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_ElecRF_VdwLJ_GeomW3P1_VF_sse4_1_double
54 * Electrostatics interaction: ReactionField
55 * VdW interaction: LennardJones
56 * Geometry: Water3-Particle
57 * Calculate force/pot: PotentialAndForce
60 nb_kernel_ElecRF_VdwLJ_GeomW3P1_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;
85 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
87 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
88 int vdwjidx0A,vdwjidx0B;
89 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
90 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
91 __m128d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
92 __m128d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
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);
101 __m128d dummy_mask,cutoff_mask;
102 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
103 __m128d one = _mm_set1_pd(1.0);
104 __m128d two = _mm_set1_pd(2.0);
110 jindex = nlist->jindex;
112 shiftidx = nlist->shift;
114 shiftvec = fr->shift_vec[0];
115 fshift = fr->fshift[0];
116 facel = _mm_set1_pd(fr->epsfac);
117 charge = mdatoms->chargeA;
118 krf = _mm_set1_pd(fr->ic->k_rf);
119 krf2 = _mm_set1_pd(fr->ic->k_rf*2.0);
120 crf = _mm_set1_pd(fr->ic->c_rf);
121 nvdwtype = fr->ntype;
123 vdwtype = mdatoms->typeA;
125 /* Setup water-specific parameters */
126 inr = nlist->iinr[0];
127 iq0 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
128 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
129 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
130 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
132 /* Avoid stupid compiler warnings */
140 /* Start outer loop over neighborlists */
141 for(iidx=0; iidx<nri; iidx++)
143 /* Load shift vector for this list */
144 i_shift_offset = DIM*shiftidx[iidx];
146 /* Load limits for loop over neighbors */
147 j_index_start = jindex[iidx];
148 j_index_end = jindex[iidx+1];
150 /* Get outer coordinate index */
152 i_coord_offset = DIM*inr;
154 /* Load i particle coords and add shift vector */
155 gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
156 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
158 fix0 = _mm_setzero_pd();
159 fiy0 = _mm_setzero_pd();
160 fiz0 = _mm_setzero_pd();
161 fix1 = _mm_setzero_pd();
162 fiy1 = _mm_setzero_pd();
163 fiz1 = _mm_setzero_pd();
164 fix2 = _mm_setzero_pd();
165 fiy2 = _mm_setzero_pd();
166 fiz2 = _mm_setzero_pd();
168 /* Reset potential sums */
169 velecsum = _mm_setzero_pd();
170 vvdwsum = _mm_setzero_pd();
172 /* Start inner kernel loop */
173 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
176 /* Get j neighbor index, and coordinate index */
179 j_coord_offsetA = DIM*jnrA;
180 j_coord_offsetB = DIM*jnrB;
182 /* load j atom coordinates */
183 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
186 /* Calculate displacement vector */
187 dx00 = _mm_sub_pd(ix0,jx0);
188 dy00 = _mm_sub_pd(iy0,jy0);
189 dz00 = _mm_sub_pd(iz0,jz0);
190 dx10 = _mm_sub_pd(ix1,jx0);
191 dy10 = _mm_sub_pd(iy1,jy0);
192 dz10 = _mm_sub_pd(iz1,jz0);
193 dx20 = _mm_sub_pd(ix2,jx0);
194 dy20 = _mm_sub_pd(iy2,jy0);
195 dz20 = _mm_sub_pd(iz2,jz0);
197 /* Calculate squared distance and things based on it */
198 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
199 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
200 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
