src/gmxlib/nonbonded/nb_kernel_avx_128_fma_double/nb_kernel_ElecGB_VdwNone_GeomP1P1_avx_128_fma_double.c

   1 /*
   2  * Note: this file was generated by the Gromacs avx_128_fma_double kernel generator.
   3  *
   4  *                This source code is part of
   5  *
   6  *                 G   R   O   M   A   C   S
   7  *
   8  * Copyright (c) 2001-2012, The GROMACS Development Team
   9  *
  10  * Gromacs is a library for molecular simulation and trajectory analysis,
  11  * written by Erik Lindahl, David van der Spoel, Berk Hess, and others - for
  12  * a full list of developers and information, check out http://www.gromacs.org
  13  *
  14  * This program is free software; you can redistribute it and/or modify it under
  15  * the terms of the GNU Lesser General Public License as published by the Free
  16  * Software Foundation; either version 2 of the License, or (at your option) any
  17  * later version.
  18  *
  19  * To help fund GROMACS development, we humbly ask that you cite
  20  * the papers people have written on it - you can find them on the website.
  21  */
  22 #ifdef HAVE_CONFIG_H
  23 #include <config.h>
  24 #endif
  25
  26 #include <math.h>
  27
  28 #include "../nb_kernel.h"
  29 #include "types/simple.h"
  30 #include "vec.h"
  31 #include "nrnb.h"
  32
  33 #include "gmx_math_x86_avx_128_fma_double.h"
  34 #include "kernelutil_x86_avx_128_fma_double.h"
  35
  36 /*
  37  * Gromacs nonbonded kernel:   nb_kernel_ElecGB_VdwNone_GeomP1P1_VF_avx_128_fma_double
  38  * Electrostatics interaction: GeneralizedBorn
  39  * VdW interaction:            None
  40  * Geometry:                   Particle-Particle
  41  * Calculate force/pot:        PotentialAndForce
  42  */
  43 void
  44 nb_kernel_ElecGB_VdwNone_GeomP1P1_VF_avx_128_fma_double
  45                     (t_nblist * gmx_restrict                nlist,
  46                      rvec * gmx_restrict                    xx,
  47                      rvec * gmx_restrict                    ff,
  48                      t_forcerec * gmx_restrict              fr,
  49                      t_mdatoms * gmx_restrict               mdatoms,
  50                      nb_kernel_data_t * gmx_restrict        kernel_data,
  51                      t_nrnb * gmx_restrict                  nrnb)
  52 {
  53     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  54      * just 0 for non-waters.
  55      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
  56      * jnr indices corresponding to data put in the four positions in the SIMD register.
  57      */
  58     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  59     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  60     int              jnrA,jnrB;
  61     int              j_coord_offsetA,j_coord_offsetB;
  62     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  63     real             rcutoff_scalar;
  64     real             *shiftvec,*fshift,*x,*f;
  65     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  66     int              vdwioffset0;
  67     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  68     int              vdwjidx0A,vdwjidx0B;
  69     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  70     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  71     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
  72     real             *charge;
  73     __m128i          gbitab;
  74     __m128d          vgb,fgb,vgbsum,dvdasum,gbscale,gbtabscale,isaprod,gbqqfactor,gbinvepsdiff,dvdaj,gbeps,twogbeps,dvdatmp;
  75     __m128d          minushalf = _mm_set1_pd(-0.5);
  76     real             *invsqrta,*dvda,*gbtab;
  77     __m128i          vfitab;
  78     __m128i          ifour       = _mm_set1_epi32(4);
  79     __m128d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
  80     real             *vftab;
  81     __m128d          dummy_mask,cutoff_mask;
  82     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
  83     __m128d          one     = _mm_set1_pd(1.0);
  84     __m128d          two     = _mm_set1_pd(2.0);
  85     x                = xx[0];
  86     f                = ff[0];
  87
  88     nri              = nlist->nri;
