1,jz2);
1150             dx20             = _mm256_sub_ps(ix2,jx0);
1151             dy20             = _mm256_sub_ps(iy2,jy0);
1152             dz20             = _mm256_sub_ps(iz2,jz0);
1153             dx21             = _mm256_sub_ps(ix2,jx1);
1154             dy21             = _mm256_sub_ps(iy2,jy1);
1155             dz21             = _mm256_sub_ps(iz2,jz1);
1156             dx22             = _mm256_sub_ps(ix2,jx2);
1157             dy22             = _mm256_sub_ps(iy2,jy2);
1158             dz22             = _mm256_sub_ps(iz2,jz2);
1159
1160             /* Calculate squared distance and things based on it */
1161             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1162             rsq01            = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
1163             rsq02            = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
1164             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
1165             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1166             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1167             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
1168             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1169             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1170
1171             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
1172             rinv01           = gmx_mm256_invsqrt_ps(rsq01);
1173             rinv02           = gmx_mm256_invsqrt_ps(rsq02);
1174             rinv10           = gmx_mm256_invsqrt_ps(rsq10);
1175             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1176             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1177             rinv20           = gmx_mm256_invsqrt_ps(rsq20);
1178             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1179             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1180
1181             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
1182             rinvsq01         = _mm256_mul_ps(rinv01,rinv01);
1183             rinvsq02         = _mm256_mul_ps(rinv02,rinv02);
1184             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
1185             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1186             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1187             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
1188             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1189             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1190
1191             fjx0             = _mm256_setzero_ps();
1192             fjy0             = _mm256_setzero_ps();
1193             fjz0             = _mm256_setzero_ps();
1194             fjx1             = _mm256_setzero_ps();
1195             fjy1             = _mm256_setzero_ps();
1196             fjz1             = _mm256_setzero_ps();
1197             fjx2             = _mm256_setzero_ps();
1198             fjy2             = _mm256_setzero_ps();
1199             fjz2             = _mm256_setzero_ps();
1200
1201             /**************************
1202              * CALCULATE INTERACTIONS *
1203              **************************/
1204
1205             /* COULOMB ELECTROSTATICS */
1206             velec            = _mm256_mul_ps(qq00,rinv00);
1207             felec            = _mm256_mul_ps(velec,rinvsq00);
1208
1209             fscal            = felec;
1210
1211             /* Calculate temporary vectorial force */
1212             tx               = _mm256_mul_ps(fscal,dx00);
1213             ty               = _mm256_mul_ps(fscal,dy00);
1214             tz               = _mm256_mul_ps(fscal,dz00);
1215
1216             /* Update vectorial force */
1217             fix0             = _mm256_add_ps(fix0,tx);
1218             fiy0             = _mm256_add_ps(fiy0,ty);
1219             fiz0             = _mm256_add_ps(fiz0,tz);
1220
1221             fjx0             = _mm256_add_ps(fjx0,tx);
1222             fjy0             = _mm256_add_ps(fjy0,ty);
1223             fjz0             = _mm256_add_ps(fjz0,tz);
1224
1225             /**************************
1226              * CALCULATE INTERACTIONS *
1227              **************************/
1228
1229             /* COULOMB ELECTROSTATICS */
1230             velec            = _mm256_mul_ps(qq01,rinv01);
1231             felec            = _mm256_mul_ps(velec,rinvsq01);
1232
1233             fscal            = felec;
1234
1235             /* Calculate temporary vectorial force */
