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.
42 #include "../nb_kernel.h"
43 #include "types/simple.h"
44 #include "gromacs/math/vec.h"
47 #include "kernelutil_sparc64_hpc_ace_double.h"
50 * Gromacs nonbonded kernel: nb_kernel_ElecEwSw_VdwNone_GeomW4P1_VF_sparc64_hpc_ace_double
51 * Electrostatics interaction: Ewald
52 * VdW interaction: None
53 * Geometry: Water4-Particle
54 * Calculate force/pot: PotentialAndForce
57 nb_kernel_ElecEwSw_VdwNone_GeomW4P1_VF_sparc64_hpc_ace_double
58 (t_nblist * gmx_restrict nlist,
59 rvec * gmx_restrict xx,
60 rvec * gmx_restrict ff,
61 t_forcerec * gmx_restrict fr,
62 t_mdatoms * gmx_restrict mdatoms,
63 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
64 t_nrnb * gmx_restrict nrnb)
66 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
67 * just 0 for non-waters.
68 * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
69 * jnr indices corresponding to data put in the four positions in the SIMD register.
71 int i_shift_offset,i_coord_offset,outeriter,inneriter;
72 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
74 int j_coord_offsetA,j_coord_offsetB;
75 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
77 real *shiftvec,*fshift,*x,*f;
78 _fjsp_v2r8 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
80 _fjsp_v2r8 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
82 _fjsp_v2r8 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
84 _fjsp_v2r8 ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
85 int vdwjidx0A,vdwjidx0B;
86 _fjsp_v2r8 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
87 _fjsp_v2r8 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
88 _fjsp_v2r8 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
89 _fjsp_v2r8 dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
90 _fjsp_v2r8 velec,felec,velecsum,facel,crf,krf,krf2;
92 _fjsp_v2r8 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
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;
116 sh_ewald = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
117 ewtab = fr->ic->tabq_coul_FDV0;
118 ewtabscale = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
119 ewtabhalfspace = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
121 /* Setup water-specific parameters */
122 inr = nlist->iinr[0];
123 iq1 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
124 iq2 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
125 iq3 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
127 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
128 rcutoff_scalar = fr->rcoulomb;
129 rcutoff = gmx_fjsp_set1_v2r8(rcutoff_scalar);
130 rcutoff2 = _fjsp_mul_v2r8(rcutoff,rcutoff);
132 rswitch_scalar = fr->rcoulomb_switch;
133 rswitch = gmx_fjsp_set1_v2r8(rswitch_scalar);
134 /* Setup switch parameters */
135 d_scalar = rcutoff_scalar-rswitch_scalar;
136 d = gmx_fjsp_set1_v2r8(d_scalar);
137 swV3 = gmx_fjsp_set1_v2r8(-10.0/(d_scalar*d_scalar*d_scalar));
138 swV4 = gmx_fjsp_set1_v2r8( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
139 swV5 = gmx_fjsp_set1_v2r8( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
140 swF2 = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar));
141 swF3 = gmx_fjsp_set1_v2r8( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
142 swF4 = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
144 /* Avoid stupid compiler warnings */
152 /* Start outer loop over neighborlists */
153 for(iidx=0; iidx<nri; iidx++)
155 /* Load shift vector for this list */
156 i_shift_offset = DIM*shiftidx[iidx];
158 /* Load limits for loop over neighbors */
159 j_index_start = jindex[iidx];
160 j_index_end = jindex[iidx+1];
162 /* Get outer coordinate index */
164 i_coord_offset = DIM*inr;
166 /* Load i particle coords and add shift vector */
167 gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
168 &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
170 fix1 = _fjsp_setzero_v2r8();
171 fiy1 = _fjsp_setzero_v2r8();
172 fiz1 = _fjsp_setzero_v2r8();
173 fix2 = _fjsp_setzero_v2r8();
174 fiy2 = _fjsp_setzero_v2r8();
175 fiz2 = _fjsp_setzero_v2r8();
176 fix3 = _fjsp_setzero_v2r8();
177 fiy3 = _fjsp_setzero_v2r8();
178 fiz3 = _fjsp_setzero_v2r8();
180 /* Reset potential sums */
181 velecsum = _fjsp_setzero_v2r8();
183 /* Start inner kernel loop */
184 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
187 /* Get j neighbor index, and coordinate index */
190 j_coord_offsetA = DIM*jnrA;
191 j_coord_offsetB = DIM*jnrB;
193 /* load j atom coordinates */
194 gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
197 /* Calculate displacement vector */
198 dx10 = _fjsp_sub_v2r8(ix1,jx0);
199 dy10 = _fjsp_sub_v2r8(iy1,jy0);
200 dz10 = _fjsp_sub_v2r8(iz1,jz0);
201 dx20 = _fjsp_sub_v2r8(ix2,jx0);
202 dy20 = _fjsp_sub_v2r8(iy2,jy0);
203 dz20 = _fjsp_sub_v2r8(iz2,jz0);
204 dx30 = _fjsp_sub_v2r8(ix3,jx0);
205 dy30 = _fjsp_sub_v2r8(iy3,jy0);
206 dz30 = _fjsp_sub_v2r8(iz3,jz0);
208 /* Calculate squared distance and things based on it */
209 rsq10 = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
210 rsq20 = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
211 rsq30 = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