  89     iinr             = nlist->iinr;
  90     jindex           = nlist->jindex;
  91     jjnr             = nlist->jjnr;
  92     shiftidx         = nlist->shift;
  93     gid              = nlist->gid;
  94     shiftvec         = fr->shift_vec[0];
  95     fshift           = fr->fshift[0];
  96     facel            = _mm_set1_pd(fr->epsfac);
  97     charge           = mdatoms->chargeA;
  98
  99     invsqrta         = fr->invsqrta;
 100     dvda             = fr->dvda;
 101     gbtabscale       = _mm_set1_pd(fr->gbtab.scale);
 102     gbtab            = fr->gbtab.data;
 103     gbinvepsdiff     = _mm_set1_pd((1.0/fr->epsilon_r) - (1.0/fr->gb_epsilon_solvent));
 104
 105     /* Avoid stupid compiler warnings */
 106     jnrA = jnrB = 0;
 107     j_coord_offsetA = 0;
 108     j_coord_offsetB = 0;
 109
 110     outeriter        = 0;
 111     inneriter        = 0;
 112
 113     /* Start outer loop over neighborlists */
 114     for(iidx=0; iidx<nri; iidx++)
 115     {
 116         /* Load shift vector for this list */
 117         i_shift_offset   = DIM*shiftidx[iidx];
 118
 119         /* Load limits for loop over neighbors */
 120         j_index_start    = jindex[iidx];
 121         j_index_end      = jindex[iidx+1];
 122
 123         /* Get outer coordinate index */
 124         inr              = iinr[iidx];
 125         i_coord_offset   = DIM*inr;
 126
 127         /* Load i particle coords and add shift vector */
 128         gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 129
 130         fix0             = _mm_setzero_pd();
 131         fiy0             = _mm_setzero_pd();
 132         fiz0             = _mm_setzero_pd();
 133
 134         /* Load parameters for i particles */
 135         iq0              = _mm_mul_pd(facel,_mm_load1_pd(charge+inr+0));
 136         isai0            = _mm_load1_pd(invsqrta+inr+0);
 137
 138         /* Reset potential sums */
 139         velecsum         = _mm_setzero_pd();
 140         vgbsum           = _mm_setzero_pd();
 141         dvdasum          = _mm_setzero_pd();
 142
 143         /* Start inner kernel loop */
 144         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 145         {
 146
 147             /* Get j neighbor index, and coordinate index */
 148             jnrA             = jjnr[jidx];
 149             jnrB             = jjnr[jidx+1];
 150             j_coord_offsetA  = DIM*jnrA;
 151             j_coord_offsetB  = DIM*jnrB;
 152
 153             /* load j atom coordinates */
 154             gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 155                                               &jx0,&jy0,&jz0);
 156
 157             /* Calculate displacement vector */
 158             dx00             = _mm_sub_pd(ix0,jx0);
 159             dy00             = _mm_sub_pd(iy0,jy0);
 160             dz00             = _mm_sub_pd(iz0,jz0);
 161
 162             /* Calculate squared distance and things based on it */
 163             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 164
 165             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 166
 167             /* Load parameters for j particles */
 168             jq0              = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
 169             isaj0            = gmx_mm_load_2real_swizzle_pd(invsqrta+jnrA+0,invsqrta+jnrB+0);
 170
 171             /**************************
 172              * CALCULATE INTERACTIONS *
 173              **************************/
 174
 175             r00              = _mm_mul_pd(rsq00,rinv00);
 176
 177             /* Compute parameters for interactions between i and j atoms */
 178             qq00             = _mm_mul_pd(iq0,jq0);
 179
 180             /* GENERALIZED BORN AND COULOMB ELECTROSTATICS */
 181             isaprod          = _mm_mul_pd(isai0,isaj0);
 182             gbqqfactor       = _mm_xor_pd(signbit,_mm_mul_pd(qq00,_mm_mul_pd(isaprod,gbinvepsdiff)));
 183             gbscale          = _mm_mul_pd(isaprod,gbtabscale);
 184
 185             /* Calculate generalized born table index - this is a separate table from the normal one,
 186              * but we use the same procedure by multiplying r with scale and truncating to integer.
