src/gromacs/gmxlib/nonbonded/nb_kernel_avx_128_fma_double/nb_kernel_ElecEwSh_VdwNone_GeomW4P1_avx_128_fma_double.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS avx_128_fma_double kernel generator.
  37  */
  38 #include "config.h"
  39
  40 #include <math.h>
  41
  42 #include "../nb_kernel.h"
  43 #include "types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "nrnb.h"
  46
  47 #include "gromacs/simd/math_x86_avx_128_fma_double.h"
  48 #include "kernelutil_x86_avx_128_fma_double.h"
  49
  50 /*
  51  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSh_VdwNone_GeomW4P1_VF_avx_128_fma_double
  52  * Electrostatics interaction: Ewald
  53  * VdW interaction:            None
  54  * Geometry:                   Water4-Particle
  55  * Calculate force/pot:        PotentialAndForce
  56  */
  57 void
  58 nb_kernel_ElecEwSh_VdwNone_GeomW4P1_VF_avx_128_fma_double
  59                     (t_nblist                    * gmx_restrict       nlist,
  60                      rvec                        * gmx_restrict          xx,
  61                      rvec                        * gmx_restrict          ff,
  62                      t_forcerec                  * gmx_restrict          fr,
  63                      t_mdatoms                   * gmx_restrict     mdatoms,
  64                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  65                      t_nrnb                      * gmx_restrict        nrnb)
  66 {
  67     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  68      * just 0 for non-waters.
  69      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
  70      * jnr indices corresponding to data put in the four positions in the SIMD register.
  71      */
  72     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  73     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  74     int              jnrA,jnrB;
  75     int              j_coord_offsetA,j_coord_offsetB;
  76     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  77     real             rcutoff_scalar;
  78     real             *shiftvec,*fshift,*x,*f;
  79     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  80     int              vdwioffset1;
  81     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  82     int              vdwioffset2;
  83     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  84     int              vdwioffset3;
  85     __m128d          ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  86     int              vdwjidx0A,vdwjidx0B;
  87     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  88     __m128d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  89     __m128d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  90     __m128d          dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
  91     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
  92     real             *charge;
  93     __m128i          ewitab;
  94     __m128d          ewtabscale,eweps,twoeweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
  95     real             *ewtab;
  96     __m128d          dummy_mask,cutoff_mask;
  97     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
  98     __m128d          one     = _mm_set1_pd(1.0);
  99     __m128d          two     = _mm_set1_pd(2.0);
 100     x                = xx[0];
 101     f                = ff[0];
 102
 103     nri              = nlist->nri;
 104     iinr             = nlist->iinr;
 105     jindex           = nlist->jindex;
 106     jjnr             = nlist->jjnr;
 107     shiftidx         = nlist->shift;
 108     gid              = nlist->gid;
 109     shiftvec         = fr->shift_vec[0];
 110     fshift           = fr->fshift[0];
 111     facel            = _mm_set1_pd(fr->epsfac);
 112     charge           = mdatoms->chargeA;
 113
 114     sh_ewald         = _mm_set1_pd(fr->ic->sh_ewald);
 115     ewtab            = fr->ic->tabq_coul_FDV0;
 116     ewtabscale       = _mm_set1_pd(fr->ic->tabq_scale);
 117     ewtabhalfspace   = _mm_set1_pd(0.5/fr->ic->tabq_scale);
 118
 119     /* Setup water-specific parameters */
 120     inr              = nlist->iinr[0];
 121     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
 122     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
 123     iq3              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
 124
 125     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 126     rcutoff_scalar   = fr->rcoulomb;
 127     rcutoff          = _mm_set1_pd(rcutoff_scalar);
 128     rcutoff2         = _mm_mul_pd(rcutoff,rcutoff);
 129
 130     /* Avoid stupid compiler warnings */
 131     jnrA = jnrB = 0;
 132     j_coord_offsetA = 0;
 133     j_coord_offsetB = 0;
 134
 135     outeriter        = 0;
 136     inneriter        = 0;
 137
 138     /* Start outer loop over neighborlists */
 139     for(iidx=0; iidx<nri; iidx++)
 140     {
 141         /* Load shift vector for this list */
 142         i_shift_offset   = DIM*shiftidx[iidx];
 143
 144         /* Load limits for loop over neighbors */
 145         j_index_start    = jindex[iidx];
 146         j_index_end      = jindex[iidx+1];
 147
 148         /* Get outer coordinate index */
 149         inr              = iinr[iidx];
 150         i_coord_offset   = DIM*inr;
 151
 152         /* Load i particle coords and add shift vector */
 153         gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
 154                                                  &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 155
 156         fix1             = _mm_setzero_pd();
 157         fiy1             = _mm_setzero_pd();
 158         fiz1             = _mm_setzero_pd();
 159         fix2             = _mm_setzero_pd();
 160         fiy2             = _mm_setzero_pd();
 161         fiz2             = _mm_setzero_pd();
 162         fix3             = _mm_setzero_pd();
 163         fiy3             = _mm_setzero_pd();
 164         fiz3             = _mm_setzero_pd();
 165
 166         /* Reset potential sums */
 167         velecsum         = _mm_setzero_pd();
 168
 169         /* Start inner kernel loop */
 170         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 171         {
