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_ElecCoul_VdwLJ_GeomP1P1_VF_sse4_1_double
54 * Electrostatics interaction: Coulomb
55 * VdW interaction: LennardJones
56 * Geometry: Particle-Particle
57 * Calculate force/pot: PotentialAndForce
60 nb_kernel_ElecCoul_VdwLJ_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 nvdwtype = fr->ntype;
114 vdwtype = mdatoms->typeA;
116 /* Avoid stupid compiler warnings */
124 /* Start outer loop over neighborlists */
125 for(iidx=0; iidx<nri; iidx++)
127 /* Load shift vector for this list */
128 i_shift_offset = DIM*shiftidx[iidx];
130 /* Load limits for loop over neighbors */
131 j_index_start = jindex[iidx];
132 j_index_end = jindex[iidx+1];
134 /* Get outer coordinate index */
136 i_coord_offset = DIM*inr;
138 /* Load i particle coords and add shift vector */
139 gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
141 fix0 = _mm_setzero_pd();
142 fiy0 = _mm_setzero_pd();
143 fiz0 = _mm_setzero_pd();
145 /* Load parameters for i particles */
146 iq0 = _mm_mul_pd(facel,_mm_load1_pd(charge+inr+0));
147 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
149 /* Reset potential sums */
150 velecsum = _mm_setzero_pd();
151 vvdwsum = _mm_setzero_pd();
153 /* Start inner kernel loop */
154 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
157 /* Get j neighbor index, and coordinate index */
160 j_coord_offsetA = DIM*jnrA;
161 j_coord_offsetB = DIM*jnrB;
163 /* load j atom coordinates */
164 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
167 /* Calculate displacement vector */
168 dx00 = _mm_sub_pd(ix0,jx0);
169 dy00 = _mm_sub_pd(iy0,jy0);
170 dz00 = _mm_sub_pd(iz0,jz0);
172 /* Calculate squared distance and things based on it */
173 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
175 rinv00 = gmx_mm_invsqrt_pd(rsq00);
177 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
179 /* Load parameters for j particles */
180 jq0 = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
181 vdwjidx0A = 2*vdwtype[jnrA+0];
182 vdwjidx0B = 2*vdwtype[jnrB+0];
184 /**************************
185 * CALCULATE INTERACTIONS *
186 **************************/
188 /* Compute parameters for interactions between i and j atoms */
189 qq00 = _mm_mul_pd(iq0,jq0);
190 gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
191 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
193 /* COULOMB ELECTROSTATICS */
194 velec = _mm_mul_pd(qq00,rinv00);
195 felec = _mm_mul_pd(velec,rinvsq00);
197 /* LENNARD-JONES DISPERSION/REPULSION */
199 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
200 vvdw6 = _mm_mul_pd(c6_00,rinvsix);
201 vvdw12 = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
202 vvdw = _mm_sub_pd( _mm_mul_pd(vvdw12,one_twelfth) , _mm_mul_pd(vvdw6,one_sixth) );
203 fvdw = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
205 /* Update potential sum for this i atom from the interaction with this j atom. */
206 velecsum = _mm_add_pd(velecsum,velec);
207 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
209 fscal = _mm_add_pd(felec,fvdw);
211 /* Calculate temporary vectorial force */
212 tx = _mm_mul_pd(fscal,dx00);
213 ty = _mm_mul_pd(fscal,dy00);
214 tz = _mm_mul_pd(fscal,dz00);
216 /* Update vectorial force */
217 fix0 = _mm_add_pd(fix0,tx);
218 fiy0 = _mm_add_pd(fiy0,ty);
219 fiz0 = _mm_add_pd(fiz0,tz);
221 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,tx,ty,tz);
223 /* Inner loop uses 40 flops */
230 j_coord_offsetA = DIM*jnrA;
232 /* load j atom coordinates */
233 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
236 /* Calculate displacement vector */
237 dx00 = _mm_sub_pd(ix0,jx0);
238 dy00 = _mm_sub_pd(iy0,jy0);
239 dz00 = _mm_sub_pd(iz0,jz0);
241 /* Calculate squared distance and things based on it */
242 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
244 rinv00 = gmx_mm_invsqrt_pd(rsq00);
246 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
248 /* Load parameters for j particles */
249 jq0 = _mm_load_sd(charge+jnrA+0);
250 vdwjidx0A = 2*vdwtype[jnrA+0];
252 /**************************
253 * CALCULATE INTERACTIONS *
254 **************************/
256 /* Compute parameters for interactions between i and j atoms */
257 qq00 = _mm_mul_pd(iq0,jq0);
