src/gromacs/gmxlib/nonbonded/nb_kernel_avx_256_single/nb_kernel_ElecCoul_VdwNone_GeomW3W3_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_GeomW3W3_VF_avx_256_single
  52  * Electrostatics interaction: Coulomb
  53  * VdW interaction:            None
  54  * Geometry:                   Water3-Water3
  55  * Calculate force/pot:        PotentialAndForce
  56  */
  57 void
  58 nb_kernel_ElecCoul_VdwNone_GeomW3W3_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 *           vdwioffsetptr0;
  87     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  88     real *           vdwioffsetptr1;
  89     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  90     real *           vdwioffsetptr2;
  91     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  92     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
  93     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  94     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
  95     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  96     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
  97     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  98     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  99     __m256           dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
 100     __m256           dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
 101     __m256           dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 102     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
 103     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
 104     __m256           dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 105     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 106     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 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     iq0              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+0]));
 130     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
 131     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
 132
 133     jq0              = _mm256_set1_ps(charge[inr+0]);
 134     jq1              = _mm256_set1_ps(charge[inr+1]);
 135     jq2              = _mm256_set1_ps(charge[inr+2]);
 136     qq00             = _mm256_mul_ps(iq0,jq0);
 137     qq01             = _mm256_mul_ps(iq0,jq1);
 138     qq02             = _mm256_mul_ps(iq0,jq2);
 139     qq10             = _mm256_mul_ps(iq1,jq0);
 140     qq11             = _mm256_mul_ps(iq1,jq1);
 141     qq12             = _mm256_mul_ps(iq1,jq2);
 142     qq20             = _mm256_mul_ps(iq2,jq0);
 143     qq21             = _mm256_mul_ps(iq2,jq1);
 144     qq22             = _mm256_mul_ps(iq2,jq2);
 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,
 181                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 182
 183         fix0             = _mm256_setzero_ps();
 184         fiy0             = _mm256_setzero_ps();
 185         fiz0             = _mm256_setzero_ps();
 186         fix1             = _mm256_setzero_ps();
 187         fiy1             = _mm256_setzero_ps();
 188         fiz1             = _mm256_setzero_ps();
 189         fix2             = _mm256_setzero_ps();
 190         fiy2             = _mm256_setzero_ps();
 191         fiz2             = _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,x+j_coord_offsetB,
 220                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 221                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 222                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 223                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 224
 225             /* Calculate displacement vector */
 226             dx00             = _mm256_sub_ps(ix0,jx0);
 227             dy00             = _mm256_sub_ps(iy0,jy0);
 228             dz00             = _mm256_sub_ps(iz0,jz0);
 229             dx01             = _mm256_sub_ps(ix0,jx1);
 230             dy01             = _mm256_sub_ps(iy0,jy1);
 231             dz01             = _mm256_sub_ps(iz0,jz1);
 232             dx02             = _mm256_sub_ps(ix0,jx2);
 233             dy02             = _mm256_sub_ps(iy0,jy2);
 234             dz02             = _mm256_sub_ps(iz0,jz2);
 235             dx10             = _mm256_sub_ps(ix1,jx0);
 236             dy10             = _mm256_sub_ps(iy1,jy0);
 237             dz10             = _mm256_sub_ps(iz1,jz0);
 238             dx11             = _mm256_sub_ps(ix1,jx1);
 239             dy11             = _mm256_sub_ps(iy1,jy1);
 240             dz11             = _mm256_sub_ps(iz1,jz1);
 241             dx12             = _mm256_sub_ps(ix1,jx2);
 242             dy12             = _mm256_sub_ps(iy1,jy2);
 243             dz12             = _mm256_sub_ps(iz1,jz2);
 244             dx20             = _mm256_sub_ps(ix2,jx0);
 245             dy20             = _mm256_sub_ps(iy2,jy0);
 246             dz20             = _mm256_sub_ps(iz2,jz0);
 247             dx21             = _mm256_sub_ps(ix2,jx1);
 248             dy21             = _mm256_sub_ps(iy2,jy1);
 249             dz21             = _mm256_sub_ps(iz2,jz1);
 250             dx22             = _mm256_sub_ps(ix2,jx2);
 251             dy22             = _mm256_sub_ps(iy2,jy2);
 252             dz22             = _mm256_sub_ps(iz2,jz2);
 253
 254             /* Calculate squared distance and things based on it */
 255             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 256             rsq01            = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
 257             rsq02            = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
 258             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
 259             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 260             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 261             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
 262             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 263             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 264
 265             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
 266             rinv01           = gmx_mm256_invsqrt_ps(rsq01);
 267             rinv02           = gmx_mm256_invsqrt_ps(rsq02);
 268             rinv10           = gmx_mm256_invsqrt_ps(rsq10);
 269             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 270             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 271             rinv20           = gmx_mm256_invsqrt_ps(rsq20);
 272             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 273             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 274
 275             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
 276             rinvsq01         = _mm256_mul_ps(rinv01,rinv01);
 277             rinvsq02         = _mm256_mul_ps(rinv02,rinv02);
 278             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
 279             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 280             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 281             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
 282             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 283             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 284
 285             fjx0             = _mm256_setzero_ps();
 286             fjy0             = _mm256_setzero_ps();
 287             fjz0             = _mm256_setzero_ps();
 288             fjx1             = _mm256_setzero_ps();
 289             fjy1             = _mm256_setzero_ps();
 290             fjz1             = _mm256_setzero_ps();
 291             fjx2             = _mm256_setzero_ps();
 292             fjy2             = _mm256_setzero_ps();
 293             fjz2             = _mm256_setzero_ps();
 294
 295             /**************************
 296              * CALCULATE INTERACTIONS *
 297              **************************/
 298
 299             /* COULOMB ELECTROSTATICS */
 300             velec            = _mm256_mul_ps(qq00,rinv00);
 301             felec            = _mm256_mul_ps(velec,rinvsq00);
 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,dx00);
 310             ty               = _mm256_mul_ps(fscal,dy00);
 311             tz               = _mm256_mul_ps(fscal,dz00);
 312
 313             /* Update vectorial force */
 314             fix0             = _mm256_add_ps(fix0,tx);
 315             fiy0             = _mm256_add_ps(fiy0,ty);
