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.
44 #include "../nb_kernel.h"
45 #include "gromacs/legacyheaders/types/simple.h"
46 #include "gromacs/math/vec.h"
47 #include "gromacs/legacyheaders/nrnb.h"
49 #include "kernelutil_sparc64_hpc_ace_double.h"
52 * Gromacs nonbonded kernel: nb_kernel_ElecEw_VdwLJEw_GeomW3P1_VF_sparc64_hpc_ace_double
53 * Electrostatics interaction: Ewald
54 * VdW interaction: LJEwald
55 * Geometry: Water3-Particle
56 * Calculate force/pot: PotentialAndForce
59 nb_kernel_ElecEw_VdwLJEw_GeomW3P1_VF_sparc64_hpc_ace_double
60 (t_nblist * gmx_restrict nlist,
61 rvec * gmx_restrict xx,
62 rvec * gmx_restrict ff,
63 t_forcerec * gmx_restrict fr,
64 t_mdatoms * gmx_restrict mdatoms,
65 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
66 t_nrnb * gmx_restrict nrnb)
68 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
69 * just 0 for non-waters.
70 * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
71 * jnr indices corresponding to data put in the four positions in the SIMD register.
73 int i_shift_offset,i_coord_offset,outeriter,inneriter;
74 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
76 int j_coord_offsetA,j_coord_offsetB;
77 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
79 real *shiftvec,*fshift,*x,*f;
80 _fjsp_v2r8 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
82 _fjsp_v2r8 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
84 _fjsp_v2r8 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
86 _fjsp_v2r8 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
87 int vdwjidx0A,vdwjidx0B;
88 _fjsp_v2r8 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
89 _fjsp_v2r8 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
90 _fjsp_v2r8 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
91 _fjsp_v2r8 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
92 _fjsp_v2r8 velec,felec,velecsum,facel,crf,krf,krf2;
95 _fjsp_v2r8 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
98 _fjsp_v2r8 one_sixth = gmx_fjsp_set1_v2r8(1.0/6.0);
99 _fjsp_v2r8 one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
100 _fjsp_v2r8 c6grid_00;
101 _fjsp_v2r8 c6grid_10;
102 _fjsp_v2r8 c6grid_20;
104 _fjsp_v2r8 ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
105 _fjsp_v2r8 one_half = gmx_fjsp_set1_v2r8(0.5);
106 _fjsp_v2r8 minus_one = gmx_fjsp_set1_v2r8(-1.0);
107 _fjsp_v2r8 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
110 _fjsp_v2r8 dummy_mask,cutoff_mask;
111 _fjsp_v2r8 one = gmx_fjsp_set1_v2r8(1.0);
112 _fjsp_v2r8 two = gmx_fjsp_set1_v2r8(2.0);
113 union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
120 jindex = nlist->jindex;
122 shiftidx = nlist->shift;
124 shiftvec = fr->shift_vec[0];
125 fshift = fr->fshift[0];
126 facel = gmx_fjsp_set1_v2r8(fr->epsfac);
127 charge = mdatoms->chargeA;
128 nvdwtype = fr->ntype;
130 vdwtype = mdatoms->typeA;
131 vdwgridparam = fr->ljpme_c6grid;
132 sh_lj_ewald = gmx_fjsp_set1_v2r8(fr->ic->sh_lj_ewald);
133 ewclj = gmx_fjsp_set1_v2r8(fr->ewaldcoeff_lj);
134 ewclj2 = _fjsp_mul_v2r8(minus_one,_fjsp_mul_v2r8(ewclj,ewclj));
136 sh_ewald = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
137 ewtab = fr->ic->tabq_coul_FDV0;
138 ewtabscale = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
139 ewtabhalfspace = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
141 /* Setup water-specific parameters */
142 inr = nlist->iinr[0];
143 iq0 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
144 iq1 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
145 iq2 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
146 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
148 /* Avoid stupid compiler warnings */
156 /* Start outer loop over neighborlists */
157 for(iidx=0; iidx<nri; iidx++)
159 /* Load shift vector for this list */
160 i_shift_offset = DIM*shiftidx[iidx];
162 /* Load limits for loop over neighbors */
163 j_index_start = jindex[iidx];
164 j_index_end = jindex[iidx+1];
166 /* Get outer coordinate index */
168 i_coord_offset = DIM*inr;
170 /* Load i particle coords and add shift vector */
171 gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
172 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
174 fix0 = _fjsp_setzero_v2r8();
175 fiy0 = _fjsp_setzero_v2r8();
176 fiz0 = _fjsp_setzero_v2r8();
177 fix1 = _fjsp_setzero_v2r8();
178 fiy1 = _fjsp_setzero_v2r8();
179 fiz1 = _fjsp_setzero_v2r8();
180 fix2 = _fjsp_setzero_v2r8();
181 fiy2 = _fjsp_setzero_v2r8();
182 fiz2 = _fjsp_setzero_v2r8();
184 /* Reset potential sums */
