2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
5 * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
6 * and including many others, as listed in the AUTHORS file in the
7 * top-level source directory and at http://www.gromacs.org.
9 * GROMACS is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU Lesser General Public License
11 * as published by the Free Software Foundation; either version 2.1
12 * of the License, or (at your option) any later version.
14 * GROMACS is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
17 * Lesser General Public License for more details.
19 * You should have received a copy of the GNU Lesser General Public
20 * License along with GROMACS; if not, see
21 * http://www.gnu.org/licenses, or write to the Free Software Foundation,
22 * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
24 * If you want to redistribute modifications to GROMACS, please
25 * consider that scientific software is very special. Version
26 * control is crucial - bugs must be traceable. We will be happy to
27 * consider code for inclusion in the official distribution, but
28 * derived work must not be called official GROMACS. Details are found
29 * in the README & COPYING files - if they are missing, get the
30 * official version at http://www.gromacs.org.
32 * To help us fund GROMACS development, we humbly ask that you cite
33 * the research papers on the package. Check out http://www.gromacs.org.
36 * Note: this file was generated by the GROMACS sse2_double kernel generator.
42 #include "../nb_kernel.h"
43 #include "gromacs/legacyheaders/types/simple.h"
44 #include "gromacs/math/vec.h"
45 #include "gromacs/legacyheaders/nrnb.h"
47 #include "gromacs/simd/math_x86_sse2_double.h"
48 #include "kernelutil_x86_sse2_double.h"
51 * Gromacs nonbonded kernel: nb_kernel_ElecEwSh_VdwLJSh_GeomW3W3_VF_sse2_double
52 * Electrostatics interaction: Ewald
53 * VdW interaction: LennardJones
54 * Geometry: Water3-Water3
55 * Calculate force/pot: PotentialAndForce
58 nb_kernel_ElecEwSh_VdwLJSh_GeomW3W3_VF_sse2_double
59 (t_nblist * gmx_restrict nlist,
60 rvec * gmx_restrict xx,
61 rvec * gmx_restrict ff,
62 t_forcerec * gmx_restrict fr,
63 t_mdatoms * gmx_restrict mdatoms,
64 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
65 t_nrnb * gmx_restrict nrnb)
67 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
68 * just 0 for non-waters.
69 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
70 * jnr indices corresponding to data put in the four positions in the SIMD register.
72 int i_shift_offset,i_coord_offset,outeriter,inneriter;
73 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
75 int j_coord_offsetA,j_coord_offsetB;
76 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
78 real *shiftvec,*fshift,*x,*f;
79 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
81 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
83 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
85 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
86 int vdwjidx0A,vdwjidx0B;
87 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
88 int vdwjidx1A,vdwjidx1B;
89 __m128d jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
90 int vdwjidx2A,vdwjidx2B;
91 __m128d jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
92 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
93 __m128d dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
94 __m128d dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
95 __m128d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
96 __m128d dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
97 __m128d dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
98 __m128d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
99 __m128d dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
100 __m128d dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
101 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
104 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
107 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
108 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
110 __m128d ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
112 __m128d dummy_mask,cutoff_mask;
113 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
114 __m128d one = _mm_set1_pd(1.0);
115 __m128d two = _mm_set1_pd(2.0);
121 jindex = nlist->jindex;
123 shiftidx = nlist->shift;
125 shiftvec = fr->shift_vec[0];
126 fshift = fr->fshift[0];
127 facel = _mm_set1_pd(fr->epsfac);
128 charge = mdatoms->chargeA;
129 nvdwtype = fr->ntype;
131 vdwtype = mdatoms->typeA;
133 sh_ewald = _mm_set1_pd(fr->ic->sh_ewald);
134 ewtab = fr->ic->tabq_coul_FDV0;
135 ewtabscale = _mm_set1_pd(fr->ic->tabq_scale);
136 ewtabhalfspace = _mm_set1_pd(0.5/fr->ic->tabq_scale);
138 /* Setup water-specific parameters */
139 inr = nlist->iinr[0];
140 iq0 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
141 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
142 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
143 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
145 jq0 = _mm_set1_pd(charge[inr+0]);
146 jq1 = _mm_set1_pd(charge[inr+1]);
147 jq2 = _mm_set1_pd(charge[inr+2]);
148 vdwjidx0A = 2*vdwtype[inr+0];
149 qq00 = _mm_mul_pd(iq0,jq0);
150 c6_00 = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
151 c12_00 = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
152 qq01 = _mm_mul_pd(iq0,jq1);
153 qq02 = _mm_mul_pd(iq0,jq2);
154 qq10 = _mm_mul_pd(iq1,jq0);
155 qq11 = _mm_mul_pd(iq1,jq1);
156 qq12 = _mm_mul_pd(iq1,jq2);
157 qq20 = _mm_mul_pd(iq2,jq0);
158 qq21 = _mm_mul_pd(iq2,jq1);
159 qq22 = _mm_mul_pd(iq2,jq2);
161 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
162 rcutoff_scalar = fr->rcoulomb;
163 rcutoff = _mm_set1_pd(rcutoff_scalar);
164 rcutoff2 = _mm_mul_pd(rcutoff,rcutoff);
166 sh_vdw_invrcut6 = _mm_set1_pd(fr->ic->sh_invrc6);
167 rvdw = _mm_set1_pd(fr->rvdw);
169 /* Avoid stupid compiler warnings */
177 /* Start outer loop over neighborlists */
178 for(iidx=0; iidx<nri; iidx++)
180 /* Load shift vector for this list */
181 i_shift_offset = DIM*shiftidx[iidx];
183 /* Load limits for loop over neighbors */
184 j_index_start = jindex[iidx];
185 j_index_end = jindex[iidx+1];
187 /* Get outer coordinate index */
189 i_coord_offset = DIM*inr;
191 /* Load i particle coords and add shift vector */
192 gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
193 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
195 fix0 = _mm_setzero_pd();
196 fiy0 = _mm_setzero_pd();
197 fiz0 = _mm_setzero_pd();
198 fix1 = _mm_setzero_pd();
199 fiy1 = _mm_setzero_pd();
200 fiz1 = _mm_setzero_pd();
201 fix2 = _mm_setzero_pd();
202 fiy2 = _mm_setzero_pd();
203 fiz2 = _mm_setzero_pd();
205 /* Reset potential sums */
206 velecsum = _mm_setzero_pd();
207 vvdwsum = _mm_setzero_pd();
209 /* Start inner kernel loop */
210 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
213 /* Get j neighbor index, and coordinate index */
216 j_coord_offsetA = DIM*jnrA;
217 j_coord_offsetB = DIM*jnrB;
219 /* load j atom coordinates */
220 gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
221 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
223 /* Calculate displacement vector */
224 dx00 = _mm_sub_pd(ix0,jx0);
225 dy00 = _mm_sub_pd(iy0,jy0);
226 dz00 = _mm_sub_pd(iz0,jz0);
227 dx01 = _mm_sub_pd(ix0,jx1);
228 dy01 = _mm_sub_pd(iy0,jy1);
229 dz01 = _mm_sub_pd(iz0,jz1);
230 dx02 = _mm_sub_pd(ix0,jx2);
231 dy02 = _mm_sub_pd(iy0,jy2);
232 dz02 = _mm_sub_pd(iz0,jz2);
233 dx10 = _mm_sub_pd(ix1,jx0);
234 dy10 = _mm_sub_pd(iy1,jy0);
235 dz10 = _mm_sub_pd(iz1,jz0);
236 dx11 = _mm_sub_pd(ix1,jx1);
237 dy11 = _mm_sub_pd(iy1,jy1);
238 dz11 = _mm_sub_pd(iz1,jz1);
239 dx12 = _mm_sub_pd(ix1,jx2);
240 dy12 = _mm_sub_pd(iy1,jy2);
241 dz12 = _mm_sub_pd(iz1,jz2);
242 dx20 = _mm_sub_pd(ix2,jx0);
243 dy20 = _mm_sub_pd(iy2,jy0);
244 dz20 = _mm_sub_pd(iz2,jz0);
245 dx21 = _mm_sub_pd(ix2,jx1);
246 dy21 = _mm_sub_pd(iy2,jy1);
247 dz21 = _mm_sub_pd(iz2,jz1);
248 dx22 = _mm_sub_pd(ix2,jx2);
249 dy22 = _mm_sub_pd(iy2,jy2);
250 dz22 = _mm_sub_pd(iz2,jz2);
252 /* Calculate squared distance and things based on it */
253 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
254 rsq01 = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
255 rsq02 = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
256 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
257 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
258 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
259 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
260 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
261 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
263 rinv00 = gmx_mm_invsqrt_pd(rsq00);
264 rinv01 = gmx_mm_invsqrt_pd(rsq01);
265 rinv02 = gmx_mm_invsqrt_pd(rsq02);
266 rinv10 = gmx_mm_invsqrt_pd(rsq10);
267 rinv11 = gmx_mm_invsqrt_pd(rsq11);
268 rinv12 = gmx_mm_invsqrt_pd(rsq12);
269 rinv20 = gmx_mm_invsqrt_pd(rsq20);
270 rinv21 = gmx_mm_invsqrt_pd(rsq21);
271 rinv22 = gmx_mm_invsqrt_pd(rsq22);
273 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
274 rinvsq01 = _mm_mul_pd(rinv01,rinv01);
275 rinvsq02 = _mm_mul_pd(rinv02,rinv02);
276 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
277 rinvsq11 = _mm_mul_pd(rinv11,rinv11);
278 rinvsq12 = _mm_mul_pd(rinv12,rinv12);
279 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
280 rinvsq21 = _mm_mul_pd(rinv21,rinv21);
281 rinvsq22 = _mm_mul_pd(rinv22,rinv22);
283 fjx0 = _mm_setzero_pd();
284 fjy0 = _mm_setzero_pd();
285 fjz0 = _mm_setzero_pd();
