src/gromacs/gmxlib/nonbonded/nb_kernel_avx_128_fma_double/nb_kernel_ElecCSTab_VdwNone_GeomW3W3_avx_128_fma_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 avx_128_fma_double kernel generator.
  37  */
  38 #include "gmxpre.h"
  39
  40 #include "config.h"
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "gromacs/gmxlib/nrnb.h"
  46
  47 #include "kernelutil_x86_avx_128_fma_double.h"
  48
  49 /*
  50  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwNone_GeomW3W3_VF_avx_128_fma_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_avx_128_fma_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,twovfeps;
 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           = avx128fma_invsqrt_d(rsq00);
 240             rinv01           = avx128fma_invsqrt_d(rsq01);
 241             rinv02           = avx128fma_invsqrt_d(rsq02);
 242             rinv10           = avx128fma_invsqrt_d(rsq10);
 243             rinv11           = avx128fma_invsqrt_d(rsq11);
 244             rinv12           = avx128fma_invsqrt_d(rsq12);
 245             rinv20           = avx128fma_invsqrt_d(rsq20);
 246             rinv21           = avx128fma_invsqrt_d(rsq21);
 247             rinv22           = avx128fma_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 #ifdef __XOP__
 269             vfeps            = _mm_frcz_pd(rt);
 270 #else
 271             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 272 #endif
 273             twovfeps         = _mm_add_pd(vfeps,vfeps);
 274             vfitab           = _mm_slli_epi32(vfitab,2);
 275
 276             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 277             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 278             F                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 279             GMX_MM_TRANSPOSE2_PD(Y,F);
 280             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
 281             H                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
 282             GMX_MM_TRANSPOSE2_PD(G,H);
 283             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
 284             VV               = _mm_macc_pd(vfeps,Fp,Y);
 285             velec            = _mm_mul_pd(qq00,VV);
 286             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
 287             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq00,FF),_mm_mul_pd(vftabscale,rinv00)));
 288
 289             /* Update potential sum for this i atom from the interaction with this j atom. */
 290             velecsum         = _mm_add_pd(velecsum,velec);
 291
 292             fscal            = felec;
 293
 294             /* Update vectorial force */
 295             fix0             = _mm_macc_pd(dx00,fscal,fix0);
 296             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
 297             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
 298
 299             fjx0             = _mm_macc_pd(dx00,fscal,fjx0);
 300             fjy0             = _mm_macc_pd(dy00,fscal,fjy0);
 301             fjz0             = _mm_macc_pd(dz00,fscal,fjz0);
 302
 303             /**************************
 304              * CALCULATE INTERACTIONS *
 305              **************************/
 306
 307             r01              = _mm_mul_pd(rsq01,rinv01);
 308
 309             /* Calculate table index by multiplying r with table scale and truncate to integer */
 310             rt               = _mm_mul_pd(r01,vftabscale);
 311             vfitab           = _mm_cvttpd_epi32(rt);
 312 #ifdef __XOP__
 313             vfeps            = _mm_frcz_pd(rt);
 314 #else
 315             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 316 #endif
 317             twovfeps         = _mm_add_pd(vfeps,vfeps);
 318             vfitab           = _mm_slli_epi32(vfitab,2);
 319
 320             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 321             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 322             F                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 323             GMX_MM_TRANSPOSE2_PD(Y,F);
 324             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
 325             H                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
 326             GMX_MM_TRANSPOSE2_PD(G,H);
 327             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
 328             VV               = _mm_macc_pd(vfeps,Fp,Y);
 329             velec            = _mm_mul_pd(qq01,VV);
 330             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
 331             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq01,FF),_mm_mul_pd(vftabscale,rinv01)));
 332
 333             /* Update potential sum for this i atom from the interaction with this j atom. */
 334             velecsum         = _mm_add_pd(velecsum,velec);
 335
 336             fscal            = felec;
 337
 338             /* Update vectorial force */
 339             fix0             = _mm_macc_pd(dx01,fscal,fix0);
 340             fiy0             = _mm_macc_pd(dy01,fscal,fiy0);
 341             fiz0             = _mm_macc_pd(dz01,fscal,fiz0);
 342
 343             fjx1             = _mm_macc_pd(dx01,fscal,fjx1);
 344             fjy1             = _mm_macc_pd(dy01,fscal,fjy1);
 345             fjz1             = _mm_macc_pd(dz01,fscal,fjz1);
 346
 347             /**************************
 348              * CALCULATE INTERACTIONS *
 349              **************************/
 350
 351             r02              = _mm_mul_pd(rsq02,rinv02);
 352
 353             /* Calculate table index by multiplying r with table scale and truncate to integer */
 354             rt               = _mm_mul_pd(r02,vftabscale);
 355             vfitab           = _mm_cvttpd_epi32(rt);
 356 #ifdef __XOP__
 357             vfeps            = _mm_frcz_pd(rt);
 358 #else
 359             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 360 #endif
 361             twovfeps         = _mm_add_pd(vfeps,vfeps);
 362             vfitab           = _mm_slli_epi32(vfitab,2);
 363
 364             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 365             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 366             F                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 367             GMX_MM_TRANSPOSE2_PD(Y,F);
 368             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
 369             H                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
 370             GMX_MM_TRANSPOSE2_PD(G,H);
 371             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
 372             VV               = _mm_macc_pd(vfeps,Fp,Y);
 373             velec            = _mm_mul_pd(qq02,VV);
 374             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
 375             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq02,FF),_mm_mul_pd(vftabscale,rinv02)));
 376
 377             /* Update potential sum for this i atom from the interaction with this j atom. */
 378             velecsum         = _mm_add_pd(velecsum,velec);
 379
 380             fscal            = felec;
 381
 382             /* Update vectorial force */
 383             fix0             = _mm_macc_pd(dx02,fscal,fix0);
 384             fiy0             = _mm_macc_pd(dy02,fscal,fiy0);
 385             fiz0             = _mm_macc_pd(dz02,fscal,fiz0);
 386
 387             fjx2             = _mm_macc_pd(dx02,fscal,fjx2);
 388             fjy2             = _mm_macc_pd(dy02,fscal,fjy2);
 389             fjz2             = _mm_macc_pd(dz02,fscal,fjz2);
 390
 391             /**************************
 392              * CALCULATE INTERACTIONS *
 393              **************************/
 394
 395             r10              = _mm_mul_pd(rsq10,rinv10);
 396
 397             /* Calculate table index by multiplying r with table scale and truncate to integer */
 398             rt               = _mm_mul_pd(r10,vftabscale);
 399             vfitab           = _mm_cvttpd_epi32(rt);
 400 #ifdef __XOP__
 401             vfeps            = _mm_frcz_pd(rt);
 402 #else
 403             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 404 #endif
 405             twovfeps         = _mm_add_pd(vfeps,vfeps);
 406             vfitab           = _mm_slli_epi32(vfitab,2);
 407
 408             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 409             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 410             F                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 411             GMX_MM_TRANSPOSE2_PD(Y,F);
 412             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
 413             H                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
 414             GMX_MM_TRANSPOSE2_PD(G,H);
 415             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
 416             VV               = _mm_macc_pd(vfeps,Fp,Y);
 417             velec            = _mm_mul_pd(qq10,VV);
 418             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
 419             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq10,FF),_mm_mul_pd(vftabscale,rinv10)));
 420
 421             /* Update potential sum for this i atom from the interaction with this j atom. */
 422             velecsum         = _mm_add_pd(velecsum,velec);
 423
 424             fscal            = felec;
 425
 426             /* Update vectorial force */
 427             fix1             = _mm_macc_pd(dx10,fscal,fix1);
 428             fiy1             = _mm_macc_pd(dy10,fscal,fiy1);
 429             fiz1             = _mm_macc_pd(dz10,fscal,fiz1);
 430
 431             fjx0             = _mm_macc_pd(dx10,fscal,fjx0);
 432             fjy0             = _mm_macc_pd(dy10,fscal,fjy0);
 433             fjz0             = _mm_macc_pd(dz10,fscal,fjz0);
 434
 435             /**************************
 436              * CALCULATE INTERACTIONS *
 437              **************************/
 438
 439             r11              = _mm_mul_pd(rsq11,rinv11);
 440
 441             /* Calculate table index by multiplying r with table scale and truncate to integer */
 442             rt               = _mm_mul_pd(r11,vftabscale);
 443             vfitab           = _mm_cvttpd_epi32(rt);
 444 #ifdef __XOP__
 445             vfeps            = _mm_frcz_pd(rt);
 446 #else
 447             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 448 #endif
 449             twovfeps         = _mm_add_pd(vfeps,vfeps);
 450             vfitab           = _mm_slli_epi32(vfitab,2);
 451
 452             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 453             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 454             F                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 455             GMX_MM_TRANSPOSE2_PD(Y,F);
 456             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
 457             H                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
 458             GMX_MM_TRANSPOSE2_PD(G,H);
 459             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
 460             VV               = _mm_macc_pd(vfeps,Fp,Y);
 461             velec            = _mm_mul_pd(qq11,VV);
 462             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
 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             /* Update vectorial force */
 471             fix1             = _mm_macc_pd(dx11,fscal,fix1);
 472             fiy1             = _mm_macc_pd(dy11,fscal,fiy1);
 473             fiz1             = _mm_macc_pd(dz11,fscal,fiz1);
 474
 475             fjx1             = _mm_macc_pd(dx11,fscal,fjx1);
 476             fjy1             = _mm_macc_pd(dy11,fscal,fjy1);
 477             fjz1             = _mm_macc_pd(dz11,fscal,fjz1);
 478
 479             /**************************
 480              * CALCULATE INTERACTIONS *
