src/gmxlib/nonbonded/nb_kernel_avx_256_single/nb_kernel_ElecCoul_VdwNone_GeomW4W4_avx_256_single.c

   1 /*
   2  * Note: this file was generated by the Gromacs avx_256_single kernel generator.
   3  *
   4  *                This source code is part of
   5  *
   6  *                 G   R   O   M   A   C   S
   7  *
   8  * Copyright (c) 2001-2012, The GROMACS Development Team
   9  *
  10  * Gromacs is a library for molecular simulation and trajectory analysis,
  11  * written by Erik Lindahl, David van der Spoel, Berk Hess, and others - for
  12  * a full list of developers and information, check out http://www.gromacs.org
  13  *
  14  * This program is free software; you can redistribute it and/or modify it under
  15  * the terms of the GNU Lesser General Public License as published by the Free
  16  * Software Foundation; either version 2 of the License, or (at your option) any
  17  * later version.
  18  *
  19  * To help fund GROMACS development, we humbly ask that you cite
  20  * the papers people have written on it - you can find them on the website.
  21  */
  22 #ifdef HAVE_CONFIG_H
  23 #include <config.h>
  24 #endif
  25
  26 #include <math.h>
  27
  28 #include "../nb_kernel.h"
  29 #include "types/simple.h"
  30 #include "vec.h"
  31 #include "nrnb.h"
  32
  33 #include "gmx_math_x86_avx_256_single.h"
  34 #include "kernelutil_x86_avx_256_single.h"
  35
  36 /*
  37  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwNone_GeomW4W4_VF_avx_256_single
  38  * Electrostatics interaction: Coulomb
  39  * VdW interaction:            None
  40  * Geometry:                   Water4-Water4
  41  * Calculate force/pot:        PotentialAndForce
  42  */
  43 void
  44 nb_kernel_ElecCoul_VdwNone_GeomW4W4_VF_avx_256_single
  45                     (t_nblist * gmx_restrict                nlist,
  46                      rvec * gmx_restrict                    xx,
  47                      rvec * gmx_restrict                    ff,
  48                      t_forcerec * gmx_restrict              fr,
  49                      t_mdatoms * gmx_restrict               mdatoms,
  50                      nb_kernel_data_t * gmx_restrict        kernel_data,
  51                      t_nrnb * gmx_restrict                  nrnb)
  52 {
  53     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  54      * just 0 for non-waters.
  55      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
  56      * jnr indices corresponding to data put in the four positions in the SIMD register.
  57      */
  58     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  59     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  60     int              jnrA,jnrB,jnrC,jnrD;
  61     int              jnrE,jnrF,jnrG,jnrH;
  62     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  63     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
  64     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  65     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
  66     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  67     real             rcutoff_scalar;
  68     real             *shiftvec,*fshift,*x,*f;
  69     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
  70     real             scratch[4*DIM];
  71     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  72     real *           vdwioffsetptr1;
  73     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  74     real *           vdwioffsetptr2;
  75     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  76     real *           vdwioffsetptr3;
  77     __m256           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  78     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
  79     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  80     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
  81     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  82     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D,vdwjidx3E,vdwjidx3F,vdwjidx3G,vdwjidx3H;
  83     __m256           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
  84     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
  85     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
  86     __m256           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
  87     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
  88     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
  89     __m256           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
  90     __m256           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
  91     __m256           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
  92     __m256           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
  93     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
  94     real             *charge;
  95     __m256           dummy_mask,cutoff_mask;
  96     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
  97     __m256           one     = _mm256_set1_ps(1.0);
  98     __m256           two     = _mm256_set1_ps(2.0);
  99     x                = xx[0];
 100     f                = ff[0];
 101
 102     nri              = nlist->nri;
 103     iinr             = nlist->iinr;
 104     jindex           = nlist->jindex;
 105     jjnr             = nlist->jjnr;
 106     shiftidx         = nlist->shift;
 107     gid              = nlist->gid;
 108     shiftvec         = fr->shift_vec[0];
 109     fshift           = fr->fshift[0];
 110     facel            = _mm256_set1_ps(fr->epsfac);
 111     charge           = mdatoms->chargeA;
 112
 113     /* Setup water-specific parameters */
 114     inr              = nlist->iinr[0];
 115     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
 116     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
 117     iq3              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+3]));
 118
 119     jq1              = _mm256_set1_ps(charge[inr+1]);
 120     jq2              = _mm256_set1_ps(charge[inr+2]);
 121     jq3              = _mm256_set1_ps(charge[inr+3]);
 122     qq11             = _mm256_mul_ps(iq1,jq1);
 123     qq12             = _mm256_mul_ps(iq1,jq2);
 124     qq13             = _mm256_mul_ps(iq1,jq3);
 125     qq21             = _mm256_mul_ps(iq2,jq1);
 126     qq22             = _mm256_mul_ps(iq2,jq2);
 127     qq23             = _mm256_mul_ps(iq2,jq3);
 128     qq31             = _mm256_mul_ps(iq3,jq1);
 129     qq32             = _mm256_mul_ps(iq3,jq2);
 130     qq33             = _mm256_mul_ps(iq3,jq3);
 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+DIM,
 167                                                     &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 168
 169         fix1             = _mm256_setzero_ps();
 170         fiy1             = _mm256_setzero_ps();
 171         fiz1             = _mm256_setzero_ps();
 172         fix2             = _mm256_setzero_ps();
 173         fiy2             = _mm256_setzero_ps();
 174         fiz2             = _mm256_setzero_ps();
 175         fix3             = _mm256_setzero_ps();
 176         fiy3             = _mm256_setzero_ps();
 177         fiz3             = _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+DIM,x+j_coord_offsetB+DIM,
 206                                                  x+j_coord_offsetC+DIM,x+j_coord_offsetD+DIM,
 207                                                  x+j_coord_offsetE+DIM,x+j_coord_offsetF+DIM,
 208                                                  x+j_coord_offsetG+DIM,x+j_coord_offsetH+DIM,
 209                                                  &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
 210
 211             /* Calculate displacement vector */
 212             dx11             = _mm256_sub_ps(ix1,jx1);
 213             dy11             = _mm256_sub_ps(iy1,jy1);
 214             dz11             = _mm256_sub_ps(iz1,jz1);
 215             dx12             = _mm256_sub_ps(ix1,jx2);
 216             dy12             = _mm256_sub_ps(iy1,jy2);
 217             dz12             = _mm256_sub_ps(iz1,jz2);
 218             dx13             = _mm256_sub_ps(ix1,jx3);
 219             dy13             = _mm256_sub_ps(iy1,jy3);
 220             dz13             = _mm256_sub_ps(iz1,jz3);
 221             dx21             = _mm256_sub_ps(ix2,jx1);
 222             dy21             = _mm256_sub_ps(iy2,jy1);
 223             dz21             = _mm256_sub_ps(iz2,jz1);
 224             dx22             = _mm256_sub_ps(ix2,jx2);
 225             dy22             = _mm256_sub_ps(iy2,jy2);
 226             dz22             = _mm256_sub_ps(iz2,jz2);
 227             dx23             = _mm256_sub_ps(ix2,jx3);
 228             dy23             = _mm256_sub_ps(iy2,jy3);
 229             dz23             = _mm256_sub_ps(iz2,jz3);
 230             dx31             = _mm256_sub_ps(ix3,jx1);
 231             dy31             = _mm256_sub_ps(iy3,jy1);
 232             dz31             = _mm256_sub_ps(iz3,jz1);
 233             dx32             = _mm256_sub_ps(ix3,jx2);
 234             dy32             = _mm256_sub_ps(iy3,jy2);
 235             dz32             = _mm256_sub_ps(iz3,jz2);
 236             dx33             = _mm256_sub_ps(ix3,jx3);
 237             dy33             = _mm256_sub_ps(iy3,jy3);
 238             dz33             = _mm256_sub_ps(iz3,jz3);
 239
 240             /* Calculate squared distance and things based on it */
 241             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 242             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 243             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
 244             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 245             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 246             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
 247             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
 248             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
 249             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
 250
 251             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 252             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 253             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
 254             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 255             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 256             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
 257             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
