2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2012,2013,2014, 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.
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.
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.
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.
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.
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.
36 * Note: this file was generated by the GROMACS sse2_double kernel generator.
42 #include "../nb_kernel.h"
43 #include "types/simple.h"
44 #include "gromacs/math/vec.h"
47 #include "gromacs/simd/math_x86_sse2_double.h"
48 #include "kernelutil_x86_sse2_double.h"
51 * Gromacs nonbonded kernel: nb_kernel_ElecEwSh_VdwNone_GeomW4P1_VF_sse2_double
52 * Electrostatics interaction: Ewald
53 * VdW interaction: None
54 * Geometry: Water4-Particle
55 * Calculate force/pot: PotentialAndForce
58 nb_kernel_ElecEwSh_VdwNone_GeomW4P1_VF_sse2_double
59 (t_nblist * gmx_restrict nlist,
60 rvec * gmx_restrict xx,
61 rvec * gmx_restrict ff,
62 t_forcerec * gmx_restrict fr,
63 t_mdatoms * gmx_restrict mdatoms,
64 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
65 t_nrnb * gmx_restrict nrnb)
67 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
68 * just 0 for non-waters.
69 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
70 * jnr indices corresponding to data put in the four positions in the SIMD register.
72 int i_shift_offset,i_coord_offset,outeriter,inneriter;
73 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
75 int j_coord_offsetA,j_coord_offsetB;
76 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
78 real *shiftvec,*fshift,*x,*f;
79 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
81 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
83 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
85 __m128d ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
86 int vdwjidx0A,vdwjidx0B;
87 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
88 __m128d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
89 __m128d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
90 __m128d dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
91 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
94 __m128d ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
96 __m128d dummy_mask,cutoff_mask;
97 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
98 __m128d one = _mm_set1_pd(1.0);
99 __m128d two = _mm_set1_pd(2.0);
105 jindex = nlist->jindex;
107 shiftidx = nlist->shift;
109 shiftvec = fr->shift_vec[0];
110 fshift = fr->fshift[0];
111 facel = _mm_set1_pd(fr->epsfac);
112 charge = mdatoms->chargeA;
114 sh_ewald = _mm_set1_pd(fr->ic->sh_ewald);
115 ewtab = fr->ic->tabq_coul_FDV0;
116 ewtabscale = _mm_set1_pd(fr->ic->tabq_scale);
117 ewtabhalfspace = _mm_set1_pd(0.5/fr->ic->tabq_scale);
119 /* Setup water-specific parameters */
120 inr = nlist->iinr[0];
121 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
122 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
123 iq3 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
125 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
126 rcutoff_scalar = fr->rcoulomb;
127 rcutoff = _mm_set1_pd(rcutoff_scalar);
128 rcutoff2 = _mm_mul_pd(rcutoff,rcutoff);
130 /* Avoid stupid compiler warnings */
138 /* Start outer loop over neighborlists */
139 for(iidx=0; iidx<nri; iidx++)
141 /* Load shift vector for this list */
142 i_shift_offset = DIM*shiftidx[iidx];
144 /* Load limits for loop over neighbors */
145 j_index_start = jindex[iidx];
146 j_index_end = jindex[iidx+1];
148 /* Get outer coordinate index */
150 i_coord_offset = DIM*inr;
152 /* Load i particle coords and add shift vector */
153 gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
154 &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
156 fix1 = _mm_setzero_pd();
157 fiy1 = _mm_setzero_pd();
158 fiz1 = _mm_setzero_pd();
159 fix2 = _mm_setzero_pd();
160 fiy2 = _mm_setzero_pd();
