?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Fix segmentation fault in minimize
[gromacs.git] / src / gromacs / gmxlib / nonbonded / nb_kernel_sse2_double / nb_kernel_ElecRF_VdwCSTab_GeomW3P1_sse2_double.cpp
blob8402dc01325c15e04ac641bdb3fdfc9bbf53bb4f
1 /*
2 * This file is part of the GROMACS molecular simulation package.
3 *
4 * Copyright (c) 2012,2013,2014,2015,2017,2018, 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.
8 *
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.
13 *
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.
18 *
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.
23 *
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.
31 *
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.
34 */
35 /*
36 * Note: this file was generated by the GROMACS sse2_double kernel generator.
37 */
38 #include "gmxpre.h"
39
40 #include "config.h"
41
42 #include <math.h>
43
44 #include "../nb_kernel.h"
45 #include "gromacs/gmxlib/nrnb.h"
46
47 #include "kernelutil_x86_sse2_double.h"
48
49 /*
50 * Gromacs nonbonded kernel: nb_kernel_ElecRF_VdwCSTab_GeomW3P1_VF_sse2_double
51 * Electrostatics interaction: ReactionField
52 * VdW interaction: CubicSplineTable
53 * Geometry: Water3-Particle
54 * Calculate force/pot: PotentialAndForce
55 */
56 void
57 nb_kernel_ElecRF_VdwCSTab_GeomW3P1_VF_sse2_double
58 (t_nblist * gmx_restrict nlist,
59 rvec * gmx_restrict xx,
60 rvec * gmx_restrict ff,
61 struct t_forcerec * gmx_restrict fr,
62 t_mdatoms * gmx_restrict mdatoms,
63 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
64 t_nrnb * gmx_restrict nrnb)
65 {
66 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
67 * just 0 for non-waters.
68 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
69 * jnr indices corresponding to data put in the four positions in the SIMD register.
70 */
71 int i_shift_offset,i_coord_offset,outeriter,inneriter;
72 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
73 int jnrA,jnrB;
74 int j_coord_offsetA,j_coord_offsetB;
75 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
76 real rcutoff_scalar;
77 real *shiftvec,*fshift,*x,*f;
78 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
79 int vdwioffset0;
80 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
81 int vdwioffset1;
82 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
83 int vdwioffset2;
84 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
85 int vdwjidx0A,vdwjidx0B;
86 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
87 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
88 __m128d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
89 __m128d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
90 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
91 real *charge;
92 int nvdwtype;
93 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
94 int *vdwtype;
95 real *vdwparam;
96 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
97 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
98 __m128i vfitab;
99 __m128i ifour = _mm_set1_epi32(4);
100 __m128d rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
101 real *vftab;
102 __m128d dummy_mask,cutoff_mask;
103 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
104 __m128d one = _mm_set1_pd(1.0);
105 __m128d two = _mm_set1_pd(2.0);
106 x = xx[0];
107 f = ff[0];
108
109 nri = nlist->nri;
110 iinr = nlist->iinr;
111 jindex = nlist->jindex;
112 jjnr = nlist->jjnr;
113 shiftidx = nlist->shift;
114 gid = nlist->gid;
115 shiftvec = fr->shift_vec[0];
116 fshift = fr->fshift[0];
117 facel = _mm_set1_pd(fr->ic->epsfac);
118 charge = mdatoms->chargeA;
119 krf = _mm_set1_pd(fr->ic->k_rf);
120 krf2 = _mm_set1_pd(fr->ic->k_rf*2.0);
121 crf = _mm_set1_pd(fr->ic->c_rf);
122 nvdwtype = fr->ntype;
123 vdwparam = fr->nbfp;
124 vdwtype = mdatoms->typeA;
125
126 vftab = kernel_data->table_vdw->data;
127 vftabscale = _mm_set1_pd(kernel_data->table_vdw->scale);
128
129 /* Setup water-specific parameters */
130 inr = nlist->iinr[0];
131 iq0 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
132 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
133 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
134 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
135
136 /* Avoid stupid compiler warnings */