286 fjx1 = _mm_setzero_pd();
287 fjy1 = _mm_setzero_pd();
288 fjz1 = _mm_setzero_pd();
289 fjx2 = _mm_setzero_pd();
290 fjy2 = _mm_setzero_pd();
291 fjz2 = _mm_setzero_pd();
293 /**************************
294 * CALCULATE INTERACTIONS *
295 **************************/
297 if (gmx_mm_any_lt(rsq00,rcutoff2))
300 r00 = _mm_mul_pd(rsq00,rinv00);
302 /* EWALD ELECTROSTATICS */
304 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
305 ewrt = _mm_mul_pd(r00,ewtabscale);
306 ewitab = _mm_cvttpd_epi32(ewrt);
307 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
308 ewitab = _mm_slli_epi32(ewitab,2);
309 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
310 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
311 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
312 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
313 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
314 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
315 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
316 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
317 velec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_sub_pd(rinv00,sh_ewald),velec));
318 felec = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
320 /* LENNARD-JONES DISPERSION/REPULSION */
322 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
323 vvdw6 = _mm_mul_pd(c6_00,rinvsix);
324 vvdw12 = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
325 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) ,
326 _mm_mul_pd( _mm_sub_pd(vvdw6,_mm_mul_pd(c6_00,sh_vdw_invrcut6)),one_sixth));
327 fvdw = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
329 cutoff_mask = _mm_cmplt_pd(rsq00,rcutoff2);
331 /* Update potential sum for this i atom from the interaction with this j atom. */
332 velec = _mm_and_pd(velec,cutoff_mask);
333 velecsum = _mm_add_pd(velecsum,velec);
334 vvdw = _mm_and_pd(vvdw,cutoff_mask);
335 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
337 fscal = _mm_add_pd(felec,fvdw);
339 fscal = _mm_and_pd(fscal,cutoff_mask);
341 /* Calculate temporary vectorial force */
342 tx = _mm_mul_pd(fscal,dx00);
343 ty = _mm_mul_pd(fscal,dy00);
344 tz = _mm_mul_pd(fscal,dz00);
346 /* Update vectorial force */
347 fix0 = _mm_add_pd(fix0,tx);
348 fiy0 = _mm_add_pd(fiy0,ty);
349 fiz0 = _mm_add_pd(fiz0,tz);
351 fjx0 = _mm_add_pd(fjx0,tx);
352 fjy0 = _mm_add_pd(fjy0,ty);
353 fjz0 = _mm_add_pd(fjz0,tz);
357 /**************************
358 * CALCULATE INTERACTIONS *
359 **************************/
361 if (gmx_mm_any_lt(rsq01,rcutoff2))
364 r01 = _mm_mul_pd(rsq01,rinv01);
366 /* EWALD ELECTROSTATICS */
368 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
369 ewrt = _mm_mul_pd(r01,ewtabscale);
370 ewitab = _mm_cvttpd_epi32(ewrt);
371 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
372 ewitab = _mm_slli_epi32(ewitab,2);
373 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
374 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
375 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
376 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
377 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
378 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
379 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
380 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
381 velec = _mm_mul_pd(qq01,_mm_sub_pd(_mm_sub_pd(rinv01,sh_ewald),velec));
382 felec = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
384 cutoff_mask = _mm_cmplt_pd(rsq01,rcutoff2);
386 /* Update potential sum for this i atom from the interaction with this j atom. */
387 velec = _mm_and_pd(velec,cutoff_mask);
388 velecsum = _mm_add_pd(velecsum,velec);
392 fscal = _mm_and_pd(fscal,cutoff_mask);
394 /* Calculate temporary vectorial force */
395 tx = _mm_mul_pd(fscal,dx01);
396 ty = _mm_mul_pd(fscal,dy01);
397 tz = _mm_mul_pd(fscal,dz01);
399 /* Update vectorial force */
400 fix0 = _mm_add_pd(fix0,tx);
401 fiy0 = _mm_add_pd(fiy0,ty);
402 fiz0 = _mm_add_pd(fiz0,tz);
404 fjx1 = _mm_add_pd(fjx1,tx);
405 fjy1 = _mm_add_pd(fjy1,ty);
406 fjz1 = _mm_add_pd(fjz1,tz);
410 /**************************
411 * CALCULATE INTERACTIONS *
412 **************************/
414 if (gmx_mm_any_lt(rsq02,rcutoff2))
417 r02 = _mm_mul_pd(rsq02,rinv02);
419 /* EWALD ELECTROSTATICS */
421 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
422 ewrt = _mm_mul_pd(r02,ewtabscale);
423 ewitab = _mm_cvttpd_epi32(ewrt);
424 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
425 ewitab = _mm_slli_epi32(ewitab,2);
426 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
427 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
428 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
429 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
430 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
431 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
432 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
433 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
434 velec = _mm_mul_pd(qq02,_mm_sub_pd(_mm_sub_pd(rinv02,sh_ewald),velec));
435 felec = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
437 cutoff_mask = _mm_cmplt_pd(rsq02,rcutoff2);
439 /* Update potential sum for this i atom from the interaction with this j atom. */
440 velec = _mm_and_pd(velec,cutoff_mask);
441 velecsum = _mm_add_pd(velecsum,velec);
445 fscal = _mm_and_pd(fscal,cutoff_mask);
447 /* Calculate temporary vectorial force */
448 tx = _mm_mul_pd(fscal,dx02);
449 ty = _mm_mul_pd(fscal,dy02);
450 tz = _mm_mul_pd(fscal,dz02);
452 /* Update vectorial force */
453 fix0 = _mm_add_pd(fix0,tx);
454 fiy0 = _mm_add_pd(fiy0,ty);
455 fiz0 = _mm_add_pd(fiz0,tz);
457 fjx2 = _mm_add_pd(fjx2,tx);
458 fjy2 = _mm_add_pd(fjy2,ty);
459 fjz2 = _mm_add_pd(fjz2,tz);
463 /**************************
464 * CALCULATE INTERACTIONS *
465 **************************/
467 if (gmx_mm_any_lt(rsq10,rcutoff2))
470 r10 = _mm_mul_pd(rsq10,rinv10);
472 /* EWALD ELECTROSTATICS */
474 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
475 ewrt = _mm_mul_pd(r10,ewtabscale);
476 ewitab = _mm_cvttpd_epi32(ewrt);
477 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
478 ewitab = _mm_slli_epi32(ewitab,2);
479 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
480 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
481 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
482 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
483 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
484 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
485 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
486 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
487 velec = _mm_mul_pd(qq10,_mm_sub_pd(_mm_sub_pd(rinv10,sh_ewald),velec));
488 felec = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
490 cutoff_mask = _mm_cmplt_pd(rsq10,rcutoff2);
492 /* Update potential sum for this i atom from the interaction with this j atom. */
493 velec = _mm_and_pd(velec,cutoff_mask);
494 velecsum = _mm_add_pd(velecsum,velec);
498 fscal = _mm_and_pd(fscal,cutoff_mask);
500 /* Calculate temporary vectorial force */
501 tx = _mm_mul_pd(fscal,dx10);
502 ty = _mm_mul_pd(fscal,dy10);
503 tz = _mm_mul_pd(fscal,dz10);
505 /* Update vectorial force */
506 fix1 = _mm_add_pd(fix1,tx);
507 fiy1 = _mm_add_pd(fiy1,ty);
508 fiz1 = _mm_add_pd(fiz1,tz);
510 fjx0 = _mm_add_pd(fjx0,tx);
511 fjy0 = _mm_add_pd(fjy0,ty);
512 fjz0 = _mm_add_pd(fjz0,tz);
516 /**************************
517 * CALCULATE INTERACTIONS *
518 **************************/
520 if (gmx_mm_any_lt(rsq11,rcutoff2))
523 r11 = _mm_mul_pd(rsq11,rinv11);
525 /* EWALD ELECTROSTATICS */
527 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
528 ewrt = _mm_mul_pd(r11,ewtabscale);
529 ewitab = _mm_cvttpd_epi32(ewrt);
530 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
531 ewitab = _mm_slli_epi32(ewitab,2);
532 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
533 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
534 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
535 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
536 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
537 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
538 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
539 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
540 velec = _mm_mul_pd(qq11,_mm_sub_pd(_mm_sub_pd(rinv11,sh_ewald),velec));
541 felec = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
543 cutoff_mask = _mm_cmplt_pd(rsq11,rcutoff2);
545 /* Update potential sum for this i atom from the interaction with this j atom. */
546 velec = _mm_and_pd(velec,cutoff_mask);
547 velecsum = _mm_add_pd(velecsum,velec);
551 fscal = _mm_and_pd(fscal,cutoff_mask);
553 /* Calculate temporary vectorial force */
554 tx = _mm_mul_pd(fscal,dx11);
555 ty = _mm_mul_pd(fscal,dy11);
556 tz = _mm_mul_pd(fscal,dz11);
558 /* Update vectorial force */
559 fix1 = _mm_add_pd(fix1,tx);
560 fiy1 = _mm_add_pd(fiy1,ty);
561 fiz1 = _mm_add_pd(fiz1,tz);
563 fjx1 = _mm_add_pd(fjx1,tx);
564 fjy1 = _mm_add_pd(fjy1,ty);
565 fjz1 = _mm_add_pd(fjz1,tz);
569 /**************************
570 * CALCULATE INTERACTIONS *
571 **************************/
573 if (gmx_mm_any_lt(rsq12,rcutoff2))
576 r12 = _mm_mul_pd(rsq12,rinv12);
578 /* EWALD ELECTROSTATICS */
580 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
581 ewrt = _mm_mul_pd(r12,ewtabscale);
582 ewitab = _mm_cvttpd_epi32(ewrt);
583 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
