src/gromacs/gmxlib/nonbonded/nb_kernel_sse2_double/nb_kernel_ElecEw_VdwNone_GeomW3W3_sse2_double.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013, 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 sse2_double kernel generator.
  37  */
  38 #ifdef HAVE_CONFIG_H
  39 #include <config.h>
  40 #endif
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "types/simple.h"
  46 #include "vec.h"
  47 #include "nrnb.h"
  48
  49 #include "gromacs/simd/math_x86_sse2_double.h"
  50 #include "kernelutil_x86_sse2_double.h"
  51
  52 /*
  53  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwNone_GeomW3W3_VF_sse2_double
  54  * Electrostatics interaction: Ewald
  55  * VdW interaction:            None
  56  * Geometry:                   Water3-Water3
  57  * Calculate force/pot:        PotentialAndForce
  58  */
  59 void
  60 nb_kernel_ElecEw_VdwNone_GeomW3W3_VF_sse2_double
  61                     (t_nblist                    * gmx_restrict       nlist,
  62                      rvec                        * gmx_restrict          xx,
  63                      rvec                        * gmx_restrict          ff,
  64                      t_forcerec                  * gmx_restrict          fr,
  65                      t_mdatoms                   * gmx_restrict     mdatoms,
  66                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  67                      t_nrnb                      * gmx_restrict        nrnb)
  68 {
  69     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  70      * just 0 for non-waters.
  71      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
  72      * jnr indices corresponding to data put in the four positions in the SIMD register.
  73      */
  74     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  75     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  76     int              jnrA,jnrB;
  77     int              j_coord_offsetA,j_coord_offsetB;
  78     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  79     real             rcutoff_scalar;
  80     real             *shiftvec,*fshift,*x,*f;
  81     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  82     int              vdwioffset0;
  83     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  84     int              vdwioffset1;
  85     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  86     int              vdwioffset2;
  87     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  88     int              vdwjidx0A,vdwjidx0B;
  89     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  90     int              vdwjidx1A,vdwjidx1B;
  91     __m128d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  92     int              vdwjidx2A,vdwjidx2B;
  93     __m128d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  94     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  95     __m128d          dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
  96     __m128d          dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
  97     __m128d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  98     __m128d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
  99     __m128d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
 100     __m128d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 101     __m128d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 102     __m128d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 103     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
 104     real             *charge;
 105     __m128i          ewitab;
 106     __m128d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 107     real             *ewtab;
 108     __m128d          dummy_mask,cutoff_mask;
 109     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 110     __m128d          one     = _mm_set1_pd(1.0);
 111     __m128d          two     = _mm_set1_pd(2.0);
 112     x                = xx[0];
 113     f                = ff[0];
 114
 115     nri              = nlist->nri;
 116     iinr             = nlist->iinr;
 117     jindex           = nlist->jindex;
 118     jjnr             = nlist->jjnr;
 119     shiftidx         = nlist->shift;
 120     gid              = nlist->gid;
 121     shiftvec         = fr->shift_vec[0];
 122     fshift           = fr->fshift[0];
 123     facel            = _mm_set1_pd(fr->epsfac);
 124     charge           = mdatoms->chargeA;
 125
 126     sh_ewald         = _mm_set1_pd(fr->ic->sh_ewald);
 127     ewtab            = fr->ic->tabq_coul_FDV0;
 128     ewtabscale       = _mm_set1_pd(fr->ic->tabq_scale);
 129     ewtabhalfspace   = _mm_set1_pd(0.5/fr->ic->tabq_scale);
 130
 131     /* Setup water-specific parameters */
 132     inr              = nlist->iinr[0];
 133     iq0              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
 134     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
 135     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
 136
 137     jq0              = _mm_set1_pd(charge[inr+0]);
 138     jq1              = _mm_set1_pd(charge[inr+1]);
 139     jq2              = _mm_set1_pd(charge[inr+2]);
 140     qq00             = _mm_mul_pd(iq0,jq0);
 141     qq01             = _mm_mul_pd(iq0,jq1);
 142     qq02             = _mm_mul_pd(iq0,jq2);
 143     qq10             = _mm_mul_pd(iq1,jq0);
 144     qq11             = _mm_mul_pd(iq1,jq1);
 145     qq12             = _mm_mul_pd(iq1,jq2);
 146     qq20             = _mm_mul_pd(iq2,jq0);
 147     qq21             = _mm_mul_pd(iq2,jq1);
 148     qq22             = _mm_mul_pd(iq2,jq2);
 149
 150     /* Avoid stupid compiler warnings */
 151     jnrA = jnrB = 0;
 152     j_coord_offsetA = 0;
 153     j_coord_offsetB = 0;
 154
 155     outeriter        = 0;
 156     inneriter        = 0;
 157
 158     /* Start outer loop over neighborlists */
 159     for(iidx=0; iidx<nri; iidx++)
 160     {
 161         /* Load shift vector for this list */
 162         i_shift_offset   = DIM*shiftidx[iidx];
 163
 164         /* Load limits for loop over neighbors */
 165         j_index_start    = jindex[iidx];
 166         j_index_end      = jindex[iidx+1];
 167
 168         /* Get outer coordinate index */
 169         inr              = iinr[iidx];
 170         i_coord_offset   = DIM*inr;
 171
 172         /* Load i particle coords and add shift vector */
 173         gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 174                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 175
 176         fix0             = _mm_setzero_pd();
 177         fiy0             = _mm_setzero_pd();
 178         fiz0             = _mm_setzero_pd();
 179         fix1             = _mm_setzero_pd();
 180         fiy1             = _mm_setzero_pd();
 181         fiz1             = _mm_setzero_pd();
 182         fix2             = _mm_setzero_pd();
 183         fiy2             = _mm_setzero_pd();
 184         fiz2             = _mm_setzero_pd();
 185
 186         /* Reset potential sums */
 187         velecsum         = _mm_setzero_pd();
 188
 189         /* Start inner kernel loop */
 190         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 191         {
 192
 193             /* Get j neighbor index, and coordinate index */
 194             jnrA             = jjnr[jidx];
 195             jnrB             = jjnr[jidx+1];
 196             j_coord_offsetA  = DIM*jnrA;
 197             j_coord_offsetB  = DIM*jnrB;
 198
 199             /* load j atom coordinates */
 200             gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 201                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 202
 203             /* Calculate displacement vector */
 204             dx00             = _mm_sub_pd(ix0,jx0);
 205             dy00             = _mm_sub_pd(iy0,jy0);
 206             dz00             = _mm_sub_pd(iz0,jz0);
 207             dx01             = _mm_sub_pd(ix0,jx1);
 208             dy01             = _mm_sub_pd(iy0,jy1);
 209             dz01             = _mm_sub_pd(iz0,jz1);
 210             dx02             = _mm_sub_pd(ix0,jx2);
 211             dy02             = _mm_sub_pd(iy0,jy2);
 212             dz02             = _mm_sub_pd(iz0,jz2);
 213             dx10             = _mm_sub_pd(ix1,jx0);
 214             dy10             = _mm_sub_pd(iy1,jy0);
 215             dz10             = _mm_sub_pd(iz1,jz0);
 216             dx11             = _mm_sub_pd(ix1,jx1);
 217             dy11             = _mm_sub_pd(iy1,jy1);
 218             dz11             = _mm_sub_pd(iz1,jz1);
 219             dx12             = _mm_sub_pd(ix1,jx2);
 220             dy12             = _mm_sub_pd(iy1,jy2);
 221             dz12             = _mm_sub_pd(iz1,jz2);
 222             dx20             = _mm_sub_pd(ix2,jx0);
 223             dy20             = _mm_sub_pd(iy2,jy0);
 224             dz20             = _mm_sub_pd(iz2,jz0);
 225             dx21             = _mm_sub_pd(ix2,jx1);
 226             dy21             = _mm_sub_pd(iy2,jy1);
 227             dz21             = _mm_sub_pd(iz2,jz1);
 228             dx22             = _mm_sub_pd(ix2,jx2);
 229             dy22             = _mm_sub_pd(iy2,jy2);
 230             dz22             = _mm_sub_pd(iz2,jz2);
 231
 232             /* Calculate squared distance and things based on it */
 233             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 234             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
 235             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
 236             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 237             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 238             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 239             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 240             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 241             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 242
 243             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 244             rinv01           = gmx_mm_invsqrt_pd(rsq01);
 245             rinv02           = gmx_mm_invsqrt_pd(rsq02);
 246             rinv10           = gmx_mm_invsqrt_pd(rsq10);
 247             rinv11           = gmx_mm_invsqrt_pd(rsq11);
 248             rinv12           = gmx_mm_invsqrt_pd(rsq12);
 249             rinv20           = gmx_mm_invsqrt_pd(rsq20);
 250             rinv21           = gmx_mm_invsqrt_pd(rsq21);
 251             rinv22           = gmx_mm_invsqrt_pd(rsq22);
 252
 253             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 254             rinvsq01         = _mm_mul_pd(rinv01,rinv01);
 255             rinvsq02         = _mm_mul_pd(rinv02,rinv02);
 256             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
 257             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
 258             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
 259             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
 260             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
 261             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
 262
 263             fjx0             = _mm_setzero_pd();
 264             fjy0             = _mm_setzero_pd();
 265             fjz0             = _mm_setzero_pd();
 266             fjx1             = _mm_setzero_pd();
 267             fjy1             = _mm_setzero_pd();
 268             fjz1             = _mm_setzero_pd();
 269             fjx2             = _mm_setzero_pd();
 270             fjy2             = _mm_setzero_pd();
 271             fjz2             = _mm_setzero_pd();
 272
 273             /**************************
 274              * CALCULATE INTERACTIONS *
 275              **************************/
 276
 277             r00              = _mm_mul_pd(rsq00,rinv00);
 278
 279             /* EWALD ELECTROSTATICS */
 280
 281             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 282             ewrt             = _mm_mul_pd(r00,ewtabscale);
 283             ewitab           = _mm_cvttpd_epi32(ewrt);
 284             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 285             ewitab           = _mm_slli_epi32(ewitab,2);
 286             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 287             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 288             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 289             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 290             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 291             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 292             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 293             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 294             velec            = _mm_mul_pd(qq00,_mm_sub_pd(rinv00,velec));