 258             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
 259             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
 260
 261             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 262             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 263             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
 264             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 265             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 266             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
 267             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
 268             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
 269             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
 270
 271             fjx1             = _mm256_setzero_ps();
 272             fjy1             = _mm256_setzero_ps();
 273             fjz1             = _mm256_setzero_ps();
 274             fjx2             = _mm256_setzero_ps();
 275             fjy2             = _mm256_setzero_ps();
 276             fjz2             = _mm256_setzero_ps();
 277             fjx3             = _mm256_setzero_ps();
 278             fjy3             = _mm256_setzero_ps();
 279             fjz3             = _mm256_setzero_ps();
 280
 281             /**************************
 282              * CALCULATE INTERACTIONS *
 283              **************************/
 284
 285             /* COULOMB ELECTROSTATICS */
 286             velec            = _mm256_mul_ps(qq11,rinv11);
 287             felec            = _mm256_mul_ps(velec,rinvsq11);
 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,dx11);
 296             ty               = _mm256_mul_ps(fscal,dy11);
 297             tz               = _mm256_mul_ps(fscal,dz11);
 298
 299             /* Update vectorial force */
 300             fix1             = _mm256_add_ps(fix1,tx);
 301             fiy1             = _mm256_add_ps(fiy1,ty);
 302             fiz1             = _mm256_add_ps(fiz1,tz);
 303
 304             fjx1             = _mm256_add_ps(fjx1,tx);
 305             fjy1             = _mm256_add_ps(fjy1,ty);
 306             fjz1             = _mm256_add_ps(fjz1,tz);
 307
 308             /**************************
 309              * CALCULATE INTERACTIONS *
 310              **************************/
 311
 312             /* COULOMB ELECTROSTATICS */
 313             velec            = _mm256_mul_ps(qq12,rinv12);
 314             felec            = _mm256_mul_ps(velec,rinvsq12);
 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,dx12);
 323             ty               = _mm256_mul_ps(fscal,dy12);
 324             tz               = _mm256_mul_ps(fscal,dz12);
 325
 326             /* Update vectorial force */
 327             fix1             = _mm256_add_ps(fix1,tx);
 328             fiy1             = _mm256_add_ps(fiy1,ty);
 329             fiz1             = _mm256_add_ps(fiz1,tz);
 330
 331             fjx2             = _mm256_add_ps(fjx2,tx);
 332             fjy2             = _mm256_add_ps(fjy2,ty);
 333             fjz2             = _mm256_add_ps(fjz2,tz);
 334
 335             /**************************
 336              * CALCULATE INTERACTIONS *
 337              **************************/
 338
 339             /* COULOMB ELECTROSTATICS */
 340             velec            = _mm256_mul_ps(qq13,rinv13);
 341             felec            = _mm256_mul_ps(velec,rinvsq13);
 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,dx13);
 350             ty               = _mm256_mul_ps(fscal,dy13);
 351             tz               = _mm256_mul_ps(fscal,dz13);
 352
 353             /* Update vectorial force */
 354             fix1             = _mm256_add_ps(fix1,tx);
 355             fiy1             = _mm256_add_ps(fiy1,ty);
 356             fiz1             = _mm256_add_ps(fiz1,tz);
 357
 358             fjx3             = _mm256_add_ps(fjx3,tx);
 359             fjy3             = _mm256_add_ps(fjy3,ty);
 360             fjz3             = _mm256_add_ps(fjz3,tz);
 361
 362             /**************************
 363              * CALCULATE INTERACTIONS *
 364              **************************/
 365
 366             /* COULOMB ELECTROSTATICS */
 367             velec            = _mm256_mul_ps(qq21,rinv21);
 368             felec            = _mm256_mul_ps(velec,rinvsq21);
 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,dx21);
 377             ty               = _mm256_mul_ps(fscal,dy21);
 378             tz               = _mm256_mul_ps(fscal,dz21);
 379
 380             /* Update vectorial force */
 381             fix2             = _mm256_add_ps(fix2,tx);
 382             fiy2             = _mm256_add_ps(fiy2,ty);
 383             fiz2             = _mm256_add_ps(fiz2,tz);
 384
 385             fjx1             = _mm256_add_ps(fjx1,tx);
 386             fjy1             = _mm256_add_ps(fjy1,ty);
 387             fjz1             = _mm256_add_ps(fjz1,tz);
 388
 389             /**************************
 390              * CALCULATE INTERACTIONS *
 391              **************************/
 392
 393             /* COULOMB ELECTROSTATICS */
 394             velec            = _mm256_mul_ps(qq22,rinv22);
 395             felec            = _mm256_mul_ps(velec,rinvsq22);
 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,dx22);
 404             ty               = _mm256_mul_ps(fscal,dy22);
 405             tz               = _mm256_mul_ps(fscal,dz22);
 406
 407             /* Update vectorial force */
 408             fix2             = _mm256_add_ps(fix2,tx);
 409             fiy2             = _mm256_add_ps(fiy2,ty);
 410             fiz2             = _mm256_add_ps(fiz2,tz);
 411
 412             fjx2             = _mm256_add_ps(fjx2,tx);
 413             fjy2             = _mm256_add_ps(fjy2,ty);
 414             fjz2             = _mm256_add_ps(fjz2,tz);
 415
 416             /**************************
 417              * CALCULATE INTERACTIONS *
 418              **************************/
 419
 420             /* COULOMB ELECTROSTATICS */
 421             velec            = _mm256_mul_ps(qq23,rinv23);
 422             felec            = _mm256_mul_ps(velec,rinvsq23);
 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,dx23);
 431             ty               = _mm256_mul_ps(fscal,dy23);
 432             tz               = _mm256_mul_ps(fscal,dz23);
 433
 434             /* Update vectorial force */
 435             fix2             = _mm256_add_ps(fix2,tx);
 436             fiy2             = _mm256_add_ps(fiy2,ty);
 437             fiz2             = _mm256_add_ps(fiz2,tz);
 438
 439             fjx3             = _mm256_add_ps(fjx3,tx);
 440             fjy3             = _mm256_add_ps(fjy3,ty);
 441             fjz3             = _mm256_add_ps(fjz3,tz);
 442
 443             /**************************
 444              * CALCULATE INTERACTIONS *
 445              **************************/
 446
 447             /* COULOMB ELECTROSTATICS */
 448             velec            = _mm256_mul_ps(qq31,rinv31);
 449             felec            = _mm256_mul_ps(velec,rinvsq31);
 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,dx31);
 458             ty               = _mm256_mul_ps(fscal,dy31);
 459             tz               = _mm256_mul_ps(fscal,dz31);
 460
 461             /* Update vectorial force */
 462             fix3             = _mm256_add_ps(fix3,tx);
 463             fiy3             = _mm256_add_ps(fiy3,ty);
 464             fiz3             = _mm256_add_ps(fiz3,tz);
 465
 466             fjx1             = _mm256_add_ps(fjx1,tx);
 467             fjy1             = _mm256_add_ps(fjy1,ty);
 468             fjz1             = _mm256_add_ps(fjz1,tz);
 469
 470             /**************************
 471              * CALCULATE INTERACTIONS *
 472              **************************/
 473
 474             /* COULOMB ELECTROSTATICS */
 475             velec            = _mm256_mul_ps(qq32,rinv32);
 476             felec            = _mm256_mul_ps(velec,rinvsq32);
 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,dx32);
 485             ty               = _mm256_mul_ps(fscal,dy32);
 486             tz               = _mm256_mul_ps(fscal,dz32);
 487
 488             /* Update vectorial force */
 489             fix3             = _mm256_add_ps(fix3,tx);
 490             fiy3             = _mm256_add_ps(fiy3,ty);
 491             fiz3             = _mm256_add_ps(fiz3,tz);
 492
 493             fjx2             = _mm256_add_ps(fjx2,tx);
 494             fjy2             = _mm256_add_ps(fjy2,ty);
 495             fjz2             = _mm256_add_ps(fjz2,tz);
 496
 497             /**************************
 498              * CALCULATE INTERACTIONS *
 499              **************************/
 500
 501             /* COULOMB ELECTROSTATICS */
 502             velec            = _mm256_mul_ps(qq33,rinv33);
 503             felec            = _mm256_mul_ps(velec,rinvsq33);
 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,dx33);
 512             ty               = _mm256_mul_ps(fscal,dy33);
 513             tz               = _mm256_mul_ps(fscal,dz33);
 514
 515             /* Update vectorial force */
 516             fix3             = _mm256_add_ps(fix3,tx);
 517             fiy3             = _mm256_add_ps(fiy3,ty);
 518             fiz3             = _mm256_add_ps(fiz3,tz);
 519
 520             fjx3             = _mm256_add_ps(fjx3,tx);
 521             fjy3             = _mm256_add_ps(fjy3,ty);
 522             fjz3             = _mm256_add_ps(fjz3,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+DIM,fjptrB+DIM,fjptrC+DIM,fjptrD+DIM,
 534                                                       fjptrE+DIM,fjptrF+DIM,fjptrG+DIM,fjptrH+DIM,
 535                                                       fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 536
 537             /* Inner loop uses 243 flops */
 538         }
 539
 540         if(jidx<j_index_end)
 541         {
 542
 543             /* Get j neighbor index, and coordinate index */
 544             jnrlistA         = jjnr[jidx];
 545             jnrlistB         = jjnr[jidx+1];
 546             jnrlistC         = jjnr[jidx+2];
 547             jnrlistD         = jjnr[jidx+3];
 548             jnrlistE         = jjnr[jidx+4];
 549             jnrlistF         = jjnr[jidx+5];
 550             jnrlistG         = jjnr[jidx+6];
 551             jnrlistH         = jjnr[jidx+7];
 552             /* Sign of each element will be negative for non-real atoms.