 316             fiz0             = _mm256_add_ps(fiz0,tz);
 317
 318             fjx0             = _mm256_add_ps(fjx0,tx);
 319             fjy0             = _mm256_add_ps(fjy0,ty);
 320             fjz0             = _mm256_add_ps(fjz0,tz);
 321
 322             /**************************
 323              * CALCULATE INTERACTIONS *
 324              **************************/
 325
 326             /* COULOMB ELECTROSTATICS */
 327             velec            = _mm256_mul_ps(qq01,rinv01);
 328             felec            = _mm256_mul_ps(velec,rinvsq01);
 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,dx01);
 337             ty               = _mm256_mul_ps(fscal,dy01);
 338             tz               = _mm256_mul_ps(fscal,dz01);
 339
 340             /* Update vectorial force */
 341             fix0             = _mm256_add_ps(fix0,tx);
 342             fiy0             = _mm256_add_ps(fiy0,ty);
 343             fiz0             = _mm256_add_ps(fiz0,tz);
 344
 345             fjx1             = _mm256_add_ps(fjx1,tx);
 346             fjy1             = _mm256_add_ps(fjy1,ty);
 347             fjz1             = _mm256_add_ps(fjz1,tz);
 348
 349             /**************************
 350              * CALCULATE INTERACTIONS *
 351              **************************/
 352
 353             /* COULOMB ELECTROSTATICS */
 354             velec            = _mm256_mul_ps(qq02,rinv02);
 355             felec            = _mm256_mul_ps(velec,rinvsq02);
 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,dx02);
 364             ty               = _mm256_mul_ps(fscal,dy02);
 365             tz               = _mm256_mul_ps(fscal,dz02);
 366
 367             /* Update vectorial force */
 368             fix0             = _mm256_add_ps(fix0,tx);
 369             fiy0             = _mm256_add_ps(fiy0,ty);
 370             fiz0             = _mm256_add_ps(fiz0,tz);
 371
 372             fjx2             = _mm256_add_ps(fjx2,tx);
 373             fjy2             = _mm256_add_ps(fjy2,ty);
 374             fjz2             = _mm256_add_ps(fjz2,tz);
 375
 376             /**************************
 377              * CALCULATE INTERACTIONS *
 378              **************************/
 379
 380             /* COULOMB ELECTROSTATICS */
 381             velec            = _mm256_mul_ps(qq10,rinv10);
 382             felec            = _mm256_mul_ps(velec,rinvsq10);
 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,dx10);
 391             ty               = _mm256_mul_ps(fscal,dy10);
 392             tz               = _mm256_mul_ps(fscal,dz10);
 393
 394             /* Update vectorial force */
 395             fix1             = _mm256_add_ps(fix1,tx);
 396             fiy1             = _mm256_add_ps(fiy1,ty);
 397             fiz1             = _mm256_add_ps(fiz1,tz);
 398
 399             fjx0             = _mm256_add_ps(fjx0,tx);
 400             fjy0             = _mm256_add_ps(fjy0,ty);
 401             fjz0             = _mm256_add_ps(fjz0,tz);
 402
 403             /**************************
 404              * CALCULATE INTERACTIONS *
 405              **************************/
 406
 407             /* COULOMB ELECTROSTATICS */
 408             velec            = _mm256_mul_ps(qq11,rinv11);
 409             felec            = _mm256_mul_ps(velec,rinvsq11);
 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,dx11);
 418             ty               = _mm256_mul_ps(fscal,dy11);
 419             tz               = _mm256_mul_ps(fscal,dz11);
 420
 421             /* Update vectorial force */
 422             fix1             = _mm256_add_ps(fix1,tx);
 423             fiy1             = _mm256_add_ps(fiy1,ty);
 424             fiz1             = _mm256_add_ps(fiz1,tz);
 425
 426             fjx1             = _mm256_add_ps(fjx1,tx);
 427             fjy1             = _mm256_add_ps(fjy1,ty);
 428             fjz1             = _mm256_add_ps(fjz1,tz);
 429
 430             /**************************
 431              * CALCULATE INTERACTIONS *
 432              **************************/
 433
 434             /* COULOMB ELECTROSTATICS */
 435             velec            = _mm256_mul_ps(qq12,rinv12);
 436             felec            = _mm256_mul_ps(velec,rinvsq12);
 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,dx12);
 445             ty               = _mm256_mul_ps(fscal,dy12);
 446             tz               = _mm256_mul_ps(fscal,dz12);
 447
 448             /* Update vectorial force */
 449             fix1             = _mm256_add_ps(fix1,tx);
 450             fiy1             = _mm256_add_ps(fiy1,ty);
 451             fiz1             = _mm256_add_ps(fiz1,tz);
 452
 453             fjx2             = _mm256_add_ps(fjx2,tx);
 454             fjy2             = _mm256_add_ps(fjy2,ty);
 455             fjz2             = _mm256_add_ps(fjz2,tz);
 456
 457             /**************************
 458              * CALCULATE INTERACTIONS *
 459              **************************/
 460
 461             /* COULOMB ELECTROSTATICS */
 462             velec            = _mm256_mul_ps(qq20,rinv20);
 463             felec            = _mm256_mul_ps(velec,rinvsq20);
 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,dx20);
 472             ty               = _mm256_mul_ps(fscal,dy20);
 473             tz               = _mm256_mul_ps(fscal,dz20);
 474
 475             /* Update vectorial force */
 476             fix2             = _mm256_add_ps(fix2,tx);
 477             fiy2             = _mm256_add_ps(fiy2,ty);
 478             fiz2             = _mm256_add_ps(fiz2,tz);
 479
 480             fjx0             = _mm256_add_ps(fjx0,tx);
 481             fjy0             = _mm256_add_ps(fjy0,ty);
 482             fjz0             = _mm256_add_ps(fjz0,tz);
 483
 484             /**************************
 485              * CALCULATE INTERACTIONS *
 486              **************************/
 487
 488             /* COULOMB ELECTROSTATICS */
 489             velec            = _mm256_mul_ps(qq21,rinv21);
 490             felec            = _mm256_mul_ps(velec,rinvsq21);
 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,dx21);
 499             ty               = _mm256_mul_ps(fscal,dy21);
 500             tz               = _mm256_mul_ps(fscal,dz21);
 501
 502             /* Update vectorial force */
 503             fix2             = _mm256_add_ps(fix2,tx);
 504             fiy2             = _mm256_add_ps(fiy2,ty);
 505             fiz2             = _mm256_add_ps(fiz2,tz);
 506
 507             fjx1             = _mm256_add_ps(fjx1,tx);
 508             fjy1             = _mm256_add_ps(fjy1,ty);
 509             fjz1             = _mm256_add_ps(fjz1,tz);
 510
 511             /**************************
 512              * CALCULATE INTERACTIONS *
 513              **************************/
 514
 515             /* COULOMB ELECTROSTATICS */
 516             velec            = _mm256_mul_ps(qq22,rinv22);
 517             felec            = _mm256_mul_ps(velec,rinvsq22);
 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,dx22);
 526             ty               = _mm256_mul_ps(fscal,dy22);
 527             tz               = _mm256_mul_ps(fscal,dz22);
 528
 529             /* Update vectorial force */
 530             fix2             = _mm256_add_ps(fix2,tx);
 531             fiy2             = _mm256_add_ps(fiy2,ty);
 532             fiz2             = _mm256_add_ps(fiz2,tz);
 533
 534             fjx2             = _mm256_add_ps(fjx2,tx);
 535             fjy2             = _mm256_add_ps(fjy2,ty);
 536             fjz2             = _mm256_add_ps(fjz2,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,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
 548                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 549
 550             /* Inner loop uses 243 flops */
 551         }
 552
 553         if(jidx<j_index_end)
 554         {
 555
 556             /* Get j neighbor index, and coordinate index */
 557             jnrlistA         = jjnr[jidx];
 558             jnrlistB         = jjnr[jidx+1];
 559             jnrlistC         = jjnr[jidx+2];
 560             jnrlistD         = jjnr[jidx+3];
 561             jnrlistE         = jjnr[jidx+4];
 562             jnrlistF         = jjnr[jidx+5];
 563             jnrlistG         = jjnr[jidx+6];
 564             jnrlistH         = jjnr[jidx+7];
 565             /* Sign of each element will be negative for non-real atoms.