1236             tx               = _mm256_mul_ps(fscal,dx01);
1237             ty               = _mm256_mul_ps(fscal,dy01);
1238             tz               = _mm256_mul_ps(fscal,dz01);
1239
1240             /* Update vectorial force */
1241             fix0             = _mm256_add_ps(fix0,tx);
1242             fiy0             = _mm256_add_ps(fiy0,ty);
1243             fiz0             = _mm256_add_ps(fiz0,tz);
1244
1245             fjx1             = _mm256_add_ps(fjx1,tx);
1246             fjy1             = _mm256_add_ps(fjy1,ty);
1247             fjz1             = _mm256_add_ps(fjz1,tz);
1248
1249             /**************************
1250              * CALCULATE INTERACTIONS *
1251              **************************/
1252
1253             /* COULOMB ELECTROSTATICS */
1254             velec            = _mm256_mul_ps(qq02,rinv02);
1255             felec            = _mm256_mul_ps(velec,rinvsq02);
1256
1257             fscal            = felec;
1258
1259             /* Calculate temporary vectorial force */
1260             tx               = _mm256_mul_ps(fscal,dx02);
1261             ty               = _mm256_mul_ps(fscal,dy02);
1262             tz               = _mm256_mul_ps(fscal,dz02);
1263
1264             /* Update vectorial force */
1265             fix0             = _mm256_add_ps(fix0,tx);
1266             fiy0             = _mm256_add_ps(fiy0,ty);
1267             fiz0             = _mm256_add_ps(fiz0,tz);
1268
1269             fjx2             = _mm256_add_ps(fjx2,tx);
1270             fjy2             = _mm256_add_ps(fjy2,ty);
1271             fjz2             = _mm256_add_ps(fjz2,tz);
1272
1273             /**************************
1274              * CALCULATE INTERACTIONS *
1275              **************************/
1276
1277             /* COULOMB ELECTROSTATICS */
1278             velec            = _mm256_mul_ps(qq10,rinv10);
1279             felec            = _mm256_mul_ps(velec,rinvsq10);
1280
1281             fscal            = felec;
1282
1283             /* Calculate temporary vectorial force */
1284             tx               = _mm256_mul_ps(fscal,dx10);
1285             ty               = _mm256_mul_ps(fscal,dy10);
1286             tz               = _mm256_mul_ps(fscal,dz10);
1287
1288             /* Update vectorial force */
1289             fix1             = _mm256_add_ps(fix1,tx);
1290             fiy1             = _mm256_add_ps(fiy1,ty);
1291             fiz1             = _mm256_add_ps(fiz1,tz);
1292
1293             fjx0             = _mm256_add_ps(fjx0,tx);
1294             fjy0             = _mm256_add_ps(fjy0,ty);
1295             fjz0             = _mm256_add_ps(fjz0,tz);
1296
1297             /**************************
1298              * CALCULATE INTERACTIONS *
1299              **************************/
1300
1301             /* COULOMB ELECTROSTATICS */
1302             velec            = _mm256_mul_ps(qq11,rinv11);
1303             felec            = _mm256_mul_ps(velec,rinvsq11);
1304
1305             fscal            = felec;
1306
1307             /* Calculate temporary vectorial force */
1308             tx               = _mm256_mul_ps(fscal,dx11);
1309             ty               = _mm256_mul_ps(fscal,dy11);
1310             tz               = _mm256_mul_ps(fscal,dz11);
1311
1312             /* Update vectorial force */
1313             fix1             = _mm256_add_ps(fix1,tx);
1314             fiy1             = _mm256_add_ps(fiy1,ty);
1315             fiz1             = _mm256_add_ps(fiz1,tz);
1316
1317             fjx1             = _mm256_add_ps(fjx1,tx);
1318             fjy1             = _mm256_add_ps(fjy1,ty);
1319             fjz1             = _mm256_add_ps(fjz1,tz);
1320
1321             /**************************
1322              * CALCULATE INTERACTIONS *
1323              **************************/
1324
1325             /* COULOMB ELECTROSTATICS */
1326             velec            = _mm256_mul_ps(qq12,rinv12);
1327             felec            = _mm256_mul_ps(velec,rinvsq12);
1328
1329             fscal            = felec;
1330
1331             /* Calculate temporary vectorial force */
1332             tx               = _mm256_mul_ps(fscal,dx12);
1333             ty               = _mm256_mul_ps(fscal,dy12);
1334             tz               = _mm256_mul_ps(fscal,dz12);
1335
1336             /* Update vectorial force */
1337             fix1             = _mm256_add_ps(fix1,tx);
1338             fiy1             = _mm256_add_ps(fiy1,ty);
1339             fiz1             = _mm256_add_ps(fiz1,tz);
1340
