2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2012,2013, by the GROMACS development team, led by
5 * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
6 * and including many others, as listed in the AUTHORS file in the
7 * top-level source directory and at http://www.gromacs.org.
9 * GROMACS is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU Lesser General Public License
11 * as published by the Free Software Foundation; either version 2.1
12 * of the License, or (at your option) any later version.
14 * GROMACS is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
17 * Lesser General Public License for more details.
19 * You should have received a copy of the GNU Lesser General Public
20 * License along with GROMACS; if not, see
21 * http://www.gnu.org/licenses, or write to the Free Software Foundation,
22 * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
24 * If you want to redistribute modifications to GROMACS, please
25 * consider that scientific software is very special. Version
26 * control is crucial - bugs must be traceable. We will be happy to
27 * consider code for inclusion in the official distribution, but
28 * derived work must not be called official GROMACS. Details are found
29 * in the README & COPYING files - if they are missing, get the
30 * official version at http://www.gromacs.org.
32 * To help us fund GROMACS development, we humbly ask that you cite
33 * the research papers on the package. Check out http://www.gromacs.org.
36 * Note: this file was generated by the GROMACS 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_ElecEw_VdwLJ_GeomP1P1_VF_avx_128_fma_double
54 * Electrostatics interaction: Ewald
55 * VdW interaction: LennardJones
56 * Geometry: Particle-Particle
57 * Calculate force/pot: PotentialAndForce
60 nb_kernel_ElecEw_VdwLJ_GeomP1P1_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;
84 int vdwjidx0A,vdwjidx0B;
85 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
86 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
87 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
90 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
93 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
94 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
96 __m128d ewtabscale,eweps,twoeweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
98 __m128d dummy_mask,cutoff_mask;
99 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
100 __m128d one = _mm_set1_pd(1.0);
101 __m128d two = _mm_set1_pd(2.0);
107 jindex = nlist->jindex;
109 shiftidx = nlist->shift;
111 shiftvec = fr->shift_vec[0];
112 fshift = fr->fshift[0];
113 facel = _mm_set1_pd(fr->epsfac);
114 charge = mdatoms->chargeA;
115 nvdwtype = fr->ntype;
117 vdwtype = mdatoms->typeA;
119 sh_ewald = _mm_set1_pd(fr->ic->sh_ewald);
120 ewtab = fr->ic->tabq_coul_FDV0;
121 ewtabscale = _mm_set1_pd(fr->ic->tabq_scale);
122 ewtabhalfspace = _mm_set1_pd(0.5/fr->ic->tabq_scale);
124 /* Avoid stupid compiler warnings */
132 /* Start outer loop over neighborlists */
133 for(iidx=0; iidx<nri; iidx++)
135 /* Load shift vector for this list */
136 i_shift_offset = DIM*shiftidx[iidx];
138 /* Load limits for loop over neighbors */
139 j_index_start = jindex[iidx];
140 j_index_end = jindex[iidx+1];
142 /* Get outer coordinate index */
144 i_coord_offset = DIM*inr;
146 /* Load i particle coords and add shift vector */
147 gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
149 fix0 = _mm_setzero_pd();
150 fiy0 = _mm_setzero_pd();
151 fiz0 = _mm_setzero_pd();
153 /* Load parameters for i particles */
154 iq0 = _mm_mul_pd(facel,_mm_load1_pd(charge+inr+0));
155 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
157 /* Reset potential sums */
158 velecsum = _mm_setzero_pd();
159 vvdwsum = _mm_setzero_pd();
161 /* Start inner kernel loop */
162 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
165 /* Get j neighbor index, and coordinate index */
168 j_coord_offsetA = DIM*jnrA;
169 j_coord_offsetB = DIM*jnrB;
171 /* load j atom coordinates */
172 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
175 /* Calculate displacement vector */
176 dx00 = _mm_sub_pd(ix0,jx0);
177 dy00 = _mm_sub_pd(iy0,jy0);
178 dz00 = _mm_sub_pd(iz0,jz0);
180 /* Calculate squared distance and things based on it */
181 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
183 rinv00 = gmx_mm_invsqrt_pd(rsq00);
185 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
187 /* Load parameters for j particles */
