?  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_sse4_1_single / nb_kernel_ElecEwSw_VdwNone_GeomW3P1_sse4_1_single.cpp
blobbe98bf3abce677ee2059e6871cfd27ee27fe4b16
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 sse4_1_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/gmxlib/nrnb.h"
46
47 #include "kernelutil_x86_sse4_1_single.h"
48
49 /*
50 * Gromacs nonbonded kernel: nb_kernel_ElecEwSw_VdwNone_GeomW3P1_VF_sse4_1_single
51 * Electrostatics interaction: Ewald
52 * VdW interaction: None
53 * Geometry: Water3-Particle
54 * Calculate force/pot: PotentialAndForce
55 */
56 void
57 nb_kernel_ElecEwSw_VdwNone_GeomW3P1_VF_sse4_1_single
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,C,D refer to j loop unrolling done with SSE, e.g. for the four 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,jnrC,jnrD;
74 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
75 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
76 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
77 real rcutoff_scalar;
78 real *shiftvec,*fshift,*x,*f;
79 real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
80 real scratch[4*DIM];
81 __m128 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
82 int vdwioffset0;
83 __m128 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
84 int vdwioffset1;
85 __m128 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
86 int vdwioffset2;
87 __m128 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
88 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
89 __m128 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
90 __m128 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
91 __m128 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
92 __m128 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
93 __m128 velec,felec,velecsum,facel,crf,krf,krf2;
94 real *charge;
95 __m128i ewitab;
96 __m128 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
97 real *ewtab;
98 __m128 rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
99 real rswitch_scalar,d_scalar;
100 __m128 dummy_mask,cutoff_mask;
101 __m128 signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
102 __m128 one = _mm_set1_ps(1.0);
103 __m128 two = _mm_set1_ps(2.0);
104 x = xx[0];
105 f = ff[0];
106
107 nri = nlist->nri;
108 iinr = nlist->iinr;
109 jindex = nlist->jindex;
110 jjnr = nlist->jjnr;
111 shiftidx = nlist->shift;
112 gid = nlist->gid;
113 shiftvec = fr->shift_vec[0];
114 fshift = fr->fshift[0];
115 facel = _mm_set1_ps(fr->ic->epsfac);
116 charge = mdatoms->chargeA;
117
118 sh_ewald = _mm_set1_ps(fr->ic->sh_ewald);
119 ewtab = fr->ic->tabq_coul_FDV0;
120 ewtabscale = _mm_set1_ps(fr->ic->tabq_scale);
121 ewtabhalfspace = _mm_set1_ps(0.5/fr->ic->tabq_scale);
122
123 /* Setup water-specific parameters */
124 inr = nlist->iinr[0];
125 iq0 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+0]));
126 iq1 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1]));
127 iq2 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2]));
128
129 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
130 rcutoff_scalar = fr->ic->rcoulomb;
131 rcutoff = _mm_set1_ps(rcutoff_scalar);
132 rcutoff2 = _mm_mul_ps(rcutoff,rcutoff);
133
134 rswitch_scalar = fr->ic->rcoulomb_switch;
135 rswitch = _mm_set1_ps(rswitch_scalar);
136 /* Setup switch parameters */
137 d_scalar = rcutoff_scalar-rswitch_scalar;
138 d = _mm_set1_ps(d_scalar);
139 swV3 = _mm_set1_ps(-10.0/(d_scalar*d_scalar*d_scalar));
140 swV4 = _mm_set1_ps( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
141 swV5 = _mm_set1_ps( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
142 swF2 = _mm_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar));
143 swF3 = _mm_set1_ps( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
144 swF4 = _mm_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
145
146 /* Avoid stupid compiler warnings */
147 jnrA = jnrB = jnrC = jnrD = 0;
148 j_coord_offsetA = 0;
149 j_coord_offsetB = 0;
150 j_coord_offsetC = 0;
151 j_coord_offsetD = 0;
152
153 outeriter = 0;
154 inneriter = 0;
155
156 for(iidx=0;iidx<4*DIM;iidx++)
157 {
158 scratch[iidx] = 0.0;
159 }
160
161 /* Start outer loop over neighborlists */
162 for(iidx=0; iidx<nri; iidx++)
163 {
164 /* Load shift vector for this list */
165 i_shift_offset = DIM*shiftidx[iidx];
166
167 /* Load limits for loop over neighbors */
168 j_index_start = jindex[iidx];
169 j_index_end = jindex[iidx+1];
170
171 /* Get outer coordinate index */
172 inr = iinr[iidx];
173 i_coord_offset = DIM*inr;
174
175 /* Load i particle coords and add shift vector */
176 gmx_mm_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
177 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
178
179 fix0 = _mm_setzero_ps();
180 fiy0 = _mm_setzero_ps();
181 fiz0 = _mm_setzero_ps();
182 fix1 = _mm_setzero_ps();
183 fiy1 = _mm_setzero_ps();
184 fiz1 = _mm_setzero_ps();
185 fix2 = _mm_setzero_ps();
186 fiy2 = _mm_setzero_ps();
187 fiz2 = _mm_setzero_ps();
188
189 /* Reset potential sums */
190 velecsum = _mm_setzero_ps();
191
192 /* Start inner kernel loop */
193 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
194 {
195
196 /* Get j neighbor index, and coordinate index */
197 jnrA = jjnr[jidx];
198 jnrB = jjnr[jidx+1];
199 jnrC = jjnr[jidx+2];
200 jnrD = jjnr[jidx+3];
201 j_coord_offsetA = DIM*jnrA;
202 j_coord_offsetB = DIM*jnrB;
203 j_coord_offsetC = DIM*jnrC;
204 j_coord_offsetD = DIM*jnrD;
205
206 /* load j atom coordinates */
207 gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
208 x+j_coord_offsetC,x+j_coord_offsetD,
209 &jx0,&jy0,&jz0);
210
211 /* Calculate displacement vector */
212 dx00 = _mm_sub_ps(ix0,jx0);
213 dy00 = _mm_sub_ps(iy0,jy0);
214 dz00 = _mm_sub_ps(iz0,jz0);
215 dx10 = _mm_sub_ps(ix1,jx0);
216 dy10 = _mm_sub_ps(iy1,jy0);
217 dz10 = _mm_sub_ps(iz1,jz0);
218 dx20 = _mm_sub_ps(ix2,jx0);
219 dy20 = _mm_sub_ps(iy2,jy0);
220 dz20 = _mm_sub_ps(iz2,jz0);
221
222 /* Calculate squared distance and things based on it */
223 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
224 rsq10 = gmx_mm_calc_rsq_ps(dx10,dy10,dz10);
