src/gromacs/gmxlib/nonbonded/nb_kernel_avx_128_fma_single/nb_kernel_ElecCSTab_VdwNone_GeomP1P1_avx_128_fma_single.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 avx_128_fma_single kernel generator.
  37  */
  38 #include "config.h"
  39
  40 #include <math.h>
  41
  42 #include "../nb_kernel.h"
  43 #include "gromacs/legacyheaders/types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "gromacs/legacyheaders/nrnb.h"
  46
  47 #include "gromacs/simd/math_x86_avx_128_fma_single.h"
  48 #include "kernelutil_x86_avx_128_fma_single.h"
  49
  50 /*
  51  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwNone_GeomP1P1_VF_avx_128_fma_single
  52  * Electrostatics interaction: CubicSplineTable
  53  * VdW interaction:            None
  54  * Geometry:                   Particle-Particle
  55  * Calculate force/pot:        PotentialAndForce
  56  */
  57 void
  58 nb_kernel_ElecCSTab_VdwNone_GeomP1P1_VF_avx_128_fma_single
  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,C,D refer to j loop unrolling done with AVX_128, e.g. for the four 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,jnrC,jnrD;
  75     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  76     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  77     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  78     real             rcutoff_scalar;
  79     real             *shiftvec,*fshift,*x,*f;
  80     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
  81     real             scratch[4*DIM];
  82     __m128           fscal,rcutoff,rcutoff2,jidxall;
  83     int              vdwioffset0;
  84     __m128           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  85     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
  86     __m128           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  87     __m128           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  88     __m128           velec,felec,velecsum,facel,crf,krf,krf2;
  89     real             *charge;
  90     __m128i          vfitab;
  91     __m128i          ifour       = _mm_set1_epi32(4);
  92     __m128           rt,vfeps,twovfeps,vftabscale,Y,F,G,H,Fp,VV,FF;
  93     real             *vftab;
  94     __m128           dummy_mask,cutoff_mask;
  95     __m128           signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
  96     __m128           one     = _mm_set1_ps(1.0);
  97     __m128           two     = _mm_set1_ps(2.0);
  98     x                = xx[0];
  99     f                = ff[0];
 100
 101     nri              = nlist->nri;
 102     iinr             = nlist->iinr;
 103     jindex           = nlist->jindex;
 104     jjnr             = nlist->jjnr;
 105     shiftidx         = nlist->shift;
 106     gid              = nlist->gid;
 107     shiftvec         = fr->shift_vec[0];
 108     fshift           = fr->fshift[0];
 109     facel            = _mm_set1_ps(fr->epsfac);
 110     charge           = mdatoms->chargeA;
 111
 112     vftab            = kernel_data->table_elec->data;
 113     vftabscale       = _mm_set1_ps(kernel_data->table_elec->scale);
 114
 115     /* Avoid stupid compiler warnings */
 116     jnrA = jnrB = jnrC = jnrD = 0;
 117     j_coord_offsetA = 0;
 118     j_coord_offsetB = 0;
 119     j_coord_offsetC = 0;
 120     j_coord_offsetD = 0;
 121
 122     outeriter        = 0;
 123     inneriter        = 0;
 124
 125     for(iidx=0;iidx<4*DIM;iidx++)
 126     {
 127         scratch[iidx] = 0.0;
 128     }
 129
 130     /* Start outer loop over neighborlists */
 131     for(iidx=0; iidx<nri; iidx++)
 132     {
 133         /* Load shift vector for this list */
 134         i_shift_offset   = DIM*shiftidx[iidx];
 135
 136         /* Load limits for loop over neighbors */
 137         j_index_start    = jindex[iidx];
 138         j_index_end      = jindex[iidx+1];
 139
 140         /* Get outer coordinate index */
 141         inr              = iinr[iidx];
 142         i_coord_offset   = DIM*inr;
 143
 144         /* Load i particle coords and add shift vector */
 145         gmx_mm_load_shift_and_1rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 146
 147         fix0             = _mm_setzero_ps();
 148         fiy0             = _mm_setzero_ps();
 149         fiz0             = _mm_setzero_ps();
 150
 151         /* Load parameters for i particles */
 152         iq0              = _mm_mul_ps(facel,_mm_load1_ps(charge+inr+0));
 153
 154         /* Reset potential sums */
 155         velecsum         = _mm_setzero_ps();
 156
 157         /* Start inner kernel loop */
 158         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 159         {
 160
