src/gromacs/gmxlib/nonbonded/nb_kernel_sse4_1_double/nb_kernel_ElecRF_VdwNone_GeomW3W3_sse4_1_double.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS sse4_1_double kernel generator.
  37  */
  38 #include "config.h"
  39
  40 #include <math.h>
  41
  42 #include "../nb_kernel.h"
  43 #include "types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "nrnb.h"
  46
  47 #include "gromacs/simd/math_x86_sse4_1_double.h"
  48 #include "kernelutil_x86_sse4_1_double.h"
  49
  50 /*
  51  * Gromacs nonbonded kernel:   nb_kernel_ElecRF_VdwNone_GeomW3W3_VF_sse4_1_double
  52  * Electrostatics interaction: ReactionField
  53  * VdW interaction:            None
  54  * Geometry:                   Water3-Water3
  55  * Calculate force/pot:        PotentialAndForce
  56  */
  57 void
  58 nb_kernel_ElecRF_VdwNone_GeomW3W3_VF_sse4_1_double
  59                     (t_nblist                    * gmx_restrict       nlist,
  60                      rvec                        * gmx_restrict          xx,
  61                      rvec                        * gmx_restrict          ff,
  62                      t_forcerec                  * gmx_restrict          fr,
  63                      t_mdatoms                   * gmx_restrict     mdatoms,
  64                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  65                      t_nrnb                      * gmx_restrict        nrnb)
  66 {
  67     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  68      * just 0 for non-waters.
  69      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
  70      * jnr indices corresponding to data put in the four positions in the SIMD register.
  71      */
  72     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  73     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  74     int              jnrA,jnrB;
  75     int              j_coord_offsetA,j_coord_offsetB;
  76     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  77     real             rcutoff_scalar;
  78     real             *shiftvec,*fshift,*x,*f;
  79     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  80     int              vdwioffset0;
  81     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  82     int              vdwioffset1;
  83     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  84     int              vdwioffset2;
  85     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  86     int              vdwjidx0A,vdwjidx0B;
  87     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  88     int              vdwjidx1A,vdwjidx1B;
  89     __m128d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  90     int              vdwjidx2A,vdwjidx2B;
  91     __m128d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  92     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  93     __m128d          dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
  94     __m128d          dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
  95     __m128d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  96     __m128d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
  97     __m128d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
  98     __m128d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  99     __m128d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 100     __m128d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 101     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
 102     real             *charge;
 103     __m128d          dummy_mask,cutoff_mask;
 104     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 105     __m128d          one     = _mm_set1_pd(1.0);
 106     __m128d          two     = _mm_set1_pd(2.0);
 107     x                = xx[0];
 108     f                = ff[0];
 109
 110     nri              = nlist->nri;
 111     iinr             = nlist->iinr;
 112     jindex           = nlist->jindex;
 113     jjnr             = nlist->jjnr;
 114     shiftidx         = nlist->shift;
 115     gid              = nlist->gid;
 116     shiftvec         = fr->shift_vec[0];
 117     fshift           = fr->fshift[0];
 118     facel            = _mm_set1_pd(fr->epsfac);
 119     charge           = mdatoms->chargeA;
 120     krf              = _mm_set1_pd(fr->ic->k_rf);
 121     krf2             = _mm_set1_pd(fr->ic->k_rf*2.0);
 122     crf              = _mm_set1_pd(fr->ic->c_rf);
 123
 124     /* Setup water-specific parameters */
 125     inr              = nlist->iinr[0];
 126     iq0              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
 127     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
 128     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
 129
 130     jq0              = _mm_set1_pd(charge[inr+0]);
 131     jq1              = _mm_set1_pd(charge[inr+1]);
 132     jq2              = _mm_set1_pd(charge[inr+2]);
 133     qq00             = _mm_mul_pd(iq0,jq0);
 134     qq01             = _mm_mul_pd(iq0,jq1);
 135     qq02             = _mm_mul_pd(iq0,jq2);
 136     qq10             = _mm_mul_pd(iq1,jq0);
 137     qq11             = _mm_mul_pd(iq1,jq1);
 138     qq12             = _mm_mul_pd(iq1,jq2);
 139     qq20             = _mm_mul_pd(iq2,jq0);
 140     qq21             = _mm_mul_pd(iq2,jq1);
 141     qq22             = _mm_mul_pd(iq2,jq2);
 142
 143     /* Avoid stupid compiler warnings */
 144     jnrA = jnrB = 0;
 145     j_coord_offsetA = 0;
 146     j_coord_offsetB = 0;
 147
 148     outeriter        = 0;
 149     inneriter        = 0;
 150
 151     /* Start outer loop over neighborlists */
 152     for(iidx=0; iidx<nri; iidx++)
 153     {
 154         /* Load shift vector for this list */
 155         i_shift_offset   = DIM*shiftidx[iidx];
 156
 157         /* Load limits for loop over neighbors */
 158         j_index_start    = jindex[iidx];
 159         j_index_end      = jindex[iidx+1];
 160
 161         /* Get outer coordinate index */
 162         inr              = iinr[iidx];
 163         i_coord_offset   = DIM*inr;
 164
 165         /* Load i particle coords and add shift vector */
 166         gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 167                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 168
 169         fix0             = _mm_setzero_pd();
 170         fiy0             = _mm_setzero_pd();
 171         fiz0             = _mm_setzero_pd();
 172         fix1             = _mm_setzero_pd();
 173         fiy1             = _mm_setzero_pd();
 174         fiz1             = _mm_setzero_pd();
 175         fix2             = _mm_setzero_pd();
 176         fiy2             = _mm_setzero_pd();
 177         fiz2             = _mm_setzero_pd();
 178
 179         /* Reset potential sums */
 180         velecsum         = _mm_setzero_pd();
 181
 182         /* Start inner kernel loop */
 183         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 184         {
 185
 186             /* Get j neighbor index, and coordinate index */
 187             jnrA             = jjnr[jidx];
 188             jnrB             = jjnr[jidx+1];
 189             j_coord_offsetA  = DIM*jnrA;
 190             j_coord_offsetB  = DIM*jnrB;
 191
 192             /* load j atom coordinates */
 193             gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 194                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 195
 196             /* Calculate displacement vector */
 197             dx00             = _mm_sub_pd(ix0,jx0);
 198             dy00             = _mm_sub_pd(iy0,jy0);
 199             dz00             = _mm_sub_pd(iz0,jz0);
 200             dx01             = _mm_sub_pd(ix0,jx1);
 201             dy01             = _mm_sub_pd(iy0,jy1);
 202             dz01             = _mm_sub_pd(iz0,jz1);
 203             dx02             = _mm_sub_pd(ix0,jx2);
 204             dy02             = _mm_sub_pd(iy0,jy2);
 205             dz02             = _mm_sub_pd(iz0,jz2);
 206             dx10             = _mm_sub_pd(ix1,jx0);
 207             dy10             = _mm_sub_pd(iy1,jy0);
 208             dz10             = _mm_sub_pd(iz1,jz0);
 209             dx11             = _mm_sub_pd(ix1,jx1);
 210             dy11             = _mm_sub_pd(iy1,jy1);
 211             dz11             = _mm_sub_pd(iz1,jz1);
 212             dx12             = _mm_sub_pd(ix1,jx2);
 213             dy12             = _mm_sub_pd(iy1,jy2);
 214             dz12             = _mm_sub_pd(iz1,jz2);
 215             dx20             = _mm_sub_pd(ix2,jx0);
 216             dy20             = _mm_sub_pd(iy2,jy0);
 217             dz20             = _mm_sub_pd(iz2,jz0);
 218             dx21             = _mm_sub_pd(ix2,jx1);
 219             dy21             = _mm_sub_pd(iy2,jy1);
 220             dz21             = _mm_sub_pd(iz2,jz1);
 221             dx22             = _mm_sub_pd(ix2,jx2);
 222             dy22             = _mm_sub_pd(iy2,jy2);
 223             dz22             = _mm_sub_pd(iz2,jz2);
 224
 225             /* Calculate squared distance and things based on it */
 226             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 227             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
 228             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