 187              */
 188             rt               = _mm_mul_pd(r00,gbscale);
 189             gbitab           = _mm_cvttpd_epi32(rt);
 190 #ifdef __XOP__
 191             gbeps            = _mm_frcz_pd(rt);
 192 #else
 193             gbeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 194 #endif
 195             gbitab           = _mm_slli_epi32(gbitab,2);
 196
 197             Y                = _mm_load_pd( gbtab + _mm_extract_epi32(gbitab,0) );
 198             F                = _mm_load_pd( gbtab + _mm_extract_epi32(gbitab,1) );
 199             GMX_MM_TRANSPOSE2_PD(Y,F);
 200             G                = _mm_load_pd( gbtab + _mm_extract_epi32(gbitab,0) +2);
 201             H                = _mm_load_pd( gbtab + _mm_extract_epi32(gbitab,1) +2);
 202             GMX_MM_TRANSPOSE2_PD(G,H);
 203             Fp               = _mm_macc_pd(gbeps,_mm_macc_pd(gbeps,H,G),F);
 204             VV               = _mm_macc_pd(gbeps,Fp,Y);
 205             vgb              = _mm_mul_pd(gbqqfactor,VV);
 206
 207             twogbeps         = _mm_add_pd(gbeps,gbeps);
 208             FF               = _mm_macc_pd(_mm_macc_pd(twogbeps,H,G),gbeps,Fp);
 209             fgb              = _mm_mul_pd(gbqqfactor,_mm_mul_pd(FF,gbscale));
 210             dvdatmp          = _mm_mul_pd(minushalf,_mm_macc_pd(fgb,r00,vgb));
 211             dvdasum          = _mm_add_pd(dvdasum,dvdatmp);
 212             gmx_mm_increment_2real_swizzle_pd(dvda+jnrA,dvda+jnrB,_mm_mul_pd(dvdatmp,_mm_mul_pd(isaj0,isaj0)));
 213             velec            = _mm_mul_pd(qq00,rinv00);
 214             felec            = _mm_mul_pd(_mm_msub_pd(velec,rinv00,fgb),rinv00);
 215
 216             /* Update potential sum for this i atom from the interaction with this j atom. */
 217             velecsum         = _mm_add_pd(velecsum,velec);
 218             vgbsum           = _mm_add_pd(vgbsum,vgb);
 219
 220             fscal            = felec;
 221
 222             /* Update vectorial force */
 223             fix0             = _mm_macc_pd(dx00,fscal,fix0);
 224             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
 225             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
 226
 227             gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,
 228                                                    _mm_mul_pd(dx00,fscal),
 229                                                    _mm_mul_pd(dy00,fscal),
 230                                                    _mm_mul_pd(dz00,fscal));
 231
 232             /* Inner loop uses 61 flops */
 233         }
 234
 235         if(jidx<j_index_end)
 236         {
 237
 238             jnrA             = jjnr[jidx];
 239             j_coord_offsetA  = DIM*jnrA;
 240
 241             /* load j atom coordinates */
 242             gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 243                                               &jx0,&jy0,&jz0);
 244
 245             /* Calculate displacement vector */
 246             dx00             = _mm_sub_pd(ix0,jx0);
 247             dy00             = _mm_sub_pd(iy0,jy0);
 248             dz00             = _mm_sub_pd(iz0,jz0);
 249
 250             /* Calculate squared distance and things based on it */
 251             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 252
 253             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 254
 255             /* Load parameters for j particles */
 256             jq0              = _mm_load_sd(charge+jnrA+0);
 257             isaj0            = _mm_load_sd(invsqrta+jnrA+0);
 258
 259             /**************************
 260              * CALCULATE INTERACTIONS *
 261              **************************/
 262
 263             r00              = _mm_mul_pd(rsq00,rinv00);
 264
 265             /* Compute parameters for interactions between i and j atoms */
 266             qq00             = _mm_mul_pd(iq0,jq0);
 267
 268             /* GENERALIZED BORN AND COULOMB ELECTROSTATICS */
 269             isaprod          = _mm_mul_pd(isai0,isaj0);
 270             gbqqfactor       = _mm_xor_pd(signbit,_mm_mul_pd(qq00,_mm_mul_pd(isaprod,gbinvepsdiff)));
 271             gbscale          = _mm_mul_pd(isaprod,gbtabscale);
 272
 273             /* Calculate generalized born table index - this is a separate table from the normal one,
 274              * but we use the same procedure by multiplying r with scale and truncating to integer.
