2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2012,2013, 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.
44 #include "../nb_kernel.h"
45 #include "types/simple.h"
49 #include "gromacs/simd/math_x86_sse2_double.h"
50 #include "kernelutil_x86_sse2_double.h"
53 * Gromacs nonbonded kernel: nb_kernel_ElecEwSh_VdwNone_GeomW4W4_VF_sse2_double
54 * Electrostatics interaction: Ewald
55 * VdW interaction: None
56 * Geometry: Water4-Water4
57 * Calculate force/pot: PotentialAndForce
60 nb_kernel_ElecEwSh_VdwNone_GeomW4W4_VF_sse2_double
61 (t_nblist * gmx_restrict nlist,
62 rvec * gmx_restrict xx,
63 rvec * gmx_restrict ff,
64 t_forcerec * gmx_restrict fr,
65 t_mdatoms * gmx_restrict mdatoms,
66 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
67 t_nrnb * gmx_restrict nrnb)
69 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
70 * just 0 for non-waters.
71 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
72 * jnr indices corresponding to data put in the four positions in the SIMD register.
74 int i_shift_offset,i_coord_offset,outeriter,inneriter;
75 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
77 int j_coord_offsetA,j_coord_offsetB;
78 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
80 real *shiftvec,*fshift,*x,*f;
81 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
83 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
85 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
87 __m128d ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
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 int vdwjidx3A,vdwjidx3B;
93 __m128d jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
94 __m128d dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
95 __m128d dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
96 __m128d dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
97 __m128d dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
98 __m128d dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
99 __m128d dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
100 __m128d dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
101 __m128d dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
102 __m128d dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
103 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
106 __m128d ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
108 __m128d dummy_mask,cutoff_mask;
109 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
110 __m128d one = _mm_set1_pd(1.0);
111 __m128d two = _mm_set1_pd(2.0);
117 jindex = nlist->jindex;
119 shiftidx = nlist->shift;
121 shiftvec = fr->shift_vec[0];
122 fshift = fr->fshift[0];
123 facel = _mm_set1_pd(fr->epsfac);
124 charge = mdatoms->chargeA;
126 sh_ewald = _mm_set1_pd(fr->ic->sh_ewald);
127 ewtab = fr->ic->tabq_coul_FDV0;
128 ewtabscale = _mm_set1_pd(fr->ic->tabq_scale);
129 ewtabhalfspace = _mm_set1_pd(0.5/fr->ic->tabq_scale);
131 /* Setup water-specific parameters */
132 inr = nlist->iinr[0];
133 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
134 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
135 iq3 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
137 jq1 = _mm_set1_pd(charge[inr+1]);
138 jq2 = _mm_set1_pd(charge[inr+2]);
139 jq3 = _mm_set1_pd(charge[inr+3]);
140 qq11 = _mm_mul_pd(iq1,jq1);
141 qq12 = _mm_mul_pd(iq1,jq2);
142 qq13 = _mm_mul_pd(iq1,jq3);
143 qq21 = _mm_mul_pd(iq2,jq1);
144 qq22 = _mm_mul_pd(iq2,jq2);
145 qq23 = _mm_mul_pd(iq2,jq3);
146 qq31 = _mm_mul_pd(iq3,jq1);
147 qq32 = _mm_mul_pd(iq3,jq2);
148 qq33 = _mm_mul_pd(iq3,jq3);
150 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
151 rcutoff_scalar = fr->rcoulomb;
152 rcutoff = _mm_set1_pd(rcutoff_scalar);
153 rcutoff2 = _mm_mul_pd(rcutoff,rcutoff);
155 /* Avoid stupid compiler warnings */
163 /* Start outer loop over neighborlists */
164 for(iidx=0; iidx<nri; iidx++)
166 /* Load shift vector for this list */
167 i_shift_offset = DIM*shiftidx[iidx];
169 /* Load limits for loop over neighbors */
170 j_index_start = jindex[iidx];
171 j_index_end = jindex[iidx+1];
173 /* Get outer coordinate index */
175 i_coord_offset = DIM*inr;
177 /* Load i particle coords and add shift vector */
178 gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
179 &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
181 fix1 = _mm_setzero_pd();
182 fiy1 = _mm_setzero_pd();
183 fiz1 = _mm_setzero_pd();
184 fix2 = _mm_setzero_pd();
185 fiy2 = _mm_setzero_pd();
186 fiz2 = _mm_setzero_pd();
187 fix3 = _mm_setzero_pd();
188 fiy3 = _mm_setzero_pd();
189 fiz3 = _mm_setzero_pd();
191 /* Reset potential sums */
192 velecsum = _mm_setzero_pd();
194 /* Start inner kernel loop */
195 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
198 /* Get j neighbor index, and coordinate index */
201 j_coord_offsetA = DIM*jnrA;
202 j_coord_offsetB = DIM*jnrB;
204 /* load j atom coordinates */
205 gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
206 &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
208 /* Calculate displacement vector */
209 dx11 = _mm_sub_pd(ix1,jx1);
210 dy11 = _mm_sub_pd(iy1,jy1);
211 dz11 = _mm_sub_pd(iz1,jz1);
212 dx12 = _mm_sub_pd(ix1,jx2);
213 dy12 = _mm_sub_pd(iy1,jy2);
214 dz12 = _mm_sub_pd(iz1,jz2);
215 dx13 = _mm_sub_pd(ix1,jx3);
216 dy13 = _mm_sub_pd(iy1,jy3);
217 dz13 = _mm_sub_pd(iz1,jz3);
218 dx21 = _mm_sub_pd(ix2,jx1);
219 dy21 = _mm_sub_pd(iy2,jy1);
220 dz21 = _mm_sub_pd(iz2,jz1);
221 dx22 = _mm_sub_pd(ix2,jx2);
222 dy22 = _mm_sub_pd(iy2,jy2);
223 dz22 = _mm_sub_pd(iz2,jz2);
224 dx23 = _mm_sub_pd(ix2,jx3);
225 dy23 = _mm_sub_pd(iy2,jy3);
226 dz23 = _mm_sub_pd(iz2,jz3);
227 dx31 = _mm_sub_pd(ix3,jx1);
228 dy31 = _mm_sub_pd(iy3,jy1);
229 dz31 = _mm_sub_pd(iz3,jz1);
230 dx32 = _mm_sub_pd(ix3,jx2);
231 dy32 = _mm_sub_pd(iy3,jy2);
232 dz32 = _mm_sub_pd(iz3,jz2);
233 dx33 = _mm_sub_pd(ix3,jx3);
234 dy33 = _mm_sub_pd(iy3,jy3);
235 dz33 = _mm_sub_pd(iz3,jz3);
237 /* Calculate squared distance and things based on it */
238 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
239 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
240 rsq13 = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
241 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
242 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
243 rsq23 = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
244 rsq31 = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
245 rsq32 = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
246 rsq33 = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
248 rinv11 = gmx_mm_invsqrt_pd(rsq11);
249 rinv12 = gmx_mm_invsqrt_pd(rsq12);
250 rinv13 = gmx_mm_invsqrt_pd(rsq13);
251 rinv21 = gmx_mm_invsqrt_pd(rsq21);
252 rinv22 = gmx_mm_invsqrt_pd(rsq22);
253 rinv23 = gmx_mm_invsqrt_pd(rsq23);
254 rinv31 = gmx_mm_invsqrt_pd(rsq31);
255 rinv32 = gmx_mm_invsqrt_pd(rsq32);
256 rinv33 = gmx_mm_invsqrt_pd(rsq33);
258 rinvsq11 = _mm_mul_pd(rinv11,rinv11);
259 rinvsq12 = _mm_mul_pd(rinv12,rinv12);
260 rinvsq13 = _mm_mul_pd(rinv13,rinv13);
261 rinvsq21 = _mm_mul_pd(rinv21,rinv21);
262 rinvsq22 = _mm_mul_pd(rinv22,rinv22);
263 rinvsq23 = _mm_mul_pd(rinv23,rinv23);
264 rinvsq31 = _mm_mul_pd(rinv31,rinv31);
265 rinvsq32 = _mm_mul_pd(rinv32,rinv32);
266 rinvsq33 = _mm_mul_pd(rinv33,rinv33);
268 fjx1 = _mm_setzero_pd();
269 fjy1 = _mm_setzero_pd();
270 fjz1 = _mm_setzero_pd();
271 fjx2 = _mm_setzero_pd();
272 fjy2 = _mm_setzero_pd();
273 fjz2 = _mm_setzero_pd();
274 fjx3 = _mm_setzero_pd();
275 fjy3 = _mm_setzero_pd();
276 fjz3 = _mm_setzero_pd();
278 /**************************
279 * CALCULATE INTERACTIONS *
280 **************************/
282 if (gmx_mm_any_lt(rsq11,rcutoff2))
285 r11 = _mm_mul_pd(rsq11,rinv11);
287 /* EWALD ELECTROSTATICS */
289 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
290 ewrt = _mm_mul_pd(r11,ewtabscale);
291 ewitab = _mm_cvttpd_epi32(ewrt);
292 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
293 ewitab = _mm_slli_epi32(ewitab,2);
294 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
295 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
296 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
297 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
298 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
299 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
300 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
301 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
302 velec = _mm_mul_pd(qq11,_mm_sub_pd(_mm_sub_pd(rinv11,sh_ewald),velec));
303 felec = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
305 cutoff_mask = _mm_cmplt_pd(rsq11,rcutoff2);
307 /* Update potential sum for this i atom from the interaction with this j atom. */
308 velec = _mm_and_pd(velec,cutoff_mask);
309 velecsum = _mm_add_pd(velecsum,velec);
313 fscal = _mm_and_pd(fscal,cutoff_mask);
315 /* Calculate temporary vectorial force */
316 tx = _mm_mul_pd(fscal,dx11);
317 ty = _mm_mul_pd(fscal,dy11);
318 tz = _mm_mul_pd(fscal,dz11);
320 /* Update vectorial force */
321 fix1 = _mm_add_pd(fix1,tx);
322 fiy1 = _mm_add_pd(fiy1,ty);
