src/gromacs/gmxlib/nonbonded/nb_kernel_sse4_1_double/nb_kernel_ElecEwSw_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 "gmxpre.h"
  39
  40 #include "config.h"
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "gromacs/legacyheaders/types/simple.h"
  46 #include "gromacs/math/vec.h"
  47 #include "gromacs/legacyheaders/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_ElecEwSw_VdwNone_GeomW3W3_VF_sse4_1_double
  54  * Electrostatics interaction: Ewald
  55  * VdW interaction:            None
  56  * Geometry:                   Water3-Water3
  57  * Calculate force/pot:        PotentialAndForce
  58  */
  59 void
  60 nb_kernel_ElecEwSw_VdwNone_GeomW3W3_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              vdwjidx0A,vdwjidx0B;
  89     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  90     int              vdwjidx1A,vdwjidx1B;
  91     __m128d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  92     int              vdwjidx2A,vdwjidx2B;
  93     __m128d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  94     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  95     __m128d          dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
  96     __m128d          dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
  97     __m128d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  98     __m128d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
  99     __m128d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
 100     __m128d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 101     __m128d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 102     __m128d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 103     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
 104     real             *charge;
 105     __m128i          ewitab;
 106     __m128d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 107     real             *ewtab;
 108     __m128d          rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
 109     real             rswitch_scalar,d_scalar;
 110     __m128d          dummy_mask,cutoff_mask;
 111     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 112     __m128d          one     = _mm_set1_pd(1.0);
 113     __m128d          two     = _mm_set1_pd(2.0);
 114     x                = xx[0];
 115     f                = ff[0];
 116
 117     nri              = nlist->nri;
 118     iinr             = nlist->iinr;
 119     jindex           = nlist->jindex;
 120     jjnr             = nlist->jjnr;
 121     shiftidx         = nlist->shift;
 122     gid              = nlist->gid;
 123     shiftvec         = fr->shift_vec[0];
 124     fshift           = fr->fshift[0];
 125     facel            = _mm_set1_pd(fr->epsfac);
 126     charge           = mdatoms->chargeA;
 127
 128     sh_ewald         = _mm_set1_pd(fr->ic->sh_ewald);
 129     ewtab            = fr->ic->tabq_coul_FDV0;
 130     ewtabscale       = _mm_set1_pd(fr->ic->tabq_scale);
 131     ewtabhalfspace   = _mm_set1_pd(0.5/fr->ic->tabq_scale);
 132
 133     /* Setup water-specific parameters */
 134     inr              = nlist->iinr[0];
 135     iq0              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
 136     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
 137     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
 138
 139     jq0              = _mm_set1_pd(charge[inr+0]);
 140     jq1              = _mm_set1_pd(charge[inr+1]);
 141     jq2              = _mm_set1_pd(charge[inr+2]);
 142     qq00             = _mm_mul_pd(iq0,jq0);
 143     qq01             = _mm_mul_pd(iq0,jq1);
 144     qq02             = _mm_mul_pd(iq0,jq2);
 145     qq10             = _mm_mul_pd(iq1,jq0);
 146     qq11             = _mm_mul_pd(iq1,jq1);
 147     qq12             = _mm_mul_pd(iq1,jq2);
 148     qq20             = _mm_mul_pd(iq2,jq0);
 149     qq21             = _mm_mul_pd(iq2,jq1);
 150     qq22             = _mm_mul_pd(iq2,jq2);
 151
 152     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 153     rcutoff_scalar   = fr->rcoulomb;
 154     rcutoff          = _mm_set1_pd(rcutoff_scalar);
 155     rcutoff2         = _mm_mul_pd(rcutoff,rcutoff);
 156
 157     rswitch_scalar   = fr->rcoulomb_switch;
 158     rswitch          = _mm_set1_pd(rswitch_scalar);
 159     /* Setup switch parameters */
 160     d_scalar         = rcutoff_scalar-rswitch_scalar;
 161     d                = _mm_set1_pd(d_scalar);
 162     swV3             = _mm_set1_pd(-10.0/(d_scalar*d_scalar*d_scalar));
 163     swV4             = _mm_set1_pd( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 164     swV5             = _mm_set1_pd( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 165     swF2             = _mm_set1_pd(-30.0/(d_scalar*d_scalar*d_scalar));
 166     swF3             = _mm_set1_pd( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 167     swF4             = _mm_set1_pd(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 168
 169     /* Avoid stupid compiler warnings */
 170     jnrA = jnrB = 0;
 171     j_coord_offsetA = 0;
 172     j_coord_offsetB = 0;
 173
 174     outeriter        = 0;
 175     inneriter        = 0;
 176
 177     /* Start outer loop over neighborlists */
 178     for(iidx=0; iidx<nri; iidx++)
 179     {
 180         /* Load shift vector for this list */
 181         i_shift_offset   = DIM*shiftidx[iidx];
 182
 183         /* Load limits for loop over neighbors */
 184         j_index_start    = jindex[iidx];
 185         j_index_end      = jindex[iidx+1];
 186
 187         /* Get outer coordinate index */
 188         inr              = iinr[iidx];
 189         i_coord_offset   = DIM*inr;
 190
 191         /* Load i particle coords and add shift vector */
 192         gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 193                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 194
 195         fix0             = _mm_setzero_pd();
 196         fiy0             = _mm_setzero_pd();
 197         fiz0             = _mm_setzero_pd();
 198         fix1             = _mm_setzero_pd();
 199         fiy1             = _mm_setzero_pd();
 200         fiz1             = _mm_setzero_pd();
 201         fix2             = _mm_setzero_pd();
 202         fiy2             = _mm_setzero_pd();
 203         fiz2             = _mm_setzero_pd();
 204
 205         /* Reset potential sums */
 206         velecsum         = _mm_setzero_pd();
 207
 208         /* Start inner kernel loop */
 209         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 210         {
 211
 212             /* Get j neighbor index, and coordinate index */
 213             jnrA             = jjnr[jidx];
 214             jnrB             = jjnr[jidx+1];
 215             j_coord_offsetA  = DIM*jnrA;
 216             j_coord_offsetB  = DIM*jnrB;
 217
 218             /* load j atom coordinates */
 219             gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 220                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 221
 222             /* Calculate displacement vector */
 223             dx00             = _mm_sub_pd(ix0,jx0);
 224             dy00             = _mm_sub_pd(iy0,jy0);
 225             dz00             = _mm_sub_pd(iz0,jz0);
 226             dx01             = _mm_sub_pd(ix0,jx1);
 227             dy01             = _mm_sub_pd(iy0,jy1);
 228             dz01             = _mm_sub_pd(iz0,jz1);
 229             dx02             = _mm_sub_pd(ix0,jx2);
 230             dy02             = _mm_sub_pd(iy0,jy2);
 231             dz02             = _mm_sub_pd(iz0,jz2);
 232             dx10             = _mm_sub_pd(ix1,jx0);
 233             dy10             = _mm_sub_pd(iy1,jy0);
 234             dz10             = _mm_sub_pd(iz1,jz0);
 235             dx11             = _mm_sub_pd(ix1,jx1);
 236             dy11             = _mm_sub_pd(iy1,jy1);
 237             dz11             = _mm_sub_pd(iz1,jz1);
 238             dx12             = _mm_sub_pd(ix1,jx2);
 239             dy12             = _mm_sub_pd(iy1,jy2);
 240             dz12             = _mm_sub_pd(iz1,jz2);
 241             dx20             = _mm_sub_pd(ix2,jx0);
 242             dy20             = _mm_sub_pd(iy2,jy0);
 243             dz20             = _mm_sub_pd(iz2,jz0);
 244             dx21             = _mm_sub_pd(ix2,jx1);
 245             dy21             = _mm_sub_pd(iy2,jy1);
 246             dz21             = _mm_sub_pd(iz2,jz1);
 247             dx22             = _mm_sub_pd(ix2,jx2);
 248             dy22             = _mm_sub_pd(iy2,jy2);
 249             dz22             = _mm_sub_pd(iz2,jz2);
 250
 251             /* Calculate squared distance and things based on it */
 252             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 253             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
 254             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
 255             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 256             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 257             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 258             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 259             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 260             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 261
 262             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 263             rinv01           = gmx_mm_invsqrt_pd(rsq01);
 264             rinv02           = gmx_mm_invsqrt_pd(rsq02);
 265             rinv10           = gmx_mm_invsqrt_pd(rsq10);
 266             rinv11           = gmx_mm_invsqrt_pd(rsq11);
 267             rinv12           = gmx_mm_invsqrt_pd(rsq12);
 268             rinv20           = gmx_mm_invsqrt_pd(rsq20);
 269             rinv21           = gmx_mm_invsqrt_pd(rsq21);
 270             rinv22           = gmx_mm_invsqrt_pd(rsq22);
 271
 272             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 273             rinvsq01         = _mm_mul_pd(rinv01,rinv01);
 274             rinvsq02         = _mm_mul_pd(rinv02,rinv02);
 275             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
 276             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
 277             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
 278             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
 279             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
 280             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
 281
 282             fjx0             = _mm_setzero_pd();
 283             fjy0             = _mm_setzero_pd();
 284             fjz0             = _mm_setzero_pd();
 285             fjx1             = _mm_setzero_pd();
 286             fjy1             = _mm_setzero_pd();
 287             fjz1             = _mm_setzero_pd();
 288             fjx2             = _mm_setzero_pd();
 289             fjy2             = _mm_setzero_pd();
 290             fjz2             = _mm_setzero_pd();
 291
 292             /**************************
 293              * CALCULATE INTERACTIONS *
 294              **************************/
 295
 296             if (gmx_mm_any_lt(rsq00,rcutoff2))
 297             {
 298
 299             r00              = _mm_mul_pd(rsq00,rinv00);
 300
 301             /* EWALD ELECTROSTATICS */
 302
 303             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 304             ewrt             = _mm_mul_pd(r00,ewtabscale);
 305             ewitab           = _mm_cvttpd_epi32(ewrt);
 306             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 307             ewitab           = _mm_slli_epi32(ewitab,2);
 308             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 309             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 310             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 311             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 312             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 313             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 314             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 315             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 316             velec            = _mm_mul_pd(qq00,_mm_sub_pd(rinv00,velec));
 317             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
 318
 319             d                = _mm_sub_pd(r00,rswitch);
 320             d                = _mm_max_pd(d,_mm_setzero_pd());
 321             d2               = _mm_mul_pd(d,d);
 322             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_add_pd(swV3,_mm_mul_pd(d,_mm_add_pd(swV4,_mm_mul_pd(d,swV5)))))));
 323
 324             dsw              = _mm_mul_pd(d2,_mm_add_pd(swF2,_mm_mul_pd(d,_mm_add_pd(swF3,_mm_mul_pd(d,swF4)))));
 325
 326             /* Evaluate switch function */
 327             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 328             felec            = _mm_sub_pd( _mm_mul_pd(felec,sw) , _mm_mul_pd(rinv00,_mm_mul_pd(velec,dsw)) );
 329             velec            = _mm_mul_pd(velec,sw);
 330             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
 331
 332             /* Update potential sum for this i atom from the interaction with this j atom. */
 333             velec            = _mm_and_pd(velec,cutoff_mask);
 334             velecsum         = _mm_add_pd(velecsum,velec);
 335
 336             fscal            = felec;
 337
 338             fscal            = _mm_and_pd(fscal,cutoff_mask);
 339
 340             /* Calculate temporary vectorial force */
 341             tx               = _mm_mul_pd(fscal,dx00);
 342             ty               = _mm_mul_pd(fscal,dy00);
 343             tz               = _mm_mul_pd(fscal,dz00);
 344
 345             /* Update vectorial force */
 346             fix0             = _mm_add_pd(fix0,tx);
 347             fiy0             = _mm_add_pd(fiy0,ty);
 348             fiz0             = _mm_add_pd(fiz0,tz);
 349
 350             fjx0             = _mm_add_pd(fjx0,tx);
 351             fjy0             = _mm_add_pd(fjy0,ty);
 352             fjz0             = _mm_add_pd(fjz0,tz);
 353
 354             }
 355
 356             /**************************
 357              * CALCULATE INTERACTIONS *
 358              **************************/
 359
 360             if (gmx_mm_any_lt(rsq01,rcutoff2))
 361             {
 362
 363             r01              = _mm_mul_pd(rsq01,rinv01);
 364
 365             /* EWALD ELECTROSTATICS */
 366
 367             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 368             ewrt             = _mm_mul_pd(r01,ewtabscale);
 369             ewitab           = _mm_cvttpd_epi32(ewrt);
 370             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 371             ewitab           = _mm_slli_epi32(ewitab,2);
 372             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 373             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 374             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 375             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 376             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 377             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 378             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 379             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 380             velec            = _mm_mul_pd(qq01,_mm_sub_pd(rinv01,velec));
 381             felec            = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
 382
 383             d                = _mm_sub_pd(r01,rswitch);
 384             d                = _mm_max_pd(d,_mm_setzero_pd());
 385             d2               = _mm_mul_pd(d,d);
 386             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_add_pd(swV3,_mm_mul_pd(d,_mm_add_pd(swV4,_mm_mul_pd(d,swV5)))))));
 387
 388             dsw              = _mm_mul_pd(d2,_mm_add_pd(swF2,_mm_mul_pd(d,_mm_add_pd(swF3,_mm_mul_pd(d,swF4)))));
 389
 390             /* Evaluate switch function */
 391             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 392             felec            = _mm_sub_pd( _mm_mul_pd(felec,sw) , _mm_mul_pd(rinv01,_mm_mul_pd(velec,dsw)) );
 393             velec            = _mm_mul_pd(velec,sw);
 394             cutoff_mask      = _mm_cmplt_pd(rsq01,rcutoff2);
 395
 396             /* Update potential sum for this i atom from the interaction with this j atom. */
 397             velec            = _mm_and_pd(velec,cutoff_mask);
 398             velecsum         = _mm_add_pd(velecsum,velec);
 399
 400             fscal            = felec;
 401
 402             fscal            = _mm_and_pd(fscal,cutoff_mask);
 403
 404             /* Calculate temporary vectorial force */
 405             tx               = _mm_mul_pd(fscal,dx01);
 406             ty               = _mm_mul_pd(fscal,dy01);
 407             tz               = _mm_mul_pd(fscal,dz01);
 408
 409             /* Update vectorial force */
 410             fix0             = _mm_add_pd(fix0,tx);
 411             fiy0             = _mm_add_pd(fiy0,ty);
 412             fiz0             = _mm_add_pd(fiz0,tz);
 413
 414             fjx1             = _mm_add_pd(fjx1,tx);
 415             fjy1             = _mm_add_pd(fjy1,ty);
 416             fjz1             = _mm_add_pd(fjz1,tz);
 417
 418             }
 419
 420             /**************************
 421              * CALCULATE INTERACTIONS *
 422              **************************/
 423
 424             if (gmx_mm_any_lt(rsq02,rcutoff2))
 425             {
 426
 427             r02              = _mm_mul_pd(rsq02,rinv02);
 428
 429             /* EWALD ELECTROSTATICS */
 430
 431             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 432             ewrt             = _mm_mul_pd(r02,ewtabscale);
 433             ewitab           = _mm_cvttpd_epi32(ewrt);
 434             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 435             ewitab           = _mm_slli_epi32(ewitab,2);
 436             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 437             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 438             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 439             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 440             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 441             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 442             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 443             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 444             velec            = _mm_mul_pd(qq02,_mm_sub_pd(rinv02,velec));
 445             felec            = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
 446
 447             d                = _mm_sub_pd(r02,rswitch);
 448             d                = _mm_max_pd(d,_mm_setzero_pd());
 449             d2               = _mm_mul_pd(d,d);
