src/gromacs/gmxlib/nonbonded/nb_kernel_avx_128_fma_single/nb_kernel_ElecEw_VdwCSTab_GeomW3W3_avx_128_fma_single.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS avx_128_fma_single kernel generator.
  37  */
  38 #include "gmxpre.h"
  39
  40 #include "config.h"
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "gromacs/legacyheaders/types/simple.h"
  46 #include "gromacs/math/vec.h"
  47 #include "gromacs/legacyheaders/nrnb.h"
  48
  49 #include "gromacs/simd/math_x86_avx_128_fma_single.h"
  50 #include "kernelutil_x86_avx_128_fma_single.h"
  51
  52 /*
  53  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwCSTab_GeomW3W3_VF_avx_128_fma_single
  54  * Electrostatics interaction: Ewald
  55  * VdW interaction:            CubicSplineTable
  56  * Geometry:                   Water3-Water3
  57  * Calculate force/pot:        PotentialAndForce
  58  */
  59 void
  60 nb_kernel_ElecEw_VdwCSTab_GeomW3W3_VF_avx_128_fma_single
  61                     (t_nblist                    * gmx_restrict       nlist,
  62                      rvec                        * gmx_restrict          xx,
  63                      rvec                        * gmx_restrict          ff,
  64                      t_forcerec                  * gmx_restrict          fr,
  65                      t_mdatoms                   * gmx_restrict     mdatoms,
  66                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  67                      t_nrnb                      * gmx_restrict        nrnb)
  68 {
  69     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  70      * just 0 for non-waters.
  71      * Suffixes A,B,C,D refer to j loop unrolling done with AVX_128, e.g. for the four different
  72      * jnr indices corresponding to data put in the four positions in the SIMD register.
  73      */
  74     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  75     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  76     int              jnrA,jnrB,jnrC,jnrD;
  77     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  78     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  79     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  80     real             rcutoff_scalar;
  81     real             *shiftvec,*fshift,*x,*f;
  82     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
  83     real             scratch[4*DIM];
  84     __m128           fscal,rcutoff,rcutoff2,jidxall;
  85     int              vdwioffset0;
  86     __m128           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  87     int              vdwioffset1;
  88     __m128           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  89     int              vdwioffset2;
  90     __m128           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  91     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
  92     __m128           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  93     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
  94     __m128           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  95     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
  96     __m128           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  97     __m128           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  98     __m128           dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
  99     __m128           dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
 100     __m128           dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 101     __m128           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
 102     __m128           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
 103     __m128           dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 104     __m128           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 105     __m128           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 106     __m128           velec,felec,velecsum,facel,crf,krf,krf2;
 107     real             *charge;
 108     int              nvdwtype;
 109     __m128           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 110     int              *vdwtype;
 111     real             *vdwparam;
 112     __m128           one_sixth   = _mm_set1_ps(1.0/6.0);
 113     __m128           one_twelfth = _mm_set1_ps(1.0/12.0);
 114     __m128i          vfitab;
 115     __m128i          ifour       = _mm_set1_epi32(4);
 116     __m128           rt,vfeps,twovfeps,vftabscale,Y,F,G,H,Fp,VV,FF;
 117     real             *vftab;
 118     __m128i          ewitab;
 119     __m128           ewtabscale,eweps,twoeweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 120     __m128           beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
 121     real             *ewtab;
 122     __m128           dummy_mask,cutoff_mask;
 123     __m128           signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
 124     __m128           one     = _mm_set1_ps(1.0);
 125     __m128           two     = _mm_set1_ps(2.0);
 126     x                = xx[0];
 127     f                = ff[0];
 128
 129     nri              = nlist->nri;
 130     iinr             = nlist->iinr;
 131     jindex           = nlist->jindex;
 132     jjnr             = nlist->jjnr;
 133     shiftidx         = nlist->shift;
 134     gid              = nlist->gid;
 135     shiftvec         = fr->shift_vec[0];
 136     fshift           = fr->fshift[0];
 137     facel            = _mm_set1_ps(fr->epsfac);
 138     charge           = mdatoms->chargeA;
 139     nvdwtype         = fr->ntype;
 140     vdwparam         = fr->nbfp;
 141     vdwtype          = mdatoms->typeA;
 142
 143     vftab            = kernel_data->table_vdw->data;
 144     vftabscale       = _mm_set1_ps(kernel_data->table_vdw->scale);
 145
 146     sh_ewald         = _mm_set1_ps(fr->ic->sh_ewald);
 147     beta             = _mm_set1_ps(fr->ic->ewaldcoeff_q);
 148     beta2            = _mm_mul_ps(beta,beta);
 149     beta3            = _mm_mul_ps(beta,beta2);
 150     ewtab            = fr->ic->tabq_coul_FDV0;
 151     ewtabscale       = _mm_set1_ps(fr->ic->tabq_scale);
 152     ewtabhalfspace   = _mm_set1_ps(0.5/fr->ic->tabq_scale);
 153
 154     /* Setup water-specific parameters */
 155     inr              = nlist->iinr[0];
 156     iq0              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+0]));
 157     iq1              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1]));
 158     iq2              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2]));
 159     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 160
 161     jq0              = _mm_set1_ps(charge[inr+0]);
 162     jq1              = _mm_set1_ps(charge[inr+1]);
 163     jq2              = _mm_set1_ps(charge[inr+2]);
 164     vdwjidx0A        = 2*vdwtype[inr+0];
 165     qq00             = _mm_mul_ps(iq0,jq0);
 166     c6_00            = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A]);
 167     c12_00           = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A+1]);
 168     qq01             = _mm_mul_ps(iq0,jq1);
 169     qq02             = _mm_mul_ps(iq0,jq2);
 170     qq10             = _mm_mul_ps(iq1,jq0);
 171     qq11             = _mm_mul_ps(iq1,jq1);
 172     qq12             = _mm_mul_ps(iq1,jq2);
 173     qq20             = _mm_mul_ps(iq2,jq0);
 174     qq21             = _mm_mul_ps(iq2,jq1);
 175     qq22             = _mm_mul_ps(iq2,jq2);
 176
 177     /* Avoid stupid compiler warnings */
 178     jnrA = jnrB = jnrC = jnrD = 0;
 179     j_coord_offsetA = 0;
 180     j_coord_offsetB = 0;
 181     j_coord_offsetC = 0;
 182     j_coord_offsetD = 0;
 183
 184     outeriter        = 0;
 185     inneriter        = 0;
 186
 187     for(iidx=0;iidx<4*DIM;iidx++)
 188     {
 189         scratch[iidx] = 0.0;
 190     }
 191
 192     /* Start outer loop over neighborlists */
 193     for(iidx=0; iidx<nri; iidx++)
 194     {
 195         /* Load shift vector for this list */
 196         i_shift_offset   = DIM*shiftidx[iidx];
 197
 198         /* Load limits for loop over neighbors */
 199         j_index_start    = jindex[iidx];
 200         j_index_end      = jindex[iidx+1];
 201
 202         /* Get outer coordinate index */
 203         inr              = iinr[iidx];
 204         i_coord_offset   = DIM*inr;
 205
 206         /* Load i particle coords and add shift vector */
 207         gmx_mm_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
 208                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 209
 210         fix0             = _mm_setzero_ps();
 211         fiy0             = _mm_setzero_ps();
 212         fiz0             = _mm_setzero_ps();
 213         fix1             = _mm_setzero_ps();
 214         fiy1             = _mm_setzero_ps();
 215         fiz1             = _mm_setzero_ps();
 216         fix2             = _mm_setzero_ps();
 217         fiy2             = _mm_setzero_ps();
 218         fiz2             = _mm_setzero_ps();
 219
 220         /* Reset potential sums */
 221         velecsum         = _mm_setzero_ps();
 222         vvdwsum          = _mm_setzero_ps();
 223
 224         /* Start inner kernel loop */
 225         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 226         {
 227
 228             /* Get j neighbor index, and coordinate index */
 229             jnrA             = jjnr[jidx];
 230             jnrB             = jjnr[jidx+1];
 231             jnrC             = jjnr[jidx+2];
 232             jnrD             = jjnr[jidx+3];
 233             j_coord_offsetA  = DIM*jnrA;
 234             j_coord_offsetB  = DIM*jnrB;
 235             j_coord_offsetC  = DIM*jnrC;
 236             j_coord_offsetD  = DIM*jnrD;
 237
 238             /* load j atom coordinates */
 239             gmx_mm_load_3rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 240                                               x+j_coord_offsetC,x+j_coord_offsetD,
 241                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 242
 243             /* Calculate displacement vector */
 244             dx00             = _mm_sub_ps(ix0,jx0);
 245             dy00             = _mm_sub_ps(iy0,jy0);
 246             dz00             = _mm_sub_ps(iz0,jz0);
 247             dx01             = _mm_sub_ps(ix0,jx1);
 248             dy01             = _mm_sub_ps(iy0,jy1);
 249             dz01             = _mm_sub_ps(iz0,jz1);
 250             dx02             = _mm_sub_ps(ix0,jx2);
 251             dy02             = _mm_sub_ps(iy0,jy2);
 252             dz02             = _mm_sub_ps(iz0,jz2);
 253             dx10             = _mm_sub_ps(ix1,jx0);
 254             dy10             = _mm_sub_ps(iy1,jy0);
 255             dz10             = _mm_sub_ps(iz1,jz0);
 256             dx11             = _mm_sub_ps(ix1,jx1);
 257             dy11             = _mm_sub_ps(iy1,jy1);
 258             dz11             = _mm_sub_ps(iz1,jz1);
 259             dx12             = _mm_sub_ps(ix1,jx2);
 260             dy12             = _mm_sub_ps(iy1,jy2);
 261             dz12             = _mm_sub_ps(iz1,jz2);
 262             dx20             = _mm_sub_ps(ix2,jx0);
 263             dy20             = _mm_sub_ps(iy2,jy0);
 264             dz20             = _mm_sub_ps(iz2,jz0);
 265             dx21             = _mm_sub_ps(ix2,jx1);
 266             dy21             = _mm_sub_ps(iy2,jy1);
 267             dz21             = _mm_sub_ps(iz2,jz1);
 268             dx22             = _mm_sub_ps(ix2,jx2);
 269             dy22             = _mm_sub_ps(iy2,jy2);
 270             dz22             = _mm_sub_ps(iz2,jz2);
 271
 272             /* Calculate squared distance and things based on it */
 273             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
 274             rsq01            = gmx_mm_calc_rsq_ps(dx01,dy01,dz01);
 275             rsq02            = gmx_mm_calc_rsq_ps(dx02,dy02,dz02);
 276             rsq10            = gmx_mm_calc_rsq_ps(dx10,dy10,dz10);
 277             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
 278             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
 279             rsq20            = gmx_mm_calc_rsq_ps(dx20,dy20,dz20);
 280             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
 281             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
 282
 283             rinv00           = gmx_mm_invsqrt_ps(rsq00);
 284             rinv01           = gmx_mm_invsqrt_ps(rsq01);
 285             rinv02           = gmx_mm_invsqrt_ps(rsq02);
 286             rinv10           = gmx_mm_invsqrt_ps(rsq10);
 287             rinv11           = gmx_mm_invsqrt_ps(rsq11);
