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 sse4_1_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_sse4_1_double.h"
48 #include "kernelutil_x86_sse4_1_double.h"
51 * Gromacs nonbonded kernel: nb_kernel_ElecEw_VdwNone_GeomW4W4_VF_sse4_1_double
52 * Electrostatics interaction: Ewald
53 * VdW interaction: None
54 * Geometry: Water4-Water4
55 * Calculate force/pot: PotentialAndForce
58 nb_kernel_ElecEw_VdwNone_GeomW4W4_VF_sse4_1_double
59 (t_nblist * gmx_restrict nlist,
60 rvec * gmx_restrict xx,
61 rvec * gmx_restrict ff,
62 t_forcerec * gmx_restrict fr,
63 t_mdatoms * gmx_restrict mdatoms,
64 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
65 t_nrnb * gmx_restrict nrnb)
67 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
68 * just 0 for non-waters.
69 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
70 * jnr indices corresponding to data put in the four positions in the SIMD register.
72 int i_shift_offset,i_coord_offset,outeriter,inneriter;
73 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
75 int j_coord_offsetA,j_coord_offsetB;
76 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
78 real *shiftvec,*fshift,*x,*f;
79 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
81 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
83 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
85 __m128d ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
86 int vdwjidx1A,vdwjidx1B;
87 __m128d jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
88 int vdwjidx2A,vdwjidx2B;
89 __m128d jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
90 int vdwjidx3A,vdwjidx3B;
91 __m128d jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
92 __m128d dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
93 __m128d dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
94 __m128d dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
95 __m128d dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
96 __m128d dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
97 __m128d dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
98 __m128d dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
99 __m128d dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
100 __m128d dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
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 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
132 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
133 iq3 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
135 jq1 = _mm_set1_pd(charge[inr+1]);
136 jq2 = _mm_set1_pd(charge[inr+2]);
137 jq3 = _mm_set1_pd(charge[inr+3]);
138 qq11 = _mm_mul_pd(iq1,jq1);
139 qq12 = _mm_mul_pd(iq1,jq2);
140 qq13 = _mm_mul_pd(iq1,jq3);
141 qq21 = _mm_mul_pd(iq2,jq1);
142 qq22 = _mm_mul_pd(iq2,jq2);
143 qq23 = _mm_mul_pd(iq2,jq3);
144 qq31 = _mm_mul_pd(iq3,jq1);
145 qq32 = _mm_mul_pd(iq3,jq2);
146 qq33 = _mm_mul_pd(iq3,jq3);
148 /* Avoid stupid compiler warnings */
156 /* Start outer loop over neighborlists */
157 for(iidx=0; iidx<nri; iidx++)
159 /* Load shift vector for this list */
160 i_shift_offset = DIM*shiftidx[iidx];
162 /* Load limits for loop over neighbors */
163 j_index_start = jindex[iidx];
164 j_index_end = jindex[iidx+1];
166 /* Get outer coordinate index */
168 i_coord_offset = DIM*inr;
170 /* Load i particle coords and add shift vector */
171 gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
172 &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
174 fix1 = _mm_setzero_pd();
175 fiy1 = _mm_setzero_pd();
176 fiz1 = _mm_setzero_pd();
177 fix2 = _mm_setzero_pd();
178 fiy2 = _mm_setzero_pd();
179 fiz2 = _mm_setzero_pd();
180 fix3 = _mm_setzero_pd();
181 fiy3 = _mm_setzero_pd();
182 fiz3 = _mm_setzero_pd();
184 /* Reset potential sums */
185 velecsum = _mm_setzero_pd();
187 /* Start inner kernel loop */
188 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
191 /* Get j neighbor index, and coordinate index */
194 j_coord_offsetA = DIM*jnrA;
195 j_coord_offsetB = DIM*jnrB;
197 /* load j atom coordinates */
198 gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
199 &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
201 /* Calculate displacement vector */
202 dx11 = _mm_sub_pd(ix1,jx1);
203 dy11 = _mm_sub_pd(iy1,jy1);
204 dz11 = _mm_sub_pd(iz1,jz1);
205 dx12 = _mm_sub_pd(ix1,jx2);
206 dy12 = _mm_sub_pd(iy1,jy2);
207 dz12 = _mm_sub_pd(iz1,jz2);
208 dx13 = _mm_sub_pd(ix1,jx3);
209 dy13 = _mm_sub_pd(iy1,jy3);
210 dz13 = _mm_sub_pd(iz1,jz3);
211 dx21 = _mm_sub_pd(ix2,jx1);
212 dy21 = _mm_sub_pd(iy2,jy1);
213 dz21 = _mm_sub_pd(iz2,jz1);
214 dx22 = _mm_sub_pd(ix2,jx2);
215 dy22 = _mm_sub_pd(iy2,jy2);
216 dz22 = _mm_sub_pd(iz2,jz2);
217 dx23 = _mm_sub_pd(ix2,jx3);
218 dy23 = _mm_sub_pd(iy2,jy3);
219 dz23 = _mm_sub_pd(iz2,jz3);
220 dx31 = _mm_sub_pd(ix3,jx1);
221 dy31 = _mm_sub_pd(iy3,jy1);
222 dz31 = _mm_sub_pd(iz3,jz1);
223 dx32 = _mm_sub_pd(ix3,jx2);
224 dy32 = _mm_sub_pd(iy3,jy2);
225 dz32 = _mm_sub_pd(iz3,jz2);
226 dx33 = _mm_sub_pd(ix3,jx3);
227 dy33 = _mm_sub_pd(iy3,jy3);
228 dz33 = _mm_sub_pd(iz3,jz3);
230 /* Calculate squared distance and things based on it */
231 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
232 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
233 rsq13 = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
234 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
235 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
236 rsq23 = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
237 rsq31 = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
238 rsq32 = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
239 rsq33 = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
241 rinv11 = gmx_mm_invsqrt_pd(rsq11);
242 rinv12 = gmx_mm_invsqrt_pd(rsq12);
243 rinv13 = gmx_mm_invsqrt_pd(rsq13);
244 rinv21 = gmx_mm_invsqrt_pd(rsq21);
245 rinv22 = gmx_mm_invsqrt_pd(rsq22);
246 rinv23 = gmx_mm_invsqrt_pd(rsq23);
247 rinv31 = gmx_mm_invsqrt_pd(rsq31);
248 rinv32 = gmx_mm_invsqrt_pd(rsq32);
249 rinv33 = gmx_mm_invsqrt_pd(rsq33);
251 rinvsq11 = _mm_mul_pd(rinv11,rinv11);
252 rinvsq12 = _mm_mul_pd(rinv12,rinv12);
253 rinvsq13 = _mm_mul_pd(rinv13,rinv13);
254 rinvsq21 = _mm_mul_pd(rinv21,rinv21);
255 rinvsq22 = _mm_mul_pd(rinv22,rinv22);
256 rinvsq23 = _mm_mul_pd(rinv23,rinv23);
257 rinvsq31 = _mm_mul_pd(rinv31,rinv31);
258 rinvsq32 = _mm_mul_pd(rinv32,rinv32);
259 rinvsq33 = _mm_mul_pd(rinv33,rinv33);
261 fjx1 = _mm_setzero_pd();
262 fjy1 = _mm_setzero_pd();
263 fjz1 = _mm_setzero_pd();
264 fjx2 = _mm_setzero_pd();
265 fjy2 = _mm_setzero_pd();
266 fjz2 = _mm_setzero_pd();
267 fjx3 = _mm_setzero_pd();
268 fjy3 = _mm_setzero_pd();
269 fjz3 = _mm_setzero_pd();
271 /**************************
272 * CALCULATE INTERACTIONS *
273 **************************/
275 r11 = _mm_mul_pd(rsq11,rinv11);
277 /* EWALD ELECTROSTATICS */
279 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
280 ewrt = _mm_mul_pd(r11,ewtabscale);
281 ewitab = _mm_cvttpd_epi32(ewrt);
282 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
283 ewitab = _mm_slli_epi32(ewitab,2);
284 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
285 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
286 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
287 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
288 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
289 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
290 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
291 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
292 velec = _mm_mul_pd(qq11,_mm_sub_pd(rinv11,velec));
293 felec = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
295 /* Update potential sum for this i atom from the interaction with this j atom. */
296 velecsum = _mm_add_pd(velecsum,velec);
300 /* Calculate temporary vectorial force */
301 tx = _mm_mul_pd(fscal,dx11);
302 ty = _mm_mul_pd(fscal,dy11);
303 tz = _mm_mul_pd(fscal,dz11);
305 /* Update vectorial force */
306 fix1 = _mm_add_pd(fix1,tx);
307 fiy1 = _mm_add_pd(fiy1,ty);
308 fiz1 = _mm_add_pd(fiz1,tz);
310 fjx1 = _mm_add_pd(fjx1,tx);
311 fjy1 = _mm_add_pd(fjy1,ty);
312 fjz1 = _mm_add_pd(fjz1,tz);
314 /**************************