137 jnrA = jnrB = 0;
138 j_coord_offsetA = 0;
139 j_coord_offsetB = 0;
140
141 outeriter = 0;
142 inneriter = 0;
143
144 /* Start outer loop over neighborlists */
145 for(iidx=0; iidx<nri; iidx++)
146 {
147 /* Load shift vector for this list */
148 i_shift_offset = DIM*shiftidx[iidx];
149
150 /* Load limits for loop over neighbors */
151 j_index_start = jindex[iidx];
152 j_index_end = jindex[iidx+1];
153
154 /* Get outer coordinate index */
155 inr = iinr[iidx];
156 i_coord_offset = DIM*inr;
157
158 /* Load i particle coords and add shift vector */
159 gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
160 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
161
162 fix0 = _mm_setzero_pd();
163 fiy0 = _mm_setzero_pd();
164 fiz0 = _mm_setzero_pd();
165 fix1 = _mm_setzero_pd();
166 fiy1 = _mm_setzero_pd();
167 fiz1 = _mm_setzero_pd();
168 fix2 = _mm_setzero_pd();
169 fiy2 = _mm_setzero_pd();
170 fiz2 = _mm_setzero_pd();
171
172 /* Reset potential sums */
173 velecsum = _mm_setzero_pd();
174 vvdwsum = _mm_setzero_pd();
175
176 /* Start inner kernel loop */
177 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
178 {
179
180 /* Get j neighbor index, and coordinate index */
181 jnrA = jjnr[jidx];
182 jnrB = jjnr[jidx+1];
183 j_coord_offsetA = DIM*jnrA;
184 j_coord_offsetB = DIM*jnrB;
185
186 /* load j atom coordinates */
187 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
188 &jx0,&jy0,&jz0);
189
190 /* Calculate displacement vector */
191 dx00 = _mm_sub_pd(ix0,jx0);
192 dy00 = _mm_sub_pd(iy0,jy0);
193 dz00 = _mm_sub_pd(iz0,jz0);
194 dx10 = _mm_sub_pd(ix1,jx0);
195 dy10 = _mm_sub_pd(iy1,jy0);
196 dz10 = _mm_sub_pd(iz1,jz0);
197 dx20 = _mm_sub_pd(ix2,jx0);
198 dy20 = _mm_sub_pd(iy2,jy0);
199 dz20 = _mm_sub_pd(iz2,jz0);
200
201 /* Calculate squared distance and things based on it */
202 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
203 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
204 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
205
206 rinv00 = sse2_invsqrt_d(rsq00);
207 rinv10 = sse2_invsqrt_d(rsq10);
208 rinv20 = sse2_invsqrt_d(rsq20);
209
210 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
211 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
212 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
213
214 /* Load parameters for j particles */
215 jq0 = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
216 vdwjidx0A = 2*vdwtype[jnrA+0];
217 vdwjidx0B = 2*vdwtype[jnrB+0];
218
219 fjx0 = _mm_setzero_pd();
220 fjy0 = _mm_setzero_pd();
221 fjz0 = _mm_setzero_pd();
222
223 /**************************
224 * CALCULATE INTERACTIONS *
225 **************************/
226
227 r00 = _mm_mul_pd(rsq00,rinv00);
228
229 /* Compute parameters for interactions between i and j atoms */
230 qq00 = _mm_mul_pd(iq0,jq0);
231 gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
232 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
233
234 /* Calculate table index by multiplying r with table scale and truncate to integer */
235 rt = _mm_mul_pd(r00,vftabscale);
236 vfitab = _mm_cvttpd_epi32(rt);
237 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
238 vfitab = _mm_slli_epi32(vfitab,3);
239
240 /* REACTION-FIELD ELECTROSTATICS */
241 velec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_add_pd(rinv00,_mm_mul_pd(krf,rsq00)),crf));
242 felec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
243
244 /* CUBIC SPLINE TABLE DISPERSION */
245 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
246 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
247 GMX_MM_TRANSPOSE2_PD(Y,F);
248 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
249 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
250 GMX_MM_TRANSPOSE2_PD(G,H);
251 Heps = _mm_mul_pd(vfeps,H);
252 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
253 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
254 vvdw6 = _mm_mul_pd(c6_00,VV);
255 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
256 fvdw6 = _mm_mul_pd(c6_00,FF);
257
258 /* CUBIC SPLINE TABLE REPULSION */
259 vfitab = _mm_add_epi32(vfitab,ifour);
260 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
261 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
262 GMX_MM_TRANSPOSE2_PD(Y,F);
