src/gromacs/gmxlib/nonbonded/nb_kernel_avx_128_fma_double/nb_kernel_ElecRFCut_VdwCSTab_GeomW4W4_avx_128_fma_double.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS avx_128_fma_double kernel generator.
  37  */
  38 #include "gmxpre.h"
  39
  40 #include "config.h"
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "gromacs/legacyheaders/types/simple.h"
  46 #include "gromacs/math/vec.h"
  47 #include "gromacs/legacyheaders/nrnb.h"
  48
  49 #include "gromacs/simd/math_x86_avx_128_fma_double.h"
  50 #include "kernelutil_x86_avx_128_fma_double.h"
  51
  52 /*
  53  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwCSTab_GeomW4W4_VF_avx_128_fma_double
  54  * Electrostatics interaction: ReactionField
  55  * VdW interaction:            CubicSplineTable
  56  * Geometry:                   Water4-Water4
  57  * Calculate force/pot:        PotentialAndForce
  58  */
  59 void
  60 nb_kernel_ElecRFCut_VdwCSTab_GeomW4W4_VF_avx_128_fma_double
  61                     (t_nblist                    * gmx_restrict       nlist,
  62                      rvec                        * gmx_restrict          xx,
  63                      rvec                        * gmx_restrict          ff,
  64                      t_forcerec                  * gmx_restrict          fr,
  65                      t_mdatoms                   * gmx_restrict     mdatoms,
  66                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  67                      t_nrnb                      * gmx_restrict        nrnb)
  68 {
  69     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  70      * just 0 for non-waters.
  71      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
  72      * jnr indices corresponding to data put in the four positions in the SIMD register.
  73      */
  74     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  75     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  76     int              jnrA,jnrB;
  77     int              j_coord_offsetA,j_coord_offsetB;
  78     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  79     real             rcutoff_scalar;
  80     real             *shiftvec,*fshift,*x,*f;
  81     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  82     int              vdwioffset0;
  83     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  84     int              vdwioffset1;
  85     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  86     int              vdwioffset2;
  87     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  88     int              vdwioffset3;
  89     __m128d          ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  90     int              vdwjidx0A,vdwjidx0B;
  91     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  92     int              vdwjidx1A,vdwjidx1B;
  93     __m128d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  94     int              vdwjidx2A,vdwjidx2B;
  95     __m128d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  96     int              vdwjidx3A,vdwjidx3B;
  97     __m128d          jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
  98     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  99     __m128d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
 100     __m128d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
 101     __m128d          dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
 102     __m128d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 103     __m128d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 104     __m128d          dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
 105     __m128d          dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
 106     __m128d          dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
 107     __m128d          dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
 108     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
 109     real             *charge;
 110     int              nvdwtype;
 111     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 112     int              *vdwtype;
 113     real             *vdwparam;
 114     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
 115     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
 116     __m128i          vfitab;
 117     __m128i          ifour       = _mm_set1_epi32(4);
 118     __m128d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 119     real             *vftab;
 120     __m128d          dummy_mask,cutoff_mask;
 121     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 122     __m128d          one     = _mm_set1_pd(1.0);
 123     __m128d          two     = _mm_set1_pd(2.0);
 124     x                = xx[0];
 125     f                = ff[0];
 126
 127     nri              = nlist->nri;
 128     iinr             = nlist->iinr;
 129     jindex           = nlist->jindex;
 130     jjnr             = nlist->jjnr;
 131     shiftidx         = nlist->shift;
 132     gid              = nlist->gid;
 133     shiftvec         = fr->shift_vec[0];
 134     fshift           = fr->fshift[0];
 135     facel            = _mm_set1_pd(fr->epsfac);
 136     charge           = mdatoms->chargeA;
 137     krf              = _mm_set1_pd(fr->ic->k_rf);
 138     krf2             = _mm_set1_pd(fr->ic->k_rf*2.0);
 139     crf              = _mm_set1_pd(fr->ic->c_rf);
 140     nvdwtype         = fr->ntype;
 141     vdwparam         = fr->nbfp;
 142     vdwtype          = mdatoms->typeA;
 143
 144     vftab            = kernel_data->table_vdw->data;
 145     vftabscale       = _mm_set1_pd(kernel_data->table_vdw->scale);
 146
 147     /* Setup water-specific parameters */
 148     inr              = nlist->iinr[0];
 149     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
 150     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
 151     iq3              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
 152     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 153
 154     jq1              = _mm_set1_pd(charge[inr+1]);
 155     jq2              = _mm_set1_pd(charge[inr+2]);
 156     jq3              = _mm_set1_pd(charge[inr+3]);
 157     vdwjidx0A        = 2*vdwtype[inr+0];
 158     c6_00            = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
 159     c12_00           = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
 160     qq11             = _mm_mul_pd(iq1,jq1);
 161     qq12             = _mm_mul_pd(iq1,jq2);
 162     qq13             = _mm_mul_pd(iq1,jq3);
 163     qq21             = _mm_mul_pd(iq2,jq1);
 164     qq22             = _mm_mul_pd(iq2,jq2);
 165     qq23             = _mm_mul_pd(iq2,jq3);
 166     qq31             = _mm_mul_pd(iq3,jq1);
 167     qq32             = _mm_mul_pd(iq3,jq2);
 168     qq33             = _mm_mul_pd(iq3,jq3);
 169
 170     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 171     rcutoff_scalar   = fr->rcoulomb;
 172     rcutoff          = _mm_set1_pd(rcutoff_scalar);
 173     rcutoff2         = _mm_mul_pd(rcutoff,rcutoff);
 174
 175     /* Avoid stupid compiler warnings */
 176     jnrA = jnrB = 0;
 177     j_coord_offsetA = 0;
 178     j_coord_offsetB = 0;
 179
 180     outeriter        = 0;
 181     inneriter        = 0;
 182
 183     /* Start outer loop over neighborlists */
 184     for(iidx=0; iidx<nri; iidx++)
 185     {
 186         /* Load shift vector for this list */
 187         i_shift_offset   = DIM*shiftidx[iidx];
 188
 189         /* Load limits for loop over neighbors */
 190         j_index_start    = jindex[iidx];
 191         j_index_end      = jindex[iidx+1];
 192
 193         /* Get outer coordinate index */
 194         inr              = iinr[iidx];
 195         i_coord_offset   = DIM*inr;
 196
 197         /* Load i particle coords and add shift vector */
 198         gmx_mm_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 199                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 200
 201         fix0             = _mm_setzero_pd();
 202         fiy0             = _mm_setzero_pd();
 203         fiz0             = _mm_setzero_pd();
 204         fix1             = _mm_setzero_pd();
 205         fiy1             = _mm_setzero_pd();
 206         fiz1             = _mm_setzero_pd();
 207         fix2             = _mm_setzero_pd();
 208         fiy2             = _mm_setzero_pd();
 209         fiz2             = _mm_setzero_pd();
 210         fix3             = _mm_setzero_pd();
 211         fiy3             = _mm_setzero_pd();
 212         fiz3             = _mm_setzero_pd();
 213
 214         /* Reset potential sums */
 215         velecsum         = _mm_setzero_pd();
 216         vvdwsum          = _mm_setzero_pd();
 217
 218         /* Start inner kernel loop */
 219         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 220         {
 221
 222             /* Get j neighbor index, and coordinate index */
 223             jnrA             = jjnr[jidx];
 224             jnrB             = jjnr[jidx+1];
 225             j_coord_offsetA  = DIM*jnrA;
 226             j_coord_offsetB  = DIM*jnrB;
 227
 228             /* load j atom coordinates */
 229             gmx_mm_load_4rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 230                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 231                                               &jy2,&jz2,&jx3,&jy3,&jz3);
 232
 233             /* Calculate displacement vector */
 234             dx00             = _mm_sub_pd(ix0,jx0);
 235             dy00             = _mm_sub_pd(iy0,jy0);
 236             dz00             = _mm_sub_pd(iz0,jz0);
 237             dx11             = _mm_sub_pd(ix1,jx1);
 238             dy11             = _mm_sub_pd(iy1,jy1);
 239             dz11             = _mm_sub_pd(iz1,jz1);
 240             dx12             = _mm_sub_pd(ix1,jx2);
 241             dy12             = _mm_sub_pd(iy1,jy2);
 242             dz12             = _mm_sub_pd(iz1,jz2);
 243             dx13             = _mm_sub_pd(ix1,jx3);
 244             dy13             = _mm_sub_pd(iy1,jy3);
 245             dz13             = _mm_sub_pd(iz1,jz3);
 246             dx21             = _mm_sub_pd(ix2,jx1);
 247             dy21             = _mm_sub_pd(iy2,jy1);
 248             dz21             = _mm_sub_pd(iz2,jz1);
 249             dx22             = _mm_sub_pd(ix2,jx2);
 250             dy22             = _mm_sub_pd(iy2,jy2);
 251             dz22             = _mm_sub_pd(iz2,jz2);
 252             dx23             = _mm_sub_pd(ix2,jx3);
 253             dy23             = _mm_sub_pd(iy2,jy3);
 254             dz23             = _mm_sub_pd(iz2,jz3);
 255             dx31             = _mm_sub_pd(ix3,jx1);
 256             dy31             = _mm_sub_pd(iy3,jy1);
 257             dz31             = _mm_sub_pd(iz3,jz1);
 258             dx32             = _mm_sub_pd(ix3,jx2);
 259             dy32             = _mm_sub_pd(iy3,jy2);
 260             dz32             = _mm_sub_pd(iz3,jz2);
 261             dx33             = _mm_sub_pd(ix3,jx3);
 262             dy33             = _mm_sub_pd(iy3,jy3);
 263             dz33             = _mm_sub_pd(iz3,jz3);
 264
 265             /* Calculate squared distance and things based on it */
 266             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 267             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 268             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 269             rsq13            = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
 270             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 271             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 272             rsq23            = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
 273             rsq31            = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
 274             rsq32            = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
 275             rsq33            = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
 276