 161             /* Get j neighbor index, and coordinate index */
 162             jnrA             = jjnr[jidx];
 163             jnrB             = jjnr[jidx+1];
 164             jnrC             = jjnr[jidx+2];
 165             jnrD             = jjnr[jidx+3];
 166             j_coord_offsetA  = DIM*jnrA;
 167             j_coord_offsetB  = DIM*jnrB;
 168             j_coord_offsetC  = DIM*jnrC;
 169             j_coord_offsetD  = DIM*jnrD;
 170
 171             /* load j atom coordinates */
 172             gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 173                                               x+j_coord_offsetC,x+j_coord_offsetD,
 174                                               &jx0,&jy0,&jz0);
 175
 176             /* Calculate displacement vector */
 177             dx00             = _mm_sub_ps(ix0,jx0);
 178             dy00             = _mm_sub_ps(iy0,jy0);
 179             dz00             = _mm_sub_ps(iz0,jz0);
 180
 181             /* Calculate squared distance and things based on it */
 182             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
 183
 184             rinv00           = gmx_mm_invsqrt_ps(rsq00);
 185
 186             /* Load parameters for j particles */
 187             jq0              = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
 188                                                               charge+jnrC+0,charge+jnrD+0);
 189
 190             /**************************
 191              * CALCULATE INTERACTIONS *
 192              **************************/
 193
 194             r00              = _mm_mul_ps(rsq00,rinv00);
 195
 196             /* Compute parameters for interactions between i and j atoms */
 197             qq00             = _mm_mul_ps(iq0,jq0);
 198
 199             /* Calculate table index by multiplying r with table scale and truncate to integer */
 200             rt               = _mm_mul_ps(r00,vftabscale);
 201             vfitab           = _mm_cvttps_epi32(rt);
 202 #ifdef __XOP__
 203             vfeps            = _mm_frcz_ps(rt);
 204 #else
 205             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
 206 #endif
 207             twovfeps         = _mm_add_ps(vfeps,vfeps);
 208             vfitab           = _mm_slli_epi32(vfitab,2);
 209
 210             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 211             Y                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
 212             F                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
 213             G                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
 214             H                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
 215             _MM_TRANSPOSE4_PS(Y,F,G,H);
 216             Fp               = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
 217             VV               = _mm_macc_ps(vfeps,Fp,Y);
 218             velec            = _mm_mul_ps(qq00,VV);
 219             FF               = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
 220             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq00,FF),_mm_mul_ps(vftabscale,rinv00)));
 221
 222             /* Update potential sum for this i atom from the interaction with this j atom. */
 223             velecsum         = _mm_add_ps(velecsum,velec);
 224
 225             fscal            = felec;
 226
 227              /* Update vectorial force */
 228             fix0             = _mm_macc_ps(dx00,fscal,fix0);
 229             fiy0             = _mm_macc_ps(dy00,fscal,fiy0);
 230             fiz0             = _mm_macc_ps(dz00,fscal,fiz0);
 231
 232             fjptrA             = f+j_coord_offsetA;
 233             fjptrB             = f+j_coord_offsetB;
 234             fjptrC             = f+j_coord_offsetC;
 235             fjptrD             = f+j_coord_offsetD;
 236             gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
 237                                                    _mm_mul_ps(dx00,fscal),
 238                                                    _mm_mul_ps(dy00,fscal),
 239                                                    _mm_mul_ps(dz00,fscal));
 240
 241             /* Inner loop uses 46 flops */
 242         }
 243
 244         if(jidx<j_index_end)
 245         {
 246
 247             /* Get j neighbor index, and coordinate index */
 248             jnrlistA         = jjnr[jidx];
 249             jnrlistB         = jjnr[jidx+1];
 250             jnrlistC         = jjnr[jidx+2];
 251             jnrlistD         = jjnr[jidx+3];
 252             /* Sign of each element will be negative for non-real atoms.
