?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Removed simple.h from nb_kernel_sse2_XX
[gromacs.git] / src / gromacs / gmxlib / nonbonded / nb_kernel_sse2_double / nb_kernel_ElecRF_VdwNone_GeomW3P1_sse2_double.c
blobd3be4325b0297b71f044a75fd51335781242abe8
1 /*
2 * This file is part of the GROMACS molecular simulation package.
3 *
4 * Copyright (c) 2012,2013,2014,2015, 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/math/vec.h"
46 #include "gromacs/legacyheaders/nrnb.h"
47
48 #include "gromacs/simd/math_x86_sse2_double.h"
49 #include "kernelutil_x86_sse2_double.h"
50
51 /*
52 * Gromacs nonbonded kernel: nb_kernel_ElecRF_VdwNone_GeomW3P1_VF_sse2_double
53 * Electrostatics interaction: ReactionField
54 * VdW interaction: None
55 * Geometry: Water3-Particle
56 * Calculate force/pot: PotentialAndForce
57 */
58 void
59 nb_kernel_ElecRF_VdwNone_GeomW3P1_VF_sse2_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)
67 {
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 SSE double precision, e.g. for the two different
71 * jnr indices corresponding to data put in the four positions in the SIMD register.
72 */
73 int i_shift_offset,i_coord_offset,outeriter,inneriter;
74 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
75 int jnrA,jnrB;
76 int j_coord_offsetA,j_coord_offsetB;
77 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
78 real rcutoff_scalar;
79 real *shiftvec,*fshift,*x,*f;
80 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
81 int vdwioffset0;
82 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
83 int vdwioffset1;
84 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
85 int vdwioffset2;
86 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
87 int vdwjidx0A,vdwjidx0B;
88 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
89 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
90 __m128d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
91 __m128d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
92 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
93 real *charge;
94 __m128d dummy_mask,cutoff_mask;
95 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
96 __m128d one = _mm_set1_pd(1.0);
97 __m128d two = _mm_set1_pd(2.0);
98 x = xx[0];
99 f = ff[0];
100
101 nri = nlist->nri;
102 iinr = nlist->iinr;
103 jindex = nlist->jindex;
104 jjnr = nlist->jjnr;
105 shiftidx = nlist->shift;
106 gid = nlist->gid;
107 shiftvec = fr->shift_vec[0];
108 fshift = fr->fshift[0];
109 facel = _mm_set1_pd(fr->epsfac);
110 charge = mdatoms->chargeA;
111 krf = _mm_set1_pd(fr->ic->k_rf);
112 krf2 = _mm_set1_pd(fr->ic->k_rf*2.0);
113 crf = _mm_set1_pd(fr->ic->c_rf);
114
115 /* Setup water-specific parameters */
116 inr = nlist->iinr[0];
117 iq0 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
118 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
119 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
120
121 /* Avoid stupid compiler warnings */
122 jnrA = jnrB = 0;
123 j_coord_offsetA = 0;
124 j_coord_offsetB = 0;
125
126 outeriter = 0;
127 inneriter = 0;
128
129 /* Start outer loop over neighborlists */
130 for(iidx=0; iidx<nri; iidx++)
131 {
132 /* Load shift vector for this list */
133 i_shift_offset = DIM*shiftidx[iidx];
134
135 /* Load limits for loop over neighbors */
136 j_index_start = jindex[iidx];
137 j_index_end = jindex[iidx+1];
138
139 /* Get outer coordinate index */
140 inr = iinr[iidx];
141 i_coord_offset = DIM*inr;
142
143 /* Load i particle coords and add shift vector */
144 gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
