src/gromacs/gmxlib/nonbonded/nb_kernel_avx_128_fma_double/nb_kernel_ElecEw_VdwLJEw_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_ElecEw_VdwLJEw_GeomW3W3_VF_avx_128_fma_double
  51  * Electrostatics interaction: Ewald
  52  * VdW interaction:            LJEwald
  53  * Geometry:                   Water3-Water3
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecEw_VdwLJEw_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     int              nvdwtype;
 103     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 104     int              *vdwtype;
 105     real             *vdwparam;
 106     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
 107     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
 108     __m128d           c6grid_00;
 109     __m128d           c6grid_01;
 110     __m128d           c6grid_02;
 111     __m128d           c6grid_10;
 112     __m128d           c6grid_11;
 113     __m128d           c6grid_12;
 114     __m128d           c6grid_20;
 115     __m128d           c6grid_21;
 116     __m128d           c6grid_22;
 117     real             *vdwgridparam;
 118     __m128d           ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
 119     __m128d           one_half  = _mm_set1_pd(0.5);
 120     __m128d           minus_one = _mm_set1_pd(-1.0);
 121     __m128i          ewitab;
 122     __m128d          ewtabscale,eweps,twoeweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 123     real             *ewtab;
 124     __m128d          dummy_mask,cutoff_mask;
 125     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 126     __m128d          one     = _mm_set1_pd(1.0);
 127     __m128d          two     = _mm_set1_pd(2.0);
 128     x                = xx[0];
 129     f                = ff[0];
 130
 131     nri              = nlist->nri;
 132     iinr             = nlist->iinr;
 133     jindex           = nlist->jindex;
 134     jjnr             = nlist->jjnr;
 135     shiftidx         = nlist->shift;
 136     gid              = nlist->gid;
 137     shiftvec         = fr->shift_vec[0];
 138     fshift           = fr->fshift[0];
 139     facel            = _mm_set1_pd(fr->ic->epsfac);
 140     charge           = mdatoms->chargeA;
 141     nvdwtype         = fr->ntype;
 142     vdwparam         = fr->nbfp;
 143     vdwtype          = mdatoms->typeA;
 144     vdwgridparam     = fr->ljpme_c6grid;
 145     sh_lj_ewald      = _mm_set1_pd(fr->ic->sh_lj_ewald);
 146     ewclj            = _mm_set1_pd(fr->ic->ewaldcoeff_lj);
 147     ewclj2           = _mm_mul_pd(minus_one,_mm_mul_pd(ewclj,ewclj));
 148
 149     sh_ewald         = _mm_set1_pd(fr->ic->sh_ewald);
 150     ewtab            = fr->ic->tabq_coul_FDV0;
 151     ewtabscale       = _mm_set1_pd(fr->ic->tabq_scale);
 152     ewtabhalfspace   = _mm_set1_pd(0.5/fr->ic->tabq_scale);
 153
 154     /* Setup water-specific parameters */
 155     inr              = nlist->iinr[0];
 156     iq0              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
 157     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
 158     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
 159     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 160
 161     jq0              = _mm_set1_pd(charge[inr+0]);
 162     jq1              = _mm_set1_pd(charge[inr+1]);
 163     jq2              = _mm_set1_pd(charge[inr+2]);
 164     vdwjidx0A        = 2*vdwtype[inr+0];
 165     qq00             = _mm_mul_pd(iq0,jq0);
 166     c6_00            = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
 167     c12_00           = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
 168     c6grid_00        = _mm_set1_pd(vdwgridparam[vdwioffset0+vdwjidx0A]);
 169     qq01             = _mm_mul_pd(iq0,jq1);
 170     qq02             = _mm_mul_pd(iq0,jq2);
 171     qq10             = _mm_mul_pd(iq1,jq0);
 172     qq11             = _mm_mul_pd(iq1,jq1);
 173     qq12             = _mm_mul_pd(iq1,jq2);
 174     qq20             = _mm_mul_pd(iq2,jq0);
 175     qq21             = _mm_mul_pd(iq2,jq1);
 176     qq22             = _mm_mul_pd(iq2,jq2);
 177
 178     /* Avoid stupid compiler warnings */
 179     jnrA = jnrB = 0;
 180     j_coord_offsetA = 0;
 181     j_coord_offsetB = 0;
 182
 183     outeriter        = 0;
 184     inneriter        = 0;
 185
 186     /* Start outer loop over neighborlists */
 187     for(iidx=0; iidx<nri; iidx++)
 188     {
 189         /* Load shift vector for this list */
 190         i_shift_offset   = DIM*shiftidx[iidx];
 191
 192         /* Load limits for loop over neighbors */
 193         j_index_start    = jindex[iidx];
 194         j_index_end      = jindex[iidx+1];
 195
 196         /* Get outer coordinate index */
 197         inr              = iinr[iidx];
 198         i_coord_offset   = DIM*inr;
 199
 200         /* Load i particle coords and add shift vector */
 201         gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 202                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 203
 204         fix0             = _mm_setzero_pd();
 205         fiy0             = _mm_setzero_pd();
 206         fiz0             = _mm_setzero_pd();
 207         fix1             = _mm_setzero_pd();
 208         fiy1             = _mm_setzero_pd();
 209         fiz1             = _mm_setzero_pd();
 210         fix2             = _mm_setzero_pd();
 211         fiy2             = _mm_setzero_pd();
 212         fiz2             = _mm_setzero_pd();
 213
 214         /* Reset potential sums */
 215         velecsum         = _mm_setzero_pd();
 216         vvdwsum          = _mm_setzero_pd();
 217
 218         /* Start inner kernel loop */
 219         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 220         {
 221
 222             /* Get j neighbor index, and coordinate index */
 223             jnrA             = jjnr[jidx];
 224             jnrB             = jjnr[jidx+1];
 225             j_coord_offsetA  = DIM*jnrA;
 226             j_coord_offsetB  = DIM*jnrB;
 227
 228             /* load j atom coordinates */
 229             gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 230                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 231
 232             /* Calculate displacement vector */
 233             dx00             = _mm_sub_pd(ix0,jx0);
 234             dy00             = _mm_sub_pd(iy0,jy0);
 235             dz00             = _mm_sub_pd(iz0,jz0);
 236             dx01             = _mm_sub_pd(ix0,jx1);
 237             dy01             = _mm_sub_pd(iy0,jy1);
 238             dz01             = _mm_sub_pd(iz0,jz1);
 239             dx02             = _mm_sub_pd(ix0,jx2);
 240             dy02             = _mm_sub_pd(iy0,jy2);
 241             dz02             = _mm_sub_pd(iz0,jz2);
 242             dx10             = _mm_sub_pd(ix1,jx0);
 243             dy10             = _mm_sub_pd(iy1,jy0);
 244             dz10             = _mm_sub_pd(iz1,jz0);
 245             dx11             = _mm_sub_pd(ix1,jx1);
 246             dy11             = _mm_sub_pd(iy1,jy1);
 247             dz11             = _mm_sub_pd(iz1,jz1);
 248             dx12             = _mm_sub_pd(ix1,jx2);
 249             dy12             = _mm_sub_pd(iy1,jy2);
 250             dz12             = _mm_sub_pd(iz1,jz2);
 251             dx20             = _mm_sub_pd(ix2,jx0);
 252             dy20             = _mm_sub_pd(iy2,jy0);
 253             dz20             = _mm_sub_pd(iz2,jz0);
 254             dx21             = _mm_sub_pd(ix2,jx1);
 255             dy21             = _mm_sub_pd(iy2,jy1);
 256             dz21             = _mm_sub_pd(iz2,jz1);
 257             dx22             = _mm_sub_pd(ix2,jx2);
 258             dy22             = _mm_sub_pd(iy2,jy2);
 259             dz22             = _mm_sub_pd(iz2,jz2);
 260
 261             /* Calculate squared distance and things based on it */
 262             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 263             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
 264             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
 265             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 266             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 267             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 268             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 269             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 270             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 271
 272             rinv00           = avx128fma_invsqrt_d(rsq00);
 273             rinv01           = avx128fma_invsqrt_d(rsq01);
 274             rinv02           = avx128fma_invsqrt_d(rsq02);
 275             rinv10           = avx128fma_invsqrt_d(rsq10);
 276             rinv11           = avx128fma_invsqrt_d(rsq11);
 277             rinv12           = avx128fma_invsqrt_d(rsq12);
 278             rinv20           = avx128fma_invsqrt_d(rsq20);
 279             rinv21           = avx128fma_invsqrt_d(rsq21);
 280             rinv22           = avx128fma_invsqrt_d(rsq22);
 281
 282             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 283             rinvsq01         = _mm_mul_pd(rinv01,rinv01);
 284             rinvsq02         = _mm_mul_pd(rinv02,rinv02);
 285             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
 286             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
 287             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
 288             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
 289             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
 290             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
 291
 292             fjx0             = _mm_setzero_pd();
 293             fjy0             = _mm_setzero_pd();
 294             fjz0             = _mm_setzero_pd();
 295             fjx1             = _mm_setzero_pd();
 296             fjy1             = _mm_setzero_pd();
 297             fjz1             = _mm_setzero_pd();
 298             fjx2             = _mm_setzero_pd();
 299             fjy2             = _mm_setzero_pd();
 300             fjz2             = _mm_setzero_pd();
 301
 302             /**************************
 303              * CALCULATE INTERACTIONS *
 304              **************************/
 305
 306             r00              = _mm_mul_pd(rsq00,rinv00);
 307
 308             /* EWALD ELECTROSTATICS */
 309
 310             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 311             ewrt             = _mm_mul_pd(r00,ewtabscale);
 312             ewitab           = _mm_cvttpd_epi32(ewrt);
 313 #ifdef __XOP__
 314             eweps            = _mm_frcz_pd(ewrt);
 315 #else
 316             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 317 #endif
 318             twoeweps         = _mm_add_pd(eweps,eweps);
 319             ewitab           = _mm_slli_epi32(ewitab,2);
 320             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 321             ewtabD           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 322             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 323             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
 324             ewtabFn          = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
 325             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 326             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
 327             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
 328             velec            = _mm_mul_pd(qq00,_mm_sub_pd(rinv00,velec));
 329             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
 330
 331             /* Analytical LJ-PME */
 332             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 333             ewcljrsq         = _mm_mul_pd(ewclj2,rsq00);
 334             ewclj6           = _mm_mul_pd(ewclj2,_mm_mul_pd(ewclj2,ewclj2));
 335             exponent         = avx128fma_exp_d(ewcljrsq);
 336             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
 337             poly             = _mm_mul_pd(exponent,_mm_macc_pd(_mm_mul_pd(ewcljrsq,ewcljrsq),one_half,_mm_sub_pd(one,ewcljrsq)));
 338             /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
 339             vvdw6            = _mm_mul_pd(_mm_macc_pd(-c6grid_00,_mm_sub_pd(one,poly),c6_00),rinvsix);
 340             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
 341             vvdw             = _mm_msub_pd(vvdw12,one_twelfth,_mm_mul_pd(vvdw6,one_sixth));
 342             /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
 343             fvdw             = _mm_mul_pd(_mm_add_pd(vvdw12,_mm_msub_pd(_mm_mul_pd(c6grid_00,one_sixth),_mm_mul_pd(exponent,ewclj6),vvdw6)),rinvsq00);
 344
 345             /* Update potential sum for this i atom from the interaction with this j atom. */
 346             velecsum         = _mm_add_pd(velecsum,velec);
 347             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 348
 349             fscal            = _mm_add_pd(felec,fvdw);
 350
 351             /* Update vectorial force */