 172
 173             /* Get j neighbor index, and coordinate index */
 174             jnrA             = jjnr[jidx];
 175             jnrB             = jjnr[jidx+1];
 176             j_coord_offsetA  = DIM*jnrA;
 177             j_coord_offsetB  = DIM*jnrB;
 178
 179             /* load j atom coordinates */
 180             gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 181                                               &jx0,&jy0,&jz0);
 182
 183             /* Calculate displacement vector */
 184             dx10             = _mm_sub_pd(ix1,jx0);
 185             dy10             = _mm_sub_pd(iy1,jy0);
 186             dz10             = _mm_sub_pd(iz1,jz0);
 187             dx20             = _mm_sub_pd(ix2,jx0);
 188             dy20             = _mm_sub_pd(iy2,jy0);
 189             dz20             = _mm_sub_pd(iz2,jz0);
 190             dx30             = _mm_sub_pd(ix3,jx0);
 191             dy30             = _mm_sub_pd(iy3,jy0);
 192             dz30             = _mm_sub_pd(iz3,jz0);
 193
 194             /* Calculate squared distance and things based on it */
 195             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 196             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 197             rsq30            = gmx_mm_calc_rsq_pd(dx30,dy30,dz30);
 198
 199             rinv10           = gmx_mm_invsqrt_pd(rsq10);
 200             rinv20           = gmx_mm_invsqrt_pd(rsq20);
 201             rinv30           = gmx_mm_invsqrt_pd(rsq30);
 202
 203             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
 204             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
 205             rinvsq30         = _mm_mul_pd(rinv30,rinv30);
 206
 207             /* Load parameters for j particles */
 208             jq0              = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
 209
 210             fjx0             = _mm_setzero_pd();
 211             fjy0             = _mm_setzero_pd();
 212             fjz0             = _mm_setzero_pd();
 213
 214             /**************************
 215              * CALCULATE INTERACTIONS *
 216              **************************/
 217
 218             if (gmx_mm_any_lt(rsq10,rcutoff2))
 219             {
 220
 221             r10              = _mm_mul_pd(rsq10,rinv10);
 222
 223             /* Compute parameters for interactions between i and j atoms */
 224             qq10             = _mm_mul_pd(iq1,jq0);
 225
 226             /* EWALD ELECTROSTATICS */
 227
 228             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 229             ewrt             = _mm_mul_pd(r10,ewtabscale);
 230             ewitab           = _mm_cvttpd_epi32(ewrt);
 231 #ifdef __XOP__
 232             eweps            = _mm_frcz_pd(ewrt);
 233 #else
 234             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 235 #endif
 236             twoeweps         = _mm_add_pd(eweps,eweps);
 237             ewitab           = _mm_slli_epi32(ewitab,2);
 238             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 239             ewtabD           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 240             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 241             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
 242             ewtabFn          = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
 243             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 244             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
 245             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
 246             velec            = _mm_mul_pd(qq10,_mm_sub_pd(_mm_sub_pd(rinv10,sh_ewald),velec));
 247             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
 248
 249             cutoff_mask      = _mm_cmplt_pd(rsq10,rcutoff2);
 250
 251             /* Update potential sum for this i atom from the interaction with this j atom. */
 252             velec            = _mm_and_pd(velec,cutoff_mask);
 253             velecsum         = _mm_add_pd(velecsum,velec);
 254
 255             fscal            = felec;
 256
 257             fscal            = _mm_and_pd(fscal,cutoff_mask);
 258
 259             /* Update vectorial force */
 260             fix1             = _mm_macc_pd(dx10,fscal,fix1);
 261             fiy1             = _mm_macc_pd(dy10,fscal,fiy1);
 262             fiz1             = _mm_macc_pd(dz10,fscal,fiz1);
 263
 264             fjx0             = _mm_macc_pd(dx10,fscal,fjx0);
 265             fjy0             = _mm_macc_pd(dy10,fscal,fjy0);
 266             fjz0             = _mm_macc_pd(dz10,fscal,fjz0);
 267
 268             }
 269
 270             /**************************
 271              * CALCULATE INTERACTIONS *
 272              **************************/
 273
 274             if (gmx_mm_any_lt(rsq20,rcutoff2))
 275             {
 276
 277             r20              = _mm_mul_pd(rsq20,rinv20);
 278
 279             /* Compute parameters for interactions between i and j atoms */
 280             qq20             = _mm_mul_pd(iq2,jq0);
 281
 282             /* EWALD ELECTROSTATICS */
 283
 284             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 285             ewrt             = _mm_mul_pd(r20,ewtabscale);
 286             ewitab           = _mm_cvttpd_epi32(ewrt);
 287 #ifdef __XOP__
 288             eweps            = _mm_frcz_pd(ewrt);
 289 #else
 290             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 291 #endif
 292             twoeweps         = _mm_add_pd(eweps,eweps);
 293             ewitab           = _mm_slli_epi32(ewitab,2);
 294             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 295             ewtabD           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 296             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 297             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
 298             ewtabFn          = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
 299             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 300             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
