src/gromacs/gmxlib/nonbonded/nb_kernel_avx_128_fma_double/nb_kernel_ElecEwSw_VdwLJSw_GeomW3W3_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 "gromacs/legacyheaders/types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "gromacs/legacyheaders/nrnb.h"
  46
  47 #include "gromacs/simd/math_x86_avx_128_fma_double.h"
  48 #include "kernelutil_x86_avx_128_fma_double.h"
  49
  50 /*
  51  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSw_VdwLJSw_GeomW3W3_VF_avx_128_fma_double
  52  * Electrostatics interaction: Ewald
  53  * VdW interaction:            LennardJones
  54  * Geometry:                   Water3-Water3
  55  * Calculate force/pot:        PotentialAndForce
  56  */
  57 void
  58 nb_kernel_ElecEwSw_VdwLJSw_GeomW3W3_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              vdwioffset0;
  81     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  82     int              vdwioffset1;
  83     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  84     int              vdwioffset2;
  85     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  86     int              vdwjidx0A,vdwjidx0B;
  87     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  88     int              vdwjidx1A,vdwjidx1B;
  89     __m128d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  90     int              vdwjidx2A,vdwjidx2B;
  91     __m128d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  92     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  93     __m128d          dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
  94     __m128d          dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
  95     __m128d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  96     __m128d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
  97     __m128d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
  98     __m128d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  99     __m128d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 100     __m128d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 101     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
 102     real             *charge;
 103     int              nvdwtype;
 104     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 105     int              *vdwtype;
 106     real             *vdwparam;
 107     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
 108     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
 109     __m128i          ewitab;
 110     __m128d          ewtabscale,eweps,twoeweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 111     real             *ewtab;
 112     __m128d          rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
 113     real             rswitch_scalar,d_scalar;
 114     __m128d          dummy_mask,cutoff_mask;
 115     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 116     __m128d          one     = _mm_set1_pd(1.0);
 117     __m128d          two     = _mm_set1_pd(2.0);
 118     x                = xx[0];
 119     f                = ff[0];
 120
 121     nri              = nlist->nri;
 122     iinr             = nlist->iinr;
 123     jindex           = nlist->jindex;
 124     jjnr             = nlist->jjnr;
 125     shiftidx         = nlist->shift;
 126     gid              = nlist->gid;
 127     shiftvec         = fr->shift_vec[0];
 128     fshift           = fr->fshift[0];
 129     facel            = _mm_set1_pd(fr->epsfac);
 130     charge           = mdatoms->chargeA;
 131     nvdwtype         = fr->ntype;
 132     vdwparam         = fr->nbfp;
 133     vdwtype          = mdatoms->typeA;
 134
 135     sh_ewald         = _mm_set1_pd(fr->ic->sh_ewald);
 136     ewtab            = fr->ic->tabq_coul_FDV0;
 137     ewtabscale       = _mm_set1_pd(fr->ic->tabq_scale);
 138     ewtabhalfspace   = _mm_set1_pd(0.5/fr->ic->tabq_scale);
 139
 140     /* Setup water-specific parameters */
 141     inr              = nlist->iinr[0];
 142     iq0              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
 143     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
 144     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
 145     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 146
 147     jq0              = _mm_set1_pd(charge[inr+0]);
 148     jq1              = _mm_set1_pd(charge[inr+1]);
 149     jq2              = _mm_set1_pd(charge[inr+2]);
 150     vdwjidx0A        = 2*vdwtype[inr+0];
 151     qq00             = _mm_mul_pd(iq0,jq0);
 152     c6_00            = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
 153     c12_00           = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
 154     qq01             = _mm_mul_pd(iq0,jq1);
 155     qq02             = _mm_mul_pd(iq0,jq2);
 156     qq10             = _mm_mul_pd(iq1,jq0);
 157     qq11             = _mm_mul_pd(iq1,jq1);
 158     qq12             = _mm_mul_pd(iq1,jq2);
 159     qq20             = _mm_mul_pd(iq2,jq0);
 160     qq21             = _mm_mul_pd(iq2,jq1);
 161     qq22             = _mm_mul_pd(iq2,jq2);
 162
 163     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 164     rcutoff_scalar   = fr->rcoulomb;
 165     rcutoff          = _mm_set1_pd(rcutoff_scalar);
 166     rcutoff2         = _mm_mul_pd(rcutoff,rcutoff);
 167
 168     rswitch_scalar   = fr->rcoulomb_switch;
 169     rswitch          = _mm_set1_pd(rswitch_scalar);
 170     /* Setup switch parameters */
 171     d_scalar         = rcutoff_scalar-rswitch_scalar;
 172     d                = _mm_set1_pd(d_scalar);
 173     swV3             = _mm_set1_pd(-10.0/(d_scalar*d_scalar*d_scalar));
 174     swV4             = _mm_set1_pd( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 175     swV5             = _mm_set1_pd( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 176     swF2             = _mm_set1_pd(-30.0/(d_scalar*d_scalar*d_scalar));
 177     swF3             = _mm_set1_pd( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 178     swF4             = _mm_set1_pd(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 179
 180     /* Avoid stupid compiler warnings */
 181     jnrA = jnrB = 0;
 182     j_coord_offsetA = 0;
 183     j_coord_offsetB = 0;
 184
 185     outeriter        = 0;
 186     inneriter        = 0;
 187
 188     /* Start outer loop over neighborlists */
 189     for(iidx=0; iidx<nri; iidx++)
 190     {
 191         /* Load shift vector for this list */
 192         i_shift_offset   = DIM*shiftidx[iidx];
 193
 194         /* Load limits for loop over neighbors */
 195         j_index_start    = jindex[iidx];
 196         j_index_end      = jindex[iidx+1];
 197
 198         /* Get outer coordinate index */
 199         inr              = iinr[iidx];
 200         i_coord_offset   = DIM*inr;
 201
 202         /* Load i particle coords and add shift vector */
 203         gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 204                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 205
 206         fix0             = _mm_setzero_pd();
 207         fiy0             = _mm_setzero_pd();
 208         fiz0             = _mm_setzero_pd();
 209         fix1             = _mm_setzero_pd();
 210         fiy1             = _mm_setzero_pd();
 211         fiz1             = _mm_setzero_pd();
 212         fix2             = _mm_setzero_pd();
 213         fiy2             = _mm_setzero_pd();
 214         fiz2             = _mm_setzero_pd();
 215
 216         /* Reset potential sums */
 217         velecsum         = _mm_setzero_pd();
 218         vvdwsum          = _mm_setzero_pd();
 219
 220         /* Start inner kernel loop */
 221         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 222         {
 223
 224             /* Get j neighbor index, and coordinate index */
 225             jnrA             = jjnr[jidx];
 226             jnrB             = jjnr[jidx+1];
 227             j_coord_offsetA  = DIM*jnrA;
 228             j_coord_offsetB  = DIM*jnrB;
 229
 230             /* load j atom coordinates */
 231             gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 232                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 233
 234             /* Calculate displacement vector */
 235             dx00             = _mm_sub_pd(ix0,jx0);
 236             dy00             = _mm_sub_pd(iy0,jy0);
 237             dz00             = _mm_sub_pd(iz0,jz0);
 238             dx01             = _mm_sub_pd(ix0,jx1);
 239             dy01             = _mm_sub_pd(iy0,jy1);
 240             dz01             = _mm_sub_pd(iz0,jz1);
 241             dx02             = _mm_sub_pd(ix0,jx2);
 242             dy02             = _mm_sub_pd(iy0,jy2);
 243             dz02             = _mm_sub_pd(iz0,jz2);
 244             dx10             = _mm_sub_pd(ix1,jx0);
 245             dy10             = _mm_sub_pd(iy1,jy0);
 246             dz10             = _mm_sub_pd(iz1,jz0);
 247             dx11             = _mm_sub_pd(ix1,jx1);
 248             dy11             = _mm_sub_pd(iy1,jy1);
 249             dz11             = _mm_sub_pd(iz1,jz1);
 250             dx12             = _mm_sub_pd(ix1,jx2);
 251             dy12             = _mm_sub_pd(iy1,jy2);
 252             dz12             = _mm_sub_pd(iz1,jz2);
 253             dx20             = _mm_sub_pd(ix2,jx0);
 254             dy20             = _mm_sub_pd(iy2,jy0);
 255             dz20             = _mm_sub_pd(iz2,jz0);
 256             dx21             = _mm_sub_pd(ix2,jx1);
 257             dy21             = _mm_sub_pd(iy2,jy1);
 258             dz21             = _mm_sub_pd(iz2,jz1);
 259             dx22             = _mm_sub_pd(ix2,jx2);
 260             dy22             = _mm_sub_pd(iy2,jy2);
 261             dz22             = _mm_sub_pd(iz2,jz2);
 262
 263             /* Calculate squared distance and things based on it */
 264             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 265             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
 266             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
 267             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 268             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 269             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 270             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 271             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 272             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 273
 274             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 275             rinv01           = gmx_mm_invsqrt_pd(rsq01);
 276             rinv02           = gmx_mm_invsqrt_pd(rsq02);
 277             rinv10           = gmx_mm_invsqrt_pd(rsq10);
 278             rinv11           = gmx_mm_invsqrt_pd(rsq11);
 279             rinv12           = gmx_mm_invsqrt_pd(rsq12);
 280             rinv20           = gmx_mm_invsqrt_pd(rsq20);
 281             rinv21           = gmx_mm_invsqrt_pd(rsq21);
 282             rinv22           = gmx_mm_invsqrt_pd(rsq22);
 283
 284             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 285             rinvsq01         = _mm_mul_pd(rinv01,rinv01);
 286             rinvsq02         = _mm_mul_pd(rinv02,rinv02);
 287             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
 288             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
 289             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
 290             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
 291             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
 292             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
 293
 294             fjx0             = _mm_setzero_pd();
 295             fjy0             = _mm_setzero_pd();
 296             fjz0             = _mm_setzero_pd();
 297             fjx1             = _mm_setzero_pd();
 298             fjy1             = _mm_setzero_pd();
 299             fjz1             = _mm_setzero_pd();
 300             fjx2             = _mm_setzero_pd();
 301             fjy2             = _mm_setzero_pd();
 302             fjz2             = _mm_setzero_pd();
 303
 304             /**************************
 305              * CALCULATE INTERACTIONS *
 306              **************************/
 307
 308             if (gmx_mm_any_lt(rsq00,rcutoff2))
 309             {
 310
 311             r00              = _mm_mul_pd(rsq00,rinv00);
 312
 313             /* EWALD ELECTROSTATICS */
 314
 315             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 316             ewrt             = _mm_mul_pd(r00,ewtabscale);
 317             ewitab           = _mm_cvttpd_epi32(ewrt);
 318 #ifdef __XOP__
 319             eweps            = _mm_frcz_pd(ewrt);
 320 #else
 321             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 322 #endif
 323             twoeweps         = _mm_add_pd(eweps,eweps);
 324             ewitab           = _mm_slli_epi32(ewitab,2);
 325             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 326             ewtabD           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 327             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 328             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
 329             ewtabFn          = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
 330             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 331             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
 332             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
 333             velec            = _mm_mul_pd(qq00,_mm_sub_pd(rinv00,velec));
 334             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
 335
 336             /* LENNARD-JONES DISPERSION/REPULSION */
 337
 338             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 339             vvdw6            = _mm_mul_pd(c6_00,rinvsix);
 340             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
 341             vvdw             = _mm_msub_pd( vvdw12,one_twelfth, _mm_mul_pd(vvdw6,one_sixth) );
 342             fvdw             = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
 343
 344             d                = _mm_sub_pd(r00,rswitch);
 345             d                = _mm_max_pd(d,_mm_setzero_pd());
 346             d2               = _mm_mul_pd(d,d);
 347             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
 348
 349             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
 350
 351             /* Evaluate switch function */
 352             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 353             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv00,_mm_mul_pd(velec,dsw)) );
 354             fvdw             = _mm_msub_pd( fvdw,sw , _mm_mul_pd(rinv00,_mm_mul_pd(vvdw,dsw)) );
 355             velec            = _mm_mul_pd(velec,sw);
 356             vvdw             = _mm_mul_pd(vvdw,sw);
 357             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
 358
 359             /* Update potential sum for this i atom from the interaction with this j atom. */
 360             velec            = _mm_and_pd(velec,cutoff_mask);
 361             velecsum         = _mm_add_pd(velecsum,velec);
 362             vvdw             = _mm_and_pd(vvdw,cutoff_mask);
 363             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 364
 365             fscal            = _mm_add_pd(felec,fvdw);
 366
 367             fscal            = _mm_and_pd(fscal,cutoff_mask);
 368
 369             /* Update vectorial force */
 370             fix0             = _mm_macc_pd(dx00,fscal,fix0);
 371             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
 372             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
 373
 374             fjx0             = _mm_macc_pd(dx00,fscal,fjx0);
 375             fjy0             = _mm_macc_pd(dy00,fscal,fjy0);
 376             fjz0             = _mm_macc_pd(dz00,fscal,fjz0);
 377
 378             }
 379
 380             /**************************
 381              * CALCULATE INTERACTIONS *
 382              **************************/
 383
 384             if (gmx_mm_any_lt(rsq01,rcutoff2))
 385             {
 386
 387             r01              = _mm_mul_pd(rsq01,rinv01);
 388
 389             /* EWALD ELECTROSTATICS */
 390
 391             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 392             ewrt             = _mm_mul_pd(r01,ewtabscale);
 393             ewitab           = _mm_cvttpd_epi32(ewrt);
 394 #ifdef __XOP__
 395             eweps            = _mm_frcz_pd(ewrt);
 396 #else
 397             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 398 #endif
 399             twoeweps         = _mm_add_pd(eweps,eweps);
 400             ewitab           = _mm_slli_epi32(ewitab,2);
 401             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 402             ewtabD           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 403             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 404             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
 405             ewtabFn          = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
 406             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 407             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
 408             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
 409             velec            = _mm_mul_pd(qq01,_mm_sub_pd(rinv01,velec));
 410             felec            = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
 411
 412             d                = _mm_sub_pd(r01,rswitch);
 413             d                = _mm_max_pd(d,_mm_setzero_pd());
 414             d2               = _mm_mul_pd(d,d);
 415             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
 416
 417             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
 418
 419             /* Evaluate switch function */
 420             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 421             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv01,_mm_mul_pd(velec,dsw)) );
 422             velec            = _mm_mul_pd(velec,sw);
 423             cutoff_mask      = _mm_cmplt_pd(rsq01,rcutoff2);
 424
 425             /* Update potential sum for this i atom from the interaction with this j atom. */
 426             velec            = _mm_and_pd(velec,cutoff_mask);
 427             velecsum         = _mm_add_pd(velecsum,velec);
 428
 429             fscal            = felec;
 430
 431             fscal            = _mm_and_pd(fscal,cutoff_mask);
 432
 433             /* Update vectorial force */
 434             fix0             = _mm_macc_pd(dx01,fscal,fix0);
 435             fiy0             = _mm_macc_pd(dy01,fscal,fiy0);
 436             fiz0             = _mm_macc_pd(dz01,fscal,fiz0);
 437
 438             fjx1             = _mm_macc_pd(dx01,fscal,fjx1);
 439             fjy1             = _mm_macc_pd(dy01,fscal,fjy1);
 440             fjz1             = _mm_macc_pd(dz01,fscal,fjz1);