 481              **************************/
 482
 483             r12              = _mm_mul_pd(rsq12,rinv12);
 484
 485             /* Calculate table index by multiplying r with table scale and truncate to integer */
 486             rt               = _mm_mul_pd(r12,vftabscale);
 487             vfitab           = _mm_cvttpd_epi32(rt);
 488 #ifdef __XOP__
 489             vfeps            = _mm_frcz_pd(rt);
 490 #else
 491             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 492 #endif
 493             twovfeps         = _mm_add_pd(vfeps,vfeps);
 494             vfitab           = _mm_slli_epi32(vfitab,2);
 495
 496             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 497             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 498             F                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 499             GMX_MM_TRANSPOSE2_PD(Y,F);
 500             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
 501             H                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
 502             GMX_MM_TRANSPOSE2_PD(G,H);
 503             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
 504             VV               = _mm_macc_pd(vfeps,Fp,Y);
 505             velec            = _mm_mul_pd(qq12,VV);
 506             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
 507             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
 508
 509             /* Update potential sum for this i atom from the interaction with this j atom. */
 510             velecsum         = _mm_add_pd(velecsum,velec);
 511
 512             fscal            = felec;
 513
 514             /* Update vectorial force */
 515             fix1             = _mm_macc_pd(dx12,fscal,fix1);
 516             fiy1             = _mm_macc_pd(dy12,fscal,fiy1);
 517             fiz1             = _mm_macc_pd(dz12,fscal,fiz1);
 518
 519             fjx2             = _mm_macc_pd(dx12,fscal,fjx2);
 520             fjy2             = _mm_macc_pd(dy12,fscal,fjy2);
 521             fjz2             = _mm_macc_pd(dz12,fscal,fjz2);
 522
 523             /**************************
 524              * CALCULATE INTERACTIONS *
 525              **************************/
 526
 527             r20              = _mm_mul_pd(rsq20,rinv20);
 528
 529             /* Calculate table index by multiplying r with table scale and truncate to integer */
 530             rt               = _mm_mul_pd(r20,vftabscale);
 531             vfitab           = _mm_cvttpd_epi32(rt);
 532 #ifdef __XOP__
 533             vfeps            = _mm_frcz_pd(rt);
 534 #else
 535             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 536 #endif
 537             twovfeps         = _mm_add_pd(vfeps,vfeps);
 538             vfitab           = _mm_slli_epi32(vfitab,2);
 539
 540             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 541             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 542             F                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 543             GMX_MM_TRANSPOSE2_PD(Y,F);
 544             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
 545             H                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
 546             GMX_MM_TRANSPOSE2_PD(G,H);
 547             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
 548             VV               = _mm_macc_pd(vfeps,Fp,Y);
 549             velec            = _mm_mul_pd(qq20,VV);
 550             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
 551             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq20,FF),_mm_mul_pd(vftabscale,rinv20)));
 552
 553             /* Update potential sum for this i atom from the interaction with this j atom. */
 554             velecsum         = _mm_add_pd(velecsum,velec);
 555
 556             fscal            = felec;
 557
 558             /* Update vectorial force */
 559             fix2             = _mm_macc_pd(dx20,fscal,fix2);
 560             fiy2             = _mm_macc_pd(dy20,fscal,fiy2);
 561             fiz2             = _mm_macc_pd(dz20,fscal,fiz2);
 562
 563             fjx0             = _mm_macc_pd(dx20,fscal,fjx0);
 564             fjy0             = _mm_macc_pd(dy20,fscal,fjy0);
 565             fjz0             = _mm_macc_pd(dz20,fscal,fjz0);
 566
 567             /**************************
 568              * CALCULATE INTERACTIONS *
 569              **************************/
 570
 571             r21              = _mm_mul_pd(rsq21,rinv21);
 572
 573             /* Calculate table index by multiplying r with table scale and truncate to integer */
 574             rt               = _mm_mul_pd(r21,vftabscale);
 575             vfitab           = _mm_cvttpd_epi32(rt);
 576 #ifdef __XOP__
 577             vfeps            = _mm_frcz_pd(rt);
 578 #else
 579             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 580 #endif
 581             twovfeps         = _mm_add_pd(vfeps,vfeps);
 582             vfitab           = _mm_slli_epi32(vfitab,2);
 583
 584             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 585             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 586             F                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 587             GMX_MM_TRANSPOSE2_PD(Y,F);
 588             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
 589             H                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
 590             GMX_MM_TRANSPOSE2_PD(G,H);
 591             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
 592             VV               = _mm_macc_pd(vfeps,Fp,Y);
 593             velec            = _mm_mul_pd(qq21,VV);
 594             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
 595             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
 596
 597             /* Update potential sum for this i atom from the interaction with this j atom. */
 598             velecsum         = _mm_add_pd(velecsum,velec);
 599
 600             fscal            = felec;
 601
 602             /* Update vectorial force */
 603             fix2             = _mm_macc_pd(dx21,fscal,fix2);
 604             fiy2             = _mm_macc_pd(dy21,fscal,fiy2);
 605             fiz2             = _mm_macc_pd(dz21,fscal,fiz2);
 606
 607             fjx1             = _mm_macc_pd(dx21,fscal,fjx1);
 608             fjy1             = _mm_macc_pd(dy21,fscal,fjy1);
 609             fjz1             = _mm_macc_pd(dz21,fscal,fjz1);
 610
 611             /**************************
 612              * CALCULATE INTERACTIONS *
 613              **************************/
 614
 615             r22              = _mm_mul_pd(rsq22,rinv22);
 616
 617             /* Calculate table index by multiplying r with table scale and truncate to integer */
 618             rt               = _mm_mul_pd(r22,vftabscale);
 619             vfitab           = _mm_cvttpd_epi32(rt);
 620 #ifdef __XOP__
 621             vfeps            = _mm_frcz_pd(rt);
 622 #else
 623             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 624 #endif
 625             twovfeps         = _mm_add_pd(vfeps,vfeps);
 626             vfitab           = _mm_slli_epi32(vfitab,2);
 627
 628             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 629             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 630             F                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 631             GMX_MM_TRANSPOSE2_PD(Y,F);
 632             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
 633             H                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
 634             GMX_MM_TRANSPOSE2_PD(G,H);
 635             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
 636             VV               = _mm_macc_pd(vfeps,Fp,Y);
 637             velec            = _mm_mul_pd(qq22,VV);
 638             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
 639             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
 640
 641             /* Update potential sum for this i atom from the interaction with this j atom. */
 642             velecsum         = _mm_add_pd(velecsum,velec);
 643
 644             fscal            = felec;
 645
 646             /* Update vectorial force */
 647             fix2             = _mm_macc_pd(dx22,fscal,fix2);
 648             fiy2             = _mm_macc_pd(dy22,fscal,fiy2);
 649             fiz2             = _mm_macc_pd(dz22,fscal,fiz2);
 650
 651             fjx2             = _mm_macc_pd(dx22,fscal,fjx2);
 652             fjy2             = _mm_macc_pd(dy22,fscal,fjy2);
 653             fjz2             = _mm_macc_pd(dz22,fscal,fjz2);
 654
 655             gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 656
 657             /* Inner loop uses 414 flops */
 658         }
 659
 660         if(jidx<j_index_end)
 661         {
 662
 663             jnrA             = jjnr[jidx];
 664             j_coord_offsetA  = DIM*jnrA;
 665
 666             /* load j atom coordinates */
 667             gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 668                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 669
 670             /* Calculate displacement vector */
 671             dx00             = _mm_sub_pd(ix0,jx0);
 672             dy00             = _mm_sub_pd(iy0,jy0);
 673             dz00             = _mm_sub_pd(iz0,jz0);
 674             dx01             = _mm_sub_pd(ix0,jx1);
 675             dy01             = _mm_sub_pd(iy0,jy1);
 676             dz01             = _mm_sub_pd(iz0,jz1);
 677             dx02             = _mm_sub_pd(ix0,jx2);
 678             dy02             = _mm_sub_pd(iy0,jy2);
 679             dz02             = _mm_sub_pd(iz0,jz2);
 680             dx10             = _mm_sub_pd(ix1,jx0);
 681             dy10             = _mm_sub_pd(iy1,jy0);
 682             dz10             = _mm_sub_pd(iz1,jz0);
 683             dx11             = _mm_sub_pd(ix1,jx1);
 684             dy11             = _mm_sub_pd(iy1,jy1);
 685             dz11             = _mm_sub_pd(iz1,jz1);
 686             dx12             = _mm_sub_pd(ix1,jx2);
 687             dy12             = _mm_sub_pd(iy1,jy2);
 688             dz12             = _mm_sub_pd(iz1,jz2);
 689             dx20             = _mm_sub_pd(ix2,jx0);
 690             dy20             = _mm_sub_pd(iy2,jy0);
 691             dz20             = _mm_sub_pd(iz2,jz0);
 692             dx21             = _mm_sub_pd(ix2,jx1);
 693             dy21             = _mm_sub_pd(iy2,jy1);
 694             dz21             = _mm_sub_pd(iz2,jz1);
 695             dx22             = _mm_sub_pd(ix2,jx2);
 696             dy22             = _mm_sub_pd(iy2,jy2);
 697             dz22             = _mm_sub_pd(iz2,jz2);
 698
 699             /* Calculate squared distance and things based on it */
 700             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 701             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
 702             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
 703             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 704             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 705             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 706             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 707             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 708             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 709
 710             rinv00           = avx128fma_invsqrt_d(rsq00);
 711             rinv01           = avx128fma_invsqrt_d(rsq01);
 712             rinv02           = avx128fma_invsqrt_d(rsq02);
 713             rinv10           = avx128fma_invsqrt_d(rsq10);