 301             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
 302             velec            = _mm_mul_pd(qq20,_mm_sub_pd(_mm_sub_pd(rinv20,sh_ewald),velec));
 303             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
 304
 305             cutoff_mask      = _mm_cmplt_pd(rsq20,rcutoff2);
 306
 307             /* Update potential sum for this i atom from the interaction with this j atom. */
 308             velec            = _mm_and_pd(velec,cutoff_mask);
 309             velecsum         = _mm_add_pd(velecsum,velec);
 310
 311             fscal            = felec;
 312
 313             fscal            = _mm_and_pd(fscal,cutoff_mask);
 314
 315             /* Update vectorial force */
 316             fix2             = _mm_macc_pd(dx20,fscal,fix2);
 317             fiy2             = _mm_macc_pd(dy20,fscal,fiy2);
 318             fiz2             = _mm_macc_pd(dz20,fscal,fiz2);
 319
 320             fjx0             = _mm_macc_pd(dx20,fscal,fjx0);
 321             fjy0             = _mm_macc_pd(dy20,fscal,fjy0);
 322             fjz0             = _mm_macc_pd(dz20,fscal,fjz0);
 323
 324             }
 325
 326             /**************************
 327              * CALCULATE INTERACTIONS *
 328              **************************/
 329
 330             if (gmx_mm_any_lt(rsq30,rcutoff2))
 331             {
 332
 333             r30              = _mm_mul_pd(rsq30,rinv30);
 334
 335             /* Compute parameters for interactions between i and j atoms */
 336             qq30             = _mm_mul_pd(iq3,jq0);
 337
 338             /* EWALD ELECTROSTATICS */
 339
 340             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 341             ewrt             = _mm_mul_pd(r30,ewtabscale);
 342             ewitab           = _mm_cvttpd_epi32(ewrt);
 343 #ifdef __XOP__
 344             eweps            = _mm_frcz_pd(ewrt);
 345 #else
 346             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 347 #endif
 348             twoeweps         = _mm_add_pd(eweps,eweps);
 349             ewitab           = _mm_slli_epi32(ewitab,2);
 350             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 351             ewtabD           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 352             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 353             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
 354             ewtabFn          = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
 355             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 356             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
 357             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
 358             velec            = _mm_mul_pd(qq30,_mm_sub_pd(_mm_sub_pd(rinv30,sh_ewald),velec));
 359             felec            = _mm_mul_pd(_mm_mul_pd(qq30,rinv30),_mm_sub_pd(rinvsq30,felec));
 360
 361             cutoff_mask      = _mm_cmplt_pd(rsq30,rcutoff2);
 362
 363             /* Update potential sum for this i atom from the interaction with this j atom. */
 364             velec            = _mm_and_pd(velec,cutoff_mask);
 365             velecsum         = _mm_add_pd(velecsum,velec);
 366
 367             fscal            = felec;
 368
 369             fscal            = _mm_and_pd(fscal,cutoff_mask);
 370
 371             /* Update vectorial force */
 372             fix3             = _mm_macc_pd(dx30,fscal,fix3);
 373             fiy3             = _mm_macc_pd(dy30,fscal,fiy3);
 374             fiz3             = _mm_macc_pd(dz30,fscal,fiz3);
 375
 376             fjx0             = _mm_macc_pd(dx30,fscal,fjx0);
 377             fjy0             = _mm_macc_pd(dy30,fscal,fjy0);
 378             fjz0             = _mm_macc_pd(dz30,fscal,fjz0);
 379
 380             }
 381
 382             gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 383
 384             /* Inner loop uses 150 flops */
 385         }
 386
 387         if(jidx<j_index_end)
 388         {
 389
 390             jnrA             = jjnr[jidx];
 391             j_coord_offsetA  = DIM*jnrA;
 392
 393             /* load j atom coordinates */
 394             gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 395                                               &jx0,&jy0,&jz0);
 396
 397             /* Calculate displacement vector */
 398             dx10             = _mm_sub_pd(ix1,jx0);
 399             dy10             = _mm_sub_pd(iy1,jy0);
 400             dz10             = _mm_sub_pd(iz1,jz0);
 401             dx20             = _mm_sub_pd(ix2,jx0);
 402             dy20             = _mm_sub_pd(iy2,jy0);
 403             dz20             = _mm_sub_pd(iz2,jz0);
 404             dx30             = _mm_sub_pd(ix3,jx0);
 405             dy30             = _mm_sub_pd(iy3,jy0);
 406             dz30             = _mm_sub_pd(iz3,jz0);
 407
 408             /* Calculate squared distance and things based on it */
 409             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 410             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 411             rsq30            = gmx_mm_calc_rsq_pd(dx30,dy30,dz30);
 412
 413             rinv10           = gmx_mm_invsqrt_pd(rsq10);
 414             rinv20           = gmx_mm_invsqrt_pd(rsq20);
 415             rinv30           = gmx_mm_invsqrt_pd(rsq30);
 416
 417             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
 418             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
 419             rinvsq30         = _mm_mul_pd(rinv30,rinv30);
 420
 421             /* Load parameters for j particles */
 422             jq0              = _mm_load_sd(charge+jnrA+0);
 423
 424             fjx0             = _mm_setzero_pd();
 425             fjy0             = _mm_setzero_pd();
 426             fjz0             = _mm_setzero_pd();
 427
 428             /**************************
 429              * CALCULATE INTERACTIONS *
 430              **************************/
 431
 432             if (gmx_mm_any_lt(rsq10,rcutoff2))
 433             {
 434
 435             r10              = _mm_mul_pd(rsq10,rinv10);
 436
