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 avx_128_fma_double kernel generator.
44 #include "../nb_kernel.h"
45 #include "types/simple.h"
49 #include "gromacs/simd/math_x86_avx_128_fma_double.h"
50 #include "kernelutil_x86_avx_128_fma_double.h"
53 * Gromacs nonbonded kernel: nb_kernel_ElecEwSh_VdwLJEwSh_GeomW4W4_VF_avx_128_fma_double
54 * Electrostatics interaction: Ewald
55 * VdW interaction: LJEwald
56 * Geometry: Water4-Water4
57 * Calculate force/pot: PotentialAndForce
60 nb_kernel_ElecEwSh_VdwLJEwSh_GeomW4W4_VF_avx_128_fma_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;
89 __m128d ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
90 int vdwjidx0A,vdwjidx0B;
91 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
92 int vdwjidx1A,vdwjidx1B;
93 __m128d jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
94 int vdwjidx2A,vdwjidx2B;
95 __m128d jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
96 int vdwjidx3A,vdwjidx3B;
97 __m128d jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
98 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
99 __m128d dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
100 __m128d dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
101 __m128d dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
102 __m128d dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
103 __m128d dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
104 __m128d dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
105 __m128d dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
106 __m128d dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
107 __m128d dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
108 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
111 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
114 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
115 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
127 __m128d ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
128 __m128d one_half = _mm_set1_pd(0.5);
129 __m128d minus_one = _mm_set1_pd(-1.0);
131 __m128d ewtabscale,eweps,twoeweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
133 __m128d dummy_mask,cutoff_mask;
134 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
135 __m128d one = _mm_set1_pd(1.0);
136 __m128d two = _mm_set1_pd(2.0);
142 jindex = nlist->jindex;
144 shiftidx = nlist->shift;
146 shiftvec = fr->shift_vec[0];
147 fshift = fr->fshift[0];
148 facel = _mm_set1_pd(fr->epsfac);
149 charge = mdatoms->chargeA;
150 nvdwtype = fr->ntype;
152 vdwtype = mdatoms->typeA;
153 vdwgridparam = fr->ljpme_c6grid;
154 sh_lj_ewald = _mm_set1_pd(fr->ic->sh_lj_ewald);
155 ewclj = _mm_set1_pd(fr->ewaldcoeff_lj);
156 ewclj2 = _mm_mul_pd(minus_one,_mm_mul_pd(ewclj,ewclj));
158 sh_ewald = _mm_set1_pd(fr->ic->sh_ewald);
159 ewtab = fr->ic->tabq_coul_FDV0;
160 ewtabscale = _mm_set1_pd(fr->ic->tabq_scale);
161 ewtabhalfspace = _mm_set1_pd(0.5/fr->ic->tabq_scale);
163 /* Setup water-specific parameters */
164 inr = nlist->iinr[0];
165 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
166 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
167 iq3 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
168 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
170 jq1 = _mm_set1_pd(charge[inr+1]);
171 jq2 = _mm_set1_pd(charge[inr+2]);
172 jq3 = _mm_set1_pd(charge[inr+3]);
173 vdwjidx0A = 2*vdwtype[inr+0];
174 c6_00 = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
175 c12_00 = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
176 c6grid_00 = _mm_set1_pd(vdwgridparam[vdwioffset0+vdwjidx0A]);
177 qq11 = _mm_mul_pd(iq1,jq1);
178 qq12 = _mm_mul_pd(iq1,jq2);
179 qq13 = _mm_mul_pd(iq1,jq3);
180 qq21 = _mm_mul_pd(iq2,jq1);
181 qq22 = _mm_mul_pd(iq2,jq2);
182 qq23 = _mm_mul_pd(iq2,jq3);
183 qq31 = _mm_mul_pd(iq3,jq1);
184 qq32 = _mm_mul_pd(iq3,jq2);
185 qq33 = _mm_mul_pd(iq3,jq3);
187 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
188 rcutoff_scalar = fr->rcoulomb;
189 rcutoff = _mm_set1_pd(rcutoff_scalar);
190 rcutoff2 = _mm_mul_pd(rcutoff,rcutoff);
192 sh_vdw_invrcut6 = _mm_set1_pd(fr->ic->sh_invrc6);
193 rvdw = _mm_set1_pd(fr->rvdw);
195 /* Avoid stupid compiler warnings */
203 /* Start outer loop over neighborlists */
204 for(iidx=0; iidx<nri; iidx++)
206 /* Load shift vector for this list */
207 i_shift_offset = DIM*shiftidx[iidx];
209 /* Load limits for loop over neighbors */
210 j_index_start = jindex[iidx];
211 j_index_end = jindex[iidx+1];
213 /* Get outer coordinate index */
215 i_coord_offset = DIM*inr;
217 /* Load i particle coords and add shift vector */
218 gmx_mm_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
219 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
221 fix0 = _mm_setzero_pd();
222 fiy0 = _mm_setzero_pd();
223 fiz0 = _mm_setzero_pd();
224 fix1 = _mm_setzero_pd();
225 fiy1 = _mm_setzero_pd();
226 fiz1 = _mm_setzero_pd();
227 fix2 = _mm_setzero_pd();
228 fiy2 = _mm_setzero_pd();
229 fiz2 = _mm_setzero_pd();
230 fix3 = _mm_setzero_pd();
231 fiy3 = _mm_setzero_pd();
232 fiz3 = _mm_setzero_pd();
234 /* Reset potential sums */
235 velecsum = _mm_setzero_pd();
236 vvdwsum = _mm_setzero_pd();
238 /* Start inner kernel loop */
239 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
242 /* Get j neighbor index, and coordinate index */
245 j_coord_offsetA = DIM*jnrA;
246 j_coord_offsetB = DIM*jnrB;
248 /* load j atom coordinates */
249 gmx_mm_load_4rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
250 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
251 &jy2,&jz2,&jx3,&jy3,&jz3);
253 /* Calculate displacement vector */
254 dx00 = _mm_sub_pd(ix0,jx0);
255 dy00 = _mm_sub_pd(iy0,jy0);
256 dz00 = _mm_sub_pd(iz0,jz0);
257 dx11 = _mm_sub_pd(ix1,jx1);
258 dy11 = _mm_sub_pd(iy1,jy1);
259 dz11 = _mm_sub_pd(iz1,jz1);
260 dx12 = _mm_sub_pd(ix1,jx2);
261 dy12 = _mm_sub_pd(iy1,jy2);
262 dz12 = _mm_sub_pd(iz1,jz2);
263 dx13 = _mm_sub_pd(ix1,jx3);
264 dy13 = _mm_sub_pd(iy1,jy3);
265 dz13 = _mm_sub_pd(iz1,jz3);
266 dx21 = _mm_sub_pd(ix2,jx1);
267 dy21 = _mm_sub_pd(iy2,jy1);
268 dz21 = _mm_sub_pd(iz2,jz1);
269 dx22 = _mm_sub_pd(ix2,jx2);
270 dy22 = _mm_sub_pd(iy2,jy2);
271 dz22 = _mm_sub_pd(iz2,jz2);
272 dx23 = _mm_sub_pd(ix2,jx3);
273 dy23 = _mm_sub_pd(iy2,jy3);
274 dz23 = _mm_sub_pd(iz2,jz3);
275 dx31 = _mm_sub_pd(ix3,jx1);
276 dy31 = _mm_sub_pd(iy3,jy1);
277 dz31 = _mm_sub_pd(iz3,jz1);
278 dx32 = _mm_sub_pd(ix3,jx2);
279 dy32 = _mm_sub_pd(iy3,jy2);
280 dz32 = _mm_sub_pd(iz3,jz2);
281 dx33 = _mm_sub_pd(ix3,jx3);
282 dy33 = _mm_sub_pd(iy3,jy3);
283 dz33 = _mm_sub_pd(iz3,jz3);
285 /* Calculate squared distance and things based on it */
286 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
287 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
288 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
289 rsq13 = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
290 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
291 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
292 rsq23 = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
293 rsq31 = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
294 rsq32 = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
295 rsq33 = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
297 rinv00 = gmx_mm_invsqrt_pd(rsq00);
298 rinv11 = gmx_mm_invsqrt_pd(rsq11);
299 rinv12 = gmx_mm_invsqrt_pd(rsq12);
300 rinv13 = gmx_mm_invsqrt_pd(rsq13);
301 rinv21 = gmx_mm_invsqrt_pd(rsq21);
302 rinv22 = gmx_mm_invsqrt_pd(rsq22);
303 rinv23 = gmx_mm_invsqrt_pd(rsq23);
304 rinv31 = gmx_mm_invsqrt_pd(rsq31);
305 rinv32 = gmx_mm_invsqrt_pd(rsq32);
306 rinv33 = gmx_mm_invsqrt_pd(rsq33);
308 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
309 rinvsq11 = _mm_mul_pd(rinv11,rinv11);
310 rinvsq12 = _mm_mul_pd(rinv12,rinv12);
311 rinvsq13 = _mm_mul_pd(rinv13,rinv13);
312 rinvsq21 = _mm_mul_pd(rinv21,rinv21);
313 rinvsq22 = _mm_mul_pd(rinv22,rinv22);
314 rinvsq23 = _mm_mul_pd(rinv23,rinv23);
315 rinvsq31 = _mm_mul_pd(rinv31,rinv31);
316 rinvsq32 = _mm_mul_pd(rinv32,rinv32);
317 rinvsq33 = _mm_mul_pd(rinv33,rinv33);
319 fjx0 = _mm_setzero_pd();
320 fjy0 = _mm_setzero_pd();
321 fjz0 = _mm_setzero_pd();
322 fjx1 = _mm_setzero_pd();
323 fjy1 = _mm_setzero_pd();
324 fjz1 = _mm_setzero_pd();
325 fjx2 = _mm_setzero_pd();
326 fjy2 = _mm_setzero_pd();
327 fjz2 = _mm_setzero_pd();
328 fjx3 = _mm_setzero_pd();
329 fjy3 = _mm_setzero_pd();
330 fjz3 = _mm_setzero_pd();
332 /**************************
333 * CALCULATE INTERACTIONS *
334 **************************/
336 if (gmx_mm_any_lt(rsq00,rcutoff2))
339 r00 = _mm_mul_pd(rsq00,rinv00);
341 /* Analytical LJ-PME */
342 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
343 ewcljrsq = _mm_mul_pd(ewclj2,rsq00);
344 ewclj6 = _mm_mul_pd(ewclj2,_mm_mul_pd(ewclj2,ewclj2));
345 exponent = gmx_simd_exp_d(ewcljrsq);
346 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
347 poly = _mm_mul_pd(exponent,_mm_macc_pd(_mm_mul_pd(ewcljrsq,ewcljrsq),one_half,_mm_sub_pd(one,ewcljrsq)));
348 /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
349 vvdw6 = _mm_mul_pd(_mm_macc_pd(-c6grid_00,_mm_sub_pd(one,poly),c6_00),rinvsix);
350 vvdw12 = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
351 vvdw = _mm_msub_pd(_mm_nmacc_pd(c12_00,_mm_mul_pd(sh_vdw_invrcut6,sh_vdw_invrcut6),vvdw12),one_twelfth,
352 _mm_mul_pd(_mm_sub_pd(vvdw6,_mm_macc_pd(c6grid_00,sh_lj_ewald,_mm_mul_pd(c6_00,sh_vdw_invrcut6))),one_sixth));
353 /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
354 fvdw = _mm_mul_pd(_mm_add_pd(vvdw12,_mm_msub_pd(_mm_mul_pd(c6grid_00,one_sixth),_mm_mul_pd(exponent,ewclj6),vvdw6)),rinvsq00);
356 cutoff_mask = _mm_cmplt_pd(rsq00,rcutoff2);
358 /* Update potential sum for this i atom from the interaction with this j atom. */
359 vvdw = _mm_and_pd(vvdw,cutoff_mask);
360 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
364 fscal = _mm_and_pd(fscal,cutoff_mask);
366 /* Update vectorial force */
367 fix0 = _mm_macc_pd(dx00,fscal,fix0);
368 fiy0 = _mm_macc_pd(dy00,fscal,fiy0);
369 fiz0 = _mm_macc_pd(dz00,fscal,fiz0);
371 fjx0 = _mm_macc_pd(dx00,fscal,fjx0);
372 fjy0 = _mm_macc_pd(dy00,fscal,fjy0);
373 fjz0 = _mm_macc_pd(dz00,fscal,fjz0);
377 /**************************
378 * CALCULATE INTERACTIONS *
379 **************************/
381 if (gmx_mm_any_lt(rsq11,rcutoff2))
384 r11 = _mm_mul_pd(rsq11,rinv11);
386 /* EWALD ELECTROSTATICS */