225 rsq20 = gmx_mm_calc_rsq_ps(dx20,dy20,dz20);
226
227 rinv00 = sse41_invsqrt_f(rsq00);
228 rinv10 = sse41_invsqrt_f(rsq10);
229 rinv20 = sse41_invsqrt_f(rsq20);
230
231 rinvsq00 = _mm_mul_ps(rinv00,rinv00);
232 rinvsq10 = _mm_mul_ps(rinv10,rinv10);
233 rinvsq20 = _mm_mul_ps(rinv20,rinv20);
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
239 fjx0 = _mm_setzero_ps();
240 fjy0 = _mm_setzero_ps();
241 fjz0 = _mm_setzero_ps();
242
243 /**************************
244 * CALCULATE INTERACTIONS *
245 **************************/
246
247 if (gmx_mm_any_lt(rsq00,rcutoff2))
248 {
249
250 r00 = _mm_mul_ps(rsq00,rinv00);
251
252 /* Compute parameters for interactions between i and j atoms */
253 qq00 = _mm_mul_ps(iq0,jq0);
254
255 /* EWALD ELECTROSTATICS */
256
257 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
258 ewrt = _mm_mul_ps(r00,ewtabscale);
259 ewitab = _mm_cvttps_epi32(ewrt);
260 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
261 ewitab = _mm_slli_epi32(ewitab,2);
262 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
263 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
264 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
265 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
266 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
267 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
268 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
269 velec = _mm_mul_ps(qq00,_mm_sub_ps(rinv00,velec));
270 felec = _mm_mul_ps(_mm_mul_ps(qq00,rinv00),_mm_sub_ps(rinvsq00,felec));
271
272 d = _mm_sub_ps(r00,rswitch);
273 d = _mm_max_ps(d,_mm_setzero_ps());
274 d2 = _mm_mul_ps(d,d);
275 sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5)))))));
276
277 dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4)))));
278
279 /* Evaluate switch function */
280 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
281 felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv00,_mm_mul_ps(velec,dsw)) );
282 velec = _mm_mul_ps(velec,sw);
283 cutoff_mask = _mm_cmplt_ps(rsq00,rcutoff2);
284
285 /* Update potential sum for this i atom from the interaction with this j atom. */
286 velec = _mm_and_ps(velec,cutoff_mask);
287 velecsum = _mm_add_ps(velecsum,velec);
288
289 fscal = felec;
290
291 fscal = _mm_and_ps(fscal,cutoff_mask);
292
293 /* Calculate temporary vectorial force */
294 tx = _mm_mul_ps(fscal,dx00);
295 ty = _mm_mul_ps(fscal,dy00);
296 tz = _mm_mul_ps(fscal,dz00);
297
298 /* Update vectorial force */
299 fix0 = _mm_add_ps(fix0,tx);
300 fiy0 = _mm_add_ps(fiy0,ty);
301 fiz0 = _mm_add_ps(fiz0,tz);
302
303 fjx0 = _mm_add_ps(fjx0,tx);
304 fjy0 = _mm_add_ps(fjy0,ty);
305 fjz0 = _mm_add_ps(fjz0,tz);
306
307 }
308
309 /**************************
310 * CALCULATE INTERACTIONS *
311 **************************/
312
313 if (gmx_mm_any_lt(rsq10,rcutoff2))
314 {
315
316 r10 = _mm_mul_ps(rsq10,rinv10);
317
318 /* Compute parameters for interactions between i and j atoms */
319 qq10 = _mm_mul_ps(iq1,jq0);
320
321 /* EWALD ELECTROSTATICS */
322
323 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
324 ewrt = _mm_mul_ps(r10,ewtabscale);
325 ewitab = _mm_cvttps_epi32(ewrt);
326 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
327 ewitab = _mm_slli_epi32(ewitab,2);
328 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
329 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
330 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
331 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
332 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
333 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
334 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
335 velec = _mm_mul_ps(qq10,_mm_sub_ps(rinv10,velec));
336 felec = _mm_mul_ps(_mm_mul_ps(qq10,rinv10),_mm_sub_ps(rinvsq10,felec));
337
338 d = _mm_sub_ps(r10,rswitch);
339 d = _mm_max_ps(d,_mm_setzero_ps());
340 d2 = _mm_mul_ps(d,d);
341 sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5)))))));
342
343 dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4)))));
344
345 /* Evaluate switch function */
346 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
347 felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv10,_mm_mul_ps(velec,dsw)) );
348 velec = _mm_mul_ps(velec,sw);
349 cutoff_mask = _mm_cmplt_ps(rsq10,rcutoff2);
350
351 /* Update potential sum for this i atom from the interaction with this j atom. */
352 velec = _mm_and_ps(velec,cutoff_mask);
353 velecsum = _mm_add_ps(velecsum,velec);
354
355 fscal = felec;
356
357 fscal = _mm_and_ps(fscal,cutoff_mask);
358
359 /* Calculate temporary vectorial force */
360 tx = _mm_mul_ps(fscal,dx10);
361 ty = _mm_mul_ps(fscal,dy10);
362 tz = _mm_mul_ps(fscal,dz10);
363
364 /* Update vectorial force */
365 fix1 = _mm_add_ps(fix1,tx);
366 fiy1 = _mm_add_ps(fiy1,ty);
367 fiz1 = _mm_add_ps(fiz1,tz);
368
369 fjx0 = _mm_add_ps(fjx0,tx);
370 fjy0 = _mm_add_ps(fjy0,ty);
371 fjz0 = _mm_add_ps(fjz0,tz);
372
373 }
374
375 /**************************
376 * CALCULATE INTERACTIONS *
377 **************************/
378
379 if (gmx_mm_any_lt(rsq20,rcutoff2))
380 {
381
382 r20 = _mm_mul_ps(rsq20,rinv20);
383
384 /* Compute parameters for interactions between i and j atoms */
385 qq20 = _mm_mul_ps(iq2,jq0);
386
387 /* EWALD ELECTROSTATICS */
388
389 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
390 ewrt = _mm_mul_ps(r20,ewtabscale);
391 ewitab = _mm_cvttps_epi32(ewrt);
392 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
393 ewitab = _mm_slli_epi32(ewitab,2);
394 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
395 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
396 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
397 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
398 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
399 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
400 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
401 velec = _mm_mul_ps(qq20,_mm_sub_ps(rinv20,velec));
402 felec = _mm_mul_ps(_mm_mul_ps(qq20,rinv20),_mm_sub_ps(rinvsq20,felec));
403
404 d = _mm_sub_ps(r20,rswitch);