 277             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 278             rinv11           = gmx_mm_invsqrt_pd(rsq11);
 279             rinv12           = gmx_mm_invsqrt_pd(rsq12);
 280             rinv13           = gmx_mm_invsqrt_pd(rsq13);
 281             rinv21           = gmx_mm_invsqrt_pd(rsq21);
 282             rinv22           = gmx_mm_invsqrt_pd(rsq22);
 283             rinv23           = gmx_mm_invsqrt_pd(rsq23);
 284             rinv31           = gmx_mm_invsqrt_pd(rsq31);
 285             rinv32           = gmx_mm_invsqrt_pd(rsq32);
 286             rinv33           = gmx_mm_invsqrt_pd(rsq33);
 287
 288             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
 289             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
 290             rinvsq13         = _mm_mul_pd(rinv13,rinv13);
 291             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
 292             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
 293             rinvsq23         = _mm_mul_pd(rinv23,rinv23);
 294             rinvsq31         = _mm_mul_pd(rinv31,rinv31);
 295             rinvsq32         = _mm_mul_pd(rinv32,rinv32);
 296             rinvsq33         = _mm_mul_pd(rinv33,rinv33);
 297
 298             fjx0             = _mm_setzero_pd();
 299             fjy0             = _mm_setzero_pd();
 300             fjz0             = _mm_setzero_pd();
 301             fjx1             = _mm_setzero_pd();
 302             fjy1             = _mm_setzero_pd();
 303             fjz1             = _mm_setzero_pd();
 304             fjx2             = _mm_setzero_pd();
 305             fjy2             = _mm_setzero_pd();
 306             fjz2             = _mm_setzero_pd();
 307             fjx3             = _mm_setzero_pd();
 308             fjy3             = _mm_setzero_pd();
 309             fjz3             = _mm_setzero_pd();
 310
 311             /**************************
 312              * CALCULATE INTERACTIONS *
 313              **************************/
 314
 315             r00              = _mm_mul_pd(rsq00,rinv00);
 316
 317             /* Calculate table index by multiplying r with table scale and truncate to integer */
 318             rt               = _mm_mul_pd(r00,vftabscale);
 319             vfitab           = _mm_cvttpd_epi32(rt);
 320 #ifdef __XOP__
 321             vfeps            = _mm_frcz_pd(rt);
 322 #else
 323             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 324 #endif
 325             twovfeps         = _mm_add_pd(vfeps,vfeps);
 326             vfitab           = _mm_slli_epi32(vfitab,3);
 327
 328             /* CUBIC SPLINE TABLE DISPERSION */
 329             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 330             F                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 331             GMX_MM_TRANSPOSE2_PD(Y,F);
 332             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
 333             H                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
 334             GMX_MM_TRANSPOSE2_PD(G,H);
 335             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(H,vfeps,G),F);
 336             VV               = _mm_macc_pd(vfeps,Fp,Y);
 337             vvdw6            = _mm_mul_pd(c6_00,VV);
 338             FF               = _mm_macc_pd(vfeps,_mm_macc_pd(twovfeps,H,G),Fp);
 339             fvdw6            = _mm_mul_pd(c6_00,FF);
 340
 341             /* CUBIC SPLINE TABLE REPULSION */
 342             vfitab           = _mm_add_epi32(vfitab,ifour);
 343             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 344             F                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 345             GMX_MM_TRANSPOSE2_PD(Y,F);
 346             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
 347             H                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
 348             GMX_MM_TRANSPOSE2_PD(G,H);
 349             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(H,vfeps,G),F);
 350             VV               = _mm_macc_pd(vfeps,Fp,Y);
 351             vvdw12           = _mm_mul_pd(c12_00,VV);
 352             FF               = _mm_macc_pd(vfeps,_mm_macc_pd(twovfeps,H,G),Fp);
 353             fvdw12           = _mm_mul_pd(c12_00,FF);
 354             vvdw             = _mm_add_pd(vvdw12,vvdw6);
 355             fvdw             = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
 356
 357             /* Update potential sum for this i atom from the interaction with this j atom. */
 358             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 359
 360             fscal            = fvdw;
 361
 362             /* Update vectorial force */
 363             fix0             = _mm_macc_pd(dx00,fscal,fix0);
 364             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
 365             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
 366
 367             fjx0             = _mm_macc_pd(dx00,fscal,fjx0);
 368             fjy0             = _mm_macc_pd(dy00,fscal,fjy0);
 369             fjz0             = _mm_macc_pd(dz00,fscal,fjz0);
 370
 371             /**************************
 372              * CALCULATE INTERACTIONS *
 373              **************************/
 374
 375             if (gmx_mm_any_lt(rsq11,rcutoff2))
 376             {
 377
 378             /* REACTION-FIELD ELECTROSTATICS */
 379             velec            = _mm_mul_pd(qq11,_mm_sub_pd(_mm_macc_pd(krf,rsq11,rinv11),crf));
 380             felec            = _mm_mul_pd(qq11,_mm_msub_pd(rinv11,rinvsq11,krf2));
 381
 382             cutoff_mask      = _mm_cmplt_pd(rsq11,rcutoff2);
 383
 384             /* Update potential sum for this i atom from the interaction with this j atom. */
 385             velec            = _mm_and_pd(velec,cutoff_mask);
 386             velecsum         = _mm_add_pd(velecsum,velec);
 387
 388             fscal            = felec;
 389
 390             fscal            = _mm_and_pd(fscal,cutoff_mask);
 391
 392             /* Update vectorial force */
 393             fix1             = _mm_macc_pd(dx11,fscal,fix1);
 394             fiy1             = _mm_macc_pd(dy11,fscal,fiy1);
 395             fiz1             = _mm_macc_pd(dz11,fscal,fiz1);
 396
 397             fjx1             = _mm_macc_pd(dx11,fscal,fjx1);
 398             fjy1             = _mm_macc_pd(dy11,fscal,fjy1);
 399             fjz1             = _mm_macc_pd(dz11,fscal,fjz1);
 400
 401             }
 402
 403             /**************************
 404              * CALCULATE INTERACTIONS *
 405              **************************/
 406
 407             if (gmx_mm_any_lt(rsq12,rcutoff2))
 408             {
 409
 410             /* REACTION-FIELD ELECTROSTATICS */
 411             velec            = _mm_mul_pd(qq12,_mm_sub_pd(_mm_macc_pd(krf,rsq12,rinv12),crf));
 412             felec            = _mm_mul_pd(qq12,_mm_msub_pd(rinv12,rinvsq12,krf2));
 413
 414             cutoff_mask      = _mm_cmplt_pd(rsq12,rcutoff2);
 415
 416             /* Update potential sum for this i atom from the interaction with this j atom. */
 417             velec            = _mm_and_pd(velec,cutoff_mask);
 418             velecsum         = _mm_add_pd(velecsum,velec);
 419
 420             fscal            = felec;
 421
 422             fscal            = _mm_and_pd(fscal,cutoff_mask);
 423
 424             /* Update vectorial force */
 425             fix1             = _mm_macc_pd(dx12,fscal,fix1);
 426             fiy1             = _mm_macc_pd(dy12,fscal,fiy1);
 427             fiz1             = _mm_macc_pd(dz12,fscal,fiz1);
 428
 429             fjx2             = _mm_macc_pd(dx12,fscal,fjx2);
 430             fjy2             = _mm_macc_pd(dy12,fscal,fjy2);
 431             fjz2             = _mm_macc_pd(dz12,fscal,fjz2);
 432
 433             }
 434
 435             /**************************
 436              * CALCULATE INTERACTIONS *
 437              **************************/
 438
 439             if (gmx_mm_any_lt(rsq13,rcutoff2))
 440             {
 441
 442             /* REACTION-FIELD ELECTROSTATICS */
 443             velec            = _mm_mul_pd(qq13,_mm_sub_pd(_mm_macc_pd(krf,rsq13,rinv13),crf));
 444             felec            = _mm_mul_pd(qq13,_mm_msub_pd(rinv13,rinvsq13,krf2));
 445
 446             cutoff_mask      = _mm_cmplt_pd(rsq13,rcutoff2);
 447
 448             /* Update potential sum for this i atom from the interaction with this j atom. */
 449             velec            = _mm_and_pd(velec,cutoff_mask);
 450             velecsum         = _mm_add_pd(velecsum,velec);
 451
 452             fscal            = felec;
 453
 454             fscal            = _mm_and_pd(fscal,cutoff_mask);
 455
 456             /* Update vectorial force */
 457             fix1             = _mm_macc_pd(dx13,fscal,fix1);
 458             fiy1             = _mm_macc_pd(dy13,fscal,fiy1);
 459             fiz1             = _mm_macc_pd(dz13,fscal,fiz1);
 460
 461             fjx3             = _mm_macc_pd(dx13,fscal,fjx3);
 462             fjy3             = _mm_macc_pd(dy13,fscal,fjy3);
 463             fjz3             = _mm_macc_pd(dz13,fscal,fjz3);
 464
 465             }
 466
 467             /**************************
 468              * CALCULATE INTERACTIONS *
 469              **************************/
 470
 471             if (gmx_mm_any_lt(rsq21,rcutoff2))
 472             {
 473
 474             /* REACTION-FIELD ELECTROSTATICS */
 475             velec            = _mm_mul_pd(qq21,_mm_sub_pd(_mm_macc_pd(krf,rsq21,rinv21),crf));
 476             felec            = _mm_mul_pd(qq21,_mm_msub_pd(rinv21,rinvsq21,krf2));
 477
 478             cutoff_mask      = _mm_cmplt_pd(rsq21,rcutoff2);
 479
 480             /* Update potential sum for this i atom from the interaction with this j atom. */
 481             velec            = _mm_and_pd(velec,cutoff_mask);
 482             velecsum         = _mm_add_pd(velecsum,velec);
 483
 484             fscal            = felec;
 485
 486             fscal            = _mm_and_pd(fscal,cutoff_mask);
 487
 488             /* Update vectorial force */
 489             fix2             = _mm_macc_pd(dx21,fscal,fix2);
 490             fiy2             = _mm_macc_pd(dy21,fscal,fiy2);
 491             fiz2             = _mm_macc_pd(dz21,fscal,fiz2);
 492
 493             fjx1             = _mm_macc_pd(dx21,fscal,fjx1);
 494             fjy1             = _mm_macc_pd(dy21,fscal,fjy1);
 495             fjz1             = _mm_macc_pd(dz21,fscal,fjz1);
 496
 497             }
 498
 499             /**************************
 500              * CALCULATE INTERACTIONS *
 501              **************************/
 502
 503             if (gmx_mm_any_lt(rsq22,rcutoff2))
 504             {
 505
 506             /* REACTION-FIELD ELECTROSTATICS */
 507             velec            = _mm_mul_pd(qq22,_mm_sub_pd(_mm_macc_pd(krf,rsq22,rinv22),crf));
 508             felec            = _mm_mul_pd(qq22,_mm_msub_pd(rinv22,rinvsq22,krf2));
 509
 510             cutoff_mask      = _mm_cmplt_pd(rsq22,rcutoff2);
 511
 512             /* Update potential sum for this i atom from the interaction with this j atom. */
 513             velec            = _mm_and_pd(velec,cutoff_mask);
 514             velecsum         = _mm_add_pd(velecsum,velec);
 515
 516             fscal            = felec;
 517
 518             fscal            = _mm_and_pd(fscal,cutoff_mask);
 519
 520             /* Update vectorial force */
 521             fix2             = _mm_macc_pd(dx22,fscal,fix2);
 522             fiy2             = _mm_macc_pd(dy22,fscal,fiy2);
 523             fiz2             = _mm_macc_pd(dz22,fscal,fiz2);
 524
 525             fjx2             = _mm_macc_pd(dx22,fscal,fjx2);
 526             fjy2             = _mm_macc_pd(dy22,fscal,fjy2);
 527             fjz2             = _mm_macc_pd(dz22,fscal,fjz2);
 528
 529             }
 530
 531             /**************************
 532              * CALCULATE INTERACTIONS *
