src/gmxlib/nonbonded/nb_kernel_sse2_single/nb_kernel_ElecRF_VdwNone_GeomP1P1_sse2_single.c

   1 /*
   2  * Note: this file was generated by the Gromacs sse2_single kernel generator.
   3  *
   4  *                This source code is part of
   5  *
   6  *                 G   R   O   M   A   C   S
   7  *
   8  * Copyright (c) 2001-2012, The GROMACS Development Team
   9  *
  10  * Gromacs is a library for molecular simulation and trajectory analysis,
  11  * written by Erik Lindahl, David van der Spoel, Berk Hess, and others - for
  12  * a full list of developers and information, check out http://www.gromacs.org
  13  *
  14  * This program is free software; you can redistribute it and/or modify it under
  15  * the terms of the GNU Lesser General Public License as published by the Free
  16  * Software Foundation; either version 2 of the License, or (at your option) any
  17  * later version.
  18  *
  19  * To help fund GROMACS development, we humbly ask that you cite
  20  * the papers people have written on it - you can find them on the website.
  21  */
  22 #ifdef HAVE_CONFIG_H
  23 #include <config.h>
  24 #endif
  25
  26 #include <math.h>
  27
  28 #include "../nb_kernel.h"
  29 #include "types/simple.h"
  30 #include "vec.h"
  31 #include "nrnb.h"
  32
  33 #include "gmx_math_x86_sse2_single.h"
  34 #include "kernelutil_x86_sse2_single.h"
  35
  36 /*
  37  * Gromacs nonbonded kernel:   nb_kernel_ElecRF_VdwNone_GeomP1P1_VF_sse2_single
  38  * Electrostatics interaction: ReactionField
  39  * VdW interaction:            None
  40  * Geometry:                   Particle-Particle
  41  * Calculate force/pot:        PotentialAndForce
  42  */
  43 void
  44 nb_kernel_ElecRF_VdwNone_GeomP1P1_VF_sse2_single
  45                     (t_nblist * gmx_restrict                nlist,
  46                      rvec * gmx_restrict                    xx,
  47                      rvec * gmx_restrict                    ff,
  48                      t_forcerec * gmx_restrict              fr,
  49                      t_mdatoms * gmx_restrict               mdatoms,
  50                      nb_kernel_data_t * gmx_restrict        kernel_data,
  51                      t_nrnb * gmx_restrict                  nrnb)
  52 {
  53     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  54      * just 0 for non-waters.
  55      * Suffixes A,B,C,D refer to j loop unrolling done with SSE, e.g. for the four different
  56      * jnr indices corresponding to data put in the four positions in the SIMD register.
  57      */
  58     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  59     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  60     int              jnrA,jnrB,jnrC,jnrD;
  61     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  62     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  63     real             shX,shY,shZ,rcutoff_scalar;
  64     real             *shiftvec,*fshift,*x,*f;
  65     __m128           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  66     int              vdwioffset0;
  67     __m128           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  68     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
  69     __m128           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  70     __m128           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  71     __m128           velec,felec,velecsum,facel,crf,krf,krf2;
  72     real             *charge;
  73     __m128           dummy_mask,cutoff_mask;
  74     __m128           signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
  75     __m128           one     = _mm_set1_ps(1.0);
  76     __m128           two     = _mm_set1_ps(2.0);
  77     x                = xx[0];
  78     f                = ff[0];
  79
  80     nri              = nlist->nri;
  81     iinr             = nlist->iinr;
  82     jindex           = nlist->jindex;
  83     jjnr             = nlist->jjnr;
  84     shiftidx         = nlist->shift;
  85     gid              = nlist->gid;
  86     shiftvec         = fr->shift_vec[0];
  87     fshift           = fr->fshift[0];
  88     facel            = _mm_set1_ps(fr->epsfac);
  89     charge           = mdatoms->chargeA;
  90     krf              = _mm_set1_ps(fr->ic->k_rf);
  91     krf2             = _mm_set1_ps(fr->ic->k_rf*2.0);