323 fiz1 = _mm_add_pd(fiz1,tz);
325 fjx1 = _mm_add_pd(fjx1,tx);
326 fjy1 = _mm_add_pd(fjy1,ty);
327 fjz1 = _mm_add_pd(fjz1,tz);
331 /**************************
332 * CALCULATE INTERACTIONS *
333 **************************/
335 if (gmx_mm_any_lt(rsq12,rcutoff2))
338 r12 = _mm_mul_pd(rsq12,rinv12);
340 /* EWALD ELECTROSTATICS */
342 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
343 ewrt = _mm_mul_pd(r12,ewtabscale);
344 ewitab = _mm_cvttpd_epi32(ewrt);
345 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
346 ewitab = _mm_slli_epi32(ewitab,2);
347 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
348 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
349 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
350 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
351 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
352 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
353 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
354 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
355 velec = _mm_mul_pd(qq12,_mm_sub_pd(_mm_sub_pd(rinv12,sh_ewald),velec));
356 felec = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
358 cutoff_mask = _mm_cmplt_pd(rsq12,rcutoff2);
360 /* Update potential sum for this i atom from the interaction with this j atom. */
361 velec = _mm_and_pd(velec,cutoff_mask);
362 velecsum = _mm_add_pd(velecsum,velec);
366 fscal = _mm_and_pd(fscal,cutoff_mask);
368 /* Calculate temporary vectorial force */
369 tx = _mm_mul_pd(fscal,dx12);
370 ty = _mm_mul_pd(fscal,dy12);
371 tz = _mm_mul_pd(fscal,dz12);
373 /* Update vectorial force */
374 fix1 = _mm_add_pd(fix1,tx);
375 fiy1 = _mm_add_pd(fiy1,ty);
376 fiz1 = _mm_add_pd(fiz1,tz);
378 fjx2 = _mm_add_pd(fjx2,tx);
379 fjy2 = _mm_add_pd(fjy2,ty);
380 fjz2 = _mm_add_pd(fjz2,tz);
384 /**************************
385 * CALCULATE INTERACTIONS *
386 **************************/
388 if (gmx_mm_any_lt(rsq13,rcutoff2))
391 r13 = _mm_mul_pd(rsq13,rinv13);
393 /* EWALD ELECTROSTATICS */
395 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
396 ewrt = _mm_mul_pd(r13,ewtabscale);
397 ewitab = _mm_cvttpd_epi32(ewrt);
398 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
399 ewitab = _mm_slli_epi32(ewitab,2);
400 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
401 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
402 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
403 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
404 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
405 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
406 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
407 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
408 velec = _mm_mul_pd(qq13,_mm_sub_pd(_mm_sub_pd(rinv13,sh_ewald),velec));
409 felec = _mm_mul_pd(_mm_mul_pd(qq13,rinv13),_mm_sub_pd(rinvsq13,felec));
411 cutoff_mask = _mm_cmplt_pd(rsq13,rcutoff2);
413 /* Update potential sum for this i atom from the interaction with this j atom. */
414 velec = _mm_and_pd(velec,cutoff_mask);
415 velecsum = _mm_add_pd(velecsum,velec);
419 fscal = _mm_and_pd(fscal,cutoff_mask);
421 /* Calculate temporary vectorial force */
422 tx = _mm_mul_pd(fscal,dx13);
423 ty = _mm_mul_pd(fscal,dy13);
424 tz = _mm_mul_pd(fscal,dz13);
426 /* Update vectorial force */
427 fix1 = _mm_add_pd(fix1,tx);
428 fiy1 = _mm_add_pd(fiy1,ty);
429 fiz1 = _mm_add_pd(fiz1,tz);
431 fjx3 = _mm_add_pd(fjx3,tx);
432 fjy3 = _mm_add_pd(fjy3,ty);
433 fjz3 = _mm_add_pd(fjz3,tz);
437 /**************************
438 * CALCULATE INTERACTIONS *
439 **************************/
441 if (gmx_mm_any_lt(rsq21,rcutoff2))
444 r21 = _mm_mul_pd(rsq21,rinv21);
446 /* EWALD ELECTROSTATICS */
448 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
449 ewrt = _mm_mul_pd(r21,ewtabscale);
450 ewitab = _mm_cvttpd_epi32(ewrt);
451 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
452 ewitab = _mm_slli_epi32(ewitab,2);
453 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
454 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
455 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
456 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
457 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
458 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
459 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
460 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
461 velec = _mm_mul_pd(qq21,_mm_sub_pd(_mm_sub_pd(rinv21,sh_ewald),velec));
462 felec = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
464 cutoff_mask = _mm_cmplt_pd(rsq21,rcutoff2);
466 /* Update potential sum for this i atom from the interaction with this j atom. */
467 velec = _mm_and_pd(velec,cutoff_mask);
468 velecsum = _mm_add_pd(velecsum,velec);
472 fscal = _mm_and_pd(fscal,cutoff_mask);
474 /* Calculate temporary vectorial force */
475 tx = _mm_mul_pd(fscal,dx21);
476 ty = _mm_mul_pd(fscal,dy21);
477 tz = _mm_mul_pd(fscal,dz21);
479 /* Update vectorial force */
480 fix2 = _mm_add_pd(fix2,tx);
481 fiy2 = _mm_add_pd(fiy2,ty);
482 fiz2 = _mm_add_pd(fiz2,tz);
484 fjx1 = _mm_add_pd(fjx1,tx);
485 fjy1 = _mm_add_pd(fjy1,ty);
486 fjz1 = _mm_add_pd(fjz1,tz);
490 /**************************
491 * CALCULATE INTERACTIONS *
492 **************************/
494 if (gmx_mm_any_lt(rsq22,rcutoff2))
497 r22 = _mm_mul_pd(rsq22,rinv22);
499 /* EWALD ELECTROSTATICS */
501 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
502 ewrt = _mm_mul_pd(r22,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_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
508 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
509 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
510 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
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(qq22,_mm_sub_pd(_mm_sub_pd(rinv22,sh_ewald),velec));
515 felec = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
517 cutoff_mask = _mm_cmplt_pd(rsq22,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 velecsum = _mm_add_pd(velecsum,velec);
525 fscal = _mm_and_pd(fscal,cutoff_mask);
527 /* Calculate temporary vectorial force */
528 tx = _mm_mul_pd(fscal,dx22);
529 ty = _mm_mul_pd(fscal,dy22);
530 tz = _mm_mul_pd(fscal,dz22);
532 /* Update vectorial force */
533 fix2 = _mm_add_pd(fix2,tx);
534 fiy2 = _mm_add_pd(fiy2,ty);
535 fiz2 = _mm_add_pd(fiz2,tz);
537 fjx2 = _mm_add_pd(fjx2,tx);
538 fjy2 = _mm_add_pd(fjy2,ty);
539 fjz2 = _mm_add_pd(fjz2,tz);
543 /**************************
544 * CALCULATE INTERACTIONS *
545 **************************/
547 if (gmx_mm_any_lt(rsq23,rcutoff2))
550 r23 = _mm_mul_pd(rsq23,rinv23);
552 /* EWALD ELECTROSTATICS */
554 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
555 ewrt = _mm_mul_pd(r23,ewtabscale);
556 ewitab = _mm_cvttpd_epi32(ewrt);
557 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
558 ewitab = _mm_slli_epi32(ewitab,2);
559 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
560 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
561 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
562 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
563 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
564 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
565 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
566 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
567 velec = _mm_mul_pd(qq23,_mm_sub_pd(_mm_sub_pd(rinv23,sh_ewald),velec));
568 felec = _mm_mul_pd(_mm_mul_pd(qq23,rinv23),_mm_sub_pd(rinvsq23,felec));
570 cutoff_mask = _mm_cmplt_pd(rsq23,rcutoff2);
572 /* Update potential sum for this i atom from the interaction with this j atom. */
573 velec = _mm_and_pd(velec,cutoff_mask);
574 velecsum = _mm_add_pd(velecsum,velec);
578 fscal = _mm_and_pd(fscal,cutoff_mask);
580 /* Calculate temporary vectorial force */
581 tx = _mm_mul_pd(fscal,dx23);
582 ty = _mm_mul_pd(fscal,dy23);
583 tz = _mm_mul_pd(fscal,dz23);
585 /* Update vectorial force */
586 fix2 = _mm_add_pd(fix2,tx);
587 fiy2 = _mm_add_pd(fiy2,ty);
588 fiz2 = _mm_add_pd(fiz2,tz);
590 fjx3 = _mm_add_pd(fjx3,tx);
591 fjy3 = _mm_add_pd(fjy3,ty);
592 fjz3 = _mm_add_pd(fjz3,tz);
596 /**************************
597 * CALCULATE INTERACTIONS *
598 **************************/
600 if (gmx_mm_any_lt(rsq31,rcutoff2))
603 r31 = _mm_mul_pd(rsq31,rinv31);
605 /* EWALD ELECTROSTATICS */
607 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
608 ewrt = _mm_mul_pd(r31,ewtabscale);
609 ewitab = _mm_cvttpd_epi32(ewrt);
610 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
611 ewitab = _mm_slli_epi32(ewitab,2);
612 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
613 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
614 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
615 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
616 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
617 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
618 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
619 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
620 velec = _mm_mul_pd(qq31,_mm_sub_pd(_mm_sub_pd(rinv31,sh_ewald),velec));
