src/gromacs/gmxlib/nonbonded/nb_kernel_sse4_1_double/nb_kernel_ElecEwSh_VdwLJSh_GeomW3W3_sse4_1_double.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS sse4_1_double kernel generator.
  37  */
  38 #ifdef HAVE_CONFIG_H
  39 #include <config.h>
  40 #endif
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "types/simple.h"
  46 #include "vec.h"
  47 #include "nrnb.h"
  48
  49 #include "gromacs/simd/math_x86_sse4_1_double.h"
  50 #include "kernelutil_x86_sse4_1_double.h"
  51
  52 /*
  53  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSh_VdwLJSh_GeomW3W3_VF_sse4_1_double
  54  * Electrostatics interaction: Ewald
  55  * VdW interaction:            LennardJones
  56  * Geometry:                   Water3-Water3
  57  * Calculate force/pot:        PotentialAndForce
  58  */
  59 void
  60 nb_kernel_ElecEwSh_VdwLJSh_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     int              nvdwtype;
 106     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 107     int              *vdwtype;
 108     real             *vdwparam;
 109     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
 110     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
 111     __m128i          ewitab;
 112     __m128d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 113     real             *ewtab;
 114     __m128d          dummy_mask,cutoff_mask;
 115     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 116     __m128d          one     = _mm_set1_pd(1.0);
 117     __m128d          two     = _mm_set1_pd(2.0);
 118     x                = xx[0];
 119     f                = ff[0];
 120
 121     nri              = nlist->nri;
 122     iinr             = nlist->iinr;
 123     jindex           = nlist->jindex;
 124     jjnr             = nlist->jjnr;
 125     shiftidx         = nlist->shift;
 126     gid              = nlist->gid;
 127     shiftvec         = fr->shift_vec[0];
 128     fshift           = fr->fshift[0];
 129     facel            = _mm_set1_pd(fr->epsfac);
 130     charge           = mdatoms->chargeA;
 131     nvdwtype         = fr->ntype;
 132     vdwparam         = fr->nbfp;
 133     vdwtype          = mdatoms->typeA;
 134
 135     sh_ewald         = _mm_set1_pd(fr->ic->sh_ewald);
 136     ewtab            = fr->ic->tabq_coul_FDV0;
 137     ewtabscale       = _mm_set1_pd(fr->ic->tabq_scale);
 138     ewtabhalfspace   = _mm_set1_pd(0.5/fr->ic->tabq_scale);
 139
 140     /* Setup water-specific parameters */
 141     inr              = nlist->iinr[0];
 142     iq0              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
 143     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
 144     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
 145     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 146
 147     jq0              = _mm_set1_pd(charge[inr+0]);
 148     jq1              = _mm_set1_pd(charge[inr+1]);
 149     jq2              = _mm_set1_pd(charge[inr+2]);
 150     vdwjidx0A        = 2*vdwtype[inr+0];
 151     qq00             = _mm_mul_pd(iq0,jq0);
 152     c6_00            = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
 153     c12_00           = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
 154     qq01             = _mm_mul_pd(iq0,jq1);
 155     qq02             = _mm_mul_pd(iq0,jq2);
 156     qq10             = _mm_mul_pd(iq1,jq0);
 157     qq11             = _mm_mul_pd(iq1,jq1);
 158     qq12             = _mm_mul_pd(iq1,jq2);
 159     qq20             = _mm_mul_pd(iq2,jq0);
 160     qq21             = _mm_mul_pd(iq2,jq1);
 161     qq22             = _mm_mul_pd(iq2,jq2);
 162
 163     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 164     rcutoff_scalar   = fr->rcoulomb;
 165     rcutoff          = _mm_set1_pd(rcutoff_scalar);
 166     rcutoff2         = _mm_mul_pd(rcutoff,rcutoff);
 167
 168     sh_vdw_invrcut6  = _mm_set1_pd(fr->ic->sh_invrc6);
 169     rvdw             = _mm_set1_pd(fr->rvdw);
 170
 171     /* Avoid stupid compiler warnings */
 172     jnrA = jnrB = 0;
 173     j_coord_offsetA = 0;
 174     j_coord_offsetB = 0;
 175
 176     outeriter        = 0;
 177     inneriter        = 0;
 178
 179     /* Start outer loop over neighborlists */
 180     for(iidx=0; iidx<nri; iidx++)
 181     {
 182         /* Load shift vector for this list */
 183         i_shift_offset   = DIM*shiftidx[iidx];
 184
 185         /* Load limits for loop over neighbors */
 186         j_index_start    = jindex[iidx];
 187         j_index_end      = jindex[iidx+1];
 188
 189         /* Get outer coordinate index */
 190         inr              = iinr[iidx];
 191         i_coord_offset   = DIM*inr;
 192
 193         /* Load i particle coords and add shift vector */
 194         gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 195                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 196
 197         fix0             = _mm_setzero_pd();
 198         fiy0             = _mm_setzero_pd();
 199         fiz0             = _mm_setzero_pd();
 200         fix1             = _mm_setzero_pd();
 201         fiy1             = _mm_setzero_pd();
 202         fiz1             = _mm_setzero_pd();
 203         fix2             = _mm_setzero_pd();
 204         fiy2             = _mm_setzero_pd();
 205         fiz2             = _mm_setzero_pd();
 206
 207         /* Reset potential sums */
 208         velecsum         = _mm_setzero_pd();
 209         vvdwsum          = _mm_setzero_pd();
 210
 211         /* Start inner kernel loop */
 212         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 213         {
 214
 215             /* Get j neighbor index, and coordinate index */
 216             jnrA             = jjnr[jidx];
 217             jnrB             = jjnr[jidx+1];
 218             j_coord_offsetA  = DIM*jnrA;
 219             j_coord_offsetB  = DIM*jnrB;
 220
 221             /* load j atom coordinates */
 222             gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 223                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 224
 225             /* Calculate displacement vector */
 226             dx00             = _mm_sub_pd(ix0,jx0);
 227             dy00             = _mm_sub_pd(iy0,jy0);
 228             dz00             = _mm_sub_pd(iz0,jz0);
 229             dx01             = _mm_sub_pd(ix0,jx1);
 230             dy01             = _mm_sub_pd(iy0,jy1);
 231             dz01             = _mm_sub_pd(iz0,jz1);
 232             dx02             = _mm_sub_pd(ix0,jx2);
 233             dy02             = _mm_sub_pd(iy0,jy2);
 234             dz02             = _mm_sub_pd(iz0,jz2);
 235             dx10             = _mm_sub_pd(ix1,jx0);
 236             dy10             = _mm_sub_pd(iy1,jy0);
 237             dz10             = _mm_sub_pd(iz1,jz0);
 238             dx11             = _mm_sub_pd(ix1,jx1);
 239             dy11             = _mm_sub_pd(iy1,jy1);
 240             dz11             = _mm_sub_pd(iz1,jz1);
 241             dx12             = _mm_sub_pd(ix1,jx2);
 242             dy12             = _mm_sub_pd(iy1,jy2);
 243             dz12             = _mm_sub_pd(iz1,jz2);
 244             dx20             = _mm_sub_pd(ix2,jx0);
 245             dy20             = _mm_sub_pd(iy2,jy0);
 246             dz20             = _mm_sub_pd(iz2,jz0);
 247             dx21             = _mm_sub_pd(ix2,jx1);
 248             dy21             = _mm_sub_pd(iy2,jy1);
 249             dz21             = _mm_sub_pd(iz2,jz1);
 250             dx22             = _mm_sub_pd(ix2,jx2);
 251             dy22             = _mm_sub_pd(iy2,jy2);
 252             dz22             = _mm_sub_pd(iz2,jz2);
 253
 254             /* Calculate squared distance and things based on it */
 255             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 256             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
 257             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
 258             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 259             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 260             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 261             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 262             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 263             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 264
 265             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 266             rinv01           = gmx_mm_invsqrt_pd(rsq01);
 267             rinv02           = gmx_mm_invsqrt_pd(rsq02);
 268             rinv10           = gmx_mm_invsqrt_pd(rsq10);
 269             rinv11           = gmx_mm_invsqrt_pd(rsq11);
 270             rinv12           = gmx_mm_invsqrt_pd(rsq12);
 271             rinv20           = gmx_mm_invsqrt_pd(rsq20);
 272             rinv21           = gmx_mm_invsqrt_pd(rsq21);
 273             rinv22           = gmx_mm_invsqrt_pd(rsq22);
 274
 275             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 276             rinvsq01         = _mm_mul_pd(rinv01,rinv01);
 277             rinvsq02         = _mm_mul_pd(rinv02,rinv02);
 278             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
 279             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
 280             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
 281             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
 282             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
 283             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
 284
 285             fjx0             = _mm_setzero_pd();
 286             fjy0             = _mm_setzero_pd();
 287             fjz0             = _mm_setzero_pd();
 288             fjx1             = _mm_setzero_pd();
 289             fjy1             = _mm_setzero_pd();
 290             fjz1             = _mm_setzero_pd();
 291             fjx2             = _mm_setzero_pd();
 292             fjy2             = _mm_setzero_pd();
 293             fjz2             = _mm_setzero_pd();
 294
 295             /**************************
 296              * CALCULATE INTERACTIONS *
 297              **************************/
 298
 299             if (gmx_mm_any_lt(rsq00,rcutoff2))
 300             {
 301
 302             r00              = _mm_mul_pd(rsq00,rinv00);
 303
 304             /* EWALD ELECTROSTATICS */
 305
 306             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 307             ewrt             = _mm_mul_pd(r00,ewtabscale);
 308             ewitab           = _mm_cvttpd_epi32(ewrt);
 309             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 310             ewitab           = _mm_slli_epi32(ewitab,2);
 311             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 312             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 313             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 314             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 315             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 316             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 317             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 318             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 319             velec            = _mm_mul_pd(qq00,_mm_sub_pd(_mm_sub_pd(rinv00,sh_ewald),velec));
 320             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
 321
 322             /* LENNARD-JONES DISPERSION/REPULSION */
 323
 324             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 325             vvdw6            = _mm_mul_pd(c6_00,rinvsix);
 326             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
 327             vvdw             = _mm_sub_pd(_mm_mul_pd( _mm_sub_pd(vvdw12 , _mm_mul_pd(c12_00,_mm_mul_pd(sh_vdw_invrcut6,sh_vdw_invrcut6))), one_twelfth) ,
 328                                           _mm_mul_pd( _mm_sub_pd(vvdw6,_mm_mul_pd(c6_00,sh_vdw_invrcut6)),one_sixth));
 329             fvdw             = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
 330
 331             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
 332
 333             /* Update potential sum for this i atom from the interaction with this j atom. */
 334             velec            = _mm_and_pd(velec,cutoff_mask);
 335             velecsum         = _mm_add_pd(velecsum,velec);
 336             vvdw             = _mm_and_pd(vvdw,cutoff_mask);
 337             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 338
 339             fscal            = _mm_add_pd(felec,fvdw);
 340
 341             fscal            = _mm_and_pd(fscal,cutoff_mask);
 342
 343             /* Calculate temporary vectorial force */
 344             tx               = _mm_mul_pd(fscal,dx00);
 345             ty               = _mm_mul_pd(fscal,dy00);
 346             tz               = _mm_mul_pd(fscal,dz00);
 347
 348             /* Update vectorial force */
 349             fix0             = _mm_add_pd(fix0,tx);
 350             fiy0             = _mm_add_pd(fiy0,ty);
 351             fiz0             = _mm_add_pd(fiz0,tz);
 352
 353             fjx0             = _mm_add_pd(fjx0,tx);
 354             fjy0             = _mm_add_pd(fjy0,ty);
 355             fjz0             = _mm_add_pd(fjz0,tz);
 356
 357             }
 358
 359             /**************************
 360              * CALCULATE INTERACTIONS *
 361              **************************/
 362
 363             if (gmx_mm_any_lt(rsq01,rcutoff2))
 364             {
 365
 366             r01              = _mm_mul_pd(rsq01,rinv01);
 367
 368             /* EWALD ELECTROSTATICS */
 369
 370             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 371             ewrt             = _mm_mul_pd(r01,ewtabscale);
 372             ewitab           = _mm_cvttpd_epi32(ewrt);
 373             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 374             ewitab           = _mm_slli_epi32(ewitab,2);