213 rinv10 = gmx_fjsp_invsqrt_v2r8(rsq10);
214 rinv20 = gmx_fjsp_invsqrt_v2r8(rsq20);
215 rinv30 = gmx_fjsp_invsqrt_v2r8(rsq30);
217 rinvsq10 = _fjsp_mul_v2r8(rinv10,rinv10);
218 rinvsq20 = _fjsp_mul_v2r8(rinv20,rinv20);
219 rinvsq30 = _fjsp_mul_v2r8(rinv30,rinv30);
221 /* Load parameters for j particles */
222 jq0 = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
224 fjx0 = _fjsp_setzero_v2r8();
225 fjy0 = _fjsp_setzero_v2r8();
226 fjz0 = _fjsp_setzero_v2r8();
228 /**************************
229 * CALCULATE INTERACTIONS *
230 **************************/
232 if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
235 r10 = _fjsp_mul_v2r8(rsq10,rinv10);
237 /* Compute parameters for interactions between i and j atoms */
238 qq10 = _fjsp_mul_v2r8(iq1,jq0);
240 /* EWALD ELECTROSTATICS */
242 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
243 ewrt = _fjsp_mul_v2r8(r10,ewtabscale);
244 itab_tmp = _fjsp_dtox_v2r8(ewrt);
245 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
246 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
248 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
249 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
250 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
251 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
252 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
253 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
254 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
255 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
256 velec = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
257 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
259 d = _fjsp_sub_v2r8(r10,rswitch);
260 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
261 d2 = _fjsp_mul_v2r8(d,d);
262 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
264 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
266 /* Evaluate switch function */
267 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
268 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv10,_fjsp_mul_v2r8(velec,dsw)) );
269 velec = _fjsp_mul_v2r8(velec,sw);
270 cutoff_mask = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
272 /* Update potential sum for this i atom from the interaction with this j atom. */
273 velec = _fjsp_and_v2r8(velec,cutoff_mask);
274 velecsum = _fjsp_add_v2r8(velecsum,velec);
278 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
280 /* Update vectorial force */
281 fix1 = _fjsp_madd_v2r8(dx10,fscal,fix1);
282 fiy1 = _fjsp_madd_v2r8(dy10,fscal,fiy1);
283 fiz1 = _fjsp_madd_v2r8(dz10,fscal,fiz1);
285 fjx0 = _fjsp_madd_v2r8(dx10,fscal,fjx0);
286 fjy0 = _fjsp_madd_v2r8(dy10,fscal,fjy0);
287 fjz0 = _fjsp_madd_v2r8(dz10,fscal,fjz0);
291 /**************************
292 * CALCULATE INTERACTIONS *
293 **************************/
295 if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
298 r20 = _fjsp_mul_v2r8(rsq20,rinv20);
300 /* Compute parameters for interactions between i and j atoms */
301 qq20 = _fjsp_mul_v2r8(iq2,jq0);
303 /* EWALD ELECTROSTATICS */
305 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
306 ewrt = _fjsp_mul_v2r8(r20,ewtabscale);
307 itab_tmp = _fjsp_dtox_v2r8(ewrt);
308 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
309 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
311 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
312 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
313 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
314 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
315 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
316 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
317 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
318 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
319 velec = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
320 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
322 d = _fjsp_sub_v2r8(r20,rswitch);
323 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
324 d2 = _fjsp_mul_v2r8(d,d);
325 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
327 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
329 /* Evaluate switch function */
330 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
331 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv20,_fjsp_mul_v2r8(velec,dsw)) );
332 velec = _fjsp_mul_v2r8(velec,sw);
333 cutoff_mask = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
335 /* Update potential sum for this i atom from the interaction with this j atom. */
336 velec = _fjsp_and_v2r8(velec,cutoff_mask);
337 velecsum = _fjsp_add_v2r8(velecsum,velec);
341 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
343 /* Update vectorial force */
344 fix2 = _fjsp_madd_v2r8(dx20,fscal,fix2);
345 fiy2 = _fjsp_madd_v2r8(dy20,fscal,fiy2);
346 fiz2 = _fjsp_madd_v2r8(dz20,fscal,fiz2);
348 fjx0 = _fjsp_madd_v2r8(dx20,fscal,fjx0);
349 fjy0 = _fjsp_madd_v2r8(dy20,fscal,fjy0);
350 fjz0 = _fjsp_madd_v2r8(dz20,fscal,fjz0);
354 /**************************
355 * CALCULATE INTERACTIONS *
356 **************************/
358 if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
361 r30 = _fjsp_mul_v2r8(rsq30,rinv30);
