src/gromacs/gmxlib/nonbonded/nb_kernel_avx_256_single/nb_kernel_ElecCoul_VdwNone_GeomW4W4_avx_256_single.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS avx_256_single kernel generator.
  37  */
  38 #include "config.h"
  39
  40 #include <math.h>
  41
  42 #include "../nb_kernel.h"
  43 #include "types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "nrnb.h"
  46
  47 #include "gromacs/simd/math_x86_avx_256_single.h"
  48 #include "kernelutil_x86_avx_256_single.h"
  49
  50 /*
  51  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwNone_GeomW4W4_VF_avx_256_single
  52  * Electrostatics interaction: Coulomb
  53  * VdW interaction:            None
  54  * Geometry:                   Water4-Water4
  55  * Calculate force/pot:        PotentialAndForce
  56  */
  57 void
  58 nb_kernel_ElecCoul_VdwNone_GeomW4W4_VF_avx_256_single
  59                     (t_nblist                    * gmx_restrict       nlist,
  60                      rvec                        * gmx_restrict          xx,
  61                      rvec                        * gmx_restrict          ff,
  62                      t_forcerec                  * gmx_restrict          fr,
  63                      t_mdatoms                   * gmx_restrict     mdatoms,
  64                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  65                      t_nrnb                      * gmx_restrict        nrnb)
  66 {
  67     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  68      * just 0 for non-waters.
  69      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
  70      * jnr indices corresponding to data put in the four positions in the SIMD register.
  71      */
  72     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  73     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  74     int              jnrA,jnrB,jnrC,jnrD;
  75     int              jnrE,jnrF,jnrG,jnrH;
  76     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  77     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
  78     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  79     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
  80     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  81     real             rcutoff_scalar;
  82     real             *shiftvec,*fshift,*x,*f;
  83     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
  84     real             scratch[4*DIM];
  85     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  86     real *           vdwioffsetptr1;
  87     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  88     real *           vdwioffsetptr2;
  89     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  90     real *           vdwioffsetptr3;
  91     __m256           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  92     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
  93     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  94     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
  95     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  96     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D,vdwjidx3E,vdwjidx3F,vdwjidx3G,vdwjidx3H;
  97     __m256           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
  98     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
  99     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
 100     __m256           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
 101     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 102     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 103     __m256           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
 104     __m256           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
 105     __m256           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
 106     __m256           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
 107     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
 108     real             *charge;
 109     __m256           dummy_mask,cutoff_mask;
 110     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
 111     __m256           one     = _mm256_set1_ps(1.0);
 112     __m256           two     = _mm256_set1_ps(2.0);
 113     x                = xx[0];
 114     f                = ff[0];
 115
 116     nri              = nlist->nri;
 117     iinr             = nlist->iinr;
 118     jindex           = nlist->jindex;
 119     jjnr             = nlist->jjnr;
 120     shiftidx         = nlist->shift;
 121     gid              = nlist->gid;
 122     shiftvec         = fr->shift_vec[0];
 123     fshift           = fr->fshift[0];
 124     facel            = _mm256_set1_ps(fr->epsfac);
 125     charge           = mdatoms->chargeA;
 126
 127     /* Setup water-specific parameters */
 128     inr              = nlist->iinr[0];
 129     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
 130     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
 131     iq3              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+3]));
 132
 133     jq1              = _mm256_set1_ps(charge[inr+1]);
 134     jq2              = _mm256_set1_ps(charge[inr+2]);
 135     jq3              = _mm256_set1_ps(charge[inr+3]);
 136     qq11             = _mm256_mul_ps(iq1,jq1);
 137     qq12             = _mm256_mul_ps(iq1,jq2);
 138     qq13             = _mm256_mul_ps(iq1,jq3);
 139     qq21             = _mm256_mul_ps(iq2,jq1);
 140     qq22             = _mm256_mul_ps(iq2,jq2);
 141     qq23             = _mm256_mul_ps(iq2,jq3);
 142     qq31             = _mm256_mul_ps(iq3,jq1);
 143     qq32             = _mm256_mul_ps(iq3,jq2);
 144     qq33             = _mm256_mul_ps(iq3,jq3);
 145
 146     /* Avoid stupid compiler warnings */
 147     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
 148     j_coord_offsetA = 0;
 149     j_coord_offsetB = 0;
 150     j_coord_offsetC = 0;
 151     j_coord_offsetD = 0;
 152     j_coord_offsetE = 0;
 153     j_coord_offsetF = 0;
 154     j_coord_offsetG = 0;
 155     j_coord_offsetH = 0;
 156
 157     outeriter        = 0;
 158     inneriter        = 0;
 159
 160     for(iidx=0;iidx<4*DIM;iidx++)
 161     {
 162         scratch[iidx] = 0.0;
 163     }
 164
 165     /* Start outer loop over neighborlists */
 166     for(iidx=0; iidx<nri; iidx++)
 167     {
 168         /* Load shift vector for this list */
 169         i_shift_offset   = DIM*shiftidx[iidx];
 170
 171         /* Load limits for loop over neighbors */
 172         j_index_start    = jindex[iidx];
 173         j_index_end      = jindex[iidx+1];
 174
 175         /* Get outer coordinate index */
 176         inr              = iinr[iidx];
 177         i_coord_offset   = DIM*inr;
 178
 179         /* Load i particle coords and add shift vector */
 180         gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
 181                                                     &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 182
 183         fix1             = _mm256_setzero_ps();
 184         fiy1             = _mm256_setzero_ps();
 185         fiz1             = _mm256_setzero_ps();
 186         fix2             = _mm256_setzero_ps();
 187         fiy2             = _mm256_setzero_ps();
 188         fiz2             = _mm256_setzero_ps();
 189         fix3             = _mm256_setzero_ps();
 190         fiy3             = _mm256_setzero_ps();
 191         fiz3             = _mm256_setzero_ps();
 192
 193         /* Reset potential sums */
 194         velecsum         = _mm256_setzero_ps();
 195
 196         /* Start inner kernel loop */
 197         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
 198         {
 199
 200             /* Get j neighbor index, and coordinate index */
 201             jnrA             = jjnr[jidx];
 202             jnrB             = jjnr[jidx+1];
 203             jnrC             = jjnr[jidx+2];
 204             jnrD             = jjnr[jidx+3];
 205             jnrE             = jjnr[jidx+4];
 206             jnrF             = jjnr[jidx+5];
 207             jnrG             = jjnr[jidx+6];
 208             jnrH             = jjnr[jidx+7];
 209             j_coord_offsetA  = DIM*jnrA;
 210             j_coord_offsetB  = DIM*jnrB;
 211             j_coord_offsetC  = DIM*jnrC;
 212             j_coord_offsetD  = DIM*jnrD;
 213             j_coord_offsetE  = DIM*jnrE;
 214             j_coord_offsetF  = DIM*jnrF;
 215             j_coord_offsetG  = DIM*jnrG;
 216             j_coord_offsetH  = DIM*jnrH;
 217
 218             /* load j atom coordinates */
 219             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
 220                                                  x+j_coord_offsetC+DIM,x+j_coord_offsetD+DIM,
 221                                                  x+j_coord_offsetE+DIM,x+j_coord_offsetF+DIM,
 222                                                  x+j_coord_offsetG+DIM,x+j_coord_offsetH+DIM,
 223                                                  &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
 224
 225             /* Calculate displacement vector */
 226             dx11             = _mm256_sub_ps(ix1,jx1);
 227             dy11             = _mm256_sub_ps(iy1,jy1);
 228             dz11             = _mm256_sub_ps(iz1,jz1);
 229             dx12             = _mm256_sub_ps(ix1,jx2);
 230             dy12             = _mm256_sub_ps(iy1,jy2);
 231             dz12             = _mm256_sub_ps(iz1,jz2);
 232             dx13             = _mm256_sub_ps(ix1,jx3);
 233             dy13             = _mm256_sub_ps(iy1,jy3);
 234             dz13             = _mm256_sub_ps(iz1,jz3);
 235             dx21             = _mm256_sub_ps(ix2,jx1);
 236             dy21             = _mm256_sub_ps(iy2,jy1);
 237             dz21             = _mm256_sub_ps(iz2,jz1);
 238             dx22             = _mm256_sub_ps(ix2,jx2);
 239             dy22             = _mm256_sub_ps(iy2,jy2);
 240             dz22             = _mm256_sub_ps(iz2,jz2);
 241             dx23             = _mm256_sub_ps(ix2,jx3);
 242             dy23             = _mm256_sub_ps(iy2,jy3);
 243             dz23             = _mm256_sub_ps(iz2,jz3);
 244             dx31             = _mm256_sub_ps(ix3,jx1);
 245             dy31             = _mm256_sub_ps(iy3,jy1);
 246             dz31             = _mm256_sub_ps(iz3,jz1);
 247             dx32             = _mm256_sub_ps(ix3,jx2);
 248             dy32             = _mm256_sub_ps(iy3,jy2);
 249             dz32             = _mm256_sub_ps(iz3,jz2);
 250             dx33             = _mm256_sub_ps(ix3,jx3);
 251             dy33             = _mm256_sub_ps(iy3,jy3);
 252             dz33             = _mm256_sub_ps(iz3,jz3);
 253
 254             /* Calculate squared distance and things based on it */
