2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
5 * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
6 * and including many others, as listed in the AUTHORS file in the
7 * top-level source directory and at http://www.gromacs.org.
9 * GROMACS is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU Lesser General Public License
11 * as published by the Free Software Foundation; either version 2.1
12 * of the License, or (at your option) any later version.
14 * GROMACS is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
17 * Lesser General Public License for more details.
19 * You should have received a copy of the GNU Lesser General Public
20 * License along with GROMACS; if not, see
21 * http://www.gnu.org/licenses, or write to the Free Software Foundation,
22 * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
24 * If you want to redistribute modifications to GROMACS, please
25 * consider that scientific software is very special. Version
26 * control is crucial - bugs must be traceable. We will be happy to
27 * consider code for inclusion in the official distribution, but
28 * derived work must not be called official GROMACS. Details are found
29 * in the README & COPYING files - if they are missing, get the
30 * official version at http://www.gromacs.org.
32 * To help us fund GROMACS development, we humbly ask that you cite
33 * the research papers on the package. Check out http://www.gromacs.org.
36 * Note: this file was generated by the GROMACS sse2_double kernel generator.
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_VdwLJEwSh_GeomW3W3_VF_sse2_double
54 * Electrostatics interaction: Ewald
55 * VdW interaction: LJEwald
56 * Geometry: Water3-Water3
57 * Calculate force/pot: PotentialAndForce
60 nb_kernel_ElecEwSh_VdwLJEwSh_GeomW3W3_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 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
85 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
87 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
88 int vdwjidx0A,vdwjidx0B;
89 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
90 int vdwjidx1A,vdwjidx1B;
91 __m128d jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
92 int vdwjidx2A,vdwjidx2B;
93 __m128d jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
94 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
95 __m128d dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
96 __m128d dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
97 __m128d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
98 __m128d dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
99 __m128d dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
100 __m128d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
101 __m128d dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
102 __m128d dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
103 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
106 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
109 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
110 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
120 __m128d ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
122 __m128d one_half = _mm_set1_pd(0.5);
123 __m128d minus_one = _mm_set1_pd(-1.0);
125 __m128d ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
127 __m128d dummy_mask,cutoff_mask;
128 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
129 __m128d one = _mm_set1_pd(1.0);
130 __m128d two = _mm_set1_pd(2.0);
136 jindex = nlist->jindex;
138 shiftidx = nlist->shift;
140 shiftvec = fr->shift_vec[0];
141 fshift = fr->fshift[0];
142 facel = _mm_set1_pd(fr->epsfac);
143 charge = mdatoms->chargeA;
144 nvdwtype = fr->ntype;
146 vdwtype = mdatoms->typeA;
147 vdwgridparam = fr->ljpme_c6grid;
148 sh_lj_ewald = _mm_set1_pd(fr->ic->sh_lj_ewald);
149 ewclj = _mm_set1_pd(fr->ewaldcoeff_lj);
150 ewclj2 = _mm_mul_pd(minus_one,_mm_mul_pd(ewclj,ewclj));
152 sh_ewald = _mm_set1_pd(fr->ic->sh_ewald);
153 ewtab = fr->ic->tabq_coul_FDV0;
154 ewtabscale = _mm_set1_pd(fr->ic->tabq_scale);
155 ewtabhalfspace = _mm_set1_pd(0.5/fr->ic->tabq_scale);
157 /* Setup water-specific parameters */
158 inr = nlist->iinr[0];
159 iq0 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
160 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
161 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
162 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
164 jq0 = _mm_set1_pd(charge[inr+0]);
165 jq1 = _mm_set1_pd(charge[inr+1]);
166 jq2 = _mm_set1_pd(charge[inr+2]);
167 vdwjidx0A = 2*vdwtype[inr+0];
168 qq00 = _mm_mul_pd(iq0,jq0);
169 c6_00 = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
170 c12_00 = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
171 c6grid_00 = _mm_set1_pd(vdwgridparam[vdwioffset0+vdwjidx0A]);
172 qq01 = _mm_mul_pd(iq0,jq1);
173 qq02 = _mm_mul_pd(iq0,jq2);
174 qq10 = _mm_mul_pd(iq1,jq0);
175 qq11 = _mm_mul_pd(iq1,jq1);
176 qq12 = _mm_mul_pd(iq1,jq2);
177 qq20 = _mm_mul_pd(iq2,jq0);
178 qq21 = _mm_mul_pd(iq2,jq1);
179 qq22 = _mm_mul_pd(iq2,jq2);
181 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
182 rcutoff_scalar = fr->rcoulomb;
183 rcutoff = _mm_set1_pd(rcutoff_scalar);
184 rcutoff2 = _mm_mul_pd(rcutoff,rcutoff);
186 sh_vdw_invrcut6 = _mm_set1_pd(fr->ic->sh_invrc6);
187 rvdw = _mm_set1_pd(fr->rvdw);
189 /* Avoid stupid compiler warnings */
197 /* Start outer loop over neighborlists */
198 for(iidx=0; iidx<nri; iidx++)
200 /* Load shift vector for this list */
201 i_shift_offset = DIM*shiftidx[iidx];
203 /* Load limits for loop over neighbors */
204 j_index_start = jindex[iidx];
205 j_index_end = jindex[iidx+1];
207 /* Get outer coordinate index */
209 i_coord_offset = DIM*inr;
211 /* Load i particle coords and add shift vector */
212 gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
213 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
215 fix0 = _mm_setzero_pd();
216 fiy0 = _mm_setzero_pd();
217 fiz0 = _mm_setzero_pd();
218 fix1 = _mm_setzero_pd();
219 fiy1 = _mm_setzero_pd();
220 fiz1 = _mm_setzero_pd();
221 fix2 = _mm_setzero_pd();
222 fiy2 = _mm_setzero_pd();
223 fiz2 = _mm_setzero_pd();
225 /* Reset potential sums */
226 velecsum = _mm_setzero_pd();
227 vvdwsum = _mm_setzero_pd();
229 /* Start inner kernel loop */
230 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
233 /* Get j neighbor index, and coordinate index */
236 j_coord_offsetA = DIM*jnrA;
237 j_coord_offsetB = DIM*jnrB;
239 /* load j atom coordinates */
240 gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
241 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
243 /* Calculate displacement vector */
244 dx00 = _mm_sub_pd(ix0,jx0);
245 dy00 = _mm_sub_pd(iy0,jy0);
246 dz00 = _mm_sub_pd(iz0,jz0);
247 dx01 = _mm_sub_pd(ix0,jx1);
248 dy01 = _mm_sub_pd(iy0,jy1);
249 dz01 = _mm_sub_pd(iz0,jz1);
250 dx02 = _mm_sub_pd(ix0,jx2);
251 dy02 = _mm_sub_pd(iy0,jy2);
252 dz02 = _mm_sub_pd(iz0,jz2);
253 dx10 = _mm_sub_pd(ix1,jx0);
254 dy10 = _mm_sub_pd(iy1,jy0);
255 dz10 = _mm_sub_pd(iz1,jz0);
256 dx11 = _mm_sub_pd(ix1,jx1);
257 dy11 = _mm_sub_pd(iy1,jy1);
258 dz11 = _mm_sub_pd(iz1,jz1);
259 dx12 = _mm_sub_pd(ix1,jx2);
260 dy12 = _mm_sub_pd(iy1,jy2);
261 dz12 = _mm_sub_pd(iz1,jz2);
262 dx20 = _mm_sub_pd(ix2,jx0);
263 dy20 = _mm_sub_pd(iy2,jy0);
264 dz20 = _mm_sub_pd(iz2,jz0);
265 dx21 = _mm_sub_pd(ix2,jx1);
266 dy21 = _mm_sub_pd(iy2,jy1);
267 dz21 = _mm_sub_pd(iz2,jz1);
268 dx22 = _mm_sub_pd(ix2,jx2);
269 dy22 = _mm_sub_pd(iy2,jy2);
270 dz22 = _mm_sub_pd(iz2,jz2);
272 /* Calculate squared distance and things based on it */
273 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
274 rsq01 = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
275 rsq02 = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
276 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
277 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
278 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
279 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
280 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
281 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
283 rinv00 = gmx_mm_invsqrt_pd(rsq00);
284 rinv01 = gmx_mm_invsqrt_pd(rsq01);
285 rinv02 = gmx_mm_invsqrt_pd(rsq02);
286 rinv10 = gmx_mm_invsqrt_pd(rsq10);
287 rinv11 = gmx_mm_invsqrt_pd(rsq11);
288 rinv12 = gmx_mm_invsqrt_pd(rsq12);
289 rinv20 = gmx_mm_invsqrt_pd(rsq20);
290 rinv21 = gmx_mm_invsqrt_pd(rsq21);
291 rinv22 = gmx_mm_invsqrt_pd(rsq22);
293 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
294 rinvsq01 = _mm_mul_pd(rinv01,rinv01);
295 rinvsq02 = _mm_mul_pd(rinv02,rinv02);
296 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
297 rinvsq11 = _mm_mul_pd(rinv11,rinv11);
298 rinvsq12 = _mm_mul_pd(rinv12,rinv12);
299 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
300 rinvsq21 = _mm_mul_pd(rinv21,rinv21);
301 rinvsq22 = _mm_mul_pd(rinv22,rinv22);
303 fjx0 = _mm_setzero_pd();
304 fjy0 = _mm_setzero_pd();
305 fjz0 = _mm_setzero_pd();
306 fjx1 = _mm_setzero_pd();
307 fjy1 = _mm_setzero_pd();
308 fjz1 = _mm_setzero_pd();
309 fjx2 = _mm_setzero_pd();
310 fjy2 = _mm_setzero_pd();
311 fjz2 = _mm_setzero_pd();
313 /**************************
314 * CALCULATE INTERACTIONS *
315 **************************/
317 if (gmx_mm_any_lt(rsq00,rcutoff2))
320 r00 = _mm_mul_pd(rsq00,rinv00);
322 /* EWALD ELECTROSTATICS */
324 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
325 ewrt = _mm_mul_pd(r00,ewtabscale);
326 ewitab = _mm_cvttpd_epi32(ewrt);
327 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
328 ewitab = _mm_slli_epi32(ewitab,2);
329 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
330 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
331 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
332 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
333 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
334 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