 441
 442             }
 443
 444             /**************************
 445              * CALCULATE INTERACTIONS *
 446              **************************/
 447
 448             if (gmx_mm_any_lt(rsq02,rcutoff2))
 449             {
 450
 451             r02              = _mm_mul_pd(rsq02,rinv02);
 452
 453             /* EWALD ELECTROSTATICS */
 454
 455             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 456             ewrt             = _mm_mul_pd(r02,ewtabscale);
 457             ewitab           = _mm_cvttpd_epi32(ewrt);
 458 #ifdef __XOP__
 459             eweps            = _mm_frcz_pd(ewrt);
 460 #else
 461             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 462 #endif
 463             twoeweps         = _mm_add_pd(eweps,eweps);
 464             ewitab           = _mm_slli_epi32(ewitab,2);
 465             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 466             ewtabD           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 467             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 468             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
 469             ewtabFn          = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
 470             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 471             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
 472             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
 473             velec            = _mm_mul_pd(qq02,_mm_sub_pd(rinv02,velec));
 474             felec            = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
 475
 476             d                = _mm_sub_pd(r02,rswitch);
 477             d                = _mm_max_pd(d,_mm_setzero_pd());
 478             d2               = _mm_mul_pd(d,d);
 479             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
 480
 481             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
 482
 483             /* Evaluate switch function */
 484             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 485             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv02,_mm_mul_pd(velec,dsw)) );
 486             velec            = _mm_mul_pd(velec,sw);
 487             cutoff_mask      = _mm_cmplt_pd(rsq02,rcutoff2);
 488
 489             /* Update potential sum for this i atom from the interaction with this j atom. */
 490             velec            = _mm_and_pd(velec,cutoff_mask);
 491             velecsum         = _mm_add_pd(velecsum,velec);
 492
 493             fscal            = felec;
 494
 495             fscal            = _mm_and_pd(fscal,cutoff_mask);
 496
 497             /* Update vectorial force */
 498             fix0             = _mm_macc_pd(dx02,fscal,fix0);
 499             fiy0             = _mm_macc_pd(dy02,fscal,fiy0);
 500             fiz0             = _mm_macc_pd(dz02,fscal,fiz0);
 501
 502             fjx2             = _mm_macc_pd(dx02,fscal,fjx2);
 503             fjy2             = _mm_macc_pd(dy02,fscal,fjy2);
 504             fjz2             = _mm_macc_pd(dz02,fscal,fjz2);
 505
 506             }
 507
 508             /**************************
 509              * CALCULATE INTERACTIONS *
 510              **************************/
 511
 512             if (gmx_mm_any_lt(rsq10,rcutoff2))
 513             {
 514
 515             r10              = _mm_mul_pd(rsq10,rinv10);
 516
 517             /* EWALD ELECTROSTATICS */
 518
 519             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 520             ewrt             = _mm_mul_pd(r10,ewtabscale);
 521             ewitab           = _mm_cvttpd_epi32(ewrt);
 522 #ifdef __XOP__
 523             eweps            = _mm_frcz_pd(ewrt);
 524 #else
 525             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 526 #endif
 527             twoeweps         = _mm_add_pd(eweps,eweps);
 528             ewitab           = _mm_slli_epi32(ewitab,2);
 529             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 530             ewtabD           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 531             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 532             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
 533             ewtabFn          = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
 534             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 535             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
 536             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
 537             velec            = _mm_mul_pd(qq10,_mm_sub_pd(rinv10,velec));
 538             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
 539
 540             d                = _mm_sub_pd(r10,rswitch);
 541             d                = _mm_max_pd(d,_mm_setzero_pd());
 542             d2               = _mm_mul_pd(d,d);
 543             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
 544
 545             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
 546
 547             /* Evaluate switch function */
 548             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 549             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv10,_mm_mul_pd(velec,dsw)) );
 550             velec            = _mm_mul_pd(velec,sw);
 551             cutoff_mask      = _mm_cmplt_pd(rsq10,rcutoff2);
 552
 553             /* Update potential sum for this i atom from the interaction with this j atom. */
 554             velec            = _mm_and_pd(velec,cutoff_mask);
 555             velecsum         = _mm_add_pd(velecsum,velec);
 556
 557             fscal            = felec;
 558
 559             fscal            = _mm_and_pd(fscal,cutoff_mask);
 560
 561             /* Update vectorial force */
 562             fix1             = _mm_macc_pd(dx10,fscal,fix1);
 563             fiy1             = _mm_macc_pd(dy10,fscal,fiy1);
 564             fiz1             = _mm_macc_pd(dz10,fscal,fiz1);
 565
 566             fjx0             = _mm_macc_pd(dx10,fscal,fjx0);
 567             fjy0             = _mm_macc_pd(dy10,fscal,fjy0);
 568             fjz0             = _mm_macc_pd(dz10,fscal,fjz0);
 569
 570             }
 571
 572             /**************************
 573              * CALCULATE INTERACTIONS *
 574              **************************/
 575
 576             if (gmx_mm_any_lt(rsq11,rcutoff2))
 577             {
 578
 579             r11              = _mm_mul_pd(rsq11,rinv11);
 580
 581             /* EWALD ELECTROSTATICS */
 582
 583             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 584             ewrt             = _mm_mul_pd(r11,ewtabscale);
 585             ewitab           = _mm_cvttpd_epi32(ewrt);
 586 #ifdef __XOP__
 587             eweps            = _mm_frcz_pd(ewrt);
 588 #else
 589             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 590 #endif
 591             twoeweps         = _mm_add_pd(eweps,eweps);
 592             ewitab           = _mm_slli_epi32(ewitab,2);
 593             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 594             ewtabD           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 595             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 596             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
 597             ewtabFn          = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
 598             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 599             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
 600             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
 601             velec            = _mm_mul_pd(qq11,_mm_sub_pd(rinv11,velec));
 602             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
 603
 604             d                = _mm_sub_pd(r11,rswitch);
 605             d                = _mm_max_pd(d,_mm_setzero_pd());
 606             d2               = _mm_mul_pd(d,d);
 607             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
 608
 609             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
 610
 611             /* Evaluate switch function */
 612             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 613             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv11,_mm_mul_pd(velec,dsw)) );
 614             velec            = _mm_mul_pd(velec,sw);
 615             cutoff_mask      = _mm_cmplt_pd(rsq11,rcutoff2);
 616
 617             /* Update potential sum for this i atom from the interaction with this j atom. */
 618             velec            = _mm_and_pd(velec,cutoff_mask);
 619             velecsum         = _mm_add_pd(velecsum,velec);
 620
 621             fscal            = felec;
 622
 623             fscal            = _mm_and_pd(fscal,cutoff_mask);
 624
 625             /* Update vectorial force */
 626             fix1             = _mm_macc_pd(dx11,fscal,fix1);
 627             fiy1             = _mm_macc_pd(dy11,fscal,fiy1);
 628             fiz1             = _mm_macc_pd(dz11,fscal,fiz1);
 629
 630             fjx1             = _mm_macc_pd(dx11,fscal,fjx1);
 631             fjy1             = _mm_macc_pd(dy11,fscal,fjy1);
 632             fjz1             = _mm_macc_pd(dz11,fscal,fjz1);
 633
 634             }
 635
 636             /**************************
 637              * CALCULATE INTERACTIONS *
 638              **************************/
 639
 640             if (gmx_mm_any_lt(rsq12,rcutoff2))
 641             {
 642
 643             r12              = _mm_mul_pd(rsq12,rinv12);
 644
 645             /* EWALD ELECTROSTATICS */
 646
 647             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 648             ewrt             = _mm_mul_pd(r12,ewtabscale);
 649             ewitab           = _mm_cvttpd_epi32(ewrt);
 650 #ifdef __XOP__
 651             eweps            = _mm_frcz_pd(ewrt);
 652 #else
 653             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 654 #endif
 655             twoeweps         = _mm_add_pd(eweps,eweps);
 656             ewitab           = _mm_slli_epi32(ewitab,2);
 657             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 658             ewtabD           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 659             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 660             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
 661             ewtabFn          = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
 662             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 663             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
 664             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
 665             velec            = _mm_mul_pd(qq12,_mm_sub_pd(rinv12,velec));
 666             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
 667
 668             d                = _mm_sub_pd(r12,rswitch);
 669             d                = _mm_max_pd(d,_mm_setzero_pd());
 670             d2               = _mm_mul_pd(d,d);
 671             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
 672
 673             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
 674
 675             /* Evaluate switch function */
 676             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 677             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv12,_mm_mul_pd(velec,dsw)) );
 678             velec            = _mm_mul_pd(velec,sw);
 679             cutoff_mask      = _mm_cmplt_pd(rsq12,rcutoff2);
 680
 681             /* Update potential sum for this i atom from the interaction with this j atom. */
 682             velec            = _mm_and_pd(velec,cutoff_mask);
 683             velecsum         = _mm_add_pd(velecsum,velec);
 684
 685             fscal            = felec;
 686
 687             fscal            = _mm_and_pd(fscal,cutoff_mask);
 688
 689             /* Update vectorial force */
 690             fix1             = _mm_macc_pd(dx12,fscal,fix1);
 691             fiy1             = _mm_macc_pd(dy12,fscal,fiy1);
 692             fiz1             = _mm_macc_pd(dz12,fscal,fiz1);
 693
 694             fjx2             = _mm_macc_pd(dx12,fscal,fjx2);
 695             fjy2             = _mm_macc_pd(dy12,fscal,fjy2);
 696             fjz2             = _mm_macc_pd(dz12,fscal,fjz2);
 697
 698             }
 699
 700             /**************************
 701              * CALCULATE INTERACTIONS *
 702              **************************/
 703
 704             if (gmx_mm_any_lt(rsq20,rcutoff2))
 705             {
 706
 707             r20              = _mm_mul_pd(rsq20,rinv20);
 708
 709             /* EWALD ELECTROSTATICS */
 710
 711             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 712             ewrt             = _mm_mul_pd(r20,ewtabscale);
 713             ewitab           = _mm_cvttpd_epi32(ewrt);
 714 #ifdef __XOP__
 715             eweps            = _mm_frcz_pd(ewrt);
 716 #else
 717             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 718 #endif
 719             twoeweps         = _mm_add_pd(eweps,eweps);
 720             ewitab           = _mm_slli_epi32(ewitab,2);
 721             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 722             ewtabD           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 723             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 724             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
 725             ewtabFn          = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
 726             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 727             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
 728             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
 729             velec            = _mm_mul_pd(qq20,_mm_sub_pd(rinv20,velec));
 730             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
 731
 732             d                = _mm_sub_pd(r20,rswitch);
 733             d                = _mm_max_pd(d,_mm_setzero_pd());
 734             d2               = _mm_mul_pd(d,d);
 735             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
 736
 737             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
 738
 739             /* Evaluate switch function */
 740             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 741             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv20,_mm_mul_pd(velec,dsw)) );
 742             velec            = _mm_mul_pd(velec,sw);
 743             cutoff_mask      = _mm_cmplt_pd(rsq20,rcutoff2);
 744
 745             /* Update potential sum for this i atom from the interaction with this j atom. */
 746             velec            = _mm_and_pd(velec,cutoff_mask);
 747             velecsum         = _mm_add_pd(velecsum,velec);
 748
 749             fscal            = felec;
 750
 751             fscal            = _mm_and_pd(fscal,cutoff_mask);
 752
 753             /* Update vectorial force */
 754             fix2             = _mm_macc_pd(dx20,fscal,fix2);
 755             fiy2             = _mm_macc_pd(dy20,fscal,fiy2);
 756             fiz2             = _mm_macc_pd(dz20,fscal,fiz2);
 757
 758             fjx0             = _mm_macc_pd(dx20,fscal,fjx0);
 759             fjy0             = _mm_macc_pd(dy20,fscal,fjy0);
 760             fjz0             = _mm_macc_pd(dz20,fscal,fjz0);
 761
 762             }
 763
 764             /**************************
 765              * CALCULATE INTERACTIONS *
 766              **************************/
 767
 768             if (gmx_mm_any_lt(rsq21,rcutoff2))
 769             {
 770
 771             r21              = _mm_mul_pd(rsq21,rinv21);
 772
 773             /* EWALD ELECTROSTATICS */
 774
 775             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 776             ewrt             = _mm_mul_pd(r21,ewtabscale);
 777             ewitab           = _mm_cvttpd_epi32(ewrt);
 778 #ifdef __XOP__
 779             eweps            = _mm_frcz_pd(ewrt);
 780 #else
 781             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 782 #endif
 783             twoeweps         = _mm_add_pd(eweps,eweps);
 784             ewitab           = _mm_slli_epi32(ewitab,2);
 785             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 786             ewtabD           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 787             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 788             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
 789             ewtabFn          = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
 790             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 791             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
 792             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
 793             velec            = _mm_mul_pd(qq21,_mm_sub_pd(rinv21,velec));
 794             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
 795
 796             d                = _mm_sub_pd(r21,rswitch);
 797             d                = _mm_max_pd(d,_mm_setzero_pd());
 798             d2               = _mm_mul_pd(d,d);
 799             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
 800
 801             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
 802
 803             /* Evaluate switch function */
 804             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 805             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv21,_mm_mul_pd(velec,dsw)) );
 806             velec            = _mm_mul_pd(velec,sw);
 807             cutoff_mask      = _mm_cmplt_pd(rsq21,rcutoff2);
 808
 809             /* Update potential sum for this i atom from the interaction with this j atom. */
 810             velec            = _mm_and_pd(velec,cutoff_mask);
 811             velecsum         = _mm_add_pd(velecsum,velec);
 812
 813             fscal            = felec;
 814
 815             fscal            = _mm_and_pd(fscal,cutoff_mask);
 816
 817             /* Update vectorial force */
 818             fix2             = _mm_macc_pd(dx21,fscal,fix2);
 819             fiy2             = _mm_macc_pd(dy21,fscal,fiy2);
 820             fiz2             = _mm_macc_pd(dz21,fscal,fiz2);
 821
 822             fjx1             = _mm_macc_pd(dx21,fscal,fjx1);
 823             fjy1             = _mm_macc_pd(dy21,fscal,fjy1);
 824             fjz1             = _mm_macc_pd(dz21,fscal,fjz1);
 825
 826             }
 827
 828             /**************************
 829              * CALCULATE INTERACTIONS *
 830              **************************/
 831
 832             if (gmx_mm_any_lt(rsq22,rcutoff2))
 833             {
 834
 835             r22              = _mm_mul_pd(rsq22,rinv22);
 836
 837             /* EWALD ELECTROSTATICS */
 838
 839             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 840             ewrt             = _mm_mul_pd(r22,ewtabscale);
 841             ewitab           = _mm_cvttpd_epi32(ewrt);
 842 #ifdef __XOP__