 375             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 376             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 377             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 378             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 379             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 380             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 381             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 382             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 383             velec            = _mm_mul_pd(qq01,_mm_sub_pd(_mm_sub_pd(rinv01,sh_ewald),velec));
 384             felec            = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
 385
 386             cutoff_mask      = _mm_cmplt_pd(rsq01,rcutoff2);
 387
 388             /* Update potential sum for this i atom from the interaction with this j atom. */
 389             velec            = _mm_and_pd(velec,cutoff_mask);
 390             velecsum         = _mm_add_pd(velecsum,velec);
 391
 392             fscal            = felec;
 393
 394             fscal            = _mm_and_pd(fscal,cutoff_mask);
 395
 396             /* Calculate temporary vectorial force */
 397             tx               = _mm_mul_pd(fscal,dx01);
 398             ty               = _mm_mul_pd(fscal,dy01);
 399             tz               = _mm_mul_pd(fscal,dz01);
 400
 401             /* Update vectorial force */
 402             fix0             = _mm_add_pd(fix0,tx);
 403             fiy0             = _mm_add_pd(fiy0,ty);
 404             fiz0             = _mm_add_pd(fiz0,tz);
 405
 406             fjx1             = _mm_add_pd(fjx1,tx);
 407             fjy1             = _mm_add_pd(fjy1,ty);
 408             fjz1             = _mm_add_pd(fjz1,tz);
 409
 410             }
 411
 412             /**************************
 413              * CALCULATE INTERACTIONS *
 414              **************************/
 415
 416             if (gmx_mm_any_lt(rsq02,rcutoff2))
 417             {
 418
 419             r02              = _mm_mul_pd(rsq02,rinv02);
 420
 421             /* EWALD ELECTROSTATICS */
 422
 423             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 424             ewrt             = _mm_mul_pd(r02,ewtabscale);
 425             ewitab           = _mm_cvttpd_epi32(ewrt);
 426             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 427             ewitab           = _mm_slli_epi32(ewitab,2);
 428             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 429             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 430             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 431             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 432             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 433             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 434             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 435             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 436             velec            = _mm_mul_pd(qq02,_mm_sub_pd(_mm_sub_pd(rinv02,sh_ewald),velec));
 437             felec            = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
 438
 439             cutoff_mask      = _mm_cmplt_pd(rsq02,rcutoff2);
 440
 441             /* Update potential sum for this i atom from the interaction with this j atom. */
 442             velec            = _mm_and_pd(velec,cutoff_mask);
 443             velecsum         = _mm_add_pd(velecsum,velec);
 444
 445             fscal            = felec;
 446
 447             fscal            = _mm_and_pd(fscal,cutoff_mask);
 448
 449             /* Calculate temporary vectorial force */
 450             tx               = _mm_mul_pd(fscal,dx02);
 451             ty               = _mm_mul_pd(fscal,dy02);
 452             tz               = _mm_mul_pd(fscal,dz02);
 453
 454             /* Update vectorial force */
 455             fix0             = _mm_add_pd(fix0,tx);
 456             fiy0             = _mm_add_pd(fiy0,ty);
 457             fiz0             = _mm_add_pd(fiz0,tz);
 458
 459             fjx2             = _mm_add_pd(fjx2,tx);
 460             fjy2             = _mm_add_pd(fjy2,ty);
 461             fjz2             = _mm_add_pd(fjz2,tz);
 462
 463             }
 464
 465             /**************************
 466              * CALCULATE INTERACTIONS *
 467              **************************/
 468
 469             if (gmx_mm_any_lt(rsq10,rcutoff2))
 470             {
 471
 472             r10              = _mm_mul_pd(rsq10,rinv10);
 473
 474             /* EWALD ELECTROSTATICS */
 475
 476             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 477             ewrt             = _mm_mul_pd(r10,ewtabscale);
 478             ewitab           = _mm_cvttpd_epi32(ewrt);
 479             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 480             ewitab           = _mm_slli_epi32(ewitab,2);
 481             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 482             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 483             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 484             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 485             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 486             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 487             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 488             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 489             velec            = _mm_mul_pd(qq10,_mm_sub_pd(_mm_sub_pd(rinv10,sh_ewald),velec));
 490             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
 491
 492             cutoff_mask      = _mm_cmplt_pd(rsq10,rcutoff2);
 493
 494             /* Update potential sum for this i atom from the interaction with this j atom. */
 495             velec            = _mm_and_pd(velec,cutoff_mask);
 496             velecsum         = _mm_add_pd(velecsum,velec);
 497
 498             fscal            = felec;
 499
 500             fscal            = _mm_and_pd(fscal,cutoff_mask);
 501
 502             /* Calculate temporary vectorial force */
 503             tx               = _mm_mul_pd(fscal,dx10);
 504             ty               = _mm_mul_pd(fscal,dy10);
 505             tz               = _mm_mul_pd(fscal,dz10);
 506
 507             /* Update vectorial force */
 508             fix1             = _mm_add_pd(fix1,tx);
 509             fiy1             = _mm_add_pd(fiy1,ty);
 510             fiz1             = _mm_add_pd(fiz1,tz);
 511
 512             fjx0             = _mm_add_pd(fjx0,tx);
 513             fjy0             = _mm_add_pd(fjy0,ty);
 514             fjz0             = _mm_add_pd(fjz0,tz);
 515
 516             }
 517
 518             /**************************
 519              * CALCULATE INTERACTIONS *
 520              **************************/
 521
 522             if (gmx_mm_any_lt(rsq11,rcutoff2))
 523             {
 524
 525             r11              = _mm_mul_pd(rsq11,rinv11);
 526
 527             /* EWALD ELECTROSTATICS */
 528
 529             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 530             ewrt             = _mm_mul_pd(r11,ewtabscale);
 531             ewitab           = _mm_cvttpd_epi32(ewrt);
 532             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 533             ewitab           = _mm_slli_epi32(ewitab,2);
 534             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 535             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 536             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 537             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 538             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 539             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 540             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 541             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 542             velec            = _mm_mul_pd(qq11,_mm_sub_pd(_mm_sub_pd(rinv11,sh_ewald),velec));
 543             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
 544
 545             cutoff_mask      = _mm_cmplt_pd(rsq11,rcutoff2);
 546
 547             /* Update potential sum for this i atom from the interaction with this j atom. */
 548             velec            = _mm_and_pd(velec,cutoff_mask);
 549             velecsum         = _mm_add_pd(velecsum,velec);
 550
 551             fscal            = felec;
 552
 553             fscal            = _mm_and_pd(fscal,cutoff_mask);
 554
 555             /* Calculate temporary vectorial force */
 556             tx               = _mm_mul_pd(fscal,dx11);
 557             ty               = _mm_mul_pd(fscal,dy11);
 558             tz               = _mm_mul_pd(fscal,dz11);
 559
 560             /* Update vectorial force */
 561             fix1             = _mm_add_pd(fix1,tx);
 562             fiy1             = _mm_add_pd(fiy1,ty);
 563             fiz1             = _mm_add_pd(fiz1,tz);
 564
 565             fjx1             = _mm_add_pd(fjx1,tx);
 566             fjy1             = _mm_add_pd(fjy1,ty);
 567             fjz1             = _mm_add_pd(fjz1,tz);
 568
 569             }
 570
 571             /**************************
 572              * CALCULATE INTERACTIONS *
 573              **************************/
 574
 575             if (gmx_mm_any_lt(rsq12,rcutoff2))
 576             {
 577
 578             r12              = _mm_mul_pd(rsq12,rinv12);
 579
 580             /* EWALD ELECTROSTATICS */
 581
 582             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 583             ewrt             = _mm_mul_pd(r12,ewtabscale);
 584             ewitab           = _mm_cvttpd_epi32(ewrt);
 585             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 586             ewitab           = _mm_slli_epi32(ewitab,2);
 587             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 588             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 589             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 590             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 591             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 592             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 593             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 594             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 595             velec            = _mm_mul_pd(qq12,_mm_sub_pd(_mm_sub_pd(rinv12,sh_ewald),velec));
 596             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
 597
 598             cutoff_mask      = _mm_cmplt_pd(rsq12,rcutoff2);
 599
 600             /* Update potential sum for this i atom from the interaction with this j atom. */
 601             velec            = _mm_and_pd(velec,cutoff_mask);
 602             velecsum         = _mm_add_pd(velecsum,velec);
 603
 604             fscal            = felec;
 605
 606             fscal            = _mm_and_pd(fscal,cutoff_mask);
 607
 608             /* Calculate temporary vectorial force */
 609             tx               = _mm_mul_pd(fscal,dx12);
 610             ty               = _mm_mul_pd(fscal,dy12);
 611             tz               = _mm_mul_pd(fscal,dz12);
 612
 613             /* Update vectorial force */
 614             fix1             = _mm_add_pd(fix1,tx);
 615             fiy1             = _mm_add_pd(fiy1,ty);
 616             fiz1             = _mm_add_pd(fiz1,tz);
 617
 618             fjx2             = _mm_add_pd(fjx2,tx);
 619             fjy2             = _mm_add_pd(fjy2,ty);
 620             fjz2             = _mm_add_pd(fjz2,tz);
 621
 622             }
 623
 624             /**************************
 625              * CALCULATE INTERACTIONS *
 626              **************************/
 627
 628             if (gmx_mm_any_lt(rsq20,rcutoff2))
 629             {
 630
 631             r20              = _mm_mul_pd(rsq20,rinv20);
 632
 633             /* EWALD ELECTROSTATICS */
 634
 635             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 636             ewrt             = _mm_mul_pd(r20,ewtabscale);
 637             ewitab           = _mm_cvttpd_epi32(ewrt);
 638             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 639             ewitab           = _mm_slli_epi32(ewitab,2);
 640             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 641             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 642             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 643             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 644             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 645             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 646             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 647             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 648             velec            = _mm_mul_pd(qq20,_mm_sub_pd(_mm_sub_pd(rinv20,sh_ewald),velec));
 649             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
 650
 651             cutoff_mask      = _mm_cmplt_pd(rsq20,rcutoff2);
 652
 653             /* Update potential sum for this i atom from the interaction with this j atom. */
 654             velec            = _mm_and_pd(velec,cutoff_mask);
 655             velecsum         = _mm_add_pd(velecsum,velec);
 656
 657             fscal            = felec;
 658
 659             fscal            = _mm_and_pd(fscal,cutoff_mask);
 660
 661             /* Calculate temporary vectorial force */
 662             tx               = _mm_mul_pd(fscal,dx20);
 663             ty               = _mm_mul_pd(fscal,dy20);
 664             tz               = _mm_mul_pd(fscal,dz20);
 665
 666             /* Update vectorial force */
 667             fix2             = _mm_add_pd(fix2,tx);
 668             fiy2             = _mm_add_pd(fiy2,ty);
 669             fiz2             = _mm_add_pd(fiz2,tz);