 275              */
 276             rt               = _mm_mul_pd(r00,gbscale);
 277             gbitab           = _mm_cvttpd_epi32(rt);
 278 #ifdef __XOP__
 279             gbeps            = _mm_frcz_pd(rt);
 280 #else
 281             gbeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 282 #endif
 283             gbitab           = _mm_slli_epi32(gbitab,2);
 284
 285             Y                = _mm_load_pd( gbtab + _mm_extract_epi32(gbitab,0) );
 286             F                = _mm_setzero_pd();
 287             GMX_MM_TRANSPOSE2_PD(Y,F);
 288             G                = _mm_load_pd( gbtab + _mm_extract_epi32(gbitab,0) +2);
 289             H                = _mm_setzero_pd();
 290             GMX_MM_TRANSPOSE2_PD(G,H);
 291             Fp               = _mm_macc_pd(gbeps,_mm_macc_pd(gbeps,H,G),F);
 292             VV               = _mm_macc_pd(gbeps,Fp,Y);
 293             vgb              = _mm_mul_pd(gbqqfactor,VV);
 294
 295             twogbeps         = _mm_add_pd(gbeps,gbeps);
 296             FF               = _mm_macc_pd(_mm_macc_pd(twogbeps,H,G),gbeps,Fp);
 297             fgb              = _mm_mul_pd(gbqqfactor,_mm_mul_pd(FF,gbscale));
 298             dvdatmp          = _mm_mul_pd(minushalf,_mm_macc_pd(fgb,r00,vgb));
 299             dvdasum          = _mm_add_pd(dvdasum,dvdatmp);
 300             gmx_mm_increment_1real_pd(dvda+jnrA,_mm_mul_pd(dvdatmp,_mm_mul_pd(isaj0,isaj0)));
 301             velec            = _mm_mul_pd(qq00,rinv00);
 302             felec            = _mm_mul_pd(_mm_msub_pd(velec,rinv00,fgb),rinv00);
 303
 304             /* Update potential sum for this i atom from the interaction with this j atom. */
 305             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 306             velecsum         = _mm_add_pd(velecsum,velec);
 307             vgb              = _mm_unpacklo_pd(vgb,_mm_setzero_pd());
 308             vgbsum           = _mm_add_pd(vgbsum,vgb);
 309
 310             fscal            = felec;
 311
 312             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 313
 314             /* Update vectorial force */
 315             fix0             = _mm_macc_pd(dx00,fscal,fix0);
 316             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
 317             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
 318
 319             gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,
 320                                                    _mm_mul_pd(dx00,fscal),
 321                                                    _mm_mul_pd(dy00,fscal),
 322                                                    _mm_mul_pd(dz00,fscal));
 323
 324             /* Inner loop uses 61 flops */
 325         }
 326
 327         /* End of innermost loop */
 328
 329         gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
 330                                               f+i_coord_offset,fshift+i_shift_offset);
 331
 332         ggid                        = gid[iidx];
 333         /* Update potential energies */
 334         gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
 335         gmx_mm_update_1pot_pd(vgbsum,kernel_data->energygrp_polarization+ggid);
 336         dvdasum = _mm_mul_pd(dvdasum, _mm_mul_pd(isai0,isai0));
 337         gmx_mm_update_1pot_pd(dvdasum,dvda+inr);
 338
 339         /* Increment number of inner iterations */
 340         inneriter                  += j_index_end - j_index_start;
 341
 342         /* Outer loop uses 9 flops */
 343     }
 344
 345     /* Increment number of outer iterations */
 346     outeriter        += nri;
 347
 348     /* Update outer/inner flops */
 349
 350     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VF,outeriter*9 + inneriter*61);
 351 }
 352 /*
 353  * Gromacs nonbonded kernel:   nb_kernel_ElecGB_VdwNone_GeomP1P1_F_avx_128_fma_double
 354  * Electrostatics interaction: GeneralizedBorn
 355  * VdW interaction:            None
 356  * Geometry:                   Particle-Particle
 357  * Calculate force/pot:        Force
 358  */
 359 void
 360 nb_kernel_ElecGB_VdwNone_GeomP1P1_F_avx_128_fma_double
 361                     (t_nblist * gmx_restrict                nlist,
 362                      rvec * gmx_restrict                    xx,
 363                      rvec * gmx_restrict                    ff,
 364                      t_forcerec * gmx_restrict              fr,
 365                      t_mdatoms * gmx_restrict               mdatoms,
 366                      nb_kernel_data_t * gmx_restrict        kernel_data,
 367                      t_nrnb * gmx_restrict                  nrnb)
 368 {
 369     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 370      * just 0 for non-waters.