 533              **************************/
 534
 535             if (gmx_mm_any_lt(rsq23,rcutoff2))
 536             {
 537
 538             /* REACTION-FIELD ELECTROSTATICS */
 539             velec            = _mm_mul_pd(qq23,_mm_sub_pd(_mm_macc_pd(krf,rsq23,rinv23),crf));
 540             felec            = _mm_mul_pd(qq23,_mm_msub_pd(rinv23,rinvsq23,krf2));
 541
 542             cutoff_mask      = _mm_cmplt_pd(rsq23,rcutoff2);
 543
 544             /* Update potential sum for this i atom from the interaction with this j atom. */
 545             velec            = _mm_and_pd(velec,cutoff_mask);
 546             velecsum         = _mm_add_pd(velecsum,velec);
 547
 548             fscal            = felec;
 549
 550             fscal            = _mm_and_pd(fscal,cutoff_mask);
 551
 552             /* Update vectorial force */
 553             fix2             = _mm_macc_pd(dx23,fscal,fix2);
 554             fiy2             = _mm_macc_pd(dy23,fscal,fiy2);
 555             fiz2             = _mm_macc_pd(dz23,fscal,fiz2);
 556
 557             fjx3             = _mm_macc_pd(dx23,fscal,fjx3);
 558             fjy3             = _mm_macc_pd(dy23,fscal,fjy3);
 559             fjz3             = _mm_macc_pd(dz23,fscal,fjz3);
 560
 561             }
 562
 563             /**************************
 564              * CALCULATE INTERACTIONS *
 565              **************************/
 566
 567             if (gmx_mm_any_lt(rsq31,rcutoff2))
 568             {
 569
 570             /* REACTION-FIELD ELECTROSTATICS */
 571             velec            = _mm_mul_pd(qq31,_mm_sub_pd(_mm_macc_pd(krf,rsq31,rinv31),crf));
 572             felec            = _mm_mul_pd(qq31,_mm_msub_pd(rinv31,rinvsq31,krf2));
 573
 574             cutoff_mask      = _mm_cmplt_pd(rsq31,rcutoff2);
 575
 576             /* Update potential sum for this i atom from the interaction with this j atom. */
 577             velec            = _mm_and_pd(velec,cutoff_mask);
 578             velecsum         = _mm_add_pd(velecsum,velec);
 579
 580             fscal            = felec;
 581
 582             fscal            = _mm_and_pd(fscal,cutoff_mask);
 583
 584             /* Update vectorial force */
 585             fix3             = _mm_macc_pd(dx31,fscal,fix3);
 586             fiy3             = _mm_macc_pd(dy31,fscal,fiy3);
 587             fiz3             = _mm_macc_pd(dz31,fscal,fiz3);
 588
 589             fjx1             = _mm_macc_pd(dx31,fscal,fjx1);
 590             fjy1             = _mm_macc_pd(dy31,fscal,fjy1);
 591             fjz1             = _mm_macc_pd(dz31,fscal,fjz1);
 592
 593             }
 594
 595             /**************************
 596              * CALCULATE INTERACTIONS *
 597              **************************/
 598
 599             if (gmx_mm_any_lt(rsq32,rcutoff2))
 600             {
 601
 602             /* REACTION-FIELD ELECTROSTATICS */
 603             velec            = _mm_mul_pd(qq32,_mm_sub_pd(_mm_macc_pd(krf,rsq32,rinv32),crf));
 604             felec            = _mm_mul_pd(qq32,_mm_msub_pd(rinv32,rinvsq32,krf2));
 605
 606             cutoff_mask      = _mm_cmplt_pd(rsq32,rcutoff2);
 607
 608             /* Update potential sum for this i atom from the interaction with this j atom. */
 609             velec            = _mm_and_pd(velec,cutoff_mask);
 610             velecsum         = _mm_add_pd(velecsum,velec);
 611
 612             fscal            = felec;
 613
 614             fscal            = _mm_and_pd(fscal,cutoff_mask);
 615
 616             /* Update vectorial force */
 617             fix3             = _mm_macc_pd(dx32,fscal,fix3);
 618             fiy3             = _mm_macc_pd(dy32,fscal,fiy3);
 619             fiz3             = _mm_macc_pd(dz32,fscal,fiz3);
 620
 621             fjx2             = _mm_macc_pd(dx32,fscal,fjx2);
 622             fjy2             = _mm_macc_pd(dy32,fscal,fjy2);
 623             fjz2             = _mm_macc_pd(dz32,fscal,fjz2);
 624
 625             }
 626
 627             /**************************
 628              * CALCULATE INTERACTIONS *
 629              **************************/
 630
 631             if (gmx_mm_any_lt(rsq33,rcutoff2))
 632             {
 633
 634             /* REACTION-FIELD ELECTROSTATICS */
 635             velec            = _mm_mul_pd(qq33,_mm_sub_pd(_mm_macc_pd(krf,rsq33,rinv33),crf));
 636             felec            = _mm_mul_pd(qq33,_mm_msub_pd(rinv33,rinvsq33,krf2));
 637
 638             cutoff_mask      = _mm_cmplt_pd(rsq33,rcutoff2);
 639
 640             /* Update potential sum for this i atom from the interaction with this j atom. */
 641             velec            = _mm_and_pd(velec,cutoff_mask);
 642             velecsum         = _mm_add_pd(velecsum,velec);
 643
 644             fscal            = felec;
 645
 646             fscal            = _mm_and_pd(fscal,cutoff_mask);
 647
 648             /* Update vectorial force */
 649             fix3             = _mm_macc_pd(dx33,fscal,fix3);
 650             fiy3             = _mm_macc_pd(dy33,fscal,fiy3);
 651             fiz3             = _mm_macc_pd(dz33,fscal,fiz3);
 652
 653             fjx3             = _mm_macc_pd(dx33,fscal,fjx3);
 654             fjy3             = _mm_macc_pd(dy33,fscal,fjy3);
 655             fjz3             = _mm_macc_pd(dz33,fscal,fjz3);
 656
 657             }
 658
 659             gmx_mm_decrement_4rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 660
 661             /* Inner loop uses 413 flops */
 662         }
 663
 664         if(jidx<j_index_end)
 665         {
 666
 667             jnrA             = jjnr[jidx];
 668             j_coord_offsetA  = DIM*jnrA;
 669
 670             /* load j atom coordinates */
 671             gmx_mm_load_4rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 672                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 673                                               &jy2,&jz2,&jx3,&jy3,&jz3);
 674
 675             /* Calculate displacement vector */
 676             dx00             = _mm_sub_pd(ix0,jx0);
 677             dy00             = _mm_sub_pd(iy0,jy0);
 678             dz00             = _mm_sub_pd(iz0,jz0);
 679             dx11             = _mm_sub_pd(ix1,jx1);
 680             dy11             = _mm_sub_pd(iy1,jy1);
 681             dz11             = _mm_sub_pd(iz1,jz1);
 682             dx12             = _mm_sub_pd(ix1,jx2);
 683             dy12             = _mm_sub_pd(iy1,jy2);
 684             dz12             = _mm_sub_pd(iz1,jz2);
 685             dx13             = _mm_sub_pd(ix1,jx3);
 686             dy13             = _mm_sub_pd(iy1,jy3);
 687             dz13             = _mm_sub_pd(iz1,jz3);
 688             dx21             = _mm_sub_pd(ix2,jx1);
 689             dy21             = _mm_sub_pd(iy2,jy1);
 690             dz21             = _mm_sub_pd(iz2,jz1);
 691             dx22             = _mm_sub_pd(ix2,jx2);
 692             dy22             = _mm_sub_pd(iy2,jy2);
 693             dz22             = _mm_sub_pd(iz2,jz2);
 694             dx23             = _mm_sub_pd(ix2,jx3);
 695             dy23             = _mm_sub_pd(iy2,jy3);
 696             dz23             = _mm_sub_pd(iz2,jz3);
 697             dx31             = _mm_sub_pd(ix3,jx1);
 698             dy31             = _mm_sub_pd(iy3,jy1);
 699             dz31             = _mm_sub_pd(iz3,jz1);
 700             dx32             = _mm_sub_pd(ix3,jx2);
 701             dy32             = _mm_sub_pd(iy3,jy2);
 702             dz32             = _mm_sub_pd(iz3,jz2);
 703             dx33             = _mm_sub_pd(ix3,jx3);
 704             dy33             = _mm_sub_pd(iy3,jy3);
 705             dz33             = _mm_sub_pd(iz3,jz3);
 706
 707             /* Calculate squared distance and things based on it */
 708             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 709             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 710             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 711             rsq13            = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
 712             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 713             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 714             rsq23            = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
 715             rsq31            = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
 716             rsq32            = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
 717             rsq33            = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
 718
 719             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 720             rinv11           = gmx_mm_invsqrt_pd(rsq11);
 721             rinv12           = gmx_mm_invsqrt_pd(rsq12);
 722             rinv13           = gmx_mm_invsqrt_pd(rsq13);
 723             rinv21           = gmx_mm_invsqrt_pd(rsq21);
 724             rinv22           = gmx_mm_invsqrt_pd(rsq22);
 725             rinv23           = gmx_mm_invsqrt_pd(rsq23);
 726             rinv31           = gmx_mm_invsqrt_pd(rsq31);
 727             rinv32           = gmx_mm_invsqrt_pd(rsq32);
 728             rinv33           = gmx_mm_invsqrt_pd(rsq33);
 729
 730             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
 731             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
 732             rinvsq13         = _mm_mul_pd(rinv13,rinv13);
 733             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
 734             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
 735             rinvsq23         = _mm_mul_pd(rinv23,rinv23);
 736             rinvsq31         = _mm_mul_pd(rinv31,rinv31);
 737             rinvsq32         = _mm_mul_pd(rinv32,rinv32);
 738             rinvsq33         = _mm_mul_pd(rinv33,rinv33);
 739
 740             fjx0             = _mm_setzero_pd();
 741             fjy0             = _mm_setzero_pd();
 742             fjz0             = _mm_setzero_pd();
 743             fjx1             = _mm_setzero_pd();
 744             fjy1             = _mm_setzero_pd();
 745             fjz1             = _mm_setzero_pd();
 746             fjx2             = _mm_setzero_pd();
 747             fjy2             = _mm_setzero_pd();
 748             fjz2             = _mm_setzero_pd();
 749             fjx3             = _mm_setzero_pd();
 750             fjy3             = _mm_setzero_pd();
 751             fjz3             = _mm_setzero_pd();
 752
 753             /**************************
 754              * CALCULATE INTERACTIONS *
 755              **************************/
 756
 757             r00              = _mm_mul_pd(rsq00,rinv00);
 758
 759             /* Calculate table index by multiplying r with table scale and truncate to integer */
 760             rt               = _mm_mul_pd(r00,vftabscale);
 761             vfitab           = _mm_cvttpd_epi32(rt);
 762 #ifdef __XOP__
 763             vfeps            = _mm_frcz_pd(rt);
 764 #else
 765             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 766 #endif
 767             twovfeps         = _mm_add_pd(vfeps,vfeps);
 768             vfitab           = _mm_slli_epi32(vfitab,3);
 769
 770             /* CUBIC SPLINE TABLE DISPERSION */
 771             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 772             F                = _mm_setzero_pd();
 773             GMX_MM_TRANSPOSE2_PD(Y,F);
 774             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
 775             H                = _mm_setzero_pd();
 776             GMX_MM_TRANSPOSE2_PD(G,H);
 777             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(H,vfeps,G),F);
 778             VV               = _mm_macc_pd(vfeps,Fp,Y);
 779             vvdw6            = _mm_mul_pd(c6_00,VV);
 780             FF               = _mm_macc_pd(vfeps,_mm_macc_pd(twovfeps,H,G),Fp);
 781             fvdw6            = _mm_mul_pd(c6_00,FF);
 782
 783             /* CUBIC SPLINE TABLE REPULSION */