 566              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 567              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 568              */
 569             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
 570                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
 571
 572             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 573             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 574             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 575             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 576             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
 577             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
 578             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
 579             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
 580             j_coord_offsetA  = DIM*jnrA;
 581             j_coord_offsetB  = DIM*jnrB;
 582             j_coord_offsetC  = DIM*jnrC;
 583             j_coord_offsetD  = DIM*jnrD;
 584             j_coord_offsetE  = DIM*jnrE;
 585             j_coord_offsetF  = DIM*jnrF;
 586             j_coord_offsetG  = DIM*jnrG;
 587             j_coord_offsetH  = DIM*jnrH;
 588
 589             /* load j atom coordinates */
 590             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 591                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 592                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 593                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 594                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 595
 596             /* Calculate displacement vector */
 597             dx00             = _mm256_sub_ps(ix0,jx0);
 598             dy00             = _mm256_sub_ps(iy0,jy0);
 599             dz00             = _mm256_sub_ps(iz0,jz0);
 600             dx01             = _mm256_sub_ps(ix0,jx1);
 601             dy01             = _mm256_sub_ps(iy0,jy1);
 602             dz01             = _mm256_sub_ps(iz0,jz1);
 603             dx02             = _mm256_sub_ps(ix0,jx2);
 604             dy02             = _mm256_sub_ps(iy0,jy2);
 605             dz02             = _mm256_sub_ps(iz0,jz2);
 606             dx10             = _mm256_sub_ps(ix1,jx0);
 607             dy10             = _mm256_sub_ps(iy1,jy0);
 608             dz10             = _mm256_sub_ps(iz1,jz0);
 609             dx11             = _mm256_sub_ps(ix1,jx1);
 610             dy11             = _mm256_sub_ps(iy1,jy1);
 611             dz11             = _mm256_sub_ps(iz1,jz1);
 612             dx12             = _mm256_sub_ps(ix1,jx2);
 613             dy12             = _mm256_sub_ps(iy1,jy2);
 614             dz12             = _mm256_sub_ps(iz1,jz2);
 615             dx20             = _mm256_sub_ps(ix2,jx0);
 616             dy20             = _mm256_sub_ps(iy2,jy0);
 617             dz20             = _mm256_sub_ps(iz2,jz0);
 618             dx21             = _mm256_sub_ps(ix2,jx1);
 619             dy21             = _mm256_sub_ps(iy2,jy1);
 620             dz21             = _mm256_sub_ps(iz2,jz1);
 621             dx22             = _mm256_sub_ps(ix2,jx2);
 622             dy22             = _mm256_sub_ps(iy2,jy2);
 623             dz22             = _mm256_sub_ps(iz2,jz2);
 624
 625             /* Calculate squared distance and things based on it */
 626             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 627             rsq01            = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
 628             rsq02            = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
 629             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
 630             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 631             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 632             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
 633             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 634             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 635
 636             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
 637             rinv01           = gmx_mm256_invsqrt_ps(rsq01);
 638             rinv02           = gmx_mm256_invsqrt_ps(rsq02);
 639             rinv10           = gmx_mm256_invsqrt_ps(rsq10);
 640             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 641             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 642             rinv20           = gmx_mm256_invsqrt_ps(rsq20);
 643             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 644             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 645
 646             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
 647             rinvsq01         = _mm256_mul_ps(rinv01,rinv01);
 648             rinvsq02         = _mm256_mul_ps(rinv02,rinv02);
 649             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
 650             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 651             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 652             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
 653             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 654             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 655
 656             fjx0             = _mm256_setzero_ps();
 657             fjy0             = _mm256_setzero_ps();
 658             fjz0             = _mm256_setzero_ps();
 659             fjx1             = _mm256_setzero_ps();
 660             fjy1             = _mm256_setzero_ps();
 661             fjz1             = _mm256_setzero_ps();
 662             fjx2             = _mm256_setzero_ps();
 663             fjy2             = _mm256_setzero_ps();
 664             fjz2             = _mm256_setzero_ps();
 665
 666             /**************************
 667              * CALCULATE INTERACTIONS *
 668              **************************/
 669
 670             /* COULOMB ELECTROSTATICS */
 671             velec            = _mm256_mul_ps(qq00,rinv00);
 672             felec            = _mm256_mul_ps(velec,rinvsq00);
 673
 674             /* Update potential sum for this i atom from the interaction with this j atom. */
 675             velec            = _mm256_andnot_ps(dummy_mask,velec);
 676             velecsum         = _mm256_add_ps(velecsum,velec);
 677
 678             fscal            = felec;
 679
 680             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 681
 682             /* Calculate temporary vectorial force */
 683             tx               = _mm256_mul_ps(fscal,dx00);
 684             ty               = _mm256_mul_ps(fscal,dy00);
 685             tz               = _mm256_mul_ps(fscal,dz00);
 686
 687             /* Update vectorial force */
 688             fix0             = _mm256_add_ps(fix0,tx);
 689             fiy0             = _mm256_add_ps(fiy0,ty);
 690             fiz0             = _mm256_add_ps(fiz0,tz);
 691
 692             fjx0             = _mm256_add_ps(fjx0,tx);
 693             fjy0             = _mm256_add_ps(fjy0,ty);
 694             fjz0             = _mm256_add_ps(fjz0,tz);
 695
 696             /**************************
 697              * CALCULATE INTERACTIONS *
 698              **************************/
 699
 700             /* COULOMB ELECTROSTATICS */
 701             velec            = _mm256_mul_ps(qq01,rinv01);
 702             felec            = _mm256_mul_ps(velec,rinvsq01);
 703
 704             /* Update potential sum for this i atom from the interaction with this j atom. */
 705             velec            = _mm256_andnot_ps(dummy_mask,velec);
 706             velecsum         = _mm256_add_ps(velecsum,velec);
 707
 708             fscal            = felec;
 709
 710             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 711
 712             /* Calculate temporary vectorial force */
 713             tx               = _mm256_mul_ps(fscal,dx01);
 714             ty               = _mm256_mul_ps(fscal,dy01);
 715             tz               = _mm256_mul_ps(fscal,dz01);
 716
 717             /* Update vectorial force */
 718             fix0             = _mm256_add_ps(fix0,tx);
 719             fiy0             = _mm256_add_ps(fiy0,ty);
 720             fiz0             = _mm256_add_ps(fiz0,tz);
 721
 722             fjx1             = _mm256_add_ps(fjx1,tx);
 723             fjy1             = _mm256_add_ps(fjy1,ty);
 724             fjz1             = _mm256_add_ps(fjz1,tz);
 725
 726             /**************************
 727              * CALCULATE INTERACTIONS *
 728              **************************/
 729
 730             /* COULOMB ELECTROSTATICS */
 731             velec            = _mm256_mul_ps(qq02,rinv02);
 732             felec            = _mm256_mul_ps(velec,rinvsq02);
 733
 734             /* Update potential sum for this i atom from the interaction with this j atom. */
 735             velec            = _mm256_andnot_ps(dummy_mask,velec);
 736             velecsum         = _mm256_add_ps(velecsum,velec);
 737
 738             fscal            = felec;
 739
 740             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 741
 742             /* Calculate temporary vectorial force */
 743             tx               = _mm256_mul_ps(fscal,dx02);
 744             ty               = _mm256_mul_ps(fscal,dy02);
 745             tz               = _mm256_mul_ps(fscal,dz02);
 746
 747             /* Update vectorial force */
 748             fix0             = _mm256_add_ps(fix0,tx);
 749             fiy0             = _mm256_add_ps(fiy0,ty);
 750             fiz0             = _mm256_add_ps(fiz0,tz);
 751
 752             fjx2             = _mm256_add_ps(fjx2,tx);
 753             fjy2             = _mm256_add_ps(fjy2,ty);
 754             fjz2             = _mm256_add_ps(fjz2,tz);
 755
 756             /**************************
 757              * CALCULATE INTERACTIONS *
 758              **************************/
 759
 760             /* COULOMB ELECTROSTATICS */
 761             velec            = _mm256_mul_ps(qq10,rinv10);
 762             felec            = _mm256_mul_ps(velec,rinvsq10);
 763
 764             /* Update potential sum for this i atom from the interaction with this j atom. */
 765             velec            = _mm256_andnot_ps(dummy_mask,velec);
