src/gromacs/gmxlib/nonbonded/nb_kernel_sse4_1_double/nb_kernel_ElecEw_VdwNone_GeomP1P1_sse4_1_double.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS sse4_1_double kernel generator.
  37  */
  38 #include "config.h"
  39
  40 #include <math.h>
  41
  42 #include "../nb_kernel.h"
  43 #include "types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "nrnb.h"
  46
  47 #include "gromacs/simd/math_x86_sse4_1_double.h"
  48 #include "kernelutil_x86_sse4_1_double.h"
  49
  50 /*
  51  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwNone_GeomP1P1_VF_sse4_1_double
  52  * Electrostatics interaction: Ewald
  53  * VdW interaction:            None
  54  * Geometry:                   Particle-Particle
  55  * Calculate force/pot:        PotentialAndForce
  56  */
  57 void
  58 nb_kernel_ElecEw_VdwNone_GeomP1P1_VF_sse4_1_double
  59                     (t_nblist                    * gmx_restrict       nlist,
  60                      rvec                        * gmx_restrict          xx,
  61                      rvec                        * gmx_restrict          ff,
  62                      t_forcerec                  * gmx_restrict          fr,
  63                      t_mdatoms                   * gmx_restrict     mdatoms,
  64                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  65                      t_nrnb                      * gmx_restrict        nrnb)
  66 {
  67     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  68      * just 0 for non-waters.
  69      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
  70      * jnr indices corresponding to data put in the four positions in the SIMD register.
  71      */
  72     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  73     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  74     int              jnrA,jnrB;
  75     int              j_coord_offsetA,j_coord_offsetB;
  76     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  77     real             rcutoff_scalar;
  78     real             *shiftvec,*fshift,*x,*f;
  79     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  80     int              vdwioffset0;
  81     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  82     int              vdwjidx0A,vdwjidx0B;
  83     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  84     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  85     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
  86     real             *charge;
  87     __m128i          ewitab;
  88     __m128d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
  89     real             *ewtab;
  90     __m128d          dummy_mask,cutoff_mask;
  91     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
  92     __m128d          one     = _mm_set1_pd(1.0);
  93     __m128d          two     = _mm_set1_pd(2.0);
  94     x                = xx[0];
  95     f                = ff[0];
  96
  97     nri              = nlist->nri;
  98     iinr             = nlist->iinr;
  99     jindex           = nlist->jindex;
 100     jjnr             = nlist->jjnr;
 101     shiftidx         = nlist->shift;
 102     gid              = nlist->gid;
 103     shiftvec         = fr->shift_vec[0];
 104     fshift           = fr->fshift[0];
 105     facel            = _mm_set1_pd(fr->epsfac);
 106     charge           = mdatoms->chargeA;
 107
 108     sh_ewald         = _mm_set1_pd(fr->ic->sh_ewald);
 109     ewtab            = fr->ic->tabq_coul_FDV0;
 110     ewtabscale       = _mm_set1_pd(fr->ic->tabq_scale);
 111     ewtabhalfspace   = _mm_set1_pd(0.5/fr->ic->tabq_scale);
 112
 113     /* Avoid stupid compiler warnings */
 114     jnrA = jnrB = 0;
 115     j_coord_offsetA = 0;
 116     j_coord_offsetB = 0;
 117
 118     outeriter        = 0;
 119     inneriter        = 0;
 120
 121     /* Start outer loop over neighborlists */
 122     for(iidx=0; iidx<nri; iidx++)
 123     {
 124         /* Load shift vector for this list */
 125         i_shift_offset   = DIM*shiftidx[iidx];
 126
 127         /* Load limits for loop over neighbors */
 128         j_index_start    = jindex[iidx];
 129         j_index_end      = jindex[iidx+1];
 130
 131         /* Get outer coordinate index */
 132         inr              = iinr[iidx];
 133         i_coord_offset   = DIM*inr;
 134
 135         /* Load i particle coords and add shift vector */
 136         gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 137
 138         fix0             = _mm_setzero_pd();
 139         fiy0             = _mm_setzero_pd();
 140         fiz0             = _mm_setzero_pd();
 141
 142         /* Load parameters for i particles */
 143         iq0              = _mm_mul_pd(facel,_mm_load1_pd(charge+inr+0));
 144
 145         /* Reset potential sums */
 146         velecsum         = _mm_setzero_pd();
 147
 148         /* Start inner kernel loop */
 149         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 150         {
 151
 152             /* Get j neighbor index, and coordinate index */
 153             jnrA             = jjnr[jidx];
 154             jnrB             = jjnr[jidx+1];
 155             j_coord_offsetA  = DIM*jnrA;
 156             j_coord_offsetB  = DIM*jnrB;
 157
 158             /* load j atom coordinates */
 159             gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 160                                               &jx0,&jy0,&jz0);
 161
 162             /* Calculate displacement vector */
 163             dx00             = _mm_sub_pd(ix0,jx0);
 164             dy00             = _mm_sub_pd(iy0,jy0);
 165             dz00             = _mm_sub_pd(iz0,jz0);
 166
 167             /* Calculate squared distance and things based on it */