621 felec = _mm_mul_pd(_mm_mul_pd(qq31,rinv31),_mm_sub_pd(rinvsq31,felec));
623 cutoff_mask = _mm_cmplt_pd(rsq31,rcutoff2);
625 /* Update potential sum for this i atom from the interaction with this j atom. */
626 velec = _mm_and_pd(velec,cutoff_mask);
627 velecsum = _mm_add_pd(velecsum,velec);
631 fscal = _mm_and_pd(fscal,cutoff_mask);
633 /* Calculate temporary vectorial force */
634 tx = _mm_mul_pd(fscal,dx31);
635 ty = _mm_mul_pd(fscal,dy31);
636 tz = _mm_mul_pd(fscal,dz31);
638 /* Update vectorial force */
639 fix3 = _mm_add_pd(fix3,tx);
640 fiy3 = _mm_add_pd(fiy3,ty);
641 fiz3 = _mm_add_pd(fiz3,tz);
643 fjx1 = _mm_add_pd(fjx1,tx);
644 fjy1 = _mm_add_pd(fjy1,ty);
645 fjz1 = _mm_add_pd(fjz1,tz);
649 /**************************
650 * CALCULATE INTERACTIONS *
651 **************************/
653 if (gmx_mm_any_lt(rsq32,rcutoff2))
656 r32 = _mm_mul_pd(rsq32,rinv32);
658 /* EWALD ELECTROSTATICS */
660 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
661 ewrt = _mm_mul_pd(r32,ewtabscale);
662 ewitab = _mm_cvttpd_epi32(ewrt);
663 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
664 ewitab = _mm_slli_epi32(ewitab,2);
665 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
666 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
667 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
668 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
669 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
670 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
671 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
672 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
673 velec = _mm_mul_pd(qq32,_mm_sub_pd(_mm_sub_pd(rinv32,sh_ewald),velec));
674 felec = _mm_mul_pd(_mm_mul_pd(qq32,rinv32),_mm_sub_pd(rinvsq32,felec));
676 cutoff_mask = _mm_cmplt_pd(rsq32,rcutoff2);
678 /* Update potential sum for this i atom from the interaction with this j atom. */
679 velec = _mm_and_pd(velec,cutoff_mask);
680 velecsum = _mm_add_pd(velecsum,velec);
684 fscal = _mm_and_pd(fscal,cutoff_mask);
686 /* Calculate temporary vectorial force */
687 tx = _mm_mul_pd(fscal,dx32);
688 ty = _mm_mul_pd(fscal,dy32);
689 tz = _mm_mul_pd(fscal,dz32);
691 /* Update vectorial force */
692 fix3 = _mm_add_pd(fix3,tx);
693 fiy3 = _mm_add_pd(fiy3,ty);
694 fiz3 = _mm_add_pd(fiz3,tz);
696 fjx2 = _mm_add_pd(fjx2,tx);
697 fjy2 = _mm_add_pd(fjy2,ty);
698 fjz2 = _mm_add_pd(fjz2,tz);
702 /**************************
703 * CALCULATE INTERACTIONS *
704 **************************/
706 if (gmx_mm_any_lt(rsq33,rcutoff2))
709 r33 = _mm_mul_pd(rsq33,rinv33);
711 /* EWALD ELECTROSTATICS */
713 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
714 ewrt = _mm_mul_pd(r33,ewtabscale);
715 ewitab = _mm_cvttpd_epi32(ewrt);
716 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
717 ewitab = _mm_slli_epi32(ewitab,2);
718 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
719 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
720 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
721 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
722 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
723 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
724 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
725 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
726 velec = _mm_mul_pd(qq33,_mm_sub_pd(_mm_sub_pd(rinv33,sh_ewald),velec));
727 felec = _mm_mul_pd(_mm_mul_pd(qq33,rinv33),_mm_sub_pd(rinvsq33,felec));
729 cutoff_mask = _mm_cmplt_pd(rsq33,rcutoff2);
731 /* Update potential sum for this i atom from the interaction with this j atom. */
732 velec = _mm_and_pd(velec,cutoff_mask);
733 velecsum = _mm_add_pd(velecsum,velec);
737 fscal = _mm_and_pd(fscal,cutoff_mask);
739 /* Calculate temporary vectorial force */
740 tx = _mm_mul_pd(fscal,dx33);
741 ty = _mm_mul_pd(fscal,dy33);
742 tz = _mm_mul_pd(fscal,dz33);
744 /* Update vectorial force */
745 fix3 = _mm_add_pd(fix3,tx);
746 fiy3 = _mm_add_pd(fiy3,ty);
747 fiz3 = _mm_add_pd(fiz3,tz);
749 fjx3 = _mm_add_pd(fjx3,tx);
750 fjy3 = _mm_add_pd(fjy3,ty);
751 fjz3 = _mm_add_pd(fjz3,tz);
755 gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA+DIM,f+j_coord_offsetB+DIM,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
757 /* Inner loop uses 414 flops */
764 j_coord_offsetA = DIM*jnrA;
766 /* load j atom coordinates */
767 gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA+DIM,
768 &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
770 /* Calculate displacement vector */
771 dx11 = _mm_sub_pd(ix1,jx1);
772 dy11 = _mm_sub_pd(iy1,jy1);
773 dz11 = _mm_sub_pd(iz1,jz1);
774 dx12 = _mm_sub_pd(ix1,jx2);
775 dy12 = _mm_sub_pd(iy1,jy2);
776 dz12 = _mm_sub_pd(iz1,jz2);
777 dx13 = _mm_sub_pd(ix1,jx3);
778 dy13 = _mm_sub_pd(iy1,jy3);
779 dz13 = _mm_sub_pd(iz1,jz3);
780 dx21 = _mm_sub_pd(ix2,jx1);
781 dy21 = _mm_sub_pd(iy2,jy1);
782 dz21 = _mm_sub_pd(iz2,jz1);
783 dx22 = _mm_sub_pd(ix2,jx2);
784 dy22 = _mm_sub_pd(iy2,jy2);
785 dz22 = _mm_sub_pd(iz2,jz2);
786 dx23 = _mm_sub_pd(ix2,jx3);
787 dy23 = _mm_sub_pd(iy2,jy3);
788 dz23 = _mm_sub_pd(iz2,jz3);
789 dx31 = _mm_sub_pd(ix3,jx1);
790 dy31 = _mm_sub_pd(iy3,jy1);
791 dz31 = _mm_sub_pd(iz3,jz1);
792 dx32 = _mm_sub_pd(ix3,jx2);
793 dy32 = _mm_sub_pd(iy3,jy2);
794 dz32 = _mm_sub_pd(iz3,jz2);
795 dx33 = _mm_sub_pd(ix3,jx3);
796 dy33 = _mm_sub_pd(iy3,jy3);
797 dz33 = _mm_sub_pd(iz3,jz3);
799 /* Calculate squared distance and things based on it */
800 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
801 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
802 rsq13 = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
803 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
804 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
805 rsq23 = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
806 rsq31 = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
807 rsq32 = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
808 rsq33 = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
810 rinv11 = gmx_mm_invsqrt_pd(rsq11);
811 rinv12 = gmx_mm_invsqrt_pd(rsq12);
812 rinv13 = gmx_mm_invsqrt_pd(rsq13);
813 rinv21 = gmx_mm_invsqrt_pd(rsq21);
814 rinv22 = gmx_mm_invsqrt_pd(rsq22);
815 rinv23 = gmx_mm_invsqrt_pd(rsq23);
816 rinv31 = gmx_mm_invsqrt_pd(rsq31);
817 rinv32 = gmx_mm_invsqrt_pd(rsq32);
818 rinv33 = gmx_mm_invsqrt_pd(rsq33);
820 rinvsq11 = _mm_mul_pd(rinv11,rinv11);
821 rinvsq12 = _mm_mul_pd(rinv12,rinv12);
822 rinvsq13 = _mm_mul_pd(rinv13,rinv13);
823 rinvsq21 = _mm_mul_pd(rinv21,rinv21);
824 rinvsq22 = _mm_mul_pd(rinv22,rinv22);
825 rinvsq23 = _mm_mul_pd(rinv23,rinv23);
826 rinvsq31 = _mm_mul_pd(rinv31,rinv31);
827 rinvsq32 = _mm_mul_pd(rinv32,rinv32);
828 rinvsq33 = _mm_mul_pd(rinv33,rinv33);
830 fjx1 = _mm_setzero_pd();
831 fjy1 = _mm_setzero_pd();
832 fjz1 = _mm_setzero_pd();
833 fjx2 = _mm_setzero_pd();
834 fjy2 = _mm_setzero_pd();
835 fjz2 = _mm_setzero_pd();
836 fjx3 = _mm_setzero_pd();
837 fjy3 = _mm_setzero_pd();
838 fjz3 = _mm_setzero_pd();
840 /**************************
841 * CALCULATE INTERACTIONS *
842 **************************/
844 if (gmx_mm_any_lt(rsq11,rcutoff2))
847 r11 = _mm_mul_pd(rsq11,rinv11);
849 /* EWALD ELECTROSTATICS */
851 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
852 ewrt = _mm_mul_pd(r11,ewtabscale);
853 ewitab = _mm_cvttpd_epi32(ewrt);
854 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
855 ewitab = _mm_slli_epi32(ewitab,2);
856 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
857 ewtabD = _mm_setzero_pd();
858 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
859 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
860 ewtabFn = _mm_setzero_pd();
861 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
862 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
863 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
864 velec = _mm_mul_pd(qq11,_mm_sub_pd(_mm_sub_pd(rinv11,sh_ewald),velec));
865 felec = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
867 cutoff_mask = _mm_cmplt_pd(rsq11,rcutoff2);
869 /* Update potential sum for this i atom from the interaction with this j atom. */
870 velec = _mm_and_pd(velec,cutoff_mask);
871 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
872 velecsum = _mm_add_pd(velecsum,velec);
876 fscal = _mm_and_pd(fscal,cutoff_mask);
878 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
880 /* Calculate temporary vectorial force */
881 tx = _mm_mul_pd(fscal,dx11);
882 ty = _mm_mul_pd(fscal,dy11);
883 tz = _mm_mul_pd(fscal,dz11);
885 /* Update vectorial force */
886 fix1 = _mm_add_pd(fix1,tx);
887 fiy1 = _mm_add_pd(fiy1,ty);
888 fiz1 = _mm_add_pd(fiz1,tz);
890 fjx1 = _mm_add_pd(fjx1,tx);
891 fjy1 = _mm_add_pd(fjy1,ty);
892 fjz1 = _mm_add_pd(fjz1,tz);
896 /**************************
897 * CALCULATE INTERACTIONS *
898 **************************/
900 if (gmx_mm_any_lt(rsq12,rcutoff2))
903 r12 = _mm_mul_pd(rsq12,rinv12);
905 /* EWALD ELECTROSTATICS */
907 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
908 ewrt = _mm_mul_pd(r12,ewtabscale);
909 ewitab = _mm_cvttpd_epi32(ewrt);