 288             rinv12           = gmx_mm_invsqrt_ps(rsq12);
 289             rinv20           = gmx_mm_invsqrt_ps(rsq20);
 290             rinv21           = gmx_mm_invsqrt_ps(rsq21);
 291             rinv22           = gmx_mm_invsqrt_ps(rsq22);
 292
 293             rinvsq00         = _mm_mul_ps(rinv00,rinv00);
 294             rinvsq01         = _mm_mul_ps(rinv01,rinv01);
 295             rinvsq02         = _mm_mul_ps(rinv02,rinv02);
 296             rinvsq10         = _mm_mul_ps(rinv10,rinv10);
 297             rinvsq11         = _mm_mul_ps(rinv11,rinv11);
 298             rinvsq12         = _mm_mul_ps(rinv12,rinv12);
 299             rinvsq20         = _mm_mul_ps(rinv20,rinv20);
 300             rinvsq21         = _mm_mul_ps(rinv21,rinv21);
 301             rinvsq22         = _mm_mul_ps(rinv22,rinv22);
 302
 303             fjx0             = _mm_setzero_ps();
 304             fjy0             = _mm_setzero_ps();
 305             fjz0             = _mm_setzero_ps();
 306             fjx1             = _mm_setzero_ps();
 307             fjy1             = _mm_setzero_ps();
 308             fjz1             = _mm_setzero_ps();
 309             fjx2             = _mm_setzero_ps();
 310             fjy2             = _mm_setzero_ps();
 311             fjz2             = _mm_setzero_ps();
 312
 313             /**************************
 314              * CALCULATE INTERACTIONS *
 315              **************************/
 316
 317             r00              = _mm_mul_ps(rsq00,rinv00);
 318
 319             /* Calculate table index by multiplying r with table scale and truncate to integer */
 320             rt               = _mm_mul_ps(r00,vftabscale);
 321             vfitab           = _mm_cvttps_epi32(rt);
 322 #ifdef __XOP__
 323             vfeps            = _mm_frcz_ps(rt);
 324 #else
 325             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
 326 #endif
 327             twovfeps         = _mm_add_ps(vfeps,vfeps);
 328             vfitab           = _mm_slli_epi32(vfitab,3);
 329
 330             /* EWALD ELECTROSTATICS */
 331
 332             /* Analytical PME correction */
 333             zeta2            = _mm_mul_ps(beta2,rsq00);
 334             rinv3            = _mm_mul_ps(rinvsq00,rinv00);
 335             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
 336             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
 337             felec            = _mm_mul_ps(qq00,felec);
 338             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
 339             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv00);
 340             velec            = _mm_mul_ps(qq00,velec);
 341
 342             /* CUBIC SPLINE TABLE DISPERSION */
 343             Y                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
 344             F                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
 345             G                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
 346             H                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
 347             _MM_TRANSPOSE4_PS(Y,F,G,H);
 348             Fp               = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
 349             VV               = _mm_macc_ps(vfeps,Fp,Y);
 350             vvdw6            = _mm_mul_ps(c6_00,VV);
 351             FF               = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
 352             fvdw6            = _mm_mul_ps(c6_00,FF);
 353
 354             /* CUBIC SPLINE TABLE REPULSION */
 355             vfitab           = _mm_add_epi32(vfitab,ifour);
 356             Y                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
 357             F                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
 358             G                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
 359             H                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
 360             _MM_TRANSPOSE4_PS(Y,F,G,H);
 361             Fp               = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
 362             VV               = _mm_macc_ps(vfeps,Fp,Y);
 363             vvdw12           = _mm_mul_ps(c12_00,VV);
 364             FF               = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
 365             fvdw12           = _mm_mul_ps(c12_00,FF);
 366             vvdw             = _mm_add_ps(vvdw12,vvdw6);
 367             fvdw             = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
 368
 369             /* Update potential sum for this i atom from the interaction with this j atom. */
 370             velecsum         = _mm_add_ps(velecsum,velec);
 371             vvdwsum          = _mm_add_ps(vvdwsum,vvdw);
 372
 373             fscal            = _mm_add_ps(felec,fvdw);
 374
 375              /* Update vectorial force */
 376             fix0             = _mm_macc_ps(dx00,fscal,fix0);
 377             fiy0             = _mm_macc_ps(dy00,fscal,fiy0);
 378             fiz0             = _mm_macc_ps(dz00,fscal,fiz0);
 379
 380             fjx0             = _mm_macc_ps(dx00,fscal,fjx0);
 381             fjy0             = _mm_macc_ps(dy00,fscal,fjy0);
 382             fjz0             = _mm_macc_ps(dz00,fscal,fjz0);
 383
 384             /**************************
 385              * CALCULATE INTERACTIONS *
 386              **************************/
 387
 388             r01              = _mm_mul_ps(rsq01,rinv01);
 389
 390             /* EWALD ELECTROSTATICS */
 391
 392             /* Analytical PME correction */
 393             zeta2            = _mm_mul_ps(beta2,rsq01);
 394             rinv3            = _mm_mul_ps(rinvsq01,rinv01);
 395             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
 396             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
 397             felec            = _mm_mul_ps(qq01,felec);
 398             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
 399             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv01);
 400             velec            = _mm_mul_ps(qq01,velec);
 401
 402             /* Update potential sum for this i atom from the interaction with this j atom. */
 403             velecsum         = _mm_add_ps(velecsum,velec);
 404
 405             fscal            = felec;
 406
 407              /* Update vectorial force */
 408             fix0             = _mm_macc_ps(dx01,fscal,fix0);
 409             fiy0             = _mm_macc_ps(dy01,fscal,fiy0);
 410             fiz0             = _mm_macc_ps(dz01,fscal,fiz0);
 411
 412             fjx1             = _mm_macc_ps(dx01,fscal,fjx1);
 413             fjy1             = _mm_macc_ps(dy01,fscal,fjy1);
 414             fjz1             = _mm_macc_ps(dz01,fscal,fjz1);
 415
 416             /**************************
 417              * CALCULATE INTERACTIONS *
 418              **************************/
 419
 420             r02              = _mm_mul_ps(rsq02,rinv02);
 421
 422             /* EWALD ELECTROSTATICS */
 423
 424             /* Analytical PME correction */
 425             zeta2            = _mm_mul_ps(beta2,rsq02);
 426             rinv3            = _mm_mul_ps(rinvsq02,rinv02);
 427             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
 428             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
 429             felec            = _mm_mul_ps(qq02,felec);
 430             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
 431             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv02);
 432             velec            = _mm_mul_ps(qq02,velec);
 433
 434             /* Update potential sum for this i atom from the interaction with this j atom. */
 435             velecsum         = _mm_add_ps(velecsum,velec);
 436
 437             fscal            = felec;
 438
 439              /* Update vectorial force */
 440             fix0             = _mm_macc_ps(dx02,fscal,fix0);
 441             fiy0             = _mm_macc_ps(dy02,fscal,fiy0);
 442             fiz0             = _mm_macc_ps(dz02,fscal,fiz0);
 443
 444             fjx2             = _mm_macc_ps(dx02,fscal,fjx2);
 445             fjy2             = _mm_macc_ps(dy02,fscal,fjy2);
 446             fjz2             = _mm_macc_ps(dz02,fscal,fjz2);
 447
 448             /**************************
 449              * CALCULATE INTERACTIONS *
 450              **************************/
 451
 452             r10              = _mm_mul_ps(rsq10,rinv10);
 453
 454             /* EWALD ELECTROSTATICS */
 455
 456             /* Analytical PME correction */
 457             zeta2            = _mm_mul_ps(beta2,rsq10);
 458             rinv3            = _mm_mul_ps(rinvsq10,rinv10);
 459             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
 460             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
 461             felec            = _mm_mul_ps(qq10,felec);
 462             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
 463             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv10);
 464             velec            = _mm_mul_ps(qq10,velec);
 465
 466             /* Update potential sum for this i atom from the interaction with this j atom. */
 467             velecsum         = _mm_add_ps(velecsum,velec);
 468
 469             fscal            = felec;
 470
 471              /* Update vectorial force */
 472             fix1             = _mm_macc_ps(dx10,fscal,fix1);
 473             fiy1             = _mm_macc_ps(dy10,fscal,fiy1);
 474             fiz1             = _mm_macc_ps(dz10,fscal,fiz1);
 475
 476             fjx0             = _mm_macc_ps(dx10,fscal,fjx0);
 477             fjy0             = _mm_macc_ps(dy10,fscal,fjy0);
 478             fjz0             = _mm_macc_ps(dz10,fscal,fjz0);
 479
 480             /**************************
 481              * CALCULATE INTERACTIONS *
 482              **************************/
 483
 484             r11              = _mm_mul_ps(rsq11,rinv11);
 485
 486             /* EWALD ELECTROSTATICS */
 487
 488             /* Analytical PME correction */
 489             zeta2            = _mm_mul_ps(beta2,rsq11);
 490             rinv3            = _mm_mul_ps(rinvsq11,rinv11);
 491             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
 492             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
 493             felec            = _mm_mul_ps(qq11,felec);
 494             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
 495             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv11);
 496             velec            = _mm_mul_ps(qq11,velec);
 497
 498             /* Update potential sum for this i atom from the interaction with this j atom. */
 499             velecsum         = _mm_add_ps(velecsum,velec);
 500
 501             fscal            = felec;
 502
 503              /* Update vectorial force */
 504             fix1             = _mm_macc_ps(dx11,fscal,fix1);
 505             fiy1             = _mm_macc_ps(dy11,fscal,fiy1);
 506             fiz1             = _mm_macc_ps(dz11,fscal,fiz1);
 507
 508             fjx1             = _mm_macc_ps(dx11,fscal,fjx1);
 509             fjy1             = _mm_macc_ps(dy11,fscal,fjy1);
 510             fjz1             = _mm_macc_ps(dz11,fscal,fjz1);
 511
 512             /**************************
 513              * CALCULATE INTERACTIONS *
 514              **************************/
 515
 516             r12              = _mm_mul_ps(rsq12,rinv12);
 517
 518             /* EWALD ELECTROSTATICS */
 519
 520             /* Analytical PME correction */
 521             zeta2            = _mm_mul_ps(beta2,rsq12);
 522             rinv3            = _mm_mul_ps(rinvsq12,rinv12);
 523             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
 524             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
 525             felec            = _mm_mul_ps(qq12,felec);
 526             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
 527             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv12);
 528             velec            = _mm_mul_ps(qq12,velec);
 529
 530             /* Update potential sum for this i atom from the interaction with this j atom. */
 531             velecsum         = _mm_add_ps(velecsum,velec);
 532
 533             fscal            = felec;
 534
 535              /* Update vectorial force */
 536             fix1             = _mm_macc_ps(dx12,fscal,fix1);
 537             fiy1             = _mm_macc_ps(dy12,fscal,fiy1);
 538             fiz1             = _mm_macc_ps(dz12,fscal,fiz1);
 539
 540             fjx2             = _mm_macc_ps(dx12,fscal,fjx2);
