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_GeomW3W3_VF_sse2_double
52 * Electrostatics interaction: Ewald
53 * VdW interaction: None
54 * Geometry: Water3-Water3
55 * Calculate force/pot: PotentialAndForce
58 nb_kernel_ElecEwSh_VdwNone_GeomW3W3_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 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
83 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
85 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
86 int vdwjidx0A,vdwjidx0B;
87 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
88 int vdwjidx1A,vdwjidx1B;
89 __m128d jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
90 int vdwjidx2A,vdwjidx2B;
91 __m128d jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
92 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
93 __m128d dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
94 __m128d dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
95 __m128d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
96 __m128d dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
97 __m128d dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
98 __m128d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
99 __m128d dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
100 __m128d dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
101 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
104 __m128d ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
106 __m128d dummy_mask,cutoff_mask;
107 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
108 __m128d one = _mm_set1_pd(1.0);
109 __m128d two = _mm_set1_pd(2.0);
115 jindex = nlist->jindex;
117 shiftidx = nlist->shift;
119 shiftvec = fr->shift_vec[0];
120 fshift = fr->fshift[0];
121 facel = _mm_set1_pd(fr->epsfac);
122 charge = mdatoms->chargeA;
124 sh_ewald = _mm_set1_pd(fr->ic->sh_ewald);
125 ewtab = fr->ic->tabq_coul_FDV0;
126 ewtabscale = _mm_set1_pd(fr->ic->tabq_scale);
127 ewtabhalfspace = _mm_set1_pd(0.5/fr->ic->tabq_scale);
129 /* Setup water-specific parameters */
130 inr = nlist->iinr[0];
131 iq0 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
132 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
133 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
135 jq0 = _mm_set1_pd(charge[inr+0]);
136 jq1 = _mm_set1_pd(charge[inr+1]);
137 jq2 = _mm_set1_pd(charge[inr+2]);
138 qq00 = _mm_mul_pd(iq0,jq0);
139 qq01 = _mm_mul_pd(iq0,jq1);
140 qq02 = _mm_mul_pd(iq0,jq2);
141 qq10 = _mm_mul_pd(iq1,jq0);
142 qq11 = _mm_mul_pd(iq1,jq1);
143 qq12 = _mm_mul_pd(iq1,jq2);
144 qq20 = _mm_mul_pd(iq2,jq0);
145 qq21 = _mm_mul_pd(iq2,jq1);
146 qq22 = _mm_mul_pd(iq2,jq2);
148 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
149 rcutoff_scalar = fr->rcoulomb;
150 rcutoff = _mm_set1_pd(rcutoff_scalar);
151 rcutoff2 = _mm_mul_pd(rcutoff,rcutoff);
153 /* Avoid stupid compiler warnings */
161 /* Start outer loop over neighborlists */
162 for(iidx=0; iidx<nri; iidx++)
164 /* Load shift vector for this list */
165 i_shift_offset = DIM*shiftidx[iidx];
167 /* Load limits for loop over neighbors */
168 j_index_start = jindex[iidx];
169 j_index_end = jindex[iidx+1];
171 /* Get outer coordinate index */
173 i_coord_offset = DIM*inr;
175 /* Load i particle coords and add shift vector */
176 gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
177 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
179 fix0 = _mm_setzero_pd();
180 fiy0 = _mm_setzero_pd();
181 fiz0 = _mm_setzero_pd();
182 fix1 = _mm_setzero_pd();
183 fiy1 = _mm_setzero_pd();
184 fiz1 = _mm_setzero_pd();
185 fix2 = _mm_setzero_pd();
186 fiy2 = _mm_setzero_pd();
187 fiz2 = _mm_setzero_pd();
189 /* Reset potential sums */
190 velecsum = _mm_setzero_pd();
192 /* Start inner kernel loop */
193 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
196 /* Get j neighbor index, and coordinate index */
199 j_coord_offsetA = DIM*jnrA;
200 j_coord_offsetB = DIM*jnrB;
202 /* load j atom coordinates */
203 gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
204 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
206 /* Calculate displacement vector */
207 dx00 = _mm_sub_pd(ix0,jx0);
208 dy00 = _mm_sub_pd(iy0,jy0);
209 dz00 = _mm_sub_pd(iz0,jz0);
210 dx01 = _mm_sub_pd(ix0,jx1);
211 dy01 = _mm_sub_pd(iy0,jy1);
212 dz01 = _mm_sub_pd(iz0,jz1);
213 dx02 = _mm_sub_pd(ix0,jx2);
214 dy02 = _mm_sub_pd(iy0,jy2);
215 dz02 = _mm_sub_pd(iz0,jz2);
216 dx10 = _mm_sub_pd(ix1,jx0);
217 dy10 = _mm_sub_pd(iy1,jy0);
218 dz10 = _mm_sub_pd(iz1,jz0);
219 dx11 = _mm_sub_pd(ix1,jx1);
220 dy11 = _mm_sub_pd(iy1,jy1);
221 dz11 = _mm_sub_pd(iz1,jz1);
222 dx12 = _mm_sub_pd(ix1,jx2);
223 dy12 = _mm_sub_pd(iy1,jy2);
224 dz12 = _mm_sub_pd(iz1,jz2);
225 dx20 = _mm_sub_pd(ix2,jx0);
226 dy20 = _mm_sub_pd(iy2,jy0);
227 dz20 = _mm_sub_pd(iz2,jz0);
228 dx21 = _mm_sub_pd(ix2,jx1);
229 dy21 = _mm_sub_pd(iy2,jy1);
230 dz21 = _mm_sub_pd(iz2,jz1);
231 dx22 = _mm_sub_pd(ix2,jx2);
232 dy22 = _mm_sub_pd(iy2,jy2);
233 dz22 = _mm_sub_pd(iz2,jz2);
235 /* Calculate squared distance and things based on it */
236 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
237 rsq01 = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
238 rsq02 = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
239 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
240 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
241 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
242 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
243 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
244 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
246 rinv00 = gmx_mm_invsqrt_pd(rsq00);
247 rinv01 = gmx_mm_invsqrt_pd(rsq01);
248 rinv02 = gmx_mm_invsqrt_pd(rsq02);
249 rinv10 = gmx_mm_invsqrt_pd(rsq10);
250 rinv11 = gmx_mm_invsqrt_pd(rsq11);
251 rinv12 = gmx_mm_invsqrt_pd(rsq12);
252 rinv20 = gmx_mm_invsqrt_pd(rsq20);
253 rinv21 = gmx_mm_invsqrt_pd(rsq21);
254 rinv22 = gmx_mm_invsqrt_pd(rsq22);
256 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
257 rinvsq01 = _mm_mul_pd(rinv01,rinv01);
258 rinvsq02 = _mm_mul_pd(rinv02,rinv02);
259 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
260 rinvsq11 = _mm_mul_pd(rinv11,rinv11);
261 rinvsq12 = _mm_mul_pd(rinv12,rinv12);
262 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
263 rinvsq21 = _mm_mul_pd(rinv21,rinv21);
264 rinvsq22 = _mm_mul_pd(rinv22,rinv22);
266 fjx0 = _mm_setzero_pd();
267 fjy0 = _mm_setzero_pd();
268 fjz0 = _mm_setzero_pd();
269 fjx1 = _mm_setzero_pd();
270 fjy1 = _mm_setzero_pd();
271 fjz1 = _mm_setzero_pd();
272 fjx2 = _mm_setzero_pd();
273 fjy2 = _mm_setzero_pd();
274 fjz2 = _mm_setzero_pd();
276 /**************************
277 * CALCULATE INTERACTIONS *
278 **************************/
280 if (gmx_mm_any_lt(rsq00,rcutoff2))
283 r00 = _mm_mul_pd(rsq00,rinv00);
285 /* EWALD ELECTROSTATICS */
287 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
288 ewrt = _mm_mul_pd(r00,ewtabscale);
289 ewitab = _mm_cvttpd_epi32(ewrt);
290 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
291 ewitab = _mm_slli_epi32(ewitab,2);
292 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
293 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
294 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
295 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
296 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
297 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
298 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
299 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
300 velec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_sub_pd(rinv00,sh_ewald),velec));
301 felec = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
303 cutoff_mask = _mm_cmplt_pd(rsq00,rcutoff2);
305 /* Update potential sum for this i atom from the interaction with this j atom. */
306 velec = _mm_and_pd(velec,cutoff_mask);
307 velecsum = _mm_add_pd(velecsum,velec);
311 fscal = _mm_and_pd(fscal,cutoff_mask);
313 /* Calculate temporary vectorial force */
314 tx = _mm_mul_pd(fscal,dx00);
315 ty = _mm_mul_pd(fscal,dy00);
316 tz = _mm_mul_pd(fscal,dz00);
318 /* Update vectorial force */
319 fix0 = _mm_add_pd(fix0,tx);
320 fiy0 = _mm_add_pd(fiy0,ty);
321 fiz0 = _mm_add_pd(fiz0,tz);
323 fjx0 = _mm_add_pd(fjx0,tx);
324 fjy0 = _mm_add_pd(fjy0,ty);
325 fjz0 = _mm_add_pd(fjz0,tz);
329 /**************************
330 * CALCULATE INTERACTIONS *
331 **************************/
333 if (gmx_mm_any_lt(rsq01,rcutoff2))
336 r01 = _mm_mul_pd(rsq01,rinv01);
338 /* EWALD ELECTROSTATICS */
340 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
341 ewrt = _mm_mul_pd(r01,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(qq01,_mm_sub_pd(_mm_sub_pd(rinv01,sh_ewald),velec));
354 felec = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
356 cutoff_mask = _mm_cmplt_pd(rsq01,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,dx01);
