?  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_single / nb_kernel_ElecRF_VdwLJ_GeomW4P1_sse2_single.c
blob05f448164c7d26d2ac352ac1b88d161ae981b7cf
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_single 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_single.h"
49 #include "kernelutil_x86_sse2_single.h"
50
51 /*
52 * Gromacs nonbonded kernel: nb_kernel_ElecRF_VdwLJ_GeomW4P1_VF_sse2_single
53 * Electrostatics interaction: ReactionField
54 * VdW interaction: LennardJones
55 * Geometry: Water4-Particle
56 * Calculate force/pot: PotentialAndForce
57 */
58 void
59 nb_kernel_ElecRF_VdwLJ_GeomW4P1_VF_sse2_single
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,C,D refer to j loop unrolling done with SSE, e.g. for the four 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,jnrC,jnrD;
76 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
77 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
78 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
79 real rcutoff_scalar;
80 real *shiftvec,*fshift,*x,*f;
81 real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
82 real scratch[4*DIM];
83 __m128 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
84 int vdwioffset0;
85 __m128 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
86 int vdwioffset1;
87 __m128 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
88 int vdwioffset2;
89 __m128 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
90 int vdwioffset3;
91 __m128 ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
92 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
93 __m128 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
94 __m128 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
95 __m128 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
96 __m128 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
97 __m128 dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
98 __m128 velec,felec,velecsum,facel,crf,krf,krf2;
99 real *charge;
100 int nvdwtype;
101 __m128 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
102 int *vdwtype;
103 real *vdwparam;
104 __m128 one_sixth = _mm_set1_ps(1.0/6.0);
105 __m128 one_twelfth = _mm_set1_ps(1.0/12.0);
106 __m128 dummy_mask,cutoff_mask;
107 __m128 signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
108 __m128 one = _mm_set1_ps(1.0);
109 __m128 two = _mm_set1_ps(2.0);
110 x = xx[0];
111 f = ff[0];
112
113 nri = nlist->nri;
114 iinr = nlist->iinr;
115 jindex = nlist->jindex;
116 jjnr = nlist->jjnr;
117 shiftidx = nlist->shift;
118 gid = nlist->gid;
119 shiftvec = fr->shift_vec[0];
120 fshift = fr->fshift[0];
121 facel = _mm_set1_ps(fr->epsfac);
122 charge = mdatoms->chargeA;
123 krf = _mm_set1_ps(fr->ic->k_rf);
124 krf2 = _mm_set1_ps(fr->ic->k_rf*2.0);
125 crf = _mm_set1_ps(fr->ic->c_rf);
126 nvdwtype = fr->ntype;
127 vdwparam = fr->nbfp;
128 vdwtype = mdatoms->typeA;
129
130 /* Setup water-specific parameters */
131 inr = nlist->iinr[0];
132 iq1 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1]));
133 iq2 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2]));
134 iq3 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+3]));
135 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
136
137 /* Avoid stupid compiler warnings */
138 jnrA = jnrB = jnrC = jnrD = 0;
139 j_coord_offsetA = 0;
140 j_coord_offsetB = 0;
141 j_coord_offsetC = 0;
142 j_coord_offsetD = 0;
143
144 outeriter = 0;
145 inneriter = 0;
146
147 for(iidx=0;iidx<4*DIM;iidx++)
148 {
149 scratch[iidx] = 0.0;
150 }
151
152 /* Start outer loop over neighborlists */
153 for(iidx=0; iidx<nri; iidx++)
154 {
155 /* Load shift vector for this list */
156 i_shift_offset = DIM*shiftidx[iidx];
157
158 /* Load limits for loop over neighbors */
159 j_index_start = jindex[iidx];
160 j_index_end = jindex[iidx+1];
161
162 /* Get outer coordinate index */
163 inr = iinr[iidx];
164 i_coord_offset = DIM*inr;
165
166 /* Load i particle coords and add shift vector */
167 gmx_mm_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
168 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
169
170 fix0 = _mm_setzero_ps();
171 fiy0 = _mm_setzero_ps();
172 fiz0 = _mm_setzero_ps();
173 fix1 = _mm_setzero_ps();
174 fiy1 = _mm_setzero_ps();
175 fiz1 = _mm_setzero_ps();
176 fix2 = _mm_setzero_ps();
177 fiy2 = _mm_setzero_ps();
178 fiz2 = _mm_setzero_ps();
179 fix3 = _mm_setzero_ps();
180 fiy3 = _mm_setzero_ps();
181 fiz3 = _mm_setzero_ps();
