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_ElecEwSh_VdwLJSh_GeomW3P1_VF_sparc64_hpc_ace_double
53 * Electrostatics interaction: Ewald
54 * VdW interaction: LennardJones
55 * Geometry: Water3-Particle
56 * Calculate force/pot: PotentialAndForce
59 nb_kernel_ElecEwSh_VdwLJSh_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 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
103 _fjsp_v2r8 dummy_mask,cutoff_mask;
104 _fjsp_v2r8 one = gmx_fjsp_set1_v2r8(1.0);
105 _fjsp_v2r8 two = gmx_fjsp_set1_v2r8(2.0);
106 union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
113 jindex = nlist->jindex;
115 shiftidx = nlist->shift;
117 shiftvec = fr->shift_vec[0];
118 fshift = fr->fshift[0];
119 facel = gmx_fjsp_set1_v2r8(fr->epsfac);
120 charge = mdatoms->chargeA;
121 nvdwtype = fr->ntype;
123 vdwtype = mdatoms->typeA;
125 sh_ewald = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
126 ewtab = fr->ic->tabq_coul_FDV0;
127 ewtabscale = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
128 ewtabhalfspace = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
130 /* Setup water-specific parameters */
131 inr = nlist->iinr[0];
132 iq0 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
133 iq1 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
134 iq2 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
135 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
137 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
138 rcutoff_scalar = fr->rcoulomb;
139 rcutoff = gmx_fjsp_set1_v2r8(rcutoff_scalar);
140 rcutoff2 = _fjsp_mul_v2r8(rcutoff,rcutoff);
142 sh_vdw_invrcut6 = gmx_fjsp_set1_v2r8(fr->ic->sh_invrc6);
143 rvdw = gmx_fjsp_set1_v2r8(fr->rvdw);
145 /* Avoid stupid compiler warnings */
153 /* Start outer loop over neighborlists */
154 for(iidx=0; iidx<nri; iidx++)
156 /* Load shift vector for this list */
157 i_shift_offset = DIM*shiftidx[iidx];
159 /* Load limits for loop over neighbors */
160 j_index_start = jindex[iidx];
161 j_index_end = jindex[iidx+1];
163 /* Get outer coordinate index */
165 i_coord_offset = DIM*inr;
167 /* Load i particle coords and add shift vector */
168 gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
169 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
171 fix0 = _fjsp_setzero_v2r8();
172 fiy0 = _fjsp_setzero_v2r8();
173 fiz0 = _fjsp_setzero_v2r8();
174 fix1 = _fjsp_setzero_v2r8();
175 fiy1 = _fjsp_setzero_v2r8();
176 fiz1 = _fjsp_setzero_v2r8();
177 fix2 = _fjsp_setzero_v2r8();
178 fiy2 = _fjsp_setzero_v2r8();
179 fiz2 = _fjsp_setzero_v2r8();
181 /* Reset potential sums */
182 velecsum = _fjsp_setzero_v2r8();
183 vvdwsum = _fjsp_setzero_v2r8();
185 /* Start inner kernel loop */
186 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
189 /* Get j neighbor index, and coordinate index */
192 j_coord_offsetA = DIM*jnrA;
193 j_coord_offsetB = DIM*jnrB;
195 /* load j atom coordinates */
196 gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
199 /* Calculate displacement vector */
200 dx00 = _fjsp_sub_v2r8(ix0,jx0);
201 dy00 = _fjsp_sub_v2r8(iy0,jy0);
202 dz00 = _fjsp_sub_v2r8(iz0,jz0);
203 dx10 = _fjsp_sub_v2r8(ix1,jx0);
204 dy10 = _fjsp_sub_v2r8(iy1,jy0);
205 dz10 = _fjsp_sub_v2r8(iz1,jz0);
206 dx20 = _fjsp_sub_v2r8(ix2,jx0);
207 dy20 = _fjsp_sub_v2r8(iy2,jy0);
208 dz20 = _fjsp_sub_v2r8(iz2,jz0);
210 /* Calculate squared distance and things based on it */
211 rsq00 = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
212 rsq10 = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
213 rsq20 = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
215 rinv00 = gmx_fjsp_invsqrt_v2r8(rsq00);
216 rinv10 = gmx_fjsp_invsqrt_v2r8(rsq10);
217 rinv20 = gmx_fjsp_invsqrt_v2r8(rsq20);
219 rinvsq00 = _fjsp_mul_v2r8(rinv00,rinv00);
220 rinvsq10 = _fjsp_mul_v2r8(rinv10,rinv10);
221 rinvsq20 = _fjsp_mul_v2r8(rinv20,rinv20);
223 /* Load parameters for j particles */
224 jq0 = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
225 vdwjidx0A = 2*vdwtype[jnrA+0];
226 vdwjidx0B = 2*vdwtype[jnrB+0];
228 fjx0 = _fjsp_setzero_v2r8();
229 fjy0 = _fjsp_setzero_v2r8();
230 fjz0 = _fjsp_setzero_v2r8();
232 /**************************
233 * CALCULATE INTERACTIONS *
234 **************************/
236 if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