 541             fjy2             = _mm_macc_ps(dy12,fscal,fjy2);
 542             fjz2             = _mm_macc_ps(dz12,fscal,fjz2);
 543
 544             /**************************
 545              * CALCULATE INTERACTIONS *
 546              **************************/
 547
 548             r20              = _mm_mul_ps(rsq20,rinv20);
 549
 550             /* EWALD ELECTROSTATICS */
 551
 552             /* Analytical PME correction */
 553             zeta2            = _mm_mul_ps(beta2,rsq20);
 554             rinv3            = _mm_mul_ps(rinvsq20,rinv20);
 555             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
 556             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
 557             felec            = _mm_mul_ps(qq20,felec);
 558             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
 559             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv20);
 560             velec            = _mm_mul_ps(qq20,velec);
 561
 562             /* Update potential sum for this i atom from the interaction with this j atom. */
 563             velecsum         = _mm_add_ps(velecsum,velec);
 564
 565             fscal            = felec;
 566
 567              /* Update vectorial force */
 568             fix2             = _mm_macc_ps(dx20,fscal,fix2);
 569             fiy2             = _mm_macc_ps(dy20,fscal,fiy2);
 570             fiz2             = _mm_macc_ps(dz20,fscal,fiz2);
 571
 572             fjx0             = _mm_macc_ps(dx20,fscal,fjx0);
 573             fjy0             = _mm_macc_ps(dy20,fscal,fjy0);
 574             fjz0             = _mm_macc_ps(dz20,fscal,fjz0);
 575
 576             /**************************
 577              * CALCULATE INTERACTIONS *
 578              **************************/
 579
 580             r21              = _mm_mul_ps(rsq21,rinv21);
 581
 582             /* EWALD ELECTROSTATICS */
 583
 584             /* Analytical PME correction */
 585             zeta2            = _mm_mul_ps(beta2,rsq21);
 586             rinv3            = _mm_mul_ps(rinvsq21,rinv21);
 587             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
 588             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
 589             felec            = _mm_mul_ps(qq21,felec);
 590             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
 591             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv21);
 592             velec            = _mm_mul_ps(qq21,velec);
 593
 594             /* Update potential sum for this i atom from the interaction with this j atom. */
 595             velecsum         = _mm_add_ps(velecsum,velec);
 596
 597             fscal            = felec;
 598
 599              /* Update vectorial force */
 600             fix2             = _mm_macc_ps(dx21,fscal,fix2);
 601             fiy2             = _mm_macc_ps(dy21,fscal,fiy2);
 602             fiz2             = _mm_macc_ps(dz21,fscal,fiz2);
 603
 604             fjx1             = _mm_macc_ps(dx21,fscal,fjx1);
 605             fjy1             = _mm_macc_ps(dy21,fscal,fjy1);
 606             fjz1             = _mm_macc_ps(dz21,fscal,fjz1);
 607
 608             /**************************
 609              * CALCULATE INTERACTIONS *
 610              **************************/
 611
 612             r22              = _mm_mul_ps(rsq22,rinv22);
 613
 614             /* EWALD ELECTROSTATICS */
 615
 616             /* Analytical PME correction */
 617             zeta2            = _mm_mul_ps(beta2,rsq22);
 618             rinv3            = _mm_mul_ps(rinvsq22,rinv22);
 619             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
 620             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
 621             felec            = _mm_mul_ps(qq22,felec);
 622             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
 623             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv22);
 624             velec            = _mm_mul_ps(qq22,velec);
 625
 626             /* Update potential sum for this i atom from the interaction with this j atom. */
 627             velecsum         = _mm_add_ps(velecsum,velec);
 628
 629             fscal            = felec;
 630
 631              /* Update vectorial force */
 632             fix2             = _mm_macc_ps(dx22,fscal,fix2);
 633             fiy2             = _mm_macc_ps(dy22,fscal,fiy2);
 634             fiz2             = _mm_macc_ps(dz22,fscal,fiz2);
 635
 636             fjx2             = _mm_macc_ps(dx22,fscal,fjx2);
 637             fjy2             = _mm_macc_ps(dy22,fscal,fjy2);
 638             fjz2             = _mm_macc_ps(dz22,fscal,fjz2);
 639
 640             fjptrA             = f+j_coord_offsetA;
 641             fjptrB             = f+j_coord_offsetB;
 642             fjptrC             = f+j_coord_offsetC;
 643             fjptrD             = f+j_coord_offsetD;
 644
 645             gmx_mm_decrement_3rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
 646                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 647
 648             /* Inner loop uses 295 flops */
 649         }
 650
 651         if(jidx<j_index_end)
 652         {
 653
 654             /* Get j neighbor index, and coordinate index */
 655             jnrlistA         = jjnr[jidx];
 656             jnrlistB         = jjnr[jidx+1];
 657             jnrlistC         = jjnr[jidx+2];
 658             jnrlistD         = jjnr[jidx+3];
 659             /* Sign of each element will be negative for non-real atoms.
 660              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 661              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 662              */
 663             dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 664             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 665             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 666             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 667             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 668             j_coord_offsetA  = DIM*jnrA;
 669             j_coord_offsetB  = DIM*jnrB;
 670             j_coord_offsetC  = DIM*jnrC;
 671             j_coord_offsetD  = DIM*jnrD;
 672
 673             /* load j atom coordinates */
 674             gmx_mm_load_3rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 675                                               x+j_coord_offsetC,x+j_coord_offsetD,
 676                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 677
 678             /* Calculate displacement vector */
 679             dx00             = _mm_sub_ps(ix0,jx0);
 680             dy00             = _mm_sub_ps(iy0,jy0);
 681             dz00             = _mm_sub_ps(iz0,jz0);
 682             dx01             = _mm_sub_ps(ix0,jx1);
 683             dy01             = _mm_sub_ps(iy0,jy1);
 684             dz01             = _mm_sub_ps(iz0,jz1);
 685             dx02             = _mm_sub_ps(ix0,jx2);
 686             dy02             = _mm_sub_ps(iy0,jy2);
 687             dz02             = _mm_sub_ps(iz0,jz2);
 688             dx10             = _mm_sub_ps(ix1,jx0);
 689             dy10             = _mm_sub_ps(iy1,jy0);
 690             dz10             = _mm_sub_ps(iz1,jz0);
 691             dx11             = _mm_sub_ps(ix1,jx1);
 692             dy11             = _mm_sub_ps(iy1,jy1);
 693             dz11             = _mm_sub_ps(iz1,jz1);
 694             dx12             = _mm_sub_ps(ix1,jx2);
 695             dy12             = _mm_sub_ps(iy1,jy2);
 696             dz12             = _mm_sub_ps(iz1,jz2);
 697             dx20             = _mm_sub_ps(ix2,jx0);
 698             dy20             = _mm_sub_ps(iy2,jy0);
 699             dz20             = _mm_sub_ps(iz2,jz0);
 700             dx21             = _mm_sub_ps(ix2,jx1);
 701             dy21             = _mm_sub_ps(iy2,jy1);
 702             dz21             = _mm_sub_ps(iz2,jz1);
 703             dx22             = _mm_sub_ps(ix2,jx2);
 704             dy22             = _mm_sub_ps(iy2,jy2);
 705             dz22             = _mm_sub_ps(iz2,jz2);
 706
 707             /* Calculate squared distance and things based on it */
 708             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
 709             rsq01            = gmx_mm_calc_rsq_ps(dx01,dy01,dz01);
 710             rsq02            = gmx_mm_calc_rsq_ps(dx02,dy02,dz02);
 711             rsq10            = gmx_mm_calc_rsq_ps(dx10,dy10,dz10);
 712             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
 713             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
 714             rsq20            = gmx_mm_calc_rsq_ps(dx20,dy20,dz20);
 715             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
 716             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
 717
 718             rinv00           = gmx_mm_invsqrt_ps(rsq00);
 719             rinv01           = gmx_mm_invsqrt_ps(rsq01);
 720             rinv02           = gmx_mm_invsqrt_ps(rsq02);
 721             rinv10           = gmx_mm_invsqrt_ps(rsq10);
 722             rinv11           = gmx_mm_invsqrt_ps(rsq11);
 723             rinv12           = gmx_mm_invsqrt_ps(rsq12);
 724             rinv20           = gmx_mm_invsqrt_ps(rsq20);
 725             rinv21           = gmx_mm_invsqrt_ps(rsq21);
 726             rinv22           = gmx_mm_invsqrt_ps(rsq22);
 727
 728             rinvsq00         = _mm_mul_ps(rinv00,rinv00);
 729             rinvsq01         = _mm_mul_ps(rinv01,rinv01);
 730             rinvsq02         = _mm_mul_ps(rinv02,rinv02);
 731             rinvsq10         = _mm_mul_ps(rinv10,rinv10);
 732             rinvsq11         = _mm_mul_ps(rinv11,rinv11);
 733             rinvsq12         = _mm_mul_ps(rinv12,rinv12);
 734             rinvsq20         = _mm_mul_ps(rinv20,rinv20);
 735             rinvsq21         = _mm_mul_ps(rinv21,rinv21);
 736             rinvsq22         = _mm_mul_ps(rinv22,rinv22);
 737
 738             fjx0             = _mm_setzero_ps();
 739             fjy0             = _mm_setzero_ps();
 740             fjz0             = _mm_setzero_ps();
 741             fjx1             = _mm_setzero_ps();
 742             fjy1             = _mm_setzero_ps();
 743             fjz1             = _mm_setzero_ps();
 744             fjx2             = _mm_setzero_ps();
 745             fjy2             = _mm_setzero_ps();
 746             fjz2             = _mm_setzero_ps();
 747
 748             /**************************
 749              * CALCULATE INTERACTIONS *
 750              **************************/
 751
 752             r00              = _mm_mul_ps(rsq00,rinv00);
 753             r00              = _mm_andnot_ps(dummy_mask,r00);
 754
 755             /* Calculate table index by multiplying r with table scale and truncate to integer */
 756             rt               = _mm_mul_ps(r00,vftabscale);
 757             vfitab           = _mm_cvttps_epi32(rt);
 758 #ifdef __XOP__
 759             vfeps            = _mm_frcz_ps(rt);
 760 #else
 761             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
 762 #endif
 763             twovfeps         = _mm_add_ps(vfeps,vfeps);
 764             vfitab           = _mm_slli_epi32(vfitab,3);
 765
 766             /* EWALD ELECTROSTATICS */
 767
 768             /* Analytical PME correction */
 769             zeta2            = _mm_mul_ps(beta2,rsq00);
 770             rinv3            = _mm_mul_ps(rinvsq00,rinv00);
 771             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
 772             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
 773             felec            = _mm_mul_ps(qq00,felec);
 774             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
 775             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv00);
 776             velec            = _mm_mul_ps(qq00,velec);
 777
 778             /* CUBIC SPLINE TABLE DISPERSION */
 779             Y                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
 780             F                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
 781             G                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
 782             H                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
 783             _MM_TRANSPOSE4_PS(Y,F,G,H);
 784             Fp               = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
 785             VV               = _mm_macc_ps(vfeps,Fp,Y);
 786             vvdw6            = _mm_mul_ps(c6_00,VV);
 787             FF               = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