182
183 /* Reset potential sums */
184 velecsum = _mm_setzero_ps();
185 vvdwsum = _mm_setzero_ps();
186
187 /* Start inner kernel loop */
188 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
189 {
190
191 /* Get j neighbor index, and coordinate index */
192 jnrA = jjnr[jidx];
193 jnrB = jjnr[jidx+1];
194 jnrC = jjnr[jidx+2];
195 jnrD = jjnr[jidx+3];
196 j_coord_offsetA = DIM*jnrA;
197 j_coord_offsetB = DIM*jnrB;
198 j_coord_offsetC = DIM*jnrC;
199 j_coord_offsetD = DIM*jnrD;
200
201 /* load j atom coordinates */
202 gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
203 x+j_coord_offsetC,x+j_coord_offsetD,
204 &jx0,&jy0,&jz0);
205
206 /* Calculate displacement vector */
207 dx00 = _mm_sub_ps(ix0,jx0);
208 dy00 = _mm_sub_ps(iy0,jy0);
209 dz00 = _mm_sub_ps(iz0,jz0);
210 dx10 = _mm_sub_ps(ix1,jx0);
211 dy10 = _mm_sub_ps(iy1,jy0);
212 dz10 = _mm_sub_ps(iz1,jz0);
213 dx20 = _mm_sub_ps(ix2,jx0);
214 dy20 = _mm_sub_ps(iy2,jy0);
215 dz20 = _mm_sub_ps(iz2,jz0);
216 dx30 = _mm_sub_ps(ix3,jx0);
217 dy30 = _mm_sub_ps(iy3,jy0);
218 dz30 = _mm_sub_ps(iz3,jz0);
219
220 /* Calculate squared distance and things based on it */
221 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
222 rsq10 = gmx_mm_calc_rsq_ps(dx10,dy10,dz10);
223 rsq20 = gmx_mm_calc_rsq_ps(dx20,dy20,dz20);
224 rsq30 = gmx_mm_calc_rsq_ps(dx30,dy30,dz30);
225
226 rinv10 = gmx_mm_invsqrt_ps(rsq10);
227 rinv20 = gmx_mm_invsqrt_ps(rsq20);
228 rinv30 = gmx_mm_invsqrt_ps(rsq30);
229
230 rinvsq00 = gmx_mm_inv_ps(rsq00);
231 rinvsq10 = _mm_mul_ps(rinv10,rinv10);
232 rinvsq20 = _mm_mul_ps(rinv20,rinv20);
233 rinvsq30 = _mm_mul_ps(rinv30,rinv30);
234
235 /* Load parameters for j particles */
236 jq0 = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
237 charge+jnrC+0,charge+jnrD+0);
238 vdwjidx0A = 2*vdwtype[jnrA+0];
239 vdwjidx0B = 2*vdwtype[jnrB+0];
240 vdwjidx0C = 2*vdwtype[jnrC+0];
241 vdwjidx0D = 2*vdwtype[jnrD+0];
242
243 fjx0 = _mm_setzero_ps();
244 fjy0 = _mm_setzero_ps();
245 fjz0 = _mm_setzero_ps();
246
247 /**************************
248 * CALCULATE INTERACTIONS *
249 **************************/
250
251 /* Compute parameters for interactions between i and j atoms */
252 gmx_mm_load_4pair_swizzle_ps(vdwparam+vdwioffset0+vdwjidx0A,
253 vdwparam+vdwioffset0+vdwjidx0B,
254 vdwparam+vdwioffset0+vdwjidx0C,
255 vdwparam+vdwioffset0+vdwjidx0D,
256 &c6_00,&c12_00);
257
258 /* LENNARD-JONES DISPERSION/REPULSION */
259
260 rinvsix = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00);
261 vvdw6 = _mm_mul_ps(c6_00,rinvsix);
262 vvdw12 = _mm_mul_ps(c12_00,_mm_mul_ps(rinvsix,rinvsix));
263 vvdw = _mm_sub_ps( _mm_mul_ps(vvdw12,one_twelfth) , _mm_mul_ps(vvdw6,one_sixth) );
264 fvdw = _mm_mul_ps(_mm_sub_ps(vvdw12,vvdw6),rinvsq00);
265
266 /* Update potential sum for this i atom from the interaction with this j atom. */
267 vvdwsum = _mm_add_ps(vvdwsum,vvdw);
268
269 fscal = fvdw;
270
271 /* Calculate temporary vectorial force */
272 tx = _mm_mul_ps(fscal,dx00);
273 ty = _mm_mul_ps(fscal,dy00);
274 tz = _mm_mul_ps(fscal,dz00);
275
276 /* Update vectorial force */
277 fix0 = _mm_add_ps(fix0,tx);
278 fiy0 = _mm_add_ps(fiy0,ty);
279 fiz0 = _mm_add_ps(fiz0,tz);
280
281 fjx0 = _mm_add_ps(fjx0,tx);
282 fjy0 = _mm_add_ps(fjy0,ty);
283 fjz0 = _mm_add_ps(fjz0,tz);
284
285 /**************************
286 * CALCULATE INTERACTIONS *
287 **************************/
288
289 /* Compute parameters for interactions between i and j atoms */
290 qq10 = _mm_mul_ps(iq1,jq0);
291
292 /* REACTION-FIELD ELECTROSTATICS */
293 velec = _mm_mul_ps(qq10,_mm_sub_ps(_mm_add_ps(rinv10,_mm_mul_ps(krf,rsq10)),crf));
294 felec = _mm_mul_ps(qq10,_mm_sub_ps(_mm_mul_ps(rinv10,rinvsq10),krf2));
295
296 /* Update potential sum for this i atom from the interaction with this j atom. */
297 velecsum = _mm_add_ps(velecsum,velec);
298
299 fscal = felec;
300
301 /* Calculate temporary vectorial force */
302 tx = _mm_mul_ps(fscal,dx10);
303 ty = _mm_mul_ps(fscal,dy10);
304 tz = _mm_mul_ps(fscal,dz10);
305
306 /* Update vectorial force */
307 fix1 = _mm_add_ps(fix1,tx);
308 fiy1 = _mm_add_ps(fiy1,ty);
309 fiz1 = _mm_add_ps(fiz1,tz);
310
311 fjx0 = _mm_add_ps(fjx0,tx);
312 fjy0 = _mm_add_ps(fjy0,ty);
313 fjz0 = _mm_add_ps(fjz0,tz);
314
315 /**************************
316 * CALCULATE INTERACTIONS *
317 **************************/
318
319 /* Compute parameters for interactions between i and j atoms */
320 qq20 = _mm_mul_ps(iq2,jq0);
