src/gmxlib/nonbonded/nb_kernel_avx_256_double/nb_kernel_ElecEwSh_VdwNone_GeomP1P1_avx_256_double.c

   1 /*
   2  * Note: this file was generated by the Gromacs avx_256_double 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_avx_256_double.h"
  34 #include "kernelutil_x86_avx_256_double.h"
  35
  36 /*
  37  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSh_VdwNone_GeomP1P1_VF_avx_256_double
  38  * Electrostatics interaction: Ewald
  39  * VdW interaction:            None
  40  * Geometry:                   Particle-Particle
  41  * Calculate force/pot:        PotentialAndForce
  42  */
  43 void
  44 nb_kernel_ElecEwSh_VdwNone_GeomP1P1_VF_avx_256_double
  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 AVX, 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              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  62     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
  63     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  64     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  65     real             rcutoff_scalar;
  66     real             *shiftvec,*fshift,*x,*f;
  67     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
  68     real             scratch[4*DIM];
  69     __m256d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  70     real *           vdwioffsetptr0;
  71     __m256d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  72     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
  73     __m256d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  74     __m256d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  75     __m256d          velec,felec,velecsum,facel,crf,krf,krf2;
  76     real             *charge;
  77     __m128i          ewitab;
  78     __m256d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
  79     __m256d          beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
  80     real             *ewtab;
  81     __m256d          dummy_mask,cutoff_mask;
  82     __m128           tmpmask0,tmpmask1;
  83     __m256d          signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
  84     __m256d          one     = _mm256_set1_pd(1.0);
  85     __m256d          two     = _mm256_set1_pd(2.0);
  86     x                = xx[0];
  87     f                = ff[0];
  88
  89     nri              = nlist->nri;
  90     iinr             = nlist->iinr;
  91     jindex           = nlist->jindex;
  92     jjnr             = nlist->jjnr;
  93     shiftidx         = nlist->shift;
  94     gid              = nlist->gid;
  95     shiftvec         = fr->shift_vec[0];
  96     fshift           = fr->fshift[0];
  97     facel            = _mm256_set1_pd(fr->epsfac);
  98     charge           = mdatoms->chargeA;
  99
 100     sh_ewald         = _mm256_set1_pd(fr->ic->sh_ewald);
 101     beta             = _mm256_set1_pd(fr->ic->ewaldcoeff);
 102     beta2            = _mm256_mul_pd(beta,beta);
 103     beta3            = _mm256_mul_pd(beta,beta2);
 104
 105     ewtab            = fr->ic->tabq_coul_FDV0;
 106     ewtabscale       = _mm256_set1_pd(fr->ic->tabq_scale);
 107     ewtabhalfspace   = _mm256_set1_pd(0.5/fr->ic->tabq_scale);
 108
 109     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 110     rcutoff_scalar   = fr->rcoulomb;
 111     rcutoff          = _mm256_set1_pd(rcutoff_scalar);
 112     rcutoff2         = _mm256_mul_pd(rcutoff,rcutoff);
 113
 114     /* Avoid stupid compiler warnings */
 115     jnrA = jnrB = jnrC = jnrD = 0;
 116     j_coord_offsetA = 0;
 117     j_coord_offsetB = 0;
 118     j_coord_offsetC = 0;
 119     j_coord_offsetD = 0;
 120
 121     outeriter        = 0;
 122     inneriter        = 0;
 123
 124     for(iidx=0;iidx<4*DIM;iidx++)
 125     {
 126         scratch[iidx] = 0.0;
 127     }
 128
 129     /* Start outer loop over neighborlists */
 130     for(iidx=0; iidx<nri; iidx++)
 131     {
 132         /* Load shift vector for this list */
 133         i_shift_offset   = DIM*shiftidx[iidx];
 134
 135         /* Load limits for loop over neighbors */
 136         j_index_start    = jindex[iidx];
 137         j_index_end      = jindex[iidx+1];
 138
 139         /* Get outer coordinate index */
 140         inr              = iinr[iidx];
 141         i_coord_offset   = DIM*inr;
 142
 143         /* Load i particle coords and add shift vector */
 144         gmx_mm256_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 145
 146         fix0             = _mm256_setzero_pd();
 147         fiy0             = _mm256_setzero_pd();
 148         fiz0             = _mm256_setzero_pd();
 149
 150         /* Load parameters for i particles */
 151         iq0              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+0]));
 152
 153         /* Reset potential sums */
 154         velecsum         = _mm256_setzero_pd();
 155
 156         /* Start inner kernel loop */
