2 * Note: this file was generated by the Gromacs avx_128_fma_double kernel generator.
4 * This source code is part of
8 * Copyright (c) 2001-2012, The GROMACS Development Team
10 * Gromacs is a library for molecular simulation and trajectory analysis,
11 * written by Erik Lindahl, David van der Spoel, Berk Hess, and others - for
12 * a full list of developers and information, check out http://www.gromacs.org
14 * This program is free software; you can redistribute it and/or modify it under
15 * the terms of the GNU Lesser General Public License as published by the Free
16 * Software Foundation; either version 2 of the License, or (at your option) any
19 * To help fund GROMACS development, we humbly ask that you cite
20 * the papers people have written on it - you can find them on the website.
28 #include "../nb_kernel.h"
29 #include "types/simple.h"
33 #include "gmx_math_x86_avx_128_fma_double.h"
34 #include "kernelutil_x86_avx_128_fma_double.h"
37 * Gromacs nonbonded kernel: nb_kernel_ElecCSTab_VdwLJ_GeomW4P1_VF_avx_128_fma_double
38 * Electrostatics interaction: CubicSplineTable
39 * VdW interaction: LennardJones
40 * Geometry: Water4-Particle
41 * Calculate force/pot: PotentialAndForce
44 nb_kernel_ElecCSTab_VdwLJ_GeomW4P1_VF_avx_128_fma_double
45 (t_nblist * gmx_restrict nlist,
46 rvec * gmx_restrict xx,
47 rvec * gmx_restrict ff,
48 t_forcerec * gmx_restrict fr,
49 t_mdatoms * gmx_restrict mdatoms,
50 nb_kernel_data_t * gmx_restrict kernel_data,
51 t_nrnb * gmx_restrict nrnb)
53 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
54 * just 0 for non-waters.
55 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
56 * jnr indices corresponding to data put in the four positions in the SIMD register.
58 int i_shift_offset,i_coord_offset,outeriter,inneriter;
59 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
61 int j_coord_offsetA,j_coord_offsetB;
62 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
64 real *shiftvec,*fshift,*x,*f;
65 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
67 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
69 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
71 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
73 __m128d ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
74 int vdwjidx0A,vdwjidx0B;
75 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
76 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
77 __m128d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
78 __m128d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
79 __m128d dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
80 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
83 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
86 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
87 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
89 __m128i ifour = _mm_set1_epi32(4);
90 __m128d rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
92 __m128d dummy_mask,cutoff_mask;
93 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
94 __m128d one = _mm_set1_pd(1.0);
95 __m128d two = _mm_set1_pd(2.0);
101 jindex = nlist->jindex;
103 shiftidx = nlist->shift;
105 shiftvec = fr->shift_vec[0];
106 fshift = fr->fshift[0];
107 facel = _mm_set1_pd(fr->epsfac);
108 charge = mdatoms->chargeA;
109 nvdwtype = fr->ntype;
111 vdwtype = mdatoms->typeA;
113 vftab = kernel_data->table_elec->data;
114 vftabscale = _mm_set1_pd(kernel_data->table_elec->scale);
116 /* Setup water-specific parameters */
117 inr = nlist->iinr[0];
118 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
119 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
120 iq3 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
121 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
123 /* Avoid stupid compiler warnings */
131 /* Start outer loop over neighborlists */
132 for(iidx=0; iidx<nri; iidx++)
134 /* Load shift vector for this list */
135 i_shift_offset = DIM*shiftidx[iidx];
137 /* Load limits for loop over neighbors */
138 j_index_start = jindex[iidx];
139 j_index_end = jindex[iidx+1];
141 /* Get outer coordinate index */
143 i_coord_offset = DIM*inr;
145 /* Load i particle coords and add shift vector */
146 gmx_mm_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
147 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
149 fix0 = _mm_setzero_pd();
150 fiy0 = _mm_setzero_pd();
151 fiz0 = _mm_setzero_pd();
152 fix1 = _mm_setzero_pd();
153 fiy1 = _mm_setzero_pd();
154 fiz1 = _mm_setzero_pd();
155 fix2 = _mm_setzero_pd();
156 fiy2 = _mm_setzero_pd();
157 fiz2 = _mm_setzero_pd();
158 fix3 = _mm_setzero_pd();
