src/gromacs/gmxlib/nonbonded/nb_kernel_avx_128_fma_double/nb_kernel_ElecEwSw_VdwLJSw_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 "gmxpre.h"
  39
  40 #include "config.h"
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "gromacs/legacyheaders/types/simple.h"
  46 #include "gromacs/math/vec.h"
  47 #include "gromacs/legacyheaders/nrnb.h"
  48
  49 #include "gromacs/simd/math_x86_avx_128_fma_double.h"
  50 #include "kernelutil_x86_avx_128_fma_double.h"
  51
  52 /*
  53  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSw_VdwLJSw_GeomW4P1_VF_avx_128_fma_double
  54  * Electrostatics interaction: Ewald
  55  * VdW interaction:            LennardJones
  56  * Geometry:                   Water4-Particle
  57  * Calculate force/pot:        PotentialAndForce
  58  */
  59 void
  60 nb_kernel_ElecEwSw_VdwLJSw_GeomW4P1_VF_avx_128_fma_double
  61                     (t_nblist                    * gmx_restrict       nlist,
  62                      rvec                        * gmx_restrict          xx,
  63                      rvec                        * gmx_restrict          ff,
  64                      t_forcerec                  * gmx_restrict          fr,
  65                      t_mdatoms                   * gmx_restrict     mdatoms,
  66                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  67                      t_nrnb                      * gmx_restrict        nrnb)
  68 {
  69     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  70      * just 0 for non-waters.
  71      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
  72      * jnr indices corresponding to data put in the four positions in the SIMD register.
  73      */
  74     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  75     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  76     int              jnrA,jnrB;
  77     int              j_coord_offsetA,j_coord_offsetB;
  78     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  79     real             rcutoff_scalar;
  80     real             *shiftvec,*fshift,*x,*f;
  81     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  82     int              vdwioffset0;
  83     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  84     int              vdwioffset1;
  85     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  86     int              vdwioffset2;
  87     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  88     int              vdwioffset3;
  89     __m128d          ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  90     int              vdwjidx0A,vdwjidx0B;
  91     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  92     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  93     __m128d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  94     __m128d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  95     __m128d          dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
  96     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
  97     real             *charge;
  98     int              nvdwtype;
  99     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 100     int              *vdwtype;
 101     real             *vdwparam;
 102     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
 103     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
 104     __m128i          ewitab;
 105     __m128d          ewtabscale,eweps,twoeweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 106     real             *ewtab;
 107     __m128d          rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
 108     real             rswitch_scalar,d_scalar;
 109     __m128d          dummy_mask,cutoff_mask;
 110     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 111     __m128d          one     = _mm_set1_pd(1.0);
 112     __m128d          two     = _mm_set1_pd(2.0);
 113     x                = xx[0];
 114     f                = ff[0];
 115
 116     nri              = nlist->nri;
 117     iinr             = nlist->iinr;
 118     jindex           = nlist->jindex;
 119     jjnr             = nlist->jjnr;
 120     shiftidx         = nlist->shift;
 121     gid              = nlist->gid;
 122     shiftvec         = fr->shift_vec[0];
 123     fshift           = fr->fshift[0];
 124     facel            = _mm_set1_pd(fr->epsfac);
 125     charge           = mdatoms->chargeA;
 126     nvdwtype         = fr->ntype;
 127     vdwparam         = fr->nbfp;
 128     vdwtype          = mdatoms->typeA;
 129
 130     sh_ewald         = _mm_set1_pd(fr->ic->sh_ewald);
 131     ewtab            = fr->ic->tabq_coul_FDV0;
 132     ewtabscale       = _mm_set1_pd(fr->ic->tabq_scale);
 133     ewtabhalfspace   = _mm_set1_pd(0.5/fr->ic->tabq_scale);
 134
 135     /* Setup water-specific parameters */
 136     inr              = nlist->iinr[0];
 137     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
 138     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
 139     iq3              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
 140     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 141
 142     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 143     rcutoff_scalar   = fr->rcoulomb;
 144     rcutoff          = _mm_set1_pd(rcutoff_scalar);
 145     rcutoff2         = _mm_mul_pd(rcutoff,rcutoff);
 146
 147     rswitch_scalar   = fr->rcoulomb_switch;
 148     rswitch          = _mm_set1_pd(rswitch_scalar);
 149     /* Setup switch parameters */
 150     d_scalar         = rcutoff_scalar-rswitch_scalar;
 151     d                = _mm_set1_pd(d_scalar);
 152     swV3             = _mm_set1_pd(-10.0/(d_scalar*d_scalar*d_scalar));
 153     swV4             = _mm_set1_pd( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 154     swV5             = _mm_set1_pd( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 155     swF2             = _mm_set1_pd(-30.0/(d_scalar*d_scalar*d_scalar));
 156     swF3             = _mm_set1_pd( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 157     swF4             = _mm_set1_pd(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 158
 159     /* Avoid stupid compiler warnings */
 160     jnrA = jnrB = 0;
 161     j_coord_offsetA = 0;
 162     j_coord_offsetB = 0;
 163
 164     outeriter        = 0;
 165     inneriter        = 0;
 166
 167     /* Start outer loop over neighborlists */
 168     for(iidx=0; iidx<nri; iidx++)
 169     {
 170         /* Load shift vector for this list */
 171         i_shift_offset   = DIM*shiftidx[iidx];
 172
 173         /* Load limits for loop over neighbors */
 174         j_index_start    = jindex[iidx];
 175         j_index_end      = jindex[iidx+1];
 176
 177         /* Get outer coordinate index */
 178         inr              = iinr[iidx];
 179         i_coord_offset   = DIM*inr;
 180
 181         /* Load i particle coords and add shift vector */
 182         gmx_mm_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 183                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 184
 185         fix0             = _mm_setzero_pd();
 186         fiy0             = _mm_setzero_pd();
 187         fiz0             = _mm_setzero_pd();
 188         fix1             = _mm_setzero_pd();
 189         fiy1             = _mm_setzero_pd();
 190         fiz1             = _mm_setzero_pd();
 191         fix2             = _mm_setzero_pd();
 192         fiy2             = _mm_setzero_pd();
 193         fiz2             = _mm_setzero_pd();
 194         fix3             = _mm_setzero_pd();
 195         fiy3             = _mm_setzero_pd();
 196         fiz3             = _mm_setzero_pd();
 197
 198         /* Reset potential sums */
 199         velecsum         = _mm_setzero_pd();
 200         vvdwsum          = _mm_setzero_pd();
 201
 202         /* Start inner kernel loop */
 203         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 204         {
 205
 206             /* Get j neighbor index, and coordinate index */
 207             jnrA             = jjnr[jidx];
 208             jnrB             = jjnr[jidx+1];
 209             j_coord_offsetA  = DIM*jnrA;
 210             j_coord_offsetB  = DIM*jnrB;
 211
 212             /* load j atom coordinates */
 213             gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 214                                               &jx0,&jy0,&jz0);
 215
 216             /* Calculate displacement vector */
 217             dx00             = _mm_sub_pd(ix0,jx0);
 218             dy00             = _mm_sub_pd(iy0,jy0);
 219             dz00             = _mm_sub_pd(iz0,jz0);
 220             dx10             = _mm_sub_pd(ix1,jx0);
 221             dy10             = _mm_sub_pd(iy1,jy0);
 222             dz10             = _mm_sub_pd(iz1,jz0);
 223             dx20             = _mm_sub_pd(ix2,jx0);
 224             dy20             = _mm_sub_pd(iy2,jy0);
 225             dz20             = _mm_sub_pd(iz2,jz0);
 226             dx30             = _mm_sub_pd(ix3,jx0);
 227             dy30             = _mm_sub_pd(iy3,jy0);
 228             dz30             = _mm_sub_pd(iz3,jz0);
 229
 230             /* Calculate squared distance and things based on it */