263 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
264 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
265 GMX_MM_TRANSPOSE2_PD(G,H);
266 Heps = _mm_mul_pd(vfeps,H);
267 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
268 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
269 vvdw12 = _mm_mul_pd(c12_00,VV);
270 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
271 fvdw12 = _mm_mul_pd(c12_00,FF);
272 vvdw = _mm_add_pd(vvdw12,vvdw6);
273 fvdw = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
274
275 /* Update potential sum for this i atom from the interaction with this j atom. */
276 velecsum = _mm_add_pd(velecsum,velec);
277 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
278
279 fscal = _mm_add_pd(felec,fvdw);
280
281 /* Calculate temporary vectorial force */
282 tx = _mm_mul_pd(fscal,dx00);
283 ty = _mm_mul_pd(fscal,dy00);
284 tz = _mm_mul_pd(fscal,dz00);
285
286 /* Update vectorial force */
287 fix0 = _mm_add_pd(fix0,tx);
288 fiy0 = _mm_add_pd(fiy0,ty);
289 fiz0 = _mm_add_pd(fiz0,tz);
290
291 fjx0 = _mm_add_pd(fjx0,tx);
292 fjy0 = _mm_add_pd(fjy0,ty);
293 fjz0 = _mm_add_pd(fjz0,tz);
294
295 /**************************
296 * CALCULATE INTERACTIONS *
297 **************************/
298
299 /* Compute parameters for interactions between i and j atoms */
300 qq10 = _mm_mul_pd(iq1,jq0);
301
302 /* REACTION-FIELD ELECTROSTATICS */
303 velec = _mm_mul_pd(qq10,_mm_sub_pd(_mm_add_pd(rinv10,_mm_mul_pd(krf,rsq10)),crf));
304 felec = _mm_mul_pd(qq10,_mm_sub_pd(_mm_mul_pd(rinv10,rinvsq10),krf2));
305
306 /* Update potential sum for this i atom from the interaction with this j atom. */
307 velecsum = _mm_add_pd(velecsum,velec);
308
309 fscal = felec;
310
311 /* Calculate temporary vectorial force */
312 tx = _mm_mul_pd(fscal,dx10);
313 ty = _mm_mul_pd(fscal,dy10);
314 tz = _mm_mul_pd(fscal,dz10);
315
316 /* Update vectorial force */
317 fix1 = _mm_add_pd(fix1,tx);
318 fiy1 = _mm_add_pd(fiy1,ty);
319 fiz1 = _mm_add_pd(fiz1,tz);
320
321 fjx0 = _mm_add_pd(fjx0,tx);
322 fjy0 = _mm_add_pd(fjy0,ty);
323 fjz0 = _mm_add_pd(fjz0,tz);
324
325 /**************************
326 * CALCULATE INTERACTIONS *
327 **************************/
328
329 /* Compute parameters for interactions between i and j atoms */
330 qq20 = _mm_mul_pd(iq2,jq0);
331
332 /* REACTION-FIELD ELECTROSTATICS */
333 velec = _mm_mul_pd(qq20,_mm_sub_pd(_mm_add_pd(rinv20,_mm_mul_pd(krf,rsq20)),crf));
334 felec = _mm_mul_pd(qq20,_mm_sub_pd(_mm_mul_pd(rinv20,rinvsq20),krf2));
335
336 /* Update potential sum for this i atom from the interaction with this j atom. */
337 velecsum = _mm_add_pd(velecsum,velec);
338
339 fscal = felec;
340
341 /* Calculate temporary vectorial force */
342 tx = _mm_mul_pd(fscal,dx20);
343 ty = _mm_mul_pd(fscal,dy20);
344 tz = _mm_mul_pd(fscal,dz20);
345
346 /* Update vectorial force */
347 fix2 = _mm_add_pd(fix2,tx);
348 fiy2 = _mm_add_pd(fiy2,ty);
349 fiz2 = _mm_add_pd(fiz2,tz);
350
351 fjx0 = _mm_add_pd(fjx0,tx);
352 fjy0 = _mm_add_pd(fjy0,ty);
353 fjz0 = _mm_add_pd(fjz0,tz);
354
355 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
356
357 /* Inner loop uses 134 flops */
358 }
359
360 if(jidx<j_index_end)
361 {
362
363 jnrA = jjnr[jidx];
364 j_coord_offsetA = DIM*jnrA;
365
366 /* load j atom coordinates */
367 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
368 &jx0,&jy0,&jz0);
369
370 /* Calculate displacement vector */
371 dx00 = _mm_sub_pd(ix0,jx0);
372 dy00 = _mm_sub_pd(iy0,jy0);
373 dz00 = _mm_sub_pd(iz0,jz0);
374 dx10 = _mm_sub_pd(ix1,jx0);
375 dy10 = _mm_sub_pd(iy1,jy0);
376 dz10 = _mm_sub_pd(iz1,jz0);
377 dx20 = _mm_sub_pd(ix2,jx0);
378 dy20 = _mm_sub_pd(iy2,jy0);
379 dz20 = _mm_sub_pd(iz2,jz0);
380
381 /* Calculate squared distance and things based on it */
382 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
383 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
384 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
385
386 rinv00 = sse2_invsqrt_d(rsq00);
387 rinv10 = sse2_invsqrt_d(rsq10);
388 rinv20 = sse2_invsqrt_d(rsq20);
389
390 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
391 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
392 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
393
