src/gromacs/gmxlib/nonbonded/nb_kernel_sse4_1_double/nb_kernel_ElecCSTab_VdwNone_GeomW3W3_sse4_1_double.cpp

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014,2015,2017,2018, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS sse4_1_double kernel generator.
  37  */
  38 #include "gmxpre.h"
  39
  40 #include "config.h"
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "gromacs/gmxlib/nrnb.h"
  46
  47 #include "kernelutil_x86_sse4_1_double.h"
  48
  49 /*
  50  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwNone_GeomW3W3_VF_sse4_1_double
  51  * Electrostatics interaction: CubicSplineTable
  52  * VdW interaction:            None
  53  * Geometry:                   Water3-Water3
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecCSTab_VdwNone_GeomW3W3_VF_sse4_1_double
  58                     (t_nblist                    * gmx_restrict       nlist,
  59                      rvec                        * gmx_restrict          xx,
  60                      rvec                        * gmx_restrict          ff,
  61                      struct t_forcerec           * gmx_restrict          fr,
  62                      t_mdatoms                   * gmx_restrict     mdatoms,
  63                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  64                      t_nrnb                      * gmx_restrict        nrnb)
  65 {
  66     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  67      * just 0 for non-waters.
  68      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
  69      * jnr indices corresponding to data put in the four positions in the SIMD register.
  70      */
  71     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  72     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  73     int              jnrA,jnrB;
  74     int              j_coord_offsetA,j_coord_offsetB;
  75     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  76     real             rcutoff_scalar;
  77     real             *shiftvec,*fshift,*x,*f;
  78     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  79     int              vdwioffset0;
  80     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  81     int              vdwioffset1;
  82     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  83     int              vdwioffset2;
  84     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  85     int              vdwjidx0A,vdwjidx0B;
  86     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  87     int              vdwjidx1A,vdwjidx1B;
  88     __m128d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  89     int              vdwjidx2A,vdwjidx2B;
  90     __m128d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  91     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  92     __m128d          dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
  93     __m128d          dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
  94     __m128d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  95     __m128d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
  96     __m128d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
  97     __m128d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  98     __m128d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
  99     __m128d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 100     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
 101     real             *charge;
 102     __m128i          vfitab;
 103     __m128i          ifour       = _mm_set1_epi32(4);
 104     __m128d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
 105     real             *vftab;
 106     __m128d          dummy_mask,cutoff_mask;
 107     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 108     __m128d          one     = _mm_set1_pd(1.0);
 109     __m128d          two     = _mm_set1_pd(2.0);
 110     x                = xx[0];
 111     f                = ff[0];
 112
 113     nri              = nlist->nri;
 114     iinr             = nlist->iinr;
 115     jindex           = nlist->jindex;
 116     jjnr             = nlist->jjnr;
 117     shiftidx         = nlist->shift;
 118     gid              = nlist->gid;
 119     shiftvec         = fr->shift_vec[0];
 120     fshift           = fr->fshift[0];
 121     facel            = _mm_set1_pd(fr->ic->epsfac);
 122     charge           = mdatoms->chargeA;
 123
 124     vftab            = kernel_data->table_elec->data;
 125     vftabscale       = _mm_set1_pd(kernel_data->table_elec->scale);
 126
 127     /* Setup water-specific parameters */
 128     inr              = nlist->iinr[0];
 129     iq0              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
 130     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
 131     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
 132
 133     jq0              = _mm_set1_pd(charge[inr+0]);
 134     jq1              = _mm_set1_pd(charge[inr+1]);
 135     jq2              = _mm_set1_pd(charge[inr+2]);
 136     qq00             = _mm_mul_pd(iq0,jq0);
 137     qq01             = _mm_mul_pd(iq0,jq1);
 138     qq02             = _mm_mul_pd(iq0,jq2);
 139     qq10             = _mm_mul_pd(iq1,jq0);
 140     qq11             = _mm_mul_pd(iq1,jq1);
 141     qq12             = _mm_mul_pd(iq1,jq2);
 142     qq20             = _mm_mul_pd(iq2,jq0);
 143     qq21             = _mm_mul_pd(iq2,jq1);
 144     qq22             = _mm_mul_pd(iq2,jq2);
 145
 146     /* Avoid stupid compiler warnings */
 147     jnrA = jnrB = 0;
 148     j_coord_offsetA = 0;
 149     j_coord_offsetB = 0;
 150
 151     outeriter        = 0;
 152     inneriter        = 0;
 153
 154     /* Start outer loop over neighborlists */
 155     for(iidx=0; iidx<nri; iidx++)
 156     {
 157         /* Load shift vector for this list */
 158         i_shift_offset   = DIM*shiftidx[iidx];
 159
 160         /* Load limits for loop over neighbors */
 161         j_index_start    = jindex[iidx];
 162         j_index_end      = jindex[iidx+1];
 163
 164         /* Get outer coordinate index */
 165         inr              = iinr[iidx];
 166         i_coord_offset   = DIM*inr;
 167
 168         /* Load i particle coords and add shift vector */
 169         gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 170                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 171
 172         fix0             = _mm_setzero_pd();
 173         fiy0             = _mm_setzero_pd();
 174         fiz0             = _mm_setzero_pd();
 175         fix1             = _mm_setzero_pd();
 176         fiy1             = _mm_setzero_pd();
 177         fiz1             = _mm_setzero_pd();
 178         fix2             = _mm_setzero_pd();
 179         fiy2             = _mm_setzero_pd();
 180         fiz2             = _mm_setzero_pd();
 181
 182         /* Reset potential sums */
 183         velecsum         = _mm_setzero_pd();
 184
 185         /* Start inner kernel loop */
 186         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 187         {
 188
 189             /* Get j neighbor index, and coordinate index */
 190             jnrA             = jjnr[jidx];
 191             jnrB             = jjnr[jidx+1];
 192             j_coord_offsetA  = DIM*jnrA;
 193             j_coord_offsetB  = DIM*jnrB;
 194
 195             /* load j atom coordinates */
 196             gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 197                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 198
 199             /* Calculate displacement vector */
 200             dx00             = _mm_sub_pd(ix0,jx0);
 201             dy00             = _mm_sub_pd(iy0,jy0);
 202             dz00             = _mm_sub_pd(iz0,jz0);
 203             dx01             = _mm_sub_pd(ix0,jx1);
 204             dy01             = _mm_sub_pd(iy0,jy1);
 205             dz01             = _mm_sub_pd(iz0,jz1);
 206             dx02             = _mm_sub_pd(ix0,jx2);
 207             dy02             = _mm_sub_pd(iy0,jy2);
 208             dz02             = _mm_sub_pd(iz0,jz2);
 209             dx10             = _mm_sub_pd(ix1,jx0);
 210             dy10             = _mm_sub_pd(iy1,jy0);
 211             dz10             = _mm_sub_pd(iz1,jz0);
 212             dx11             = _mm_sub_pd(ix1,jx1);
 213             dy11             = _mm_sub_pd(iy1,jy1);
 214             dz11             = _mm_sub_pd(iz1,jz1);
 215             dx12             = _mm_sub_pd(ix1,jx2);
 216             dy12             = _mm_sub_pd(iy1,jy2);
 217             dz12             = _mm_sub_pd(iz1,jz2);
 218             dx20             = _mm_sub_pd(ix2,jx0);
 219             dy20             = _mm_sub_pd(iy2,jy0);
 220             dz20             = _mm_sub_pd(iz2,jz0);
 221             dx21             = _mm_sub_pd(ix2,jx1);
 222             dy21             = _mm_sub_pd(iy2,jy1);
 223             dz21             = _mm_sub_pd(iz2,jz1);
 224             dx22             = _mm_sub_pd(ix2,jx2);
 225             dy22             = _mm_sub_pd(iy2,jy2);
 226             dz22             = _mm_sub_pd(iz2,jz2);
 227
 228             /* Calculate squared distance and things based on it */
 229             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 230             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
 231             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
 232             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 233             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 234             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 235             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 236             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 237             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 238
 239             rinv00           = sse41_invsqrt_d(rsq00);
 240             rinv01           = sse41_invsqrt_d(rsq01);
 241             rinv02           = sse41_invsqrt_d(rsq02);
 242             rinv10           = sse41_invsqrt_d(rsq10);
 243             rinv11           = sse41_invsqrt_d(rsq11);
 244             rinv12           = sse41_invsqrt_d(rsq12);
 245             rinv20           = sse41_invsqrt_d(rsq20);
 246             rinv21           = sse41_invsqrt_d(rsq21);
 247             rinv22           = sse41_invsqrt_d(rsq22);
 248
 249             fjx0             = _mm_setzero_pd();
 250             fjy0             = _mm_setzero_pd();
 251             fjz0             = _mm_setzero_pd();
 252             fjx1             = _mm_setzero_pd();
 253             fjy1             = _mm_setzero_pd();
 254             fjz1             = _mm_setzero_pd();
 255             fjx2             = _mm_setzero_pd();
 256             fjy2             = _mm_setzero_pd();
 257             fjz2             = _mm_setzero_pd();
 258
 259             /**************************
 260              * CALCULATE INTERACTIONS *
 261              **************************/
 262
 263             r00              = _mm_mul_pd(rsq00,rinv00);
 264
 265             /* Calculate table index by multiplying r with table scale and truncate to integer */
 266             rt               = _mm_mul_pd(r00,vftabscale);
 267             vfitab           = _mm_cvttpd_epi32(rt);
 268             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 269             vfitab           = _mm_slli_epi32(vfitab,2);
 270
 271             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 272             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 273             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 274             GMX_MM_TRANSPOSE2_PD(Y,F);
 275             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 276             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 277             GMX_MM_TRANSPOSE2_PD(G,H);
 278             Heps             = _mm_mul_pd(vfeps,H);
 279             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 280             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 281             velec            = _mm_mul_pd(qq00,VV);
 282             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 283             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq00,FF),_mm_mul_pd(vftabscale,rinv00)));
 284
 285             /* Update potential sum for this i atom from the interaction with this j atom. */
 286             velecsum         = _mm_add_pd(velecsum,velec);
 287
 288             fscal            = felec;
 289
 290             /* Calculate temporary vectorial force */
 291             tx               = _mm_mul_pd(fscal,dx00);
 292             ty               = _mm_mul_pd(fscal,dy00);
 293             tz               = _mm_mul_pd(fscal,dz00);
 294
 295             /* Update vectorial force */
 296             fix0             = _mm_add_pd(fix0,tx);
 297             fiy0             = _mm_add_pd(fiy0,ty);
 298             fiz0             = _mm_add_pd(fiz0,tz);
 299
 300             fjx0             = _mm_add_pd(fjx0,tx);
 301             fjy0             = _mm_add_pd(fjy0,ty);
 302             fjz0             = _mm_add_pd(fjz0,tz);
 303
 304             /**************************
 305              * CALCULATE INTERACTIONS *
 306              **************************/
 307
 308             r01              = _mm_mul_pd(rsq01,rinv01);
 309
 310             /* Calculate table index by multiplying r with table scale and truncate to integer */
 311             rt               = _mm_mul_pd(r01,vftabscale);
 312             vfitab           = _mm_cvttpd_epi32(rt);
 313             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 314             vfitab           = _mm_slli_epi32(vfitab,2);
 315
 316             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 317             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 318             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 319             GMX_MM_TRANSPOSE2_PD(Y,F);
 320             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 321             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 322             GMX_MM_TRANSPOSE2_PD(G,H);
 323             Heps             = _mm_mul_pd(vfeps,H);
 324             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 325             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 326             velec            = _mm_mul_pd(qq01,VV);
 327             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 328             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq01,FF),_mm_mul_pd(vftabscale,rinv01)));
 329
 330             /* Update potential sum for this i atom from the interaction with this j atom. */