 670
 671             fjx0             = _mm_add_pd(fjx0,tx);
 672             fjy0             = _mm_add_pd(fjy0,ty);
 673             fjz0             = _mm_add_pd(fjz0,tz);
 674
 675             }
 676
 677             /**************************
 678              * CALCULATE INTERACTIONS *
 679              **************************/
 680
 681             if (gmx_mm_any_lt(rsq21,rcutoff2))
 682             {
 683
 684             r21              = _mm_mul_pd(rsq21,rinv21);
 685
 686             /* EWALD ELECTROSTATICS */
 687
 688             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 689             ewrt             = _mm_mul_pd(r21,ewtabscale);
 690             ewitab           = _mm_cvttpd_epi32(ewrt);
 691             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 692             ewitab           = _mm_slli_epi32(ewitab,2);
 693             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 694             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 695             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 696             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 697             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 698             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 699             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 700             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 701             velec            = _mm_mul_pd(qq21,_mm_sub_pd(_mm_sub_pd(rinv21,sh_ewald),velec));
 702             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
 703
 704             cutoff_mask      = _mm_cmplt_pd(rsq21,rcutoff2);
 705
 706             /* Update potential sum for this i atom from the interaction with this j atom. */
 707             velec            = _mm_and_pd(velec,cutoff_mask);
 708             velecsum         = _mm_add_pd(velecsum,velec);
 709
 710             fscal            = felec;
 711
 712             fscal            = _mm_and_pd(fscal,cutoff_mask);
 713
 714             /* Calculate temporary vectorial force */
 715             tx               = _mm_mul_pd(fscal,dx21);
 716             ty               = _mm_mul_pd(fscal,dy21);
 717             tz               = _mm_mul_pd(fscal,dz21);
 718
 719             /* Update vectorial force */
 720             fix2             = _mm_add_pd(fix2,tx);
 721             fiy2             = _mm_add_pd(fiy2,ty);
 722             fiz2             = _mm_add_pd(fiz2,tz);
 723
 724             fjx1             = _mm_add_pd(fjx1,tx);
 725             fjy1             = _mm_add_pd(fjy1,ty);
 726             fjz1             = _mm_add_pd(fjz1,tz);
 727
 728             }
 729
 730             /**************************
 731              * CALCULATE INTERACTIONS *
 732              **************************/
 733
 734             if (gmx_mm_any_lt(rsq22,rcutoff2))
 735             {
 736
 737             r22              = _mm_mul_pd(rsq22,rinv22);
 738
 739             /* EWALD ELECTROSTATICS */
 740
 741             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 742             ewrt             = _mm_mul_pd(r22,ewtabscale);
 743             ewitab           = _mm_cvttpd_epi32(ewrt);
 744             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 745             ewitab           = _mm_slli_epi32(ewitab,2);
 746             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 747             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 748             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 749             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 750             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 751             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 752             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 753             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 754             velec            = _mm_mul_pd(qq22,_mm_sub_pd(_mm_sub_pd(rinv22,sh_ewald),velec));
 755             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
 756
 757             cutoff_mask      = _mm_cmplt_pd(rsq22,rcutoff2);
 758
 759             /* Update potential sum for this i atom from the interaction with this j atom. */
 760             velec            = _mm_and_pd(velec,cutoff_mask);
 761             velecsum         = _mm_add_pd(velecsum,velec);
 762
 763             fscal            = felec;
 764
 765             fscal            = _mm_and_pd(fscal,cutoff_mask);
 766
 767             /* Calculate temporary vectorial force */
 768             tx               = _mm_mul_pd(fscal,dx22);
 769             ty               = _mm_mul_pd(fscal,dy22);
 770             tz               = _mm_mul_pd(fscal,dz22);
 771
 772             /* Update vectorial force */
 773             fix2             = _mm_add_pd(fix2,tx);
 774             fiy2             = _mm_add_pd(fiy2,ty);
 775             fiz2             = _mm_add_pd(fiz2,tz);
 776
 777             fjx2             = _mm_add_pd(fjx2,tx);
 778             fjy2             = _mm_add_pd(fjy2,ty);
 779             fjz2             = _mm_add_pd(fjz2,tz);
 780
 781             }
 782
 783             gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 784
 785             /* Inner loop uses 432 flops */
 786         }
 787
 788         if(jidx<j_index_end)
 789         {
 790
 791             jnrA             = jjnr[jidx];
 792             j_coord_offsetA  = DIM*jnrA;
 793
 794             /* load j atom coordinates */
 795             gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 796                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 797
 798             /* Calculate displacement vector */
 799             dx00             = _mm_sub_pd(ix0,jx0);
 800             dy00             = _mm_sub_pd(iy0,jy0);
 801             dz00             = _mm_sub_pd(iz0,jz0);
 802             dx01             = _mm_sub_pd(ix0,jx1);
 803             dy01             = _mm_sub_pd(iy0,jy1);
 804             dz01             = _mm_sub_pd(iz0,jz1);
 805             dx02             = _mm_sub_pd(ix0,jx2);
 806             dy02             = _mm_sub_pd(iy0,jy2);
 807             dz02             = _mm_sub_pd(iz0,jz2);
 808             dx10             = _mm_sub_pd(ix1,jx0);
 809             dy10             = _mm_sub_pd(iy1,jy0);
 810             dz10             = _mm_sub_pd(iz1,jz0);
 811             dx11             = _mm_sub_pd(ix1,jx1);
 812             dy11             = _mm_sub_pd(iy1,jy1);
 813             dz11             = _mm_sub_pd(iz1,jz1);
 814             dx12             = _mm_sub_pd(ix1,jx2);
 815             dy12             = _mm_sub_pd(iy1,jy2);
 816             dz12             = _mm_sub_pd(iz1,jz2);
 817             dx20             = _mm_sub_pd(ix2,jx0);
 818             dy20             = _mm_sub_pd(iy2,jy0);
 819             dz20             = _mm_sub_pd(iz2,jz0);
 820             dx21             = _mm_sub_pd(ix2,jx1);
 821             dy21             = _mm_sub_pd(iy2,jy1);
 822             dz21             = _mm_sub_pd(iz2,jz1);
 823             dx22             = _mm_sub_pd(ix2,jx2);
 824             dy22             = _mm_sub_pd(iy2,jy2);
 825             dz22             = _mm_sub_pd(iz2,jz2);
 826
 827             /* Calculate squared distance and things based on it */
 828             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 829             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
 830             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
 831             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 832             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 833             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 834             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 835             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 836             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 837
 838             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 839             rinv01           = gmx_mm_invsqrt_pd(rsq01);
 840             rinv02           = gmx_mm_invsqrt_pd(rsq02);
 841             rinv10           = gmx_mm_invsqrt_pd(rsq10);
 842             rinv11           = gmx_mm_invsqrt_pd(rsq11);
 843             rinv12           = gmx_mm_invsqrt_pd(rsq12);
 844             rinv20           = gmx_mm_invsqrt_pd(rsq20);
 845             rinv21           = gmx_mm_invsqrt_pd(rsq21);
 846             rinv22           = gmx_mm_invsqrt_pd(rsq22);
 847
 848             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 849             rinvsq01         = _mm_mul_pd(rinv01,rinv01);
 850             rinvsq02         = _mm_mul_pd(rinv02,rinv02);
 851             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
 852             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
 853             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
 854             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
 855             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
 856             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
 857
 858             fjx0             = _mm_setzero_pd();
 859             fjy0             = _mm_setzero_pd();
 860             fjz0             = _mm_setzero_pd();
 861             fjx1             = _mm_setzero_pd();
 862             fjy1             = _mm_setzero_pd();
 863             fjz1             = _mm_setzero_pd();
 864             fjx2             = _mm_setzero_pd();
 865             fjy2             = _mm_setzero_pd();
 866             fjz2             = _mm_setzero_pd();
 867
 868             /**************************
 869              * CALCULATE INTERACTIONS *
 870              **************************/
 871
 872             if (gmx_mm_any_lt(rsq00,rcutoff2))
 873             {
 874
 875             r00              = _mm_mul_pd(rsq00,rinv00);
 876
 877             /* EWALD ELECTROSTATICS */
 878
 879             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 880             ewrt             = _mm_mul_pd(r00,ewtabscale);
 881             ewitab           = _mm_cvttpd_epi32(ewrt);
 882             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 883             ewitab           = _mm_slli_epi32(ewitab,2);
 884             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 885             ewtabD           = _mm_setzero_pd();
 886             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 887             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 888             ewtabFn          = _mm_setzero_pd();
 889             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 890             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 891             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 892             velec            = _mm_mul_pd(qq00,_mm_sub_pd(_mm_sub_pd(rinv00,sh_ewald),velec));
 893             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
 894
 895             /* LENNARD-JONES DISPERSION/REPULSION */
 896
 897             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 898             vvdw6            = _mm_mul_pd(c6_00,rinvsix);
 899             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
 900             vvdw             = _mm_sub_pd(_mm_mul_pd( _mm_sub_pd(vvdw12 , _mm_mul_pd(c12_00,_mm_mul_pd(sh_vdw_invrcut6,sh_vdw_invrcut6))), one_twelfth) ,
 901                                           _mm_mul_pd( _mm_sub_pd(vvdw6,_mm_mul_pd(c6_00,sh_vdw_invrcut6)),one_sixth));
 902             fvdw             = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
 903
 904             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
 905
 906             /* Update potential sum for this i atom from the interaction with this j atom. */
 907             velec            = _mm_and_pd(velec,cutoff_mask);
 908             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 909             velecsum         = _mm_add_pd(velecsum,velec);
 910             vvdw             = _mm_and_pd(vvdw,cutoff_mask);
 911             vvdw             = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
 912             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 913
 914             fscal            = _mm_add_pd(felec,fvdw);
 915
 916             fscal            = _mm_and_pd(fscal,cutoff_mask);
 917
 918             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 919
 920             /* Calculate temporary vectorial force */
 921             tx               = _mm_mul_pd(fscal,dx00);
 922             ty               = _mm_mul_pd(fscal,dy00);
 923             tz               = _mm_mul_pd(fscal,dz00);
 924
 925             /* Update vectorial force */
 926             fix0             = _mm_add_pd(fix0,tx);
 927             fiy0             = _mm_add_pd(fiy0,ty);
 928             fiz0             = _mm_add_pd(fiz0,tz);
 929
 930             fjx0             = _mm_add_pd(fjx0,tx);
 931             fjy0             = _mm_add_pd(fjy0,ty);
 932             fjz0             = _mm_add_pd(fjz0,tz);
 933
 934             }
 935
 936             /**************************
 937              * CALCULATE INTERACTIONS *
 938              **************************/
 939
 940             if (gmx_mm_any_lt(rsq01,rcutoff2))
 941             {
 942
 943             r01              = _mm_mul_pd(rsq01,rinv01);
 944
 945             /* EWALD ELECTROSTATICS */
 946
 947             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 948             ewrt             = _mm_mul_pd(r01,ewtabscale);
 949             ewitab           = _mm_cvttpd_epi32(ewrt);
 950             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 951             ewitab           = _mm_slli_epi32(ewitab,2);
 952             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 953             ewtabD           = _mm_setzero_pd();
