2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2012,2013, by the GROMACS development team, led by
5 * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
6 * and including many others, as listed in the AUTHORS file in the
7 * top-level source directory and at http://www.gromacs.org.
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 sse4_1_double kernel generator.
44 #include "../nb_kernel.h"
45 #include "types/simple.h"
49 #include "gromacs/simd/math_x86_sse4_1_double.h"
50 #include "kernelutil_x86_sse4_1_double.h"
53 * Gromacs nonbonded kernel: nb_kernel_ElecRFCut_VdwLJSh_GeomP1P1_VF_sse4_1_double
54 * Electrostatics interaction: ReactionField
55 * VdW interaction: LennardJones
56 * Geometry: Particle-Particle
57 * Calculate force/pot: PotentialAndForce
60 nb_kernel_ElecRFCut_VdwLJSh_GeomP1P1_VF_sse4_1_double
61 (t_nblist * gmx_restrict nlist,
62 rvec * gmx_restrict xx,
63 rvec * gmx_restrict ff,
64 t_forcerec * gmx_restrict fr,
65 t_mdatoms * gmx_restrict mdatoms,
66 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
67 t_nrnb * gmx_restrict nrnb)
69 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
70 * just 0 for non-waters.
71 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
72 * jnr indices corresponding to data put in the four positions in the SIMD register.
74 int i_shift_offset,i_coord_offset,outeriter,inneriter;
75 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
77 int j_coord_offsetA,j_coord_offsetB;
78 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
80 real *shiftvec,*fshift,*x,*f;
81 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
83 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
84 int vdwjidx0A,vdwjidx0B;
85 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
86 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
87 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
90 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
93 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
94 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
95 __m128d dummy_mask,cutoff_mask;
96 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
97 __m128d one = _mm_set1_pd(1.0);
98 __m128d two = _mm_set1_pd(2.0);
104 jindex = nlist->jindex;
106 shiftidx = nlist->shift;
108 shiftvec = fr->shift_vec[0];
109 fshift = fr->fshift[0];
110 facel = _mm_set1_pd(fr->epsfac);
111 charge = mdatoms->chargeA;
112 krf = _mm_set1_pd(fr->ic->k_rf);
113 krf2 = _mm_set1_pd(fr->ic->k_rf*2.0);
114 crf = _mm_set1_pd(fr->ic->c_rf);
115 nvdwtype = fr->ntype;
117 vdwtype = mdatoms->typeA;
119 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
120 rcutoff_scalar = fr->rcoulomb;
121 rcutoff = _mm_set1_pd(rcutoff_scalar);
122 rcutoff2 = _mm_mul_pd(rcutoff,rcutoff);
124 sh_vdw_invrcut6 = _mm_set1_pd(fr->ic->sh_invrc6);
125 rvdw = _mm_set1_pd(fr->rvdw);
127 /* Avoid stupid compiler warnings */
135 /* Start outer loop over neighborlists */
136 for(iidx=0; iidx<nri; iidx++)
138 /* Load shift vector for this list */
139 i_shift_offset = DIM*shiftidx[iidx];
141 /* Load limits for loop over neighbors */
142 j_index_start = jindex[iidx];
143 j_index_end = jindex[iidx+1];
145 /* Get outer coordinate index */
147 i_coord_offset = DIM*inr;
149 /* Load i particle coords and add shift vector */
150 gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
152 fix0 = _mm_setzero_pd();
153 fiy0 = _mm_setzero_pd();
154 fiz0 = _mm_setzero_pd();
156 /* Load parameters for i particles */
157 iq0 = _mm_mul_pd(facel,_mm_load1_pd(charge+inr+0));
158 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
160 /* Reset potential sums */
161 velecsum = _mm_setzero_pd();
162 vvdwsum = _mm_setzero_pd();
164 /* Start inner kernel loop */
165 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
168 /* Get j neighbor index, and coordinate index */
171 j_coord_offsetA = DIM*jnrA;
172 j_coord_offsetB = DIM*jnrB;
174 /* load j atom coordinates */
175 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
178 /* Calculate displacement vector */
179 dx00 = _mm_sub_pd(ix0,jx0);
180 dy00 = _mm_sub_pd(iy0,jy0);
181 dz00 = _mm_sub_pd(iz0,jz0);
183 /* Calculate squared distance and things based on it */
