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_ElecEwSw_VdwLJSw_GeomW4W4_VF_sparc64_hpc_ace_double
53 * Electrostatics interaction: Ewald
54 * VdW interaction: LennardJones
55 * Geometry: Water4-Water4
56 * Calculate force/pot: PotentialAndForce
59 nb_kernel_ElecEwSw_VdwLJSw_GeomW4W4_VF_sparc64_hpc_ace_double
60 (t_nblist * gmx_restrict nlist,
61 rvec * gmx_restrict xx,
62 rvec * gmx_restrict ff,
63 t_forcerec * gmx_restrict fr,
64 t_mdatoms * gmx_restrict mdatoms,
65 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
66 t_nrnb * gmx_restrict nrnb)
68 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
69 * just 0 for non-waters.
70 * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
71 * jnr indices corresponding to data put in the four positions in the SIMD register.
73 int i_shift_offset,i_coord_offset,outeriter,inneriter;
74 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
76 int j_coord_offsetA,j_coord_offsetB;
77 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
79 real *shiftvec,*fshift,*x,*f;
80 _fjsp_v2r8 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
82 _fjsp_v2r8 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
84 _fjsp_v2r8 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
86 _fjsp_v2r8 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
88 _fjsp_v2r8 ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
89 int vdwjidx0A,vdwjidx0B;
90 _fjsp_v2r8 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
91 int vdwjidx1A,vdwjidx1B;
92 _fjsp_v2r8 jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
93 int vdwjidx2A,vdwjidx2B;
94 _fjsp_v2r8 jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
95 int vdwjidx3A,vdwjidx3B;
96 _fjsp_v2r8 jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
97 _fjsp_v2r8 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
98 _fjsp_v2r8 dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
99 _fjsp_v2r8 dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
100 _fjsp_v2r8 dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
101 _fjsp_v2r8 dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
102 _fjsp_v2r8 dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
103 _fjsp_v2r8 dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
104 _fjsp_v2r8 dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
105 _fjsp_v2r8 dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
106 _fjsp_v2r8 dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
107 _fjsp_v2r8 velec,felec,velecsum,facel,crf,krf,krf2;
110 _fjsp_v2r8 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
113 _fjsp_v2r8 one_sixth = gmx_fjsp_set1_v2r8(1.0/6.0);
114 _fjsp_v2r8 one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
115 _fjsp_v2r8 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
117 _fjsp_v2r8 rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
118 real rswitch_scalar,d_scalar;
120 _fjsp_v2r8 dummy_mask,cutoff_mask;
121 _fjsp_v2r8 one = gmx_fjsp_set1_v2r8(1.0);
122 _fjsp_v2r8 two = gmx_fjsp_set1_v2r8(2.0);
123 union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
130 jindex = nlist->jindex;
132 shiftidx = nlist->shift;
134 shiftvec = fr->shift_vec[0];
135 fshift = fr->fshift[0];
136 facel = gmx_fjsp_set1_v2r8(fr->epsfac);
137 charge = mdatoms->chargeA;
138 nvdwtype = fr->ntype;
140 vdwtype = mdatoms->typeA;
142 sh_ewald = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
143 ewtab = fr->ic->tabq_coul_FDV0;
144 ewtabscale = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
145 ewtabhalfspace = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
147 /* Setup water-specific parameters */
148 inr = nlist->iinr[0];
149 iq1 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
150 iq2 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
151 iq3 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
152 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
154 jq1 = gmx_fjsp_set1_v2r8(charge[inr+1]);
155 jq2 = gmx_fjsp_set1_v2r8(charge[inr+2]);
156 jq3 = gmx_fjsp_set1_v2r8(charge[inr+3]);
157 vdwjidx0A = 2*vdwtype[inr+0];
158 c6_00 = gmx_fjsp_set1_v2r8(vdwparam[vdwioffset0+vdwjidx0A]);
159 c12_00 = gmx_fjsp_set1_v2r8(vdwparam[vdwioffset0+vdwjidx0A+1]);
160 qq11 = _fjsp_mul_v2r8(iq1,jq1);
161 qq12 = _fjsp_mul_v2r8(iq1,jq2);
162 qq13 = _fjsp_mul_v2r8(iq1,jq3);
163 qq21 = _fjsp_mul_v2r8(iq2,jq1);
164 qq22 = _fjsp_mul_v2r8(iq2,jq2);
165 qq23 = _fjsp_mul_v2r8(iq2,jq3);
166 qq31 = _fjsp_mul_v2r8(iq3,jq1);
167 qq32 = _fjsp_mul_v2r8(iq3,jq2);
168 qq33 = _fjsp_mul_v2r8(iq3,jq3);
170 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
171 rcutoff_scalar = fr->rcoulomb;
172 rcutoff = gmx_fjsp_set1_v2r8(rcutoff_scalar);
173 rcutoff2 = _fjsp_mul_v2r8(rcutoff,rcutoff);
175 rswitch_scalar = fr->rcoulomb_switch;
176 rswitch = gmx_fjsp_set1_v2r8(rswitch_scalar);
177 /* Setup switch parameters */
178 d_scalar = rcutoff_scalar-rswitch_scalar;
179 d = gmx_fjsp_set1_v2r8(d_scalar);
180 swV3 = gmx_fjsp_set1_v2r8(-10.0/(d_scalar*d_scalar*d_scalar));
181 swV4 = gmx_fjsp_set1_v2r8( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
182 swV5 = gmx_fjsp_set1_v2r8( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
183 swF2 = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar));
184 swF3 = gmx_fjsp_set1_v2r8( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
185 swF4 = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
187 /* Avoid stupid compiler warnings */
195 /* Start outer loop over neighborlists */
196 for(iidx=0; iidx<nri; iidx++)
198 /* Load shift vector for this list */
199 i_shift_offset = DIM*shiftidx[iidx];
201 /* Load limits for loop over neighbors */
202 j_index_start = jindex[iidx];
203 j_index_end = jindex[iidx+1];
205 /* Get outer coordinate index */
207 i_coord_offset = DIM*inr;
209 /* Load i particle coords and add shift vector */
210 gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
211 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
213 fix0 = _fjsp_setzero_v2r8();
214 fiy0 = _fjsp_setzero_v2r8();
215 fiz0 = _fjsp_setzero_v2r8();
216 fix1 = _fjsp_setzero_v2r8();
217 fiy1 = _fjsp_setzero_v2r8();
218 fiz1 = _fjsp_setzero_v2r8();
219 fix2 = _fjsp_setzero_v2r8();
220 fiy2 = _fjsp_setzero_v2r8();
221 fiz2 = _fjsp_setzero_v2r8();
222 fix3 = _fjsp_setzero_v2r8();
223 fiy3 = _fjsp_setzero_v2r8();
224 fiz3 = _fjsp_setzero_v2r8();
226 /* Reset potential sums */
227 velecsum = _fjsp_setzero_v2r8();
228 vvdwsum = _fjsp_setzero_v2r8();
230 /* Start inner kernel loop */
231 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
234 /* Get j neighbor index, and coordinate index */
237 j_coord_offsetA = DIM*jnrA;
238 j_coord_offsetB = DIM*jnrB;
240 /* load j atom coordinates */
241 gmx_fjsp_load_4rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
242 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
243 &jy2,&jz2,&jx3,&jy3,&jz3);
245 /* Calculate displacement vector */
246 dx00 = _fjsp_sub_v2r8(ix0,jx0);
247 dy00 = _fjsp_sub_v2r8(iy0,jy0);
248 dz00 = _fjsp_sub_v2r8(iz0,jz0);
249 dx11 = _fjsp_sub_v2r8(ix1,jx1);
250 dy11 = _fjsp_sub_v2r8(iy1,jy1);
251 dz11 = _fjsp_sub_v2r8(iz1,jz1);
252 dx12 = _fjsp_sub_v2r8(ix1,jx2);
253 dy12 = _fjsp_sub_v2r8(iy1,jy2);
254 dz12 = _fjsp_sub_v2r8(iz1,jz2);
255 dx13 = _fjsp_sub_v2r8(ix1,jx3);
256 dy13 = _fjsp_sub_v2r8(iy1,jy3);
257 dz13 = _fjsp_sub_v2r8(iz1,jz3);
258 dx21 = _fjsp_sub_v2r8(ix2,jx1);
259 dy21 = _fjsp_sub_v2r8(iy2,jy1);
260 dz21 = _fjsp_sub_v2r8(iz2,jz1);
261 dx22 = _fjsp_sub_v2r8(ix2,jx2);
262 dy22 = _fjsp_sub_v2r8(iy2,jy2);
263 dz22 = _fjsp_sub_v2r8(iz2,jz2);
264 dx23 = _fjsp_sub_v2r8(ix2,jx3);
265 dy23 = _fjsp_sub_v2r8(iy2,jy3);
266 dz23 = _fjsp_sub_v2r8(iz2,jz3);
267 dx31 = _fjsp_sub_v2r8(ix3,jx1);
268 dy31 = _fjsp_sub_v2r8(iy3,jy1);
269 dz31 = _fjsp_sub_v2r8(iz3,jz1);
270 dx32 = _fjsp_sub_v2r8(ix3,jx2);
271 dy32 = _fjsp_sub_v2r8(iy3,jy2);
272 dz32 = _fjsp_sub_v2r8(iz3,jz2);
273 dx33 = _fjsp_sub_v2r8(ix3,jx3);
274 dy33 = _fjsp_sub_v2r8(iy3,jy3);
275 dz33 = _fjsp_sub_v2r8(iz3,jz3);
277 /* Calculate squared distance and things based on it */
278 rsq00 = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
279 rsq11 = gmx_fjsp_calc_rsq_v2r8(dx11,dy11,dz11);
280 rsq12 = gmx_fjsp_calc_rsq_v2r8(dx12,dy12,dz12);
281 rsq13 = gmx_fjsp_calc_rsq_v2r8(dx13,dy13,dz13);
282 rsq21 = gmx_fjsp_calc_rsq_v2r8(dx21,dy21,dz21);
283 rsq22 = gmx_fjsp_calc_rsq_v2r8(dx22,dy22,dz22);
284 rsq23 = gmx_fjsp_calc_rsq_v2r8(dx23,dy23,dz23);
285 rsq31 = gmx_fjsp_calc_rsq_v2r8(dx31,dy31,dz31);
286 rsq32 = gmx_fjsp_calc_rsq_v2r8(dx32,dy32,dz32);
287 rsq33 = gmx_fjsp_calc_rsq_v2r8(dx33,dy33,dz33);
289 rinv00 = gmx_fjsp_invsqrt_v2r8(rsq00);
290 rinv11 = gmx_fjsp_invsqrt_v2r8(rsq11);
291 rinv12 = gmx_fjsp_invsqrt_v2r8(rsq12);
292 rinv13 = gmx_fjsp_invsqrt_v2r8(rsq13);
293 rinv21 = gmx_fjsp_invsqrt_v2r8(rsq21);
294 rinv22 = gmx_fjsp_invsqrt_v2r8(rsq22);
295 rinv23 = gmx_fjsp_invsqrt_v2r8(rsq23);
296 rinv31 = gmx_fjsp_invsqrt_v2r8(rsq31);
297 rinv32 = gmx_fjsp_invsqrt_v2r8(rsq32);
298 rinv33 = gmx_fjsp_invsqrt_v2r8(rsq33);
300 rinvsq00 = _fjsp_mul_v2r8(rinv00,rinv00);
301 rinvsq11 = _fjsp_mul_v2r8(rinv11,rinv11);
302 rinvsq12 = _fjsp_mul_v2r8(rinv12,rinv12);
303 rinvsq13 = _fjsp_mul_v2r8(rinv13,rinv13);
304 rinvsq21 = _fjsp_mul_v2r8(rinv21,rinv21);
305 rinvsq22 = _fjsp_mul_v2r8(rinv22,rinv22);
306 rinvsq23 = _fjsp_mul_v2r8(rinv23,rinv23);
307 rinvsq31 = _fjsp_mul_v2r8(rinv31,rinv31);
308 rinvsq32 = _fjsp_mul_v2r8(rinv32,rinv32);
309 rinvsq33 = _fjsp_mul_v2r8(rinv33,rinv33);
311 fjx0 = _fjsp_setzero_v2r8();
312 fjy0 = _fjsp_setzero_v2r8();
313 fjz0 = _fjsp_setzero_v2r8();
314 fjx1 = _fjsp_setzero_v2r8();
315 fjy1 = _fjsp_setzero_v2r8();
316 fjz1 = _fjsp_setzero_v2r8();
317 fjx2 = _fjsp_setzero_v2r8();
318 fjy2 = _fjsp_setzero_v2r8();
319 fjz2 = _fjsp_setzero_v2r8();
320 fjx3 = _fjsp_setzero_v2r8();
321 fjy3 = _fjsp_setzero_v2r8();
322 fjz3 = _fjsp_setzero_v2r8();
324 /**************************
325 * CALCULATE INTERACTIONS *
326 **************************/
328 if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
331 r00 = _fjsp_mul_v2r8(rsq00,rinv00);
333 /* LENNARD-JONES DISPERSION/REPULSION */
335 rinvsix = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
336 vvdw6 = _fjsp_mul_v2r8(c6_00,rinvsix);
337 vvdw12 = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
338 vvdw = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
339 fvdw = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
341 d = _fjsp_sub_v2r8(r00,rswitch);
342 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
343 d2 = _fjsp_mul_v2r8(d,d);
344 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
346 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
348 /* Evaluate switch function */
349 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
350 fvdw = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
351 vvdw = _fjsp_mul_v2r8(vvdw,sw);
352 cutoff_mask = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
354 /* Update potential sum for this i atom from the interaction with this j atom. */
355 vvdw = _fjsp_and_v2r8(vvdw,cutoff_mask);
356 vvdwsum = _fjsp_add_v2r8(vvdwsum,vvdw);
360 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
362 /* Update vectorial force */
363 fix0 = _fjsp_madd_v2r8(dx00,fscal,fix0);
364 fiy0 = _fjsp_madd_v2r8(dy00,fscal,fiy0);
365 fiz0 = _fjsp_madd_v2r8(dz00,fscal,fiz0);
367 fjx0 = _fjsp_madd_v2r8(dx00,fscal,fjx0);
368 fjy0 = _fjsp_madd_v2r8(dy00,fscal,fjy0);
369 fjz0 = _fjsp_madd_v2r8(dz00,fscal,fjz0);
373 /**************************
374 * CALCULATE INTERACTIONS *
375 **************************/
377 if (gmx_fjsp_any_lt_v2r8(rsq11,rcutoff2))
380 r11 = _fjsp_mul_v2r8(rsq11,rinv11);
382 /* EWALD ELECTROSTATICS */
384 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
385 ewrt = _fjsp_mul_v2r8(r11,ewtabscale);
386 itab_tmp = _fjsp_dtox_v2r8(ewrt);
387 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
388 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
390 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
391 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
392 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
393 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
394 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
395 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
396 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
397 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
398 velec = _fjsp_mul_v2r8(qq11,_fjsp_sub_v2r8(rinv11,velec));
399 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq11,rinv11),_fjsp_sub_v2r8(rinvsq11,felec));
401 d = _fjsp_sub_v2r8(r11,rswitch);
402 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
403 d2 = _fjsp_mul_v2r8(d,d);
404 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
406 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
408 /* Evaluate switch function */
409 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
410 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv11,_fjsp_mul_v2r8(velec,dsw)) );
411 velec = _fjsp_mul_v2r8(velec,sw);
412 cutoff_mask = _fjsp_cmplt_v2r8(rsq11,rcutoff2);
414 /* Update potential sum for this i atom from the interaction with this j atom. */
415 velec = _fjsp_and_v2r8(velec,cutoff_mask);
416 velecsum = _fjsp_add_v2r8(velecsum,velec);
420 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
422 /* Update vectorial force */
423 fix1 = _fjsp_madd_v2r8(dx11,fscal,fix1);
424 fiy1 = _fjsp_madd_v2r8(dy11,fscal,fiy1);
425 fiz1 = _fjsp_madd_v2r8(dz11,fscal,fiz1);
427 fjx1 = _fjsp_madd_v2r8(dx11,fscal,fjx1);
428 fjy1 = _fjsp_madd_v2r8(dy11,fscal,fjy1);
429 fjz1 = _fjsp_madd_v2r8(dz11,fscal,fjz1);
433 /**************************
434 * CALCULATE INTERACTIONS *
435 **************************/
437 if (gmx_fjsp_any_lt_v2r8(rsq12,rcutoff2))
440 r12 = _fjsp_mul_v2r8(rsq12,rinv12);
442 /* EWALD ELECTROSTATICS */
444 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
445 ewrt = _fjsp_mul_v2r8(r12,ewtabscale);
446 itab_tmp = _fjsp_dtox_v2r8(ewrt);
447 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
448 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
450 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
451 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
452 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
453 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
454 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
455 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