 450             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_add_pd(swV3,_mm_mul_pd(d,_mm_add_pd(swV4,_mm_mul_pd(d,swV5)))))));
 451
 452             dsw              = _mm_mul_pd(d2,_mm_add_pd(swF2,_mm_mul_pd(d,_mm_add_pd(swF3,_mm_mul_pd(d,swF4)))));
 453
 454             /* Evaluate switch function */
 455             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 456             felec            = _mm_sub_pd( _mm_mul_pd(felec,sw) , _mm_mul_pd(rinv02,_mm_mul_pd(velec,dsw)) );
 457             velec            = _mm_mul_pd(velec,sw);
 458             cutoff_mask      = _mm_cmplt_pd(rsq02,rcutoff2);
 459
 460             /* Update potential sum for this i atom from the interaction with this j atom. */
 461             velec            = _mm_and_pd(velec,cutoff_mask);
 462             velecsum         = _mm_add_pd(velecsum,velec);
 463
 464             fscal            = felec;
 465
 466             fscal            = _mm_and_pd(fscal,cutoff_mask);
 467
 468             /* Calculate temporary vectorial force */
 469             tx               = _mm_mul_pd(fscal,dx02);
 470             ty               = _mm_mul_pd(fscal,dy02);
 471             tz               = _mm_mul_pd(fscal,dz02);
 472
 473             /* Update vectorial force */
 474             fix0             = _mm_add_pd(fix0,tx);
 475             fiy0             = _mm_add_pd(fiy0,ty);
 476             fiz0             = _mm_add_pd(fiz0,tz);
 477
 478             fjx2             = _mm_add_pd(fjx2,tx);
 479             fjy2             = _mm_add_pd(fjy2,ty);
 480             fjz2             = _mm_add_pd(fjz2,tz);
 481
 482             }
 483
 484             /**************************
 485              * CALCULATE INTERACTIONS *
 486              **************************/
 487
 488             if (gmx_mm_any_lt(rsq10,rcutoff2))
 489             {
 490
 491             r10              = _mm_mul_pd(rsq10,rinv10);
 492
 493             /* EWALD ELECTROSTATICS */
 494
 495             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 496             ewrt             = _mm_mul_pd(r10,ewtabscale);
 497             ewitab           = _mm_cvttpd_epi32(ewrt);
 498             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 499             ewitab           = _mm_slli_epi32(ewitab,2);
 500             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 501             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 502             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 503             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 504             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 505             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 506             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 507             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 508             velec            = _mm_mul_pd(qq10,_mm_sub_pd(rinv10,velec));
 509             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
 510
 511             d                = _mm_sub_pd(r10,rswitch);
 512             d                = _mm_max_pd(d,_mm_setzero_pd());
 513             d2               = _mm_mul_pd(d,d);
 514             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_add_pd(swV3,_mm_mul_pd(d,_mm_add_pd(swV4,_mm_mul_pd(d,swV5)))))));
 515
 516             dsw              = _mm_mul_pd(d2,_mm_add_pd(swF2,_mm_mul_pd(d,_mm_add_pd(swF3,_mm_mul_pd(d,swF4)))));
 517
 518             /* Evaluate switch function */
 519             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 520             felec            = _mm_sub_pd( _mm_mul_pd(felec,sw) , _mm_mul_pd(rinv10,_mm_mul_pd(velec,dsw)) );
 521             velec            = _mm_mul_pd(velec,sw);
 522             cutoff_mask      = _mm_cmplt_pd(rsq10,rcutoff2);
 523
 524             /* Update potential sum for this i atom from the interaction with this j atom. */
 525             velec            = _mm_and_pd(velec,cutoff_mask);
 526             velecsum         = _mm_add_pd(velecsum,velec);
 527
 528             fscal            = felec;
 529
 530             fscal            = _mm_and_pd(fscal,cutoff_mask);
 531
 532             /* Calculate temporary vectorial force */
 533             tx               = _mm_mul_pd(fscal,dx10);
 534             ty               = _mm_mul_pd(fscal,dy10);
 535             tz               = _mm_mul_pd(fscal,dz10);
 536
 537             /* Update vectorial force */
 538             fix1             = _mm_add_pd(fix1,tx);
 539             fiy1             = _mm_add_pd(fiy1,ty);
 540             fiz1             = _mm_add_pd(fiz1,tz);
 541
 542             fjx0             = _mm_add_pd(fjx0,tx);
 543             fjy0             = _mm_add_pd(fjy0,ty);
 544             fjz0             = _mm_add_pd(fjz0,tz);
 545
 546             }
 547
 548             /**************************
 549              * CALCULATE INTERACTIONS *
 550              **************************/
 551
 552             if (gmx_mm_any_lt(rsq11,rcutoff2))
 553             {
 554
 555             r11              = _mm_mul_pd(rsq11,rinv11);
 556
 557             /* EWALD ELECTROSTATICS */
 558
 559             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 560             ewrt             = _mm_mul_pd(r11,ewtabscale);
 561             ewitab           = _mm_cvttpd_epi32(ewrt);
 562             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 563             ewitab           = _mm_slli_epi32(ewitab,2);
 564             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 565             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 566             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 567             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 568             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 569             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 570             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 571             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 572             velec            = _mm_mul_pd(qq11,_mm_sub_pd(rinv11,velec));
 573             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
 574
 575             d                = _mm_sub_pd(r11,rswitch);
 576             d                = _mm_max_pd(d,_mm_setzero_pd());
 577             d2               = _mm_mul_pd(d,d);
 578             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_add_pd(swV3,_mm_mul_pd(d,_mm_add_pd(swV4,_mm_mul_pd(d,swV5)))))));
 579
 580             dsw              = _mm_mul_pd(d2,_mm_add_pd(swF2,_mm_mul_pd(d,_mm_add_pd(swF3,_mm_mul_pd(d,swF4)))));
 581
 582             /* Evaluate switch function */
 583             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 584             felec            = _mm_sub_pd( _mm_mul_pd(felec,sw) , _mm_mul_pd(rinv11,_mm_mul_pd(velec,dsw)) );
 585             velec            = _mm_mul_pd(velec,sw);
 586             cutoff_mask      = _mm_cmplt_pd(rsq11,rcutoff2);
 587
 588             /* Update potential sum for this i atom from the interaction with this j atom. */
 589             velec            = _mm_and_pd(velec,cutoff_mask);
 590             velecsum         = _mm_add_pd(velecsum,velec);
 591
 592             fscal            = felec;
 593
 594             fscal            = _mm_and_pd(fscal,cutoff_mask);
 595
 596             /* Calculate temporary vectorial force */
 597             tx               = _mm_mul_pd(fscal,dx11);
 598             ty               = _mm_mul_pd(fscal,dy11);
 599             tz               = _mm_mul_pd(fscal,dz11);
 600
 601             /* Update vectorial force */
 602             fix1             = _mm_add_pd(fix1,tx);
 603             fiy1             = _mm_add_pd(fiy1,ty);
 604             fiz1             = _mm_add_pd(fiz1,tz);
 605
 606             fjx1             = _mm_add_pd(fjx1,tx);
 607             fjy1             = _mm_add_pd(fjy1,ty);
 608             fjz1             = _mm_add_pd(fjz1,tz);
 609
 610             }
 611
 612             /**************************
 613              * CALCULATE INTERACTIONS *
 614              **************************/
 615
 616             if (gmx_mm_any_lt(rsq12,rcutoff2))
 617             {
 618
 619             r12              = _mm_mul_pd(rsq12,rinv12);
 620
 621             /* EWALD ELECTROSTATICS */
 622
 623             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 624             ewrt             = _mm_mul_pd(r12,ewtabscale);
 625             ewitab           = _mm_cvttpd_epi32(ewrt);
 626             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 627             ewitab           = _mm_slli_epi32(ewitab,2);
 628             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 629             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 630             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 631             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 632             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 633             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 634             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 635             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 636             velec            = _mm_mul_pd(qq12,_mm_sub_pd(rinv12,velec));
 637             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
 638
 639             d                = _mm_sub_pd(r12,rswitch);
 640             d                = _mm_max_pd(d,_mm_setzero_pd());
 641             d2               = _mm_mul_pd(d,d);
 642             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_add_pd(swV3,_mm_mul_pd(d,_mm_add_pd(swV4,_mm_mul_pd(d,swV5)))))));
 643
 644             dsw              = _mm_mul_pd(d2,_mm_add_pd(swF2,_mm_mul_pd(d,_mm_add_pd(swF3,_mm_mul_pd(d,swF4)))));
 645
 646             /* Evaluate switch function */
 647             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 648             felec            = _mm_sub_pd( _mm_mul_pd(felec,sw) , _mm_mul_pd(rinv12,_mm_mul_pd(velec,dsw)) );
 649             velec            = _mm_mul_pd(velec,sw);
 650             cutoff_mask      = _mm_cmplt_pd(rsq12,rcutoff2);
 651
 652             /* Update potential sum for this i atom from the interaction with this j atom. */
 653             velec            = _mm_and_pd(velec,cutoff_mask);
 654             velecsum         = _mm_add_pd(velecsum,velec);
 655
 656             fscal            = felec;
 657
 658             fscal            = _mm_and_pd(fscal,cutoff_mask);
 659
 660             /* Calculate temporary vectorial force */
 661             tx               = _mm_mul_pd(fscal,dx12);
 662             ty               = _mm_mul_pd(fscal,dy12);
 663             tz               = _mm_mul_pd(fscal,dz12);
 664
 665             /* Update vectorial force */
 666             fix1             = _mm_add_pd(fix1,tx);
 667             fiy1             = _mm_add_pd(fiy1,ty);
 668             fiz1             = _mm_add_pd(fiz1,tz);
 669
 670             fjx2             = _mm_add_pd(fjx2,tx);
 671             fjy2             = _mm_add_pd(fjy2,ty);
 672             fjz2             = _mm_add_pd(fjz2,tz);
 673
 674             }
 675
 676             /**************************
 677              * CALCULATE INTERACTIONS *
 678              **************************/
 679
 680             if (gmx_mm_any_lt(rsq20,rcutoff2))
 681             {
 682
 683             r20              = _mm_mul_pd(rsq20,rinv20);
 684
 685             /* EWALD ELECTROSTATICS */
 686
 687             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 688             ewrt             = _mm_mul_pd(r20,ewtabscale);
 689             ewitab           = _mm_cvttpd_epi32(ewrt);
 690             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 691             ewitab           = _mm_slli_epi32(ewitab,2);
 692             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 693             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 694             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 695             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 696             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 697             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 698             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 699             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 700             velec            = _mm_mul_pd(qq20,_mm_sub_pd(rinv20,velec));
 701             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
 702
 703             d                = _mm_sub_pd(r20,rswitch);
 704             d                = _mm_max_pd(d,_mm_setzero_pd());
 705             d2               = _mm_mul_pd(d,d);
 706             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_add_pd(swV3,_mm_mul_pd(d,_mm_add_pd(swV4,_mm_mul_pd(d,swV5)))))));
 707
 708             dsw              = _mm_mul_pd(d2,_mm_add_pd(swF2,_mm_mul_pd(d,_mm_add_pd(swF3,_mm_mul_pd(d,swF4)))));
 709
 710             /* Evaluate switch function */
 711             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 712             felec            = _mm_sub_pd( _mm_mul_pd(felec,sw) , _mm_mul_pd(rinv20,_mm_mul_pd(velec,dsw)) );
 713             velec            = _mm_mul_pd(velec,sw);
 714             cutoff_mask      = _mm_cmplt_pd(rsq20,rcutoff2);
 715
 716             /* Update potential sum for this i atom from the interaction with this j atom. */
 717             velec            = _mm_and_pd(velec,cutoff_mask);
 718             velecsum         = _mm_add_pd(velecsum,velec);
 719
 720             fscal            = felec;
 721
 722             fscal            = _mm_and_pd(fscal,cutoff_mask);
 723
 724             /* Calculate temporary vectorial force */
 725             tx               = _mm_mul_pd(fscal,dx20);
 726             ty               = _mm_mul_pd(fscal,dy20);
 727             tz               = _mm_mul_pd(fscal,dz20);
 728
 729             /* Update vectorial force */
 730             fix2             = _mm_add_pd(fix2,tx);
 731             fiy2             = _mm_add_pd(fiy2,ty);
 732             fiz2             = _mm_add_pd(fiz2,tz);
 733
 734             fjx0             = _mm_add_pd(fjx0,tx);
 735             fjy0             = _mm_add_pd(fjy0,ty);
 736             fjz0             = _mm_add_pd(fjz0,tz);
 737
 738             }
 739
 740             /**************************
 741              * CALCULATE INTERACTIONS *
 742              **************************/
 743
 744             if (gmx_mm_any_lt(rsq21,rcutoff2))
 745             {
 746
 747             r21              = _mm_mul_pd(rsq21,rinv21);
 748
 749             /* EWALD ELECTROSTATICS */
 750
 751             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 752             ewrt             = _mm_mul_pd(r21,ewtabscale);
 753             ewitab           = _mm_cvttpd_epi32(ewrt);
 754             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 755             ewitab           = _mm_slli_epi32(ewitab,2);
 756             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 757             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 758             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 759             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 760             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 761             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 762             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 763             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 764             velec            = _mm_mul_pd(qq21,_mm_sub_pd(rinv21,velec));
 765             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
 766
 767             d                = _mm_sub_pd(r21,rswitch);
 768             d                = _mm_max_pd(d,_mm_setzero_pd());
 769             d2               = _mm_mul_pd(d,d);
 770             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_add_pd(swV3,_mm_mul_pd(d,_mm_add_pd(swV4,_mm_mul_pd(d,swV5)))))));
 771
 772             dsw              = _mm_mul_pd(d2,_mm_add_pd(swF2,_mm_mul_pd(d,_mm_add_pd(swF3,_mm_mul_pd(d,swF4)))));
 773
 774             /* Evaluate switch function */
 775             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 776             felec            = _mm_sub_pd( _mm_mul_pd(felec,sw) , _mm_mul_pd(rinv21,_mm_mul_pd(velec,dsw)) );
 777             velec            = _mm_mul_pd(velec,sw);
 778             cutoff_mask      = _mm_cmplt_pd(rsq21,rcutoff2);
 779
 780             /* Update potential sum for this i atom from the interaction with this j atom. */
 781             velec            = _mm_and_pd(velec,cutoff_mask);
 782             velecsum         = _mm_add_pd(velecsum,velec);
 783
 784             fscal            = felec;
 785
 786             fscal            = _mm_and_pd(fscal,cutoff_mask);
 787
 788             /* Calculate temporary vectorial force */
 789             tx               = _mm_mul_pd(fscal,dx21);
 790             ty               = _mm_mul_pd(fscal,dy21);
 791             tz               = _mm_mul_pd(fscal,dz21);
 792
 793             /* Update vectorial force */
 794             fix2             = _mm_add_pd(fix2,tx);
 795             fiy2             = _mm_add_pd(fiy2,ty);
 796             fiz2             = _mm_add_pd(fiz2,tz);
 797
 798             fjx1             = _mm_add_pd(fjx1,tx);
 799             fjy1             = _mm_add_pd(fjy1,ty);
 800             fjz1             = _mm_add_pd(fjz1,tz);
 801
 802             }
 803
 804             /**************************
 805              * CALCULATE INTERACTIONS *
 806              **************************/
 807
 808             if (gmx_mm_any_lt(rsq22,rcutoff2))
 809             {
 810
 811             r22              = _mm_mul_pd(rsq22,rinv22);
 812
 813             /* EWALD ELECTROSTATICS */
 814
 815             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 816             ewrt             = _mm_mul_pd(r22,ewtabscale);
 817             ewitab           = _mm_cvttpd_epi32(ewrt);
 818             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 819             ewitab           = _mm_slli_epi32(ewitab,2);
 820             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 821             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 822             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 823             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 824             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 825             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 826             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 827             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 828             velec            = _mm_mul_pd(qq22,_mm_sub_pd(rinv22,velec));
 829             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