 788             fvdw6            = _mm_mul_ps(c6_00,FF);
 789
 790             /* CUBIC SPLINE TABLE REPULSION */
 791             vfitab           = _mm_add_epi32(vfitab,ifour);
 792             Y                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
 793             F                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
 794             G                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
 795             H                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
 796             _MM_TRANSPOSE4_PS(Y,F,G,H);
 797             Fp               = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
 798             VV               = _mm_macc_ps(vfeps,Fp,Y);
 799             vvdw12           = _mm_mul_ps(c12_00,VV);
 800             FF               = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
 801             fvdw12           = _mm_mul_ps(c12_00,FF);
 802             vvdw             = _mm_add_ps(vvdw12,vvdw6);
 803             fvdw             = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
 804
 805             /* Update potential sum for this i atom from the interaction with this j atom. */
 806             velec            = _mm_andnot_ps(dummy_mask,velec);
 807             velecsum         = _mm_add_ps(velecsum,velec);
 808             vvdw             = _mm_andnot_ps(dummy_mask,vvdw);
 809             vvdwsum          = _mm_add_ps(vvdwsum,vvdw);
 810
 811             fscal            = _mm_add_ps(felec,fvdw);
 812
 813             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 814
 815              /* Update vectorial force */
 816             fix0             = _mm_macc_ps(dx00,fscal,fix0);
 817             fiy0             = _mm_macc_ps(dy00,fscal,fiy0);
 818             fiz0             = _mm_macc_ps(dz00,fscal,fiz0);
 819
 820             fjx0             = _mm_macc_ps(dx00,fscal,fjx0);
 821             fjy0             = _mm_macc_ps(dy00,fscal,fjy0);
 822             fjz0             = _mm_macc_ps(dz00,fscal,fjz0);
 823
 824             /**************************
 825              * CALCULATE INTERACTIONS *
 826              **************************/
 827
 828             r01              = _mm_mul_ps(rsq01,rinv01);
 829             r01              = _mm_andnot_ps(dummy_mask,r01);
 830
 831             /* EWALD ELECTROSTATICS */
 832
 833             /* Analytical PME correction */
 834             zeta2            = _mm_mul_ps(beta2,rsq01);
 835             rinv3            = _mm_mul_ps(rinvsq01,rinv01);
 836             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
 837             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
 838             felec            = _mm_mul_ps(qq01,felec);
 839             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
 840             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv01);
 841             velec            = _mm_mul_ps(qq01,velec);
 842
 843             /* Update potential sum for this i atom from the interaction with this j atom. */
 844             velec            = _mm_andnot_ps(dummy_mask,velec);
 845             velecsum         = _mm_add_ps(velecsum,velec);
 846
 847             fscal            = felec;
 848
 849             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 850
 851              /* Update vectorial force */
 852             fix0             = _mm_macc_ps(dx01,fscal,fix0);
 853             fiy0             = _mm_macc_ps(dy01,fscal,fiy0);
 854             fiz0             = _mm_macc_ps(dz01,fscal,fiz0);
 855
 856             fjx1             = _mm_macc_ps(dx01,fscal,fjx1);
 857             fjy1             = _mm_macc_ps(dy01,fscal,fjy1);
 858             fjz1             = _mm_macc_ps(dz01,fscal,fjz1);
 859
 860             /**************************
 861              * CALCULATE INTERACTIONS *
 862              **************************/
 863
 864             r02              = _mm_mul_ps(rsq02,rinv02);
 865             r02              = _mm_andnot_ps(dummy_mask,r02);
 866
 867             /* EWALD ELECTROSTATICS */
 868
 869             /* Analytical PME correction */
 870             zeta2            = _mm_mul_ps(beta2,rsq02);
 871             rinv3            = _mm_mul_ps(rinvsq02,rinv02);
 872             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
 873             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
 874             felec            = _mm_mul_ps(qq02,felec);
 875             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
 876             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv02);
 877             velec            = _mm_mul_ps(qq02,velec);
 878
 879             /* Update potential sum for this i atom from the interaction with this j atom. */
 880             velec            = _mm_andnot_ps(dummy_mask,velec);
 881             velecsum         = _mm_add_ps(velecsum,velec);
 882
 883             fscal            = felec;
 884
 885             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 886
 887              /* Update vectorial force */
 888             fix0             = _mm_macc_ps(dx02,fscal,fix0);
 889             fiy0             = _mm_macc_ps(dy02,fscal,fiy0);
 890             fiz0             = _mm_macc_ps(dz02,fscal,fiz0);
 891
 892             fjx2             = _mm_macc_ps(dx02,fscal,fjx2);
 893             fjy2             = _mm_macc_ps(dy02,fscal,fjy2);
 894             fjz2             = _mm_macc_ps(dz02,fscal,fjz2);
 895
 896             /**************************
 897              * CALCULATE INTERACTIONS *
 898              **************************/
 899
 900             r10              = _mm_mul_ps(rsq10,rinv10);
 901             r10              = _mm_andnot_ps(dummy_mask,r10);
 902
 903             /* EWALD ELECTROSTATICS */
 904
 905             /* Analytical PME correction */
 906             zeta2            = _mm_mul_ps(beta2,rsq10);
 907             rinv3            = _mm_mul_ps(rinvsq10,rinv10);
 908             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
 909             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
 910             felec            = _mm_mul_ps(qq10,felec);
 911             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
 912             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv10);
 913             velec            = _mm_mul_ps(qq10,velec);
 914
 915             /* Update potential sum for this i atom from the interaction with this j atom. */
 916             velec            = _mm_andnot_ps(dummy_mask,velec);
 917             velecsum         = _mm_add_ps(velecsum,velec);
 918
 919             fscal            = felec;
 920
 921             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 922
 923              /* Update vectorial force */
 924             fix1             = _mm_macc_ps(dx10,fscal,fix1);
 925             fiy1             = _mm_macc_ps(dy10,fscal,fiy1);
 926             fiz1             = _mm_macc_ps(dz10,fscal,fiz1);
 927
 928             fjx0             = _mm_macc_ps(dx10,fscal,fjx0);
 929             fjy0             = _mm_macc_ps(dy10,fscal,fjy0);
 930             fjz0             = _mm_macc_ps(dz10,fscal,fjz0);
 931
 932             /**************************
 933              * CALCULATE INTERACTIONS *
 934              **************************/
 935
 936             r11              = _mm_mul_ps(rsq11,rinv11);
 937             r11              = _mm_andnot_ps(dummy_mask,r11);
 938
 939             /* EWALD ELECTROSTATICS */
 940
 941             /* Analytical PME correction */
 942             zeta2            = _mm_mul_ps(beta2,rsq11);
 943             rinv3            = _mm_mul_ps(rinvsq11,rinv11);
 944             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
 945             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
 946             felec            = _mm_mul_ps(qq11,felec);
 947             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
 948             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv11);
 949             velec            = _mm_mul_ps(qq11,velec);
 950
 951             /* Update potential sum for this i atom from the interaction with this j atom. */
 952             velec            = _mm_andnot_ps(dummy_mask,velec);
 953             velecsum         = _mm_add_ps(velecsum,velec);
 954
 955             fscal            = felec;
 956
 957             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 958
 959              /* Update vectorial force */
 960             fix1             = _mm_macc_ps(dx11,fscal,fix1);
 961             fiy1             = _mm_macc_ps(dy11,fscal,fiy1);
 962             fiz1             = _mm_macc_ps(dz11,fscal,fiz1);
 963
 964             fjx1             = _mm_macc_ps(dx11,fscal,fjx1);
 965             fjy1             = _mm_macc_ps(dy11,fscal,fjy1);
 966             fjz1             = _mm_macc_ps(dz11,fscal,fjz1);
 967
 968             /**************************
 969              * CALCULATE INTERACTIONS *
 970              **************************/
 971
 972             r12              = _mm_mul_ps(rsq12,rinv12);
 973             r12              = _mm_andnot_ps(dummy_mask,r12);
 974
 975             /* EWALD ELECTROSTATICS */
 976
 977             /* Analytical PME correction */
 978             zeta2            = _mm_mul_ps(beta2,rsq12);
 979             rinv3            = _mm_mul_ps(rinvsq12,rinv12);
 980             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
 981             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
 982             felec            = _mm_mul_ps(qq12,felec);
 983             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
 984             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv12);
 985             velec            = _mm_mul_ps(qq12,velec);
 986
 987             /* Update potential sum for this i atom from the interaction with this j atom. */
 988             velec            = _mm_andnot_ps(dummy_mask,velec);
 989             velecsum         = _mm_add_ps(velecsum,velec);
 990
 991             fscal            = felec;
 992
 993             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 994
 995              /* Update vectorial force */
 996             fix1             = _mm_macc_ps(dx12,fscal,fix1);
 997             fiy1             = _mm_macc_ps(dy12,fscal,fiy1);
 998             fiz1             = _mm_macc_ps(dz12,fscal,fiz1);
 999
1000             fjx2             = _mm_macc_ps(dx12,fscal,fjx2);
1001             fjy2             = _mm_macc_ps(dy12,fscal,fjy2);
1002             fjz2             = _mm_macc_ps(dz12,fscal,fjz2);
1003
1004             /**************************
1005              * CALCULATE INTERACTIONS *
1006              **************************/
1007
1008             r20              = _mm_mul_ps(rsq20,rinv20);
1009             r20              = _mm_andnot_ps(dummy_mask,r20);
1010
1011             /* EWALD ELECTROSTATICS */
1012
1013             /* Analytical PME correction */
1014             zeta2            = _mm_mul_ps(beta2,rsq20);
1015             rinv3            = _mm_mul_ps(rinvsq20,rinv20);
1016             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
1017             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
1018             felec            = _mm_mul_ps(qq20,felec);
1019             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
1020             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv20);
1021             velec            = _mm_mul_ps(qq20,velec);
1022
1023             /* Update potential sum for this i atom from the interaction with this j atom. */
1024             velec            = _mm_andnot_ps(dummy_mask,velec);
1025             velecsum         = _mm_add_ps(velecsum,velec);
1026
1027             fscal            = felec;
1028
1029             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1030
1031              /* Update vectorial force */
1032             fix2             = _mm_macc_ps(dx20,fscal,fix2);
1033             fiy2             = _mm_macc_ps(dy20,fscal,fiy2);
1034             fiz2             = _mm_macc_ps(dz20,fscal,fiz2);
1035
1036             fjx0             = _mm_macc_ps(dx20,fscal,fjx0);
1037             fjy0             = _mm_macc_ps(dy20,fscal,fjy0);
1038             fjz0             = _mm_macc_ps(dz20,fscal,fjz0);
1039
1040             /**************************
1041              * CALCULATE INTERACTIONS *
1042              **************************/
1043
1044             r21              = _mm_mul_ps(rsq21,rinv21);
1045             r21              = _mm_andnot_ps(dummy_mask,r21);
1046
1047             /* EWALD ELECTROSTATICS */