321
322 /* REACTION-FIELD ELECTROSTATICS */
323 velec = _mm_mul_ps(qq20,_mm_sub_ps(_mm_add_ps(rinv20,_mm_mul_ps(krf,rsq20)),crf));
324 felec = _mm_mul_ps(qq20,_mm_sub_ps(_mm_mul_ps(rinv20,rinvsq20),krf2));
325
326 /* Update potential sum for this i atom from the interaction with this j atom. */
327 velecsum = _mm_add_ps(velecsum,velec);
328
329 fscal = felec;
330
331 /* Calculate temporary vectorial force */
332 tx = _mm_mul_ps(fscal,dx20);
333 ty = _mm_mul_ps(fscal,dy20);
334 tz = _mm_mul_ps(fscal,dz20);
335
336 /* Update vectorial force */
337 fix2 = _mm_add_ps(fix2,tx);
338 fiy2 = _mm_add_ps(fiy2,ty);
339 fiz2 = _mm_add_ps(fiz2,tz);
340
341 fjx0 = _mm_add_ps(fjx0,tx);
342 fjy0 = _mm_add_ps(fjy0,ty);
343 fjz0 = _mm_add_ps(fjz0,tz);
344
345 /**************************
346 * CALCULATE INTERACTIONS *
347 **************************/
348
349 /* Compute parameters for interactions between i and j atoms */
350 qq30 = _mm_mul_ps(iq3,jq0);
351
352 /* REACTION-FIELD ELECTROSTATICS */
353 velec = _mm_mul_ps(qq30,_mm_sub_ps(_mm_add_ps(rinv30,_mm_mul_ps(krf,rsq30)),crf));
354 felec = _mm_mul_ps(qq30,_mm_sub_ps(_mm_mul_ps(rinv30,rinvsq30),krf2));
355
356 /* Update potential sum for this i atom from the interaction with this j atom. */
357 velecsum = _mm_add_ps(velecsum,velec);
358
359 fscal = felec;
360
361 /* Calculate temporary vectorial force */
362 tx = _mm_mul_ps(fscal,dx30);
363 ty = _mm_mul_ps(fscal,dy30);
364 tz = _mm_mul_ps(fscal,dz30);
365
366 /* Update vectorial force */
367 fix3 = _mm_add_ps(fix3,tx);
368 fiy3 = _mm_add_ps(fiy3,ty);
369 fiz3 = _mm_add_ps(fiz3,tz);
370
371 fjx0 = _mm_add_ps(fjx0,tx);
372 fjy0 = _mm_add_ps(fjy0,ty);
373 fjz0 = _mm_add_ps(fjz0,tz);
374
375 fjptrA = f+j_coord_offsetA;
376 fjptrB = f+j_coord_offsetB;
377 fjptrC = f+j_coord_offsetC;
378 fjptrD = f+j_coord_offsetD;
379
380 gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjx0,fjy0,fjz0);
381
382 /* Inner loop uses 128 flops */
383 }
384
385 if(jidx<j_index_end)
386 {
387
388 /* Get j neighbor index, and coordinate index */
389 jnrlistA = jjnr[jidx];
390 jnrlistB = jjnr[jidx+1];
391 jnrlistC = jjnr[jidx+2];
392 jnrlistD = jjnr[jidx+3];
393 /* Sign of each element will be negative for non-real atoms.
394 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
395 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
396 */
397 dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
398 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
399 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
400 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
401 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
402 j_coord_offsetA = DIM*jnrA;
403 j_coord_offsetB = DIM*jnrB;
404 j_coord_offsetC = DIM*jnrC;
405 j_coord_offsetD = DIM*jnrD;
406
407 /* load j atom coordinates */
408 gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
409 x+j_coord_offsetC,x+j_coord_offsetD,
410 &jx0,&jy0,&jz0);
411
412 /* Calculate displacement vector */
413 dx00 = _mm_sub_ps(ix0,jx0);
414 dy00 = _mm_sub_ps(iy0,jy0);
415 dz00 = _mm_sub_ps(iz0,jz0);
416 dx10 = _mm_sub_ps(ix1,jx0);
417 dy10 = _mm_sub_ps(iy1,jy0);
418 dz10 = _mm_sub_ps(iz1,jz0);
419 dx20 = _mm_sub_ps(ix2,jx0);
420 dy20 = _mm_sub_ps(iy2,jy0);
421 dz20 = _mm_sub_ps(iz2,jz0);
422 dx30 = _mm_sub_ps(ix3,jx0);
423 dy30 = _mm_sub_ps(iy3,jy0);
424 dz30 = _mm_sub_ps(iz3,jz0);
425
426 /* Calculate squared distance and things based on it */
427 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
428 rsq10 = gmx_mm_calc_rsq_ps(dx10,dy10,dz10);
429 rsq20 = gmx_mm_calc_rsq_ps(dx20,dy20,dz20);
430 rsq30 = gmx_mm_calc_rsq_ps(dx30,dy30,dz30);
431
432 rinv10 = gmx_mm_invsqrt_ps(rsq10);
433 rinv20 = gmx_mm_invsqrt_ps(rsq20);
434 rinv30 = gmx_mm_invsqrt_ps(rsq30);
435
436 rinvsq00 = gmx_mm_inv_ps(rsq00);
437 rinvsq10 = _mm_mul_ps(rinv10,rinv10);
438 rinvsq20 = _mm_mul_ps(rinv20,rinv20);
439 rinvsq30 = _mm_mul_ps(rinv30,rinv30);
440
441 /* Load parameters for j particles */
442 jq0 = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
443 charge+jnrC+0,charge+jnrD+0);
444 vdwjidx0A = 2*vdwtype[jnrA+0];
445 vdwjidx0B = 2*vdwtype[jnrB+0];
446 vdwjidx0C = 2*vdwtype[jnrC+0];
447 vdwjidx0D = 2*vdwtype[jnrD+0];
448
449 fjx0 = _mm_setzero_ps();
450 fjy0 = _mm_setzero_ps();
451 fjz0 = _mm_setzero_ps();
452
453 /**************************
454 * CALCULATE INTERACTIONS *
455 **************************/
456
457 /* Compute parameters for interactions between i and j atoms */
