2 * This file is part of the GROMACS molecular simulation package.
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.
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.
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.
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.
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.
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.
36 * Note: this file was generated by the GROMACS sparc64_hpc_ace_double kernel generator.
44 #include "../nb_kernel.h"
45 #include "types/simple.h"
46 #include "gromacs/legacyheaders/vec.h"
49 #include "kernelutil_sparc64_hpc_ace_double.h"
52 * Gromacs nonbonded kernel: nb_kernel_ElecGB_VdwLJ_GeomP1P1_VF_sparc64_hpc_ace_double
53 * Electrostatics interaction: GeneralizedBorn
54 * VdW interaction: LennardJones
55 * Geometry: Particle-Particle
56 * Calculate force/pot: PotentialAndForce
59 nb_kernel_ElecGB_VdwLJ_GeomP1P1_VF_sparc64_hpc_ace_double
60 (t_nblist * gmx_restrict nlist,
61 rvec * gmx_restrict xx,
62 rvec * gmx_restrict ff,
63 t_forcerec * gmx_restrict fr,
64 t_mdatoms * gmx_restrict mdatoms,
65 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
66 t_nrnb * gmx_restrict nrnb)
68 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
69 * just 0 for non-waters.
70 * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
71 * jnr indices corresponding to data put in the four positions in the SIMD register.
73 int i_shift_offset,i_coord_offset,outeriter,inneriter;
74 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
76 int j_coord_offsetA,j_coord_offsetB;
77 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
79 real *shiftvec,*fshift,*x,*f;
80 _fjsp_v2r8 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
82 _fjsp_v2r8 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
83 int vdwjidx0A,vdwjidx0B;
84 _fjsp_v2r8 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
85 _fjsp_v2r8 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
86 _fjsp_v2r8 velec,felec,velecsum,facel,crf,krf,krf2;
88 _fjsp_v2r8 vgb,fgb,vgbsum,dvdasum,gbscale,gbtabscale,isaprod,gbqqfactor,gbinvepsdiff,dvdaj,gbeps,twogbeps,dvdatmp;
89 _fjsp_v2r8 minushalf = gmx_fjsp_set1_v2r8(-0.5);
90 real *invsqrta,*dvda,*gbtab;
92 _fjsp_v2r8 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
95 _fjsp_v2r8 one_sixth = gmx_fjsp_set1_v2r8(1.0/6.0);
96 _fjsp_v2r8 one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
97 _fjsp_v2r8 rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
100 _fjsp_v2r8 dummy_mask,cutoff_mask;
101 _fjsp_v2r8 one = gmx_fjsp_set1_v2r8(1.0);
102 _fjsp_v2r8 two = gmx_fjsp_set1_v2r8(2.0);
103 union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
110 jindex = nlist->jindex;
112 shiftidx = nlist->shift;
114 shiftvec = fr->shift_vec[0];
115 fshift = fr->fshift[0];
116 facel = gmx_fjsp_set1_v2r8(fr->epsfac);
117 charge = mdatoms->chargeA;
118 nvdwtype = fr->ntype;
120 vdwtype = mdatoms->typeA;
122 invsqrta = fr->invsqrta;
124 gbtabscale = gmx_fjsp_set1_v2r8(fr->gbtab.scale);
125 gbtab = fr->gbtab.data;
126 gbinvepsdiff = gmx_fjsp_set1_v2r8((1.0/fr->epsilon_r) - (1.0/fr->gb_epsilon_solvent));