335 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
336 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
337 velec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_sub_pd(rinv00,sh_ewald),velec));
338 felec = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
340 /* Analytical LJ-PME */
341 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
342 ewcljrsq = _mm_mul_pd(ewclj2,rsq00);
343 ewclj6 = _mm_mul_pd(ewclj2,_mm_mul_pd(ewclj2,ewclj2));
344 exponent = gmx_simd_exp_d(ewcljrsq);
345 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
346 poly = _mm_mul_pd(exponent,_mm_add_pd(_mm_sub_pd(one,ewcljrsq),_mm_mul_pd(_mm_mul_pd(ewcljrsq,ewcljrsq),one_half)));
347 /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
348 vvdw6 = _mm_mul_pd(_mm_sub_pd(c6_00,_mm_mul_pd(c6grid_00,_mm_sub_pd(one,poly))),rinvsix);
349 vvdw12 = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
350 vvdw = _mm_sub_pd(_mm_mul_pd( _mm_sub_pd(vvdw12 , _mm_mul_pd(c12_00,_mm_mul_pd(sh_vdw_invrcut6,sh_vdw_invrcut6))),one_twelfth),
351 _mm_mul_pd( _mm_sub_pd(vvdw6,_mm_add_pd(_mm_mul_pd(c6_00,sh_vdw_invrcut6),_mm_mul_pd(c6grid_00,sh_lj_ewald))),one_sixth));
352 /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
353 fvdw = _mm_mul_pd(_mm_sub_pd(vvdw12,_mm_sub_pd(vvdw6,_mm_mul_pd(_mm_mul_pd(c6grid_00,one_sixth),_mm_mul_pd(exponent,ewclj6)))),rinvsq00);
355 cutoff_mask = _mm_cmplt_pd(rsq00,rcutoff2);
357 /* Update potential sum for this i atom from the interaction with this j atom. */
358 velec = _mm_and_pd(velec,cutoff_mask);
359 velecsum = _mm_add_pd(velecsum,velec);
360 vvdw = _mm_and_pd(vvdw,cutoff_mask);
361 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
363 fscal = _mm_add_pd(felec,fvdw);
365 fscal = _mm_and_pd(fscal,cutoff_mask);
367 /* Calculate temporary vectorial force */
368 tx = _mm_mul_pd(fscal,dx00);
369 ty = _mm_mul_pd(fscal,dy00);
370 tz = _mm_mul_pd(fscal,dz00);
372 /* Update vectorial force */
373 fix0 = _mm_add_pd(fix0,tx);
374 fiy0 = _mm_add_pd(fiy0,ty);
375 fiz0 = _mm_add_pd(fiz0,tz);
377 fjx0 = _mm_add_pd(fjx0,tx);
378 fjy0 = _mm_add_pd(fjy0,ty);
379 fjz0 = _mm_add_pd(fjz0,tz);
383 /**************************
384 * CALCULATE INTERACTIONS *
385 **************************/
387 if (gmx_mm_any_lt(rsq01,rcutoff2))
390 r01 = _mm_mul_pd(rsq01,rinv01);
392 /* EWALD ELECTROSTATICS */
394 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
395 ewrt = _mm_mul_pd(r01,ewtabscale);
396 ewitab = _mm_cvttpd_epi32(ewrt);
397 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
398 ewitab = _mm_slli_epi32(ewitab,2);
399 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
400 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
401 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
402 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
403 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
404 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
405 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
406 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
407 velec = _mm_mul_pd(qq01,_mm_sub_pd(_mm_sub_pd(rinv01,sh_ewald),velec));
408 felec = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
410 cutoff_mask = _mm_cmplt_pd(rsq01,rcutoff2);
412 /* Update potential sum for this i atom from the interaction with this j atom. */
413 velec = _mm_and_pd(velec,cutoff_mask);
414 velecsum = _mm_add_pd(velecsum,velec);
418 fscal = _mm_and_pd(fscal,cutoff_mask);
420 /* Calculate temporary vectorial force */
421 tx = _mm_mul_pd(fscal,dx01);
422 ty = _mm_mul_pd(fscal,dy01);
423 tz = _mm_mul_pd(fscal,dz01);
425 /* Update vectorial force */
426 fix0 = _mm_add_pd(fix0,tx);
427 fiy0 = _mm_add_pd(fiy0,ty);
428 fiz0 = _mm_add_pd(fiz0,tz);
430 fjx1 = _mm_add_pd(fjx1,tx);
431 fjy1 = _mm_add_pd(fjy1,ty);
432 fjz1 = _mm_add_pd(fjz1,tz);
436 /**************************
437 * CALCULATE INTERACTIONS *
438 **************************/
440 if (gmx_mm_any_lt(rsq02,rcutoff2))
443 r02 = _mm_mul_pd(rsq02,rinv02);
445 /* EWALD ELECTROSTATICS */
447 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
448 ewrt = _mm_mul_pd(r02,ewtabscale);
449 ewitab = _mm_cvttpd_epi32(ewrt);
450 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
451 ewitab = _mm_slli_epi32(ewitab,2);
452 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
453 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
454 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
455 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
456 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
457 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
458 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
459 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
460 velec = _mm_mul_pd(qq02,_mm_sub_pd(_mm_sub_pd(rinv02,sh_ewald),velec));
461 felec = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
463 cutoff_mask = _mm_cmplt_pd(rsq02,rcutoff2);
465 /* Update potential sum for this i atom from the interaction with this j atom. */
466 velec = _mm_and_pd(velec,cutoff_mask);
467 velecsum = _mm_add_pd(velecsum,velec);
471 fscal = _mm_and_pd(fscal,cutoff_mask);
473 /* Calculate temporary vectorial force */
474 tx = _mm_mul_pd(fscal,dx02);
475 ty = _mm_mul_pd(fscal,dy02);
476 tz = _mm_mul_pd(fscal,dz02);
478 /* Update vectorial force */
479 fix0 = _mm_add_pd(fix0,tx);
480 fiy0 = _mm_add_pd(fiy0,ty);
481 fiz0 = _mm_add_pd(fiz0,tz);
483 fjx2 = _mm_add_pd(fjx2,tx);
484 fjy2 = _mm_add_pd(fjy2,ty);
485 fjz2 = _mm_add_pd(fjz2,tz);
489 /**************************
490 * CALCULATE INTERACTIONS *
491 **************************/
493 if (gmx_mm_any_lt(rsq10,rcutoff2))
496 r10 = _mm_mul_pd(rsq10,rinv10);
498 /* EWALD ELECTROSTATICS */
500 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
501 ewrt = _mm_mul_pd(r10,ewtabscale);
502 ewitab = _mm_cvttpd_epi32(ewrt);
503 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
504 ewitab = _mm_slli_epi32(ewitab,2);
505 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
506 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
507 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
508 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
509 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
510 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
511 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
512 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
513 velec = _mm_mul_pd(qq10,_mm_sub_pd(_mm_sub_pd(rinv10,sh_ewald),velec));
514 felec = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
516 cutoff_mask = _mm_cmplt_pd(rsq10,rcutoff2);
518 /* Update potential sum for this i atom from the interaction with this j atom. */
519 velec = _mm_and_pd(velec,cutoff_mask);
520 velecsum = _mm_add_pd(velecsum,velec);
524 fscal = _mm_and_pd(fscal,cutoff_mask);
526 /* Calculate temporary vectorial force */
527 tx = _mm_mul_pd(fscal,dx10);
528 ty = _mm_mul_pd(fscal,dy10);
529 tz = _mm_mul_pd(fscal,dz10);
531 /* Update vectorial force */
532 fix1 = _mm_add_pd(fix1,tx);
533 fiy1 = _mm_add_pd(fiy1,ty);
534 fiz1 = _mm_add_pd(fiz1,tz);
536 fjx0 = _mm_add_pd(fjx0,tx);
537 fjy0 = _mm_add_pd(fjy0,ty);
538 fjz0 = _mm_add_pd(fjz0,tz);
542 /**************************
543 * CALCULATE INTERACTIONS *
544 **************************/
546 if (gmx_mm_any_lt(rsq11,rcutoff2))
549 r11 = _mm_mul_pd(rsq11,rinv11);
551 /* EWALD ELECTROSTATICS */
553 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
554 ewrt = _mm_mul_pd(r11,ewtabscale);
555 ewitab = _mm_cvttpd_epi32(ewrt);
556 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
557 ewitab = _mm_slli_epi32(ewitab,2);
558 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
559 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
560 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
561 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
562 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
563 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
564 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
565 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
566 velec = _mm_mul_pd(qq11,_mm_sub_pd(_mm_sub_pd(rinv11,sh_ewald),velec));
567 felec = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
569 cutoff_mask = _mm_cmplt_pd(rsq11,rcutoff2);
571 /* Update potential sum for this i atom from the interaction with this j atom. */
572 velec = _mm_and_pd(velec,cutoff_mask);
573 velecsum = _mm_add_pd(velecsum,velec);
577 fscal = _mm_and_pd(fscal,cutoff_mask);
579 /* Calculate temporary vectorial force */
580 tx = _mm_mul_pd(fscal,dx11);
581 ty = _mm_mul_pd(fscal,dy11);
582 tz = _mm_mul_pd(fscal,dz11);
584 /* Update vectorial force */
585 fix1 = _mm_add_pd(fix1,tx);
586 fiy1 = _mm_add_pd(fiy1,ty);
587 fiz1 = _mm_add_pd(fiz1,tz);
589 fjx1 = _mm_add_pd(fjx1,tx);
590 fjy1 = _mm_add_pd(fjy1,ty);
591 fjz1 = _mm_add_pd(fjz1,tz);
595 /**************************
596 * CALCULATE INTERACTIONS *
597 **************************/
599 if (gmx_mm_any_lt(rsq12,rcutoff2))
602 r12 = _mm_mul_pd(rsq12,rinv12);
604 /* EWALD ELECTROSTATICS */
606 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
607 ewrt = _mm_mul_pd(r12,ewtabscale);
608 ewitab = _mm_cvttpd_epi32(ewrt);
609 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
610 ewitab = _mm_slli_epi32(ewitab,2);
611 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
612 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
613 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
614 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
615 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
616 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
617 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
618 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
619 velec = _mm_mul_pd(qq12,_mm_sub_pd(_mm_sub_pd(rinv12,sh_ewald),velec));
620 felec = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
622 cutoff_mask = _mm_cmplt_pd(rsq12,rcutoff2);