458 gmx_mm_load_4pair_swizzle_ps(vdwparam+vdwioffset0+vdwjidx0A,
459 vdwparam+vdwioffset0+vdwjidx0B,
460 vdwparam+vdwioffset0+vdwjidx0C,
461 vdwparam+vdwioffset0+vdwjidx0D,
462 &c6_00,&c12_00);
463
464 /* LENNARD-JONES DISPERSION/REPULSION */
465
466 rinvsix = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00);
467 vvdw6 = _mm_mul_ps(c6_00,rinvsix);
468 vvdw12 = _mm_mul_ps(c12_00,_mm_mul_ps(rinvsix,rinvsix));
469 vvdw = _mm_sub_ps( _mm_mul_ps(vvdw12,one_twelfth) , _mm_mul_ps(vvdw6,one_sixth) );
470 fvdw = _mm_mul_ps(_mm_sub_ps(vvdw12,vvdw6),rinvsq00);
471
472 /* Update potential sum for this i atom from the interaction with this j atom. */
473 vvdw = _mm_andnot_ps(dummy_mask,vvdw);
474 vvdwsum = _mm_add_ps(vvdwsum,vvdw);
475
476 fscal = fvdw;
477
478 fscal = _mm_andnot_ps(dummy_mask,fscal);
479
480 /* Calculate temporary vectorial force */
481 tx = _mm_mul_ps(fscal,dx00);
482 ty = _mm_mul_ps(fscal,dy00);
483 tz = _mm_mul_ps(fscal,dz00);
484
485 /* Update vectorial force */
486 fix0 = _mm_add_ps(fix0,tx);
487 fiy0 = _mm_add_ps(fiy0,ty);
488 fiz0 = _mm_add_ps(fiz0,tz);
489
490 fjx0 = _mm_add_ps(fjx0,tx);
491 fjy0 = _mm_add_ps(fjy0,ty);
492 fjz0 = _mm_add_ps(fjz0,tz);
493
494 /**************************
495 * CALCULATE INTERACTIONS *
496 **************************/
497
498 /* Compute parameters for interactions between i and j atoms */
499 qq10 = _mm_mul_ps(iq1,jq0);
500
501 /* REACTION-FIELD ELECTROSTATICS */
502 velec = _mm_mul_ps(qq10,_mm_sub_ps(_mm_add_ps(rinv10,_mm_mul_ps(krf,rsq10)),crf));
503 felec = _mm_mul_ps(qq10,_mm_sub_ps(_mm_mul_ps(rinv10,rinvsq10),krf2));
504
505 /* Update potential sum for this i atom from the interaction with this j atom. */
506 velec = _mm_andnot_ps(dummy_mask,velec);
507 velecsum = _mm_add_ps(velecsum,velec);
508
509 fscal = felec;
510
511 fscal = _mm_andnot_ps(dummy_mask,fscal);
512
513 /* Calculate temporary vectorial force */
514 tx = _mm_mul_ps(fscal,dx10);
515 ty = _mm_mul_ps(fscal,dy10);
516 tz = _mm_mul_ps(fscal,dz10);
517
518 /* Update vectorial force */
519 fix1 = _mm_add_ps(fix1,tx);
520 fiy1 = _mm_add_ps(fiy1,ty);
521 fiz1 = _mm_add_ps(fiz1,tz);
522
523 fjx0 = _mm_add_ps(fjx0,tx);
524 fjy0 = _mm_add_ps(fjy0,ty);
525 fjz0 = _mm_add_ps(fjz0,tz);
526
527 /**************************
528 * CALCULATE INTERACTIONS *
529 **************************/
530
531 /* Compute parameters for interactions between i and j atoms */
532 qq20 = _mm_mul_ps(iq2,jq0);
533
534 /* REACTION-FIELD ELECTROSTATICS */
535 velec = _mm_mul_ps(qq20,_mm_sub_ps(_mm_add_ps(rinv20,_mm_mul_ps(krf,rsq20)),crf));
536 felec = _mm_mul_ps(qq20,_mm_sub_ps(_mm_mul_ps(rinv20,rinvsq20),krf2));
537
538 /* Update potential sum for this i atom from the interaction with this j atom. */
539 velec = _mm_andnot_ps(dummy_mask,velec);
540 velecsum = _mm_add_ps(velecsum,velec);
541
542 fscal = felec;
543
544 fscal = _mm_andnot_ps(dummy_mask,fscal);
545
546 /* Calculate temporary vectorial force */
547 tx = _mm_mul_ps(fscal,dx20);
548 ty = _mm_mul_ps(fscal,dy20);
549 tz = _mm_mul_ps(fscal,dz20);
550
551 /* Update vectorial force */
552 fix2 = _mm_add_ps(fix2,tx);
553 fiy2 = _mm_add_ps(fiy2,ty);
554 fiz2 = _mm_add_ps(fiz2,tz);
555
556 fjx0 = _mm_add_ps(fjx0,tx);
557 fjy0 = _mm_add_ps(fjy0,ty);
558 fjz0 = _mm_add_ps(fjz0,tz);
559
560 /**************************
561 * CALCULATE INTERACTIONS *
562 **************************/
563
564 /* Compute parameters for interactions between i and j atoms */
565 qq30 = _mm_mul_ps(iq3,jq0);
566
567 /* REACTION-FIELD ELECTROSTATICS */
568 velec = _mm_mul_ps(qq30,_mm_sub_ps(_mm_add_ps(rinv30,_mm_mul_ps(krf,rsq30)),crf));
569 felec = _mm_mul_ps(qq30,_mm_sub_ps(_mm_mul_ps(rinv30,rinvsq30),krf2));
570
571 /* Update potential sum for this i atom from the interaction with this j atom. */
572 velec = _mm_andnot_ps(dummy_mask,velec);
573 velecsum = _mm_add_ps(velecsum,velec);
574
575 fscal = felec;
576
577 fscal = _mm_andnot_ps(dummy_mask,fscal);
578
579 /* Calculate temporary vectorial force */
580 tx = _mm_mul_ps(fscal,dx30);
581 ty = _mm_mul_ps(fscal,dy30);
582 tz = _mm_mul_ps(fscal,dz30);
583
584 /* Update vectorial force */
585 fix3 = _mm_add_ps(fix3,tx);
586 fiy3 = _mm_add_ps(fiy3,ty);
587 fiz3 = _mm_add_ps(fiz3,tz);
588
589 fjx0 = _mm_add_ps(fjx0,tx);
590 fjy0 = _mm_add_ps(fjy0,ty);
591 fjz0 = _mm_add_ps(fjz0,tz);
592
593 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