 784             vfitab           = _mm_add_epi32(vfitab,ifour);
 785             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 786             F                = _mm_setzero_pd();
 787             GMX_MM_TRANSPOSE2_PD(Y,F);
 788             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
 789             H                = _mm_setzero_pd();
 790             GMX_MM_TRANSPOSE2_PD(G,H);
 791             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(H,vfeps,G),F);
 792             VV               = _mm_macc_pd(vfeps,Fp,Y);
 793             vvdw12           = _mm_mul_pd(c12_00,VV);
 794             FF               = _mm_macc_pd(vfeps,_mm_macc_pd(twovfeps,H,G),Fp);
 795             fvdw12           = _mm_mul_pd(c12_00,FF);
 796             vvdw             = _mm_add_pd(vvdw12,vvdw6);
 797             fvdw             = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
 798
 799             /* Update potential sum for this i atom from the interaction with this j atom. */
 800             vvdw             = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
 801             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 802
 803             fscal            = fvdw;
 804
 805             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 806
 807             /* Update vectorial force */
 808             fix0             = _mm_macc_pd(dx00,fscal,fix0);
 809             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
 810             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
 811
 812             fjx0             = _mm_macc_pd(dx00,fscal,fjx0);
 813             fjy0             = _mm_macc_pd(dy00,fscal,fjy0);
 814             fjz0             = _mm_macc_pd(dz00,fscal,fjz0);
 815
 816             /**************************
 817              * CALCULATE INTERACTIONS *
 818              **************************/
 819
 820             if (gmx_mm_any_lt(rsq11,rcutoff2))
 821             {
 822
 823             /* REACTION-FIELD ELECTROSTATICS */
 824             velec            = _mm_mul_pd(qq11,_mm_sub_pd(_mm_macc_pd(krf,rsq11,rinv11),crf));
 825             felec            = _mm_mul_pd(qq11,_mm_msub_pd(rinv11,rinvsq11,krf2));
 826
 827             cutoff_mask      = _mm_cmplt_pd(rsq11,rcutoff2);
 828
 829             /* Update potential sum for this i atom from the interaction with this j atom. */
 830             velec            = _mm_and_pd(velec,cutoff_mask);
 831             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 832             velecsum         = _mm_add_pd(velecsum,velec);
 833
 834             fscal            = felec;
 835
 836             fscal            = _mm_and_pd(fscal,cutoff_mask);
 837
 838             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 839
 840             /* Update vectorial force */
 841             fix1             = _mm_macc_pd(dx11,fscal,fix1);
 842             fiy1             = _mm_macc_pd(dy11,fscal,fiy1);
 843             fiz1             = _mm_macc_pd(dz11,fscal,fiz1);
 844
 845             fjx1             = _mm_macc_pd(dx11,fscal,fjx1);
 846             fjy1             = _mm_macc_pd(dy11,fscal,fjy1);
 847             fjz1             = _mm_macc_pd(dz11,fscal,fjz1);
 848
 849             }
 850
 851             /**************************
 852              * CALCULATE INTERACTIONS *
 853              **************************/
 854
 855             if (gmx_mm_any_lt(rsq12,rcutoff2))
 856             {
 857
 858             /* REACTION-FIELD ELECTROSTATICS */
 859             velec            = _mm_mul_pd(qq12,_mm_sub_pd(_mm_macc_pd(krf,rsq12,rinv12),crf));
 860             felec            = _mm_mul_pd(qq12,_mm_msub_pd(rinv12,rinvsq12,krf2));
 861
 862             cutoff_mask      = _mm_cmplt_pd(rsq12,rcutoff2);
 863
 864             /* Update potential sum for this i atom from the interaction with this j atom. */
 865             velec            = _mm_and_pd(velec,cutoff_mask);
 866             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 867             velecsum         = _mm_add_pd(velecsum,velec);
 868
 869             fscal            = felec;
 870
 871             fscal            = _mm_and_pd(fscal,cutoff_mask);
 872
 873             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 874
 875             /* Update vectorial force */
 876             fix1             = _mm_macc_pd(dx12,fscal,fix1);
 877             fiy1             = _mm_macc_pd(dy12,fscal,fiy1);
 878             fiz1             = _mm_macc_pd(dz12,fscal,fiz1);
 879
 880             fjx2             = _mm_macc_pd(dx12,fscal,fjx2);
 881             fjy2             = _mm_macc_pd(dy12,fscal,fjy2);
 882             fjz2             = _mm_macc_pd(dz12,fscal,fjz2);
 883
 884             }
 885
 886             /**************************
 887              * CALCULATE INTERACTIONS *
 888              **************************/
 889
 890             if (gmx_mm_any_lt(rsq13,rcutoff2))
 891             {
 892
 893             /* REACTION-FIELD ELECTROSTATICS */
 894             velec            = _mm_mul_pd(qq13,_mm_sub_pd(_mm_macc_pd(krf,rsq13,rinv13),crf));
 895             felec            = _mm_mul_pd(qq13,_mm_msub_pd(rinv13,rinvsq13,krf2));
 896
 897             cutoff_mask      = _mm_cmplt_pd(rsq13,rcutoff2);
 898
 899             /* Update potential sum for this i atom from the interaction with this j atom. */
 900             velec            = _mm_and_pd(velec,cutoff_mask);
 901             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 902             velecsum         = _mm_add_pd(velecsum,velec);
 903
 904             fscal            = felec;
 905
 906             fscal            = _mm_and_pd(fscal,cutoff_mask);
 907
 908             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 909
 910             /* Update vectorial force */
 911             fix1             = _mm_macc_pd(dx13,fscal,fix1);
 912             fiy1             = _mm_macc_pd(dy13,fscal,fiy1);
 913             fiz1             = _mm_macc_pd(dz13,fscal,fiz1);
 914
 915             fjx3             = _mm_macc_pd(dx13,fscal,fjx3);
 916             fjy3             = _mm_macc_pd(dy13,fscal,fjy3);
 917             fjz3             = _mm_macc_pd(dz13,fscal,fjz3);
 918
 919             }
 920
 921             /**************************
 922              * CALCULATE INTERACTIONS *
 923              **************************/
 924
 925             if (gmx_mm_any_lt(rsq21,rcutoff2))
 926             {
 927
 928             /* REACTION-FIELD ELECTROSTATICS */
 929             velec            = _mm_mul_pd(qq21,_mm_sub_pd(_mm_macc_pd(krf,rsq21,rinv21),crf));
 930             felec            = _mm_mul_pd(qq21,_mm_msub_pd(rinv21,rinvsq21,krf2));
 931
 932             cutoff_mask      = _mm_cmplt_pd(rsq21,rcutoff2);
 933
 934             /* Update potential sum for this i atom from the interaction with this j atom. */
 935             velec            = _mm_and_pd(velec,cutoff_mask);
 936             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 937             velecsum         = _mm_add_pd(velecsum,velec);
 938
 939             fscal            = felec;
 940
 941             fscal            = _mm_and_pd(fscal,cutoff_mask);
 942
 943             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 944
 945             /* Update vectorial force */
 946             fix2             = _mm_macc_pd(dx21,fscal,fix2);
 947             fiy2             = _mm_macc_pd(dy21,fscal,fiy2);
 948             fiz2             = _mm_macc_pd(dz21,fscal,fiz2);
 949
 950             fjx1             = _mm_macc_pd(dx21,fscal,fjx1);
 951             fjy1             = _mm_macc_pd(dy21,fscal,fjy1);
 952             fjz1             = _mm_macc_pd(dz21,fscal,fjz1);
 953
 954             }
 955
 956             /**************************
 957              * CALCULATE INTERACTIONS *
 958              **************************/
 959
 960             if (gmx_mm_any_lt(rsq22,rcutoff2))
 961             {
 962
 963             /* REACTION-FIELD ELECTROSTATICS */
 964             velec            = _mm_mul_pd(qq22,_mm_sub_pd(_mm_macc_pd(krf,rsq22,rinv22),crf));
 965             felec            = _mm_mul_pd(qq22,_mm_msub_pd(rinv22,rinvsq22,krf2));
 966
 967             cutoff_mask      = _mm_cmplt_pd(rsq22,rcutoff2);
 968
 969             /* Update potential sum for this i atom from the interaction with this j atom. */
 970             velec            = _mm_and_pd(velec,cutoff_mask);
 971             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 972             velecsum         = _mm_add_pd(velecsum,velec);
 973
 974             fscal            = felec;
 975
 976             fscal            = _mm_and_pd(fscal,cutoff_mask);
 977
 978             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 979
 980             /* Update vectorial force */
 981             fix2             = _mm_macc_pd(dx22,fscal,fix2);
 982             fiy2             = _mm_macc_pd(dy22,fscal,fiy2);
 983             fiz2             = _mm_macc_pd(dz22,fscal,fiz2);
 984
 985             fjx2             = _mm_macc_pd(dx22,fscal,fjx2);
 986             fjy2             = _mm_macc_pd(dy22,fscal,fjy2);
 987             fjz2             = _mm_macc_pd(dz22,fscal,fjz2);
 988
 989             }
 990
 991             /**************************
 992              * CALCULATE INTERACTIONS *
 993              **************************/
 994
 995             if (gmx_mm_any_lt(rsq23,rcutoff2))
 996             {
 997
 998             /* REACTION-FIELD ELECTROSTATICS */
 999             velec            = _mm_mul_pd(qq23,_mm_sub_pd(_mm_macc_pd(krf,rsq23,rinv23),crf));
1000             felec            = _mm_mul_pd(qq23,_mm_msub_pd(rinv23,rinvsq23,krf2));
1001
1002             cutoff_mask      = _mm_cmplt_pd(rsq23,rcutoff2);
1003
1004             /* Update potential sum for this i atom from the interaction with this j atom. */
1005             velec            = _mm_and_pd(velec,cutoff_mask);
1006             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1007             velecsum         = _mm_add_pd(velecsum,velec);
1008
1009             fscal            = felec;
1010
1011             fscal            = _mm_and_pd(fscal,cutoff_mask);
1012
1013             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1014
1015             /* Update vectorial force */
1016             fix2             = _mm_macc_pd(dx23,fscal,fix2);
1017             fiy2             = _mm_macc_pd(dy23,fscal,fiy2);
1018             fiz2             = _mm_macc_pd(dz23,fscal,fiz2);
1019
1020             fjx3             = _mm_macc_pd(dx23,fscal,fjx3);
1021             fjy3             = _mm_macc_pd(dy23,fscal,fjy3);
1022             fjz3             = _mm_macc_pd(dz23,fscal,fjz3);
1023
1024             }
1025
1026             /**************************
1027              * CALCULATE INTERACTIONS *
1028              **************************/
1029
1030             if (gmx_mm_any_lt(rsq31,rcutoff2))
1031             {
1032
1033             /* REACTION-FIELD ELECTROSTATICS */
1034             velec            = _mm_mul_pd(qq31,_mm_sub_pd(_mm_macc_pd(krf,rsq31,rinv31),crf));
1035             felec            = _mm_mul_pd(qq31,_mm_msub_pd(rinv31,rinvsq31,krf2));
1036
1037             cutoff_mask      = _mm_cmplt_pd(rsq31,rcutoff2);
1038
1039             /* Update potential sum for this i atom from the interaction with this j atom. */
1040             velec            = _mm_and_pd(velec,cutoff_mask);
1041             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1042             velecsum         = _mm_add_pd(velecsum,velec);
1043
1044             fscal            = felec;
1045
1046             fscal            = _mm_and_pd(fscal,cutoff_mask);
1047
1048             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1049
1050             /* Update vectorial force */
1051             fix3             = _mm_macc_pd(dx31,fscal,fix3);
1052             fiy3             = _mm_macc_pd(dy31,fscal,fiy3);
