src/gromacs/gmxlib/nonbonded/nb_kernel_sse4_1_double/nb_kernel_ElecRF_VdwLJ_GeomW4W4_sse4_1_double.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013, 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 #ifdef HAVE_CONFIG_H
  39 #include <config.h>
  40 #endif
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "types/simple.h"
  46 #include "vec.h"
  47 #include "nrnb.h"
  48
  49 #include "gromacs/simd/math_x86_sse4_1_double.h"
  50 #include "kernelutil_x86_sse4_1_double.h"
  51
  52 /*
  53  * Gromacs nonbonded kernel:   nb_kernel_ElecRF_VdwLJ_GeomW4W4_VF_sse4_1_double
  54  * Electrostatics interaction: ReactionField
  55  * VdW interaction:            LennardJones
  56  * Geometry:                   Water4-Water4
  57  * Calculate force/pot:        PotentialAndForce
  58  */
  59 void
  60 nb_kernel_ElecRF_VdwLJ_GeomW4W4_VF_sse4_1_double
  61                     (t_nblist                    * gmx_restrict       nlist,
  62                      rvec                        * gmx_restrict          xx,
  63                      rvec                        * gmx_restrict          ff,
  64                      t_forcerec                  * gmx_restrict          fr,
  65                      t_mdatoms                   * gmx_restrict     mdatoms,
  66                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  67                      t_nrnb                      * gmx_restrict        nrnb)
  68 {
  69     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  70      * just 0 for non-waters.
  71      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
  72      * jnr indices corresponding to data put in the four positions in the SIMD register.
  73      */
  74     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  75     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  76     int              jnrA,jnrB;
  77     int              j_coord_offsetA,j_coord_offsetB;
  78     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  79     real             rcutoff_scalar;
  80     real             *shiftvec,*fshift,*x,*f;
  81     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  82     int              vdwioffset0;
  83     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  84     int              vdwioffset1;
  85     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  86     int              vdwioffset2;
  87     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  88     int              vdwioffset3;
  89     __m128d          ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  90     int              vdwjidx0A,vdwjidx0B;
  91     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  92     int              vdwjidx1A,vdwjidx1B;
  93     __m128d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  94     int              vdwjidx2A,vdwjidx2B;
  95     __m128d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  96     int              vdwjidx3A,vdwjidx3B;
  97     __m128d          jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
  98     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  99     __m128d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
 100     __m128d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
 101     __m128d          dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
 102     __m128d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 103     __m128d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 104     __m128d          dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
 105     __m128d          dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
 106     __m128d          dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
 107     __m128d          dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
 108     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
 109     real             *charge;
 110     int              nvdwtype;
 111     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 112     int              *vdwtype;
 113     real             *vdwparam;
 114     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
 115     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
 116     __m128d          dummy_mask,cutoff_mask;
 117     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 118     __m128d          one     = _mm_set1_pd(1.0);
 119     __m128d          two     = _mm_set1_pd(2.0);
 120     x                = xx[0];
 121     f                = ff[0];
 122
 123     nri              = nlist->nri;
 124     iinr             = nlist->iinr;
 125     jindex           = nlist->jindex;
 126     jjnr             = nlist->jjnr;
 127     shiftidx         = nlist->shift;
 128     gid              = nlist->gid;
 129     shiftvec         = fr->shift_vec[0];
 130     fshift           = fr->fshift[0];
 131     facel            = _mm_set1_pd(fr->epsfac);
 132     charge           = mdatoms->chargeA;
 133     krf              = _mm_set1_pd(fr->ic->k_rf);
 134     krf2             = _mm_set1_pd(fr->ic->k_rf*2.0);
 135     crf              = _mm_set1_pd(fr->ic->c_rf);
 136     nvdwtype         = fr->ntype;
 137     vdwparam         = fr->nbfp;
 138     vdwtype          = mdatoms->typeA;
 139
 140     /* Setup water-specific parameters */
 141     inr              = nlist->iinr[0];
 142     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
 143     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
 144     iq3              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
 145     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 146
 147     jq1              = _mm_set1_pd(charge[inr+1]);
 148     jq2              = _mm_set1_pd(charge[inr+2]);
 149     jq3              = _mm_set1_pd(charge[inr+3]);
 150     vdwjidx0A        = 2*vdwtype[inr+0];
 151     c6_00            = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
 152     c12_00           = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
 153     qq11             = _mm_mul_pd(iq1,jq1);
 154     qq12             = _mm_mul_pd(iq1,jq2);
 155     qq13             = _mm_mul_pd(iq1,jq3);
 156     qq21             = _mm_mul_pd(iq2,jq1);
 157     qq22             = _mm_mul_pd(iq2,jq2);
 158     qq23             = _mm_mul_pd(iq2,jq3);
 159     qq31             = _mm_mul_pd(iq3,jq1);
 160     qq32             = _mm_mul_pd(iq3,jq2);
 161     qq33             = _mm_mul_pd(iq3,jq3);
 162
 163     /* Avoid stupid compiler warnings */
 164     jnrA = jnrB = 0;
 165     j_coord_offsetA = 0;
 166     j_coord_offsetB = 0;
 167
 168     outeriter        = 0;
 169     inneriter        = 0;
 170
 171     /* Start outer loop over neighborlists */
 172     for(iidx=0; iidx<nri; iidx++)
 173     {
 174         /* Load shift vector for this list */
 175         i_shift_offset   = DIM*shiftidx[iidx];
 176
 177         /* Load limits for loop over neighbors */
 178         j_index_start    = jindex[iidx];
 179         j_index_end      = jindex[iidx+1];
 180
 181         /* Get outer coordinate index */
 182         inr              = iinr[iidx];
 183         i_coord_offset   = DIM*inr;
 184
 185         /* Load i particle coords and add shift vector */
 186         gmx_mm_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 187                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 188
 189         fix0             = _mm_setzero_pd();
 190         fiy0             = _mm_setzero_pd();
 191         fiz0             = _mm_setzero_pd();
 192         fix1             = _mm_setzero_pd();
 193         fiy1             = _mm_setzero_pd();
 194         fiz1             = _mm_setzero_pd();
 195         fix2             = _mm_setzero_pd();
 196         fiy2             = _mm_setzero_pd();
 197         fiz2             = _mm_setzero_pd();
 198         fix3             = _mm_setzero_pd();
 199         fiy3             = _mm_setzero_pd();
 200         fiz3             = _mm_setzero_pd();
 201
 202         /* Reset potential sums */
 203         velecsum         = _mm_setzero_pd();
 204         vvdwsum          = _mm_setzero_pd();
 205
 206         /* Start inner kernel loop */
 207         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 208         {
 209
 210             /* Get j neighbor index, and coordinate index */
 211             jnrA             = jjnr[jidx];
 212             jnrB             = jjnr[jidx+1];
 213             j_coord_offsetA  = DIM*jnrA;
 214             j_coord_offsetB  = DIM*jnrB;
 215
 216             /* load j atom coordinates */
 217             gmx_mm_load_4rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 218                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 219                                               &jy2,&jz2,&jx3,&jy3,&jz3);
 220
 221             /* Calculate displacement vector */
 222             dx00             = _mm_sub_pd(ix0,jx0);
 223             dy00             = _mm_sub_pd(iy0,jy0);
 224             dz00             = _mm_sub_pd(iz0,jz0);
 225             dx11             = _mm_sub_pd(ix1,jx1);
 226             dy11             = _mm_sub_pd(iy1,jy1);
 227             dz11             = _mm_sub_pd(iz1,jz1);
 228             dx12             = _mm_sub_pd(ix1,jx2);
 229             dy12             = _mm_sub_pd(iy1,jy2);
 230             dz12             = _mm_sub_pd(iz1,jz2);
 231             dx13             = _mm_sub_pd(ix1,jx3);
 232             dy13             = _mm_sub_pd(iy1,jy3);
 233             dz13             = _mm_sub_pd(iz1,jz3);
 234             dx21             = _mm_sub_pd(ix2,jx1);
 235             dy21             = _mm_sub_pd(iy2,jy1);
 236             dz21             = _mm_sub_pd(iz2,jz1);
 237             dx22             = _mm_sub_pd(ix2,jx2);
 238             dy22             = _mm_sub_pd(iy2,jy2);
 239             dz22             = _mm_sub_pd(iz2,jz2);
 240             dx23             = _mm_sub_pd(ix2,jx3);
 241             dy23             = _mm_sub_pd(iy2,jy3);
 242             dz23             = _mm_sub_pd(iz2,jz3);
 243             dx31             = _mm_sub_pd(ix3,jx1);
 244             dy31             = _mm_sub_pd(iy3,jy1);
 245             dz31             = _mm_sub_pd(iz3,jz1);
 246             dx32             = _mm_sub_pd(ix3,jx2);
 247             dy32             = _mm_sub_pd(iy3,jy2);
 248             dz32             = _mm_sub_pd(iz3,jz2);
 249             dx33             = _mm_sub_pd(ix3,jx3);
 250             dy33             = _mm_sub_pd(iy3,jy3);
 251             dz33             = _mm_sub_pd(iz3,jz3);
 252
 253             /* Calculate squared distance and things based on it */