315 * CALCULATE INTERACTIONS *
316 **************************/
318 r12 = _mm_mul_pd(rsq12,rinv12);
320 /* EWALD ELECTROSTATICS */
322 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
323 ewrt = _mm_mul_pd(r12,ewtabscale);
324 ewitab = _mm_cvttpd_epi32(ewrt);
325 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
326 ewitab = _mm_slli_epi32(ewitab,2);
327 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
328 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
329 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
330 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
331 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
332 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
333 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
334 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
335 velec = _mm_mul_pd(qq12,_mm_sub_pd(rinv12,velec));
336 felec = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
338 /* Update potential sum for this i atom from the interaction with this j atom. */
339 velecsum = _mm_add_pd(velecsum,velec);
343 /* Calculate temporary vectorial force */
344 tx = _mm_mul_pd(fscal,dx12);
345 ty = _mm_mul_pd(fscal,dy12);
346 tz = _mm_mul_pd(fscal,dz12);
348 /* Update vectorial force */
349 fix1 = _mm_add_pd(fix1,tx);
350 fiy1 = _mm_add_pd(fiy1,ty);
351 fiz1 = _mm_add_pd(fiz1,tz);
353 fjx2 = _mm_add_pd(fjx2,tx);
354 fjy2 = _mm_add_pd(fjy2,ty);
355 fjz2 = _mm_add_pd(fjz2,tz);
357 /**************************
358 * CALCULATE INTERACTIONS *
359 **************************/
361 r13 = _mm_mul_pd(rsq13,rinv13);
363 /* EWALD ELECTROSTATICS */
365 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
366 ewrt = _mm_mul_pd(r13,ewtabscale);
367 ewitab = _mm_cvttpd_epi32(ewrt);
368 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
369 ewitab = _mm_slli_epi32(ewitab,2);
370 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
371 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
372 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
373 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
374 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
375 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
376 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
377 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
378 velec = _mm_mul_pd(qq13,_mm_sub_pd(rinv13,velec));
379 felec = _mm_mul_pd(_mm_mul_pd(qq13,rinv13),_mm_sub_pd(rinvsq13,felec));
381 /* Update potential sum for this i atom from the interaction with this j atom. */
382 velecsum = _mm_add_pd(velecsum,velec);
386 /* Calculate temporary vectorial force */
387 tx = _mm_mul_pd(fscal,dx13);
388 ty = _mm_mul_pd(fscal,dy13);
389 tz = _mm_mul_pd(fscal,dz13);
391 /* Update vectorial force */
392 fix1 = _mm_add_pd(fix1,tx);
393 fiy1 = _mm_add_pd(fiy1,ty);
394 fiz1 = _mm_add_pd(fiz1,tz);
396 fjx3 = _mm_add_pd(fjx3,tx);
397 fjy3 = _mm_add_pd(fjy3,ty);
398 fjz3 = _mm_add_pd(fjz3,tz);
400 /**************************
401 * CALCULATE INTERACTIONS *
402 **************************/
404 r21 = _mm_mul_pd(rsq21,rinv21);
406 /* EWALD ELECTROSTATICS */
408 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
409 ewrt = _mm_mul_pd(r21,ewtabscale);
410 ewitab = _mm_cvttpd_epi32(ewrt);
411 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
412 ewitab = _mm_slli_epi32(ewitab,2);
413 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
414 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
415 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
416 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
417 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
418 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
419 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
420 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
421 velec = _mm_mul_pd(qq21,_mm_sub_pd(rinv21,velec));
422 felec = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
424 /* Update potential sum for this i atom from the interaction with this j atom. */
425 velecsum = _mm_add_pd(velecsum,velec);
429 /* Calculate temporary vectorial force */
430 tx = _mm_mul_pd(fscal,dx21);
431 ty = _mm_mul_pd(fscal,dy21);
432 tz = _mm_mul_pd(fscal,dz21);
434 /* Update vectorial force */
435 fix2 = _mm_add_pd(fix2,tx);
436 fiy2 = _mm_add_pd(fiy2,ty);
437 fiz2 = _mm_add_pd(fiz2,tz);
439 fjx1 = _mm_add_pd(fjx1,tx);
440 fjy1 = _mm_add_pd(fjy1,ty);
441 fjz1 = _mm_add_pd(fjz1,tz);
443 /**************************
444 * CALCULATE INTERACTIONS *
445 **************************/
447 r22 = _mm_mul_pd(rsq22,rinv22);
449 /* EWALD ELECTROSTATICS */
451 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
452 ewrt = _mm_mul_pd(r22,ewtabscale);
453 ewitab = _mm_cvttpd_epi32(ewrt);
454 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
455 ewitab = _mm_slli_epi32(ewitab,2);
456 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
457 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
458 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
459 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
460 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
461 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
462 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
463 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
464 velec = _mm_mul_pd(qq22,_mm_sub_pd(rinv22,velec));
465 felec = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
467 /* Update potential sum for this i atom from the interaction with this j atom. */
468 velecsum = _mm_add_pd(velecsum,velec);
472 /* Calculate temporary vectorial force */
473 tx = _mm_mul_pd(fscal,dx22);
474 ty = _mm_mul_pd(fscal,dy22);
475 tz = _mm_mul_pd(fscal,dz22);
477 /* Update vectorial force */
478 fix2 = _mm_add_pd(fix2,tx);
479 fiy2 = _mm_add_pd(fiy2,ty);
480 fiz2 = _mm_add_pd(fiz2,tz);
482 fjx2 = _mm_add_pd(fjx2,tx);
483 fjy2 = _mm_add_pd(fjy2,ty);
484 fjz2 = _mm_add_pd(fjz2,tz);
486 /**************************
487 * CALCULATE INTERACTIONS *
488 **************************/
490 r23 = _mm_mul_pd(rsq23,rinv23);
492 /* EWALD ELECTROSTATICS */
494 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
495 ewrt = _mm_mul_pd(r23,ewtabscale);
496 ewitab = _mm_cvttpd_epi32(ewrt);
497 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
498 ewitab = _mm_slli_epi32(ewitab,2);
499 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
500 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
501 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
502 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
503 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
504 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
505 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
506 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
507 velec = _mm_mul_pd(qq23,_mm_sub_pd(rinv23,velec));
508 felec = _mm_mul_pd(_mm_mul_pd(qq23,rinv23),_mm_sub_pd(rinvsq23,felec));
510 /* Update potential sum for this i atom from the interaction with this j atom. */
511 velecsum = _mm_add_pd(velecsum,velec);
515 /* Calculate temporary vectorial force */
516 tx = _mm_mul_pd(fscal,dx23);
517 ty = _mm_mul_pd(fscal,dy23);
518 tz = _mm_mul_pd(fscal,dz23);
520 /* Update vectorial force */
521 fix2 = _mm_add_pd(fix2,tx);
522 fiy2 = _mm_add_pd(fiy2,ty);
523 fiz2 = _mm_add_pd(fiz2,tz);
525 fjx3 = _mm_add_pd(fjx3,tx);
526 fjy3 = _mm_add_pd(fjy3,ty);
527 fjz3 = _mm_add_pd(fjz3,tz);
529 /**************************
530 * CALCULATE INTERACTIONS *
531 **************************/
533 r31 = _mm_mul_pd(rsq31,rinv31);
535 /* EWALD ELECTROSTATICS */
537 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
538 ewrt = _mm_mul_pd(r31,ewtabscale);
539 ewitab = _mm_cvttpd_epi32(ewrt);
540 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
541 ewitab = _mm_slli_epi32(ewitab,2);
542 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
543 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
544 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
545 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
546 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
547 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
548 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