1053             fiz3             = _mm_macc_pd(dz31,fscal,fiz3);
1054
1055             fjx1             = _mm_macc_pd(dx31,fscal,fjx1);
1056             fjy1             = _mm_macc_pd(dy31,fscal,fjy1);
1057             fjz1             = _mm_macc_pd(dz31,fscal,fjz1);
1058
1059             }
1060
1061             /**************************
1062              * CALCULATE INTERACTIONS *
1063              **************************/
1064
1065             if (gmx_mm_any_lt(rsq32,rcutoff2))
1066             {
1067
1068             /* REACTION-FIELD ELECTROSTATICS */
1069             velec            = _mm_mul_pd(qq32,_mm_sub_pd(_mm_macc_pd(krf,rsq32,rinv32),crf));
1070             felec            = _mm_mul_pd(qq32,_mm_msub_pd(rinv32,rinvsq32,krf2));
1071
1072             cutoff_mask      = _mm_cmplt_pd(rsq32,rcutoff2);
1073
1074             /* Update potential sum for this i atom from the interaction with this j atom. */
1075             velec            = _mm_and_pd(velec,cutoff_mask);
1076             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1077             velecsum         = _mm_add_pd(velecsum,velec);
1078
1079             fscal            = felec;
1080
1081             fscal            = _mm_and_pd(fscal,cutoff_mask);
1082
1083             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1084
1085             /* Update vectorial force */
1086             fix3             = _mm_macc_pd(dx32,fscal,fix3);
1087             fiy3             = _mm_macc_pd(dy32,fscal,fiy3);
1088             fiz3             = _mm_macc_pd(dz32,fscal,fiz3);
1089
1090             fjx2             = _mm_macc_pd(dx32,fscal,fjx2);
1091             fjy2             = _mm_macc_pd(dy32,fscal,fjy2);
1092             fjz2             = _mm_macc_pd(dz32,fscal,fjz2);
1093
1094             }
1095
1096             /**************************
1097              * CALCULATE INTERACTIONS *
1098              **************************/
1099
1100             if (gmx_mm_any_lt(rsq33,rcutoff2))
1101             {
1102
1103             /* REACTION-FIELD ELECTROSTATICS */
1104             velec            = _mm_mul_pd(qq33,_mm_sub_pd(_mm_macc_pd(krf,rsq33,rinv33),crf));
1105             felec            = _mm_mul_pd(qq33,_mm_msub_pd(rinv33,rinvsq33,krf2));
1106
1107             cutoff_mask      = _mm_cmplt_pd(rsq33,rcutoff2);
1108
1109             /* Update potential sum for this i atom from the interaction with this j atom. */
1110             velec            = _mm_and_pd(velec,cutoff_mask);
1111             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1112             velecsum         = _mm_add_pd(velecsum,velec);
1113
1114             fscal            = felec;
1115
1116             fscal            = _mm_and_pd(fscal,cutoff_mask);
1117
1118             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1119
1120             /* Update vectorial force */
1121             fix3             = _mm_macc_pd(dx33,fscal,fix3);
1122             fiy3             = _mm_macc_pd(dy33,fscal,fiy3);
1123             fiz3             = _mm_macc_pd(dz33,fscal,fiz3);
1124
1125             fjx3             = _mm_macc_pd(dx33,fscal,fjx3);
1126             fjy3             = _mm_macc_pd(dy33,fscal,fjy3);
1127             fjz3             = _mm_macc_pd(dz33,fscal,fjz3);
1128
1129             }
1130
1131             gmx_mm_decrement_4rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1132
1133             /* Inner loop uses 413 flops */
1134         }
1135
1136         /* End of innermost loop */
1137
1138         gmx_mm_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1139                                               f+i_coord_offset,fshift+i_shift_offset);
1140
1141         ggid                        = gid[iidx];
1142         /* Update potential energies */
1143         gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1144         gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1145
1146         /* Increment number of inner iterations */
1147         inneriter                  += j_index_end - j_index_start;
1148
1149         /* Outer loop uses 26 flops */
1150     }
1151
1152     /* Increment number of outer iterations */
1153     outeriter        += nri;
1154
1155     /* Update outer/inner flops */
1156
1157     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*413);
1158 }
1159 /*
1160  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwCSTab_GeomW4W4_F_avx_128_fma_double
1161  * Electrostatics interaction: ReactionField
1162  * VdW interaction:            CubicSplineTable
1163  * Geometry:                   Water4-Water4
1164  * Calculate force/pot:        Force
1165  */
1166 void
1167 nb_kernel_ElecRFCut_VdwCSTab_GeomW4W4_F_avx_128_fma_double
1168                     (t_nblist                    * gmx_restrict       nlist,
1169                      rvec                        * gmx_restrict          xx,
1170                      rvec                        * gmx_restrict          ff,
1171                      t_forcerec                  * gmx_restrict          fr,
1172                      t_mdatoms                   * gmx_restrict     mdatoms,
1173                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1174                      t_nrnb                      * gmx_restrict        nrnb)
1175 {
1176     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1177      * just 0 for non-waters.
1178      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
1179      * jnr indices corresponding to data put in the four positions in the SIMD register.
1180      */
1181     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1182     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1183     int              jnrA,jnrB;
1184     int              j_coord_offsetA,j_coord_offsetB;
1185     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1186     real             rcutoff_scalar;
1187     real             *shiftvec,*fshift,*x,*f;
1188     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1189     int              vdwioffset0;
1190     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1191     int              vdwioffset1;
1192     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1193     int              vdwioffset2;
1194     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1195     int              vdwioffset3;
1196     __m128d          ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1197     int              vdwjidx0A,vdwjidx0B;
1198     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1199     int              vdwjidx1A,vdwjidx1B;
1200     __m128d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1201     int              vdwjidx2A,vdwjidx2B;
1202     __m128d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1203     int              vdwjidx3A,vdwjidx3B;
1204     __m128d          jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1205     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1206     __m128d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1207     __m128d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1208     __m128d          dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1209     __m128d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1210     __m128d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1211     __m128d          dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1212     __m128d          dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1213     __m128d          dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1214     __m128d          dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1215     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
1216     real             *charge;
1217     int              nvdwtype;
1218     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1219     int              *vdwtype;
1220     real             *vdwparam;
1221     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
1222     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
1223     __m128i          vfitab;
1224     __m128i          ifour       = _mm_set1_epi32(4);
1225     __m128d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
1226     real             *vftab;
1227     __m128d          dummy_mask,cutoff_mask;
1228     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
1229     __m128d          one     = _mm_set1_pd(1.0);
1230     __m128d          two     = _mm_set1_pd(2.0);
1231     x                = xx[0];
1232     f                = ff[0];
1233
1234     nri              = nlist->nri;
1235     iinr             = nlist->iinr;
1236     jindex           = nlist->jindex;
1237     jjnr             = nlist->jjnr;
1238     shiftidx         = nlist->shift;
1239     gid              = nlist->gid;
1240     shiftvec         = fr->shift_vec[0];
1241     fshift           = fr->fshift[0];
1242     facel            = _mm_set1_pd(fr->epsfac);
1243     charge           = mdatoms->chargeA;
1244     krf              = _mm_set1_pd(fr->ic->k_rf);
1245     krf2             = _mm_set1_pd(fr->ic->k_rf*2.0);
1246     crf              = _mm_set1_pd(fr->ic->c_rf);
1247     nvdwtype         = fr->ntype;
1248     vdwparam         = fr->nbfp;
1249     vdwtype          = mdatoms->typeA;
1250
1251     vftab            = kernel_data->table_vdw->data;
1252     vftabscale       = _mm_set1_pd(kernel_data->table_vdw->scale);
1253
1254     /* Setup water-specific parameters */
1255     inr              = nlist->iinr[0];
1256     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
1257     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
1258     iq3              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
1259     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
1260
1261     jq1              = _mm_set1_pd(charge[inr+1]);
1262     jq2              = _mm_set1_pd(charge[inr+2]);
1263     jq3              = _mm_set1_pd(charge[inr+3]);
1264     vdwjidx0A        = 2*vdwtype[inr+0];
1265     c6_00            = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
1266     c12_00           = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
1267     qq11             = _mm_mul_pd(iq1,jq1);
1268     qq12             = _mm_mul_pd(iq1,jq2);
1269     qq13             = _mm_mul_pd(iq1,jq3);
1270     qq21             = _mm_mul_pd(iq2,jq1);
1271     qq22             = _mm_mul_pd(iq2,jq2);
1272     qq23             = _mm_mul_pd(iq2,jq3);
1273     qq31             = _mm_mul_pd(iq3,jq1);
1274     qq32             = _mm_mul_pd(iq3,jq2);
1275     qq33             = _mm_mul_pd(iq3,jq3);
1276
1277     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
1278     rcutoff_scalar   = fr->rcoulomb;
1279     rcutoff          = _mm_set1_pd(rcutoff_scalar);
1280     rcutoff2         = _mm_mul_pd(rcutoff,rcutoff);
1281
1282     /* Avoid stupid compiler warnings */
1283     jnrA = jnrB = 0;
1284     j_coord_offsetA = 0;
1285     j_coord_offsetB = 0;
1286
1287     outeriter        = 0;
1288     inneriter        = 0;
1289
1290     /* Start outer loop over neighborlists */
1291     for(iidx=0; iidx<nri; iidx++)
1292     {
1293         /* Load shift vector for this list */
1294         i_shift_offset   = DIM*shiftidx[iidx];
1295
1296         /* Load limits for loop over neighbors */
1297         j_index_start    = jindex[iidx];
1298         j_index_end      = jindex[iidx+1];
1299
1300         /* Get outer coordinate index */
1301         inr              = iinr[iidx];
1302         i_coord_offset   = DIM*inr;
1303
1304         /* Load i particle coords and add shift vector */
1305         gmx_mm_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1306                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1307
1308         fix0             = _mm_setzero_pd();
1309         fiy0             = _mm_setzero_pd();
1310         fiz0             = _mm_setzero_pd();
1311         fix1             = _mm_setzero_pd();
1312         fiy1             = _mm_setzero_pd();
1313         fiz1             = _mm_setzero_pd();
