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_GeomW4P1_VF_sparc64_hpc_ace_double
53 * Electrostatics interaction: Ewald
54 * VdW interaction: LJEwald
55 * Geometry: Water4-Particle
56 * Calculate force/pot: PotentialAndForce
59 nb_kernel_ElecEw_VdwLJEw_GeomW4P1_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;
88 _fjsp_v2r8 ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
89 int vdwjidx0A,vdwjidx0B;
90 _fjsp_v2r8 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
91 _fjsp_v2r8 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
92 _fjsp_v2r8 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
93 _fjsp_v2r8 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
94 _fjsp_v2r8 dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
95 _fjsp_v2r8 velec,felec,velecsum,facel,crf,krf,krf2;
98 _fjsp_v2r8 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
101 _fjsp_v2r8 one_sixth = gmx_fjsp_set1_v2r8(1.0/6.0);
102 _fjsp_v2r8 one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
103 _fjsp_v2r8 c6grid_00;
104 _fjsp_v2r8 c6grid_10;
105 _fjsp_v2r8 c6grid_20;
106 _fjsp_v2r8 c6grid_30;
108 _fjsp_v2r8 ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
109 _fjsp_v2r8 one_half = gmx_fjsp_set1_v2r8(0.5);
110 _fjsp_v2r8 minus_one = gmx_fjsp_set1_v2r8(-1.0);
111 _fjsp_v2r8 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
114 _fjsp_v2r8 dummy_mask,cutoff_mask;
115 _fjsp_v2r8 one = gmx_fjsp_set1_v2r8(1.0);
116 _fjsp_v2r8 two = gmx_fjsp_set1_v2r8(2.0);
117 union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
124 jindex = nlist->jindex;
126 shiftidx = nlist->shift;
128 shiftvec = fr->shift_vec[0];
129 fshift = fr->fshift[0];
130 facel = gmx_fjsp_set1_v2r8(fr->epsfac);
131 charge = mdatoms->chargeA;
132 nvdwtype = fr->ntype;
134 vdwtype = mdatoms->typeA;
135 vdwgridparam = fr->ljpme_c6grid;
136 sh_lj_ewald = gmx_fjsp_set1_v2r8(fr->ic->sh_lj_ewald);
137 ewclj = gmx_fjsp_set1_v2r8(fr->ewaldcoeff_lj);
138 ewclj2 = _fjsp_mul_v2r8(minus_one,_fjsp_mul_v2r8(ewclj,ewclj));
140 sh_ewald = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
141 ewtab = fr->ic->tabq_coul_FDV0;
142 ewtabscale = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
143 ewtabhalfspace = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
145 /* Setup water-specific parameters */
146 inr = nlist->iinr[0];
147 iq1 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
148 iq2 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
149 iq3 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
150 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
152 /* Avoid stupid compiler warnings */
160 /* Start outer loop over neighborlists */
161 for(iidx=0; iidx<nri; iidx++)
163 /* Load shift vector for this list */
164 i_shift_offset = DIM*shiftidx[iidx];
166 /* Load limits for loop over neighbors */
167 j_index_start = jindex[iidx];
168 j_index_end = jindex[iidx+1];
170 /* Get outer coordinate index */
172 i_coord_offset = DIM*inr;
174 /* Load i particle coords and add shift vector */
175 gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
176 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
178 fix0 = _fjsp_setzero_v2r8();
179 fiy0 = _fjsp_setzero_v2r8();
180 fiz0 = _fjsp_setzero_v2r8();
181 fix1 = _fjsp_setzero_v2r8();
182 fiy1 = _fjsp_setzero_v2r8();
183 fiz1 = _fjsp_setzero_v2r8();
184 fix2 = _fjsp_setzero_v2r8();
185 fiy2 = _fjsp_setzero_v2r8();
186 fiz2 = _fjsp_setzero_v2r8();
187 fix3 = _fjsp_setzero_v2r8();
188 fiy3 = _fjsp_setzero_v2r8();
189 fiz3 = _fjsp_setzero_v2r8();
191 /* Reset potential sums */
192 velecsum = _fjsp_setzero_v2r8();
193 vvdwsum = _fjsp_setzero_v2r8();
195 /* Start inner kernel loop */
196 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
199 /* Get j neighbor index, and coordinate index */
202 j_coord_offsetA = DIM*jnrA;
203 j_coord_offsetB = DIM*jnrB;
205 /* load j atom coordinates */
206 gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
209 /* Calculate displacement vector */
210 dx00 = _fjsp_sub_v2r8(ix0,jx0);
211 dy00 = _fjsp_sub_v2r8(iy0,jy0);
212 dz00 = _fjsp_sub_v2r8(iz0,jz0);
213 dx10 = _fjsp_sub_v2r8(ix1,jx0);
214 dy10 = _fjsp_sub_v2r8(iy1,jy0);
215 dz10 = _fjsp_sub_v2r8(iz1,jz0);
216 dx20 = _fjsp_sub_v2r8(ix2,jx0);
217 dy20 = _fjsp_sub_v2r8(iy2,jy0);
218 dz20 = _fjsp_sub_v2r8(iz2,jz0);
219 dx30 = _fjsp_sub_v2r8(ix3,jx0);
220 dy30 = _fjsp_sub_v2r8(iy3,jy0);
221 dz30 = _fjsp_sub_v2r8(iz3,jz0);
223 /* Calculate squared distance and things based on it */
224 rsq00 = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
225 rsq10 = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
226 rsq20 = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
227 rsq30 = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
229 rinv00 = gmx_fjsp_invsqrt_v2r8(rsq00);
230 rinv10 = gmx_fjsp_invsqrt_v2r8(rsq10);
231 rinv20 = gmx_fjsp_invsqrt_v2r8(rsq20);
232 rinv30 = gmx_fjsp_invsqrt_v2r8(rsq30);