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 /* EWALD ELECTROSTATICS */
248 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
249 ewrt = _fjsp_mul_v2r8(r00,ewtabscale);
250 itab_tmp = _fjsp_dtox_v2r8(ewrt);
251 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
252 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
254 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
255 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
256 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
257 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
258 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
259 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
260 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
261 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
262 velec = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv00,sh_ewald),velec));
263 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
265 /* LENNARD-JONES DISPERSION/REPULSION */
267 rinvsix = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
268 vvdw6 = _fjsp_mul_v2r8(c6_00,rinvsix);
269 vvdw12 = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
270 vvdw = _fjsp_msub_v2r8(_fjsp_nmsub_v2r8(c12_00,_fjsp_mul_v2r8(sh_vdw_invrcut6,sh_vdw_invrcut6),vvdw12),one_twelfth,
271 _fjsp_mul_v2r8(_fjsp_nmsub_v2r8( c6_00,sh_vdw_invrcut6,vvdw6),one_sixth));
272 fvdw = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
274 cutoff_mask = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
276 /* Update potential sum for this i atom from the interaction with this j atom. */
277 velec = _fjsp_and_v2r8(velec,cutoff_mask);
278 velecsum = _fjsp_add_v2r8(velecsum,velec);
279 vvdw = _fjsp_and_v2r8(vvdw,cutoff_mask);
280 vvdwsum = _fjsp_add_v2r8(vvdwsum,vvdw);
282 fscal = _fjsp_add_v2r8(felec,fvdw);
284 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
286 /* Update vectorial force */
287 fix0 = _fjsp_madd_v2r8(dx00,fscal,fix0);
288 fiy0 = _fjsp_madd_v2r8(dy00,fscal,fiy0);
289 fiz0 = _fjsp_madd_v2r8(dz00,fscal,fiz0);
291 fjx0 = _fjsp_madd_v2r8(dx00,fscal,fjx0);
292 fjy0 = _fjsp_madd_v2r8(dy00,fscal,fjy0);
293 fjz0 = _fjsp_madd_v2r8(dz00,fscal,fjz0);
297 /**************************
298 * CALCULATE INTERACTIONS *
299 **************************/
301 if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
304 r10 = _fjsp_mul_v2r8(rsq10,rinv10);
306 /* Compute parameters for interactions between i and j atoms */
307 qq10 = _fjsp_mul_v2r8(iq1,jq0);
309 /* EWALD ELECTROSTATICS */
311 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
312 ewrt = _fjsp_mul_v2r8(r10,ewtabscale);
313 itab_tmp = _fjsp_dtox_v2r8(ewrt);
314 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
315 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
317 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
318 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
319 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
320 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
321 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
322 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
323 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
324 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
325 velec = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv10,sh_ewald),velec));
326 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
328 cutoff_mask = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
330 /* Update potential sum for this i atom from the interaction with this j atom. */
331 velec = _fjsp_and_v2r8(velec,cutoff_mask);
332 velecsum = _fjsp_add_v2r8(velecsum,velec);
336 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
338 /* Update vectorial force */
339 fix1 = _fjsp_madd_v2r8(dx10,fscal,fix1);
340 fiy1 = _fjsp_madd_v2r8(dy10,fscal,fiy1);
341 fiz1 = _fjsp_madd_v2r8(dz10,fscal,fiz1);
343 fjx0 = _fjsp_madd_v2r8(dx10,fscal,fjx0);
344 fjy0 = _fjsp_madd_v2r8(dy10,fscal,fjy0);
345 fjz0 = _fjsp_madd_v2r8(dz10,fscal,fjz0);
349 /**************************
350 * CALCULATE INTERACTIONS *
351 **************************/
353 if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
356 r20 = _fjsp_mul_v2r8(rsq20,rinv20);
358 /* Compute parameters for interactions between i and j atoms */
359 qq20 = _fjsp_mul_v2r8(iq2,jq0);
361 /* EWALD ELECTROSTATICS */
