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 sse4_1_double kernel generator.
44 #include "../nb_kernel.h"
45 #include "gromacs/legacyheaders/types/simple.h"
46 #include "gromacs/math/vec.h"
47 #include "gromacs/legacyheaders/nrnb.h"
49 #include "gromacs/simd/math_x86_sse4_1_double.h"
50 #include "kernelutil_x86_sse4_1_double.h"
53 * Gromacs nonbonded kernel: nb_kernel_ElecEw_VdwNone_GeomP1P1_VF_sse4_1_double
54 * Electrostatics interaction: Ewald
55 * VdW interaction: None
56 * Geometry: Particle-Particle
57 * Calculate force/pot: PotentialAndForce
60 nb_kernel_ElecEw_VdwNone_GeomP1P1_VF_sse4_1_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 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
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;
110 sh_ewald = _mm_set1_pd(fr->ic->sh_ewald);
111 ewtab = fr->ic->tabq_coul_FDV0;
112 ewtabscale = _mm_set1_pd(fr->ic->tabq_scale);
113 ewtabhalfspace = _mm_set1_pd(0.5/fr->ic->tabq_scale);
115 /* Avoid stupid compiler warnings */
123 /* Start outer loop over neighborlists */
124 for(iidx=0; iidx<nri; iidx++)
126 /* Load shift vector for this list */
127 i_shift_offset = DIM*shiftidx[iidx];
129 /* Load limits for loop over neighbors */
130 j_index_start = jindex[iidx];
131 j_index_end = jindex[iidx+1];
133 /* Get outer coordinate index */
135 i_coord_offset = DIM*inr;
137 /* Load i particle coords and add shift vector */
138 gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
140 fix0 = _mm_setzero_pd();
141 fiy0 = _mm_setzero_pd();
142 fiz0 = _mm_setzero_pd();
144 /* Load parameters for i particles */
145 iq0 = _mm_mul_pd(facel,_mm_load1_pd(charge+inr+0));
147 /* Reset potential sums */
148 velecsum = _mm_setzero_pd();
150 /* Start inner kernel loop */
151 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
154 /* Get j neighbor index, and coordinate index */
157 j_coord_offsetA = DIM*jnrA;
158 j_coord_offsetB = DIM*jnrB;
160 /* load j atom coordinates */
161 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
164 /* Calculate displacement vector */
165 dx00 = _mm_sub_pd(ix0,jx0);
166 dy00 = _mm_sub_pd(iy0,jy0);
167 dz00 = _mm_sub_pd(iz0,jz0);
169 /* Calculate squared distance and things based on it */
170 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
172 rinv00 = gmx_mm_invsqrt_pd(rsq00);
174 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
176 /* Load parameters for j particles */
177 jq0 = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
179 /**************************
180 * CALCULATE INTERACTIONS *
181 **************************/
183 r00 = _mm_mul_pd(rsq00,rinv00);
185 /* Compute parameters for interactions between i and j atoms */
186 qq00 = _mm_mul_pd(iq0,jq0);
188 /* EWALD ELECTROSTATICS */
190 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
191 ewrt = _mm_mul_pd(r00,ewtabscale);
192 ewitab = _mm_cvttpd_epi32(ewrt);
193 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
194 ewitab = _mm_slli_epi32(ewitab,2);
195 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
196 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
197 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
198 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
199 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
200 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
201 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
202 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
203 velec = _mm_mul_pd(qq00,_mm_sub_pd(rinv00,velec));
204 felec = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
206 /* Update potential sum for this i atom from the interaction with this j atom. */
207 velecsum = _mm_add_pd(velecsum,velec);
211 /* Calculate temporary vectorial force */
212 tx = _mm_mul_pd(fscal,dx00);
213 ty = _mm_mul_pd(fscal,dy00);