910 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
911 ewitab = _mm_slli_epi32(ewitab,2);
912 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
913 ewtabD = _mm_setzero_pd();
914 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
915 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
916 ewtabFn = _mm_setzero_pd();
917 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
918 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
919 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
920 velec = _mm_mul_pd(qq12,_mm_sub_pd(_mm_sub_pd(rinv12,sh_ewald),velec));
921 felec = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
923 cutoff_mask = _mm_cmplt_pd(rsq12,rcutoff2);
925 /* Update potential sum for this i atom from the interaction with this j atom. */
926 velec = _mm_and_pd(velec,cutoff_mask);
927 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
928 velecsum = _mm_add_pd(velecsum,velec);
932 fscal = _mm_and_pd(fscal,cutoff_mask);
934 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
936 /* Calculate temporary vectorial force */
937 tx = _mm_mul_pd(fscal,dx12);
938 ty = _mm_mul_pd(fscal,dy12);
939 tz = _mm_mul_pd(fscal,dz12);
941 /* Update vectorial force */
942 fix1 = _mm_add_pd(fix1,tx);
943 fiy1 = _mm_add_pd(fiy1,ty);
944 fiz1 = _mm_add_pd(fiz1,tz);
946 fjx2 = _mm_add_pd(fjx2,tx);
947 fjy2 = _mm_add_pd(fjy2,ty);
948 fjz2 = _mm_add_pd(fjz2,tz);
952 /**************************
953 * CALCULATE INTERACTIONS *
954 **************************/
956 if (gmx_mm_any_lt(rsq13,rcutoff2))
959 r13 = _mm_mul_pd(rsq13,rinv13);
961 /* EWALD ELECTROSTATICS */
963 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
964 ewrt = _mm_mul_pd(r13,ewtabscale);
965 ewitab = _mm_cvttpd_epi32(ewrt);
966 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
967 ewitab = _mm_slli_epi32(ewitab,2);
968 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
969 ewtabD = _mm_setzero_pd();
970 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
971 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
972 ewtabFn = _mm_setzero_pd();
973 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
974 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
975 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
976 velec = _mm_mul_pd(qq13,_mm_sub_pd(_mm_sub_pd(rinv13,sh_ewald),velec));
977 felec = _mm_mul_pd(_mm_mul_pd(qq13,rinv13),_mm_sub_pd(rinvsq13,felec));
979 cutoff_mask = _mm_cmplt_pd(rsq13,rcutoff2);
981 /* Update potential sum for this i atom from the interaction with this j atom. */
982 velec = _mm_and_pd(velec,cutoff_mask);
983 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
984 velecsum = _mm_add_pd(velecsum,velec);
988 fscal = _mm_and_pd(fscal,cutoff_mask);
990 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
992 /* Calculate temporary vectorial force */
993 tx = _mm_mul_pd(fscal,dx13);
994 ty = _mm_mul_pd(fscal,dy13);
995 tz = _mm_mul_pd(fscal,dz13);
997 /* Update vectorial force */
998 fix1 = _mm_add_pd(fix1,tx);
999 fiy1 = _mm_add_pd(fiy1,ty);
1000 fiz1 = _mm_add_pd(fiz1,tz);
1002 fjx3 = _mm_add_pd(fjx3,tx);
1003 fjy3 = _mm_add_pd(fjy3,ty);
1004 fjz3 = _mm_add_pd(fjz3,tz);
1008 /**************************
1009 * CALCULATE INTERACTIONS *
1010 **************************/
1012 if (gmx_mm_any_lt(rsq21,rcutoff2))
1015 r21 = _mm_mul_pd(rsq21,rinv21);
1017 /* EWALD ELECTROSTATICS */
1019 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1020 ewrt = _mm_mul_pd(r21,ewtabscale);
1021 ewitab = _mm_cvttpd_epi32(ewrt);
1022 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1023 ewitab = _mm_slli_epi32(ewitab,2);
1024 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1025 ewtabD = _mm_setzero_pd();
1026 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1027 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1028 ewtabFn = _mm_setzero_pd();
1029 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1030 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1031 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1032 velec = _mm_mul_pd(qq21,_mm_sub_pd(_mm_sub_pd(rinv21,sh_ewald),velec));
1033 felec = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
1035 cutoff_mask = _mm_cmplt_pd(rsq21,rcutoff2);
1037 /* Update potential sum for this i atom from the interaction with this j atom. */
1038 velec = _mm_and_pd(velec,cutoff_mask);
1039 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1040 velecsum = _mm_add_pd(velecsum,velec);
1044 fscal = _mm_and_pd(fscal,cutoff_mask);
1046 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1048 /* Calculate temporary vectorial force */
1049 tx = _mm_mul_pd(fscal,dx21);
1050 ty = _mm_mul_pd(fscal,dy21);
1051 tz = _mm_mul_pd(fscal,dz21);
1053 /* Update vectorial force */
1054 fix2 = _mm_add_pd(fix2,tx);
1055 fiy2 = _mm_add_pd(fiy2,ty);
1056 fiz2 = _mm_add_pd(fiz2,tz);
1058 fjx1 = _mm_add_pd(fjx1,tx);
1059 fjy1 = _mm_add_pd(fjy1,ty);
1060 fjz1 = _mm_add_pd(fjz1,tz);
1064 /**************************
1065 * CALCULATE INTERACTIONS *
1066 **************************/
1068 if (gmx_mm_any_lt(rsq22,rcutoff2))
1071 r22 = _mm_mul_pd(rsq22,rinv22);
1073 /* EWALD ELECTROSTATICS */
1075 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1076 ewrt = _mm_mul_pd(r22,ewtabscale);
1077 ewitab = _mm_cvttpd_epi32(ewrt);
1078 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1079 ewitab = _mm_slli_epi32(ewitab,2);
1080 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1081 ewtabD = _mm_setzero_pd();
1082 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1083 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1084 ewtabFn = _mm_setzero_pd();
1085 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1086 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1087 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1088 velec = _mm_mul_pd(qq22,_mm_sub_pd(_mm_sub_pd(rinv22,sh_ewald),velec));
1089 felec = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
1091 cutoff_mask = _mm_cmplt_pd(rsq22,rcutoff2);
1093 /* Update potential sum for this i atom from the interaction with this j atom. */
1094 velec = _mm_and_pd(velec,cutoff_mask);
1095 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1096 velecsum = _mm_add_pd(velecsum,velec);
1100 fscal = _mm_and_pd(fscal,cutoff_mask);
1102 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1104 /* Calculate temporary vectorial force */
1105 tx = _mm_mul_pd(fscal,dx22);
1106 ty = _mm_mul_pd(fscal,dy22);
1107 tz = _mm_mul_pd(fscal,dz22);
1109 /* Update vectorial force */
1110 fix2 = _mm_add_pd(fix2,tx);
1111 fiy2 = _mm_add_pd(fiy2,ty);
1112 fiz2 = _mm_add_pd(fiz2,tz);
1114 fjx2 = _mm_add_pd(fjx2,tx);
1115 fjy2 = _mm_add_pd(fjy2,ty);
1116 fjz2 = _mm_add_pd(fjz2,tz);
1120 /**************************
1121 * CALCULATE INTERACTIONS *
1122 **************************/
1124 if (gmx_mm_any_lt(rsq23,rcutoff2))
1127 r23 = _mm_mul_pd(rsq23,rinv23);
1129 /* EWALD ELECTROSTATICS */
1131 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1132 ewrt = _mm_mul_pd(r23,ewtabscale);
1133 ewitab = _mm_cvttpd_epi32(ewrt);
1134 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1135 ewitab = _mm_slli_epi32(ewitab,2);
1136 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1137 ewtabD = _mm_setzero_pd();
1138 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1139 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1140 ewtabFn = _mm_setzero_pd();
1141 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1142 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1143 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1144 velec = _mm_mul_pd(qq23,_mm_sub_pd(_mm_sub_pd(rinv23,sh_ewald),velec));
1145 felec = _mm_mul_pd(_mm_mul_pd(qq23,rinv23),_mm_sub_pd(rinvsq23,felec));
1147 cutoff_mask = _mm_cmplt_pd(rsq23,rcutoff2);
1149 /* Update potential sum for this i atom from the interaction with this j atom. */
1150 velec = _mm_and_pd(velec,cutoff_mask);
1151 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1152 velecsum = _mm_add_pd(velecsum,velec);
1156 fscal = _mm_and_pd(fscal,cutoff_mask);
1158 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1160 /* Calculate temporary vectorial force */
1161 tx = _mm_mul_pd(fscal,dx23);
1162 ty = _mm_mul_pd(fscal,dy23);
1163 tz = _mm_mul_pd(fscal,dz23);
1165 /* Update vectorial force */
1166 fix2 = _mm_add_pd(fix2,tx);
1167 fiy2 = _mm_add_pd(fiy2,ty);
1168 fiz2 = _mm_add_pd(fiz2,tz);
1170 fjx3 = _mm_add_pd(fjx3,tx);
1171 fjy3 = _mm_add_pd(fjy3,ty);
1172 fjz3 = _mm_add_pd(fjz3,tz);
1176 /**************************
1177 * CALCULATE INTERACTIONS *
1178 **************************/
1180 if (gmx_mm_any_lt(rsq31,rcutoff2))
1183 r31 = _mm_mul_pd(rsq31,rinv31);
1185 /* EWALD ELECTROSTATICS */
1187 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1188 ewrt = _mm_mul_pd(r31,ewtabscale);
1189 ewitab = _mm_cvttpd_epi32(ewrt);
1190 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1191 ewitab = _mm_slli_epi32(ewitab,2);
1192 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1193 ewtabD = _mm_setzero_pd();