594 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
595 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
596 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
597
598 gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjx0,fjy0,fjz0);
599
600 /* Inner loop uses 128 flops */
601 }
602
603 /* End of innermost loop */
604
605 gmx_mm_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
606 f+i_coord_offset,fshift+i_shift_offset);
607
608 ggid = gid[iidx];
609 /* Update potential energies */
610 gmx_mm_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
611 gmx_mm_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
612
613 /* Increment number of inner iterations */
614 inneriter += j_index_end - j_index_start;
615
616 /* Outer loop uses 26 flops */
617 }
618
619 /* Increment number of outer iterations */
620 outeriter += nri;
621
622 /* Update outer/inner flops */
623
624 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_VF,outeriter*26 + inneriter*128);
625 }
626 /*
627 * Gromacs nonbonded kernel: nb_kernel_ElecRF_VdwLJ_GeomW4P1_F_sse2_single
628 * Electrostatics interaction: ReactionField
629 * VdW interaction: LennardJones
630 * Geometry: Water4-Particle
631 * Calculate force/pot: Force
632 */
633 void
634 nb_kernel_ElecRF_VdwLJ_GeomW4P1_F_sse2_single
635 (t_nblist * gmx_restrict nlist,
636 rvec * gmx_restrict xx,
637 rvec * gmx_restrict ff,
638 t_forcerec * gmx_restrict fr,
639 t_mdatoms * gmx_restrict mdatoms,
640 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
641 t_nrnb * gmx_restrict nrnb)
642 {
643 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
644 * just 0 for non-waters.
645 * Suffixes A,B,C,D refer to j loop unrolling done with SSE, e.g. for the four different
646 * jnr indices corresponding to data put in the four positions in the SIMD register.
647 */
648 int i_shift_offset,i_coord_offset,outeriter,inneriter;
649 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
650 int jnrA,jnrB,jnrC,jnrD;
651 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
652 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
653 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
654 real rcutoff_scalar;
655 real *shiftvec,*fshift,*x,*f;
656 real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
657 real scratch[4*DIM];
658 __m128 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
659 int vdwioffset0;
660 __m128 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
661 int vdwioffset1;
662 __m128 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
663 int vdwioffset2;
664 __m128 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
665 int vdwioffset3;
666 __m128 ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
667 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
668 __m128 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
669 __m128 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
670 __m128 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
671 __m128 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
672 __m128 dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
673 __m128 velec,felec,velecsum,facel,crf,krf,krf2;
674 real *charge;
675 int nvdwtype;
676 __m128 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
677 int *vdwtype;
678 real *vdwparam;
679 __m128 one_sixth = _mm_set1_ps(1.0/6.0);
680 __m128 one_twelfth = _mm_set1_ps(1.0/12.0);
681 __m128 dummy_mask,cutoff_mask;
682 __m128 signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
683 __m128 one = _mm_set1_ps(1.0);
684 __m128 two = _mm_set1_ps(2.0);
685 x = xx[0];
686 f = ff[0];
687
688 nri = nlist->nri;
689 iinr = nlist->iinr;
690 jindex = nlist->jindex;
691 jjnr = nlist->jjnr;
692 shiftidx = nlist->shift;
693 gid = nlist->gid;
694 shiftvec = fr->shift_vec[0];
695 fshift = fr->fshift[0];
696 facel = _mm_set1_ps(fr->epsfac);
697 charge = mdatoms->chargeA;
698 krf = _mm_set1_ps(fr->ic->k_rf);
699 krf2 = _mm_set1_ps(fr->ic->k_rf*2.0);
700 crf = _mm_set1_ps(fr->ic->c_rf);
701 nvdwtype = fr->ntype;
702 vdwparam = fr->nbfp;
703 vdwtype = mdatoms->typeA;
704
705 /* Setup water-specific parameters */
706 inr = nlist->iinr[0];
707 iq1 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1]));
708 iq2 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2]));
709 iq3 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+3]));
710 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
711
712 /* Avoid stupid compiler warnings */