 843             eweps            = _mm_frcz_pd(ewrt);
 844 #else
 845             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 846 #endif
 847             twoeweps         = _mm_add_pd(eweps,eweps);
 848             ewitab           = _mm_slli_epi32(ewitab,2);
 849             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 850             ewtabD           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 851             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 852             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
 853             ewtabFn          = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
 854             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 855             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
 856             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
 857             velec            = _mm_mul_pd(qq22,_mm_sub_pd(rinv22,velec));
 858             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
 859
 860             d                = _mm_sub_pd(r22,rswitch);
 861             d                = _mm_max_pd(d,_mm_setzero_pd());
 862             d2               = _mm_mul_pd(d,d);
 863             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
 864
 865             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
 866
 867             /* Evaluate switch function */
 868             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 869             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv22,_mm_mul_pd(velec,dsw)) );
 870             velec            = _mm_mul_pd(velec,sw);
 871             cutoff_mask      = _mm_cmplt_pd(rsq22,rcutoff2);
 872
 873             /* Update potential sum for this i atom from the interaction with this j atom. */
 874             velec            = _mm_and_pd(velec,cutoff_mask);
 875             velecsum         = _mm_add_pd(velecsum,velec);
 876
 877             fscal            = felec;
 878
 879             fscal            = _mm_and_pd(fscal,cutoff_mask);
 880
 881             /* Update vectorial force */
 882             fix2             = _mm_macc_pd(dx22,fscal,fix2);
 883             fiy2             = _mm_macc_pd(dy22,fscal,fiy2);
 884             fiz2             = _mm_macc_pd(dz22,fscal,fiz2);
 885
 886             fjx2             = _mm_macc_pd(dx22,fscal,fjx2);
 887             fjy2             = _mm_macc_pd(dy22,fscal,fjy2);
 888             fjz2             = _mm_macc_pd(dz22,fscal,fjz2);
 889
 890             }
 891
 892             gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 893
 894             /* Inner loop uses 630 flops */
 895         }
 896
 897         if(jidx<j_index_end)
 898         {
 899
 900             jnrA             = jjnr[jidx];
 901             j_coord_offsetA  = DIM*jnrA;
 902
 903             /* load j atom coordinates */
 904             gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 905                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 906
 907             /* Calculate displacement vector */
 908             dx00             = _mm_sub_pd(ix0,jx0);
 909             dy00             = _mm_sub_pd(iy0,jy0);
 910             dz00             = _mm_sub_pd(iz0,jz0);
 911             dx01             = _mm_sub_pd(ix0,jx1);
 912             dy01             = _mm_sub_pd(iy0,jy1);
 913             dz01             = _mm_sub_pd(iz0,jz1);
 914             dx02             = _mm_sub_pd(ix0,jx2);
 915             dy02             = _mm_sub_pd(iy0,jy2);
 916             dz02             = _mm_sub_pd(iz0,jz2);
 917             dx10             = _mm_sub_pd(ix1,jx0);
 918             dy10             = _mm_sub_pd(iy1,jy0);
 919             dz10             = _mm_sub_pd(iz1,jz0);
 920             dx11             = _mm_sub_pd(ix1,jx1);
 921             dy11             = _mm_sub_pd(iy1,jy1);
 922             dz11             = _mm_sub_pd(iz1,jz1);
 923             dx12             = _mm_sub_pd(ix1,jx2);
 924             dy12             = _mm_sub_pd(iy1,jy2);
 925             dz12             = _mm_sub_pd(iz1,jz2);
 926             dx20             = _mm_sub_pd(ix2,jx0);
 927             dy20             = _mm_sub_pd(iy2,jy0);
 928             dz20             = _mm_sub_pd(iz2,jz0);
 929             dx21             = _mm_sub_pd(ix2,jx1);
 930             dy21             = _mm_sub_pd(iy2,jy1);
 931             dz21             = _mm_sub_pd(iz2,jz1);
 932             dx22             = _mm_sub_pd(ix2,jx2);
 933             dy22             = _mm_sub_pd(iy2,jy2);
 934             dz22             = _mm_sub_pd(iz2,jz2);
 935
 936             /* Calculate squared distance and things based on it */
 937             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 938             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
 939             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
 940             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 941             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 942             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 943             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 944             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 945             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 946
 947             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 948             rinv01           = gmx_mm_invsqrt_pd(rsq01);
 949             rinv02           = gmx_mm_invsqrt_pd(rsq02);
 950             rinv10           = gmx_mm_invsqrt_pd(rsq10);
 951             rinv11           = gmx_mm_invsqrt_pd(rsq11);
 952             rinv12           = gmx_mm_invsqrt_pd(rsq12);
 953             rinv20           = gmx_mm_invsqrt_pd(rsq20);
 954             rinv21           = gmx_mm_invsqrt_pd(rsq21);
 955             rinv22           = gmx_mm_invsqrt_pd(rsq22);
 956
 957             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 958             rinvsq01         = _mm_mul_pd(rinv01,rinv01);
 959             rinvsq02         = _mm_mul_pd(rinv02,rinv02);
 960             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
 961             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
 962             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
 963             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
 964             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
 965             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
 966
 967             fjx0             = _mm_setzero_pd();
 968             fjy0             = _mm_setzero_pd();
 969             fjz0             = _mm_setzero_pd();
 970             fjx1             = _mm_setzero_pd();
 971             fjy1             = _mm_setzero_pd();
 972             fjz1             = _mm_setzero_pd();
 973             fjx2             = _mm_setzero_pd();
 974             fjy2             = _mm_setzero_pd();
 975             fjz2             = _mm_setzero_pd();
 976
 977             /**************************
 978              * CALCULATE INTERACTIONS *
 979              **************************/
 980
 981             if (gmx_mm_any_lt(rsq00,rcutoff2))
 982             {
 983
 984             r00              = _mm_mul_pd(rsq00,rinv00);
 985
 986             /* EWALD ELECTROSTATICS */
 987
 988             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 989             ewrt             = _mm_mul_pd(r00,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             ewitab           = _mm_slli_epi32(ewitab,2);
 998             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 999             ewtabD           = _mm_setzero_pd();
1000             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1001             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
1002             ewtabFn          = _mm_setzero_pd();
1003             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1004             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
1005             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
1006             velec            = _mm_mul_pd(qq00,_mm_sub_pd(rinv00,velec));
1007             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
1008
1009             /* LENNARD-JONES DISPERSION/REPULSION */
1010
1011             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1012             vvdw6            = _mm_mul_pd(c6_00,rinvsix);
1013             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
1014             vvdw             = _mm_msub_pd( vvdw12,one_twelfth, _mm_mul_pd(vvdw6,one_sixth) );
1015             fvdw             = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
1016
1017             d                = _mm_sub_pd(r00,rswitch);
1018             d                = _mm_max_pd(d,_mm_setzero_pd());
1019             d2               = _mm_mul_pd(d,d);
1020             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
1021
1022             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
1023
1024             /* Evaluate switch function */
1025             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1026             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv00,_mm_mul_pd(velec,dsw)) );
1027             fvdw             = _mm_msub_pd( fvdw,sw , _mm_mul_pd(rinv00,_mm_mul_pd(vvdw,dsw)) );
1028             velec            = _mm_mul_pd(velec,sw);
1029             vvdw             = _mm_mul_pd(vvdw,sw);
1030             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
1031
1032             /* Update potential sum for this i atom from the interaction with this j atom. */
1033             velec            = _mm_and_pd(velec,cutoff_mask);
1034             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1035             velecsum         = _mm_add_pd(velecsum,velec);
1036             vvdw             = _mm_and_pd(vvdw,cutoff_mask);
1037             vvdw             = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
1038             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
1039
1040             fscal            = _mm_add_pd(felec,fvdw);
1041
1042             fscal            = _mm_and_pd(fscal,cutoff_mask);
1043
1044             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1045
1046             /* Update vectorial force */
1047             fix0             = _mm_macc_pd(dx00,fscal,fix0);
1048             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
1049             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
1050
1051             fjx0             = _mm_macc_pd(dx00,fscal,fjx0);
1052             fjy0             = _mm_macc_pd(dy00,fscal,fjy0);
1053             fjz0             = _mm_macc_pd(dz00,fscal,fjz0);
1054
1055             }
1056
1057             /**************************
1058              * CALCULATE INTERACTIONS *
1059              **************************/
1060
1061             if (gmx_mm_any_lt(rsq01,rcutoff2))
1062             {
1063
1064             r01              = _mm_mul_pd(rsq01,rinv01);
1065
1066             /* EWALD ELECTROSTATICS */
1067
1068             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1069             ewrt             = _mm_mul_pd(r01,ewtabscale);
1070             ewitab           = _mm_cvttpd_epi32(ewrt);
1071 #ifdef __XOP__
1072             eweps            = _mm_frcz_pd(ewrt);
1073 #else
1074             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1075 #endif
1076             twoeweps         = _mm_add_pd(eweps,eweps);
1077             ewitab           = _mm_slli_epi32(ewitab,2);
1078             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1079             ewtabD           = _mm_setzero_pd();
1080             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1081             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
1082             ewtabFn          = _mm_setzero_pd();
1083             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1084             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
1085             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
1086             velec            = _mm_mul_pd(qq01,_mm_sub_pd(rinv01,velec));
1087             felec            = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
1088
1089             d                = _mm_sub_pd(r01,rswitch);
1090             d                = _mm_max_pd(d,_mm_setzero_pd());
1091             d2               = _mm_mul_pd(d,d);
1092             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
1093
1094             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
1095
1096             /* Evaluate switch function */
1097             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1098             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv01,_mm_mul_pd(velec,dsw)) );
1099             velec            = _mm_mul_pd(velec,sw);
1100             cutoff_mask      = _mm_cmplt_pd(rsq01,rcutoff2);
1101
1102             /* Update potential sum for this i atom from the interaction with this j atom. */
1103             velec            = _mm_and_pd(velec,cutoff_mask);
1104             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1105             velecsum         = _mm_add_pd(velecsum,velec);
1106
1107             fscal            = felec;
1108
1109             fscal            = _mm_and_pd(fscal,cutoff_mask);
1110
1111             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1112
1113             /* Update vectorial force */
1114             fix0             = _mm_macc_pd(dx01,fscal,fix0);
1115             fiy0             = _mm_macc_pd(dy01,fscal,fiy0);
1116             fiz0             = _mm_macc_pd(dz01,fscal,fiz0);
1117
1118             fjx1             = _mm_macc_pd(dx01,fscal,fjx1);
1119             fjy1             = _mm_macc_pd(dy01,fscal,fjy1);
1120             fjz1             = _mm_macc_pd(dz01,fscal,fjz1);
1121
1122             }
1123
1124             /**************************
1125              * CALCULATE INTERACTIONS *
1126              **************************/
1127
1128             if (gmx_mm_any_lt(rsq02,rcutoff2))
1129             {
1130
1131             r02              = _mm_mul_pd(rsq02,rinv02);
1132
1133             /* EWALD ELECTROSTATICS */
1134
1135             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1136             ewrt             = _mm_mul_pd(r02,ewtabscale);
1137             ewitab           = _mm_cvttpd_epi32(ewrt);
1138 #ifdef __XOP__
1139             eweps            = _mm_frcz_pd(ewrt);
1140 #else
1141             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1142 #endif
1143             twoeweps         = _mm_add_pd(eweps,eweps);
1144             ewitab           = _mm_slli_epi32(ewitab,2);
1145             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1146             ewtabD           = _mm_setzero_pd();
1147             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1148             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
1149             ewtabFn          = _mm_setzero_pd();
1150             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1151             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
1152             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
1153             velec            = _mm_mul_pd(qq02,_mm_sub_pd(rinv02,velec));
1154             felec            = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
1155
1156             d                = _mm_sub_pd(r02,rswitch);
1157             d                = _mm_max_pd(d,_mm_setzero_pd());
1158             d2               = _mm_mul_pd(d,d);
1159             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
1160
1161             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
1162
1163             /* Evaluate switch function */
1164             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1165             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv02,_mm_mul_pd(velec,dsw)) );
1166             velec            = _mm_mul_pd(velec,sw);
1167             cutoff_mask      = _mm_cmplt_pd(rsq02,rcutoff2);
1168
1169             /* Update potential sum for this i atom from the interaction with this j atom. */
1170             velec            = _mm_and_pd(velec,cutoff_mask);
1171             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1172             velecsum         = _mm_add_pd(velecsum,velec);
1173
1174             fscal            = felec;
1175
1176             fscal            = _mm_and_pd(fscal,cutoff_mask);
1177
1178             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1179
1180             /* Update vectorial force */
1181             fix0             = _mm_macc_pd(dx02,fscal,fix0);
1182             fiy0             = _mm_macc_pd(dy02,fscal,fiy0);
1183             fiz0             = _mm_macc_pd(dz02,fscal,fiz0);
1184
1185             fjx2             = _mm_macc_pd(dx02,fscal,fjx2);
1186             fjy2             = _mm_macc_pd(dy02,fscal,fjy2);
1187             fjz2             = _mm_macc_pd(dz02,fscal,fjz2);
1188
1189             }
1190
1191             /**************************
1192              * CALCULATE INTERACTIONS *
1193              **************************/
1194
1195             if (gmx_mm_any_lt(rsq10,rcutoff2))
1196             {
1197
1198             r10              = _mm_mul_pd(rsq10,rinv10);
1199
1200             /* EWALD ELECTROSTATICS */
1201
1202             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1203             ewrt             = _mm_mul_pd(r10,ewtabscale);
1204             ewitab           = _mm_cvttpd_epi32(ewrt);
1205 #ifdef __XOP__
1206             eweps            = _mm_frcz_pd(ewrt);
1207 #else
1208             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1209 #endif
1210             twoeweps         = _mm_add_pd(eweps,eweps);
1211             ewitab           = _mm_slli_epi32(ewitab,2);
1212             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1213             ewtabD           = _mm_setzero_pd();
1214             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1215             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
1216             ewtabFn          = _mm_setzero_pd();
1217             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1218             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
1219             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
1220             velec            = _mm_mul_pd(qq10,_mm_sub_pd(rinv10,velec));
1221             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
1222
1223             d                = _mm_sub_pd(r10,rswitch);
1224             d                = _mm_max_pd(d,_mm_setzero_pd());
1225             d2               = _mm_mul_pd(d,d);
1226             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
1227
1228             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
1229
1230             /* Evaluate switch function */