624 /* Update potential sum for this i atom from the interaction with this j atom. */
625 velec = _mm_and_pd(velec,cutoff_mask);
626 velecsum = _mm_add_pd(velecsum,velec);
630 fscal = _mm_and_pd(fscal,cutoff_mask);
632 /* Calculate temporary vectorial force */
633 tx = _mm_mul_pd(fscal,dx12);
634 ty = _mm_mul_pd(fscal,dy12);
635 tz = _mm_mul_pd(fscal,dz12);
637 /* Update vectorial force */
638 fix1 = _mm_add_pd(fix1,tx);
639 fiy1 = _mm_add_pd(fiy1,ty);
640 fiz1 = _mm_add_pd(fiz1,tz);
642 fjx2 = _mm_add_pd(fjx2,tx);
643 fjy2 = _mm_add_pd(fjy2,ty);
644 fjz2 = _mm_add_pd(fjz2,tz);
648 /**************************
649 * CALCULATE INTERACTIONS *
650 **************************/
652 if (gmx_mm_any_lt(rsq20,rcutoff2))
655 r20 = _mm_mul_pd(rsq20,rinv20);
657 /* EWALD ELECTROSTATICS */
659 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
660 ewrt = _mm_mul_pd(r20,ewtabscale);
661 ewitab = _mm_cvttpd_epi32(ewrt);
662 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
663 ewitab = _mm_slli_epi32(ewitab,2);
664 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
665 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
666 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
667 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
668 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
669 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
670 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
671 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
672 velec = _mm_mul_pd(qq20,_mm_sub_pd(_mm_sub_pd(rinv20,sh_ewald),velec));
673 felec = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
675 cutoff_mask = _mm_cmplt_pd(rsq20,rcutoff2);
677 /* Update potential sum for this i atom from the interaction with this j atom. */
678 velec = _mm_and_pd(velec,cutoff_mask);
679 velecsum = _mm_add_pd(velecsum,velec);
683 fscal = _mm_and_pd(fscal,cutoff_mask);
685 /* Calculate temporary vectorial force */
686 tx = _mm_mul_pd(fscal,dx20);
687 ty = _mm_mul_pd(fscal,dy20);
688 tz = _mm_mul_pd(fscal,dz20);
690 /* Update vectorial force */
691 fix2 = _mm_add_pd(fix2,tx);
692 fiy2 = _mm_add_pd(fiy2,ty);
693 fiz2 = _mm_add_pd(fiz2,tz);
695 fjx0 = _mm_add_pd(fjx0,tx);
696 fjy0 = _mm_add_pd(fjy0,ty);
697 fjz0 = _mm_add_pd(fjz0,tz);
701 /**************************
702 * CALCULATE INTERACTIONS *
703 **************************/
705 if (gmx_mm_any_lt(rsq21,rcutoff2))
708 r21 = _mm_mul_pd(rsq21,rinv21);
710 /* EWALD ELECTROSTATICS */
712 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
713 ewrt = _mm_mul_pd(r21,ewtabscale);
714 ewitab = _mm_cvttpd_epi32(ewrt);
715 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
716 ewitab = _mm_slli_epi32(ewitab,2);
717 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
718 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
719 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
720 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
721 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
722 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
723 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
724 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
725 velec = _mm_mul_pd(qq21,_mm_sub_pd(_mm_sub_pd(rinv21,sh_ewald),velec));
726 felec = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
728 cutoff_mask = _mm_cmplt_pd(rsq21,rcutoff2);
730 /* Update potential sum for this i atom from the interaction with this j atom. */
731 velec = _mm_and_pd(velec,cutoff_mask);
732 velecsum = _mm_add_pd(velecsum,velec);
736 fscal = _mm_and_pd(fscal,cutoff_mask);
738 /* Calculate temporary vectorial force */
739 tx = _mm_mul_pd(fscal,dx21);
740 ty = _mm_mul_pd(fscal,dy21);
741 tz = _mm_mul_pd(fscal,dz21);
743 /* Update vectorial force */
744 fix2 = _mm_add_pd(fix2,tx);
745 fiy2 = _mm_add_pd(fiy2,ty);
746 fiz2 = _mm_add_pd(fiz2,tz);
748 fjx1 = _mm_add_pd(fjx1,tx);
749 fjy1 = _mm_add_pd(fjy1,ty);
750 fjz1 = _mm_add_pd(fjz1,tz);
754 /**************************
755 * CALCULATE INTERACTIONS *
756 **************************/
758 if (gmx_mm_any_lt(rsq22,rcutoff2))
761 r22 = _mm_mul_pd(rsq22,rinv22);
763 /* EWALD ELECTROSTATICS */
765 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
766 ewrt = _mm_mul_pd(r22,ewtabscale);
767 ewitab = _mm_cvttpd_epi32(ewrt);
768 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
769 ewitab = _mm_slli_epi32(ewitab,2);
770 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
771 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
772 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
773 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
774 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
775 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
776 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
777 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
778 velec = _mm_mul_pd(qq22,_mm_sub_pd(_mm_sub_pd(rinv22,sh_ewald),velec));
779 felec = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
781 cutoff_mask = _mm_cmplt_pd(rsq22,rcutoff2);
783 /* Update potential sum for this i atom from the interaction with this j atom. */
784 velec = _mm_and_pd(velec,cutoff_mask);
785 velecsum = _mm_add_pd(velecsum,velec);
789 fscal = _mm_and_pd(fscal,cutoff_mask);
791 /* Calculate temporary vectorial force */
792 tx = _mm_mul_pd(fscal,dx22);
793 ty = _mm_mul_pd(fscal,dy22);
794 tz = _mm_mul_pd(fscal,dz22);
796 /* Update vectorial force */
797 fix2 = _mm_add_pd(fix2,tx);
798 fiy2 = _mm_add_pd(fiy2,ty);
799 fiz2 = _mm_add_pd(fiz2,tz);
801 fjx2 = _mm_add_pd(fjx2,tx);
802 fjy2 = _mm_add_pd(fjy2,ty);
803 fjz2 = _mm_add_pd(fjz2,tz);
807 gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
809 /* Inner loop uses 450 flops */
816 j_coord_offsetA = DIM*jnrA;
818 /* load j atom coordinates */
819 gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
820 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
822 /* Calculate displacement vector */
823 dx00 = _mm_sub_pd(ix0,jx0);
824 dy00 = _mm_sub_pd(iy0,jy0);
825 dz00 = _mm_sub_pd(iz0,jz0);
826 dx01 = _mm_sub_pd(ix0,jx1);
827 dy01 = _mm_sub_pd(iy0,jy1);
828 dz01 = _mm_sub_pd(iz0,jz1);
829 dx02 = _mm_sub_pd(ix0,jx2);
830 dy02 = _mm_sub_pd(iy0,jy2);
831 dz02 = _mm_sub_pd(iz0,jz2);
832 dx10 = _mm_sub_pd(ix1,jx0);
833 dy10 = _mm_sub_pd(iy1,jy0);
834 dz10 = _mm_sub_pd(iz1,jz0);
835 dx11 = _mm_sub_pd(ix1,jx1);
836 dy11 = _mm_sub_pd(iy1,jy1);
837 dz11 = _mm_sub_pd(iz1,jz1);
838 dx12 = _mm_sub_pd(ix1,jx2);
839 dy12 = _mm_sub_pd(iy1,jy2);
840 dz12 = _mm_sub_pd(iz1,jz2);
841 dx20 = _mm_sub_pd(ix2,jx0);
842 dy20 = _mm_sub_pd(iy2,jy0);
843 dz20 = _mm_sub_pd(iz2,jz0);
844 dx21 = _mm_sub_pd(ix2,jx1);
845 dy21 = _mm_sub_pd(iy2,jy1);
846 dz21 = _mm_sub_pd(iz2,jz1);
847 dx22 = _mm_sub_pd(ix2,jx2);
848 dy22 = _mm_sub_pd(iy2,jy2);
849 dz22 = _mm_sub_pd(iz2,jz2);
851 /* Calculate squared distance and things based on it */
852 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
853 rsq01 = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
854 rsq02 = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
855 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
856 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
857 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
858 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
859 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
860 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
862 rinv00 = gmx_mm_invsqrt_pd(rsq00);
863 rinv01 = gmx_mm_invsqrt_pd(rsq01);
864 rinv02 = gmx_mm_invsqrt_pd(rsq02);
865 rinv10 = gmx_mm_invsqrt_pd(rsq10);
866 rinv11 = gmx_mm_invsqrt_pd(rsq11);
867 rinv12 = gmx_mm_invsqrt_pd(rsq12);
868 rinv20 = gmx_mm_invsqrt_pd(rsq20);
869 rinv21 = gmx_mm_invsqrt_pd(rsq21);
870 rinv22 = gmx_mm_invsqrt_pd(rsq22);
872 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
873 rinvsq01 = _mm_mul_pd(rinv01,rinv01);
874 rinvsq02 = _mm_mul_pd(rinv02,rinv02);
875 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
876 rinvsq11 = _mm_mul_pd(rinv11,rinv11);
877 rinvsq12 = _mm_mul_pd(rinv12,rinv12);
878 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
879 rinvsq21 = _mm_mul_pd(rinv21,rinv21);
880 rinvsq22 = _mm_mul_pd(rinv22,rinv22);
882 fjx0 = _mm_setzero_pd();
883 fjy0 = _mm_setzero_pd();
884 fjz0 = _mm_setzero_pd();
885 fjx1 = _mm_setzero_pd();
886 fjy1 = _mm_setzero_pd();
887 fjz1 = _mm_setzero_pd();
888 fjx2 = _mm_setzero_pd();
889 fjy2 = _mm_setzero_pd();
890 fjz2 = _mm_setzero_pd();
892 /**************************
893 * CALCULATE INTERACTIONS *
894 **************************/
896 if (gmx_mm_any_lt(rsq00,rcutoff2))
899 r00 = _mm_mul_pd(rsq00,rinv00);
901 /* EWALD ELECTROSTATICS */
903 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
904 ewrt = _mm_mul_pd(r00,ewtabscale);
905 ewitab = _mm_cvttpd_epi32(ewrt);
906 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
907 ewitab = _mm_slli_epi32(ewitab,2);
908 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
909 ewtabD = _mm_setzero_pd();
910 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
911 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
912 ewtabFn = _mm_setzero_pd();
913 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
914 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
915 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
916 velec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_sub_pd(rinv00,sh_ewald),velec));
917 felec = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
919 /* Analytical LJ-PME */
920 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
921 ewcljrsq = _mm_mul_pd(ewclj2,rsq00);
922 ewclj6 = _mm_mul_pd(ewclj2,_mm_mul_pd(ewclj2,ewclj2));
923 exponent = gmx_simd_exp_d(ewcljrsq);
924 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
925 poly = _mm_mul_pd(exponent,_mm_add_pd(_mm_sub_pd(one,ewcljrsq),_mm_mul_pd(_mm_mul_pd(ewcljrsq,ewcljrsq),one_half)));
926 /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
927 vvdw6 = _mm_mul_pd(_mm_sub_pd(c6_00,_mm_mul_pd(c6grid_00,_mm_sub_pd(one,poly))),rinvsix);
