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 sse2_double kernel generator.
42 #include "../nb_kernel.h"
43 #include "types/simple.h"
44 #include "gromacs/math/vec.h"
47 #include "gromacs/simd/math_x86_sse2_double.h"
48 #include "kernelutil_x86_sse2_double.h"
51 * Gromacs nonbonded kernel: nb_kernel_ElecRF_VdwLJ_GeomP1P1_VF_sse2_double
52 * Electrostatics interaction: ReactionField
53 * VdW interaction: LennardJones
54 * Geometry: Particle-Particle
55 * Calculate force/pot: PotentialAndForce
58 nb_kernel_ElecRF_VdwLJ_GeomP1P1_VF_sse2_double
59 (t_nblist * gmx_restrict nlist,
60 rvec * gmx_restrict xx,
61 rvec * gmx_restrict ff,
62 t_forcerec * gmx_restrict fr,
63 t_mdatoms * gmx_restrict mdatoms,
64 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
65 t_nrnb * gmx_restrict nrnb)
67 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
68 * just 0 for non-waters.
69 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
70 * jnr indices corresponding to data put in the four positions in the SIMD register.
72 int i_shift_offset,i_coord_offset,outeriter,inneriter;
73 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
75 int j_coord_offsetA,j_coord_offsetB;
76 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
78 real *shiftvec,*fshift,*x,*f;
79 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
81 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
82 int vdwjidx0A,vdwjidx0B;
83 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
84 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
85 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
88 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
91 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
92 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
93 __m128d dummy_mask,cutoff_mask;
94 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
95 __m128d one = _mm_set1_pd(1.0);
96 __m128d two = _mm_set1_pd(2.0);
102 jindex = nlist->jindex;
104 shiftidx = nlist->shift;
106 shiftvec = fr->shift_vec[0];
107 fshift = fr->fshift[0];
108 facel = _mm_set1_pd(fr->epsfac);
109 charge = mdatoms->chargeA;
110 krf = _mm_set1_pd(fr->ic->k_rf);
111 krf2 = _mm_set1_pd(fr->ic->k_rf*2.0);
112 crf = _mm_set1_pd(fr->ic->c_rf);
113 nvdwtype = fr->ntype;
115 vdwtype = mdatoms->typeA;
117 /* Avoid stupid compiler warnings */
125 /* Start outer loop over neighborlists */
126 for(iidx=0; iidx<nri; iidx++)
128 /* Load shift vector for this list */
129 i_shift_offset = DIM*shiftidx[iidx];
131 /* Load limits for loop over neighbors */
132 j_index_start = jindex[iidx];
133 j_index_end = jindex[iidx+1];
135 /* Get outer coordinate index */
137 i_coord_offset = DIM*inr;
139 /* Load i particle coords and add shift vector */
140 gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
142 fix0 = _mm_setzero_pd();
143 fiy0 = _mm_setzero_pd();
144 fiz0 = _mm_setzero_pd();
146 /* Load parameters for i particles */
147 iq0 = _mm_mul_pd(facel,_mm_load1_pd(charge+inr+0));
148 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
150 /* Reset potential sums */
151 velecsum = _mm_setzero_pd();
152 vvdwsum = _mm_setzero_pd();
154 /* Start inner kernel loop */
155 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
158 /* Get j neighbor index, and coordinate index */
161 j_coord_offsetA = DIM*jnrA;
162 j_coord_offsetB = DIM*jnrB;
164 /* load j atom coordinates */
165 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
168 /* Calculate displacement vector */
169 dx00 = _mm_sub_pd(ix0,jx0);
170 dy00 = _mm_sub_pd(iy0,jy0);
171 dz00 = _mm_sub_pd(iz0,jz0);
173 /* Calculate squared distance and things based on it */
174 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
176 rinv00 = gmx_mm_invsqrt_pd(rsq00);
178 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
180 /* Load parameters for j particles */
181 jq0 = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
182 vdwjidx0A = 2*vdwtype[jnrA+0];
183 vdwjidx0B = 2*vdwtype[jnrB+0];
185 /**************************
186 * CALCULATE INTERACTIONS *
187 **************************/
189 /* Compute parameters for interactions between i and j atoms */
190 qq00 = _mm_mul_pd(iq0,jq0);
191 gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
192 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
194 /* REACTION-FIELD ELECTROSTATICS */
195 velec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_add_pd(rinv00,_mm_mul_pd(krf,rsq00)),crf));