363 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
364 ewrt = _fjsp_mul_v2r8(r20,ewtabscale);
365 itab_tmp = _fjsp_dtox_v2r8(ewrt);
366 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
367 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
369 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
370 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
371 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
372 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
373 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
374 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
375 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
376 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
377 velec = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv20,sh_ewald),velec));
378 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
380 cutoff_mask = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
382 /* Update potential sum for this i atom from the interaction with this j atom. */
383 velec = _fjsp_and_v2r8(velec,cutoff_mask);
384 velecsum = _fjsp_add_v2r8(velecsum,velec);
388 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
390 /* Update vectorial force */
391 fix2 = _fjsp_madd_v2r8(dx20,fscal,fix2);
392 fiy2 = _fjsp_madd_v2r8(dy20,fscal,fiy2);
393 fiz2 = _fjsp_madd_v2r8(dz20,fscal,fiz2);
395 fjx0 = _fjsp_madd_v2r8(dx20,fscal,fjx0);
396 fjy0 = _fjsp_madd_v2r8(dy20,fscal,fjy0);
397 fjz0 = _fjsp_madd_v2r8(dz20,fscal,fjz0);
401 gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
403 /* Inner loop uses 168 flops */
410 j_coord_offsetA = DIM*jnrA;
412 /* load j atom coordinates */
413 gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
416 /* Calculate displacement vector */
417 dx00 = _fjsp_sub_v2r8(ix0,jx0);
418 dy00 = _fjsp_sub_v2r8(iy0,jy0);
419 dz00 = _fjsp_sub_v2r8(iz0,jz0);
420 dx10 = _fjsp_sub_v2r8(ix1,jx0);
421 dy10 = _fjsp_sub_v2r8(iy1,jy0);
422 dz10 = _fjsp_sub_v2r8(iz1,jz0);
423 dx20 = _fjsp_sub_v2r8(ix2,jx0);
424 dy20 = _fjsp_sub_v2r8(iy2,jy0);
425 dz20 = _fjsp_sub_v2r8(iz2,jz0);
427 /* Calculate squared distance and things based on it */
428 rsq00 = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
429 rsq10 = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
430 rsq20 = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
432 rinv00 = gmx_fjsp_invsqrt_v2r8(rsq00);
433 rinv10 = gmx_fjsp_invsqrt_v2r8(rsq10);
434 rinv20 = gmx_fjsp_invsqrt_v2r8(rsq20);
436 rinvsq00 = _fjsp_mul_v2r8(rinv00,rinv00);
437 rinvsq10 = _fjsp_mul_v2r8(rinv10,rinv10);
438 rinvsq20 = _fjsp_mul_v2r8(rinv20,rinv20);
440 /* Load parameters for j particles */
441 jq0 = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
442 vdwjidx0A = 2*vdwtype[jnrA+0];
444 fjx0 = _fjsp_setzero_v2r8();
445 fjy0 = _fjsp_setzero_v2r8();
446 fjz0 = _fjsp_setzero_v2r8();
448 /**************************
449 * CALCULATE INTERACTIONS *
450 **************************/
452 if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
455 r00 = _fjsp_mul_v2r8(rsq00,rinv00);
457 /* Compute parameters for interactions between i and j atoms */
458 qq00 = _fjsp_mul_v2r8(iq0,jq0);
459 gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
460 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
462 /* EWALD ELECTROSTATICS */
464 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
465 ewrt = _fjsp_mul_v2r8(r00,ewtabscale);
466 itab_tmp = _fjsp_dtox_v2r8(ewrt);
467 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
468 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
470 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
471 ewtabD = _fjsp_setzero_v2r8();
472 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
473 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
474 ewtabFn = _fjsp_setzero_v2r8();
475 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
476 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
477 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
478 velec = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv00,sh_ewald),velec));
479 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
481 /* LENNARD-JONES DISPERSION/REPULSION */
483 rinvsix = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
484 vvdw6 = _fjsp_mul_v2r8(c6_00,rinvsix);
485 vvdw12 = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