185 velecsum = _fjsp_setzero_v2r8();
186 vvdwsum = _fjsp_setzero_v2r8();
188 /* Start inner kernel loop */
189 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
192 /* Get j neighbor index, and coordinate index */
195 j_coord_offsetA = DIM*jnrA;
196 j_coord_offsetB = DIM*jnrB;
198 /* load j atom coordinates */
199 gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
202 /* Calculate displacement vector */
203 dx00 = _fjsp_sub_v2r8(ix0,jx0);
204 dy00 = _fjsp_sub_v2r8(iy0,jy0);
205 dz00 = _fjsp_sub_v2r8(iz0,jz0);
206 dx10 = _fjsp_sub_v2r8(ix1,jx0);
207 dy10 = _fjsp_sub_v2r8(iy1,jy0);
208 dz10 = _fjsp_sub_v2r8(iz1,jz0);
209 dx20 = _fjsp_sub_v2r8(ix2,jx0);
210 dy20 = _fjsp_sub_v2r8(iy2,jy0);
211 dz20 = _fjsp_sub_v2r8(iz2,jz0);
213 /* Calculate squared distance and things based on it */
214 rsq00 = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
215 rsq10 = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
216 rsq20 = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
218 rinv00 = gmx_fjsp_invsqrt_v2r8(rsq00);
219 rinv10 = gmx_fjsp_invsqrt_v2r8(rsq10);
220 rinv20 = gmx_fjsp_invsqrt_v2r8(rsq20);
222 rinvsq00 = _fjsp_mul_v2r8(rinv00,rinv00);
223 rinvsq10 = _fjsp_mul_v2r8(rinv10,rinv10);
224 rinvsq20 = _fjsp_mul_v2r8(rinv20,rinv20);
226 /* Load parameters for j particles */
227 jq0 = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
228 vdwjidx0A = 2*vdwtype[jnrA+0];
229 vdwjidx0B = 2*vdwtype[jnrB+0];
231 fjx0 = _fjsp_setzero_v2r8();
232 fjy0 = _fjsp_setzero_v2r8();
233 fjz0 = _fjsp_setzero_v2r8();
235 /**************************
236 * CALCULATE INTERACTIONS *
237 **************************/
239 r00 = _fjsp_mul_v2r8(rsq00,rinv00);
241 /* Compute parameters for interactions between i and j atoms */
242 qq00 = _fjsp_mul_v2r8(iq0,jq0);
243 gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
244 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
246 c6grid_00 = gmx_fjsp_load_2real_swizzle_v2r8(vdwgridparam+vdwioffset0+vdwjidx0A,
247 vdwgridparam+vdwioffset0+vdwjidx0B);
249 /* EWALD ELECTROSTATICS */
251 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
252 ewrt = _fjsp_mul_v2r8(r00,ewtabscale);
253 itab_tmp = _fjsp_dtox_v2r8(ewrt);
254 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
255 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
257 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
258 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
259 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
260 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
261 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
262 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
263 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
264 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
265 velec = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(rinv00,velec));
266 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
268 /* Analytical LJ-PME */
269 rinvsix = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
270 ewcljrsq = _fjsp_mul_v2r8(ewclj2,rsq00);
271 ewclj6 = _fjsp_mul_v2r8(ewclj2,_fjsp_mul_v2r8(ewclj2,ewclj2));
272 exponent = gmx_simd_exp_d(-ewcljrsq);
273 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
274 poly = _fjsp_mul_v2r8(exponent,_fjsp_madd_v2r8(_fjsp_mul_v2r8(ewcljrsq,ewcljrsq),one_half,_fjsp_sub_v2r8(one,ewcljrsq)));
275 /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
276 vvdw6 = _fjsp_mul_v2r8(_fjsp_madd_v2r8(-c6grid_00,_fjsp_sub_v2r8(one,poly),c6_00),rinvsix);
277 vvdw12 = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
278 vvdw = _fjsp_msub_v2r8(vvdw12,one_twelfth,_fjsp_mul_v2r8(vvdw6,one_sixth));
279 /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
280 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);
282 /* Update potential sum for this i atom from the interaction with this j atom. */
283 velecsum = _fjsp_add_v2r8(velecsum,velec);
284 vvdwsum = _fjsp_add_v2r8(vvdwsum,vvdw);
286 fscal = _fjsp_add_v2r8(felec,fvdw);
288 /* Update vectorial force */
289 fix0 = _fjsp_madd_v2r8(dx00,fscal,fix0);
290 fiy0 = _fjsp_madd_v2r8(dy00,fscal,fiy0);
291 fiz0 = _fjsp_madd_v2r8(dz00,fscal,fiz0);
293 fjx0 = _fjsp_madd_v2r8(dx00,fscal,fjx0);
294 fjy0 = _fjsp_madd_v2r8(dy00,fscal,fjy0);
295 fjz0 = _fjsp_madd_v2r8(dz00,fscal,fjz0);
297 /**************************
298 * CALCULATE INTERACTIONS *
299 **************************/
301 r10 = _fjsp_mul_v2r8(rsq10,rinv10);
303 /* Compute parameters for interactions between i and j atoms */
304 qq10 = _fjsp_mul_v2r8(iq1,jq0);
306 /* EWALD ELECTROSTATICS */