456 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
457 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
458 velec = _fjsp_mul_v2r8(qq12,_fjsp_sub_v2r8(rinv12,velec));
459 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq12,rinv12),_fjsp_sub_v2r8(rinvsq12,felec));
461 d = _fjsp_sub_v2r8(r12,rswitch);
462 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
463 d2 = _fjsp_mul_v2r8(d,d);
464 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
466 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
468 /* Evaluate switch function */
469 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
470 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv12,_fjsp_mul_v2r8(velec,dsw)) );
471 velec = _fjsp_mul_v2r8(velec,sw);
472 cutoff_mask = _fjsp_cmplt_v2r8(rsq12,rcutoff2);
474 /* Update potential sum for this i atom from the interaction with this j atom. */
475 velec = _fjsp_and_v2r8(velec,cutoff_mask);
476 velecsum = _fjsp_add_v2r8(velecsum,velec);
480 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
482 /* Update vectorial force */
483 fix1 = _fjsp_madd_v2r8(dx12,fscal,fix1);
484 fiy1 = _fjsp_madd_v2r8(dy12,fscal,fiy1);
485 fiz1 = _fjsp_madd_v2r8(dz12,fscal,fiz1);
487 fjx2 = _fjsp_madd_v2r8(dx12,fscal,fjx2);
488 fjy2 = _fjsp_madd_v2r8(dy12,fscal,fjy2);
489 fjz2 = _fjsp_madd_v2r8(dz12,fscal,fjz2);
493 /**************************
494 * CALCULATE INTERACTIONS *
495 **************************/
497 if (gmx_fjsp_any_lt_v2r8(rsq13,rcutoff2))
500 r13 = _fjsp_mul_v2r8(rsq13,rinv13);
502 /* EWALD ELECTROSTATICS */
504 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
505 ewrt = _fjsp_mul_v2r8(r13,ewtabscale);
506 itab_tmp = _fjsp_dtox_v2r8(ewrt);
507 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
508 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
510 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
511 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
512 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
513 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
514 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
515 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
516 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
517 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
518 velec = _fjsp_mul_v2r8(qq13,_fjsp_sub_v2r8(rinv13,velec));
519 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq13,rinv13),_fjsp_sub_v2r8(rinvsq13,felec));
521 d = _fjsp_sub_v2r8(r13,rswitch);
522 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
523 d2 = _fjsp_mul_v2r8(d,d);
524 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
526 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
528 /* Evaluate switch function */
529 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
530 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv13,_fjsp_mul_v2r8(velec,dsw)) );
531 velec = _fjsp_mul_v2r8(velec,sw);
532 cutoff_mask = _fjsp_cmplt_v2r8(rsq13,rcutoff2);
534 /* Update potential sum for this i atom from the interaction with this j atom. */
535 velec = _fjsp_and_v2r8(velec,cutoff_mask);
536 velecsum = _fjsp_add_v2r8(velecsum,velec);
540 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
542 /* Update vectorial force */
543 fix1 = _fjsp_madd_v2r8(dx13,fscal,fix1);
544 fiy1 = _fjsp_madd_v2r8(dy13,fscal,fiy1);
545 fiz1 = _fjsp_madd_v2r8(dz13,fscal,fiz1);
547 fjx3 = _fjsp_madd_v2r8(dx13,fscal,fjx3);
548 fjy3 = _fjsp_madd_v2r8(dy13,fscal,fjy3);
549 fjz3 = _fjsp_madd_v2r8(dz13,fscal,fjz3);
553 /**************************
554 * CALCULATE INTERACTIONS *
555 **************************/
557 if (gmx_fjsp_any_lt_v2r8(rsq21,rcutoff2))
560 r21 = _fjsp_mul_v2r8(rsq21,rinv21);
562 /* EWALD ELECTROSTATICS */
564 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
565 ewrt = _fjsp_mul_v2r8(r21,ewtabscale);
566 itab_tmp = _fjsp_dtox_v2r8(ewrt);
567 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
568 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
570 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
571 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
572 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
573 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
574 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
575 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
576 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
577 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
578 velec = _fjsp_mul_v2r8(qq21,_fjsp_sub_v2r8(rinv21,velec));
579 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq21,rinv21),_fjsp_sub_v2r8(rinvsq21,felec));
581 d = _fjsp_sub_v2r8(r21,rswitch);
582 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
583 d2 = _fjsp_mul_v2r8(d,d);
584 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
586 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
588 /* Evaluate switch function */
589 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
590 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv21,_fjsp_mul_v2r8(velec,dsw)) );
591 velec = _fjsp_mul_v2r8(velec,sw);
592 cutoff_mask = _fjsp_cmplt_v2r8(rsq21,rcutoff2);
594 /* Update potential sum for this i atom from the interaction with this j atom. */
595 velec = _fjsp_and_v2r8(velec,cutoff_mask);
596 velecsum = _fjsp_add_v2r8(velecsum,velec);
600 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
602 /* Update vectorial force */
603 fix2 = _fjsp_madd_v2r8(dx21,fscal,fix2);
604 fiy2 = _fjsp_madd_v2r8(dy21,fscal,fiy2);
605 fiz2 = _fjsp_madd_v2r8(dz21,fscal,fiz2);
607 fjx1 = _fjsp_madd_v2r8(dx21,fscal,fjx1);
608 fjy1 = _fjsp_madd_v2r8(dy21,fscal,fjy1);
609 fjz1 = _fjsp_madd_v2r8(dz21,fscal,fjz1);
613 /**************************
614 * CALCULATE INTERACTIONS *
615 **************************/
617 if (gmx_fjsp_any_lt_v2r8(rsq22,rcutoff2))
620 r22 = _fjsp_mul_v2r8(rsq22,rinv22);
622 /* EWALD ELECTROSTATICS */
624 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
625 ewrt = _fjsp_mul_v2r8(r22,ewtabscale);
626 itab_tmp = _fjsp_dtox_v2r8(ewrt);
627 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
628 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
630 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
631 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
632 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
633 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
634 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
635 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
636 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
637 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
638 velec = _fjsp_mul_v2r8(qq22,_fjsp_sub_v2r8(rinv22,velec));
639 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq22,rinv22),_fjsp_sub_v2r8(rinvsq22,felec));
641 d = _fjsp_sub_v2r8(r22,rswitch);
642 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
643 d2 = _fjsp_mul_v2r8(d,d);
644 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
646 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
648 /* Evaluate switch function */
649 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
650 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv22,_fjsp_mul_v2r8(velec,dsw)) );
651 velec = _fjsp_mul_v2r8(velec,sw);
652 cutoff_mask = _fjsp_cmplt_v2r8(rsq22,rcutoff2);
654 /* Update potential sum for this i atom from the interaction with this j atom. */
655 velec = _fjsp_and_v2r8(velec,cutoff_mask);
656 velecsum = _fjsp_add_v2r8(velecsum,velec);
660 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
662 /* Update vectorial force */
663 fix2 = _fjsp_madd_v2r8(dx22,fscal,fix2);
664 fiy2 = _fjsp_madd_v2r8(dy22,fscal,fiy2);
665 fiz2 = _fjsp_madd_v2r8(dz22,fscal,fiz2);
667 fjx2 = _fjsp_madd_v2r8(dx22,fscal,fjx2);
668 fjy2 = _fjsp_madd_v2r8(dy22,fscal,fjy2);
669 fjz2 = _fjsp_madd_v2r8(dz22,fscal,fjz2);
673 /**************************
674 * CALCULATE INTERACTIONS *
675 **************************/
677 if (gmx_fjsp_any_lt_v2r8(rsq23,rcutoff2))
680 r23 = _fjsp_mul_v2r8(rsq23,rinv23);
682 /* EWALD ELECTROSTATICS */
684 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
685 ewrt = _fjsp_mul_v2r8(r23,ewtabscale);
686 itab_tmp = _fjsp_dtox_v2r8(ewrt);
687 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
688 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
690 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
691 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
692 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
693 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
694 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
695 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
696 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
697 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
698 velec = _fjsp_mul_v2r8(qq23,_fjsp_sub_v2r8(rinv23,velec));
699 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq23,rinv23),_fjsp_sub_v2r8(rinvsq23,felec));
701 d = _fjsp_sub_v2r8(r23,rswitch);
702 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
703 d2 = _fjsp_mul_v2r8(d,d);
704 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
706 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
708 /* Evaluate switch function */
709 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
710 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv23,_fjsp_mul_v2r8(velec,dsw)) );
711 velec = _fjsp_mul_v2r8(velec,sw);
712 cutoff_mask = _fjsp_cmplt_v2r8(rsq23,rcutoff2);
714 /* Update potential sum for this i atom from the interaction with this j atom. */
715 velec = _fjsp_and_v2r8(velec,cutoff_mask);
716 velecsum = _fjsp_add_v2r8(velecsum,velec);
720 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
722 /* Update vectorial force */
723 fix2 = _fjsp_madd_v2r8(dx23,fscal,fix2);
724 fiy2 = _fjsp_madd_v2r8(dy23,fscal,fiy2);
725 fiz2 = _fjsp_madd_v2r8(dz23,fscal,fiz2);
727 fjx3 = _fjsp_madd_v2r8(dx23,fscal,fjx3);
728 fjy3 = _fjsp_madd_v2r8(dy23,fscal,fjy3);
729 fjz3 = _fjsp_madd_v2r8(dz23,fscal,fjz3);
733 /**************************
734 * CALCULATE INTERACTIONS *
735 **************************/
737 if (gmx_fjsp_any_lt_v2r8(rsq31,rcutoff2))
740 r31 = _fjsp_mul_v2r8(rsq31,rinv31);
742 /* EWALD ELECTROSTATICS */
744 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
745 ewrt = _fjsp_mul_v2r8(r31,ewtabscale);
746 itab_tmp = _fjsp_dtox_v2r8(ewrt);
747 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
748 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
750 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
751 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
752 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
753 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
754 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
755 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
756 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
757 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
758 velec = _fjsp_mul_v2r8(qq31,_fjsp_sub_v2r8(rinv31,velec));
759 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq31,rinv31),_fjsp_sub_v2r8(rinvsq31,felec));
761 d = _fjsp_sub_v2r8(r31,rswitch);
762 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
763 d2 = _fjsp_mul_v2r8(d,d);
764 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
766 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
768 /* Evaluate switch function */
769 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
770 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv31,_fjsp_mul_v2r8(velec,dsw)) );
771 velec = _fjsp_mul_v2r8(velec,sw);
772 cutoff_mask = _fjsp_cmplt_v2r8(rsq31,rcutoff2);
774 /* Update potential sum for this i atom from the interaction with this j atom. */
775 velec = _fjsp_and_v2r8(velec,cutoff_mask);
776 velecsum = _fjsp_add_v2r8(velecsum,velec);
780 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
782 /* Update vectorial force */
783 fix3 = _fjsp_madd_v2r8(dx31,fscal,fix3);
784 fiy3 = _fjsp_madd_v2r8(dy31,fscal,fiy3);
785 fiz3 = _fjsp_madd_v2r8(dz31,fscal,fiz3);
787 fjx1 = _fjsp_madd_v2r8(dx31,fscal,fjx1);
788 fjy1 = _fjsp_madd_v2r8(dy31,fscal,fjy1);
789 fjz1 = _fjsp_madd_v2r8(dz31,fscal,fjz1);
793 /**************************
794 * CALCULATE INTERACTIONS *
795 **************************/
797 if (gmx_fjsp_any_lt_v2r8(rsq32,rcutoff2))
800 r32 = _fjsp_mul_v2r8(rsq32,rinv32);
802 /* EWALD ELECTROSTATICS */
804 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
805 ewrt = _fjsp_mul_v2r8(r32,ewtabscale);
806 itab_tmp = _fjsp_dtox_v2r8(ewrt);
807 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
808 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
810 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
811 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
812 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
813 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
814 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
815 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
816 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
817 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
818 velec = _fjsp_mul_v2r8(qq32,_fjsp_sub_v2r8(rinv32,velec));
819 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq32,rinv32),_fjsp_sub_v2r8(rinvsq32,felec));
821 d = _fjsp_sub_v2r8(r32,rswitch);
822 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
823 d2 = _fjsp_mul_v2r8(d,d);
824 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
826 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
828 /* Evaluate switch function */
829 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
830 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv32,_fjsp_mul_v2r8(velec,dsw)) );
831 velec = _fjsp_mul_v2r8(velec,sw);
832 cutoff_mask = _fjsp_cmplt_v2r8(rsq32,rcutoff2);
834 /* Update potential sum for this i atom from the interaction with this j atom. */
835 velec = _fjsp_and_v2r8(velec,cutoff_mask);
836 velecsum = _fjsp_add_v2r8(velecsum,velec);
840 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
842 /* Update vectorial force */
843 fix3 = _fjsp_madd_v2r8(dx32,fscal,fix3);
844 fiy3 = _fjsp_madd_v2r8(dy32,fscal,fiy3);
845 fiz3 = _fjsp_madd_v2r8(dz32,fscal,fiz3);
847 fjx2 = _fjsp_madd_v2r8(dx32,fscal,fjx2);
848 fjy2 = _fjsp_madd_v2r8(dy32,fscal,fjy2);
849 fjz2 = _fjsp_madd_v2r8(dz32,fscal,fjz2);
853 /**************************
854 * CALCULATE INTERACTIONS *
855 **************************/
857 if (gmx_fjsp_any_lt_v2r8(rsq33,rcutoff2))
860 r33 = _fjsp_mul_v2r8(rsq33,rinv33);