1231             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1232             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv10,_mm_mul_pd(velec,dsw)) );
1233             velec            = _mm_mul_pd(velec,sw);
1234             cutoff_mask      = _mm_cmplt_pd(rsq10,rcutoff2);
1235
1236             /* Update potential sum for this i atom from the interaction with this j atom. */
1237             velec            = _mm_and_pd(velec,cutoff_mask);
1238             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1239             velecsum         = _mm_add_pd(velecsum,velec);
1240
1241             fscal            = felec;
1242
1243             fscal            = _mm_and_pd(fscal,cutoff_mask);
1244
1245             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1246
1247             /* Update vectorial force */
1248             fix1             = _mm_macc_pd(dx10,fscal,fix1);
1249             fiy1             = _mm_macc_pd(dy10,fscal,fiy1);
1250             fiz1             = _mm_macc_pd(dz10,fscal,fiz1);
1251
1252             fjx0             = _mm_macc_pd(dx10,fscal,fjx0);
1253             fjy0             = _mm_macc_pd(dy10,fscal,fjy0);
1254             fjz0             = _mm_macc_pd(dz10,fscal,fjz0);
1255
1256             }
1257
1258             /**************************
1259              * CALCULATE INTERACTIONS *
1260              **************************/
1261
1262             if (gmx_mm_any_lt(rsq11,rcutoff2))
1263             {
1264
1265             r11              = _mm_mul_pd(rsq11,rinv11);
1266
1267             /* EWALD ELECTROSTATICS */
1268
1269             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1270             ewrt             = _mm_mul_pd(r11,ewtabscale);
1271             ewitab           = _mm_cvttpd_epi32(ewrt);
1272 #ifdef __XOP__
1273             eweps            = _mm_frcz_pd(ewrt);
1274 #else
1275             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1276 #endif
1277             twoeweps         = _mm_add_pd(eweps,eweps);
1278             ewitab           = _mm_slli_epi32(ewitab,2);
1279             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1280             ewtabD           = _mm_setzero_pd();
1281             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1282             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
1283             ewtabFn          = _mm_setzero_pd();
1284             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1285             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
1286             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
1287             velec            = _mm_mul_pd(qq11,_mm_sub_pd(rinv11,velec));
1288             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
1289
1290             d                = _mm_sub_pd(r11,rswitch);
1291             d                = _mm_max_pd(d,_mm_setzero_pd());
1292             d2               = _mm_mul_pd(d,d);
1293             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
1294
1295             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
1296
1297             /* Evaluate switch function */
1298             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1299             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv11,_mm_mul_pd(velec,dsw)) );
1300             velec            = _mm_mul_pd(velec,sw);
1301             cutoff_mask      = _mm_cmplt_pd(rsq11,rcutoff2);
1302
1303             /* Update potential sum for this i atom from the interaction with this j atom. */
1304             velec            = _mm_and_pd(velec,cutoff_mask);
1305             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1306             velecsum         = _mm_add_pd(velecsum,velec);
1307
1308             fscal            = felec;
1309
1310             fscal            = _mm_and_pd(fscal,cutoff_mask);
1311
1312             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1313
1314             /* Update vectorial force */
1315             fix1             = _mm_macc_pd(dx11,fscal,fix1);
1316             fiy1             = _mm_macc_pd(dy11,fscal,fiy1);
1317             fiz1             = _mm_macc_pd(dz11,fscal,fiz1);
1318
1319             fjx1             = _mm_macc_pd(dx11,fscal,fjx1);
1320             fjy1             = _mm_macc_pd(dy11,fscal,fjy1);
1321             fjz1             = _mm_macc_pd(dz11,fscal,fjz1);
1322
1323             }
1324
1325             /**************************
1326              * CALCULATE INTERACTIONS *
1327              **************************/
1328
1329             if (gmx_mm_any_lt(rsq12,rcutoff2))
1330             {
1331
1332             r12              = _mm_mul_pd(rsq12,rinv12);
1333
1334             /* EWALD ELECTROSTATICS */
1335
1336             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1337             ewrt             = _mm_mul_pd(r12,ewtabscale);
1338             ewitab           = _mm_cvttpd_epi32(ewrt);
1339 #ifdef __XOP__
1340             eweps            = _mm_frcz_pd(ewrt);
1341 #else
1342             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1343 #endif
1344             twoeweps         = _mm_add_pd(eweps,eweps);
1345             ewitab           = _mm_slli_epi32(ewitab,2);
1346             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1347             ewtabD           = _mm_setzero_pd();
1348             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1349             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
1350             ewtabFn          = _mm_setzero_pd();
1351             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1352             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
1353             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
1354             velec            = _mm_mul_pd(qq12,_mm_sub_pd(rinv12,velec));
1355             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
1356
1357             d                = _mm_sub_pd(r12,rswitch);
1358             d                = _mm_max_pd(d,_mm_setzero_pd());
1359             d2               = _mm_mul_pd(d,d);
1360             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
1361
1362             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
1363
1364             /* Evaluate switch function */
1365             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1366             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv12,_mm_mul_pd(velec,dsw)) );
1367             velec            = _mm_mul_pd(velec,sw);
1368             cutoff_mask      = _mm_cmplt_pd(rsq12,rcutoff2);
1369
1370             /* Update potential sum for this i atom from the interaction with this j atom. */
1371             velec            = _mm_and_pd(velec,cutoff_mask);
1372             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1373             velecsum         = _mm_add_pd(velecsum,velec);
1374
1375             fscal            = felec;
1376
1377             fscal            = _mm_and_pd(fscal,cutoff_mask);
1378
1379             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1380
1381             /* Update vectorial force */
1382             fix1             = _mm_macc_pd(dx12,fscal,fix1);
1383             fiy1             = _mm_macc_pd(dy12,fscal,fiy1);
1384             fiz1             = _mm_macc_pd(dz12,fscal,fiz1);
1385
1386             fjx2             = _mm_macc_pd(dx12,fscal,fjx2);
1387             fjy2             = _mm_macc_pd(dy12,fscal,fjy2);
1388             fjz2             = _mm_macc_pd(dz12,fscal,fjz2);
1389
1390             }
1391
1392             /**************************
1393              * CALCULATE INTERACTIONS *
1394              **************************/
1395
1396             if (gmx_mm_any_lt(rsq20,rcutoff2))
1397             {
1398
1399             r20              = _mm_mul_pd(rsq20,rinv20);
1400
1401             /* EWALD ELECTROSTATICS */
1402
1403             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1404             ewrt             = _mm_mul_pd(r20,ewtabscale);
1405             ewitab           = _mm_cvttpd_epi32(ewrt);
1406 #ifdef __XOP__
1407             eweps            = _mm_frcz_pd(ewrt);
1408 #else
1409             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1410 #endif
1411             twoeweps         = _mm_add_pd(eweps,eweps);
1412             ewitab           = _mm_slli_epi32(ewitab,2);
1413             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1414             ewtabD           = _mm_setzero_pd();
1415             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1416             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
1417             ewtabFn          = _mm_setzero_pd();
1418             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1419             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
1420             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
1421             velec            = _mm_mul_pd(qq20,_mm_sub_pd(rinv20,velec));
1422             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
1423
1424             d                = _mm_sub_pd(r20,rswitch);
1425             d                = _mm_max_pd(d,_mm_setzero_pd());
1426             d2               = _mm_mul_pd(d,d);
1427             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
1428
1429             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
1430
1431             /* Evaluate switch function */
1432             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1433             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv20,_mm_mul_pd(velec,dsw)) );
1434             velec            = _mm_mul_pd(velec,sw);
1435             cutoff_mask      = _mm_cmplt_pd(rsq20,rcutoff2);
1436
1437             /* Update potential sum for this i atom from the interaction with this j atom. */
1438             velec            = _mm_and_pd(velec,cutoff_mask);
1439             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1440             velecsum         = _mm_add_pd(velecsum,velec);
1441
1442             fscal            = felec;
1443
1444             fscal            = _mm_and_pd(fscal,cutoff_mask);
1445
1446             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1447
1448             /* Update vectorial force */
1449             fix2             = _mm_macc_pd(dx20,fscal,fix2);
1450             fiy2             = _mm_macc_pd(dy20,fscal,fiy2);
1451             fiz2             = _mm_macc_pd(dz20,fscal,fiz2);
1452
1453             fjx0             = _mm_macc_pd(dx20,fscal,fjx0);
1454             fjy0             = _mm_macc_pd(dy20,fscal,fjy0);
1455             fjz0             = _mm_macc_pd(dz20,fscal,fjz0);
1456
1457             }
1458
1459             /**************************
1460              * CALCULATE INTERACTIONS *
1461              **************************/
1462
1463             if (gmx_mm_any_lt(rsq21,rcutoff2))
1464             {
1465
1466             r21              = _mm_mul_pd(rsq21,rinv21);
1467
1468             /* EWALD ELECTROSTATICS */
1469
1470             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1471             ewrt             = _mm_mul_pd(r21,ewtabscale);
1472             ewitab           = _mm_cvttpd_epi32(ewrt);
1473 #ifdef __XOP__
1474             eweps            = _mm_frcz_pd(ewrt);
1475 #else
1476             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1477 #endif
1478             twoeweps         = _mm_add_pd(eweps,eweps);
1479             ewitab           = _mm_slli_epi32(ewitab,2);
1480             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1481             ewtabD           = _mm_setzero_pd();
1482             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1483             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
1484             ewtabFn          = _mm_setzero_pd();
1485             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1486             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
1487             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
1488             velec            = _mm_mul_pd(qq21,_mm_sub_pd(rinv21,velec));
1489             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
1490
1491             d                = _mm_sub_pd(r21,rswitch);
1492             d                = _mm_max_pd(d,_mm_setzero_pd());
1493             d2               = _mm_mul_pd(d,d);
1494             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
1495
1496             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
1497
1498             /* Evaluate switch function */
1499             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1500             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv21,_mm_mul_pd(velec,dsw)) );
1501             velec            = _mm_mul_pd(velec,sw);
1502             cutoff_mask      = _mm_cmplt_pd(rsq21,rcutoff2);
1503
1504             /* Update potential sum for this i atom from the interaction with this j atom. */
1505             velec            = _mm_and_pd(velec,cutoff_mask);
1506             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1507             velecsum         = _mm_add_pd(velecsum,velec);
1508
1509             fscal            = felec;
1510
1511             fscal            = _mm_and_pd(fscal,cutoff_mask);
1512
1513             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1514
1515             /* Update vectorial force */
1516             fix2             = _mm_macc_pd(dx21,fscal,fix2);
1517             fiy2             = _mm_macc_pd(dy21,fscal,fiy2);
1518             fiz2             = _mm_macc_pd(dz21,fscal,fiz2);
1519
1520             fjx1             = _mm_macc_pd(dx21,fscal,fjx1);
1521             fjy1             = _mm_macc_pd(dy21,fscal,fjy1);
1522             fjz1             = _mm_macc_pd(dz21,fscal,fjz1);
1523
1524             }
1525
1526             /**************************
1527              * CALCULATE INTERACTIONS *
1528              **************************/
1529
1530             if (gmx_mm_any_lt(rsq22,rcutoff2))
1531             {
1532
1533             r22              = _mm_mul_pd(rsq22,rinv22);
1534
1535             /* EWALD ELECTROSTATICS */
1536
1537             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1538             ewrt             = _mm_mul_pd(r22,ewtabscale);
1539             ewitab           = _mm_cvttpd_epi32(ewrt);
1540 #ifdef __XOP__
1541             eweps            = _mm_frcz_pd(ewrt);
1542 #else
1543             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1544 #endif
1545             twoeweps         = _mm_add_pd(eweps,eweps);
1546             ewitab           = _mm_slli_epi32(ewitab,2);
1547             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1548             ewtabD           = _mm_setzero_pd();
1549             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1550             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
1551             ewtabFn          = _mm_setzero_pd();
1552             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1553             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
1554             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
1555             velec            = _mm_mul_pd(qq22,_mm_sub_pd(rinv22,velec));
1556             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
1557
1558             d                = _mm_sub_pd(r22,rswitch);
1559             d                = _mm_max_pd(d,_mm_setzero_pd());
1560             d2               = _mm_mul_pd(d,d);
1561             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
1562
1563             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
1564
1565             /* Evaluate switch function */
1566             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1567             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv22,_mm_mul_pd(velec,dsw)) );
1568             velec            = _mm_mul_pd(velec,sw);
1569             cutoff_mask      = _mm_cmplt_pd(rsq22,rcutoff2);
1570
1571             /* Update potential sum for this i atom from the interaction with this j atom. */
1572             velec            = _mm_and_pd(velec,cutoff_mask);
1573             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1574             velecsum         = _mm_add_pd(velecsum,velec);
1575
1576             fscal            = felec;
1577
1578             fscal            = _mm_and_pd(fscal,cutoff_mask);
1579
1580             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1581
1582             /* Update vectorial force */
1583             fix2             = _mm_macc_pd(dx22,fscal,fix2);
1584             fiy2             = _mm_macc_pd(dy22,fscal,fiy2);
1585             fiz2             = _mm_macc_pd(dz22,fscal,fiz2);
1586
1587             fjx2             = _mm_macc_pd(dx22,fscal,fjx2);
1588             fjy2             = _mm_macc_pd(dy22,fscal,fjy2);
1589             fjz2             = _mm_macc_pd(dz22,fscal,fjz2);
1590
1591             }
1592
1593             gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1594
1595             /* Inner loop uses 630 flops */
1596         }
1597
1598         /* End of innermost loop */
1599
1600         gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1601                                               f+i_coord_offset,fshift+i_shift_offset);
1602
1603         ggid                        = gid[iidx];
1604         /* Update potential energies */
1605         gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1606         gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1607
1608         /* Increment number of inner iterations */
1609         inneriter                  += j_index_end - j_index_start;
1610
1611         /* Outer loop uses 20 flops */
1612     }
1613
1614     /* Increment number of outer iterations */
1615     outeriter        += nri;
1616
1617     /* Update outer/inner flops */
1618
1619     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_VF,outeriter*20 + inneriter*630);
1620 }
1621 /*
1622  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSw_VdwLJSw_GeomW3W3_F_avx_128_fma_double
1623  * Electrostatics interaction: Ewald
1624  * VdW interaction:            LennardJones
1625  * Geometry:                   Water3-Water3
1626  * Calculate force/pot:        Force
1627  */
1628 void
1629 nb_kernel_ElecEwSw_VdwLJSw_GeomW3W3_F_avx_128_fma_double
1630                     (t_nblist                    * gmx_restrict       nlist,
1631                      rvec                        * gmx_restrict          xx,
1632                      rvec                        * gmx_restrict          ff,
1633                      t_forcerec                  * gmx_restrict          fr,
1634                      t_mdatoms                   * gmx_restrict     mdatoms,
1635                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1636                      t_nrnb                      * gmx_restrict        nrnb)
1637 {
1638     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1639      * just 0 for non-waters.
