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) 2012,2013,2014, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS sse2_double kernel generator.
  37  */
  38 #include "config.h"
  39
  40 #include <math.h>
  41
  42 #include "../nb_kernel.h"
  43 #include "gromacs/legacyheaders/types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "gromacs/legacyheaders/nrnb.h"
  46
  47 #include "gromacs/simd/math_x86_sse2_double.h"
  48 #include "kernelutil_x86_sse2_double.h"
  49
  50 /*
  51  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwNone_GeomW3W3_VF_sse2_double
  52  * Electrostatics interaction: Ewald
  53  * VdW interaction:            None
  54  * Geometry:                   Water3-Water3
  55  * Calculate force/pot:        PotentialAndForce
  56  */
  57 void
  58 nb_kernel_ElecEw_VdwNone_GeomW3W3_VF_sse2_double
  59                     (t_nblist                    * gmx_restrict       nlist,
  60                      rvec                        * gmx_restrict          xx,
  61                      rvec                        * gmx_restrict          ff,
  62                      t_forcerec                  * gmx_restrict          fr,
  63                      t_mdatoms                   * gmx_restrict     mdatoms,
  64                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  65                      t_nrnb                      * gmx_restrict        nrnb)
  66 {
  67     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  68      * just 0 for non-waters.
  69      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
  70      * jnr indices corresponding to data put in the four positions in the SIMD register.
  71      */
  72     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  73     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  74     int              jnrA,jnrB;
  75     int              j_coord_offsetA,j_coord_offsetB;
  76     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  77     real             rcutoff_scalar;
  78     real             *shiftvec,*fshift,*x,*f;
  79     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  80     int              vdwioffset0;
  81     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  82     int              vdwioffset1;
  83     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  84     int              vdwioffset2;
  85     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  86     int              vdwjidx0A,vdwjidx0B;
  87     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  88     int              vdwjidx1A,vdwjidx1B;
  89     __m128d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  90     int              vdwjidx2A,vdwjidx2B;
  91     __m128d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  92     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  93     __m128d          dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
  94     __m128d          dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
  95     __m128d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  96     __m128d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
  97     __m128d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
  98     __m128d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  99     __m128d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 100     __m128d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 101     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
 102     real             *charge;
 103     __m128i          ewitab;
 104     __m128d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 105     real             *ewtab;
 106     __m128d          dummy_mask,cutoff_mask;
 107     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 108     __m128d          one     = _mm_set1_pd(1.0);
 109     __m128d          two     = _mm_set1_pd(2.0);
 110     x                = xx[0];
 111     f                = ff[0];
 112
 113     nri              = nlist->nri;
 114     iinr             = nlist->iinr;
 115     jindex           = nlist->jindex;
 116     jjnr             = nlist->jjnr;
 117     shiftidx         = nlist->shift;
 118     gid              = nlist->gid;
 119     shiftvec         = fr->shift_vec[0];
 120     fshift           = fr->fshift[0];
 121     facel            = _mm_set1_pd(fr->epsfac);
 122     charge           = mdatoms->chargeA;
 123
 124     sh_ewald         = _mm_set1_pd(fr->ic->sh_ewald);
 125     ewtab            = fr->ic->tabq_coul_FDV0;
 126     ewtabscale       = _mm_set1_pd(fr->ic->tabq_scale);
 127     ewtabhalfspace   = _mm_set1_pd(0.5/fr->ic->tabq_scale);
 128
 129     /* Setup water-specific parameters */
 130     inr              = nlist->iinr[0];
 131     iq0              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
 132     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
 133     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
 134
 135     jq0              = _mm_set1_pd(charge[inr+0]);
 136     jq1              = _mm_set1_pd(charge[inr+1]);
 137     jq2              = _mm_set1_pd(charge[inr+2]);
 138     qq00             = _mm_mul_pd(iq0,jq0);
 139     qq01             = _mm_mul_pd(iq0,jq1);
 140     qq02             = _mm_mul_pd(iq0,jq2);
 141     qq10             = _mm_mul_pd(iq1,jq0);
 142     qq11             = _mm_mul_pd(iq1,jq1);
 143     qq12             = _mm_mul_pd(iq1,jq2);
 144     qq20             = _mm_mul_pd(iq2,jq0);
 145     qq21             = _mm_mul_pd(iq2,jq1);
 146     qq22             = _mm_mul_pd(iq2,jq2);
 147
 148     /* Avoid stupid compiler warnings */
 149     jnrA = jnrB = 0;
 150     j_coord_offsetA = 0;
 151     j_coord_offsetB = 0;
 152
 153     outeriter        = 0;
 154     inneriter        = 0;
 155
 156     /* Start outer loop over neighborlists */
 157     for(iidx=0; iidx<nri; iidx++)
 158     {
 159         /* Load shift vector for this list */
 160         i_shift_offset   = DIM*shiftidx[iidx];
 161
 162         /* Load limits for loop over neighbors */
 163         j_index_start    = jindex[iidx];
 164         j_index_end      = jindex[iidx+1];
 165
 166         /* Get outer coordinate index */
 167         inr              = iinr[iidx];
 168         i_coord_offset   = DIM*inr;
 169
 170         /* Load i particle coords and add shift vector */
 171         gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 172                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 173
 174         fix0             = _mm_setzero_pd();
 175         fiy0             = _mm_setzero_pd();
 176         fiz0             = _mm_setzero_pd();
 177         fix1             = _mm_setzero_pd();
 178         fiy1             = _mm_setzero_pd();
 179         fiz1             = _mm_setzero_pd();
 180         fix2             = _mm_setzero_pd();
 181         fiy2             = _mm_setzero_pd();
 182         fiz2             = _mm_setzero_pd();
 183
 184         /* Reset potential sums */
 185         velecsum         = _mm_setzero_pd();
 186
 187         /* Start inner kernel loop */
 188         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 189         {
 190
 191             /* Get j neighbor index, and coordinate index */
 192             jnrA             = jjnr[jidx];
 193             jnrB             = jjnr[jidx+1];
 194             j_coord_offsetA  = DIM*jnrA;
 195             j_coord_offsetB  = DIM*jnrB;
 196
 197             /* load j atom coordinates */
 198             gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 199                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 200
 201             /* Calculate displacement vector */
 202             dx00             = _mm_sub_pd(ix0,jx0);
 203             dy00             = _mm_sub_pd(iy0,jy0);
 204             dz00             = _mm_sub_pd(iz0,jz0);
 205             dx01             = _mm_sub_pd(ix0,jx1);
 206             dy01             = _mm_sub_pd(iy0,jy1);
 207             dz01             = _mm_sub_pd(iz0,jz1);
 208             dx02             = _mm_sub_pd(ix0,jx2);
 209             dy02             = _mm_sub_pd(iy0,jy2);
 210             dz02             = _mm_sub_pd(iz0,jz2);
 211             dx10             = _mm_sub_pd(ix1,jx0);
 212             dy10             = _mm_sub_pd(iy1,jy0);
 213             dz10             = _mm_sub_pd(iz1,jz0);
 214             dx11             = _mm_sub_pd(ix1,jx1);
 215             dy11             = _mm_sub_pd(iy1,jy1);
 216             dz11             = _mm_sub_pd(iz1,jz1);
 217             dx12             = _mm_sub_pd(ix1,jx2);
 218             dy12             = _mm_sub_pd(iy1,jy2);
 219             dz12             = _mm_sub_pd(iz1,jz2);
 220             dx20             = _mm_sub_pd(ix2,jx0);
 221             dy20             = _mm_sub_pd(iy2,jy0);
 222             dz20             = _mm_sub_pd(iz2,jz0);
 223             dx21             = _mm_sub_pd(ix2,jx1);
 224             dy21             = _mm_sub_pd(iy2,jy1);
 225             dz21             = _mm_sub_pd(iz2,jz1);
 226             dx22             = _mm_sub_pd(ix2,jx2);
 227             dy22             = _mm_sub_pd(iy2,jy2);
 228             dz22             = _mm_sub_pd(iz2,jz2);
 229
 230             /* Calculate squared distance and things based on it */
 231             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 232             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
 233             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
 234             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 235             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 236             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 237             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 238             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 239             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 240
 241             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 242             rinv01           = gmx_mm_invsqrt_pd(rsq01);
 243             rinv02           = gmx_mm_invsqrt_pd(rsq02);
 244             rinv10           = gmx_mm_invsqrt_pd(rsq10);
 245             rinv11           = gmx_mm_invsqrt_pd(rsq11);
 246             rinv12           = gmx_mm_invsqrt_pd(rsq12);
 247             rinv20           = gmx_mm_invsqrt_pd(rsq20);
 248             rinv21           = gmx_mm_invsqrt_pd(rsq21);
 249             rinv22           = gmx_mm_invsqrt_pd(rsq22);
 250
 251             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 252             rinvsq01         = _mm_mul_pd(rinv01,rinv01);
 253             rinvsq02         = _mm_mul_pd(rinv02,rinv02);
 254             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
 255             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
 256             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
 257             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
 258             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
 259             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
 260
 261             fjx0             = _mm_setzero_pd();
 262             fjy0             = _mm_setzero_pd();
 263             fjz0             = _mm_setzero_pd();
 264             fjx1             = _mm_setzero_pd();
 265             fjy1             = _mm_setzero_pd();
 266             fjz1             = _mm_setzero_pd();
 267             fjx2             = _mm_setzero_pd();
 268             fjy2             = _mm_setzero_pd();
 269             fjz2             = _mm_setzero_pd();
 270
 271             /**************************
 272              * CALCULATE INTERACTIONS *
 273              **************************/
 274
 275             r00              = _mm_mul_pd(rsq00,rinv00);
 276
 277             /* EWALD ELECTROSTATICS */
 278
 279             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 280             ewrt             = _mm_mul_pd(r00,ewtabscale);
 281             ewitab           = _mm_cvttpd_epi32(ewrt);
 282             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 283             ewitab           = _mm_slli_epi32(ewitab,2);
 284             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 285             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 286             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 287             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 288             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 289             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 290             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 291             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 292             velec            = _mm_mul_pd(qq00,_mm_sub_pd(rinv00,velec));