 954             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 955             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 956             ewtabFn          = _mm_setzero_pd();
 957             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 958             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 959             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 960             velec            = _mm_mul_pd(qq01,_mm_sub_pd(_mm_sub_pd(rinv01,sh_ewald),velec));
 961             felec            = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
 962
 963             cutoff_mask      = _mm_cmplt_pd(rsq01,rcutoff2);
 964
 965             /* Update potential sum for this i atom from the interaction with this j atom. */
 966             velec            = _mm_and_pd(velec,cutoff_mask);
 967             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 968             velecsum         = _mm_add_pd(velecsum,velec);
 969
 970             fscal            = felec;
 971
 972             fscal            = _mm_and_pd(fscal,cutoff_mask);
 973
 974             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 975
 976             /* Calculate temporary vectorial force */
 977             tx               = _mm_mul_pd(fscal,dx01);
 978             ty               = _mm_mul_pd(fscal,dy01);
 979             tz               = _mm_mul_pd(fscal,dz01);
 980
 981             /* Update vectorial force */
 982             fix0             = _mm_add_pd(fix0,tx);
 983             fiy0             = _mm_add_pd(fiy0,ty);
 984             fiz0             = _mm_add_pd(fiz0,tz);
 985
 986             fjx1             = _mm_add_pd(fjx1,tx);
 987             fjy1             = _mm_add_pd(fjy1,ty);
 988             fjz1             = _mm_add_pd(fjz1,tz);
 989
 990             }
 991
 992             /**************************
 993              * CALCULATE INTERACTIONS *
 994              **************************/
 995
 996             if (gmx_mm_any_lt(rsq02,rcutoff2))
 997             {
 998
 999             r02              = _mm_mul_pd(rsq02,rinv02);
1000
1001             /* EWALD ELECTROSTATICS */
1002
1003             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1004             ewrt             = _mm_mul_pd(r02,ewtabscale);
1005             ewitab           = _mm_cvttpd_epi32(ewrt);
1006             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1007             ewitab           = _mm_slli_epi32(ewitab,2);
1008             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1009             ewtabD           = _mm_setzero_pd();
1010             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1011             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1012             ewtabFn          = _mm_setzero_pd();
1013             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1014             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1015             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1016             velec            = _mm_mul_pd(qq02,_mm_sub_pd(_mm_sub_pd(rinv02,sh_ewald),velec));
1017             felec            = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
1018
1019             cutoff_mask      = _mm_cmplt_pd(rsq02,rcutoff2);
1020
1021             /* Update potential sum for this i atom from the interaction with this j atom. */
1022             velec            = _mm_and_pd(velec,cutoff_mask);
1023             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1024             velecsum         = _mm_add_pd(velecsum,velec);
1025
1026             fscal            = felec;
1027
1028             fscal            = _mm_and_pd(fscal,cutoff_mask);
1029
1030             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1031
1032             /* Calculate temporary vectorial force */
1033             tx               = _mm_mul_pd(fscal,dx02);
1034             ty               = _mm_mul_pd(fscal,dy02);
1035             tz               = _mm_mul_pd(fscal,dz02);
1036
1037             /* Update vectorial force */
1038             fix0             = _mm_add_pd(fix0,tx);
1039             fiy0             = _mm_add_pd(fiy0,ty);
1040             fiz0             = _mm_add_pd(fiz0,tz);
1041
1042             fjx2             = _mm_add_pd(fjx2,tx);
1043             fjy2             = _mm_add_pd(fjy2,ty);
1044             fjz2             = _mm_add_pd(fjz2,tz);
1045
1046             }
1047
1048             /**************************
1049              * CALCULATE INTERACTIONS *
1050              **************************/
1051
1052             if (gmx_mm_any_lt(rsq10,rcutoff2))
1053             {
1054
1055             r10              = _mm_mul_pd(rsq10,rinv10);
1056
1057             /* EWALD ELECTROSTATICS */
1058
1059             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1060             ewrt             = _mm_mul_pd(r10,ewtabscale);
1061             ewitab           = _mm_cvttpd_epi32(ewrt);
1062             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1063             ewitab           = _mm_slli_epi32(ewitab,2);
1064             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1065             ewtabD           = _mm_setzero_pd();
1066             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1067             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1068             ewtabFn          = _mm_setzero_pd();
1069             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1070             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1071             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1072             velec            = _mm_mul_pd(qq10,_mm_sub_pd(_mm_sub_pd(rinv10,sh_ewald),velec));
1073             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
1074
1075             cutoff_mask      = _mm_cmplt_pd(rsq10,rcutoff2);
1076
1077             /* Update potential sum for this i atom from the interaction with this j atom. */
1078             velec            = _mm_and_pd(velec,cutoff_mask);
1079             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1080             velecsum         = _mm_add_pd(velecsum,velec);
1081
1082             fscal            = felec;
1083
1084             fscal            = _mm_and_pd(fscal,cutoff_mask);
1085
1086             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1087
1088             /* Calculate temporary vectorial force */
1089             tx               = _mm_mul_pd(fscal,dx10);
1090             ty               = _mm_mul_pd(fscal,dy10);
1091             tz               = _mm_mul_pd(fscal,dz10);
1092
1093             /* Update vectorial force */
1094             fix1             = _mm_add_pd(fix1,tx);
1095             fiy1             = _mm_add_pd(fiy1,ty);
1096             fiz1             = _mm_add_pd(fiz1,tz);
1097
1098             fjx0             = _mm_add_pd(fjx0,tx);
1099             fjy0             = _mm_add_pd(fjy0,ty);
1100             fjz0             = _mm_add_pd(fjz0,tz);
1101
1102             }
1103
1104             /**************************
1105              * CALCULATE INTERACTIONS *
1106              **************************/
1107
1108             if (gmx_mm_any_lt(rsq11,rcutoff2))
1109             {
1110
1111             r11              = _mm_mul_pd(rsq11,rinv11);
1112
1113             /* EWALD ELECTROSTATICS */
1114
1115             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1116             ewrt             = _mm_mul_pd(r11,ewtabscale);
1117             ewitab           = _mm_cvttpd_epi32(ewrt);
1118             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1119             ewitab           = _mm_slli_epi32(ewitab,2);
1120             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1121             ewtabD           = _mm_setzero_pd();
1122             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1123             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1124             ewtabFn          = _mm_setzero_pd();
1125             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1126             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1127             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1128             velec            = _mm_mul_pd(qq11,_mm_sub_pd(_mm_sub_pd(rinv11,sh_ewald),velec));
1129             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
1130
1131             cutoff_mask      = _mm_cmplt_pd(rsq11,rcutoff2);
1132
1133             /* Update potential sum for this i atom from the interaction with this j atom. */
1134             velec            = _mm_and_pd(velec,cutoff_mask);
1135             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1136             velecsum         = _mm_add_pd(velecsum,velec);
1137
1138             fscal            = felec;
1139
1140             fscal            = _mm_and_pd(fscal,cutoff_mask);
1141
1142             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1143
1144             /* Calculate temporary vectorial force */
1145             tx               = _mm_mul_pd(fscal,dx11);
1146             ty               = _mm_mul_pd(fscal,dy11);
1147             tz               = _mm_mul_pd(fscal,dz11);
1148
1149             /* Update vectorial force */
1150             fix1             = _mm_add_pd(fix1,tx);
1151             fiy1             = _mm_add_pd(fiy1,ty);
1152             fiz1             = _mm_add_pd(fiz1,tz);
1153
1154             fjx1             = _mm_add_pd(fjx1,tx);
1155             fjy1             = _mm_add_pd(fjy1,ty);
1156             fjz1             = _mm_add_pd(fjz1,tz);
1157
1158             }
1159
1160             /**************************
1161              * CALCULATE INTERACTIONS *
1162              **************************/
1163
1164             if (gmx_mm_any_lt(rsq12,rcutoff2))
1165             {
1166
1167             r12              = _mm_mul_pd(rsq12,rinv12);
1168
1169             /* EWALD ELECTROSTATICS */
1170
1171             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1172             ewrt             = _mm_mul_pd(r12,ewtabscale);
1173             ewitab           = _mm_cvttpd_epi32(ewrt);
1174             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1175             ewitab           = _mm_slli_epi32(ewitab,2);
1176             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1177             ewtabD           = _mm_setzero_pd();
1178             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1179             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1180             ewtabFn          = _mm_setzero_pd();
1181             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1182             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1183             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1184             velec            = _mm_mul_pd(qq12,_mm_sub_pd(_mm_sub_pd(rinv12,sh_ewald),velec));
1185             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
1186
1187             cutoff_mask      = _mm_cmplt_pd(rsq12,rcutoff2);
1188
1189             /* Update potential sum for this i atom from the interaction with this j atom. */
1190             velec            = _mm_and_pd(velec,cutoff_mask);
1191             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1192             velecsum         = _mm_add_pd(velecsum,velec);
1193
1194             fscal            = felec;
1195
1196             fscal            = _mm_and_pd(fscal,cutoff_mask);
1197
1198             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1199
1200             /* Calculate temporary vectorial force */
1201             tx               = _mm_mul_pd(fscal,dx12);
1202             ty               = _mm_mul_pd(fscal,dy12);
1203             tz               = _mm_mul_pd(fscal,dz12);
1204
1205             /* Update vectorial force */
1206             fix1             = _mm_add_pd(fix1,tx);
1207             fiy1             = _mm_add_pd(fiy1,ty);
1208             fiz1             = _mm_add_pd(fiz1,tz);
1209
1210             fjx2             = _mm_add_pd(fjx2,tx);
1211             fjy2             = _mm_add_pd(fjy2,ty);
1212             fjz2             = _mm_add_pd(fjz2,tz);
1213
1214             }
1215
1216             /**************************
1217              * CALCULATE INTERACTIONS *
1218              **************************/
1219
1220             if (gmx_mm_any_lt(rsq20,rcutoff2))
1221             {
1222
1223             r20              = _mm_mul_pd(rsq20,rinv20);
1224
1225             /* EWALD ELECTROSTATICS */
1226
1227             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1228             ewrt             = _mm_mul_pd(r20,ewtabscale);
1229             ewitab           = _mm_cvttpd_epi32(ewrt);
1230             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1231             ewitab           = _mm_slli_epi32(ewitab,2);
1232             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1233             ewtabD           = _mm_setzero_pd();
1234             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1235             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1236             ewtabFn          = _mm_setzero_pd();
1237             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1238             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1239             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1240             velec            = _mm_mul_pd(qq20,_mm_sub_pd(_mm_sub_pd(rinv20,sh_ewald),velec));
1241             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
1242
1243             cutoff_mask      = _mm_cmplt_pd(rsq20,rcutoff2);
1244
1245             /* Update potential sum for this i atom from the interaction with this j atom. */
1246             velec            = _mm_and_pd(velec,cutoff_mask);
1247             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1248             velecsum         = _mm_add_pd(velecsum,velec);
1249
1250             fscal            = felec;
1251
1252             fscal            = _mm_and_pd(fscal,cutoff_mask);
1253