549 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
550 velec = _mm_mul_pd(qq31,_mm_sub_pd(rinv31,velec));
551 felec = _mm_mul_pd(_mm_mul_pd(qq31,rinv31),_mm_sub_pd(rinvsq31,felec));
553 /* Update potential sum for this i atom from the interaction with this j atom. */
554 velecsum = _mm_add_pd(velecsum,velec);
558 /* Calculate temporary vectorial force */
559 tx = _mm_mul_pd(fscal,dx31);
560 ty = _mm_mul_pd(fscal,dy31);
561 tz = _mm_mul_pd(fscal,dz31);
563 /* Update vectorial force */
564 fix3 = _mm_add_pd(fix3,tx);
565 fiy3 = _mm_add_pd(fiy3,ty);
566 fiz3 = _mm_add_pd(fiz3,tz);
568 fjx1 = _mm_add_pd(fjx1,tx);
569 fjy1 = _mm_add_pd(fjy1,ty);
570 fjz1 = _mm_add_pd(fjz1,tz);
572 /**************************
573 * CALCULATE INTERACTIONS *
574 **************************/
576 r32 = _mm_mul_pd(rsq32,rinv32);
578 /* EWALD ELECTROSTATICS */
580 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
581 ewrt = _mm_mul_pd(r32,ewtabscale);
582 ewitab = _mm_cvttpd_epi32(ewrt);
583 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
584 ewitab = _mm_slli_epi32(ewitab,2);
585 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
586 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
587 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
588 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
589 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
590 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
591 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
592 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
593 velec = _mm_mul_pd(qq32,_mm_sub_pd(rinv32,velec));
594 felec = _mm_mul_pd(_mm_mul_pd(qq32,rinv32),_mm_sub_pd(rinvsq32,felec));
596 /* Update potential sum for this i atom from the interaction with this j atom. */
597 velecsum = _mm_add_pd(velecsum,velec);
601 /* Calculate temporary vectorial force */
602 tx = _mm_mul_pd(fscal,dx32);
603 ty = _mm_mul_pd(fscal,dy32);
604 tz = _mm_mul_pd(fscal,dz32);
606 /* Update vectorial force */
607 fix3 = _mm_add_pd(fix3,tx);
608 fiy3 = _mm_add_pd(fiy3,ty);
609 fiz3 = _mm_add_pd(fiz3,tz);
611 fjx2 = _mm_add_pd(fjx2,tx);
612 fjy2 = _mm_add_pd(fjy2,ty);
613 fjz2 = _mm_add_pd(fjz2,tz);
615 /**************************
616 * CALCULATE INTERACTIONS *
617 **************************/
619 r33 = _mm_mul_pd(rsq33,rinv33);
621 /* EWALD ELECTROSTATICS */
623 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
624 ewrt = _mm_mul_pd(r33,ewtabscale);
625 ewitab = _mm_cvttpd_epi32(ewrt);
626 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
627 ewitab = _mm_slli_epi32(ewitab,2);
628 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
629 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
630 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
631 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
632 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
633 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
634 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
635 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
636 velec = _mm_mul_pd(qq33,_mm_sub_pd(rinv33,velec));
637 felec = _mm_mul_pd(_mm_mul_pd(qq33,rinv33),_mm_sub_pd(rinvsq33,felec));
639 /* Update potential sum for this i atom from the interaction with this j atom. */
640 velecsum = _mm_add_pd(velecsum,velec);
644 /* Calculate temporary vectorial force */
645 tx = _mm_mul_pd(fscal,dx33);
646 ty = _mm_mul_pd(fscal,dy33);
647 tz = _mm_mul_pd(fscal,dz33);
649 /* Update vectorial force */
650 fix3 = _mm_add_pd(fix3,tx);
651 fiy3 = _mm_add_pd(fiy3,ty);
652 fiz3 = _mm_add_pd(fiz3,tz);
654 fjx3 = _mm_add_pd(fjx3,tx);
655 fjy3 = _mm_add_pd(fjy3,ty);
656 fjz3 = _mm_add_pd(fjz3,tz);
658 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);
660 /* Inner loop uses 369 flops */
667 j_coord_offsetA = DIM*jnrA;
669 /* load j atom coordinates */
670 gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA+DIM,
671 &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
673 /* Calculate displacement vector */
674 dx11 = _mm_sub_pd(ix1,jx1);
675 dy11 = _mm_sub_pd(iy1,jy1);
676 dz11 = _mm_sub_pd(iz1,jz1);
677 dx12 = _mm_sub_pd(ix1,jx2);
678 dy12 = _mm_sub_pd(iy1,jy2);
679 dz12 = _mm_sub_pd(iz1,jz2);
680 dx13 = _mm_sub_pd(ix1,jx3);
681 dy13 = _mm_sub_pd(iy1,jy3);
682 dz13 = _mm_sub_pd(iz1,jz3);
683 dx21 = _mm_sub_pd(ix2,jx1);
684 dy21 = _mm_sub_pd(iy2,jy1);
685 dz21 = _mm_sub_pd(iz2,jz1);
686 dx22 = _mm_sub_pd(ix2,jx2);
687 dy22 = _mm_sub_pd(iy2,jy2);
688 dz22 = _mm_sub_pd(iz2,jz2);
689 dx23 = _mm_sub_pd(ix2,jx3);
690 dy23 = _mm_sub_pd(iy2,jy3);
691 dz23 = _mm_sub_pd(iz2,jz3);
692 dx31 = _mm_sub_pd(ix3,jx1);
693 dy31 = _mm_sub_pd(iy3,jy1);
694 dz31 = _mm_sub_pd(iz3,jz1);
695 dx32 = _mm_sub_pd(ix3,jx2);
696 dy32 = _mm_sub_pd(iy3,jy2);
697 dz32 = _mm_sub_pd(iz3,jz2);
698 dx33 = _mm_sub_pd(ix3,jx3);
699 dy33 = _mm_sub_pd(iy3,jy3);
700 dz33 = _mm_sub_pd(iz3,jz3);
702 /* Calculate squared distance and things based on it */
703 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
704 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
705 rsq13 = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
706 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
707 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
708 rsq23 = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
709 rsq31 = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
710 rsq32 = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
711 rsq33 = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
713 rinv11 = gmx_mm_invsqrt_pd(rsq11);
714 rinv12 = gmx_mm_invsqrt_pd(rsq12);
715 rinv13 = gmx_mm_invsqrt_pd(rsq13);
716 rinv21 = gmx_mm_invsqrt_pd(rsq21);
717 rinv22 = gmx_mm_invsqrt_pd(rsq22);
718 rinv23 = gmx_mm_invsqrt_pd(rsq23);
719 rinv31 = gmx_mm_invsqrt_pd(rsq31);
720 rinv32 = gmx_mm_invsqrt_pd(rsq32);
721 rinv33 = gmx_mm_invsqrt_pd(rsq33);
723 rinvsq11 = _mm_mul_pd(rinv11,rinv11);
724 rinvsq12 = _mm_mul_pd(rinv12,rinv12);
725 rinvsq13 = _mm_mul_pd(rinv13,rinv13);
726 rinvsq21 = _mm_mul_pd(rinv21,rinv21);
727 rinvsq22 = _mm_mul_pd(rinv22,rinv22);
728 rinvsq23 = _mm_mul_pd(rinv23,rinv23);
729 rinvsq31 = _mm_mul_pd(rinv31,rinv31);
730 rinvsq32 = _mm_mul_pd(rinv32,rinv32);
731 rinvsq33 = _mm_mul_pd(rinv33,rinv33);
733 fjx1 = _mm_setzero_pd();
734 fjy1 = _mm_setzero_pd();
735 fjz1 = _mm_setzero_pd();
736 fjx2 = _mm_setzero_pd();
737 fjy2 = _mm_setzero_pd();
738 fjz2 = _mm_setzero_pd();
739 fjx3 = _mm_setzero_pd();
740 fjy3 = _mm_setzero_pd();
741 fjz3 = _mm_setzero_pd();
743 /**************************
744 * CALCULATE INTERACTIONS *
745 **************************/
747 r11 = _mm_mul_pd(rsq11,rinv11);
749 /* EWALD ELECTROSTATICS */
751 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
752 ewrt = _mm_mul_pd(r11,ewtabscale);
753 ewitab = _mm_cvttpd_epi32(ewrt);
754 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
755 ewitab = _mm_slli_epi32(ewitab,2);
756 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
757 ewtabD = _mm_setzero_pd();
758 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
759 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
760 ewtabFn = _mm_setzero_pd();
761 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
762 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
763 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
764 velec = _mm_mul_pd(qq11,_mm_sub_pd(rinv11,velec));
765 felec = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
767 /* Update potential sum for this i atom from the interaction with this j atom. */
768 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
769 velecsum = _mm_add_pd(velecsum,velec);
773 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
775 /* Calculate temporary vectorial force */
776 tx = _mm_mul_pd(fscal,dx11);
777 ty = _mm_mul_pd(fscal,dy11);
778 tz = _mm_mul_pd(fscal,dz11);
780 /* Update vectorial force */
781 fix1 = _mm_add_pd(fix1,tx);
782 fiy1 = _mm_add_pd(fiy1,ty);
783 fiz1 = _mm_add_pd(fiz1,tz);
785 fjx1 = _mm_add_pd(fjx1,tx);
786 fjy1 = _mm_add_pd(fjy1,ty);
787 fjz1 = _mm_add_pd(fjz1,tz);
789 /**************************
790 * CALCULATE INTERACTIONS *
791 **************************/
793 r12 = _mm_mul_pd(rsq12,rinv12);
