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_GeomP1P1_VF_sparc64_hpc_ace_double
53 * Electrostatics interaction: Ewald
54 * VdW interaction: LJEwald
55 * Geometry: Particle-Particle
56 * Calculate force/pot: PotentialAndForce
59 nb_kernel_ElecEw_VdwLJEw_GeomP1P1_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;
83 int vdwjidx0A,vdwjidx0B;
84 _fjsp_v2r8 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
85 _fjsp_v2r8 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
86 _fjsp_v2r8 velec,felec,velecsum,facel,crf,krf,krf2;
89 _fjsp_v2r8 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
92 _fjsp_v2r8 one_sixth = gmx_fjsp_set1_v2r8(1.0/6.0);
93 _fjsp_v2r8 one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
96 _fjsp_v2r8 ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
97 _fjsp_v2r8 one_half = gmx_fjsp_set1_v2r8(0.5);
98 _fjsp_v2r8 minus_one = gmx_fjsp_set1_v2r8(-1.0);
99 _fjsp_v2r8 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
102 _fjsp_v2r8 dummy_mask,cutoff_mask;
103 _fjsp_v2r8 one = gmx_fjsp_set1_v2r8(1.0);
104 _fjsp_v2r8 two = gmx_fjsp_set1_v2r8(2.0);
105 union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
112 jindex = nlist->jindex;
114 shiftidx = nlist->shift;
116 shiftvec = fr->shift_vec[0];
117 fshift = fr->fshift[0];
118 facel = gmx_fjsp_set1_v2r8(fr->epsfac);
119 charge = mdatoms->chargeA;
120 nvdwtype = fr->ntype;
122 vdwtype = mdatoms->typeA;
123 vdwgridparam = fr->ljpme_c6grid;
124 sh_lj_ewald = gmx_fjsp_set1_v2r8(fr->ic->sh_lj_ewald);
125 ewclj = gmx_fjsp_set1_v2r8(fr->ewaldcoeff_lj);
126 ewclj2 = _fjsp_mul_v2r8(minus_one,_fjsp_mul_v2r8(ewclj,ewclj));
128 sh_ewald = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
129 ewtab = fr->ic->tabq_coul_FDV0;
130 ewtabscale = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
131 ewtabhalfspace = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
133 /* Avoid stupid compiler warnings */
141 /* Start outer loop over neighborlists */
142 for(iidx=0; iidx<nri; iidx++)
144 /* Load shift vector for this list */
145 i_shift_offset = DIM*shiftidx[iidx];
147 /* Load limits for loop over neighbors */
148 j_index_start = jindex[iidx];
149 j_index_end = jindex[iidx+1];
151 /* Get outer coordinate index */
153 i_coord_offset = DIM*inr;
155 /* Load i particle coords and add shift vector */
156 gmx_fjsp_load_shift_and_1rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
158 fix0 = _fjsp_setzero_v2r8();
159 fiy0 = _fjsp_setzero_v2r8();
160 fiz0 = _fjsp_setzero_v2r8();
162 /* Load parameters for i particles */
163 iq0 = _fjsp_mul_v2r8(facel,gmx_fjsp_load1_v2r8(charge+inr+0));
164 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
166 /* Reset potential sums */
167 velecsum = _fjsp_setzero_v2r8();
168 vvdwsum = _fjsp_setzero_v2r8();
170 /* Start inner kernel loop */
171 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
174 /* Get j neighbor index, and coordinate index */
177 j_coord_offsetA = DIM*jnrA;
178 j_coord_offsetB = DIM*jnrB;
180 /* load j atom coordinates */
181 gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
184 /* Calculate displacement vector */
185 dx00 = _fjsp_sub_v2r8(ix0,jx0);
186 dy00 = _fjsp_sub_v2r8(iy0,jy0);
187 dz00 = _fjsp_sub_v2r8(iz0,jz0);
189 /* Calculate squared distance and things based on it */
190 rsq00 = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
192 rinv00 = gmx_fjsp_invsqrt_v2r8(rsq00);
194 rinvsq00 = _fjsp_mul_v2r8(rinv00,rinv00);
196 /* Load parameters for j particles */
197 jq0 = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
198 vdwjidx0A = 2*vdwtype[jnrA+0];
199 vdwjidx0B = 2*vdwtype[jnrB+0];
201 /**************************
202 * CALCULATE INTERACTIONS *
203 **************************/
205 r00 = _fjsp_mul_v2r8(rsq00,rinv00);
207 /* Compute parameters for interactions between i and j atoms */
208 qq00 = _fjsp_mul_v2r8(iq0,jq0);
209 gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
210 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
212 c6grid_00 = gmx_fjsp_load_2real_swizzle_v2r8(vdwgridparam+vdwioffset0+vdwjidx0A,