214 tz = _mm_mul_pd(fscal,dz00);
216 /* Update vectorial force */
217 fix0 = _mm_add_pd(fix0,tx);
218 fiy0 = _mm_add_pd(fiy0,ty);
219 fiz0 = _mm_add_pd(fiz0,tz);
221 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,tx,ty,tz);
223 /* Inner loop uses 41 flops */
230 j_coord_offsetA = DIM*jnrA;
232 /* load j atom coordinates */
233 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
236 /* Calculate displacement vector */
237 dx00 = _mm_sub_pd(ix0,jx0);
238 dy00 = _mm_sub_pd(iy0,jy0);
239 dz00 = _mm_sub_pd(iz0,jz0);
241 /* Calculate squared distance and things based on it */
242 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
244 rinv00 = gmx_mm_invsqrt_pd(rsq00);
246 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
248 /* Load parameters for j particles */
249 jq0 = _mm_load_sd(charge+jnrA+0);
251 /**************************
252 * CALCULATE INTERACTIONS *
253 **************************/
255 r00 = _mm_mul_pd(rsq00,rinv00);
257 /* Compute parameters for interactions between i and j atoms */
258 qq00 = _mm_mul_pd(iq0,jq0);
260 /* EWALD ELECTROSTATICS */
262 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
263 ewrt = _mm_mul_pd(r00,ewtabscale);
264 ewitab = _mm_cvttpd_epi32(ewrt);
265 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
266 ewitab = _mm_slli_epi32(ewitab,2);
267 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
268 ewtabD = _mm_setzero_pd();
269 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
270 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
271 ewtabFn = _mm_setzero_pd();
272 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
273 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
274 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
275 velec = _mm_mul_pd(qq00,_mm_sub_pd(rinv00,velec));
276 felec = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
278 /* Update potential sum for this i atom from the interaction with this j atom. */
279 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
280 velecsum = _mm_add_pd(velecsum,velec);
284 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
286 /* Calculate temporary vectorial force */
287 tx = _mm_mul_pd(fscal,dx00);
288 ty = _mm_mul_pd(fscal,dy00);
289 tz = _mm_mul_pd(fscal,dz00);
291 /* Update vectorial force */
292 fix0 = _mm_add_pd(fix0,tx);
293 fiy0 = _mm_add_pd(fiy0,ty);
294 fiz0 = _mm_add_pd(fiz0,tz);
296 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,tx,ty,tz);
298 /* Inner loop uses 41 flops */
301 /* End of innermost loop */
303 gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
304 f+i_coord_offset,fshift+i_shift_offset);
307 /* Update potential energies */
308 gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
310 /* Increment number of inner iterations */
311 inneriter += j_index_end - j_index_start;
313 /* Outer loop uses 8 flops */
316 /* Increment number of outer iterations */
319 /* Update outer/inner flops */
321 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VF,outeriter*8 + inneriter*41);
324 * Gromacs nonbonded kernel: nb_kernel_ElecEw_VdwNone_GeomP1P1_F_sse4_1_double
325 * Electrostatics interaction: Ewald
326 * VdW interaction: None
327 * Geometry: Particle-Particle
328 * Calculate force/pot: Force
331 nb_kernel_ElecEw_VdwNone_GeomP1P1_F_sse4_1_double
332 (t_nblist * gmx_restrict nlist,
333 rvec * gmx_restrict xx,
334 rvec * gmx_restrict ff,
335 t_forcerec * gmx_restrict fr,
336 t_mdatoms * gmx_restrict mdatoms,
337 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
338 t_nrnb * gmx_restrict nrnb)
340 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
341 * just 0 for non-waters.
342 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
343 * jnr indices corresponding to data put in the four positions in the SIMD register.