 295             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
 296
 297             /* Update potential sum for this i atom from the interaction with this j atom. */
 298             velecsum         = _mm_add_pd(velecsum,velec);
 299
 300             fscal            = felec;
 301
 302             /* Calculate temporary vectorial force */
 303             tx               = _mm_mul_pd(fscal,dx00);
 304             ty               = _mm_mul_pd(fscal,dy00);
 305             tz               = _mm_mul_pd(fscal,dz00);
 306
 307             /* Update vectorial force */
 308             fix0             = _mm_add_pd(fix0,tx);
 309             fiy0             = _mm_add_pd(fiy0,ty);
 310             fiz0             = _mm_add_pd(fiz0,tz);
 311
 312             fjx0             = _mm_add_pd(fjx0,tx);
 313             fjy0             = _mm_add_pd(fjy0,ty);
 314             fjz0             = _mm_add_pd(fjz0,tz);
 315
 316             /**************************
 317              * CALCULATE INTERACTIONS *
 318              **************************/
 319
 320             r01              = _mm_mul_pd(rsq01,rinv01);
 321
 322             /* EWALD ELECTROSTATICS */
 323
 324             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 325             ewrt             = _mm_mul_pd(r01,ewtabscale);
 326             ewitab           = _mm_cvttpd_epi32(ewrt);
 327             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 328             ewitab           = _mm_slli_epi32(ewitab,2);
 329             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 330             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 331             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 332             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 333             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 334             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 335             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 336             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 337             velec            = _mm_mul_pd(qq01,_mm_sub_pd(rinv01,velec));
 338             felec            = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
 339
 340             /* Update potential sum for this i atom from the interaction with this j atom. */
 341             velecsum         = _mm_add_pd(velecsum,velec);
 342
 343             fscal            = felec;
 344
 345             /* Calculate temporary vectorial force */
 346             tx               = _mm_mul_pd(fscal,dx01);
 347             ty               = _mm_mul_pd(fscal,dy01);
 348             tz               = _mm_mul_pd(fscal,dz01);
 349
 350             /* Update vectorial force */
 351             fix0             = _mm_add_pd(fix0,tx);
 352             fiy0             = _mm_add_pd(fiy0,ty);
 353             fiz0             = _mm_add_pd(fiz0,tz);
 354
 355             fjx1             = _mm_add_pd(fjx1,tx);
 356             fjy1             = _mm_add_pd(fjy1,ty);
 357             fjz1             = _mm_add_pd(fjz1,tz);
 358
 359             /**************************
 360              * CALCULATE INTERACTIONS *
 361              **************************/
 362
 363             r02              = _mm_mul_pd(rsq02,rinv02);
 364
 365             /* EWALD ELECTROSTATICS */
 366
 367             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 368             ewrt             = _mm_mul_pd(r02,ewtabscale);
 369             ewitab           = _mm_cvttpd_epi32(ewrt);
 370             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 371             ewitab           = _mm_slli_epi32(ewitab,2);
 372             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 373             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 374             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 375             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 376             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 377             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 378             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 379             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 380             velec            = _mm_mul_pd(qq02,_mm_sub_pd(rinv02,velec));
 381             felec            = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
 382
 383             /* Update potential sum for this i atom from the interaction with this j atom. */
 384             velecsum         = _mm_add_pd(velecsum,velec);
 385
 386             fscal            = felec;
 387
 388             /* Calculate temporary vectorial force */
 389             tx               = _mm_mul_pd(fscal,dx02);
 390             ty               = _mm_mul_pd(fscal,dy02);
 391             tz               = _mm_mul_pd(fscal,dz02);
 392
 393             /* Update vectorial force */
 394             fix0             = _mm_add_pd(fix0,tx);
 395             fiy0             = _mm_add_pd(fiy0,ty);
 396             fiz0             = _mm_add_pd(fiz0,tz);
 397
 398             fjx2             = _mm_add_pd(fjx2,tx);
 399             fjy2             = _mm_add_pd(fjy2,ty);
 400             fjz2             = _mm_add_pd(fjz2,tz);
 401
 402             /**************************
 403              * CALCULATE INTERACTIONS *
 404              **************************/
 405
 406             r10              = _mm_mul_pd(rsq10,rinv10);
 407
 408             /* EWALD ELECTROSTATICS */
 409
 410             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 411             ewrt             = _mm_mul_pd(r10,ewtabscale);
 412             ewitab           = _mm_cvttpd_epi32(ewrt);
 413             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 414             ewitab           = _mm_slli_epi32(ewitab,2);
 415             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 416             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 417             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 418             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 419             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 420             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 421             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 422             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 423             velec            = _mm_mul_pd(qq10,_mm_sub_pd(rinv10,velec));
 424             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
 425
 426             /* Update potential sum for this i atom from the interaction with this j atom. */
 427             velecsum         = _mm_add_pd(velecsum,velec);
 428
 429             fscal            = felec;
 430
 431             /* Calculate temporary vectorial force */
 432             tx               = _mm_mul_pd(fscal,dx10);
 433             ty               = _mm_mul_pd(fscal,dy10);
 434             tz               = _mm_mul_pd(fscal,dz10);
 435
 436             /* Update vectorial force */
 437             fix1             = _mm_add_pd(fix1,tx);
 438             fiy1             = _mm_add_pd(fiy1,ty);
 439             fiz1             = _mm_add_pd(fiz1,tz);
 440
 441             fjx0             = _mm_add_pd(fjx0,tx);
 442             fjy0             = _mm_add_pd(fjy0,ty);
 443             fjz0             = _mm_add_pd(fjz0,tz);
 444
 445             /**************************
 446              * CALCULATE INTERACTIONS *
 447              **************************/
 448
 449             r11              = _mm_mul_pd(rsq11,rinv11);
 450
 451             /* EWALD ELECTROSTATICS */
 452
 453             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 454             ewrt             = _mm_mul_pd(r11,ewtabscale);
 455             ewitab           = _mm_cvttpd_epi32(ewrt);
 456             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 457             ewitab           = _mm_slli_epi32(ewitab,2);
 458             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 459             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 460             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 461             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 462             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 463             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 464             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 465             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 466             velec            = _mm_mul_pd(qq11,_mm_sub_pd(rinv11,velec));
 467             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
 468
 469             /* Update potential sum for this i atom from the interaction with this j atom. */
 470             velecsum         = _mm_add_pd(velecsum,velec);
 471
 472             fscal            = felec;
 473
 474             /* Calculate temporary vectorial force */
 475             tx               = _mm_mul_pd(fscal,dx11);
 476             ty               = _mm_mul_pd(fscal,dy11);
 477             tz               = _mm_mul_pd(fscal,dz11);
 478
 479             /* Update vectorial force */
 480             fix1             = _mm_add_pd(fix1,tx);
 481             fiy1             = _mm_add_pd(fiy1,ty);
 482             fiz1             = _mm_add_pd(fiz1,tz);
 483
 484             fjx1             = _mm_add_pd(fjx1,tx);
 485             fjy1             = _mm_add_pd(fjy1,ty);
 486             fjz1             = _mm_add_pd(fjz1,tz);
 487
 488             /**************************
 489              * CALCULATE INTERACTIONS *
 490              **************************/
 491
 492             r12              = _mm_mul_pd(rsq12,rinv12);
 493
 494             /* EWALD ELECTROSTATICS */
 495
 496             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 497             ewrt             = _mm_mul_pd(r12,ewtabscale);
 498             ewitab           = _mm_cvttpd_epi32(ewrt);
 499             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 500             ewitab           = _mm_slli_epi32(ewitab,2);
 501             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 502             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 503             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 504             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 505             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 506             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 507             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 508             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 509             velec            = _mm_mul_pd(qq12,_mm_sub_pd(rinv12,velec));
 510             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
 511
 512             /* Update potential sum for this i atom from the interaction with this j atom. */
 513             velecsum         = _mm_add_pd(velecsum,velec);
 514
 515             fscal            = felec;
 516
 517             /* Calculate temporary vectorial force */
 518             tx               = _mm_mul_pd(fscal,dx12);
 519             ty               = _mm_mul_pd(fscal,dy12);
 520             tz               = _mm_mul_pd(fscal,dz12);
 521
 522             /* Update vectorial force */
 523             fix1             = _mm_add_pd(fix1,tx);
 524             fiy1             = _mm_add_pd(fiy1,ty);
 525             fiz1             = _mm_add_pd(fiz1,tz);
 526
 527             fjx2             = _mm_add_pd(fjx2,tx);
 528             fjy2             = _mm_add_pd(fjy2,ty);
 529             fjz2             = _mm_add_pd(fjz2,tz);
 530
 531             /**************************
 532              * CALCULATE INTERACTIONS *
 533              **************************/
 534
 535             r20              = _mm_mul_pd(rsq20,rinv20);
 536
 537             /* EWALD ELECTROSTATICS */
 538
 539             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 540             ewrt             = _mm_mul_pd(r20,ewtabscale);
 541             ewitab           = _mm_cvttpd_epi32(ewrt);