405 d = _mm_max_ps(d,_mm_setzero_ps());
406 d2 = _mm_mul_ps(d,d);
407 sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5)))))));
408
409 dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4)))));
410
411 /* Evaluate switch function */
412 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
413 felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv20,_mm_mul_ps(velec,dsw)) );
414 velec = _mm_mul_ps(velec,sw);
415 cutoff_mask = _mm_cmplt_ps(rsq20,rcutoff2);
416
417 /* Update potential sum for this i atom from the interaction with this j atom. */
418 velec = _mm_and_ps(velec,cutoff_mask);
419 velecsum = _mm_add_ps(velecsum,velec);
420
421 fscal = felec;
422
423 fscal = _mm_and_ps(fscal,cutoff_mask);
424
425 /* Calculate temporary vectorial force */
426 tx = _mm_mul_ps(fscal,dx20);
427 ty = _mm_mul_ps(fscal,dy20);
428 tz = _mm_mul_ps(fscal,dz20);
429
430 /* Update vectorial force */
431 fix2 = _mm_add_ps(fix2,tx);
432 fiy2 = _mm_add_ps(fiy2,ty);
433 fiz2 = _mm_add_ps(fiz2,tz);
434
435 fjx0 = _mm_add_ps(fjx0,tx);
436 fjy0 = _mm_add_ps(fjy0,ty);
437 fjz0 = _mm_add_ps(fjz0,tz);
438
439 }
440
441 fjptrA = f+j_coord_offsetA;
442 fjptrB = f+j_coord_offsetB;
443 fjptrC = f+j_coord_offsetC;
444 fjptrD = f+j_coord_offsetD;
445
446 gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjx0,fjy0,fjz0);
447
448 /* Inner loop uses 195 flops */
449 }
450
451 if(jidx<j_index_end)
452 {
453
454 /* Get j neighbor index, and coordinate index */
455 jnrlistA = jjnr[jidx];
456 jnrlistB = jjnr[jidx+1];
457 jnrlistC = jjnr[jidx+2];
458 jnrlistD = jjnr[jidx+3];
459 /* Sign of each element will be negative for non-real atoms.
460 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
461 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
462 */
463 dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
464 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
465 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
466 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
467 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
468 j_coord_offsetA = DIM*jnrA;
469 j_coord_offsetB = DIM*jnrB;
470 j_coord_offsetC = DIM*jnrC;
471 j_coord_offsetD = DIM*jnrD;
472
473 /* load j atom coordinates */
474 gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
475 x+j_coord_offsetC,x+j_coord_offsetD,
476 &jx0,&jy0,&jz0);
477
478 /* Calculate displacement vector */
479 dx00 = _mm_sub_ps(ix0,jx0);
480 dy00 = _mm_sub_ps(iy0,jy0);
481 dz00 = _mm_sub_ps(iz0,jz0);
482 dx10 = _mm_sub_ps(ix1,jx0);
483 dy10 = _mm_sub_ps(iy1,jy0);
484 dz10 = _mm_sub_ps(iz1,jz0);
485 dx20 = _mm_sub_ps(ix2,jx0);
486 dy20 = _mm_sub_ps(iy2,jy0);
487 dz20 = _mm_sub_ps(iz2,jz0);
488
489 /* Calculate squared distance and things based on it */
490 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
491 rsq10 = gmx_mm_calc_rsq_ps(dx10,dy10,dz10);
492 rsq20 = gmx_mm_calc_rsq_ps(dx20,dy20,dz20);
493
494 rinv00 = sse41_invsqrt_f(rsq00);
495 rinv10 = sse41_invsqrt_f(rsq10);
496 rinv20 = sse41_invsqrt_f(rsq20);
497
498 rinvsq00 = _mm_mul_ps(rinv00,rinv00);
499 rinvsq10 = _mm_mul_ps(rinv10,rinv10);
500 rinvsq20 = _mm_mul_ps(rinv20,rinv20);
501
502 /* Load parameters for j particles */
503 jq0 = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
504 charge+jnrC+0,charge+jnrD+0);
505
506 fjx0 = _mm_setzero_ps();
507 fjy0 = _mm_setzero_ps();
508 fjz0 = _mm_setzero_ps();
509
510 /**************************
511 * CALCULATE INTERACTIONS *
512 **************************/
513
514 if (gmx_mm_any_lt(rsq00,rcutoff2))
515 {
516
517 r00 = _mm_mul_ps(rsq00,rinv00);
518 r00 = _mm_andnot_ps(dummy_mask,r00);
519
520 /* Compute parameters for interactions between i and j atoms */
521 qq00 = _mm_mul_ps(iq0,jq0);
522
523 /* EWALD ELECTROSTATICS */
524
525 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
526 ewrt = _mm_mul_ps(r00,ewtabscale);
527 ewitab = _mm_cvttps_epi32(ewrt);
528 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
529 ewitab = _mm_slli_epi32(ewitab,2);
530 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
531 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
532 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
533 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
534 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
535 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
536 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
537 velec = _mm_mul_ps(qq00,_mm_sub_ps(rinv00,velec));
538 felec = _mm_mul_ps(_mm_mul_ps(qq00,rinv00),_mm_sub_ps(rinvsq00,felec));
539
540 d = _mm_sub_ps(r00,rswitch);
541 d = _mm_max_ps(d,_mm_setzero_ps());
542 d2 = _mm_mul_ps(d,d);
543 sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5)))))));
544
545 dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4)))));
546
547 /* Evaluate switch function */
548 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
549 felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv00,_mm_mul_ps(velec,dsw)) );
550 velec = _mm_mul_ps(velec,sw);
551 cutoff_mask = _mm_cmplt_ps(rsq00,rcutoff2);
552
553 /* Update potential sum for this i atom from the interaction with this j atom. */
554 velec = _mm_and_ps(velec,cutoff_mask);
555 velec = _mm_andnot_ps(dummy_mask,velec);
556 velecsum = _mm_add_ps(velecsum,velec);
557
558 fscal = felec;
559
560 fscal = _mm_and_ps(fscal,cutoff_mask);
561
562 fscal = _mm_andnot_ps(dummy_mask,fscal);
563
564 /* Calculate temporary vectorial force */
565 tx = _mm_mul_ps(fscal,dx00);
566 ty = _mm_mul_ps(fscal,dy00);
567 tz = _mm_mul_ps(fscal,dz00);
568
569 /* Update vectorial force */
570 fix0 = _mm_add_ps(fix0,tx);
571 fiy0 = _mm_add_ps(fiy0,ty);
572 fiz0 = _mm_add_ps(fiz0,tz);
573
574 fjx0 = _mm_add_ps(fjx0,tx);
575 fjy0 = _mm_add_ps(fjy0,ty);
576 fjz0 = _mm_add_ps(fjz0,tz);
577
578 }
579
580 /**************************
581 * CALCULATE INTERACTIONS *
582 **************************/
583
584 if (gmx_mm_any_lt(rsq10,rcutoff2))
585 {
586
587 r10 = _mm_mul_ps(rsq10,rinv10);
588 r10 = _mm_andnot_ps(dummy_mask,r10);
589
590 /* Compute parameters for interactions between i and j atoms */
591 qq10 = _mm_mul_ps(iq1,jq0);
592