862 /* EWALD ELECTROSTATICS */
864 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
865 ewrt = _fjsp_mul_v2r8(r33,ewtabscale);
866 itab_tmp = _fjsp_dtox_v2r8(ewrt);
867 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
868 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
870 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
871 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
872 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
873 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
874 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
875 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
876 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
877 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
878 velec = _fjsp_mul_v2r8(qq33,_fjsp_sub_v2r8(rinv33,velec));
879 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq33,rinv33),_fjsp_sub_v2r8(rinvsq33,felec));
881 d = _fjsp_sub_v2r8(r33,rswitch);
882 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
883 d2 = _fjsp_mul_v2r8(d,d);
884 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
886 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
888 /* Evaluate switch function */
889 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
890 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv33,_fjsp_mul_v2r8(velec,dsw)) );
891 velec = _fjsp_mul_v2r8(velec,sw);
892 cutoff_mask = _fjsp_cmplt_v2r8(rsq33,rcutoff2);
894 /* Update potential sum for this i atom from the interaction with this j atom. */
895 velec = _fjsp_and_v2r8(velec,cutoff_mask);
896 velecsum = _fjsp_add_v2r8(velecsum,velec);
900 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
902 /* Update vectorial force */
903 fix3 = _fjsp_madd_v2r8(dx33,fscal,fix3);
904 fiy3 = _fjsp_madd_v2r8(dy33,fscal,fiy3);
905 fiz3 = _fjsp_madd_v2r8(dz33,fscal,fiz3);
907 fjx3 = _fjsp_madd_v2r8(dx33,fscal,fjx3);
908 fjy3 = _fjsp_madd_v2r8(dy33,fscal,fjy3);
909 fjz3 = _fjsp_madd_v2r8(dz33,fscal,fjz3);
913 gmx_fjsp_decrement_4rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
915 /* Inner loop uses 677 flops */
922 j_coord_offsetA = DIM*jnrA;
924 /* load j atom coordinates */
925 gmx_fjsp_load_4rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
926 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
927 &jy2,&jz2,&jx3,&jy3,&jz3);
929 /* Calculate displacement vector */
930 dx00 = _fjsp_sub_v2r8(ix0,jx0);
931 dy00 = _fjsp_sub_v2r8(iy0,jy0);
932 dz00 = _fjsp_sub_v2r8(iz0,jz0);
933 dx11 = _fjsp_sub_v2r8(ix1,jx1);
934 dy11 = _fjsp_sub_v2r8(iy1,jy1);
935 dz11 = _fjsp_sub_v2r8(iz1,jz1);
936 dx12 = _fjsp_sub_v2r8(ix1,jx2);
937 dy12 = _fjsp_sub_v2r8(iy1,jy2);
938 dz12 = _fjsp_sub_v2r8(iz1,jz2);
939 dx13 = _fjsp_sub_v2r8(ix1,jx3);
940 dy13 = _fjsp_sub_v2r8(iy1,jy3);
941 dz13 = _fjsp_sub_v2r8(iz1,jz3);
942 dx21 = _fjsp_sub_v2r8(ix2,jx1);
943 dy21 = _fjsp_sub_v2r8(iy2,jy1);
944 dz21 = _fjsp_sub_v2r8(iz2,jz1);
945 dx22 = _fjsp_sub_v2r8(ix2,jx2);
946 dy22 = _fjsp_sub_v2r8(iy2,jy2);
947 dz22 = _fjsp_sub_v2r8(iz2,jz2);
948 dx23 = _fjsp_sub_v2r8(ix2,jx3);
949 dy23 = _fjsp_sub_v2r8(iy2,jy3);
950 dz23 = _fjsp_sub_v2r8(iz2,jz3);
951 dx31 = _fjsp_sub_v2r8(ix3,jx1);
952 dy31 = _fjsp_sub_v2r8(iy3,jy1);
953 dz31 = _fjsp_sub_v2r8(iz3,jz1);
954 dx32 = _fjsp_sub_v2r8(ix3,jx2);
955 dy32 = _fjsp_sub_v2r8(iy3,jy2);
956 dz32 = _fjsp_sub_v2r8(iz3,jz2);
957 dx33 = _fjsp_sub_v2r8(ix3,jx3);
958 dy33 = _fjsp_sub_v2r8(iy3,jy3);
959 dz33 = _fjsp_sub_v2r8(iz3,jz3);
961 /* Calculate squared distance and things based on it */
962 rsq00 = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
963 rsq11 = gmx_fjsp_calc_rsq_v2r8(dx11,dy11,dz11);
964 rsq12 = gmx_fjsp_calc_rsq_v2r8(dx12,dy12,dz12);
965 rsq13 = gmx_fjsp_calc_rsq_v2r8(dx13,dy13,dz13);
966 rsq21 = gmx_fjsp_calc_rsq_v2r8(dx21,dy21,dz21);
967 rsq22 = gmx_fjsp_calc_rsq_v2r8(dx22,dy22,dz22);
968 rsq23 = gmx_fjsp_calc_rsq_v2r8(dx23,dy23,dz23);
969 rsq31 = gmx_fjsp_calc_rsq_v2r8(dx31,dy31,dz31);
970 rsq32 = gmx_fjsp_calc_rsq_v2r8(dx32,dy32,dz32);
971 rsq33 = gmx_fjsp_calc_rsq_v2r8(dx33,dy33,dz33);
973 rinv00 = gmx_fjsp_invsqrt_v2r8(rsq00);
974 rinv11 = gmx_fjsp_invsqrt_v2r8(rsq11);
975 rinv12 = gmx_fjsp_invsqrt_v2r8(rsq12);
976 rinv13 = gmx_fjsp_invsqrt_v2r8(rsq13);
977 rinv21 = gmx_fjsp_invsqrt_v2r8(rsq21);
978 rinv22 = gmx_fjsp_invsqrt_v2r8(rsq22);
979 rinv23 = gmx_fjsp_invsqrt_v2r8(rsq23);
980 rinv31 = gmx_fjsp_invsqrt_v2r8(rsq31);
981 rinv32 = gmx_fjsp_invsqrt_v2r8(rsq32);
982 rinv33 = gmx_fjsp_invsqrt_v2r8(rsq33);
984 rinvsq00 = _fjsp_mul_v2r8(rinv00,rinv00);
985 rinvsq11 = _fjsp_mul_v2r8(rinv11,rinv11);
986 rinvsq12 = _fjsp_mul_v2r8(rinv12,rinv12);
987 rinvsq13 = _fjsp_mul_v2r8(rinv13,rinv13);
988 rinvsq21 = _fjsp_mul_v2r8(rinv21,rinv21);
989 rinvsq22 = _fjsp_mul_v2r8(rinv22,rinv22);
990 rinvsq23 = _fjsp_mul_v2r8(rinv23,rinv23);
991 rinvsq31 = _fjsp_mul_v2r8(rinv31,rinv31);
992 rinvsq32 = _fjsp_mul_v2r8(rinv32,rinv32);
993 rinvsq33 = _fjsp_mul_v2r8(rinv33,rinv33);
995 fjx0 = _fjsp_setzero_v2r8();
996 fjy0 = _fjsp_setzero_v2r8();
997 fjz0 = _fjsp_setzero_v2r8();
998 fjx1 = _fjsp_setzero_v2r8();
999 fjy1 = _fjsp_setzero_v2r8();
1000 fjz1 = _fjsp_setzero_v2r8();
1001 fjx2 = _fjsp_setzero_v2r8();
1002 fjy2 = _fjsp_setzero_v2r8();
1003 fjz2 = _fjsp_setzero_v2r8();
1004 fjx3 = _fjsp_setzero_v2r8();
1005 fjy3 = _fjsp_setzero_v2r8();
1006 fjz3 = _fjsp_setzero_v2r8();
1008 /**************************
1009 * CALCULATE INTERACTIONS *
1010 **************************/
1012 if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
1015 r00 = _fjsp_mul_v2r8(rsq00,rinv00);
1017 /* LENNARD-JONES DISPERSION/REPULSION */
1019 rinvsix = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
1020 vvdw6 = _fjsp_mul_v2r8(c6_00,rinvsix);
1021 vvdw12 = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
1022 vvdw = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
1023 fvdw = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
1025 d = _fjsp_sub_v2r8(r00,rswitch);
1026 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1027 d2 = _fjsp_mul_v2r8(d,d);
1028 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1030 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1032 /* Evaluate switch function */
1033 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1034 fvdw = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
1035 vvdw = _fjsp_mul_v2r8(vvdw,sw);
1036 cutoff_mask = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
1038 /* Update potential sum for this i atom from the interaction with this j atom. */
1039 vvdw = _fjsp_and_v2r8(vvdw,cutoff_mask);
1040 vvdw = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
1041 vvdwsum = _fjsp_add_v2r8(vvdwsum,vvdw);
1045 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
1047 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1049 /* Update vectorial force */
1050 fix0 = _fjsp_madd_v2r8(dx00,fscal,fix0);
1051 fiy0 = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1052 fiz0 = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1054 fjx0 = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1055 fjy0 = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1056 fjz0 = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1060 /**************************
1061 * CALCULATE INTERACTIONS *
1062 **************************/
1064 if (gmx_fjsp_any_lt_v2r8(rsq11,rcutoff2))
1067 r11 = _fjsp_mul_v2r8(rsq11,rinv11);
1069 /* EWALD ELECTROSTATICS */
1071 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1072 ewrt = _fjsp_mul_v2r8(r11,ewtabscale);
1073 itab_tmp = _fjsp_dtox_v2r8(ewrt);
1074 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1075 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1077 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1078 ewtabD = _fjsp_setzero_v2r8();
1079 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1080 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1081 ewtabFn = _fjsp_setzero_v2r8();
1082 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1083 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1084 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1085 velec = _fjsp_mul_v2r8(qq11,_fjsp_sub_v2r8(rinv11,velec));
1086 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq11,rinv11),_fjsp_sub_v2r8(rinvsq11,felec));
1088 d = _fjsp_sub_v2r8(r11,rswitch);
1089 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1090 d2 = _fjsp_mul_v2r8(d,d);
1091 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1093 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1095 /* Evaluate switch function */
1096 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1097 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv11,_fjsp_mul_v2r8(velec,dsw)) );
1098 velec = _fjsp_mul_v2r8(velec,sw);
1099 cutoff_mask = _fjsp_cmplt_v2r8(rsq11,rcutoff2);
1101 /* Update potential sum for this i atom from the interaction with this j atom. */
1102 velec = _fjsp_and_v2r8(velec,cutoff_mask);
1103 velec = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
1104 velecsum = _fjsp_add_v2r8(velecsum,velec);
1108 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
1110 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1112 /* Update vectorial force */
1113 fix1 = _fjsp_madd_v2r8(dx11,fscal,fix1);
1114 fiy1 = _fjsp_madd_v2r8(dy11,fscal,fiy1);
1115 fiz1 = _fjsp_madd_v2r8(dz11,fscal,fiz1);
1117 fjx1 = _fjsp_madd_v2r8(dx11,fscal,fjx1);
1118 fjy1 = _fjsp_madd_v2r8(dy11,fscal,fjy1);
1119 fjz1 = _fjsp_madd_v2r8(dz11,fscal,fjz1);
1123 /**************************
1124 * CALCULATE INTERACTIONS *
1125 **************************/
1127 if (gmx_fjsp_any_lt_v2r8(rsq12,rcutoff2))
1130 r12 = _fjsp_mul_v2r8(rsq12,rinv12);
1132 /* EWALD ELECTROSTATICS */
1134 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1135 ewrt = _fjsp_mul_v2r8(r12,ewtabscale);
1136 itab_tmp = _fjsp_dtox_v2r8(ewrt);
1137 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1138 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1140 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1141 ewtabD = _fjsp_setzero_v2r8();
1142 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1143 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1144 ewtabFn = _fjsp_setzero_v2r8();
1145 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1146 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1147 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1148 velec = _fjsp_mul_v2r8(qq12,_fjsp_sub_v2r8(rinv12,velec));
1149 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq12,rinv12),_fjsp_sub_v2r8(rinvsq12,felec));
1151 d = _fjsp_sub_v2r8(r12,rswitch);
1152 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1153 d2 = _fjsp_mul_v2r8(d,d);
1154 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1156 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1158 /* Evaluate switch function */
1159 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1160 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv12,_fjsp_mul_v2r8(velec,dsw)) );
1161 velec = _fjsp_mul_v2r8(velec,sw);
1162 cutoff_mask = _fjsp_cmplt_v2r8(rsq12,rcutoff2);
1164 /* Update potential sum for this i atom from the interaction with this j atom. */
1165 velec = _fjsp_and_v2r8(velec,cutoff_mask);
1166 velec = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
1167 velecsum = _fjsp_add_v2r8(velecsum,velec);
1171 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
1173 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1175 /* Update vectorial force */
1176 fix1 = _fjsp_madd_v2r8(dx12,fscal,fix1);
1177 fiy1 = _fjsp_madd_v2r8(dy12,fscal,fiy1);
1178 fiz1 = _fjsp_madd_v2r8(dz12,fscal,fiz1);
1180 fjx2 = _fjsp_madd_v2r8(dx12,fscal,fjx2);
1181 fjy2 = _fjsp_madd_v2r8(dy12,fscal,fjy2);
1182 fjz2 = _fjsp_madd_v2r8(dz12,fscal,fjz2);
1186 /**************************
1187 * CALCULATE INTERACTIONS *
1188 **************************/
1190 if (gmx_fjsp_any_lt_v2r8(rsq13,rcutoff2))
1193 r13 = _fjsp_mul_v2r8(rsq13,rinv13);
1195 /* EWALD ELECTROSTATICS */
1197 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1198 ewrt = _fjsp_mul_v2r8(r13,ewtabscale);
1199 itab_tmp = _fjsp_dtox_v2r8(ewrt);
1200 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1201 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1203 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1204 ewtabD = _fjsp_setzero_v2r8();
1205 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1206 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1207 ewtabFn = _fjsp_setzero_v2r8();
1208 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1209 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1210 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1211 velec = _fjsp_mul_v2r8(qq13,_fjsp_sub_v2r8(rinv13,velec));
1212 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq13,rinv13),_fjsp_sub_v2r8(rinvsq13,felec));
1214 d = _fjsp_sub_v2r8(r13,rswitch);
1215 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1216 d2 = _fjsp_mul_v2r8(d,d);
1217 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1219 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1221 /* Evaluate switch function */
1222 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1223 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv13,_fjsp_mul_v2r8(velec,dsw)) );
1224 velec = _fjsp_mul_v2r8(velec,sw);
1225 cutoff_mask = _fjsp_cmplt_v2r8(rsq13,rcutoff2);
1227 /* Update potential sum for this i atom from the interaction with this j atom. */
1228 velec = _fjsp_and_v2r8(velec,cutoff_mask);
1229 velec = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
1230 velecsum = _fjsp_add_v2r8(velecsum,velec);
1234 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
1236 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1238 /* Update vectorial force */
1239 fix1 = _fjsp_madd_v2r8(dx13,fscal,fix1);
1240 fiy1 = _fjsp_madd_v2r8(dy13,fscal,fiy1);
1241 fiz1 = _fjsp_madd_v2r8(dz13,fscal,fiz1);
1243 fjx3 = _fjsp_madd_v2r8(dx13,fscal,fjx3);
1244 fjy3 = _fjsp_madd_v2r8(dy13,fscal,fjy3);
1245 fjz3 = _fjsp_madd_v2r8(dz13,fscal,fjz3);
1249 /**************************
1250 * CALCULATE INTERACTIONS *
1251 **************************/
1253 if (gmx_fjsp_any_lt_v2r8(rsq21,rcutoff2))
1256 r21 = _fjsp_mul_v2r8(rsq21,rinv21);
1258 /* EWALD ELECTROSTATICS */
1260 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1261 ewrt = _fjsp_mul_v2r8(r21,ewtabscale);
1262 itab_tmp = _fjsp_dtox_v2r8(ewrt);
1263 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1264 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1266 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1267 ewtabD = _fjsp_setzero_v2r8();