145 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
146
147 fix0 = _mm_setzero_pd();
148 fiy0 = _mm_setzero_pd();
149 fiz0 = _mm_setzero_pd();
150 fix1 = _mm_setzero_pd();
151 fiy1 = _mm_setzero_pd();
152 fiz1 = _mm_setzero_pd();
153 fix2 = _mm_setzero_pd();
154 fiy2 = _mm_setzero_pd();
155 fiz2 = _mm_setzero_pd();
156
157 /* Reset potential sums */
158 velecsum = _mm_setzero_pd();
159
160 /* Start inner kernel loop */
161 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
162 {
163
164 /* Get j neighbor index, and coordinate index */
165 jnrA = jjnr[jidx];
166 jnrB = jjnr[jidx+1];
167 j_coord_offsetA = DIM*jnrA;
168 j_coord_offsetB = DIM*jnrB;
169
170 /* load j atom coordinates */
171 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
172 &jx0,&jy0,&jz0);
173
174 /* Calculate displacement vector */
175 dx00 = _mm_sub_pd(ix0,jx0);
176 dy00 = _mm_sub_pd(iy0,jy0);
177 dz00 = _mm_sub_pd(iz0,jz0);
178 dx10 = _mm_sub_pd(ix1,jx0);
179 dy10 = _mm_sub_pd(iy1,jy0);
180 dz10 = _mm_sub_pd(iz1,jz0);
181 dx20 = _mm_sub_pd(ix2,jx0);
182 dy20 = _mm_sub_pd(iy2,jy0);
183 dz20 = _mm_sub_pd(iz2,jz0);
184
185 /* Calculate squared distance and things based on it */
186 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
187 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
188 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
189
190 rinv00 = gmx_mm_invsqrt_pd(rsq00);
191 rinv10 = gmx_mm_invsqrt_pd(rsq10);
192 rinv20 = gmx_mm_invsqrt_pd(rsq20);
193
194 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
195 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
196 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
197
198 /* Load parameters for j particles */
199 jq0 = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
200
201 fjx0 = _mm_setzero_pd();
202 fjy0 = _mm_setzero_pd();
203 fjz0 = _mm_setzero_pd();
204
205 /**************************
206 * CALCULATE INTERACTIONS *
207 **************************/
208
209 /* Compute parameters for interactions between i and j atoms */
210 qq00 = _mm_mul_pd(iq0,jq0);
211
212 /* REACTION-FIELD ELECTROSTATICS */
213 velec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_add_pd(rinv00,_mm_mul_pd(krf,rsq00)),crf));
214 felec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
215
216 /* Update potential sum for this i atom from the interaction with this j atom. */
217 velecsum = _mm_add_pd(velecsum,velec);
218
219 fscal = felec;
220
221 /* Calculate temporary vectorial force */
222 tx = _mm_mul_pd(fscal,dx00);
223 ty = _mm_mul_pd(fscal,dy00);
224 tz = _mm_mul_pd(fscal,dz00);
225
226 /* Update vectorial force */
227 fix0 = _mm_add_pd(fix0,tx);
228 fiy0 = _mm_add_pd(fiy0,ty);
229 fiz0 = _mm_add_pd(fiz0,tz);
230
231 fjx0 = _mm_add_pd(fjx0,tx);
232 fjy0 = _mm_add_pd(fjy0,ty);
233 fjz0 = _mm_add_pd(fjz0,tz);
234
235 /**************************
236 * CALCULATE INTERACTIONS *
237 **************************/
238
239 /* Compute parameters for interactions between i and j atoms */
240 qq10 = _mm_mul_pd(iq1,jq0);
241
242 /* REACTION-FIELD ELECTROSTATICS */
243 velec = _mm_mul_pd(qq10,_mm_sub_pd(_mm_add_pd(rinv10,_mm_mul_pd(krf,rsq10)),crf));
244 felec = _mm_mul_pd(qq10,_mm_sub_pd(_mm_mul_pd(rinv10,rinvsq10),krf2));
245
246 /* Update potential sum for this i atom from the interaction with this j atom. */
247 velecsum = _mm_add_pd(velecsum,velec);
248
249 fscal = felec;
250
251 /* Calculate temporary vectorial force */
252 tx = _mm_mul_pd(fscal,dx10);
253 ty = _mm_mul_pd(fscal,dy10);
254 tz = _mm_mul_pd(fscal,dz10);
255
256 /* Update vectorial force */
257 fix1 = _mm_add_pd(fix1,tx);