 231             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 232             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 233             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 234             rsq30            = gmx_mm_calc_rsq_pd(dx30,dy30,dz30);
 235
 236             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 237             rinv10           = gmx_mm_invsqrt_pd(rsq10);
 238             rinv20           = gmx_mm_invsqrt_pd(rsq20);
 239             rinv30           = gmx_mm_invsqrt_pd(rsq30);
 240
 241             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 242             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
 243             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
 244             rinvsq30         = _mm_mul_pd(rinv30,rinv30);
 245
 246             /* Load parameters for j particles */
 247             jq0              = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
 248             vdwjidx0A        = 2*vdwtype[jnrA+0];
 249             vdwjidx0B        = 2*vdwtype[jnrB+0];
 250
 251             fjx0             = _mm_setzero_pd();
 252             fjy0             = _mm_setzero_pd();
 253             fjz0             = _mm_setzero_pd();
 254
 255             /**************************
 256              * CALCULATE INTERACTIONS *
 257              **************************/
 258
 259             if (gmx_mm_any_lt(rsq00,rcutoff2))
 260             {
 261
 262             r00              = _mm_mul_pd(rsq00,rinv00);
 263
 264             /* Compute parameters for interactions between i and j atoms */
 265             gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
 266                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 267
 268             /* LENNARD-JONES DISPERSION/REPULSION */
 269
 270             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 271             vvdw6            = _mm_mul_pd(c6_00,rinvsix);
 272             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
 273             vvdw             = _mm_msub_pd( vvdw12,one_twelfth, _mm_mul_pd(vvdw6,one_sixth) );
 274             fvdw             = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
 275
 276             d                = _mm_sub_pd(r00,rswitch);
 277             d                = _mm_max_pd(d,_mm_setzero_pd());
 278             d2               = _mm_mul_pd(d,d);
 279             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
 280
 281             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
 282
 283             /* Evaluate switch function */
 284             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 285             fvdw             = _mm_msub_pd( fvdw,sw , _mm_mul_pd(rinv00,_mm_mul_pd(vvdw,dsw)) );
 286             vvdw             = _mm_mul_pd(vvdw,sw);
 287             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
 288
 289             /* Update potential sum for this i atom from the interaction with this j atom. */
 290             vvdw             = _mm_and_pd(vvdw,cutoff_mask);
 291             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 292
 293             fscal            = fvdw;
 294
 295             fscal            = _mm_and_pd(fscal,cutoff_mask);
 296
 297             /* Update vectorial force */
 298             fix0             = _mm_macc_pd(dx00,fscal,fix0);
 299             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
 300             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
 301
 302             fjx0             = _mm_macc_pd(dx00,fscal,fjx0);
 303             fjy0             = _mm_macc_pd(dy00,fscal,fjy0);
 304             fjz0             = _mm_macc_pd(dz00,fscal,fjz0);
 305
 306             }
 307
 308             /**************************
 309              * CALCULATE INTERACTIONS *
 310              **************************/
 311
 312             if (gmx_mm_any_lt(rsq10,rcutoff2))
 313             {
 314
 315             r10              = _mm_mul_pd(rsq10,rinv10);
 316
 317             /* Compute parameters for interactions between i and j atoms */
 318             qq10             = _mm_mul_pd(iq1,jq0);
 319
 320             /* EWALD ELECTROSTATICS */
 321
 322             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 323             ewrt             = _mm_mul_pd(r10,ewtabscale);
 324             ewitab           = _mm_cvttpd_epi32(ewrt);
 325 #ifdef __XOP__
 326             eweps            = _mm_frcz_pd(ewrt);
 327 #else
 328             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 329 #endif
 330             twoeweps         = _mm_add_pd(eweps,eweps);
 331             ewitab           = _mm_slli_epi32(ewitab,2);
 332             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 333             ewtabD           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 334             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 335             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
 336             ewtabFn          = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
 337             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 338             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
 339             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
 340             velec            = _mm_mul_pd(qq10,_mm_sub_pd(rinv10,velec));
 341             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
 342
 343             d                = _mm_sub_pd(r10,rswitch);
 344             d                = _mm_max_pd(d,_mm_setzero_pd());
 345             d2               = _mm_mul_pd(d,d);
 346             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
 347
 348             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
 349
 350             /* Evaluate switch function */
 351             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 352             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv10,_mm_mul_pd(velec,dsw)) );
 353             velec            = _mm_mul_pd(velec,sw);
 354             cutoff_mask      = _mm_cmplt_pd(rsq10,rcutoff2);
 355
 356             /* Update potential sum for this i atom from the interaction with this j atom. */
 357             velec            = _mm_and_pd(velec,cutoff_mask);
 358             velecsum         = _mm_add_pd(velecsum,velec);
 359
 360             fscal            = felec;
 361
 362             fscal            = _mm_and_pd(fscal,cutoff_mask);
 363
 364             /* Update vectorial force */
 365             fix1             = _mm_macc_pd(dx10,fscal,fix1);
 366             fiy1             = _mm_macc_pd(dy10,fscal,fiy1);
 367             fiz1             = _mm_macc_pd(dz10,fscal,fiz1);
 368
 369             fjx0             = _mm_macc_pd(dx10,fscal,fjx0);
 370             fjy0             = _mm_macc_pd(dy10,fscal,fjy0);
 371             fjz0             = _mm_macc_pd(dz10,fscal,fjz0);
 372
 373             }
 374
 375             /**************************
 376              * CALCULATE INTERACTIONS *
 377              **************************/
 378
 379             if (gmx_mm_any_lt(rsq20,rcutoff2))
 380             {
 381
 382             r20              = _mm_mul_pd(rsq20,rinv20);
 383
 384             /* Compute parameters for interactions between i and j atoms */
 385             qq20             = _mm_mul_pd(iq2,jq0);
 386
 387             /* EWALD ELECTROSTATICS */
 388
 389             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 390             ewrt             = _mm_mul_pd(r20,ewtabscale);
 391             ewitab           = _mm_cvttpd_epi32(ewrt);
 392 #ifdef __XOP__
 393             eweps            = _mm_frcz_pd(ewrt);
 394 #else
 395             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 396 #endif
 397             twoeweps         = _mm_add_pd(eweps,eweps);
 398             ewitab           = _mm_slli_epi32(ewitab,2);
 399             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 400             ewtabD           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 401             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 402             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
 403             ewtabFn          = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
 404             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 405             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
 406             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
 407             velec            = _mm_mul_pd(qq20,_mm_sub_pd(rinv20,velec));
 408             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
 409
 410             d                = _mm_sub_pd(r20,rswitch);
 411             d                = _mm_max_pd(d,_mm_setzero_pd());
 412             d2               = _mm_mul_pd(d,d);
 413             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
 414
 415             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
 416
 417             /* Evaluate switch function */
 418             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 419             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv20,_mm_mul_pd(velec,dsw)) );
 420             velec            = _mm_mul_pd(velec,sw);