713 jnrA = jnrB = jnrC = jnrD = 0;
714 j_coord_offsetA = 0;
715 j_coord_offsetB = 0;
716 j_coord_offsetC = 0;
717 j_coord_offsetD = 0;
718
719 outeriter = 0;
720 inneriter = 0;
721
722 for(iidx=0;iidx<4*DIM;iidx++)
723 {
724 scratch[iidx] = 0.0;
725 }
726
727 /* Start outer loop over neighborlists */
728 for(iidx=0; iidx<nri; iidx++)
729 {
730 /* Load shift vector for this list */
731 i_shift_offset = DIM*shiftidx[iidx];
732
733 /* Load limits for loop over neighbors */
734 j_index_start = jindex[iidx];
735 j_index_end = jindex[iidx+1];
736
737 /* Get outer coordinate index */
738 inr = iinr[iidx];
739 i_coord_offset = DIM*inr;
740
741 /* Load i particle coords and add shift vector */
742 gmx_mm_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
743 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
744
745 fix0 = _mm_setzero_ps();
746 fiy0 = _mm_setzero_ps();
747 fiz0 = _mm_setzero_ps();
748 fix1 = _mm_setzero_ps();
749 fiy1 = _mm_setzero_ps();
750 fiz1 = _mm_setzero_ps();
751 fix2 = _mm_setzero_ps();
752 fiy2 = _mm_setzero_ps();
753 fiz2 = _mm_setzero_ps();
754 fix3 = _mm_setzero_ps();
755 fiy3 = _mm_setzero_ps();
756 fiz3 = _mm_setzero_ps();
757
758 /* Start inner kernel loop */
759 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
760 {
761
762 /* Get j neighbor index, and coordinate index */
763 jnrA = jjnr[jidx];
764 jnrB = jjnr[jidx+1];
765 jnrC = jjnr[jidx+2];
766 jnrD = jjnr[jidx+3];
767 j_coord_offsetA = DIM*jnrA;
768 j_coord_offsetB = DIM*jnrB;
769 j_coord_offsetC = DIM*jnrC;
770 j_coord_offsetD = DIM*jnrD;
771
772 /* load j atom coordinates */
773 gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
774 x+j_coord_offsetC,x+j_coord_offsetD,
775 &jx0,&jy0,&jz0);
776
777 /* Calculate displacement vector */
778 dx00 = _mm_sub_ps(ix0,jx0);
779 dy00 = _mm_sub_ps(iy0,jy0);
780 dz00 = _mm_sub_ps(iz0,jz0);
781 dx10 = _mm_sub_ps(ix1,jx0);
782 dy10 = _mm_sub_ps(iy1,jy0);
783 dz10 = _mm_sub_ps(iz1,jz0);
784 dx20 = _mm_sub_ps(ix2,jx0);
785 dy20 = _mm_sub_ps(iy2,jy0);
786 dz20 = _mm_sub_ps(iz2,jz0);
787 dx30 = _mm_sub_ps(ix3,jx0);
788 dy30 = _mm_sub_ps(iy3,jy0);
789 dz30 = _mm_sub_ps(iz3,jz0);
790
791 /* Calculate squared distance and things based on it */
792 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
793 rsq10 = gmx_mm_calc_rsq_ps(dx10,dy10,dz10);
794 rsq20 = gmx_mm_calc_rsq_ps(dx20,dy20,dz20);
795 rsq30 = gmx_mm_calc_rsq_ps(dx30,dy30,dz30);
796
797 rinv10 = gmx_mm_invsqrt_ps(rsq10);
798 rinv20 = gmx_mm_invsqrt_ps(rsq20);
799 rinv30 = gmx_mm_invsqrt_ps(rsq30);
800
801 rinvsq00 = gmx_mm_inv_ps(rsq00);
802 rinvsq10 = _mm_mul_ps(rinv10,rinv10);
803 rinvsq20 = _mm_mul_ps(rinv20,rinv20);
804 rinvsq30 = _mm_mul_ps(rinv30,rinv30);
805
806 /* Load parameters for j particles */
807 jq0 = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
808 charge+jnrC+0,charge+jnrD+0);
809 vdwjidx0A = 2*vdwtype[jnrA+0];
810 vdwjidx0B = 2*vdwtype[jnrB+0];
811 vdwjidx0C = 2*vdwtype[jnrC+0];
812 vdwjidx0D = 2*vdwtype[jnrD+0];
813
814 fjx0 = _mm_setzero_ps();
815 fjy0 = _mm_setzero_ps();
816 fjz0 = _mm_setzero_ps();
817
818 /**************************
819 * CALCULATE INTERACTIONS *
820 **************************/
821
822 /* Compute parameters for interactions between i and j atoms */
823 gmx_mm_load_4pair_swizzle_ps(vdwparam+vdwioffset0+vdwjidx0A,
824 vdwparam+vdwioffset0+vdwjidx0B,
825 vdwparam+vdwioffset0+vdwjidx0C,
826 vdwparam+vdwioffset0+vdwjidx0D,
827 &c6_00,&c12_00);
828
829 /* LENNARD-JONES DISPERSION/REPULSION */
830
831 rinvsix = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00);
832 fvdw = _mm_mul_ps(_mm_sub_ps(_mm_mul_ps(c12_00,rinvsix),c6_00),_mm_mul_ps(rinvsix,rinvsq00));
833
834 fscal = fvdw;
835
836 /* Calculate temporary vectorial force */
837 tx = _mm_mul_ps(fscal,dx00);
838 ty = _mm_mul_ps(fscal,dy00);
839 tz = _mm_mul_ps(fscal,dz00);
840
841 /* Update vectorial force */
842 fix0 = _mm_add_ps(fix0,tx);
843 fiy0 = _mm_add_ps(fiy0,ty);
844 fiz0 = _mm_add_ps(fiz0,tz);
845
846 fjx0 = _mm_add_ps(fjx0,tx);
847 fjy0 = _mm_add_ps(fjy0,ty);
848 fjz0 = _mm_add_ps(fjz0,tz);
849
850 /**************************
851 * CALCULATE INTERACTIONS *
852 **************************/
853
854 /* Compute parameters for interactions between i and j atoms */
855 qq10 = _mm_mul_ps(iq1,jq0);
856
857 /* REACTION-FIELD ELECTROSTATICS */
858 felec = _mm_mul_ps(qq10,_mm_sub_ps(_mm_mul_ps(rinv10,rinvsq10),krf2));
859
860 fscal = felec;
861