 331             velecsum         = _mm_add_pd(velecsum,velec);
 332
 333             fscal            = felec;
 334
 335             /* Calculate temporary vectorial force */
 336             tx               = _mm_mul_pd(fscal,dx01);
 337             ty               = _mm_mul_pd(fscal,dy01);
 338             tz               = _mm_mul_pd(fscal,dz01);
 339
 340             /* Update vectorial force */
 341             fix0             = _mm_add_pd(fix0,tx);
 342             fiy0             = _mm_add_pd(fiy0,ty);
 343             fiz0             = _mm_add_pd(fiz0,tz);
 344
 345             fjx1             = _mm_add_pd(fjx1,tx);
 346             fjy1             = _mm_add_pd(fjy1,ty);
 347             fjz1             = _mm_add_pd(fjz1,tz);
 348
 349             /**************************
 350              * CALCULATE INTERACTIONS *
 351              **************************/
 352
 353             r02              = _mm_mul_pd(rsq02,rinv02);
 354
 355             /* Calculate table index by multiplying r with table scale and truncate to integer */
 356             rt               = _mm_mul_pd(r02,vftabscale);
 357             vfitab           = _mm_cvttpd_epi32(rt);
 358             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 359             vfitab           = _mm_slli_epi32(vfitab,2);
 360
 361             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 362             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 363             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 364             GMX_MM_TRANSPOSE2_PD(Y,F);
 365             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 366             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 367             GMX_MM_TRANSPOSE2_PD(G,H);
 368             Heps             = _mm_mul_pd(vfeps,H);
 369             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 370             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 371             velec            = _mm_mul_pd(qq02,VV);
 372             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 373             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq02,FF),_mm_mul_pd(vftabscale,rinv02)));
 374
 375             /* Update potential sum for this i atom from the interaction with this j atom. */
 376             velecsum         = _mm_add_pd(velecsum,velec);
 377
 378             fscal            = felec;
 379
 380             /* Calculate temporary vectorial force */
 381             tx               = _mm_mul_pd(fscal,dx02);
 382             ty               = _mm_mul_pd(fscal,dy02);
 383             tz               = _mm_mul_pd(fscal,dz02);
 384
 385             /* Update vectorial force */
 386             fix0             = _mm_add_pd(fix0,tx);
 387             fiy0             = _mm_add_pd(fiy0,ty);
 388             fiz0             = _mm_add_pd(fiz0,tz);
 389
 390             fjx2             = _mm_add_pd(fjx2,tx);
 391             fjy2             = _mm_add_pd(fjy2,ty);
 392             fjz2             = _mm_add_pd(fjz2,tz);
 393
 394             /**************************
 395              * CALCULATE INTERACTIONS *
 396              **************************/
 397
 398             r10              = _mm_mul_pd(rsq10,rinv10);
 399
 400             /* Calculate table index by multiplying r with table scale and truncate to integer */
 401             rt               = _mm_mul_pd(r10,vftabscale);
 402             vfitab           = _mm_cvttpd_epi32(rt);
 403             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 404             vfitab           = _mm_slli_epi32(vfitab,2);
 405
 406             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 407             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 408             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 409             GMX_MM_TRANSPOSE2_PD(Y,F);
 410             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 411             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 412             GMX_MM_TRANSPOSE2_PD(G,H);
 413             Heps             = _mm_mul_pd(vfeps,H);
 414             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 415             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 416             velec            = _mm_mul_pd(qq10,VV);
 417             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 418             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq10,FF),_mm_mul_pd(vftabscale,rinv10)));
 419
 420             /* Update potential sum for this i atom from the interaction with this j atom. */
 421             velecsum         = _mm_add_pd(velecsum,velec);
 422
 423             fscal            = felec;
 424
 425             /* Calculate temporary vectorial force */
 426             tx               = _mm_mul_pd(fscal,dx10);
 427             ty               = _mm_mul_pd(fscal,dy10);
 428             tz               = _mm_mul_pd(fscal,dz10);
 429
 430             /* Update vectorial force */
 431             fix1             = _mm_add_pd(fix1,tx);
 432             fiy1             = _mm_add_pd(fiy1,ty);
 433             fiz1             = _mm_add_pd(fiz1,tz);
 434
 435             fjx0             = _mm_add_pd(fjx0,tx);
 436             fjy0             = _mm_add_pd(fjy0,ty);
 437             fjz0             = _mm_add_pd(fjz0,tz);
 438
 439             /**************************
 440              * CALCULATE INTERACTIONS *
 441              **************************/
 442
 443             r11              = _mm_mul_pd(rsq11,rinv11);
 444
 445             /* Calculate table index by multiplying r with table scale and truncate to integer */
 446             rt               = _mm_mul_pd(r11,vftabscale);
 447             vfitab           = _mm_cvttpd_epi32(rt);
 448             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 449             vfitab           = _mm_slli_epi32(vfitab,2);
 450
 451             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 452             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 453             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 454             GMX_MM_TRANSPOSE2_PD(Y,F);
 455             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 456             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 457             GMX_MM_TRANSPOSE2_PD(G,H);
 458             Heps             = _mm_mul_pd(vfeps,H);
 459             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 460             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 461             velec            = _mm_mul_pd(qq11,VV);
 462             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 463             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
 464
 465             /* Update potential sum for this i atom from the interaction with this j atom. */
 466             velecsum         = _mm_add_pd(velecsum,velec);
 467
 468             fscal            = felec;
 469
 470             /* Calculate temporary vectorial force */
 471             tx               = _mm_mul_pd(fscal,dx11);
 472             ty               = _mm_mul_pd(fscal,dy11);
 473             tz               = _mm_mul_pd(fscal,dz11);
 474
 475             /* Update vectorial force */
 476             fix1             = _mm_add_pd(fix1,tx);
 477             fiy1             = _mm_add_pd(fiy1,ty);
 478             fiz1             = _mm_add_pd(fiz1,tz);
 479
 480             fjx1             = _mm_add_pd(fjx1,tx);
 481             fjy1             = _mm_add_pd(fjy1,ty);
 482             fjz1             = _mm_add_pd(fjz1,tz);
 483
 484             /**************************
 485              * CALCULATE INTERACTIONS *
 486              **************************/
 487
 488             r12              = _mm_mul_pd(rsq12,rinv12);
 489
 490             /* Calculate table index by multiplying r with table scale and truncate to integer */
 491             rt               = _mm_mul_pd(r12,vftabscale);
 492             vfitab           = _mm_cvttpd_epi32(rt);
 493             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 494             vfitab           = _mm_slli_epi32(vfitab,2);
 495
 496             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 497             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 498             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 499             GMX_MM_TRANSPOSE2_PD(Y,F);
 500             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 501             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 502             GMX_MM_TRANSPOSE2_PD(G,H);
 503             Heps             = _mm_mul_pd(vfeps,H);
 504             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 505             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 506             velec            = _mm_mul_pd(qq12,VV);
 507             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 508             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
 509
 510             /* Update potential sum for this i atom from the interaction with this j atom. */
 511             velecsum         = _mm_add_pd(velecsum,velec);
 512
 513             fscal            = felec;
 514
 515             /* Calculate temporary vectorial force */
 516             tx               = _mm_mul_pd(fscal,dx12);
 517             ty               = _mm_mul_pd(fscal,dy12);
 518             tz               = _mm_mul_pd(fscal,dz12);
 519
 520             /* Update vectorial force */
 521             fix1             = _mm_add_pd(fix1,tx);
 522             fiy1             = _mm_add_pd(fiy1,ty);
 523             fiz1             = _mm_add_pd(fiz1,tz);
 524
 525             fjx2             = _mm_add_pd(fjx2,tx);
 526             fjy2             = _mm_add_pd(fjy2,ty);
 527             fjz2             = _mm_add_pd(fjz2,tz);
 528
 529             /**************************
 530              * CALCULATE INTERACTIONS *
 531              **************************/
 532
 533             r20              = _mm_mul_pd(rsq20,rinv20);
 534
 535             /* Calculate table index by multiplying r with table scale and truncate to integer */
 536             rt               = _mm_mul_pd(r20,vftabscale);
 537             vfitab           = _mm_cvttpd_epi32(rt);
 538             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 539             vfitab           = _mm_slli_epi32(vfitab,2);
 540
 541             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 542             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 543             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 544             GMX_MM_TRANSPOSE2_PD(Y,F);
 545             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 546             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 547             GMX_MM_TRANSPOSE2_PD(G,H);
 548             Heps             = _mm_mul_pd(vfeps,H);
 549             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 550             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 551             velec            = _mm_mul_pd(qq20,VV);
 552             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 553             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq20,FF),_mm_mul_pd(vftabscale,rinv20)));
 554
 555             /* Update potential sum for this i atom from the interaction with this j atom. */
 556             velecsum         = _mm_add_pd(velecsum,velec);
 557
 558             fscal            = felec;
 559
 560             /* Calculate temporary vectorial force */
 561             tx               = _mm_mul_pd(fscal,dx20);
 562             ty               = _mm_mul_pd(fscal,dy20);
 563             tz               = _mm_mul_pd(fscal,dz20);
 564
 565             /* Update vectorial force */
 566             fix2             = _mm_add_pd(fix2,tx);
 567             fiy2             = _mm_add_pd(fiy2,ty);
 568             fiz2             = _mm_add_pd(fiz2,tz);
 569
 570             fjx0             = _mm_add_pd(fjx0,tx);
 571             fjy0             = _mm_add_pd(fjy0,ty);
 572             fjz0             = _mm_add_pd(fjz0,tz);
 573
 574             /**************************
 575              * CALCULATE INTERACTIONS *
 576              **************************/
 577
 578             r21              = _mm_mul_pd(rsq21,rinv21);
 579
 580             /* Calculate table index by multiplying r with table scale and truncate to integer */
 581             rt               = _mm_mul_pd(r21,vftabscale);
 582             vfitab           = _mm_cvttpd_epi32(rt);
 583             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 584             vfitab           = _mm_slli_epi32(vfitab,2);
 585
 586             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 587             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 588             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 589             GMX_MM_TRANSPOSE2_PD(Y,F);
 590             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 591             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 592             GMX_MM_TRANSPOSE2_PD(G,H);
 593             Heps             = _mm_mul_pd(vfeps,H);
 594             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 595             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 596             velec            = _mm_mul_pd(qq21,VV);
 597             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 598             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
 599
 600             /* Update potential sum for this i atom from the interaction with this j atom. */
 601             velecsum         = _mm_add_pd(velecsum,velec);
 602
 603             fscal            = felec;
 604
 605             /* Calculate temporary vectorial force */
 606             tx               = _mm_mul_pd(fscal,dx21);
 607             ty               = _mm_mul_pd(fscal,dy21);
 608             tz               = _mm_mul_pd(fscal,dz21);
 609
 610             /* Update vectorial force */
 611             fix2             = _mm_add_pd(fix2,tx);
 612             fiy2             = _mm_add_pd(fiy2,ty);
 613             fiz2             = _mm_add_pd(fiz2,tz);
 614
 615             fjx1             = _mm_add_pd(fjx1,tx);
 616             fjy1             = _mm_add_pd(fjy1,ty);
 617             fjz1             = _mm_add_pd(fjz1,tz);
 618
 619             /**************************
 620              * CALCULATE INTERACTIONS *
 621              **************************/
 622
 623             r22              = _mm_mul_pd(rsq22,rinv22);
 624
 625             /* Calculate table index by multiplying r with table scale and truncate to integer */