202 rinv00 = gmx_mm_invsqrt_pd(rsq00);
203 rinv10 = gmx_mm_invsqrt_pd(rsq10);
204 rinv20 = gmx_mm_invsqrt_pd(rsq20);
206 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
207 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
208 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
210 /* Load parameters for j particles */
211 jq0 = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
212 vdwjidx0A = 2*vdwtype[jnrA+0];
213 vdwjidx0B = 2*vdwtype[jnrB+0];
215 fjx0 = _mm_setzero_pd();
216 fjy0 = _mm_setzero_pd();
217 fjz0 = _mm_setzero_pd();
219 /**************************
220 * CALCULATE INTERACTIONS *
221 **************************/
223 /* Compute parameters for interactions between i and j atoms */
224 qq00 = _mm_mul_pd(iq0,jq0);
225 gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
226 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
228 /* REACTION-FIELD ELECTROSTATICS */
229 velec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_add_pd(rinv00,_mm_mul_pd(krf,rsq00)),crf));
230 felec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
232 /* LENNARD-JONES DISPERSION/REPULSION */
234 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
235 vvdw6 = _mm_mul_pd(c6_00,rinvsix);
236 vvdw12 = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
237 vvdw = _mm_sub_pd( _mm_mul_pd(vvdw12,one_twelfth) , _mm_mul_pd(vvdw6,one_sixth) );
238 fvdw = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
240 /* Update potential sum for this i atom from the interaction with this j atom. */
241 velecsum = _mm_add_pd(velecsum,velec);
242 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
244 fscal = _mm_add_pd(felec,fvdw);
246 /* Calculate temporary vectorial force */
247 tx = _mm_mul_pd(fscal,dx00);
248 ty = _mm_mul_pd(fscal,dy00);
249 tz = _mm_mul_pd(fscal,dz00);
251 /* Update vectorial force */
252 fix0 = _mm_add_pd(fix0,tx);
253 fiy0 = _mm_add_pd(fiy0,ty);
254 fiz0 = _mm_add_pd(fiz0,tz);
256 fjx0 = _mm_add_pd(fjx0,tx);
257 fjy0 = _mm_add_pd(fjy0,ty);
258 fjz0 = _mm_add_pd(fjz0,tz);
260 /**************************
261 * CALCULATE INTERACTIONS *
262 **************************/
264 /* Compute parameters for interactions between i and j atoms */
265 qq10 = _mm_mul_pd(iq1,jq0);
267 /* REACTION-FIELD ELECTROSTATICS */
268 velec = _mm_mul_pd(qq10,_mm_sub_pd(_mm_add_pd(rinv10,_mm_mul_pd(krf,rsq10)),crf));
269 felec = _mm_mul_pd(qq10,_mm_sub_pd(_mm_mul_pd(rinv10,rinvsq10),krf2));
271 /* Update potential sum for this i atom from the interaction with this j atom. */
272 velecsum = _mm_add_pd(velecsum,velec);
276 /* Calculate temporary vectorial force */
277 tx = _mm_mul_pd(fscal,dx10);
278 ty = _mm_mul_pd(fscal,dy10);
279 tz = _mm_mul_pd(fscal,dz10);
281 /* Update vectorial force */
282 fix1 = _mm_add_pd(fix1,tx);
283 fiy1 = _mm_add_pd(fiy1,ty);
284 fiz1 = _mm_add_pd(fiz1,tz);
286 fjx0 = _mm_add_pd(fjx0,tx);
287 fjy0 = _mm_add_pd(fjy0,ty);
288 fjz0 = _mm_add_pd(fjz0,tz);
290 /**************************
291 * CALCULATE INTERACTIONS *
292 **************************/
294 /* Compute parameters for interactions between i and j atoms */
295 qq20 = _mm_mul_pd(iq2,jq0);
297 /* REACTION-FIELD ELECTROSTATICS */
298 velec = _mm_mul_pd(qq20,_mm_sub_pd(_mm_add_pd(rinv20,_mm_mul_pd(krf,rsq20)),crf));
299 felec = _mm_mul_pd(qq20,_mm_sub_pd(_mm_mul_pd(rinv20,rinvsq20),krf2));
301 /* Update potential sum for this i atom from the interaction with this j atom. */
