src/gromacs/gmxlib/nonbonded/nb_kernel_sse4_1_double/nb_kernel_ElecRF_VdwNone_GeomW4W4_sse4_1_double.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014,2015,2017, 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 sse4_1_double kernel generator.
  37  */
  38 #include "gmxpre.h"
  39
  40 #include "config.h"
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "gromacs/gmxlib/nrnb.h"
  46
  47 #include "kernelutil_x86_sse4_1_double.h"
  48
  49 /*
  50  * Gromacs nonbonded kernel:   nb_kernel_ElecRF_VdwNone_GeomW4W4_VF_sse4_1_double
  51  * Electrostatics interaction: ReactionField
  52  * VdW interaction:            None
  53  * Geometry:                   Water4-Water4
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecRF_VdwNone_GeomW4W4_VF_sse4_1_double
  58                     (t_nblist                    * gmx_restrict       nlist,
  59                      rvec                        * gmx_restrict          xx,
  60                      rvec                        * gmx_restrict          ff,
  61                      struct t_forcerec           * gmx_restrict          fr,
  62                      t_mdatoms                   * gmx_restrict     mdatoms,
  63                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  64                      t_nrnb                      * gmx_restrict        nrnb)
  65 {
  66     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  67      * just 0 for non-waters.
  68      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
  69      * jnr indices corresponding to data put in the four positions in the SIMD register.
  70      */
  71     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  72     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  73     int              jnrA,jnrB;
  74     int              j_coord_offsetA,j_coord_offsetB;
  75     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  76     real             rcutoff_scalar;
  77     real             *shiftvec,*fshift,*x,*f;
  78     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  79     int              vdwioffset1;
  80     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  81     int              vdwioffset2;
  82     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  83     int              vdwioffset3;
  84     __m128d          ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  85     int              vdwjidx1A,vdwjidx1B;
  86     __m128d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  87     int              vdwjidx2A,vdwjidx2B;
  88     __m128d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  89     int              vdwjidx3A,vdwjidx3B;
  90     __m128d          jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
  91     __m128d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
  92     __m128d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
  93     __m128d          dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
  94     __m128d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
  95     __m128d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
  96     __m128d          dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
  97     __m128d          dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
  98     __m128d          dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
  99     __m128d          dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
 100     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
 101     real             *charge;
 102     __m128d          dummy_mask,cutoff_mask;
 103     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 104     __m128d          one     = _mm_set1_pd(1.0);
 105     __m128d          two     = _mm_set1_pd(2.0);
 106     x                = xx[0];
 107     f                = ff[0];
 108
 109     nri              = nlist->nri;
 110     iinr             = nlist->iinr;
 111     jindex           = nlist->jindex;
 112     jjnr             = nlist->jjnr;
 113     shiftidx         = nlist->shift;
 114     gid              = nlist->gid;
 115     shiftvec         = fr->shift_vec[0];
 116     fshift           = fr->fshift[0];
 117     facel            = _mm_set1_pd(fr->ic->epsfac);
 118     charge           = mdatoms->chargeA;
 119     krf              = _mm_set1_pd(fr->ic->k_rf);
 120     krf2             = _mm_set1_pd(fr->ic->k_rf*2.0);
 121     crf              = _mm_set1_pd(fr->ic->c_rf);
 122
 123     /* Setup water-specific parameters */
 124     inr              = nlist->iinr[0];
 125     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
 126     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
 127     iq3              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
 128
 129     jq1              = _mm_set1_pd(charge[inr+1]);
 130     jq2              = _mm_set1_pd(charge[inr+2]);
 131     jq3              = _mm_set1_pd(charge[inr+3]);
 132     qq11             = _mm_mul_pd(iq1,jq1);
 133     qq12             = _mm_mul_pd(iq1,jq2);
 134     qq13             = _mm_mul_pd(iq1,jq3);
 135     qq21             = _mm_mul_pd(iq2,jq1);
 136     qq22             = _mm_mul_pd(iq2,jq2);
 137     qq23             = _mm_mul_pd(iq2,jq3);
 138     qq31             = _mm_mul_pd(iq3,jq1);
 139     qq32             = _mm_mul_pd(iq3,jq2);
 140     qq33             = _mm_mul_pd(iq3,jq3);
 141
 142     /* Avoid stupid compiler warnings */
 143     jnrA = jnrB = 0;
 144     j_coord_offsetA = 0;
 145     j_coord_offsetB = 0;
 146
 147     outeriter        = 0;
 148     inneriter        = 0;
 149
 150     /* Start outer loop over neighborlists */
 151     for(iidx=0; iidx<nri; iidx++)
 152     {
 153         /* Load shift vector for this list */
 154         i_shift_offset   = DIM*shiftidx[iidx];
 155
 156         /* Load limits for loop over neighbors */
 157         j_index_start    = jindex[iidx];
 158         j_index_end      = jindex[iidx+1];
 159
 160         /* Get outer coordinate index */
 161         inr              = iinr[iidx];
 162         i_coord_offset   = DIM*inr;
 163
 164         /* Load i particle coords and add shift vector */
 165         gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
 166                                                  &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 167
 168         fix1             = _mm_setzero_pd();
 169         fiy1             = _mm_setzero_pd();
 170         fiz1             = _mm_setzero_pd();
 171         fix2             = _mm_setzero_pd();
 172         fiy2             = _mm_setzero_pd();
 173         fiz2             = _mm_setzero_pd();
 174         fix3             = _mm_setzero_pd();
 175         fiy3             = _mm_setzero_pd();
 176         fiz3             = _mm_setzero_pd();
 177
 178         /* Reset potential sums */
 179         velecsum         = _mm_setzero_pd();
 180
 181         /* Start inner kernel loop */
 182         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 183         {
 184
 185             /* Get j neighbor index, and coordinate index */
 186             jnrA             = jjnr[jidx];
 187             jnrB             = jjnr[jidx+1];
 188             j_coord_offsetA  = DIM*jnrA;
 189             j_coord_offsetB  = DIM*jnrB;
 190
 191             /* load j atom coordinates */
 192             gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
 193                                               &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
 194
 195             /* Calculate displacement vector */
 196             dx11             = _mm_sub_pd(ix1,jx1);
 197             dy11             = _mm_sub_pd(iy1,jy1);
 198             dz11             = _mm_sub_pd(iz1,jz1);
 199             dx12             = _mm_sub_pd(ix1,jx2);
 200             dy12             = _mm_sub_pd(iy1,jy2);
 201             dz12             = _mm_sub_pd(iz1,jz2);
 202             dx13             = _mm_sub_pd(ix1,jx3);
 203             dy13             = _mm_sub_pd(iy1,jy3);
 204             dz13             = _mm_sub_pd(iz1,jz3);
 205             dx21             = _mm_sub_pd(ix2,jx1);
 206             dy21             = _mm_sub_pd(iy2,jy1);
 207             dz21             = _mm_sub_pd(iz2,jz1);
 208             dx22             = _mm_sub_pd(ix2,jx2);
 209             dy22             = _mm_sub_pd(iy2,jy2);
 210             dz22             = _mm_sub_pd(iz2,jz2);
 211             dx23             = _mm_sub_pd(ix2,jx3);
 212             dy23             = _mm_sub_pd(iy2,jy3);
 213             dz23             = _mm_sub_pd(iz2,jz3);
 214             dx31             = _mm_sub_pd(ix3,jx1);
 215             dy31             = _mm_sub_pd(iy3,jy1);
 216             dz31             = _mm_sub_pd(iz3,jz1);
 217             dx32             = _mm_sub_pd(ix3,jx2);
 218             dy32             = _mm_sub_pd(iy3,jy2);
 219             dz32             = _mm_sub_pd(iz3,jz2);
 220             dx33             = _mm_sub_pd(ix3,jx3);
 221             dy33             = _mm_sub_pd(iy3,jy3);
 222             dz33             = _mm_sub_pd(iz3,jz3);
 223
 224             /* Calculate squared distance and things based on it */
 225             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 226             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 227             rsq13            = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
 228             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 229             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 230             rsq23            = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