1194 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1195 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1196 ewtabFn = _mm_setzero_pd();
1197 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1198 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1199 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1200 velec = _mm_mul_pd(qq31,_mm_sub_pd(_mm_sub_pd(rinv31,sh_ewald),velec));
1201 felec = _mm_mul_pd(_mm_mul_pd(qq31,rinv31),_mm_sub_pd(rinvsq31,felec));
1203 cutoff_mask = _mm_cmplt_pd(rsq31,rcutoff2);
1205 /* Update potential sum for this i atom from the interaction with this j atom. */
1206 velec = _mm_and_pd(velec,cutoff_mask);
1207 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1208 velecsum = _mm_add_pd(velecsum,velec);
1212 fscal = _mm_and_pd(fscal,cutoff_mask);
1214 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1216 /* Calculate temporary vectorial force */
1217 tx = _mm_mul_pd(fscal,dx31);
1218 ty = _mm_mul_pd(fscal,dy31);
1219 tz = _mm_mul_pd(fscal,dz31);
1221 /* Update vectorial force */
1222 fix3 = _mm_add_pd(fix3,tx);
1223 fiy3 = _mm_add_pd(fiy3,ty);
1224 fiz3 = _mm_add_pd(fiz3,tz);
1226 fjx1 = _mm_add_pd(fjx1,tx);
1227 fjy1 = _mm_add_pd(fjy1,ty);
1228 fjz1 = _mm_add_pd(fjz1,tz);
1232 /**************************
1233 * CALCULATE INTERACTIONS *
1234 **************************/
1236 if (gmx_mm_any_lt(rsq32,rcutoff2))
1239 r32 = _mm_mul_pd(rsq32,rinv32);
1241 /* EWALD ELECTROSTATICS */
1243 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1244 ewrt = _mm_mul_pd(r32,ewtabscale);
1245 ewitab = _mm_cvttpd_epi32(ewrt);
1246 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1247 ewitab = _mm_slli_epi32(ewitab,2);
1248 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1249 ewtabD = _mm_setzero_pd();
1250 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1251 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1252 ewtabFn = _mm_setzero_pd();
1253 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1254 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1255 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1256 velec = _mm_mul_pd(qq32,_mm_sub_pd(_mm_sub_pd(rinv32,sh_ewald),velec));
1257 felec = _mm_mul_pd(_mm_mul_pd(qq32,rinv32),_mm_sub_pd(rinvsq32,felec));
1259 cutoff_mask = _mm_cmplt_pd(rsq32,rcutoff2);
1261 /* Update potential sum for this i atom from the interaction with this j atom. */
1262 velec = _mm_and_pd(velec,cutoff_mask);
1263 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1264 velecsum = _mm_add_pd(velecsum,velec);
1268 fscal = _mm_and_pd(fscal,cutoff_mask);
1270 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1272 /* Calculate temporary vectorial force */
1273 tx = _mm_mul_pd(fscal,dx32);
1274 ty = _mm_mul_pd(fscal,dy32);
1275 tz = _mm_mul_pd(fscal,dz32);
1277 /* Update vectorial force */
1278 fix3 = _mm_add_pd(fix3,tx);
1279 fiy3 = _mm_add_pd(fiy3,ty);
1280 fiz3 = _mm_add_pd(fiz3,tz);
1282 fjx2 = _mm_add_pd(fjx2,tx);
1283 fjy2 = _mm_add_pd(fjy2,ty);
1284 fjz2 = _mm_add_pd(fjz2,tz);
1288 /**************************
1289 * CALCULATE INTERACTIONS *
1290 **************************/
1292 if (gmx_mm_any_lt(rsq33,rcutoff2))
1295 r33 = _mm_mul_pd(rsq33,rinv33);
1297 /* EWALD ELECTROSTATICS */
1299 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1300 ewrt = _mm_mul_pd(r33,ewtabscale);
1301 ewitab = _mm_cvttpd_epi32(ewrt);
1302 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1303 ewitab = _mm_slli_epi32(ewitab,2);
1304 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1305 ewtabD = _mm_setzero_pd();
1306 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1307 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1308 ewtabFn = _mm_setzero_pd();
1309 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1310 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1311 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1312 velec = _mm_mul_pd(qq33,_mm_sub_pd(_mm_sub_pd(rinv33,sh_ewald),velec));
1313 felec = _mm_mul_pd(_mm_mul_pd(qq33,rinv33),_mm_sub_pd(rinvsq33,felec));
1315 cutoff_mask = _mm_cmplt_pd(rsq33,rcutoff2);
1317 /* Update potential sum for this i atom from the interaction with this j atom. */
1318 velec = _mm_and_pd(velec,cutoff_mask);
1319 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1320 velecsum = _mm_add_pd(velecsum,velec);
1324 fscal = _mm_and_pd(fscal,cutoff_mask);
1326 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1328 /* Calculate temporary vectorial force */
1329 tx = _mm_mul_pd(fscal,dx33);
1330 ty = _mm_mul_pd(fscal,dy33);
1331 tz = _mm_mul_pd(fscal,dz33);
1333 /* Update vectorial force */
1334 fix3 = _mm_add_pd(fix3,tx);
1335 fiy3 = _mm_add_pd(fiy3,ty);
1336 fiz3 = _mm_add_pd(fiz3,tz);
1338 fjx3 = _mm_add_pd(fjx3,tx);
1339 fjy3 = _mm_add_pd(fjy3,ty);
1340 fjz3 = _mm_add_pd(fjz3,tz);
1344 gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA+DIM,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1346 /* Inner loop uses 414 flops */
1349 /* End of innermost loop */
1351 gmx_mm_update_iforce_3atom_swizzle_pd(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1352 f+i_coord_offset+DIM,fshift+i_shift_offset);
1355 /* Update potential energies */
1356 gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1358 /* Increment number of inner iterations */
1359 inneriter += j_index_end - j_index_start;
1361 /* Outer loop uses 19 flops */
1364 /* Increment number of outer iterations */
1367 /* Update outer/inner flops */
1369 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4W4_VF,outeriter*19 + inneriter*414);
1372 * Gromacs nonbonded kernel: nb_kernel_ElecEwSh_VdwNone_GeomW4W4_F_sse2_double
1373 * Electrostatics interaction: Ewald
1374 * VdW interaction: None
1375 * Geometry: Water4-Water4
1376 * Calculate force/pot: Force
1379 nb_kernel_ElecEwSh_VdwNone_GeomW4W4_F_sse2_double
1380 (t_nblist * gmx_restrict nlist,
1381 rvec * gmx_restrict xx,
1382 rvec * gmx_restrict ff,
1383 t_forcerec * gmx_restrict fr,
1384 t_mdatoms * gmx_restrict mdatoms,
1385 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1386 t_nrnb * gmx_restrict nrnb)
1388 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1389 * just 0 for non-waters.
1390 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
1391 * jnr indices corresponding to data put in the four positions in the SIMD register.
1393 int i_shift_offset,i_coord_offset,outeriter,inneriter;
1394 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1396 int j_coord_offsetA,j_coord_offsetB;
1397 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
1398 real rcutoff_scalar;
1399 real *shiftvec,*fshift,*x,*f;
1400 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1402 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1404 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1406 __m128d ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
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 int vdwjidx3A,vdwjidx3B;
1412 __m128d jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1413 __m128d dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1414 __m128d dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1415 __m128d dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1416 __m128d dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1417 __m128d dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1418 __m128d dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1419 __m128d dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1420 __m128d dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1421 __m128d dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1422 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
1425 __m128d ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1427 __m128d dummy_mask,cutoff_mask;
1428 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
1429 __m128d one = _mm_set1_pd(1.0);
1430 __m128d two = _mm_set1_pd(2.0);
1436 jindex = nlist->jindex;
1438 shiftidx = nlist->shift;
1440 shiftvec = fr->shift_vec[0];
1441 fshift = fr->fshift[0];
1442 facel = _mm_set1_pd(fr->epsfac);
1443 charge = mdatoms->chargeA;
1445 sh_ewald = _mm_set1_pd(fr->ic->sh_ewald);
1446 ewtab = fr->ic->tabq_coul_F;
1447 ewtabscale = _mm_set1_pd(fr->ic->tabq_scale);
1448 ewtabhalfspace = _mm_set1_pd(0.5/fr->ic->tabq_scale);
1450 /* Setup water-specific parameters */
1451 inr = nlist->iinr[0];
1452 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
1453 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
1454 iq3 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
1456 jq1 = _mm_set1_pd(charge[inr+1]);
1457 jq2 = _mm_set1_pd(charge[inr+2]);
1458 jq3 = _mm_set1_pd(charge[inr+3]);
1459 qq11 = _mm_mul_pd(iq1,jq1);
1460 qq12 = _mm_mul_pd(iq1,jq2);
1461 qq13 = _mm_mul_pd(iq1,jq3);
1462 qq21 = _mm_mul_pd(iq2,jq1);
1463 qq22 = _mm_mul_pd(iq2,jq2);
1464 qq23 = _mm_mul_pd(iq2,jq3);
1465 qq31 = _mm_mul_pd(iq3,jq1);
1466 qq32 = _mm_mul_pd(iq3,jq2);
1467 qq33 = _mm_mul_pd(iq3,jq3);
1469 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
1470 rcutoff_scalar = fr->rcoulomb;
1471 rcutoff = _mm_set1_pd(rcutoff_scalar);
1472 rcutoff2 = _mm_mul_pd(rcutoff,rcutoff);