302 velecsum = _mm_add_pd(velecsum,velec);
306 /* Calculate temporary vectorial force */
307 tx = _mm_mul_pd(fscal,dx20);
308 ty = _mm_mul_pd(fscal,dy20);
309 tz = _mm_mul_pd(fscal,dz20);
311 /* Update vectorial force */
312 fix2 = _mm_add_pd(fix2,tx);
313 fiy2 = _mm_add_pd(fiy2,ty);
314 fiz2 = _mm_add_pd(fiz2,tz);
316 fjx0 = _mm_add_pd(fjx0,tx);
317 fjy0 = _mm_add_pd(fjy0,ty);
318 fjz0 = _mm_add_pd(fjz0,tz);
320 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
322 /* Inner loop uses 111 flops */
329 j_coord_offsetA = DIM*jnrA;
331 /* load j atom coordinates */
332 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
335 /* Calculate displacement vector */
336 dx00 = _mm_sub_pd(ix0,jx0);
337 dy00 = _mm_sub_pd(iy0,jy0);
338 dz00 = _mm_sub_pd(iz0,jz0);
339 dx10 = _mm_sub_pd(ix1,jx0);
340 dy10 = _mm_sub_pd(iy1,jy0);
341 dz10 = _mm_sub_pd(iz1,jz0);
342 dx20 = _mm_sub_pd(ix2,jx0);
343 dy20 = _mm_sub_pd(iy2,jy0);
344 dz20 = _mm_sub_pd(iz2,jz0);
346 /* Calculate squared distance and things based on it */
347 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
348 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
349 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
351 rinv00 = gmx_mm_invsqrt_pd(rsq00);
352 rinv10 = gmx_mm_invsqrt_pd(rsq10);
353 rinv20 = gmx_mm_invsqrt_pd(rsq20);
355 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
356 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
357 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
359 /* Load parameters for j particles */
360 jq0 = _mm_load_sd(charge+jnrA+0);
361 vdwjidx0A = 2*vdwtype[jnrA+0];
363 fjx0 = _mm_setzero_pd();
364 fjy0 = _mm_setzero_pd();
365 fjz0 = _mm_setzero_pd();
367 /**************************
368 * CALCULATE INTERACTIONS *
369 **************************/
371 /* Compute parameters for interactions between i and j atoms */
372 qq00 = _mm_mul_pd(iq0,jq0);
373 gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
375 /* REACTION-FIELD ELECTROSTATICS */
376 velec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_add_pd(rinv00,_mm_mul_pd(krf,rsq00)),crf));
377 felec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
379 /* LENNARD-JONES DISPERSION/REPULSION */
381 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
382 vvdw6 = _mm_mul_pd(c6_00,rinvsix);
383 vvdw12 = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
384 vvdw = _mm_sub_pd( _mm_mul_pd(vvdw12,one_twelfth) , _mm_mul_pd(vvdw6,one_sixth) );
385 fvdw = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
387 /* Update potential sum for this i atom from the interaction with this j atom. */
388 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
389 velecsum = _mm_add_pd(velecsum,velec);
390 vvdw = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
391 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
393 fscal = _mm_add_pd(felec,fvdw);
395 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
397 /* Calculate temporary vectorial force */
398 tx = _mm_mul_pd(fscal,dx00);
399 ty = _mm_mul_pd(fscal,dy00);
400 tz = _mm_mul_pd(fscal,dz00);
402 /* Update vectorial force */
403 fix0 = _mm_add_pd(fix0,tx);
404 fiy0 = _mm_add_pd(fiy0,ty);
405 fiz0 = _mm_add_pd(fiz0,tz);
407 fjx0 = _mm_add_pd(fjx0,tx);
408 fjy0 = _mm_add_pd(fjy0,ty);
409 fjz0 = _mm_add_pd(fjz0,tz);
411 /**************************
412 * CALCULATE INTERACTIONS *