1048
1049             /* Analytical PME correction */
1050             zeta2            = _mm_mul_ps(beta2,rsq21);
1051             rinv3            = _mm_mul_ps(rinvsq21,rinv21);
1052             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
1053             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
1054             felec            = _mm_mul_ps(qq21,felec);
1055             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
1056             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv21);
1057             velec            = _mm_mul_ps(qq21,velec);
1058
1059             /* Update potential sum for this i atom from the interaction with this j atom. */
1060             velec            = _mm_andnot_ps(dummy_mask,velec);
1061             velecsum         = _mm_add_ps(velecsum,velec);
1062
1063             fscal            = felec;
1064
1065             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1066
1067              /* Update vectorial force */
1068             fix2             = _mm_macc_ps(dx21,fscal,fix2);
1069             fiy2             = _mm_macc_ps(dy21,fscal,fiy2);
1070             fiz2             = _mm_macc_ps(dz21,fscal,fiz2);
1071
1072             fjx1             = _mm_macc_ps(dx21,fscal,fjx1);
1073             fjy1             = _mm_macc_ps(dy21,fscal,fjy1);
1074             fjz1             = _mm_macc_ps(dz21,fscal,fjz1);
1075
1076             /**************************
1077              * CALCULATE INTERACTIONS *
1078              **************************/
1079
1080             r22              = _mm_mul_ps(rsq22,rinv22);
1081             r22              = _mm_andnot_ps(dummy_mask,r22);
1082
1083             /* EWALD ELECTROSTATICS */
1084
1085             /* Analytical PME correction */
1086             zeta2            = _mm_mul_ps(beta2,rsq22);
1087             rinv3            = _mm_mul_ps(rinvsq22,rinv22);
1088             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
1089             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
1090             felec            = _mm_mul_ps(qq22,felec);
1091             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
1092             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv22);
1093             velec            = _mm_mul_ps(qq22,velec);
1094
1095             /* Update potential sum for this i atom from the interaction with this j atom. */
1096             velec            = _mm_andnot_ps(dummy_mask,velec);
1097             velecsum         = _mm_add_ps(velecsum,velec);
1098
1099             fscal            = felec;
1100
1101             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1102
1103              /* Update vectorial force */
1104             fix2             = _mm_macc_ps(dx22,fscal,fix2);
1105             fiy2             = _mm_macc_ps(dy22,fscal,fiy2);
1106             fiz2             = _mm_macc_ps(dz22,fscal,fiz2);
1107
1108             fjx2             = _mm_macc_ps(dx22,fscal,fjx2);
1109             fjy2             = _mm_macc_ps(dy22,fscal,fjy2);
1110             fjz2             = _mm_macc_ps(dz22,fscal,fjz2);
1111
1112             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1113             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1114             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1115             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1116
1117             gmx_mm_decrement_3rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
1118                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1119
1120             /* Inner loop uses 304 flops */
1121         }
1122
1123         /* End of innermost loop */
1124
1125         gmx_mm_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1126                                               f+i_coord_offset,fshift+i_shift_offset);
1127
1128         ggid                        = gid[iidx];
1129         /* Update potential energies */
1130         gmx_mm_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
1131         gmx_mm_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
1132
1133         /* Increment number of inner iterations */
1134         inneriter                  += j_index_end - j_index_start;
1135
1136         /* Outer loop uses 20 flops */
1137     }
1138
1139     /* Increment number of outer iterations */
1140     outeriter        += nri;
1141
1142     /* Update outer/inner flops */
1143
1144     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_VF,outeriter*20 + inneriter*304);
1145 }
1146 /*
1147  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwCSTab_GeomW3W3_F_avx_128_fma_single
1148  * Electrostatics interaction: Ewald
1149  * VdW interaction:            CubicSplineTable
1150  * Geometry:                   Water3-Water3
1151  * Calculate force/pot:        Force
1152  */
1153 void
1154 nb_kernel_ElecEw_VdwCSTab_GeomW3W3_F_avx_128_fma_single
1155                     (t_nblist                    * gmx_restrict       nlist,
1156                      rvec                        * gmx_restrict          xx,
1157                      rvec                        * gmx_restrict          ff,
1158                      t_forcerec                  * gmx_restrict          fr,
1159                      t_mdatoms                   * gmx_restrict     mdatoms,
1160                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1161                      t_nrnb                      * gmx_restrict        nrnb)
1162 {
1163     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1164      * just 0 for non-waters.
1165      * Suffixes A,B,C,D refer to j loop unrolling done with AVX_128, e.g. for the four different
1166      * jnr indices corresponding to data put in the four positions in the SIMD register.
1167      */
1168     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1169     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1170     int              jnrA,jnrB,jnrC,jnrD;
1171     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1172     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1173     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1174     real             rcutoff_scalar;
1175     real             *shiftvec,*fshift,*x,*f;
1176     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
1177     real             scratch[4*DIM];
1178     __m128           fscal,rcutoff,rcutoff2,jidxall;
1179     int              vdwioffset0;
1180     __m128           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1181     int              vdwioffset1;
1182     __m128           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1183     int              vdwioffset2;
1184     __m128           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1185     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
1186     __m128           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1187     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
1188     __m128           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1189     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
1190     __m128           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1191     __m128           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1192     __m128           dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1193     __m128           dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1194     __m128           dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1195     __m128           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1196     __m128           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1197     __m128           dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1198     __m128           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1199     __m128           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1200     __m128           velec,felec,velecsum,facel,crf,krf,krf2;
1201     real             *charge;
1202     int              nvdwtype;
1203     __m128           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1204     int              *vdwtype;
1205     real             *vdwparam;
1206     __m128           one_sixth   = _mm_set1_ps(1.0/6.0);
1207     __m128           one_twelfth = _mm_set1_ps(1.0/12.0);
1208     __m128i          vfitab;
1209     __m128i          ifour       = _mm_set1_epi32(4);
1210     __m128           rt,vfeps,twovfeps,vftabscale,Y,F,G,H,Fp,VV,FF;
1211     real             *vftab;
1212     __m128i          ewitab;
1213     __m128           ewtabscale,eweps,twoeweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1214     __m128           beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
1215     real             *ewtab;
1216     __m128           dummy_mask,cutoff_mask;
1217     __m128           signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
1218     __m128           one     = _mm_set1_ps(1.0);
1219     __m128           two     = _mm_set1_ps(2.0);
1220     x                = xx[0];
1221     f                = ff[0];
1222
1223     nri              = nlist->nri;
1224     iinr             = nlist->iinr;
1225     jindex           = nlist->jindex;
1226     jjnr             = nlist->jjnr;
1227     shiftidx         = nlist->shift;
1228     gid              = nlist->gid;
1229     shiftvec         = fr->shift_vec[0];
1230     fshift           = fr->fshift[0];
1231     facel            = _mm_set1_ps(fr->epsfac);
1232     charge           = mdatoms->chargeA;
1233     nvdwtype         = fr->ntype;
1234     vdwparam         = fr->nbfp;
1235     vdwtype          = mdatoms->typeA;
1236
1237     vftab            = kernel_data->table_vdw->data;
1238     vftabscale       = _mm_set1_ps(kernel_data->table_vdw->scale);
1239
1240     sh_ewald         = _mm_set1_ps(fr->ic->sh_ewald);
1241     beta             = _mm_set1_ps(fr->ic->ewaldcoeff_q);
1242     beta2            = _mm_mul_ps(beta,beta);
1243     beta3            = _mm_mul_ps(beta,beta2);
1244     ewtab            = fr->ic->tabq_coul_F;
1245     ewtabscale       = _mm_set1_ps(fr->ic->tabq_scale);
1246     ewtabhalfspace   = _mm_set1_ps(0.5/fr->ic->tabq_scale);
1247
1248     /* Setup water-specific parameters */
1249     inr              = nlist->iinr[0];
1250     iq0              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+0]));
1251     iq1              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1]));
1252     iq2              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2]));
1253     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
1254
1255     jq0              = _mm_set1_ps(charge[inr+0]);
1256     jq1              = _mm_set1_ps(charge[inr+1]);
1257     jq2              = _mm_set1_ps(charge[inr+2]);
1258     vdwjidx0A        = 2*vdwtype[inr+0];
1259     qq00             = _mm_mul_ps(iq0,jq0);
1260     c6_00            = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A]);
1261     c12_00           = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A+1]);
1262     qq01             = _mm_mul_ps(iq0,jq1);
1263     qq02             = _mm_mul_ps(iq0,jq2);
1264     qq10             = _mm_mul_ps(iq1,jq0);
1265     qq11             = _mm_mul_ps(iq1,jq1);
1266     qq12             = _mm_mul_ps(iq1,jq2);
1267     qq20             = _mm_mul_ps(iq2,jq0);
1268     qq21             = _mm_mul_ps(iq2,jq1);
1269     qq22             = _mm_mul_ps(iq2,jq2);
1270
1271     /* Avoid stupid compiler warnings */
1272     jnrA = jnrB = jnrC = jnrD = 0;
1273     j_coord_offsetA = 0;
1274     j_coord_offsetB = 0;
1275     j_coord_offsetC = 0;
1276     j_coord_offsetD = 0;
1277
1278     outeriter        = 0;
1279     inneriter        = 0;
1280
1281     for(iidx=0;iidx<4*DIM;iidx++)
1282     {
1283         scratch[iidx] = 0.0;
1284     }
1285
1286     /* Start outer loop over neighborlists */
1287     for(iidx=0; iidx<nri; iidx++)
1288     {
1289         /* Load shift vector for this list */
1290         i_shift_offset   = DIM*shiftidx[iidx];
1291
1292         /* Load limits for loop over neighbors */
1293         j_index_start    = jindex[iidx];
1294         j_index_end      = jindex[iidx+1];
1295
1296         /* Get outer coordinate index */
1297         inr              = iinr[iidx];
1298         i_coord_offset   = DIM*inr;
1299
1300         /* Load i particle coords and add shift vector */
1301         gmx_mm_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1302                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1303
1304         fix0             = _mm_setzero_ps();
1305         fiy0             = _mm_setzero_ps();
1306         fiz0             = _mm_setzero_ps();