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(qq12,_mm_sub_pd(_mm_sub_pd(rinv12,sh_ewald),velec));
594 felec = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
596 cutoff_mask = _mm_cmplt_pd(rsq12,rcutoff2);
598 /* Update potential sum for this i atom from the interaction with this j atom. */
599 velec = _mm_and_pd(velec,cutoff_mask);
600 velecsum = _mm_add_pd(velecsum,velec);
604 fscal = _mm_and_pd(fscal,cutoff_mask);
606 /* Calculate temporary vectorial force */
607 tx = _mm_mul_pd(fscal,dx12);
608 ty = _mm_mul_pd(fscal,dy12);
609 tz = _mm_mul_pd(fscal,dz12);
611 /* Update vectorial force */
612 fix1 = _mm_add_pd(fix1,tx);
613 fiy1 = _mm_add_pd(fiy1,ty);
614 fiz1 = _mm_add_pd(fiz1,tz);
616 fjx2 = _mm_add_pd(fjx2,tx);
617 fjy2 = _mm_add_pd(fjy2,ty);
618 fjz2 = _mm_add_pd(fjz2,tz);
622 /**************************
623 * CALCULATE INTERACTIONS *
624 **************************/
626 if (gmx_mm_any_lt(rsq20,rcutoff2))
629 r20 = _mm_mul_pd(rsq20,rinv20);
631 /* EWALD ELECTROSTATICS */
633 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
634 ewrt = _mm_mul_pd(r20,ewtabscale);
635 ewitab = _mm_cvttpd_epi32(ewrt);
636 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
637 ewitab = _mm_slli_epi32(ewitab,2);
638 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
639 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
640 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
641 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
642 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
643 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
644 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
645 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
646 velec = _mm_mul_pd(qq20,_mm_sub_pd(_mm_sub_pd(rinv20,sh_ewald),velec));
647 felec = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
649 cutoff_mask = _mm_cmplt_pd(rsq20,rcutoff2);
651 /* Update potential sum for this i atom from the interaction with this j atom. */
652 velec = _mm_and_pd(velec,cutoff_mask);
653 velecsum = _mm_add_pd(velecsum,velec);
657 fscal = _mm_and_pd(fscal,cutoff_mask);
659 /* Calculate temporary vectorial force */
660 tx = _mm_mul_pd(fscal,dx20);
661 ty = _mm_mul_pd(fscal,dy20);
662 tz = _mm_mul_pd(fscal,dz20);
664 /* Update vectorial force */
665 fix2 = _mm_add_pd(fix2,tx);
666 fiy2 = _mm_add_pd(fiy2,ty);
667 fiz2 = _mm_add_pd(fiz2,tz);
669 fjx0 = _mm_add_pd(fjx0,tx);
670 fjy0 = _mm_add_pd(fjy0,ty);
671 fjz0 = _mm_add_pd(fjz0,tz);
675 /**************************
676 * CALCULATE INTERACTIONS *
677 **************************/
679 if (gmx_mm_any_lt(rsq21,rcutoff2))
682 r21 = _mm_mul_pd(rsq21,rinv21);
684 /* EWALD ELECTROSTATICS */
686 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
687 ewrt = _mm_mul_pd(r21,ewtabscale);
688 ewitab = _mm_cvttpd_epi32(ewrt);
689 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
690 ewitab = _mm_slli_epi32(ewitab,2);
691 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
692 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
693 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
694 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
695 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
696 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
697 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
698 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
699 velec = _mm_mul_pd(qq21,_mm_sub_pd(_mm_sub_pd(rinv21,sh_ewald),velec));
700 felec = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
702 cutoff_mask = _mm_cmplt_pd(rsq21,rcutoff2);
704 /* Update potential sum for this i atom from the interaction with this j atom. */
705 velec = _mm_and_pd(velec,cutoff_mask);
706 velecsum = _mm_add_pd(velecsum,velec);
710 fscal = _mm_and_pd(fscal,cutoff_mask);
712 /* Calculate temporary vectorial force */
713 tx = _mm_mul_pd(fscal,dx21);
714 ty = _mm_mul_pd(fscal,dy21);
715 tz = _mm_mul_pd(fscal,dz21);
717 /* Update vectorial force */
718 fix2 = _mm_add_pd(fix2,tx);
719 fiy2 = _mm_add_pd(fiy2,ty);
720 fiz2 = _mm_add_pd(fiz2,tz);
722 fjx1 = _mm_add_pd(fjx1,tx);
723 fjy1 = _mm_add_pd(fjy1,ty);
724 fjz1 = _mm_add_pd(fjz1,tz);
728 /**************************
729 * CALCULATE INTERACTIONS *
730 **************************/
732 if (gmx_mm_any_lt(rsq22,rcutoff2))
735 r22 = _mm_mul_pd(rsq22,rinv22);
737 /* EWALD ELECTROSTATICS */
739 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
740 ewrt = _mm_mul_pd(r22,ewtabscale);
741 ewitab = _mm_cvttpd_epi32(ewrt);
742 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
743 ewitab = _mm_slli_epi32(ewitab,2);
744 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
745 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
746 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
747 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
748 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
749 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
750 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
751 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
752 velec = _mm_mul_pd(qq22,_mm_sub_pd(_mm_sub_pd(rinv22,sh_ewald),velec));
753 felec = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
755 cutoff_mask = _mm_cmplt_pd(rsq22,rcutoff2);
757 /* Update potential sum for this i atom from the interaction with this j atom. */
758 velec = _mm_and_pd(velec,cutoff_mask);
759 velecsum = _mm_add_pd(velecsum,velec);
763 fscal = _mm_and_pd(fscal,cutoff_mask);
765 /* Calculate temporary vectorial force */
766 tx = _mm_mul_pd(fscal,dx22);
767 ty = _mm_mul_pd(fscal,dy22);
768 tz = _mm_mul_pd(fscal,dz22);
770 /* Update vectorial force */
771 fix2 = _mm_add_pd(fix2,tx);
772 fiy2 = _mm_add_pd(fiy2,ty);
773 fiz2 = _mm_add_pd(fiz2,tz);
775 fjx2 = _mm_add_pd(fjx2,tx);
776 fjy2 = _mm_add_pd(fjy2,ty);
777 fjz2 = _mm_add_pd(fjz2,tz);
781 gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
783 /* Inner loop uses 432 flops */
790 j_coord_offsetA = DIM*jnrA;
792 /* load j atom coordinates */
793 gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
794 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
796 /* Calculate displacement vector */
797 dx00 = _mm_sub_pd(ix0,jx0);
798 dy00 = _mm_sub_pd(iy0,jy0);
799 dz00 = _mm_sub_pd(iz0,jz0);
800 dx01 = _mm_sub_pd(ix0,jx1);
801 dy01 = _mm_sub_pd(iy0,jy1);
802 dz01 = _mm_sub_pd(iz0,jz1);
803 dx02 = _mm_sub_pd(ix0,jx2);
804 dy02 = _mm_sub_pd(iy0,jy2);
805 dz02 = _mm_sub_pd(iz0,jz2);
806 dx10 = _mm_sub_pd(ix1,jx0);
807 dy10 = _mm_sub_pd(iy1,jy0);
808 dz10 = _mm_sub_pd(iz1,jz0);
809 dx11 = _mm_sub_pd(ix1,jx1);
810 dy11 = _mm_sub_pd(iy1,jy1);
811 dz11 = _mm_sub_pd(iz1,jz1);
812 dx12 = _mm_sub_pd(ix1,jx2);
813 dy12 = _mm_sub_pd(iy1,jy2);
814 dz12 = _mm_sub_pd(iz1,jz2);
815 dx20 = _mm_sub_pd(ix2,jx0);
816 dy20 = _mm_sub_pd(iy2,jy0);
817 dz20 = _mm_sub_pd(iz2,jz0);
818 dx21 = _mm_sub_pd(ix2,jx1);
819 dy21 = _mm_sub_pd(iy2,jy1);
820 dz21 = _mm_sub_pd(iz2,jz1);
821 dx22 = _mm_sub_pd(ix2,jx2);
822 dy22 = _mm_sub_pd(iy2,jy2);
823 dz22 = _mm_sub_pd(iz2,jz2);
825 /* Calculate squared distance and things based on it */
826 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
827 rsq01 = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
828 rsq02 = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
829 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
830 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
831 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
832 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
833 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
834 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
836 rinv00 = gmx_mm_invsqrt_pd(rsq00);
837 rinv01 = gmx_mm_invsqrt_pd(rsq01);
838 rinv02 = gmx_mm_invsqrt_pd(rsq02);
839 rinv10 = gmx_mm_invsqrt_pd(rsq10);
840 rinv11 = gmx_mm_invsqrt_pd(rsq11);
841 rinv12 = gmx_mm_invsqrt_pd(rsq12);
842 rinv20 = gmx_mm_invsqrt_pd(rsq20);
843 rinv21 = gmx_mm_invsqrt_pd(rsq21);
844 rinv22 = gmx_mm_invsqrt_pd(rsq22);
846 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
847 rinvsq01 = _mm_mul_pd(rinv01,rinv01);
848 rinvsq02 = _mm_mul_pd(rinv02,rinv02);
849 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
850 rinvsq11 = _mm_mul_pd(rinv11,rinv11);
851 rinvsq12 = _mm_mul_pd(rinv12,rinv12);
852 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
853 rinvsq21 = _mm_mul_pd(rinv21,rinv21);
854 rinvsq22 = _mm_mul_pd(rinv22,rinv22);
856 fjx0 = _mm_setzero_pd();
857 fjy0 = _mm_setzero_pd();
858 fjz0 = _mm_setzero_pd();
859 fjx1 = _mm_setzero_pd();
860 fjy1 = _mm_setzero_pd();
861 fjz1 = _mm_setzero_pd();
862 fjx2 = _mm_setzero_pd();
863 fjy2 = _mm_setzero_pd();
864 fjz2 = _mm_setzero_pd();
866 /**************************
867 * CALCULATE INTERACTIONS *
868 **************************/
870 if (gmx_mm_any_lt(rsq00,rcutoff2))
873 r00 = _mm_mul_pd(rsq00,rinv00);
875 /* EWALD ELECTROSTATICS */
877 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
878 ewrt = _mm_mul_pd(r00,ewtabscale);
879 ewitab = _mm_cvttpd_epi32(ewrt);
880 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
881 ewitab = _mm_slli_epi32(ewitab,2);
882 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
883 ewtabD = _mm_setzero_pd();
884 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
885 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
886 ewtabFn = _mm_setzero_pd();
887 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
888 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
889 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
890 velec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_sub_pd(rinv00,sh_ewald),velec));