184 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
186 rinv00 = gmx_mm_invsqrt_pd(rsq00);
188 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
190 /* Load parameters for j particles */
191 jq0 = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
192 vdwjidx0A = 2*vdwtype[jnrA+0];
193 vdwjidx0B = 2*vdwtype[jnrB+0];
195 /**************************
196 * CALCULATE INTERACTIONS *
197 **************************/
199 if (gmx_mm_any_lt(rsq00,rcutoff2))
202 /* Compute parameters for interactions between i and j atoms */
203 qq00 = _mm_mul_pd(iq0,jq0);
204 gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
205 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
207 /* REACTION-FIELD ELECTROSTATICS */
208 velec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_add_pd(rinv00,_mm_mul_pd(krf,rsq00)),crf));
209 felec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
211 /* LENNARD-JONES DISPERSION/REPULSION */
213 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
214 vvdw6 = _mm_mul_pd(c6_00,rinvsix);
215 vvdw12 = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
216 vvdw = _mm_sub_pd(_mm_mul_pd( _mm_sub_pd(vvdw12 , _mm_mul_pd(c12_00,_mm_mul_pd(sh_vdw_invrcut6,sh_vdw_invrcut6))), one_twelfth) ,
217 _mm_mul_pd( _mm_sub_pd(vvdw6,_mm_mul_pd(c6_00,sh_vdw_invrcut6)),one_sixth));
218 fvdw = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
220 cutoff_mask = _mm_cmplt_pd(rsq00,rcutoff2);
222 /* Update potential sum for this i atom from the interaction with this j atom. */
223 velec = _mm_and_pd(velec,cutoff_mask);
224 velecsum = _mm_add_pd(velecsum,velec);
225 vvdw = _mm_and_pd(vvdw,cutoff_mask);
226 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
228 fscal = _mm_add_pd(felec,fvdw);
230 fscal = _mm_and_pd(fscal,cutoff_mask);
232 /* Calculate temporary vectorial force */
233 tx = _mm_mul_pd(fscal,dx00);
234 ty = _mm_mul_pd(fscal,dy00);
235 tz = _mm_mul_pd(fscal,dz00);
237 /* Update vectorial force */
238 fix0 = _mm_add_pd(fix0,tx);
239 fiy0 = _mm_add_pd(fiy0,ty);
240 fiz0 = _mm_add_pd(fiz0,tz);
242 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,tx,ty,tz);
246 /* Inner loop uses 54 flops */
253 j_coord_offsetA = DIM*jnrA;
255 /* load j atom coordinates */
256 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
259 /* Calculate displacement vector */
260 dx00 = _mm_sub_pd(ix0,jx0);
261 dy00 = _mm_sub_pd(iy0,jy0);
262 dz00 = _mm_sub_pd(iz0,jz0);
264 /* Calculate squared distance and things based on it */
265 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
267 rinv00 = gmx_mm_invsqrt_pd(rsq00);
269 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
271 /* Load parameters for j particles */
272 jq0 = _mm_load_sd(charge+jnrA+0);
273 vdwjidx0A = 2*vdwtype[jnrA+0];
275 /**************************
276 * CALCULATE INTERACTIONS *
277 **************************/
279 if (gmx_mm_any_lt(rsq00,rcutoff2))
282 /* Compute parameters for interactions between i and j atoms */
283 qq00 = _mm_mul_pd(iq0,jq0);
284 gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
286 /* REACTION-FIELD ELECTROSTATICS */
287 velec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_add_pd(rinv00,_mm_mul_pd(krf,rsq00)),crf));
288 felec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
290 /* LENNARD-JONES DISPERSION/REPULSION */
292 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
293 vvdw6 = _mm_mul_pd(c6_00,rinvsix);
294 vvdw12 = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
295 vvdw = _mm_sub_pd(_mm_mul_pd( _mm_sub_pd(vvdw12 , _mm_mul_pd(c12_00,_mm_mul_pd(sh_vdw_invrcut6,sh_vdw_invrcut6))), one_twelfth) ,
296 _mm_mul_pd( _mm_sub_pd(vvdw6,_mm_mul_pd(c6_00,sh_vdw_invrcut6)),one_sixth));
297 fvdw = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
299 cutoff_mask = _mm_cmplt_pd(rsq00,rcutoff2);
301 /* Update potential sum for this i atom from the interaction with this j atom. */
302 velec = _mm_and_pd(velec,cutoff_mask);
303 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
304 velecsum = _mm_add_pd(velecsum,velec);
305 vvdw = _mm_and_pd(vvdw,cutoff_mask);
306 vvdw = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