795 /* EWALD ELECTROSTATICS */
797 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
798 ewrt = _mm_mul_pd(r12,ewtabscale);
799 ewitab = _mm_cvttpd_epi32(ewrt);
800 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
801 ewitab = _mm_slli_epi32(ewitab,2);
802 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
803 ewtabD = _mm_setzero_pd();
804 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
805 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
806 ewtabFn = _mm_setzero_pd();
807 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
808 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
809 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
810 velec = _mm_mul_pd(qq12,_mm_sub_pd(rinv12,velec));
811 felec = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
813 /* Update potential sum for this i atom from the interaction with this j atom. */
814 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
815 velecsum = _mm_add_pd(velecsum,velec);
819 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
821 /* Calculate temporary vectorial force */
822 tx = _mm_mul_pd(fscal,dx12);
823 ty = _mm_mul_pd(fscal,dy12);
824 tz = _mm_mul_pd(fscal,dz12);
826 /* Update vectorial force */
827 fix1 = _mm_add_pd(fix1,tx);
828 fiy1 = _mm_add_pd(fiy1,ty);
829 fiz1 = _mm_add_pd(fiz1,tz);
831 fjx2 = _mm_add_pd(fjx2,tx);
832 fjy2 = _mm_add_pd(fjy2,ty);
833 fjz2 = _mm_add_pd(fjz2,tz);
835 /**************************
836 * CALCULATE INTERACTIONS *
837 **************************/
839 r13 = _mm_mul_pd(rsq13,rinv13);
841 /* EWALD ELECTROSTATICS */
843 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
844 ewrt = _mm_mul_pd(r13,ewtabscale);
845 ewitab = _mm_cvttpd_epi32(ewrt);
846 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
847 ewitab = _mm_slli_epi32(ewitab,2);
848 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
849 ewtabD = _mm_setzero_pd();
850 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
851 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
852 ewtabFn = _mm_setzero_pd();
853 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
854 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
855 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
856 velec = _mm_mul_pd(qq13,_mm_sub_pd(rinv13,velec));
857 felec = _mm_mul_pd(_mm_mul_pd(qq13,rinv13),_mm_sub_pd(rinvsq13,felec));
859 /* Update potential sum for this i atom from the interaction with this j atom. */
860 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
861 velecsum = _mm_add_pd(velecsum,velec);
865 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
867 /* Calculate temporary vectorial force */
868 tx = _mm_mul_pd(fscal,dx13);
869 ty = _mm_mul_pd(fscal,dy13);
870 tz = _mm_mul_pd(fscal,dz13);
872 /* Update vectorial force */
873 fix1 = _mm_add_pd(fix1,tx);
874 fiy1 = _mm_add_pd(fiy1,ty);
875 fiz1 = _mm_add_pd(fiz1,tz);
877 fjx3 = _mm_add_pd(fjx3,tx);
878 fjy3 = _mm_add_pd(fjy3,ty);
879 fjz3 = _mm_add_pd(fjz3,tz);
881 /**************************
882 * CALCULATE INTERACTIONS *
883 **************************/
885 r21 = _mm_mul_pd(rsq21,rinv21);
887 /* EWALD ELECTROSTATICS */
889 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
890 ewrt = _mm_mul_pd(r21,ewtabscale);
891 ewitab = _mm_cvttpd_epi32(ewrt);
892 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
893 ewitab = _mm_slli_epi32(ewitab,2);
894 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
895 ewtabD = _mm_setzero_pd();
896 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
897 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
898 ewtabFn = _mm_setzero_pd();
899 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
900 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
901 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
902 velec = _mm_mul_pd(qq21,_mm_sub_pd(rinv21,velec));
903 felec = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
905 /* Update potential sum for this i atom from the interaction with this j atom. */
906 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
907 velecsum = _mm_add_pd(velecsum,velec);
911 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
913 /* Calculate temporary vectorial force */
914 tx = _mm_mul_pd(fscal,dx21);
915 ty = _mm_mul_pd(fscal,dy21);
916 tz = _mm_mul_pd(fscal,dz21);
918 /* Update vectorial force */
919 fix2 = _mm_add_pd(fix2,tx);
920 fiy2 = _mm_add_pd(fiy2,ty);
921 fiz2 = _mm_add_pd(fiz2,tz);
923 fjx1 = _mm_add_pd(fjx1,tx);
924 fjy1 = _mm_add_pd(fjy1,ty);
925 fjz1 = _mm_add_pd(fjz1,tz);
927 /**************************
928 * CALCULATE INTERACTIONS *
929 **************************/
931 r22 = _mm_mul_pd(rsq22,rinv22);
933 /* EWALD ELECTROSTATICS */
935 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
936 ewrt = _mm_mul_pd(r22,ewtabscale);
937 ewitab = _mm_cvttpd_epi32(ewrt);
938 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
939 ewitab = _mm_slli_epi32(ewitab,2);
940 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
941 ewtabD = _mm_setzero_pd();
942 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
943 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
944 ewtabFn = _mm_setzero_pd();
945 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
946 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
947 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
948 velec = _mm_mul_pd(qq22,_mm_sub_pd(rinv22,velec));
949 felec = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
951 /* Update potential sum for this i atom from the interaction with this j atom. */
952 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
953 velecsum = _mm_add_pd(velecsum,velec);
957 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
959 /* Calculate temporary vectorial force */
960 tx = _mm_mul_pd(fscal,dx22);
961 ty = _mm_mul_pd(fscal,dy22);
962 tz = _mm_mul_pd(fscal,dz22);
964 /* Update vectorial force */
965 fix2 = _mm_add_pd(fix2,tx);
966 fiy2 = _mm_add_pd(fiy2,ty);
967 fiz2 = _mm_add_pd(fiz2,tz);
969 fjx2 = _mm_add_pd(fjx2,tx);
970 fjy2 = _mm_add_pd(fjy2,ty);
971 fjz2 = _mm_add_pd(fjz2,tz);
973 /**************************
974 * CALCULATE INTERACTIONS *
975 **************************/
977 r23 = _mm_mul_pd(rsq23,rinv23);
979 /* EWALD ELECTROSTATICS */
981 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
982 ewrt = _mm_mul_pd(r23,ewtabscale);
983 ewitab = _mm_cvttpd_epi32(ewrt);
984 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
985 ewitab = _mm_slli_epi32(ewitab,2);
986 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
987 ewtabD = _mm_setzero_pd();
988 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
989 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
990 ewtabFn = _mm_setzero_pd();
991 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
992 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
993 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
994 velec = _mm_mul_pd(qq23,_mm_sub_pd(rinv23,velec));
995 felec = _mm_mul_pd(_mm_mul_pd(qq23,rinv23),_mm_sub_pd(rinvsq23,felec));
997 /* Update potential sum for this i atom from the interaction with this j atom. */
998 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
999 velecsum = _mm_add_pd(velecsum,velec);
1003 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1005 /* Calculate temporary vectorial force */
1006 tx = _mm_mul_pd(fscal,dx23);
1007 ty = _mm_mul_pd(fscal,dy23);
1008 tz = _mm_mul_pd(fscal,dz23);
1010 /* Update vectorial force */
1011 fix2 = _mm_add_pd(fix2,tx);
1012 fiy2 = _mm_add_pd(fiy2,ty);
1013 fiz2 = _mm_add_pd(fiz2,tz);
1015 fjx3 = _mm_add_pd(fjx3,tx);
1016 fjy3 = _mm_add_pd(fjy3,ty);
1017 fjz3 = _mm_add_pd(fjz3,tz);
1019 /**************************
1020 * CALCULATE INTERACTIONS *
1021 **************************/
1023 r31 = _mm_mul_pd(rsq31,rinv31);
1025 /* EWALD ELECTROSTATICS */
1027 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1028 ewrt = _mm_mul_pd(r31,ewtabscale);
1029 ewitab = _mm_cvttpd_epi32(ewrt);
1030 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1031 ewitab = _mm_slli_epi32(ewitab,2);
1032 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1033 ewtabD = _mm_setzero_pd();
1034 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1035 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1036 ewtabFn = _mm_setzero_pd();