  92     crf              = _mm_set1_ps(fr->ic->c_rf);
  93
  94     /* Avoid stupid compiler warnings */
  95     jnrA = jnrB = jnrC = jnrD = 0;
  96     j_coord_offsetA = 0;
  97     j_coord_offsetB = 0;
  98     j_coord_offsetC = 0;
  99     j_coord_offsetD = 0;
 100
 101     outeriter        = 0;
 102     inneriter        = 0;
 103
 104     /* Start outer loop over neighborlists */
 105     for(iidx=0; iidx<nri; iidx++)
 106     {
 107         /* Load shift vector for this list */
 108         i_shift_offset   = DIM*shiftidx[iidx];
 109         shX              = shiftvec[i_shift_offset+XX];
 110         shY              = shiftvec[i_shift_offset+YY];
 111         shZ              = shiftvec[i_shift_offset+ZZ];
 112
 113         /* Load limits for loop over neighbors */
 114         j_index_start    = jindex[iidx];
 115         j_index_end      = jindex[iidx+1];
 116
 117         /* Get outer coordinate index */
 118         inr              = iinr[iidx];
 119         i_coord_offset   = DIM*inr;
 120
 121         /* Load i particle coords and add shift vector */
 122         ix0              = _mm_set1_ps(shX + x[i_coord_offset+DIM*0+XX]);
 123         iy0              = _mm_set1_ps(shY + x[i_coord_offset+DIM*0+YY]);
 124         iz0              = _mm_set1_ps(shZ + x[i_coord_offset+DIM*0+ZZ]);
 125
 126         fix0             = _mm_setzero_ps();
 127         fiy0             = _mm_setzero_ps();
 128         fiz0             = _mm_setzero_ps();
 129
 130         /* Load parameters for i particles */
 131         iq0              = _mm_mul_ps(facel,_mm_load1_ps(charge+inr+0));
 132
 133         /* Reset potential sums */
 134         velecsum         = _mm_setzero_ps();
 135
 136         /* Start inner kernel loop */
 137         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 138         {
 139
 140             /* Get j neighbor index, and coordinate index */
 141             jnrA             = jjnr[jidx];
 142             jnrB             = jjnr[jidx+1];
 143             jnrC             = jjnr[jidx+2];
 144             jnrD             = jjnr[jidx+3];
 145
 146             j_coord_offsetA  = DIM*jnrA;
 147             j_coord_offsetB  = DIM*jnrB;
 148             j_coord_offsetC  = DIM*jnrC;
 149             j_coord_offsetD  = DIM*jnrD;
 150
 151             /* load j atom coordinates */
 152             gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 153                                               x+j_coord_offsetC,x+j_coord_offsetD,
 154                                               &jx0,&jy0,&jz0);
 155
 156             /* Calculate displacement vector */
 157             dx00             = _mm_sub_ps(ix0,jx0);
 158             dy00             = _mm_sub_ps(iy0,jy0);
 159             dz00             = _mm_sub_ps(iz0,jz0);
 160
 161             /* Calculate squared distance and things based on it */
 162             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
 163
 164             rinv00           = gmx_mm_invsqrt_ps(rsq00);
 165
 166             rinvsq00         = _mm_mul_ps(rinv00,rinv00);
 167
 168             /* Load parameters for j particles */
 169             jq0              = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
 170                                                               charge+jnrC+0,charge+jnrD+0);
 171
 172             /**************************
 173              * CALCULATE INTERACTIONS *
 174              **************************/
 175
 176             /* Compute parameters for interactions between i and j atoms */
 177             qq00             = _mm_mul_ps(iq0,jq0);
 178
 179             /* REACTION-FIELD ELECTROSTATICS */
 180             velec            = _mm_mul_ps(qq00,_mm_sub_ps(_mm_add_ps(rinv00,_mm_mul_ps(krf,rsq00)),crf));
 181             felec            = _mm_mul_ps(qq00,_mm_sub_ps(_mm_mul_ps(rinv00,rinvsq00),krf2));
 182
 183             /* Update potential sum for this i atom from the interaction with this j atom. */
 184             velecsum         = _mm_add_ps(velecsum,velec);
 185
 186             fscal            = felec;
 187
 188             /* Calculate temporary vectorial force */
 189             tx               = _mm_mul_ps(fscal,dx00);
 190             ty               = _mm_mul_ps(fscal,dy00);
 191             tz               = _mm_mul_ps(fscal,dz00);