891 felec = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
893 /* LENNARD-JONES DISPERSION/REPULSION */
895 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
896 vvdw6 = _mm_mul_pd(c6_00,rinvsix);
897 vvdw12 = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
898 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) ,
899 _mm_mul_pd( _mm_sub_pd(vvdw6,_mm_mul_pd(c6_00,sh_vdw_invrcut6)),one_sixth));
900 fvdw = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
902 cutoff_mask = _mm_cmplt_pd(rsq00,rcutoff2);
904 /* Update potential sum for this i atom from the interaction with this j atom. */
905 velec = _mm_and_pd(velec,cutoff_mask);
906 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
907 velecsum = _mm_add_pd(velecsum,velec);
908 vvdw = _mm_and_pd(vvdw,cutoff_mask);
909 vvdw = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
910 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
912 fscal = _mm_add_pd(felec,fvdw);
914 fscal = _mm_and_pd(fscal,cutoff_mask);
916 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
918 /* Calculate temporary vectorial force */
919 tx = _mm_mul_pd(fscal,dx00);
920 ty = _mm_mul_pd(fscal,dy00);
921 tz = _mm_mul_pd(fscal,dz00);
923 /* Update vectorial force */
924 fix0 = _mm_add_pd(fix0,tx);
925 fiy0 = _mm_add_pd(fiy0,ty);
926 fiz0 = _mm_add_pd(fiz0,tz);
928 fjx0 = _mm_add_pd(fjx0,tx);
929 fjy0 = _mm_add_pd(fjy0,ty);
930 fjz0 = _mm_add_pd(fjz0,tz);
934 /**************************
935 * CALCULATE INTERACTIONS *
936 **************************/
938 if (gmx_mm_any_lt(rsq01,rcutoff2))
941 r01 = _mm_mul_pd(rsq01,rinv01);
943 /* EWALD ELECTROSTATICS */
945 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
946 ewrt = _mm_mul_pd(r01,ewtabscale);
947 ewitab = _mm_cvttpd_epi32(ewrt);
948 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
949 ewitab = _mm_slli_epi32(ewitab,2);
950 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
951 ewtabD = _mm_setzero_pd();
952 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
953 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
954 ewtabFn = _mm_setzero_pd();
955 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
956 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
957 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
958 velec = _mm_mul_pd(qq01,_mm_sub_pd(_mm_sub_pd(rinv01,sh_ewald),velec));
959 felec = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
961 cutoff_mask = _mm_cmplt_pd(rsq01,rcutoff2);
963 /* Update potential sum for this i atom from the interaction with this j atom. */
964 velec = _mm_and_pd(velec,cutoff_mask);
965 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
966 velecsum = _mm_add_pd(velecsum,velec);
970 fscal = _mm_and_pd(fscal,cutoff_mask);
972 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
974 /* Calculate temporary vectorial force */
975 tx = _mm_mul_pd(fscal,dx01);
976 ty = _mm_mul_pd(fscal,dy01);
977 tz = _mm_mul_pd(fscal,dz01);
979 /* Update vectorial force */
980 fix0 = _mm_add_pd(fix0,tx);
981 fiy0 = _mm_add_pd(fiy0,ty);
982 fiz0 = _mm_add_pd(fiz0,tz);
984 fjx1 = _mm_add_pd(fjx1,tx);
985 fjy1 = _mm_add_pd(fjy1,ty);
986 fjz1 = _mm_add_pd(fjz1,tz);
990 /**************************
991 * CALCULATE INTERACTIONS *
992 **************************/
994 if (gmx_mm_any_lt(rsq02,rcutoff2))
997 r02 = _mm_mul_pd(rsq02,rinv02);
999 /* EWALD ELECTROSTATICS */
1001 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1002 ewrt = _mm_mul_pd(r02,ewtabscale);
1003 ewitab = _mm_cvttpd_epi32(ewrt);
1004 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1005 ewitab = _mm_slli_epi32(ewitab,2);
1006 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1007 ewtabD = _mm_setzero_pd();
1008 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1009 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1010 ewtabFn = _mm_setzero_pd();
1011 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1012 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1013 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1014 velec = _mm_mul_pd(qq02,_mm_sub_pd(_mm_sub_pd(rinv02,sh_ewald),velec));
1015 felec = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
1017 cutoff_mask = _mm_cmplt_pd(rsq02,rcutoff2);
1019 /* Update potential sum for this i atom from the interaction with this j atom. */
1020 velec = _mm_and_pd(velec,cutoff_mask);
1021 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1022 velecsum = _mm_add_pd(velecsum,velec);
1026 fscal = _mm_and_pd(fscal,cutoff_mask);
1028 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1030 /* Calculate temporary vectorial force */
1031 tx = _mm_mul_pd(fscal,dx02);
1032 ty = _mm_mul_pd(fscal,dy02);
1033 tz = _mm_mul_pd(fscal,dz02);
1035 /* Update vectorial force */
1036 fix0 = _mm_add_pd(fix0,tx);
1037 fiy0 = _mm_add_pd(fiy0,ty);
1038 fiz0 = _mm_add_pd(fiz0,tz);
1040 fjx2 = _mm_add_pd(fjx2,tx);
1041 fjy2 = _mm_add_pd(fjy2,ty);
1042 fjz2 = _mm_add_pd(fjz2,tz);
1046 /**************************
1047 * CALCULATE INTERACTIONS *
1048 **************************/
1050 if (gmx_mm_any_lt(rsq10,rcutoff2))
1053 r10 = _mm_mul_pd(rsq10,rinv10);
1055 /* EWALD ELECTROSTATICS */
1057 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1058 ewrt = _mm_mul_pd(r10,ewtabscale);
1059 ewitab = _mm_cvttpd_epi32(ewrt);
1060 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1061 ewitab = _mm_slli_epi32(ewitab,2);
1062 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1063 ewtabD = _mm_setzero_pd();
1064 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1065 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1066 ewtabFn = _mm_setzero_pd();
1067 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1068 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1069 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1070 velec = _mm_mul_pd(qq10,_mm_sub_pd(_mm_sub_pd(rinv10,sh_ewald),velec));
1071 felec = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
1073 cutoff_mask = _mm_cmplt_pd(rsq10,rcutoff2);
1075 /* Update potential sum for this i atom from the interaction with this j atom. */
1076 velec = _mm_and_pd(velec,cutoff_mask);
1077 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1078 velecsum = _mm_add_pd(velecsum,velec);
1082 fscal = _mm_and_pd(fscal,cutoff_mask);
1084 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1086 /* Calculate temporary vectorial force */
1087 tx = _mm_mul_pd(fscal,dx10);
1088 ty = _mm_mul_pd(fscal,dy10);
1089 tz = _mm_mul_pd(fscal,dz10);
1091 /* Update vectorial force */
1092 fix1 = _mm_add_pd(fix1,tx);
1093 fiy1 = _mm_add_pd(fiy1,ty);
1094 fiz1 = _mm_add_pd(fiz1,tz);
1096 fjx0 = _mm_add_pd(fjx0,tx);
1097 fjy0 = _mm_add_pd(fjy0,ty);
1098 fjz0 = _mm_add_pd(fjz0,tz);
1102 /**************************
1103 * CALCULATE INTERACTIONS *
1104 **************************/
1106 if (gmx_mm_any_lt(rsq11,rcutoff2))
1109 r11 = _mm_mul_pd(rsq11,rinv11);
1111 /* EWALD ELECTROSTATICS */
1113 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1114 ewrt = _mm_mul_pd(r11,ewtabscale);
1115 ewitab = _mm_cvttpd_epi32(ewrt);
1116 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1117 ewitab = _mm_slli_epi32(ewitab,2);
1118 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1119 ewtabD = _mm_setzero_pd();
1120 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1121 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1122 ewtabFn = _mm_setzero_pd();
1123 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1124 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1125 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1126 velec = _mm_mul_pd(qq11,_mm_sub_pd(_mm_sub_pd(rinv11,sh_ewald),velec));
1127 felec = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
1129 cutoff_mask = _mm_cmplt_pd(rsq11,rcutoff2);
1131 /* Update potential sum for this i atom from the interaction with this j atom. */
1132 velec = _mm_and_pd(velec,cutoff_mask);
1133 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1134 velecsum = _mm_add_pd(velecsum,velec);
1138 fscal = _mm_and_pd(fscal,cutoff_mask);
1140 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1142 /* Calculate temporary vectorial force */
1143 tx = _mm_mul_pd(fscal,dx11);
1144 ty = _mm_mul_pd(fscal,dy11);
1145 tz = _mm_mul_pd(fscal,dz11);
1147 /* Update vectorial force */
1148 fix1 = _mm_add_pd(fix1,tx);
1149 fiy1 = _mm_add_pd(fiy1,ty);
1150 fiz1 = _mm_add_pd(fiz1,tz);
1152 fjx1 = _mm_add_pd(fjx1,tx);
1153 fjy1 = _mm_add_pd(fjy1,ty);
1154 fjz1 = _mm_add_pd(fjz1,tz);
1158 /**************************
1159 * CALCULATE INTERACTIONS *
1160 **************************/
1162 if (gmx_mm_any_lt(rsq12,rcutoff2))
1165 r12 = _mm_mul_pd(rsq12,rinv12);
1167 /* EWALD ELECTROSTATICS */
1169 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1170 ewrt = _mm_mul_pd(r12,ewtabscale);
1171 ewitab = _mm_cvttpd_epi32(ewrt);
1172 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1173 ewitab = _mm_slli_epi32(ewitab,2);
1174 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1175 ewtabD = _mm_setzero_pd();
1176 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1177 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1178 ewtabFn = _mm_setzero_pd();
1179 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1180 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1181 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1182 velec = _mm_mul_pd(qq12,_mm_sub_pd(_mm_sub_pd(rinv12,sh_ewald),velec));
1183 felec = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
1185 cutoff_mask = _mm_cmplt_pd(rsq12,rcutoff2);
1187 /* Update potential sum for this i atom from the interaction with this j atom. */