159 fiy3 = _mm_setzero_pd();
160 fiz3 = _mm_setzero_pd();
162 /* Reset potential sums */
163 velecsum = _mm_setzero_pd();
164 vvdwsum = _mm_setzero_pd();
166 /* Start inner kernel loop */
167 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
170 /* Get j neighbor index, and coordinate index */
173 j_coord_offsetA = DIM*jnrA;
174 j_coord_offsetB = DIM*jnrB;
176 /* load j atom coordinates */
177 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
180 /* Calculate displacement vector */
181 dx00 = _mm_sub_pd(ix0,jx0);
182 dy00 = _mm_sub_pd(iy0,jy0);
183 dz00 = _mm_sub_pd(iz0,jz0);
184 dx10 = _mm_sub_pd(ix1,jx0);
185 dy10 = _mm_sub_pd(iy1,jy0);
186 dz10 = _mm_sub_pd(iz1,jz0);
187 dx20 = _mm_sub_pd(ix2,jx0);
188 dy20 = _mm_sub_pd(iy2,jy0);
189 dz20 = _mm_sub_pd(iz2,jz0);
190 dx30 = _mm_sub_pd(ix3,jx0);
191 dy30 = _mm_sub_pd(iy3,jy0);
192 dz30 = _mm_sub_pd(iz3,jz0);
194 /* Calculate squared distance and things based on it */
195 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
196 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
197 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
198 rsq30 = gmx_mm_calc_rsq_pd(dx30,dy30,dz30);
200 rinv10 = gmx_mm_invsqrt_pd(rsq10);
201 rinv20 = gmx_mm_invsqrt_pd(rsq20);
202 rinv30 = gmx_mm_invsqrt_pd(rsq30);
204 rinvsq00 = gmx_mm_inv_pd(rsq00);
206 /* Load parameters for j particles */
207 jq0 = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
208 vdwjidx0A = 2*vdwtype[jnrA+0];
209 vdwjidx0B = 2*vdwtype[jnrB+0];
211 fjx0 = _mm_setzero_pd();
212 fjy0 = _mm_setzero_pd();
213 fjz0 = _mm_setzero_pd();
215 /**************************
216 * CALCULATE INTERACTIONS *
217 **************************/
219 /* Compute parameters for interactions between i and j atoms */
220 gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
221 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
223 /* LENNARD-JONES DISPERSION/REPULSION */
225 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
226 vvdw6 = _mm_mul_pd(c6_00,rinvsix);
227 vvdw12 = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
228 vvdw = _mm_msub_pd( vvdw12,one_twelfth, _mm_mul_pd(vvdw6,one_sixth) );
229 fvdw = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
231 /* Update potential sum for this i atom from the interaction with this j atom. */
232 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
236 /* Update vectorial force */
237 fix0 = _mm_macc_pd(dx00,fscal,fix0);
238 fiy0 = _mm_macc_pd(dy00,fscal,fiy0);
239 fiz0 = _mm_macc_pd(dz00,fscal,fiz0);
241 fjx0 = _mm_macc_pd(dx00,fscal,fjx0);
242 fjy0 = _mm_macc_pd(dy00,fscal,fjy0);
243 fjz0 = _mm_macc_pd(dz00,fscal,fjz0);
245 /**************************
246 * CALCULATE INTERACTIONS *
247 **************************/
249 r10 = _mm_mul_pd(rsq10,rinv10);
251 /* Compute parameters for interactions between i and j atoms */
252 qq10 = _mm_mul_pd(iq1,jq0);
254 /* Calculate table index by multiplying r with table scale and truncate to integer */
255 rt = _mm_mul_pd(r10,vftabscale);
256 vfitab = _mm_cvttpd_epi32(rt);
258 vfeps = _mm_frcz_pd(rt);
260 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
262 twovfeps = _mm_add_pd(vfeps,vfeps);
263 vfitab = _mm_slli_epi32(vfitab,2);
265 /* CUBIC SPLINE TABLE ELECTROSTATICS */
266 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
267 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
268 GMX_MM_TRANSPOSE2_PD(Y,F);
269 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
270 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
271 GMX_MM_TRANSPOSE2_PD(G,H);
272 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
273 VV = _mm_macc_pd(vfeps,Fp,Y);
274 velec = _mm_mul_pd(qq10,VV);
275 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
276 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq10,FF),_mm_mul_pd(vftabscale,rinv10)));
278 /* Update potential sum for this i atom from the interaction with this j atom. */
279 velecsum = _mm_add_pd(velecsum,velec);
283 /* Update vectorial force */
284 fix1 = _mm_macc_pd(dx10,fscal,fix1);
285 fiy1 = _mm_macc_pd(dy10,fscal,fiy1);
286 fiz1 = _mm_macc_pd(dz10,fscal,fiz1);
288 fjx0 = _mm_macc_pd(dx10,fscal,fjx0);
289 fjy0 = _mm_macc_pd(dy10,fscal,fjy0);
290 fjz0 = _mm_macc_pd(dz10,fscal,fjz0);
292 /**************************
293 * CALCULATE INTERACTIONS *
294 **************************/
296 r20 = _mm_mul_pd(rsq20,rinv20);
298 /* Compute parameters for interactions between i and j atoms */
299 qq20 = _mm_mul_pd(iq2,jq0);
301 /* Calculate table index by multiplying r with table scale and truncate to integer */
302 rt = _mm_mul_pd(r20,vftabscale);
303 vfitab = _mm_cvttpd_epi32(rt);
305 vfeps = _mm_frcz_pd(rt);