 293             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
 294
 295             /* Update potential sum for this i atom from the interaction with this j atom. */
 296             velecsum         = _mm_add_pd(velecsum,velec);
 297
 298             fscal            = felec;
 299
 300             /* Calculate temporary vectorial force */
 301             tx               = _mm_mul_pd(fscal,dx00);
 302             ty               = _mm_mul_pd(fscal,dy00);
 303             tz               = _mm_mul_pd(fscal,dz00);
 304
 305             /* Update vectorial force */
 306             fix0             = _mm_add_pd(fix0,tx);
 307             fiy0             = _mm_add_pd(fiy0,ty);
 308             fiz0             = _mm_add_pd(fiz0,tz);
 309
 310             fjx0             = _mm_add_pd(fjx0,tx);
 311             fjy0             = _mm_add_pd(fjy0,ty);
 312             fjz0             = _mm_add_pd(fjz0,tz);
 313
 314             /**************************
 315              * CALCULATE INTERACTIONS *
 316              **************************/
 317
 318             r01              = _mm_mul_pd(rsq01,rinv01);
 319
 320             /* EWALD ELECTROSTATICS */
 321
 322             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 323             ewrt             = _mm_mul_pd(r01,ewtabscale);
 324             ewitab           = _mm_cvttpd_epi32(ewrt);
 325             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 326             ewitab           = _mm_slli_epi32(ewitab,2);
 327             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 328             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 329             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 330             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 331             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 332             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 333             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 334             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 335             velec            = _mm_mul_pd(qq01,_mm_sub_pd(rinv01,velec));
 336             felec            = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
 337
 338             /* Update potential sum for this i atom from the interaction with this j atom. */
 339             velecsum         = _mm_add_pd(velecsum,velec);
 340
 341             fscal            = felec;
 342
 343             /* Calculate temporary vectorial force */
 344             tx               = _mm_mul_pd(fscal,dx01);
 345             ty               = _mm_mul_pd(fscal,dy01);
 346             tz               = _mm_mul_pd(fscal,dz01);
 347
 348             /* Update vectorial force */
 349             fix0             = _mm_add_pd(fix0,tx);
 350             fiy0             = _mm_add_pd(fiy0,ty);
 351             fiz0             = _mm_add_pd(fiz0,tz);
 352
 353             fjx1             = _mm_add_pd(fjx1,tx);
 354             fjy1             = _mm_add_pd(fjy1,ty);
 355             fjz1             = _mm_add_pd(fjz1,tz);
 356
 357             /**************************
 358              * CALCULATE INTERACTIONS *
 359              **************************/
 360
 361             r02              = _mm_mul_pd(rsq02,rinv02);
 362
 363             /* EWALD ELECTROSTATICS */
 364
 365             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 366             ewrt             = _mm_mul_pd(r02,ewtabscale);
 367             ewitab           = _mm_cvttpd_epi32(ewrt);
 368             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 369             ewitab           = _mm_slli_epi32(ewitab,2);
 370             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 371             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 372             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 373             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 374             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 375             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 376             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 377             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 378             velec            = _mm_mul_pd(qq02,_mm_sub_pd(rinv02,velec));
 379             felec            = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
 380
 381             /* Update potential sum for this i atom from the interaction with this j atom. */
 382             velecsum         = _mm_add_pd(velecsum,velec);
 383
 384             fscal            = felec;
 385
 386             /* Calculate temporary vectorial force */
 387             tx               = _mm_mul_pd(fscal,dx02);
 388             ty               = _mm_mul_pd(fscal,dy02);
 389             tz               = _mm_mul_pd(fscal,dz02);
 390
 391             /* Update vectorial force */
 392             fix0             = _mm_add_pd(fix0,tx);
 393             fiy0             = _mm_add_pd(fiy0,ty);
 394             fiz0             = _mm_add_pd(fiz0,tz);
 395
 396             fjx2             = _mm_add_pd(fjx2,tx);
 397             fjy2             = _mm_add_pd(fjy2,ty);
 398             fjz2             = _mm_add_pd(fjz2,tz);
 399
 400             /**************************
 401              * CALCULATE INTERACTIONS *
 402              **************************/
 403
 404             r10              = _mm_mul_pd(rsq10,rinv10);
 405
 406             /* EWALD ELECTROSTATICS */
 407
 408             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 409             ewrt             = _mm_mul_pd(r10,ewtabscale);
 410             ewitab           = _mm_cvttpd_epi32(ewrt);
 411             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 412             ewitab           = _mm_slli_epi32(ewitab,2);
 413             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 414             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 415             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 416             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 417             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 418             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 419             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 420             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 421             velec            = _mm_mul_pd(qq10,_mm_sub_pd(rinv10,velec));
 422             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
 423
 424             /* Update potential sum for this i atom from the interaction with this j atom. */
 425             velecsum         = _mm_add_pd(velecsum,velec);
 426
 427             fscal            = felec;
 428
 429             /* Calculate temporary vectorial force */
 430             tx               = _mm_mul_pd(fscal,dx10);
 431             ty               = _mm_mul_pd(fscal,dy10);
 432             tz               = _mm_mul_pd(fscal,dz10);
 433
 434             /* Update vectorial force */
 435             fix1             = _mm_add_pd(fix1,tx);
 436             fiy1             = _mm_add_pd(fiy1,ty);
 437             fiz1             = _mm_add_pd(fiz1,tz);
 438
 439             fjx0             = _mm_add_pd(fjx0,tx);
 440             fjy0             = _mm_add_pd(fjy0,ty);
 441             fjz0             = _mm_add_pd(fjz0,tz);
 442
 443             /**************************
 444              * CALCULATE INTERACTIONS *
 445              **************************/
 446
 447             r11              = _mm_mul_pd(rsq11,rinv11);
 448
 449             /* EWALD ELECTROSTATICS */
 450
 451             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 452             ewrt             = _mm_mul_pd(r11,ewtabscale);
 453             ewitab           = _mm_cvttpd_epi32(ewrt);
 454             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 455             ewitab           = _mm_slli_epi32(ewitab,2);
 456             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 457             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 458             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 459             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 460             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 461             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 462             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 463             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 464             velec            = _mm_mul_pd(qq11,_mm_sub_pd(rinv11,velec));
 465             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
 466
 467             /* Update potential sum for this i atom from the interaction with this j atom. */
 468             velecsum         = _mm_add_pd(velecsum,velec);
 469
 470             fscal            = felec;
 471
 472             /* Calculate temporary vectorial force */
 473             tx               = _mm_mul_pd(fscal,dx11);
 474             ty               = _mm_mul_pd(fscal,dy11);
 475             tz               = _mm_mul_pd(fscal,dz11);
 476
 477             /* Update vectorial force */
 478             fix1             = _mm_add_pd(fix1,tx);
 479             fiy1             = _mm_add_pd(fiy1,ty);
 480             fiz1             = _mm_add_pd(fiz1,tz);
 481
 482             fjx1             = _mm_add_pd(fjx1,tx);
 483             fjy1             = _mm_add_pd(fjy1,ty);
 484             fjz1             = _mm_add_pd(fjz1,tz);
 485
 486             /**************************
 487              * CALCULATE INTERACTIONS *
 488              **************************/
 489
 490             r12              = _mm_mul_pd(rsq12,rinv12);
 491
 492             /* EWALD ELECTROSTATICS */
 493
 494             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 495             ewrt             = _mm_mul_pd(r12,ewtabscale);
 496             ewitab           = _mm_cvttpd_epi32(ewrt);
 497             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 498             ewitab           = _mm_slli_epi32(ewitab,2);
 499             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 500             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 501             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 502             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 503             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 504             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 505             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 506             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 507             velec            = _mm_mul_pd(qq12,_mm_sub_pd(rinv12,velec));
 508             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
 509
 510             /* Update potential sum for this i atom from the interaction with this j atom. */
 511             velecsum         = _mm_add_pd(velecsum,velec);
 512
 513             fscal            = felec;
 514
 515             /* Calculate temporary vectorial force */
 516             tx               = _mm_mul_pd(fscal,dx12);
 517             ty               = _mm_mul_pd(fscal,dy12);
 518             tz               = _mm_mul_pd(fscal,dz12);
 519
 520             /* Update vectorial force */
 521             fix1             = _mm_add_pd(fix1,tx);
 522             fiy1             = _mm_add_pd(fiy1,ty);
 523             fiz1             = _mm_add_pd(fiz1,tz);
 524
 525             fjx2             = _mm_add_pd(fjx2,tx);
 526             fjy2             = _mm_add_pd(fjy2,ty);
 527             fjz2             = _mm_add_pd(fjz2,tz);
 528
 529             /**************************
 530              * CALCULATE INTERACTIONS *
 531              **************************/
 532
 533             r20              = _mm_mul_pd(rsq20,rinv20);
 534
 535             /* EWALD ELECTROSTATICS */
 536
 537             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 538             ewrt             = _mm_mul_pd(r20,ewtabscale);
 539             ewitab           = _mm_cvttpd_epi32(ewrt);