 766             velecsum         = _mm256_add_ps(velecsum,velec);
 767
 768             fscal            = felec;
 769
 770             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 771
 772             /* Calculate temporary vectorial force */
 773             tx               = _mm256_mul_ps(fscal,dx10);
 774             ty               = _mm256_mul_ps(fscal,dy10);
 775             tz               = _mm256_mul_ps(fscal,dz10);
 776
 777             /* Update vectorial force */
 778             fix1             = _mm256_add_ps(fix1,tx);
 779             fiy1             = _mm256_add_ps(fiy1,ty);
 780             fiz1             = _mm256_add_ps(fiz1,tz);
 781
 782             fjx0             = _mm256_add_ps(fjx0,tx);
 783             fjy0             = _mm256_add_ps(fjy0,ty);
 784             fjz0             = _mm256_add_ps(fjz0,tz);
 785
 786             /**************************
 787              * CALCULATE INTERACTIONS *
 788              **************************/
 789
 790             /* COULOMB ELECTROSTATICS */
 791             velec            = _mm256_mul_ps(qq11,rinv11);
 792             felec            = _mm256_mul_ps(velec,rinvsq11);
 793
 794             /* Update potential sum for this i atom from the interaction with this j atom. */
 795             velec            = _mm256_andnot_ps(dummy_mask,velec);
 796             velecsum         = _mm256_add_ps(velecsum,velec);
 797
 798             fscal            = felec;
 799
 800             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 801
 802             /* Calculate temporary vectorial force */
 803             tx               = _mm256_mul_ps(fscal,dx11);
 804             ty               = _mm256_mul_ps(fscal,dy11);
 805             tz               = _mm256_mul_ps(fscal,dz11);
 806
 807             /* Update vectorial force */
 808             fix1             = _mm256_add_ps(fix1,tx);
 809             fiy1             = _mm256_add_ps(fiy1,ty);
 810             fiz1             = _mm256_add_ps(fiz1,tz);
 811
 812             fjx1             = _mm256_add_ps(fjx1,tx);
 813             fjy1             = _mm256_add_ps(fjy1,ty);
 814             fjz1             = _mm256_add_ps(fjz1,tz);
 815
 816             /**************************
 817              * CALCULATE INTERACTIONS *
 818              **************************/
 819
 820             /* COULOMB ELECTROSTATICS */
 821             velec            = _mm256_mul_ps(qq12,rinv12);
 822             felec            = _mm256_mul_ps(velec,rinvsq12);
 823
 824             /* Update potential sum for this i atom from the interaction with this j atom. */
 825             velec            = _mm256_andnot_ps(dummy_mask,velec);
 826             velecsum         = _mm256_add_ps(velecsum,velec);
 827
 828             fscal            = felec;
 829
 830             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 831
 832             /* Calculate temporary vectorial force */
 833             tx               = _mm256_mul_ps(fscal,dx12);
 834             ty               = _mm256_mul_ps(fscal,dy12);
 835             tz               = _mm256_mul_ps(fscal,dz12);
 836
 837             /* Update vectorial force */
 838             fix1             = _mm256_add_ps(fix1,tx);
 839             fiy1             = _mm256_add_ps(fiy1,ty);
 840             fiz1             = _mm256_add_ps(fiz1,tz);
 841
 842             fjx2             = _mm256_add_ps(fjx2,tx);
 843             fjy2             = _mm256_add_ps(fjy2,ty);
 844             fjz2             = _mm256_add_ps(fjz2,tz);
 845
 846             /**************************
 847              * CALCULATE INTERACTIONS *
 848              **************************/
 849
 850             /* COULOMB ELECTROSTATICS */
 851             velec            = _mm256_mul_ps(qq20,rinv20);
 852             felec            = _mm256_mul_ps(velec,rinvsq20);
 853
 854             /* Update potential sum for this i atom from the interaction with this j atom. */
 855             velec            = _mm256_andnot_ps(dummy_mask,velec);
 856             velecsum         = _mm256_add_ps(velecsum,velec);
 857
 858             fscal            = felec;
 859
 860             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 861
 862             /* Calculate temporary vectorial force */
 863             tx               = _mm256_mul_ps(fscal,dx20);
 864             ty               = _mm256_mul_ps(fscal,dy20);
 865             tz               = _mm256_mul_ps(fscal,dz20);
 866
 867             /* Update vectorial force */
 868             fix2             = _mm256_add_ps(fix2,tx);
 869             fiy2             = _mm256_add_ps(fiy2,ty);
 870             fiz2             = _mm256_add_ps(fiz2,tz);
 871
 872             fjx0             = _mm256_add_ps(fjx0,tx);
 873             fjy0             = _mm256_add_ps(fjy0,ty);
 874             fjz0             = _mm256_add_ps(fjz0,tz);
 875
 876             /**************************
 877              * CALCULATE INTERACTIONS *
 878              **************************/
 879
 880             /* COULOMB ELECTROSTATICS */
 881             velec            = _mm256_mul_ps(qq21,rinv21);
 882             felec            = _mm256_mul_ps(velec,rinvsq21);
 883
 884             /* Update potential sum for this i atom from the interaction with this j atom. */
 885             velec            = _mm256_andnot_ps(dummy_mask,velec);
 886             velecsum         = _mm256_add_ps(velecsum,velec);
 887
 888             fscal            = felec;
 889
 890             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 891
 892             /* Calculate temporary vectorial force */
 893             tx               = _mm256_mul_ps(fscal,dx21);
 894             ty               = _mm256_mul_ps(fscal,dy21);
 895             tz               = _mm256_mul_ps(fscal,dz21);
 896
 897             /* Update vectorial force */
 898             fix2             = _mm256_add_ps(fix2,tx);
 899             fiy2             = _mm256_add_ps(fiy2,ty);
 900             fiz2             = _mm256_add_ps(fiz2,tz);
 901
 902             fjx1             = _mm256_add_ps(fjx1,tx);
 903             fjy1             = _mm256_add_ps(fjy1,ty);
 904             fjz1             = _mm256_add_ps(fjz1,tz);
 905
 906             /**************************
 907              * CALCULATE INTERACTIONS *
 908              **************************/
 909
 910             /* COULOMB ELECTROSTATICS */
 911             velec            = _mm256_mul_ps(qq22,rinv22);
 912             felec            = _mm256_mul_ps(velec,rinvsq22);
 913
 914             /* Update potential sum for this i atom from the interaction with this j atom. */
 915             velec            = _mm256_andnot_ps(dummy_mask,velec);
 916             velecsum         = _mm256_add_ps(velecsum,velec);
 917
 918             fscal            = felec;
 919
 920             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 921
 922             /* Calculate temporary vectorial force */
 923             tx               = _mm256_mul_ps(fscal,dx22);
 924             ty               = _mm256_mul_ps(fscal,dy22);
 925             tz               = _mm256_mul_ps(fscal,dz22);
 926
 927             /* Update vectorial force */
 928             fix2             = _mm256_add_ps(fix2,tx);
 929             fiy2             = _mm256_add_ps(fiy2,ty);
 930             fiz2             = _mm256_add_ps(fiz2,tz);
 931
 932             fjx2             = _mm256_add_ps(fjx2,tx);
 933             fjy2             = _mm256_add_ps(fjy2,ty);
 934             fjz2             = _mm256_add_ps(fjz2,tz);
 935
 936             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
 937             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
 938             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
 939             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
 940             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
 941             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
 942             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
 943             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
 944
 945             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
 946                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 947
 948             /* Inner loop uses 243 flops */
 949         }
 950
 951         /* End of innermost loop */
 952
 953         gmx_mm256_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 954                                                  f+i_coord_offset,fshift+i_shift_offset);
 955
 956         ggid                        = gid[iidx];
 957         /* Update potential energies */
 958         gmx_mm256_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
 959
 960         /* Increment number of inner iterations */
 961         inneriter                  += j_index_end - j_index_start;
 962
 963         /* Outer loop uses 19 flops */
 964     }
 965
 966     /* Increment number of outer iterations */
 967     outeriter        += nri;
 968
 969     /* Update outer/inner flops */
 970
 971     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3W3_VF,outeriter*19 + inneriter*243);
 972 }
 973 /*
 974  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwNone_GeomW3W3_F_avx_256_single
 975  * Electrostatics interaction: Coulomb
 976  * VdW interaction:            None
 977  * Geometry:                   Water3-Water3
 978  * Calculate force/pot:        Force
 979  */
 980 void
 981 nb_kernel_ElecCoul_VdwNone_GeomW3W3_F_avx_256_single
 982                     (t_nblist                    * gmx_restrict       nlist,
 983                      rvec                        * gmx_restrict          xx,
 984                      rvec                        * gmx_restrict          ff,
 985                      t_forcerec                  * gmx_restrict          fr,
 986                      t_mdatoms                   * gmx_restrict     mdatoms,
 987                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 988                      t_nrnb                      * gmx_restrict        nrnb)
 989 {
 990     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 991      * just 0 for non-waters.
 992      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
 993      * jnr indices corresponding to data put in the four positions in the SIMD register.