 437             /* Compute parameters for interactions between i and j atoms */
 438             qq10             = _mm_mul_pd(iq1,jq0);
 439
 440             /* EWALD ELECTROSTATICS */
 441
 442             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 443             ewrt             = _mm_mul_pd(r10,ewtabscale);
 444             ewitab           = _mm_cvttpd_epi32(ewrt);
 445 #ifdef __XOP__
 446             eweps            = _mm_frcz_pd(ewrt);
 447 #else
 448             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 449 #endif
 450             twoeweps         = _mm_add_pd(eweps,eweps);
 451             ewitab           = _mm_slli_epi32(ewitab,2);
 452             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 453             ewtabD           = _mm_setzero_pd();
 454             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 455             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
 456             ewtabFn          = _mm_setzero_pd();
 457             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 458             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
 459             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
 460             velec            = _mm_mul_pd(qq10,_mm_sub_pd(_mm_sub_pd(rinv10,sh_ewald),velec));
 461             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
 462
 463             cutoff_mask      = _mm_cmplt_pd(rsq10,rcutoff2);
 464
 465             /* Update potential sum for this i atom from the interaction with this j atom. */
 466             velec            = _mm_and_pd(velec,cutoff_mask);
 467             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 468             velecsum         = _mm_add_pd(velecsum,velec);
 469
 470             fscal            = felec;
 471
 472             fscal            = _mm_and_pd(fscal,cutoff_mask);
 473
 474             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 475
 476             /* Update vectorial force */
 477             fix1             = _mm_macc_pd(dx10,fscal,fix1);
 478             fiy1             = _mm_macc_pd(dy10,fscal,fiy1);
 479             fiz1             = _mm_macc_pd(dz10,fscal,fiz1);
 480
 481             fjx0             = _mm_macc_pd(dx10,fscal,fjx0);
 482             fjy0             = _mm_macc_pd(dy10,fscal,fjy0);
 483             fjz0             = _mm_macc_pd(dz10,fscal,fjz0);
 484
 485             }
 486
 487             /**************************
 488              * CALCULATE INTERACTIONS *
 489              **************************/
 490
 491             if (gmx_mm_any_lt(rsq20,rcutoff2))
 492             {
 493
 494             r20              = _mm_mul_pd(rsq20,rinv20);
 495
 496             /* Compute parameters for interactions between i and j atoms */
 497             qq20             = _mm_mul_pd(iq2,jq0);
 498
 499             /* EWALD ELECTROSTATICS */
 500
 501             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 502             ewrt             = _mm_mul_pd(r20,ewtabscale);
 503             ewitab           = _mm_cvttpd_epi32(ewrt);
 504 #ifdef __XOP__
 505             eweps            = _mm_frcz_pd(ewrt);
 506 #else
 507             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 508 #endif
 509             twoeweps         = _mm_add_pd(eweps,eweps);
 510             ewitab           = _mm_slli_epi32(ewitab,2);
 511             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 512             ewtabD           = _mm_setzero_pd();
 513             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 514             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
 515             ewtabFn          = _mm_setzero_pd();
 516             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 517             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
 518             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
 519             velec            = _mm_mul_pd(qq20,_mm_sub_pd(_mm_sub_pd(rinv20,sh_ewald),velec));
 520             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
 521
 522             cutoff_mask      = _mm_cmplt_pd(rsq20,rcutoff2);
 523
 524             /* Update potential sum for this i atom from the interaction with this j atom. */
 525             velec            = _mm_and_pd(velec,cutoff_mask);
 526             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 527             velecsum         = _mm_add_pd(velecsum,velec);
 528
 529             fscal            = felec;
 530
 531             fscal            = _mm_and_pd(fscal,cutoff_mask);
 532
 533             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 534
 535             /* Update vectorial force */
 536             fix2             = _mm_macc_pd(dx20,fscal,fix2);
 537             fiy2             = _mm_macc_pd(dy20,fscal,fiy2);
 538             fiz2             = _mm_macc_pd(dz20,fscal,fiz2);
 539
 540             fjx0             = _mm_macc_pd(dx20,fscal,fjx0);
 541             fjy0             = _mm_macc_pd(dy20,fscal,fjy0);
 542             fjz0             = _mm_macc_pd(dz20,fscal,fjz0);
 543
 544             }
 545
 546             /**************************
 547              * CALCULATE INTERACTIONS *
 548              **************************/
 549
 550             if (gmx_mm_any_lt(rsq30,rcutoff2))
 551             {
 552
 553             r30              = _mm_mul_pd(rsq30,rinv30);
 554
 555             /* Compute parameters for interactions between i and j atoms */
 556             qq30             = _mm_mul_pd(iq3,jq0);
 557
 558             /* EWALD ELECTROSTATICS */
 559
 560             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 561             ewrt             = _mm_mul_pd(r30,ewtabscale);
 562             ewitab           = _mm_cvttpd_epi32(ewrt);
 563 #ifdef __XOP__
 564             eweps            = _mm_frcz_pd(ewrt);
 565 #else
 566             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 567 #endif
 568             twoeweps         = _mm_add_pd(eweps,eweps);
 569             ewitab           = _mm_slli_epi32(ewitab,2);
 570             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 571             ewtabD           = _mm_setzero_pd();