363 /* Compute parameters for interactions between i and j atoms */
364 qq30 = _fjsp_mul_v2r8(iq3,jq0);
366 /* EWALD ELECTROSTATICS */
368 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
369 ewrt = _fjsp_mul_v2r8(r30,ewtabscale);
370 itab_tmp = _fjsp_dtox_v2r8(ewrt);
371 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
372 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
374 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
375 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
376 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
377 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
378 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
379 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
380 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
381 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
382 velec = _fjsp_mul_v2r8(qq30,_fjsp_sub_v2r8(rinv30,velec));
383 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,rinv30),_fjsp_sub_v2r8(rinvsq30,felec));
385 d = _fjsp_sub_v2r8(r30,rswitch);
386 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
387 d2 = _fjsp_mul_v2r8(d,d);
388 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
390 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
392 /* Evaluate switch function */
393 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
394 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv30,_fjsp_mul_v2r8(velec,dsw)) );
395 velec = _fjsp_mul_v2r8(velec,sw);
396 cutoff_mask = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
398 /* Update potential sum for this i atom from the interaction with this j atom. */
399 velec = _fjsp_and_v2r8(velec,cutoff_mask);
400 velecsum = _fjsp_add_v2r8(velecsum,velec);
404 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
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);
417 gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
419 /* Inner loop uses 207 flops */
426 j_coord_offsetA = DIM*jnrA;
428 /* load j atom coordinates */
429 gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
432 /* Calculate displacement vector */
433 dx10 = _fjsp_sub_v2r8(ix1,jx0);
434 dy10 = _fjsp_sub_v2r8(iy1,jy0);
435 dz10 = _fjsp_sub_v2r8(iz1,jz0);
436 dx20 = _fjsp_sub_v2r8(ix2,jx0);
437 dy20 = _fjsp_sub_v2r8(iy2,jy0);
438 dz20 = _fjsp_sub_v2r8(iz2,jz0);
439 dx30 = _fjsp_sub_v2r8(ix3,jx0);
440 dy30 = _fjsp_sub_v2r8(iy3,jy0);
441 dz30 = _fjsp_sub_v2r8(iz3,jz0);
443 /* Calculate squared distance and things based on it */
444 rsq10 = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
445 rsq20 = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
446 rsq30 = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
448 rinv10 = gmx_fjsp_invsqrt_v2r8(rsq10);
449 rinv20 = gmx_fjsp_invsqrt_v2r8(rsq20);
450 rinv30 = gmx_fjsp_invsqrt_v2r8(rsq30);
452 rinvsq10 = _fjsp_mul_v2r8(rinv10,rinv10);
453 rinvsq20 = _fjsp_mul_v2r8(rinv20,rinv20);
454 rinvsq30 = _fjsp_mul_v2r8(rinv30,rinv30);
456 /* Load parameters for j particles */
457 jq0 = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
459 fjx0 = _fjsp_setzero_v2r8();
460 fjy0 = _fjsp_setzero_v2r8();
461 fjz0 = _fjsp_setzero_v2r8();
463 /**************************
464 * CALCULATE INTERACTIONS *
465 **************************/
467 if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
470 r10 = _fjsp_mul_v2r8(rsq10,rinv10);
472 /* Compute parameters for interactions between i and j atoms */
473 qq10 = _fjsp_mul_v2r8(iq1,jq0);
475 /* EWALD ELECTROSTATICS */
477 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
478 ewrt = _fjsp_mul_v2r8(r10,ewtabscale);
479 itab_tmp = _fjsp_dtox_v2r8(ewrt);
480 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
481 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
483 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
484 ewtabD = _fjsp_setzero_v2r8();
485 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
486 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
487 ewtabFn = _fjsp_setzero_v2r8();
488 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
489 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
490 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
491 velec = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
492 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
494 d = _fjsp_sub_v2r8(r10,rswitch);
495 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
496 d2 = _fjsp_mul_v2r8(d,d);
497 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
499 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
501 /* Evaluate switch function */
502 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
503 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv10,_fjsp_mul_v2r8(velec,dsw)) );
504 velec = _fjsp_mul_v2r8(velec,sw);
505 cutoff_mask = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
507 /* Update potential sum for this i atom from the interaction with this j atom. */
508 velec = _fjsp_and_v2r8(velec,cutoff_mask);
509 velec = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
510 velecsum = _fjsp_add_v2r8(velecsum,velec);
514 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
516 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
