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 "types/simple.h"
46 #include "gromacs/math/vec.h"
49 #include "kernelutil_sparc64_hpc_ace_double.h"
52 * Gromacs nonbonded kernel: nb_kernel_ElecEwSw_VdwNone_GeomW3P1_VF_sparc64_hpc_ace_double
53 * Electrostatics interaction: Ewald
54 * VdW interaction: None
55 * Geometry: Water3-Particle
56 * Calculate force/pot: PotentialAndForce
59 nb_kernel_ElecEwSw_VdwNone_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;
94 _fjsp_v2r8 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
96 _fjsp_v2r8 rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
97 real rswitch_scalar,d_scalar;
99 _fjsp_v2r8 dummy_mask,cutoff_mask;
100 _fjsp_v2r8 one = gmx_fjsp_set1_v2r8(1.0);
101 _fjsp_v2r8 two = gmx_fjsp_set1_v2r8(2.0);
102 union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
109 jindex = nlist->jindex;
111 shiftidx = nlist->shift;
113 shiftvec = fr->shift_vec[0];
114 fshift = fr->fshift[0];
115 facel = gmx_fjsp_set1_v2r8(fr->epsfac);
116 charge = mdatoms->chargeA;
118 sh_ewald = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
119 ewtab = fr->ic->tabq_coul_FDV0;
120 ewtabscale = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
121 ewtabhalfspace = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
123 /* Setup water-specific parameters */
124 inr = nlist->iinr[0];
125 iq0 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
126 iq1 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
127 iq2 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
129 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
130 rcutoff_scalar = fr->rcoulomb;
131 rcutoff = gmx_fjsp_set1_v2r8(rcutoff_scalar);
132 rcutoff2 = _fjsp_mul_v2r8(rcutoff,rcutoff);
134 rswitch_scalar = fr->rcoulomb_switch;
135 rswitch = gmx_fjsp_set1_v2r8(rswitch_scalar);
136 /* Setup switch parameters */
137 d_scalar = rcutoff_scalar-rswitch_scalar;
138 d = gmx_fjsp_set1_v2r8(d_scalar);
139 swV3 = gmx_fjsp_set1_v2r8(-10.0/(d_scalar*d_scalar*d_scalar));
140 swV4 = gmx_fjsp_set1_v2r8( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
141 swV5 = gmx_fjsp_set1_v2r8( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
142 swF2 = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar));
143 swF3 = gmx_fjsp_set1_v2r8( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
144 swF4 = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
146 /* Avoid stupid compiler warnings */
154 /* Start outer loop over neighborlists */
155 for(iidx=0; iidx<nri; iidx++)
157 /* Load shift vector for this list */
158 i_shift_offset = DIM*shiftidx[iidx];
160 /* Load limits for loop over neighbors */
161 j_index_start = jindex[iidx];
162 j_index_end = jindex[iidx+1];
164 /* Get outer coordinate index */
166 i_coord_offset = DIM*inr;
168 /* Load i particle coords and add shift vector */
169 gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
170 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
172 fix0 = _fjsp_setzero_v2r8();
173 fiy0 = _fjsp_setzero_v2r8();
174 fiz0 = _fjsp_setzero_v2r8();
175 fix1 = _fjsp_setzero_v2r8();
176 fiy1 = _fjsp_setzero_v2r8();
177 fiz1 = _fjsp_setzero_v2r8();
178 fix2 = _fjsp_setzero_v2r8();
179 fiy2 = _fjsp_setzero_v2r8();
180 fiz2 = _fjsp_setzero_v2r8();
182 /* Reset potential sums */
183 velecsum = _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);
226 fjx0 = _fjsp_setzero_v2r8();
227 fjy0 = _fjsp_setzero_v2r8();
228 fjz0 = _fjsp_setzero_v2r8();
230 /**************************
231 * CALCULATE INTERACTIONS *
232 **************************/
234 if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
237 r00 = _fjsp_mul_v2r8(rsq00,rinv00);
239 /* Compute parameters for interactions between i and j atoms */
240 qq00 = _fjsp_mul_v2r8(iq0,jq0);
242 /* EWALD ELECTROSTATICS */
244 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
245 ewrt = _fjsp_mul_v2r8(r00,ewtabscale);
246 itab_tmp = _fjsp_dtox_v2r8(ewrt);
247 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
248 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
250 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
251 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
252 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
253 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
254 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
255 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
256 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
257 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
258 velec = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(rinv00,velec));
259 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
261 d = _fjsp_sub_v2r8(r00,rswitch);
262 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
263 d2 = _fjsp_mul_v2r8(d,d);
264 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
266 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
268 /* Evaluate switch function */