 157         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 158         {
 159
 160             /* Get j neighbor index, and coordinate index */
 161             jnrA             = jjnr[jidx];
 162             jnrB             = jjnr[jidx+1];
 163             jnrC             = jjnr[jidx+2];
 164             jnrD             = jjnr[jidx+3];
 165             j_coord_offsetA  = DIM*jnrA;
 166             j_coord_offsetB  = DIM*jnrB;
 167             j_coord_offsetC  = DIM*jnrC;
 168             j_coord_offsetD  = DIM*jnrD;
 169
 170             /* load j atom coordinates */
 171             gmx_mm256_load_1rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 172                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 173                                                  &jx0,&jy0,&jz0);
 174
 175             /* Calculate displacement vector */
 176             dx00             = _mm256_sub_pd(ix0,jx0);
 177             dy00             = _mm256_sub_pd(iy0,jy0);
 178             dz00             = _mm256_sub_pd(iz0,jz0);
 179
 180             /* Calculate squared distance and things based on it */
 181             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 182
 183             rinv00           = gmx_mm256_invsqrt_pd(rsq00);
 184
 185             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
 186
 187             /* Load parameters for j particles */
 188             jq0              = gmx_mm256_load_4real_swizzle_pd(charge+jnrA+0,charge+jnrB+0,
 189                                                                  charge+jnrC+0,charge+jnrD+0);
 190
 191             /**************************
 192              * CALCULATE INTERACTIONS *
 193              **************************/
 194
 195             if (gmx_mm256_any_lt(rsq00,rcutoff2))
 196             {
 197
 198             r00              = _mm256_mul_pd(rsq00,rinv00);
 199
 200             /* Compute parameters for interactions between i and j atoms */
 201             qq00             = _mm256_mul_pd(iq0,jq0);
 202
 203             /* EWALD ELECTROSTATICS */
 204
 205             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 206             ewrt             = _mm256_mul_pd(r00,ewtabscale);
 207             ewitab           = _mm256_cvttpd_epi32(ewrt);
 208             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 209             ewitab           = _mm_slli_epi32(ewitab,2);
 210             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 211             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 212             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 213             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 214             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 215             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 216             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 217             velec            = _mm256_mul_pd(qq00,_mm256_sub_pd(_mm256_sub_pd(rinv00,sh_ewald),velec));
 218             felec            = _mm256_mul_pd(_mm256_mul_pd(qq00,rinv00),_mm256_sub_pd(rinvsq00,felec));
 219
 220             cutoff_mask      = _mm256_cmp_pd(rsq00,rcutoff2,_CMP_LT_OQ);
 221
 222             /* Update potential sum for this i atom from the interaction with this j atom. */
 223             velec            = _mm256_and_pd(velec,cutoff_mask);
 224             velecsum         = _mm256_add_pd(velecsum,velec);
 225
 226             fscal            = felec;
 227
 228             fscal            = _mm256_and_pd(fscal,cutoff_mask);
 229
 230             /* Calculate temporary vectorial force */
 231             tx               = _mm256_mul_pd(fscal,dx00);
 232             ty               = _mm256_mul_pd(fscal,dy00);
 233             tz               = _mm256_mul_pd(fscal,dz00);
 234
 235             /* Update vectorial force */
 236             fix0             = _mm256_add_pd(fix0,tx);
 237             fiy0             = _mm256_add_pd(fiy0,ty);
 238             fiz0             = _mm256_add_pd(fiz0,tz);
 239
 240             fjptrA             = f+j_coord_offsetA;
 241             fjptrB             = f+j_coord_offsetB;
 242             fjptrC             = f+j_coord_offsetC;
 243             fjptrD             = f+j_coord_offsetD;
 244             gmx_mm256_decrement_1rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,tx,ty,tz);
 245
 246             }
 247
 248             /* Inner loop uses 46 flops */
 249         }
 250
 251         if(jidx<j_index_end)
 252         {
 253
 254             /* Get j neighbor index, and coordinate index */
 255             jnrlistA         = jjnr[jidx];
 256             jnrlistB         = jjnr[jidx+1];
 257             jnrlistC         = jjnr[jidx+2];
 258             jnrlistD         = jjnr[jidx+3];
 259             /* Sign of each element will be negative for non-real atoms.