 830
 831             d                = _mm_sub_pd(r22,rswitch);
 832             d                = _mm_max_pd(d,_mm_setzero_pd());
 833             d2               = _mm_mul_pd(d,d);
 834             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_add_pd(swV3,_mm_mul_pd(d,_mm_add_pd(swV4,_mm_mul_pd(d,swV5)))))));
 835
 836             dsw              = _mm_mul_pd(d2,_mm_add_pd(swF2,_mm_mul_pd(d,_mm_add_pd(swF3,_mm_mul_pd(d,swF4)))));
 837
 838             /* Evaluate switch function */
 839             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 840             felec            = _mm_sub_pd( _mm_mul_pd(felec,sw) , _mm_mul_pd(rinv22,_mm_mul_pd(velec,dsw)) );
 841             velec            = _mm_mul_pd(velec,sw);
 842             cutoff_mask      = _mm_cmplt_pd(rsq22,rcutoff2);
 843
 844             /* Update potential sum for this i atom from the interaction with this j atom. */
 845             velec            = _mm_and_pd(velec,cutoff_mask);
 846             velecsum         = _mm_add_pd(velecsum,velec);
 847
 848             fscal            = felec;
 849
 850             fscal            = _mm_and_pd(fscal,cutoff_mask);
 851
 852             /* Calculate temporary vectorial force */
 853             tx               = _mm_mul_pd(fscal,dx22);
 854             ty               = _mm_mul_pd(fscal,dy22);
 855             tz               = _mm_mul_pd(fscal,dz22);
 856
 857             /* Update vectorial force */
 858             fix2             = _mm_add_pd(fix2,tx);
 859             fiy2             = _mm_add_pd(fiy2,ty);
 860             fiz2             = _mm_add_pd(fiz2,tz);
 861
 862             fjx2             = _mm_add_pd(fjx2,tx);
 863             fjy2             = _mm_add_pd(fjy2,ty);
 864             fjz2             = _mm_add_pd(fjz2,tz);
 865
 866             }
 867
 868             gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 869
 870             /* Inner loop uses 585 flops */
 871         }
 872
 873         if(jidx<j_index_end)
 874         {
 875
 876             jnrA             = jjnr[jidx];
 877             j_coord_offsetA  = DIM*jnrA;
 878
 879             /* load j atom coordinates */
 880             gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 881                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 882
 883             /* Calculate displacement vector */
 884             dx00             = _mm_sub_pd(ix0,jx0);
 885             dy00             = _mm_sub_pd(iy0,jy0);
 886             dz00             = _mm_sub_pd(iz0,jz0);
 887             dx01             = _mm_sub_pd(ix0,jx1);
 888             dy01             = _mm_sub_pd(iy0,jy1);
 889             dz01             = _mm_sub_pd(iz0,jz1);
 890             dx02             = _mm_sub_pd(ix0,jx2);
 891             dy02             = _mm_sub_pd(iy0,jy2);
 892             dz02             = _mm_sub_pd(iz0,jz2);
 893             dx10             = _mm_sub_pd(ix1,jx0);
 894             dy10             = _mm_sub_pd(iy1,jy0);
 895             dz10             = _mm_sub_pd(iz1,jz0);
 896             dx11             = _mm_sub_pd(ix1,jx1);
 897             dy11             = _mm_sub_pd(iy1,jy1);
 898             dz11             = _mm_sub_pd(iz1,jz1);
 899             dx12             = _mm_sub_pd(ix1,jx2);
 900             dy12             = _mm_sub_pd(iy1,jy2);
 901             dz12             = _mm_sub_pd(iz1,jz2);
 902             dx20             = _mm_sub_pd(ix2,jx0);
 903             dy20             = _mm_sub_pd(iy2,jy0);
 904             dz20             = _mm_sub_pd(iz2,jz0);
 905             dx21             = _mm_sub_pd(ix2,jx1);
 906             dy21             = _mm_sub_pd(iy2,jy1);
 907             dz21             = _mm_sub_pd(iz2,jz1);
 908             dx22             = _mm_sub_pd(ix2,jx2);
 909             dy22             = _mm_sub_pd(iy2,jy2);
 910             dz22             = _mm_sub_pd(iz2,jz2);
 911
 912             /* Calculate squared distance and things based on it */
 913             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 914             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
 915             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
 916             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 917             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 918             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 919             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 920             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 921             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 922
 923             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 924             rinv01           = gmx_mm_invsqrt_pd(rsq01);
 925             rinv02           = gmx_mm_invsqrt_pd(rsq02);
 926             rinv10           = gmx_mm_invsqrt_pd(rsq10);
 927             rinv11           = gmx_mm_invsqrt_pd(rsq11);
 928             rinv12           = gmx_mm_invsqrt_pd(rsq12);
 929             rinv20           = gmx_mm_invsqrt_pd(rsq20);
 930             rinv21           = gmx_mm_invsqrt_pd(rsq21);
 931             rinv22           = gmx_mm_invsqrt_pd(rsq22);
 932
 933             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 934             rinvsq01         = _mm_mul_pd(rinv01,rinv01);
 935             rinvsq02         = _mm_mul_pd(rinv02,rinv02);
 936             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
 937             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
 938             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
 939             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
 940             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
 941             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
 942
 943             fjx0             = _mm_setzero_pd();
 944             fjy0             = _mm_setzero_pd();
 945             fjz0             = _mm_setzero_pd();
 946             fjx1             = _mm_setzero_pd();
 947             fjy1             = _mm_setzero_pd();
 948             fjz1             = _mm_setzero_pd();
 949             fjx2             = _mm_setzero_pd();
 950             fjy2             = _mm_setzero_pd();
 951             fjz2             = _mm_setzero_pd();
 952
 953             /**************************
 954              * CALCULATE INTERACTIONS *
 955              **************************/
 956
 957             if (gmx_mm_any_lt(rsq00,rcutoff2))
 958             {
 959
 960             r00              = _mm_mul_pd(rsq00,rinv00);
 961
 962             /* EWALD ELECTROSTATICS */
 963
 964             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 965             ewrt             = _mm_mul_pd(r00,ewtabscale);
 966             ewitab           = _mm_cvttpd_epi32(ewrt);
 967             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 968             ewitab           = _mm_slli_epi32(ewitab,2);
 969             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 970             ewtabD           = _mm_setzero_pd();
 971             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 972             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 973             ewtabFn          = _mm_setzero_pd();
 974             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 975             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 976             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 977             velec            = _mm_mul_pd(qq00,_mm_sub_pd(rinv00,velec));
 978             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
 979
 980             d                = _mm_sub_pd(r00,rswitch);
 981             d                = _mm_max_pd(d,_mm_setzero_pd());
 982             d2               = _mm_mul_pd(d,d);
 983             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_add_pd(swV3,_mm_mul_pd(d,_mm_add_pd(swV4,_mm_mul_pd(d,swV5)))))));
 984
 985             dsw              = _mm_mul_pd(d2,_mm_add_pd(swF2,_mm_mul_pd(d,_mm_add_pd(swF3,_mm_mul_pd(d,swF4)))));
 986
 987             /* Evaluate switch function */
 988             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 989             felec            = _mm_sub_pd( _mm_mul_pd(felec,sw) , _mm_mul_pd(rinv00,_mm_mul_pd(velec,dsw)) );
 990             velec            = _mm_mul_pd(velec,sw);
 991             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
 992
 993             /* Update potential sum for this i atom from the interaction with this j atom. */
 994             velec            = _mm_and_pd(velec,cutoff_mask);
 995             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 996             velecsum         = _mm_add_pd(velecsum,velec);
 997
 998             fscal            = felec;
 999
1000             fscal            = _mm_and_pd(fscal,cutoff_mask);
1001
1002             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1003
1004             /* Calculate temporary vectorial force */
1005             tx               = _mm_mul_pd(fscal,dx00);
1006             ty               = _mm_mul_pd(fscal,dy00);
1007             tz               = _mm_mul_pd(fscal,dz00);
1008
1009             /* Update vectorial force */
1010             fix0             = _mm_add_pd(fix0,tx);
1011             fiy0             = _mm_add_pd(fiy0,ty);
1012             fiz0             = _mm_add_pd(fiz0,tz);
1013
1014             fjx0             = _mm_add_pd(fjx0,tx);
1015             fjy0             = _mm_add_pd(fjy0,ty);
1016             fjz0             = _mm_add_pd(fjz0,tz);
1017
1018             }
1019
1020             /**************************
1021              * CALCULATE INTERACTIONS *
1022              **************************/
1023
1024             if (gmx_mm_any_lt(rsq01,rcutoff2))
1025             {
1026
1027             r01              = _mm_mul_pd(rsq01,rinv01);
1028
1029             /* EWALD ELECTROSTATICS */
1030
1031             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1032             ewrt             = _mm_mul_pd(r01,ewtabscale);
1033             ewitab           = _mm_cvttpd_epi32(ewrt);
1034             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1035             ewitab           = _mm_slli_epi32(ewitab,2);
1036             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1037             ewtabD           = _mm_setzero_pd();
1038             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1039             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1040             ewtabFn          = _mm_setzero_pd();
1041             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1042             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1043             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1044             velec            = _mm_mul_pd(qq01,_mm_sub_pd(rinv01,velec));
1045             felec            = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
1046
1047             d                = _mm_sub_pd(r01,rswitch);
1048             d                = _mm_max_pd(d,_mm_setzero_pd());
1049             d2               = _mm_mul_pd(d,d);
1050             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_add_pd(swV3,_mm_mul_pd(d,_mm_add_pd(swV4,_mm_mul_pd(d,swV5)))))));
1051
1052             dsw              = _mm_mul_pd(d2,_mm_add_pd(swF2,_mm_mul_pd(d,_mm_add_pd(swF3,_mm_mul_pd(d,swF4)))));
1053
1054             /* Evaluate switch function */
1055             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1056             felec            = _mm_sub_pd( _mm_mul_pd(felec,sw) , _mm_mul_pd(rinv01,_mm_mul_pd(velec,dsw)) );
1057             velec            = _mm_mul_pd(velec,sw);
1058             cutoff_mask      = _mm_cmplt_pd(rsq01,rcutoff2);
1059
1060             /* Update potential sum for this i atom from the interaction with this j atom. */
1061             velec            = _mm_and_pd(velec,cutoff_mask);
1062             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1063             velecsum         = _mm_add_pd(velecsum,velec);
1064
1065             fscal            = felec;
1066
1067             fscal            = _mm_and_pd(fscal,cutoff_mask);
1068
1069             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1070
1071             /* Calculate temporary vectorial force */
1072             tx               = _mm_mul_pd(fscal,dx01);
1073             ty               = _mm_mul_pd(fscal,dy01);
1074             tz               = _mm_mul_pd(fscal,dz01);
1075
1076             /* Update vectorial force */
1077             fix0             = _mm_add_pd(fix0,tx);
1078             fiy0             = _mm_add_pd(fiy0,ty);
1079             fiz0             = _mm_add_pd(fiz0,tz);
1080
1081             fjx1             = _mm_add_pd(fjx1,tx);
1082             fjy1             = _mm_add_pd(fjy1,ty);
1083             fjz1             = _mm_add_pd(fjz1,tz);
1084
1085             }
1086
1087             /**************************
1088              * CALCULATE INTERACTIONS *
1089              **************************/
1090
1091             if (gmx_mm_any_lt(rsq02,rcutoff2))
1092             {
1093
1094             r02              = _mm_mul_pd(rsq02,rinv02);
1095
1096             /* EWALD ELECTROSTATICS */
1097
1098             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1099             ewrt             = _mm_mul_pd(r02,ewtabscale);
1100             ewitab           = _mm_cvttpd_epi32(ewrt);
1101             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1102             ewitab           = _mm_slli_epi32(ewitab,2);
1103             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1104             ewtabD           = _mm_setzero_pd();
1105             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1106             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1107             ewtabFn          = _mm_setzero_pd();
1108             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1109             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1110             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1111             velec            = _mm_mul_pd(qq02,_mm_sub_pd(rinv02,velec));
1112             felec            = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
1113
1114             d                = _mm_sub_pd(r02,rswitch);
1115             d                = _mm_max_pd(d,_mm_setzero_pd());
1116             d2               = _mm_mul_pd(d,d);
1117             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_add_pd(swV3,_mm_mul_pd(d,_mm_add_pd(swV4,_mm_mul_pd(d,swV5)))))));
1118
1119             dsw              = _mm_mul_pd(d2,_mm_add_pd(swF2,_mm_mul_pd(d,_mm_add_pd(swF3,_mm_mul_pd(d,swF4)))));
1120
1121             /* Evaluate switch function */
1122             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1123             felec            = _mm_sub_pd( _mm_mul_pd(felec,sw) , _mm_mul_pd(rinv02,_mm_mul_pd(velec,dsw)) );
1124             velec            = _mm_mul_pd(velec,sw);
1125             cutoff_mask      = _mm_cmplt_pd(rsq02,rcutoff2);
1126
1127             /* Update potential sum for this i atom from the interaction with this j atom. */
1128             velec            = _mm_and_pd(velec,cutoff_mask);
1129             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1130             velecsum         = _mm_add_pd(velecsum,velec);
1131
1132             fscal            = felec;
1133
1134             fscal            = _mm_and_pd(fscal,cutoff_mask);
1135
1136             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1137
1138             /* Calculate temporary vectorial force */
1139             tx               = _mm_mul_pd(fscal,dx02);
1140             ty               = _mm_mul_pd(fscal,dy02);
1141             tz               = _mm_mul_pd(fscal,dz02);
1142
1143             /* Update vectorial force */
1144             fix0             = _mm_add_pd(fix0,tx);
1145             fiy0             = _mm_add_pd(fiy0,ty);
1146             fiz0             = _mm_add_pd(fiz0,tz);
1147
1148             fjx2             = _mm_add_pd(fjx2,tx);
1149             fjy2             = _mm_add_pd(fjy2,ty);
1150             fjz2             = _mm_add_pd(fjz2,tz);
1151
1152             }
1153
1154             /**************************
1155              * CALCULATE INTERACTIONS *
1156              **************************/
1157
1158             if (gmx_mm_any_lt(rsq10,rcutoff2))
1159             {
1160
1161             r10              = _mm_mul_pd(rsq10,rinv10);
1162
1163             /* EWALD ELECTROSTATICS */
1164
1165             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1166             ewrt             = _mm_mul_pd(r10,ewtabscale);
1167             ewitab           = _mm_cvttpd_epi32(ewrt);
1168             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1169             ewitab           = _mm_slli_epi32(ewitab,2);
1170             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1171             ewtabD           = _mm_setzero_pd();
1172             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1173             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1174             ewtabFn          = _mm_setzero_pd();
1175             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1176             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1177             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1178             velec            = _mm_mul_pd(qq10,_mm_sub_pd(rinv10,velec));
1179             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
1180
1181             d                = _mm_sub_pd(r10,rswitch);
1182             d                = _mm_max_pd(d,_mm_setzero_pd());
1183             d2               = _mm_mul_pd(d,d);
1184             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_add_pd(swV3,_mm_mul_pd(d,_mm_add_pd(swV4,_mm_mul_pd(d,swV5)))))));
1185
1186             dsw              = _mm_mul_pd(d2,_mm_add_pd(swF2,_mm_mul_pd(d,_mm_add_pd(swF3,_mm_mul_pd(d,swF4)))));
1187
1188             /* Evaluate switch function */
1189             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1190             felec            = _mm_sub_pd( _mm_mul_pd(felec,sw) , _mm_mul_pd(rinv10,_mm_mul_pd(velec,dsw)) );
1191             velec            = _mm_mul_pd(velec,sw);
1192             cutoff_mask      = _mm_cmplt_pd(rsq10,rcutoff2);
1193
1194             /* Update potential sum for this i atom from the interaction with this j atom. */
1195             velec            = _mm_and_pd(velec,cutoff_mask);
1196             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1197             velecsum         = _mm_add_pd(velecsum,velec);
1198
1199             fscal            = felec;
1200
1201             fscal            = _mm_and_pd(fscal,cutoff_mask);
1202
1203             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1204
1205             /* Calculate temporary vectorial force */