413 **************************/
415 /* Compute parameters for interactions between i and j atoms */
416 qq10 = _mm_mul_pd(iq1,jq0);
418 /* REACTION-FIELD ELECTROSTATICS */
419 velec = _mm_mul_pd(qq10,_mm_sub_pd(_mm_add_pd(rinv10,_mm_mul_pd(krf,rsq10)),crf));
420 felec = _mm_mul_pd(qq10,_mm_sub_pd(_mm_mul_pd(rinv10,rinvsq10),krf2));
422 /* Update potential sum for this i atom from the interaction with this j atom. */
423 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
424 velecsum = _mm_add_pd(velecsum,velec);
428 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
430 /* Calculate temporary vectorial force */
431 tx = _mm_mul_pd(fscal,dx10);
432 ty = _mm_mul_pd(fscal,dy10);
433 tz = _mm_mul_pd(fscal,dz10);
435 /* Update vectorial force */
436 fix1 = _mm_add_pd(fix1,tx);
437 fiy1 = _mm_add_pd(fiy1,ty);
438 fiz1 = _mm_add_pd(fiz1,tz);
440 fjx0 = _mm_add_pd(fjx0,tx);
441 fjy0 = _mm_add_pd(fjy0,ty);
442 fjz0 = _mm_add_pd(fjz0,tz);
444 /**************************
445 * CALCULATE INTERACTIONS *
446 **************************/
448 /* Compute parameters for interactions between i and j atoms */
449 qq20 = _mm_mul_pd(iq2,jq0);
451 /* REACTION-FIELD ELECTROSTATICS */
452 velec = _mm_mul_pd(qq20,_mm_sub_pd(_mm_add_pd(rinv20,_mm_mul_pd(krf,rsq20)),crf));
453 felec = _mm_mul_pd(qq20,_mm_sub_pd(_mm_mul_pd(rinv20,rinvsq20),krf2));
455 /* Update potential sum for this i atom from the interaction with this j atom. */
456 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
457 velecsum = _mm_add_pd(velecsum,velec);
461 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
463 /* Calculate temporary vectorial force */
464 tx = _mm_mul_pd(fscal,dx20);
465 ty = _mm_mul_pd(fscal,dy20);
466 tz = _mm_mul_pd(fscal,dz20);
468 /* Update vectorial force */
469 fix2 = _mm_add_pd(fix2,tx);
470 fiy2 = _mm_add_pd(fiy2,ty);
471 fiz2 = _mm_add_pd(fiz2,tz);
473 fjx0 = _mm_add_pd(fjx0,tx);
474 fjy0 = _mm_add_pd(fjy0,ty);
475 fjz0 = _mm_add_pd(fjz0,tz);
477 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0);
479 /* Inner loop uses 111 flops */
482 /* End of innermost loop */
484 gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
485 f+i_coord_offset,fshift+i_shift_offset);
488 /* Update potential energies */
489 gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
490 gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
492 /* Increment number of inner iterations */
493 inneriter += j_index_end - j_index_start;
495 /* Outer loop uses 20 flops */
498 /* Increment number of outer iterations */
501 /* Update outer/inner flops */
503 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*111);
506 * Gromacs nonbonded kernel: nb_kernel_ElecRF_VdwLJ_GeomW3P1_F_sse4_1_double
507 * Electrostatics interaction: ReactionField
508 * VdW interaction: LennardJones
509 * Geometry: Water3-Particle
510 * Calculate force/pot: Force
513 nb_kernel_ElecRF_VdwLJ_GeomW3P1_F_sse4_1_double
514 (t_nblist * gmx_restrict nlist,
515 rvec * gmx_restrict xx,
516 rvec * gmx_restrict ff,
517 t_forcerec * gmx_restrict fr,
518 t_mdatoms * gmx_restrict mdatoms,
519 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
520 t_nrnb * gmx_restrict nrnb)
522 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
523 * just 0 for non-waters.
524 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
525 * jnr indices corresponding to data put in the four positions in the SIMD register.