 553              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 554              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 555              */
 556             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
 557                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
 558
 559             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 560             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 561             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 562             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 563             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
 564             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
 565             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
 566             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
 567             j_coord_offsetA  = DIM*jnrA;
 568             j_coord_offsetB  = DIM*jnrB;
 569             j_coord_offsetC  = DIM*jnrC;
 570             j_coord_offsetD  = DIM*jnrD;
 571             j_coord_offsetE  = DIM*jnrE;
 572             j_coord_offsetF  = DIM*jnrF;
 573             j_coord_offsetG  = DIM*jnrG;
 574             j_coord_offsetH  = DIM*jnrH;
 575
 576             /* load j atom coordinates */
 577             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
 578                                                  x+j_coord_offsetC+DIM,x+j_coord_offsetD+DIM,
 579                                                  x+j_coord_offsetE+DIM,x+j_coord_offsetF+DIM,
 580                                                  x+j_coord_offsetG+DIM,x+j_coord_offsetH+DIM,
 581                                                  &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
 582
 583             /* Calculate displacement vector */
 584             dx11             = _mm256_sub_ps(ix1,jx1);
 585             dy11             = _mm256_sub_ps(iy1,jy1);
 586             dz11             = _mm256_sub_ps(iz1,jz1);
 587             dx12             = _mm256_sub_ps(ix1,jx2);
 588             dy12             = _mm256_sub_ps(iy1,jy2);
 589             dz12             = _mm256_sub_ps(iz1,jz2);
 590             dx13             = _mm256_sub_ps(ix1,jx3);
 591             dy13             = _mm256_sub_ps(iy1,jy3);
 592             dz13             = _mm256_sub_ps(iz1,jz3);
 593             dx21             = _mm256_sub_ps(ix2,jx1);
 594             dy21             = _mm256_sub_ps(iy2,jy1);
 595             dz21             = _mm256_sub_ps(iz2,jz1);
 596             dx22             = _mm256_sub_ps(ix2,jx2);
 597             dy22             = _mm256_sub_ps(iy2,jy2);
 598             dz22             = _mm256_sub_ps(iz2,jz2);
 599             dx23             = _mm256_sub_ps(ix2,jx3);
 600             dy23             = _mm256_sub_ps(iy2,jy3);
 601             dz23             = _mm256_sub_ps(iz2,jz3);
 602             dx31             = _mm256_sub_ps(ix3,jx1);
 603             dy31             = _mm256_sub_ps(iy3,jy1);
 604             dz31             = _mm256_sub_ps(iz3,jz1);
 605             dx32             = _mm256_sub_ps(ix3,jx2);
 606             dy32             = _mm256_sub_ps(iy3,jy2);
 607             dz32             = _mm256_sub_ps(iz3,jz2);
 608             dx33             = _mm256_sub_ps(ix3,jx3);
 609             dy33             = _mm256_sub_ps(iy3,jy3);
 610             dz33             = _mm256_sub_ps(iz3,jz3);
 611
 612             /* Calculate squared distance and things based on it */
 613             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 614             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 615             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
 616             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 617             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 618             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
 619             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
 620             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
 621             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
 622
 623             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 624             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 625             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
 626             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 627             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 628             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
 629             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
 630             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
 631             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
 632
 633             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 634             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 635             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
 636             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 637             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 638             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
 639             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
 640             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
 641             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
 642
 643             fjx1             = _mm256_setzero_ps();
 644             fjy1             = _mm256_setzero_ps();
 645             fjz1             = _mm256_setzero_ps();
 646             fjx2             = _mm256_setzero_ps();
 647             fjy2             = _mm256_setzero_ps();
 648             fjz2             = _mm256_setzero_ps();
 649             fjx3             = _mm256_setzero_ps();
 650             fjy3             = _mm256_setzero_ps();
 651             fjz3             = _mm256_setzero_ps();
 652
 653             /**************************
 654              * CALCULATE INTERACTIONS *
 655              **************************/
 656
 657             /* COULOMB ELECTROSTATICS */
 658             velec            = _mm256_mul_ps(qq11,rinv11);
 659             felec            = _mm256_mul_ps(velec,rinvsq11);
 660
 661             /* Update potential sum for this i atom from the interaction with this j atom. */
 662             velec            = _mm256_andnot_ps(dummy_mask,velec);
 663             velecsum         = _mm256_add_ps(velecsum,velec);
 664
 665             fscal            = felec;
 666
 667             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 668
 669             /* Calculate temporary vectorial force */
 670             tx               = _mm256_mul_ps(fscal,dx11);
 671             ty               = _mm256_mul_ps(fscal,dy11);
 672             tz               = _mm256_mul_ps(fscal,dz11);
 673
 674             /* Update vectorial force */
 675             fix1             = _mm256_add_ps(fix1,tx);
 676             fiy1             = _mm256_add_ps(fiy1,ty);
 677             fiz1             = _mm256_add_ps(fiz1,tz);
 678
 679             fjx1             = _mm256_add_ps(fjx1,tx);
 680             fjy1             = _mm256_add_ps(fjy1,ty);
 681             fjz1             = _mm256_add_ps(fjz1,tz);
 682
 683             /**************************
 684              * CALCULATE INTERACTIONS *
 685              **************************/
 686
 687             /* COULOMB ELECTROSTATICS */
 688             velec            = _mm256_mul_ps(qq12,rinv12);
 689             felec            = _mm256_mul_ps(velec,rinvsq12);
 690
 691             /* Update potential sum for this i atom from the interaction with this j atom. */
 692             velec            = _mm256_andnot_ps(dummy_mask,velec);
 693             velecsum         = _mm256_add_ps(velecsum,velec);
 694
 695             fscal            = felec;
 696
 697             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 698
 699             /* Calculate temporary vectorial force */
 700             tx               = _mm256_mul_ps(fscal,dx12);
 701             ty               = _mm256_mul_ps(fscal,dy12);
 702             tz               = _mm256_mul_ps(fscal,dz12);
 703
 704             /* Update vectorial force */
 705             fix1             = _mm256_add_ps(fix1,tx);
 706             fiy1             = _mm256_add_ps(fiy1,ty);
 707             fiz1             = _mm256_add_ps(fiz1,tz);
 708
 709             fjx2             = _mm256_add_ps(fjx2,tx);
 710             fjy2             = _mm256_add_ps(fjy2,ty);
 711             fjz2             = _mm256_add_ps(fjz2,tz);
 712
 713             /**************************
 714              * CALCULATE INTERACTIONS *
 715              **************************/
 716
 717             /* COULOMB ELECTROSTATICS */
 718             velec            = _mm256_mul_ps(qq13,rinv13);
 719             felec            = _mm256_mul_ps(velec,rinvsq13);
 720
 721             /* Update potential sum for this i atom from the interaction with this j atom. */
 722             velec            = _mm256_andnot_ps(dummy_mask,velec);
 723             velecsum         = _mm256_add_ps(velecsum,velec);
 724
 725             fscal            = felec;
 726
 727             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 728
 729             /* Calculate temporary vectorial force */
 730             tx               = _mm256_mul_ps(fscal,dx13);
 731             ty               = _mm256_mul_ps(fscal,dy13);
 732             tz               = _mm256_mul_ps(fscal,dz13);
 733
 734             /* Update vectorial force */
 735             fix1             = _mm256_add_ps(fix1,tx);
 736             fiy1             = _mm256_add_ps(fiy1,ty);
 737             fiz1             = _mm256_add_ps(fiz1,tz);
 738
 739             fjx3             = _mm256_add_ps(fjx3,tx);
 740             fjy3             = _mm256_add_ps(fjy3,ty);
 741             fjz3             = _mm256_add_ps(fjz3,tz);
 742
 743             /**************************
 744              * CALCULATE INTERACTIONS *
 745              **************************/
 746
 747             /* COULOMB ELECTROSTATICS */
 748             velec            = _mm256_mul_ps(qq21,rinv21);
 749             felec            = _mm256_mul_ps(velec,rinvsq21);
 750
 751             /* Update potential sum for this i atom from the interaction with this j atom. */