394 /* Load parameters for j particles */
395 jq0 = _mm_load_sd(charge+jnrA+0);
396 vdwjidx0A = 2*vdwtype[jnrA+0];
397
398 fjx0 = _mm_setzero_pd();
399 fjy0 = _mm_setzero_pd();
400 fjz0 = _mm_setzero_pd();
401
402 /**************************
403 * CALCULATE INTERACTIONS *
404 **************************/
405
406 r00 = _mm_mul_pd(rsq00,rinv00);
407
408 /* Compute parameters for interactions between i and j atoms */
409 qq00 = _mm_mul_pd(iq0,jq0);
410 gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
411
412 /* Calculate table index by multiplying r with table scale and truncate to integer */
413 rt = _mm_mul_pd(r00,vftabscale);
414 vfitab = _mm_cvttpd_epi32(rt);
415 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
416 vfitab = _mm_slli_epi32(vfitab,3);
417
418 /* REACTION-FIELD ELECTROSTATICS */
419 velec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_add_pd(rinv00,_mm_mul_pd(krf,rsq00)),crf));
420 felec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
421
422 /* CUBIC SPLINE TABLE DISPERSION */
423 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
424 F = _mm_setzero_pd();
425 GMX_MM_TRANSPOSE2_PD(Y,F);
426 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
427 H = _mm_setzero_pd();
428 GMX_MM_TRANSPOSE2_PD(G,H);
429 Heps = _mm_mul_pd(vfeps,H);
430 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
431 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
432 vvdw6 = _mm_mul_pd(c6_00,VV);
433 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
434 fvdw6 = _mm_mul_pd(c6_00,FF);
435
436 /* CUBIC SPLINE TABLE REPULSION */
437 vfitab = _mm_add_epi32(vfitab,ifour);
438 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
439 F = _mm_setzero_pd();
440 GMX_MM_TRANSPOSE2_PD(Y,F);
441 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
442 H = _mm_setzero_pd();
443 GMX_MM_TRANSPOSE2_PD(G,H);
444 Heps = _mm_mul_pd(vfeps,H);
445 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
446 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
447 vvdw12 = _mm_mul_pd(c12_00,VV);
448 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
449 fvdw12 = _mm_mul_pd(c12_00,FF);
450 vvdw = _mm_add_pd(vvdw12,vvdw6);
451 fvdw = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
452
453 /* Update potential sum for this i atom from the interaction with this j atom. */
454 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
455 velecsum = _mm_add_pd(velecsum,velec);
456 vvdw = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
457 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
458
459 fscal = _mm_add_pd(felec,fvdw);
460
461 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
462
463 /* Calculate temporary vectorial force */
464 tx = _mm_mul_pd(fscal,dx00);
465 ty = _mm_mul_pd(fscal,dy00);
466 tz = _mm_mul_pd(fscal,dz00);
467
468 /* Update vectorial force */
469 fix0 = _mm_add_pd(fix0,tx);
470 fiy0 = _mm_add_pd(fiy0,ty);
471 fiz0 = _mm_add_pd(fiz0,tz);
472
473 fjx0 = _mm_add_pd(fjx0,tx);
474 fjy0 = _mm_add_pd(fjy0,ty);
475 fjz0 = _mm_add_pd(fjz0,tz);
476
477 /**************************
478 * CALCULATE INTERACTIONS *
479 **************************/
480
481 /* Compute parameters for interactions between i and j atoms */
482 qq10 = _mm_mul_pd(iq1,jq0);
483
484 /* REACTION-FIELD ELECTROSTATICS */
485 velec = _mm_mul_pd(qq10,_mm_sub_pd(_mm_add_pd(rinv10,_mm_mul_pd(krf,rsq10)),crf));
486 felec = _mm_mul_pd(qq10,_mm_sub_pd(_mm_mul_pd(rinv10,rinvsq10),krf2));
487
488 /* Update potential sum for this i atom from the interaction with this j atom. */
489 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
490 velecsum = _mm_add_pd(velecsum,velec);
491
492 fscal = felec;
493
494 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
495
496 /* Calculate temporary vectorial force */
497 tx = _mm_mul_pd(fscal,dx10);
498 ty = _mm_mul_pd(fscal,dy10);
499 tz = _mm_mul_pd(fscal,dz10);
500
501 /* Update vectorial force */
502 fix1 = _mm_add_pd(fix1,tx);
503 fiy1 = _mm_add_pd(fiy1,ty);
504 fiz1 = _mm_add_pd(fiz1,tz);
505
506 fjx0 = _mm_add_pd(fjx0,tx);
507 fjy0 = _mm_add_pd(fjy0,ty);
508 fjz0 = _mm_add_pd(fjz0,tz);
509
510 /**************************
511 * CALCULATE INTERACTIONS *
512 **************************/
513
514 /* Compute parameters for interactions between i and j atoms */