1307         fix1             = _mm_setzero_ps();
1308         fiy1             = _mm_setzero_ps();
1309         fiz1             = _mm_setzero_ps();
1310         fix2             = _mm_setzero_ps();
1311         fiy2             = _mm_setzero_ps();
1312         fiz2             = _mm_setzero_ps();
1313
1314         /* Start inner kernel loop */
1315         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
1316         {
1317
1318             /* Get j neighbor index, and coordinate index */
1319             jnrA             = jjnr[jidx];
1320             jnrB             = jjnr[jidx+1];
1321             jnrC             = jjnr[jidx+2];
1322             jnrD             = jjnr[jidx+3];
1323             j_coord_offsetA  = DIM*jnrA;
1324             j_coord_offsetB  = DIM*jnrB;
1325             j_coord_offsetC  = DIM*jnrC;
1326             j_coord_offsetD  = DIM*jnrD;
1327
1328             /* load j atom coordinates */
1329             gmx_mm_load_3rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1330                                               x+j_coord_offsetC,x+j_coord_offsetD,
1331                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1332
1333             /* Calculate displacement vector */
1334             dx00             = _mm_sub_ps(ix0,jx0);
1335             dy00             = _mm_sub_ps(iy0,jy0);
1336             dz00             = _mm_sub_ps(iz0,jz0);
1337             dx01             = _mm_sub_ps(ix0,jx1);
1338             dy01             = _mm_sub_ps(iy0,jy1);
1339             dz01             = _mm_sub_ps(iz0,jz1);
1340             dx02             = _mm_sub_ps(ix0,jx2);
1341             dy02             = _mm_sub_ps(iy0,jy2);
1342             dz02             = _mm_sub_ps(iz0,jz2);
1343             dx10             = _mm_sub_ps(ix1,jx0);
1344             dy10             = _mm_sub_ps(iy1,jy0);
1345             dz10             = _mm_sub_ps(iz1,jz0);
1346             dx11             = _mm_sub_ps(ix1,jx1);
1347             dy11             = _mm_sub_ps(iy1,jy1);
1348             dz11             = _mm_sub_ps(iz1,jz1);
1349             dx12             = _mm_sub_ps(ix1,jx2);
1350             dy12             = _mm_sub_ps(iy1,jy2);
1351             dz12             = _mm_sub_ps(iz1,jz2);
1352             dx20             = _mm_sub_ps(ix2,jx0);
1353             dy20             = _mm_sub_ps(iy2,jy0);
1354             dz20             = _mm_sub_ps(iz2,jz0);
1355             dx21             = _mm_sub_ps(ix2,jx1);
1356             dy21             = _mm_sub_ps(iy2,jy1);
1357             dz21             = _mm_sub_ps(iz2,jz1);
1358             dx22             = _mm_sub_ps(ix2,jx2);
1359             dy22             = _mm_sub_ps(iy2,jy2);
1360             dz22             = _mm_sub_ps(iz2,jz2);
1361
1362             /* Calculate squared distance and things based on it */
1363             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
1364             rsq01            = gmx_mm_calc_rsq_ps(dx01,dy01,dz01);
1365             rsq02            = gmx_mm_calc_rsq_ps(dx02,dy02,dz02);
1366             rsq10            = gmx_mm_calc_rsq_ps(dx10,dy10,dz10);
1367             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
1368             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
1369             rsq20            = gmx_mm_calc_rsq_ps(dx20,dy20,dz20);
1370             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
1371             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
1372
1373             rinv00           = gmx_mm_invsqrt_ps(rsq00);
1374             rinv01           = gmx_mm_invsqrt_ps(rsq01);
1375             rinv02           = gmx_mm_invsqrt_ps(rsq02);
1376             rinv10           = gmx_mm_invsqrt_ps(rsq10);
1377             rinv11           = gmx_mm_invsqrt_ps(rsq11);
1378             rinv12           = gmx_mm_invsqrt_ps(rsq12);
1379             rinv20           = gmx_mm_invsqrt_ps(rsq20);
1380             rinv21           = gmx_mm_invsqrt_ps(rsq21);
1381             rinv22           = gmx_mm_invsqrt_ps(rsq22);
1382
1383             rinvsq00         = _mm_mul_ps(rinv00,rinv00);
1384             rinvsq01         = _mm_mul_ps(rinv01,rinv01);
1385             rinvsq02         = _mm_mul_ps(rinv02,rinv02);
1386             rinvsq10         = _mm_mul_ps(rinv10,rinv10);
1387             rinvsq11         = _mm_mul_ps(rinv11,rinv11);
1388             rinvsq12         = _mm_mul_ps(rinv12,rinv12);
1389             rinvsq20         = _mm_mul_ps(rinv20,rinv20);
1390             rinvsq21         = _mm_mul_ps(rinv21,rinv21);
1391             rinvsq22         = _mm_mul_ps(rinv22,rinv22);
1392
1393             fjx0             = _mm_setzero_ps();
1394             fjy0             = _mm_setzero_ps();
1395             fjz0             = _mm_setzero_ps();
1396             fjx1             = _mm_setzero_ps();
1397             fjy1             = _mm_setzero_ps();
1398             fjz1             = _mm_setzero_ps();
1399             fjx2             = _mm_setzero_ps();
1400             fjy2             = _mm_setzero_ps();
1401             fjz2             = _mm_setzero_ps();
1402
1403             /**************************
1404              * CALCULATE INTERACTIONS *
1405              **************************/
1406
1407             r00              = _mm_mul_ps(rsq00,rinv00);
1408
1409             /* Calculate table index by multiplying r with table scale and truncate to integer */
1410             rt               = _mm_mul_ps(r00,vftabscale);
1411             vfitab           = _mm_cvttps_epi32(rt);
1412 #ifdef __XOP__
1413             vfeps            = _mm_frcz_ps(rt);
1414 #else
1415             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1416 #endif
1417             twovfeps         = _mm_add_ps(vfeps,vfeps);
1418             vfitab           = _mm_slli_epi32(vfitab,3);
1419
1420             /* EWALD ELECTROSTATICS */
1421
1422             /* Analytical PME correction */
1423             zeta2            = _mm_mul_ps(beta2,rsq00);
1424             rinv3            = _mm_mul_ps(rinvsq00,rinv00);
1425             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
1426             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
1427             felec            = _mm_mul_ps(qq00,felec);
1428
1429             /* CUBIC SPLINE TABLE DISPERSION */
1430             Y                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
1431             F                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
1432             G                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
1433             H                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
1434             _MM_TRANSPOSE4_PS(Y,F,G,H);
1435             Fp               = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
1436             FF               = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
1437             fvdw6            = _mm_mul_ps(c6_00,FF);
1438
1439             /* CUBIC SPLINE TABLE REPULSION */
1440             vfitab           = _mm_add_epi32(vfitab,ifour);
1441             Y                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
1442             F                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
1443             G                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
1444             H                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
1445             _MM_TRANSPOSE4_PS(Y,F,G,H);
1446             Fp               = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
1447             FF               = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
1448             fvdw12           = _mm_mul_ps(c12_00,FF);
1449             fvdw             = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
1450
1451             fscal            = _mm_add_ps(felec,fvdw);
1452
1453              /* Update vectorial force */
1454             fix0             = _mm_macc_ps(dx00,fscal,fix0);
1455             fiy0             = _mm_macc_ps(dy00,fscal,fiy0);
1456             fiz0             = _mm_macc_ps(dz00,fscal,fiz0);
1457
1458             fjx0             = _mm_macc_ps(dx00,fscal,fjx0);
1459             fjy0             = _mm_macc_ps(dy00,fscal,fjy0);
1460             fjz0             = _mm_macc_ps(dz00,fscal,fjz0);
1461
1462             /**************************
1463              * CALCULATE INTERACTIONS *
1464              **************************/
1465
1466             r01              = _mm_mul_ps(rsq01,rinv01);
1467
1468             /* EWALD ELECTROSTATICS */
1469
1470             /* Analytical PME correction */
1471             zeta2            = _mm_mul_ps(beta2,rsq01);
1472             rinv3            = _mm_mul_ps(rinvsq01,rinv01);
1473             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
1474             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
1475             felec            = _mm_mul_ps(qq01,felec);
1476
1477             fscal            = felec;
1478
1479              /* Update vectorial force */
1480             fix0             = _mm_macc_ps(dx01,fscal,fix0);
1481             fiy0             = _mm_macc_ps(dy01,fscal,fiy0);
1482             fiz0             = _mm_macc_ps(dz01,fscal,fiz0);
1483
1484             fjx1             = _mm_macc_ps(dx01,fscal,fjx1);
1485             fjy1             = _mm_macc_ps(dy01,fscal,fjy1);
1486             fjz1             = _mm_macc_ps(dz01,fscal,fjz1);
1487
1488             /**************************
1489              * CALCULATE INTERACTIONS *
1490              **************************/
1491
1492             r02              = _mm_mul_ps(rsq02,rinv02);
1493
1494             /* EWALD ELECTROSTATICS */
1495
1496             /* Analytical PME correction */
1497             zeta2            = _mm_mul_ps(beta2,rsq02);
1498             rinv3            = _mm_mul_ps(rinvsq02,rinv02);
1499             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
1500             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
1501             felec            = _mm_mul_ps(qq02,felec);
1502
1503             fscal            = felec;
1504
1505              /* Update vectorial force */
1506             fix0             = _mm_macc_ps(dx02,fscal,fix0);
1507             fiy0             = _mm_macc_ps(dy02,fscal,fiy0);
1508             fiz0             = _mm_macc_ps(dz02,fscal,fiz0);
1509
1510             fjx2             = _mm_macc_ps(dx02,fscal,fjx2);
1511             fjy2             = _mm_macc_ps(dy02,fscal,fjy2);
1512             fjz2             = _mm_macc_ps(dz02,fscal,fjz2);
1513
1514             /**************************
1515              * CALCULATE INTERACTIONS *
1516              **************************/
1517
1518             r10              = _mm_mul_ps(rsq10,rinv10);
1519
1520             /* EWALD ELECTROSTATICS */
1521
1522             /* Analytical PME correction */
1523             zeta2            = _mm_mul_ps(beta2,rsq10);
1524             rinv3            = _mm_mul_ps(rinvsq10,rinv10);
1525             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
1526             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
1527             felec            = _mm_mul_ps(qq10,felec);
1528
1529             fscal            = felec;
1530
1531              /* Update vectorial force */
1532             fix1             = _mm_macc_ps(dx10,fscal,fix1);
1533             fiy1             = _mm_macc_ps(dy10,fscal,fiy1);
1534             fiz1             = _mm_macc_ps(dz10,fscal,fiz1);
1535
1536             fjx0             = _mm_macc_ps(dx10,fscal,fjx0);
1537             fjy0             = _mm_macc_ps(dy10,fscal,fjy0);
1538             fjz0             = _mm_macc_ps(dz10,fscal,fjz0);
1539
1540             /**************************
1541              * CALCULATE INTERACTIONS *
1542              **************************/
1543
1544             r11              = _mm_mul_ps(rsq11,rinv11);
1545
1546             /* EWALD ELECTROSTATICS */
1547
1548             /* Analytical PME correction */
1549             zeta2            = _mm_mul_ps(beta2,rsq11);
1550             rinv3            = _mm_mul_ps(rinvsq11,rinv11);
1551             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
1552             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
1553             felec            = _mm_mul_ps(qq11,felec);
1554
1555             fscal            = felec;
1556
1557              /* Update vectorial force */
1558             fix1             = _mm_macc_ps(dx11,fscal,fix1);
1559             fiy1             = _mm_macc_ps(dy11,fscal,fiy1);
1560             fiz1             = _mm_macc_ps(dz11,fscal,fiz1);
