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 sparc64_hpc_ace_double kernel generator.
42 #include "../nb_kernel.h"
43 #include "types/simple.h"
44 #include "gromacs/math/vec.h"
47 #include "kernelutil_sparc64_hpc_ace_double.h"
50 * Gromacs nonbonded kernel: nb_kernel_ElecNone_VdwLJEwSh_GeomP1P1_VF_sparc64_hpc_ace_double
51 * Electrostatics interaction: None
52 * VdW interaction: LJEwald
53 * Geometry: Particle-Particle
54 * Calculate force/pot: PotentialAndForce
57 nb_kernel_ElecNone_VdwLJEwSh_GeomP1P1_VF_sparc64_hpc_ace_double
58 (t_nblist * gmx_restrict nlist,
59 rvec * gmx_restrict xx,
60 rvec * gmx_restrict ff,
61 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 double precision SIMD, 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 _fjsp_v2r8 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
80 _fjsp_v2r8 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
81 int vdwjidx0A,vdwjidx0B;
82 _fjsp_v2r8 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
83 _fjsp_v2r8 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
85 _fjsp_v2r8 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
88 _fjsp_v2r8 one_sixth = gmx_fjsp_set1_v2r8(1.0/6.0);
89 _fjsp_v2r8 one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
92 _fjsp_v2r8 ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
93 _fjsp_v2r8 one_half = gmx_fjsp_set1_v2r8(0.5);
94 _fjsp_v2r8 minus_one = gmx_fjsp_set1_v2r8(-1.0);
96 _fjsp_v2r8 dummy_mask,cutoff_mask;
97 _fjsp_v2r8 one = gmx_fjsp_set1_v2r8(1.0);
98 _fjsp_v2r8 two = gmx_fjsp_set1_v2r8(2.0);
99 union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
106 jindex = nlist->jindex;
108 shiftidx = nlist->shift;
110 shiftvec = fr->shift_vec[0];
111 fshift = fr->fshift[0];
112 nvdwtype = fr->ntype;
114 vdwtype = mdatoms->typeA;
115 vdwgridparam = fr->ljpme_c6grid;
116 sh_lj_ewald = gmx_fjsp_set1_v2r8(fr->ic->sh_lj_ewald);
117 ewclj = gmx_fjsp_set1_v2r8(fr->ewaldcoeff_lj);
118 ewclj2 = _fjsp_mul_v2r8(minus_one,_fjsp_mul_v2r8(ewclj,ewclj));
120 rcutoff_scalar = fr->rvdw;
121 rcutoff = gmx_fjsp_set1_v2r8(rcutoff_scalar);
122 rcutoff2 = _fjsp_mul_v2r8(rcutoff,rcutoff);
124 sh_vdw_invrcut6 = gmx_fjsp_set1_v2r8(fr->ic->sh_invrc6);
125 rvdw = gmx_fjsp_set1_v2r8(fr->rvdw);
127 /* Avoid stupid compiler warnings */
135 /* Start outer loop over neighborlists */
136 for(iidx=0; iidx<nri; iidx++)
138 /* Load shift vector for this list */
139 i_shift_offset = DIM*shiftidx[iidx];
141 /* Load limits for loop over neighbors */
142 j_index_start = jindex[iidx];
143 j_index_end = jindex[iidx+1];
145 /* Get outer coordinate index */
147 i_coord_offset = DIM*inr;
149 /* Load i particle coords and add shift vector */
150 gmx_fjsp_load_shift_and_1rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
152 fix0 = _fjsp_setzero_v2r8();
153 fiy0 = _fjsp_setzero_v2r8();
154 fiz0 = _fjsp_setzero_v2r8();
156 /* Load parameters for i particles */
157 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
159 /* Reset potential sums */
160 vvdwsum = _fjsp_setzero_v2r8();
162 /* Start inner kernel loop */
163 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
166 /* Get j neighbor index, and coordinate index */
169 j_coord_offsetA = DIM*jnrA;
170 j_coord_offsetB = DIM*jnrB;
172 /* load j atom coordinates */
173 gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
176 /* Calculate displacement vector */
177 dx00 = _fjsp_sub_v2r8(ix0,jx0);