 255             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 256             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 257             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
 258             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 259             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 260             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
 261             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
 262             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
 263             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
 264
 265             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 266             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 267             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
 268             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 269             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 270             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
 271             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
 272             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
 273             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
 274
 275             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 276             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 277             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
 278             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 279             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 280             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
 281             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
 282             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
 283             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
 284
 285             fjx1             = _mm256_setzero_ps();
 286             fjy1             = _mm256_setzero_ps();
 287             fjz1             = _mm256_setzero_ps();
 288             fjx2             = _mm256_setzero_ps();
 289             fjy2             = _mm256_setzero_ps();
 290             fjz2             = _mm256_setzero_ps();
 291             fjx3             = _mm256_setzero_ps();
 292             fjy3             = _mm256_setzero_ps();
 293             fjz3             = _mm256_setzero_ps();
 294
 295             /**************************
 296              * CALCULATE INTERACTIONS *
 297              **************************/
 298
 299             /* COULOMB ELECTROSTATICS */
 300             velec            = _mm256_mul_ps(qq11,rinv11);
 301             felec            = _mm256_mul_ps(velec,rinvsq11);
 302
 303             /* Update potential sum for this i atom from the interaction with this j atom. */
 304             velecsum         = _mm256_add_ps(velecsum,velec);
 305
 306             fscal            = felec;
 307
 308             /* Calculate temporary vectorial force */
 309             tx               = _mm256_mul_ps(fscal,dx11);
 310             ty               = _mm256_mul_ps(fscal,dy11);
 311             tz               = _mm256_mul_ps(fscal,dz11);
 312
 313             /* Update vectorial force */
 314             fix1             = _mm256_add_ps(fix1,tx);
 315             fiy1             = _mm256_add_ps(fiy1,ty);
 316             fiz1             = _mm256_add_ps(fiz1,tz);
 317
 318             fjx1             = _mm256_add_ps(fjx1,tx);
 319             fjy1             = _mm256_add_ps(fjy1,ty);
 320             fjz1             = _mm256_add_ps(fjz1,tz);
 321
 322             /**************************
 323              * CALCULATE INTERACTIONS *
 324              **************************/
 325
 326             /* COULOMB ELECTROSTATICS */
 327             velec            = _mm256_mul_ps(qq12,rinv12);
 328             felec            = _mm256_mul_ps(velec,rinvsq12);
 329
 330             /* Update potential sum for this i atom from the interaction with this j atom. */
 331             velecsum         = _mm256_add_ps(velecsum,velec);
 332
 333             fscal            = felec;
 334
 335             /* Calculate temporary vectorial force */
 336             tx               = _mm256_mul_ps(fscal,dx12);
 337             ty               = _mm256_mul_ps(fscal,dy12);
 338             tz               = _mm256_mul_ps(fscal,dz12);
 339
 340             /* Update vectorial force */
 341             fix1             = _mm256_add_ps(fix1,tx);
 342             fiy1             = _mm256_add_ps(fiy1,ty);
 343             fiz1             = _mm256_add_ps(fiz1,tz);
 344
 345             fjx2             = _mm256_add_ps(fjx2,tx);
 346             fjy2             = _mm256_add_ps(fjy2,ty);
 347             fjz2             = _mm256_add_ps(fjz2,tz);
 348
 349             /**************************
 350              * CALCULATE INTERACTIONS *
 351              **************************/
 352
 353             /* COULOMB ELECTROSTATICS */
 354             velec            = _mm256_mul_ps(qq13,rinv13);
 355             felec            = _mm256_mul_ps(velec,rinvsq13);
 356
 357             /* Update potential sum for this i atom from the interaction with this j atom. */
 358             velecsum         = _mm256_add_ps(velecsum,velec);
 359
 360             fscal            = felec;
 361
 362             /* Calculate temporary vectorial force */
 363             tx               = _mm256_mul_ps(fscal,dx13);
 364             ty               = _mm256_mul_ps(fscal,dy13);
 365             tz               = _mm256_mul_ps(fscal,dz13);
 366
 367             /* Update vectorial force */
 368             fix1             = _mm256_add_ps(fix1,tx);
 369             fiy1             = _mm256_add_ps(fiy1,ty);
 370             fiz1             = _mm256_add_ps(fiz1,tz);
 371
 372             fjx3             = _mm256_add_ps(fjx3,tx);
 373             fjy3             = _mm256_add_ps(fjy3,ty);
 374             fjz3             = _mm256_add_ps(fjz3,tz);
 375
 376             /**************************
 377              * CALCULATE INTERACTIONS *
 378              **************************/
 379
 380             /* COULOMB ELECTROSTATICS */
 381             velec            = _mm256_mul_ps(qq21,rinv21);
 382             felec            = _mm256_mul_ps(velec,rinvsq21);
 383
 384             /* Update potential sum for this i atom from the interaction with this j atom. */
 385             velecsum         = _mm256_add_ps(velecsum,velec);
 386
 387             fscal            = felec;
 388
 389             /* Calculate temporary vectorial force */
 390             tx               = _mm256_mul_ps(fscal,dx21);
 391             ty               = _mm256_mul_ps(fscal,dy21);
 392             tz               = _mm256_mul_ps(fscal,dz21);
 393
 394             /* Update vectorial force */
 395             fix2             = _mm256_add_ps(fix2,tx);
 396             fiy2             = _mm256_add_ps(fiy2,ty);
 397             fiz2             = _mm256_add_ps(fiz2,tz);
 398
 399             fjx1             = _mm256_add_ps(fjx1,tx);
 400             fjy1             = _mm256_add_ps(fjy1,ty);
 401             fjz1             = _mm256_add_ps(fjz1,tz);
 402
 403             /**************************
 404              * CALCULATE INTERACTIONS *
 405              **************************/
 406
 407             /* COULOMB ELECTROSTATICS */
 408             velec            = _mm256_mul_ps(qq22,rinv22);
 409             felec            = _mm256_mul_ps(velec,rinvsq22);
 410
 411             /* Update potential sum for this i atom from the interaction with this j atom. */
 412             velecsum         = _mm256_add_ps(velecsum,velec);
 413
 414             fscal            = felec;
 415
 416             /* Calculate temporary vectorial force */
 417             tx               = _mm256_mul_ps(fscal,dx22);
 418             ty               = _mm256_mul_ps(fscal,dy22);
 419             tz               = _mm256_mul_ps(fscal,dz22);
 420
 421             /* Update vectorial force */
 422             fix2             = _mm256_add_ps(fix2,tx);
 423             fiy2             = _mm256_add_ps(fiy2,ty);
 424             fiz2             = _mm256_add_ps(fiz2,tz);
 425
 426             fjx2             = _mm256_add_ps(fjx2,tx);
 427             fjy2             = _mm256_add_ps(fjy2,ty);
 428             fjz2             = _mm256_add_ps(fjz2,tz);
 429
 430             /**************************
 431              * CALCULATE INTERACTIONS *
 432              **************************/
 433
 434             /* COULOMB ELECTROSTATICS */
 435             velec            = _mm256_mul_ps(qq23,rinv23);
 436             felec            = _mm256_mul_ps(velec,rinvsq23);
 437
 438             /* Update potential sum for this i atom from the interaction with this j atom. */
 439             velecsum         = _mm256_add_ps(velecsum,velec);
 440
 441             fscal            = felec;
 442
 443             /* Calculate temporary vectorial force */
 444             tx               = _mm256_mul_ps(fscal,dx23);
 445             ty               = _mm256_mul_ps(fscal,dy23);
 446             tz               = _mm256_mul_ps(fscal,dz23);
 447
 448             /* Update vectorial force */
 449             fix2             = _mm256_add_ps(fix2,tx);
 450             fiy2             = _mm256_add_ps(fiy2,ty);
 451             fiz2             = _mm256_add_ps(fiz2,tz);
 452
 453             fjx3             = _mm256_add_ps(fjx3,tx);
 454             fjy3             = _mm256_add_ps(fjy3,ty);
 455             fjz3             = _mm256_add_ps(fjz3,tz);
 456
 457             /**************************
 458              * CALCULATE INTERACTIONS *
 459              **************************/
 460
 461             /* COULOMB ELECTROSTATICS */
 462             velec            = _mm256_mul_ps(qq31,rinv31);
 463             felec            = _mm256_mul_ps(velec,rinvsq31);
 464
 465             /* Update potential sum for this i atom from the interaction with this j atom. */
 466             velecsum         = _mm256_add_ps(velecsum,velec);
 467
 468             fscal            = felec;
 469
 470             /* Calculate temporary vectorial force */
 471             tx               = _mm256_mul_ps(fscal,dx31);
 472             ty               = _mm256_mul_ps(fscal,dy31);
 473             tz               = _mm256_mul_ps(fscal,dz31);
 474
 475             /* Update vectorial force */
 476             fix3             = _mm256_add_ps(fix3,tx);
 477             fiy3             = _mm256_add_ps(fiy3,ty);
 478             fiz3             = _mm256_add_ps(fiz3,tz);
 479
 480             fjx1             = _mm256_add_ps(fjx1,tx);
 481             fjy1             = _mm256_add_ps(fjy1,ty);
 482             fjz1             = _mm256_add_ps(fjz1,tz);