 260              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 261              * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
 262              */
 263             tmpmask0 = gmx_mm_castsi128_pd(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 264
 265             tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
 266             tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
 267             dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
 268
 269             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 270             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 271             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 272             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 273             j_coord_offsetA  = DIM*jnrA;
 274             j_coord_offsetB  = DIM*jnrB;
 275             j_coord_offsetC  = DIM*jnrC;
 276             j_coord_offsetD  = DIM*jnrD;
 277
 278             /* load j atom coordinates */
 279             gmx_mm256_load_1rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 280                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 281                                                  &jx0,&jy0,&jz0);
 282
 283             /* Calculate displacement vector */
 284             dx00             = _mm256_sub_pd(ix0,jx0);
 285             dy00             = _mm256_sub_pd(iy0,jy0);
 286             dz00             = _mm256_sub_pd(iz0,jz0);
 287
 288             /* Calculate squared distance and things based on it */
 289             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 290
 291             rinv00           = gmx_mm256_invsqrt_pd(rsq00);
 292
 293             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
 294
 295             /* Load parameters for j particles */
 296             jq0              = gmx_mm256_load_4real_swizzle_pd(charge+jnrA+0,charge+jnrB+0,
 297                                                                  charge+jnrC+0,charge+jnrD+0);
 298
 299             /**************************
 300              * CALCULATE INTERACTIONS *
 301              **************************/
 302
 303             if (gmx_mm256_any_lt(rsq00,rcutoff2))
 304             {
 305
 306             r00              = _mm256_mul_pd(rsq00,rinv00);
 307             r00              = _mm256_andnot_pd(dummy_mask,r00);
 308
 309             /* Compute parameters for interactions between i and j atoms */
 310             qq00             = _mm256_mul_pd(iq0,jq0);
 311
 312             /* EWALD ELECTROSTATICS */
 313
 314             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 315             ewrt             = _mm256_mul_pd(r00,ewtabscale);
 316             ewitab           = _mm256_cvttpd_epi32(ewrt);
 317             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 318             ewitab           = _mm_slli_epi32(ewitab,2);
 319             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 320             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 321             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 322             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 323             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 324             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 325             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 326             velec            = _mm256_mul_pd(qq00,_mm256_sub_pd(_mm256_sub_pd(rinv00,sh_ewald),velec));
 327             felec            = _mm256_mul_pd(_mm256_mul_pd(qq00,rinv00),_mm256_sub_pd(rinvsq00,felec));
 328
 329             cutoff_mask      = _mm256_cmp_pd(rsq00,rcutoff2,_CMP_LT_OQ);
 330
 331             /* Update potential sum for this i atom from the interaction with this j atom. */
 332             velec            = _mm256_and_pd(velec,cutoff_mask);
 333             velec            = _mm256_andnot_pd(dummy_mask,velec);
 334             velecsum         = _mm256_add_pd(velecsum,velec);
 335
 336             fscal            = felec;
 337
 338             fscal            = _mm256_and_pd(fscal,cutoff_mask);
 339
 340             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 341
 342             /* Calculate temporary vectorial force */
 343             tx               = _mm256_mul_pd(fscal,dx00);
 344             ty               = _mm256_mul_pd(fscal,dy00);
 345             tz               = _mm256_mul_pd(fscal,dz00);
 346
 347             /* Update vectorial force */
 348             fix0             = _mm256_add_pd(fix0,tx);
 349             fiy0             = _mm256_add_pd(fiy0,ty);
 350             fiz0             = _mm256_add_pd(fiz0,tz);
 351
 352             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
 353             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
 354             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
 355             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
 356             gmx_mm256_decrement_1rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,tx,ty,tz);
 357
 358             }
 359
 360             /* Inner loop uses 47 flops */
 361         }
 362