 192
 193             /* Update vectorial force */
 194             fix0             = _mm_add_ps(fix0,tx);
 195             fiy0             = _mm_add_ps(fiy0,ty);
 196             fiz0             = _mm_add_ps(fiz0,tz);
 197
 198             gmx_mm_decrement_1rvec_4ptr_swizzle_ps(f+j_coord_offsetA,f+j_coord_offsetB,
 199                                                    f+j_coord_offsetC,f+j_coord_offsetD,
 200                                                    tx,ty,tz);
 201
 202             /* Inner loop uses 32 flops */
 203         }
 204
 205         if(jidx<j_index_end)
 206         {
 207
 208             /* Get j neighbor index, and coordinate index */
 209             jnrA             = jjnr[jidx];
 210             jnrB             = jjnr[jidx+1];
 211             jnrC             = jjnr[jidx+2];
 212             jnrD             = jjnr[jidx+3];
 213
 214             /* Sign of each element will be negative for non-real atoms.
 215              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 216              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 217              */
 218             dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 219             jnrA       = (jnrA>=0) ? jnrA : 0;
 220             jnrB       = (jnrB>=0) ? jnrB : 0;
 221             jnrC       = (jnrC>=0) ? jnrC : 0;
 222             jnrD       = (jnrD>=0) ? jnrD : 0;
 223
 224             j_coord_offsetA  = DIM*jnrA;
 225             j_coord_offsetB  = DIM*jnrB;
 226             j_coord_offsetC  = DIM*jnrC;
 227             j_coord_offsetD  = DIM*jnrD;
 228
 229             /* load j atom coordinates */
 230             gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 231                                               x+j_coord_offsetC,x+j_coord_offsetD,
 232                                               &jx0,&jy0,&jz0);
 233
 234             /* Calculate displacement vector */
 235             dx00             = _mm_sub_ps(ix0,jx0);
 236             dy00             = _mm_sub_ps(iy0,jy0);
 237             dz00             = _mm_sub_ps(iz0,jz0);
 238
 239             /* Calculate squared distance and things based on it */
 240             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
 241
 242             rinv00           = gmx_mm_invsqrt_ps(rsq00);
 243
 244             rinvsq00         = _mm_mul_ps(rinv00,rinv00);
 245
 246             /* Load parameters for j particles */
 247             jq0              = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
 248                                                               charge+jnrC+0,charge+jnrD+0);
 249
 250             /**************************
 251              * CALCULATE INTERACTIONS *
 252              **************************/
 253
 254             /* Compute parameters for interactions between i and j atoms */
 255             qq00             = _mm_mul_ps(iq0,jq0);
 256
 257             /* REACTION-FIELD ELECTROSTATICS */
 258             velec            = _mm_mul_ps(qq00,_mm_sub_ps(_mm_add_ps(rinv00,_mm_mul_ps(krf,rsq00)),crf));
 259             felec            = _mm_mul_ps(qq00,_mm_sub_ps(_mm_mul_ps(rinv00,rinvsq00),krf2));
 260
 261             /* Update potential sum for this i atom from the interaction with this j atom. */
 262             velec            = _mm_andnot_ps(dummy_mask,velec);
 263             velecsum         = _mm_add_ps(velecsum,velec);
 264
 265             fscal            = felec;
 266
 267             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 268
 269             /* Calculate temporary vectorial force */
 270             tx               = _mm_mul_ps(fscal,dx00);
 271             ty               = _mm_mul_ps(fscal,dy00);
 272             tz               = _mm_mul_ps(fscal,dz00);
 273
 274             /* Update vectorial force */
 275             fix0             = _mm_add_ps(fix0,tx);
 276             fiy0             = _mm_add_ps(fiy0,ty);
 277             fiz0             = _mm_add_ps(fiz0,tz);
 278
 279             gmx_mm_decrement_1rvec_4ptr_swizzle_ps(f+j_coord_offsetA,f+j_coord_offsetB,
 280                                                    f+j_coord_offsetC,f+j_coord_offsetD,
 281                                                    tx,ty,tz);
 282
 283             /* Inner loop uses 32 flops */
 284         }
 285