 253              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 254              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 255              */
 256             dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 257             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 258             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 259             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 260             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 261             j_coord_offsetA  = DIM*jnrA;
 262             j_coord_offsetB  = DIM*jnrB;
 263             j_coord_offsetC  = DIM*jnrC;
 264             j_coord_offsetD  = DIM*jnrD;
 265
 266             /* load j atom coordinates */
 267             gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 268                                               x+j_coord_offsetC,x+j_coord_offsetD,
 269                                               &jx0,&jy0,&jz0);
 270
 271             /* Calculate displacement vector */
 272             dx00             = _mm_sub_ps(ix0,jx0);
 273             dy00             = _mm_sub_ps(iy0,jy0);
 274             dz00             = _mm_sub_ps(iz0,jz0);
 275
 276             /* Calculate squared distance and things based on it */
 277             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
 278
 279             rinv00           = gmx_mm_invsqrt_ps(rsq00);
 280
 281             /* Load parameters for j particles */
 282             jq0              = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
 283                                                               charge+jnrC+0,charge+jnrD+0);
 284
 285             /**************************
 286              * CALCULATE INTERACTIONS *
 287              **************************/
 288
 289             r00              = _mm_mul_ps(rsq00,rinv00);
 290             r00              = _mm_andnot_ps(dummy_mask,r00);
 291
 292             /* Compute parameters for interactions between i and j atoms */
 293             qq00             = _mm_mul_ps(iq0,jq0);
 294
 295             /* Calculate table index by multiplying r with table scale and truncate to integer */
 296             rt               = _mm_mul_ps(r00,vftabscale);
 297             vfitab           = _mm_cvttps_epi32(rt);
 298 #ifdef __XOP__
 299             vfeps            = _mm_frcz_ps(rt);
 300 #else
 301             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
 302 #endif
 303             twovfeps         = _mm_add_ps(vfeps,vfeps);
 304             vfitab           = _mm_slli_epi32(vfitab,2);
 305
 306             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 307             Y                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
 308             F                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
 309             G                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
 310             H                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
 311             _MM_TRANSPOSE4_PS(Y,F,G,H);
 312             Fp               = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
 313             VV               = _mm_macc_ps(vfeps,Fp,Y);
 314             velec            = _mm_mul_ps(qq00,VV);
 315             FF               = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
 316             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq00,FF),_mm_mul_ps(vftabscale,rinv00)));
 317
 318             /* Update potential sum for this i atom from the interaction with this j atom. */
 319             velec            = _mm_andnot_ps(dummy_mask,velec);
 320             velecsum         = _mm_add_ps(velecsum,velec);
 321
 322             fscal            = felec;
 323
 324             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 325
 326              /* Update vectorial force */
 327             fix0             = _mm_macc_ps(dx00,fscal,fix0);
 328             fiy0             = _mm_macc_ps(dy00,fscal,fiy0);
 329             fiz0             = _mm_macc_ps(dz00,fscal,fiz0);
 330
 331             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
 332             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
 333             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
 334             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
 335             gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
 336                                                    _mm_mul_ps(dx00,fscal),
 337                                                    _mm_mul_ps(dy00,fscal),
 338                                                    _mm_mul_ps(dz00,fscal));
 339
 340             /* Inner loop uses 47 flops */
 341         }
 342
 343         /* End of innermost loop */
 344
 345         gmx_mm_update_iforce_1atom_swizzle_ps(fix0,fiy0,fiz0,