862 /* Calculate temporary vectorial force */
863 tx = _mm_mul_ps(fscal,dx10);
864 ty = _mm_mul_ps(fscal,dy10);
865 tz = _mm_mul_ps(fscal,dz10);
866
867 /* Update vectorial force */
868 fix1 = _mm_add_ps(fix1,tx);
869 fiy1 = _mm_add_ps(fiy1,ty);
870 fiz1 = _mm_add_ps(fiz1,tz);
871
872 fjx0 = _mm_add_ps(fjx0,tx);
873 fjy0 = _mm_add_ps(fjy0,ty);
874 fjz0 = _mm_add_ps(fjz0,tz);
875
876 /**************************
877 * CALCULATE INTERACTIONS *
878 **************************/
879
880 /* Compute parameters for interactions between i and j atoms */
881 qq20 = _mm_mul_ps(iq2,jq0);
882
883 /* REACTION-FIELD ELECTROSTATICS */
884 felec = _mm_mul_ps(qq20,_mm_sub_ps(_mm_mul_ps(rinv20,rinvsq20),krf2));
885
886 fscal = felec;
887
888 /* Calculate temporary vectorial force */
889 tx = _mm_mul_ps(fscal,dx20);
890 ty = _mm_mul_ps(fscal,dy20);
891 tz = _mm_mul_ps(fscal,dz20);
892
893 /* Update vectorial force */
894 fix2 = _mm_add_ps(fix2,tx);
895 fiy2 = _mm_add_ps(fiy2,ty);
896 fiz2 = _mm_add_ps(fiz2,tz);
897
898 fjx0 = _mm_add_ps(fjx0,tx);
899 fjy0 = _mm_add_ps(fjy0,ty);
900 fjz0 = _mm_add_ps(fjz0,tz);
901
902 /**************************
903 * CALCULATE INTERACTIONS *
904 **************************/
905
906 /* Compute parameters for interactions between i and j atoms */
907 qq30 = _mm_mul_ps(iq3,jq0);
908
909 /* REACTION-FIELD ELECTROSTATICS */
910 felec = _mm_mul_ps(qq30,_mm_sub_ps(_mm_mul_ps(rinv30,rinvsq30),krf2));
911
912 fscal = felec;
913
914 /* Calculate temporary vectorial force */
915 tx = _mm_mul_ps(fscal,dx30);
916 ty = _mm_mul_ps(fscal,dy30);
917 tz = _mm_mul_ps(fscal,dz30);
918
919 /* Update vectorial force */
920 fix3 = _mm_add_ps(fix3,tx);
921 fiy3 = _mm_add_ps(fiy3,ty);
922 fiz3 = _mm_add_ps(fiz3,tz);
923
924 fjx0 = _mm_add_ps(fjx0,tx);
925 fjy0 = _mm_add_ps(fjy0,ty);
926 fjz0 = _mm_add_ps(fjz0,tz);
927
928 fjptrA = f+j_coord_offsetA;
929 fjptrB = f+j_coord_offsetB;
930 fjptrC = f+j_coord_offsetC;
931 fjptrD = f+j_coord_offsetD;
932
933 gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjx0,fjy0,fjz0);
934
935 /* Inner loop uses 108 flops */
936 }
937
938 if(jidx<j_index_end)
939 {
940
941 /* Get j neighbor index, and coordinate index */
942 jnrlistA = jjnr[jidx];
943 jnrlistB = jjnr[jidx+1];
944 jnrlistC = jjnr[jidx+2];
945 jnrlistD = jjnr[jidx+3];
946 /* Sign of each element will be negative for non-real atoms.
947 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
948 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
949 */
950 dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
951 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
952 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
953 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
954 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
955 j_coord_offsetA = DIM*jnrA;
956 j_coord_offsetB = DIM*jnrB;
957 j_coord_offsetC = DIM*jnrC;
958 j_coord_offsetD = DIM*jnrD;
959
960 /* load j atom coordinates */
961 gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
962 x+j_coord_offsetC,x+j_coord_offsetD,
963 &jx0,&jy0,&jz0);
964
965 /* Calculate displacement vector */
966 dx00 = _mm_sub_ps(ix0,jx0);
967 dy00 = _mm_sub_ps(iy0,jy0);
968 dz00 = _mm_sub_ps(iz0,jz0);
969 dx10 = _mm_sub_ps(ix1,jx0);
970 dy10 = _mm_sub_ps(iy1,jy0);
971 dz10 = _mm_sub_ps(iz1,jz0);
972 dx20 = _mm_sub_ps(ix2,jx0);
973 dy20 = _mm_sub_ps(iy2,jy0);
974 dz20 = _mm_sub_ps(iz2,jz0);
975 dx30 = _mm_sub_ps(ix3,jx0);
976 dy30 = _mm_sub_ps(iy3,jy0);
977 dz30 = _mm_sub_ps(iz3,jz0);
978
979 /* Calculate squared distance and things based on it */
980 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
981 rsq10 = gmx_mm_calc_rsq_ps(dx10,dy10,dz10);
982 rsq20 = gmx_mm_calc_rsq_ps(dx20,dy20,dz20);
983 rsq30 = gmx_mm_calc_rsq_ps(dx30,dy30,dz30);
984
985 rinv10 = gmx_mm_invsqrt_ps(rsq10);
986 rinv20 = gmx_mm_invsqrt_ps(rsq20);
987 rinv30 = gmx_mm_invsqrt_ps(rsq30);
988
989 rinvsq00 = gmx_mm_inv_ps(rsq00);
990 rinvsq10 = _mm_mul_ps(rinv10,rinv10);
991 rinvsq20 = _mm_mul_ps(rinv20,rinv20);
992 rinvsq30 = _mm_mul_ps(rinv30,rinv30);
993
994 /* Load parameters for j particles */
995 jq0 = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
996 charge+jnrC+0,charge+jnrD+0);
997 vdwjidx0A = 2*vdwtype[jnrA+0];
998 vdwjidx0B = 2*vdwtype[jnrB+0];
999 vdwjidx0C = 2*vdwtype[jnrC+0];
1000 vdwjidx0D = 2*vdwtype[jnrD+0];
1001
1002 fjx0 = _mm_setzero_ps();
1003 fjy0 = _mm_setzero_ps();
1004 fjz0 = _mm_setzero_ps();
1005
