2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2012,2013,2014,2015,2017,2018, 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_ElecEwSh_VdwLJSh_GeomP1P1_VF_sse4_1_double
51 * Electrostatics interaction: Ewald
52 * VdW interaction: LennardJones
53 * Geometry: Particle-Particle
54 * Calculate force/pot: PotentialAndForce
57 nb_kernel_ElecEwSh_VdwLJSh_GeomP1P1_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;
81 int vdwjidx0A,vdwjidx0B;
82 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
83 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
84 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
87 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
90 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
91 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
93 __m128d ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
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->ic->epsfac);
111 charge = mdatoms->chargeA;
112 nvdwtype = fr->ntype;
114 vdwtype = mdatoms->typeA;
116 sh_ewald = _mm_set1_pd(fr->ic->sh_ewald);
117 ewtab = fr->ic->tabq_coul_FDV0;
118 ewtabscale = _mm_set1_pd(fr->ic->tabq_scale);
119 ewtabhalfspace = _mm_set1_pd(0.5/fr->ic->tabq_scale);
121 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
122 rcutoff_scalar = fr->ic->rcoulomb;
123 rcutoff = _mm_set1_pd(rcutoff_scalar);
124 rcutoff2 = _mm_mul_pd(rcutoff,rcutoff);
126 sh_vdw_invrcut6 = _mm_set1_pd(fr->ic->sh_invrc6);
127 rvdw = _mm_set1_pd(fr->ic->rvdw);
129 /* Avoid stupid compiler warnings */
137 /* Start outer loop over neighborlists */
138 for(iidx=0; iidx<nri; iidx++)
140 /* Load shift vector for this list */
141 i_shift_offset = DIM*shiftidx[iidx];
143 /* Load limits for loop over neighbors */
144 j_index_start = jindex[iidx];
145 j_index_end = jindex[iidx+1];
147 /* Get outer coordinate index */
149 i_coord_offset = DIM*inr;
151 /* Load i particle coords and add shift vector */
152 gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
154 fix0 = _mm_setzero_pd();
155 fiy0 = _mm_setzero_pd();
156 fiz0 = _mm_setzero_pd();
158 /* Load parameters for i particles */
159 iq0 = _mm_mul_pd(facel,_mm_load1_pd(charge+inr+0));
160 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
162 /* Reset potential sums */
163 velecsum = _mm_setzero_pd();
164 vvdwsum = _mm_setzero_pd();
166 /* Start inner kernel loop */
167 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
170 /* Get j neighbor index, and coordinate index */
173 j_coord_offsetA = DIM*jnrA;
174 j_coord_offsetB = DIM*jnrB;
176 /* load j atom coordinates */
177 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
180 /* Calculate displacement vector */
181 dx00 = _mm_sub_pd(ix0,jx0);
182 dy00 = _mm_sub_pd(iy0,jy0);
183 dz00 = _mm_sub_pd(iz0,jz0);
185 /* Calculate squared distance and things based on it */
186 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
188 rinv00 = sse41_invsqrt_d(rsq00);
190 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
192 /* Load parameters for j particles */
193 jq0 = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
194 vdwjidx0A = 2*vdwtype[jnrA+0];
195 vdwjidx0B = 2*vdwtype[jnrB+0];
197 /**************************
198 * CALCULATE INTERACTIONS *
199 **************************/
201 if (gmx_mm_any_lt(rsq00,rcutoff2))
204 r00 = _mm_mul_pd(rsq00,rinv00);
206 /* Compute parameters for interactions between i and j atoms */
207 qq00 = _mm_mul_pd(iq0,jq0);
208 gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
209 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
211 /* EWALD ELECTROSTATICS */
213 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
214 ewrt = _mm_mul_pd(r00,ewtabscale);
215 ewitab = _mm_cvttpd_epi32(ewrt);
216 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
217 ewitab = _mm_slli_epi32(ewitab,2);
218 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
219 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
220 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
221 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
222 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
223 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
224 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
225 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
226 velec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_sub_pd(rinv00,sh_ewald),velec));