388 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
389 ewrt = _mm_mul_pd(r11,ewtabscale);
390 ewitab = _mm_cvttpd_epi32(ewrt);
392 eweps = _mm_frcz_pd(ewrt);
394 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
396 twoeweps = _mm_add_pd(eweps,eweps);
397 ewitab = _mm_slli_epi32(ewitab,2);
398 ewtabF = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
399 ewtabD = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
400 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
401 ewtabV = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
402 ewtabFn = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
403 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
404 felec = _mm_macc_pd(eweps,ewtabD,ewtabF);
405 velec = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
406 velec = _mm_mul_pd(qq11,_mm_sub_pd(_mm_sub_pd(rinv11,sh_ewald),velec));
407 felec = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
409 cutoff_mask = _mm_cmplt_pd(rsq11,rcutoff2);
411 /* Update potential sum for this i atom from the interaction with this j atom. */
412 velec = _mm_and_pd(velec,cutoff_mask);
413 velecsum = _mm_add_pd(velecsum,velec);
417 fscal = _mm_and_pd(fscal,cutoff_mask);
419 /* Update vectorial force */
420 fix1 = _mm_macc_pd(dx11,fscal,fix1);
421 fiy1 = _mm_macc_pd(dy11,fscal,fiy1);
422 fiz1 = _mm_macc_pd(dz11,fscal,fiz1);
424 fjx1 = _mm_macc_pd(dx11,fscal,fjx1);
425 fjy1 = _mm_macc_pd(dy11,fscal,fjy1);
426 fjz1 = _mm_macc_pd(dz11,fscal,fjz1);
430 /**************************
431 * CALCULATE INTERACTIONS *
432 **************************/
434 if (gmx_mm_any_lt(rsq12,rcutoff2))
437 r12 = _mm_mul_pd(rsq12,rinv12);
439 /* EWALD ELECTROSTATICS */
441 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
442 ewrt = _mm_mul_pd(r12,ewtabscale);
443 ewitab = _mm_cvttpd_epi32(ewrt);
445 eweps = _mm_frcz_pd(ewrt);
447 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
449 twoeweps = _mm_add_pd(eweps,eweps);
450 ewitab = _mm_slli_epi32(ewitab,2);
451 ewtabF = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
452 ewtabD = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
453 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
454 ewtabV = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
455 ewtabFn = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
456 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
457 felec = _mm_macc_pd(eweps,ewtabD,ewtabF);
458 velec = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
459 velec = _mm_mul_pd(qq12,_mm_sub_pd(_mm_sub_pd(rinv12,sh_ewald),velec));
460 felec = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
462 cutoff_mask = _mm_cmplt_pd(rsq12,rcutoff2);
464 /* Update potential sum for this i atom from the interaction with this j atom. */
465 velec = _mm_and_pd(velec,cutoff_mask);
466 velecsum = _mm_add_pd(velecsum,velec);
470 fscal = _mm_and_pd(fscal,cutoff_mask);
472 /* Update vectorial force */
473 fix1 = _mm_macc_pd(dx12,fscal,fix1);
474 fiy1 = _mm_macc_pd(dy12,fscal,fiy1);
475 fiz1 = _mm_macc_pd(dz12,fscal,fiz1);
477 fjx2 = _mm_macc_pd(dx12,fscal,fjx2);
478 fjy2 = _mm_macc_pd(dy12,fscal,fjy2);
479 fjz2 = _mm_macc_pd(dz12,fscal,fjz2);
483 /**************************
484 * CALCULATE INTERACTIONS *
485 **************************/
487 if (gmx_mm_any_lt(rsq13,rcutoff2))
490 r13 = _mm_mul_pd(rsq13,rinv13);
492 /* EWALD ELECTROSTATICS */
494 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
495 ewrt = _mm_mul_pd(r13,ewtabscale);
496 ewitab = _mm_cvttpd_epi32(ewrt);
498 eweps = _mm_frcz_pd(ewrt);
500 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
502 twoeweps = _mm_add_pd(eweps,eweps);
503 ewitab = _mm_slli_epi32(ewitab,2);
504 ewtabF = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
505 ewtabD = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
506 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
507 ewtabV = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
508 ewtabFn = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
509 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
510 felec = _mm_macc_pd(eweps,ewtabD,ewtabF);
511 velec = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
512 velec = _mm_mul_pd(qq13,_mm_sub_pd(_mm_sub_pd(rinv13,sh_ewald),velec));
513 felec = _mm_mul_pd(_mm_mul_pd(qq13,rinv13),_mm_sub_pd(rinvsq13,felec));
515 cutoff_mask = _mm_cmplt_pd(rsq13,rcutoff2);
517 /* Update potential sum for this i atom from the interaction with this j atom. */
518 velec = _mm_and_pd(velec,cutoff_mask);
519 velecsum = _mm_add_pd(velecsum,velec);
523 fscal = _mm_and_pd(fscal,cutoff_mask);
525 /* Update vectorial force */
526 fix1 = _mm_macc_pd(dx13,fscal,fix1);
527 fiy1 = _mm_macc_pd(dy13,fscal,fiy1);
528 fiz1 = _mm_macc_pd(dz13,fscal,fiz1);
530 fjx3 = _mm_macc_pd(dx13,fscal,fjx3);
531 fjy3 = _mm_macc_pd(dy13,fscal,fjy3);
532 fjz3 = _mm_macc_pd(dz13,fscal,fjz3);
536 /**************************
537 * CALCULATE INTERACTIONS *
538 **************************/
540 if (gmx_mm_any_lt(rsq21,rcutoff2))
543 r21 = _mm_mul_pd(rsq21,rinv21);
545 /* EWALD ELECTROSTATICS */
547 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
548 ewrt = _mm_mul_pd(r21,ewtabscale);
549 ewitab = _mm_cvttpd_epi32(ewrt);
551 eweps = _mm_frcz_pd(ewrt);
553 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
555 twoeweps = _mm_add_pd(eweps,eweps);
556 ewitab = _mm_slli_epi32(ewitab,2);
557 ewtabF = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
558 ewtabD = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
559 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
560 ewtabV = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
561 ewtabFn = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
562 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
563 felec = _mm_macc_pd(eweps,ewtabD,ewtabF);
564 velec = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
565 velec = _mm_mul_pd(qq21,_mm_sub_pd(_mm_sub_pd(rinv21,sh_ewald),velec));
566 felec = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
568 cutoff_mask = _mm_cmplt_pd(rsq21,rcutoff2);
570 /* Update potential sum for this i atom from the interaction with this j atom. */
571 velec = _mm_and_pd(velec,cutoff_mask);
572 velecsum = _mm_add_pd(velecsum,velec);
576 fscal = _mm_and_pd(fscal,cutoff_mask);
578 /* Update vectorial force */
579 fix2 = _mm_macc_pd(dx21,fscal,fix2);
580 fiy2 = _mm_macc_pd(dy21,fscal,fiy2);
581 fiz2 = _mm_macc_pd(dz21,fscal,fiz2);
583 fjx1 = _mm_macc_pd(dx21,fscal,fjx1);
584 fjy1 = _mm_macc_pd(dy21,fscal,fjy1);
585 fjz1 = _mm_macc_pd(dz21,fscal,fjz1);
589 /**************************
590 * CALCULATE INTERACTIONS *
591 **************************/
593 if (gmx_mm_any_lt(rsq22,rcutoff2))
596 r22 = _mm_mul_pd(rsq22,rinv22);
598 /* EWALD ELECTROSTATICS */
600 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
601 ewrt = _mm_mul_pd(r22,ewtabscale);
602 ewitab = _mm_cvttpd_epi32(ewrt);
604 eweps = _mm_frcz_pd(ewrt);
606 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
608 twoeweps = _mm_add_pd(eweps,eweps);
609 ewitab = _mm_slli_epi32(ewitab,2);
610 ewtabF = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
611 ewtabD = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
612 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
613 ewtabV = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
614 ewtabFn = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
615 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
616 felec = _mm_macc_pd(eweps,ewtabD,ewtabF);
617 velec = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
618 velec = _mm_mul_pd(qq22,_mm_sub_pd(_mm_sub_pd(rinv22,sh_ewald),velec));
619 felec = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
621 cutoff_mask = _mm_cmplt_pd(rsq22,rcutoff2);
623 /* Update potential sum for this i atom from the interaction with this j atom. */
624 velec = _mm_and_pd(velec,cutoff_mask);
625 velecsum = _mm_add_pd(velecsum,velec);
629 fscal = _mm_and_pd(fscal,cutoff_mask);
631 /* Update vectorial force */
632 fix2 = _mm_macc_pd(dx22,fscal,fix2);
633 fiy2 = _mm_macc_pd(dy22,fscal,fiy2);
634 fiz2 = _mm_macc_pd(dz22,fscal,fiz2);
636 fjx2 = _mm_macc_pd(dx22,fscal,fjx2);
637 fjy2 = _mm_macc_pd(dy22,fscal,fjy2);
638 fjz2 = _mm_macc_pd(dz22,fscal,fjz2);
642 /**************************
643 * CALCULATE INTERACTIONS *
644 **************************/
646 if (gmx_mm_any_lt(rsq23,rcutoff2))
649 r23 = _mm_mul_pd(rsq23,rinv23);
651 /* EWALD ELECTROSTATICS */
653 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
654 ewrt = _mm_mul_pd(r23,ewtabscale);
655 ewitab = _mm_cvttpd_epi32(ewrt);
657 eweps = _mm_frcz_pd(ewrt);
659 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
661 twoeweps = _mm_add_pd(eweps,eweps);
662 ewitab = _mm_slli_epi32(ewitab,2);
663 ewtabF = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
664 ewtabD = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
665 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
666 ewtabV = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
667 ewtabFn = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
668 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
669 felec = _mm_macc_pd(eweps,ewtabD,ewtabF);
670 velec = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
671 velec = _mm_mul_pd(qq23,_mm_sub_pd(_mm_sub_pd(rinv23,sh_ewald),velec));
672 felec = _mm_mul_pd(_mm_mul_pd(qq23,rinv23),_mm_sub_pd(rinvsq23,felec));
674 cutoff_mask = _mm_cmplt_pd(rsq23,rcutoff2);
676 /* Update potential sum for this i atom from the interaction with this j atom. */
677 velec = _mm_and_pd(velec,cutoff_mask);
678 velecsum = _mm_add_pd(velecsum,velec);
682 fscal = _mm_and_pd(fscal,cutoff_mask);
684 /* Update vectorial force */
685 fix2 = _mm_macc_pd(dx23,fscal,fix2);
686 fiy2 = _mm_macc_pd(dy23,fscal,fiy2);
687 fiz2 = _mm_macc_pd(dz23,fscal,fiz2);
689 fjx3 = _mm_macc_pd(dx23,fscal,fjx3);
690 fjy3 = _mm_macc_pd(dy23,fscal,fjy3);
691 fjz3 = _mm_macc_pd(dz23,fscal,fjz3);
695 /**************************
696 * CALCULATE INTERACTIONS *
697 **************************/
699 if (gmx_mm_any_lt(rsq31,rcutoff2))
702 r31 = _mm_mul_pd(rsq31,rinv31);
704 /* EWALD ELECTROSTATICS */
706 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
707 ewrt = _mm_mul_pd(r31,ewtabscale);
708 ewitab = _mm_cvttpd_epi32(ewrt);
710 eweps = _mm_frcz_pd(ewrt);
712 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