234 rinvsq00 = _fjsp_mul_v2r8(rinv00,rinv00);
235 rinvsq10 = _fjsp_mul_v2r8(rinv10,rinv10);
236 rinvsq20 = _fjsp_mul_v2r8(rinv20,rinv20);
237 rinvsq30 = _fjsp_mul_v2r8(rinv30,rinv30);
239 /* Load parameters for j particles */
240 jq0 = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
241 vdwjidx0A = 2*vdwtype[jnrA+0];
242 vdwjidx0B = 2*vdwtype[jnrB+0];
244 fjx0 = _fjsp_setzero_v2r8();
245 fjy0 = _fjsp_setzero_v2r8();
246 fjz0 = _fjsp_setzero_v2r8();
248 /**************************
249 * CALCULATE INTERACTIONS *
250 **************************/
252 r00 = _fjsp_mul_v2r8(rsq00,rinv00);
254 /* Compute parameters for interactions between i and j atoms */
255 gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
256 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
258 c6grid_00 = gmx_fjsp_load_2real_swizzle_v2r8(vdwgridparam+vdwioffset0+vdwjidx0A,
259 vdwgridparam+vdwioffset0+vdwjidx0B);
261 /* Analytical LJ-PME */
262 rinvsix = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
263 ewcljrsq = _fjsp_mul_v2r8(ewclj2,rsq00);
264 ewclj6 = _fjsp_mul_v2r8(ewclj2,_fjsp_mul_v2r8(ewclj2,ewclj2));
265 exponent = gmx_simd_exp_d(ewcljrsq);
266 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
267 poly = _fjsp_mul_v2r8(exponent,_fjsp_madd_v2r8(_fjsp_mul_v2r8(ewcljrsq,ewcljrsq),one_half,_fjsp_sub_v2r8(one,ewcljrsq)));
268 /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
269 vvdw6 = _fjsp_mul_v2r8(_fjsp_madd_v2r8(c6grid_00,_fjsp_sub_v2r8(poly,one),c6_00),rinvsix);
270 vvdw12 = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
271 vvdw = _fjsp_msub_v2r8(vvdw12,one_twelfth,_fjsp_mul_v2r8(vvdw6,one_sixth));
272 /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
273 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);
275 /* Update potential sum for this i atom from the interaction with this j atom. */
276 vvdwsum = _fjsp_add_v2r8(vvdwsum,vvdw);
280 /* Update vectorial force */
281 fix0 = _fjsp_madd_v2r8(dx00,fscal,fix0);
282 fiy0 = _fjsp_madd_v2r8(dy00,fscal,fiy0);
283 fiz0 = _fjsp_madd_v2r8(dz00,fscal,fiz0);
285 fjx0 = _fjsp_madd_v2r8(dx00,fscal,fjx0);
286 fjy0 = _fjsp_madd_v2r8(dy00,fscal,fjy0);
287 fjz0 = _fjsp_madd_v2r8(dz00,fscal,fjz0);
289 /**************************
290 * CALCULATE INTERACTIONS *
291 **************************/
293 r10 = _fjsp_mul_v2r8(rsq10,rinv10);
295 /* Compute parameters for interactions between i and j atoms */
296 qq10 = _fjsp_mul_v2r8(iq1,jq0);
298 /* EWALD ELECTROSTATICS */
300 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
301 ewrt = _fjsp_mul_v2r8(r10,ewtabscale);
302 itab_tmp = _fjsp_dtox_v2r8(ewrt);
303 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
304 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
306 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
307 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
308 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
309 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
310 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
311 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
312 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
313 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
314 velec = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
315 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
317 /* Update potential sum for this i atom from the interaction with this j atom. */
318 velecsum = _fjsp_add_v2r8(velecsum,velec);
322 /* Update vectorial force */
323 fix1 = _fjsp_madd_v2r8(dx10,fscal,fix1);
324 fiy1 = _fjsp_madd_v2r8(dy10,fscal,fiy1);
325 fiz1 = _fjsp_madd_v2r8(dz10,fscal,fiz1);
327 fjx0 = _fjsp_madd_v2r8(dx10,fscal,fjx0);
328 fjy0 = _fjsp_madd_v2r8(dy10,fscal,fjy0);
329 fjz0 = _fjsp_madd_v2r8(dz10,fscal,fjz0);
331 /**************************
332 * CALCULATE INTERACTIONS *
333 **************************/
335 r20 = _fjsp_mul_v2r8(rsq20,rinv20);
337 /* Compute parameters for interactions between i and j atoms */
338 qq20 = _fjsp_mul_v2r8(iq2,jq0);
340 /* EWALD ELECTROSTATICS */
342 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
343 ewrt = _fjsp_mul_v2r8(r20,ewtabscale);
344 itab_tmp = _fjsp_dtox_v2r8(ewrt);
345 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
346 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
348 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
349 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
350 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
351 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
352 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
353 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
354 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
355 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
356 velec = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
357 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
359 /* Update potential sum for this i atom from the interaction with this j atom. */