 626             rt               = _mm_mul_pd(r22,vftabscale);
 627             vfitab           = _mm_cvttpd_epi32(rt);
 628             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 629             vfitab           = _mm_slli_epi32(vfitab,2);
 630
 631             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 632             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 633             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 634             GMX_MM_TRANSPOSE2_PD(Y,F);
 635             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 636             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 637             GMX_MM_TRANSPOSE2_PD(G,H);
 638             Heps             = _mm_mul_pd(vfeps,H);
 639             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 640             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 641             velec            = _mm_mul_pd(qq22,VV);
 642             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 643             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
 644
 645             /* Update potential sum for this i atom from the interaction with this j atom. */
 646             velecsum         = _mm_add_pd(velecsum,velec);
 647
 648             fscal            = felec;
 649
 650             /* Calculate temporary vectorial force */
 651             tx               = _mm_mul_pd(fscal,dx22);
 652             ty               = _mm_mul_pd(fscal,dy22);
 653             tz               = _mm_mul_pd(fscal,dz22);
 654
 655             /* Update vectorial force */
 656             fix2             = _mm_add_pd(fix2,tx);
 657             fiy2             = _mm_add_pd(fiy2,ty);
 658             fiz2             = _mm_add_pd(fiz2,tz);
 659
 660             fjx2             = _mm_add_pd(fjx2,tx);
 661             fjy2             = _mm_add_pd(fjy2,ty);
 662             fjz2             = _mm_add_pd(fjz2,tz);
 663
 664             gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 665
 666             /* Inner loop uses 387 flops */
 667         }
 668
 669         if(jidx<j_index_end)
 670         {
 671
 672             jnrA             = jjnr[jidx];
 673             j_coord_offsetA  = DIM*jnrA;
 674
 675             /* load j atom coordinates */
 676             gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 677                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 678
 679             /* Calculate displacement vector */
 680             dx00             = _mm_sub_pd(ix0,jx0);
 681             dy00             = _mm_sub_pd(iy0,jy0);
 682             dz00             = _mm_sub_pd(iz0,jz0);
 683             dx01             = _mm_sub_pd(ix0,jx1);
 684             dy01             = _mm_sub_pd(iy0,jy1);
 685             dz01             = _mm_sub_pd(iz0,jz1);
 686             dx02             = _mm_sub_pd(ix0,jx2);
 687             dy02             = _mm_sub_pd(iy0,jy2);
 688             dz02             = _mm_sub_pd(iz0,jz2);
 689             dx10             = _mm_sub_pd(ix1,jx0);
 690             dy10             = _mm_sub_pd(iy1,jy0);
 691             dz10             = _mm_sub_pd(iz1,jz0);
 692             dx11             = _mm_sub_pd(ix1,jx1);
 693             dy11             = _mm_sub_pd(iy1,jy1);
 694             dz11             = _mm_sub_pd(iz1,jz1);
 695             dx12             = _mm_sub_pd(ix1,jx2);
 696             dy12             = _mm_sub_pd(iy1,jy2);
 697             dz12             = _mm_sub_pd(iz1,jz2);
 698             dx20             = _mm_sub_pd(ix2,jx0);
 699             dy20             = _mm_sub_pd(iy2,jy0);
 700             dz20             = _mm_sub_pd(iz2,jz0);
 701             dx21             = _mm_sub_pd(ix2,jx1);
 702             dy21             = _mm_sub_pd(iy2,jy1);
 703             dz21             = _mm_sub_pd(iz2,jz1);
 704             dx22             = _mm_sub_pd(ix2,jx2);
 705             dy22             = _mm_sub_pd(iy2,jy2);
 706             dz22             = _mm_sub_pd(iz2,jz2);
 707
 708             /* Calculate squared distance and things based on it */
 709             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 710             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
 711             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
 712             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 713             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 714             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 715             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 716             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 717             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 718
 719             rinv00           = sse41_invsqrt_d(rsq00);
 720             rinv01           = sse41_invsqrt_d(rsq01);
 721             rinv02           = sse41_invsqrt_d(rsq02);
 722             rinv10           = sse41_invsqrt_d(rsq10);
 723             rinv11           = sse41_invsqrt_d(rsq11);
 724             rinv12           = sse41_invsqrt_d(rsq12);
 725             rinv20           = sse41_invsqrt_d(rsq20);
 726             rinv21           = sse41_invsqrt_d(rsq21);
 727             rinv22           = sse41_invsqrt_d(rsq22);
 728
 729             fjx0             = _mm_setzero_pd();
 730             fjy0             = _mm_setzero_pd();
 731             fjz0             = _mm_setzero_pd();
 732             fjx1             = _mm_setzero_pd();
 733             fjy1             = _mm_setzero_pd();
 734             fjz1             = _mm_setzero_pd();
 735             fjx2             = _mm_setzero_pd();
 736             fjy2             = _mm_setzero_pd();
 737             fjz2             = _mm_setzero_pd();
 738
 739             /**************************
 740              * CALCULATE INTERACTIONS *
 741              **************************/
 742
 743             r00              = _mm_mul_pd(rsq00,rinv00);
 744
 745             /* Calculate table index by multiplying r with table scale and truncate to integer */
 746             rt               = _mm_mul_pd(r00,vftabscale);
 747             vfitab           = _mm_cvttpd_epi32(rt);
 748             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 749             vfitab           = _mm_slli_epi32(vfitab,2);
 750
 751             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 752             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 753             F                = _mm_setzero_pd();
 754             GMX_MM_TRANSPOSE2_PD(Y,F);
 755             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 756             H                = _mm_setzero_pd();
 757             GMX_MM_TRANSPOSE2_PD(G,H);
 758             Heps             = _mm_mul_pd(vfeps,H);
 759             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 760             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 761             velec            = _mm_mul_pd(qq00,VV);
 762             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 763             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq00,FF),_mm_mul_pd(vftabscale,rinv00)));
 764
 765             /* Update potential sum for this i atom from the interaction with this j atom. */
 766             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 767             velecsum         = _mm_add_pd(velecsum,velec);
 768
 769             fscal            = felec;
 770
 771             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 772
 773             /* Calculate temporary vectorial force */
 774             tx               = _mm_mul_pd(fscal,dx00);
 775             ty               = _mm_mul_pd(fscal,dy00);
 776             tz               = _mm_mul_pd(fscal,dz00);
 777
 778             /* Update vectorial force */
 779             fix0             = _mm_add_pd(fix0,tx);
 780             fiy0             = _mm_add_pd(fiy0,ty);
 781             fiz0             = _mm_add_pd(fiz0,tz);
 782
 783             fjx0             = _mm_add_pd(fjx0,tx);
 784             fjy0             = _mm_add_pd(fjy0,ty);
 785             fjz0             = _mm_add_pd(fjz0,tz);
 786
 787             /**************************
 788              * CALCULATE INTERACTIONS *
 789              **************************/
 790
 791             r01              = _mm_mul_pd(rsq01,rinv01);
 792
 793             /* Calculate table index by multiplying r with table scale and truncate to integer */
 794             rt               = _mm_mul_pd(r01,vftabscale);
 795             vfitab           = _mm_cvttpd_epi32(rt);
 796             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 797             vfitab           = _mm_slli_epi32(vfitab,2);
 798
 799             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 800             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 801             F                = _mm_setzero_pd();
 802             GMX_MM_TRANSPOSE2_PD(Y,F);
 803             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 804             H                = _mm_setzero_pd();
 805             GMX_MM_TRANSPOSE2_PD(G,H);
 806             Heps             = _mm_mul_pd(vfeps,H);
 807             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 808             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 809             velec            = _mm_mul_pd(qq01,VV);
 810             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 811             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq01,FF),_mm_mul_pd(vftabscale,rinv01)));
 812
 813             /* Update potential sum for this i atom from the interaction with this j atom. */
 814             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 815             velecsum         = _mm_add_pd(velecsum,velec);
 816
 817             fscal            = felec;
 818
 819             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 820
 821             /* Calculate temporary vectorial force */
 822             tx               = _mm_mul_pd(fscal,dx01);
 823             ty               = _mm_mul_pd(fscal,dy01);
 824             tz               = _mm_mul_pd(fscal,dz01);
 825
 826             /* Update vectorial force */
 827             fix0             = _mm_add_pd(fix0,tx);
 828             fiy0             = _mm_add_pd(fiy0,ty);
 829             fiz0             = _mm_add_pd(fiz0,tz);
 830
 831             fjx1             = _mm_add_pd(fjx1,tx);
 832             fjy1             = _mm_add_pd(fjy1,ty);
 833             fjz1             = _mm_add_pd(fjz1,tz);
 834
 835             /**************************
 836              * CALCULATE INTERACTIONS *
 837              **************************/
 838
 839             r02              = _mm_mul_pd(rsq02,rinv02);
 840
 841             /* Calculate table index by multiplying r with table scale and truncate to integer */
 842             rt               = _mm_mul_pd(r02,vftabscale);
 843             vfitab           = _mm_cvttpd_epi32(rt);
 844             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 845             vfitab           = _mm_slli_epi32(vfitab,2);
 846
 847             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 848             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 849             F                = _mm_setzero_pd();
 850             GMX_MM_TRANSPOSE2_PD(Y,F);
 851             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 852             H                = _mm_setzero_pd();
 853             GMX_MM_TRANSPOSE2_PD(G,H);
 854             Heps             = _mm_mul_pd(vfeps,H);
 855             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 856             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 857             velec            = _mm_mul_pd(qq02,VV);
 858             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 859             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq02,FF),_mm_mul_pd(vftabscale,rinv02)));
 860
 861             /* Update potential sum for this i atom from the interaction with this j atom. */
 862             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 863             velecsum         = _mm_add_pd(velecsum,velec);
 864
 865             fscal            = felec;
 866
 867             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 868
 869             /* Calculate temporary vectorial force */
 870             tx               = _mm_mul_pd(fscal,dx02);
 871             ty               = _mm_mul_pd(fscal,dy02);
 872             tz               = _mm_mul_pd(fscal,dz02);
 873
 874             /* Update vectorial force */
 875             fix0             = _mm_add_pd(fix0,tx);
 876             fiy0             = _mm_add_pd(fiy0,ty);
 877             fiz0             = _mm_add_pd(fiz0,tz);
 878
 879             fjx2             = _mm_add_pd(fjx2,tx);
 880             fjy2             = _mm_add_pd(fjy2,ty);
 881             fjz2             = _mm_add_pd(fjz2,tz);
 882
 883             /**************************
 884              * CALCULATE INTERACTIONS *
 885              **************************/
 886
 887             r10              = _mm_mul_pd(rsq10,rinv10);
 888
 889             /* Calculate table index by multiplying r with table scale and truncate to integer */
 890             rt               = _mm_mul_pd(r10,vftabscale);
 891             vfitab           = _mm_cvttpd_epi32(rt);
 892             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 893             vfitab           = _mm_slli_epi32(vfitab,2);
 894
 895             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 896             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 897             F                = _mm_setzero_pd();
 898             GMX_MM_TRANSPOSE2_PD(Y,F);
 899             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 900             H                = _mm_setzero_pd();
 901             GMX_MM_TRANSPOSE2_PD(G,H);
 902             Heps             = _mm_mul_pd(vfeps,H);
 903             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 904             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 905             velec            = _mm_mul_pd(qq10,VV);
 906             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 907             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq10,FF),_mm_mul_pd(vftabscale,rinv10)));
 908
 909             /* Update potential sum for this i atom from the interaction with this j atom. */
 910             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 911             velecsum         = _mm_add_pd(velecsum,velec);
