src/gromacs/gmxlib/nonbonded/nb_kernel_sse4_1_double/nb_kernel_ElecRFCut_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_ElecRFCut_VdwNone_GeomP1P1_VF_sse4_1_double
  52  * Electrostatics interaction: ReactionField
  53  * VdW interaction:            None
  54  * Geometry:                   Particle-Particle
  55  * Calculate force/pot:        PotentialAndForce
  56  */
  57 void
  58 nb_kernel_ElecRFCut_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     __m128d          dummy_mask,cutoff_mask;
  88     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
  89     __m128d          one     = _mm_set1_pd(1.0);
  90     __m128d          two     = _mm_set1_pd(2.0);
  91     x                = xx[0];
  92     f                = ff[0];
  93
  94     nri              = nlist->nri;
  95     iinr             = nlist->iinr;
  96     jindex           = nlist->jindex;
  97     jjnr             = nlist->jjnr;
  98     shiftidx         = nlist->shift;
  99     gid              = nlist->gid;
 100     shiftvec         = fr->shift_vec[0];
 101     fshift           = fr->fshift[0];
 102     facel            = _mm_set1_pd(fr->epsfac);
 103     charge           = mdatoms->chargeA;
 104     krf              = _mm_set1_pd(fr->ic->k_rf);
 105     krf2             = _mm_set1_pd(fr->ic->k_rf*2.0);
 106     crf              = _mm_set1_pd(fr->ic->c_rf);
 107
 108     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 109     rcutoff_scalar   = fr->rcoulomb;
 110     rcutoff          = _mm_set1_pd(rcutoff_scalar);
 111     rcutoff2         = _mm_mul_pd(rcutoff,rcutoff);
 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             if (gmx_mm_any_lt(rsq00,rcutoff2))
 182             {
 183
 184             /* Compute parameters for interactions between i and j atoms */
 185             qq00             = _mm_mul_pd(iq0,jq0);
 186
 187             /* REACTION-FIELD ELECTROSTATICS */
 188             velec            = _mm_mul_pd(qq00,_mm_sub_pd(_mm_add_pd(rinv00,_mm_mul_pd(krf,rsq00)),crf));
 189             felec            = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
 190
 191             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
 192
 193             /* Update potential sum for this i atom from the interaction with this j atom. */
 194             velec            = _mm_and_pd(velec,cutoff_mask);
 195             velecsum         = _mm_add_pd(velecsum,velec);
 196
 197             fscal            = felec;
 198
 199             fscal            = _mm_and_pd(fscal,cutoff_mask);
 200
 201             /* Calculate temporary vectorial force */
 202             tx               = _mm_mul_pd(fscal,dx00);
 203             ty               = _mm_mul_pd(fscal,dy00);
 204             tz               = _mm_mul_pd(fscal,dz00);
 205
 206             /* Update vectorial force */
 207             fix0             = _mm_add_pd(fix0,tx);
 208             fiy0             = _mm_add_pd(fiy0,ty);
 209             fiz0             = _mm_add_pd(fiz0,tz);
 210
 211             gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,tx,ty,tz);
 212
 213             }
 214
 215             /* Inner loop uses 36 flops */
 216         }
 217
 218         if(jidx<j_index_end)
 219         {
 220
 221             jnrA             = jjnr[jidx];
 222             j_coord_offsetA  = DIM*jnrA;
 223
 224             /* load j atom coordinates */
 225             gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 226                                               &jx0,&jy0,&jz0);
 227
 228             /* Calculate displacement vector */
 229             dx00             = _mm_sub_pd(ix0,jx0);
 230             dy00             = _mm_sub_pd(iy0,jy0);
 231             dz00             = _mm_sub_pd(iz0,jz0);
 232
 233             /* Calculate squared distance and things based on it */
 234             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 235
 236             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 237
 238             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 239
 240             /* Load parameters for j particles */
 241             jq0              = _mm_load_sd(charge+jnrA+0);
 242
 243             /**************************
 244              * CALCULATE INTERACTIONS *
 245              **************************/
 246
 247             if (gmx_mm_any_lt(rsq00,rcutoff2))
 248             {
 249
 250             /* Compute parameters for interactions between i and j atoms */
 251             qq00             = _mm_mul_pd(iq0,jq0);
 252
 253             /* REACTION-FIELD ELECTROSTATICS */
 254             velec            = _mm_mul_pd(qq00,_mm_sub_pd(_mm_add_pd(rinv00,_mm_mul_pd(krf,rsq00)),crf));
 255             felec            = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
 256
 257             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
 258
 259             /* Update potential sum for this i atom from the interaction with this j atom. */
 260             velec            = _mm_and_pd(velec,cutoff_mask);
 261             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 262             velecsum         = _mm_add_pd(velecsum,velec);
 263
 264             fscal            = felec;
 265
 266             fscal            = _mm_and_pd(fscal,cutoff_mask);
 267
 268             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 269
 270             /* Calculate temporary vectorial force */
 271             tx               = _mm_mul_pd(fscal,dx00);
 272             ty               = _mm_mul_pd(fscal,dy00);