 254             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 255             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 256             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 257             rsq13            = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
 258             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 259             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 260             rsq23            = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
 261             rsq31            = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
 262             rsq32            = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
 263             rsq33            = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
 264
 265             rinv11           = gmx_mm_invsqrt_pd(rsq11);
 266             rinv12           = gmx_mm_invsqrt_pd(rsq12);
 267             rinv13           = gmx_mm_invsqrt_pd(rsq13);
 268             rinv21           = gmx_mm_invsqrt_pd(rsq21);
 269             rinv22           = gmx_mm_invsqrt_pd(rsq22);
 270             rinv23           = gmx_mm_invsqrt_pd(rsq23);
 271             rinv31           = gmx_mm_invsqrt_pd(rsq31);
 272             rinv32           = gmx_mm_invsqrt_pd(rsq32);
 273             rinv33           = gmx_mm_invsqrt_pd(rsq33);
 274
 275             rinvsq00         = gmx_mm_inv_pd(rsq00);
 276             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
 277             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
 278             rinvsq13         = _mm_mul_pd(rinv13,rinv13);
 279             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
 280             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
 281             rinvsq23         = _mm_mul_pd(rinv23,rinv23);
 282             rinvsq31         = _mm_mul_pd(rinv31,rinv31);
 283             rinvsq32         = _mm_mul_pd(rinv32,rinv32);
 284             rinvsq33         = _mm_mul_pd(rinv33,rinv33);
 285
 286             fjx0             = _mm_setzero_pd();
 287             fjy0             = _mm_setzero_pd();
 288             fjz0             = _mm_setzero_pd();
 289             fjx1             = _mm_setzero_pd();
 290             fjy1             = _mm_setzero_pd();
 291             fjz1             = _mm_setzero_pd();
 292             fjx2             = _mm_setzero_pd();
 293             fjy2             = _mm_setzero_pd();
 294             fjz2             = _mm_setzero_pd();
 295             fjx3             = _mm_setzero_pd();
 296             fjy3             = _mm_setzero_pd();
 297             fjz3             = _mm_setzero_pd();
 298
 299             /**************************
 300              * CALCULATE INTERACTIONS *
 301              **************************/
 302
 303             /* LENNARD-JONES DISPERSION/REPULSION */
 304
 305             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 306             vvdw6            = _mm_mul_pd(c6_00,rinvsix);
 307             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
 308             vvdw             = _mm_sub_pd( _mm_mul_pd(vvdw12,one_twelfth) , _mm_mul_pd(vvdw6,one_sixth) );
 309             fvdw             = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
 310
 311             /* Update potential sum for this i atom from the interaction with this j atom. */
 312             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 313
 314             fscal            = fvdw;
 315
 316             /* Calculate temporary vectorial force */
 317             tx               = _mm_mul_pd(fscal,dx00);
 318             ty               = _mm_mul_pd(fscal,dy00);
 319             tz               = _mm_mul_pd(fscal,dz00);
 320
 321             /* Update vectorial force */
 322             fix0             = _mm_add_pd(fix0,tx);
 323             fiy0             = _mm_add_pd(fiy0,ty);
 324             fiz0             = _mm_add_pd(fiz0,tz);
 325
 326             fjx0             = _mm_add_pd(fjx0,tx);
 327             fjy0             = _mm_add_pd(fjy0,ty);
 328             fjz0             = _mm_add_pd(fjz0,tz);
 329
 330             /**************************
 331              * CALCULATE INTERACTIONS *
 332              **************************/
 333
 334             /* REACTION-FIELD ELECTROSTATICS */
 335             velec            = _mm_mul_pd(qq11,_mm_sub_pd(_mm_add_pd(rinv11,_mm_mul_pd(krf,rsq11)),crf));
 336             felec            = _mm_mul_pd(qq11,_mm_sub_pd(_mm_mul_pd(rinv11,rinvsq11),krf2));
 337
 338             /* Update potential sum for this i atom from the interaction with this j atom. */
 339             velecsum         = _mm_add_pd(velecsum,velec);
 340
 341             fscal            = felec;
 342
 343             /* Calculate temporary vectorial force */
 344             tx               = _mm_mul_pd(fscal,dx11);
 345             ty               = _mm_mul_pd(fscal,dy11);
 346             tz               = _mm_mul_pd(fscal,dz11);
 347
 348             /* Update vectorial force */
 349             fix1             = _mm_add_pd(fix1,tx);
 350             fiy1             = _mm_add_pd(fiy1,ty);
 351             fiz1             = _mm_add_pd(fiz1,tz);
 352
 353             fjx1             = _mm_add_pd(fjx1,tx);
 354             fjy1             = _mm_add_pd(fjy1,ty);
 355             fjz1             = _mm_add_pd(fjz1,tz);
 356
 357             /**************************
 358              * CALCULATE INTERACTIONS *
 359              **************************/
 360
 361             /* REACTION-FIELD ELECTROSTATICS */
 362             velec            = _mm_mul_pd(qq12,_mm_sub_pd(_mm_add_pd(rinv12,_mm_mul_pd(krf,rsq12)),crf));
 363             felec            = _mm_mul_pd(qq12,_mm_sub_pd(_mm_mul_pd(rinv12,rinvsq12),krf2));
 364
 365             /* Update potential sum for this i atom from the interaction with this j atom. */
 366             velecsum         = _mm_add_pd(velecsum,velec);
 367
 368             fscal            = felec;
 369
 370             /* Calculate temporary vectorial force */
 371             tx               = _mm_mul_pd(fscal,dx12);
 372             ty               = _mm_mul_pd(fscal,dy12);
 373             tz               = _mm_mul_pd(fscal,dz12);
 374
 375             /* Update vectorial force */
 376             fix1             = _mm_add_pd(fix1,tx);
 377             fiy1             = _mm_add_pd(fiy1,ty);
 378             fiz1             = _mm_add_pd(fiz1,tz);
 379
 380             fjx2             = _mm_add_pd(fjx2,tx);
 381             fjy2             = _mm_add_pd(fjy2,ty);
 382             fjz2             = _mm_add_pd(fjz2,tz);
 383
 384             /**************************
 385              * CALCULATE INTERACTIONS *
 386              **************************/
 387
 388             /* REACTION-FIELD ELECTROSTATICS */
 389             velec            = _mm_mul_pd(qq13,_mm_sub_pd(_mm_add_pd(rinv13,_mm_mul_pd(krf,rsq13)),crf));
 390             felec            = _mm_mul_pd(qq13,_mm_sub_pd(_mm_mul_pd(rinv13,rinvsq13),krf2));
 391
 392             /* Update potential sum for this i atom from the interaction with this j atom. */
 393             velecsum         = _mm_add_pd(velecsum,velec);
 394
 395             fscal            = felec;
 396
 397             /* Calculate temporary vectorial force */
 398             tx               = _mm_mul_pd(fscal,dx13);
 399             ty               = _mm_mul_pd(fscal,dy13);
 400             tz               = _mm_mul_pd(fscal,dz13);
 401
 402             /* Update vectorial force */
 403             fix1             = _mm_add_pd(fix1,tx);
 404             fiy1             = _mm_add_pd(fiy1,ty);
 405             fiz1             = _mm_add_pd(fiz1,tz);
 406
 407             fjx3             = _mm_add_pd(fjx3,tx);
 408             fjy3             = _mm_add_pd(fjy3,ty);
 409             fjz3             = _mm_add_pd(fjz3,tz);
 410
 411             /**************************
 412              * CALCULATE INTERACTIONS *
 413              **************************/
 414
 415             /* REACTION-FIELD ELECTROSTATICS */
 416             velec            = _mm_mul_pd(qq21,_mm_sub_pd(_mm_add_pd(rinv21,_mm_mul_pd(krf,rsq21)),crf));
 417             felec            = _mm_mul_pd(qq21,_mm_sub_pd(_mm_mul_pd(rinv21,rinvsq21),krf2));
 418
 419             /* Update potential sum for this i atom from the interaction with this j atom. */
 420             velecsum         = _mm_add_pd(velecsum,velec);
 421
 422             fscal            = felec;
 423
 424             /* Calculate temporary vectorial force */
 425             tx               = _mm_mul_pd(fscal,dx21);
 426             ty               = _mm_mul_pd(fscal,dy21);
 427             tz               = _mm_mul_pd(fscal,dz21);
 428
 429             /* Update vectorial force */
 430             fix2             = _mm_add_pd(fix2,tx);
 431             fiy2             = _mm_add_pd(fiy2,ty);
 432             fiz2             = _mm_add_pd(fiz2,tz);
 433
 434             fjx1             = _mm_add_pd(fjx1,tx);
 435             fjy1             = _mm_add_pd(fjy1,ty);
 436             fjz1             = _mm_add_pd(fjz1,tz);
 437
 438             /**************************
 439              * CALCULATE INTERACTIONS *
 440              **************************/
 441
 442             /* REACTION-FIELD ELECTROSTATICS */
 443             velec            = _mm_mul_pd(qq22,_mm_sub_pd(_mm_add_pd(rinv22,_mm_mul_pd(krf,rsq22)),crf));
 444             felec            = _mm_mul_pd(qq22,_mm_sub_pd(_mm_mul_pd(rinv22,rinvsq22),krf2));
 445
 446             /* Update potential sum for this i atom from the interaction with this j atom. */
 447             velecsum         = _mm_add_pd(velecsum,velec);
 448
 449             fscal            = felec;
 450
 451             /* Calculate temporary vectorial force */
 452             tx               = _mm_mul_pd(fscal,dx22);
 453             ty               = _mm_mul_pd(fscal,dy22);
 454             tz               = _mm_mul_pd(fscal,dz22);
 455
 456             /* Update vectorial force */
 457             fix2             = _mm_add_pd(fix2,tx);
 458             fiy2             = _mm_add_pd(fiy2,ty);
 459             fiz2             = _mm_add_pd(fiz2,tz);
 460
 461             fjx2             = _mm_add_pd(fjx2,tx);
 462             fjy2             = _mm_add_pd(fjy2,ty);
 463             fjz2             = _mm_add_pd(fjz2,tz);
 464
 465             /**************************
 466              * CALCULATE INTERACTIONS *
 467              **************************/
 468
 469             /* REACTION-FIELD ELECTROSTATICS */
 470             velec            = _mm_mul_pd(qq23,_mm_sub_pd(_mm_add_pd(rinv23,_mm_mul_pd(krf,rsq23)),crf));