928 vvdw12 = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
929 vvdw = _mm_sub_pd(_mm_mul_pd( _mm_sub_pd(vvdw12 , _mm_mul_pd(c12_00,_mm_mul_pd(sh_vdw_invrcut6,sh_vdw_invrcut6))),one_twelfth),
930 _mm_mul_pd( _mm_sub_pd(vvdw6,_mm_add_pd(_mm_mul_pd(c6_00,sh_vdw_invrcut6),_mm_mul_pd(c6grid_00,sh_lj_ewald))),one_sixth));
931 /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
932 fvdw = _mm_mul_pd(_mm_sub_pd(vvdw12,_mm_sub_pd(vvdw6,_mm_mul_pd(_mm_mul_pd(c6grid_00,one_sixth),_mm_mul_pd(exponent,ewclj6)))),rinvsq00);
934 cutoff_mask = _mm_cmplt_pd(rsq00,rcutoff2);
936 /* Update potential sum for this i atom from the interaction with this j atom. */
937 velec = _mm_and_pd(velec,cutoff_mask);
938 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
939 velecsum = _mm_add_pd(velecsum,velec);
940 vvdw = _mm_and_pd(vvdw,cutoff_mask);
941 vvdw = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
942 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
944 fscal = _mm_add_pd(felec,fvdw);
946 fscal = _mm_and_pd(fscal,cutoff_mask);
948 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
950 /* Calculate temporary vectorial force */
951 tx = _mm_mul_pd(fscal,dx00);
952 ty = _mm_mul_pd(fscal,dy00);
953 tz = _mm_mul_pd(fscal,dz00);
955 /* Update vectorial force */
956 fix0 = _mm_add_pd(fix0,tx);
957 fiy0 = _mm_add_pd(fiy0,ty);
958 fiz0 = _mm_add_pd(fiz0,tz);
960 fjx0 = _mm_add_pd(fjx0,tx);
961 fjy0 = _mm_add_pd(fjy0,ty);
962 fjz0 = _mm_add_pd(fjz0,tz);
966 /**************************
967 * CALCULATE INTERACTIONS *
968 **************************/
970 if (gmx_mm_any_lt(rsq01,rcutoff2))
973 r01 = _mm_mul_pd(rsq01,rinv01);
975 /* EWALD ELECTROSTATICS */
977 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
978 ewrt = _mm_mul_pd(r01,ewtabscale);
979 ewitab = _mm_cvttpd_epi32(ewrt);
980 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
981 ewitab = _mm_slli_epi32(ewitab,2);
982 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
983 ewtabD = _mm_setzero_pd();
984 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
985 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
986 ewtabFn = _mm_setzero_pd();
987 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
988 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
989 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
990 velec = _mm_mul_pd(qq01,_mm_sub_pd(_mm_sub_pd(rinv01,sh_ewald),velec));
991 felec = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
993 cutoff_mask = _mm_cmplt_pd(rsq01,rcutoff2);
995 /* Update potential sum for this i atom from the interaction with this j atom. */
996 velec = _mm_and_pd(velec,cutoff_mask);
997 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
998 velecsum = _mm_add_pd(velecsum,velec);
1002 fscal = _mm_and_pd(fscal,cutoff_mask);
1004 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1006 /* Calculate temporary vectorial force */
1007 tx = _mm_mul_pd(fscal,dx01);
1008 ty = _mm_mul_pd(fscal,dy01);
1009 tz = _mm_mul_pd(fscal,dz01);
1011 /* Update vectorial force */
1012 fix0 = _mm_add_pd(fix0,tx);
1013 fiy0 = _mm_add_pd(fiy0,ty);
1014 fiz0 = _mm_add_pd(fiz0,tz);
1016 fjx1 = _mm_add_pd(fjx1,tx);
1017 fjy1 = _mm_add_pd(fjy1,ty);
1018 fjz1 = _mm_add_pd(fjz1,tz);
1022 /**************************
1023 * CALCULATE INTERACTIONS *
1024 **************************/
1026 if (gmx_mm_any_lt(rsq02,rcutoff2))
1029 r02 = _mm_mul_pd(rsq02,rinv02);
1031 /* EWALD ELECTROSTATICS */
1033 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1034 ewrt = _mm_mul_pd(r02,ewtabscale);
1035 ewitab = _mm_cvttpd_epi32(ewrt);
1036 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1037 ewitab = _mm_slli_epi32(ewitab,2);
1038 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1039 ewtabD = _mm_setzero_pd();
1040 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1041 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1042 ewtabFn = _mm_setzero_pd();
1043 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1044 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1045 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1046 velec = _mm_mul_pd(qq02,_mm_sub_pd(_mm_sub_pd(rinv02,sh_ewald),velec));
1047 felec = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
1049 cutoff_mask = _mm_cmplt_pd(rsq02,rcutoff2);
1051 /* Update potential sum for this i atom from the interaction with this j atom. */
1052 velec = _mm_and_pd(velec,cutoff_mask);
1053 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1054 velecsum = _mm_add_pd(velecsum,velec);
1058 fscal = _mm_and_pd(fscal,cutoff_mask);
1060 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1062 /* Calculate temporary vectorial force */
1063 tx = _mm_mul_pd(fscal,dx02);
1064 ty = _mm_mul_pd(fscal,dy02);
1065 tz = _mm_mul_pd(fscal,dz02);
1067 /* Update vectorial force */
1068 fix0 = _mm_add_pd(fix0,tx);
1069 fiy0 = _mm_add_pd(fiy0,ty);
1070 fiz0 = _mm_add_pd(fiz0,tz);
1072 fjx2 = _mm_add_pd(fjx2,tx);
1073 fjy2 = _mm_add_pd(fjy2,ty);
1074 fjz2 = _mm_add_pd(fjz2,tz);
1078 /**************************
1079 * CALCULATE INTERACTIONS *
1080 **************************/
1082 if (gmx_mm_any_lt(rsq10,rcutoff2))
1085 r10 = _mm_mul_pd(rsq10,rinv10);
1087 /* EWALD ELECTROSTATICS */
1089 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1090 ewrt = _mm_mul_pd(r10,ewtabscale);
1091 ewitab = _mm_cvttpd_epi32(ewrt);
1092 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1093 ewitab = _mm_slli_epi32(ewitab,2);
1094 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1095 ewtabD = _mm_setzero_pd();
1096 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1097 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1098 ewtabFn = _mm_setzero_pd();
1099 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1100 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1101 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1102 velec = _mm_mul_pd(qq10,_mm_sub_pd(_mm_sub_pd(rinv10,sh_ewald),velec));
1103 felec = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
1105 cutoff_mask = _mm_cmplt_pd(rsq10,rcutoff2);
1107 /* Update potential sum for this i atom from the interaction with this j atom. */
1108 velec = _mm_and_pd(velec,cutoff_mask);
1109 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1110 velecsum = _mm_add_pd(velecsum,velec);
1114 fscal = _mm_and_pd(fscal,cutoff_mask);
1116 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1118 /* Calculate temporary vectorial force */
1119 tx = _mm_mul_pd(fscal,dx10);
1120 ty = _mm_mul_pd(fscal,dy10);
1121 tz = _mm_mul_pd(fscal,dz10);
1123 /* Update vectorial force */
1124 fix1 = _mm_add_pd(fix1,tx);
1125 fiy1 = _mm_add_pd(fiy1,ty);
1126 fiz1 = _mm_add_pd(fiz1,tz);
1128 fjx0 = _mm_add_pd(fjx0,tx);
1129 fjy0 = _mm_add_pd(fjy0,ty);
1130 fjz0 = _mm_add_pd(fjz0,tz);
1134 /**************************
1135 * CALCULATE INTERACTIONS *
1136 **************************/
1138 if (gmx_mm_any_lt(rsq11,rcutoff2))
1141 r11 = _mm_mul_pd(rsq11,rinv11);
1143 /* EWALD ELECTROSTATICS */
1145 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1146 ewrt = _mm_mul_pd(r11,ewtabscale);
1147 ewitab = _mm_cvttpd_epi32(ewrt);
1148 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1149 ewitab = _mm_slli_epi32(ewitab,2);
1150 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1151 ewtabD = _mm_setzero_pd();
1152 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1153 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1154 ewtabFn = _mm_setzero_pd();
1155 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1156 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1157 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1158 velec = _mm_mul_pd(qq11,_mm_sub_pd(_mm_sub_pd(rinv11,sh_ewald),velec));
1159 felec = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
1161 cutoff_mask = _mm_cmplt_pd(rsq11,rcutoff2);
1163 /* Update potential sum for this i atom from the interaction with this j atom. */
1164 velec = _mm_and_pd(velec,cutoff_mask);
1165 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1166 velecsum = _mm_add_pd(velecsum,velec);
1170 fscal = _mm_and_pd(fscal,cutoff_mask);
1172 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1174 /* Calculate temporary vectorial force */
1175 tx = _mm_mul_pd(fscal,dx11);
1176 ty = _mm_mul_pd(fscal,dy11);
1177 tz = _mm_mul_pd(fscal,dz11);
1179 /* Update vectorial force */
1180 fix1 = _mm_add_pd(fix1,tx);
1181 fiy1 = _mm_add_pd(fiy1,ty);
1182 fiz1 = _mm_add_pd(fiz1,tz);
1184 fjx1 = _mm_add_pd(fjx1,tx);
1185 fjy1 = _mm_add_pd(fjy1,ty);
1186 fjz1 = _mm_add_pd(fjz1,tz);
1190 /**************************
1191 * CALCULATE INTERACTIONS *
1192 **************************/
1194 if (gmx_mm_any_lt(rsq12,rcutoff2))
1197 r12 = _mm_mul_pd(rsq12,rinv12);
1199 /* EWALD ELECTROSTATICS */
1201 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1202 ewrt = _mm_mul_pd(r12,ewtabscale);
1203 ewitab = _mm_cvttpd_epi32(ewrt);
1204 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1205 ewitab = _mm_slli_epi32(ewitab,2);
1206 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1207 ewtabD = _mm_setzero_pd();
1208 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1209 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1210 ewtabFn = _mm_setzero_pd();
1211 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1212 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1213 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1214 velec = _mm_mul_pd(qq12,_mm_sub_pd(_mm_sub_pd(rinv12,sh_ewald),velec));
1215 felec = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
1217 cutoff_mask = _mm_cmplt_pd(rsq12,rcutoff2);
1219 /* Update potential sum for this i atom from the interaction with this j atom. */
1220 velec = _mm_and_pd(velec,cutoff_mask);
1221 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1222 velecsum = _mm_add_pd(velecsum,velec);
1226 fscal = _mm_and_pd(fscal,cutoff_mask);
1228 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1230 /* Calculate temporary vectorial force */