1037 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1038 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1039 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1040 velec = _mm_mul_pd(qq31,_mm_sub_pd(rinv31,velec));
1041 felec = _mm_mul_pd(_mm_mul_pd(qq31,rinv31),_mm_sub_pd(rinvsq31,felec));
1043 /* Update potential sum for this i atom from the interaction with this j atom. */
1044 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1045 velecsum = _mm_add_pd(velecsum,velec);
1049 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1051 /* Calculate temporary vectorial force */
1052 tx = _mm_mul_pd(fscal,dx31);
1053 ty = _mm_mul_pd(fscal,dy31);
1054 tz = _mm_mul_pd(fscal,dz31);
1056 /* Update vectorial force */
1057 fix3 = _mm_add_pd(fix3,tx);
1058 fiy3 = _mm_add_pd(fiy3,ty);
1059 fiz3 = _mm_add_pd(fiz3,tz);
1061 fjx1 = _mm_add_pd(fjx1,tx);
1062 fjy1 = _mm_add_pd(fjy1,ty);
1063 fjz1 = _mm_add_pd(fjz1,tz);
1065 /**************************
1066 * CALCULATE INTERACTIONS *
1067 **************************/
1069 r32 = _mm_mul_pd(rsq32,rinv32);
1071 /* EWALD ELECTROSTATICS */
1073 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1074 ewrt = _mm_mul_pd(r32,ewtabscale);
1075 ewitab = _mm_cvttpd_epi32(ewrt);
1076 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
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(qq32,_mm_sub_pd(rinv32,velec));
1087 felec = _mm_mul_pd(_mm_mul_pd(qq32,rinv32),_mm_sub_pd(rinvsq32,felec));
1089 /* Update potential sum for this i atom from the interaction with this j atom. */
1090 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1091 velecsum = _mm_add_pd(velecsum,velec);
1095 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1097 /* Calculate temporary vectorial force */
1098 tx = _mm_mul_pd(fscal,dx32);
1099 ty = _mm_mul_pd(fscal,dy32);
1100 tz = _mm_mul_pd(fscal,dz32);
1102 /* Update vectorial force */
1103 fix3 = _mm_add_pd(fix3,tx);
1104 fiy3 = _mm_add_pd(fiy3,ty);
1105 fiz3 = _mm_add_pd(fiz3,tz);
1107 fjx2 = _mm_add_pd(fjx2,tx);
1108 fjy2 = _mm_add_pd(fjy2,ty);
1109 fjz2 = _mm_add_pd(fjz2,tz);
1111 /**************************
1112 * CALCULATE INTERACTIONS *
1113 **************************/
1115 r33 = _mm_mul_pd(rsq33,rinv33);
1117 /* EWALD ELECTROSTATICS */
1119 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1120 ewrt = _mm_mul_pd(r33,ewtabscale);
1121 ewitab = _mm_cvttpd_epi32(ewrt);
1122 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1123 ewitab = _mm_slli_epi32(ewitab,2);
1124 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1125 ewtabD = _mm_setzero_pd();
1126 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1127 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1128 ewtabFn = _mm_setzero_pd();
1129 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1130 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1131 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1132 velec = _mm_mul_pd(qq33,_mm_sub_pd(rinv33,velec));
1133 felec = _mm_mul_pd(_mm_mul_pd(qq33,rinv33),_mm_sub_pd(rinvsq33,felec));
1135 /* Update potential sum for this i atom from the interaction with this j atom. */
1136 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1137 velecsum = _mm_add_pd(velecsum,velec);
1141 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1143 /* Calculate temporary vectorial force */
1144 tx = _mm_mul_pd(fscal,dx33);
1145 ty = _mm_mul_pd(fscal,dy33);
1146 tz = _mm_mul_pd(fscal,dz33);
1148 /* Update vectorial force */
1149 fix3 = _mm_add_pd(fix3,tx);
1150 fiy3 = _mm_add_pd(fiy3,ty);
1151 fiz3 = _mm_add_pd(fiz3,tz);
1153 fjx3 = _mm_add_pd(fjx3,tx);
1154 fjy3 = _mm_add_pd(fjy3,ty);
1155 fjz3 = _mm_add_pd(fjz3,tz);
1157 gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA+DIM,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1159 /* Inner loop uses 369 flops */
1162 /* End of innermost loop */
1164 gmx_mm_update_iforce_3atom_swizzle_pd(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1165 f+i_coord_offset+DIM,fshift+i_shift_offset);
1168 /* Update potential energies */
1169 gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1171 /* Increment number of inner iterations */
1172 inneriter += j_index_end - j_index_start;
1174 /* Outer loop uses 19 flops */
1177 /* Increment number of outer iterations */
1180 /* Update outer/inner flops */
1182 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4W4_VF,outeriter*19 + inneriter*369);
1185 * Gromacs nonbonded kernel: nb_kernel_ElecEw_VdwNone_GeomW4W4_F_sse4_1_double
1186 * Electrostatics interaction: Ewald
1187 * VdW interaction: None
1188 * Geometry: Water4-Water4
1189 * Calculate force/pot: Force
1192 nb_kernel_ElecEw_VdwNone_GeomW4W4_F_sse4_1_double
1193 (t_nblist * gmx_restrict nlist,
1194 rvec * gmx_restrict xx,
1195 rvec * gmx_restrict ff,
1196 t_forcerec * gmx_restrict fr,
1197 t_mdatoms * gmx_restrict mdatoms,
1198 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1199 t_nrnb * gmx_restrict nrnb)
1201 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1202 * just 0 for non-waters.
1203 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
1204 * jnr indices corresponding to data put in the four positions in the SIMD register.
1206 int i_shift_offset,i_coord_offset,outeriter,inneriter;
1207 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1209 int j_coord_offsetA,j_coord_offsetB;
1210 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
1211 real rcutoff_scalar;
1212 real *shiftvec,*fshift,*x,*f;
1213 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1215 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1217 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1219 __m128d ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1220 int vdwjidx1A,vdwjidx1B;
1221 __m128d jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1222 int vdwjidx2A,vdwjidx2B;
1223 __m128d jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1224 int vdwjidx3A,vdwjidx3B;
1225 __m128d jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1226 __m128d dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1227 __m128d dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1228 __m128d dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1229 __m128d dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1230 __m128d dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1231 __m128d dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1232 __m128d dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1233 __m128d dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1234 __m128d dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1235 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
1238 __m128d ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1240 __m128d dummy_mask,cutoff_mask;
1241 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
1242 __m128d one = _mm_set1_pd(1.0);
1243 __m128d two = _mm_set1_pd(2.0);
1249 jindex = nlist->jindex;
1251 shiftidx = nlist->shift;
1253 shiftvec = fr->shift_vec[0];
1254 fshift = fr->fshift[0];
1255 facel = _mm_set1_pd(fr->epsfac);
1256 charge = mdatoms->chargeA;
1258 sh_ewald = _mm_set1_pd(fr->ic->sh_ewald);
1259 ewtab = fr->ic->tabq_coul_F;
1260 ewtabscale = _mm_set1_pd(fr->ic->tabq_scale);
1261 ewtabhalfspace = _mm_set1_pd(0.5/fr->ic->tabq_scale);
1263 /* Setup water-specific parameters */
1264 inr = nlist->iinr[0];
1265 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
1266 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
1267 iq3 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
1269 jq1 = _mm_set1_pd(charge[inr+1]);
1270 jq2 = _mm_set1_pd(charge[inr+2]);
1271 jq3 = _mm_set1_pd(charge[inr+3]);
1272 qq11 = _mm_mul_pd(iq1,jq1);
1273 qq12 = _mm_mul_pd(iq1,jq2);
1274 qq13 = _mm_mul_pd(iq1,jq3);
1275 qq21 = _mm_mul_pd(iq2,jq1);
1276 qq22 = _mm_mul_pd(iq2,jq2);
1277 qq23 = _mm_mul_pd(iq2,jq3);
1278 qq31 = _mm_mul_pd(iq3,jq1);
1279 qq32 = _mm_mul_pd(iq3,jq2);
1280 qq33 = _mm_mul_pd(iq3,jq3);
1282 /* Avoid stupid compiler warnings */
1284 j_coord_offsetA = 0;
1285 j_coord_offsetB = 0;
1290 /* Start outer loop over neighborlists */
1291 for(iidx=0; iidx<nri; iidx++)