 714             rinv11           = avx128fma_invsqrt_d(rsq11);
 715             rinv12           = avx128fma_invsqrt_d(rsq12);
 716             rinv20           = avx128fma_invsqrt_d(rsq20);
 717             rinv21           = avx128fma_invsqrt_d(rsq21);
 718             rinv22           = avx128fma_invsqrt_d(rsq22);
 719
 720             fjx0             = _mm_setzero_pd();
 721             fjy0             = _mm_setzero_pd();
 722             fjz0             = _mm_setzero_pd();
 723             fjx1             = _mm_setzero_pd();
 724             fjy1             = _mm_setzero_pd();
 725             fjz1             = _mm_setzero_pd();
 726             fjx2             = _mm_setzero_pd();
 727             fjy2             = _mm_setzero_pd();
 728             fjz2             = _mm_setzero_pd();
 729
 730             /**************************
 731              * CALCULATE INTERACTIONS *
 732              **************************/
 733
 734             r00              = _mm_mul_pd(rsq00,rinv00);
 735
 736             /* Calculate table index by multiplying r with table scale and truncate to integer */
 737             rt               = _mm_mul_pd(r00,vftabscale);
 738             vfitab           = _mm_cvttpd_epi32(rt);
 739 #ifdef __XOP__
 740             vfeps            = _mm_frcz_pd(rt);
 741 #else
 742             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 743 #endif
 744             twovfeps         = _mm_add_pd(vfeps,vfeps);
 745             vfitab           = _mm_slli_epi32(vfitab,2);
 746
 747             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 748             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 749             F                = _mm_setzero_pd();
 750             GMX_MM_TRANSPOSE2_PD(Y,F);
 751             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
 752             H                = _mm_setzero_pd();
 753             GMX_MM_TRANSPOSE2_PD(G,H);
 754             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
 755             VV               = _mm_macc_pd(vfeps,Fp,Y);
 756             velec            = _mm_mul_pd(qq00,VV);
 757             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
 758             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq00,FF),_mm_mul_pd(vftabscale,rinv00)));
 759
 760             /* Update potential sum for this i atom from the interaction with this j atom. */
 761             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 762             velecsum         = _mm_add_pd(velecsum,velec);
 763
 764             fscal            = felec;
 765
 766             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 767
 768             /* Update vectorial force */
 769             fix0             = _mm_macc_pd(dx00,fscal,fix0);
 770             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
 771             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
 772
 773             fjx0             = _mm_macc_pd(dx00,fscal,fjx0);
 774             fjy0             = _mm_macc_pd(dy00,fscal,fjy0);
 775             fjz0             = _mm_macc_pd(dz00,fscal,fjz0);
 776
 777             /**************************
 778              * CALCULATE INTERACTIONS *
 779              **************************/
 780
 781             r01              = _mm_mul_pd(rsq01,rinv01);
 782
 783             /* Calculate table index by multiplying r with table scale and truncate to integer */
 784             rt               = _mm_mul_pd(r01,vftabscale);
 785             vfitab           = _mm_cvttpd_epi32(rt);
 786 #ifdef __XOP__
 787             vfeps            = _mm_frcz_pd(rt);
 788 #else
 789             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 790 #endif
 791             twovfeps         = _mm_add_pd(vfeps,vfeps);
 792             vfitab           = _mm_slli_epi32(vfitab,2);
 793
 794             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 795             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 796             F                = _mm_setzero_pd();
 797             GMX_MM_TRANSPOSE2_PD(Y,F);
 798             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
 799             H                = _mm_setzero_pd();
 800             GMX_MM_TRANSPOSE2_PD(G,H);
 801             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
 802             VV               = _mm_macc_pd(vfeps,Fp,Y);
 803             velec            = _mm_mul_pd(qq01,VV);
 804             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
 805             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq01,FF),_mm_mul_pd(vftabscale,rinv01)));
 806
 807             /* Update potential sum for this i atom from the interaction with this j atom. */
 808             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 809             velecsum         = _mm_add_pd(velecsum,velec);
 810
 811             fscal            = felec;
 812
 813             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 814
 815             /* Update vectorial force */
 816             fix0             = _mm_macc_pd(dx01,fscal,fix0);
 817             fiy0             = _mm_macc_pd(dy01,fscal,fiy0);
 818             fiz0             = _mm_macc_pd(dz01,fscal,fiz0);
 819
 820             fjx1             = _mm_macc_pd(dx01,fscal,fjx1);
 821             fjy1             = _mm_macc_pd(dy01,fscal,fjy1);
 822             fjz1             = _mm_macc_pd(dz01,fscal,fjz1);
 823
 824             /**************************
 825              * CALCULATE INTERACTIONS *
 826              **************************/
 827
 828             r02              = _mm_mul_pd(rsq02,rinv02);
 829
 830             /* Calculate table index by multiplying r with table scale and truncate to integer */
 831             rt               = _mm_mul_pd(r02,vftabscale);
 832             vfitab           = _mm_cvttpd_epi32(rt);
 833 #ifdef __XOP__
 834             vfeps            = _mm_frcz_pd(rt);
 835 #else
 836             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 837 #endif
 838             twovfeps         = _mm_add_pd(vfeps,vfeps);
 839             vfitab           = _mm_slli_epi32(vfitab,2);
 840
 841             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 842             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 843             F                = _mm_setzero_pd();
 844             GMX_MM_TRANSPOSE2_PD(Y,F);
 845             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
 846             H                = _mm_setzero_pd();
 847             GMX_MM_TRANSPOSE2_PD(G,H);
 848             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
 849             VV               = _mm_macc_pd(vfeps,Fp,Y);
 850             velec            = _mm_mul_pd(qq02,VV);
 851             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
 852             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq02,FF),_mm_mul_pd(vftabscale,rinv02)));
 853
 854             /* Update potential sum for this i atom from the interaction with this j atom. */
 855             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 856             velecsum         = _mm_add_pd(velecsum,velec);
 857
 858             fscal            = felec;
 859
 860             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 861
 862             /* Update vectorial force */
 863             fix0             = _mm_macc_pd(dx02,fscal,fix0);
 864             fiy0             = _mm_macc_pd(dy02,fscal,fiy0);
 865             fiz0             = _mm_macc_pd(dz02,fscal,fiz0);
 866
 867             fjx2             = _mm_macc_pd(dx02,fscal,fjx2);
 868             fjy2             = _mm_macc_pd(dy02,fscal,fjy2);
 869             fjz2             = _mm_macc_pd(dz02,fscal,fjz2);
 870
 871             /**************************
 872              * CALCULATE INTERACTIONS *
 873              **************************/
 874
 875             r10              = _mm_mul_pd(rsq10,rinv10);
 876
 877             /* Calculate table index by multiplying r with table scale and truncate to integer */
 878             rt               = _mm_mul_pd(r10,vftabscale);
 879             vfitab           = _mm_cvttpd_epi32(rt);
 880 #ifdef __XOP__
 881             vfeps            = _mm_frcz_pd(rt);
 882 #else
 883             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 884 #endif
 885             twovfeps         = _mm_add_pd(vfeps,vfeps);
 886             vfitab           = _mm_slli_epi32(vfitab,2);
 887
 888             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 889             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 890             F                = _mm_setzero_pd();
 891             GMX_MM_TRANSPOSE2_PD(Y,F);
 892             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
 893             H                = _mm_setzero_pd();
 894             GMX_MM_TRANSPOSE2_PD(G,H);
 895             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
 896             VV               = _mm_macc_pd(vfeps,Fp,Y);
 897             velec            = _mm_mul_pd(qq10,VV);
 898             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
 899             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq10,FF),_mm_mul_pd(vftabscale,rinv10)));
 900
 901             /* Update potential sum for this i atom from the interaction with this j atom. */
 902             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 903             velecsum         = _mm_add_pd(velecsum,velec);
 904
 905             fscal            = felec;
 906
 907             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 908
 909             /* Update vectorial force */
 910             fix1             = _mm_macc_pd(dx10,fscal,fix1);
 911             fiy1             = _mm_macc_pd(dy10,fscal,fiy1);
 912             fiz1             = _mm_macc_pd(dz10,fscal,fiz1);
 913
 914             fjx0             = _mm_macc_pd(dx10,fscal,fjx0);
 915             fjy0             = _mm_macc_pd(dy10,fscal,fjy0);
 916             fjz0             = _mm_macc_pd(dz10,fscal,fjz0);
 917
 918             /**************************
 919              * CALCULATE INTERACTIONS *
 920              **************************/
 921
 922             r11              = _mm_mul_pd(rsq11,rinv11);
 923
 924             /* Calculate table index by multiplying r with table scale and truncate to integer */
 925             rt               = _mm_mul_pd(r11,vftabscale);
 926             vfitab           = _mm_cvttpd_epi32(rt);
 927 #ifdef __XOP__
 928             vfeps            = _mm_frcz_pd(rt);
 929 #else
 930             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 931 #endif
 932             twovfeps         = _mm_add_pd(vfeps,vfeps);
 933             vfitab           = _mm_slli_epi32(vfitab,2);
 934
 935             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 936             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 937             F                = _mm_setzero_pd();
 938             GMX_MM_TRANSPOSE2_PD(Y,F);
 939             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
 940             H                = _mm_setzero_pd();
 941             GMX_MM_TRANSPOSE2_PD(G,H);
 942             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
 943             VV               = _mm_macc_pd(vfeps,Fp,Y);
 944             velec            = _mm_mul_pd(qq11,VV);
 945             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
 946             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
 947
 948             /* Update potential sum for this i atom from the interaction with this j atom. */
 949             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 950             velecsum         = _mm_add_pd(velecsum,velec);
 951
 952             fscal            = felec;