 229             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 230             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 231             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 232             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 233             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 234             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 235
 236             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 237             rinv01           = gmx_mm_invsqrt_pd(rsq01);
 238             rinv02           = gmx_mm_invsqrt_pd(rsq02);
 239             rinv10           = gmx_mm_invsqrt_pd(rsq10);
 240             rinv11           = gmx_mm_invsqrt_pd(rsq11);
 241             rinv12           = gmx_mm_invsqrt_pd(rsq12);
 242             rinv20           = gmx_mm_invsqrt_pd(rsq20);
 243             rinv21           = gmx_mm_invsqrt_pd(rsq21);
 244             rinv22           = gmx_mm_invsqrt_pd(rsq22);
 245
 246             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 247             rinvsq01         = _mm_mul_pd(rinv01,rinv01);
 248             rinvsq02         = _mm_mul_pd(rinv02,rinv02);
 249             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
 250             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
 251             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
 252             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
 253             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
 254             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
 255
 256             fjx0             = _mm_setzero_pd();
 257             fjy0             = _mm_setzero_pd();
 258             fjz0             = _mm_setzero_pd();
 259             fjx1             = _mm_setzero_pd();
 260             fjy1             = _mm_setzero_pd();
 261             fjz1             = _mm_setzero_pd();
 262             fjx2             = _mm_setzero_pd();
 263             fjy2             = _mm_setzero_pd();
 264             fjz2             = _mm_setzero_pd();
 265
 266             /**************************
 267              * CALCULATE INTERACTIONS *
 268              **************************/
 269
 270             /* REACTION-FIELD ELECTROSTATICS */
 271             velec            = _mm_mul_pd(qq00,_mm_sub_pd(_mm_add_pd(rinv00,_mm_mul_pd(krf,rsq00)),crf));
 272             felec            = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
 273
 274             /* Update potential sum for this i atom from the interaction with this j atom. */
 275             velecsum         = _mm_add_pd(velecsum,velec);
 276
 277             fscal            = felec;
 278
 279             /* Calculate temporary vectorial force */
 280             tx               = _mm_mul_pd(fscal,dx00);
 281             ty               = _mm_mul_pd(fscal,dy00);
 282             tz               = _mm_mul_pd(fscal,dz00);
 283
 284             /* Update vectorial force */
 285             fix0             = _mm_add_pd(fix0,tx);
 286             fiy0             = _mm_add_pd(fiy0,ty);
 287             fiz0             = _mm_add_pd(fiz0,tz);
 288
 289             fjx0             = _mm_add_pd(fjx0,tx);
 290             fjy0             = _mm_add_pd(fjy0,ty);
 291             fjz0             = _mm_add_pd(fjz0,tz);
 292
 293             /**************************
 294              * CALCULATE INTERACTIONS *
 295              **************************/
 296
 297             /* REACTION-FIELD ELECTROSTATICS */
 298             velec            = _mm_mul_pd(qq01,_mm_sub_pd(_mm_add_pd(rinv01,_mm_mul_pd(krf,rsq01)),crf));
 299             felec            = _mm_mul_pd(qq01,_mm_sub_pd(_mm_mul_pd(rinv01,rinvsq01),krf2));
 300
 301             /* Update potential sum for this i atom from the interaction with this j atom. */
 302             velecsum         = _mm_add_pd(velecsum,velec);
 303
 304             fscal            = felec;
 305
 306             /* Calculate temporary vectorial force */
 307             tx               = _mm_mul_pd(fscal,dx01);
 308             ty               = _mm_mul_pd(fscal,dy01);
 309             tz               = _mm_mul_pd(fscal,dz01);
 310
 311             /* Update vectorial force */
 312             fix0             = _mm_add_pd(fix0,tx);
 313             fiy0             = _mm_add_pd(fiy0,ty);
 314             fiz0             = _mm_add_pd(fiz0,tz);
 315
 316             fjx1             = _mm_add_pd(fjx1,tx);
 317             fjy1             = _mm_add_pd(fjy1,ty);
 318             fjz1             = _mm_add_pd(fjz1,tz);
 319
 320             /**************************
 321              * CALCULATE INTERACTIONS *
 322              **************************/
 323
 324             /* REACTION-FIELD ELECTROSTATICS */
 325             velec            = _mm_mul_pd(qq02,_mm_sub_pd(_mm_add_pd(rinv02,_mm_mul_pd(krf,rsq02)),crf));
 326             felec            = _mm_mul_pd(qq02,_mm_sub_pd(_mm_mul_pd(rinv02,rinvsq02),krf2));
 327
 328             /* Update potential sum for this i atom from the interaction with this j atom. */
 329             velecsum         = _mm_add_pd(velecsum,velec);
 330
 331             fscal            = felec;
 332
 333             /* Calculate temporary vectorial force */
 334             tx               = _mm_mul_pd(fscal,dx02);
 335             ty               = _mm_mul_pd(fscal,dy02);
 336             tz               = _mm_mul_pd(fscal,dz02);
 337
 338             /* Update vectorial force */
 339             fix0             = _mm_add_pd(fix0,tx);
 340             fiy0             = _mm_add_pd(fiy0,ty);
 341             fiz0             = _mm_add_pd(fiz0,tz);
 342
 343             fjx2             = _mm_add_pd(fjx2,tx);
 344             fjy2             = _mm_add_pd(fjy2,ty);
 345             fjz2             = _mm_add_pd(fjz2,tz);
 346
 347             /**************************
 348              * CALCULATE INTERACTIONS *
 349              **************************/
 350
 351             /* REACTION-FIELD ELECTROSTATICS */
 352             velec            = _mm_mul_pd(qq10,_mm_sub_pd(_mm_add_pd(rinv10,_mm_mul_pd(krf,rsq10)),crf));
 353             felec            = _mm_mul_pd(qq10,_mm_sub_pd(_mm_mul_pd(rinv10,rinvsq10),krf2));
 354
 355             /* Update potential sum for this i atom from the interaction with this j atom. */
 356             velecsum         = _mm_add_pd(velecsum,velec);
 357
 358             fscal            = felec;
 359
 360             /* Calculate temporary vectorial force */
 361             tx               = _mm_mul_pd(fscal,dx10);
 362             ty               = _mm_mul_pd(fscal,dy10);
 363             tz               = _mm_mul_pd(fscal,dz10);
 364
 365             /* Update vectorial force */
 366             fix1             = _mm_add_pd(fix1,tx);
 367             fiy1             = _mm_add_pd(fiy1,ty);
 368             fiz1             = _mm_add_pd(fiz1,tz);
 369
 370             fjx0             = _mm_add_pd(fjx0,tx);
 371             fjy0             = _mm_add_pd(fjy0,ty);
 372             fjz0             = _mm_add_pd(fjz0,tz);
 373
 374             /**************************
 375              * CALCULATE INTERACTIONS *
 376              **************************/
 377
 378             /* REACTION-FIELD ELECTROSTATICS */
 379             velec            = _mm_mul_pd(qq11,_mm_sub_pd(_mm_add_pd(rinv11,_mm_mul_pd(krf,rsq11)),crf));
 380             felec            = _mm_mul_pd(qq11,_mm_sub_pd(_mm_mul_pd(rinv11,rinvsq11),krf2));
 381
 382             /* Update potential sum for this i atom from the interaction with this j atom. */
 383             velecsum         = _mm_add_pd(velecsum,velec);
 384
 385             fscal            = felec;
 386
 387             /* Calculate temporary vectorial force */
 388             tx               = _mm_mul_pd(fscal,dx11);
 389             ty               = _mm_mul_pd(fscal,dy11);
 390             tz               = _mm_mul_pd(fscal,dz11);
 391
 392             /* Update vectorial force */
 393             fix1             = _mm_add_pd(fix1,tx);
 394             fiy1             = _mm_add_pd(fiy1,ty);
 395             fiz1             = _mm_add_pd(fiz1,tz);
 396
 397             fjx1             = _mm_add_pd(fjx1,tx);
 398             fjy1             = _mm_add_pd(fjy1,ty);
 399             fjz1             = _mm_add_pd(fjz1,tz);
 400
 401             /**************************
 402              * CALCULATE INTERACTIONS *
 403              **************************/
 404
 405             /* REACTION-FIELD ELECTROSTATICS */
 406             velec            = _mm_mul_pd(qq12,_mm_sub_pd(_mm_add_pd(rinv12,_mm_mul_pd(krf,rsq12)),crf));
 407             felec            = _mm_mul_pd(qq12,_mm_sub_pd(_mm_mul_pd(rinv12,rinvsq12),krf2));
 408
 409             /* Update potential sum for this i atom from the interaction with this j atom. */
 410             velecsum         = _mm_add_pd(velecsum,velec);
 411
 412             fscal            = felec;
 413
 414             /* Calculate temporary vectorial force */
 415             tx               = _mm_mul_pd(fscal,dx12);
 416             ty               = _mm_mul_pd(fscal,dy12);
 417             tz               = _mm_mul_pd(fscal,dz12);
 418
 419             /* Update vectorial force */
 420             fix1             = _mm_add_pd(fix1,tx);
 421             fiy1             = _mm_add_pd(fiy1,ty);
 422             fiz1             = _mm_add_pd(fiz1,tz);
 423
 424             fjx2             = _mm_add_pd(fjx2,tx);