 483
 484             /**************************
 485              * CALCULATE INTERACTIONS *
 486              **************************/
 487
 488             /* COULOMB ELECTROSTATICS */
 489             velec            = _mm256_mul_ps(qq32,rinv32);
 490             felec            = _mm256_mul_ps(velec,rinvsq32);
 491
 492             /* Update potential sum for this i atom from the interaction with this j atom. */
 493             velecsum         = _mm256_add_ps(velecsum,velec);
 494
 495             fscal            = felec;
 496
 497             /* Calculate temporary vectorial force */
 498             tx               = _mm256_mul_ps(fscal,dx32);
 499             ty               = _mm256_mul_ps(fscal,dy32);
 500             tz               = _mm256_mul_ps(fscal,dz32);
 501
 502             /* Update vectorial force */
 503             fix3             = _mm256_add_ps(fix3,tx);
 504             fiy3             = _mm256_add_ps(fiy3,ty);
 505             fiz3             = _mm256_add_ps(fiz3,tz);
 506
 507             fjx2             = _mm256_add_ps(fjx2,tx);
 508             fjy2             = _mm256_add_ps(fjy2,ty);
 509             fjz2             = _mm256_add_ps(fjz2,tz);
 510
 511             /**************************
 512              * CALCULATE INTERACTIONS *
 513              **************************/
 514
 515             /* COULOMB ELECTROSTATICS */
 516             velec            = _mm256_mul_ps(qq33,rinv33);
 517             felec            = _mm256_mul_ps(velec,rinvsq33);
 518
 519             /* Update potential sum for this i atom from the interaction with this j atom. */
 520             velecsum         = _mm256_add_ps(velecsum,velec);
 521
 522             fscal            = felec;
 523
 524             /* Calculate temporary vectorial force */
 525             tx               = _mm256_mul_ps(fscal,dx33);
 526             ty               = _mm256_mul_ps(fscal,dy33);
 527             tz               = _mm256_mul_ps(fscal,dz33);
 528
 529             /* Update vectorial force */
 530             fix3             = _mm256_add_ps(fix3,tx);
 531             fiy3             = _mm256_add_ps(fiy3,ty);
 532             fiz3             = _mm256_add_ps(fiz3,tz);
 533
 534             fjx3             = _mm256_add_ps(fjx3,tx);
 535             fjy3             = _mm256_add_ps(fjy3,ty);
 536             fjz3             = _mm256_add_ps(fjz3,tz);
 537
 538             fjptrA             = f+j_coord_offsetA;
 539             fjptrB             = f+j_coord_offsetB;
 540             fjptrC             = f+j_coord_offsetC;
 541             fjptrD             = f+j_coord_offsetD;
 542             fjptrE             = f+j_coord_offsetE;
 543             fjptrF             = f+j_coord_offsetF;
 544             fjptrG             = f+j_coord_offsetG;
 545             fjptrH             = f+j_coord_offsetH;
 546
 547             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA+DIM,fjptrB+DIM,fjptrC+DIM,fjptrD+DIM,
 548                                                       fjptrE+DIM,fjptrF+DIM,fjptrG+DIM,fjptrH+DIM,
 549                                                       fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 550
 551             /* Inner loop uses 243 flops */
 552         }
 553
 554         if(jidx<j_index_end)
 555         {
 556
 557             /* Get j neighbor index, and coordinate index */
 558             jnrlistA         = jjnr[jidx];
 559             jnrlistB         = jjnr[jidx+1];
 560             jnrlistC         = jjnr[jidx+2];
 561             jnrlistD         = jjnr[jidx+3];
 562             jnrlistE         = jjnr[jidx+4];
 563             jnrlistF         = jjnr[jidx+5];
 564             jnrlistG         = jjnr[jidx+6];
 565             jnrlistH         = jjnr[jidx+7];
 566             /* Sign of each element will be negative for non-real atoms.
 567              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 568              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 569              */
 570             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
 571                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
 572
 573             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 574             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 575             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 576             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 577             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
 578             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
 579             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
 580             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
 581             j_coord_offsetA  = DIM*jnrA;
 582             j_coord_offsetB  = DIM*jnrB;
 583             j_coord_offsetC  = DIM*jnrC;
 584             j_coord_offsetD  = DIM*jnrD;
 585             j_coord_offsetE  = DIM*jnrE;
 586             j_coord_offsetF  = DIM*jnrF;
 587             j_coord_offsetG  = DIM*jnrG;
 588             j_coord_offsetH  = DIM*jnrH;
 589
 590             /* load j atom coordinates */
 591             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
 592                                                  x+j_coord_offsetC+DIM,x+j_coord_offsetD+DIM,
 593                                                  x+j_coord_offsetE+DIM,x+j_coord_offsetF+DIM,
 594                                                  x+j_coord_offsetG+DIM,x+j_coord_offsetH+DIM,
 595                                                  &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
 596
 597             /* Calculate displacement vector */
 598             dx11             = _mm256_sub_ps(ix1,jx1);
 599             dy11             = _mm256_sub_ps(iy1,jy1);
 600             dz11             = _mm256_sub_ps(iz1,jz1);
 601             dx12             = _mm256_sub_ps(ix1,jx2);
 602             dy12             = _mm256_sub_ps(iy1,jy2);
 603             dz12             = _mm256_sub_ps(iz1,jz2);
 604             dx13             = _mm256_sub_ps(ix1,jx3);
 605             dy13             = _mm256_sub_ps(iy1,jy3);
 606             dz13             = _mm256_sub_ps(iz1,jz3);
 607             dx21             = _mm256_sub_ps(ix2,jx1);
 608             dy21             = _mm256_sub_ps(iy2,jy1);
 609             dz21             = _mm256_sub_ps(iz2,jz1);
 610             dx22             = _mm256_sub_ps(ix2,jx2);
 611             dy22             = _mm256_sub_ps(iy2,jy2);
 612             dz22             = _mm256_sub_ps(iz2,jz2);
 613             dx23             = _mm256_sub_ps(ix2,jx3);
 614             dy23             = _mm256_sub_ps(iy2,jy3);
 615             dz23             = _mm256_sub_ps(iz2,jz3);
 616             dx31             = _mm256_sub_ps(ix3,jx1);
 617             dy31             = _mm256_sub_ps(iy3,jy1);
 618             dz31             = _mm256_sub_ps(iz3,jz1);
 619             dx32             = _mm256_sub_ps(ix3,jx2);
 620             dy32             = _mm256_sub_ps(iy3,jy2);
 621             dz32             = _mm256_sub_ps(iz3,jz2);
 622             dx33             = _mm256_sub_ps(ix3,jx3);
 623             dy33             = _mm256_sub_ps(iy3,jy3);
 624             dz33             = _mm256_sub_ps(iz3,jz3);
 625
 626             /* Calculate squared distance and things based on it */
 627             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 628             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 629             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
 630             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 631             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 632             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
 633             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
 634             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
 635             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
 636
 637             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 638             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 639             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
 640             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 641             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 642             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
 643             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
 644             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
 645             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
 646
 647             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 648             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 649             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
 650             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 651             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 652             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
 653             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
 654             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
 655             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
 656
 657             fjx1             = _mm256_setzero_ps();
 658             fjy1             = _mm256_setzero_ps();
 659             fjz1             = _mm256_setzero_ps();
 660             fjx2             = _mm256_setzero_ps();
 661             fjy2             = _mm256_setzero_ps();
 662             fjz2             = _mm256_setzero_ps();
 663             fjx3             = _mm256_setzero_ps();
 664             fjy3             = _mm256_setzero_ps();
 665             fjz3             = _mm256_setzero_ps();
 666
 667             /**************************
 668              * CALCULATE INTERACTIONS *
 669              **************************/
 670
 671             /* COULOMB ELECTROSTATICS */
 672             velec            = _mm256_mul_ps(qq11,rinv11);
 673             felec            = _mm256_mul_ps(velec,rinvsq11);
 674
 675             /* Update potential sum for this i atom from the interaction with this j atom. */
 676             velec            = _mm256_andnot_ps(dummy_mask,velec);
 677             velecsum         = _mm256_add_ps(velecsum,velec);
 678
 679             fscal            = felec;
 680
 681             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 682
 683             /* Calculate temporary vectorial force */
 684             tx               = _mm256_mul_ps(fscal,dx11);
 685             ty               = _mm256_mul_ps(fscal,dy11);
 686             tz               = _mm256_mul_ps(fscal,dz11);
 687
 688             /* Update vectorial force */