213 vdwgridparam+vdwioffset0+vdwjidx0B);
215 /* EWALD ELECTROSTATICS */
217 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
218 ewrt = _fjsp_mul_v2r8(r00,ewtabscale);
219 itab_tmp = _fjsp_dtox_v2r8(ewrt);
220 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
221 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
223 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
224 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
225 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
226 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
227 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
228 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
229 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
230 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
231 velec = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(rinv00,velec));
232 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
234 /* Analytical LJ-PME */
235 rinvsix = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
236 ewcljrsq = _fjsp_mul_v2r8(ewclj2,rsq00);
237 ewclj6 = _fjsp_mul_v2r8(ewclj2,_fjsp_mul_v2r8(ewclj2,ewclj2));
238 exponent = gmx_simd_exp_d(-ewcljrsq);
239 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
240 poly = _fjsp_mul_v2r8(exponent,_fjsp_madd_v2r8(_fjsp_mul_v2r8(ewcljrsq,ewcljrsq),one_half,_fjsp_sub_v2r8(one,ewcljrsq)));
241 /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
242 vvdw6 = _fjsp_mul_v2r8(_fjsp_madd_v2r8(-c6grid_00,_fjsp_sub_v2r8(one,poly),c6_00),rinvsix);
243 vvdw12 = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
244 vvdw = _fjsp_msub_v2r8(vvdw12,one_twelfth,_fjsp_mul_v2r8(vvdw6,one_sixth));
245 /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
246 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);
248 /* Update potential sum for this i atom from the interaction with this j atom. */
249 velecsum = _fjsp_add_v2r8(velecsum,velec);
250 vvdwsum = _fjsp_add_v2r8(vvdwsum,vvdw);
252 fscal = _fjsp_add_v2r8(felec,fvdw);
254 /* Update vectorial force */
255 fix0 = _fjsp_madd_v2r8(dx00,fscal,fix0);
256 fiy0 = _fjsp_madd_v2r8(dy00,fscal,fiy0);
257 fiz0 = _fjsp_madd_v2r8(dz00,fscal,fiz0);
259 gmx_fjsp_decrement_fma_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fscal,dx00,dy00,dz00);
261 /* Inner loop uses 68 flops */
268 j_coord_offsetA = DIM*jnrA;
270 /* load j atom coordinates */
271 gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
274 /* Calculate displacement vector */
275 dx00 = _fjsp_sub_v2r8(ix0,jx0);
276 dy00 = _fjsp_sub_v2r8(iy0,jy0);
277 dz00 = _fjsp_sub_v2r8(iz0,jz0);
279 /* Calculate squared distance and things based on it */
280 rsq00 = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
282 rinv00 = gmx_fjsp_invsqrt_v2r8(rsq00);
284 rinvsq00 = _fjsp_mul_v2r8(rinv00,rinv00);
286 /* Load parameters for j particles */
287 jq0 = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
288 vdwjidx0A = 2*vdwtype[jnrA+0];
290 /**************************
291 * CALCULATE INTERACTIONS *
292 **************************/
294 r00 = _fjsp_mul_v2r8(rsq00,rinv00);
296 /* Compute parameters for interactions between i and j atoms */
297 qq00 = _fjsp_mul_v2r8(iq0,jq0);
298 gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
300 c6grid_00 = gmx_fjsp_load_1real_swizzle_v2r8(vdwgridparam+vdwioffset0+vdwjidx0A);
302 /* EWALD ELECTROSTATICS */
304 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
305 ewrt = _fjsp_mul_v2r8(r00,ewtabscale);
306 itab_tmp = _fjsp_dtox_v2r8(ewrt);
307 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
308 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
310 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
311 ewtabD = _fjsp_setzero_v2r8();
312 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
313 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
314 ewtabFn = _fjsp_setzero_v2r8();
315 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
316 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
