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_ElecRFCut_VdwLJSw_GeomP1P1_VF_sparc64_hpc_ace_double
53 * Electrostatics interaction: ReactionField
54 * VdW interaction: LennardJones
55 * Geometry: Particle-Particle
56 * Calculate force/pot: PotentialAndForce
59 nb_kernel_ElecRFCut_VdwLJSw_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);
94 _fjsp_v2r8 rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
95 real rswitch_scalar,d_scalar;
97 _fjsp_v2r8 dummy_mask,cutoff_mask;
98 _fjsp_v2r8 one = gmx_fjsp_set1_v2r8(1.0);
99 _fjsp_v2r8 two = gmx_fjsp_set1_v2r8(2.0);
100 union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
107 jindex = nlist->jindex;
109 shiftidx = nlist->shift;
111 shiftvec = fr->shift_vec[0];
112 fshift = fr->fshift[0];
113 facel = gmx_fjsp_set1_v2r8(fr->epsfac);
114 charge = mdatoms->chargeA;
115 krf = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
116 krf2 = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
117 crf = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
118 nvdwtype = fr->ntype;
120 vdwtype = mdatoms->typeA;
122 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
123 rcutoff_scalar = fr->rcoulomb;
124 rcutoff = gmx_fjsp_set1_v2r8(rcutoff_scalar);
125 rcutoff2 = _fjsp_mul_v2r8(rcutoff,rcutoff);
127 rswitch_scalar = fr->rvdw_switch;
128 rswitch = gmx_fjsp_set1_v2r8(rswitch_scalar);
129 /* Setup switch parameters */
130 d_scalar = rcutoff_scalar-rswitch_scalar;
131 d = gmx_fjsp_set1_v2r8(d_scalar);
132 swV3 = gmx_fjsp_set1_v2r8(-10.0/(d_scalar*d_scalar*d_scalar));
133 swV4 = gmx_fjsp_set1_v2r8( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
134 swV5 = gmx_fjsp_set1_v2r8( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
135 swF2 = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar));
136 swF3 = gmx_fjsp_set1_v2r8( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
137 swF4 = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
139 /* Avoid stupid compiler warnings */
147 /* Start outer loop over neighborlists */
148 for(iidx=0; iidx<nri; iidx++)
150 /* Load shift vector for this list */
151 i_shift_offset = DIM*shiftidx[iidx];
153 /* Load limits for loop over neighbors */
154 j_index_start = jindex[iidx];
155 j_index_end = jindex[iidx+1];
157 /* Get outer coordinate index */
159 i_coord_offset = DIM*inr;
161 /* Load i particle coords and add shift vector */
162 gmx_fjsp_load_shift_and_1rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
164 fix0 = _fjsp_setzero_v2r8();
165 fiy0 = _fjsp_setzero_v2r8();
166 fiz0 = _fjsp_setzero_v2r8();
168 /* Load parameters for i particles */
169 iq0 = _fjsp_mul_v2r8(facel,gmx_fjsp_load1_v2r8(charge+inr+0));
170 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
172 /* Reset potential sums */
173 velecsum = _fjsp_setzero_v2r8();
174 vvdwsum = _fjsp_setzero_v2r8();
176 /* Start inner kernel loop */
177 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
180 /* Get j neighbor index, and coordinate index */
183 j_coord_offsetA = DIM*jnrA;
184 j_coord_offsetB = DIM*jnrB;
186 /* load j atom coordinates */
187 gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
190 /* Calculate displacement vector */
191 dx00 = _fjsp_sub_v2r8(ix0,jx0);
192 dy00 = _fjsp_sub_v2r8(iy0,jy0);
193 dz00 = _fjsp_sub_v2r8(iz0,jz0);
195 /* Calculate squared distance and things based on it */
196 rsq00 = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
198 rinv00 = gmx_fjsp_invsqrt_v2r8(rsq00);
200 rinvsq00 = _fjsp_mul_v2r8(rinv00,rinv00);
202 /* Load parameters for j particles */
203 jq0 = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
204 vdwjidx0A = 2*vdwtype[jnrA+0];
205 vdwjidx0B = 2*vdwtype[jnrB+0];
207 /**************************
208 * CALCULATE INTERACTIONS *
209 **************************/
211 if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
214 r00 = _fjsp_mul_v2r8(rsq00,rinv00);
216 /* Compute parameters for interactions between i and j atoms */
217 qq00 = _fjsp_mul_v2r8(iq0,jq0);
218 gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
219 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
221 /* REACTION-FIELD ELECTROSTATICS */
222 velec = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq00,rinv00),crf));
223 felec = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
225 /* LENNARD-JONES DISPERSION/REPULSION */
227 rinvsix = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
228 vvdw6 = _fjsp_mul_v2r8(c6_00,rinvsix);
229 vvdw12 = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
230 vvdw = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
231 fvdw = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
233 d = _fjsp_sub_v2r8(r00,rswitch);
234 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
235 d2 = _fjsp_mul_v2r8(d,d);
236 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
238 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
240 /* Evaluate switch function */
241 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