360 velecsum = _fjsp_add_v2r8(velecsum,velec);
364 /* Update vectorial force */
365 fix2 = _fjsp_madd_v2r8(dx20,fscal,fix2);
366 fiy2 = _fjsp_madd_v2r8(dy20,fscal,fiy2);
367 fiz2 = _fjsp_madd_v2r8(dz20,fscal,fiz2);
369 fjx0 = _fjsp_madd_v2r8(dx20,fscal,fjx0);
370 fjy0 = _fjsp_madd_v2r8(dy20,fscal,fjy0);
371 fjz0 = _fjsp_madd_v2r8(dz20,fscal,fjz0);
373 /**************************
374 * CALCULATE INTERACTIONS *
375 **************************/
377 r30 = _fjsp_mul_v2r8(rsq30,rinv30);
379 /* Compute parameters for interactions between i and j atoms */
380 qq30 = _fjsp_mul_v2r8(iq3,jq0);
382 /* EWALD ELECTROSTATICS */
384 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
385 ewrt = _fjsp_mul_v2r8(r30,ewtabscale);
386 itab_tmp = _fjsp_dtox_v2r8(ewrt);
387 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
388 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
390 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
391 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
392 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
393 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
394 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
395 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
396 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
397 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
398 velec = _fjsp_mul_v2r8(qq30,_fjsp_sub_v2r8(rinv30,velec));
399 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,rinv30),_fjsp_sub_v2r8(rinvsq30,felec));
401 /* Update potential sum for this i atom from the interaction with this j atom. */
402 velecsum = _fjsp_add_v2r8(velecsum,velec);
406 /* Update vectorial force */
407 fix3 = _fjsp_madd_v2r8(dx30,fscal,fix3);
408 fiy3 = _fjsp_madd_v2r8(dy30,fscal,fiy3);
409 fiz3 = _fjsp_madd_v2r8(dz30,fscal,fiz3);
411 fjx0 = _fjsp_madd_v2r8(dx30,fscal,fjx0);
412 fjy0 = _fjsp_madd_v2r8(dy30,fscal,fjy0);
413 fjz0 = _fjsp_madd_v2r8(dz30,fscal,fjz0);
415 gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
417 /* Inner loop uses 185 flops */
424 j_coord_offsetA = DIM*jnrA;
426 /* load j atom coordinates */
427 gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
430 /* Calculate displacement vector */
431 dx00 = _fjsp_sub_v2r8(ix0,jx0);
432 dy00 = _fjsp_sub_v2r8(iy0,jy0);
433 dz00 = _fjsp_sub_v2r8(iz0,jz0);
434 dx10 = _fjsp_sub_v2r8(ix1,jx0);
435 dy10 = _fjsp_sub_v2r8(iy1,jy0);
436 dz10 = _fjsp_sub_v2r8(iz1,jz0);
437 dx20 = _fjsp_sub_v2r8(ix2,jx0);
438 dy20 = _fjsp_sub_v2r8(iy2,jy0);
439 dz20 = _fjsp_sub_v2r8(iz2,jz0);
440 dx30 = _fjsp_sub_v2r8(ix3,jx0);
441 dy30 = _fjsp_sub_v2r8(iy3,jy0);
442 dz30 = _fjsp_sub_v2r8(iz3,jz0);
444 /* Calculate squared distance and things based on it */
445 rsq00 = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
446 rsq10 = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
447 rsq20 = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
448 rsq30 = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
450 rinv00 = gmx_fjsp_invsqrt_v2r8(rsq00);
451 rinv10 = gmx_fjsp_invsqrt_v2r8(rsq10);
452 rinv20 = gmx_fjsp_invsqrt_v2r8(rsq20);
453 rinv30 = gmx_fjsp_invsqrt_v2r8(rsq30);
455 rinvsq00 = _fjsp_mul_v2r8(rinv00,rinv00);
456 rinvsq10 = _fjsp_mul_v2r8(rinv10,rinv10);
457 rinvsq20 = _fjsp_mul_v2r8(rinv20,rinv20);
458 rinvsq30 = _fjsp_mul_v2r8(rinv30,rinv30);
460 /* Load parameters for j particles */
461 jq0 = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
462 vdwjidx0A = 2*vdwtype[jnrA+0];
464 fjx0 = _fjsp_setzero_v2r8();
465 fjy0 = _fjsp_setzero_v2r8();
466 fjz0 = _fjsp_setzero_v2r8();
468 /**************************
469 * CALCULATE INTERACTIONS *
470 **************************/
472 r00 = _fjsp_mul_v2r8(rsq00,rinv00);
474 /* Compute parameters for interactions between i and j atoms */
475 gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
476 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
478 c6grid_00 = gmx_fjsp_load_2real_swizzle_v2r8(vdwgridparam+vdwioffset0+vdwjidx0A,
479 vdwgridparam+vdwioffset0+vdwjidx0B);
481 /* Analytical LJ-PME */
482 rinvsix = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
483 ewcljrsq = _fjsp_mul_v2r8(ewclj2,rsq00);
484 ewclj6 = _fjsp_mul_v2r8(ewclj2,_fjsp_mul_v2r8(ewclj2,ewclj2));
485 exponent = gmx_simd_exp_d(ewcljrsq);
486 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
487 poly = _fjsp_mul_v2r8(exponent,_fjsp_madd_v2r8(_fjsp_mul_v2r8(ewcljrsq,ewcljrsq),one_half,_fjsp_sub_v2r8(one,ewcljrsq)));
488 /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
489 vvdw6 = _fjsp_mul_v2r8(_fjsp_madd_v2r8(c6grid_00,_fjsp_sub_v2r8(poly,one),c6_00),rinvsix);
490 vvdw12 = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
491 vvdw = _fjsp_msub_v2r8(vvdw12,one_twelfth,_fjsp_mul_v2r8(vvdw6,one_sixth));
492 /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
493 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);
495 /* Update potential sum for this i atom from the interaction with this j atom. */