227 felec = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
229 /* LENNARD-JONES DISPERSION/REPULSION */
231 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
232 vvdw6 = _mm_mul_pd(c6_00,rinvsix);
233 vvdw12 = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
234 vvdw = _mm_sub_pd(_mm_mul_pd( _mm_sub_pd(vvdw12 , _mm_mul_pd(c12_00,_mm_mul_pd(sh_vdw_invrcut6,sh_vdw_invrcut6))), one_twelfth) ,
235 _mm_mul_pd( _mm_sub_pd(vvdw6,_mm_mul_pd(c6_00,sh_vdw_invrcut6)),one_sixth));
236 fvdw = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
238 cutoff_mask = _mm_cmplt_pd(rsq00,rcutoff2);
240 /* Update potential sum for this i atom from the interaction with this j atom. */
241 velec = _mm_and_pd(velec,cutoff_mask);
242 velecsum = _mm_add_pd(velecsum,velec);
243 vvdw = _mm_and_pd(vvdw,cutoff_mask);
244 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
246 fscal = _mm_add_pd(felec,fvdw);
248 fscal = _mm_and_pd(fscal,cutoff_mask);
250 /* Calculate temporary vectorial force */
251 tx = _mm_mul_pd(fscal,dx00);
252 ty = _mm_mul_pd(fscal,dy00);
253 tz = _mm_mul_pd(fscal,dz00);
255 /* Update vectorial force */
256 fix0 = _mm_add_pd(fix0,tx);
257 fiy0 = _mm_add_pd(fiy0,ty);
258 fiz0 = _mm_add_pd(fiz0,tz);
260 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,tx,ty,tz);
264 /* Inner loop uses 64 flops */
271 j_coord_offsetA = DIM*jnrA;
273 /* load j atom coordinates */
274 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
277 /* Calculate displacement vector */
278 dx00 = _mm_sub_pd(ix0,jx0);
279 dy00 = _mm_sub_pd(iy0,jy0);
280 dz00 = _mm_sub_pd(iz0,jz0);
282 /* Calculate squared distance and things based on it */
283 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
285 rinv00 = sse41_invsqrt_d(rsq00);
287 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
289 /* Load parameters for j particles */
290 jq0 = _mm_load_sd(charge+jnrA+0);
291 vdwjidx0A = 2*vdwtype[jnrA+0];
293 /**************************
294 * CALCULATE INTERACTIONS *
295 **************************/
297 if (gmx_mm_any_lt(rsq00,rcutoff2))
300 r00 = _mm_mul_pd(rsq00,rinv00);
302 /* Compute parameters for interactions between i and j atoms */
303 qq00 = _mm_mul_pd(iq0,jq0);
304 gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
306 /* EWALD ELECTROSTATICS */
308 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
309 ewrt = _mm_mul_pd(r00,ewtabscale);
310 ewitab = _mm_cvttpd_epi32(ewrt);
311 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
312 ewitab = _mm_slli_epi32(ewitab,2);
313 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
314 ewtabD = _mm_setzero_pd();
315 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
316 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
317 ewtabFn = _mm_setzero_pd();
318 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
319 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
320 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
321 velec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_sub_pd(rinv00,sh_ewald),velec));
322 felec = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
324 /* LENNARD-JONES DISPERSION/REPULSION */
326 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
327 vvdw6 = _mm_mul_pd(c6_00,rinvsix);
328 vvdw12 = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
329 vvdw = _mm_sub_pd(_mm_mul_pd( _mm_sub_pd(vvdw12 , _mm_mul_pd(c12_00,_mm_mul_pd(sh_vdw_invrcut6,sh_vdw_invrcut6))), one_twelfth) ,
330 _mm_mul_pd( _mm_sub_pd(vvdw6,_mm_mul_pd(c6_00,sh_vdw_invrcut6)),one_sixth));
331 fvdw = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
333 cutoff_mask = _mm_cmplt_pd(rsq00,rcutoff2);
335 /* Update potential sum for this i atom from the interaction with this j atom. */
336 velec = _mm_and_pd(velec,cutoff_mask);
337 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
338 velecsum = _mm_add_pd(velecsum,velec);
339 vvdw = _mm_and_pd(vvdw,cutoff_mask);
340 vvdw = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
341 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
343 fscal = _mm_add_pd(felec,fvdw);
345 fscal = _mm_and_pd(fscal,cutoff_mask);
347 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
349 /* Calculate temporary vectorial force */
350 tx = _mm_mul_pd(fscal,dx00);
351 ty = _mm_mul_pd(fscal,dy00);
352 tz = _mm_mul_pd(fscal,dz00);
354 /* Update vectorial force */
355 fix0 = _mm_add_pd(fix0,tx);