515 qq20 = _mm_mul_pd(iq2,jq0);
516
517 /* REACTION-FIELD ELECTROSTATICS */
518 velec = _mm_mul_pd(qq20,_mm_sub_pd(_mm_add_pd(rinv20,_mm_mul_pd(krf,rsq20)),crf));
519 felec = _mm_mul_pd(qq20,_mm_sub_pd(_mm_mul_pd(rinv20,rinvsq20),krf2));
520
521 /* Update potential sum for this i atom from the interaction with this j atom. */
522 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
523 velecsum = _mm_add_pd(velecsum,velec);
524
525 fscal = felec;
526
527 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
528
529 /* Calculate temporary vectorial force */
530 tx = _mm_mul_pd(fscal,dx20);
531 ty = _mm_mul_pd(fscal,dy20);
532 tz = _mm_mul_pd(fscal,dz20);
533
534 /* Update vectorial force */
535 fix2 = _mm_add_pd(fix2,tx);
536 fiy2 = _mm_add_pd(fiy2,ty);
537 fiz2 = _mm_add_pd(fiz2,tz);
538
539 fjx0 = _mm_add_pd(fjx0,tx);
540 fjy0 = _mm_add_pd(fjy0,ty);
541 fjz0 = _mm_add_pd(fjz0,tz);
542
543 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0);
544
545 /* Inner loop uses 134 flops */
546 }
547
548 /* End of innermost loop */
549
550 gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
551 f+i_coord_offset,fshift+i_shift_offset);
552
553 ggid = gid[iidx];
554 /* Update potential energies */
555 gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
556 gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
557
558 /* Increment number of inner iterations */
559 inneriter += j_index_end - j_index_start;
560
561 /* Outer loop uses 20 flops */
562 }
563
564 /* Increment number of outer iterations */
565 outeriter += nri;
566
567 /* Update outer/inner flops */
568
569 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*134);
570 }
571 /*
572 * Gromacs nonbonded kernel: nb_kernel_ElecRF_VdwCSTab_GeomW3P1_F_sse2_double
573 * Electrostatics interaction: ReactionField
574 * VdW interaction: CubicSplineTable
575 * Geometry: Water3-Particle
576 * Calculate force/pot: Force
577 */
578 void
579 nb_kernel_ElecRF_VdwCSTab_GeomW3P1_F_sse2_double
580 (t_nblist * gmx_restrict nlist,
581 rvec * gmx_restrict xx,
582 rvec * gmx_restrict ff,
583 struct t_forcerec * gmx_restrict fr,
584 t_mdatoms * gmx_restrict mdatoms,
585 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
586 t_nrnb * gmx_restrict nrnb)
587 {
588 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
589 * just 0 for non-waters.
590 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
591 * jnr indices corresponding to data put in the four positions in the SIMD register.
592 */
593 int i_shift_offset,i_coord_offset,outeriter,inneriter;
594 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
595 int jnrA,jnrB;
596 int j_coord_offsetA,j_coord_offsetB;
597 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
598 real rcutoff_scalar;
599 real *shiftvec,*fshift,*x,*f;
600 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
601 int vdwioffset0;
602 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
603 int vdwioffset1;
604 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
605 int vdwioffset2;
606 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
607 int vdwjidx0A,vdwjidx0B;
608 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
609 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
610 __m128d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
611 __m128d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
612 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
613 real *charge;
614 int nvdwtype;
615 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
616 int *vdwtype;
617 real *vdwparam;
618 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
619 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
620 __m128i vfitab;
621 __m128i ifour = _mm_set1_epi32(4);
622 __m128d rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
623 real *vftab;
624 __m128d dummy_mask,cutoff_mask;
625 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
626 __m128d one = _mm_set1_pd(1.0);
627 __m128d two = _mm_set1_pd(2.0);
628 x = xx[0];
629 f = ff[0];
630
631 nri = nlist->nri;
632 iinr = nlist->iinr;
633 jindex = nlist->jindex;
634 jjnr = nlist->jjnr;
635 shiftidx = nlist->shift;
636 gid = nlist->gid;
637 shiftvec = fr->shift_vec[0];
638 fshift = fr->fshift[0];
639 facel = _mm_set1_pd(fr->ic->epsfac);