368 ty = _mm_mul_pd(fscal,dy01);
369 tz = _mm_mul_pd(fscal,dz01);
371 /* Update vectorial force */
372 fix0 = _mm_add_pd(fix0,tx);
373 fiy0 = _mm_add_pd(fiy0,ty);
374 fiz0 = _mm_add_pd(fiz0,tz);
376 fjx1 = _mm_add_pd(fjx1,tx);
377 fjy1 = _mm_add_pd(fjy1,ty);
378 fjz1 = _mm_add_pd(fjz1,tz);
382 /**************************
383 * CALCULATE INTERACTIONS *
384 **************************/
386 if (gmx_mm_any_lt(rsq02,rcutoff2))
389 r02 = _mm_mul_pd(rsq02,rinv02);
391 /* EWALD ELECTROSTATICS */
393 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
394 ewrt = _mm_mul_pd(r02,ewtabscale);
395 ewitab = _mm_cvttpd_epi32(ewrt);
396 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
397 ewitab = _mm_slli_epi32(ewitab,2);
398 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
399 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
400 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
401 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
402 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
403 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
404 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
405 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
406 velec = _mm_mul_pd(qq02,_mm_sub_pd(_mm_sub_pd(rinv02,sh_ewald),velec));
407 felec = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
409 cutoff_mask = _mm_cmplt_pd(rsq02,rcutoff2);
411 /* Update potential sum for this i atom from the interaction with this j atom. */
412 velec = _mm_and_pd(velec,cutoff_mask);
413 velecsum = _mm_add_pd(velecsum,velec);
417 fscal = _mm_and_pd(fscal,cutoff_mask);
419 /* Calculate temporary vectorial force */
420 tx = _mm_mul_pd(fscal,dx02);
421 ty = _mm_mul_pd(fscal,dy02);
422 tz = _mm_mul_pd(fscal,dz02);
424 /* Update vectorial force */
425 fix0 = _mm_add_pd(fix0,tx);
426 fiy0 = _mm_add_pd(fiy0,ty);
427 fiz0 = _mm_add_pd(fiz0,tz);
429 fjx2 = _mm_add_pd(fjx2,tx);
430 fjy2 = _mm_add_pd(fjy2,ty);
431 fjz2 = _mm_add_pd(fjz2,tz);
435 /**************************
436 * CALCULATE INTERACTIONS *
437 **************************/
439 if (gmx_mm_any_lt(rsq10,rcutoff2))
442 r10 = _mm_mul_pd(rsq10,rinv10);
444 /* EWALD ELECTROSTATICS */
446 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
447 ewrt = _mm_mul_pd(r10,ewtabscale);
448 ewitab = _mm_cvttpd_epi32(ewrt);
449 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
450 ewitab = _mm_slli_epi32(ewitab,2);
451 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
452 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
453 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
454 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
455 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
456 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
457 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
458 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
459 velec = _mm_mul_pd(qq10,_mm_sub_pd(_mm_sub_pd(rinv10,sh_ewald),velec));
460 felec = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
462 cutoff_mask = _mm_cmplt_pd(rsq10,rcutoff2);
464 /* Update potential sum for this i atom from the interaction with this j atom. */
465 velec = _mm_and_pd(velec,cutoff_mask);
466 velecsum = _mm_add_pd(velecsum,velec);
470 fscal = _mm_and_pd(fscal,cutoff_mask);
472 /* Calculate temporary vectorial force */
473 tx = _mm_mul_pd(fscal,dx10);
474 ty = _mm_mul_pd(fscal,dy10);
475 tz = _mm_mul_pd(fscal,dz10);
477 /* Update vectorial force */
478 fix1 = _mm_add_pd(fix1,tx);
479 fiy1 = _mm_add_pd(fiy1,ty);
480 fiz1 = _mm_add_pd(fiz1,tz);
482 fjx0 = _mm_add_pd(fjx0,tx);
483 fjy0 = _mm_add_pd(fjy0,ty);
484 fjz0 = _mm_add_pd(fjz0,tz);
488 /**************************
489 * CALCULATE INTERACTIONS *
490 **************************/
492 if (gmx_mm_any_lt(rsq11,rcutoff2))
495 r11 = _mm_mul_pd(rsq11,rinv11);
497 /* EWALD ELECTROSTATICS */
499 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
500 ewrt = _mm_mul_pd(r11,ewtabscale);
501 ewitab = _mm_cvttpd_epi32(ewrt);
502 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
503 ewitab = _mm_slli_epi32(ewitab,2);
504 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
505 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
506 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
507 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
508 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
509 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
510 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
511 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
512 velec = _mm_mul_pd(qq11,_mm_sub_pd(_mm_sub_pd(rinv11,sh_ewald),velec));
513 felec = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
515 cutoff_mask = _mm_cmplt_pd(rsq11,rcutoff2);
517 /* Update potential sum for this i atom from the interaction with this j atom. */
518 velec = _mm_and_pd(velec,cutoff_mask);
519 velecsum = _mm_add_pd(velecsum,velec);
523 fscal = _mm_and_pd(fscal,cutoff_mask);
525 /* Calculate temporary vectorial force */
526 tx = _mm_mul_pd(fscal,dx11);
527 ty = _mm_mul_pd(fscal,dy11);
528 tz = _mm_mul_pd(fscal,dz11);
530 /* Update vectorial force */
531 fix1 = _mm_add_pd(fix1,tx);
532 fiy1 = _mm_add_pd(fiy1,ty);
533 fiz1 = _mm_add_pd(fiz1,tz);
535 fjx1 = _mm_add_pd(fjx1,tx);
536 fjy1 = _mm_add_pd(fjy1,ty);
537 fjz1 = _mm_add_pd(fjz1,tz);
541 /**************************
542 * CALCULATE INTERACTIONS *
543 **************************/
545 if (gmx_mm_any_lt(rsq12,rcutoff2))
548 r12 = _mm_mul_pd(rsq12,rinv12);
550 /* EWALD ELECTROSTATICS */
552 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
553 ewrt = _mm_mul_pd(r12,ewtabscale);
554 ewitab = _mm_cvttpd_epi32(ewrt);
555 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
556 ewitab = _mm_slli_epi32(ewitab,2);
557 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
558 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
559 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
560 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
561 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
562 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
563 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
564 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
565 velec = _mm_mul_pd(qq12,_mm_sub_pd(_mm_sub_pd(rinv12,sh_ewald),velec));
566 felec = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
568 cutoff_mask = _mm_cmplt_pd(rsq12,rcutoff2);
570 /* Update potential sum for this i atom from the interaction with this j atom. */
571 velec = _mm_and_pd(velec,cutoff_mask);
572 velecsum = _mm_add_pd(velecsum,velec);
576 fscal = _mm_and_pd(fscal,cutoff_mask);
578 /* Calculate temporary vectorial force */
579 tx = _mm_mul_pd(fscal,dx12);
580 ty = _mm_mul_pd(fscal,dy12);
581 tz = _mm_mul_pd(fscal,dz12);
583 /* Update vectorial force */
584 fix1 = _mm_add_pd(fix1,tx);
585 fiy1 = _mm_add_pd(fiy1,ty);
586 fiz1 = _mm_add_pd(fiz1,tz);
588 fjx2 = _mm_add_pd(fjx2,tx);
589 fjy2 = _mm_add_pd(fjy2,ty);
590 fjz2 = _mm_add_pd(fjz2,tz);
594 /**************************
595 * CALCULATE INTERACTIONS *
596 **************************/
598 if (gmx_mm_any_lt(rsq20,rcutoff2))
601 r20 = _mm_mul_pd(rsq20,rinv20);
603 /* EWALD ELECTROSTATICS */
605 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
606 ewrt = _mm_mul_pd(r20,ewtabscale);
607 ewitab = _mm_cvttpd_epi32(ewrt);
608 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
609 ewitab = _mm_slli_epi32(ewitab,2);
610 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
611 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
612 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
613 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
614 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
615 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
616 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
617 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
618 velec = _mm_mul_pd(qq20,_mm_sub_pd(_mm_sub_pd(rinv20,sh_ewald),velec));
619 felec = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
621 cutoff_mask = _mm_cmplt_pd(rsq20,rcutoff2);
623 /* Update potential sum for this i atom from the interaction with this j atom. */
624 velec = _mm_and_pd(velec,cutoff_mask);
625 velecsum = _mm_add_pd(velecsum,velec);
629 fscal = _mm_and_pd(fscal,cutoff_mask);
631 /* Calculate temporary vectorial force */
632 tx = _mm_mul_pd(fscal,dx20);
633 ty = _mm_mul_pd(fscal,dy20);
634 tz = _mm_mul_pd(fscal,dz20);
636 /* Update vectorial force */
637 fix2 = _mm_add_pd(fix2,tx);
638 fiy2 = _mm_add_pd(fiy2,ty);
639 fiz2 = _mm_add_pd(fiz2,tz);
641 fjx0 = _mm_add_pd(fjx0,tx);
642 fjy0 = _mm_add_pd(fjy0,ty);
643 fjz0 = _mm_add_pd(fjz0,tz);
647 /**************************
648 * CALCULATE INTERACTIONS *
649 **************************/
651 if (gmx_mm_any_lt(rsq21,rcutoff2))
654 r21 = _mm_mul_pd(rsq21,rinv21);