1231 tx = _mm_mul_pd(fscal,dx12);
1232 ty = _mm_mul_pd(fscal,dy12);
1233 tz = _mm_mul_pd(fscal,dz12);
1235 /* Update vectorial force */
1236 fix1 = _mm_add_pd(fix1,tx);
1237 fiy1 = _mm_add_pd(fiy1,ty);
1238 fiz1 = _mm_add_pd(fiz1,tz);
1240 fjx2 = _mm_add_pd(fjx2,tx);
1241 fjy2 = _mm_add_pd(fjy2,ty);
1242 fjz2 = _mm_add_pd(fjz2,tz);
1246 /**************************
1247 * CALCULATE INTERACTIONS *
1248 **************************/
1250 if (gmx_mm_any_lt(rsq20,rcutoff2))
1253 r20 = _mm_mul_pd(rsq20,rinv20);
1255 /* EWALD ELECTROSTATICS */
1257 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1258 ewrt = _mm_mul_pd(r20,ewtabscale);
1259 ewitab = _mm_cvttpd_epi32(ewrt);
1260 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1261 ewitab = _mm_slli_epi32(ewitab,2);
1262 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1263 ewtabD = _mm_setzero_pd();
1264 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1265 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1266 ewtabFn = _mm_setzero_pd();
1267 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1268 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1269 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1270 velec = _mm_mul_pd(qq20,_mm_sub_pd(_mm_sub_pd(rinv20,sh_ewald),velec));
1271 felec = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
1273 cutoff_mask = _mm_cmplt_pd(rsq20,rcutoff2);
1275 /* Update potential sum for this i atom from the interaction with this j atom. */
1276 velec = _mm_and_pd(velec,cutoff_mask);
1277 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1278 velecsum = _mm_add_pd(velecsum,velec);
1282 fscal = _mm_and_pd(fscal,cutoff_mask);
1284 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1286 /* Calculate temporary vectorial force */
1287 tx = _mm_mul_pd(fscal,dx20);
1288 ty = _mm_mul_pd(fscal,dy20);
1289 tz = _mm_mul_pd(fscal,dz20);
1291 /* Update vectorial force */
1292 fix2 = _mm_add_pd(fix2,tx);
1293 fiy2 = _mm_add_pd(fiy2,ty);
1294 fiz2 = _mm_add_pd(fiz2,tz);
1296 fjx0 = _mm_add_pd(fjx0,tx);
1297 fjy0 = _mm_add_pd(fjy0,ty);
1298 fjz0 = _mm_add_pd(fjz0,tz);
1302 /**************************
1303 * CALCULATE INTERACTIONS *
1304 **************************/
1306 if (gmx_mm_any_lt(rsq21,rcutoff2))
1309 r21 = _mm_mul_pd(rsq21,rinv21);
1311 /* EWALD ELECTROSTATICS */
1313 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1314 ewrt = _mm_mul_pd(r21,ewtabscale);
1315 ewitab = _mm_cvttpd_epi32(ewrt);
1316 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1317 ewitab = _mm_slli_epi32(ewitab,2);
1318 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1319 ewtabD = _mm_setzero_pd();
1320 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1321 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1322 ewtabFn = _mm_setzero_pd();
1323 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1324 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1325 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1326 velec = _mm_mul_pd(qq21,_mm_sub_pd(_mm_sub_pd(rinv21,sh_ewald),velec));
1327 felec = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
1329 cutoff_mask = _mm_cmplt_pd(rsq21,rcutoff2);
1331 /* Update potential sum for this i atom from the interaction with this j atom. */
1332 velec = _mm_and_pd(velec,cutoff_mask);
1333 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1334 velecsum = _mm_add_pd(velecsum,velec);
1338 fscal = _mm_and_pd(fscal,cutoff_mask);
1340 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1342 /* Calculate temporary vectorial force */
1343 tx = _mm_mul_pd(fscal,dx21);
1344 ty = _mm_mul_pd(fscal,dy21);
1345 tz = _mm_mul_pd(fscal,dz21);
1347 /* Update vectorial force */
1348 fix2 = _mm_add_pd(fix2,tx);
1349 fiy2 = _mm_add_pd(fiy2,ty);
1350 fiz2 = _mm_add_pd(fiz2,tz);
1352 fjx1 = _mm_add_pd(fjx1,tx);
1353 fjy1 = _mm_add_pd(fjy1,ty);
1354 fjz1 = _mm_add_pd(fjz1,tz);
1358 /**************************
1359 * CALCULATE INTERACTIONS *
1360 **************************/
1362 if (gmx_mm_any_lt(rsq22,rcutoff2))
1365 r22 = _mm_mul_pd(rsq22,rinv22);
1367 /* EWALD ELECTROSTATICS */
1369 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1370 ewrt = _mm_mul_pd(r22,ewtabscale);
1371 ewitab = _mm_cvttpd_epi32(ewrt);
1372 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1373 ewitab = _mm_slli_epi32(ewitab,2);
1374 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1375 ewtabD = _mm_setzero_pd();
1376 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1377 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1378 ewtabFn = _mm_setzero_pd();
1379 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1380 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1381 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1382 velec = _mm_mul_pd(qq22,_mm_sub_pd(_mm_sub_pd(rinv22,sh_ewald),velec));
1383 felec = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
1385 cutoff_mask = _mm_cmplt_pd(rsq22,rcutoff2);
1387 /* Update potential sum for this i atom from the interaction with this j atom. */
1388 velec = _mm_and_pd(velec,cutoff_mask);
1389 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1390 velecsum = _mm_add_pd(velecsum,velec);
1394 fscal = _mm_and_pd(fscal,cutoff_mask);
1396 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1398 /* Calculate temporary vectorial force */
1399 tx = _mm_mul_pd(fscal,dx22);
1400 ty = _mm_mul_pd(fscal,dy22);
1401 tz = _mm_mul_pd(fscal,dz22);
1403 /* Update vectorial force */
1404 fix2 = _mm_add_pd(fix2,tx);
1405 fiy2 = _mm_add_pd(fiy2,ty);
1406 fiz2 = _mm_add_pd(fiz2,tz);
1408 fjx2 = _mm_add_pd(fjx2,tx);
1409 fjy2 = _mm_add_pd(fjy2,ty);
1410 fjz2 = _mm_add_pd(fjz2,tz);
1414 gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1416 /* Inner loop uses 450 flops */
1419 /* End of innermost loop */
1421 gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1422 f+i_coord_offset,fshift+i_shift_offset);
1425 /* Update potential energies */
1426 gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1427 gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1429 /* Increment number of inner iterations */
1430 inneriter += j_index_end - j_index_start;
1432 /* Outer loop uses 20 flops */
1435 /* Increment number of outer iterations */
1438 /* Update outer/inner flops */
1440 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_VF,outeriter*20 + inneriter*450);
1443 * Gromacs nonbonded kernel: nb_kernel_ElecEwSh_VdwLJEwSh_GeomW3W3_F_sse2_double
1444 * Electrostatics interaction: Ewald
1445 * VdW interaction: LJEwald
1446 * Geometry: Water3-Water3
1447 * Calculate force/pot: Force
1450 nb_kernel_ElecEwSh_VdwLJEwSh_GeomW3W3_F_sse2_double
1451 (t_nblist * gmx_restrict nlist,
1452 rvec * gmx_restrict xx,
1453 rvec * gmx_restrict ff,
1454 t_forcerec * gmx_restrict fr,
1455 t_mdatoms * gmx_restrict mdatoms,
1456 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1457 t_nrnb * gmx_restrict nrnb)
1459 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1460 * just 0 for non-waters.
1461 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
1462 * jnr indices corresponding to data put in the four positions in the SIMD register.
1464 int i_shift_offset,i_coord_offset,outeriter,inneriter;
1465 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1467 int j_coord_offsetA,j_coord_offsetB;
1468 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
1469 real rcutoff_scalar;
1470 real *shiftvec,*fshift,*x,*f;
1471 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1473 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1475 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1477 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1478 int vdwjidx0A,vdwjidx0B;
1479 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1480 int vdwjidx1A,vdwjidx1B;
1481 __m128d jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1482 int vdwjidx2A,vdwjidx2B;
1483 __m128d jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1484 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1485 __m128d dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1486 __m128d dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1487 __m128d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1488 __m128d dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1489 __m128d dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1490 __m128d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1491 __m128d dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1492 __m128d dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1493 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
1496 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1499 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
1500 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
1510 __m128d ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
1512 __m128d one_half = _mm_set1_pd(0.5);
1513 __m128d minus_one = _mm_set1_pd(-1.0);
1515 __m128d ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1517 __m128d dummy_mask,cutoff_mask;
1518 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
1519 __m128d one = _mm_set1_pd(1.0);
1520 __m128d two = _mm_set1_pd(2.0);
1526 jindex = nlist->jindex;
1528 shiftidx = nlist->shift;
1530 shiftvec = fr->shift_vec[0];
1531 fshift = fr->fshift[0];
1532 facel = _mm_set1_pd(fr->epsfac);
1533 charge = mdatoms->chargeA;
1534 nvdwtype = fr->ntype;
1535 vdwparam = fr->nbfp;
1536 vdwtype = mdatoms->typeA;
1537 vdwgridparam = fr->ljpme_c6grid;
1538 sh_lj_ewald = _mm_set1_pd(fr->ic->sh_lj_ewald);
1539 ewclj = _mm_set1_pd(fr->ewaldcoeff_lj);
1540 ewclj2 = _mm_mul_pd(minus_one,_mm_mul_pd(ewclj,ewclj));
1542 sh_ewald = _mm_set1_pd(fr->ic->sh_ewald);
1543 ewtab = fr->ic->tabq_coul_F;
1544 ewtabscale = _mm_set1_pd(fr->ic->tabq_scale);
1545 ewtabhalfspace = _mm_set1_pd(0.5/fr->ic->tabq_scale);
1547 /* Setup water-specific parameters */
1548 inr = nlist->iinr[0];
1549 iq0 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