518 /* Update vectorial force */
519 fix1 = _fjsp_madd_v2r8(dx10,fscal,fix1);
520 fiy1 = _fjsp_madd_v2r8(dy10,fscal,fiy1);
521 fiz1 = _fjsp_madd_v2r8(dz10,fscal,fiz1);
523 fjx0 = _fjsp_madd_v2r8(dx10,fscal,fjx0);
524 fjy0 = _fjsp_madd_v2r8(dy10,fscal,fjy0);
525 fjz0 = _fjsp_madd_v2r8(dz10,fscal,fjz0);
529 /**************************
530 * CALCULATE INTERACTIONS *
531 **************************/
533 if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
536 r20 = _fjsp_mul_v2r8(rsq20,rinv20);
538 /* Compute parameters for interactions between i and j atoms */
539 qq20 = _fjsp_mul_v2r8(iq2,jq0);
541 /* EWALD ELECTROSTATICS */
543 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
544 ewrt = _fjsp_mul_v2r8(r20,ewtabscale);
545 itab_tmp = _fjsp_dtox_v2r8(ewrt);
546 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
547 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
549 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
550 ewtabD = _fjsp_setzero_v2r8();
551 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
552 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
553 ewtabFn = _fjsp_setzero_v2r8();
554 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
555 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
556 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
557 velec = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
558 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
560 d = _fjsp_sub_v2r8(r20,rswitch);
561 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
562 d2 = _fjsp_mul_v2r8(d,d);
563 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
565 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
567 /* Evaluate switch function */
568 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
569 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv20,_fjsp_mul_v2r8(velec,dsw)) );
570 velec = _fjsp_mul_v2r8(velec,sw);
571 cutoff_mask = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
573 /* Update potential sum for this i atom from the interaction with this j atom. */
574 velec = _fjsp_and_v2r8(velec,cutoff_mask);
575 velec = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
576 velecsum = _fjsp_add_v2r8(velecsum,velec);
580 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
582 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
584 /* Update vectorial force */
585 fix2 = _fjsp_madd_v2r8(dx20,fscal,fix2);
586 fiy2 = _fjsp_madd_v2r8(dy20,fscal,fiy2);
587 fiz2 = _fjsp_madd_v2r8(dz20,fscal,fiz2);
589 fjx0 = _fjsp_madd_v2r8(dx20,fscal,fjx0);
590 fjy0 = _fjsp_madd_v2r8(dy20,fscal,fjy0);
591 fjz0 = _fjsp_madd_v2r8(dz20,fscal,fjz0);
595 /**************************
596 * CALCULATE INTERACTIONS *
597 **************************/
599 if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
602 r30 = _fjsp_mul_v2r8(rsq30,rinv30);
604 /* Compute parameters for interactions between i and j atoms */
605 qq30 = _fjsp_mul_v2r8(iq3,jq0);
607 /* EWALD ELECTROSTATICS */
609 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
610 ewrt = _fjsp_mul_v2r8(r30,ewtabscale);
611 itab_tmp = _fjsp_dtox_v2r8(ewrt);
612 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
613 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
615 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
616 ewtabD = _fjsp_setzero_v2r8();
617 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
618 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
619 ewtabFn = _fjsp_setzero_v2r8();
620 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
621 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
622 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
623 velec = _fjsp_mul_v2r8(qq30,_fjsp_sub_v2r8(rinv30,velec));
624 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,rinv30),_fjsp_sub_v2r8(rinvsq30,felec));
626 d = _fjsp_sub_v2r8(r30,rswitch);
627 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
628 d2 = _fjsp_mul_v2r8(d,d);
629 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
631 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
633 /* Evaluate switch function */
634 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
635 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv30,_fjsp_mul_v2r8(velec,dsw)) );
636 velec = _fjsp_mul_v2r8(velec,sw);
637 cutoff_mask = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
639 /* Update potential sum for this i atom from the interaction with this j atom. */
640 velec = _fjsp_and_v2r8(velec,cutoff_mask);
641 velec = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
642 velecsum = _fjsp_add_v2r8(velecsum,velec);
646 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
648 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
650 /* Update vectorial force */
651 fix3 = _fjsp_madd_v2r8(dx30,fscal,fix3);
652 fiy3 = _fjsp_madd_v2r8(dy30,fscal,fiy3);
653 fiz3 = _fjsp_madd_v2r8(dz30,fscal,fiz3);
655 fjx0 = _fjsp_madd_v2r8(dx30,fscal,fjx0);
656 fjy0 = _fjsp_madd_v2r8(dy30,fscal,fjy0);
657 fjz0 = _fjsp_madd_v2r8(dz30,fscal,fjz0);
661 gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
663 /* Inner loop uses 207 flops */
666 /* End of innermost loop */