345 int i_shift_offset,i_coord_offset,outeriter,inneriter;
346 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
348 int j_coord_offsetA,j_coord_offsetB;
349 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
351 real *shiftvec,*fshift,*x,*f;
352 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
354 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
355 int vdwjidx0A,vdwjidx0B;
356 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
357 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
358 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
361 __m128d ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
363 __m128d dummy_mask,cutoff_mask;
364 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
365 __m128d one = _mm_set1_pd(1.0);
366 __m128d two = _mm_set1_pd(2.0);
372 jindex = nlist->jindex;
374 shiftidx = nlist->shift;
376 shiftvec = fr->shift_vec[0];
377 fshift = fr->fshift[0];
378 facel = _mm_set1_pd(fr->epsfac);
379 charge = mdatoms->chargeA;
381 sh_ewald = _mm_set1_pd(fr->ic->sh_ewald);
382 ewtab = fr->ic->tabq_coul_F;
383 ewtabscale = _mm_set1_pd(fr->ic->tabq_scale);
384 ewtabhalfspace = _mm_set1_pd(0.5/fr->ic->tabq_scale);
386 /* Avoid stupid compiler warnings */
394 /* Start outer loop over neighborlists */
395 for(iidx=0; iidx<nri; iidx++)
397 /* Load shift vector for this list */
398 i_shift_offset = DIM*shiftidx[iidx];
400 /* Load limits for loop over neighbors */
401 j_index_start = jindex[iidx];
402 j_index_end = jindex[iidx+1];
404 /* Get outer coordinate index */
406 i_coord_offset = DIM*inr;
408 /* Load i particle coords and add shift vector */
409 gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
411 fix0 = _mm_setzero_pd();
412 fiy0 = _mm_setzero_pd();
413 fiz0 = _mm_setzero_pd();
415 /* Load parameters for i particles */
416 iq0 = _mm_mul_pd(facel,_mm_load1_pd(charge+inr+0));
418 /* Start inner kernel loop */
419 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
422 /* Get j neighbor index, and coordinate index */
425 j_coord_offsetA = DIM*jnrA;
426 j_coord_offsetB = DIM*jnrB;
428 /* load j atom coordinates */
429 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
432 /* Calculate displacement vector */
433 dx00 = _mm_sub_pd(ix0,jx0);
434 dy00 = _mm_sub_pd(iy0,jy0);
435 dz00 = _mm_sub_pd(iz0,jz0);
437 /* Calculate squared distance and things based on it */
438 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
440 rinv00 = gmx_mm_invsqrt_pd(rsq00);
442 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
444 /* Load parameters for j particles */
445 jq0 = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
447 /**************************
448 * CALCULATE INTERACTIONS *
449 **************************/
451 r00 = _mm_mul_pd(rsq00,rinv00);
453 /* Compute parameters for interactions between i and j atoms */
454 qq00 = _mm_mul_pd(iq0,jq0);
456 /* EWALD ELECTROSTATICS */
458 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
459 ewrt = _mm_mul_pd(r00,ewtabscale);
460 ewitab = _mm_cvttpd_epi32(ewrt);
461 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
462 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
464 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
465 felec = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
469 /* Calculate temporary vectorial force */
470 tx = _mm_mul_pd(fscal,dx00);
471 ty = _mm_mul_pd(fscal,dy00);
472 tz = _mm_mul_pd(fscal,dz00);
474 /* Update vectorial force */
475 fix0 = _mm_add_pd(fix0,tx);
476 fiy0 = _mm_add_pd(fiy0,ty);
477 fiz0 = _mm_add_pd(fiz0,tz);
479 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,tx,ty,tz);
481 /* Inner loop uses 36 flops */
488 j_coord_offsetA = DIM*jnrA;
490 /* load j atom coordinates */
491 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
494 /* Calculate displacement vector */
495 dx00 = _mm_sub_pd(ix0,jx0);
496 dy00 = _mm_sub_pd(iy0,jy0);
497 dz00 = _mm_sub_pd(iz0,jz0);
499 /* Calculate squared distance and things based on it */
500 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
502 rinv00 = gmx_mm_invsqrt_pd(rsq00);
504 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
506 /* Load parameters for j particles */
507 jq0 = _mm_load_sd(charge+jnrA+0);
509 /**************************
510 * CALCULATE INTERACTIONS *
511 **************************/
513 r00 = _mm_mul_pd(rsq00,rinv00);
515 /* Compute parameters for interactions between i and j atoms */
516 qq00 = _mm_mul_pd(iq0,jq0);
518 /* EWALD ELECTROSTATICS */
520 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
521 ewrt = _mm_mul_pd(r00,ewtabscale);
522 ewitab = _mm_cvttpd_epi32(ewrt);
523 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
524 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
525 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
526 felec = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
530 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
532 /* Calculate temporary vectorial force */
533 tx = _mm_mul_pd(fscal,dx00);
534 ty = _mm_mul_pd(fscal,dy00);
535 tz = _mm_mul_pd(fscal,dz00);
537 /* Update vectorial force */
538 fix0 = _mm_add_pd(fix0,tx);
539 fiy0 = _mm_add_pd(fiy0,ty);
540 fiz0 = _mm_add_pd(fiz0,tz);
542 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,tx,ty,tz);
544 /* Inner loop uses 36 flops */
547 /* End of innermost loop */
549 gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
550 f+i_coord_offset,fshift+i_shift_offset);
552 /* Increment number of inner iterations */
553 inneriter += j_index_end - j_index_start;
555 /* Outer loop uses 7 flops */
558 /* Increment number of outer iterations */
561 /* Update outer/inner flops */
563 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_F,outeriter*7 + inneriter*36);