 953
 954             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 955
 956             /* Update vectorial force */
 957             fix1             = _mm_macc_pd(dx11,fscal,fix1);
 958             fiy1             = _mm_macc_pd(dy11,fscal,fiy1);
 959             fiz1             = _mm_macc_pd(dz11,fscal,fiz1);
 960
 961             fjx1             = _mm_macc_pd(dx11,fscal,fjx1);
 962             fjy1             = _mm_macc_pd(dy11,fscal,fjy1);
 963             fjz1             = _mm_macc_pd(dz11,fscal,fjz1);
 964
 965             /**************************
 966              * CALCULATE INTERACTIONS *
 967              **************************/
 968
 969             r12              = _mm_mul_pd(rsq12,rinv12);
 970
 971             /* Calculate table index by multiplying r with table scale and truncate to integer */
 972             rt               = _mm_mul_pd(r12,vftabscale);
 973             vfitab           = _mm_cvttpd_epi32(rt);
 974 #ifdef __XOP__
 975             vfeps            = _mm_frcz_pd(rt);
 976 #else
 977             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 978 #endif
 979             twovfeps         = _mm_add_pd(vfeps,vfeps);
 980             vfitab           = _mm_slli_epi32(vfitab,2);
 981
 982             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 983             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 984             F                = _mm_setzero_pd();
 985             GMX_MM_TRANSPOSE2_PD(Y,F);
 986             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
 987             H                = _mm_setzero_pd();
 988             GMX_MM_TRANSPOSE2_PD(G,H);
 989             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
 990             VV               = _mm_macc_pd(vfeps,Fp,Y);
 991             velec            = _mm_mul_pd(qq12,VV);
 992             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
 993             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
 994
 995             /* Update potential sum for this i atom from the interaction with this j atom. */
 996             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 997             velecsum         = _mm_add_pd(velecsum,velec);
 998
 999             fscal            = felec;
1000
1001             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1002
1003             /* Update vectorial force */
1004             fix1             = _mm_macc_pd(dx12,fscal,fix1);
1005             fiy1             = _mm_macc_pd(dy12,fscal,fiy1);
1006             fiz1             = _mm_macc_pd(dz12,fscal,fiz1);
1007
1008             fjx2             = _mm_macc_pd(dx12,fscal,fjx2);
1009             fjy2             = _mm_macc_pd(dy12,fscal,fjy2);
1010             fjz2             = _mm_macc_pd(dz12,fscal,fjz2);
1011
1012             /**************************
1013              * CALCULATE INTERACTIONS *
1014              **************************/
1015
1016             r20              = _mm_mul_pd(rsq20,rinv20);
1017
1018             /* Calculate table index by multiplying r with table scale and truncate to integer */
1019             rt               = _mm_mul_pd(r20,vftabscale);
1020             vfitab           = _mm_cvttpd_epi32(rt);
1021 #ifdef __XOP__
1022             vfeps            = _mm_frcz_pd(rt);
1023 #else
1024             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1025 #endif
1026             twovfeps         = _mm_add_pd(vfeps,vfeps);
1027             vfitab           = _mm_slli_epi32(vfitab,2);
1028
1029             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1030             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1031             F                = _mm_setzero_pd();
1032             GMX_MM_TRANSPOSE2_PD(Y,F);
1033             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1034             H                = _mm_setzero_pd();
1035             GMX_MM_TRANSPOSE2_PD(G,H);
1036             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1037             VV               = _mm_macc_pd(vfeps,Fp,Y);
1038             velec            = _mm_mul_pd(qq20,VV);
1039             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1040             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq20,FF),_mm_mul_pd(vftabscale,rinv20)));
1041
1042             /* Update potential sum for this i atom from the interaction with this j atom. */
1043             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1044             velecsum         = _mm_add_pd(velecsum,velec);
1045
1046             fscal            = felec;
1047
1048             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1049
1050             /* Update vectorial force */
1051             fix2             = _mm_macc_pd(dx20,fscal,fix2);
1052             fiy2             = _mm_macc_pd(dy20,fscal,fiy2);
1053             fiz2             = _mm_macc_pd(dz20,fscal,fiz2);
1054
1055             fjx0             = _mm_macc_pd(dx20,fscal,fjx0);
1056             fjy0             = _mm_macc_pd(dy20,fscal,fjy0);
1057             fjz0             = _mm_macc_pd(dz20,fscal,fjz0);
1058
1059             /**************************
1060              * CALCULATE INTERACTIONS *
1061              **************************/
1062
1063             r21              = _mm_mul_pd(rsq21,rinv21);
1064
1065             /* Calculate table index by multiplying r with table scale and truncate to integer */
1066             rt               = _mm_mul_pd(r21,vftabscale);
1067             vfitab           = _mm_cvttpd_epi32(rt);
1068 #ifdef __XOP__
1069             vfeps            = _mm_frcz_pd(rt);
1070 #else
1071             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1072 #endif
1073             twovfeps         = _mm_add_pd(vfeps,vfeps);
1074             vfitab           = _mm_slli_epi32(vfitab,2);
1075
1076             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1077             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1078             F                = _mm_setzero_pd();
1079             GMX_MM_TRANSPOSE2_PD(Y,F);
1080             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1081             H                = _mm_setzero_pd();
1082             GMX_MM_TRANSPOSE2_PD(G,H);
1083             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1084             VV               = _mm_macc_pd(vfeps,Fp,Y);
1085             velec            = _mm_mul_pd(qq21,VV);
1086             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1087             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
1088
1089             /* Update potential sum for this i atom from the interaction with this j atom. */
1090             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1091             velecsum         = _mm_add_pd(velecsum,velec);
1092
1093             fscal            = felec;
1094
1095             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1096
1097             /* Update vectorial force */
1098             fix2             = _mm_macc_pd(dx21,fscal,fix2);
1099             fiy2             = _mm_macc_pd(dy21,fscal,fiy2);
1100             fiz2             = _mm_macc_pd(dz21,fscal,fiz2);
1101
1102             fjx1             = _mm_macc_pd(dx21,fscal,fjx1);
1103             fjy1             = _mm_macc_pd(dy21,fscal,fjy1);
1104             fjz1             = _mm_macc_pd(dz21,fscal,fjz1);
1105
1106             /**************************
1107              * CALCULATE INTERACTIONS *
1108              **************************/
1109
1110             r22              = _mm_mul_pd(rsq22,rinv22);
1111
1112             /* Calculate table index by multiplying r with table scale and truncate to integer */
1113             rt               = _mm_mul_pd(r22,vftabscale);
1114             vfitab           = _mm_cvttpd_epi32(rt);
1115 #ifdef __XOP__
1116             vfeps            = _mm_frcz_pd(rt);
1117 #else
1118             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1119 #endif
1120             twovfeps         = _mm_add_pd(vfeps,vfeps);
1121             vfitab           = _mm_slli_epi32(vfitab,2);
1122
1123             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1124             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1125             F                = _mm_setzero_pd();
1126             GMX_MM_TRANSPOSE2_PD(Y,F);
1127             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1128             H                = _mm_setzero_pd();
1129             GMX_MM_TRANSPOSE2_PD(G,H);
1130             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1131             VV               = _mm_macc_pd(vfeps,Fp,Y);
1132             velec            = _mm_mul_pd(qq22,VV);
1133             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1134             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
1135
1136             /* Update potential sum for this i atom from the interaction with this j atom. */
1137             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1138             velecsum         = _mm_add_pd(velecsum,velec);
1139
1140             fscal            = felec;
1141
1142             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1143
1144             /* Update vectorial force */
1145             fix2             = _mm_macc_pd(dx22,fscal,fix2);
1146             fiy2             = _mm_macc_pd(dy22,fscal,fiy2);
1147             fiz2             = _mm_macc_pd(dz22,fscal,fiz2);
1148
1149             fjx2             = _mm_macc_pd(dx22,fscal,fjx2);
1150             fjy2             = _mm_macc_pd(dy22,fscal,fjy2);
1151             fjz2             = _mm_macc_pd(dz22,fscal,fjz2);
1152
1153             gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1154
1155             /* Inner loop uses 414 flops */
1156         }
1157
1158         /* End of innermost loop */
1159
1160         gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1161                                               f+i_coord_offset,fshift+i_shift_offset);
1162
1163         ggid                        = gid[iidx];
1164         /* Update potential energies */
1165         gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1166
1167         /* Increment number of inner iterations */
1168         inneriter                  += j_index_end - j_index_start;
1169
1170         /* Outer loop uses 19 flops */
1171     }
1172
1173     /* Increment number of outer iterations */
1174     outeriter        += nri;
1175
1176     /* Update outer/inner flops */
1177
1178     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3W3_VF,outeriter*19 + inneriter*414);
1179 }
1180 /*
1181  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwNone_GeomW3W3_F_avx_128_fma_double
1182  * Electrostatics interaction: CubicSplineTable
1183  * VdW interaction:            None
1184  * Geometry:                   Water3-Water3
1185  * Calculate force/pot:        Force
1186  */
1187 void
1188 nb_kernel_ElecCSTab_VdwNone_GeomW3W3_F_avx_128_fma_double
1189                     (t_nblist                    * gmx_restrict       nlist,
1190                      rvec                        * gmx_restrict          xx,
1191                      rvec                        * gmx_restrict          ff,
1192                      struct t_forcerec           * gmx_restrict          fr,
1193                      t_mdatoms                   * gmx_restrict     mdatoms,
1194                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1195                      t_nrnb                      * gmx_restrict        nrnb)
1196 {
1197     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1198      * just 0 for non-waters.
1199      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
1200      * jnr indices corresponding to data put in the four positions in the SIMD register.