 168             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 169
 170             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 171
 172             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 173
 174             /* Load parameters for j particles */
 175             jq0              = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
 176
 177             /**************************
 178              * CALCULATE INTERACTIONS *
 179              **************************/
 180
 181             r00              = _mm_mul_pd(rsq00,rinv00);
 182
 183             /* Compute parameters for interactions between i and j atoms */
 184             qq00             = _mm_mul_pd(iq0,jq0);
 185
 186             /* EWALD ELECTROSTATICS */
 187
 188             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 189             ewrt             = _mm_mul_pd(r00,ewtabscale);
 190             ewitab           = _mm_cvttpd_epi32(ewrt);
 191             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 192             ewitab           = _mm_slli_epi32(ewitab,2);
 193             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 194             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 195             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 196             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 197             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 198             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 199             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 200             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 201             velec            = _mm_mul_pd(qq00,_mm_sub_pd(rinv00,velec));
 202             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
 203
 204             /* Update potential sum for this i atom from the interaction with this j atom. */
 205             velecsum         = _mm_add_pd(velecsum,velec);
 206
 207             fscal            = felec;
 208
 209             /* Calculate temporary vectorial force */
 210             tx               = _mm_mul_pd(fscal,dx00);
 211             ty               = _mm_mul_pd(fscal,dy00);
 212             tz               = _mm_mul_pd(fscal,dz00);
 213
 214             /* Update vectorial force */
 215             fix0             = _mm_add_pd(fix0,tx);
 216             fiy0             = _mm_add_pd(fiy0,ty);
 217             fiz0             = _mm_add_pd(fiz0,tz);
 218
 219             gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,tx,ty,tz);
 220
 221             /* Inner loop uses 41 flops */
 222         }
 223
 224         if(jidx<j_index_end)
 225         {
 226
 227             jnrA             = jjnr[jidx];
 228             j_coord_offsetA  = DIM*jnrA;
 229
 230             /* load j atom coordinates */
 231             gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 232                                               &jx0,&jy0,&jz0);
 233
 234             /* Calculate displacement vector */
 235             dx00             = _mm_sub_pd(ix0,jx0);
 236             dy00             = _mm_sub_pd(iy0,jy0);
 237             dz00             = _mm_sub_pd(iz0,jz0);
 238
 239             /* Calculate squared distance and things based on it */
 240             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 241
 242             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 243
 244             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 245
 246             /* Load parameters for j particles */
 247             jq0              = _mm_load_sd(charge+jnrA+0);
 248
 249             /**************************
 250              * CALCULATE INTERACTIONS *
 251              **************************/
 252
 253             r00              = _mm_mul_pd(rsq00,rinv00);
 254
 255             /* Compute parameters for interactions between i and j atoms */
 256             qq00             = _mm_mul_pd(iq0,jq0);
 257
 258             /* EWALD ELECTROSTATICS */
 259
 260             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 261             ewrt             = _mm_mul_pd(r00,ewtabscale);
 262             ewitab           = _mm_cvttpd_epi32(ewrt);
 263             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 264             ewitab           = _mm_slli_epi32(ewitab,2);
 265             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 266             ewtabD           = _mm_setzero_pd();
 267             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 268             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 269             ewtabFn          = _mm_setzero_pd();
 270             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 271             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 272             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 273             velec            = _mm_mul_pd(qq00,_mm_sub_pd(rinv00,velec));
 274             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
 275
 276             /* Update potential sum for this i atom from the interaction with this j atom. */
 277             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 278             velecsum         = _mm_add_pd(velecsum,velec);
 279
 280             fscal            = felec;
 281
 282             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 283
 284             /* Calculate temporary vectorial force */
 285             tx               = _mm_mul_pd(fscal,dx00);
 286             ty               = _mm_mul_pd(fscal,dy00);
 287             tz               = _mm_mul_pd(fscal,dz00);
 288
 289             /* Update vectorial force */
 290             fix0             = _mm_add_pd(fix0,tx);
