2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2012,2013,2014,2015,2017, 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 "gromacs/gmxlib/nrnb.h"
47 #include "kernelutil_x86_sse4_1_double.h"
50 * Gromacs nonbonded kernel: nb_kernel_ElecCoul_VdwCSTab_GeomW4P1_VF_sse4_1_double
51 * Electrostatics interaction: Coulomb
52 * VdW interaction: CubicSplineTable
53 * Geometry: Water4-Particle
54 * Calculate force/pot: PotentialAndForce
57 nb_kernel_ElecCoul_VdwCSTab_GeomW4P1_VF_sse4_1_double
58 (t_nblist * gmx_restrict nlist,
59 rvec * gmx_restrict xx,
60 rvec * gmx_restrict ff,
61 struct t_forcerec * gmx_restrict fr,
62 t_mdatoms * gmx_restrict mdatoms,
63 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
64 t_nrnb * gmx_restrict nrnb)
66 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
67 * just 0 for non-waters.
68 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
69 * jnr indices corresponding to data put in the four positions in the SIMD register.
71 int i_shift_offset,i_coord_offset,outeriter,inneriter;
72 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
74 int j_coord_offsetA,j_coord_offsetB;
75 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
77 real *shiftvec,*fshift,*x,*f;
78 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
80 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
82 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
84 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
86 __m128d ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
87 int vdwjidx0A,vdwjidx0B;
88 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
89 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
90 __m128d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
91 __m128d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
92 __m128d dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
93 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
96 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
99 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
100 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
102 __m128i ifour = _mm_set1_epi32(4);
103 __m128d rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
105 __m128d dummy_mask,cutoff_mask;
106 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
107 __m128d one = _mm_set1_pd(1.0);
108 __m128d two = _mm_set1_pd(2.0);
114 jindex = nlist->jindex;
116 shiftidx = nlist->shift;
118 shiftvec = fr->shift_vec[0];
119 fshift = fr->fshift[0];
120 facel = _mm_set1_pd(fr->ic->epsfac);
121 charge = mdatoms->chargeA;
122 nvdwtype = fr->ntype;
124 vdwtype = mdatoms->typeA;
126 vftab = kernel_data->table_vdw->data;
127 vftabscale = _mm_set1_pd(kernel_data->table_vdw->scale);
129 /* Setup water-specific parameters */
130 inr = nlist->iinr[0];
131 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
132 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
133 iq3 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
134 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
136 /* Avoid stupid compiler warnings */
144 /* Start outer loop over neighborlists */
145 for(iidx=0; iidx<nri; iidx++)
147 /* Load shift vector for this list */
148 i_shift_offset = DIM*shiftidx[iidx];
150 /* Load limits for loop over neighbors */
151 j_index_start = jindex[iidx];
152 j_index_end = jindex[iidx+1];
154 /* Get outer coordinate index */
156 i_coord_offset = DIM*inr;
158 /* Load i particle coords and add shift vector */
159 gmx_mm_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
160 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
162 fix0 = _mm_setzero_pd();
163 fiy0 = _mm_setzero_pd();
164 fiz0 = _mm_setzero_pd();
165 fix1 = _mm_setzero_pd();
166 fiy1 = _mm_setzero_pd();
167 fiz1 = _mm_setzero_pd();
168 fix2 = _mm_setzero_pd();
169 fiy2 = _mm_setzero_pd();
170 fiz2 = _mm_setzero_pd();
171 fix3 = _mm_setzero_pd();
172 fiy3 = _mm_setzero_pd();
173 fiz3 = _mm_setzero_pd();
175 /* Reset potential sums */
176 velecsum = _mm_setzero_pd();
177 vvdwsum = _mm_setzero_pd();
179 /* Start inner kernel loop */
180 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
183 /* Get j neighbor index, and coordinate index */
