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 sse4_1_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_sse4_1_double.h"
48 #include "kernelutil_x86_sse4_1_double.h"
51 * Gromacs nonbonded kernel: nb_kernel_ElecNone_VdwLJEw_GeomP1P1_VF_sse4_1_double
52 * Electrostatics interaction: None
53 * VdW interaction: LJEwald
54 * Geometry: Particle-Particle
55 * Calculate force/pot: PotentialAndForce
58 nb_kernel_ElecNone_VdwLJEw_GeomP1P1_VF_sse4_1_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;
86 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
89 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
90 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
92 __m128d ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
94 __m128d one_half = _mm_set1_pd(0.5);
95 __m128d minus_one = _mm_set1_pd(-1.0);
96 __m128d dummy_mask,cutoff_mask;
97 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
98 __m128d one = _mm_set1_pd(1.0);
99 __m128d two = _mm_set1_pd(2.0);
105 jindex = nlist->jindex;
107 shiftidx = nlist->shift;
109 shiftvec = fr->shift_vec[0];
110 fshift = fr->fshift[0];
111 nvdwtype = fr->ntype;
113 vdwtype = mdatoms->typeA;
114 vdwgridparam = fr->ljpme_c6grid;
115 sh_lj_ewald = _mm_set1_pd(fr->ic->sh_lj_ewald);
116 ewclj = _mm_set1_pd(fr->ewaldcoeff_lj);
117 ewclj2 = _mm_mul_pd(minus_one,_mm_mul_pd(ewclj,ewclj));
119 /* Avoid stupid compiler warnings */
127 /* Start outer loop over neighborlists */
128 for(iidx=0; iidx<nri; iidx++)
130 /* Load shift vector for this list */
131 i_shift_offset = DIM*shiftidx[iidx];
133 /* Load limits for loop over neighbors */
134 j_index_start = jindex[iidx];
135 j_index_end = jindex[iidx+1];
137 /* Get outer coordinate index */
139 i_coord_offset = DIM*inr;
141 /* Load i particle coords and add shift vector */
142 gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
144 fix0 = _mm_setzero_pd();
145 fiy0 = _mm_setzero_pd();
146 fiz0 = _mm_setzero_pd();
148 /* Load parameters for i particles */
149 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
151 /* Reset potential sums */
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 vdwjidx0A = 2*vdwtype[jnrA+0];
182 vdwjidx0B = 2*vdwtype[jnrB+0];
184 /**************************
185 * CALCULATE INTERACTIONS *
186 **************************/
188 r00 = _mm_mul_pd(rsq00,rinv00);
190 /* Compute parameters for interactions between i and j atoms */
191 gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
192 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
193 c6grid_00 = gmx_mm_load_2real_swizzle_pd(vdwgridparam+vdwioffset0+vdwjidx0A,
194 vdwgridparam+vdwioffset0+vdwjidx0B);
196 /* Analytical LJ-PME */
197 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
198 ewcljrsq = _mm_mul_pd(ewclj2,rsq00);
199 ewclj6 = _mm_mul_pd(ewclj2,_mm_mul_pd(ewclj2,ewclj2));
200 exponent = gmx_simd_exp_d(ewcljrsq);
201 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
202 poly = _mm_mul_pd(exponent,_mm_add_pd(_mm_sub_pd(one,ewcljrsq),_mm_mul_pd(_mm_mul_pd(ewcljrsq,ewcljrsq),one_half)));
203 /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
204 vvdw6 = _mm_mul_pd(_mm_sub_pd(c6_00,_mm_mul_pd(c6grid_00,_mm_sub_pd(one,poly))),rinvsix);
205 vvdw12 = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
206 vvdw = _mm_sub_pd(_mm_mul_pd(vvdw12,one_twelfth),_mm_mul_pd(vvdw6,one_sixth));
207 /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
208 fvdw = _mm_mul_pd(_mm_sub_pd(vvdw12,_mm_sub_pd(vvdw6,_mm_mul_pd(_mm_mul_pd(c6grid_00,one_sixth),_mm_mul_pd(exponent,ewclj6)))),rinvsq00);
210 /* Update potential sum for this i atom from the interaction with this j atom. */
211 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
215 /* Calculate temporary vectorial force */
216 tx = _mm_mul_pd(fscal,dx00);
217 ty = _mm_mul_pd(fscal,dy00);