 540             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 541             ewitab           = _mm_slli_epi32(ewitab,2);
 542             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 543             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 544             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 545             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 546             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 547             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 548             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 549             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 550             velec            = _mm_mul_pd(qq20,_mm_sub_pd(rinv20,velec));
 551             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
 552
 553             /* Update potential sum for this i atom from the interaction with this j atom. */
 554             velecsum         = _mm_add_pd(velecsum,velec);
 555
 556             fscal            = felec;
 557
 558             /* Calculate temporary vectorial force */
 559             tx               = _mm_mul_pd(fscal,dx20);
 560             ty               = _mm_mul_pd(fscal,dy20);
 561             tz               = _mm_mul_pd(fscal,dz20);
 562
 563             /* Update vectorial force */
 564             fix2             = _mm_add_pd(fix2,tx);
 565             fiy2             = _mm_add_pd(fiy2,ty);
 566             fiz2             = _mm_add_pd(fiz2,tz);
 567
 568             fjx0             = _mm_add_pd(fjx0,tx);
 569             fjy0             = _mm_add_pd(fjy0,ty);
 570             fjz0             = _mm_add_pd(fjz0,tz);
 571
 572             /**************************
 573              * CALCULATE INTERACTIONS *
 574              **************************/
 575
 576             r21              = _mm_mul_pd(rsq21,rinv21);
 577
 578             /* EWALD ELECTROSTATICS */
 579
 580             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 581             ewrt             = _mm_mul_pd(r21,ewtabscale);
 582             ewitab           = _mm_cvttpd_epi32(ewrt);
 583             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 584             ewitab           = _mm_slli_epi32(ewitab,2);
 585             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 586             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 587             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 588             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 589             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 590             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 591             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 592             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 593             velec            = _mm_mul_pd(qq21,_mm_sub_pd(rinv21,velec));
 594             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
 595
 596             /* Update potential sum for this i atom from the interaction with this j atom. */
 597             velecsum         = _mm_add_pd(velecsum,velec);
 598
 599             fscal            = felec;
 600
 601             /* Calculate temporary vectorial force */
 602             tx               = _mm_mul_pd(fscal,dx21);
 603             ty               = _mm_mul_pd(fscal,dy21);
 604             tz               = _mm_mul_pd(fscal,dz21);
 605
 606             /* Update vectorial force */
 607             fix2             = _mm_add_pd(fix2,tx);
 608             fiy2             = _mm_add_pd(fiy2,ty);
 609             fiz2             = _mm_add_pd(fiz2,tz);
 610
 611             fjx1             = _mm_add_pd(fjx1,tx);
 612             fjy1             = _mm_add_pd(fjy1,ty);
 613             fjz1             = _mm_add_pd(fjz1,tz);
 614
 615             /**************************
 616              * CALCULATE INTERACTIONS *
 617              **************************/
 618
 619             r22              = _mm_mul_pd(rsq22,rinv22);
 620
 621             /* EWALD ELECTROSTATICS */
 622
 623             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 624             ewrt             = _mm_mul_pd(r22,ewtabscale);
 625             ewitab           = _mm_cvttpd_epi32(ewrt);
 626             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 627             ewitab           = _mm_slli_epi32(ewitab,2);
 628             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 629             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 630             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 631             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 632             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 633             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 634             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 635             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 636             velec            = _mm_mul_pd(qq22,_mm_sub_pd(rinv22,velec));
 637             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
 638
 639             /* Update potential sum for this i atom from the interaction with this j atom. */
 640             velecsum         = _mm_add_pd(velecsum,velec);
 641
 642             fscal            = felec;
 643
 644             /* Calculate temporary vectorial force */
 645             tx               = _mm_mul_pd(fscal,dx22);
 646             ty               = _mm_mul_pd(fscal,dy22);
 647             tz               = _mm_mul_pd(fscal,dz22);
 648
 649             /* Update vectorial force */
 650             fix2             = _mm_add_pd(fix2,tx);
 651             fiy2             = _mm_add_pd(fiy2,ty);
 652             fiz2             = _mm_add_pd(fiz2,tz);
 653
 654             fjx2             = _mm_add_pd(fjx2,tx);
 655             fjy2             = _mm_add_pd(fjy2,ty);
 656             fjz2             = _mm_add_pd(fjz2,tz);
 657
 658             gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 659
 660             /* Inner loop uses 369 flops */
 661         }
 662
 663         if(jidx<j_index_end)
 664         {
 665
 666             jnrA             = jjnr[jidx];
 667             j_coord_offsetA  = DIM*jnrA;
 668
 669             /* load j atom coordinates */
 670             gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 671                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 672
 673             /* Calculate displacement vector */
 674             dx00             = _mm_sub_pd(ix0,jx0);
 675             dy00             = _mm_sub_pd(iy0,jy0);
 676             dz00             = _mm_sub_pd(iz0,jz0);
 677             dx01             = _mm_sub_pd(ix0,jx1);
 678             dy01             = _mm_sub_pd(iy0,jy1);
 679             dz01             = _mm_sub_pd(iz0,jz1);
 680             dx02             = _mm_sub_pd(ix0,jx2);
 681             dy02             = _mm_sub_pd(iy0,jy2);
 682             dz02             = _mm_sub_pd(iz0,jz2);
 683             dx10             = _mm_sub_pd(ix1,jx0);
 684             dy10             = _mm_sub_pd(iy1,jy0);
 685             dz10             = _mm_sub_pd(iz1,jz0);
 686             dx11             = _mm_sub_pd(ix1,jx1);
 687             dy11             = _mm_sub_pd(iy1,jy1);
 688             dz11             = _mm_sub_pd(iz1,jz1);
 689             dx12             = _mm_sub_pd(ix1,jx2);
 690             dy12             = _mm_sub_pd(iy1,jy2);
 691             dz12             = _mm_sub_pd(iz1,jz2);
 692             dx20             = _mm_sub_pd(ix2,jx0);
 693             dy20             = _mm_sub_pd(iy2,jy0);
 694             dz20             = _mm_sub_pd(iz2,jz0);
 695             dx21             = _mm_sub_pd(ix2,jx1);
 696             dy21             = _mm_sub_pd(iy2,jy1);
 697             dz21             = _mm_sub_pd(iz2,jz1);
 698             dx22             = _mm_sub_pd(ix2,jx2);
 699             dy22             = _mm_sub_pd(iy2,jy2);
 700             dz22             = _mm_sub_pd(iz2,jz2);
 701
 702             /* Calculate squared distance and things based on it */
 703             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 704             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
 705             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
 706             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 707             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 708             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 709             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 710             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 711             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 712
 713             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 714             rinv01           = gmx_mm_invsqrt_pd(rsq01);
 715             rinv02           = gmx_mm_invsqrt_pd(rsq02);
 716             rinv10           = gmx_mm_invsqrt_pd(rsq10);
 717             rinv11           = gmx_mm_invsqrt_pd(rsq11);
 718             rinv12           = gmx_mm_invsqrt_pd(rsq12);
 719             rinv20           = gmx_mm_invsqrt_pd(rsq20);
 720             rinv21           = gmx_mm_invsqrt_pd(rsq21);
 721             rinv22           = gmx_mm_invsqrt_pd(rsq22);
 722
 723             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 724             rinvsq01         = _mm_mul_pd(rinv01,rinv01);
 725             rinvsq02         = _mm_mul_pd(rinv02,rinv02);
 726             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
 727             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
 728             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
 729             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
 730             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
 731             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
 732
 733             fjx0             = _mm_setzero_pd();
 734             fjy0             = _mm_setzero_pd();
 735             fjz0             = _mm_setzero_pd();
 736             fjx1             = _mm_setzero_pd();
 737             fjy1             = _mm_setzero_pd();
 738             fjz1             = _mm_setzero_pd();
 739             fjx2             = _mm_setzero_pd();
 740             fjy2             = _mm_setzero_pd();
 741             fjz2             = _mm_setzero_pd();
 742
 743             /**************************
 744              * CALCULATE INTERACTIONS *
 745              **************************/
 746
 747             r00              = _mm_mul_pd(rsq00,rinv00);
 748
 749             /* EWALD ELECTROSTATICS */
 750
 751             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 752             ewrt             = _mm_mul_pd(r00,ewtabscale);
 753             ewitab           = _mm_cvttpd_epi32(ewrt);
 754             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 755             ewitab           = _mm_slli_epi32(ewitab,2);
 756             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 757             ewtabD           = _mm_setzero_pd();
 758             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 759             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 760             ewtabFn          = _mm_setzero_pd();
 761             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 762             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 763             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 764             velec            = _mm_mul_pd(qq00,_mm_sub_pd(rinv00,velec));
 765             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
 766
 767             /* Update potential sum for this i atom from the interaction with this j atom. */
 768             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 769             velecsum         = _mm_add_pd(velecsum,velec);
 770
 771             fscal            = felec;
 772
 773             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 774
 775             /* Calculate temporary vectorial force */
 776             tx               = _mm_mul_pd(fscal,dx00);
 777             ty               = _mm_mul_pd(fscal,dy00);
 778             tz               = _mm_mul_pd(fscal,dz00);
 779
 780             /* Update vectorial force */
 781             fix0             = _mm_add_pd(fix0,tx);
 782             fiy0             = _mm_add_pd(fiy0,ty);
 783             fiz0             = _mm_add_pd(fiz0,tz);