317 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
318 velec = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(rinv00,velec));
319 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
321 /* Analytical LJ-PME */
322 rinvsix = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
323 ewcljrsq = _fjsp_mul_v2r8(ewclj2,rsq00);
324 ewclj6 = _fjsp_mul_v2r8(ewclj2,_fjsp_mul_v2r8(ewclj2,ewclj2));
325 exponent = gmx_simd_exp_d(-ewcljrsq);
326 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
327 poly = _fjsp_mul_v2r8(exponent,_fjsp_madd_v2r8(_fjsp_mul_v2r8(ewcljrsq,ewcljrsq),one_half,_fjsp_sub_v2r8(one,ewcljrsq)));
328 /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
329 vvdw6 = _fjsp_mul_v2r8(_fjsp_madd_v2r8(-c6grid_00,_fjsp_sub_v2r8(one,poly),c6_00),rinvsix);
330 vvdw12 = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
331 vvdw = _fjsp_msub_v2r8(vvdw12,one_twelfth,_fjsp_mul_v2r8(vvdw6,one_sixth));
332 /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
333 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);
335 /* Update potential sum for this i atom from the interaction with this j atom. */
336 velec = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
337 velecsum = _fjsp_add_v2r8(velecsum,velec);
338 vvdw = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
339 vvdwsum = _fjsp_add_v2r8(vvdwsum,vvdw);
341 fscal = _fjsp_add_v2r8(felec,fvdw);
343 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
345 /* Update vectorial force */
346 fix0 = _fjsp_madd_v2r8(dx00,fscal,fix0);
347 fiy0 = _fjsp_madd_v2r8(dy00,fscal,fiy0);
348 fiz0 = _fjsp_madd_v2r8(dz00,fscal,fiz0);
350 gmx_fjsp_decrement_fma_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fscal,dx00,dy00,dz00);
352 /* Inner loop uses 68 flops */
355 /* End of innermost loop */
357 gmx_fjsp_update_iforce_1atom_swizzle_v2r8(fix0,fiy0,fiz0,
358 f+i_coord_offset,fshift+i_shift_offset);
361 /* Update potential energies */
362 gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
363 gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
365 /* Increment number of inner iterations */
366 inneriter += j_index_end - j_index_start;
368 /* Outer loop uses 9 flops */
371 /* Increment number of outer iterations */
374 /* Update outer/inner flops */
376 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_VF,outeriter*9 + inneriter*68);
379 * Gromacs nonbonded kernel: nb_kernel_ElecEw_VdwLJEw_GeomP1P1_F_sparc64_hpc_ace_double
380 * Electrostatics interaction: Ewald
381 * VdW interaction: LJEwald
382 * Geometry: Particle-Particle
383 * Calculate force/pot: Force
386 nb_kernel_ElecEw_VdwLJEw_GeomP1P1_F_sparc64_hpc_ace_double
387 (t_nblist * gmx_restrict nlist,
388 rvec * gmx_restrict xx,
389 rvec * gmx_restrict ff,
390 t_forcerec * gmx_restrict fr,
391 t_mdatoms * gmx_restrict mdatoms,
392 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
393 t_nrnb * gmx_restrict nrnb)
395 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
396 * just 0 for non-waters.
397 * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
398 * jnr indices corresponding to data put in the four positions in the SIMD register.
400 int i_shift_offset,i_coord_offset,outeriter,inneriter;
401 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
403 int j_coord_offsetA,j_coord_offsetB;
404 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
406 real *shiftvec,*fshift,*x,*f;
407 _fjsp_v2r8 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
409 _fjsp_v2r8 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
410 int vdwjidx0A,vdwjidx0B;
411 _fjsp_v2r8 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
412 _fjsp_v2r8 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
413 _fjsp_v2r8 velec,felec,velecsum,facel,crf,krf,krf2;
416 _fjsp_v2r8 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
419 _fjsp_v2r8 one_sixth = gmx_fjsp_set1_v2r8(1.0/6.0);
420 _fjsp_v2r8 one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
421 _fjsp_v2r8 c6grid_00;
423 _fjsp_v2r8 ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