 371      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
 372      * jnr indices corresponding to data put in the four positions in the SIMD register.
 373      */
 374     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 375     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 376     int              jnrA,jnrB;
 377     int              j_coord_offsetA,j_coord_offsetB;
 378     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 379     real             rcutoff_scalar;
 380     real             *shiftvec,*fshift,*x,*f;
 381     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 382     int              vdwioffset0;
 383     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 384     int              vdwjidx0A,vdwjidx0B;
 385     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 386     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 387     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
 388     real             *charge;
 389     __m128i          gbitab;
 390     __m128d          vgb,fgb,vgbsum,dvdasum,gbscale,gbtabscale,isaprod,gbqqfactor,gbinvepsdiff,dvdaj,gbeps,twogbeps,dvdatmp;
 391     __m128d          minushalf = _mm_set1_pd(-0.5);
 392     real             *invsqrta,*dvda,*gbtab;
 393     __m128i          vfitab;
 394     __m128i          ifour       = _mm_set1_epi32(4);
 395     __m128d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 396     real             *vftab;
 397     __m128d          dummy_mask,cutoff_mask;
 398     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 399     __m128d          one     = _mm_set1_pd(1.0);
 400     __m128d          two     = _mm_set1_pd(2.0);
 401     x                = xx[0];
 402     f                = ff[0];
 403
 404     nri              = nlist->nri;
 405     iinr             = nlist->iinr;
 406     jindex           = nlist->jindex;
 407     jjnr             = nlist->jjnr;
 408     shiftidx         = nlist->shift;
 409     gid              = nlist->gid;
 410     shiftvec         = fr->shift_vec[0];
 411     fshift           = fr->fshift[0];
 412     facel            = _mm_set1_pd(fr->epsfac);
 413     charge           = mdatoms->chargeA;
 414
 415     invsqrta         = fr->invsqrta;
 416     dvda             = fr->dvda;
 417     gbtabscale       = _mm_set1_pd(fr->gbtab.scale);
 418     gbtab            = fr->gbtab.data;
 419     gbinvepsdiff     = _mm_set1_pd((1.0/fr->epsilon_r) - (1.0/fr->gb_epsilon_solvent));
 420
 421     /* Avoid stupid compiler warnings */
 422     jnrA = jnrB = 0;
 423     j_coord_offsetA = 0;
 424     j_coord_offsetB = 0;
 425
 426     outeriter        = 0;
 427     inneriter        = 0;
 428
 429     /* Start outer loop over neighborlists */
 430     for(iidx=0; iidx<nri; iidx++)
 431     {
 432         /* Load shift vector for this list */
 433         i_shift_offset   = DIM*shiftidx[iidx];
 434
 435         /* Load limits for loop over neighbors */
 436         j_index_start    = jindex[iidx];
 437         j_index_end      = jindex[iidx+1];
 438
 439         /* Get outer coordinate index */
 440         inr              = iinr[iidx];
 441         i_coord_offset   = DIM*inr;
 442
 443         /* Load i particle coords and add shift vector */
 444         gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 445
 446         fix0             = _mm_setzero_pd();
 447         fiy0             = _mm_setzero_pd();
 448         fiz0             = _mm_setzero_pd();
 449
 450         /* Load parameters for i particles */
 451         iq0              = _mm_mul_pd(facel,_mm_load1_pd(charge+inr+0));
 452         isai0            = _mm_load1_pd(invsqrta+inr+0);
 453
 454         dvdasum          = _mm_setzero_pd();
 455
 456         /* Start inner kernel loop */
 457         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 458         {
 459
 460             /* Get j neighbor index, and coordinate index */
 461             jnrA             = jjnr[jidx];
 462             jnrB             = jjnr[jidx+1];
 463             j_coord_offsetA  = DIM*jnrA;
 464             j_coord_offsetB  = DIM*jnrB;
 465
 466             /* load j atom coordinates */
 467             gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 468                                               &jx0,&jy0,&jz0);
 469
 470             /* Calculate displacement vector */
 471             dx00             = _mm_sub_pd(ix0,jx0);
 472             dy00             = _mm_sub_pd(iy0,jy0);
 473             dz00             = _mm_sub_pd(iz0,jz0);
 474
 475             /* Calculate squared distance and things based on it */
 476             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 477
 478             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 479
 480             /* Load parameters for j particles */
 481             jq0              = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
 482             isaj0            = gmx_mm_load_2real_swizzle_pd(invsqrta+jnrA+0,invsqrta+jnrB+0);
 483
 484             /**************************
 485              * CALCULATE INTERACTIONS *
 486              **************************/
 487
 488             r00              = _mm_mul_pd(rsq00,rinv00);
 489
 490             /* Compute parameters for interactions between i and j atoms */
 491             qq00             = _mm_mul_pd(iq0,jq0);
 492
 493             /* GENERALIZED BORN AND COULOMB ELECTROSTATICS */
 494             isaprod          = _mm_mul_pd(isai0,isaj0);
 495             gbqqfactor       = _mm_xor_pd(signbit,_mm_mul_pd(qq00,_mm_mul_pd(isaprod,gbinvepsdiff)));
 496             gbscale          = _mm_mul_pd(isaprod,gbtabscale);
 497
 498             /* Calculate generalized born table index - this is a separate table from the normal one,
 499              * but we use the same procedure by multiplying r with scale and truncating to integer.