593 /* EWALD ELECTROSTATICS */
594
595 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
596 ewrt = _mm_mul_ps(r10,ewtabscale);
597 ewitab = _mm_cvttps_epi32(ewrt);
598 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
599 ewitab = _mm_slli_epi32(ewitab,2);
600 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
601 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
602 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
603 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
604 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
605 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
606 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
607 velec = _mm_mul_ps(qq10,_mm_sub_ps(rinv10,velec));
608 felec = _mm_mul_ps(_mm_mul_ps(qq10,rinv10),_mm_sub_ps(rinvsq10,felec));
609
610 d = _mm_sub_ps(r10,rswitch);
611 d = _mm_max_ps(d,_mm_setzero_ps());
612 d2 = _mm_mul_ps(d,d);
613 sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5)))))));
614
615 dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4)))));
616
617 /* Evaluate switch function */
618 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
619 felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv10,_mm_mul_ps(velec,dsw)) );
620 velec = _mm_mul_ps(velec,sw);
621 cutoff_mask = _mm_cmplt_ps(rsq10,rcutoff2);
622
623 /* Update potential sum for this i atom from the interaction with this j atom. */
624 velec = _mm_and_ps(velec,cutoff_mask);
625 velec = _mm_andnot_ps(dummy_mask,velec);
626 velecsum = _mm_add_ps(velecsum,velec);
627
628 fscal = felec;
629
630 fscal = _mm_and_ps(fscal,cutoff_mask);
631
632 fscal = _mm_andnot_ps(dummy_mask,fscal);
633
634 /* Calculate temporary vectorial force */
635 tx = _mm_mul_ps(fscal,dx10);
636 ty = _mm_mul_ps(fscal,dy10);
637 tz = _mm_mul_ps(fscal,dz10);
638
639 /* Update vectorial force */
640 fix1 = _mm_add_ps(fix1,tx);
641 fiy1 = _mm_add_ps(fiy1,ty);
642 fiz1 = _mm_add_ps(fiz1,tz);
643
644 fjx0 = _mm_add_ps(fjx0,tx);
645 fjy0 = _mm_add_ps(fjy0,ty);
646 fjz0 = _mm_add_ps(fjz0,tz);
647
648 }
649
650 /**************************
651 * CALCULATE INTERACTIONS *
652 **************************/
653
654 if (gmx_mm_any_lt(rsq20,rcutoff2))
655 {
656
657 r20 = _mm_mul_ps(rsq20,rinv20);
658 r20 = _mm_andnot_ps(dummy_mask,r20);
659
660 /* Compute parameters for interactions between i and j atoms */
661 qq20 = _mm_mul_ps(iq2,jq0);
662
663 /* EWALD ELECTROSTATICS */
664
665 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
666 ewrt = _mm_mul_ps(r20,ewtabscale);
667 ewitab = _mm_cvttps_epi32(ewrt);
668 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
669 ewitab = _mm_slli_epi32(ewitab,2);
670 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
671 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
672 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
673 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
674 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
675 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
676 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
677 velec = _mm_mul_ps(qq20,_mm_sub_ps(rinv20,velec));
678 felec = _mm_mul_ps(_mm_mul_ps(qq20,rinv20),_mm_sub_ps(rinvsq20,felec));
679
680 d = _mm_sub_ps(r20,rswitch);
681 d = _mm_max_ps(d,_mm_setzero_ps());
682 d2 = _mm_mul_ps(d,d);
683 sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5)))))));
684
685 dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4)))));
686
687 /* Evaluate switch function */
688 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
689 felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv20,_mm_mul_ps(velec,dsw)) );
690 velec = _mm_mul_ps(velec,sw);
691 cutoff_mask = _mm_cmplt_ps(rsq20,rcutoff2);
692
693 /* Update potential sum for this i atom from the interaction with this j atom. */
694 velec = _mm_and_ps(velec,cutoff_mask);
695 velec = _mm_andnot_ps(dummy_mask,velec);
696 velecsum = _mm_add_ps(velecsum,velec);
697
698 fscal = felec;
699
700 fscal = _mm_and_ps(fscal,cutoff_mask);
701
702 fscal = _mm_andnot_ps(dummy_mask,fscal);
703
704 /* Calculate temporary vectorial force */
705 tx = _mm_mul_ps(fscal,dx20);
706 ty = _mm_mul_ps(fscal,dy20);
707 tz = _mm_mul_ps(fscal,dz20);
708
709 /* Update vectorial force */
710 fix2 = _mm_add_ps(fix2,tx);
711 fiy2 = _mm_add_ps(fiy2,ty);
712 fiz2 = _mm_add_ps(fiz2,tz);
713
714 fjx0 = _mm_add_ps(fjx0,tx);
715 fjy0 = _mm_add_ps(fjy0,ty);
716 fjz0 = _mm_add_ps(fjz0,tz);
717
718 }
719
720 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
721 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
722 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
723 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
724
725 gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjx0,fjy0,fjz0);
726
727 /* Inner loop uses 198 flops */
728 }
729
730 /* End of innermost loop */
731
732 gmx_mm_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
733 f+i_coord_offset,fshift+i_shift_offset);
734
735 ggid = gid[iidx];
736 /* Update potential energies */
737 gmx_mm_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
738
739 /* Increment number of inner iterations */
740 inneriter += j_index_end - j_index_start;
741
742 /* Outer loop uses 19 flops */
743 }
744
745 /* Increment number of outer iterations */
746 outeriter += nri;
747
748 /* Update outer/inner flops */
749
750 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3_VF,outeriter*19 + inneriter*198);
751 }
752 /*
753 * Gromacs nonbonded kernel: nb_kernel_ElecEwSw_VdwNone_GeomW3P1_F_sse4_1_single
754 * Electrostatics interaction: Ewald
755 * VdW interaction: None
756 * Geometry: Water3-Particle
757 * Calculate force/pot: Force
758 */
759 void
760 nb_kernel_ElecEwSw_VdwNone_GeomW3P1_F_sse4_1_single
761 (t_nblist * gmx_restrict nlist,
762 rvec * gmx_restrict xx,
763 rvec * gmx_restrict ff,
764 struct t_forcerec * gmx_restrict fr,
765 t_mdatoms * gmx_restrict mdatoms,
766 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
767 t_nrnb * gmx_restrict nrnb)
768 {
769 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
770 * just 0 for non-waters.
