2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
5 * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
6 * and including many others, as listed in the AUTHORS file in the
7 * top-level source directory and at http://www.gromacs.org.
9 * GROMACS is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU Lesser General Public License
11 * as published by the Free Software Foundation; either version 2.1
12 * of the License, or (at your option) any later version.
14 * GROMACS is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
17 * Lesser General Public License for more details.
19 * You should have received a copy of the GNU Lesser General Public
20 * License along with GROMACS; if not, see
21 * http://www.gnu.org/licenses, or write to the Free Software Foundation,
22 * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
24 * If you want to redistribute modifications to GROMACS, please
25 * consider that scientific software is very special. Version
26 * control is crucial - bugs must be traceable. We will be happy to
27 * consider code for inclusion in the official distribution, but
28 * derived work must not be called official GROMACS. Details are found
29 * in the README & COPYING files - if they are missing, get the
30 * official version at http://www.gromacs.org.
32 * To help us fund GROMACS development, we humbly ask that you cite
33 * the research papers on the package. Check out http://www.gromacs.org.
36 * Note: this file was generated by the GROMACS sse2_double kernel generator.
42 #include "../nb_kernel.h"
43 #include "types/simple.h"
44 #include "gromacs/math/vec.h"
47 #include "gromacs/simd/math_x86_sse2_double.h"
48 #include "kernelutil_x86_sse2_double.h"
51 * Gromacs nonbonded kernel: nb_kernel_ElecRFCut_VdwNone_GeomP1P1_VF_sse2_double
52 * Electrostatics interaction: ReactionField
53 * VdW interaction: None
54 * Geometry: Particle-Particle
55 * Calculate force/pot: PotentialAndForce
58 nb_kernel_ElecRFCut_VdwNone_GeomP1P1_VF_sse2_double
59 (t_nblist * gmx_restrict nlist,
60 rvec * gmx_restrict xx,
61 rvec * gmx_restrict ff,
62 t_forcerec * gmx_restrict fr,
63 t_mdatoms * gmx_restrict mdatoms,
64 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
65 t_nrnb * gmx_restrict nrnb)
67 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
68 * just 0 for non-waters.
69 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
70 * jnr indices corresponding to data put in the four positions in the SIMD register.
72 int i_shift_offset,i_coord_offset,outeriter,inneriter;
73 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
75 int j_coord_offsetA,j_coord_offsetB;
76 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
78 real *shiftvec,*fshift,*x,*f;
79 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
81 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
82 int vdwjidx0A,vdwjidx0B;
83 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
84 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
85 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
87 __m128d dummy_mask,cutoff_mask;
88 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
89 __m128d one = _mm_set1_pd(1.0);
90 __m128d two = _mm_set1_pd(2.0);
96 jindex = nlist->jindex;
98 shiftidx = nlist->shift;
100 shiftvec = fr->shift_vec[0];
101 fshift = fr->fshift[0];
102 facel = _mm_set1_pd(fr->epsfac);
103 charge = mdatoms->chargeA;
104 krf = _mm_set1_pd(fr->ic->k_rf);
105 krf2 = _mm_set1_pd(fr->ic->k_rf*2.0);
106 crf = _mm_set1_pd(fr->ic->c_rf);
108 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
109 rcutoff_scalar = fr->rcoulomb;
110 rcutoff = _mm_set1_pd(rcutoff_scalar);
111 rcutoff2 = _mm_mul_pd(rcutoff,rcutoff);
113 /* Avoid stupid compiler warnings */
121 /* Start outer loop over neighborlists */
122 for(iidx=0; iidx<nri; iidx++)
124 /* Load shift vector for this list */
125 i_shift_offset = DIM*shiftidx[iidx];
127 /* Load limits for loop over neighbors */
128 j_index_start = jindex[iidx];
129 j_index_end = jindex[iidx+1];
131 /* Get outer coordinate index */
133 i_coord_offset = DIM*inr;
135 /* Load i particle coords and add shift vector */
136 gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
138 fix0 = _mm_setzero_pd();
139 fiy0 = _mm_setzero_pd();
140 fiz0 = _mm_setzero_pd();
142 /* Load parameters for i particles */
143 iq0 = _mm_mul_pd(facel,_mm_load1_pd(charge+inr+0));
145 /* Reset potential sums */
146 velecsum = _mm_setzero_pd();
148 /* Start inner kernel loop */
149 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
152 /* Get j neighbor index, and coordinate index */
155 j_coord_offsetA = DIM*jnrA;
156 j_coord_offsetB = DIM*jnrB;
158 /* load j atom coordinates */