308 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
309 ewrt = _fjsp_mul_v2r8(r10,ewtabscale);
310 itab_tmp = _fjsp_dtox_v2r8(ewrt);
311 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
312 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
314 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
315 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
316 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
317 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
318 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
319 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
320 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
321 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
322 velec = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
323 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
325 /* Update potential sum for this i atom from the interaction with this j atom. */
326 velecsum = _fjsp_add_v2r8(velecsum,velec);
330 /* Update vectorial force */
331 fix1 = _fjsp_madd_v2r8(dx10,fscal,fix1);
332 fiy1 = _fjsp_madd_v2r8(dy10,fscal,fiy1);
333 fiz1 = _fjsp_madd_v2r8(dz10,fscal,fiz1);
335 fjx0 = _fjsp_madd_v2r8(dx10,fscal,fjx0);
336 fjy0 = _fjsp_madd_v2r8(dy10,fscal,fjy0);
337 fjz0 = _fjsp_madd_v2r8(dz10,fscal,fjz0);
339 /**************************
340 * CALCULATE INTERACTIONS *
341 **************************/
343 r20 = _fjsp_mul_v2r8(rsq20,rinv20);
345 /* Compute parameters for interactions between i and j atoms */
346 qq20 = _fjsp_mul_v2r8(iq2,jq0);
348 /* EWALD ELECTROSTATICS */
350 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
351 ewrt = _fjsp_mul_v2r8(r20,ewtabscale);
352 itab_tmp = _fjsp_dtox_v2r8(ewrt);
353 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
354 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
356 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
357 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
358 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
359 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
360 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
361 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
362 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
363 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
364 velec = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
365 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
367 /* Update potential sum for this i atom from the interaction with this j atom. */
368 velecsum = _fjsp_add_v2r8(velecsum,velec);
372 /* Update vectorial force */
373 fix2 = _fjsp_madd_v2r8(dx20,fscal,fix2);
374 fiy2 = _fjsp_madd_v2r8(dy20,fscal,fiy2);
375 fiz2 = _fjsp_madd_v2r8(dz20,fscal,fiz2);
377 fjx0 = _fjsp_madd_v2r8(dx20,fscal,fjx0);
378 fjy0 = _fjsp_madd_v2r8(dy20,fscal,fjy0);
379 fjz0 = _fjsp_madd_v2r8(dz20,fscal,fjz0);
381 gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
383 /* Inner loop uses 159 flops */
390 j_coord_offsetA = DIM*jnrA;
392 /* load j atom coordinates */
393 gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
396 /* Calculate displacement vector */
397 dx00 = _fjsp_sub_v2r8(ix0,jx0);
398 dy00 = _fjsp_sub_v2r8(iy0,jy0);
399 dz00 = _fjsp_sub_v2r8(iz0,jz0);
400 dx10 = _fjsp_sub_v2r8(ix1,jx0);
401 dy10 = _fjsp_sub_v2r8(iy1,jy0);
402 dz10 = _fjsp_sub_v2r8(iz1,jz0);
403 dx20 = _fjsp_sub_v2r8(ix2,jx0);
404 dy20 = _fjsp_sub_v2r8(iy2,jy0);
405 dz20 = _fjsp_sub_v2r8(iz2,jz0);
407 /* Calculate squared distance and things based on it */
408 rsq00 = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
409 rsq10 = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
410 rsq20 = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
412 rinv00 = gmx_fjsp_invsqrt_v2r8(rsq00);
413 rinv10 = gmx_fjsp_invsqrt_v2r8(rsq10);
414 rinv20 = gmx_fjsp_invsqrt_v2r8(rsq20);
416 rinvsq00 = _fjsp_mul_v2r8(rinv00,rinv00);
417 rinvsq10 = _fjsp_mul_v2r8(rinv10,rinv10);
418 rinvsq20 = _fjsp_mul_v2r8(rinv20,rinv20);
420 /* Load parameters for j particles */
421 jq0 = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
422 vdwjidx0A = 2*vdwtype[jnrA+0];
424 fjx0 = _fjsp_setzero_v2r8();
425 fjy0 = _fjsp_setzero_v2r8();
426 fjz0 = _fjsp_setzero_v2r8();
428 /**************************
429 * CALCULATE INTERACTIONS *
430 **************************/
432 r00 = _fjsp_mul_v2r8(rsq00,rinv00);
434 /* Compute parameters for interactions between i and j atoms */
435 qq00 = _fjsp_mul_v2r8(iq0,jq0);
436 gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
438 c6grid_00 = gmx_fjsp_load_1real_swizzle_v2r8(vdwgridparam+vdwioffset0+vdwjidx0A);