714 twoeweps = _mm_add_pd(eweps,eweps);
715 ewitab = _mm_slli_epi32(ewitab,2);
716 ewtabF = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
717 ewtabD = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
718 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
719 ewtabV = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
720 ewtabFn = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
721 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
722 felec = _mm_macc_pd(eweps,ewtabD,ewtabF);
723 velec = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
724 velec = _mm_mul_pd(qq31,_mm_sub_pd(_mm_sub_pd(rinv31,sh_ewald),velec));
725 felec = _mm_mul_pd(_mm_mul_pd(qq31,rinv31),_mm_sub_pd(rinvsq31,felec));
727 cutoff_mask = _mm_cmplt_pd(rsq31,rcutoff2);
729 /* Update potential sum for this i atom from the interaction with this j atom. */
730 velec = _mm_and_pd(velec,cutoff_mask);
731 velecsum = _mm_add_pd(velecsum,velec);
735 fscal = _mm_and_pd(fscal,cutoff_mask);
737 /* Update vectorial force */
738 fix3 = _mm_macc_pd(dx31,fscal,fix3);
739 fiy3 = _mm_macc_pd(dy31,fscal,fiy3);
740 fiz3 = _mm_macc_pd(dz31,fscal,fiz3);
742 fjx1 = _mm_macc_pd(dx31,fscal,fjx1);
743 fjy1 = _mm_macc_pd(dy31,fscal,fjy1);
744 fjz1 = _mm_macc_pd(dz31,fscal,fjz1);
748 /**************************
749 * CALCULATE INTERACTIONS *
750 **************************/
752 if (gmx_mm_any_lt(rsq32,rcutoff2))
755 r32 = _mm_mul_pd(rsq32,rinv32);
757 /* EWALD ELECTROSTATICS */
759 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
760 ewrt = _mm_mul_pd(r32,ewtabscale);
761 ewitab = _mm_cvttpd_epi32(ewrt);
763 eweps = _mm_frcz_pd(ewrt);
765 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
767 twoeweps = _mm_add_pd(eweps,eweps);
768 ewitab = _mm_slli_epi32(ewitab,2);
769 ewtabF = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
770 ewtabD = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
771 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
772 ewtabV = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
773 ewtabFn = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
774 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
775 felec = _mm_macc_pd(eweps,ewtabD,ewtabF);
776 velec = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
777 velec = _mm_mul_pd(qq32,_mm_sub_pd(_mm_sub_pd(rinv32,sh_ewald),velec));
778 felec = _mm_mul_pd(_mm_mul_pd(qq32,rinv32),_mm_sub_pd(rinvsq32,felec));
780 cutoff_mask = _mm_cmplt_pd(rsq32,rcutoff2);
782 /* Update potential sum for this i atom from the interaction with this j atom. */
783 velec = _mm_and_pd(velec,cutoff_mask);
784 velecsum = _mm_add_pd(velecsum,velec);
788 fscal = _mm_and_pd(fscal,cutoff_mask);
790 /* Update vectorial force */
791 fix3 = _mm_macc_pd(dx32,fscal,fix3);
792 fiy3 = _mm_macc_pd(dy32,fscal,fiy3);
793 fiz3 = _mm_macc_pd(dz32,fscal,fiz3);
795 fjx2 = _mm_macc_pd(dx32,fscal,fjx2);
796 fjy2 = _mm_macc_pd(dy32,fscal,fjy2);
797 fjz2 = _mm_macc_pd(dz32,fscal,fjz2);
801 /**************************
802 * CALCULATE INTERACTIONS *
803 **************************/
805 if (gmx_mm_any_lt(rsq33,rcutoff2))
808 r33 = _mm_mul_pd(rsq33,rinv33);
810 /* EWALD ELECTROSTATICS */
812 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
813 ewrt = _mm_mul_pd(r33,ewtabscale);
814 ewitab = _mm_cvttpd_epi32(ewrt);
816 eweps = _mm_frcz_pd(ewrt);
818 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
820 twoeweps = _mm_add_pd(eweps,eweps);
821 ewitab = _mm_slli_epi32(ewitab,2);
822 ewtabF = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
823 ewtabD = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
824 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
825 ewtabV = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
826 ewtabFn = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
827 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
828 felec = _mm_macc_pd(eweps,ewtabD,ewtabF);
829 velec = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
830 velec = _mm_mul_pd(qq33,_mm_sub_pd(_mm_sub_pd(rinv33,sh_ewald),velec));
831 felec = _mm_mul_pd(_mm_mul_pd(qq33,rinv33),_mm_sub_pd(rinvsq33,felec));
833 cutoff_mask = _mm_cmplt_pd(rsq33,rcutoff2);
835 /* Update potential sum for this i atom from the interaction with this j atom. */
836 velec = _mm_and_pd(velec,cutoff_mask);
837 velecsum = _mm_add_pd(velecsum,velec);
841 fscal = _mm_and_pd(fscal,cutoff_mask);
843 /* Update vectorial force */
844 fix3 = _mm_macc_pd(dx33,fscal,fix3);
845 fiy3 = _mm_macc_pd(dy33,fscal,fiy3);
846 fiz3 = _mm_macc_pd(dz33,fscal,fiz3);
848 fjx3 = _mm_macc_pd(dx33,fscal,fjx3);
849 fjy3 = _mm_macc_pd(dy33,fscal,fjy3);
850 fjz3 = _mm_macc_pd(dz33,fscal,fjz3);
854 gmx_mm_decrement_4rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
856 /* Inner loop uses 502 flops */
863 j_coord_offsetA = DIM*jnrA;
865 /* load j atom coordinates */
866 gmx_mm_load_4rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
867 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
868 &jy2,&jz2,&jx3,&jy3,&jz3);
870 /* Calculate displacement vector */
871 dx00 = _mm_sub_pd(ix0,jx0);
872 dy00 = _mm_sub_pd(iy0,jy0);
873 dz00 = _mm_sub_pd(iz0,jz0);
874 dx11 = _mm_sub_pd(ix1,jx1);
875 dy11 = _mm_sub_pd(iy1,jy1);
876 dz11 = _mm_sub_pd(iz1,jz1);
877 dx12 = _mm_sub_pd(ix1,jx2);
878 dy12 = _mm_sub_pd(iy1,jy2);
879 dz12 = _mm_sub_pd(iz1,jz2);
880 dx13 = _mm_sub_pd(ix1,jx3);
881 dy13 = _mm_sub_pd(iy1,jy3);
882 dz13 = _mm_sub_pd(iz1,jz3);
883 dx21 = _mm_sub_pd(ix2,jx1);
884 dy21 = _mm_sub_pd(iy2,jy1);
885 dz21 = _mm_sub_pd(iz2,jz1);
886 dx22 = _mm_sub_pd(ix2,jx2);
887 dy22 = _mm_sub_pd(iy2,jy2);
888 dz22 = _mm_sub_pd(iz2,jz2);
889 dx23 = _mm_sub_pd(ix2,jx3);
890 dy23 = _mm_sub_pd(iy2,jy3);
891 dz23 = _mm_sub_pd(iz2,jz3);
892 dx31 = _mm_sub_pd(ix3,jx1);
893 dy31 = _mm_sub_pd(iy3,jy1);
894 dz31 = _mm_sub_pd(iz3,jz1);
895 dx32 = _mm_sub_pd(ix3,jx2);
896 dy32 = _mm_sub_pd(iy3,jy2);
897 dz32 = _mm_sub_pd(iz3,jz2);
898 dx33 = _mm_sub_pd(ix3,jx3);
899 dy33 = _mm_sub_pd(iy3,jy3);
900 dz33 = _mm_sub_pd(iz3,jz3);
902 /* Calculate squared distance and things based on it */
903 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
904 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
905 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
906 rsq13 = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
907 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
908 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
909 rsq23 = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
910 rsq31 = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
911 rsq32 = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
912 rsq33 = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
914 rinv00 = gmx_mm_invsqrt_pd(rsq00);
915 rinv11 = gmx_mm_invsqrt_pd(rsq11);
916 rinv12 = gmx_mm_invsqrt_pd(rsq12);
917 rinv13 = gmx_mm_invsqrt_pd(rsq13);
918 rinv21 = gmx_mm_invsqrt_pd(rsq21);
919 rinv22 = gmx_mm_invsqrt_pd(rsq22);
920 rinv23 = gmx_mm_invsqrt_pd(rsq23);
921 rinv31 = gmx_mm_invsqrt_pd(rsq31);
922 rinv32 = gmx_mm_invsqrt_pd(rsq32);
923 rinv33 = gmx_mm_invsqrt_pd(rsq33);
925 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
926 rinvsq11 = _mm_mul_pd(rinv11,rinv11);
927 rinvsq12 = _mm_mul_pd(rinv12,rinv12);
928 rinvsq13 = _mm_mul_pd(rinv13,rinv13);
929 rinvsq21 = _mm_mul_pd(rinv21,rinv21);
930 rinvsq22 = _mm_mul_pd(rinv22,rinv22);
931 rinvsq23 = _mm_mul_pd(rinv23,rinv23);
932 rinvsq31 = _mm_mul_pd(rinv31,rinv31);
933 rinvsq32 = _mm_mul_pd(rinv32,rinv32);
934 rinvsq33 = _mm_mul_pd(rinv33,rinv33);
936 fjx0 = _mm_setzero_pd();
937 fjy0 = _mm_setzero_pd();
938 fjz0 = _mm_setzero_pd();
939 fjx1 = _mm_setzero_pd();
940 fjy1 = _mm_setzero_pd();
941 fjz1 = _mm_setzero_pd();
942 fjx2 = _mm_setzero_pd();
943 fjy2 = _mm_setzero_pd();
944 fjz2 = _mm_setzero_pd();
945 fjx3 = _mm_setzero_pd();
946 fjy3 = _mm_setzero_pd();
947 fjz3 = _mm_setzero_pd();
949 /**************************
950 * CALCULATE INTERACTIONS *
951 **************************/
953 if (gmx_mm_any_lt(rsq00,rcutoff2))
956 r00 = _mm_mul_pd(rsq00,rinv00);
958 /* Analytical LJ-PME */
959 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
960 ewcljrsq = _mm_mul_pd(ewclj2,rsq00);
961 ewclj6 = _mm_mul_pd(ewclj2,_mm_mul_pd(ewclj2,ewclj2));
962 exponent = gmx_simd_exp_d(ewcljrsq);
963 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
964 poly = _mm_mul_pd(exponent,_mm_macc_pd(_mm_mul_pd(ewcljrsq,ewcljrsq),one_half,_mm_sub_pd(one,ewcljrsq)));
965 /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
966 vvdw6 = _mm_mul_pd(_mm_macc_pd(-c6grid_00,_mm_sub_pd(one,poly),c6_00),rinvsix);
967 vvdw12 = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
968 vvdw = _mm_msub_pd(_mm_nmacc_pd(c12_00,_mm_mul_pd(sh_vdw_invrcut6,sh_vdw_invrcut6),vvdw12),one_twelfth,
969 _mm_mul_pd(_mm_sub_pd(vvdw6,_mm_macc_pd(c6grid_00,sh_lj_ewald,_mm_mul_pd(c6_00,sh_vdw_invrcut6))),one_sixth));
970 /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
971 fvdw = _mm_mul_pd(_mm_add_pd(vvdw12,_mm_msub_pd(_mm_mul_pd(c6grid_00,one_sixth),_mm_mul_pd(exponent,ewclj6),vvdw6)),rinvsq00);
973 cutoff_mask = _mm_cmplt_pd(rsq00,rcutoff2);
975 /* Update potential sum for this i atom from the interaction with this j atom. */
976 vvdw = _mm_and_pd(vvdw,cutoff_mask);
977 vvdw = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
978 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
982 fscal = _mm_and_pd(fscal,cutoff_mask);
984 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
986 /* Update vectorial force */
987 fix0 = _mm_macc_pd(dx00,fscal,fix0);
988 fiy0 = _mm_macc_pd(dy00,fscal,fiy0);
989 fiz0 = _mm_macc_pd(dz00,fscal,fiz0);
991 fjx0 = _mm_macc_pd(dx00,fscal,fjx0);
992 fjy0 = _mm_macc_pd(dy00,fscal,fjy0);
993 fjz0 = _mm_macc_pd(dz00,fscal,fjz0);
997 /**************************
998 * CALCULATE INTERACTIONS *
999 **************************/
1001 if (gmx_mm_any_lt(rsq11,rcutoff2))
1004 r11 = _mm_mul_pd(rsq11,rinv11);
1006 /* EWALD ELECTROSTATICS */
1008 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1009 ewrt = _mm_mul_pd(r11,ewtabscale);
1010 ewitab = _mm_cvttpd_epi32(ewrt);
1012 eweps = _mm_frcz_pd(ewrt);
1014 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1016 twoeweps = _mm_add_pd(eweps,eweps);
1017 ewitab = _mm_slli_epi32(ewitab,2);
1018 ewtabF = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1019 ewtabD = _mm_setzero_pd();