1293 /* Load shift vector for this list */
1294 i_shift_offset = DIM*shiftidx[iidx];
1296 /* Load limits for loop over neighbors */
1297 j_index_start = jindex[iidx];
1298 j_index_end = jindex[iidx+1];
1300 /* Get outer coordinate index */
1302 i_coord_offset = DIM*inr;
1304 /* Load i particle coords and add shift vector */
1305 gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
1306 &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1308 fix1 = _mm_setzero_pd();
1309 fiy1 = _mm_setzero_pd();
1310 fiz1 = _mm_setzero_pd();
1311 fix2 = _mm_setzero_pd();
1312 fiy2 = _mm_setzero_pd();
1313 fiz2 = _mm_setzero_pd();
1314 fix3 = _mm_setzero_pd();
1315 fiy3 = _mm_setzero_pd();
1316 fiz3 = _mm_setzero_pd();
1318 /* Start inner kernel loop */
1319 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
1322 /* Get j neighbor index, and coordinate index */
1324 jnrB = jjnr[jidx+1];
1325 j_coord_offsetA = DIM*jnrA;
1326 j_coord_offsetB = DIM*jnrB;
1328 /* load j atom coordinates */
1329 gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
1330 &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
1332 /* Calculate displacement vector */
1333 dx11 = _mm_sub_pd(ix1,jx1);
1334 dy11 = _mm_sub_pd(iy1,jy1);
1335 dz11 = _mm_sub_pd(iz1,jz1);
1336 dx12 = _mm_sub_pd(ix1,jx2);
1337 dy12 = _mm_sub_pd(iy1,jy2);
1338 dz12 = _mm_sub_pd(iz1,jz2);
1339 dx13 = _mm_sub_pd(ix1,jx3);
1340 dy13 = _mm_sub_pd(iy1,jy3);
1341 dz13 = _mm_sub_pd(iz1,jz3);
1342 dx21 = _mm_sub_pd(ix2,jx1);
1343 dy21 = _mm_sub_pd(iy2,jy1);
1344 dz21 = _mm_sub_pd(iz2,jz1);
1345 dx22 = _mm_sub_pd(ix2,jx2);
1346 dy22 = _mm_sub_pd(iy2,jy2);
1347 dz22 = _mm_sub_pd(iz2,jz2);
1348 dx23 = _mm_sub_pd(ix2,jx3);
1349 dy23 = _mm_sub_pd(iy2,jy3);
1350 dz23 = _mm_sub_pd(iz2,jz3);
1351 dx31 = _mm_sub_pd(ix3,jx1);
1352 dy31 = _mm_sub_pd(iy3,jy1);
1353 dz31 = _mm_sub_pd(iz3,jz1);
1354 dx32 = _mm_sub_pd(ix3,jx2);
1355 dy32 = _mm_sub_pd(iy3,jy2);
1356 dz32 = _mm_sub_pd(iz3,jz2);
1357 dx33 = _mm_sub_pd(ix3,jx3);
1358 dy33 = _mm_sub_pd(iy3,jy3);
1359 dz33 = _mm_sub_pd(iz3,jz3);
1361 /* Calculate squared distance and things based on it */
1362 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1363 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1364 rsq13 = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
1365 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1366 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1367 rsq23 = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
1368 rsq31 = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
1369 rsq32 = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
1370 rsq33 = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
1372 rinv11 = gmx_mm_invsqrt_pd(rsq11);
1373 rinv12 = gmx_mm_invsqrt_pd(rsq12);
1374 rinv13 = gmx_mm_invsqrt_pd(rsq13);
1375 rinv21 = gmx_mm_invsqrt_pd(rsq21);
1376 rinv22 = gmx_mm_invsqrt_pd(rsq22);
1377 rinv23 = gmx_mm_invsqrt_pd(rsq23);
1378 rinv31 = gmx_mm_invsqrt_pd(rsq31);
1379 rinv32 = gmx_mm_invsqrt_pd(rsq32);
1380 rinv33 = gmx_mm_invsqrt_pd(rsq33);
1382 rinvsq11 = _mm_mul_pd(rinv11,rinv11);
1383 rinvsq12 = _mm_mul_pd(rinv12,rinv12);
1384 rinvsq13 = _mm_mul_pd(rinv13,rinv13);
1385 rinvsq21 = _mm_mul_pd(rinv21,rinv21);
1386 rinvsq22 = _mm_mul_pd(rinv22,rinv22);
1387 rinvsq23 = _mm_mul_pd(rinv23,rinv23);
1388 rinvsq31 = _mm_mul_pd(rinv31,rinv31);
1389 rinvsq32 = _mm_mul_pd(rinv32,rinv32);
1390 rinvsq33 = _mm_mul_pd(rinv33,rinv33);
1392 fjx1 = _mm_setzero_pd();
1393 fjy1 = _mm_setzero_pd();
1394 fjz1 = _mm_setzero_pd();
1395 fjx2 = _mm_setzero_pd();
1396 fjy2 = _mm_setzero_pd();
1397 fjz2 = _mm_setzero_pd();
1398 fjx3 = _mm_setzero_pd();
1399 fjy3 = _mm_setzero_pd();
1400 fjz3 = _mm_setzero_pd();
1402 /**************************
1403 * CALCULATE INTERACTIONS *
1404 **************************/
1406 r11 = _mm_mul_pd(rsq11,rinv11);
1408 /* EWALD ELECTROSTATICS */
1410 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1411 ewrt = _mm_mul_pd(r11,ewtabscale);
1412 ewitab = _mm_cvttpd_epi32(ewrt);
1413 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1414 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1416 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1417 felec = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
1421 /* Calculate temporary vectorial force */
1422 tx = _mm_mul_pd(fscal,dx11);
1423 ty = _mm_mul_pd(fscal,dy11);
1424 tz = _mm_mul_pd(fscal,dz11);
1426 /* Update vectorial force */
1427 fix1 = _mm_add_pd(fix1,tx);
1428 fiy1 = _mm_add_pd(fiy1,ty);
1429 fiz1 = _mm_add_pd(fiz1,tz);
1431 fjx1 = _mm_add_pd(fjx1,tx);
1432 fjy1 = _mm_add_pd(fjy1,ty);
1433 fjz1 = _mm_add_pd(fjz1,tz);
1435 /**************************
1436 * CALCULATE INTERACTIONS *
1437 **************************/
1439 r12 = _mm_mul_pd(rsq12,rinv12);
1441 /* EWALD ELECTROSTATICS */
1443 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1444 ewrt = _mm_mul_pd(r12,ewtabscale);
1445 ewitab = _mm_cvttpd_epi32(ewrt);
1446 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1447 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1449 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1450 felec = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
1454 /* Calculate temporary vectorial force */
1455 tx = _mm_mul_pd(fscal,dx12);
1456 ty = _mm_mul_pd(fscal,dy12);
1457 tz = _mm_mul_pd(fscal,dz12);
1459 /* Update vectorial force */
1460 fix1 = _mm_add_pd(fix1,tx);
1461 fiy1 = _mm_add_pd(fiy1,ty);
1462 fiz1 = _mm_add_pd(fiz1,tz);
1464 fjx2 = _mm_add_pd(fjx2,tx);
1465 fjy2 = _mm_add_pd(fjy2,ty);
1466 fjz2 = _mm_add_pd(fjz2,tz);
1468 /**************************
1469 * CALCULATE INTERACTIONS *
1470 **************************/
1472 r13 = _mm_mul_pd(rsq13,rinv13);
1474 /* EWALD ELECTROSTATICS */
1476 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1477 ewrt = _mm_mul_pd(r13,ewtabscale);
1478 ewitab = _mm_cvttpd_epi32(ewrt);
1479 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1480 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1482 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1483 felec = _mm_mul_pd(_mm_mul_pd(qq13,rinv13),_mm_sub_pd(rinvsq13,felec));
1487 /* Calculate temporary vectorial force */
1488 tx = _mm_mul_pd(fscal,dx13);
1489 ty = _mm_mul_pd(fscal,dy13);
1490 tz = _mm_mul_pd(fscal,dz13);
1492 /* Update vectorial force */
1493 fix1 = _mm_add_pd(fix1,tx);
1494 fiy1 = _mm_add_pd(fiy1,ty);
1495 fiz1 = _mm_add_pd(fiz1,tz);
1497 fjx3 = _mm_add_pd(fjx3,tx);
1498 fjy3 = _mm_add_pd(fjy3,ty);
1499 fjz3 = _mm_add_pd(fjz3,tz);
1501 /**************************
1502 * CALCULATE INTERACTIONS *
1503 **************************/
1505 r21 = _mm_mul_pd(rsq21,rinv21);
1507 /* EWALD ELECTROSTATICS */
1509 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1510 ewrt = _mm_mul_pd(r21,ewtabscale);
1511 ewitab = _mm_cvttpd_epi32(ewrt);
1512 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1513 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1515 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1516 felec = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
1520 /* Calculate temporary vectorial force */
1521 tx = _mm_mul_pd(fscal,dx21);
1522 ty = _mm_mul_pd(fscal,dy21);
1523 tz = _mm_mul_pd(fscal,dz21);
1525 /* Update vectorial force */
1526 fix2 = _mm_add_pd(fix2,tx);
1527 fiy2 = _mm_add_pd(fiy2,ty);
1528 fiz2 = _mm_add_pd(fiz2,tz);
1530 fjx1 = _mm_add_pd(fjx1,tx);
1531 fjy1 = _mm_add_pd(fjy1,ty);
1532 fjz1 = _mm_add_pd(fjz1,tz);
1534 /**************************
1535 * CALCULATE INTERACTIONS *
1536 **************************/
1538 r22 = _mm_mul_pd(rsq22,rinv22);
1540 /* EWALD ELECTROSTATICS */
1542 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1543 ewrt = _mm_mul_pd(r22,ewtabscale);
1544 ewitab = _mm_cvttpd_epi32(ewrt);
1545 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1546 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1548 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1549 felec = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
1553 /* Calculate temporary vectorial force */
1554 tx = _mm_mul_pd(fscal,dx22);
1555 ty = _mm_mul_pd(fscal,dy22);
1556 tz = _mm_mul_pd(fscal,dz22);
1558 /* Update vectorial force */
1559 fix2 = _mm_add_pd(fix2,tx);