 425             fjy2             = _mm_add_pd(fjy2,ty);
 426             fjz2             = _mm_add_pd(fjz2,tz);
 427
 428             /**************************
 429              * CALCULATE INTERACTIONS *
 430              **************************/
 431
 432             /* REACTION-FIELD ELECTROSTATICS */
 433             velec            = _mm_mul_pd(qq20,_mm_sub_pd(_mm_add_pd(rinv20,_mm_mul_pd(krf,rsq20)),crf));
 434             felec            = _mm_mul_pd(qq20,_mm_sub_pd(_mm_mul_pd(rinv20,rinvsq20),krf2));
 435
 436             /* Update potential sum for this i atom from the interaction with this j atom. */
 437             velecsum         = _mm_add_pd(velecsum,velec);
 438
 439             fscal            = felec;
 440
 441             /* Calculate temporary vectorial force */
 442             tx               = _mm_mul_pd(fscal,dx20);
 443             ty               = _mm_mul_pd(fscal,dy20);
 444             tz               = _mm_mul_pd(fscal,dz20);
 445
 446             /* Update vectorial force */
 447             fix2             = _mm_add_pd(fix2,tx);
 448             fiy2             = _mm_add_pd(fiy2,ty);
 449             fiz2             = _mm_add_pd(fiz2,tz);
 450
 451             fjx0             = _mm_add_pd(fjx0,tx);
 452             fjy0             = _mm_add_pd(fjy0,ty);
 453             fjz0             = _mm_add_pd(fjz0,tz);
 454
 455             /**************************
 456              * CALCULATE INTERACTIONS *
 457              **************************/
 458
 459             /* REACTION-FIELD ELECTROSTATICS */
 460             velec            = _mm_mul_pd(qq21,_mm_sub_pd(_mm_add_pd(rinv21,_mm_mul_pd(krf,rsq21)),crf));
 461             felec            = _mm_mul_pd(qq21,_mm_sub_pd(_mm_mul_pd(rinv21,rinvsq21),krf2));
 462
 463             /* Update potential sum for this i atom from the interaction with this j atom. */
 464             velecsum         = _mm_add_pd(velecsum,velec);
 465
 466             fscal            = felec;
 467
 468             /* Calculate temporary vectorial force */
 469             tx               = _mm_mul_pd(fscal,dx21);
 470             ty               = _mm_mul_pd(fscal,dy21);
 471             tz               = _mm_mul_pd(fscal,dz21);
 472
 473             /* Update vectorial force */
 474             fix2             = _mm_add_pd(fix2,tx);
 475             fiy2             = _mm_add_pd(fiy2,ty);
 476             fiz2             = _mm_add_pd(fiz2,tz);
 477
 478             fjx1             = _mm_add_pd(fjx1,tx);
 479             fjy1             = _mm_add_pd(fjy1,ty);
 480             fjz1             = _mm_add_pd(fjz1,tz);
 481
 482             /**************************
 483              * CALCULATE INTERACTIONS *
 484              **************************/
 485
 486             /* REACTION-FIELD ELECTROSTATICS */
 487             velec            = _mm_mul_pd(qq22,_mm_sub_pd(_mm_add_pd(rinv22,_mm_mul_pd(krf,rsq22)),crf));
 488             felec            = _mm_mul_pd(qq22,_mm_sub_pd(_mm_mul_pd(rinv22,rinvsq22),krf2));
 489
 490             /* Update potential sum for this i atom from the interaction with this j atom. */
 491             velecsum         = _mm_add_pd(velecsum,velec);
 492
 493             fscal            = felec;
 494
 495             /* Calculate temporary vectorial force */
 496             tx               = _mm_mul_pd(fscal,dx22);
 497             ty               = _mm_mul_pd(fscal,dy22);
 498             tz               = _mm_mul_pd(fscal,dz22);
 499
 500             /* Update vectorial force */
 501             fix2             = _mm_add_pd(fix2,tx);
 502             fiy2             = _mm_add_pd(fiy2,ty);
 503             fiz2             = _mm_add_pd(fiz2,tz);
 504
 505             fjx2             = _mm_add_pd(fjx2,tx);
 506             fjy2             = _mm_add_pd(fjy2,ty);
 507             fjz2             = _mm_add_pd(fjz2,tz);
 508
 509             gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 510
 511             /* Inner loop uses 288 flops */
 512         }
 513
 514         if(jidx<j_index_end)
 515         {
 516
 517             jnrA             = jjnr[jidx];
 518             j_coord_offsetA  = DIM*jnrA;
 519
 520             /* load j atom coordinates */
 521             gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 522                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 523
 524             /* Calculate displacement vector */
 525             dx00             = _mm_sub_pd(ix0,jx0);
 526             dy00             = _mm_sub_pd(iy0,jy0);
 527             dz00             = _mm_sub_pd(iz0,jz0);
 528             dx01             = _mm_sub_pd(ix0,jx1);
 529             dy01             = _mm_sub_pd(iy0,jy1);
 530             dz01             = _mm_sub_pd(iz0,jz1);
 531             dx02             = _mm_sub_pd(ix0,jx2);
 532             dy02             = _mm_sub_pd(iy0,jy2);
 533             dz02             = _mm_sub_pd(iz0,jz2);
 534             dx10             = _mm_sub_pd(ix1,jx0);
 535             dy10             = _mm_sub_pd(iy1,jy0);
 536             dz10             = _mm_sub_pd(iz1,jz0);
 537             dx11             = _mm_sub_pd(ix1,jx1);
 538             dy11             = _mm_sub_pd(iy1,jy1);
 539             dz11             = _mm_sub_pd(iz1,jz1);
 540             dx12             = _mm_sub_pd(ix1,jx2);
 541             dy12             = _mm_sub_pd(iy1,jy2);
 542             dz12             = _mm_sub_pd(iz1,jz2);
 543             dx20             = _mm_sub_pd(ix2,jx0);
 544             dy20             = _mm_sub_pd(iy2,jy0);
 545             dz20             = _mm_sub_pd(iz2,jz0);
 546             dx21             = _mm_sub_pd(ix2,jx1);
 547             dy21             = _mm_sub_pd(iy2,jy1);
 548             dz21             = _mm_sub_pd(iz2,jz1);
 549             dx22             = _mm_sub_pd(ix2,jx2);
 550             dy22             = _mm_sub_pd(iy2,jy2);
 551             dz22             = _mm_sub_pd(iz2,jz2);
 552
 553             /* Calculate squared distance and things based on it */
 554             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 555             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
 556             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
 557             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 558             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 559             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 560             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 561             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 562             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 563
 564             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 565             rinv01           = gmx_mm_invsqrt_pd(rsq01);
 566             rinv02           = gmx_mm_invsqrt_pd(rsq02);
 567             rinv10           = gmx_mm_invsqrt_pd(rsq10);
 568             rinv11           = gmx_mm_invsqrt_pd(rsq11);
 569             rinv12           = gmx_mm_invsqrt_pd(rsq12);
 570             rinv20           = gmx_mm_invsqrt_pd(rsq20);
 571             rinv21           = gmx_mm_invsqrt_pd(rsq21);
 572             rinv22           = gmx_mm_invsqrt_pd(rsq22);
 573
 574             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 575             rinvsq01         = _mm_mul_pd(rinv01,rinv01);
 576             rinvsq02         = _mm_mul_pd(rinv02,rinv02);
 577             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
 578             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
 579             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
 580             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
 581             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
 582             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
 583
 584             fjx0             = _mm_setzero_pd();
 585             fjy0             = _mm_setzero_pd();
 586             fjz0             = _mm_setzero_pd();
 587             fjx1             = _mm_setzero_pd();
 588             fjy1             = _mm_setzero_pd();
 589             fjz1             = _mm_setzero_pd();
 590             fjx2             = _mm_setzero_pd();
 591             fjy2             = _mm_setzero_pd();
 592             fjz2             = _mm_setzero_pd();
 593
 594             /**************************
 595              * CALCULATE INTERACTIONS *
 596              **************************/
 597
 598             /* REACTION-FIELD ELECTROSTATICS */
 599             velec            = _mm_mul_pd(qq00,_mm_sub_pd(_mm_add_pd(rinv00,_mm_mul_pd(krf,rsq00)),crf));
 600             felec            = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
 601
 602             /* Update potential sum for this i atom from the interaction with this j atom. */
 603             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 604             velecsum         = _mm_add_pd(velecsum,velec);
 605
 606             fscal            = felec;
 607
 608             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 609
 610             /* Calculate temporary vectorial force */
 611             tx               = _mm_mul_pd(fscal,dx00);
 612             ty               = _mm_mul_pd(fscal,dy00);
 613             tz               = _mm_mul_pd(fscal,dz00);
 614