496 vvdw = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
497 vvdwsum = _fjsp_add_v2r8(vvdwsum,vvdw);
501 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
503 /* Update vectorial force */
504 fix0 = _fjsp_madd_v2r8(dx00,fscal,fix0);
505 fiy0 = _fjsp_madd_v2r8(dy00,fscal,fiy0);
506 fiz0 = _fjsp_madd_v2r8(dz00,fscal,fiz0);
508 fjx0 = _fjsp_madd_v2r8(dx00,fscal,fjx0);
509 fjy0 = _fjsp_madd_v2r8(dy00,fscal,fjy0);
510 fjz0 = _fjsp_madd_v2r8(dz00,fscal,fjz0);
512 /**************************
513 * CALCULATE INTERACTIONS *
514 **************************/
516 r10 = _fjsp_mul_v2r8(rsq10,rinv10);
518 /* Compute parameters for interactions between i and j atoms */
519 qq10 = _fjsp_mul_v2r8(iq1,jq0);
521 /* EWALD ELECTROSTATICS */
523 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
524 ewrt = _fjsp_mul_v2r8(r10,ewtabscale);
525 itab_tmp = _fjsp_dtox_v2r8(ewrt);
526 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
527 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
529 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
530 ewtabD = _fjsp_setzero_v2r8();
531 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
532 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
533 ewtabFn = _fjsp_setzero_v2r8();
534 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
535 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
536 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
537 velec = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
538 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
540 /* Update potential sum for this i atom from the interaction with this j atom. */
541 velec = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
542 velecsum = _fjsp_add_v2r8(velecsum,velec);
546 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
548 /* Update vectorial force */
549 fix1 = _fjsp_madd_v2r8(dx10,fscal,fix1);
550 fiy1 = _fjsp_madd_v2r8(dy10,fscal,fiy1);
551 fiz1 = _fjsp_madd_v2r8(dz10,fscal,fiz1);
553 fjx0 = _fjsp_madd_v2r8(dx10,fscal,fjx0);
554 fjy0 = _fjsp_madd_v2r8(dy10,fscal,fjy0);
555 fjz0 = _fjsp_madd_v2r8(dz10,fscal,fjz0);
557 /**************************
558 * CALCULATE INTERACTIONS *
559 **************************/
561 r20 = _fjsp_mul_v2r8(rsq20,rinv20);
563 /* Compute parameters for interactions between i and j atoms */
564 qq20 = _fjsp_mul_v2r8(iq2,jq0);
566 /* EWALD ELECTROSTATICS */
568 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
569 ewrt = _fjsp_mul_v2r8(r20,ewtabscale);
570 itab_tmp = _fjsp_dtox_v2r8(ewrt);
571 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
572 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
574 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
575 ewtabD = _fjsp_setzero_v2r8();
576 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
577 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
578 ewtabFn = _fjsp_setzero_v2r8();
579 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
580 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
581 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
582 velec = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
583 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
585 /* Update potential sum for this i atom from the interaction with this j atom. */
586 velec = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
587 velecsum = _fjsp_add_v2r8(velecsum,velec);
591 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
593 /* Update vectorial force */
594 fix2 = _fjsp_madd_v2r8(dx20,fscal,fix2);
595 fiy2 = _fjsp_madd_v2r8(dy20,fscal,fiy2);
596 fiz2 = _fjsp_madd_v2r8(dz20,fscal,fiz2);
598 fjx0 = _fjsp_madd_v2r8(dx20,fscal,fjx0);
599 fjy0 = _fjsp_madd_v2r8(dy20,fscal,fjy0);
600 fjz0 = _fjsp_madd_v2r8(dz20,fscal,fjz0);
602 /**************************
603 * CALCULATE INTERACTIONS *
604 **************************/
606 r30 = _fjsp_mul_v2r8(rsq30,rinv30);
608 /* Compute parameters for interactions between i and j atoms */
609 qq30 = _fjsp_mul_v2r8(iq3,jq0);
611 /* EWALD ELECTROSTATICS */
613 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
614 ewrt = _fjsp_mul_v2r8(r30,ewtabscale);
615 itab_tmp = _fjsp_dtox_v2r8(ewrt);
616 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
617 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
619 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
620 ewtabD = _fjsp_setzero_v2r8();
621 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
622 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
623 ewtabFn = _fjsp_setzero_v2r8();
624 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
625 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
626 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
627 velec = _fjsp_mul_v2r8(qq30,_fjsp_sub_v2r8(rinv30,velec));
628 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,rinv30),_fjsp_sub_v2r8(rinvsq30,felec));