 994      */
 995     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 996     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 997     int              jnrA,jnrB,jnrC,jnrD;
 998     int              jnrE,jnrF,jnrG,jnrH;
 999     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1000     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1001     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1002     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
1003     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1004     real             rcutoff_scalar;
1005     real             *shiftvec,*fshift,*x,*f;
1006     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
1007     real             scratch[4*DIM];
1008     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1009     real *           vdwioffsetptr0;
1010     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
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     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
1016     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
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     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1022     __m256           dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1023     __m256           dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1024     __m256           dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1025     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1026     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1027     __m256           dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1028     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1029     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1030     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
1031     real             *charge;
1032     __m256           dummy_mask,cutoff_mask;
1033     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
1034     __m256           one     = _mm256_set1_ps(1.0);
1035     __m256           two     = _mm256_set1_ps(2.0);
1036     x                = xx[0];
1037     f                = ff[0];
1038
1039     nri              = nlist->nri;
1040     iinr             = nlist->iinr;
1041     jindex           = nlist->jindex;
1042     jjnr             = nlist->jjnr;
1043     shiftidx         = nlist->shift;
1044     gid              = nlist->gid;
1045     shiftvec         = fr->shift_vec[0];
1046     fshift           = fr->fshift[0];
1047     facel            = _mm256_set1_ps(fr->epsfac);
1048     charge           = mdatoms->chargeA;
1049
1050     /* Setup water-specific parameters */
1051     inr              = nlist->iinr[0];
1052     iq0              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+0]));
1053     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
1054     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
1055
1056     jq0              = _mm256_set1_ps(charge[inr+0]);
1057     jq1              = _mm256_set1_ps(charge[inr+1]);
1058     jq2              = _mm256_set1_ps(charge[inr+2]);
1059     qq00             = _mm256_mul_ps(iq0,jq0);
1060     qq01             = _mm256_mul_ps(iq0,jq1);
1061     qq02             = _mm256_mul_ps(iq0,jq2);
1062     qq10             = _mm256_mul_ps(iq1,jq0);
1063     qq11             = _mm256_mul_ps(iq1,jq1);
1064     qq12             = _mm256_mul_ps(iq1,jq2);
1065     qq20             = _mm256_mul_ps(iq2,jq0);
1066     qq21             = _mm256_mul_ps(iq2,jq1);
1067     qq22             = _mm256_mul_ps(iq2,jq2);
1068
1069     /* Avoid stupid compiler warnings */
1070     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
1071     j_coord_offsetA = 0;
1072     j_coord_offsetB = 0;
1073     j_coord_offsetC = 0;
1074     j_coord_offsetD = 0;
1075     j_coord_offsetE = 0;
1076     j_coord_offsetF = 0;
1077     j_coord_offsetG = 0;
1078     j_coord_offsetH = 0;
1079
1080     outeriter        = 0;
1081     inneriter        = 0;
1082
1083     for(iidx=0;iidx<4*DIM;iidx++)
1084     {
1085         scratch[iidx] = 0.0;
1086     }
1087
1088     /* Start outer loop over neighborlists */
1089     for(iidx=0; iidx<nri; iidx++)
1090     {
1091         /* Load shift vector for this list */
1092         i_shift_offset   = DIM*shiftidx[iidx];
1093
1094         /* Load limits for loop over neighbors */
1095         j_index_start    = jindex[iidx];
1096         j_index_end      = jindex[iidx+1];
1097
1098         /* Get outer coordinate index */
1099         inr              = iinr[iidx];
1100         i_coord_offset   = DIM*inr;
1101
1102         /* Load i particle coords and add shift vector */
1103         gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1104                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1105
1106         fix0             = _mm256_setzero_ps();
1107         fiy0             = _mm256_setzero_ps();
1108         fiz0             = _mm256_setzero_ps();
1109         fix1             = _mm256_setzero_ps();
1110         fiy1             = _mm256_setzero_ps();
1111         fiz1             = _mm256_setzero_ps();
1112         fix2             = _mm256_setzero_ps();
1113         fiy2             = _mm256_setzero_ps();
1114         fiz2             = _mm256_setzero_ps();
1115
1116         /* Start inner kernel loop */
1117         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
1118         {
1119
1120             /* Get j neighbor index, and coordinate index */
1121             jnrA             = jjnr[jidx];
1122             jnrB             = jjnr[jidx+1];
1123             jnrC             = jjnr[jidx+2];
1124             jnrD             = jjnr[jidx+3];
1125             jnrE             = jjnr[jidx+4];
1126             jnrF             = jjnr[jidx+5];
1127             jnrG             = jjnr[jidx+6];
1128             jnrH             = jjnr[jidx+7];
1129             j_coord_offsetA  = DIM*jnrA;
1130             j_coord_offsetB  = DIM*jnrB;
1131             j_coord_offsetC  = DIM*jnrC;
1132             j_coord_offsetD  = DIM*jnrD;
1133             j_coord_offsetE  = DIM*jnrE;
1134             j_coord_offsetF  = DIM*jnrF;
1135             j_coord_offsetG  = DIM*jnrG;
1136             j_coord_offsetH  = DIM*jnrH;
1137
1138             /* load j atom coordinates */
1139             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1140                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1141                                                  x+j_coord_offsetE,x+j_coord_offsetF,
1142                                                  x+j_coord_offsetG,x+j_coord_offsetH,
1143                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1144
1145             /* Calculate displacement vector */
1146             dx00             = _mm256_sub_ps(ix0,jx0);
1147             dy00             = _mm256_sub_ps(iy0,jy0);
1148             dz00             = _mm256_sub_ps(iz0,jz0);
1149             dx01             = _mm256_sub_ps(ix0,jx1);
1150             dy01             = _mm256_sub_ps(iy0,jy1);
1151             dz01             = _mm256_sub_ps(iz0,jz1);
1152             dx02             = _mm256_sub_ps(ix0,jx2);
1153             dy02             = _mm256_sub_ps(iy0,jy2);
1154             dz02             = _mm256_sub_ps(iz0,jz2);
1155             dx10             = _mm256_sub_ps(ix1,jx0);
1156             dy10             = _mm256_sub_ps(iy1,jy0);
1157             dz10             = _mm256_sub_ps(iz1,jz0);
1158             dx11             = _mm256_sub_ps(ix1,jx1);
1159             dy11             = _mm256_sub_ps(iy1,jy1);
1160             dz11             = _mm256_sub_ps(iz1,jz1);
1161             dx12             = _mm256_sub_ps(ix1,jx2);
1162             dy12             = _mm256_sub_ps(iy1,jy2);
1163             dz12             = _mm256_sub_ps(iz1,jz2);
1164             dx20             = _mm256_sub_ps(ix2,jx0);
1165             dy20             = _mm256_sub_ps(iy2,jy0);
1166             dz20             = _mm256_sub_ps(iz2,jz0);
1167             dx21             = _mm256_sub_ps(ix2,jx1);
1168             dy21             = _mm256_sub_ps(iy2,jy1);
1169             dz21             = _mm256_sub_ps(iz2,jz1);
1170             dx22             = _mm256_sub_ps(ix2,jx2);
1171             dy22             = _mm256_sub_ps(iy2,jy2);
1172             dz22             = _mm256_sub_ps(iz2,jz2);
1173
1174             /* Calculate squared distance and things based on it */
1175             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1176             rsq01            = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
1177             rsq02            = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
1178             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
1179             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1180             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1181             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
1182             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1183             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1184
1185             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
1186             rinv01           = gmx_mm256_invsqrt_ps(rsq01);
1187             rinv02           = gmx_mm256_invsqrt_ps(rsq02);
1188             rinv10           = gmx_mm256_invsqrt_ps(rsq10);
1189             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1190             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1191             rinv20           = gmx_mm256_invsqrt_ps(rsq20);
1192             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1193             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1194
1195             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
1196             rinvsq01         = _mm256_mul_ps(rinv01,rinv01);
1197             rinvsq02         = _mm256_mul_ps(rinv02,rinv02);
1198             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
1199             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1200             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1201             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
1202             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1203             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1204
1205             fjx0             = _mm256_setzero_ps();
1206             fjy0             = _mm256_setzero_ps();
1207             fjz0             = _mm256_setzero_ps();
1208             fjx1             = _mm256_setzero_ps();
1209             fjy1             = _mm256_setzero_ps();
1210             fjz1             = _mm256_setzero_ps();
1211             fjx2             = _mm256_setzero_ps();