 500              */
 501             rt               = _mm_mul_pd(r00,gbscale);
 502             gbitab           = _mm_cvttpd_epi32(rt);
 503 #ifdef __XOP__
 504             gbeps            = _mm_frcz_pd(rt);
 505 #else
 506             gbeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 507 #endif
 508             gbitab           = _mm_slli_epi32(gbitab,2);
 509
 510             Y                = _mm_load_pd( gbtab + _mm_extract_epi32(gbitab,0) );
 511             F                = _mm_load_pd( gbtab + _mm_extract_epi32(gbitab,1) );
 512             GMX_MM_TRANSPOSE2_PD(Y,F);
 513             G                = _mm_load_pd( gbtab + _mm_extract_epi32(gbitab,0) +2);
 514             H                = _mm_load_pd( gbtab + _mm_extract_epi32(gbitab,1) +2);
 515             GMX_MM_TRANSPOSE2_PD(G,H);
 516             Fp               = _mm_macc_pd(gbeps,_mm_macc_pd(gbeps,H,G),F);
 517             VV               = _mm_macc_pd(gbeps,Fp,Y);
 518             vgb              = _mm_mul_pd(gbqqfactor,VV);
 519
 520             twogbeps         = _mm_add_pd(gbeps,gbeps);
 521             FF               = _mm_macc_pd(_mm_macc_pd(twogbeps,H,G),gbeps,Fp);
 522             fgb              = _mm_mul_pd(gbqqfactor,_mm_mul_pd(FF,gbscale));
 523             dvdatmp          = _mm_mul_pd(minushalf,_mm_macc_pd(fgb,r00,vgb));
 524             dvdasum          = _mm_add_pd(dvdasum,dvdatmp);
 525             gmx_mm_increment_2real_swizzle_pd(dvda+jnrA,dvda+jnrB,_mm_mul_pd(dvdatmp,_mm_mul_pd(isaj0,isaj0)));
 526             velec            = _mm_mul_pd(qq00,rinv00);
 527             felec            = _mm_mul_pd(_mm_msub_pd(velec,rinv00,fgb),rinv00);
 528
 529             fscal            = felec;
 530
 531             /* Update vectorial force */
 532             fix0             = _mm_macc_pd(dx00,fscal,fix0);
 533             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
 534             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
 535
 536             gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,
 537                                                    _mm_mul_pd(dx00,fscal),
 538                                                    _mm_mul_pd(dy00,fscal),
 539                                                    _mm_mul_pd(dz00,fscal));
 540
 541             /* Inner loop uses 59 flops */
 542         }
 543
 544         if(jidx<j_index_end)
 545         {
 546
 547             jnrA             = jjnr[jidx];
 548             j_coord_offsetA  = DIM*jnrA;
 549
 550             /* load j atom coordinates */
 551             gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 552                                               &jx0,&jy0,&jz0);
 553
 554             /* Calculate displacement vector */
 555             dx00             = _mm_sub_pd(ix0,jx0);
 556             dy00             = _mm_sub_pd(iy0,jy0);
 557             dz00             = _mm_sub_pd(iz0,jz0);
 558
 559             /* Calculate squared distance and things based on it */
 560             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 561
 562             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 563
 564             /* Load parameters for j particles */
 565             jq0              = _mm_load_sd(charge+jnrA+0);
 566             isaj0            = _mm_load_sd(invsqrta+jnrA+0);
 567
 568             /**************************
 569              * CALCULATE INTERACTIONS *
 570              **************************/
 571
 572             r00              = _mm_mul_pd(rsq00,rinv00);
 573
 574             /* Compute parameters for interactions between i and j atoms */
 575             qq00             = _mm_mul_pd(iq0,jq0);
 576
 577             /* GENERALIZED BORN AND COULOMB ELECTROSTATICS */
 578             isaprod          = _mm_mul_pd(isai0,isaj0);
 579             gbqqfactor       = _mm_xor_pd(signbit,_mm_mul_pd(qq00,_mm_mul_pd(isaprod,gbinvepsdiff)));
 580             gbscale          = _mm_mul_pd(isaprod,gbtabscale);
 581
 582             /* Calculate generalized born table index - this is a separate table from the normal one,
 583              * but we use the same procedure by multiplying r with scale and truncating to integer.