 273             tz               = _mm_mul_pd(fscal,dz00);
 274
 275             /* Update vectorial force */
 276             fix0             = _mm_add_pd(fix0,tx);
 277             fiy0             = _mm_add_pd(fiy0,ty);
 278             fiz0             = _mm_add_pd(fiz0,tz);
 279
 280             gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,tx,ty,tz);
 281
 282             }
 283
 284             /* Inner loop uses 36 flops */
 285         }
 286
 287         /* End of innermost loop */
 288
 289         gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
 290                                               f+i_coord_offset,fshift+i_shift_offset);
 291
 292         ggid                        = gid[iidx];
 293         /* Update potential energies */
 294         gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
 295
 296         /* Increment number of inner iterations */
 297         inneriter                  += j_index_end - j_index_start;
 298
 299         /* Outer loop uses 8 flops */
 300     }
 301
 302     /* Increment number of outer iterations */
 303     outeriter        += nri;
 304
 305     /* Update outer/inner flops */
 306
 307     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VF,outeriter*8 + inneriter*36);
 308 }
 309 /*
 310  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwNone_GeomP1P1_F_sse4_1_double
 311  * Electrostatics interaction: ReactionField
 312  * VdW interaction:            None
 313  * Geometry:                   Particle-Particle
 314  * Calculate force/pot:        Force
 315  */
 316 void
 317 nb_kernel_ElecRFCut_VdwNone_GeomP1P1_F_sse4_1_double
 318                     (t_nblist                    * gmx_restrict       nlist,
 319                      rvec                        * gmx_restrict          xx,
 320                      rvec                        * gmx_restrict          ff,
 321                      t_forcerec                  * gmx_restrict          fr,
 322                      t_mdatoms                   * gmx_restrict     mdatoms,
 323                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 324                      t_nrnb                      * gmx_restrict        nrnb)
 325 {
 326     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 327      * just 0 for non-waters.
 328      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
 329      * jnr indices corresponding to data put in the four positions in the SIMD register.
 330      */
 331     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 332     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 333     int              jnrA,jnrB;
 334     int              j_coord_offsetA,j_coord_offsetB;
 335     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 336     real             rcutoff_scalar;
 337     real             *shiftvec,*fshift,*x,*f;
 338     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 339     int              vdwioffset0;
 340     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 341     int              vdwjidx0A,vdwjidx0B;
 342     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 343     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 344     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
 345     real             *charge;
 346     __m128d          dummy_mask,cutoff_mask;
 347     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 348     __m128d          one     = _mm_set1_pd(1.0);
 349     __m128d          two     = _mm_set1_pd(2.0);
 350     x                = xx[0];
 351     f                = ff[0];
 352
 353     nri              = nlist->nri;
 354     iinr             = nlist->iinr;
 355     jindex           = nlist->jindex;
 356     jjnr             = nlist->jjnr;
 357     shiftidx         = nlist->shift;
 358     gid              = nlist->gid;
 359     shiftvec         = fr->shift_vec[0];
 360     fshift           = fr->fshift[0];
 361     facel            = _mm_set1_pd(fr->epsfac);
 362     charge           = mdatoms->chargeA;
 363     krf              = _mm_set1_pd(fr->ic->k_rf);
 364     krf2             = _mm_set1_pd(fr->ic->k_rf*2.0);
 365     crf              = _mm_set1_pd(fr->ic->c_rf);
 366
 367     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 368     rcutoff_scalar   = fr->rcoulomb;
 369     rcutoff          = _mm_set1_pd(rcutoff_scalar);
 370     rcutoff2         = _mm_mul_pd(rcutoff,rcutoff);
 371
 372     /* Avoid stupid compiler warnings */
 373     jnrA = jnrB = 0;
 374     j_coord_offsetA = 0;
 375     j_coord_offsetB = 0;
 376
 377     outeriter        = 0;
 378     inneriter        = 0;
 379
 380     /* Start outer loop over neighborlists */
 381     for(iidx=0; iidx<nri; iidx++)
 382     {
 383         /* Load shift vector for this list */
 384         i_shift_offset   = DIM*shiftidx[iidx];
 385
 386         /* Load limits for loop over neighbors */
 387         j_index_start    = jindex[iidx];
 388         j_index_end      = jindex[iidx+1];
 389
 390         /* Get outer coordinate index */
 391         inr              = iinr[iidx];
 392         i_coord_offset   = DIM*inr;
 393
 394         /* Load i particle coords and add shift vector */
 395         gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 396
 397         fix0             = _mm_setzero_pd();
 398         fiy0             = _mm_setzero_pd();
 399         fiz0             = _mm_setzero_pd();
 400
 401         /* Load parameters for i particles */
 402         iq0              = _mm_mul_pd(facel,_mm_load1_pd(charge+inr+0));
 403
 404         /* Start inner kernel loop */