1561
1562             fjx1             = _mm_macc_ps(dx11,fscal,fjx1);
1563             fjy1             = _mm_macc_ps(dy11,fscal,fjy1);
1564             fjz1             = _mm_macc_ps(dz11,fscal,fjz1);
1565
1566             /**************************
1567              * CALCULATE INTERACTIONS *
1568              **************************/
1569
1570             r12              = _mm_mul_ps(rsq12,rinv12);
1571
1572             /* EWALD ELECTROSTATICS */
1573
1574             /* Analytical PME correction */
1575             zeta2            = _mm_mul_ps(beta2,rsq12);
1576             rinv3            = _mm_mul_ps(rinvsq12,rinv12);
1577             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
1578             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
1579             felec            = _mm_mul_ps(qq12,felec);
1580
1581             fscal            = felec;
1582
1583              /* Update vectorial force */
1584             fix1             = _mm_macc_ps(dx12,fscal,fix1);
1585             fiy1             = _mm_macc_ps(dy12,fscal,fiy1);
1586             fiz1             = _mm_macc_ps(dz12,fscal,fiz1);
1587
1588             fjx2             = _mm_macc_ps(dx12,fscal,fjx2);
1589             fjy2             = _mm_macc_ps(dy12,fscal,fjy2);
1590             fjz2             = _mm_macc_ps(dz12,fscal,fjz2);
1591
1592             /**************************
1593              * CALCULATE INTERACTIONS *
1594              **************************/
1595
1596             r20              = _mm_mul_ps(rsq20,rinv20);
1597
1598             /* EWALD ELECTROSTATICS */
1599
1600             /* Analytical PME correction */
1601             zeta2            = _mm_mul_ps(beta2,rsq20);
1602             rinv3            = _mm_mul_ps(rinvsq20,rinv20);
1603             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
1604             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
1605             felec            = _mm_mul_ps(qq20,felec);
1606
1607             fscal            = felec;
1608
1609              /* Update vectorial force */
1610             fix2             = _mm_macc_ps(dx20,fscal,fix2);
1611             fiy2             = _mm_macc_ps(dy20,fscal,fiy2);
1612             fiz2             = _mm_macc_ps(dz20,fscal,fiz2);
1613
1614             fjx0             = _mm_macc_ps(dx20,fscal,fjx0);
1615             fjy0             = _mm_macc_ps(dy20,fscal,fjy0);
1616             fjz0             = _mm_macc_ps(dz20,fscal,fjz0);
1617
1618             /**************************
1619              * CALCULATE INTERACTIONS *
1620              **************************/
1621
1622             r21              = _mm_mul_ps(rsq21,rinv21);
1623
1624             /* EWALD ELECTROSTATICS */
1625
1626             /* Analytical PME correction */
1627             zeta2            = _mm_mul_ps(beta2,rsq21);
1628             rinv3            = _mm_mul_ps(rinvsq21,rinv21);
1629             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
1630             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
1631             felec            = _mm_mul_ps(qq21,felec);
1632
1633             fscal            = felec;
1634
1635              /* Update vectorial force */
1636             fix2             = _mm_macc_ps(dx21,fscal,fix2);
1637             fiy2             = _mm_macc_ps(dy21,fscal,fiy2);
1638             fiz2             = _mm_macc_ps(dz21,fscal,fiz2);
1639
1640             fjx1             = _mm_macc_ps(dx21,fscal,fjx1);
1641             fjy1             = _mm_macc_ps(dy21,fscal,fjy1);
1642             fjz1             = _mm_macc_ps(dz21,fscal,fjz1);
1643
1644             /**************************
1645              * CALCULATE INTERACTIONS *
1646              **************************/
1647
1648             r22              = _mm_mul_ps(rsq22,rinv22);
1649
1650             /* EWALD ELECTROSTATICS */
1651
1652             /* Analytical PME correction */
1653             zeta2            = _mm_mul_ps(beta2,rsq22);
1654             rinv3            = _mm_mul_ps(rinvsq22,rinv22);
1655             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
1656             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
1657             felec            = _mm_mul_ps(qq22,felec);
1658
1659             fscal            = felec;
1660
1661              /* Update vectorial force */
1662             fix2             = _mm_macc_ps(dx22,fscal,fix2);
1663             fiy2             = _mm_macc_ps(dy22,fscal,fiy2);
1664             fiz2             = _mm_macc_ps(dz22,fscal,fiz2);
1665
1666             fjx2             = _mm_macc_ps(dx22,fscal,fjx2);
1667             fjy2             = _mm_macc_ps(dy22,fscal,fjy2);
1668             fjz2             = _mm_macc_ps(dz22,fscal,fjz2);
1669
1670             fjptrA             = f+j_coord_offsetA;
1671             fjptrB             = f+j_coord_offsetB;
1672             fjptrC             = f+j_coord_offsetC;
1673             fjptrD             = f+j_coord_offsetD;
1674
1675             gmx_mm_decrement_3rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
1676                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1677
1678             /* Inner loop uses 278 flops */
1679         }
1680
1681         if(jidx<j_index_end)
1682         {
1683
1684             /* Get j neighbor index, and coordinate index */
1685             jnrlistA         = jjnr[jidx];
1686             jnrlistB         = jjnr[jidx+1];
1687             jnrlistC         = jjnr[jidx+2];
1688             jnrlistD         = jjnr[jidx+3];
1689             /* Sign of each element will be negative for non-real atoms.
1690              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1691              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1692              */
1693             dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
1694             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1695             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1696             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1697             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1698             j_coord_offsetA  = DIM*jnrA;
1699             j_coord_offsetB  = DIM*jnrB;
1700             j_coord_offsetC  = DIM*jnrC;
1701             j_coord_offsetD  = DIM*jnrD;
1702
1703             /* load j atom coordinates */
1704             gmx_mm_load_3rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1705                                               x+j_coord_offsetC,x+j_coord_offsetD,
1706                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1707
1708             /* Calculate displacement vector */
1709             dx00             = _mm_sub_ps(ix0,jx0);
1710             dy00             = _mm_sub_ps(iy0,jy0);
1711             dz00             = _mm_sub_ps(iz0,jz0);
1712             dx01             = _mm_sub_ps(ix0,jx1);
1713             dy01             = _mm_sub_ps(iy0,jy1);
1714             dz01             = _mm_sub_ps(iz0,jz1);
1715             dx02             = _mm_sub_ps(ix0,jx2);
1716             dy02             = _mm_sub_ps(iy0,jy2);
1717             dz02             = _mm_sub_ps(iz0,jz2);
1718             dx10             = _mm_sub_ps(ix1,jx0);
1719             dy10             = _mm_sub_ps(iy1,jy0);
1720             dz10             = _mm_sub_ps(iz1,jz0);
1721             dx11             = _mm_sub_ps(ix1,jx1);
1722             dy11             = _mm_sub_ps(iy1,jy1);
1723             dz11             = _mm_sub_ps(iz1,jz1);
1724             dx12             = _mm_sub_ps(ix1,jx2);
1725             dy12             = _mm_sub_ps(iy1,jy2);
1726             dz12             = _mm_sub_ps(iz1,jz2);
1727             dx20             = _mm_sub_ps(ix2,jx0);
1728             dy20             = _mm_sub_ps(iy2,jy0);
1729             dz20             = _mm_sub_ps(iz2,jz0);
1730             dx21             = _mm_sub_ps(ix2,jx1);
1731             dy21             = _mm_sub_ps(iy2,jy1);
1732             dz21             = _mm_sub_ps(iz2,jz1);
1733             dx22             = _mm_sub_ps(ix2,jx2);
1734             dy22             = _mm_sub_ps(iy2,jy2);
1735             dz22             = _mm_sub_ps(iz2,jz2);
1736
1737             /* Calculate squared distance and things based on it */
1738             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
1739             rsq01            = gmx_mm_calc_rsq_ps(dx01,dy01,dz01);
1740             rsq02            = gmx_mm_calc_rsq_ps(dx02,dy02,dz02);
1741             rsq10            = gmx_mm_calc_rsq_ps(dx10,dy10,dz10);
1742             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
1743             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
1744             rsq20            = gmx_mm_calc_rsq_ps(dx20,dy20,dz20);
1745             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
1746             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
1747
1748             rinv00           = gmx_mm_invsqrt_ps(rsq00);
1749             rinv01           = gmx_mm_invsqrt_ps(rsq01);
1750             rinv02           = gmx_mm_invsqrt_ps(rsq02);
1751             rinv10           = gmx_mm_invsqrt_ps(rsq10);
1752             rinv11           = gmx_mm_invsqrt_ps(rsq11);
1753             rinv12           = gmx_mm_invsqrt_ps(rsq12);
1754             rinv20           = gmx_mm_invsqrt_ps(rsq20);
1755             rinv21           = gmx_mm_invsqrt_ps(rsq21);
1756             rinv22           = gmx_mm_invsqrt_ps(rsq22);
1757
1758             rinvsq00         = _mm_mul_ps(rinv00,rinv00);
1759             rinvsq01         = _mm_mul_ps(rinv01,rinv01);
1760             rinvsq02         = _mm_mul_ps(rinv02,rinv02);
1761             rinvsq10         = _mm_mul_ps(rinv10,rinv10);
1762             rinvsq11         = _mm_mul_ps(rinv11,rinv11);
1763             rinvsq12         = _mm_mul_ps(rinv12,rinv12);
1764             rinvsq20         = _mm_mul_ps(rinv20,rinv20);
1765             rinvsq21         = _mm_mul_ps(rinv21,rinv21);
1766             rinvsq22         = _mm_mul_ps(rinv22,rinv22);
1767
1768             fjx0             = _mm_setzero_ps();
1769             fjy0             = _mm_setzero_ps();
1770             fjz0             = _mm_setzero_ps();
1771             fjx1             = _mm_setzero_ps();
1772             fjy1             = _mm_setzero_ps();
1773             fjz1             = _mm_setzero_ps();
1774             fjx2             = _mm_setzero_ps();
1775             fjy2             = _mm_setzero_ps();
1776             fjz2             = _mm_setzero_ps();
1777
1778             /**************************
1779              * CALCULATE INTERACTIONS *
1780              **************************/
1781
1782             r00              = _mm_mul_ps(rsq00,rinv00);
1783             r00              = _mm_andnot_ps(dummy_mask,r00);
1784
1785             /* Calculate table index by multiplying r with table scale and truncate to integer */
1786             rt               = _mm_mul_ps(r00,vftabscale);
1787             vfitab           = _mm_cvttps_epi32(rt);
1788 #ifdef __XOP__
1789             vfeps            = _mm_frcz_ps(rt);
1790 #else
1791             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1792 #endif
1793             twovfeps         = _mm_add_ps(vfeps,vfeps);
1794             vfitab           = _mm_slli_epi32(vfitab,3);
1795
1796             /* EWALD ELECTROSTATICS */
1797
1798             /* Analytical PME correction */
1799             zeta2            = _mm_mul_ps(beta2,rsq00);
1800             rinv3            = _mm_mul_ps(rinvsq00,rinv00);
1801             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
1802             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
1803             felec            = _mm_mul_ps(qq00,felec);
1804
1805             /* CUBIC SPLINE TABLE DISPERSION */
1806             Y                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
1807             F                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
1808             G                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
1809             H                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
1810             _MM_TRANSPOSE4_PS(Y,F,G,H);
1811             Fp               = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
1812             FF               = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
1813             fvdw6            = _mm_mul_ps(c6_00,FF);
1814
1815             /* CUBIC SPLINE TABLE REPULSION */
1816             vfitab           = _mm_add_epi32(vfitab,ifour);
1817             Y                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