 471             felec            = _mm_mul_pd(qq23,_mm_sub_pd(_mm_mul_pd(rinv23,rinvsq23),krf2));
 472
 473             /* Update potential sum for this i atom from the interaction with this j atom. */
 474             velecsum         = _mm_add_pd(velecsum,velec);
 475
 476             fscal            = felec;
 477
 478             /* Calculate temporary vectorial force */
 479             tx               = _mm_mul_pd(fscal,dx23);
 480             ty               = _mm_mul_pd(fscal,dy23);
 481             tz               = _mm_mul_pd(fscal,dz23);
 482
 483             /* Update vectorial force */
 484             fix2             = _mm_add_pd(fix2,tx);
 485             fiy2             = _mm_add_pd(fiy2,ty);
 486             fiz2             = _mm_add_pd(fiz2,tz);
 487
 488             fjx3             = _mm_add_pd(fjx3,tx);
 489             fjy3             = _mm_add_pd(fjy3,ty);
 490             fjz3             = _mm_add_pd(fjz3,tz);
 491
 492             /**************************
 493              * CALCULATE INTERACTIONS *
 494              **************************/
 495
 496             /* REACTION-FIELD ELECTROSTATICS */
 497             velec            = _mm_mul_pd(qq31,_mm_sub_pd(_mm_add_pd(rinv31,_mm_mul_pd(krf,rsq31)),crf));
 498             felec            = _mm_mul_pd(qq31,_mm_sub_pd(_mm_mul_pd(rinv31,rinvsq31),krf2));
 499
 500             /* Update potential sum for this i atom from the interaction with this j atom. */
 501             velecsum         = _mm_add_pd(velecsum,velec);
 502
 503             fscal            = felec;
 504
 505             /* Calculate temporary vectorial force */
 506             tx               = _mm_mul_pd(fscal,dx31);
 507             ty               = _mm_mul_pd(fscal,dy31);
 508             tz               = _mm_mul_pd(fscal,dz31);
 509
 510             /* Update vectorial force */
 511             fix3             = _mm_add_pd(fix3,tx);
 512             fiy3             = _mm_add_pd(fiy3,ty);
 513             fiz3             = _mm_add_pd(fiz3,tz);
 514
 515             fjx1             = _mm_add_pd(fjx1,tx);
 516             fjy1             = _mm_add_pd(fjy1,ty);
 517             fjz1             = _mm_add_pd(fjz1,tz);
 518
 519             /**************************
 520              * CALCULATE INTERACTIONS *
 521              **************************/
 522
 523             /* REACTION-FIELD ELECTROSTATICS */
 524             velec            = _mm_mul_pd(qq32,_mm_sub_pd(_mm_add_pd(rinv32,_mm_mul_pd(krf,rsq32)),crf));
 525             felec            = _mm_mul_pd(qq32,_mm_sub_pd(_mm_mul_pd(rinv32,rinvsq32),krf2));
 526
 527             /* Update potential sum for this i atom from the interaction with this j atom. */
 528             velecsum         = _mm_add_pd(velecsum,velec);
 529
 530             fscal            = felec;
 531
 532             /* Calculate temporary vectorial force */
 533             tx               = _mm_mul_pd(fscal,dx32);
 534             ty               = _mm_mul_pd(fscal,dy32);
 535             tz               = _mm_mul_pd(fscal,dz32);
 536
 537             /* Update vectorial force */
 538             fix3             = _mm_add_pd(fix3,tx);
 539             fiy3             = _mm_add_pd(fiy3,ty);
 540             fiz3             = _mm_add_pd(fiz3,tz);
 541
 542             fjx2             = _mm_add_pd(fjx2,tx);
 543             fjy2             = _mm_add_pd(fjy2,ty);
 544             fjz2             = _mm_add_pd(fjz2,tz);
 545
 546             /**************************
 547              * CALCULATE INTERACTIONS *
 548              **************************/
 549
 550             /* REACTION-FIELD ELECTROSTATICS */
 551             velec            = _mm_mul_pd(qq33,_mm_sub_pd(_mm_add_pd(rinv33,_mm_mul_pd(krf,rsq33)),crf));
 552             felec            = _mm_mul_pd(qq33,_mm_sub_pd(_mm_mul_pd(rinv33,rinvsq33),krf2));
 553
 554             /* Update potential sum for this i atom from the interaction with this j atom. */
 555             velecsum         = _mm_add_pd(velecsum,velec);
 556
 557             fscal            = felec;
 558
 559             /* Calculate temporary vectorial force */
 560             tx               = _mm_mul_pd(fscal,dx33);
 561             ty               = _mm_mul_pd(fscal,dy33);
 562             tz               = _mm_mul_pd(fscal,dz33);
 563
 564             /* Update vectorial force */
 565             fix3             = _mm_add_pd(fix3,tx);
 566             fiy3             = _mm_add_pd(fiy3,ty);
 567             fiz3             = _mm_add_pd(fiz3,tz);
 568
 569             fjx3             = _mm_add_pd(fjx3,tx);
 570             fjy3             = _mm_add_pd(fjy3,ty);
 571             fjz3             = _mm_add_pd(fjz3,tz);
 572
 573             gmx_mm_decrement_4rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 574
 575             /* Inner loop uses 323 flops */
 576         }
 577
 578         if(jidx<j_index_end)
 579         {
 580
 581             jnrA             = jjnr[jidx];
 582             j_coord_offsetA  = DIM*jnrA;
 583
 584             /* load j atom coordinates */
 585             gmx_mm_load_4rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 586                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 587                                               &jy2,&jz2,&jx3,&jy3,&jz3);
 588
 589             /* Calculate displacement vector */
 590             dx00             = _mm_sub_pd(ix0,jx0);
 591             dy00             = _mm_sub_pd(iy0,jy0);
 592             dz00             = _mm_sub_pd(iz0,jz0);
 593             dx11             = _mm_sub_pd(ix1,jx1);
 594             dy11             = _mm_sub_pd(iy1,jy1);
 595             dz11             = _mm_sub_pd(iz1,jz1);
 596             dx12             = _mm_sub_pd(ix1,jx2);
 597             dy12             = _mm_sub_pd(iy1,jy2);
 598             dz12             = _mm_sub_pd(iz1,jz2);
 599             dx13             = _mm_sub_pd(ix1,jx3);
 600             dy13             = _mm_sub_pd(iy1,jy3);
 601             dz13             = _mm_sub_pd(iz1,jz3);
 602             dx21             = _mm_sub_pd(ix2,jx1);
 603             dy21             = _mm_sub_pd(iy2,jy1);
 604             dz21             = _mm_sub_pd(iz2,jz1);
 605             dx22             = _mm_sub_pd(ix2,jx2);
 606             dy22             = _mm_sub_pd(iy2,jy2);
 607             dz22             = _mm_sub_pd(iz2,jz2);
 608             dx23             = _mm_sub_pd(ix2,jx3);
 609             dy23             = _mm_sub_pd(iy2,jy3);
 610             dz23             = _mm_sub_pd(iz2,jz3);
 611             dx31             = _mm_sub_pd(ix3,jx1);
 612             dy31             = _mm_sub_pd(iy3,jy1);
 613             dz31             = _mm_sub_pd(iz3,jz1);
 614             dx32             = _mm_sub_pd(ix3,jx2);
 615             dy32             = _mm_sub_pd(iy3,jy2);
 616             dz32             = _mm_sub_pd(iz3,jz2);
 617             dx33             = _mm_sub_pd(ix3,jx3);
 618             dy33             = _mm_sub_pd(iy3,jy3);
 619             dz33             = _mm_sub_pd(iz3,jz3);
 620
 621             /* Calculate squared distance and things based on it */
 622             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 623             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 624             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 625             rsq13            = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
 626             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 627             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 628             rsq23            = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
 629             rsq31            = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
 630             rsq32            = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
 631             rsq33            = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
 632
 633             rinv11           = gmx_mm_invsqrt_pd(rsq11);
 634             rinv12           = gmx_mm_invsqrt_pd(rsq12);
 635             rinv13           = gmx_mm_invsqrt_pd(rsq13);
 636             rinv21           = gmx_mm_invsqrt_pd(rsq21);
 637             rinv22           = gmx_mm_invsqrt_pd(rsq22);
 638             rinv23           = gmx_mm_invsqrt_pd(rsq23);
 639             rinv31           = gmx_mm_invsqrt_pd(rsq31);
 640             rinv32           = gmx_mm_invsqrt_pd(rsq32);
 641             rinv33           = gmx_mm_invsqrt_pd(rsq33);
 642
 643             rinvsq00         = gmx_mm_inv_pd(rsq00);
 644             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
 645             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
 646             rinvsq13         = _mm_mul_pd(rinv13,rinv13);
 647             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
 648             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
 649             rinvsq23         = _mm_mul_pd(rinv23,rinv23);
 650             rinvsq31         = _mm_mul_pd(rinv31,rinv31);
 651             rinvsq32         = _mm_mul_pd(rinv32,rinv32);
 652             rinvsq33         = _mm_mul_pd(rinv33,rinv33);
 653
 654             fjx0             = _mm_setzero_pd();
 655             fjy0             = _mm_setzero_pd();
 656             fjz0             = _mm_setzero_pd();
 657             fjx1             = _mm_setzero_pd();
 658             fjy1             = _mm_setzero_pd();
 659             fjz1             = _mm_setzero_pd();
 660             fjx2             = _mm_setzero_pd();
 661             fjy2             = _mm_setzero_pd();
 662             fjz2             = _mm_setzero_pd();
 663             fjx3             = _mm_setzero_pd();
 664             fjy3             = _mm_setzero_pd();
 665             fjz3             = _mm_setzero_pd();
 666
 667             /**************************
 668              * CALCULATE INTERACTIONS *
 669              **************************/
 670
 671             /* LENNARD-JONES DISPERSION/REPULSION */
 672
 673             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 674             vvdw6            = _mm_mul_pd(c6_00,rinvsix);
 675             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
 676             vvdw             = _mm_sub_pd( _mm_mul_pd(vvdw12,one_twelfth) , _mm_mul_pd(vvdw6,one_sixth) );
 677             fvdw             = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
 678
 679             /* Update potential sum for this i atom from the interaction with this j atom. */
 680             vvdw             = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
 681             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 682