 352             fix0             = _mm_macc_pd(dx00,fscal,fix0);
 353             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
 354             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
 355
 356             fjx0             = _mm_macc_pd(dx00,fscal,fjx0);
 357             fjy0             = _mm_macc_pd(dy00,fscal,fjy0);
 358             fjz0             = _mm_macc_pd(dz00,fscal,fjz0);
 359
 360             /**************************
 361              * CALCULATE INTERACTIONS *
 362              **************************/
 363
 364             r01              = _mm_mul_pd(rsq01,rinv01);
 365
 366             /* EWALD ELECTROSTATICS */
 367
 368             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 369             ewrt             = _mm_mul_pd(r01,ewtabscale);
 370             ewitab           = _mm_cvttpd_epi32(ewrt);
 371 #ifdef __XOP__
 372             eweps            = _mm_frcz_pd(ewrt);
 373 #else
 374             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 375 #endif
 376             twoeweps         = _mm_add_pd(eweps,eweps);
 377             ewitab           = _mm_slli_epi32(ewitab,2);
 378             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 379             ewtabD           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 380             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 381             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
 382             ewtabFn          = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
 383             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 384             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
 385             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
 386             velec            = _mm_mul_pd(qq01,_mm_sub_pd(rinv01,velec));
 387             felec            = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
 388
 389             /* Update potential sum for this i atom from the interaction with this j atom. */
 390             velecsum         = _mm_add_pd(velecsum,velec);
 391
 392             fscal            = felec;
 393
 394             /* Update vectorial force */
 395             fix0             = _mm_macc_pd(dx01,fscal,fix0);
 396             fiy0             = _mm_macc_pd(dy01,fscal,fiy0);
 397             fiz0             = _mm_macc_pd(dz01,fscal,fiz0);
 398
 399             fjx1             = _mm_macc_pd(dx01,fscal,fjx1);
 400             fjy1             = _mm_macc_pd(dy01,fscal,fjy1);
 401             fjz1             = _mm_macc_pd(dz01,fscal,fjz1);
 402
 403             /**************************
 404              * CALCULATE INTERACTIONS *
 405              **************************/
 406
 407             r02              = _mm_mul_pd(rsq02,rinv02);
 408
 409             /* EWALD ELECTROSTATICS */
 410
 411             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 412             ewrt             = _mm_mul_pd(r02,ewtabscale);
 413             ewitab           = _mm_cvttpd_epi32(ewrt);
 414 #ifdef __XOP__
 415             eweps            = _mm_frcz_pd(ewrt);
 416 #else
 417             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 418 #endif
 419             twoeweps         = _mm_add_pd(eweps,eweps);
 420             ewitab           = _mm_slli_epi32(ewitab,2);
 421             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 422             ewtabD           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 423             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 424             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
 425             ewtabFn          = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
 426             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 427             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
 428             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
 429             velec            = _mm_mul_pd(qq02,_mm_sub_pd(rinv02,velec));
 430             felec            = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
 431
 432             /* Update potential sum for this i atom from the interaction with this j atom. */
 433             velecsum         = _mm_add_pd(velecsum,velec);
 434
 435             fscal            = felec;
 436
 437             /* Update vectorial force */
 438             fix0             = _mm_macc_pd(dx02,fscal,fix0);
 439             fiy0             = _mm_macc_pd(dy02,fscal,fiy0);
 440             fiz0             = _mm_macc_pd(dz02,fscal,fiz0);
 441
 442             fjx2             = _mm_macc_pd(dx02,fscal,fjx2);
 443             fjy2             = _mm_macc_pd(dy02,fscal,fjy2);
 444             fjz2             = _mm_macc_pd(dz02,fscal,fjz2);
 445
 446             /**************************
 447              * CALCULATE INTERACTIONS *
 448              **************************/
 449
 450             r10              = _mm_mul_pd(rsq10,rinv10);
 451
 452             /* EWALD ELECTROSTATICS */
 453
 454             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 455             ewrt             = _mm_mul_pd(r10,ewtabscale);
 456             ewitab           = _mm_cvttpd_epi32(ewrt);
 457 #ifdef __XOP__
 458             eweps            = _mm_frcz_pd(ewrt);
 459 #else
 460             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 461 #endif
 462             twoeweps         = _mm_add_pd(eweps,eweps);
 463             ewitab           = _mm_slli_epi32(ewitab,2);
 464             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 465             ewtabD           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 466             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 467             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
 468             ewtabFn          = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
 469             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 470             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
 471             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
 472             velec            = _mm_mul_pd(qq10,_mm_sub_pd(rinv10,velec));
 473             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
 474
 475             /* Update potential sum for this i atom from the interaction with this j atom. */
 476             velecsum         = _mm_add_pd(velecsum,velec);
 477
 478             fscal            = felec;
 479
 480             /* Update vectorial force */
 481             fix1             = _mm_macc_pd(dx10,fscal,fix1);
 482             fiy1             = _mm_macc_pd(dy10,fscal,fiy1);
 483             fiz1             = _mm_macc_pd(dz10,fscal,fiz1);
 484
 485             fjx0             = _mm_macc_pd(dx10,fscal,fjx0);
 486             fjy0             = _mm_macc_pd(dy10,fscal,fjy0);
 487             fjz0             = _mm_macc_pd(dz10,fscal,fjz0);
 488
 489             /**************************
 490              * CALCULATE INTERACTIONS *
 491              **************************/
 492
 493             r11              = _mm_mul_pd(rsq11,rinv11);
 494
 495             /* EWALD ELECTROSTATICS */
 496
 497             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 498             ewrt             = _mm_mul_pd(r11,ewtabscale);
 499             ewitab           = _mm_cvttpd_epi32(ewrt);
 500 #ifdef __XOP__
 501             eweps            = _mm_frcz_pd(ewrt);
 502 #else
 503             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 504 #endif
 505             twoeweps         = _mm_add_pd(eweps,eweps);
 506             ewitab           = _mm_slli_epi32(ewitab,2);
 507             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 508             ewtabD           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 509             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 510             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
 511             ewtabFn          = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
 512             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 513             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
 514             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
 515             velec            = _mm_mul_pd(qq11,_mm_sub_pd(rinv11,velec));
 516             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
 517
 518             /* Update potential sum for this i atom from the interaction with this j atom. */
 519             velecsum         = _mm_add_pd(velecsum,velec);
 520
 521             fscal            = felec;
 522
 523             /* Update vectorial force */
 524             fix1             = _mm_macc_pd(dx11,fscal,fix1);
 525             fiy1             = _mm_macc_pd(dy11,fscal,fiy1);
 526             fiz1             = _mm_macc_pd(dz11,fscal,fiz1);
 527
 528             fjx1             = _mm_macc_pd(dx11,fscal,fjx1);
 529             fjy1             = _mm_macc_pd(dy11,fscal,fjy1);
 530             fjz1             = _mm_macc_pd(dz11,fscal,fjz1);
 531
 532             /**************************
 533              * CALCULATE INTERACTIONS *
 534              **************************/
 535
 536             r12              = _mm_mul_pd(rsq12,rinv12);
 537
 538             /* EWALD ELECTROSTATICS */
 539
 540             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 541             ewrt             = _mm_mul_pd(r12,ewtabscale);
 542             ewitab           = _mm_cvttpd_epi32(ewrt);
 543 #ifdef __XOP__
 544             eweps            = _mm_frcz_pd(ewrt);
 545 #else
 546             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 547 #endif
 548             twoeweps         = _mm_add_pd(eweps,eweps);
 549             ewitab           = _mm_slli_epi32(ewitab,2);
 550             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 551             ewtabD           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 552             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 553             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
 554             ewtabFn          = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
 555             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 556             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
 557             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
 558             velec            = _mm_mul_pd(qq12,_mm_sub_pd(rinv12,velec));
 559             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
 560
 561             /* Update potential sum for this i atom from the interaction with this j atom. */
 562             velecsum         = _mm_add_pd(velecsum,velec);
 563
 564             fscal            = felec;
 565
 566             /* Update vectorial force */
 567             fix1             = _mm_macc_pd(dx12,fscal,fix1);
 568             fiy1             = _mm_macc_pd(dy12,fscal,fiy1);
 569             fiz1             = _mm_macc_pd(dz12,fscal,fiz1);
 570
 571             fjx2             = _mm_macc_pd(dx12,fscal,fjx2);
 572             fjy2             = _mm_macc_pd(dy12,fscal,fjy2);
 573             fjz2             = _mm_macc_pd(dz12,fscal,fjz2);
 574
 575             /**************************
 576              * CALCULATE INTERACTIONS *
 577              **************************/
 578
 579             r20              = _mm_mul_pd(rsq20,rinv20);
 580
 581             /* EWALD ELECTROSTATICS */
 582
 583             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 584             ewrt             = _mm_mul_pd(r20,ewtabscale);
 585             ewitab           = _mm_cvttpd_epi32(ewrt);
 586 #ifdef __XOP__
 587             eweps            = _mm_frcz_pd(ewrt);
 588 #else
 589             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 590 #endif
 591             twoeweps         = _mm_add_pd(eweps,eweps);
 592             ewitab           = _mm_slli_epi32(ewitab,2);
 593             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 594             ewtabD           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 595             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 596             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
 597             ewtabFn          = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
 598             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 599             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
 600             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
 601             velec            = _mm_mul_pd(qq20,_mm_sub_pd(rinv20,velec));
 602             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
 603
 604             /* Update potential sum for this i atom from the interaction with this j atom. */
 605             velecsum         = _mm_add_pd(velecsum,velec);
 606
 607             fscal            = felec;
 608
 609             /* Update vectorial force */
 610             fix2             = _mm_macc_pd(dx20,fscal,fix2);
 611             fiy2             = _mm_macc_pd(dy20,fscal,fiy2);
 612             fiz2             = _mm_macc_pd(dz20,fscal,fiz2);