269 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
270 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(velec,dsw)) );
271 velec = _fjsp_mul_v2r8(velec,sw);
272 cutoff_mask = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
274 /* Update potential sum for this i atom from the interaction with this j atom. */
275 velec = _fjsp_and_v2r8(velec,cutoff_mask);
276 velecsum = _fjsp_add_v2r8(velecsum,velec);
280 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
282 /* Update vectorial force */
283 fix0 = _fjsp_madd_v2r8(dx00,fscal,fix0);
284 fiy0 = _fjsp_madd_v2r8(dy00,fscal,fiy0);
285 fiz0 = _fjsp_madd_v2r8(dz00,fscal,fiz0);
287 fjx0 = _fjsp_madd_v2r8(dx00,fscal,fjx0);
288 fjy0 = _fjsp_madd_v2r8(dy00,fscal,fjy0);
289 fjz0 = _fjsp_madd_v2r8(dz00,fscal,fjz0);
293 /**************************
294 * CALCULATE INTERACTIONS *
295 **************************/
297 if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
300 r10 = _fjsp_mul_v2r8(rsq10,rinv10);
302 /* Compute parameters for interactions between i and j atoms */
303 qq10 = _fjsp_mul_v2r8(iq1,jq0);
305 /* EWALD ELECTROSTATICS */
307 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
308 ewrt = _fjsp_mul_v2r8(r10,ewtabscale);
309 itab_tmp = _fjsp_dtox_v2r8(ewrt);
310 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
311 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
313 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
314 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
315 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
316 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
317 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
318 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
319 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
320 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
321 velec = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
322 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
324 d = _fjsp_sub_v2r8(r10,rswitch);
325 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
326 d2 = _fjsp_mul_v2r8(d,d);
327 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
329 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
331 /* Evaluate switch function */
332 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
333 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv10,_fjsp_mul_v2r8(velec,dsw)) );
334 velec = _fjsp_mul_v2r8(velec,sw);
335 cutoff_mask = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
337 /* Update potential sum for this i atom from the interaction with this j atom. */
338 velec = _fjsp_and_v2r8(velec,cutoff_mask);
339 velecsum = _fjsp_add_v2r8(velecsum,velec);
343 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
345 /* Update vectorial force */
346 fix1 = _fjsp_madd_v2r8(dx10,fscal,fix1);
347 fiy1 = _fjsp_madd_v2r8(dy10,fscal,fiy1);
348 fiz1 = _fjsp_madd_v2r8(dz10,fscal,fiz1);
350 fjx0 = _fjsp_madd_v2r8(dx10,fscal,fjx0);
351 fjy0 = _fjsp_madd_v2r8(dy10,fscal,fjy0);
352 fjz0 = _fjsp_madd_v2r8(dz10,fscal,fjz0);
356 /**************************
357 * CALCULATE INTERACTIONS *
358 **************************/
360 if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
363 r20 = _fjsp_mul_v2r8(rsq20,rinv20);
365 /* Compute parameters for interactions between i and j atoms */
366 qq20 = _fjsp_mul_v2r8(iq2,jq0);
368 /* EWALD ELECTROSTATICS */
370 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
371 ewrt = _fjsp_mul_v2r8(r20,ewtabscale);
372 itab_tmp = _fjsp_dtox_v2r8(ewrt);
373 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
374 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
376 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
377 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
378 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
379 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
380 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
381 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
382 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
383 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
384 velec = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
385 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
387 d = _fjsp_sub_v2r8(r20,rswitch);
388 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
389 d2 = _fjsp_mul_v2r8(d,d);
390 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
392 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
394 /* Evaluate switch function */
395 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
396 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv20,_fjsp_mul_v2r8(velec,dsw)) );
397 velec = _fjsp_mul_v2r8(velec,sw);
398 cutoff_mask = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
400 /* Update potential sum for this i atom from the interaction with this j atom. */
401 velec = _fjsp_and_v2r8(velec,cutoff_mask);
402 velecsum = _fjsp_add_v2r8(velecsum,velec);
406 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