440 /* EWALD ELECTROSTATICS */
442 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
443 ewrt = _fjsp_mul_v2r8(r00,ewtabscale);
444 itab_tmp = _fjsp_dtox_v2r8(ewrt);
445 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
446 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
448 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
449 ewtabD = _fjsp_setzero_v2r8();
450 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
451 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
452 ewtabFn = _fjsp_setzero_v2r8();
453 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
454 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
455 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
456 velec = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(rinv00,velec));
457 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
459 /* Analytical LJ-PME */
460 rinvsix = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
461 ewcljrsq = _fjsp_mul_v2r8(ewclj2,rsq00);
462 ewclj6 = _fjsp_mul_v2r8(ewclj2,_fjsp_mul_v2r8(ewclj2,ewclj2));
463 exponent = gmx_simd_exp_d(-ewcljrsq);
464 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
465 poly = _fjsp_mul_v2r8(exponent,_fjsp_madd_v2r8(_fjsp_mul_v2r8(ewcljrsq,ewcljrsq),one_half,_fjsp_sub_v2r8(one,ewcljrsq)));
466 /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
467 vvdw6 = _fjsp_mul_v2r8(_fjsp_madd_v2r8(-c6grid_00,_fjsp_sub_v2r8(one,poly),c6_00),rinvsix);
468 vvdw12 = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
469 vvdw = _fjsp_msub_v2r8(vvdw12,one_twelfth,_fjsp_mul_v2r8(vvdw6,one_sixth));
470 /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
471 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);
473 /* Update potential sum for this i atom from the interaction with this j atom. */
474 velec = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
475 velecsum = _fjsp_add_v2r8(velecsum,velec);
476 vvdw = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
477 vvdwsum = _fjsp_add_v2r8(vvdwsum,vvdw);
479 fscal = _fjsp_add_v2r8(felec,fvdw);
481 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
483 /* Update vectorial force */
484 fix0 = _fjsp_madd_v2r8(dx00,fscal,fix0);
485 fiy0 = _fjsp_madd_v2r8(dy00,fscal,fiy0);
486 fiz0 = _fjsp_madd_v2r8(dz00,fscal,fiz0);
488 fjx0 = _fjsp_madd_v2r8(dx00,fscal,fjx0);
489 fjy0 = _fjsp_madd_v2r8(dy00,fscal,fjy0);
490 fjz0 = _fjsp_madd_v2r8(dz00,fscal,fjz0);
492 /**************************
493 * CALCULATE INTERACTIONS *
494 **************************/
496 r10 = _fjsp_mul_v2r8(rsq10,rinv10);
498 /* Compute parameters for interactions between i and j atoms */
499 qq10 = _fjsp_mul_v2r8(iq1,jq0);
501 /* EWALD ELECTROSTATICS */
503 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
504 ewrt = _fjsp_mul_v2r8(r10,ewtabscale);
505 itab_tmp = _fjsp_dtox_v2r8(ewrt);
506 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
507 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
509 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
510 ewtabD = _fjsp_setzero_v2r8();
511 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
512 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
513 ewtabFn = _fjsp_setzero_v2r8();
514 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
515 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
516 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
517 velec = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
518 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
520 /* Update potential sum for this i atom from the interaction with this j atom. */
521 velec = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
522 velecsum = _fjsp_add_v2r8(velecsum,velec);
526 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
528 /* Update vectorial force */
529 fix1 = _fjsp_madd_v2r8(dx10,fscal,fix1);
530 fiy1 = _fjsp_madd_v2r8(dy10,fscal,fiy1);
531 fiz1 = _fjsp_madd_v2r8(dz10,fscal,fiz1);
533 fjx0 = _fjsp_madd_v2r8(dx10,fscal,fjx0);
534 fjy0 = _fjsp_madd_v2r8(dy10,fscal,fjy0);
535 fjz0 = _fjsp_madd_v2r8(dz10,fscal,fjz0);
537 /**************************
538 * CALCULATE INTERACTIONS *
539 **************************/
541 r20 = _fjsp_mul_v2r8(rsq20,rinv20);
543 /* Compute parameters for interactions between i and j atoms */
544 qq20 = _fjsp_mul_v2r8(iq2,jq0);
546 /* EWALD ELECTROSTATICS */
548 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
549 ewrt = _fjsp_mul_v2r8(r20,ewtabscale);
550 itab_tmp = _fjsp_dtox_v2r8(ewrt);
551 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