668 gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
669 f+i_coord_offset+DIM,fshift+i_shift_offset);
672 /* Update potential energies */
673 gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
675 /* Increment number of inner iterations */
676 inneriter += j_index_end - j_index_start;
678 /* Outer loop uses 19 flops */
681 /* Increment number of outer iterations */
684 /* Update outer/inner flops */
686 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4_VF,outeriter*19 + inneriter*207);
689 * Gromacs nonbonded kernel: nb_kernel_ElecEwSw_VdwNone_GeomW4P1_F_sparc64_hpc_ace_double
690 * Electrostatics interaction: Ewald
691 * VdW interaction: None
692 * Geometry: Water4-Particle
693 * Calculate force/pot: Force
696 nb_kernel_ElecEwSw_VdwNone_GeomW4P1_F_sparc64_hpc_ace_double
697 (t_nblist * gmx_restrict nlist,
698 rvec * gmx_restrict xx,
699 rvec * gmx_restrict ff,
700 t_forcerec * gmx_restrict fr,
701 t_mdatoms * gmx_restrict mdatoms,
702 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
703 t_nrnb * gmx_restrict nrnb)
705 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
706 * just 0 for non-waters.
707 * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
708 * jnr indices corresponding to data put in the four positions in the SIMD register.
710 int i_shift_offset,i_coord_offset,outeriter,inneriter;
711 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
713 int j_coord_offsetA,j_coord_offsetB;
714 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
716 real *shiftvec,*fshift,*x,*f;
717 _fjsp_v2r8 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
719 _fjsp_v2r8 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
721 _fjsp_v2r8 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
723 _fjsp_v2r8 ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
724 int vdwjidx0A,vdwjidx0B;
725 _fjsp_v2r8 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
726 _fjsp_v2r8 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
727 _fjsp_v2r8 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
728 _fjsp_v2r8 dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
729 _fjsp_v2r8 velec,felec,velecsum,facel,crf,krf,krf2;
731 _fjsp_v2r8 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
733 _fjsp_v2r8 rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
734 real rswitch_scalar,d_scalar;
736 _fjsp_v2r8 dummy_mask,cutoff_mask;
737 _fjsp_v2r8 one = gmx_fjsp_set1_v2r8(1.0);
738 _fjsp_v2r8 two = gmx_fjsp_set1_v2r8(2.0);
739 union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
746 jindex = nlist->jindex;
748 shiftidx = nlist->shift;
750 shiftvec = fr->shift_vec[0];
751 fshift = fr->fshift[0];
752 facel = gmx_fjsp_set1_v2r8(fr->epsfac);
753 charge = mdatoms->chargeA;
755 sh_ewald = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
756 ewtab = fr->ic->tabq_coul_FDV0;
757 ewtabscale = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
758 ewtabhalfspace = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
760 /* Setup water-specific parameters */
761 inr = nlist->iinr[0];
762 iq1 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
763 iq2 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
764 iq3 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
766 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
767 rcutoff_scalar = fr->rcoulomb;
768 rcutoff = gmx_fjsp_set1_v2r8(rcutoff_scalar);
769 rcutoff2 = _fjsp_mul_v2r8(rcutoff,rcutoff);
771 rswitch_scalar = fr->rcoulomb_switch;
772 rswitch = gmx_fjsp_set1_v2r8(rswitch_scalar);
773 /* Setup switch parameters */
774 d_scalar = rcutoff_scalar-rswitch_scalar;
775 d = gmx_fjsp_set1_v2r8(d_scalar);
776 swV3 = gmx_fjsp_set1_v2r8(-10.0/(d_scalar*d_scalar*d_scalar));
777 swV4 = gmx_fjsp_set1_v2r8( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
778 swV5 = gmx_fjsp_set1_v2r8( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
779 swF2 = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar));
780 swF3 = gmx_fjsp_set1_v2r8( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
781 swF4 = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
783 /* Avoid stupid compiler warnings */
791 /* Start outer loop over neighborlists */
792 for(iidx=0; iidx<nri; iidx++)
794 /* Load shift vector for this list */
795 i_shift_offset = DIM*shiftidx[iidx];
797 /* Load limits for loop over neighbors */
798 j_index_start = jindex[iidx];
799 j_index_end = jindex[iidx+1];
801 /* Get outer coordinate index */
803 i_coord_offset = DIM*inr;
805 /* Load i particle coords and add shift vector */
806 gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
807 &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
809 fix1 = _fjsp_setzero_v2r8();
810 fiy1 = _fjsp_setzero_v2r8();
811 fiz1 = _fjsp_setzero_v2r8();
812 fix2 = _fjsp_setzero_v2r8();
813 fiy2 = _fjsp_setzero_v2r8();
814 fiz2 = _fjsp_setzero_v2r8();
815 fix3 = _fjsp_setzero_v2r8();
816 fiy3 = _fjsp_setzero_v2r8();
817 fiz3 = _fjsp_setzero_v2r8();
819 /* Start inner kernel loop */
820 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
823 /* Get j neighbor index, and coordinate index */
826 j_coord_offsetA = DIM*jnrA;
827 j_coord_offsetB = DIM*jnrB;
829 /* load j atom coordinates */