1268 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1269 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1270 ewtabFn = _fjsp_setzero_v2r8();
1271 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1272 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1273 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1274 velec = _fjsp_mul_v2r8(qq21,_fjsp_sub_v2r8(rinv21,velec));
1275 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq21,rinv21),_fjsp_sub_v2r8(rinvsq21,felec));
1277 d = _fjsp_sub_v2r8(r21,rswitch);
1278 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1279 d2 = _fjsp_mul_v2r8(d,d);
1280 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1282 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1284 /* Evaluate switch function */
1285 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1286 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv21,_fjsp_mul_v2r8(velec,dsw)) );
1287 velec = _fjsp_mul_v2r8(velec,sw);
1288 cutoff_mask = _fjsp_cmplt_v2r8(rsq21,rcutoff2);
1290 /* Update potential sum for this i atom from the interaction with this j atom. */
1291 velec = _fjsp_and_v2r8(velec,cutoff_mask);
1292 velec = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
1293 velecsum = _fjsp_add_v2r8(velecsum,velec);
1297 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
1299 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1301 /* Update vectorial force */
1302 fix2 = _fjsp_madd_v2r8(dx21,fscal,fix2);
1303 fiy2 = _fjsp_madd_v2r8(dy21,fscal,fiy2);
1304 fiz2 = _fjsp_madd_v2r8(dz21,fscal,fiz2);
1306 fjx1 = _fjsp_madd_v2r8(dx21,fscal,fjx1);
1307 fjy1 = _fjsp_madd_v2r8(dy21,fscal,fjy1);
1308 fjz1 = _fjsp_madd_v2r8(dz21,fscal,fjz1);
1312 /**************************
1313 * CALCULATE INTERACTIONS *
1314 **************************/
1316 if (gmx_fjsp_any_lt_v2r8(rsq22,rcutoff2))
1319 r22 = _fjsp_mul_v2r8(rsq22,rinv22);
1321 /* EWALD ELECTROSTATICS */
1323 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1324 ewrt = _fjsp_mul_v2r8(r22,ewtabscale);
1325 itab_tmp = _fjsp_dtox_v2r8(ewrt);
1326 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1327 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1329 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1330 ewtabD = _fjsp_setzero_v2r8();
1331 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1332 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1333 ewtabFn = _fjsp_setzero_v2r8();
1334 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1335 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1336 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1337 velec = _fjsp_mul_v2r8(qq22,_fjsp_sub_v2r8(rinv22,velec));
1338 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq22,rinv22),_fjsp_sub_v2r8(rinvsq22,felec));
1340 d = _fjsp_sub_v2r8(r22,rswitch);
1341 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1342 d2 = _fjsp_mul_v2r8(d,d);
1343 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1345 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1347 /* Evaluate switch function */
1348 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1349 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv22,_fjsp_mul_v2r8(velec,dsw)) );
1350 velec = _fjsp_mul_v2r8(velec,sw);
1351 cutoff_mask = _fjsp_cmplt_v2r8(rsq22,rcutoff2);
1353 /* Update potential sum for this i atom from the interaction with this j atom. */
1354 velec = _fjsp_and_v2r8(velec,cutoff_mask);
1355 velec = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
1356 velecsum = _fjsp_add_v2r8(velecsum,velec);
1360 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
1362 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1364 /* Update vectorial force */
1365 fix2 = _fjsp_madd_v2r8(dx22,fscal,fix2);
1366 fiy2 = _fjsp_madd_v2r8(dy22,fscal,fiy2);
1367 fiz2 = _fjsp_madd_v2r8(dz22,fscal,fiz2);
1369 fjx2 = _fjsp_madd_v2r8(dx22,fscal,fjx2);
1370 fjy2 = _fjsp_madd_v2r8(dy22,fscal,fjy2);
1371 fjz2 = _fjsp_madd_v2r8(dz22,fscal,fjz2);
1375 /**************************
1376 * CALCULATE INTERACTIONS *
1377 **************************/
1379 if (gmx_fjsp_any_lt_v2r8(rsq23,rcutoff2))
1382 r23 = _fjsp_mul_v2r8(rsq23,rinv23);
1384 /* EWALD ELECTROSTATICS */
1386 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1387 ewrt = _fjsp_mul_v2r8(r23,ewtabscale);
1388 itab_tmp = _fjsp_dtox_v2r8(ewrt);
1389 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1390 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1392 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1393 ewtabD = _fjsp_setzero_v2r8();
1394 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1395 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1396 ewtabFn = _fjsp_setzero_v2r8();
1397 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1398 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1399 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1400 velec = _fjsp_mul_v2r8(qq23,_fjsp_sub_v2r8(rinv23,velec));
1401 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq23,rinv23),_fjsp_sub_v2r8(rinvsq23,felec));
1403 d = _fjsp_sub_v2r8(r23,rswitch);
1404 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1405 d2 = _fjsp_mul_v2r8(d,d);
1406 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1408 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1410 /* Evaluate switch function */
1411 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1412 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv23,_fjsp_mul_v2r8(velec,dsw)) );
1413 velec = _fjsp_mul_v2r8(velec,sw);
1414 cutoff_mask = _fjsp_cmplt_v2r8(rsq23,rcutoff2);
1416 /* Update potential sum for this i atom from the interaction with this j atom. */
1417 velec = _fjsp_and_v2r8(velec,cutoff_mask);
1418 velec = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
1419 velecsum = _fjsp_add_v2r8(velecsum,velec);
1423 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
1425 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1427 /* Update vectorial force */
1428 fix2 = _fjsp_madd_v2r8(dx23,fscal,fix2);
1429 fiy2 = _fjsp_madd_v2r8(dy23,fscal,fiy2);
1430 fiz2 = _fjsp_madd_v2r8(dz23,fscal,fiz2);
1432 fjx3 = _fjsp_madd_v2r8(dx23,fscal,fjx3);
1433 fjy3 = _fjsp_madd_v2r8(dy23,fscal,fjy3);
1434 fjz3 = _fjsp_madd_v2r8(dz23,fscal,fjz3);
1438 /**************************
1439 * CALCULATE INTERACTIONS *
1440 **************************/
1442 if (gmx_fjsp_any_lt_v2r8(rsq31,rcutoff2))
1445 r31 = _fjsp_mul_v2r8(rsq31,rinv31);
1447 /* EWALD ELECTROSTATICS */
1449 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1450 ewrt = _fjsp_mul_v2r8(r31,ewtabscale);
1451 itab_tmp = _fjsp_dtox_v2r8(ewrt);
1452 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1453 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1455 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1456 ewtabD = _fjsp_setzero_v2r8();
1457 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1458 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1459 ewtabFn = _fjsp_setzero_v2r8();
1460 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1461 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1462 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1463 velec = _fjsp_mul_v2r8(qq31,_fjsp_sub_v2r8(rinv31,velec));
1464 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq31,rinv31),_fjsp_sub_v2r8(rinvsq31,felec));
1466 d = _fjsp_sub_v2r8(r31,rswitch);
1467 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1468 d2 = _fjsp_mul_v2r8(d,d);
1469 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1471 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1473 /* Evaluate switch function */
1474 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1475 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv31,_fjsp_mul_v2r8(velec,dsw)) );
1476 velec = _fjsp_mul_v2r8(velec,sw);
1477 cutoff_mask = _fjsp_cmplt_v2r8(rsq31,rcutoff2);
1479 /* Update potential sum for this i atom from the interaction with this j atom. */
1480 velec = _fjsp_and_v2r8(velec,cutoff_mask);
1481 velec = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
1482 velecsum = _fjsp_add_v2r8(velecsum,velec);
1486 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
1488 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1490 /* Update vectorial force */
1491 fix3 = _fjsp_madd_v2r8(dx31,fscal,fix3);
1492 fiy3 = _fjsp_madd_v2r8(dy31,fscal,fiy3);
1493 fiz3 = _fjsp_madd_v2r8(dz31,fscal,fiz3);
1495 fjx1 = _fjsp_madd_v2r8(dx31,fscal,fjx1);
1496 fjy1 = _fjsp_madd_v2r8(dy31,fscal,fjy1);
1497 fjz1 = _fjsp_madd_v2r8(dz31,fscal,fjz1);
1501 /**************************
1502 * CALCULATE INTERACTIONS *
1503 **************************/
1505 if (gmx_fjsp_any_lt_v2r8(rsq32,rcutoff2))
1508 r32 = _fjsp_mul_v2r8(rsq32,rinv32);
1510 /* EWALD ELECTROSTATICS */
1512 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1513 ewrt = _fjsp_mul_v2r8(r32,ewtabscale);
1514 itab_tmp = _fjsp_dtox_v2r8(ewrt);
1515 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1516 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1518 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1519 ewtabD = _fjsp_setzero_v2r8();
1520 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1521 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1522 ewtabFn = _fjsp_setzero_v2r8();
1523 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1524 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1525 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1526 velec = _fjsp_mul_v2r8(qq32,_fjsp_sub_v2r8(rinv32,velec));
1527 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq32,rinv32),_fjsp_sub_v2r8(rinvsq32,felec));
1529 d = _fjsp_sub_v2r8(r32,rswitch);
1530 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1531 d2 = _fjsp_mul_v2r8(d,d);
1532 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1534 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1536 /* Evaluate switch function */
1537 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1538 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv32,_fjsp_mul_v2r8(velec,dsw)) );
1539 velec = _fjsp_mul_v2r8(velec,sw);
1540 cutoff_mask = _fjsp_cmplt_v2r8(rsq32,rcutoff2);
1542 /* Update potential sum for this i atom from the interaction with this j atom. */
1543 velec = _fjsp_and_v2r8(velec,cutoff_mask);
1544 velec = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
1545 velecsum = _fjsp_add_v2r8(velecsum,velec);
1549 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
1551 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1553 /* Update vectorial force */
1554 fix3 = _fjsp_madd_v2r8(dx32,fscal,fix3);
1555 fiy3 = _fjsp_madd_v2r8(dy32,fscal,fiy3);
1556 fiz3 = _fjsp_madd_v2r8(dz32,fscal,fiz3);
1558 fjx2 = _fjsp_madd_v2r8(dx32,fscal,fjx2);
1559 fjy2 = _fjsp_madd_v2r8(dy32,fscal,fjy2);
1560 fjz2 = _fjsp_madd_v2r8(dz32,fscal,fjz2);
1564 /**************************
1565 * CALCULATE INTERACTIONS *
1566 **************************/
1568 if (gmx_fjsp_any_lt_v2r8(rsq33,rcutoff2))
1571 r33 = _fjsp_mul_v2r8(rsq33,rinv33);
1573 /* EWALD ELECTROSTATICS */
1575 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1576 ewrt = _fjsp_mul_v2r8(r33,ewtabscale);
1577 itab_tmp = _fjsp_dtox_v2r8(ewrt);
1578 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1579 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1581 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1582 ewtabD = _fjsp_setzero_v2r8();
1583 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1584 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1585 ewtabFn = _fjsp_setzero_v2r8();
1586 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1587 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1588 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1589 velec = _fjsp_mul_v2r8(qq33,_fjsp_sub_v2r8(rinv33,velec));
1590 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq33,rinv33),_fjsp_sub_v2r8(rinvsq33,felec));
1592 d = _fjsp_sub_v2r8(r33,rswitch);
1593 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1594 d2 = _fjsp_mul_v2r8(d,d);
1595 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1597 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1599 /* Evaluate switch function */
1600 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1601 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv33,_fjsp_mul_v2r8(velec,dsw)) );
1602 velec = _fjsp_mul_v2r8(velec,sw);
1603 cutoff_mask = _fjsp_cmplt_v2r8(rsq33,rcutoff2);
1605 /* Update potential sum for this i atom from the interaction with this j atom. */
1606 velec = _fjsp_and_v2r8(velec,cutoff_mask);
1607 velec = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
1608 velecsum = _fjsp_add_v2r8(velecsum,velec);
1612 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
1614 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1616 /* Update vectorial force */
1617 fix3 = _fjsp_madd_v2r8(dx33,fscal,fix3);
1618 fiy3 = _fjsp_madd_v2r8(dy33,fscal,fiy3);
1619 fiz3 = _fjsp_madd_v2r8(dz33,fscal,fiz3);
1621 fjx3 = _fjsp_madd_v2r8(dx33,fscal,fjx3);
1622 fjy3 = _fjsp_madd_v2r8(dy33,fscal,fjy3);
1623 fjz3 = _fjsp_madd_v2r8(dz33,fscal,fjz3);
1627 gmx_fjsp_decrement_4rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1629 /* Inner loop uses 677 flops */
1632 /* End of innermost loop */
1634 gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1635 f+i_coord_offset,fshift+i_shift_offset);
1638 /* Update potential energies */
1639 gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
1640 gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
1642 /* Increment number of inner iterations */
1643 inneriter += j_index_end - j_index_start;
1645 /* Outer loop uses 26 flops */
1648 /* Increment number of outer iterations */
1651 /* Update outer/inner flops */
1653 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*677);
1656 * Gromacs nonbonded kernel: nb_kernel_ElecEwSw_VdwLJSw_GeomW4W4_F_sparc64_hpc_ace_double
1657 * Electrostatics interaction: Ewald
1658 * VdW interaction: LennardJones
1659 * Geometry: Water4-Water4
1660 * Calculate force/pot: Force
1663 nb_kernel_ElecEwSw_VdwLJSw_GeomW4W4_F_sparc64_hpc_ace_double
1664 (t_nblist * gmx_restrict nlist,
1665 rvec * gmx_restrict xx,
1666 rvec * gmx_restrict ff,
1667 t_forcerec * gmx_restrict fr,
1668 t_mdatoms * gmx_restrict mdatoms,
1669 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1670 t_nrnb * gmx_restrict nrnb)
1672 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1673 * just 0 for non-waters.