 572             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 573             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
 574             ewtabFn          = _mm_setzero_pd();
 575             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 576             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
 577             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
 578             velec            = _mm_mul_pd(qq30,_mm_sub_pd(_mm_sub_pd(rinv30,sh_ewald),velec));
 579             felec            = _mm_mul_pd(_mm_mul_pd(qq30,rinv30),_mm_sub_pd(rinvsq30,felec));
 580
 581             cutoff_mask      = _mm_cmplt_pd(rsq30,rcutoff2);
 582
 583             /* Update potential sum for this i atom from the interaction with this j atom. */
 584             velec            = _mm_and_pd(velec,cutoff_mask);
 585             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 586             velecsum         = _mm_add_pd(velecsum,velec);
 587
 588             fscal            = felec;
 589
 590             fscal            = _mm_and_pd(fscal,cutoff_mask);
 591
 592             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 593
 594             /* Update vectorial force */
 595             fix3             = _mm_macc_pd(dx30,fscal,fix3);
 596             fiy3             = _mm_macc_pd(dy30,fscal,fiy3);
 597             fiz3             = _mm_macc_pd(dz30,fscal,fiz3);
 598
 599             fjx0             = _mm_macc_pd(dx30,fscal,fjx0);
 600             fjy0             = _mm_macc_pd(dy30,fscal,fjy0);
 601             fjz0             = _mm_macc_pd(dz30,fscal,fjz0);
 602
 603             }
 604
 605             gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 606
 607             /* Inner loop uses 150 flops */
 608         }
 609
 610         /* End of innermost loop */
 611
 612         gmx_mm_update_iforce_3atom_swizzle_pd(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 613                                               f+i_coord_offset+DIM,fshift+i_shift_offset);
 614
 615         ggid                        = gid[iidx];
 616         /* Update potential energies */
 617         gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
 618
 619         /* Increment number of inner iterations */
 620         inneriter                  += j_index_end - j_index_start;
 621
 622         /* Outer loop uses 19 flops */
 623     }
 624
 625     /* Increment number of outer iterations */
 626     outeriter        += nri;
 627
 628     /* Update outer/inner flops */
 629
 630     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4_VF,outeriter*19 + inneriter*150);
 631 }
 632 /*
 633  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSh_VdwNone_GeomW4P1_F_avx_128_fma_double
 634  * Electrostatics interaction: Ewald
 635  * VdW interaction:            None
 636  * Geometry:                   Water4-Particle
 637  * Calculate force/pot:        Force
 638  */
 639 void
 640 nb_kernel_ElecEwSh_VdwNone_GeomW4P1_F_avx_128_fma_double
 641                     (t_nblist                    * gmx_restrict       nlist,
 642                      rvec                        * gmx_restrict          xx,
 643                      rvec                        * gmx_restrict          ff,
 644                      t_forcerec                  * gmx_restrict          fr,
 645                      t_mdatoms                   * gmx_restrict     mdatoms,
 646                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 647                      t_nrnb                      * gmx_restrict        nrnb)
 648 {
 649     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 650      * just 0 for non-waters.
 651      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
 652      * jnr indices corresponding to data put in the four positions in the SIMD register.
 653      */
 654     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 655     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 656     int              jnrA,jnrB;
 657     int              j_coord_offsetA,j_coord_offsetB;
 658     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 659     real             rcutoff_scalar;
 660     real             *shiftvec,*fshift,*x,*f;
 661     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 662     int              vdwioffset1;
 663     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 664     int              vdwioffset2;
 665     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 666     int              vdwioffset3;
 667     __m128d          ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
 668     int              vdwjidx0A,vdwjidx0B;
 669     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 670     __m128d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 671     __m128d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 672     __m128d          dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
 673     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
 674     real             *charge;
 675     __m128i          ewitab;
 676     __m128d          ewtabscale,eweps,twoeweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 677     real             *ewtab;
 678     __m128d          dummy_mask,cutoff_mask;
 679     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 680     __m128d          one     = _mm_set1_pd(1.0);
 681     __m128d          two     = _mm_set1_pd(2.0);
 682     x                = xx[0];
 683     f                = ff[0];
 684
 685     nri              = nlist->nri;
 686     iinr             = nlist->iinr;
 687     jindex           = nlist->jindex;
 688     jjnr             = nlist->jjnr;
 689     shiftidx         = nlist->shift;
 690     gid              = nlist->gid;
 691     shiftvec         = fr->shift_vec[0];
 692     fshift           = fr->fshift[0];
 693     facel            = _mm_set1_pd(fr->epsfac);
 694     charge           = mdatoms->chargeA;
 695
 696     sh_ewald         = _mm_set1_pd(fr->ic->sh_ewald);
 697     ewtab            = fr->ic->tabq_coul_F;
 698     ewtabscale       = _mm_set1_pd(fr->ic->tabq_scale);