 542             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 543             ewitab           = _mm_slli_epi32(ewitab,2);
 544             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 545             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 546             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 547             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 548             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 549             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 550             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 551             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 552             velec            = _mm_mul_pd(qq20,_mm_sub_pd(rinv20,velec));
 553             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
 554
 555             /* Update potential sum for this i atom from the interaction with this j atom. */
 556             velecsum         = _mm_add_pd(velecsum,velec);
 557
 558             fscal            = felec;
 559
 560             /* Calculate temporary vectorial force */
 561             tx               = _mm_mul_pd(fscal,dx20);
 562             ty               = _mm_mul_pd(fscal,dy20);
 563             tz               = _mm_mul_pd(fscal,dz20);
 564
 565             /* Update vectorial force */
 566             fix2             = _mm_add_pd(fix2,tx);
 567             fiy2             = _mm_add_pd(fiy2,ty);
 568             fiz2             = _mm_add_pd(fiz2,tz);
 569
 570             fjx0             = _mm_add_pd(fjx0,tx);
 571             fjy0             = _mm_add_pd(fjy0,ty);
 572             fjz0             = _mm_add_pd(fjz0,tz);
 573
 574             /**************************
 575              * CALCULATE INTERACTIONS *
 576              **************************/
 577
 578             r21              = _mm_mul_pd(rsq21,rinv21);
 579
 580             /* EWALD ELECTROSTATICS */
 581
 582             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 583             ewrt             = _mm_mul_pd(r21,ewtabscale);
 584             ewitab           = _mm_cvttpd_epi32(ewrt);
 585             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 586             ewitab           = _mm_slli_epi32(ewitab,2);
 587             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 588             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 589             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 590             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 591             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 592             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 593             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 594             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 595             velec            = _mm_mul_pd(qq21,_mm_sub_pd(rinv21,velec));
 596             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
 597
 598             /* Update potential sum for this i atom from the interaction with this j atom. */
 599             velecsum         = _mm_add_pd(velecsum,velec);
 600
 601             fscal            = felec;
 602
 603             /* Calculate temporary vectorial force */
 604             tx               = _mm_mul_pd(fscal,dx21);
 605             ty               = _mm_mul_pd(fscal,dy21);
 606             tz               = _mm_mul_pd(fscal,dz21);
 607
 608             /* Update vectorial force */
 609             fix2             = _mm_add_pd(fix2,tx);
 610             fiy2             = _mm_add_pd(fiy2,ty);
 611             fiz2             = _mm_add_pd(fiz2,tz);
 612
 613             fjx1             = _mm_add_pd(fjx1,tx);
 614             fjy1             = _mm_add_pd(fjy1,ty);
 615             fjz1             = _mm_add_pd(fjz1,tz);
 616
 617             /**************************
 618              * CALCULATE INTERACTIONS *
 619              **************************/
 620
 621             r22              = _mm_mul_pd(rsq22,rinv22);
 622
 623             /* EWALD ELECTROSTATICS */
 624
 625             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 626             ewrt             = _mm_mul_pd(r22,ewtabscale);
 627             ewitab           = _mm_cvttpd_epi32(ewrt);
 628             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 629             ewitab           = _mm_slli_epi32(ewitab,2);
 630             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 631             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 632             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 633             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 634             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 635             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 636             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 637             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 638             velec            = _mm_mul_pd(qq22,_mm_sub_pd(rinv22,velec));
 639             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
 640
 641             /* Update potential sum for this i atom from the interaction with this j atom. */
 642             velecsum         = _mm_add_pd(velecsum,velec);
 643
 644             fscal            = felec;
 645
 646             /* Calculate temporary vectorial force */
 647             tx               = _mm_mul_pd(fscal,dx22);
 648             ty               = _mm_mul_pd(fscal,dy22);
 649             tz               = _mm_mul_pd(fscal,dz22);
 650
 651             /* Update vectorial force */
 652             fix2             = _mm_add_pd(fix2,tx);
 653             fiy2             = _mm_add_pd(fiy2,ty);
 654             fiz2             = _mm_add_pd(fiz2,tz);
 655
 656             fjx2             = _mm_add_pd(fjx2,tx);
 657             fjy2             = _mm_add_pd(fjy2,ty);
 658             fjz2             = _mm_add_pd(fjz2,tz);
 659
 660             gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 661
 662             /* Inner loop uses 369 flops */
 663         }
 664
 665         if(jidx<j_index_end)
 666         {
 667
 668             jnrA             = jjnr[jidx];
 669             j_coord_offsetA  = DIM*jnrA;
 670
 671             /* load j atom coordinates */
 672             gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 673                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 674
 675             /* Calculate displacement vector */
 676             dx00             = _mm_sub_pd(ix0,jx0);
 677             dy00             = _mm_sub_pd(iy0,jy0);
 678             dz00             = _mm_sub_pd(iz0,jz0);
 679             dx01             = _mm_sub_pd(ix0,jx1);
 680             dy01             = _mm_sub_pd(iy0,jy1);
 681             dz01             = _mm_sub_pd(iz0,jz1);
 682             dx02             = _mm_sub_pd(ix0,jx2);
 683             dy02             = _mm_sub_pd(iy0,jy2);
 684             dz02             = _mm_sub_pd(iz0,jz2);
 685             dx10             = _mm_sub_pd(ix1,jx0);
 686             dy10             = _mm_sub_pd(iy1,jy0);
 687             dz10             = _mm_sub_pd(iz1,jz0);
 688             dx11             = _mm_sub_pd(ix1,jx1);
 689             dy11             = _mm_sub_pd(iy1,jy1);
 690             dz11             = _mm_sub_pd(iz1,jz1);
 691             dx12             = _mm_sub_pd(ix1,jx2);
 692             dy12             = _mm_sub_pd(iy1,jy2);
 693             dz12             = _mm_sub_pd(iz1,jz2);
 694             dx20             = _mm_sub_pd(ix2,jx0);
 695             dy20             = _mm_sub_pd(iy2,jy0);
 696             dz20             = _mm_sub_pd(iz2,jz0);
 697             dx21             = _mm_sub_pd(ix2,jx1);
 698             dy21             = _mm_sub_pd(iy2,jy1);
 699             dz21             = _mm_sub_pd(iz2,jz1);
 700             dx22             = _mm_sub_pd(ix2,jx2);
 701             dy22             = _mm_sub_pd(iy2,jy2);
 702             dz22             = _mm_sub_pd(iz2,jz2);
 703
 704             /* Calculate squared distance and things based on it */
 705             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 706             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
 707             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
 708             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 709             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 710             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 711             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 712             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 713             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 714
 715             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 716             rinv01           = gmx_mm_invsqrt_pd(rsq01);
 717             rinv02           = gmx_mm_invsqrt_pd(rsq02);
 718             rinv10           = gmx_mm_invsqrt_pd(rsq10);
 719             rinv11           = gmx_mm_invsqrt_pd(rsq11);
 720             rinv12           = gmx_mm_invsqrt_pd(rsq12);
 721             rinv20           = gmx_mm_invsqrt_pd(rsq20);
 722             rinv21           = gmx_mm_invsqrt_pd(rsq21);
 723             rinv22           = gmx_mm_invsqrt_pd(rsq22);
 724
 725             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 726             rinvsq01         = _mm_mul_pd(rinv01,rinv01);
 727             rinvsq02         = _mm_mul_pd(rinv02,rinv02);
 728             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
 729             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
 730             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
 731             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
 732             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
 733             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
 734
 735             fjx0             = _mm_setzero_pd();
 736             fjy0             = _mm_setzero_pd();
 737             fjz0             = _mm_setzero_pd();
 738             fjx1             = _mm_setzero_pd();
 739             fjy1             = _mm_setzero_pd();
 740             fjz1             = _mm_setzero_pd();
 741             fjx2             = _mm_setzero_pd();
 742             fjy2             = _mm_setzero_pd();
 743             fjz2             = _mm_setzero_pd();
 744
 745             /**************************
 746              * CALCULATE INTERACTIONS *
 747              **************************/
 748
 749             r00              = _mm_mul_pd(rsq00,rinv00);
 750
 751             /* EWALD ELECTROSTATICS */
 752
 753             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 754             ewrt             = _mm_mul_pd(r00,ewtabscale);
 755             ewitab           = _mm_cvttpd_epi32(ewrt);
 756             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 757             ewitab           = _mm_slli_epi32(ewitab,2);
 758             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 759             ewtabD           = _mm_setzero_pd();
 760             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 761             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 762             ewtabFn          = _mm_setzero_pd();
 763             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 764             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 765             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 766             velec            = _mm_mul_pd(qq00,_mm_sub_pd(rinv00,velec));
 767             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
 768
 769             /* Update potential sum for this i atom from the interaction with this j atom. */
 770             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 771             velecsum         = _mm_add_pd(velecsum,velec);
 772
 773             fscal            = felec;
 774
 775             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 776
 777             /* Calculate temporary vectorial force */
 778             tx               = _mm_mul_pd(fscal,dx00);
 779             ty               = _mm_mul_pd(fscal,dy00);
 780             tz               = _mm_mul_pd(fscal,dz00);
 781
 782             /* Update vectorial force */
 783             fix0             = _mm_add_pd(fix0,tx);
 784             fiy0             = _mm_add_pd(fiy0,ty);