 421             cutoff_mask      = _mm_cmplt_pd(rsq20,rcutoff2);
 422
 423             /* Update potential sum for this i atom from the interaction with this j atom. */
 424             velec            = _mm_and_pd(velec,cutoff_mask);
 425             velecsum         = _mm_add_pd(velecsum,velec);
 426
 427             fscal            = felec;
 428
 429             fscal            = _mm_and_pd(fscal,cutoff_mask);
 430
 431             /* Update vectorial force */
 432             fix2             = _mm_macc_pd(dx20,fscal,fix2);
 433             fiy2             = _mm_macc_pd(dy20,fscal,fiy2);
 434             fiz2             = _mm_macc_pd(dz20,fscal,fiz2);
 435
 436             fjx0             = _mm_macc_pd(dx20,fscal,fjx0);
 437             fjy0             = _mm_macc_pd(dy20,fscal,fjy0);
 438             fjz0             = _mm_macc_pd(dz20,fscal,fjz0);
 439
 440             }
 441
 442             /**************************
 443              * CALCULATE INTERACTIONS *
 444              **************************/
 445
 446             if (gmx_mm_any_lt(rsq30,rcutoff2))
 447             {
 448
 449             r30              = _mm_mul_pd(rsq30,rinv30);
 450
 451             /* Compute parameters for interactions between i and j atoms */
 452             qq30             = _mm_mul_pd(iq3,jq0);
 453
 454             /* EWALD ELECTROSTATICS */
 455
 456             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 457             ewrt             = _mm_mul_pd(r30,ewtabscale);
 458             ewitab           = _mm_cvttpd_epi32(ewrt);
 459 #ifdef __XOP__
 460             eweps            = _mm_frcz_pd(ewrt);
 461 #else
 462             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 463 #endif
 464             twoeweps         = _mm_add_pd(eweps,eweps);
 465             ewitab           = _mm_slli_epi32(ewitab,2);
 466             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 467             ewtabD           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 468             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 469             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
 470             ewtabFn          = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
 471             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 472             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
 473             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
 474             velec            = _mm_mul_pd(qq30,_mm_sub_pd(rinv30,velec));
 475             felec            = _mm_mul_pd(_mm_mul_pd(qq30,rinv30),_mm_sub_pd(rinvsq30,felec));
 476
 477             d                = _mm_sub_pd(r30,rswitch);
 478             d                = _mm_max_pd(d,_mm_setzero_pd());
 479             d2               = _mm_mul_pd(d,d);
 480             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
 481
 482             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
 483
 484             /* Evaluate switch function */
 485             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 486             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv30,_mm_mul_pd(velec,dsw)) );
 487             velec            = _mm_mul_pd(velec,sw);
 488             cutoff_mask      = _mm_cmplt_pd(rsq30,rcutoff2);
 489
 490             /* Update potential sum for this i atom from the interaction with this j atom. */
 491             velec            = _mm_and_pd(velec,cutoff_mask);
 492             velecsum         = _mm_add_pd(velecsum,velec);
 493
 494             fscal            = felec;
 495
 496             fscal            = _mm_and_pd(fscal,cutoff_mask);
 497
 498             /* Update vectorial force */
 499             fix3             = _mm_macc_pd(dx30,fscal,fix3);
 500             fiy3             = _mm_macc_pd(dy30,fscal,fiy3);
 501             fiz3             = _mm_macc_pd(dz30,fscal,fiz3);
 502
 503             fjx0             = _mm_macc_pd(dx30,fscal,fjx0);
 504             fjy0             = _mm_macc_pd(dy30,fscal,fjy0);
 505             fjz0             = _mm_macc_pd(dz30,fscal,fjz0);
 506
 507             }
 508
 509             gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 510
 511             /* Inner loop uses 269 flops */
 512         }
 513
 514         if(jidx<j_index_end)
 515         {
 516
 517             jnrA             = jjnr[jidx];
 518             j_coord_offsetA  = DIM*jnrA;
 519
 520             /* load j atom coordinates */
 521             gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 522                                               &jx0,&jy0,&jz0);
 523
 524             /* Calculate displacement vector */
 525             dx00             = _mm_sub_pd(ix0,jx0);
 526             dy00             = _mm_sub_pd(iy0,jy0);
 527             dz00             = _mm_sub_pd(iz0,jz0);
 528             dx10             = _mm_sub_pd(ix1,jx0);
 529             dy10             = _mm_sub_pd(iy1,jy0);
 530             dz10             = _mm_sub_pd(iz1,jz0);
 531             dx20             = _mm_sub_pd(ix2,jx0);
 532             dy20             = _mm_sub_pd(iy2,jy0);
 533             dz20             = _mm_sub_pd(iz2,jz0);
 534             dx30             = _mm_sub_pd(ix3,jx0);
 535             dy30             = _mm_sub_pd(iy3,jy0);
 536             dz30             = _mm_sub_pd(iz3,jz0);
 537
 538             /* Calculate squared distance and things based on it */
 539             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 540             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 541             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 542             rsq30            = gmx_mm_calc_rsq_pd(dx30,dy30,dz30);
 543
 544             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 545             rinv10           = gmx_mm_invsqrt_pd(rsq10);
 546             rinv20           = gmx_mm_invsqrt_pd(rsq20);
 547             rinv30           = gmx_mm_invsqrt_pd(rsq30);
 548
 549             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 550             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
 551             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
 552             rinvsq30         = _mm_mul_pd(rinv30,rinv30);
 553
 554             /* Load parameters for j particles */
 555             jq0              = _mm_load_sd(charge+jnrA+0);
 556             vdwjidx0A        = 2*vdwtype[jnrA+0];
 557
 558             fjx0             = _mm_setzero_pd();
 559             fjy0             = _mm_setzero_pd();
 560             fjz0             = _mm_setzero_pd();
 561
 562             /**************************
 563              * CALCULATE INTERACTIONS *
 564              **************************/
 565
 566             if (gmx_mm_any_lt(rsq00,rcutoff2))
 567             {
 568
 569             r00              = _mm_mul_pd(rsq00,rinv00);
 570
 571             /* Compute parameters for interactions between i and j atoms */
 572             gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 573
 574             /* LENNARD-JONES DISPERSION/REPULSION */
 575
 576             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 577             vvdw6            = _mm_mul_pd(c6_00,rinvsix);
 578             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
 579             vvdw             = _mm_msub_pd( vvdw12,one_twelfth, _mm_mul_pd(vvdw6,one_sixth) );
 580             fvdw             = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
 581
 582             d                = _mm_sub_pd(r00,rswitch);
 583             d                = _mm_max_pd(d,_mm_setzero_pd());
 584             d2               = _mm_mul_pd(d,d);
 585             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
 586
 587             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
 588
 589             /* Evaluate switch function */
 590             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 591             fvdw             = _mm_msub_pd( fvdw,sw , _mm_mul_pd(rinv00,_mm_mul_pd(vvdw,dsw)) );
 592             vvdw             = _mm_mul_pd(vvdw,sw);
 593             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
 594
 595             /* Update potential sum for this i atom from the interaction with this j atom. */
 596             vvdw             = _mm_and_pd(vvdw,cutoff_mask);
 597             vvdw             = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
 598             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 599
 600             fscal            = fvdw;
 601
 602             fscal            = _mm_and_pd(fscal,cutoff_mask);
 603
 604             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 605
 606             /* Update vectorial force */
 607             fix0             = _mm_macc_pd(dx00,fscal,fix0);
 608             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
 609             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
 610
 611             fjx0             = _mm_macc_pd(dx00,fscal,fjx0);
 612             fjy0             = _mm_macc_pd(dy00,fscal,fjy0);
 613             fjz0             = _mm_macc_pd(dz00,fscal,fjz0);
 614