 912
 913             fscal            = felec;
 914
 915             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 916
 917             /* Calculate temporary vectorial force */
 918             tx               = _mm_mul_pd(fscal,dx10);
 919             ty               = _mm_mul_pd(fscal,dy10);
 920             tz               = _mm_mul_pd(fscal,dz10);
 921
 922             /* Update vectorial force */
 923             fix1             = _mm_add_pd(fix1,tx);
 924             fiy1             = _mm_add_pd(fiy1,ty);
 925             fiz1             = _mm_add_pd(fiz1,tz);
 926
 927             fjx0             = _mm_add_pd(fjx0,tx);
 928             fjy0             = _mm_add_pd(fjy0,ty);
 929             fjz0             = _mm_add_pd(fjz0,tz);
 930
 931             /**************************
 932              * CALCULATE INTERACTIONS *
 933              **************************/
 934
 935             r11              = _mm_mul_pd(rsq11,rinv11);
 936
 937             /* Calculate table index by multiplying r with table scale and truncate to integer */
 938             rt               = _mm_mul_pd(r11,vftabscale);
 939             vfitab           = _mm_cvttpd_epi32(rt);
 940             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 941             vfitab           = _mm_slli_epi32(vfitab,2);
 942
 943             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 944             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 945             F                = _mm_setzero_pd();
 946             GMX_MM_TRANSPOSE2_PD(Y,F);
 947             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 948             H                = _mm_setzero_pd();
 949             GMX_MM_TRANSPOSE2_PD(G,H);
 950             Heps             = _mm_mul_pd(vfeps,H);
 951             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 952             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 953             velec            = _mm_mul_pd(qq11,VV);
 954             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 955             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
 956
 957             /* Update potential sum for this i atom from the interaction with this j atom. */
 958             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 959             velecsum         = _mm_add_pd(velecsum,velec);
 960
 961             fscal            = felec;
 962
 963             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 964
 965             /* Calculate temporary vectorial force */
 966             tx               = _mm_mul_pd(fscal,dx11);
 967             ty               = _mm_mul_pd(fscal,dy11);
 968             tz               = _mm_mul_pd(fscal,dz11);
 969
 970             /* Update vectorial force */
 971             fix1             = _mm_add_pd(fix1,tx);
 972             fiy1             = _mm_add_pd(fiy1,ty);
 973             fiz1             = _mm_add_pd(fiz1,tz);
 974
 975             fjx1             = _mm_add_pd(fjx1,tx);
 976             fjy1             = _mm_add_pd(fjy1,ty);
 977             fjz1             = _mm_add_pd(fjz1,tz);
 978
 979             /**************************
 980              * CALCULATE INTERACTIONS *
 981              **************************/
 982
 983             r12              = _mm_mul_pd(rsq12,rinv12);
 984
 985             /* Calculate table index by multiplying r with table scale and truncate to integer */
 986             rt               = _mm_mul_pd(r12,vftabscale);
 987             vfitab           = _mm_cvttpd_epi32(rt);
 988             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 989             vfitab           = _mm_slli_epi32(vfitab,2);
 990
 991             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 992             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 993             F                = _mm_setzero_pd();
 994             GMX_MM_TRANSPOSE2_PD(Y,F);
 995             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 996             H                = _mm_setzero_pd();
 997             GMX_MM_TRANSPOSE2_PD(G,H);
 998             Heps             = _mm_mul_pd(vfeps,H);
 999             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1000             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1001             velec            = _mm_mul_pd(qq12,VV);
1002             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1003             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
1004
1005             /* Update potential sum for this i atom from the interaction with this j atom. */
1006             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1007             velecsum         = _mm_add_pd(velecsum,velec);
1008
1009             fscal            = felec;
1010
1011             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1012
1013             /* Calculate temporary vectorial force */
1014             tx               = _mm_mul_pd(fscal,dx12);
1015             ty               = _mm_mul_pd(fscal,dy12);
1016             tz               = _mm_mul_pd(fscal,dz12);
1017
1018             /* Update vectorial force */
1019             fix1             = _mm_add_pd(fix1,tx);
1020             fiy1             = _mm_add_pd(fiy1,ty);
1021             fiz1             = _mm_add_pd(fiz1,tz);
1022
1023             fjx2             = _mm_add_pd(fjx2,tx);
1024             fjy2             = _mm_add_pd(fjy2,ty);
1025             fjz2             = _mm_add_pd(fjz2,tz);
1026
1027             /**************************
1028              * CALCULATE INTERACTIONS *
1029              **************************/
1030
1031             r20              = _mm_mul_pd(rsq20,rinv20);
1032
1033             /* Calculate table index by multiplying r with table scale and truncate to integer */
1034             rt               = _mm_mul_pd(r20,vftabscale);
1035             vfitab           = _mm_cvttpd_epi32(rt);
1036             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1037             vfitab           = _mm_slli_epi32(vfitab,2);
1038
1039             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1040             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1041             F                = _mm_setzero_pd();
1042             GMX_MM_TRANSPOSE2_PD(Y,F);
1043             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1044             H                = _mm_setzero_pd();
1045             GMX_MM_TRANSPOSE2_PD(G,H);
1046             Heps             = _mm_mul_pd(vfeps,H);
1047             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1048             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1049             velec            = _mm_mul_pd(qq20,VV);
1050             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1051             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq20,FF),_mm_mul_pd(vftabscale,rinv20)));
1052
1053             /* Update potential sum for this i atom from the interaction with this j atom. */
1054             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1055             velecsum         = _mm_add_pd(velecsum,velec);
1056
1057             fscal            = felec;
1058
1059             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1060
1061             /* Calculate temporary vectorial force */
1062             tx               = _mm_mul_pd(fscal,dx20);
1063             ty               = _mm_mul_pd(fscal,dy20);
1064             tz               = _mm_mul_pd(fscal,dz20);
1065
1066             /* Update vectorial force */
1067             fix2             = _mm_add_pd(fix2,tx);
1068             fiy2             = _mm_add_pd(fiy2,ty);
1069             fiz2             = _mm_add_pd(fiz2,tz);
1070
1071             fjx0             = _mm_add_pd(fjx0,tx);
1072             fjy0             = _mm_add_pd(fjy0,ty);
1073             fjz0             = _mm_add_pd(fjz0,tz);
1074
1075             /**************************
1076              * CALCULATE INTERACTIONS *
1077              **************************/
1078
1079             r21              = _mm_mul_pd(rsq21,rinv21);
1080
1081             /* Calculate table index by multiplying r with table scale and truncate to integer */
1082             rt               = _mm_mul_pd(r21,vftabscale);
1083             vfitab           = _mm_cvttpd_epi32(rt);
1084             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1085             vfitab           = _mm_slli_epi32(vfitab,2);
1086
1087             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1088             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1089             F                = _mm_setzero_pd();
1090             GMX_MM_TRANSPOSE2_PD(Y,F);
1091             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1092             H                = _mm_setzero_pd();
1093             GMX_MM_TRANSPOSE2_PD(G,H);
1094             Heps             = _mm_mul_pd(vfeps,H);
1095             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1096             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1097             velec            = _mm_mul_pd(qq21,VV);
1098             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1099             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
1100
1101             /* Update potential sum for this i atom from the interaction with this j atom. */
1102             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1103             velecsum         = _mm_add_pd(velecsum,velec);
1104
1105             fscal            = felec;
1106
1107             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1108
1109             /* Calculate temporary vectorial force */
1110             tx               = _mm_mul_pd(fscal,dx21);
1111             ty               = _mm_mul_pd(fscal,dy21);
1112             tz               = _mm_mul_pd(fscal,dz21);
1113
1114             /* Update vectorial force */
1115             fix2             = _mm_add_pd(fix2,tx);
1116             fiy2             = _mm_add_pd(fiy2,ty);
1117             fiz2             = _mm_add_pd(fiz2,tz);
1118
1119             fjx1             = _mm_add_pd(fjx1,tx);
1120             fjy1             = _mm_add_pd(fjy1,ty);
1121             fjz1             = _mm_add_pd(fjz1,tz);
1122
1123             /**************************
1124              * CALCULATE INTERACTIONS *
1125              **************************/
1126
1127             r22              = _mm_mul_pd(rsq22,rinv22);
1128
1129             /* Calculate table index by multiplying r with table scale and truncate to integer */
1130             rt               = _mm_mul_pd(r22,vftabscale);
1131             vfitab           = _mm_cvttpd_epi32(rt);
1132             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1133             vfitab           = _mm_slli_epi32(vfitab,2);
1134
1135             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1136             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1137             F                = _mm_setzero_pd();
1138             GMX_MM_TRANSPOSE2_PD(Y,F);
1139             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1140             H                = _mm_setzero_pd();
1141             GMX_MM_TRANSPOSE2_PD(G,H);
1142             Heps             = _mm_mul_pd(vfeps,H);
1143             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1144             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1145             velec            = _mm_mul_pd(qq22,VV);
1146             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1147             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
1148
1149             /* Update potential sum for this i atom from the interaction with this j atom. */
1150             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1151             velecsum         = _mm_add_pd(velecsum,velec);
1152
1153             fscal            = felec;
1154
1155             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1156
1157             /* Calculate temporary vectorial force */
1158             tx               = _mm_mul_pd(fscal,dx22);
1159             ty               = _mm_mul_pd(fscal,dy22);
1160             tz               = _mm_mul_pd(fscal,dz22);
1161
1162             /* Update vectorial force */
1163             fix2             = _mm_add_pd(fix2,tx);
1164             fiy2             = _mm_add_pd(fiy2,ty);
1165             fiz2             = _mm_add_pd(fiz2,tz);
1166
1167             fjx2             = _mm_add_pd(fjx2,tx);
1168             fjy2             = _mm_add_pd(fjy2,ty);
1169             fjz2             = _mm_add_pd(fjz2,tz);
1170
1171             gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1172
1173             /* Inner loop uses 387 flops */
1174         }
1175
1176         /* End of innermost loop */
1177
1178         gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1179                                               f+i_coord_offset,fshift+i_shift_offset);
1180
1181         ggid                        = gid[iidx];
1182         /* Update potential energies */
1183         gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1184
1185         /* Increment number of inner iterations */
1186         inneriter                  += j_index_end - j_index_start;
1187
1188         /* Outer loop uses 19 flops */
1189     }
1190
1191     /* Increment number of outer iterations */
1192     outeriter        += nri;
1193
1194     /* Update outer/inner flops */
1195
1196     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3W3_VF,outeriter*19 + inneriter*387);
1197 }
1198 /*
1199  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwNone_GeomW3W3_F_sse4_1_double
1200  * Electrostatics interaction: CubicSplineTable
1201  * VdW interaction:            None
1202  * Geometry:                   Water3-Water3
1203  * Calculate force/pot:        Force
1204  */
1205 void
1206 nb_kernel_ElecCSTab_VdwNone_GeomW3W3_F_sse4_1_double
1207                     (t_nblist                    * gmx_restrict       nlist,
1208                      rvec                        * gmx_restrict          xx,
1209                      rvec                        * gmx_restrict          ff,
1210                      struct t_forcerec           * gmx_restrict          fr,
1211                      t_mdatoms                   * gmx_restrict     mdatoms,
1212                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1213                      t_nrnb                      * gmx_restrict        nrnb)
1214 {
1215     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1216      * just 0 for non-waters.