656 /* EWALD ELECTROSTATICS */
658 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
659 ewrt = _mm_mul_pd(r21,ewtabscale);
660 ewitab = _mm_cvttpd_epi32(ewrt);
661 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
662 ewitab = _mm_slli_epi32(ewitab,2);
663 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
664 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
665 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
666 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
667 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
668 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
669 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
670 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
671 velec = _mm_mul_pd(qq21,_mm_sub_pd(_mm_sub_pd(rinv21,sh_ewald),velec));
672 felec = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
674 cutoff_mask = _mm_cmplt_pd(rsq21,rcutoff2);
676 /* Update potential sum for this i atom from the interaction with this j atom. */
677 velec = _mm_and_pd(velec,cutoff_mask);
678 velecsum = _mm_add_pd(velecsum,velec);
682 fscal = _mm_and_pd(fscal,cutoff_mask);
684 /* Calculate temporary vectorial force */
685 tx = _mm_mul_pd(fscal,dx21);
686 ty = _mm_mul_pd(fscal,dy21);
687 tz = _mm_mul_pd(fscal,dz21);
689 /* Update vectorial force */
690 fix2 = _mm_add_pd(fix2,tx);
691 fiy2 = _mm_add_pd(fiy2,ty);
692 fiz2 = _mm_add_pd(fiz2,tz);
694 fjx1 = _mm_add_pd(fjx1,tx);
695 fjy1 = _mm_add_pd(fjy1,ty);
696 fjz1 = _mm_add_pd(fjz1,tz);
700 /**************************
701 * CALCULATE INTERACTIONS *
702 **************************/
704 if (gmx_mm_any_lt(rsq22,rcutoff2))
707 r22 = _mm_mul_pd(rsq22,rinv22);
709 /* EWALD ELECTROSTATICS */
711 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
712 ewrt = _mm_mul_pd(r22,ewtabscale);
713 ewitab = _mm_cvttpd_epi32(ewrt);
714 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
715 ewitab = _mm_slli_epi32(ewitab,2);
716 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
717 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
718 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
719 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
720 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
721 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
722 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
723 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
724 velec = _mm_mul_pd(qq22,_mm_sub_pd(_mm_sub_pd(rinv22,sh_ewald),velec));
725 felec = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
727 cutoff_mask = _mm_cmplt_pd(rsq22,rcutoff2);
729 /* Update potential sum for this i atom from the interaction with this j atom. */
730 velec = _mm_and_pd(velec,cutoff_mask);
731 velecsum = _mm_add_pd(velecsum,velec);
735 fscal = _mm_and_pd(fscal,cutoff_mask);
737 /* Calculate temporary vectorial force */
738 tx = _mm_mul_pd(fscal,dx22);
739 ty = _mm_mul_pd(fscal,dy22);
740 tz = _mm_mul_pd(fscal,dz22);
742 /* Update vectorial force */
743 fix2 = _mm_add_pd(fix2,tx);
744 fiy2 = _mm_add_pd(fiy2,ty);
745 fiz2 = _mm_add_pd(fiz2,tz);
747 fjx2 = _mm_add_pd(fjx2,tx);
748 fjy2 = _mm_add_pd(fjy2,ty);
749 fjz2 = _mm_add_pd(fjz2,tz);
753 gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
755 /* Inner loop uses 414 flops */
762 j_coord_offsetA = DIM*jnrA;
764 /* load j atom coordinates */
765 gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
766 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
768 /* Calculate displacement vector */
769 dx00 = _mm_sub_pd(ix0,jx0);
770 dy00 = _mm_sub_pd(iy0,jy0);
771 dz00 = _mm_sub_pd(iz0,jz0);
772 dx01 = _mm_sub_pd(ix0,jx1);
773 dy01 = _mm_sub_pd(iy0,jy1);
774 dz01 = _mm_sub_pd(iz0,jz1);
775 dx02 = _mm_sub_pd(ix0,jx2);
776 dy02 = _mm_sub_pd(iy0,jy2);
777 dz02 = _mm_sub_pd(iz0,jz2);
778 dx10 = _mm_sub_pd(ix1,jx0);
779 dy10 = _mm_sub_pd(iy1,jy0);
780 dz10 = _mm_sub_pd(iz1,jz0);
781 dx11 = _mm_sub_pd(ix1,jx1);
782 dy11 = _mm_sub_pd(iy1,jy1);
783 dz11 = _mm_sub_pd(iz1,jz1);
784 dx12 = _mm_sub_pd(ix1,jx2);
785 dy12 = _mm_sub_pd(iy1,jy2);
786 dz12 = _mm_sub_pd(iz1,jz2);
787 dx20 = _mm_sub_pd(ix2,jx0);
788 dy20 = _mm_sub_pd(iy2,jy0);
789 dz20 = _mm_sub_pd(iz2,jz0);
790 dx21 = _mm_sub_pd(ix2,jx1);
791 dy21 = _mm_sub_pd(iy2,jy1);
792 dz21 = _mm_sub_pd(iz2,jz1);
793 dx22 = _mm_sub_pd(ix2,jx2);
794 dy22 = _mm_sub_pd(iy2,jy2);
795 dz22 = _mm_sub_pd(iz2,jz2);
797 /* Calculate squared distance and things based on it */
798 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
799 rsq01 = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
800 rsq02 = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
801 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
802 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
803 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
804 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
805 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
806 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
808 rinv00 = gmx_mm_invsqrt_pd(rsq00);
809 rinv01 = gmx_mm_invsqrt_pd(rsq01);
810 rinv02 = gmx_mm_invsqrt_pd(rsq02);
811 rinv10 = gmx_mm_invsqrt_pd(rsq10);
812 rinv11 = gmx_mm_invsqrt_pd(rsq11);
813 rinv12 = gmx_mm_invsqrt_pd(rsq12);
814 rinv20 = gmx_mm_invsqrt_pd(rsq20);
815 rinv21 = gmx_mm_invsqrt_pd(rsq21);
816 rinv22 = gmx_mm_invsqrt_pd(rsq22);
818 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
819 rinvsq01 = _mm_mul_pd(rinv01,rinv01);
820 rinvsq02 = _mm_mul_pd(rinv02,rinv02);
821 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
822 rinvsq11 = _mm_mul_pd(rinv11,rinv11);
823 rinvsq12 = _mm_mul_pd(rinv12,rinv12);
824 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
825 rinvsq21 = _mm_mul_pd(rinv21,rinv21);
826 rinvsq22 = _mm_mul_pd(rinv22,rinv22);
828 fjx0 = _mm_setzero_pd();
829 fjy0 = _mm_setzero_pd();
830 fjz0 = _mm_setzero_pd();
831 fjx1 = _mm_setzero_pd();
832 fjy1 = _mm_setzero_pd();
833 fjz1 = _mm_setzero_pd();
834 fjx2 = _mm_setzero_pd();
835 fjy2 = _mm_setzero_pd();
836 fjz2 = _mm_setzero_pd();
838 /**************************
839 * CALCULATE INTERACTIONS *
840 **************************/
842 if (gmx_mm_any_lt(rsq00,rcutoff2))
845 r00 = _mm_mul_pd(rsq00,rinv00);
847 /* EWALD ELECTROSTATICS */
849 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
850 ewrt = _mm_mul_pd(r00,ewtabscale);
851 ewitab = _mm_cvttpd_epi32(ewrt);
852 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
853 ewitab = _mm_slli_epi32(ewitab,2);
854 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
855 ewtabD = _mm_setzero_pd();
856 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
857 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
858 ewtabFn = _mm_setzero_pd();
859 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
860 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
861 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
862 velec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_sub_pd(rinv00,sh_ewald),velec));
863 felec = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
865 cutoff_mask = _mm_cmplt_pd(rsq00,rcutoff2);
867 /* Update potential sum for this i atom from the interaction with this j atom. */
868 velec = _mm_and_pd(velec,cutoff_mask);
869 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
870 velecsum = _mm_add_pd(velecsum,velec);
874 fscal = _mm_and_pd(fscal,cutoff_mask);
876 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
878 /* Calculate temporary vectorial force */
879 tx = _mm_mul_pd(fscal,dx00);
880 ty = _mm_mul_pd(fscal,dy00);
881 tz = _mm_mul_pd(fscal,dz00);
883 /* Update vectorial force */
884 fix0 = _mm_add_pd(fix0,tx);
885 fiy0 = _mm_add_pd(fiy0,ty);
886 fiz0 = _mm_add_pd(fiz0,tz);
888 fjx0 = _mm_add_pd(fjx0,tx);
889 fjy0 = _mm_add_pd(fjy0,ty);
890 fjz0 = _mm_add_pd(fjz0,tz);
894 /**************************
895 * CALCULATE INTERACTIONS *
896 **************************/
898 if (gmx_mm_any_lt(rsq01,rcutoff2))
901 r01 = _mm_mul_pd(rsq01,rinv01);
903 /* EWALD ELECTROSTATICS */
905 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
906 ewrt = _mm_mul_pd(r01,ewtabscale);
907 ewitab = _mm_cvttpd_epi32(ewrt);
908 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
909 ewitab = _mm_slli_epi32(ewitab,2);
910 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
911 ewtabD = _mm_setzero_pd();
912 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
913 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
914 ewtabFn = _mm_setzero_pd();
915 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
916 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
917 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
918 velec = _mm_mul_pd(qq01,_mm_sub_pd(_mm_sub_pd(rinv01,sh_ewald),velec));
919 felec = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
921 cutoff_mask = _mm_cmplt_pd(rsq01,rcutoff2);
923 /* Update potential sum for this i atom from the interaction with this j atom. */