1640      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
1641      * jnr indices corresponding to data put in the four positions in the SIMD register.
1642      */
1643     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1644     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1645     int              jnrA,jnrB;
1646     int              j_coord_offsetA,j_coord_offsetB;
1647     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1648     real             rcutoff_scalar;
1649     real             *shiftvec,*fshift,*x,*f;
1650     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1651     int              vdwioffset0;
1652     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1653     int              vdwioffset1;
1654     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1655     int              vdwioffset2;
1656     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1657     int              vdwjidx0A,vdwjidx0B;
1658     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1659     int              vdwjidx1A,vdwjidx1B;
1660     __m128d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1661     int              vdwjidx2A,vdwjidx2B;
1662     __m128d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1663     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1664     __m128d          dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1665     __m128d          dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1666     __m128d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1667     __m128d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1668     __m128d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1669     __m128d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1670     __m128d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1671     __m128d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1672     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
1673     real             *charge;
1674     int              nvdwtype;
1675     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1676     int              *vdwtype;
1677     real             *vdwparam;
1678     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
1679     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
1680     __m128i          ewitab;
1681     __m128d          ewtabscale,eweps,twoeweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1682     real             *ewtab;
1683     __m128d          rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
1684     real             rswitch_scalar,d_scalar;
1685     __m128d          dummy_mask,cutoff_mask;
1686     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
1687     __m128d          one     = _mm_set1_pd(1.0);
1688     __m128d          two     = _mm_set1_pd(2.0);
1689     x                = xx[0];
1690     f                = ff[0];
1691
1692     nri              = nlist->nri;
1693     iinr             = nlist->iinr;
1694     jindex           = nlist->jindex;
1695     jjnr             = nlist->jjnr;
1696     shiftidx         = nlist->shift;
1697     gid              = nlist->gid;
1698     shiftvec         = fr->shift_vec[0];
1699     fshift           = fr->fshift[0];
1700     facel            = _mm_set1_pd(fr->epsfac);
1701     charge           = mdatoms->chargeA;
1702     nvdwtype         = fr->ntype;
1703     vdwparam         = fr->nbfp;
1704     vdwtype          = mdatoms->typeA;
1705
1706     sh_ewald         = _mm_set1_pd(fr->ic->sh_ewald);
1707     ewtab            = fr->ic->tabq_coul_FDV0;
1708     ewtabscale       = _mm_set1_pd(fr->ic->tabq_scale);
1709     ewtabhalfspace   = _mm_set1_pd(0.5/fr->ic->tabq_scale);
1710
1711     /* Setup water-specific parameters */
1712     inr              = nlist->iinr[0];
1713     iq0              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
1714     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
1715     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
1716     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
1717
1718     jq0              = _mm_set1_pd(charge[inr+0]);
1719     jq1              = _mm_set1_pd(charge[inr+1]);
1720     jq2              = _mm_set1_pd(charge[inr+2]);
1721     vdwjidx0A        = 2*vdwtype[inr+0];
1722     qq00             = _mm_mul_pd(iq0,jq0);
1723     c6_00            = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
1724     c12_00           = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
1725     qq01             = _mm_mul_pd(iq0,jq1);
1726     qq02             = _mm_mul_pd(iq0,jq2);
1727     qq10             = _mm_mul_pd(iq1,jq0);
1728     qq11             = _mm_mul_pd(iq1,jq1);
1729     qq12             = _mm_mul_pd(iq1,jq2);
1730     qq20             = _mm_mul_pd(iq2,jq0);
1731     qq21             = _mm_mul_pd(iq2,jq1);
1732     qq22             = _mm_mul_pd(iq2,jq2);
1733
1734     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
1735     rcutoff_scalar   = fr->rcoulomb;
1736     rcutoff          = _mm_set1_pd(rcutoff_scalar);
1737     rcutoff2         = _mm_mul_pd(rcutoff,rcutoff);
1738
1739     rswitch_scalar   = fr->rcoulomb_switch;
1740     rswitch          = _mm_set1_pd(rswitch_scalar);
1741     /* Setup switch parameters */
1742     d_scalar         = rcutoff_scalar-rswitch_scalar;
1743     d                = _mm_set1_pd(d_scalar);
1744     swV3             = _mm_set1_pd(-10.0/(d_scalar*d_scalar*d_scalar));
1745     swV4             = _mm_set1_pd( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
1746     swV5             = _mm_set1_pd( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
1747     swF2             = _mm_set1_pd(-30.0/(d_scalar*d_scalar*d_scalar));
1748     swF3             = _mm_set1_pd( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
1749     swF4             = _mm_set1_pd(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
1750
1751     /* Avoid stupid compiler warnings */
1752     jnrA = jnrB = 0;
1753     j_coord_offsetA = 0;
1754     j_coord_offsetB = 0;
1755
1756     outeriter        = 0;
1757     inneriter        = 0;
1758
1759     /* Start outer loop over neighborlists */
1760     for(iidx=0; iidx<nri; iidx++)
1761     {
1762         /* Load shift vector for this list */
1763         i_shift_offset   = DIM*shiftidx[iidx];
1764
1765         /* Load limits for loop over neighbors */
1766         j_index_start    = jindex[iidx];
1767         j_index_end      = jindex[iidx+1];
1768
1769         /* Get outer coordinate index */
1770         inr              = iinr[iidx];
1771         i_coord_offset   = DIM*inr;
1772
1773         /* Load i particle coords and add shift vector */
1774         gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1775                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1776
1777         fix0             = _mm_setzero_pd();
1778         fiy0             = _mm_setzero_pd();
1779         fiz0             = _mm_setzero_pd();
1780         fix1             = _mm_setzero_pd();
1781         fiy1             = _mm_setzero_pd();
1782         fiz1             = _mm_setzero_pd();
1783         fix2             = _mm_setzero_pd();
1784         fiy2             = _mm_setzero_pd();
1785         fiz2             = _mm_setzero_pd();
1786
1787         /* Start inner kernel loop */
1788         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
1789         {
1790
1791             /* Get j neighbor index, and coordinate index */
1792             jnrA             = jjnr[jidx];
1793             jnrB             = jjnr[jidx+1];
1794             j_coord_offsetA  = DIM*jnrA;
1795             j_coord_offsetB  = DIM*jnrB;
1796
1797             /* load j atom coordinates */
1798             gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1799                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1800
1801             /* Calculate displacement vector */
1802             dx00             = _mm_sub_pd(ix0,jx0);
1803             dy00             = _mm_sub_pd(iy0,jy0);
1804             dz00             = _mm_sub_pd(iz0,jz0);
1805             dx01             = _mm_sub_pd(ix0,jx1);
1806             dy01             = _mm_sub_pd(iy0,jy1);
1807             dz01             = _mm_sub_pd(iz0,jz1);
1808             dx02             = _mm_sub_pd(ix0,jx2);
1809             dy02             = _mm_sub_pd(iy0,jy2);
1810             dz02             = _mm_sub_pd(iz0,jz2);
1811             dx10             = _mm_sub_pd(ix1,jx0);
1812             dy10             = _mm_sub_pd(iy1,jy0);
1813             dz10             = _mm_sub_pd(iz1,jz0);
1814             dx11             = _mm_sub_pd(ix1,jx1);
1815             dy11             = _mm_sub_pd(iy1,jy1);
1816             dz11             = _mm_sub_pd(iz1,jz1);
1817             dx12             = _mm_sub_pd(ix1,jx2);
1818             dy12             = _mm_sub_pd(iy1,jy2);
1819             dz12             = _mm_sub_pd(iz1,jz2);
1820             dx20             = _mm_sub_pd(ix2,jx0);
1821             dy20             = _mm_sub_pd(iy2,jy0);
1822             dz20             = _mm_sub_pd(iz2,jz0);
1823             dx21             = _mm_sub_pd(ix2,jx1);
1824             dy21             = _mm_sub_pd(iy2,jy1);
1825             dz21             = _mm_sub_pd(iz2,jz1);
1826             dx22             = _mm_sub_pd(ix2,jx2);
1827             dy22             = _mm_sub_pd(iy2,jy2);
1828             dz22             = _mm_sub_pd(iz2,jz2);
1829
1830             /* Calculate squared distance and things based on it */
1831             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1832             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
1833             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
1834             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
1835             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1836             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1837             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
1838             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1839             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1840
1841             rinv00           = gmx_mm_invsqrt_pd(rsq00);
1842             rinv01           = gmx_mm_invsqrt_pd(rsq01);
1843             rinv02           = gmx_mm_invsqrt_pd(rsq02);
1844             rinv10           = gmx_mm_invsqrt_pd(rsq10);
1845             rinv11           = gmx_mm_invsqrt_pd(rsq11);
1846             rinv12           = gmx_mm_invsqrt_pd(rsq12);
1847             rinv20           = gmx_mm_invsqrt_pd(rsq20);
1848             rinv21           = gmx_mm_invsqrt_pd(rsq21);
1849             rinv22           = gmx_mm_invsqrt_pd(rsq22);
1850
1851             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
1852             rinvsq01         = _mm_mul_pd(rinv01,rinv01);
1853             rinvsq02         = _mm_mul_pd(rinv02,rinv02);
1854             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
1855             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
1856             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
1857             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
1858             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
1859             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
1860
1861             fjx0             = _mm_setzero_pd();
1862             fjy0             = _mm_setzero_pd();
1863             fjz0             = _mm_setzero_pd();
1864             fjx1             = _mm_setzero_pd();
1865             fjy1             = _mm_setzero_pd();
1866             fjz1             = _mm_setzero_pd();
1867             fjx2             = _mm_setzero_pd();
1868             fjy2             = _mm_setzero_pd();
1869             fjz2             = _mm_setzero_pd();
1870
1871             /**************************
1872              * CALCULATE INTERACTIONS *
1873              **************************/
1874
1875             if (gmx_mm_any_lt(rsq00,rcutoff2))
1876             {
1877
1878             r00              = _mm_mul_pd(rsq00,rinv00);
1879
1880             /* EWALD ELECTROSTATICS */
1881
1882             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1883             ewrt             = _mm_mul_pd(r00,ewtabscale);
1884             ewitab           = _mm_cvttpd_epi32(ewrt);
1885 #ifdef __XOP__
1886             eweps            = _mm_frcz_pd(ewrt);
1887 #else
1888             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1889 #endif
1890             twoeweps         = _mm_add_pd(eweps,eweps);
1891             ewitab           = _mm_slli_epi32(ewitab,2);
1892             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1893             ewtabD           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1894             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1895             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
1896             ewtabFn          = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
1897             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1898             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
1899             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
1900             velec            = _mm_mul_pd(qq00,_mm_sub_pd(rinv00,velec));
1901             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
1902
1903             /* LENNARD-JONES DISPERSION/REPULSION */
1904
1905             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1906             vvdw6            = _mm_mul_pd(c6_00,rinvsix);
1907             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
1908             vvdw             = _mm_msub_pd( vvdw12,one_twelfth, _mm_mul_pd(vvdw6,one_sixth) );
1909             fvdw             = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
1910
1911             d                = _mm_sub_pd(r00,rswitch);
1912             d                = _mm_max_pd(d,_mm_setzero_pd());
1913             d2               = _mm_mul_pd(d,d);
1914             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
1915
1916             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
1917
1918             /* Evaluate switch function */
1919             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1920             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv00,_mm_mul_pd(velec,dsw)) );
1921             fvdw             = _mm_msub_pd( fvdw,sw , _mm_mul_pd(rinv00,_mm_mul_pd(vvdw,dsw)) );
1922             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
1923
1924             fscal            = _mm_add_pd(felec,fvdw);
1925
1926             fscal            = _mm_and_pd(fscal,cutoff_mask);
1927
1928             /* Update vectorial force */
1929             fix0             = _mm_macc_pd(dx00,fscal,fix0);
1930             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
1931             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
1932
1933             fjx0             = _mm_macc_pd(dx00,fscal,fjx0);
1934             fjy0             = _mm_macc_pd(dy00,fscal,fjy0);
1935             fjz0             = _mm_macc_pd(dz00,fscal,fjz0);
1936
1937             }
1938
1939             /**************************
1940              * CALCULATE INTERACTIONS *
1941              **************************/
1942
1943             if (gmx_mm_any_lt(rsq01,rcutoff2))
1944             {
1945
1946             r01              = _mm_mul_pd(rsq01,rinv01);
1947
1948             /* EWALD ELECTROSTATICS */
1949
1950             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1951             ewrt             = _mm_mul_pd(r01,ewtabscale);
1952             ewitab           = _mm_cvttpd_epi32(ewrt);
1953 #ifdef __XOP__
1954             eweps            = _mm_frcz_pd(ewrt);
1955 #else
1956             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1957 #endif
1958             twoeweps         = _mm_add_pd(eweps,eweps);
1959             ewitab           = _mm_slli_epi32(ewitab,2);
1960             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1961             ewtabD           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1962             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1963             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
1964             ewtabFn          = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
1965             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1966             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
1967             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
1968             velec            = _mm_mul_pd(qq01,_mm_sub_pd(rinv01,velec));
1969             felec            = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
1970
1971             d                = _mm_sub_pd(r01,rswitch);
1972             d                = _mm_max_pd(d,_mm_setzero_pd());
1973             d2               = _mm_mul_pd(d,d);
1974             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
1975
1976             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
1977
1978             /* Evaluate switch function */