188 jq0 = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
189 vdwjidx0A = 2*vdwtype[jnrA+0];
190 vdwjidx0B = 2*vdwtype[jnrB+0];
192 /**************************
193 * CALCULATE INTERACTIONS *
194 **************************/
196 r00 = _mm_mul_pd(rsq00,rinv00);
198 /* Compute parameters for interactions between i and j atoms */
199 qq00 = _mm_mul_pd(iq0,jq0);
200 gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
201 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
203 /* EWALD ELECTROSTATICS */
205 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
206 ewrt = _mm_mul_pd(r00,ewtabscale);
207 ewitab = _mm_cvttpd_epi32(ewrt);
209 eweps = _mm_frcz_pd(ewrt);
211 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
213 twoeweps = _mm_add_pd(eweps,eweps);
214 ewitab = _mm_slli_epi32(ewitab,2);
215 ewtabF = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
216 ewtabD = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
217 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
218 ewtabV = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
219 ewtabFn = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
220 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
221 felec = _mm_macc_pd(eweps,ewtabD,ewtabF);
222 velec = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
223 velec = _mm_mul_pd(qq00,_mm_sub_pd(rinv00,velec));
224 felec = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
226 /* LENNARD-JONES DISPERSION/REPULSION */
228 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
229 vvdw6 = _mm_mul_pd(c6_00,rinvsix);
230 vvdw12 = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
231 vvdw = _mm_msub_pd( vvdw12,one_twelfth, _mm_mul_pd(vvdw6,one_sixth) );
232 fvdw = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
234 /* Update potential sum for this i atom from the interaction with this j atom. */
235 velecsum = _mm_add_pd(velecsum,velec);
236 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
238 fscal = _mm_add_pd(felec,fvdw);
240 /* Update vectorial force */
241 fix0 = _mm_macc_pd(dx00,fscal,fix0);
242 fiy0 = _mm_macc_pd(dy00,fscal,fiy0);
243 fiz0 = _mm_macc_pd(dz00,fscal,fiz0);
245 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,
246 _mm_mul_pd(dx00,fscal),
247 _mm_mul_pd(dy00,fscal),
248 _mm_mul_pd(dz00,fscal));
250 /* Inner loop uses 56 flops */
257 j_coord_offsetA = DIM*jnrA;
259 /* load j atom coordinates */
260 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
263 /* Calculate displacement vector */
264 dx00 = _mm_sub_pd(ix0,jx0);
265 dy00 = _mm_sub_pd(iy0,jy0);
266 dz00 = _mm_sub_pd(iz0,jz0);
268 /* Calculate squared distance and things based on it */
269 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
271 rinv00 = gmx_mm_invsqrt_pd(rsq00);
273 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
275 /* Load parameters for j particles */
276 jq0 = _mm_load_sd(charge+jnrA+0);
277 vdwjidx0A = 2*vdwtype[jnrA+0];
279 /**************************
280 * CALCULATE INTERACTIONS *
281 **************************/
283 r00 = _mm_mul_pd(rsq00,rinv00);
285 /* Compute parameters for interactions between i and j atoms */
286 qq00 = _mm_mul_pd(iq0,jq0);
287 gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
289 /* EWALD ELECTROSTATICS */
291 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
292 ewrt = _mm_mul_pd(r00,ewtabscale);
293 ewitab = _mm_cvttpd_epi32(ewrt);
295 eweps = _mm_frcz_pd(ewrt);
297 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
299 twoeweps = _mm_add_pd(eweps,eweps);
300 ewitab = _mm_slli_epi32(ewitab,2);
301 ewtabF = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
302 ewtabD = _mm_setzero_pd();
303 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
304 ewtabV = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
305 ewtabFn = _mm_setzero_pd();
306 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
307 felec = _mm_macc_pd(eweps,ewtabD,ewtabF);
308 velec = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
309 velec = _mm_mul_pd(qq00,_mm_sub_pd(rinv00,velec));
310 felec = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
312 /* LENNARD-JONES DISPERSION/REPULSION */
314 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
315 vvdw6 = _mm_mul_pd(c6_00,rinvsix);
316 vvdw12 = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
317 vvdw = _mm_msub_pd( vvdw12,one_twelfth, _mm_mul_pd(vvdw6,one_sixth) );
318 fvdw = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