356 fiy0 = _mm_add_pd(fiy0,ty);
357 fiz0 = _mm_add_pd(fiz0,tz);
359 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,tx,ty,tz);
363 /* Inner loop uses 64 flops */
366 /* End of innermost loop */
368 gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
369 f+i_coord_offset,fshift+i_shift_offset);
372 /* Update potential energies */
373 gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
374 gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
376 /* Increment number of inner iterations */
377 inneriter += j_index_end - j_index_start;
379 /* Outer loop uses 9 flops */
382 /* Increment number of outer iterations */
385 /* Update outer/inner flops */
387 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_VF,outeriter*9 + inneriter*64);
390 * Gromacs nonbonded kernel: nb_kernel_ElecEwSh_VdwLJSh_GeomP1P1_F_sse4_1_double
391 * Electrostatics interaction: Ewald
392 * VdW interaction: LennardJones
393 * Geometry: Particle-Particle
394 * Calculate force/pot: Force
397 nb_kernel_ElecEwSh_VdwLJSh_GeomP1P1_F_sse4_1_double
398 (t_nblist * gmx_restrict nlist,
399 rvec * gmx_restrict xx,
400 rvec * gmx_restrict ff,
401 struct t_forcerec * gmx_restrict fr,
402 t_mdatoms * gmx_restrict mdatoms,
403 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
404 t_nrnb * gmx_restrict nrnb)
406 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
407 * just 0 for non-waters.
408 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
409 * jnr indices corresponding to data put in the four positions in the SIMD register.
411 int i_shift_offset,i_coord_offset,outeriter,inneriter;
412 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
414 int j_coord_offsetA,j_coord_offsetB;
415 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
417 real *shiftvec,*fshift,*x,*f;
418 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
420 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
421 int vdwjidx0A,vdwjidx0B;
422 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
423 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
424 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
427 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
430 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
431 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
433 __m128d ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
435 __m128d dummy_mask,cutoff_mask;
436 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
437 __m128d one = _mm_set1_pd(1.0);
438 __m128d two = _mm_set1_pd(2.0);
444 jindex = nlist->jindex;
446 shiftidx = nlist->shift;
448 shiftvec = fr->shift_vec[0];
449 fshift = fr->fshift[0];
450 facel = _mm_set1_pd(fr->ic->epsfac);
451 charge = mdatoms->chargeA;
452 nvdwtype = fr->ntype;
454 vdwtype = mdatoms->typeA;
456 sh_ewald = _mm_set1_pd(fr->ic->sh_ewald);
457 ewtab = fr->ic->tabq_coul_F;
458 ewtabscale = _mm_set1_pd(fr->ic->tabq_scale);
459 ewtabhalfspace = _mm_set1_pd(0.5/fr->ic->tabq_scale);
461 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
462 rcutoff_scalar = fr->ic->rcoulomb;
463 rcutoff = _mm_set1_pd(rcutoff_scalar);
464 rcutoff2 = _mm_mul_pd(rcutoff,rcutoff);
466 sh_vdw_invrcut6 = _mm_set1_pd(fr->ic->sh_invrc6);
467 rvdw = _mm_set1_pd(fr->ic->rvdw);
469 /* Avoid stupid compiler warnings */
477 /* Start outer loop over neighborlists */
478 for(iidx=0; iidx<nri; iidx++)
480 /* Load shift vector for this list */
481 i_shift_offset = DIM*shiftidx[iidx];
483 /* Load limits for loop over neighbors */
484 j_index_start = jindex[iidx];
485 j_index_end = jindex[iidx+1];
487 /* Get outer coordinate index */
489 i_coord_offset = DIM*inr;
491 /* Load i particle coords and add shift vector */
492 gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
494 fix0 = _mm_setzero_pd();
495 fiy0 = _mm_setzero_pd();
496 fiz0 = _mm_setzero_pd();
498 /* Load parameters for i particles */
499 iq0 = _mm_mul_pd(facel,_mm_load1_pd(charge+inr+0));
500 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
502 /* Start inner kernel loop */
503 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
506 /* Get j neighbor index, and coordinate index */
509 j_coord_offsetA = DIM*jnrA;
510 j_coord_offsetB = DIM*jnrB;
512 /* load j atom coordinates */
513 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
516 /* Calculate displacement vector */
517 dx00 = _mm_sub_pd(ix0,jx0);
518 dy00 = _mm_sub_pd(iy0,jy0);