1188 velec = _mm_and_pd(velec,cutoff_mask);
1189 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1190 velecsum = _mm_add_pd(velecsum,velec);
1194 fscal = _mm_and_pd(fscal,cutoff_mask);
1196 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1198 /* Calculate temporary vectorial force */
1199 tx = _mm_mul_pd(fscal,dx12);
1200 ty = _mm_mul_pd(fscal,dy12);
1201 tz = _mm_mul_pd(fscal,dz12);
1203 /* Update vectorial force */
1204 fix1 = _mm_add_pd(fix1,tx);
1205 fiy1 = _mm_add_pd(fiy1,ty);
1206 fiz1 = _mm_add_pd(fiz1,tz);
1208 fjx2 = _mm_add_pd(fjx2,tx);
1209 fjy2 = _mm_add_pd(fjy2,ty);
1210 fjz2 = _mm_add_pd(fjz2,tz);
1214 /**************************
1215 * CALCULATE INTERACTIONS *
1216 **************************/
1218 if (gmx_mm_any_lt(rsq20,rcutoff2))
1221 r20 = _mm_mul_pd(rsq20,rinv20);
1223 /* EWALD ELECTROSTATICS */
1225 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1226 ewrt = _mm_mul_pd(r20,ewtabscale);
1227 ewitab = _mm_cvttpd_epi32(ewrt);
1228 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1229 ewitab = _mm_slli_epi32(ewitab,2);
1230 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1231 ewtabD = _mm_setzero_pd();
1232 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1233 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1234 ewtabFn = _mm_setzero_pd();
1235 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1236 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1237 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1238 velec = _mm_mul_pd(qq20,_mm_sub_pd(_mm_sub_pd(rinv20,sh_ewald),velec));
1239 felec = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
1241 cutoff_mask = _mm_cmplt_pd(rsq20,rcutoff2);
1243 /* Update potential sum for this i atom from the interaction with this j atom. */
1244 velec = _mm_and_pd(velec,cutoff_mask);
1245 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1246 velecsum = _mm_add_pd(velecsum,velec);
1250 fscal = _mm_and_pd(fscal,cutoff_mask);
1252 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1254 /* Calculate temporary vectorial force */
1255 tx = _mm_mul_pd(fscal,dx20);
1256 ty = _mm_mul_pd(fscal,dy20);
1257 tz = _mm_mul_pd(fscal,dz20);
1259 /* Update vectorial force */
1260 fix2 = _mm_add_pd(fix2,tx);
1261 fiy2 = _mm_add_pd(fiy2,ty);
1262 fiz2 = _mm_add_pd(fiz2,tz);
1264 fjx0 = _mm_add_pd(fjx0,tx);
1265 fjy0 = _mm_add_pd(fjy0,ty);
1266 fjz0 = _mm_add_pd(fjz0,tz);
1270 /**************************
1271 * CALCULATE INTERACTIONS *
1272 **************************/
1274 if (gmx_mm_any_lt(rsq21,rcutoff2))
1277 r21 = _mm_mul_pd(rsq21,rinv21);
1279 /* EWALD ELECTROSTATICS */
1281 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1282 ewrt = _mm_mul_pd(r21,ewtabscale);
1283 ewitab = _mm_cvttpd_epi32(ewrt);
1284 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1285 ewitab = _mm_slli_epi32(ewitab,2);
1286 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1287 ewtabD = _mm_setzero_pd();
1288 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1289 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1290 ewtabFn = _mm_setzero_pd();
1291 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1292 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1293 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1294 velec = _mm_mul_pd(qq21,_mm_sub_pd(_mm_sub_pd(rinv21,sh_ewald),velec));
1295 felec = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
1297 cutoff_mask = _mm_cmplt_pd(rsq21,rcutoff2);
1299 /* Update potential sum for this i atom from the interaction with this j atom. */
1300 velec = _mm_and_pd(velec,cutoff_mask);
1301 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1302 velecsum = _mm_add_pd(velecsum,velec);
1306 fscal = _mm_and_pd(fscal,cutoff_mask);
1308 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1310 /* Calculate temporary vectorial force */
1311 tx = _mm_mul_pd(fscal,dx21);
1312 ty = _mm_mul_pd(fscal,dy21);
1313 tz = _mm_mul_pd(fscal,dz21);
1315 /* Update vectorial force */
1316 fix2 = _mm_add_pd(fix2,tx);
1317 fiy2 = _mm_add_pd(fiy2,ty);
1318 fiz2 = _mm_add_pd(fiz2,tz);
1320 fjx1 = _mm_add_pd(fjx1,tx);
1321 fjy1 = _mm_add_pd(fjy1,ty);
1322 fjz1 = _mm_add_pd(fjz1,tz);
1326 /**************************
1327 * CALCULATE INTERACTIONS *
1328 **************************/
1330 if (gmx_mm_any_lt(rsq22,rcutoff2))
1333 r22 = _mm_mul_pd(rsq22,rinv22);
1335 /* EWALD ELECTROSTATICS */
1337 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1338 ewrt = _mm_mul_pd(r22,ewtabscale);
1339 ewitab = _mm_cvttpd_epi32(ewrt);
1340 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1341 ewitab = _mm_slli_epi32(ewitab,2);
1342 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1343 ewtabD = _mm_setzero_pd();
1344 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1345 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1346 ewtabFn = _mm_setzero_pd();
1347 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1348 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1349 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1350 velec = _mm_mul_pd(qq22,_mm_sub_pd(_mm_sub_pd(rinv22,sh_ewald),velec));
1351 felec = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
1353 cutoff_mask = _mm_cmplt_pd(rsq22,rcutoff2);
1355 /* Update potential sum for this i atom from the interaction with this j atom. */
1356 velec = _mm_and_pd(velec,cutoff_mask);
1357 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1358 velecsum = _mm_add_pd(velecsum,velec);
1362 fscal = _mm_and_pd(fscal,cutoff_mask);
1364 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1366 /* Calculate temporary vectorial force */
1367 tx = _mm_mul_pd(fscal,dx22);
1368 ty = _mm_mul_pd(fscal,dy22);
1369 tz = _mm_mul_pd(fscal,dz22);
1371 /* Update vectorial force */
1372 fix2 = _mm_add_pd(fix2,tx);
1373 fiy2 = _mm_add_pd(fiy2,ty);
1374 fiz2 = _mm_add_pd(fiz2,tz);
1376 fjx2 = _mm_add_pd(fjx2,tx);
1377 fjy2 = _mm_add_pd(fjy2,ty);
1378 fjz2 = _mm_add_pd(fjz2,tz);
1382 gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1384 /* Inner loop uses 432 flops */
1387 /* End of innermost loop */
1389 gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1390 f+i_coord_offset,fshift+i_shift_offset);
1393 /* Update potential energies */
1394 gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1395 gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1397 /* Increment number of inner iterations */
1398 inneriter += j_index_end - j_index_start;
1400 /* Outer loop uses 20 flops */
1403 /* Increment number of outer iterations */
1406 /* Update outer/inner flops */
1408 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_VF,outeriter*20 + inneriter*432);
1411 * Gromacs nonbonded kernel: nb_kernel_ElecEwSh_VdwLJSh_GeomW3W3_F_sse2_double
1412 * Electrostatics interaction: Ewald
1413 * VdW interaction: LennardJones
1414 * Geometry: Water3-Water3
1415 * Calculate force/pot: Force
1418 nb_kernel_ElecEwSh_VdwLJSh_GeomW3W3_F_sse2_double
1419 (t_nblist * gmx_restrict nlist,
1420 rvec * gmx_restrict xx,
1421 rvec * gmx_restrict ff,
1422 t_forcerec * gmx_restrict fr,
1423 t_mdatoms * gmx_restrict mdatoms,
1424 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1425 t_nrnb * gmx_restrict nrnb)
1427 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1428 * just 0 for non-waters.
1429 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
1430 * jnr indices corresponding to data put in the four positions in the SIMD register.
1432 int i_shift_offset,i_coord_offset,outeriter,inneriter;
1433 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1435 int j_coord_offsetA,j_coord_offsetB;
1436 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
1437 real rcutoff_scalar;
1438 real *shiftvec,*fshift,*x,*f;
1439 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1441 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1443 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1445 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1446 int vdwjidx0A,vdwjidx0B;
1447 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1448 int vdwjidx1A,vdwjidx1B;
1449 __m128d jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1450 int vdwjidx2A,vdwjidx2B;
1451 __m128d jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1452 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1453 __m128d dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1454 __m128d dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1455 __m128d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1456 __m128d dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1457 __m128d dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1458 __m128d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1459 __m128d dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1460 __m128d dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1461 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
1464 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1467 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
1468 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
1470 __m128d ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1472 __m128d dummy_mask,cutoff_mask;
1473 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
1474 __m128d one = _mm_set1_pd(1.0);
1475 __m128d two = _mm_set1_pd(2.0);
1481 jindex = nlist->jindex;
1483 shiftidx = nlist->shift;
1485 shiftvec = fr->shift_vec[0];
1486 fshift = fr->fshift[0];
1487 facel = _mm_set1_pd(fr->epsfac);
1488 charge = mdatoms->chargeA;
1489 nvdwtype = fr->ntype;
1490 vdwparam = fr->nbfp;