178 dy00 = _fjsp_sub_v2r8(iy0,jy0);
179 dz00 = _fjsp_sub_v2r8(iz0,jz0);
181 /* Calculate squared distance and things based on it */
182 rsq00 = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
184 rinv00 = gmx_fjsp_invsqrt_v2r8(rsq00);
186 rinvsq00 = _fjsp_mul_v2r8(rinv00,rinv00);
188 /* Load parameters for j particles */
189 vdwjidx0A = 2*vdwtype[jnrA+0];
190 vdwjidx0B = 2*vdwtype[jnrB+0];
192 /**************************
193 * CALCULATE INTERACTIONS *
194 **************************/
196 if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
199 r00 = _fjsp_mul_v2r8(rsq00,rinv00);
201 /* Compute parameters for interactions between i and j atoms */
202 gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
203 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
205 c6grid_00 = gmx_fjsp_load_2real_swizzle_v2r8(vdwgridparam+vdwioffset0+vdwjidx0A,
206 vdwgridparam+vdwioffset0+vdwjidx0B);
208 /* Analytical LJ-PME */
209 rinvsix = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
210 ewcljrsq = _fjsp_mul_v2r8(ewclj2,rsq00);
211 ewclj6 = _fjsp_mul_v2r8(ewclj2,_fjsp_mul_v2r8(ewclj2,ewclj2));
212 exponent = gmx_simd_exp_d(-ewcljrsq);
213 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
214 poly = _fjsp_mul_v2r8(exponent,_fjsp_madd_v2r8(_fjsp_mul_v2r8(ewcljrsq,ewcljrsq),one_half,_fjsp_sub_v2r8(one,ewcljrsq)));
215 /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
216 vvdw6 = _fjsp_mul_v2r8(_fjsp_madd_v2r8(-c6grid_00,_fjsp_sub_v2r8(one,poly),c6_00),rinvsix);
217 vvdw12 = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
218 vvdw = _fjsp_msub_v2r8(_fjsp_nmsub_v2r8(c12_00,_fjsp_mul_v2r8(sh_vdw_invrcut6,sh_vdw_invrcut6),vvdw12),one_twelfth,
219 _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw6,_fjsp_madd_v2r8(c6grid_00,sh_lj_ewald,_fjsp_mul_v2r8(c6_00,sh_vdw_invrcut6))),one_sixth));
220 /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
221 fvdw = _fjsp_mul_v2r8(_fjsp_add_v2r8(vvdw12,_fjsp_msub_v2r8(_fjsp_mul_v2r8(c6grid_00,one_sixth),_fjsp_mul_v2r8(exponent,ewclj6),vvdw6)),rinvsq00);
223 cutoff_mask = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
225 /* Update potential sum for this i atom from the interaction with this j atom. */
226 vvdw = _fjsp_and_v2r8(vvdw,cutoff_mask);
227 vvdwsum = _fjsp_add_v2r8(vvdwsum,vvdw);
231 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
233 /* Update vectorial force */
234 fix0 = _fjsp_madd_v2r8(dx00,fscal,fix0);
235 fiy0 = _fjsp_madd_v2r8(dy00,fscal,fiy0);
236 fiz0 = _fjsp_madd_v2r8(dz00,fscal,fiz0);
238 gmx_fjsp_decrement_fma_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fscal,dx00,dy00,dz00);
242 /* Inner loop uses 59 flops */
249 j_coord_offsetA = DIM*jnrA;
251 /* load j atom coordinates */
252 gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
255 /* Calculate displacement vector */
256 dx00 = _fjsp_sub_v2r8(ix0,jx0);
257 dy00 = _fjsp_sub_v2r8(iy0,jy0);
258 dz00 = _fjsp_sub_v2r8(iz0,jz0);
260 /* Calculate squared distance and things based on it */
261 rsq00 = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
263 rinv00 = gmx_fjsp_invsqrt_v2r8(rsq00);
265 rinvsq00 = _fjsp_mul_v2r8(rinv00,rinv00);
267 /* Load parameters for j particles */
268 vdwjidx0A = 2*vdwtype[jnrA+0];
270 /**************************
271 * CALCULATE INTERACTIONS *
272 **************************/
274 if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
277 r00 = _fjsp_mul_v2r8(rsq00,rinv00);