186 j_coord_offsetA = DIM*jnrA;
187 j_coord_offsetB = DIM*jnrB;
189 /* load j atom coordinates */
190 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
193 /* Calculate displacement vector */
194 dx00 = _mm_sub_pd(ix0,jx0);
195 dy00 = _mm_sub_pd(iy0,jy0);
196 dz00 = _mm_sub_pd(iz0,jz0);
197 dx10 = _mm_sub_pd(ix1,jx0);
198 dy10 = _mm_sub_pd(iy1,jy0);
199 dz10 = _mm_sub_pd(iz1,jz0);
200 dx20 = _mm_sub_pd(ix2,jx0);
201 dy20 = _mm_sub_pd(iy2,jy0);
202 dz20 = _mm_sub_pd(iz2,jz0);
203 dx30 = _mm_sub_pd(ix3,jx0);
204 dy30 = _mm_sub_pd(iy3,jy0);
205 dz30 = _mm_sub_pd(iz3,jz0);
207 /* Calculate squared distance and things based on it */
208 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
209 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
210 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
211 rsq30 = gmx_mm_calc_rsq_pd(dx30,dy30,dz30);
213 rinv00 = sse41_invsqrt_d(rsq00);
214 rinv10 = sse41_invsqrt_d(rsq10);
215 rinv20 = sse41_invsqrt_d(rsq20);
216 rinv30 = sse41_invsqrt_d(rsq30);
218 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
219 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
220 rinvsq30 = _mm_mul_pd(rinv30,rinv30);
222 /* Load parameters for j particles */
223 jq0 = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
224 vdwjidx0A = 2*vdwtype[jnrA+0];
225 vdwjidx0B = 2*vdwtype[jnrB+0];
227 fjx0 = _mm_setzero_pd();
228 fjy0 = _mm_setzero_pd();
229 fjz0 = _mm_setzero_pd();
231 /**************************
232 * CALCULATE INTERACTIONS *
233 **************************/
235 r00 = _mm_mul_pd(rsq00,rinv00);
237 /* Compute parameters for interactions between i and j atoms */
238 gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
239 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
241 /* Calculate table index by multiplying r with table scale and truncate to integer */
242 rt = _mm_mul_pd(r00,vftabscale);
243 vfitab = _mm_cvttpd_epi32(rt);
244 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
245 vfitab = _mm_slli_epi32(vfitab,3);
247 /* CUBIC SPLINE TABLE DISPERSION */
248 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
249 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
250 GMX_MM_TRANSPOSE2_PD(Y,F);
251 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
252 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
253 GMX_MM_TRANSPOSE2_PD(G,H);
254 Heps = _mm_mul_pd(vfeps,H);
255 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
256 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
257 vvdw6 = _mm_mul_pd(c6_00,VV);
258 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
259 fvdw6 = _mm_mul_pd(c6_00,FF);
261 /* CUBIC SPLINE TABLE REPULSION */
262 vfitab = _mm_add_epi32(vfitab,ifour);
263 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
264 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
265 GMX_MM_TRANSPOSE2_PD(Y,F);
266 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
267 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
268 GMX_MM_TRANSPOSE2_PD(G,H);
269 Heps = _mm_mul_pd(vfeps,H);
270 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
271 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
272 vvdw12 = _mm_mul_pd(c12_00,VV);
273 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
274 fvdw12 = _mm_mul_pd(c12_00,FF);
275 vvdw = _mm_add_pd(vvdw12,vvdw6);
276 fvdw = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
278 /* Update potential sum for this i atom from the interaction with this j atom. */
279 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
283 /* Calculate temporary vectorial force */
284 tx = _mm_mul_pd(fscal,dx00);
285 ty = _mm_mul_pd(fscal,dy00);
286 tz = _mm_mul_pd(fscal,dz00);
288 /* Update vectorial force */
289 fix0 = _mm_add_pd(fix0,tx);
290 fiy0 = _mm_add_pd(fiy0,ty);
291 fiz0 = _mm_add_pd(fiz0,tz);
293 fjx0 = _mm_add_pd(fjx0,tx);
294 fjy0 = _mm_add_pd(fjy0,ty);
295 fjz0 = _mm_add_pd(fjz0,tz);
297 /**************************
298 * CALCULATE INTERACTIONS *
299 **************************/
301 /* Compute parameters for interactions between i and j atoms */
302 qq10 = _mm_mul_pd(iq1,jq0);
304 /* COULOMB ELECTROSTATICS */
305 velec = _mm_mul_pd(qq10,rinv10);
306 felec = _mm_mul_pd(velec,rinvsq10);