 752             velec            = _mm256_andnot_ps(dummy_mask,velec);
 753             velecsum         = _mm256_add_ps(velecsum,velec);
 754
 755             fscal            = felec;
 756
 757             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 758
 759             /* Calculate temporary vectorial force */
 760             tx               = _mm256_mul_ps(fscal,dx21);
 761             ty               = _mm256_mul_ps(fscal,dy21);
 762             tz               = _mm256_mul_ps(fscal,dz21);
 763
 764             /* Update vectorial force */
 765             fix2             = _mm256_add_ps(fix2,tx);
 766             fiy2             = _mm256_add_ps(fiy2,ty);
 767             fiz2             = _mm256_add_ps(fiz2,tz);
 768
 769             fjx1             = _mm256_add_ps(fjx1,tx);
 770             fjy1             = _mm256_add_ps(fjy1,ty);
 771             fjz1             = _mm256_add_ps(fjz1,tz);
 772
 773             /**************************
 774              * CALCULATE INTERACTIONS *
 775              **************************/
 776
 777             /* COULOMB ELECTROSTATICS */
 778             velec            = _mm256_mul_ps(qq22,rinv22);
 779             felec            = _mm256_mul_ps(velec,rinvsq22);
 780
 781             /* Update potential sum for this i atom from the interaction with this j atom. */
 782             velec            = _mm256_andnot_ps(dummy_mask,velec);
 783             velecsum         = _mm256_add_ps(velecsum,velec);
 784
 785             fscal            = felec;
 786
 787             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 788
 789             /* Calculate temporary vectorial force */
 790             tx               = _mm256_mul_ps(fscal,dx22);
 791             ty               = _mm256_mul_ps(fscal,dy22);
 792             tz               = _mm256_mul_ps(fscal,dz22);
 793
 794             /* Update vectorial force */
 795             fix2             = _mm256_add_ps(fix2,tx);
 796             fiy2             = _mm256_add_ps(fiy2,ty);
 797             fiz2             = _mm256_add_ps(fiz2,tz);
 798
 799             fjx2             = _mm256_add_ps(fjx2,tx);
 800             fjy2             = _mm256_add_ps(fjy2,ty);
 801             fjz2             = _mm256_add_ps(fjz2,tz);
 802
 803             /**************************
 804              * CALCULATE INTERACTIONS *
 805              **************************/
 806
 807             /* COULOMB ELECTROSTATICS */
 808             velec            = _mm256_mul_ps(qq23,rinv23);
 809             felec            = _mm256_mul_ps(velec,rinvsq23);
 810
 811             /* Update potential sum for this i atom from the interaction with this j atom. */
 812             velec            = _mm256_andnot_ps(dummy_mask,velec);
 813             velecsum         = _mm256_add_ps(velecsum,velec);
 814
 815             fscal            = felec;
 816
 817             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 818
 819             /* Calculate temporary vectorial force */
 820             tx               = _mm256_mul_ps(fscal,dx23);
 821             ty               = _mm256_mul_ps(fscal,dy23);
 822             tz               = _mm256_mul_ps(fscal,dz23);
 823
 824             /* Update vectorial force */
 825             fix2             = _mm256_add_ps(fix2,tx);
 826             fiy2             = _mm256_add_ps(fiy2,ty);
 827             fiz2             = _mm256_add_ps(fiz2,tz);
 828
 829             fjx3             = _mm256_add_ps(fjx3,tx);
 830             fjy3             = _mm256_add_ps(fjy3,ty);
 831             fjz3             = _mm256_add_ps(fjz3,tz);
 832
 833             /**************************
 834              * CALCULATE INTERACTIONS *
 835              **************************/
 836
 837             /* COULOMB ELECTROSTATICS */
 838             velec            = _mm256_mul_ps(qq31,rinv31);
 839             felec            = _mm256_mul_ps(velec,rinvsq31);
 840
 841             /* Update potential sum for this i atom from the interaction with this j atom. */
 842             velec            = _mm256_andnot_ps(dummy_mask,velec);
 843             velecsum         = _mm256_add_ps(velecsum,velec);
 844
 845             fscal            = felec;
 846
 847             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 848
 849             /* Calculate temporary vectorial force */
 850             tx               = _mm256_mul_ps(fscal,dx31);
 851             ty               = _mm256_mul_ps(fscal,dy31);
 852             tz               = _mm256_mul_ps(fscal,dz31);
 853
 854             /* Update vectorial force */
 855             fix3             = _mm256_add_ps(fix3,tx);
 856             fiy3             = _mm256_add_ps(fiy3,ty);
 857             fiz3             = _mm256_add_ps(fiz3,tz);
 858
 859             fjx1             = _mm256_add_ps(fjx1,tx);
 860             fjy1             = _mm256_add_ps(fjy1,ty);
 861             fjz1             = _mm256_add_ps(fjz1,tz);
 862
 863             /**************************
 864              * CALCULATE INTERACTIONS *
 865              **************************/
 866
 867             /* COULOMB ELECTROSTATICS */
 868             velec            = _mm256_mul_ps(qq32,rinv32);
 869             felec            = _mm256_mul_ps(velec,rinvsq32);
 870
 871             /* Update potential sum for this i atom from the interaction with this j atom. */
 872             velec            = _mm256_andnot_ps(dummy_mask,velec);
 873             velecsum         = _mm256_add_ps(velecsum,velec);
 874
 875             fscal            = felec;
 876
 877             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 878
 879             /* Calculate temporary vectorial force */
 880             tx               = _mm256_mul_ps(fscal,dx32);
 881             ty               = _mm256_mul_ps(fscal,dy32);
 882             tz               = _mm256_mul_ps(fscal,dz32);
 883
 884             /* Update vectorial force */
 885             fix3             = _mm256_add_ps(fix3,tx);
 886             fiy3             = _mm256_add_ps(fiy3,ty);
 887             fiz3             = _mm256_add_ps(fiz3,tz);
 888
 889             fjx2             = _mm256_add_ps(fjx2,tx);
 890             fjy2             = _mm256_add_ps(fjy2,ty);
 891             fjz2             = _mm256_add_ps(fjz2,tz);
 892
 893             /**************************
 894              * CALCULATE INTERACTIONS *
 895              **************************/
 896
 897             /* COULOMB ELECTROSTATICS */
 898             velec            = _mm256_mul_ps(qq33,rinv33);
 899             felec            = _mm256_mul_ps(velec,rinvsq33);
 900
 901             /* Update potential sum for this i atom from the interaction with this j atom. */
 902             velec            = _mm256_andnot_ps(dummy_mask,velec);
 903             velecsum         = _mm256_add_ps(velecsum,velec);
 904
 905             fscal            = felec;
 906
 907             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 908
 909             /* Calculate temporary vectorial force */
 910             tx               = _mm256_mul_ps(fscal,dx33);
 911             ty               = _mm256_mul_ps(fscal,dy33);
 912             tz               = _mm256_mul_ps(fscal,dz33);
 913
 914             /* Update vectorial force */
 915             fix3             = _mm256_add_ps(fix3,tx);
 916             fiy3             = _mm256_add_ps(fiy3,ty);
 917             fiz3             = _mm256_add_ps(fiz3,tz);
 918
 919             fjx3             = _mm256_add_ps(fjx3,tx);
 920             fjy3             = _mm256_add_ps(fjy3,ty);
 921             fjz3             = _mm256_add_ps(fjz3,tz);
 922
 923             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
 924             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
 925             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
 926             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
 927             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
 928             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
 929             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
 930             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
 931
 932             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA+DIM,fjptrB+DIM,fjptrC+DIM,fjptrD+DIM,
 933                                                       fjptrE+DIM,fjptrF+DIM,fjptrG+DIM,fjptrH+DIM,
 934                                                       fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 935
 936             /* Inner loop uses 243 flops */
 937         }
 938
 939         /* End of innermost loop */
 940
 941         gmx_mm256_update_iforce_3atom_swizzle_ps(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 942                                                  f+i_coord_offset+DIM,fshift+i_shift_offset);
 943
 944         ggid                        = gid[iidx];
 945         /* Update potential energies */
 946         gmx_mm256_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
 947
 948         /* Increment number of inner iterations */
 949         inneriter                  += j_index_end - j_index_start;
 950
 951         /* Outer loop uses 19 flops */
 952     }
 953
 954     /* Increment number of outer iterations */
 955     outeriter        += nri;
 956
 957     /* Update outer/inner flops */
 958
 959     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4W4_VF,outeriter*19 + inneriter*243);
 960 }
 961 /*
 962  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwNone_GeomW4W4_F_avx_256_single
 963  * Electrostatics interaction: Coulomb
 964  * VdW interaction:            None
 965  * Geometry:                   Water4-Water4
 966  * Calculate force/pot:        Force
 967  */
 968 void
 969 nb_kernel_ElecCoul_VdwNone_GeomW4W4_F_avx_256_single
 970                     (t_nblist * gmx_restrict                nlist,
 971                      rvec * gmx_restrict                    xx,
 972                      rvec * gmx_restrict                    ff,
 973                      t_forcerec * gmx_restrict              fr,
 974                      t_mdatoms * gmx_restrict               mdatoms,
 975                      nb_kernel_data_t * gmx_restrict        kernel_data,
 976                      t_nrnb * gmx_restrict                  nrnb)
 977 {
 978     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 979      * just 0 for non-waters.