924 velec = _mm_and_pd(velec,cutoff_mask);
925 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
926 velecsum = _mm_add_pd(velecsum,velec);
930 fscal = _mm_and_pd(fscal,cutoff_mask);
932 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
934 /* Calculate temporary vectorial force */
935 tx = _mm_mul_pd(fscal,dx01);
936 ty = _mm_mul_pd(fscal,dy01);
937 tz = _mm_mul_pd(fscal,dz01);
939 /* Update vectorial force */
940 fix0 = _mm_add_pd(fix0,tx);
941 fiy0 = _mm_add_pd(fiy0,ty);
942 fiz0 = _mm_add_pd(fiz0,tz);
944 fjx1 = _mm_add_pd(fjx1,tx);
945 fjy1 = _mm_add_pd(fjy1,ty);
946 fjz1 = _mm_add_pd(fjz1,tz);
950 /**************************
951 * CALCULATE INTERACTIONS *
952 **************************/
954 if (gmx_mm_any_lt(rsq02,rcutoff2))
957 r02 = _mm_mul_pd(rsq02,rinv02);
959 /* EWALD ELECTROSTATICS */
961 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
962 ewrt = _mm_mul_pd(r02,ewtabscale);
963 ewitab = _mm_cvttpd_epi32(ewrt);
964 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
965 ewitab = _mm_slli_epi32(ewitab,2);
966 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
967 ewtabD = _mm_setzero_pd();
968 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
969 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
970 ewtabFn = _mm_setzero_pd();
971 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
972 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
973 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
974 velec = _mm_mul_pd(qq02,_mm_sub_pd(_mm_sub_pd(rinv02,sh_ewald),velec));
975 felec = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
977 cutoff_mask = _mm_cmplt_pd(rsq02,rcutoff2);
979 /* Update potential sum for this i atom from the interaction with this j atom. */
980 velec = _mm_and_pd(velec,cutoff_mask);
981 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
982 velecsum = _mm_add_pd(velecsum,velec);
986 fscal = _mm_and_pd(fscal,cutoff_mask);
988 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
990 /* Calculate temporary vectorial force */
991 tx = _mm_mul_pd(fscal,dx02);
992 ty = _mm_mul_pd(fscal,dy02);
993 tz = _mm_mul_pd(fscal,dz02);
995 /* Update vectorial force */
996 fix0 = _mm_add_pd(fix0,tx);
997 fiy0 = _mm_add_pd(fiy0,ty);
998 fiz0 = _mm_add_pd(fiz0,tz);
1000 fjx2 = _mm_add_pd(fjx2,tx);
1001 fjy2 = _mm_add_pd(fjy2,ty);
1002 fjz2 = _mm_add_pd(fjz2,tz);
1006 /**************************
1007 * CALCULATE INTERACTIONS *
1008 **************************/
1010 if (gmx_mm_any_lt(rsq10,rcutoff2))
1013 r10 = _mm_mul_pd(rsq10,rinv10);
1015 /* EWALD ELECTROSTATICS */
1017 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1018 ewrt = _mm_mul_pd(r10,ewtabscale);
1019 ewitab = _mm_cvttpd_epi32(ewrt);
1020 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1021 ewitab = _mm_slli_epi32(ewitab,2);
1022 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1023 ewtabD = _mm_setzero_pd();
1024 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1025 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1026 ewtabFn = _mm_setzero_pd();
1027 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1028 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1029 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1030 velec = _mm_mul_pd(qq10,_mm_sub_pd(_mm_sub_pd(rinv10,sh_ewald),velec));
1031 felec = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
1033 cutoff_mask = _mm_cmplt_pd(rsq10,rcutoff2);
1035 /* Update potential sum for this i atom from the interaction with this j atom. */
1036 velec = _mm_and_pd(velec,cutoff_mask);
1037 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1038 velecsum = _mm_add_pd(velecsum,velec);
1042 fscal = _mm_and_pd(fscal,cutoff_mask);
1044 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1046 /* Calculate temporary vectorial force */
1047 tx = _mm_mul_pd(fscal,dx10);
1048 ty = _mm_mul_pd(fscal,dy10);
1049 tz = _mm_mul_pd(fscal,dz10);
1051 /* Update vectorial force */
1052 fix1 = _mm_add_pd(fix1,tx);
1053 fiy1 = _mm_add_pd(fiy1,ty);
1054 fiz1 = _mm_add_pd(fiz1,tz);
1056 fjx0 = _mm_add_pd(fjx0,tx);
1057 fjy0 = _mm_add_pd(fjy0,ty);
1058 fjz0 = _mm_add_pd(fjz0,tz);
1062 /**************************
1063 * CALCULATE INTERACTIONS *
1064 **************************/
1066 if (gmx_mm_any_lt(rsq11,rcutoff2))
1069 r11 = _mm_mul_pd(rsq11,rinv11);
1071 /* EWALD ELECTROSTATICS */
1073 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1074 ewrt = _mm_mul_pd(r11,ewtabscale);
1075 ewitab = _mm_cvttpd_epi32(ewrt);
1076 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1077 ewitab = _mm_slli_epi32(ewitab,2);
1078 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1079 ewtabD = _mm_setzero_pd();
1080 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1081 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1082 ewtabFn = _mm_setzero_pd();
1083 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1084 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1085 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1086 velec = _mm_mul_pd(qq11,_mm_sub_pd(_mm_sub_pd(rinv11,sh_ewald),velec));
1087 felec = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
1089 cutoff_mask = _mm_cmplt_pd(rsq11,rcutoff2);
1091 /* Update potential sum for this i atom from the interaction with this j atom. */
1092 velec = _mm_and_pd(velec,cutoff_mask);
1093 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1094 velecsum = _mm_add_pd(velecsum,velec);
1098 fscal = _mm_and_pd(fscal,cutoff_mask);
1100 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1102 /* Calculate temporary vectorial force */
1103 tx = _mm_mul_pd(fscal,dx11);
1104 ty = _mm_mul_pd(fscal,dy11);
1105 tz = _mm_mul_pd(fscal,dz11);
1107 /* Update vectorial force */
1108 fix1 = _mm_add_pd(fix1,tx);
1109 fiy1 = _mm_add_pd(fiy1,ty);
1110 fiz1 = _mm_add_pd(fiz1,tz);
1112 fjx1 = _mm_add_pd(fjx1,tx);
1113 fjy1 = _mm_add_pd(fjy1,ty);
1114 fjz1 = _mm_add_pd(fjz1,tz);
1118 /**************************
1119 * CALCULATE INTERACTIONS *
1120 **************************/
1122 if (gmx_mm_any_lt(rsq12,rcutoff2))
1125 r12 = _mm_mul_pd(rsq12,rinv12);
1127 /* EWALD ELECTROSTATICS */
1129 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1130 ewrt = _mm_mul_pd(r12,ewtabscale);
1131 ewitab = _mm_cvttpd_epi32(ewrt);
1132 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1133 ewitab = _mm_slli_epi32(ewitab,2);
1134 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1135 ewtabD = _mm_setzero_pd();
1136 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1137 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1138 ewtabFn = _mm_setzero_pd();
1139 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1140 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1141 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1142 velec = _mm_mul_pd(qq12,_mm_sub_pd(_mm_sub_pd(rinv12,sh_ewald),velec));
1143 felec = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
1145 cutoff_mask = _mm_cmplt_pd(rsq12,rcutoff2);
1147 /* Update potential sum for this i atom from the interaction with this j atom. */
1148 velec = _mm_and_pd(velec,cutoff_mask);
1149 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1150 velecsum = _mm_add_pd(velecsum,velec);
1154 fscal = _mm_and_pd(fscal,cutoff_mask);
1156 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1158 /* Calculate temporary vectorial force */
1159 tx = _mm_mul_pd(fscal,dx12);
1160 ty = _mm_mul_pd(fscal,dy12);
1161 tz = _mm_mul_pd(fscal,dz12);
1163 /* Update vectorial force */
1164 fix1 = _mm_add_pd(fix1,tx);
1165 fiy1 = _mm_add_pd(fiy1,ty);
1166 fiz1 = _mm_add_pd(fiz1,tz);
1168 fjx2 = _mm_add_pd(fjx2,tx);
1169 fjy2 = _mm_add_pd(fjy2,ty);
1170 fjz2 = _mm_add_pd(fjz2,tz);
1174 /**************************
1175 * CALCULATE INTERACTIONS *
1176 **************************/
1178 if (gmx_mm_any_lt(rsq20,rcutoff2))
1181 r20 = _mm_mul_pd(rsq20,rinv20);
1183 /* EWALD ELECTROSTATICS */
1185 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1186 ewrt = _mm_mul_pd(r20,ewtabscale);
1187 ewitab = _mm_cvttpd_epi32(ewrt);
1188 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1189 ewitab = _mm_slli_epi32(ewitab,2);
1190 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1191 ewtabD = _mm_setzero_pd();
1192 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1193 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1194 ewtabFn = _mm_setzero_pd();
1195 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1196 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1197 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1198 velec = _mm_mul_pd(qq20,_mm_sub_pd(_mm_sub_pd(rinv20,sh_ewald),velec));
1199 felec = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
1201 cutoff_mask = _mm_cmplt_pd(rsq20,rcutoff2);
1203 /* Update potential sum for this i atom from the interaction with this j atom. */
1204 velec = _mm_and_pd(velec,cutoff_mask);