 615             /* Update vectorial force */
 616             fix0             = _mm_add_pd(fix0,tx);
 617             fiy0             = _mm_add_pd(fiy0,ty);
 618             fiz0             = _mm_add_pd(fiz0,tz);
 619
 620             fjx0             = _mm_add_pd(fjx0,tx);
 621             fjy0             = _mm_add_pd(fjy0,ty);
 622             fjz0             = _mm_add_pd(fjz0,tz);
 623
 624             /**************************
 625              * CALCULATE INTERACTIONS *
 626              **************************/
 627
 628             /* REACTION-FIELD ELECTROSTATICS */
 629             velec            = _mm_mul_pd(qq01,_mm_sub_pd(_mm_add_pd(rinv01,_mm_mul_pd(krf,rsq01)),crf));
 630             felec            = _mm_mul_pd(qq01,_mm_sub_pd(_mm_mul_pd(rinv01,rinvsq01),krf2));
 631
 632             /* Update potential sum for this i atom from the interaction with this j atom. */
 633             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 634             velecsum         = _mm_add_pd(velecsum,velec);
 635
 636             fscal            = felec;
 637
 638             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 639
 640             /* Calculate temporary vectorial force */
 641             tx               = _mm_mul_pd(fscal,dx01);
 642             ty               = _mm_mul_pd(fscal,dy01);
 643             tz               = _mm_mul_pd(fscal,dz01);
 644
 645             /* Update vectorial force */
 646             fix0             = _mm_add_pd(fix0,tx);
 647             fiy0             = _mm_add_pd(fiy0,ty);
 648             fiz0             = _mm_add_pd(fiz0,tz);
 649
 650             fjx1             = _mm_add_pd(fjx1,tx);
 651             fjy1             = _mm_add_pd(fjy1,ty);
 652             fjz1             = _mm_add_pd(fjz1,tz);
 653
 654             /**************************
 655              * CALCULATE INTERACTIONS *
 656              **************************/
 657
 658             /* REACTION-FIELD ELECTROSTATICS */
 659             velec            = _mm_mul_pd(qq02,_mm_sub_pd(_mm_add_pd(rinv02,_mm_mul_pd(krf,rsq02)),crf));
 660             felec            = _mm_mul_pd(qq02,_mm_sub_pd(_mm_mul_pd(rinv02,rinvsq02),krf2));
 661
 662             /* Update potential sum for this i atom from the interaction with this j atom. */
 663             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 664             velecsum         = _mm_add_pd(velecsum,velec);
 665
 666             fscal            = felec;
 667
 668             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 669
 670             /* Calculate temporary vectorial force */
 671             tx               = _mm_mul_pd(fscal,dx02);
 672             ty               = _mm_mul_pd(fscal,dy02);
 673             tz               = _mm_mul_pd(fscal,dz02);
 674
 675             /* Update vectorial force */
 676             fix0             = _mm_add_pd(fix0,tx);
 677             fiy0             = _mm_add_pd(fiy0,ty);
 678             fiz0             = _mm_add_pd(fiz0,tz);
 679
 680             fjx2             = _mm_add_pd(fjx2,tx);
 681             fjy2             = _mm_add_pd(fjy2,ty);
 682             fjz2             = _mm_add_pd(fjz2,tz);
 683
 684             /**************************
 685              * CALCULATE INTERACTIONS *
 686              **************************/
 687
 688             /* REACTION-FIELD ELECTROSTATICS */
 689             velec            = _mm_mul_pd(qq10,_mm_sub_pd(_mm_add_pd(rinv10,_mm_mul_pd(krf,rsq10)),crf));
 690             felec            = _mm_mul_pd(qq10,_mm_sub_pd(_mm_mul_pd(rinv10,rinvsq10),krf2));
 691
 692             /* Update potential sum for this i atom from the interaction with this j atom. */
 693             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 694             velecsum         = _mm_add_pd(velecsum,velec);
 695
 696             fscal            = felec;
 697
 698             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 699
 700             /* Calculate temporary vectorial force */
 701             tx               = _mm_mul_pd(fscal,dx10);
 702             ty               = _mm_mul_pd(fscal,dy10);
 703             tz               = _mm_mul_pd(fscal,dz10);
 704
 705             /* Update vectorial force */
 706             fix1             = _mm_add_pd(fix1,tx);
 707             fiy1             = _mm_add_pd(fiy1,ty);
 708             fiz1             = _mm_add_pd(fiz1,tz);
 709
 710             fjx0             = _mm_add_pd(fjx0,tx);
 711             fjy0             = _mm_add_pd(fjy0,ty);
 712             fjz0             = _mm_add_pd(fjz0,tz);
 713
 714             /**************************
 715              * CALCULATE INTERACTIONS *
 716              **************************/
 717
 718             /* REACTION-FIELD ELECTROSTATICS */
 719             velec            = _mm_mul_pd(qq11,_mm_sub_pd(_mm_add_pd(rinv11,_mm_mul_pd(krf,rsq11)),crf));
 720             felec            = _mm_mul_pd(qq11,_mm_sub_pd(_mm_mul_pd(rinv11,rinvsq11),krf2));
 721
 722             /* Update potential sum for this i atom from the interaction with this j atom. */
 723             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 724             velecsum         = _mm_add_pd(velecsum,velec);
 725
 726             fscal            = felec;
 727
 728             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 729
 730             /* Calculate temporary vectorial force */
 731             tx               = _mm_mul_pd(fscal,dx11);
 732             ty               = _mm_mul_pd(fscal,dy11);
 733             tz               = _mm_mul_pd(fscal,dz11);
 734
 735             /* Update vectorial force */
 736             fix1             = _mm_add_pd(fix1,tx);
 737             fiy1             = _mm_add_pd(fiy1,ty);
 738             fiz1             = _mm_add_pd(fiz1,tz);
 739
 740             fjx1             = _mm_add_pd(fjx1,tx);
 741             fjy1             = _mm_add_pd(fjy1,ty);
 742             fjz1             = _mm_add_pd(fjz1,tz);
 743
 744             /**************************
 745              * CALCULATE INTERACTIONS *
 746              **************************/
 747
 748             /* REACTION-FIELD ELECTROSTATICS */
 749             velec            = _mm_mul_pd(qq12,_mm_sub_pd(_mm_add_pd(rinv12,_mm_mul_pd(krf,rsq12)),crf));
 750             felec            = _mm_mul_pd(qq12,_mm_sub_pd(_mm_mul_pd(rinv12,rinvsq12),krf2));
 751
 752             /* Update potential sum for this i atom from the interaction with this j atom. */
 753             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 754             velecsum         = _mm_add_pd(velecsum,velec);
 755
 756             fscal            = felec;
 757
 758             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 759
 760             /* Calculate temporary vectorial force */
 761             tx               = _mm_mul_pd(fscal,dx12);
 762             ty               = _mm_mul_pd(fscal,dy12);
 763             tz               = _mm_mul_pd(fscal,dz12);
 764
 765             /* Update vectorial force */
 766             fix1             = _mm_add_pd(fix1,tx);
 767             fiy1             = _mm_add_pd(fiy1,ty);
 768             fiz1             = _mm_add_pd(fiz1,tz);
 769
 770             fjx2             = _mm_add_pd(fjx2,tx);
 771             fjy2             = _mm_add_pd(fjy2,ty);
 772             fjz2             = _mm_add_pd(fjz2,tz);
 773
 774             /**************************
 775              * CALCULATE INTERACTIONS *
 776              **************************/
 777
 778             /* REACTION-FIELD ELECTROSTATICS */
 779             velec            = _mm_mul_pd(qq20,_mm_sub_pd(_mm_add_pd(rinv20,_mm_mul_pd(krf,rsq20)),crf));
 780             felec            = _mm_mul_pd(qq20,_mm_sub_pd(_mm_mul_pd(rinv20,rinvsq20),krf2));
 781
 782             /* Update potential sum for this i atom from the interaction with this j atom. */
 783             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 784             velecsum         = _mm_add_pd(velecsum,velec);
 785
 786             fscal            = felec;
 787
 788             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 789
 790             /* Calculate temporary vectorial force */
 791             tx               = _mm_mul_pd(fscal,dx20);
 792             ty               = _mm_mul_pd(fscal,dy20);
 793             tz               = _mm_mul_pd(fscal,dz20);
 794
 795             /* Update vectorial force */
 796             fix2             = _mm_add_pd(fix2,tx);
 797             fiy2             = _mm_add_pd(fiy2,ty);
 798             fiz2             = _mm_add_pd(fiz2,tz);
 799
 800             fjx0             = _mm_add_pd(fjx0,tx);
 801             fjy0             = _mm_add_pd(fjy0,ty);
 802             fjz0             = _mm_add_pd(fjz0,tz);
 803
 804             /**************************
 805              * CALCULATE INTERACTIONS *
 806              **************************/
 807
 808             /* REACTION-FIELD ELECTROSTATICS */
 809             velec            = _mm_mul_pd(qq21,_mm_sub_pd(_mm_add_pd(rinv21,_mm_mul_pd(krf,rsq21)),crf));
 810             felec            = _mm_mul_pd(qq21,_mm_sub_pd(_mm_mul_pd(rinv21,rinvsq21),krf2));
 811
 812             /* Update potential sum for this i atom from the interaction with this j atom. */
 813             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 814             velecsum         = _mm_add_pd(velecsum,velec);