 689             fix1             = _mm256_add_ps(fix1,tx);
 690             fiy1             = _mm256_add_ps(fiy1,ty);
 691             fiz1             = _mm256_add_ps(fiz1,tz);
 692
 693             fjx1             = _mm256_add_ps(fjx1,tx);
 694             fjy1             = _mm256_add_ps(fjy1,ty);
 695             fjz1             = _mm256_add_ps(fjz1,tz);
 696
 697             /**************************
 698              * CALCULATE INTERACTIONS *
 699              **************************/
 700
 701             /* COULOMB ELECTROSTATICS */
 702             velec            = _mm256_mul_ps(qq12,rinv12);
 703             felec            = _mm256_mul_ps(velec,rinvsq12);
 704
 705             /* Update potential sum for this i atom from the interaction with this j atom. */
 706             velec            = _mm256_andnot_ps(dummy_mask,velec);
 707             velecsum         = _mm256_add_ps(velecsum,velec);
 708
 709             fscal            = felec;
 710
 711             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 712
 713             /* Calculate temporary vectorial force */
 714             tx               = _mm256_mul_ps(fscal,dx12);
 715             ty               = _mm256_mul_ps(fscal,dy12);
 716             tz               = _mm256_mul_ps(fscal,dz12);
 717
 718             /* Update vectorial force */
 719             fix1             = _mm256_add_ps(fix1,tx);
 720             fiy1             = _mm256_add_ps(fiy1,ty);
 721             fiz1             = _mm256_add_ps(fiz1,tz);
 722
 723             fjx2             = _mm256_add_ps(fjx2,tx);
 724             fjy2             = _mm256_add_ps(fjy2,ty);
 725             fjz2             = _mm256_add_ps(fjz2,tz);
 726
 727             /**************************
 728              * CALCULATE INTERACTIONS *
 729              **************************/
 730
 731             /* COULOMB ELECTROSTATICS */
 732             velec            = _mm256_mul_ps(qq13,rinv13);
 733             felec            = _mm256_mul_ps(velec,rinvsq13);
 734
 735             /* Update potential sum for this i atom from the interaction with this j atom. */
 736             velec            = _mm256_andnot_ps(dummy_mask,velec);
 737             velecsum         = _mm256_add_ps(velecsum,velec);
 738
 739             fscal            = felec;
 740
 741             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 742
 743             /* Calculate temporary vectorial force */
 744             tx               = _mm256_mul_ps(fscal,dx13);
 745             ty               = _mm256_mul_ps(fscal,dy13);
 746             tz               = _mm256_mul_ps(fscal,dz13);
 747
 748             /* Update vectorial force */
 749             fix1             = _mm256_add_ps(fix1,tx);
 750             fiy1             = _mm256_add_ps(fiy1,ty);
 751             fiz1             = _mm256_add_ps(fiz1,tz);
 752
 753             fjx3             = _mm256_add_ps(fjx3,tx);
 754             fjy3             = _mm256_add_ps(fjy3,ty);
 755             fjz3             = _mm256_add_ps(fjz3,tz);
 756
 757             /**************************
 758              * CALCULATE INTERACTIONS *
 759              **************************/
 760
 761             /* COULOMB ELECTROSTATICS */
 762             velec            = _mm256_mul_ps(qq21,rinv21);
 763             felec            = _mm256_mul_ps(velec,rinvsq21);
 764
 765             /* Update potential sum for this i atom from the interaction with this j atom. */
 766             velec            = _mm256_andnot_ps(dummy_mask,velec);
 767             velecsum         = _mm256_add_ps(velecsum,velec);
 768
 769             fscal            = felec;
 770
 771             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 772
 773             /* Calculate temporary vectorial force */
 774             tx               = _mm256_mul_ps(fscal,dx21);
 775             ty               = _mm256_mul_ps(fscal,dy21);
 776             tz               = _mm256_mul_ps(fscal,dz21);
 777
 778             /* Update vectorial force */
 779             fix2             = _mm256_add_ps(fix2,tx);
 780             fiy2             = _mm256_add_ps(fiy2,ty);
 781             fiz2             = _mm256_add_ps(fiz2,tz);
 782
 783             fjx1             = _mm256_add_ps(fjx1,tx);
 784             fjy1             = _mm256_add_ps(fjy1,ty);
 785             fjz1             = _mm256_add_ps(fjz1,tz);
 786
 787             /**************************
 788              * CALCULATE INTERACTIONS *
 789              **************************/
 790
 791             /* COULOMB ELECTROSTATICS */
 792             velec            = _mm256_mul_ps(qq22,rinv22);
 793             felec            = _mm256_mul_ps(velec,rinvsq22);
 794
 795             /* Update potential sum for this i atom from the interaction with this j atom. */
 796             velec            = _mm256_andnot_ps(dummy_mask,velec);
 797             velecsum         = _mm256_add_ps(velecsum,velec);
 798
 799             fscal            = felec;
 800
 801             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 802
 803             /* Calculate temporary vectorial force */
 804             tx               = _mm256_mul_ps(fscal,dx22);
 805             ty               = _mm256_mul_ps(fscal,dy22);
 806             tz               = _mm256_mul_ps(fscal,dz22);
 807
 808             /* Update vectorial force */
 809             fix2             = _mm256_add_ps(fix2,tx);
 810             fiy2             = _mm256_add_ps(fiy2,ty);
 811             fiz2             = _mm256_add_ps(fiz2,tz);
 812
 813             fjx2             = _mm256_add_ps(fjx2,tx);
 814             fjy2             = _mm256_add_ps(fjy2,ty);
 815             fjz2             = _mm256_add_ps(fjz2,tz);
 816
 817             /**************************
 818              * CALCULATE INTERACTIONS *
 819              **************************/
 820
 821             /* COULOMB ELECTROSTATICS */
 822             velec            = _mm256_mul_ps(qq23,rinv23);
 823             felec            = _mm256_mul_ps(velec,rinvsq23);
 824
 825             /* Update potential sum for this i atom from the interaction with this j atom. */
 826             velec            = _mm256_andnot_ps(dummy_mask,velec);
 827             velecsum         = _mm256_add_ps(velecsum,velec);
 828
 829             fscal            = felec;
 830
 831             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 832
 833             /* Calculate temporary vectorial force */
 834             tx               = _mm256_mul_ps(fscal,dx23);
 835             ty               = _mm256_mul_ps(fscal,dy23);
 836             tz               = _mm256_mul_ps(fscal,dz23);
 837
 838             /* Update vectorial force */
 839             fix2             = _mm256_add_ps(fix2,tx);
 840             fiy2             = _mm256_add_ps(fiy2,ty);
 841             fiz2             = _mm256_add_ps(fiz2,tz);
 842
 843             fjx3             = _mm256_add_ps(fjx3,tx);
 844             fjy3             = _mm256_add_ps(fjy3,ty);
 845             fjz3             = _mm256_add_ps(fjz3,tz);
 846
 847             /**************************
 848              * CALCULATE INTERACTIONS *
 849              **************************/
 850
 851             /* COULOMB ELECTROSTATICS */
 852             velec            = _mm256_mul_ps(qq31,rinv31);
 853             felec            = _mm256_mul_ps(velec,rinvsq31);
 854
 855             /* Update potential sum for this i atom from the interaction with this j atom. */
 856             velec            = _mm256_andnot_ps(dummy_mask,velec);
 857             velecsum         = _mm256_add_ps(velecsum,velec);
 858
 859             fscal            = felec;
 860
 861             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 862
 863             /* Calculate temporary vectorial force */
 864             tx               = _mm256_mul_ps(fscal,dx31);
 865             ty               = _mm256_mul_ps(fscal,dy31);
 866             tz               = _mm256_mul_ps(fscal,dz31);
 867
 868             /* Update vectorial force */
 869             fix3             = _mm256_add_ps(fix3,tx);
 870             fiy3             = _mm256_add_ps(fiy3,ty);
 871             fiz3             = _mm256_add_ps(fiz3,tz);
 872
 873             fjx1             = _mm256_add_ps(fjx1,tx);
 874             fjy1             = _mm256_add_ps(fjy1,ty);
 875             fjz1             = _mm256_add_ps(fjz1,tz);
 876
 877             /**************************
 878              * CALCULATE INTERACTIONS *
 879              **************************/
 880
 881             /* COULOMB ELECTROSTATICS */
 882             velec            = _mm256_mul_ps(qq32,rinv32);
 883             felec            = _mm256_mul_ps(velec,rinvsq32);
 884
 885             /* Update potential sum for this i atom from the interaction with this j atom. */
 886             velec            = _mm256_andnot_ps(dummy_mask,velec);
 887             velecsum         = _mm256_add_ps(velecsum,velec);
 888
 889             fscal            = felec;
 890
 891             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 892
 893             /* Calculate temporary vectorial force */
 894             tx               = _mm256_mul_ps(fscal,dx32);
 895             ty               = _mm256_mul_ps(fscal,dy32);
 896             tz               = _mm256_mul_ps(fscal,dz32);
 897
 898             /* Update vectorial force */
 899             fix3             = _mm256_add_ps(fix3,tx);
 900             fiy3             = _mm256_add_ps(fiy3,ty);
 901             fiz3             = _mm256_add_ps(fiz3,tz);
 902
 903             fjx2             = _mm256_add_ps(fjx2,tx);
 904             fjy2             = _mm256_add_ps(fjy2,ty);
 905             fjz2             = _mm256_add_ps(fjz2,tz);
 906
 907             /**************************
 908              * CALCULATE INTERACTIONS *
 909              **************************/
 910
 911             /* COULOMB ELECTROSTATICS */
 912             velec            = _mm256_mul_ps(qq33,rinv33);
 913             felec            = _mm256_mul_ps(velec,rinvsq33);
 914
 915             /* Update potential sum for this i atom from the interaction with this j atom. */
 916             velec            = _mm256_andnot_ps(dummy_mask,velec);
 917             velecsum         = _mm256_add_ps(velecsum,velec);
 918
 919             fscal            = felec;
 920
 921             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 922
 923             /* Calculate temporary vectorial force */
 924             tx               = _mm256_mul_ps(fscal,dx33);
 925             ty               = _mm256_mul_ps(fscal,dy33);
 926             tz               = _mm256_mul_ps(fscal,dz33);
 927