 785             fiz0             = _mm_add_pd(fiz0,tz);
 786
 787             fjx0             = _mm_add_pd(fjx0,tx);
 788             fjy0             = _mm_add_pd(fjy0,ty);
 789             fjz0             = _mm_add_pd(fjz0,tz);
 790
 791             /**************************
 792              * CALCULATE INTERACTIONS *
 793              **************************/
 794
 795             r01              = _mm_mul_pd(rsq01,rinv01);
 796
 797             /* EWALD ELECTROSTATICS */
 798
 799             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 800             ewrt             = _mm_mul_pd(r01,ewtabscale);
 801             ewitab           = _mm_cvttpd_epi32(ewrt);
 802             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 803             ewitab           = _mm_slli_epi32(ewitab,2);
 804             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 805             ewtabD           = _mm_setzero_pd();
 806             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 807             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 808             ewtabFn          = _mm_setzero_pd();
 809             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 810             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 811             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 812             velec            = _mm_mul_pd(qq01,_mm_sub_pd(rinv01,velec));
 813             felec            = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
 814
 815             /* Update potential sum for this i atom from the interaction with this j atom. */
 816             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 817             velecsum         = _mm_add_pd(velecsum,velec);
 818
 819             fscal            = felec;
 820
 821             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 822
 823             /* Calculate temporary vectorial force */
 824             tx               = _mm_mul_pd(fscal,dx01);
 825             ty               = _mm_mul_pd(fscal,dy01);
 826             tz               = _mm_mul_pd(fscal,dz01);
 827
 828             /* Update vectorial force */
 829             fix0             = _mm_add_pd(fix0,tx);
 830             fiy0             = _mm_add_pd(fiy0,ty);
 831             fiz0             = _mm_add_pd(fiz0,tz);
 832
 833             fjx1             = _mm_add_pd(fjx1,tx);
 834             fjy1             = _mm_add_pd(fjy1,ty);
 835             fjz1             = _mm_add_pd(fjz1,tz);
 836
 837             /**************************
 838              * CALCULATE INTERACTIONS *
 839              **************************/
 840
 841             r02              = _mm_mul_pd(rsq02,rinv02);
 842
 843             /* EWALD ELECTROSTATICS */
 844
 845             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 846             ewrt             = _mm_mul_pd(r02,ewtabscale);
 847             ewitab           = _mm_cvttpd_epi32(ewrt);
 848             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 849             ewitab           = _mm_slli_epi32(ewitab,2);
 850             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 851             ewtabD           = _mm_setzero_pd();
 852             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 853             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 854             ewtabFn          = _mm_setzero_pd();
 855             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 856             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 857             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 858             velec            = _mm_mul_pd(qq02,_mm_sub_pd(rinv02,velec));
 859             felec            = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
 860
 861             /* Update potential sum for this i atom from the interaction with this j atom. */
 862             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 863             velecsum         = _mm_add_pd(velecsum,velec);
 864
 865             fscal            = felec;
 866
 867             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 868
 869             /* Calculate temporary vectorial force */
 870             tx               = _mm_mul_pd(fscal,dx02);
 871             ty               = _mm_mul_pd(fscal,dy02);
 872             tz               = _mm_mul_pd(fscal,dz02);
 873
 874             /* Update vectorial force */
 875             fix0             = _mm_add_pd(fix0,tx);
 876             fiy0             = _mm_add_pd(fiy0,ty);
 877             fiz0             = _mm_add_pd(fiz0,tz);
 878
 879             fjx2             = _mm_add_pd(fjx2,tx);
 880             fjy2             = _mm_add_pd(fjy2,ty);
 881             fjz2             = _mm_add_pd(fjz2,tz);
 882
 883             /**************************
 884              * CALCULATE INTERACTIONS *
 885              **************************/
 886
 887             r10              = _mm_mul_pd(rsq10,rinv10);
 888
 889             /* EWALD ELECTROSTATICS */
 890
 891             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 892             ewrt             = _mm_mul_pd(r10,ewtabscale);
 893             ewitab           = _mm_cvttpd_epi32(ewrt);
 894             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 895             ewitab           = _mm_slli_epi32(ewitab,2);
 896             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 897             ewtabD           = _mm_setzero_pd();
 898             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 899             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 900             ewtabFn          = _mm_setzero_pd();
 901             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 902             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 903             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 904             velec            = _mm_mul_pd(qq10,_mm_sub_pd(rinv10,velec));
 905             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
 906
 907             /* Update potential sum for this i atom from the interaction with this j atom. */
 908             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 909             velecsum         = _mm_add_pd(velecsum,velec);
 910
 911             fscal            = felec;
 912
 913             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 914
 915             /* Calculate temporary vectorial force */
 916             tx               = _mm_mul_pd(fscal,dx10);
 917             ty               = _mm_mul_pd(fscal,dy10);
 918             tz               = _mm_mul_pd(fscal,dz10);
 919
 920             /* Update vectorial force */
 921             fix1             = _mm_add_pd(fix1,tx);
 922             fiy1             = _mm_add_pd(fiy1,ty);
 923             fiz1             = _mm_add_pd(fiz1,tz);
 924
 925             fjx0             = _mm_add_pd(fjx0,tx);
 926             fjy0             = _mm_add_pd(fjy0,ty);
 927             fjz0             = _mm_add_pd(fjz0,tz);
 928
 929             /**************************
 930              * CALCULATE INTERACTIONS *
 931              **************************/
 932
 933             r11              = _mm_mul_pd(rsq11,rinv11);
 934
 935             /* EWALD ELECTROSTATICS */
 936
 937             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 938             ewrt             = _mm_mul_pd(r11,ewtabscale);
 939             ewitab           = _mm_cvttpd_epi32(ewrt);
 940             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 941             ewitab           = _mm_slli_epi32(ewitab,2);
 942             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 943             ewtabD           = _mm_setzero_pd();
 944             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 945             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 946             ewtabFn          = _mm_setzero_pd();
 947             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 948             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 949             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 950             velec            = _mm_mul_pd(qq11,_mm_sub_pd(rinv11,velec));
 951             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
 952
 953             /* Update potential sum for this i atom from the interaction with this j atom. */
 954             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 955             velecsum         = _mm_add_pd(velecsum,velec);
 956
 957             fscal            = felec;
 958
 959             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 960
 961             /* Calculate temporary vectorial force */
 962             tx               = _mm_mul_pd(fscal,dx11);
 963             ty               = _mm_mul_pd(fscal,dy11);
 964             tz               = _mm_mul_pd(fscal,dz11);
 965
 966             /* Update vectorial force */
 967             fix1             = _mm_add_pd(fix1,tx);
 968             fiy1             = _mm_add_pd(fiy1,ty);
 969             fiz1             = _mm_add_pd(fiz1,tz);
 970
 971             fjx1             = _mm_add_pd(fjx1,tx);
 972             fjy1             = _mm_add_pd(fjy1,ty);
 973             fjz1             = _mm_add_pd(fjz1,tz);
 974
 975             /**************************
 976              * CALCULATE INTERACTIONS *
 977              **************************/
 978
 979             r12              = _mm_mul_pd(rsq12,rinv12);
 980
 981             /* EWALD ELECTROSTATICS */
 982
 983             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 984             ewrt             = _mm_mul_pd(r12,ewtabscale);
 985             ewitab           = _mm_cvttpd_epi32(ewrt);
 986             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 987             ewitab           = _mm_slli_epi32(ewitab,2);
 988             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 989             ewtabD           = _mm_setzero_pd();
 990             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 991             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 992             ewtabFn          = _mm_setzero_pd();
 993             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 994             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 995             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 996             velec            = _mm_mul_pd(qq12,_mm_sub_pd(rinv12,velec));
 997             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
 998
 999             /* Update potential sum for this i atom from the interaction with this j atom. */
1000             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1001             velecsum         = _mm_add_pd(velecsum,velec);
1002
1003             fscal            = felec;
1004
1005             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1006
1007             /* Calculate temporary vectorial force */
1008             tx               = _mm_mul_pd(fscal,dx12);
1009             ty               = _mm_mul_pd(fscal,dy12);
1010             tz               = _mm_mul_pd(fscal,dz12);
1011
1012             /* Update vectorial force */
1013             fix1             = _mm_add_pd(fix1,tx);
1014             fiy1             = _mm_add_pd(fiy1,ty);
1015             fiz1             = _mm_add_pd(fiz1,tz);
1016
1017             fjx2             = _mm_add_pd(fjx2,tx);
1018             fjy2             = _mm_add_pd(fjy2,ty);
1019             fjz2             = _mm_add_pd(fjz2,tz);
1020
1021             /**************************
1022              * CALCULATE INTERACTIONS *
1023              **************************/
1024
1025             r20              = _mm_mul_pd(rsq20,rinv20);
1026
1027             /* EWALD ELECTROSTATICS */
1028
1029             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1030             ewrt             = _mm_mul_pd(r20,ewtabscale);
1031             ewitab           = _mm_cvttpd_epi32(ewrt);
1032             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1033             ewitab           = _mm_slli_epi32(ewitab,2);
1034             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1035             ewtabD           = _mm_setzero_pd();
1036             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1037             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1038             ewtabFn          = _mm_setzero_pd();