1314         fix2             = _mm_setzero_pd();
1315         fiy2             = _mm_setzero_pd();
1316         fiz2             = _mm_setzero_pd();
1317         fix3             = _mm_setzero_pd();
1318         fiy3             = _mm_setzero_pd();
1319         fiz3             = _mm_setzero_pd();
1320
1321         /* Start inner kernel loop */
1322         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
1323         {
1324
1325             /* Get j neighbor index, and coordinate index */
1326             jnrA             = jjnr[jidx];
1327             jnrB             = jjnr[jidx+1];
1328             j_coord_offsetA  = DIM*jnrA;
1329             j_coord_offsetB  = DIM*jnrB;
1330
1331             /* load j atom coordinates */
1332             gmx_mm_load_4rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1333                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1334                                               &jy2,&jz2,&jx3,&jy3,&jz3);
1335
1336             /* Calculate displacement vector */
1337             dx00             = _mm_sub_pd(ix0,jx0);
1338             dy00             = _mm_sub_pd(iy0,jy0);
1339             dz00             = _mm_sub_pd(iz0,jz0);
1340             dx11             = _mm_sub_pd(ix1,jx1);
1341             dy11             = _mm_sub_pd(iy1,jy1);
1342             dz11             = _mm_sub_pd(iz1,jz1);
1343             dx12             = _mm_sub_pd(ix1,jx2);
1344             dy12             = _mm_sub_pd(iy1,jy2);
1345             dz12             = _mm_sub_pd(iz1,jz2);
1346             dx13             = _mm_sub_pd(ix1,jx3);
1347             dy13             = _mm_sub_pd(iy1,jy3);
1348             dz13             = _mm_sub_pd(iz1,jz3);
1349             dx21             = _mm_sub_pd(ix2,jx1);
1350             dy21             = _mm_sub_pd(iy2,jy1);
1351             dz21             = _mm_sub_pd(iz2,jz1);
1352             dx22             = _mm_sub_pd(ix2,jx2);
1353             dy22             = _mm_sub_pd(iy2,jy2);
1354             dz22             = _mm_sub_pd(iz2,jz2);
1355             dx23             = _mm_sub_pd(ix2,jx3);
1356             dy23             = _mm_sub_pd(iy2,jy3);
1357             dz23             = _mm_sub_pd(iz2,jz3);
1358             dx31             = _mm_sub_pd(ix3,jx1);
1359             dy31             = _mm_sub_pd(iy3,jy1);
1360             dz31             = _mm_sub_pd(iz3,jz1);
1361             dx32             = _mm_sub_pd(ix3,jx2);
1362             dy32             = _mm_sub_pd(iy3,jy2);
1363             dz32             = _mm_sub_pd(iz3,jz2);
1364             dx33             = _mm_sub_pd(ix3,jx3);
1365             dy33             = _mm_sub_pd(iy3,jy3);
1366             dz33             = _mm_sub_pd(iz3,jz3);
1367
1368             /* Calculate squared distance and things based on it */
1369             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1370             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1371             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1372             rsq13            = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
1373             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1374             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1375             rsq23            = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
1376             rsq31            = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
1377             rsq32            = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
1378             rsq33            = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
1379
1380             rinv00           = gmx_mm_invsqrt_pd(rsq00);
1381             rinv11           = gmx_mm_invsqrt_pd(rsq11);
1382             rinv12           = gmx_mm_invsqrt_pd(rsq12);
1383             rinv13           = gmx_mm_invsqrt_pd(rsq13);
1384             rinv21           = gmx_mm_invsqrt_pd(rsq21);
1385             rinv22           = gmx_mm_invsqrt_pd(rsq22);
1386             rinv23           = gmx_mm_invsqrt_pd(rsq23);
1387             rinv31           = gmx_mm_invsqrt_pd(rsq31);
1388             rinv32           = gmx_mm_invsqrt_pd(rsq32);
1389             rinv33           = gmx_mm_invsqrt_pd(rsq33);
1390
1391             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
1392             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
1393             rinvsq13         = _mm_mul_pd(rinv13,rinv13);
1394             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
1395             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
1396             rinvsq23         = _mm_mul_pd(rinv23,rinv23);
1397             rinvsq31         = _mm_mul_pd(rinv31,rinv31);
1398             rinvsq32         = _mm_mul_pd(rinv32,rinv32);
1399             rinvsq33         = _mm_mul_pd(rinv33,rinv33);
1400
1401             fjx0             = _mm_setzero_pd();
1402             fjy0             = _mm_setzero_pd();
1403             fjz0             = _mm_setzero_pd();
1404             fjx1             = _mm_setzero_pd();
1405             fjy1             = _mm_setzero_pd();
1406             fjz1             = _mm_setzero_pd();
1407             fjx2             = _mm_setzero_pd();
1408             fjy2             = _mm_setzero_pd();
1409             fjz2             = _mm_setzero_pd();
1410             fjx3             = _mm_setzero_pd();
1411             fjy3             = _mm_setzero_pd();
1412             fjz3             = _mm_setzero_pd();
1413
1414             /**************************
1415              * CALCULATE INTERACTIONS *
1416              **************************/
1417
1418             r00              = _mm_mul_pd(rsq00,rinv00);
1419
1420             /* Calculate table index by multiplying r with table scale and truncate to integer */
1421             rt               = _mm_mul_pd(r00,vftabscale);
1422             vfitab           = _mm_cvttpd_epi32(rt);
1423 #ifdef __XOP__
1424             vfeps            = _mm_frcz_pd(rt);
1425 #else
1426             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1427 #endif
1428             twovfeps         = _mm_add_pd(vfeps,vfeps);
1429             vfitab           = _mm_slli_epi32(vfitab,3);
1430
1431             /* CUBIC SPLINE TABLE DISPERSION */
1432             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1433             F                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1434             GMX_MM_TRANSPOSE2_PD(Y,F);
1435             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1436             H                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1437             GMX_MM_TRANSPOSE2_PD(G,H);
1438             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(H,vfeps,G),F);
1439             FF               = _mm_macc_pd(vfeps,_mm_macc_pd(twovfeps,H,G),Fp);
1440             fvdw6            = _mm_mul_pd(c6_00,FF);
1441
1442             /* CUBIC SPLINE TABLE REPULSION */
1443             vfitab           = _mm_add_epi32(vfitab,ifour);
1444             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1445             F                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1446             GMX_MM_TRANSPOSE2_PD(Y,F);
1447             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1448             H                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1449             GMX_MM_TRANSPOSE2_PD(G,H);
1450             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(H,vfeps,G),F);
1451             FF               = _mm_macc_pd(vfeps,_mm_macc_pd(twovfeps,H,G),Fp);
1452             fvdw12           = _mm_mul_pd(c12_00,FF);
1453             fvdw             = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
1454
1455             fscal            = fvdw;
1456
1457             /* Update vectorial force */
1458             fix0             = _mm_macc_pd(dx00,fscal,fix0);
1459             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
1460             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
1461
1462             fjx0             = _mm_macc_pd(dx00,fscal,fjx0);
1463             fjy0             = _mm_macc_pd(dy00,fscal,fjy0);
1464             fjz0             = _mm_macc_pd(dz00,fscal,fjz0);
1465
1466             /**************************
1467              * CALCULATE INTERACTIONS *
1468              **************************/
1469
1470             if (gmx_mm_any_lt(rsq11,rcutoff2))
1471             {
1472
1473             /* REACTION-FIELD ELECTROSTATICS */
1474             felec            = _mm_mul_pd(qq11,_mm_msub_pd(rinv11,rinvsq11,krf2));
1475
1476             cutoff_mask      = _mm_cmplt_pd(rsq11,rcutoff2);
1477
1478             fscal            = felec;
1479
1480             fscal            = _mm_and_pd(fscal,cutoff_mask);
1481
1482             /* Update vectorial force */
1483             fix1             = _mm_macc_pd(dx11,fscal,fix1);
1484             fiy1             = _mm_macc_pd(dy11,fscal,fiy1);
1485             fiz1             = _mm_macc_pd(dz11,fscal,fiz1);
1486
1487             fjx1             = _mm_macc_pd(dx11,fscal,fjx1);
1488             fjy1             = _mm_macc_pd(dy11,fscal,fjy1);
1489             fjz1             = _mm_macc_pd(dz11,fscal,fjz1);
1490
1491             }
1492
1493             /**************************
1494              * CALCULATE INTERACTIONS *
1495              **************************/
1496
1497             if (gmx_mm_any_lt(rsq12,rcutoff2))
1498             {
1499
1500             /* REACTION-FIELD ELECTROSTATICS */
1501             felec            = _mm_mul_pd(qq12,_mm_msub_pd(rinv12,rinvsq12,krf2));
1502
1503             cutoff_mask      = _mm_cmplt_pd(rsq12,rcutoff2);
1504
1505             fscal            = felec;
1506
1507             fscal            = _mm_and_pd(fscal,cutoff_mask);
1508
1509             /* Update vectorial force */
1510             fix1             = _mm_macc_pd(dx12,fscal,fix1);
1511             fiy1             = _mm_macc_pd(dy12,fscal,fiy1);
1512             fiz1             = _mm_macc_pd(dz12,fscal,fiz1);
1513
1514             fjx2             = _mm_macc_pd(dx12,fscal,fjx2);
1515             fjy2             = _mm_macc_pd(dy12,fscal,fjy2);
1516             fjz2             = _mm_macc_pd(dz12,fscal,fjz2);
1517
1518             }
1519
1520             /**************************
1521              * CALCULATE INTERACTIONS *
1522              **************************/
1523
1524             if (gmx_mm_any_lt(rsq13,rcutoff2))
1525             {
1526
1527             /* REACTION-FIELD ELECTROSTATICS */
1528             felec            = _mm_mul_pd(qq13,_mm_msub_pd(rinv13,rinvsq13,krf2));
1529
1530             cutoff_mask      = _mm_cmplt_pd(rsq13,rcutoff2);
1531
1532             fscal            = felec;
1533
1534             fscal            = _mm_and_pd(fscal,cutoff_mask);
1535
1536             /* Update vectorial force */
1537             fix1             = _mm_macc_pd(dx13,fscal,fix1);
1538             fiy1             = _mm_macc_pd(dy13,fscal,fiy1);
1539             fiz1             = _mm_macc_pd(dz13,fscal,fiz1);
1540
1541             fjx3             = _mm_macc_pd(dx13,fscal,fjx3);
1542             fjy3             = _mm_macc_pd(dy13,fscal,fjy3);
1543             fjz3             = _mm_macc_pd(dz13,fscal,fjz3);
1544
1545             }
1546
1547             /**************************
1548              * CALCULATE INTERACTIONS *
1549              **************************/
1550
1551             if (gmx_mm_any_lt(rsq21,rcutoff2))
1552             {
1553
1554             /* REACTION-FIELD ELECTROSTATICS */
1555             felec            = _mm_mul_pd(qq21,_mm_msub_pd(rinv21,rinvsq21,krf2));
1556
1557             cutoff_mask      = _mm_cmplt_pd(rsq21,rcutoff2);
1558
1559             fscal            = felec;
1560
1561             fscal            = _mm_and_pd(fscal,cutoff_mask);
1562
1563             /* Update vectorial force */
1564             fix2             = _mm_macc_pd(dx21,fscal,fix2);