258 fiy1 = _mm_add_pd(fiy1,ty);
259 fiz1 = _mm_add_pd(fiz1,tz);
260
261 fjx0 = _mm_add_pd(fjx0,tx);
262 fjy0 = _mm_add_pd(fjy0,ty);
263 fjz0 = _mm_add_pd(fjz0,tz);
264
265 /**************************
266 * CALCULATE INTERACTIONS *
267 **************************/
268
269 /* Compute parameters for interactions between i and j atoms */
270 qq20 = _mm_mul_pd(iq2,jq0);
271
272 /* REACTION-FIELD ELECTROSTATICS */
273 velec = _mm_mul_pd(qq20,_mm_sub_pd(_mm_add_pd(rinv20,_mm_mul_pd(krf,rsq20)),crf));
274 felec = _mm_mul_pd(qq20,_mm_sub_pd(_mm_mul_pd(rinv20,rinvsq20),krf2));
275
276 /* Update potential sum for this i atom from the interaction with this j atom. */
277 velecsum = _mm_add_pd(velecsum,velec);
278
279 fscal = felec;
280
281 /* Calculate temporary vectorial force */
282 tx = _mm_mul_pd(fscal,dx20);
283 ty = _mm_mul_pd(fscal,dy20);
284 tz = _mm_mul_pd(fscal,dz20);
285
286 /* Update vectorial force */
287 fix2 = _mm_add_pd(fix2,tx);
288 fiy2 = _mm_add_pd(fiy2,ty);
289 fiz2 = _mm_add_pd(fiz2,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 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
296
297 /* Inner loop uses 99 flops */
298 }
299
300 if(jidx<j_index_end)
301 {
302
303 jnrA = jjnr[jidx];
304 j_coord_offsetA = DIM*jnrA;
305
306 /* load j atom coordinates */
307 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
308 &jx0,&jy0,&jz0);
309
310 /* Calculate displacement vector */
311 dx00 = _mm_sub_pd(ix0,jx0);
312 dy00 = _mm_sub_pd(iy0,jy0);
313 dz00 = _mm_sub_pd(iz0,jz0);
314 dx10 = _mm_sub_pd(ix1,jx0);
315 dy10 = _mm_sub_pd(iy1,jy0);
316 dz10 = _mm_sub_pd(iz1,jz0);
317 dx20 = _mm_sub_pd(ix2,jx0);
318 dy20 = _mm_sub_pd(iy2,jy0);
319 dz20 = _mm_sub_pd(iz2,jz0);
320
321 /* Calculate squared distance and things based on it */
322 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
323 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
324 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
325
326 rinv00 = gmx_mm_invsqrt_pd(rsq00);
327 rinv10 = gmx_mm_invsqrt_pd(rsq10);
328 rinv20 = gmx_mm_invsqrt_pd(rsq20);
329
330 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
331 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
332 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
333
334 /* Load parameters for j particles */
335 jq0 = _mm_load_sd(charge+jnrA+0);
336
337 fjx0 = _mm_setzero_pd();
338 fjy0 = _mm_setzero_pd();
339 fjz0 = _mm_setzero_pd();
340
341 /**************************
342 * CALCULATE INTERACTIONS *
343 **************************/
344
345 /* Compute parameters for interactions between i and j atoms */
346 qq00 = _mm_mul_pd(iq0,jq0);
347
348 /* REACTION-FIELD ELECTROSTATICS */
349 velec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_add_pd(rinv00,_mm_mul_pd(krf,rsq00)),crf));
350 felec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
351
352 /* Update potential sum for this i atom from the interaction with this j atom. */
353 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
354 velecsum = _mm_add_pd(velecsum,velec);
355
356 fscal = felec;
357
358 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
359
360 /* Calculate temporary vectorial force */
361 tx = _mm_mul_pd(fscal,dx00);
362 ty = _mm_mul_pd(fscal,dy00);
363 tz = _mm_mul_pd(fscal,dz00);
364
365 /* Update vectorial force */
366 fix0 = _mm_add_pd(fix0,tx);
367 fiy0 = _mm_add_pd(fiy0,ty);
368 fiz0 = _mm_add_pd(fiz0,tz);
369
370 fjx0 = _mm_add_pd(fjx0,tx);
371 fjy0 = _mm_add_pd(fjy0,ty);
372 fjz0 = _mm_add_pd(fjz0,tz);
373
374 /**************************
375 * CALCULATE INTERACTIONS *
376 **************************/