242 fvdw = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
243 vvdw = _fjsp_mul_v2r8(vvdw,sw);
244 cutoff_mask = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
246 /* Update potential sum for this i atom from the interaction with this j atom. */
247 velec = _fjsp_and_v2r8(velec,cutoff_mask);
248 velecsum = _fjsp_add_v2r8(velecsum,velec);
249 vvdw = _fjsp_and_v2r8(vvdw,cutoff_mask);
250 vvdwsum = _fjsp_add_v2r8(vvdwsum,vvdw);
252 fscal = _fjsp_add_v2r8(felec,fvdw);
254 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
256 /* Update vectorial force */
257 fix0 = _fjsp_madd_v2r8(dx00,fscal,fix0);
258 fiy0 = _fjsp_madd_v2r8(dy00,fscal,fiy0);
259 fiz0 = _fjsp_madd_v2r8(dz00,fscal,fiz0);
261 gmx_fjsp_decrement_fma_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fscal,dx00,dy00,dz00);
265 /* Inner loop uses 73 flops */
272 j_coord_offsetA = DIM*jnrA;
274 /* load j atom coordinates */
275 gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
278 /* Calculate displacement vector */
279 dx00 = _fjsp_sub_v2r8(ix0,jx0);
280 dy00 = _fjsp_sub_v2r8(iy0,jy0);
281 dz00 = _fjsp_sub_v2r8(iz0,jz0);
283 /* Calculate squared distance and things based on it */
284 rsq00 = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
286 rinv00 = gmx_fjsp_invsqrt_v2r8(rsq00);
288 rinvsq00 = _fjsp_mul_v2r8(rinv00,rinv00);
290 /* Load parameters for j particles */
291 jq0 = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
292 vdwjidx0A = 2*vdwtype[jnrA+0];
294 /**************************
295 * CALCULATE INTERACTIONS *
296 **************************/
298 if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
301 r00 = _fjsp_mul_v2r8(rsq00,rinv00);
303 /* Compute parameters for interactions between i and j atoms */
304 qq00 = _fjsp_mul_v2r8(iq0,jq0);
305 gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
306 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
308 /* REACTION-FIELD ELECTROSTATICS */
309 velec = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq00,rinv00),crf));
310 felec = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
312 /* LENNARD-JONES DISPERSION/REPULSION */
314 rinvsix = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
315 vvdw6 = _fjsp_mul_v2r8(c6_00,rinvsix);
316 vvdw12 = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
317 vvdw = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
318 fvdw = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
320 d = _fjsp_sub_v2r8(r00,rswitch);
321 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
322 d2 = _fjsp_mul_v2r8(d,d);
323 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
325 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
327 /* Evaluate switch function */
328 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
329 fvdw = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
330 vvdw = _fjsp_mul_v2r8(vvdw,sw);
331 cutoff_mask = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
333 /* Update potential sum for this i atom from the interaction with this j atom. */
334 velec = _fjsp_and_v2r8(velec,cutoff_mask);
335 velec = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
336 velecsum = _fjsp_add_v2r8(velecsum,velec);
337 vvdw = _fjsp_and_v2r8(vvdw,cutoff_mask);
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_and_v2r8(fscal,cutoff_mask);
345 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
347 /* Update vectorial force */
348 fix0 = _fjsp_madd_v2r8(dx00,fscal,fix0);
349 fiy0 = _fjsp_madd_v2r8(dy00,fscal,fiy0);
350 fiz0 = _fjsp_madd_v2r8(dz00,fscal,fiz0);
352 gmx_fjsp_decrement_fma_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fscal,dx00,dy00,dz00);
356 /* Inner loop uses 73 flops */
359 /* End of innermost loop */
361 gmx_fjsp_update_iforce_1atom_swizzle_v2r8(fix0,fiy0,fiz0,
362 f+i_coord_offset,fshift+i_shift_offset);
365 /* Update potential energies */
366 gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
367 gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
369 /* Increment number of inner iterations */
370 inneriter += j_index_end - j_index_start;
372 /* Outer loop uses 9 flops */
375 /* Increment number of outer iterations */
378 /* Update outer/inner flops */
380 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_VF,outeriter*9 + inneriter*73);
383 * Gromacs nonbonded kernel: nb_kernel_ElecRFCut_VdwLJSw_GeomP1P1_F_sparc64_hpc_ace_double
384 * Electrostatics interaction: ReactionField
385 * VdW interaction: LennardJones
386 * Geometry: Particle-Particle
387 * Calculate force/pot: Force
390 nb_kernel_ElecRFCut_VdwLJSw_GeomP1P1_F_sparc64_hpc_ace_double
391 (t_nblist * gmx_restrict nlist,
392 rvec * gmx_restrict xx,
393 rvec * gmx_restrict ff,
394 t_forcerec * gmx_restrict fr,
395 t_mdatoms * gmx_restrict mdatoms,
396 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
397 t_nrnb * gmx_restrict nrnb)
399 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
400 * just 0 for non-waters.