486 vvdw = _fjsp_msub_v2r8(_fjsp_nmsub_v2r8(c12_00,_fjsp_mul_v2r8(sh_vdw_invrcut6,sh_vdw_invrcut6),vvdw12),one_twelfth,
487 _fjsp_mul_v2r8(_fjsp_nmsub_v2r8( c6_00,sh_vdw_invrcut6,vvdw6),one_sixth));
488 fvdw = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
490 cutoff_mask = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
492 /* Update potential sum for this i atom from the interaction with this j atom. */
493 velec = _fjsp_and_v2r8(velec,cutoff_mask);
494 velec = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
495 velecsum = _fjsp_add_v2r8(velecsum,velec);
496 vvdw = _fjsp_and_v2r8(vvdw,cutoff_mask);
497 vvdw = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
498 vvdwsum = _fjsp_add_v2r8(vvdwsum,vvdw);
500 fscal = _fjsp_add_v2r8(felec,fvdw);
502 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
504 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
506 /* Update vectorial force */
507 fix0 = _fjsp_madd_v2r8(dx00,fscal,fix0);
508 fiy0 = _fjsp_madd_v2r8(dy00,fscal,fiy0);
509 fiz0 = _fjsp_madd_v2r8(dz00,fscal,fiz0);
511 fjx0 = _fjsp_madd_v2r8(dx00,fscal,fjx0);
512 fjy0 = _fjsp_madd_v2r8(dy00,fscal,fjy0);
513 fjz0 = _fjsp_madd_v2r8(dz00,fscal,fjz0);
517 /**************************
518 * CALCULATE INTERACTIONS *
519 **************************/
521 if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
524 r10 = _fjsp_mul_v2r8(rsq10,rinv10);
526 /* Compute parameters for interactions between i and j atoms */
527 qq10 = _fjsp_mul_v2r8(iq1,jq0);
529 /* EWALD ELECTROSTATICS */
531 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
532 ewrt = _fjsp_mul_v2r8(r10,ewtabscale);
533 itab_tmp = _fjsp_dtox_v2r8(ewrt);
534 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
535 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
537 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
538 ewtabD = _fjsp_setzero_v2r8();
539 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
540 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
541 ewtabFn = _fjsp_setzero_v2r8();
542 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
543 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
544 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
545 velec = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv10,sh_ewald),velec));
546 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
548 cutoff_mask = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
550 /* Update potential sum for this i atom from the interaction with this j atom. */
551 velec = _fjsp_and_v2r8(velec,cutoff_mask);
552 velec = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
553 velecsum = _fjsp_add_v2r8(velecsum,velec);
557 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
559 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
561 /* Update vectorial force */
562 fix1 = _fjsp_madd_v2r8(dx10,fscal,fix1);
563 fiy1 = _fjsp_madd_v2r8(dy10,fscal,fiy1);
564 fiz1 = _fjsp_madd_v2r8(dz10,fscal,fiz1);
566 fjx0 = _fjsp_madd_v2r8(dx10,fscal,fjx0);
567 fjy0 = _fjsp_madd_v2r8(dy10,fscal,fjy0);
568 fjz0 = _fjsp_madd_v2r8(dz10,fscal,fjz0);
572 /**************************
573 * CALCULATE INTERACTIONS *
574 **************************/
576 if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
579 r20 = _fjsp_mul_v2r8(rsq20,rinv20);
581 /* Compute parameters for interactions between i and j atoms */
582 qq20 = _fjsp_mul_v2r8(iq2,jq0);
584 /* EWALD ELECTROSTATICS */
586 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
587 ewrt = _fjsp_mul_v2r8(r20,ewtabscale);
588 itab_tmp = _fjsp_dtox_v2r8(ewrt);
589 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
590 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
592 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
593 ewtabD = _fjsp_setzero_v2r8();
594 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
595 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
596 ewtabFn = _fjsp_setzero_v2r8();
597 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
598 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
599 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
600 velec = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv20,sh_ewald),velec));
601 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
603 cutoff_mask = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