128 /* Avoid stupid compiler warnings */
136 /* Start outer loop over neighborlists */
137 for(iidx=0; iidx<nri; iidx++)
139 /* Load shift vector for this list */
140 i_shift_offset = DIM*shiftidx[iidx];
142 /* Load limits for loop over neighbors */
143 j_index_start = jindex[iidx];
144 j_index_end = jindex[iidx+1];
146 /* Get outer coordinate index */
148 i_coord_offset = DIM*inr;
150 /* Load i particle coords and add shift vector */
151 gmx_fjsp_load_shift_and_1rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
153 fix0 = _fjsp_setzero_v2r8();
154 fiy0 = _fjsp_setzero_v2r8();
155 fiz0 = _fjsp_setzero_v2r8();
157 /* Load parameters for i particles */
158 iq0 = _fjsp_mul_v2r8(facel,gmx_fjsp_load1_v2r8(charge+inr+0));
159 isai0 = gmx_fjsp_load1_v2r8(invsqrta+inr+0);
160 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
162 /* Reset potential sums */
163 velecsum = _fjsp_setzero_v2r8();
164 vgbsum = _fjsp_setzero_v2r8();
165 vvdwsum = _fjsp_setzero_v2r8();
166 dvdasum = _fjsp_setzero_v2r8();
168 /* Start inner kernel loop */
169 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
172 /* Get j neighbor index, and coordinate index */
175 j_coord_offsetA = DIM*jnrA;
176 j_coord_offsetB = DIM*jnrB;
178 /* load j atom coordinates */
179 gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
182 /* Calculate displacement vector */
183 dx00 = _fjsp_sub_v2r8(ix0,jx0);
184 dy00 = _fjsp_sub_v2r8(iy0,jy0);
185 dz00 = _fjsp_sub_v2r8(iz0,jz0);
187 /* Calculate squared distance and things based on it */
188 rsq00 = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
190 rinv00 = gmx_fjsp_invsqrt_v2r8(rsq00);
192 rinvsq00 = _fjsp_mul_v2r8(rinv00,rinv00);
194 /* Load parameters for j particles */
195 jq0 = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
196 isaj0 = gmx_fjsp_load_2real_swizzle_v2r8(invsqrta+jnrA+0,invsqrta+jnrB+0);
197 vdwjidx0A = 2*vdwtype[jnrA+0];
198 vdwjidx0B = 2*vdwtype[jnrB+0];
200 /**************************
201 * CALCULATE INTERACTIONS *
202 **************************/
204 r00 = _fjsp_mul_v2r8(rsq00,rinv00);
206 /* Compute parameters for interactions between i and j atoms */
207 qq00 = _fjsp_mul_v2r8(iq0,jq0);
208 gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
209 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
211 /* GENERALIZED BORN AND COULOMB ELECTROSTATICS */
212 isaprod = _fjsp_mul_v2r8(isai0,isaj0);
213 gbqqfactor = _fjsp_neg_v2r8(_fjsp_mul_v2r8(qq00,_fjsp_mul_v2r8(isaprod,gbinvepsdiff)));
214 gbscale = _fjsp_mul_v2r8(isaprod,gbtabscale);
216 /* Calculate generalized born table index - this is a separate table from the normal one,
217 * but we use the same procedure by multiplying r with scale and truncating to integer.
219 rt = _fjsp_mul_v2r8(r00,gbscale);
220 itab_tmp = _fjsp_dtox_v2r8(rt);
221 gbeps = _fjsp_sub_v2r8(rt,_fjsp_xtod_v2r8(itab_tmp));
222 _fjsp_store_v2r8(&gbconv.simd,itab_tmp);
224 Y = _fjsp_load_v2r8( gbtab + 4*gbconv.i[0] );
225 F = _fjsp_load_v2r8( gbtab + 4*gbconv.i[1] );
226 GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
227 G = _fjsp_load_v2r8( gbtab + 4*gbconv.i[0] +2);
228 H = _fjsp_load_v2r8( gbtab + 4*gbconv.i[1] +2);