1560 fiy2 = _mm_add_pd(fiy2,ty);
1561 fiz2 = _mm_add_pd(fiz2,tz);
1563 fjx2 = _mm_add_pd(fjx2,tx);
1564 fjy2 = _mm_add_pd(fjy2,ty);
1565 fjz2 = _mm_add_pd(fjz2,tz);
1567 /**************************
1568 * CALCULATE INTERACTIONS *
1569 **************************/
1571 r23 = _mm_mul_pd(rsq23,rinv23);
1573 /* EWALD ELECTROSTATICS */
1575 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1576 ewrt = _mm_mul_pd(r23,ewtabscale);
1577 ewitab = _mm_cvttpd_epi32(ewrt);
1578 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1579 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1581 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1582 felec = _mm_mul_pd(_mm_mul_pd(qq23,rinv23),_mm_sub_pd(rinvsq23,felec));
1586 /* Calculate temporary vectorial force */
1587 tx = _mm_mul_pd(fscal,dx23);
1588 ty = _mm_mul_pd(fscal,dy23);
1589 tz = _mm_mul_pd(fscal,dz23);
1591 /* Update vectorial force */
1592 fix2 = _mm_add_pd(fix2,tx);
1593 fiy2 = _mm_add_pd(fiy2,ty);
1594 fiz2 = _mm_add_pd(fiz2,tz);
1596 fjx3 = _mm_add_pd(fjx3,tx);
1597 fjy3 = _mm_add_pd(fjy3,ty);
1598 fjz3 = _mm_add_pd(fjz3,tz);
1600 /**************************
1601 * CALCULATE INTERACTIONS *
1602 **************************/
1604 r31 = _mm_mul_pd(rsq31,rinv31);
1606 /* EWALD ELECTROSTATICS */
1608 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1609 ewrt = _mm_mul_pd(r31,ewtabscale);
1610 ewitab = _mm_cvttpd_epi32(ewrt);
1611 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1612 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1614 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1615 felec = _mm_mul_pd(_mm_mul_pd(qq31,rinv31),_mm_sub_pd(rinvsq31,felec));
1619 /* Calculate temporary vectorial force */
1620 tx = _mm_mul_pd(fscal,dx31);
1621 ty = _mm_mul_pd(fscal,dy31);
1622 tz = _mm_mul_pd(fscal,dz31);
1624 /* Update vectorial force */
1625 fix3 = _mm_add_pd(fix3,tx);
1626 fiy3 = _mm_add_pd(fiy3,ty);
1627 fiz3 = _mm_add_pd(fiz3,tz);
1629 fjx1 = _mm_add_pd(fjx1,tx);
1630 fjy1 = _mm_add_pd(fjy1,ty);
1631 fjz1 = _mm_add_pd(fjz1,tz);
1633 /**************************
1634 * CALCULATE INTERACTIONS *
1635 **************************/
1637 r32 = _mm_mul_pd(rsq32,rinv32);
1639 /* EWALD ELECTROSTATICS */
1641 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1642 ewrt = _mm_mul_pd(r32,ewtabscale);
1643 ewitab = _mm_cvttpd_epi32(ewrt);
1644 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1645 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1647 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1648 felec = _mm_mul_pd(_mm_mul_pd(qq32,rinv32),_mm_sub_pd(rinvsq32,felec));
1652 /* Calculate temporary vectorial force */
1653 tx = _mm_mul_pd(fscal,dx32);
1654 ty = _mm_mul_pd(fscal,dy32);
1655 tz = _mm_mul_pd(fscal,dz32);
1657 /* Update vectorial force */
1658 fix3 = _mm_add_pd(fix3,tx);
1659 fiy3 = _mm_add_pd(fiy3,ty);
1660 fiz3 = _mm_add_pd(fiz3,tz);
1662 fjx2 = _mm_add_pd(fjx2,tx);
1663 fjy2 = _mm_add_pd(fjy2,ty);
1664 fjz2 = _mm_add_pd(fjz2,tz);
1666 /**************************
1667 * CALCULATE INTERACTIONS *
1668 **************************/
1670 r33 = _mm_mul_pd(rsq33,rinv33);
1672 /* EWALD ELECTROSTATICS */
1674 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1675 ewrt = _mm_mul_pd(r33,ewtabscale);
1676 ewitab = _mm_cvttpd_epi32(ewrt);
1677 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1678 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1680 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1681 felec = _mm_mul_pd(_mm_mul_pd(qq33,rinv33),_mm_sub_pd(rinvsq33,felec));
1685 /* Calculate temporary vectorial force */
1686 tx = _mm_mul_pd(fscal,dx33);
1687 ty = _mm_mul_pd(fscal,dy33);
1688 tz = _mm_mul_pd(fscal,dz33);
1690 /* Update vectorial force */
1691 fix3 = _mm_add_pd(fix3,tx);
1692 fiy3 = _mm_add_pd(fiy3,ty);
1693 fiz3 = _mm_add_pd(fiz3,tz);
1695 fjx3 = _mm_add_pd(fjx3,tx);
1696 fjy3 = _mm_add_pd(fjy3,ty);
1697 fjz3 = _mm_add_pd(fjz3,tz);
1699 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);
1701 /* Inner loop uses 324 flops */
1704 if(jidx<j_index_end)
1708 j_coord_offsetA = DIM*jnrA;
1710 /* load j atom coordinates */
1711 gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA+DIM,
1712 &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
1714 /* Calculate displacement vector */
1715 dx11 = _mm_sub_pd(ix1,jx1);
1716 dy11 = _mm_sub_pd(iy1,jy1);
1717 dz11 = _mm_sub_pd(iz1,jz1);
1718 dx12 = _mm_sub_pd(ix1,jx2);
1719 dy12 = _mm_sub_pd(iy1,jy2);
1720 dz12 = _mm_sub_pd(iz1,jz2);
1721 dx13 = _mm_sub_pd(ix1,jx3);
1722 dy13 = _mm_sub_pd(iy1,jy3);
1723 dz13 = _mm_sub_pd(iz1,jz3);
1724 dx21 = _mm_sub_pd(ix2,jx1);
1725 dy21 = _mm_sub_pd(iy2,jy1);
1726 dz21 = _mm_sub_pd(iz2,jz1);
1727 dx22 = _mm_sub_pd(ix2,jx2);
1728 dy22 = _mm_sub_pd(iy2,jy2);
1729 dz22 = _mm_sub_pd(iz2,jz2);
1730 dx23 = _mm_sub_pd(ix2,jx3);
1731 dy23 = _mm_sub_pd(iy2,jy3);
1732 dz23 = _mm_sub_pd(iz2,jz3);
1733 dx31 = _mm_sub_pd(ix3,jx1);
1734 dy31 = _mm_sub_pd(iy3,jy1);
1735 dz31 = _mm_sub_pd(iz3,jz1);
1736 dx32 = _mm_sub_pd(ix3,jx2);
1737 dy32 = _mm_sub_pd(iy3,jy2);
1738 dz32 = _mm_sub_pd(iz3,jz2);
1739 dx33 = _mm_sub_pd(ix3,jx3);
1740 dy33 = _mm_sub_pd(iy3,jy3);
1741 dz33 = _mm_sub_pd(iz3,jz3);
1743 /* Calculate squared distance and things based on it */
1744 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1745 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1746 rsq13 = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
1747 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1748 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1749 rsq23 = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
1750 rsq31 = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
1751 rsq32 = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
1752 rsq33 = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
1754 rinv11 = gmx_mm_invsqrt_pd(rsq11);
1755 rinv12 = gmx_mm_invsqrt_pd(rsq12);
1756 rinv13 = gmx_mm_invsqrt_pd(rsq13);
1757 rinv21 = gmx_mm_invsqrt_pd(rsq21);
1758 rinv22 = gmx_mm_invsqrt_pd(rsq22);
1759 rinv23 = gmx_mm_invsqrt_pd(rsq23);
1760 rinv31 = gmx_mm_invsqrt_pd(rsq31);
1761 rinv32 = gmx_mm_invsqrt_pd(rsq32);
1762 rinv33 = gmx_mm_invsqrt_pd(rsq33);
1764 rinvsq11 = _mm_mul_pd(rinv11,rinv11);
1765 rinvsq12 = _mm_mul_pd(rinv12,rinv12);
1766 rinvsq13 = _mm_mul_pd(rinv13,rinv13);
1767 rinvsq21 = _mm_mul_pd(rinv21,rinv21);
1768 rinvsq22 = _mm_mul_pd(rinv22,rinv22);
1769 rinvsq23 = _mm_mul_pd(rinv23,rinv23);
1770 rinvsq31 = _mm_mul_pd(rinv31,rinv31);
1771 rinvsq32 = _mm_mul_pd(rinv32,rinv32);
1772 rinvsq33 = _mm_mul_pd(rinv33,rinv33);
1774 fjx1 = _mm_setzero_pd();
1775 fjy1 = _mm_setzero_pd();
1776 fjz1 = _mm_setzero_pd();
1777 fjx2 = _mm_setzero_pd();
1778 fjy2 = _mm_setzero_pd();
1779 fjz2 = _mm_setzero_pd();
1780 fjx3 = _mm_setzero_pd();
1781 fjy3 = _mm_setzero_pd();
1782 fjz3 = _mm_setzero_pd();
1784 /**************************
1785 * CALCULATE INTERACTIONS *
1786 **************************/
1788 r11 = _mm_mul_pd(rsq11,rinv11);
1790 /* EWALD ELECTROSTATICS */
1792 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1793 ewrt = _mm_mul_pd(r11,ewtabscale);
1794 ewitab = _mm_cvttpd_epi32(ewrt);
1795 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1796 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1797 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1798 felec = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
1802 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1804 /* Calculate temporary vectorial force */
1805 tx = _mm_mul_pd(fscal,dx11);
1806 ty = _mm_mul_pd(fscal,dy11);
1807 tz = _mm_mul_pd(fscal,dz11);
1809 /* Update vectorial force */
1810 fix1 = _mm_add_pd(fix1,tx);
1811 fiy1 = _mm_add_pd(fiy1,ty);
1812 fiz1 = _mm_add_pd(fiz1,tz);
1814 fjx1 = _mm_add_pd(fjx1,tx);
1815 fjy1 = _mm_add_pd(fjy1,ty);
1816 fjz1 = _mm_add_pd(fjz1,tz);
1818 /**************************
1819 * CALCULATE INTERACTIONS *
1820 **************************/
1822 r12 = _mm_mul_pd(rsq12,rinv12);
1824 /* EWALD ELECTROSTATICS */
1826 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1827 ewrt = _mm_mul_pd(r12,ewtabscale);
1828 ewitab = _mm_cvttpd_epi32(ewrt);