1550 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
1551 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
1552 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
1554 jq0 = _mm_set1_pd(charge[inr+0]);
1555 jq1 = _mm_set1_pd(charge[inr+1]);
1556 jq2 = _mm_set1_pd(charge[inr+2]);
1557 vdwjidx0A = 2*vdwtype[inr+0];
1558 qq00 = _mm_mul_pd(iq0,jq0);
1559 c6_00 = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
1560 c12_00 = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
1561 c6grid_00 = _mm_set1_pd(vdwgridparam[vdwioffset0+vdwjidx0A]);
1562 qq01 = _mm_mul_pd(iq0,jq1);
1563 qq02 = _mm_mul_pd(iq0,jq2);
1564 qq10 = _mm_mul_pd(iq1,jq0);
1565 qq11 = _mm_mul_pd(iq1,jq1);
1566 qq12 = _mm_mul_pd(iq1,jq2);
1567 qq20 = _mm_mul_pd(iq2,jq0);
1568 qq21 = _mm_mul_pd(iq2,jq1);
1569 qq22 = _mm_mul_pd(iq2,jq2);
1571 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
1572 rcutoff_scalar = fr->rcoulomb;
1573 rcutoff = _mm_set1_pd(rcutoff_scalar);
1574 rcutoff2 = _mm_mul_pd(rcutoff,rcutoff);
1576 sh_vdw_invrcut6 = _mm_set1_pd(fr->ic->sh_invrc6);
1577 rvdw = _mm_set1_pd(fr->rvdw);
1579 /* Avoid stupid compiler warnings */
1581 j_coord_offsetA = 0;
1582 j_coord_offsetB = 0;
1587 /* Start outer loop over neighborlists */
1588 for(iidx=0; iidx<nri; iidx++)
1590 /* Load shift vector for this list */
1591 i_shift_offset = DIM*shiftidx[iidx];
1593 /* Load limits for loop over neighbors */
1594 j_index_start = jindex[iidx];
1595 j_index_end = jindex[iidx+1];
1597 /* Get outer coordinate index */
1599 i_coord_offset = DIM*inr;
1601 /* Load i particle coords and add shift vector */
1602 gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1603 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1605 fix0 = _mm_setzero_pd();
1606 fiy0 = _mm_setzero_pd();
1607 fiz0 = _mm_setzero_pd();
1608 fix1 = _mm_setzero_pd();
1609 fiy1 = _mm_setzero_pd();
1610 fiz1 = _mm_setzero_pd();
1611 fix2 = _mm_setzero_pd();
1612 fiy2 = _mm_setzero_pd();
1613 fiz2 = _mm_setzero_pd();
1615 /* Start inner kernel loop */
1616 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
1619 /* Get j neighbor index, and coordinate index */
1621 jnrB = jjnr[jidx+1];
1622 j_coord_offsetA = DIM*jnrA;
1623 j_coord_offsetB = DIM*jnrB;
1625 /* load j atom coordinates */
1626 gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1627 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1629 /* Calculate displacement vector */
1630 dx00 = _mm_sub_pd(ix0,jx0);
1631 dy00 = _mm_sub_pd(iy0,jy0);
1632 dz00 = _mm_sub_pd(iz0,jz0);
1633 dx01 = _mm_sub_pd(ix0,jx1);
1634 dy01 = _mm_sub_pd(iy0,jy1);
1635 dz01 = _mm_sub_pd(iz0,jz1);
1636 dx02 = _mm_sub_pd(ix0,jx2);
1637 dy02 = _mm_sub_pd(iy0,jy2);
1638 dz02 = _mm_sub_pd(iz0,jz2);
1639 dx10 = _mm_sub_pd(ix1,jx0);
1640 dy10 = _mm_sub_pd(iy1,jy0);
1641 dz10 = _mm_sub_pd(iz1,jz0);
1642 dx11 = _mm_sub_pd(ix1,jx1);
1643 dy11 = _mm_sub_pd(iy1,jy1);
1644 dz11 = _mm_sub_pd(iz1,jz1);
1645 dx12 = _mm_sub_pd(ix1,jx2);
1646 dy12 = _mm_sub_pd(iy1,jy2);
1647 dz12 = _mm_sub_pd(iz1,jz2);
1648 dx20 = _mm_sub_pd(ix2,jx0);
1649 dy20 = _mm_sub_pd(iy2,jy0);
1650 dz20 = _mm_sub_pd(iz2,jz0);
1651 dx21 = _mm_sub_pd(ix2,jx1);
1652 dy21 = _mm_sub_pd(iy2,jy1);
1653 dz21 = _mm_sub_pd(iz2,jz1);
1654 dx22 = _mm_sub_pd(ix2,jx2);
1655 dy22 = _mm_sub_pd(iy2,jy2);
1656 dz22 = _mm_sub_pd(iz2,jz2);
1658 /* Calculate squared distance and things based on it */
1659 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1660 rsq01 = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
1661 rsq02 = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
1662 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
1663 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1664 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1665 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
1666 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1667 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1669 rinv00 = gmx_mm_invsqrt_pd(rsq00);
1670 rinv01 = gmx_mm_invsqrt_pd(rsq01);
1671 rinv02 = gmx_mm_invsqrt_pd(rsq02);
1672 rinv10 = gmx_mm_invsqrt_pd(rsq10);
1673 rinv11 = gmx_mm_invsqrt_pd(rsq11);
1674 rinv12 = gmx_mm_invsqrt_pd(rsq12);
1675 rinv20 = gmx_mm_invsqrt_pd(rsq20);
1676 rinv21 = gmx_mm_invsqrt_pd(rsq21);
1677 rinv22 = gmx_mm_invsqrt_pd(rsq22);
1679 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
1680 rinvsq01 = _mm_mul_pd(rinv01,rinv01);
1681 rinvsq02 = _mm_mul_pd(rinv02,rinv02);
1682 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
1683 rinvsq11 = _mm_mul_pd(rinv11,rinv11);
1684 rinvsq12 = _mm_mul_pd(rinv12,rinv12);
1685 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
1686 rinvsq21 = _mm_mul_pd(rinv21,rinv21);
1687 rinvsq22 = _mm_mul_pd(rinv22,rinv22);
1689 fjx0 = _mm_setzero_pd();
1690 fjy0 = _mm_setzero_pd();
1691 fjz0 = _mm_setzero_pd();
1692 fjx1 = _mm_setzero_pd();
1693 fjy1 = _mm_setzero_pd();
1694 fjz1 = _mm_setzero_pd();
1695 fjx2 = _mm_setzero_pd();
1696 fjy2 = _mm_setzero_pd();
1697 fjz2 = _mm_setzero_pd();
1699 /**************************
1700 * CALCULATE INTERACTIONS *
1701 **************************/
1703 if (gmx_mm_any_lt(rsq00,rcutoff2))
1706 r00 = _mm_mul_pd(rsq00,rinv00);
1708 /* EWALD ELECTROSTATICS */
1710 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1711 ewrt = _mm_mul_pd(r00,ewtabscale);
1712 ewitab = _mm_cvttpd_epi32(ewrt);
1713 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1714 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1716 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1717 felec = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
1719 /* Analytical LJ-PME */
1720 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1721 ewcljrsq = _mm_mul_pd(ewclj2,rsq00);
1722 ewclj6 = _mm_mul_pd(ewclj2,_mm_mul_pd(ewclj2,ewclj2));
1723 exponent = gmx_simd_exp_d(ewcljrsq);
1724 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
1725 poly = _mm_mul_pd(exponent,_mm_add_pd(_mm_sub_pd(one,ewcljrsq),_mm_mul_pd(_mm_mul_pd(ewcljrsq,ewcljrsq),one_half)));
1726 /* f6A = 6 * C6grid * (1 - poly) */
1727 f6A = _mm_mul_pd(c6grid_00,_mm_sub_pd(one,poly));
1728 /* f6B = C6grid * exponent * beta^6 */
1729 f6B = _mm_mul_pd(_mm_mul_pd(c6grid_00,one_sixth),_mm_mul_pd(exponent,ewclj6));
1730 /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
1731 fvdw = _mm_mul_pd(_mm_add_pd(_mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),_mm_sub_pd(c6_00,f6A)),rinvsix),f6B),rinvsq00);
1733 cutoff_mask = _mm_cmplt_pd(rsq00,rcutoff2);
1735 fscal = _mm_add_pd(felec,fvdw);
1737 fscal = _mm_and_pd(fscal,cutoff_mask);
1739 /* Calculate temporary vectorial force */
1740 tx = _mm_mul_pd(fscal,dx00);
1741 ty = _mm_mul_pd(fscal,dy00);
1742 tz = _mm_mul_pd(fscal,dz00);
1744 /* Update vectorial force */
1745 fix0 = _mm_add_pd(fix0,tx);
1746 fiy0 = _mm_add_pd(fiy0,ty);
1747 fiz0 = _mm_add_pd(fiz0,tz);
1749 fjx0 = _mm_add_pd(fjx0,tx);
1750 fjy0 = _mm_add_pd(fjy0,ty);
1751 fjz0 = _mm_add_pd(fjz0,tz);
1755 /**************************
1756 * CALCULATE INTERACTIONS *
1757 **************************/
1759 if (gmx_mm_any_lt(rsq01,rcutoff2))
1762 r01 = _mm_mul_pd(rsq01,rinv01);
1764 /* EWALD ELECTROSTATICS */
1766 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1767 ewrt = _mm_mul_pd(r01,ewtabscale);
1768 ewitab = _mm_cvttpd_epi32(ewrt);
1769 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1770 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1772 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1773 felec = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
1775 cutoff_mask = _mm_cmplt_pd(rsq01,rcutoff2);
1779 fscal = _mm_and_pd(fscal,cutoff_mask);
1781 /* Calculate temporary vectorial force */
1782 tx = _mm_mul_pd(fscal,dx01);
1783 ty = _mm_mul_pd(fscal,dy01);
1784 tz = _mm_mul_pd(fscal,dz01);
1786 /* Update vectorial force */
1787 fix0 = _mm_add_pd(fix0,tx);
1788 fiy0 = _mm_add_pd(fiy0,ty);
1789 fiz0 = _mm_add_pd(fiz0,tz);
1791 fjx1 = _mm_add_pd(fjx1,tx);
1792 fjy1 = _mm_add_pd(fjy1,ty);
1793 fjz1 = _mm_add_pd(fjz1,tz);
1797 /**************************
1798 * CALCULATE INTERACTIONS *
1799 **************************/
1801 if (gmx_mm_any_lt(rsq02,rcutoff2))
1804 r02 = _mm_mul_pd(rsq02,rinv02);
1806 /* EWALD ELECTROSTATICS */
1808 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1809 ewrt = _mm_mul_pd(r02,ewtabscale);
1810 ewitab = _mm_cvttpd_epi32(ewrt);
1811 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1812 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1814 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1815 felec = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
1817 cutoff_mask = _mm_cmplt_pd(rsq02,rcutoff2);
1821 fscal = _mm_and_pd(fscal,cutoff_mask);
1823 /* Calculate temporary vectorial force */
1824 tx = _mm_mul_pd(fscal,dx02);
1825 ty = _mm_mul_pd(fscal,dy02);
1826 tz = _mm_mul_pd(fscal,dz02);
1828 /* Update vectorial force */
1829 fix0 = _mm_add_pd(fix0,tx);
1830 fiy0 = _mm_add_pd(fiy0,ty);
1831 fiz0 = _mm_add_pd(fiz0,tz);
1833 fjx2 = _mm_add_pd(fjx2,tx);
1834 fjy2 = _mm_add_pd(fjy2,ty);
1835 fjz2 = _mm_add_pd(fjz2,tz);
1839 /**************************
1840 * CALCULATE INTERACTIONS *
1841 **************************/
1843 if (gmx_mm_any_lt(rsq10,rcutoff2))
1846 r10 = _mm_mul_pd(rsq10,rinv10);
1848 /* EWALD ELECTROSTATICS */
1850 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1851 ewrt = _mm_mul_pd(r10,ewtabscale);
1852 ewitab = _mm_cvttpd_epi32(ewrt);
1853 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1854 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1856 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1857 felec = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
1859 cutoff_mask = _mm_cmplt_pd(rsq10,rcutoff2);
1863 fscal = _mm_and_pd(fscal,cutoff_mask);
1865 /* Calculate temporary vectorial force */
1866 tx = _mm_mul_pd(fscal,dx10);
1867 ty = _mm_mul_pd(fscal,dy10);
1868 tz = _mm_mul_pd(fscal,dz10);
1870 /* Update vectorial force */