258 gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
260 /* COULOMB ELECTROSTATICS */
261 velec = _mm_mul_pd(qq00,rinv00);
262 felec = _mm_mul_pd(velec,rinvsq00);
264 /* LENNARD-JONES DISPERSION/REPULSION */
266 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
267 vvdw6 = _mm_mul_pd(c6_00,rinvsix);
268 vvdw12 = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
269 vvdw = _mm_sub_pd( _mm_mul_pd(vvdw12,one_twelfth) , _mm_mul_pd(vvdw6,one_sixth) );
270 fvdw = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
272 /* Update potential sum for this i atom from the interaction with this j atom. */
273 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
274 velecsum = _mm_add_pd(velecsum,velec);
275 vvdw = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
276 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
278 fscal = _mm_add_pd(felec,fvdw);
280 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
282 /* Calculate temporary vectorial force */
283 tx = _mm_mul_pd(fscal,dx00);
284 ty = _mm_mul_pd(fscal,dy00);
285 tz = _mm_mul_pd(fscal,dz00);
287 /* Update vectorial force */
288 fix0 = _mm_add_pd(fix0,tx);
289 fiy0 = _mm_add_pd(fiy0,ty);
290 fiz0 = _mm_add_pd(fiz0,tz);
292 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,tx,ty,tz);
294 /* Inner loop uses 40 flops */
297 /* End of innermost loop */
299 gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
300 f+i_coord_offset,fshift+i_shift_offset);
303 /* Update potential energies */
304 gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
305 gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
307 /* Increment number of inner iterations */
308 inneriter += j_index_end - j_index_start;
310 /* Outer loop uses 9 flops */
313 /* Increment number of outer iterations */
316 /* Update outer/inner flops */
318 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_VF,outeriter*9 + inneriter*40);
321 * Gromacs nonbonded kernel: nb_kernel_ElecCoul_VdwLJ_GeomP1P1_F_sse4_1_double
322 * Electrostatics interaction: Coulomb
323 * VdW interaction: LennardJones
324 * Geometry: Particle-Particle
325 * Calculate force/pot: Force
328 nb_kernel_ElecCoul_VdwLJ_GeomP1P1_F_sse4_1_double
329 (t_nblist * gmx_restrict nlist,
330 rvec * gmx_restrict xx,
331 rvec * gmx_restrict ff,
332 t_forcerec * gmx_restrict fr,
333 t_mdatoms * gmx_restrict mdatoms,
334 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
335 t_nrnb * gmx_restrict nrnb)
337 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
338 * just 0 for non-waters.
339 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
340 * jnr indices corresponding to data put in the four positions in the SIMD register.
342 int i_shift_offset,i_coord_offset,outeriter,inneriter;
343 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
345 int j_coord_offsetA,j_coord_offsetB;
346 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
348 real *shiftvec,*fshift,*x,*f;
349 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
351 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
352 int vdwjidx0A,vdwjidx0B;
353 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
354 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
355 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
358 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
361 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
362 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
363 __m128d dummy_mask,cutoff_mask;
364 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
365 __m128d one = _mm_set1_pd(1.0);
366 __m128d two = _mm_set1_pd(2.0);
372 jindex = nlist->jindex;
374 shiftidx = nlist->shift;
376 shiftvec = fr->shift_vec[0];
377 fshift = fr->fshift[0];
378 facel = _mm_set1_pd(fr->epsfac);
379 charge = mdatoms->chargeA;
380 nvdwtype = fr->ntype;
382 vdwtype = mdatoms->typeA;
384 /* Avoid stupid compiler warnings */
392 /* Start outer loop over neighborlists */
393 for(iidx=0; iidx<nri; iidx++)
395 /* Load shift vector for this list */
396 i_shift_offset = DIM*shiftidx[iidx];
398 /* Load limits for loop over neighbors */
399 j_index_start = jindex[iidx];
400 j_index_end = jindex[iidx+1];
402 /* Get outer coordinate index */
404 i_coord_offset = DIM*inr;
