?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Removed simple.h from nb_kernel_sse4_1_XX
[gromacs.git] / src / gromacs / gmxlib / nonbonded / nb_kernel_sse4_1_double / nb_kernel_ElecRF_VdwNone_GeomP1P1_sse4_1_double.c
blob57737dba6dc813d51c67ca00c542cb86b966b24a
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 sse4_1_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_sse4_1_double.h"
49 #include "kernelutil_x86_sse4_1_double.h"
50
51 /*
52 * Gromacs nonbonded kernel: nb_kernel_ElecRF_VdwNone_GeomP1P1_VF_sse4_1_double
53 * Electrostatics interaction: ReactionField
54 * VdW interaction: None
55 * Geometry: Particle-Particle
56 * Calculate force/pot: PotentialAndForce
57 */
58 void
59 nb_kernel_ElecRF_VdwNone_GeomP1P1_VF_sse4_1_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 vdwjidx0A,vdwjidx0B;
84 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
85 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
86 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
87 real *charge;
88 __m128d dummy_mask,cutoff_mask;
89 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
90 __m128d one = _mm_set1_pd(1.0);
91 __m128d two = _mm_set1_pd(2.0);
92 x = xx[0];
93 f = ff[0];
94
95 nri = nlist->nri;
96 iinr = nlist->iinr;
97 jindex = nlist->jindex;
98 jjnr = nlist->jjnr;
99 shiftidx = nlist->shift;
100 gid = nlist->gid;
101 shiftvec = fr->shift_vec[0];
102 fshift = fr->fshift[0];
103 facel = _mm_set1_pd(fr->epsfac);
104 charge = mdatoms->chargeA;
105 krf = _mm_set1_pd(fr->ic->k_rf);
106 krf2 = _mm_set1_pd(fr->ic->k_rf*2.0);
107 crf = _mm_set1_pd(fr->ic->c_rf);
108
109 /* Avoid stupid compiler warnings */
110 jnrA = jnrB = 0;
111 j_coord_offsetA = 0;
112 j_coord_offsetB = 0;
113
114 outeriter = 0;
115 inneriter = 0;
116
117 /* Start outer loop over neighborlists */
118 for(iidx=0; iidx<nri; iidx++)
119 {
120 /* Load shift vector for this list */
121 i_shift_offset = DIM*shiftidx[iidx];
122
123 /* Load limits for loop over neighbors */
124 j_index_start = jindex[iidx];
125 j_index_end = jindex[iidx+1];
126
127 /* Get outer coordinate index */
128 inr = iinr[iidx];
129 i_coord_offset = DIM*inr;
130
131 /* Load i particle coords and add shift vector */
132 gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
133
134 fix0 = _mm_setzero_pd();
135 fiy0 = _mm_setzero_pd();
136 fiz0 = _mm_setzero_pd();
137
138 /* Load parameters for i particles */
139 iq0 = _mm_mul_pd(facel,_mm_load1_pd(charge+inr+0));
140
141 /* Reset potential sums */
142 velecsum = _mm_setzero_pd();
143
144 /* Start inner kernel loop */
145 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
146 {
147
148 /* Get j neighbor index, and coordinate index */
149 jnrA = jjnr[jidx];
150 jnrB = jjnr[jidx+1];
151 j_coord_offsetA = DIM*jnrA;
152 j_coord_offsetB = DIM*jnrB;
153
154 /* load j atom coordinates */
155 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
156 &jx0,&jy0,&jz0);
157
158 /* Calculate displacement vector */
159 dx00 = _mm_sub_pd(ix0,jx0);
160 dy00 = _mm_sub_pd(iy0,jy0);
161 dz00 = _mm_sub_pd(iz0,jz0);
162
163 /* Calculate squared distance and things based on it */
164 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
165
166 rinv00 = gmx_mm_invsqrt_pd(rsq00);
167
168 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
169
170 /* Load parameters for j particles */
171 jq0 = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
172
173 /**************************
174 * CALCULATE INTERACTIONS *
175 **************************/
176
177 /* Compute parameters for interactions between i and j atoms */
178 qq00 = _mm_mul_pd(iq0,jq0);
179
180 /* REACTION-FIELD ELECTROSTATICS */
181 velec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_add_pd(rinv00,_mm_mul_pd(krf,rsq00)),crf));