 291             fiy0             = _mm_add_pd(fiy0,ty);
 292             fiz0             = _mm_add_pd(fiz0,tz);
 293
 294             gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,tx,ty,tz);
 295
 296             /* Inner loop uses 41 flops */
 297         }
 298
 299         /* End of innermost loop */
 300
 301         gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
 302                                               f+i_coord_offset,fshift+i_shift_offset);
 303
 304         ggid                        = gid[iidx];
 305         /* Update potential energies */
 306         gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
 307
 308         /* Increment number of inner iterations */
 309         inneriter                  += j_index_end - j_index_start;
 310
 311         /* Outer loop uses 8 flops */
 312     }
 313
 314     /* Increment number of outer iterations */
 315     outeriter        += nri;
 316
 317     /* Update outer/inner flops */
 318
 319     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VF,outeriter*8 + inneriter*41);
 320 }
 321 /*
 322  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwNone_GeomP1P1_F_sse4_1_double
 323  * Electrostatics interaction: Ewald
 324  * VdW interaction:            None
 325  * Geometry:                   Particle-Particle
 326  * Calculate force/pot:        Force
 327  */
 328 void
 329 nb_kernel_ElecEw_VdwNone_GeomP1P1_F_sse4_1_double
 330                     (t_nblist                    * gmx_restrict       nlist,
 331                      rvec                        * gmx_restrict          xx,
 332                      rvec                        * gmx_restrict          ff,
 333                      t_forcerec                  * gmx_restrict          fr,
 334                      t_mdatoms                   * gmx_restrict     mdatoms,
 335                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 336                      t_nrnb                      * gmx_restrict        nrnb)
 337 {
 338     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 339      * just 0 for non-waters.
 340      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
 341      * jnr indices corresponding to data put in the four positions in the SIMD register.
 342      */
 343     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 344     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 345     int              jnrA,jnrB;
 346     int              j_coord_offsetA,j_coord_offsetB;
 347     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 348     real             rcutoff_scalar;
 349     real             *shiftvec,*fshift,*x,*f;
 350     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 351     int              vdwioffset0;
 352     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 353     int              vdwjidx0A,vdwjidx0B;
 354     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 355     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 356     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
 357     real             *charge;
 358     __m128i          ewitab;
 359     __m128d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 360     real             *ewtab;
 361     __m128d          dummy_mask,cutoff_mask;
 362     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 363     __m128d          one     = _mm_set1_pd(1.0);
 364     __m128d          two     = _mm_set1_pd(2.0);
 365     x                = xx[0];
 366     f                = ff[0];
 367
 368     nri              = nlist->nri;
 369     iinr             = nlist->iinr;
 370     jindex           = nlist->jindex;
 371     jjnr             = nlist->jjnr;
 372     shiftidx         = nlist->shift;
 373     gid              = nlist->gid;
 374     shiftvec         = fr->shift_vec[0];
 375     fshift           = fr->fshift[0];
 376     facel            = _mm_set1_pd(fr->epsfac);
 377     charge           = mdatoms->chargeA;
 378
 379     sh_ewald         = _mm_set1_pd(fr->ic->sh_ewald);
 380     ewtab            = fr->ic->tabq_coul_F;
 381     ewtabscale       = _mm_set1_pd(fr->ic->tabq_scale);
 382     ewtabhalfspace   = _mm_set1_pd(0.5/fr->ic->tabq_scale);
 383
 384     /* Avoid stupid compiler warnings */
 385     jnrA = jnrB = 0;
 386     j_coord_offsetA = 0;
 387     j_coord_offsetB = 0;
 388
 389     outeriter        = 0;
 390     inneriter        = 0;
 391
 392     /* Start outer loop over neighborlists */
 393     for(iidx=0; iidx<nri; iidx++)
 394     {
 395         /* Load shift vector for this list */
 396         i_shift_offset   = DIM*shiftidx[iidx];
 397
 398         /* Load limits for loop over neighbors */
 399         j_index_start    = jindex[iidx];
 400         j_index_end      = jindex[iidx+1];
 401
 402         /* Get outer coordinate index */
 403         inr              = iinr[iidx];
 404         i_coord_offset   = DIM*inr;
 405
 406         /* Load i particle coords and add shift vector */
 407         gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 408
 409         fix0             = _mm_setzero_pd();
 410         fiy0             = _mm_setzero_pd();
 411         fiz0             = _mm_setzero_pd();
 412
 413         /* Load parameters for i particles */
 414         iq0              = _mm_mul_pd(facel,_mm_load1_pd(charge+inr+0));
 415
 416         /* Start inner kernel loop */
 417         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 418         {
 419
 420             /* Get j neighbor index, and coordinate index */
 421             jnrA             = jjnr[jidx];
 422             jnrB             = jjnr[jidx+1];
 423             j_coord_offsetA  = DIM*jnrA;
 424             j_coord_offsetB  = DIM*jnrB;
 425