161 fiz2 = _mm_setzero_pd();
162 fix3 = _mm_setzero_pd();
163 fiy3 = _mm_setzero_pd();
164 fiz3 = _mm_setzero_pd();
166 /* Reset potential sums */
167 velecsum = _mm_setzero_pd();
169 /* Start inner kernel loop */
170 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
173 /* Get j neighbor index, and coordinate index */
176 j_coord_offsetA = DIM*jnrA;
177 j_coord_offsetB = DIM*jnrB;
179 /* load j atom coordinates */
180 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
183 /* Calculate displacement vector */
184 dx10 = _mm_sub_pd(ix1,jx0);
185 dy10 = _mm_sub_pd(iy1,jy0);
186 dz10 = _mm_sub_pd(iz1,jz0);
187 dx20 = _mm_sub_pd(ix2,jx0);
188 dy20 = _mm_sub_pd(iy2,jy0);
189 dz20 = _mm_sub_pd(iz2,jz0);
190 dx30 = _mm_sub_pd(ix3,jx0);
191 dy30 = _mm_sub_pd(iy3,jy0);
192 dz30 = _mm_sub_pd(iz3,jz0);
194 /* Calculate squared distance and things based on it */
195 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
196 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
197 rsq30 = gmx_mm_calc_rsq_pd(dx30,dy30,dz30);
199 rinv10 = gmx_mm_invsqrt_pd(rsq10);
200 rinv20 = gmx_mm_invsqrt_pd(rsq20);
201 rinv30 = gmx_mm_invsqrt_pd(rsq30);
203 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
204 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
205 rinvsq30 = _mm_mul_pd(rinv30,rinv30);
207 /* Load parameters for j particles */
208 jq0 = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
210 fjx0 = _mm_setzero_pd();
211 fjy0 = _mm_setzero_pd();
212 fjz0 = _mm_setzero_pd();
214 /**************************
215 * CALCULATE INTERACTIONS *
216 **************************/
218 if (gmx_mm_any_lt(rsq10,rcutoff2))
221 r10 = _mm_mul_pd(rsq10,rinv10);
223 /* Compute parameters for interactions between i and j atoms */
224 qq10 = _mm_mul_pd(iq1,jq0);
226 /* EWALD ELECTROSTATICS */
228 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
229 ewrt = _mm_mul_pd(r10,ewtabscale);
230 ewitab = _mm_cvttpd_epi32(ewrt);
231 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
232 ewitab = _mm_slli_epi32(ewitab,2);
233 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
234 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
235 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
236 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
237 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
238 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
239 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
240 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
241 velec = _mm_mul_pd(qq10,_mm_sub_pd(_mm_sub_pd(rinv10,sh_ewald),velec));
242 felec = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
244 cutoff_mask = _mm_cmplt_pd(rsq10,rcutoff2);
246 /* Update potential sum for this i atom from the interaction with this j atom. */
247 velec = _mm_and_pd(velec,cutoff_mask);
248 velecsum = _mm_add_pd(velecsum,velec);
252 fscal = _mm_and_pd(fscal,cutoff_mask);
254 /* Calculate temporary vectorial force */
255 tx = _mm_mul_pd(fscal,dx10);
256 ty = _mm_mul_pd(fscal,dy10);
257 tz = _mm_mul_pd(fscal,dz10);
259 /* Update vectorial force */
260 fix1 = _mm_add_pd(fix1,tx);
261 fiy1 = _mm_add_pd(fiy1,ty);
262 fiz1 = _mm_add_pd(fiz1,tz);
264 fjx0 = _mm_add_pd(fjx0,tx);
265 fjy0 = _mm_add_pd(fjy0,ty);
266 fjz0 = _mm_add_pd(fjz0,tz);
270 /**************************
271 * CALCULATE INTERACTIONS *
272 **************************/
274 if (gmx_mm_any_lt(rsq20,rcutoff2))
277 r20 = _mm_mul_pd(rsq20,rinv20);
279 /* Compute parameters for interactions between i and j atoms */
280 qq20 = _mm_mul_pd(iq2,jq0);
282 /* EWALD ELECTROSTATICS */
284 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
285 ewrt = _mm_mul_pd(r20,ewtabscale);
286 ewitab = _mm_cvttpd_epi32(ewrt);
287 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
288 ewitab = _mm_slli_epi32(ewitab,2);
289 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
290 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
291 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