279 /* Compute parameters for interactions between i and j atoms */
280 gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
282 c6grid_00 = gmx_fjsp_load_1real_swizzle_v2r8(vdwgridparam+vdwioffset0+vdwjidx0A);
284 /* Analytical LJ-PME */
285 rinvsix = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
286 ewcljrsq = _fjsp_mul_v2r8(ewclj2,rsq00);
287 ewclj6 = _fjsp_mul_v2r8(ewclj2,_fjsp_mul_v2r8(ewclj2,ewclj2));
288 exponent = gmx_simd_exp_d(-ewcljrsq);
289 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
290 poly = _fjsp_mul_v2r8(exponent,_fjsp_madd_v2r8(_fjsp_mul_v2r8(ewcljrsq,ewcljrsq),one_half,_fjsp_sub_v2r8(one,ewcljrsq)));
291 /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
292 vvdw6 = _fjsp_mul_v2r8(_fjsp_madd_v2r8(-c6grid_00,_fjsp_sub_v2r8(one,poly),c6_00),rinvsix);
293 vvdw12 = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
294 vvdw = _fjsp_msub_v2r8(_fjsp_nmsub_v2r8(c12_00,_fjsp_mul_v2r8(sh_vdw_invrcut6,sh_vdw_invrcut6),vvdw12),one_twelfth,
295 _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw6,_fjsp_madd_v2r8(c6grid_00,sh_lj_ewald,_fjsp_mul_v2r8(c6_00,sh_vdw_invrcut6))),one_sixth));
296 /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
297 fvdw = _fjsp_mul_v2r8(_fjsp_add_v2r8(vvdw12,_fjsp_msub_v2r8(_fjsp_mul_v2r8(c6grid_00,one_sixth),_fjsp_mul_v2r8(exponent,ewclj6),vvdw6)),rinvsq00);
299 cutoff_mask = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
301 /* Update potential sum for this i atom from the interaction with this j atom. */
302 vvdw = _fjsp_and_v2r8(vvdw,cutoff_mask);
303 vvdw = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
304 vvdwsum = _fjsp_add_v2r8(vvdwsum,vvdw);
308 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
310 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
312 /* Update vectorial force */
313 fix0 = _fjsp_madd_v2r8(dx00,fscal,fix0);
314 fiy0 = _fjsp_madd_v2r8(dy00,fscal,fiy0);
315 fiz0 = _fjsp_madd_v2r8(dz00,fscal,fiz0);
317 gmx_fjsp_decrement_fma_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fscal,dx00,dy00,dz00);
321 /* Inner loop uses 59 flops */
324 /* End of innermost loop */
326 gmx_fjsp_update_iforce_1atom_swizzle_v2r8(fix0,fiy0,fiz0,
327 f+i_coord_offset,fshift+i_shift_offset);
330 /* Update potential energies */
331 gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
333 /* Increment number of inner iterations */
334 inneriter += j_index_end - j_index_start;
336 /* Outer loop uses 7 flops */
339 /* Increment number of outer iterations */
342 /* Update outer/inner flops */
344 inc_nrnb(nrnb,eNR_NBKERNEL_VDW_VF,outeriter*7 + inneriter*59);
347 * Gromacs nonbonded kernel: nb_kernel_ElecNone_VdwLJEwSh_GeomP1P1_F_sparc64_hpc_ace_double
348 * Electrostatics interaction: None
349 * VdW interaction: LJEwald
350 * Geometry: Particle-Particle
351 * Calculate force/pot: Force
354 nb_kernel_ElecNone_VdwLJEwSh_GeomP1P1_F_sparc64_hpc_ace_double
355 (t_nblist * gmx_restrict nlist,
356 rvec * gmx_restrict xx,
357 rvec * gmx_restrict ff,
358 t_forcerec * gmx_restrict fr,
359 t_mdatoms * gmx_restrict mdatoms,
360 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
361 t_nrnb * gmx_restrict nrnb)
363 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
364 * just 0 for non-waters.
365 * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
366 * jnr indices corresponding to data put in the four positions in the SIMD register.