830 gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
833 /* Calculate displacement vector */
834 dx10 = _fjsp_sub_v2r8(ix1,jx0);
835 dy10 = _fjsp_sub_v2r8(iy1,jy0);
836 dz10 = _fjsp_sub_v2r8(iz1,jz0);
837 dx20 = _fjsp_sub_v2r8(ix2,jx0);
838 dy20 = _fjsp_sub_v2r8(iy2,jy0);
839 dz20 = _fjsp_sub_v2r8(iz2,jz0);
840 dx30 = _fjsp_sub_v2r8(ix3,jx0);
841 dy30 = _fjsp_sub_v2r8(iy3,jy0);
842 dz30 = _fjsp_sub_v2r8(iz3,jz0);
844 /* Calculate squared distance and things based on it */
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 rinv10 = gmx_fjsp_invsqrt_v2r8(rsq10);
850 rinv20 = gmx_fjsp_invsqrt_v2r8(rsq20);
851 rinv30 = gmx_fjsp_invsqrt_v2r8(rsq30);
853 rinvsq10 = _fjsp_mul_v2r8(rinv10,rinv10);
854 rinvsq20 = _fjsp_mul_v2r8(rinv20,rinv20);
855 rinvsq30 = _fjsp_mul_v2r8(rinv30,rinv30);
857 /* Load parameters for j particles */
858 jq0 = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
860 fjx0 = _fjsp_setzero_v2r8();
861 fjy0 = _fjsp_setzero_v2r8();
862 fjz0 = _fjsp_setzero_v2r8();
864 /**************************
865 * CALCULATE INTERACTIONS *
866 **************************/
868 if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
871 r10 = _fjsp_mul_v2r8(rsq10,rinv10);
873 /* Compute parameters for interactions between i and j atoms */
874 qq10 = _fjsp_mul_v2r8(iq1,jq0);
876 /* EWALD ELECTROSTATICS */
878 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
879 ewrt = _fjsp_mul_v2r8(r10,ewtabscale);
880 itab_tmp = _fjsp_dtox_v2r8(ewrt);
881 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
882 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
884 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
885 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
886 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
887 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
888 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
889 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
890 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
891 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
892 velec = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
893 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
895 d = _fjsp_sub_v2r8(r10,rswitch);
896 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
897 d2 = _fjsp_mul_v2r8(d,d);
898 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
900 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
902 /* Evaluate switch function */
903 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
904 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv10,_fjsp_mul_v2r8(velec,dsw)) );
905 cutoff_mask = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
909 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
911 /* Update vectorial force */
912 fix1 = _fjsp_madd_v2r8(dx10,fscal,fix1);
913 fiy1 = _fjsp_madd_v2r8(dy10,fscal,fiy1);
914 fiz1 = _fjsp_madd_v2r8(dz10,fscal,fiz1);
916 fjx0 = _fjsp_madd_v2r8(dx10,fscal,fjx0);
917 fjy0 = _fjsp_madd_v2r8(dy10,fscal,fjy0);
918 fjz0 = _fjsp_madd_v2r8(dz10,fscal,fjz0);
922 /**************************
923 * CALCULATE INTERACTIONS *
924 **************************/
926 if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
929 r20 = _fjsp_mul_v2r8(rsq20,rinv20);
931 /* Compute parameters for interactions between i and j atoms */
932 qq20 = _fjsp_mul_v2r8(iq2,jq0);
934 /* EWALD ELECTROSTATICS */
936 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
937 ewrt = _fjsp_mul_v2r8(r20,ewtabscale);
938 itab_tmp = _fjsp_dtox_v2r8(ewrt);
939 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
940 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
942 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
943 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
944 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
945 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
946 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
947 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
948 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
949 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
950 velec = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
951 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
953 d = _fjsp_sub_v2r8(r20,rswitch);
954 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
955 d2 = _fjsp_mul_v2r8(d,d);
956 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
958 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
960 /* Evaluate switch function */
961 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
962 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv20,_fjsp_mul_v2r8(velec,dsw)) );
963 cutoff_mask = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
967 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
969 /* Update vectorial force */
970 fix2 = _fjsp_madd_v2r8(dx20,fscal,fix2);
971 fiy2 = _fjsp_madd_v2r8(dy20,fscal,fiy2);
972 fiz2 = _fjsp_madd_v2r8(dz20,fscal,fiz2);
974 fjx0 = _fjsp_madd_v2r8(dx20,fscal,fjx0);
975 fjy0 = _fjsp_madd_v2r8(dy20,fscal,fjy0);
976 fjz0 = _fjsp_madd_v2r8(dz20,fscal,fjz0);
980 /**************************
981 * CALCULATE INTERACTIONS *
982 **************************/