1674 * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
1675 * jnr indices corresponding to data put in the four positions in the SIMD register.
1677 int i_shift_offset,i_coord_offset,outeriter,inneriter;
1678 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1680 int j_coord_offsetA,j_coord_offsetB;
1681 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
1682 real rcutoff_scalar;
1683 real *shiftvec,*fshift,*x,*f;
1684 _fjsp_v2r8 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1686 _fjsp_v2r8 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1688 _fjsp_v2r8 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1690 _fjsp_v2r8 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1692 _fjsp_v2r8 ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1693 int vdwjidx0A,vdwjidx0B;
1694 _fjsp_v2r8 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1695 int vdwjidx1A,vdwjidx1B;
1696 _fjsp_v2r8 jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1697 int vdwjidx2A,vdwjidx2B;
1698 _fjsp_v2r8 jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1699 int vdwjidx3A,vdwjidx3B;
1700 _fjsp_v2r8 jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1701 _fjsp_v2r8 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1702 _fjsp_v2r8 dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1703 _fjsp_v2r8 dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1704 _fjsp_v2r8 dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1705 _fjsp_v2r8 dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1706 _fjsp_v2r8 dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1707 _fjsp_v2r8 dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1708 _fjsp_v2r8 dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1709 _fjsp_v2r8 dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1710 _fjsp_v2r8 dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1711 _fjsp_v2r8 velec,felec,velecsum,facel,crf,krf,krf2;
1714 _fjsp_v2r8 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1717 _fjsp_v2r8 one_sixth = gmx_fjsp_set1_v2r8(1.0/6.0);
1718 _fjsp_v2r8 one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
1719 _fjsp_v2r8 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1721 _fjsp_v2r8 rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
1722 real rswitch_scalar,d_scalar;
1723 _fjsp_v2r8 itab_tmp;
1724 _fjsp_v2r8 dummy_mask,cutoff_mask;
1725 _fjsp_v2r8 one = gmx_fjsp_set1_v2r8(1.0);
1726 _fjsp_v2r8 two = gmx_fjsp_set1_v2r8(2.0);
1727 union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
1734 jindex = nlist->jindex;
1736 shiftidx = nlist->shift;
1738 shiftvec = fr->shift_vec[0];
1739 fshift = fr->fshift[0];
1740 facel = gmx_fjsp_set1_v2r8(fr->epsfac);
1741 charge = mdatoms->chargeA;
1742 nvdwtype = fr->ntype;
1743 vdwparam = fr->nbfp;
1744 vdwtype = mdatoms->typeA;
1746 sh_ewald = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
1747 ewtab = fr->ic->tabq_coul_FDV0;
1748 ewtabscale = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
1749 ewtabhalfspace = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
1751 /* Setup water-specific parameters */
1752 inr = nlist->iinr[0];
1753 iq1 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
1754 iq2 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
1755 iq3 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
1756 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
1758 jq1 = gmx_fjsp_set1_v2r8(charge[inr+1]);
1759 jq2 = gmx_fjsp_set1_v2r8(charge[inr+2]);
1760 jq3 = gmx_fjsp_set1_v2r8(charge[inr+3]);
1761 vdwjidx0A = 2*vdwtype[inr+0];
1762 c6_00 = gmx_fjsp_set1_v2r8(vdwparam[vdwioffset0+vdwjidx0A]);
1763 c12_00 = gmx_fjsp_set1_v2r8(vdwparam[vdwioffset0+vdwjidx0A+1]);
1764 qq11 = _fjsp_mul_v2r8(iq1,jq1);
1765 qq12 = _fjsp_mul_v2r8(iq1,jq2);
1766 qq13 = _fjsp_mul_v2r8(iq1,jq3);
1767 qq21 = _fjsp_mul_v2r8(iq2,jq1);
1768 qq22 = _fjsp_mul_v2r8(iq2,jq2);
1769 qq23 = _fjsp_mul_v2r8(iq2,jq3);
1770 qq31 = _fjsp_mul_v2r8(iq3,jq1);
1771 qq32 = _fjsp_mul_v2r8(iq3,jq2);
1772 qq33 = _fjsp_mul_v2r8(iq3,jq3);
1774 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
1775 rcutoff_scalar = fr->rcoulomb;
1776 rcutoff = gmx_fjsp_set1_v2r8(rcutoff_scalar);
1777 rcutoff2 = _fjsp_mul_v2r8(rcutoff,rcutoff);
1779 rswitch_scalar = fr->rcoulomb_switch;
1780 rswitch = gmx_fjsp_set1_v2r8(rswitch_scalar);
1781 /* Setup switch parameters */
1782 d_scalar = rcutoff_scalar-rswitch_scalar;
1783 d = gmx_fjsp_set1_v2r8(d_scalar);
1784 swV3 = gmx_fjsp_set1_v2r8(-10.0/(d_scalar*d_scalar*d_scalar));
1785 swV4 = gmx_fjsp_set1_v2r8( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
1786 swV5 = gmx_fjsp_set1_v2r8( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
1787 swF2 = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar));
1788 swF3 = gmx_fjsp_set1_v2r8( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
1789 swF4 = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
1791 /* Avoid stupid compiler warnings */
1793 j_coord_offsetA = 0;
1794 j_coord_offsetB = 0;
1799 /* Start outer loop over neighborlists */
1800 for(iidx=0; iidx<nri; iidx++)
1802 /* Load shift vector for this list */
1803 i_shift_offset = DIM*shiftidx[iidx];
1805 /* Load limits for loop over neighbors */
1806 j_index_start = jindex[iidx];
1807 j_index_end = jindex[iidx+1];
1809 /* Get outer coordinate index */
1811 i_coord_offset = DIM*inr;
1813 /* Load i particle coords and add shift vector */
1814 gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
1815 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1817 fix0 = _fjsp_setzero_v2r8();
1818 fiy0 = _fjsp_setzero_v2r8();
1819 fiz0 = _fjsp_setzero_v2r8();
1820 fix1 = _fjsp_setzero_v2r8();
1821 fiy1 = _fjsp_setzero_v2r8();
1822 fiz1 = _fjsp_setzero_v2r8();
1823 fix2 = _fjsp_setzero_v2r8();
1824 fiy2 = _fjsp_setzero_v2r8();
1825 fiz2 = _fjsp_setzero_v2r8();
1826 fix3 = _fjsp_setzero_v2r8();
1827 fiy3 = _fjsp_setzero_v2r8();
1828 fiz3 = _fjsp_setzero_v2r8();
1830 /* Start inner kernel loop */
1831 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
1834 /* Get j neighbor index, and coordinate index */
1836 jnrB = jjnr[jidx+1];
1837 j_coord_offsetA = DIM*jnrA;
1838 j_coord_offsetB = DIM*jnrB;
1840 /* load j atom coordinates */
1841 gmx_fjsp_load_4rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
1842 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1843 &jy2,&jz2,&jx3,&jy3,&jz3);
1845 /* Calculate displacement vector */
1846 dx00 = _fjsp_sub_v2r8(ix0,jx0);
1847 dy00 = _fjsp_sub_v2r8(iy0,jy0);
1848 dz00 = _fjsp_sub_v2r8(iz0,jz0);
1849 dx11 = _fjsp_sub_v2r8(ix1,jx1);
1850 dy11 = _fjsp_sub_v2r8(iy1,jy1);
1851 dz11 = _fjsp_sub_v2r8(iz1,jz1);
1852 dx12 = _fjsp_sub_v2r8(ix1,jx2);
1853 dy12 = _fjsp_sub_v2r8(iy1,jy2);
1854 dz12 = _fjsp_sub_v2r8(iz1,jz2);
1855 dx13 = _fjsp_sub_v2r8(ix1,jx3);
1856 dy13 = _fjsp_sub_v2r8(iy1,jy3);
1857 dz13 = _fjsp_sub_v2r8(iz1,jz3);
1858 dx21 = _fjsp_sub_v2r8(ix2,jx1);
1859 dy21 = _fjsp_sub_v2r8(iy2,jy1);
1860 dz21 = _fjsp_sub_v2r8(iz2,jz1);
1861 dx22 = _fjsp_sub_v2r8(ix2,jx2);
1862 dy22 = _fjsp_sub_v2r8(iy2,jy2);
1863 dz22 = _fjsp_sub_v2r8(iz2,jz2);
1864 dx23 = _fjsp_sub_v2r8(ix2,jx3);
1865 dy23 = _fjsp_sub_v2r8(iy2,jy3);
1866 dz23 = _fjsp_sub_v2r8(iz2,jz3);
1867 dx31 = _fjsp_sub_v2r8(ix3,jx1);
1868 dy31 = _fjsp_sub_v2r8(iy3,jy1);
1869 dz31 = _fjsp_sub_v2r8(iz3,jz1);
1870 dx32 = _fjsp_sub_v2r8(ix3,jx2);
1871 dy32 = _fjsp_sub_v2r8(iy3,jy2);
1872 dz32 = _fjsp_sub_v2r8(iz3,jz2);
1873 dx33 = _fjsp_sub_v2r8(ix3,jx3);
1874 dy33 = _fjsp_sub_v2r8(iy3,jy3);
1875 dz33 = _fjsp_sub_v2r8(iz3,jz3);
1877 /* Calculate squared distance and things based on it */
1878 rsq00 = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
1879 rsq11 = gmx_fjsp_calc_rsq_v2r8(dx11,dy11,dz11);
1880 rsq12 = gmx_fjsp_calc_rsq_v2r8(dx12,dy12,dz12);
1881 rsq13 = gmx_fjsp_calc_rsq_v2r8(dx13,dy13,dz13);
1882 rsq21 = gmx_fjsp_calc_rsq_v2r8(dx21,dy21,dz21);
1883 rsq22 = gmx_fjsp_calc_rsq_v2r8(dx22,dy22,dz22);
1884 rsq23 = gmx_fjsp_calc_rsq_v2r8(dx23,dy23,dz23);
1885 rsq31 = gmx_fjsp_calc_rsq_v2r8(dx31,dy31,dz31);
1886 rsq32 = gmx_fjsp_calc_rsq_v2r8(dx32,dy32,dz32);
1887 rsq33 = gmx_fjsp_calc_rsq_v2r8(dx33,dy33,dz33);
1889 rinv00 = gmx_fjsp_invsqrt_v2r8(rsq00);
1890 rinv11 = gmx_fjsp_invsqrt_v2r8(rsq11);
1891 rinv12 = gmx_fjsp_invsqrt_v2r8(rsq12);
1892 rinv13 = gmx_fjsp_invsqrt_v2r8(rsq13);
1893 rinv21 = gmx_fjsp_invsqrt_v2r8(rsq21);
1894 rinv22 = gmx_fjsp_invsqrt_v2r8(rsq22);
1895 rinv23 = gmx_fjsp_invsqrt_v2r8(rsq23);
1896 rinv31 = gmx_fjsp_invsqrt_v2r8(rsq31);
1897 rinv32 = gmx_fjsp_invsqrt_v2r8(rsq32);
1898 rinv33 = gmx_fjsp_invsqrt_v2r8(rsq33);
1900 rinvsq00 = _fjsp_mul_v2r8(rinv00,rinv00);
1901 rinvsq11 = _fjsp_mul_v2r8(rinv11,rinv11);
1902 rinvsq12 = _fjsp_mul_v2r8(rinv12,rinv12);
1903 rinvsq13 = _fjsp_mul_v2r8(rinv13,rinv13);
1904 rinvsq21 = _fjsp_mul_v2r8(rinv21,rinv21);
1905 rinvsq22 = _fjsp_mul_v2r8(rinv22,rinv22);
1906 rinvsq23 = _fjsp_mul_v2r8(rinv23,rinv23);
1907 rinvsq31 = _fjsp_mul_v2r8(rinv31,rinv31);
1908 rinvsq32 = _fjsp_mul_v2r8(rinv32,rinv32);
1909 rinvsq33 = _fjsp_mul_v2r8(rinv33,rinv33);
1911 fjx0 = _fjsp_setzero_v2r8();
1912 fjy0 = _fjsp_setzero_v2r8();
1913 fjz0 = _fjsp_setzero_v2r8();
1914 fjx1 = _fjsp_setzero_v2r8();
1915 fjy1 = _fjsp_setzero_v2r8();
1916 fjz1 = _fjsp_setzero_v2r8();
1917 fjx2 = _fjsp_setzero_v2r8();
1918 fjy2 = _fjsp_setzero_v2r8();
1919 fjz2 = _fjsp_setzero_v2r8();
1920 fjx3 = _fjsp_setzero_v2r8();
1921 fjy3 = _fjsp_setzero_v2r8();
1922 fjz3 = _fjsp_setzero_v2r8();
1924 /**************************
1925 * CALCULATE INTERACTIONS *
1926 **************************/
1928 if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
1931 r00 = _fjsp_mul_v2r8(rsq00,rinv00);
1933 /* LENNARD-JONES DISPERSION/REPULSION */
1935 rinvsix = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
1936 vvdw6 = _fjsp_mul_v2r8(c6_00,rinvsix);
1937 vvdw12 = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
1938 vvdw = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
1939 fvdw = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
1941 d = _fjsp_sub_v2r8(r00,rswitch);
1942 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1943 d2 = _fjsp_mul_v2r8(d,d);
1944 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1946 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1948 /* Evaluate switch function */
1949 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1950 fvdw = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
1951 cutoff_mask = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
1955 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
1957 /* Update vectorial force */
1958 fix0 = _fjsp_madd_v2r8(dx00,fscal,fix0);
1959 fiy0 = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1960 fiz0 = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1962 fjx0 = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1963 fjy0 = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1964 fjz0 = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1968 /**************************
1969 * CALCULATE INTERACTIONS *
1970 **************************/
1972 if (gmx_fjsp_any_lt_v2r8(rsq11,rcutoff2))
1975 r11 = _fjsp_mul_v2r8(rsq11,rinv11);
1977 /* EWALD ELECTROSTATICS */
1979 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1980 ewrt = _fjsp_mul_v2r8(r11,ewtabscale);
1981 itab_tmp = _fjsp_dtox_v2r8(ewrt);
1982 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1983 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1985 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1986 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
1987 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1988 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1989 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
1990 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1991 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1992 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1993 velec = _fjsp_mul_v2r8(qq11,_fjsp_sub_v2r8(rinv11,velec));
1994 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq11,rinv11),_fjsp_sub_v2r8(rinvsq11,felec));
1996 d = _fjsp_sub_v2r8(r11,rswitch);
1997 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1998 d2 = _fjsp_mul_v2r8(d,d);
1999 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
2001 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
2003 /* Evaluate switch function */
2004 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2005 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv11,_fjsp_mul_v2r8(velec,dsw)) );
2006 cutoff_mask = _fjsp_cmplt_v2r8(rsq11,rcutoff2);
2010 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
2012 /* Update vectorial force */
2013 fix1 = _fjsp_madd_v2r8(dx11,fscal,fix1);
2014 fiy1 = _fjsp_madd_v2r8(dy11,fscal,fiy1);
2015 fiz1 = _fjsp_madd_v2r8(dz11,fscal,fiz1);
2017 fjx1 = _fjsp_madd_v2r8(dx11,fscal,fjx1);
2018 fjy1 = _fjsp_madd_v2r8(dy11,fscal,fjy1);
2019 fjz1 = _fjsp_madd_v2r8(dz11,fscal,fjz1);
2023 /**************************
2024 * CALCULATE INTERACTIONS *
2025 **************************/