1341             fjx2             = _mm256_add_ps(fjx2,tx);
1342             fjy2             = _mm256_add_ps(fjy2,ty);
1343             fjz2             = _mm256_add_ps(fjz2,tz);
1344
1345             /**************************
1346              * CALCULATE INTERACTIONS *
1347              **************************/
1348
1349             /* COULOMB ELECTROSTATICS */
1350             velec            = _mm256_mul_ps(qq20,rinv20);
1351             felec            = _mm256_mul_ps(velec,rinvsq20);
1352
1353             fscal            = felec;
1354
1355             /* Calculate temporary vectorial force */
1356             tx               = _mm256_mul_ps(fscal,dx20);
1357             ty               = _mm256_mul_ps(fscal,dy20);
1358             tz               = _mm256_mul_ps(fscal,dz20);
1359
1360             /* Update vectorial force */
1361             fix2             = _mm256_add_ps(fix2,tx);
1362             fiy2             = _mm256_add_ps(fiy2,ty);
1363             fiz2             = _mm256_add_ps(fiz2,tz);
1364
1365             fjx0             = _mm256_add_ps(fjx0,tx);
1366             fjy0             = _mm256_add_ps(fjy0,ty);
1367             fjz0             = _mm256_add_ps(fjz0,tz);
1368
1369             /**************************
1370              * CALCULATE INTERACTIONS *
1371              **************************/
1372
1373             /* COULOMB ELECTROSTATICS */
1374             velec            = _mm256_mul_ps(qq21,rinv21);
1375             felec            = _mm256_mul_ps(velec,rinvsq21);
1376
1377             fscal            = felec;
1378
1379             /* Calculate temporary vectorial force */
1380             tx               = _mm256_mul_ps(fscal,dx21);
1381             ty               = _mm256_mul_ps(fscal,dy21);
1382             tz               = _mm256_mul_ps(fscal,dz21);
1383
1384             /* Update vectorial force */
1385             fix2             = _mm256_add_ps(fix2,tx);
1386             fiy2             = _mm256_add_ps(fiy2,ty);
1387             fiz2             = _mm256_add_ps(fiz2,tz);
1388
1389             fjx1             = _mm256_add_ps(fjx1,tx);
1390             fjy1             = _mm256_add_ps(fjy1,ty);
1391             fjz1             = _mm256_add_ps(fjz1,tz);
1392
1393             /**************************
1394              * CALCULATE INTERACTIONS *
1395              **************************/
1396
1397             /* COULOMB ELECTROSTATICS */
1398             velec            = _mm256_mul_ps(qq22,rinv22);
1399             felec            = _mm256_mul_ps(velec,rinvsq22);
1400
1401             fscal            = felec;
1402
1403             /* Calculate temporary vectorial force */
1404             tx               = _mm256_mul_ps(fscal,dx22);
1405             ty               = _mm256_mul_ps(fscal,dy22);
1406             tz               = _mm256_mul_ps(fscal,dz22);
1407
1408             /* Update vectorial force */
1409             fix2             = _mm256_add_ps(fix2,tx);
1410             fiy2             = _mm256_add_ps(fiy2,ty);
1411             fiz2             = _mm256_add_ps(fiz2,tz);
1412
1413             fjx2             = _mm256_add_ps(fjx2,tx);
1414             fjy2             = _mm256_add_ps(fjy2,ty);
1415             fjz2             = _mm256_add_ps(fjz2,tz);
1416
1417             fjptrA             = f+j_coord_offsetA;
1418             fjptrB             = f+j_coord_offsetB;
1419             fjptrC             = f+j_coord_offsetC;
1420             fjptrD             = f+j_coord_offsetD;
1421             fjptrE             = f+j_coord_offsetE;
1422             fjptrF             = f+j_coord_offsetF;
1423             fjptrG             = f+j_coord_offsetG;
1424             fjptrH             = f+j_coord_offsetH;
1425
1426             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1427                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1428
1429             /* Inner loop uses 234 flops */
1430         }
1431
1432         if(jidx<j_index_end)
1433         {
1434
1435             /* Get j neighbor index, and coordinate index */
1436             jnrlistA         = jjnr[jidx];
1437             jnrlistB         = jjnr[jidx+1];
1438             jnrlistC         = jjnr[jidx+2];
1439             jnrlistD         = jjnr[jidx+3];
1440             jnrlistE         = jjnr[jidx+4];
1441             jnrlistF         = jjnr[jidx+5];
1442             jnrlistG         = jjnr[jidx+6];
1443             jnrlistH         = jjnr[jidx+7];
1444             /* Sign of each element will be negative for non-real atoms.