771 * Suffixes A,B,C,D refer to j loop unrolling done with SSE, e.g. for the four different
772 * jnr indices corresponding to data put in the four positions in the SIMD register.
773 */
774 int i_shift_offset,i_coord_offset,outeriter,inneriter;
775 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
776 int jnrA,jnrB,jnrC,jnrD;
777 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
778 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
779 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
780 real rcutoff_scalar;
781 real *shiftvec,*fshift,*x,*f;
782 real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
783 real scratch[4*DIM];
784 __m128 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
785 int vdwioffset0;
786 __m128 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
787 int vdwioffset1;
788 __m128 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
789 int vdwioffset2;
790 __m128 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
791 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
792 __m128 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
793 __m128 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
794 __m128 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
795 __m128 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
796 __m128 velec,felec,velecsum,facel,crf,krf,krf2;
797 real *charge;
798 __m128i ewitab;
799 __m128 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
800 real *ewtab;
801 __m128 rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
802 real rswitch_scalar,d_scalar;
803 __m128 dummy_mask,cutoff_mask;
804 __m128 signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
805 __m128 one = _mm_set1_ps(1.0);
806 __m128 two = _mm_set1_ps(2.0);
807 x = xx[0];
808 f = ff[0];
809
810 nri = nlist->nri;
811 iinr = nlist->iinr;
812 jindex = nlist->jindex;
813 jjnr = nlist->jjnr;
814 shiftidx = nlist->shift;
815 gid = nlist->gid;
816 shiftvec = fr->shift_vec[0];
817 fshift = fr->fshift[0];
818 facel = _mm_set1_ps(fr->ic->epsfac);
819 charge = mdatoms->chargeA;
820
821 sh_ewald = _mm_set1_ps(fr->ic->sh_ewald);
822 ewtab = fr->ic->tabq_coul_FDV0;
823 ewtabscale = _mm_set1_ps(fr->ic->tabq_scale);
824 ewtabhalfspace = _mm_set1_ps(0.5/fr->ic->tabq_scale);
825
826 /* Setup water-specific parameters */
827 inr = nlist->iinr[0];
828 iq0 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+0]));
829 iq1 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1]));
830 iq2 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2]));
831
832 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
833 rcutoff_scalar = fr->ic->rcoulomb;
834 rcutoff = _mm_set1_ps(rcutoff_scalar);
835 rcutoff2 = _mm_mul_ps(rcutoff,rcutoff);
836
837 rswitch_scalar = fr->ic->rcoulomb_switch;
838 rswitch = _mm_set1_ps(rswitch_scalar);
839 /* Setup switch parameters */
840 d_scalar = rcutoff_scalar-rswitch_scalar;
841 d = _mm_set1_ps(d_scalar);
842 swV3 = _mm_set1_ps(-10.0/(d_scalar*d_scalar*d_scalar));
843 swV4 = _mm_set1_ps( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
844 swV5 = _mm_set1_ps( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
845 swF2 = _mm_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar));
846 swF3 = _mm_set1_ps( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
847 swF4 = _mm_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
848
849 /* Avoid stupid compiler warnings */
850 jnrA = jnrB = jnrC = jnrD = 0;
851 j_coord_offsetA = 0;
852 j_coord_offsetB = 0;
853 j_coord_offsetC = 0;
854 j_coord_offsetD = 0;
855
856 outeriter = 0;
857 inneriter = 0;
858
859 for(iidx=0;iidx<4*DIM;iidx++)
860 {
861 scratch[iidx] = 0.0;
862 }
863
864 /* Start outer loop over neighborlists */
865 for(iidx=0; iidx<nri; iidx++)
866 {
867 /* Load shift vector for this list */
868 i_shift_offset = DIM*shiftidx[iidx];
869
870 /* Load limits for loop over neighbors */
871 j_index_start = jindex[iidx];
872 j_index_end = jindex[iidx+1];
873
874 /* Get outer coordinate index */
875 inr = iinr[iidx];
876 i_coord_offset = DIM*inr;
877
878 /* Load i particle coords and add shift vector */
879 gmx_mm_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
880 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
881
882 fix0 = _mm_setzero_ps();
883 fiy0 = _mm_setzero_ps();
884 fiz0 = _mm_setzero_ps();
885 fix1 = _mm_setzero_ps();
886 fiy1 = _mm_setzero_ps();
887 fiz1 = _mm_setzero_ps();
888 fix2 = _mm_setzero_ps();
889 fiy2 = _mm_setzero_ps();
890 fiz2 = _mm_setzero_ps();
891
892 /* Start inner kernel loop */
893 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
894 {
895
896 /* Get j neighbor index, and coordinate index */
897 jnrA = jjnr[jidx];
898 jnrB = jjnr[jidx+1];
899 jnrC = jjnr[jidx+2];
900 jnrD = jjnr[jidx+3];
901 j_coord_offsetA = DIM*jnrA;
902 j_coord_offsetB = DIM*jnrB;
903 j_coord_offsetC = DIM*jnrC;
904 j_coord_offsetD = DIM*jnrD;
905
906 /* load j atom coordinates */
907 gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
908 x+j_coord_offsetC,x+j_coord_offsetD,
909 &jx0,&jy0,&jz0);
910
911 /* Calculate displacement vector */
912 dx00 = _mm_sub_ps(ix0,jx0);
913 dy00 = _mm_sub_ps(iy0,jy0);
914 dz00 = _mm_sub_ps(iz0,jz0);
915 dx10 = _mm_sub_ps(ix1,jx0);
916 dy10 = _mm_sub_ps(iy1,jy0);
917 dz10 = _mm_sub_ps(iz1,jz0);
918 dx20 = _mm_sub_ps(ix2,jx0);
919 dy20 = _mm_sub_ps(iy2,jy0);
920 dz20 = _mm_sub_ps(iz2,jz0);
921
922 /* Calculate squared distance and things based on it */
923 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
924 rsq10 = gmx_mm_calc_rsq_ps(dx10,dy10,dz10);