2027 if (gmx_fjsp_any_lt_v2r8(rsq12,rcutoff2))
2030 r12 = _fjsp_mul_v2r8(rsq12,rinv12);
2032 /* EWALD ELECTROSTATICS */
2034 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2035 ewrt = _fjsp_mul_v2r8(r12,ewtabscale);
2036 itab_tmp = _fjsp_dtox_v2r8(ewrt);
2037 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
2038 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
2040 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
2041 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
2042 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
2043 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
2044 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
2045 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
2046 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
2047 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
2048 velec = _fjsp_mul_v2r8(qq12,_fjsp_sub_v2r8(rinv12,velec));
2049 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq12,rinv12),_fjsp_sub_v2r8(rinvsq12,felec));
2051 d = _fjsp_sub_v2r8(r12,rswitch);
2052 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
2053 d2 = _fjsp_mul_v2r8(d,d);
2054 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
2056 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
2058 /* Evaluate switch function */
2059 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2060 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv12,_fjsp_mul_v2r8(velec,dsw)) );
2061 cutoff_mask = _fjsp_cmplt_v2r8(rsq12,rcutoff2);
2065 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
2067 /* Update vectorial force */
2068 fix1 = _fjsp_madd_v2r8(dx12,fscal,fix1);
2069 fiy1 = _fjsp_madd_v2r8(dy12,fscal,fiy1);
2070 fiz1 = _fjsp_madd_v2r8(dz12,fscal,fiz1);
2072 fjx2 = _fjsp_madd_v2r8(dx12,fscal,fjx2);
2073 fjy2 = _fjsp_madd_v2r8(dy12,fscal,fjy2);
2074 fjz2 = _fjsp_madd_v2r8(dz12,fscal,fjz2);
2078 /**************************
2079 * CALCULATE INTERACTIONS *
2080 **************************/
2082 if (gmx_fjsp_any_lt_v2r8(rsq13,rcutoff2))
2085 r13 = _fjsp_mul_v2r8(rsq13,rinv13);
2087 /* EWALD ELECTROSTATICS */
2089 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2090 ewrt = _fjsp_mul_v2r8(r13,ewtabscale);
2091 itab_tmp = _fjsp_dtox_v2r8(ewrt);
2092 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
2093 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
2095 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
2096 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
2097 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
2098 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
2099 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
2100 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
2101 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
2102 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
2103 velec = _fjsp_mul_v2r8(qq13,_fjsp_sub_v2r8(rinv13,velec));
2104 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq13,rinv13),_fjsp_sub_v2r8(rinvsq13,felec));
2106 d = _fjsp_sub_v2r8(r13,rswitch);
2107 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
2108 d2 = _fjsp_mul_v2r8(d,d);
2109 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
2111 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
2113 /* Evaluate switch function */
2114 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2115 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv13,_fjsp_mul_v2r8(velec,dsw)) );
2116 cutoff_mask = _fjsp_cmplt_v2r8(rsq13,rcutoff2);
2120 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
2122 /* Update vectorial force */
2123 fix1 = _fjsp_madd_v2r8(dx13,fscal,fix1);
2124 fiy1 = _fjsp_madd_v2r8(dy13,fscal,fiy1);
2125 fiz1 = _fjsp_madd_v2r8(dz13,fscal,fiz1);
2127 fjx3 = _fjsp_madd_v2r8(dx13,fscal,fjx3);
2128 fjy3 = _fjsp_madd_v2r8(dy13,fscal,fjy3);
2129 fjz3 = _fjsp_madd_v2r8(dz13,fscal,fjz3);
2133 /**************************
2134 * CALCULATE INTERACTIONS *
2135 **************************/
2137 if (gmx_fjsp_any_lt_v2r8(rsq21,rcutoff2))
2140 r21 = _fjsp_mul_v2r8(rsq21,rinv21);
2142 /* EWALD ELECTROSTATICS */
2144 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2145 ewrt = _fjsp_mul_v2r8(r21,ewtabscale);
2146 itab_tmp = _fjsp_dtox_v2r8(ewrt);
2147 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
2148 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
2150 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
2151 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
2152 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
2153 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
2154 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
2155 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
2156 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
2157 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
2158 velec = _fjsp_mul_v2r8(qq21,_fjsp_sub_v2r8(rinv21,velec));
2159 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq21,rinv21),_fjsp_sub_v2r8(rinvsq21,felec));
2161 d = _fjsp_sub_v2r8(r21,rswitch);
2162 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
2163 d2 = _fjsp_mul_v2r8(d,d);
2164 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
2166 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
2168 /* Evaluate switch function */
2169 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2170 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv21,_fjsp_mul_v2r8(velec,dsw)) );
2171 cutoff_mask = _fjsp_cmplt_v2r8(rsq21,rcutoff2);
2175 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
2177 /* Update vectorial force */
2178 fix2 = _fjsp_madd_v2r8(dx21,fscal,fix2);
2179 fiy2 = _fjsp_madd_v2r8(dy21,fscal,fiy2);
2180 fiz2 = _fjsp_madd_v2r8(dz21,fscal,fiz2);
2182 fjx1 = _fjsp_madd_v2r8(dx21,fscal,fjx1);
2183 fjy1 = _fjsp_madd_v2r8(dy21,fscal,fjy1);
2184 fjz1 = _fjsp_madd_v2r8(dz21,fscal,fjz1);
2188 /**************************
2189 * CALCULATE INTERACTIONS *
2190 **************************/
2192 if (gmx_fjsp_any_lt_v2r8(rsq22,rcutoff2))
2195 r22 = _fjsp_mul_v2r8(rsq22,rinv22);
2197 /* EWALD ELECTROSTATICS */
2199 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2200 ewrt = _fjsp_mul_v2r8(r22,ewtabscale);
2201 itab_tmp = _fjsp_dtox_v2r8(ewrt);
2202 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
2203 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
2205 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
2206 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
2207 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
2208 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
2209 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
2210 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
2211 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
2212 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
2213 velec = _fjsp_mul_v2r8(qq22,_fjsp_sub_v2r8(rinv22,velec));
2214 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq22,rinv22),_fjsp_sub_v2r8(rinvsq22,felec));
2216 d = _fjsp_sub_v2r8(r22,rswitch);
2217 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
2218 d2 = _fjsp_mul_v2r8(d,d);
2219 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
2221 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
2223 /* Evaluate switch function */
2224 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2225 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv22,_fjsp_mul_v2r8(velec,dsw)) );
2226 cutoff_mask = _fjsp_cmplt_v2r8(rsq22,rcutoff2);
2230 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
2232 /* Update vectorial force */
2233 fix2 = _fjsp_madd_v2r8(dx22,fscal,fix2);
2234 fiy2 = _fjsp_madd_v2r8(dy22,fscal,fiy2);
2235 fiz2 = _fjsp_madd_v2r8(dz22,fscal,fiz2);
2237 fjx2 = _fjsp_madd_v2r8(dx22,fscal,fjx2);
2238 fjy2 = _fjsp_madd_v2r8(dy22,fscal,fjy2);
2239 fjz2 = _fjsp_madd_v2r8(dz22,fscal,fjz2);
2243 /**************************
2244 * CALCULATE INTERACTIONS *
2245 **************************/
2247 if (gmx_fjsp_any_lt_v2r8(rsq23,rcutoff2))
2250 r23 = _fjsp_mul_v2r8(rsq23,rinv23);
2252 /* EWALD ELECTROSTATICS */
2254 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2255 ewrt = _fjsp_mul_v2r8(r23,ewtabscale);
2256 itab_tmp = _fjsp_dtox_v2r8(ewrt);
2257 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
2258 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
2260 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
2261 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
2262 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
2263 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
2264 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
2265 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
2266 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
2267 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
2268 velec = _fjsp_mul_v2r8(qq23,_fjsp_sub_v2r8(rinv23,velec));
2269 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq23,rinv23),_fjsp_sub_v2r8(rinvsq23,felec));
2271 d = _fjsp_sub_v2r8(r23,rswitch);
2272 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
2273 d2 = _fjsp_mul_v2r8(d,d);
2274 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
2276 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
2278 /* Evaluate switch function */
2279 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2280 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv23,_fjsp_mul_v2r8(velec,dsw)) );
2281 cutoff_mask = _fjsp_cmplt_v2r8(rsq23,rcutoff2);
2285 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
2287 /* Update vectorial force */
2288 fix2 = _fjsp_madd_v2r8(dx23,fscal,fix2);
2289 fiy2 = _fjsp_madd_v2r8(dy23,fscal,fiy2);
2290 fiz2 = _fjsp_madd_v2r8(dz23,fscal,fiz2);
2292 fjx3 = _fjsp_madd_v2r8(dx23,fscal,fjx3);
2293 fjy3 = _fjsp_madd_v2r8(dy23,fscal,fjy3);
2294 fjz3 = _fjsp_madd_v2r8(dz23,fscal,fjz3);
2298 /**************************
2299 * CALCULATE INTERACTIONS *
2300 **************************/
2302 if (gmx_fjsp_any_lt_v2r8(rsq31,rcutoff2))
2305 r31 = _fjsp_mul_v2r8(rsq31,rinv31);
2307 /* EWALD ELECTROSTATICS */
2309 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2310 ewrt = _fjsp_mul_v2r8(r31,ewtabscale);
2311 itab_tmp = _fjsp_dtox_v2r8(ewrt);
2312 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
2313 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
2315 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
2316 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
2317 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
2318 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
2319 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
2320 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
2321 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
2322 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
2323 velec = _fjsp_mul_v2r8(qq31,_fjsp_sub_v2r8(rinv31,velec));
2324 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq31,rinv31),_fjsp_sub_v2r8(rinvsq31,felec));
2326 d = _fjsp_sub_v2r8(r31,rswitch);
2327 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
2328 d2 = _fjsp_mul_v2r8(d,d);
2329 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
2331 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
2333 /* Evaluate switch function */
2334 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2335 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv31,_fjsp_mul_v2r8(velec,dsw)) );
2336 cutoff_mask = _fjsp_cmplt_v2r8(rsq31,rcutoff2);
2340 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
2342 /* Update vectorial force */
2343 fix3 = _fjsp_madd_v2r8(dx31,fscal,fix3);
2344 fiy3 = _fjsp_madd_v2r8(dy31,fscal,fiy3);
2345 fiz3 = _fjsp_madd_v2r8(dz31,fscal,fiz3);
2347 fjx1 = _fjsp_madd_v2r8(dx31,fscal,fjx1);
2348 fjy1 = _fjsp_madd_v2r8(dy31,fscal,fjy1);
2349 fjz1 = _fjsp_madd_v2r8(dz31,fscal,fjz1);
2353 /**************************
2354 * CALCULATE INTERACTIONS *
2355 **************************/
2357 if (gmx_fjsp_any_lt_v2r8(rsq32,rcutoff2))
2360 r32 = _fjsp_mul_v2r8(rsq32,rinv32);
2362 /* EWALD ELECTROSTATICS */
2364 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2365 ewrt = _fjsp_mul_v2r8(r32,ewtabscale);
2366 itab_tmp = _fjsp_dtox_v2r8(ewrt);
2367 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
2368 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
2370 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
2371 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
2372 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
2373 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
2374 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
2375 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
2376 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
2377 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
2378 velec = _fjsp_mul_v2r8(qq32,_fjsp_sub_v2r8(rinv32,velec));
2379 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq32,rinv32),_fjsp_sub_v2r8(rinvsq32,felec));
2381 d = _fjsp_sub_v2r8(r32,rswitch);
2382 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
2383 d2 = _fjsp_mul_v2r8(d,d);
2384 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