 231             rsq31            = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
 232             rsq32            = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
 233             rsq33            = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
 234
 235             rinv11           = sse41_invsqrt_d(rsq11);
 236             rinv12           = sse41_invsqrt_d(rsq12);
 237             rinv13           = sse41_invsqrt_d(rsq13);
 238             rinv21           = sse41_invsqrt_d(rsq21);
 239             rinv22           = sse41_invsqrt_d(rsq22);
 240             rinv23           = sse41_invsqrt_d(rsq23);
 241             rinv31           = sse41_invsqrt_d(rsq31);
 242             rinv32           = sse41_invsqrt_d(rsq32);
 243             rinv33           = sse41_invsqrt_d(rsq33);
 244
 245             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
 246             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
 247             rinvsq13         = _mm_mul_pd(rinv13,rinv13);
 248             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
 249             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
 250             rinvsq23         = _mm_mul_pd(rinv23,rinv23);
 251             rinvsq31         = _mm_mul_pd(rinv31,rinv31);
 252             rinvsq32         = _mm_mul_pd(rinv32,rinv32);
 253             rinvsq33         = _mm_mul_pd(rinv33,rinv33);
 254
 255             fjx1             = _mm_setzero_pd();
 256             fjy1             = _mm_setzero_pd();
 257             fjz1             = _mm_setzero_pd();
 258             fjx2             = _mm_setzero_pd();
 259             fjy2             = _mm_setzero_pd();
 260             fjz2             = _mm_setzero_pd();
 261             fjx3             = _mm_setzero_pd();
 262             fjy3             = _mm_setzero_pd();
 263             fjz3             = _mm_setzero_pd();
 264
 265             /**************************
 266              * CALCULATE INTERACTIONS *
 267              **************************/
 268
 269             /* REACTION-FIELD ELECTROSTATICS */
 270             velec            = _mm_mul_pd(qq11,_mm_sub_pd(_mm_add_pd(rinv11,_mm_mul_pd(krf,rsq11)),crf));
 271             felec            = _mm_mul_pd(qq11,_mm_sub_pd(_mm_mul_pd(rinv11,rinvsq11),krf2));
 272
 273             /* Update potential sum for this i atom from the interaction with this j atom. */
 274             velecsum         = _mm_add_pd(velecsum,velec);
 275
 276             fscal            = felec;
 277
 278             /* Calculate temporary vectorial force */
 279             tx               = _mm_mul_pd(fscal,dx11);
 280             ty               = _mm_mul_pd(fscal,dy11);
 281             tz               = _mm_mul_pd(fscal,dz11);
 282
 283             /* Update vectorial force */
 284             fix1             = _mm_add_pd(fix1,tx);
 285             fiy1             = _mm_add_pd(fiy1,ty);
 286             fiz1             = _mm_add_pd(fiz1,tz);
 287
 288             fjx1             = _mm_add_pd(fjx1,tx);
 289             fjy1             = _mm_add_pd(fjy1,ty);
 290             fjz1             = _mm_add_pd(fjz1,tz);
 291
 292             /**************************
 293              * CALCULATE INTERACTIONS *
 294              **************************/
 295
 296             /* REACTION-FIELD ELECTROSTATICS */
 297             velec            = _mm_mul_pd(qq12,_mm_sub_pd(_mm_add_pd(rinv12,_mm_mul_pd(krf,rsq12)),crf));
 298             felec            = _mm_mul_pd(qq12,_mm_sub_pd(_mm_mul_pd(rinv12,rinvsq12),krf2));
 299
 300             /* Update potential sum for this i atom from the interaction with this j atom. */
 301             velecsum         = _mm_add_pd(velecsum,velec);
 302
 303             fscal            = felec;
 304
 305             /* Calculate temporary vectorial force */
 306             tx               = _mm_mul_pd(fscal,dx12);
 307             ty               = _mm_mul_pd(fscal,dy12);
 308             tz               = _mm_mul_pd(fscal,dz12);
 309
 310             /* Update vectorial force */
 311             fix1             = _mm_add_pd(fix1,tx);
 312             fiy1             = _mm_add_pd(fiy1,ty);
 313             fiz1             = _mm_add_pd(fiz1,tz);
 314
 315             fjx2             = _mm_add_pd(fjx2,tx);
 316             fjy2             = _mm_add_pd(fjy2,ty);
 317             fjz2             = _mm_add_pd(fjz2,tz);
 318
 319             /**************************
 320              * CALCULATE INTERACTIONS *
 321              **************************/
 322
 323             /* REACTION-FIELD ELECTROSTATICS */
 324             velec            = _mm_mul_pd(qq13,_mm_sub_pd(_mm_add_pd(rinv13,_mm_mul_pd(krf,rsq13)),crf));
 325             felec            = _mm_mul_pd(qq13,_mm_sub_pd(_mm_mul_pd(rinv13,rinvsq13),krf2));
 326
 327             /* Update potential sum for this i atom from the interaction with this j atom. */
 328             velecsum         = _mm_add_pd(velecsum,velec);
 329
 330             fscal            = felec;
 331
 332             /* Calculate temporary vectorial force */
 333             tx               = _mm_mul_pd(fscal,dx13);
 334             ty               = _mm_mul_pd(fscal,dy13);
 335             tz               = _mm_mul_pd(fscal,dz13);
 336
 337             /* Update vectorial force */
 338             fix1             = _mm_add_pd(fix1,tx);
 339             fiy1             = _mm_add_pd(fiy1,ty);
 340             fiz1             = _mm_add_pd(fiz1,tz);
 341
 342             fjx3             = _mm_add_pd(fjx3,tx);
 343             fjy3             = _mm_add_pd(fjy3,ty);
 344             fjz3             = _mm_add_pd(fjz3,tz);
 345
 346             /**************************
 347              * CALCULATE INTERACTIONS *
 348              **************************/
 349
 350             /* REACTION-FIELD ELECTROSTATICS */
 351             velec            = _mm_mul_pd(qq21,_mm_sub_pd(_mm_add_pd(rinv21,_mm_mul_pd(krf,rsq21)),crf));
 352             felec            = _mm_mul_pd(qq21,_mm_sub_pd(_mm_mul_pd(rinv21,rinvsq21),krf2));
 353
 354             /* Update potential sum for this i atom from the interaction with this j atom. */
 355             velecsum         = _mm_add_pd(velecsum,velec);
 356
 357             fscal            = felec;
 358
 359             /* Calculate temporary vectorial force */
 360             tx               = _mm_mul_pd(fscal,dx21);
 361             ty               = _mm_mul_pd(fscal,dy21);
 362             tz               = _mm_mul_pd(fscal,dz21);
 363
 364             /* Update vectorial force */
 365             fix2             = _mm_add_pd(fix2,tx);
 366             fiy2             = _mm_add_pd(fiy2,ty);
 367             fiz2             = _mm_add_pd(fiz2,tz);
 368
 369             fjx1             = _mm_add_pd(fjx1,tx);
 370             fjy1             = _mm_add_pd(fjy1,ty);
 371             fjz1             = _mm_add_pd(fjz1,tz);
 372
 373             /**************************
 374              * CALCULATE INTERACTIONS *
 375              **************************/
 376
 377             /* REACTION-FIELD ELECTROSTATICS */
 378             velec            = _mm_mul_pd(qq22,_mm_sub_pd(_mm_add_pd(rinv22,_mm_mul_pd(krf,rsq22)),crf));
 379             felec            = _mm_mul_pd(qq22,_mm_sub_pd(_mm_mul_pd(rinv22,rinvsq22),krf2));
 380
 381             /* Update potential sum for this i atom from the interaction with this j atom. */
 382             velecsum         = _mm_add_pd(velecsum,velec);
 383
 384             fscal            = felec;
 385
 386             /* Calculate temporary vectorial force */
 387             tx               = _mm_mul_pd(fscal,dx22);
 388             ty               = _mm_mul_pd(fscal,dy22);
 389             tz               = _mm_mul_pd(fscal,dz22);
 390
 391             /* Update vectorial force */
 392             fix2             = _mm_add_pd(fix2,tx);
 393             fiy2             = _mm_add_pd(fiy2,ty);
 394             fiz2             = _mm_add_pd(fiz2,tz);
 395
 396             fjx2             = _mm_add_pd(fjx2,tx);
 397             fjy2             = _mm_add_pd(fjy2,ty);
 398             fjz2             = _mm_add_pd(fjz2,tz);
 399
 400             /**************************
 401              * CALCULATE INTERACTIONS *
 402              **************************/
 403
 404             /* REACTION-FIELD ELECTROSTATICS */
 405             velec            = _mm_mul_pd(qq23,_mm_sub_pd(_mm_add_pd(rinv23,_mm_mul_pd(krf,rsq23)),crf));
 406             felec            = _mm_mul_pd(qq23,_mm_sub_pd(_mm_mul_pd(rinv23,rinvsq23),krf2));
 407
 408             /* Update potential sum for this i atom from the interaction with this j atom. */
 409             velecsum         = _mm_add_pd(velecsum,velec);
 410
 411             fscal            = felec;
 412
 413             /* Calculate temporary vectorial force */
 414             tx               = _mm_mul_pd(fscal,dx23);
 415             ty               = _mm_mul_pd(fscal,dy23);
 416             tz               = _mm_mul_pd(fscal,dz23);
 417
 418             /* Update vectorial force */
 419             fix2             = _mm_add_pd(fix2,tx);
 420             fiy2             = _mm_add_pd(fiy2,ty);
 421             fiz2             = _mm_add_pd(fiz2,tz);
 422
 423             fjx3             = _mm_add_pd(fjx3,tx);
 424             fjy3             = _mm_add_pd(fjy3,ty);
 425             fjz3             = _mm_add_pd(fjz3,tz);