1212             fjy2             = _mm256_setzero_ps();
1213             fjz2             = _mm256_setzero_ps();
1214
1215             /**************************
1216              * CALCULATE INTERACTIONS *
1217              **************************/
1218
1219             /* COULOMB ELECTROSTATICS */
1220             velec            = _mm256_mul_ps(qq00,rinv00);
1221             felec            = _mm256_mul_ps(velec,rinvsq00);
1222
1223             fscal            = felec;
1224
1225             /* Calculate temporary vectorial force */
1226             tx               = _mm256_mul_ps(fscal,dx00);
1227             ty               = _mm256_mul_ps(fscal,dy00);
1228             tz               = _mm256_mul_ps(fscal,dz00);
1229
1230             /* Update vectorial force */
1231             fix0             = _mm256_add_ps(fix0,tx);
1232             fiy0             = _mm256_add_ps(fiy0,ty);
1233             fiz0             = _mm256_add_ps(fiz0,tz);
1234
1235             fjx0             = _mm256_add_ps(fjx0,tx);
1236             fjy0             = _mm256_add_ps(fjy0,ty);
1237             fjz0             = _mm256_add_ps(fjz0,tz);
1238
1239             /**************************
1240              * CALCULATE INTERACTIONS *
1241              **************************/
1242
1243             /* COULOMB ELECTROSTATICS */
1244             velec            = _mm256_mul_ps(qq01,rinv01);
1245             felec            = _mm256_mul_ps(velec,rinvsq01);
1246
1247             fscal            = felec;
1248
1249             /* Calculate temporary vectorial force */
1250             tx               = _mm256_mul_ps(fscal,dx01);
1251             ty               = _mm256_mul_ps(fscal,dy01);
1252             tz               = _mm256_mul_ps(fscal,dz01);
1253
1254             /* Update vectorial force */
1255             fix0             = _mm256_add_ps(fix0,tx);
1256             fiy0             = _mm256_add_ps(fiy0,ty);
1257             fiz0             = _mm256_add_ps(fiz0,tz);
1258
1259             fjx1             = _mm256_add_ps(fjx1,tx);
1260             fjy1             = _mm256_add_ps(fjy1,ty);
1261             fjz1             = _mm256_add_ps(fjz1,tz);
1262
1263             /**************************
1264              * CALCULATE INTERACTIONS *
1265              **************************/
1266
1267             /* COULOMB ELECTROSTATICS */
1268             velec            = _mm256_mul_ps(qq02,rinv02);
1269             felec            = _mm256_mul_ps(velec,rinvsq02);
1270
1271             fscal            = felec;
1272
1273             /* Calculate temporary vectorial force */
1274             tx               = _mm256_mul_ps(fscal,dx02);
1275             ty               = _mm256_mul_ps(fscal,dy02);
1276             tz               = _mm256_mul_ps(fscal,dz02);
1277
1278             /* Update vectorial force */
1279             fix0             = _mm256_add_ps(fix0,tx);
1280             fiy0             = _mm256_add_ps(fiy0,ty);
1281             fiz0             = _mm256_add_ps(fiz0,tz);
1282
1283             fjx2             = _mm256_add_ps(fjx2,tx);
1284             fjy2             = _mm256_add_ps(fjy2,ty);
1285             fjz2             = _mm256_add_ps(fjz2,tz);
1286
1287             /**************************
1288              * CALCULATE INTERACTIONS *
1289              **************************/
1290
1291             /* COULOMB ELECTROSTATICS */
1292             velec            = _mm256_mul_ps(qq10,rinv10);
1293             felec            = _mm256_mul_ps(velec,rinvsq10);
1294
1295             fscal            = felec;
1296
1297             /* Calculate temporary vectorial force */
1298             tx               = _mm256_mul_ps(fscal,dx10);
1299             ty               = _mm256_mul_ps(fscal,dy10);
1300             tz               = _mm256_mul_ps(fscal,dz10);
1301
1302             /* Update vectorial force */
1303             fix1             = _mm256_add_ps(fix1,tx);
1304             fiy1             = _mm256_add_ps(fiy1,ty);
1305             fiz1             = _mm256_add_ps(fiz1,tz);
1306
1307             fjx0             = _mm256_add_ps(fjx0,tx);
1308             fjy0             = _mm256_add_ps(fjy0,ty);
1309             fjz0             = _mm256_add_ps(fjz0,tz);
1310
1311             /**************************
1312              * CALCULATE INTERACTIONS *
1313              **************************/
1314
1315             /* COULOMB ELECTROSTATICS */
1316             velec            = _mm256_mul_ps(qq11,rinv11);
1317             felec            = _mm256_mul_ps(velec,rinvsq11);
1318
1319             fscal            = felec;
1320
1321             /* Calculate temporary vectorial force */
1322             tx               = _mm256_mul_ps(fscal,dx11);
1323             ty               = _mm256_mul_ps(fscal,dy11);
1324             tz               = _mm256_mul_ps(fscal,dz11);
1325
1326             /* Update vectorial force */
1327             fix1             = _mm256_add_ps(fix1,tx);
1328             fiy1             = _mm256_add_ps(fiy1,ty);
1329             fiz1             = _mm256_add_ps(fiz1,tz);
1330
1331             fjx1             = _mm256_add_ps(fjx1,tx);
1332             fjy1             = _mm256_add_ps(fjy1,ty);
1333             fjz1             = _mm256_add_ps(fjz1,tz);
1334
1335             /**************************
1336              * CALCULATE INTERACTIONS *
1337              **************************/
1338
1339             /* COULOMB ELECTROSTATICS */
1340             velec            = _mm256_mul_ps(qq12,rinv12);
1341             felec            = _mm256_mul_ps(velec,rinvsq12);
1342
1343             fscal            = felec;
1344
1345             /* Calculate temporary vectorial force */
1346             tx               = _mm256_mul_ps(fscal,dx12);
1347             ty               = _mm256_mul_ps(fscal,dy12);
1348             tz               = _mm256_mul_ps(fscal,dz12);
1349
1350             /* Update vectorial force */
1351             fix1             = _mm256_add_ps(fix1,tx);
1352             fiy1             = _mm256_add_ps(fiy1,ty);
1353             fiz1             = _mm256_add_ps(fiz1,tz);
1354
1355             fjx2             = _mm256_add_ps(fjx2,tx);
1356             fjy2             = _mm256_add_ps(fjy2,ty);
1357             fjz2             = _mm256_add_ps(fjz2,tz);
1358
1359             /**************************
1360              * CALCULATE INTERACTIONS *
1361              **************************/
1362
1363             /* COULOMB ELECTROSTATICS */
1364             velec            = _mm256_mul_ps(qq20,rinv20);
1365             felec            = _mm256_mul_ps(velec,rinvsq20);
1366
1367             fscal            = felec;
1368
1369             /* Calculate temporary vectorial force */
1370             tx               = _mm256_mul_ps(fscal,dx20);
1371             ty               = _mm256_mul_ps(fscal,dy20);
1372             tz               = _mm256_mul_ps(fscal,dz20);
1373
1374             /* Update vectorial force */
1375             fix2             = _mm256_add_ps(fix2,tx);
1376             fiy2             = _mm256_add_ps(fiy2,ty);
1377             fiz2             = _mm256_add_ps(fiz2,tz);
1378
1379             fjx0             = _mm256_add_ps(fjx0,tx);
1380             fjy0             = _mm256_add_ps(fjy0,ty);
1381             fjz0             = _mm256_add_ps(fjz0,tz);
1382
1383             /**************************
1384              * CALCULATE INTERACTIONS *
1385              **************************/
1386
1387             /* COULOMB ELECTROSTATICS */
1388             velec            = _mm256_mul_ps(qq21,rinv21);
1389             felec            = _mm256_mul_ps(velec,rinvsq21);
1390
1391             fscal            = felec;
1392
1393             /* Calculate temporary vectorial force */
1394             tx               = _mm256_mul_ps(fscal,dx21);
1395             ty               = _mm256_mul_ps(fscal,dy21);
1396             tz               = _mm256_mul_ps(fscal,dz21);
1397
1398             /* Update vectorial force */
1399             fix2             = _mm256_add_ps(fix2,tx);
1400             fiy2             = _mm256_add_ps(fiy2,ty);
1401             fiz2             = _mm256_add_ps(fiz2,tz);
1402
1403             fjx1             = _mm256_add_ps(fjx1,tx);
1404             fjy1             = _mm256_add_ps(fjy1,ty);
1405             fjz1             = _mm256_add_ps(fjz1,tz);
1406
1407             /**************************
1408              * CALCULATE INTERACTIONS *
1409              **************************/
1410
1411             /* COULOMB ELECTROSTATICS */
1412             velec            = _mm256_mul_ps(qq22,rinv22);
1413             felec            = _mm256_mul_ps(velec,rinvsq22);
1414
1415             fscal            = felec;
1416
1417             /* Calculate temporary vectorial force */
1418             tx               = _mm256_mul_ps(fscal,dx22);
1419             ty               = _mm256_mul_ps(fscal,dy22);
1420             tz               = _mm256_mul_ps(fscal,dz22);
1421
1422             /* Update vectorial force */
1423             fix2             = _mm256_add_ps(fix2,tx);
1424             fiy2             = _mm256_add_ps(fiy2,ty);
1425             fiz2             = _mm256_add_ps(fiz2,tz);
1426
1427             fjx2             = _mm256_add_ps(fjx2,tx);
1428             fjy2             = _mm256_add_ps(fjy2,ty);
1429             fjz2             = _mm256_add_ps(fjz2,tz);
1430
1431             fjptrA             = f+j_coord_offsetA;
1432             fjptrB             = f+j_coord_offsetB;
1433             fjptrC             = f+j_coord_offsetC;
1434             fjptrD             = f+j_coord_offsetD;
1435             fjptrE             = f+j_coord_offsetE;
1436             fjptrF             = f+j_coord_offsetF;
1437             fjptrG             = f+j_coord_offsetG;
1438             fjptrH             = f+j_coord_offsetH;
1439
1440             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1441                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1442
1443             /* Inner loop uses 234 flops */
1444         }
1445
1446         if(jidx<j_index_end)
1447         {
1448
1449             /* Get j neighbor index, and coordinate index */
1450             jnrlistA         = jjnr[jidx];
1451             jnrlistB         = jjnr[jidx+1];
1452             jnrlistC         = jjnr[jidx+2];
1453             jnrlistD         = jjnr[jidx+3];
1454             jnrlistE         = jjnr[jidx+4];
1455             jnrlistF         = jjnr[jidx+5];
1456             jnrlistG         = jjnr[jidx+6];
1457             jnrlistH         = jjnr[jidx+7];
1458             /* Sign of each element will be negative for non-real atoms.