 980      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
 981      * jnr indices corresponding to data put in the four positions in the SIMD register.
 982      */
 983     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 984     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 985     int              jnrA,jnrB,jnrC,jnrD;
 986     int              jnrE,jnrF,jnrG,jnrH;
 987     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
 988     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
 989     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
 990     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
 991     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 992     real             rcutoff_scalar;
 993     real             *shiftvec,*fshift,*x,*f;
 994     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
 995     real             scratch[4*DIM];
 996     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 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     real *           vdwioffsetptr3;
1002     __m256           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
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     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D,vdwjidx3E,vdwjidx3F,vdwjidx3G,vdwjidx3H;
1008     __m256           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1009     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1010     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1011     __m256           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1012     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1013     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1014     __m256           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1015     __m256           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1016     __m256           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1017     __m256           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1018     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
1019     real             *charge;
1020     __m256           dummy_mask,cutoff_mask;
1021     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
1022     __m256           one     = _mm256_set1_ps(1.0);
1023     __m256           two     = _mm256_set1_ps(2.0);
1024     x                = xx[0];
1025     f                = ff[0];
1026
1027     nri              = nlist->nri;
1028     iinr             = nlist->iinr;
1029     jindex           = nlist->jindex;
1030     jjnr             = nlist->jjnr;
1031     shiftidx         = nlist->shift;
1032     gid              = nlist->gid;
1033     shiftvec         = fr->shift_vec[0];
1034     fshift           = fr->fshift[0];
1035     facel            = _mm256_set1_ps(fr->epsfac);
1036     charge           = mdatoms->chargeA;
1037
1038     /* Setup water-specific parameters */
1039     inr              = nlist->iinr[0];
1040     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
1041     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
1042     iq3              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+3]));
1043
1044     jq1              = _mm256_set1_ps(charge[inr+1]);
1045     jq2              = _mm256_set1_ps(charge[inr+2]);
1046     jq3              = _mm256_set1_ps(charge[inr+3]);
1047     qq11             = _mm256_mul_ps(iq1,jq1);
1048     qq12             = _mm256_mul_ps(iq1,jq2);
1049     qq13             = _mm256_mul_ps(iq1,jq3);
1050     qq21             = _mm256_mul_ps(iq2,jq1);
1051     qq22             = _mm256_mul_ps(iq2,jq2);
1052     qq23             = _mm256_mul_ps(iq2,jq3);
1053     qq31             = _mm256_mul_ps(iq3,jq1);
1054     qq32             = _mm256_mul_ps(iq3,jq2);
1055     qq33             = _mm256_mul_ps(iq3,jq3);
1056
1057     /* Avoid stupid compiler warnings */
1058     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
1059     j_coord_offsetA = 0;
1060     j_coord_offsetB = 0;
1061     j_coord_offsetC = 0;
1062     j_coord_offsetD = 0;
1063     j_coord_offsetE = 0;
1064     j_coord_offsetF = 0;
1065     j_coord_offsetG = 0;
1066     j_coord_offsetH = 0;
1067
1068     outeriter        = 0;
1069     inneriter        = 0;
1070
1071     for(iidx=0;iidx<4*DIM;iidx++)
1072     {
1073         scratch[iidx] = 0.0;
1074     }
1075
1076     /* Start outer loop over neighborlists */
1077     for(iidx=0; iidx<nri; iidx++)
1078     {
1079         /* Load shift vector for this list */
1080         i_shift_offset   = DIM*shiftidx[iidx];
1081
1082         /* Load limits for loop over neighbors */
1083         j_index_start    = jindex[iidx];
1084         j_index_end      = jindex[iidx+1];
1085
1086         /* Get outer coordinate index */
1087         inr              = iinr[iidx];
1088         i_coord_offset   = DIM*inr;
1089
1090         /* Load i particle coords and add shift vector */
1091         gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
1092                                                     &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1093
1094         fix1             = _mm256_setzero_ps();
1095         fiy1             = _mm256_setzero_ps();
1096         fiz1             = _mm256_setzero_ps();
1097         fix2             = _mm256_setzero_ps();
1098         fiy2             = _mm256_setzero_ps();
1099         fiz2             = _mm256_setzero_ps();
1100         fix3             = _mm256_setzero_ps();
1101         fiy3             = _mm256_setzero_ps();
1102         fiz3             = _mm256_setzero_ps();
1103
1104         /* Start inner kernel loop */
1105         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
1106         {
1107
1108             /* Get j neighbor index, and coordinate index */
1109             jnrA             = jjnr[jidx];
1110             jnrB             = jjnr[jidx+1];
1111             jnrC             = jjnr[jidx+2];
1112             jnrD             = jjnr[jidx+3];
1113             jnrE             = jjnr[jidx+4];
1114             jnrF             = jjnr[jidx+5];
1115             jnrG             = jjnr[jidx+6];
1116             jnrH             = jjnr[jidx+7];
1117             j_coord_offsetA  = DIM*jnrA;
1118             j_coord_offsetB  = DIM*jnrB;
1119             j_coord_offsetC  = DIM*jnrC;
1120             j_coord_offsetD  = DIM*jnrD;
1121             j_coord_offsetE  = DIM*jnrE;
1122             j_coord_offsetF  = DIM*jnrF;
1123             j_coord_offsetG  = DIM*jnrG;
1124             j_coord_offsetH  = DIM*jnrH;
1125
1126             /* load j atom coordinates */
1127             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
1128                                                  x+j_coord_offsetC+DIM,x+j_coord_offsetD+DIM,
1129                                                  x+j_coord_offsetE+DIM,x+j_coord_offsetF+DIM,
1130                                                  x+j_coord_offsetG+DIM,x+j_coord_offsetH+DIM,
1131                                                  &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
1132
1133             /* Calculate displacement vector */
1134             dx11             = _mm256_sub_ps(ix1,jx1);
1135             dy11             = _mm256_sub_ps(iy1,jy1);
1136             dz11             = _mm256_sub_ps(iz1,jz1);
1137             dx12             = _mm256_sub_ps(ix1,jx2);
1138             dy12             = _mm256_sub_ps(iy1,jy2);
1139             dz12             = _mm256_sub_ps(iz1,jz2);
1140             dx13             = _mm256_sub_ps(ix1,jx3);
1141             dy13             = _mm256_sub_ps(iy1,jy3);
1142             dz13             = _mm256_sub_ps(iz1,jz3);
1143             dx21             = _mm256_sub_ps(ix2,jx1);
1144             dy21             = _mm256_sub_ps(iy2,jy1);
1145             dz21             = _mm256_sub_ps(iz2,jz1);
1146             dx22             = _mm256_sub_ps(ix2,jx2);
1147             dy22             = _mm256_sub_ps(iy2,jy2);
1148             dz22             = _mm256_sub_ps(iz2,jz2);
1149             dx23             = _mm256_sub_ps(ix2,jx3);
1150             dy23             = _mm256_sub_ps(iy2,jy3);
1151             dz23             = _mm256_sub_ps(iz2,jz3);
1152             dx31             = _mm256_sub_ps(ix3,jx1);
1153             dy31             = _mm256_sub_ps(iy3,jy1);
1154             dz31             = _mm256_sub_ps(iz3,jz1);
1155             dx32             = _mm256_sub_ps(ix3,jx2);
1156             dy32             = _mm256_sub_ps(iy3,jy2);
1157             dz32             = _mm256_sub_ps(iz3,jz2);
1158             dx33             = _mm256_sub_ps(ix3,jx3);
1159             dy33             = _mm256_sub_ps(iy3,jy3);
1160             dz33             = _mm256_sub_ps(iz3,jz3);
1161
1162             /* Calculate squared distance and things based on it */
1163             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1164             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1165             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
1166             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1167             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1168             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
1169             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
1170             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
1171             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
1172
1173             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1174             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1175             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
1176             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1177             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1178             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