1217      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
1218      * jnr indices corresponding to data put in the four positions in the SIMD register.
1219      */
1220     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1221     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1222     int              jnrA,jnrB;
1223     int              j_coord_offsetA,j_coord_offsetB;
1224     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1225     real             rcutoff_scalar;
1226     real             *shiftvec,*fshift,*x,*f;
1227     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1228     int              vdwioffset0;
1229     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1230     int              vdwioffset1;
1231     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1232     int              vdwioffset2;
1233     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1234     int              vdwjidx0A,vdwjidx0B;
1235     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1236     int              vdwjidx1A,vdwjidx1B;
1237     __m128d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1238     int              vdwjidx2A,vdwjidx2B;
1239     __m128d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1240     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1241     __m128d          dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1242     __m128d          dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1243     __m128d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1244     __m128d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1245     __m128d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1246     __m128d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1247     __m128d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1248     __m128d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1249     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
1250     real             *charge;
1251     __m128i          vfitab;
1252     __m128i          ifour       = _mm_set1_epi32(4);
1253     __m128d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
1254     real             *vftab;
1255     __m128d          dummy_mask,cutoff_mask;
1256     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
1257     __m128d          one     = _mm_set1_pd(1.0);
1258     __m128d          two     = _mm_set1_pd(2.0);
1259     x                = xx[0];
1260     f                = ff[0];
1261
1262     nri              = nlist->nri;
1263     iinr             = nlist->iinr;
1264     jindex           = nlist->jindex;
1265     jjnr             = nlist->jjnr;
1266     shiftidx         = nlist->shift;
1267     gid              = nlist->gid;
1268     shiftvec         = fr->shift_vec[0];
1269     fshift           = fr->fshift[0];
1270     facel            = _mm_set1_pd(fr->ic->epsfac);
1271     charge           = mdatoms->chargeA;
1272
1273     vftab            = kernel_data->table_elec->data;
1274     vftabscale       = _mm_set1_pd(kernel_data->table_elec->scale);
1275
1276     /* Setup water-specific parameters */
1277     inr              = nlist->iinr[0];
1278     iq0              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
1279     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
1280     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
1281
1282     jq0              = _mm_set1_pd(charge[inr+0]);
1283     jq1              = _mm_set1_pd(charge[inr+1]);
1284     jq2              = _mm_set1_pd(charge[inr+2]);
1285     qq00             = _mm_mul_pd(iq0,jq0);
1286     qq01             = _mm_mul_pd(iq0,jq1);
1287     qq02             = _mm_mul_pd(iq0,jq2);
1288     qq10             = _mm_mul_pd(iq1,jq0);
1289     qq11             = _mm_mul_pd(iq1,jq1);
1290     qq12             = _mm_mul_pd(iq1,jq2);
1291     qq20             = _mm_mul_pd(iq2,jq0);
1292     qq21             = _mm_mul_pd(iq2,jq1);
1293     qq22             = _mm_mul_pd(iq2,jq2);
1294
1295     /* Avoid stupid compiler warnings */
1296     jnrA = jnrB = 0;
1297     j_coord_offsetA = 0;
1298     j_coord_offsetB = 0;
1299
1300     outeriter        = 0;
1301     inneriter        = 0;
1302
1303     /* Start outer loop over neighborlists */
1304     for(iidx=0; iidx<nri; iidx++)
1305     {
1306         /* Load shift vector for this list */
1307         i_shift_offset   = DIM*shiftidx[iidx];
1308
1309         /* Load limits for loop over neighbors */
1310         j_index_start    = jindex[iidx];
1311         j_index_end      = jindex[iidx+1];
1312
1313         /* Get outer coordinate index */
1314         inr              = iinr[iidx];
1315         i_coord_offset   = DIM*inr;
1316
1317         /* Load i particle coords and add shift vector */
1318         gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1319                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1320
1321         fix0             = _mm_setzero_pd();
1322         fiy0             = _mm_setzero_pd();
1323         fiz0             = _mm_setzero_pd();
1324         fix1             = _mm_setzero_pd();
1325         fiy1             = _mm_setzero_pd();
1326         fiz1             = _mm_setzero_pd();
1327         fix2             = _mm_setzero_pd();
1328         fiy2             = _mm_setzero_pd();
1329         fiz2             = _mm_setzero_pd();
1330
1331         /* Start inner kernel loop */
1332         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
1333         {
1334
1335             /* Get j neighbor index, and coordinate index */
1336             jnrA             = jjnr[jidx];
1337             jnrB             = jjnr[jidx+1];
1338             j_coord_offsetA  = DIM*jnrA;
1339             j_coord_offsetB  = DIM*jnrB;
1340
1341             /* load j atom coordinates */
1342             gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1343                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1344
1345             /* Calculate displacement vector */
1346             dx00             = _mm_sub_pd(ix0,jx0);
1347             dy00             = _mm_sub_pd(iy0,jy0);
1348             dz00             = _mm_sub_pd(iz0,jz0);
1349             dx01             = _mm_sub_pd(ix0,jx1);
1350             dy01             = _mm_sub_pd(iy0,jy1);
1351             dz01             = _mm_sub_pd(iz0,jz1);
1352             dx02             = _mm_sub_pd(ix0,jx2);
1353             dy02             = _mm_sub_pd(iy0,jy2);
1354             dz02             = _mm_sub_pd(iz0,jz2);
1355             dx10             = _mm_sub_pd(ix1,jx0);
1356             dy10             = _mm_sub_pd(iy1,jy0);
1357             dz10             = _mm_sub_pd(iz1,jz0);
1358             dx11             = _mm_sub_pd(ix1,jx1);
1359             dy11             = _mm_sub_pd(iy1,jy1);
1360             dz11             = _mm_sub_pd(iz1,jz1);
1361             dx12             = _mm_sub_pd(ix1,jx2);
1362             dy12             = _mm_sub_pd(iy1,jy2);
1363             dz12             = _mm_sub_pd(iz1,jz2);
1364             dx20             = _mm_sub_pd(ix2,jx0);
1365             dy20             = _mm_sub_pd(iy2,jy0);
1366             dz20             = _mm_sub_pd(iz2,jz0);
1367             dx21             = _mm_sub_pd(ix2,jx1);
1368             dy21             = _mm_sub_pd(iy2,jy1);
1369             dz21             = _mm_sub_pd(iz2,jz1);
1370             dx22             = _mm_sub_pd(ix2,jx2);
1371             dy22             = _mm_sub_pd(iy2,jy2);
1372             dz22             = _mm_sub_pd(iz2,jz2);
1373
1374             /* Calculate squared distance and things based on it */
1375             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1376             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
1377             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
1378             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
1379             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1380             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1381             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
1382             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1383             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1384
1385             rinv00           = sse41_invsqrt_d(rsq00);
1386             rinv01           = sse41_invsqrt_d(rsq01);
1387             rinv02           = sse41_invsqrt_d(rsq02);
1388             rinv10           = sse41_invsqrt_d(rsq10);
1389             rinv11           = sse41_invsqrt_d(rsq11);
1390             rinv12           = sse41_invsqrt_d(rsq12);
1391             rinv20           = sse41_invsqrt_d(rsq20);
1392             rinv21           = sse41_invsqrt_d(rsq21);
1393             rinv22           = sse41_invsqrt_d(rsq22);
1394
1395             fjx0             = _mm_setzero_pd();
1396             fjy0             = _mm_setzero_pd();
1397             fjz0             = _mm_setzero_pd();
1398             fjx1             = _mm_setzero_pd();
1399             fjy1             = _mm_setzero_pd();
1400             fjz1             = _mm_setzero_pd();
1401             fjx2             = _mm_setzero_pd();
1402             fjy2             = _mm_setzero_pd();
1403             fjz2             = _mm_setzero_pd();
1404
1405             /**************************
1406              * CALCULATE INTERACTIONS *
1407              **************************/
1408
1409             r00              = _mm_mul_pd(rsq00,rinv00);
1410
1411             /* Calculate table index by multiplying r with table scale and truncate to integer */
1412             rt               = _mm_mul_pd(r00,vftabscale);
1413             vfitab           = _mm_cvttpd_epi32(rt);
1414             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1415             vfitab           = _mm_slli_epi32(vfitab,2);
1416
1417             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1418             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1419             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1420             GMX_MM_TRANSPOSE2_PD(Y,F);
1421             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1422             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1423             GMX_MM_TRANSPOSE2_PD(G,H);
1424             Heps             = _mm_mul_pd(vfeps,H);
1425             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1426             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1427             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq00,FF),_mm_mul_pd(vftabscale,rinv00)));
1428
1429             fscal            = felec;
1430
1431             /* Calculate temporary vectorial force */
1432             tx               = _mm_mul_pd(fscal,dx00);
1433             ty               = _mm_mul_pd(fscal,dy00);
1434             tz               = _mm_mul_pd(fscal,dz00);
1435
1436             /* Update vectorial force */
1437             fix0             = _mm_add_pd(fix0,tx);
1438             fiy0             = _mm_add_pd(fiy0,ty);
1439             fiz0             = _mm_add_pd(fiz0,tz);
1440
1441             fjx0             = _mm_add_pd(fjx0,tx);
1442             fjy0             = _mm_add_pd(fjy0,ty);
1443             fjz0             = _mm_add_pd(fjz0,tz);
1444
1445             /**************************
1446              * CALCULATE INTERACTIONS *
1447              **************************/
1448
1449             r01              = _mm_mul_pd(rsq01,rinv01);
1450
1451             /* Calculate table index by multiplying r with table scale and truncate to integer */
1452             rt               = _mm_mul_pd(r01,vftabscale);
1453             vfitab           = _mm_cvttpd_epi32(rt);
1454             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1455             vfitab           = _mm_slli_epi32(vfitab,2);
1456
1457             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1458             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1459             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1460             GMX_MM_TRANSPOSE2_PD(Y,F);
1461             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1462             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1463             GMX_MM_TRANSPOSE2_PD(G,H);
1464             Heps             = _mm_mul_pd(vfeps,H);
1465             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1466             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1467             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq01,FF),_mm_mul_pd(vftabscale,rinv01)));