640 charge = mdatoms->chargeA;
641 krf = _mm_set1_pd(fr->ic->k_rf);
642 krf2 = _mm_set1_pd(fr->ic->k_rf*2.0);
643 crf = _mm_set1_pd(fr->ic->c_rf);
644 nvdwtype = fr->ntype;
645 vdwparam = fr->nbfp;
646 vdwtype = mdatoms->typeA;
647
648 vftab = kernel_data->table_vdw->data;
649 vftabscale = _mm_set1_pd(kernel_data->table_vdw->scale);
650
651 /* Setup water-specific parameters */
652 inr = nlist->iinr[0];
653 iq0 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
654 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
655 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
656 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
657
658 /* Avoid stupid compiler warnings */
659 jnrA = jnrB = 0;
660 j_coord_offsetA = 0;
661 j_coord_offsetB = 0;
662
663 outeriter = 0;
664 inneriter = 0;
665
666 /* Start outer loop over neighborlists */
667 for(iidx=0; iidx<nri; iidx++)
668 {
669 /* Load shift vector for this list */
670 i_shift_offset = DIM*shiftidx[iidx];
671
672 /* Load limits for loop over neighbors */
673 j_index_start = jindex[iidx];
674 j_index_end = jindex[iidx+1];
675
676 /* Get outer coordinate index */
677 inr = iinr[iidx];
678 i_coord_offset = DIM*inr;
679
680 /* Load i particle coords and add shift vector */
681 gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
682 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
683
684 fix0 = _mm_setzero_pd();
685 fiy0 = _mm_setzero_pd();
686 fiz0 = _mm_setzero_pd();
687 fix1 = _mm_setzero_pd();
688 fiy1 = _mm_setzero_pd();
689 fiz1 = _mm_setzero_pd();
690 fix2 = _mm_setzero_pd();
691 fiy2 = _mm_setzero_pd();
692 fiz2 = _mm_setzero_pd();
693
694 /* Start inner kernel loop */
695 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
696 {
697
698 /* Get j neighbor index, and coordinate index */
699 jnrA = jjnr[jidx];
700 jnrB = jjnr[jidx+1];
701 j_coord_offsetA = DIM*jnrA;
702 j_coord_offsetB = DIM*jnrB;
703
704 /* load j atom coordinates */
705 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
706 &jx0,&jy0,&jz0);
707
708 /* Calculate displacement vector */
709 dx00 = _mm_sub_pd(ix0,jx0);
710 dy00 = _mm_sub_pd(iy0,jy0);
711 dz00 = _mm_sub_pd(iz0,jz0);
712 dx10 = _mm_sub_pd(ix1,jx0);
713 dy10 = _mm_sub_pd(iy1,jy0);
714 dz10 = _mm_sub_pd(iz1,jz0);
715 dx20 = _mm_sub_pd(ix2,jx0);
716 dy20 = _mm_sub_pd(iy2,jy0);
717 dz20 = _mm_sub_pd(iz2,jz0);
718
719 /* Calculate squared distance and things based on it */
720 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
721 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
722 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
723
724 rinv00 = sse2_invsqrt_d(rsq00);
725 rinv10 = sse2_invsqrt_d(rsq10);
726 rinv20 = sse2_invsqrt_d(rsq20);
727
728 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
729 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
730 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
731
732 /* Load parameters for j particles */
733 jq0 = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
734 vdwjidx0A = 2*vdwtype[jnrA+0];
735 vdwjidx0B = 2*vdwtype[jnrB+0];
736
737 fjx0 = _mm_setzero_pd();
738 fjy0 = _mm_setzero_pd();
739 fjz0 = _mm_setzero_pd();
740
741 /**************************
742 * CALCULATE INTERACTIONS *
743 **************************/
744
745 r00 = _mm_mul_pd(rsq00,rinv00);
746
747 /* Compute parameters for interactions between i and j atoms */
748 qq00 = _mm_mul_pd(iq0,jq0);
749 gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
750 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
751
752 /* Calculate table index by multiplying r with table scale and truncate to integer */
753 rt = _mm_mul_pd(r00,vftabscale);
754 vfitab = _mm_cvttpd_epi32(rt);
755 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
756 vfitab = _mm_slli_epi32(vfitab,3);
757
758 /* REACTION-FIELD ELECTROSTATICS */
759 felec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
760
761 /* CUBIC SPLINE TABLE DISPERSION */
762 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
763 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
764 GMX_MM_TRANSPOSE2_PD(Y,F);
765 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
766 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
767 GMX_MM_TRANSPOSE2_PD(G,H);
768 Heps = _mm_mul_pd(vfeps,H);