605 /* Update potential sum for this i atom from the interaction with this j atom. */
606 velec = _fjsp_and_v2r8(velec,cutoff_mask);
607 velec = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
608 velecsum = _fjsp_add_v2r8(velecsum,velec);
612 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
614 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
616 /* Update vectorial force */
617 fix2 = _fjsp_madd_v2r8(dx20,fscal,fix2);
618 fiy2 = _fjsp_madd_v2r8(dy20,fscal,fiy2);
619 fiz2 = _fjsp_madd_v2r8(dz20,fscal,fiz2);
621 fjx0 = _fjsp_madd_v2r8(dx20,fscal,fjx0);
622 fjy0 = _fjsp_madd_v2r8(dy20,fscal,fjy0);
623 fjz0 = _fjsp_madd_v2r8(dz20,fscal,fjz0);
627 gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
629 /* Inner loop uses 168 flops */
632 /* End of innermost loop */
634 gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
635 f+i_coord_offset,fshift+i_shift_offset);
638 /* Update potential energies */
639 gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
640 gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
642 /* Increment number of inner iterations */
643 inneriter += j_index_end - j_index_start;
645 /* Outer loop uses 20 flops */
648 /* Increment number of outer iterations */
651 /* Update outer/inner flops */
653 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*168);
656 * Gromacs nonbonded kernel: nb_kernel_ElecEwSh_VdwLJSh_GeomW3P1_F_sparc64_hpc_ace_double
657 * Electrostatics interaction: Ewald
658 * VdW interaction: LennardJones
659 * Geometry: Water3-Particle
660 * Calculate force/pot: Force
663 nb_kernel_ElecEwSh_VdwLJSh_GeomW3P1_F_sparc64_hpc_ace_double
664 (t_nblist * gmx_restrict nlist,
665 rvec * gmx_restrict xx,
666 rvec * gmx_restrict ff,
667 t_forcerec * gmx_restrict fr,
668 t_mdatoms * gmx_restrict mdatoms,
669 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
670 t_nrnb * gmx_restrict nrnb)
672 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
673 * just 0 for non-waters.
674 * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
675 * jnr indices corresponding to data put in the four positions in the SIMD register.
677 int i_shift_offset,i_coord_offset,outeriter,inneriter;
678 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
680 int j_coord_offsetA,j_coord_offsetB;
681 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
683 real *shiftvec,*fshift,*x,*f;
684 _fjsp_v2r8 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
686 _fjsp_v2r8 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
688 _fjsp_v2r8 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
690 _fjsp_v2r8 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
691 int vdwjidx0A,vdwjidx0B;
692 _fjsp_v2r8 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
693 _fjsp_v2r8 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
694 _fjsp_v2r8 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
695 _fjsp_v2r8 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
696 _fjsp_v2r8 velec,felec,velecsum,facel,crf,krf,krf2;
699 _fjsp_v2r8 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
702 _fjsp_v2r8 one_sixth = gmx_fjsp_set1_v2r8(1.0/6.0);
703 _fjsp_v2r8 one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
704 _fjsp_v2r8 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
707 _fjsp_v2r8 dummy_mask,cutoff_mask;
708 _fjsp_v2r8 one = gmx_fjsp_set1_v2r8(1.0);
709 _fjsp_v2r8 two = gmx_fjsp_set1_v2r8(2.0);
710 union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
717 jindex = nlist->jindex;
719 shiftidx = nlist->shift;
721 shiftvec = fr->shift_vec[0];
722 fshift = fr->fshift[0];
723 facel = gmx_fjsp_set1_v2r8(fr->epsfac);
724 charge = mdatoms->chargeA;
725 nvdwtype = fr->ntype;
727 vdwtype = mdatoms->typeA;
729 sh_ewald = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
730 ewtab = fr->ic->tabq_coul_F;
731 ewtabscale = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
732 ewtabhalfspace = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
734 /* Setup water-specific parameters */
735 inr = nlist->iinr[0];
736 iq0 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
737 iq1 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
738 iq2 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
739 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