308 /* Update potential sum for this i atom from the interaction with this j atom. */
309 velecsum = _mm_add_pd(velecsum,velec);
313 /* Calculate temporary vectorial force */
314 tx = _mm_mul_pd(fscal,dx10);
315 ty = _mm_mul_pd(fscal,dy10);
316 tz = _mm_mul_pd(fscal,dz10);
318 /* Update vectorial force */
319 fix1 = _mm_add_pd(fix1,tx);
320 fiy1 = _mm_add_pd(fiy1,ty);
321 fiz1 = _mm_add_pd(fiz1,tz);
323 fjx0 = _mm_add_pd(fjx0,tx);
324 fjy0 = _mm_add_pd(fjy0,ty);
325 fjz0 = _mm_add_pd(fjz0,tz);
327 /**************************
328 * CALCULATE INTERACTIONS *
329 **************************/
331 /* Compute parameters for interactions between i and j atoms */
332 qq20 = _mm_mul_pd(iq2,jq0);
334 /* COULOMB ELECTROSTATICS */
335 velec = _mm_mul_pd(qq20,rinv20);
336 felec = _mm_mul_pd(velec,rinvsq20);
338 /* Update potential sum for this i atom from the interaction with this j atom. */
339 velecsum = _mm_add_pd(velecsum,velec);
343 /* Calculate temporary vectorial force */
344 tx = _mm_mul_pd(fscal,dx20);
345 ty = _mm_mul_pd(fscal,dy20);
346 tz = _mm_mul_pd(fscal,dz20);
348 /* Update vectorial force */
349 fix2 = _mm_add_pd(fix2,tx);
350 fiy2 = _mm_add_pd(fiy2,ty);
351 fiz2 = _mm_add_pd(fiz2,tz);
353 fjx0 = _mm_add_pd(fjx0,tx);
354 fjy0 = _mm_add_pd(fjy0,ty);
355 fjz0 = _mm_add_pd(fjz0,tz);
357 /**************************
358 * CALCULATE INTERACTIONS *
359 **************************/
361 /* Compute parameters for interactions between i and j atoms */
362 qq30 = _mm_mul_pd(iq3,jq0);
364 /* COULOMB ELECTROSTATICS */
365 velec = _mm_mul_pd(qq30,rinv30);
366 felec = _mm_mul_pd(velec,rinvsq30);
368 /* Update potential sum for this i atom from the interaction with this j atom. */
369 velecsum = _mm_add_pd(velecsum,velec);
373 /* Calculate temporary vectorial force */
374 tx = _mm_mul_pd(fscal,dx30);
375 ty = _mm_mul_pd(fscal,dy30);
376 tz = _mm_mul_pd(fscal,dz30);
378 /* Update vectorial force */
379 fix3 = _mm_add_pd(fix3,tx);
380 fiy3 = _mm_add_pd(fiy3,ty);
381 fiz3 = _mm_add_pd(fiz3,tz);
383 fjx0 = _mm_add_pd(fjx0,tx);
384 fjy0 = _mm_add_pd(fjy0,ty);
385 fjz0 = _mm_add_pd(fjz0,tz);
387 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
389 /* Inner loop uses 143 flops */
396 j_coord_offsetA = DIM*jnrA;
398 /* load j atom coordinates */
399 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
402 /* Calculate displacement vector */
403 dx00 = _mm_sub_pd(ix0,jx0);
404 dy00 = _mm_sub_pd(iy0,jy0);
405 dz00 = _mm_sub_pd(iz0,jz0);
406 dx10 = _mm_sub_pd(ix1,jx0);
407 dy10 = _mm_sub_pd(iy1,jy0);
408 dz10 = _mm_sub_pd(iz1,jz0);
409 dx20 = _mm_sub_pd(ix2,jx0);
410 dy20 = _mm_sub_pd(iy2,jy0);
411 dz20 = _mm_sub_pd(iz2,jz0);
412 dx30 = _mm_sub_pd(ix3,jx0);
413 dy30 = _mm_sub_pd(iy3,jy0);
414 dz30 = _mm_sub_pd(iz3,jz0);
416 /* Calculate squared distance and things based on it */
417 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
418 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
419 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
420 rsq30 = gmx_mm_calc_rsq_pd(dx30,dy30,dz30);
422 rinv00 = sse41_invsqrt_d(rsq00);
423 rinv10 = sse41_invsqrt_d(rsq10);
424 rinv20 = sse41_invsqrt_d(rsq20);
425 rinv30 = sse41_invsqrt_d(rsq30);
427 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
428 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
429 rinvsq30 = _mm_mul_pd(rinv30,rinv30);
431 /* Load parameters for j particles */
432 jq0 = _mm_load_sd(charge+jnrA+0);
433 vdwjidx0A = 2*vdwtype[jnrA+0];
435 fjx0 = _mm_setzero_pd();
436 fjy0 = _mm_setzero_pd();
437 fjz0 = _mm_setzero_pd();
439 /**************************
440 * CALCULATE INTERACTIONS *
441 **************************/
443 r00 = _mm_mul_pd(rsq00,rinv00);
445 /* Compute parameters for interactions between i and j atoms */
446 gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
448 /* Calculate table index by multiplying r with table scale and truncate to integer */
449 rt = _mm_mul_pd(r00,vftabscale);
450 vfitab = _mm_cvttpd_epi32(rt);
451 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
452 vfitab = _mm_slli_epi32(vfitab,3);
454 /* CUBIC SPLINE TABLE DISPERSION */
455 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
456 F = _mm_setzero_pd();