 613
 614             fjx0             = _mm_macc_pd(dx20,fscal,fjx0);
 615             fjy0             = _mm_macc_pd(dy20,fscal,fjy0);
 616             fjz0             = _mm_macc_pd(dz20,fscal,fjz0);
 617
 618             /**************************
 619              * CALCULATE INTERACTIONS *
 620              **************************/
 621
 622             r21              = _mm_mul_pd(rsq21,rinv21);
 623
 624             /* EWALD ELECTROSTATICS */
 625
 626             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 627             ewrt             = _mm_mul_pd(r21,ewtabscale);
 628             ewitab           = _mm_cvttpd_epi32(ewrt);
 629 #ifdef __XOP__
 630             eweps            = _mm_frcz_pd(ewrt);
 631 #else
 632             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 633 #endif
 634             twoeweps         = _mm_add_pd(eweps,eweps);
 635             ewitab           = _mm_slli_epi32(ewitab,2);
 636             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 637             ewtabD           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 638             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 639             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
 640             ewtabFn          = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
 641             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 642             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
 643             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
 644             velec            = _mm_mul_pd(qq21,_mm_sub_pd(rinv21,velec));
 645             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
 646
 647             /* Update potential sum for this i atom from the interaction with this j atom. */
 648             velecsum         = _mm_add_pd(velecsum,velec);
 649
 650             fscal            = felec;
 651
 652             /* Update vectorial force */
 653             fix2             = _mm_macc_pd(dx21,fscal,fix2);
 654             fiy2             = _mm_macc_pd(dy21,fscal,fiy2);
 655             fiz2             = _mm_macc_pd(dz21,fscal,fiz2);
 656
 657             fjx1             = _mm_macc_pd(dx21,fscal,fjx1);
 658             fjy1             = _mm_macc_pd(dy21,fscal,fjy1);
 659             fjz1             = _mm_macc_pd(dz21,fscal,fjz1);
 660
 661             /**************************
 662              * CALCULATE INTERACTIONS *
 663              **************************/
 664
 665             r22              = _mm_mul_pd(rsq22,rinv22);
 666
 667             /* EWALD ELECTROSTATICS */
 668
 669             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 670             ewrt             = _mm_mul_pd(r22,ewtabscale);
 671             ewitab           = _mm_cvttpd_epi32(ewrt);
 672 #ifdef __XOP__
 673             eweps            = _mm_frcz_pd(ewrt);
 674 #else
 675             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 676 #endif
 677             twoeweps         = _mm_add_pd(eweps,eweps);
 678             ewitab           = _mm_slli_epi32(ewitab,2);
 679             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 680             ewtabD           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 681             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 682             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
 683             ewtabFn          = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
 684             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 685             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
 686             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
 687             velec            = _mm_mul_pd(qq22,_mm_sub_pd(rinv22,velec));
 688             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
 689
 690             /* Update potential sum for this i atom from the interaction with this j atom. */
 691             velecsum         = _mm_add_pd(velecsum,velec);
 692
 693             fscal            = felec;
 694
 695             /* Update vectorial force */
 696             fix2             = _mm_macc_pd(dx22,fscal,fix2);
 697             fiy2             = _mm_macc_pd(dy22,fscal,fiy2);
 698             fiz2             = _mm_macc_pd(dz22,fscal,fiz2);
 699
 700             fjx2             = _mm_macc_pd(dx22,fscal,fjx2);
 701             fjy2             = _mm_macc_pd(dy22,fscal,fjy2);
 702             fjz2             = _mm_macc_pd(dz22,fscal,fjz2);
 703
 704             gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 705
 706             /* Inner loop uses 420 flops */
 707         }
 708
 709         if(jidx<j_index_end)
 710         {
 711
 712             jnrA             = jjnr[jidx];
 713             j_coord_offsetA  = DIM*jnrA;
 714
 715             /* load j atom coordinates */
 716             gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 717                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 718
 719             /* Calculate displacement vector */
 720             dx00             = _mm_sub_pd(ix0,jx0);
 721             dy00             = _mm_sub_pd(iy0,jy0);
 722             dz00             = _mm_sub_pd(iz0,jz0);
 723             dx01             = _mm_sub_pd(ix0,jx1);
 724             dy01             = _mm_sub_pd(iy0,jy1);
 725             dz01             = _mm_sub_pd(iz0,jz1);
 726             dx02             = _mm_sub_pd(ix0,jx2);
 727             dy02             = _mm_sub_pd(iy0,jy2);
 728             dz02             = _mm_sub_pd(iz0,jz2);
 729             dx10             = _mm_sub_pd(ix1,jx0);
 730             dy10             = _mm_sub_pd(iy1,jy0);
 731             dz10             = _mm_sub_pd(iz1,jz0);
 732             dx11             = _mm_sub_pd(ix1,jx1);
 733             dy11             = _mm_sub_pd(iy1,jy1);
 734             dz11             = _mm_sub_pd(iz1,jz1);
 735             dx12             = _mm_sub_pd(ix1,jx2);
 736             dy12             = _mm_sub_pd(iy1,jy2);
 737             dz12             = _mm_sub_pd(iz1,jz2);
 738             dx20             = _mm_sub_pd(ix2,jx0);
 739             dy20             = _mm_sub_pd(iy2,jy0);
 740             dz20             = _mm_sub_pd(iz2,jz0);
 741             dx21             = _mm_sub_pd(ix2,jx1);
 742             dy21             = _mm_sub_pd(iy2,jy1);
 743             dz21             = _mm_sub_pd(iz2,jz1);
 744             dx22             = _mm_sub_pd(ix2,jx2);
 745             dy22             = _mm_sub_pd(iy2,jy2);
 746             dz22             = _mm_sub_pd(iz2,jz2);
 747
 748             /* Calculate squared distance and things based on it */
 749             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 750             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
 751             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
 752             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 753             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 754             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 755             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 756             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 757             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 758
 759             rinv00           = avx128fma_invsqrt_d(rsq00);
 760             rinv01           = avx128fma_invsqrt_d(rsq01);
 761             rinv02           = avx128fma_invsqrt_d(rsq02);
 762             rinv10           = avx128fma_invsqrt_d(rsq10);
 763             rinv11           = avx128fma_invsqrt_d(rsq11);
 764             rinv12           = avx128fma_invsqrt_d(rsq12);
 765             rinv20           = avx128fma_invsqrt_d(rsq20);
 766             rinv21           = avx128fma_invsqrt_d(rsq21);
 767             rinv22           = avx128fma_invsqrt_d(rsq22);
 768
 769             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 770             rinvsq01         = _mm_mul_pd(rinv01,rinv01);
 771             rinvsq02         = _mm_mul_pd(rinv02,rinv02);
 772             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
 773             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
 774             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
 775             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
 776             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
 777             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
 778
 779             fjx0             = _mm_setzero_pd();
 780             fjy0             = _mm_setzero_pd();
 781             fjz0             = _mm_setzero_pd();
 782             fjx1             = _mm_setzero_pd();
 783             fjy1             = _mm_setzero_pd();
 784             fjz1             = _mm_setzero_pd();
 785             fjx2             = _mm_setzero_pd();
 786             fjy2             = _mm_setzero_pd();
 787             fjz2             = _mm_setzero_pd();
 788
 789             /**************************
 790              * CALCULATE INTERACTIONS *
 791              **************************/
 792
 793             r00              = _mm_mul_pd(rsq00,rinv00);
 794
 795             /* EWALD ELECTROSTATICS */
 796
 797             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 798             ewrt             = _mm_mul_pd(r00,ewtabscale);
 799             ewitab           = _mm_cvttpd_epi32(ewrt);
 800 #ifdef __XOP__
 801             eweps            = _mm_frcz_pd(ewrt);
 802 #else
 803             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 804 #endif
 805             twoeweps         = _mm_add_pd(eweps,eweps);
 806             ewitab           = _mm_slli_epi32(ewitab,2);
 807             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 808             ewtabD           = _mm_setzero_pd();
 809             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 810             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
 811             ewtabFn          = _mm_setzero_pd();
 812             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 813             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
 814             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
 815             velec            = _mm_mul_pd(qq00,_mm_sub_pd(rinv00,velec));
 816             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
 817
 818             /* Analytical LJ-PME */
 819             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 820             ewcljrsq         = _mm_mul_pd(ewclj2,rsq00);
 821             ewclj6           = _mm_mul_pd(ewclj2,_mm_mul_pd(ewclj2,ewclj2));
 822             exponent         = avx128fma_exp_d(ewcljrsq);
 823             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
 824             poly             = _mm_mul_pd(exponent,_mm_macc_pd(_mm_mul_pd(ewcljrsq,ewcljrsq),one_half,_mm_sub_pd(one,ewcljrsq)));
 825             /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
 826             vvdw6            = _mm_mul_pd(_mm_macc_pd(-c6grid_00,_mm_sub_pd(one,poly),c6_00),rinvsix);
 827             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
 828             vvdw             = _mm_msub_pd(vvdw12,one_twelfth,_mm_mul_pd(vvdw6,one_sixth));
 829             /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
 830             fvdw             = _mm_mul_pd(_mm_add_pd(vvdw12,_mm_msub_pd(_mm_mul_pd(c6grid_00,one_sixth),_mm_mul_pd(exponent,ewclj6),vvdw6)),rinvsq00);
 831
 832             /* Update potential sum for this i atom from the interaction with this j atom. */
 833             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 834             velecsum         = _mm_add_pd(velecsum,velec);
 835             vvdw             = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
 836             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 837
 838             fscal            = _mm_add_pd(felec,fvdw);
 839
 840             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 841
 842             /* Update vectorial force */
 843             fix0             = _mm_macc_pd(dx00,fscal,fix0);
 844             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
 845             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
 846
 847             fjx0             = _mm_macc_pd(dx00,fscal,fjx0);
 848             fjy0             = _mm_macc_pd(dy00,fscal,fjy0);
 849             fjz0             = _mm_macc_pd(dz00,fscal,fjz0);
 850
 851             /**************************
 852              * CALCULATE INTERACTIONS *
 853              **************************/
 854
 855             r01              = _mm_mul_pd(rsq01,rinv01);
 856
 857             /* EWALD ELECTROSTATICS */
 858
 859             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 860             ewrt             = _mm_mul_pd(r01,ewtabscale);
 861             ewitab           = _mm_cvttpd_epi32(ewrt);
 862 #ifdef __XOP__
 863             eweps            = _mm_frcz_pd(ewrt);
 864 #else