527 int i_shift_offset,i_coord_offset,outeriter,inneriter;
528 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
530 int j_coord_offsetA,j_coord_offsetB;
531 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
533 real *shiftvec,*fshift,*x,*f;
534 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
536 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
538 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
540 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
541 int vdwjidx0A,vdwjidx0B;
542 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
543 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
544 __m128d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
545 __m128d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
546 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
549 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
552 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
553 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
554 __m128d dummy_mask,cutoff_mask;
555 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
556 __m128d one = _mm_set1_pd(1.0);
557 __m128d two = _mm_set1_pd(2.0);
563 jindex = nlist->jindex;
565 shiftidx = nlist->shift;
567 shiftvec = fr->shift_vec[0];
568 fshift = fr->fshift[0];
569 facel = _mm_set1_pd(fr->epsfac);
570 charge = mdatoms->chargeA;
571 krf = _mm_set1_pd(fr->ic->k_rf);
572 krf2 = _mm_set1_pd(fr->ic->k_rf*2.0);
573 crf = _mm_set1_pd(fr->ic->c_rf);
574 nvdwtype = fr->ntype;
576 vdwtype = mdatoms->typeA;
578 /* Setup water-specific parameters */
579 inr = nlist->iinr[0];
580 iq0 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
581 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
582 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
583 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
585 /* Avoid stupid compiler warnings */
593 /* Start outer loop over neighborlists */
594 for(iidx=0; iidx<nri; iidx++)
596 /* Load shift vector for this list */
597 i_shift_offset = DIM*shiftidx[iidx];
599 /* Load limits for loop over neighbors */
600 j_index_start = jindex[iidx];
601 j_index_end = jindex[iidx+1];
603 /* Get outer coordinate index */
605 i_coord_offset = DIM*inr;
607 /* Load i particle coords and add shift vector */
608 gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
609 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
611 fix0 = _mm_setzero_pd();
612 fiy0 = _mm_setzero_pd();
613 fiz0 = _mm_setzero_pd();
614 fix1 = _mm_setzero_pd();
615 fiy1 = _mm_setzero_pd();
616 fiz1 = _mm_setzero_pd();
617 fix2 = _mm_setzero_pd();
618 fiy2 = _mm_setzero_pd();
619 fiz2 = _mm_setzero_pd();
621 /* Start inner kernel loop */
622 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
625 /* Get j neighbor index, and coordinate index */
628 j_coord_offsetA = DIM*jnrA;
629 j_coord_offsetB = DIM*jnrB;
631 /* load j atom coordinates */
632 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
635 /* Calculate displacement vector */
636 dx00 = _mm_sub_pd(ix0,jx0);
637 dy00 = _mm_sub_pd(iy0,jy0);
638 dz00 = _mm_sub_pd(iz0,jz0);
639 dx10 = _mm_sub_pd(ix1,jx0);
640 dy10 = _mm_sub_pd(iy1,jy0);
641 dz10 = _mm_sub_pd(iz1,jz0);
642 dx20 = _mm_sub_pd(ix2,jx0);
643 dy20 = _mm_sub_pd(iy2,jy0);
644 dz20 = _mm_sub_pd(iz2,jz0);
646 /* Calculate squared distance and things based on it */
647 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
648 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
649 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
651 rinv00 = gmx_mm_invsqrt_pd(rsq00);