 405         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 406         {
 407
 408             /* Get j neighbor index, and coordinate index */
 409             jnrA             = jjnr[jidx];
 410             jnrB             = jjnr[jidx+1];
 411             j_coord_offsetA  = DIM*jnrA;
 412             j_coord_offsetB  = DIM*jnrB;
 413
 414             /* load j atom coordinates */
 415             gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 416                                               &jx0,&jy0,&jz0);
 417
 418             /* Calculate displacement vector */
 419             dx00             = _mm_sub_pd(ix0,jx0);
 420             dy00             = _mm_sub_pd(iy0,jy0);
 421             dz00             = _mm_sub_pd(iz0,jz0);
 422
 423             /* Calculate squared distance and things based on it */
 424             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 425
 426             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 427
 428             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 429
 430             /* Load parameters for j particles */
 431             jq0              = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
 432
 433             /**************************
 434              * CALCULATE INTERACTIONS *
 435              **************************/
 436
 437             if (gmx_mm_any_lt(rsq00,rcutoff2))
 438             {
 439
 440             /* Compute parameters for interactions between i and j atoms */
 441             qq00             = _mm_mul_pd(iq0,jq0);
 442
 443             /* REACTION-FIELD ELECTROSTATICS */
 444             felec            = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
 445
 446             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
 447
 448             fscal            = felec;
 449
 450             fscal            = _mm_and_pd(fscal,cutoff_mask);
 451
 452             /* Calculate temporary vectorial force */
 453             tx               = _mm_mul_pd(fscal,dx00);
 454             ty               = _mm_mul_pd(fscal,dy00);
 455             tz               = _mm_mul_pd(fscal,dz00);
 456
 457             /* Update vectorial force */
 458             fix0             = _mm_add_pd(fix0,tx);
 459             fiy0             = _mm_add_pd(fiy0,ty);
 460             fiz0             = _mm_add_pd(fiz0,tz);
 461
 462             gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,tx,ty,tz);
 463
 464             }
 465
 466             /* Inner loop uses 30 flops */
 467         }
 468
 469         if(jidx<j_index_end)
 470         {
 471
 472             jnrA             = jjnr[jidx];
 473             j_coord_offsetA  = DIM*jnrA;
 474
 475             /* load j atom coordinates */
 476             gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 477                                               &jx0,&jy0,&jz0);
 478
 479             /* Calculate displacement vector */
 480             dx00             = _mm_sub_pd(ix0,jx0);
 481             dy00             = _mm_sub_pd(iy0,jy0);
 482             dz00             = _mm_sub_pd(iz0,jz0);
 483
 484             /* Calculate squared distance and things based on it */
 485             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 486
 487             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 488
 489             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 490
 491             /* Load parameters for j particles */
 492             jq0              = _mm_load_sd(charge+jnrA+0);
 493
 494             /**************************
 495              * CALCULATE INTERACTIONS *
 496              **************************/
 497
 498             if (gmx_mm_any_lt(rsq00,rcutoff2))
 499             {
 500
 501             /* Compute parameters for interactions between i and j atoms */
 502             qq00             = _mm_mul_pd(iq0,jq0);
 503
 504             /* REACTION-FIELD ELECTROSTATICS */
 505             felec            = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
 506
 507             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
 508
 509             fscal            = felec;
 510
 511             fscal            = _mm_and_pd(fscal,cutoff_mask);
 512
 513             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 514
 515             /* Calculate temporary vectorial force */
 516             tx               = _mm_mul_pd(fscal,dx00);
 517             ty               = _mm_mul_pd(fscal,dy00);
 518             tz               = _mm_mul_pd(fscal,dz00);
 519
 520             /* Update vectorial force */
 521             fix0             = _mm_add_pd(fix0,tx);
 522             fiy0             = _mm_add_pd(fiy0,ty);
 523             fiz0             = _mm_add_pd(fiz0,tz);
 524
 525             gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,tx,ty,tz);
 526
 527             }
 528
 529             /* Inner loop uses 30 flops */
 530         }
 531
 532         /* End of innermost loop */
 533
 534         gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
 535                                               f+i_coord_offset,fshift+i_shift_offset);
 536
 537         /* Increment number of inner iterations */
 538         inneriter                  += j_index_end - j_index_start;
 539
 540         /* Outer loop uses 7 flops */
 541     }
 542
 543     /* Increment number of outer iterations */
 544     outeriter        += nri;
 545
 546     /* Update outer/inner flops */
 547
 548     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_F,outeriter*7 + inneriter*30);
 549 }