1871 fix1 = _mm_add_pd(fix1,tx);
1872 fiy1 = _mm_add_pd(fiy1,ty);
1873 fiz1 = _mm_add_pd(fiz1,tz);
1875 fjx0 = _mm_add_pd(fjx0,tx);
1876 fjy0 = _mm_add_pd(fjy0,ty);
1877 fjz0 = _mm_add_pd(fjz0,tz);
1881 /**************************
1882 * CALCULATE INTERACTIONS *
1883 **************************/
1885 if (gmx_mm_any_lt(rsq11,rcutoff2))
1888 r11 = _mm_mul_pd(rsq11,rinv11);
1890 /* EWALD ELECTROSTATICS */
1892 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1893 ewrt = _mm_mul_pd(r11,ewtabscale);
1894 ewitab = _mm_cvttpd_epi32(ewrt);
1895 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1896 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1898 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1899 felec = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
1901 cutoff_mask = _mm_cmplt_pd(rsq11,rcutoff2);
1905 fscal = _mm_and_pd(fscal,cutoff_mask);
1907 /* Calculate temporary vectorial force */
1908 tx = _mm_mul_pd(fscal,dx11);
1909 ty = _mm_mul_pd(fscal,dy11);
1910 tz = _mm_mul_pd(fscal,dz11);
1912 /* Update vectorial force */
1913 fix1 = _mm_add_pd(fix1,tx);
1914 fiy1 = _mm_add_pd(fiy1,ty);
1915 fiz1 = _mm_add_pd(fiz1,tz);
1917 fjx1 = _mm_add_pd(fjx1,tx);
1918 fjy1 = _mm_add_pd(fjy1,ty);
1919 fjz1 = _mm_add_pd(fjz1,tz);
1923 /**************************
1924 * CALCULATE INTERACTIONS *
1925 **************************/
1927 if (gmx_mm_any_lt(rsq12,rcutoff2))
1930 r12 = _mm_mul_pd(rsq12,rinv12);
1932 /* EWALD ELECTROSTATICS */
1934 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1935 ewrt = _mm_mul_pd(r12,ewtabscale);
1936 ewitab = _mm_cvttpd_epi32(ewrt);
1937 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1938 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1940 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1941 felec = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
1943 cutoff_mask = _mm_cmplt_pd(rsq12,rcutoff2);
1947 fscal = _mm_and_pd(fscal,cutoff_mask);
1949 /* Calculate temporary vectorial force */
1950 tx = _mm_mul_pd(fscal,dx12);
1951 ty = _mm_mul_pd(fscal,dy12);
1952 tz = _mm_mul_pd(fscal,dz12);
1954 /* Update vectorial force */
1955 fix1 = _mm_add_pd(fix1,tx);
1956 fiy1 = _mm_add_pd(fiy1,ty);
1957 fiz1 = _mm_add_pd(fiz1,tz);
1959 fjx2 = _mm_add_pd(fjx2,tx);
1960 fjy2 = _mm_add_pd(fjy2,ty);
1961 fjz2 = _mm_add_pd(fjz2,tz);
1965 /**************************
1966 * CALCULATE INTERACTIONS *
1967 **************************/
1969 if (gmx_mm_any_lt(rsq20,rcutoff2))
1972 r20 = _mm_mul_pd(rsq20,rinv20);
1974 /* EWALD ELECTROSTATICS */
1976 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1977 ewrt = _mm_mul_pd(r20,ewtabscale);
1978 ewitab = _mm_cvttpd_epi32(ewrt);
1979 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1980 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1982 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1983 felec = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
1985 cutoff_mask = _mm_cmplt_pd(rsq20,rcutoff2);
1989 fscal = _mm_and_pd(fscal,cutoff_mask);
1991 /* Calculate temporary vectorial force */
1992 tx = _mm_mul_pd(fscal,dx20);
1993 ty = _mm_mul_pd(fscal,dy20);
1994 tz = _mm_mul_pd(fscal,dz20);
1996 /* Update vectorial force */
1997 fix2 = _mm_add_pd(fix2,tx);
1998 fiy2 = _mm_add_pd(fiy2,ty);
1999 fiz2 = _mm_add_pd(fiz2,tz);
2001 fjx0 = _mm_add_pd(fjx0,tx);
2002 fjy0 = _mm_add_pd(fjy0,ty);
2003 fjz0 = _mm_add_pd(fjz0,tz);
2007 /**************************
2008 * CALCULATE INTERACTIONS *
2009 **************************/
2011 if (gmx_mm_any_lt(rsq21,rcutoff2))
2014 r21 = _mm_mul_pd(rsq21,rinv21);
2016 /* EWALD ELECTROSTATICS */
2018 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2019 ewrt = _mm_mul_pd(r21,ewtabscale);
2020 ewitab = _mm_cvttpd_epi32(ewrt);
2021 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2022 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
2024 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2025 felec = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
2027 cutoff_mask = _mm_cmplt_pd(rsq21,rcutoff2);
2031 fscal = _mm_and_pd(fscal,cutoff_mask);
2033 /* Calculate temporary vectorial force */
2034 tx = _mm_mul_pd(fscal,dx21);
2035 ty = _mm_mul_pd(fscal,dy21);
2036 tz = _mm_mul_pd(fscal,dz21);
2038 /* Update vectorial force */
2039 fix2 = _mm_add_pd(fix2,tx);
2040 fiy2 = _mm_add_pd(fiy2,ty);
2041 fiz2 = _mm_add_pd(fiz2,tz);
2043 fjx1 = _mm_add_pd(fjx1,tx);
2044 fjy1 = _mm_add_pd(fjy1,ty);
2045 fjz1 = _mm_add_pd(fjz1,tz);
2049 /**************************
2050 * CALCULATE INTERACTIONS *
2051 **************************/
2053 if (gmx_mm_any_lt(rsq22,rcutoff2))
2056 r22 = _mm_mul_pd(rsq22,rinv22);
2058 /* EWALD ELECTROSTATICS */
2060 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2061 ewrt = _mm_mul_pd(r22,ewtabscale);
2062 ewitab = _mm_cvttpd_epi32(ewrt);
2063 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2064 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
2066 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2067 felec = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
2069 cutoff_mask = _mm_cmplt_pd(rsq22,rcutoff2);
2073 fscal = _mm_and_pd(fscal,cutoff_mask);
2075 /* Calculate temporary vectorial force */
2076 tx = _mm_mul_pd(fscal,dx22);
2077 ty = _mm_mul_pd(fscal,dy22);
2078 tz = _mm_mul_pd(fscal,dz22);
2080 /* Update vectorial force */
2081 fix2 = _mm_add_pd(fix2,tx);
2082 fiy2 = _mm_add_pd(fiy2,ty);
2083 fiz2 = _mm_add_pd(fiz2,tz);
2085 fjx2 = _mm_add_pd(fjx2,tx);
2086 fjy2 = _mm_add_pd(fjy2,ty);
2087 fjz2 = _mm_add_pd(fjz2,tz);
2091 gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
2093 /* Inner loop uses 374 flops */
2096 if(jidx<j_index_end)
2100 j_coord_offsetA = DIM*jnrA;
2102 /* load j atom coordinates */
2103 gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
2104 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
2106 /* Calculate displacement vector */
2107 dx00 = _mm_sub_pd(ix0,jx0);
2108 dy00 = _mm_sub_pd(iy0,jy0);
2109 dz00 = _mm_sub_pd(iz0,jz0);
2110 dx01 = _mm_sub_pd(ix0,jx1);
2111 dy01 = _mm_sub_pd(iy0,jy1);
2112 dz01 = _mm_sub_pd(iz0,jz1);
2113 dx02 = _mm_sub_pd(ix0,jx2);
2114 dy02 = _mm_sub_pd(iy0,jy2);
2115 dz02 = _mm_sub_pd(iz0,jz2);
2116 dx10 = _mm_sub_pd(ix1,jx0);
2117 dy10 = _mm_sub_pd(iy1,jy0);
2118 dz10 = _mm_sub_pd(iz1,jz0);
2119 dx11 = _mm_sub_pd(ix1,jx1);
2120 dy11 = _mm_sub_pd(iy1,jy1);
2121 dz11 = _mm_sub_pd(iz1,jz1);
2122 dx12 = _mm_sub_pd(ix1,jx2);
2123 dy12 = _mm_sub_pd(iy1,jy2);
2124 dz12 = _mm_sub_pd(iz1,jz2);
2125 dx20 = _mm_sub_pd(ix2,jx0);
2126 dy20 = _mm_sub_pd(iy2,jy0);
2127 dz20 = _mm_sub_pd(iz2,jz0);
2128 dx21 = _mm_sub_pd(ix2,jx1);
2129 dy21 = _mm_sub_pd(iy2,jy1);
2130 dz21 = _mm_sub_pd(iz2,jz1);
2131 dx22 = _mm_sub_pd(ix2,jx2);
2132 dy22 = _mm_sub_pd(iy2,jy2);
2133 dz22 = _mm_sub_pd(iz2,jz2);
2135 /* Calculate squared distance and things based on it */
2136 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
2137 rsq01 = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
2138 rsq02 = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
2139 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
2140 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
2141 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
2142 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
2143 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
2144 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
2146 rinv00 = gmx_mm_invsqrt_pd(rsq00);
2147 rinv01 = gmx_mm_invsqrt_pd(rsq01);
2148 rinv02 = gmx_mm_invsqrt_pd(rsq02);
2149 rinv10 = gmx_mm_invsqrt_pd(rsq10);
2150 rinv11 = gmx_mm_invsqrt_pd(rsq11);
2151 rinv12 = gmx_mm_invsqrt_pd(rsq12);
2152 rinv20 = gmx_mm_invsqrt_pd(rsq20);
2153 rinv21 = gmx_mm_invsqrt_pd(rsq21);
2154 rinv22 = gmx_mm_invsqrt_pd(rsq22);
2156 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
2157 rinvsq01 = _mm_mul_pd(rinv01,rinv01);
2158 rinvsq02 = _mm_mul_pd(rinv02,rinv02);
2159 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
2160 rinvsq11 = _mm_mul_pd(rinv11,rinv11);
2161 rinvsq12 = _mm_mul_pd(rinv12,rinv12);
2162 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
2163 rinvsq21 = _mm_mul_pd(rinv21,rinv21);
2164 rinvsq22 = _mm_mul_pd(rinv22,rinv22);
2166 fjx0 = _mm_setzero_pd();
2167 fjy0 = _mm_setzero_pd();
2168 fjz0 = _mm_setzero_pd();
2169 fjx1 = _mm_setzero_pd();
2170 fjy1 = _mm_setzero_pd();
2171 fjz1 = _mm_setzero_pd();
2172 fjx2 = _mm_setzero_pd();
2173 fjy2 = _mm_setzero_pd();
2174 fjz2 = _mm_setzero_pd();
2176 /**************************
2177 * CALCULATE INTERACTIONS *
2178 **************************/
2180 if (gmx_mm_any_lt(rsq00,rcutoff2))
2183 r00 = _mm_mul_pd(rsq00,rinv00);
2185 /* EWALD ELECTROSTATICS */
2187 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2188 ewrt = _mm_mul_pd(r00,ewtabscale);
2189 ewitab = _mm_cvttpd_epi32(ewrt);
2190 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2191 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2192 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2193 felec = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
2195 /* Analytical LJ-PME */
2196 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
2197 ewcljrsq = _mm_mul_pd(ewclj2,rsq00);
2198 ewclj6 = _mm_mul_pd(ewclj2,_mm_mul_pd(ewclj2,ewclj2));
2199 exponent = gmx_simd_exp_d(ewcljrsq);
2200 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
2201 poly = _mm_mul_pd(exponent,_mm_add_pd(_mm_sub_pd(one,ewcljrsq),_mm_mul_pd(_mm_mul_pd(ewcljrsq,ewcljrsq),one_half)));
2202 /* f6A = 6 * C6grid * (1 - poly) */
2203 f6A = _mm_mul_pd(c6grid_00,_mm_sub_pd(one,poly));
2204 /* f6B = C6grid * exponent * beta^6 */
2205 f6B = _mm_mul_pd(_mm_mul_pd(c6grid_00,one_sixth),_mm_mul_pd(exponent,ewclj6));
2206 /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
2207 fvdw = _mm_mul_pd(_mm_add_pd(_mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),_mm_sub_pd(c6_00,f6A)),rinvsix),f6B),rinvsq00);
2209 cutoff_mask = _mm_cmplt_pd(rsq00,rcutoff2);
2211 fscal = _mm_add_pd(felec,fvdw);
2213 fscal = _mm_and_pd(fscal,cutoff_mask);