1039             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1040             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1041             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1042             velec            = _mm_mul_pd(qq20,_mm_sub_pd(rinv20,velec));
1043             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
1044
1045             /* Update potential sum for this i atom from the interaction with this j atom. */
1046             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1047             velecsum         = _mm_add_pd(velecsum,velec);
1048
1049             fscal            = felec;
1050
1051             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1052
1053             /* Calculate temporary vectorial force */
1054             tx               = _mm_mul_pd(fscal,dx20);
1055             ty               = _mm_mul_pd(fscal,dy20);
1056             tz               = _mm_mul_pd(fscal,dz20);
1057
1058             /* Update vectorial force */
1059             fix2             = _mm_add_pd(fix2,tx);
1060             fiy2             = _mm_add_pd(fiy2,ty);
1061             fiz2             = _mm_add_pd(fiz2,tz);
1062
1063             fjx0             = _mm_add_pd(fjx0,tx);
1064             fjy0             = _mm_add_pd(fjy0,ty);
1065             fjz0             = _mm_add_pd(fjz0,tz);
1066
1067             /**************************
1068              * CALCULATE INTERACTIONS *
1069              **************************/
1070
1071             r21              = _mm_mul_pd(rsq21,rinv21);
1072
1073             /* EWALD ELECTROSTATICS */
1074
1075             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1076             ewrt             = _mm_mul_pd(r21,ewtabscale);
1077             ewitab           = _mm_cvttpd_epi32(ewrt);
1078             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1079             ewitab           = _mm_slli_epi32(ewitab,2);
1080             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1081             ewtabD           = _mm_setzero_pd();
1082             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1083             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1084             ewtabFn          = _mm_setzero_pd();
1085             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1086             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1087             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1088             velec            = _mm_mul_pd(qq21,_mm_sub_pd(rinv21,velec));
1089             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
1090
1091             /* Update potential sum for this i atom from the interaction with this j atom. */
1092             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1093             velecsum         = _mm_add_pd(velecsum,velec);
1094
1095             fscal            = felec;
1096
1097             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1098
1099             /* Calculate temporary vectorial force */
1100             tx               = _mm_mul_pd(fscal,dx21);
1101             ty               = _mm_mul_pd(fscal,dy21);
1102             tz               = _mm_mul_pd(fscal,dz21);
1103
1104             /* Update vectorial force */
1105             fix2             = _mm_add_pd(fix2,tx);
1106             fiy2             = _mm_add_pd(fiy2,ty);
1107             fiz2             = _mm_add_pd(fiz2,tz);
1108
1109             fjx1             = _mm_add_pd(fjx1,tx);
1110             fjy1             = _mm_add_pd(fjy1,ty);
1111             fjz1             = _mm_add_pd(fjz1,tz);
1112
1113             /**************************
1114              * CALCULATE INTERACTIONS *
1115              **************************/
1116
1117             r22              = _mm_mul_pd(rsq22,rinv22);
1118
1119             /* EWALD ELECTROSTATICS */
1120
1121             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1122             ewrt             = _mm_mul_pd(r22,ewtabscale);
1123             ewitab           = _mm_cvttpd_epi32(ewrt);
1124             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1125             ewitab           = _mm_slli_epi32(ewitab,2);
1126             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1127             ewtabD           = _mm_setzero_pd();
1128             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1129             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1130             ewtabFn          = _mm_setzero_pd();
1131             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1132             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1133             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1134             velec            = _mm_mul_pd(qq22,_mm_sub_pd(rinv22,velec));
1135             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
1136
1137             /* Update potential sum for this i atom from the interaction with this j atom. */
1138             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1139             velecsum         = _mm_add_pd(velecsum,velec);
1140
1141             fscal            = felec;
1142
1143             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1144
1145             /* Calculate temporary vectorial force */
1146             tx               = _mm_mul_pd(fscal,dx22);
1147             ty               = _mm_mul_pd(fscal,dy22);
1148             tz               = _mm_mul_pd(fscal,dz22);
1149
1150             /* Update vectorial force */
1151             fix2             = _mm_add_pd(fix2,tx);
1152             fiy2             = _mm_add_pd(fiy2,ty);
1153             fiz2             = _mm_add_pd(fiz2,tz);
1154
1155             fjx2             = _mm_add_pd(fjx2,tx);
1156             fjy2             = _mm_add_pd(fjy2,ty);
1157             fjz2             = _mm_add_pd(fjz2,tz);
1158
1159             gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1160
1161             /* Inner loop uses 369 flops */
1162         }
1163
1164         /* End of innermost loop */
1165
1166         gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1167                                               f+i_coord_offset,fshift+i_shift_offset);
1168
1169         ggid                        = gid[iidx];
1170         /* Update potential energies */
1171         gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1172
1173         /* Increment number of inner iterations */
1174         inneriter                  += j_index_end - j_index_start;
1175
1176         /* Outer loop uses 19 flops */
1177     }
1178
1179     /* Increment number of outer iterations */
1180     outeriter        += nri;
1181
1182     /* Update outer/inner flops */
1183
1184     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3W3_VF,outeriter*19 + inneriter*369);
1185 }
1186 /*
1187  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwNone_GeomW3W3_F_sse2_double
1188  * Electrostatics interaction: Ewald
1189  * VdW interaction:            None
1190  * Geometry:                   Water3-Water3
1191  * Calculate force/pot:        Force
1192  */
1193 void
1194 nb_kernel_ElecEw_VdwNone_GeomW3W3_F_sse2_double
1195                     (t_nblist                    * gmx_restrict       nlist,
1196                      rvec                        * gmx_restrict          xx,
1197                      rvec                        * gmx_restrict          ff,
1198                      t_forcerec                  * gmx_restrict          fr,
1199                      t_mdatoms                   * gmx_restrict     mdatoms,
1200                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1201                      t_nrnb                      * gmx_restrict        nrnb)
1202 {
1203     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1204      * just 0 for non-waters.
1205      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
1206      * jnr indices corresponding to data put in the four positions in the SIMD register.
1207      */
1208     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1209     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1210     int              jnrA,jnrB;
1211     int              j_coord_offsetA,j_coord_offsetB;
1212     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1213     real             rcutoff_scalar;
1214     real             *shiftvec,*fshift,*x,*f;
1215     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1216     int              vdwioffset0;
1217     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1218     int              vdwioffset1;
1219     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1220     int              vdwioffset2;
1221     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1222     int              vdwjidx0A,vdwjidx0B;
1223     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1224     int              vdwjidx1A,vdwjidx1B;
1225     __m128d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1226     int              vdwjidx2A,vdwjidx2B;
1227     __m128d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1228     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1229     __m128d          dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1230     __m128d          dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1231     __m128d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1232     __m128d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1233     __m128d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1234     __m128d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1235     __m128d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1236     __m128d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1237     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
1238     real             *charge;
1239     __m128i          ewitab;
1240     __m128d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1241     real             *ewtab;
1242     __m128d          dummy_mask,cutoff_mask;
1243     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
1244     __m128d          one     = _mm_set1_pd(1.0);
1245     __m128d          two     = _mm_set1_pd(2.0);
1246     x                = xx[0];
1247     f                = ff[0];
1248
1249     nri              = nlist->nri;
1250     iinr             = nlist->iinr;
1251     jindex           = nlist->jindex;
1252     jjnr             = nlist->jjnr;
1253     shiftidx         = nlist->shift;
1254     gid              = nlist->gid;
1255     shiftvec         = fr->shift_vec[0];
1256     fshift           = fr->fshift[0];
1257     facel            = _mm_set1_pd(fr->epsfac);
1258     charge           = mdatoms->chargeA;
1259
1260     sh_ewald         = _mm_set1_pd(fr->ic->sh_ewald);
1261     ewtab            = fr->ic->tabq_coul_F;
1262     ewtabscale       = _mm_set1_pd(fr->ic->tabq_scale);
1263     ewtabhalfspace   = _mm_set1_pd(0.5/fr->ic->tabq_scale);
1264
1265     /* Setup water-specific parameters */
1266     inr              = nlist->iinr[0];
1267     iq0              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
1268     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
1269     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
1270
1271     jq0              = _mm_set1_pd(charge[inr+0]);
1272     jq1              = _mm_set1_pd(charge[inr+1]);
1273     jq2              = _mm_set1_pd(charge[inr+2]);
1274     qq00             = _mm_mul_pd(iq0,jq0);
1275     qq01             = _mm_mul_pd(iq0,jq1);
1276     qq02             = _mm_mul_pd(iq0,jq2);
1277     qq10             = _mm_mul_pd(iq1,jq0);
1278     qq11             = _mm_mul_pd(iq1,jq1);
1279     qq12             = _mm_mul_pd(iq1,jq2);
1280     qq20             = _mm_mul_pd(iq2,jq0);
1281     qq21             = _mm_mul_pd(iq2,jq1);
1282     qq22             = _mm_mul_pd(iq2,jq2);
1283
1284     /* Avoid stupid compiler warnings */
1285     jnrA = jnrB = 0;
1286     j_coord_offsetA = 0;
1287     j_coord_offsetB = 0;
1288
1289     outeriter        = 0;
1290     inneriter        = 0;
1291
1292     /* Start outer loop over neighborlists */
1293     for(iidx=0; iidx<nri; iidx++)
1294     {
1295         /* Load shift vector for this list */
1296         i_shift_offset   = DIM*shiftidx[iidx];
1297
1298         /* Load limits for loop over neighbors */
1299         j_index_start    = jindex[iidx];
1300         j_index_end      = jindex[iidx+1];
1301
1302         /* Get outer coordinate index */
1303         inr              = iinr[iidx];