1205 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1206 velecsum = _mm_add_pd(velecsum,velec);
1210 fscal = _mm_and_pd(fscal,cutoff_mask);
1212 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1214 /* Calculate temporary vectorial force */
1215 tx = _mm_mul_pd(fscal,dx20);
1216 ty = _mm_mul_pd(fscal,dy20);
1217 tz = _mm_mul_pd(fscal,dz20);
1219 /* Update vectorial force */
1220 fix2 = _mm_add_pd(fix2,tx);
1221 fiy2 = _mm_add_pd(fiy2,ty);
1222 fiz2 = _mm_add_pd(fiz2,tz);
1224 fjx0 = _mm_add_pd(fjx0,tx);
1225 fjy0 = _mm_add_pd(fjy0,ty);
1226 fjz0 = _mm_add_pd(fjz0,tz);
1230 /**************************
1231 * CALCULATE INTERACTIONS *
1232 **************************/
1234 if (gmx_mm_any_lt(rsq21,rcutoff2))
1237 r21 = _mm_mul_pd(rsq21,rinv21);
1239 /* EWALD ELECTROSTATICS */
1241 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1242 ewrt = _mm_mul_pd(r21,ewtabscale);
1243 ewitab = _mm_cvttpd_epi32(ewrt);
1244 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1245 ewitab = _mm_slli_epi32(ewitab,2);
1246 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1247 ewtabD = _mm_setzero_pd();
1248 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1249 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1250 ewtabFn = _mm_setzero_pd();
1251 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1252 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1253 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1254 velec = _mm_mul_pd(qq21,_mm_sub_pd(_mm_sub_pd(rinv21,sh_ewald),velec));
1255 felec = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
1257 cutoff_mask = _mm_cmplt_pd(rsq21,rcutoff2);
1259 /* Update potential sum for this i atom from the interaction with this j atom. */
1260 velec = _mm_and_pd(velec,cutoff_mask);
1261 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1262 velecsum = _mm_add_pd(velecsum,velec);
1266 fscal = _mm_and_pd(fscal,cutoff_mask);
1268 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1270 /* Calculate temporary vectorial force */
1271 tx = _mm_mul_pd(fscal,dx21);
1272 ty = _mm_mul_pd(fscal,dy21);
1273 tz = _mm_mul_pd(fscal,dz21);
1275 /* Update vectorial force */
1276 fix2 = _mm_add_pd(fix2,tx);
1277 fiy2 = _mm_add_pd(fiy2,ty);
1278 fiz2 = _mm_add_pd(fiz2,tz);
1280 fjx1 = _mm_add_pd(fjx1,tx);
1281 fjy1 = _mm_add_pd(fjy1,ty);
1282 fjz1 = _mm_add_pd(fjz1,tz);
1286 /**************************
1287 * CALCULATE INTERACTIONS *
1288 **************************/
1290 if (gmx_mm_any_lt(rsq22,rcutoff2))
1293 r22 = _mm_mul_pd(rsq22,rinv22);
1295 /* EWALD ELECTROSTATICS */
1297 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1298 ewrt = _mm_mul_pd(r22,ewtabscale);
1299 ewitab = _mm_cvttpd_epi32(ewrt);
1300 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1301 ewitab = _mm_slli_epi32(ewitab,2);
1302 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1303 ewtabD = _mm_setzero_pd();
1304 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1305 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1306 ewtabFn = _mm_setzero_pd();
1307 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1308 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1309 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1310 velec = _mm_mul_pd(qq22,_mm_sub_pd(_mm_sub_pd(rinv22,sh_ewald),velec));
1311 felec = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
1313 cutoff_mask = _mm_cmplt_pd(rsq22,rcutoff2);
1315 /* Update potential sum for this i atom from the interaction with this j atom. */
1316 velec = _mm_and_pd(velec,cutoff_mask);
1317 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1318 velecsum = _mm_add_pd(velecsum,velec);
1322 fscal = _mm_and_pd(fscal,cutoff_mask);
1324 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1326 /* Calculate temporary vectorial force */
1327 tx = _mm_mul_pd(fscal,dx22);
1328 ty = _mm_mul_pd(fscal,dy22);
1329 tz = _mm_mul_pd(fscal,dz22);
1331 /* Update vectorial force */
1332 fix2 = _mm_add_pd(fix2,tx);
1333 fiy2 = _mm_add_pd(fiy2,ty);
1334 fiz2 = _mm_add_pd(fiz2,tz);
1336 fjx2 = _mm_add_pd(fjx2,tx);
1337 fjy2 = _mm_add_pd(fjy2,ty);
1338 fjz2 = _mm_add_pd(fjz2,tz);
1342 gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1344 /* Inner loop uses 414 flops */
1347 /* End of innermost loop */
1349 gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1350 f+i_coord_offset,fshift+i_shift_offset);
1353 /* Update potential energies */
1354 gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1356 /* Increment number of inner iterations */
1357 inneriter += j_index_end - j_index_start;
1359 /* Outer loop uses 19 flops */
1362 /* Increment number of outer iterations */
1365 /* Update outer/inner flops */
1367 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3W3_VF,outeriter*19 + inneriter*414);
1370 * Gromacs nonbonded kernel: nb_kernel_ElecEwSh_VdwNone_GeomW3W3_F_sse2_double
1371 * Electrostatics interaction: Ewald
1372 * VdW interaction: None
1373 * Geometry: Water3-Water3
1374 * Calculate force/pot: Force
1377 nb_kernel_ElecEwSh_VdwNone_GeomW3W3_F_sse2_double
1378 (t_nblist * gmx_restrict nlist,
1379 rvec * gmx_restrict xx,
1380 rvec * gmx_restrict ff,
1381 t_forcerec * gmx_restrict fr,
1382 t_mdatoms * gmx_restrict mdatoms,
1383 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1384 t_nrnb * gmx_restrict nrnb)
1386 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1387 * just 0 for non-waters.
1388 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
1389 * jnr indices corresponding to data put in the four positions in the SIMD register.
1391 int i_shift_offset,i_coord_offset,outeriter,inneriter;
1392 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1394 int j_coord_offsetA,j_coord_offsetB;
1395 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
1396 real rcutoff_scalar;
1397 real *shiftvec,*fshift,*x,*f;
1398 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1400 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1402 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1404 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1405 int vdwjidx0A,vdwjidx0B;
1406 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1407 int vdwjidx1A,vdwjidx1B;
1408 __m128d jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1409 int vdwjidx2A,vdwjidx2B;
1410 __m128d jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1411 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1412 __m128d dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1413 __m128d dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1414 __m128d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1415 __m128d dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1416 __m128d dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1417 __m128d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1418 __m128d dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1419 __m128d dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1420 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
1423 __m128d ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1425 __m128d dummy_mask,cutoff_mask;
1426 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
1427 __m128d one = _mm_set1_pd(1.0);
1428 __m128d two = _mm_set1_pd(2.0);
1434 jindex = nlist->jindex;
1436 shiftidx = nlist->shift;
1438 shiftvec = fr->shift_vec[0];
1439 fshift = fr->fshift[0];
1440 facel = _mm_set1_pd(fr->epsfac);
1441 charge = mdatoms->chargeA;
1443 sh_ewald = _mm_set1_pd(fr->ic->sh_ewald);
1444 ewtab = fr->ic->tabq_coul_F;
1445 ewtabscale = _mm_set1_pd(fr->ic->tabq_scale);
1446 ewtabhalfspace = _mm_set1_pd(0.5/fr->ic->tabq_scale);
1448 /* Setup water-specific parameters */
1449 inr = nlist->iinr[0];
1450 iq0 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
1451 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
1452 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
1454 jq0 = _mm_set1_pd(charge[inr+0]);
1455 jq1 = _mm_set1_pd(charge[inr+1]);
1456 jq2 = _mm_set1_pd(charge[inr+2]);
1457 qq00 = _mm_mul_pd(iq0,jq0);
1458 qq01 = _mm_mul_pd(iq0,jq1);
1459 qq02 = _mm_mul_pd(iq0,jq2);
1460 qq10 = _mm_mul_pd(iq1,jq0);
1461 qq11 = _mm_mul_pd(iq1,jq1);
1462 qq12 = _mm_mul_pd(iq1,jq2);
1463 qq20 = _mm_mul_pd(iq2,jq0);
1464 qq21 = _mm_mul_pd(iq2,jq1);
1465 qq22 = _mm_mul_pd(iq2,jq2);
1467 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
1468 rcutoff_scalar = fr->rcoulomb;
1469 rcutoff = _mm_set1_pd(rcutoff_scalar);
1470 rcutoff2 = _mm_mul_pd(rcutoff,rcutoff);
1472 /* Avoid stupid compiler warnings */
1474 j_coord_offsetA = 0;
1475 j_coord_offsetB = 0;
1480 /* Start outer loop over neighborlists */
1481 for(iidx=0; iidx<nri; iidx++)
1483 /* Load shift vector for this list */
1484 i_shift_offset = DIM*shiftidx[iidx];
1486 /* Load limits for loop over neighbors */
1487 j_index_start = jindex[iidx];
1488 j_index_end = jindex[iidx+1];