307 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
309 twovfeps = _mm_add_pd(vfeps,vfeps);
310 vfitab = _mm_slli_epi32(vfitab,2);
312 /* CUBIC SPLINE TABLE ELECTROSTATICS */
313 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
314 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
315 GMX_MM_TRANSPOSE2_PD(Y,F);
316 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
317 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
318 GMX_MM_TRANSPOSE2_PD(G,H);
319 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
320 VV = _mm_macc_pd(vfeps,Fp,Y);
321 velec = _mm_mul_pd(qq20,VV);
322 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
323 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq20,FF),_mm_mul_pd(vftabscale,rinv20)));
325 /* Update potential sum for this i atom from the interaction with this j atom. */
326 velecsum = _mm_add_pd(velecsum,velec);
330 /* Update vectorial force */
331 fix2 = _mm_macc_pd(dx20,fscal,fix2);
332 fiy2 = _mm_macc_pd(dy20,fscal,fiy2);
333 fiz2 = _mm_macc_pd(dz20,fscal,fiz2);
335 fjx0 = _mm_macc_pd(dx20,fscal,fjx0);
336 fjy0 = _mm_macc_pd(dy20,fscal,fjy0);
337 fjz0 = _mm_macc_pd(dz20,fscal,fjz0);
339 /**************************
340 * CALCULATE INTERACTIONS *
341 **************************/
343 r30 = _mm_mul_pd(rsq30,rinv30);
345 /* Compute parameters for interactions between i and j atoms */
346 qq30 = _mm_mul_pd(iq3,jq0);
348 /* Calculate table index by multiplying r with table scale and truncate to integer */
349 rt = _mm_mul_pd(r30,vftabscale);
350 vfitab = _mm_cvttpd_epi32(rt);
352 vfeps = _mm_frcz_pd(rt);
354 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
356 twovfeps = _mm_add_pd(vfeps,vfeps);
357 vfitab = _mm_slli_epi32(vfitab,2);
359 /* CUBIC SPLINE TABLE ELECTROSTATICS */
360 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
361 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
362 GMX_MM_TRANSPOSE2_PD(Y,F);
363 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
364 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
365 GMX_MM_TRANSPOSE2_PD(G,H);
366 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
367 VV = _mm_macc_pd(vfeps,Fp,Y);
368 velec = _mm_mul_pd(qq30,VV);
369 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
370 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq30,FF),_mm_mul_pd(vftabscale,rinv30)));
372 /* Update potential sum for this i atom from the interaction with this j atom. */
373 velecsum = _mm_add_pd(velecsum,velec);
377 /* Update vectorial force */
378 fix3 = _mm_macc_pd(dx30,fscal,fix3);
379 fiy3 = _mm_macc_pd(dy30,fscal,fiy3);
380 fiz3 = _mm_macc_pd(dz30,fscal,fiz3);
382 fjx0 = _mm_macc_pd(dx30,fscal,fjx0);
383 fjy0 = _mm_macc_pd(dy30,fscal,fjy0);
384 fjz0 = _mm_macc_pd(dz30,fscal,fjz0);
386 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
388 /* Inner loop uses 176 flops */
395 j_coord_offsetA = DIM*jnrA;
397 /* load j atom coordinates */
398 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
401 /* Calculate displacement vector */
402 dx00 = _mm_sub_pd(ix0,jx0);
403 dy00 = _mm_sub_pd(iy0,jy0);
404 dz00 = _mm_sub_pd(iz0,jz0);
405 dx10 = _mm_sub_pd(ix1,jx0);
406 dy10 = _mm_sub_pd(iy1,jy0);
407 dz10 = _mm_sub_pd(iz1,jz0);
408 dx20 = _mm_sub_pd(ix2,jx0);
409 dy20 = _mm_sub_pd(iy2,jy0);
410 dz20 = _mm_sub_pd(iz2,jz0);
411 dx30 = _mm_sub_pd(ix3,jx0);
412 dy30 = _mm_sub_pd(iy3,jy0);
413 dz30 = _mm_sub_pd(iz3,jz0);
415 /* Calculate squared distance and things based on it */
416 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
417 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
418 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
419 rsq30 = gmx_mm_calc_rsq_pd(dx30,dy30,dz30);
421 rinv10 = gmx_mm_invsqrt_pd(rsq10);
422 rinv20 = gmx_mm_invsqrt_pd(rsq20);
423 rinv30 = gmx_mm_invsqrt_pd(rsq30);
425 rinvsq00 = gmx_mm_inv_pd(rsq00);
427 /* Load parameters for j particles */
428 jq0 = _mm_load_sd(charge+jnrA+0);
429 vdwjidx0A = 2*vdwtype[jnrA+0];
431 fjx0 = _mm_setzero_pd();
432 fjy0 = _mm_setzero_pd();
433 fjz0 = _mm_setzero_pd();
435 /**************************
436 * CALCULATE INTERACTIONS *
437 **************************/
439 /* Compute parameters for interactions between i and j atoms */
440 gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
442 /* LENNARD-JONES DISPERSION/REPULSION */
444 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
445 vvdw6 = _mm_mul_pd(c6_00,rinvsix);
446 vvdw12 = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
447 vvdw = _mm_msub_pd( vvdw12,one_twelfth, _mm_mul_pd(vvdw6,one_sixth) );
448 fvdw = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