377
378 /* Compute parameters for interactions between i and j atoms */
379 qq10 = _mm_mul_pd(iq1,jq0);
380
381 /* REACTION-FIELD ELECTROSTATICS */
382 velec = _mm_mul_pd(qq10,_mm_sub_pd(_mm_add_pd(rinv10,_mm_mul_pd(krf,rsq10)),crf));
383 felec = _mm_mul_pd(qq10,_mm_sub_pd(_mm_mul_pd(rinv10,rinvsq10),krf2));
384
385 /* Update potential sum for this i atom from the interaction with this j atom. */
386 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
387 velecsum = _mm_add_pd(velecsum,velec);
388
389 fscal = felec;
390
391 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
392
393 /* Calculate temporary vectorial force */
394 tx = _mm_mul_pd(fscal,dx10);
395 ty = _mm_mul_pd(fscal,dy10);
396 tz = _mm_mul_pd(fscal,dz10);
397
398 /* Update vectorial force */
399 fix1 = _mm_add_pd(fix1,tx);
400 fiy1 = _mm_add_pd(fiy1,ty);
401 fiz1 = _mm_add_pd(fiz1,tz);
402
403 fjx0 = _mm_add_pd(fjx0,tx);
404 fjy0 = _mm_add_pd(fjy0,ty);
405 fjz0 = _mm_add_pd(fjz0,tz);
406
407 /**************************
408 * CALCULATE INTERACTIONS *
409 **************************/
410
411 /* Compute parameters for interactions between i and j atoms */
412 qq20 = _mm_mul_pd(iq2,jq0);
413
414 /* REACTION-FIELD ELECTROSTATICS */
415 velec = _mm_mul_pd(qq20,_mm_sub_pd(_mm_add_pd(rinv20,_mm_mul_pd(krf,rsq20)),crf));
416 felec = _mm_mul_pd(qq20,_mm_sub_pd(_mm_mul_pd(rinv20,rinvsq20),krf2));
417
418 /* Update potential sum for this i atom from the interaction with this j atom. */
419 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
420 velecsum = _mm_add_pd(velecsum,velec);
421
422 fscal = felec;
423
424 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
425
426 /* Calculate temporary vectorial force */
427 tx = _mm_mul_pd(fscal,dx20);
428 ty = _mm_mul_pd(fscal,dy20);
429 tz = _mm_mul_pd(fscal,dz20);
430
431 /* Update vectorial force */
432 fix2 = _mm_add_pd(fix2,tx);
433 fiy2 = _mm_add_pd(fiy2,ty);
434 fiz2 = _mm_add_pd(fiz2,tz);
435
436 fjx0 = _mm_add_pd(fjx0,tx);
437 fjy0 = _mm_add_pd(fjy0,ty);
438 fjz0 = _mm_add_pd(fjz0,tz);
439
440 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0);
441
442 /* Inner loop uses 99 flops */
443 }
444
445 /* End of innermost loop */
446
447 gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
448 f+i_coord_offset,fshift+i_shift_offset);
449
450 ggid = gid[iidx];
451 /* Update potential energies */
452 gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
453
454 /* Increment number of inner iterations */
455 inneriter += j_index_end - j_index_start;
456
457 /* Outer loop uses 19 flops */
458 }
459
460 /* Increment number of outer iterations */
461 outeriter += nri;
462
463 /* Update outer/inner flops */
464
465 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3_VF,outeriter*19 + inneriter*99);
466 }
467 /*
468 * Gromacs nonbonded kernel: nb_kernel_ElecRF_VdwNone_GeomW3P1_F_sse2_double
469 * Electrostatics interaction: ReactionField
470 * VdW interaction: None
471 * Geometry: Water3-Particle
472 * Calculate force/pot: Force
473 */
474 void
475 nb_kernel_ElecRF_VdwNone_GeomW3P1_F_sse2_double
476 (t_nblist * gmx_restrict nlist,
477 rvec * gmx_restrict xx,
478 rvec * gmx_restrict ff,
479 t_forcerec * gmx_restrict fr,
480 t_mdatoms * gmx_restrict mdatoms,
481 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
482 t_nrnb * gmx_restrict nrnb)
483 {
484 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
485 * just 0 for non-waters.
486 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
487 * jnr indices corresponding to data put in the four positions in the SIMD register.