1818             F                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
1819             G                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
1820             H                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
1821             _MM_TRANSPOSE4_PS(Y,F,G,H);
1822             Fp               = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
1823             FF               = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
1824             fvdw12           = _mm_mul_ps(c12_00,FF);
1825             fvdw             = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
1826
1827             fscal            = _mm_add_ps(felec,fvdw);
1828
1829             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1830
1831              /* Update vectorial force */
1832             fix0             = _mm_macc_ps(dx00,fscal,fix0);
1833             fiy0             = _mm_macc_ps(dy00,fscal,fiy0);
1834             fiz0             = _mm_macc_ps(dz00,fscal,fiz0);
1835
1836             fjx0             = _mm_macc_ps(dx00,fscal,fjx0);
1837             fjy0             = _mm_macc_ps(dy00,fscal,fjy0);
1838             fjz0             = _mm_macc_ps(dz00,fscal,fjz0);
1839
1840             /**************************
1841              * CALCULATE INTERACTIONS *
1842              **************************/
1843
1844             r01              = _mm_mul_ps(rsq01,rinv01);
1845             r01              = _mm_andnot_ps(dummy_mask,r01);
1846
1847             /* EWALD ELECTROSTATICS */
1848
1849             /* Analytical PME correction */
1850             zeta2            = _mm_mul_ps(beta2,rsq01);
1851             rinv3            = _mm_mul_ps(rinvsq01,rinv01);
1852             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
1853             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
1854             felec            = _mm_mul_ps(qq01,felec);
1855
1856             fscal            = felec;
1857
1858             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1859
1860              /* Update vectorial force */
1861             fix0             = _mm_macc_ps(dx01,fscal,fix0);
1862             fiy0             = _mm_macc_ps(dy01,fscal,fiy0);
1863             fiz0             = _mm_macc_ps(dz01,fscal,fiz0);
1864
1865             fjx1             = _mm_macc_ps(dx01,fscal,fjx1);
1866             fjy1             = _mm_macc_ps(dy01,fscal,fjy1);
1867             fjz1             = _mm_macc_ps(dz01,fscal,fjz1);
1868
1869             /**************************
1870              * CALCULATE INTERACTIONS *
1871              **************************/
1872
1873             r02              = _mm_mul_ps(rsq02,rinv02);
1874             r02              = _mm_andnot_ps(dummy_mask,r02);
1875
1876             /* EWALD ELECTROSTATICS */
1877
1878             /* Analytical PME correction */
1879             zeta2            = _mm_mul_ps(beta2,rsq02);
1880             rinv3            = _mm_mul_ps(rinvsq02,rinv02);
1881             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
1882             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
1883             felec            = _mm_mul_ps(qq02,felec);
1884
1885             fscal            = felec;
1886
1887             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1888
1889              /* Update vectorial force */
1890             fix0             = _mm_macc_ps(dx02,fscal,fix0);
1891             fiy0             = _mm_macc_ps(dy02,fscal,fiy0);
1892             fiz0             = _mm_macc_ps(dz02,fscal,fiz0);
1893
1894             fjx2             = _mm_macc_ps(dx02,fscal,fjx2);
1895             fjy2             = _mm_macc_ps(dy02,fscal,fjy2);
1896             fjz2             = _mm_macc_ps(dz02,fscal,fjz2);
1897
1898             /**************************
1899              * CALCULATE INTERACTIONS *
1900              **************************/
1901
1902             r10              = _mm_mul_ps(rsq10,rinv10);
1903             r10              = _mm_andnot_ps(dummy_mask,r10);
1904
1905             /* EWALD ELECTROSTATICS */
1906
1907             /* Analytical PME correction */
1908             zeta2            = _mm_mul_ps(beta2,rsq10);
1909             rinv3            = _mm_mul_ps(rinvsq10,rinv10);
1910             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
1911             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
1912             felec            = _mm_mul_ps(qq10,felec);
1913
1914             fscal            = felec;
1915
1916             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1917
1918              /* Update vectorial force */
1919             fix1             = _mm_macc_ps(dx10,fscal,fix1);
1920             fiy1             = _mm_macc_ps(dy10,fscal,fiy1);
1921             fiz1             = _mm_macc_ps(dz10,fscal,fiz1);
1922
1923             fjx0             = _mm_macc_ps(dx10,fscal,fjx0);
1924             fjy0             = _mm_macc_ps(dy10,fscal,fjy0);
1925             fjz0             = _mm_macc_ps(dz10,fscal,fjz0);
1926
1927             /**************************
1928              * CALCULATE INTERACTIONS *
1929              **************************/
1930
1931             r11              = _mm_mul_ps(rsq11,rinv11);
1932             r11              = _mm_andnot_ps(dummy_mask,r11);
1933
1934             /* EWALD ELECTROSTATICS */
1935
1936             /* Analytical PME correction */
1937             zeta2            = _mm_mul_ps(beta2,rsq11);
1938             rinv3            = _mm_mul_ps(rinvsq11,rinv11);
1939             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
1940             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
1941             felec            = _mm_mul_ps(qq11,felec);
1942
1943             fscal            = felec;
1944
1945             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1946
1947              /* Update vectorial force */
1948             fix1             = _mm_macc_ps(dx11,fscal,fix1);
1949             fiy1             = _mm_macc_ps(dy11,fscal,fiy1);
1950             fiz1             = _mm_macc_ps(dz11,fscal,fiz1);
1951
1952             fjx1             = _mm_macc_ps(dx11,fscal,fjx1);
1953             fjy1             = _mm_macc_ps(dy11,fscal,fjy1);
1954             fjz1             = _mm_macc_ps(dz11,fscal,fjz1);
1955
1956             /**************************
1957              * CALCULATE INTERACTIONS *
1958              **************************/
1959
1960             r12              = _mm_mul_ps(rsq12,rinv12);
1961             r12              = _mm_andnot_ps(dummy_mask,r12);
1962
1963             /* EWALD ELECTROSTATICS */
1964
1965             /* Analytical PME correction */
1966             zeta2            = _mm_mul_ps(beta2,rsq12);
1967             rinv3            = _mm_mul_ps(rinvsq12,rinv12);
1968             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
1969             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
1970             felec            = _mm_mul_ps(qq12,felec);
1971
1972             fscal            = felec;
1973
1974             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1975
1976              /* Update vectorial force */
1977             fix1             = _mm_macc_ps(dx12,fscal,fix1);
1978             fiy1             = _mm_macc_ps(dy12,fscal,fiy1);
1979             fiz1             = _mm_macc_ps(dz12,fscal,fiz1);
1980
1981             fjx2             = _mm_macc_ps(dx12,fscal,fjx2);
1982             fjy2             = _mm_macc_ps(dy12,fscal,fjy2);
1983             fjz2             = _mm_macc_ps(dz12,fscal,fjz2);
1984
1985             /**************************
1986              * CALCULATE INTERACTIONS *
1987              **************************/
1988
1989             r20              = _mm_mul_ps(rsq20,rinv20);
1990             r20              = _mm_andnot_ps(dummy_mask,r20);
1991
1992             /* EWALD ELECTROSTATICS */
1993
1994             /* Analytical PME correction */
1995             zeta2            = _mm_mul_ps(beta2,rsq20);
1996             rinv3            = _mm_mul_ps(rinvsq20,rinv20);
1997             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
1998             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
1999             felec            = _mm_mul_ps(qq20,felec);
2000
2001             fscal            = felec;
2002
2003             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2004
2005              /* Update vectorial force */
2006             fix2             = _mm_macc_ps(dx20,fscal,fix2);
2007             fiy2             = _mm_macc_ps(dy20,fscal,fiy2);
2008             fiz2             = _mm_macc_ps(dz20,fscal,fiz2);
2009
2010             fjx0             = _mm_macc_ps(dx20,fscal,fjx0);
2011             fjy0             = _mm_macc_ps(dy20,fscal,fjy0);
2012             fjz0             = _mm_macc_ps(dz20,fscal,fjz0);
2013
2014             /**************************
2015              * CALCULATE INTERACTIONS *
2016              **************************/
2017
2018             r21              = _mm_mul_ps(rsq21,rinv21);
2019             r21              = _mm_andnot_ps(dummy_mask,r21);
2020
2021             /* EWALD ELECTROSTATICS */
2022
2023             /* Analytical PME correction */
2024             zeta2            = _mm_mul_ps(beta2,rsq21);
2025             rinv3            = _mm_mul_ps(rinvsq21,rinv21);
2026             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
2027             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
2028             felec            = _mm_mul_ps(qq21,felec);
2029
2030             fscal            = felec;
2031
2032             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2033
2034              /* Update vectorial force */
2035             fix2             = _mm_macc_ps(dx21,fscal,fix2);
2036             fiy2             = _mm_macc_ps(dy21,fscal,fiy2);
2037             fiz2             = _mm_macc_ps(dz21,fscal,fiz2);
2038
2039             fjx1             = _mm_macc_ps(dx21,fscal,fjx1);
2040             fjy1             = _mm_macc_ps(dy21,fscal,fjy1);
2041             fjz1             = _mm_macc_ps(dz21,fscal,fjz1);
2042
2043             /**************************
2044              * CALCULATE INTERACTIONS *
2045              **************************/
2046
2047             r22              = _mm_mul_ps(rsq22,rinv22);
2048             r22              = _mm_andnot_ps(dummy_mask,r22);
2049
2050             /* EWALD ELECTROSTATICS */
2051
2052             /* Analytical PME correction */
2053             zeta2            = _mm_mul_ps(beta2,rsq22);
2054             rinv3            = _mm_mul_ps(rinvsq22,rinv22);
2055             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
2056             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
2057             felec            = _mm_mul_ps(qq22,felec);
2058
2059             fscal            = felec;
2060
2061             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2062
2063              /* Update vectorial force */
2064             fix2             = _mm_macc_ps(dx22,fscal,fix2);
2065             fiy2             = _mm_macc_ps(dy22,fscal,fiy2);
2066             fiz2             = _mm_macc_ps(dz22,fscal,fiz2);
2067
2068             fjx2             = _mm_macc_ps(dx22,fscal,fjx2);
2069             fjy2             = _mm_macc_ps(dy22,fscal,fjy2);
2070             fjz2             = _mm_macc_ps(dz22,fscal,fjz2);
2071
2072             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
2073             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
2074             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
2075             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
2076
2077             gmx_mm_decrement_3rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
2078                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
2079
2080             /* Inner loop uses 287 flops */
2081         }
2082
2083         /* End of innermost loop */
2084
2085         gmx_mm_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
2086                                               f+i_coord_offset,fshift+i_shift_offset);
2087
2088         /* Increment number of inner iterations */
2089         inneriter                  += j_index_end - j_index_start;
2090
2091         /* Outer loop uses 18 flops */
2092     }
2093
2094     /* Increment number of outer iterations */
2095     outeriter        += nri;
2096
2097     /* Update outer/inner flops */
2098
2099     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_F,outeriter*18 + inneriter*287);
2100 }