2386 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
2388 /* Evaluate switch function */
2389 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2390 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv32,_fjsp_mul_v2r8(velec,dsw)) );
2391 cutoff_mask = _fjsp_cmplt_v2r8(rsq32,rcutoff2);
2395 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
2397 /* Update vectorial force */
2398 fix3 = _fjsp_madd_v2r8(dx32,fscal,fix3);
2399 fiy3 = _fjsp_madd_v2r8(dy32,fscal,fiy3);
2400 fiz3 = _fjsp_madd_v2r8(dz32,fscal,fiz3);
2402 fjx2 = _fjsp_madd_v2r8(dx32,fscal,fjx2);
2403 fjy2 = _fjsp_madd_v2r8(dy32,fscal,fjy2);
2404 fjz2 = _fjsp_madd_v2r8(dz32,fscal,fjz2);
2408 /**************************
2409 * CALCULATE INTERACTIONS *
2410 **************************/
2412 if (gmx_fjsp_any_lt_v2r8(rsq33,rcutoff2))
2415 r33 = _fjsp_mul_v2r8(rsq33,rinv33);
2417 /* EWALD ELECTROSTATICS */
2419 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2420 ewrt = _fjsp_mul_v2r8(r33,ewtabscale);
2421 itab_tmp = _fjsp_dtox_v2r8(ewrt);
2422 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
2423 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
2425 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
2426 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
2427 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
2428 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
2429 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
2430 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
2431 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
2432 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
2433 velec = _fjsp_mul_v2r8(qq33,_fjsp_sub_v2r8(rinv33,velec));
2434 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq33,rinv33),_fjsp_sub_v2r8(rinvsq33,felec));
2436 d = _fjsp_sub_v2r8(r33,rswitch);
2437 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
2438 d2 = _fjsp_mul_v2r8(d,d);
2439 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
2441 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
2443 /* Evaluate switch function */
2444 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2445 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv33,_fjsp_mul_v2r8(velec,dsw)) );
2446 cutoff_mask = _fjsp_cmplt_v2r8(rsq33,rcutoff2);
2450 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
2452 /* Update vectorial force */
2453 fix3 = _fjsp_madd_v2r8(dx33,fscal,fix3);
2454 fiy3 = _fjsp_madd_v2r8(dy33,fscal,fiy3);
2455 fiz3 = _fjsp_madd_v2r8(dz33,fscal,fiz3);
2457 fjx3 = _fjsp_madd_v2r8(dx33,fscal,fjx3);
2458 fjy3 = _fjsp_madd_v2r8(dy33,fscal,fjy3);
2459 fjz3 = _fjsp_madd_v2r8(dz33,fscal,fjz3);
2463 gmx_fjsp_decrement_4rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2465 /* Inner loop uses 647 flops */
2468 if(jidx<j_index_end)
2472 j_coord_offsetA = DIM*jnrA;
2474 /* load j atom coordinates */
2475 gmx_fjsp_load_4rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
2476 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
2477 &jy2,&jz2,&jx3,&jy3,&jz3);
2479 /* Calculate displacement vector */
2480 dx00 = _fjsp_sub_v2r8(ix0,jx0);
2481 dy00 = _fjsp_sub_v2r8(iy0,jy0);
2482 dz00 = _fjsp_sub_v2r8(iz0,jz0);
2483 dx11 = _fjsp_sub_v2r8(ix1,jx1);
2484 dy11 = _fjsp_sub_v2r8(iy1,jy1);
2485 dz11 = _fjsp_sub_v2r8(iz1,jz1);
2486 dx12 = _fjsp_sub_v2r8(ix1,jx2);
2487 dy12 = _fjsp_sub_v2r8(iy1,jy2);
2488 dz12 = _fjsp_sub_v2r8(iz1,jz2);
2489 dx13 = _fjsp_sub_v2r8(ix1,jx3);
2490 dy13 = _fjsp_sub_v2r8(iy1,jy3);
2491 dz13 = _fjsp_sub_v2r8(iz1,jz3);
2492 dx21 = _fjsp_sub_v2r8(ix2,jx1);
2493 dy21 = _fjsp_sub_v2r8(iy2,jy1);
2494 dz21 = _fjsp_sub_v2r8(iz2,jz1);
2495 dx22 = _fjsp_sub_v2r8(ix2,jx2);
2496 dy22 = _fjsp_sub_v2r8(iy2,jy2);
2497 dz22 = _fjsp_sub_v2r8(iz2,jz2);
2498 dx23 = _fjsp_sub_v2r8(ix2,jx3);
2499 dy23 = _fjsp_sub_v2r8(iy2,jy3);
2500 dz23 = _fjsp_sub_v2r8(iz2,jz3);
2501 dx31 = _fjsp_sub_v2r8(ix3,jx1);
2502 dy31 = _fjsp_sub_v2r8(iy3,jy1);
2503 dz31 = _fjsp_sub_v2r8(iz3,jz1);
2504 dx32 = _fjsp_sub_v2r8(ix3,jx2);
2505 dy32 = _fjsp_sub_v2r8(iy3,jy2);
2506 dz32 = _fjsp_sub_v2r8(iz3,jz2);
2507 dx33 = _fjsp_sub_v2r8(ix3,jx3);
2508 dy33 = _fjsp_sub_v2r8(iy3,jy3);
2509 dz33 = _fjsp_sub_v2r8(iz3,jz3);
2511 /* Calculate squared distance and things based on it */
2512 rsq00 = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
2513 rsq11 = gmx_fjsp_calc_rsq_v2r8(dx11,dy11,dz11);
2514 rsq12 = gmx_fjsp_calc_rsq_v2r8(dx12,dy12,dz12);
2515 rsq13 = gmx_fjsp_calc_rsq_v2r8(dx13,dy13,dz13);
2516 rsq21 = gmx_fjsp_calc_rsq_v2r8(dx21,dy21,dz21);
2517 rsq22 = gmx_fjsp_calc_rsq_v2r8(dx22,dy22,dz22);
2518 rsq23 = gmx_fjsp_calc_rsq_v2r8(dx23,dy23,dz23);
2519 rsq31 = gmx_fjsp_calc_rsq_v2r8(dx31,dy31,dz31);
2520 rsq32 = gmx_fjsp_calc_rsq_v2r8(dx32,dy32,dz32);
2521 rsq33 = gmx_fjsp_calc_rsq_v2r8(dx33,dy33,dz33);
2523 rinv00 = gmx_fjsp_invsqrt_v2r8(rsq00);
2524 rinv11 = gmx_fjsp_invsqrt_v2r8(rsq11);
2525 rinv12 = gmx_fjsp_invsqrt_v2r8(rsq12);
2526 rinv13 = gmx_fjsp_invsqrt_v2r8(rsq13);
2527 rinv21 = gmx_fjsp_invsqrt_v2r8(rsq21);
2528 rinv22 = gmx_fjsp_invsqrt_v2r8(rsq22);
2529 rinv23 = gmx_fjsp_invsqrt_v2r8(rsq23);
2530 rinv31 = gmx_fjsp_invsqrt_v2r8(rsq31);
2531 rinv32 = gmx_fjsp_invsqrt_v2r8(rsq32);
2532 rinv33 = gmx_fjsp_invsqrt_v2r8(rsq33);
2534 rinvsq00 = _fjsp_mul_v2r8(rinv00,rinv00);
2535 rinvsq11 = _fjsp_mul_v2r8(rinv11,rinv11);
2536 rinvsq12 = _fjsp_mul_v2r8(rinv12,rinv12);
2537 rinvsq13 = _fjsp_mul_v2r8(rinv13,rinv13);
2538 rinvsq21 = _fjsp_mul_v2r8(rinv21,rinv21);
2539 rinvsq22 = _fjsp_mul_v2r8(rinv22,rinv22);
2540 rinvsq23 = _fjsp_mul_v2r8(rinv23,rinv23);
2541 rinvsq31 = _fjsp_mul_v2r8(rinv31,rinv31);
2542 rinvsq32 = _fjsp_mul_v2r8(rinv32,rinv32);
2543 rinvsq33 = _fjsp_mul_v2r8(rinv33,rinv33);
2545 fjx0 = _fjsp_setzero_v2r8();
2546 fjy0 = _fjsp_setzero_v2r8();
2547 fjz0 = _fjsp_setzero_v2r8();
2548 fjx1 = _fjsp_setzero_v2r8();
2549 fjy1 = _fjsp_setzero_v2r8();
2550 fjz1 = _fjsp_setzero_v2r8();
2551 fjx2 = _fjsp_setzero_v2r8();
2552 fjy2 = _fjsp_setzero_v2r8();
2553 fjz2 = _fjsp_setzero_v2r8();
2554 fjx3 = _fjsp_setzero_v2r8();
2555 fjy3 = _fjsp_setzero_v2r8();
2556 fjz3 = _fjsp_setzero_v2r8();
2558 /**************************
2559 * CALCULATE INTERACTIONS *
2560 **************************/
2562 if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
2565 r00 = _fjsp_mul_v2r8(rsq00,rinv00);
2567 /* LENNARD-JONES DISPERSION/REPULSION */
2569 rinvsix = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
2570 vvdw6 = _fjsp_mul_v2r8(c6_00,rinvsix);
2571 vvdw12 = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
2572 vvdw = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
2573 fvdw = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
2575 d = _fjsp_sub_v2r8(r00,rswitch);
2576 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
2577 d2 = _fjsp_mul_v2r8(d,d);
2578 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
2580 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
2582 /* Evaluate switch function */
2583 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2584 fvdw = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
2585 cutoff_mask = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
2589 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
2591 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
2593 /* Update vectorial force */
2594 fix0 = _fjsp_madd_v2r8(dx00,fscal,fix0);
2595 fiy0 = _fjsp_madd_v2r8(dy00,fscal,fiy0);
2596 fiz0 = _fjsp_madd_v2r8(dz00,fscal,fiz0);
2598 fjx0 = _fjsp_madd_v2r8(dx00,fscal,fjx0);
2599 fjy0 = _fjsp_madd_v2r8(dy00,fscal,fjy0);
2600 fjz0 = _fjsp_madd_v2r8(dz00,fscal,fjz0);
2604 /**************************
2605 * CALCULATE INTERACTIONS *
2606 **************************/
2608 if (gmx_fjsp_any_lt_v2r8(rsq11,rcutoff2))
2611 r11 = _fjsp_mul_v2r8(rsq11,rinv11);
2613 /* EWALD ELECTROSTATICS */
2615 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2616 ewrt = _fjsp_mul_v2r8(r11,ewtabscale);
2617 itab_tmp = _fjsp_dtox_v2r8(ewrt);
2618 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
2619 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
2621 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
2622 ewtabD = _fjsp_setzero_v2r8();
2623 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
2624 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
2625 ewtabFn = _fjsp_setzero_v2r8();
2626 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
2627 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
2628 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
2629 velec = _fjsp_mul_v2r8(qq11,_fjsp_sub_v2r8(rinv11,velec));
2630 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq11,rinv11),_fjsp_sub_v2r8(rinvsq11,felec));
2632 d = _fjsp_sub_v2r8(r11,rswitch);
2633 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
2634 d2 = _fjsp_mul_v2r8(d,d);
2635 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
2637 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
2639 /* Evaluate switch function */
2640 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2641 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv11,_fjsp_mul_v2r8(velec,dsw)) );
2642 cutoff_mask = _fjsp_cmplt_v2r8(rsq11,rcutoff2);
2646 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
2648 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
2650 /* Update vectorial force */
2651 fix1 = _fjsp_madd_v2r8(dx11,fscal,fix1);
2652 fiy1 = _fjsp_madd_v2r8(dy11,fscal,fiy1);
2653 fiz1 = _fjsp_madd_v2r8(dz11,fscal,fiz1);
2655 fjx1 = _fjsp_madd_v2r8(dx11,fscal,fjx1);
2656 fjy1 = _fjsp_madd_v2r8(dy11,fscal,fjy1);
2657 fjz1 = _fjsp_madd_v2r8(dz11,fscal,fjz1);
2661 /**************************
2662 * CALCULATE INTERACTIONS *
2663 **************************/
2665 if (gmx_fjsp_any_lt_v2r8(rsq12,rcutoff2))
2668 r12 = _fjsp_mul_v2r8(rsq12,rinv12);
2670 /* EWALD ELECTROSTATICS */
2672 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2673 ewrt = _fjsp_mul_v2r8(r12,ewtabscale);
2674 itab_tmp = _fjsp_dtox_v2r8(ewrt);
2675 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
2676 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
2678 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
2679 ewtabD = _fjsp_setzero_v2r8();
2680 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
2681 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
2682 ewtabFn = _fjsp_setzero_v2r8();
2683 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
2684 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
2685 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
2686 velec = _fjsp_mul_v2r8(qq12,_fjsp_sub_v2r8(rinv12,velec));
2687 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq12,rinv12),_fjsp_sub_v2r8(rinvsq12,felec));
2689 d = _fjsp_sub_v2r8(r12,rswitch);
2690 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
2691 d2 = _fjsp_mul_v2r8(d,d);
2692 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
2694 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
2696 /* Evaluate switch function */
2697 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2698 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv12,_fjsp_mul_v2r8(velec,dsw)) );
2699 cutoff_mask = _fjsp_cmplt_v2r8(rsq12,rcutoff2);
2703 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
2705 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
2707 /* Update vectorial force */
2708 fix1 = _fjsp_madd_v2r8(dx12,fscal,fix1);
2709 fiy1 = _fjsp_madd_v2r8(dy12,fscal,fiy1);
2710 fiz1 = _fjsp_madd_v2r8(dz12,fscal,fiz1);
2712 fjx2 = _fjsp_madd_v2r8(dx12,fscal,fjx2);
2713 fjy2 = _fjsp_madd_v2r8(dy12,fscal,fjy2);
2714 fjz2 = _fjsp_madd_v2r8(dz12,fscal,fjz2);
2718 /**************************
2719 * CALCULATE INTERACTIONS *
2720 **************************/
2722 if (gmx_fjsp_any_lt_v2r8(rsq13,rcutoff2))
2725 r13 = _fjsp_mul_v2r8(rsq13,rinv13);
2727 /* EWALD ELECTROSTATICS */
2729 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2730 ewrt = _fjsp_mul_v2r8(r13,ewtabscale);
2731 itab_tmp = _fjsp_dtox_v2r8(ewrt);
2732 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
2733 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
2735 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
2736 ewtabD = _fjsp_setzero_v2r8();
2737 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
2738 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
2739 ewtabFn = _fjsp_setzero_v2r8();
2740 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
2741 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
2742 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
2743 velec = _fjsp_mul_v2r8(qq13,_fjsp_sub_v2r8(rinv13,velec));
2744 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq13,rinv13),_fjsp_sub_v2r8(rinvsq13,felec));
2746 d = _fjsp_sub_v2r8(r13,rswitch);
2747 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
2748 d2 = _fjsp_mul_v2r8(d,d);
2749 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
2751 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
2753 /* Evaluate switch function */
2754 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2755 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv13,_fjsp_mul_v2r8(velec,dsw)) );