 865             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 866 #endif
 867             twoeweps         = _mm_add_pd(eweps,eweps);
 868             ewitab           = _mm_slli_epi32(ewitab,2);
 869             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 870             ewtabD           = _mm_setzero_pd();
 871             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 872             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
 873             ewtabFn          = _mm_setzero_pd();
 874             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 875             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
 876             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
 877             velec            = _mm_mul_pd(qq01,_mm_sub_pd(rinv01,velec));
 878             felec            = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
 879
 880             /* Update potential sum for this i atom from the interaction with this j atom. */
 881             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 882             velecsum         = _mm_add_pd(velecsum,velec);
 883
 884             fscal            = felec;
 885
 886             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 887
 888             /* Update vectorial force */
 889             fix0             = _mm_macc_pd(dx01,fscal,fix0);
 890             fiy0             = _mm_macc_pd(dy01,fscal,fiy0);
 891             fiz0             = _mm_macc_pd(dz01,fscal,fiz0);
 892
 893             fjx1             = _mm_macc_pd(dx01,fscal,fjx1);
 894             fjy1             = _mm_macc_pd(dy01,fscal,fjy1);
 895             fjz1             = _mm_macc_pd(dz01,fscal,fjz1);
 896
 897             /**************************
 898              * CALCULATE INTERACTIONS *
 899              **************************/
 900
 901             r02              = _mm_mul_pd(rsq02,rinv02);
 902
 903             /* EWALD ELECTROSTATICS */
 904
 905             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 906             ewrt             = _mm_mul_pd(r02,ewtabscale);
 907             ewitab           = _mm_cvttpd_epi32(ewrt);
 908 #ifdef __XOP__
 909             eweps            = _mm_frcz_pd(ewrt);
 910 #else
 911             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 912 #endif
 913             twoeweps         = _mm_add_pd(eweps,eweps);
 914             ewitab           = _mm_slli_epi32(ewitab,2);
 915             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 916             ewtabD           = _mm_setzero_pd();
 917             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 918             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
 919             ewtabFn          = _mm_setzero_pd();
 920             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 921             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
 922             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
 923             velec            = _mm_mul_pd(qq02,_mm_sub_pd(rinv02,velec));
 924             felec            = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
 925
 926             /* Update potential sum for this i atom from the interaction with this j atom. */
 927             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 928             velecsum         = _mm_add_pd(velecsum,velec);
 929
 930             fscal            = felec;
 931
 932             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 933
 934             /* Update vectorial force */
 935             fix0             = _mm_macc_pd(dx02,fscal,fix0);
 936             fiy0             = _mm_macc_pd(dy02,fscal,fiy0);
 937             fiz0             = _mm_macc_pd(dz02,fscal,fiz0);
 938
 939             fjx2             = _mm_macc_pd(dx02,fscal,fjx2);
 940             fjy2             = _mm_macc_pd(dy02,fscal,fjy2);
 941             fjz2             = _mm_macc_pd(dz02,fscal,fjz2);
 942
 943             /**************************
 944              * CALCULATE INTERACTIONS *
 945              **************************/
 946
 947             r10              = _mm_mul_pd(rsq10,rinv10);
 948
 949             /* EWALD ELECTROSTATICS */
 950
 951             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 952             ewrt             = _mm_mul_pd(r10,ewtabscale);
 953             ewitab           = _mm_cvttpd_epi32(ewrt);
 954 #ifdef __XOP__
 955             eweps            = _mm_frcz_pd(ewrt);
 956 #else
 957             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 958 #endif
 959             twoeweps         = _mm_add_pd(eweps,eweps);
 960             ewitab           = _mm_slli_epi32(ewitab,2);
 961             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 962             ewtabD           = _mm_setzero_pd();
 963             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 964             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
 965             ewtabFn          = _mm_setzero_pd();
 966             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 967             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
 968             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
 969             velec            = _mm_mul_pd(qq10,_mm_sub_pd(rinv10,velec));
 970             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
 971
 972             /* Update potential sum for this i atom from the interaction with this j atom. */
 973             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 974             velecsum         = _mm_add_pd(velecsum,velec);
 975
 976             fscal            = felec;
 977
 978             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 979
 980             /* Update vectorial force */
 981             fix1             = _mm_macc_pd(dx10,fscal,fix1);
 982             fiy1             = _mm_macc_pd(dy10,fscal,fiy1);
 983             fiz1             = _mm_macc_pd(dz10,fscal,fiz1);
 984
 985             fjx0             = _mm_macc_pd(dx10,fscal,fjx0);
 986             fjy0             = _mm_macc_pd(dy10,fscal,fjy0);
 987             fjz0             = _mm_macc_pd(dz10,fscal,fjz0);
 988
 989             /**************************
 990              * CALCULATE INTERACTIONS *
 991              **************************/
 992
 993             r11              = _mm_mul_pd(rsq11,rinv11);
 994
 995             /* EWALD ELECTROSTATICS */
 996
 997             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 998             ewrt             = _mm_mul_pd(r11,ewtabscale);
 999             ewitab           = _mm_cvttpd_epi32(ewrt);
1000 #ifdef __XOP__
1001             eweps            = _mm_frcz_pd(ewrt);
1002 #else
1003             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1004 #endif
1005             twoeweps         = _mm_add_pd(eweps,eweps);
1006             ewitab           = _mm_slli_epi32(ewitab,2);
1007             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1008             ewtabD           = _mm_setzero_pd();
1009             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1010             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
1011             ewtabFn          = _mm_setzero_pd();
1012             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1013             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
1014             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
1015             velec            = _mm_mul_pd(qq11,_mm_sub_pd(rinv11,velec));
1016             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
1017
1018             /* Update potential sum for this i atom from the interaction with this j atom. */
1019             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1020             velecsum         = _mm_add_pd(velecsum,velec);
1021
1022             fscal            = felec;
1023
1024             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1025
1026             /* Update vectorial force */
1027             fix1             = _mm_macc_pd(dx11,fscal,fix1);
1028             fiy1             = _mm_macc_pd(dy11,fscal,fiy1);
1029             fiz1             = _mm_macc_pd(dz11,fscal,fiz1);
1030
1031             fjx1             = _mm_macc_pd(dx11,fscal,fjx1);
1032             fjy1             = _mm_macc_pd(dy11,fscal,fjy1);
1033             fjz1             = _mm_macc_pd(dz11,fscal,fjz1);
1034
1035             /**************************
1036              * CALCULATE INTERACTIONS *
1037              **************************/
1038
1039             r12              = _mm_mul_pd(rsq12,rinv12);
1040
1041             /* EWALD ELECTROSTATICS */
1042
1043             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1044             ewrt             = _mm_mul_pd(r12,ewtabscale);
1045             ewitab           = _mm_cvttpd_epi32(ewrt);
1046 #ifdef __XOP__
1047             eweps            = _mm_frcz_pd(ewrt);
1048 #else
1049             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1050 #endif
1051             twoeweps         = _mm_add_pd(eweps,eweps);
1052             ewitab           = _mm_slli_epi32(ewitab,2);
1053             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1054             ewtabD           = _mm_setzero_pd();
1055             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1056             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
1057             ewtabFn          = _mm_setzero_pd();
1058             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1059             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
1060             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
1061             velec            = _mm_mul_pd(qq12,_mm_sub_pd(rinv12,velec));
1062             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
1063
1064             /* Update potential sum for this i atom from the interaction with this j atom. */
1065             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1066             velecsum         = _mm_add_pd(velecsum,velec);
1067
1068             fscal            = felec;
1069
1070             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1071
1072             /* Update vectorial force */
1073             fix1             = _mm_macc_pd(dx12,fscal,fix1);
1074             fiy1             = _mm_macc_pd(dy12,fscal,fiy1);
1075             fiz1             = _mm_macc_pd(dz12,fscal,fiz1);
1076
1077             fjx2             = _mm_macc_pd(dx12,fscal,fjx2);
1078             fjy2             = _mm_macc_pd(dy12,fscal,fjy2);
1079             fjz2             = _mm_macc_pd(dz12,fscal,fjz2);
1080
1081             /**************************
1082              * CALCULATE INTERACTIONS *
1083              **************************/
1084
1085             r20              = _mm_mul_pd(rsq20,rinv20);
1086
1087             /* EWALD ELECTROSTATICS */
1088
1089             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1090             ewrt             = _mm_mul_pd(r20,ewtabscale);
1091             ewitab           = _mm_cvttpd_epi32(ewrt);
1092 #ifdef __XOP__
1093             eweps            = _mm_frcz_pd(ewrt);
1094 #else
1095             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1096 #endif
1097             twoeweps         = _mm_add_pd(eweps,eweps);
1098             ewitab           = _mm_slli_epi32(ewitab,2);
1099             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1100             ewtabD           = _mm_setzero_pd();
1101             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1102             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
1103             ewtabFn          = _mm_setzero_pd();
1104             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1105             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
1106             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
1107             velec            = _mm_mul_pd(qq20,_mm_sub_pd(rinv20,velec));
1108             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
1109
1110             /* Update potential sum for this i atom from the interaction with this j atom. */
1111             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1112             velecsum         = _mm_add_pd(velecsum,velec);
1113
1114             fscal            = felec;
1115
1116             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1117
1118             /* Update vectorial force */
1119             fix2             = _mm_macc_pd(dx20,fscal,fix2);
1120             fiy2             = _mm_macc_pd(dy20,fscal,fiy2);
1121             fiz2             = _mm_macc_pd(dz20,fscal,fiz2);
1122
1123             fjx0             = _mm_macc_pd(dx20,fscal,fjx0);
1124             fjy0             = _mm_macc_pd(dy20,fscal,fjy0);
1125             fjz0             = _mm_macc_pd(dz20,fscal,fjz0);
1126
1127             /**************************
1128              * CALCULATE INTERACTIONS *
1129              **************************/
1130
1131             r21              = _mm_mul_pd(rsq21,rinv21);
1132
1133             /* EWALD ELECTROSTATICS */
1134