292 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
293 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
294 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
295 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
296 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
297 velec = _mm_mul_pd(qq20,_mm_sub_pd(_mm_sub_pd(rinv20,sh_ewald),velec));
298 felec = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
300 cutoff_mask = _mm_cmplt_pd(rsq20,rcutoff2);
302 /* Update potential sum for this i atom from the interaction with this j atom. */
303 velec = _mm_and_pd(velec,cutoff_mask);
304 velecsum = _mm_add_pd(velecsum,velec);
308 fscal = _mm_and_pd(fscal,cutoff_mask);
310 /* Calculate temporary vectorial force */
311 tx = _mm_mul_pd(fscal,dx20);
312 ty = _mm_mul_pd(fscal,dy20);
313 tz = _mm_mul_pd(fscal,dz20);
315 /* Update vectorial force */
316 fix2 = _mm_add_pd(fix2,tx);
317 fiy2 = _mm_add_pd(fiy2,ty);
318 fiz2 = _mm_add_pd(fiz2,tz);
320 fjx0 = _mm_add_pd(fjx0,tx);
321 fjy0 = _mm_add_pd(fjy0,ty);
322 fjz0 = _mm_add_pd(fjz0,tz);
326 /**************************
327 * CALCULATE INTERACTIONS *
328 **************************/
330 if (gmx_mm_any_lt(rsq30,rcutoff2))
333 r30 = _mm_mul_pd(rsq30,rinv30);
335 /* Compute parameters for interactions between i and j atoms */
336 qq30 = _mm_mul_pd(iq3,jq0);
338 /* EWALD ELECTROSTATICS */
340 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
341 ewrt = _mm_mul_pd(r30,ewtabscale);
342 ewitab = _mm_cvttpd_epi32(ewrt);
343 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
344 ewitab = _mm_slli_epi32(ewitab,2);
345 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
346 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
347 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
348 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
349 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
350 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
351 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
352 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
353 velec = _mm_mul_pd(qq30,_mm_sub_pd(_mm_sub_pd(rinv30,sh_ewald),velec));
354 felec = _mm_mul_pd(_mm_mul_pd(qq30,rinv30),_mm_sub_pd(rinvsq30,felec));
356 cutoff_mask = _mm_cmplt_pd(rsq30,rcutoff2);
358 /* Update potential sum for this i atom from the interaction with this j atom. */
359 velec = _mm_and_pd(velec,cutoff_mask);
360 velecsum = _mm_add_pd(velecsum,velec);
364 fscal = _mm_and_pd(fscal,cutoff_mask);
366 /* Calculate temporary vectorial force */
367 tx = _mm_mul_pd(fscal,dx30);
368 ty = _mm_mul_pd(fscal,dy30);
369 tz = _mm_mul_pd(fscal,dz30);
371 /* Update vectorial force */
372 fix3 = _mm_add_pd(fix3,tx);
373 fiy3 = _mm_add_pd(fiy3,ty);
374 fiz3 = _mm_add_pd(fiz3,tz);
376 fjx0 = _mm_add_pd(fjx0,tx);
377 fjy0 = _mm_add_pd(fjy0,ty);
378 fjz0 = _mm_add_pd(fjz0,tz);
382 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
384 /* Inner loop uses 141 flops */
391 j_coord_offsetA = DIM*jnrA;
393 /* load j atom coordinates */
394 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
397 /* Calculate displacement vector */
398 dx10 = _mm_sub_pd(ix1,jx0);
399 dy10 = _mm_sub_pd(iy1,jy0);
400 dz10 = _mm_sub_pd(iz1,jz0);
401 dx20 = _mm_sub_pd(ix2,jx0);
402 dy20 = _mm_sub_pd(iy2,jy0);
403 dz20 = _mm_sub_pd(iz2,jz0);
404 dx30 = _mm_sub_pd(ix3,jx0);
405 dy30 = _mm_sub_pd(iy3,jy0);
406 dz30 = _mm_sub_pd(iz3,jz0);
408 /* Calculate squared distance and things based on it */
409 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
410 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
411 rsq30 = gmx_mm_calc_rsq_pd(dx30,dy30,dz30);
413 rinv10 = gmx_mm_invsqrt_pd(rsq10);
414 rinv20 = gmx_mm_invsqrt_pd(rsq20);
415 rinv30 = gmx_mm_invsqrt_pd(rsq30);
417 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
418 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
419 rinvsq30 = _mm_mul_pd(rinv30,rinv30);
421 /* Load parameters for j particles */
422 jq0 = _mm_load_sd(charge+jnrA+0);