450 /* Update potential sum for this i atom from the interaction with this j atom. */
451 vvdw = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
452 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
456 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
458 /* Update vectorial force */
459 fix0 = _mm_macc_pd(dx00,fscal,fix0);
460 fiy0 = _mm_macc_pd(dy00,fscal,fiy0);
461 fiz0 = _mm_macc_pd(dz00,fscal,fiz0);
463 fjx0 = _mm_macc_pd(dx00,fscal,fjx0);
464 fjy0 = _mm_macc_pd(dy00,fscal,fjy0);
465 fjz0 = _mm_macc_pd(dz00,fscal,fjz0);
467 /**************************
468 * CALCULATE INTERACTIONS *
469 **************************/
471 r10 = _mm_mul_pd(rsq10,rinv10);
473 /* Compute parameters for interactions between i and j atoms */
474 qq10 = _mm_mul_pd(iq1,jq0);
476 /* Calculate table index by multiplying r with table scale and truncate to integer */
477 rt = _mm_mul_pd(r10,vftabscale);
478 vfitab = _mm_cvttpd_epi32(rt);
480 vfeps = _mm_frcz_pd(rt);
482 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
484 twovfeps = _mm_add_pd(vfeps,vfeps);
485 vfitab = _mm_slli_epi32(vfitab,2);
487 /* CUBIC SPLINE TABLE ELECTROSTATICS */
488 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
489 F = _mm_setzero_pd();
490 GMX_MM_TRANSPOSE2_PD(Y,F);
491 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
492 H = _mm_setzero_pd();
493 GMX_MM_TRANSPOSE2_PD(G,H);
494 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
495 VV = _mm_macc_pd(vfeps,Fp,Y);
496 velec = _mm_mul_pd(qq10,VV);
497 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
498 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq10,FF),_mm_mul_pd(vftabscale,rinv10)));
500 /* Update potential sum for this i atom from the interaction with this j atom. */
501 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
502 velecsum = _mm_add_pd(velecsum,velec);
506 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
508 /* Update vectorial force */
509 fix1 = _mm_macc_pd(dx10,fscal,fix1);
510 fiy1 = _mm_macc_pd(dy10,fscal,fiy1);
511 fiz1 = _mm_macc_pd(dz10,fscal,fiz1);
513 fjx0 = _mm_macc_pd(dx10,fscal,fjx0);
514 fjy0 = _mm_macc_pd(dy10,fscal,fjy0);
515 fjz0 = _mm_macc_pd(dz10,fscal,fjz0);
517 /**************************
518 * CALCULATE INTERACTIONS *
519 **************************/
521 r20 = _mm_mul_pd(rsq20,rinv20);
523 /* Compute parameters for interactions between i and j atoms */
524 qq20 = _mm_mul_pd(iq2,jq0);
526 /* Calculate table index by multiplying r with table scale and truncate to integer */
527 rt = _mm_mul_pd(r20,vftabscale);
528 vfitab = _mm_cvttpd_epi32(rt);
530 vfeps = _mm_frcz_pd(rt);
532 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
534 twovfeps = _mm_add_pd(vfeps,vfeps);
535 vfitab = _mm_slli_epi32(vfitab,2);
537 /* CUBIC SPLINE TABLE ELECTROSTATICS */
538 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
539 F = _mm_setzero_pd();
540 GMX_MM_TRANSPOSE2_PD(Y,F);
541 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
542 H = _mm_setzero_pd();
543 GMX_MM_TRANSPOSE2_PD(G,H);
544 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
545 VV = _mm_macc_pd(vfeps,Fp,Y);
546 velec = _mm_mul_pd(qq20,VV);
547 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
548 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq20,FF),_mm_mul_pd(vftabscale,rinv20)));
550 /* Update potential sum for this i atom from the interaction with this j atom. */
551 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
552 velecsum = _mm_add_pd(velecsum,velec);
556 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
558 /* Update vectorial force */
559 fix2 = _mm_macc_pd(dx20,fscal,fix2);
560 fiy2 = _mm_macc_pd(dy20,fscal,fiy2);
561 fiz2 = _mm_macc_pd(dz20,fscal,fiz2);
563 fjx0 = _mm_macc_pd(dx20,fscal,fjx0);
564 fjy0 = _mm_macc_pd(dy20,fscal,fjy0);
565 fjz0 = _mm_macc_pd(dz20,fscal,fjz0);
567 /**************************
568 * CALCULATE INTERACTIONS *
569 **************************/
571 r30 = _mm_mul_pd(rsq30,rinv30);
573 /* Compute parameters for interactions between i and j atoms */
574 qq30 = _mm_mul_pd(iq3,jq0);
576 /* Calculate table index by multiplying r with table scale and truncate to integer */
577 rt = _mm_mul_pd(r30,vftabscale);
578 vfitab = _mm_cvttpd_epi32(rt);
580 vfeps = _mm_frcz_pd(rt);
582 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
584 twovfeps = _mm_add_pd(vfeps,vfeps);
585 vfitab = _mm_slli_epi32(vfitab,2);
587 /* CUBIC SPLINE TABLE ELECTROSTATICS */
588 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
589 F = _mm_setzero_pd();
590 GMX_MM_TRANSPOSE2_PD(Y,F);
591 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
592 H = _mm_setzero_pd();