630 /* Update potential sum for this i atom from the interaction with this j atom. */
631 velec = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
632 velecsum = _fjsp_add_v2r8(velecsum,velec);
636 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
638 /* Update vectorial force */
639 fix3 = _fjsp_madd_v2r8(dx30,fscal,fix3);
640 fiy3 = _fjsp_madd_v2r8(dy30,fscal,fiy3);
641 fiz3 = _fjsp_madd_v2r8(dz30,fscal,fiz3);
643 fjx0 = _fjsp_madd_v2r8(dx30,fscal,fjx0);
644 fjy0 = _fjsp_madd_v2r8(dy30,fscal,fjy0);
645 fjz0 = _fjsp_madd_v2r8(dz30,fscal,fjz0);
647 gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
649 /* Inner loop uses 185 flops */
652 /* End of innermost loop */
654 gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
655 f+i_coord_offset,fshift+i_shift_offset);
658 /* Update potential energies */
659 gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
660 gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
662 /* Increment number of inner iterations */
663 inneriter += j_index_end - j_index_start;
665 /* Outer loop uses 26 flops */
668 /* Increment number of outer iterations */
671 /* Update outer/inner flops */
673 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_VF,outeriter*26 + inneriter*185);
676 * Gromacs nonbonded kernel: nb_kernel_ElecEw_VdwLJEw_GeomW4P1_F_sparc64_hpc_ace_double
677 * Electrostatics interaction: Ewald
678 * VdW interaction: LJEwald
679 * Geometry: Water4-Particle
680 * Calculate force/pot: Force
683 nb_kernel_ElecEw_VdwLJEw_GeomW4P1_F_sparc64_hpc_ace_double
684 (t_nblist * gmx_restrict nlist,
685 rvec * gmx_restrict xx,
686 rvec * gmx_restrict ff,
687 t_forcerec * gmx_restrict fr,
688 t_mdatoms * gmx_restrict mdatoms,
689 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
690 t_nrnb * gmx_restrict nrnb)
692 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
693 * just 0 for non-waters.
694 * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
695 * jnr indices corresponding to data put in the four positions in the SIMD register.
697 int i_shift_offset,i_coord_offset,outeriter,inneriter;
698 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
700 int j_coord_offsetA,j_coord_offsetB;
701 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
703 real *shiftvec,*fshift,*x,*f;
704 _fjsp_v2r8 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
706 _fjsp_v2r8 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
708 _fjsp_v2r8 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
710 _fjsp_v2r8 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
712 _fjsp_v2r8 ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
713 int vdwjidx0A,vdwjidx0B;
714 _fjsp_v2r8 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
715 _fjsp_v2r8 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
716 _fjsp_v2r8 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
717 _fjsp_v2r8 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
718 _fjsp_v2r8 dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
719 _fjsp_v2r8 velec,felec,velecsum,facel,crf,krf,krf2;
722 _fjsp_v2r8 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
725 _fjsp_v2r8 one_sixth = gmx_fjsp_set1_v2r8(1.0/6.0);
726 _fjsp_v2r8 one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
727 _fjsp_v2r8 c6grid_00;
728 _fjsp_v2r8 c6grid_10;
729 _fjsp_v2r8 c6grid_20;
730 _fjsp_v2r8 c6grid_30;
732 _fjsp_v2r8 ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
733 _fjsp_v2r8 one_half = gmx_fjsp_set1_v2r8(0.5);
734 _fjsp_v2r8 minus_one = gmx_fjsp_set1_v2r8(-1.0);
735 _fjsp_v2r8 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
738 _fjsp_v2r8 dummy_mask,cutoff_mask;
739 _fjsp_v2r8 one = gmx_fjsp_set1_v2r8(1.0);
740 _fjsp_v2r8 two = gmx_fjsp_set1_v2r8(2.0);
741 union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
748 jindex = nlist->jindex;
750 shiftidx = nlist->shift;
752 shiftvec = fr->shift_vec[0];
753 fshift = fr->fshift[0];
754 facel = gmx_fjsp_set1_v2r8(fr->epsfac);
755 charge = mdatoms->chargeA;
756 nvdwtype = fr->ntype;
758 vdwtype = mdatoms->typeA;
759 vdwgridparam = fr->ljpme_c6grid;
760 sh_lj_ewald = gmx_fjsp_set1_v2r8(fr->ic->sh_lj_ewald);
761 ewclj = gmx_fjsp_set1_v2r8(fr->ewaldcoeff_lj);
762 ewclj2 = _fjsp_mul_v2r8(minus_one,_fjsp_mul_v2r8(ewclj,ewclj));
764 sh_ewald = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
765 ewtab = fr->ic->tabq_coul_F;
766 ewtabscale = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
767 ewtabhalfspace = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
769 /* Setup water-specific parameters */
770 inr = nlist->iinr[0];
771 iq1 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
772 iq2 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
773 iq3 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
774 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
776 /* Avoid stupid compiler warnings */
784 /* Start outer loop over neighborlists */
785 for(iidx=0; iidx<nri; iidx++)