1445              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1446              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1447              */
1448             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
1449                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
1450
1451             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1452             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1453             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1454             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1455             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
1456             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
1457             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
1458             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
1459             j_coord_offsetA  = DIM*jnrA;
1460             j_coord_offsetB  = DIM*jnrB;
1461             j_coord_offsetC  = DIM*jnrC;
1462             j_coord_offsetD  = DIM*jnrD;
1463             j_coord_offsetE  = DIM*jnrE;
1464             j_coord_offsetF  = DIM*jnrF;
1465             j_coord_offsetG  = DIM*jnrG;
1466             j_coord_offsetH  = DIM*jnrH;
1467
1468             /* load j atom coordinates */
1469             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1470                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1471                                                  x+j_coord_offsetE,x+j_coord_offsetF,
1472                                                  x+j_coord_offsetG,x+j_coord_offsetH,
1473                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1474
1475             /* Calculate displacement vector */
1476             dx00             = _mm256_sub_ps(ix0,jx0);
1477             dy00             = _mm256_sub_ps(iy0,jy0);
1478             dz00             = _mm256_sub_ps(iz0,jz0);
1479             dx01             = _mm256_sub_ps(ix0,jx1);
1480             dy01             = _mm256_sub_ps(iy0,jy1);
1481             dz01             = _mm256_sub_ps(iz0,jz1);
1482             dx02             = _mm256_sub_ps(ix0,jx2);
1483             dy02             = _mm256_sub_ps(iy0,jy2);
1484             dz02             = _mm256_sub_ps(iz0,jz2);
1485             dx10             = _mm256_sub_ps(ix1,jx0);
1486             dy10             = _mm256_sub_ps(iy1,jy0);
1487             dz10             = _mm256_sub_ps(iz1,jz0);
1488             dx11             = _mm256_sub_ps(ix1,jx1);
1489             dy11             = _mm256_sub_ps(iy1,jy1);
1490             dz11             = _mm256_sub_ps(iz1,jz1);
1491             dx12             = _mm256_sub_ps(ix1,jx2);
1492             dy12             = _mm256_sub_ps(iy1,jy2);
1493             dz12             = _mm256_sub_ps(iz1,jz2);
1494             dx20             = _mm256_sub_ps(ix2,jx0);
1495             dy20             = _mm256_sub_ps(iy2,jy0);
1496             dz20             = _mm256_sub_ps(iz2,jz0);
1497             dx21             = _mm256_sub_ps(ix2,jx1);
1498             dy21             = _mm256_sub_ps(iy2,jy1);
1499             dz21             = _mm256_sub_ps(iz2,jz1);
1500             dx22             = _mm256_sub_ps(ix2,jx2);
1501             dy22             = _mm256_sub_ps(iy2,jy2);
1502             dz22             = _mm256_sub_ps(iz2,jz2);
1503
1504             /* Calculate squared distance and things based on it */
1505             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1506             rsq01            = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
1507             rsq02            = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
1508             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
1509             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1510             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1511             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
1512             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1513             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1514
1515             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
1516             rinv01           = gmx_mm256_invsqrt_ps(rsq01);
1517             rinv02           = gmx_mm256_invsqrt_ps(rsq02);