1006 /**************************
1007 * CALCULATE INTERACTIONS *
1008 **************************/
1009
1010 /* Compute parameters for interactions between i and j atoms */
1011 gmx_mm_load_4pair_swizzle_ps(vdwparam+vdwioffset0+vdwjidx0A,
1012 vdwparam+vdwioffset0+vdwjidx0B,
1013 vdwparam+vdwioffset0+vdwjidx0C,
1014 vdwparam+vdwioffset0+vdwjidx0D,
1015 &c6_00,&c12_00);
1016
1017 /* LENNARD-JONES DISPERSION/REPULSION */
1018
1019 rinvsix = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00);
1020 fvdw = _mm_mul_ps(_mm_sub_ps(_mm_mul_ps(c12_00,rinvsix),c6_00),_mm_mul_ps(rinvsix,rinvsq00));
1021
1022 fscal = fvdw;
1023
1024 fscal = _mm_andnot_ps(dummy_mask,fscal);
1025
1026 /* Calculate temporary vectorial force */
1027 tx = _mm_mul_ps(fscal,dx00);
1028 ty = _mm_mul_ps(fscal,dy00);
1029 tz = _mm_mul_ps(fscal,dz00);
1030
1031 /* Update vectorial force */
1032 fix0 = _mm_add_ps(fix0,tx);
1033 fiy0 = _mm_add_ps(fiy0,ty);
1034 fiz0 = _mm_add_ps(fiz0,tz);
1035
1036 fjx0 = _mm_add_ps(fjx0,tx);
1037 fjy0 = _mm_add_ps(fjy0,ty);
1038 fjz0 = _mm_add_ps(fjz0,tz);
1039
1040 /**************************
1041 * CALCULATE INTERACTIONS *
1042 **************************/
1043
1044 /* Compute parameters for interactions between i and j atoms */
1045 qq10 = _mm_mul_ps(iq1,jq0);
1046
1047 /* REACTION-FIELD ELECTROSTATICS */
1048 felec = _mm_mul_ps(qq10,_mm_sub_ps(_mm_mul_ps(rinv10,rinvsq10),krf2));
1049
1050 fscal = felec;
1051
1052 fscal = _mm_andnot_ps(dummy_mask,fscal);
1053
1054 /* Calculate temporary vectorial force */
1055 tx = _mm_mul_ps(fscal,dx10);
1056 ty = _mm_mul_ps(fscal,dy10);
1057 tz = _mm_mul_ps(fscal,dz10);
1058
1059 /* Update vectorial force */
1060 fix1 = _mm_add_ps(fix1,tx);
1061 fiy1 = _mm_add_ps(fiy1,ty);
1062 fiz1 = _mm_add_ps(fiz1,tz);
1063
1064 fjx0 = _mm_add_ps(fjx0,tx);
1065 fjy0 = _mm_add_ps(fjy0,ty);
1066 fjz0 = _mm_add_ps(fjz0,tz);
1067
1068 /**************************
1069 * CALCULATE INTERACTIONS *
1070 **************************/
1071
1072 /* Compute parameters for interactions between i and j atoms */
1073 qq20 = _mm_mul_ps(iq2,jq0);
1074
1075 /* REACTION-FIELD ELECTROSTATICS */
1076 felec = _mm_mul_ps(qq20,_mm_sub_ps(_mm_mul_ps(rinv20,rinvsq20),krf2));
1077
1078 fscal = felec;
1079
1080 fscal = _mm_andnot_ps(dummy_mask,fscal);
1081
1082 /* Calculate temporary vectorial force */
1083 tx = _mm_mul_ps(fscal,dx20);
1084 ty = _mm_mul_ps(fscal,dy20);
1085 tz = _mm_mul_ps(fscal,dz20);
1086
1087 /* Update vectorial force */
1088 fix2 = _mm_add_ps(fix2,tx);
1089 fiy2 = _mm_add_ps(fiy2,ty);
1090 fiz2 = _mm_add_ps(fiz2,tz);
1091
1092 fjx0 = _mm_add_ps(fjx0,tx);
1093 fjy0 = _mm_add_ps(fjy0,ty);
1094 fjz0 = _mm_add_ps(fjz0,tz);
1095
1096 /**************************
1097 * CALCULATE INTERACTIONS *
1098 **************************/
1099
1100 /* Compute parameters for interactions between i and j atoms */
1101 qq30 = _mm_mul_ps(iq3,jq0);
1102
1103 /* REACTION-FIELD ELECTROSTATICS */
1104 felec = _mm_mul_ps(qq30,_mm_sub_ps(_mm_mul_ps(rinv30,rinvsq30),krf2));
1105
1106 fscal = felec;
1107
1108 fscal = _mm_andnot_ps(dummy_mask,fscal);
1109
1110 /* Calculate temporary vectorial force */
1111 tx = _mm_mul_ps(fscal,dx30);
1112 ty = _mm_mul_ps(fscal,dy30);
1113 tz = _mm_mul_ps(fscal,dz30);
1114
1115 /* Update vectorial force */
1116 fix3 = _mm_add_ps(fix3,tx);
1117 fiy3 = _mm_add_ps(fiy3,ty);
1118 fiz3 = _mm_add_ps(fiz3,tz);
1119
1120 fjx0 = _mm_add_ps(fjx0,tx);
1121 fjy0 = _mm_add_ps(fjy0,ty);
1122 fjz0 = _mm_add_ps(fjz0,tz);
1123
1124 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1125 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1126 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1127 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1128
1129 gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjx0,fjy0,fjz0);
1130
1131 /* Inner loop uses 108 flops */
1132 }
1133
1134 /* End of innermost loop */
1135
1136 gmx_mm_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1137 f+i_coord_offset,fshift+i_shift_offset);
1138
1139 /* Increment number of inner iterations */
1140 inneriter += j_index_end - j_index_start;
1141
1142 /* Outer loop uses 24 flops */
1143 }
1144
1145 /* Increment number of outer iterations */
1146 outeriter += nri;
1147
1148 /* Update outer/inner flops */
1149
1150 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_F,outeriter*24 + inneriter*108);
1151 }
The ultimate molecular dynamics simulation package