787 /* Load shift vector for this list */
788 i_shift_offset = DIM*shiftidx[iidx];
790 /* Load limits for loop over neighbors */
791 j_index_start = jindex[iidx];
792 j_index_end = jindex[iidx+1];
794 /* Get outer coordinate index */
796 i_coord_offset = DIM*inr;
798 /* Load i particle coords and add shift vector */
799 gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
800 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
802 fix0 = _fjsp_setzero_v2r8();
803 fiy0 = _fjsp_setzero_v2r8();
804 fiz0 = _fjsp_setzero_v2r8();
805 fix1 = _fjsp_setzero_v2r8();
806 fiy1 = _fjsp_setzero_v2r8();
807 fiz1 = _fjsp_setzero_v2r8();
808 fix2 = _fjsp_setzero_v2r8();
809 fiy2 = _fjsp_setzero_v2r8();
810 fiz2 = _fjsp_setzero_v2r8();
811 fix3 = _fjsp_setzero_v2r8();
812 fiy3 = _fjsp_setzero_v2r8();
813 fiz3 = _fjsp_setzero_v2r8();
815 /* Start inner kernel loop */
816 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
819 /* Get j neighbor index, and coordinate index */
822 j_coord_offsetA = DIM*jnrA;
823 j_coord_offsetB = DIM*jnrB;
825 /* load j atom coordinates */
826 gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
829 /* Calculate displacement vector */
830 dx00 = _fjsp_sub_v2r8(ix0,jx0);
831 dy00 = _fjsp_sub_v2r8(iy0,jy0);
832 dz00 = _fjsp_sub_v2r8(iz0,jz0);
833 dx10 = _fjsp_sub_v2r8(ix1,jx0);
834 dy10 = _fjsp_sub_v2r8(iy1,jy0);
835 dz10 = _fjsp_sub_v2r8(iz1,jz0);
836 dx20 = _fjsp_sub_v2r8(ix2,jx0);
837 dy20 = _fjsp_sub_v2r8(iy2,jy0);
838 dz20 = _fjsp_sub_v2r8(iz2,jz0);
839 dx30 = _fjsp_sub_v2r8(ix3,jx0);
840 dy30 = _fjsp_sub_v2r8(iy3,jy0);
841 dz30 = _fjsp_sub_v2r8(iz3,jz0);
843 /* Calculate squared distance and things based on it */
844 rsq00 = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
845 rsq10 = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
846 rsq20 = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
847 rsq30 = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
849 rinv00 = gmx_fjsp_invsqrt_v2r8(rsq00);
850 rinv10 = gmx_fjsp_invsqrt_v2r8(rsq10);
851 rinv20 = gmx_fjsp_invsqrt_v2r8(rsq20);
852 rinv30 = gmx_fjsp_invsqrt_v2r8(rsq30);
854 rinvsq00 = _fjsp_mul_v2r8(rinv00,rinv00);
855 rinvsq10 = _fjsp_mul_v2r8(rinv10,rinv10);
856 rinvsq20 = _fjsp_mul_v2r8(rinv20,rinv20);
857 rinvsq30 = _fjsp_mul_v2r8(rinv30,rinv30);
859 /* Load parameters for j particles */
860 jq0 = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
861 vdwjidx0A = 2*vdwtype[jnrA+0];
862 vdwjidx0B = 2*vdwtype[jnrB+0];
864 fjx0 = _fjsp_setzero_v2r8();
865 fjy0 = _fjsp_setzero_v2r8();
866 fjz0 = _fjsp_setzero_v2r8();
868 /**************************
869 * CALCULATE INTERACTIONS *
870 **************************/
872 r00 = _fjsp_mul_v2r8(rsq00,rinv00);
874 /* Compute parameters for interactions between i and j atoms */
875 gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
876 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
878 c6grid_00 = gmx_fjsp_load_2real_swizzle_v2r8(vdwgridparam+vdwioffset0+vdwjidx0A,
879 vdwgridparam+vdwioffset0+vdwjidx0B);
881 /* Analytical LJ-PME */
882 rinvsix = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
883 ewcljrsq = _fjsp_mul_v2r8(ewclj2,rsq00);
884 ewclj6 = _fjsp_mul_v2r8(ewclj2,_fjsp_mul_v2r8(ewclj2,ewclj2));
885 exponent = gmx_simd_exp_d(ewcljrsq);
886 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
887 poly = _fjsp_mul_v2r8(exponent,_fjsp_madd_v2r8(_fjsp_mul_v2r8(ewcljrsq,ewcljrsq),one_half,_fjsp_sub_v2r8(one,ewcljrsq)));
888 /* f6A = 6 * C6grid * (1 - poly) */
889 f6A = _fjsp_mul_v2r8(c6grid_00,_fjsp_sub_v2r8(one,poly));
890 /* f6B = C6grid * exponent * beta^6 */
891 f6B = _fjsp_mul_v2r8(_fjsp_mul_v2r8(c6grid_00,one_sixth),_fjsp_mul_v2r8(exponent,ewclj6));
892 /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
893 fvdw = _fjsp_mul_v2r8(_fjsp_madd_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,_fjsp_sub_v2r8(c6_00,f6A)),rinvsix,f6B),rinvsq00);
897 /* Update vectorial force */
898 fix0 = _fjsp_madd_v2r8(dx00,fscal,fix0);
899 fiy0 = _fjsp_madd_v2r8(dy00,fscal,fiy0);
900 fiz0 = _fjsp_madd_v2r8(dz00,fscal,fiz0);
902 fjx0 = _fjsp_madd_v2r8(dx00,fscal,fjx0);
903 fjy0 = _fjsp_madd_v2r8(dy00,fscal,fjy0);
904 fjz0 = _fjsp_madd_v2r8(dz00,fscal,fjz0);
906 /**************************
907 * CALCULATE INTERACTIONS *
908 **************************/
910 r10 = _fjsp_mul_v2r8(rsq10,rinv10);
912 /* Compute parameters for interactions between i and j atoms */
913 qq10 = _fjsp_mul_v2r8(iq1,jq0);
915 /* EWALD ELECTROSTATICS */
917 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
918 ewrt = _fjsp_mul_v2r8(r10,ewtabscale);
919 itab_tmp = _fjsp_dtox_v2r8(ewrt);
920 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
921 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
923 gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
925 felec = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
926 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
930 /* Update vectorial force */