 346                                               f+i_coord_offset,fshift+i_shift_offset);
 347
 348         ggid                        = gid[iidx];
 349         /* Update potential energies */
 350         gmx_mm_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
 351
 352         /* Increment number of inner iterations */
 353         inneriter                  += j_index_end - j_index_start;
 354
 355         /* Outer loop uses 8 flops */
 356     }
 357
 358     /* Increment number of outer iterations */
 359     outeriter        += nri;
 360
 361     /* Update outer/inner flops */
 362
 363     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VF,outeriter*8 + inneriter*47);
 364 }
 365 /*
 366  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwNone_GeomP1P1_F_avx_128_fma_single
 367  * Electrostatics interaction: CubicSplineTable
 368  * VdW interaction:            None
 369  * Geometry:                   Particle-Particle
 370  * Calculate force/pot:        Force
 371  */
 372 void
 373 nb_kernel_ElecCSTab_VdwNone_GeomP1P1_F_avx_128_fma_single
 374                     (t_nblist                    * gmx_restrict       nlist,
 375                      rvec                        * gmx_restrict          xx,
 376                      rvec                        * gmx_restrict          ff,
 377                      t_forcerec                  * gmx_restrict          fr,
 378                      t_mdatoms                   * gmx_restrict     mdatoms,
 379                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 380                      t_nrnb                      * gmx_restrict        nrnb)
 381 {
 382     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 383      * just 0 for non-waters.
 384      * Suffixes A,B,C,D refer to j loop unrolling done with AVX_128, e.g. for the four different
 385      * jnr indices corresponding to data put in the four positions in the SIMD register.
 386      */
 387     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 388     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 389     int              jnrA,jnrB,jnrC,jnrD;
 390     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
 391     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
 392     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 393     real             rcutoff_scalar;
 394     real             *shiftvec,*fshift,*x,*f;
 395     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
 396     real             scratch[4*DIM];
 397     __m128           fscal,rcutoff,rcutoff2,jidxall;
 398     int              vdwioffset0;
 399     __m128           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 400     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
 401     __m128           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 402     __m128           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 403     __m128           velec,felec,velecsum,facel,crf,krf,krf2;
 404     real             *charge;
 405     __m128i          vfitab;
 406     __m128i          ifour       = _mm_set1_epi32(4);
 407     __m128           rt,vfeps,twovfeps,vftabscale,Y,F,G,H,Fp,VV,FF;
 408     real             *vftab;
 409     __m128           dummy_mask,cutoff_mask;
 410     __m128           signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
 411     __m128           one     = _mm_set1_ps(1.0);
 412     __m128           two     = _mm_set1_ps(2.0);
 413     x                = xx[0];
 414     f                = ff[0];
 415
 416     nri              = nlist->nri;
 417     iinr             = nlist->iinr;
 418     jindex           = nlist->jindex;
 419     jjnr             = nlist->jjnr;
 420     shiftidx         = nlist->shift;
 421     gid              = nlist->gid;
 422     shiftvec         = fr->shift_vec[0];
 423     fshift           = fr->fshift[0];
 424     facel            = _mm_set1_ps(fr->epsfac);
 425     charge           = mdatoms->chargeA;
 426
 427     vftab            = kernel_data->table_elec->data;
 428     vftabscale       = _mm_set1_ps(kernel_data->table_elec->scale);
 429
 430     /* Avoid stupid compiler warnings */
 431     jnrA = jnrB = jnrC = jnrD = 0;
 432     j_coord_offsetA = 0;
 433     j_coord_offsetB = 0;
 434     j_coord_offsetC = 0;
 435     j_coord_offsetD = 0;
 436
 437     outeriter        = 0;
 438     inneriter        = 0;
 439
 440     for(iidx=0;iidx<4*DIM;iidx++)
 441     {
 442         scratch[iidx] = 0.0;
 443     }
 444
 445     /* Start outer loop over neighborlists */
 446     for(iidx=0; iidx<nri; iidx++)
 447     {
 448         /* Load shift vector for this list */
 449         i_shift_offset   = DIM*shiftidx[iidx];
 450
 451         /* Load limits for loop over neighbors */
 452         j_index_start    = jindex[iidx];
 453         j_index_end      = jindex[iidx+1];