1020 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1021 ewtabV = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
1022 ewtabFn = _mm_setzero_pd();
1023 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1024 felec = _mm_macc_pd(eweps,ewtabD,ewtabF);
1025 velec = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
1026 velec = _mm_mul_pd(qq11,_mm_sub_pd(_mm_sub_pd(rinv11,sh_ewald),velec));
1027 felec = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
1029 cutoff_mask = _mm_cmplt_pd(rsq11,rcutoff2);
1031 /* Update potential sum for this i atom from the interaction with this j atom. */
1032 velec = _mm_and_pd(velec,cutoff_mask);
1033 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1034 velecsum = _mm_add_pd(velecsum,velec);
1038 fscal = _mm_and_pd(fscal,cutoff_mask);
1040 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1042 /* Update vectorial force */
1043 fix1 = _mm_macc_pd(dx11,fscal,fix1);
1044 fiy1 = _mm_macc_pd(dy11,fscal,fiy1);
1045 fiz1 = _mm_macc_pd(dz11,fscal,fiz1);
1047 fjx1 = _mm_macc_pd(dx11,fscal,fjx1);
1048 fjy1 = _mm_macc_pd(dy11,fscal,fjy1);
1049 fjz1 = _mm_macc_pd(dz11,fscal,fjz1);
1053 /**************************
1054 * CALCULATE INTERACTIONS *
1055 **************************/
1057 if (gmx_mm_any_lt(rsq12,rcutoff2))
1060 r12 = _mm_mul_pd(rsq12,rinv12);
1062 /* EWALD ELECTROSTATICS */
1064 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1065 ewrt = _mm_mul_pd(r12,ewtabscale);
1066 ewitab = _mm_cvttpd_epi32(ewrt);
1068 eweps = _mm_frcz_pd(ewrt);
1070 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1072 twoeweps = _mm_add_pd(eweps,eweps);
1073 ewitab = _mm_slli_epi32(ewitab,2);
1074 ewtabF = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1075 ewtabD = _mm_setzero_pd();
1076 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1077 ewtabV = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
1078 ewtabFn = _mm_setzero_pd();
1079 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1080 felec = _mm_macc_pd(eweps,ewtabD,ewtabF);
1081 velec = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
1082 velec = _mm_mul_pd(qq12,_mm_sub_pd(_mm_sub_pd(rinv12,sh_ewald),velec));
1083 felec = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
1085 cutoff_mask = _mm_cmplt_pd(rsq12,rcutoff2);
1087 /* Update potential sum for this i atom from the interaction with this j atom. */
1088 velec = _mm_and_pd(velec,cutoff_mask);
1089 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1090 velecsum = _mm_add_pd(velecsum,velec);
1094 fscal = _mm_and_pd(fscal,cutoff_mask);
1096 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1098 /* Update vectorial force */
1099 fix1 = _mm_macc_pd(dx12,fscal,fix1);
1100 fiy1 = _mm_macc_pd(dy12,fscal,fiy1);
1101 fiz1 = _mm_macc_pd(dz12,fscal,fiz1);
1103 fjx2 = _mm_macc_pd(dx12,fscal,fjx2);
1104 fjy2 = _mm_macc_pd(dy12,fscal,fjy2);
1105 fjz2 = _mm_macc_pd(dz12,fscal,fjz2);
1109 /**************************
1110 * CALCULATE INTERACTIONS *
1111 **************************/
1113 if (gmx_mm_any_lt(rsq13,rcutoff2))
1116 r13 = _mm_mul_pd(rsq13,rinv13);
1118 /* EWALD ELECTROSTATICS */
1120 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1121 ewrt = _mm_mul_pd(r13,ewtabscale);
1122 ewitab = _mm_cvttpd_epi32(ewrt);
1124 eweps = _mm_frcz_pd(ewrt);
1126 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1128 twoeweps = _mm_add_pd(eweps,eweps);
1129 ewitab = _mm_slli_epi32(ewitab,2);
1130 ewtabF = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1131 ewtabD = _mm_setzero_pd();
1132 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1133 ewtabV = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
1134 ewtabFn = _mm_setzero_pd();
1135 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1136 felec = _mm_macc_pd(eweps,ewtabD,ewtabF);
1137 velec = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
1138 velec = _mm_mul_pd(qq13,_mm_sub_pd(_mm_sub_pd(rinv13,sh_ewald),velec));
1139 felec = _mm_mul_pd(_mm_mul_pd(qq13,rinv13),_mm_sub_pd(rinvsq13,felec));
1141 cutoff_mask = _mm_cmplt_pd(rsq13,rcutoff2);
1143 /* Update potential sum for this i atom from the interaction with this j atom. */
1144 velec = _mm_and_pd(velec,cutoff_mask);
1145 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1146 velecsum = _mm_add_pd(velecsum,velec);
1150 fscal = _mm_and_pd(fscal,cutoff_mask);
1152 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1154 /* Update vectorial force */
1155 fix1 = _mm_macc_pd(dx13,fscal,fix1);
1156 fiy1 = _mm_macc_pd(dy13,fscal,fiy1);
1157 fiz1 = _mm_macc_pd(dz13,fscal,fiz1);
1159 fjx3 = _mm_macc_pd(dx13,fscal,fjx3);
1160 fjy3 = _mm_macc_pd(dy13,fscal,fjy3);
1161 fjz3 = _mm_macc_pd(dz13,fscal,fjz3);
1165 /**************************
1166 * CALCULATE INTERACTIONS *
1167 **************************/
1169 if (gmx_mm_any_lt(rsq21,rcutoff2))
1172 r21 = _mm_mul_pd(rsq21,rinv21);
1174 /* EWALD ELECTROSTATICS */
1176 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1177 ewrt = _mm_mul_pd(r21,ewtabscale);
1178 ewitab = _mm_cvttpd_epi32(ewrt);
1180 eweps = _mm_frcz_pd(ewrt);
1182 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1184 twoeweps = _mm_add_pd(eweps,eweps);
1185 ewitab = _mm_slli_epi32(ewitab,2);
1186 ewtabF = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1187 ewtabD = _mm_setzero_pd();
1188 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1189 ewtabV = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
1190 ewtabFn = _mm_setzero_pd();
1191 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1192 felec = _mm_macc_pd(eweps,ewtabD,ewtabF);
1193 velec = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
1194 velec = _mm_mul_pd(qq21,_mm_sub_pd(_mm_sub_pd(rinv21,sh_ewald),velec));
1195 felec = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
1197 cutoff_mask = _mm_cmplt_pd(rsq21,rcutoff2);
1199 /* Update potential sum for this i atom from the interaction with this j atom. */
1200 velec = _mm_and_pd(velec,cutoff_mask);
1201 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1202 velecsum = _mm_add_pd(velecsum,velec);
1206 fscal = _mm_and_pd(fscal,cutoff_mask);
1208 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1210 /* Update vectorial force */
1211 fix2 = _mm_macc_pd(dx21,fscal,fix2);
1212 fiy2 = _mm_macc_pd(dy21,fscal,fiy2);
1213 fiz2 = _mm_macc_pd(dz21,fscal,fiz2);
1215 fjx1 = _mm_macc_pd(dx21,fscal,fjx1);
1216 fjy1 = _mm_macc_pd(dy21,fscal,fjy1);
1217 fjz1 = _mm_macc_pd(dz21,fscal,fjz1);
1221 /**************************
1222 * CALCULATE INTERACTIONS *
1223 **************************/
1225 if (gmx_mm_any_lt(rsq22,rcutoff2))
1228 r22 = _mm_mul_pd(rsq22,rinv22);
1230 /* EWALD ELECTROSTATICS */
1232 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1233 ewrt = _mm_mul_pd(r22,ewtabscale);
1234 ewitab = _mm_cvttpd_epi32(ewrt);
1236 eweps = _mm_frcz_pd(ewrt);
1238 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1240 twoeweps = _mm_add_pd(eweps,eweps);
1241 ewitab = _mm_slli_epi32(ewitab,2);
1242 ewtabF = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1243 ewtabD = _mm_setzero_pd();
1244 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1245 ewtabV = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
1246 ewtabFn = _mm_setzero_pd();
1247 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1248 felec = _mm_macc_pd(eweps,ewtabD,ewtabF);
1249 velec = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
1250 velec = _mm_mul_pd(qq22,_mm_sub_pd(_mm_sub_pd(rinv22,sh_ewald),velec));
1251 felec = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
1253 cutoff_mask = _mm_cmplt_pd(rsq22,rcutoff2);
1255 /* Update potential sum for this i atom from the interaction with this j atom. */
1256 velec = _mm_and_pd(velec,cutoff_mask);
1257 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1258 velecsum = _mm_add_pd(velecsum,velec);
1262 fscal = _mm_and_pd(fscal,cutoff_mask);
1264 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1266 /* Update vectorial force */
1267 fix2 = _mm_macc_pd(dx22,fscal,fix2);
1268 fiy2 = _mm_macc_pd(dy22,fscal,fiy2);
1269 fiz2 = _mm_macc_pd(dz22,fscal,fiz2);
1271 fjx2 = _mm_macc_pd(dx22,fscal,fjx2);
1272 fjy2 = _mm_macc_pd(dy22,fscal,fjy2);
1273 fjz2 = _mm_macc_pd(dz22,fscal,fjz2);
1277 /**************************
1278 * CALCULATE INTERACTIONS *
1279 **************************/
1281 if (gmx_mm_any_lt(rsq23,rcutoff2))
1284 r23 = _mm_mul_pd(rsq23,rinv23);
1286 /* EWALD ELECTROSTATICS */
1288 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1289 ewrt = _mm_mul_pd(r23,ewtabscale);
1290 ewitab = _mm_cvttpd_epi32(ewrt);
1292 eweps = _mm_frcz_pd(ewrt);
1294 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1296 twoeweps = _mm_add_pd(eweps,eweps);
1297 ewitab = _mm_slli_epi32(ewitab,2);
1298 ewtabF = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1299 ewtabD = _mm_setzero_pd();
1300 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1301 ewtabV = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
1302 ewtabFn = _mm_setzero_pd();
1303 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1304 felec = _mm_macc_pd(eweps,ewtabD,ewtabF);
1305 velec = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
1306 velec = _mm_mul_pd(qq23,_mm_sub_pd(_mm_sub_pd(rinv23,sh_ewald),velec));
1307 felec = _mm_mul_pd(_mm_mul_pd(qq23,rinv23),_mm_sub_pd(rinvsq23,felec));
1309 cutoff_mask = _mm_cmplt_pd(rsq23,rcutoff2);
1311 /* Update potential sum for this i atom from the interaction with this j atom. */
1312 velec = _mm_and_pd(velec,cutoff_mask);
1313 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1314 velecsum = _mm_add_pd(velecsum,velec);
1318 fscal = _mm_and_pd(fscal,cutoff_mask);
1320 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1322 /* Update vectorial force */
1323 fix2 = _mm_macc_pd(dx23,fscal,fix2);
1324 fiy2 = _mm_macc_pd(dy23,fscal,fiy2);
1325 fiz2 = _mm_macc_pd(dz23,fscal,fiz2);
1327 fjx3 = _mm_macc_pd(dx23,fscal,fjx3);
1328 fjy3 = _mm_macc_pd(dy23,fscal,fjy3);
1329 fjz3 = _mm_macc_pd(dz23,fscal,fjz3);
1333 /**************************
1334 * CALCULATE INTERACTIONS *
1335 **************************/
1337 if (gmx_mm_any_lt(rsq31,rcutoff2))
1340 r31 = _mm_mul_pd(rsq31,rinv31);
1342 /* EWALD ELECTROSTATICS */
1344 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1345 ewrt = _mm_mul_pd(r31,ewtabscale);
1346 ewitab = _mm_cvttpd_epi32(ewrt);
1348 eweps = _mm_frcz_pd(ewrt);
1350 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1352 twoeweps = _mm_add_pd(eweps,eweps);