 615             }
 616
 617             /**************************
 618              * CALCULATE INTERACTIONS *
 619              **************************/
 620
 621             if (gmx_mm_any_lt(rsq10,rcutoff2))
 622             {
 623
 624             r10              = _mm_mul_pd(rsq10,rinv10);
 625
 626             /* Compute parameters for interactions between i and j atoms */
 627             qq10             = _mm_mul_pd(iq1,jq0);
 628
 629             /* EWALD ELECTROSTATICS */
 630
 631             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 632             ewrt             = _mm_mul_pd(r10,ewtabscale);
 633             ewitab           = _mm_cvttpd_epi32(ewrt);
 634 #ifdef __XOP__
 635             eweps            = _mm_frcz_pd(ewrt);
 636 #else
 637             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 638 #endif
 639             twoeweps         = _mm_add_pd(eweps,eweps);
 640             ewitab           = _mm_slli_epi32(ewitab,2);
 641             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 642             ewtabD           = _mm_setzero_pd();
 643             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 644             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
 645             ewtabFn          = _mm_setzero_pd();
 646             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 647             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
 648             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
 649             velec            = _mm_mul_pd(qq10,_mm_sub_pd(rinv10,velec));
 650             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
 651
 652             d                = _mm_sub_pd(r10,rswitch);
 653             d                = _mm_max_pd(d,_mm_setzero_pd());
 654             d2               = _mm_mul_pd(d,d);
 655             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
 656
 657             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
 658
 659             /* Evaluate switch function */
 660             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 661             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv10,_mm_mul_pd(velec,dsw)) );
 662             velec            = _mm_mul_pd(velec,sw);
 663             cutoff_mask      = _mm_cmplt_pd(rsq10,rcutoff2);
 664
 665             /* Update potential sum for this i atom from the interaction with this j atom. */
 666             velec            = _mm_and_pd(velec,cutoff_mask);
 667             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 668             velecsum         = _mm_add_pd(velecsum,velec);
 669
 670             fscal            = felec;
 671
 672             fscal            = _mm_and_pd(fscal,cutoff_mask);
 673
 674             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 675
 676             /* Update vectorial force */
 677             fix1             = _mm_macc_pd(dx10,fscal,fix1);
 678             fiy1             = _mm_macc_pd(dy10,fscal,fiy1);
 679             fiz1             = _mm_macc_pd(dz10,fscal,fiz1);
 680
 681             fjx0             = _mm_macc_pd(dx10,fscal,fjx0);
 682             fjy0             = _mm_macc_pd(dy10,fscal,fjy0);
 683             fjz0             = _mm_macc_pd(dz10,fscal,fjz0);
 684
 685             }
 686
 687             /**************************
 688              * CALCULATE INTERACTIONS *
 689              **************************/
 690
 691             if (gmx_mm_any_lt(rsq20,rcutoff2))
 692             {
 693
 694             r20              = _mm_mul_pd(rsq20,rinv20);
 695
 696             /* Compute parameters for interactions between i and j atoms */
 697             qq20             = _mm_mul_pd(iq2,jq0);
 698
 699             /* EWALD ELECTROSTATICS */
 700
 701             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 702             ewrt             = _mm_mul_pd(r20,ewtabscale);
 703             ewitab           = _mm_cvttpd_epi32(ewrt);
 704 #ifdef __XOP__
 705             eweps            = _mm_frcz_pd(ewrt);
 706 #else
 707             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 708 #endif
 709             twoeweps         = _mm_add_pd(eweps,eweps);
 710             ewitab           = _mm_slli_epi32(ewitab,2);
 711             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 712             ewtabD           = _mm_setzero_pd();
 713             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 714             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
 715             ewtabFn          = _mm_setzero_pd();
 716             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 717             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
 718             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
 719             velec            = _mm_mul_pd(qq20,_mm_sub_pd(rinv20,velec));
 720             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
 721
 722             d                = _mm_sub_pd(r20,rswitch);
 723             d                = _mm_max_pd(d,_mm_setzero_pd());
 724             d2               = _mm_mul_pd(d,d);
 725             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
 726
 727             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
 728
 729             /* Evaluate switch function */
 730             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 731             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv20,_mm_mul_pd(velec,dsw)) );
 732             velec            = _mm_mul_pd(velec,sw);
 733             cutoff_mask      = _mm_cmplt_pd(rsq20,rcutoff2);
 734
 735             /* Update potential sum for this i atom from the interaction with this j atom. */
 736             velec            = _mm_and_pd(velec,cutoff_mask);
 737             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 738             velecsum         = _mm_add_pd(velecsum,velec);
 739
 740             fscal            = felec;
 741
 742             fscal            = _mm_and_pd(fscal,cutoff_mask);
 743
 744             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 745
 746             /* Update vectorial force */
 747             fix2             = _mm_macc_pd(dx20,fscal,fix2);
 748             fiy2             = _mm_macc_pd(dy20,fscal,fiy2);
 749             fiz2             = _mm_macc_pd(dz20,fscal,fiz2);
 750
 751             fjx0             = _mm_macc_pd(dx20,fscal,fjx0);
 752             fjy0             = _mm_macc_pd(dy20,fscal,fjy0);
 753             fjz0             = _mm_macc_pd(dz20,fscal,fjz0);
 754
 755             }
 756
 757             /**************************
 758              * CALCULATE INTERACTIONS *
 759              **************************/
 760
 761             if (gmx_mm_any_lt(rsq30,rcutoff2))
 762             {
 763
 764             r30              = _mm_mul_pd(rsq30,rinv30);
 765
 766             /* Compute parameters for interactions between i and j atoms */
 767             qq30             = _mm_mul_pd(iq3,jq0);
 768
 769             /* EWALD ELECTROSTATICS */
 770
 771             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 772             ewrt             = _mm_mul_pd(r30,ewtabscale);
 773             ewitab           = _mm_cvttpd_epi32(ewrt);
 774 #ifdef __XOP__
 775             eweps            = _mm_frcz_pd(ewrt);
 776 #else
 777             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 778 #endif
 779             twoeweps         = _mm_add_pd(eweps,eweps);
 780             ewitab           = _mm_slli_epi32(ewitab,2);
 781             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 782             ewtabD           = _mm_setzero_pd();
 783             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 784             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
 785             ewtabFn          = _mm_setzero_pd();
 786             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 787             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
 788             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
 789             velec            = _mm_mul_pd(qq30,_mm_sub_pd(rinv30,velec));
 790             felec            = _mm_mul_pd(_mm_mul_pd(qq30,rinv30),_mm_sub_pd(rinvsq30,felec));
 791
 792             d                = _mm_sub_pd(r30,rswitch);
 793             d                = _mm_max_pd(d,_mm_setzero_pd());
 794             d2               = _mm_mul_pd(d,d);
 795             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
 796
 797             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
 798
 799             /* Evaluate switch function */
 800             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 801             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv30,_mm_mul_pd(velec,dsw)) );
 802             velec            = _mm_mul_pd(velec,sw);
 803             cutoff_mask      = _mm_cmplt_pd(rsq30,rcutoff2);
 804
 805             /* Update potential sum for this i atom from the interaction with this j atom. */
 806             velec            = _mm_and_pd(velec,cutoff_mask);
 807             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 808             velecsum         = _mm_add_pd(velecsum,velec);