 363         /* End of innermost loop */
 364
 365         gmx_mm256_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
 366                                                  f+i_coord_offset,fshift+i_shift_offset);
 367
 368         ggid                        = gid[iidx];
 369         /* Update potential energies */
 370         gmx_mm256_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
 371
 372         /* Increment number of inner iterations */
 373         inneriter                  += j_index_end - j_index_start;
 374
 375         /* Outer loop uses 8 flops */
 376     }
 377
 378     /* Increment number of outer iterations */
 379     outeriter        += nri;
 380
 381     /* Update outer/inner flops */
 382
 383     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VF,outeriter*8 + inneriter*47);
 384 }
 385 /*
 386  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSh_VdwNone_GeomP1P1_F_avx_256_double
 387  * Electrostatics interaction: Ewald
 388  * VdW interaction:            None
 389  * Geometry:                   Particle-Particle
 390  * Calculate force/pot:        Force
 391  */
 392 void
 393 nb_kernel_ElecEwSh_VdwNone_GeomP1P1_F_avx_256_double
 394                     (t_nblist * gmx_restrict                nlist,
 395                      rvec * gmx_restrict                    xx,
 396                      rvec * gmx_restrict                    ff,
 397                      t_forcerec * gmx_restrict              fr,
 398                      t_mdatoms * gmx_restrict               mdatoms,
 399                      nb_kernel_data_t * gmx_restrict        kernel_data,
 400                      t_nrnb * gmx_restrict                  nrnb)
 401 {
 402     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 403      * just 0 for non-waters.
 404      * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
 405      * jnr indices corresponding to data put in the four positions in the SIMD register.
 406      */
 407     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 408     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 409     int              jnrA,jnrB,jnrC,jnrD;
 410     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
 411     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
 412     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
 413     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 414     real             rcutoff_scalar;
 415     real             *shiftvec,*fshift,*x,*f;
 416     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
 417     real             scratch[4*DIM];
 418     __m256d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 419     real *           vdwioffsetptr0;
 420     __m256d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 421     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
 422     __m256d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 423     __m256d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 424     __m256d          velec,felec,velecsum,facel,crf,krf,krf2;
 425     real             *charge;
 426     __m128i          ewitab;
 427     __m256d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 428     __m256d          beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
 429     real             *ewtab;
 430     __m256d          dummy_mask,cutoff_mask;
 431     __m128           tmpmask0,tmpmask1;
 432     __m256d          signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
 433     __m256d          one     = _mm256_set1_pd(1.0);
 434     __m256d          two     = _mm256_set1_pd(2.0);
 435     x                = xx[0];
 436     f                = ff[0];
 437
 438     nri              = nlist->nri;
 439     iinr             = nlist->iinr;
 440     jindex           = nlist->jindex;
 441     jjnr             = nlist->jjnr;
 442     shiftidx         = nlist->shift;
 443     gid              = nlist->gid;
 444     shiftvec         = fr->shift_vec[0];
 445     fshift           = fr->fshift[0];
 446     facel            = _mm256_set1_pd(fr->epsfac);
 447     charge           = mdatoms->chargeA;
 448
 449     sh_ewald         = _mm256_set1_pd(fr->ic->sh_ewald);
 450     beta             = _mm256_set1_pd(fr->ic->ewaldcoeff);
 451     beta2            = _mm256_mul_pd(beta,beta);
 452     beta3            = _mm256_mul_pd(beta,beta2);
 453
 454     ewtab            = fr->ic->tabq_coul_F;
 455     ewtabscale       = _mm256_set1_pd(fr->ic->tabq_scale);
 456     ewtabhalfspace   = _mm256_set1_pd(0.5/fr->ic->tabq_scale);
 457
 458     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 459     rcutoff_scalar   = fr->rcoulomb;
 460     rcutoff          = _mm256_set1_pd(rcutoff_scalar);
 461     rcutoff2         = _mm256_mul_pd(rcutoff,rcutoff);
 462
 463     /* Avoid stupid compiler warnings */
 464     jnrA = jnrB = jnrC = jnrD = 0;
 465     j_coord_offsetA = 0;
 466     j_coord_offsetB = 0;
 467     j_coord_offsetC = 0;
 468     j_coord_offsetD = 0;
 469
 470     outeriter        = 0;
 471     inneriter        = 0;
 472
 473     for(iidx=0;iidx<4*DIM;iidx++)
 474     {
 475         scratch[iidx] = 0.0;
 476     }
 477
 478     /* Start outer loop over neighborlists */
 479     for(iidx=0; iidx<nri; iidx++)