408 /* Update vectorial force */
409 fix2 = _fjsp_madd_v2r8(dx20,fscal,fix2);
410 fiy2 = _fjsp_madd_v2r8(dy20,fscal,fiy2);
411 fiz2 = _fjsp_madd_v2r8(dz20,fscal,fiz2);
413 fjx0 = _fjsp_madd_v2r8(dx20,fscal,fjx0);
414 fjy0 = _fjsp_madd_v2r8(dy20,fscal,fjy0);
415 fjz0 = _fjsp_madd_v2r8(dz20,fscal,fjz0);
419 gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
421 /* Inner loop uses 207 flops */
428 j_coord_offsetA = DIM*jnrA;
430 /* load j atom coordinates */
431 gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
434 /* Calculate displacement vector */
435 dx00 = _fjsp_sub_v2r8(ix0,jx0);
436 dy00 = _fjsp_sub_v2r8(iy0,jy0);
437 dz00 = _fjsp_sub_v2r8(iz0,jz0);
438 dx10 = _fjsp_sub_v2r8(ix1,jx0);
439 dy10 = _fjsp_sub_v2r8(iy1,jy0);
440 dz10 = _fjsp_sub_v2r8(iz1,jz0);
441 dx20 = _fjsp_sub_v2r8(ix2,jx0);
442 dy20 = _fjsp_sub_v2r8(iy2,jy0);
443 dz20 = _fjsp_sub_v2r8(iz2,jz0);
445 /* Calculate squared distance and things based on it */
446 rsq00 = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
447 rsq10 = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
448 rsq20 = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
450 rinv00 = gmx_fjsp_invsqrt_v2r8(rsq00);
451 rinv10 = gmx_fjsp_invsqrt_v2r8(rsq10);
452 rinv20 = gmx_fjsp_invsqrt_v2r8(rsq20);
454 rinvsq00 = _fjsp_mul_v2r8(rinv00,rinv00);
455 rinvsq10 = _fjsp_mul_v2r8(rinv10,rinv10);
456 rinvsq20 = _fjsp_mul_v2r8(rinv20,rinv20);
458 /* Load parameters for j particles */
459 jq0 = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
461 fjx0 = _fjsp_setzero_v2r8();
462 fjy0 = _fjsp_setzero_v2r8();
463 fjz0 = _fjsp_setzero_v2r8();
465 /**************************
466 * CALCULATE INTERACTIONS *
467 **************************/
469 if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
472 r00 = _fjsp_mul_v2r8(rsq00,rinv00);
474 /* Compute parameters for interactions between i and j atoms */
475 qq00 = _fjsp_mul_v2r8(iq0,jq0);
477 /* EWALD ELECTROSTATICS */
479 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
480 ewrt = _fjsp_mul_v2r8(r00,ewtabscale);
481 itab_tmp = _fjsp_dtox_v2r8(ewrt);
482 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
483 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
485 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
486 ewtabD = _fjsp_setzero_v2r8();
487 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
488 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
489 ewtabFn = _fjsp_setzero_v2r8();
490 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
491 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
492 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
493 velec = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(rinv00,velec));
494 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
496 d = _fjsp_sub_v2r8(r00,rswitch);
497 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
498 d2 = _fjsp_mul_v2r8(d,d);
499 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
501 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
503 /* Evaluate switch function */
504 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
505 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(velec,dsw)) );
506 velec = _fjsp_mul_v2r8(velec,sw);
507 cutoff_mask = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
509 /* Update potential sum for this i atom from the interaction with this j atom. */
510 velec = _fjsp_and_v2r8(velec,cutoff_mask);
511 velec = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
512 velecsum = _fjsp_add_v2r8(velecsum,velec);
516 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
518 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
520 /* Update vectorial force */
521 fix0 = _fjsp_madd_v2r8(dx00,fscal,fix0);
522 fiy0 = _fjsp_madd_v2r8(dy00,fscal,fiy0);
523 fiz0 = _fjsp_madd_v2r8(dz00,fscal,fiz0);
525 fjx0 = _fjsp_madd_v2r8(dx00,fscal,fjx0);
526 fjy0 = _fjsp_madd_v2r8(dy00,fscal,fjy0);
527 fjz0 = _fjsp_madd_v2r8(dz00,fscal,fjz0);
531 /**************************
532 * CALCULATE INTERACTIONS *
533 **************************/
535 if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
538 r10 = _fjsp_mul_v2r8(rsq10,rinv10);
540 /* Compute parameters for interactions between i and j atoms */
541 qq10 = _fjsp_mul_v2r8(iq1,jq0);
543 /* EWALD ELECTROSTATICS */
545 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
546 ewrt = _fjsp_mul_v2r8(r10,ewtabscale);
547 itab_tmp = _fjsp_dtox_v2r8(ewrt);
548 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
549 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
551 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
552 ewtabD = _fjsp_setzero_v2r8();
553 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
554 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
555 ewtabFn = _fjsp_setzero_v2r8();
556 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
557 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
558 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