593 GMX_MM_TRANSPOSE2_PD(G,H);
594 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
595 VV = _mm_macc_pd(vfeps,Fp,Y);
596 velec = _mm_mul_pd(qq30,VV);
597 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
598 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq30,FF),_mm_mul_pd(vftabscale,rinv30)));
600 /* Update potential sum for this i atom from the interaction with this j atom. */
601 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
602 velecsum = _mm_add_pd(velecsum,velec);
606 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
608 /* Update vectorial force */
609 fix3 = _mm_macc_pd(dx30,fscal,fix3);
610 fiy3 = _mm_macc_pd(dy30,fscal,fiy3);
611 fiz3 = _mm_macc_pd(dz30,fscal,fiz3);
613 fjx0 = _mm_macc_pd(dx30,fscal,fjx0);
614 fjy0 = _mm_macc_pd(dy30,fscal,fjy0);
615 fjz0 = _mm_macc_pd(dz30,fscal,fjz0);
617 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0);
619 /* Inner loop uses 176 flops */
622 /* End of innermost loop */
624 gmx_mm_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
625 f+i_coord_offset,fshift+i_shift_offset);
628 /* Update potential energies */
629 gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
630 gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
632 /* Increment number of inner iterations */
633 inneriter += j_index_end - j_index_start;
635 /* Outer loop uses 26 flops */
638 /* Increment number of outer iterations */
641 /* Update outer/inner flops */
643 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_VF,outeriter*26 + inneriter*176);
646 * Gromacs nonbonded kernel: nb_kernel_ElecCSTab_VdwLJ_GeomW4P1_F_avx_128_fma_double
647 * Electrostatics interaction: CubicSplineTable
648 * VdW interaction: LennardJones
649 * Geometry: Water4-Particle
650 * Calculate force/pot: Force
653 nb_kernel_ElecCSTab_VdwLJ_GeomW4P1_F_avx_128_fma_double
654 (t_nblist * gmx_restrict nlist,
655 rvec * gmx_restrict xx,
656 rvec * gmx_restrict ff,
657 t_forcerec * gmx_restrict fr,
658 t_mdatoms * gmx_restrict mdatoms,
659 nb_kernel_data_t * gmx_restrict kernel_data,
660 t_nrnb * gmx_restrict nrnb)
662 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
663 * just 0 for non-waters.
664 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
665 * jnr indices corresponding to data put in the four positions in the SIMD register.
667 int i_shift_offset,i_coord_offset,outeriter,inneriter;
668 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
670 int j_coord_offsetA,j_coord_offsetB;
671 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
673 real *shiftvec,*fshift,*x,*f;
674 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
676 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
678 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
680 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
682 __m128d ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
683 int vdwjidx0A,vdwjidx0B;
684 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
685 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
686 __m128d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
687 __m128d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
688 __m128d dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
689 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
692 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
695 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
696 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
698 __m128i ifour = _mm_set1_epi32(4);
699 __m128d rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
701 __m128d dummy_mask,cutoff_mask;
702 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
703 __m128d one = _mm_set1_pd(1.0);
704 __m128d two = _mm_set1_pd(2.0);
710 jindex = nlist->jindex;
712 shiftidx = nlist->shift;
714 shiftvec = fr->shift_vec[0];
715 fshift = fr->fshift[0];
716 facel = _mm_set1_pd(fr->epsfac);
717 charge = mdatoms->chargeA;
718 nvdwtype = fr->ntype;
720 vdwtype = mdatoms->typeA;
722 vftab = kernel_data->table_elec->data;
723 vftabscale = _mm_set1_pd(kernel_data->table_elec->scale);
725 /* Setup water-specific parameters */
726 inr = nlist->iinr[0];
727 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
728 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
729 iq3 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
730 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
732 /* Avoid stupid compiler warnings */
740 /* Start outer loop over neighborlists */
741 for(iidx=0; iidx<nri; iidx++)
743 /* Load shift vector for this list */
744 i_shift_offset = DIM*shiftidx[iidx];
746 /* Load limits for loop over neighbors */
747 j_index_start = jindex[iidx];