 286         /* End of innermost loop */
 287
 288         gmx_mm_update_iforce_1atom_swizzle_ps(fix0,fiy0,fiz0,
 289                                               f+i_coord_offset,fshift+i_shift_offset);
 290
 291         ggid                        = gid[iidx];
 292         /* Update potential energies */
 293         gmx_mm_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
 294
 295         /* Increment number of inner iterations */
 296         inneriter                  += j_index_end - j_index_start;
 297
 298         /* Outer loop uses 11 flops */
 299     }
 300
 301     /* Increment number of outer iterations */
 302     outeriter        += nri;
 303
 304     /* Update outer/inner flops */
 305
 306     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VF,outeriter*11 + inneriter*32);
 307 }
 308 /*
 309  * Gromacs nonbonded kernel:   nb_kernel_ElecRF_VdwNone_GeomP1P1_F_sse2_single
 310  * Electrostatics interaction: ReactionField
 311  * VdW interaction:            None
 312  * Geometry:                   Particle-Particle
 313  * Calculate force/pot:        Force
 314  */
 315 void
 316 nb_kernel_ElecRF_VdwNone_GeomP1P1_F_sse2_single
 317                     (t_nblist * gmx_restrict                nlist,
 318                      rvec * gmx_restrict                    xx,
 319                      rvec * gmx_restrict                    ff,
 320                      t_forcerec * gmx_restrict              fr,
 321                      t_mdatoms * gmx_restrict               mdatoms,
 322                      nb_kernel_data_t * gmx_restrict        kernel_data,
 323                      t_nrnb * gmx_restrict                  nrnb)
 324 {
 325     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 326      * just 0 for non-waters.
 327      * Suffixes A,B,C,D refer to j loop unrolling done with SSE, e.g. for the four different
 328      * jnr indices corresponding to data put in the four positions in the SIMD register.
 329      */
 330     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 331     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 332     int              jnrA,jnrB,jnrC,jnrD;
 333     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
 334     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 335     real             shX,shY,shZ,rcutoff_scalar;
 336     real             *shiftvec,*fshift,*x,*f;
 337     __m128           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 338     int              vdwioffset0;
 339     __m128           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 340     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
 341     __m128           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 342     __m128           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 343     __m128           velec,felec,velecsum,facel,crf,krf,krf2;
 344     real             *charge;
 345     __m128           dummy_mask,cutoff_mask;
 346     __m128           signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
 347     __m128           one     = _mm_set1_ps(1.0);
 348     __m128           two     = _mm_set1_ps(2.0);
 349     x                = xx[0];
 350     f                = ff[0];
 351
 352     nri              = nlist->nri;
 353     iinr             = nlist->iinr;
 354     jindex           = nlist->jindex;
 355     jjnr             = nlist->jjnr;
 356     shiftidx         = nlist->shift;
 357     gid              = nlist->gid;
 358     shiftvec         = fr->shift_vec[0];
 359     fshift           = fr->fshift[0];
 360     facel            = _mm_set1_ps(fr->epsfac);
 361     charge           = mdatoms->chargeA;
 362     krf              = _mm_set1_ps(fr->ic->k_rf);
 363     krf2             = _mm_set1_ps(fr->ic->k_rf*2.0);
 364     crf              = _mm_set1_ps(fr->ic->c_rf);
 365
 366     /* Avoid stupid compiler warnings */
 367     jnrA = jnrB = jnrC = jnrD = 0;
 368     j_coord_offsetA = 0;
 369     j_coord_offsetB = 0;
 370     j_coord_offsetC = 0;
 371     j_coord_offsetD = 0;
 372
 373     outeriter        = 0;
 374     inneriter        = 0;
 375
 376     /* Start outer loop over neighborlists */
 377     for(iidx=0; iidx<nri; iidx++)
 378     {
 379         /* Load shift vector for this list */
 380         i_shift_offset   = DIM*shiftidx[iidx];
 381         shX              = shiftvec[i_shift_offset+XX];