401 * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
402 * jnr indices corresponding to data put in the four positions in the SIMD register.
404 int i_shift_offset,i_coord_offset,outeriter,inneriter;
405 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
407 int j_coord_offsetA,j_coord_offsetB;
408 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
410 real *shiftvec,*fshift,*x,*f;
411 _fjsp_v2r8 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
413 _fjsp_v2r8 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
414 int vdwjidx0A,vdwjidx0B;
415 _fjsp_v2r8 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
416 _fjsp_v2r8 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
417 _fjsp_v2r8 velec,felec,velecsum,facel,crf,krf,krf2;
420 _fjsp_v2r8 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
423 _fjsp_v2r8 one_sixth = gmx_fjsp_set1_v2r8(1.0/6.0);
424 _fjsp_v2r8 one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
425 _fjsp_v2r8 rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
426 real rswitch_scalar,d_scalar;
428 _fjsp_v2r8 dummy_mask,cutoff_mask;
429 _fjsp_v2r8 one = gmx_fjsp_set1_v2r8(1.0);
430 _fjsp_v2r8 two = gmx_fjsp_set1_v2r8(2.0);
431 union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
438 jindex = nlist->jindex;
440 shiftidx = nlist->shift;
442 shiftvec = fr->shift_vec[0];
443 fshift = fr->fshift[0];
444 facel = gmx_fjsp_set1_v2r8(fr->epsfac);
445 charge = mdatoms->chargeA;
446 krf = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
447 krf2 = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
448 crf = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
449 nvdwtype = fr->ntype;
451 vdwtype = mdatoms->typeA;
453 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
454 rcutoff_scalar = fr->rcoulomb;
455 rcutoff = gmx_fjsp_set1_v2r8(rcutoff_scalar);
456 rcutoff2 = _fjsp_mul_v2r8(rcutoff,rcutoff);
458 rswitch_scalar = fr->rvdw_switch;
459 rswitch = gmx_fjsp_set1_v2r8(rswitch_scalar);
460 /* Setup switch parameters */
461 d_scalar = rcutoff_scalar-rswitch_scalar;
462 d = gmx_fjsp_set1_v2r8(d_scalar);
463 swV3 = gmx_fjsp_set1_v2r8(-10.0/(d_scalar*d_scalar*d_scalar));
464 swV4 = gmx_fjsp_set1_v2r8( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
465 swV5 = gmx_fjsp_set1_v2r8( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
466 swF2 = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar));
467 swF3 = gmx_fjsp_set1_v2r8( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
468 swF4 = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
470 /* Avoid stupid compiler warnings */
478 /* Start outer loop over neighborlists */
479 for(iidx=0; iidx<nri; iidx++)
481 /* Load shift vector for this list */
482 i_shift_offset = DIM*shiftidx[iidx];
484 /* Load limits for loop over neighbors */
485 j_index_start = jindex[iidx];
486 j_index_end = jindex[iidx+1];
488 /* Get outer coordinate index */
490 i_coord_offset = DIM*inr;
492 /* Load i particle coords and add shift vector */
493 gmx_fjsp_load_shift_and_1rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
495 fix0 = _fjsp_setzero_v2r8();
496 fiy0 = _fjsp_setzero_v2r8();
497 fiz0 = _fjsp_setzero_v2r8();
499 /* Load parameters for i particles */
500 iq0 = _fjsp_mul_v2r8(facel,gmx_fjsp_load1_v2r8(charge+inr+0));
501 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
503 /* Start inner kernel loop */
504 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
507 /* Get j neighbor index, and coordinate index */
510 j_coord_offsetA = DIM*jnrA;
511 j_coord_offsetB = DIM*jnrB;
513 /* load j atom coordinates */
514 gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
517 /* Calculate displacement vector */
518 dx00 = _fjsp_sub_v2r8(ix0,jx0);
519 dy00 = _fjsp_sub_v2r8(iy0,jy0);
520 dz00 = _fjsp_sub_v2r8(iz0,jz0);
522 /* Calculate squared distance and things based on it */
523 rsq00 = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
525 rinv00 = gmx_fjsp_invsqrt_v2r8(rsq00);
527 rinvsq00 = _fjsp_mul_v2r8(rinv00,rinv00);
529 /* Load parameters for j particles */
530 jq0 = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
531 vdwjidx0A = 2*vdwtype[jnrA+0];
532 vdwjidx0B = 2*vdwtype[jnrB+0];
534 /**************************
535 * CALCULATE INTERACTIONS *
536 **************************/
538 if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
541 r00 = _fjsp_mul_v2r8(rsq00,rinv00);