1135             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1136             ewrt             = _mm_mul_pd(r21,ewtabscale);
1137             ewitab           = _mm_cvttpd_epi32(ewrt);
1138 #ifdef __XOP__
1139             eweps            = _mm_frcz_pd(ewrt);
1140 #else
1141             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1142 #endif
1143             twoeweps         = _mm_add_pd(eweps,eweps);
1144             ewitab           = _mm_slli_epi32(ewitab,2);
1145             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1146             ewtabD           = _mm_setzero_pd();
1147             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1148             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
1149             ewtabFn          = _mm_setzero_pd();
1150             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1151             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
1152             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
1153             velec            = _mm_mul_pd(qq21,_mm_sub_pd(rinv21,velec));
1154             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
1155
1156             /* Update potential sum for this i atom from the interaction with this j atom. */
1157             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1158             velecsum         = _mm_add_pd(velecsum,velec);
1159
1160             fscal            = felec;
1161
1162             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1163
1164             /* Update vectorial force */
1165             fix2             = _mm_macc_pd(dx21,fscal,fix2);
1166             fiy2             = _mm_macc_pd(dy21,fscal,fiy2);
1167             fiz2             = _mm_macc_pd(dz21,fscal,fiz2);
1168
1169             fjx1             = _mm_macc_pd(dx21,fscal,fjx1);
1170             fjy1             = _mm_macc_pd(dy21,fscal,fjy1);
1171             fjz1             = _mm_macc_pd(dz21,fscal,fjz1);
1172
1173             /**************************
1174              * CALCULATE INTERACTIONS *
1175              **************************/
1176
1177             r22              = _mm_mul_pd(rsq22,rinv22);
1178
1179             /* EWALD ELECTROSTATICS */
1180
1181             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1182             ewrt             = _mm_mul_pd(r22,ewtabscale);
1183             ewitab           = _mm_cvttpd_epi32(ewrt);
1184 #ifdef __XOP__
1185             eweps            = _mm_frcz_pd(ewrt);
1186 #else
1187             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1188 #endif
1189             twoeweps         = _mm_add_pd(eweps,eweps);
1190             ewitab           = _mm_slli_epi32(ewitab,2);
1191             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1192             ewtabD           = _mm_setzero_pd();
1193             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1194             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
1195             ewtabFn          = _mm_setzero_pd();
1196             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1197             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
1198             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
1199             velec            = _mm_mul_pd(qq22,_mm_sub_pd(rinv22,velec));
1200             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
1201
1202             /* Update potential sum for this i atom from the interaction with this j atom. */
1203             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1204             velecsum         = _mm_add_pd(velecsum,velec);
1205
1206             fscal            = felec;
1207
1208             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1209
1210             /* Update vectorial force */
1211             fix2             = _mm_macc_pd(dx22,fscal,fix2);
1212             fiy2             = _mm_macc_pd(dy22,fscal,fiy2);
1213             fiz2             = _mm_macc_pd(dz22,fscal,fiz2);
1214
1215             fjx2             = _mm_macc_pd(dx22,fscal,fjx2);
1216             fjy2             = _mm_macc_pd(dy22,fscal,fjy2);
1217             fjz2             = _mm_macc_pd(dz22,fscal,fjz2);
1218
1219             gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1220
1221             /* Inner loop uses 420 flops */
1222         }
1223
1224         /* End of innermost loop */
1225
1226         gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1227                                               f+i_coord_offset,fshift+i_shift_offset);
1228
1229         ggid                        = gid[iidx];
1230         /* Update potential energies */
1231         gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1232         gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1233
1234         /* Increment number of inner iterations */
1235         inneriter                  += j_index_end - j_index_start;
1236
1237         /* Outer loop uses 20 flops */
1238     }
1239
1240     /* Increment number of outer iterations */
1241     outeriter        += nri;
1242
1243     /* Update outer/inner flops */
1244
1245     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_VF,outeriter*20 + inneriter*420);
1246 }
1247 /*
1248  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwLJEw_GeomW3W3_F_avx_128_fma_double
1249  * Electrostatics interaction: Ewald
1250  * VdW interaction:            LJEwald
1251  * Geometry:                   Water3-Water3
1252  * Calculate force/pot:        Force
1253  */
1254 void
1255 nb_kernel_ElecEw_VdwLJEw_GeomW3W3_F_avx_128_fma_double
1256                     (t_nblist                    * gmx_restrict       nlist,
1257                      rvec                        * gmx_restrict          xx,
1258                      rvec                        * gmx_restrict          ff,
1259                      struct t_forcerec           * gmx_restrict          fr,
1260                      t_mdatoms                   * gmx_restrict     mdatoms,
1261                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1262                      t_nrnb                      * gmx_restrict        nrnb)
1263 {
1264     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1265      * just 0 for non-waters.
1266      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
1267      * jnr indices corresponding to data put in the four positions in the SIMD register.
1268      */
1269     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1270     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1271     int              jnrA,jnrB;
1272     int              j_coord_offsetA,j_coord_offsetB;
1273     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1274     real             rcutoff_scalar;
1275     real             *shiftvec,*fshift,*x,*f;
1276     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1277     int              vdwioffset0;
1278     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1279     int              vdwioffset1;
1280     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1281     int              vdwioffset2;
1282     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1283     int              vdwjidx0A,vdwjidx0B;
1284     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1285     int              vdwjidx1A,vdwjidx1B;
1286     __m128d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1287     int              vdwjidx2A,vdwjidx2B;
1288     __m128d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1289     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1290     __m128d          dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1291     __m128d          dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1292     __m128d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1293     __m128d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1294     __m128d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1295     __m128d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1296     __m128d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1297     __m128d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1298     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
1299     real             *charge;
1300     int              nvdwtype;
1301     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1302     int              *vdwtype;
1303     real             *vdwparam;
1304     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
1305     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
1306     __m128d           c6grid_00;
1307     __m128d           c6grid_01;
1308     __m128d           c6grid_02;
1309     __m128d           c6grid_10;
1310     __m128d           c6grid_11;
1311     __m128d           c6grid_12;
1312     __m128d           c6grid_20;
1313     __m128d           c6grid_21;
1314     __m128d           c6grid_22;
1315     real             *vdwgridparam;
1316     __m128d           ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
1317     __m128d           one_half  = _mm_set1_pd(0.5);
1318     __m128d           minus_one = _mm_set1_pd(-1.0);
1319     __m128i          ewitab;
1320     __m128d          ewtabscale,eweps,twoeweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1321     real             *ewtab;
1322     __m128d          dummy_mask,cutoff_mask;
1323     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
1324     __m128d          one     = _mm_set1_pd(1.0);
1325     __m128d          two     = _mm_set1_pd(2.0);
1326     x                = xx[0];
1327     f                = ff[0];
1328
1329     nri              = nlist->nri;
1330     iinr             = nlist->iinr;
1331     jindex           = nlist->jindex;
1332     jjnr             = nlist->jjnr;
1333     shiftidx         = nlist->shift;
1334     gid              = nlist->gid;
1335     shiftvec         = fr->shift_vec[0];
1336     fshift           = fr->fshift[0];
1337     facel            = _mm_set1_pd(fr->ic->epsfac);
1338     charge           = mdatoms->chargeA;
1339     nvdwtype         = fr->ntype;
1340     vdwparam         = fr->nbfp;
1341     vdwtype          = mdatoms->typeA;
1342     vdwgridparam     = fr->ljpme_c6grid;
1343     sh_lj_ewald      = _mm_set1_pd(fr->ic->sh_lj_ewald);
1344     ewclj            = _mm_set1_pd(fr->ic->ewaldcoeff_lj);
1345     ewclj2           = _mm_mul_pd(minus_one,_mm_mul_pd(ewclj,ewclj));
1346
1347     sh_ewald         = _mm_set1_pd(fr->ic->sh_ewald);
1348     ewtab            = fr->ic->tabq_coul_F;
1349     ewtabscale       = _mm_set1_pd(fr->ic->tabq_scale);
1350     ewtabhalfspace   = _mm_set1_pd(0.5/fr->ic->tabq_scale);
1351
1352     /* Setup water-specific parameters */
1353     inr              = nlist->iinr[0];
1354     iq0              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
1355     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
1356     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
1357     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
1358
1359     jq0              = _mm_set1_pd(charge[inr+0]);
1360     jq1              = _mm_set1_pd(charge[inr+1]);
1361     jq2              = _mm_set1_pd(charge[inr+2]);
1362     vdwjidx0A        = 2*vdwtype[inr+0];
1363     qq00             = _mm_mul_pd(iq0,jq0);
1364     c6_00            = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
1365     c12_00           = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
1366     c6grid_00        = _mm_set1_pd(vdwgridparam[vdwioffset0+vdwjidx0A]);
1367     qq01             = _mm_mul_pd(iq0,jq1);
1368     qq02             = _mm_mul_pd(iq0,jq2);
1369     qq10             = _mm_mul_pd(iq1,jq0);
1370     qq11             = _mm_mul_pd(iq1,jq1);
1371     qq12             = _mm_mul_pd(iq1,jq2);
1372     qq20             = _mm_mul_pd(iq2,jq0);
1373     qq21             = _mm_mul_pd(iq2,jq1);
1374     qq22             = _mm_mul_pd(iq2,jq2);
1375
1376     /* Avoid stupid compiler warnings */
1377     jnrA = jnrB = 0;
1378     j_coord_offsetA = 0;
1379     j_coord_offsetB = 0;
1380
1381     outeriter        = 0;
1382     inneriter        = 0;
1383
1384     /* Start outer loop over neighborlists */
1385     for(iidx=0; iidx<nri; iidx++)
1386     {
1387         /* Load shift vector for this list */
1388         i_shift_offset   = DIM*shiftidx[iidx];
1389
1390         /* Load limits for loop over neighbors */
1391         j_index_start    = jindex[iidx];
1392         j_index_end      = jindex[iidx+1];
1393
1394         /* Get outer coordinate index */
1395         inr              = iinr[iidx];
1396         i_coord_offset   = DIM*inr;
1397
1398         /* Load i particle coords and add shift vector */
1399         gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1400                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1401
1402         fix0             = _mm_setzero_pd();
1403         fiy0             = _mm_setzero_pd();
1404         fiz0             = _mm_setzero_pd();
1405         fix1             = _mm_setzero_pd();
1406         fiy1             = _mm_setzero_pd();
1407         fiz1             = _mm_setzero_pd();
1408         fix2             = _mm_setzero_pd();