424 fjx0 = _mm_setzero_pd();
425 fjy0 = _mm_setzero_pd();
426 fjz0 = _mm_setzero_pd();
428 /**************************
429 * CALCULATE INTERACTIONS *
430 **************************/
432 if (gmx_mm_any_lt(rsq10,rcutoff2))
435 r10 = _mm_mul_pd(rsq10,rinv10);
437 /* Compute parameters for interactions between i and j atoms */
438 qq10 = _mm_mul_pd(iq1,jq0);
440 /* EWALD ELECTROSTATICS */
442 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
443 ewrt = _mm_mul_pd(r10,ewtabscale);
444 ewitab = _mm_cvttpd_epi32(ewrt);
445 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
446 ewitab = _mm_slli_epi32(ewitab,2);
447 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
448 ewtabD = _mm_setzero_pd();
449 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
450 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
451 ewtabFn = _mm_setzero_pd();
452 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
453 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
454 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
455 velec = _mm_mul_pd(qq10,_mm_sub_pd(_mm_sub_pd(rinv10,sh_ewald),velec));
456 felec = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
458 cutoff_mask = _mm_cmplt_pd(rsq10,rcutoff2);
460 /* Update potential sum for this i atom from the interaction with this j atom. */
461 velec = _mm_and_pd(velec,cutoff_mask);
462 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
463 velecsum = _mm_add_pd(velecsum,velec);
467 fscal = _mm_and_pd(fscal,cutoff_mask);
469 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
471 /* Calculate temporary vectorial force */
472 tx = _mm_mul_pd(fscal,dx10);
473 ty = _mm_mul_pd(fscal,dy10);
474 tz = _mm_mul_pd(fscal,dz10);
476 /* Update vectorial force */
477 fix1 = _mm_add_pd(fix1,tx);
478 fiy1 = _mm_add_pd(fiy1,ty);
479 fiz1 = _mm_add_pd(fiz1,tz);
481 fjx0 = _mm_add_pd(fjx0,tx);
482 fjy0 = _mm_add_pd(fjy0,ty);
483 fjz0 = _mm_add_pd(fjz0,tz);
487 /**************************
488 * CALCULATE INTERACTIONS *
489 **************************/
491 if (gmx_mm_any_lt(rsq20,rcutoff2))
494 r20 = _mm_mul_pd(rsq20,rinv20);
496 /* Compute parameters for interactions between i and j atoms */
497 qq20 = _mm_mul_pd(iq2,jq0);
499 /* EWALD ELECTROSTATICS */
501 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
502 ewrt = _mm_mul_pd(r20,ewtabscale);
503 ewitab = _mm_cvttpd_epi32(ewrt);
504 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
505 ewitab = _mm_slli_epi32(ewitab,2);
506 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
507 ewtabD = _mm_setzero_pd();
508 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
509 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
510 ewtabFn = _mm_setzero_pd();
511 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
512 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
513 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
514 velec = _mm_mul_pd(qq20,_mm_sub_pd(_mm_sub_pd(rinv20,sh_ewald),velec));
515 felec = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
517 cutoff_mask = _mm_cmplt_pd(rsq20,rcutoff2);
519 /* Update potential sum for this i atom from the interaction with this j atom. */
520 velec = _mm_and_pd(velec,cutoff_mask);
521 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
522 velecsum = _mm_add_pd(velecsum,velec);
526 fscal = _mm_and_pd(fscal,cutoff_mask);
528 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
530 /* Calculate temporary vectorial force */
531 tx = _mm_mul_pd(fscal,dx20);
532 ty = _mm_mul_pd(fscal,dy20);
533 tz = _mm_mul_pd(fscal,dz20);
535 /* Update vectorial force */
536 fix2 = _mm_add_pd(fix2,tx);
537 fiy2 = _mm_add_pd(fiy2,ty);
538 fiz2 = _mm_add_pd(fiz2,tz);
540 fjx0 = _mm_add_pd(fjx0,tx);
541 fjy0 = _mm_add_pd(fjy0,ty);
542 fjz0 = _mm_add_pd(fjz0,tz);
546 /**************************
547 * CALCULATE INTERACTIONS *
548 **************************/
550 if (gmx_mm_any_lt(rsq30,rcutoff2))
553 r30 = _mm_mul_pd(rsq30,rinv30);
555 /* Compute parameters for interactions between i and j atoms */