 809
 810             fscal            = felec;
 811
 812             fscal            = _mm_and_pd(fscal,cutoff_mask);
 813
 814             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 815
 816             /* Update vectorial force */
 817             fix3             = _mm_macc_pd(dx30,fscal,fix3);
 818             fiy3             = _mm_macc_pd(dy30,fscal,fiy3);
 819             fiz3             = _mm_macc_pd(dz30,fscal,fiz3);
 820
 821             fjx0             = _mm_macc_pd(dx30,fscal,fjx0);
 822             fjy0             = _mm_macc_pd(dy30,fscal,fjy0);
 823             fjz0             = _mm_macc_pd(dz30,fscal,fjz0);
 824
 825             }
 826
 827             gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 828
 829             /* Inner loop uses 269 flops */
 830         }
 831
 832         /* End of innermost loop */
 833
 834         gmx_mm_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 835                                               f+i_coord_offset,fshift+i_shift_offset);
 836
 837         ggid                        = gid[iidx];
 838         /* Update potential energies */
 839         gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
 840         gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
 841
 842         /* Increment number of inner iterations */
 843         inneriter                  += j_index_end - j_index_start;
 844
 845         /* Outer loop uses 26 flops */
 846     }
 847
 848     /* Increment number of outer iterations */
 849     outeriter        += nri;
 850
 851     /* Update outer/inner flops */
 852
 853     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_VF,outeriter*26 + inneriter*269);
 854 }
 855 /*
 856  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSw_VdwLJSw_GeomW4P1_F_avx_128_fma_double
 857  * Electrostatics interaction: Ewald
 858  * VdW interaction:            LennardJones
 859  * Geometry:                   Water4-Particle
 860  * Calculate force/pot:        Force
 861  */
 862 void
 863 nb_kernel_ElecEwSw_VdwLJSw_GeomW4P1_F_avx_128_fma_double
 864                     (t_nblist                    * gmx_restrict       nlist,
 865                      rvec                        * gmx_restrict          xx,
 866                      rvec                        * gmx_restrict          ff,
 867                      t_forcerec                  * gmx_restrict          fr,
 868                      t_mdatoms                   * gmx_restrict     mdatoms,
 869                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 870                      t_nrnb                      * gmx_restrict        nrnb)
 871 {
 872     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 873      * just 0 for non-waters.
 874      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
 875      * jnr indices corresponding to data put in the four positions in the SIMD register.
 876      */
 877     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 878     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 879     int              jnrA,jnrB;
 880     int              j_coord_offsetA,j_coord_offsetB;
 881     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 882     real             rcutoff_scalar;
 883     real             *shiftvec,*fshift,*x,*f;
 884     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 885     int              vdwioffset0;
 886     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 887     int              vdwioffset1;
 888     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 889     int              vdwioffset2;
 890     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 891     int              vdwioffset3;
 892     __m128d          ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
 893     int              vdwjidx0A,vdwjidx0B;
 894     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 895     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 896     __m128d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 897     __m128d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 898     __m128d          dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
 899     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
 900     real             *charge;
 901     int              nvdwtype;
 902     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 903     int              *vdwtype;
 904     real             *vdwparam;
 905     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
 906     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
 907     __m128i          ewitab;
 908     __m128d          ewtabscale,eweps,twoeweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 909     real             *ewtab;
 910     __m128d          rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
 911     real             rswitch_scalar,d_scalar;
 912     __m128d          dummy_mask,cutoff_mask;
 913     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 914     __m128d          one     = _mm_set1_pd(1.0);
 915     __m128d          two     = _mm_set1_pd(2.0);
 916     x                = xx[0];
 917     f                = ff[0];
 918
 919     nri              = nlist->nri;
 920     iinr             = nlist->iinr;
 921     jindex           = nlist->jindex;
 922     jjnr             = nlist->jjnr;
 923     shiftidx         = nlist->shift;
 924     gid              = nlist->gid;
 925     shiftvec         = fr->shift_vec[0];
 926     fshift           = fr->fshift[0];
 927     facel            = _mm_set1_pd(fr->epsfac);
 928     charge           = mdatoms->chargeA;
 929     nvdwtype         = fr->ntype;
 930     vdwparam         = fr->nbfp;
 931     vdwtype          = mdatoms->typeA;
 932
 933     sh_ewald         = _mm_set1_pd(fr->ic->sh_ewald);
 934     ewtab            = fr->ic->tabq_coul_FDV0;
 935     ewtabscale       = _mm_set1_pd(fr->ic->tabq_scale);
 936     ewtabhalfspace   = _mm_set1_pd(0.5/fr->ic->tabq_scale);
 937
 938     /* Setup water-specific parameters */
 939     inr              = nlist->iinr[0];
 940     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
 941     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
 942     iq3              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
 943     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 944
 945     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 946     rcutoff_scalar   = fr->rcoulomb;
 947     rcutoff          = _mm_set1_pd(rcutoff_scalar);
 948     rcutoff2         = _mm_mul_pd(rcutoff,rcutoff);
 949
 950     rswitch_scalar   = fr->rcoulomb_switch;
 951     rswitch          = _mm_set1_pd(rswitch_scalar);
 952     /* Setup switch parameters */
 953     d_scalar         = rcutoff_scalar-rswitch_scalar;
 954     d                = _mm_set1_pd(d_scalar);
 955     swV3             = _mm_set1_pd(-10.0/(d_scalar*d_scalar*d_scalar));
 956     swV4             = _mm_set1_pd( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 957     swV5             = _mm_set1_pd( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 958     swF2             = _mm_set1_pd(-30.0/(d_scalar*d_scalar*d_scalar));
 959     swF3             = _mm_set1_pd( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 960     swF4             = _mm_set1_pd(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 961
 962     /* Avoid stupid compiler warnings */
 963     jnrA = jnrB = 0;
 964     j_coord_offsetA = 0;
 965     j_coord_offsetB = 0;
 966
 967     outeriter        = 0;
 968     inneriter        = 0;
 969
 970     /* Start outer loop over neighborlists */
 971     for(iidx=0; iidx<nri; iidx++)
 972     {
 973         /* Load shift vector for this list */
 974         i_shift_offset   = DIM*shiftidx[iidx];
 975
 976         /* Load limits for loop over neighbors */
 977         j_index_start    = jindex[iidx];
 978         j_index_end      = jindex[iidx+1];
 979
 980         /* Get outer coordinate index */
 981         inr              = iinr[iidx];
 982         i_coord_offset   = DIM*inr;
 983
 984         /* Load i particle coords and add shift vector */
 985         gmx_mm_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 986                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 987
 988         fix0             = _mm_setzero_pd();
 989         fiy0             = _mm_setzero_pd();
 990         fiz0             = _mm_setzero_pd();
 991         fix1             = _mm_setzero_pd();
 992         fiy1             = _mm_setzero_pd();
 993         fiz1             = _mm_setzero_pd();
 994         fix2             = _mm_setzero_pd();
 995         fiy2             = _mm_setzero_pd();
 996         fiz2             = _mm_setzero_pd();
 997         fix3             = _mm_setzero_pd();
 998         fiy3             = _mm_setzero_pd();
 999         fiz3             = _mm_setzero_pd();
1000
1001         /* Start inner kernel loop */
1002         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
1003         {
1004
1005             /* Get j neighbor index, and coordinate index */
1006             jnrA             = jjnr[jidx];
1007             jnrB             = jjnr[jidx+1];
1008             j_coord_offsetA  = DIM*jnrA;