1468
1469             fscal            = felec;
1470
1471             /* Calculate temporary vectorial force */
1472             tx               = _mm_mul_pd(fscal,dx01);
1473             ty               = _mm_mul_pd(fscal,dy01);
1474             tz               = _mm_mul_pd(fscal,dz01);
1475
1476             /* Update vectorial force */
1477             fix0             = _mm_add_pd(fix0,tx);
1478             fiy0             = _mm_add_pd(fiy0,ty);
1479             fiz0             = _mm_add_pd(fiz0,tz);
1480
1481             fjx1             = _mm_add_pd(fjx1,tx);
1482             fjy1             = _mm_add_pd(fjy1,ty);
1483             fjz1             = _mm_add_pd(fjz1,tz);
1484
1485             /**************************
1486              * CALCULATE INTERACTIONS *
1487              **************************/
1488
1489             r02              = _mm_mul_pd(rsq02,rinv02);
1490
1491             /* Calculate table index by multiplying r with table scale and truncate to integer */
1492             rt               = _mm_mul_pd(r02,vftabscale);
1493             vfitab           = _mm_cvttpd_epi32(rt);
1494             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1495             vfitab           = _mm_slli_epi32(vfitab,2);
1496
1497             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1498             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1499             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1500             GMX_MM_TRANSPOSE2_PD(Y,F);
1501             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1502             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1503             GMX_MM_TRANSPOSE2_PD(G,H);
1504             Heps             = _mm_mul_pd(vfeps,H);
1505             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1506             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1507             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq02,FF),_mm_mul_pd(vftabscale,rinv02)));
1508
1509             fscal            = felec;
1510
1511             /* Calculate temporary vectorial force */
1512             tx               = _mm_mul_pd(fscal,dx02);
1513             ty               = _mm_mul_pd(fscal,dy02);
1514             tz               = _mm_mul_pd(fscal,dz02);
1515
1516             /* Update vectorial force */
1517             fix0             = _mm_add_pd(fix0,tx);
1518             fiy0             = _mm_add_pd(fiy0,ty);
1519             fiz0             = _mm_add_pd(fiz0,tz);
1520
1521             fjx2             = _mm_add_pd(fjx2,tx);
1522             fjy2             = _mm_add_pd(fjy2,ty);
1523             fjz2             = _mm_add_pd(fjz2,tz);
1524
1525             /**************************
1526              * CALCULATE INTERACTIONS *
1527              **************************/
1528
1529             r10              = _mm_mul_pd(rsq10,rinv10);
1530
1531             /* Calculate table index by multiplying r with table scale and truncate to integer */
1532             rt               = _mm_mul_pd(r10,vftabscale);
1533             vfitab           = _mm_cvttpd_epi32(rt);
1534             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1535             vfitab           = _mm_slli_epi32(vfitab,2);
1536
1537             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1538             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1539             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1540             GMX_MM_TRANSPOSE2_PD(Y,F);
1541             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1542             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1543             GMX_MM_TRANSPOSE2_PD(G,H);
1544             Heps             = _mm_mul_pd(vfeps,H);
1545             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1546             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1547             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq10,FF),_mm_mul_pd(vftabscale,rinv10)));
1548
1549             fscal            = felec;
1550
1551             /* Calculate temporary vectorial force */
1552             tx               = _mm_mul_pd(fscal,dx10);
1553             ty               = _mm_mul_pd(fscal,dy10);
1554             tz               = _mm_mul_pd(fscal,dz10);
1555
1556             /* Update vectorial force */
1557             fix1             = _mm_add_pd(fix1,tx);
1558             fiy1             = _mm_add_pd(fiy1,ty);
1559             fiz1             = _mm_add_pd(fiz1,tz);
1560
1561             fjx0             = _mm_add_pd(fjx0,tx);
1562             fjy0             = _mm_add_pd(fjy0,ty);
1563             fjz0             = _mm_add_pd(fjz0,tz);
1564
1565             /**************************
1566              * CALCULATE INTERACTIONS *
1567              **************************/
1568
1569             r11              = _mm_mul_pd(rsq11,rinv11);
1570
1571             /* Calculate table index by multiplying r with table scale and truncate to integer */
1572             rt               = _mm_mul_pd(r11,vftabscale);
1573             vfitab           = _mm_cvttpd_epi32(rt);
1574             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1575             vfitab           = _mm_slli_epi32(vfitab,2);
1576
1577             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1578             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1579             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1580             GMX_MM_TRANSPOSE2_PD(Y,F);
1581             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1582             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1583             GMX_MM_TRANSPOSE2_PD(G,H);
1584             Heps             = _mm_mul_pd(vfeps,H);
1585             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1586             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1587             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
1588
1589             fscal            = felec;
1590
1591             /* Calculate temporary vectorial force */
1592             tx               = _mm_mul_pd(fscal,dx11);
1593             ty               = _mm_mul_pd(fscal,dy11);
1594             tz               = _mm_mul_pd(fscal,dz11);
1595
1596             /* Update vectorial force */
1597             fix1             = _mm_add_pd(fix1,tx);
1598             fiy1             = _mm_add_pd(fiy1,ty);
1599             fiz1             = _mm_add_pd(fiz1,tz);
1600
1601             fjx1             = _mm_add_pd(fjx1,tx);
1602             fjy1             = _mm_add_pd(fjy1,ty);
1603             fjz1             = _mm_add_pd(fjz1,tz);
1604
1605             /**************************
1606              * CALCULATE INTERACTIONS *
1607              **************************/
1608
1609             r12              = _mm_mul_pd(rsq12,rinv12);
1610
1611             /* Calculate table index by multiplying r with table scale and truncate to integer */
1612             rt               = _mm_mul_pd(r12,vftabscale);
1613             vfitab           = _mm_cvttpd_epi32(rt);
1614             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1615             vfitab           = _mm_slli_epi32(vfitab,2);
1616
1617             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1618             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1619             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1620             GMX_MM_TRANSPOSE2_PD(Y,F);
1621             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1622             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1623             GMX_MM_TRANSPOSE2_PD(G,H);
1624             Heps             = _mm_mul_pd(vfeps,H);
1625             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1626             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1627             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
1628
1629             fscal            = felec;
1630
1631             /* Calculate temporary vectorial force */
1632             tx               = _mm_mul_pd(fscal,dx12);
1633             ty               = _mm_mul_pd(fscal,dy12);
1634             tz               = _mm_mul_pd(fscal,dz12);
1635
1636             /* Update vectorial force */
1637             fix1             = _mm_add_pd(fix1,tx);
1638             fiy1             = _mm_add_pd(fiy1,ty);
1639             fiz1             = _mm_add_pd(fiz1,tz);
1640
1641             fjx2             = _mm_add_pd(fjx2,tx);
1642             fjy2             = _mm_add_pd(fjy2,ty);
1643             fjz2             = _mm_add_pd(fjz2,tz);
1644
1645             /**************************
1646              * CALCULATE INTERACTIONS *
1647              **************************/
1648
1649             r20              = _mm_mul_pd(rsq20,rinv20);
1650
1651             /* Calculate table index by multiplying r with table scale and truncate to integer */
1652             rt               = _mm_mul_pd(r20,vftabscale);
1653             vfitab           = _mm_cvttpd_epi32(rt);
1654             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1655             vfitab           = _mm_slli_epi32(vfitab,2);
1656
1657             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1658             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1659             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1660             GMX_MM_TRANSPOSE2_PD(Y,F);
1661             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1662             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1663             GMX_MM_TRANSPOSE2_PD(G,H);
1664             Heps             = _mm_mul_pd(vfeps,H);
1665             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1666             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1667             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq20,FF),_mm_mul_pd(vftabscale,rinv20)));
1668
1669             fscal            = felec;
1670
1671             /* Calculate temporary vectorial force */
1672             tx               = _mm_mul_pd(fscal,dx20);
1673             ty               = _mm_mul_pd(fscal,dy20);
1674             tz               = _mm_mul_pd(fscal,dz20);
1675
1676             /* Update vectorial force */
1677             fix2             = _mm_add_pd(fix2,tx);
1678             fiy2             = _mm_add_pd(fiy2,ty);
1679             fiz2             = _mm_add_pd(fiz2,tz);
1680
1681             fjx0             = _mm_add_pd(fjx0,tx);
1682             fjy0             = _mm_add_pd(fjy0,ty);
1683             fjz0             = _mm_add_pd(fjz0,tz);
1684
1685             /**************************
1686              * CALCULATE INTERACTIONS *
1687              **************************/
1688
1689             r21              = _mm_mul_pd(rsq21,rinv21);
1690
1691             /* Calculate table index by multiplying r with table scale and truncate to integer */
1692             rt               = _mm_mul_pd(r21,vftabscale);
1693             vfitab           = _mm_cvttpd_epi32(rt);
1694             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1695             vfitab           = _mm_slli_epi32(vfitab,2);
1696
1697             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1698             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1699             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1700             GMX_MM_TRANSPOSE2_PD(Y,F);
1701             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1702             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1703             GMX_MM_TRANSPOSE2_PD(G,H);
1704             Heps             = _mm_mul_pd(vfeps,H);
1705             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1706             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1707             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
1708
1709             fscal            = felec;
1710
1711             /* Calculate temporary vectorial force */
1712             tx               = _mm_mul_pd(fscal,dx21);
1713             ty               = _mm_mul_pd(fscal,dy21);
1714             tz               = _mm_mul_pd(fscal,dz21);
1715
1716             /* Update vectorial force */
1717             fix2             = _mm_add_pd(fix2,tx);
1718             fiy2             = _mm_add_pd(fiy2,ty);
1719             fiz2             = _mm_add_pd(fiz2,tz);
1720