1491 vdwtype = mdatoms->typeA;
1493 sh_ewald = _mm_set1_pd(fr->ic->sh_ewald);
1494 ewtab = fr->ic->tabq_coul_F;
1495 ewtabscale = _mm_set1_pd(fr->ic->tabq_scale);
1496 ewtabhalfspace = _mm_set1_pd(0.5/fr->ic->tabq_scale);
1498 /* Setup water-specific parameters */
1499 inr = nlist->iinr[0];
1500 iq0 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
1501 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
1502 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
1503 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
1505 jq0 = _mm_set1_pd(charge[inr+0]);
1506 jq1 = _mm_set1_pd(charge[inr+1]);
1507 jq2 = _mm_set1_pd(charge[inr+2]);
1508 vdwjidx0A = 2*vdwtype[inr+0];
1509 qq00 = _mm_mul_pd(iq0,jq0);
1510 c6_00 = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
1511 c12_00 = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
1512 qq01 = _mm_mul_pd(iq0,jq1);
1513 qq02 = _mm_mul_pd(iq0,jq2);
1514 qq10 = _mm_mul_pd(iq1,jq0);
1515 qq11 = _mm_mul_pd(iq1,jq1);
1516 qq12 = _mm_mul_pd(iq1,jq2);
1517 qq20 = _mm_mul_pd(iq2,jq0);
1518 qq21 = _mm_mul_pd(iq2,jq1);
1519 qq22 = _mm_mul_pd(iq2,jq2);
1521 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
1522 rcutoff_scalar = fr->rcoulomb;
1523 rcutoff = _mm_set1_pd(rcutoff_scalar);
1524 rcutoff2 = _mm_mul_pd(rcutoff,rcutoff);
1526 sh_vdw_invrcut6 = _mm_set1_pd(fr->ic->sh_invrc6);
1527 rvdw = _mm_set1_pd(fr->rvdw);
1529 /* Avoid stupid compiler warnings */
1531 j_coord_offsetA = 0;
1532 j_coord_offsetB = 0;
1537 /* Start outer loop over neighborlists */
1538 for(iidx=0; iidx<nri; iidx++)
1540 /* Load shift vector for this list */
1541 i_shift_offset = DIM*shiftidx[iidx];
1543 /* Load limits for loop over neighbors */
1544 j_index_start = jindex[iidx];
1545 j_index_end = jindex[iidx+1];
1547 /* Get outer coordinate index */
1549 i_coord_offset = DIM*inr;
1551 /* Load i particle coords and add shift vector */
1552 gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1553 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1555 fix0 = _mm_setzero_pd();
1556 fiy0 = _mm_setzero_pd();
1557 fiz0 = _mm_setzero_pd();
1558 fix1 = _mm_setzero_pd();
1559 fiy1 = _mm_setzero_pd();
1560 fiz1 = _mm_setzero_pd();
1561 fix2 = _mm_setzero_pd();
1562 fiy2 = _mm_setzero_pd();
1563 fiz2 = _mm_setzero_pd();
1565 /* Start inner kernel loop */
1566 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
1569 /* Get j neighbor index, and coordinate index */
1571 jnrB = jjnr[jidx+1];
1572 j_coord_offsetA = DIM*jnrA;
1573 j_coord_offsetB = DIM*jnrB;
1575 /* load j atom coordinates */
1576 gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1577 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1579 /* Calculate displacement vector */
1580 dx00 = _mm_sub_pd(ix0,jx0);
1581 dy00 = _mm_sub_pd(iy0,jy0);
1582 dz00 = _mm_sub_pd(iz0,jz0);
1583 dx01 = _mm_sub_pd(ix0,jx1);
1584 dy01 = _mm_sub_pd(iy0,jy1);
1585 dz01 = _mm_sub_pd(iz0,jz1);
1586 dx02 = _mm_sub_pd(ix0,jx2);
1587 dy02 = _mm_sub_pd(iy0,jy2);
1588 dz02 = _mm_sub_pd(iz0,jz2);
1589 dx10 = _mm_sub_pd(ix1,jx0);
1590 dy10 = _mm_sub_pd(iy1,jy0);
1591 dz10 = _mm_sub_pd(iz1,jz0);
1592 dx11 = _mm_sub_pd(ix1,jx1);
1593 dy11 = _mm_sub_pd(iy1,jy1);
1594 dz11 = _mm_sub_pd(iz1,jz1);
1595 dx12 = _mm_sub_pd(ix1,jx2);
1596 dy12 = _mm_sub_pd(iy1,jy2);
1597 dz12 = _mm_sub_pd(iz1,jz2);
1598 dx20 = _mm_sub_pd(ix2,jx0);
1599 dy20 = _mm_sub_pd(iy2,jy0);
1600 dz20 = _mm_sub_pd(iz2,jz0);
1601 dx21 = _mm_sub_pd(ix2,jx1);
1602 dy21 = _mm_sub_pd(iy2,jy1);
1603 dz21 = _mm_sub_pd(iz2,jz1);
1604 dx22 = _mm_sub_pd(ix2,jx2);
1605 dy22 = _mm_sub_pd(iy2,jy2);
1606 dz22 = _mm_sub_pd(iz2,jz2);
1608 /* Calculate squared distance and things based on it */
1609 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1610 rsq01 = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
1611 rsq02 = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
1612 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
1613 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1614 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1615 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
1616 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1617 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1619 rinv00 = gmx_mm_invsqrt_pd(rsq00);
1620 rinv01 = gmx_mm_invsqrt_pd(rsq01);
1621 rinv02 = gmx_mm_invsqrt_pd(rsq02);
1622 rinv10 = gmx_mm_invsqrt_pd(rsq10);
1623 rinv11 = gmx_mm_invsqrt_pd(rsq11);
1624 rinv12 = gmx_mm_invsqrt_pd(rsq12);
1625 rinv20 = gmx_mm_invsqrt_pd(rsq20);
1626 rinv21 = gmx_mm_invsqrt_pd(rsq21);
1627 rinv22 = gmx_mm_invsqrt_pd(rsq22);
1629 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
1630 rinvsq01 = _mm_mul_pd(rinv01,rinv01);
1631 rinvsq02 = _mm_mul_pd(rinv02,rinv02);
1632 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
1633 rinvsq11 = _mm_mul_pd(rinv11,rinv11);
1634 rinvsq12 = _mm_mul_pd(rinv12,rinv12);
1635 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
1636 rinvsq21 = _mm_mul_pd(rinv21,rinv21);
1637 rinvsq22 = _mm_mul_pd(rinv22,rinv22);
1639 fjx0 = _mm_setzero_pd();
1640 fjy0 = _mm_setzero_pd();
1641 fjz0 = _mm_setzero_pd();
1642 fjx1 = _mm_setzero_pd();
1643 fjy1 = _mm_setzero_pd();
1644 fjz1 = _mm_setzero_pd();
1645 fjx2 = _mm_setzero_pd();
1646 fjy2 = _mm_setzero_pd();
1647 fjz2 = _mm_setzero_pd();
1649 /**************************
1650 * CALCULATE INTERACTIONS *
1651 **************************/
1653 if (gmx_mm_any_lt(rsq00,rcutoff2))
1656 r00 = _mm_mul_pd(rsq00,rinv00);
1658 /* EWALD ELECTROSTATICS */
1660 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1661 ewrt = _mm_mul_pd(r00,ewtabscale);
1662 ewitab = _mm_cvttpd_epi32(ewrt);
1663 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1664 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1666 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1667 felec = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
1669 /* LENNARD-JONES DISPERSION/REPULSION */
1671 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1672 fvdw = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
1674 cutoff_mask = _mm_cmplt_pd(rsq00,rcutoff2);
1676 fscal = _mm_add_pd(felec,fvdw);
1678 fscal = _mm_and_pd(fscal,cutoff_mask);
1680 /* Calculate temporary vectorial force */
1681 tx = _mm_mul_pd(fscal,dx00);
1682 ty = _mm_mul_pd(fscal,dy00);
1683 tz = _mm_mul_pd(fscal,dz00);
1685 /* Update vectorial force */
1686 fix0 = _mm_add_pd(fix0,tx);
1687 fiy0 = _mm_add_pd(fiy0,ty);
1688 fiz0 = _mm_add_pd(fiz0,tz);
1690 fjx0 = _mm_add_pd(fjx0,tx);
1691 fjy0 = _mm_add_pd(fjy0,ty);
1692 fjz0 = _mm_add_pd(fjz0,tz);
1696 /**************************
1697 * CALCULATE INTERACTIONS *
1698 **************************/
1700 if (gmx_mm_any_lt(rsq01,rcutoff2))
1703 r01 = _mm_mul_pd(rsq01,rinv01);
1705 /* EWALD ELECTROSTATICS */
1707 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1708 ewrt = _mm_mul_pd(r01,ewtabscale);
1709 ewitab = _mm_cvttpd_epi32(ewrt);
1710 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1711 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1713 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1714 felec = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
1716 cutoff_mask = _mm_cmplt_pd(rsq01,rcutoff2);
1720 fscal = _mm_and_pd(fscal,cutoff_mask);
1722 /* Calculate temporary vectorial force */
1723 tx = _mm_mul_pd(fscal,dx01);
1724 ty = _mm_mul_pd(fscal,dy01);
1725 tz = _mm_mul_pd(fscal,dz01);
1727 /* Update vectorial force */
1728 fix0 = _mm_add_pd(fix0,tx);
1729 fiy0 = _mm_add_pd(fiy0,ty);
1730 fiz0 = _mm_add_pd(fiz0,tz);
1732 fjx1 = _mm_add_pd(fjx1,tx);
1733 fjy1 = _mm_add_pd(fjy1,ty);
1734 fjz1 = _mm_add_pd(fjz1,tz);
1738 /**************************
1739 * CALCULATE INTERACTIONS *
1740 **************************/
1742 if (gmx_mm_any_lt(rsq02,rcutoff2))
1745 r02 = _mm_mul_pd(rsq02,rinv02);
1747 /* EWALD ELECTROSTATICS */
1749 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1750 ewrt = _mm_mul_pd(r02,ewtabscale);
1751 ewitab = _mm_cvttpd_epi32(ewrt);
1752 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1753 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1755 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1756 felec = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
1758 cutoff_mask = _mm_cmplt_pd(rsq02,rcutoff2);
1762 fscal = _mm_and_pd(fscal,cutoff_mask);
1764 /* Calculate temporary vectorial force */
1765 tx = _mm_mul_pd(fscal,dx02);
1766 ty = _mm_mul_pd(fscal,dy02);
1767 tz = _mm_mul_pd(fscal,dz02);
1769 /* Update vectorial force */
1770 fix0 = _mm_add_pd(fix0,tx);
1771 fiy0 = _mm_add_pd(fiy0,ty);
1772 fiz0 = _mm_add_pd(fiz0,tz);
1774 fjx2 = _mm_add_pd(fjx2,tx);
1775 fjy2 = _mm_add_pd(fjy2,ty);
1776 fjz2 = _mm_add_pd(fjz2,tz);
1780 /**************************
1781 * CALCULATE INTERACTIONS *
1782 **************************/
1784 if (gmx_mm_any_lt(rsq10,rcutoff2))
1787 r10 = _mm_mul_pd(rsq10,rinv10);
1789 /* EWALD ELECTROSTATICS */
1791 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1792 ewrt = _mm_mul_pd(r10,ewtabscale);
1793 ewitab = _mm_cvttpd_epi32(ewrt);
1794 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1795 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
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(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
1800 cutoff_mask = _mm_cmplt_pd(rsq10,rcutoff2);
1804 fscal = _mm_and_pd(fscal,cutoff_mask);
1806 /* Calculate temporary vectorial force */
1807 tx = _mm_mul_pd(fscal,dx10);
1808 ty = _mm_mul_pd(fscal,dy10);
1809 tz = _mm_mul_pd(fscal,dz10);
1811 /* Update vectorial force */
1812 fix1 = _mm_add_pd(fix1,tx);