1254             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1255
1256             /* Calculate temporary vectorial force */
1257             tx               = _mm_mul_pd(fscal,dx20);
1258             ty               = _mm_mul_pd(fscal,dy20);
1259             tz               = _mm_mul_pd(fscal,dz20);
1260
1261             /* Update vectorial force */
1262             fix2             = _mm_add_pd(fix2,tx);
1263             fiy2             = _mm_add_pd(fiy2,ty);
1264             fiz2             = _mm_add_pd(fiz2,tz);
1265
1266             fjx0             = _mm_add_pd(fjx0,tx);
1267             fjy0             = _mm_add_pd(fjy0,ty);
1268             fjz0             = _mm_add_pd(fjz0,tz);
1269
1270             }
1271
1272             /**************************
1273              * CALCULATE INTERACTIONS *
1274              **************************/
1275
1276             if (gmx_mm_any_lt(rsq21,rcutoff2))
1277             {
1278
1279             r21              = _mm_mul_pd(rsq21,rinv21);
1280
1281             /* EWALD ELECTROSTATICS */
1282
1283             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1284             ewrt             = _mm_mul_pd(r21,ewtabscale);
1285             ewitab           = _mm_cvttpd_epi32(ewrt);
1286             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1287             ewitab           = _mm_slli_epi32(ewitab,2);
1288             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1289             ewtabD           = _mm_setzero_pd();
1290             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1291             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1292             ewtabFn          = _mm_setzero_pd();
1293             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1294             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1295             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1296             velec            = _mm_mul_pd(qq21,_mm_sub_pd(_mm_sub_pd(rinv21,sh_ewald),velec));
1297             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
1298
1299             cutoff_mask      = _mm_cmplt_pd(rsq21,rcutoff2);
1300
1301             /* Update potential sum for this i atom from the interaction with this j atom. */
1302             velec            = _mm_and_pd(velec,cutoff_mask);
1303             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1304             velecsum         = _mm_add_pd(velecsum,velec);
1305
1306             fscal            = felec;
1307
1308             fscal            = _mm_and_pd(fscal,cutoff_mask);
1309
1310             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1311
1312             /* Calculate temporary vectorial force */
1313             tx               = _mm_mul_pd(fscal,dx21);
1314             ty               = _mm_mul_pd(fscal,dy21);
1315             tz               = _mm_mul_pd(fscal,dz21);
1316
1317             /* Update vectorial force */
1318             fix2             = _mm_add_pd(fix2,tx);
1319             fiy2             = _mm_add_pd(fiy2,ty);
1320             fiz2             = _mm_add_pd(fiz2,tz);
1321
1322             fjx1             = _mm_add_pd(fjx1,tx);
1323             fjy1             = _mm_add_pd(fjy1,ty);
1324             fjz1             = _mm_add_pd(fjz1,tz);
1325
1326             }
1327
1328             /**************************
1329              * CALCULATE INTERACTIONS *
1330              **************************/
1331
1332             if (gmx_mm_any_lt(rsq22,rcutoff2))
1333             {
1334
1335             r22              = _mm_mul_pd(rsq22,rinv22);
1336
1337             /* EWALD ELECTROSTATICS */
1338
1339             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1340             ewrt             = _mm_mul_pd(r22,ewtabscale);
1341             ewitab           = _mm_cvttpd_epi32(ewrt);
1342             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1343             ewitab           = _mm_slli_epi32(ewitab,2);
1344             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1345             ewtabD           = _mm_setzero_pd();
1346             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1347             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1348             ewtabFn          = _mm_setzero_pd();
1349             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1350             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1351             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1352             velec            = _mm_mul_pd(qq22,_mm_sub_pd(_mm_sub_pd(rinv22,sh_ewald),velec));
1353             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
1354
1355             cutoff_mask      = _mm_cmplt_pd(rsq22,rcutoff2);
1356
1357             /* Update potential sum for this i atom from the interaction with this j atom. */
1358             velec            = _mm_and_pd(velec,cutoff_mask);
1359             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1360             velecsum         = _mm_add_pd(velecsum,velec);
1361
1362             fscal            = felec;
1363
1364             fscal            = _mm_and_pd(fscal,cutoff_mask);
1365
1366             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1367
1368             /* Calculate temporary vectorial force */
1369             tx               = _mm_mul_pd(fscal,dx22);
1370             ty               = _mm_mul_pd(fscal,dy22);
1371             tz               = _mm_mul_pd(fscal,dz22);
1372
1373             /* Update vectorial force */
1374             fix2             = _mm_add_pd(fix2,tx);
1375             fiy2             = _mm_add_pd(fiy2,ty);
1376             fiz2             = _mm_add_pd(fiz2,tz);
1377
1378             fjx2             = _mm_add_pd(fjx2,tx);
1379             fjy2             = _mm_add_pd(fjy2,ty);
1380             fjz2             = _mm_add_pd(fjz2,tz);
1381
1382             }
1383
1384             gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1385
1386             /* Inner loop uses 432 flops */
1387         }
1388
1389         /* End of innermost loop */
1390
1391         gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1392                                               f+i_coord_offset,fshift+i_shift_offset);
1393
1394         ggid                        = gid[iidx];
1395         /* Update potential energies */
1396         gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1397         gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1398
1399         /* Increment number of inner iterations */
1400         inneriter                  += j_index_end - j_index_start;
1401
1402         /* Outer loop uses 20 flops */
1403     }
1404
1405     /* Increment number of outer iterations */
1406     outeriter        += nri;
1407
1408     /* Update outer/inner flops */
1409
1410     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_VF,outeriter*20 + inneriter*432);
1411 }
1412 /*
1413  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSh_VdwLJSh_GeomW3W3_F_sse4_1_double
1414  * Electrostatics interaction: Ewald
1415  * VdW interaction:            LennardJones
1416  * Geometry:                   Water3-Water3
1417  * Calculate force/pot:        Force
1418  */
1419 void
1420 nb_kernel_ElecEwSh_VdwLJSh_GeomW3W3_F_sse4_1_double
1421                     (t_nblist                    * gmx_restrict       nlist,
1422                      rvec                        * gmx_restrict          xx,
1423                      rvec                        * gmx_restrict          ff,
1424                      t_forcerec                  * gmx_restrict          fr,
1425                      t_mdatoms                   * gmx_restrict     mdatoms,
1426                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1427                      t_nrnb                      * gmx_restrict        nrnb)
1428 {
1429     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1430      * just 0 for non-waters.
1431      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
1432      * jnr indices corresponding to data put in the four positions in the SIMD register.
1433      */
1434     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1435     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1436     int              jnrA,jnrB;
1437     int              j_coord_offsetA,j_coord_offsetB;
1438     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1439     real             rcutoff_scalar;
1440     real             *shiftvec,*fshift,*x,*f;
1441     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1442     int              vdwioffset0;
1443     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1444     int              vdwioffset1;
1445     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1446     int              vdwioffset2;
1447     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1448     int              vdwjidx0A,vdwjidx0B;
1449     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1450     int              vdwjidx1A,vdwjidx1B;
1451     __m128d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1452     int              vdwjidx2A,vdwjidx2B;
1453     __m128d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1454     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1455     __m128d          dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1456     __m128d          dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1457     __m128d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1458     __m128d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1459     __m128d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1460     __m128d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1461     __m128d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1462     __m128d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1463     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
1464     real             *charge;
1465     int              nvdwtype;
1466     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1467     int              *vdwtype;
1468     real             *vdwparam;
1469     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
1470     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
1471     __m128i          ewitab;
1472     __m128d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1473     real             *ewtab;
1474     __m128d          dummy_mask,cutoff_mask;
1475     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
1476     __m128d          one     = _mm_set1_pd(1.0);
1477     __m128d          two     = _mm_set1_pd(2.0);
1478     x                = xx[0];
1479     f                = ff[0];
1480
1481     nri              = nlist->nri;
1482     iinr             = nlist->iinr;
1483     jindex           = nlist->jindex;
1484     jjnr             = nlist->jjnr;
1485     shiftidx         = nlist->shift;
1486     gid              = nlist->gid;
1487     shiftvec         = fr->shift_vec[0];
1488     fshift           = fr->fshift[0];
1489     facel            = _mm_set1_pd(fr->epsfac);
1490     charge           = mdatoms->chargeA;
1491     nvdwtype         = fr->ntype;
1492     vdwparam         = fr->nbfp;
1493     vdwtype          = mdatoms->typeA;
1494
1495     sh_ewald         = _mm_set1_pd(fr->ic->sh_ewald);
1496     ewtab            = fr->ic->tabq_coul_F;
1497     ewtabscale       = _mm_set1_pd(fr->ic->tabq_scale);
1498     ewtabhalfspace   = _mm_set1_pd(0.5/fr->ic->tabq_scale);
1499
1500     /* Setup water-specific parameters */
1501     inr              = nlist->iinr[0];
1502     iq0              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
1503     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
1504     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
1505     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
1506
1507     jq0              = _mm_set1_pd(charge[inr+0]);
1508     jq1              = _mm_set1_pd(charge[inr+1]);
1509     jq2              = _mm_set1_pd(charge[inr+2]);
1510     vdwjidx0A        = 2*vdwtype[inr+0];
1511     qq00             = _mm_mul_pd(iq0,jq0);
1512     c6_00            = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
1513     c12_00           = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
1514     qq01             = _mm_mul_pd(iq0,jq1);
1515     qq02             = _mm_mul_pd(iq0,jq2);
1516     qq10             = _mm_mul_pd(iq1,jq0);
1517     qq11             = _mm_mul_pd(iq1,jq1);
1518     qq12             = _mm_mul_pd(iq1,jq2);
1519     qq20             = _mm_mul_pd(iq2,jq0);
1520     qq21             = _mm_mul_pd(iq2,jq1);
1521     qq22             = _mm_mul_pd(iq2,jq2);
1522
1523     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
1524     rcutoff_scalar   = fr->rcoulomb;
1525     rcutoff          = _mm_set1_pd(rcutoff_scalar);
1526     rcutoff2         = _mm_mul_pd(rcutoff,rcutoff);
1527
1528     sh_vdw_invrcut6  = _mm_set1_pd(fr->ic->sh_invrc6);
1529     rvdw             = _mm_set1_pd(fr->rvdw);
1530
1531     /* Avoid stupid compiler warnings */
1532     jnrA = jnrB = 0;
1533     j_coord_offsetA = 0;
1534     j_coord_offsetB = 0;
1535
1536     outeriter        = 0;
1537     inneriter        = 0;
1538
1539     /* Start outer loop over neighborlists */
1540     for(iidx=0; iidx<nri; iidx++)
1541     {
1542         /* Load shift vector for this list */
1543         i_shift_offset   = DIM*shiftidx[iidx];
1544
1545         /* Load limits for loop over neighbors */
1546         j_index_start    = jindex[iidx];
1547         j_index_end      = jindex[iidx+1];
1548
1549         /* Get outer coordinate index */
1550         inr              = iinr[iidx];
1551         i_coord_offset   = DIM*inr;
1552
1553         /* Load i particle coords and add shift vector */
1554         gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1555                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1556
1557         fix0             = _mm_setzero_pd();
1558         fiy0             = _mm_setzero_pd();