769 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
770 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
771 fvdw6 = _mm_mul_pd(c6_00,FF);
772
773 /* CUBIC SPLINE TABLE REPULSION */
774 vfitab = _mm_add_epi32(vfitab,ifour);
775 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
776 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
777 GMX_MM_TRANSPOSE2_PD(Y,F);
778 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
779 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
780 GMX_MM_TRANSPOSE2_PD(G,H);
781 Heps = _mm_mul_pd(vfeps,H);
782 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
783 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
784 fvdw12 = _mm_mul_pd(c12_00,FF);
785 fvdw = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
786
787 fscal = _mm_add_pd(felec,fvdw);
788
789 /* Calculate temporary vectorial force */
790 tx = _mm_mul_pd(fscal,dx00);
791 ty = _mm_mul_pd(fscal,dy00);
792 tz = _mm_mul_pd(fscal,dz00);
793
794 /* Update vectorial force */
795 fix0 = _mm_add_pd(fix0,tx);
796 fiy0 = _mm_add_pd(fiy0,ty);
797 fiz0 = _mm_add_pd(fiz0,tz);
798
799 fjx0 = _mm_add_pd(fjx0,tx);
800 fjy0 = _mm_add_pd(fjy0,ty);
801 fjz0 = _mm_add_pd(fjz0,tz);
802
803 /**************************
804 * CALCULATE INTERACTIONS *
805 **************************/
806
807 /* Compute parameters for interactions between i and j atoms */
808 qq10 = _mm_mul_pd(iq1,jq0);
809
810 /* REACTION-FIELD ELECTROSTATICS */
811 felec = _mm_mul_pd(qq10,_mm_sub_pd(_mm_mul_pd(rinv10,rinvsq10),krf2));
812
813 fscal = felec;
814
815 /* Calculate temporary vectorial force */
816 tx = _mm_mul_pd(fscal,dx10);
817 ty = _mm_mul_pd(fscal,dy10);
818 tz = _mm_mul_pd(fscal,dz10);
819
820 /* Update vectorial force */
821 fix1 = _mm_add_pd(fix1,tx);
822 fiy1 = _mm_add_pd(fiy1,ty);
823 fiz1 = _mm_add_pd(fiz1,tz);
824
825 fjx0 = _mm_add_pd(fjx0,tx);
826 fjy0 = _mm_add_pd(fjy0,ty);
827 fjz0 = _mm_add_pd(fjz0,tz);
828
829 /**************************
830 * CALCULATE INTERACTIONS *
831 **************************/
832
833 /* Compute parameters for interactions between i and j atoms */
834 qq20 = _mm_mul_pd(iq2,jq0);
835
836 /* REACTION-FIELD ELECTROSTATICS */
837 felec = _mm_mul_pd(qq20,_mm_sub_pd(_mm_mul_pd(rinv20,rinvsq20),krf2));
838
839 fscal = felec;
840
841 /* Calculate temporary vectorial force */
842 tx = _mm_mul_pd(fscal,dx20);
843 ty = _mm_mul_pd(fscal,dy20);
844 tz = _mm_mul_pd(fscal,dz20);
845
846 /* Update vectorial force */
847 fix2 = _mm_add_pd(fix2,tx);
848 fiy2 = _mm_add_pd(fiy2,ty);
849 fiz2 = _mm_add_pd(fiz2,tz);
850
851 fjx0 = _mm_add_pd(fjx0,tx);
852 fjy0 = _mm_add_pd(fjy0,ty);
853 fjz0 = _mm_add_pd(fjz0,tz);
854
855 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
856
857 /* Inner loop uses 111 flops */
858 }
859
860 if(jidx<j_index_end)
861 {
862
863 jnrA = jjnr[jidx];
864 j_coord_offsetA = DIM*jnrA;
865
866 /* load j atom coordinates */
867 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
868 &jx0,&jy0,&jz0);
869
870 /* Calculate displacement vector */
871 dx00 = _mm_sub_pd(ix0,jx0);
872 dy00 = _mm_sub_pd(iy0,jy0);
873 dz00 = _mm_sub_pd(iz0,jz0);
874 dx10 = _mm_sub_pd(ix1,jx0);
875 dy10 = _mm_sub_pd(iy1,jy0);
876 dz10 = _mm_sub_pd(iz1,jz0);
877 dx20 = _mm_sub_pd(ix2,jx0);
878 dy20 = _mm_sub_pd(iy2,jy0);
879 dz20 = _mm_sub_pd(iz2,jz0);
880
881 /* Calculate squared distance and things based on it */
882 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
883 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
884 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
885
886 rinv00 = sse2_invsqrt_d(rsq00);
887 rinv10 = sse2_invsqrt_d(rsq10);
888 rinv20 = sse2_invsqrt_d(rsq20);
889
890 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
891 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
892 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
893
894 /* Load parameters for j particles */
895 jq0 = _mm_load_sd(charge+jnrA+0);
896 vdwjidx0A = 2*vdwtype[jnrA+0];
897
898 fjx0 = _mm_setzero_pd();
899 fjy0 = _mm_setzero_pd();
900 fjz0 = _mm_setzero_pd();
901
902 /**************************
903 * CALCULATE INTERACTIONS *
904 **************************/
905
906 r00 = _mm_mul_pd(rsq00,rinv00);
907
908 /* Compute parameters for interactions between i and j atoms */