 584              */
 585             rt               = _mm_mul_pd(r00,gbscale);
 586             gbitab           = _mm_cvttpd_epi32(rt);
 587 #ifdef __XOP__
 588             gbeps            = _mm_frcz_pd(rt);
 589 #else
 590             gbeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 591 #endif
 592             gbitab           = _mm_slli_epi32(gbitab,2);
 593
 594             Y                = _mm_load_pd( gbtab + _mm_extract_epi32(gbitab,0) );
 595             F                = _mm_setzero_pd();
 596             GMX_MM_TRANSPOSE2_PD(Y,F);
 597             G                = _mm_load_pd( gbtab + _mm_extract_epi32(gbitab,0) +2);
 598             H                = _mm_setzero_pd();
 599             GMX_MM_TRANSPOSE2_PD(G,H);
 600             Fp               = _mm_macc_pd(gbeps,_mm_macc_pd(gbeps,H,G),F);
 601             VV               = _mm_macc_pd(gbeps,Fp,Y);
 602             vgb              = _mm_mul_pd(gbqqfactor,VV);
 603
 604             twogbeps         = _mm_add_pd(gbeps,gbeps);
 605             FF               = _mm_macc_pd(_mm_macc_pd(twogbeps,H,G),gbeps,Fp);
 606             fgb              = _mm_mul_pd(gbqqfactor,_mm_mul_pd(FF,gbscale));
 607             dvdatmp          = _mm_mul_pd(minushalf,_mm_macc_pd(fgb,r00,vgb));
 608             dvdasum          = _mm_add_pd(dvdasum,dvdatmp);
 609             gmx_mm_increment_1real_pd(dvda+jnrA,_mm_mul_pd(dvdatmp,_mm_mul_pd(isaj0,isaj0)));
 610             velec            = _mm_mul_pd(qq00,rinv00);
 611             felec            = _mm_mul_pd(_mm_msub_pd(velec,rinv00,fgb),rinv00);
 612
 613             fscal            = felec;
 614
 615             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 616
 617             /* Update vectorial force */
 618             fix0             = _mm_macc_pd(dx00,fscal,fix0);
 619             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
 620             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
 621
 622             gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,
 623                                                    _mm_mul_pd(dx00,fscal),
 624                                                    _mm_mul_pd(dy00,fscal),
 625                                                    _mm_mul_pd(dz00,fscal));
 626
 627             /* Inner loop uses 59 flops */
 628         }
 629
 630         /* End of innermost loop */
 631
 632         gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
 633                                               f+i_coord_offset,fshift+i_shift_offset);
 634
 635         dvdasum = _mm_mul_pd(dvdasum, _mm_mul_pd(isai0,isai0));
 636         gmx_mm_update_1pot_pd(dvdasum,dvda+inr);
 637
 638         /* Increment number of inner iterations */
 639         inneriter                  += j_index_end - j_index_start;
 640
 641         /* Outer loop uses 7 flops */
 642     }
 643
 644     /* Increment number of outer iterations */
 645     outeriter        += nri;
 646
 647     /* Update outer/inner flops */
 648
 649     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_F,outeriter*7 + inneriter*59);
 650 }