1353 ewitab = _mm_slli_epi32(ewitab,2);
1354 ewtabF = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1355 ewtabD = _mm_setzero_pd();
1356 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1357 ewtabV = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
1358 ewtabFn = _mm_setzero_pd();
1359 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1360 felec = _mm_macc_pd(eweps,ewtabD,ewtabF);
1361 velec = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
1362 velec = _mm_mul_pd(qq31,_mm_sub_pd(_mm_sub_pd(rinv31,sh_ewald),velec));
1363 felec = _mm_mul_pd(_mm_mul_pd(qq31,rinv31),_mm_sub_pd(rinvsq31,felec));
1365 cutoff_mask = _mm_cmplt_pd(rsq31,rcutoff2);
1367 /* Update potential sum for this i atom from the interaction with this j atom. */
1368 velec = _mm_and_pd(velec,cutoff_mask);
1369 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1370 velecsum = _mm_add_pd(velecsum,velec);
1374 fscal = _mm_and_pd(fscal,cutoff_mask);
1376 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1378 /* Update vectorial force */
1379 fix3 = _mm_macc_pd(dx31,fscal,fix3);
1380 fiy3 = _mm_macc_pd(dy31,fscal,fiy3);
1381 fiz3 = _mm_macc_pd(dz31,fscal,fiz3);
1383 fjx1 = _mm_macc_pd(dx31,fscal,fjx1);
1384 fjy1 = _mm_macc_pd(dy31,fscal,fjy1);
1385 fjz1 = _mm_macc_pd(dz31,fscal,fjz1);
1389 /**************************
1390 * CALCULATE INTERACTIONS *
1391 **************************/
1393 if (gmx_mm_any_lt(rsq32,rcutoff2))
1396 r32 = _mm_mul_pd(rsq32,rinv32);
1398 /* EWALD ELECTROSTATICS */
1400 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1401 ewrt = _mm_mul_pd(r32,ewtabscale);
1402 ewitab = _mm_cvttpd_epi32(ewrt);
1404 eweps = _mm_frcz_pd(ewrt);
1406 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1408 twoeweps = _mm_add_pd(eweps,eweps);
1409 ewitab = _mm_slli_epi32(ewitab,2);
1410 ewtabF = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1411 ewtabD = _mm_setzero_pd();
1412 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1413 ewtabV = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
1414 ewtabFn = _mm_setzero_pd();
1415 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1416 felec = _mm_macc_pd(eweps,ewtabD,ewtabF);
1417 velec = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
1418 velec = _mm_mul_pd(qq32,_mm_sub_pd(_mm_sub_pd(rinv32,sh_ewald),velec));
1419 felec = _mm_mul_pd(_mm_mul_pd(qq32,rinv32),_mm_sub_pd(rinvsq32,felec));
1421 cutoff_mask = _mm_cmplt_pd(rsq32,rcutoff2);
1423 /* Update potential sum for this i atom from the interaction with this j atom. */
1424 velec = _mm_and_pd(velec,cutoff_mask);
1425 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1426 velecsum = _mm_add_pd(velecsum,velec);
1430 fscal = _mm_and_pd(fscal,cutoff_mask);
1432 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1434 /* Update vectorial force */
1435 fix3 = _mm_macc_pd(dx32,fscal,fix3);
1436 fiy3 = _mm_macc_pd(dy32,fscal,fiy3);
1437 fiz3 = _mm_macc_pd(dz32,fscal,fiz3);
1439 fjx2 = _mm_macc_pd(dx32,fscal,fjx2);
1440 fjy2 = _mm_macc_pd(dy32,fscal,fjy2);
1441 fjz2 = _mm_macc_pd(dz32,fscal,fjz2);
1445 /**************************
1446 * CALCULATE INTERACTIONS *
1447 **************************/
1449 if (gmx_mm_any_lt(rsq33,rcutoff2))
1452 r33 = _mm_mul_pd(rsq33,rinv33);
1454 /* EWALD ELECTROSTATICS */
1456 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1457 ewrt = _mm_mul_pd(r33,ewtabscale);
1458 ewitab = _mm_cvttpd_epi32(ewrt);
1460 eweps = _mm_frcz_pd(ewrt);
1462 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1464 twoeweps = _mm_add_pd(eweps,eweps);
1465 ewitab = _mm_slli_epi32(ewitab,2);
1466 ewtabF = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1467 ewtabD = _mm_setzero_pd();
1468 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1469 ewtabV = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
1470 ewtabFn = _mm_setzero_pd();
1471 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1472 felec = _mm_macc_pd(eweps,ewtabD,ewtabF);
1473 velec = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
1474 velec = _mm_mul_pd(qq33,_mm_sub_pd(_mm_sub_pd(rinv33,sh_ewald),velec));
1475 felec = _mm_mul_pd(_mm_mul_pd(qq33,rinv33),_mm_sub_pd(rinvsq33,felec));
1477 cutoff_mask = _mm_cmplt_pd(rsq33,rcutoff2);
1479 /* Update potential sum for this i atom from the interaction with this j atom. */
1480 velec = _mm_and_pd(velec,cutoff_mask);
1481 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1482 velecsum = _mm_add_pd(velecsum,velec);
1486 fscal = _mm_and_pd(fscal,cutoff_mask);
1488 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1490 /* Update vectorial force */
1491 fix3 = _mm_macc_pd(dx33,fscal,fix3);
1492 fiy3 = _mm_macc_pd(dy33,fscal,fiy3);
1493 fiz3 = _mm_macc_pd(dz33,fscal,fiz3);
1495 fjx3 = _mm_macc_pd(dx33,fscal,fjx3);
1496 fjy3 = _mm_macc_pd(dy33,fscal,fjy3);
1497 fjz3 = _mm_macc_pd(dz33,fscal,fjz3);
1501 gmx_mm_decrement_4rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1503 /* Inner loop uses 502 flops */
1506 /* End of innermost loop */
1508 gmx_mm_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1509 f+i_coord_offset,fshift+i_shift_offset);
1512 /* Update potential energies */
1513 gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1514 gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1516 /* Increment number of inner iterations */
1517 inneriter += j_index_end - j_index_start;
1519 /* Outer loop uses 26 flops */
1522 /* Increment number of outer iterations */
1525 /* Update outer/inner flops */
1527 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*502);
1530 * Gromacs nonbonded kernel: nb_kernel_ElecEwSh_VdwLJEwSh_GeomW4W4_F_avx_128_fma_double
1531 * Electrostatics interaction: Ewald
1532 * VdW interaction: LJEwald
1533 * Geometry: Water4-Water4
1534 * Calculate force/pot: Force
1537 nb_kernel_ElecEwSh_VdwLJEwSh_GeomW4W4_F_avx_128_fma_double
1538 (t_nblist * gmx_restrict nlist,
1539 rvec * gmx_restrict xx,
1540 rvec * gmx_restrict ff,
1541 t_forcerec * gmx_restrict fr,
1542 t_mdatoms * gmx_restrict mdatoms,
1543 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1544 t_nrnb * gmx_restrict nrnb)
1546 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1547 * just 0 for non-waters.
1548 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
1549 * jnr indices corresponding to data put in the four positions in the SIMD register.
1551 int i_shift_offset,i_coord_offset,outeriter,inneriter;
1552 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1554 int j_coord_offsetA,j_coord_offsetB;
1555 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
1556 real rcutoff_scalar;
1557 real *shiftvec,*fshift,*x,*f;
1558 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1560 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1562 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1564 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1566 __m128d ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1567 int vdwjidx0A,vdwjidx0B;
1568 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1569 int vdwjidx1A,vdwjidx1B;
1570 __m128d jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1571 int vdwjidx2A,vdwjidx2B;
1572 __m128d jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1573 int vdwjidx3A,vdwjidx3B;
1574 __m128d jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1575 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1576 __m128d dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1577 __m128d dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1578 __m128d dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1579 __m128d dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1580 __m128d dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1581 __m128d dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1582 __m128d dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1583 __m128d dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1584 __m128d dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1585 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
1588 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1591 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
1592 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
1604 __m128d ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
1605 __m128d one_half = _mm_set1_pd(0.5);
1606 __m128d minus_one = _mm_set1_pd(-1.0);
1608 __m128d ewtabscale,eweps,twoeweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1610 __m128d dummy_mask,cutoff_mask;
1611 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
1612 __m128d one = _mm_set1_pd(1.0);
1613 __m128d two = _mm_set1_pd(2.0);
1619 jindex = nlist->jindex;
1621 shiftidx = nlist->shift;
1623 shiftvec = fr->shift_vec[0];
1624 fshift = fr->fshift[0];
1625 facel = _mm_set1_pd(fr->epsfac);
1626 charge = mdatoms->chargeA;
1627 nvdwtype = fr->ntype;
1628 vdwparam = fr->nbfp;
1629 vdwtype = mdatoms->typeA;
1630 vdwgridparam = fr->ljpme_c6grid;
1631 sh_lj_ewald = _mm_set1_pd(fr->ic->sh_lj_ewald);
1632 ewclj = _mm_set1_pd(fr->ewaldcoeff_lj);
1633 ewclj2 = _mm_mul_pd(minus_one,_mm_mul_pd(ewclj,ewclj));
1635 sh_ewald = _mm_set1_pd(fr->ic->sh_ewald);
1636 ewtab = fr->ic->tabq_coul_F;
1637 ewtabscale = _mm_set1_pd(fr->ic->tabq_scale);
1638 ewtabhalfspace = _mm_set1_pd(0.5/fr->ic->tabq_scale);
1640 /* Setup water-specific parameters */
1641 inr = nlist->iinr[0];
1642 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
1643 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
1644 iq3 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
1645 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
1647 jq1 = _mm_set1_pd(charge[inr+1]);
1648 jq2 = _mm_set1_pd(charge[inr+2]);
1649 jq3 = _mm_set1_pd(charge[inr+3]);
1650 vdwjidx0A = 2*vdwtype[inr+0];
1651 c6_00 = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
1652 c12_00 = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
1653 c6grid_00 = _mm_set1_pd(vdwgridparam[vdwioffset0+vdwjidx0A]);
1654 qq11 = _mm_mul_pd(iq1,jq1);
1655 qq12 = _mm_mul_pd(iq1,jq2);
1656 qq13 = _mm_mul_pd(iq1,jq3);
1657 qq21 = _mm_mul_pd(iq2,jq1);
1658 qq22 = _mm_mul_pd(iq2,jq2);
1659 qq23 = _mm_mul_pd(iq2,jq3);
1660 qq31 = _mm_mul_pd(iq3,jq1);
1661 qq32 = _mm_mul_pd(iq3,jq2);
1662 qq33 = _mm_mul_pd(iq3,jq3);
1664 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
1665 rcutoff_scalar = fr->rcoulomb;
1666 rcutoff = _mm_set1_pd(rcutoff_scalar);