748 j_index_end = jindex[iidx+1];
750 /* Get outer coordinate index */
752 i_coord_offset = DIM*inr;
754 /* Load i particle coords and add shift vector */
755 gmx_mm_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
756 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
758 fix0 = _mm_setzero_pd();
759 fiy0 = _mm_setzero_pd();
760 fiz0 = _mm_setzero_pd();
761 fix1 = _mm_setzero_pd();
762 fiy1 = _mm_setzero_pd();
763 fiz1 = _mm_setzero_pd();
764 fix2 = _mm_setzero_pd();
765 fiy2 = _mm_setzero_pd();
766 fiz2 = _mm_setzero_pd();
767 fix3 = _mm_setzero_pd();
768 fiy3 = _mm_setzero_pd();
769 fiz3 = _mm_setzero_pd();
771 /* Start inner kernel loop */
772 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
775 /* Get j neighbor index, and coordinate index */
778 j_coord_offsetA = DIM*jnrA;
779 j_coord_offsetB = DIM*jnrB;
781 /* load j atom coordinates */
782 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
785 /* Calculate displacement vector */
786 dx00 = _mm_sub_pd(ix0,jx0);
787 dy00 = _mm_sub_pd(iy0,jy0);
788 dz00 = _mm_sub_pd(iz0,jz0);
789 dx10 = _mm_sub_pd(ix1,jx0);
790 dy10 = _mm_sub_pd(iy1,jy0);
791 dz10 = _mm_sub_pd(iz1,jz0);
792 dx20 = _mm_sub_pd(ix2,jx0);
793 dy20 = _mm_sub_pd(iy2,jy0);
794 dz20 = _mm_sub_pd(iz2,jz0);
795 dx30 = _mm_sub_pd(ix3,jx0);
796 dy30 = _mm_sub_pd(iy3,jy0);
797 dz30 = _mm_sub_pd(iz3,jz0);
799 /* Calculate squared distance and things based on it */
800 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
801 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
802 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
803 rsq30 = gmx_mm_calc_rsq_pd(dx30,dy30,dz30);
805 rinv10 = gmx_mm_invsqrt_pd(rsq10);
806 rinv20 = gmx_mm_invsqrt_pd(rsq20);
807 rinv30 = gmx_mm_invsqrt_pd(rsq30);
809 rinvsq00 = gmx_mm_inv_pd(rsq00);
811 /* Load parameters for j particles */
812 jq0 = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
813 vdwjidx0A = 2*vdwtype[jnrA+0];
814 vdwjidx0B = 2*vdwtype[jnrB+0];
816 fjx0 = _mm_setzero_pd();
817 fjy0 = _mm_setzero_pd();
818 fjz0 = _mm_setzero_pd();
820 /**************************
821 * CALCULATE INTERACTIONS *
822 **************************/
824 /* Compute parameters for interactions between i and j atoms */
825 gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
826 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
828 /* LENNARD-JONES DISPERSION/REPULSION */
830 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
831 fvdw = _mm_mul_pd(_mm_msub_pd(c12_00,rinvsix,c6_00),_mm_mul_pd(rinvsix,rinvsq00));
835 /* Update vectorial force */
836 fix0 = _mm_macc_pd(dx00,fscal,fix0);
837 fiy0 = _mm_macc_pd(dy00,fscal,fiy0);
838 fiz0 = _mm_macc_pd(dz00,fscal,fiz0);
840 fjx0 = _mm_macc_pd(dx00,fscal,fjx0);
841 fjy0 = _mm_macc_pd(dy00,fscal,fjy0);
842 fjz0 = _mm_macc_pd(dz00,fscal,fjz0);
844 /**************************
845 * CALCULATE INTERACTIONS *
846 **************************/
848 r10 = _mm_mul_pd(rsq10,rinv10);
850 /* Compute parameters for interactions between i and j atoms */
851 qq10 = _mm_mul_pd(iq1,jq0);
853 /* Calculate table index by multiplying r with table scale and truncate to integer */
854 rt = _mm_mul_pd(r10,vftabscale);
855 vfitab = _mm_cvttpd_epi32(rt);
857 vfeps = _mm_frcz_pd(rt);
859 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
861 twovfeps = _mm_add_pd(vfeps,vfeps);
862 vfitab = _mm_slli_epi32(vfitab,2);
864 /* CUBIC SPLINE TABLE ELECTROSTATICS */
865 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
866 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
867 GMX_MM_TRANSPOSE2_PD(Y,F);
868 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
869 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
870 GMX_MM_TRANSPOSE2_PD(G,H);
871 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
872 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
873 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq10,FF),_mm_mul_pd(vftabscale,rinv10)));
877 /* Update vectorial force */
878 fix1 = _mm_macc_pd(dx10,fscal,fix1);
879 fiy1 = _mm_macc_pd(dy10,fscal,fiy1);
880 fiz1 = _mm_macc_pd(dz10,fscal,fiz1);
882 fjx0 = _mm_macc_pd(dx10,fscal,fjx0);
883 fjy0 = _mm_macc_pd(dy10,fscal,fjy0);
884 fjz0 = _mm_macc_pd(dz10,fscal,fjz0);
886 /**************************
887 * CALCULATE INTERACTIONS *
888 **************************/
890 r20 = _mm_mul_pd(rsq20,rinv20);
892 /* Compute parameters for interactions between i and j atoms */
893 qq20 = _mm_mul_pd(iq2,jq0);
895 /* Calculate table index by multiplying r with table scale and truncate to integer */
896 rt = _mm_mul_pd(r20,vftabscale);
897 vfitab = _mm_cvttpd_epi32(rt);
899 vfeps = _mm_frcz_pd(rt);