1459              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1460              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1461              */
1462             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
1463                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
1464
1465             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1466             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1467             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1468             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1469             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
1470             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
1471             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
1472             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
1473             j_coord_offsetA  = DIM*jnrA;
1474             j_coord_offsetB  = DIM*jnrB;
1475             j_coord_offsetC  = DIM*jnrC;
1476             j_coord_offsetD  = DIM*jnrD;
1477             j_coord_offsetE  = DIM*jnrE;
1478             j_coord_offsetF  = DIM*jnrF;
1479             j_coord_offsetG  = DIM*jnrG;
1480             j_coord_offsetH  = DIM*jnrH;
1481
1482             /* load j atom coordinates */
1483             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1484                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1485                                                  x+j_coord_offsetE,x+j_coord_offsetF,
1486                                                  x+j_coord_offsetG,x+j_coord_offsetH,
1487                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1488
1489             /* Calculate displacement vector */
1490             dx00             = _mm256_sub_ps(ix0,jx0);
1491             dy00             = _mm256_sub_ps(iy0,jy0);
1492             dz00             = _mm256_sub_ps(iz0,jz0);
1493             dx01             = _mm256_sub_ps(ix0,jx1);
1494             dy01             = _mm256_sub_ps(iy0,jy1);
1495             dz01             = _mm256_sub_ps(iz0,jz1);
1496             dx02             = _mm256_sub_ps(ix0,jx2);
1497             dy02             = _mm256_sub_ps(iy0,jy2);
1498             dz02             = _mm256_sub_ps(iz0,jz2);
1499             dx10             = _mm256_sub_ps(ix1,jx0);
1500             dy10             = _mm256_sub_ps(iy1,jy0);
1501             dz10             = _mm256_sub_ps(iz1,jz0);
1502             dx11             = _mm256_sub_ps(ix1,jx1);
1503             dy11             = _mm256_sub_ps(iy1,jy1);
1504             dz11             = _mm256_sub_ps(iz1,jz1);
1505             dx12             = _mm256_sub_ps(ix1,jx2);
1506             dy12             = _mm256_sub_ps(iy1,jy2);
1507             dz12             = _mm256_sub_ps(iz1,jz2);
1508             dx20             = _mm256_sub_ps(ix2,jx0);
1509             dy20             = _mm256_sub_ps(iy2,jy0);
1510             dz20             = _mm256_sub_ps(iz2,jz0);
1511             dx21             = _mm256_sub_ps(ix2,jx1);
1512             dy21             = _mm256_sub_ps(iy2,jy1);
1513             dz21             = _mm256_sub_ps(iz2,jz1);
1514             dx22             = _mm256_sub_ps(ix2,jx2);
1515             dy22             = _mm256_sub_ps(iy2,jy2);
1516             dz22             = _mm256_sub_ps(iz2,jz2);
1517
1518             /* Calculate squared distance and things based on it */
1519             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1520             rsq01            = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
1521             rsq02            = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
1522             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
1523             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1524             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1525             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
1526             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1527             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1528
1529             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
1530             rinv01           = gmx_mm256_invsqrt_ps(rsq01);
1531             rinv02           = gmx_mm256_invsqrt_ps(rsq02);
1532             rinv10           = gmx_mm256_invsqrt_ps(rsq10);
1533             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1534             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1535             rinv20           = gmx_mm256_invsqrt_ps(rsq20);
1536             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1537             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1538
1539             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
1540             rinvsq01         = _mm256_mul_ps(rinv01,rinv01);
1541             rinvsq02         = _mm256_mul_ps(rinv02,rinv02);
1542             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
1543             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1544             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1545             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
1546             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1547             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1548
1549             fjx0             = _mm256_setzero_ps();
1550             fjy0             = _mm256_setzero_ps();
1551             fjz0             = _mm256_setzero_ps();
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
1559             /**************************
1560              * CALCULATE INTERACTIONS *
1561              **************************/
1562
1563             /* COULOMB ELECTROSTATICS */
1564             velec            = _mm256_mul_ps(qq00,rinv00);
1565             felec            = _mm256_mul_ps(velec,rinvsq00);
1566
1567             fscal            = felec;
1568
1569             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1570
1571             /* Calculate temporary vectorial force */
1572             tx               = _mm256_mul_ps(fscal,dx00);
1573             ty               = _mm256_mul_ps(fscal,dy00);
1574             tz               = _mm256_mul_ps(fscal,dz00);
1575
1576             /* Update vectorial force */
1577             fix0             = _mm256_add_ps(fix0,tx);
1578             fiy0             = _mm256_add_ps(fiy0,ty);
1579             fiz0             = _mm256_add_ps(fiz0,tz);
1580
1581             fjx0             = _mm256_add_ps(fjx0,tx);
1582             fjy0             = _mm256_add_ps(fjy0,ty);
1583             fjz0             = _mm256_add_ps(fjz0,tz);
1584
1585             /**************************
1586              * CALCULATE INTERACTIONS *
1587              **************************/
1588
1589             /* COULOMB ELECTROSTATICS */
1590             velec            = _mm256_mul_ps(qq01,rinv01);
1591             felec            = _mm256_mul_ps(velec,rinvsq01);
1592
1593             fscal            = felec;
1594
1595             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1596
1597             /* Calculate temporary vectorial force */
1598             tx               = _mm256_mul_ps(fscal,dx01);
1599             ty               = _mm256_mul_ps(fscal,dy01);
1600             tz               = _mm256_mul_ps(fscal,dz01);
1601
1602             /* Update vectorial force */
1603             fix0             = _mm256_add_ps(fix0,tx);
1604             fiy0             = _mm256_add_ps(fiy0,ty);
1605             fiz0             = _mm256_add_ps(fiz0,tz);
1606
1607             fjx1             = _mm256_add_ps(fjx1,tx);
1608             fjy1             = _mm256_add_ps(fjy1,ty);
1609             fjz1             = _mm256_add_ps(fjz1,tz);
1610
1611             /**************************
1612              * CALCULATE INTERACTIONS *
1613              **************************/
1614
1615             /* COULOMB ELECTROSTATICS */
1616             velec            = _mm256_mul_ps(qq02,rinv02);
1617             felec            = _mm256_mul_ps(velec,rinvsq02);
1618
1619             fscal            = felec;
1620
1621             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1622
1623             /* Calculate temporary vectorial force */
1624             tx               = _mm256_mul_ps(fscal,dx02);
1625             ty               = _mm256_mul_ps(fscal,dy02);
1626             tz               = _mm256_mul_ps(fscal,dz02);
1627