 480     {
 481         /* Load shift vector for this list */
 482         i_shift_offset   = DIM*shiftidx[iidx];
 483
 484         /* Load limits for loop over neighbors */
 485         j_index_start    = jindex[iidx];
 486         j_index_end      = jindex[iidx+1];
 487
 488         /* Get outer coordinate index */
 489         inr              = iinr[iidx];
 490         i_coord_offset   = DIM*inr;
 491
 492         /* Load i particle coords and add shift vector */
 493         gmx_mm256_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 494
 495         fix0             = _mm256_setzero_pd();
 496         fiy0             = _mm256_setzero_pd();
 497         fiz0             = _mm256_setzero_pd();
 498
 499         /* Load parameters for i particles */
 500         iq0              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+0]));
 501
 502         /* Start inner kernel loop */
 503         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 504         {
 505
 506             /* Get j neighbor index, and coordinate index */
 507             jnrA             = jjnr[jidx];
 508             jnrB             = jjnr[jidx+1];
 509             jnrC             = jjnr[jidx+2];
 510             jnrD             = jjnr[jidx+3];
 511             j_coord_offsetA  = DIM*jnrA;
 512             j_coord_offsetB  = DIM*jnrB;
 513             j_coord_offsetC  = DIM*jnrC;
 514             j_coord_offsetD  = DIM*jnrD;
 515
 516             /* load j atom coordinates */
 517             gmx_mm256_load_1rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 518                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 519                                                  &jx0,&jy0,&jz0);
 520
 521             /* Calculate displacement vector */
 522             dx00             = _mm256_sub_pd(ix0,jx0);
 523             dy00             = _mm256_sub_pd(iy0,jy0);
 524             dz00             = _mm256_sub_pd(iz0,jz0);
 525
 526             /* Calculate squared distance and things based on it */
 527             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 528
 529             rinv00           = gmx_mm256_invsqrt_pd(rsq00);
 530
 531             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
 532
 533             /* Load parameters for j particles */
 534             jq0              = gmx_mm256_load_4real_swizzle_pd(charge+jnrA+0,charge+jnrB+0,
 535                                                                  charge+jnrC+0,charge+jnrD+0);
 536
 537             /**************************
 538              * CALCULATE INTERACTIONS *
 539              **************************/
 540
 541             if (gmx_mm256_any_lt(rsq00,rcutoff2))
 542             {
 543
 544             r00              = _mm256_mul_pd(rsq00,rinv00);
 545
 546             /* Compute parameters for interactions between i and j atoms */
 547             qq00             = _mm256_mul_pd(iq0,jq0);
 548
 549             /* EWALD ELECTROSTATICS */
 550
 551             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 552             ewrt             = _mm256_mul_pd(r00,ewtabscale);
 553             ewitab           = _mm256_cvttpd_epi32(ewrt);
 554             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 555             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
 556                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
 557                                             &ewtabF,&ewtabFn);
 558             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
 559             felec            = _mm256_mul_pd(_mm256_mul_pd(qq00,rinv00),_mm256_sub_pd(rinvsq00,felec));
 560
 561             cutoff_mask      = _mm256_cmp_pd(rsq00,rcutoff2,_CMP_LT_OQ);
 562
 563             fscal            = felec;
 564
 565             fscal            = _mm256_and_pd(fscal,cutoff_mask);
 566
 567             /* Calculate temporary vectorial force */
 568             tx               = _mm256_mul_pd(fscal,dx00);
 569             ty               = _mm256_mul_pd(fscal,dy00);
 570             tz               = _mm256_mul_pd(fscal,dz00);
 571
 572             /* Update vectorial force */
 573             fix0             = _mm256_add_pd(fix0,tx);
 574             fiy0             = _mm256_add_pd(fiy0,ty);
 575             fiz0             = _mm256_add_pd(fiz0,tz);
 576
 577             fjptrA             = f+j_coord_offsetA;
 578             fjptrB             = f+j_coord_offsetB;
 579             fjptrC             = f+j_coord_offsetC;
 580             fjptrD             = f+j_coord_offsetD;
 581             gmx_mm256_decrement_1rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,tx,ty,tz);
 582
 583             }
 584
 585             /* Inner loop uses 39 flops */
 586         }
 587
 588         if(jidx<j_index_end)
 589         {
 590
 591             /* Get j neighbor index, and coordinate index */
 592             jnrlistA         = jjnr[jidx];
 593             jnrlistB         = jjnr[jidx+1];
 594             jnrlistC         = jjnr[jidx+2];
 595             jnrlistD         = jjnr[jidx+3];
 596             /* Sign of each element will be negative for non-real atoms.