925 rsq20 = gmx_mm_calc_rsq_ps(dx20,dy20,dz20);
926
927 rinv00 = sse41_invsqrt_f(rsq00);
928 rinv10 = sse41_invsqrt_f(rsq10);
929 rinv20 = sse41_invsqrt_f(rsq20);
930
931 rinvsq00 = _mm_mul_ps(rinv00,rinv00);
932 rinvsq10 = _mm_mul_ps(rinv10,rinv10);
933 rinvsq20 = _mm_mul_ps(rinv20,rinv20);
934
935 /* Load parameters for j particles */
936 jq0 = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
937 charge+jnrC+0,charge+jnrD+0);
938
939 fjx0 = _mm_setzero_ps();
940 fjy0 = _mm_setzero_ps();
941 fjz0 = _mm_setzero_ps();
942
943 /**************************
944 * CALCULATE INTERACTIONS *
945 **************************/
946
947 if (gmx_mm_any_lt(rsq00,rcutoff2))
948 {
949
950 r00 = _mm_mul_ps(rsq00,rinv00);
951
952 /* Compute parameters for interactions between i and j atoms */
953 qq00 = _mm_mul_ps(iq0,jq0);
954
955 /* EWALD ELECTROSTATICS */
956
957 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
958 ewrt = _mm_mul_ps(r00,ewtabscale);
959 ewitab = _mm_cvttps_epi32(ewrt);
960 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
961 ewitab = _mm_slli_epi32(ewitab,2);
962 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
963 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
964 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
965 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
966 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
967 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
968 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
969 velec = _mm_mul_ps(qq00,_mm_sub_ps(rinv00,velec));
970 felec = _mm_mul_ps(_mm_mul_ps(qq00,rinv00),_mm_sub_ps(rinvsq00,felec));
971
972 d = _mm_sub_ps(r00,rswitch);
973 d = _mm_max_ps(d,_mm_setzero_ps());
974 d2 = _mm_mul_ps(d,d);
975 sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5)))))));
976
977 dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4)))));
978
979 /* Evaluate switch function */
980 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
981 felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv00,_mm_mul_ps(velec,dsw)) );
982 cutoff_mask = _mm_cmplt_ps(rsq00,rcutoff2);
983
984 fscal = felec;
985
986 fscal = _mm_and_ps(fscal,cutoff_mask);
987
988 /* Calculate temporary vectorial force */
989 tx = _mm_mul_ps(fscal,dx00);
990 ty = _mm_mul_ps(fscal,dy00);
991 tz = _mm_mul_ps(fscal,dz00);
992
993 /* Update vectorial force */
994 fix0 = _mm_add_ps(fix0,tx);
995 fiy0 = _mm_add_ps(fiy0,ty);
996 fiz0 = _mm_add_ps(fiz0,tz);
997
998 fjx0 = _mm_add_ps(fjx0,tx);
999 fjy0 = _mm_add_ps(fjy0,ty);
1000 fjz0 = _mm_add_ps(fjz0,tz);
1001
1002 }
1003
1004 /**************************
1005 * CALCULATE INTERACTIONS *
1006 **************************/
1007
1008 if (gmx_mm_any_lt(rsq10,rcutoff2))
1009 {
1010
1011 r10 = _mm_mul_ps(rsq10,rinv10);
1012
1013 /* Compute parameters for interactions between i and j atoms */
1014 qq10 = _mm_mul_ps(iq1,jq0);
1015
1016 /* EWALD ELECTROSTATICS */
1017
1018 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1019 ewrt = _mm_mul_ps(r10,ewtabscale);
1020 ewitab = _mm_cvttps_epi32(ewrt);
1021 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1022 ewitab = _mm_slli_epi32(ewitab,2);
1023 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1024 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1025 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1026 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1027 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1028 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1029 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1030 velec = _mm_mul_ps(qq10,_mm_sub_ps(rinv10,velec));
1031 felec = _mm_mul_ps(_mm_mul_ps(qq10,rinv10),_mm_sub_ps(rinvsq10,felec));
1032
1033 d = _mm_sub_ps(r10,rswitch);
1034 d = _mm_max_ps(d,_mm_setzero_ps());
1035 d2 = _mm_mul_ps(d,d);
1036 sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5)))))));
1037
1038 dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4)))));
1039
1040 /* Evaluate switch function */
1041 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1042 felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv10,_mm_mul_ps(velec,dsw)) );
1043 cutoff_mask = _mm_cmplt_ps(rsq10,rcutoff2);
1044
1045 fscal = felec;
1046
1047 fscal = _mm_and_ps(fscal,cutoff_mask);
1048
1049 /* Calculate temporary vectorial force */
1050 tx = _mm_mul_ps(fscal,dx10);
1051 ty = _mm_mul_ps(fscal,dy10);
1052 tz = _mm_mul_ps(fscal,dz10);
1053
1054 /* Update vectorial force */
1055 fix1 = _mm_add_ps(fix1,tx);
1056 fiy1 = _mm_add_ps(fiy1,ty);
1057 fiz1 = _mm_add_ps(fiz1,tz);
1058
1059 fjx0 = _mm_add_ps(fjx0,tx);
1060 fjy0 = _mm_add_ps(fjy0,ty);
1061 fjz0 = _mm_add_ps(fjz0,tz);
1062
1063 }
1064
1065 /**************************
1066 * CALCULATE INTERACTIONS *
1067 **************************/
1068
1069 if (gmx_mm_any_lt(rsq20,rcutoff2))
1070 {
1071
1072 r20 = _mm_mul_ps(rsq20,rinv20);
1073
1074 /* Compute parameters for interactions between i and j atoms */
1075 qq20 = _mm_mul_ps(iq2,jq0);
1076
1077 /* EWALD ELECTROSTATICS */
1078
1079 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1080 ewrt = _mm_mul_ps(r20,ewtabscale);
1081 ewitab = _mm_cvttps_epi32(ewrt);
1082 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1083 ewitab = _mm_slli_epi32(ewitab,2);
1084 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1085 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1086 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1087 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1088 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1089 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1090 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1091 velec = _mm_mul_ps(qq20,_mm_sub_ps(rinv20,velec));