1667 rcutoff2 = _mm_mul_pd(rcutoff,rcutoff);
1669 sh_vdw_invrcut6 = _mm_set1_pd(fr->ic->sh_invrc6);
1670 rvdw = _mm_set1_pd(fr->rvdw);
1672 /* Avoid stupid compiler warnings */
1674 j_coord_offsetA = 0;
1675 j_coord_offsetB = 0;
1680 /* Start outer loop over neighborlists */
1681 for(iidx=0; iidx<nri; iidx++)
1683 /* Load shift vector for this list */
1684 i_shift_offset = DIM*shiftidx[iidx];
1686 /* Load limits for loop over neighbors */
1687 j_index_start = jindex[iidx];
1688 j_index_end = jindex[iidx+1];
1690 /* Get outer coordinate index */
1692 i_coord_offset = DIM*inr;
1694 /* Load i particle coords and add shift vector */
1695 gmx_mm_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1696 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1698 fix0 = _mm_setzero_pd();
1699 fiy0 = _mm_setzero_pd();
1700 fiz0 = _mm_setzero_pd();
1701 fix1 = _mm_setzero_pd();
1702 fiy1 = _mm_setzero_pd();
1703 fiz1 = _mm_setzero_pd();
1704 fix2 = _mm_setzero_pd();
1705 fiy2 = _mm_setzero_pd();
1706 fiz2 = _mm_setzero_pd();
1707 fix3 = _mm_setzero_pd();
1708 fiy3 = _mm_setzero_pd();
1709 fiz3 = _mm_setzero_pd();
1711 /* Start inner kernel loop */
1712 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
1715 /* Get j neighbor index, and coordinate index */
1717 jnrB = jjnr[jidx+1];
1718 j_coord_offsetA = DIM*jnrA;
1719 j_coord_offsetB = DIM*jnrB;
1721 /* load j atom coordinates */
1722 gmx_mm_load_4rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1723 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1724 &jy2,&jz2,&jx3,&jy3,&jz3);
1726 /* Calculate displacement vector */
1727 dx00 = _mm_sub_pd(ix0,jx0);
1728 dy00 = _mm_sub_pd(iy0,jy0);
1729 dz00 = _mm_sub_pd(iz0,jz0);
1730 dx11 = _mm_sub_pd(ix1,jx1);
1731 dy11 = _mm_sub_pd(iy1,jy1);
1732 dz11 = _mm_sub_pd(iz1,jz1);
1733 dx12 = _mm_sub_pd(ix1,jx2);
1734 dy12 = _mm_sub_pd(iy1,jy2);
1735 dz12 = _mm_sub_pd(iz1,jz2);
1736 dx13 = _mm_sub_pd(ix1,jx3);
1737 dy13 = _mm_sub_pd(iy1,jy3);
1738 dz13 = _mm_sub_pd(iz1,jz3);
1739 dx21 = _mm_sub_pd(ix2,jx1);
1740 dy21 = _mm_sub_pd(iy2,jy1);
1741 dz21 = _mm_sub_pd(iz2,jz1);
1742 dx22 = _mm_sub_pd(ix2,jx2);
1743 dy22 = _mm_sub_pd(iy2,jy2);
1744 dz22 = _mm_sub_pd(iz2,jz2);
1745 dx23 = _mm_sub_pd(ix2,jx3);
1746 dy23 = _mm_sub_pd(iy2,jy3);
1747 dz23 = _mm_sub_pd(iz2,jz3);
1748 dx31 = _mm_sub_pd(ix3,jx1);
1749 dy31 = _mm_sub_pd(iy3,jy1);
1750 dz31 = _mm_sub_pd(iz3,jz1);
1751 dx32 = _mm_sub_pd(ix3,jx2);
1752 dy32 = _mm_sub_pd(iy3,jy2);
1753 dz32 = _mm_sub_pd(iz3,jz2);
1754 dx33 = _mm_sub_pd(ix3,jx3);
1755 dy33 = _mm_sub_pd(iy3,jy3);
1756 dz33 = _mm_sub_pd(iz3,jz3);
1758 /* Calculate squared distance and things based on it */
1759 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1760 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1761 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1762 rsq13 = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
1763 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1764 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1765 rsq23 = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
1766 rsq31 = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
1767 rsq32 = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
1768 rsq33 = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
1770 rinv00 = gmx_mm_invsqrt_pd(rsq00);
1771 rinv11 = gmx_mm_invsqrt_pd(rsq11);
1772 rinv12 = gmx_mm_invsqrt_pd(rsq12);
1773 rinv13 = gmx_mm_invsqrt_pd(rsq13);
1774 rinv21 = gmx_mm_invsqrt_pd(rsq21);
1775 rinv22 = gmx_mm_invsqrt_pd(rsq22);
1776 rinv23 = gmx_mm_invsqrt_pd(rsq23);
1777 rinv31 = gmx_mm_invsqrt_pd(rsq31);
1778 rinv32 = gmx_mm_invsqrt_pd(rsq32);
1779 rinv33 = gmx_mm_invsqrt_pd(rsq33);
1781 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
1782 rinvsq11 = _mm_mul_pd(rinv11,rinv11);
1783 rinvsq12 = _mm_mul_pd(rinv12,rinv12);
1784 rinvsq13 = _mm_mul_pd(rinv13,rinv13);
1785 rinvsq21 = _mm_mul_pd(rinv21,rinv21);
1786 rinvsq22 = _mm_mul_pd(rinv22,rinv22);
1787 rinvsq23 = _mm_mul_pd(rinv23,rinv23);
1788 rinvsq31 = _mm_mul_pd(rinv31,rinv31);
1789 rinvsq32 = _mm_mul_pd(rinv32,rinv32);
1790 rinvsq33 = _mm_mul_pd(rinv33,rinv33);
1792 fjx0 = _mm_setzero_pd();
1793 fjy0 = _mm_setzero_pd();
1794 fjz0 = _mm_setzero_pd();
1795 fjx1 = _mm_setzero_pd();
1796 fjy1 = _mm_setzero_pd();
1797 fjz1 = _mm_setzero_pd();
1798 fjx2 = _mm_setzero_pd();
1799 fjy2 = _mm_setzero_pd();
1800 fjz2 = _mm_setzero_pd();
1801 fjx3 = _mm_setzero_pd();
1802 fjy3 = _mm_setzero_pd();
1803 fjz3 = _mm_setzero_pd();
1805 /**************************
1806 * CALCULATE INTERACTIONS *
1807 **************************/
1809 if (gmx_mm_any_lt(rsq00,rcutoff2))
1812 r00 = _mm_mul_pd(rsq00,rinv00);
1814 /* Analytical LJ-PME */
1815 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1816 ewcljrsq = _mm_mul_pd(ewclj2,rsq00);
1817 ewclj6 = _mm_mul_pd(ewclj2,_mm_mul_pd(ewclj2,ewclj2));
1818 exponent = gmx_simd_exp_d(ewcljrsq);
1819 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
1820 poly = _mm_mul_pd(exponent,_mm_macc_pd(_mm_mul_pd(ewcljrsq,ewcljrsq),one_half,_mm_sub_pd(one,ewcljrsq)));
1821 /* f6A = 6 * C6grid * (1 - poly) */
1822 f6A = _mm_mul_pd(c6grid_00,_mm_sub_pd(one,poly));
1823 /* f6B = C6grid * exponent * beta^6 */
1824 f6B = _mm_mul_pd(_mm_mul_pd(c6grid_00,one_sixth),_mm_mul_pd(exponent,ewclj6));
1825 /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
1826 fvdw = _mm_mul_pd(_mm_macc_pd(_mm_msub_pd(c12_00,rinvsix,_mm_sub_pd(c6_00,f6A)),rinvsix,f6B),rinvsq00);
1828 cutoff_mask = _mm_cmplt_pd(rsq00,rcutoff2);
1832 fscal = _mm_and_pd(fscal,cutoff_mask);
1834 /* Update vectorial force */
1835 fix0 = _mm_macc_pd(dx00,fscal,fix0);
1836 fiy0 = _mm_macc_pd(dy00,fscal,fiy0);
1837 fiz0 = _mm_macc_pd(dz00,fscal,fiz0);
1839 fjx0 = _mm_macc_pd(dx00,fscal,fjx0);
1840 fjy0 = _mm_macc_pd(dy00,fscal,fjy0);
1841 fjz0 = _mm_macc_pd(dz00,fscal,fjz0);
1845 /**************************
1846 * CALCULATE INTERACTIONS *
1847 **************************/
1849 if (gmx_mm_any_lt(rsq11,rcutoff2))
1852 r11 = _mm_mul_pd(rsq11,rinv11);
1854 /* EWALD ELECTROSTATICS */
1856 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1857 ewrt = _mm_mul_pd(r11,ewtabscale);
1858 ewitab = _mm_cvttpd_epi32(ewrt);
1860 eweps = _mm_frcz_pd(ewrt);
1862 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1864 twoeweps = _mm_add_pd(eweps,eweps);
1865 gmx_mm_load_2pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),ewtab+_mm_extract_epi32(ewitab,1),
1867 felec = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
1868 felec = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
1870 cutoff_mask = _mm_cmplt_pd(rsq11,rcutoff2);
1874 fscal = _mm_and_pd(fscal,cutoff_mask);
1876 /* Update vectorial force */
1877 fix1 = _mm_macc_pd(dx11,fscal,fix1);
1878 fiy1 = _mm_macc_pd(dy11,fscal,fiy1);
1879 fiz1 = _mm_macc_pd(dz11,fscal,fiz1);
1881 fjx1 = _mm_macc_pd(dx11,fscal,fjx1);
1882 fjy1 = _mm_macc_pd(dy11,fscal,fjy1);
1883 fjz1 = _mm_macc_pd(dz11,fscal,fjz1);
1887 /**************************
1888 * CALCULATE INTERACTIONS *
1889 **************************/
1891 if (gmx_mm_any_lt(rsq12,rcutoff2))
1894 r12 = _mm_mul_pd(rsq12,rinv12);
1896 /* EWALD ELECTROSTATICS */
1898 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1899 ewrt = _mm_mul_pd(r12,ewtabscale);
1900 ewitab = _mm_cvttpd_epi32(ewrt);
1902 eweps = _mm_frcz_pd(ewrt);
1904 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1906 twoeweps = _mm_add_pd(eweps,eweps);
1907 gmx_mm_load_2pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),ewtab+_mm_extract_epi32(ewitab,1),
1909 felec = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
1910 felec = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
1912 cutoff_mask = _mm_cmplt_pd(rsq12,rcutoff2);
1916 fscal = _mm_and_pd(fscal,cutoff_mask);
1918 /* Update vectorial force */
1919 fix1 = _mm_macc_pd(dx12,fscal,fix1);
1920 fiy1 = _mm_macc_pd(dy12,fscal,fiy1);
1921 fiz1 = _mm_macc_pd(dz12,fscal,fiz1);
1923 fjx2 = _mm_macc_pd(dx12,fscal,fjx2);
1924 fjy2 = _mm_macc_pd(dy12,fscal,fjy2);
1925 fjz2 = _mm_macc_pd(dz12,fscal,fjz2);
1929 /**************************
1930 * CALCULATE INTERACTIONS *
1931 **************************/
1933 if (gmx_mm_any_lt(rsq13,rcutoff2))
1936 r13 = _mm_mul_pd(rsq13,rinv13);
1938 /* EWALD ELECTROSTATICS */
1940 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1941 ewrt = _mm_mul_pd(r13,ewtabscale);
1942 ewitab = _mm_cvttpd_epi32(ewrt);
1944 eweps = _mm_frcz_pd(ewrt);
1946 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1948 twoeweps = _mm_add_pd(eweps,eweps);
1949 gmx_mm_load_2pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),ewtab+_mm_extract_epi32(ewitab,1),
1951 felec = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
1952 felec = _mm_mul_pd(_mm_mul_pd(qq13,rinv13),_mm_sub_pd(rinvsq13,felec));
1954 cutoff_mask = _mm_cmplt_pd(rsq13,rcutoff2);
1958 fscal = _mm_and_pd(fscal,cutoff_mask);
1960 /* Update vectorial force */
1961 fix1 = _mm_macc_pd(dx13,fscal,fix1);
1962 fiy1 = _mm_macc_pd(dy13,fscal,fiy1);
1963 fiz1 = _mm_macc_pd(dz13,fscal,fiz1);
1965 fjx3 = _mm_macc_pd(dx13,fscal,fjx3);
1966 fjy3 = _mm_macc_pd(dy13,fscal,fjy3);
1967 fjz3 = _mm_macc_pd(dz13,fscal,fjz3);
1971 /**************************
1972 * CALCULATE INTERACTIONS *
1973 **************************/
1975 if (gmx_mm_any_lt(rsq21,rcutoff2))
1978 r21 = _mm_mul_pd(rsq21,rinv21);
1980 /* EWALD ELECTROSTATICS */
1982 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1983 ewrt = _mm_mul_pd(r21,ewtabscale);
1984 ewitab = _mm_cvttpd_epi32(ewrt);
1986 eweps = _mm_frcz_pd(ewrt);
1988 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1990 twoeweps = _mm_add_pd(eweps,eweps);
1991 gmx_mm_load_2pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),ewtab+_mm_extract_epi32(ewitab,1),
1993 felec = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
1994 felec = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
1996 cutoff_mask = _mm_cmplt_pd(rsq21,rcutoff2);
2000 fscal = _mm_and_pd(fscal,cutoff_mask);
2002 /* Update vectorial force */
2003 fix2 = _mm_macc_pd(dx21,fscal,fix2);
2004 fiy2 = _mm_macc_pd(dy21,fscal,fiy2);
2005 fiz2 = _mm_macc_pd(dz21,fscal,fiz2);
2007 fjx1 = _mm_macc_pd(dx21,fscal,fjx1);
2008 fjy1 = _mm_macc_pd(dy21,fscal,fjy1);
2009 fjz1 = _mm_macc_pd(dz21,fscal,fjz1);
2013 /**************************
2014 * CALCULATE INTERACTIONS *
2015 **************************/
2017 if (gmx_mm_any_lt(rsq22,rcutoff2))