556 qq30 = _mm_mul_pd(iq3,jq0);
558 /* EWALD ELECTROSTATICS */
560 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
561 ewrt = _mm_mul_pd(r30,ewtabscale);
562 ewitab = _mm_cvttpd_epi32(ewrt);
563 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
564 ewitab = _mm_slli_epi32(ewitab,2);
565 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
566 ewtabD = _mm_setzero_pd();
567 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
568 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
569 ewtabFn = _mm_setzero_pd();
570 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
571 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
572 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
573 velec = _mm_mul_pd(qq30,_mm_sub_pd(_mm_sub_pd(rinv30,sh_ewald),velec));
574 felec = _mm_mul_pd(_mm_mul_pd(qq30,rinv30),_mm_sub_pd(rinvsq30,felec));
576 cutoff_mask = _mm_cmplt_pd(rsq30,rcutoff2);
578 /* Update potential sum for this i atom from the interaction with this j atom. */
579 velec = _mm_and_pd(velec,cutoff_mask);
580 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
581 velecsum = _mm_add_pd(velecsum,velec);
585 fscal = _mm_and_pd(fscal,cutoff_mask);
587 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
589 /* Calculate temporary vectorial force */
590 tx = _mm_mul_pd(fscal,dx30);
591 ty = _mm_mul_pd(fscal,dy30);
592 tz = _mm_mul_pd(fscal,dz30);
594 /* Update vectorial force */
595 fix3 = _mm_add_pd(fix3,tx);
596 fiy3 = _mm_add_pd(fiy3,ty);
597 fiz3 = _mm_add_pd(fiz3,tz);
599 fjx0 = _mm_add_pd(fjx0,tx);
600 fjy0 = _mm_add_pd(fjy0,ty);
601 fjz0 = _mm_add_pd(fjz0,tz);
605 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0);
607 /* Inner loop uses 141 flops */
610 /* End of innermost loop */
612 gmx_mm_update_iforce_3atom_swizzle_pd(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
613 f+i_coord_offset+DIM,fshift+i_shift_offset);
616 /* Update potential energies */
617 gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
619 /* Increment number of inner iterations */
620 inneriter += j_index_end - j_index_start;
622 /* Outer loop uses 19 flops */
625 /* Increment number of outer iterations */
628 /* Update outer/inner flops */
630 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4_VF,outeriter*19 + inneriter*141);
633 * Gromacs nonbonded kernel: nb_kernel_ElecEwSh_VdwNone_GeomW4P1_F_sse2_double
634 * Electrostatics interaction: Ewald
635 * VdW interaction: None
636 * Geometry: Water4-Particle
637 * Calculate force/pot: Force
640 nb_kernel_ElecEwSh_VdwNone_GeomW4P1_F_sse2_double
641 (t_nblist * gmx_restrict nlist,
642 rvec * gmx_restrict xx,
643 rvec * gmx_restrict ff,
644 t_forcerec * gmx_restrict fr,
645 t_mdatoms * gmx_restrict mdatoms,
646 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
647 t_nrnb * gmx_restrict nrnb)
649 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
650 * just 0 for non-waters.
651 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
652 * jnr indices corresponding to data put in the four positions in the SIMD register.
654 int i_shift_offset,i_coord_offset,outeriter,inneriter;
655 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
657 int j_coord_offsetA,j_coord_offsetB;
658 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
660 real *shiftvec,*fshift,*x,*f;
661 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
663 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
665 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
667 __m128d ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
668 int vdwjidx0A,vdwjidx0B;
669 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
670 __m128d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
671 __m128d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
672 __m128d dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
673 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
676 __m128d ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
678 __m128d dummy_mask,cutoff_mask;