 784
 785             fjx0             = _mm_add_pd(fjx0,tx);
 786             fjy0             = _mm_add_pd(fjy0,ty);
 787             fjz0             = _mm_add_pd(fjz0,tz);
 788
 789             /**************************
 790              * CALCULATE INTERACTIONS *
 791              **************************/
 792
 793             r01              = _mm_mul_pd(rsq01,rinv01);
 794
 795             /* EWALD ELECTROSTATICS */
 796
 797             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 798             ewrt             = _mm_mul_pd(r01,ewtabscale);
 799             ewitab           = _mm_cvttpd_epi32(ewrt);
 800             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 801             ewitab           = _mm_slli_epi32(ewitab,2);
 802             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 803             ewtabD           = _mm_setzero_pd();
 804             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 805             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 806             ewtabFn          = _mm_setzero_pd();
 807             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 808             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 809             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 810             velec            = _mm_mul_pd(qq01,_mm_sub_pd(rinv01,velec));
 811             felec            = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
 812
 813             /* Update potential sum for this i atom from the interaction with this j atom. */
 814             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 815             velecsum         = _mm_add_pd(velecsum,velec);
 816
 817             fscal            = felec;
 818
 819             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 820
 821             /* Calculate temporary vectorial force */
 822             tx               = _mm_mul_pd(fscal,dx01);
 823             ty               = _mm_mul_pd(fscal,dy01);
 824             tz               = _mm_mul_pd(fscal,dz01);
 825
 826             /* Update vectorial force */
 827             fix0             = _mm_add_pd(fix0,tx);
 828             fiy0             = _mm_add_pd(fiy0,ty);
 829             fiz0             = _mm_add_pd(fiz0,tz);
 830
 831             fjx1             = _mm_add_pd(fjx1,tx);
 832             fjy1             = _mm_add_pd(fjy1,ty);
 833             fjz1             = _mm_add_pd(fjz1,tz);
 834
 835             /**************************
 836              * CALCULATE INTERACTIONS *
 837              **************************/
 838
 839             r02              = _mm_mul_pd(rsq02,rinv02);
 840
 841             /* EWALD ELECTROSTATICS */
 842
 843             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 844             ewrt             = _mm_mul_pd(r02,ewtabscale);
 845             ewitab           = _mm_cvttpd_epi32(ewrt);
 846             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 847             ewitab           = _mm_slli_epi32(ewitab,2);
 848             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 849             ewtabD           = _mm_setzero_pd();
 850             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 851             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 852             ewtabFn          = _mm_setzero_pd();
 853             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 854             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 855             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 856             velec            = _mm_mul_pd(qq02,_mm_sub_pd(rinv02,velec));
 857             felec            = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
 858
 859             /* Update potential sum for this i atom from the interaction with this j atom. */
 860             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 861             velecsum         = _mm_add_pd(velecsum,velec);
 862
 863             fscal            = felec;
 864
 865             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 866
 867             /* Calculate temporary vectorial force */
 868             tx               = _mm_mul_pd(fscal,dx02);
 869             ty               = _mm_mul_pd(fscal,dy02);
 870             tz               = _mm_mul_pd(fscal,dz02);
 871
 872             /* Update vectorial force */
 873             fix0             = _mm_add_pd(fix0,tx);
 874             fiy0             = _mm_add_pd(fiy0,ty);
 875             fiz0             = _mm_add_pd(fiz0,tz);
 876
 877             fjx2             = _mm_add_pd(fjx2,tx);
 878             fjy2             = _mm_add_pd(fjy2,ty);
 879             fjz2             = _mm_add_pd(fjz2,tz);
 880
 881             /**************************
 882              * CALCULATE INTERACTIONS *
 883              **************************/
 884
 885             r10              = _mm_mul_pd(rsq10,rinv10);
 886
 887             /* EWALD ELECTROSTATICS */
 888
 889             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 890             ewrt             = _mm_mul_pd(r10,ewtabscale);
 891             ewitab           = _mm_cvttpd_epi32(ewrt);
 892             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 893             ewitab           = _mm_slli_epi32(ewitab,2);
 894             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 895             ewtabD           = _mm_setzero_pd();
 896             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 897             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 898             ewtabFn          = _mm_setzero_pd();
 899             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 900             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 901             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 902             velec            = _mm_mul_pd(qq10,_mm_sub_pd(rinv10,velec));
 903             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
 904
 905             /* Update potential sum for this i atom from the interaction with this j atom. */
 906             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 907             velecsum         = _mm_add_pd(velecsum,velec);
 908
 909             fscal            = felec;
 910
 911             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 912
 913             /* Calculate temporary vectorial force */
 914             tx               = _mm_mul_pd(fscal,dx10);
 915             ty               = _mm_mul_pd(fscal,dy10);
 916             tz               = _mm_mul_pd(fscal,dz10);
 917
 918             /* Update vectorial force */
 919             fix1             = _mm_add_pd(fix1,tx);
 920             fiy1             = _mm_add_pd(fiy1,ty);
 921             fiz1             = _mm_add_pd(fiz1,tz);
 922
 923             fjx0             = _mm_add_pd(fjx0,tx);
 924             fjy0             = _mm_add_pd(fjy0,ty);
 925             fjz0             = _mm_add_pd(fjz0,tz);
 926
 927             /**************************
 928              * CALCULATE INTERACTIONS *
 929              **************************/
 930
 931             r11              = _mm_mul_pd(rsq11,rinv11);
 932
 933             /* EWALD ELECTROSTATICS */
 934
 935             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 936             ewrt             = _mm_mul_pd(r11,ewtabscale);
 937             ewitab           = _mm_cvttpd_epi32(ewrt);
 938             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 939             ewitab           = _mm_slli_epi32(ewitab,2);
 940             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 941             ewtabD           = _mm_setzero_pd();
 942             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 943             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 944             ewtabFn          = _mm_setzero_pd();
 945             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 946             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 947             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 948             velec            = _mm_mul_pd(qq11,_mm_sub_pd(rinv11,velec));
 949             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
 950
 951             /* Update potential sum for this i atom from the interaction with this j atom. */
 952             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 953             velecsum         = _mm_add_pd(velecsum,velec);
 954
 955             fscal            = felec;
 956
 957             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 958
 959             /* Calculate temporary vectorial force */
 960             tx               = _mm_mul_pd(fscal,dx11);
 961             ty               = _mm_mul_pd(fscal,dy11);
 962             tz               = _mm_mul_pd(fscal,dz11);
 963
 964             /* Update vectorial force */
 965             fix1             = _mm_add_pd(fix1,tx);
 966             fiy1             = _mm_add_pd(fiy1,ty);
 967             fiz1             = _mm_add_pd(fiz1,tz);
 968
 969             fjx1             = _mm_add_pd(fjx1,tx);
 970             fjy1             = _mm_add_pd(fjy1,ty);
 971             fjz1             = _mm_add_pd(fjz1,tz);
 972
 973             /**************************
 974              * CALCULATE INTERACTIONS *
 975              **************************/
 976
 977             r12              = _mm_mul_pd(rsq12,rinv12);
 978
 979             /* EWALD ELECTROSTATICS */
 980
 981             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 982             ewrt             = _mm_mul_pd(r12,ewtabscale);
 983             ewitab           = _mm_cvttpd_epi32(ewrt);
 984             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 985             ewitab           = _mm_slli_epi32(ewitab,2);
 986             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 987             ewtabD           = _mm_setzero_pd();
 988             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 989             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 990             ewtabFn          = _mm_setzero_pd();
 991             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 992             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 993             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 994             velec            = _mm_mul_pd(qq12,_mm_sub_pd(rinv12,velec));
 995             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
 996
 997             /* Update potential sum for this i atom from the interaction with this j atom. */
 998             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 999             velecsum         = _mm_add_pd(velecsum,velec);
1000
1001             fscal            = felec;
1002
1003             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1004
1005             /* Calculate temporary vectorial force */
1006             tx               = _mm_mul_pd(fscal,dx12);
1007             ty               = _mm_mul_pd(fscal,dy12);
1008             tz               = _mm_mul_pd(fscal,dz12);
1009
1010             /* Update vectorial force */
1011             fix1             = _mm_add_pd(fix1,tx);
1012             fiy1             = _mm_add_pd(fiy1,ty);
1013             fiz1             = _mm_add_pd(fiz1,tz);
1014
1015             fjx2             = _mm_add_pd(fjx2,tx);
1016             fjy2             = _mm_add_pd(fjy2,ty);
1017             fjz2             = _mm_add_pd(fjz2,tz);
1018
1019             /**************************
1020              * CALCULATE INTERACTIONS *
1021              **************************/
1022
1023             r20              = _mm_mul_pd(rsq20,rinv20);
1024
1025             /* EWALD ELECTROSTATICS */
1026
1027             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1028             ewrt             = _mm_mul_pd(r20,ewtabscale);
1029             ewitab           = _mm_cvttpd_epi32(ewrt);
1030             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1031             ewitab           = _mm_slli_epi32(ewitab,2);
1032             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1033             ewtabD           = _mm_setzero_pd();
1034             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1035             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1036             ewtabFn          = _mm_setzero_pd();