 426             /* load j atom coordinates */
 427             gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 428                                               &jx0,&jy0,&jz0);
 429
 430             /* Calculate displacement vector */
 431             dx00             = _mm_sub_pd(ix0,jx0);
 432             dy00             = _mm_sub_pd(iy0,jy0);
 433             dz00             = _mm_sub_pd(iz0,jz0);
 434
 435             /* Calculate squared distance and things based on it */
 436             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 437
 438             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 439
 440             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 441
 442             /* Load parameters for j particles */
 443             jq0              = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
 444
 445             /**************************
 446              * CALCULATE INTERACTIONS *
 447              **************************/
 448
 449             r00              = _mm_mul_pd(rsq00,rinv00);
 450
 451             /* Compute parameters for interactions between i and j atoms */
 452             qq00             = _mm_mul_pd(iq0,jq0);
 453
 454             /* EWALD ELECTROSTATICS */
 455
 456             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 457             ewrt             = _mm_mul_pd(r00,ewtabscale);
 458             ewitab           = _mm_cvttpd_epi32(ewrt);
 459             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 460             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
 461                                          &ewtabF,&ewtabFn);
 462             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
 463             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
 464
 465             fscal            = felec;
 466
 467             /* Calculate temporary vectorial force */
 468             tx               = _mm_mul_pd(fscal,dx00);
 469             ty               = _mm_mul_pd(fscal,dy00);
 470             tz               = _mm_mul_pd(fscal,dz00);
 471
 472             /* Update vectorial force */
 473             fix0             = _mm_add_pd(fix0,tx);
 474             fiy0             = _mm_add_pd(fiy0,ty);
 475             fiz0             = _mm_add_pd(fiz0,tz);
 476
 477             gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,tx,ty,tz);
 478
 479             /* Inner loop uses 36 flops */
 480         }
 481
 482         if(jidx<j_index_end)
 483         {
 484
 485             jnrA             = jjnr[jidx];
 486             j_coord_offsetA  = DIM*jnrA;
 487
 488             /* load j atom coordinates */
 489             gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 490                                               &jx0,&jy0,&jz0);
 491
 492             /* Calculate displacement vector */
 493             dx00             = _mm_sub_pd(ix0,jx0);
 494             dy00             = _mm_sub_pd(iy0,jy0);
 495             dz00             = _mm_sub_pd(iz0,jz0);
 496
 497             /* Calculate squared distance and things based on it */
 498             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 499
 500             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 501
 502             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 503
 504             /* Load parameters for j particles */
 505             jq0              = _mm_load_sd(charge+jnrA+0);
 506
 507             /**************************
 508              * CALCULATE INTERACTIONS *
 509              **************************/
 510
 511             r00              = _mm_mul_pd(rsq00,rinv00);
 512
 513             /* Compute parameters for interactions between i and j atoms */
 514             qq00             = _mm_mul_pd(iq0,jq0);
 515
 516             /* EWALD ELECTROSTATICS */
 517
 518             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 519             ewrt             = _mm_mul_pd(r00,ewtabscale);
 520             ewitab           = _mm_cvttpd_epi32(ewrt);
 521             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 522             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
 523             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
 524             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
 525
 526             fscal            = felec;
 527
 528             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 529
 530             /* Calculate temporary vectorial force */
 531             tx               = _mm_mul_pd(fscal,dx00);
 532             ty               = _mm_mul_pd(fscal,dy00);
 533             tz               = _mm_mul_pd(fscal,dz00);
 534
 535             /* Update vectorial force */
 536             fix0             = _mm_add_pd(fix0,tx);
 537             fiy0             = _mm_add_pd(fiy0,ty);
 538             fiz0             = _mm_add_pd(fiz0,tz);
 539
 540             gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,tx,ty,tz);
 541
 542             /* Inner loop uses 36 flops */
 543         }
 544
 545         /* End of innermost loop */
 546
 547         gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
 548                                               f+i_coord_offset,fshift+i_shift_offset);
 549
 550         /* Increment number of inner iterations */
 551         inneriter                  += j_index_end - j_index_start;
 552
 553         /* Outer loop uses 7 flops */
 554     }
 555
 556     /* Increment number of outer iterations */
 557     outeriter        += nri;
 558
 559     /* Update outer/inner flops */
 560
 561     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_F,outeriter*7 + inneriter*36);
 562 }