 928             /* Update vectorial force */
 929             fix3             = _mm256_add_ps(fix3,tx);
 930             fiy3             = _mm256_add_ps(fiy3,ty);
 931             fiz3             = _mm256_add_ps(fiz3,tz);
 932
 933             fjx3             = _mm256_add_ps(fjx3,tx);
 934             fjy3             = _mm256_add_ps(fjy3,ty);
 935             fjz3             = _mm256_add_ps(fjz3,tz);
 936
 937             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
 938             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
 939             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
 940             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
 941             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
 942             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
 943             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
 944             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
 945
 946             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA+DIM,fjptrB+DIM,fjptrC+DIM,fjptrD+DIM,
 947                                                       fjptrE+DIM,fjptrF+DIM,fjptrG+DIM,fjptrH+DIM,
 948                                                       fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 949
 950             /* Inner loop uses 243 flops */
 951         }
 952
 953         /* End of innermost loop */
 954
 955         gmx_mm256_update_iforce_3atom_swizzle_ps(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 956                                                  f+i_coord_offset+DIM,fshift+i_shift_offset);
 957
 958         ggid                        = gid[iidx];
 959         /* Update potential energies */
 960         gmx_mm256_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
 961
 962         /* Increment number of inner iterations */
 963         inneriter                  += j_index_end - j_index_start;
 964
 965         /* Outer loop uses 19 flops */
 966     }
 967
 968     /* Increment number of outer iterations */
 969     outeriter        += nri;
 970
 971     /* Update outer/inner flops */
 972
 973     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4W4_VF,outeriter*19 + inneriter*243);
 974 }
 975 /*
 976  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwNone_GeomW4W4_F_avx_256_single
 977  * Electrostatics interaction: Coulomb
 978  * VdW interaction:            None
 979  * Geometry:                   Water4-Water4
 980  * Calculate force/pot:        Force
 981  */
 982 void
 983 nb_kernel_ElecCoul_VdwNone_GeomW4W4_F_avx_256_single
 984                     (t_nblist                    * gmx_restrict       nlist,
 985                      rvec                        * gmx_restrict          xx,
 986                      rvec                        * gmx_restrict          ff,
 987                      t_forcerec                  * gmx_restrict          fr,
 988                      t_mdatoms                   * gmx_restrict     mdatoms,
 989                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 990                      t_nrnb                      * gmx_restrict        nrnb)
 991 {
 992     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 993      * just 0 for non-waters.
 994      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
 995      * jnr indices corresponding to data put in the four positions in the SIMD register.
 996      */
 997     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 998     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 999     int              jnrA,jnrB,jnrC,jnrD;
1000     int              jnrE,jnrF,jnrG,jnrH;
1001     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1002     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1003     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1004     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
1005     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1006     real             rcutoff_scalar;
1007     real             *shiftvec,*fshift,*x,*f;
1008     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
1009     real             scratch[4*DIM];
1010     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1011     real *           vdwioffsetptr1;
1012     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1013     real *           vdwioffsetptr2;
1014     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1015     real *           vdwioffsetptr3;
1016     __m256           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1017     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
1018     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1019     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
1020     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1021     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D,vdwjidx3E,vdwjidx3F,vdwjidx3G,vdwjidx3H;
1022     __m256           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1023     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1024     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1025     __m256           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1026     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1027     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1028     __m256           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1029     __m256           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1030     __m256           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1031     __m256           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1032     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
1033     real             *charge;
1034     __m256           dummy_mask,cutoff_mask;
1035     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
1036     __m256           one     = _mm256_set1_ps(1.0);
1037     __m256           two     = _mm256_set1_ps(2.0);
1038     x                = xx[0];
1039     f                = ff[0];
1040
1041     nri              = nlist->nri;
1042     iinr             = nlist->iinr;
1043     jindex           = nlist->jindex;
1044     jjnr             = nlist->jjnr;
1045     shiftidx         = nlist->shift;
1046     gid              = nlist->gid;
1047     shiftvec         = fr->shift_vec[0];
1048     fshift           = fr->fshift[0];
1049     facel            = _mm256_set1_ps(fr->epsfac);
1050     charge           = mdatoms->chargeA;
1051
1052     /* Setup water-specific parameters */
1053     inr              = nlist->iinr[0];
1054     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
1055     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
1056     iq3              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+3]));
1057
1058     jq1              = _mm256_set1_ps(charge[inr+1]);
1059     jq2              = _mm256_set1_ps(charge[inr+2]);
1060     jq3              = _mm256_set1_ps(charge[inr+3]);
1061     qq11             = _mm256_mul_ps(iq1,jq1);
1062     qq12             = _mm256_mul_ps(iq1,jq2);
1063     qq13             = _mm256_mul_ps(iq1,jq3);
1064     qq21             = _mm256_mul_ps(iq2,jq1);
1065     qq22             = _mm256_mul_ps(iq2,jq2);
1066     qq23             = _mm256_mul_ps(iq2,jq3);
1067     qq31             = _mm256_mul_ps(iq3,jq1);
1068     qq32             = _mm256_mul_ps(iq3,jq2);
1069     qq33             = _mm256_mul_ps(iq3,jq3);
1070
1071     /* Avoid stupid compiler warnings */
1072     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
1073     j_coord_offsetA = 0;
1074     j_coord_offsetB = 0;
1075     j_coord_offsetC = 0;
1076     j_coord_offsetD = 0;
1077     j_coord_offsetE = 0;
1078     j_coord_offsetF = 0;
1079     j_coord_offsetG = 0;
1080     j_coord_offsetH = 0;
1081
1082     outeriter        = 0;
1083     inneriter        = 0;
1084
1085     for(iidx=0;iidx<4*DIM;iidx++)
1086     {
1087         scratch[iidx] = 0.0;
1088     }
1089
1090     /* Start outer loop over neighborlists */
1091     for(iidx=0; iidx<nri; iidx++)
1092     {
1093         /* Load shift vector for this list */
1094         i_shift_offset   = DIM*shiftidx[iidx];
1095
1096         /* Load limits for loop over neighbors */
1097         j_index_start    = jindex[iidx];
1098         j_index_end      = jindex[iidx+1];
1099
1100         /* Get outer coordinate index */
1101         inr              = iinr[iidx];
1102         i_coord_offset   = DIM*inr;
1103
1104         /* Load i particle coords and add shift vector */
1105         gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
1106                                                     &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1107
1108         fix1             = _mm256_setzero_ps();
1109         fiy1             = _mm256_setzero_ps();
1110         fiz1             = _mm256_setzero_ps();
1111         fix2             = _mm256_setzero_ps();
1112         fiy2             = _mm256_setzero_ps();
1113         fiz2             = _mm256_setzero_ps();
1114         fix3             = _mm256_setzero_ps();
1115         fiy3             = _mm256_setzero_ps();
1116         fiz3             = _mm256_setzero_ps();
1117
1118         /* Start inner kernel loop */
1119         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
1120         {
1121
1122             /* Get j neighbor index, and coordinate index */
1123             jnrA             = jjnr[jidx];
1124             jnrB             = jjnr[jidx+1];
1125             jnrC             = jjnr[jidx+2];
1126             jnrD             = jjnr[jidx+3];
1127             jnrE             = jjnr[jidx+4];
1128             jnrF             = jjnr[jidx+5];
1129             jnrG             = jjnr[jidx+6];
1130             jnrH             = jjnr[jidx+7];
1131             j_coord_offsetA  = DIM*jnrA;
1132             j_coord_offsetB  = DIM*jnrB;
1133             j_coord_offsetC  = DIM*jnrC;
1134             j_coord_offsetD  = DIM*jnrD;
1135             j_coord_offsetE  = DIM*jnrE;
1136             j_coord_offsetF  = DIM*jnrF;
1137             j_coord_offsetG  = DIM*jnrG;
1138             j_coord_offsetH  = DIM*jnrH;
1139
1140             /* load j atom coordinates */
1141             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
1142                                                  x+j_coord_offsetC+DIM,x+j_coord_offsetD+DIM,
1143                                                  x+j_coord_offsetE+DIM,x+j_coord_offsetF+DIM,
1144                                                  x+j_coord_offsetG+DIM,x+j_coord_offsetH+DIM,
1145                                                  &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
1146
1147             /* Calculate displacement vector */
1148             dx11             = _mm256_sub_ps(ix1,jx1);
1149             dy11             = _mm256_sub_ps(iy1,jy1);