 426
 427             /**************************
 428              * CALCULATE INTERACTIONS *
 429              **************************/
 430
 431             /* REACTION-FIELD ELECTROSTATICS */
 432             velec            = _mm_mul_pd(qq31,_mm_sub_pd(_mm_add_pd(rinv31,_mm_mul_pd(krf,rsq31)),crf));
 433             felec            = _mm_mul_pd(qq31,_mm_sub_pd(_mm_mul_pd(rinv31,rinvsq31),krf2));
 434
 435             /* Update potential sum for this i atom from the interaction with this j atom. */
 436             velecsum         = _mm_add_pd(velecsum,velec);
 437
 438             fscal            = felec;
 439
 440             /* Calculate temporary vectorial force */
 441             tx               = _mm_mul_pd(fscal,dx31);
 442             ty               = _mm_mul_pd(fscal,dy31);
 443             tz               = _mm_mul_pd(fscal,dz31);
 444
 445             /* Update vectorial force */
 446             fix3             = _mm_add_pd(fix3,tx);
 447             fiy3             = _mm_add_pd(fiy3,ty);
 448             fiz3             = _mm_add_pd(fiz3,tz);
 449
 450             fjx1             = _mm_add_pd(fjx1,tx);
 451             fjy1             = _mm_add_pd(fjy1,ty);
 452             fjz1             = _mm_add_pd(fjz1,tz);
 453
 454             /**************************
 455              * CALCULATE INTERACTIONS *
 456              **************************/
 457
 458             /* REACTION-FIELD ELECTROSTATICS */
 459             velec            = _mm_mul_pd(qq32,_mm_sub_pd(_mm_add_pd(rinv32,_mm_mul_pd(krf,rsq32)),crf));
 460             felec            = _mm_mul_pd(qq32,_mm_sub_pd(_mm_mul_pd(rinv32,rinvsq32),krf2));
 461
 462             /* Update potential sum for this i atom from the interaction with this j atom. */
 463             velecsum         = _mm_add_pd(velecsum,velec);
 464
 465             fscal            = felec;
 466
 467             /* Calculate temporary vectorial force */
 468             tx               = _mm_mul_pd(fscal,dx32);
 469             ty               = _mm_mul_pd(fscal,dy32);
 470             tz               = _mm_mul_pd(fscal,dz32);
 471
 472             /* Update vectorial force */
 473             fix3             = _mm_add_pd(fix3,tx);
 474             fiy3             = _mm_add_pd(fiy3,ty);
 475             fiz3             = _mm_add_pd(fiz3,tz);
 476
 477             fjx2             = _mm_add_pd(fjx2,tx);
 478             fjy2             = _mm_add_pd(fjy2,ty);
 479             fjz2             = _mm_add_pd(fjz2,tz);
 480
 481             /**************************
 482              * CALCULATE INTERACTIONS *
 483              **************************/
 484
 485             /* REACTION-FIELD ELECTROSTATICS */
 486             velec            = _mm_mul_pd(qq33,_mm_sub_pd(_mm_add_pd(rinv33,_mm_mul_pd(krf,rsq33)),crf));
 487             felec            = _mm_mul_pd(qq33,_mm_sub_pd(_mm_mul_pd(rinv33,rinvsq33),krf2));
 488
 489             /* Update potential sum for this i atom from the interaction with this j atom. */
 490             velecsum         = _mm_add_pd(velecsum,velec);
 491
 492             fscal            = felec;
 493
 494             /* Calculate temporary vectorial force */
 495             tx               = _mm_mul_pd(fscal,dx33);
 496             ty               = _mm_mul_pd(fscal,dy33);
 497             tz               = _mm_mul_pd(fscal,dz33);
 498
 499             /* Update vectorial force */
 500             fix3             = _mm_add_pd(fix3,tx);
 501             fiy3             = _mm_add_pd(fiy3,ty);
 502             fiz3             = _mm_add_pd(fiz3,tz);
 503
 504             fjx3             = _mm_add_pd(fjx3,tx);
 505             fjy3             = _mm_add_pd(fjy3,ty);
 506             fjz3             = _mm_add_pd(fjz3,tz);
 507
 508             gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA+DIM,f+j_coord_offsetB+DIM,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 509
 510             /* Inner loop uses 288 flops */
 511         }
 512
 513         if(jidx<j_index_end)
 514         {
 515
 516             jnrA             = jjnr[jidx];
 517             j_coord_offsetA  = DIM*jnrA;
 518
 519             /* load j atom coordinates */
 520             gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA+DIM,
 521                                               &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
 522
 523             /* Calculate displacement vector */
 524             dx11             = _mm_sub_pd(ix1,jx1);
 525             dy11             = _mm_sub_pd(iy1,jy1);
 526             dz11             = _mm_sub_pd(iz1,jz1);
 527             dx12             = _mm_sub_pd(ix1,jx2);
 528             dy12             = _mm_sub_pd(iy1,jy2);
 529             dz12             = _mm_sub_pd(iz1,jz2);
 530             dx13             = _mm_sub_pd(ix1,jx3);
 531             dy13             = _mm_sub_pd(iy1,jy3);
 532             dz13             = _mm_sub_pd(iz1,jz3);
 533             dx21             = _mm_sub_pd(ix2,jx1);
 534             dy21             = _mm_sub_pd(iy2,jy1);
 535             dz21             = _mm_sub_pd(iz2,jz1);
 536             dx22             = _mm_sub_pd(ix2,jx2);
 537             dy22             = _mm_sub_pd(iy2,jy2);
 538             dz22             = _mm_sub_pd(iz2,jz2);
 539             dx23             = _mm_sub_pd(ix2,jx3);
 540             dy23             = _mm_sub_pd(iy2,jy3);
 541             dz23             = _mm_sub_pd(iz2,jz3);
 542             dx31             = _mm_sub_pd(ix3,jx1);
 543             dy31             = _mm_sub_pd(iy3,jy1);
 544             dz31             = _mm_sub_pd(iz3,jz1);
 545             dx32             = _mm_sub_pd(ix3,jx2);
 546             dy32             = _mm_sub_pd(iy3,jy2);
 547             dz32             = _mm_sub_pd(iz3,jz2);
 548             dx33             = _mm_sub_pd(ix3,jx3);
 549             dy33             = _mm_sub_pd(iy3,jy3);
 550             dz33             = _mm_sub_pd(iz3,jz3);
 551
 552             /* Calculate squared distance and things based on it */
 553             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 554             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 555             rsq13            = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
 556             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 557             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 558             rsq23            = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
 559             rsq31            = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
 560             rsq32            = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
 561             rsq33            = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
 562
 563             rinv11           = sse41_invsqrt_d(rsq11);
 564             rinv12           = sse41_invsqrt_d(rsq12);
 565             rinv13           = sse41_invsqrt_d(rsq13);
 566             rinv21           = sse41_invsqrt_d(rsq21);
 567             rinv22           = sse41_invsqrt_d(rsq22);
 568             rinv23           = sse41_invsqrt_d(rsq23);
 569             rinv31           = sse41_invsqrt_d(rsq31);
 570             rinv32           = sse41_invsqrt_d(rsq32);
 571             rinv33           = sse41_invsqrt_d(rsq33);
 572
 573             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
 574             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
 575             rinvsq13         = _mm_mul_pd(rinv13,rinv13);
 576             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
 577             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
 578             rinvsq23         = _mm_mul_pd(rinv23,rinv23);
 579             rinvsq31         = _mm_mul_pd(rinv31,rinv31);
 580             rinvsq32         = _mm_mul_pd(rinv32,rinv32);
 581             rinvsq33         = _mm_mul_pd(rinv33,rinv33);
 582
 583             fjx1             = _mm_setzero_pd();
 584             fjy1             = _mm_setzero_pd();
 585             fjz1             = _mm_setzero_pd();
 586             fjx2             = _mm_setzero_pd();
 587             fjy2             = _mm_setzero_pd();
 588             fjz2             = _mm_setzero_pd();
 589             fjx3             = _mm_setzero_pd();
 590             fjy3             = _mm_setzero_pd();
 591             fjz3             = _mm_setzero_pd();
 592
 593             /**************************
 594              * CALCULATE INTERACTIONS *
 595              **************************/
 596
 597             /* REACTION-FIELD ELECTROSTATICS */
 598             velec            = _mm_mul_pd(qq11,_mm_sub_pd(_mm_add_pd(rinv11,_mm_mul_pd(krf,rsq11)),crf));
 599             felec            = _mm_mul_pd(qq11,_mm_sub_pd(_mm_mul_pd(rinv11,rinvsq11),krf2));
 600
 601             /* Update potential sum for this i atom from the interaction with this j atom. */
 602             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 603             velecsum         = _mm_add_pd(velecsum,velec);
 604
 605             fscal            = felec;
 606
 607             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 608
 609             /* Calculate temporary vectorial force */
 610             tx               = _mm_mul_pd(fscal,dx11);
 611             ty               = _mm_mul_pd(fscal,dy11);
 612             tz               = _mm_mul_pd(fscal,dz11);
 613
 614             /* Update vectorial force */