 382         shY              = shiftvec[i_shift_offset+YY];
 383         shZ              = shiftvec[i_shift_offset+ZZ];
 384
 385         /* Load limits for loop over neighbors */
 386         j_index_start    = jindex[iidx];
 387         j_index_end      = jindex[iidx+1];
 388
 389         /* Get outer coordinate index */
 390         inr              = iinr[iidx];
 391         i_coord_offset   = DIM*inr;
 392
 393         /* Load i particle coords and add shift vector */
 394         ix0              = _mm_set1_ps(shX + x[i_coord_offset+DIM*0+XX]);
 395         iy0              = _mm_set1_ps(shY + x[i_coord_offset+DIM*0+YY]);
 396         iz0              = _mm_set1_ps(shZ + x[i_coord_offset+DIM*0+ZZ]);
 397
 398         fix0             = _mm_setzero_ps();
 399         fiy0             = _mm_setzero_ps();
 400         fiz0             = _mm_setzero_ps();
 401
 402         /* Load parameters for i particles */
 403         iq0              = _mm_mul_ps(facel,_mm_load1_ps(charge+inr+0));
 404
 405         /* Start inner kernel loop */
 406         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 407         {
 408
 409             /* Get j neighbor index, and coordinate index */
 410             jnrA             = jjnr[jidx];
 411             jnrB             = jjnr[jidx+1];
 412             jnrC             = jjnr[jidx+2];
 413             jnrD             = jjnr[jidx+3];
 414
 415             j_coord_offsetA  = DIM*jnrA;
 416             j_coord_offsetB  = DIM*jnrB;
 417             j_coord_offsetC  = DIM*jnrC;
 418             j_coord_offsetD  = DIM*jnrD;
 419
 420             /* load j atom coordinates */
 421             gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 422                                               x+j_coord_offsetC,x+j_coord_offsetD,
 423                                               &jx0,&jy0,&jz0);
 424
 425             /* Calculate displacement vector */
 426             dx00             = _mm_sub_ps(ix0,jx0);
 427             dy00             = _mm_sub_ps(iy0,jy0);
 428             dz00             = _mm_sub_ps(iz0,jz0);
 429
 430             /* Calculate squared distance and things based on it */
 431             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
 432
 433             rinv00           = gmx_mm_invsqrt_ps(rsq00);
 434
 435             rinvsq00         = _mm_mul_ps(rinv00,rinv00);
 436
 437             /* Load parameters for j particles */
 438             jq0              = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
 439                                                               charge+jnrC+0,charge+jnrD+0);
 440
 441             /**************************
 442              * CALCULATE INTERACTIONS *
 443              **************************/
 444
 445             /* Compute parameters for interactions between i and j atoms */
 446             qq00             = _mm_mul_ps(iq0,jq0);
 447
 448             /* REACTION-FIELD ELECTROSTATICS */
 449             felec            = _mm_mul_ps(qq00,_mm_sub_ps(_mm_mul_ps(rinv00,rinvsq00),krf2));
 450
 451             fscal            = felec;
 452
 453             /* Calculate temporary vectorial force */
 454             tx               = _mm_mul_ps(fscal,dx00);
 455             ty               = _mm_mul_ps(fscal,dy00);
 456             tz               = _mm_mul_ps(fscal,dz00);
 457
 458             /* Update vectorial force */
 459             fix0             = _mm_add_ps(fix0,tx);
 460             fiy0             = _mm_add_ps(fiy0,ty);
 461             fiz0             = _mm_add_ps(fiz0,tz);
 462
 463             gmx_mm_decrement_1rvec_4ptr_swizzle_ps(f+j_coord_offsetA,f+j_coord_offsetB,
 464                                                    f+j_coord_offsetC,f+j_coord_offsetD,
 465                                                    tx,ty,tz);
 466
 467             /* Inner loop uses 27 flops */
 468         }
 469
 470         if(jidx<j_index_end)
 471         {
 472
 473             /* Get j neighbor index, and coordinate index */
 474             jnrA             = jjnr[jidx];
 475             jnrB             = jjnr[jidx+1];
 476             jnrC             = jjnr[jidx+2];
 477             jnrD             = jjnr[jidx+3];
 478
 479             /* Sign of each element will be negative for non-real atoms.