1037             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1038             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1039             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1040             velec            = _mm_mul_pd(qq20,_mm_sub_pd(rinv20,velec));
1041             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
1042
1043             /* Update potential sum for this i atom from the interaction with this j atom. */
1044             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1045             velecsum         = _mm_add_pd(velecsum,velec);
1046
1047             fscal            = felec;
1048
1049             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1050
1051             /* Calculate temporary vectorial force */
1052             tx               = _mm_mul_pd(fscal,dx20);
1053             ty               = _mm_mul_pd(fscal,dy20);
1054             tz               = _mm_mul_pd(fscal,dz20);
1055
1056             /* Update vectorial force */
1057             fix2             = _mm_add_pd(fix2,tx);
1058             fiy2             = _mm_add_pd(fiy2,ty);
1059             fiz2             = _mm_add_pd(fiz2,tz);
1060
1061             fjx0             = _mm_add_pd(fjx0,tx);
1062             fjy0             = _mm_add_pd(fjy0,ty);
1063             fjz0             = _mm_add_pd(fjz0,tz);
1064
1065             /**************************
1066              * CALCULATE INTERACTIONS *
1067              **************************/
1068
1069             r21              = _mm_mul_pd(rsq21,rinv21);
1070
1071             /* EWALD ELECTROSTATICS */
1072
1073             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1074             ewrt             = _mm_mul_pd(r21,ewtabscale);
1075             ewitab           = _mm_cvttpd_epi32(ewrt);
1076             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1077             ewitab           = _mm_slli_epi32(ewitab,2);
1078             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1079             ewtabD           = _mm_setzero_pd();
1080             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1081             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1082             ewtabFn          = _mm_setzero_pd();
1083             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1084             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1085             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1086             velec            = _mm_mul_pd(qq21,_mm_sub_pd(rinv21,velec));
1087             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
1088
1089             /* Update potential sum for this i atom from the interaction with this j atom. */
1090             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1091             velecsum         = _mm_add_pd(velecsum,velec);
1092
1093             fscal            = felec;
1094
1095             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1096
1097             /* Calculate temporary vectorial force */
1098             tx               = _mm_mul_pd(fscal,dx21);
1099             ty               = _mm_mul_pd(fscal,dy21);
1100             tz               = _mm_mul_pd(fscal,dz21);
1101
1102             /* Update vectorial force */
1103             fix2             = _mm_add_pd(fix2,tx);
1104             fiy2             = _mm_add_pd(fiy2,ty);
1105             fiz2             = _mm_add_pd(fiz2,tz);
1106
1107             fjx1             = _mm_add_pd(fjx1,tx);
1108             fjy1             = _mm_add_pd(fjy1,ty);
1109             fjz1             = _mm_add_pd(fjz1,tz);
1110
1111             /**************************
1112              * CALCULATE INTERACTIONS *
1113              **************************/
1114
1115             r22              = _mm_mul_pd(rsq22,rinv22);
1116
1117             /* EWALD ELECTROSTATICS */
1118
1119             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1120             ewrt             = _mm_mul_pd(r22,ewtabscale);
1121             ewitab           = _mm_cvttpd_epi32(ewrt);
1122             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1123             ewitab           = _mm_slli_epi32(ewitab,2);
1124             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1125             ewtabD           = _mm_setzero_pd();
1126             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1127             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1128             ewtabFn          = _mm_setzero_pd();
1129             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1130             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1131             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1132             velec            = _mm_mul_pd(qq22,_mm_sub_pd(rinv22,velec));
1133             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
1134
1135             /* Update potential sum for this i atom from the interaction with this j atom. */
1136             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1137             velecsum         = _mm_add_pd(velecsum,velec);
1138
1139             fscal            = felec;
1140
1141             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1142
1143             /* Calculate temporary vectorial force */
1144             tx               = _mm_mul_pd(fscal,dx22);
1145             ty               = _mm_mul_pd(fscal,dy22);
1146             tz               = _mm_mul_pd(fscal,dz22);
1147
1148             /* Update vectorial force */
1149             fix2             = _mm_add_pd(fix2,tx);
1150             fiy2             = _mm_add_pd(fiy2,ty);
1151             fiz2             = _mm_add_pd(fiz2,tz);
1152
1153             fjx2             = _mm_add_pd(fjx2,tx);
1154             fjy2             = _mm_add_pd(fjy2,ty);
1155             fjz2             = _mm_add_pd(fjz2,tz);
1156
1157             gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1158
1159             /* Inner loop uses 369 flops */
1160         }
1161
1162         /* End of innermost loop */
1163
1164         gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1165                                               f+i_coord_offset,fshift+i_shift_offset);
1166
1167         ggid                        = gid[iidx];
1168         /* Update potential energies */
1169         gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1170
1171         /* Increment number of inner iterations */
1172         inneriter                  += j_index_end - j_index_start;
1173
1174         /* Outer loop uses 19 flops */
1175     }
1176
1177     /* Increment number of outer iterations */
1178     outeriter        += nri;
1179
1180     /* Update outer/inner flops */
1181
1182     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3W3_VF,outeriter*19 + inneriter*369);
1183 }
1184 /*
1185  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwNone_GeomW3W3_F_sse2_double
1186  * Electrostatics interaction: Ewald
1187  * VdW interaction:            None
1188  * Geometry:                   Water3-Water3
1189  * Calculate force/pot:        Force
1190  */
1191 void
1192 nb_kernel_ElecEw_VdwNone_GeomW3W3_F_sse2_double
1193                     (t_nblist                    * gmx_restrict       nlist,
1194                      rvec                        * gmx_restrict          xx,
1195                      rvec                        * gmx_restrict          ff,
1196                      t_forcerec                  * gmx_restrict          fr,
1197                      t_mdatoms                   * gmx_restrict     mdatoms,
1198                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1199                      t_nrnb                      * gmx_restrict        nrnb)
1200 {
1201     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1202      * just 0 for non-waters.
1203      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
1204      * jnr indices corresponding to data put in the four positions in the SIMD register.
1205      */
1206     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1207     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1208     int              jnrA,jnrB;
1209     int              j_coord_offsetA,j_coord_offsetB;
1210     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1211     real             rcutoff_scalar;
1212     real             *shiftvec,*fshift,*x,*f;
1213     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1214     int              vdwioffset0;
1215     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1216     int              vdwioffset1;
1217     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1218     int              vdwioffset2;
1219     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1220     int              vdwjidx0A,vdwjidx0B;
1221     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1222     int              vdwjidx1A,vdwjidx1B;
1223     __m128d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1224     int              vdwjidx2A,vdwjidx2B;
1225     __m128d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1226     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1227     __m128d          dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1228     __m128d          dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1229     __m128d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1230     __m128d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1231     __m128d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1232     __m128d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1233     __m128d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1234     __m128d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1235     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
1236     real             *charge;
1237     __m128i          ewitab;
1238     __m128d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1239     real             *ewtab;
1240     __m128d          dummy_mask,cutoff_mask;
1241     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
1242     __m128d          one     = _mm_set1_pd(1.0);
1243     __m128d          two     = _mm_set1_pd(2.0);
1244     x                = xx[0];
1245     f                = ff[0];
1246
1247     nri              = nlist->nri;
1248     iinr             = nlist->iinr;
1249     jindex           = nlist->jindex;
1250     jjnr             = nlist->jjnr;
1251     shiftidx         = nlist->shift;
1252     gid              = nlist->gid;
1253     shiftvec         = fr->shift_vec[0];
1254     fshift           = fr->fshift[0];
1255     facel            = _mm_set1_pd(fr->epsfac);
1256     charge           = mdatoms->chargeA;
1257
1258     sh_ewald         = _mm_set1_pd(fr->ic->sh_ewald);
1259     ewtab            = fr->ic->tabq_coul_F;
1260     ewtabscale       = _mm_set1_pd(fr->ic->tabq_scale);
1261     ewtabhalfspace   = _mm_set1_pd(0.5/fr->ic->tabq_scale);
1262
1263     /* Setup water-specific parameters */
1264     inr              = nlist->iinr[0];
1265     iq0              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
1266     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
1267     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
1268
1269     jq0              = _mm_set1_pd(charge[inr+0]);
1270     jq1              = _mm_set1_pd(charge[inr+1]);
1271     jq2              = _mm_set1_pd(charge[inr+2]);
1272     qq00             = _mm_mul_pd(iq0,jq0);
1273     qq01             = _mm_mul_pd(iq0,jq1);
1274     qq02             = _mm_mul_pd(iq0,jq2);
1275     qq10             = _mm_mul_pd(iq1,jq0);
1276     qq11             = _mm_mul_pd(iq1,jq1);
1277     qq12             = _mm_mul_pd(iq1,jq2);
1278     qq20             = _mm_mul_pd(iq2,jq0);
1279     qq21             = _mm_mul_pd(iq2,jq1);
1280     qq22             = _mm_mul_pd(iq2,jq2);
1281
1282     /* Avoid stupid compiler warnings */
1283     jnrA = jnrB = 0;
1284     j_coord_offsetA = 0;
1285     j_coord_offsetB = 0;
1286
1287     outeriter        = 0;
1288     inneriter        = 0;
1289
1290     /* Start outer loop over neighborlists */
1291     for(iidx=0; iidx<nri; iidx++)
1292     {
1293         /* Load shift vector for this list */
1294         i_shift_offset   = DIM*shiftidx[iidx];
1295
1296         /* Load limits for loop over neighbors */
1297         j_index_start    = jindex[iidx];
1298         j_index_end      = jindex[iidx+1];
1299
1300         /* Get outer coordinate index */
1301         inr              = iinr[iidx];