307 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
309 fscal = _mm_add_pd(felec,fvdw);
311 fscal = _mm_and_pd(fscal,cutoff_mask);
313 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
315 /* Calculate temporary vectorial force */
316 tx = _mm_mul_pd(fscal,dx00);
317 ty = _mm_mul_pd(fscal,dy00);
318 tz = _mm_mul_pd(fscal,dz00);
320 /* Update vectorial force */
321 fix0 = _mm_add_pd(fix0,tx);
322 fiy0 = _mm_add_pd(fiy0,ty);
323 fiz0 = _mm_add_pd(fiz0,tz);
325 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,tx,ty,tz);
329 /* Inner loop uses 54 flops */
332 /* End of innermost loop */
334 gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
335 f+i_coord_offset,fshift+i_shift_offset);
338 /* Update potential energies */
339 gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
340 gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
342 /* Increment number of inner iterations */
343 inneriter += j_index_end - j_index_start;
345 /* Outer loop uses 9 flops */
348 /* Increment number of outer iterations */
351 /* Update outer/inner flops */
353 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_VF,outeriter*9 + inneriter*54);
356 * Gromacs nonbonded kernel: nb_kernel_ElecRFCut_VdwLJSh_GeomP1P1_F_sse4_1_double
357 * Electrostatics interaction: ReactionField
358 * VdW interaction: LennardJones
359 * Geometry: Particle-Particle
360 * Calculate force/pot: Force
363 nb_kernel_ElecRFCut_VdwLJSh_GeomP1P1_F_sse4_1_double
364 (t_nblist * gmx_restrict nlist,
365 rvec * gmx_restrict xx,
366 rvec * gmx_restrict ff,
367 t_forcerec * gmx_restrict fr,
368 t_mdatoms * gmx_restrict mdatoms,
369 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
370 t_nrnb * gmx_restrict nrnb)
372 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
373 * just 0 for non-waters.
374 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
375 * jnr indices corresponding to data put in the four positions in the SIMD register.
377 int i_shift_offset,i_coord_offset,outeriter,inneriter;
378 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
380 int j_coord_offsetA,j_coord_offsetB;
381 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
383 real *shiftvec,*fshift,*x,*f;
384 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
386 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
387 int vdwjidx0A,vdwjidx0B;
388 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
389 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
390 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
393 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
396 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
397 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
398 __m128d dummy_mask,cutoff_mask;
399 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
400 __m128d one = _mm_set1_pd(1.0);
401 __m128d two = _mm_set1_pd(2.0);
407 jindex = nlist->jindex;
409 shiftidx = nlist->shift;
411 shiftvec = fr->shift_vec[0];
412 fshift = fr->fshift[0];
413 facel = _mm_set1_pd(fr->epsfac);
414 charge = mdatoms->chargeA;
415 krf = _mm_set1_pd(fr->ic->k_rf);
416 krf2 = _mm_set1_pd(fr->ic->k_rf*2.0);
417 crf = _mm_set1_pd(fr->ic->c_rf);
418 nvdwtype = fr->ntype;
420 vdwtype = mdatoms->typeA;
422 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
423 rcutoff_scalar = fr->rcoulomb;
424 rcutoff = _mm_set1_pd(rcutoff_scalar);
425 rcutoff2 = _mm_mul_pd(rcutoff,rcutoff);
427 sh_vdw_invrcut6 = _mm_set1_pd(fr->ic->sh_invrc6);
428 rvdw = _mm_set1_pd(fr->rvdw);
430 /* Avoid stupid compiler warnings */
438 /* Start outer loop over neighborlists */
439 for(iidx=0; iidx<nri; iidx++)
441 /* Load shift vector for this list */
442 i_shift_offset = DIM*shiftidx[iidx];
444 /* Load limits for loop over neighbors */
445 j_index_start = jindex[iidx];
446 j_index_end = jindex[iidx+1];
448 /* Get outer coordinate index */
450 i_coord_offset = DIM*inr;
452 /* Load i particle coords and add shift vector */
453 gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
455 fix0 = _mm_setzero_pd();
456 fiy0 = _mm_setzero_pd();
457 fiz0 = _mm_setzero_pd();
459 /* Load parameters for i particles */
460 iq0 = _mm_mul_pd(facel,_mm_load1_pd(charge+inr+0));