1474 /* Avoid stupid compiler warnings */
1476 j_coord_offsetA = 0;
1477 j_coord_offsetB = 0;
1482 /* Start outer loop over neighborlists */
1483 for(iidx=0; iidx<nri; iidx++)
1485 /* Load shift vector for this list */
1486 i_shift_offset = DIM*shiftidx[iidx];
1488 /* Load limits for loop over neighbors */
1489 j_index_start = jindex[iidx];
1490 j_index_end = jindex[iidx+1];
1492 /* Get outer coordinate index */
1494 i_coord_offset = DIM*inr;
1496 /* Load i particle coords and add shift vector */
1497 gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
1498 &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1500 fix1 = _mm_setzero_pd();
1501 fiy1 = _mm_setzero_pd();
1502 fiz1 = _mm_setzero_pd();
1503 fix2 = _mm_setzero_pd();
1504 fiy2 = _mm_setzero_pd();
1505 fiz2 = _mm_setzero_pd();
1506 fix3 = _mm_setzero_pd();
1507 fiy3 = _mm_setzero_pd();
1508 fiz3 = _mm_setzero_pd();
1510 /* Start inner kernel loop */
1511 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
1514 /* Get j neighbor index, and coordinate index */
1516 jnrB = jjnr[jidx+1];
1517 j_coord_offsetA = DIM*jnrA;
1518 j_coord_offsetB = DIM*jnrB;
1520 /* load j atom coordinates */
1521 gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
1522 &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
1524 /* Calculate displacement vector */
1525 dx11 = _mm_sub_pd(ix1,jx1);
1526 dy11 = _mm_sub_pd(iy1,jy1);
1527 dz11 = _mm_sub_pd(iz1,jz1);
1528 dx12 = _mm_sub_pd(ix1,jx2);
1529 dy12 = _mm_sub_pd(iy1,jy2);
1530 dz12 = _mm_sub_pd(iz1,jz2);
1531 dx13 = _mm_sub_pd(ix1,jx3);
1532 dy13 = _mm_sub_pd(iy1,jy3);
1533 dz13 = _mm_sub_pd(iz1,jz3);
1534 dx21 = _mm_sub_pd(ix2,jx1);
1535 dy21 = _mm_sub_pd(iy2,jy1);
1536 dz21 = _mm_sub_pd(iz2,jz1);
1537 dx22 = _mm_sub_pd(ix2,jx2);
1538 dy22 = _mm_sub_pd(iy2,jy2);
1539 dz22 = _mm_sub_pd(iz2,jz2);
1540 dx23 = _mm_sub_pd(ix2,jx3);
1541 dy23 = _mm_sub_pd(iy2,jy3);
1542 dz23 = _mm_sub_pd(iz2,jz3);
1543 dx31 = _mm_sub_pd(ix3,jx1);
1544 dy31 = _mm_sub_pd(iy3,jy1);
1545 dz31 = _mm_sub_pd(iz3,jz1);
1546 dx32 = _mm_sub_pd(ix3,jx2);
1547 dy32 = _mm_sub_pd(iy3,jy2);
1548 dz32 = _mm_sub_pd(iz3,jz2);
1549 dx33 = _mm_sub_pd(ix3,jx3);
1550 dy33 = _mm_sub_pd(iy3,jy3);
1551 dz33 = _mm_sub_pd(iz3,jz3);
1553 /* Calculate squared distance and things based on it */
1554 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1555 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1556 rsq13 = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
1557 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1558 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1559 rsq23 = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
1560 rsq31 = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
1561 rsq32 = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
1562 rsq33 = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
1564 rinv11 = gmx_mm_invsqrt_pd(rsq11);
1565 rinv12 = gmx_mm_invsqrt_pd(rsq12);
1566 rinv13 = gmx_mm_invsqrt_pd(rsq13);
1567 rinv21 = gmx_mm_invsqrt_pd(rsq21);
1568 rinv22 = gmx_mm_invsqrt_pd(rsq22);
1569 rinv23 = gmx_mm_invsqrt_pd(rsq23);
1570 rinv31 = gmx_mm_invsqrt_pd(rsq31);
1571 rinv32 = gmx_mm_invsqrt_pd(rsq32);
1572 rinv33 = gmx_mm_invsqrt_pd(rsq33);
1574 rinvsq11 = _mm_mul_pd(rinv11,rinv11);
1575 rinvsq12 = _mm_mul_pd(rinv12,rinv12);
1576 rinvsq13 = _mm_mul_pd(rinv13,rinv13);
1577 rinvsq21 = _mm_mul_pd(rinv21,rinv21);
1578 rinvsq22 = _mm_mul_pd(rinv22,rinv22);
1579 rinvsq23 = _mm_mul_pd(rinv23,rinv23);
1580 rinvsq31 = _mm_mul_pd(rinv31,rinv31);
1581 rinvsq32 = _mm_mul_pd(rinv32,rinv32);
1582 rinvsq33 = _mm_mul_pd(rinv33,rinv33);
1584 fjx1 = _mm_setzero_pd();
1585 fjy1 = _mm_setzero_pd();
1586 fjz1 = _mm_setzero_pd();
1587 fjx2 = _mm_setzero_pd();
1588 fjy2 = _mm_setzero_pd();
1589 fjz2 = _mm_setzero_pd();
1590 fjx3 = _mm_setzero_pd();
1591 fjy3 = _mm_setzero_pd();
1592 fjz3 = _mm_setzero_pd();
1594 /**************************
1595 * CALCULATE INTERACTIONS *
1596 **************************/
1598 if (gmx_mm_any_lt(rsq11,rcutoff2))
1601 r11 = _mm_mul_pd(rsq11,rinv11);
1603 /* EWALD ELECTROSTATICS */
1605 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1606 ewrt = _mm_mul_pd(r11,ewtabscale);
1607 ewitab = _mm_cvttpd_epi32(ewrt);
1608 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1609 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1611 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1612 felec = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
1614 cutoff_mask = _mm_cmplt_pd(rsq11,rcutoff2);
1618 fscal = _mm_and_pd(fscal,cutoff_mask);
1620 /* Calculate temporary vectorial force */
1621 tx = _mm_mul_pd(fscal,dx11);
1622 ty = _mm_mul_pd(fscal,dy11);
1623 tz = _mm_mul_pd(fscal,dz11);
1625 /* Update vectorial force */
1626 fix1 = _mm_add_pd(fix1,tx);
1627 fiy1 = _mm_add_pd(fiy1,ty);
1628 fiz1 = _mm_add_pd(fiz1,tz);
1630 fjx1 = _mm_add_pd(fjx1,tx);
1631 fjy1 = _mm_add_pd(fjy1,ty);
1632 fjz1 = _mm_add_pd(fjz1,tz);
1636 /**************************
1637 * CALCULATE INTERACTIONS *
1638 **************************/
1640 if (gmx_mm_any_lt(rsq12,rcutoff2))
1643 r12 = _mm_mul_pd(rsq12,rinv12);
1645 /* EWALD ELECTROSTATICS */
1647 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1648 ewrt = _mm_mul_pd(r12,ewtabscale);
1649 ewitab = _mm_cvttpd_epi32(ewrt);
1650 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1651 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1653 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1654 felec = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
1656 cutoff_mask = _mm_cmplt_pd(rsq12,rcutoff2);
1660 fscal = _mm_and_pd(fscal,cutoff_mask);
1662 /* Calculate temporary vectorial force */
1663 tx = _mm_mul_pd(fscal,dx12);
1664 ty = _mm_mul_pd(fscal,dy12);
1665 tz = _mm_mul_pd(fscal,dz12);
1667 /* Update vectorial force */
1668 fix1 = _mm_add_pd(fix1,tx);
1669 fiy1 = _mm_add_pd(fiy1,ty);
1670 fiz1 = _mm_add_pd(fiz1,tz);
1672 fjx2 = _mm_add_pd(fjx2,tx);
1673 fjy2 = _mm_add_pd(fjy2,ty);
1674 fjz2 = _mm_add_pd(fjz2,tz);
1678 /**************************
1679 * CALCULATE INTERACTIONS *
1680 **************************/
1682 if (gmx_mm_any_lt(rsq13,rcutoff2))
1685 r13 = _mm_mul_pd(rsq13,rinv13);
1687 /* EWALD ELECTROSTATICS */
1689 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1690 ewrt = _mm_mul_pd(r13,ewtabscale);
1691 ewitab = _mm_cvttpd_epi32(ewrt);
1692 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1693 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1695 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1696 felec = _mm_mul_pd(_mm_mul_pd(qq13,rinv13),_mm_sub_pd(rinvsq13,felec));
1698 cutoff_mask = _mm_cmplt_pd(rsq13,rcutoff2);
1702 fscal = _mm_and_pd(fscal,cutoff_mask);
1704 /* Calculate temporary vectorial force */
1705 tx = _mm_mul_pd(fscal,dx13);
1706 ty = _mm_mul_pd(fscal,dy13);
1707 tz = _mm_mul_pd(fscal,dz13);
1709 /* Update vectorial force */
1710 fix1 = _mm_add_pd(fix1,tx);
1711 fiy1 = _mm_add_pd(fiy1,ty);
1712 fiz1 = _mm_add_pd(fiz1,tz);
1714 fjx3 = _mm_add_pd(fjx3,tx);
1715 fjy3 = _mm_add_pd(fjy3,ty);
1716 fjz3 = _mm_add_pd(fjz3,tz);
1720 /**************************
1721 * CALCULATE INTERACTIONS *
1722 **************************/
1724 if (gmx_mm_any_lt(rsq21,rcutoff2))
1727 r21 = _mm_mul_pd(rsq21,rinv21);
1729 /* EWALD ELECTROSTATICS */
1731 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1732 ewrt = _mm_mul_pd(r21,ewtabscale);
1733 ewitab = _mm_cvttpd_epi32(ewrt);
1734 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1735 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1737 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1738 felec = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
1740 cutoff_mask = _mm_cmplt_pd(rsq21,rcutoff2);
1744 fscal = _mm_and_pd(fscal,cutoff_mask);
1746 /* Calculate temporary vectorial force */
1747 tx = _mm_mul_pd(fscal,dx21);
1748 ty = _mm_mul_pd(fscal,dy21);
1749 tz = _mm_mul_pd(fscal,dz21);
1751 /* Update vectorial force */
1752 fix2 = _mm_add_pd(fix2,tx);
1753 fiy2 = _mm_add_pd(fiy2,ty);
1754 fiz2 = _mm_add_pd(fiz2,tz);
1756 fjx1 = _mm_add_pd(fjx1,tx);
1757 fjy1 = _mm_add_pd(fjy1,ty);
1758 fjz1 = _mm_add_pd(fjz1,tz);
1762 /**************************
1763 * CALCULATE INTERACTIONS *
1764 **************************/
1766 if (gmx_mm_any_lt(rsq22,rcutoff2))
1769 r22 = _mm_mul_pd(rsq22,rinv22);
1771 /* EWALD ELECTROSTATICS */
1773 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1774 ewrt = _mm_mul_pd(r22,ewtabscale);
1775 ewitab = _mm_cvttpd_epi32(ewrt);
1776 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1777 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1779 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1780 felec = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
1782 cutoff_mask = _mm_cmplt_pd(rsq22,rcutoff2);
1786 fscal = _mm_and_pd(fscal,cutoff_mask);
1788 /* Calculate temporary vectorial force */
1789 tx = _mm_mul_pd(fscal,dx22);
1790 ty = _mm_mul_pd(fscal,dy22);
1791 tz = _mm_mul_pd(fscal,dz22);
1793 /* Update vectorial force */
1794 fix2 = _mm_add_pd(fix2,tx);
1795 fiy2 = _mm_add_pd(fiy2,ty);
1796 fiz2 = _mm_add_pd(fiz2,tz);