1009             j_coord_offsetB  = DIM*jnrB;
1010
1011             /* load j atom coordinates */
1012             gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1013                                               &jx0,&jy0,&jz0);
1014
1015             /* Calculate displacement vector */
1016             dx00             = _mm_sub_pd(ix0,jx0);
1017             dy00             = _mm_sub_pd(iy0,jy0);
1018             dz00             = _mm_sub_pd(iz0,jz0);
1019             dx10             = _mm_sub_pd(ix1,jx0);
1020             dy10             = _mm_sub_pd(iy1,jy0);
1021             dz10             = _mm_sub_pd(iz1,jz0);
1022             dx20             = _mm_sub_pd(ix2,jx0);
1023             dy20             = _mm_sub_pd(iy2,jy0);
1024             dz20             = _mm_sub_pd(iz2,jz0);
1025             dx30             = _mm_sub_pd(ix3,jx0);
1026             dy30             = _mm_sub_pd(iy3,jy0);
1027             dz30             = _mm_sub_pd(iz3,jz0);
1028
1029             /* Calculate squared distance and things based on it */
1030             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1031             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
1032             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
1033             rsq30            = gmx_mm_calc_rsq_pd(dx30,dy30,dz30);
1034
1035             rinv00           = gmx_mm_invsqrt_pd(rsq00);
1036             rinv10           = gmx_mm_invsqrt_pd(rsq10);
1037             rinv20           = gmx_mm_invsqrt_pd(rsq20);
1038             rinv30           = gmx_mm_invsqrt_pd(rsq30);
1039
1040             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
1041             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
1042             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
1043             rinvsq30         = _mm_mul_pd(rinv30,rinv30);
1044
1045             /* Load parameters for j particles */
1046             jq0              = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
1047             vdwjidx0A        = 2*vdwtype[jnrA+0];
1048             vdwjidx0B        = 2*vdwtype[jnrB+0];
1049
1050             fjx0             = _mm_setzero_pd();
1051             fjy0             = _mm_setzero_pd();
1052             fjz0             = _mm_setzero_pd();
1053
1054             /**************************
1055              * CALCULATE INTERACTIONS *
1056              **************************/
1057
1058             if (gmx_mm_any_lt(rsq00,rcutoff2))
1059             {
1060
1061             r00              = _mm_mul_pd(rsq00,rinv00);
1062
1063             /* Compute parameters for interactions between i and j atoms */
1064             gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
1065                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
1066
1067             /* LENNARD-JONES DISPERSION/REPULSION */
1068
1069             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1070             vvdw6            = _mm_mul_pd(c6_00,rinvsix);
1071             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
1072             vvdw             = _mm_msub_pd( vvdw12,one_twelfth, _mm_mul_pd(vvdw6,one_sixth) );
1073             fvdw             = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
1074
1075             d                = _mm_sub_pd(r00,rswitch);
1076             d                = _mm_max_pd(d,_mm_setzero_pd());
1077             d2               = _mm_mul_pd(d,d);
1078             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
1079
1080             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
1081
1082             /* Evaluate switch function */
1083             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1084             fvdw             = _mm_msub_pd( fvdw,sw , _mm_mul_pd(rinv00,_mm_mul_pd(vvdw,dsw)) );
1085             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
1086
1087             fscal            = fvdw;
1088
1089             fscal            = _mm_and_pd(fscal,cutoff_mask);
1090
1091             /* Update vectorial force */
1092             fix0             = _mm_macc_pd(dx00,fscal,fix0);
1093             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
1094             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
1095
1096             fjx0             = _mm_macc_pd(dx00,fscal,fjx0);
1097             fjy0             = _mm_macc_pd(dy00,fscal,fjy0);
1098             fjz0             = _mm_macc_pd(dz00,fscal,fjz0);
1099
1100             }
1101
1102             /**************************
1103              * CALCULATE INTERACTIONS *
1104              **************************/
1105
1106             if (gmx_mm_any_lt(rsq10,rcutoff2))
1107             {
1108
1109             r10              = _mm_mul_pd(rsq10,rinv10);
1110
1111             /* Compute parameters for interactions between i and j atoms */
1112             qq10             = _mm_mul_pd(iq1,jq0);
1113
1114             /* EWALD ELECTROSTATICS */
1115
1116             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1117             ewrt             = _mm_mul_pd(r10,ewtabscale);
1118             ewitab           = _mm_cvttpd_epi32(ewrt);
1119 #ifdef __XOP__
1120             eweps            = _mm_frcz_pd(ewrt);
1121 #else
1122             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1123 #endif
1124             twoeweps         = _mm_add_pd(eweps,eweps);
1125             ewitab           = _mm_slli_epi32(ewitab,2);
1126             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1127             ewtabD           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1128             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1129             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
1130             ewtabFn          = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
1131             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1132             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
1133             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
1134             velec            = _mm_mul_pd(qq10,_mm_sub_pd(rinv10,velec));
1135             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
1136
1137             d                = _mm_sub_pd(r10,rswitch);
1138             d                = _mm_max_pd(d,_mm_setzero_pd());
1139             d2               = _mm_mul_pd(d,d);
1140             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
1141
1142             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
1143
1144             /* Evaluate switch function */
1145             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1146             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv10,_mm_mul_pd(velec,dsw)) );
1147             cutoff_mask      = _mm_cmplt_pd(rsq10,rcutoff2);
1148
1149             fscal            = felec;
1150
1151             fscal            = _mm_and_pd(fscal,cutoff_mask);
1152
1153             /* Update vectorial force */
1154             fix1             = _mm_macc_pd(dx10,fscal,fix1);
1155             fiy1             = _mm_macc_pd(dy10,fscal,fiy1);
1156             fiz1             = _mm_macc_pd(dz10,fscal,fiz1);
1157
1158             fjx0             = _mm_macc_pd(dx10,fscal,fjx0);
1159             fjy0             = _mm_macc_pd(dy10,fscal,fjy0);
1160             fjz0             = _mm_macc_pd(dz10,fscal,fjz0);
1161
1162             }
1163
1164             /**************************
1165              * CALCULATE INTERACTIONS *
1166              **************************/
1167
1168             if (gmx_mm_any_lt(rsq20,rcutoff2))
1169             {
1170
1171             r20              = _mm_mul_pd(rsq20,rinv20);
1172
1173             /* Compute parameters for interactions between i and j atoms */
1174             qq20             = _mm_mul_pd(iq2,jq0);
1175
1176             /* EWALD ELECTROSTATICS */
1177
1178             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1179             ewrt             = _mm_mul_pd(r20,ewtabscale);
1180             ewitab           = _mm_cvttpd_epi32(ewrt);
1181 #ifdef __XOP__
1182             eweps            = _mm_frcz_pd(ewrt);
1183 #else
1184             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1185 #endif
1186             twoeweps         = _mm_add_pd(eweps,eweps);
1187             ewitab           = _mm_slli_epi32(ewitab,2);
1188             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1189             ewtabD           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1190             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1191             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
1192             ewtabFn          = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
1193             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1194             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
1195             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
1196             velec            = _mm_mul_pd(qq20,_mm_sub_pd(rinv20,velec));
1197             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
1198
1199             d                = _mm_sub_pd(r20,rswitch);