1206             tx               = _mm_mul_pd(fscal,dx10);
1207             ty               = _mm_mul_pd(fscal,dy10);
1208             tz               = _mm_mul_pd(fscal,dz10);
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             fjx0             = _mm_add_pd(fjx0,tx);
1216             fjy0             = _mm_add_pd(fjy0,ty);
1217             fjz0             = _mm_add_pd(fjz0,tz);
1218
1219             }
1220
1221             /**************************
1222              * CALCULATE INTERACTIONS *
1223              **************************/
1224
1225             if (gmx_mm_any_lt(rsq11,rcutoff2))
1226             {
1227
1228             r11              = _mm_mul_pd(rsq11,rinv11);
1229
1230             /* EWALD ELECTROSTATICS */
1231
1232             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1233             ewrt             = _mm_mul_pd(r11,ewtabscale);
1234             ewitab           = _mm_cvttpd_epi32(ewrt);
1235             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1236             ewitab           = _mm_slli_epi32(ewitab,2);
1237             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1238             ewtabD           = _mm_setzero_pd();
1239             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1240             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1241             ewtabFn          = _mm_setzero_pd();
1242             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1243             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1244             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1245             velec            = _mm_mul_pd(qq11,_mm_sub_pd(rinv11,velec));
1246             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
1247
1248             d                = _mm_sub_pd(r11,rswitch);
1249             d                = _mm_max_pd(d,_mm_setzero_pd());
1250             d2               = _mm_mul_pd(d,d);
1251             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_add_pd(swV3,_mm_mul_pd(d,_mm_add_pd(swV4,_mm_mul_pd(d,swV5)))))));
1252
1253             dsw              = _mm_mul_pd(d2,_mm_add_pd(swF2,_mm_mul_pd(d,_mm_add_pd(swF3,_mm_mul_pd(d,swF4)))));
1254
1255             /* Evaluate switch function */
1256             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1257             felec            = _mm_sub_pd( _mm_mul_pd(felec,sw) , _mm_mul_pd(rinv11,_mm_mul_pd(velec,dsw)) );
1258             velec            = _mm_mul_pd(velec,sw);
1259             cutoff_mask      = _mm_cmplt_pd(rsq11,rcutoff2);
1260
1261             /* Update potential sum for this i atom from the interaction with this j atom. */
1262             velec            = _mm_and_pd(velec,cutoff_mask);
1263             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1264             velecsum         = _mm_add_pd(velecsum,velec);
1265
1266             fscal            = felec;
1267
1268             fscal            = _mm_and_pd(fscal,cutoff_mask);
1269
1270             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1271
1272             /* Calculate temporary vectorial force */
1273             tx               = _mm_mul_pd(fscal,dx11);
1274             ty               = _mm_mul_pd(fscal,dy11);
1275             tz               = _mm_mul_pd(fscal,dz11);
1276
1277             /* Update vectorial force */
1278             fix1             = _mm_add_pd(fix1,tx);
1279             fiy1             = _mm_add_pd(fiy1,ty);
1280             fiz1             = _mm_add_pd(fiz1,tz);
1281
1282             fjx1             = _mm_add_pd(fjx1,tx);
1283             fjy1             = _mm_add_pd(fjy1,ty);
1284             fjz1             = _mm_add_pd(fjz1,tz);
1285
1286             }
1287
1288             /**************************
1289              * CALCULATE INTERACTIONS *
1290              **************************/
1291
1292             if (gmx_mm_any_lt(rsq12,rcutoff2))
1293             {
1294
1295             r12              = _mm_mul_pd(rsq12,rinv12);
1296
1297             /* EWALD ELECTROSTATICS */
1298
1299             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1300             ewrt             = _mm_mul_pd(r12,ewtabscale);
1301             ewitab           = _mm_cvttpd_epi32(ewrt);
1302             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1303             ewitab           = _mm_slli_epi32(ewitab,2);
1304             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1305             ewtabD           = _mm_setzero_pd();
1306             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1307             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1308             ewtabFn          = _mm_setzero_pd();
1309             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1310             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1311             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1312             velec            = _mm_mul_pd(qq12,_mm_sub_pd(rinv12,velec));
1313             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
1314
1315             d                = _mm_sub_pd(r12,rswitch);
1316             d                = _mm_max_pd(d,_mm_setzero_pd());
1317             d2               = _mm_mul_pd(d,d);
1318             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_add_pd(swV3,_mm_mul_pd(d,_mm_add_pd(swV4,_mm_mul_pd(d,swV5)))))));
1319
1320             dsw              = _mm_mul_pd(d2,_mm_add_pd(swF2,_mm_mul_pd(d,_mm_add_pd(swF3,_mm_mul_pd(d,swF4)))));
1321
1322             /* Evaluate switch function */
1323             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1324             felec            = _mm_sub_pd( _mm_mul_pd(felec,sw) , _mm_mul_pd(rinv12,_mm_mul_pd(velec,dsw)) );
1325             velec            = _mm_mul_pd(velec,sw);
1326             cutoff_mask      = _mm_cmplt_pd(rsq12,rcutoff2);
1327
1328             /* Update potential sum for this i atom from the interaction with this j atom. */
1329             velec            = _mm_and_pd(velec,cutoff_mask);
1330             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1331             velecsum         = _mm_add_pd(velecsum,velec);
1332
1333             fscal            = felec;
1334
1335             fscal            = _mm_and_pd(fscal,cutoff_mask);
1336
1337             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1338
1339             /* Calculate temporary vectorial force */
1340             tx               = _mm_mul_pd(fscal,dx12);
1341             ty               = _mm_mul_pd(fscal,dy12);
1342             tz               = _mm_mul_pd(fscal,dz12);
1343
1344             /* Update vectorial force */
1345             fix1             = _mm_add_pd(fix1,tx);
1346             fiy1             = _mm_add_pd(fiy1,ty);
1347             fiz1             = _mm_add_pd(fiz1,tz);
1348
1349             fjx2             = _mm_add_pd(fjx2,tx);
1350             fjy2             = _mm_add_pd(fjy2,ty);
1351             fjz2             = _mm_add_pd(fjz2,tz);
1352
1353             }
1354
1355             /**************************
1356              * CALCULATE INTERACTIONS *
1357              **************************/
1358
1359             if (gmx_mm_any_lt(rsq20,rcutoff2))
1360             {
1361
1362             r20              = _mm_mul_pd(rsq20,rinv20);
1363
1364             /* EWALD ELECTROSTATICS */
1365
1366             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1367             ewrt             = _mm_mul_pd(r20,ewtabscale);
1368             ewitab           = _mm_cvttpd_epi32(ewrt);
1369             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1370             ewitab           = _mm_slli_epi32(ewitab,2);
1371             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1372             ewtabD           = _mm_setzero_pd();
1373             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1374             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1375             ewtabFn          = _mm_setzero_pd();
1376             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1377             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1378             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1379             velec            = _mm_mul_pd(qq20,_mm_sub_pd(rinv20,velec));
1380             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
1381
1382             d                = _mm_sub_pd(r20,rswitch);
1383             d                = _mm_max_pd(d,_mm_setzero_pd());
1384             d2               = _mm_mul_pd(d,d);
1385             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_add_pd(swV3,_mm_mul_pd(d,_mm_add_pd(swV4,_mm_mul_pd(d,swV5)))))));
1386
1387             dsw              = _mm_mul_pd(d2,_mm_add_pd(swF2,_mm_mul_pd(d,_mm_add_pd(swF3,_mm_mul_pd(d,swF4)))));
1388
1389             /* Evaluate switch function */
1390             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1391             felec            = _mm_sub_pd( _mm_mul_pd(felec,sw) , _mm_mul_pd(rinv20,_mm_mul_pd(velec,dsw)) );
1392             velec            = _mm_mul_pd(velec,sw);
1393             cutoff_mask      = _mm_cmplt_pd(rsq20,rcutoff2);
1394
1395             /* Update potential sum for this i atom from the interaction with this j atom. */
1396             velec            = _mm_and_pd(velec,cutoff_mask);
1397             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1398             velecsum         = _mm_add_pd(velecsum,velec);
1399
1400             fscal            = felec;
1401
1402             fscal            = _mm_and_pd(fscal,cutoff_mask);
1403
1404             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1405
1406             /* Calculate temporary vectorial force */
1407             tx               = _mm_mul_pd(fscal,dx20);
1408             ty               = _mm_mul_pd(fscal,dy20);
1409             tz               = _mm_mul_pd(fscal,dz20);
1410
1411             /* Update vectorial force */
1412             fix2             = _mm_add_pd(fix2,tx);
1413             fiy2             = _mm_add_pd(fiy2,ty);
1414             fiz2             = _mm_add_pd(fiz2,tz);
1415
1416             fjx0             = _mm_add_pd(fjx0,tx);
1417             fjy0             = _mm_add_pd(fjy0,ty);
1418             fjz0             = _mm_add_pd(fjz0,tz);
1419
1420             }
1421
1422             /**************************
1423              * CALCULATE INTERACTIONS *
1424              **************************/
1425
1426             if (gmx_mm_any_lt(rsq21,rcutoff2))
1427             {
1428
1429             r21              = _mm_mul_pd(rsq21,rinv21);
1430
1431             /* EWALD ELECTROSTATICS */
1432
1433             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1434             ewrt             = _mm_mul_pd(r21,ewtabscale);
1435             ewitab           = _mm_cvttpd_epi32(ewrt);
1436             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1437             ewitab           = _mm_slli_epi32(ewitab,2);
1438             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1439             ewtabD           = _mm_setzero_pd();
1440             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1441             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1442             ewtabFn          = _mm_setzero_pd();
1443             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1444             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1445             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1446             velec            = _mm_mul_pd(qq21,_mm_sub_pd(rinv21,velec));
1447             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
1448
1449             d                = _mm_sub_pd(r21,rswitch);
1450             d                = _mm_max_pd(d,_mm_setzero_pd());
1451             d2               = _mm_mul_pd(d,d);
1452             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_add_pd(swV3,_mm_mul_pd(d,_mm_add_pd(swV4,_mm_mul_pd(d,swV5)))))));
1453
1454             dsw              = _mm_mul_pd(d2,_mm_add_pd(swF2,_mm_mul_pd(d,_mm_add_pd(swF3,_mm_mul_pd(d,swF4)))));
1455
1456             /* Evaluate switch function */
1457             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1458             felec            = _mm_sub_pd( _mm_mul_pd(felec,sw) , _mm_mul_pd(rinv21,_mm_mul_pd(velec,dsw)) );
1459             velec            = _mm_mul_pd(velec,sw);
1460             cutoff_mask      = _mm_cmplt_pd(rsq21,rcutoff2);
1461
1462             /* Update potential sum for this i atom from the interaction with this j atom. */
1463             velec            = _mm_and_pd(velec,cutoff_mask);
1464             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1465             velecsum         = _mm_add_pd(velecsum,velec);
1466
1467             fscal            = felec;
1468
1469             fscal            = _mm_and_pd(fscal,cutoff_mask);
1470
1471             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1472
1473             /* Calculate temporary vectorial force */
1474             tx               = _mm_mul_pd(fscal,dx21);
1475             ty               = _mm_mul_pd(fscal,dy21);
1476             tz               = _mm_mul_pd(fscal,dz21);
1477
1478             /* Update vectorial force */
1479             fix2             = _mm_add_pd(fix2,tx);
1480             fiy2             = _mm_add_pd(fiy2,ty);
1481             fiz2             = _mm_add_pd(fiz2,tz);
1482
1483             fjx1             = _mm_add_pd(fjx1,tx);
1484             fjy1             = _mm_add_pd(fjy1,ty);
1485             fjz1             = _mm_add_pd(fjz1,tz);
1486
1487             }
1488
1489             /**************************
1490              * CALCULATE INTERACTIONS *
1491              **************************/
1492
1493             if (gmx_mm_any_lt(rsq22,rcutoff2))
1494             {
1495
1496             r22              = _mm_mul_pd(rsq22,rinv22);
1497
1498             /* EWALD ELECTROSTATICS */
1499
1500             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1501             ewrt             = _mm_mul_pd(r22,ewtabscale);
1502             ewitab           = _mm_cvttpd_epi32(ewrt);
1503             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1504             ewitab           = _mm_slli_epi32(ewitab,2);
1505             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1506             ewtabD           = _mm_setzero_pd();
1507             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1508             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1509             ewtabFn          = _mm_setzero_pd();
1510             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1511             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1512             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1513             velec            = _mm_mul_pd(qq22,_mm_sub_pd(rinv22,velec));
1514             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
1515
1516             d                = _mm_sub_pd(r22,rswitch);
1517             d                = _mm_max_pd(d,_mm_setzero_pd());
1518             d2               = _mm_mul_pd(d,d);
1519             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_add_pd(swV3,_mm_mul_pd(d,_mm_add_pd(swV4,_mm_mul_pd(d,swV5)))))));
1520
1521             dsw              = _mm_mul_pd(d2,_mm_add_pd(swF2,_mm_mul_pd(d,_mm_add_pd(swF3,_mm_mul_pd(d,swF4)))));
1522
1523             /* Evaluate switch function */
1524             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1525             felec            = _mm_sub_pd( _mm_mul_pd(felec,sw) , _mm_mul_pd(rinv22,_mm_mul_pd(velec,dsw)) );
1526             velec            = _mm_mul_pd(velec,sw);
1527             cutoff_mask      = _mm_cmplt_pd(rsq22,rcutoff2);
1528
1529             /* Update potential sum for this i atom from the interaction with this j atom. */
1530             velec            = _mm_and_pd(velec,cutoff_mask);
1531             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1532             velecsum         = _mm_add_pd(velecsum,velec);
1533
1534             fscal            = felec;
1535
1536             fscal            = _mm_and_pd(fscal,cutoff_mask);
1537
1538             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1539
1540             /* Calculate temporary vectorial force */
1541             tx               = _mm_mul_pd(fscal,dx22);
1542             ty               = _mm_mul_pd(fscal,dy22);
1543             tz               = _mm_mul_pd(fscal,dz22);
1544
1545             /* Update vectorial force */
1546             fix2             = _mm_add_pd(fix2,tx);
1547             fiy2             = _mm_add_pd(fiy2,ty);
1548             fiz2             = _mm_add_pd(fiz2,tz);
1549
1550             fjx2             = _mm_add_pd(fjx2,tx);
1551             fjy2             = _mm_add_pd(fjy2,ty);
1552             fjz2             = _mm_add_pd(fjz2,tz);
1553
1554             }
1555
1556             gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1557
1558             /* Inner loop uses 585 flops */
1559         }
1560
1561         /* End of innermost loop */
1562
1563         gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1564                                               f+i_coord_offset,fshift+i_shift_offset);
1565
1566         ggid                        = gid[iidx];
1567         /* Update potential energies */
1568         gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1569
1570         /* Increment number of inner iterations */
1571         inneriter                  += j_index_end - j_index_start;
1572
1573         /* Outer loop uses 19 flops */
1574     }
1575
1576     /* Increment number of outer iterations */
1577     outeriter        += nri;
1578
1579     /* Update outer/inner flops */
1580
1581     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3W3_VF,outeriter*19 + inneriter*585);
1582 }
1583 /*
1584  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSw_VdwNone_GeomW3W3_F_sse4_1_double
1585  * Electrostatics interaction: Ewald
1586  * VdW interaction:            None
1587  * Geometry:                   Water3-Water3
1588  * Calculate force/pot:        Force
1589  */
1590 void
1591 nb_kernel_ElecEwSw_VdwNone_GeomW3W3_F_sse4_1_double
1592                     (t_nblist                    * gmx_restrict       nlist,
1593                      rvec                        * gmx_restrict          xx,
1594                      rvec                        * gmx_restrict          ff,
1595                      t_forcerec                  * gmx_restrict          fr,
1596                      t_mdatoms                   * gmx_restrict     mdatoms,
1597                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1598                      t_nrnb                      * gmx_restrict        nrnb)
1599 {
1600     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1601      * just 0 for non-waters.