218 tz = _mm_mul_pd(fscal,dz00);
220 /* Update vectorial force */
221 fix0 = _mm_add_pd(fix0,tx);
222 fiy0 = _mm_add_pd(fiy0,ty);
223 fiz0 = _mm_add_pd(fiz0,tz);
225 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,tx,ty,tz);
227 /* Inner loop uses 51 flops */
234 j_coord_offsetA = DIM*jnrA;
236 /* load j atom coordinates */
237 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
240 /* Calculate displacement vector */
241 dx00 = _mm_sub_pd(ix0,jx0);
242 dy00 = _mm_sub_pd(iy0,jy0);
243 dz00 = _mm_sub_pd(iz0,jz0);
245 /* Calculate squared distance and things based on it */
246 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
248 rinv00 = gmx_mm_invsqrt_pd(rsq00);
250 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
252 /* Load parameters for j particles */
253 vdwjidx0A = 2*vdwtype[jnrA+0];
255 /**************************
256 * CALCULATE INTERACTIONS *
257 **************************/
259 r00 = _mm_mul_pd(rsq00,rinv00);
261 /* Compute parameters for interactions between i and j atoms */
262 gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
264 c6grid_00 = gmx_mm_load_1real_pd(vdwgridparam+vdwioffset0+vdwjidx0A);
266 /* Analytical LJ-PME */
267 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
268 ewcljrsq = _mm_mul_pd(ewclj2,rsq00);
269 ewclj6 = _mm_mul_pd(ewclj2,_mm_mul_pd(ewclj2,ewclj2));
270 exponent = gmx_simd_exp_d(ewcljrsq);
271 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
272 poly = _mm_mul_pd(exponent,_mm_add_pd(_mm_sub_pd(one,ewcljrsq),_mm_mul_pd(_mm_mul_pd(ewcljrsq,ewcljrsq),one_half)));
273 /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
274 vvdw6 = _mm_mul_pd(_mm_sub_pd(c6_00,_mm_mul_pd(c6grid_00,_mm_sub_pd(one,poly))),rinvsix);
275 vvdw12 = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
276 vvdw = _mm_sub_pd(_mm_mul_pd(vvdw12,one_twelfth),_mm_mul_pd(vvdw6,one_sixth));
277 /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
278 fvdw = _mm_mul_pd(_mm_sub_pd(vvdw12,_mm_sub_pd(vvdw6,_mm_mul_pd(_mm_mul_pd(c6grid_00,one_sixth),_mm_mul_pd(exponent,ewclj6)))),rinvsq00);
280 /* Update potential sum for this i atom from the interaction with this j atom. */
281 vvdw = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
282 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
286 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
288 /* Calculate temporary vectorial force */
289 tx = _mm_mul_pd(fscal,dx00);
290 ty = _mm_mul_pd(fscal,dy00);
291 tz = _mm_mul_pd(fscal,dz00);
293 /* Update vectorial force */
294 fix0 = _mm_add_pd(fix0,tx);
295 fiy0 = _mm_add_pd(fiy0,ty);
296 fiz0 = _mm_add_pd(fiz0,tz);
298 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,tx,ty,tz);
300 /* Inner loop uses 51 flops */
303 /* End of innermost loop */
305 gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
306 f+i_coord_offset,fshift+i_shift_offset);
309 /* Update potential energies */
310 gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
312 /* Increment number of inner iterations */
313 inneriter += j_index_end - j_index_start;
315 /* Outer loop uses 7 flops */
318 /* Increment number of outer iterations */
321 /* Update outer/inner flops */
323 inc_nrnb(nrnb,eNR_NBKERNEL_VDW_VF,outeriter*7 + inneriter*51);
326 * Gromacs nonbonded kernel: nb_kernel_ElecNone_VdwLJEw_GeomP1P1_F_sse4_1_double
327 * Electrostatics interaction: None
328 * VdW interaction: LJEwald
329 * Geometry: Particle-Particle
330 * Calculate force/pot: Force
333 nb_kernel_ElecNone_VdwLJEw_GeomP1P1_F_sse4_1_double
334 (t_nblist * gmx_restrict nlist,
335 rvec * gmx_restrict xx,
336 rvec * gmx_restrict ff,
337 t_forcerec * gmx_restrict fr,
338 t_mdatoms * gmx_restrict mdatoms,
339 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
340 t_nrnb * gmx_restrict nrnb)
342 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
343 * just 0 for non-waters.