1518             rinv10           = gmx_mm256_invsqrt_ps(rsq10);
1519             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1520             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1521             rinv20           = gmx_mm256_invsqrt_ps(rsq20);
1522             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1523             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1524
1525             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
1526             rinvsq01         = _mm256_mul_ps(rinv01,rinv01);
1527             rinvsq02         = _mm256_mul_ps(rinv02,rinv02);
1528             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
1529             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1530             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1531             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
1532             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1533             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1534
1535             fjx0             = _mm256_setzero_ps();
1536             fjy0             = _mm256_setzero_ps();
1537             fjz0             = _mm256_setzero_ps();
1538             fjx1             = _mm256_setzero_ps();
1539             fjy1             = _mm256_setzero_ps();
1540             fjz1             = _mm256_setzero_ps();
1541             fjx2             = _mm256_setzero_ps();
1542             fjy2             = _mm256_setzero_ps();
1543             fjz2             = _mm256_setzero_ps();
1544
1545             /**************************
1546              * CALCULATE INTERACTIONS *
1547              **************************/
1548
1549             /* COULOMB ELECTROSTATICS */
1550             velec            = _mm256_mul_ps(qq00,rinv00);
1551             felec            = _mm256_mul_ps(velec,rinvsq00);
1552
1553             fscal            = felec;
1554
1555             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1556
1557             /* Calculate temporary vectorial force */
1558             tx               = _mm256_mul_ps(fscal,dx00);
1559             ty               = _mm256_mul_ps(fscal,dy00);
1560             tz               = _mm256_mul_ps(fscal,dz00);
1561
1562             /* Update vectorial force */
1563             fix0             = _mm256_add_ps(fix0,tx);
1564             fiy0             = _mm256_add_ps(fiy0,ty);
1565             fiz0             = _mm256_add_ps(fiz0,tz);
1566
1567             fjx0             = _mm256_add_ps(fjx0,tx);
1568             fjy0             = _mm256_add_ps(fjy0,ty);
1569             fjz0             = _mm256_add_ps(fjz0,tz);
1570
1571             /**************************
1572              * CALCULATE INTERACTIONS *
1573              **************************/
1574
1575             /* COULOMB ELECTROSTATICS */
1576             velec            = _mm256_mul_ps(qq01,rinv01);
1577             felec            = _mm256_mul_ps(velec,rinvsq01);
1578
1579             fscal            = felec;
1580
1581             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1582
1583             /* Calculate temporary vectorial force */
1584             tx               = _mm256_mul_ps(fscal,dx01);
1585             ty               = _mm256_mul_ps(fscal,dy01);
1586             tz               = _mm256_mul_ps(fscal,dz01);
1587
1588             /* Update vectorial force */
1589             fix0             = _mm256_add_ps(fix0,tx);
1590             fiy0             = _mm256_add_ps(fiy0,ty);
1591             fiz0             = _mm256_add_ps(fiz0,tz);
1592
1593             fjx1             = _mm256_add_ps(fjx1,tx);
1594             fjy1             = _mm256_add_ps(fjy1,ty);
1595             fjz1             = _mm256_add_ps(fjz1,tz);
1596
1597             /**************************
1598              * CALCULATE INTERACTIONS *
1599              **************************/
1600
1601             /* COULOMB ELECTROSTATICS */
1602             velec            = _mm256_mul_ps(qq02,rinv02);
1603             felec            = _mm256_mul_ps(velec,rinvsq02);
1604
1605             fscal            = felec;
1606
1607             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1608
1609             /* Calculate temporary vectorial force */
1610             tx               = _mm256_mul_ps(fscal,dx02);