1559         fiz0             = _mm_setzero_pd();
1560         fix1             = _mm_setzero_pd();
1561         fiy1             = _mm_setzero_pd();
1562         fiz1             = _mm_setzero_pd();
1563         fix2             = _mm_setzero_pd();
1564         fiy2             = _mm_setzero_pd();
1565         fiz2             = _mm_setzero_pd();
1566
1567         /* Start inner kernel loop */
1568         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
1569         {
1570
1571             /* Get j neighbor index, and coordinate index */
1572             jnrA             = jjnr[jidx];
1573             jnrB             = jjnr[jidx+1];
1574             j_coord_offsetA  = DIM*jnrA;
1575             j_coord_offsetB  = DIM*jnrB;
1576
1577             /* load j atom coordinates */
1578             gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1579                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1580
1581             /* Calculate displacement vector */
1582             dx00             = _mm_sub_pd(ix0,jx0);
1583             dy00             = _mm_sub_pd(iy0,jy0);
1584             dz00             = _mm_sub_pd(iz0,jz0);
1585             dx01             = _mm_sub_pd(ix0,jx1);
1586             dy01             = _mm_sub_pd(iy0,jy1);
1587             dz01             = _mm_sub_pd(iz0,jz1);
1588             dx02             = _mm_sub_pd(ix0,jx2);
1589             dy02             = _mm_sub_pd(iy0,jy2);
1590             dz02             = _mm_sub_pd(iz0,jz2);
1591             dx10             = _mm_sub_pd(ix1,jx0);
1592             dy10             = _mm_sub_pd(iy1,jy0);
1593             dz10             = _mm_sub_pd(iz1,jz0);
1594             dx11             = _mm_sub_pd(ix1,jx1);
1595             dy11             = _mm_sub_pd(iy1,jy1);
1596             dz11             = _mm_sub_pd(iz1,jz1);
1597             dx12             = _mm_sub_pd(ix1,jx2);
1598             dy12             = _mm_sub_pd(iy1,jy2);
1599             dz12             = _mm_sub_pd(iz1,jz2);
1600             dx20             = _mm_sub_pd(ix2,jx0);
1601             dy20             = _mm_sub_pd(iy2,jy0);
1602             dz20             = _mm_sub_pd(iz2,jz0);
1603             dx21             = _mm_sub_pd(ix2,jx1);
1604             dy21             = _mm_sub_pd(iy2,jy1);
1605             dz21             = _mm_sub_pd(iz2,jz1);
1606             dx22             = _mm_sub_pd(ix2,jx2);
1607             dy22             = _mm_sub_pd(iy2,jy2);
1608             dz22             = _mm_sub_pd(iz2,jz2);
1609
1610             /* Calculate squared distance and things based on it */
1611             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1612             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
1613             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
1614             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
1615             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1616             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1617             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
1618             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1619             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1620
1621             rinv00           = gmx_mm_invsqrt_pd(rsq00);
1622             rinv01           = gmx_mm_invsqrt_pd(rsq01);
1623             rinv02           = gmx_mm_invsqrt_pd(rsq02);
1624             rinv10           = gmx_mm_invsqrt_pd(rsq10);
1625             rinv11           = gmx_mm_invsqrt_pd(rsq11);
1626             rinv12           = gmx_mm_invsqrt_pd(rsq12);
1627             rinv20           = gmx_mm_invsqrt_pd(rsq20);
1628             rinv21           = gmx_mm_invsqrt_pd(rsq21);
1629             rinv22           = gmx_mm_invsqrt_pd(rsq22);
1630
1631             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
1632             rinvsq01         = _mm_mul_pd(rinv01,rinv01);
1633             rinvsq02         = _mm_mul_pd(rinv02,rinv02);
1634             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
1635             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
1636             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
1637             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
1638             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
1639             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
1640
1641             fjx0             = _mm_setzero_pd();
1642             fjy0             = _mm_setzero_pd();
1643             fjz0             = _mm_setzero_pd();
1644             fjx1             = _mm_setzero_pd();
1645             fjy1             = _mm_setzero_pd();
1646             fjz1             = _mm_setzero_pd();
1647             fjx2             = _mm_setzero_pd();
1648             fjy2             = _mm_setzero_pd();
1649             fjz2             = _mm_setzero_pd();
1650
1651             /**************************
1652              * CALCULATE INTERACTIONS *
1653              **************************/
1654
1655             if (gmx_mm_any_lt(rsq00,rcutoff2))
1656             {
1657
1658             r00              = _mm_mul_pd(rsq00,rinv00);
1659
1660             /* EWALD ELECTROSTATICS */
1661
1662             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1663             ewrt             = _mm_mul_pd(r00,ewtabscale);
1664             ewitab           = _mm_cvttpd_epi32(ewrt);
1665             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1666             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1667                                          &ewtabF,&ewtabFn);
1668             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1669             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
1670
1671             /* LENNARD-JONES DISPERSION/REPULSION */
1672
1673             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1674             fvdw             = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
1675
1676             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
1677
1678             fscal            = _mm_add_pd(felec,fvdw);
1679
1680             fscal            = _mm_and_pd(fscal,cutoff_mask);
1681
1682             /* Calculate temporary vectorial force */
1683             tx               = _mm_mul_pd(fscal,dx00);
1684             ty               = _mm_mul_pd(fscal,dy00);
1685             tz               = _mm_mul_pd(fscal,dz00);
1686
1687             /* Update vectorial force */
1688             fix0             = _mm_add_pd(fix0,tx);
1689             fiy0             = _mm_add_pd(fiy0,ty);
1690             fiz0             = _mm_add_pd(fiz0,tz);
1691
1692             fjx0             = _mm_add_pd(fjx0,tx);
1693             fjy0             = _mm_add_pd(fjy0,ty);
1694             fjz0             = _mm_add_pd(fjz0,tz);
1695
1696             }
1697
1698             /**************************
1699              * CALCULATE INTERACTIONS *
1700              **************************/
1701
1702             if (gmx_mm_any_lt(rsq01,rcutoff2))
1703             {
1704
1705             r01              = _mm_mul_pd(rsq01,rinv01);
1706
1707             /* EWALD ELECTROSTATICS */
1708
1709             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1710             ewrt             = _mm_mul_pd(r01,ewtabscale);
1711             ewitab           = _mm_cvttpd_epi32(ewrt);
1712             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1713             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1714                                          &ewtabF,&ewtabFn);
1715             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1716             felec            = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
1717
1718             cutoff_mask      = _mm_cmplt_pd(rsq01,rcutoff2);
1719
1720             fscal            = felec;
1721
1722             fscal            = _mm_and_pd(fscal,cutoff_mask);
1723
1724             /* Calculate temporary vectorial force */
1725             tx               = _mm_mul_pd(fscal,dx01);
1726             ty               = _mm_mul_pd(fscal,dy01);
1727             tz               = _mm_mul_pd(fscal,dz01);
1728
1729             /* Update vectorial force */
1730             fix0             = _mm_add_pd(fix0,tx);
1731             fiy0             = _mm_add_pd(fiy0,ty);
1732             fiz0             = _mm_add_pd(fiz0,tz);
1733
1734             fjx1             = _mm_add_pd(fjx1,tx);
1735             fjy1             = _mm_add_pd(fjy1,ty);
1736             fjz1             = _mm_add_pd(fjz1,tz);
1737
1738             }
1739
1740             /**************************
1741              * CALCULATE INTERACTIONS *
1742              **************************/
1743
1744             if (gmx_mm_any_lt(rsq02,rcutoff2))
1745             {
1746
1747             r02              = _mm_mul_pd(rsq02,rinv02);
1748
1749             /* EWALD ELECTROSTATICS */
1750
1751             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1752             ewrt             = _mm_mul_pd(r02,ewtabscale);
1753             ewitab           = _mm_cvttpd_epi32(ewrt);
1754             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1755             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1756                                          &ewtabF,&ewtabFn);
1757             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1758             felec            = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
1759
1760             cutoff_mask      = _mm_cmplt_pd(rsq02,rcutoff2);
1761
1762             fscal            = felec;
1763
1764             fscal            = _mm_and_pd(fscal,cutoff_mask);
1765
1766             /* Calculate temporary vectorial force */
1767             tx               = _mm_mul_pd(fscal,dx02);
1768             ty               = _mm_mul_pd(fscal,dy02);
1769             tz               = _mm_mul_pd(fscal,dz02);
1770
1771             /* Update vectorial force */
1772             fix0             = _mm_add_pd(fix0,tx);
1773             fiy0             = _mm_add_pd(fiy0,ty);
1774             fiz0             = _mm_add_pd(fiz0,tz);
1775
1776             fjx2             = _mm_add_pd(fjx2,tx);
1777             fjy2             = _mm_add_pd(fjy2,ty);
1778             fjz2             = _mm_add_pd(fjz2,tz);
1779
1780             }
1781
1782             /**************************
1783              * CALCULATE INTERACTIONS *
1784              **************************/
1785
1786             if (gmx_mm_any_lt(rsq10,rcutoff2))
1787             {
1788
1789             r10              = _mm_mul_pd(rsq10,rinv10);
1790
1791             /* EWALD ELECTROSTATICS */
1792
1793             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1794             ewrt             = _mm_mul_pd(r10,ewtabscale);
1795             ewitab           = _mm_cvttpd_epi32(ewrt);
1796             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1797             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1798                                          &ewtabF,&ewtabFn);
1799             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1800             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
1801
1802             cutoff_mask      = _mm_cmplt_pd(rsq10,rcutoff2);
1803
1804             fscal            = felec;
1805
1806             fscal            = _mm_and_pd(fscal,cutoff_mask);
1807
1808             /* Calculate temporary vectorial force */
1809             tx               = _mm_mul_pd(fscal,dx10);
1810             ty               = _mm_mul_pd(fscal,dy10);
1811             tz               = _mm_mul_pd(fscal,dz10);
1812
1813             /* Update vectorial force */
1814             fix1             = _mm_add_pd(fix1,tx);
1815             fiy1             = _mm_add_pd(fiy1,ty);
1816             fiz1             = _mm_add_pd(fiz1,tz);
1817
1818             fjx0             = _mm_add_pd(fjx0,tx);
1819             fjy0             = _mm_add_pd(fjy0,ty);
1820             fjz0             = _mm_add_pd(fjz0,tz);
1821
1822             }
1823
1824             /**************************
1825              * CALCULATE INTERACTIONS *
1826              **************************/
1827
1828             if (gmx_mm_any_lt(rsq11,rcutoff2))
1829             {
1830
1831             r11              = _mm_mul_pd(rsq11,rinv11);
1832
1833             /* EWALD ELECTROSTATICS */
1834
1835             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1836             ewrt             = _mm_mul_pd(r11,ewtabscale);
1837             ewitab           = _mm_cvttpd_epi32(ewrt);
1838             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1839             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1840                                          &ewtabF,&ewtabFn);
1841             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1842             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
1843
1844             cutoff_mask      = _mm_cmplt_pd(rsq11,rcutoff2);
1845
1846             fscal            = felec;
1847
1848             fscal            = _mm_and_pd(fscal,cutoff_mask);
1849
1850             /* Calculate temporary vectorial force */
1851             tx               = _mm_mul_pd(fscal,dx11);
1852             ty               = _mm_mul_pd(fscal,dy11);
1853             tz               = _mm_mul_pd(fscal,dz11);
1854
1855             /* Update vectorial force */
1856             fix1             = _mm_add_pd(fix1,tx);
1857             fiy1             = _mm_add_pd(fiy1,ty);
1858             fiz1             = _mm_add_pd(fiz1,tz);
1859
1860             fjx1             = _mm_add_pd(fjx1,tx);
1861             fjy1             = _mm_add_pd(fjy1,ty);
1862             fjz1             = _mm_add_pd(fjz1,tz);
1863
1864             }
1865
1866             /**************************
1867              * CALCULATE INTERACTIONS *
1868              **************************/
1869