344 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
345 * jnr indices corresponding to data put in the four positions in the SIMD register.
347 int i_shift_offset,i_coord_offset,outeriter,inneriter;
348 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
350 int j_coord_offsetA,j_coord_offsetB;
351 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
353 real *shiftvec,*fshift,*x,*f;
354 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
356 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
357 int vdwjidx0A,vdwjidx0B;
358 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
359 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
361 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
364 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
365 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
367 __m128d ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
369 __m128d one_half = _mm_set1_pd(0.5);
370 __m128d minus_one = _mm_set1_pd(-1.0);
371 __m128d dummy_mask,cutoff_mask;
372 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
373 __m128d one = _mm_set1_pd(1.0);
374 __m128d two = _mm_set1_pd(2.0);
380 jindex = nlist->jindex;
382 shiftidx = nlist->shift;
384 shiftvec = fr->shift_vec[0];
385 fshift = fr->fshift[0];
386 nvdwtype = fr->ntype;
388 vdwtype = mdatoms->typeA;
389 vdwgridparam = fr->ljpme_c6grid;
390 sh_lj_ewald = _mm_set1_pd(fr->ic->sh_lj_ewald);
391 ewclj = _mm_set1_pd(fr->ewaldcoeff_lj);
392 ewclj2 = _mm_mul_pd(minus_one,_mm_mul_pd(ewclj,ewclj));
394 /* Avoid stupid compiler warnings */
402 /* Start outer loop over neighborlists */
403 for(iidx=0; iidx<nri; iidx++)
405 /* Load shift vector for this list */
406 i_shift_offset = DIM*shiftidx[iidx];
408 /* Load limits for loop over neighbors */
409 j_index_start = jindex[iidx];
410 j_index_end = jindex[iidx+1];
412 /* Get outer coordinate index */
414 i_coord_offset = DIM*inr;
416 /* Load i particle coords and add shift vector */
417 gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
419 fix0 = _mm_setzero_pd();
420 fiy0 = _mm_setzero_pd();
421 fiz0 = _mm_setzero_pd();
423 /* Load parameters for i particles */
424 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
426 /* Start inner kernel loop */
427 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
430 /* Get j neighbor index, and coordinate index */
433 j_coord_offsetA = DIM*jnrA;
434 j_coord_offsetB = DIM*jnrB;
436 /* load j atom coordinates */
437 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
440 /* Calculate displacement vector */
441 dx00 = _mm_sub_pd(ix0,jx0);
442 dy00 = _mm_sub_pd(iy0,jy0);
443 dz00 = _mm_sub_pd(iz0,jz0);
445 /* Calculate squared distance and things based on it */
446 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
448 rinv00 = gmx_mm_invsqrt_pd(rsq00);
450 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
452 /* Load parameters for j particles */
453 vdwjidx0A = 2*vdwtype[jnrA+0];
454 vdwjidx0B = 2*vdwtype[jnrB+0];
456 /**************************
457 * CALCULATE INTERACTIONS *
458 **************************/
460 r00 = _mm_mul_pd(rsq00,rinv00);
462 /* Compute parameters for interactions between i and j atoms */
463 gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
464 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
465 c6grid_00 = gmx_mm_load_2real_swizzle_pd(vdwgridparam+vdwioffset0+vdwjidx0A,
466 vdwgridparam+vdwioffset0+vdwjidx0B);
468 /* Analytical LJ-PME */
469 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
470 ewcljrsq = _mm_mul_pd(ewclj2,rsq00);
471 ewclj6 = _mm_mul_pd(ewclj2,_mm_mul_pd(ewclj2,ewclj2));