984 if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
987 r30 = _fjsp_mul_v2r8(rsq30,rinv30);
989 /* Compute parameters for interactions between i and j atoms */
990 qq30 = _fjsp_mul_v2r8(iq3,jq0);
992 /* EWALD ELECTROSTATICS */
994 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
995 ewrt = _fjsp_mul_v2r8(r30,ewtabscale);
996 itab_tmp = _fjsp_dtox_v2r8(ewrt);
997 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
998 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1000 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1001 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
1002 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1003 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1004 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
1005 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1006 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1007 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1008 velec = _fjsp_mul_v2r8(qq30,_fjsp_sub_v2r8(rinv30,velec));
1009 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,rinv30),_fjsp_sub_v2r8(rinvsq30,felec));
1011 d = _fjsp_sub_v2r8(r30,rswitch);
1012 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1013 d2 = _fjsp_mul_v2r8(d,d);
1014 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1016 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1018 /* Evaluate switch function */
1019 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1020 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv30,_fjsp_mul_v2r8(velec,dsw)) );
1021 cutoff_mask = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
1025 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
1027 /* Update vectorial force */
1028 fix3 = _fjsp_madd_v2r8(dx30,fscal,fix3);
1029 fiy3 = _fjsp_madd_v2r8(dy30,fscal,fiy3);
1030 fiz3 = _fjsp_madd_v2r8(dz30,fscal,fiz3);
1032 fjx0 = _fjsp_madd_v2r8(dx30,fscal,fjx0);
1033 fjy0 = _fjsp_madd_v2r8(dy30,fscal,fjy0);
1034 fjz0 = _fjsp_madd_v2r8(dz30,fscal,fjz0);
1038 gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
1040 /* Inner loop uses 198 flops */
1043 if(jidx<j_index_end)
1047 j_coord_offsetA = DIM*jnrA;
1049 /* load j atom coordinates */
1050 gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
1053 /* Calculate displacement vector */
1054 dx10 = _fjsp_sub_v2r8(ix1,jx0);
1055 dy10 = _fjsp_sub_v2r8(iy1,jy0);
1056 dz10 = _fjsp_sub_v2r8(iz1,jz0);
1057 dx20 = _fjsp_sub_v2r8(ix2,jx0);
1058 dy20 = _fjsp_sub_v2r8(iy2,jy0);
1059 dz20 = _fjsp_sub_v2r8(iz2,jz0);
1060 dx30 = _fjsp_sub_v2r8(ix3,jx0);
1061 dy30 = _fjsp_sub_v2r8(iy3,jy0);
1062 dz30 = _fjsp_sub_v2r8(iz3,jz0);
1064 /* Calculate squared distance and things based on it */
1065 rsq10 = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
1066 rsq20 = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
1067 rsq30 = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
1069 rinv10 = gmx_fjsp_invsqrt_v2r8(rsq10);
1070 rinv20 = gmx_fjsp_invsqrt_v2r8(rsq20);
1071 rinv30 = gmx_fjsp_invsqrt_v2r8(rsq30);
1073 rinvsq10 = _fjsp_mul_v2r8(rinv10,rinv10);
1074 rinvsq20 = _fjsp_mul_v2r8(rinv20,rinv20);
1075 rinvsq30 = _fjsp_mul_v2r8(rinv30,rinv30);
1077 /* Load parameters for j particles */
1078 jq0 = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
1080 fjx0 = _fjsp_setzero_v2r8();
1081 fjy0 = _fjsp_setzero_v2r8();
1082 fjz0 = _fjsp_setzero_v2r8();
1084 /**************************
1085 * CALCULATE INTERACTIONS *
1086 **************************/
1088 if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
1091 r10 = _fjsp_mul_v2r8(rsq10,rinv10);
1093 /* Compute parameters for interactions between i and j atoms */
1094 qq10 = _fjsp_mul_v2r8(iq1,jq0);
1096 /* EWALD ELECTROSTATICS */
1098 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1099 ewrt = _fjsp_mul_v2r8(r10,ewtabscale);
1100 itab_tmp = _fjsp_dtox_v2r8(ewrt);
1101 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1102 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1104 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1105 ewtabD = _fjsp_setzero_v2r8();
1106 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1107 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1108 ewtabFn = _fjsp_setzero_v2r8();
1109 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1110 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1111 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1112 velec = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
1113 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
1115 d = _fjsp_sub_v2r8(r10,rswitch);
1116 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1117 d2 = _fjsp_mul_v2r8(d,d);
1118 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1120 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1122 /* Evaluate switch function */
1123 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1124 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv10,_fjsp_mul_v2r8(velec,dsw)) );
1125 cutoff_mask = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
1129 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
1131 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1133 /* Update vectorial force */