1092 felec = _mm_mul_ps(_mm_mul_ps(qq20,rinv20),_mm_sub_ps(rinvsq20,felec));
1093
1094 d = _mm_sub_ps(r20,rswitch);
1095 d = _mm_max_ps(d,_mm_setzero_ps());
1096 d2 = _mm_mul_ps(d,d);
1097 sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5)))))));
1098
1099 dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4)))));
1100
1101 /* Evaluate switch function */
1102 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1103 felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv20,_mm_mul_ps(velec,dsw)) );
1104 cutoff_mask = _mm_cmplt_ps(rsq20,rcutoff2);
1105
1106 fscal = felec;
1107
1108 fscal = _mm_and_ps(fscal,cutoff_mask);
1109
1110 /* Calculate temporary vectorial force */
1111 tx = _mm_mul_ps(fscal,dx20);
1112 ty = _mm_mul_ps(fscal,dy20);
1113 tz = _mm_mul_ps(fscal,dz20);
1114
1115 /* Update vectorial force */
1116 fix2 = _mm_add_ps(fix2,tx);
1117 fiy2 = _mm_add_ps(fiy2,ty);
1118 fiz2 = _mm_add_ps(fiz2,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 }
1125
1126 fjptrA = f+j_coord_offsetA;
1127 fjptrB = f+j_coord_offsetB;
1128 fjptrC = f+j_coord_offsetC;
1129 fjptrD = f+j_coord_offsetD;
1130
1131 gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjx0,fjy0,fjz0);
1132
1133 /* Inner loop uses 186 flops */
1134 }
1135
1136 if(jidx<j_index_end)
1137 {
1138
1139 /* Get j neighbor index, and coordinate index */
1140 jnrlistA = jjnr[jidx];
1141 jnrlistB = jjnr[jidx+1];
1142 jnrlistC = jjnr[jidx+2];
1143 jnrlistD = jjnr[jidx+3];
1144 /* Sign of each element will be negative for non-real atoms.
1145 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1146 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1147 */
1148 dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
1149 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
1150 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
1151 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
1152 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
1153 j_coord_offsetA = DIM*jnrA;
1154 j_coord_offsetB = DIM*jnrB;
1155 j_coord_offsetC = DIM*jnrC;
1156 j_coord_offsetD = DIM*jnrD;
1157
1158 /* load j atom coordinates */
1159 gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1160 x+j_coord_offsetC,x+j_coord_offsetD,
1161 &jx0,&jy0,&jz0);
1162
1163 /* Calculate displacement vector */
1164 dx00 = _mm_sub_ps(ix0,jx0);
1165 dy00 = _mm_sub_ps(iy0,jy0);
1166 dz00 = _mm_sub_ps(iz0,jz0);
1167 dx10 = _mm_sub_ps(ix1,jx0);
1168 dy10 = _mm_sub_ps(iy1,jy0);
1169 dz10 = _mm_sub_ps(iz1,jz0);
1170 dx20 = _mm_sub_ps(ix2,jx0);
1171 dy20 = _mm_sub_ps(iy2,jy0);
1172 dz20 = _mm_sub_ps(iz2,jz0);
1173
1174 /* Calculate squared distance and things based on it */
1175 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
1176 rsq10 = gmx_mm_calc_rsq_ps(dx10,dy10,dz10);
1177 rsq20 = gmx_mm_calc_rsq_ps(dx20,dy20,dz20);
1178
1179 rinv00 = sse41_invsqrt_f(rsq00);
1180 rinv10 = sse41_invsqrt_f(rsq10);
1181 rinv20 = sse41_invsqrt_f(rsq20);
1182
1183 rinvsq00 = _mm_mul_ps(rinv00,rinv00);
1184 rinvsq10 = _mm_mul_ps(rinv10,rinv10);
1185 rinvsq20 = _mm_mul_ps(rinv20,rinv20);
1186
1187 /* Load parameters for j particles */
1188 jq0 = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
1189 charge+jnrC+0,charge+jnrD+0);
1190
1191 fjx0 = _mm_setzero_ps();
1192 fjy0 = _mm_setzero_ps();
1193 fjz0 = _mm_setzero_ps();
1194
1195 /**************************
1196 * CALCULATE INTERACTIONS *
1197 **************************/
1198
1199 if (gmx_mm_any_lt(rsq00,rcutoff2))
1200 {
1201
1202 r00 = _mm_mul_ps(rsq00,rinv00);
1203 r00 = _mm_andnot_ps(dummy_mask,r00);
1204
1205 /* Compute parameters for interactions between i and j atoms */
1206 qq00 = _mm_mul_ps(iq0,jq0);
1207
1208 /* EWALD ELECTROSTATICS */
1209
1210 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1211 ewrt = _mm_mul_ps(r00,ewtabscale);
1212 ewitab = _mm_cvttps_epi32(ewrt);
1213 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1214 ewitab = _mm_slli_epi32(ewitab,2);
1215 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1216 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1217 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1218 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1219 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1220 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1221 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1222 velec = _mm_mul_ps(qq00,_mm_sub_ps(rinv00,velec));
1223 felec = _mm_mul_ps(_mm_mul_ps(qq00,rinv00),_mm_sub_ps(rinvsq00,felec));
1224
1225 d = _mm_sub_ps(r00,rswitch);
1226 d = _mm_max_ps(d,_mm_setzero_ps());
1227 d2 = _mm_mul_ps(d,d);
1228 sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5)))))));
1229
1230 dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4)))));
1231
1232 /* Evaluate switch function */
1233 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1234 felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv00,_mm_mul_ps(velec,dsw)) );
1235 cutoff_mask = _mm_cmplt_ps(rsq00,rcutoff2);
1236
1237 fscal = felec;
1238
1239 fscal = _mm_and_ps(fscal,cutoff_mask);
1240
1241 fscal = _mm_andnot_ps(dummy_mask,fscal);
1242
1243 /* Calculate temporary vectorial force */
1244 tx = _mm_mul_ps(fscal,dx00);
1245 ty = _mm_mul_ps(fscal,dy00);
1246 tz = _mm_mul_ps(fscal,dz00);
1247
1248 /* Update vectorial force */