1721             fjx1             = _mm_add_pd(fjx1,tx);
1722             fjy1             = _mm_add_pd(fjy1,ty);
1723             fjz1             = _mm_add_pd(fjz1,tz);
1724
1725             /**************************
1726              * CALCULATE INTERACTIONS *
1727              **************************/
1728
1729             r22              = _mm_mul_pd(rsq22,rinv22);
1730
1731             /* Calculate table index by multiplying r with table scale and truncate to integer */
1732             rt               = _mm_mul_pd(r22,vftabscale);
1733             vfitab           = _mm_cvttpd_epi32(rt);
1734             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1735             vfitab           = _mm_slli_epi32(vfitab,2);
1736
1737             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1738             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1739             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1740             GMX_MM_TRANSPOSE2_PD(Y,F);
1741             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1742             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1743             GMX_MM_TRANSPOSE2_PD(G,H);
1744             Heps             = _mm_mul_pd(vfeps,H);
1745             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1746             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1747             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
1748
1749             fscal            = felec;
1750
1751             /* Calculate temporary vectorial force */
1752             tx               = _mm_mul_pd(fscal,dx22);
1753             ty               = _mm_mul_pd(fscal,dy22);
1754             tz               = _mm_mul_pd(fscal,dz22);
1755
1756             /* Update vectorial force */
1757             fix2             = _mm_add_pd(fix2,tx);
1758             fiy2             = _mm_add_pd(fiy2,ty);
1759             fiz2             = _mm_add_pd(fiz2,tz);
1760
1761             fjx2             = _mm_add_pd(fjx2,tx);
1762             fjy2             = _mm_add_pd(fjy2,ty);
1763             fjz2             = _mm_add_pd(fjz2,tz);
1764
1765             gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1766
1767             /* Inner loop uses 351 flops */
1768         }
1769
1770         if(jidx<j_index_end)
1771         {
1772
1773             jnrA             = jjnr[jidx];
1774             j_coord_offsetA  = DIM*jnrA;
1775
1776             /* load j atom coordinates */
1777             gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
1778                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1779
1780             /* Calculate displacement vector */
1781             dx00             = _mm_sub_pd(ix0,jx0);
1782             dy00             = _mm_sub_pd(iy0,jy0);
1783             dz00             = _mm_sub_pd(iz0,jz0);
1784             dx01             = _mm_sub_pd(ix0,jx1);
1785             dy01             = _mm_sub_pd(iy0,jy1);
1786             dz01             = _mm_sub_pd(iz0,jz1);
1787             dx02             = _mm_sub_pd(ix0,jx2);
1788             dy02             = _mm_sub_pd(iy0,jy2);
1789             dz02             = _mm_sub_pd(iz0,jz2);
1790             dx10             = _mm_sub_pd(ix1,jx0);
1791             dy10             = _mm_sub_pd(iy1,jy0);
1792             dz10             = _mm_sub_pd(iz1,jz0);
1793             dx11             = _mm_sub_pd(ix1,jx1);
1794             dy11             = _mm_sub_pd(iy1,jy1);
1795             dz11             = _mm_sub_pd(iz1,jz1);
1796             dx12             = _mm_sub_pd(ix1,jx2);
1797             dy12             = _mm_sub_pd(iy1,jy2);
1798             dz12             = _mm_sub_pd(iz1,jz2);
1799             dx20             = _mm_sub_pd(ix2,jx0);
1800             dy20             = _mm_sub_pd(iy2,jy0);
1801             dz20             = _mm_sub_pd(iz2,jz0);
1802             dx21             = _mm_sub_pd(ix2,jx1);
1803             dy21             = _mm_sub_pd(iy2,jy1);
1804             dz21             = _mm_sub_pd(iz2,jz1);
1805             dx22             = _mm_sub_pd(ix2,jx2);
1806             dy22             = _mm_sub_pd(iy2,jy2);
1807             dz22             = _mm_sub_pd(iz2,jz2);
1808
1809             /* Calculate squared distance and things based on it */
1810             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1811             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
1812             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
1813             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
1814             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1815             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1816             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
1817             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1818             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1819
1820             rinv00           = sse41_invsqrt_d(rsq00);
1821             rinv01           = sse41_invsqrt_d(rsq01);
1822             rinv02           = sse41_invsqrt_d(rsq02);
1823             rinv10           = sse41_invsqrt_d(rsq10);
1824             rinv11           = sse41_invsqrt_d(rsq11);
1825             rinv12           = sse41_invsqrt_d(rsq12);
1826             rinv20           = sse41_invsqrt_d(rsq20);
1827             rinv21           = sse41_invsqrt_d(rsq21);
1828             rinv22           = sse41_invsqrt_d(rsq22);
1829
1830             fjx0             = _mm_setzero_pd();
1831             fjy0             = _mm_setzero_pd();
1832             fjz0             = _mm_setzero_pd();
1833             fjx1             = _mm_setzero_pd();
1834             fjy1             = _mm_setzero_pd();
1835             fjz1             = _mm_setzero_pd();
1836             fjx2             = _mm_setzero_pd();
1837             fjy2             = _mm_setzero_pd();
1838             fjz2             = _mm_setzero_pd();
1839
1840             /**************************
1841              * CALCULATE INTERACTIONS *
1842              **************************/
1843
1844             r00              = _mm_mul_pd(rsq00,rinv00);
1845
1846             /* Calculate table index by multiplying r with table scale and truncate to integer */
1847             rt               = _mm_mul_pd(r00,vftabscale);
1848             vfitab           = _mm_cvttpd_epi32(rt);
1849             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1850             vfitab           = _mm_slli_epi32(vfitab,2);
1851
1852             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1853             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1854             F                = _mm_setzero_pd();
1855             GMX_MM_TRANSPOSE2_PD(Y,F);
1856             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1857             H                = _mm_setzero_pd();
1858             GMX_MM_TRANSPOSE2_PD(G,H);
1859             Heps             = _mm_mul_pd(vfeps,H);
1860             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1861             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1862             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq00,FF),_mm_mul_pd(vftabscale,rinv00)));
1863
1864             fscal            = felec;
1865
1866             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1867
1868             /* Calculate temporary vectorial force */
1869             tx               = _mm_mul_pd(fscal,dx00);
1870             ty               = _mm_mul_pd(fscal,dy00);
1871             tz               = _mm_mul_pd(fscal,dz00);
1872
1873             /* Update vectorial force */
1874             fix0             = _mm_add_pd(fix0,tx);
1875             fiy0             = _mm_add_pd(fiy0,ty);
1876             fiz0             = _mm_add_pd(fiz0,tz);
1877
1878             fjx0             = _mm_add_pd(fjx0,tx);
1879             fjy0             = _mm_add_pd(fjy0,ty);
1880             fjz0             = _mm_add_pd(fjz0,tz);
1881
1882             /**************************
1883              * CALCULATE INTERACTIONS *
1884              **************************/
1885
1886             r01              = _mm_mul_pd(rsq01,rinv01);
1887
1888             /* Calculate table index by multiplying r with table scale and truncate to integer */
1889             rt               = _mm_mul_pd(r01,vftabscale);
1890             vfitab           = _mm_cvttpd_epi32(rt);
1891             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1892             vfitab           = _mm_slli_epi32(vfitab,2);
1893
1894             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1895             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1896             F                = _mm_setzero_pd();
1897             GMX_MM_TRANSPOSE2_PD(Y,F);
1898             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1899             H                = _mm_setzero_pd();
1900             GMX_MM_TRANSPOSE2_PD(G,H);
1901             Heps             = _mm_mul_pd(vfeps,H);
1902             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1903             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1904             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq01,FF),_mm_mul_pd(vftabscale,rinv01)));
1905
1906             fscal            = felec;
1907
1908             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1909
1910             /* Calculate temporary vectorial force */
1911             tx               = _mm_mul_pd(fscal,dx01);
1912             ty               = _mm_mul_pd(fscal,dy01);
1913             tz               = _mm_mul_pd(fscal,dz01);
1914
1915             /* Update vectorial force */
1916             fix0             = _mm_add_pd(fix0,tx);
1917             fiy0             = _mm_add_pd(fiy0,ty);
1918             fiz0             = _mm_add_pd(fiz0,tz);
1919
1920             fjx1             = _mm_add_pd(fjx1,tx);
1921             fjy1             = _mm_add_pd(fjy1,ty);
1922             fjz1             = _mm_add_pd(fjz1,tz);
1923
1924             /**************************
1925              * CALCULATE INTERACTIONS *
1926              **************************/
1927
1928             r02              = _mm_mul_pd(rsq02,rinv02);
1929
1930             /* Calculate table index by multiplying r with table scale and truncate to integer */
1931             rt               = _mm_mul_pd(r02,vftabscale);
1932             vfitab           = _mm_cvttpd_epi32(rt);
1933             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1934             vfitab           = _mm_slli_epi32(vfitab,2);
1935
1936             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1937             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1938             F                = _mm_setzero_pd();
1939             GMX_MM_TRANSPOSE2_PD(Y,F);
1940             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1941             H                = _mm_setzero_pd();
1942             GMX_MM_TRANSPOSE2_PD(G,H);
1943             Heps             = _mm_mul_pd(vfeps,H);
1944             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1945             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1946             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq02,FF),_mm_mul_pd(vftabscale,rinv02)));
1947
1948             fscal            = felec;
1949
1950             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1951
1952             /* Calculate temporary vectorial force */
1953             tx               = _mm_mul_pd(fscal,dx02);
1954             ty               = _mm_mul_pd(fscal,dy02);
1955             tz               = _mm_mul_pd(fscal,dz02);
1956
1957             /* Update vectorial force */
1958             fix0             = _mm_add_pd(fix0,tx);
1959             fiy0             = _mm_add_pd(fiy0,ty);
1960             fiz0             = _mm_add_pd(fiz0,tz);
1961
1962             fjx2             = _mm_add_pd(fjx2,tx);
1963             fjy2             = _mm_add_pd(fjy2,ty);
1964             fjz2             = _mm_add_pd(fjz2,tz);
1965
1966             /**************************
1967              * CALCULATE INTERACTIONS *
1968              **************************/
1969
1970             r10              = _mm_mul_pd(rsq10,rinv10);
1971
1972             /* Calculate table index by multiplying r with table scale and truncate to integer */
1973             rt               = _mm_mul_pd(r10,vftabscale);
1974             vfitab           = _mm_cvttpd_epi32(rt);