 815
 816             fscal            = felec;
 817
 818             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 819
 820             /* Calculate temporary vectorial force */
 821             tx               = _mm_mul_pd(fscal,dx21);
 822             ty               = _mm_mul_pd(fscal,dy21);
 823             tz               = _mm_mul_pd(fscal,dz21);
 824
 825             /* Update vectorial force */
 826             fix2             = _mm_add_pd(fix2,tx);
 827             fiy2             = _mm_add_pd(fiy2,ty);
 828             fiz2             = _mm_add_pd(fiz2,tz);
 829
 830             fjx1             = _mm_add_pd(fjx1,tx);
 831             fjy1             = _mm_add_pd(fjy1,ty);
 832             fjz1             = _mm_add_pd(fjz1,tz);
 833
 834             /**************************
 835              * CALCULATE INTERACTIONS *
 836              **************************/
 837
 838             /* REACTION-FIELD ELECTROSTATICS */
 839             velec            = _mm_mul_pd(qq22,_mm_sub_pd(_mm_add_pd(rinv22,_mm_mul_pd(krf,rsq22)),crf));
 840             felec            = _mm_mul_pd(qq22,_mm_sub_pd(_mm_mul_pd(rinv22,rinvsq22),krf2));
 841
 842             /* Update potential sum for this i atom from the interaction with this j atom. */
 843             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 844             velecsum         = _mm_add_pd(velecsum,velec);
 845
 846             fscal            = felec;
 847
 848             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 849
 850             /* Calculate temporary vectorial force */
 851             tx               = _mm_mul_pd(fscal,dx22);
 852             ty               = _mm_mul_pd(fscal,dy22);
 853             tz               = _mm_mul_pd(fscal,dz22);
 854
 855             /* Update vectorial force */
 856             fix2             = _mm_add_pd(fix2,tx);
 857             fiy2             = _mm_add_pd(fiy2,ty);
 858             fiz2             = _mm_add_pd(fiz2,tz);
 859
 860             fjx2             = _mm_add_pd(fjx2,tx);
 861             fjy2             = _mm_add_pd(fjy2,ty);
 862             fjz2             = _mm_add_pd(fjz2,tz);
 863
 864             gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 865
 866             /* Inner loop uses 288 flops */
 867         }
 868
 869         /* End of innermost loop */
 870
 871         gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 872                                               f+i_coord_offset,fshift+i_shift_offset);
 873
 874         ggid                        = gid[iidx];
 875         /* Update potential energies */
 876         gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
 877
 878         /* Increment number of inner iterations */
 879         inneriter                  += j_index_end - j_index_start;
 880
 881         /* Outer loop uses 19 flops */
 882     }
 883
 884     /* Increment number of outer iterations */
 885     outeriter        += nri;
 886
 887     /* Update outer/inner flops */
 888
 889     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3W3_VF,outeriter*19 + inneriter*288);
 890 }
 891 /*
 892  * Gromacs nonbonded kernel:   nb_kernel_ElecRF_VdwNone_GeomW3W3_F_sse4_1_double
 893  * Electrostatics interaction: ReactionField
 894  * VdW interaction:            None
 895  * Geometry:                   Water3-Water3
 896  * Calculate force/pot:        Force
 897  */
 898 void
 899 nb_kernel_ElecRF_VdwNone_GeomW3W3_F_sse4_1_double
 900                     (t_nblist                    * gmx_restrict       nlist,
 901                      rvec                        * gmx_restrict          xx,
 902                      rvec                        * gmx_restrict          ff,
 903                      t_forcerec                  * gmx_restrict          fr,
 904                      t_mdatoms                   * gmx_restrict     mdatoms,
 905                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 906                      t_nrnb                      * gmx_restrict        nrnb)
 907 {
 908     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 909      * just 0 for non-waters.
 910      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
 911      * jnr indices corresponding to data put in the four positions in the SIMD register.
 912      */
 913     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 914     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 915     int              jnrA,jnrB;
 916     int              j_coord_offsetA,j_coord_offsetB;
 917     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 918     real             rcutoff_scalar;
 919     real             *shiftvec,*fshift,*x,*f;
 920     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 921     int              vdwioffset0;
 922     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 923     int              vdwioffset1;
 924     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 925     int              vdwioffset2;
 926     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 927     int              vdwjidx0A,vdwjidx0B;
 928     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 929     int              vdwjidx1A,vdwjidx1B;
 930     __m128d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
 931     int              vdwjidx2A,vdwjidx2B;
 932     __m128d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
 933     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 934     __m128d          dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
 935     __m128d          dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
 936     __m128d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 937     __m128d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
 938     __m128d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
 939     __m128d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 940     __m128d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 941     __m128d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 942     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
 943     real             *charge;
 944     __m128d          dummy_mask,cutoff_mask;
 945     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 946     __m128d          one     = _mm_set1_pd(1.0);
 947     __m128d          two     = _mm_set1_pd(2.0);
 948     x                = xx[0];
 949     f                = ff[0];
 950
 951     nri              = nlist->nri;
 952     iinr             = nlist->iinr;
 953     jindex           = nlist->jindex;
 954     jjnr             = nlist->jjnr;
 955     shiftidx         = nlist->shift;
 956     gid              = nlist->gid;
 957     shiftvec         = fr->shift_vec[0];
 958     fshift           = fr->fshift[0];
 959     facel            = _mm_set1_pd(fr->epsfac);
 960     charge           = mdatoms->chargeA;
 961     krf              = _mm_set1_pd(fr->ic->k_rf);
 962     krf2             = _mm_set1_pd(fr->ic->k_rf*2.0);
 963     crf              = _mm_set1_pd(fr->ic->c_rf);
 964
 965     /* Setup water-specific parameters */
 966     inr              = nlist->iinr[0];
 967     iq0              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
 968     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
 969     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
 970
 971     jq0              = _mm_set1_pd(charge[inr+0]);
 972     jq1              = _mm_set1_pd(charge[inr+1]);
 973     jq2              = _mm_set1_pd(charge[inr+2]);
 974     qq00             = _mm_mul_pd(iq0,jq0);
 975     qq01             = _mm_mul_pd(iq0,jq1);
 976     qq02             = _mm_mul_pd(iq0,jq2);
 977     qq10             = _mm_mul_pd(iq1,jq0);
 978     qq11             = _mm_mul_pd(iq1,jq1);
 979     qq12             = _mm_mul_pd(iq1,jq2);
 980     qq20             = _mm_mul_pd(iq2,jq0);
 981     qq21             = _mm_mul_pd(iq2,jq1);
 982     qq22             = _mm_mul_pd(iq2,jq2);
 983
 984     /* Avoid stupid compiler warnings */
 985     jnrA = jnrB = 0;
 986     j_coord_offsetA = 0;
 987     j_coord_offsetB = 0;
 988
 989     outeriter        = 0;
 990     inneriter        = 0;
 991
 992     /* Start outer loop over neighborlists */
 993     for(iidx=0; iidx<nri; iidx++)
 994     {
 995         /* Load shift vector for this list */
 996         i_shift_offset   = DIM*shiftidx[iidx];
 997
 998         /* Load limits for loop over neighbors */
 999         j_index_start    = jindex[iidx];
1000         j_index_end      = jindex[iidx+1];
1001
1002         /* Get outer coordinate index */
1003         inr              = iinr[iidx];
1004         i_coord_offset   = DIM*inr;
1005
1006         /* Load i particle coords and add shift vector */
1007         gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1008                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1009
1010         fix0             = _mm_setzero_pd();
1011         fiy0             = _mm_setzero_pd();
1012         fiz0             = _mm_setzero_pd();
1013         fix1             = _mm_setzero_pd();
1014         fiy1             = _mm_setzero_pd();
1015         fiz1             = _mm_setzero_pd();
1016         fix2             = _mm_setzero_pd();
1017         fiy2             = _mm_setzero_pd();
1018         fiz2             = _mm_setzero_pd();
1019