1200             d                = _mm_max_pd(d,_mm_setzero_pd());
1201             d2               = _mm_mul_pd(d,d);
1202             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
1203
1204             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
1205
1206             /* Evaluate switch function */
1207             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1208             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv20,_mm_mul_pd(velec,dsw)) );
1209             cutoff_mask      = _mm_cmplt_pd(rsq20,rcutoff2);
1210
1211             fscal            = felec;
1212
1213             fscal            = _mm_and_pd(fscal,cutoff_mask);
1214
1215             /* Update vectorial force */
1216             fix2             = _mm_macc_pd(dx20,fscal,fix2);
1217             fiy2             = _mm_macc_pd(dy20,fscal,fiy2);
1218             fiz2             = _mm_macc_pd(dz20,fscal,fiz2);
1219
1220             fjx0             = _mm_macc_pd(dx20,fscal,fjx0);
1221             fjy0             = _mm_macc_pd(dy20,fscal,fjy0);
1222             fjz0             = _mm_macc_pd(dz20,fscal,fjz0);
1223
1224             }
1225
1226             /**************************
1227              * CALCULATE INTERACTIONS *
1228              **************************/
1229
1230             if (gmx_mm_any_lt(rsq30,rcutoff2))
1231             {
1232
1233             r30              = _mm_mul_pd(rsq30,rinv30);
1234
1235             /* Compute parameters for interactions between i and j atoms */
1236             qq30             = _mm_mul_pd(iq3,jq0);
1237
1238             /* EWALD ELECTROSTATICS */
1239
1240             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1241             ewrt             = _mm_mul_pd(r30,ewtabscale);
1242             ewitab           = _mm_cvttpd_epi32(ewrt);
1243 #ifdef __XOP__
1244             eweps            = _mm_frcz_pd(ewrt);
1245 #else
1246             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1247 #endif
1248             twoeweps         = _mm_add_pd(eweps,eweps);
1249             ewitab           = _mm_slli_epi32(ewitab,2);
1250             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1251             ewtabD           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1252             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1253             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
1254             ewtabFn          = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
1255             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1256             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
1257             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
1258             velec            = _mm_mul_pd(qq30,_mm_sub_pd(rinv30,velec));
1259             felec            = _mm_mul_pd(_mm_mul_pd(qq30,rinv30),_mm_sub_pd(rinvsq30,felec));
1260
1261             d                = _mm_sub_pd(r30,rswitch);
1262             d                = _mm_max_pd(d,_mm_setzero_pd());
1263             d2               = _mm_mul_pd(d,d);
1264             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
1265
1266             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
1267
1268             /* Evaluate switch function */
1269             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1270             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv30,_mm_mul_pd(velec,dsw)) );
1271             cutoff_mask      = _mm_cmplt_pd(rsq30,rcutoff2);
1272
1273             fscal            = felec;
1274
1275             fscal            = _mm_and_pd(fscal,cutoff_mask);
1276
1277             /* Update vectorial force */
1278             fix3             = _mm_macc_pd(dx30,fscal,fix3);
1279             fiy3             = _mm_macc_pd(dy30,fscal,fiy3);
1280             fiz3             = _mm_macc_pd(dz30,fscal,fiz3);
1281
1282             fjx0             = _mm_macc_pd(dx30,fscal,fjx0);
1283             fjy0             = _mm_macc_pd(dy30,fscal,fjy0);
1284             fjz0             = _mm_macc_pd(dz30,fscal,fjz0);
1285
1286             }
1287
1288             gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
1289
1290             /* Inner loop uses 257 flops */
1291         }
1292
1293         if(jidx<j_index_end)
1294         {
1295
1296             jnrA             = jjnr[jidx];
1297             j_coord_offsetA  = DIM*jnrA;
1298
1299             /* load j atom coordinates */
1300             gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
1301                                               &jx0,&jy0,&jz0);
1302
1303             /* Calculate displacement vector */
1304             dx00             = _mm_sub_pd(ix0,jx0);
1305             dy00             = _mm_sub_pd(iy0,jy0);
1306             dz00             = _mm_sub_pd(iz0,jz0);
1307             dx10             = _mm_sub_pd(ix1,jx0);
1308             dy10             = _mm_sub_pd(iy1,jy0);
1309             dz10             = _mm_sub_pd(iz1,jz0);
1310             dx20             = _mm_sub_pd(ix2,jx0);
1311             dy20             = _mm_sub_pd(iy2,jy0);
1312             dz20             = _mm_sub_pd(iz2,jz0);
1313             dx30             = _mm_sub_pd(ix3,jx0);
1314             dy30             = _mm_sub_pd(iy3,jy0);
1315             dz30             = _mm_sub_pd(iz3,jz0);
1316
1317             /* Calculate squared distance and things based on it */
1318             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1319             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
1320             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
1321             rsq30            = gmx_mm_calc_rsq_pd(dx30,dy30,dz30);
1322
1323             rinv00           = gmx_mm_invsqrt_pd(rsq00);
1324             rinv10           = gmx_mm_invsqrt_pd(rsq10);
1325             rinv20           = gmx_mm_invsqrt_pd(rsq20);
1326             rinv30           = gmx_mm_invsqrt_pd(rsq30);
1327
1328             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
1329             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
1330             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
1331             rinvsq30         = _mm_mul_pd(rinv30,rinv30);
1332
1333             /* Load parameters for j particles */
1334             jq0              = _mm_load_sd(charge+jnrA+0);
1335             vdwjidx0A        = 2*vdwtype[jnrA+0];
1336
1337             fjx0             = _mm_setzero_pd();
1338             fjy0             = _mm_setzero_pd();
1339             fjz0             = _mm_setzero_pd();
1340
1341             /**************************
1342              * CALCULATE INTERACTIONS *
1343              **************************/
1344
1345             if (gmx_mm_any_lt(rsq00,rcutoff2))
1346             {
1347
1348             r00              = _mm_mul_pd(rsq00,rinv00);
1349
1350             /* Compute parameters for interactions between i and j atoms */
1351             gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
1352
1353             /* LENNARD-JONES DISPERSION/REPULSION */
1354
1355             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1356             vvdw6            = _mm_mul_pd(c6_00,rinvsix);
1357             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
1358             vvdw             = _mm_msub_pd( vvdw12,one_twelfth, _mm_mul_pd(vvdw6,one_sixth) );
1359             fvdw             = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
1360
1361             d                = _mm_sub_pd(r00,rswitch);
1362             d                = _mm_max_pd(d,_mm_setzero_pd());
1363             d2               = _mm_mul_pd(d,d);
1364             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
1365
1366             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
1367
1368             /* Evaluate switch function */
1369             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1370             fvdw             = _mm_msub_pd( fvdw,sw , _mm_mul_pd(rinv00,_mm_mul_pd(vvdw,dsw)) );
1371             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
1372
1373             fscal            = fvdw;
1374
1375             fscal            = _mm_and_pd(fscal,cutoff_mask);
1376
1377             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1378
1379             /* Update vectorial force */
1380             fix0             = _mm_macc_pd(dx00,fscal,fix0);
1381             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
1382             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
1383
1384             fjx0             = _mm_macc_pd(dx00,fscal,fjx0);
1385             fjy0             = _mm_macc_pd(dy00,fscal,fjy0);
1386             fjz0             = _mm_macc_pd(dz00,fscal,fjz0);
1387
1388             }
1389
1390             /**************************
1391              * CALCULATE INTERACTIONS *
1392              **************************/
1393
1394             if (gmx_mm_any_lt(rsq10,rcutoff2))
1395             {
1396
1397             r10              = _mm_mul_pd(rsq10,rinv10);
1398
1399             /* Compute parameters for interactions between i and j atoms */
1400             qq10             = _mm_mul_pd(iq1,jq0);