1134 fix1 = _fjsp_madd_v2r8(dx10,fscal,fix1);
1135 fiy1 = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1136 fiz1 = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1138 fjx0 = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1139 fjy0 = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1140 fjz0 = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1144 /**************************
1145 * CALCULATE INTERACTIONS *
1146 **************************/
1148 if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
1151 r20 = _fjsp_mul_v2r8(rsq20,rinv20);
1153 /* Compute parameters for interactions between i and j atoms */
1154 qq20 = _fjsp_mul_v2r8(iq2,jq0);
1156 /* EWALD ELECTROSTATICS */
1158 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1159 ewrt = _fjsp_mul_v2r8(r20,ewtabscale);
1160 itab_tmp = _fjsp_dtox_v2r8(ewrt);
1161 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1162 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1164 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1165 ewtabD = _fjsp_setzero_v2r8();
1166 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1167 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1168 ewtabFn = _fjsp_setzero_v2r8();
1169 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1170 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1171 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1172 velec = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
1173 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
1175 d = _fjsp_sub_v2r8(r20,rswitch);
1176 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1177 d2 = _fjsp_mul_v2r8(d,d);
1178 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1180 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1182 /* Evaluate switch function */
1183 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1184 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv20,_fjsp_mul_v2r8(velec,dsw)) );
1185 cutoff_mask = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
1189 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
1191 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1193 /* Update vectorial force */
1194 fix2 = _fjsp_madd_v2r8(dx20,fscal,fix2);
1195 fiy2 = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1196 fiz2 = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1198 fjx0 = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1199 fjy0 = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1200 fjz0 = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1204 /**************************
1205 * CALCULATE INTERACTIONS *
1206 **************************/
1208 if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
1211 r30 = _fjsp_mul_v2r8(rsq30,rinv30);
1213 /* Compute parameters for interactions between i and j atoms */
1214 qq30 = _fjsp_mul_v2r8(iq3,jq0);
1216 /* EWALD ELECTROSTATICS */
1218 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1219 ewrt = _fjsp_mul_v2r8(r30,ewtabscale);
1220 itab_tmp = _fjsp_dtox_v2r8(ewrt);
1221 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1222 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1224 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1225 ewtabD = _fjsp_setzero_v2r8();
1226 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1227 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1228 ewtabFn = _fjsp_setzero_v2r8();
1229 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1230 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1231 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1232 velec = _fjsp_mul_v2r8(qq30,_fjsp_sub_v2r8(rinv30,velec));
1233 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,rinv30),_fjsp_sub_v2r8(rinvsq30,felec));
1235 d = _fjsp_sub_v2r8(r30,rswitch);
1236 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1237 d2 = _fjsp_mul_v2r8(d,d);
1238 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1240 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1242 /* Evaluate switch function */
1243 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1244 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv30,_fjsp_mul_v2r8(velec,dsw)) );
1245 cutoff_mask = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
1249 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
1251 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1253 /* Update vectorial force */
1254 fix3 = _fjsp_madd_v2r8(dx30,fscal,fix3);
1255 fiy3 = _fjsp_madd_v2r8(dy30,fscal,fiy3);
1256 fiz3 = _fjsp_madd_v2r8(dz30,fscal,fiz3);
1258 fjx0 = _fjsp_madd_v2r8(dx30,fscal,fjx0);
1259 fjy0 = _fjsp_madd_v2r8(dy30,fscal,fjy0);
1260 fjz0 = _fjsp_madd_v2r8(dz30,fscal,fjz0);
1264 gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1266 /* Inner loop uses 198 flops */
1269 /* End of innermost loop */
1271 gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1272 f+i_coord_offset+DIM,fshift+i_shift_offset);
1274 /* Increment number of inner iterations */
1275 inneriter += j_index_end - j_index_start;
1277 /* Outer loop uses 18 flops */
1280 /* Increment number of outer iterations */
1283 /* Update outer/inner flops */
1285 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4_F,outeriter*18 + inneriter*198);