1302         i_coord_offset   = DIM*inr;
1303
1304         /* Load i particle coords and add shift vector */
1305         gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1306                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1307
1308         fix0             = _mm_setzero_pd();
1309         fiy0             = _mm_setzero_pd();
1310         fiz0             = _mm_setzero_pd();
1311         fix1             = _mm_setzero_pd();
1312         fiy1             = _mm_setzero_pd();
1313         fiz1             = _mm_setzero_pd();
1314         fix2             = _mm_setzero_pd();
1315         fiy2             = _mm_setzero_pd();
1316         fiz2             = _mm_setzero_pd();
1317
1318         /* Start inner kernel loop */
1319         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
1320         {
1321
1322             /* Get j neighbor index, and coordinate index */
1323             jnrA             = jjnr[jidx];
1324             jnrB             = jjnr[jidx+1];
1325             j_coord_offsetA  = DIM*jnrA;
1326             j_coord_offsetB  = DIM*jnrB;
1327
1328             /* load j atom coordinates */
1329             gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1330                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1331
1332             /* Calculate displacement vector */
1333             dx00             = _mm_sub_pd(ix0,jx0);
1334             dy00             = _mm_sub_pd(iy0,jy0);
1335             dz00             = _mm_sub_pd(iz0,jz0);
1336             dx01             = _mm_sub_pd(ix0,jx1);
1337             dy01             = _mm_sub_pd(iy0,jy1);
1338             dz01             = _mm_sub_pd(iz0,jz1);
1339             dx02             = _mm_sub_pd(ix0,jx2);
1340             dy02             = _mm_sub_pd(iy0,jy2);
1341             dz02             = _mm_sub_pd(iz0,jz2);
1342             dx10             = _mm_sub_pd(ix1,jx0);
1343             dy10             = _mm_sub_pd(iy1,jy0);
1344             dz10             = _mm_sub_pd(iz1,jz0);
1345             dx11             = _mm_sub_pd(ix1,jx1);
1346             dy11             = _mm_sub_pd(iy1,jy1);
1347             dz11             = _mm_sub_pd(iz1,jz1);
1348             dx12             = _mm_sub_pd(ix1,jx2);
1349             dy12             = _mm_sub_pd(iy1,jy2);
1350             dz12             = _mm_sub_pd(iz1,jz2);
1351             dx20             = _mm_sub_pd(ix2,jx0);
1352             dy20             = _mm_sub_pd(iy2,jy0);
1353             dz20             = _mm_sub_pd(iz2,jz0);
1354             dx21             = _mm_sub_pd(ix2,jx1);
1355             dy21             = _mm_sub_pd(iy2,jy1);
1356             dz21             = _mm_sub_pd(iz2,jz1);
1357             dx22             = _mm_sub_pd(ix2,jx2);
1358             dy22             = _mm_sub_pd(iy2,jy2);
1359             dz22             = _mm_sub_pd(iz2,jz2);
1360
1361             /* Calculate squared distance and things based on it */
1362             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1363             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
1364             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
1365             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
1366             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1367             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1368             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
1369             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1370             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1371
1372             rinv00           = gmx_mm_invsqrt_pd(rsq00);
1373             rinv01           = gmx_mm_invsqrt_pd(rsq01);
1374             rinv02           = gmx_mm_invsqrt_pd(rsq02);
1375             rinv10           = gmx_mm_invsqrt_pd(rsq10);
1376             rinv11           = gmx_mm_invsqrt_pd(rsq11);
1377             rinv12           = gmx_mm_invsqrt_pd(rsq12);
1378             rinv20           = gmx_mm_invsqrt_pd(rsq20);
1379             rinv21           = gmx_mm_invsqrt_pd(rsq21);
1380             rinv22           = gmx_mm_invsqrt_pd(rsq22);
1381
1382             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
1383             rinvsq01         = _mm_mul_pd(rinv01,rinv01);
1384             rinvsq02         = _mm_mul_pd(rinv02,rinv02);
1385             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
1386             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
1387             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
1388             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
1389             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
1390             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
1391
1392             fjx0             = _mm_setzero_pd();
1393             fjy0             = _mm_setzero_pd();
1394             fjz0             = _mm_setzero_pd();
1395             fjx1             = _mm_setzero_pd();
1396             fjy1             = _mm_setzero_pd();
1397             fjz1             = _mm_setzero_pd();
1398             fjx2             = _mm_setzero_pd();
1399             fjy2             = _mm_setzero_pd();
1400             fjz2             = _mm_setzero_pd();
1401
1402             /**************************
1403              * CALCULATE INTERACTIONS *
1404              **************************/
1405
1406             r00              = _mm_mul_pd(rsq00,rinv00);
1407
1408             /* EWALD ELECTROSTATICS */
1409
1410             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1411             ewrt             = _mm_mul_pd(r00,ewtabscale);
1412             ewitab           = _mm_cvttpd_epi32(ewrt);
1413             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1414             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1415                                          &ewtabF,&ewtabFn);
1416             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1417             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
1418
1419             fscal            = felec;
1420
1421             /* Calculate temporary vectorial force */
1422             tx               = _mm_mul_pd(fscal,dx00);
1423             ty               = _mm_mul_pd(fscal,dy00);
1424             tz               = _mm_mul_pd(fscal,dz00);
1425
1426             /* Update vectorial force */
1427             fix0             = _mm_add_pd(fix0,tx);
1428             fiy0             = _mm_add_pd(fiy0,ty);
1429             fiz0             = _mm_add_pd(fiz0,tz);
1430
1431             fjx0             = _mm_add_pd(fjx0,tx);
1432             fjy0             = _mm_add_pd(fjy0,ty);
1433             fjz0             = _mm_add_pd(fjz0,tz);
1434
1435             /**************************
1436              * CALCULATE INTERACTIONS *
1437              **************************/
1438
1439             r01              = _mm_mul_pd(rsq01,rinv01);
1440
1441             /* EWALD ELECTROSTATICS */
1442
1443             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1444             ewrt             = _mm_mul_pd(r01,ewtabscale);
1445             ewitab           = _mm_cvttpd_epi32(ewrt);
1446             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1447             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1448                                          &ewtabF,&ewtabFn);
1449             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1450             felec            = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
1451
1452             fscal            = felec;
1453
1454             /* Calculate temporary vectorial force */
1455             tx               = _mm_mul_pd(fscal,dx01);
1456             ty               = _mm_mul_pd(fscal,dy01);
1457             tz               = _mm_mul_pd(fscal,dz01);
1458
1459             /* Update vectorial force */
1460             fix0             = _mm_add_pd(fix0,tx);
1461             fiy0             = _mm_add_pd(fiy0,ty);
1462             fiz0             = _mm_add_pd(fiz0,tz);
1463
1464             fjx1             = _mm_add_pd(fjx1,tx);
1465             fjy1             = _mm_add_pd(fjy1,ty);
1466             fjz1             = _mm_add_pd(fjz1,tz);
1467
1468             /**************************
1469              * CALCULATE INTERACTIONS *
1470              **************************/
1471
1472             r02              = _mm_mul_pd(rsq02,rinv02);
1473
1474             /* EWALD ELECTROSTATICS */
1475
1476             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1477             ewrt             = _mm_mul_pd(r02,ewtabscale);
1478             ewitab           = _mm_cvttpd_epi32(ewrt);
1479             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1480             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1481                                          &ewtabF,&ewtabFn);
1482             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1483             felec            = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
1484
1485             fscal            = felec;
1486
1487             /* Calculate temporary vectorial force */
1488             tx               = _mm_mul_pd(fscal,dx02);
1489             ty               = _mm_mul_pd(fscal,dy02);
1490             tz               = _mm_mul_pd(fscal,dz02);
1491
1492             /* Update vectorial force */
1493             fix0             = _mm_add_pd(fix0,tx);
1494             fiy0             = _mm_add_pd(fiy0,ty);
1495             fiz0             = _mm_add_pd(fiz0,tz);
1496
1497             fjx2             = _mm_add_pd(fjx2,tx);
1498             fjy2             = _mm_add_pd(fjy2,ty);
1499             fjz2             = _mm_add_pd(fjz2,tz);
1500
1501             /**************************
1502              * CALCULATE INTERACTIONS *
1503              **************************/
1504
1505             r10              = _mm_mul_pd(rsq10,rinv10);
1506
1507             /* EWALD ELECTROSTATICS */
1508
1509             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1510             ewrt             = _mm_mul_pd(r10,ewtabscale);
1511             ewitab           = _mm_cvttpd_epi32(ewrt);
1512             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1513             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1514                                          &ewtabF,&ewtabFn);
1515             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1516             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
1517
1518             fscal            = felec;
1519
1520             /* Calculate temporary vectorial force */
1521             tx               = _mm_mul_pd(fscal,dx10);
1522             ty               = _mm_mul_pd(fscal,dy10);
1523             tz               = _mm_mul_pd(fscal,dz10);
1524
1525             /* Update vectorial force */
1526             fix1             = _mm_add_pd(fix1,tx);
1527             fiy1             = _mm_add_pd(fiy1,ty);
1528             fiz1             = _mm_add_pd(fiz1,tz);
1529
1530             fjx0             = _mm_add_pd(fjx0,tx);
1531             fjy0             = _mm_add_pd(fjy0,ty);
1532             fjz0             = _mm_add_pd(fjz0,tz);
1533
1534             /**************************
1535              * CALCULATE INTERACTIONS *
1536              **************************/
1537
1538             r11              = _mm_mul_pd(rsq11,rinv11);
1539
1540             /* EWALD ELECTROSTATICS */
1541
1542             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1543             ewrt             = _mm_mul_pd(r11,ewtabscale);
1544             ewitab           = _mm_cvttpd_epi32(ewrt);
1545             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1546             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1547                                          &ewtabF,&ewtabFn);
1548             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1549             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
1550
1551             fscal            = felec;
1552
1553             /* Calculate temporary vectorial force */
1554             tx               = _mm_mul_pd(fscal,dx11);
1555             ty               = _mm_mul_pd(fscal,dy11);
1556             tz               = _mm_mul_pd(fscal,dz11);
1557
1558             /* Update vectorial force */
1559             fix1             = _mm_add_pd(fix1,tx);
1560             fiy1             = _mm_add_pd(fiy1,ty);