1150             dz11             = _mm256_sub_ps(iz1,jz1);
1151             dx12             = _mm256_sub_ps(ix1,jx2);
1152             dy12             = _mm256_sub_ps(iy1,jy2);
1153             dz12             = _mm256_sub_ps(iz1,jz2);
1154             dx13             = _mm256_sub_ps(ix1,jx3);
1155             dy13             = _mm256_sub_ps(iy1,jy3);
1156             dz13             = _mm256_sub_ps(iz1,jz3);
1157             dx21             = _mm256_sub_ps(ix2,jx1);
1158             dy21             = _mm256_sub_ps(iy2,jy1);
1159             dz21             = _mm256_sub_ps(iz2,jz1);
1160             dx22             = _mm256_sub_ps(ix2,jx2);
1161             dy22             = _mm256_sub_ps(iy2,jy2);
1162             dz22             = _mm256_sub_ps(iz2,jz2);
1163             dx23             = _mm256_sub_ps(ix2,jx3);
1164             dy23             = _mm256_sub_ps(iy2,jy3);
1165             dz23             = _mm256_sub_ps(iz2,jz3);
1166             dx31             = _mm256_sub_ps(ix3,jx1);
1167             dy31             = _mm256_sub_ps(iy3,jy1);
1168             dz31             = _mm256_sub_ps(iz3,jz1);
1169             dx32             = _mm256_sub_ps(ix3,jx2);
1170             dy32             = _mm256_sub_ps(iy3,jy2);
1171             dz32             = _mm256_sub_ps(iz3,jz2);
1172             dx33             = _mm256_sub_ps(ix3,jx3);
1173             dy33             = _mm256_sub_ps(iy3,jy3);
1174             dz33             = _mm256_sub_ps(iz3,jz3);
1175
1176             /* Calculate squared distance and things based on it */
1177             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1178             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1179             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
1180             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1181             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1182             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
1183             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
1184             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
1185             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
1186
1187             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1188             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1189             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
1190             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1191             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1192             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
1193             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
1194             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
1195             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
1196
1197             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1198             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1199             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
1200             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1201             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1202             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
1203             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
1204             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
1205             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
1206
1207             fjx1             = _mm256_setzero_ps();
1208             fjy1             = _mm256_setzero_ps();
1209             fjz1             = _mm256_setzero_ps();
1210             fjx2             = _mm256_setzero_ps();
1211             fjy2             = _mm256_setzero_ps();
1212             fjz2             = _mm256_setzero_ps();
1213             fjx3             = _mm256_setzero_ps();
1214             fjy3             = _mm256_setzero_ps();
1215             fjz3             = _mm256_setzero_ps();
1216
1217             /**************************
1218              * CALCULATE INTERACTIONS *
1219              **************************/
1220
1221             /* COULOMB ELECTROSTATICS */
1222             velec            = _mm256_mul_ps(qq11,rinv11);
1223             felec            = _mm256_mul_ps(velec,rinvsq11);
1224
1225             fscal            = felec;
1226
1227             /* Calculate temporary vectorial force */
1228             tx               = _mm256_mul_ps(fscal,dx11);
1229             ty               = _mm256_mul_ps(fscal,dy11);
1230             tz               = _mm256_mul_ps(fscal,dz11);
1231
1232             /* Update vectorial force */
1233             fix1             = _mm256_add_ps(fix1,tx);
1234             fiy1             = _mm256_add_ps(fiy1,ty);
1235             fiz1             = _mm256_add_ps(fiz1,tz);
1236
1237             fjx1             = _mm256_add_ps(fjx1,tx);
1238             fjy1             = _mm256_add_ps(fjy1,ty);
1239             fjz1             = _mm256_add_ps(fjz1,tz);
1240
1241             /**************************
1242              * CALCULATE INTERACTIONS *
1243              **************************/
1244
1245             /* COULOMB ELECTROSTATICS */
1246             velec            = _mm256_mul_ps(qq12,rinv12);
1247             felec            = _mm256_mul_ps(velec,rinvsq12);
1248
1249             fscal            = felec;
1250
1251             /* Calculate temporary vectorial force */
1252             tx               = _mm256_mul_ps(fscal,dx12);
1253             ty               = _mm256_mul_ps(fscal,dy12);
1254             tz               = _mm256_mul_ps(fscal,dz12);
1255
1256             /* Update vectorial force */
1257             fix1             = _mm256_add_ps(fix1,tx);
1258             fiy1             = _mm256_add_ps(fiy1,ty);
1259             fiz1             = _mm256_add_ps(fiz1,tz);
1260
1261             fjx2             = _mm256_add_ps(fjx2,tx);
1262             fjy2             = _mm256_add_ps(fjy2,ty);
1263             fjz2             = _mm256_add_ps(fjz2,tz);
1264
1265             /**************************
1266              * CALCULATE INTERACTIONS *
1267              **************************/
1268
1269             /* COULOMB ELECTROSTATICS */
1270             velec            = _mm256_mul_ps(qq13,rinv13);
1271             felec            = _mm256_mul_ps(velec,rinvsq13);
1272
1273             fscal            = felec;
1274
1275             /* Calculate temporary vectorial force */
1276             tx               = _mm256_mul_ps(fscal,dx13);
1277             ty               = _mm256_mul_ps(fscal,dy13);
1278             tz               = _mm256_mul_ps(fscal,dz13);
1279
1280             /* Update vectorial force */
1281             fix1             = _mm256_add_ps(fix1,tx);
1282             fiy1             = _mm256_add_ps(fiy1,ty);
1283             fiz1             = _mm256_add_ps(fiz1,tz);
1284
1285             fjx3             = _mm256_add_ps(fjx3,tx);
1286             fjy3             = _mm256_add_ps(fjy3,ty);
1287             fjz3             = _mm256_add_ps(fjz3,tz);
1288
1289             /**************************
1290              * CALCULATE INTERACTIONS *
1291              **************************/
1292
1293             /* COULOMB ELECTROSTATICS */
1294             velec            = _mm256_mul_ps(qq21,rinv21);
1295             felec            = _mm256_mul_ps(velec,rinvsq21);
1296
1297             fscal            = felec;
1298
1299             /* Calculate temporary vectorial force */
1300             tx               = _mm256_mul_ps(fscal,dx21);
1301             ty               = _mm256_mul_ps(fscal,dy21);
1302             tz               = _mm256_mul_ps(fscal,dz21);
1303
1304             /* Update vectorial force */
1305             fix2             = _mm256_add_ps(fix2,tx);
1306             fiy2             = _mm256_add_ps(fiy2,ty);
1307             fiz2             = _mm256_add_ps(fiz2,tz);
1308
1309             fjx1             = _mm256_add_ps(fjx1,tx);
1310             fjy1             = _mm256_add_ps(fjy1,ty);
1311             fjz1             = _mm256_add_ps(fjz1,tz);
1312
1313             /**************************
1314              * CALCULATE INTERACTIONS *
1315              **************************/
1316
1317             /* COULOMB ELECTROSTATICS */
1318             velec            = _mm256_mul_ps(qq22,rinv22);
1319             felec            = _mm256_mul_ps(velec,rinvsq22);
1320
1321             fscal            = felec;
1322
1323             /* Calculate temporary vectorial force */
1324             tx               = _mm256_mul_ps(fscal,dx22);
1325             ty               = _mm256_mul_ps(fscal,dy22);
1326             tz               = _mm256_mul_ps(fscal,dz22);
1327
1328             /* Update vectorial force */
1329             fix2             = _mm256_add_ps(fix2,tx);
1330             fiy2             = _mm256_add_ps(fiy2,ty);
1331             fiz2             = _mm256_add_ps(fiz2,tz);
1332
1333             fjx2             = _mm256_add_ps(fjx2,tx);
1334             fjy2             = _mm256_add_ps(fjy2,ty);
1335             fjz2             = _mm256_add_ps(fjz2,tz);
1336
1337             /**************************
1338              * CALCULATE INTERACTIONS *
1339              **************************/
1340
1341             /* COULOMB ELECTROSTATICS */
1342             velec            = _mm256_mul_ps(qq23,rinv23);
1343             felec            = _mm256_mul_ps(velec,rinvsq23);
1344
1345             fscal            = felec;
1346
1347             /* Calculate temporary vectorial force */
1348             tx               = _mm256_mul_ps(fscal,dx23);
1349             ty               = _mm256_mul_ps(fscal,dy23);
1350             tz               = _mm256_mul_ps(fscal,dz23);
1351
1352             /* Update vectorial force */
1353             fix2             = _mm256_add_ps(fix2,tx);
1354             fiy2             = _mm256_add_ps(fiy2,ty);
1355             fiz2             = _mm256_add_ps(fiz2,tz);
1356
1357             fjx3             = _mm256_add_ps(fjx3,tx);
1358             fjy3             = _mm256_add_ps(fjy3,ty);
1359             fjz3             = _mm256_add_ps(fjz3,tz);
1360
1361             /**************************
1362              * CALCULATE INTERACTIONS *
1363              **************************/
1364
1365             /* COULOMB ELECTROSTATICS */
1366             velec            = _mm256_mul_ps(qq31,rinv31);
1367             felec            = _mm256_mul_ps(velec,rinvsq31);
1368
1369             fscal            = felec;
1370
1371             /* Calculate temporary vectorial force */
1372             tx               = _mm256_mul_ps(fscal,dx31);
1373             ty               = _mm256_mul_ps(fscal,dy31);