1304         i_coord_offset   = DIM*inr;
1305
1306         /* Load i particle coords and add shift vector */
1307         gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1308                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1309
1310         fix0             = _mm_setzero_pd();
1311         fiy0             = _mm_setzero_pd();
1312         fiz0             = _mm_setzero_pd();
1313         fix1             = _mm_setzero_pd();
1314         fiy1             = _mm_setzero_pd();
1315         fiz1             = _mm_setzero_pd();
1316         fix2             = _mm_setzero_pd();
1317         fiy2             = _mm_setzero_pd();
1318         fiz2             = _mm_setzero_pd();
1319
1320         /* Start inner kernel loop */
1321         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
1322         {
1323
1324             /* Get j neighbor index, and coordinate index */
1325             jnrA             = jjnr[jidx];
1326             jnrB             = jjnr[jidx+1];
1327             j_coord_offsetA  = DIM*jnrA;
1328             j_coord_offsetB  = DIM*jnrB;
1329
1330             /* load j atom coordinates */
1331             gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1332                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1333
1334             /* Calculate displacement vector */
1335             dx00             = _mm_sub_pd(ix0,jx0);
1336             dy00             = _mm_sub_pd(iy0,jy0);
1337             dz00             = _mm_sub_pd(iz0,jz0);
1338             dx01             = _mm_sub_pd(ix0,jx1);
1339             dy01             = _mm_sub_pd(iy0,jy1);
1340             dz01             = _mm_sub_pd(iz0,jz1);
1341             dx02             = _mm_sub_pd(ix0,jx2);
1342             dy02             = _mm_sub_pd(iy0,jy2);
1343             dz02             = _mm_sub_pd(iz0,jz2);
1344             dx10             = _mm_sub_pd(ix1,jx0);
1345             dy10             = _mm_sub_pd(iy1,jy0);
1346             dz10             = _mm_sub_pd(iz1,jz0);
1347             dx11             = _mm_sub_pd(ix1,jx1);
1348             dy11             = _mm_sub_pd(iy1,jy1);
1349             dz11             = _mm_sub_pd(iz1,jz1);
1350             dx12             = _mm_sub_pd(ix1,jx2);
1351             dy12             = _mm_sub_pd(iy1,jy2);
1352             dz12             = _mm_sub_pd(iz1,jz2);
1353             dx20             = _mm_sub_pd(ix2,jx0);
1354             dy20             = _mm_sub_pd(iy2,jy0);
1355             dz20             = _mm_sub_pd(iz2,jz0);
1356             dx21             = _mm_sub_pd(ix2,jx1);
1357             dy21             = _mm_sub_pd(iy2,jy1);
1358             dz21             = _mm_sub_pd(iz2,jz1);
1359             dx22             = _mm_sub_pd(ix2,jx2);
1360             dy22             = _mm_sub_pd(iy2,jy2);
1361             dz22             = _mm_sub_pd(iz2,jz2);
1362
1363             /* Calculate squared distance and things based on it */
1364             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1365             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
1366             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
1367             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
1368             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1369             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1370             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
1371             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1372             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1373
1374             rinv00           = gmx_mm_invsqrt_pd(rsq00);
1375             rinv01           = gmx_mm_invsqrt_pd(rsq01);
1376             rinv02           = gmx_mm_invsqrt_pd(rsq02);
1377             rinv10           = gmx_mm_invsqrt_pd(rsq10);
1378             rinv11           = gmx_mm_invsqrt_pd(rsq11);
1379             rinv12           = gmx_mm_invsqrt_pd(rsq12);
1380             rinv20           = gmx_mm_invsqrt_pd(rsq20);
1381             rinv21           = gmx_mm_invsqrt_pd(rsq21);
1382             rinv22           = gmx_mm_invsqrt_pd(rsq22);
1383
1384             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
1385             rinvsq01         = _mm_mul_pd(rinv01,rinv01);
1386             rinvsq02         = _mm_mul_pd(rinv02,rinv02);
1387             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
1388             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
1389             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
1390             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
1391             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
1392             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
1393
1394             fjx0             = _mm_setzero_pd();
1395             fjy0             = _mm_setzero_pd();
1396             fjz0             = _mm_setzero_pd();
1397             fjx1             = _mm_setzero_pd();
1398             fjy1             = _mm_setzero_pd();
1399             fjz1             = _mm_setzero_pd();
1400             fjx2             = _mm_setzero_pd();
1401             fjy2             = _mm_setzero_pd();
1402             fjz2             = _mm_setzero_pd();
1403
1404             /**************************
1405              * CALCULATE INTERACTIONS *
1406              **************************/
1407
1408             r00              = _mm_mul_pd(rsq00,rinv00);
1409
1410             /* EWALD ELECTROSTATICS */
1411
1412             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1413             ewrt             = _mm_mul_pd(r00,ewtabscale);
1414             ewitab           = _mm_cvttpd_epi32(ewrt);
1415             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1416             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1417                                          &ewtabF,&ewtabFn);
1418             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1419             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
1420
1421             fscal            = felec;
1422
1423             /* Calculate temporary vectorial force */
1424             tx               = _mm_mul_pd(fscal,dx00);
1425             ty               = _mm_mul_pd(fscal,dy00);
1426             tz               = _mm_mul_pd(fscal,dz00);
1427
1428             /* Update vectorial force */
1429             fix0             = _mm_add_pd(fix0,tx);
1430             fiy0             = _mm_add_pd(fiy0,ty);
1431             fiz0             = _mm_add_pd(fiz0,tz);
1432
1433             fjx0             = _mm_add_pd(fjx0,tx);
1434             fjy0             = _mm_add_pd(fjy0,ty);
1435             fjz0             = _mm_add_pd(fjz0,tz);
1436
1437             /**************************
1438              * CALCULATE INTERACTIONS *
1439              **************************/
1440
1441             r01              = _mm_mul_pd(rsq01,rinv01);
1442
1443             /* EWALD ELECTROSTATICS */
1444
1445             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1446             ewrt             = _mm_mul_pd(r01,ewtabscale);
1447             ewitab           = _mm_cvttpd_epi32(ewrt);
1448             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1449             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1450                                          &ewtabF,&ewtabFn);
1451             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1452             felec            = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
1453
1454             fscal            = felec;
1455
1456             /* Calculate temporary vectorial force */
1457             tx               = _mm_mul_pd(fscal,dx01);
1458             ty               = _mm_mul_pd(fscal,dy01);
1459             tz               = _mm_mul_pd(fscal,dz01);
1460
1461             /* Update vectorial force */
1462             fix0             = _mm_add_pd(fix0,tx);
1463             fiy0             = _mm_add_pd(fiy0,ty);
1464             fiz0             = _mm_add_pd(fiz0,tz);
1465
1466             fjx1             = _mm_add_pd(fjx1,tx);
1467             fjy1             = _mm_add_pd(fjy1,ty);
1468             fjz1             = _mm_add_pd(fjz1,tz);
1469
1470             /**************************
1471              * CALCULATE INTERACTIONS *
1472              **************************/
1473
1474             r02              = _mm_mul_pd(rsq02,rinv02);
1475
1476             /* EWALD ELECTROSTATICS */
1477
1478             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1479             ewrt             = _mm_mul_pd(r02,ewtabscale);
1480             ewitab           = _mm_cvttpd_epi32(ewrt);
1481             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1482             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1483                                          &ewtabF,&ewtabFn);
1484             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1485             felec            = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
1486
1487             fscal            = felec;
1488
1489             /* Calculate temporary vectorial force */
1490             tx               = _mm_mul_pd(fscal,dx02);
1491             ty               = _mm_mul_pd(fscal,dy02);
1492             tz               = _mm_mul_pd(fscal,dz02);
1493
1494             /* Update vectorial force */
1495             fix0             = _mm_add_pd(fix0,tx);
1496             fiy0             = _mm_add_pd(fiy0,ty);
1497             fiz0             = _mm_add_pd(fiz0,tz);
1498
1499             fjx2             = _mm_add_pd(fjx2,tx);
1500             fjy2             = _mm_add_pd(fjy2,ty);
1501             fjz2             = _mm_add_pd(fjz2,tz);
1502
1503             /**************************
1504              * CALCULATE INTERACTIONS *
1505              **************************/
1506
1507             r10              = _mm_mul_pd(rsq10,rinv10);
1508
1509             /* EWALD ELECTROSTATICS */
1510
1511             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1512             ewrt             = _mm_mul_pd(r10,ewtabscale);
1513             ewitab           = _mm_cvttpd_epi32(ewrt);
1514             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1515             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1516                                          &ewtabF,&ewtabFn);
1517             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1518             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
1519
1520             fscal            = felec;
1521
1522             /* Calculate temporary vectorial force */
1523             tx               = _mm_mul_pd(fscal,dx10);
1524             ty               = _mm_mul_pd(fscal,dy10);
1525             tz               = _mm_mul_pd(fscal,dz10);
1526
1527             /* Update vectorial force */
1528             fix1             = _mm_add_pd(fix1,tx);
1529             fiy1             = _mm_add_pd(fiy1,ty);
1530             fiz1             = _mm_add_pd(fiz1,tz);
1531
1532             fjx0             = _mm_add_pd(fjx0,tx);
1533             fjy0             = _mm_add_pd(fjy0,ty);
1534             fjz0             = _mm_add_pd(fjz0,tz);
1535
1536             /**************************
1537              * CALCULATE INTERACTIONS *
1538              **************************/
1539
1540             r11              = _mm_mul_pd(rsq11,rinv11);
1541
1542             /* EWALD ELECTROSTATICS */
1543
1544             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1545             ewrt             = _mm_mul_pd(r11,ewtabscale);
1546             ewitab           = _mm_cvttpd_epi32(ewrt);
1547             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1548             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1549                                          &ewtabF,&ewtabFn);
1550             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1551             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
1552
1553             fscal            = felec;
1554
1555             /* Calculate temporary vectorial force */
1556             tx               = _mm_mul_pd(fscal,dx11);
1557             ty               = _mm_mul_pd(fscal,dy11);
1558             tz               = _mm_mul_pd(fscal,dz11);
1559
1560             /* Update vectorial force */
1561             fix1             = _mm_add_pd(fix1,tx);
1562             fiy1             = _mm_add_pd(fiy1,ty);