461 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
463 /* Start inner kernel loop */
464 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
467 /* Get j neighbor index, and coordinate index */
470 j_coord_offsetA = DIM*jnrA;
471 j_coord_offsetB = DIM*jnrB;
473 /* load j atom coordinates */
474 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
477 /* Calculate displacement vector */
478 dx00 = _mm_sub_pd(ix0,jx0);
479 dy00 = _mm_sub_pd(iy0,jy0);
480 dz00 = _mm_sub_pd(iz0,jz0);
482 /* Calculate squared distance and things based on it */
483 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
485 rinv00 = gmx_mm_invsqrt_pd(rsq00);
487 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
489 /* Load parameters for j particles */
490 jq0 = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
491 vdwjidx0A = 2*vdwtype[jnrA+0];
492 vdwjidx0B = 2*vdwtype[jnrB+0];
494 /**************************
495 * CALCULATE INTERACTIONS *
496 **************************/
498 if (gmx_mm_any_lt(rsq00,rcutoff2))
501 /* Compute parameters for interactions between i and j atoms */
502 qq00 = _mm_mul_pd(iq0,jq0);
503 gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
504 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
506 /* REACTION-FIELD ELECTROSTATICS */
507 felec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
509 /* LENNARD-JONES DISPERSION/REPULSION */
511 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
512 fvdw = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
514 cutoff_mask = _mm_cmplt_pd(rsq00,rcutoff2);
516 fscal = _mm_add_pd(felec,fvdw);
518 fscal = _mm_and_pd(fscal,cutoff_mask);
520 /* Calculate temporary vectorial force */
521 tx = _mm_mul_pd(fscal,dx00);
522 ty = _mm_mul_pd(fscal,dy00);
523 tz = _mm_mul_pd(fscal,dz00);
525 /* Update vectorial force */
526 fix0 = _mm_add_pd(fix0,tx);
527 fiy0 = _mm_add_pd(fiy0,ty);
528 fiz0 = _mm_add_pd(fiz0,tz);
530 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,tx,ty,tz);
534 /* Inner loop uses 37 flops */
541 j_coord_offsetA = DIM*jnrA;
543 /* load j atom coordinates */
544 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
547 /* Calculate displacement vector */
548 dx00 = _mm_sub_pd(ix0,jx0);
549 dy00 = _mm_sub_pd(iy0,jy0);
550 dz00 = _mm_sub_pd(iz0,jz0);
552 /* Calculate squared distance and things based on it */
553 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
555 rinv00 = gmx_mm_invsqrt_pd(rsq00);
557 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
559 /* Load parameters for j particles */
560 jq0 = _mm_load_sd(charge+jnrA+0);
561 vdwjidx0A = 2*vdwtype[jnrA+0];
563 /**************************
564 * CALCULATE INTERACTIONS *
565 **************************/
567 if (gmx_mm_any_lt(rsq00,rcutoff2))
570 /* Compute parameters for interactions between i and j atoms */
571 qq00 = _mm_mul_pd(iq0,jq0);
572 gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
574 /* REACTION-FIELD ELECTROSTATICS */
575 felec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
577 /* LENNARD-JONES DISPERSION/REPULSION */
579 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
580 fvdw = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
582 cutoff_mask = _mm_cmplt_pd(rsq00,rcutoff2);
584 fscal = _mm_add_pd(felec,fvdw);
586 fscal = _mm_and_pd(fscal,cutoff_mask);
588 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
590 /* Calculate temporary vectorial force */
591 tx = _mm_mul_pd(fscal,dx00);
592 ty = _mm_mul_pd(fscal,dy00);
593 tz = _mm_mul_pd(fscal,dz00);
595 /* Update vectorial force */
596 fix0 = _mm_add_pd(fix0,tx);
597 fiy0 = _mm_add_pd(fiy0,ty);
598 fiz0 = _mm_add_pd(fiz0,tz);
600 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,tx,ty,tz);
604 /* Inner loop uses 37 flops */
607 /* End of innermost loop */
609 gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
610 f+i_coord_offset,fshift+i_shift_offset);
612 /* Increment number of inner iterations */
613 inneriter += j_index_end - j_index_start;
615 /* Outer loop uses 7 flops */
618 /* Increment number of outer iterations */
621 /* Update outer/inner flops */
623 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_F,outeriter*7 + inneriter*37);