 683             fscal            = fvdw;
 684
 685             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 686
 687             /* Calculate temporary vectorial force */
 688             tx               = _mm_mul_pd(fscal,dx00);
 689             ty               = _mm_mul_pd(fscal,dy00);
 690             tz               = _mm_mul_pd(fscal,dz00);
 691
 692             /* Update vectorial force */
 693             fix0             = _mm_add_pd(fix0,tx);
 694             fiy0             = _mm_add_pd(fiy0,ty);
 695             fiz0             = _mm_add_pd(fiz0,tz);
 696
 697             fjx0             = _mm_add_pd(fjx0,tx);
 698             fjy0             = _mm_add_pd(fjy0,ty);
 699             fjz0             = _mm_add_pd(fjz0,tz);
 700
 701             /**************************
 702              * CALCULATE INTERACTIONS *
 703              **************************/
 704
 705             /* REACTION-FIELD ELECTROSTATICS */
 706             velec            = _mm_mul_pd(qq11,_mm_sub_pd(_mm_add_pd(rinv11,_mm_mul_pd(krf,rsq11)),crf));
 707             felec            = _mm_mul_pd(qq11,_mm_sub_pd(_mm_mul_pd(rinv11,rinvsq11),krf2));
 708
 709             /* Update potential sum for this i atom from the interaction with this j atom. */
 710             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 711             velecsum         = _mm_add_pd(velecsum,velec);
 712
 713             fscal            = felec;
 714
 715             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 716
 717             /* Calculate temporary vectorial force */
 718             tx               = _mm_mul_pd(fscal,dx11);
 719             ty               = _mm_mul_pd(fscal,dy11);
 720             tz               = _mm_mul_pd(fscal,dz11);
 721
 722             /* Update vectorial force */
 723             fix1             = _mm_add_pd(fix1,tx);
 724             fiy1             = _mm_add_pd(fiy1,ty);
 725             fiz1             = _mm_add_pd(fiz1,tz);
 726
 727             fjx1             = _mm_add_pd(fjx1,tx);
 728             fjy1             = _mm_add_pd(fjy1,ty);
 729             fjz1             = _mm_add_pd(fjz1,tz);
 730
 731             /**************************
 732              * CALCULATE INTERACTIONS *
 733              **************************/
 734
 735             /* REACTION-FIELD ELECTROSTATICS */
 736             velec            = _mm_mul_pd(qq12,_mm_sub_pd(_mm_add_pd(rinv12,_mm_mul_pd(krf,rsq12)),crf));
 737             felec            = _mm_mul_pd(qq12,_mm_sub_pd(_mm_mul_pd(rinv12,rinvsq12),krf2));
 738
 739             /* Update potential sum for this i atom from the interaction with this j atom. */
 740             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 741             velecsum         = _mm_add_pd(velecsum,velec);
 742
 743             fscal            = felec;
 744
 745             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 746
 747             /* Calculate temporary vectorial force */
 748             tx               = _mm_mul_pd(fscal,dx12);
 749             ty               = _mm_mul_pd(fscal,dy12);
 750             tz               = _mm_mul_pd(fscal,dz12);
 751
 752             /* Update vectorial force */
 753             fix1             = _mm_add_pd(fix1,tx);
 754             fiy1             = _mm_add_pd(fiy1,ty);
 755             fiz1             = _mm_add_pd(fiz1,tz);
 756
 757             fjx2             = _mm_add_pd(fjx2,tx);
 758             fjy2             = _mm_add_pd(fjy2,ty);
 759             fjz2             = _mm_add_pd(fjz2,tz);
 760
 761             /**************************
 762              * CALCULATE INTERACTIONS *
 763              **************************/
 764
 765             /* REACTION-FIELD ELECTROSTATICS */
 766             velec            = _mm_mul_pd(qq13,_mm_sub_pd(_mm_add_pd(rinv13,_mm_mul_pd(krf,rsq13)),crf));
 767             felec            = _mm_mul_pd(qq13,_mm_sub_pd(_mm_mul_pd(rinv13,rinvsq13),krf2));
 768
 769             /* Update potential sum for this i atom from the interaction with this j atom. */
 770             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 771             velecsum         = _mm_add_pd(velecsum,velec);
 772
 773             fscal            = felec;
 774
 775             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 776
 777             /* Calculate temporary vectorial force */
 778             tx               = _mm_mul_pd(fscal,dx13);
 779             ty               = _mm_mul_pd(fscal,dy13);
 780             tz               = _mm_mul_pd(fscal,dz13);
 781
 782             /* Update vectorial force */
 783             fix1             = _mm_add_pd(fix1,tx);
 784             fiy1             = _mm_add_pd(fiy1,ty);
 785             fiz1             = _mm_add_pd(fiz1,tz);
 786
 787             fjx3             = _mm_add_pd(fjx3,tx);
 788             fjy3             = _mm_add_pd(fjy3,ty);
 789             fjz3             = _mm_add_pd(fjz3,tz);
 790
 791             /**************************
 792              * CALCULATE INTERACTIONS *
 793              **************************/
 794
 795             /* REACTION-FIELD ELECTROSTATICS */
 796             velec            = _mm_mul_pd(qq21,_mm_sub_pd(_mm_add_pd(rinv21,_mm_mul_pd(krf,rsq21)),crf));
 797             felec            = _mm_mul_pd(qq21,_mm_sub_pd(_mm_mul_pd(rinv21,rinvsq21),krf2));
 798
 799             /* Update potential sum for this i atom from the interaction with this j atom. */
 800             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 801             velecsum         = _mm_add_pd(velecsum,velec);
 802
 803             fscal            = felec;
 804
 805             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 806
 807             /* Calculate temporary vectorial force */
 808             tx               = _mm_mul_pd(fscal,dx21);
 809             ty               = _mm_mul_pd(fscal,dy21);
 810             tz               = _mm_mul_pd(fscal,dz21);
 811
 812             /* Update vectorial force */
 813             fix2             = _mm_add_pd(fix2,tx);
 814             fiy2             = _mm_add_pd(fiy2,ty);
 815             fiz2             = _mm_add_pd(fiz2,tz);
 816
 817             fjx1             = _mm_add_pd(fjx1,tx);
 818             fjy1             = _mm_add_pd(fjy1,ty);
 819             fjz1             = _mm_add_pd(fjz1,tz);
 820
 821             /**************************
 822              * CALCULATE INTERACTIONS *
 823              **************************/
 824
 825             /* REACTION-FIELD ELECTROSTATICS */
 826             velec            = _mm_mul_pd(qq22,_mm_sub_pd(_mm_add_pd(rinv22,_mm_mul_pd(krf,rsq22)),crf));
 827             felec            = _mm_mul_pd(qq22,_mm_sub_pd(_mm_mul_pd(rinv22,rinvsq22),krf2));
 828
 829             /* Update potential sum for this i atom from the interaction with this j atom. */
 830             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 831             velecsum         = _mm_add_pd(velecsum,velec);
 832
 833             fscal            = felec;
 834
 835             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 836
 837             /* Calculate temporary vectorial force */
 838             tx               = _mm_mul_pd(fscal,dx22);
 839             ty               = _mm_mul_pd(fscal,dy22);
 840             tz               = _mm_mul_pd(fscal,dz22);
 841
 842             /* Update vectorial force */
 843             fix2             = _mm_add_pd(fix2,tx);
 844             fiy2             = _mm_add_pd(fiy2,ty);
 845             fiz2             = _mm_add_pd(fiz2,tz);
 846
 847             fjx2             = _mm_add_pd(fjx2,tx);
 848             fjy2             = _mm_add_pd(fjy2,ty);
 849             fjz2             = _mm_add_pd(fjz2,tz);
 850
 851             /**************************
 852              * CALCULATE INTERACTIONS *
 853              **************************/
 854
 855             /* REACTION-FIELD ELECTROSTATICS */
 856             velec            = _mm_mul_pd(qq23,_mm_sub_pd(_mm_add_pd(rinv23,_mm_mul_pd(krf,rsq23)),crf));
 857             felec            = _mm_mul_pd(qq23,_mm_sub_pd(_mm_mul_pd(rinv23,rinvsq23),krf2));
 858
 859             /* Update potential sum for this i atom from the interaction with this j atom. */
 860             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 861             velecsum         = _mm_add_pd(velecsum,velec);
 862
 863             fscal            = felec;
 864
 865             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 866
 867             /* Calculate temporary vectorial force */
 868             tx               = _mm_mul_pd(fscal,dx23);
 869             ty               = _mm_mul_pd(fscal,dy23);
 870             tz               = _mm_mul_pd(fscal,dz23);
 871
 872             /* Update vectorial force */
 873             fix2             = _mm_add_pd(fix2,tx);
 874             fiy2             = _mm_add_pd(fiy2,ty);
 875             fiz2             = _mm_add_pd(fiz2,tz);
 876
 877             fjx3             = _mm_add_pd(fjx3,tx);
 878             fjy3             = _mm_add_pd(fjy3,ty);
 879             fjz3             = _mm_add_pd(fjz3,tz);
 880
 881             /**************************
 882              * CALCULATE INTERACTIONS *
 883              **************************/
 884
 885             /* REACTION-FIELD ELECTROSTATICS */
 886             velec            = _mm_mul_pd(qq31,_mm_sub_pd(_mm_add_pd(rinv31,_mm_mul_pd(krf,rsq31)),crf));
 887             felec            = _mm_mul_pd(qq31,_mm_sub_pd(_mm_mul_pd(rinv31,rinvsq31),krf2));
 888
 889             /* Update potential sum for this i atom from the interaction with this j atom. */
 890             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 891             velecsum         = _mm_add_pd(velecsum,velec);
 892
 893             fscal            = felec;
 894
 895             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 896
 897             /* Calculate temporary vectorial force */
 898             tx               = _mm_mul_pd(fscal,dx31);
 899             ty               = _mm_mul_pd(fscal,dy31);
 900             tz               = _mm_mul_pd(fscal,dz31);
 901
 902             /* Update vectorial force */
 903             fix3             = _mm_add_pd(fix3,tx);
 904             fiy3             = _mm_add_pd(fiy3,ty);
 905             fiz3             = _mm_add_pd(fiz3,tz);
 906
 907             fjx1             = _mm_add_pd(fjx1,tx);
 908             fjy1             = _mm_add_pd(fjy1,ty);
 909             fjz1             = _mm_add_pd(fjz1,tz);
 910