552 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
554 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
555 ewtabD = _fjsp_setzero_v2r8();
556 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
557 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
558 ewtabFn = _fjsp_setzero_v2r8();
559 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
560 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
561 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
562 velec = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
563 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
565 /* Update potential sum for this i atom from the interaction with this j atom. */
566 velec = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
567 velecsum = _fjsp_add_v2r8(velecsum,velec);
571 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
573 /* Update vectorial force */
574 fix2 = _fjsp_madd_v2r8(dx20,fscal,fix2);
575 fiy2 = _fjsp_madd_v2r8(dy20,fscal,fiy2);
576 fiz2 = _fjsp_madd_v2r8(dz20,fscal,fiz2);
578 fjx0 = _fjsp_madd_v2r8(dx20,fscal,fjx0);
579 fjy0 = _fjsp_madd_v2r8(dy20,fscal,fjy0);
580 fjz0 = _fjsp_madd_v2r8(dz20,fscal,fjz0);
582 gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
584 /* Inner loop uses 159 flops */
587 /* End of innermost loop */
589 gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
590 f+i_coord_offset,fshift+i_shift_offset);
593 /* Update potential energies */
594 gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
595 gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
597 /* Increment number of inner iterations */
598 inneriter += j_index_end - j_index_start;
600 /* Outer loop uses 20 flops */
603 /* Increment number of outer iterations */
606 /* Update outer/inner flops */
608 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*159);
611 * Gromacs nonbonded kernel: nb_kernel_ElecEw_VdwLJEw_GeomW3P1_F_sparc64_hpc_ace_double
612 * Electrostatics interaction: Ewald
613 * VdW interaction: LJEwald
614 * Geometry: Water3-Particle
615 * Calculate force/pot: Force
618 nb_kernel_ElecEw_VdwLJEw_GeomW3P1_F_sparc64_hpc_ace_double
619 (t_nblist * gmx_restrict nlist,
620 rvec * gmx_restrict xx,
621 rvec * gmx_restrict ff,
622 t_forcerec * gmx_restrict fr,
623 t_mdatoms * gmx_restrict mdatoms,
624 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
625 t_nrnb * gmx_restrict nrnb)
627 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
628 * just 0 for non-waters.
629 * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
630 * jnr indices corresponding to data put in the four positions in the SIMD register.
632 int i_shift_offset,i_coord_offset,outeriter,inneriter;
633 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
635 int j_coord_offsetA,j_coord_offsetB;
636 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
638 real *shiftvec,*fshift,*x,*f;
639 _fjsp_v2r8 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
641 _fjsp_v2r8 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
643 _fjsp_v2r8 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
645 _fjsp_v2r8 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
646 int vdwjidx0A,vdwjidx0B;
647 _fjsp_v2r8 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
648 _fjsp_v2r8 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
649 _fjsp_v2r8 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
650 _fjsp_v2r8 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
651 _fjsp_v2r8 velec,felec,velecsum,facel,crf,krf,krf2;
654 _fjsp_v2r8 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
657 _fjsp_v2r8 one_sixth = gmx_fjsp_set1_v2r8(1.0/6.0);
658 _fjsp_v2r8 one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
659 _fjsp_v2r8 c6grid_00;
660 _fjsp_v2r8 c6grid_10;
661 _fjsp_v2r8 c6grid_20;
663 _fjsp_v2r8 ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
664 _fjsp_v2r8 one_half = gmx_fjsp_set1_v2r8(0.5);
665 _fjsp_v2r8 minus_one = gmx_fjsp_set1_v2r8(-1.0);
666 _fjsp_v2r8 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
669 _fjsp_v2r8 dummy_mask,cutoff_mask;
670 _fjsp_v2r8 one = gmx_fjsp_set1_v2r8(1.0);
671 _fjsp_v2r8 two = gmx_fjsp_set1_v2r8(2.0);
672 union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
679 jindex = nlist->jindex;
681 shiftidx = nlist->shift;
683 shiftvec = fr->shift_vec[0];
684 fshift = fr->fshift[0];
685 facel = gmx_fjsp_set1_v2r8(fr->epsfac);
686 charge = mdatoms->chargeA;
687 nvdwtype = fr->ntype;
689 vdwtype = mdatoms->typeA;
690 vdwgridparam = fr->ljpme_c6grid;
691 sh_lj_ewald = gmx_fjsp_set1_v2r8(fr->ic->sh_lj_ewald);