182 felec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
183
184 /* Update potential sum for this i atom from the interaction with this j atom. */
185 velecsum = _mm_add_pd(velecsum,velec);
186
187 fscal = felec;
188
189 /* Calculate temporary vectorial force */
190 tx = _mm_mul_pd(fscal,dx00);
191 ty = _mm_mul_pd(fscal,dy00);
192 tz = _mm_mul_pd(fscal,dz00);
193
194 /* Update vectorial force */
195 fix0 = _mm_add_pd(fix0,tx);
196 fiy0 = _mm_add_pd(fiy0,ty);
197 fiz0 = _mm_add_pd(fiz0,tz);
198
199 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,tx,ty,tz);
200
201 /* Inner loop uses 32 flops */
202 }
203
204 if(jidx<j_index_end)
205 {
206
207 jnrA = jjnr[jidx];
208 j_coord_offsetA = DIM*jnrA;
209
210 /* load j atom coordinates */
211 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
212 &jx0,&jy0,&jz0);
213
214 /* Calculate displacement vector */
215 dx00 = _mm_sub_pd(ix0,jx0);
216 dy00 = _mm_sub_pd(iy0,jy0);
217 dz00 = _mm_sub_pd(iz0,jz0);
218
219 /* Calculate squared distance and things based on it */
220 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
221
222 rinv00 = gmx_mm_invsqrt_pd(rsq00);
223
224 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
225
226 /* Load parameters for j particles */
227 jq0 = _mm_load_sd(charge+jnrA+0);
228
229 /**************************
230 * CALCULATE INTERACTIONS *
231 **************************/
232
233 /* Compute parameters for interactions between i and j atoms */
234 qq00 = _mm_mul_pd(iq0,jq0);
235
236 /* REACTION-FIELD ELECTROSTATICS */
237 velec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_add_pd(rinv00,_mm_mul_pd(krf,rsq00)),crf));
238 felec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
239
240 /* Update potential sum for this i atom from the interaction with this j atom. */
241 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
242 velecsum = _mm_add_pd(velecsum,velec);
243
244 fscal = felec;
245
246 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
247
248 /* Calculate temporary vectorial force */
249 tx = _mm_mul_pd(fscal,dx00);
250 ty = _mm_mul_pd(fscal,dy00);
251 tz = _mm_mul_pd(fscal,dz00);
252
253 /* Update vectorial force */
254 fix0 = _mm_add_pd(fix0,tx);
255 fiy0 = _mm_add_pd(fiy0,ty);
256 fiz0 = _mm_add_pd(fiz0,tz);
257
258 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,tx,ty,tz);
259
260 /* Inner loop uses 32 flops */
261 }
262
263 /* End of innermost loop */
264
265 gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
266 f+i_coord_offset,fshift+i_shift_offset);
267
268 ggid = gid[iidx];
269 /* Update potential energies */
270 gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
271
272 /* Increment number of inner iterations */
273 inneriter += j_index_end - j_index_start;
274
275 /* Outer loop uses 8 flops */
276 }
277
278 /* Increment number of outer iterations */
279 outeriter += nri;
280
281 /* Update outer/inner flops */
282
283 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VF,outeriter*8 + inneriter*32);
284 }
285 /*
286 * Gromacs nonbonded kernel: nb_kernel_ElecRF_VdwNone_GeomP1P1_F_sse4_1_double
287 * Electrostatics interaction: ReactionField
288 * VdW interaction: None
289 * Geometry: Particle-Particle
290 * Calculate force/pot: Force
291 */
292 void
293 nb_kernel_ElecRF_VdwNone_GeomP1P1_F_sse4_1_double
294 (t_nblist * gmx_restrict nlist,
295 rvec * gmx_restrict xx,
296 rvec * gmx_restrict ff,
297 t_forcerec * gmx_restrict fr,
298 t_mdatoms * gmx_restrict mdatoms,
299 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
300 t_nrnb * gmx_restrict nrnb)
301 {
302 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
303 * just 0 for non-waters.
304 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
305 * jnr indices corresponding to data put in the four positions in the SIMD register.