320 /* Update potential sum for this i atom from the interaction with this j atom. */
321 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
322 velecsum = _mm_add_pd(velecsum,velec);
323 vvdw = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
324 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
326 fscal = _mm_add_pd(felec,fvdw);
328 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
330 /* Update vectorial force */
331 fix0 = _mm_macc_pd(dx00,fscal,fix0);
332 fiy0 = _mm_macc_pd(dy00,fscal,fiy0);
333 fiz0 = _mm_macc_pd(dz00,fscal,fiz0);
335 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,
336 _mm_mul_pd(dx00,fscal),
337 _mm_mul_pd(dy00,fscal),
338 _mm_mul_pd(dz00,fscal));
340 /* Inner loop uses 56 flops */
343 /* End of innermost loop */
345 gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
346 f+i_coord_offset,fshift+i_shift_offset);
349 /* Update potential energies */
350 gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
351 gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
353 /* Increment number of inner iterations */
354 inneriter += j_index_end - j_index_start;
356 /* Outer loop uses 9 flops */
359 /* Increment number of outer iterations */
362 /* Update outer/inner flops */
364 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_VF,outeriter*9 + inneriter*56);
367 * Gromacs nonbonded kernel: nb_kernel_ElecEw_VdwLJ_GeomP1P1_F_avx_128_fma_double
368 * Electrostatics interaction: Ewald
369 * VdW interaction: LennardJones
370 * Geometry: Particle-Particle
371 * Calculate force/pot: Force
374 nb_kernel_ElecEw_VdwLJ_GeomP1P1_F_avx_128_fma_double
375 (t_nblist * gmx_restrict nlist,
376 rvec * gmx_restrict xx,
377 rvec * gmx_restrict ff,
378 t_forcerec * gmx_restrict fr,
379 t_mdatoms * gmx_restrict mdatoms,
380 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
381 t_nrnb * gmx_restrict nrnb)
383 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
384 * just 0 for non-waters.
385 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
386 * jnr indices corresponding to data put in the four positions in the SIMD register.
388 int i_shift_offset,i_coord_offset,outeriter,inneriter;
389 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
391 int j_coord_offsetA,j_coord_offsetB;
392 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
394 real *shiftvec,*fshift,*x,*f;
395 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
397 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
398 int vdwjidx0A,vdwjidx0B;
399 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
400 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
401 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
404 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
407 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
408 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
410 __m128d ewtabscale,eweps,twoeweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
412 __m128d dummy_mask,cutoff_mask;
413 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
414 __m128d one = _mm_set1_pd(1.0);
415 __m128d two = _mm_set1_pd(2.0);
421 jindex = nlist->jindex;
423 shiftidx = nlist->shift;
425 shiftvec = fr->shift_vec[0];
426 fshift = fr->fshift[0];
427 facel = _mm_set1_pd(fr->epsfac);
428 charge = mdatoms->chargeA;
429 nvdwtype = fr->ntype;
431 vdwtype = mdatoms->typeA;
433 sh_ewald = _mm_set1_pd(fr->ic->sh_ewald);
434 ewtab = fr->ic->tabq_coul_F;
435 ewtabscale = _mm_set1_pd(fr->ic->tabq_scale);
436 ewtabhalfspace = _mm_set1_pd(0.5/fr->ic->tabq_scale);
438 /* Avoid stupid compiler warnings */
446 /* Start outer loop over neighborlists */
447 for(iidx=0; iidx<nri; iidx++)
449 /* Load shift vector for this list */
450 i_shift_offset = DIM*shiftidx[iidx];
452 /* Load limits for loop over neighbors */
453 j_index_start = jindex[iidx];
454 j_index_end = jindex[iidx+1];
456 /* Get outer coordinate index */
458 i_coord_offset = DIM*inr;
460 /* Load i particle coords and add shift vector */
461 gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
463 fix0 = _mm_setzero_pd();
464 fiy0 = _mm_setzero_pd();
465 fiz0 = _mm_setzero_pd();
467 /* Load parameters for i particles */
468 iq0 = _mm_mul_pd(facel,_mm_load1_pd(charge+inr+0));
469 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
471 /* Start inner kernel loop */
472 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
475 /* Get j neighbor index, and coordinate index */
478 j_coord_offsetA = DIM*jnrA;
479 j_coord_offsetB = DIM*jnrB;
481 /* load j atom coordinates */