457 GMX_MM_TRANSPOSE2_PD(Y,F);
458 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
459 H = _mm_setzero_pd();
460 GMX_MM_TRANSPOSE2_PD(G,H);
461 Heps = _mm_mul_pd(vfeps,H);
462 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
463 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
464 vvdw6 = _mm_mul_pd(c6_00,VV);
465 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
466 fvdw6 = _mm_mul_pd(c6_00,FF);
468 /* CUBIC SPLINE TABLE REPULSION */
469 vfitab = _mm_add_epi32(vfitab,ifour);
470 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
471 F = _mm_setzero_pd();
472 GMX_MM_TRANSPOSE2_PD(Y,F);
473 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
474 H = _mm_setzero_pd();
475 GMX_MM_TRANSPOSE2_PD(G,H);
476 Heps = _mm_mul_pd(vfeps,H);
477 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
478 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
479 vvdw12 = _mm_mul_pd(c12_00,VV);
480 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
481 fvdw12 = _mm_mul_pd(c12_00,FF);
482 vvdw = _mm_add_pd(vvdw12,vvdw6);
483 fvdw = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
485 /* Update potential sum for this i atom from the interaction with this j atom. */
486 vvdw = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
487 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
491 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
493 /* Calculate temporary vectorial force */
494 tx = _mm_mul_pd(fscal,dx00);
495 ty = _mm_mul_pd(fscal,dy00);
496 tz = _mm_mul_pd(fscal,dz00);
498 /* Update vectorial force */
499 fix0 = _mm_add_pd(fix0,tx);
500 fiy0 = _mm_add_pd(fiy0,ty);
501 fiz0 = _mm_add_pd(fiz0,tz);
503 fjx0 = _mm_add_pd(fjx0,tx);
504 fjy0 = _mm_add_pd(fjy0,ty);
505 fjz0 = _mm_add_pd(fjz0,tz);
507 /**************************
508 * CALCULATE INTERACTIONS *
509 **************************/
511 /* Compute parameters for interactions between i and j atoms */
512 qq10 = _mm_mul_pd(iq1,jq0);
514 /* COULOMB ELECTROSTATICS */
515 velec = _mm_mul_pd(qq10,rinv10);
516 felec = _mm_mul_pd(velec,rinvsq10);
518 /* Update potential sum for this i atom from the interaction with this j atom. */
519 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
520 velecsum = _mm_add_pd(velecsum,velec);
524 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
526 /* Calculate temporary vectorial force */
527 tx = _mm_mul_pd(fscal,dx10);
528 ty = _mm_mul_pd(fscal,dy10);
529 tz = _mm_mul_pd(fscal,dz10);
531 /* Update vectorial force */
532 fix1 = _mm_add_pd(fix1,tx);
533 fiy1 = _mm_add_pd(fiy1,ty);
534 fiz1 = _mm_add_pd(fiz1,tz);
536 fjx0 = _mm_add_pd(fjx0,tx);
537 fjy0 = _mm_add_pd(fjy0,ty);
538 fjz0 = _mm_add_pd(fjz0,tz);
540 /**************************
541 * CALCULATE INTERACTIONS *
542 **************************/
544 /* Compute parameters for interactions between i and j atoms */
545 qq20 = _mm_mul_pd(iq2,jq0);
547 /* COULOMB ELECTROSTATICS */
548 velec = _mm_mul_pd(qq20,rinv20);
549 felec = _mm_mul_pd(velec,rinvsq20);
551 /* Update potential sum for this i atom from the interaction with this j atom. */
552 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
553 velecsum = _mm_add_pd(velecsum,velec);
557 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
559 /* Calculate temporary vectorial force */
560 tx = _mm_mul_pd(fscal,dx20);
561 ty = _mm_mul_pd(fscal,dy20);
562 tz = _mm_mul_pd(fscal,dz20);
564 /* Update vectorial force */
565 fix2 = _mm_add_pd(fix2,tx);
566 fiy2 = _mm_add_pd(fiy2,ty);
567 fiz2 = _mm_add_pd(fiz2,tz);
569 fjx0 = _mm_add_pd(fjx0,tx);
570 fjy0 = _mm_add_pd(fjy0,ty);
571 fjz0 = _mm_add_pd(fjz0,tz);
573 /**************************
574 * CALCULATE INTERACTIONS *
575 **************************/
577 /* Compute parameters for interactions between i and j atoms */
578 qq30 = _mm_mul_pd(iq3,jq0);
580 /* COULOMB ELECTROSTATICS */
581 velec = _mm_mul_pd(qq30,rinv30);
582 felec = _mm_mul_pd(velec,rinvsq30);
584 /* Update potential sum for this i atom from the interaction with this j atom. */
585 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
586 velecsum = _mm_add_pd(velecsum,velec);
590 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
592 /* Calculate temporary vectorial force */