2756 cutoff_mask = _fjsp_cmplt_v2r8(rsq13,rcutoff2);
2760 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
2762 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
2764 /* Update vectorial force */
2765 fix1 = _fjsp_madd_v2r8(dx13,fscal,fix1);
2766 fiy1 = _fjsp_madd_v2r8(dy13,fscal,fiy1);
2767 fiz1 = _fjsp_madd_v2r8(dz13,fscal,fiz1);
2769 fjx3 = _fjsp_madd_v2r8(dx13,fscal,fjx3);
2770 fjy3 = _fjsp_madd_v2r8(dy13,fscal,fjy3);
2771 fjz3 = _fjsp_madd_v2r8(dz13,fscal,fjz3);
2775 /**************************
2776 * CALCULATE INTERACTIONS *
2777 **************************/
2779 if (gmx_fjsp_any_lt_v2r8(rsq21,rcutoff2))
2782 r21 = _fjsp_mul_v2r8(rsq21,rinv21);
2784 /* EWALD ELECTROSTATICS */
2786 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2787 ewrt = _fjsp_mul_v2r8(r21,ewtabscale);
2788 itab_tmp = _fjsp_dtox_v2r8(ewrt);
2789 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
2790 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
2792 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
2793 ewtabD = _fjsp_setzero_v2r8();
2794 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
2795 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
2796 ewtabFn = _fjsp_setzero_v2r8();
2797 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
2798 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
2799 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
2800 velec = _fjsp_mul_v2r8(qq21,_fjsp_sub_v2r8(rinv21,velec));
2801 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq21,rinv21),_fjsp_sub_v2r8(rinvsq21,felec));
2803 d = _fjsp_sub_v2r8(r21,rswitch);
2804 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
2805 d2 = _fjsp_mul_v2r8(d,d);
2806 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
2808 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
2810 /* Evaluate switch function */
2811 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2812 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv21,_fjsp_mul_v2r8(velec,dsw)) );
2813 cutoff_mask = _fjsp_cmplt_v2r8(rsq21,rcutoff2);
2817 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
2819 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
2821 /* Update vectorial force */
2822 fix2 = _fjsp_madd_v2r8(dx21,fscal,fix2);
2823 fiy2 = _fjsp_madd_v2r8(dy21,fscal,fiy2);
2824 fiz2 = _fjsp_madd_v2r8(dz21,fscal,fiz2);
2826 fjx1 = _fjsp_madd_v2r8(dx21,fscal,fjx1);
2827 fjy1 = _fjsp_madd_v2r8(dy21,fscal,fjy1);
2828 fjz1 = _fjsp_madd_v2r8(dz21,fscal,fjz1);
2832 /**************************
2833 * CALCULATE INTERACTIONS *
2834 **************************/
2836 if (gmx_fjsp_any_lt_v2r8(rsq22,rcutoff2))
2839 r22 = _fjsp_mul_v2r8(rsq22,rinv22);
2841 /* EWALD ELECTROSTATICS */
2843 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2844 ewrt = _fjsp_mul_v2r8(r22,ewtabscale);
2845 itab_tmp = _fjsp_dtox_v2r8(ewrt);
2846 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
2847 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
2849 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
2850 ewtabD = _fjsp_setzero_v2r8();
2851 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
2852 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
2853 ewtabFn = _fjsp_setzero_v2r8();
2854 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
2855 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
2856 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
2857 velec = _fjsp_mul_v2r8(qq22,_fjsp_sub_v2r8(rinv22,velec));
2858 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq22,rinv22),_fjsp_sub_v2r8(rinvsq22,felec));
2860 d = _fjsp_sub_v2r8(r22,rswitch);
2861 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
2862 d2 = _fjsp_mul_v2r8(d,d);
2863 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
2865 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
2867 /* Evaluate switch function */
2868 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2869 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv22,_fjsp_mul_v2r8(velec,dsw)) );
2870 cutoff_mask = _fjsp_cmplt_v2r8(rsq22,rcutoff2);
2874 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
2876 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
2878 /* Update vectorial force */
2879 fix2 = _fjsp_madd_v2r8(dx22,fscal,fix2);
2880 fiy2 = _fjsp_madd_v2r8(dy22,fscal,fiy2);
2881 fiz2 = _fjsp_madd_v2r8(dz22,fscal,fiz2);
2883 fjx2 = _fjsp_madd_v2r8(dx22,fscal,fjx2);
2884 fjy2 = _fjsp_madd_v2r8(dy22,fscal,fjy2);
2885 fjz2 = _fjsp_madd_v2r8(dz22,fscal,fjz2);
2889 /**************************
2890 * CALCULATE INTERACTIONS *
2891 **************************/
2893 if (gmx_fjsp_any_lt_v2r8(rsq23,rcutoff2))
2896 r23 = _fjsp_mul_v2r8(rsq23,rinv23);
2898 /* EWALD ELECTROSTATICS */
2900 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2901 ewrt = _fjsp_mul_v2r8(r23,ewtabscale);
2902 itab_tmp = _fjsp_dtox_v2r8(ewrt);
2903 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
2904 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
2906 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
2907 ewtabD = _fjsp_setzero_v2r8();
2908 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
2909 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
2910 ewtabFn = _fjsp_setzero_v2r8();
2911 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
2912 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
2913 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
2914 velec = _fjsp_mul_v2r8(qq23,_fjsp_sub_v2r8(rinv23,velec));
2915 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq23,rinv23),_fjsp_sub_v2r8(rinvsq23,felec));
2917 d = _fjsp_sub_v2r8(r23,rswitch);
2918 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
2919 d2 = _fjsp_mul_v2r8(d,d);
2920 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
2922 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
2924 /* Evaluate switch function */
2925 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2926 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv23,_fjsp_mul_v2r8(velec,dsw)) );
2927 cutoff_mask = _fjsp_cmplt_v2r8(rsq23,rcutoff2);
2931 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
2933 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
2935 /* Update vectorial force */
2936 fix2 = _fjsp_madd_v2r8(dx23,fscal,fix2);
2937 fiy2 = _fjsp_madd_v2r8(dy23,fscal,fiy2);
2938 fiz2 = _fjsp_madd_v2r8(dz23,fscal,fiz2);
2940 fjx3 = _fjsp_madd_v2r8(dx23,fscal,fjx3);
2941 fjy3 = _fjsp_madd_v2r8(dy23,fscal,fjy3);
2942 fjz3 = _fjsp_madd_v2r8(dz23,fscal,fjz3);
2946 /**************************
2947 * CALCULATE INTERACTIONS *
2948 **************************/
2950 if (gmx_fjsp_any_lt_v2r8(rsq31,rcutoff2))
2953 r31 = _fjsp_mul_v2r8(rsq31,rinv31);
2955 /* EWALD ELECTROSTATICS */
2957 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2958 ewrt = _fjsp_mul_v2r8(r31,ewtabscale);
2959 itab_tmp = _fjsp_dtox_v2r8(ewrt);
2960 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
2961 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
2963 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
2964 ewtabD = _fjsp_setzero_v2r8();
2965 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
2966 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
2967 ewtabFn = _fjsp_setzero_v2r8();
2968 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
2969 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
2970 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
2971 velec = _fjsp_mul_v2r8(qq31,_fjsp_sub_v2r8(rinv31,velec));
2972 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq31,rinv31),_fjsp_sub_v2r8(rinvsq31,felec));
2974 d = _fjsp_sub_v2r8(r31,rswitch);
2975 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
2976 d2 = _fjsp_mul_v2r8(d,d);
2977 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
2979 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
2981 /* Evaluate switch function */
2982 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2983 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv31,_fjsp_mul_v2r8(velec,dsw)) );
2984 cutoff_mask = _fjsp_cmplt_v2r8(rsq31,rcutoff2);
2988 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
2990 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
2992 /* Update vectorial force */
2993 fix3 = _fjsp_madd_v2r8(dx31,fscal,fix3);
2994 fiy3 = _fjsp_madd_v2r8(dy31,fscal,fiy3);
2995 fiz3 = _fjsp_madd_v2r8(dz31,fscal,fiz3);
2997 fjx1 = _fjsp_madd_v2r8(dx31,fscal,fjx1);
2998 fjy1 = _fjsp_madd_v2r8(dy31,fscal,fjy1);
2999 fjz1 = _fjsp_madd_v2r8(dz31,fscal,fjz1);
3003 /**************************
3004 * CALCULATE INTERACTIONS *
3005 **************************/
3007 if (gmx_fjsp_any_lt_v2r8(rsq32,rcutoff2))
3010 r32 = _fjsp_mul_v2r8(rsq32,rinv32);
3012 /* EWALD ELECTROSTATICS */
3014 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
3015 ewrt = _fjsp_mul_v2r8(r32,ewtabscale);
3016 itab_tmp = _fjsp_dtox_v2r8(ewrt);
3017 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
3018 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
3020 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
3021 ewtabD = _fjsp_setzero_v2r8();
3022 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
3023 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
3024 ewtabFn = _fjsp_setzero_v2r8();
3025 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
3026 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
3027 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
3028 velec = _fjsp_mul_v2r8(qq32,_fjsp_sub_v2r8(rinv32,velec));
3029 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq32,rinv32),_fjsp_sub_v2r8(rinvsq32,felec));
3031 d = _fjsp_sub_v2r8(r32,rswitch);
3032 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
3033 d2 = _fjsp_mul_v2r8(d,d);
3034 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
3036 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
3038 /* Evaluate switch function */
3039 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
3040 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv32,_fjsp_mul_v2r8(velec,dsw)) );
3041 cutoff_mask = _fjsp_cmplt_v2r8(rsq32,rcutoff2);
3045 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
3047 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
3049 /* Update vectorial force */
3050 fix3 = _fjsp_madd_v2r8(dx32,fscal,fix3);
3051 fiy3 = _fjsp_madd_v2r8(dy32,fscal,fiy3);
3052 fiz3 = _fjsp_madd_v2r8(dz32,fscal,fiz3);
3054 fjx2 = _fjsp_madd_v2r8(dx32,fscal,fjx2);
3055 fjy2 = _fjsp_madd_v2r8(dy32,fscal,fjy2);
3056 fjz2 = _fjsp_madd_v2r8(dz32,fscal,fjz2);
3060 /**************************
3061 * CALCULATE INTERACTIONS *
3062 **************************/
3064 if (gmx_fjsp_any_lt_v2r8(rsq33,rcutoff2))
3067 r33 = _fjsp_mul_v2r8(rsq33,rinv33);
3069 /* EWALD ELECTROSTATICS */
3071 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
3072 ewrt = _fjsp_mul_v2r8(r33,ewtabscale);
3073 itab_tmp = _fjsp_dtox_v2r8(ewrt);
3074 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
3075 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
3077 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
3078 ewtabD = _fjsp_setzero_v2r8();
3079 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
3080 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
3081 ewtabFn = _fjsp_setzero_v2r8();
3082 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
3083 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
3084 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
3085 velec = _fjsp_mul_v2r8(qq33,_fjsp_sub_v2r8(rinv33,velec));
3086 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq33,rinv33),_fjsp_sub_v2r8(rinvsq33,felec));
3088 d = _fjsp_sub_v2r8(r33,rswitch);
3089 d = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
3090 d2 = _fjsp_mul_v2r8(d,d);
3091 sw = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
3093 dsw = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
3095 /* Evaluate switch function */
3096 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
3097 felec = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv33,_fjsp_mul_v2r8(velec,dsw)) );
3098 cutoff_mask = _fjsp_cmplt_v2r8(rsq33,rcutoff2);
3102 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
3104 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
3106 /* Update vectorial force */
3107 fix3 = _fjsp_madd_v2r8(dx33,fscal,fix3);
3108 fiy3 = _fjsp_madd_v2r8(dy33,fscal,fiy3);
3109 fiz3 = _fjsp_madd_v2r8(dz33,fscal,fiz3);
3111 fjx3 = _fjsp_madd_v2r8(dx33,fscal,fjx3);
3112 fjy3 = _fjsp_madd_v2r8(dy33,fscal,fjy3);
3113 fjz3 = _fjsp_madd_v2r8(dz33,fscal,fjz3);
3117 gmx_fjsp_decrement_4rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
3119 /* Inner loop uses 647 flops */
3122 /* End of innermost loop */
3124 gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
3125 f+i_coord_offset,fshift+i_shift_offset);
3127 /* Increment number of inner iterations */
3128 inneriter += j_index_end - j_index_start;
3130 /* Outer loop uses 24 flops */
3133 /* Increment number of outer iterations */
3136 /* Update outer/inner flops */
3138 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*647);