196 felec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
198 /* LENNARD-JONES DISPERSION/REPULSION */
200 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
201 vvdw6 = _mm_mul_pd(c6_00,rinvsix);
202 vvdw12 = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
203 vvdw = _mm_sub_pd( _mm_mul_pd(vvdw12,one_twelfth) , _mm_mul_pd(vvdw6,one_sixth) );
204 fvdw = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
206 /* Update potential sum for this i atom from the interaction with this j atom. */
207 velecsum = _mm_add_pd(velecsum,velec);
208 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
210 fscal = _mm_add_pd(felec,fvdw);
212 /* Calculate temporary vectorial force */
213 tx = _mm_mul_pd(fscal,dx00);
214 ty = _mm_mul_pd(fscal,dy00);
215 tz = _mm_mul_pd(fscal,dz00);
217 /* Update vectorial force */
218 fix0 = _mm_add_pd(fix0,tx);
219 fiy0 = _mm_add_pd(fiy0,ty);
220 fiz0 = _mm_add_pd(fiz0,tz);
222 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,tx,ty,tz);
224 /* Inner loop uses 44 flops */
231 j_coord_offsetA = DIM*jnrA;
233 /* load j atom coordinates */
234 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
237 /* Calculate displacement vector */
238 dx00 = _mm_sub_pd(ix0,jx0);
239 dy00 = _mm_sub_pd(iy0,jy0);
240 dz00 = _mm_sub_pd(iz0,jz0);
242 /* Calculate squared distance and things based on it */
243 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
245 rinv00 = gmx_mm_invsqrt_pd(rsq00);
247 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
249 /* Load parameters for j particles */
250 jq0 = _mm_load_sd(charge+jnrA+0);
251 vdwjidx0A = 2*vdwtype[jnrA+0];
253 /**************************
254 * CALCULATE INTERACTIONS *
255 **************************/
257 /* Compute parameters for interactions between i and j atoms */
258 qq00 = _mm_mul_pd(iq0,jq0);
259 gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
261 /* REACTION-FIELD ELECTROSTATICS */
262 velec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_add_pd(rinv00,_mm_mul_pd(krf,rsq00)),crf));
263 felec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
265 /* LENNARD-JONES DISPERSION/REPULSION */
267 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
268 vvdw6 = _mm_mul_pd(c6_00,rinvsix);
269 vvdw12 = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
270 vvdw = _mm_sub_pd( _mm_mul_pd(vvdw12,one_twelfth) , _mm_mul_pd(vvdw6,one_sixth) );
271 fvdw = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
273 /* Update potential sum for this i atom from the interaction with this j atom. */
274 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
275 velecsum = _mm_add_pd(velecsum,velec);
276 vvdw = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
277 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
279 fscal = _mm_add_pd(felec,fvdw);
281 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
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 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,tx,ty,tz);
295 /* Inner loop uses 44 flops */
298 /* End of innermost loop */
300 gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
301 f+i_coord_offset,fshift+i_shift_offset);
304 /* Update potential energies */
305 gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
306 gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
308 /* Increment number of inner iterations */
309 inneriter += j_index_end - j_index_start;
311 /* Outer loop uses 9 flops */
314 /* Increment number of outer iterations */
317 /* Update outer/inner flops */
319 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_VF,outeriter*9 + inneriter*44);
322 * Gromacs nonbonded kernel: nb_kernel_ElecRF_VdwLJ_GeomP1P1_F_sse2_double
323 * Electrostatics interaction: ReactionField
324 * VdW interaction: LennardJones
325 * Geometry: Particle-Particle
326 * Calculate force/pot: Force
329 nb_kernel_ElecRF_VdwLJ_GeomP1P1_F_sse2_double
330 (t_nblist * gmx_restrict nlist,
331 rvec * gmx_restrict xx,
332 rvec * gmx_restrict ff,
333 t_forcerec * gmx_restrict fr,
334 t_mdatoms * gmx_restrict mdatoms,
335 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
336 t_nrnb * gmx_restrict nrnb)
338 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
339 * just 0 for non-waters.
340 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
341 * jnr indices corresponding to data put in the four positions in the SIMD register.