1409         fiy2             = _mm_setzero_pd();
1410         fiz2             = _mm_setzero_pd();
1411
1412         /* Start inner kernel loop */
1413         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
1414         {
1415
1416             /* Get j neighbor index, and coordinate index */
1417             jnrA             = jjnr[jidx];
1418             jnrB             = jjnr[jidx+1];
1419             j_coord_offsetA  = DIM*jnrA;
1420             j_coord_offsetB  = DIM*jnrB;
1421
1422             /* load j atom coordinates */
1423             gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1424                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1425
1426             /* Calculate displacement vector */
1427             dx00             = _mm_sub_pd(ix0,jx0);
1428             dy00             = _mm_sub_pd(iy0,jy0);
1429             dz00             = _mm_sub_pd(iz0,jz0);
1430             dx01             = _mm_sub_pd(ix0,jx1);
1431             dy01             = _mm_sub_pd(iy0,jy1);
1432             dz01             = _mm_sub_pd(iz0,jz1);
1433             dx02             = _mm_sub_pd(ix0,jx2);
1434             dy02             = _mm_sub_pd(iy0,jy2);
1435             dz02             = _mm_sub_pd(iz0,jz2);
1436             dx10             = _mm_sub_pd(ix1,jx0);
1437             dy10             = _mm_sub_pd(iy1,jy0);
1438             dz10             = _mm_sub_pd(iz1,jz0);
1439             dx11             = _mm_sub_pd(ix1,jx1);
1440             dy11             = _mm_sub_pd(iy1,jy1);
1441             dz11             = _mm_sub_pd(iz1,jz1);
1442             dx12             = _mm_sub_pd(ix1,jx2);
1443             dy12             = _mm_sub_pd(iy1,jy2);
1444             dz12             = _mm_sub_pd(iz1,jz2);
1445             dx20             = _mm_sub_pd(ix2,jx0);
1446             dy20             = _mm_sub_pd(iy2,jy0);
1447             dz20             = _mm_sub_pd(iz2,jz0);
1448             dx21             = _mm_sub_pd(ix2,jx1);
1449             dy21             = _mm_sub_pd(iy2,jy1);
1450             dz21             = _mm_sub_pd(iz2,jz1);
1451             dx22             = _mm_sub_pd(ix2,jx2);
1452             dy22             = _mm_sub_pd(iy2,jy2);
1453             dz22             = _mm_sub_pd(iz2,jz2);
1454
1455             /* Calculate squared distance and things based on it */
1456             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1457             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
1458             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
1459             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
1460             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1461             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1462             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
1463             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1464             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1465
1466             rinv00           = avx128fma_invsqrt_d(rsq00);
1467             rinv01           = avx128fma_invsqrt_d(rsq01);
1468             rinv02           = avx128fma_invsqrt_d(rsq02);
1469             rinv10           = avx128fma_invsqrt_d(rsq10);
1470             rinv11           = avx128fma_invsqrt_d(rsq11);
1471             rinv12           = avx128fma_invsqrt_d(rsq12);
1472             rinv20           = avx128fma_invsqrt_d(rsq20);
1473             rinv21           = avx128fma_invsqrt_d(rsq21);
1474             rinv22           = avx128fma_invsqrt_d(rsq22);
1475
1476             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
1477             rinvsq01         = _mm_mul_pd(rinv01,rinv01);
1478             rinvsq02         = _mm_mul_pd(rinv02,rinv02);
1479             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
1480             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
1481             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
1482             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
1483             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
1484             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
1485
1486             fjx0             = _mm_setzero_pd();
1487             fjy0             = _mm_setzero_pd();
1488             fjz0             = _mm_setzero_pd();
1489             fjx1             = _mm_setzero_pd();
1490             fjy1             = _mm_setzero_pd();
1491             fjz1             = _mm_setzero_pd();
1492             fjx2             = _mm_setzero_pd();
1493             fjy2             = _mm_setzero_pd();
1494             fjz2             = _mm_setzero_pd();
1495
1496             /**************************
1497              * CALCULATE INTERACTIONS *
1498              **************************/
1499
1500             r00              = _mm_mul_pd(rsq00,rinv00);
1501
1502             /* EWALD ELECTROSTATICS */
1503
1504             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1505             ewrt             = _mm_mul_pd(r00,ewtabscale);
1506             ewitab           = _mm_cvttpd_epi32(ewrt);
1507 #ifdef __XOP__
1508             eweps            = _mm_frcz_pd(ewrt);
1509 #else
1510             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1511 #endif
1512             twoeweps         = _mm_add_pd(eweps,eweps);
1513             gmx_mm_load_2pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),ewtab+_mm_extract_epi32(ewitab,1),
1514                                          &ewtabF,&ewtabFn);
1515             felec            = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
1516             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
1517
1518             /* Analytical LJ-PME */
1519             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1520             ewcljrsq         = _mm_mul_pd(ewclj2,rsq00);
1521             ewclj6           = _mm_mul_pd(ewclj2,_mm_mul_pd(ewclj2,ewclj2));
1522             exponent         = avx128fma_exp_d(ewcljrsq);
1523             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
1524             poly             = _mm_mul_pd(exponent,_mm_macc_pd(_mm_mul_pd(ewcljrsq,ewcljrsq),one_half,_mm_sub_pd(one,ewcljrsq)));
1525             /* f6A = 6 * C6grid * (1 - poly) */
1526             f6A              = _mm_mul_pd(c6grid_00,_mm_sub_pd(one,poly));
1527             /* f6B = C6grid * exponent * beta^6 */
1528             f6B              = _mm_mul_pd(_mm_mul_pd(c6grid_00,one_sixth),_mm_mul_pd(exponent,ewclj6));
1529             /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
1530             fvdw              = _mm_mul_pd(_mm_macc_pd(_mm_msub_pd(c12_00,rinvsix,_mm_sub_pd(c6_00,f6A)),rinvsix,f6B),rinvsq00);
1531
1532             fscal            = _mm_add_pd(felec,fvdw);
1533
1534             /* Update vectorial force */
1535             fix0             = _mm_macc_pd(dx00,fscal,fix0);
1536             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
1537             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
1538
1539             fjx0             = _mm_macc_pd(dx00,fscal,fjx0);
1540             fjy0             = _mm_macc_pd(dy00,fscal,fjy0);
1541             fjz0             = _mm_macc_pd(dz00,fscal,fjz0);
1542
1543             /**************************
1544              * CALCULATE INTERACTIONS *
1545              **************************/
1546
1547             r01              = _mm_mul_pd(rsq01,rinv01);
1548
1549             /* EWALD ELECTROSTATICS */
1550
1551             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1552             ewrt             = _mm_mul_pd(r01,ewtabscale);
1553             ewitab           = _mm_cvttpd_epi32(ewrt);
1554 #ifdef __XOP__
1555             eweps            = _mm_frcz_pd(ewrt);
1556 #else
1557             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1558 #endif
1559             twoeweps         = _mm_add_pd(eweps,eweps);
1560             gmx_mm_load_2pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),ewtab+_mm_extract_epi32(ewitab,1),
1561                                          &ewtabF,&ewtabFn);
1562             felec            = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
1563             felec            = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
1564
1565             fscal            = felec;
1566
1567             /* Update vectorial force */
1568             fix0             = _mm_macc_pd(dx01,fscal,fix0);
1569             fiy0             = _mm_macc_pd(dy01,fscal,fiy0);
1570             fiz0             = _mm_macc_pd(dz01,fscal,fiz0);
1571
1572             fjx1             = _mm_macc_pd(dx01,fscal,fjx1);
1573             fjy1             = _mm_macc_pd(dy01,fscal,fjy1);
1574             fjz1             = _mm_macc_pd(dz01,fscal,fjz1);
1575
1576             /**************************
1577              * CALCULATE INTERACTIONS *
1578              **************************/
1579
1580             r02              = _mm_mul_pd(rsq02,rinv02);
1581
1582             /* EWALD ELECTROSTATICS */
1583
1584             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1585             ewrt             = _mm_mul_pd(r02,ewtabscale);
1586             ewitab           = _mm_cvttpd_epi32(ewrt);
1587 #ifdef __XOP__
1588             eweps            = _mm_frcz_pd(ewrt);
1589 #else
1590             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1591 #endif
1592             twoeweps         = _mm_add_pd(eweps,eweps);
1593             gmx_mm_load_2pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),ewtab+_mm_extract_epi32(ewitab,1),
1594                                          &ewtabF,&ewtabFn);
1595             felec            = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
1596             felec            = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
1597
1598             fscal            = felec;
1599
1600             /* Update vectorial force */
1601             fix0             = _mm_macc_pd(dx02,fscal,fix0);
1602             fiy0             = _mm_macc_pd(dy02,fscal,fiy0);
1603             fiz0             = _mm_macc_pd(dz02,fscal,fiz0);
1604
1605             fjx2             = _mm_macc_pd(dx02,fscal,fjx2);
1606             fjy2             = _mm_macc_pd(dy02,fscal,fjy2);
1607             fjz2             = _mm_macc_pd(dz02,fscal,fjz2);
1608
1609             /**************************
1610              * CALCULATE INTERACTIONS *
1611              **************************/
1612
1613             r10              = _mm_mul_pd(rsq10,rinv10);
1614
1615             /* EWALD ELECTROSTATICS */
1616
1617             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1618             ewrt             = _mm_mul_pd(r10,ewtabscale);
1619             ewitab           = _mm_cvttpd_epi32(ewrt);
1620 #ifdef __XOP__
1621             eweps            = _mm_frcz_pd(ewrt);
1622 #else
1623             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1624 #endif
1625             twoeweps         = _mm_add_pd(eweps,eweps);
1626             gmx_mm_load_2pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),ewtab+_mm_extract_epi32(ewitab,1),
1627                                          &ewtabF,&ewtabFn);
1628             felec            = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
1629             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
1630
1631             fscal            = felec;
1632
1633             /* Update vectorial force */
1634             fix1             = _mm_macc_pd(dx10,fscal,fix1);
1635             fiy1             = _mm_macc_pd(dy10,fscal,fiy1);
1636             fiz1             = _mm_macc_pd(dz10,fscal,fiz1);
1637
1638             fjx0             = _mm_macc_pd(dx10,fscal,fjx0);
1639             fjy0             = _mm_macc_pd(dy10,fscal,fjy0);
1640             fjz0             = _mm_macc_pd(dz10,fscal,fjz0);
1641
1642             /**************************
1643              * CALCULATE INTERACTIONS *
1644              **************************/
1645
1646             r11              = _mm_mul_pd(rsq11,rinv11);
1647
1648             /* EWALD ELECTROSTATICS */
1649
1650             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1651             ewrt             = _mm_mul_pd(r11,ewtabscale);
1652             ewitab           = _mm_cvttpd_epi32(ewrt);
1653 #ifdef __XOP__
1654             eweps            = _mm_frcz_pd(ewrt);
1655 #else
1656             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1657 #endif
1658             twoeweps         = _mm_add_pd(eweps,eweps);
1659             gmx_mm_load_2pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),ewtab+_mm_extract_epi32(ewitab,1),
1660                                          &ewtabF,&ewtabFn);
1661             felec            = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
1662             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
1663
1664             fscal            = felec;
1665
1666             /* Update vectorial force */
1667             fix1             = _mm_macc_pd(dx11,fscal,fix1);
1668             fiy1             = _mm_macc_pd(dy11,fscal,fiy1);