1490 /* Get outer coordinate index */
1492 i_coord_offset = DIM*inr;
1494 /* Load i particle coords and add shift vector */
1495 gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1496 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1498 fix0 = _mm_setzero_pd();
1499 fiy0 = _mm_setzero_pd();
1500 fiz0 = _mm_setzero_pd();
1501 fix1 = _mm_setzero_pd();
1502 fiy1 = _mm_setzero_pd();
1503 fiz1 = _mm_setzero_pd();
1504 fix2 = _mm_setzero_pd();
1505 fiy2 = _mm_setzero_pd();
1506 fiz2 = _mm_setzero_pd();
1508 /* Start inner kernel loop */
1509 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
1512 /* Get j neighbor index, and coordinate index */
1514 jnrB = jjnr[jidx+1];
1515 j_coord_offsetA = DIM*jnrA;
1516 j_coord_offsetB = DIM*jnrB;
1518 /* load j atom coordinates */
1519 gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1520 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1522 /* Calculate displacement vector */
1523 dx00 = _mm_sub_pd(ix0,jx0);
1524 dy00 = _mm_sub_pd(iy0,jy0);
1525 dz00 = _mm_sub_pd(iz0,jz0);
1526 dx01 = _mm_sub_pd(ix0,jx1);
1527 dy01 = _mm_sub_pd(iy0,jy1);
1528 dz01 = _mm_sub_pd(iz0,jz1);
1529 dx02 = _mm_sub_pd(ix0,jx2);
1530 dy02 = _mm_sub_pd(iy0,jy2);
1531 dz02 = _mm_sub_pd(iz0,jz2);
1532 dx10 = _mm_sub_pd(ix1,jx0);
1533 dy10 = _mm_sub_pd(iy1,jy0);
1534 dz10 = _mm_sub_pd(iz1,jz0);
1535 dx11 = _mm_sub_pd(ix1,jx1);
1536 dy11 = _mm_sub_pd(iy1,jy1);
1537 dz11 = _mm_sub_pd(iz1,jz1);
1538 dx12 = _mm_sub_pd(ix1,jx2);
1539 dy12 = _mm_sub_pd(iy1,jy2);
1540 dz12 = _mm_sub_pd(iz1,jz2);
1541 dx20 = _mm_sub_pd(ix2,jx0);
1542 dy20 = _mm_sub_pd(iy2,jy0);
1543 dz20 = _mm_sub_pd(iz2,jz0);
1544 dx21 = _mm_sub_pd(ix2,jx1);
1545 dy21 = _mm_sub_pd(iy2,jy1);
1546 dz21 = _mm_sub_pd(iz2,jz1);
1547 dx22 = _mm_sub_pd(ix2,jx2);
1548 dy22 = _mm_sub_pd(iy2,jy2);
1549 dz22 = _mm_sub_pd(iz2,jz2);
1551 /* Calculate squared distance and things based on it */
1552 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1553 rsq01 = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
1554 rsq02 = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
1555 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
1556 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1557 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1558 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
1559 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1560 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1562 rinv00 = gmx_mm_invsqrt_pd(rsq00);
1563 rinv01 = gmx_mm_invsqrt_pd(rsq01);
1564 rinv02 = gmx_mm_invsqrt_pd(rsq02);
1565 rinv10 = gmx_mm_invsqrt_pd(rsq10);
1566 rinv11 = gmx_mm_invsqrt_pd(rsq11);
1567 rinv12 = gmx_mm_invsqrt_pd(rsq12);
1568 rinv20 = gmx_mm_invsqrt_pd(rsq20);
1569 rinv21 = gmx_mm_invsqrt_pd(rsq21);
1570 rinv22 = gmx_mm_invsqrt_pd(rsq22);
1572 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
1573 rinvsq01 = _mm_mul_pd(rinv01,rinv01);
1574 rinvsq02 = _mm_mul_pd(rinv02,rinv02);
1575 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
1576 rinvsq11 = _mm_mul_pd(rinv11,rinv11);
1577 rinvsq12 = _mm_mul_pd(rinv12,rinv12);
1578 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
1579 rinvsq21 = _mm_mul_pd(rinv21,rinv21);
1580 rinvsq22 = _mm_mul_pd(rinv22,rinv22);
1582 fjx0 = _mm_setzero_pd();
1583 fjy0 = _mm_setzero_pd();
1584 fjz0 = _mm_setzero_pd();
1585 fjx1 = _mm_setzero_pd();
1586 fjy1 = _mm_setzero_pd();
1587 fjz1 = _mm_setzero_pd();
1588 fjx2 = _mm_setzero_pd();
1589 fjy2 = _mm_setzero_pd();
1590 fjz2 = _mm_setzero_pd();
1592 /**************************
1593 * CALCULATE INTERACTIONS *
1594 **************************/
1596 if (gmx_mm_any_lt(rsq00,rcutoff2))
1599 r00 = _mm_mul_pd(rsq00,rinv00);
1601 /* EWALD ELECTROSTATICS */
1603 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1604 ewrt = _mm_mul_pd(r00,ewtabscale);
1605 ewitab = _mm_cvttpd_epi32(ewrt);
1606 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1607 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1609 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1610 felec = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
1612 cutoff_mask = _mm_cmplt_pd(rsq00,rcutoff2);
1616 fscal = _mm_and_pd(fscal,cutoff_mask);
1618 /* Calculate temporary vectorial force */
1619 tx = _mm_mul_pd(fscal,dx00);
1620 ty = _mm_mul_pd(fscal,dy00);
1621 tz = _mm_mul_pd(fscal,dz00);
1623 /* Update vectorial force */
1624 fix0 = _mm_add_pd(fix0,tx);
1625 fiy0 = _mm_add_pd(fiy0,ty);
1626 fiz0 = _mm_add_pd(fiz0,tz);
1628 fjx0 = _mm_add_pd(fjx0,tx);
1629 fjy0 = _mm_add_pd(fjy0,ty);
1630 fjz0 = _mm_add_pd(fjz0,tz);
1634 /**************************
1635 * CALCULATE INTERACTIONS *
1636 **************************/
1638 if (gmx_mm_any_lt(rsq01,rcutoff2))
1641 r01 = _mm_mul_pd(rsq01,rinv01);
1643 /* EWALD ELECTROSTATICS */
1645 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1646 ewrt = _mm_mul_pd(r01,ewtabscale);
1647 ewitab = _mm_cvttpd_epi32(ewrt);
1648 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1649 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1651 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1652 felec = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
1654 cutoff_mask = _mm_cmplt_pd(rsq01,rcutoff2);
1658 fscal = _mm_and_pd(fscal,cutoff_mask);
1660 /* Calculate temporary vectorial force */
1661 tx = _mm_mul_pd(fscal,dx01);
1662 ty = _mm_mul_pd(fscal,dy01);
1663 tz = _mm_mul_pd(fscal,dz01);
1665 /* Update vectorial force */
1666 fix0 = _mm_add_pd(fix0,tx);
1667 fiy0 = _mm_add_pd(fiy0,ty);
1668 fiz0 = _mm_add_pd(fiz0,tz);
1670 fjx1 = _mm_add_pd(fjx1,tx);
1671 fjy1 = _mm_add_pd(fjy1,ty);
1672 fjz1 = _mm_add_pd(fjz1,tz);
1676 /**************************
1677 * CALCULATE INTERACTIONS *
1678 **************************/
1680 if (gmx_mm_any_lt(rsq02,rcutoff2))
1683 r02 = _mm_mul_pd(rsq02,rinv02);
1685 /* EWALD ELECTROSTATICS */
1687 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1688 ewrt = _mm_mul_pd(r02,ewtabscale);
1689 ewitab = _mm_cvttpd_epi32(ewrt);
1690 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1691 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1693 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1694 felec = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
1696 cutoff_mask = _mm_cmplt_pd(rsq02,rcutoff2);
1700 fscal = _mm_and_pd(fscal,cutoff_mask);
1702 /* Calculate temporary vectorial force */
1703 tx = _mm_mul_pd(fscal,dx02);
1704 ty = _mm_mul_pd(fscal,dy02);
1705 tz = _mm_mul_pd(fscal,dz02);
1707 /* Update vectorial force */
1708 fix0 = _mm_add_pd(fix0,tx);
1709 fiy0 = _mm_add_pd(fiy0,ty);
1710 fiz0 = _mm_add_pd(fiz0,tz);
1712 fjx2 = _mm_add_pd(fjx2,tx);
1713 fjy2 = _mm_add_pd(fjy2,ty);
1714 fjz2 = _mm_add_pd(fjz2,tz);
1718 /**************************
1719 * CALCULATE INTERACTIONS *
1720 **************************/
1722 if (gmx_mm_any_lt(rsq10,rcutoff2))
1725 r10 = _mm_mul_pd(rsq10,rinv10);
1727 /* EWALD ELECTROSTATICS */
1729 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1730 ewrt = _mm_mul_pd(r10,ewtabscale);
1731 ewitab = _mm_cvttpd_epi32(ewrt);
1732 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1733 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1735 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1736 felec = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
1738 cutoff_mask = _mm_cmplt_pd(rsq10,rcutoff2);
1742 fscal = _mm_and_pd(fscal,cutoff_mask);
1744 /* Calculate temporary vectorial force */
1745 tx = _mm_mul_pd(fscal,dx10);
1746 ty = _mm_mul_pd(fscal,dy10);
1747 tz = _mm_mul_pd(fscal,dz10);
1749 /* Update vectorial force */
1750 fix1 = _mm_add_pd(fix1,tx);
1751 fiy1 = _mm_add_pd(fiy1,ty);
1752 fiz1 = _mm_add_pd(fiz1,tz);
1754 fjx0 = _mm_add_pd(fjx0,tx);
1755 fjy0 = _mm_add_pd(fjy0,ty);
1756 fjz0 = _mm_add_pd(fjz0,tz);
1760 /**************************
1761 * CALCULATE INTERACTIONS *
1762 **************************/
1764 if (gmx_mm_any_lt(rsq11,rcutoff2))
1767 r11 = _mm_mul_pd(rsq11,rinv11);
1769 /* EWALD ELECTROSTATICS */
1771 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1772 ewrt = _mm_mul_pd(r11,ewtabscale);
1773 ewitab = _mm_cvttpd_epi32(ewrt);
1774 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1775 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1777 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1778 felec = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
1780 cutoff_mask = _mm_cmplt_pd(rsq11,rcutoff2);
1784 fscal = _mm_and_pd(fscal,cutoff_mask);
1786 /* Calculate temporary vectorial force */
1787 tx = _mm_mul_pd(fscal,dx11);
1788 ty = _mm_mul_pd(fscal,dy11);
1789 tz = _mm_mul_pd(fscal,dz11);
1791 /* Update vectorial force */
1792 fix1 = _mm_add_pd(fix1,tx);
1793 fiy1 = _mm_add_pd(fiy1,ty);
1794 fiz1 = _mm_add_pd(fiz1,tz);
1796 fjx1 = _mm_add_pd(fjx1,tx);
1797 fjy1 = _mm_add_pd(fjy1,ty);
1798 fjz1 = _mm_add_pd(fjz1,tz);
1802 /**************************
1803 * CALCULATE INTERACTIONS *