519 dz00 = _mm_sub_pd(iz0,jz0);
521 /* Calculate squared distance and things based on it */
522 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
524 rinv00 = sse41_invsqrt_d(rsq00);
526 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
528 /* Load parameters for j particles */
529 jq0 = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
530 vdwjidx0A = 2*vdwtype[jnrA+0];
531 vdwjidx0B = 2*vdwtype[jnrB+0];
533 /**************************
534 * CALCULATE INTERACTIONS *
535 **************************/
537 if (gmx_mm_any_lt(rsq00,rcutoff2))
540 r00 = _mm_mul_pd(rsq00,rinv00);
542 /* Compute parameters for interactions between i and j atoms */
543 qq00 = _mm_mul_pd(iq0,jq0);
544 gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
545 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
547 /* EWALD ELECTROSTATICS */
549 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
550 ewrt = _mm_mul_pd(r00,ewtabscale);
551 ewitab = _mm_cvttpd_epi32(ewrt);
552 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
553 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
555 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
556 felec = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
558 /* LENNARD-JONES DISPERSION/REPULSION */
560 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
561 fvdw = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
563 cutoff_mask = _mm_cmplt_pd(rsq00,rcutoff2);
565 fscal = _mm_add_pd(felec,fvdw);
567 fscal = _mm_and_pd(fscal,cutoff_mask);
569 /* Calculate temporary vectorial force */
570 tx = _mm_mul_pd(fscal,dx00);
571 ty = _mm_mul_pd(fscal,dy00);
572 tz = _mm_mul_pd(fscal,dz00);
574 /* Update vectorial force */
575 fix0 = _mm_add_pd(fix0,tx);
576 fiy0 = _mm_add_pd(fiy0,ty);
577 fiz0 = _mm_add_pd(fiz0,tz);
579 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,tx,ty,tz);
583 /* Inner loop uses 46 flops */
590 j_coord_offsetA = DIM*jnrA;
592 /* load j atom coordinates */
593 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
596 /* Calculate displacement vector */
597 dx00 = _mm_sub_pd(ix0,jx0);
598 dy00 = _mm_sub_pd(iy0,jy0);
599 dz00 = _mm_sub_pd(iz0,jz0);
601 /* Calculate squared distance and things based on it */
602 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
604 rinv00 = sse41_invsqrt_d(rsq00);
606 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
608 /* Load parameters for j particles */
609 jq0 = _mm_load_sd(charge+jnrA+0);
610 vdwjidx0A = 2*vdwtype[jnrA+0];
612 /**************************
613 * CALCULATE INTERACTIONS *
614 **************************/
616 if (gmx_mm_any_lt(rsq00,rcutoff2))
619 r00 = _mm_mul_pd(rsq00,rinv00);
621 /* Compute parameters for interactions between i and j atoms */
622 qq00 = _mm_mul_pd(iq0,jq0);
623 gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
625 /* EWALD ELECTROSTATICS */
627 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
628 ewrt = _mm_mul_pd(r00,ewtabscale);
629 ewitab = _mm_cvttpd_epi32(ewrt);
630 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
631 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
632 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
633 felec = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
635 /* LENNARD-JONES DISPERSION/REPULSION */
637 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
638 fvdw = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
640 cutoff_mask = _mm_cmplt_pd(rsq00,rcutoff2);
642 fscal = _mm_add_pd(felec,fvdw);
644 fscal = _mm_and_pd(fscal,cutoff_mask);
646 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
648 /* Calculate temporary vectorial force */
649 tx = _mm_mul_pd(fscal,dx00);
650 ty = _mm_mul_pd(fscal,dy00);
651 tz = _mm_mul_pd(fscal,dz00);
653 /* Update vectorial force */
654 fix0 = _mm_add_pd(fix0,tx);
655 fiy0 = _mm_add_pd(fiy0,ty);
656 fiz0 = _mm_add_pd(fiz0,tz);
658 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,tx,ty,tz);
662 /* Inner loop uses 46 flops */
665 /* End of innermost loop */
667 gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
668 f+i_coord_offset,fshift+i_shift_offset);
670 /* Increment number of inner iterations */
671 inneriter += j_index_end - j_index_start;
673 /* Outer loop uses 7 flops */
676 /* Increment number of outer iterations */
679 /* Update outer/inner flops */
681 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_F,outeriter*7 + inneriter*46);