1798 fjx2 = _mm_add_pd(fjx2,tx);
1799 fjy2 = _mm_add_pd(fjy2,ty);
1800 fjz2 = _mm_add_pd(fjz2,tz);
1804 /**************************
1805 * CALCULATE INTERACTIONS *
1806 **************************/
1808 if (gmx_mm_any_lt(rsq23,rcutoff2))
1811 r23 = _mm_mul_pd(rsq23,rinv23);
1813 /* EWALD ELECTROSTATICS */
1815 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1816 ewrt = _mm_mul_pd(r23,ewtabscale);
1817 ewitab = _mm_cvttpd_epi32(ewrt);
1818 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1819 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1821 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1822 felec = _mm_mul_pd(_mm_mul_pd(qq23,rinv23),_mm_sub_pd(rinvsq23,felec));
1824 cutoff_mask = _mm_cmplt_pd(rsq23,rcutoff2);
1828 fscal = _mm_and_pd(fscal,cutoff_mask);
1830 /* Calculate temporary vectorial force */
1831 tx = _mm_mul_pd(fscal,dx23);
1832 ty = _mm_mul_pd(fscal,dy23);
1833 tz = _mm_mul_pd(fscal,dz23);
1835 /* Update vectorial force */
1836 fix2 = _mm_add_pd(fix2,tx);
1837 fiy2 = _mm_add_pd(fiy2,ty);
1838 fiz2 = _mm_add_pd(fiz2,tz);
1840 fjx3 = _mm_add_pd(fjx3,tx);
1841 fjy3 = _mm_add_pd(fjy3,ty);
1842 fjz3 = _mm_add_pd(fjz3,tz);
1846 /**************************
1847 * CALCULATE INTERACTIONS *
1848 **************************/
1850 if (gmx_mm_any_lt(rsq31,rcutoff2))
1853 r31 = _mm_mul_pd(rsq31,rinv31);
1855 /* EWALD ELECTROSTATICS */
1857 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1858 ewrt = _mm_mul_pd(r31,ewtabscale);
1859 ewitab = _mm_cvttpd_epi32(ewrt);
1860 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1861 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1863 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1864 felec = _mm_mul_pd(_mm_mul_pd(qq31,rinv31),_mm_sub_pd(rinvsq31,felec));
1866 cutoff_mask = _mm_cmplt_pd(rsq31,rcutoff2);
1870 fscal = _mm_and_pd(fscal,cutoff_mask);
1872 /* Calculate temporary vectorial force */
1873 tx = _mm_mul_pd(fscal,dx31);
1874 ty = _mm_mul_pd(fscal,dy31);
1875 tz = _mm_mul_pd(fscal,dz31);
1877 /* Update vectorial force */
1878 fix3 = _mm_add_pd(fix3,tx);
1879 fiy3 = _mm_add_pd(fiy3,ty);
1880 fiz3 = _mm_add_pd(fiz3,tz);
1882 fjx1 = _mm_add_pd(fjx1,tx);
1883 fjy1 = _mm_add_pd(fjy1,ty);
1884 fjz1 = _mm_add_pd(fjz1,tz);
1888 /**************************
1889 * CALCULATE INTERACTIONS *
1890 **************************/
1892 if (gmx_mm_any_lt(rsq32,rcutoff2))
1895 r32 = _mm_mul_pd(rsq32,rinv32);
1897 /* EWALD ELECTROSTATICS */
1899 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1900 ewrt = _mm_mul_pd(r32,ewtabscale);
1901 ewitab = _mm_cvttpd_epi32(ewrt);
1902 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1903 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1905 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1906 felec = _mm_mul_pd(_mm_mul_pd(qq32,rinv32),_mm_sub_pd(rinvsq32,felec));
1908 cutoff_mask = _mm_cmplt_pd(rsq32,rcutoff2);
1912 fscal = _mm_and_pd(fscal,cutoff_mask);
1914 /* Calculate temporary vectorial force */
1915 tx = _mm_mul_pd(fscal,dx32);
1916 ty = _mm_mul_pd(fscal,dy32);
1917 tz = _mm_mul_pd(fscal,dz32);
1919 /* Update vectorial force */
1920 fix3 = _mm_add_pd(fix3,tx);
1921 fiy3 = _mm_add_pd(fiy3,ty);
1922 fiz3 = _mm_add_pd(fiz3,tz);
1924 fjx2 = _mm_add_pd(fjx2,tx);
1925 fjy2 = _mm_add_pd(fjy2,ty);
1926 fjz2 = _mm_add_pd(fjz2,tz);
1930 /**************************
1931 * CALCULATE INTERACTIONS *
1932 **************************/
1934 if (gmx_mm_any_lt(rsq33,rcutoff2))
1937 r33 = _mm_mul_pd(rsq33,rinv33);
1939 /* EWALD ELECTROSTATICS */
1941 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1942 ewrt = _mm_mul_pd(r33,ewtabscale);
1943 ewitab = _mm_cvttpd_epi32(ewrt);
1944 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1945 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1947 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1948 felec = _mm_mul_pd(_mm_mul_pd(qq33,rinv33),_mm_sub_pd(rinvsq33,felec));
1950 cutoff_mask = _mm_cmplt_pd(rsq33,rcutoff2);
1954 fscal = _mm_and_pd(fscal,cutoff_mask);
1956 /* Calculate temporary vectorial force */
1957 tx = _mm_mul_pd(fscal,dx33);
1958 ty = _mm_mul_pd(fscal,dy33);
1959 tz = _mm_mul_pd(fscal,dz33);
1961 /* Update vectorial force */
1962 fix3 = _mm_add_pd(fix3,tx);
1963 fiy3 = _mm_add_pd(fiy3,ty);
1964 fiz3 = _mm_add_pd(fiz3,tz);
1966 fjx3 = _mm_add_pd(fjx3,tx);
1967 fjy3 = _mm_add_pd(fjy3,ty);
1968 fjz3 = _mm_add_pd(fjz3,tz);
1972 gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA+DIM,f+j_coord_offsetB+DIM,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1974 /* Inner loop uses 351 flops */
1977 if(jidx<j_index_end)
1981 j_coord_offsetA = DIM*jnrA;
1983 /* load j atom coordinates */
1984 gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA+DIM,
1985 &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
1987 /* Calculate displacement vector */
1988 dx11 = _mm_sub_pd(ix1,jx1);
1989 dy11 = _mm_sub_pd(iy1,jy1);
1990 dz11 = _mm_sub_pd(iz1,jz1);
1991 dx12 = _mm_sub_pd(ix1,jx2);
1992 dy12 = _mm_sub_pd(iy1,jy2);
1993 dz12 = _mm_sub_pd(iz1,jz2);
1994 dx13 = _mm_sub_pd(ix1,jx3);
1995 dy13 = _mm_sub_pd(iy1,jy3);
1996 dz13 = _mm_sub_pd(iz1,jz3);
1997 dx21 = _mm_sub_pd(ix2,jx1);
1998 dy21 = _mm_sub_pd(iy2,jy1);
1999 dz21 = _mm_sub_pd(iz2,jz1);
2000 dx22 = _mm_sub_pd(ix2,jx2);
2001 dy22 = _mm_sub_pd(iy2,jy2);
2002 dz22 = _mm_sub_pd(iz2,jz2);
2003 dx23 = _mm_sub_pd(ix2,jx3);
2004 dy23 = _mm_sub_pd(iy2,jy3);
2005 dz23 = _mm_sub_pd(iz2,jz3);
2006 dx31 = _mm_sub_pd(ix3,jx1);
2007 dy31 = _mm_sub_pd(iy3,jy1);
2008 dz31 = _mm_sub_pd(iz3,jz1);
2009 dx32 = _mm_sub_pd(ix3,jx2);
2010 dy32 = _mm_sub_pd(iy3,jy2);
2011 dz32 = _mm_sub_pd(iz3,jz2);
2012 dx33 = _mm_sub_pd(ix3,jx3);
2013 dy33 = _mm_sub_pd(iy3,jy3);
2014 dz33 = _mm_sub_pd(iz3,jz3);
2016 /* Calculate squared distance and things based on it */
2017 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
2018 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
2019 rsq13 = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
2020 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
2021 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
2022 rsq23 = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
2023 rsq31 = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
2024 rsq32 = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
2025 rsq33 = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
2027 rinv11 = gmx_mm_invsqrt_pd(rsq11);
2028 rinv12 = gmx_mm_invsqrt_pd(rsq12);
2029 rinv13 = gmx_mm_invsqrt_pd(rsq13);
2030 rinv21 = gmx_mm_invsqrt_pd(rsq21);
2031 rinv22 = gmx_mm_invsqrt_pd(rsq22);
2032 rinv23 = gmx_mm_invsqrt_pd(rsq23);
2033 rinv31 = gmx_mm_invsqrt_pd(rsq31);
2034 rinv32 = gmx_mm_invsqrt_pd(rsq32);
2035 rinv33 = gmx_mm_invsqrt_pd(rsq33);
2037 rinvsq11 = _mm_mul_pd(rinv11,rinv11);
2038 rinvsq12 = _mm_mul_pd(rinv12,rinv12);
2039 rinvsq13 = _mm_mul_pd(rinv13,rinv13);
2040 rinvsq21 = _mm_mul_pd(rinv21,rinv21);
2041 rinvsq22 = _mm_mul_pd(rinv22,rinv22);
2042 rinvsq23 = _mm_mul_pd(rinv23,rinv23);
2043 rinvsq31 = _mm_mul_pd(rinv31,rinv31);
2044 rinvsq32 = _mm_mul_pd(rinv32,rinv32);
2045 rinvsq33 = _mm_mul_pd(rinv33,rinv33);
2047 fjx1 = _mm_setzero_pd();
2048 fjy1 = _mm_setzero_pd();
2049 fjz1 = _mm_setzero_pd();
2050 fjx2 = _mm_setzero_pd();
2051 fjy2 = _mm_setzero_pd();
2052 fjz2 = _mm_setzero_pd();
2053 fjx3 = _mm_setzero_pd();
2054 fjy3 = _mm_setzero_pd();
2055 fjz3 = _mm_setzero_pd();
2057 /**************************
2058 * CALCULATE INTERACTIONS *
2059 **************************/
2061 if (gmx_mm_any_lt(rsq11,rcutoff2))
2064 r11 = _mm_mul_pd(rsq11,rinv11);
2066 /* EWALD ELECTROSTATICS */
2068 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2069 ewrt = _mm_mul_pd(r11,ewtabscale);
2070 ewitab = _mm_cvttpd_epi32(ewrt);
2071 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2072 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2073 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2074 felec = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
2076 cutoff_mask = _mm_cmplt_pd(rsq11,rcutoff2);
2080 fscal = _mm_and_pd(fscal,cutoff_mask);
2082 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2084 /* Calculate temporary vectorial force */
2085 tx = _mm_mul_pd(fscal,dx11);
2086 ty = _mm_mul_pd(fscal,dy11);
2087 tz = _mm_mul_pd(fscal,dz11);
2089 /* Update vectorial force */
2090 fix1 = _mm_add_pd(fix1,tx);
2091 fiy1 = _mm_add_pd(fiy1,ty);
2092 fiz1 = _mm_add_pd(fiz1,tz);
2094 fjx1 = _mm_add_pd(fjx1,tx);
2095 fjy1 = _mm_add_pd(fjy1,ty);
2096 fjz1 = _mm_add_pd(fjz1,tz);
2100 /**************************
2101 * CALCULATE INTERACTIONS *
2102 **************************/
2104 if (gmx_mm_any_lt(rsq12,rcutoff2))
2107 r12 = _mm_mul_pd(rsq12,rinv12);
2109 /* EWALD ELECTROSTATICS */
2111 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2112 ewrt = _mm_mul_pd(r12,ewtabscale);
2113 ewitab = _mm_cvttpd_epi32(ewrt);
2114 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2115 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2116 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2117 felec = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