901 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
903 twovfeps = _mm_add_pd(vfeps,vfeps);
904 vfitab = _mm_slli_epi32(vfitab,2);
906 /* CUBIC SPLINE TABLE ELECTROSTATICS */
907 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
908 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
909 GMX_MM_TRANSPOSE2_PD(Y,F);
910 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
911 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
912 GMX_MM_TRANSPOSE2_PD(G,H);
913 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
914 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
915 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq20,FF),_mm_mul_pd(vftabscale,rinv20)));
919 /* Update vectorial force */
920 fix2 = _mm_macc_pd(dx20,fscal,fix2);
921 fiy2 = _mm_macc_pd(dy20,fscal,fiy2);
922 fiz2 = _mm_macc_pd(dz20,fscal,fiz2);
924 fjx0 = _mm_macc_pd(dx20,fscal,fjx0);
925 fjy0 = _mm_macc_pd(dy20,fscal,fjy0);
926 fjz0 = _mm_macc_pd(dz20,fscal,fjz0);
928 /**************************
929 * CALCULATE INTERACTIONS *
930 **************************/
932 r30 = _mm_mul_pd(rsq30,rinv30);
934 /* Compute parameters for interactions between i and j atoms */
935 qq30 = _mm_mul_pd(iq3,jq0);
937 /* Calculate table index by multiplying r with table scale and truncate to integer */
938 rt = _mm_mul_pd(r30,vftabscale);
939 vfitab = _mm_cvttpd_epi32(rt);
941 vfeps = _mm_frcz_pd(rt);
943 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
945 twovfeps = _mm_add_pd(vfeps,vfeps);
946 vfitab = _mm_slli_epi32(vfitab,2);
948 /* CUBIC SPLINE TABLE ELECTROSTATICS */
949 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
950 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
951 GMX_MM_TRANSPOSE2_PD(Y,F);
952 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
953 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
954 GMX_MM_TRANSPOSE2_PD(G,H);
955 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
956 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
957 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq30,FF),_mm_mul_pd(vftabscale,rinv30)));
961 /* Update vectorial force */
962 fix3 = _mm_macc_pd(dx30,fscal,fix3);
963 fiy3 = _mm_macc_pd(dy30,fscal,fiy3);
964 fiz3 = _mm_macc_pd(dz30,fscal,fiz3);
966 fjx0 = _mm_macc_pd(dx30,fscal,fjx0);
967 fjy0 = _mm_macc_pd(dy30,fscal,fjy0);
968 fjz0 = _mm_macc_pd(dz30,fscal,fjz0);
970 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
972 /* Inner loop uses 159 flops */
979 j_coord_offsetA = DIM*jnrA;
981 /* load j atom coordinates */
982 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
985 /* Calculate displacement vector */
986 dx00 = _mm_sub_pd(ix0,jx0);
987 dy00 = _mm_sub_pd(iy0,jy0);
988 dz00 = _mm_sub_pd(iz0,jz0);
989 dx10 = _mm_sub_pd(ix1,jx0);
990 dy10 = _mm_sub_pd(iy1,jy0);
991 dz10 = _mm_sub_pd(iz1,jz0);
992 dx20 = _mm_sub_pd(ix2,jx0);
993 dy20 = _mm_sub_pd(iy2,jy0);
994 dz20 = _mm_sub_pd(iz2,jz0);
995 dx30 = _mm_sub_pd(ix3,jx0);
996 dy30 = _mm_sub_pd(iy3,jy0);
997 dz30 = _mm_sub_pd(iz3,jz0);
999 /* Calculate squared distance and things based on it */
1000 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1001 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
1002 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
1003 rsq30 = gmx_mm_calc_rsq_pd(dx30,dy30,dz30);
1005 rinv10 = gmx_mm_invsqrt_pd(rsq10);
1006 rinv20 = gmx_mm_invsqrt_pd(rsq20);
1007 rinv30 = gmx_mm_invsqrt_pd(rsq30);
1009 rinvsq00 = gmx_mm_inv_pd(rsq00);
1011 /* Load parameters for j particles */
1012 jq0 = _mm_load_sd(charge+jnrA+0);
1013 vdwjidx0A = 2*vdwtype[jnrA+0];
1015 fjx0 = _mm_setzero_pd();
1016 fjy0 = _mm_setzero_pd();
1017 fjz0 = _mm_setzero_pd();
1019 /**************************
1020 * CALCULATE INTERACTIONS *
1021 **************************/
1023 /* Compute parameters for interactions between i and j atoms */
1024 gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
1026 /* LENNARD-JONES DISPERSION/REPULSION */
1028 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1029 fvdw = _mm_mul_pd(_mm_msub_pd(c12_00,rinvsix,c6_00),_mm_mul_pd(rinvsix,rinvsq00));
1033 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1035 /* Update vectorial force */
1036 fix0 = _mm_macc_pd(dx00,fscal,fix0);
1037 fiy0 = _mm_macc_pd(dy00,fscal,fiy0);
1038 fiz0 = _mm_macc_pd(dz00,fscal,fiz0);
1040 fjx0 = _mm_macc_pd(dx00,fscal,fjx0);
1041 fjy0 = _mm_macc_pd(dy00,fscal,fjy0);
1042 fjz0 = _mm_macc_pd(dz00,fscal,fjz0);
1044 /**************************
1045 * CALCULATE INTERACTIONS *
1046 **************************/
1048 r10 = _mm_mul_pd(rsq10,rinv10);
1050 /* Compute parameters for interactions between i and j atoms */