229 GMX_FJSP_TRANSPOSE2_V2R8(G,H);
230 Fp = _fjsp_madd_v2r8(gbeps,_fjsp_madd_v2r8(gbeps,H,G),F);
231 VV = _fjsp_madd_v2r8(gbeps,Fp,Y);
232 vgb = _fjsp_mul_v2r8(gbqqfactor,VV);
234 twogbeps = _fjsp_add_v2r8(gbeps,gbeps);
235 FF = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twogbeps,H,G),gbeps,Fp);
236 fgb = _fjsp_mul_v2r8(gbqqfactor,_fjsp_mul_v2r8(FF,gbscale));
237 dvdatmp = _fjsp_mul_v2r8(minushalf,_fjsp_madd_v2r8(fgb,r00,vgb));
238 dvdasum = _fjsp_add_v2r8(dvdasum,dvdatmp);
239 gmx_fjsp_increment_2real_swizzle_v2r8(dvda+jnrA,dvda+jnrB,_fjsp_mul_v2r8(dvdatmp,_fjsp_mul_v2r8(isaj0,isaj0)));
240 velec = _fjsp_mul_v2r8(qq00,rinv00);
241 felec = _fjsp_mul_v2r8(_fjsp_msub_v2r8(velec,rinv00,fgb),rinv00);
243 /* LENNARD-JONES DISPERSION/REPULSION */
245 rinvsix = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
246 vvdw6 = _fjsp_mul_v2r8(c6_00,rinvsix);
247 vvdw12 = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
248 vvdw = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
249 fvdw = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
251 /* Update potential sum for this i atom from the interaction with this j atom. */
252 velecsum = _fjsp_add_v2r8(velecsum,velec);
253 vgbsum = _fjsp_add_v2r8(vgbsum,vgb);
254 vvdwsum = _fjsp_add_v2r8(vvdwsum,vvdw);
256 fscal = _fjsp_add_v2r8(felec,fvdw);
258 /* Update vectorial force */
259 fix0 = _fjsp_madd_v2r8(dx00,fscal,fix0);
260 fiy0 = _fjsp_madd_v2r8(dy00,fscal,fiy0);
261 fiz0 = _fjsp_madd_v2r8(dz00,fscal,fiz0);
263 gmx_fjsp_decrement_fma_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fscal,dx00,dy00,dz00);
265 /* Inner loop uses 74 flops */
272 j_coord_offsetA = DIM*jnrA;
274 /* load j atom coordinates */
275 gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
278 /* Calculate displacement vector */
279 dx00 = _fjsp_sub_v2r8(ix0,jx0);
280 dy00 = _fjsp_sub_v2r8(iy0,jy0);
281 dz00 = _fjsp_sub_v2r8(iz0,jz0);
283 /* Calculate squared distance and things based on it */
284 rsq00 = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
286 rinv00 = gmx_fjsp_invsqrt_v2r8(rsq00);
288 rinvsq00 = _fjsp_mul_v2r8(rinv00,rinv00);
290 /* Load parameters for j particles */
291 jq0 = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
292 isaj0 = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),invsqrta+jnrA+0);
293 vdwjidx0A = 2*vdwtype[jnrA+0];
295 /**************************
296 * CALCULATE INTERACTIONS *
297 **************************/
299 r00 = _fjsp_mul_v2r8(rsq00,rinv00);
301 /* Compute parameters for interactions between i and j atoms */
302 qq00 = _fjsp_mul_v2r8(iq0,jq0);
303 gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
304 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
306 /* GENERALIZED BORN AND COULOMB ELECTROSTATICS */
307 isaprod = _fjsp_mul_v2r8(isai0,isaj0);
308 gbqqfactor = _fjsp_neg_v2r8(_fjsp_mul_v2r8(qq00,_fjsp_mul_v2r8(isaprod,gbinvepsdiff)));
309 gbscale = _fjsp_mul_v2r8(isaprod,gbtabscale);
311 /* Calculate generalized born table index - this is a separate table from the normal one,
312 * but we use the same procedure by multiplying r with scale and truncating to integer.