2215 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2217 /* Calculate temporary vectorial force */
2218 tx = _mm_mul_pd(fscal,dx00);
2219 ty = _mm_mul_pd(fscal,dy00);
2220 tz = _mm_mul_pd(fscal,dz00);
2222 /* Update vectorial force */
2223 fix0 = _mm_add_pd(fix0,tx);
2224 fiy0 = _mm_add_pd(fiy0,ty);
2225 fiz0 = _mm_add_pd(fiz0,tz);
2227 fjx0 = _mm_add_pd(fjx0,tx);
2228 fjy0 = _mm_add_pd(fjy0,ty);
2229 fjz0 = _mm_add_pd(fjz0,tz);
2233 /**************************
2234 * CALCULATE INTERACTIONS *
2235 **************************/
2237 if (gmx_mm_any_lt(rsq01,rcutoff2))
2240 r01 = _mm_mul_pd(rsq01,rinv01);
2242 /* EWALD ELECTROSTATICS */
2244 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2245 ewrt = _mm_mul_pd(r01,ewtabscale);
2246 ewitab = _mm_cvttpd_epi32(ewrt);
2247 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2248 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2249 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2250 felec = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
2252 cutoff_mask = _mm_cmplt_pd(rsq01,rcutoff2);
2256 fscal = _mm_and_pd(fscal,cutoff_mask);
2258 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2260 /* Calculate temporary vectorial force */
2261 tx = _mm_mul_pd(fscal,dx01);
2262 ty = _mm_mul_pd(fscal,dy01);
2263 tz = _mm_mul_pd(fscal,dz01);
2265 /* Update vectorial force */
2266 fix0 = _mm_add_pd(fix0,tx);
2267 fiy0 = _mm_add_pd(fiy0,ty);
2268 fiz0 = _mm_add_pd(fiz0,tz);
2270 fjx1 = _mm_add_pd(fjx1,tx);
2271 fjy1 = _mm_add_pd(fjy1,ty);
2272 fjz1 = _mm_add_pd(fjz1,tz);
2276 /**************************
2277 * CALCULATE INTERACTIONS *
2278 **************************/
2280 if (gmx_mm_any_lt(rsq02,rcutoff2))
2283 r02 = _mm_mul_pd(rsq02,rinv02);
2285 /* EWALD ELECTROSTATICS */
2287 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2288 ewrt = _mm_mul_pd(r02,ewtabscale);
2289 ewitab = _mm_cvttpd_epi32(ewrt);
2290 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2291 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2292 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2293 felec = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
2295 cutoff_mask = _mm_cmplt_pd(rsq02,rcutoff2);
2299 fscal = _mm_and_pd(fscal,cutoff_mask);
2301 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2303 /* Calculate temporary vectorial force */
2304 tx = _mm_mul_pd(fscal,dx02);
2305 ty = _mm_mul_pd(fscal,dy02);
2306 tz = _mm_mul_pd(fscal,dz02);
2308 /* Update vectorial force */
2309 fix0 = _mm_add_pd(fix0,tx);
2310 fiy0 = _mm_add_pd(fiy0,ty);
2311 fiz0 = _mm_add_pd(fiz0,tz);
2313 fjx2 = _mm_add_pd(fjx2,tx);
2314 fjy2 = _mm_add_pd(fjy2,ty);
2315 fjz2 = _mm_add_pd(fjz2,tz);
2319 /**************************
2320 * CALCULATE INTERACTIONS *
2321 **************************/
2323 if (gmx_mm_any_lt(rsq10,rcutoff2))
2326 r10 = _mm_mul_pd(rsq10,rinv10);
2328 /* EWALD ELECTROSTATICS */
2330 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2331 ewrt = _mm_mul_pd(r10,ewtabscale);
2332 ewitab = _mm_cvttpd_epi32(ewrt);
2333 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2334 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2335 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2336 felec = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
2338 cutoff_mask = _mm_cmplt_pd(rsq10,rcutoff2);
2342 fscal = _mm_and_pd(fscal,cutoff_mask);
2344 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2346 /* Calculate temporary vectorial force */
2347 tx = _mm_mul_pd(fscal,dx10);
2348 ty = _mm_mul_pd(fscal,dy10);
2349 tz = _mm_mul_pd(fscal,dz10);
2351 /* Update vectorial force */
2352 fix1 = _mm_add_pd(fix1,tx);
2353 fiy1 = _mm_add_pd(fiy1,ty);
2354 fiz1 = _mm_add_pd(fiz1,tz);
2356 fjx0 = _mm_add_pd(fjx0,tx);
2357 fjy0 = _mm_add_pd(fjy0,ty);
2358 fjz0 = _mm_add_pd(fjz0,tz);
2362 /**************************
2363 * CALCULATE INTERACTIONS *
2364 **************************/
2366 if (gmx_mm_any_lt(rsq11,rcutoff2))
2369 r11 = _mm_mul_pd(rsq11,rinv11);
2371 /* EWALD ELECTROSTATICS */
2373 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2374 ewrt = _mm_mul_pd(r11,ewtabscale);
2375 ewitab = _mm_cvttpd_epi32(ewrt);
2376 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2377 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2378 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2379 felec = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
2381 cutoff_mask = _mm_cmplt_pd(rsq11,rcutoff2);
2385 fscal = _mm_and_pd(fscal,cutoff_mask);
2387 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2389 /* Calculate temporary vectorial force */
2390 tx = _mm_mul_pd(fscal,dx11);
2391 ty = _mm_mul_pd(fscal,dy11);
2392 tz = _mm_mul_pd(fscal,dz11);
2394 /* Update vectorial force */
2395 fix1 = _mm_add_pd(fix1,tx);
2396 fiy1 = _mm_add_pd(fiy1,ty);
2397 fiz1 = _mm_add_pd(fiz1,tz);
2399 fjx1 = _mm_add_pd(fjx1,tx);
2400 fjy1 = _mm_add_pd(fjy1,ty);
2401 fjz1 = _mm_add_pd(fjz1,tz);
2405 /**************************
2406 * CALCULATE INTERACTIONS *
2407 **************************/
2409 if (gmx_mm_any_lt(rsq12,rcutoff2))
2412 r12 = _mm_mul_pd(rsq12,rinv12);
2414 /* EWALD ELECTROSTATICS */
2416 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2417 ewrt = _mm_mul_pd(r12,ewtabscale);
2418 ewitab = _mm_cvttpd_epi32(ewrt);
2419 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2420 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2421 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2422 felec = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
2424 cutoff_mask = _mm_cmplt_pd(rsq12,rcutoff2);
2428 fscal = _mm_and_pd(fscal,cutoff_mask);
2430 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2432 /* Calculate temporary vectorial force */
2433 tx = _mm_mul_pd(fscal,dx12);
2434 ty = _mm_mul_pd(fscal,dy12);
2435 tz = _mm_mul_pd(fscal,dz12);
2437 /* Update vectorial force */
2438 fix1 = _mm_add_pd(fix1,tx);
2439 fiy1 = _mm_add_pd(fiy1,ty);
2440 fiz1 = _mm_add_pd(fiz1,tz);
2442 fjx2 = _mm_add_pd(fjx2,tx);
2443 fjy2 = _mm_add_pd(fjy2,ty);
2444 fjz2 = _mm_add_pd(fjz2,tz);
2448 /**************************
2449 * CALCULATE INTERACTIONS *
2450 **************************/
2452 if (gmx_mm_any_lt(rsq20,rcutoff2))
2455 r20 = _mm_mul_pd(rsq20,rinv20);
2457 /* EWALD ELECTROSTATICS */
2459 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2460 ewrt = _mm_mul_pd(r20,ewtabscale);
2461 ewitab = _mm_cvttpd_epi32(ewrt);
2462 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2463 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2464 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2465 felec = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
2467 cutoff_mask = _mm_cmplt_pd(rsq20,rcutoff2);
2471 fscal = _mm_and_pd(fscal,cutoff_mask);
2473 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2475 /* Calculate temporary vectorial force */
2476 tx = _mm_mul_pd(fscal,dx20);
2477 ty = _mm_mul_pd(fscal,dy20);
2478 tz = _mm_mul_pd(fscal,dz20);
2480 /* Update vectorial force */
2481 fix2 = _mm_add_pd(fix2,tx);
2482 fiy2 = _mm_add_pd(fiy2,ty);
2483 fiz2 = _mm_add_pd(fiz2,tz);
2485 fjx0 = _mm_add_pd(fjx0,tx);
2486 fjy0 = _mm_add_pd(fjy0,ty);
2487 fjz0 = _mm_add_pd(fjz0,tz);
2491 /**************************
2492 * CALCULATE INTERACTIONS *
2493 **************************/
2495 if (gmx_mm_any_lt(rsq21,rcutoff2))
2498 r21 = _mm_mul_pd(rsq21,rinv21);
2500 /* EWALD ELECTROSTATICS */
2502 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2503 ewrt = _mm_mul_pd(r21,ewtabscale);
2504 ewitab = _mm_cvttpd_epi32(ewrt);
2505 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2506 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2507 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2508 felec = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
2510 cutoff_mask = _mm_cmplt_pd(rsq21,rcutoff2);
2514 fscal = _mm_and_pd(fscal,cutoff_mask);
2516 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2518 /* Calculate temporary vectorial force */
2519 tx = _mm_mul_pd(fscal,dx21);
2520 ty = _mm_mul_pd(fscal,dy21);
2521 tz = _mm_mul_pd(fscal,dz21);
2523 /* Update vectorial force */
2524 fix2 = _mm_add_pd(fix2,tx);
2525 fiy2 = _mm_add_pd(fiy2,ty);
2526 fiz2 = _mm_add_pd(fiz2,tz);
2528 fjx1 = _mm_add_pd(fjx1,tx);
2529 fjy1 = _mm_add_pd(fjy1,ty);
2530 fjz1 = _mm_add_pd(fjz1,tz);
2534 /**************************
2535 * CALCULATE INTERACTIONS *
2536 **************************/
2538 if (gmx_mm_any_lt(rsq22,rcutoff2))
2541 r22 = _mm_mul_pd(rsq22,rinv22);
2543 /* EWALD ELECTROSTATICS */
2545 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2546 ewrt = _mm_mul_pd(r22,ewtabscale);
2547 ewitab = _mm_cvttpd_epi32(ewrt);
2548 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2549 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2550 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2551 felec = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
2553 cutoff_mask = _mm_cmplt_pd(rsq22,rcutoff2);
2557 fscal = _mm_and_pd(fscal,cutoff_mask);
2559 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2561 /* Calculate temporary vectorial force */
2562 tx = _mm_mul_pd(fscal,dx22);
2563 ty = _mm_mul_pd(fscal,dy22);
2564 tz = _mm_mul_pd(fscal,dz22);
2566 /* Update vectorial force */
2567 fix2 = _mm_add_pd(fix2,tx);
2568 fiy2 = _mm_add_pd(fiy2,ty);
2569 fiz2 = _mm_add_pd(fiz2,tz);
2571 fjx2 = _mm_add_pd(fjx2,tx);
2572 fjy2 = _mm_add_pd(fjy2,ty);
2573 fjz2 = _mm_add_pd(fjz2,tz);
2577 gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
2579 /* Inner loop uses 374 flops */
2582 /* End of innermost loop */
2584 gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
2585 f+i_coord_offset,fshift+i_shift_offset);
2587 /* Increment number of inner iterations */
2588 inneriter += j_index_end - j_index_start;
2590 /* Outer loop uses 18 flops */
2593 /* Increment number of outer iterations */
2596 /* Update outer/inner flops */
2598 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_F,outeriter*18 + inneriter*374);