692 ewclj = gmx_fjsp_set1_v2r8(fr->ewaldcoeff_lj);
693 ewclj2 = _fjsp_mul_v2r8(minus_one,_fjsp_mul_v2r8(ewclj,ewclj));
695 sh_ewald = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
696 ewtab = fr->ic->tabq_coul_F;
697 ewtabscale = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
698 ewtabhalfspace = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
700 /* Setup water-specific parameters */
701 inr = nlist->iinr[0];
702 iq0 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
703 iq1 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
704 iq2 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
705 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
707 /* Avoid stupid compiler warnings */
715 /* Start outer loop over neighborlists */
716 for(iidx=0; iidx<nri; iidx++)
718 /* Load shift vector for this list */
719 i_shift_offset = DIM*shiftidx[iidx];
721 /* Load limits for loop over neighbors */
722 j_index_start = jindex[iidx];
723 j_index_end = jindex[iidx+1];
725 /* Get outer coordinate index */
727 i_coord_offset = DIM*inr;
729 /* Load i particle coords and add shift vector */
730 gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
731 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
733 fix0 = _fjsp_setzero_v2r8();
734 fiy0 = _fjsp_setzero_v2r8();
735 fiz0 = _fjsp_setzero_v2r8();
736 fix1 = _fjsp_setzero_v2r8();
737 fiy1 = _fjsp_setzero_v2r8();
738 fiz1 = _fjsp_setzero_v2r8();
739 fix2 = _fjsp_setzero_v2r8();
740 fiy2 = _fjsp_setzero_v2r8();
741 fiz2 = _fjsp_setzero_v2r8();
743 /* Start inner kernel loop */
744 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
747 /* Get j neighbor index, and coordinate index */
750 j_coord_offsetA = DIM*jnrA;
751 j_coord_offsetB = DIM*jnrB;
753 /* load j atom coordinates */
754 gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
757 /* Calculate displacement vector */
758 dx00 = _fjsp_sub_v2r8(ix0,jx0);
759 dy00 = _fjsp_sub_v2r8(iy0,jy0);
760 dz00 = _fjsp_sub_v2r8(iz0,jz0);
761 dx10 = _fjsp_sub_v2r8(ix1,jx0);
762 dy10 = _fjsp_sub_v2r8(iy1,jy0);
763 dz10 = _fjsp_sub_v2r8(iz1,jz0);
764 dx20 = _fjsp_sub_v2r8(ix2,jx0);
765 dy20 = _fjsp_sub_v2r8(iy2,jy0);
766 dz20 = _fjsp_sub_v2r8(iz2,jz0);
768 /* Calculate squared distance and things based on it */
769 rsq00 = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
770 rsq10 = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
771 rsq20 = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
773 rinv00 = gmx_fjsp_invsqrt_v2r8(rsq00);
774 rinv10 = gmx_fjsp_invsqrt_v2r8(rsq10);
775 rinv20 = gmx_fjsp_invsqrt_v2r8(rsq20);
777 rinvsq00 = _fjsp_mul_v2r8(rinv00,rinv00);
778 rinvsq10 = _fjsp_mul_v2r8(rinv10,rinv10);
779 rinvsq20 = _fjsp_mul_v2r8(rinv20,rinv20);
781 /* Load parameters for j particles */
782 jq0 = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
783 vdwjidx0A = 2*vdwtype[jnrA+0];
784 vdwjidx0B = 2*vdwtype[jnrB+0];
786 fjx0 = _fjsp_setzero_v2r8();
787 fjy0 = _fjsp_setzero_v2r8();
788 fjz0 = _fjsp_setzero_v2r8();
790 /**************************
791 * CALCULATE INTERACTIONS *
792 **************************/
794 r00 = _fjsp_mul_v2r8(rsq00,rinv00);
796 /* Compute parameters for interactions between i and j atoms */
797 qq00 = _fjsp_mul_v2r8(iq0,jq0);
798 gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
799 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
801 c6grid_00 = gmx_fjsp_load_2real_swizzle_v2r8(vdwgridparam+vdwioffset0+vdwjidx0A,
802 vdwgridparam+vdwioffset0+vdwjidx0B);
804 /* EWALD ELECTROSTATICS */
806 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
807 ewrt = _fjsp_mul_v2r8(r00,ewtabscale);
808 itab_tmp = _fjsp_dtox_v2r8(ewrt);
809 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
810 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
812 gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
814 felec = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
815 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
817 /* Analytical LJ-PME */
818 rinvsix = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
819 ewcljrsq = _fjsp_mul_v2r8(ewclj2,rsq00);
820 ewclj6 = _fjsp_mul_v2r8(ewclj2,_fjsp_mul_v2r8(ewclj2,ewclj2));
821 exponent = gmx_simd_exp_d(-ewcljrsq);
822 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
823 poly = _fjsp_mul_v2r8(exponent,_fjsp_madd_v2r8(_fjsp_mul_v2r8(ewcljrsq,ewcljrsq),one_half,_fjsp_sub_v2r8(one,ewcljrsq)));
824 /* f6A = 6 * C6grid * (1 - poly) */
825 f6A = _fjsp_mul_v2r8(c6grid_00,_fjsp_msub_v2r8(one,poly));
826 /* f6B = C6grid * exponent * beta^6 */