368 int i_shift_offset,i_coord_offset,outeriter,inneriter;
369 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
371 int j_coord_offsetA,j_coord_offsetB;
372 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
374 real *shiftvec,*fshift,*x,*f;
375 _fjsp_v2r8 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
377 _fjsp_v2r8 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
378 int vdwjidx0A,vdwjidx0B;
379 _fjsp_v2r8 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
380 _fjsp_v2r8 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
382 _fjsp_v2r8 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
385 _fjsp_v2r8 one_sixth = gmx_fjsp_set1_v2r8(1.0/6.0);
386 _fjsp_v2r8 one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
387 _fjsp_v2r8 c6grid_00;
389 _fjsp_v2r8 ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
390 _fjsp_v2r8 one_half = gmx_fjsp_set1_v2r8(0.5);
391 _fjsp_v2r8 minus_one = gmx_fjsp_set1_v2r8(-1.0);
393 _fjsp_v2r8 dummy_mask,cutoff_mask;
394 _fjsp_v2r8 one = gmx_fjsp_set1_v2r8(1.0);
395 _fjsp_v2r8 two = gmx_fjsp_set1_v2r8(2.0);
396 union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
403 jindex = nlist->jindex;
405 shiftidx = nlist->shift;
407 shiftvec = fr->shift_vec[0];
408 fshift = fr->fshift[0];
409 nvdwtype = fr->ntype;
411 vdwtype = mdatoms->typeA;
412 vdwgridparam = fr->ljpme_c6grid;
413 sh_lj_ewald = gmx_fjsp_set1_v2r8(fr->ic->sh_lj_ewald);
414 ewclj = gmx_fjsp_set1_v2r8(fr->ewaldcoeff_lj);
415 ewclj2 = _fjsp_mul_v2r8(minus_one,_fjsp_mul_v2r8(ewclj,ewclj));
417 rcutoff_scalar = fr->rvdw;
418 rcutoff = gmx_fjsp_set1_v2r8(rcutoff_scalar);
419 rcutoff2 = _fjsp_mul_v2r8(rcutoff,rcutoff);
421 sh_vdw_invrcut6 = gmx_fjsp_set1_v2r8(fr->ic->sh_invrc6);
422 rvdw = gmx_fjsp_set1_v2r8(fr->rvdw);
424 /* Avoid stupid compiler warnings */
432 /* Start outer loop over neighborlists */
433 for(iidx=0; iidx<nri; iidx++)
435 /* Load shift vector for this list */
436 i_shift_offset = DIM*shiftidx[iidx];
438 /* Load limits for loop over neighbors */
439 j_index_start = jindex[iidx];
440 j_index_end = jindex[iidx+1];
442 /* Get outer coordinate index */
444 i_coord_offset = DIM*inr;
446 /* Load i particle coords and add shift vector */
447 gmx_fjsp_load_shift_and_1rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
449 fix0 = _fjsp_setzero_v2r8();
450 fiy0 = _fjsp_setzero_v2r8();
451 fiz0 = _fjsp_setzero_v2r8();
453 /* Load parameters for i particles */
454 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
456 /* Start inner kernel loop */
457 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
460 /* Get j neighbor index, and coordinate index */
463 j_coord_offsetA = DIM*jnrA;
464 j_coord_offsetB = DIM*jnrB;
466 /* load j atom coordinates */
467 gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
470 /* Calculate displacement vector */
471 dx00 = _fjsp_sub_v2r8(ix0,jx0);
472 dy00 = _fjsp_sub_v2r8(iy0,jy0);
473 dz00 = _fjsp_sub_v2r8(iz0,jz0);
475 /* Calculate squared distance and things based on it */
476 rsq00 = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
478 rinv00 = gmx_fjsp_invsqrt_v2r8(rsq00);
480 rinvsq00 = _fjsp_mul_v2r8(rinv00,rinv00);
482 /* Load parameters for j particles */
483 vdwjidx0A = 2*vdwtype[jnrA+0];
484 vdwjidx0B = 2*vdwtype[jnrB+0];
486 /**************************
487 * CALCULATE INTERACTIONS *
488 **************************/
490 if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
493 r00 = _fjsp_mul_v2r8(rsq00,rinv00);
495 /* Compute parameters for interactions between i and j atoms */
496 gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
497 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
499 c6grid_00 = gmx_fjsp_load_2real_swizzle_v2r8(vdwgridparam+vdwioffset0+vdwjidx0A,
500 vdwgridparam+vdwioffset0+vdwjidx0B);
502 /* Analytical LJ-PME */
503 rinvsix = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
504 ewcljrsq = _fjsp_mul_v2r8(ewclj2,rsq00);