2020 r22 = _mm_mul_pd(rsq22,rinv22);
2022 /* EWALD ELECTROSTATICS */
2024 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2025 ewrt = _mm_mul_pd(r22,ewtabscale);
2026 ewitab = _mm_cvttpd_epi32(ewrt);
2028 eweps = _mm_frcz_pd(ewrt);
2030 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2032 twoeweps = _mm_add_pd(eweps,eweps);
2033 gmx_mm_load_2pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),ewtab+_mm_extract_epi32(ewitab,1),
2035 felec = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
2036 felec = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
2038 cutoff_mask = _mm_cmplt_pd(rsq22,rcutoff2);
2042 fscal = _mm_and_pd(fscal,cutoff_mask);
2044 /* Update vectorial force */
2045 fix2 = _mm_macc_pd(dx22,fscal,fix2);
2046 fiy2 = _mm_macc_pd(dy22,fscal,fiy2);
2047 fiz2 = _mm_macc_pd(dz22,fscal,fiz2);
2049 fjx2 = _mm_macc_pd(dx22,fscal,fjx2);
2050 fjy2 = _mm_macc_pd(dy22,fscal,fjy2);
2051 fjz2 = _mm_macc_pd(dz22,fscal,fjz2);
2055 /**************************
2056 * CALCULATE INTERACTIONS *
2057 **************************/
2059 if (gmx_mm_any_lt(rsq23,rcutoff2))
2062 r23 = _mm_mul_pd(rsq23,rinv23);
2064 /* EWALD ELECTROSTATICS */
2066 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2067 ewrt = _mm_mul_pd(r23,ewtabscale);
2068 ewitab = _mm_cvttpd_epi32(ewrt);
2070 eweps = _mm_frcz_pd(ewrt);
2072 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2074 twoeweps = _mm_add_pd(eweps,eweps);
2075 gmx_mm_load_2pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),ewtab+_mm_extract_epi32(ewitab,1),
2077 felec = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
2078 felec = _mm_mul_pd(_mm_mul_pd(qq23,rinv23),_mm_sub_pd(rinvsq23,felec));
2080 cutoff_mask = _mm_cmplt_pd(rsq23,rcutoff2);
2084 fscal = _mm_and_pd(fscal,cutoff_mask);
2086 /* Update vectorial force */
2087 fix2 = _mm_macc_pd(dx23,fscal,fix2);
2088 fiy2 = _mm_macc_pd(dy23,fscal,fiy2);
2089 fiz2 = _mm_macc_pd(dz23,fscal,fiz2);
2091 fjx3 = _mm_macc_pd(dx23,fscal,fjx3);
2092 fjy3 = _mm_macc_pd(dy23,fscal,fjy3);
2093 fjz3 = _mm_macc_pd(dz23,fscal,fjz3);
2097 /**************************
2098 * CALCULATE INTERACTIONS *
2099 **************************/
2101 if (gmx_mm_any_lt(rsq31,rcutoff2))
2104 r31 = _mm_mul_pd(rsq31,rinv31);
2106 /* EWALD ELECTROSTATICS */
2108 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2109 ewrt = _mm_mul_pd(r31,ewtabscale);
2110 ewitab = _mm_cvttpd_epi32(ewrt);
2112 eweps = _mm_frcz_pd(ewrt);
2114 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2116 twoeweps = _mm_add_pd(eweps,eweps);
2117 gmx_mm_load_2pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),ewtab+_mm_extract_epi32(ewitab,1),
2119 felec = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
2120 felec = _mm_mul_pd(_mm_mul_pd(qq31,rinv31),_mm_sub_pd(rinvsq31,felec));
2122 cutoff_mask = _mm_cmplt_pd(rsq31,rcutoff2);
2126 fscal = _mm_and_pd(fscal,cutoff_mask);
2128 /* Update vectorial force */
2129 fix3 = _mm_macc_pd(dx31,fscal,fix3);
2130 fiy3 = _mm_macc_pd(dy31,fscal,fiy3);
2131 fiz3 = _mm_macc_pd(dz31,fscal,fiz3);
2133 fjx1 = _mm_macc_pd(dx31,fscal,fjx1);
2134 fjy1 = _mm_macc_pd(dy31,fscal,fjy1);
2135 fjz1 = _mm_macc_pd(dz31,fscal,fjz1);
2139 /**************************
2140 * CALCULATE INTERACTIONS *
2141 **************************/
2143 if (gmx_mm_any_lt(rsq32,rcutoff2))
2146 r32 = _mm_mul_pd(rsq32,rinv32);
2148 /* EWALD ELECTROSTATICS */
2150 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2151 ewrt = _mm_mul_pd(r32,ewtabscale);
2152 ewitab = _mm_cvttpd_epi32(ewrt);
2154 eweps = _mm_frcz_pd(ewrt);
2156 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2158 twoeweps = _mm_add_pd(eweps,eweps);
2159 gmx_mm_load_2pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),ewtab+_mm_extract_epi32(ewitab,1),
2161 felec = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
2162 felec = _mm_mul_pd(_mm_mul_pd(qq32,rinv32),_mm_sub_pd(rinvsq32,felec));
2164 cutoff_mask = _mm_cmplt_pd(rsq32,rcutoff2);
2168 fscal = _mm_and_pd(fscal,cutoff_mask);
2170 /* Update vectorial force */
2171 fix3 = _mm_macc_pd(dx32,fscal,fix3);
2172 fiy3 = _mm_macc_pd(dy32,fscal,fiy3);
2173 fiz3 = _mm_macc_pd(dz32,fscal,fiz3);
2175 fjx2 = _mm_macc_pd(dx32,fscal,fjx2);
2176 fjy2 = _mm_macc_pd(dy32,fscal,fjy2);
2177 fjz2 = _mm_macc_pd(dz32,fscal,fjz2);
2181 /**************************
2182 * CALCULATE INTERACTIONS *
2183 **************************/
2185 if (gmx_mm_any_lt(rsq33,rcutoff2))
2188 r33 = _mm_mul_pd(rsq33,rinv33);
2190 /* EWALD ELECTROSTATICS */
2192 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2193 ewrt = _mm_mul_pd(r33,ewtabscale);
2194 ewitab = _mm_cvttpd_epi32(ewrt);
2196 eweps = _mm_frcz_pd(ewrt);
2198 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2200 twoeweps = _mm_add_pd(eweps,eweps);
2201 gmx_mm_load_2pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),ewtab+_mm_extract_epi32(ewitab,1),
2203 felec = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
2204 felec = _mm_mul_pd(_mm_mul_pd(qq33,rinv33),_mm_sub_pd(rinvsq33,felec));
2206 cutoff_mask = _mm_cmplt_pd(rsq33,rcutoff2);
2210 fscal = _mm_and_pd(fscal,cutoff_mask);
2212 /* Update vectorial force */
2213 fix3 = _mm_macc_pd(dx33,fscal,fix3);
2214 fiy3 = _mm_macc_pd(dy33,fscal,fiy3);
2215 fiz3 = _mm_macc_pd(dz33,fscal,fiz3);
2217 fjx3 = _mm_macc_pd(dx33,fscal,fjx3);
2218 fjy3 = _mm_macc_pd(dy33,fscal,fjy3);
2219 fjz3 = _mm_macc_pd(dz33,fscal,fjz3);
2223 gmx_mm_decrement_4rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2225 /* Inner loop uses 431 flops */
2228 if(jidx<j_index_end)
2232 j_coord_offsetA = DIM*jnrA;
2234 /* load j atom coordinates */
2235 gmx_mm_load_4rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
2236 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
2237 &jy2,&jz2,&jx3,&jy3,&jz3);
2239 /* Calculate displacement vector */
2240 dx00 = _mm_sub_pd(ix0,jx0);
2241 dy00 = _mm_sub_pd(iy0,jy0);
2242 dz00 = _mm_sub_pd(iz0,jz0);
2243 dx11 = _mm_sub_pd(ix1,jx1);
2244 dy11 = _mm_sub_pd(iy1,jy1);
2245 dz11 = _mm_sub_pd(iz1,jz1);
2246 dx12 = _mm_sub_pd(ix1,jx2);
2247 dy12 = _mm_sub_pd(iy1,jy2);
2248 dz12 = _mm_sub_pd(iz1,jz2);
2249 dx13 = _mm_sub_pd(ix1,jx3);
2250 dy13 = _mm_sub_pd(iy1,jy3);
2251 dz13 = _mm_sub_pd(iz1,jz3);
2252 dx21 = _mm_sub_pd(ix2,jx1);
2253 dy21 = _mm_sub_pd(iy2,jy1);
2254 dz21 = _mm_sub_pd(iz2,jz1);
2255 dx22 = _mm_sub_pd(ix2,jx2);
2256 dy22 = _mm_sub_pd(iy2,jy2);
2257 dz22 = _mm_sub_pd(iz2,jz2);
2258 dx23 = _mm_sub_pd(ix2,jx3);
2259 dy23 = _mm_sub_pd(iy2,jy3);
2260 dz23 = _mm_sub_pd(iz2,jz3);
2261 dx31 = _mm_sub_pd(ix3,jx1);
2262 dy31 = _mm_sub_pd(iy3,jy1);
2263 dz31 = _mm_sub_pd(iz3,jz1);
2264 dx32 = _mm_sub_pd(ix3,jx2);
2265 dy32 = _mm_sub_pd(iy3,jy2);
2266 dz32 = _mm_sub_pd(iz3,jz2);
2267 dx33 = _mm_sub_pd(ix3,jx3);
2268 dy33 = _mm_sub_pd(iy3,jy3);
2269 dz33 = _mm_sub_pd(iz3,jz3);
2271 /* Calculate squared distance and things based on it */
2272 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
2273 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
2274 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
2275 rsq13 = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
2276 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
2277 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
2278 rsq23 = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
2279 rsq31 = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
2280 rsq32 = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
2281 rsq33 = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
2283 rinv00 = gmx_mm_invsqrt_pd(rsq00);
2284 rinv11 = gmx_mm_invsqrt_pd(rsq11);
2285 rinv12 = gmx_mm_invsqrt_pd(rsq12);
2286 rinv13 = gmx_mm_invsqrt_pd(rsq13);
2287 rinv21 = gmx_mm_invsqrt_pd(rsq21);
2288 rinv22 = gmx_mm_invsqrt_pd(rsq22);
2289 rinv23 = gmx_mm_invsqrt_pd(rsq23);
2290 rinv31 = gmx_mm_invsqrt_pd(rsq31);
2291 rinv32 = gmx_mm_invsqrt_pd(rsq32);
2292 rinv33 = gmx_mm_invsqrt_pd(rsq33);
2294 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
2295 rinvsq11 = _mm_mul_pd(rinv11,rinv11);
2296 rinvsq12 = _mm_mul_pd(rinv12,rinv12);
2297 rinvsq13 = _mm_mul_pd(rinv13,rinv13);
2298 rinvsq21 = _mm_mul_pd(rinv21,rinv21);
2299 rinvsq22 = _mm_mul_pd(rinv22,rinv22);
2300 rinvsq23 = _mm_mul_pd(rinv23,rinv23);
2301 rinvsq31 = _mm_mul_pd(rinv31,rinv31);
2302 rinvsq32 = _mm_mul_pd(rinv32,rinv32);
2303 rinvsq33 = _mm_mul_pd(rinv33,rinv33);
2305 fjx0 = _mm_setzero_pd();
2306 fjy0 = _mm_setzero_pd();
2307 fjz0 = _mm_setzero_pd();
2308 fjx1 = _mm_setzero_pd();
2309 fjy1 = _mm_setzero_pd();
2310 fjz1 = _mm_setzero_pd();
2311 fjx2 = _mm_setzero_pd();
2312 fjy2 = _mm_setzero_pd();
2313 fjz2 = _mm_setzero_pd();
2314 fjx3 = _mm_setzero_pd();
2315 fjy3 = _mm_setzero_pd();
2316 fjz3 = _mm_setzero_pd();
2318 /**************************
2319 * CALCULATE INTERACTIONS *
2320 **************************/
2322 if (gmx_mm_any_lt(rsq00,rcutoff2))
2325 r00 = _mm_mul_pd(rsq00,rinv00);
2327 /* Analytical LJ-PME */
2328 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
2329 ewcljrsq = _mm_mul_pd(ewclj2,rsq00);
2330 ewclj6 = _mm_mul_pd(ewclj2,_mm_mul_pd(ewclj2,ewclj2));
2331 exponent = gmx_simd_exp_d(ewcljrsq);
2332 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
2333 poly = _mm_mul_pd(exponent,_mm_macc_pd(_mm_mul_pd(ewcljrsq,ewcljrsq),one_half,_mm_sub_pd(one,ewcljrsq)));
2334 /* f6A = 6 * C6grid * (1 - poly) */
2335 f6A = _mm_mul_pd(c6grid_00,_mm_sub_pd(one,poly));
2336 /* f6B = C6grid * exponent * beta^6 */
2337 f6B = _mm_mul_pd(_mm_mul_pd(c6grid_00,one_sixth),_mm_mul_pd(exponent,ewclj6));
2338 /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
2339 fvdw = _mm_mul_pd(_mm_macc_pd(_mm_msub_pd(c12_00,rinvsix,_mm_sub_pd(c6_00,f6A)),rinvsix,f6B),rinvsq00);
2341 cutoff_mask = _mm_cmplt_pd(rsq00,rcutoff2);
2345 fscal = _mm_and_pd(fscal,cutoff_mask);
2347 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2349 /* Update vectorial force */
2350 fix0 = _mm_macc_pd(dx00,fscal,fix0);
2351 fiy0 = _mm_macc_pd(dy00,fscal,fiy0);
2352 fiz0 = _mm_macc_pd(dz00,fscal,fiz0);
2354 fjx0 = _mm_macc_pd(dx00,fscal,fjx0);
2355 fjy0 = _mm_macc_pd(dy00,fscal,fjy0);
2356 fjz0 = _mm_macc_pd(dz00,fscal,fjz0);
2360 /**************************
2361 * CALCULATE INTERACTIONS *
2362 **************************/
2364 if (gmx_mm_any_lt(rsq11,rcutoff2))
2367 r11 = _mm_mul_pd(rsq11,rinv11);
2369 /* EWALD ELECTROSTATICS */
2371 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2372 ewrt = _mm_mul_pd(r11,ewtabscale);