314 rt = _fjsp_mul_v2r8(r00,gbscale);
315 itab_tmp = _fjsp_dtox_v2r8(rt);
316 gbeps = _fjsp_sub_v2r8(rt,_fjsp_xtod_v2r8(itab_tmp));
317 _fjsp_store_v2r8(&gbconv.simd,itab_tmp);
319 Y = _fjsp_load_v2r8( gbtab + 4*gbconv.i[0] );
320 F = _fjsp_setzero_v2r8();
321 GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
322 G = _fjsp_load_v2r8( gbtab + 4*gbconv.i[0] +2);
323 H = _fjsp_setzero_v2r8();
324 GMX_FJSP_TRANSPOSE2_V2R8(G,H);
325 Fp = _fjsp_madd_v2r8(gbeps,_fjsp_madd_v2r8(gbeps,H,G),F);
326 VV = _fjsp_madd_v2r8(gbeps,Fp,Y);
327 vgb = _fjsp_mul_v2r8(gbqqfactor,VV);
329 twogbeps = _fjsp_add_v2r8(gbeps,gbeps);
330 FF = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twogbeps,H,G),gbeps,Fp);
331 fgb = _fjsp_mul_v2r8(gbqqfactor,_fjsp_mul_v2r8(FF,gbscale));
332 dvdatmp = _fjsp_mul_v2r8(minushalf,_fjsp_madd_v2r8(fgb,r00,vgb));
333 dvdasum = _fjsp_add_v2r8(dvdasum,dvdatmp);
334 gmx_fjsp_increment_1real_v2r8(dvda+jnrA,_fjsp_mul_v2r8(dvdatmp,_fjsp_mul_v2r8(isaj0,isaj0)));
335 velec = _fjsp_mul_v2r8(qq00,rinv00);
336 felec = _fjsp_mul_v2r8(_fjsp_msub_v2r8(velec,rinv00,fgb),rinv00);
338 /* LENNARD-JONES DISPERSION/REPULSION */
340 rinvsix = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
341 vvdw6 = _fjsp_mul_v2r8(c6_00,rinvsix);
342 vvdw12 = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
343 vvdw = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
344 fvdw = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
346 /* Update potential sum for this i atom from the interaction with this j atom. */
347 velec = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
348 velecsum = _fjsp_add_v2r8(velecsum,velec);
349 vgb = _fjsp_unpacklo_v2r8(vgb,_fjsp_setzero_v2r8());
350 vgbsum = _fjsp_add_v2r8(vgbsum,vgb);
351 vvdw = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
352 vvdwsum = _fjsp_add_v2r8(vvdwsum,vvdw);
354 fscal = _fjsp_add_v2r8(felec,fvdw);
356 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
358 /* Update vectorial force */
359 fix0 = _fjsp_madd_v2r8(dx00,fscal,fix0);
360 fiy0 = _fjsp_madd_v2r8(dy00,fscal,fiy0);
361 fiz0 = _fjsp_madd_v2r8(dz00,fscal,fiz0);
363 gmx_fjsp_decrement_fma_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fscal,dx00,dy00,dz00);
365 /* Inner loop uses 74 flops */
368 /* End of innermost loop */
370 gmx_fjsp_update_iforce_1atom_swizzle_v2r8(fix0,fiy0,fiz0,
371 f+i_coord_offset,fshift+i_shift_offset);
374 /* Update potential energies */
375 gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
376 gmx_fjsp_update_1pot_v2r8(vgbsum,kernel_data->energygrp_polarization+ggid);
377 gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
378 dvdasum = _fjsp_mul_v2r8(dvdasum, _fjsp_mul_v2r8(isai0,isai0));
379 gmx_fjsp_update_1pot_v2r8(dvdasum,dvda+inr);
381 /* Increment number of inner iterations */
382 inneriter += j_index_end - j_index_start;
384 /* Outer loop uses 10 flops */
387 /* Increment number of outer iterations */
390 /* Update outer/inner flops */
392 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_VF,outeriter*10 + inneriter*74);
395 * Gromacs nonbonded kernel: nb_kernel_ElecGB_VdwLJ_GeomP1P1_F_sparc64_hpc_ace_double
396 * Electrostatics interaction: GeneralizedBorn
397 * VdW interaction: LennardJones
398 * Geometry: Particle-Particle
399 * Calculate force/pot: Force
402 nb_kernel_ElecGB_VdwLJ_GeomP1P1_F_sparc64_hpc_ace_double
403 (t_nblist * gmx_restrict nlist,
404 rvec * gmx_restrict xx,
405 rvec * gmx_restrict ff,
406 t_forcerec * gmx_restrict fr,
407 t_mdatoms * gmx_restrict mdatoms,
408 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
409 t_nrnb * gmx_restrict nrnb)
411 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
412 * just 0 for non-waters.
413 * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
414 * jnr indices corresponding to data put in the four positions in the SIMD register.