1669             fiz1             = _mm_macc_pd(dz11,fscal,fiz1);
1670
1671             fjx1             = _mm_macc_pd(dx11,fscal,fjx1);
1672             fjy1             = _mm_macc_pd(dy11,fscal,fjy1);
1673             fjz1             = _mm_macc_pd(dz11,fscal,fjz1);
1674
1675             /**************************
1676              * CALCULATE INTERACTIONS *
1677              **************************/
1678
1679             r12              = _mm_mul_pd(rsq12,rinv12);
1680
1681             /* EWALD ELECTROSTATICS */
1682
1683             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1684             ewrt             = _mm_mul_pd(r12,ewtabscale);
1685             ewitab           = _mm_cvttpd_epi32(ewrt);
1686 #ifdef __XOP__
1687             eweps            = _mm_frcz_pd(ewrt);
1688 #else
1689             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1690 #endif
1691             twoeweps         = _mm_add_pd(eweps,eweps);
1692             gmx_mm_load_2pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),ewtab+_mm_extract_epi32(ewitab,1),
1693                                          &ewtabF,&ewtabFn);
1694             felec            = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
1695             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
1696
1697             fscal            = felec;
1698
1699             /* Update vectorial force */
1700             fix1             = _mm_macc_pd(dx12,fscal,fix1);
1701             fiy1             = _mm_macc_pd(dy12,fscal,fiy1);
1702             fiz1             = _mm_macc_pd(dz12,fscal,fiz1);
1703
1704             fjx2             = _mm_macc_pd(dx12,fscal,fjx2);
1705             fjy2             = _mm_macc_pd(dy12,fscal,fjy2);
1706             fjz2             = _mm_macc_pd(dz12,fscal,fjz2);
1707
1708             /**************************
1709              * CALCULATE INTERACTIONS *
1710              **************************/
1711
1712             r20              = _mm_mul_pd(rsq20,rinv20);
1713
1714             /* EWALD ELECTROSTATICS */
1715
1716             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1717             ewrt             = _mm_mul_pd(r20,ewtabscale);
1718             ewitab           = _mm_cvttpd_epi32(ewrt);
1719 #ifdef __XOP__
1720             eweps            = _mm_frcz_pd(ewrt);
1721 #else
1722             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1723 #endif
1724             twoeweps         = _mm_add_pd(eweps,eweps);
1725             gmx_mm_load_2pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),ewtab+_mm_extract_epi32(ewitab,1),
1726                                          &ewtabF,&ewtabFn);
1727             felec            = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
1728             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
1729
1730             fscal            = felec;
1731
1732             /* Update vectorial force */
1733             fix2             = _mm_macc_pd(dx20,fscal,fix2);
1734             fiy2             = _mm_macc_pd(dy20,fscal,fiy2);
1735             fiz2             = _mm_macc_pd(dz20,fscal,fiz2);
1736
1737             fjx0             = _mm_macc_pd(dx20,fscal,fjx0);
1738             fjy0             = _mm_macc_pd(dy20,fscal,fjy0);
1739             fjz0             = _mm_macc_pd(dz20,fscal,fjz0);
1740
1741             /**************************
1742              * CALCULATE INTERACTIONS *
1743              **************************/
1744
1745             r21              = _mm_mul_pd(rsq21,rinv21);
1746
1747             /* EWALD ELECTROSTATICS */
1748
1749             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1750             ewrt             = _mm_mul_pd(r21,ewtabscale);
1751             ewitab           = _mm_cvttpd_epi32(ewrt);
1752 #ifdef __XOP__
1753             eweps            = _mm_frcz_pd(ewrt);
1754 #else
1755             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1756 #endif
1757             twoeweps         = _mm_add_pd(eweps,eweps);
1758             gmx_mm_load_2pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),ewtab+_mm_extract_epi32(ewitab,1),
1759                                          &ewtabF,&ewtabFn);
1760             felec            = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
1761             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
1762
1763             fscal            = felec;
1764
1765             /* Update vectorial force */
1766             fix2             = _mm_macc_pd(dx21,fscal,fix2);
1767             fiy2             = _mm_macc_pd(dy21,fscal,fiy2);
1768             fiz2             = _mm_macc_pd(dz21,fscal,fiz2);
1769
1770             fjx1             = _mm_macc_pd(dx21,fscal,fjx1);
1771             fjy1             = _mm_macc_pd(dy21,fscal,fjy1);
1772             fjz1             = _mm_macc_pd(dz21,fscal,fjz1);
1773
1774             /**************************
1775              * CALCULATE INTERACTIONS *
1776              **************************/
1777
1778             r22              = _mm_mul_pd(rsq22,rinv22);
1779
1780             /* EWALD ELECTROSTATICS */
1781
1782             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1783             ewrt             = _mm_mul_pd(r22,ewtabscale);
1784             ewitab           = _mm_cvttpd_epi32(ewrt);
1785 #ifdef __XOP__
1786             eweps            = _mm_frcz_pd(ewrt);
1787 #else
1788             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1789 #endif
1790             twoeweps         = _mm_add_pd(eweps,eweps);
1791             gmx_mm_load_2pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),ewtab+_mm_extract_epi32(ewitab,1),
1792                                          &ewtabF,&ewtabFn);
1793             felec            = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
1794             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
1795
1796             fscal            = felec;
1797
1798             /* Update vectorial force */
1799             fix2             = _mm_macc_pd(dx22,fscal,fix2);
1800             fiy2             = _mm_macc_pd(dy22,fscal,fiy2);
1801             fiz2             = _mm_macc_pd(dz22,fscal,fiz2);
1802
1803             fjx2             = _mm_macc_pd(dx22,fscal,fjx2);
1804             fjy2             = _mm_macc_pd(dy22,fscal,fjy2);
1805             fjz2             = _mm_macc_pd(dz22,fscal,fjz2);
1806
1807             gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1808
1809             /* Inner loop uses 372 flops */
1810         }
1811
1812         if(jidx<j_index_end)
1813         {
1814
1815             jnrA             = jjnr[jidx];
1816             j_coord_offsetA  = DIM*jnrA;
1817
1818             /* load j atom coordinates */
1819             gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
1820                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1821
1822             /* Calculate displacement vector */
1823             dx00             = _mm_sub_pd(ix0,jx0);
1824             dy00             = _mm_sub_pd(iy0,jy0);
1825             dz00             = _mm_sub_pd(iz0,jz0);
1826             dx01             = _mm_sub_pd(ix0,jx1);
1827             dy01             = _mm_sub_pd(iy0,jy1);
1828             dz01             = _mm_sub_pd(iz0,jz1);
1829             dx02             = _mm_sub_pd(ix0,jx2);
1830             dy02             = _mm_sub_pd(iy0,jy2);
1831             dz02             = _mm_sub_pd(iz0,jz2);
1832             dx10             = _mm_sub_pd(ix1,jx0);
1833             dy10             = _mm_sub_pd(iy1,jy0);
1834             dz10             = _mm_sub_pd(iz1,jz0);
1835             dx11             = _mm_sub_pd(ix1,jx1);
1836             dy11             = _mm_sub_pd(iy1,jy1);
1837             dz11             = _mm_sub_pd(iz1,jz1);
1838             dx12             = _mm_sub_pd(ix1,jx2);
1839             dy12             = _mm_sub_pd(iy1,jy2);
1840             dz12             = _mm_sub_pd(iz1,jz2);
1841             dx20             = _mm_sub_pd(ix2,jx0);
1842             dy20             = _mm_sub_pd(iy2,jy0);
1843             dz20             = _mm_sub_pd(iz2,jz0);
1844             dx21             = _mm_sub_pd(ix2,jx1);
1845             dy21             = _mm_sub_pd(iy2,jy1);
1846             dz21             = _mm_sub_pd(iz2,jz1);
1847             dx22             = _mm_sub_pd(ix2,jx2);
1848             dy22             = _mm_sub_pd(iy2,jy2);
1849             dz22             = _mm_sub_pd(iz2,jz2);
1850
1851             /* Calculate squared distance and things based on it */
1852             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1853             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
1854             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
1855             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
1856             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1857             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1858             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
1859             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1860             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1861
1862             rinv00           = avx128fma_invsqrt_d(rsq00);
1863             rinv01           = avx128fma_invsqrt_d(rsq01);
1864             rinv02           = avx128fma_invsqrt_d(rsq02);
1865             rinv10           = avx128fma_invsqrt_d(rsq10);
1866             rinv11           = avx128fma_invsqrt_d(rsq11);
1867             rinv12           = avx128fma_invsqrt_d(rsq12);
1868             rinv20           = avx128fma_invsqrt_d(rsq20);
1869             rinv21           = avx128fma_invsqrt_d(rsq21);
1870             rinv22           = avx128fma_invsqrt_d(rsq22);
1871
1872             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
1873             rinvsq01         = _mm_mul_pd(rinv01,rinv01);
1874             rinvsq02         = _mm_mul_pd(rinv02,rinv02);
1875             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
1876             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
1877             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
1878             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
1879             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
1880             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
1881
1882             fjx0             = _mm_setzero_pd();
1883             fjy0             = _mm_setzero_pd();
1884             fjz0             = _mm_setzero_pd();
1885             fjx1             = _mm_setzero_pd();
1886             fjy1             = _mm_setzero_pd();
1887             fjz1             = _mm_setzero_pd();
1888             fjx2             = _mm_setzero_pd();
1889             fjy2             = _mm_setzero_pd();
1890             fjz2             = _mm_setzero_pd();
1891
1892             /**************************
1893              * CALCULATE INTERACTIONS *
1894              **************************/
1895
1896             r00              = _mm_mul_pd(rsq00,rinv00);
1897
1898             /* EWALD ELECTROSTATICS */
1899
1900             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1901             ewrt             = _mm_mul_pd(r00,ewtabscale);
1902             ewitab           = _mm_cvttpd_epi32(ewrt);
1903 #ifdef __XOP__
1904             eweps            = _mm_frcz_pd(ewrt);
1905 #else
1906             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1907 #endif
1908             twoeweps         = _mm_add_pd(eweps,eweps);
1909             gmx_mm_load_1pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1910             felec            = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
1911             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
1912
1913             /* Analytical LJ-PME */
1914             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1915             ewcljrsq         = _mm_mul_pd(ewclj2,rsq00);
1916             ewclj6           = _mm_mul_pd(ewclj2,_mm_mul_pd(ewclj2,ewclj2));
1917             exponent         = avx128fma_exp_d(ewcljrsq);
1918             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
1919             poly             = _mm_mul_pd(exponent,_mm_macc_pd(_mm_mul_pd(ewcljrsq,ewcljrsq),one_half,_mm_sub_pd(one,ewcljrsq)));
1920             /* f6A = 6 * C6grid * (1 - poly) */
1921             f6A              = _mm_mul_pd(c6grid_00,_mm_sub_pd(one,poly));
1922             /* f6B = C6grid * exponent * beta^6 */
1923             f6B              = _mm_mul_pd(_mm_mul_pd(c6grid_00,one_sixth),_mm_mul_pd(exponent,ewclj6));
1924             /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
1925             fvdw              = _mm_mul_pd(_mm_macc_pd(_mm_msub_pd(c12_00,rinvsix,_mm_sub_pd(c6_00,f6A)),rinvsix,f6B),rinvsq00);
1926
1927             fscal            = _mm_add_pd(felec,fvdw);
1928
1929             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1930
1931             /* Update vectorial force */
1932             fix0             = _mm_macc_pd(dx00,fscal,fix0);