1813 fiy1 = _mm_add_pd(fiy1,ty);
1814 fiz1 = _mm_add_pd(fiz1,tz);
1816 fjx0 = _mm_add_pd(fjx0,tx);
1817 fjy0 = _mm_add_pd(fjy0,ty);
1818 fjz0 = _mm_add_pd(fjz0,tz);
1822 /**************************
1823 * CALCULATE INTERACTIONS *
1824 **************************/
1826 if (gmx_mm_any_lt(rsq11,rcutoff2))
1829 r11 = _mm_mul_pd(rsq11,rinv11);
1831 /* EWALD ELECTROSTATICS */
1833 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1834 ewrt = _mm_mul_pd(r11,ewtabscale);
1835 ewitab = _mm_cvttpd_epi32(ewrt);
1836 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1837 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1839 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1840 felec = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
1842 cutoff_mask = _mm_cmplt_pd(rsq11,rcutoff2);
1846 fscal = _mm_and_pd(fscal,cutoff_mask);
1848 /* Calculate temporary vectorial force */
1849 tx = _mm_mul_pd(fscal,dx11);
1850 ty = _mm_mul_pd(fscal,dy11);
1851 tz = _mm_mul_pd(fscal,dz11);
1853 /* Update vectorial force */
1854 fix1 = _mm_add_pd(fix1,tx);
1855 fiy1 = _mm_add_pd(fiy1,ty);
1856 fiz1 = _mm_add_pd(fiz1,tz);
1858 fjx1 = _mm_add_pd(fjx1,tx);
1859 fjy1 = _mm_add_pd(fjy1,ty);
1860 fjz1 = _mm_add_pd(fjz1,tz);
1864 /**************************
1865 * CALCULATE INTERACTIONS *
1866 **************************/
1868 if (gmx_mm_any_lt(rsq12,rcutoff2))
1871 r12 = _mm_mul_pd(rsq12,rinv12);
1873 /* EWALD ELECTROSTATICS */
1875 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1876 ewrt = _mm_mul_pd(r12,ewtabscale);
1877 ewitab = _mm_cvttpd_epi32(ewrt);
1878 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1879 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1881 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1882 felec = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
1884 cutoff_mask = _mm_cmplt_pd(rsq12,rcutoff2);
1888 fscal = _mm_and_pd(fscal,cutoff_mask);
1890 /* Calculate temporary vectorial force */
1891 tx = _mm_mul_pd(fscal,dx12);
1892 ty = _mm_mul_pd(fscal,dy12);
1893 tz = _mm_mul_pd(fscal,dz12);
1895 /* Update vectorial force */
1896 fix1 = _mm_add_pd(fix1,tx);
1897 fiy1 = _mm_add_pd(fiy1,ty);
1898 fiz1 = _mm_add_pd(fiz1,tz);
1900 fjx2 = _mm_add_pd(fjx2,tx);
1901 fjy2 = _mm_add_pd(fjy2,ty);
1902 fjz2 = _mm_add_pd(fjz2,tz);
1906 /**************************
1907 * CALCULATE INTERACTIONS *
1908 **************************/
1910 if (gmx_mm_any_lt(rsq20,rcutoff2))
1913 r20 = _mm_mul_pd(rsq20,rinv20);
1915 /* EWALD ELECTROSTATICS */
1917 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1918 ewrt = _mm_mul_pd(r20,ewtabscale);
1919 ewitab = _mm_cvttpd_epi32(ewrt);
1920 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1921 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1923 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1924 felec = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
1926 cutoff_mask = _mm_cmplt_pd(rsq20,rcutoff2);
1930 fscal = _mm_and_pd(fscal,cutoff_mask);
1932 /* Calculate temporary vectorial force */
1933 tx = _mm_mul_pd(fscal,dx20);
1934 ty = _mm_mul_pd(fscal,dy20);
1935 tz = _mm_mul_pd(fscal,dz20);
1937 /* Update vectorial force */
1938 fix2 = _mm_add_pd(fix2,tx);
1939 fiy2 = _mm_add_pd(fiy2,ty);
1940 fiz2 = _mm_add_pd(fiz2,tz);
1942 fjx0 = _mm_add_pd(fjx0,tx);
1943 fjy0 = _mm_add_pd(fjy0,ty);
1944 fjz0 = _mm_add_pd(fjz0,tz);
1948 /**************************
1949 * CALCULATE INTERACTIONS *
1950 **************************/
1952 if (gmx_mm_any_lt(rsq21,rcutoff2))
1955 r21 = _mm_mul_pd(rsq21,rinv21);
1957 /* EWALD ELECTROSTATICS */
1959 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1960 ewrt = _mm_mul_pd(r21,ewtabscale);
1961 ewitab = _mm_cvttpd_epi32(ewrt);
1962 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1963 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1965 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1966 felec = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
1968 cutoff_mask = _mm_cmplt_pd(rsq21,rcutoff2);
1972 fscal = _mm_and_pd(fscal,cutoff_mask);
1974 /* Calculate temporary vectorial force */
1975 tx = _mm_mul_pd(fscal,dx21);
1976 ty = _mm_mul_pd(fscal,dy21);
1977 tz = _mm_mul_pd(fscal,dz21);
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 fjx1 = _mm_add_pd(fjx1,tx);
1985 fjy1 = _mm_add_pd(fjy1,ty);
1986 fjz1 = _mm_add_pd(fjz1,tz);
1990 /**************************
1991 * CALCULATE INTERACTIONS *
1992 **************************/
1994 if (gmx_mm_any_lt(rsq22,rcutoff2))
1997 r22 = _mm_mul_pd(rsq22,rinv22);
1999 /* EWALD ELECTROSTATICS */
2001 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2002 ewrt = _mm_mul_pd(r22,ewtabscale);
2003 ewitab = _mm_cvttpd_epi32(ewrt);
2004 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2005 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
2007 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2008 felec = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
2010 cutoff_mask = _mm_cmplt_pd(rsq22,rcutoff2);
2014 fscal = _mm_and_pd(fscal,cutoff_mask);
2016 /* Calculate temporary vectorial force */
2017 tx = _mm_mul_pd(fscal,dx22);
2018 ty = _mm_mul_pd(fscal,dy22);
2019 tz = _mm_mul_pd(fscal,dz22);
2021 /* Update vectorial force */
2022 fix2 = _mm_add_pd(fix2,tx);
2023 fiy2 = _mm_add_pd(fiy2,ty);
2024 fiz2 = _mm_add_pd(fiz2,tz);
2026 fjx2 = _mm_add_pd(fjx2,tx);
2027 fjy2 = _mm_add_pd(fjy2,ty);
2028 fjz2 = _mm_add_pd(fjz2,tz);
2032 gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
2034 /* Inner loop uses 358 flops */
2037 if(jidx<j_index_end)
2041 j_coord_offsetA = DIM*jnrA;
2043 /* load j atom coordinates */
2044 gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
2045 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
2047 /* Calculate displacement vector */
2048 dx00 = _mm_sub_pd(ix0,jx0);
2049 dy00 = _mm_sub_pd(iy0,jy0);
2050 dz00 = _mm_sub_pd(iz0,jz0);
2051 dx01 = _mm_sub_pd(ix0,jx1);
2052 dy01 = _mm_sub_pd(iy0,jy1);
2053 dz01 = _mm_sub_pd(iz0,jz1);
2054 dx02 = _mm_sub_pd(ix0,jx2);
2055 dy02 = _mm_sub_pd(iy0,jy2);
2056 dz02 = _mm_sub_pd(iz0,jz2);
2057 dx10 = _mm_sub_pd(ix1,jx0);
2058 dy10 = _mm_sub_pd(iy1,jy0);
2059 dz10 = _mm_sub_pd(iz1,jz0);
2060 dx11 = _mm_sub_pd(ix1,jx1);
2061 dy11 = _mm_sub_pd(iy1,jy1);
2062 dz11 = _mm_sub_pd(iz1,jz1);
2063 dx12 = _mm_sub_pd(ix1,jx2);
2064 dy12 = _mm_sub_pd(iy1,jy2);
2065 dz12 = _mm_sub_pd(iz1,jz2);
2066 dx20 = _mm_sub_pd(ix2,jx0);
2067 dy20 = _mm_sub_pd(iy2,jy0);
2068 dz20 = _mm_sub_pd(iz2,jz0);
2069 dx21 = _mm_sub_pd(ix2,jx1);
2070 dy21 = _mm_sub_pd(iy2,jy1);
2071 dz21 = _mm_sub_pd(iz2,jz1);
2072 dx22 = _mm_sub_pd(ix2,jx2);
2073 dy22 = _mm_sub_pd(iy2,jy2);
2074 dz22 = _mm_sub_pd(iz2,jz2);
2076 /* Calculate squared distance and things based on it */
2077 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
2078 rsq01 = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
2079 rsq02 = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
2080 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
2081 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
2082 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
2083 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
2084 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
2085 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
2087 rinv00 = gmx_mm_invsqrt_pd(rsq00);
2088 rinv01 = gmx_mm_invsqrt_pd(rsq01);
2089 rinv02 = gmx_mm_invsqrt_pd(rsq02);
2090 rinv10 = gmx_mm_invsqrt_pd(rsq10);
2091 rinv11 = gmx_mm_invsqrt_pd(rsq11);
2092 rinv12 = gmx_mm_invsqrt_pd(rsq12);
2093 rinv20 = gmx_mm_invsqrt_pd(rsq20);
2094 rinv21 = gmx_mm_invsqrt_pd(rsq21);
2095 rinv22 = gmx_mm_invsqrt_pd(rsq22);
2097 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
2098 rinvsq01 = _mm_mul_pd(rinv01,rinv01);
2099 rinvsq02 = _mm_mul_pd(rinv02,rinv02);
2100 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
2101 rinvsq11 = _mm_mul_pd(rinv11,rinv11);
2102 rinvsq12 = _mm_mul_pd(rinv12,rinv12);
2103 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
2104 rinvsq21 = _mm_mul_pd(rinv21,rinv21);
2105 rinvsq22 = _mm_mul_pd(rinv22,rinv22);
2107 fjx0 = _mm_setzero_pd();
2108 fjy0 = _mm_setzero_pd();
2109 fjz0 = _mm_setzero_pd();
2110 fjx1 = _mm_setzero_pd();
2111 fjy1 = _mm_setzero_pd();
2112 fjz1 = _mm_setzero_pd();
2113 fjx2 = _mm_setzero_pd();
2114 fjy2 = _mm_setzero_pd();
2115 fjz2 = _mm_setzero_pd();
2117 /**************************
2118 * CALCULATE INTERACTIONS *
2119 **************************/
2121 if (gmx_mm_any_lt(rsq00,rcutoff2))
2124 r00 = _mm_mul_pd(rsq00,rinv00);
2126 /* EWALD ELECTROSTATICS */
2128 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2129 ewrt = _mm_mul_pd(r00,ewtabscale);
2130 ewitab = _mm_cvttpd_epi32(ewrt);
2131 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2132 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2133 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2134 felec = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
2136 /* LENNARD-JONES DISPERSION/REPULSION */
2138 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
2139 fvdw = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
2141 cutoff_mask = _mm_cmplt_pd(rsq00,rcutoff2);
2143 fscal = _mm_add_pd(felec,fvdw);
2145 fscal = _mm_and_pd(fscal,cutoff_mask);
2147 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2149 /* Calculate temporary vectorial force */
2150 tx = _mm_mul_pd(fscal,dx00);
2151 ty = _mm_mul_pd(fscal,dy00);
2152 tz = _mm_mul_pd(fscal,dz00);
2154 /* Update vectorial force */
2155 fix0 = _mm_add_pd(fix0,tx);
2156 fiy0 = _mm_add_pd(fiy0,ty);
2157 fiz0 = _mm_add_pd(fiz0,tz);
2159 fjx0 = _mm_add_pd(fjx0,tx);