159 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
162 /* Calculate displacement vector */
163 dx00 = _mm_sub_pd(ix0,jx0);
164 dy00 = _mm_sub_pd(iy0,jy0);
165 dz00 = _mm_sub_pd(iz0,jz0);
167 /* Calculate squared distance and things based on it */
168 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
170 rinv00 = gmx_mm_invsqrt_pd(rsq00);
172 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
174 /* Load parameters for j particles */
175 jq0 = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
177 /**************************
178 * CALCULATE INTERACTIONS *
179 **************************/
181 if (gmx_mm_any_lt(rsq00,rcutoff2))
184 /* Compute parameters for interactions between i and j atoms */
185 qq00 = _mm_mul_pd(iq0,jq0);
187 /* REACTION-FIELD ELECTROSTATICS */
188 velec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_add_pd(rinv00,_mm_mul_pd(krf,rsq00)),crf));
189 felec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
191 cutoff_mask = _mm_cmplt_pd(rsq00,rcutoff2);
193 /* Update potential sum for this i atom from the interaction with this j atom. */
194 velec = _mm_and_pd(velec,cutoff_mask);
195 velecsum = _mm_add_pd(velecsum,velec);
199 fscal = _mm_and_pd(fscal,cutoff_mask);
201 /* Calculate temporary vectorial force */
202 tx = _mm_mul_pd(fscal,dx00);
203 ty = _mm_mul_pd(fscal,dy00);
204 tz = _mm_mul_pd(fscal,dz00);
206 /* Update vectorial force */
207 fix0 = _mm_add_pd(fix0,tx);
208 fiy0 = _mm_add_pd(fiy0,ty);
209 fiz0 = _mm_add_pd(fiz0,tz);
211 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,tx,ty,tz);
215 /* Inner loop uses 36 flops */
222 j_coord_offsetA = DIM*jnrA;
224 /* load j atom coordinates */
225 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
228 /* Calculate displacement vector */
229 dx00 = _mm_sub_pd(ix0,jx0);
230 dy00 = _mm_sub_pd(iy0,jy0);
231 dz00 = _mm_sub_pd(iz0,jz0);
233 /* Calculate squared distance and things based on it */
234 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
236 rinv00 = gmx_mm_invsqrt_pd(rsq00);
238 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
240 /* Load parameters for j particles */
241 jq0 = _mm_load_sd(charge+jnrA+0);
243 /**************************
244 * CALCULATE INTERACTIONS *
245 **************************/
247 if (gmx_mm_any_lt(rsq00,rcutoff2))
250 /* Compute parameters for interactions between i and j atoms */
251 qq00 = _mm_mul_pd(iq0,jq0);
253 /* REACTION-FIELD ELECTROSTATICS */
254 velec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_add_pd(rinv00,_mm_mul_pd(krf,rsq00)),crf));
255 felec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
257 cutoff_mask = _mm_cmplt_pd(rsq00,rcutoff2);
259 /* Update potential sum for this i atom from the interaction with this j atom. */
260 velec = _mm_and_pd(velec,cutoff_mask);
261 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
262 velecsum = _mm_add_pd(velecsum,velec);
266 fscal = _mm_and_pd(fscal,cutoff_mask);
268 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
270 /* Calculate temporary vectorial force */
271 tx = _mm_mul_pd(fscal,dx00);
272 ty = _mm_mul_pd(fscal,dy00);
273 tz = _mm_mul_pd(fscal,dz00);
275 /* Update vectorial force */
276 fix0 = _mm_add_pd(fix0,tx);
277 fiy0 = _mm_add_pd(fiy0,ty);
278 fiz0 = _mm_add_pd(fiz0,tz);
280 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,tx,ty,tz);
284 /* Inner loop uses 36 flops */
287 /* End of innermost loop */
289 gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
290 f+i_coord_offset,fshift+i_shift_offset);
293 /* Update potential energies */
294 gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
296 /* Increment number of inner iterations */
297 inneriter += j_index_end - j_index_start;
299 /* Outer loop uses 8 flops */
302 /* Increment number of outer iterations */
305 /* Update outer/inner flops */
307 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VF,outeriter*8 + inneriter*36);
310 * Gromacs nonbonded kernel: nb_kernel_ElecRFCut_VdwNone_GeomP1P1_F_sse2_double
311 * Electrostatics interaction: ReactionField
312 * VdW interaction: None
313 * Geometry: Particle-Particle
314 * Calculate force/pot: Force
317 nb_kernel_ElecRFCut_VdwNone_GeomP1P1_F_sse2_double
318 (t_nblist * gmx_restrict nlist,
319 rvec * gmx_restrict xx,
320 rvec * gmx_restrict ff,
321 t_forcerec * gmx_restrict fr,
322 t_mdatoms * gmx_restrict mdatoms,
323 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
324 t_nrnb * gmx_restrict nrnb)
326 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
327 * just 0 for non-waters.
328 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
329 * jnr indices corresponding to data put in the four positions in the SIMD register.