 615             fix1             = _mm_add_pd(fix1,tx);
 616             fiy1             = _mm_add_pd(fiy1,ty);
 617             fiz1             = _mm_add_pd(fiz1,tz);
 618
 619             fjx1             = _mm_add_pd(fjx1,tx);
 620             fjy1             = _mm_add_pd(fjy1,ty);
 621             fjz1             = _mm_add_pd(fjz1,tz);
 622
 623             /**************************
 624              * CALCULATE INTERACTIONS *
 625              **************************/
 626
 627             /* REACTION-FIELD ELECTROSTATICS */
 628             velec            = _mm_mul_pd(qq12,_mm_sub_pd(_mm_add_pd(rinv12,_mm_mul_pd(krf,rsq12)),crf));
 629             felec            = _mm_mul_pd(qq12,_mm_sub_pd(_mm_mul_pd(rinv12,rinvsq12),krf2));
 630
 631             /* Update potential sum for this i atom from the interaction with this j atom. */
 632             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 633             velecsum         = _mm_add_pd(velecsum,velec);
 634
 635             fscal            = felec;
 636
 637             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 638
 639             /* Calculate temporary vectorial force */
 640             tx               = _mm_mul_pd(fscal,dx12);
 641             ty               = _mm_mul_pd(fscal,dy12);
 642             tz               = _mm_mul_pd(fscal,dz12);
 643
 644             /* Update vectorial force */
 645             fix1             = _mm_add_pd(fix1,tx);
 646             fiy1             = _mm_add_pd(fiy1,ty);
 647             fiz1             = _mm_add_pd(fiz1,tz);
 648
 649             fjx2             = _mm_add_pd(fjx2,tx);
 650             fjy2             = _mm_add_pd(fjy2,ty);
 651             fjz2             = _mm_add_pd(fjz2,tz);
 652
 653             /**************************
 654              * CALCULATE INTERACTIONS *
 655              **************************/
 656
 657             /* REACTION-FIELD ELECTROSTATICS */
 658             velec            = _mm_mul_pd(qq13,_mm_sub_pd(_mm_add_pd(rinv13,_mm_mul_pd(krf,rsq13)),crf));
 659             felec            = _mm_mul_pd(qq13,_mm_sub_pd(_mm_mul_pd(rinv13,rinvsq13),krf2));
 660
 661             /* Update potential sum for this i atom from the interaction with this j atom. */
 662             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 663             velecsum         = _mm_add_pd(velecsum,velec);
 664
 665             fscal            = felec;
 666
 667             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 668
 669             /* Calculate temporary vectorial force */
 670             tx               = _mm_mul_pd(fscal,dx13);
 671             ty               = _mm_mul_pd(fscal,dy13);
 672             tz               = _mm_mul_pd(fscal,dz13);
 673
 674             /* Update vectorial force */
 675             fix1             = _mm_add_pd(fix1,tx);
 676             fiy1             = _mm_add_pd(fiy1,ty);
 677             fiz1             = _mm_add_pd(fiz1,tz);
 678
 679             fjx3             = _mm_add_pd(fjx3,tx);
 680             fjy3             = _mm_add_pd(fjy3,ty);
 681             fjz3             = _mm_add_pd(fjz3,tz);
 682
 683             /**************************
 684              * CALCULATE INTERACTIONS *
 685              **************************/
 686
 687             /* REACTION-FIELD ELECTROSTATICS */
 688             velec            = _mm_mul_pd(qq21,_mm_sub_pd(_mm_add_pd(rinv21,_mm_mul_pd(krf,rsq21)),crf));
 689             felec            = _mm_mul_pd(qq21,_mm_sub_pd(_mm_mul_pd(rinv21,rinvsq21),krf2));
 690
 691             /* Update potential sum for this i atom from the interaction with this j atom. */
 692             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 693             velecsum         = _mm_add_pd(velecsum,velec);
 694
 695             fscal            = felec;
 696
 697             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 698
 699             /* Calculate temporary vectorial force */
 700             tx               = _mm_mul_pd(fscal,dx21);
 701             ty               = _mm_mul_pd(fscal,dy21);
 702             tz               = _mm_mul_pd(fscal,dz21);
 703
 704             /* Update vectorial force */
 705             fix2             = _mm_add_pd(fix2,tx);
 706             fiy2             = _mm_add_pd(fiy2,ty);
 707             fiz2             = _mm_add_pd(fiz2,tz);
 708
 709             fjx1             = _mm_add_pd(fjx1,tx);
 710             fjy1             = _mm_add_pd(fjy1,ty);
 711             fjz1             = _mm_add_pd(fjz1,tz);
 712
 713             /**************************
 714              * CALCULATE INTERACTIONS *
 715              **************************/
 716
 717             /* REACTION-FIELD ELECTROSTATICS */
 718             velec            = _mm_mul_pd(qq22,_mm_sub_pd(_mm_add_pd(rinv22,_mm_mul_pd(krf,rsq22)),crf));
 719             felec            = _mm_mul_pd(qq22,_mm_sub_pd(_mm_mul_pd(rinv22,rinvsq22),krf2));
 720
 721             /* Update potential sum for this i atom from the interaction with this j atom. */
 722             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 723             velecsum         = _mm_add_pd(velecsum,velec);
 724
 725             fscal            = felec;
 726
 727             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 728
 729             /* Calculate temporary vectorial force */
 730             tx               = _mm_mul_pd(fscal,dx22);
 731             ty               = _mm_mul_pd(fscal,dy22);
 732             tz               = _mm_mul_pd(fscal,dz22);
 733
 734             /* Update vectorial force */
 735             fix2             = _mm_add_pd(fix2,tx);
 736             fiy2             = _mm_add_pd(fiy2,ty);
 737             fiz2             = _mm_add_pd(fiz2,tz);
 738
 739             fjx2             = _mm_add_pd(fjx2,tx);
 740             fjy2             = _mm_add_pd(fjy2,ty);
 741             fjz2             = _mm_add_pd(fjz2,tz);
 742
 743             /**************************
 744              * CALCULATE INTERACTIONS *
 745              **************************/
 746
 747             /* REACTION-FIELD ELECTROSTATICS */
 748             velec            = _mm_mul_pd(qq23,_mm_sub_pd(_mm_add_pd(rinv23,_mm_mul_pd(krf,rsq23)),crf));
 749             felec            = _mm_mul_pd(qq23,_mm_sub_pd(_mm_mul_pd(rinv23,rinvsq23),krf2));
 750
 751             /* Update potential sum for this i atom from the interaction with this j atom. */
 752             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 753             velecsum         = _mm_add_pd(velecsum,velec);
 754
 755             fscal            = felec;
 756
 757             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 758
 759             /* Calculate temporary vectorial force */
 760             tx               = _mm_mul_pd(fscal,dx23);
 761             ty               = _mm_mul_pd(fscal,dy23);
 762             tz               = _mm_mul_pd(fscal,dz23);
 763
 764             /* Update vectorial force */
 765             fix2             = _mm_add_pd(fix2,tx);
 766             fiy2             = _mm_add_pd(fiy2,ty);
 767             fiz2             = _mm_add_pd(fiz2,tz);
 768
 769             fjx3             = _mm_add_pd(fjx3,tx);
 770             fjy3             = _mm_add_pd(fjy3,ty);
 771             fjz3             = _mm_add_pd(fjz3,tz);
 772
 773             /**************************
 774              * CALCULATE INTERACTIONS *
 775              **************************/
 776
 777             /* REACTION-FIELD ELECTROSTATICS */
 778             velec            = _mm_mul_pd(qq31,_mm_sub_pd(_mm_add_pd(rinv31,_mm_mul_pd(krf,rsq31)),crf));
 779             felec            = _mm_mul_pd(qq31,_mm_sub_pd(_mm_mul_pd(rinv31,rinvsq31),krf2));
 780
 781             /* Update potential sum for this i atom from the interaction with this j atom. */
 782             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 783             velecsum         = _mm_add_pd(velecsum,velec);
 784
 785             fscal            = felec;
 786
 787             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 788
 789             /* Calculate temporary vectorial force */
 790             tx               = _mm_mul_pd(fscal,dx31);
 791             ty               = _mm_mul_pd(fscal,dy31);
 792             tz               = _mm_mul_pd(fscal,dz31);
 793
 794             /* Update vectorial force */
 795             fix3             = _mm_add_pd(fix3,tx);
 796             fiy3             = _mm_add_pd(fiy3,ty);
 797             fiz3             = _mm_add_pd(fiz3,tz);
 798
 799             fjx1             = _mm_add_pd(fjx1,tx);
 800             fjy1             = _mm_add_pd(fjy1,ty);
 801             fjz1             = _mm_add_pd(fjz1,tz);
 802
 803             /**************************
 804              * CALCULATE INTERACTIONS *
 805              **************************/
 806
 807             /* REACTION-FIELD ELECTROSTATICS */
 808             velec            = _mm_mul_pd(qq32,_mm_sub_pd(_mm_add_pd(rinv32,_mm_mul_pd(krf,rsq32)),crf));
 809             felec            = _mm_mul_pd(qq32,_mm_sub_pd(_mm_mul_pd(rinv32,rinvsq32),krf2));
 810
 811             /* Update potential sum for this i atom from the interaction with this j atom. */
 812             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 813             velecsum         = _mm_add_pd(velecsum,velec);
 814
 815             fscal            = felec;
 816