559 velec = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
560 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
562 d = _fjsp_sub_v2r8(r10,rswitch);
563 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
564 d2 = _fjsp_mul_v2r8(d,d);
565 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
567 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
569 /* Evaluate switch function */
570 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
571 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv10,_fjsp_mul_v2r8(velec,dsw)) );
572 velec = _fjsp_mul_v2r8(velec,sw);
573 cutoff_mask = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
575 /* Update potential sum for this i atom from the interaction with this j atom. */
576 velec = _fjsp_and_v2r8(velec,cutoff_mask);
577 velec = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
578 velecsum = _fjsp_add_v2r8(velecsum,velec);
582 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
584 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
586 /* Update vectorial force */
587 fix1 = _fjsp_madd_v2r8(dx10,fscal,fix1);
588 fiy1 = _fjsp_madd_v2r8(dy10,fscal,fiy1);
589 fiz1 = _fjsp_madd_v2r8(dz10,fscal,fiz1);
591 fjx0 = _fjsp_madd_v2r8(dx10,fscal,fjx0);
592 fjy0 = _fjsp_madd_v2r8(dy10,fscal,fjy0);
593 fjz0 = _fjsp_madd_v2r8(dz10,fscal,fjz0);
597 /**************************
598 * CALCULATE INTERACTIONS *
599 **************************/
601 if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
604 r20 = _fjsp_mul_v2r8(rsq20,rinv20);
606 /* Compute parameters for interactions between i and j atoms */
607 qq20 = _fjsp_mul_v2r8(iq2,jq0);
609 /* EWALD ELECTROSTATICS */
611 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
612 ewrt = _fjsp_mul_v2r8(r20,ewtabscale);
613 itab_tmp = _fjsp_dtox_v2r8(ewrt);
614 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
615 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
617 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
618 ewtabD = _fjsp_setzero_v2r8();
619 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
620 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
621 ewtabFn = _fjsp_setzero_v2r8();
622 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
623 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
624 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
625 velec = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
626 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
628 d = _fjsp_sub_v2r8(r20,rswitch);
629 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
630 d2 = _fjsp_mul_v2r8(d,d);
631 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
633 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
635 /* Evaluate switch function */
636 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
637 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv20,_fjsp_mul_v2r8(velec,dsw)) );
638 velec = _fjsp_mul_v2r8(velec,sw);
639 cutoff_mask = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
641 /* Update potential sum for this i atom from the interaction with this j atom. */
642 velec = _fjsp_and_v2r8(velec,cutoff_mask);
643 velec = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
644 velecsum = _fjsp_add_v2r8(velecsum,velec);
648 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
650 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
652 /* Update vectorial force */
653 fix2 = _fjsp_madd_v2r8(dx20,fscal,fix2);
654 fiy2 = _fjsp_madd_v2r8(dy20,fscal,fiy2);
655 fiz2 = _fjsp_madd_v2r8(dz20,fscal,fiz2);
657 fjx0 = _fjsp_madd_v2r8(dx20,fscal,fjx0);
658 fjy0 = _fjsp_madd_v2r8(dy20,fscal,fjy0);
659 fjz0 = _fjsp_madd_v2r8(dz20,fscal,fjz0);
663 gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
665 /* Inner loop uses 207 flops */
668 /* End of innermost loop */
670 gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
671 f+i_coord_offset,fshift+i_shift_offset);
674 /* Update potential energies */
675 gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
677 /* Increment number of inner iterations */
678 inneriter += j_index_end - j_index_start;
680 /* Outer loop uses 19 flops */
683 /* Increment number of outer iterations */
686 /* Update outer/inner flops */
688 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3_VF,outeriter*19 + inneriter*207);
691 * Gromacs nonbonded kernel: nb_kernel_ElecEwSw_VdwNone_GeomW3P1_F_sparc64_hpc_ace_double
692 * Electrostatics interaction: Ewald
693 * VdW interaction: None
694 * Geometry: Water3-Particle
695 * Calculate force/pot: Force
698 nb_kernel_ElecEwSw_VdwNone_GeomW3P1_F_sparc64_hpc_ace_double
699 (t_nblist * gmx_restrict nlist,
700 rvec * gmx_restrict xx,
701 rvec * gmx_restrict ff,
702 t_forcerec * gmx_restrict fr,
703 t_mdatoms * gmx_restrict mdatoms,
704 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
705 t_nrnb * gmx_restrict nrnb)
707 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
708 * just 0 for non-waters.