1565             fiy2             = _mm_macc_pd(dy21,fscal,fiy2);
1566             fiz2             = _mm_macc_pd(dz21,fscal,fiz2);
1567
1568             fjx1             = _mm_macc_pd(dx21,fscal,fjx1);
1569             fjy1             = _mm_macc_pd(dy21,fscal,fjy1);
1570             fjz1             = _mm_macc_pd(dz21,fscal,fjz1);
1571
1572             }
1573
1574             /**************************
1575              * CALCULATE INTERACTIONS *
1576              **************************/
1577
1578             if (gmx_mm_any_lt(rsq22,rcutoff2))
1579             {
1580
1581             /* REACTION-FIELD ELECTROSTATICS */
1582             felec            = _mm_mul_pd(qq22,_mm_msub_pd(rinv22,rinvsq22,krf2));
1583
1584             cutoff_mask      = _mm_cmplt_pd(rsq22,rcutoff2);
1585
1586             fscal            = felec;
1587
1588             fscal            = _mm_and_pd(fscal,cutoff_mask);
1589
1590             /* Update vectorial force */
1591             fix2             = _mm_macc_pd(dx22,fscal,fix2);
1592             fiy2             = _mm_macc_pd(dy22,fscal,fiy2);
1593             fiz2             = _mm_macc_pd(dz22,fscal,fiz2);
1594
1595             fjx2             = _mm_macc_pd(dx22,fscal,fjx2);
1596             fjy2             = _mm_macc_pd(dy22,fscal,fjy2);
1597             fjz2             = _mm_macc_pd(dz22,fscal,fjz2);
1598
1599             }
1600
1601             /**************************
1602              * CALCULATE INTERACTIONS *
1603              **************************/
1604
1605             if (gmx_mm_any_lt(rsq23,rcutoff2))
1606             {
1607
1608             /* REACTION-FIELD ELECTROSTATICS */
1609             felec            = _mm_mul_pd(qq23,_mm_msub_pd(rinv23,rinvsq23,krf2));
1610
1611             cutoff_mask      = _mm_cmplt_pd(rsq23,rcutoff2);
1612
1613             fscal            = felec;
1614
1615             fscal            = _mm_and_pd(fscal,cutoff_mask);
1616
1617             /* Update vectorial force */
1618             fix2             = _mm_macc_pd(dx23,fscal,fix2);
1619             fiy2             = _mm_macc_pd(dy23,fscal,fiy2);
1620             fiz2             = _mm_macc_pd(dz23,fscal,fiz2);
1621
1622             fjx3             = _mm_macc_pd(dx23,fscal,fjx3);
1623             fjy3             = _mm_macc_pd(dy23,fscal,fjy3);
1624             fjz3             = _mm_macc_pd(dz23,fscal,fjz3);
1625
1626             }
1627
1628             /**************************
1629              * CALCULATE INTERACTIONS *
1630              **************************/
1631
1632             if (gmx_mm_any_lt(rsq31,rcutoff2))
1633             {
1634
1635             /* REACTION-FIELD ELECTROSTATICS */
1636             felec            = _mm_mul_pd(qq31,_mm_msub_pd(rinv31,rinvsq31,krf2));
1637
1638             cutoff_mask      = _mm_cmplt_pd(rsq31,rcutoff2);
1639
1640             fscal            = felec;
1641
1642             fscal            = _mm_and_pd(fscal,cutoff_mask);
1643
1644             /* Update vectorial force */
1645             fix3             = _mm_macc_pd(dx31,fscal,fix3);
1646             fiy3             = _mm_macc_pd(dy31,fscal,fiy3);
1647             fiz3             = _mm_macc_pd(dz31,fscal,fiz3);
1648
1649             fjx1             = _mm_macc_pd(dx31,fscal,fjx1);
1650             fjy1             = _mm_macc_pd(dy31,fscal,fjy1);
1651             fjz1             = _mm_macc_pd(dz31,fscal,fjz1);
1652
1653             }
1654
1655             /**************************
1656              * CALCULATE INTERACTIONS *
1657              **************************/
1658
1659             if (gmx_mm_any_lt(rsq32,rcutoff2))
1660             {
1661
1662             /* REACTION-FIELD ELECTROSTATICS */
1663             felec            = _mm_mul_pd(qq32,_mm_msub_pd(rinv32,rinvsq32,krf2));
1664
1665             cutoff_mask      = _mm_cmplt_pd(rsq32,rcutoff2);
1666
1667             fscal            = felec;
1668
1669             fscal            = _mm_and_pd(fscal,cutoff_mask);
1670
1671             /* Update vectorial force */
1672             fix3             = _mm_macc_pd(dx32,fscal,fix3);
1673             fiy3             = _mm_macc_pd(dy32,fscal,fiy3);
1674             fiz3             = _mm_macc_pd(dz32,fscal,fiz3);
1675
1676             fjx2             = _mm_macc_pd(dx32,fscal,fjx2);
1677             fjy2             = _mm_macc_pd(dy32,fscal,fjy2);
1678             fjz2             = _mm_macc_pd(dz32,fscal,fjz2);
1679
1680             }
1681
1682             /**************************
1683              * CALCULATE INTERACTIONS *
1684              **************************/
1685
1686             if (gmx_mm_any_lt(rsq33,rcutoff2))
1687             {
1688
1689             /* REACTION-FIELD ELECTROSTATICS */
1690             felec            = _mm_mul_pd(qq33,_mm_msub_pd(rinv33,rinvsq33,krf2));
1691
1692             cutoff_mask      = _mm_cmplt_pd(rsq33,rcutoff2);
1693
1694             fscal            = felec;
1695
1696             fscal            = _mm_and_pd(fscal,cutoff_mask);
1697
1698             /* Update vectorial force */
1699             fix3             = _mm_macc_pd(dx33,fscal,fix3);
1700             fiy3             = _mm_macc_pd(dy33,fscal,fiy3);
1701             fiz3             = _mm_macc_pd(dz33,fscal,fiz3);
1702
1703             fjx3             = _mm_macc_pd(dx33,fscal,fjx3);
1704             fjy3             = _mm_macc_pd(dy33,fscal,fjy3);
1705             fjz3             = _mm_macc_pd(dz33,fscal,fjz3);
1706
1707             }
1708
1709             gmx_mm_decrement_4rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1710
1711             /* Inner loop uses 351 flops */
1712         }
1713
1714         if(jidx<j_index_end)
1715         {
1716
1717             jnrA             = jjnr[jidx];
1718             j_coord_offsetA  = DIM*jnrA;
1719
1720             /* load j atom coordinates */
1721             gmx_mm_load_4rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
1722                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1723                                               &jy2,&jz2,&jx3,&jy3,&jz3);
1724
1725             /* Calculate displacement vector */
1726             dx00             = _mm_sub_pd(ix0,jx0);
1727             dy00             = _mm_sub_pd(iy0,jy0);
1728             dz00             = _mm_sub_pd(iz0,jz0);
1729             dx11             = _mm_sub_pd(ix1,jx1);
1730             dy11             = _mm_sub_pd(iy1,jy1);
1731             dz11             = _mm_sub_pd(iz1,jz1);
1732             dx12             = _mm_sub_pd(ix1,jx2);
1733             dy12             = _mm_sub_pd(iy1,jy2);
1734             dz12             = _mm_sub_pd(iz1,jz2);
1735             dx13             = _mm_sub_pd(ix1,jx3);
1736             dy13             = _mm_sub_pd(iy1,jy3);
1737             dz13             = _mm_sub_pd(iz1,jz3);
1738             dx21             = _mm_sub_pd(ix2,jx1);
1739             dy21             = _mm_sub_pd(iy2,jy1);
1740             dz21             = _mm_sub_pd(iz2,jz1);
1741             dx22             = _mm_sub_pd(ix2,jx2);
1742             dy22             = _mm_sub_pd(iy2,jy2);
1743             dz22             = _mm_sub_pd(iz2,jz2);
1744             dx23             = _mm_sub_pd(ix2,jx3);
1745             dy23             = _mm_sub_pd(iy2,jy3);
1746             dz23             = _mm_sub_pd(iz2,jz3);
1747             dx31             = _mm_sub_pd(ix3,jx1);
1748             dy31             = _mm_sub_pd(iy3,jy1);
1749             dz31             = _mm_sub_pd(iz3,jz1);
1750             dx32             = _mm_sub_pd(ix3,jx2);
1751             dy32             = _mm_sub_pd(iy3,jy2);
1752             dz32             = _mm_sub_pd(iz3,jz2);
1753             dx33             = _mm_sub_pd(ix3,jx3);
1754             dy33             = _mm_sub_pd(iy3,jy3);
1755             dz33             = _mm_sub_pd(iz3,jz3);
1756
1757             /* Calculate squared distance and things based on it */
1758             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1759             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1760             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1761             rsq13            = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
1762             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1763             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1764             rsq23            = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
1765             rsq31            = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
1766             rsq32            = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
1767             rsq33            = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
1768
1769             rinv00           = gmx_mm_invsqrt_pd(rsq00);
1770             rinv11           = gmx_mm_invsqrt_pd(rsq11);
1771             rinv12           = gmx_mm_invsqrt_pd(rsq12);
1772             rinv13           = gmx_mm_invsqrt_pd(rsq13);
1773             rinv21           = gmx_mm_invsqrt_pd(rsq21);
1774             rinv22           = gmx_mm_invsqrt_pd(rsq22);
1775             rinv23           = gmx_mm_invsqrt_pd(rsq23);
1776             rinv31           = gmx_mm_invsqrt_pd(rsq31);
1777             rinv32           = gmx_mm_invsqrt_pd(rsq32);
1778             rinv33           = gmx_mm_invsqrt_pd(rsq33);
1779
1780             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
1781             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
1782             rinvsq13         = _mm_mul_pd(rinv13,rinv13);
1783             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
1784             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
1785             rinvsq23         = _mm_mul_pd(rinv23,rinv23);
1786             rinvsq31         = _mm_mul_pd(rinv31,rinv31);
1787             rinvsq32         = _mm_mul_pd(rinv32,rinv32);
1788             rinvsq33         = _mm_mul_pd(rinv33,rinv33);
1789
1790             fjx0             = _mm_setzero_pd();
1791             fjy0             = _mm_setzero_pd();
1792             fjz0             = _mm_setzero_pd();
1793             fjx1             = _mm_setzero_pd();
1794             fjy1             = _mm_setzero_pd();
1795             fjz1             = _mm_setzero_pd();
1796             fjx2             = _mm_setzero_pd();
1797             fjy2             = _mm_setzero_pd();
1798             fjz2             = _mm_setzero_pd();
1799             fjx3             = _mm_setzero_pd();
1800             fjy3             = _mm_setzero_pd();
1801             fjz3             = _mm_setzero_pd();
1802
1803             /**************************
1804              * CALCULATE INTERACTIONS *
1805              **************************/
1806
1807             r00              = _mm_mul_pd(rsq00,rinv00);
1808
1809             /* Calculate table index by multiplying r with table scale and truncate to integer */
1810             rt               = _mm_mul_pd(r00,vftabscale);
1811             vfitab           = _mm_cvttpd_epi32(rt);
1812 #ifdef __XOP__
1813             vfeps            = _mm_frcz_pd(rt);
1814 #else
1815             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1816 #endif
1817             twovfeps         = _mm_add_pd(vfeps,vfeps);
1818             vfitab           = _mm_slli_epi32(vfitab,3);
1819
1820             /* CUBIC SPLINE TABLE DISPERSION */
1821             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1822             F                = _mm_setzero_pd();
1823             GMX_MM_TRANSPOSE2_PD(Y,F);
1824             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1825             H                = _mm_setzero_pd();
1826             GMX_MM_TRANSPOSE2_PD(G,H);
1827             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(H,vfeps,G),F);
1828             FF               = _mm_macc_pd(vfeps,_mm_macc_pd(twovfeps,H,G),Fp);
1829             fvdw6            = _mm_mul_pd(c6_00,FF);
1830
1831             /* CUBIC SPLINE TABLE REPULSION */