1602      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
1603      * jnr indices corresponding to data put in the four positions in the SIMD register.
1604      */
1605     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1606     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1607     int              jnrA,jnrB;
1608     int              j_coord_offsetA,j_coord_offsetB;
1609     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1610     real             rcutoff_scalar;
1611     real             *shiftvec,*fshift,*x,*f;
1612     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1613     int              vdwioffset0;
1614     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1615     int              vdwioffset1;
1616     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1617     int              vdwioffset2;
1618     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1619     int              vdwjidx0A,vdwjidx0B;
1620     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1621     int              vdwjidx1A,vdwjidx1B;
1622     __m128d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1623     int              vdwjidx2A,vdwjidx2B;
1624     __m128d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1625     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1626     __m128d          dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1627     __m128d          dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1628     __m128d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1629     __m128d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1630     __m128d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1631     __m128d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1632     __m128d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1633     __m128d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1634     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
1635     real             *charge;
1636     __m128i          ewitab;
1637     __m128d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1638     real             *ewtab;
1639     __m128d          rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
1640     real             rswitch_scalar,d_scalar;
1641     __m128d          dummy_mask,cutoff_mask;
1642     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
1643     __m128d          one     = _mm_set1_pd(1.0);
1644     __m128d          two     = _mm_set1_pd(2.0);
1645     x                = xx[0];
1646     f                = ff[0];
1647
1648     nri              = nlist->nri;
1649     iinr             = nlist->iinr;
1650     jindex           = nlist->jindex;
1651     jjnr             = nlist->jjnr;
1652     shiftidx         = nlist->shift;
1653     gid              = nlist->gid;
1654     shiftvec         = fr->shift_vec[0];
1655     fshift           = fr->fshift[0];
1656     facel            = _mm_set1_pd(fr->epsfac);
1657     charge           = mdatoms->chargeA;
1658
1659     sh_ewald         = _mm_set1_pd(fr->ic->sh_ewald);
1660     ewtab            = fr->ic->tabq_coul_FDV0;
1661     ewtabscale       = _mm_set1_pd(fr->ic->tabq_scale);
1662     ewtabhalfspace   = _mm_set1_pd(0.5/fr->ic->tabq_scale);
1663
1664     /* Setup water-specific parameters */
1665     inr              = nlist->iinr[0];
1666     iq0              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
1667     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
1668     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
1669
1670     jq0              = _mm_set1_pd(charge[inr+0]);
1671     jq1              = _mm_set1_pd(charge[inr+1]);
1672     jq2              = _mm_set1_pd(charge[inr+2]);
1673     qq00             = _mm_mul_pd(iq0,jq0);
1674     qq01             = _mm_mul_pd(iq0,jq1);
1675     qq02             = _mm_mul_pd(iq0,jq2);
1676     qq10             = _mm_mul_pd(iq1,jq0);
1677     qq11             = _mm_mul_pd(iq1,jq1);
1678     qq12             = _mm_mul_pd(iq1,jq2);
1679     qq20             = _mm_mul_pd(iq2,jq0);
1680     qq21             = _mm_mul_pd(iq2,jq1);
1681     qq22             = _mm_mul_pd(iq2,jq2);
1682
1683     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
1684     rcutoff_scalar   = fr->rcoulomb;
1685     rcutoff          = _mm_set1_pd(rcutoff_scalar);
1686     rcutoff2         = _mm_mul_pd(rcutoff,rcutoff);
1687
1688     rswitch_scalar   = fr->rcoulomb_switch;
1689     rswitch          = _mm_set1_pd(rswitch_scalar);
1690     /* Setup switch parameters */
1691     d_scalar         = rcutoff_scalar-rswitch_scalar;
1692     d                = _mm_set1_pd(d_scalar);
1693     swV3             = _mm_set1_pd(-10.0/(d_scalar*d_scalar*d_scalar));
1694     swV4             = _mm_set1_pd( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
1695     swV5             = _mm_set1_pd( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
1696     swF2             = _mm_set1_pd(-30.0/(d_scalar*d_scalar*d_scalar));
1697     swF3             = _mm_set1_pd( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
1698     swF4             = _mm_set1_pd(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
1699
1700     /* Avoid stupid compiler warnings */
1701     jnrA = jnrB = 0;
1702     j_coord_offsetA = 0;
1703     j_coord_offsetB = 0;
1704
1705     outeriter        = 0;
1706     inneriter        = 0;
1707
1708     /* Start outer loop over neighborlists */
1709     for(iidx=0; iidx<nri; iidx++)
1710     {
1711         /* Load shift vector for this list */
1712         i_shift_offset   = DIM*shiftidx[iidx];
1713
1714         /* Load limits for loop over neighbors */
1715         j_index_start    = jindex[iidx];
1716         j_index_end      = jindex[iidx+1];
1717
1718         /* Get outer coordinate index */
1719         inr              = iinr[iidx];
1720         i_coord_offset   = DIM*inr;
1721
1722         /* Load i particle coords and add shift vector */
1723         gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1724                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1725
1726         fix0             = _mm_setzero_pd();
1727         fiy0             = _mm_setzero_pd();
1728         fiz0             = _mm_setzero_pd();
1729         fix1             = _mm_setzero_pd();
1730         fiy1             = _mm_setzero_pd();
1731         fiz1             = _mm_setzero_pd();
1732         fix2             = _mm_setzero_pd();
1733         fiy2             = _mm_setzero_pd();
1734         fiz2             = _mm_setzero_pd();
1735
1736         /* Start inner kernel loop */
1737         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
1738         {
1739
1740             /* Get j neighbor index, and coordinate index */
1741             jnrA             = jjnr[jidx];
1742             jnrB             = jjnr[jidx+1];
1743             j_coord_offsetA  = DIM*jnrA;
1744             j_coord_offsetB  = DIM*jnrB;
1745
1746             /* load j atom coordinates */
1747             gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1748                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1749
1750             /* Calculate displacement vector */
1751             dx00             = _mm_sub_pd(ix0,jx0);
1752             dy00             = _mm_sub_pd(iy0,jy0);
1753             dz00             = _mm_sub_pd(iz0,jz0);
1754             dx01             = _mm_sub_pd(ix0,jx1);
1755             dy01             = _mm_sub_pd(iy0,jy1);
1756             dz01             = _mm_sub_pd(iz0,jz1);
1757             dx02             = _mm_sub_pd(ix0,jx2);
1758             dy02             = _mm_sub_pd(iy0,jy2);
1759             dz02             = _mm_sub_pd(iz0,jz2);
1760             dx10             = _mm_sub_pd(ix1,jx0);
1761             dy10             = _mm_sub_pd(iy1,jy0);
1762             dz10             = _mm_sub_pd(iz1,jz0);
1763             dx11             = _mm_sub_pd(ix1,jx1);
1764             dy11             = _mm_sub_pd(iy1,jy1);
1765             dz11             = _mm_sub_pd(iz1,jz1);
1766             dx12             = _mm_sub_pd(ix1,jx2);
1767             dy12             = _mm_sub_pd(iy1,jy2);
1768             dz12             = _mm_sub_pd(iz1,jz2);
1769             dx20             = _mm_sub_pd(ix2,jx0);
1770             dy20             = _mm_sub_pd(iy2,jy0);
1771             dz20             = _mm_sub_pd(iz2,jz0);
1772             dx21             = _mm_sub_pd(ix2,jx1);
1773             dy21             = _mm_sub_pd(iy2,jy1);
1774             dz21             = _mm_sub_pd(iz2,jz1);
1775             dx22             = _mm_sub_pd(ix2,jx2);
1776             dy22             = _mm_sub_pd(iy2,jy2);
1777             dz22             = _mm_sub_pd(iz2,jz2);
1778
1779             /* Calculate squared distance and things based on it */
1780             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1781             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
1782             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
1783             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
1784             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1785             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1786             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
1787             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1788             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1789
1790             rinv00           = gmx_mm_invsqrt_pd(rsq00);
1791             rinv01           = gmx_mm_invsqrt_pd(rsq01);
1792             rinv02           = gmx_mm_invsqrt_pd(rsq02);
1793             rinv10           = gmx_mm_invsqrt_pd(rsq10);
1794             rinv11           = gmx_mm_invsqrt_pd(rsq11);
1795             rinv12           = gmx_mm_invsqrt_pd(rsq12);
1796             rinv20           = gmx_mm_invsqrt_pd(rsq20);
1797             rinv21           = gmx_mm_invsqrt_pd(rsq21);
1798             rinv22           = gmx_mm_invsqrt_pd(rsq22);
1799
1800             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
1801             rinvsq01         = _mm_mul_pd(rinv01,rinv01);
1802             rinvsq02         = _mm_mul_pd(rinv02,rinv02);
1803             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
1804             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
1805             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
1806             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
1807             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
1808             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
1809
1810             fjx0             = _mm_setzero_pd();
1811             fjy0             = _mm_setzero_pd();
1812             fjz0             = _mm_setzero_pd();
1813             fjx1             = _mm_setzero_pd();
1814             fjy1             = _mm_setzero_pd();
1815             fjz1             = _mm_setzero_pd();
1816             fjx2             = _mm_setzero_pd();
1817             fjy2             = _mm_setzero_pd();
1818             fjz2             = _mm_setzero_pd();
1819
1820             /**************************
1821              * CALCULATE INTERACTIONS *
1822              **************************/
1823
1824             if (gmx_mm_any_lt(rsq00,rcutoff2))
1825             {
1826
1827             r00              = _mm_mul_pd(rsq00,rinv00);
1828
1829             /* EWALD ELECTROSTATICS */
1830
1831             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1832             ewrt             = _mm_mul_pd(r00,ewtabscale);
1833             ewitab           = _mm_cvttpd_epi32(ewrt);
1834             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1835             ewitab           = _mm_slli_epi32(ewitab,2);
1836             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1837             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1838             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1839             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1840             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
1841             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1842             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1843             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1844             velec            = _mm_mul_pd(qq00,_mm_sub_pd(rinv00,velec));
1845             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
1846
1847             d                = _mm_sub_pd(r00,rswitch);
1848             d                = _mm_max_pd(d,_mm_setzero_pd());
1849             d2               = _mm_mul_pd(d,d);
1850             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_add_pd(swV3,_mm_mul_pd(d,_mm_add_pd(swV4,_mm_mul_pd(d,swV5)))))));
1851
1852             dsw              = _mm_mul_pd(d2,_mm_add_pd(swF2,_mm_mul_pd(d,_mm_add_pd(swF3,_mm_mul_pd(d,swF4)))));
1853
1854             /* Evaluate switch function */
1855             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1856             felec            = _mm_sub_pd( _mm_mul_pd(felec,sw) , _mm_mul_pd(rinv00,_mm_mul_pd(velec,dsw)) );
1857             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
1858
1859             fscal            = felec;
1860
1861             fscal            = _mm_and_pd(fscal,cutoff_mask);
1862
1863             /* Calculate temporary vectorial force */
1864             tx               = _mm_mul_pd(fscal,dx00);
1865             ty               = _mm_mul_pd(fscal,dy00);
1866             tz               = _mm_mul_pd(fscal,dz00);
1867
1868             /* Update vectorial force */
1869             fix0             = _mm_add_pd(fix0,tx);
1870             fiy0             = _mm_add_pd(fiy0,ty);
1871             fiz0             = _mm_add_pd(fiz0,tz);
1872
1873             fjx0             = _mm_add_pd(fjx0,tx);
1874             fjy0             = _mm_add_pd(fjy0,ty);
1875             fjz0             = _mm_add_pd(fjz0,tz);
1876
1877             }
1878
1879             /**************************
1880              * CALCULATE INTERACTIONS *
1881              **************************/
1882
1883             if (gmx_mm_any_lt(rsq01,rcutoff2))
1884             {
1885
1886             r01              = _mm_mul_pd(rsq01,rinv01);
1887
1888             /* EWALD ELECTROSTATICS */
1889
1890             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1891             ewrt             = _mm_mul_pd(r01,ewtabscale);
1892             ewitab           = _mm_cvttpd_epi32(ewrt);
1893             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1894             ewitab           = _mm_slli_epi32(ewitab,2);
1895             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1896             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1897             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1898             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1899             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
1900             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1901             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1902             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1903             velec            = _mm_mul_pd(qq01,_mm_sub_pd(rinv01,velec));
1904             felec            = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
1905
1906             d                = _mm_sub_pd(r01,rswitch);
1907             d                = _mm_max_pd(d,_mm_setzero_pd());
1908             d2               = _mm_mul_pd(d,d);
1909             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_add_pd(swV3,_mm_mul_pd(d,_mm_add_pd(swV4,_mm_mul_pd(d,swV5)))))));
1910
1911             dsw              = _mm_mul_pd(d2,_mm_add_pd(swF2,_mm_mul_pd(d,_mm_add_pd(swF3,_mm_mul_pd(d,swF4)))));
1912
1913             /* Evaluate switch function */
1914             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1915             felec            = _mm_sub_pd( _mm_mul_pd(felec,sw) , _mm_mul_pd(rinv01,_mm_mul_pd(velec,dsw)) );
1916             cutoff_mask      = _mm_cmplt_pd(rsq01,rcutoff2);
1917
1918             fscal            = felec;
1919
1920             fscal            = _mm_and_pd(fscal,cutoff_mask);
1921
1922             /* Calculate temporary vectorial force */
1923             tx               = _mm_mul_pd(fscal,dx01);
1924             ty               = _mm_mul_pd(fscal,dy01);
1925             tz               = _mm_mul_pd(fscal,dz01);
1926
1927             /* Update vectorial force */
1928             fix0             = _mm_add_pd(fix0,tx);
1929             fiy0             = _mm_add_pd(fiy0,ty);
1930             fiz0             = _mm_add_pd(fiz0,tz);
1931
1932             fjx1             = _mm_add_pd(fjx1,tx);
1933             fjy1             = _mm_add_pd(fjy1,ty);
1934             fjz1             = _mm_add_pd(fjz1,tz);
1935
1936             }
1937
1938             /**************************
1939              * CALCULATE INTERACTIONS *
1940              **************************/
1941
1942             if (gmx_mm_any_lt(rsq02,rcutoff2))
1943             {
1944
1945             r02              = _mm_mul_pd(rsq02,rinv02);
1946
1947             /* EWALD ELECTROSTATICS */
1948