709 * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
710 * jnr indices corresponding to data put in the four positions in the SIMD register.
712 int i_shift_offset,i_coord_offset,outeriter,inneriter;
713 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
715 int j_coord_offsetA,j_coord_offsetB;
716 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
718 real *shiftvec,*fshift,*x,*f;
719 _fjsp_v2r8 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
721 _fjsp_v2r8 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
723 _fjsp_v2r8 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
725 _fjsp_v2r8 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
726 int vdwjidx0A,vdwjidx0B;
727 _fjsp_v2r8 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
728 _fjsp_v2r8 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
729 _fjsp_v2r8 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
730 _fjsp_v2r8 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
731 _fjsp_v2r8 velec,felec,velecsum,facel,crf,krf,krf2;
733 _fjsp_v2r8 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
735 _fjsp_v2r8 rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
736 real rswitch_scalar,d_scalar;
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;
757 sh_ewald = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
758 ewtab = fr->ic->tabq_coul_FDV0;
759 ewtabscale = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
760 ewtabhalfspace = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
762 /* Setup water-specific parameters */
763 inr = nlist->iinr[0];
764 iq0 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
765 iq1 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
766 iq2 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
768 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
769 rcutoff_scalar = fr->rcoulomb;
770 rcutoff = gmx_fjsp_set1_v2r8(rcutoff_scalar);
771 rcutoff2 = _fjsp_mul_v2r8(rcutoff,rcutoff);
773 rswitch_scalar = fr->rcoulomb_switch;
774 rswitch = gmx_fjsp_set1_v2r8(rswitch_scalar);
775 /* Setup switch parameters */
776 d_scalar = rcutoff_scalar-rswitch_scalar;
777 d = gmx_fjsp_set1_v2r8(d_scalar);
778 swV3 = gmx_fjsp_set1_v2r8(-10.0/(d_scalar*d_scalar*d_scalar));
779 swV4 = gmx_fjsp_set1_v2r8( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
780 swV5 = gmx_fjsp_set1_v2r8( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
781 swF2 = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar));
782 swF3 = gmx_fjsp_set1_v2r8( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
783 swF4 = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
785 /* Avoid stupid compiler warnings */
793 /* Start outer loop over neighborlists */
794 for(iidx=0; iidx<nri; iidx++)
796 /* Load shift vector for this list */
797 i_shift_offset = DIM*shiftidx[iidx];
799 /* Load limits for loop over neighbors */
800 j_index_start = jindex[iidx];
801 j_index_end = jindex[iidx+1];
803 /* Get outer coordinate index */
805 i_coord_offset = DIM*inr;
807 /* Load i particle coords and add shift vector */
808 gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
809 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
811 fix0 = _fjsp_setzero_v2r8();
812 fiy0 = _fjsp_setzero_v2r8();
813 fiz0 = _fjsp_setzero_v2r8();
814 fix1 = _fjsp_setzero_v2r8();
815 fiy1 = _fjsp_setzero_v2r8();
816 fiz1 = _fjsp_setzero_v2r8();
817 fix2 = _fjsp_setzero_v2r8();
818 fiy2 = _fjsp_setzero_v2r8();
819 fiz2 = _fjsp_setzero_v2r8();
821 /* Start inner kernel loop */
822 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
825 /* Get j neighbor index, and coordinate index */
828 j_coord_offsetA = DIM*jnrA;
829 j_coord_offsetB = DIM*jnrB;
831 /* load j atom coordinates */
832 gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
835 /* Calculate displacement vector */
836 dx00 = _fjsp_sub_v2r8(ix0,jx0);
837 dy00 = _fjsp_sub_v2r8(iy0,jy0);
838 dz00 = _fjsp_sub_v2r8(iz0,jz0);
839 dx10 = _fjsp_sub_v2r8(ix1,jx0);
840 dy10 = _fjsp_sub_v2r8(iy1,jy0);
841 dz10 = _fjsp_sub_v2r8(iz1,jz0);
842 dx20 = _fjsp_sub_v2r8(ix2,jx0);
843 dy20 = _fjsp_sub_v2r8(iy2,jy0);
844 dz20 = _fjsp_sub_v2r8(iz2,jz0);
846 /* Calculate squared distance and things based on it */
847 rsq00 = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
848 rsq10 = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
849 rsq20 = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
851 rinv00 = gmx_fjsp_invsqrt_v2r8(rsq00);
852 rinv10 = gmx_fjsp_invsqrt_v2r8(rsq10);
853 rinv20 = gmx_fjsp_invsqrt_v2r8(rsq20);
855 rinvsq00 = _fjsp_mul_v2r8(rinv00,rinv00);
856 rinvsq10 = _fjsp_mul_v2r8(rinv10,rinv10);
857 rinvsq20 = _fjsp_mul_v2r8(rinv20,rinv20);
859 /* Load parameters for j particles */
860 jq0 = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
862 fjx0 = _fjsp_setzero_v2r8();
863 fjy0 = _fjsp_setzero_v2r8();
864 fjz0 = _fjsp_setzero_v2r8();