1201      */
1202     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1203     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1204     int              jnrA,jnrB;
1205     int              j_coord_offsetA,j_coord_offsetB;
1206     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1207     real             rcutoff_scalar;
1208     real             *shiftvec,*fshift,*x,*f;
1209     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1210     int              vdwioffset0;
1211     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1212     int              vdwioffset1;
1213     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1214     int              vdwioffset2;
1215     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1216     int              vdwjidx0A,vdwjidx0B;
1217     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1218     int              vdwjidx1A,vdwjidx1B;
1219     __m128d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1220     int              vdwjidx2A,vdwjidx2B;
1221     __m128d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1222     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1223     __m128d          dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1224     __m128d          dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1225     __m128d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1226     __m128d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1227     __m128d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1228     __m128d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1229     __m128d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1230     __m128d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1231     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
1232     real             *charge;
1233     __m128i          vfitab;
1234     __m128i          ifour       = _mm_set1_epi32(4);
1235     __m128d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
1236     real             *vftab;
1237     __m128d          dummy_mask,cutoff_mask;
1238     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
1239     __m128d          one     = _mm_set1_pd(1.0);
1240     __m128d          two     = _mm_set1_pd(2.0);
1241     x                = xx[0];
1242     f                = ff[0];
1243
1244     nri              = nlist->nri;
1245     iinr             = nlist->iinr;
1246     jindex           = nlist->jindex;
1247     jjnr             = nlist->jjnr;
1248     shiftidx         = nlist->shift;
1249     gid              = nlist->gid;
1250     shiftvec         = fr->shift_vec[0];
1251     fshift           = fr->fshift[0];
1252     facel            = _mm_set1_pd(fr->ic->epsfac);
1253     charge           = mdatoms->chargeA;
1254
1255     vftab            = kernel_data->table_elec->data;
1256     vftabscale       = _mm_set1_pd(kernel_data->table_elec->scale);
1257
1258     /* Setup water-specific parameters */
1259     inr              = nlist->iinr[0];
1260     iq0              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
1261     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
1262     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
1263
1264     jq0              = _mm_set1_pd(charge[inr+0]);
1265     jq1              = _mm_set1_pd(charge[inr+1]);
1266     jq2              = _mm_set1_pd(charge[inr+2]);
1267     qq00             = _mm_mul_pd(iq0,jq0);
1268     qq01             = _mm_mul_pd(iq0,jq1);
1269     qq02             = _mm_mul_pd(iq0,jq2);
1270     qq10             = _mm_mul_pd(iq1,jq0);
1271     qq11             = _mm_mul_pd(iq1,jq1);
1272     qq12             = _mm_mul_pd(iq1,jq2);
1273     qq20             = _mm_mul_pd(iq2,jq0);
1274     qq21             = _mm_mul_pd(iq2,jq1);
1275     qq22             = _mm_mul_pd(iq2,jq2);
1276
1277     /* Avoid stupid compiler warnings */
1278     jnrA = jnrB = 0;
1279     j_coord_offsetA = 0;
1280     j_coord_offsetB = 0;
1281
1282     outeriter        = 0;
1283     inneriter        = 0;
1284
1285     /* Start outer loop over neighborlists */
1286     for(iidx=0; iidx<nri; iidx++)
1287     {
1288         /* Load shift vector for this list */
1289         i_shift_offset   = DIM*shiftidx[iidx];
1290
1291         /* Load limits for loop over neighbors */
1292         j_index_start    = jindex[iidx];
1293         j_index_end      = jindex[iidx+1];
1294
1295         /* Get outer coordinate index */
1296         inr              = iinr[iidx];
1297         i_coord_offset   = DIM*inr;
1298
1299         /* Load i particle coords and add shift vector */
1300         gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1301                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1302
1303         fix0             = _mm_setzero_pd();
1304         fiy0             = _mm_setzero_pd();
1305         fiz0             = _mm_setzero_pd();
1306         fix1             = _mm_setzero_pd();
1307         fiy1             = _mm_setzero_pd();
1308         fiz1             = _mm_setzero_pd();
1309         fix2             = _mm_setzero_pd();
1310         fiy2             = _mm_setzero_pd();
1311         fiz2             = _mm_setzero_pd();
1312
1313         /* Start inner kernel loop */
1314         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
1315         {
1316
1317             /* Get j neighbor index, and coordinate index */
1318             jnrA             = jjnr[jidx];
1319             jnrB             = jjnr[jidx+1];
1320             j_coord_offsetA  = DIM*jnrA;
1321             j_coord_offsetB  = DIM*jnrB;
1322
1323             /* load j atom coordinates */
1324             gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1325                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1326
1327             /* Calculate displacement vector */
1328             dx00             = _mm_sub_pd(ix0,jx0);
1329             dy00             = _mm_sub_pd(iy0,jy0);
1330             dz00             = _mm_sub_pd(iz0,jz0);
1331             dx01             = _mm_sub_pd(ix0,jx1);
1332             dy01             = _mm_sub_pd(iy0,jy1);
1333             dz01             = _mm_sub_pd(iz0,jz1);
1334             dx02             = _mm_sub_pd(ix0,jx2);
1335             dy02             = _mm_sub_pd(iy0,jy2);
1336             dz02             = _mm_sub_pd(iz0,jz2);
1337             dx10             = _mm_sub_pd(ix1,jx0);
1338             dy10             = _mm_sub_pd(iy1,jy0);
1339             dz10             = _mm_sub_pd(iz1,jz0);
1340             dx11             = _mm_sub_pd(ix1,jx1);
1341             dy11             = _mm_sub_pd(iy1,jy1);
1342             dz11             = _mm_sub_pd(iz1,jz1);
1343             dx12             = _mm_sub_pd(ix1,jx2);
1344             dy12             = _mm_sub_pd(iy1,jy2);
1345             dz12             = _mm_sub_pd(iz1,jz2);
1346             dx20             = _mm_sub_pd(ix2,jx0);
1347             dy20             = _mm_sub_pd(iy2,jy0);
1348             dz20             = _mm_sub_pd(iz2,jz0);
1349             dx21             = _mm_sub_pd(ix2,jx1);
1350             dy21             = _mm_sub_pd(iy2,jy1);
1351             dz21             = _mm_sub_pd(iz2,jz1);
1352             dx22             = _mm_sub_pd(ix2,jx2);
1353             dy22             = _mm_sub_pd(iy2,jy2);
1354             dz22             = _mm_sub_pd(iz2,jz2);
1355
1356             /* Calculate squared distance and things based on it */
1357             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1358             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
1359             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
1360             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
1361             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1362             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1363             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
1364             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1365             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1366
1367             rinv00           = avx128fma_invsqrt_d(rsq00);
1368             rinv01           = avx128fma_invsqrt_d(rsq01);
1369             rinv02           = avx128fma_invsqrt_d(rsq02);
1370             rinv10           = avx128fma_invsqrt_d(rsq10);
1371             rinv11           = avx128fma_invsqrt_d(rsq11);
1372             rinv12           = avx128fma_invsqrt_d(rsq12);
1373             rinv20           = avx128fma_invsqrt_d(rsq20);
1374             rinv21           = avx128fma_invsqrt_d(rsq21);
1375             rinv22           = avx128fma_invsqrt_d(rsq22);
1376
1377             fjx0             = _mm_setzero_pd();
1378             fjy0             = _mm_setzero_pd();
1379             fjz0             = _mm_setzero_pd();
1380             fjx1             = _mm_setzero_pd();
1381             fjy1             = _mm_setzero_pd();
1382             fjz1             = _mm_setzero_pd();
1383             fjx2             = _mm_setzero_pd();
1384             fjy2             = _mm_setzero_pd();
1385             fjz2             = _mm_setzero_pd();
1386
1387             /**************************
1388              * CALCULATE INTERACTIONS *
1389              **************************/
1390
1391             r00              = _mm_mul_pd(rsq00,rinv00);
1392
1393             /* Calculate table index by multiplying r with table scale and truncate to integer */
1394             rt               = _mm_mul_pd(r00,vftabscale);
1395             vfitab           = _mm_cvttpd_epi32(rt);
1396 #ifdef __XOP__
1397             vfeps            = _mm_frcz_pd(rt);
1398 #else
1399             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1400 #endif
1401             twovfeps         = _mm_add_pd(vfeps,vfeps);
1402             vfitab           = _mm_slli_epi32(vfitab,2);
1403
1404             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1405             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1406             F                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1407             GMX_MM_TRANSPOSE2_PD(Y,F);
1408             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1409             H                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1410             GMX_MM_TRANSPOSE2_PD(G,H);
1411             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1412             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1413             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq00,FF),_mm_mul_pd(vftabscale,rinv00)));
1414
1415             fscal            = felec;
1416
1417             /* Update vectorial force */
1418             fix0             = _mm_macc_pd(dx00,fscal,fix0);
1419             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
1420             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
1421
1422             fjx0             = _mm_macc_pd(dx00,fscal,fjx0);
1423             fjy0             = _mm_macc_pd(dy00,fscal,fjy0);
1424             fjz0             = _mm_macc_pd(dz00,fscal,fjz0);
1425
1426             /**************************
1427              * CALCULATE INTERACTIONS *
1428              **************************/
1429
1430             r01              = _mm_mul_pd(rsq01,rinv01);
1431
1432             /* Calculate table index by multiplying r with table scale and truncate to integer */
1433             rt               = _mm_mul_pd(r01,vftabscale);
1434             vfitab           = _mm_cvttpd_epi32(rt);
1435 #ifdef __XOP__
1436             vfeps            = _mm_frcz_pd(rt);
1437 #else
1438             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1439 #endif
1440             twovfeps         = _mm_add_pd(vfeps,vfeps);
1441             vfitab           = _mm_slli_epi32(vfitab,2);
1442
1443             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1444             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1445             F                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1446             GMX_MM_TRANSPOSE2_PD(Y,F);