 480              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 481              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 482              */
 483             dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 484             jnrA       = (jnrA>=0) ? jnrA : 0;
 485             jnrB       = (jnrB>=0) ? jnrB : 0;
 486             jnrC       = (jnrC>=0) ? jnrC : 0;
 487             jnrD       = (jnrD>=0) ? jnrD : 0;
 488
 489             j_coord_offsetA  = DIM*jnrA;
 490             j_coord_offsetB  = DIM*jnrB;
 491             j_coord_offsetC  = DIM*jnrC;
 492             j_coord_offsetD  = DIM*jnrD;
 493
 494             /* load j atom coordinates */
 495             gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 496                                               x+j_coord_offsetC,x+j_coord_offsetD,
 497                                               &jx0,&jy0,&jz0);
 498
 499             /* Calculate displacement vector */
 500             dx00             = _mm_sub_ps(ix0,jx0);
 501             dy00             = _mm_sub_ps(iy0,jy0);
 502             dz00             = _mm_sub_ps(iz0,jz0);
 503
 504             /* Calculate squared distance and things based on it */
 505             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
 506
 507             rinv00           = gmx_mm_invsqrt_ps(rsq00);
 508
 509             rinvsq00         = _mm_mul_ps(rinv00,rinv00);
 510
 511             /* Load parameters for j particles */
 512             jq0              = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
 513                                                               charge+jnrC+0,charge+jnrD+0);
 514
 515             /**************************
 516              * CALCULATE INTERACTIONS *
 517              **************************/
 518
 519             /* Compute parameters for interactions between i and j atoms */
 520             qq00             = _mm_mul_ps(iq0,jq0);
 521
 522             /* REACTION-FIELD ELECTROSTATICS */
 523             felec            = _mm_mul_ps(qq00,_mm_sub_ps(_mm_mul_ps(rinv00,rinvsq00),krf2));
 524
 525             fscal            = felec;
 526
 527             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 528
 529             /* Calculate temporary vectorial force */
 530             tx               = _mm_mul_ps(fscal,dx00);
 531             ty               = _mm_mul_ps(fscal,dy00);
 532             tz               = _mm_mul_ps(fscal,dz00);
 533
 534             /* Update vectorial force */
 535             fix0             = _mm_add_ps(fix0,tx);
 536             fiy0             = _mm_add_ps(fiy0,ty);
 537             fiz0             = _mm_add_ps(fiz0,tz);
 538
 539             gmx_mm_decrement_1rvec_4ptr_swizzle_ps(f+j_coord_offsetA,f+j_coord_offsetB,
 540                                                    f+j_coord_offsetC,f+j_coord_offsetD,
 541                                                    tx,ty,tz);
 542
 543             /* Inner loop uses 27 flops */
 544         }
 545
 546         /* End of innermost loop */
 547
 548         gmx_mm_update_iforce_1atom_swizzle_ps(fix0,fiy0,fiz0,
 549                                               f+i_coord_offset,fshift+i_shift_offset);
 550
 551         /* Increment number of inner iterations */
 552         inneriter                  += j_index_end - j_index_start;
 553
 554         /* Outer loop uses 10 flops */
 555     }
 556
 557     /* Increment number of outer iterations */
 558     outeriter        += nri;
 559
 560     /* Update outer/inner flops */
 561
 562     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_F,outeriter*10 + inneriter*27);
 563 }