1611             ty               = _mm256_mul_ps(fscal,dy02);
1612             tz               = _mm256_mul_ps(fscal,dz02);
1613
1614             /* Update vectorial force */
1615             fix0             = _mm256_add_ps(fix0,tx);
1616             fiy0             = _mm256_add_ps(fiy0,ty);
1617             fiz0             = _mm256_add_ps(fiz0,tz);
1618
1619             fjx2             = _mm256_add_ps(fjx2,tx);
1620             fjy2             = _mm256_add_ps(fjy2,ty);
1621             fjz2             = _mm256_add_ps(fjz2,tz);
1622
1623             /**************************
1624              * CALCULATE INTERACTIONS *
1625              **************************/
1626
1627             /* COULOMB ELECTROSTATICS */
1628             velec            = _mm256_mul_ps(qq10,rinv10);
1629             felec            = _mm256_mul_ps(velec,rinvsq10);
1630
1631             fscal            = felec;
1632
1633             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1634
1635             /* Calculate temporary vectorial force */
1636             tx               = _mm256_mul_ps(fscal,dx10);
1637             ty               = _mm256_mul_ps(fscal,dy10);
1638             tz               = _mm256_mul_ps(fscal,dz10);
1639
1640             /* Update vectorial force */
1641             fix1             = _mm256_add_ps(fix1,tx);
1642             fiy1             = _mm256_add_ps(fiy1,ty);
1643             fiz1             = _mm256_add_ps(fiz1,tz);
1644
1645             fjx0             = _mm256_add_ps(fjx0,tx);
1646             fjy0             = _mm256_add_ps(fjy0,ty);
1647             fjz0             = _mm256_add_ps(fjz0,tz);
1648
1649             /**************************
1650              * CALCULATE INTERACTIONS *
1651              **************************/
1652
1653             /* COULOMB ELECTROSTATICS */
1654             velec            = _mm256_mul_ps(qq11,rinv11);
1655             felec            = _mm256_mul_ps(velec,rinvsq11);
1656
1657             fscal            = felec;
1658
1659             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1660
1661             /* Calculate temporary vectorial force */
1662             tx               = _mm256_mul_ps(fscal,dx11);
1663             ty               = _mm256_mul_ps(fscal,dy11);
1664             tz               = _mm256_mul_ps(fscal,dz11);
1665
1666             /* Update vectorial force */
1667             fix1             = _mm256_add_ps(fix1,tx);
1668             fiy1             = _mm256_add_ps(fiy1,ty);
1669             fiz1             = _mm256_add_ps(fiz1,tz);
1670
1671             fjx1             = _mm256_add_ps(fjx1,tx);
1672             fjy1             = _mm256_add_ps(fjy1,ty);
1673             fjz1             = _mm256_add_ps(fjz1,tz);
1674
1675             /**************************
1676              * CALCULATE INTERACTIONS *
1677              **************************/
1678
1679             /* COULOMB ELECTROSTATICS */
1680             velec            = _mm256_mul_ps(qq12,rinv12);
1681             felec            = _mm256_mul_ps(velec,rinvsq12);
1682
1683             fscal            = felec;
1684
1685             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1686
1687             /* Calculate temporary vectorial force */
1688             tx               = _mm256_mul_ps(fscal,dx12);
1689             ty               = _mm256_mul_ps(fscal,dy12);
1690             tz               = _mm256_mul_ps(fscal,dz12);
1691
1692             /* Update vectorial force */
1693             fix1             = _mm256_add_ps(fix1,tx);
1694             fiy1             = _mm256_add_ps(fiy1,ty);
1695             fiz1             = _mm256_add_ps(fiz1,tz);
1696
1697             fjx2             = _mm256_add_ps(fjx2,tx);
1698             fjy2             = _mm256_add_ps(fjy2,ty);
1699             fjz2             = _mm256_add_ps(fjz2,tz);
1700
1701             /**************************
1702              * CALCULATE INTERACTIONS *
1703              **************************/
1704
1705             /* COULOMB ELECTROSTATICS */
1706             velec            = _mm256_mul_ps(qq20,rinv20);
1707             felec            = _mm256_mul_ps(velec,rinvsq20);
1708
1709             fscal            = felec;
1710
1711             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1712