1870             if (gmx_mm_any_lt(rsq12,rcutoff2))
1871             {
1872
1873             r12              = _mm_mul_pd(rsq12,rinv12);
1874
1875             /* EWALD ELECTROSTATICS */
1876
1877             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1878             ewrt             = _mm_mul_pd(r12,ewtabscale);
1879             ewitab           = _mm_cvttpd_epi32(ewrt);
1880             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1881             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1882                                          &ewtabF,&ewtabFn);
1883             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1884             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
1885
1886             cutoff_mask      = _mm_cmplt_pd(rsq12,rcutoff2);
1887
1888             fscal            = felec;
1889
1890             fscal            = _mm_and_pd(fscal,cutoff_mask);
1891
1892             /* Calculate temporary vectorial force */
1893             tx               = _mm_mul_pd(fscal,dx12);
1894             ty               = _mm_mul_pd(fscal,dy12);
1895             tz               = _mm_mul_pd(fscal,dz12);
1896
1897             /* Update vectorial force */
1898             fix1             = _mm_add_pd(fix1,tx);
1899             fiy1             = _mm_add_pd(fiy1,ty);
1900             fiz1             = _mm_add_pd(fiz1,tz);
1901
1902             fjx2             = _mm_add_pd(fjx2,tx);
1903             fjy2             = _mm_add_pd(fjy2,ty);
1904             fjz2             = _mm_add_pd(fjz2,tz);
1905
1906             }
1907
1908             /**************************
1909              * CALCULATE INTERACTIONS *
1910              **************************/
1911
1912             if (gmx_mm_any_lt(rsq20,rcutoff2))
1913             {
1914
1915             r20              = _mm_mul_pd(rsq20,rinv20);
1916
1917             /* EWALD ELECTROSTATICS */
1918
1919             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1920             ewrt             = _mm_mul_pd(r20,ewtabscale);
1921             ewitab           = _mm_cvttpd_epi32(ewrt);
1922             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1923             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1924                                          &ewtabF,&ewtabFn);
1925             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1926             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
1927
1928             cutoff_mask      = _mm_cmplt_pd(rsq20,rcutoff2);
1929
1930             fscal            = felec;
1931
1932             fscal            = _mm_and_pd(fscal,cutoff_mask);
1933
1934             /* Calculate temporary vectorial force */
1935             tx               = _mm_mul_pd(fscal,dx20);
1936             ty               = _mm_mul_pd(fscal,dy20);
1937             tz               = _mm_mul_pd(fscal,dz20);
1938
1939             /* Update vectorial force */
1940             fix2             = _mm_add_pd(fix2,tx);
1941             fiy2             = _mm_add_pd(fiy2,ty);
1942             fiz2             = _mm_add_pd(fiz2,tz);
1943
1944             fjx0             = _mm_add_pd(fjx0,tx);
1945             fjy0             = _mm_add_pd(fjy0,ty);
1946             fjz0             = _mm_add_pd(fjz0,tz);
1947
1948             }
1949
1950             /**************************
1951              * CALCULATE INTERACTIONS *
1952              **************************/
1953
1954             if (gmx_mm_any_lt(rsq21,rcutoff2))
1955             {
1956
1957             r21              = _mm_mul_pd(rsq21,rinv21);
1958
1959             /* EWALD ELECTROSTATICS */
1960
1961             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1962             ewrt             = _mm_mul_pd(r21,ewtabscale);
1963             ewitab           = _mm_cvttpd_epi32(ewrt);
1964             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1965             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1966                                          &ewtabF,&ewtabFn);
1967             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1968             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
1969
1970             cutoff_mask      = _mm_cmplt_pd(rsq21,rcutoff2);
1971
1972             fscal            = felec;
1973
1974             fscal            = _mm_and_pd(fscal,cutoff_mask);
1975
1976             /* Calculate temporary vectorial force */
1977             tx               = _mm_mul_pd(fscal,dx21);
1978             ty               = _mm_mul_pd(fscal,dy21);
1979             tz               = _mm_mul_pd(fscal,dz21);
1980
1981             /* Update vectorial force */
1982             fix2             = _mm_add_pd(fix2,tx);
1983             fiy2             = _mm_add_pd(fiy2,ty);
1984             fiz2             = _mm_add_pd(fiz2,tz);
1985
1986             fjx1             = _mm_add_pd(fjx1,tx);
1987             fjy1             = _mm_add_pd(fjy1,ty);
1988             fjz1             = _mm_add_pd(fjz1,tz);
1989
1990             }
1991
1992             /**************************
1993              * CALCULATE INTERACTIONS *
1994              **************************/
1995
1996             if (gmx_mm_any_lt(rsq22,rcutoff2))
1997             {
1998
1999             r22              = _mm_mul_pd(rsq22,rinv22);
2000
2001             /* EWALD ELECTROSTATICS */
2002
2003             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2004             ewrt             = _mm_mul_pd(r22,ewtabscale);
2005             ewitab           = _mm_cvttpd_epi32(ewrt);
2006             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2007             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
2008                                          &ewtabF,&ewtabFn);
2009             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2010             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
2011
2012             cutoff_mask      = _mm_cmplt_pd(rsq22,rcutoff2);
2013
2014             fscal            = felec;
2015
2016             fscal            = _mm_and_pd(fscal,cutoff_mask);
2017
2018             /* Calculate temporary vectorial force */
2019             tx               = _mm_mul_pd(fscal,dx22);
2020             ty               = _mm_mul_pd(fscal,dy22);
2021             tz               = _mm_mul_pd(fscal,dz22);
2022
2023             /* Update vectorial force */
2024             fix2             = _mm_add_pd(fix2,tx);
2025             fiy2             = _mm_add_pd(fiy2,ty);
2026             fiz2             = _mm_add_pd(fiz2,tz);
2027
2028             fjx2             = _mm_add_pd(fjx2,tx);
2029             fjy2             = _mm_add_pd(fjy2,ty);
2030             fjz2             = _mm_add_pd(fjz2,tz);
2031
2032             }
2033
2034             gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
2035
2036             /* Inner loop uses 358 flops */
2037         }
2038
2039         if(jidx<j_index_end)
2040         {
2041
2042             jnrA             = jjnr[jidx];
2043             j_coord_offsetA  = DIM*jnrA;
2044
2045             /* load j atom coordinates */
2046             gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
2047                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
2048
2049             /* Calculate displacement vector */
2050             dx00             = _mm_sub_pd(ix0,jx0);
2051             dy00             = _mm_sub_pd(iy0,jy0);
2052             dz00             = _mm_sub_pd(iz0,jz0);
2053             dx01             = _mm_sub_pd(ix0,jx1);
2054             dy01             = _mm_sub_pd(iy0,jy1);
2055             dz01             = _mm_sub_pd(iz0,jz1);
2056             dx02             = _mm_sub_pd(ix0,jx2);
2057             dy02             = _mm_sub_pd(iy0,jy2);
2058             dz02             = _mm_sub_pd(iz0,jz2);
2059             dx10             = _mm_sub_pd(ix1,jx0);
2060             dy10             = _mm_sub_pd(iy1,jy0);
2061             dz10             = _mm_sub_pd(iz1,jz0);
2062             dx11             = _mm_sub_pd(ix1,jx1);
2063             dy11             = _mm_sub_pd(iy1,jy1);
2064             dz11             = _mm_sub_pd(iz1,jz1);
2065             dx12             = _mm_sub_pd(ix1,jx2);
2066             dy12             = _mm_sub_pd(iy1,jy2);
2067             dz12             = _mm_sub_pd(iz1,jz2);
2068             dx20             = _mm_sub_pd(ix2,jx0);
2069             dy20             = _mm_sub_pd(iy2,jy0);
2070             dz20             = _mm_sub_pd(iz2,jz0);
2071             dx21             = _mm_sub_pd(ix2,jx1);
2072             dy21             = _mm_sub_pd(iy2,jy1);
2073             dz21             = _mm_sub_pd(iz2,jz1);
2074             dx22             = _mm_sub_pd(ix2,jx2);
2075             dy22             = _mm_sub_pd(iy2,jy2);
2076             dz22             = _mm_sub_pd(iz2,jz2);
2077
2078             /* Calculate squared distance and things based on it */
2079             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
2080             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
2081             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
2082             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
2083             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
2084             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
2085             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
2086             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
2087             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
2088
2089             rinv00           = gmx_mm_invsqrt_pd(rsq00);
2090             rinv01           = gmx_mm_invsqrt_pd(rsq01);
2091             rinv02           = gmx_mm_invsqrt_pd(rsq02);
2092             rinv10           = gmx_mm_invsqrt_pd(rsq10);
2093             rinv11           = gmx_mm_invsqrt_pd(rsq11);
2094             rinv12           = gmx_mm_invsqrt_pd(rsq12);
2095             rinv20           = gmx_mm_invsqrt_pd(rsq20);
2096             rinv21           = gmx_mm_invsqrt_pd(rsq21);
2097             rinv22           = gmx_mm_invsqrt_pd(rsq22);
2098
2099             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
2100             rinvsq01         = _mm_mul_pd(rinv01,rinv01);
2101             rinvsq02         = _mm_mul_pd(rinv02,rinv02);
2102             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
2103             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
2104             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
2105             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
2106             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
2107             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
2108
2109             fjx0             = _mm_setzero_pd();
2110             fjy0             = _mm_setzero_pd();
2111             fjz0             = _mm_setzero_pd();
2112             fjx1             = _mm_setzero_pd();
2113             fjy1             = _mm_setzero_pd();
2114             fjz1             = _mm_setzero_pd();
2115             fjx2             = _mm_setzero_pd();
2116             fjy2             = _mm_setzero_pd();
2117             fjz2             = _mm_setzero_pd();
2118
2119             /**************************
2120              * CALCULATE INTERACTIONS *
2121              **************************/
2122
2123             if (gmx_mm_any_lt(rsq00,rcutoff2))
2124             {
2125
2126             r00              = _mm_mul_pd(rsq00,rinv00);
2127
2128             /* EWALD ELECTROSTATICS */
2129
2130             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2131             ewrt             = _mm_mul_pd(r00,ewtabscale);
2132             ewitab           = _mm_cvttpd_epi32(ewrt);
2133             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2134             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2135             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2136             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
2137
2138             /* LENNARD-JONES DISPERSION/REPULSION */
2139
2140             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
2141             fvdw             = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
2142
2143             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
2144
2145             fscal            = _mm_add_pd(felec,fvdw);
2146
2147             fscal            = _mm_and_pd(fscal,cutoff_mask);
2148
2149             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2150
2151             /* Calculate temporary vectorial force */
2152             tx               = _mm_mul_pd(fscal,dx00);
2153             ty               = _mm_mul_pd(fscal,dy00);
2154             tz               = _mm_mul_pd(fscal,dz00);
2155
2156             /* Update vectorial force */
2157             fix0             = _mm_add_pd(fix0,tx);
2158             fiy0             = _mm_add_pd(fiy0,ty);
2159             fiz0             = _mm_add_pd(fiz0,tz);
2160
2161             fjx0             = _mm_add_pd(fjx0,tx);
2162             fjy0             = _mm_add_pd(fjy0,ty);
2163             fjz0             = _mm_add_pd(fjz0,tz);
2164
2165             }
2166
2167             /**************************
2168              * CALCULATE INTERACTIONS *
2169              **************************/
2170
2171             if (gmx_mm_any_lt(rsq01,rcutoff2))
2172             {
2173
2174             r01              = _mm_mul_pd(rsq01,rinv01);
2175
2176             /* EWALD ELECTROSTATICS */
2177
2178             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2179             ewrt             = _mm_mul_pd(r01,ewtabscale);
2180             ewitab           = _mm_cvttpd_epi32(ewrt);
2181             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2182             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2183             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2184             felec            = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