1975             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1976             vfitab           = _mm_slli_epi32(vfitab,2);
1977
1978             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1979             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1980             F                = _mm_setzero_pd();
1981             GMX_MM_TRANSPOSE2_PD(Y,F);
1982             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1983             H                = _mm_setzero_pd();
1984             GMX_MM_TRANSPOSE2_PD(G,H);
1985             Heps             = _mm_mul_pd(vfeps,H);
1986             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1987             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1988             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq10,FF),_mm_mul_pd(vftabscale,rinv10)));
1989
1990             fscal            = felec;
1991
1992             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1993
1994             /* Calculate temporary vectorial force */
1995             tx               = _mm_mul_pd(fscal,dx10);
1996             ty               = _mm_mul_pd(fscal,dy10);
1997             tz               = _mm_mul_pd(fscal,dz10);
1998
1999             /* Update vectorial force */
2000             fix1             = _mm_add_pd(fix1,tx);
2001             fiy1             = _mm_add_pd(fiy1,ty);
2002             fiz1             = _mm_add_pd(fiz1,tz);
2003
2004             fjx0             = _mm_add_pd(fjx0,tx);
2005             fjy0             = _mm_add_pd(fjy0,ty);
2006             fjz0             = _mm_add_pd(fjz0,tz);
2007
2008             /**************************
2009              * CALCULATE INTERACTIONS *
2010              **************************/
2011
2012             r11              = _mm_mul_pd(rsq11,rinv11);
2013
2014             /* Calculate table index by multiplying r with table scale and truncate to integer */
2015             rt               = _mm_mul_pd(r11,vftabscale);
2016             vfitab           = _mm_cvttpd_epi32(rt);
2017             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2018             vfitab           = _mm_slli_epi32(vfitab,2);
2019
2020             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2021             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2022             F                = _mm_setzero_pd();
2023             GMX_MM_TRANSPOSE2_PD(Y,F);
2024             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2025             H                = _mm_setzero_pd();
2026             GMX_MM_TRANSPOSE2_PD(G,H);
2027             Heps             = _mm_mul_pd(vfeps,H);
2028             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2029             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2030             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
2031
2032             fscal            = felec;
2033
2034             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2035
2036             /* Calculate temporary vectorial force */
2037             tx               = _mm_mul_pd(fscal,dx11);
2038             ty               = _mm_mul_pd(fscal,dy11);
2039             tz               = _mm_mul_pd(fscal,dz11);
2040
2041             /* Update vectorial force */
2042             fix1             = _mm_add_pd(fix1,tx);
2043             fiy1             = _mm_add_pd(fiy1,ty);
2044             fiz1             = _mm_add_pd(fiz1,tz);
2045
2046             fjx1             = _mm_add_pd(fjx1,tx);
2047             fjy1             = _mm_add_pd(fjy1,ty);
2048             fjz1             = _mm_add_pd(fjz1,tz);
2049
2050             /**************************
2051              * CALCULATE INTERACTIONS *
2052              **************************/
2053
2054             r12              = _mm_mul_pd(rsq12,rinv12);
2055
2056             /* Calculate table index by multiplying r with table scale and truncate to integer */
2057             rt               = _mm_mul_pd(r12,vftabscale);
2058             vfitab           = _mm_cvttpd_epi32(rt);
2059             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2060             vfitab           = _mm_slli_epi32(vfitab,2);
2061
2062             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2063             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2064             F                = _mm_setzero_pd();
2065             GMX_MM_TRANSPOSE2_PD(Y,F);
2066             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2067             H                = _mm_setzero_pd();
2068             GMX_MM_TRANSPOSE2_PD(G,H);
2069             Heps             = _mm_mul_pd(vfeps,H);
2070             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2071             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2072             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
2073
2074             fscal            = felec;
2075
2076             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2077
2078             /* Calculate temporary vectorial force */
2079             tx               = _mm_mul_pd(fscal,dx12);
2080             ty               = _mm_mul_pd(fscal,dy12);
2081             tz               = _mm_mul_pd(fscal,dz12);
2082
2083             /* Update vectorial force */
2084             fix1             = _mm_add_pd(fix1,tx);
2085             fiy1             = _mm_add_pd(fiy1,ty);
2086             fiz1             = _mm_add_pd(fiz1,tz);
2087
2088             fjx2             = _mm_add_pd(fjx2,tx);
2089             fjy2             = _mm_add_pd(fjy2,ty);
2090             fjz2             = _mm_add_pd(fjz2,tz);
2091
2092             /**************************
2093              * CALCULATE INTERACTIONS *
2094              **************************/
2095
2096             r20              = _mm_mul_pd(rsq20,rinv20);
2097
2098             /* Calculate table index by multiplying r with table scale and truncate to integer */
2099             rt               = _mm_mul_pd(r20,vftabscale);
2100             vfitab           = _mm_cvttpd_epi32(rt);
2101             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2102             vfitab           = _mm_slli_epi32(vfitab,2);
2103
2104             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2105             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2106             F                = _mm_setzero_pd();
2107             GMX_MM_TRANSPOSE2_PD(Y,F);
2108             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2109             H                = _mm_setzero_pd();
2110             GMX_MM_TRANSPOSE2_PD(G,H);
2111             Heps             = _mm_mul_pd(vfeps,H);
2112             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2113             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2114             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq20,FF),_mm_mul_pd(vftabscale,rinv20)));
2115
2116             fscal            = felec;
2117
2118             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2119
2120             /* Calculate temporary vectorial force */
2121             tx               = _mm_mul_pd(fscal,dx20);
2122             ty               = _mm_mul_pd(fscal,dy20);
2123             tz               = _mm_mul_pd(fscal,dz20);
2124
2125             /* Update vectorial force */
2126             fix2             = _mm_add_pd(fix2,tx);
2127             fiy2             = _mm_add_pd(fiy2,ty);
2128             fiz2             = _mm_add_pd(fiz2,tz);
2129
2130             fjx0             = _mm_add_pd(fjx0,tx);
2131             fjy0             = _mm_add_pd(fjy0,ty);
2132             fjz0             = _mm_add_pd(fjz0,tz);
2133
2134             /**************************
2135              * CALCULATE INTERACTIONS *
2136              **************************/
2137
2138             r21              = _mm_mul_pd(rsq21,rinv21);
2139
2140             /* Calculate table index by multiplying r with table scale and truncate to integer */
2141             rt               = _mm_mul_pd(r21,vftabscale);
2142             vfitab           = _mm_cvttpd_epi32(rt);
2143             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2144             vfitab           = _mm_slli_epi32(vfitab,2);
2145
2146             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2147             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2148             F                = _mm_setzero_pd();
2149             GMX_MM_TRANSPOSE2_PD(Y,F);
2150             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2151             H                = _mm_setzero_pd();
2152             GMX_MM_TRANSPOSE2_PD(G,H);
2153             Heps             = _mm_mul_pd(vfeps,H);
2154             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2155             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2156             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
2157
2158             fscal            = felec;
2159
2160             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2161
2162             /* Calculate temporary vectorial force */
2163             tx               = _mm_mul_pd(fscal,dx21);
2164             ty               = _mm_mul_pd(fscal,dy21);
2165             tz               = _mm_mul_pd(fscal,dz21);
2166
2167             /* Update vectorial force */
2168             fix2             = _mm_add_pd(fix2,tx);
2169             fiy2             = _mm_add_pd(fiy2,ty);
2170             fiz2             = _mm_add_pd(fiz2,tz);
2171
2172             fjx1             = _mm_add_pd(fjx1,tx);
2173             fjy1             = _mm_add_pd(fjy1,ty);
2174             fjz1             = _mm_add_pd(fjz1,tz);
2175
2176             /**************************
2177              * CALCULATE INTERACTIONS *
2178              **************************/
2179
2180             r22              = _mm_mul_pd(rsq22,rinv22);
2181
2182             /* Calculate table index by multiplying r with table scale and truncate to integer */
2183             rt               = _mm_mul_pd(r22,vftabscale);
2184             vfitab           = _mm_cvttpd_epi32(rt);
2185             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2186             vfitab           = _mm_slli_epi32(vfitab,2);
2187
2188             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2189             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2190             F                = _mm_setzero_pd();
2191             GMX_MM_TRANSPOSE2_PD(Y,F);
2192             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2193             H                = _mm_setzero_pd();
2194             GMX_MM_TRANSPOSE2_PD(G,H);
2195             Heps             = _mm_mul_pd(vfeps,H);
2196             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2197             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2198             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
2199
2200             fscal            = felec;
2201
2202             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2203
2204             /* Calculate temporary vectorial force */
2205             tx               = _mm_mul_pd(fscal,dx22);
2206             ty               = _mm_mul_pd(fscal,dy22);
2207             tz               = _mm_mul_pd(fscal,dz22);
2208
2209             /* Update vectorial force */
2210             fix2             = _mm_add_pd(fix2,tx);
2211             fiy2             = _mm_add_pd(fiy2,ty);
2212             fiz2             = _mm_add_pd(fiz2,tz);
2213
2214             fjx2             = _mm_add_pd(fjx2,tx);
2215             fjy2             = _mm_add_pd(fjy2,ty);
2216             fjz2             = _mm_add_pd(fjz2,tz);
2217
2218             gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
2219
2220             /* Inner loop uses 351 flops */
2221         }
2222
2223         /* End of innermost loop */
2224
2225         gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
2226                                               f+i_coord_offset,fshift+i_shift_offset);
2227
2228         /* Increment number of inner iterations */
2229         inneriter                  += j_index_end - j_index_start;
2230
2231         /* Outer loop uses 18 flops */
2232     }
2233
2234     /* Increment number of outer iterations */
2235     outeriter        += nri;
2236
2237     /* Update outer/inner flops */
2238
2239     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3W3_F,outeriter*18 + inneriter*351);
2240 }