 911             /**************************
 912              * CALCULATE INTERACTIONS *
 913              **************************/
 914
 915             /* REACTION-FIELD ELECTROSTATICS */
 916             velec            = _mm_mul_pd(qq32,_mm_sub_pd(_mm_add_pd(rinv32,_mm_mul_pd(krf,rsq32)),crf));
 917             felec            = _mm_mul_pd(qq32,_mm_sub_pd(_mm_mul_pd(rinv32,rinvsq32),krf2));
 918
 919             /* Update potential sum for this i atom from the interaction with this j atom. */
 920             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 921             velecsum         = _mm_add_pd(velecsum,velec);
 922
 923             fscal            = felec;
 924
 925             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 926
 927             /* Calculate temporary vectorial force */
 928             tx               = _mm_mul_pd(fscal,dx32);
 929             ty               = _mm_mul_pd(fscal,dy32);
 930             tz               = _mm_mul_pd(fscal,dz32);
 931
 932             /* Update vectorial force */
 933             fix3             = _mm_add_pd(fix3,tx);
 934             fiy3             = _mm_add_pd(fiy3,ty);
 935             fiz3             = _mm_add_pd(fiz3,tz);
 936
 937             fjx2             = _mm_add_pd(fjx2,tx);
 938             fjy2             = _mm_add_pd(fjy2,ty);
 939             fjz2             = _mm_add_pd(fjz2,tz);
 940
 941             /**************************
 942              * CALCULATE INTERACTIONS *
 943              **************************/
 944
 945             /* REACTION-FIELD ELECTROSTATICS */
 946             velec            = _mm_mul_pd(qq33,_mm_sub_pd(_mm_add_pd(rinv33,_mm_mul_pd(krf,rsq33)),crf));
 947             felec            = _mm_mul_pd(qq33,_mm_sub_pd(_mm_mul_pd(rinv33,rinvsq33),krf2));
 948
 949             /* Update potential sum for this i atom from the interaction with this j atom. */
 950             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 951             velecsum         = _mm_add_pd(velecsum,velec);
 952
 953             fscal            = felec;
 954
 955             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 956
 957             /* Calculate temporary vectorial force */
 958             tx               = _mm_mul_pd(fscal,dx33);
 959             ty               = _mm_mul_pd(fscal,dy33);
 960             tz               = _mm_mul_pd(fscal,dz33);
 961
 962             /* Update vectorial force */
 963             fix3             = _mm_add_pd(fix3,tx);
 964             fiy3             = _mm_add_pd(fiy3,ty);
 965             fiz3             = _mm_add_pd(fiz3,tz);
 966
 967             fjx3             = _mm_add_pd(fjx3,tx);
 968             fjy3             = _mm_add_pd(fjy3,ty);
 969             fjz3             = _mm_add_pd(fjz3,tz);
 970
 971             gmx_mm_decrement_4rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 972
 973             /* Inner loop uses 323 flops */
 974         }
 975
 976         /* End of innermost loop */
 977
 978         gmx_mm_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 979                                               f+i_coord_offset,fshift+i_shift_offset);
 980
 981         ggid                        = gid[iidx];
 982         /* Update potential energies */
 983         gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
 984         gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
 985
 986         /* Increment number of inner iterations */
 987         inneriter                  += j_index_end - j_index_start;
 988
 989         /* Outer loop uses 26 flops */
 990     }
 991
 992     /* Increment number of outer iterations */
 993     outeriter        += nri;
 994
 995     /* Update outer/inner flops */
 996
 997     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*323);
 998 }
 999 /*
1000  * Gromacs nonbonded kernel:   nb_kernel_ElecRF_VdwLJ_GeomW4W4_F_sse4_1_double
1001  * Electrostatics interaction: ReactionField
1002  * VdW interaction:            LennardJones
1003  * Geometry:                   Water4-Water4
1004  * Calculate force/pot:        Force
1005  */
1006 void
1007 nb_kernel_ElecRF_VdwLJ_GeomW4W4_F_sse4_1_double
1008                     (t_nblist                    * gmx_restrict       nlist,
1009                      rvec                        * gmx_restrict          xx,
1010                      rvec                        * gmx_restrict          ff,
1011                      t_forcerec                  * gmx_restrict          fr,
1012                      t_mdatoms                   * gmx_restrict     mdatoms,
1013                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1014                      t_nrnb                      * gmx_restrict        nrnb)
1015 {
1016     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1017      * just 0 for non-waters.
1018      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
1019      * jnr indices corresponding to data put in the four positions in the SIMD register.
1020      */
1021     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1022     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1023     int              jnrA,jnrB;
1024     int              j_coord_offsetA,j_coord_offsetB;
1025     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1026     real             rcutoff_scalar;
1027     real             *shiftvec,*fshift,*x,*f;
1028     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1029     int              vdwioffset0;
1030     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1031     int              vdwioffset1;
1032     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1033     int              vdwioffset2;
1034     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1035     int              vdwioffset3;
1036     __m128d          ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1037     int              vdwjidx0A,vdwjidx0B;
1038     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1039     int              vdwjidx1A,vdwjidx1B;
1040     __m128d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1041     int              vdwjidx2A,vdwjidx2B;
1042     __m128d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1043     int              vdwjidx3A,vdwjidx3B;
1044     __m128d          jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1045     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1046     __m128d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1047     __m128d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1048     __m128d          dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1049     __m128d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1050     __m128d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1051     __m128d          dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1052     __m128d          dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1053     __m128d          dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1054     __m128d          dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1055     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
1056     real             *charge;
1057     int              nvdwtype;
1058     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1059     int              *vdwtype;
1060     real             *vdwparam;
1061     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
1062     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
1063     __m128d          dummy_mask,cutoff_mask;
1064     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
1065     __m128d          one     = _mm_set1_pd(1.0);
1066     __m128d          two     = _mm_set1_pd(2.0);
1067     x                = xx[0];
1068     f                = ff[0];
1069
1070     nri              = nlist->nri;
1071     iinr             = nlist->iinr;
1072     jindex           = nlist->jindex;
1073     jjnr             = nlist->jjnr;
1074     shiftidx         = nlist->shift;
1075     gid              = nlist->gid;
1076     shiftvec         = fr->shift_vec[0];
1077     fshift           = fr->fshift[0];
1078     facel            = _mm_set1_pd(fr->epsfac);
1079     charge           = mdatoms->chargeA;
1080     krf              = _mm_set1_pd(fr->ic->k_rf);
1081     krf2             = _mm_set1_pd(fr->ic->k_rf*2.0);
1082     crf              = _mm_set1_pd(fr->ic->c_rf);
1083     nvdwtype         = fr->ntype;
1084     vdwparam         = fr->nbfp;
1085     vdwtype          = mdatoms->typeA;
1086
1087     /* Setup water-specific parameters */
1088     inr              = nlist->iinr[0];
1089     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
1090     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
1091     iq3              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
1092     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
1093
1094     jq1              = _mm_set1_pd(charge[inr+1]);
1095     jq2              = _mm_set1_pd(charge[inr+2]);
1096     jq3              = _mm_set1_pd(charge[inr+3]);
1097     vdwjidx0A        = 2*vdwtype[inr+0];
1098     c6_00            = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
1099     c12_00           = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
1100     qq11             = _mm_mul_pd(iq1,jq1);
1101     qq12             = _mm_mul_pd(iq1,jq2);
1102     qq13             = _mm_mul_pd(iq1,jq3);
1103     qq21             = _mm_mul_pd(iq2,jq1);
1104     qq22             = _mm_mul_pd(iq2,jq2);
1105     qq23             = _mm_mul_pd(iq2,jq3);
1106     qq31             = _mm_mul_pd(iq3,jq1);
1107     qq32             = _mm_mul_pd(iq3,jq2);
1108     qq33             = _mm_mul_pd(iq3,jq3);
1109
1110     /* Avoid stupid compiler warnings */
1111     jnrA = jnrB = 0;
1112     j_coord_offsetA = 0;
1113     j_coord_offsetB = 0;
1114
1115     outeriter        = 0;
1116     inneriter        = 0;
1117
1118     /* Start outer loop over neighborlists */
1119     for(iidx=0; iidx<nri; iidx++)
1120     {
1121         /* Load shift vector for this list */
1122         i_shift_offset   = DIM*shiftidx[iidx];
1123
1124         /* Load limits for loop over neighbors */
1125         j_index_start    = jindex[iidx];
1126         j_index_end      = jindex[iidx+1];
1127
1128         /* Get outer coordinate index */
1129         inr              = iinr[iidx];
1130         i_coord_offset   = DIM*inr;
1131
1132         /* Load i particle coords and add shift vector */
1133         gmx_mm_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1134                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1135
1136         fix0             = _mm_setzero_pd();