543 /* Compute parameters for interactions between i and j atoms */
544 qq00 = _fjsp_mul_v2r8(iq0,jq0);
545 gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
546 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
548 /* REACTION-FIELD ELECTROSTATICS */
549 felec = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
551 /* LENNARD-JONES DISPERSION/REPULSION */
553 rinvsix = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
554 vvdw6 = _fjsp_mul_v2r8(c6_00,rinvsix);
555 vvdw12 = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
556 vvdw = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
557 fvdw = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
559 d = _fjsp_sub_v2r8(r00,rswitch);
560 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
561 d2 = _fjsp_mul_v2r8(d,d);
562 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
564 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
566 /* Evaluate switch function */
567 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
568 fvdw = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
569 cutoff_mask = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
571 fscal = _fjsp_add_v2r8(felec,fvdw);
573 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
575 /* Update vectorial force */
576 fix0 = _fjsp_madd_v2r8(dx00,fscal,fix0);
577 fiy0 = _fjsp_madd_v2r8(dy00,fscal,fiy0);
578 fiz0 = _fjsp_madd_v2r8(dz00,fscal,fiz0);
580 gmx_fjsp_decrement_fma_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fscal,dx00,dy00,dz00);
584 /* Inner loop uses 64 flops */
591 j_coord_offsetA = DIM*jnrA;
593 /* load j atom coordinates */
594 gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
597 /* Calculate displacement vector */
598 dx00 = _fjsp_sub_v2r8(ix0,jx0);
599 dy00 = _fjsp_sub_v2r8(iy0,jy0);
600 dz00 = _fjsp_sub_v2r8(iz0,jz0);
602 /* Calculate squared distance and things based on it */
603 rsq00 = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
605 rinv00 = gmx_fjsp_invsqrt_v2r8(rsq00);
607 rinvsq00 = _fjsp_mul_v2r8(rinv00,rinv00);
609 /* Load parameters for j particles */
610 jq0 = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
611 vdwjidx0A = 2*vdwtype[jnrA+0];
613 /**************************
614 * CALCULATE INTERACTIONS *
615 **************************/
617 if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
620 r00 = _fjsp_mul_v2r8(rsq00,rinv00);
622 /* Compute parameters for interactions between i and j atoms */
623 qq00 = _fjsp_mul_v2r8(iq0,jq0);
624 gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
625 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
627 /* REACTION-FIELD ELECTROSTATICS */
628 felec = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
630 /* LENNARD-JONES DISPERSION/REPULSION */
632 rinvsix = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
633 vvdw6 = _fjsp_mul_v2r8(c6_00,rinvsix);
634 vvdw12 = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
635 vvdw = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
636 fvdw = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
638 d = _fjsp_sub_v2r8(r00,rswitch);
639 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
640 d2 = _fjsp_mul_v2r8(d,d);
641 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
643 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
645 /* Evaluate switch function */
646 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
647 fvdw = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
648 cutoff_mask = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
650 fscal = _fjsp_add_v2r8(felec,fvdw);
652 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
654 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
656 /* Update vectorial force */
657 fix0 = _fjsp_madd_v2r8(dx00,fscal,fix0);
658 fiy0 = _fjsp_madd_v2r8(dy00,fscal,fiy0);
659 fiz0 = _fjsp_madd_v2r8(dz00,fscal,fiz0);
661 gmx_fjsp_decrement_fma_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fscal,dx00,dy00,dz00);
665 /* Inner loop uses 64 flops */
668 /* End of innermost loop */
670 gmx_fjsp_update_iforce_1atom_swizzle_v2r8(fix0,fiy0,fiz0,
671 f+i_coord_offset,fshift+i_shift_offset);
673 /* Increment number of inner iterations */
674 inneriter += j_index_end - j_index_start;
676 /* Outer loop uses 7 flops */
679 /* Increment number of outer iterations */
682 /* Update outer/inner flops */
684 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_F,outeriter*7 + inneriter*64);