406 /* Load i particle coords and add shift vector */
407 gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
409 fix0 = _mm_setzero_pd();
410 fiy0 = _mm_setzero_pd();
411 fiz0 = _mm_setzero_pd();
413 /* Load parameters for i particles */
414 iq0 = _mm_mul_pd(facel,_mm_load1_pd(charge+inr+0));
415 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
417 /* Start inner kernel loop */
418 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
421 /* Get j neighbor index, and coordinate index */
424 j_coord_offsetA = DIM*jnrA;
425 j_coord_offsetB = DIM*jnrB;
427 /* load j atom coordinates */
428 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
431 /* Calculate displacement vector */
432 dx00 = _mm_sub_pd(ix0,jx0);
433 dy00 = _mm_sub_pd(iy0,jy0);
434 dz00 = _mm_sub_pd(iz0,jz0);
436 /* Calculate squared distance and things based on it */
437 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
439 rinv00 = gmx_mm_invsqrt_pd(rsq00);
441 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
443 /* Load parameters for j particles */
444 jq0 = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
445 vdwjidx0A = 2*vdwtype[jnrA+0];
446 vdwjidx0B = 2*vdwtype[jnrB+0];
448 /**************************
449 * CALCULATE INTERACTIONS *
450 **************************/
452 /* Compute parameters for interactions between i and j atoms */
453 qq00 = _mm_mul_pd(iq0,jq0);
454 gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
455 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
457 /* COULOMB ELECTROSTATICS */
458 velec = _mm_mul_pd(qq00,rinv00);
459 felec = _mm_mul_pd(velec,rinvsq00);
461 /* LENNARD-JONES DISPERSION/REPULSION */
463 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
464 fvdw = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
466 fscal = _mm_add_pd(felec,fvdw);
468 /* Calculate temporary vectorial force */
469 tx = _mm_mul_pd(fscal,dx00);
470 ty = _mm_mul_pd(fscal,dy00);
471 tz = _mm_mul_pd(fscal,dz00);
473 /* Update vectorial force */
474 fix0 = _mm_add_pd(fix0,tx);
475 fiy0 = _mm_add_pd(fiy0,ty);
476 fiz0 = _mm_add_pd(fiz0,tz);
478 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,tx,ty,tz);
480 /* Inner loop uses 34 flops */
487 j_coord_offsetA = DIM*jnrA;
489 /* load j atom coordinates */
490 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
493 /* Calculate displacement vector */
494 dx00 = _mm_sub_pd(ix0,jx0);
495 dy00 = _mm_sub_pd(iy0,jy0);
496 dz00 = _mm_sub_pd(iz0,jz0);
498 /* Calculate squared distance and things based on it */
499 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
501 rinv00 = gmx_mm_invsqrt_pd(rsq00);
503 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
505 /* Load parameters for j particles */
506 jq0 = _mm_load_sd(charge+jnrA+0);
507 vdwjidx0A = 2*vdwtype[jnrA+0];
509 /**************************
510 * CALCULATE INTERACTIONS *
511 **************************/
513 /* Compute parameters for interactions between i and j atoms */
514 qq00 = _mm_mul_pd(iq0,jq0);
515 gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
517 /* COULOMB ELECTROSTATICS */
518 velec = _mm_mul_pd(qq00,rinv00);
519 felec = _mm_mul_pd(velec,rinvsq00);
521 /* LENNARD-JONES DISPERSION/REPULSION */
523 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
524 fvdw = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
526 fscal = _mm_add_pd(felec,fvdw);
528 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
530 /* Calculate temporary vectorial force */
531 tx = _mm_mul_pd(fscal,dx00);
532 ty = _mm_mul_pd(fscal,dy00);
533 tz = _mm_mul_pd(fscal,dz00);
535 /* Update vectorial force */
536 fix0 = _mm_add_pd(fix0,tx);
537 fiy0 = _mm_add_pd(fiy0,ty);
538 fiz0 = _mm_add_pd(fiz0,tz);
540 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,tx,ty,tz);
542 /* Inner loop uses 34 flops */
545 /* End of innermost loop */
547 gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
548 f+i_coord_offset,fshift+i_shift_offset);
550 /* Increment number of inner iterations */
551 inneriter += j_index_end - j_index_start;
553 /* Outer loop uses 7 flops */
556 /* Increment number of outer iterations */
559 /* Update outer/inner flops */
561 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_F,outeriter*7 + inneriter*34);