482 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
485 /* Calculate displacement vector */
486 dx00 = _mm_sub_pd(ix0,jx0);
487 dy00 = _mm_sub_pd(iy0,jy0);
488 dz00 = _mm_sub_pd(iz0,jz0);
490 /* Calculate squared distance and things based on it */
491 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
493 rinv00 = gmx_mm_invsqrt_pd(rsq00);
495 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
497 /* Load parameters for j particles */
498 jq0 = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
499 vdwjidx0A = 2*vdwtype[jnrA+0];
500 vdwjidx0B = 2*vdwtype[jnrB+0];
502 /**************************
503 * CALCULATE INTERACTIONS *
504 **************************/
506 r00 = _mm_mul_pd(rsq00,rinv00);
508 /* Compute parameters for interactions between i and j atoms */
509 qq00 = _mm_mul_pd(iq0,jq0);
510 gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
511 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
513 /* EWALD ELECTROSTATICS */
515 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
516 ewrt = _mm_mul_pd(r00,ewtabscale);
517 ewitab = _mm_cvttpd_epi32(ewrt);
519 eweps = _mm_frcz_pd(ewrt);
521 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
523 twoeweps = _mm_add_pd(eweps,eweps);
524 gmx_mm_load_2pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),ewtab+_mm_extract_epi32(ewitab,1),
526 felec = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
527 felec = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
529 /* LENNARD-JONES DISPERSION/REPULSION */
531 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
532 fvdw = _mm_mul_pd(_mm_msub_pd(c12_00,rinvsix,c6_00),_mm_mul_pd(rinvsix,rinvsq00));
534 fscal = _mm_add_pd(felec,fvdw);
536 /* Update vectorial force */
537 fix0 = _mm_macc_pd(dx00,fscal,fix0);
538 fiy0 = _mm_macc_pd(dy00,fscal,fiy0);
539 fiz0 = _mm_macc_pd(dz00,fscal,fiz0);
541 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,
542 _mm_mul_pd(dx00,fscal),
543 _mm_mul_pd(dy00,fscal),
544 _mm_mul_pd(dz00,fscal));
546 /* Inner loop uses 46 flops */
553 j_coord_offsetA = DIM*jnrA;
555 /* load j atom coordinates */
556 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
559 /* Calculate displacement vector */
560 dx00 = _mm_sub_pd(ix0,jx0);
561 dy00 = _mm_sub_pd(iy0,jy0);
562 dz00 = _mm_sub_pd(iz0,jz0);
564 /* Calculate squared distance and things based on it */
565 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
567 rinv00 = gmx_mm_invsqrt_pd(rsq00);
569 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
571 /* Load parameters for j particles */
572 jq0 = _mm_load_sd(charge+jnrA+0);
573 vdwjidx0A = 2*vdwtype[jnrA+0];
575 /**************************
576 * CALCULATE INTERACTIONS *
577 **************************/
579 r00 = _mm_mul_pd(rsq00,rinv00);
581 /* Compute parameters for interactions between i and j atoms */
582 qq00 = _mm_mul_pd(iq0,jq0);
583 gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
585 /* EWALD ELECTROSTATICS */
587 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
588 ewrt = _mm_mul_pd(r00,ewtabscale);
589 ewitab = _mm_cvttpd_epi32(ewrt);
591 eweps = _mm_frcz_pd(ewrt);
593 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
595 twoeweps = _mm_add_pd(eweps,eweps);
596 gmx_mm_load_1pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
597 felec = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
598 felec = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
600 /* LENNARD-JONES DISPERSION/REPULSION */
602 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
603 fvdw = _mm_mul_pd(_mm_msub_pd(c12_00,rinvsix,c6_00),_mm_mul_pd(rinvsix,rinvsq00));
605 fscal = _mm_add_pd(felec,fvdw);
607 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
609 /* Update vectorial force */
610 fix0 = _mm_macc_pd(dx00,fscal,fix0);
611 fiy0 = _mm_macc_pd(dy00,fscal,fiy0);
612 fiz0 = _mm_macc_pd(dz00,fscal,fiz0);
614 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,
615 _mm_mul_pd(dx00,fscal),
616 _mm_mul_pd(dy00,fscal),
617 _mm_mul_pd(dz00,fscal));
619 /* Inner loop uses 46 flops */
622 /* End of innermost loop */
624 gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
625 f+i_coord_offset,fshift+i_shift_offset);
627 /* Increment number of inner iterations */
628 inneriter += j_index_end - j_index_start;
630 /* Outer loop uses 7 flops */
633 /* Increment number of outer iterations */
636 /* Update outer/inner flops */
638 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_F,outeriter*7 + inneriter*46);