 817             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 818
 819             /* Calculate temporary vectorial force */
 820             tx               = _mm_mul_pd(fscal,dx32);
 821             ty               = _mm_mul_pd(fscal,dy32);
 822             tz               = _mm_mul_pd(fscal,dz32);
 823
 824             /* Update vectorial force */
 825             fix3             = _mm_add_pd(fix3,tx);
 826             fiy3             = _mm_add_pd(fiy3,ty);
 827             fiz3             = _mm_add_pd(fiz3,tz);
 828
 829             fjx2             = _mm_add_pd(fjx2,tx);
 830             fjy2             = _mm_add_pd(fjy2,ty);
 831             fjz2             = _mm_add_pd(fjz2,tz);
 832
 833             /**************************
 834              * CALCULATE INTERACTIONS *
 835              **************************/
 836
 837             /* REACTION-FIELD ELECTROSTATICS */
 838             velec            = _mm_mul_pd(qq33,_mm_sub_pd(_mm_add_pd(rinv33,_mm_mul_pd(krf,rsq33)),crf));
 839             felec            = _mm_mul_pd(qq33,_mm_sub_pd(_mm_mul_pd(rinv33,rinvsq33),krf2));
 840
 841             /* Update potential sum for this i atom from the interaction with this j atom. */
 842             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 843             velecsum         = _mm_add_pd(velecsum,velec);
 844
 845             fscal            = felec;
 846
 847             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 848
 849             /* Calculate temporary vectorial force */
 850             tx               = _mm_mul_pd(fscal,dx33);
 851             ty               = _mm_mul_pd(fscal,dy33);
 852             tz               = _mm_mul_pd(fscal,dz33);
 853
 854             /* Update vectorial force */
 855             fix3             = _mm_add_pd(fix3,tx);
 856             fiy3             = _mm_add_pd(fiy3,ty);
 857             fiz3             = _mm_add_pd(fiz3,tz);
 858
 859             fjx3             = _mm_add_pd(fjx3,tx);
 860             fjy3             = _mm_add_pd(fjy3,ty);
 861             fjz3             = _mm_add_pd(fjz3,tz);
 862
 863             gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA+DIM,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 864
 865             /* Inner loop uses 288 flops */
 866         }
 867
 868         /* End of innermost loop */
 869
 870         gmx_mm_update_iforce_3atom_swizzle_pd(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 871                                               f+i_coord_offset+DIM,fshift+i_shift_offset);
 872
 873         ggid                        = gid[iidx];
 874         /* Update potential energies */
 875         gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
 876
 877         /* Increment number of inner iterations */
 878         inneriter                  += j_index_end - j_index_start;
 879
 880         /* Outer loop uses 19 flops */
 881     }
 882
 883     /* Increment number of outer iterations */
 884     outeriter        += nri;
 885
 886     /* Update outer/inner flops */
 887
 888     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4W4_VF,outeriter*19 + inneriter*288);
 889 }
 890 /*
 891  * Gromacs nonbonded kernel:   nb_kernel_ElecRF_VdwNone_GeomW4W4_F_sse4_1_double
 892  * Electrostatics interaction: ReactionField
 893  * VdW interaction:            None
 894  * Geometry:                   Water4-Water4
 895  * Calculate force/pot:        Force
 896  */
 897 void
 898 nb_kernel_ElecRF_VdwNone_GeomW4W4_F_sse4_1_double
 899                     (t_nblist                    * gmx_restrict       nlist,
 900                      rvec                        * gmx_restrict          xx,
 901                      rvec                        * gmx_restrict          ff,
 902                      struct t_forcerec           * gmx_restrict          fr,
 903                      t_mdatoms                   * gmx_restrict     mdatoms,
 904                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 905                      t_nrnb                      * gmx_restrict        nrnb)
 906 {
 907     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 908      * just 0 for non-waters.
 909      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
 910      * jnr indices corresponding to data put in the four positions in the SIMD register.
 911      */
 912     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 913     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 914     int              jnrA,jnrB;
 915     int              j_coord_offsetA,j_coord_offsetB;
 916     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 917     real             rcutoff_scalar;
 918     real             *shiftvec,*fshift,*x,*f;
 919     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 920     int              vdwioffset1;
 921     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 922     int              vdwioffset2;
 923     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 924     int              vdwioffset3;
 925     __m128d          ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
 926     int              vdwjidx1A,vdwjidx1B;
 927     __m128d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
 928     int              vdwjidx2A,vdwjidx2B;
 929     __m128d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
 930     int              vdwjidx3A,vdwjidx3B;
 931     __m128d          jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
 932     __m128d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
 933     __m128d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
 934     __m128d          dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
 935     __m128d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 936     __m128d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 937     __m128d          dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
 938     __m128d          dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
 939     __m128d          dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
 940     __m128d          dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
 941     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
 942     real             *charge;
 943     __m128d          dummy_mask,cutoff_mask;
 944     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 945     __m128d          one     = _mm_set1_pd(1.0);
 946     __m128d          two     = _mm_set1_pd(2.0);
 947     x                = xx[0];
 948     f                = ff[0];
 949
 950     nri              = nlist->nri;
 951     iinr             = nlist->iinr;
 952     jindex           = nlist->jindex;
 953     jjnr             = nlist->jjnr;
 954     shiftidx         = nlist->shift;
 955     gid              = nlist->gid;
 956     shiftvec         = fr->shift_vec[0];
 957     fshift           = fr->fshift[0];
 958     facel            = _mm_set1_pd(fr->ic->epsfac);
 959     charge           = mdatoms->chargeA;
 960     krf              = _mm_set1_pd(fr->ic->k_rf);
 961     krf2             = _mm_set1_pd(fr->ic->k_rf*2.0);
 962     crf              = _mm_set1_pd(fr->ic->c_rf);
 963
 964     /* Setup water-specific parameters */
 965     inr              = nlist->iinr[0];
 966     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
 967     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
 968     iq3              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
 969
 970     jq1              = _mm_set1_pd(charge[inr+1]);
 971     jq2              = _mm_set1_pd(charge[inr+2]);
 972     jq3              = _mm_set1_pd(charge[inr+3]);
 973     qq11             = _mm_mul_pd(iq1,jq1);
 974     qq12             = _mm_mul_pd(iq1,jq2);
 975     qq13             = _mm_mul_pd(iq1,jq3);
 976     qq21             = _mm_mul_pd(iq2,jq1);
 977     qq22             = _mm_mul_pd(iq2,jq2);
 978     qq23             = _mm_mul_pd(iq2,jq3);
 979     qq31             = _mm_mul_pd(iq3,jq1);
 980     qq32             = _mm_mul_pd(iq3,jq2);
 981     qq33             = _mm_mul_pd(iq3,jq3);
 982
 983     /* Avoid stupid compiler warnings */
 984     jnrA = jnrB = 0;
 985     j_coord_offsetA = 0;
 986     j_coord_offsetB = 0;
 987
 988     outeriter        = 0;
 989     inneriter        = 0;
 990
 991     /* Start outer loop over neighborlists */
 992     for(iidx=0; iidx<nri; iidx++)
 993     {
 994         /* Load shift vector for this list */
 995         i_shift_offset   = DIM*shiftidx[iidx];
 996
 997         /* Load limits for loop over neighbors */
 998         j_index_start    = jindex[iidx];
 999         j_index_end      = jindex[iidx+1];
1000
1001         /* Get outer coordinate index */
1002         inr              = iinr[iidx];
1003         i_coord_offset   = DIM*inr;
1004
1005         /* Load i particle coords and add shift vector */
1006         gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
1007                                                  &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1008
1009         fix1             = _mm_setzero_pd();
1010         fiy1             = _mm_setzero_pd();
1011         fiz1             = _mm_setzero_pd();
1012         fix2             = _mm_setzero_pd();
1013         fiy2             = _mm_setzero_pd();
1014         fiz2             = _mm_setzero_pd();
1015         fix3             = _mm_setzero_pd();
1016         fiy3             = _mm_setzero_pd();
1017         fiz3             = _mm_setzero_pd();
1018
1019         /* Start inner kernel loop */