 454
 455         /* Get outer coordinate index */
 456         inr              = iinr[iidx];
 457         i_coord_offset   = DIM*inr;
 458
 459         /* Load i particle coords and add shift vector */
 460         gmx_mm_load_shift_and_1rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 461
 462         fix0             = _mm_setzero_ps();
 463         fiy0             = _mm_setzero_ps();
 464         fiz0             = _mm_setzero_ps();
 465
 466         /* Load parameters for i particles */
 467         iq0              = _mm_mul_ps(facel,_mm_load1_ps(charge+inr+0));
 468
 469         /* Start inner kernel loop */
 470         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 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             j_coord_offsetA  = DIM*jnrA;
 479             j_coord_offsetB  = DIM*jnrB;
 480             j_coord_offsetC  = DIM*jnrC;
 481             j_coord_offsetD  = DIM*jnrD;
 482
 483             /* load j atom coordinates */
 484             gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 485                                               x+j_coord_offsetC,x+j_coord_offsetD,
 486                                               &jx0,&jy0,&jz0);
 487
 488             /* Calculate displacement vector */
 489             dx00             = _mm_sub_ps(ix0,jx0);
 490             dy00             = _mm_sub_ps(iy0,jy0);
 491             dz00             = _mm_sub_ps(iz0,jz0);
 492
 493             /* Calculate squared distance and things based on it */
 494             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
 495
 496             rinv00           = gmx_mm_invsqrt_ps(rsq00);
 497
 498             /* Load parameters for j particles */
 499             jq0              = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
 500                                                               charge+jnrC+0,charge+jnrD+0);
 501
 502             /**************************
 503              * CALCULATE INTERACTIONS *
 504              **************************/
 505
 506             r00              = _mm_mul_ps(rsq00,rinv00);
 507
 508             /* Compute parameters for interactions between i and j atoms */
 509             qq00             = _mm_mul_ps(iq0,jq0);
 510
 511             /* Calculate table index by multiplying r with table scale and truncate to integer */
 512             rt               = _mm_mul_ps(r00,vftabscale);
 513             vfitab           = _mm_cvttps_epi32(rt);
 514 #ifdef __XOP__
 515             vfeps            = _mm_frcz_ps(rt);
 516 #else
 517             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
 518 #endif
 519             twovfeps         = _mm_add_ps(vfeps,vfeps);
 520             vfitab           = _mm_slli_epi32(vfitab,2);
 521
 522             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 523             Y                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
 524             F                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
 525             G                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
 526             H                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
 527             _MM_TRANSPOSE4_PS(Y,F,G,H);
 528             Fp               = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
 529             FF               = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
 530             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq00,FF),_mm_mul_ps(vftabscale,rinv00)));
 531
 532             fscal            = felec;
 533
 534              /* Update vectorial force */
 535             fix0             = _mm_macc_ps(dx00,fscal,fix0);
 536             fiy0             = _mm_macc_ps(dy00,fscal,fiy0);
 537             fiz0             = _mm_macc_ps(dz00,fscal,fiz0);
 538
 539             fjptrA             = f+j_coord_offsetA;
 540             fjptrB             = f+j_coord_offsetB;
 541             fjptrC             = f+j_coord_offsetC;
 542             fjptrD             = f+j_coord_offsetD;
 543             gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
 544                                                    _mm_mul_ps(dx00,fscal),
 545                                                    _mm_mul_ps(dy00,fscal),
 546                                                    _mm_mul_ps(dz00,fscal));
 547
 548             /* Inner loop uses 42 flops */
 549         }
 550
 551         if(jidx<j_index_end)
 552         {
 553
 554             /* Get j neighbor index, and coordinate index */
 555             jnrlistA         = jjnr[jidx];
 556             jnrlistB         = jjnr[jidx+1];
 557             jnrlistC         = jjnr[jidx+2];
 558             jnrlistD         = jjnr[jidx+3];
 559             /* Sign of each element will be negative for non-real atoms.