424 _fjsp_v2r8 one_half = gmx_fjsp_set1_v2r8(0.5);
425 _fjsp_v2r8 minus_one = gmx_fjsp_set1_v2r8(-1.0);
426 _fjsp_v2r8 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
429 _fjsp_v2r8 dummy_mask,cutoff_mask;
430 _fjsp_v2r8 one = gmx_fjsp_set1_v2r8(1.0);
431 _fjsp_v2r8 two = gmx_fjsp_set1_v2r8(2.0);
432 union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
439 jindex = nlist->jindex;
441 shiftidx = nlist->shift;
443 shiftvec = fr->shift_vec[0];
444 fshift = fr->fshift[0];
445 facel = gmx_fjsp_set1_v2r8(fr->epsfac);
446 charge = mdatoms->chargeA;
447 nvdwtype = fr->ntype;
449 vdwtype = mdatoms->typeA;
450 vdwgridparam = fr->ljpme_c6grid;
451 sh_lj_ewald = gmx_fjsp_set1_v2r8(fr->ic->sh_lj_ewald);
452 ewclj = gmx_fjsp_set1_v2r8(fr->ewaldcoeff_lj);
453 ewclj2 = _fjsp_mul_v2r8(minus_one,_fjsp_mul_v2r8(ewclj,ewclj));
455 sh_ewald = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
456 ewtab = fr->ic->tabq_coul_F;
457 ewtabscale = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
458 ewtabhalfspace = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
460 /* Avoid stupid compiler warnings */
468 /* Start outer loop over neighborlists */
469 for(iidx=0; iidx<nri; iidx++)
471 /* Load shift vector for this list */
472 i_shift_offset = DIM*shiftidx[iidx];
474 /* Load limits for loop over neighbors */
475 j_index_start = jindex[iidx];
476 j_index_end = jindex[iidx+1];
478 /* Get outer coordinate index */
480 i_coord_offset = DIM*inr;
482 /* Load i particle coords and add shift vector */
483 gmx_fjsp_load_shift_and_1rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
485 fix0 = _fjsp_setzero_v2r8();
486 fiy0 = _fjsp_setzero_v2r8();
487 fiz0 = _fjsp_setzero_v2r8();
489 /* Load parameters for i particles */
490 iq0 = _fjsp_mul_v2r8(facel,gmx_fjsp_load1_v2r8(charge+inr+0));
491 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
493 /* Start inner kernel loop */
494 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
497 /* Get j neighbor index, and coordinate index */
500 j_coord_offsetA = DIM*jnrA;
501 j_coord_offsetB = DIM*jnrB;
503 /* load j atom coordinates */
504 gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
507 /* Calculate displacement vector */
508 dx00 = _fjsp_sub_v2r8(ix0,jx0);
509 dy00 = _fjsp_sub_v2r8(iy0,jy0);
510 dz00 = _fjsp_sub_v2r8(iz0,jz0);
512 /* Calculate squared distance and things based on it */
513 rsq00 = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
515 rinv00 = gmx_fjsp_invsqrt_v2r8(rsq00);
517 rinvsq00 = _fjsp_mul_v2r8(rinv00,rinv00);
519 /* Load parameters for j particles */
520 jq0 = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
521 vdwjidx0A = 2*vdwtype[jnrA+0];
522 vdwjidx0B = 2*vdwtype[jnrB+0];
524 /**************************
525 * CALCULATE INTERACTIONS *
526 **************************/
528 r00 = _fjsp_mul_v2r8(rsq00,rinv00);
530 /* Compute parameters for interactions between i and j atoms */
531 qq00 = _fjsp_mul_v2r8(iq0,jq0);
532 gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
533 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
535 c6grid_00 = gmx_fjsp_load_2real_swizzle_v2r8(vdwgridparam+vdwioffset0+vdwjidx0A,
536 vdwgridparam+vdwioffset0+vdwjidx0B);
538 /* EWALD ELECTROSTATICS */
540 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
541 ewrt = _fjsp_mul_v2r8(r00,ewtabscale);
542 itab_tmp = _fjsp_dtox_v2r8(ewrt);
543 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
544 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
546 gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
548 felec = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
549 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
551 /* Analytical LJ-PME */
552 rinvsix = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
553 ewcljrsq = _fjsp_mul_v2r8(ewclj2,rsq00);
554 ewclj6 = _fjsp_mul_v2r8(ewclj2,_fjsp_mul_v2r8(ewclj2,ewclj2));
555 exponent = gmx_simd_exp_d(-ewcljrsq);
556 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
557 poly = _fjsp_mul_v2r8(exponent,_fjsp_madd_v2r8(_fjsp_mul_v2r8(ewcljrsq,ewcljrsq),one_half,_fjsp_sub_v2r8(one,ewcljrsq)));