1020         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
1021         {
1022
1023             /* Get j neighbor index, and coordinate index */
1024             jnrA             = jjnr[jidx];
1025             jnrB             = jjnr[jidx+1];
1026             j_coord_offsetA  = DIM*jnrA;
1027             j_coord_offsetB  = DIM*jnrB;
1028
1029             /* load j atom coordinates */
1030             gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
1031                                               &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
1032
1033             /* Calculate displacement vector */
1034             dx11             = _mm_sub_pd(ix1,jx1);
1035             dy11             = _mm_sub_pd(iy1,jy1);
1036             dz11             = _mm_sub_pd(iz1,jz1);
1037             dx12             = _mm_sub_pd(ix1,jx2);
1038             dy12             = _mm_sub_pd(iy1,jy2);
1039             dz12             = _mm_sub_pd(iz1,jz2);
1040             dx13             = _mm_sub_pd(ix1,jx3);
1041             dy13             = _mm_sub_pd(iy1,jy3);
1042             dz13             = _mm_sub_pd(iz1,jz3);
1043             dx21             = _mm_sub_pd(ix2,jx1);
1044             dy21             = _mm_sub_pd(iy2,jy1);
1045             dz21             = _mm_sub_pd(iz2,jz1);
1046             dx22             = _mm_sub_pd(ix2,jx2);
1047             dy22             = _mm_sub_pd(iy2,jy2);
1048             dz22             = _mm_sub_pd(iz2,jz2);
1049             dx23             = _mm_sub_pd(ix2,jx3);
1050             dy23             = _mm_sub_pd(iy2,jy3);
1051             dz23             = _mm_sub_pd(iz2,jz3);
1052             dx31             = _mm_sub_pd(ix3,jx1);
1053             dy31             = _mm_sub_pd(iy3,jy1);
1054             dz31             = _mm_sub_pd(iz3,jz1);
1055             dx32             = _mm_sub_pd(ix3,jx2);
1056             dy32             = _mm_sub_pd(iy3,jy2);
1057             dz32             = _mm_sub_pd(iz3,jz2);
1058             dx33             = _mm_sub_pd(ix3,jx3);
1059             dy33             = _mm_sub_pd(iy3,jy3);
1060             dz33             = _mm_sub_pd(iz3,jz3);
1061
1062             /* Calculate squared distance and things based on it */
1063             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1064             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1065             rsq13            = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
1066             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1067             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1068             rsq23            = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
1069             rsq31            = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
1070             rsq32            = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
1071             rsq33            = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
1072
1073             rinv11           = sse41_invsqrt_d(rsq11);
1074             rinv12           = sse41_invsqrt_d(rsq12);
1075             rinv13           = sse41_invsqrt_d(rsq13);
1076             rinv21           = sse41_invsqrt_d(rsq21);
1077             rinv22           = sse41_invsqrt_d(rsq22);
1078             rinv23           = sse41_invsqrt_d(rsq23);
1079             rinv31           = sse41_invsqrt_d(rsq31);
1080             rinv32           = sse41_invsqrt_d(rsq32);
1081             rinv33           = sse41_invsqrt_d(rsq33);
1082
1083             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
1084             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
1085             rinvsq13         = _mm_mul_pd(rinv13,rinv13);
1086             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
1087             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
1088             rinvsq23         = _mm_mul_pd(rinv23,rinv23);
1089             rinvsq31         = _mm_mul_pd(rinv31,rinv31);
1090             rinvsq32         = _mm_mul_pd(rinv32,rinv32);
1091             rinvsq33         = _mm_mul_pd(rinv33,rinv33);
1092
1093             fjx1             = _mm_setzero_pd();
1094             fjy1             = _mm_setzero_pd();
1095             fjz1             = _mm_setzero_pd();
1096             fjx2             = _mm_setzero_pd();
1097             fjy2             = _mm_setzero_pd();
1098             fjz2             = _mm_setzero_pd();
1099             fjx3             = _mm_setzero_pd();
1100             fjy3             = _mm_setzero_pd();
1101             fjz3             = _mm_setzero_pd();
1102
1103             /**************************
1104              * CALCULATE INTERACTIONS *
1105              **************************/
1106
1107             /* REACTION-FIELD ELECTROSTATICS */
1108             felec            = _mm_mul_pd(qq11,_mm_sub_pd(_mm_mul_pd(rinv11,rinvsq11),krf2));
1109
1110             fscal            = felec;
1111
1112             /* Calculate temporary vectorial force */
1113             tx               = _mm_mul_pd(fscal,dx11);
1114             ty               = _mm_mul_pd(fscal,dy11);
1115             tz               = _mm_mul_pd(fscal,dz11);
1116
1117             /* Update vectorial force */
1118             fix1             = _mm_add_pd(fix1,tx);
1119             fiy1             = _mm_add_pd(fiy1,ty);
1120             fiz1             = _mm_add_pd(fiz1,tz);
1121
1122             fjx1             = _mm_add_pd(fjx1,tx);
1123             fjy1             = _mm_add_pd(fjy1,ty);
1124             fjz1             = _mm_add_pd(fjz1,tz);
1125
1126             /**************************
1127              * CALCULATE INTERACTIONS *
1128              **************************/
1129
1130             /* REACTION-FIELD ELECTROSTATICS */
1131             felec            = _mm_mul_pd(qq12,_mm_sub_pd(_mm_mul_pd(rinv12,rinvsq12),krf2));
1132
1133             fscal            = felec;
1134
1135             /* Calculate temporary vectorial force */
1136             tx               = _mm_mul_pd(fscal,dx12);
1137             ty               = _mm_mul_pd(fscal,dy12);
1138             tz               = _mm_mul_pd(fscal,dz12);
1139
1140             /* Update vectorial force */
1141             fix1             = _mm_add_pd(fix1,tx);
1142             fiy1             = _mm_add_pd(fiy1,ty);
1143             fiz1             = _mm_add_pd(fiz1,tz);
1144
1145             fjx2             = _mm_add_pd(fjx2,tx);
1146             fjy2             = _mm_add_pd(fjy2,ty);
1147             fjz2             = _mm_add_pd(fjz2,tz);
1148
1149             /**************************
1150              * CALCULATE INTERACTIONS *
1151              **************************/
1152
1153             /* REACTION-FIELD ELECTROSTATICS */
1154             felec            = _mm_mul_pd(qq13,_mm_sub_pd(_mm_mul_pd(rinv13,rinvsq13),krf2));
1155
1156             fscal            = felec;
1157
1158             /* Calculate temporary vectorial force */
1159             tx               = _mm_mul_pd(fscal,dx13);
1160             ty               = _mm_mul_pd(fscal,dy13);
1161             tz               = _mm_mul_pd(fscal,dz13);
1162
1163             /* Update vectorial force */
1164             fix1             = _mm_add_pd(fix1,tx);
1165             fiy1             = _mm_add_pd(fiy1,ty);
1166             fiz1             = _mm_add_pd(fiz1,tz);
1167
1168             fjx3             = _mm_add_pd(fjx3,tx);
1169             fjy3             = _mm_add_pd(fjy3,ty);
1170             fjz3             = _mm_add_pd(fjz3,tz);
1171
1172             /**************************
1173              * CALCULATE INTERACTIONS *
1174              **************************/
1175
1176             /* REACTION-FIELD ELECTROSTATICS */
1177             felec            = _mm_mul_pd(qq21,_mm_sub_pd(_mm_mul_pd(rinv21,rinvsq21),krf2));
1178
1179             fscal            = felec;
1180
1181             /* Calculate temporary vectorial force */
1182             tx               = _mm_mul_pd(fscal,dx21);
1183             ty               = _mm_mul_pd(fscal,dy21);
1184             tz               = _mm_mul_pd(fscal,dz21);
1185
1186             /* Update vectorial force */
1187             fix2             = _mm_add_pd(fix2,tx);
1188             fiy2             = _mm_add_pd(fiy2,ty);
1189             fiz2             = _mm_add_pd(fiz2,tz);
1190
1191             fjx1             = _mm_add_pd(fjx1,tx);
1192             fjy1             = _mm_add_pd(fjy1,ty);
1193             fjz1             = _mm_add_pd(fjz1,tz);
1194
1195             /**************************
1196              * CALCULATE INTERACTIONS *
1197              **************************/
1198
1199             /* REACTION-FIELD ELECTROSTATICS */
1200             felec            = _mm_mul_pd(qq22,_mm_sub_pd(_mm_mul_pd(rinv22,rinvsq22),krf2));
1201
1202             fscal            = felec;
1203
1204             /* Calculate temporary vectorial force */
1205             tx               = _mm_mul_pd(fscal,dx22);
1206             ty               = _mm_mul_pd(fscal,dy22);
1207             tz               = _mm_mul_pd(fscal,dz22);
1208
1209             /* Update vectorial force */
1210             fix2             = _mm_add_pd(fix2,tx);
1211             fiy2             = _mm_add_pd(fiy2,ty);
1212             fiz2             = _mm_add_pd(fiz2,tz);
1213
1214             fjx2             = _mm_add_pd(fjx2,tx);
1215             fjy2             = _mm_add_pd(fjy2,ty);
1216             fjz2             = _mm_add_pd(fjz2,tz);