1179             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
1180             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
1181             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
1182
1183             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1184             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1185             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
1186             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1187             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1188             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
1189             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
1190             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
1191             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
1192
1193             fjx1             = _mm256_setzero_ps();
1194             fjy1             = _mm256_setzero_ps();
1195             fjz1             = _mm256_setzero_ps();
1196             fjx2             = _mm256_setzero_ps();
1197             fjy2             = _mm256_setzero_ps();
1198             fjz2             = _mm256_setzero_ps();
1199             fjx3             = _mm256_setzero_ps();
1200             fjy3             = _mm256_setzero_ps();
1201             fjz3             = _mm256_setzero_ps();
1202
1203             /**************************
1204              * CALCULATE INTERACTIONS *
1205              **************************/
1206
1207             /* COULOMB ELECTROSTATICS */
1208             velec            = _mm256_mul_ps(qq11,rinv11);
1209             felec            = _mm256_mul_ps(velec,rinvsq11);
1210
1211             fscal            = felec;
1212
1213             /* Calculate temporary vectorial force */
1214             tx               = _mm256_mul_ps(fscal,dx11);
1215             ty               = _mm256_mul_ps(fscal,dy11);
1216             tz               = _mm256_mul_ps(fscal,dz11);
1217
1218             /* Update vectorial force */
1219             fix1             = _mm256_add_ps(fix1,tx);
1220             fiy1             = _mm256_add_ps(fiy1,ty);
1221             fiz1             = _mm256_add_ps(fiz1,tz);
1222
1223             fjx1             = _mm256_add_ps(fjx1,tx);
1224             fjy1             = _mm256_add_ps(fjy1,ty);
1225             fjz1             = _mm256_add_ps(fjz1,tz);
1226
1227             /**************************
1228              * CALCULATE INTERACTIONS *
1229              **************************/
1230
1231             /* COULOMB ELECTROSTATICS */
1232             velec            = _mm256_mul_ps(qq12,rinv12);
1233             felec            = _mm256_mul_ps(velec,rinvsq12);
1234
1235             fscal            = felec;
1236
1237             /* Calculate temporary vectorial force */
1238             tx               = _mm256_mul_ps(fscal,dx12);
1239             ty               = _mm256_mul_ps(fscal,dy12);
1240             tz               = _mm256_mul_ps(fscal,dz12);
1241
1242             /* Update vectorial force */
1243             fix1             = _mm256_add_ps(fix1,tx);
1244             fiy1             = _mm256_add_ps(fiy1,ty);
1245             fiz1             = _mm256_add_ps(fiz1,tz);
1246
1247             fjx2             = _mm256_add_ps(fjx2,tx);
1248             fjy2             = _mm256_add_ps(fjy2,ty);
1249             fjz2             = _mm256_add_ps(fjz2,tz);
1250
1251             /**************************
1252              * CALCULATE INTERACTIONS *
1253              **************************/
1254
1255             /* COULOMB ELECTROSTATICS */
1256             velec            = _mm256_mul_ps(qq13,rinv13);
1257             felec            = _mm256_mul_ps(velec,rinvsq13);
1258
1259             fscal            = felec;
1260
1261             /* Calculate temporary vectorial force */
1262             tx               = _mm256_mul_ps(fscal,dx13);
1263             ty               = _mm256_mul_ps(fscal,dy13);
1264             tz               = _mm256_mul_ps(fscal,dz13);
1265
1266             /* Update vectorial force */
1267             fix1             = _mm256_add_ps(fix1,tx);
1268             fiy1             = _mm256_add_ps(fiy1,ty);
1269             fiz1             = _mm256_add_ps(fiz1,tz);
1270
1271             fjx3             = _mm256_add_ps(fjx3,tx);
1272             fjy3             = _mm256_add_ps(fjy3,ty);
1273             fjz3             = _mm256_add_ps(fjz3,tz);
1274
1275             /**************************
1276              * CALCULATE INTERACTIONS *
1277              **************************/
1278
1279             /* COULOMB ELECTROSTATICS */
1280             velec            = _mm256_mul_ps(qq21,rinv21);
1281             felec            = _mm256_mul_ps(velec,rinvsq21);
1282
1283             fscal            = felec;
1284
1285             /* Calculate temporary vectorial force */
1286             tx               = _mm256_mul_ps(fscal,dx21);
1287             ty               = _mm256_mul_ps(fscal,dy21);
1288             tz               = _mm256_mul_ps(fscal,dz21);
1289
1290             /* Update vectorial force */
1291             fix2             = _mm256_add_ps(fix2,tx);
1292             fiy2             = _mm256_add_ps(fiy2,ty);
1293             fiz2             = _mm256_add_ps(fiz2,tz);
1294
1295             fjx1             = _mm256_add_ps(fjx1,tx);
1296             fjy1             = _mm256_add_ps(fjy1,ty);
1297             fjz1             = _mm256_add_ps(fjz1,tz);
1298
1299             /**************************
1300              * CALCULATE INTERACTIONS *
1301              **************************/
1302
1303             /* COULOMB ELECTROSTATICS */
1304             velec            = _mm256_mul_ps(qq22,rinv22);
1305             felec            = _mm256_mul_ps(velec,rinvsq22);
1306
1307             fscal            = felec;
1308
1309             /* Calculate temporary vectorial force */
1310             tx               = _mm256_mul_ps(fscal,dx22);
1311             ty               = _mm256_mul_ps(fscal,dy22);
1312             tz               = _mm256_mul_ps(fscal,dz22);
1313
1314             /* Update vectorial force */
1315             fix2             = _mm256_add_ps(fix2,tx);
1316             fiy2             = _mm256_add_ps(fiy2,ty);
1317             fiz2             = _mm256_add_ps(fiz2,tz);
1318
1319             fjx2             = _mm256_add_ps(fjx2,tx);
1320             fjy2             = _mm256_add_ps(fjy2,ty);
1321             fjz2             = _mm256_add_ps(fjz2,tz);
1322
1323             /**************************
1324              * CALCULATE INTERACTIONS *
1325              **************************/
1326
1327             /* COULOMB ELECTROSTATICS */
1328             velec            = _mm256_mul_ps(qq23,rinv23);
1329             felec            = _mm256_mul_ps(velec,rinvsq23);
1330
1331             fscal            = felec;
1332
1333             /* Calculate temporary vectorial force */
1334             tx               = _mm256_mul_ps(fscal,dx23);
1335             ty               = _mm256_mul_ps(fscal,dy23);
1336             tz               = _mm256_mul_ps(fscal,dz23);
1337
1338             /* Update vectorial force */
1339             fix2             = _mm256_add_ps(fix2,tx);
1340             fiy2             = _mm256_add_ps(fiy2,ty);
1341             fiz2             = _mm256_add_ps(fiz2,tz);
1342
1343             fjx3             = _mm256_add_ps(fjx3,tx);
1344             fjy3             = _mm256_add_ps(fjy3,ty);
1345             fjz3             = _mm256_add_ps(fjz3,tz);
1346
1347             /**************************
1348              * CALCULATE INTERACTIONS *
1349              **************************/
1350
1351             /* COULOMB ELECTROSTATICS */
1352             velec            = _mm256_mul_ps(qq31,rinv31);
1353             felec            = _mm256_mul_ps(velec,rinvsq31);
1354
1355             fscal            = felec;
1356
1357             /* Calculate temporary vectorial force */
1358             tx               = _mm256_mul_ps(fscal,dx31);
1359             ty               = _mm256_mul_ps(fscal,dy31);
1360             tz               = _mm256_mul_ps(fscal,dz31);
1361
1362             /* Update vectorial force */
1363             fix3             = _mm256_add_ps(fix3,tx);
1364             fiy3             = _mm256_add_ps(fiy3,ty);
1365             fiz3             = _mm256_add_ps(fiz3,tz);
1366
1367             fjx1             = _mm256_add_ps(fjx1,tx);
1368             fjy1             = _mm256_add_ps(fjy1,ty);
1369             fjz1             = _mm256_add_ps(fjz1,tz);
1370
1371             /**************************
1372              * CALCULATE INTERACTIONS *
1373              **************************/
1374
1375             /* COULOMB ELECTROSTATICS */
1376             velec            = _mm256_mul_ps(qq32,rinv32);
1377             felec            = _mm256_mul_ps(velec,rinvsq32);
1378
1379             fscal            = felec;
1380
1381             /* Calculate temporary vectorial force */
1382             tx               = _mm256_mul_ps(fscal,dx32);
1383             ty               = _mm256_mul_ps(fscal,dy32);
1384             tz               = _mm256_mul_ps(fscal,dz32);
1385
1386             /* Update vectorial force */
1387             fix3             = _mm256_add_ps(fix3,tx);
1388             fiy3             = _mm256_add_ps(fiy3,ty);
1389             fiz3             = _mm256_add_ps(fiz3,tz);
1390
1391             fjx2             = _mm256_add_ps(fjx2,tx);
1392             fjy2             = _mm256_add_ps(fjy2,ty);
1393             fjz2             = _mm256_add_ps(fjz2,tz);
1394
1395             /**************************
1396              * CALCULATE INTERACTIONS *
1397              **************************/
1398