741 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
742 rcutoff_scalar = fr->rcoulomb;
743 rcutoff = gmx_fjsp_set1_v2r8(rcutoff_scalar);
744 rcutoff2 = _fjsp_mul_v2r8(rcutoff,rcutoff);
746 sh_vdw_invrcut6 = gmx_fjsp_set1_v2r8(fr->ic->sh_invrc6);
747 rvdw = gmx_fjsp_set1_v2r8(fr->rvdw);
749 /* Avoid stupid compiler warnings */
757 /* Start outer loop over neighborlists */
758 for(iidx=0; iidx<nri; iidx++)
760 /* Load shift vector for this list */
761 i_shift_offset = DIM*shiftidx[iidx];
763 /* Load limits for loop over neighbors */
764 j_index_start = jindex[iidx];
765 j_index_end = jindex[iidx+1];
767 /* Get outer coordinate index */
769 i_coord_offset = DIM*inr;
771 /* Load i particle coords and add shift vector */
772 gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
773 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
775 fix0 = _fjsp_setzero_v2r8();
776 fiy0 = _fjsp_setzero_v2r8();
777 fiz0 = _fjsp_setzero_v2r8();
778 fix1 = _fjsp_setzero_v2r8();
779 fiy1 = _fjsp_setzero_v2r8();
780 fiz1 = _fjsp_setzero_v2r8();
781 fix2 = _fjsp_setzero_v2r8();
782 fiy2 = _fjsp_setzero_v2r8();
783 fiz2 = _fjsp_setzero_v2r8();
785 /* Start inner kernel loop */
786 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
789 /* Get j neighbor index, and coordinate index */
792 j_coord_offsetA = DIM*jnrA;
793 j_coord_offsetB = DIM*jnrB;
795 /* load j atom coordinates */
796 gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
799 /* Calculate displacement vector */
800 dx00 = _fjsp_sub_v2r8(ix0,jx0);
801 dy00 = _fjsp_sub_v2r8(iy0,jy0);
802 dz00 = _fjsp_sub_v2r8(iz0,jz0);
803 dx10 = _fjsp_sub_v2r8(ix1,jx0);
804 dy10 = _fjsp_sub_v2r8(iy1,jy0);
805 dz10 = _fjsp_sub_v2r8(iz1,jz0);
806 dx20 = _fjsp_sub_v2r8(ix2,jx0);
807 dy20 = _fjsp_sub_v2r8(iy2,jy0);
808 dz20 = _fjsp_sub_v2r8(iz2,jz0);
810 /* Calculate squared distance and things based on it */
811 rsq00 = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
812 rsq10 = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
813 rsq20 = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
815 rinv00 = gmx_fjsp_invsqrt_v2r8(rsq00);
816 rinv10 = gmx_fjsp_invsqrt_v2r8(rsq10);
817 rinv20 = gmx_fjsp_invsqrt_v2r8(rsq20);
819 rinvsq00 = _fjsp_mul_v2r8(rinv00,rinv00);
820 rinvsq10 = _fjsp_mul_v2r8(rinv10,rinv10);
821 rinvsq20 = _fjsp_mul_v2r8(rinv20,rinv20);
823 /* Load parameters for j particles */
824 jq0 = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
825 vdwjidx0A = 2*vdwtype[jnrA+0];
826 vdwjidx0B = 2*vdwtype[jnrB+0];
828 fjx0 = _fjsp_setzero_v2r8();
829 fjy0 = _fjsp_setzero_v2r8();
830 fjz0 = _fjsp_setzero_v2r8();
832 /**************************
833 * CALCULATE INTERACTIONS *
834 **************************/
836 if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
839 r00 = _fjsp_mul_v2r8(rsq00,rinv00);
841 /* Compute parameters for interactions between i and j atoms */
842 qq00 = _fjsp_mul_v2r8(iq0,jq0);
843 gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
844 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
846 /* EWALD ELECTROSTATICS */
848 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
849 ewrt = _fjsp_mul_v2r8(r00,ewtabscale);
850 itab_tmp = _fjsp_dtox_v2r8(ewrt);
851 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
852 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
854 gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
856 felec = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
857 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
859 /* LENNARD-JONES DISPERSION/REPULSION */
861 rinvsix = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
862 fvdw = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
864 cutoff_mask = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
866 fscal = _fjsp_add_v2r8(felec,fvdw);
868 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
870 /* Update vectorial force */
871 fix0 = _fjsp_madd_v2r8(dx00,fscal,fix0);
872 fiy0 = _fjsp_madd_v2r8(dy00,fscal,fiy0);
873 fiz0 = _fjsp_madd_v2r8(dz00,fscal,fiz0);
875 fjx0 = _fjsp_madd_v2r8(dx00,fscal,fjx0);
876 fjy0 = _fjsp_madd_v2r8(dy00,fscal,fjy0);
877 fjz0 = _fjsp_madd_v2r8(dz00,fscal,fjz0);
881 /**************************
882 * CALCULATE INTERACTIONS *
883 **************************/
885 if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