1713             /* Calculate temporary vectorial force */
1714             tx               = _mm256_mul_ps(fscal,dx20);
1715             ty               = _mm256_mul_ps(fscal,dy20);
1716             tz               = _mm256_mul_ps(fscal,dz20);
1717
1718             /* Update vectorial force */
1719             fix2             = _mm256_add_ps(fix2,tx);
1720             fiy2             = _mm256_add_ps(fiy2,ty);
1721             fiz2             = _mm256_add_ps(fiz2,tz);
1722
1723             fjx0             = _mm256_add_ps(fjx0,tx);
1724             fjy0             = _mm256_add_ps(fjy0,ty);
1725             fjz0             = _mm256_add_ps(fjz0,tz);
1726
1727             /**************************
1728              * CALCULATE INTERACTIONS *
1729              **************************/
1730
1731             /* COULOMB ELECTROSTATICS */
1732             velec            = _mm256_mul_ps(qq21,rinv21);
1733             felec            = _mm256_mul_ps(velec,rinvsq21);
1734
1735             fscal            = felec;
1736
1737             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1738
1739             /* Calculate temporary vectorial force */
1740             tx               = _mm256_mul_ps(fscal,dx21);
1741             ty               = _mm256_mul_ps(fscal,dy21);
1742             tz               = _mm256_mul_ps(fscal,dz21);
1743
1744             /* Update vectorial force */
1745             fix2             = _mm256_add_ps(fix2,tx);
1746             fiy2             = _mm256_add_ps(fiy2,ty);
1747             fiz2             = _mm256_add_ps(fiz2,tz);
1748
1749             fjx1             = _mm256_add_ps(fjx1,tx);
1750             fjy1             = _mm256_add_ps(fjy1,ty);
1751             fjz1             = _mm256_add_ps(fjz1,tz);
1752
1753             /**************************
1754              * CALCULATE INTERACTIONS *
1755              **************************/
1756
1757             /* COULOMB ELECTROSTATICS */
1758             velec            = _mm256_mul_ps(qq22,rinv22);
1759             felec            = _mm256_mul_ps(velec,rinvsq22);
1760
1761             fscal            = felec;
1762
1763             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1764
1765             /* Calculate temporary vectorial force */
1766             tx               = _mm256_mul_ps(fscal,dx22);
1767             ty               = _mm256_mul_ps(fscal,dy22);
1768             tz               = _mm256_mul_ps(fscal,dz22);
1769
1770             /* Update vectorial force */
1771             fix2             = _mm256_add_ps(fix2,tx);
1772             fiy2             = _mm256_add_ps(fiy2,ty);
1773             fiz2             = _mm256_add_ps(fiz2,tz);
1774
1775             fjx2             = _mm256_add_ps(fjx2,tx);
1776             fjy2             = _mm256_add_ps(fjy2,ty);
1777             fjz2             = _mm256_add_ps(fjz2,tz);
1778
1779             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1780             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1781             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1782             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1783             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1784             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1785             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
1786             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
1787
1788             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1789                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1790
1791             /* Inner loop uses 234 flops */
1792         }
1793
1794         /* End of innermost loop */
1795
1796         gmx_mm256_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1797                                                  f+i_coord_offset,fshift+i_shift_offset);
1798
1799         /* Increment number of inner iterations */
1800         inneriter                  += j_index_end - j_index_start;
1801
1802         /* Outer loop uses 18 flops */
1803     }
1804
1805     /* Increment number of outer iterations */
1806     outeriter        += nri;
1807
1808     /* Update outer/inner flops */
1809
1810     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3W3_F,outeriter*18 + inneriter*234);
1811 }