1137         fiy0             = _mm_setzero_pd();
1138         fiz0             = _mm_setzero_pd();
1139         fix1             = _mm_setzero_pd();
1140         fiy1             = _mm_setzero_pd();
1141         fiz1             = _mm_setzero_pd();
1142         fix2             = _mm_setzero_pd();
1143         fiy2             = _mm_setzero_pd();
1144         fiz2             = _mm_setzero_pd();
1145         fix3             = _mm_setzero_pd();
1146         fiy3             = _mm_setzero_pd();
1147         fiz3             = _mm_setzero_pd();
1148
1149         /* Start inner kernel loop */
1150         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
1151         {
1152
1153             /* Get j neighbor index, and coordinate index */
1154             jnrA             = jjnr[jidx];
1155             jnrB             = jjnr[jidx+1];
1156             j_coord_offsetA  = DIM*jnrA;
1157             j_coord_offsetB  = DIM*jnrB;
1158
1159             /* load j atom coordinates */
1160             gmx_mm_load_4rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1161                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1162                                               &jy2,&jz2,&jx3,&jy3,&jz3);
1163
1164             /* Calculate displacement vector */
1165             dx00             = _mm_sub_pd(ix0,jx0);
1166             dy00             = _mm_sub_pd(iy0,jy0);
1167             dz00             = _mm_sub_pd(iz0,jz0);
1168             dx11             = _mm_sub_pd(ix1,jx1);
1169             dy11             = _mm_sub_pd(iy1,jy1);
1170             dz11             = _mm_sub_pd(iz1,jz1);
1171             dx12             = _mm_sub_pd(ix1,jx2);
1172             dy12             = _mm_sub_pd(iy1,jy2);
1173             dz12             = _mm_sub_pd(iz1,jz2);
1174             dx13             = _mm_sub_pd(ix1,jx3);
1175             dy13             = _mm_sub_pd(iy1,jy3);
1176             dz13             = _mm_sub_pd(iz1,jz3);
1177             dx21             = _mm_sub_pd(ix2,jx1);
1178             dy21             = _mm_sub_pd(iy2,jy1);
1179             dz21             = _mm_sub_pd(iz2,jz1);
1180             dx22             = _mm_sub_pd(ix2,jx2);
1181             dy22             = _mm_sub_pd(iy2,jy2);
1182             dz22             = _mm_sub_pd(iz2,jz2);
1183             dx23             = _mm_sub_pd(ix2,jx3);
1184             dy23             = _mm_sub_pd(iy2,jy3);
1185             dz23             = _mm_sub_pd(iz2,jz3);
1186             dx31             = _mm_sub_pd(ix3,jx1);
1187             dy31             = _mm_sub_pd(iy3,jy1);
1188             dz31             = _mm_sub_pd(iz3,jz1);
1189             dx32             = _mm_sub_pd(ix3,jx2);
1190             dy32             = _mm_sub_pd(iy3,jy2);
1191             dz32             = _mm_sub_pd(iz3,jz2);
1192             dx33             = _mm_sub_pd(ix3,jx3);
1193             dy33             = _mm_sub_pd(iy3,jy3);
1194             dz33             = _mm_sub_pd(iz3,jz3);
1195
1196             /* Calculate squared distance and things based on it */
1197             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1198             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1199             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1200             rsq13            = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
1201             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1202             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1203             rsq23            = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
1204             rsq31            = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
1205             rsq32            = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
1206             rsq33            = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
1207
1208             rinv11           = gmx_mm_invsqrt_pd(rsq11);
1209             rinv12           = gmx_mm_invsqrt_pd(rsq12);
1210             rinv13           = gmx_mm_invsqrt_pd(rsq13);
1211             rinv21           = gmx_mm_invsqrt_pd(rsq21);
1212             rinv22           = gmx_mm_invsqrt_pd(rsq22);
1213             rinv23           = gmx_mm_invsqrt_pd(rsq23);
1214             rinv31           = gmx_mm_invsqrt_pd(rsq31);
1215             rinv32           = gmx_mm_invsqrt_pd(rsq32);
1216             rinv33           = gmx_mm_invsqrt_pd(rsq33);
1217
1218             rinvsq00         = gmx_mm_inv_pd(rsq00);
1219             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
1220             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
1221             rinvsq13         = _mm_mul_pd(rinv13,rinv13);
1222             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
1223             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
1224             rinvsq23         = _mm_mul_pd(rinv23,rinv23);
1225             rinvsq31         = _mm_mul_pd(rinv31,rinv31);
1226             rinvsq32         = _mm_mul_pd(rinv32,rinv32);
1227             rinvsq33         = _mm_mul_pd(rinv33,rinv33);
1228
1229             fjx0             = _mm_setzero_pd();
1230             fjy0             = _mm_setzero_pd();
1231             fjz0             = _mm_setzero_pd();
1232             fjx1             = _mm_setzero_pd();
1233             fjy1             = _mm_setzero_pd();
1234             fjz1             = _mm_setzero_pd();
1235             fjx2             = _mm_setzero_pd();
1236             fjy2             = _mm_setzero_pd();
1237             fjz2             = _mm_setzero_pd();
1238             fjx3             = _mm_setzero_pd();
1239             fjy3             = _mm_setzero_pd();
1240             fjz3             = _mm_setzero_pd();
1241
1242             /**************************
1243              * CALCULATE INTERACTIONS *
1244              **************************/
1245
1246             /* LENNARD-JONES DISPERSION/REPULSION */
1247
1248             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1249             fvdw             = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
1250
1251             fscal            = fvdw;
1252
1253             /* Calculate temporary vectorial force */
1254             tx               = _mm_mul_pd(fscal,dx00);
1255             ty               = _mm_mul_pd(fscal,dy00);
1256             tz               = _mm_mul_pd(fscal,dz00);
1257
1258             /* Update vectorial force */
1259             fix0             = _mm_add_pd(fix0,tx);
1260             fiy0             = _mm_add_pd(fiy0,ty);
1261             fiz0             = _mm_add_pd(fiz0,tz);
1262
1263             fjx0             = _mm_add_pd(fjx0,tx);
1264             fjy0             = _mm_add_pd(fjy0,ty);
1265             fjz0             = _mm_add_pd(fjz0,tz);
1266
1267             /**************************
1268              * CALCULATE INTERACTIONS *
1269              **************************/
1270
1271             /* REACTION-FIELD ELECTROSTATICS */
1272             felec            = _mm_mul_pd(qq11,_mm_sub_pd(_mm_mul_pd(rinv11,rinvsq11),krf2));
1273
1274             fscal            = felec;
1275
1276             /* Calculate temporary vectorial force */
1277             tx               = _mm_mul_pd(fscal,dx11);
1278             ty               = _mm_mul_pd(fscal,dy11);
1279             tz               = _mm_mul_pd(fscal,dz11);
1280
1281             /* Update vectorial force */
1282             fix1             = _mm_add_pd(fix1,tx);
1283             fiy1             = _mm_add_pd(fiy1,ty);
1284             fiz1             = _mm_add_pd(fiz1,tz);
1285
1286             fjx1             = _mm_add_pd(fjx1,tx);
1287             fjy1             = _mm_add_pd(fjy1,ty);
1288             fjz1             = _mm_add_pd(fjz1,tz);
1289
1290             /**************************
1291              * CALCULATE INTERACTIONS *
1292              **************************/
1293
1294             /* REACTION-FIELD ELECTROSTATICS */
1295             felec            = _mm_mul_pd(qq12,_mm_sub_pd(_mm_mul_pd(rinv12,rinvsq12),krf2));
1296
1297             fscal            = felec;
1298
1299             /* Calculate temporary vectorial force */
1300             tx               = _mm_mul_pd(fscal,dx12);
1301             ty               = _mm_mul_pd(fscal,dy12);
1302             tz               = _mm_mul_pd(fscal,dz12);
1303
1304             /* Update vectorial force */
1305             fix1             = _mm_add_pd(fix1,tx);
1306             fiy1             = _mm_add_pd(fiy1,ty);
1307             fiz1             = _mm_add_pd(fiz1,tz);
1308
1309             fjx2             = _mm_add_pd(fjx2,tx);
1310             fjy2             = _mm_add_pd(fjy2,ty);
1311             fjz2             = _mm_add_pd(fjz2,tz);
1312
1313             /**************************
1314              * CALCULATE INTERACTIONS *
1315              **************************/
1316
1317             /* REACTION-FIELD ELECTROSTATICS */
1318             felec            = _mm_mul_pd(qq13,_mm_sub_pd(_mm_mul_pd(rinv13,rinvsq13),krf2));
1319
1320             fscal            = felec;
1321
1322             /* Calculate temporary vectorial force */
1323             tx               = _mm_mul_pd(fscal,dx13);
1324             ty               = _mm_mul_pd(fscal,dy13);
1325             tz               = _mm_mul_pd(fscal,dz13);
1326
1327             /* Update vectorial force */
1328             fix1             = _mm_add_pd(fix1,tx);
1329             fiy1             = _mm_add_pd(fiy1,ty);
1330             fiz1             = _mm_add_pd(fiz1,tz);
1331
1332             fjx3             = _mm_add_pd(fjx3,tx);
1333             fjy3             = _mm_add_pd(fjy3,ty);
1334             fjz3             = _mm_add_pd(fjz3,tz);
1335
1336             /**************************
1337              * CALCULATE INTERACTIONS *
1338              **************************/
1339
1340             /* REACTION-FIELD ELECTROSTATICS */
1341             felec            = _mm_mul_pd(qq21,_mm_sub_pd(_mm_mul_pd(rinv21,rinvsq21),krf2));
1342
1343             fscal            = felec;
1344
1345             /* Calculate temporary vectorial force */
1346             tx               = _mm_mul_pd(fscal,dx21);
1347             ty               = _mm_mul_pd(fscal,dy21);
1348             tz               = _mm_mul_pd(fscal,dz21);
1349
1350             /* Update vectorial force */
1351             fix2             = _mm_add_pd(fix2,tx);
1352             fiy2             = _mm_add_pd(fiy2,ty);
1353             fiz2             = _mm_add_pd(fiz2,tz);
1354
1355             fjx1             = _mm_add_pd(fjx1,tx);
1356             fjy1             = _mm_add_pd(fjy1,ty);