1051 qq10 = _mm_mul_pd(iq1,jq0);
1053 /* Calculate table index by multiplying r with table scale and truncate to integer */
1054 rt = _mm_mul_pd(r10,vftabscale);
1055 vfitab = _mm_cvttpd_epi32(rt);
1057 vfeps = _mm_frcz_pd(rt);
1059 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1061 twovfeps = _mm_add_pd(vfeps,vfeps);
1062 vfitab = _mm_slli_epi32(vfitab,2);
1064 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1065 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1066 F = _mm_setzero_pd();
1067 GMX_MM_TRANSPOSE2_PD(Y,F);
1068 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1069 H = _mm_setzero_pd();
1070 GMX_MM_TRANSPOSE2_PD(G,H);
1071 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1072 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1073 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq10,FF),_mm_mul_pd(vftabscale,rinv10)));
1077 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1079 /* Update vectorial force */
1080 fix1 = _mm_macc_pd(dx10,fscal,fix1);
1081 fiy1 = _mm_macc_pd(dy10,fscal,fiy1);
1082 fiz1 = _mm_macc_pd(dz10,fscal,fiz1);
1084 fjx0 = _mm_macc_pd(dx10,fscal,fjx0);
1085 fjy0 = _mm_macc_pd(dy10,fscal,fjy0);
1086 fjz0 = _mm_macc_pd(dz10,fscal,fjz0);
1088 /**************************
1089 * CALCULATE INTERACTIONS *
1090 **************************/
1092 r20 = _mm_mul_pd(rsq20,rinv20);
1094 /* Compute parameters for interactions between i and j atoms */
1095 qq20 = _mm_mul_pd(iq2,jq0);
1097 /* Calculate table index by multiplying r with table scale and truncate to integer */
1098 rt = _mm_mul_pd(r20,vftabscale);
1099 vfitab = _mm_cvttpd_epi32(rt);
1101 vfeps = _mm_frcz_pd(rt);
1103 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1105 twovfeps = _mm_add_pd(vfeps,vfeps);
1106 vfitab = _mm_slli_epi32(vfitab,2);
1108 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1109 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1110 F = _mm_setzero_pd();
1111 GMX_MM_TRANSPOSE2_PD(Y,F);
1112 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1113 H = _mm_setzero_pd();
1114 GMX_MM_TRANSPOSE2_PD(G,H);
1115 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1116 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1117 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq20,FF),_mm_mul_pd(vftabscale,rinv20)));
1121 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1123 /* Update vectorial force */
1124 fix2 = _mm_macc_pd(dx20,fscal,fix2);
1125 fiy2 = _mm_macc_pd(dy20,fscal,fiy2);
1126 fiz2 = _mm_macc_pd(dz20,fscal,fiz2);
1128 fjx0 = _mm_macc_pd(dx20,fscal,fjx0);
1129 fjy0 = _mm_macc_pd(dy20,fscal,fjy0);
1130 fjz0 = _mm_macc_pd(dz20,fscal,fjz0);
1132 /**************************
1133 * CALCULATE INTERACTIONS *
1134 **************************/
1136 r30 = _mm_mul_pd(rsq30,rinv30);
1138 /* Compute parameters for interactions between i and j atoms */
1139 qq30 = _mm_mul_pd(iq3,jq0);
1141 /* Calculate table index by multiplying r with table scale and truncate to integer */
1142 rt = _mm_mul_pd(r30,vftabscale);
1143 vfitab = _mm_cvttpd_epi32(rt);
1145 vfeps = _mm_frcz_pd(rt);
1147 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1149 twovfeps = _mm_add_pd(vfeps,vfeps);
1150 vfitab = _mm_slli_epi32(vfitab,2);
1152 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1153 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1154 F = _mm_setzero_pd();
1155 GMX_MM_TRANSPOSE2_PD(Y,F);
1156 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1157 H = _mm_setzero_pd();
1158 GMX_MM_TRANSPOSE2_PD(G,H);
1159 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1160 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1161 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq30,FF),_mm_mul_pd(vftabscale,rinv30)));
1165 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1167 /* Update vectorial force */
1168 fix3 = _mm_macc_pd(dx30,fscal,fix3);
1169 fiy3 = _mm_macc_pd(dy30,fscal,fiy3);
1170 fiz3 = _mm_macc_pd(dz30,fscal,fiz3);
1172 fjx0 = _mm_macc_pd(dx30,fscal,fjx0);
1173 fjy0 = _mm_macc_pd(dy30,fscal,fjy0);
1174 fjz0 = _mm_macc_pd(dz30,fscal,fjz0);
1176 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1178 /* Inner loop uses 159 flops */
1181 /* End of innermost loop */
1183 gmx_mm_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1184 f+i_coord_offset,fshift+i_shift_offset);
1186 /* Increment number of inner iterations */
1187 inneriter += j_index_end - j_index_start;
1189 /* Outer loop uses 24 flops */
1192 /* Increment number of outer iterations */
1195 /* Update outer/inner flops */
1197 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_F,outeriter*24 + inneriter*159);