1563             fiz1             = _mm_add_pd(fiz1,tz);
1564
1565             fjx1             = _mm_add_pd(fjx1,tx);
1566             fjy1             = _mm_add_pd(fjy1,ty);
1567             fjz1             = _mm_add_pd(fjz1,tz);
1568
1569             /**************************
1570              * CALCULATE INTERACTIONS *
1571              **************************/
1572
1573             r12              = _mm_mul_pd(rsq12,rinv12);
1574
1575             /* EWALD ELECTROSTATICS */
1576
1577             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1578             ewrt             = _mm_mul_pd(r12,ewtabscale);
1579             ewitab           = _mm_cvttpd_epi32(ewrt);
1580             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1581             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1582                                          &ewtabF,&ewtabFn);
1583             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1584             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
1585
1586             fscal            = felec;
1587
1588             /* Calculate temporary vectorial force */
1589             tx               = _mm_mul_pd(fscal,dx12);
1590             ty               = _mm_mul_pd(fscal,dy12);
1591             tz               = _mm_mul_pd(fscal,dz12);
1592
1593             /* Update vectorial force */
1594             fix1             = _mm_add_pd(fix1,tx);
1595             fiy1             = _mm_add_pd(fiy1,ty);
1596             fiz1             = _mm_add_pd(fiz1,tz);
1597
1598             fjx2             = _mm_add_pd(fjx2,tx);
1599             fjy2             = _mm_add_pd(fjy2,ty);
1600             fjz2             = _mm_add_pd(fjz2,tz);
1601
1602             /**************************
1603              * CALCULATE INTERACTIONS *
1604              **************************/
1605
1606             r20              = _mm_mul_pd(rsq20,rinv20);
1607
1608             /* EWALD ELECTROSTATICS */
1609
1610             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1611             ewrt             = _mm_mul_pd(r20,ewtabscale);
1612             ewitab           = _mm_cvttpd_epi32(ewrt);
1613             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1614             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1615                                          &ewtabF,&ewtabFn);
1616             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1617             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
1618
1619             fscal            = felec;
1620
1621             /* Calculate temporary vectorial force */
1622             tx               = _mm_mul_pd(fscal,dx20);
1623             ty               = _mm_mul_pd(fscal,dy20);
1624             tz               = _mm_mul_pd(fscal,dz20);
1625
1626             /* Update vectorial force */
1627             fix2             = _mm_add_pd(fix2,tx);
1628             fiy2             = _mm_add_pd(fiy2,ty);
1629             fiz2             = _mm_add_pd(fiz2,tz);
1630
1631             fjx0             = _mm_add_pd(fjx0,tx);
1632             fjy0             = _mm_add_pd(fjy0,ty);
1633             fjz0             = _mm_add_pd(fjz0,tz);
1634
1635             /**************************
1636              * CALCULATE INTERACTIONS *
1637              **************************/
1638
1639             r21              = _mm_mul_pd(rsq21,rinv21);
1640
1641             /* EWALD ELECTROSTATICS */
1642
1643             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1644             ewrt             = _mm_mul_pd(r21,ewtabscale);
1645             ewitab           = _mm_cvttpd_epi32(ewrt);
1646             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1647             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1648                                          &ewtabF,&ewtabFn);
1649             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1650             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
1651
1652             fscal            = felec;
1653
1654             /* Calculate temporary vectorial force */
1655             tx               = _mm_mul_pd(fscal,dx21);
1656             ty               = _mm_mul_pd(fscal,dy21);
1657             tz               = _mm_mul_pd(fscal,dz21);
1658
1659             /* Update vectorial force */
1660             fix2             = _mm_add_pd(fix2,tx);
1661             fiy2             = _mm_add_pd(fiy2,ty);
1662             fiz2             = _mm_add_pd(fiz2,tz);
1663
1664             fjx1             = _mm_add_pd(fjx1,tx);
1665             fjy1             = _mm_add_pd(fjy1,ty);
1666             fjz1             = _mm_add_pd(fjz1,tz);
1667
1668             /**************************
1669              * CALCULATE INTERACTIONS *
1670              **************************/
1671
1672             r22              = _mm_mul_pd(rsq22,rinv22);
1673
1674             /* EWALD ELECTROSTATICS */
1675
1676             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1677             ewrt             = _mm_mul_pd(r22,ewtabscale);
1678             ewitab           = _mm_cvttpd_epi32(ewrt);
1679             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1680             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1681                                          &ewtabF,&ewtabFn);
1682             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1683             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
1684
1685             fscal            = felec;
1686
1687             /* Calculate temporary vectorial force */
1688             tx               = _mm_mul_pd(fscal,dx22);
1689             ty               = _mm_mul_pd(fscal,dy22);
1690             tz               = _mm_mul_pd(fscal,dz22);
1691
1692             /* Update vectorial force */
1693             fix2             = _mm_add_pd(fix2,tx);
1694             fiy2             = _mm_add_pd(fiy2,ty);
1695             fiz2             = _mm_add_pd(fiz2,tz);
1696
1697             fjx2             = _mm_add_pd(fjx2,tx);
1698             fjy2             = _mm_add_pd(fjy2,ty);
1699             fjz2             = _mm_add_pd(fjz2,tz);
1700
1701             gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1702
1703             /* Inner loop uses 324 flops */
1704         }
1705
1706         if(jidx<j_index_end)
1707         {
1708
1709             jnrA             = jjnr[jidx];
1710             j_coord_offsetA  = DIM*jnrA;
1711
1712             /* load j atom coordinates */
1713             gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
1714                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1715
1716             /* Calculate displacement vector */
1717             dx00             = _mm_sub_pd(ix0,jx0);
1718             dy00             = _mm_sub_pd(iy0,jy0);
1719             dz00             = _mm_sub_pd(iz0,jz0);
1720             dx01             = _mm_sub_pd(ix0,jx1);
1721             dy01             = _mm_sub_pd(iy0,jy1);
1722             dz01             = _mm_sub_pd(iz0,jz1);
1723             dx02             = _mm_sub_pd(ix0,jx2);
1724             dy02             = _mm_sub_pd(iy0,jy2);
1725             dz02             = _mm_sub_pd(iz0,jz2);
1726             dx10             = _mm_sub_pd(ix1,jx0);
1727             dy10             = _mm_sub_pd(iy1,jy0);
1728             dz10             = _mm_sub_pd(iz1,jz0);
1729             dx11             = _mm_sub_pd(ix1,jx1);
1730             dy11             = _mm_sub_pd(iy1,jy1);
1731             dz11             = _mm_sub_pd(iz1,jz1);
1732             dx12             = _mm_sub_pd(ix1,jx2);
1733             dy12             = _mm_sub_pd(iy1,jy2);
1734             dz12             = _mm_sub_pd(iz1,jz2);
1735             dx20             = _mm_sub_pd(ix2,jx0);
1736             dy20             = _mm_sub_pd(iy2,jy0);
1737             dz20             = _mm_sub_pd(iz2,jz0);
1738             dx21             = _mm_sub_pd(ix2,jx1);
1739             dy21             = _mm_sub_pd(iy2,jy1);
1740             dz21             = _mm_sub_pd(iz2,jz1);
1741             dx22             = _mm_sub_pd(ix2,jx2);
1742             dy22             = _mm_sub_pd(iy2,jy2);
1743             dz22             = _mm_sub_pd(iz2,jz2);
1744
1745             /* Calculate squared distance and things based on it */
1746             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1747             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
1748             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
1749             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
1750             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1751             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1752             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
1753             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1754             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1755
1756             rinv00           = gmx_mm_invsqrt_pd(rsq00);
1757             rinv01           = gmx_mm_invsqrt_pd(rsq01);
1758             rinv02           = gmx_mm_invsqrt_pd(rsq02);
1759             rinv10           = gmx_mm_invsqrt_pd(rsq10);
1760             rinv11           = gmx_mm_invsqrt_pd(rsq11);
1761             rinv12           = gmx_mm_invsqrt_pd(rsq12);
1762             rinv20           = gmx_mm_invsqrt_pd(rsq20);
1763             rinv21           = gmx_mm_invsqrt_pd(rsq21);
1764             rinv22           = gmx_mm_invsqrt_pd(rsq22);
1765
1766             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
1767             rinvsq01         = _mm_mul_pd(rinv01,rinv01);
1768             rinvsq02         = _mm_mul_pd(rinv02,rinv02);
1769             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
1770             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
1771             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
1772             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
1773             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
1774             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
1775
1776             fjx0             = _mm_setzero_pd();
1777             fjy0             = _mm_setzero_pd();
1778             fjz0             = _mm_setzero_pd();
1779             fjx1             = _mm_setzero_pd();
1780             fjy1             = _mm_setzero_pd();
1781             fjz1             = _mm_setzero_pd();
1782             fjx2             = _mm_setzero_pd();
1783             fjy2             = _mm_setzero_pd();
1784             fjz2             = _mm_setzero_pd();
1785
1786             /**************************
1787              * CALCULATE INTERACTIONS *
1788              **************************/
1789
1790             r00              = _mm_mul_pd(rsq00,rinv00);
1791
1792             /* EWALD ELECTROSTATICS */
1793
1794             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1795             ewrt             = _mm_mul_pd(r00,ewtabscale);
1796             ewitab           = _mm_cvttpd_epi32(ewrt);
1797             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1798             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1799             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1800             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
1801
1802             fscal            = felec;
1803
1804             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1805
1806             /* Calculate temporary vectorial force */
1807             tx               = _mm_mul_pd(fscal,dx00);
1808             ty               = _mm_mul_pd(fscal,dy00);
1809             tz               = _mm_mul_pd(fscal,dz00);
1810
1811             /* Update vectorial force */
1812             fix0             = _mm_add_pd(fix0,tx);
1813             fiy0             = _mm_add_pd(fiy0,ty);
1814             fiz0             = _mm_add_pd(fiz0,tz);
1815
1816             fjx0             = _mm_add_pd(fjx0,tx);
1817             fjy0             = _mm_add_pd(fjy0,ty);
1818             fjz0             = _mm_add_pd(fjz0,tz);
1819
1820             /**************************
1821              * CALCULATE INTERACTIONS *
1822              **************************/
1823
1824             r01              = _mm_mul_pd(rsq01,rinv01);
1825
1826             /* EWALD ELECTROSTATICS */
1827