1804 **************************/
1806 if (gmx_mm_any_lt(rsq12,rcutoff2))
1809 r12 = _mm_mul_pd(rsq12,rinv12);
1811 /* EWALD ELECTROSTATICS */
1813 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1814 ewrt = _mm_mul_pd(r12,ewtabscale);
1815 ewitab = _mm_cvttpd_epi32(ewrt);
1816 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1817 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1819 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1820 felec = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
1822 cutoff_mask = _mm_cmplt_pd(rsq12,rcutoff2);
1826 fscal = _mm_and_pd(fscal,cutoff_mask);
1828 /* Calculate temporary vectorial force */
1829 tx = _mm_mul_pd(fscal,dx12);
1830 ty = _mm_mul_pd(fscal,dy12);
1831 tz = _mm_mul_pd(fscal,dz12);
1833 /* Update vectorial force */
1834 fix1 = _mm_add_pd(fix1,tx);
1835 fiy1 = _mm_add_pd(fiy1,ty);
1836 fiz1 = _mm_add_pd(fiz1,tz);
1838 fjx2 = _mm_add_pd(fjx2,tx);
1839 fjy2 = _mm_add_pd(fjy2,ty);
1840 fjz2 = _mm_add_pd(fjz2,tz);
1844 /**************************
1845 * CALCULATE INTERACTIONS *
1846 **************************/
1848 if (gmx_mm_any_lt(rsq20,rcutoff2))
1851 r20 = _mm_mul_pd(rsq20,rinv20);
1853 /* EWALD ELECTROSTATICS */
1855 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1856 ewrt = _mm_mul_pd(r20,ewtabscale);
1857 ewitab = _mm_cvttpd_epi32(ewrt);
1858 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1859 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1861 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1862 felec = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
1864 cutoff_mask = _mm_cmplt_pd(rsq20,rcutoff2);
1868 fscal = _mm_and_pd(fscal,cutoff_mask);
1870 /* Calculate temporary vectorial force */
1871 tx = _mm_mul_pd(fscal,dx20);
1872 ty = _mm_mul_pd(fscal,dy20);
1873 tz = _mm_mul_pd(fscal,dz20);
1875 /* Update vectorial force */
1876 fix2 = _mm_add_pd(fix2,tx);
1877 fiy2 = _mm_add_pd(fiy2,ty);
1878 fiz2 = _mm_add_pd(fiz2,tz);
1880 fjx0 = _mm_add_pd(fjx0,tx);
1881 fjy0 = _mm_add_pd(fjy0,ty);
1882 fjz0 = _mm_add_pd(fjz0,tz);
1886 /**************************
1887 * CALCULATE INTERACTIONS *
1888 **************************/
1890 if (gmx_mm_any_lt(rsq21,rcutoff2))
1893 r21 = _mm_mul_pd(rsq21,rinv21);
1895 /* EWALD ELECTROSTATICS */
1897 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1898 ewrt = _mm_mul_pd(r21,ewtabscale);
1899 ewitab = _mm_cvttpd_epi32(ewrt);
1900 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1901 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1903 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1904 felec = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
1906 cutoff_mask = _mm_cmplt_pd(rsq21,rcutoff2);
1910 fscal = _mm_and_pd(fscal,cutoff_mask);
1912 /* Calculate temporary vectorial force */
1913 tx = _mm_mul_pd(fscal,dx21);
1914 ty = _mm_mul_pd(fscal,dy21);
1915 tz = _mm_mul_pd(fscal,dz21);
1917 /* Update vectorial force */
1918 fix2 = _mm_add_pd(fix2,tx);
1919 fiy2 = _mm_add_pd(fiy2,ty);
1920 fiz2 = _mm_add_pd(fiz2,tz);
1922 fjx1 = _mm_add_pd(fjx1,tx);
1923 fjy1 = _mm_add_pd(fjy1,ty);
1924 fjz1 = _mm_add_pd(fjz1,tz);
1928 /**************************
1929 * CALCULATE INTERACTIONS *
1930 **************************/
1932 if (gmx_mm_any_lt(rsq22,rcutoff2))
1935 r22 = _mm_mul_pd(rsq22,rinv22);
1937 /* EWALD ELECTROSTATICS */
1939 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1940 ewrt = _mm_mul_pd(r22,ewtabscale);
1941 ewitab = _mm_cvttpd_epi32(ewrt);
1942 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1943 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1945 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1946 felec = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
1948 cutoff_mask = _mm_cmplt_pd(rsq22,rcutoff2);
1952 fscal = _mm_and_pd(fscal,cutoff_mask);
1954 /* Calculate temporary vectorial force */
1955 tx = _mm_mul_pd(fscal,dx22);
1956 ty = _mm_mul_pd(fscal,dy22);
1957 tz = _mm_mul_pd(fscal,dz22);
1959 /* Update vectorial force */
1960 fix2 = _mm_add_pd(fix2,tx);
1961 fiy2 = _mm_add_pd(fiy2,ty);
1962 fiz2 = _mm_add_pd(fiz2,tz);
1964 fjx2 = _mm_add_pd(fjx2,tx);
1965 fjy2 = _mm_add_pd(fjy2,ty);
1966 fjz2 = _mm_add_pd(fjz2,tz);
1970 gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1972 /* Inner loop uses 351 flops */
1975 if(jidx<j_index_end)
1979 j_coord_offsetA = DIM*jnrA;
1981 /* load j atom coordinates */
1982 gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
1983 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1985 /* Calculate displacement vector */
1986 dx00 = _mm_sub_pd(ix0,jx0);
1987 dy00 = _mm_sub_pd(iy0,jy0);
1988 dz00 = _mm_sub_pd(iz0,jz0);
1989 dx01 = _mm_sub_pd(ix0,jx1);
1990 dy01 = _mm_sub_pd(iy0,jy1);
1991 dz01 = _mm_sub_pd(iz0,jz1);
1992 dx02 = _mm_sub_pd(ix0,jx2);
1993 dy02 = _mm_sub_pd(iy0,jy2);
1994 dz02 = _mm_sub_pd(iz0,jz2);
1995 dx10 = _mm_sub_pd(ix1,jx0);
1996 dy10 = _mm_sub_pd(iy1,jy0);
1997 dz10 = _mm_sub_pd(iz1,jz0);
1998 dx11 = _mm_sub_pd(ix1,jx1);
1999 dy11 = _mm_sub_pd(iy1,jy1);
2000 dz11 = _mm_sub_pd(iz1,jz1);
2001 dx12 = _mm_sub_pd(ix1,jx2);
2002 dy12 = _mm_sub_pd(iy1,jy2);
2003 dz12 = _mm_sub_pd(iz1,jz2);
2004 dx20 = _mm_sub_pd(ix2,jx0);
2005 dy20 = _mm_sub_pd(iy2,jy0);
2006 dz20 = _mm_sub_pd(iz2,jz0);
2007 dx21 = _mm_sub_pd(ix2,jx1);
2008 dy21 = _mm_sub_pd(iy2,jy1);
2009 dz21 = _mm_sub_pd(iz2,jz1);
2010 dx22 = _mm_sub_pd(ix2,jx2);
2011 dy22 = _mm_sub_pd(iy2,jy2);
2012 dz22 = _mm_sub_pd(iz2,jz2);
2014 /* Calculate squared distance and things based on it */
2015 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
2016 rsq01 = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
2017 rsq02 = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
2018 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
2019 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
2020 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
2021 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
2022 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
2023 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
2025 rinv00 = gmx_mm_invsqrt_pd(rsq00);
2026 rinv01 = gmx_mm_invsqrt_pd(rsq01);
2027 rinv02 = gmx_mm_invsqrt_pd(rsq02);
2028 rinv10 = gmx_mm_invsqrt_pd(rsq10);
2029 rinv11 = gmx_mm_invsqrt_pd(rsq11);
2030 rinv12 = gmx_mm_invsqrt_pd(rsq12);
2031 rinv20 = gmx_mm_invsqrt_pd(rsq20);
2032 rinv21 = gmx_mm_invsqrt_pd(rsq21);
2033 rinv22 = gmx_mm_invsqrt_pd(rsq22);
2035 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
2036 rinvsq01 = _mm_mul_pd(rinv01,rinv01);
2037 rinvsq02 = _mm_mul_pd(rinv02,rinv02);
2038 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
2039 rinvsq11 = _mm_mul_pd(rinv11,rinv11);
2040 rinvsq12 = _mm_mul_pd(rinv12,rinv12);
2041 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
2042 rinvsq21 = _mm_mul_pd(rinv21,rinv21);
2043 rinvsq22 = _mm_mul_pd(rinv22,rinv22);
2045 fjx0 = _mm_setzero_pd();
2046 fjy0 = _mm_setzero_pd();
2047 fjz0 = _mm_setzero_pd();
2048 fjx1 = _mm_setzero_pd();
2049 fjy1 = _mm_setzero_pd();
2050 fjz1 = _mm_setzero_pd();
2051 fjx2 = _mm_setzero_pd();
2052 fjy2 = _mm_setzero_pd();
2053 fjz2 = _mm_setzero_pd();
2055 /**************************
2056 * CALCULATE INTERACTIONS *
2057 **************************/
2059 if (gmx_mm_any_lt(rsq00,rcutoff2))
2062 r00 = _mm_mul_pd(rsq00,rinv00);
2064 /* EWALD ELECTROSTATICS */
2066 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2067 ewrt = _mm_mul_pd(r00,ewtabscale);
2068 ewitab = _mm_cvttpd_epi32(ewrt);
2069 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2070 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2071 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2072 felec = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
2074 cutoff_mask = _mm_cmplt_pd(rsq00,rcutoff2);
2078 fscal = _mm_and_pd(fscal,cutoff_mask);
2080 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2082 /* Calculate temporary vectorial force */
2083 tx = _mm_mul_pd(fscal,dx00);
2084 ty = _mm_mul_pd(fscal,dy00);
2085 tz = _mm_mul_pd(fscal,dz00);
2087 /* Update vectorial force */
2088 fix0 = _mm_add_pd(fix0,tx);
2089 fiy0 = _mm_add_pd(fiy0,ty);
2090 fiz0 = _mm_add_pd(fiz0,tz);
2092 fjx0 = _mm_add_pd(fjx0,tx);
2093 fjy0 = _mm_add_pd(fjy0,ty);
2094 fjz0 = _mm_add_pd(fjz0,tz);
2098 /**************************
2099 * CALCULATE INTERACTIONS *
2100 **************************/
2102 if (gmx_mm_any_lt(rsq01,rcutoff2))
2105 r01 = _mm_mul_pd(rsq01,rinv01);
2107 /* EWALD ELECTROSTATICS */
2109 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2110 ewrt = _mm_mul_pd(r01,ewtabscale);
2111 ewitab = _mm_cvttpd_epi32(ewrt);
2112 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2113 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2114 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2115 felec = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
2117 cutoff_mask = _mm_cmplt_pd(rsq01,rcutoff2);