1401
1402             /* EWALD ELECTROSTATICS */
1403
1404             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1405             ewrt             = _mm_mul_pd(r10,ewtabscale);
1406             ewitab           = _mm_cvttpd_epi32(ewrt);
1407 #ifdef __XOP__
1408             eweps            = _mm_frcz_pd(ewrt);
1409 #else
1410             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1411 #endif
1412             twoeweps         = _mm_add_pd(eweps,eweps);
1413             ewitab           = _mm_slli_epi32(ewitab,2);
1414             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1415             ewtabD           = _mm_setzero_pd();
1416             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1417             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
1418             ewtabFn          = _mm_setzero_pd();
1419             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1420             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
1421             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
1422             velec            = _mm_mul_pd(qq10,_mm_sub_pd(rinv10,velec));
1423             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
1424
1425             d                = _mm_sub_pd(r10,rswitch);
1426             d                = _mm_max_pd(d,_mm_setzero_pd());
1427             d2               = _mm_mul_pd(d,d);
1428             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
1429
1430             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
1431
1432             /* Evaluate switch function */
1433             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1434             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv10,_mm_mul_pd(velec,dsw)) );
1435             cutoff_mask      = _mm_cmplt_pd(rsq10,rcutoff2);
1436
1437             fscal            = felec;
1438
1439             fscal            = _mm_and_pd(fscal,cutoff_mask);
1440
1441             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1442
1443             /* Update vectorial force */
1444             fix1             = _mm_macc_pd(dx10,fscal,fix1);
1445             fiy1             = _mm_macc_pd(dy10,fscal,fiy1);
1446             fiz1             = _mm_macc_pd(dz10,fscal,fiz1);
1447
1448             fjx0             = _mm_macc_pd(dx10,fscal,fjx0);
1449             fjy0             = _mm_macc_pd(dy10,fscal,fjy0);
1450             fjz0             = _mm_macc_pd(dz10,fscal,fjz0);
1451
1452             }
1453
1454             /**************************
1455              * CALCULATE INTERACTIONS *
1456              **************************/
1457
1458             if (gmx_mm_any_lt(rsq20,rcutoff2))
1459             {
1460
1461             r20              = _mm_mul_pd(rsq20,rinv20);
1462
1463             /* Compute parameters for interactions between i and j atoms */
1464             qq20             = _mm_mul_pd(iq2,jq0);
1465
1466             /* EWALD ELECTROSTATICS */
1467
1468             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1469             ewrt             = _mm_mul_pd(r20,ewtabscale);
1470             ewitab           = _mm_cvttpd_epi32(ewrt);
1471 #ifdef __XOP__
1472             eweps            = _mm_frcz_pd(ewrt);
1473 #else
1474             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1475 #endif
1476             twoeweps         = _mm_add_pd(eweps,eweps);
1477             ewitab           = _mm_slli_epi32(ewitab,2);
1478             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1479             ewtabD           = _mm_setzero_pd();
1480             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1481             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
1482             ewtabFn          = _mm_setzero_pd();
1483             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1484             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
1485             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
1486             velec            = _mm_mul_pd(qq20,_mm_sub_pd(rinv20,velec));
1487             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
1488
1489             d                = _mm_sub_pd(r20,rswitch);
1490             d                = _mm_max_pd(d,_mm_setzero_pd());
1491             d2               = _mm_mul_pd(d,d);
1492             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
1493
1494             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
1495
1496             /* Evaluate switch function */
1497             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1498             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv20,_mm_mul_pd(velec,dsw)) );
1499             cutoff_mask      = _mm_cmplt_pd(rsq20,rcutoff2);
1500
1501             fscal            = felec;
1502
1503             fscal            = _mm_and_pd(fscal,cutoff_mask);
1504
1505             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1506
1507             /* Update vectorial force */
1508             fix2             = _mm_macc_pd(dx20,fscal,fix2);
1509             fiy2             = _mm_macc_pd(dy20,fscal,fiy2);
1510             fiz2             = _mm_macc_pd(dz20,fscal,fiz2);
1511
1512             fjx0             = _mm_macc_pd(dx20,fscal,fjx0);
1513             fjy0             = _mm_macc_pd(dy20,fscal,fjy0);
1514             fjz0             = _mm_macc_pd(dz20,fscal,fjz0);
1515
1516             }
1517
1518             /**************************
1519              * CALCULATE INTERACTIONS *
1520              **************************/
1521
1522             if (gmx_mm_any_lt(rsq30,rcutoff2))
1523             {
1524
1525             r30              = _mm_mul_pd(rsq30,rinv30);
1526
1527             /* Compute parameters for interactions between i and j atoms */
1528             qq30             = _mm_mul_pd(iq3,jq0);
1529
1530             /* EWALD ELECTROSTATICS */
1531
1532             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1533             ewrt             = _mm_mul_pd(r30,ewtabscale);
1534             ewitab           = _mm_cvttpd_epi32(ewrt);
1535 #ifdef __XOP__
1536             eweps            = _mm_frcz_pd(ewrt);
1537 #else
1538             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1539 #endif
1540             twoeweps         = _mm_add_pd(eweps,eweps);
1541             ewitab           = _mm_slli_epi32(ewitab,2);
1542             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1543             ewtabD           = _mm_setzero_pd();
1544             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1545             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
1546             ewtabFn          = _mm_setzero_pd();
1547             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1548             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
1549             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
1550             velec            = _mm_mul_pd(qq30,_mm_sub_pd(rinv30,velec));
1551             felec            = _mm_mul_pd(_mm_mul_pd(qq30,rinv30),_mm_sub_pd(rinvsq30,felec));
1552
1553             d                = _mm_sub_pd(r30,rswitch);
1554             d                = _mm_max_pd(d,_mm_setzero_pd());
1555             d2               = _mm_mul_pd(d,d);
1556             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
1557
1558             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
1559
1560             /* Evaluate switch function */
1561             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1562             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv30,_mm_mul_pd(velec,dsw)) );
1563             cutoff_mask      = _mm_cmplt_pd(rsq30,rcutoff2);
1564
1565             fscal            = felec;
1566
1567             fscal            = _mm_and_pd(fscal,cutoff_mask);
1568
1569             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1570
1571             /* Update vectorial force */
1572             fix3             = _mm_macc_pd(dx30,fscal,fix3);
1573             fiy3             = _mm_macc_pd(dy30,fscal,fiy3);
1574             fiz3             = _mm_macc_pd(dz30,fscal,fiz3);
1575
1576             fjx0             = _mm_macc_pd(dx30,fscal,fjx0);
1577             fjy0             = _mm_macc_pd(dy30,fscal,fjy0);
1578             fjz0             = _mm_macc_pd(dz30,fscal,fjz0);
1579
1580             }
1581
1582             gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1583
1584             /* Inner loop uses 257 flops */
1585         }
1586
1587         /* End of innermost loop */
1588
1589         gmx_mm_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1590                                               f+i_coord_offset,fshift+i_shift_offset);
1591
1592         /* Increment number of inner iterations */
1593         inneriter                  += j_index_end - j_index_start;
1594
1595         /* Outer loop uses 24 flops */
1596     }
1597
1598     /* Increment number of outer iterations */
1599     outeriter        += nri;
1600
1601     /* Update outer/inner flops */
1602
1603     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_F,outeriter*24 + inneriter*257);
1604 }