488 */
489 int i_shift_offset,i_coord_offset,outeriter,inneriter;
490 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
491 int jnrA,jnrB;
492 int j_coord_offsetA,j_coord_offsetB;
493 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
494 real rcutoff_scalar;
495 real *shiftvec,*fshift,*x,*f;
496 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
497 int vdwioffset0;
498 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
499 int vdwioffset1;
500 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
501 int vdwioffset2;
502 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
503 int vdwjidx0A,vdwjidx0B;
504 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
505 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
506 __m128d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
507 __m128d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
508 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
509 real *charge;
510 __m128d dummy_mask,cutoff_mask;
511 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
512 __m128d one = _mm_set1_pd(1.0);
513 __m128d two = _mm_set1_pd(2.0);
514 x = xx[0];
515 f = ff[0];
516
517 nri = nlist->nri;
518 iinr = nlist->iinr;
519 jindex = nlist->jindex;
520 jjnr = nlist->jjnr;
521 shiftidx = nlist->shift;
522 gid = nlist->gid;
523 shiftvec = fr->shift_vec[0];
524 fshift = fr->fshift[0];
525 facel = _mm_set1_pd(fr->epsfac);
526 charge = mdatoms->chargeA;
527 krf = _mm_set1_pd(fr->ic->k_rf);
528 krf2 = _mm_set1_pd(fr->ic->k_rf*2.0);
529 crf = _mm_set1_pd(fr->ic->c_rf);
530
531 /* Setup water-specific parameters */
532 inr = nlist->iinr[0];
533 iq0 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
534 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
535 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
536
537 /* Avoid stupid compiler warnings */
538 jnrA = jnrB = 0;
539 j_coord_offsetA = 0;
540 j_coord_offsetB = 0;
541
542 outeriter = 0;
543 inneriter = 0;
544
545 /* Start outer loop over neighborlists */
546 for(iidx=0; iidx<nri; iidx++)
547 {
548 /* Load shift vector for this list */
549 i_shift_offset = DIM*shiftidx[iidx];
550
551 /* Load limits for loop over neighbors */
552 j_index_start = jindex[iidx];
553 j_index_end = jindex[iidx+1];
554
555 /* Get outer coordinate index */
556 inr = iinr[iidx];
557 i_coord_offset = DIM*inr;
558
559 /* Load i particle coords and add shift vector */
560 gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
561 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
562
563 fix0 = _mm_setzero_pd();
564 fiy0 = _mm_setzero_pd();
565 fiz0 = _mm_setzero_pd();
566 fix1 = _mm_setzero_pd();
567 fiy1 = _mm_setzero_pd();
568 fiz1 = _mm_setzero_pd();
569 fix2 = _mm_setzero_pd();
570 fiy2 = _mm_setzero_pd();
571 fiz2 = _mm_setzero_pd();
572
573 /* Start inner kernel loop */
574 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
575 {
576
577 /* Get j neighbor index, and coordinate index */
578 jnrA = jjnr[jidx];
579 jnrB = jjnr[jidx+1];
580 j_coord_offsetA = DIM*jnrA;
581 j_coord_offsetB = DIM*jnrB;
582
583 /* load j atom coordinates */
584 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
585 &jx0,&jy0,&jz0);
586
587 /* Calculate displacement vector */
588 dx00 = _mm_sub_pd(ix0,jx0);
589 dy00 = _mm_sub_pd(iy0,jy0);
590 dz00 = _mm_sub_pd(iz0,jz0);
591 dx10 = _mm_sub_pd(ix1,jx0);
592 dy10 = _mm_sub_pd(iy1,jy0);
593 dz10 = _mm_sub_pd(iz1,jz0);
594 dx20 = _mm_sub_pd(ix2,jx0);
595 dy20 = _mm_sub_pd(iy2,jy0);
596 dz20 = _mm_sub_pd(iz2,jz0);
597
598 /* Calculate squared distance and things based on it */
599 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