343 int i_shift_offset,i_coord_offset,outeriter,inneriter;
344 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
346 int j_coord_offsetA,j_coord_offsetB;
347 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
349 real *shiftvec,*fshift,*x,*f;
350 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
352 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
353 int vdwjidx0A,vdwjidx0B;
354 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
355 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
356 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
359 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
362 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
363 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
364 __m128d dummy_mask,cutoff_mask;
365 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
366 __m128d one = _mm_set1_pd(1.0);
367 __m128d two = _mm_set1_pd(2.0);
373 jindex = nlist->jindex;
375 shiftidx = nlist->shift;
377 shiftvec = fr->shift_vec[0];
378 fshift = fr->fshift[0];
379 facel = _mm_set1_pd(fr->epsfac);
380 charge = mdatoms->chargeA;
381 krf = _mm_set1_pd(fr->ic->k_rf);
382 krf2 = _mm_set1_pd(fr->ic->k_rf*2.0);
383 crf = _mm_set1_pd(fr->ic->c_rf);
384 nvdwtype = fr->ntype;
386 vdwtype = mdatoms->typeA;
388 /* Avoid stupid compiler warnings */
396 /* Start outer loop over neighborlists */
397 for(iidx=0; iidx<nri; iidx++)
399 /* Load shift vector for this list */
400 i_shift_offset = DIM*shiftidx[iidx];
402 /* Load limits for loop over neighbors */
403 j_index_start = jindex[iidx];
404 j_index_end = jindex[iidx+1];
406 /* Get outer coordinate index */
408 i_coord_offset = DIM*inr;
410 /* Load i particle coords and add shift vector */
411 gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
413 fix0 = _mm_setzero_pd();
414 fiy0 = _mm_setzero_pd();
415 fiz0 = _mm_setzero_pd();
417 /* Load parameters for i particles */
418 iq0 = _mm_mul_pd(facel,_mm_load1_pd(charge+inr+0));
419 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
421 /* Start inner kernel loop */
422 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
425 /* Get j neighbor index, and coordinate index */
428 j_coord_offsetA = DIM*jnrA;
429 j_coord_offsetB = DIM*jnrB;
431 /* load j atom coordinates */
432 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
435 /* Calculate displacement vector */
436 dx00 = _mm_sub_pd(ix0,jx0);
437 dy00 = _mm_sub_pd(iy0,jy0);
438 dz00 = _mm_sub_pd(iz0,jz0);
440 /* Calculate squared distance and things based on it */
441 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
443 rinv00 = gmx_mm_invsqrt_pd(rsq00);
445 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
447 /* Load parameters for j particles */
448 jq0 = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
449 vdwjidx0A = 2*vdwtype[jnrA+0];
450 vdwjidx0B = 2*vdwtype[jnrB+0];
452 /**************************
453 * CALCULATE INTERACTIONS *
454 **************************/
456 /* Compute parameters for interactions between i and j atoms */
457 qq00 = _mm_mul_pd(iq0,jq0);
458 gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
459 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
461 /* REACTION-FIELD ELECTROSTATICS */
462 felec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
464 /* LENNARD-JONES DISPERSION/REPULSION */
466 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
467 fvdw = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
469 fscal = _mm_add_pd(felec,fvdw);
471 /* Calculate temporary vectorial force */
472 tx = _mm_mul_pd(fscal,dx00);
473 ty = _mm_mul_pd(fscal,dy00);
474 tz = _mm_mul_pd(fscal,dz00);
476 /* Update vectorial force */
477 fix0 = _mm_add_pd(fix0,tx);
478 fiy0 = _mm_add_pd(fiy0,ty);
479 fiz0 = _mm_add_pd(fiz0,tz);
481 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,tx,ty,tz);
483 /* Inner loop uses 34 flops */
490 j_coord_offsetA = DIM*jnrA;
492 /* load j atom coordinates */
493 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
496 /* Calculate displacement vector */
497 dx00 = _mm_sub_pd(ix0,jx0);
498 dy00 = _mm_sub_pd(iy0,jy0);
499 dz00 = _mm_sub_pd(iz0,jz0);
501 /* Calculate squared distance and things based on it */
502 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
504 rinv00 = gmx_mm_invsqrt_pd(rsq00);
506 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
508 /* Load parameters for j particles */
509 jq0 = _mm_load_sd(charge+jnrA+0);
510 vdwjidx0A = 2*vdwtype[jnrA+0];
512 /**************************
513 * CALCULATE INTERACTIONS *
514 **************************/
516 /* Compute parameters for interactions between i and j atoms */
517 qq00 = _mm_mul_pd(iq0,jq0);
518 gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
520 /* REACTION-FIELD ELECTROSTATICS */
521 felec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
523 /* LENNARD-JONES DISPERSION/REPULSION */
525 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
526 fvdw = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
528 fscal = _mm_add_pd(felec,fvdw);
530 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
532 /* Calculate temporary vectorial force */
533 tx = _mm_mul_pd(fscal,dx00);
534 ty = _mm_mul_pd(fscal,dy00);
535 tz = _mm_mul_pd(fscal,dz00);
537 /* Update vectorial force */
538 fix0 = _mm_add_pd(fix0,tx);
539 fiy0 = _mm_add_pd(fiy0,ty);
540 fiz0 = _mm_add_pd(fiz0,tz);
542 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,tx,ty,tz);
544 /* Inner loop uses 34 flops */
547 /* End of innermost loop */
549 gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
550 f+i_coord_offset,fshift+i_shift_offset);
552 /* Increment number of inner iterations */
553 inneriter += j_index_end - j_index_start;
555 /* Outer loop uses 7 flops */
558 /* Increment number of outer iterations */
561 /* Update outer/inner flops */
563 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_F,outeriter*7 + inneriter*34);