2119 cutoff_mask = _mm_cmplt_pd(rsq12,rcutoff2);
2123 fscal = _mm_and_pd(fscal,cutoff_mask);
2125 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2127 /* Calculate temporary vectorial force */
2128 tx = _mm_mul_pd(fscal,dx12);
2129 ty = _mm_mul_pd(fscal,dy12);
2130 tz = _mm_mul_pd(fscal,dz12);
2132 /* Update vectorial force */
2133 fix1 = _mm_add_pd(fix1,tx);
2134 fiy1 = _mm_add_pd(fiy1,ty);
2135 fiz1 = _mm_add_pd(fiz1,tz);
2137 fjx2 = _mm_add_pd(fjx2,tx);
2138 fjy2 = _mm_add_pd(fjy2,ty);
2139 fjz2 = _mm_add_pd(fjz2,tz);
2143 /**************************
2144 * CALCULATE INTERACTIONS *
2145 **************************/
2147 if (gmx_mm_any_lt(rsq13,rcutoff2))
2150 r13 = _mm_mul_pd(rsq13,rinv13);
2152 /* EWALD ELECTROSTATICS */
2154 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2155 ewrt = _mm_mul_pd(r13,ewtabscale);
2156 ewitab = _mm_cvttpd_epi32(ewrt);
2157 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2158 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2159 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2160 felec = _mm_mul_pd(_mm_mul_pd(qq13,rinv13),_mm_sub_pd(rinvsq13,felec));
2162 cutoff_mask = _mm_cmplt_pd(rsq13,rcutoff2);
2166 fscal = _mm_and_pd(fscal,cutoff_mask);
2168 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2170 /* Calculate temporary vectorial force */
2171 tx = _mm_mul_pd(fscal,dx13);
2172 ty = _mm_mul_pd(fscal,dy13);
2173 tz = _mm_mul_pd(fscal,dz13);
2175 /* Update vectorial force */
2176 fix1 = _mm_add_pd(fix1,tx);
2177 fiy1 = _mm_add_pd(fiy1,ty);
2178 fiz1 = _mm_add_pd(fiz1,tz);
2180 fjx3 = _mm_add_pd(fjx3,tx);
2181 fjy3 = _mm_add_pd(fjy3,ty);
2182 fjz3 = _mm_add_pd(fjz3,tz);
2186 /**************************
2187 * CALCULATE INTERACTIONS *
2188 **************************/
2190 if (gmx_mm_any_lt(rsq21,rcutoff2))
2193 r21 = _mm_mul_pd(rsq21,rinv21);
2195 /* EWALD ELECTROSTATICS */
2197 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2198 ewrt = _mm_mul_pd(r21,ewtabscale);
2199 ewitab = _mm_cvttpd_epi32(ewrt);
2200 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2201 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2202 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2203 felec = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
2205 cutoff_mask = _mm_cmplt_pd(rsq21,rcutoff2);
2209 fscal = _mm_and_pd(fscal,cutoff_mask);
2211 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2213 /* Calculate temporary vectorial force */
2214 tx = _mm_mul_pd(fscal,dx21);
2215 ty = _mm_mul_pd(fscal,dy21);
2216 tz = _mm_mul_pd(fscal,dz21);
2218 /* Update vectorial force */
2219 fix2 = _mm_add_pd(fix2,tx);
2220 fiy2 = _mm_add_pd(fiy2,ty);
2221 fiz2 = _mm_add_pd(fiz2,tz);
2223 fjx1 = _mm_add_pd(fjx1,tx);
2224 fjy1 = _mm_add_pd(fjy1,ty);
2225 fjz1 = _mm_add_pd(fjz1,tz);
2229 /**************************
2230 * CALCULATE INTERACTIONS *
2231 **************************/
2233 if (gmx_mm_any_lt(rsq22,rcutoff2))
2236 r22 = _mm_mul_pd(rsq22,rinv22);
2238 /* EWALD ELECTROSTATICS */
2240 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2241 ewrt = _mm_mul_pd(r22,ewtabscale);
2242 ewitab = _mm_cvttpd_epi32(ewrt);
2243 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2244 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2245 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2246 felec = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
2248 cutoff_mask = _mm_cmplt_pd(rsq22,rcutoff2);
2252 fscal = _mm_and_pd(fscal,cutoff_mask);
2254 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2256 /* Calculate temporary vectorial force */
2257 tx = _mm_mul_pd(fscal,dx22);
2258 ty = _mm_mul_pd(fscal,dy22);
2259 tz = _mm_mul_pd(fscal,dz22);
2261 /* Update vectorial force */
2262 fix2 = _mm_add_pd(fix2,tx);
2263 fiy2 = _mm_add_pd(fiy2,ty);
2264 fiz2 = _mm_add_pd(fiz2,tz);
2266 fjx2 = _mm_add_pd(fjx2,tx);
2267 fjy2 = _mm_add_pd(fjy2,ty);
2268 fjz2 = _mm_add_pd(fjz2,tz);
2272 /**************************
2273 * CALCULATE INTERACTIONS *
2274 **************************/
2276 if (gmx_mm_any_lt(rsq23,rcutoff2))
2279 r23 = _mm_mul_pd(rsq23,rinv23);
2281 /* EWALD ELECTROSTATICS */
2283 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2284 ewrt = _mm_mul_pd(r23,ewtabscale);
2285 ewitab = _mm_cvttpd_epi32(ewrt);
2286 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2287 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2288 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2289 felec = _mm_mul_pd(_mm_mul_pd(qq23,rinv23),_mm_sub_pd(rinvsq23,felec));
2291 cutoff_mask = _mm_cmplt_pd(rsq23,rcutoff2);
2295 fscal = _mm_and_pd(fscal,cutoff_mask);
2297 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2299 /* Calculate temporary vectorial force */
2300 tx = _mm_mul_pd(fscal,dx23);
2301 ty = _mm_mul_pd(fscal,dy23);
2302 tz = _mm_mul_pd(fscal,dz23);
2304 /* Update vectorial force */
2305 fix2 = _mm_add_pd(fix2,tx);
2306 fiy2 = _mm_add_pd(fiy2,ty);
2307 fiz2 = _mm_add_pd(fiz2,tz);
2309 fjx3 = _mm_add_pd(fjx3,tx);
2310 fjy3 = _mm_add_pd(fjy3,ty);
2311 fjz3 = _mm_add_pd(fjz3,tz);
2315 /**************************
2316 * CALCULATE INTERACTIONS *
2317 **************************/
2319 if (gmx_mm_any_lt(rsq31,rcutoff2))
2322 r31 = _mm_mul_pd(rsq31,rinv31);
2324 /* EWALD ELECTROSTATICS */
2326 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2327 ewrt = _mm_mul_pd(r31,ewtabscale);
2328 ewitab = _mm_cvttpd_epi32(ewrt);
2329 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2330 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2331 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2332 felec = _mm_mul_pd(_mm_mul_pd(qq31,rinv31),_mm_sub_pd(rinvsq31,felec));
2334 cutoff_mask = _mm_cmplt_pd(rsq31,rcutoff2);
2338 fscal = _mm_and_pd(fscal,cutoff_mask);
2340 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2342 /* Calculate temporary vectorial force */
2343 tx = _mm_mul_pd(fscal,dx31);
2344 ty = _mm_mul_pd(fscal,dy31);
2345 tz = _mm_mul_pd(fscal,dz31);
2347 /* Update vectorial force */
2348 fix3 = _mm_add_pd(fix3,tx);
2349 fiy3 = _mm_add_pd(fiy3,ty);
2350 fiz3 = _mm_add_pd(fiz3,tz);
2352 fjx1 = _mm_add_pd(fjx1,tx);
2353 fjy1 = _mm_add_pd(fjy1,ty);
2354 fjz1 = _mm_add_pd(fjz1,tz);
2358 /**************************
2359 * CALCULATE INTERACTIONS *
2360 **************************/
2362 if (gmx_mm_any_lt(rsq32,rcutoff2))
2365 r32 = _mm_mul_pd(rsq32,rinv32);
2367 /* EWALD ELECTROSTATICS */
2369 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2370 ewrt = _mm_mul_pd(r32,ewtabscale);
2371 ewitab = _mm_cvttpd_epi32(ewrt);
2372 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2373 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2374 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2375 felec = _mm_mul_pd(_mm_mul_pd(qq32,rinv32),_mm_sub_pd(rinvsq32,felec));
2377 cutoff_mask = _mm_cmplt_pd(rsq32,rcutoff2);
2381 fscal = _mm_and_pd(fscal,cutoff_mask);
2383 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2385 /* Calculate temporary vectorial force */
2386 tx = _mm_mul_pd(fscal,dx32);
2387 ty = _mm_mul_pd(fscal,dy32);
2388 tz = _mm_mul_pd(fscal,dz32);
2390 /* Update vectorial force */
2391 fix3 = _mm_add_pd(fix3,tx);
2392 fiy3 = _mm_add_pd(fiy3,ty);
2393 fiz3 = _mm_add_pd(fiz3,tz);
2395 fjx2 = _mm_add_pd(fjx2,tx);
2396 fjy2 = _mm_add_pd(fjy2,ty);
2397 fjz2 = _mm_add_pd(fjz2,tz);
2401 /**************************
2402 * CALCULATE INTERACTIONS *
2403 **************************/
2405 if (gmx_mm_any_lt(rsq33,rcutoff2))
2408 r33 = _mm_mul_pd(rsq33,rinv33);
2410 /* EWALD ELECTROSTATICS */
2412 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2413 ewrt = _mm_mul_pd(r33,ewtabscale);
2414 ewitab = _mm_cvttpd_epi32(ewrt);
2415 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2416 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2417 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2418 felec = _mm_mul_pd(_mm_mul_pd(qq33,rinv33),_mm_sub_pd(rinvsq33,felec));
2420 cutoff_mask = _mm_cmplt_pd(rsq33,rcutoff2);
2424 fscal = _mm_and_pd(fscal,cutoff_mask);
2426 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2428 /* Calculate temporary vectorial force */
2429 tx = _mm_mul_pd(fscal,dx33);
2430 ty = _mm_mul_pd(fscal,dy33);
2431 tz = _mm_mul_pd(fscal,dz33);
2433 /* Update vectorial force */
2434 fix3 = _mm_add_pd(fix3,tx);
2435 fiy3 = _mm_add_pd(fiy3,ty);
2436 fiz3 = _mm_add_pd(fiz3,tz);
2438 fjx3 = _mm_add_pd(fjx3,tx);
2439 fjy3 = _mm_add_pd(fjy3,ty);
2440 fjz3 = _mm_add_pd(fjz3,tz);
2444 gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA+DIM,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2446 /* Inner loop uses 351 flops */
2449 /* End of innermost loop */
2451 gmx_mm_update_iforce_3atom_swizzle_pd(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2452 f+i_coord_offset+DIM,fshift+i_shift_offset);
2454 /* Increment number of inner iterations */
2455 inneriter += j_index_end - j_index_start;
2457 /* Outer loop uses 18 flops */
2460 /* Increment number of outer iterations */
2463 /* Update outer/inner flops */
2465 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4W4_F,outeriter*18 + inneriter*351);