2121 fscal = _mm_and_pd(fscal,cutoff_mask);
2123 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2125 /* Calculate temporary vectorial force */
2126 tx = _mm_mul_pd(fscal,dx01);
2127 ty = _mm_mul_pd(fscal,dy01);
2128 tz = _mm_mul_pd(fscal,dz01);
2130 /* Update vectorial force */
2131 fix0 = _mm_add_pd(fix0,tx);
2132 fiy0 = _mm_add_pd(fiy0,ty);
2133 fiz0 = _mm_add_pd(fiz0,tz);
2135 fjx1 = _mm_add_pd(fjx1,tx);
2136 fjy1 = _mm_add_pd(fjy1,ty);
2137 fjz1 = _mm_add_pd(fjz1,tz);
2141 /**************************
2142 * CALCULATE INTERACTIONS *
2143 **************************/
2145 if (gmx_mm_any_lt(rsq02,rcutoff2))
2148 r02 = _mm_mul_pd(rsq02,rinv02);
2150 /* EWALD ELECTROSTATICS */
2152 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2153 ewrt = _mm_mul_pd(r02,ewtabscale);
2154 ewitab = _mm_cvttpd_epi32(ewrt);
2155 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2156 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2157 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2158 felec = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
2160 cutoff_mask = _mm_cmplt_pd(rsq02,rcutoff2);
2164 fscal = _mm_and_pd(fscal,cutoff_mask);
2166 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2168 /* Calculate temporary vectorial force */
2169 tx = _mm_mul_pd(fscal,dx02);
2170 ty = _mm_mul_pd(fscal,dy02);
2171 tz = _mm_mul_pd(fscal,dz02);
2173 /* Update vectorial force */
2174 fix0 = _mm_add_pd(fix0,tx);
2175 fiy0 = _mm_add_pd(fiy0,ty);
2176 fiz0 = _mm_add_pd(fiz0,tz);
2178 fjx2 = _mm_add_pd(fjx2,tx);
2179 fjy2 = _mm_add_pd(fjy2,ty);
2180 fjz2 = _mm_add_pd(fjz2,tz);
2184 /**************************
2185 * CALCULATE INTERACTIONS *
2186 **************************/
2188 if (gmx_mm_any_lt(rsq10,rcutoff2))
2191 r10 = _mm_mul_pd(rsq10,rinv10);
2193 /* EWALD ELECTROSTATICS */
2195 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2196 ewrt = _mm_mul_pd(r10,ewtabscale);
2197 ewitab = _mm_cvttpd_epi32(ewrt);
2198 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2199 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2200 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2201 felec = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
2203 cutoff_mask = _mm_cmplt_pd(rsq10,rcutoff2);
2207 fscal = _mm_and_pd(fscal,cutoff_mask);
2209 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2211 /* Calculate temporary vectorial force */
2212 tx = _mm_mul_pd(fscal,dx10);
2213 ty = _mm_mul_pd(fscal,dy10);
2214 tz = _mm_mul_pd(fscal,dz10);
2216 /* Update vectorial force */
2217 fix1 = _mm_add_pd(fix1,tx);
2218 fiy1 = _mm_add_pd(fiy1,ty);
2219 fiz1 = _mm_add_pd(fiz1,tz);
2221 fjx0 = _mm_add_pd(fjx0,tx);
2222 fjy0 = _mm_add_pd(fjy0,ty);
2223 fjz0 = _mm_add_pd(fjz0,tz);
2227 /**************************
2228 * CALCULATE INTERACTIONS *
2229 **************************/
2231 if (gmx_mm_any_lt(rsq11,rcutoff2))
2234 r11 = _mm_mul_pd(rsq11,rinv11);
2236 /* EWALD ELECTROSTATICS */
2238 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2239 ewrt = _mm_mul_pd(r11,ewtabscale);
2240 ewitab = _mm_cvttpd_epi32(ewrt);
2241 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2242 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2243 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2244 felec = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
2246 cutoff_mask = _mm_cmplt_pd(rsq11,rcutoff2);
2250 fscal = _mm_and_pd(fscal,cutoff_mask);
2252 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2254 /* Calculate temporary vectorial force */
2255 tx = _mm_mul_pd(fscal,dx11);
2256 ty = _mm_mul_pd(fscal,dy11);
2257 tz = _mm_mul_pd(fscal,dz11);
2259 /* Update vectorial force */
2260 fix1 = _mm_add_pd(fix1,tx);
2261 fiy1 = _mm_add_pd(fiy1,ty);
2262 fiz1 = _mm_add_pd(fiz1,tz);
2264 fjx1 = _mm_add_pd(fjx1,tx);
2265 fjy1 = _mm_add_pd(fjy1,ty);
2266 fjz1 = _mm_add_pd(fjz1,tz);
2270 /**************************
2271 * CALCULATE INTERACTIONS *
2272 **************************/
2274 if (gmx_mm_any_lt(rsq12,rcutoff2))
2277 r12 = _mm_mul_pd(rsq12,rinv12);
2279 /* EWALD ELECTROSTATICS */
2281 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2282 ewrt = _mm_mul_pd(r12,ewtabscale);
2283 ewitab = _mm_cvttpd_epi32(ewrt);
2284 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2285 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2286 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2287 felec = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
2289 cutoff_mask = _mm_cmplt_pd(rsq12,rcutoff2);
2293 fscal = _mm_and_pd(fscal,cutoff_mask);
2295 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2297 /* Calculate temporary vectorial force */
2298 tx = _mm_mul_pd(fscal,dx12);
2299 ty = _mm_mul_pd(fscal,dy12);
2300 tz = _mm_mul_pd(fscal,dz12);
2302 /* Update vectorial force */
2303 fix1 = _mm_add_pd(fix1,tx);
2304 fiy1 = _mm_add_pd(fiy1,ty);
2305 fiz1 = _mm_add_pd(fiz1,tz);
2307 fjx2 = _mm_add_pd(fjx2,tx);
2308 fjy2 = _mm_add_pd(fjy2,ty);
2309 fjz2 = _mm_add_pd(fjz2,tz);
2313 /**************************
2314 * CALCULATE INTERACTIONS *
2315 **************************/
2317 if (gmx_mm_any_lt(rsq20,rcutoff2))
2320 r20 = _mm_mul_pd(rsq20,rinv20);
2322 /* EWALD ELECTROSTATICS */
2324 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2325 ewrt = _mm_mul_pd(r20,ewtabscale);
2326 ewitab = _mm_cvttpd_epi32(ewrt);
2327 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2328 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2329 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2330 felec = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
2332 cutoff_mask = _mm_cmplt_pd(rsq20,rcutoff2);
2336 fscal = _mm_and_pd(fscal,cutoff_mask);
2338 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2340 /* Calculate temporary vectorial force */
2341 tx = _mm_mul_pd(fscal,dx20);
2342 ty = _mm_mul_pd(fscal,dy20);
2343 tz = _mm_mul_pd(fscal,dz20);
2345 /* Update vectorial force */
2346 fix2 = _mm_add_pd(fix2,tx);
2347 fiy2 = _mm_add_pd(fiy2,ty);
2348 fiz2 = _mm_add_pd(fiz2,tz);
2350 fjx0 = _mm_add_pd(fjx0,tx);
2351 fjy0 = _mm_add_pd(fjy0,ty);
2352 fjz0 = _mm_add_pd(fjz0,tz);
2356 /**************************
2357 * CALCULATE INTERACTIONS *
2358 **************************/
2360 if (gmx_mm_any_lt(rsq21,rcutoff2))
2363 r21 = _mm_mul_pd(rsq21,rinv21);
2365 /* EWALD ELECTROSTATICS */
2367 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2368 ewrt = _mm_mul_pd(r21,ewtabscale);
2369 ewitab = _mm_cvttpd_epi32(ewrt);
2370 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2371 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2372 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2373 felec = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
2375 cutoff_mask = _mm_cmplt_pd(rsq21,rcutoff2);
2379 fscal = _mm_and_pd(fscal,cutoff_mask);
2381 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2383 /* Calculate temporary vectorial force */
2384 tx = _mm_mul_pd(fscal,dx21);
2385 ty = _mm_mul_pd(fscal,dy21);
2386 tz = _mm_mul_pd(fscal,dz21);
2388 /* Update vectorial force */
2389 fix2 = _mm_add_pd(fix2,tx);
2390 fiy2 = _mm_add_pd(fiy2,ty);
2391 fiz2 = _mm_add_pd(fiz2,tz);
2393 fjx1 = _mm_add_pd(fjx1,tx);
2394 fjy1 = _mm_add_pd(fjy1,ty);
2395 fjz1 = _mm_add_pd(fjz1,tz);
2399 /**************************
2400 * CALCULATE INTERACTIONS *
2401 **************************/
2403 if (gmx_mm_any_lt(rsq22,rcutoff2))
2406 r22 = _mm_mul_pd(rsq22,rinv22);
2408 /* EWALD ELECTROSTATICS */
2410 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2411 ewrt = _mm_mul_pd(r22,ewtabscale);
2412 ewitab = _mm_cvttpd_epi32(ewrt);
2413 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2414 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2415 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2416 felec = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
2418 cutoff_mask = _mm_cmplt_pd(rsq22,rcutoff2);
2422 fscal = _mm_and_pd(fscal,cutoff_mask);
2424 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2426 /* Calculate temporary vectorial force */
2427 tx = _mm_mul_pd(fscal,dx22);
2428 ty = _mm_mul_pd(fscal,dy22);
2429 tz = _mm_mul_pd(fscal,dz22);
2431 /* Update vectorial force */
2432 fix2 = _mm_add_pd(fix2,tx);
2433 fiy2 = _mm_add_pd(fiy2,ty);
2434 fiz2 = _mm_add_pd(fiz2,tz);
2436 fjx2 = _mm_add_pd(fjx2,tx);
2437 fjy2 = _mm_add_pd(fjy2,ty);
2438 fjz2 = _mm_add_pd(fjz2,tz);
2442 gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
2444 /* Inner loop uses 351 flops */
2447 /* End of innermost loop */
2449 gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
2450 f+i_coord_offset,fshift+i_shift_offset);
2452 /* Increment number of inner iterations */
2453 inneriter += j_index_end - j_index_start;
2455 /* Outer loop uses 18 flops */
2458 /* Increment number of outer iterations */
2461 /* Update outer/inner flops */
2463 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3W3_F,outeriter*18 + inneriter*351);