306 */
307 int i_shift_offset,i_coord_offset,outeriter,inneriter;
308 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
309 int jnrA,jnrB;
310 int j_coord_offsetA,j_coord_offsetB;
311 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
312 real rcutoff_scalar;
313 real *shiftvec,*fshift,*x,*f;
314 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
315 int vdwioffset0;
316 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
317 int vdwjidx0A,vdwjidx0B;
318 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
319 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
320 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
321 real *charge;
322 __m128d dummy_mask,cutoff_mask;
323 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
324 __m128d one = _mm_set1_pd(1.0);
325 __m128d two = _mm_set1_pd(2.0);
326 x = xx[0];
327 f = ff[0];
328
329 nri = nlist->nri;
330 iinr = nlist->iinr;
331 jindex = nlist->jindex;
332 jjnr = nlist->jjnr;
333 shiftidx = nlist->shift;
334 gid = nlist->gid;
335 shiftvec = fr->shift_vec[0];
336 fshift = fr->fshift[0];
337 facel = _mm_set1_pd(fr->epsfac);
338 charge = mdatoms->chargeA;
339 krf = _mm_set1_pd(fr->ic->k_rf);
340 krf2 = _mm_set1_pd(fr->ic->k_rf*2.0);
341 crf = _mm_set1_pd(fr->ic->c_rf);
342
343 /* Avoid stupid compiler warnings */
344 jnrA = jnrB = 0;
345 j_coord_offsetA = 0;
346 j_coord_offsetB = 0;
347
348 outeriter = 0;
349 inneriter = 0;
350
351 /* Start outer loop over neighborlists */
352 for(iidx=0; iidx<nri; iidx++)
353 {
354 /* Load shift vector for this list */
355 i_shift_offset = DIM*shiftidx[iidx];
356
357 /* Load limits for loop over neighbors */
358 j_index_start = jindex[iidx];
359 j_index_end = jindex[iidx+1];
360
361 /* Get outer coordinate index */
362 inr = iinr[iidx];
363 i_coord_offset = DIM*inr;
364
365 /* Load i particle coords and add shift vector */
366 gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
367
368 fix0 = _mm_setzero_pd();
369 fiy0 = _mm_setzero_pd();
370 fiz0 = _mm_setzero_pd();
371
372 /* Load parameters for i particles */
373 iq0 = _mm_mul_pd(facel,_mm_load1_pd(charge+inr+0));
374
375 /* Start inner kernel loop */
376 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
377 {
378
379 /* Get j neighbor index, and coordinate index */
380 jnrA = jjnr[jidx];
381 jnrB = jjnr[jidx+1];
382 j_coord_offsetA = DIM*jnrA;
383 j_coord_offsetB = DIM*jnrB;
384
385 /* load j atom coordinates */
386 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
387 &jx0,&jy0,&jz0);
388
389 /* Calculate displacement vector */
390 dx00 = _mm_sub_pd(ix0,jx0);
391 dy00 = _mm_sub_pd(iy0,jy0);
392 dz00 = _mm_sub_pd(iz0,jz0);
393
394 /* Calculate squared distance and things based on it */
395 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
396
397 rinv00 = gmx_mm_invsqrt_pd(rsq00);
398
399 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
400
401 /* Load parameters for j particles */
402 jq0 = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
403
404 /**************************
405 * CALCULATE INTERACTIONS *
406 **************************/
407
408 /* Compute parameters for interactions between i and j atoms */
409 qq00 = _mm_mul_pd(iq0,jq0);
410
411 /* REACTION-FIELD ELECTROSTATICS */
412 felec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
413
414 fscal = felec;
415
416 /* Calculate temporary vectorial force */
417 tx = _mm_mul_pd(fscal,dx00);
418 ty = _mm_mul_pd(fscal,dy00);
419 tz = _mm_mul_pd(fscal,dz00);
420
421 /* Update vectorial force */
422 fix0 = _mm_add_pd(fix0,tx);
423 fiy0 = _mm_add_pd(fiy0,ty);
424 fiz0 = _mm_add_pd(fiz0,tz);
425
426 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,tx,ty,tz);
427
428 /* Inner loop uses 27 flops */
429 }
430
431 if(jidx<j_index_end)
432 {
433
434 jnrA = jjnr[jidx];
435 j_coord_offsetA = DIM*jnrA;
436
437 /* load j atom coordinates */
438 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
439 &jx0,&jy0,&jz0);
440
441 /* Calculate displacement vector */
442 dx00 = _mm_sub_pd(ix0,jx0);
443 dy00 = _mm_sub_pd(iy0,jy0);
444 dz00 = _mm_sub_pd(iz0,jz0);
445
446 /* Calculate squared distance and things based on it */
447 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
448
449 rinv00 = gmx_mm_invsqrt_pd(rsq00);
450
451 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
452
453 /* Load parameters for j particles */
454 jq0 = _mm_load_sd(charge+jnrA+0);
455
456 /**************************
457 * CALCULATE INTERACTIONS *
458 **************************/
459
460 /* Compute parameters for interactions between i and j atoms */
461 qq00 = _mm_mul_pd(iq0,jq0);
462
463 /* REACTION-FIELD ELECTROSTATICS */
464 felec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
465
466 fscal = felec;
467
468 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
469
470 /* Calculate temporary vectorial force */
471 tx = _mm_mul_pd(fscal,dx00);
472 ty = _mm_mul_pd(fscal,dy00);
473 tz = _mm_mul_pd(fscal,dz00);
474
475 /* Update vectorial force */
476 fix0 = _mm_add_pd(fix0,tx);
477 fiy0 = _mm_add_pd(fiy0,ty);
478 fiz0 = _mm_add_pd(fiz0,tz);
479
480 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,tx,ty,tz);
481
482 /* Inner loop uses 27 flops */
483 }
484
485 /* End of innermost loop */
486
487 gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
488 f+i_coord_offset,fshift+i_shift_offset);
489
490 /* Increment number of inner iterations */
491 inneriter += j_index_end - j_index_start;
492
493 /* Outer loop uses 7 flops */
494 }
495
496 /* Increment number of outer iterations */
497 outeriter += nri;
498
499 /* Update outer/inner flops */
500
501 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_F,outeriter*7 + inneriter*27);
502 }
The ultimate molecular dynamics simulation package