1399             /* COULOMB ELECTROSTATICS */
1400             velec            = _mm256_mul_ps(qq33,rinv33);
1401             felec            = _mm256_mul_ps(velec,rinvsq33);
1402
1403             fscal            = felec;
1404
1405             /* Calculate temporary vectorial force */
1406             tx               = _mm256_mul_ps(fscal,dx33);
1407             ty               = _mm256_mul_ps(fscal,dy33);
1408             tz               = _mm256_mul_ps(fscal,dz33);
1409
1410             /* Update vectorial force */
1411             fix3             = _mm256_add_ps(fix3,tx);
1412             fiy3             = _mm256_add_ps(fiy3,ty);
1413             fiz3             = _mm256_add_ps(fiz3,tz);
1414
1415             fjx3             = _mm256_add_ps(fjx3,tx);
1416             fjy3             = _mm256_add_ps(fjy3,ty);
1417             fjz3             = _mm256_add_ps(fjz3,tz);
1418
1419             fjptrA             = f+j_coord_offsetA;
1420             fjptrB             = f+j_coord_offsetB;
1421             fjptrC             = f+j_coord_offsetC;
1422             fjptrD             = f+j_coord_offsetD;
1423             fjptrE             = f+j_coord_offsetE;
1424             fjptrF             = f+j_coord_offsetF;
1425             fjptrG             = f+j_coord_offsetG;
1426             fjptrH             = f+j_coord_offsetH;
1427
1428             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA+DIM,fjptrB+DIM,fjptrC+DIM,fjptrD+DIM,
1429                                                       fjptrE+DIM,fjptrF+DIM,fjptrG+DIM,fjptrH+DIM,
1430                                                       fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1431
1432             /* Inner loop uses 234 flops */
1433         }
1434
1435         if(jidx<j_index_end)
1436         {
1437
1438             /* Get j neighbor index, and coordinate index */
1439             jnrlistA         = jjnr[jidx];
1440             jnrlistB         = jjnr[jidx+1];
1441             jnrlistC         = jjnr[jidx+2];
1442             jnrlistD         = jjnr[jidx+3];
1443             jnrlistE         = jjnr[jidx+4];
1444             jnrlistF         = jjnr[jidx+5];
1445             jnrlistG         = jjnr[jidx+6];
1446             jnrlistH         = jjnr[jidx+7];
1447             /* Sign of each element will be negative for non-real atoms.
1448              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1449              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1450              */
1451             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
1452                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
1453
1454             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1455             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1456             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1457             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1458             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
1459             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
1460             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
1461             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
1462             j_coord_offsetA  = DIM*jnrA;
1463             j_coord_offsetB  = DIM*jnrB;
1464             j_coord_offsetC  = DIM*jnrC;
1465             j_coord_offsetD  = DIM*jnrD;
1466             j_coord_offsetE  = DIM*jnrE;
1467             j_coord_offsetF  = DIM*jnrF;
1468             j_coord_offsetG  = DIM*jnrG;
1469             j_coord_offsetH  = DIM*jnrH;
1470
1471             /* load j atom coordinates */
1472             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
1473                                                  x+j_coord_offsetC+DIM,x+j_coord_offsetD+DIM,
1474                                                  x+j_coord_offsetE+DIM,x+j_coord_offsetF+DIM,
1475                                                  x+j_coord_offsetG+DIM,x+j_coord_offsetH+DIM,
1476                                                  &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
1477
1478             /* Calculate displacement vector */
1479             dx11             = _mm256_sub_ps(ix1,jx1);
1480             dy11             = _mm256_sub_ps(iy1,jy1);
1481             dz11             = _mm256_sub_ps(iz1,jz1);
1482             dx12             = _mm256_sub_ps(ix1,jx2);
1483             dy12             = _mm256_sub_ps(iy1,jy2);
1484             dz12             = _mm256_sub_ps(iz1,jz2);
1485             dx13             = _mm256_sub_ps(ix1,jx3);
1486             dy13             = _mm256_sub_ps(iy1,jy3);
1487             dz13             = _mm256_sub_ps(iz1,jz3);
1488             dx21             = _mm256_sub_ps(ix2,jx1);
1489             dy21             = _mm256_sub_ps(iy2,jy1);
1490             dz21             = _mm256_sub_ps(iz2,jz1);
1491             dx22             = _mm256_sub_ps(ix2,jx2);
1492             dy22             = _mm256_sub_ps(iy2,jy2);
1493             dz22             = _mm256_sub_ps(iz2,jz2);
1494             dx23             = _mm256_sub_ps(ix2,jx3);
1495             dy23             = _mm256_sub_ps(iy2,jy3);
1496             dz23             = _mm256_sub_ps(iz2,jz3);
1497             dx31             = _mm256_sub_ps(ix3,jx1);
1498             dy31             = _mm256_sub_ps(iy3,jy1);
1499             dz31             = _mm256_sub_ps(iz3,jz1);
1500             dx32             = _mm256_sub_ps(ix3,jx2);
1501             dy32             = _mm256_sub_ps(iy3,jy2);
1502             dz32             = _mm256_sub_ps(iz3,jz2);
1503             dx33             = _mm256_sub_ps(ix3,jx3);
1504             dy33             = _mm256_sub_ps(iy3,jy3);
1505             dz33             = _mm256_sub_ps(iz3,jz3);
1506
1507             /* Calculate squared distance and things based on it */
1508             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1509             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1510             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
1511             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1512             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1513             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
1514             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
1515             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
1516             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
1517
1518             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1519             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1520             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
1521             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1522             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1523             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
1524             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
1525             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
1526             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
1527
1528             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1529             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1530             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
1531             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1532             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1533             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
1534             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
1535             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
1536             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
1537
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             fjx3             = _mm256_setzero_ps();
1545             fjy3             = _mm256_setzero_ps();
1546             fjz3             = _mm256_setzero_ps();
1547
1548             /**************************
1549              * CALCULATE INTERACTIONS *
1550              **************************/
1551
1552             /* COULOMB ELECTROSTATICS */
1553             velec            = _mm256_mul_ps(qq11,rinv11);
1554             felec            = _mm256_mul_ps(velec,rinvsq11);
1555
1556             fscal            = felec;
1557
1558             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1559
1560             /* Calculate temporary vectorial force */
1561             tx               = _mm256_mul_ps(fscal,dx11);
1562             ty               = _mm256_mul_ps(fscal,dy11);
1563             tz               = _mm256_mul_ps(fscal,dz11);
1564
1565             /* Update vectorial force */
1566             fix1             = _mm256_add_ps(fix1,tx);
1567             fiy1             = _mm256_add_ps(fiy1,ty);
1568             fiz1             = _mm256_add_ps(fiz1,tz);
1569
1570             fjx1             = _mm256_add_ps(fjx1,tx);
1571             fjy1             = _mm256_add_ps(fjy1,ty);
1572             fjz1             = _mm256_add_ps(fjz1,tz);
1573
1574             /**************************
1575              * CALCULATE INTERACTIONS *
1576              **************************/
1577
1578             /* COULOMB ELECTROSTATICS */
1579             velec            = _mm256_mul_ps(qq12,rinv12);
1580             felec            = _mm256_mul_ps(velec,rinvsq12);
1581
1582             fscal            = felec;
1583
1584             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1585
1586             /* Calculate temporary vectorial force */
1587             tx               = _mm256_mul_ps(fscal,dx12);
1588             ty               = _mm256_mul_ps(fscal,dy12);
1589             tz               = _mm256_mul_ps(fscal,dz12);
1590