1628             /* Update vectorial force */
1629             fix0             = _mm256_add_ps(fix0,tx);
1630             fiy0             = _mm256_add_ps(fiy0,ty);
1631             fiz0             = _mm256_add_ps(fiz0,tz);
1632
1633             fjx2             = _mm256_add_ps(fjx2,tx);
1634             fjy2             = _mm256_add_ps(fjy2,ty);
1635             fjz2             = _mm256_add_ps(fjz2,tz);
1636
1637             /**************************
1638              * CALCULATE INTERACTIONS *
1639              **************************/
1640
1641             /* COULOMB ELECTROSTATICS */
1642             velec            = _mm256_mul_ps(qq10,rinv10);
1643             felec            = _mm256_mul_ps(velec,rinvsq10);
1644
1645             fscal            = felec;
1646
1647             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1648
1649             /* Calculate temporary vectorial force */
1650             tx               = _mm256_mul_ps(fscal,dx10);
1651             ty               = _mm256_mul_ps(fscal,dy10);
1652             tz               = _mm256_mul_ps(fscal,dz10);
1653
1654             /* Update vectorial force */
1655             fix1             = _mm256_add_ps(fix1,tx);
1656             fiy1             = _mm256_add_ps(fiy1,ty);
1657             fiz1             = _mm256_add_ps(fiz1,tz);
1658
1659             fjx0             = _mm256_add_ps(fjx0,tx);
1660             fjy0             = _mm256_add_ps(fjy0,ty);
1661             fjz0             = _mm256_add_ps(fjz0,tz);
1662
1663             /**************************
1664              * CALCULATE INTERACTIONS *
1665              **************************/
1666
1667             /* COULOMB ELECTROSTATICS */
1668             velec            = _mm256_mul_ps(qq11,rinv11);
1669             felec            = _mm256_mul_ps(velec,rinvsq11);
1670
1671             fscal            = felec;
1672
1673             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1674
1675             /* Calculate temporary vectorial force */
1676             tx               = _mm256_mul_ps(fscal,dx11);
1677             ty               = _mm256_mul_ps(fscal,dy11);
1678             tz               = _mm256_mul_ps(fscal,dz11);
1679
1680             /* Update vectorial force */
1681             fix1             = _mm256_add_ps(fix1,tx);
1682             fiy1             = _mm256_add_ps(fiy1,ty);
1683             fiz1             = _mm256_add_ps(fiz1,tz);
1684
1685             fjx1             = _mm256_add_ps(fjx1,tx);
1686             fjy1             = _mm256_add_ps(fjy1,ty);
1687             fjz1             = _mm256_add_ps(fjz1,tz);
1688
1689             /**************************
1690              * CALCULATE INTERACTIONS *
1691              **************************/
1692
1693             /* COULOMB ELECTROSTATICS */
1694             velec            = _mm256_mul_ps(qq12,rinv12);
1695             felec            = _mm256_mul_ps(velec,rinvsq12);
1696
1697             fscal            = felec;
1698
1699             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1700
1701             /* Calculate temporary vectorial force */
1702             tx               = _mm256_mul_ps(fscal,dx12);
1703             ty               = _mm256_mul_ps(fscal,dy12);
1704             tz               = _mm256_mul_ps(fscal,dz12);
1705
1706             /* Update vectorial force */
1707             fix1             = _mm256_add_ps(fix1,tx);
1708             fiy1             = _mm256_add_ps(fiy1,ty);
1709             fiz1             = _mm256_add_ps(fiz1,tz);
1710
1711             fjx2             = _mm256_add_ps(fjx2,tx);
1712             fjy2             = _mm256_add_ps(fjy2,ty);
1713             fjz2             = _mm256_add_ps(fjz2,tz);
1714
1715             /**************************
1716              * CALCULATE INTERACTIONS *
1717              **************************/
1718
1719             /* COULOMB ELECTROSTATICS */
1720             velec            = _mm256_mul_ps(qq20,rinv20);
1721             felec            = _mm256_mul_ps(velec,rinvsq20);
1722
1723             fscal            = felec;
1724
1725             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1726
1727             /* Calculate temporary vectorial force */
1728             tx               = _mm256_mul_ps(fscal,dx20);
1729             ty               = _mm256_mul_ps(fscal,dy20);
1730             tz               = _mm256_mul_ps(fscal,dz20);
1731
1732             /* Update vectorial force */
1733             fix2             = _mm256_add_ps(fix2,tx);
1734             fiy2             = _mm256_add_ps(fiy2,ty);
1735             fiz2             = _mm256_add_ps(fiz2,tz);
1736
1737             fjx0             = _mm256_add_ps(fjx0,tx);
1738             fjy0             = _mm256_add_ps(fjy0,ty);
1739             fjz0             = _mm256_add_ps(fjz0,tz);
1740
1741             /**************************
1742              * CALCULATE INTERACTIONS *
1743              **************************/
1744
1745             /* COULOMB ELECTROSTATICS */
1746             velec            = _mm256_mul_ps(qq21,rinv21);
1747             felec            = _mm256_mul_ps(velec,rinvsq21);
1748
1749             fscal            = felec;
1750
1751             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1752
1753             /* Calculate temporary vectorial force */
1754             tx               = _mm256_mul_ps(fscal,dx21);
1755             ty               = _mm256_mul_ps(fscal,dy21);
1756             tz               = _mm256_mul_ps(fscal,dz21);
1757
1758             /* Update vectorial force */
1759             fix2             = _mm256_add_ps(fix2,tx);
1760             fiy2             = _mm256_add_ps(fiy2,ty);
1761             fiz2             = _mm256_add_ps(fiz2,tz);
1762
1763             fjx1             = _mm256_add_ps(fjx1,tx);
1764             fjy1             = _mm256_add_ps(fjy1,ty);
1765             fjz1             = _mm256_add_ps(fjz1,tz);
1766
1767             /**************************
1768              * CALCULATE INTERACTIONS *
1769              **************************/
1770
1771             /* COULOMB ELECTROSTATICS */
1772             velec            = _mm256_mul_ps(qq22,rinv22);
1773             felec            = _mm256_mul_ps(velec,rinvsq22);
1774
1775             fscal            = felec;
1776
1777             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1778
1779             /* Calculate temporary vectorial force */
1780             tx               = _mm256_mul_ps(fscal,dx22);
1781             ty               = _mm256_mul_ps(fscal,dy22);
1782             tz               = _mm256_mul_ps(fscal,dz22);
1783
1784             /* Update vectorial force */
1785             fix2             = _mm256_add_ps(fix2,tx);
1786             fiy2             = _mm256_add_ps(fiy2,ty);
1787             fiz2             = _mm256_add_ps(fiz2,tz);
1788
1789             fjx2             = _mm256_add_ps(fjx2,tx);
1790             fjy2             = _mm256_add_ps(fjy2,ty);
1791             fjz2             = _mm256_add_ps(fjz2,tz);
1792
1793             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1794             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1795             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1796             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1797             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1798             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1799             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
1800             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
1801
1802             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1803                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1804
1805             /* Inner loop uses 234 flops */
1806         }
1807
1808         /* End of innermost loop */
1809
1810         gmx_mm256_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1811                                                  f+i_coord_offset,fshift+i_shift_offset);
1812
1813         /* Increment number of inner iterations */
1814         inneriter                  += j_index_end - j_index_start;
1815
1816         /* Outer loop uses 18 flops */
1817     }
1818
1819     /* Increment number of outer iterations */
1820     outeriter        += nri;
1821
1822     /* Update outer/inner flops */
1823
1824     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3W3_F,outeriter*18 + inneriter*234);
1825 }