679 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
680 __m128d one = _mm_set1_pd(1.0);
681 __m128d two = _mm_set1_pd(2.0);
687 jindex = nlist->jindex;
689 shiftidx = nlist->shift;
691 shiftvec = fr->shift_vec[0];
692 fshift = fr->fshift[0];
693 facel = _mm_set1_pd(fr->epsfac);
694 charge = mdatoms->chargeA;
696 sh_ewald = _mm_set1_pd(fr->ic->sh_ewald);
697 ewtab = fr->ic->tabq_coul_F;
698 ewtabscale = _mm_set1_pd(fr->ic->tabq_scale);
699 ewtabhalfspace = _mm_set1_pd(0.5/fr->ic->tabq_scale);
701 /* Setup water-specific parameters */
702 inr = nlist->iinr[0];
703 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
704 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
705 iq3 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
707 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
708 rcutoff_scalar = fr->rcoulomb;
709 rcutoff = _mm_set1_pd(rcutoff_scalar);
710 rcutoff2 = _mm_mul_pd(rcutoff,rcutoff);
712 /* Avoid stupid compiler warnings */
720 /* Start outer loop over neighborlists */
721 for(iidx=0; iidx<nri; iidx++)
723 /* Load shift vector for this list */
724 i_shift_offset = DIM*shiftidx[iidx];
726 /* Load limits for loop over neighbors */
727 j_index_start = jindex[iidx];
728 j_index_end = jindex[iidx+1];
730 /* Get outer coordinate index */
732 i_coord_offset = DIM*inr;
734 /* Load i particle coords and add shift vector */
735 gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
736 &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
738 fix1 = _mm_setzero_pd();
739 fiy1 = _mm_setzero_pd();
740 fiz1 = _mm_setzero_pd();
741 fix2 = _mm_setzero_pd();
742 fiy2 = _mm_setzero_pd();
743 fiz2 = _mm_setzero_pd();
744 fix3 = _mm_setzero_pd();
745 fiy3 = _mm_setzero_pd();
746 fiz3 = _mm_setzero_pd();
748 /* Start inner kernel loop */
749 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
752 /* Get j neighbor index, and coordinate index */
755 j_coord_offsetA = DIM*jnrA;
756 j_coord_offsetB = DIM*jnrB;
758 /* load j atom coordinates */
759 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
762 /* Calculate displacement vector */
763 dx10 = _mm_sub_pd(ix1,jx0);
764 dy10 = _mm_sub_pd(iy1,jy0);
765 dz10 = _mm_sub_pd(iz1,jz0);
766 dx20 = _mm_sub_pd(ix2,jx0);
767 dy20 = _mm_sub_pd(iy2,jy0);
768 dz20 = _mm_sub_pd(iz2,jz0);
769 dx30 = _mm_sub_pd(ix3,jx0);
770 dy30 = _mm_sub_pd(iy3,jy0);
771 dz30 = _mm_sub_pd(iz3,jz0);
773 /* Calculate squared distance and things based on it */
774 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
775 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
776 rsq30 = gmx_mm_calc_rsq_pd(dx30,dy30,dz30);
778 rinv10 = gmx_mm_invsqrt_pd(rsq10);
779 rinv20 = gmx_mm_invsqrt_pd(rsq20);
780 rinv30 = gmx_mm_invsqrt_pd(rsq30);
782 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
783 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
784 rinvsq30 = _mm_mul_pd(rinv30,rinv30);
786 /* Load parameters for j particles */
787 jq0 = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
789 fjx0 = _mm_setzero_pd();
790 fjy0 = _mm_setzero_pd();
791 fjz0 = _mm_setzero_pd();
793 /**************************
794 * CALCULATE INTERACTIONS *
795 **************************/
797 if (gmx_mm_any_lt(rsq10,rcutoff2))
800 r10 = _mm_mul_pd(rsq10,rinv10);
802 /* Compute parameters for interactions between i and j atoms */
803 qq10 = _mm_mul_pd(iq1,jq0);
805 /* EWALD ELECTROSTATICS */
807 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
808 ewrt = _mm_mul_pd(r10,ewtabscale);
809 ewitab = _mm_cvttpd_epi32(ewrt);
810 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
811 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
813 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
814 felec = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
816 cutoff_mask = _mm_cmplt_pd(rsq10,rcutoff2);
820 fscal = _mm_and_pd(fscal,cutoff_mask);
822 /* Calculate temporary vectorial force */
823 tx = _mm_mul_pd(fscal,dx10);
824 ty = _mm_mul_pd(fscal,dy10);
825 tz = _mm_mul_pd(fscal,dz10);