866 /**************************
867 * CALCULATE INTERACTIONS *
868 **************************/
870 if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
873 r00 = _fjsp_mul_v2r8(rsq00,rinv00);
875 /* Compute parameters for interactions between i and j atoms */
876 qq00 = _fjsp_mul_v2r8(iq0,jq0);
878 /* EWALD ELECTROSTATICS */
880 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
881 ewrt = _fjsp_mul_v2r8(r00,ewtabscale);
882 itab_tmp = _fjsp_dtox_v2r8(ewrt);
883 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
884 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
886 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
887 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
888 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
889 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
890 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
891 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
892 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
893 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
894 velec = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(rinv00,velec));
895 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
897 d = _fjsp_sub_v2r8(r00,rswitch);
898 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
899 d2 = _fjsp_mul_v2r8(d,d);
900 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
902 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
904 /* Evaluate switch function */
905 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
906 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(velec,dsw)) );
907 cutoff_mask = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
911 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
913 /* Update vectorial force */
914 fix0 = _fjsp_madd_v2r8(dx00,fscal,fix0);
915 fiy0 = _fjsp_madd_v2r8(dy00,fscal,fiy0);
916 fiz0 = _fjsp_madd_v2r8(dz00,fscal,fiz0);
918 fjx0 = _fjsp_madd_v2r8(dx00,fscal,fjx0);
919 fjy0 = _fjsp_madd_v2r8(dy00,fscal,fjy0);
920 fjz0 = _fjsp_madd_v2r8(dz00,fscal,fjz0);
924 /**************************
925 * CALCULATE INTERACTIONS *
926 **************************/
928 if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
931 r10 = _fjsp_mul_v2r8(rsq10,rinv10);
933 /* Compute parameters for interactions between i and j atoms */
934 qq10 = _fjsp_mul_v2r8(iq1,jq0);
936 /* EWALD ELECTROSTATICS */
938 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
939 ewrt = _fjsp_mul_v2r8(r10,ewtabscale);
940 itab_tmp = _fjsp_dtox_v2r8(ewrt);
941 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
942 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
944 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
945 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
946 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
947 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
948 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
949 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
950 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
951 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
952 velec = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
953 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
955 d = _fjsp_sub_v2r8(r10,rswitch);
956 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
957 d2 = _fjsp_mul_v2r8(d,d);
958 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
960 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
962 /* Evaluate switch function */
963 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
964 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv10,_fjsp_mul_v2r8(velec,dsw)) );
965 cutoff_mask = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
969 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
971 /* Update vectorial force */
972 fix1 = _fjsp_madd_v2r8(dx10,fscal,fix1);
973 fiy1 = _fjsp_madd_v2r8(dy10,fscal,fiy1);
974 fiz1 = _fjsp_madd_v2r8(dz10,fscal,fiz1);
976 fjx0 = _fjsp_madd_v2r8(dx10,fscal,fjx0);
977 fjy0 = _fjsp_madd_v2r8(dy10,fscal,fjy0);
978 fjz0 = _fjsp_madd_v2r8(dz10,fscal,fjz0);
982 /**************************
983 * CALCULATE INTERACTIONS *
984 **************************/
986 if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
989 r20 = _fjsp_mul_v2r8(rsq20,rinv20);
991 /* Compute parameters for interactions between i and j atoms */
992 qq20 = _fjsp_mul_v2r8(iq2,jq0);
994 /* EWALD ELECTROSTATICS */
996 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
997 ewrt = _fjsp_mul_v2r8(r20,ewtabscale);
998 itab_tmp = _fjsp_dtox_v2r8(ewrt);
999 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1000 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1002 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1003 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
1004 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1005 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1006 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