2373 ewitab = _mm_cvttpd_epi32(ewrt);
2375 eweps = _mm_frcz_pd(ewrt);
2377 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2379 twoeweps = _mm_add_pd(eweps,eweps);
2380 gmx_mm_load_1pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2381 felec = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
2382 felec = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
2384 cutoff_mask = _mm_cmplt_pd(rsq11,rcutoff2);
2388 fscal = _mm_and_pd(fscal,cutoff_mask);
2390 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2392 /* Update vectorial force */
2393 fix1 = _mm_macc_pd(dx11,fscal,fix1);
2394 fiy1 = _mm_macc_pd(dy11,fscal,fiy1);
2395 fiz1 = _mm_macc_pd(dz11,fscal,fiz1);
2397 fjx1 = _mm_macc_pd(dx11,fscal,fjx1);
2398 fjy1 = _mm_macc_pd(dy11,fscal,fjy1);
2399 fjz1 = _mm_macc_pd(dz11,fscal,fjz1);
2403 /**************************
2404 * CALCULATE INTERACTIONS *
2405 **************************/
2407 if (gmx_mm_any_lt(rsq12,rcutoff2))
2410 r12 = _mm_mul_pd(rsq12,rinv12);
2412 /* EWALD ELECTROSTATICS */
2414 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2415 ewrt = _mm_mul_pd(r12,ewtabscale);
2416 ewitab = _mm_cvttpd_epi32(ewrt);
2418 eweps = _mm_frcz_pd(ewrt);
2420 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2422 twoeweps = _mm_add_pd(eweps,eweps);
2423 gmx_mm_load_1pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2424 felec = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
2425 felec = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
2427 cutoff_mask = _mm_cmplt_pd(rsq12,rcutoff2);
2431 fscal = _mm_and_pd(fscal,cutoff_mask);
2433 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2435 /* Update vectorial force */
2436 fix1 = _mm_macc_pd(dx12,fscal,fix1);
2437 fiy1 = _mm_macc_pd(dy12,fscal,fiy1);
2438 fiz1 = _mm_macc_pd(dz12,fscal,fiz1);
2440 fjx2 = _mm_macc_pd(dx12,fscal,fjx2);
2441 fjy2 = _mm_macc_pd(dy12,fscal,fjy2);
2442 fjz2 = _mm_macc_pd(dz12,fscal,fjz2);
2446 /**************************
2447 * CALCULATE INTERACTIONS *
2448 **************************/
2450 if (gmx_mm_any_lt(rsq13,rcutoff2))
2453 r13 = _mm_mul_pd(rsq13,rinv13);
2455 /* EWALD ELECTROSTATICS */
2457 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2458 ewrt = _mm_mul_pd(r13,ewtabscale);
2459 ewitab = _mm_cvttpd_epi32(ewrt);
2461 eweps = _mm_frcz_pd(ewrt);
2463 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2465 twoeweps = _mm_add_pd(eweps,eweps);
2466 gmx_mm_load_1pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2467 felec = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
2468 felec = _mm_mul_pd(_mm_mul_pd(qq13,rinv13),_mm_sub_pd(rinvsq13,felec));
2470 cutoff_mask = _mm_cmplt_pd(rsq13,rcutoff2);
2474 fscal = _mm_and_pd(fscal,cutoff_mask);
2476 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2478 /* Update vectorial force */
2479 fix1 = _mm_macc_pd(dx13,fscal,fix1);
2480 fiy1 = _mm_macc_pd(dy13,fscal,fiy1);
2481 fiz1 = _mm_macc_pd(dz13,fscal,fiz1);
2483 fjx3 = _mm_macc_pd(dx13,fscal,fjx3);
2484 fjy3 = _mm_macc_pd(dy13,fscal,fjy3);
2485 fjz3 = _mm_macc_pd(dz13,fscal,fjz3);
2489 /**************************
2490 * CALCULATE INTERACTIONS *
2491 **************************/
2493 if (gmx_mm_any_lt(rsq21,rcutoff2))
2496 r21 = _mm_mul_pd(rsq21,rinv21);
2498 /* EWALD ELECTROSTATICS */
2500 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2501 ewrt = _mm_mul_pd(r21,ewtabscale);
2502 ewitab = _mm_cvttpd_epi32(ewrt);
2504 eweps = _mm_frcz_pd(ewrt);
2506 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2508 twoeweps = _mm_add_pd(eweps,eweps);
2509 gmx_mm_load_1pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2510 felec = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
2511 felec = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
2513 cutoff_mask = _mm_cmplt_pd(rsq21,rcutoff2);
2517 fscal = _mm_and_pd(fscal,cutoff_mask);
2519 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2521 /* Update vectorial force */
2522 fix2 = _mm_macc_pd(dx21,fscal,fix2);
2523 fiy2 = _mm_macc_pd(dy21,fscal,fiy2);
2524 fiz2 = _mm_macc_pd(dz21,fscal,fiz2);
2526 fjx1 = _mm_macc_pd(dx21,fscal,fjx1);
2527 fjy1 = _mm_macc_pd(dy21,fscal,fjy1);
2528 fjz1 = _mm_macc_pd(dz21,fscal,fjz1);
2532 /**************************
2533 * CALCULATE INTERACTIONS *
2534 **************************/
2536 if (gmx_mm_any_lt(rsq22,rcutoff2))
2539 r22 = _mm_mul_pd(rsq22,rinv22);
2541 /* EWALD ELECTROSTATICS */
2543 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2544 ewrt = _mm_mul_pd(r22,ewtabscale);
2545 ewitab = _mm_cvttpd_epi32(ewrt);
2547 eweps = _mm_frcz_pd(ewrt);
2549 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2551 twoeweps = _mm_add_pd(eweps,eweps);
2552 gmx_mm_load_1pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2553 felec = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
2554 felec = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
2556 cutoff_mask = _mm_cmplt_pd(rsq22,rcutoff2);
2560 fscal = _mm_and_pd(fscal,cutoff_mask);
2562 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2564 /* Update vectorial force */
2565 fix2 = _mm_macc_pd(dx22,fscal,fix2);
2566 fiy2 = _mm_macc_pd(dy22,fscal,fiy2);
2567 fiz2 = _mm_macc_pd(dz22,fscal,fiz2);
2569 fjx2 = _mm_macc_pd(dx22,fscal,fjx2);
2570 fjy2 = _mm_macc_pd(dy22,fscal,fjy2);
2571 fjz2 = _mm_macc_pd(dz22,fscal,fjz2);
2575 /**************************
2576 * CALCULATE INTERACTIONS *
2577 **************************/
2579 if (gmx_mm_any_lt(rsq23,rcutoff2))
2582 r23 = _mm_mul_pd(rsq23,rinv23);
2584 /* EWALD ELECTROSTATICS */
2586 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2587 ewrt = _mm_mul_pd(r23,ewtabscale);
2588 ewitab = _mm_cvttpd_epi32(ewrt);
2590 eweps = _mm_frcz_pd(ewrt);
2592 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2594 twoeweps = _mm_add_pd(eweps,eweps);
2595 gmx_mm_load_1pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2596 felec = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
2597 felec = _mm_mul_pd(_mm_mul_pd(qq23,rinv23),_mm_sub_pd(rinvsq23,felec));
2599 cutoff_mask = _mm_cmplt_pd(rsq23,rcutoff2);
2603 fscal = _mm_and_pd(fscal,cutoff_mask);
2605 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2607 /* Update vectorial force */
2608 fix2 = _mm_macc_pd(dx23,fscal,fix2);
2609 fiy2 = _mm_macc_pd(dy23,fscal,fiy2);
2610 fiz2 = _mm_macc_pd(dz23,fscal,fiz2);
2612 fjx3 = _mm_macc_pd(dx23,fscal,fjx3);
2613 fjy3 = _mm_macc_pd(dy23,fscal,fjy3);
2614 fjz3 = _mm_macc_pd(dz23,fscal,fjz3);
2618 /**************************
2619 * CALCULATE INTERACTIONS *
2620 **************************/
2622 if (gmx_mm_any_lt(rsq31,rcutoff2))
2625 r31 = _mm_mul_pd(rsq31,rinv31);
2627 /* EWALD ELECTROSTATICS */
2629 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2630 ewrt = _mm_mul_pd(r31,ewtabscale);
2631 ewitab = _mm_cvttpd_epi32(ewrt);
2633 eweps = _mm_frcz_pd(ewrt);
2635 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2637 twoeweps = _mm_add_pd(eweps,eweps);
2638 gmx_mm_load_1pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2639 felec = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
2640 felec = _mm_mul_pd(_mm_mul_pd(qq31,rinv31),_mm_sub_pd(rinvsq31,felec));
2642 cutoff_mask = _mm_cmplt_pd(rsq31,rcutoff2);
2646 fscal = _mm_and_pd(fscal,cutoff_mask);
2648 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2650 /* Update vectorial force */
2651 fix3 = _mm_macc_pd(dx31,fscal,fix3);
2652 fiy3 = _mm_macc_pd(dy31,fscal,fiy3);
2653 fiz3 = _mm_macc_pd(dz31,fscal,fiz3);
2655 fjx1 = _mm_macc_pd(dx31,fscal,fjx1);
2656 fjy1 = _mm_macc_pd(dy31,fscal,fjy1);
2657 fjz1 = _mm_macc_pd(dz31,fscal,fjz1);
2661 /**************************
2662 * CALCULATE INTERACTIONS *
2663 **************************/
2665 if (gmx_mm_any_lt(rsq32,rcutoff2))
2668 r32 = _mm_mul_pd(rsq32,rinv32);
2670 /* EWALD ELECTROSTATICS */
2672 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2673 ewrt = _mm_mul_pd(r32,ewtabscale);
2674 ewitab = _mm_cvttpd_epi32(ewrt);
2676 eweps = _mm_frcz_pd(ewrt);
2678 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2680 twoeweps = _mm_add_pd(eweps,eweps);
2681 gmx_mm_load_1pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2682 felec = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
2683 felec = _mm_mul_pd(_mm_mul_pd(qq32,rinv32),_mm_sub_pd(rinvsq32,felec));
2685 cutoff_mask = _mm_cmplt_pd(rsq32,rcutoff2);
2689 fscal = _mm_and_pd(fscal,cutoff_mask);
2691 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2693 /* Update vectorial force */
2694 fix3 = _mm_macc_pd(dx32,fscal,fix3);
2695 fiy3 = _mm_macc_pd(dy32,fscal,fiy3);
2696 fiz3 = _mm_macc_pd(dz32,fscal,fiz3);
2698 fjx2 = _mm_macc_pd(dx32,fscal,fjx2);
2699 fjy2 = _mm_macc_pd(dy32,fscal,fjy2);
2700 fjz2 = _mm_macc_pd(dz32,fscal,fjz2);
2704 /**************************
2705 * CALCULATE INTERACTIONS *
2706 **************************/
2708 if (gmx_mm_any_lt(rsq33,rcutoff2))
2711 r33 = _mm_mul_pd(rsq33,rinv33);
2713 /* EWALD ELECTROSTATICS */
2715 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2716 ewrt = _mm_mul_pd(r33,ewtabscale);
2717 ewitab = _mm_cvttpd_epi32(ewrt);
2719 eweps = _mm_frcz_pd(ewrt);
2721 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2723 twoeweps = _mm_add_pd(eweps,eweps);
2724 gmx_mm_load_1pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2725 felec = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
2726 felec = _mm_mul_pd(_mm_mul_pd(qq33,rinv33),_mm_sub_pd(rinvsq33,felec));
2728 cutoff_mask = _mm_cmplt_pd(rsq33,rcutoff2);
2732 fscal = _mm_and_pd(fscal,cutoff_mask);
2734 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2736 /* Update vectorial force */
2737 fix3 = _mm_macc_pd(dx33,fscal,fix3);
2738 fiy3 = _mm_macc_pd(dy33,fscal,fiy3);
2739 fiz3 = _mm_macc_pd(dz33,fscal,fiz3);
2741 fjx3 = _mm_macc_pd(dx33,fscal,fjx3);
2742 fjy3 = _mm_macc_pd(dy33,fscal,fjy3);
2743 fjz3 = _mm_macc_pd(dz33,fscal,fjz3);
2747 gmx_mm_decrement_4rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2749 /* Inner loop uses 431 flops */
2752 /* End of innermost loop */
2754 gmx_mm_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2755 f+i_coord_offset,fshift+i_shift_offset);
2757 /* Increment number of inner iterations */
2758 inneriter += j_index_end - j_index_start;
2760 /* Outer loop uses 24 flops */
2763 /* Increment number of outer iterations */
2766 /* Update outer/inner flops */
2768 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*431);