1357             fjz1             = _mm_add_pd(fjz1,tz);
1358
1359             /**************************
1360              * CALCULATE INTERACTIONS *
1361              **************************/
1362
1363             /* REACTION-FIELD ELECTROSTATICS */
1364             felec            = _mm_mul_pd(qq22,_mm_sub_pd(_mm_mul_pd(rinv22,rinvsq22),krf2));
1365
1366             fscal            = felec;
1367
1368             /* Calculate temporary vectorial force */
1369             tx               = _mm_mul_pd(fscal,dx22);
1370             ty               = _mm_mul_pd(fscal,dy22);
1371             tz               = _mm_mul_pd(fscal,dz22);
1372
1373             /* Update vectorial force */
1374             fix2             = _mm_add_pd(fix2,tx);
1375             fiy2             = _mm_add_pd(fiy2,ty);
1376             fiz2             = _mm_add_pd(fiz2,tz);
1377
1378             fjx2             = _mm_add_pd(fjx2,tx);
1379             fjy2             = _mm_add_pd(fjy2,ty);
1380             fjz2             = _mm_add_pd(fjz2,tz);
1381
1382             /**************************
1383              * CALCULATE INTERACTIONS *
1384              **************************/
1385
1386             /* REACTION-FIELD ELECTROSTATICS */
1387             felec            = _mm_mul_pd(qq23,_mm_sub_pd(_mm_mul_pd(rinv23,rinvsq23),krf2));
1388
1389             fscal            = felec;
1390
1391             /* Calculate temporary vectorial force */
1392             tx               = _mm_mul_pd(fscal,dx23);
1393             ty               = _mm_mul_pd(fscal,dy23);
1394             tz               = _mm_mul_pd(fscal,dz23);
1395
1396             /* Update vectorial force */
1397             fix2             = _mm_add_pd(fix2,tx);
1398             fiy2             = _mm_add_pd(fiy2,ty);
1399             fiz2             = _mm_add_pd(fiz2,tz);
1400
1401             fjx3             = _mm_add_pd(fjx3,tx);
1402             fjy3             = _mm_add_pd(fjy3,ty);
1403             fjz3             = _mm_add_pd(fjz3,tz);
1404
1405             /**************************
1406              * CALCULATE INTERACTIONS *
1407              **************************/
1408
1409             /* REACTION-FIELD ELECTROSTATICS */
1410             felec            = _mm_mul_pd(qq31,_mm_sub_pd(_mm_mul_pd(rinv31,rinvsq31),krf2));
1411
1412             fscal            = felec;
1413
1414             /* Calculate temporary vectorial force */
1415             tx               = _mm_mul_pd(fscal,dx31);
1416             ty               = _mm_mul_pd(fscal,dy31);
1417             tz               = _mm_mul_pd(fscal,dz31);
1418
1419             /* Update vectorial force */
1420             fix3             = _mm_add_pd(fix3,tx);
1421             fiy3             = _mm_add_pd(fiy3,ty);
1422             fiz3             = _mm_add_pd(fiz3,tz);
1423
1424             fjx1             = _mm_add_pd(fjx1,tx);
1425             fjy1             = _mm_add_pd(fjy1,ty);
1426             fjz1             = _mm_add_pd(fjz1,tz);
1427
1428             /**************************
1429              * CALCULATE INTERACTIONS *
1430              **************************/
1431
1432             /* REACTION-FIELD ELECTROSTATICS */
1433             felec            = _mm_mul_pd(qq32,_mm_sub_pd(_mm_mul_pd(rinv32,rinvsq32),krf2));
1434
1435             fscal            = felec;
1436
1437             /* Calculate temporary vectorial force */
1438             tx               = _mm_mul_pd(fscal,dx32);
1439             ty               = _mm_mul_pd(fscal,dy32);
1440             tz               = _mm_mul_pd(fscal,dz32);
1441
1442             /* Update vectorial force */
1443             fix3             = _mm_add_pd(fix3,tx);
1444             fiy3             = _mm_add_pd(fiy3,ty);
1445             fiz3             = _mm_add_pd(fiz3,tz);
1446
1447             fjx2             = _mm_add_pd(fjx2,tx);
1448             fjy2             = _mm_add_pd(fjy2,ty);
1449             fjz2             = _mm_add_pd(fjz2,tz);
1450
1451             /**************************
1452              * CALCULATE INTERACTIONS *
1453              **************************/
1454
1455             /* REACTION-FIELD ELECTROSTATICS */
1456             felec            = _mm_mul_pd(qq33,_mm_sub_pd(_mm_mul_pd(rinv33,rinvsq33),krf2));
1457
1458             fscal            = felec;
1459
1460             /* Calculate temporary vectorial force */
1461             tx               = _mm_mul_pd(fscal,dx33);
1462             ty               = _mm_mul_pd(fscal,dy33);
1463             tz               = _mm_mul_pd(fscal,dz33);
1464
1465             /* Update vectorial force */
1466             fix3             = _mm_add_pd(fix3,tx);
1467             fiy3             = _mm_add_pd(fiy3,ty);
1468             fiz3             = _mm_add_pd(fiz3,tz);
1469
1470             fjx3             = _mm_add_pd(fjx3,tx);
1471             fjy3             = _mm_add_pd(fjy3,ty);
1472             fjz3             = _mm_add_pd(fjz3,tz);
1473
1474             gmx_mm_decrement_4rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1475
1476             /* Inner loop uses 273 flops */
1477         }
1478
1479         if(jidx<j_index_end)
1480         {
1481
1482             jnrA             = jjnr[jidx];
1483             j_coord_offsetA  = DIM*jnrA;
1484
1485             /* load j atom coordinates */
1486             gmx_mm_load_4rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
1487                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1488                                               &jy2,&jz2,&jx3,&jy3,&jz3);
1489
1490             /* Calculate displacement vector */
1491             dx00             = _mm_sub_pd(ix0,jx0);
1492             dy00             = _mm_sub_pd(iy0,jy0);
1493             dz00             = _mm_sub_pd(iz0,jz0);
1494             dx11             = _mm_sub_pd(ix1,jx1);
1495             dy11             = _mm_sub_pd(iy1,jy1);
1496             dz11             = _mm_sub_pd(iz1,jz1);
1497             dx12             = _mm_sub_pd(ix1,jx2);
1498             dy12             = _mm_sub_pd(iy1,jy2);
1499             dz12             = _mm_sub_pd(iz1,jz2);
1500             dx13             = _mm_sub_pd(ix1,jx3);
1501             dy13             = _mm_sub_pd(iy1,jy3);
1502             dz13             = _mm_sub_pd(iz1,jz3);
1503             dx21             = _mm_sub_pd(ix2,jx1);
1504             dy21             = _mm_sub_pd(iy2,jy1);
1505             dz21             = _mm_sub_pd(iz2,jz1);
1506             dx22             = _mm_sub_pd(ix2,jx2);
1507             dy22             = _mm_sub_pd(iy2,jy2);
1508             dz22             = _mm_sub_pd(iz2,jz2);
1509             dx23             = _mm_sub_pd(ix2,jx3);
1510             dy23             = _mm_sub_pd(iy2,jy3);
1511             dz23             = _mm_sub_pd(iz2,jz3);
1512             dx31             = _mm_sub_pd(ix3,jx1);
1513             dy31             = _mm_sub_pd(iy3,jy1);
1514             dz31             = _mm_sub_pd(iz3,jz1);
1515             dx32             = _mm_sub_pd(ix3,jx2);
1516             dy32             = _mm_sub_pd(iy3,jy2);
1517             dz32             = _mm_sub_pd(iz3,jz2);
1518             dx33             = _mm_sub_pd(ix3,jx3);
1519             dy33             = _mm_sub_pd(iy3,jy3);
1520             dz33             = _mm_sub_pd(iz3,jz3);
1521
1522             /* Calculate squared distance and things based on it */
1523             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1524             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1525             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1526             rsq13            = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
1527             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1528             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1529             rsq23            = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
1530             rsq31            = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
1531             rsq32            = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
1532             rsq33            = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
1533
1534             rinv11           = gmx_mm_invsqrt_pd(rsq11);
1535             rinv12           = gmx_mm_invsqrt_pd(rsq12);
1536             rinv13           = gmx_mm_invsqrt_pd(rsq13);
1537             rinv21           = gmx_mm_invsqrt_pd(rsq21);
1538             rinv22           = gmx_mm_invsqrt_pd(rsq22);
1539             rinv23           = gmx_mm_invsqrt_pd(rsq23);
1540             rinv31           = gmx_mm_invsqrt_pd(rsq31);
1541             rinv32           = gmx_mm_invsqrt_pd(rsq32);
1542             rinv33           = gmx_mm_invsqrt_pd(rsq33);
1543
1544             rinvsq00         = gmx_mm_inv_pd(rsq00);
1545             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
1546             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
1547             rinvsq13         = _mm_mul_pd(rinv13,rinv13);
1548             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
1549             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
1550             rinvsq23         = _mm_mul_pd(rinv23,rinv23);
1551             rinvsq31         = _mm_mul_pd(rinv31,rinv31);
1552             rinvsq32         = _mm_mul_pd(rinv32,rinv32);
1553             rinvsq33         = _mm_mul_pd(rinv33,rinv33);
1554
1555             fjx0             = _mm_setzero_pd();
1556             fjy0             = _mm_setzero_pd();
1557             fjz0             = _mm_setzero_pd();
1558             fjx1             = _mm_setzero_pd();
1559             fjy1             = _mm_setzero_pd();
1560             fjz1             = _mm_setzero_pd();
1561             fjx2             = _mm_setzero_pd();
1562             fjy2             = _mm_setzero_pd();
1563             fjz2             = _mm_setzero_pd();
1564             fjx3             = _mm_setzero_pd();
1565             fjy3             = _mm_setzero_pd();
1566             fjz3             = _mm_setzero_pd();
1567
1568             /**************************
1569              * CALCULATE INTERACTIONS *
1570              **************************/
1571
1572             /* LENNARD-JONES DISPERSION/REPULSION */
1573
1574             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1575             fvdw             = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
1576
1577             fscal            = fvdw;
1578
1579             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1580
1581             /* Calculate temporary vectorial force */
1582             tx               = _mm_mul_pd(fscal,dx00);