827 f6B = _fjsp_mul_v2r8(_fjsp_mul_v2r8(c6grid_00,one_sixth),_fjsp_mul_v2r8(exponent,ewclj6));
828 /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
829 fvdw = _fjsp_mul_v2r8(_fjsp_madd_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,_fjsp_sub_v2r8(c6_00,f6A)),rinvsix,f6B),rinvsq00);
831 fscal = _fjsp_add_v2r8(felec,fvdw);
833 /* Update vectorial force */
834 fix0 = _fjsp_madd_v2r8(dx00,fscal,fix0);
835 fiy0 = _fjsp_madd_v2r8(dy00,fscal,fiy0);
836 fiz0 = _fjsp_madd_v2r8(dz00,fscal,fiz0);
838 fjx0 = _fjsp_madd_v2r8(dx00,fscal,fjx0);
839 fjy0 = _fjsp_madd_v2r8(dy00,fscal,fjy0);
840 fjz0 = _fjsp_madd_v2r8(dz00,fscal,fjz0);
842 /**************************
843 * CALCULATE INTERACTIONS *
844 **************************/
846 r10 = _fjsp_mul_v2r8(rsq10,rinv10);
848 /* Compute parameters for interactions between i and j atoms */
849 qq10 = _fjsp_mul_v2r8(iq1,jq0);
851 /* EWALD ELECTROSTATICS */
853 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
854 ewrt = _fjsp_mul_v2r8(r10,ewtabscale);
855 itab_tmp = _fjsp_dtox_v2r8(ewrt);
856 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
857 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
859 gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
861 felec = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
862 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
866 /* Update vectorial force */
867 fix1 = _fjsp_madd_v2r8(dx10,fscal,fix1);
868 fiy1 = _fjsp_madd_v2r8(dy10,fscal,fiy1);
869 fiz1 = _fjsp_madd_v2r8(dz10,fscal,fiz1);
871 fjx0 = _fjsp_madd_v2r8(dx10,fscal,fjx0);
872 fjy0 = _fjsp_madd_v2r8(dy10,fscal,fjy0);
873 fjz0 = _fjsp_madd_v2r8(dz10,fscal,fjz0);
875 /**************************
876 * CALCULATE INTERACTIONS *
877 **************************/
879 r20 = _fjsp_mul_v2r8(rsq20,rinv20);
881 /* Compute parameters for interactions between i and j atoms */
882 qq20 = _fjsp_mul_v2r8(iq2,jq0);
884 /* EWALD ELECTROSTATICS */
886 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
887 ewrt = _fjsp_mul_v2r8(r20,ewtabscale);
888 itab_tmp = _fjsp_dtox_v2r8(ewrt);
889 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
890 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
892 gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
894 felec = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
895 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
899 /* Update vectorial force */
900 fix2 = _fjsp_madd_v2r8(dx20,fscal,fix2);
901 fiy2 = _fjsp_madd_v2r8(dy20,fscal,fiy2);
902 fiz2 = _fjsp_madd_v2r8(dz20,fscal,fiz2);
904 fjx0 = _fjsp_madd_v2r8(dx20,fscal,fjx0);
905 fjy0 = _fjsp_madd_v2r8(dy20,fscal,fjy0);
906 fjz0 = _fjsp_madd_v2r8(dz20,fscal,fjz0);
908 gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
910 /* Inner loop uses 142 flops */
917 j_coord_offsetA = DIM*jnrA;
919 /* load j atom coordinates */
920 gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
923 /* Calculate displacement vector */
924 dx00 = _fjsp_sub_v2r8(ix0,jx0);
925 dy00 = _fjsp_sub_v2r8(iy0,jy0);
926 dz00 = _fjsp_sub_v2r8(iz0,jz0);
927 dx10 = _fjsp_sub_v2r8(ix1,jx0);
928 dy10 = _fjsp_sub_v2r8(iy1,jy0);
929 dz10 = _fjsp_sub_v2r8(iz1,jz0);
930 dx20 = _fjsp_sub_v2r8(ix2,jx0);
931 dy20 = _fjsp_sub_v2r8(iy2,jy0);
932 dz20 = _fjsp_sub_v2r8(iz2,jz0);
934 /* Calculate squared distance and things based on it */
935 rsq00 = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
936 rsq10 = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
937 rsq20 = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
939 rinv00 = gmx_fjsp_invsqrt_v2r8(rsq00);
940 rinv10 = gmx_fjsp_invsqrt_v2r8(rsq10);
941 rinv20 = gmx_fjsp_invsqrt_v2r8(rsq20);
943 rinvsq00 = _fjsp_mul_v2r8(rinv00,rinv00);
944 rinvsq10 = _fjsp_mul_v2r8(rinv10,rinv10);
945 rinvsq20 = _fjsp_mul_v2r8(rinv20,rinv20);
947 /* Load parameters for j particles */
948 jq0 = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
949 vdwjidx0A = 2*vdwtype[jnrA+0];
951 fjx0 = _fjsp_setzero_v2r8();
952 fjy0 = _fjsp_setzero_v2r8();
953 fjz0 = _fjsp_setzero_v2r8();
955 /**************************
956 * CALCULATE INTERACTIONS *
957 **************************/
959 r00 = _fjsp_mul_v2r8(rsq00,rinv00);
961 /* Compute parameters for interactions between i and j atoms */
962 qq00 = _fjsp_mul_v2r8(iq0,jq0);
963 gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
965 c6grid_00 = gmx_fjsp_load_1real_swizzle_v2r8(vdwgridparam+vdwioffset0+vdwjidx0A);
967 /* EWALD ELECTROSTATICS */
969 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
970 ewrt = _fjsp_mul_v2r8(r00,ewtabscale);
971 itab_tmp = _fjsp_dtox_v2r8(ewrt);