931 fix1 = _fjsp_madd_v2r8(dx10,fscal,fix1);
932 fiy1 = _fjsp_madd_v2r8(dy10,fscal,fiy1);
933 fiz1 = _fjsp_madd_v2r8(dz10,fscal,fiz1);
935 fjx0 = _fjsp_madd_v2r8(dx10,fscal,fjx0);
936 fjy0 = _fjsp_madd_v2r8(dy10,fscal,fjy0);
937 fjz0 = _fjsp_madd_v2r8(dz10,fscal,fjz0);
939 /**************************
940 * CALCULATE INTERACTIONS *
941 **************************/
943 r20 = _fjsp_mul_v2r8(rsq20,rinv20);
945 /* Compute parameters for interactions between i and j atoms */
946 qq20 = _fjsp_mul_v2r8(iq2,jq0);
948 /* EWALD ELECTROSTATICS */
950 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
951 ewrt = _fjsp_mul_v2r8(r20,ewtabscale);
952 itab_tmp = _fjsp_dtox_v2r8(ewrt);
953 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
954 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
956 gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
958 felec = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
959 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
963 /* Update vectorial force */
964 fix2 = _fjsp_madd_v2r8(dx20,fscal,fix2);
965 fiy2 = _fjsp_madd_v2r8(dy20,fscal,fiy2);
966 fiz2 = _fjsp_madd_v2r8(dz20,fscal,fiz2);
968 fjx0 = _fjsp_madd_v2r8(dx20,fscal,fjx0);
969 fjy0 = _fjsp_madd_v2r8(dy20,fscal,fjy0);
970 fjz0 = _fjsp_madd_v2r8(dz20,fscal,fjz0);
972 /**************************
973 * CALCULATE INTERACTIONS *
974 **************************/
976 r30 = _fjsp_mul_v2r8(rsq30,rinv30);
978 /* Compute parameters for interactions between i and j atoms */
979 qq30 = _fjsp_mul_v2r8(iq3,jq0);
981 /* EWALD ELECTROSTATICS */
983 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
984 ewrt = _fjsp_mul_v2r8(r30,ewtabscale);
985 itab_tmp = _fjsp_dtox_v2r8(ewrt);
986 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
987 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
989 gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
991 felec = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
992 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,rinv30),_fjsp_sub_v2r8(rinvsq30,felec));
996 /* Update vectorial force */
997 fix3 = _fjsp_madd_v2r8(dx30,fscal,fix3);
998 fiy3 = _fjsp_madd_v2r8(dy30,fscal,fiy3);
999 fiz3 = _fjsp_madd_v2r8(dz30,fscal,fiz3);
1001 fjx0 = _fjsp_madd_v2r8(dx30,fscal,fjx0);
1002 fjy0 = _fjsp_madd_v2r8(dy30,fscal,fjy0);
1003 fjz0 = _fjsp_madd_v2r8(dz30,fscal,fjz0);
1005 gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
1007 /* Inner loop uses 168 flops */
1010 if(jidx<j_index_end)
1014 j_coord_offsetA = DIM*jnrA;
1016 /* load j atom coordinates */
1017 gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
1020 /* Calculate displacement vector */
1021 dx00 = _fjsp_sub_v2r8(ix0,jx0);
1022 dy00 = _fjsp_sub_v2r8(iy0,jy0);
1023 dz00 = _fjsp_sub_v2r8(iz0,jz0);
1024 dx10 = _fjsp_sub_v2r8(ix1,jx0);
1025 dy10 = _fjsp_sub_v2r8(iy1,jy0);
1026 dz10 = _fjsp_sub_v2r8(iz1,jz0);
1027 dx20 = _fjsp_sub_v2r8(ix2,jx0);
1028 dy20 = _fjsp_sub_v2r8(iy2,jy0);
1029 dz20 = _fjsp_sub_v2r8(iz2,jz0);
1030 dx30 = _fjsp_sub_v2r8(ix3,jx0);
1031 dy30 = _fjsp_sub_v2r8(iy3,jy0);
1032 dz30 = _fjsp_sub_v2r8(iz3,jz0);
1034 /* Calculate squared distance and things based on it */
1035 rsq00 = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
1036 rsq10 = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
1037 rsq20 = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
1038 rsq30 = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
1040 rinv00 = gmx_fjsp_invsqrt_v2r8(rsq00);
1041 rinv10 = gmx_fjsp_invsqrt_v2r8(rsq10);
1042 rinv20 = gmx_fjsp_invsqrt_v2r8(rsq20);
1043 rinv30 = gmx_fjsp_invsqrt_v2r8(rsq30);
1045 rinvsq00 = _fjsp_mul_v2r8(rinv00,rinv00);
1046 rinvsq10 = _fjsp_mul_v2r8(rinv10,rinv10);
1047 rinvsq20 = _fjsp_mul_v2r8(rinv20,rinv20);
1048 rinvsq30 = _fjsp_mul_v2r8(rinv30,rinv30);
1050 /* Load parameters for j particles */
1051 jq0 = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
1052 vdwjidx0A = 2*vdwtype[jnrA+0];
1054 fjx0 = _fjsp_setzero_v2r8();
1055 fjy0 = _fjsp_setzero_v2r8();
1056 fjz0 = _fjsp_setzero_v2r8();
1058 /**************************
1059 * CALCULATE INTERACTIONS *
1060 **************************/
1062 r00 = _fjsp_mul_v2r8(rsq00,rinv00);
1064 /* Compute parameters for interactions between i and j atoms */
1065 gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
1066 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
1068 c6grid_00 = gmx_fjsp_load_2real_swizzle_v2r8(vdwgridparam+vdwioffset0+vdwjidx0A,
1069 vdwgridparam+vdwioffset0+vdwjidx0B);
1071 /* Analytical LJ-PME */
1072 rinvsix = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
1073 ewcljrsq = _fjsp_mul_v2r8(ewclj2,rsq00);
1074 ewclj6 = _fjsp_mul_v2r8(ewclj2,_fjsp_mul_v2r8(ewclj2,ewclj2));
1075 exponent = gmx_simd_exp_d(ewcljrsq);
1076 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
1077 poly = _fjsp_mul_v2r8(exponent,_fjsp_madd_v2r8(_fjsp_mul_v2r8(ewcljrsq,ewcljrsq),one_half,_fjsp_sub_v2r8(one,ewcljrsq)));