600 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
601 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
602
603 rinv00 = gmx_mm_invsqrt_pd(rsq00);
604 rinv10 = gmx_mm_invsqrt_pd(rsq10);
605 rinv20 = gmx_mm_invsqrt_pd(rsq20);
606
607 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
608 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
609 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
610
611 /* Load parameters for j particles */
612 jq0 = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
613
614 fjx0 = _mm_setzero_pd();
615 fjy0 = _mm_setzero_pd();
616 fjz0 = _mm_setzero_pd();
617
618 /**************************
619 * CALCULATE INTERACTIONS *
620 **************************/
621
622 /* Compute parameters for interactions between i and j atoms */
623 qq00 = _mm_mul_pd(iq0,jq0);
624
625 /* REACTION-FIELD ELECTROSTATICS */
626 felec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
627
628 fscal = felec;
629
630 /* Calculate temporary vectorial force */
631 tx = _mm_mul_pd(fscal,dx00);
632 ty = _mm_mul_pd(fscal,dy00);
633 tz = _mm_mul_pd(fscal,dz00);
634
635 /* Update vectorial force */
636 fix0 = _mm_add_pd(fix0,tx);
637 fiy0 = _mm_add_pd(fiy0,ty);
638 fiz0 = _mm_add_pd(fiz0,tz);
639
640 fjx0 = _mm_add_pd(fjx0,tx);
641 fjy0 = _mm_add_pd(fjy0,ty);
642 fjz0 = _mm_add_pd(fjz0,tz);
643
644 /**************************
645 * CALCULATE INTERACTIONS *
646 **************************/
647
648 /* Compute parameters for interactions between i and j atoms */
649 qq10 = _mm_mul_pd(iq1,jq0);
650
651 /* REACTION-FIELD ELECTROSTATICS */
652 felec = _mm_mul_pd(qq10,_mm_sub_pd(_mm_mul_pd(rinv10,rinvsq10),krf2));
653
654 fscal = felec;
655
656 /* Calculate temporary vectorial force */
657 tx = _mm_mul_pd(fscal,dx10);
658 ty = _mm_mul_pd(fscal,dy10);
659 tz = _mm_mul_pd(fscal,dz10);
660
661 /* Update vectorial force */
662 fix1 = _mm_add_pd(fix1,tx);
663 fiy1 = _mm_add_pd(fiy1,ty);
664 fiz1 = _mm_add_pd(fiz1,tz);
665
666 fjx0 = _mm_add_pd(fjx0,tx);
667 fjy0 = _mm_add_pd(fjy0,ty);
668 fjz0 = _mm_add_pd(fjz0,tz);
669
670 /**************************
671 * CALCULATE INTERACTIONS *
672 **************************/
673
674 /* Compute parameters for interactions between i and j atoms */
675 qq20 = _mm_mul_pd(iq2,jq0);
676
677 /* REACTION-FIELD ELECTROSTATICS */
678 felec = _mm_mul_pd(qq20,_mm_sub_pd(_mm_mul_pd(rinv20,rinvsq20),krf2));
679
680 fscal = felec;
681
682 /* Calculate temporary vectorial force */
683 tx = _mm_mul_pd(fscal,dx20);
684 ty = _mm_mul_pd(fscal,dy20);
685 tz = _mm_mul_pd(fscal,dz20);
686
687 /* Update vectorial force */
688 fix2 = _mm_add_pd(fix2,tx);
689 fiy2 = _mm_add_pd(fiy2,ty);
690 fiz2 = _mm_add_pd(fiz2,tz);
691
692 fjx0 = _mm_add_pd(fjx0,tx);
693 fjy0 = _mm_add_pd(fjy0,ty);
694 fjz0 = _mm_add_pd(fjz0,tz);
695
696 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
697
698 /* Inner loop uses 84 flops */
699 }
700
701 if(jidx<j_index_end)
702 {
703
704 jnrA = jjnr[jidx];
705 j_coord_offsetA = DIM*jnrA;
706
707 /* load j atom coordinates */
708 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
709 &jx0,&jy0,&jz0);
710
711 /* Calculate displacement vector */
712 dx00 = _mm_sub_pd(ix0,jx0);
713 dy00 = _mm_sub_pd(iy0,jy0);
714 dz00 = _mm_sub_pd(iz0,jz0);
715 dx10 = _mm_sub_pd(ix1,jx0);
716 dy10 = _mm_sub_pd(iy1,jy0);
717 dz10 = _mm_sub_pd(iz1,jz0);
718 dx20 = _mm_sub_pd(ix2,jx0);
719 dy20 = _mm_sub_pd(iy2,jy0);
720 dz20 = _mm_sub_pd(iz2,jz0);
721
722 /* Calculate squared distance and things based on it */
723 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