1561             fiz1             = _mm_add_pd(fiz1,tz);
1562
1563             fjx1             = _mm_add_pd(fjx1,tx);
1564             fjy1             = _mm_add_pd(fjy1,ty);
1565             fjz1             = _mm_add_pd(fjz1,tz);
1566
1567             /**************************
1568              * CALCULATE INTERACTIONS *
1569              **************************/
1570
1571             r12              = _mm_mul_pd(rsq12,rinv12);
1572
1573             /* EWALD ELECTROSTATICS */
1574
1575             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1576             ewrt             = _mm_mul_pd(r12,ewtabscale);
1577             ewitab           = _mm_cvttpd_epi32(ewrt);
1578             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1579             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1580                                          &ewtabF,&ewtabFn);
1581             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1582             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
1583
1584             fscal            = felec;
1585
1586             /* Calculate temporary vectorial force */
1587             tx               = _mm_mul_pd(fscal,dx12);
1588             ty               = _mm_mul_pd(fscal,dy12);
1589             tz               = _mm_mul_pd(fscal,dz12);
1590
1591             /* Update vectorial force */
1592             fix1             = _mm_add_pd(fix1,tx);
1593             fiy1             = _mm_add_pd(fiy1,ty);
1594             fiz1             = _mm_add_pd(fiz1,tz);
1595
1596             fjx2             = _mm_add_pd(fjx2,tx);
1597             fjy2             = _mm_add_pd(fjy2,ty);
1598             fjz2             = _mm_add_pd(fjz2,tz);
1599
1600             /**************************
1601              * CALCULATE INTERACTIONS *
1602              **************************/
1603
1604             r20              = _mm_mul_pd(rsq20,rinv20);
1605
1606             /* EWALD ELECTROSTATICS */
1607
1608             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1609             ewrt             = _mm_mul_pd(r20,ewtabscale);
1610             ewitab           = _mm_cvttpd_epi32(ewrt);
1611             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1612             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1613                                          &ewtabF,&ewtabFn);
1614             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1615             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
1616
1617             fscal            = felec;
1618
1619             /* Calculate temporary vectorial force */
1620             tx               = _mm_mul_pd(fscal,dx20);
1621             ty               = _mm_mul_pd(fscal,dy20);
1622             tz               = _mm_mul_pd(fscal,dz20);
1623
1624             /* Update vectorial force */
1625             fix2             = _mm_add_pd(fix2,tx);
1626             fiy2             = _mm_add_pd(fiy2,ty);
1627             fiz2             = _mm_add_pd(fiz2,tz);
1628
1629             fjx0             = _mm_add_pd(fjx0,tx);
1630             fjy0             = _mm_add_pd(fjy0,ty);
1631             fjz0             = _mm_add_pd(fjz0,tz);
1632
1633             /**************************
1634              * CALCULATE INTERACTIONS *
1635              **************************/
1636
1637             r21              = _mm_mul_pd(rsq21,rinv21);
1638
1639             /* EWALD ELECTROSTATICS */
1640
1641             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1642             ewrt             = _mm_mul_pd(r21,ewtabscale);
1643             ewitab           = _mm_cvttpd_epi32(ewrt);
1644             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1645             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1646                                          &ewtabF,&ewtabFn);
1647             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1648             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
1649
1650             fscal            = felec;
1651
1652             /* Calculate temporary vectorial force */
1653             tx               = _mm_mul_pd(fscal,dx21);
1654             ty               = _mm_mul_pd(fscal,dy21);
1655             tz               = _mm_mul_pd(fscal,dz21);
1656
1657             /* Update vectorial force */
1658             fix2             = _mm_add_pd(fix2,tx);
1659             fiy2             = _mm_add_pd(fiy2,ty);
1660             fiz2             = _mm_add_pd(fiz2,tz);
1661
1662             fjx1             = _mm_add_pd(fjx1,tx);
1663             fjy1             = _mm_add_pd(fjy1,ty);
1664             fjz1             = _mm_add_pd(fjz1,tz);
1665
1666             /**************************
1667              * CALCULATE INTERACTIONS *
1668              **************************/
1669
1670             r22              = _mm_mul_pd(rsq22,rinv22);
1671
1672             /* EWALD ELECTROSTATICS */
1673
1674             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1675             ewrt             = _mm_mul_pd(r22,ewtabscale);
1676             ewitab           = _mm_cvttpd_epi32(ewrt);
1677             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1678             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1679                                          &ewtabF,&ewtabFn);
1680             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1681             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
1682
1683             fscal            = felec;
1684
1685             /* Calculate temporary vectorial force */
1686             tx               = _mm_mul_pd(fscal,dx22);
1687             ty               = _mm_mul_pd(fscal,dy22);
1688             tz               = _mm_mul_pd(fscal,dz22);
1689
1690             /* Update vectorial force */
1691             fix2             = _mm_add_pd(fix2,tx);
1692             fiy2             = _mm_add_pd(fiy2,ty);
1693             fiz2             = _mm_add_pd(fiz2,tz);
1694
1695             fjx2             = _mm_add_pd(fjx2,tx);
1696             fjy2             = _mm_add_pd(fjy2,ty);
1697             fjz2             = _mm_add_pd(fjz2,tz);
1698
1699             gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1700
1701             /* Inner loop uses 324 flops */
1702         }
1703
1704         if(jidx<j_index_end)
1705         {
1706
1707             jnrA             = jjnr[jidx];
1708             j_coord_offsetA  = DIM*jnrA;
1709
1710             /* load j atom coordinates */
1711             gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
1712                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1713
1714             /* Calculate displacement vector */
1715             dx00             = _mm_sub_pd(ix0,jx0);
1716             dy00             = _mm_sub_pd(iy0,jy0);
1717             dz00             = _mm_sub_pd(iz0,jz0);
1718             dx01             = _mm_sub_pd(ix0,jx1);
1719             dy01             = _mm_sub_pd(iy0,jy1);
1720             dz01             = _mm_sub_pd(iz0,jz1);
1721             dx02             = _mm_sub_pd(ix0,jx2);
1722             dy02             = _mm_sub_pd(iy0,jy2);
1723             dz02             = _mm_sub_pd(iz0,jz2);
1724             dx10             = _mm_sub_pd(ix1,jx0);
1725             dy10             = _mm_sub_pd(iy1,jy0);
1726             dz10             = _mm_sub_pd(iz1,jz0);
1727             dx11             = _mm_sub_pd(ix1,jx1);
1728             dy11             = _mm_sub_pd(iy1,jy1);
1729             dz11             = _mm_sub_pd(iz1,jz1);
1730             dx12             = _mm_sub_pd(ix1,jx2);
1731             dy12             = _mm_sub_pd(iy1,jy2);
1732             dz12             = _mm_sub_pd(iz1,jz2);
1733             dx20             = _mm_sub_pd(ix2,jx0);
1734             dy20             = _mm_sub_pd(iy2,jy0);
1735             dz20             = _mm_sub_pd(iz2,jz0);
1736             dx21             = _mm_sub_pd(ix2,jx1);
1737             dy21             = _mm_sub_pd(iy2,jy1);
1738             dz21             = _mm_sub_pd(iz2,jz1);
1739             dx22             = _mm_sub_pd(ix2,jx2);
1740             dy22             = _mm_sub_pd(iy2,jy2);
1741             dz22             = _mm_sub_pd(iz2,jz2);
1742
1743             /* Calculate squared distance and things based on it */
1744             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1745             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
1746             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
1747             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
1748             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1749             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1750             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
1751             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1752             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1753
1754             rinv00           = gmx_mm_invsqrt_pd(rsq00);
1755             rinv01           = gmx_mm_invsqrt_pd(rsq01);
1756             rinv02           = gmx_mm_invsqrt_pd(rsq02);
1757             rinv10           = gmx_mm_invsqrt_pd(rsq10);
1758             rinv11           = gmx_mm_invsqrt_pd(rsq11);
1759             rinv12           = gmx_mm_invsqrt_pd(rsq12);
1760             rinv20           = gmx_mm_invsqrt_pd(rsq20);
1761             rinv21           = gmx_mm_invsqrt_pd(rsq21);
1762             rinv22           = gmx_mm_invsqrt_pd(rsq22);
1763
1764             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
1765             rinvsq01         = _mm_mul_pd(rinv01,rinv01);
1766             rinvsq02         = _mm_mul_pd(rinv02,rinv02);
1767             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
1768             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
1769             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
1770             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
1771             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
1772             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
1773
1774             fjx0             = _mm_setzero_pd();
1775             fjy0             = _mm_setzero_pd();
1776             fjz0             = _mm_setzero_pd();
1777             fjx1             = _mm_setzero_pd();
1778             fjy1             = _mm_setzero_pd();
1779             fjz1             = _mm_setzero_pd();
1780             fjx2             = _mm_setzero_pd();
1781             fjy2             = _mm_setzero_pd();
1782             fjz2             = _mm_setzero_pd();
1783
1784             /**************************
1785              * CALCULATE INTERACTIONS *
1786              **************************/
1787
1788             r00              = _mm_mul_pd(rsq00,rinv00);
1789
1790             /* EWALD ELECTROSTATICS */
1791
1792             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1793             ewrt             = _mm_mul_pd(r00,ewtabscale);
1794             ewitab           = _mm_cvttpd_epi32(ewrt);
1795             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1796             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1797             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1798             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
1799
1800             fscal            = felec;
1801
1802             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1803
1804             /* Calculate temporary vectorial force */
1805             tx               = _mm_mul_pd(fscal,dx00);
1806             ty               = _mm_mul_pd(fscal,dy00);
1807             tz               = _mm_mul_pd(fscal,dz00);
1808
1809             /* Update vectorial force */
1810             fix0             = _mm_add_pd(fix0,tx);
1811             fiy0             = _mm_add_pd(fiy0,ty);
1812             fiz0             = _mm_add_pd(fiz0,tz);
1813
1814             fjx0             = _mm_add_pd(fjx0,tx);
1815             fjy0             = _mm_add_pd(fjy0,ty);
1816             fjz0             = _mm_add_pd(fjz0,tz);
1817
1818             /**************************
1819              * CALCULATE INTERACTIONS *
1820              **************************/
1821
1822             r01              = _mm_mul_pd(rsq01,rinv01);
1823
1824             /* EWALD ELECTROSTATICS */
1825