1447             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1448             H                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1449             GMX_MM_TRANSPOSE2_PD(G,H);
1450             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1451             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1452             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq01,FF),_mm_mul_pd(vftabscale,rinv01)));
1453
1454             fscal            = felec;
1455
1456             /* Update vectorial force */
1457             fix0             = _mm_macc_pd(dx01,fscal,fix0);
1458             fiy0             = _mm_macc_pd(dy01,fscal,fiy0);
1459             fiz0             = _mm_macc_pd(dz01,fscal,fiz0);
1460
1461             fjx1             = _mm_macc_pd(dx01,fscal,fjx1);
1462             fjy1             = _mm_macc_pd(dy01,fscal,fjy1);
1463             fjz1             = _mm_macc_pd(dz01,fscal,fjz1);
1464
1465             /**************************
1466              * CALCULATE INTERACTIONS *
1467              **************************/
1468
1469             r02              = _mm_mul_pd(rsq02,rinv02);
1470
1471             /* Calculate table index by multiplying r with table scale and truncate to integer */
1472             rt               = _mm_mul_pd(r02,vftabscale);
1473             vfitab           = _mm_cvttpd_epi32(rt);
1474 #ifdef __XOP__
1475             vfeps            = _mm_frcz_pd(rt);
1476 #else
1477             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1478 #endif
1479             twovfeps         = _mm_add_pd(vfeps,vfeps);
1480             vfitab           = _mm_slli_epi32(vfitab,2);
1481
1482             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1483             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1484             F                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1485             GMX_MM_TRANSPOSE2_PD(Y,F);
1486             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1487             H                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1488             GMX_MM_TRANSPOSE2_PD(G,H);
1489             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1490             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1491             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq02,FF),_mm_mul_pd(vftabscale,rinv02)));
1492
1493             fscal            = felec;
1494
1495             /* Update vectorial force */
1496             fix0             = _mm_macc_pd(dx02,fscal,fix0);
1497             fiy0             = _mm_macc_pd(dy02,fscal,fiy0);
1498             fiz0             = _mm_macc_pd(dz02,fscal,fiz0);
1499
1500             fjx2             = _mm_macc_pd(dx02,fscal,fjx2);
1501             fjy2             = _mm_macc_pd(dy02,fscal,fjy2);
1502             fjz2             = _mm_macc_pd(dz02,fscal,fjz2);
1503
1504             /**************************
1505              * CALCULATE INTERACTIONS *
1506              **************************/
1507
1508             r10              = _mm_mul_pd(rsq10,rinv10);
1509
1510             /* Calculate table index by multiplying r with table scale and truncate to integer */
1511             rt               = _mm_mul_pd(r10,vftabscale);
1512             vfitab           = _mm_cvttpd_epi32(rt);
1513 #ifdef __XOP__
1514             vfeps            = _mm_frcz_pd(rt);
1515 #else
1516             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1517 #endif
1518             twovfeps         = _mm_add_pd(vfeps,vfeps);
1519             vfitab           = _mm_slli_epi32(vfitab,2);
1520
1521             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1522             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1523             F                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1524             GMX_MM_TRANSPOSE2_PD(Y,F);
1525             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1526             H                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1527             GMX_MM_TRANSPOSE2_PD(G,H);
1528             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1529             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1530             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq10,FF),_mm_mul_pd(vftabscale,rinv10)));
1531
1532             fscal            = felec;
1533
1534             /* Update vectorial force */
1535             fix1             = _mm_macc_pd(dx10,fscal,fix1);
1536             fiy1             = _mm_macc_pd(dy10,fscal,fiy1);
1537             fiz1             = _mm_macc_pd(dz10,fscal,fiz1);
1538
1539             fjx0             = _mm_macc_pd(dx10,fscal,fjx0);
1540             fjy0             = _mm_macc_pd(dy10,fscal,fjy0);
1541             fjz0             = _mm_macc_pd(dz10,fscal,fjz0);
1542
1543             /**************************
1544              * CALCULATE INTERACTIONS *
1545              **************************/
1546
1547             r11              = _mm_mul_pd(rsq11,rinv11);
1548
1549             /* Calculate table index by multiplying r with table scale and truncate to integer */
1550             rt               = _mm_mul_pd(r11,vftabscale);
1551             vfitab           = _mm_cvttpd_epi32(rt);
1552 #ifdef __XOP__
1553             vfeps            = _mm_frcz_pd(rt);
1554 #else
1555             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1556 #endif
1557             twovfeps         = _mm_add_pd(vfeps,vfeps);
1558             vfitab           = _mm_slli_epi32(vfitab,2);
1559
1560             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1561             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1562             F                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1563             GMX_MM_TRANSPOSE2_PD(Y,F);
1564             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1565             H                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1566             GMX_MM_TRANSPOSE2_PD(G,H);
1567             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1568             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1569             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
1570
1571             fscal            = felec;
1572
1573             /* Update vectorial force */
1574             fix1             = _mm_macc_pd(dx11,fscal,fix1);
1575             fiy1             = _mm_macc_pd(dy11,fscal,fiy1);
1576             fiz1             = _mm_macc_pd(dz11,fscal,fiz1);
1577
1578             fjx1             = _mm_macc_pd(dx11,fscal,fjx1);
1579             fjy1             = _mm_macc_pd(dy11,fscal,fjy1);
1580             fjz1             = _mm_macc_pd(dz11,fscal,fjz1);
1581
1582             /**************************
1583              * CALCULATE INTERACTIONS *
1584              **************************/
1585
1586             r12              = _mm_mul_pd(rsq12,rinv12);
1587
1588             /* Calculate table index by multiplying r with table scale and truncate to integer */
1589             rt               = _mm_mul_pd(r12,vftabscale);
1590             vfitab           = _mm_cvttpd_epi32(rt);
1591 #ifdef __XOP__
1592             vfeps            = _mm_frcz_pd(rt);
1593 #else
1594             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1595 #endif
1596             twovfeps         = _mm_add_pd(vfeps,vfeps);
1597             vfitab           = _mm_slli_epi32(vfitab,2);
1598
1599             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1600             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1601             F                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1602             GMX_MM_TRANSPOSE2_PD(Y,F);
1603             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1604             H                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1605             GMX_MM_TRANSPOSE2_PD(G,H);
1606             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1607             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1608             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
1609
1610             fscal            = felec;
1611
1612             /* Update vectorial force */
1613             fix1             = _mm_macc_pd(dx12,fscal,fix1);
1614             fiy1             = _mm_macc_pd(dy12,fscal,fiy1);
1615             fiz1             = _mm_macc_pd(dz12,fscal,fiz1);
1616
1617             fjx2             = _mm_macc_pd(dx12,fscal,fjx2);
1618             fjy2             = _mm_macc_pd(dy12,fscal,fjy2);
1619             fjz2             = _mm_macc_pd(dz12,fscal,fjz2);
1620
1621             /**************************
1622              * CALCULATE INTERACTIONS *
1623              **************************/
1624
1625             r20              = _mm_mul_pd(rsq20,rinv20);
1626
1627             /* Calculate table index by multiplying r with table scale and truncate to integer */
1628             rt               = _mm_mul_pd(r20,vftabscale);
1629             vfitab           = _mm_cvttpd_epi32(rt);
1630 #ifdef __XOP__
1631             vfeps            = _mm_frcz_pd(rt);
1632 #else
1633             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1634 #endif
1635             twovfeps         = _mm_add_pd(vfeps,vfeps);
1636             vfitab           = _mm_slli_epi32(vfitab,2);
1637
1638             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1639             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1640             F                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1641             GMX_MM_TRANSPOSE2_PD(Y,F);
1642             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1643             H                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1644             GMX_MM_TRANSPOSE2_PD(G,H);
1645             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1646             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1647             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq20,FF),_mm_mul_pd(vftabscale,rinv20)));
1648
1649             fscal            = felec;
1650
1651             /* Update vectorial force */
1652             fix2             = _mm_macc_pd(dx20,fscal,fix2);
1653             fiy2             = _mm_macc_pd(dy20,fscal,fiy2);
1654             fiz2             = _mm_macc_pd(dz20,fscal,fiz2);
1655
1656             fjx0             = _mm_macc_pd(dx20,fscal,fjx0);
1657             fjy0             = _mm_macc_pd(dy20,fscal,fjy0);
1658             fjz0             = _mm_macc_pd(dz20,fscal,fjz0);
1659
1660             /**************************
1661              * CALCULATE INTERACTIONS *
1662              **************************/
1663
1664             r21              = _mm_mul_pd(rsq21,rinv21);
1665
1666             /* Calculate table index by multiplying r with table scale and truncate to integer */
1667             rt               = _mm_mul_pd(r21,vftabscale);
1668             vfitab           = _mm_cvttpd_epi32(rt);
1669 #ifdef __XOP__
1670             vfeps            = _mm_frcz_pd(rt);
1671 #else
1672             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1673 #endif
1674             twovfeps         = _mm_add_pd(vfeps,vfeps);
1675             vfitab           = _mm_slli_epi32(vfitab,2);
1676
1677             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1678             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1679             F                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1680             GMX_MM_TRANSPOSE2_PD(Y,F);
1681             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1682             H                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1683             GMX_MM_TRANSPOSE2_PD(G,H);
1684             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1685             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1686             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
1687
1688             fscal            = felec;
1689
1690             /* Update vectorial force */
1691             fix2             = _mm_macc_pd(dx21,fscal,fix2);
1692             fiy2             = _mm_macc_pd(dy21,fscal,fiy2);