2185
2186             cutoff_mask      = _mm_cmplt_pd(rsq01,rcutoff2);
2187
2188             fscal            = felec;
2189
2190             fscal            = _mm_and_pd(fscal,cutoff_mask);
2191
2192             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2193
2194             /* Calculate temporary vectorial force */
2195             tx               = _mm_mul_pd(fscal,dx01);
2196             ty               = _mm_mul_pd(fscal,dy01);
2197             tz               = _mm_mul_pd(fscal,dz01);
2198
2199             /* Update vectorial force */
2200             fix0             = _mm_add_pd(fix0,tx);
2201             fiy0             = _mm_add_pd(fiy0,ty);
2202             fiz0             = _mm_add_pd(fiz0,tz);
2203
2204             fjx1             = _mm_add_pd(fjx1,tx);
2205             fjy1             = _mm_add_pd(fjy1,ty);
2206             fjz1             = _mm_add_pd(fjz1,tz);
2207
2208             }
2209
2210             /**************************
2211              * CALCULATE INTERACTIONS *
2212              **************************/
2213
2214             if (gmx_mm_any_lt(rsq02,rcutoff2))
2215             {
2216
2217             r02              = _mm_mul_pd(rsq02,rinv02);
2218
2219             /* EWALD ELECTROSTATICS */
2220
2221             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2222             ewrt             = _mm_mul_pd(r02,ewtabscale);
2223             ewitab           = _mm_cvttpd_epi32(ewrt);
2224             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2225             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2226             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2227             felec            = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
2228
2229             cutoff_mask      = _mm_cmplt_pd(rsq02,rcutoff2);
2230
2231             fscal            = felec;
2232
2233             fscal            = _mm_and_pd(fscal,cutoff_mask);
2234
2235             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2236
2237             /* Calculate temporary vectorial force */
2238             tx               = _mm_mul_pd(fscal,dx02);
2239             ty               = _mm_mul_pd(fscal,dy02);
2240             tz               = _mm_mul_pd(fscal,dz02);
2241
2242             /* Update vectorial force */
2243             fix0             = _mm_add_pd(fix0,tx);
2244             fiy0             = _mm_add_pd(fiy0,ty);
2245             fiz0             = _mm_add_pd(fiz0,tz);
2246
2247             fjx2             = _mm_add_pd(fjx2,tx);
2248             fjy2             = _mm_add_pd(fjy2,ty);
2249             fjz2             = _mm_add_pd(fjz2,tz);
2250
2251             }
2252
2253             /**************************
2254              * CALCULATE INTERACTIONS *
2255              **************************/
2256
2257             if (gmx_mm_any_lt(rsq10,rcutoff2))
2258             {
2259
2260             r10              = _mm_mul_pd(rsq10,rinv10);
2261
2262             /* EWALD ELECTROSTATICS */
2263
2264             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2265             ewrt             = _mm_mul_pd(r10,ewtabscale);
2266             ewitab           = _mm_cvttpd_epi32(ewrt);
2267             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2268             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2269             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2270             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
2271
2272             cutoff_mask      = _mm_cmplt_pd(rsq10,rcutoff2);
2273
2274             fscal            = felec;
2275
2276             fscal            = _mm_and_pd(fscal,cutoff_mask);
2277
2278             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2279
2280             /* Calculate temporary vectorial force */
2281             tx               = _mm_mul_pd(fscal,dx10);
2282             ty               = _mm_mul_pd(fscal,dy10);
2283             tz               = _mm_mul_pd(fscal,dz10);
2284
2285             /* Update vectorial force */
2286             fix1             = _mm_add_pd(fix1,tx);
2287             fiy1             = _mm_add_pd(fiy1,ty);
2288             fiz1             = _mm_add_pd(fiz1,tz);
2289
2290             fjx0             = _mm_add_pd(fjx0,tx);
2291             fjy0             = _mm_add_pd(fjy0,ty);
2292             fjz0             = _mm_add_pd(fjz0,tz);
2293
2294             }
2295
2296             /**************************
2297              * CALCULATE INTERACTIONS *
2298              **************************/
2299
2300             if (gmx_mm_any_lt(rsq11,rcutoff2))
2301             {
2302
2303             r11              = _mm_mul_pd(rsq11,rinv11);
2304
2305             /* EWALD ELECTROSTATICS */
2306
2307             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2308             ewrt             = _mm_mul_pd(r11,ewtabscale);
2309             ewitab           = _mm_cvttpd_epi32(ewrt);
2310             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2311             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2312             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2313             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
2314
2315             cutoff_mask      = _mm_cmplt_pd(rsq11,rcutoff2);
2316
2317             fscal            = felec;
2318
2319             fscal            = _mm_and_pd(fscal,cutoff_mask);
2320
2321             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2322
2323             /* Calculate temporary vectorial force */
2324             tx               = _mm_mul_pd(fscal,dx11);
2325             ty               = _mm_mul_pd(fscal,dy11);
2326             tz               = _mm_mul_pd(fscal,dz11);
2327
2328             /* Update vectorial force */
2329             fix1             = _mm_add_pd(fix1,tx);
2330             fiy1             = _mm_add_pd(fiy1,ty);
2331             fiz1             = _mm_add_pd(fiz1,tz);
2332
2333             fjx1             = _mm_add_pd(fjx1,tx);
2334             fjy1             = _mm_add_pd(fjy1,ty);
2335             fjz1             = _mm_add_pd(fjz1,tz);
2336
2337             }
2338
2339             /**************************
2340              * CALCULATE INTERACTIONS *
2341              **************************/
2342
2343             if (gmx_mm_any_lt(rsq12,rcutoff2))
2344             {
2345
2346             r12              = _mm_mul_pd(rsq12,rinv12);
2347
2348             /* EWALD ELECTROSTATICS */
2349
2350             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2351             ewrt             = _mm_mul_pd(r12,ewtabscale);
2352             ewitab           = _mm_cvttpd_epi32(ewrt);
2353             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2354             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2355             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2356             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
2357
2358             cutoff_mask      = _mm_cmplt_pd(rsq12,rcutoff2);
2359
2360             fscal            = felec;
2361
2362             fscal            = _mm_and_pd(fscal,cutoff_mask);
2363
2364             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2365
2366             /* Calculate temporary vectorial force */
2367             tx               = _mm_mul_pd(fscal,dx12);
2368             ty               = _mm_mul_pd(fscal,dy12);
2369             tz               = _mm_mul_pd(fscal,dz12);
2370
2371             /* Update vectorial force */
2372             fix1             = _mm_add_pd(fix1,tx);
2373             fiy1             = _mm_add_pd(fiy1,ty);
2374             fiz1             = _mm_add_pd(fiz1,tz);
2375
2376             fjx2             = _mm_add_pd(fjx2,tx);
2377             fjy2             = _mm_add_pd(fjy2,ty);
2378             fjz2             = _mm_add_pd(fjz2,tz);
2379
2380             }
2381
2382             /**************************
2383              * CALCULATE INTERACTIONS *
2384              **************************/
2385
2386             if (gmx_mm_any_lt(rsq20,rcutoff2))
2387             {
2388
2389             r20              = _mm_mul_pd(rsq20,rinv20);
2390
2391             /* EWALD ELECTROSTATICS */
2392
2393             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2394             ewrt             = _mm_mul_pd(r20,ewtabscale);
2395             ewitab           = _mm_cvttpd_epi32(ewrt);
2396             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2397             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2398             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2399             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
2400
2401             cutoff_mask      = _mm_cmplt_pd(rsq20,rcutoff2);
2402
2403             fscal            = felec;
2404
2405             fscal            = _mm_and_pd(fscal,cutoff_mask);
2406
2407             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2408
2409             /* Calculate temporary vectorial force */
2410             tx               = _mm_mul_pd(fscal,dx20);
2411             ty               = _mm_mul_pd(fscal,dy20);
2412             tz               = _mm_mul_pd(fscal,dz20);
2413
2414             /* Update vectorial force */
2415             fix2             = _mm_add_pd(fix2,tx);
2416             fiy2             = _mm_add_pd(fiy2,ty);
2417             fiz2             = _mm_add_pd(fiz2,tz);
2418
2419             fjx0             = _mm_add_pd(fjx0,tx);
2420             fjy0             = _mm_add_pd(fjy0,ty);
2421             fjz0             = _mm_add_pd(fjz0,tz);
2422
2423             }
2424
2425             /**************************
2426              * CALCULATE INTERACTIONS *
2427              **************************/
2428
2429             if (gmx_mm_any_lt(rsq21,rcutoff2))
2430             {
2431
2432             r21              = _mm_mul_pd(rsq21,rinv21);
2433
2434             /* EWALD ELECTROSTATICS */
2435
2436             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2437             ewrt             = _mm_mul_pd(r21,ewtabscale);
2438             ewitab           = _mm_cvttpd_epi32(ewrt);
2439             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2440             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2441             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2442             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
2443
2444             cutoff_mask      = _mm_cmplt_pd(rsq21,rcutoff2);
2445
2446             fscal            = felec;
2447
2448             fscal            = _mm_and_pd(fscal,cutoff_mask);
2449
2450             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2451
2452             /* Calculate temporary vectorial force */
2453             tx               = _mm_mul_pd(fscal,dx21);
2454             ty               = _mm_mul_pd(fscal,dy21);
2455             tz               = _mm_mul_pd(fscal,dz21);
2456
2457             /* Update vectorial force */
2458             fix2             = _mm_add_pd(fix2,tx);
2459             fiy2             = _mm_add_pd(fiy2,ty);
2460             fiz2             = _mm_add_pd(fiz2,tz);
2461
2462             fjx1             = _mm_add_pd(fjx1,tx);
2463             fjy1             = _mm_add_pd(fjy1,ty);
2464             fjz1             = _mm_add_pd(fjz1,tz);
2465
2466             }
2467
2468             /**************************
2469              * CALCULATE INTERACTIONS *
2470              **************************/
2471
2472             if (gmx_mm_any_lt(rsq22,rcutoff2))
2473             {
2474
2475             r22              = _mm_mul_pd(rsq22,rinv22);
2476
2477             /* EWALD ELECTROSTATICS */
2478
2479             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2480             ewrt             = _mm_mul_pd(r22,ewtabscale);
2481             ewitab           = _mm_cvttpd_epi32(ewrt);
2482             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2483             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2484             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2485             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
2486
2487             cutoff_mask      = _mm_cmplt_pd(rsq22,rcutoff2);
2488
2489             fscal            = felec;
2490
2491             fscal            = _mm_and_pd(fscal,cutoff_mask);
2492
2493             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2494
2495             /* Calculate temporary vectorial force */
2496             tx               = _mm_mul_pd(fscal,dx22);
2497             ty               = _mm_mul_pd(fscal,dy22);
2498             tz               = _mm_mul_pd(fscal,dz22);
2499
2500             /* Update vectorial force */
2501             fix2             = _mm_add_pd(fix2,tx);
2502             fiy2             = _mm_add_pd(fiy2,ty);
2503             fiz2             = _mm_add_pd(fiz2,tz);
2504
2505             fjx2             = _mm_add_pd(fjx2,tx);
2506             fjy2             = _mm_add_pd(fjy2,ty);
2507             fjz2             = _mm_add_pd(fjz2,tz);
2508
2509             }
2510
2511             gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
2512
2513             /* Inner loop uses 358 flops */
2514         }
2515
2516         /* End of innermost loop */
2517
2518         gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
2519                                               f+i_coord_offset,fshift+i_shift_offset);
2520
2521         /* Increment number of inner iterations */
2522         inneriter                  += j_index_end - j_index_start;
2523
2524         /* Outer loop uses 18 flops */
2525     }
2526
2527     /* Increment number of outer iterations */
2528     outeriter        += nri;
2529
2530     /* Update outer/inner flops */
2531
2532     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_F,outeriter*18 + inneriter*358);
2533 }