1374             tz               = _mm256_mul_ps(fscal,dz31);
1375
1376             /* Update vectorial force */
1377             fix3             = _mm256_add_ps(fix3,tx);
1378             fiy3             = _mm256_add_ps(fiy3,ty);
1379             fiz3             = _mm256_add_ps(fiz3,tz);
1380
1381             fjx1             = _mm256_add_ps(fjx1,tx);
1382             fjy1             = _mm256_add_ps(fjy1,ty);
1383             fjz1             = _mm256_add_ps(fjz1,tz);
1384
1385             /**************************
1386              * CALCULATE INTERACTIONS *
1387              **************************/
1388
1389             /* COULOMB ELECTROSTATICS */
1390             velec            = _mm256_mul_ps(qq32,rinv32);
1391             felec            = _mm256_mul_ps(velec,rinvsq32);
1392
1393             fscal            = felec;
1394
1395             /* Calculate temporary vectorial force */
1396             tx               = _mm256_mul_ps(fscal,dx32);
1397             ty               = _mm256_mul_ps(fscal,dy32);
1398             tz               = _mm256_mul_ps(fscal,dz32);
1399
1400             /* Update vectorial force */
1401             fix3             = _mm256_add_ps(fix3,tx);
1402             fiy3             = _mm256_add_ps(fiy3,ty);
1403             fiz3             = _mm256_add_ps(fiz3,tz);
1404
1405             fjx2             = _mm256_add_ps(fjx2,tx);
1406             fjy2             = _mm256_add_ps(fjy2,ty);
1407             fjz2             = _mm256_add_ps(fjz2,tz);
1408
1409             /**************************
1410              * CALCULATE INTERACTIONS *
1411              **************************/
1412
1413             /* COULOMB ELECTROSTATICS */
1414             velec            = _mm256_mul_ps(qq33,rinv33);
1415             felec            = _mm256_mul_ps(velec,rinvsq33);
1416
1417             fscal            = felec;
1418
1419             /* Calculate temporary vectorial force */
1420             tx               = _mm256_mul_ps(fscal,dx33);
1421             ty               = _mm256_mul_ps(fscal,dy33);
1422             tz               = _mm256_mul_ps(fscal,dz33);
1423
1424             /* Update vectorial force */
1425             fix3             = _mm256_add_ps(fix3,tx);
1426             fiy3             = _mm256_add_ps(fiy3,ty);
1427             fiz3             = _mm256_add_ps(fiz3,tz);
1428
1429             fjx3             = _mm256_add_ps(fjx3,tx);
1430             fjy3             = _mm256_add_ps(fjy3,ty);
1431             fjz3             = _mm256_add_ps(fjz3,tz);
1432
1433             fjptrA             = f+j_coord_offsetA;
1434             fjptrB             = f+j_coord_offsetB;
1435             fjptrC             = f+j_coord_offsetC;
1436             fjptrD             = f+j_coord_offsetD;
1437             fjptrE             = f+j_coord_offsetE;
1438             fjptrF             = f+j_coord_offsetF;
1439             fjptrG             = f+j_coord_offsetG;
1440             fjptrH             = f+j_coord_offsetH;
1441
1442             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA+DIM,fjptrB+DIM,fjptrC+DIM,fjptrD+DIM,
1443                                                       fjptrE+DIM,fjptrF+DIM,fjptrG+DIM,fjptrH+DIM,
1444                                                       fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1445
1446             /* Inner loop uses 234 flops */
1447         }
1448
1449         if(jidx<j_index_end)
1450         {
1451
1452             /* Get j neighbor index, and coordinate index */
1453             jnrlistA         = jjnr[jidx];
1454             jnrlistB         = jjnr[jidx+1];
1455             jnrlistC         = jjnr[jidx+2];
1456             jnrlistD         = jjnr[jidx+3];
1457             jnrlistE         = jjnr[jidx+4];
1458             jnrlistF         = jjnr[jidx+5];
1459             jnrlistG         = jjnr[jidx+6];
1460             jnrlistH         = jjnr[jidx+7];
1461             /* Sign of each element will be negative for non-real atoms.
1462              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1463              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1464              */
1465             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
1466                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
1467
1468             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1469             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1470             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1471             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1472             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
1473             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
1474             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
1475             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
1476             j_coord_offsetA  = DIM*jnrA;
1477             j_coord_offsetB  = DIM*jnrB;
1478             j_coord_offsetC  = DIM*jnrC;
1479             j_coord_offsetD  = DIM*jnrD;
1480             j_coord_offsetE  = DIM*jnrE;
1481             j_coord_offsetF  = DIM*jnrF;
1482             j_coord_offsetG  = DIM*jnrG;
1483             j_coord_offsetH  = DIM*jnrH;
1484
1485             /* load j atom coordinates */
1486             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
1487                                                  x+j_coord_offsetC+DIM,x+j_coord_offsetD+DIM,
1488                                                  x+j_coord_offsetE+DIM,x+j_coord_offsetF+DIM,
1489                                                  x+j_coord_offsetG+DIM,x+j_coord_offsetH+DIM,
1490                                                  &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
1491
1492             /* Calculate displacement vector */
1493             dx11             = _mm256_sub_ps(ix1,jx1);
1494             dy11             = _mm256_sub_ps(iy1,jy1);
1495             dz11             = _mm256_sub_ps(iz1,jz1);
1496             dx12             = _mm256_sub_ps(ix1,jx2);
1497             dy12             = _mm256_sub_ps(iy1,jy2);
1498             dz12             = _mm256_sub_ps(iz1,jz2);
1499             dx13             = _mm256_sub_ps(ix1,jx3);
1500             dy13             = _mm256_sub_ps(iy1,jy3);
1501             dz13             = _mm256_sub_ps(iz1,jz3);
1502             dx21             = _mm256_sub_ps(ix2,jx1);
1503             dy21             = _mm256_sub_ps(iy2,jy1);
1504             dz21             = _mm256_sub_ps(iz2,jz1);
1505             dx22             = _mm256_sub_ps(ix2,jx2);
1506             dy22             = _mm256_sub_ps(iy2,jy2);
1507             dz22             = _mm256_sub_ps(iz2,jz2);
1508             dx23             = _mm256_sub_ps(ix2,jx3);
1509             dy23             = _mm256_sub_ps(iy2,jy3);
1510             dz23             = _mm256_sub_ps(iz2,jz3);
1511             dx31             = _mm256_sub_ps(ix3,jx1);
1512             dy31             = _mm256_sub_ps(iy3,jy1);
1513             dz31             = _mm256_sub_ps(iz3,jz1);
1514             dx32             = _mm256_sub_ps(ix3,jx2);
1515             dy32             = _mm256_sub_ps(iy3,jy2);
1516             dz32             = _mm256_sub_ps(iz3,jz2);
1517             dx33             = _mm256_sub_ps(ix3,jx3);
1518             dy33             = _mm256_sub_ps(iy3,jy3);
1519             dz33             = _mm256_sub_ps(iz3,jz3);
1520
1521             /* Calculate squared distance and things based on it */
1522             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1523             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1524             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
1525             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1526             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1527             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
1528             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
1529             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
1530             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
1531
1532             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1533             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1534             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
1535             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1536             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1537             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
1538             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
1539             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
1540             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
1541
1542             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1543             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1544             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
1545             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1546             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1547             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
1548             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
1549             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
1550             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
1551
1552             fjx1             = _mm256_setzero_ps();
1553             fjy1             = _mm256_setzero_ps();
1554             fjz1             = _mm256_setzero_ps();
1555             fjx2             = _mm256_setzero_ps();
1556             fjy2             = _mm256_setzero_ps();
1557             fjz2             = _mm256_setzero_ps();
1558             fjx3             = _mm256_setzero_ps();
1559             fjy3             = _mm256_setzero_ps();
1560             fjz3             = _mm256_setzero_ps();
1561
1562             /**************************
1563              * CALCULATE INTERACTIONS *
1564              **************************/
1565
1566             /* COULOMB ELECTROSTATICS */
1567             velec            = _mm256_mul_ps(qq11,rinv11);
1568             felec            = _mm256_mul_ps(velec,rinvsq11);
1569
1570             fscal            = felec;
1571
1572             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1573
1574             /* Calculate temporary vectorial force */
1575             tx               = _mm256_mul_ps(fscal,dx11);
1576             ty               = _mm256_mul_ps(fscal,dy11);
1577             tz               = _mm256_mul_ps(fscal,dz11);
1578
1579             /* Update vectorial force */
1580             fix1             = _mm256_add_ps(fix1,tx);
1581             fiy1             = _mm256_add_ps(fiy1,ty);