1826             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1827             ewrt             = _mm_mul_pd(r01,ewtabscale);
1828             ewitab           = _mm_cvttpd_epi32(ewrt);
1829             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1830             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1831             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1832             felec            = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
1833
1834             fscal            = felec;
1835
1836             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1837
1838             /* Calculate temporary vectorial force */
1839             tx               = _mm_mul_pd(fscal,dx01);
1840             ty               = _mm_mul_pd(fscal,dy01);
1841             tz               = _mm_mul_pd(fscal,dz01);
1842
1843             /* Update vectorial force */
1844             fix0             = _mm_add_pd(fix0,tx);
1845             fiy0             = _mm_add_pd(fiy0,ty);
1846             fiz0             = _mm_add_pd(fiz0,tz);
1847
1848             fjx1             = _mm_add_pd(fjx1,tx);
1849             fjy1             = _mm_add_pd(fjy1,ty);
1850             fjz1             = _mm_add_pd(fjz1,tz);
1851
1852             /**************************
1853              * CALCULATE INTERACTIONS *
1854              **************************/
1855
1856             r02              = _mm_mul_pd(rsq02,rinv02);
1857
1858             /* EWALD ELECTROSTATICS */
1859
1860             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1861             ewrt             = _mm_mul_pd(r02,ewtabscale);
1862             ewitab           = _mm_cvttpd_epi32(ewrt);
1863             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1864             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1865             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1866             felec            = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
1867
1868             fscal            = felec;
1869
1870             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1871
1872             /* Calculate temporary vectorial force */
1873             tx               = _mm_mul_pd(fscal,dx02);
1874             ty               = _mm_mul_pd(fscal,dy02);
1875             tz               = _mm_mul_pd(fscal,dz02);
1876
1877             /* Update vectorial force */
1878             fix0             = _mm_add_pd(fix0,tx);
1879             fiy0             = _mm_add_pd(fiy0,ty);
1880             fiz0             = _mm_add_pd(fiz0,tz);
1881
1882             fjx2             = _mm_add_pd(fjx2,tx);
1883             fjy2             = _mm_add_pd(fjy2,ty);
1884             fjz2             = _mm_add_pd(fjz2,tz);
1885
1886             /**************************
1887              * CALCULATE INTERACTIONS *
1888              **************************/
1889
1890             r10              = _mm_mul_pd(rsq10,rinv10);
1891
1892             /* EWALD ELECTROSTATICS */
1893
1894             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1895             ewrt             = _mm_mul_pd(r10,ewtabscale);
1896             ewitab           = _mm_cvttpd_epi32(ewrt);
1897             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1898             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1899             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1900             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
1901
1902             fscal            = felec;
1903
1904             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1905
1906             /* Calculate temporary vectorial force */
1907             tx               = _mm_mul_pd(fscal,dx10);
1908             ty               = _mm_mul_pd(fscal,dy10);
1909             tz               = _mm_mul_pd(fscal,dz10);
1910
1911             /* Update vectorial force */
1912             fix1             = _mm_add_pd(fix1,tx);
1913             fiy1             = _mm_add_pd(fiy1,ty);
1914             fiz1             = _mm_add_pd(fiz1,tz);
1915
1916             fjx0             = _mm_add_pd(fjx0,tx);
1917             fjy0             = _mm_add_pd(fjy0,ty);
1918             fjz0             = _mm_add_pd(fjz0,tz);
1919
1920             /**************************
1921              * CALCULATE INTERACTIONS *
1922              **************************/
1923
1924             r11              = _mm_mul_pd(rsq11,rinv11);
1925
1926             /* EWALD ELECTROSTATICS */
1927
1928             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1929             ewrt             = _mm_mul_pd(r11,ewtabscale);
1930             ewitab           = _mm_cvttpd_epi32(ewrt);
1931             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1932             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1933             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1934             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
1935
1936             fscal            = felec;
1937
1938             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1939
1940             /* Calculate temporary vectorial force */
1941             tx               = _mm_mul_pd(fscal,dx11);
1942             ty               = _mm_mul_pd(fscal,dy11);
1943             tz               = _mm_mul_pd(fscal,dz11);
1944
1945             /* Update vectorial force */
1946             fix1             = _mm_add_pd(fix1,tx);
1947             fiy1             = _mm_add_pd(fiy1,ty);
1948             fiz1             = _mm_add_pd(fiz1,tz);
1949
1950             fjx1             = _mm_add_pd(fjx1,tx);
1951             fjy1             = _mm_add_pd(fjy1,ty);
1952             fjz1             = _mm_add_pd(fjz1,tz);
1953
1954             /**************************
1955              * CALCULATE INTERACTIONS *
1956              **************************/
1957
1958             r12              = _mm_mul_pd(rsq12,rinv12);
1959
1960             /* EWALD ELECTROSTATICS */
1961
1962             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1963             ewrt             = _mm_mul_pd(r12,ewtabscale);
1964             ewitab           = _mm_cvttpd_epi32(ewrt);
1965             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1966             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1967             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1968             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
1969
1970             fscal            = felec;
1971
1972             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1973
1974             /* Calculate temporary vectorial force */
1975             tx               = _mm_mul_pd(fscal,dx12);
1976             ty               = _mm_mul_pd(fscal,dy12);
1977             tz               = _mm_mul_pd(fscal,dz12);
1978
1979             /* Update vectorial force */
1980             fix1             = _mm_add_pd(fix1,tx);
1981             fiy1             = _mm_add_pd(fiy1,ty);
1982             fiz1             = _mm_add_pd(fiz1,tz);
1983
1984             fjx2             = _mm_add_pd(fjx2,tx);
1985             fjy2             = _mm_add_pd(fjy2,ty);
1986             fjz2             = _mm_add_pd(fjz2,tz);
1987
1988             /**************************
1989              * CALCULATE INTERACTIONS *
1990              **************************/
1991
1992             r20              = _mm_mul_pd(rsq20,rinv20);
1993
1994             /* EWALD ELECTROSTATICS */
1995
1996             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1997             ewrt             = _mm_mul_pd(r20,ewtabscale);
1998             ewitab           = _mm_cvttpd_epi32(ewrt);
1999             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2000             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2001             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2002             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
2003
2004             fscal            = felec;
2005
2006             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2007
2008             /* Calculate temporary vectorial force */
2009             tx               = _mm_mul_pd(fscal,dx20);
2010             ty               = _mm_mul_pd(fscal,dy20);
2011             tz               = _mm_mul_pd(fscal,dz20);
2012
2013             /* Update vectorial force */
2014             fix2             = _mm_add_pd(fix2,tx);
2015             fiy2             = _mm_add_pd(fiy2,ty);
2016             fiz2             = _mm_add_pd(fiz2,tz);
2017
2018             fjx0             = _mm_add_pd(fjx0,tx);
2019             fjy0             = _mm_add_pd(fjy0,ty);
2020             fjz0             = _mm_add_pd(fjz0,tz);
2021
2022             /**************************
2023              * CALCULATE INTERACTIONS *
2024              **************************/
2025
2026             r21              = _mm_mul_pd(rsq21,rinv21);
2027
2028             /* EWALD ELECTROSTATICS */
2029
2030             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2031             ewrt             = _mm_mul_pd(r21,ewtabscale);
2032             ewitab           = _mm_cvttpd_epi32(ewrt);
2033             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2034             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2035             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2036             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
2037
2038             fscal            = felec;
2039
2040             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2041
2042             /* Calculate temporary vectorial force */
2043             tx               = _mm_mul_pd(fscal,dx21);
2044             ty               = _mm_mul_pd(fscal,dy21);
2045             tz               = _mm_mul_pd(fscal,dz21);
2046
2047             /* Update vectorial force */
2048             fix2             = _mm_add_pd(fix2,tx);
2049             fiy2             = _mm_add_pd(fiy2,ty);
2050             fiz2             = _mm_add_pd(fiz2,tz);
2051
2052             fjx1             = _mm_add_pd(fjx1,tx);
2053             fjy1             = _mm_add_pd(fjy1,ty);
2054             fjz1             = _mm_add_pd(fjz1,tz);
2055
2056             /**************************
2057              * CALCULATE INTERACTIONS *
2058              **************************/
2059
2060             r22              = _mm_mul_pd(rsq22,rinv22);
2061
2062             /* EWALD ELECTROSTATICS */
2063
2064             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2065             ewrt             = _mm_mul_pd(r22,ewtabscale);
2066             ewitab           = _mm_cvttpd_epi32(ewrt);
2067             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2068             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2069             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2070             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
2071
2072             fscal            = felec;
2073
2074             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2075
2076             /* Calculate temporary vectorial force */
2077             tx               = _mm_mul_pd(fscal,dx22);
2078             ty               = _mm_mul_pd(fscal,dy22);
2079             tz               = _mm_mul_pd(fscal,dz22);
2080
2081             /* Update vectorial force */
2082             fix2             = _mm_add_pd(fix2,tx);
2083             fiy2             = _mm_add_pd(fiy2,ty);
2084             fiz2             = _mm_add_pd(fiz2,tz);
2085
2086             fjx2             = _mm_add_pd(fjx2,tx);
2087             fjy2             = _mm_add_pd(fjy2,ty);
2088             fjz2             = _mm_add_pd(fjz2,tz);
2089
2090             gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
2091
2092             /* Inner loop uses 324 flops */
2093         }
2094
2095         /* End of innermost loop */
2096
2097         gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
2098                                               f+i_coord_offset,fshift+i_shift_offset);
2099
2100         /* Increment number of inner iterations */
2101         inneriter                  += j_index_end - j_index_start;
2102
2103         /* Outer loop uses 18 flops */
2104     }
2105
2106     /* Increment number of outer iterations */
2107     outeriter        += nri;
2108
2109     /* Update outer/inner flops */
2110
2111     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3W3_F,outeriter*18 + inneriter*324);
2112 }