1217
1218             /**************************
1219              * CALCULATE INTERACTIONS *
1220              **************************/
1221
1222             /* REACTION-FIELD ELECTROSTATICS */
1223             felec            = _mm_mul_pd(qq23,_mm_sub_pd(_mm_mul_pd(rinv23,rinvsq23),krf2));
1224
1225             fscal            = felec;
1226
1227             /* Calculate temporary vectorial force */
1228             tx               = _mm_mul_pd(fscal,dx23);
1229             ty               = _mm_mul_pd(fscal,dy23);
1230             tz               = _mm_mul_pd(fscal,dz23);
1231
1232             /* Update vectorial force */
1233             fix2             = _mm_add_pd(fix2,tx);
1234             fiy2             = _mm_add_pd(fiy2,ty);
1235             fiz2             = _mm_add_pd(fiz2,tz);
1236
1237             fjx3             = _mm_add_pd(fjx3,tx);
1238             fjy3             = _mm_add_pd(fjy3,ty);
1239             fjz3             = _mm_add_pd(fjz3,tz);
1240
1241             /**************************
1242              * CALCULATE INTERACTIONS *
1243              **************************/
1244
1245             /* REACTION-FIELD ELECTROSTATICS */
1246             felec            = _mm_mul_pd(qq31,_mm_sub_pd(_mm_mul_pd(rinv31,rinvsq31),krf2));
1247
1248             fscal            = felec;
1249
1250             /* Calculate temporary vectorial force */
1251             tx               = _mm_mul_pd(fscal,dx31);
1252             ty               = _mm_mul_pd(fscal,dy31);
1253             tz               = _mm_mul_pd(fscal,dz31);
1254
1255             /* Update vectorial force */
1256             fix3             = _mm_add_pd(fix3,tx);
1257             fiy3             = _mm_add_pd(fiy3,ty);
1258             fiz3             = _mm_add_pd(fiz3,tz);
1259
1260             fjx1             = _mm_add_pd(fjx1,tx);
1261             fjy1             = _mm_add_pd(fjy1,ty);
1262             fjz1             = _mm_add_pd(fjz1,tz);
1263
1264             /**************************
1265              * CALCULATE INTERACTIONS *
1266              **************************/
1267
1268             /* REACTION-FIELD ELECTROSTATICS */
1269             felec            = _mm_mul_pd(qq32,_mm_sub_pd(_mm_mul_pd(rinv32,rinvsq32),krf2));
1270
1271             fscal            = felec;
1272
1273             /* Calculate temporary vectorial force */
1274             tx               = _mm_mul_pd(fscal,dx32);
1275             ty               = _mm_mul_pd(fscal,dy32);
1276             tz               = _mm_mul_pd(fscal,dz32);
1277
1278             /* Update vectorial force */
1279             fix3             = _mm_add_pd(fix3,tx);
1280             fiy3             = _mm_add_pd(fiy3,ty);
1281             fiz3             = _mm_add_pd(fiz3,tz);
1282
1283             fjx2             = _mm_add_pd(fjx2,tx);
1284             fjy2             = _mm_add_pd(fjy2,ty);
1285             fjz2             = _mm_add_pd(fjz2,tz);
1286
1287             /**************************
1288              * CALCULATE INTERACTIONS *
1289              **************************/
1290
1291             /* REACTION-FIELD ELECTROSTATICS */
1292             felec            = _mm_mul_pd(qq33,_mm_sub_pd(_mm_mul_pd(rinv33,rinvsq33),krf2));
1293
1294             fscal            = felec;
1295
1296             /* Calculate temporary vectorial force */
1297             tx               = _mm_mul_pd(fscal,dx33);
1298             ty               = _mm_mul_pd(fscal,dy33);
1299             tz               = _mm_mul_pd(fscal,dz33);
1300
1301             /* Update vectorial force */
1302             fix3             = _mm_add_pd(fix3,tx);
1303             fiy3             = _mm_add_pd(fiy3,ty);
1304             fiz3             = _mm_add_pd(fiz3,tz);
1305
1306             fjx3             = _mm_add_pd(fjx3,tx);
1307             fjy3             = _mm_add_pd(fjy3,ty);
1308             fjz3             = _mm_add_pd(fjz3,tz);
1309
1310             gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA+DIM,f+j_coord_offsetB+DIM,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1311
1312             /* Inner loop uses 243 flops */
1313         }
1314
1315         if(jidx<j_index_end)
1316         {
1317
1318             jnrA             = jjnr[jidx];
1319             j_coord_offsetA  = DIM*jnrA;
1320
1321             /* load j atom coordinates */
1322             gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA+DIM,
1323                                               &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
1324
1325             /* Calculate displacement vector */
1326             dx11             = _mm_sub_pd(ix1,jx1);
1327             dy11             = _mm_sub_pd(iy1,jy1);
1328             dz11             = _mm_sub_pd(iz1,jz1);
1329             dx12             = _mm_sub_pd(ix1,jx2);
1330             dy12             = _mm_sub_pd(iy1,jy2);
1331             dz12             = _mm_sub_pd(iz1,jz2);
1332             dx13             = _mm_sub_pd(ix1,jx3);
1333             dy13             = _mm_sub_pd(iy1,jy3);
1334             dz13             = _mm_sub_pd(iz1,jz3);
1335             dx21             = _mm_sub_pd(ix2,jx1);
1336             dy21             = _mm_sub_pd(iy2,jy1);
1337             dz21             = _mm_sub_pd(iz2,jz1);
1338             dx22             = _mm_sub_pd(ix2,jx2);
1339             dy22             = _mm_sub_pd(iy2,jy2);
1340             dz22             = _mm_sub_pd(iz2,jz2);
1341             dx23             = _mm_sub_pd(ix2,jx3);
1342             dy23             = _mm_sub_pd(iy2,jy3);
1343             dz23             = _mm_sub_pd(iz2,jz3);
1344             dx31             = _mm_sub_pd(ix3,jx1);
1345             dy31             = _mm_sub_pd(iy3,jy1);
1346             dz31             = _mm_sub_pd(iz3,jz1);
1347             dx32             = _mm_sub_pd(ix3,jx2);
1348             dy32             = _mm_sub_pd(iy3,jy2);
1349             dz32             = _mm_sub_pd(iz3,jz2);
1350             dx33             = _mm_sub_pd(ix3,jx3);
1351             dy33             = _mm_sub_pd(iy3,jy3);
1352             dz33             = _mm_sub_pd(iz3,jz3);
1353
1354             /* Calculate squared distance and things based on it */
1355             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1356             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1357             rsq13            = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
1358             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1359             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1360             rsq23            = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
1361             rsq31            = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
1362             rsq32            = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
1363             rsq33            = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
1364
1365             rinv11           = sse41_invsqrt_d(rsq11);
1366             rinv12           = sse41_invsqrt_d(rsq12);
1367             rinv13           = sse41_invsqrt_d(rsq13);
1368             rinv21           = sse41_invsqrt_d(rsq21);
1369             rinv22           = sse41_invsqrt_d(rsq22);
1370             rinv23           = sse41_invsqrt_d(rsq23);
1371             rinv31           = sse41_invsqrt_d(rsq31);
1372             rinv32           = sse41_invsqrt_d(rsq32);
1373             rinv33           = sse41_invsqrt_d(rsq33);
1374
1375             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
1376             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
1377             rinvsq13         = _mm_mul_pd(rinv13,rinv13);
1378             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
1379             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
1380             rinvsq23         = _mm_mul_pd(rinv23,rinv23);
1381             rinvsq31         = _mm_mul_pd(rinv31,rinv31);
1382             rinvsq32         = _mm_mul_pd(rinv32,rinv32);
1383             rinvsq33         = _mm_mul_pd(rinv33,rinv33);
1384
1385             fjx1             = _mm_setzero_pd();
1386             fjy1             = _mm_setzero_pd();
1387             fjz1             = _mm_setzero_pd();
1388             fjx2             = _mm_setzero_pd();
1389             fjy2             = _mm_setzero_pd();
1390             fjz2             = _mm_setzero_pd();
1391             fjx3             = _mm_setzero_pd();
1392             fjy3             = _mm_setzero_pd();
1393             fjz3             = _mm_setzero_pd();
1394
1395             /**************************
1396              * CALCULATE INTERACTIONS *
1397              **************************/
1398
1399             /* REACTION-FIELD ELECTROSTATICS */
1400             felec            = _mm_mul_pd(qq11,_mm_sub_pd(_mm_mul_pd(rinv11,rinvsq11),krf2));
1401
1402             fscal            = felec;
1403
1404             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1405
1406             /* Calculate temporary vectorial force */
1407             tx               = _mm_mul_pd(fscal,dx11);
1408             ty               = _mm_mul_pd(fscal,dy11);
1409             tz               = _mm_mul_pd(fscal,dz11);
1410
1411             /* Update vectorial force */
1412             fix1             = _mm_add_pd(fix1,tx);
1413             fiy1             = _mm_add_pd(fiy1,ty);
1414             fiz1             = _mm_add_pd(fiz1,tz);
1415
1416             fjx1             = _mm_add_pd(fjx1,tx);