1933             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
1934             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
1935
1936             fjx0             = _mm_macc_pd(dx00,fscal,fjx0);
1937             fjy0             = _mm_macc_pd(dy00,fscal,fjy0);
1938             fjz0             = _mm_macc_pd(dz00,fscal,fjz0);
1939
1940             /**************************
1941              * CALCULATE INTERACTIONS *
1942              **************************/
1943
1944             r01              = _mm_mul_pd(rsq01,rinv01);
1945
1946             /* EWALD ELECTROSTATICS */
1947
1948             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1949             ewrt             = _mm_mul_pd(r01,ewtabscale);
1950             ewitab           = _mm_cvttpd_epi32(ewrt);
1951 #ifdef __XOP__
1952             eweps            = _mm_frcz_pd(ewrt);
1953 #else
1954             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1955 #endif
1956             twoeweps         = _mm_add_pd(eweps,eweps);
1957             gmx_mm_load_1pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1958             felec            = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
1959             felec            = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
1960
1961             fscal            = felec;
1962
1963             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1964
1965             /* Update vectorial force */
1966             fix0             = _mm_macc_pd(dx01,fscal,fix0);
1967             fiy0             = _mm_macc_pd(dy01,fscal,fiy0);
1968             fiz0             = _mm_macc_pd(dz01,fscal,fiz0);
1969
1970             fjx1             = _mm_macc_pd(dx01,fscal,fjx1);
1971             fjy1             = _mm_macc_pd(dy01,fscal,fjy1);
1972             fjz1             = _mm_macc_pd(dz01,fscal,fjz1);
1973
1974             /**************************
1975              * CALCULATE INTERACTIONS *
1976              **************************/
1977
1978             r02              = _mm_mul_pd(rsq02,rinv02);
1979
1980             /* EWALD ELECTROSTATICS */
1981
1982             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1983             ewrt             = _mm_mul_pd(r02,ewtabscale);
1984             ewitab           = _mm_cvttpd_epi32(ewrt);
1985 #ifdef __XOP__
1986             eweps            = _mm_frcz_pd(ewrt);
1987 #else
1988             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1989 #endif
1990             twoeweps         = _mm_add_pd(eweps,eweps);
1991             gmx_mm_load_1pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1992             felec            = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
1993             felec            = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
1994
1995             fscal            = felec;
1996
1997             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1998
1999             /* Update vectorial force */
2000             fix0             = _mm_macc_pd(dx02,fscal,fix0);
2001             fiy0             = _mm_macc_pd(dy02,fscal,fiy0);
2002             fiz0             = _mm_macc_pd(dz02,fscal,fiz0);
2003
2004             fjx2             = _mm_macc_pd(dx02,fscal,fjx2);
2005             fjy2             = _mm_macc_pd(dy02,fscal,fjy2);
2006             fjz2             = _mm_macc_pd(dz02,fscal,fjz2);
2007
2008             /**************************
2009              * CALCULATE INTERACTIONS *
2010              **************************/
2011
2012             r10              = _mm_mul_pd(rsq10,rinv10);
2013
2014             /* EWALD ELECTROSTATICS */
2015
2016             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2017             ewrt             = _mm_mul_pd(r10,ewtabscale);
2018             ewitab           = _mm_cvttpd_epi32(ewrt);
2019 #ifdef __XOP__
2020             eweps            = _mm_frcz_pd(ewrt);
2021 #else
2022             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2023 #endif
2024             twoeweps         = _mm_add_pd(eweps,eweps);
2025             gmx_mm_load_1pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2026             felec            = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
2027             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
2028
2029             fscal            = felec;
2030
2031             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2032
2033             /* Update vectorial force */
2034             fix1             = _mm_macc_pd(dx10,fscal,fix1);
2035             fiy1             = _mm_macc_pd(dy10,fscal,fiy1);
2036             fiz1             = _mm_macc_pd(dz10,fscal,fiz1);
2037
2038             fjx0             = _mm_macc_pd(dx10,fscal,fjx0);
2039             fjy0             = _mm_macc_pd(dy10,fscal,fjy0);
2040             fjz0             = _mm_macc_pd(dz10,fscal,fjz0);
2041
2042             /**************************
2043              * CALCULATE INTERACTIONS *
2044              **************************/
2045
2046             r11              = _mm_mul_pd(rsq11,rinv11);
2047
2048             /* EWALD ELECTROSTATICS */
2049
2050             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2051             ewrt             = _mm_mul_pd(r11,ewtabscale);
2052             ewitab           = _mm_cvttpd_epi32(ewrt);
2053 #ifdef __XOP__
2054             eweps            = _mm_frcz_pd(ewrt);
2055 #else
2056             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2057 #endif
2058             twoeweps         = _mm_add_pd(eweps,eweps);
2059             gmx_mm_load_1pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2060             felec            = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
2061             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
2062
2063             fscal            = felec;
2064
2065             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2066
2067             /* Update vectorial force */
2068             fix1             = _mm_macc_pd(dx11,fscal,fix1);
2069             fiy1             = _mm_macc_pd(dy11,fscal,fiy1);
2070             fiz1             = _mm_macc_pd(dz11,fscal,fiz1);
2071
2072             fjx1             = _mm_macc_pd(dx11,fscal,fjx1);
2073             fjy1             = _mm_macc_pd(dy11,fscal,fjy1);
2074             fjz1             = _mm_macc_pd(dz11,fscal,fjz1);
2075
2076             /**************************
2077              * CALCULATE INTERACTIONS *
2078              **************************/
2079
2080             r12              = _mm_mul_pd(rsq12,rinv12);
2081
2082             /* EWALD ELECTROSTATICS */
2083
2084             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2085             ewrt             = _mm_mul_pd(r12,ewtabscale);
2086             ewitab           = _mm_cvttpd_epi32(ewrt);
2087 #ifdef __XOP__
2088             eweps            = _mm_frcz_pd(ewrt);
2089 #else
2090             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2091 #endif
2092             twoeweps         = _mm_add_pd(eweps,eweps);
2093             gmx_mm_load_1pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2094             felec            = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
2095             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
2096
2097             fscal            = felec;
2098
2099             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2100
2101             /* Update vectorial force */
2102             fix1             = _mm_macc_pd(dx12,fscal,fix1);
2103             fiy1             = _mm_macc_pd(dy12,fscal,fiy1);
2104             fiz1             = _mm_macc_pd(dz12,fscal,fiz1);
2105
2106             fjx2             = _mm_macc_pd(dx12,fscal,fjx2);
2107             fjy2             = _mm_macc_pd(dy12,fscal,fjy2);
2108             fjz2             = _mm_macc_pd(dz12,fscal,fjz2);
2109
2110             /**************************
2111              * CALCULATE INTERACTIONS *
2112              **************************/
2113
2114             r20              = _mm_mul_pd(rsq20,rinv20);
2115
2116             /* EWALD ELECTROSTATICS */
2117
2118             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2119             ewrt             = _mm_mul_pd(r20,ewtabscale);
2120             ewitab           = _mm_cvttpd_epi32(ewrt);
2121 #ifdef __XOP__
2122             eweps            = _mm_frcz_pd(ewrt);
2123 #else
2124             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2125 #endif
2126             twoeweps         = _mm_add_pd(eweps,eweps);
2127             gmx_mm_load_1pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2128             felec            = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
2129             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
2130
2131             fscal            = felec;
2132
2133             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2134
2135             /* Update vectorial force */
2136             fix2             = _mm_macc_pd(dx20,fscal,fix2);
2137             fiy2             = _mm_macc_pd(dy20,fscal,fiy2);
2138             fiz2             = _mm_macc_pd(dz20,fscal,fiz2);
2139
2140             fjx0             = _mm_macc_pd(dx20,fscal,fjx0);
2141             fjy0             = _mm_macc_pd(dy20,fscal,fjy0);
2142             fjz0             = _mm_macc_pd(dz20,fscal,fjz0);
2143
2144             /**************************
2145              * CALCULATE INTERACTIONS *
2146              **************************/
2147
2148             r21              = _mm_mul_pd(rsq21,rinv21);
2149
2150             /* EWALD ELECTROSTATICS */
2151
2152             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2153             ewrt             = _mm_mul_pd(r21,ewtabscale);
2154             ewitab           = _mm_cvttpd_epi32(ewrt);
2155 #ifdef __XOP__
2156             eweps            = _mm_frcz_pd(ewrt);
2157 #else
2158             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2159 #endif
2160             twoeweps         = _mm_add_pd(eweps,eweps);
2161             gmx_mm_load_1pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2162             felec            = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
2163             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
2164
2165             fscal            = felec;
2166
2167             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2168
2169             /* Update vectorial force */
2170             fix2             = _mm_macc_pd(dx21,fscal,fix2);
2171             fiy2             = _mm_macc_pd(dy21,fscal,fiy2);
2172             fiz2             = _mm_macc_pd(dz21,fscal,fiz2);
2173
2174             fjx1             = _mm_macc_pd(dx21,fscal,fjx1);
2175             fjy1             = _mm_macc_pd(dy21,fscal,fjy1);
2176             fjz1             = _mm_macc_pd(dz21,fscal,fjz1);
2177
2178             /**************************
2179              * CALCULATE INTERACTIONS *
2180              **************************/
2181
2182             r22              = _mm_mul_pd(rsq22,rinv22);
2183
2184             /* EWALD ELECTROSTATICS */
2185
2186             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2187             ewrt             = _mm_mul_pd(r22,ewtabscale);
2188             ewitab           = _mm_cvttpd_epi32(ewrt);
2189 #ifdef __XOP__
2190             eweps            = _mm_frcz_pd(ewrt);
2191 #else
2192             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2193 #endif
2194             twoeweps         = _mm_add_pd(eweps,eweps);
2195             gmx_mm_load_1pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2196             felec            = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
2197             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
2198
2199             fscal            = felec;
2200
2201             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2202
2203             /* Update vectorial force */
2204             fix2             = _mm_macc_pd(dx22,fscal,fix2);
2205             fiy2             = _mm_macc_pd(dy22,fscal,fiy2);
2206             fiz2             = _mm_macc_pd(dz22,fscal,fiz2);
2207
2208             fjx2             = _mm_macc_pd(dx22,fscal,fjx2);
2209             fjy2             = _mm_macc_pd(dy22,fscal,fjy2);
2210             fjz2             = _mm_macc_pd(dz22,fscal,fjz2);
2211
2212             gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
2213
2214             /* Inner loop uses 372 flops */
2215         }
2216
2217         /* End of innermost loop */
2218
2219         gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
2220                                               f+i_coord_offset,fshift+i_shift_offset);
2221
2222         /* Increment number of inner iterations */
2223         inneriter                  += j_index_end - j_index_start;
2224
2225         /* Outer loop uses 18 flops */
2226     }
2227
2228     /* Increment number of outer iterations */
2229     outeriter        += nri;
2230
2231     /* Update outer/inner flops */
2232
2233     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_F,outeriter*18 + inneriter*372);
2234 }