909 qq00 = _mm_mul_pd(iq0,jq0);
910 gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
911
912 /* Calculate table index by multiplying r with table scale and truncate to integer */
913 rt = _mm_mul_pd(r00,vftabscale);
914 vfitab = _mm_cvttpd_epi32(rt);
915 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
916 vfitab = _mm_slli_epi32(vfitab,3);
917
918 /* REACTION-FIELD ELECTROSTATICS */
919 felec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
920
921 /* CUBIC SPLINE TABLE DISPERSION */
922 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
923 F = _mm_setzero_pd();
924 GMX_MM_TRANSPOSE2_PD(Y,F);
925 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
926 H = _mm_setzero_pd();
927 GMX_MM_TRANSPOSE2_PD(G,H);
928 Heps = _mm_mul_pd(vfeps,H);
929 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
930 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
931 fvdw6 = _mm_mul_pd(c6_00,FF);
932
933 /* CUBIC SPLINE TABLE REPULSION */
934 vfitab = _mm_add_epi32(vfitab,ifour);
935 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
936 F = _mm_setzero_pd();
937 GMX_MM_TRANSPOSE2_PD(Y,F);
938 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
939 H = _mm_setzero_pd();
940 GMX_MM_TRANSPOSE2_PD(G,H);
941 Heps = _mm_mul_pd(vfeps,H);
942 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
943 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
944 fvdw12 = _mm_mul_pd(c12_00,FF);
945 fvdw = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
946
947 fscal = _mm_add_pd(felec,fvdw);
948
949 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
950
951 /* Calculate temporary vectorial force */
952 tx = _mm_mul_pd(fscal,dx00);
953 ty = _mm_mul_pd(fscal,dy00);
954 tz = _mm_mul_pd(fscal,dz00);
955
956 /* Update vectorial force */
957 fix0 = _mm_add_pd(fix0,tx);
958 fiy0 = _mm_add_pd(fiy0,ty);
959 fiz0 = _mm_add_pd(fiz0,tz);
960
961 fjx0 = _mm_add_pd(fjx0,tx);
962 fjy0 = _mm_add_pd(fjy0,ty);
963 fjz0 = _mm_add_pd(fjz0,tz);
964
965 /**************************
966 * CALCULATE INTERACTIONS *
967 **************************/
968
969 /* Compute parameters for interactions between i and j atoms */
970 qq10 = _mm_mul_pd(iq1,jq0);
971
972 /* REACTION-FIELD ELECTROSTATICS */
973 felec = _mm_mul_pd(qq10,_mm_sub_pd(_mm_mul_pd(rinv10,rinvsq10),krf2));
974
975 fscal = felec;
976
977 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
978
979 /* Calculate temporary vectorial force */
980 tx = _mm_mul_pd(fscal,dx10);
981 ty = _mm_mul_pd(fscal,dy10);
982 tz = _mm_mul_pd(fscal,dz10);
983
984 /* Update vectorial force */
985 fix1 = _mm_add_pd(fix1,tx);
986 fiy1 = _mm_add_pd(fiy1,ty);
987 fiz1 = _mm_add_pd(fiz1,tz);
988
989 fjx0 = _mm_add_pd(fjx0,tx);
990 fjy0 = _mm_add_pd(fjy0,ty);
991 fjz0 = _mm_add_pd(fjz0,tz);
992
993 /**************************
994 * CALCULATE INTERACTIONS *
995 **************************/
996
997 /* Compute parameters for interactions between i and j atoms */
998 qq20 = _mm_mul_pd(iq2,jq0);
999
1000 /* REACTION-FIELD ELECTROSTATICS */
1001 felec = _mm_mul_pd(qq20,_mm_sub_pd(_mm_mul_pd(rinv20,rinvsq20),krf2));
1002
1003 fscal = felec;
1004
1005 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1006
1007 /* Calculate temporary vectorial force */
1008 tx = _mm_mul_pd(fscal,dx20);
1009 ty = _mm_mul_pd(fscal,dy20);
1010 tz = _mm_mul_pd(fscal,dz20);
1011
1012 /* Update vectorial force */
1013 fix2 = _mm_add_pd(fix2,tx);
1014 fiy2 = _mm_add_pd(fiy2,ty);
1015 fiz2 = _mm_add_pd(fiz2,tz);
1016
1017 fjx0 = _mm_add_pd(fjx0,tx);
1018 fjy0 = _mm_add_pd(fjy0,ty);
1019 fjz0 = _mm_add_pd(fjz0,tz);
1020
1021 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1022
1023 /* Inner loop uses 111 flops */
1024 }
1025
1026 /* End of innermost loop */
1027
1028 gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1029 f+i_coord_offset,fshift+i_shift_offset);
1030
1031 /* Increment number of inner iterations */
1032 inneriter += j_index_end - j_index_start;
1033
1034 /* Outer loop uses 18 flops */
1035 }
1036
1037 /* Increment number of outer iterations */
1038 outeriter += nri;
1039
1040 /* Update outer/inner flops */
1041
1042 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_F,outeriter*18 + inneriter*111);
1043 }
The ultimate molecular dynamics simulation package