593 tx = _mm_mul_pd(fscal,dx30);
594 ty = _mm_mul_pd(fscal,dy30);
595 tz = _mm_mul_pd(fscal,dz30);
597 /* Update vectorial force */
598 fix3 = _mm_add_pd(fix3,tx);
599 fiy3 = _mm_add_pd(fiy3,ty);
600 fiz3 = _mm_add_pd(fiz3,tz);
602 fjx0 = _mm_add_pd(fjx0,tx);
603 fjy0 = _mm_add_pd(fjy0,ty);
604 fjz0 = _mm_add_pd(fjz0,tz);
606 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0);
608 /* Inner loop uses 143 flops */
611 /* End of innermost loop */
613 gmx_mm_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
614 f+i_coord_offset,fshift+i_shift_offset);
617 /* Update potential energies */
618 gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
619 gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
621 /* Increment number of inner iterations */
622 inneriter += j_index_end - j_index_start;
624 /* Outer loop uses 26 flops */
627 /* Increment number of outer iterations */
630 /* Update outer/inner flops */
632 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_VF,outeriter*26 + inneriter*143);
635 * Gromacs nonbonded kernel: nb_kernel_ElecCoul_VdwCSTab_GeomW4P1_F_sse4_1_double
636 * Electrostatics interaction: Coulomb
637 * VdW interaction: CubicSplineTable
638 * Geometry: Water4-Particle
639 * Calculate force/pot: Force
642 nb_kernel_ElecCoul_VdwCSTab_GeomW4P1_F_sse4_1_double
643 (t_nblist * gmx_restrict nlist,
644 rvec * gmx_restrict xx,
645 rvec * gmx_restrict ff,
646 struct t_forcerec * gmx_restrict fr,
647 t_mdatoms * gmx_restrict mdatoms,
648 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
649 t_nrnb * gmx_restrict nrnb)
651 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
652 * just 0 for non-waters.
653 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
654 * jnr indices corresponding to data put in the four positions in the SIMD register.
656 int i_shift_offset,i_coord_offset,outeriter,inneriter;
657 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
659 int j_coord_offsetA,j_coord_offsetB;
660 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
662 real *shiftvec,*fshift,*x,*f;
663 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
665 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
667 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
669 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
671 __m128d ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
672 int vdwjidx0A,vdwjidx0B;
673 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
674 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
675 __m128d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
676 __m128d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
677 __m128d dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
678 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
681 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
684 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
685 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
687 __m128i ifour = _mm_set1_epi32(4);
688 __m128d rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
690 __m128d dummy_mask,cutoff_mask;
691 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
692 __m128d one = _mm_set1_pd(1.0);
693 __m128d two = _mm_set1_pd(2.0);
699 jindex = nlist->jindex;
701 shiftidx = nlist->shift;
703 shiftvec = fr->shift_vec[0];
704 fshift = fr->fshift[0];
705 facel = _mm_set1_pd(fr->ic->epsfac);
706 charge = mdatoms->chargeA;
707 nvdwtype = fr->ntype;
709 vdwtype = mdatoms->typeA;
711 vftab = kernel_data->table_vdw->data;
712 vftabscale = _mm_set1_pd(kernel_data->table_vdw->scale);
714 /* Setup water-specific parameters */
715 inr = nlist->iinr[0];
716 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
717 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
718 iq3 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
719 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
721 /* Avoid stupid compiler warnings */
729 /* Start outer loop over neighborlists */
730 for(iidx=0; iidx<nri; iidx++)
732 /* Load shift vector for this list */
733 i_shift_offset = DIM*shiftidx[iidx];
735 /* Load limits for loop over neighbors */
736 j_index_start = jindex[iidx];
737 j_index_end = jindex[iidx+1];