1007 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1008 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1009 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1010 velec = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
1011 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
1013 d = _fjsp_sub_v2r8(r20,rswitch);
1014 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1015 d2 = _fjsp_mul_v2r8(d,d);
1016 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1018 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1020 /* Evaluate switch function */
1021 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1022 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv20,_fjsp_mul_v2r8(velec,dsw)) );
1023 cutoff_mask = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
1027 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
1029 /* Update vectorial force */
1030 fix2 = _fjsp_madd_v2r8(dx20,fscal,fix2);
1031 fiy2 = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1032 fiz2 = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1034 fjx0 = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1035 fjy0 = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1036 fjz0 = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1040 gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
1042 /* Inner loop uses 198 flops */
1045 if(jidx<j_index_end)
1049 j_coord_offsetA = DIM*jnrA;
1051 /* load j atom coordinates */
1052 gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
1055 /* Calculate displacement vector */
1056 dx00 = _fjsp_sub_v2r8(ix0,jx0);
1057 dy00 = _fjsp_sub_v2r8(iy0,jy0);
1058 dz00 = _fjsp_sub_v2r8(iz0,jz0);
1059 dx10 = _fjsp_sub_v2r8(ix1,jx0);
1060 dy10 = _fjsp_sub_v2r8(iy1,jy0);
1061 dz10 = _fjsp_sub_v2r8(iz1,jz0);
1062 dx20 = _fjsp_sub_v2r8(ix2,jx0);
1063 dy20 = _fjsp_sub_v2r8(iy2,jy0);
1064 dz20 = _fjsp_sub_v2r8(iz2,jz0);
1066 /* Calculate squared distance and things based on it */
1067 rsq00 = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
1068 rsq10 = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
1069 rsq20 = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
1071 rinv00 = gmx_fjsp_invsqrt_v2r8(rsq00);
1072 rinv10 = gmx_fjsp_invsqrt_v2r8(rsq10);
1073 rinv20 = gmx_fjsp_invsqrt_v2r8(rsq20);
1075 rinvsq00 = _fjsp_mul_v2r8(rinv00,rinv00);
1076 rinvsq10 = _fjsp_mul_v2r8(rinv10,rinv10);
1077 rinvsq20 = _fjsp_mul_v2r8(rinv20,rinv20);
1079 /* Load parameters for j particles */
1080 jq0 = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
1082 fjx0 = _fjsp_setzero_v2r8();
1083 fjy0 = _fjsp_setzero_v2r8();
1084 fjz0 = _fjsp_setzero_v2r8();
1086 /**************************
1087 * CALCULATE INTERACTIONS *
1088 **************************/
1090 if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
1093 r00 = _fjsp_mul_v2r8(rsq00,rinv00);
1095 /* Compute parameters for interactions between i and j atoms */
1096 qq00 = _fjsp_mul_v2r8(iq0,jq0);
1098 /* EWALD ELECTROSTATICS */
1100 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1101 ewrt = _fjsp_mul_v2r8(r00,ewtabscale);
1102 itab_tmp = _fjsp_dtox_v2r8(ewrt);
1103 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1104 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1106 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1107 ewtabD = _fjsp_setzero_v2r8();
1108 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1109 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1110 ewtabFn = _fjsp_setzero_v2r8();
1111 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1112 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1113 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1114 velec = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(rinv00,velec));
1115 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
1117 d = _fjsp_sub_v2r8(r00,rswitch);
1118 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1119 d2 = _fjsp_mul_v2r8(d,d);
1120 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1122 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1124 /* Evaluate switch function */
1125 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1126 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(velec,dsw)) );
1127 cutoff_mask = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
1131 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
1133 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1135 /* Update vectorial force */
1136 fix0 = _fjsp_madd_v2r8(dx00,fscal,fix0);
1137 fiy0 = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1138 fiz0 = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1140 fjx0 = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1141 fjy0 = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1142 fjz0 = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1146 /**************************