972 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
973 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
975 gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
976 felec = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
977 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
979 /* Analytical LJ-PME */
980 rinvsix = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
981 ewcljrsq = _fjsp_mul_v2r8(ewclj2,rsq00);
982 ewclj6 = _fjsp_mul_v2r8(ewclj2,_fjsp_mul_v2r8(ewclj2,ewclj2));
983 exponent = gmx_simd_exp_d(-ewcljrsq);
984 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
985 poly = _fjsp_mul_v2r8(exponent,_fjsp_madd_v2r8(_fjsp_mul_v2r8(ewcljrsq,ewcljrsq),one_half,_fjsp_sub_v2r8(one,ewcljrsq)));
986 /* f6A = 6 * C6grid * (1 - poly) */
987 f6A = _fjsp_mul_v2r8(c6grid_00,_fjsp_msub_v2r8(one,poly));
988 /* f6B = C6grid * exponent * beta^6 */
989 f6B = _fjsp_mul_v2r8(_fjsp_mul_v2r8(c6grid_00,one_sixth),_fjsp_mul_v2r8(exponent,ewclj6));
990 /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
991 fvdw = _fjsp_mul_v2r8(_fjsp_madd_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,_fjsp_sub_v2r8(c6_00,f6A)),rinvsix,f6B),rinvsq00);
993 fscal = _fjsp_add_v2r8(felec,fvdw);
995 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
997 /* Update vectorial force */
998 fix0 = _fjsp_madd_v2r8(dx00,fscal,fix0);
999 fiy0 = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1000 fiz0 = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1002 fjx0 = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1003 fjy0 = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1004 fjz0 = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1006 /**************************
1007 * CALCULATE INTERACTIONS *
1008 **************************/
1010 r10 = _fjsp_mul_v2r8(rsq10,rinv10);
1012 /* Compute parameters for interactions between i and j atoms */
1013 qq10 = _fjsp_mul_v2r8(iq1,jq0);
1015 /* EWALD ELECTROSTATICS */
1017 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1018 ewrt = _fjsp_mul_v2r8(r10,ewtabscale);
1019 itab_tmp = _fjsp_dtox_v2r8(ewrt);
1020 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1021 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1023 gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
1024 felec = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1025 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
1029 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1031 /* Update vectorial force */
1032 fix1 = _fjsp_madd_v2r8(dx10,fscal,fix1);
1033 fiy1 = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1034 fiz1 = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1036 fjx0 = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1037 fjy0 = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1038 fjz0 = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1040 /**************************
1041 * CALCULATE INTERACTIONS *
1042 **************************/
1044 r20 = _fjsp_mul_v2r8(rsq20,rinv20);
1046 /* Compute parameters for interactions between i and j atoms */
1047 qq20 = _fjsp_mul_v2r8(iq2,jq0);
1049 /* EWALD ELECTROSTATICS */
1051 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1052 ewrt = _fjsp_mul_v2r8(r20,ewtabscale);
1053 itab_tmp = _fjsp_dtox_v2r8(ewrt);
1054 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1055 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1057 gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
1058 felec = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1059 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
1063 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1065 /* Update vectorial force */
1066 fix2 = _fjsp_madd_v2r8(dx20,fscal,fix2);
1067 fiy2 = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1068 fiz2 = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1070 fjx0 = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1071 fjy0 = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1072 fjz0 = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1074 gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1076 /* Inner loop uses 142 flops */
1079 /* End of innermost loop */
1081 gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1082 f+i_coord_offset,fshift+i_shift_offset);
1084 /* Increment number of inner iterations */
1085 inneriter += j_index_end - j_index_start;
1087 /* Outer loop uses 18 flops */
1090 /* Increment number of outer iterations */
1093 /* Update outer/inner flops */
1095 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_F,outeriter*18 + inneriter*142);