739 /* Get outer coordinate index */
741 i_coord_offset = DIM*inr;
743 /* Load i particle coords and add shift vector */
744 gmx_mm_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
745 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
747 fix0 = _mm_setzero_pd();
748 fiy0 = _mm_setzero_pd();
749 fiz0 = _mm_setzero_pd();
750 fix1 = _mm_setzero_pd();
751 fiy1 = _mm_setzero_pd();
752 fiz1 = _mm_setzero_pd();
753 fix2 = _mm_setzero_pd();
754 fiy2 = _mm_setzero_pd();
755 fiz2 = _mm_setzero_pd();
756 fix3 = _mm_setzero_pd();
757 fiy3 = _mm_setzero_pd();
758 fiz3 = _mm_setzero_pd();
760 /* Start inner kernel loop */
761 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
764 /* Get j neighbor index, and coordinate index */
767 j_coord_offsetA = DIM*jnrA;
768 j_coord_offsetB = DIM*jnrB;
770 /* load j atom coordinates */
771 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
774 /* Calculate displacement vector */
775 dx00 = _mm_sub_pd(ix0,jx0);
776 dy00 = _mm_sub_pd(iy0,jy0);
777 dz00 = _mm_sub_pd(iz0,jz0);
778 dx10 = _mm_sub_pd(ix1,jx0);
779 dy10 = _mm_sub_pd(iy1,jy0);
780 dz10 = _mm_sub_pd(iz1,jz0);
781 dx20 = _mm_sub_pd(ix2,jx0);
782 dy20 = _mm_sub_pd(iy2,jy0);
783 dz20 = _mm_sub_pd(iz2,jz0);
784 dx30 = _mm_sub_pd(ix3,jx0);
785 dy30 = _mm_sub_pd(iy3,jy0);
786 dz30 = _mm_sub_pd(iz3,jz0);
788 /* Calculate squared distance and things based on it */
789 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
790 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
791 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
792 rsq30 = gmx_mm_calc_rsq_pd(dx30,dy30,dz30);
794 rinv00 = sse41_invsqrt_d(rsq00);
795 rinv10 = sse41_invsqrt_d(rsq10);
796 rinv20 = sse41_invsqrt_d(rsq20);
797 rinv30 = sse41_invsqrt_d(rsq30);
799 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
800 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
801 rinvsq30 = _mm_mul_pd(rinv30,rinv30);
803 /* Load parameters for j particles */
804 jq0 = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
805 vdwjidx0A = 2*vdwtype[jnrA+0];
806 vdwjidx0B = 2*vdwtype[jnrB+0];
808 fjx0 = _mm_setzero_pd();
809 fjy0 = _mm_setzero_pd();
810 fjz0 = _mm_setzero_pd();
812 /**************************
813 * CALCULATE INTERACTIONS *
814 **************************/
816 r00 = _mm_mul_pd(rsq00,rinv00);
818 /* Compute parameters for interactions between i and j atoms */
819 gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
820 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
822 /* Calculate table index by multiplying r with table scale and truncate to integer */
823 rt = _mm_mul_pd(r00,vftabscale);
824 vfitab = _mm_cvttpd_epi32(rt);
825 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
826 vfitab = _mm_slli_epi32(vfitab,3);
828 /* CUBIC SPLINE TABLE DISPERSION */
829 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
830 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
831 GMX_MM_TRANSPOSE2_PD(Y,F);
832 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
833 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
834 GMX_MM_TRANSPOSE2_PD(G,H);
835 Heps = _mm_mul_pd(vfeps,H);
836 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
837 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
838 fvdw6 = _mm_mul_pd(c6_00,FF);
840 /* CUBIC SPLINE TABLE REPULSION */
841 vfitab = _mm_add_epi32(vfitab,ifour);
842 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
843 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
844 GMX_MM_TRANSPOSE2_PD(Y,F);
845 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
846 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
847 GMX_MM_TRANSPOSE2_PD(G,H);
848 Heps = _mm_mul_pd(vfeps,H);
849 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
850 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
851 fvdw12 = _mm_mul_pd(c12_00,FF);
852 fvdw = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
856 /* Calculate temporary vectorial force */
857 tx = _mm_mul_pd(fscal,dx00);
858 ty = _mm_mul_pd(fscal,dy00);
859 tz = _mm_mul_pd(fscal,dz00);
861 /* Update vectorial force */
862 fix0 = _mm_add_pd(fix0,tx);
863 fiy0 = _mm_add_pd(fiy0,ty);
864 fiz0 = _mm_add_pd(fiz0,tz);