1979             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1980             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv01,_mm_mul_pd(velec,dsw)) );
1981             cutoff_mask      = _mm_cmplt_pd(rsq01,rcutoff2);
1982
1983             fscal            = felec;
1984
1985             fscal            = _mm_and_pd(fscal,cutoff_mask);
1986
1987             /* Update vectorial force */
1988             fix0             = _mm_macc_pd(dx01,fscal,fix0);
1989             fiy0             = _mm_macc_pd(dy01,fscal,fiy0);
1990             fiz0             = _mm_macc_pd(dz01,fscal,fiz0);
1991
1992             fjx1             = _mm_macc_pd(dx01,fscal,fjx1);
1993             fjy1             = _mm_macc_pd(dy01,fscal,fjy1);
1994             fjz1             = _mm_macc_pd(dz01,fscal,fjz1);
1995
1996             }
1997
1998             /**************************
1999              * CALCULATE INTERACTIONS *
2000              **************************/
2001
2002             if (gmx_mm_any_lt(rsq02,rcutoff2))
2003             {
2004
2005             r02              = _mm_mul_pd(rsq02,rinv02);
2006
2007             /* EWALD ELECTROSTATICS */
2008
2009             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2010             ewrt             = _mm_mul_pd(r02,ewtabscale);
2011             ewitab           = _mm_cvttpd_epi32(ewrt);
2012 #ifdef __XOP__
2013             eweps            = _mm_frcz_pd(ewrt);
2014 #else
2015             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2016 #endif
2017             twoeweps         = _mm_add_pd(eweps,eweps);
2018             ewitab           = _mm_slli_epi32(ewitab,2);
2019             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
2020             ewtabD           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
2021             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
2022             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
2023             ewtabFn          = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
2024             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
2025             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
2026             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
2027             velec            = _mm_mul_pd(qq02,_mm_sub_pd(rinv02,velec));
2028             felec            = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
2029
2030             d                = _mm_sub_pd(r02,rswitch);
2031             d                = _mm_max_pd(d,_mm_setzero_pd());
2032             d2               = _mm_mul_pd(d,d);
2033             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
2034
2035             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
2036
2037             /* Evaluate switch function */
2038             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2039             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv02,_mm_mul_pd(velec,dsw)) );
2040             cutoff_mask      = _mm_cmplt_pd(rsq02,rcutoff2);
2041
2042             fscal            = felec;
2043
2044             fscal            = _mm_and_pd(fscal,cutoff_mask);
2045
2046             /* Update vectorial force */
2047             fix0             = _mm_macc_pd(dx02,fscal,fix0);
2048             fiy0             = _mm_macc_pd(dy02,fscal,fiy0);
2049             fiz0             = _mm_macc_pd(dz02,fscal,fiz0);
2050
2051             fjx2             = _mm_macc_pd(dx02,fscal,fjx2);
2052             fjy2             = _mm_macc_pd(dy02,fscal,fjy2);
2053             fjz2             = _mm_macc_pd(dz02,fscal,fjz2);
2054
2055             }
2056
2057             /**************************
2058              * CALCULATE INTERACTIONS *
2059              **************************/
2060
2061             if (gmx_mm_any_lt(rsq10,rcutoff2))
2062             {
2063
2064             r10              = _mm_mul_pd(rsq10,rinv10);
2065
2066             /* EWALD ELECTROSTATICS */
2067
2068             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2069             ewrt             = _mm_mul_pd(r10,ewtabscale);
2070             ewitab           = _mm_cvttpd_epi32(ewrt);
2071 #ifdef __XOP__
2072             eweps            = _mm_frcz_pd(ewrt);
2073 #else
2074             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2075 #endif
2076             twoeweps         = _mm_add_pd(eweps,eweps);
2077             ewitab           = _mm_slli_epi32(ewitab,2);
2078             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
2079             ewtabD           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
2080             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
2081             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
2082             ewtabFn          = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
2083             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
2084             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
2085             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
2086             velec            = _mm_mul_pd(qq10,_mm_sub_pd(rinv10,velec));
2087             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
2088
2089             d                = _mm_sub_pd(r10,rswitch);
2090             d                = _mm_max_pd(d,_mm_setzero_pd());
2091             d2               = _mm_mul_pd(d,d);
2092             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
2093
2094             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
2095
2096             /* Evaluate switch function */
2097             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2098             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv10,_mm_mul_pd(velec,dsw)) );
2099             cutoff_mask      = _mm_cmplt_pd(rsq10,rcutoff2);
2100
2101             fscal            = felec;
2102
2103             fscal            = _mm_and_pd(fscal,cutoff_mask);
2104
2105             /* Update vectorial force */
2106             fix1             = _mm_macc_pd(dx10,fscal,fix1);
2107             fiy1             = _mm_macc_pd(dy10,fscal,fiy1);
2108             fiz1             = _mm_macc_pd(dz10,fscal,fiz1);
2109
2110             fjx0             = _mm_macc_pd(dx10,fscal,fjx0);
2111             fjy0             = _mm_macc_pd(dy10,fscal,fjy0);
2112             fjz0             = _mm_macc_pd(dz10,fscal,fjz0);
2113
2114             }
2115
2116             /**************************
2117              * CALCULATE INTERACTIONS *
2118              **************************/
2119
2120             if (gmx_mm_any_lt(rsq11,rcutoff2))
2121             {
2122
2123             r11              = _mm_mul_pd(rsq11,rinv11);
2124
2125             /* EWALD ELECTROSTATICS */
2126
2127             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2128             ewrt             = _mm_mul_pd(r11,ewtabscale);
2129             ewitab           = _mm_cvttpd_epi32(ewrt);
2130 #ifdef __XOP__
2131             eweps            = _mm_frcz_pd(ewrt);
2132 #else
2133             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2134 #endif
2135             twoeweps         = _mm_add_pd(eweps,eweps);
2136             ewitab           = _mm_slli_epi32(ewitab,2);
2137             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
2138             ewtabD           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
2139             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
2140             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
2141             ewtabFn          = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
2142             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
2143             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
2144             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
2145             velec            = _mm_mul_pd(qq11,_mm_sub_pd(rinv11,velec));
2146             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
2147
2148             d                = _mm_sub_pd(r11,rswitch);
2149             d                = _mm_max_pd(d,_mm_setzero_pd());
2150             d2               = _mm_mul_pd(d,d);
2151             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
2152
2153             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
2154
2155             /* Evaluate switch function */
2156             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2157             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv11,_mm_mul_pd(velec,dsw)) );
2158             cutoff_mask      = _mm_cmplt_pd(rsq11,rcutoff2);
2159
2160             fscal            = felec;
2161
2162             fscal            = _mm_and_pd(fscal,cutoff_mask);
2163
2164             /* Update vectorial force */
2165             fix1             = _mm_macc_pd(dx11,fscal,fix1);
2166             fiy1             = _mm_macc_pd(dy11,fscal,fiy1);
2167             fiz1             = _mm_macc_pd(dz11,fscal,fiz1);
2168
2169             fjx1             = _mm_macc_pd(dx11,fscal,fjx1);
2170             fjy1             = _mm_macc_pd(dy11,fscal,fjy1);
2171             fjz1             = _mm_macc_pd(dz11,fscal,fjz1);
2172
2173             }
2174
2175             /**************************
2176              * CALCULATE INTERACTIONS *
2177              **************************/
2178
2179             if (gmx_mm_any_lt(rsq12,rcutoff2))
2180             {
2181
2182             r12              = _mm_mul_pd(rsq12,rinv12);
2183
2184             /* EWALD ELECTROSTATICS */
2185
2186             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2187             ewrt             = _mm_mul_pd(r12,ewtabscale);
2188             ewitab           = _mm_cvttpd_epi32(ewrt);
2189 #ifdef __XOP__
2190             eweps            = _mm_frcz_pd(ewrt);
2191 #else
2192             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2193 #endif
2194             twoeweps         = _mm_add_pd(eweps,eweps);
2195             ewitab           = _mm_slli_epi32(ewitab,2);
2196             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
2197             ewtabD           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
2198             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
2199             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
2200             ewtabFn          = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
2201             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
2202             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
2203             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
2204             velec            = _mm_mul_pd(qq12,_mm_sub_pd(rinv12,velec));
2205             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
2206
2207             d                = _mm_sub_pd(r12,rswitch);
2208             d                = _mm_max_pd(d,_mm_setzero_pd());
2209             d2               = _mm_mul_pd(d,d);
2210             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
2211
2212             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
2213
2214             /* Evaluate switch function */
2215             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2216             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv12,_mm_mul_pd(velec,dsw)) );
2217             cutoff_mask      = _mm_cmplt_pd(rsq12,rcutoff2);
2218
2219             fscal            = felec;
2220
2221             fscal            = _mm_and_pd(fscal,cutoff_mask);
2222
2223             /* Update vectorial force */
2224             fix1             = _mm_macc_pd(dx12,fscal,fix1);
2225             fiy1             = _mm_macc_pd(dy12,fscal,fiy1);
2226             fiz1             = _mm_macc_pd(dz12,fscal,fiz1);
2227
2228             fjx2             = _mm_macc_pd(dx12,fscal,fjx2);
2229             fjy2             = _mm_macc_pd(dy12,fscal,fjy2);
2230             fjz2             = _mm_macc_pd(dz12,fscal,fjz2);
2231
2232             }
2233
2234             /**************************
2235              * CALCULATE INTERACTIONS *
2236              **************************/
2237
2238             if (gmx_mm_any_lt(rsq20,rcutoff2))
2239             {
2240
2241             r20              = _mm_mul_pd(rsq20,rinv20);
2242
2243             /* EWALD ELECTROSTATICS */
2244
2245             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2246             ewrt             = _mm_mul_pd(r20,ewtabscale);
2247             ewitab           = _mm_cvttpd_epi32(ewrt);
2248 #ifdef __XOP__
2249             eweps            = _mm_frcz_pd(ewrt);
2250 #else
2251             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2252 #endif
2253             twoeweps         = _mm_add_pd(eweps,eweps);
2254             ewitab           = _mm_slli_epi32(ewitab,2);
2255             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
2256             ewtabD           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
2257             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
2258             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
2259             ewtabFn          = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
2260             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
2261             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
2262             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
2263             velec            = _mm_mul_pd(qq20,_mm_sub_pd(rinv20,velec));
2264             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
2265
2266             d                = _mm_sub_pd(r20,rswitch);
2267             d                = _mm_max_pd(d,_mm_setzero_pd());
2268             d2               = _mm_mul_pd(d,d);
2269             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
2270
2271             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
2272
2273             /* Evaluate switch function */
2274             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2275             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv20,_mm_mul_pd(velec,dsw)) );
2276             cutoff_mask      = _mm_cmplt_pd(rsq20,rcutoff2);
2277
2278             fscal            = felec;
2279
2280             fscal            = _mm_and_pd(fscal,cutoff_mask);
2281
2282             /* Update vectorial force */
2283             fix2             = _mm_macc_pd(dx20,fscal,fix2);
2284             fiy2             = _mm_macc_pd(dy20,fscal,fiy2);
2285             fiz2             = _mm_macc_pd(dz20,fscal,fiz2);
2286
2287             fjx0             = _mm_macc_pd(dx20,fscal,fjx0);
2288             fjy0             = _mm_macc_pd(dy20,fscal,fjy0);
2289             fjz0             = _mm_macc_pd(dz20,fscal,fjz0);
2290
2291             }
2292
2293             /**************************
2294              * CALCULATE INTERACTIONS *
2295              **************************/
2296
2297             if (gmx_mm_any_lt(rsq21,rcutoff2))
2298             {
2299
2300             r21              = _mm_mul_pd(rsq21,rinv21);
2301
2302             /* EWALD ELECTROSTATICS */
2303
2304             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2305             ewrt             = _mm_mul_pd(r21,ewtabscale);
2306             ewitab           = _mm_cvttpd_epi32(ewrt);
2307 #ifdef __XOP__
2308             eweps            = _mm_frcz_pd(ewrt);
2309 #else
2310             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2311 #endif
2312             twoeweps         = _mm_add_pd(eweps,eweps);
2313             ewitab           = _mm_slli_epi32(ewitab,2);
2314             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
2315             ewtabD           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
2316             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
2317             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
2318             ewtabFn          = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
2319             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
2320             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
2321             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
2322             velec            = _mm_mul_pd(qq21,_mm_sub_pd(rinv21,velec));
2323             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
2324
2325             d                = _mm_sub_pd(r21,rswitch);
2326             d                = _mm_max_pd(d,_mm_setzero_pd());
2327             d2               = _mm_mul_pd(d,d);
2328             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
2329
2330             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
2331
2332             /* Evaluate switch function */
2333             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2334             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv21,_mm_mul_pd(velec,dsw)) );
2335             cutoff_mask      = _mm_cmplt_pd(rsq21,rcutoff2);
2336
2337             fscal            = felec;
2338
2339             fscal            = _mm_and_pd(fscal,cutoff_mask);
2340
2341             /* Update vectorial force */