724 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
725 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
726
727 rinv00 = gmx_mm_invsqrt_pd(rsq00);
728 rinv10 = gmx_mm_invsqrt_pd(rsq10);
729 rinv20 = gmx_mm_invsqrt_pd(rsq20);
730
731 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
732 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
733 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
734
735 /* Load parameters for j particles */
736 jq0 = _mm_load_sd(charge+jnrA+0);
737
738 fjx0 = _mm_setzero_pd();
739 fjy0 = _mm_setzero_pd();
740 fjz0 = _mm_setzero_pd();
741
742 /**************************
743 * CALCULATE INTERACTIONS *
744 **************************/
745
746 /* Compute parameters for interactions between i and j atoms */
747 qq00 = _mm_mul_pd(iq0,jq0);
748
749 /* REACTION-FIELD ELECTROSTATICS */
750 felec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
751
752 fscal = felec;
753
754 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
755
756 /* Calculate temporary vectorial force */
757 tx = _mm_mul_pd(fscal,dx00);
758 ty = _mm_mul_pd(fscal,dy00);
759 tz = _mm_mul_pd(fscal,dz00);
760
761 /* Update vectorial force */
762 fix0 = _mm_add_pd(fix0,tx);
763 fiy0 = _mm_add_pd(fiy0,ty);
764 fiz0 = _mm_add_pd(fiz0,tz);
765
766 fjx0 = _mm_add_pd(fjx0,tx);
767 fjy0 = _mm_add_pd(fjy0,ty);
768 fjz0 = _mm_add_pd(fjz0,tz);
769
770 /**************************
771 * CALCULATE INTERACTIONS *
772 **************************/
773
774 /* Compute parameters for interactions between i and j atoms */
775 qq10 = _mm_mul_pd(iq1,jq0);
776
777 /* REACTION-FIELD ELECTROSTATICS */
778 felec = _mm_mul_pd(qq10,_mm_sub_pd(_mm_mul_pd(rinv10,rinvsq10),krf2));
779
780 fscal = felec;
781
782 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
783
784 /* Calculate temporary vectorial force */
785 tx = _mm_mul_pd(fscal,dx10);
786 ty = _mm_mul_pd(fscal,dy10);
787 tz = _mm_mul_pd(fscal,dz10);
788
789 /* Update vectorial force */
790 fix1 = _mm_add_pd(fix1,tx);
791 fiy1 = _mm_add_pd(fiy1,ty);
792 fiz1 = _mm_add_pd(fiz1,tz);
793
794 fjx0 = _mm_add_pd(fjx0,tx);
795 fjy0 = _mm_add_pd(fjy0,ty);
796 fjz0 = _mm_add_pd(fjz0,tz);
797
798 /**************************
799 * CALCULATE INTERACTIONS *
800 **************************/
801
802 /* Compute parameters for interactions between i and j atoms */
803 qq20 = _mm_mul_pd(iq2,jq0);
804
805 /* REACTION-FIELD ELECTROSTATICS */
806 felec = _mm_mul_pd(qq20,_mm_sub_pd(_mm_mul_pd(rinv20,rinvsq20),krf2));
807
808 fscal = felec;
809
810 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
811
812 /* Calculate temporary vectorial force */
813 tx = _mm_mul_pd(fscal,dx20);
814 ty = _mm_mul_pd(fscal,dy20);
815 tz = _mm_mul_pd(fscal,dz20);
816
817 /* Update vectorial force */
818 fix2 = _mm_add_pd(fix2,tx);
819 fiy2 = _mm_add_pd(fiy2,ty);
820 fiz2 = _mm_add_pd(fiz2,tz);
821
822 fjx0 = _mm_add_pd(fjx0,tx);
823 fjy0 = _mm_add_pd(fjy0,ty);
824 fjz0 = _mm_add_pd(fjz0,tz);
825
826 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0);
827
828 /* Inner loop uses 84 flops */
829 }
830
831 /* End of innermost loop */
832
833 gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
834 f+i_coord_offset,fshift+i_shift_offset);
835
836 /* Increment number of inner iterations */
837 inneriter += j_index_end - j_index_start;
838
839 /* Outer loop uses 18 flops */
840 }
841
842 /* Increment number of outer iterations */
843 outeriter += nri;
844
845 /* Update outer/inner flops */
846
847 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3_F,outeriter*18 + inneriter*84);
848 }
The ultimate molecular dynamics simulation package