 560              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 561              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 562              */
 563             dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 564             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 565             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 566             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 567             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 568             j_coord_offsetA  = DIM*jnrA;
 569             j_coord_offsetB  = DIM*jnrB;
 570             j_coord_offsetC  = DIM*jnrC;
 571             j_coord_offsetD  = DIM*jnrD;
 572
 573             /* load j atom coordinates */
 574             gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 575                                               x+j_coord_offsetC,x+j_coord_offsetD,
 576                                               &jx0,&jy0,&jz0);
 577
 578             /* Calculate displacement vector */
 579             dx00             = _mm_sub_ps(ix0,jx0);
 580             dy00             = _mm_sub_ps(iy0,jy0);
 581             dz00             = _mm_sub_ps(iz0,jz0);
 582
 583             /* Calculate squared distance and things based on it */
 584             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
 585
 586             rinv00           = gmx_mm_invsqrt_ps(rsq00);
 587
 588             /* Load parameters for j particles */
 589             jq0              = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
 590                                                               charge+jnrC+0,charge+jnrD+0);
 591
 592             /**************************
 593              * CALCULATE INTERACTIONS *
 594              **************************/
 595
 596             r00              = _mm_mul_ps(rsq00,rinv00);
 597             r00              = _mm_andnot_ps(dummy_mask,r00);
 598
 599             /* Compute parameters for interactions between i and j atoms */
 600             qq00             = _mm_mul_ps(iq0,jq0);
 601
 602             /* Calculate table index by multiplying r with table scale and truncate to integer */
 603             rt               = _mm_mul_ps(r00,vftabscale);
 604             vfitab           = _mm_cvttps_epi32(rt);
 605 #ifdef __XOP__
 606             vfeps            = _mm_frcz_ps(rt);
 607 #else
 608             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
 609 #endif
 610             twovfeps         = _mm_add_ps(vfeps,vfeps);
 611             vfitab           = _mm_slli_epi32(vfitab,2);
 612
 613             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 614             Y                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
 615             F                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
 616             G                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
 617             H                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
 618             _MM_TRANSPOSE4_PS(Y,F,G,H);
 619             Fp               = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
 620             FF               = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
 621             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq00,FF),_mm_mul_ps(vftabscale,rinv00)));
 622
 623             fscal            = felec;
 624
 625             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 626
 627              /* Update vectorial force */
 628             fix0             = _mm_macc_ps(dx00,fscal,fix0);
 629             fiy0             = _mm_macc_ps(dy00,fscal,fiy0);
 630             fiz0             = _mm_macc_ps(dz00,fscal,fiz0);
 631
 632             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
 633             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
 634             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
 635             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
 636             gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
 637                                                    _mm_mul_ps(dx00,fscal),
 638                                                    _mm_mul_ps(dy00,fscal),
 639                                                    _mm_mul_ps(dz00,fscal));
 640
 641             /* Inner loop uses 43 flops */
 642         }
 643
 644         /* End of innermost loop */
 645
 646         gmx_mm_update_iforce_1atom_swizzle_ps(fix0,fiy0,fiz0,
 647                                               f+i_coord_offset,fshift+i_shift_offset);
 648
 649         /* Increment number of inner iterations */
 650         inneriter                  += j_index_end - j_index_start;
 651
 652         /* Outer loop uses 7 flops */
 653     }
 654
 655     /* Increment number of outer iterations */
 656     outeriter        += nri;
 657
 658     /* Update outer/inner flops */
 659
 660     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_F,outeriter*7 + inneriter*43);
 661 }