1020         /* Start inner kernel loop */
1021         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
1022         {
1023
1024             /* Get j neighbor index, and coordinate index */
1025             jnrA             = jjnr[jidx];
1026             jnrB             = jjnr[jidx+1];
1027             j_coord_offsetA  = DIM*jnrA;
1028             j_coord_offsetB  = DIM*jnrB;
1029
1030             /* load j atom coordinates */
1031             gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1032                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1033
1034             /* Calculate displacement vector */
1035             dx00             = _mm_sub_pd(ix0,jx0);
1036             dy00             = _mm_sub_pd(iy0,jy0);
1037             dz00             = _mm_sub_pd(iz0,jz0);
1038             dx01             = _mm_sub_pd(ix0,jx1);
1039             dy01             = _mm_sub_pd(iy0,jy1);
1040             dz01             = _mm_sub_pd(iz0,jz1);
1041             dx02             = _mm_sub_pd(ix0,jx2);
1042             dy02             = _mm_sub_pd(iy0,jy2);
1043             dz02             = _mm_sub_pd(iz0,jz2);
1044             dx10             = _mm_sub_pd(ix1,jx0);
1045             dy10             = _mm_sub_pd(iy1,jy0);
1046             dz10             = _mm_sub_pd(iz1,jz0);
1047             dx11             = _mm_sub_pd(ix1,jx1);
1048             dy11             = _mm_sub_pd(iy1,jy1);
1049             dz11             = _mm_sub_pd(iz1,jz1);
1050             dx12             = _mm_sub_pd(ix1,jx2);
1051             dy12             = _mm_sub_pd(iy1,jy2);
1052             dz12             = _mm_sub_pd(iz1,jz2);
1053             dx20             = _mm_sub_pd(ix2,jx0);
1054             dy20             = _mm_sub_pd(iy2,jy0);
1055             dz20             = _mm_sub_pd(iz2,jz0);
1056             dx21             = _mm_sub_pd(ix2,jx1);
1057             dy21             = _mm_sub_pd(iy2,jy1);
1058             dz21             = _mm_sub_pd(iz2,jz1);
1059             dx22             = _mm_sub_pd(ix2,jx2);
1060             dy22             = _mm_sub_pd(iy2,jy2);
1061             dz22             = _mm_sub_pd(iz2,jz2);
1062
1063             /* Calculate squared distance and things based on it */
1064             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1065             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
1066             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
1067             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
1068             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1069             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1070             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
1071             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1072             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1073
1074             rinv00           = gmx_mm_invsqrt_pd(rsq00);
1075             rinv01           = gmx_mm_invsqrt_pd(rsq01);
1076             rinv02           = gmx_mm_invsqrt_pd(rsq02);
1077             rinv10           = gmx_mm_invsqrt_pd(rsq10);
1078             rinv11           = gmx_mm_invsqrt_pd(rsq11);
1079             rinv12           = gmx_mm_invsqrt_pd(rsq12);
1080             rinv20           = gmx_mm_invsqrt_pd(rsq20);
1081             rinv21           = gmx_mm_invsqrt_pd(rsq21);
1082             rinv22           = gmx_mm_invsqrt_pd(rsq22);
1083
1084             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
1085             rinvsq01         = _mm_mul_pd(rinv01,rinv01);
1086             rinvsq02         = _mm_mul_pd(rinv02,rinv02);
1087             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
1088             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
1089             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
1090             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
1091             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
1092             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
1093
1094             fjx0             = _mm_setzero_pd();
1095             fjy0             = _mm_setzero_pd();
1096             fjz0             = _mm_setzero_pd();
1097             fjx1             = _mm_setzero_pd();
1098             fjy1             = _mm_setzero_pd();
1099             fjz1             = _mm_setzero_pd();
1100             fjx2             = _mm_setzero_pd();
1101             fjy2             = _mm_setzero_pd();
1102             fjz2             = _mm_setzero_pd();
1103
1104             /**************************
1105              * CALCULATE INTERACTIONS *
1106              **************************/
1107
1108             /* REACTION-FIELD ELECTROSTATICS */
1109             felec            = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
1110
1111             fscal            = felec;
1112
1113             /* Calculate temporary vectorial force */
1114             tx               = _mm_mul_pd(fscal,dx00);
1115             ty               = _mm_mul_pd(fscal,dy00);
1116             tz               = _mm_mul_pd(fscal,dz00);
1117
1118             /* Update vectorial force */
1119             fix0             = _mm_add_pd(fix0,tx);
1120             fiy0             = _mm_add_pd(fiy0,ty);
1121             fiz0             = _mm_add_pd(fiz0,tz);
1122
1123             fjx0             = _mm_add_pd(fjx0,tx);
1124             fjy0             = _mm_add_pd(fjy0,ty);
1125             fjz0             = _mm_add_pd(fjz0,tz);
1126
1127             /**************************
1128              * CALCULATE INTERACTIONS *
1129              **************************/
1130
1131             /* REACTION-FIELD ELECTROSTATICS */
1132             felec            = _mm_mul_pd(qq01,_mm_sub_pd(_mm_mul_pd(rinv01,rinvsq01),krf2));
1133
1134             fscal            = felec;
1135
1136             /* Calculate temporary vectorial force */
1137             tx               = _mm_mul_pd(fscal,dx01);
1138             ty               = _mm_mul_pd(fscal,dy01);
1139             tz               = _mm_mul_pd(fscal,dz01);
1140
1141             /* Update vectorial force */
1142             fix0             = _mm_add_pd(fix0,tx);
1143             fiy0             = _mm_add_pd(fiy0,ty);
1144             fiz0             = _mm_add_pd(fiz0,tz);
1145
1146             fjx1             = _mm_add_pd(fjx1,tx);
1147             fjy1             = _mm_add_pd(fjy1,ty);
1148             fjz1             = _mm_add_pd(fjz1,tz);
1149
1150             /**************************
1151              * CALCULATE INTERACTIONS *
1152              **************************/
1153
1154             /* REACTION-FIELD ELECTROSTATICS */
1155             felec            = _mm_mul_pd(qq02,_mm_sub_pd(_mm_mul_pd(rinv02,rinvsq02),krf2));
1156
1157             fscal            = felec;
1158
1159             /* Calculate temporary vectorial force */
1160             tx               = _mm_mul_pd(fscal,dx02);
1161             ty               = _mm_mul_pd(fscal,dy02);
1162             tz               = _mm_mul_pd(fscal,dz02);
1163
1164             /* Update vectorial force */
1165             fix0             = _mm_add_pd(fix0,tx);
1166             fiy0             = _mm_add_pd(fiy0,ty);
1167             fiz0             = _mm_add_pd(fiz0,tz);
1168
1169             fjx2             = _mm_add_pd(fjx2,tx);
1170             fjy2             = _mm_add_pd(fjy2,ty);
1171             fjz2             = _mm_add_pd(fjz2,tz);
1172
1173             /**************************
1174              * CALCULATE INTERACTIONS *
1175              **************************/
1176
1177             /* REACTION-FIELD ELECTROSTATICS */
1178             felec            = _mm_mul_pd(qq10,_mm_sub_pd(_mm_mul_pd(rinv10,rinvsq10),krf2));
1179
1180             fscal            = felec;
1181
1182             /* Calculate temporary vectorial force */
1183             tx               = _mm_mul_pd(fscal,dx10);
1184             ty               = _mm_mul_pd(fscal,dy10);
1185             tz               = _mm_mul_pd(fscal,dz10);
1186
1187             /* Update vectorial force */
1188             fix1             = _mm_add_pd(fix1,tx);
1189             fiy1             = _mm_add_pd(fiy1,ty);
1190             fiz1             = _mm_add_pd(fiz1,tz);
1191
1192             fjx0             = _mm_add_pd(fjx0,tx);
1193             fjy0             = _mm_add_pd(fjy0,ty);
1194             fjz0             = _mm_add_pd(fjz0,tz);
1195
1196             /**************************
1197              * CALCULATE INTERACTIONS *
1198              **************************/
1199
1200             /* REACTION-FIELD ELECTROSTATICS */
1201             felec            = _mm_mul_pd(qq11,_mm_sub_pd(_mm_mul_pd(rinv11,rinvsq11),krf2));
1202
1203             fscal            = felec;
1204
1205             /* Calculate temporary vectorial force */
1206             tx               = _mm_mul_pd(fscal,dx11);
1207             ty               = _mm_mul_pd(fscal,dy11);
1208             tz               = _mm_mul_pd(fscal,dz11);
1209
1210             /* Update vectorial force */
1211             fix1             = _mm_add_pd(fix1,tx);
1212             fiy1             = _mm_add_pd(fiy1,ty);
1213             fiz1             = _mm_add_pd(fiz1,tz);
1214
1215             fjx1             = _mm_add_pd(fjx1,tx);
1216             fjy1             = _mm_add_pd(fjy1,ty);