416 int i_shift_offset,i_coord_offset,outeriter,inneriter;
417 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
419 int j_coord_offsetA,j_coord_offsetB;
420 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
422 real *shiftvec,*fshift,*x,*f;
423 _fjsp_v2r8 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
425 _fjsp_v2r8 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
426 int vdwjidx0A,vdwjidx0B;
427 _fjsp_v2r8 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
428 _fjsp_v2r8 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
429 _fjsp_v2r8 velec,felec,velecsum,facel,crf,krf,krf2;
431 _fjsp_v2r8 vgb,fgb,vgbsum,dvdasum,gbscale,gbtabscale,isaprod,gbqqfactor,gbinvepsdiff,dvdaj,gbeps,twogbeps,dvdatmp;
432 _fjsp_v2r8 minushalf = gmx_fjsp_set1_v2r8(-0.5);
433 real *invsqrta,*dvda,*gbtab;
435 _fjsp_v2r8 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
438 _fjsp_v2r8 one_sixth = gmx_fjsp_set1_v2r8(1.0/6.0);
439 _fjsp_v2r8 one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
440 _fjsp_v2r8 rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
443 _fjsp_v2r8 dummy_mask,cutoff_mask;
444 _fjsp_v2r8 one = gmx_fjsp_set1_v2r8(1.0);
445 _fjsp_v2r8 two = gmx_fjsp_set1_v2r8(2.0);
446 union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
453 jindex = nlist->jindex;
455 shiftidx = nlist->shift;
457 shiftvec = fr->shift_vec[0];
458 fshift = fr->fshift[0];
459 facel = gmx_fjsp_set1_v2r8(fr->epsfac);
460 charge = mdatoms->chargeA;
461 nvdwtype = fr->ntype;
463 vdwtype = mdatoms->typeA;
465 invsqrta = fr->invsqrta;
467 gbtabscale = gmx_fjsp_set1_v2r8(fr->gbtab.scale);
468 gbtab = fr->gbtab.data;
469 gbinvepsdiff = gmx_fjsp_set1_v2r8((1.0/fr->epsilon_r) - (1.0/fr->gb_epsilon_solvent));
471 /* Avoid stupid compiler warnings */
479 /* Start outer loop over neighborlists */
480 for(iidx=0; iidx<nri; iidx++)
482 /* Load shift vector for this list */
483 i_shift_offset = DIM*shiftidx[iidx];
485 /* Load limits for loop over neighbors */
486 j_index_start = jindex[iidx];
487 j_index_end = jindex[iidx+1];
489 /* Get outer coordinate index */
491 i_coord_offset = DIM*inr;
493 /* Load i particle coords and add shift vector */
494 gmx_fjsp_load_shift_and_1rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
496 fix0 = _fjsp_setzero_v2r8();
497 fiy0 = _fjsp_setzero_v2r8();
498 fiz0 = _fjsp_setzero_v2r8();
500 /* Load parameters for i particles */
501 iq0 = _fjsp_mul_v2r8(facel,gmx_fjsp_load1_v2r8(charge+inr+0));
502 isai0 = gmx_fjsp_load1_v2r8(invsqrta+inr+0);
503 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
505 dvdasum = _fjsp_setzero_v2r8();
507 /* Start inner kernel loop */
508 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
511 /* Get j neighbor index, and coordinate index */
514 j_coord_offsetA = DIM*jnrA;
515 j_coord_offsetB = DIM*jnrB;
517 /* load j atom coordinates */
518 gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
521 /* Calculate displacement vector */
522 dx00 = _fjsp_sub_v2r8(ix0,jx0);
523 dy00 = _fjsp_sub_v2r8(iy0,jy0);
524 dz00 = _fjsp_sub_v2r8(iz0,jz0);
526 /* Calculate squared distance and things based on it */
527 rsq00 = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
529 rinv00 = gmx_fjsp_invsqrt_v2r8(rsq00);
531 rinvsq00 = _fjsp_mul_v2r8(rinv00,rinv00);
533 /* Load parameters for j particles */
534 jq0 = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
535 isaj0 = gmx_fjsp_load_2real_swizzle_v2r8(invsqrta+jnrA+0,invsqrta+jnrB+0);
536 vdwjidx0A = 2*vdwtype[jnrA+0];
537 vdwjidx0B = 2*vdwtype[jnrB+0];
539 /**************************
540 * CALCULATE INTERACTIONS *
541 **************************/
543 r00 = _fjsp_mul_v2r8(rsq00,rinv00);
545 /* Compute parameters for interactions between i and j atoms */
546 qq00 = _fjsp_mul_v2r8(iq0,jq0);
547 gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
548 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
550 /* GENERALIZED BORN AND COULOMB ELECTROSTATICS */
551 isaprod = _fjsp_mul_v2r8(isai0,isaj0);
552 gbqqfactor = _fjsp_neg_v2r8(_fjsp_mul_v2r8(qq00,_fjsp_mul_v2r8(isaprod,gbinvepsdiff)));
553 gbscale = _fjsp_mul_v2r8(isaprod,gbtabscale);
555 /* Calculate generalized born table index - this is a separate table from the normal one,
556 * but we use the same procedure by multiplying r with scale and truncating to integer.