1693             fiz2             = _mm_macc_pd(dz21,fscal,fiz2);
1694
1695             fjx1             = _mm_macc_pd(dx21,fscal,fjx1);
1696             fjy1             = _mm_macc_pd(dy21,fscal,fjy1);
1697             fjz1             = _mm_macc_pd(dz21,fscal,fjz1);
1698
1699             /**************************
1700              * CALCULATE INTERACTIONS *
1701              **************************/
1702
1703             r22              = _mm_mul_pd(rsq22,rinv22);
1704
1705             /* Calculate table index by multiplying r with table scale and truncate to integer */
1706             rt               = _mm_mul_pd(r22,vftabscale);
1707             vfitab           = _mm_cvttpd_epi32(rt);
1708 #ifdef __XOP__
1709             vfeps            = _mm_frcz_pd(rt);
1710 #else
1711             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1712 #endif
1713             twovfeps         = _mm_add_pd(vfeps,vfeps);
1714             vfitab           = _mm_slli_epi32(vfitab,2);
1715
1716             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1717             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1718             F                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1719             GMX_MM_TRANSPOSE2_PD(Y,F);
1720             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1721             H                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1722             GMX_MM_TRANSPOSE2_PD(G,H);
1723             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1724             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1725             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
1726
1727             fscal            = felec;
1728
1729             /* Update vectorial force */
1730             fix2             = _mm_macc_pd(dx22,fscal,fix2);
1731             fiy2             = _mm_macc_pd(dy22,fscal,fiy2);
1732             fiz2             = _mm_macc_pd(dz22,fscal,fiz2);
1733
1734             fjx2             = _mm_macc_pd(dx22,fscal,fjx2);
1735             fjy2             = _mm_macc_pd(dy22,fscal,fjy2);
1736             fjz2             = _mm_macc_pd(dz22,fscal,fjz2);
1737
1738             gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1739
1740             /* Inner loop uses 378 flops */
1741         }
1742
1743         if(jidx<j_index_end)
1744         {
1745
1746             jnrA             = jjnr[jidx];
1747             j_coord_offsetA  = DIM*jnrA;
1748
1749             /* load j atom coordinates */
1750             gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
1751                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1752
1753             /* Calculate displacement vector */
1754             dx00             = _mm_sub_pd(ix0,jx0);
1755             dy00             = _mm_sub_pd(iy0,jy0);
1756             dz00             = _mm_sub_pd(iz0,jz0);
1757             dx01             = _mm_sub_pd(ix0,jx1);
1758             dy01             = _mm_sub_pd(iy0,jy1);
1759             dz01             = _mm_sub_pd(iz0,jz1);
1760             dx02             = _mm_sub_pd(ix0,jx2);
1761             dy02             = _mm_sub_pd(iy0,jy2);
1762             dz02             = _mm_sub_pd(iz0,jz2);
1763             dx10             = _mm_sub_pd(ix1,jx0);
1764             dy10             = _mm_sub_pd(iy1,jy0);
1765             dz10             = _mm_sub_pd(iz1,jz0);
1766             dx11             = _mm_sub_pd(ix1,jx1);
1767             dy11             = _mm_sub_pd(iy1,jy1);
1768             dz11             = _mm_sub_pd(iz1,jz1);
1769             dx12             = _mm_sub_pd(ix1,jx2);
1770             dy12             = _mm_sub_pd(iy1,jy2);
1771             dz12             = _mm_sub_pd(iz1,jz2);
1772             dx20             = _mm_sub_pd(ix2,jx0);
1773             dy20             = _mm_sub_pd(iy2,jy0);
1774             dz20             = _mm_sub_pd(iz2,jz0);
1775             dx21             = _mm_sub_pd(ix2,jx1);
1776             dy21             = _mm_sub_pd(iy2,jy1);
1777             dz21             = _mm_sub_pd(iz2,jz1);
1778             dx22             = _mm_sub_pd(ix2,jx2);
1779             dy22             = _mm_sub_pd(iy2,jy2);
1780             dz22             = _mm_sub_pd(iz2,jz2);
1781
1782             /* Calculate squared distance and things based on it */
1783             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1784             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
1785             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
1786             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
1787             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1788             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1789             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
1790             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1791             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1792
1793             rinv00           = avx128fma_invsqrt_d(rsq00);
1794             rinv01           = avx128fma_invsqrt_d(rsq01);
1795             rinv02           = avx128fma_invsqrt_d(rsq02);
1796             rinv10           = avx128fma_invsqrt_d(rsq10);
1797             rinv11           = avx128fma_invsqrt_d(rsq11);
1798             rinv12           = avx128fma_invsqrt_d(rsq12);
1799             rinv20           = avx128fma_invsqrt_d(rsq20);
1800             rinv21           = avx128fma_invsqrt_d(rsq21);
1801             rinv22           = avx128fma_invsqrt_d(rsq22);
1802
1803             fjx0             = _mm_setzero_pd();
1804             fjy0             = _mm_setzero_pd();
1805             fjz0             = _mm_setzero_pd();
1806             fjx1             = _mm_setzero_pd();
1807             fjy1             = _mm_setzero_pd();
1808             fjz1             = _mm_setzero_pd();
1809             fjx2             = _mm_setzero_pd();
1810             fjy2             = _mm_setzero_pd();
1811             fjz2             = _mm_setzero_pd();
1812
1813             /**************************
1814              * CALCULATE INTERACTIONS *
1815              **************************/
1816
1817             r00              = _mm_mul_pd(rsq00,rinv00);
1818
1819             /* Calculate table index by multiplying r with table scale and truncate to integer */
1820             rt               = _mm_mul_pd(r00,vftabscale);
1821             vfitab           = _mm_cvttpd_epi32(rt);
1822 #ifdef __XOP__
1823             vfeps            = _mm_frcz_pd(rt);
1824 #else
1825             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1826 #endif
1827             twovfeps         = _mm_add_pd(vfeps,vfeps);
1828             vfitab           = _mm_slli_epi32(vfitab,2);
1829
1830             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1831             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1832             F                = _mm_setzero_pd();
1833             GMX_MM_TRANSPOSE2_PD(Y,F);
1834             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1835             H                = _mm_setzero_pd();
1836             GMX_MM_TRANSPOSE2_PD(G,H);
1837             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1838             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1839             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq00,FF),_mm_mul_pd(vftabscale,rinv00)));
1840
1841             fscal            = felec;
1842
1843             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1844
1845             /* Update vectorial force */
1846             fix0             = _mm_macc_pd(dx00,fscal,fix0);
1847             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
1848             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
1849
1850             fjx0             = _mm_macc_pd(dx00,fscal,fjx0);
1851             fjy0             = _mm_macc_pd(dy00,fscal,fjy0);
1852             fjz0             = _mm_macc_pd(dz00,fscal,fjz0);
1853
1854             /**************************
1855              * CALCULATE INTERACTIONS *
1856              **************************/
1857
1858             r01              = _mm_mul_pd(rsq01,rinv01);
1859
1860             /* Calculate table index by multiplying r with table scale and truncate to integer */
1861             rt               = _mm_mul_pd(r01,vftabscale);
1862             vfitab           = _mm_cvttpd_epi32(rt);
1863 #ifdef __XOP__
1864             vfeps            = _mm_frcz_pd(rt);
1865 #else
1866             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1867 #endif
1868             twovfeps         = _mm_add_pd(vfeps,vfeps);
1869             vfitab           = _mm_slli_epi32(vfitab,2);
1870
1871             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1872             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1873             F                = _mm_setzero_pd();
1874             GMX_MM_TRANSPOSE2_PD(Y,F);
1875             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1876             H                = _mm_setzero_pd();
1877             GMX_MM_TRANSPOSE2_PD(G,H);
1878             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1879             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1880             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq01,FF),_mm_mul_pd(vftabscale,rinv01)));
1881
1882             fscal            = felec;
1883
1884             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1885
1886             /* Update vectorial force */
1887             fix0             = _mm_macc_pd(dx01,fscal,fix0);
1888             fiy0             = _mm_macc_pd(dy01,fscal,fiy0);
1889             fiz0             = _mm_macc_pd(dz01,fscal,fiz0);
1890
1891             fjx1             = _mm_macc_pd(dx01,fscal,fjx1);
1892             fjy1             = _mm_macc_pd(dy01,fscal,fjy1);
1893             fjz1             = _mm_macc_pd(dz01,fscal,fjz1);
1894
1895             /**************************
1896              * CALCULATE INTERACTIONS *
1897              **************************/
1898
1899             r02              = _mm_mul_pd(rsq02,rinv02);
1900
1901             /* Calculate table index by multiplying r with table scale and truncate to integer */
1902             rt               = _mm_mul_pd(r02,vftabscale);
1903             vfitab           = _mm_cvttpd_epi32(rt);
1904 #ifdef __XOP__
1905             vfeps            = _mm_frcz_pd(rt);
1906 #else
1907             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1908 #endif
1909             twovfeps         = _mm_add_pd(vfeps,vfeps);
1910             vfitab           = _mm_slli_epi32(vfitab,2);
1911
1912             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1913             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1914             F                = _mm_setzero_pd();
1915             GMX_MM_TRANSPOSE2_PD(Y,F);
1916             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1917             H                = _mm_setzero_pd();
1918             GMX_MM_TRANSPOSE2_PD(G,H);
1919             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1920             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1921             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq02,FF),_mm_mul_pd(vftabscale,rinv02)));
1922
1923             fscal            = felec;
1924
1925             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1926
1927             /* Update vectorial force */
1928             fix0             = _mm_macc_pd(dx02,fscal,fix0);
1929             fiy0             = _mm_macc_pd(dy02,fscal,fiy0);
1930             fiz0             = _mm_macc_pd(dz02,fscal,fiz0);
1931
1932             fjx2             = _mm_macc_pd(dx02,fscal,fjx2);
1933             fjy2             = _mm_macc_pd(dy02,fscal,fjy2);
1934             fjz2             = _mm_macc_pd(dz02,fscal,fjz2);
1935
1936             /**************************
1937              * CALCULATE INTERACTIONS *
1938              **************************/
1939
1940             r10              = _mm_mul_pd(rsq10,rinv10);
1941