331 int i_shift_offset,i_coord_offset,outeriter,inneriter;
332 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
334 int j_coord_offsetA,j_coord_offsetB;
335 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
337 real *shiftvec,*fshift,*x,*f;
338 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
340 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
341 int vdwjidx0A,vdwjidx0B;
342 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
343 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
344 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
346 __m128d dummy_mask,cutoff_mask;
347 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
348 __m128d one = _mm_set1_pd(1.0);
349 __m128d two = _mm_set1_pd(2.0);
355 jindex = nlist->jindex;
357 shiftidx = nlist->shift;
359 shiftvec = fr->shift_vec[0];
360 fshift = fr->fshift[0];
361 facel = _mm_set1_pd(fr->epsfac);
362 charge = mdatoms->chargeA;
363 krf = _mm_set1_pd(fr->ic->k_rf);
364 krf2 = _mm_set1_pd(fr->ic->k_rf*2.0);
365 crf = _mm_set1_pd(fr->ic->c_rf);
367 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
368 rcutoff_scalar = fr->rcoulomb;
369 rcutoff = _mm_set1_pd(rcutoff_scalar);
370 rcutoff2 = _mm_mul_pd(rcutoff,rcutoff);
372 /* Avoid stupid compiler warnings */
380 /* Start outer loop over neighborlists */
381 for(iidx=0; iidx<nri; iidx++)
383 /* Load shift vector for this list */
384 i_shift_offset = DIM*shiftidx[iidx];
386 /* Load limits for loop over neighbors */
387 j_index_start = jindex[iidx];
388 j_index_end = jindex[iidx+1];
390 /* Get outer coordinate index */
392 i_coord_offset = DIM*inr;
394 /* Load i particle coords and add shift vector */
395 gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
397 fix0 = _mm_setzero_pd();
398 fiy0 = _mm_setzero_pd();
399 fiz0 = _mm_setzero_pd();
401 /* Load parameters for i particles */
402 iq0 = _mm_mul_pd(facel,_mm_load1_pd(charge+inr+0));
404 /* Start inner kernel loop */
405 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
408 /* Get j neighbor index, and coordinate index */
411 j_coord_offsetA = DIM*jnrA;
412 j_coord_offsetB = DIM*jnrB;
414 /* load j atom coordinates */
415 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
418 /* Calculate displacement vector */
419 dx00 = _mm_sub_pd(ix0,jx0);
420 dy00 = _mm_sub_pd(iy0,jy0);
421 dz00 = _mm_sub_pd(iz0,jz0);
423 /* Calculate squared distance and things based on it */
424 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
426 rinv00 = gmx_mm_invsqrt_pd(rsq00);
428 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
430 /* Load parameters for j particles */
431 jq0 = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
433 /**************************
434 * CALCULATE INTERACTIONS *
435 **************************/
437 if (gmx_mm_any_lt(rsq00,rcutoff2))
440 /* Compute parameters for interactions between i and j atoms */
441 qq00 = _mm_mul_pd(iq0,jq0);
443 /* REACTION-FIELD ELECTROSTATICS */
444 felec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
446 cutoff_mask = _mm_cmplt_pd(rsq00,rcutoff2);
450 fscal = _mm_and_pd(fscal,cutoff_mask);
452 /* Calculate temporary vectorial force */
453 tx = _mm_mul_pd(fscal,dx00);
454 ty = _mm_mul_pd(fscal,dy00);
455 tz = _mm_mul_pd(fscal,dz00);
457 /* Update vectorial force */
458 fix0 = _mm_add_pd(fix0,tx);
459 fiy0 = _mm_add_pd(fiy0,ty);
460 fiz0 = _mm_add_pd(fiz0,tz);
462 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,tx,ty,tz);
466 /* Inner loop uses 30 flops */
473 j_coord_offsetA = DIM*jnrA;
475 /* load j atom coordinates */
476 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
479 /* Calculate displacement vector */
480 dx00 = _mm_sub_pd(ix0,jx0);
481 dy00 = _mm_sub_pd(iy0,jy0);
482 dz00 = _mm_sub_pd(iz0,jz0);
484 /* Calculate squared distance and things based on it */
485 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
487 rinv00 = gmx_mm_invsqrt_pd(rsq00);
489 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
491 /* Load parameters for j particles */
492 jq0 = _mm_load_sd(charge+jnrA+0);
494 /**************************
495 * CALCULATE INTERACTIONS *
496 **************************/
498 if (gmx_mm_any_lt(rsq00,rcutoff2))
501 /* Compute parameters for interactions between i and j atoms */
502 qq00 = _mm_mul_pd(iq0,jq0);
504 /* REACTION-FIELD ELECTROSTATICS */
505 felec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
507 cutoff_mask = _mm_cmplt_pd(rsq00,rcutoff2);
511 fscal = _mm_and_pd(fscal,cutoff_mask);
513 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
515 /* Calculate temporary vectorial force */
516 tx = _mm_mul_pd(fscal,dx00);
517 ty = _mm_mul_pd(fscal,dy00);
518 tz = _mm_mul_pd(fscal,dz00);
520 /* Update vectorial force */
521 fix0 = _mm_add_pd(fix0,tx);
522 fiy0 = _mm_add_pd(fiy0,ty);
523 fiz0 = _mm_add_pd(fiz0,tz);
525 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,tx,ty,tz);
529 /* Inner loop uses 30 flops */
532 /* End of innermost loop */
534 gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
535 f+i_coord_offset,fshift+i_shift_offset);
537 /* Increment number of inner iterations */
538 inneriter += j_index_end - j_index_start;
540 /* Outer loop uses 7 flops */
543 /* Increment number of outer iterations */
546 /* Update outer/inner flops */
548 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_F,outeriter*7 + inneriter*30);