1249 fix0 = _mm_add_ps(fix0,tx);
1250 fiy0 = _mm_add_ps(fiy0,ty);
1251 fiz0 = _mm_add_ps(fiz0,tz);
1252
1253 fjx0 = _mm_add_ps(fjx0,tx);
1254 fjy0 = _mm_add_ps(fjy0,ty);
1255 fjz0 = _mm_add_ps(fjz0,tz);
1256
1257 }
1258
1259 /**************************
1260 * CALCULATE INTERACTIONS *
1261 **************************/
1262
1263 if (gmx_mm_any_lt(rsq10,rcutoff2))
1264 {
1265
1266 r10 = _mm_mul_ps(rsq10,rinv10);
1267 r10 = _mm_andnot_ps(dummy_mask,r10);
1268
1269 /* Compute parameters for interactions between i and j atoms */
1270 qq10 = _mm_mul_ps(iq1,jq0);
1271
1272 /* EWALD ELECTROSTATICS */
1273
1274 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1275 ewrt = _mm_mul_ps(r10,ewtabscale);
1276 ewitab = _mm_cvttps_epi32(ewrt);
1277 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1278 ewitab = _mm_slli_epi32(ewitab,2);
1279 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1280 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1281 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1282 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1283 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1284 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1285 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1286 velec = _mm_mul_ps(qq10,_mm_sub_ps(rinv10,velec));
1287 felec = _mm_mul_ps(_mm_mul_ps(qq10,rinv10),_mm_sub_ps(rinvsq10,felec));
1288
1289 d = _mm_sub_ps(r10,rswitch);
1290 d = _mm_max_ps(d,_mm_setzero_ps());
1291 d2 = _mm_mul_ps(d,d);
1292 sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5)))))));
1293
1294 dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4)))));
1295
1296 /* Evaluate switch function */
1297 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1298 felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv10,_mm_mul_ps(velec,dsw)) );
1299 cutoff_mask = _mm_cmplt_ps(rsq10,rcutoff2);
1300
1301 fscal = felec;
1302
1303 fscal = _mm_and_ps(fscal,cutoff_mask);
1304
1305 fscal = _mm_andnot_ps(dummy_mask,fscal);
1306
1307 /* Calculate temporary vectorial force */
1308 tx = _mm_mul_ps(fscal,dx10);
1309 ty = _mm_mul_ps(fscal,dy10);
1310 tz = _mm_mul_ps(fscal,dz10);
1311
1312 /* Update vectorial force */
1313 fix1 = _mm_add_ps(fix1,tx);
1314 fiy1 = _mm_add_ps(fiy1,ty);
1315 fiz1 = _mm_add_ps(fiz1,tz);
1316
1317 fjx0 = _mm_add_ps(fjx0,tx);
1318 fjy0 = _mm_add_ps(fjy0,ty);
1319 fjz0 = _mm_add_ps(fjz0,tz);
1320
1321 }
1322
1323 /**************************
1324 * CALCULATE INTERACTIONS *
1325 **************************/
1326
1327 if (gmx_mm_any_lt(rsq20,rcutoff2))
1328 {
1329
1330 r20 = _mm_mul_ps(rsq20,rinv20);
1331 r20 = _mm_andnot_ps(dummy_mask,r20);
1332
1333 /* Compute parameters for interactions between i and j atoms */
1334 qq20 = _mm_mul_ps(iq2,jq0);
1335
1336 /* EWALD ELECTROSTATICS */
1337
1338 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1339 ewrt = _mm_mul_ps(r20,ewtabscale);
1340 ewitab = _mm_cvttps_epi32(ewrt);
1341 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1342 ewitab = _mm_slli_epi32(ewitab,2);
1343 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1344 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1345 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1346 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1347 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1348 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1349 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1350 velec = _mm_mul_ps(qq20,_mm_sub_ps(rinv20,velec));
1351 felec = _mm_mul_ps(_mm_mul_ps(qq20,rinv20),_mm_sub_ps(rinvsq20,felec));
1352
1353 d = _mm_sub_ps(r20,rswitch);
1354 d = _mm_max_ps(d,_mm_setzero_ps());
1355 d2 = _mm_mul_ps(d,d);
1356 sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5)))))));
1357
1358 dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4)))));
1359
1360 /* Evaluate switch function */
1361 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1362 felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv20,_mm_mul_ps(velec,dsw)) );
1363 cutoff_mask = _mm_cmplt_ps(rsq20,rcutoff2);
1364
1365 fscal = felec;
1366
1367 fscal = _mm_and_ps(fscal,cutoff_mask);
1368
1369 fscal = _mm_andnot_ps(dummy_mask,fscal);
1370
1371 /* Calculate temporary vectorial force */
1372 tx = _mm_mul_ps(fscal,dx20);
1373 ty = _mm_mul_ps(fscal,dy20);
1374 tz = _mm_mul_ps(fscal,dz20);
1375
1376 /* Update vectorial force */
1377 fix2 = _mm_add_ps(fix2,tx);
1378 fiy2 = _mm_add_ps(fiy2,ty);
1379 fiz2 = _mm_add_ps(fiz2,tz);
1380
1381 fjx0 = _mm_add_ps(fjx0,tx);
1382 fjy0 = _mm_add_ps(fjy0,ty);
1383 fjz0 = _mm_add_ps(fjz0,tz);
1384
1385 }
1386
1387 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1388 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1389 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1390 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1391
1392 gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjx0,fjy0,fjz0);
1393
1394 /* Inner loop uses 189 flops */
1395 }
1396
1397 /* End of innermost loop */
1398
1399 gmx_mm_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1400 f+i_coord_offset,fshift+i_shift_offset);
1401
1402 /* Increment number of inner iterations */
1403 inneriter += j_index_end - j_index_start;
1404
1405 /* Outer loop uses 18 flops */
1406 }
1407
1408 /* Increment number of outer iterations */
1409 outeriter += nri;
1410
1411 /* Update outer/inner flops */
1412
1413 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3_F,outeriter*18 + inneriter*189);
1414 }
The ultimate molecular dynamics simulation package