1832             vfitab           = _mm_add_epi32(vfitab,ifour);
1833             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1834             F                = _mm_setzero_pd();
1835             GMX_MM_TRANSPOSE2_PD(Y,F);
1836             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1837             H                = _mm_setzero_pd();
1838             GMX_MM_TRANSPOSE2_PD(G,H);
1839             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(H,vfeps,G),F);
1840             FF               = _mm_macc_pd(vfeps,_mm_macc_pd(twovfeps,H,G),Fp);
1841             fvdw12           = _mm_mul_pd(c12_00,FF);
1842             fvdw             = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
1843
1844             fscal            = fvdw;
1845
1846             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1847
1848             /* Update vectorial force */
1849             fix0             = _mm_macc_pd(dx00,fscal,fix0);
1850             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
1851             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
1852
1853             fjx0             = _mm_macc_pd(dx00,fscal,fjx0);
1854             fjy0             = _mm_macc_pd(dy00,fscal,fjy0);
1855             fjz0             = _mm_macc_pd(dz00,fscal,fjz0);
1856
1857             /**************************
1858              * CALCULATE INTERACTIONS *
1859              **************************/
1860
1861             if (gmx_mm_any_lt(rsq11,rcutoff2))
1862             {
1863
1864             /* REACTION-FIELD ELECTROSTATICS */
1865             felec            = _mm_mul_pd(qq11,_mm_msub_pd(rinv11,rinvsq11,krf2));
1866
1867             cutoff_mask      = _mm_cmplt_pd(rsq11,rcutoff2);
1868
1869             fscal            = felec;
1870
1871             fscal            = _mm_and_pd(fscal,cutoff_mask);
1872
1873             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1874
1875             /* Update vectorial force */
1876             fix1             = _mm_macc_pd(dx11,fscal,fix1);
1877             fiy1             = _mm_macc_pd(dy11,fscal,fiy1);
1878             fiz1             = _mm_macc_pd(dz11,fscal,fiz1);
1879
1880             fjx1             = _mm_macc_pd(dx11,fscal,fjx1);
1881             fjy1             = _mm_macc_pd(dy11,fscal,fjy1);
1882             fjz1             = _mm_macc_pd(dz11,fscal,fjz1);
1883
1884             }
1885
1886             /**************************
1887              * CALCULATE INTERACTIONS *
1888              **************************/
1889
1890             if (gmx_mm_any_lt(rsq12,rcutoff2))
1891             {
1892
1893             /* REACTION-FIELD ELECTROSTATICS */
1894             felec            = _mm_mul_pd(qq12,_mm_msub_pd(rinv12,rinvsq12,krf2));
1895
1896             cutoff_mask      = _mm_cmplt_pd(rsq12,rcutoff2);
1897
1898             fscal            = felec;
1899
1900             fscal            = _mm_and_pd(fscal,cutoff_mask);
1901
1902             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1903
1904             /* Update vectorial force */
1905             fix1             = _mm_macc_pd(dx12,fscal,fix1);
1906             fiy1             = _mm_macc_pd(dy12,fscal,fiy1);
1907             fiz1             = _mm_macc_pd(dz12,fscal,fiz1);
1908
1909             fjx2             = _mm_macc_pd(dx12,fscal,fjx2);
1910             fjy2             = _mm_macc_pd(dy12,fscal,fjy2);
1911             fjz2             = _mm_macc_pd(dz12,fscal,fjz2);
1912
1913             }
1914
1915             /**************************
1916              * CALCULATE INTERACTIONS *
1917              **************************/
1918
1919             if (gmx_mm_any_lt(rsq13,rcutoff2))
1920             {
1921
1922             /* REACTION-FIELD ELECTROSTATICS */
1923             felec            = _mm_mul_pd(qq13,_mm_msub_pd(rinv13,rinvsq13,krf2));
1924
1925             cutoff_mask      = _mm_cmplt_pd(rsq13,rcutoff2);
1926
1927             fscal            = felec;
1928
1929             fscal            = _mm_and_pd(fscal,cutoff_mask);
1930
1931             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1932
1933             /* Update vectorial force */
1934             fix1             = _mm_macc_pd(dx13,fscal,fix1);
1935             fiy1             = _mm_macc_pd(dy13,fscal,fiy1);
1936             fiz1             = _mm_macc_pd(dz13,fscal,fiz1);
1937
1938             fjx3             = _mm_macc_pd(dx13,fscal,fjx3);
1939             fjy3             = _mm_macc_pd(dy13,fscal,fjy3);
1940             fjz3             = _mm_macc_pd(dz13,fscal,fjz3);
1941
1942             }
1943
1944             /**************************
1945              * CALCULATE INTERACTIONS *
1946              **************************/
1947
1948             if (gmx_mm_any_lt(rsq21,rcutoff2))
1949             {
1950
1951             /* REACTION-FIELD ELECTROSTATICS */
1952             felec            = _mm_mul_pd(qq21,_mm_msub_pd(rinv21,rinvsq21,krf2));
1953
1954             cutoff_mask      = _mm_cmplt_pd(rsq21,rcutoff2);
1955
1956             fscal            = felec;
1957
1958             fscal            = _mm_and_pd(fscal,cutoff_mask);
1959
1960             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1961
1962             /* Update vectorial force */
1963             fix2             = _mm_macc_pd(dx21,fscal,fix2);
1964             fiy2             = _mm_macc_pd(dy21,fscal,fiy2);
1965             fiz2             = _mm_macc_pd(dz21,fscal,fiz2);
1966
1967             fjx1             = _mm_macc_pd(dx21,fscal,fjx1);
1968             fjy1             = _mm_macc_pd(dy21,fscal,fjy1);
1969             fjz1             = _mm_macc_pd(dz21,fscal,fjz1);
1970
1971             }
1972
1973             /**************************
1974              * CALCULATE INTERACTIONS *
1975              **************************/
1976
1977             if (gmx_mm_any_lt(rsq22,rcutoff2))
1978             {
1979
1980             /* REACTION-FIELD ELECTROSTATICS */
1981             felec            = _mm_mul_pd(qq22,_mm_msub_pd(rinv22,rinvsq22,krf2));
1982
1983             cutoff_mask      = _mm_cmplt_pd(rsq22,rcutoff2);
1984
1985             fscal            = felec;
1986
1987             fscal            = _mm_and_pd(fscal,cutoff_mask);
1988
1989             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1990
1991             /* Update vectorial force */
1992             fix2             = _mm_macc_pd(dx22,fscal,fix2);
1993             fiy2             = _mm_macc_pd(dy22,fscal,fiy2);
1994             fiz2             = _mm_macc_pd(dz22,fscal,fiz2);
1995
1996             fjx2             = _mm_macc_pd(dx22,fscal,fjx2);
1997             fjy2             = _mm_macc_pd(dy22,fscal,fjy2);
1998             fjz2             = _mm_macc_pd(dz22,fscal,fjz2);
1999
2000             }
2001
2002             /**************************
2003              * CALCULATE INTERACTIONS *
2004              **************************/
2005
2006             if (gmx_mm_any_lt(rsq23,rcutoff2))
2007             {
2008
2009             /* REACTION-FIELD ELECTROSTATICS */
2010             felec            = _mm_mul_pd(qq23,_mm_msub_pd(rinv23,rinvsq23,krf2));
2011
2012             cutoff_mask      = _mm_cmplt_pd(rsq23,rcutoff2);
2013
2014             fscal            = felec;
2015
2016             fscal            = _mm_and_pd(fscal,cutoff_mask);
2017
2018             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2019
2020             /* Update vectorial force */
2021             fix2             = _mm_macc_pd(dx23,fscal,fix2);
2022             fiy2             = _mm_macc_pd(dy23,fscal,fiy2);
2023             fiz2             = _mm_macc_pd(dz23,fscal,fiz2);
2024
2025             fjx3             = _mm_macc_pd(dx23,fscal,fjx3);
2026             fjy3             = _mm_macc_pd(dy23,fscal,fjy3);
2027             fjz3             = _mm_macc_pd(dz23,fscal,fjz3);
2028
2029             }
2030
2031             /**************************
2032              * CALCULATE INTERACTIONS *
2033              **************************/
2034
2035             if (gmx_mm_any_lt(rsq31,rcutoff2))
2036             {
2037
2038             /* REACTION-FIELD ELECTROSTATICS */
2039             felec            = _mm_mul_pd(qq31,_mm_msub_pd(rinv31,rinvsq31,krf2));
2040
2041             cutoff_mask      = _mm_cmplt_pd(rsq31,rcutoff2);
2042
2043             fscal            = felec;
2044
2045             fscal            = _mm_and_pd(fscal,cutoff_mask);
2046
2047             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2048
2049             /* Update vectorial force */
2050             fix3             = _mm_macc_pd(dx31,fscal,fix3);
2051             fiy3             = _mm_macc_pd(dy31,fscal,fiy3);
2052             fiz3             = _mm_macc_pd(dz31,fscal,fiz3);
2053
2054             fjx1             = _mm_macc_pd(dx31,fscal,fjx1);
2055             fjy1             = _mm_macc_pd(dy31,fscal,fjy1);
2056             fjz1             = _mm_macc_pd(dz31,fscal,fjz1);
2057
2058             }
2059
2060             /**************************
2061              * CALCULATE INTERACTIONS *
2062              **************************/
2063
2064             if (gmx_mm_any_lt(rsq32,rcutoff2))
2065             {
2066
2067             /* REACTION-FIELD ELECTROSTATICS */
2068             felec            = _mm_mul_pd(qq32,_mm_msub_pd(rinv32,rinvsq32,krf2));
2069
2070             cutoff_mask      = _mm_cmplt_pd(rsq32,rcutoff2);
2071
2072             fscal            = felec;
2073
2074             fscal            = _mm_and_pd(fscal,cutoff_mask);
2075
2076             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2077
2078             /* Update vectorial force */
2079             fix3             = _mm_macc_pd(dx32,fscal,fix3);
2080             fiy3             = _mm_macc_pd(dy32,fscal,fiy3);
2081             fiz3             = _mm_macc_pd(dz32,fscal,fiz3);
2082
2083             fjx2             = _mm_macc_pd(dx32,fscal,fjx2);
2084             fjy2             = _mm_macc_pd(dy32,fscal,fjy2);
2085             fjz2             = _mm_macc_pd(dz32,fscal,fjz2);
2086
2087             }
2088
2089             /**************************
2090              * CALCULATE INTERACTIONS *
2091              **************************/
2092
2093             if (gmx_mm_any_lt(rsq33,rcutoff2))
2094             {
2095
2096             /* REACTION-FIELD ELECTROSTATICS */
2097             felec            = _mm_mul_pd(qq33,_mm_msub_pd(rinv33,rinvsq33,krf2));
2098
2099             cutoff_mask      = _mm_cmplt_pd(rsq33,rcutoff2);
2100
2101             fscal            = felec;
2102
2103             fscal            = _mm_and_pd(fscal,cutoff_mask);
2104
2105             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2106
2107             /* Update vectorial force */
2108             fix3             = _mm_macc_pd(dx33,fscal,fix3);
2109             fiy3             = _mm_macc_pd(dy33,fscal,fiy3);
2110             fiz3             = _mm_macc_pd(dz33,fscal,fiz3);
2111
2112             fjx3             = _mm_macc_pd(dx33,fscal,fjx3);
2113             fjy3             = _mm_macc_pd(dy33,fscal,fjy3);
2114             fjz3             = _mm_macc_pd(dz33,fscal,fjz3);
2115
2116             }
2117
2118             gmx_mm_decrement_4rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2119
2120             /* Inner loop uses 351 flops */
2121         }
2122
2123         /* End of innermost loop */
2124
2125         gmx_mm_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2126                                               f+i_coord_offset,fshift+i_shift_offset);
2127
2128         /* Increment number of inner iterations */
2129         inneriter                  += j_index_end - j_index_start;
2130
2131         /* Outer loop uses 24 flops */
2132     }
2133
2134     /* Increment number of outer iterations */
2135     outeriter        += nri;
2136
2137     /* Update outer/inner flops */
2138
2139     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*351);
2140 }