1217             fjz1             = _mm_add_pd(fjz1,tz);
1218
1219             /**************************
1220              * CALCULATE INTERACTIONS *
1221              **************************/
1222
1223             /* REACTION-FIELD ELECTROSTATICS */
1224             felec            = _mm_mul_pd(qq12,_mm_sub_pd(_mm_mul_pd(rinv12,rinvsq12),krf2));
1225
1226             fscal            = felec;
1227
1228             /* Calculate temporary vectorial force */
1229             tx               = _mm_mul_pd(fscal,dx12);
1230             ty               = _mm_mul_pd(fscal,dy12);
1231             tz               = _mm_mul_pd(fscal,dz12);
1232
1233             /* Update vectorial force */
1234             fix1             = _mm_add_pd(fix1,tx);
1235             fiy1             = _mm_add_pd(fiy1,ty);
1236             fiz1             = _mm_add_pd(fiz1,tz);
1237
1238             fjx2             = _mm_add_pd(fjx2,tx);
1239             fjy2             = _mm_add_pd(fjy2,ty);
1240             fjz2             = _mm_add_pd(fjz2,tz);
1241
1242             /**************************
1243              * CALCULATE INTERACTIONS *
1244              **************************/
1245
1246             /* REACTION-FIELD ELECTROSTATICS */
1247             felec            = _mm_mul_pd(qq20,_mm_sub_pd(_mm_mul_pd(rinv20,rinvsq20),krf2));
1248
1249             fscal            = felec;
1250
1251             /* Calculate temporary vectorial force */
1252             tx               = _mm_mul_pd(fscal,dx20);
1253             ty               = _mm_mul_pd(fscal,dy20);
1254             tz               = _mm_mul_pd(fscal,dz20);
1255
1256             /* Update vectorial force */
1257             fix2             = _mm_add_pd(fix2,tx);
1258             fiy2             = _mm_add_pd(fiy2,ty);
1259             fiz2             = _mm_add_pd(fiz2,tz);
1260
1261             fjx0             = _mm_add_pd(fjx0,tx);
1262             fjy0             = _mm_add_pd(fjy0,ty);
1263             fjz0             = _mm_add_pd(fjz0,tz);
1264
1265             /**************************
1266              * CALCULATE INTERACTIONS *
1267              **************************/
1268
1269             /* REACTION-FIELD ELECTROSTATICS */
1270             felec            = _mm_mul_pd(qq21,_mm_sub_pd(_mm_mul_pd(rinv21,rinvsq21),krf2));
1271
1272             fscal            = felec;
1273
1274             /* Calculate temporary vectorial force */
1275             tx               = _mm_mul_pd(fscal,dx21);
1276             ty               = _mm_mul_pd(fscal,dy21);
1277             tz               = _mm_mul_pd(fscal,dz21);
1278
1279             /* Update vectorial force */
1280             fix2             = _mm_add_pd(fix2,tx);
1281             fiy2             = _mm_add_pd(fiy2,ty);
1282             fiz2             = _mm_add_pd(fiz2,tz);
1283
1284             fjx1             = _mm_add_pd(fjx1,tx);
1285             fjy1             = _mm_add_pd(fjy1,ty);
1286             fjz1             = _mm_add_pd(fjz1,tz);
1287
1288             /**************************
1289              * CALCULATE INTERACTIONS *
1290              **************************/
1291
1292             /* REACTION-FIELD ELECTROSTATICS */
1293             felec            = _mm_mul_pd(qq22,_mm_sub_pd(_mm_mul_pd(rinv22,rinvsq22),krf2));
1294
1295             fscal            = felec;
1296
1297             /* Calculate temporary vectorial force */
1298             tx               = _mm_mul_pd(fscal,dx22);
1299             ty               = _mm_mul_pd(fscal,dy22);
1300             tz               = _mm_mul_pd(fscal,dz22);
1301
1302             /* Update vectorial force */
1303             fix2             = _mm_add_pd(fix2,tx);
1304             fiy2             = _mm_add_pd(fiy2,ty);
1305             fiz2             = _mm_add_pd(fiz2,tz);
1306
1307             fjx2             = _mm_add_pd(fjx2,tx);
1308             fjy2             = _mm_add_pd(fjy2,ty);
1309             fjz2             = _mm_add_pd(fjz2,tz);
1310
1311             gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1312
1313             /* Inner loop uses 243 flops */
1314         }
1315
1316         if(jidx<j_index_end)
1317         {
1318
1319             jnrA             = jjnr[jidx];
1320             j_coord_offsetA  = DIM*jnrA;
1321
1322             /* load j atom coordinates */
1323             gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
1324                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1325
1326             /* Calculate displacement vector */
1327             dx00             = _mm_sub_pd(ix0,jx0);
1328             dy00             = _mm_sub_pd(iy0,jy0);
1329             dz00             = _mm_sub_pd(iz0,jz0);
1330             dx01             = _mm_sub_pd(ix0,jx1);
1331             dy01             = _mm_sub_pd(iy0,jy1);
1332             dz01             = _mm_sub_pd(iz0,jz1);
1333             dx02             = _mm_sub_pd(ix0,jx2);
1334             dy02             = _mm_sub_pd(iy0,jy2);
1335             dz02             = _mm_sub_pd(iz0,jz2);
1336             dx10             = _mm_sub_pd(ix1,jx0);
1337             dy10             = _mm_sub_pd(iy1,jy0);
1338             dz10             = _mm_sub_pd(iz1,jz0);
1339             dx11             = _mm_sub_pd(ix1,jx1);
1340             dy11             = _mm_sub_pd(iy1,jy1);
1341             dz11             = _mm_sub_pd(iz1,jz1);
1342             dx12             = _mm_sub_pd(ix1,jx2);
1343             dy12             = _mm_sub_pd(iy1,jy2);
1344             dz12             = _mm_sub_pd(iz1,jz2);
1345             dx20             = _mm_sub_pd(ix2,jx0);
1346             dy20             = _mm_sub_pd(iy2,jy0);
1347             dz20             = _mm_sub_pd(iz2,jz0);
1348             dx21             = _mm_sub_pd(ix2,jx1);
1349             dy21             = _mm_sub_pd(iy2,jy1);
1350             dz21             = _mm_sub_pd(iz2,jz1);
1351             dx22             = _mm_sub_pd(ix2,jx2);
1352             dy22             = _mm_sub_pd(iy2,jy2);
1353             dz22             = _mm_sub_pd(iz2,jz2);
1354
1355             /* Calculate squared distance and things based on it */
1356             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1357             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
1358             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
1359             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
1360             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1361             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1362             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
1363             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1364             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1365
1366             rinv00           = gmx_mm_invsqrt_pd(rsq00);
1367             rinv01           = gmx_mm_invsqrt_pd(rsq01);
1368             rinv02           = gmx_mm_invsqrt_pd(rsq02);
1369             rinv10           = gmx_mm_invsqrt_pd(rsq10);
1370             rinv11           = gmx_mm_invsqrt_pd(rsq11);
1371             rinv12           = gmx_mm_invsqrt_pd(rsq12);
1372             rinv20           = gmx_mm_invsqrt_pd(rsq20);
1373             rinv21           = gmx_mm_invsqrt_pd(rsq21);
1374             rinv22           = gmx_mm_invsqrt_pd(rsq22);
1375
1376             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
1377             rinvsq01         = _mm_mul_pd(rinv01,rinv01);
1378             rinvsq02         = _mm_mul_pd(rinv02,rinv02);
1379             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
1380             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
1381             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
1382             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
1383             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
1384             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
1385
1386             fjx0             = _mm_setzero_pd();
1387             fjy0             = _mm_setzero_pd();
1388             fjz0             = _mm_setzero_pd();
1389             fjx1             = _mm_setzero_pd();
1390             fjy1             = _mm_setzero_pd();
1391             fjz1             = _mm_setzero_pd();
1392             fjx2             = _mm_setzero_pd();
1393             fjy2             = _mm_setzero_pd();
1394             fjz2             = _mm_setzero_pd();
1395
1396             /**************************
1397              * CALCULATE INTERACTIONS *
1398              **************************/
1399
1400             /* REACTION-FIELD ELECTROSTATICS */
1401             felec            = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
1402
1403             fscal            = felec;
1404
1405             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1406
1407             /* Calculate temporary vectorial force */
1408             tx               = _mm_mul_pd(fscal,dx00);
1409             ty               = _mm_mul_pd(fscal,dy00);
1410             tz               = _mm_mul_pd(fscal,dz00);
1411
1412             /* Update vectorial force */
1413             fix0             = _mm_add_pd(fix0,tx);
1414             fiy0             = _mm_add_pd(fiy0,ty);
1415             fiz0             = _mm_add_pd(fiz0,tz);
1416
1417             fjx0             = _mm_add_pd(fjx0,tx);