 597              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 598              * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
 599              */
 600             tmpmask0 = gmx_mm_castsi128_pd(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 601
 602             tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
 603             tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
 604             dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
 605
 606             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 607             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 608             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 609             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 610             j_coord_offsetA  = DIM*jnrA;
 611             j_coord_offsetB  = DIM*jnrB;
 612             j_coord_offsetC  = DIM*jnrC;
 613             j_coord_offsetD  = DIM*jnrD;
 614
 615             /* load j atom coordinates */
 616             gmx_mm256_load_1rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 617                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 618                                                  &jx0,&jy0,&jz0);
 619
 620             /* Calculate displacement vector */
 621             dx00             = _mm256_sub_pd(ix0,jx0);
 622             dy00             = _mm256_sub_pd(iy0,jy0);
 623             dz00             = _mm256_sub_pd(iz0,jz0);
 624
 625             /* Calculate squared distance and things based on it */
 626             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 627
 628             rinv00           = gmx_mm256_invsqrt_pd(rsq00);
 629
 630             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
 631
 632             /* Load parameters for j particles */
 633             jq0              = gmx_mm256_load_4real_swizzle_pd(charge+jnrA+0,charge+jnrB+0,
 634                                                                  charge+jnrC+0,charge+jnrD+0);
 635
 636             /**************************
 637              * CALCULATE INTERACTIONS *
 638              **************************/
 639
 640             if (gmx_mm256_any_lt(rsq00,rcutoff2))
 641             {
 642
 643             r00              = _mm256_mul_pd(rsq00,rinv00);
 644             r00              = _mm256_andnot_pd(dummy_mask,r00);
 645
 646             /* Compute parameters for interactions between i and j atoms */
 647             qq00             = _mm256_mul_pd(iq0,jq0);
 648
 649             /* EWALD ELECTROSTATICS */
 650
 651             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 652             ewrt             = _mm256_mul_pd(r00,ewtabscale);
 653             ewitab           = _mm256_cvttpd_epi32(ewrt);
 654             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 655             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
 656                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
 657                                             &ewtabF,&ewtabFn);
 658             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
 659             felec            = _mm256_mul_pd(_mm256_mul_pd(qq00,rinv00),_mm256_sub_pd(rinvsq00,felec));
 660
 661             cutoff_mask      = _mm256_cmp_pd(rsq00,rcutoff2,_CMP_LT_OQ);
 662
 663             fscal            = felec;
 664
 665             fscal            = _mm256_and_pd(fscal,cutoff_mask);
 666
 667             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 668
 669             /* Calculate temporary vectorial force */
 670             tx               = _mm256_mul_pd(fscal,dx00);
 671             ty               = _mm256_mul_pd(fscal,dy00);
 672             tz               = _mm256_mul_pd(fscal,dz00);
 673
 674             /* Update vectorial force */
 675             fix0             = _mm256_add_pd(fix0,tx);
 676             fiy0             = _mm256_add_pd(fiy0,ty);
 677             fiz0             = _mm256_add_pd(fiz0,tz);
 678
 679             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
 680             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
 681             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
 682             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
 683             gmx_mm256_decrement_1rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,tx,ty,tz);
 684
 685             }
 686
 687             /* Inner loop uses 40 flops */
 688         }
 689
 690         /* End of innermost loop */
 691
 692         gmx_mm256_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
 693                                                  f+i_coord_offset,fshift+i_shift_offset);
 694
 695         /* Increment number of inner iterations */
 696         inneriter                  += j_index_end - j_index_start;
 697
 698         /* Outer loop uses 7 flops */
 699     }
 700
 701     /* Increment number of outer iterations */
 702     outeriter        += nri;
 703
 704     /* Update outer/inner flops */
 705
 706     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_F,outeriter*7 + inneriter*40);
 707 }