866 fjx0 = _mm_add_pd(fjx0,tx);
867 fjy0 = _mm_add_pd(fjy0,ty);
868 fjz0 = _mm_add_pd(fjz0,tz);
870 /**************************
871 * CALCULATE INTERACTIONS *
872 **************************/
874 /* Compute parameters for interactions between i and j atoms */
875 qq10 = _mm_mul_pd(iq1,jq0);
877 /* COULOMB ELECTROSTATICS */
878 velec = _mm_mul_pd(qq10,rinv10);
879 felec = _mm_mul_pd(velec,rinvsq10);
883 /* Calculate temporary vectorial force */
884 tx = _mm_mul_pd(fscal,dx10);
885 ty = _mm_mul_pd(fscal,dy10);
886 tz = _mm_mul_pd(fscal,dz10);
888 /* Update vectorial force */
889 fix1 = _mm_add_pd(fix1,tx);
890 fiy1 = _mm_add_pd(fiy1,ty);
891 fiz1 = _mm_add_pd(fiz1,tz);
893 fjx0 = _mm_add_pd(fjx0,tx);
894 fjy0 = _mm_add_pd(fjy0,ty);
895 fjz0 = _mm_add_pd(fjz0,tz);
897 /**************************
898 * CALCULATE INTERACTIONS *
899 **************************/
901 /* Compute parameters for interactions between i and j atoms */
902 qq20 = _mm_mul_pd(iq2,jq0);
904 /* COULOMB ELECTROSTATICS */
905 velec = _mm_mul_pd(qq20,rinv20);
906 felec = _mm_mul_pd(velec,rinvsq20);
910 /* Calculate temporary vectorial force */
911 tx = _mm_mul_pd(fscal,dx20);
912 ty = _mm_mul_pd(fscal,dy20);
913 tz = _mm_mul_pd(fscal,dz20);
915 /* Update vectorial force */
916 fix2 = _mm_add_pd(fix2,tx);
917 fiy2 = _mm_add_pd(fiy2,ty);
918 fiz2 = _mm_add_pd(fiz2,tz);
920 fjx0 = _mm_add_pd(fjx0,tx);
921 fjy0 = _mm_add_pd(fjy0,ty);
922 fjz0 = _mm_add_pd(fjz0,tz);
924 /**************************
925 * CALCULATE INTERACTIONS *
926 **************************/
928 /* Compute parameters for interactions between i and j atoms */
929 qq30 = _mm_mul_pd(iq3,jq0);
931 /* COULOMB ELECTROSTATICS */
932 velec = _mm_mul_pd(qq30,rinv30);
933 felec = _mm_mul_pd(velec,rinvsq30);
937 /* Calculate temporary vectorial force */
938 tx = _mm_mul_pd(fscal,dx30);
939 ty = _mm_mul_pd(fscal,dy30);
940 tz = _mm_mul_pd(fscal,dz30);
942 /* Update vectorial force */
943 fix3 = _mm_add_pd(fix3,tx);
944 fiy3 = _mm_add_pd(fiy3,ty);
945 fiz3 = _mm_add_pd(fiz3,tz);
947 fjx0 = _mm_add_pd(fjx0,tx);
948 fjy0 = _mm_add_pd(fjy0,ty);
949 fjz0 = _mm_add_pd(fjz0,tz);
951 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
953 /* Inner loop uses 132 flops */
960 j_coord_offsetA = DIM*jnrA;
962 /* load j atom coordinates */
963 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
966 /* Calculate displacement vector */
967 dx00 = _mm_sub_pd(ix0,jx0);
968 dy00 = _mm_sub_pd(iy0,jy0);
969 dz00 = _mm_sub_pd(iz0,jz0);
970 dx10 = _mm_sub_pd(ix1,jx0);
971 dy10 = _mm_sub_pd(iy1,jy0);
972 dz10 = _mm_sub_pd(iz1,jz0);
973 dx20 = _mm_sub_pd(ix2,jx0);
974 dy20 = _mm_sub_pd(iy2,jy0);
975 dz20 = _mm_sub_pd(iz2,jz0);
976 dx30 = _mm_sub_pd(ix3,jx0);
977 dy30 = _mm_sub_pd(iy3,jy0);
978 dz30 = _mm_sub_pd(iz3,jz0);
980 /* Calculate squared distance and things based on it */
981 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
982 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
983 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
984 rsq30 = gmx_mm_calc_rsq_pd(dx30,dy30,dz30);
986 rinv00 = sse41_invsqrt_d(rsq00);
987 rinv10 = sse41_invsqrt_d(rsq10);
988 rinv20 = sse41_invsqrt_d(rsq20);
989 rinv30 = sse41_invsqrt_d(rsq30);
991 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
992 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
993 rinvsq30 = _mm_mul_pd(rinv30,rinv30);
995 /* Load parameters for j particles */
996 jq0 = _mm_load_sd(charge+jnrA+0);
997 vdwjidx0A = 2*vdwtype[jnrA+0];
999 fjx0 = _mm_setzero_pd();
1000 fjy0 = _mm_setzero_pd();
1001 fjz0 = _mm_setzero_pd();
1003 /**************************
1004 * CALCULATE INTERACTIONS *
1005 **************************/
1007 r00 = _mm_mul_pd(rsq00,rinv00);
1009 /* Compute parameters for interactions between i and j atoms */
1010 gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
1012 /* Calculate table index by multiplying r with table scale and truncate to integer */
1013 rt = _mm_mul_pd(r00,vftabscale);
1014 vfitab = _mm_cvttpd_epi32(rt);
1015 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1016 vfitab = _mm_slli_epi32(vfitab,3);
1018 /* CUBIC SPLINE TABLE DISPERSION */
1019 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1020 F = _mm_setzero_pd();