1942             /* Calculate table index by multiplying r with table scale and truncate to integer */
1943             rt               = _mm_mul_pd(r10,vftabscale);
1944             vfitab           = _mm_cvttpd_epi32(rt);
1945 #ifdef __XOP__
1946             vfeps            = _mm_frcz_pd(rt);
1947 #else
1948             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1949 #endif
1950             twovfeps         = _mm_add_pd(vfeps,vfeps);
1951             vfitab           = _mm_slli_epi32(vfitab,2);
1952
1953             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1954             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1955             F                = _mm_setzero_pd();
1956             GMX_MM_TRANSPOSE2_PD(Y,F);
1957             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1958             H                = _mm_setzero_pd();
1959             GMX_MM_TRANSPOSE2_PD(G,H);
1960             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1961             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1962             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq10,FF),_mm_mul_pd(vftabscale,rinv10)));
1963
1964             fscal            = felec;
1965
1966             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1967
1968             /* Update vectorial force */
1969             fix1             = _mm_macc_pd(dx10,fscal,fix1);
1970             fiy1             = _mm_macc_pd(dy10,fscal,fiy1);
1971             fiz1             = _mm_macc_pd(dz10,fscal,fiz1);
1972
1973             fjx0             = _mm_macc_pd(dx10,fscal,fjx0);
1974             fjy0             = _mm_macc_pd(dy10,fscal,fjy0);
1975             fjz0             = _mm_macc_pd(dz10,fscal,fjz0);
1976
1977             /**************************
1978              * CALCULATE INTERACTIONS *
1979              **************************/
1980
1981             r11              = _mm_mul_pd(rsq11,rinv11);
1982
1983             /* Calculate table index by multiplying r with table scale and truncate to integer */
1984             rt               = _mm_mul_pd(r11,vftabscale);
1985             vfitab           = _mm_cvttpd_epi32(rt);
1986 #ifdef __XOP__
1987             vfeps            = _mm_frcz_pd(rt);
1988 #else
1989             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1990 #endif
1991             twovfeps         = _mm_add_pd(vfeps,vfeps);
1992             vfitab           = _mm_slli_epi32(vfitab,2);
1993
1994             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1995             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1996             F                = _mm_setzero_pd();
1997             GMX_MM_TRANSPOSE2_PD(Y,F);
1998             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1999             H                = _mm_setzero_pd();
2000             GMX_MM_TRANSPOSE2_PD(G,H);
2001             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
2002             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
2003             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
2004
2005             fscal            = felec;
2006
2007             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2008
2009             /* Update vectorial force */
2010             fix1             = _mm_macc_pd(dx11,fscal,fix1);
2011             fiy1             = _mm_macc_pd(dy11,fscal,fiy1);
2012             fiz1             = _mm_macc_pd(dz11,fscal,fiz1);
2013
2014             fjx1             = _mm_macc_pd(dx11,fscal,fjx1);
2015             fjy1             = _mm_macc_pd(dy11,fscal,fjy1);
2016             fjz1             = _mm_macc_pd(dz11,fscal,fjz1);
2017
2018             /**************************
2019              * CALCULATE INTERACTIONS *
2020              **************************/
2021
2022             r12              = _mm_mul_pd(rsq12,rinv12);
2023
2024             /* Calculate table index by multiplying r with table scale and truncate to integer */
2025             rt               = _mm_mul_pd(r12,vftabscale);
2026             vfitab           = _mm_cvttpd_epi32(rt);
2027 #ifdef __XOP__
2028             vfeps            = _mm_frcz_pd(rt);
2029 #else
2030             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2031 #endif
2032             twovfeps         = _mm_add_pd(vfeps,vfeps);
2033             vfitab           = _mm_slli_epi32(vfitab,2);
2034
2035             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2036             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2037             F                = _mm_setzero_pd();
2038             GMX_MM_TRANSPOSE2_PD(Y,F);
2039             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
2040             H                = _mm_setzero_pd();
2041             GMX_MM_TRANSPOSE2_PD(G,H);
2042             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
2043             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
2044             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
2045
2046             fscal            = felec;
2047
2048             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2049
2050             /* Update vectorial force */
2051             fix1             = _mm_macc_pd(dx12,fscal,fix1);
2052             fiy1             = _mm_macc_pd(dy12,fscal,fiy1);
2053             fiz1             = _mm_macc_pd(dz12,fscal,fiz1);
2054
2055             fjx2             = _mm_macc_pd(dx12,fscal,fjx2);
2056             fjy2             = _mm_macc_pd(dy12,fscal,fjy2);
2057             fjz2             = _mm_macc_pd(dz12,fscal,fjz2);
2058
2059             /**************************
2060              * CALCULATE INTERACTIONS *
2061              **************************/
2062
2063             r20              = _mm_mul_pd(rsq20,rinv20);
2064
2065             /* Calculate table index by multiplying r with table scale and truncate to integer */
2066             rt               = _mm_mul_pd(r20,vftabscale);
2067             vfitab           = _mm_cvttpd_epi32(rt);
2068 #ifdef __XOP__
2069             vfeps            = _mm_frcz_pd(rt);
2070 #else
2071             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2072 #endif
2073             twovfeps         = _mm_add_pd(vfeps,vfeps);
2074             vfitab           = _mm_slli_epi32(vfitab,2);
2075
2076             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2077             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2078             F                = _mm_setzero_pd();
2079             GMX_MM_TRANSPOSE2_PD(Y,F);
2080             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
2081             H                = _mm_setzero_pd();
2082             GMX_MM_TRANSPOSE2_PD(G,H);
2083             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
2084             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
2085             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq20,FF),_mm_mul_pd(vftabscale,rinv20)));
2086
2087             fscal            = felec;
2088
2089             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2090
2091             /* Update vectorial force */
2092             fix2             = _mm_macc_pd(dx20,fscal,fix2);
2093             fiy2             = _mm_macc_pd(dy20,fscal,fiy2);
2094             fiz2             = _mm_macc_pd(dz20,fscal,fiz2);
2095
2096             fjx0             = _mm_macc_pd(dx20,fscal,fjx0);
2097             fjy0             = _mm_macc_pd(dy20,fscal,fjy0);
2098             fjz0             = _mm_macc_pd(dz20,fscal,fjz0);
2099
2100             /**************************
2101              * CALCULATE INTERACTIONS *
2102              **************************/
2103
2104             r21              = _mm_mul_pd(rsq21,rinv21);
2105
2106             /* Calculate table index by multiplying r with table scale and truncate to integer */
2107             rt               = _mm_mul_pd(r21,vftabscale);
2108             vfitab           = _mm_cvttpd_epi32(rt);
2109 #ifdef __XOP__
2110             vfeps            = _mm_frcz_pd(rt);
2111 #else
2112             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2113 #endif
2114             twovfeps         = _mm_add_pd(vfeps,vfeps);
2115             vfitab           = _mm_slli_epi32(vfitab,2);
2116
2117             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2118             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2119             F                = _mm_setzero_pd();
2120             GMX_MM_TRANSPOSE2_PD(Y,F);
2121             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
2122             H                = _mm_setzero_pd();
2123             GMX_MM_TRANSPOSE2_PD(G,H);
2124             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
2125             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
2126             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
2127
2128             fscal            = felec;
2129
2130             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2131
2132             /* Update vectorial force */
2133             fix2             = _mm_macc_pd(dx21,fscal,fix2);
2134             fiy2             = _mm_macc_pd(dy21,fscal,fiy2);
2135             fiz2             = _mm_macc_pd(dz21,fscal,fiz2);
2136
2137             fjx1             = _mm_macc_pd(dx21,fscal,fjx1);
2138             fjy1             = _mm_macc_pd(dy21,fscal,fjy1);
2139             fjz1             = _mm_macc_pd(dz21,fscal,fjz1);
2140
2141             /**************************
2142              * CALCULATE INTERACTIONS *
2143              **************************/
2144
2145             r22              = _mm_mul_pd(rsq22,rinv22);
2146
2147             /* Calculate table index by multiplying r with table scale and truncate to integer */
2148             rt               = _mm_mul_pd(r22,vftabscale);
2149             vfitab           = _mm_cvttpd_epi32(rt);
2150 #ifdef __XOP__
2151             vfeps            = _mm_frcz_pd(rt);
2152 #else
2153             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2154 #endif
2155             twovfeps         = _mm_add_pd(vfeps,vfeps);
2156             vfitab           = _mm_slli_epi32(vfitab,2);
2157
2158             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2159             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2160             F                = _mm_setzero_pd();
2161             GMX_MM_TRANSPOSE2_PD(Y,F);
2162             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
2163             H                = _mm_setzero_pd();
2164             GMX_MM_TRANSPOSE2_PD(G,H);
2165             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
2166             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
2167             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
2168
2169             fscal            = felec;
2170
2171             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2172
2173             /* Update vectorial force */
2174             fix2             = _mm_macc_pd(dx22,fscal,fix2);
2175             fiy2             = _mm_macc_pd(dy22,fscal,fiy2);
2176             fiz2             = _mm_macc_pd(dz22,fscal,fiz2);
2177
2178             fjx2             = _mm_macc_pd(dx22,fscal,fjx2);
2179             fjy2             = _mm_macc_pd(dy22,fscal,fjy2);
2180             fjz2             = _mm_macc_pd(dz22,fscal,fjz2);
2181
2182             gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
2183
2184             /* Inner loop uses 378 flops */
2185         }
2186
2187         /* End of innermost loop */
2188
2189         gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
2190                                               f+i_coord_offset,fshift+i_shift_offset);
2191
2192         /* Increment number of inner iterations */
2193         inneriter                  += j_index_end - j_index_start;
2194
2195         /* Outer loop uses 18 flops */
2196     }
2197
2198     /* Increment number of outer iterations */
2199     outeriter        += nri;
2200
2201     /* Update outer/inner flops */
2202
2203     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3W3_F,outeriter*18 + inneriter*378);
2204 }