1078 /* f6A = 6 * C6grid * (1 - poly) */
1079 f6A = _fjsp_mul_v2r8(c6grid_00,_fjsp_sub_v2r8(one,poly));
1080 /* f6B = C6grid * exponent * beta^6 */
1081 f6B = _fjsp_mul_v2r8(_fjsp_mul_v2r8(c6grid_00,one_sixth),_fjsp_mul_v2r8(exponent,ewclj6));
1082 /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
1083 fvdw = _fjsp_mul_v2r8(_fjsp_madd_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,_fjsp_sub_v2r8(c6_00,f6A)),rinvsix,f6B),rinvsq00);
1087 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1089 /* Update vectorial force */
1090 fix0 = _fjsp_madd_v2r8(dx00,fscal,fix0);
1091 fiy0 = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1092 fiz0 = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1094 fjx0 = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1095 fjy0 = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1096 fjz0 = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1098 /**************************
1099 * CALCULATE INTERACTIONS *
1100 **************************/
1102 r10 = _fjsp_mul_v2r8(rsq10,rinv10);
1104 /* Compute parameters for interactions between i and j atoms */
1105 qq10 = _fjsp_mul_v2r8(iq1,jq0);
1107 /* EWALD ELECTROSTATICS */
1109 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1110 ewrt = _fjsp_mul_v2r8(r10,ewtabscale);
1111 itab_tmp = _fjsp_dtox_v2r8(ewrt);
1112 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1113 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1115 gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
1116 felec = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1117 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
1121 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1123 /* Update vectorial force */
1124 fix1 = _fjsp_madd_v2r8(dx10,fscal,fix1);
1125 fiy1 = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1126 fiz1 = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1128 fjx0 = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1129 fjy0 = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1130 fjz0 = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1132 /**************************
1133 * CALCULATE INTERACTIONS *
1134 **************************/
1136 r20 = _fjsp_mul_v2r8(rsq20,rinv20);
1138 /* Compute parameters for interactions between i and j atoms */
1139 qq20 = _fjsp_mul_v2r8(iq2,jq0);
1141 /* EWALD ELECTROSTATICS */
1143 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1144 ewrt = _fjsp_mul_v2r8(r20,ewtabscale);
1145 itab_tmp = _fjsp_dtox_v2r8(ewrt);
1146 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1147 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1149 gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
1150 felec = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1151 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
1155 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1157 /* Update vectorial force */
1158 fix2 = _fjsp_madd_v2r8(dx20,fscal,fix2);
1159 fiy2 = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1160 fiz2 = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1162 fjx0 = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1163 fjy0 = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1164 fjz0 = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1166 /**************************
1167 * CALCULATE INTERACTIONS *
1168 **************************/
1170 r30 = _fjsp_mul_v2r8(rsq30,rinv30);
1172 /* Compute parameters for interactions between i and j atoms */
1173 qq30 = _fjsp_mul_v2r8(iq3,jq0);
1175 /* EWALD ELECTROSTATICS */
1177 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1178 ewrt = _fjsp_mul_v2r8(r30,ewtabscale);
1179 itab_tmp = _fjsp_dtox_v2r8(ewrt);
1180 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1181 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1183 gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
1184 felec = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1185 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,rinv30),_fjsp_sub_v2r8(rinvsq30,felec));
1189 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1191 /* Update vectorial force */
1192 fix3 = _fjsp_madd_v2r8(dx30,fscal,fix3);
1193 fiy3 = _fjsp_madd_v2r8(dy30,fscal,fiy3);
1194 fiz3 = _fjsp_madd_v2r8(dz30,fscal,fiz3);
1196 fjx0 = _fjsp_madd_v2r8(dx30,fscal,fjx0);
1197 fjy0 = _fjsp_madd_v2r8(dy30,fscal,fjy0);
1198 fjz0 = _fjsp_madd_v2r8(dz30,fscal,fjz0);
1200 gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1202 /* Inner loop uses 168 flops */
1205 /* End of innermost loop */
1207 gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1208 f+i_coord_offset,fshift+i_shift_offset);
1210 /* Increment number of inner iterations */
1211 inneriter += j_index_end - j_index_start;
1213 /* Outer loop uses 24 flops */
1216 /* Increment number of outer iterations */
1219 /* Update outer/inner flops */
1221 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_F,outeriter*24 + inneriter*168);