 699     ewtabhalfspace   = _mm_set1_pd(0.5/fr->ic->tabq_scale);
 700
 701     /* Setup water-specific parameters */
 702     inr              = nlist->iinr[0];
 703     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
 704     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
 705     iq3              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
 706
 707     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 708     rcutoff_scalar   = fr->rcoulomb;
 709     rcutoff          = _mm_set1_pd(rcutoff_scalar);
 710     rcutoff2         = _mm_mul_pd(rcutoff,rcutoff);
 711
 712     /* Avoid stupid compiler warnings */
 713     jnrA = jnrB = 0;
 714     j_coord_offsetA = 0;
 715     j_coord_offsetB = 0;
 716
 717     outeriter        = 0;
 718     inneriter        = 0;
 719
 720     /* Start outer loop over neighborlists */
 721     for(iidx=0; iidx<nri; iidx++)
 722     {
 723         /* Load shift vector for this list */
 724         i_shift_offset   = DIM*shiftidx[iidx];
 725
 726         /* Load limits for loop over neighbors */
 727         j_index_start    = jindex[iidx];
 728         j_index_end      = jindex[iidx+1];
 729
 730         /* Get outer coordinate index */
 731         inr              = iinr[iidx];
 732         i_coord_offset   = DIM*inr;
 733
 734         /* Load i particle coords and add shift vector */
 735         gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
 736                                                  &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 737
 738         fix1             = _mm_setzero_pd();
 739         fiy1             = _mm_setzero_pd();
 740         fiz1             = _mm_setzero_pd();
 741         fix2             = _mm_setzero_pd();
 742         fiy2             = _mm_setzero_pd();
 743         fiz2             = _mm_setzero_pd();
 744         fix3             = _mm_setzero_pd();
 745         fiy3             = _mm_setzero_pd();
 746         fiz3             = _mm_setzero_pd();
 747
 748         /* Start inner kernel loop */
 749         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 750         {
 751
 752             /* Get j neighbor index, and coordinate index */
 753             jnrA             = jjnr[jidx];
 754             jnrB             = jjnr[jidx+1];
 755             j_coord_offsetA  = DIM*jnrA;
 756             j_coord_offsetB  = DIM*jnrB;
 757
 758             /* load j atom coordinates */
 759             gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 760                                               &jx0,&jy0,&jz0);
 761
 762             /* Calculate displacement vector */
 763             dx10             = _mm_sub_pd(ix1,jx0);
 764             dy10             = _mm_sub_pd(iy1,jy0);
 765             dz10             = _mm_sub_pd(iz1,jz0);
 766             dx20             = _mm_sub_pd(ix2,jx0);
 767             dy20             = _mm_sub_pd(iy2,jy0);
 768             dz20             = _mm_sub_pd(iz2,jz0);
 769             dx30             = _mm_sub_pd(ix3,jx0);
 770             dy30             = _mm_sub_pd(iy3,jy0);
 771             dz30             = _mm_sub_pd(iz3,jz0);
 772
 773             /* Calculate squared distance and things based on it */
 774             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 775             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 776             rsq30            = gmx_mm_calc_rsq_pd(dx30,dy30,dz30);
 777
 778             rinv10           = gmx_mm_invsqrt_pd(rsq10);
 779             rinv20           = gmx_mm_invsqrt_pd(rsq20);
 780             rinv30           = gmx_mm_invsqrt_pd(rsq30);
 781
 782             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
 783             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
 784             rinvsq30         = _mm_mul_pd(rinv30,rinv30);
 785
 786             /* Load parameters for j particles */
 787             jq0              = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
 788
 789             fjx0             = _mm_setzero_pd();
 790             fjy0             = _mm_setzero_pd();
 791             fjz0             = _mm_setzero_pd();
 792
 793             /**************************
 794              * CALCULATE INTERACTIONS *
 795              **************************/
 796
 797             if (gmx_mm_any_lt(rsq10,rcutoff2))
 798             {
 799
 800             r10              = _mm_mul_pd(rsq10,rinv10);
 801
 802             /* Compute parameters for interactions between i and j atoms */
 803             qq10             = _mm_mul_pd(iq1,jq0);
 804
 805             /* EWALD ELECTROSTATICS */
 806
 807             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 808             ewrt             = _mm_mul_pd(r10,ewtabscale);
 809             ewitab           = _mm_cvttpd_epi32(ewrt);
 810 #ifdef __XOP__
 811             eweps            = _mm_frcz_pd(ewrt);
 812 #else
 813             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 814 #endif
 815             twoeweps         = _mm_add_pd(eweps,eweps);
 816             gmx_mm_load_2pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),ewtab+_mm_extract_epi32(ewitab,1),
 817                                          &ewtabF,&ewtabFn);
 818             felec            = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
 819             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
 820
 821             cutoff_mask      = _mm_cmplt_pd(rsq10,rcutoff2);
 822
 823             fscal            = felec;
 824
 825             fscal            = _mm_and_pd(fscal,cutoff_mask);
 826
 827             /* Update vectorial force */
 828             fix1             = _mm_macc_pd(dx10,fscal,fix1);
 829             fiy1             = _mm_macc_pd(dy10,fscal,fiy1);
 830             fiz1             = _mm_macc_pd(dz10,fscal,fiz1);
 831
 832             fjx0             = _mm_macc_pd(dx10,fscal,fjx0);
 833             fjy0             = _mm_macc_pd(dy10,fscal,fjy0);
 834             fjz0             = _mm_macc_pd(dz10,fscal,fjz0);
 835
 836             }
 837
 838             /**************************
 839              * CALCULATE INTERACTIONS *