1021 GMX_MM_TRANSPOSE2_PD(Y,F);
1022 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1023 H = _mm_setzero_pd();
1024 GMX_MM_TRANSPOSE2_PD(G,H);
1025 Heps = _mm_mul_pd(vfeps,H);
1026 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1027 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1028 fvdw6 = _mm_mul_pd(c6_00,FF);
1030 /* CUBIC SPLINE TABLE REPULSION */
1031 vfitab = _mm_add_epi32(vfitab,ifour);
1032 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1033 F = _mm_setzero_pd();
1034 GMX_MM_TRANSPOSE2_PD(Y,F);
1035 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1036 H = _mm_setzero_pd();
1037 GMX_MM_TRANSPOSE2_PD(G,H);
1038 Heps = _mm_mul_pd(vfeps,H);
1039 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1040 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1041 fvdw12 = _mm_mul_pd(c12_00,FF);
1042 fvdw = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
1046 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1048 /* Calculate temporary vectorial force */
1049 tx = _mm_mul_pd(fscal,dx00);
1050 ty = _mm_mul_pd(fscal,dy00);
1051 tz = _mm_mul_pd(fscal,dz00);
1053 /* Update vectorial force */
1054 fix0 = _mm_add_pd(fix0,tx);
1055 fiy0 = _mm_add_pd(fiy0,ty);
1056 fiz0 = _mm_add_pd(fiz0,tz);
1058 fjx0 = _mm_add_pd(fjx0,tx);
1059 fjy0 = _mm_add_pd(fjy0,ty);
1060 fjz0 = _mm_add_pd(fjz0,tz);
1062 /**************************
1063 * CALCULATE INTERACTIONS *
1064 **************************/
1066 /* Compute parameters for interactions between i and j atoms */
1067 qq10 = _mm_mul_pd(iq1,jq0);
1069 /* COULOMB ELECTROSTATICS */
1070 velec = _mm_mul_pd(qq10,rinv10);
1071 felec = _mm_mul_pd(velec,rinvsq10);
1075 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1077 /* Calculate temporary vectorial force */
1078 tx = _mm_mul_pd(fscal,dx10);
1079 ty = _mm_mul_pd(fscal,dy10);
1080 tz = _mm_mul_pd(fscal,dz10);
1082 /* Update vectorial force */
1083 fix1 = _mm_add_pd(fix1,tx);
1084 fiy1 = _mm_add_pd(fiy1,ty);
1085 fiz1 = _mm_add_pd(fiz1,tz);
1087 fjx0 = _mm_add_pd(fjx0,tx);
1088 fjy0 = _mm_add_pd(fjy0,ty);
1089 fjz0 = _mm_add_pd(fjz0,tz);
1091 /**************************
1092 * CALCULATE INTERACTIONS *
1093 **************************/
1095 /* Compute parameters for interactions between i and j atoms */
1096 qq20 = _mm_mul_pd(iq2,jq0);
1098 /* COULOMB ELECTROSTATICS */
1099 velec = _mm_mul_pd(qq20,rinv20);
1100 felec = _mm_mul_pd(velec,rinvsq20);
1104 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1106 /* Calculate temporary vectorial force */
1107 tx = _mm_mul_pd(fscal,dx20);
1108 ty = _mm_mul_pd(fscal,dy20);
1109 tz = _mm_mul_pd(fscal,dz20);
1111 /* Update vectorial force */
1112 fix2 = _mm_add_pd(fix2,tx);
1113 fiy2 = _mm_add_pd(fiy2,ty);
1114 fiz2 = _mm_add_pd(fiz2,tz);
1116 fjx0 = _mm_add_pd(fjx0,tx);
1117 fjy0 = _mm_add_pd(fjy0,ty);
1118 fjz0 = _mm_add_pd(fjz0,tz);
1120 /**************************
1121 * CALCULATE INTERACTIONS *
1122 **************************/
1124 /* Compute parameters for interactions between i and j atoms */
1125 qq30 = _mm_mul_pd(iq3,jq0);
1127 /* COULOMB ELECTROSTATICS */
1128 velec = _mm_mul_pd(qq30,rinv30);
1129 felec = _mm_mul_pd(velec,rinvsq30);
1133 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1135 /* Calculate temporary vectorial force */
1136 tx = _mm_mul_pd(fscal,dx30);
1137 ty = _mm_mul_pd(fscal,dy30);
1138 tz = _mm_mul_pd(fscal,dz30);
1140 /* Update vectorial force */
1141 fix3 = _mm_add_pd(fix3,tx);
1142 fiy3 = _mm_add_pd(fiy3,ty);
1143 fiz3 = _mm_add_pd(fiz3,tz);
1145 fjx0 = _mm_add_pd(fjx0,tx);
1146 fjy0 = _mm_add_pd(fjy0,ty);
1147 fjz0 = _mm_add_pd(fjz0,tz);
1149 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1151 /* Inner loop uses 132 flops */
1154 /* End of innermost loop */
1156 gmx_mm_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1157 f+i_coord_offset,fshift+i_shift_offset);
1159 /* Increment number of inner iterations */
1160 inneriter += j_index_end - j_index_start;
1162 /* Outer loop uses 24 flops */
1165 /* Increment number of outer iterations */
1168 /* Update outer/inner flops */
1170 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_F,outeriter*24 + inneriter*132);