472 exponent = gmx_simd_exp_d(ewcljrsq);
473 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
474 poly = _mm_mul_pd(exponent,_mm_add_pd(_mm_sub_pd(one,ewcljrsq),_mm_mul_pd(_mm_mul_pd(ewcljrsq,ewcljrsq),one_half)));
475 /* f6A = 6 * C6grid * (1 - poly) */
476 f6A = _mm_mul_pd(c6grid_00,_mm_sub_pd(one,poly));
477 /* f6B = C6grid * exponent * beta^6 */
478 f6B = _mm_mul_pd(_mm_mul_pd(c6grid_00,one_sixth),_mm_mul_pd(exponent,ewclj6));
479 /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
480 fvdw = _mm_mul_pd(_mm_add_pd(_mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),_mm_sub_pd(c6_00,f6A)),rinvsix),f6B),rinvsq00);
484 /* Calculate temporary vectorial force */
485 tx = _mm_mul_pd(fscal,dx00);
486 ty = _mm_mul_pd(fscal,dy00);
487 tz = _mm_mul_pd(fscal,dz00);
489 /* Update vectorial force */
490 fix0 = _mm_add_pd(fix0,tx);
491 fiy0 = _mm_add_pd(fiy0,ty);
492 fiz0 = _mm_add_pd(fiz0,tz);
494 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,tx,ty,tz);
496 /* Inner loop uses 46 flops */
503 j_coord_offsetA = DIM*jnrA;
505 /* load j atom coordinates */
506 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
509 /* Calculate displacement vector */
510 dx00 = _mm_sub_pd(ix0,jx0);
511 dy00 = _mm_sub_pd(iy0,jy0);
512 dz00 = _mm_sub_pd(iz0,jz0);
514 /* Calculate squared distance and things based on it */
515 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
517 rinv00 = gmx_mm_invsqrt_pd(rsq00);
519 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
521 /* Load parameters for j particles */
522 vdwjidx0A = 2*vdwtype[jnrA+0];
524 /**************************
525 * CALCULATE INTERACTIONS *
526 **************************/
528 r00 = _mm_mul_pd(rsq00,rinv00);
530 /* Compute parameters for interactions between i and j atoms */
531 gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
533 c6grid_00 = gmx_mm_load_1real_pd(vdwgridparam+vdwioffset0+vdwjidx0A);
535 /* Analytical LJ-PME */
536 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
537 ewcljrsq = _mm_mul_pd(ewclj2,rsq00);
538 ewclj6 = _mm_mul_pd(ewclj2,_mm_mul_pd(ewclj2,ewclj2));
539 exponent = gmx_simd_exp_d(ewcljrsq);
540 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
541 poly = _mm_mul_pd(exponent,_mm_add_pd(_mm_sub_pd(one,ewcljrsq),_mm_mul_pd(_mm_mul_pd(ewcljrsq,ewcljrsq),one_half)));
542 /* f6A = 6 * C6grid * (1 - poly) */
543 f6A = _mm_mul_pd(c6grid_00,_mm_sub_pd(one,poly));
544 /* f6B = C6grid * exponent * beta^6 */
545 f6B = _mm_mul_pd(_mm_mul_pd(c6grid_00,one_sixth),_mm_mul_pd(exponent,ewclj6));
546 /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
547 fvdw = _mm_mul_pd(_mm_add_pd(_mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),_mm_sub_pd(c6_00,f6A)),rinvsix),f6B),rinvsq00);
551 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
553 /* Calculate temporary vectorial force */
554 tx = _mm_mul_pd(fscal,dx00);
555 ty = _mm_mul_pd(fscal,dy00);
556 tz = _mm_mul_pd(fscal,dz00);
558 /* Update vectorial force */
559 fix0 = _mm_add_pd(fix0,tx);
560 fiy0 = _mm_add_pd(fiy0,ty);
561 fiz0 = _mm_add_pd(fiz0,tz);
563 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,tx,ty,tz);
565 /* Inner loop uses 46 flops */
568 /* End of innermost loop */
570 gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
571 f+i_coord_offset,fshift+i_shift_offset);
573 /* Increment number of inner iterations */
574 inneriter += j_index_end - j_index_start;
576 /* Outer loop uses 6 flops */
579 /* Increment number of outer iterations */
582 /* Update outer/inner flops */
584 inc_nrnb(nrnb,eNR_NBKERNEL_VDW_F,outeriter*6 + inneriter*46);