827 /* Update vectorial force */
828 fix1 = _mm_add_pd(fix1,tx);
829 fiy1 = _mm_add_pd(fiy1,ty);
830 fiz1 = _mm_add_pd(fiz1,tz);
832 fjx0 = _mm_add_pd(fjx0,tx);
833 fjy0 = _mm_add_pd(fjy0,ty);
834 fjz0 = _mm_add_pd(fjz0,tz);
838 /**************************
839 * CALCULATE INTERACTIONS *
840 **************************/
842 if (gmx_mm_any_lt(rsq20,rcutoff2))
845 r20 = _mm_mul_pd(rsq20,rinv20);
847 /* Compute parameters for interactions between i and j atoms */
848 qq20 = _mm_mul_pd(iq2,jq0);
850 /* EWALD ELECTROSTATICS */
852 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
853 ewrt = _mm_mul_pd(r20,ewtabscale);
854 ewitab = _mm_cvttpd_epi32(ewrt);
855 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
856 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
858 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
859 felec = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
861 cutoff_mask = _mm_cmplt_pd(rsq20,rcutoff2);
865 fscal = _mm_and_pd(fscal,cutoff_mask);
867 /* Calculate temporary vectorial force */
868 tx = _mm_mul_pd(fscal,dx20);
869 ty = _mm_mul_pd(fscal,dy20);
870 tz = _mm_mul_pd(fscal,dz20);
872 /* Update vectorial force */
873 fix2 = _mm_add_pd(fix2,tx);
874 fiy2 = _mm_add_pd(fiy2,ty);
875 fiz2 = _mm_add_pd(fiz2,tz);
877 fjx0 = _mm_add_pd(fjx0,tx);
878 fjy0 = _mm_add_pd(fjy0,ty);
879 fjz0 = _mm_add_pd(fjz0,tz);
883 /**************************
884 * CALCULATE INTERACTIONS *
885 **************************/
887 if (gmx_mm_any_lt(rsq30,rcutoff2))
890 r30 = _mm_mul_pd(rsq30,rinv30);
892 /* Compute parameters for interactions between i and j atoms */
893 qq30 = _mm_mul_pd(iq3,jq0);
895 /* EWALD ELECTROSTATICS */
897 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
898 ewrt = _mm_mul_pd(r30,ewtabscale);
899 ewitab = _mm_cvttpd_epi32(ewrt);
900 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
901 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
903 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
904 felec = _mm_mul_pd(_mm_mul_pd(qq30,rinv30),_mm_sub_pd(rinvsq30,felec));
906 cutoff_mask = _mm_cmplt_pd(rsq30,rcutoff2);
910 fscal = _mm_and_pd(fscal,cutoff_mask);
912 /* Calculate temporary vectorial force */
913 tx = _mm_mul_pd(fscal,dx30);
914 ty = _mm_mul_pd(fscal,dy30);
915 tz = _mm_mul_pd(fscal,dz30);
917 /* Update vectorial force */
918 fix3 = _mm_add_pd(fix3,tx);
919 fiy3 = _mm_add_pd(fiy3,ty);
920 fiz3 = _mm_add_pd(fiz3,tz);
922 fjx0 = _mm_add_pd(fjx0,tx);
923 fjy0 = _mm_add_pd(fjy0,ty);
924 fjz0 = _mm_add_pd(fjz0,tz);
928 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
930 /* Inner loop uses 120 flops */
937 j_coord_offsetA = DIM*jnrA;
939 /* load j atom coordinates */
940 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
943 /* Calculate displacement vector */
944 dx10 = _mm_sub_pd(ix1,jx0);
945 dy10 = _mm_sub_pd(iy1,jy0);
946 dz10 = _mm_sub_pd(iz1,jz0);
947 dx20 = _mm_sub_pd(ix2,jx0);
948 dy20 = _mm_sub_pd(iy2,jy0);
949 dz20 = _mm_sub_pd(iz2,jz0);
950 dx30 = _mm_sub_pd(ix3,jx0);
951 dy30 = _mm_sub_pd(iy3,jy0);
952 dz30 = _mm_sub_pd(iz3,jz0);
954 /* Calculate squared distance and things based on it */
955 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
956 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
957 rsq30 = gmx_mm_calc_rsq_pd(dx30,dy30,dz30);
959 rinv10 = gmx_mm_invsqrt_pd(rsq10);
960 rinv20 = gmx_mm_invsqrt_pd(rsq20);
961 rinv30 = gmx_mm_invsqrt_pd(rsq30);
963 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
964 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
965 rinvsq30 = _mm_mul_pd(rinv30,rinv30);
967 /* Load parameters for j particles */
968 jq0 = _mm_load_sd(charge+jnrA+0);
970 fjx0 = _mm_setzero_pd();
971 fjy0 = _mm_setzero_pd();
972 fjz0 = _mm_setzero_pd();
974 /**************************
975 * CALCULATE INTERACTIONS *
976 **************************/
978 if (gmx_mm_any_lt(rsq10,rcutoff2))
981 r10 = _mm_mul_pd(rsq10,rinv10);
983 /* Compute parameters for interactions between i and j atoms */
984 qq10 = _mm_mul_pd(iq1,jq0);