558 rt = _fjsp_mul_v2r8(r00,gbscale);
559 itab_tmp = _fjsp_dtox_v2r8(rt);
560 gbeps = _fjsp_sub_v2r8(rt,_fjsp_xtod_v2r8(itab_tmp));
561 _fjsp_store_v2r8(&gbconv.simd,itab_tmp);
563 Y = _fjsp_load_v2r8( gbtab + 4*gbconv.i[0] );
564 F = _fjsp_load_v2r8( gbtab + 4*gbconv.i[1] );
565 GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
566 G = _fjsp_load_v2r8( gbtab + 4*gbconv.i[0] +2);
567 H = _fjsp_load_v2r8( gbtab + 4*gbconv.i[1] +2);
568 GMX_FJSP_TRANSPOSE2_V2R8(G,H);
569 Fp = _fjsp_madd_v2r8(gbeps,_fjsp_madd_v2r8(gbeps,H,G),F);
570 VV = _fjsp_madd_v2r8(gbeps,Fp,Y);
571 vgb = _fjsp_mul_v2r8(gbqqfactor,VV);
573 twogbeps = _fjsp_add_v2r8(gbeps,gbeps);
574 FF = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twogbeps,H,G),gbeps,Fp);
575 fgb = _fjsp_mul_v2r8(gbqqfactor,_fjsp_mul_v2r8(FF,gbscale));
576 dvdatmp = _fjsp_mul_v2r8(minushalf,_fjsp_madd_v2r8(fgb,r00,vgb));
577 dvdasum = _fjsp_add_v2r8(dvdasum,dvdatmp);
578 gmx_fjsp_increment_2real_swizzle_v2r8(dvda+jnrA,dvda+jnrB,_fjsp_mul_v2r8(dvdatmp,_fjsp_mul_v2r8(isaj0,isaj0)));
579 velec = _fjsp_mul_v2r8(qq00,rinv00);
580 felec = _fjsp_mul_v2r8(_fjsp_msub_v2r8(velec,rinv00,fgb),rinv00);
582 /* LENNARD-JONES DISPERSION/REPULSION */
584 rinvsix = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
585 fvdw = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
587 fscal = _fjsp_add_v2r8(felec,fvdw);
589 /* Update vectorial force */
590 fix0 = _fjsp_madd_v2r8(dx00,fscal,fix0);
591 fiy0 = _fjsp_madd_v2r8(dy00,fscal,fiy0);
592 fiz0 = _fjsp_madd_v2r8(dz00,fscal,fiz0);
594 gmx_fjsp_decrement_fma_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fscal,dx00,dy00,dz00);
596 /* Inner loop uses 67 flops */
603 j_coord_offsetA = DIM*jnrA;
605 /* load j atom coordinates */
606 gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
609 /* Calculate displacement vector */
610 dx00 = _fjsp_sub_v2r8(ix0,jx0);
611 dy00 = _fjsp_sub_v2r8(iy0,jy0);
612 dz00 = _fjsp_sub_v2r8(iz0,jz0);
614 /* Calculate squared distance and things based on it */
615 rsq00 = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
617 rinv00 = gmx_fjsp_invsqrt_v2r8(rsq00);
619 rinvsq00 = _fjsp_mul_v2r8(rinv00,rinv00);
621 /* Load parameters for j particles */
622 jq0 = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
623 isaj0 = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),invsqrta+jnrA+0);
624 vdwjidx0A = 2*vdwtype[jnrA+0];
626 /**************************
627 * CALCULATE INTERACTIONS *
628 **************************/
630 r00 = _fjsp_mul_v2r8(rsq00,rinv00);
632 /* Compute parameters for interactions between i and j atoms */
633 qq00 = _fjsp_mul_v2r8(iq0,jq0);
634 gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
635 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
637 /* GENERALIZED BORN AND COULOMB ELECTROSTATICS */
638 isaprod = _fjsp_mul_v2r8(isai0,isaj0);
639 gbqqfactor = _fjsp_neg_v2r8(_fjsp_mul_v2r8(qq00,_fjsp_mul_v2r8(isaprod,gbinvepsdiff)));
640 gbscale = _fjsp_mul_v2r8(isaprod,gbtabscale);
642 /* Calculate generalized born table index - this is a separate table from the normal one,
643 * but we use the same procedure by multiplying r with scale and truncating to integer.