2342             fix2             = _mm_macc_pd(dx21,fscal,fix2);
2343             fiy2             = _mm_macc_pd(dy21,fscal,fiy2);
2344             fiz2             = _mm_macc_pd(dz21,fscal,fiz2);
2345
2346             fjx1             = _mm_macc_pd(dx21,fscal,fjx1);
2347             fjy1             = _mm_macc_pd(dy21,fscal,fjy1);
2348             fjz1             = _mm_macc_pd(dz21,fscal,fjz1);
2349
2350             }
2351
2352             /**************************
2353              * CALCULATE INTERACTIONS *
2354              **************************/
2355
2356             if (gmx_mm_any_lt(rsq22,rcutoff2))
2357             {
2358
2359             r22              = _mm_mul_pd(rsq22,rinv22);
2360
2361             /* EWALD ELECTROSTATICS */
2362
2363             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2364             ewrt             = _mm_mul_pd(r22,ewtabscale);
2365             ewitab           = _mm_cvttpd_epi32(ewrt);
2366 #ifdef __XOP__
2367             eweps            = _mm_frcz_pd(ewrt);
2368 #else
2369             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2370 #endif
2371             twoeweps         = _mm_add_pd(eweps,eweps);
2372             ewitab           = _mm_slli_epi32(ewitab,2);
2373             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
2374             ewtabD           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
2375             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
2376             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
2377             ewtabFn          = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
2378             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
2379             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
2380             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
2381             velec            = _mm_mul_pd(qq22,_mm_sub_pd(rinv22,velec));
2382             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
2383
2384             d                = _mm_sub_pd(r22,rswitch);
2385             d                = _mm_max_pd(d,_mm_setzero_pd());
2386             d2               = _mm_mul_pd(d,d);
2387             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
2388
2389             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
2390
2391             /* Evaluate switch function */
2392             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2393             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv22,_mm_mul_pd(velec,dsw)) );
2394             cutoff_mask      = _mm_cmplt_pd(rsq22,rcutoff2);
2395
2396             fscal            = felec;
2397
2398             fscal            = _mm_and_pd(fscal,cutoff_mask);
2399
2400             /* Update vectorial force */
2401             fix2             = _mm_macc_pd(dx22,fscal,fix2);
2402             fiy2             = _mm_macc_pd(dy22,fscal,fiy2);
2403             fiz2             = _mm_macc_pd(dz22,fscal,fiz2);
2404
2405             fjx2             = _mm_macc_pd(dx22,fscal,fjx2);
2406             fjy2             = _mm_macc_pd(dy22,fscal,fjy2);
2407             fjz2             = _mm_macc_pd(dz22,fscal,fjz2);
2408
2409             }
2410
2411             gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
2412
2413             /* Inner loop uses 600 flops */
2414         }
2415
2416         if(jidx<j_index_end)
2417         {
2418
2419             jnrA             = jjnr[jidx];
2420             j_coord_offsetA  = DIM*jnrA;
2421
2422             /* load j atom coordinates */
2423             gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
2424                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
2425
2426             /* Calculate displacement vector */
2427             dx00             = _mm_sub_pd(ix0,jx0);
2428             dy00             = _mm_sub_pd(iy0,jy0);
2429             dz00             = _mm_sub_pd(iz0,jz0);
2430             dx01             = _mm_sub_pd(ix0,jx1);
2431             dy01             = _mm_sub_pd(iy0,jy1);
2432             dz01             = _mm_sub_pd(iz0,jz1);
2433             dx02             = _mm_sub_pd(ix0,jx2);
2434             dy02             = _mm_sub_pd(iy0,jy2);
2435             dz02             = _mm_sub_pd(iz0,jz2);
2436             dx10             = _mm_sub_pd(ix1,jx0);
2437             dy10             = _mm_sub_pd(iy1,jy0);
2438             dz10             = _mm_sub_pd(iz1,jz0);
2439             dx11             = _mm_sub_pd(ix1,jx1);
2440             dy11             = _mm_sub_pd(iy1,jy1);
2441             dz11             = _mm_sub_pd(iz1,jz1);
2442             dx12             = _mm_sub_pd(ix1,jx2);
2443             dy12             = _mm_sub_pd(iy1,jy2);
2444             dz12             = _mm_sub_pd(iz1,jz2);
2445             dx20             = _mm_sub_pd(ix2,jx0);
2446             dy20             = _mm_sub_pd(iy2,jy0);
2447             dz20             = _mm_sub_pd(iz2,jz0);
2448             dx21             = _mm_sub_pd(ix2,jx1);
2449             dy21             = _mm_sub_pd(iy2,jy1);
2450             dz21             = _mm_sub_pd(iz2,jz1);
2451             dx22             = _mm_sub_pd(ix2,jx2);
2452             dy22             = _mm_sub_pd(iy2,jy2);
2453             dz22             = _mm_sub_pd(iz2,jz2);
2454
2455             /* Calculate squared distance and things based on it */
2456             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
2457             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
2458             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
2459             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
2460             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
2461             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
2462             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
2463             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
2464             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
2465
2466             rinv00           = gmx_mm_invsqrt_pd(rsq00);
2467             rinv01           = gmx_mm_invsqrt_pd(rsq01);
2468             rinv02           = gmx_mm_invsqrt_pd(rsq02);
2469             rinv10           = gmx_mm_invsqrt_pd(rsq10);
2470             rinv11           = gmx_mm_invsqrt_pd(rsq11);
2471             rinv12           = gmx_mm_invsqrt_pd(rsq12);
2472             rinv20           = gmx_mm_invsqrt_pd(rsq20);
2473             rinv21           = gmx_mm_invsqrt_pd(rsq21);
2474             rinv22           = gmx_mm_invsqrt_pd(rsq22);
2475
2476             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
2477             rinvsq01         = _mm_mul_pd(rinv01,rinv01);
2478             rinvsq02         = _mm_mul_pd(rinv02,rinv02);
2479             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
2480             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
2481             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
2482             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
2483             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
2484             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
2485
2486             fjx0             = _mm_setzero_pd();
2487             fjy0             = _mm_setzero_pd();
2488             fjz0             = _mm_setzero_pd();
2489             fjx1             = _mm_setzero_pd();
2490             fjy1             = _mm_setzero_pd();
2491             fjz1             = _mm_setzero_pd();
2492             fjx2             = _mm_setzero_pd();
2493             fjy2             = _mm_setzero_pd();
2494             fjz2             = _mm_setzero_pd();
2495
2496             /**************************
2497              * CALCULATE INTERACTIONS *
2498              **************************/
2499
2500             if (gmx_mm_any_lt(rsq00,rcutoff2))
2501             {
2502
2503             r00              = _mm_mul_pd(rsq00,rinv00);
2504
2505             /* EWALD ELECTROSTATICS */
2506
2507             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2508             ewrt             = _mm_mul_pd(r00,ewtabscale);
2509             ewitab           = _mm_cvttpd_epi32(ewrt);
2510 #ifdef __XOP__
2511             eweps            = _mm_frcz_pd(ewrt);
2512 #else
2513             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2514 #endif
2515             twoeweps         = _mm_add_pd(eweps,eweps);
2516             ewitab           = _mm_slli_epi32(ewitab,2);
2517             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
2518             ewtabD           = _mm_setzero_pd();
2519             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
2520             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
2521             ewtabFn          = _mm_setzero_pd();
2522             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
2523             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
2524             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
2525             velec            = _mm_mul_pd(qq00,_mm_sub_pd(rinv00,velec));
2526             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
2527
2528             /* LENNARD-JONES DISPERSION/REPULSION */
2529
2530             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
2531             vvdw6            = _mm_mul_pd(c6_00,rinvsix);
2532             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
2533             vvdw             = _mm_msub_pd( vvdw12,one_twelfth, _mm_mul_pd(vvdw6,one_sixth) );
2534             fvdw             = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
2535
2536             d                = _mm_sub_pd(r00,rswitch);
2537             d                = _mm_max_pd(d,_mm_setzero_pd());
2538             d2               = _mm_mul_pd(d,d);
2539             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
2540
2541             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
2542
2543             /* Evaluate switch function */
2544             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2545             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv00,_mm_mul_pd(velec,dsw)) );
2546             fvdw             = _mm_msub_pd( fvdw,sw , _mm_mul_pd(rinv00,_mm_mul_pd(vvdw,dsw)) );
2547             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
2548
2549             fscal            = _mm_add_pd(felec,fvdw);
2550
2551             fscal            = _mm_and_pd(fscal,cutoff_mask);
2552
2553             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2554
2555             /* Update vectorial force */
2556             fix0             = _mm_macc_pd(dx00,fscal,fix0);
2557             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
2558             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
2559
2560             fjx0             = _mm_macc_pd(dx00,fscal,fjx0);
2561             fjy0             = _mm_macc_pd(dy00,fscal,fjy0);
2562             fjz0             = _mm_macc_pd(dz00,fscal,fjz0);
2563
2564             }
2565
2566             /**************************
2567              * CALCULATE INTERACTIONS *
2568              **************************/
2569
2570             if (gmx_mm_any_lt(rsq01,rcutoff2))
2571             {
2572
2573             r01              = _mm_mul_pd(rsq01,rinv01);
2574
2575             /* EWALD ELECTROSTATICS */
2576
2577             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2578             ewrt             = _mm_mul_pd(r01,ewtabscale);
2579             ewitab           = _mm_cvttpd_epi32(ewrt);
2580 #ifdef __XOP__
2581             eweps            = _mm_frcz_pd(ewrt);
2582 #else
2583             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2584 #endif
2585             twoeweps         = _mm_add_pd(eweps,eweps);
2586             ewitab           = _mm_slli_epi32(ewitab,2);
2587             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
2588             ewtabD           = _mm_setzero_pd();
2589             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
2590             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
2591             ewtabFn          = _mm_setzero_pd();
2592             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
2593             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
2594             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
2595             velec            = _mm_mul_pd(qq01,_mm_sub_pd(rinv01,velec));
2596             felec            = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
2597
2598             d                = _mm_sub_pd(r01,rswitch);
2599             d                = _mm_max_pd(d,_mm_setzero_pd());
2600             d2               = _mm_mul_pd(d,d);
2601             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
2602
2603             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
2604
2605             /* Evaluate switch function */
2606             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2607             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv01,_mm_mul_pd(velec,dsw)) );
2608             cutoff_mask      = _mm_cmplt_pd(rsq01,rcutoff2);
2609
2610             fscal            = felec;
2611
2612             fscal            = _mm_and_pd(fscal,cutoff_mask);
2613
2614             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2615
2616             /* Update vectorial force */
2617             fix0             = _mm_macc_pd(dx01,fscal,fix0);
2618             fiy0             = _mm_macc_pd(dy01,fscal,fiy0);
2619             fiz0             = _mm_macc_pd(dz01,fscal,fiz0);
2620
2621             fjx1             = _mm_macc_pd(dx01,fscal,fjx1);
2622             fjy1             = _mm_macc_pd(dy01,fscal,fjy1);
2623             fjz1             = _mm_macc_pd(dz01,fscal,fjz1);
2624
2625             }
2626
2627             /**************************
2628              * CALCULATE INTERACTIONS *
2629              **************************/
2630
2631             if (gmx_mm_any_lt(rsq02,rcutoff2))
2632             {
2633
2634             r02              = _mm_mul_pd(rsq02,rinv02);
2635
2636             /* EWALD ELECTROSTATICS */
2637
2638             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2639             ewrt             = _mm_mul_pd(r02,ewtabscale);
2640             ewitab           = _mm_cvttpd_epi32(ewrt);
2641 #ifdef __XOP__
2642             eweps            = _mm_frcz_pd(ewrt);
2643 #else
2644             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2645 #endif
2646             twoeweps         = _mm_add_pd(eweps,eweps);
2647             ewitab           = _mm_slli_epi32(ewitab,2);
2648             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
2649             ewtabD           = _mm_setzero_pd();
2650             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
2651             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
2652             ewtabFn          = _mm_setzero_pd();
2653             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
2654             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
2655             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
2656             velec            = _mm_mul_pd(qq02,_mm_sub_pd(rinv02,velec));
2657             felec            = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
2658
2659             d                = _mm_sub_pd(r02,rswitch);
2660             d                = _mm_max_pd(d,_mm_setzero_pd());
2661             d2               = _mm_mul_pd(d,d);
2662             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
2663
2664             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
2665
2666             /* Evaluate switch function */
2667             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2668             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv02,_mm_mul_pd(velec,dsw)) );
2669             cutoff_mask      = _mm_cmplt_pd(rsq02,rcutoff2);
2670
2671             fscal            = felec;
2672
2673             fscal            = _mm_and_pd(fscal,cutoff_mask);
2674
2675             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2676
2677             /* Update vectorial force */
2678             fix0             = _mm_macc_pd(dx02,fscal,fix0);
2679             fiy0             = _mm_macc_pd(dy02,fscal,fiy0);
2680             fiz0             = _mm_macc_pd(dz02,fscal,fiz0);
2681
2682             fjx2             = _mm_macc_pd(dx02,fscal,fjx2);
2683             fjy2             = _mm_macc_pd(dy02,fscal,fjy2);
2684             fjz2             = _mm_macc_pd(dz02,fscal,fjz2);
2685
2686             }
2687
2688             /**************************
2689              * CALCULATE INTERACTIONS *
2690              **************************/
2691
2692             if (gmx_mm_any_lt(rsq10,rcutoff2))
2693             {
2694
2695             r10              = _mm_mul_pd(rsq10,rinv10);
2696
2697             /* EWALD ELECTROSTATICS */
2698
2699             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2700             ewrt             = _mm_mul_pd(r10,ewtabscale);