1949             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1950             ewrt             = _mm_mul_pd(r02,ewtabscale);
1951             ewitab           = _mm_cvttpd_epi32(ewrt);
1952             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1953             ewitab           = _mm_slli_epi32(ewitab,2);
1954             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1955             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1956             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1957             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1958             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
1959             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1960             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1961             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1962             velec            = _mm_mul_pd(qq02,_mm_sub_pd(rinv02,velec));
1963             felec            = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
1964
1965             d                = _mm_sub_pd(r02,rswitch);
1966             d                = _mm_max_pd(d,_mm_setzero_pd());
1967             d2               = _mm_mul_pd(d,d);
1968             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_add_pd(swV3,_mm_mul_pd(d,_mm_add_pd(swV4,_mm_mul_pd(d,swV5)))))));
1969
1970             dsw              = _mm_mul_pd(d2,_mm_add_pd(swF2,_mm_mul_pd(d,_mm_add_pd(swF3,_mm_mul_pd(d,swF4)))));
1971
1972             /* Evaluate switch function */
1973             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1974             felec            = _mm_sub_pd( _mm_mul_pd(felec,sw) , _mm_mul_pd(rinv02,_mm_mul_pd(velec,dsw)) );
1975             cutoff_mask      = _mm_cmplt_pd(rsq02,rcutoff2);
1976
1977             fscal            = felec;
1978
1979             fscal            = _mm_and_pd(fscal,cutoff_mask);
1980
1981             /* Calculate temporary vectorial force */
1982             tx               = _mm_mul_pd(fscal,dx02);
1983             ty               = _mm_mul_pd(fscal,dy02);
1984             tz               = _mm_mul_pd(fscal,dz02);
1985
1986             /* Update vectorial force */
1987             fix0             = _mm_add_pd(fix0,tx);
1988             fiy0             = _mm_add_pd(fiy0,ty);
1989             fiz0             = _mm_add_pd(fiz0,tz);
1990
1991             fjx2             = _mm_add_pd(fjx2,tx);
1992             fjy2             = _mm_add_pd(fjy2,ty);
1993             fjz2             = _mm_add_pd(fjz2,tz);
1994
1995             }
1996
1997             /**************************
1998              * CALCULATE INTERACTIONS *
1999              **************************/
2000
2001             if (gmx_mm_any_lt(rsq10,rcutoff2))
2002             {
2003
2004             r10              = _mm_mul_pd(rsq10,rinv10);
2005
2006             /* EWALD ELECTROSTATICS */
2007
2008             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2009             ewrt             = _mm_mul_pd(r10,ewtabscale);
2010             ewitab           = _mm_cvttpd_epi32(ewrt);
2011             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2012             ewitab           = _mm_slli_epi32(ewitab,2);
2013             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
2014             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
2015             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
2016             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
2017             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
2018             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
2019             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
2020             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
2021             velec            = _mm_mul_pd(qq10,_mm_sub_pd(rinv10,velec));
2022             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
2023
2024             d                = _mm_sub_pd(r10,rswitch);
2025             d                = _mm_max_pd(d,_mm_setzero_pd());
2026             d2               = _mm_mul_pd(d,d);
2027             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_add_pd(swV3,_mm_mul_pd(d,_mm_add_pd(swV4,_mm_mul_pd(d,swV5)))))));
2028
2029             dsw              = _mm_mul_pd(d2,_mm_add_pd(swF2,_mm_mul_pd(d,_mm_add_pd(swF3,_mm_mul_pd(d,swF4)))));
2030
2031             /* Evaluate switch function */
2032             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2033             felec            = _mm_sub_pd( _mm_mul_pd(felec,sw) , _mm_mul_pd(rinv10,_mm_mul_pd(velec,dsw)) );
2034             cutoff_mask      = _mm_cmplt_pd(rsq10,rcutoff2);
2035
2036             fscal            = felec;
2037
2038             fscal            = _mm_and_pd(fscal,cutoff_mask);
2039
2040             /* Calculate temporary vectorial force */
2041             tx               = _mm_mul_pd(fscal,dx10);
2042             ty               = _mm_mul_pd(fscal,dy10);
2043             tz               = _mm_mul_pd(fscal,dz10);
2044
2045             /* Update vectorial force */
2046             fix1             = _mm_add_pd(fix1,tx);
2047             fiy1             = _mm_add_pd(fiy1,ty);
2048             fiz1             = _mm_add_pd(fiz1,tz);
2049
2050             fjx0             = _mm_add_pd(fjx0,tx);
2051             fjy0             = _mm_add_pd(fjy0,ty);
2052             fjz0             = _mm_add_pd(fjz0,tz);
2053
2054             }
2055
2056             /**************************
2057              * CALCULATE INTERACTIONS *
2058              **************************/
2059
2060             if (gmx_mm_any_lt(rsq11,rcutoff2))
2061             {
2062
2063             r11              = _mm_mul_pd(rsq11,rinv11);
2064
2065             /* EWALD ELECTROSTATICS */
2066
2067             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2068             ewrt             = _mm_mul_pd(r11,ewtabscale);
2069             ewitab           = _mm_cvttpd_epi32(ewrt);
2070             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2071             ewitab           = _mm_slli_epi32(ewitab,2);
2072             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
2073             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
2074             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
2075             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
2076             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
2077             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
2078             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
2079             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
2080             velec            = _mm_mul_pd(qq11,_mm_sub_pd(rinv11,velec));
2081             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
2082
2083             d                = _mm_sub_pd(r11,rswitch);
2084             d                = _mm_max_pd(d,_mm_setzero_pd());
2085             d2               = _mm_mul_pd(d,d);
2086             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_add_pd(swV3,_mm_mul_pd(d,_mm_add_pd(swV4,_mm_mul_pd(d,swV5)))))));
2087
2088             dsw              = _mm_mul_pd(d2,_mm_add_pd(swF2,_mm_mul_pd(d,_mm_add_pd(swF3,_mm_mul_pd(d,swF4)))));
2089
2090             /* Evaluate switch function */
2091             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2092             felec            = _mm_sub_pd( _mm_mul_pd(felec,sw) , _mm_mul_pd(rinv11,_mm_mul_pd(velec,dsw)) );
2093             cutoff_mask      = _mm_cmplt_pd(rsq11,rcutoff2);
2094
2095             fscal            = felec;
2096
2097             fscal            = _mm_and_pd(fscal,cutoff_mask);
2098
2099             /* Calculate temporary vectorial force */
2100             tx               = _mm_mul_pd(fscal,dx11);
2101             ty               = _mm_mul_pd(fscal,dy11);
2102             tz               = _mm_mul_pd(fscal,dz11);
2103
2104             /* Update vectorial force */
2105             fix1             = _mm_add_pd(fix1,tx);
2106             fiy1             = _mm_add_pd(fiy1,ty);
2107             fiz1             = _mm_add_pd(fiz1,tz);
2108
2109             fjx1             = _mm_add_pd(fjx1,tx);
2110             fjy1             = _mm_add_pd(fjy1,ty);
2111             fjz1             = _mm_add_pd(fjz1,tz);
2112
2113             }
2114
2115             /**************************
2116              * CALCULATE INTERACTIONS *
2117              **************************/
2118
2119             if (gmx_mm_any_lt(rsq12,rcutoff2))
2120             {
2121
2122             r12              = _mm_mul_pd(rsq12,rinv12);
2123
2124             /* EWALD ELECTROSTATICS */
2125
2126             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2127             ewrt             = _mm_mul_pd(r12,ewtabscale);
2128             ewitab           = _mm_cvttpd_epi32(ewrt);
2129             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2130             ewitab           = _mm_slli_epi32(ewitab,2);
2131             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
2132             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
2133             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
2134             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
2135             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
2136             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
2137             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
2138             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
2139             velec            = _mm_mul_pd(qq12,_mm_sub_pd(rinv12,velec));
2140             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
2141
2142             d                = _mm_sub_pd(r12,rswitch);
2143             d                = _mm_max_pd(d,_mm_setzero_pd());
2144             d2               = _mm_mul_pd(d,d);
2145             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_add_pd(swV3,_mm_mul_pd(d,_mm_add_pd(swV4,_mm_mul_pd(d,swV5)))))));
2146
2147             dsw              = _mm_mul_pd(d2,_mm_add_pd(swF2,_mm_mul_pd(d,_mm_add_pd(swF3,_mm_mul_pd(d,swF4)))));
2148
2149             /* Evaluate switch function */
2150             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2151             felec            = _mm_sub_pd( _mm_mul_pd(felec,sw) , _mm_mul_pd(rinv12,_mm_mul_pd(velec,dsw)) );
2152             cutoff_mask      = _mm_cmplt_pd(rsq12,rcutoff2);
2153
2154             fscal            = felec;
2155
2156             fscal            = _mm_and_pd(fscal,cutoff_mask);
2157
2158             /* Calculate temporary vectorial force */
2159             tx               = _mm_mul_pd(fscal,dx12);
2160             ty               = _mm_mul_pd(fscal,dy12);
2161             tz               = _mm_mul_pd(fscal,dz12);
2162
2163             /* Update vectorial force */
2164             fix1             = _mm_add_pd(fix1,tx);
2165             fiy1             = _mm_add_pd(fiy1,ty);
2166             fiz1             = _mm_add_pd(fiz1,tz);
2167
2168             fjx2             = _mm_add_pd(fjx2,tx);
2169             fjy2             = _mm_add_pd(fjy2,ty);
2170             fjz2             = _mm_add_pd(fjz2,tz);
2171
2172             }
2173
2174             /**************************
2175              * CALCULATE INTERACTIONS *
2176              **************************/
2177
2178             if (gmx_mm_any_lt(rsq20,rcutoff2))
2179             {
2180
2181             r20              = _mm_mul_pd(rsq20,rinv20);
2182
2183             /* EWALD ELECTROSTATICS */
2184
2185             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2186             ewrt             = _mm_mul_pd(r20,ewtabscale);
2187             ewitab           = _mm_cvttpd_epi32(ewrt);
2188             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2189             ewitab           = _mm_slli_epi32(ewitab,2);
2190             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
2191             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
2192             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
2193             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
2194             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
2195             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
2196             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
2197             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
2198             velec            = _mm_mul_pd(qq20,_mm_sub_pd(rinv20,velec));
2199             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
2200
2201             d                = _mm_sub_pd(r20,rswitch);
2202             d                = _mm_max_pd(d,_mm_setzero_pd());
2203             d2               = _mm_mul_pd(d,d);
2204             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_add_pd(swV3,_mm_mul_pd(d,_mm_add_pd(swV4,_mm_mul_pd(d,swV5)))))));
2205
2206             dsw              = _mm_mul_pd(d2,_mm_add_pd(swF2,_mm_mul_pd(d,_mm_add_pd(swF3,_mm_mul_pd(d,swF4)))));
2207
2208             /* Evaluate switch function */
2209             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2210             felec            = _mm_sub_pd( _mm_mul_pd(felec,sw) , _mm_mul_pd(rinv20,_mm_mul_pd(velec,dsw)) );
2211             cutoff_mask      = _mm_cmplt_pd(rsq20,rcutoff2);
2212
2213             fscal            = felec;
2214
2215             fscal            = _mm_and_pd(fscal,cutoff_mask);
2216
2217             /* Calculate temporary vectorial force */
2218             tx               = _mm_mul_pd(fscal,dx20);
2219             ty               = _mm_mul_pd(fscal,dy20);
2220             tz               = _mm_mul_pd(fscal,dz20);
2221
2222             /* Update vectorial force */
2223             fix2             = _mm_add_pd(fix2,tx);
2224             fiy2             = _mm_add_pd(fiy2,ty);
2225             fiz2             = _mm_add_pd(fiz2,tz);
2226
2227             fjx0             = _mm_add_pd(fjx0,tx);
2228             fjy0             = _mm_add_pd(fjy0,ty);
2229             fjz0             = _mm_add_pd(fjz0,tz);
2230
2231             }
2232
2233             /**************************
2234              * CALCULATE INTERACTIONS *
2235              **************************/
2236
2237             if (gmx_mm_any_lt(rsq21,rcutoff2))
2238             {
2239
2240             r21              = _mm_mul_pd(rsq21,rinv21);
2241
2242             /* EWALD ELECTROSTATICS */
2243
2244             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2245             ewrt             = _mm_mul_pd(r21,ewtabscale);
2246             ewitab           = _mm_cvttpd_epi32(ewrt);
2247             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2248             ewitab           = _mm_slli_epi32(ewitab,2);
2249             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
2250             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
2251             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
2252             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
2253             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
2254             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
2255             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
2256             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
2257             velec            = _mm_mul_pd(qq21,_mm_sub_pd(rinv21,velec));
2258             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
2259
2260             d                = _mm_sub_pd(r21,rswitch);
2261             d                = _mm_max_pd(d,_mm_setzero_pd());
2262             d2               = _mm_mul_pd(d,d);
2263             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_add_pd(swV3,_mm_mul_pd(d,_mm_add_pd(swV4,_mm_mul_pd(d,swV5)))))));
2264
2265             dsw              = _mm_mul_pd(d2,_mm_add_pd(swF2,_mm_mul_pd(d,_mm_add_pd(swF3,_mm_mul_pd(d,swF4)))));
2266
2267             /* Evaluate switch function */
2268             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2269             felec            = _mm_sub_pd( _mm_mul_pd(felec,sw) , _mm_mul_pd(rinv21,_mm_mul_pd(velec,dsw)) );
2270             cutoff_mask      = _mm_cmplt_pd(rsq21,rcutoff2);
2271
2272             fscal            = felec;
2273
2274             fscal            = _mm_and_pd(fscal,cutoff_mask);
2275
2276             /* Calculate temporary vectorial force */
2277             tx               = _mm_mul_pd(fscal,dx21);
2278             ty               = _mm_mul_pd(fscal,dy21);
2279             tz               = _mm_mul_pd(fscal,dz21);
2280
2281             /* Update vectorial force */
2282             fix2             = _mm_add_pd(fix2,tx);
2283             fiy2             = _mm_add_pd(fiy2,ty);
2284             fiz2             = _mm_add_pd(fiz2,tz);
2285
2286             fjx1             = _mm_add_pd(fjx1,tx);
2287             fjy1             = _mm_add_pd(fjy1,ty);
2288             fjz1             = _mm_add_pd(fjz1,tz);
2289
2290             }
2291
2292             /**************************
2293              * CALCULATE INTERACTIONS *
2294              **************************/
2295
2296             if (gmx_mm_any_lt(rsq22,rcutoff2))
2297             {