1829 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1830 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1831 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1832 felec = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
1836 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1838 /* Calculate temporary vectorial force */
1839 tx = _mm_mul_pd(fscal,dx12);
1840 ty = _mm_mul_pd(fscal,dy12);
1841 tz = _mm_mul_pd(fscal,dz12);
1843 /* Update vectorial force */
1844 fix1 = _mm_add_pd(fix1,tx);
1845 fiy1 = _mm_add_pd(fiy1,ty);
1846 fiz1 = _mm_add_pd(fiz1,tz);
1848 fjx2 = _mm_add_pd(fjx2,tx);
1849 fjy2 = _mm_add_pd(fjy2,ty);
1850 fjz2 = _mm_add_pd(fjz2,tz);
1852 /**************************
1853 * CALCULATE INTERACTIONS *
1854 **************************/
1856 r13 = _mm_mul_pd(rsq13,rinv13);
1858 /* EWALD ELECTROSTATICS */
1860 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1861 ewrt = _mm_mul_pd(r13,ewtabscale);
1862 ewitab = _mm_cvttpd_epi32(ewrt);
1863 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1864 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1865 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1866 felec = _mm_mul_pd(_mm_mul_pd(qq13,rinv13),_mm_sub_pd(rinvsq13,felec));
1870 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1872 /* Calculate temporary vectorial force */
1873 tx = _mm_mul_pd(fscal,dx13);
1874 ty = _mm_mul_pd(fscal,dy13);
1875 tz = _mm_mul_pd(fscal,dz13);
1877 /* Update vectorial force */
1878 fix1 = _mm_add_pd(fix1,tx);
1879 fiy1 = _mm_add_pd(fiy1,ty);
1880 fiz1 = _mm_add_pd(fiz1,tz);
1882 fjx3 = _mm_add_pd(fjx3,tx);
1883 fjy3 = _mm_add_pd(fjy3,ty);
1884 fjz3 = _mm_add_pd(fjz3,tz);
1886 /**************************
1887 * CALCULATE INTERACTIONS *
1888 **************************/
1890 r21 = _mm_mul_pd(rsq21,rinv21);
1892 /* EWALD ELECTROSTATICS */
1894 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1895 ewrt = _mm_mul_pd(r21,ewtabscale);
1896 ewitab = _mm_cvttpd_epi32(ewrt);
1897 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1898 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1899 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1900 felec = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
1904 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1906 /* Calculate temporary vectorial force */
1907 tx = _mm_mul_pd(fscal,dx21);
1908 ty = _mm_mul_pd(fscal,dy21);
1909 tz = _mm_mul_pd(fscal,dz21);
1911 /* Update vectorial force */
1912 fix2 = _mm_add_pd(fix2,tx);
1913 fiy2 = _mm_add_pd(fiy2,ty);
1914 fiz2 = _mm_add_pd(fiz2,tz);
1916 fjx1 = _mm_add_pd(fjx1,tx);
1917 fjy1 = _mm_add_pd(fjy1,ty);
1918 fjz1 = _mm_add_pd(fjz1,tz);
1920 /**************************
1921 * CALCULATE INTERACTIONS *
1922 **************************/
1924 r22 = _mm_mul_pd(rsq22,rinv22);
1926 /* EWALD ELECTROSTATICS */
1928 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1929 ewrt = _mm_mul_pd(r22,ewtabscale);
1930 ewitab = _mm_cvttpd_epi32(ewrt);
1931 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1932 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1933 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1934 felec = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
1938 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1940 /* Calculate temporary vectorial force */
1941 tx = _mm_mul_pd(fscal,dx22);
1942 ty = _mm_mul_pd(fscal,dy22);
1943 tz = _mm_mul_pd(fscal,dz22);
1945 /* Update vectorial force */
1946 fix2 = _mm_add_pd(fix2,tx);
1947 fiy2 = _mm_add_pd(fiy2,ty);
1948 fiz2 = _mm_add_pd(fiz2,tz);
1950 fjx2 = _mm_add_pd(fjx2,tx);
1951 fjy2 = _mm_add_pd(fjy2,ty);
1952 fjz2 = _mm_add_pd(fjz2,tz);
1954 /**************************
1955 * CALCULATE INTERACTIONS *
1956 **************************/
1958 r23 = _mm_mul_pd(rsq23,rinv23);
1960 /* EWALD ELECTROSTATICS */
1962 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1963 ewrt = _mm_mul_pd(r23,ewtabscale);
1964 ewitab = _mm_cvttpd_epi32(ewrt);
1965 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1966 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1967 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1968 felec = _mm_mul_pd(_mm_mul_pd(qq23,rinv23),_mm_sub_pd(rinvsq23,felec));
1972 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1974 /* Calculate temporary vectorial force */
1975 tx = _mm_mul_pd(fscal,dx23);
1976 ty = _mm_mul_pd(fscal,dy23);
1977 tz = _mm_mul_pd(fscal,dz23);
1979 /* Update vectorial force */
1980 fix2 = _mm_add_pd(fix2,tx);
1981 fiy2 = _mm_add_pd(fiy2,ty);
1982 fiz2 = _mm_add_pd(fiz2,tz);
1984 fjx3 = _mm_add_pd(fjx3,tx);
1985 fjy3 = _mm_add_pd(fjy3,ty);
1986 fjz3 = _mm_add_pd(fjz3,tz);
1988 /**************************
1989 * CALCULATE INTERACTIONS *
1990 **************************/
1992 r31 = _mm_mul_pd(rsq31,rinv31);
1994 /* EWALD ELECTROSTATICS */
1996 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1997 ewrt = _mm_mul_pd(r31,ewtabscale);
1998 ewitab = _mm_cvttpd_epi32(ewrt);
1999 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2000 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2001 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2002 felec = _mm_mul_pd(_mm_mul_pd(qq31,rinv31),_mm_sub_pd(rinvsq31,felec));
2006 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2008 /* Calculate temporary vectorial force */
2009 tx = _mm_mul_pd(fscal,dx31);
2010 ty = _mm_mul_pd(fscal,dy31);
2011 tz = _mm_mul_pd(fscal,dz31);
2013 /* Update vectorial force */
2014 fix3 = _mm_add_pd(fix3,tx);
2015 fiy3 = _mm_add_pd(fiy3,ty);
2016 fiz3 = _mm_add_pd(fiz3,tz);
2018 fjx1 = _mm_add_pd(fjx1,tx);
2019 fjy1 = _mm_add_pd(fjy1,ty);
2020 fjz1 = _mm_add_pd(fjz1,tz);
2022 /**************************
2023 * CALCULATE INTERACTIONS *
2024 **************************/
2026 r32 = _mm_mul_pd(rsq32,rinv32);
2028 /* EWALD ELECTROSTATICS */
2030 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2031 ewrt = _mm_mul_pd(r32,ewtabscale);
2032 ewitab = _mm_cvttpd_epi32(ewrt);
2033 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2034 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2035 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2036 felec = _mm_mul_pd(_mm_mul_pd(qq32,rinv32),_mm_sub_pd(rinvsq32,felec));
2040 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2042 /* Calculate temporary vectorial force */
2043 tx = _mm_mul_pd(fscal,dx32);
2044 ty = _mm_mul_pd(fscal,dy32);
2045 tz = _mm_mul_pd(fscal,dz32);
2047 /* Update vectorial force */
2048 fix3 = _mm_add_pd(fix3,tx);
2049 fiy3 = _mm_add_pd(fiy3,ty);
2050 fiz3 = _mm_add_pd(fiz3,tz);
2052 fjx2 = _mm_add_pd(fjx2,tx);
2053 fjy2 = _mm_add_pd(fjy2,ty);
2054 fjz2 = _mm_add_pd(fjz2,tz);
2056 /**************************
2057 * CALCULATE INTERACTIONS *
2058 **************************/
2060 r33 = _mm_mul_pd(rsq33,rinv33);
2062 /* EWALD ELECTROSTATICS */
2064 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2065 ewrt = _mm_mul_pd(r33,ewtabscale);
2066 ewitab = _mm_cvttpd_epi32(ewrt);
2067 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2068 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2069 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2070 felec = _mm_mul_pd(_mm_mul_pd(qq33,rinv33),_mm_sub_pd(rinvsq33,felec));
2074 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2076 /* Calculate temporary vectorial force */
2077 tx = _mm_mul_pd(fscal,dx33);
2078 ty = _mm_mul_pd(fscal,dy33);
2079 tz = _mm_mul_pd(fscal,dz33);
2081 /* Update vectorial force */
2082 fix3 = _mm_add_pd(fix3,tx);
2083 fiy3 = _mm_add_pd(fiy3,ty);
2084 fiz3 = _mm_add_pd(fiz3,tz);
2086 fjx3 = _mm_add_pd(fjx3,tx);
2087 fjy3 = _mm_add_pd(fjy3,ty);
2088 fjz3 = _mm_add_pd(fjz3,tz);
2090 gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA+DIM,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2092 /* Inner loop uses 324 flops */
2095 /* End of innermost loop */
2097 gmx_mm_update_iforce_3atom_swizzle_pd(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2098 f+i_coord_offset+DIM,fshift+i_shift_offset);
2100 /* Increment number of inner iterations */
2101 inneriter += j_index_end - j_index_start;
2103 /* Outer loop uses 18 flops */
2106 /* Increment number of outer iterations */
2109 /* Update outer/inner flops */
2111 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4W4_F,outeriter*18 + inneriter*324);