1582             fiz1             = _mm256_add_ps(fiz1,tz);
1583
1584             fjx1             = _mm256_add_ps(fjx1,tx);
1585             fjy1             = _mm256_add_ps(fjy1,ty);
1586             fjz1             = _mm256_add_ps(fjz1,tz);
1587
1588             /**************************
1589              * CALCULATE INTERACTIONS *
1590              **************************/
1591
1592             /* COULOMB ELECTROSTATICS */
1593             velec            = _mm256_mul_ps(qq12,rinv12);
1594             felec            = _mm256_mul_ps(velec,rinvsq12);
1595
1596             fscal            = felec;
1597
1598             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1599
1600             /* Calculate temporary vectorial force */
1601             tx               = _mm256_mul_ps(fscal,dx12);
1602             ty               = _mm256_mul_ps(fscal,dy12);
1603             tz               = _mm256_mul_ps(fscal,dz12);
1604
1605             /* Update vectorial force */
1606             fix1             = _mm256_add_ps(fix1,tx);
1607             fiy1             = _mm256_add_ps(fiy1,ty);
1608             fiz1             = _mm256_add_ps(fiz1,tz);
1609
1610             fjx2             = _mm256_add_ps(fjx2,tx);
1611             fjy2             = _mm256_add_ps(fjy2,ty);
1612             fjz2             = _mm256_add_ps(fjz2,tz);
1613
1614             /**************************
1615              * CALCULATE INTERACTIONS *
1616              **************************/
1617
1618             /* COULOMB ELECTROSTATICS */
1619             velec            = _mm256_mul_ps(qq13,rinv13);
1620             felec            = _mm256_mul_ps(velec,rinvsq13);
1621
1622             fscal            = felec;
1623
1624             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1625
1626             /* Calculate temporary vectorial force */
1627             tx               = _mm256_mul_ps(fscal,dx13);
1628             ty               = _mm256_mul_ps(fscal,dy13);
1629             tz               = _mm256_mul_ps(fscal,dz13);
1630
1631             /* Update vectorial force */
1632             fix1             = _mm256_add_ps(fix1,tx);
1633             fiy1             = _mm256_add_ps(fiy1,ty);
1634             fiz1             = _mm256_add_ps(fiz1,tz);
1635
1636             fjx3             = _mm256_add_ps(fjx3,tx);
1637             fjy3             = _mm256_add_ps(fjy3,ty);
1638             fjz3             = _mm256_add_ps(fjz3,tz);
1639
1640             /**************************
1641              * CALCULATE INTERACTIONS *
1642              **************************/
1643
1644             /* COULOMB ELECTROSTATICS */
1645             velec            = _mm256_mul_ps(qq21,rinv21);
1646             felec            = _mm256_mul_ps(velec,rinvsq21);
1647
1648             fscal            = felec;
1649
1650             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1651
1652             /* Calculate temporary vectorial force */
1653             tx               = _mm256_mul_ps(fscal,dx21);
1654             ty               = _mm256_mul_ps(fscal,dy21);
1655             tz               = _mm256_mul_ps(fscal,dz21);
1656
1657             /* Update vectorial force */
1658             fix2             = _mm256_add_ps(fix2,tx);
1659             fiy2             = _mm256_add_ps(fiy2,ty);
1660             fiz2             = _mm256_add_ps(fiz2,tz);
1661
1662             fjx1             = _mm256_add_ps(fjx1,tx);
1663             fjy1             = _mm256_add_ps(fjy1,ty);
1664             fjz1             = _mm256_add_ps(fjz1,tz);
1665
1666             /**************************
1667              * CALCULATE INTERACTIONS *
1668              **************************/
1669
1670             /* COULOMB ELECTROSTATICS */
1671             velec            = _mm256_mul_ps(qq22,rinv22);
1672             felec            = _mm256_mul_ps(velec,rinvsq22);
1673
1674             fscal            = felec;
1675
1676             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1677
1678             /* Calculate temporary vectorial force */
1679             tx               = _mm256_mul_ps(fscal,dx22);
1680             ty               = _mm256_mul_ps(fscal,dy22);
1681             tz               = _mm256_mul_ps(fscal,dz22);
1682
1683             /* Update vectorial force */
1684             fix2             = _mm256_add_ps(fix2,tx);
1685             fiy2             = _mm256_add_ps(fiy2,ty);
1686             fiz2             = _mm256_add_ps(fiz2,tz);
1687
1688             fjx2             = _mm256_add_ps(fjx2,tx);
1689             fjy2             = _mm256_add_ps(fjy2,ty);
1690             fjz2             = _mm256_add_ps(fjz2,tz);
1691
1692             /**************************
1693              * CALCULATE INTERACTIONS *
1694              **************************/
1695
1696             /* COULOMB ELECTROSTATICS */
1697             velec            = _mm256_mul_ps(qq23,rinv23);
1698             felec            = _mm256_mul_ps(velec,rinvsq23);
1699
1700             fscal            = felec;
1701
1702             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1703
1704             /* Calculate temporary vectorial force */
1705             tx               = _mm256_mul_ps(fscal,dx23);
1706             ty               = _mm256_mul_ps(fscal,dy23);
1707             tz               = _mm256_mul_ps(fscal,dz23);
1708
1709             /* Update vectorial force */
1710             fix2             = _mm256_add_ps(fix2,tx);
1711             fiy2             = _mm256_add_ps(fiy2,ty);
1712             fiz2             = _mm256_add_ps(fiz2,tz);
1713
1714             fjx3             = _mm256_add_ps(fjx3,tx);
1715             fjy3             = _mm256_add_ps(fjy3,ty);
1716             fjz3             = _mm256_add_ps(fjz3,tz);
1717
1718             /**************************
1719              * CALCULATE INTERACTIONS *
1720              **************************/
1721
1722             /* COULOMB ELECTROSTATICS */
1723             velec            = _mm256_mul_ps(qq31,rinv31);
1724             felec            = _mm256_mul_ps(velec,rinvsq31);
1725
1726             fscal            = felec;
1727
1728             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1729
1730             /* Calculate temporary vectorial force */
1731             tx               = _mm256_mul_ps(fscal,dx31);
1732             ty               = _mm256_mul_ps(fscal,dy31);
1733             tz               = _mm256_mul_ps(fscal,dz31);
1734
1735             /* Update vectorial force */
1736             fix3             = _mm256_add_ps(fix3,tx);
1737             fiy3             = _mm256_add_ps(fiy3,ty);
1738             fiz3             = _mm256_add_ps(fiz3,tz);
1739
1740             fjx1             = _mm256_add_ps(fjx1,tx);
1741             fjy1             = _mm256_add_ps(fjy1,ty);
1742             fjz1             = _mm256_add_ps(fjz1,tz);
1743
1744             /**************************
1745              * CALCULATE INTERACTIONS *
1746              **************************/
1747
1748             /* COULOMB ELECTROSTATICS */
1749             velec            = _mm256_mul_ps(qq32,rinv32);
1750             felec            = _mm256_mul_ps(velec,rinvsq32);
1751
1752             fscal            = felec;
1753
1754             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1755
1756             /* Calculate temporary vectorial force */
1757             tx               = _mm256_mul_ps(fscal,dx32);
1758             ty               = _mm256_mul_ps(fscal,dy32);
1759             tz               = _mm256_mul_ps(fscal,dz32);
1760
1761             /* Update vectorial force */
1762             fix3             = _mm256_add_ps(fix3,tx);
1763             fiy3             = _mm256_add_ps(fiy3,ty);
1764             fiz3             = _mm256_add_ps(fiz3,tz);
1765
1766             fjx2             = _mm256_add_ps(fjx2,tx);
1767             fjy2             = _mm256_add_ps(fjy2,ty);
1768             fjz2             = _mm256_add_ps(fjz2,tz);
1769
1770             /**************************
1771              * CALCULATE INTERACTIONS *
1772              **************************/
1773
1774             /* COULOMB ELECTROSTATICS */
1775             velec            = _mm256_mul_ps(qq33,rinv33);
1776             felec            = _mm256_mul_ps(velec,rinvsq33);
1777
1778             fscal            = felec;
1779
1780             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1781
1782             /* Calculate temporary vectorial force */
1783             tx               = _mm256_mul_ps(fscal,dx33);
1784             ty               = _mm256_mul_ps(fscal,dy33);
1785             tz               = _mm256_mul_ps(fscal,dz33);
1786
1787             /* Update vectorial force */
1788             fix3             = _mm256_add_ps(fix3,tx);
1789             fiy3             = _mm256_add_ps(fiy3,ty);
1790             fiz3             = _mm256_add_ps(fiz3,tz);
1791
1792             fjx3             = _mm256_add_ps(fjx3,tx);
1793             fjy3             = _mm256_add_ps(fjy3,ty);
1794             fjz3             = _mm256_add_ps(fjz3,tz);
1795
1796             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1797             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1798             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1799             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1800             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1801             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1802             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
1803             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
1804
1805             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA+DIM,fjptrB+DIM,fjptrC+DIM,fjptrD+DIM,
1806                                                       fjptrE+DIM,fjptrF+DIM,fjptrG+DIM,fjptrH+DIM,
1807                                                       fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1808
1809             /* Inner loop uses 234 flops */
1810         }
1811
1812         /* End of innermost loop */
1813
1814         gmx_mm256_update_iforce_3atom_swizzle_ps(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1815                                                  f+i_coord_offset+DIM,fshift+i_shift_offset);
1816
1817         /* Increment number of inner iterations */
1818         inneriter                  += j_index_end - j_index_start;
1819
1820         /* Outer loop uses 18 flops */
1821     }
1822
1823     /* Increment number of outer iterations */
1824     outeriter        += nri;
1825
1826     /* Update outer/inner flops */
1827
1828     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4W4_F,outeriter*18 + inneriter*234);
1829 }