2298
2299             r22              = _mm_mul_pd(rsq22,rinv22);
2300
2301             /* EWALD ELECTROSTATICS */
2302
2303             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2304             ewrt             = _mm_mul_pd(r22,ewtabscale);
2305             ewitab           = _mm_cvttpd_epi32(ewrt);
2306             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2307             ewitab           = _mm_slli_epi32(ewitab,2);
2308             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
2309             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
2310             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
2311             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
2312             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
2313             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
2314             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
2315             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
2316             velec            = _mm_mul_pd(qq22,_mm_sub_pd(rinv22,velec));
2317             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
2318
2319             d                = _mm_sub_pd(r22,rswitch);
2320             d                = _mm_max_pd(d,_mm_setzero_pd());
2321             d2               = _mm_mul_pd(d,d);
2322             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_add_pd(swV3,_mm_mul_pd(d,_mm_add_pd(swV4,_mm_mul_pd(d,swV5)))))));
2323
2324             dsw              = _mm_mul_pd(d2,_mm_add_pd(swF2,_mm_mul_pd(d,_mm_add_pd(swF3,_mm_mul_pd(d,swF4)))));
2325
2326             /* Evaluate switch function */
2327             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2328             felec            = _mm_sub_pd( _mm_mul_pd(felec,sw) , _mm_mul_pd(rinv22,_mm_mul_pd(velec,dsw)) );
2329             cutoff_mask      = _mm_cmplt_pd(rsq22,rcutoff2);
2330
2331             fscal            = felec;
2332
2333             fscal            = _mm_and_pd(fscal,cutoff_mask);
2334
2335             /* Calculate temporary vectorial force */
2336             tx               = _mm_mul_pd(fscal,dx22);
2337             ty               = _mm_mul_pd(fscal,dy22);
2338             tz               = _mm_mul_pd(fscal,dz22);
2339
2340             /* Update vectorial force */
2341             fix2             = _mm_add_pd(fix2,tx);
2342             fiy2             = _mm_add_pd(fiy2,ty);
2343             fiz2             = _mm_add_pd(fiz2,tz);
2344
2345             fjx2             = _mm_add_pd(fjx2,tx);
2346             fjy2             = _mm_add_pd(fjy2,ty);
2347             fjz2             = _mm_add_pd(fjz2,tz);
2348
2349             }
2350
2351             gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
2352
2353             /* Inner loop uses 558 flops */
2354         }
2355
2356         if(jidx<j_index_end)
2357         {
2358
2359             jnrA             = jjnr[jidx];
2360             j_coord_offsetA  = DIM*jnrA;
2361
2362             /* load j atom coordinates */
2363             gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
2364                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
2365
2366             /* Calculate displacement vector */
2367             dx00             = _mm_sub_pd(ix0,jx0);
2368             dy00             = _mm_sub_pd(iy0,jy0);
2369             dz00             = _mm_sub_pd(iz0,jz0);
2370             dx01             = _mm_sub_pd(ix0,jx1);
2371             dy01             = _mm_sub_pd(iy0,jy1);
2372             dz01             = _mm_sub_pd(iz0,jz1);
2373             dx02             = _mm_sub_pd(ix0,jx2);
2374             dy02             = _mm_sub_pd(iy0,jy2);
2375             dz02             = _mm_sub_pd(iz0,jz2);
2376             dx10             = _mm_sub_pd(ix1,jx0);
2377             dy10             = _mm_sub_pd(iy1,jy0);
2378             dz10             = _mm_sub_pd(iz1,jz0);
2379             dx11             = _mm_sub_pd(ix1,jx1);
2380             dy11             = _mm_sub_pd(iy1,jy1);
2381             dz11             = _mm_sub_pd(iz1,jz1);
2382             dx12             = _mm_sub_pd(ix1,jx2);
2383             dy12             = _mm_sub_pd(iy1,jy2);
2384             dz12             = _mm_sub_pd(iz1,jz2);
2385             dx20             = _mm_sub_pd(ix2,jx0);
2386             dy20             = _mm_sub_pd(iy2,jy0);
2387             dz20             = _mm_sub_pd(iz2,jz0);
2388             dx21             = _mm_sub_pd(ix2,jx1);
2389             dy21             = _mm_sub_pd(iy2,jy1);
2390             dz21             = _mm_sub_pd(iz2,jz1);
2391             dx22             = _mm_sub_pd(ix2,jx2);
2392             dy22             = _mm_sub_pd(iy2,jy2);
2393             dz22             = _mm_sub_pd(iz2,jz2);
2394
2395             /* Calculate squared distance and things based on it */
2396             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
2397             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
2398             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
2399             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
2400             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
2401             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
2402             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
2403             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
2404             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
2405
2406             rinv00           = gmx_mm_invsqrt_pd(rsq00);
2407             rinv01           = gmx_mm_invsqrt_pd(rsq01);
2408             rinv02           = gmx_mm_invsqrt_pd(rsq02);
2409             rinv10           = gmx_mm_invsqrt_pd(rsq10);
2410             rinv11           = gmx_mm_invsqrt_pd(rsq11);
2411             rinv12           = gmx_mm_invsqrt_pd(rsq12);
2412             rinv20           = gmx_mm_invsqrt_pd(rsq20);
2413             rinv21           = gmx_mm_invsqrt_pd(rsq21);
2414             rinv22           = gmx_mm_invsqrt_pd(rsq22);
2415
2416             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
2417             rinvsq01         = _mm_mul_pd(rinv01,rinv01);
2418             rinvsq02         = _mm_mul_pd(rinv02,rinv02);
2419             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
2420             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
2421             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
2422             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
2423             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
2424             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
2425
2426             fjx0             = _mm_setzero_pd();
2427             fjy0             = _mm_setzero_pd();
2428             fjz0             = _mm_setzero_pd();
2429             fjx1             = _mm_setzero_pd();
2430             fjy1             = _mm_setzero_pd();
2431             fjz1             = _mm_setzero_pd();
2432             fjx2             = _mm_setzero_pd();
2433             fjy2             = _mm_setzero_pd();
2434             fjz2             = _mm_setzero_pd();
2435
2436             /**************************
2437              * CALCULATE INTERACTIONS *
2438              **************************/
2439
2440             if (gmx_mm_any_lt(rsq00,rcutoff2))
2441             {
2442
2443             r00              = _mm_mul_pd(rsq00,rinv00);
2444
2445             /* EWALD ELECTROSTATICS */
2446
2447             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2448             ewrt             = _mm_mul_pd(r00,ewtabscale);
2449             ewitab           = _mm_cvttpd_epi32(ewrt);
2450             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2451             ewitab           = _mm_slli_epi32(ewitab,2);
2452             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
2453             ewtabD           = _mm_setzero_pd();
2454             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
2455             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
2456             ewtabFn          = _mm_setzero_pd();
2457             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
2458             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
2459             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
2460             velec            = _mm_mul_pd(qq00,_mm_sub_pd(rinv00,velec));
2461             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
2462
2463             d                = _mm_sub_pd(r00,rswitch);
2464             d                = _mm_max_pd(d,_mm_setzero_pd());
2465             d2               = _mm_mul_pd(d,d);
2466             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_add_pd(swV3,_mm_mul_pd(d,_mm_add_pd(swV4,_mm_mul_pd(d,swV5)))))));
2467
2468             dsw              = _mm_mul_pd(d2,_mm_add_pd(swF2,_mm_mul_pd(d,_mm_add_pd(swF3,_mm_mul_pd(d,swF4)))));
2469
2470             /* Evaluate switch function */
2471             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2472             felec            = _mm_sub_pd( _mm_mul_pd(felec,sw) , _mm_mul_pd(rinv00,_mm_mul_pd(velec,dsw)) );
2473             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
2474
2475             fscal            = felec;
2476
2477             fscal            = _mm_and_pd(fscal,cutoff_mask);
2478
2479             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2480
2481             /* Calculate temporary vectorial force */
2482             tx               = _mm_mul_pd(fscal,dx00);
2483             ty               = _mm_mul_pd(fscal,dy00);
2484             tz               = _mm_mul_pd(fscal,dz00);
2485
2486             /* Update vectorial force */
2487             fix0             = _mm_add_pd(fix0,tx);
2488             fiy0             = _mm_add_pd(fiy0,ty);
2489             fiz0             = _mm_add_pd(fiz0,tz);
2490
2491             fjx0             = _mm_add_pd(fjx0,tx);
2492             fjy0             = _mm_add_pd(fjy0,ty);
2493             fjz0             = _mm_add_pd(fjz0,tz);
2494
2495             }
2496
2497             /**************************
2498              * CALCULATE INTERACTIONS *
2499              **************************/
2500
2501             if (gmx_mm_any_lt(rsq01,rcutoff2))
2502             {
2503
2504             r01              = _mm_mul_pd(rsq01,rinv01);
2505
2506             /* EWALD ELECTROSTATICS */
2507
2508             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2509             ewrt             = _mm_mul_pd(r01,ewtabscale);
2510             ewitab           = _mm_cvttpd_epi32(ewrt);
2511             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2512             ewitab           = _mm_slli_epi32(ewitab,2);
2513             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
2514             ewtabD           = _mm_setzero_pd();
2515             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
2516             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
2517             ewtabFn          = _mm_setzero_pd();
2518             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
2519             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
2520             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
2521             velec            = _mm_mul_pd(qq01,_mm_sub_pd(rinv01,velec));
2522             felec            = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
2523
2524             d                = _mm_sub_pd(r01,rswitch);
2525             d                = _mm_max_pd(d,_mm_setzero_pd());
2526             d2               = _mm_mul_pd(d,d);
2527             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_add_pd(swV3,_mm_mul_pd(d,_mm_add_pd(swV4,_mm_mul_pd(d,swV5)))))));
2528
2529             dsw              = _mm_mul_pd(d2,_mm_add_pd(swF2,_mm_mul_pd(d,_mm_add_pd(swF3,_mm_mul_pd(d,swF4)))));
2530
2531             /* Evaluate switch function */
2532             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2533             felec            = _mm_sub_pd( _mm_mul_pd(felec,sw) , _mm_mul_pd(rinv01,_mm_mul_pd(velec,dsw)) );
2534             cutoff_mask      = _mm_cmplt_pd(rsq01,rcutoff2);
2535
2536             fscal            = felec;
2537
2538             fscal            = _mm_and_pd(fscal,cutoff_mask);
2539
2540             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2541
2542             /* Calculate temporary vectorial force */
2543             tx               = _mm_mul_pd(fscal,dx01);
2544             ty               = _mm_mul_pd(fscal,dy01);
2545             tz               = _mm_mul_pd(fscal,dz01);
2546
2547             /* Update vectorial force */
2548             fix0             = _mm_add_pd(fix0,tx);
2549             fiy0             = _mm_add_pd(fiy0,ty);
2550             fiz0             = _mm_add_pd(fiz0,tz);
2551
2552             fjx1             = _mm_add_pd(fjx1,tx);
2553             fjy1             = _mm_add_pd(fjy1,ty);
2554             fjz1             = _mm_add_pd(fjz1,tz);
2555
2556             }
2557
2558             /**************************
2559              * CALCULATE INTERACTIONS *
2560              **************************/
2561
2562             if (gmx_mm_any_lt(rsq02,rcutoff2))
2563             {
2564
2565             r02              = _mm_mul_pd(rsq02,rinv02);
2566
2567             /* EWALD ELECTROSTATICS */
2568
2569             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2570             ewrt             = _mm_mul_pd(r02,ewtabscale);
2571             ewitab           = _mm_cvttpd_epi32(ewrt);
2572             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2573             ewitab           = _mm_slli_epi32(ewitab,2);
2574             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
2575             ewtabD           = _mm_setzero_pd();
2576             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
2577             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
2578             ewtabFn          = _mm_setzero_pd();
2579             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
2580             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
2581             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
2582             velec            = _mm_mul_pd(qq02,_mm_sub_pd(rinv02,velec));
2583             felec            = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
2584
2585             d                = _mm_sub_pd(r02,rswitch);
2586             d                = _mm_max_pd(d,_mm_setzero_pd());
2587             d2               = _mm_mul_pd(d,d);
2588             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_add_pd(swV3,_mm_mul_pd(d,_mm_add_pd(swV4,_mm_mul_pd(d,swV5)))))));
2589
2590             dsw              = _mm_mul_pd(d2,_mm_add_pd(swF2,_mm_mul_pd(d,_mm_add_pd(swF3,_mm_mul_pd(d,swF4)))));
2591
2592             /* Evaluate switch function */
2593             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2594             felec            = _mm_sub_pd( _mm_mul_pd(felec,sw) , _mm_mul_pd(rinv02,_mm_mul_pd(velec,dsw)) );
2595             cutoff_mask      = _mm_cmplt_pd(rsq02,rcutoff2);
2596
2597             fscal            = felec;
2598
2599             fscal            = _mm_and_pd(fscal,cutoff_mask);
2600
2601             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2602
2603             /* Calculate temporary vectorial force */
2604             tx               = _mm_mul_pd(fscal,dx02);
2605             ty               = _mm_mul_pd(fscal,dy02);
2606             tz               = _mm_mul_pd(fscal,dz02);
2607
2608             /* Update vectorial force */
2609             fix0             = _mm_add_pd(fix0,tx);
2610             fiy0             = _mm_add_pd(fiy0,ty);
2611             fiz0             = _mm_add_pd(fiz0,tz);
2612
2613             fjx2             = _mm_add_pd(fjx2,tx);
2614             fjy2             = _mm_add_pd(fjy2,ty);
2615             fjz2             = _mm_add_pd(fjz2,tz);
2616
2617             }
2618
2619             /**************************
2620              * CALCULATE INTERACTIONS *
2621              **************************/
2622
2623             if (gmx_mm_any_lt(rsq10,rcutoff2))
2624             {
2625
2626             r10              = _mm_mul_pd(rsq10,rinv10);
2627
2628             /* EWALD ELECTROSTATICS */
2629
2630             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2631             ewrt             = _mm_mul_pd(r10,ewtabscale);
2632             ewitab           = _mm_cvttpd_epi32(ewrt);
2633             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2634             ewitab           = _mm_slli_epi32(ewitab,2);
2635             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
2636             ewtabD           = _mm_setzero_pd();
2637             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
2638             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
2639             ewtabFn          = _mm_setzero_pd();
2640             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
2641             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
2642             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