652 rinv10 = gmx_mm_invsqrt_pd(rsq10);
653 rinv20 = gmx_mm_invsqrt_pd(rsq20);
655 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
656 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
657 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
659 /* Load parameters for j particles */
660 jq0 = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
661 vdwjidx0A = 2*vdwtype[jnrA+0];
662 vdwjidx0B = 2*vdwtype[jnrB+0];
664 fjx0 = _mm_setzero_pd();
665 fjy0 = _mm_setzero_pd();
666 fjz0 = _mm_setzero_pd();
668 /**************************
669 * CALCULATE INTERACTIONS *
670 **************************/
672 /* Compute parameters for interactions between i and j atoms */
673 qq00 = _mm_mul_pd(iq0,jq0);
674 gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
675 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
677 /* REACTION-FIELD ELECTROSTATICS */
678 felec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
680 /* LENNARD-JONES DISPERSION/REPULSION */
682 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
683 fvdw = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
685 fscal = _mm_add_pd(felec,fvdw);
687 /* Calculate temporary vectorial force */
688 tx = _mm_mul_pd(fscal,dx00);
689 ty = _mm_mul_pd(fscal,dy00);
690 tz = _mm_mul_pd(fscal,dz00);
692 /* Update vectorial force */
693 fix0 = _mm_add_pd(fix0,tx);
694 fiy0 = _mm_add_pd(fiy0,ty);
695 fiz0 = _mm_add_pd(fiz0,tz);
697 fjx0 = _mm_add_pd(fjx0,tx);
698 fjy0 = _mm_add_pd(fjy0,ty);
699 fjz0 = _mm_add_pd(fjz0,tz);
701 /**************************
702 * CALCULATE INTERACTIONS *
703 **************************/
705 /* Compute parameters for interactions between i and j atoms */
706 qq10 = _mm_mul_pd(iq1,jq0);
708 /* REACTION-FIELD ELECTROSTATICS */
709 felec = _mm_mul_pd(qq10,_mm_sub_pd(_mm_mul_pd(rinv10,rinvsq10),krf2));
713 /* Calculate temporary vectorial force */
714 tx = _mm_mul_pd(fscal,dx10);
715 ty = _mm_mul_pd(fscal,dy10);
716 tz = _mm_mul_pd(fscal,dz10);
718 /* Update vectorial force */
719 fix1 = _mm_add_pd(fix1,tx);
720 fiy1 = _mm_add_pd(fiy1,ty);
721 fiz1 = _mm_add_pd(fiz1,tz);
723 fjx0 = _mm_add_pd(fjx0,tx);
724 fjy0 = _mm_add_pd(fjy0,ty);
725 fjz0 = _mm_add_pd(fjz0,tz);
727 /**************************
728 * CALCULATE INTERACTIONS *
729 **************************/
731 /* Compute parameters for interactions between i and j atoms */
732 qq20 = _mm_mul_pd(iq2,jq0);
734 /* REACTION-FIELD ELECTROSTATICS */
735 felec = _mm_mul_pd(qq20,_mm_sub_pd(_mm_mul_pd(rinv20,rinvsq20),krf2));
739 /* Calculate temporary vectorial force */
740 tx = _mm_mul_pd(fscal,dx20);
741 ty = _mm_mul_pd(fscal,dy20);
742 tz = _mm_mul_pd(fscal,dz20);
744 /* Update vectorial force */
745 fix2 = _mm_add_pd(fix2,tx);
746 fiy2 = _mm_add_pd(fiy2,ty);
747 fiz2 = _mm_add_pd(fiz2,tz);
749 fjx0 = _mm_add_pd(fjx0,tx);
750 fjy0 = _mm_add_pd(fjy0,ty);
751 fjz0 = _mm_add_pd(fjz0,tz);
753 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
755 /* Inner loop uses 91 flops */
762 j_coord_offsetA = DIM*jnrA;
764 /* load j atom coordinates */
765 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
768 /* Calculate displacement vector */
769 dx00 = _mm_sub_pd(ix0,jx0);
770 dy00 = _mm_sub_pd(iy0,jy0);
771 dz00 = _mm_sub_pd(iz0,jz0);
772 dx10 = _mm_sub_pd(ix1,jx0);
773 dy10 = _mm_sub_pd(iy1,jy0);
774 dz10 = _mm_sub_pd(iz1,jz0);
775 dx20 = _mm_sub_pd(ix2,jx0);
776 dy20 = _mm_sub_pd(iy2,jy0);
777 dz20 = _mm_sub_pd(iz2,jz0);
779 /* Calculate squared distance and things based on it */
780 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
781 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