1417             fjy1             = _mm_add_pd(fjy1,ty);
1418             fjz1             = _mm_add_pd(fjz1,tz);
1419
1420             /**************************
1421              * CALCULATE INTERACTIONS *
1422              **************************/
1423
1424             /* REACTION-FIELD ELECTROSTATICS */
1425             felec            = _mm_mul_pd(qq12,_mm_sub_pd(_mm_mul_pd(rinv12,rinvsq12),krf2));
1426
1427             fscal            = felec;
1428
1429             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1430
1431             /* Calculate temporary vectorial force */
1432             tx               = _mm_mul_pd(fscal,dx12);
1433             ty               = _mm_mul_pd(fscal,dy12);
1434             tz               = _mm_mul_pd(fscal,dz12);
1435
1436             /* Update vectorial force */
1437             fix1             = _mm_add_pd(fix1,tx);
1438             fiy1             = _mm_add_pd(fiy1,ty);
1439             fiz1             = _mm_add_pd(fiz1,tz);
1440
1441             fjx2             = _mm_add_pd(fjx2,tx);
1442             fjy2             = _mm_add_pd(fjy2,ty);
1443             fjz2             = _mm_add_pd(fjz2,tz);
1444
1445             /**************************
1446              * CALCULATE INTERACTIONS *
1447              **************************/
1448
1449             /* REACTION-FIELD ELECTROSTATICS */
1450             felec            = _mm_mul_pd(qq13,_mm_sub_pd(_mm_mul_pd(rinv13,rinvsq13),krf2));
1451
1452             fscal            = felec;
1453
1454             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1455
1456             /* Calculate temporary vectorial force */
1457             tx               = _mm_mul_pd(fscal,dx13);
1458             ty               = _mm_mul_pd(fscal,dy13);
1459             tz               = _mm_mul_pd(fscal,dz13);
1460
1461             /* Update vectorial force */
1462             fix1             = _mm_add_pd(fix1,tx);
1463             fiy1             = _mm_add_pd(fiy1,ty);
1464             fiz1             = _mm_add_pd(fiz1,tz);
1465
1466             fjx3             = _mm_add_pd(fjx3,tx);
1467             fjy3             = _mm_add_pd(fjy3,ty);
1468             fjz3             = _mm_add_pd(fjz3,tz);
1469
1470             /**************************
1471              * CALCULATE INTERACTIONS *
1472              **************************/
1473
1474             /* REACTION-FIELD ELECTROSTATICS */
1475             felec            = _mm_mul_pd(qq21,_mm_sub_pd(_mm_mul_pd(rinv21,rinvsq21),krf2));
1476
1477             fscal            = felec;
1478
1479             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1480
1481             /* Calculate temporary vectorial force */
1482             tx               = _mm_mul_pd(fscal,dx21);
1483             ty               = _mm_mul_pd(fscal,dy21);
1484             tz               = _mm_mul_pd(fscal,dz21);
1485
1486             /* Update vectorial force */
1487             fix2             = _mm_add_pd(fix2,tx);
1488             fiy2             = _mm_add_pd(fiy2,ty);
1489             fiz2             = _mm_add_pd(fiz2,tz);
1490
1491             fjx1             = _mm_add_pd(fjx1,tx);
1492             fjy1             = _mm_add_pd(fjy1,ty);
1493             fjz1             = _mm_add_pd(fjz1,tz);
1494
1495             /**************************
1496              * CALCULATE INTERACTIONS *
1497              **************************/
1498
1499             /* REACTION-FIELD ELECTROSTATICS */
1500             felec            = _mm_mul_pd(qq22,_mm_sub_pd(_mm_mul_pd(rinv22,rinvsq22),krf2));
1501
1502             fscal            = felec;
1503
1504             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1505
1506             /* Calculate temporary vectorial force */
1507             tx               = _mm_mul_pd(fscal,dx22);
1508             ty               = _mm_mul_pd(fscal,dy22);
1509             tz               = _mm_mul_pd(fscal,dz22);
1510
1511             /* Update vectorial force */
1512             fix2             = _mm_add_pd(fix2,tx);
1513             fiy2             = _mm_add_pd(fiy2,ty);
1514             fiz2             = _mm_add_pd(fiz2,tz);
1515
1516             fjx2             = _mm_add_pd(fjx2,tx);
1517             fjy2             = _mm_add_pd(fjy2,ty);
1518             fjz2             = _mm_add_pd(fjz2,tz);
1519
1520             /**************************
1521              * CALCULATE INTERACTIONS *
1522              **************************/
1523
1524             /* REACTION-FIELD ELECTROSTATICS */
1525             felec            = _mm_mul_pd(qq23,_mm_sub_pd(_mm_mul_pd(rinv23,rinvsq23),krf2));
1526
1527             fscal            = felec;
1528
1529             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1530
1531             /* Calculate temporary vectorial force */
1532             tx               = _mm_mul_pd(fscal,dx23);
1533             ty               = _mm_mul_pd(fscal,dy23);
1534             tz               = _mm_mul_pd(fscal,dz23);
1535
1536             /* Update vectorial force */
1537             fix2             = _mm_add_pd(fix2,tx);
1538             fiy2             = _mm_add_pd(fiy2,ty);
1539             fiz2             = _mm_add_pd(fiz2,tz);
1540
1541             fjx3             = _mm_add_pd(fjx3,tx);
1542             fjy3             = _mm_add_pd(fjy3,ty);
1543             fjz3             = _mm_add_pd(fjz3,tz);
1544
1545             /**************************
1546              * CALCULATE INTERACTIONS *
1547              **************************/
1548
1549             /* REACTION-FIELD ELECTROSTATICS */
1550             felec            = _mm_mul_pd(qq31,_mm_sub_pd(_mm_mul_pd(rinv31,rinvsq31),krf2));
1551
1552             fscal            = felec;
1553
1554             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1555
1556             /* Calculate temporary vectorial force */
1557             tx               = _mm_mul_pd(fscal,dx31);
1558             ty               = _mm_mul_pd(fscal,dy31);
1559             tz               = _mm_mul_pd(fscal,dz31);
1560
1561             /* Update vectorial force */
1562             fix3             = _mm_add_pd(fix3,tx);
1563             fiy3             = _mm_add_pd(fiy3,ty);
1564             fiz3             = _mm_add_pd(fiz3,tz);
1565
1566             fjx1             = _mm_add_pd(fjx1,tx);
1567             fjy1             = _mm_add_pd(fjy1,ty);
1568             fjz1             = _mm_add_pd(fjz1,tz);
1569
1570             /**************************
1571              * CALCULATE INTERACTIONS *
1572              **************************/
1573
1574             /* REACTION-FIELD ELECTROSTATICS */
1575             felec            = _mm_mul_pd(qq32,_mm_sub_pd(_mm_mul_pd(rinv32,rinvsq32),krf2));
1576
1577             fscal            = felec;
1578
1579             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1580
1581             /* Calculate temporary vectorial force */
1582             tx               = _mm_mul_pd(fscal,dx32);
1583             ty               = _mm_mul_pd(fscal,dy32);
1584             tz               = _mm_mul_pd(fscal,dz32);
1585
1586             /* Update vectorial force */
1587             fix3             = _mm_add_pd(fix3,tx);
1588             fiy3             = _mm_add_pd(fiy3,ty);
1589             fiz3             = _mm_add_pd(fiz3,tz);
1590
1591             fjx2             = _mm_add_pd(fjx2,tx);
1592             fjy2             = _mm_add_pd(fjy2,ty);
1593             fjz2             = _mm_add_pd(fjz2,tz);
1594
1595             /**************************
1596              * CALCULATE INTERACTIONS *
1597              **************************/
1598
1599             /* REACTION-FIELD ELECTROSTATICS */
1600             felec            = _mm_mul_pd(qq33,_mm_sub_pd(_mm_mul_pd(rinv33,rinvsq33),krf2));
1601
1602             fscal            = felec;
1603
1604             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1605
1606             /* Calculate temporary vectorial force */
1607             tx               = _mm_mul_pd(fscal,dx33);
1608             ty               = _mm_mul_pd(fscal,dy33);
1609             tz               = _mm_mul_pd(fscal,dz33);
1610
1611             /* Update vectorial force */
1612             fix3             = _mm_add_pd(fix3,tx);
1613             fiy3             = _mm_add_pd(fiy3,ty);
1614             fiz3             = _mm_add_pd(fiz3,tz);
1615
1616             fjx3             = _mm_add_pd(fjx3,tx);
1617             fjy3             = _mm_add_pd(fjy3,ty);
1618             fjz3             = _mm_add_pd(fjz3,tz);
1619
1620             gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA+DIM,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1621
1622             /* Inner loop uses 243 flops */
1623         }
1624
1625         /* End of innermost loop */
1626
1627         gmx_mm_update_iforce_3atom_swizzle_pd(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1628                                               f+i_coord_offset+DIM,fshift+i_shift_offset);
1629
1630         /* Increment number of inner iterations */
1631         inneriter                  += j_index_end - j_index_start;
1632
1633         /* Outer loop uses 18 flops */
1634     }
1635
1636     /* Increment number of outer iterations */
1637     outeriter        += nri;
1638
1639     /* Update outer/inner flops */
1640
1641     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4W4_F,outeriter*18 + inneriter*243);
1642 }