2160 fjy0 = _mm_add_pd(fjy0,ty);
2161 fjz0 = _mm_add_pd(fjz0,tz);
2165 /**************************
2166 * CALCULATE INTERACTIONS *
2167 **************************/
2169 if (gmx_mm_any_lt(rsq01,rcutoff2))
2172 r01 = _mm_mul_pd(rsq01,rinv01);
2174 /* EWALD ELECTROSTATICS */
2176 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2177 ewrt = _mm_mul_pd(r01,ewtabscale);
2178 ewitab = _mm_cvttpd_epi32(ewrt);
2179 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2180 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2181 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2182 felec = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
2184 cutoff_mask = _mm_cmplt_pd(rsq01,rcutoff2);
2188 fscal = _mm_and_pd(fscal,cutoff_mask);
2190 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2192 /* Calculate temporary vectorial force */
2193 tx = _mm_mul_pd(fscal,dx01);
2194 ty = _mm_mul_pd(fscal,dy01);
2195 tz = _mm_mul_pd(fscal,dz01);
2197 /* Update vectorial force */
2198 fix0 = _mm_add_pd(fix0,tx);
2199 fiy0 = _mm_add_pd(fiy0,ty);
2200 fiz0 = _mm_add_pd(fiz0,tz);
2202 fjx1 = _mm_add_pd(fjx1,tx);
2203 fjy1 = _mm_add_pd(fjy1,ty);
2204 fjz1 = _mm_add_pd(fjz1,tz);
2208 /**************************
2209 * CALCULATE INTERACTIONS *
2210 **************************/
2212 if (gmx_mm_any_lt(rsq02,rcutoff2))
2215 r02 = _mm_mul_pd(rsq02,rinv02);
2217 /* EWALD ELECTROSTATICS */
2219 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2220 ewrt = _mm_mul_pd(r02,ewtabscale);
2221 ewitab = _mm_cvttpd_epi32(ewrt);
2222 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2223 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2224 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2225 felec = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
2227 cutoff_mask = _mm_cmplt_pd(rsq02,rcutoff2);
2231 fscal = _mm_and_pd(fscal,cutoff_mask);
2233 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2235 /* Calculate temporary vectorial force */
2236 tx = _mm_mul_pd(fscal,dx02);
2237 ty = _mm_mul_pd(fscal,dy02);
2238 tz = _mm_mul_pd(fscal,dz02);
2240 /* Update vectorial force */
2241 fix0 = _mm_add_pd(fix0,tx);
2242 fiy0 = _mm_add_pd(fiy0,ty);
2243 fiz0 = _mm_add_pd(fiz0,tz);
2245 fjx2 = _mm_add_pd(fjx2,tx);
2246 fjy2 = _mm_add_pd(fjy2,ty);
2247 fjz2 = _mm_add_pd(fjz2,tz);
2251 /**************************
2252 * CALCULATE INTERACTIONS *
2253 **************************/
2255 if (gmx_mm_any_lt(rsq10,rcutoff2))
2258 r10 = _mm_mul_pd(rsq10,rinv10);
2260 /* EWALD ELECTROSTATICS */
2262 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2263 ewrt = _mm_mul_pd(r10,ewtabscale);
2264 ewitab = _mm_cvttpd_epi32(ewrt);
2265 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2266 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2267 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2268 felec = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
2270 cutoff_mask = _mm_cmplt_pd(rsq10,rcutoff2);
2274 fscal = _mm_and_pd(fscal,cutoff_mask);
2276 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2278 /* Calculate temporary vectorial force */
2279 tx = _mm_mul_pd(fscal,dx10);
2280 ty = _mm_mul_pd(fscal,dy10);
2281 tz = _mm_mul_pd(fscal,dz10);
2283 /* Update vectorial force */
2284 fix1 = _mm_add_pd(fix1,tx);
2285 fiy1 = _mm_add_pd(fiy1,ty);
2286 fiz1 = _mm_add_pd(fiz1,tz);
2288 fjx0 = _mm_add_pd(fjx0,tx);
2289 fjy0 = _mm_add_pd(fjy0,ty);
2290 fjz0 = _mm_add_pd(fjz0,tz);
2294 /**************************
2295 * CALCULATE INTERACTIONS *
2296 **************************/
2298 if (gmx_mm_any_lt(rsq11,rcutoff2))
2301 r11 = _mm_mul_pd(rsq11,rinv11);
2303 /* EWALD ELECTROSTATICS */
2305 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2306 ewrt = _mm_mul_pd(r11,ewtabscale);
2307 ewitab = _mm_cvttpd_epi32(ewrt);
2308 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2309 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2310 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2311 felec = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
2313 cutoff_mask = _mm_cmplt_pd(rsq11,rcutoff2);
2317 fscal = _mm_and_pd(fscal,cutoff_mask);
2319 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2321 /* Calculate temporary vectorial force */
2322 tx = _mm_mul_pd(fscal,dx11);
2323 ty = _mm_mul_pd(fscal,dy11);
2324 tz = _mm_mul_pd(fscal,dz11);
2326 /* Update vectorial force */
2327 fix1 = _mm_add_pd(fix1,tx);
2328 fiy1 = _mm_add_pd(fiy1,ty);
2329 fiz1 = _mm_add_pd(fiz1,tz);
2331 fjx1 = _mm_add_pd(fjx1,tx);
2332 fjy1 = _mm_add_pd(fjy1,ty);
2333 fjz1 = _mm_add_pd(fjz1,tz);
2337 /**************************
2338 * CALCULATE INTERACTIONS *
2339 **************************/
2341 if (gmx_mm_any_lt(rsq12,rcutoff2))
2344 r12 = _mm_mul_pd(rsq12,rinv12);
2346 /* EWALD ELECTROSTATICS */
2348 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2349 ewrt = _mm_mul_pd(r12,ewtabscale);
2350 ewitab = _mm_cvttpd_epi32(ewrt);
2351 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2352 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2353 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2354 felec = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
2356 cutoff_mask = _mm_cmplt_pd(rsq12,rcutoff2);
2360 fscal = _mm_and_pd(fscal,cutoff_mask);
2362 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2364 /* Calculate temporary vectorial force */
2365 tx = _mm_mul_pd(fscal,dx12);
2366 ty = _mm_mul_pd(fscal,dy12);
2367 tz = _mm_mul_pd(fscal,dz12);
2369 /* Update vectorial force */
2370 fix1 = _mm_add_pd(fix1,tx);
2371 fiy1 = _mm_add_pd(fiy1,ty);
2372 fiz1 = _mm_add_pd(fiz1,tz);
2374 fjx2 = _mm_add_pd(fjx2,tx);
2375 fjy2 = _mm_add_pd(fjy2,ty);
2376 fjz2 = _mm_add_pd(fjz2,tz);
2380 /**************************
2381 * CALCULATE INTERACTIONS *
2382 **************************/
2384 if (gmx_mm_any_lt(rsq20,rcutoff2))
2387 r20 = _mm_mul_pd(rsq20,rinv20);
2389 /* EWALD ELECTROSTATICS */
2391 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2392 ewrt = _mm_mul_pd(r20,ewtabscale);
2393 ewitab = _mm_cvttpd_epi32(ewrt);
2394 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2395 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2396 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2397 felec = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
2399 cutoff_mask = _mm_cmplt_pd(rsq20,rcutoff2);
2403 fscal = _mm_and_pd(fscal,cutoff_mask);
2405 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2407 /* Calculate temporary vectorial force */
2408 tx = _mm_mul_pd(fscal,dx20);
2409 ty = _mm_mul_pd(fscal,dy20);
2410 tz = _mm_mul_pd(fscal,dz20);
2412 /* Update vectorial force */
2413 fix2 = _mm_add_pd(fix2,tx);
2414 fiy2 = _mm_add_pd(fiy2,ty);
2415 fiz2 = _mm_add_pd(fiz2,tz);
2417 fjx0 = _mm_add_pd(fjx0,tx);
2418 fjy0 = _mm_add_pd(fjy0,ty);
2419 fjz0 = _mm_add_pd(fjz0,tz);
2423 /**************************
2424 * CALCULATE INTERACTIONS *
2425 **************************/
2427 if (gmx_mm_any_lt(rsq21,rcutoff2))
2430 r21 = _mm_mul_pd(rsq21,rinv21);
2432 /* EWALD ELECTROSTATICS */
2434 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2435 ewrt = _mm_mul_pd(r21,ewtabscale);
2436 ewitab = _mm_cvttpd_epi32(ewrt);
2437 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2438 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2439 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2440 felec = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
2442 cutoff_mask = _mm_cmplt_pd(rsq21,rcutoff2);
2446 fscal = _mm_and_pd(fscal,cutoff_mask);
2448 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2450 /* Calculate temporary vectorial force */
2451 tx = _mm_mul_pd(fscal,dx21);
2452 ty = _mm_mul_pd(fscal,dy21);
2453 tz = _mm_mul_pd(fscal,dz21);
2455 /* Update vectorial force */
2456 fix2 = _mm_add_pd(fix2,tx);
2457 fiy2 = _mm_add_pd(fiy2,ty);
2458 fiz2 = _mm_add_pd(fiz2,tz);
2460 fjx1 = _mm_add_pd(fjx1,tx);
2461 fjy1 = _mm_add_pd(fjy1,ty);
2462 fjz1 = _mm_add_pd(fjz1,tz);
2466 /**************************
2467 * CALCULATE INTERACTIONS *
2468 **************************/
2470 if (gmx_mm_any_lt(rsq22,rcutoff2))
2473 r22 = _mm_mul_pd(rsq22,rinv22);
2475 /* EWALD ELECTROSTATICS */
2477 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2478 ewrt = _mm_mul_pd(r22,ewtabscale);
2479 ewitab = _mm_cvttpd_epi32(ewrt);
2480 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2481 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2482 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2483 felec = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
2485 cutoff_mask = _mm_cmplt_pd(rsq22,rcutoff2);
2489 fscal = _mm_and_pd(fscal,cutoff_mask);
2491 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2493 /* Calculate temporary vectorial force */
2494 tx = _mm_mul_pd(fscal,dx22);
2495 ty = _mm_mul_pd(fscal,dy22);
2496 tz = _mm_mul_pd(fscal,dz22);
2498 /* Update vectorial force */
2499 fix2 = _mm_add_pd(fix2,tx);
2500 fiy2 = _mm_add_pd(fiy2,ty);
2501 fiz2 = _mm_add_pd(fiz2,tz);
2503 fjx2 = _mm_add_pd(fjx2,tx);
2504 fjy2 = _mm_add_pd(fjy2,ty);
2505 fjz2 = _mm_add_pd(fjz2,tz);
2509 gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
2511 /* Inner loop uses 358 flops */
2514 /* End of innermost loop */
2516 gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
2517 f+i_coord_offset,fshift+i_shift_offset);
2519 /* Increment number of inner iterations */
2520 inneriter += j_index_end - j_index_start;
2522 /* Outer loop uses 18 flops */
2525 /* Increment number of outer iterations */
2528 /* Update outer/inner flops */
2530 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_F,outeriter*18 + inneriter*358);