1591             /* Update vectorial force */
1592             fix1             = _mm256_add_ps(fix1,tx);
1593             fiy1             = _mm256_add_ps(fiy1,ty);
1594             fiz1             = _mm256_add_ps(fiz1,tz);
1595
1596             fjx2             = _mm256_add_ps(fjx2,tx);
1597             fjy2             = _mm256_add_ps(fjy2,ty);
1598             fjz2             = _mm256_add_ps(fjz2,tz);
1599
1600             /**************************
1601              * CALCULATE INTERACTIONS *
1602              **************************/
1603
1604             /* COULOMB ELECTROSTATICS */
1605             velec            = _mm256_mul_ps(qq13,rinv13);
1606             felec            = _mm256_mul_ps(velec,rinvsq13);
1607
1608             fscal            = felec;
1609
1610             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1611
1612             /* Calculate temporary vectorial force */
1613             tx               = _mm256_mul_ps(fscal,dx13);
1614             ty               = _mm256_mul_ps(fscal,dy13);
1615             tz               = _mm256_mul_ps(fscal,dz13);
1616
1617             /* Update vectorial force */
1618             fix1             = _mm256_add_ps(fix1,tx);
1619             fiy1             = _mm256_add_ps(fiy1,ty);
1620             fiz1             = _mm256_add_ps(fiz1,tz);
1621
1622             fjx3             = _mm256_add_ps(fjx3,tx);
1623             fjy3             = _mm256_add_ps(fjy3,ty);
1624             fjz3             = _mm256_add_ps(fjz3,tz);
1625
1626             /**************************
1627              * CALCULATE INTERACTIONS *
1628              **************************/
1629
1630             /* COULOMB ELECTROSTATICS */
1631             velec            = _mm256_mul_ps(qq21,rinv21);
1632             felec            = _mm256_mul_ps(velec,rinvsq21);
1633
1634             fscal            = felec;
1635
1636             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1637
1638             /* Calculate temporary vectorial force */
1639             tx               = _mm256_mul_ps(fscal,dx21);
1640             ty               = _mm256_mul_ps(fscal,dy21);
1641             tz               = _mm256_mul_ps(fscal,dz21);
1642
1643             /* Update vectorial force */
1644             fix2             = _mm256_add_ps(fix2,tx);
1645             fiy2             = _mm256_add_ps(fiy2,ty);
1646             fiz2             = _mm256_add_ps(fiz2,tz);
1647
1648             fjx1             = _mm256_add_ps(fjx1,tx);
1649             fjy1             = _mm256_add_ps(fjy1,ty);
1650             fjz1             = _mm256_add_ps(fjz1,tz);
1651
1652             /**************************
1653              * CALCULATE INTERACTIONS *
1654              **************************/
1655
1656             /* COULOMB ELECTROSTATICS */
1657             velec            = _mm256_mul_ps(qq22,rinv22);
1658             felec            = _mm256_mul_ps(velec,rinvsq22);
1659
1660             fscal            = felec;
1661
1662             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1663
1664             /* Calculate temporary vectorial force */
1665             tx               = _mm256_mul_ps(fscal,dx22);
1666             ty               = _mm256_mul_ps(fscal,dy22);
1667             tz               = _mm256_mul_ps(fscal,dz22);
1668
1669             /* Update vectorial force */
1670             fix2             = _mm256_add_ps(fix2,tx);
1671             fiy2             = _mm256_add_ps(fiy2,ty);
1672             fiz2             = _mm256_add_ps(fiz2,tz);
1673
1674             fjx2             = _mm256_add_ps(fjx2,tx);
1675             fjy2             = _mm256_add_ps(fjy2,ty);
1676             fjz2             = _mm256_add_ps(fjz2,tz);
1677
1678             /**************************
1679              * CALCULATE INTERACTIONS *
1680              **************************/
1681
1682             /* COULOMB ELECTROSTATICS */
1683             velec            = _mm256_mul_ps(qq23,rinv23);
1684             felec            = _mm256_mul_ps(velec,rinvsq23);
1685
1686             fscal            = felec;
1687
1688             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1689
1690             /* Calculate temporary vectorial force */
1691             tx               = _mm256_mul_ps(fscal,dx23);
1692             ty               = _mm256_mul_ps(fscal,dy23);
1693             tz               = _mm256_mul_ps(fscal,dz23);
1694
1695             /* Update vectorial force */
1696             fix2             = _mm256_add_ps(fix2,tx);
1697             fiy2             = _mm256_add_ps(fiy2,ty);
1698             fiz2             = _mm256_add_ps(fiz2,tz);
1699
1700             fjx3             = _mm256_add_ps(fjx3,tx);
1701             fjy3             = _mm256_add_ps(fjy3,ty);
1702             fjz3             = _mm256_add_ps(fjz3,tz);
1703
1704             /**************************
1705              * CALCULATE INTERACTIONS *
1706              **************************/
1707
1708             /* COULOMB ELECTROSTATICS */
1709             velec            = _mm256_mul_ps(qq31,rinv31);
1710             felec            = _mm256_mul_ps(velec,rinvsq31);
1711
1712             fscal            = felec;
1713
1714             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1715
1716             /* Calculate temporary vectorial force */
1717             tx               = _mm256_mul_ps(fscal,dx31);
1718             ty               = _mm256_mul_ps(fscal,dy31);
1719             tz               = _mm256_mul_ps(fscal,dz31);
1720
1721             /* Update vectorial force */
1722             fix3             = _mm256_add_ps(fix3,tx);
1723             fiy3             = _mm256_add_ps(fiy3,ty);
1724             fiz3             = _mm256_add_ps(fiz3,tz);
1725
1726             fjx1             = _mm256_add_ps(fjx1,tx);
1727             fjy1             = _mm256_add_ps(fjy1,ty);
1728             fjz1             = _mm256_add_ps(fjz1,tz);
1729
1730             /**************************
1731              * CALCULATE INTERACTIONS *
1732              **************************/
1733
1734             /* COULOMB ELECTROSTATICS */
1735             velec            = _mm256_mul_ps(qq32,rinv32);
1736             felec            = _mm256_mul_ps(velec,rinvsq32);
1737
1738             fscal            = felec;
1739
1740             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1741
1742             /* Calculate temporary vectorial force */
1743             tx               = _mm256_mul_ps(fscal,dx32);
1744             ty               = _mm256_mul_ps(fscal,dy32);
1745             tz               = _mm256_mul_ps(fscal,dz32);
1746
1747             /* Update vectorial force */
1748             fix3             = _mm256_add_ps(fix3,tx);
1749             fiy3             = _mm256_add_ps(fiy3,ty);
1750             fiz3             = _mm256_add_ps(fiz3,tz);
1751
1752             fjx2             = _mm256_add_ps(fjx2,tx);
1753             fjy2             = _mm256_add_ps(fjy2,ty);
1754             fjz2             = _mm256_add_ps(fjz2,tz);
1755
1756             /**************************
1757              * CALCULATE INTERACTIONS *
1758              **************************/
1759
1760             /* COULOMB ELECTROSTATICS */
1761             velec            = _mm256_mul_ps(qq33,rinv33);
1762             felec            = _mm256_mul_ps(velec,rinvsq33);
1763
1764             fscal            = felec;
1765
1766             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1767
1768             /* Calculate temporary vectorial force */
1769             tx               = _mm256_mul_ps(fscal,dx33);
1770             ty               = _mm256_mul_ps(fscal,dy33);
1771             tz               = _mm256_mul_ps(fscal,dz33);
1772
1773             /* Update vectorial force */
1774             fix3             = _mm256_add_ps(fix3,tx);
1775             fiy3             = _mm256_add_ps(fiy3,ty);
1776             fiz3             = _mm256_add_ps(fiz3,tz);
1777
1778             fjx3             = _mm256_add_ps(fjx3,tx);
1779             fjy3             = _mm256_add_ps(fjy3,ty);
1780             fjz3             = _mm256_add_ps(fjz3,tz);
1781
1782             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1783             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1784             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1785             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1786             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1787             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1788             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
1789             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
1790
1791             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA+DIM,fjptrB+DIM,fjptrC+DIM,fjptrD+DIM,
1792                                                       fjptrE+DIM,fjptrF+DIM,fjptrG+DIM,fjptrH+DIM,
1793                                                       fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1794
1795             /* Inner loop uses 234 flops */
1796         }
1797
1798         /* End of innermost loop */
1799
1800         gmx_mm256_update_iforce_3atom_swizzle_ps(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1801                                                  f+i_coord_offset+DIM,fshift+i_shift_offset);
1802
1803         /* Increment number of inner iterations */
1804         inneriter                  += j_index_end - j_index_start;
1805
1806         /* Outer loop uses 18 flops */
1807     }
1808
1809     /* Increment number of outer iterations */
1810     outeriter        += nri;
1811
1812     /* Update outer/inner flops */
1813
1814     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4W4_F,outeriter*18 + inneriter*234);
1815 }