2643             velec            = _mm_mul_pd(qq10,_mm_sub_pd(rinv10,velec));
2644             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
2645
2646             d                = _mm_sub_pd(r10,rswitch);
2647             d                = _mm_max_pd(d,_mm_setzero_pd());
2648             d2               = _mm_mul_pd(d,d);
2649             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_add_pd(swV3,_mm_mul_pd(d,_mm_add_pd(swV4,_mm_mul_pd(d,swV5)))))));
2650
2651             dsw              = _mm_mul_pd(d2,_mm_add_pd(swF2,_mm_mul_pd(d,_mm_add_pd(swF3,_mm_mul_pd(d,swF4)))));
2652
2653             /* Evaluate switch function */
2654             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2655             felec            = _mm_sub_pd( _mm_mul_pd(felec,sw) , _mm_mul_pd(rinv10,_mm_mul_pd(velec,dsw)) );
2656             cutoff_mask      = _mm_cmplt_pd(rsq10,rcutoff2);
2657
2658             fscal            = felec;
2659
2660             fscal            = _mm_and_pd(fscal,cutoff_mask);
2661
2662             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2663
2664             /* Calculate temporary vectorial force */
2665             tx               = _mm_mul_pd(fscal,dx10);
2666             ty               = _mm_mul_pd(fscal,dy10);
2667             tz               = _mm_mul_pd(fscal,dz10);
2668
2669             /* Update vectorial force */
2670             fix1             = _mm_add_pd(fix1,tx);
2671             fiy1             = _mm_add_pd(fiy1,ty);
2672             fiz1             = _mm_add_pd(fiz1,tz);
2673
2674             fjx0             = _mm_add_pd(fjx0,tx);
2675             fjy0             = _mm_add_pd(fjy0,ty);
2676             fjz0             = _mm_add_pd(fjz0,tz);
2677
2678             }
2679
2680             /**************************
2681              * CALCULATE INTERACTIONS *
2682              **************************/
2683
2684             if (gmx_mm_any_lt(rsq11,rcutoff2))
2685             {
2686
2687             r11              = _mm_mul_pd(rsq11,rinv11);
2688
2689             /* EWALD ELECTROSTATICS */
2690
2691             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2692             ewrt             = _mm_mul_pd(r11,ewtabscale);
2693             ewitab           = _mm_cvttpd_epi32(ewrt);
2694             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2695             ewitab           = _mm_slli_epi32(ewitab,2);
2696             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
2697             ewtabD           = _mm_setzero_pd();
2698             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
2699             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
2700             ewtabFn          = _mm_setzero_pd();
2701             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
2702             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
2703             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
2704             velec            = _mm_mul_pd(qq11,_mm_sub_pd(rinv11,velec));
2705             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
2706
2707             d                = _mm_sub_pd(r11,rswitch);
2708             d                = _mm_max_pd(d,_mm_setzero_pd());
2709             d2               = _mm_mul_pd(d,d);
2710             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_add_pd(swV3,_mm_mul_pd(d,_mm_add_pd(swV4,_mm_mul_pd(d,swV5)))))));
2711
2712             dsw              = _mm_mul_pd(d2,_mm_add_pd(swF2,_mm_mul_pd(d,_mm_add_pd(swF3,_mm_mul_pd(d,swF4)))));
2713
2714             /* Evaluate switch function */
2715             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2716             felec            = _mm_sub_pd( _mm_mul_pd(felec,sw) , _mm_mul_pd(rinv11,_mm_mul_pd(velec,dsw)) );
2717             cutoff_mask      = _mm_cmplt_pd(rsq11,rcutoff2);
2718
2719             fscal            = felec;
2720
2721             fscal            = _mm_and_pd(fscal,cutoff_mask);
2722
2723             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2724
2725             /* Calculate temporary vectorial force */
2726             tx               = _mm_mul_pd(fscal,dx11);
2727             ty               = _mm_mul_pd(fscal,dy11);
2728             tz               = _mm_mul_pd(fscal,dz11);
2729
2730             /* Update vectorial force */
2731             fix1             = _mm_add_pd(fix1,tx);
2732             fiy1             = _mm_add_pd(fiy1,ty);
2733             fiz1             = _mm_add_pd(fiz1,tz);
2734
2735             fjx1             = _mm_add_pd(fjx1,tx);
2736             fjy1             = _mm_add_pd(fjy1,ty);
2737             fjz1             = _mm_add_pd(fjz1,tz);
2738
2739             }
2740
2741             /**************************
2742              * CALCULATE INTERACTIONS *
2743              **************************/
2744
2745             if (gmx_mm_any_lt(rsq12,rcutoff2))
2746             {
2747
2748             r12              = _mm_mul_pd(rsq12,rinv12);
2749
2750             /* EWALD ELECTROSTATICS */
2751
2752             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2753             ewrt             = _mm_mul_pd(r12,ewtabscale);
2754             ewitab           = _mm_cvttpd_epi32(ewrt);
2755             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2756             ewitab           = _mm_slli_epi32(ewitab,2);
2757             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
2758             ewtabD           = _mm_setzero_pd();
2759             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
2760             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
2761             ewtabFn          = _mm_setzero_pd();
2762             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
2763             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
2764             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
2765             velec            = _mm_mul_pd(qq12,_mm_sub_pd(rinv12,velec));
2766             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
2767
2768             d                = _mm_sub_pd(r12,rswitch);
2769             d                = _mm_max_pd(d,_mm_setzero_pd());
2770             d2               = _mm_mul_pd(d,d);
2771             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_add_pd(swV3,_mm_mul_pd(d,_mm_add_pd(swV4,_mm_mul_pd(d,swV5)))))));
2772
2773             dsw              = _mm_mul_pd(d2,_mm_add_pd(swF2,_mm_mul_pd(d,_mm_add_pd(swF3,_mm_mul_pd(d,swF4)))));
2774
2775             /* Evaluate switch function */
2776             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2777             felec            = _mm_sub_pd( _mm_mul_pd(felec,sw) , _mm_mul_pd(rinv12,_mm_mul_pd(velec,dsw)) );
2778             cutoff_mask      = _mm_cmplt_pd(rsq12,rcutoff2);
2779
2780             fscal            = felec;
2781
2782             fscal            = _mm_and_pd(fscal,cutoff_mask);
2783
2784             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2785
2786             /* Calculate temporary vectorial force */
2787             tx               = _mm_mul_pd(fscal,dx12);
2788             ty               = _mm_mul_pd(fscal,dy12);
2789             tz               = _mm_mul_pd(fscal,dz12);
2790
2791             /* Update vectorial force */
2792             fix1             = _mm_add_pd(fix1,tx);
2793             fiy1             = _mm_add_pd(fiy1,ty);
2794             fiz1             = _mm_add_pd(fiz1,tz);
2795
2796             fjx2             = _mm_add_pd(fjx2,tx);
2797             fjy2             = _mm_add_pd(fjy2,ty);
2798             fjz2             = _mm_add_pd(fjz2,tz);
2799
2800             }
2801
2802             /**************************
2803              * CALCULATE INTERACTIONS *
2804              **************************/
2805
2806             if (gmx_mm_any_lt(rsq20,rcutoff2))
2807             {
2808
2809             r20              = _mm_mul_pd(rsq20,rinv20);
2810
2811             /* EWALD ELECTROSTATICS */
2812
2813             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2814             ewrt             = _mm_mul_pd(r20,ewtabscale);
2815             ewitab           = _mm_cvttpd_epi32(ewrt);
2816             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2817             ewitab           = _mm_slli_epi32(ewitab,2);
2818             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
2819             ewtabD           = _mm_setzero_pd();
2820             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
2821             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
2822             ewtabFn          = _mm_setzero_pd();
2823             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
2824             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
2825             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
2826             velec            = _mm_mul_pd(qq20,_mm_sub_pd(rinv20,velec));
2827             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
2828
2829             d                = _mm_sub_pd(r20,rswitch);
2830             d                = _mm_max_pd(d,_mm_setzero_pd());
2831             d2               = _mm_mul_pd(d,d);
2832             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_add_pd(swV3,_mm_mul_pd(d,_mm_add_pd(swV4,_mm_mul_pd(d,swV5)))))));
2833
2834             dsw              = _mm_mul_pd(d2,_mm_add_pd(swF2,_mm_mul_pd(d,_mm_add_pd(swF3,_mm_mul_pd(d,swF4)))));
2835
2836             /* Evaluate switch function */
2837             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2838             felec            = _mm_sub_pd( _mm_mul_pd(felec,sw) , _mm_mul_pd(rinv20,_mm_mul_pd(velec,dsw)) );
2839             cutoff_mask      = _mm_cmplt_pd(rsq20,rcutoff2);
2840
2841             fscal            = felec;
2842
2843             fscal            = _mm_and_pd(fscal,cutoff_mask);
2844
2845             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2846
2847             /* Calculate temporary vectorial force */
2848             tx               = _mm_mul_pd(fscal,dx20);
2849             ty               = _mm_mul_pd(fscal,dy20);
2850             tz               = _mm_mul_pd(fscal,dz20);
2851
2852             /* Update vectorial force */
2853             fix2             = _mm_add_pd(fix2,tx);
2854             fiy2             = _mm_add_pd(fiy2,ty);
2855             fiz2             = _mm_add_pd(fiz2,tz);
2856
2857             fjx0             = _mm_add_pd(fjx0,tx);
2858             fjy0             = _mm_add_pd(fjy0,ty);
2859             fjz0             = _mm_add_pd(fjz0,tz);
2860
2861             }
2862
2863             /**************************
2864              * CALCULATE INTERACTIONS *
2865              **************************/
2866
2867             if (gmx_mm_any_lt(rsq21,rcutoff2))
2868             {
2869
2870             r21              = _mm_mul_pd(rsq21,rinv21);
2871
2872             /* EWALD ELECTROSTATICS */
2873
2874             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2875             ewrt             = _mm_mul_pd(r21,ewtabscale);
2876             ewitab           = _mm_cvttpd_epi32(ewrt);
2877             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2878             ewitab           = _mm_slli_epi32(ewitab,2);
2879             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
2880             ewtabD           = _mm_setzero_pd();
2881             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
2882             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
2883             ewtabFn          = _mm_setzero_pd();
2884             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
2885             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
2886             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
2887             velec            = _mm_mul_pd(qq21,_mm_sub_pd(rinv21,velec));
2888             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
2889
2890             d                = _mm_sub_pd(r21,rswitch);
2891             d                = _mm_max_pd(d,_mm_setzero_pd());
2892             d2               = _mm_mul_pd(d,d);
2893             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_add_pd(swV3,_mm_mul_pd(d,_mm_add_pd(swV4,_mm_mul_pd(d,swV5)))))));
2894
2895             dsw              = _mm_mul_pd(d2,_mm_add_pd(swF2,_mm_mul_pd(d,_mm_add_pd(swF3,_mm_mul_pd(d,swF4)))));
2896
2897             /* Evaluate switch function */
2898             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2899             felec            = _mm_sub_pd( _mm_mul_pd(felec,sw) , _mm_mul_pd(rinv21,_mm_mul_pd(velec,dsw)) );
2900             cutoff_mask      = _mm_cmplt_pd(rsq21,rcutoff2);
2901
2902             fscal            = felec;
2903
2904             fscal            = _mm_and_pd(fscal,cutoff_mask);
2905
2906             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2907
2908             /* Calculate temporary vectorial force */
2909             tx               = _mm_mul_pd(fscal,dx21);
2910             ty               = _mm_mul_pd(fscal,dy21);
2911             tz               = _mm_mul_pd(fscal,dz21);
2912
2913             /* Update vectorial force */
2914             fix2             = _mm_add_pd(fix2,tx);
2915             fiy2             = _mm_add_pd(fiy2,ty);
2916             fiz2             = _mm_add_pd(fiz2,tz);
2917
2918             fjx1             = _mm_add_pd(fjx1,tx);
2919             fjy1             = _mm_add_pd(fjy1,ty);
2920             fjz1             = _mm_add_pd(fjz1,tz);
2921
2922             }
2923
2924             /**************************
2925              * CALCULATE INTERACTIONS *
2926              **************************/
2927
2928             if (gmx_mm_any_lt(rsq22,rcutoff2))
2929             {
2930
2931             r22              = _mm_mul_pd(rsq22,rinv22);
2932
2933             /* EWALD ELECTROSTATICS */
2934
2935             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2936             ewrt             = _mm_mul_pd(r22,ewtabscale);
2937             ewitab           = _mm_cvttpd_epi32(ewrt);
2938             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2939             ewitab           = _mm_slli_epi32(ewitab,2);
2940             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
2941             ewtabD           = _mm_setzero_pd();
2942             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
2943             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
2944             ewtabFn          = _mm_setzero_pd();
2945             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
2946             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
2947             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
2948             velec            = _mm_mul_pd(qq22,_mm_sub_pd(rinv22,velec));
2949             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
2950
2951             d                = _mm_sub_pd(r22,rswitch);
2952             d                = _mm_max_pd(d,_mm_setzero_pd());
2953             d2               = _mm_mul_pd(d,d);
2954             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_add_pd(swV3,_mm_mul_pd(d,_mm_add_pd(swV4,_mm_mul_pd(d,swV5)))))));
2955
2956             dsw              = _mm_mul_pd(d2,_mm_add_pd(swF2,_mm_mul_pd(d,_mm_add_pd(swF3,_mm_mul_pd(d,swF4)))));
2957
2958             /* Evaluate switch function */
2959             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2960             felec            = _mm_sub_pd( _mm_mul_pd(felec,sw) , _mm_mul_pd(rinv22,_mm_mul_pd(velec,dsw)) );
2961             cutoff_mask      = _mm_cmplt_pd(rsq22,rcutoff2);
2962
2963             fscal            = felec;
2964
2965             fscal            = _mm_and_pd(fscal,cutoff_mask);
2966
2967             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2968
2969             /* Calculate temporary vectorial force */
2970             tx               = _mm_mul_pd(fscal,dx22);
2971             ty               = _mm_mul_pd(fscal,dy22);
2972             tz               = _mm_mul_pd(fscal,dz22);
2973
2974             /* Update vectorial force */
2975             fix2             = _mm_add_pd(fix2,tx);
2976             fiy2             = _mm_add_pd(fiy2,ty);
2977             fiz2             = _mm_add_pd(fiz2,tz);
2978
2979             fjx2             = _mm_add_pd(fjx2,tx);
2980             fjy2             = _mm_add_pd(fjy2,ty);
2981             fjz2             = _mm_add_pd(fjz2,tz);
2982
2983             }
2984
2985             gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
2986
2987             /* Inner loop uses 558 flops */
2988         }
2989
2990         /* End of innermost loop */
2991
2992         gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
2993                                               f+i_coord_offset,fshift+i_shift_offset);
2994
2995         /* Increment number of inner iterations */
2996         inneriter                  += j_index_end - j_index_start;
2997
2998         /* Outer loop uses 18 flops */
2999     }
3000
3001     /* Increment number of outer iterations */
3002     outeriter        += nri;
3003
3004     /* Update outer/inner flops */
3005
3006     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3W3_F,outeriter*18 + inneriter*558);
3007 }