 840              **************************/
 841
 842             if (gmx_mm_any_lt(rsq20,rcutoff2))
 843             {
 844
 845             r20              = _mm_mul_pd(rsq20,rinv20);
 846
 847             /* Compute parameters for interactions between i and j atoms */
 848             qq20             = _mm_mul_pd(iq2,jq0);
 849
 850             /* EWALD ELECTROSTATICS */
 851
 852             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 853             ewrt             = _mm_mul_pd(r20,ewtabscale);
 854             ewitab           = _mm_cvttpd_epi32(ewrt);
 855 #ifdef __XOP__
 856             eweps            = _mm_frcz_pd(ewrt);
 857 #else
 858             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 859 #endif
 860             twoeweps         = _mm_add_pd(eweps,eweps);
 861             gmx_mm_load_2pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),ewtab+_mm_extract_epi32(ewitab,1),
 862                                          &ewtabF,&ewtabFn);
 863             felec            = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
 864             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
 865
 866             cutoff_mask      = _mm_cmplt_pd(rsq20,rcutoff2);
 867
 868             fscal            = felec;
 869
 870             fscal            = _mm_and_pd(fscal,cutoff_mask);
 871
 872             /* Update vectorial force */
 873             fix2             = _mm_macc_pd(dx20,fscal,fix2);
 874             fiy2             = _mm_macc_pd(dy20,fscal,fiy2);
 875             fiz2             = _mm_macc_pd(dz20,fscal,fiz2);
 876
 877             fjx0             = _mm_macc_pd(dx20,fscal,fjx0);
 878             fjy0             = _mm_macc_pd(dy20,fscal,fjy0);
 879             fjz0             = _mm_macc_pd(dz20,fscal,fjz0);
 880
 881             }
 882
 883             /**************************
 884              * CALCULATE INTERACTIONS *
 885              **************************/
 886
 887             if (gmx_mm_any_lt(rsq30,rcutoff2))
 888             {
 889
 890             r30              = _mm_mul_pd(rsq30,rinv30);
 891
 892             /* Compute parameters for interactions between i and j atoms */
 893             qq30             = _mm_mul_pd(iq3,jq0);
 894
 895             /* EWALD ELECTROSTATICS */
 896
 897             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 898             ewrt             = _mm_mul_pd(r30,ewtabscale);
 899             ewitab           = _mm_cvttpd_epi32(ewrt);
 900 #ifdef __XOP__
 901             eweps            = _mm_frcz_pd(ewrt);
 902 #else
 903             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 904 #endif
 905             twoeweps         = _mm_add_pd(eweps,eweps);
 906             gmx_mm_load_2pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),ewtab+_mm_extract_epi32(ewitab,1),
 907                                          &ewtabF,&ewtabFn);
 908             felec            = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
 909             felec            = _mm_mul_pd(_mm_mul_pd(qq30,rinv30),_mm_sub_pd(rinvsq30,felec));
 910
 911             cutoff_mask      = _mm_cmplt_pd(rsq30,rcutoff2);
 912
 913             fscal            = felec;
 914
 915             fscal            = _mm_and_pd(fscal,cutoff_mask);
 916
 917             /* Update vectorial force */
 918             fix3             = _mm_macc_pd(dx30,fscal,fix3);
 919             fiy3             = _mm_macc_pd(dy30,fscal,fiy3);
 920             fiz3             = _mm_macc_pd(dz30,fscal,fiz3);
 921
 922             fjx0             = _mm_macc_pd(dx30,fscal,fjx0);
 923             fjy0             = _mm_macc_pd(dy30,fscal,fjy0);
 924             fjz0             = _mm_macc_pd(dz30,fscal,fjz0);
 925
 926             }
 927
 928             gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 929
 930             /* Inner loop uses 129 flops */
 931         }
 932
 933         if(jidx<j_index_end)
 934         {
 935
 936             jnrA             = jjnr[jidx];
 937             j_coord_offsetA  = DIM*jnrA;
 938
 939             /* load j atom coordinates */
 940             gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 941                                               &jx0,&jy0,&jz0);
 942
 943             /* Calculate displacement vector */
 944             dx10             = _mm_sub_pd(ix1,jx0);
 945             dy10             = _mm_sub_pd(iy1,jy0);
 946             dz10             = _mm_sub_pd(iz1,jz0);
 947             dx20             = _mm_sub_pd(ix2,jx0);
 948             dy20             = _mm_sub_pd(iy2,jy0);
 949             dz20             = _mm_sub_pd(iz2,jz0);
 950             dx30             = _mm_sub_pd(ix3,jx0);
 951             dy30             = _mm_sub_pd(iy3,jy0);
 952             dz30             = _mm_sub_pd(iz3,jz0);
 953
 954             /* Calculate squared distance and things based on it */
 955             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 956             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 957             rsq30            = gmx_mm_calc_rsq_pd(dx30,dy30,dz30);
 958
 959             rinv10           = gmx_mm_invsqrt_pd(rsq10);
 960             rinv20           = gmx_mm_invsqrt_pd(rsq20);
 961             rinv30           = gmx_mm_invsqrt_pd(rsq30);
 962
 963             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
 964             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
 965             rinvsq30         = _mm_mul_pd(rinv30,rinv30);
 966
 967             /* Load parameters for j particles */
 968             jq0              = _mm_load_sd(charge+jnrA+0);
 969
 970             fjx0             = _mm_setzero_pd();
 971             fjy0             = _mm_setzero_pd();
 972             fjz0             = _mm_setzero_pd();
 973
 974             /**************************
 975              * CALCULATE INTERACTIONS *
 976              **************************/
 977
 978             if (gmx_mm_any_lt(rsq10,rcutoff2))
 979             {
 980
 981             r10              = _mm_mul_pd(rsq10,rinv10);
 982
 983             /* Compute parameters for interactions between i and j atoms */