1418             fjy0             = _mm_add_pd(fjy0,ty);
1419             fjz0             = _mm_add_pd(fjz0,tz);
1420
1421             /**************************
1422              * CALCULATE INTERACTIONS *
1423              **************************/
1424
1425             /* REACTION-FIELD ELECTROSTATICS */
1426             felec            = _mm_mul_pd(qq01,_mm_sub_pd(_mm_mul_pd(rinv01,rinvsq01),krf2));
1427
1428             fscal            = felec;
1429
1430             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1431
1432             /* Calculate temporary vectorial force */
1433             tx               = _mm_mul_pd(fscal,dx01);
1434             ty               = _mm_mul_pd(fscal,dy01);
1435             tz               = _mm_mul_pd(fscal,dz01);
1436
1437             /* Update vectorial force */
1438             fix0             = _mm_add_pd(fix0,tx);
1439             fiy0             = _mm_add_pd(fiy0,ty);
1440             fiz0             = _mm_add_pd(fiz0,tz);
1441
1442             fjx1             = _mm_add_pd(fjx1,tx);
1443             fjy1             = _mm_add_pd(fjy1,ty);
1444             fjz1             = _mm_add_pd(fjz1,tz);
1445
1446             /**************************
1447              * CALCULATE INTERACTIONS *
1448              **************************/
1449
1450             /* REACTION-FIELD ELECTROSTATICS */
1451             felec            = _mm_mul_pd(qq02,_mm_sub_pd(_mm_mul_pd(rinv02,rinvsq02),krf2));
1452
1453             fscal            = felec;
1454
1455             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1456
1457             /* Calculate temporary vectorial force */
1458             tx               = _mm_mul_pd(fscal,dx02);
1459             ty               = _mm_mul_pd(fscal,dy02);
1460             tz               = _mm_mul_pd(fscal,dz02);
1461
1462             /* Update vectorial force */
1463             fix0             = _mm_add_pd(fix0,tx);
1464             fiy0             = _mm_add_pd(fiy0,ty);
1465             fiz0             = _mm_add_pd(fiz0,tz);
1466
1467             fjx2             = _mm_add_pd(fjx2,tx);
1468             fjy2             = _mm_add_pd(fjy2,ty);
1469             fjz2             = _mm_add_pd(fjz2,tz);
1470
1471             /**************************
1472              * CALCULATE INTERACTIONS *
1473              **************************/
1474
1475             /* REACTION-FIELD ELECTROSTATICS */
1476             felec            = _mm_mul_pd(qq10,_mm_sub_pd(_mm_mul_pd(rinv10,rinvsq10),krf2));
1477
1478             fscal            = felec;
1479
1480             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1481
1482             /* Calculate temporary vectorial force */
1483             tx               = _mm_mul_pd(fscal,dx10);
1484             ty               = _mm_mul_pd(fscal,dy10);
1485             tz               = _mm_mul_pd(fscal,dz10);
1486
1487             /* Update vectorial force */
1488             fix1             = _mm_add_pd(fix1,tx);
1489             fiy1             = _mm_add_pd(fiy1,ty);
1490             fiz1             = _mm_add_pd(fiz1,tz);
1491
1492             fjx0             = _mm_add_pd(fjx0,tx);
1493             fjy0             = _mm_add_pd(fjy0,ty);
1494             fjz0             = _mm_add_pd(fjz0,tz);
1495
1496             /**************************
1497              * CALCULATE INTERACTIONS *
1498              **************************/
1499
1500             /* REACTION-FIELD ELECTROSTATICS */
1501             felec            = _mm_mul_pd(qq11,_mm_sub_pd(_mm_mul_pd(rinv11,rinvsq11),krf2));
1502
1503             fscal            = felec;
1504
1505             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1506
1507             /* Calculate temporary vectorial force */
1508             tx               = _mm_mul_pd(fscal,dx11);
1509             ty               = _mm_mul_pd(fscal,dy11);
1510             tz               = _mm_mul_pd(fscal,dz11);
1511
1512             /* Update vectorial force */
1513             fix1             = _mm_add_pd(fix1,tx);
1514             fiy1             = _mm_add_pd(fiy1,ty);
1515             fiz1             = _mm_add_pd(fiz1,tz);
1516
1517             fjx1             = _mm_add_pd(fjx1,tx);
1518             fjy1             = _mm_add_pd(fjy1,ty);
1519             fjz1             = _mm_add_pd(fjz1,tz);
1520
1521             /**************************
1522              * CALCULATE INTERACTIONS *
1523              **************************/
1524
1525             /* REACTION-FIELD ELECTROSTATICS */
1526             felec            = _mm_mul_pd(qq12,_mm_sub_pd(_mm_mul_pd(rinv12,rinvsq12),krf2));
1527
1528             fscal            = felec;
1529
1530             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1531
1532             /* Calculate temporary vectorial force */
1533             tx               = _mm_mul_pd(fscal,dx12);
1534             ty               = _mm_mul_pd(fscal,dy12);
1535             tz               = _mm_mul_pd(fscal,dz12);
1536
1537             /* Update vectorial force */
1538             fix1             = _mm_add_pd(fix1,tx);
1539             fiy1             = _mm_add_pd(fiy1,ty);
1540             fiz1             = _mm_add_pd(fiz1,tz);
1541
1542             fjx2             = _mm_add_pd(fjx2,tx);
1543             fjy2             = _mm_add_pd(fjy2,ty);
1544             fjz2             = _mm_add_pd(fjz2,tz);
1545
1546             /**************************
1547              * CALCULATE INTERACTIONS *
1548              **************************/
1549
1550             /* REACTION-FIELD ELECTROSTATICS */
1551             felec            = _mm_mul_pd(qq20,_mm_sub_pd(_mm_mul_pd(rinv20,rinvsq20),krf2));
1552
1553             fscal            = felec;
1554
1555             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1556
1557             /* Calculate temporary vectorial force */
1558             tx               = _mm_mul_pd(fscal,dx20);
1559             ty               = _mm_mul_pd(fscal,dy20);
1560             tz               = _mm_mul_pd(fscal,dz20);
1561
1562             /* Update vectorial force */
1563             fix2             = _mm_add_pd(fix2,tx);
1564             fiy2             = _mm_add_pd(fiy2,ty);
1565             fiz2             = _mm_add_pd(fiz2,tz);
1566
1567             fjx0             = _mm_add_pd(fjx0,tx);
1568             fjy0             = _mm_add_pd(fjy0,ty);
1569             fjz0             = _mm_add_pd(fjz0,tz);
1570
1571             /**************************
1572              * CALCULATE INTERACTIONS *
1573              **************************/
1574
1575             /* REACTION-FIELD ELECTROSTATICS */
1576             felec            = _mm_mul_pd(qq21,_mm_sub_pd(_mm_mul_pd(rinv21,rinvsq21),krf2));
1577
1578             fscal            = felec;
1579
1580             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1581
1582             /* Calculate temporary vectorial force */
1583             tx               = _mm_mul_pd(fscal,dx21);
1584             ty               = _mm_mul_pd(fscal,dy21);
1585             tz               = _mm_mul_pd(fscal,dz21);
1586
1587             /* Update vectorial force */
1588             fix2             = _mm_add_pd(fix2,tx);
1589             fiy2             = _mm_add_pd(fiy2,ty);
1590             fiz2             = _mm_add_pd(fiz2,tz);
1591
1592             fjx1             = _mm_add_pd(fjx1,tx);
1593             fjy1             = _mm_add_pd(fjy1,ty);
1594             fjz1             = _mm_add_pd(fjz1,tz);
1595
1596             /**************************
1597              * CALCULATE INTERACTIONS *
1598              **************************/
1599
1600             /* REACTION-FIELD ELECTROSTATICS */
1601             felec            = _mm_mul_pd(qq22,_mm_sub_pd(_mm_mul_pd(rinv22,rinvsq22),krf2));
1602
1603             fscal            = felec;
1604
1605             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1606
1607             /* Calculate temporary vectorial force */
1608             tx               = _mm_mul_pd(fscal,dx22);
1609             ty               = _mm_mul_pd(fscal,dy22);
1610             tz               = _mm_mul_pd(fscal,dz22);
1611
1612             /* Update vectorial force */
1613             fix2             = _mm_add_pd(fix2,tx);
1614             fiy2             = _mm_add_pd(fiy2,ty);
1615             fiz2             = _mm_add_pd(fiz2,tz);
1616
1617             fjx2             = _mm_add_pd(fjx2,tx);
1618             fjy2             = _mm_add_pd(fjy2,ty);
1619             fjz2             = _mm_add_pd(fjz2,tz);
1620
1621             gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1622
1623             /* Inner loop uses 243 flops */
1624         }
1625
1626         /* End of innermost loop */
1627
1628         gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1629                                               f+i_coord_offset,fshift+i_shift_offset);
1630
1631         /* Increment number of inner iterations */
1632         inneriter                  += j_index_end - j_index_start;
1633
1634         /* Outer loop uses 18 flops */
1635     }
1636
1637     /* Increment number of outer iterations */
1638     outeriter        += nri;
1639
1640     /* Update outer/inner flops */
1641
1642     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3W3_F,outeriter*18 + inneriter*243);
1643 }