888 r10 = _fjsp_mul_v2r8(rsq10,rinv10);
890 /* Compute parameters for interactions between i and j atoms */
891 qq10 = _fjsp_mul_v2r8(iq1,jq0);
893 /* EWALD ELECTROSTATICS */
895 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
896 ewrt = _fjsp_mul_v2r8(r10,ewtabscale);
897 itab_tmp = _fjsp_dtox_v2r8(ewrt);
898 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
899 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
901 gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
903 felec = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
904 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
906 cutoff_mask = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
910 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
912 /* Update vectorial force */
913 fix1 = _fjsp_madd_v2r8(dx10,fscal,fix1);
914 fiy1 = _fjsp_madd_v2r8(dy10,fscal,fiy1);
915 fiz1 = _fjsp_madd_v2r8(dz10,fscal,fiz1);
917 fjx0 = _fjsp_madd_v2r8(dx10,fscal,fjx0);
918 fjy0 = _fjsp_madd_v2r8(dy10,fscal,fjy0);
919 fjz0 = _fjsp_madd_v2r8(dz10,fscal,fjz0);
923 /**************************
924 * CALCULATE INTERACTIONS *
925 **************************/
927 if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
930 r20 = _fjsp_mul_v2r8(rsq20,rinv20);
932 /* Compute parameters for interactions between i and j atoms */
933 qq20 = _fjsp_mul_v2r8(iq2,jq0);
935 /* EWALD ELECTROSTATICS */
937 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
938 ewrt = _fjsp_mul_v2r8(r20,ewtabscale);
939 itab_tmp = _fjsp_dtox_v2r8(ewrt);
940 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
941 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
943 gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
945 felec = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
946 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
948 cutoff_mask = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
952 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
954 /* Update vectorial force */
955 fix2 = _fjsp_madd_v2r8(dx20,fscal,fix2);
956 fiy2 = _fjsp_madd_v2r8(dy20,fscal,fiy2);
957 fiz2 = _fjsp_madd_v2r8(dz20,fscal,fiz2);
959 fjx0 = _fjsp_madd_v2r8(dx20,fscal,fjx0);
960 fjy0 = _fjsp_madd_v2r8(dy20,fscal,fjy0);
961 fjz0 = _fjsp_madd_v2r8(dz20,fscal,fjz0);
965 gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
967 /* Inner loop uses 136 flops */
974 j_coord_offsetA = DIM*jnrA;
976 /* load j atom coordinates */
977 gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
980 /* Calculate displacement vector */
981 dx00 = _fjsp_sub_v2r8(ix0,jx0);
982 dy00 = _fjsp_sub_v2r8(iy0,jy0);
983 dz00 = _fjsp_sub_v2r8(iz0,jz0);
984 dx10 = _fjsp_sub_v2r8(ix1,jx0);
985 dy10 = _fjsp_sub_v2r8(iy1,jy0);
986 dz10 = _fjsp_sub_v2r8(iz1,jz0);
987 dx20 = _fjsp_sub_v2r8(ix2,jx0);
988 dy20 = _fjsp_sub_v2r8(iy2,jy0);
989 dz20 = _fjsp_sub_v2r8(iz2,jz0);
991 /* Calculate squared distance and things based on it */
992 rsq00 = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
993 rsq10 = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
994 rsq20 = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
996 rinv00 = gmx_fjsp_invsqrt_v2r8(rsq00);
997 rinv10 = gmx_fjsp_invsqrt_v2r8(rsq10);
998 rinv20 = gmx_fjsp_invsqrt_v2r8(rsq20);
1000 rinvsq00 = _fjsp_mul_v2r8(rinv00,rinv00);
1001 rinvsq10 = _fjsp_mul_v2r8(rinv10,rinv10);
1002 rinvsq20 = _fjsp_mul_v2r8(rinv20,rinv20);
1004 /* Load parameters for j particles */
1005 jq0 = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
1006 vdwjidx0A = 2*vdwtype[jnrA+0];
1008 fjx0 = _fjsp_setzero_v2r8();
1009 fjy0 = _fjsp_setzero_v2r8();
1010 fjz0 = _fjsp_setzero_v2r8();
1012 /**************************
1013 * CALCULATE INTERACTIONS *
1014 **************************/
1016 if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
1019 r00 = _fjsp_mul_v2r8(rsq00,rinv00);
1021 /* Compute parameters for interactions between i and j atoms */
1022 qq00 = _fjsp_mul_v2r8(iq0,jq0);
1023 gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
1024 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
1026 /* EWALD ELECTROSTATICS */
1028 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1029 ewrt = _fjsp_mul_v2r8(r00,ewtabscale);
1030 itab_tmp = _fjsp_dtox_v2r8(ewrt);
1031 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1032 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1034 gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