 984             qq10             = _mm_mul_pd(iq1,jq0);
 985
 986             /* EWALD ELECTROSTATICS */
 987
 988             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 989             ewrt             = _mm_mul_pd(r10,ewtabscale);
 990             ewitab           = _mm_cvttpd_epi32(ewrt);
 991 #ifdef __XOP__
 992             eweps            = _mm_frcz_pd(ewrt);
 993 #else
 994             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 995 #endif
 996             twoeweps         = _mm_add_pd(eweps,eweps);
 997             gmx_mm_load_1pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
 998             felec            = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
 999             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
1000
1001             cutoff_mask      = _mm_cmplt_pd(rsq10,rcutoff2);
1002
1003             fscal            = felec;
1004
1005             fscal            = _mm_and_pd(fscal,cutoff_mask);
1006
1007             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1008
1009             /* Update vectorial force */
1010             fix1             = _mm_macc_pd(dx10,fscal,fix1);
1011             fiy1             = _mm_macc_pd(dy10,fscal,fiy1);
1012             fiz1             = _mm_macc_pd(dz10,fscal,fiz1);
1013
1014             fjx0             = _mm_macc_pd(dx10,fscal,fjx0);
1015             fjy0             = _mm_macc_pd(dy10,fscal,fjy0);
1016             fjz0             = _mm_macc_pd(dz10,fscal,fjz0);
1017
1018             }
1019
1020             /**************************
1021              * CALCULATE INTERACTIONS *
1022              **************************/
1023
1024             if (gmx_mm_any_lt(rsq20,rcutoff2))
1025             {
1026
1027             r20              = _mm_mul_pd(rsq20,rinv20);
1028
1029             /* Compute parameters for interactions between i and j atoms */
1030             qq20             = _mm_mul_pd(iq2,jq0);
1031
1032             /* EWALD ELECTROSTATICS */
1033
1034             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1035             ewrt             = _mm_mul_pd(r20,ewtabscale);
1036             ewitab           = _mm_cvttpd_epi32(ewrt);
1037 #ifdef __XOP__
1038             eweps            = _mm_frcz_pd(ewrt);
1039 #else
1040             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1041 #endif
1042             twoeweps         = _mm_add_pd(eweps,eweps);
1043             gmx_mm_load_1pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1044             felec            = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
1045             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
1046
1047             cutoff_mask      = _mm_cmplt_pd(rsq20,rcutoff2);
1048
1049             fscal            = felec;
1050
1051             fscal            = _mm_and_pd(fscal,cutoff_mask);
1052
1053             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1054
1055             /* Update vectorial force */
1056             fix2             = _mm_macc_pd(dx20,fscal,fix2);
1057             fiy2             = _mm_macc_pd(dy20,fscal,fiy2);
1058             fiz2             = _mm_macc_pd(dz20,fscal,fiz2);
1059
1060             fjx0             = _mm_macc_pd(dx20,fscal,fjx0);
1061             fjy0             = _mm_macc_pd(dy20,fscal,fjy0);
1062             fjz0             = _mm_macc_pd(dz20,fscal,fjz0);
1063
1064             }
1065
1066             /**************************
1067              * CALCULATE INTERACTIONS *
1068              **************************/
1069
1070             if (gmx_mm_any_lt(rsq30,rcutoff2))
1071             {
1072
1073             r30              = _mm_mul_pd(rsq30,rinv30);
1074
1075             /* Compute parameters for interactions between i and j atoms */
1076             qq30             = _mm_mul_pd(iq3,jq0);
1077
1078             /* EWALD ELECTROSTATICS */
1079
1080             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1081             ewrt             = _mm_mul_pd(r30,ewtabscale);
1082             ewitab           = _mm_cvttpd_epi32(ewrt);
1083 #ifdef __XOP__
1084             eweps            = _mm_frcz_pd(ewrt);
1085 #else
1086             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1087 #endif
1088             twoeweps         = _mm_add_pd(eweps,eweps);
1089             gmx_mm_load_1pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1090             felec            = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
1091             felec            = _mm_mul_pd(_mm_mul_pd(qq30,rinv30),_mm_sub_pd(rinvsq30,felec));
1092
1093             cutoff_mask      = _mm_cmplt_pd(rsq30,rcutoff2);
1094
1095             fscal            = felec;
1096
1097             fscal            = _mm_and_pd(fscal,cutoff_mask);
1098
1099             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1100
1101             /* Update vectorial force */
1102             fix3             = _mm_macc_pd(dx30,fscal,fix3);
1103             fiy3             = _mm_macc_pd(dy30,fscal,fiy3);
1104             fiz3             = _mm_macc_pd(dz30,fscal,fiz3);
1105
1106             fjx0             = _mm_macc_pd(dx30,fscal,fjx0);
1107             fjy0             = _mm_macc_pd(dy30,fscal,fjy0);
1108             fjz0             = _mm_macc_pd(dz30,fscal,fjz0);
1109
1110             }
1111
1112             gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1113
1114             /* Inner loop uses 129 flops */
1115         }
1116
1117         /* End of innermost loop */
1118
1119         gmx_mm_update_iforce_3atom_swizzle_pd(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1120                                               f+i_coord_offset+DIM,fshift+i_shift_offset);
1121
1122         /* Increment number of inner iterations */
1123         inneriter                  += j_index_end - j_index_start;
1124
1125         /* Outer loop uses 18 flops */
1126     }
1127
1128     /* Increment number of outer iterations */
1129     outeriter        += nri;
1130
1131     /* Update outer/inner flops */
1132
1133     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4_F,outeriter*18 + inneriter*129);
1134 }