1828             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1829             ewrt             = _mm_mul_pd(r01,ewtabscale);
1830             ewitab           = _mm_cvttpd_epi32(ewrt);
1831             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1832             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1833             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1834             felec            = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
1835
1836             fscal            = felec;
1837
1838             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1839
1840             /* Calculate temporary vectorial force */
1841             tx               = _mm_mul_pd(fscal,dx01);
1842             ty               = _mm_mul_pd(fscal,dy01);
1843             tz               = _mm_mul_pd(fscal,dz01);
1844
1845             /* Update vectorial force */
1846             fix0             = _mm_add_pd(fix0,tx);
1847             fiy0             = _mm_add_pd(fiy0,ty);
1848             fiz0             = _mm_add_pd(fiz0,tz);
1849
1850             fjx1             = _mm_add_pd(fjx1,tx);
1851             fjy1             = _mm_add_pd(fjy1,ty);
1852             fjz1             = _mm_add_pd(fjz1,tz);
1853
1854             /**************************
1855              * CALCULATE INTERACTIONS *
1856              **************************/
1857
1858             r02              = _mm_mul_pd(rsq02,rinv02);
1859
1860             /* EWALD ELECTROSTATICS */
1861
1862             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1863             ewrt             = _mm_mul_pd(r02,ewtabscale);
1864             ewitab           = _mm_cvttpd_epi32(ewrt);
1865             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1866             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1867             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1868             felec            = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
1869
1870             fscal            = felec;
1871
1872             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1873
1874             /* Calculate temporary vectorial force */
1875             tx               = _mm_mul_pd(fscal,dx02);
1876             ty               = _mm_mul_pd(fscal,dy02);
1877             tz               = _mm_mul_pd(fscal,dz02);
1878
1879             /* Update vectorial force */
1880             fix0             = _mm_add_pd(fix0,tx);
1881             fiy0             = _mm_add_pd(fiy0,ty);
1882             fiz0             = _mm_add_pd(fiz0,tz);
1883
1884             fjx2             = _mm_add_pd(fjx2,tx);
1885             fjy2             = _mm_add_pd(fjy2,ty);
1886             fjz2             = _mm_add_pd(fjz2,tz);
1887
1888             /**************************
1889              * CALCULATE INTERACTIONS *
1890              **************************/
1891
1892             r10              = _mm_mul_pd(rsq10,rinv10);
1893
1894             /* EWALD ELECTROSTATICS */
1895
1896             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1897             ewrt             = _mm_mul_pd(r10,ewtabscale);
1898             ewitab           = _mm_cvttpd_epi32(ewrt);
1899             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1900             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1901             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1902             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
1903
1904             fscal            = felec;
1905
1906             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1907
1908             /* Calculate temporary vectorial force */
1909             tx               = _mm_mul_pd(fscal,dx10);
1910             ty               = _mm_mul_pd(fscal,dy10);
1911             tz               = _mm_mul_pd(fscal,dz10);
1912
1913             /* Update vectorial force */
1914             fix1             = _mm_add_pd(fix1,tx);
1915             fiy1             = _mm_add_pd(fiy1,ty);
1916             fiz1             = _mm_add_pd(fiz1,tz);
1917
1918             fjx0             = _mm_add_pd(fjx0,tx);
1919             fjy0             = _mm_add_pd(fjy0,ty);
1920             fjz0             = _mm_add_pd(fjz0,tz);
1921
1922             /**************************
1923              * CALCULATE INTERACTIONS *
1924              **************************/
1925
1926             r11              = _mm_mul_pd(rsq11,rinv11);
1927
1928             /* EWALD ELECTROSTATICS */
1929
1930             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1931             ewrt             = _mm_mul_pd(r11,ewtabscale);
1932             ewitab           = _mm_cvttpd_epi32(ewrt);
1933             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1934             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1935             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1936             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
1937
1938             fscal            = felec;
1939
1940             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1941
1942             /* Calculate temporary vectorial force */
1943             tx               = _mm_mul_pd(fscal,dx11);
1944             ty               = _mm_mul_pd(fscal,dy11);
1945             tz               = _mm_mul_pd(fscal,dz11);
1946
1947             /* Update vectorial force */
1948             fix1             = _mm_add_pd(fix1,tx);
1949             fiy1             = _mm_add_pd(fiy1,ty);
1950             fiz1             = _mm_add_pd(fiz1,tz);
1951
1952             fjx1             = _mm_add_pd(fjx1,tx);
1953             fjy1             = _mm_add_pd(fjy1,ty);
1954             fjz1             = _mm_add_pd(fjz1,tz);
1955
1956             /**************************
1957              * CALCULATE INTERACTIONS *
1958              **************************/
1959
1960             r12              = _mm_mul_pd(rsq12,rinv12);
1961
1962             /* EWALD ELECTROSTATICS */
1963
1964             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1965             ewrt             = _mm_mul_pd(r12,ewtabscale);
1966             ewitab           = _mm_cvttpd_epi32(ewrt);
1967             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1968             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1969             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1970             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
1971
1972             fscal            = felec;
1973
1974             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1975
1976             /* Calculate temporary vectorial force */
1977             tx               = _mm_mul_pd(fscal,dx12);
1978             ty               = _mm_mul_pd(fscal,dy12);
1979             tz               = _mm_mul_pd(fscal,dz12);
1980
1981             /* Update vectorial force */
1982             fix1             = _mm_add_pd(fix1,tx);
1983             fiy1             = _mm_add_pd(fiy1,ty);
1984             fiz1             = _mm_add_pd(fiz1,tz);
1985
1986             fjx2             = _mm_add_pd(fjx2,tx);
1987             fjy2             = _mm_add_pd(fjy2,ty);
1988             fjz2             = _mm_add_pd(fjz2,tz);
1989
1990             /**************************
1991              * CALCULATE INTERACTIONS *
1992              **************************/
1993
1994             r20              = _mm_mul_pd(rsq20,rinv20);
1995
1996             /* EWALD ELECTROSTATICS */
1997
1998             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1999             ewrt             = _mm_mul_pd(r20,ewtabscale);
2000             ewitab           = _mm_cvttpd_epi32(ewrt);
2001             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2002             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2003             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2004             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
2005
2006             fscal            = felec;
2007
2008             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2009
2010             /* Calculate temporary vectorial force */
2011             tx               = _mm_mul_pd(fscal,dx20);
2012             ty               = _mm_mul_pd(fscal,dy20);
2013             tz               = _mm_mul_pd(fscal,dz20);
2014
2015             /* Update vectorial force */
2016             fix2             = _mm_add_pd(fix2,tx);
2017             fiy2             = _mm_add_pd(fiy2,ty);
2018             fiz2             = _mm_add_pd(fiz2,tz);
2019
2020             fjx0             = _mm_add_pd(fjx0,tx);
2021             fjy0             = _mm_add_pd(fjy0,ty);
2022             fjz0             = _mm_add_pd(fjz0,tz);
2023
2024             /**************************
2025              * CALCULATE INTERACTIONS *
2026              **************************/
2027
2028             r21              = _mm_mul_pd(rsq21,rinv21);
2029
2030             /* EWALD ELECTROSTATICS */
2031
2032             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2033             ewrt             = _mm_mul_pd(r21,ewtabscale);
2034             ewitab           = _mm_cvttpd_epi32(ewrt);
2035             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2036             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2037             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2038             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
2039
2040             fscal            = felec;
2041
2042             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2043
2044             /* Calculate temporary vectorial force */
2045             tx               = _mm_mul_pd(fscal,dx21);
2046             ty               = _mm_mul_pd(fscal,dy21);
2047             tz               = _mm_mul_pd(fscal,dz21);
2048
2049             /* Update vectorial force */
2050             fix2             = _mm_add_pd(fix2,tx);
2051             fiy2             = _mm_add_pd(fiy2,ty);
2052             fiz2             = _mm_add_pd(fiz2,tz);
2053
2054             fjx1             = _mm_add_pd(fjx1,tx);
2055             fjy1             = _mm_add_pd(fjy1,ty);
2056             fjz1             = _mm_add_pd(fjz1,tz);
2057
2058             /**************************
2059              * CALCULATE INTERACTIONS *
2060              **************************/
2061
2062             r22              = _mm_mul_pd(rsq22,rinv22);
2063
2064             /* EWALD ELECTROSTATICS */
2065
2066             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2067             ewrt             = _mm_mul_pd(r22,ewtabscale);
2068             ewitab           = _mm_cvttpd_epi32(ewrt);
2069             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2070             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2071             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2072             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
2073
2074             fscal            = felec;
2075
2076             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2077
2078             /* Calculate temporary vectorial force */
2079             tx               = _mm_mul_pd(fscal,dx22);
2080             ty               = _mm_mul_pd(fscal,dy22);
2081             tz               = _mm_mul_pd(fscal,dz22);
2082
2083             /* Update vectorial force */
2084             fix2             = _mm_add_pd(fix2,tx);
2085             fiy2             = _mm_add_pd(fiy2,ty);
2086             fiz2             = _mm_add_pd(fiz2,tz);
2087
2088             fjx2             = _mm_add_pd(fjx2,tx);
2089             fjy2             = _mm_add_pd(fjy2,ty);
2090             fjz2             = _mm_add_pd(fjz2,tz);
2091
2092             gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
2093
2094             /* Inner loop uses 324 flops */
2095         }
2096
2097         /* End of innermost loop */
2098
2099         gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
2100                                               f+i_coord_offset,fshift+i_shift_offset);
2101
2102         /* Increment number of inner iterations */
2103         inneriter                  += j_index_end - j_index_start;
2104
2105         /* Outer loop uses 18 flops */
2106     }
2107
2108     /* Increment number of outer iterations */
2109     outeriter        += nri;
2110
2111     /* Update outer/inner flops */
2112
2113     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3W3_F,outeriter*18 + inneriter*324);
2114 }