505 ewclj6 = _fjsp_mul_v2r8(ewclj2,_fjsp_mul_v2r8(ewclj2,ewclj2));
506 exponent = gmx_simd_exp_d(-ewcljrsq);
507 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
508 poly = _fjsp_mul_v2r8(exponent,_fjsp_madd_v2r8(_fjsp_mul_v2r8(ewcljrsq,ewcljrsq),one_half,_fjsp_sub_v2r8(one,ewcljrsq)));
509 /* f6A = 6 * C6grid * (1 - poly) */
510 f6A = _fjsp_mul_v2r8(c6grid_00,_fjsp_msub_v2r8(one,poly));
511 /* f6B = C6grid * exponent * beta^6 */
512 f6B = _fjsp_mul_v2r8(_fjsp_mul_v2r8(c6grid_00,one_sixth),_fjsp_mul_v2r8(exponent,ewclj6));
513 /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
514 fvdw = _fjsp_mul_v2r8(_fjsp_madd_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,_fjsp_sub_v2r8(c6_00,f6A)),rinvsix,f6B),rinvsq00);
516 cutoff_mask = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
520 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
522 /* Update vectorial force */
523 fix0 = _fjsp_madd_v2r8(dx00,fscal,fix0);
524 fiy0 = _fjsp_madd_v2r8(dy00,fscal,fiy0);
525 fiz0 = _fjsp_madd_v2r8(dz00,fscal,fiz0);
527 gmx_fjsp_decrement_fma_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fscal,dx00,dy00,dz00);
531 /* Inner loop uses 51 flops */
538 j_coord_offsetA = DIM*jnrA;
540 /* load j atom coordinates */
541 gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
544 /* Calculate displacement vector */
545 dx00 = _fjsp_sub_v2r8(ix0,jx0);
546 dy00 = _fjsp_sub_v2r8(iy0,jy0);
547 dz00 = _fjsp_sub_v2r8(iz0,jz0);
549 /* Calculate squared distance and things based on it */
550 rsq00 = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
552 rinv00 = gmx_fjsp_invsqrt_v2r8(rsq00);
554 rinvsq00 = _fjsp_mul_v2r8(rinv00,rinv00);
556 /* Load parameters for j particles */
557 vdwjidx0A = 2*vdwtype[jnrA+0];
559 /**************************
560 * CALCULATE INTERACTIONS *
561 **************************/
563 if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
566 r00 = _fjsp_mul_v2r8(rsq00,rinv00);
568 /* Compute parameters for interactions between i and j atoms */
569 gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
571 c6grid_00 = gmx_fjsp_load_1real_swizzle_v2r8(vdwgridparam+vdwioffset0+vdwjidx0A);
573 /* Analytical LJ-PME */
574 rinvsix = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
575 ewcljrsq = _fjsp_mul_v2r8(ewclj2,rsq00);
576 ewclj6 = _fjsp_mul_v2r8(ewclj2,_fjsp_mul_v2r8(ewclj2,ewclj2));
577 exponent = gmx_simd_exp_d(-ewcljrsq);
578 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
579 poly = _fjsp_mul_v2r8(exponent,_fjsp_madd_v2r8(_fjsp_mul_v2r8(ewcljrsq,ewcljrsq),one_half,_fjsp_sub_v2r8(one,ewcljrsq)));
580 /* f6A = 6 * C6grid * (1 - poly) */
581 f6A = _fjsp_mul_v2r8(c6grid_00,_fjsp_msub_v2r8(one,poly));
582 /* f6B = C6grid * exponent * beta^6 */
583 f6B = _fjsp_mul_v2r8(_fjsp_mul_v2r8(c6grid_00,one_sixth),_fjsp_mul_v2r8(exponent,ewclj6));
584 /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
585 fvdw = _fjsp_mul_v2r8(_fjsp_madd_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,_fjsp_sub_v2r8(c6_00,f6A)),rinvsix,f6B),rinvsq00);
587 cutoff_mask = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
591 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
593 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
595 /* Update vectorial force */
596 fix0 = _fjsp_madd_v2r8(dx00,fscal,fix0);
597 fiy0 = _fjsp_madd_v2r8(dy00,fscal,fiy0);
598 fiz0 = _fjsp_madd_v2r8(dz00,fscal,fiz0);
600 gmx_fjsp_decrement_fma_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fscal,dx00,dy00,dz00);
604 /* Inner loop uses 51 flops */
607 /* End of innermost loop */
609 gmx_fjsp_update_iforce_1atom_swizzle_v2r8(fix0,fiy0,fiz0,
610 f+i_coord_offset,fshift+i_shift_offset);
612 /* Increment number of inner iterations */
613 inneriter += j_index_end - j_index_start;
615 /* Outer loop uses 6 flops */
618 /* Increment number of outer iterations */
621 /* Update outer/inner flops */
623 inc_nrnb(nrnb,eNR_NBKERNEL_VDW_F,outeriter*6 + inneriter*51);