1035 felec = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1036 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
1038 /* LENNARD-JONES DISPERSION/REPULSION */
1040 rinvsix = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
1041 fvdw = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
1043 cutoff_mask = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
1045 fscal = _fjsp_add_v2r8(felec,fvdw);
1047 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
1049 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1051 /* Update vectorial force */
1052 fix0 = _fjsp_madd_v2r8(dx00,fscal,fix0);
1053 fiy0 = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1054 fiz0 = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1056 fjx0 = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1057 fjy0 = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1058 fjz0 = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1062 /**************************
1063 * CALCULATE INTERACTIONS *
1064 **************************/
1066 if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
1069 r10 = _fjsp_mul_v2r8(rsq10,rinv10);
1071 /* Compute parameters for interactions between i and j atoms */
1072 qq10 = _fjsp_mul_v2r8(iq1,jq0);
1074 /* EWALD ELECTROSTATICS */
1076 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1077 ewrt = _fjsp_mul_v2r8(r10,ewtabscale);
1078 itab_tmp = _fjsp_dtox_v2r8(ewrt);
1079 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1080 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1082 gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
1083 felec = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1084 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
1086 cutoff_mask = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
1090 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
1092 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1094 /* Update vectorial force */
1095 fix1 = _fjsp_madd_v2r8(dx10,fscal,fix1);
1096 fiy1 = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1097 fiz1 = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1099 fjx0 = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1100 fjy0 = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1101 fjz0 = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1105 /**************************
1106 * CALCULATE INTERACTIONS *
1107 **************************/
1109 if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
1112 r20 = _fjsp_mul_v2r8(rsq20,rinv20);
1114 /* Compute parameters for interactions between i and j atoms */
1115 qq20 = _fjsp_mul_v2r8(iq2,jq0);
1117 /* EWALD ELECTROSTATICS */
1119 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1120 ewrt = _fjsp_mul_v2r8(r20,ewtabscale);
1121 itab_tmp = _fjsp_dtox_v2r8(ewrt);
1122 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1123 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1125 gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
1126 felec = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1127 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
1129 cutoff_mask = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
1133 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
1135 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1137 /* Update vectorial force */
1138 fix2 = _fjsp_madd_v2r8(dx20,fscal,fix2);
1139 fiy2 = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1140 fiz2 = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1142 fjx0 = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1143 fjy0 = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1144 fjz0 = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1148 gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1150 /* Inner loop uses 136 flops */
1153 /* End of innermost loop */
1155 gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1156 f+i_coord_offset,fshift+i_shift_offset);
1158 /* Increment number of inner iterations */
1159 inneriter += j_index_end - j_index_start;
1161 /* Outer loop uses 18 flops */
1164 /* Increment number of outer iterations */
1167 /* Update outer/inner flops */
1169 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_F,outeriter*18 + inneriter*136);