2701             ewitab           = _mm_cvttpd_epi32(ewrt);
2702 #ifdef __XOP__
2703             eweps            = _mm_frcz_pd(ewrt);
2704 #else
2705             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2706 #endif
2707             twoeweps         = _mm_add_pd(eweps,eweps);
2708             ewitab           = _mm_slli_epi32(ewitab,2);
2709             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
2710             ewtabD           = _mm_setzero_pd();
2711             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
2712             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
2713             ewtabFn          = _mm_setzero_pd();
2714             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
2715             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
2716             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
2717             velec            = _mm_mul_pd(qq10,_mm_sub_pd(rinv10,velec));
2718             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
2719
2720             d                = _mm_sub_pd(r10,rswitch);
2721             d                = _mm_max_pd(d,_mm_setzero_pd());
2722             d2               = _mm_mul_pd(d,d);
2723             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
2724
2725             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
2726
2727             /* Evaluate switch function */
2728             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2729             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv10,_mm_mul_pd(velec,dsw)) );
2730             cutoff_mask      = _mm_cmplt_pd(rsq10,rcutoff2);
2731
2732             fscal            = felec;
2733
2734             fscal            = _mm_and_pd(fscal,cutoff_mask);
2735
2736             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2737
2738             /* Update vectorial force */
2739             fix1             = _mm_macc_pd(dx10,fscal,fix1);
2740             fiy1             = _mm_macc_pd(dy10,fscal,fiy1);
2741             fiz1             = _mm_macc_pd(dz10,fscal,fiz1);
2742
2743             fjx0             = _mm_macc_pd(dx10,fscal,fjx0);
2744             fjy0             = _mm_macc_pd(dy10,fscal,fjy0);
2745             fjz0             = _mm_macc_pd(dz10,fscal,fjz0);
2746
2747             }
2748
2749             /**************************
2750              * CALCULATE INTERACTIONS *
2751              **************************/
2752
2753             if (gmx_mm_any_lt(rsq11,rcutoff2))
2754             {
2755
2756             r11              = _mm_mul_pd(rsq11,rinv11);
2757
2758             /* EWALD ELECTROSTATICS */
2759
2760             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2761             ewrt             = _mm_mul_pd(r11,ewtabscale);
2762             ewitab           = _mm_cvttpd_epi32(ewrt);
2763 #ifdef __XOP__
2764             eweps            = _mm_frcz_pd(ewrt);
2765 #else
2766             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2767 #endif
2768             twoeweps         = _mm_add_pd(eweps,eweps);
2769             ewitab           = _mm_slli_epi32(ewitab,2);
2770             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
2771             ewtabD           = _mm_setzero_pd();
2772             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
2773             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
2774             ewtabFn          = _mm_setzero_pd();
2775             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
2776             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
2777             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
2778             velec            = _mm_mul_pd(qq11,_mm_sub_pd(rinv11,velec));
2779             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
2780
2781             d                = _mm_sub_pd(r11,rswitch);
2782             d                = _mm_max_pd(d,_mm_setzero_pd());
2783             d2               = _mm_mul_pd(d,d);
2784             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
2785
2786             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
2787
2788             /* Evaluate switch function */
2789             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2790             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv11,_mm_mul_pd(velec,dsw)) );
2791             cutoff_mask      = _mm_cmplt_pd(rsq11,rcutoff2);
2792
2793             fscal            = felec;
2794
2795             fscal            = _mm_and_pd(fscal,cutoff_mask);
2796
2797             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2798
2799             /* Update vectorial force */
2800             fix1             = _mm_macc_pd(dx11,fscal,fix1);
2801             fiy1             = _mm_macc_pd(dy11,fscal,fiy1);
2802             fiz1             = _mm_macc_pd(dz11,fscal,fiz1);
2803
2804             fjx1             = _mm_macc_pd(dx11,fscal,fjx1);
2805             fjy1             = _mm_macc_pd(dy11,fscal,fjy1);
2806             fjz1             = _mm_macc_pd(dz11,fscal,fjz1);
2807
2808             }
2809
2810             /**************************
2811              * CALCULATE INTERACTIONS *
2812              **************************/
2813
2814             if (gmx_mm_any_lt(rsq12,rcutoff2))
2815             {
2816
2817             r12              = _mm_mul_pd(rsq12,rinv12);
2818
2819             /* EWALD ELECTROSTATICS */
2820
2821             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2822             ewrt             = _mm_mul_pd(r12,ewtabscale);
2823             ewitab           = _mm_cvttpd_epi32(ewrt);
2824 #ifdef __XOP__
2825             eweps            = _mm_frcz_pd(ewrt);
2826 #else
2827             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2828 #endif
2829             twoeweps         = _mm_add_pd(eweps,eweps);
2830             ewitab           = _mm_slli_epi32(ewitab,2);
2831             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
2832             ewtabD           = _mm_setzero_pd();
2833             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
2834             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
2835             ewtabFn          = _mm_setzero_pd();
2836             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
2837             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
2838             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
2839             velec            = _mm_mul_pd(qq12,_mm_sub_pd(rinv12,velec));
2840             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
2841
2842             d                = _mm_sub_pd(r12,rswitch);
2843             d                = _mm_max_pd(d,_mm_setzero_pd());
2844             d2               = _mm_mul_pd(d,d);
2845             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
2846
2847             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
2848
2849             /* Evaluate switch function */
2850             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2851             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv12,_mm_mul_pd(velec,dsw)) );
2852             cutoff_mask      = _mm_cmplt_pd(rsq12,rcutoff2);
2853
2854             fscal            = felec;
2855
2856             fscal            = _mm_and_pd(fscal,cutoff_mask);
2857
2858             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2859
2860             /* Update vectorial force */
2861             fix1             = _mm_macc_pd(dx12,fscal,fix1);
2862             fiy1             = _mm_macc_pd(dy12,fscal,fiy1);
2863             fiz1             = _mm_macc_pd(dz12,fscal,fiz1);
2864
2865             fjx2             = _mm_macc_pd(dx12,fscal,fjx2);
2866             fjy2             = _mm_macc_pd(dy12,fscal,fjy2);
2867             fjz2             = _mm_macc_pd(dz12,fscal,fjz2);
2868
2869             }
2870
2871             /**************************
2872              * CALCULATE INTERACTIONS *
2873              **************************/
2874
2875             if (gmx_mm_any_lt(rsq20,rcutoff2))
2876             {
2877
2878             r20              = _mm_mul_pd(rsq20,rinv20);
2879
2880             /* EWALD ELECTROSTATICS */
2881
2882             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2883             ewrt             = _mm_mul_pd(r20,ewtabscale);
2884             ewitab           = _mm_cvttpd_epi32(ewrt);
2885 #ifdef __XOP__
2886             eweps            = _mm_frcz_pd(ewrt);
2887 #else
2888             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2889 #endif
2890             twoeweps         = _mm_add_pd(eweps,eweps);
2891             ewitab           = _mm_slli_epi32(ewitab,2);
2892             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
2893             ewtabD           = _mm_setzero_pd();
2894             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
2895             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
2896             ewtabFn          = _mm_setzero_pd();
2897             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
2898             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
2899             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
2900             velec            = _mm_mul_pd(qq20,_mm_sub_pd(rinv20,velec));
2901             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
2902
2903             d                = _mm_sub_pd(r20,rswitch);
2904             d                = _mm_max_pd(d,_mm_setzero_pd());
2905             d2               = _mm_mul_pd(d,d);
2906             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
2907
2908             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
2909
2910             /* Evaluate switch function */
2911             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2912             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv20,_mm_mul_pd(velec,dsw)) );
2913             cutoff_mask      = _mm_cmplt_pd(rsq20,rcutoff2);
2914
2915             fscal            = felec;
2916
2917             fscal            = _mm_and_pd(fscal,cutoff_mask);
2918
2919             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2920
2921             /* Update vectorial force */
2922             fix2             = _mm_macc_pd(dx20,fscal,fix2);
2923             fiy2             = _mm_macc_pd(dy20,fscal,fiy2);
2924             fiz2             = _mm_macc_pd(dz20,fscal,fiz2);
2925
2926             fjx0             = _mm_macc_pd(dx20,fscal,fjx0);
2927             fjy0             = _mm_macc_pd(dy20,fscal,fjy0);
2928             fjz0             = _mm_macc_pd(dz20,fscal,fjz0);
2929
2930             }
2931
2932             /**************************
2933              * CALCULATE INTERACTIONS *
2934              **************************/
2935
2936             if (gmx_mm_any_lt(rsq21,rcutoff2))
2937             {
2938
2939             r21              = _mm_mul_pd(rsq21,rinv21);
2940
2941             /* EWALD ELECTROSTATICS */
2942
2943             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2944             ewrt             = _mm_mul_pd(r21,ewtabscale);
2945             ewitab           = _mm_cvttpd_epi32(ewrt);
2946 #ifdef __XOP__
2947             eweps            = _mm_frcz_pd(ewrt);
2948 #else
2949             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2950 #endif
2951             twoeweps         = _mm_add_pd(eweps,eweps);
2952             ewitab           = _mm_slli_epi32(ewitab,2);
2953             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
2954             ewtabD           = _mm_setzero_pd();
2955             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
2956             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
2957             ewtabFn          = _mm_setzero_pd();
2958             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
2959             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
2960             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
2961             velec            = _mm_mul_pd(qq21,_mm_sub_pd(rinv21,velec));
2962             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
2963
2964             d                = _mm_sub_pd(r21,rswitch);
2965             d                = _mm_max_pd(d,_mm_setzero_pd());
2966             d2               = _mm_mul_pd(d,d);
2967             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
2968
2969             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
2970
2971             /* Evaluate switch function */
2972             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2973             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv21,_mm_mul_pd(velec,dsw)) );
2974             cutoff_mask      = _mm_cmplt_pd(rsq21,rcutoff2);
2975
2976             fscal            = felec;
2977
2978             fscal            = _mm_and_pd(fscal,cutoff_mask);
2979
2980             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2981
2982             /* Update vectorial force */
2983             fix2             = _mm_macc_pd(dx21,fscal,fix2);
2984             fiy2             = _mm_macc_pd(dy21,fscal,fiy2);
2985             fiz2             = _mm_macc_pd(dz21,fscal,fiz2);
2986
2987             fjx1             = _mm_macc_pd(dx21,fscal,fjx1);
2988             fjy1             = _mm_macc_pd(dy21,fscal,fjy1);
2989             fjz1             = _mm_macc_pd(dz21,fscal,fjz1);
2990
2991             }
2992
2993             /**************************
2994              * CALCULATE INTERACTIONS *
2995              **************************/
2996
2997             if (gmx_mm_any_lt(rsq22,rcutoff2))
2998             {
2999
3000             r22              = _mm_mul_pd(rsq22,rinv22);
3001
3002             /* EWALD ELECTROSTATICS */
3003
3004             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
3005             ewrt             = _mm_mul_pd(r22,ewtabscale);
3006             ewitab           = _mm_cvttpd_epi32(ewrt);
3007 #ifdef __XOP__
3008             eweps            = _mm_frcz_pd(ewrt);
3009 #else
3010             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
3011 #endif
3012             twoeweps         = _mm_add_pd(eweps,eweps);
3013             ewitab           = _mm_slli_epi32(ewitab,2);
3014             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
3015             ewtabD           = _mm_setzero_pd();
3016             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
3017             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
3018             ewtabFn          = _mm_setzero_pd();
3019             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
3020             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
3021             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
3022             velec            = _mm_mul_pd(qq22,_mm_sub_pd(rinv22,velec));
3023             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
3024
3025             d                = _mm_sub_pd(r22,rswitch);
3026             d                = _mm_max_pd(d,_mm_setzero_pd());
3027             d2               = _mm_mul_pd(d,d);
3028             sw               = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
3029
3030             dsw              = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
3031
3032             /* Evaluate switch function */
3033             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
3034             felec            = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv22,_mm_mul_pd(velec,dsw)) );
3035             cutoff_mask      = _mm_cmplt_pd(rsq22,rcutoff2);
3036
3037             fscal            = felec;
3038
3039             fscal            = _mm_and_pd(fscal,cutoff_mask);
3040
3041             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
3042
3043             /* Update vectorial force */
3044             fix2             = _mm_macc_pd(dx22,fscal,fix2);
3045             fiy2             = _mm_macc_pd(dy22,fscal,fiy2);
3046             fiz2             = _mm_macc_pd(dz22,fscal,fiz2);
3047
3048             fjx2             = _mm_macc_pd(dx22,fscal,fjx2);
3049             fjy2             = _mm_macc_pd(dy22,fscal,fjy2);
3050             fjz2             = _mm_macc_pd(dz22,fscal,fjz2);
3051
3052             }
3053
3054             gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
3055
3056             /* Inner loop uses 600 flops */
3057         }
3058
3059         /* End of innermost loop */
3060
3061         gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
3062                                               f+i_coord_offset,fshift+i_shift_offset);
3063
3064         /* Increment number of inner iterations */
3065         inneriter                  += j_index_end - j_index_start;
3066
3067         /* Outer loop uses 18 flops */
3068     }
3069
3070     /* Increment number of outer iterations */
3071     outeriter        += nri;
3072
3073     /* Update outer/inner flops */
3074
3075     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_F,outeriter*18 + inneriter*600);
3076 }