1583             ty               = _mm_mul_pd(fscal,dy00);
1584             tz               = _mm_mul_pd(fscal,dz00);
1585
1586             /* Update vectorial force */
1587             fix0             = _mm_add_pd(fix0,tx);
1588             fiy0             = _mm_add_pd(fiy0,ty);
1589             fiz0             = _mm_add_pd(fiz0,tz);
1590
1591             fjx0             = _mm_add_pd(fjx0,tx);
1592             fjy0             = _mm_add_pd(fjy0,ty);
1593             fjz0             = _mm_add_pd(fjz0,tz);
1594
1595             /**************************
1596              * CALCULATE INTERACTIONS *
1597              **************************/
1598
1599             /* REACTION-FIELD ELECTROSTATICS */
1600             felec            = _mm_mul_pd(qq11,_mm_sub_pd(_mm_mul_pd(rinv11,rinvsq11),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,dx11);
1608             ty               = _mm_mul_pd(fscal,dy11);
1609             tz               = _mm_mul_pd(fscal,dz11);
1610
1611             /* Update vectorial force */
1612             fix1             = _mm_add_pd(fix1,tx);
1613             fiy1             = _mm_add_pd(fiy1,ty);
1614             fiz1             = _mm_add_pd(fiz1,tz);
1615
1616             fjx1             = _mm_add_pd(fjx1,tx);
1617             fjy1             = _mm_add_pd(fjy1,ty);
1618             fjz1             = _mm_add_pd(fjz1,tz);
1619
1620             /**************************
1621              * CALCULATE INTERACTIONS *
1622              **************************/
1623
1624             /* REACTION-FIELD ELECTROSTATICS */
1625             felec            = _mm_mul_pd(qq12,_mm_sub_pd(_mm_mul_pd(rinv12,rinvsq12),krf2));
1626
1627             fscal            = felec;
1628
1629             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1630
1631             /* Calculate temporary vectorial force */
1632             tx               = _mm_mul_pd(fscal,dx12);
1633             ty               = _mm_mul_pd(fscal,dy12);
1634             tz               = _mm_mul_pd(fscal,dz12);
1635
1636             /* Update vectorial force */
1637             fix1             = _mm_add_pd(fix1,tx);
1638             fiy1             = _mm_add_pd(fiy1,ty);
1639             fiz1             = _mm_add_pd(fiz1,tz);
1640
1641             fjx2             = _mm_add_pd(fjx2,tx);
1642             fjy2             = _mm_add_pd(fjy2,ty);
1643             fjz2             = _mm_add_pd(fjz2,tz);
1644
1645             /**************************
1646              * CALCULATE INTERACTIONS *
1647              **************************/
1648
1649             /* REACTION-FIELD ELECTROSTATICS */
1650             felec            = _mm_mul_pd(qq13,_mm_sub_pd(_mm_mul_pd(rinv13,rinvsq13),krf2));
1651
1652             fscal            = felec;
1653
1654             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1655
1656             /* Calculate temporary vectorial force */
1657             tx               = _mm_mul_pd(fscal,dx13);
1658             ty               = _mm_mul_pd(fscal,dy13);
1659             tz               = _mm_mul_pd(fscal,dz13);
1660
1661             /* Update vectorial force */
1662             fix1             = _mm_add_pd(fix1,tx);
1663             fiy1             = _mm_add_pd(fiy1,ty);
1664             fiz1             = _mm_add_pd(fiz1,tz);
1665
1666             fjx3             = _mm_add_pd(fjx3,tx);
1667             fjy3             = _mm_add_pd(fjy3,ty);
1668             fjz3             = _mm_add_pd(fjz3,tz);
1669
1670             /**************************
1671              * CALCULATE INTERACTIONS *
1672              **************************/
1673
1674             /* REACTION-FIELD ELECTROSTATICS */
1675             felec            = _mm_mul_pd(qq21,_mm_sub_pd(_mm_mul_pd(rinv21,rinvsq21),krf2));
1676
1677             fscal            = felec;
1678
1679             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1680
1681             /* Calculate temporary vectorial force */
1682             tx               = _mm_mul_pd(fscal,dx21);
1683             ty               = _mm_mul_pd(fscal,dy21);
1684             tz               = _mm_mul_pd(fscal,dz21);
1685
1686             /* Update vectorial force */
1687             fix2             = _mm_add_pd(fix2,tx);
1688             fiy2             = _mm_add_pd(fiy2,ty);
1689             fiz2             = _mm_add_pd(fiz2,tz);
1690
1691             fjx1             = _mm_add_pd(fjx1,tx);
1692             fjy1             = _mm_add_pd(fjy1,ty);
1693             fjz1             = _mm_add_pd(fjz1,tz);
1694
1695             /**************************
1696              * CALCULATE INTERACTIONS *
1697              **************************/
1698
1699             /* REACTION-FIELD ELECTROSTATICS */
1700             felec            = _mm_mul_pd(qq22,_mm_sub_pd(_mm_mul_pd(rinv22,rinvsq22),krf2));
1701
1702             fscal            = felec;
1703
1704             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1705
1706             /* Calculate temporary vectorial force */
1707             tx               = _mm_mul_pd(fscal,dx22);
1708             ty               = _mm_mul_pd(fscal,dy22);
1709             tz               = _mm_mul_pd(fscal,dz22);
1710
1711             /* Update vectorial force */
1712             fix2             = _mm_add_pd(fix2,tx);
1713             fiy2             = _mm_add_pd(fiy2,ty);
1714             fiz2             = _mm_add_pd(fiz2,tz);
1715
1716             fjx2             = _mm_add_pd(fjx2,tx);
1717             fjy2             = _mm_add_pd(fjy2,ty);
1718             fjz2             = _mm_add_pd(fjz2,tz);
1719
1720             /**************************
1721              * CALCULATE INTERACTIONS *
1722              **************************/
1723
1724             /* REACTION-FIELD ELECTROSTATICS */
1725             felec            = _mm_mul_pd(qq23,_mm_sub_pd(_mm_mul_pd(rinv23,rinvsq23),krf2));
1726
1727             fscal            = felec;
1728
1729             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1730
1731             /* Calculate temporary vectorial force */
1732             tx               = _mm_mul_pd(fscal,dx23);
1733             ty               = _mm_mul_pd(fscal,dy23);
1734             tz               = _mm_mul_pd(fscal,dz23);
1735
1736             /* Update vectorial force */
1737             fix2             = _mm_add_pd(fix2,tx);
1738             fiy2             = _mm_add_pd(fiy2,ty);
1739             fiz2             = _mm_add_pd(fiz2,tz);
1740
1741             fjx3             = _mm_add_pd(fjx3,tx);
1742             fjy3             = _mm_add_pd(fjy3,ty);
1743             fjz3             = _mm_add_pd(fjz3,tz);
1744
1745             /**************************
1746              * CALCULATE INTERACTIONS *
1747              **************************/
1748
1749             /* REACTION-FIELD ELECTROSTATICS */
1750             felec            = _mm_mul_pd(qq31,_mm_sub_pd(_mm_mul_pd(rinv31,rinvsq31),krf2));
1751
1752             fscal            = felec;
1753
1754             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1755
1756             /* Calculate temporary vectorial force */
1757             tx               = _mm_mul_pd(fscal,dx31);
1758             ty               = _mm_mul_pd(fscal,dy31);
1759             tz               = _mm_mul_pd(fscal,dz31);
1760
1761             /* Update vectorial force */
1762             fix3             = _mm_add_pd(fix3,tx);
1763             fiy3             = _mm_add_pd(fiy3,ty);
1764             fiz3             = _mm_add_pd(fiz3,tz);
1765
1766             fjx1             = _mm_add_pd(fjx1,tx);
1767             fjy1             = _mm_add_pd(fjy1,ty);
1768             fjz1             = _mm_add_pd(fjz1,tz);
1769
1770             /**************************
1771              * CALCULATE INTERACTIONS *
1772              **************************/
1773
1774             /* REACTION-FIELD ELECTROSTATICS */
1775             felec            = _mm_mul_pd(qq32,_mm_sub_pd(_mm_mul_pd(rinv32,rinvsq32),krf2));
1776
1777             fscal            = felec;
1778
1779             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1780
1781             /* Calculate temporary vectorial force */
1782             tx               = _mm_mul_pd(fscal,dx32);
1783             ty               = _mm_mul_pd(fscal,dy32);
1784             tz               = _mm_mul_pd(fscal,dz32);
1785
1786             /* Update vectorial force */
1787             fix3             = _mm_add_pd(fix3,tx);
1788             fiy3             = _mm_add_pd(fiy3,ty);
1789             fiz3             = _mm_add_pd(fiz3,tz);
1790
1791             fjx2             = _mm_add_pd(fjx2,tx);
1792             fjy2             = _mm_add_pd(fjy2,ty);
1793             fjz2             = _mm_add_pd(fjz2,tz);
1794
1795             /**************************
1796              * CALCULATE INTERACTIONS *
1797              **************************/
1798
1799             /* REACTION-FIELD ELECTROSTATICS */
1800             felec            = _mm_mul_pd(qq33,_mm_sub_pd(_mm_mul_pd(rinv33,rinvsq33),krf2));
1801
1802             fscal            = felec;
1803
1804             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1805
1806             /* Calculate temporary vectorial force */
1807             tx               = _mm_mul_pd(fscal,dx33);
1808             ty               = _mm_mul_pd(fscal,dy33);
1809             tz               = _mm_mul_pd(fscal,dz33);
1810
1811             /* Update vectorial force */
1812             fix3             = _mm_add_pd(fix3,tx);
1813             fiy3             = _mm_add_pd(fiy3,ty);
1814             fiz3             = _mm_add_pd(fiz3,tz);
1815
1816             fjx3             = _mm_add_pd(fjx3,tx);
1817             fjy3             = _mm_add_pd(fjy3,ty);
1818             fjz3             = _mm_add_pd(fjz3,tz);
1819
1820             gmx_mm_decrement_4rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1821
1822             /* Inner loop uses 273 flops */
1823         }
1824
1825         /* End of innermost loop */
1826
1827         gmx_mm_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1828                                               f+i_coord_offset,fshift+i_shift_offset);
1829
1830         /* Increment number of inner iterations */
1831         inneriter                  += j_index_end - j_index_start;
1832
1833         /* Outer loop uses 24 flops */
1834     }
1835
1836     /* Increment number of outer iterations */
1837     outeriter        += nri;
1838
1839     /* Update outer/inner flops */
1840
1841     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*273);
1842 }