986 /* EWALD ELECTROSTATICS */
988 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
989 ewrt = _mm_mul_pd(r10,ewtabscale);
990 ewitab = _mm_cvttpd_epi32(ewrt);
991 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
992 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
993 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
994 felec = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
996 cutoff_mask = _mm_cmplt_pd(rsq10,rcutoff2);
1000 fscal = _mm_and_pd(fscal,cutoff_mask);
1002 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1004 /* Calculate temporary vectorial force */
1005 tx = _mm_mul_pd(fscal,dx10);
1006 ty = _mm_mul_pd(fscal,dy10);
1007 tz = _mm_mul_pd(fscal,dz10);
1009 /* Update vectorial force */
1010 fix1 = _mm_add_pd(fix1,tx);
1011 fiy1 = _mm_add_pd(fiy1,ty);
1012 fiz1 = _mm_add_pd(fiz1,tz);
1014 fjx0 = _mm_add_pd(fjx0,tx);
1015 fjy0 = _mm_add_pd(fjy0,ty);
1016 fjz0 = _mm_add_pd(fjz0,tz);
1020 /**************************
1021 * CALCULATE INTERACTIONS *
1022 **************************/
1024 if (gmx_mm_any_lt(rsq20,rcutoff2))
1027 r20 = _mm_mul_pd(rsq20,rinv20);
1029 /* Compute parameters for interactions between i and j atoms */
1030 qq20 = _mm_mul_pd(iq2,jq0);
1032 /* EWALD ELECTROSTATICS */
1034 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1035 ewrt = _mm_mul_pd(r20,ewtabscale);
1036 ewitab = _mm_cvttpd_epi32(ewrt);
1037 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1038 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1039 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1040 felec = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
1042 cutoff_mask = _mm_cmplt_pd(rsq20,rcutoff2);
1046 fscal = _mm_and_pd(fscal,cutoff_mask);
1048 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1050 /* Calculate temporary vectorial force */
1051 tx = _mm_mul_pd(fscal,dx20);
1052 ty = _mm_mul_pd(fscal,dy20);
1053 tz = _mm_mul_pd(fscal,dz20);
1055 /* Update vectorial force */
1056 fix2 = _mm_add_pd(fix2,tx);
1057 fiy2 = _mm_add_pd(fiy2,ty);
1058 fiz2 = _mm_add_pd(fiz2,tz);
1060 fjx0 = _mm_add_pd(fjx0,tx);
1061 fjy0 = _mm_add_pd(fjy0,ty);
1062 fjz0 = _mm_add_pd(fjz0,tz);
1066 /**************************
1067 * CALCULATE INTERACTIONS *
1068 **************************/
1070 if (gmx_mm_any_lt(rsq30,rcutoff2))
1073 r30 = _mm_mul_pd(rsq30,rinv30);
1075 /* Compute parameters for interactions between i and j atoms */
1076 qq30 = _mm_mul_pd(iq3,jq0);
1078 /* EWALD ELECTROSTATICS */
1080 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1081 ewrt = _mm_mul_pd(r30,ewtabscale);
1082 ewitab = _mm_cvttpd_epi32(ewrt);
1083 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1084 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1085 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1086 felec = _mm_mul_pd(_mm_mul_pd(qq30,rinv30),_mm_sub_pd(rinvsq30,felec));
1088 cutoff_mask = _mm_cmplt_pd(rsq30,rcutoff2);
1092 fscal = _mm_and_pd(fscal,cutoff_mask);
1094 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1096 /* Calculate temporary vectorial force */
1097 tx = _mm_mul_pd(fscal,dx30);
1098 ty = _mm_mul_pd(fscal,dy30);
1099 tz = _mm_mul_pd(fscal,dz30);
1101 /* Update vectorial force */
1102 fix3 = _mm_add_pd(fix3,tx);
1103 fiy3 = _mm_add_pd(fiy3,ty);
1104 fiz3 = _mm_add_pd(fiz3,tz);
1106 fjx0 = _mm_add_pd(fjx0,tx);
1107 fjy0 = _mm_add_pd(fjy0,ty);
1108 fjz0 = _mm_add_pd(fjz0,tz);
1112 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1114 /* Inner loop uses 120 flops */
1117 /* End of innermost loop */
1119 gmx_mm_update_iforce_3atom_swizzle_pd(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1120 f+i_coord_offset+DIM,fshift+i_shift_offset);
1122 /* Increment number of inner iterations */
1123 inneriter += j_index_end - j_index_start;
1125 /* Outer loop uses 18 flops */
1128 /* Increment number of outer iterations */
1131 /* Update outer/inner flops */
1133 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4_F,outeriter*18 + inneriter*120);