558 /* f6A = 6 * C6grid * (1 - poly) */
559 f6A = _fjsp_mul_v2r8(c6grid_00,_fjsp_msub_v2r8(one,poly));
560 /* f6B = C6grid * exponent * beta^6 */
561 f6B = _fjsp_mul_v2r8(_fjsp_mul_v2r8(c6grid_00,one_sixth),_fjsp_mul_v2r8(exponent,ewclj6));
562 /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
563 fvdw = _fjsp_mul_v2r8(_fjsp_madd_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,_fjsp_sub_v2r8(c6_00,f6A)),rinvsix,f6B),rinvsq00);
565 fscal = _fjsp_add_v2r8(felec,fvdw);
567 /* Update vectorial force */
568 fix0 = _fjsp_madd_v2r8(dx00,fscal,fix0);
569 fiy0 = _fjsp_madd_v2r8(dy00,fscal,fiy0);
570 fiz0 = _fjsp_madd_v2r8(dz00,fscal,fiz0);
572 gmx_fjsp_decrement_fma_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fscal,dx00,dy00,dz00);
574 /* Inner loop uses 61 flops */
581 j_coord_offsetA = DIM*jnrA;
583 /* load j atom coordinates */
584 gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
587 /* Calculate displacement vector */
588 dx00 = _fjsp_sub_v2r8(ix0,jx0);
589 dy00 = _fjsp_sub_v2r8(iy0,jy0);
590 dz00 = _fjsp_sub_v2r8(iz0,jz0);
592 /* Calculate squared distance and things based on it */
593 rsq00 = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
595 rinv00 = gmx_fjsp_invsqrt_v2r8(rsq00);
597 rinvsq00 = _fjsp_mul_v2r8(rinv00,rinv00);
599 /* Load parameters for j particles */
600 jq0 = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
601 vdwjidx0A = 2*vdwtype[jnrA+0];
603 /**************************
604 * CALCULATE INTERACTIONS *
605 **************************/
607 r00 = _fjsp_mul_v2r8(rsq00,rinv00);
609 /* Compute parameters for interactions between i and j atoms */
610 qq00 = _fjsp_mul_v2r8(iq0,jq0);
611 gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
613 c6grid_00 = gmx_fjsp_load_1real_swizzle_v2r8(vdwgridparam+vdwioffset0+vdwjidx0A);
615 /* EWALD ELECTROSTATICS */
617 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
618 ewrt = _fjsp_mul_v2r8(r00,ewtabscale);
619 itab_tmp = _fjsp_dtox_v2r8(ewrt);
620 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
621 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
623 gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
624 felec = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
625 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
627 /* Analytical LJ-PME */
628 rinvsix = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
629 ewcljrsq = _fjsp_mul_v2r8(ewclj2,rsq00);
630 ewclj6 = _fjsp_mul_v2r8(ewclj2,_fjsp_mul_v2r8(ewclj2,ewclj2));
631 exponent = gmx_simd_exp_d(-ewcljrsq);
632 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
633 poly = _fjsp_mul_v2r8(exponent,_fjsp_madd_v2r8(_fjsp_mul_v2r8(ewcljrsq,ewcljrsq),one_half,_fjsp_sub_v2r8(one,ewcljrsq)));
634 /* f6A = 6 * C6grid * (1 - poly) */
635 f6A = _fjsp_mul_v2r8(c6grid_00,_fjsp_msub_v2r8(one,poly));
636 /* f6B = C6grid * exponent * beta^6 */
637 f6B = _fjsp_mul_v2r8(_fjsp_mul_v2r8(c6grid_00,one_sixth),_fjsp_mul_v2r8(exponent,ewclj6));
638 /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
639 fvdw = _fjsp_mul_v2r8(_fjsp_madd_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,_fjsp_sub_v2r8(c6_00,f6A)),rinvsix,f6B),rinvsq00);
641 fscal = _fjsp_add_v2r8(felec,fvdw);
643 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
645 /* Update vectorial force */
646 fix0 = _fjsp_madd_v2r8(dx00,fscal,fix0);
647 fiy0 = _fjsp_madd_v2r8(dy00,fscal,fiy0);
648 fiz0 = _fjsp_madd_v2r8(dz00,fscal,fiz0);
650 gmx_fjsp_decrement_fma_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fscal,dx00,dy00,dz00);
652 /* Inner loop uses 61 flops */
655 /* End of innermost loop */
657 gmx_fjsp_update_iforce_1atom_swizzle_v2r8(fix0,fiy0,fiz0,
658 f+i_coord_offset,fshift+i_shift_offset);
660 /* Increment number of inner iterations */
661 inneriter += j_index_end - j_index_start;
663 /* Outer loop uses 7 flops */
666 /* Increment number of outer iterations */
669 /* Update outer/inner flops */
671 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_F,outeriter*7 + inneriter*61);