782 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
784 rinv00 = gmx_mm_invsqrt_pd(rsq00);
785 rinv10 = gmx_mm_invsqrt_pd(rsq10);
786 rinv20 = gmx_mm_invsqrt_pd(rsq20);
788 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
789 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
790 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
792 /* Load parameters for j particles */
793 jq0 = _mm_load_sd(charge+jnrA+0);
794 vdwjidx0A = 2*vdwtype[jnrA+0];
796 fjx0 = _mm_setzero_pd();
797 fjy0 = _mm_setzero_pd();
798 fjz0 = _mm_setzero_pd();
800 /**************************
801 * CALCULATE INTERACTIONS *
802 **************************/
804 /* Compute parameters for interactions between i and j atoms */
805 qq00 = _mm_mul_pd(iq0,jq0);
806 gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
808 /* REACTION-FIELD ELECTROSTATICS */
809 felec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
811 /* LENNARD-JONES DISPERSION/REPULSION */
813 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
814 fvdw = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
816 fscal = _mm_add_pd(felec,fvdw);
818 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
820 /* Calculate temporary vectorial force */
821 tx = _mm_mul_pd(fscal,dx00);
822 ty = _mm_mul_pd(fscal,dy00);
823 tz = _mm_mul_pd(fscal,dz00);
825 /* Update vectorial force */
826 fix0 = _mm_add_pd(fix0,tx);
827 fiy0 = _mm_add_pd(fiy0,ty);
828 fiz0 = _mm_add_pd(fiz0,tz);
830 fjx0 = _mm_add_pd(fjx0,tx);
831 fjy0 = _mm_add_pd(fjy0,ty);
832 fjz0 = _mm_add_pd(fjz0,tz);
834 /**************************
835 * CALCULATE INTERACTIONS *
836 **************************/
838 /* Compute parameters for interactions between i and j atoms */
839 qq10 = _mm_mul_pd(iq1,jq0);
841 /* REACTION-FIELD ELECTROSTATICS */
842 felec = _mm_mul_pd(qq10,_mm_sub_pd(_mm_mul_pd(rinv10,rinvsq10),krf2));
846 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
848 /* Calculate temporary vectorial force */
849 tx = _mm_mul_pd(fscal,dx10);
850 ty = _mm_mul_pd(fscal,dy10);
851 tz = _mm_mul_pd(fscal,dz10);
853 /* Update vectorial force */
854 fix1 = _mm_add_pd(fix1,tx);
855 fiy1 = _mm_add_pd(fiy1,ty);
856 fiz1 = _mm_add_pd(fiz1,tz);
858 fjx0 = _mm_add_pd(fjx0,tx);
859 fjy0 = _mm_add_pd(fjy0,ty);
860 fjz0 = _mm_add_pd(fjz0,tz);
862 /**************************
863 * CALCULATE INTERACTIONS *
864 **************************/
866 /* Compute parameters for interactions between i and j atoms */
867 qq20 = _mm_mul_pd(iq2,jq0);
869 /* REACTION-FIELD ELECTROSTATICS */
870 felec = _mm_mul_pd(qq20,_mm_sub_pd(_mm_mul_pd(rinv20,rinvsq20),krf2));
874 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
876 /* Calculate temporary vectorial force */
877 tx = _mm_mul_pd(fscal,dx20);
878 ty = _mm_mul_pd(fscal,dy20);
879 tz = _mm_mul_pd(fscal,dz20);
881 /* Update vectorial force */
882 fix2 = _mm_add_pd(fix2,tx);
883 fiy2 = _mm_add_pd(fiy2,ty);
884 fiz2 = _mm_add_pd(fiz2,tz);
886 fjx0 = _mm_add_pd(fjx0,tx);
887 fjy0 = _mm_add_pd(fjy0,ty);
888 fjz0 = _mm_add_pd(fjz0,tz);
890 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0);
892 /* Inner loop uses 91 flops */
895 /* End of innermost loop */
897 gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
898 f+i_coord_offset,fshift+i_shift_offset);
900 /* Increment number of inner iterations */
901 inneriter += j_index_end - j_index_start;
903 /* Outer loop uses 18 flops */
906 /* Increment number of outer iterations */
909 /* Update outer/inner flops */
911 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_F,outeriter*18 + inneriter*91);