1147 * CALCULATE INTERACTIONS *
1148 **************************/
1150 if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
1153 r10 = _fjsp_mul_v2r8(rsq10,rinv10);
1155 /* Compute parameters for interactions between i and j atoms */
1156 qq10 = _fjsp_mul_v2r8(iq1,jq0);
1158 /* EWALD ELECTROSTATICS */
1160 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1161 ewrt = _fjsp_mul_v2r8(r10,ewtabscale);
1162 itab_tmp = _fjsp_dtox_v2r8(ewrt);
1163 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1164 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1166 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1167 ewtabD = _fjsp_setzero_v2r8();
1168 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1169 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1170 ewtabFn = _fjsp_setzero_v2r8();
1171 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1172 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1173 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1174 velec = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
1175 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
1177 d = _fjsp_sub_v2r8(r10,rswitch);
1178 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1179 d2 = _fjsp_mul_v2r8(d,d);
1180 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1182 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1184 /* Evaluate switch function */
1185 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1186 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv10,_fjsp_mul_v2r8(velec,dsw)) );
1187 cutoff_mask = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
1191 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
1193 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1195 /* Update vectorial force */
1196 fix1 = _fjsp_madd_v2r8(dx10,fscal,fix1);
1197 fiy1 = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1198 fiz1 = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1200 fjx0 = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1201 fjy0 = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1202 fjz0 = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1206 /**************************
1207 * CALCULATE INTERACTIONS *
1208 **************************/
1210 if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
1213 r20 = _fjsp_mul_v2r8(rsq20,rinv20);
1215 /* Compute parameters for interactions between i and j atoms */
1216 qq20 = _fjsp_mul_v2r8(iq2,jq0);
1218 /* EWALD ELECTROSTATICS */
1220 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1221 ewrt = _fjsp_mul_v2r8(r20,ewtabscale);
1222 itab_tmp = _fjsp_dtox_v2r8(ewrt);
1223 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1224 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1226 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1227 ewtabD = _fjsp_setzero_v2r8();
1228 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1229 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1230 ewtabFn = _fjsp_setzero_v2r8();
1231 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1232 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1233 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1234 velec = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
1235 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
1237 d = _fjsp_sub_v2r8(r20,rswitch);
1238 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1239 d2 = _fjsp_mul_v2r8(d,d);
1240 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1242 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1244 /* Evaluate switch function */
1245 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1246 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv20,_fjsp_mul_v2r8(velec,dsw)) );
1247 cutoff_mask = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
1251 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
1253 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1255 /* Update vectorial force */
1256 fix2 = _fjsp_madd_v2r8(dx20,fscal,fix2);
1257 fiy2 = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1258 fiz2 = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1260 fjx0 = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1261 fjy0 = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1262 fjz0 = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1266 gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1268 /* Inner loop uses 198 flops */
1271 /* End of innermost loop */
1273 gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1274 f+i_coord_offset,fshift+i_shift_offset);
1276 /* Increment number of inner iterations */
1277 inneriter += j_index_end - j_index_start;
1279 /* Outer loop uses 18 flops */
1282 /* Increment number of outer iterations */
1285 /* Update outer/inner flops */
1287 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3_F,outeriter*18 + inneriter*198);