645 rt = _fjsp_mul_v2r8(r00,gbscale);
646 itab_tmp = _fjsp_dtox_v2r8(rt);
647 gbeps = _fjsp_sub_v2r8(rt,_fjsp_xtod_v2r8(itab_tmp));
648 _fjsp_store_v2r8(&gbconv.simd,itab_tmp);
650 Y = _fjsp_load_v2r8( gbtab + 4*gbconv.i[0] );
651 F = _fjsp_setzero_v2r8();
652 GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
653 G = _fjsp_load_v2r8( gbtab + 4*gbconv.i[0] +2);
654 H = _fjsp_setzero_v2r8();
655 GMX_FJSP_TRANSPOSE2_V2R8(G,H);
656 Fp = _fjsp_madd_v2r8(gbeps,_fjsp_madd_v2r8(gbeps,H,G),F);
657 VV = _fjsp_madd_v2r8(gbeps,Fp,Y);
658 vgb = _fjsp_mul_v2r8(gbqqfactor,VV);
660 twogbeps = _fjsp_add_v2r8(gbeps,gbeps);
661 FF = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twogbeps,H,G),gbeps,Fp);
662 fgb = _fjsp_mul_v2r8(gbqqfactor,_fjsp_mul_v2r8(FF,gbscale));
663 dvdatmp = _fjsp_mul_v2r8(minushalf,_fjsp_madd_v2r8(fgb,r00,vgb));
664 dvdasum = _fjsp_add_v2r8(dvdasum,dvdatmp);
665 gmx_fjsp_increment_1real_v2r8(dvda+jnrA,_fjsp_mul_v2r8(dvdatmp,_fjsp_mul_v2r8(isaj0,isaj0)));
666 velec = _fjsp_mul_v2r8(qq00,rinv00);
667 felec = _fjsp_mul_v2r8(_fjsp_msub_v2r8(velec,rinv00,fgb),rinv00);
669 /* LENNARD-JONES DISPERSION/REPULSION */
671 rinvsix = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
672 fvdw = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
674 fscal = _fjsp_add_v2r8(felec,fvdw);
676 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
678 /* Update vectorial force */
679 fix0 = _fjsp_madd_v2r8(dx00,fscal,fix0);
680 fiy0 = _fjsp_madd_v2r8(dy00,fscal,fiy0);
681 fiz0 = _fjsp_madd_v2r8(dz00,fscal,fiz0);
683 gmx_fjsp_decrement_fma_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fscal,dx00,dy00,dz00);
685 /* Inner loop uses 67 flops */
688 /* End of innermost loop */
690 gmx_fjsp_update_iforce_1atom_swizzle_v2r8(fix0,fiy0,fiz0,
691 f+i_coord_offset,fshift+i_shift_offset);
693 dvdasum = _fjsp_mul_v2r8(dvdasum, _fjsp_mul_v2r8(isai0,isai0));
694 gmx_fjsp_update_1pot_v2r8(dvdasum,dvda+inr);
696 /* Increment number of inner iterations */
697 inneriter += j_index_end - j_index_start;
699 /* Outer loop uses 7 flops */
702 /* Increment number of outer iterations */
705 /* Update outer/inner flops */
707 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_F,outeriter*7 + inneriter*67);