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 sse4_1_double kernel generator.
44 #include "../nb_kernel.h"
45 #include "types/simple.h"
49 #include "gmx_math_x86_sse4_1_double.h"
50 #include "kernelutil_x86_sse4_1_double.h"
53 * Gromacs nonbonded kernel: nb_kernel_ElecCoul_VdwNone_GeomP1P1_VF_sse4_1_double
54 * Electrostatics interaction: Coulomb
55 * VdW interaction: None
56 * Geometry: Particle-Particle
57 * Calculate force/pot: PotentialAndForce
60 nb_kernel_ElecCoul_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;
89 __m128d dummy_mask,cutoff_mask;
90 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
91 __m128d one = _mm_set1_pd(1.0);
92 __m128d two = _mm_set1_pd(2.0);
98 jindex = nlist->jindex;
100 shiftidx = nlist->shift;
102 shiftvec = fr->shift_vec[0];
103 fshift = fr->fshift[0];
104 facel = _mm_set1_pd(fr->epsfac);
105 charge = mdatoms->chargeA;
107 /* Avoid stupid compiler warnings */
115 /* Start outer loop over neighborlists */
116 for(iidx=0; iidx<nri; iidx++)
118 /* Load shift vector for this list */
119 i_shift_offset = DIM*shiftidx[iidx];
121 /* Load limits for loop over neighbors */
122 j_index_start = jindex[iidx];
123 j_index_end = jindex[iidx+1];
125 /* Get outer coordinate index */
127 i_coord_offset = DIM*inr;
129 /* Load i particle coords and add shift vector */
130 gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
132 fix0 = _mm_setzero_pd();
133 fiy0 = _mm_setzero_pd();
134 fiz0 = _mm_setzero_pd();
136 /* Load parameters for i particles */
137 iq0 = _mm_mul_pd(facel,_mm_load1_pd(charge+inr+0));
139 /* Reset potential sums */
140 velecsum = _mm_setzero_pd();
142 /* Start inner kernel loop */
143 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
146 /* Get j neighbor index, and coordinate index */
149 j_coord_offsetA = DIM*jnrA;
150 j_coord_offsetB = DIM*jnrB;
152 /* load j atom coordinates */
153 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
156 /* Calculate displacement vector */
157 dx00 = _mm_sub_pd(ix0,jx0);
158 dy00 = _mm_sub_pd(iy0,jy0);
159 dz00 = _mm_sub_pd(iz0,jz0);
161 /* Calculate squared distance and things based on it */
162 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
164 rinv00 = gmx_mm_invsqrt_pd(rsq00);
166 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
168 /* Load parameters for j particles */
169 jq0 = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
171 /**************************
172 * CALCULATE INTERACTIONS *
173 **************************/
175 /* Compute parameters for interactions between i and j atoms */
176 qq00 = _mm_mul_pd(iq0,jq0);
178 /* COULOMB ELECTROSTATICS */
179 velec = _mm_mul_pd(qq00,rinv00);
180 felec = _mm_mul_pd(velec,rinvsq00);
182 /* Update potential sum for this i atom from the interaction with this j atom. */
183 velecsum = _mm_add_pd(velecsum,velec);
187 /* Calculate temporary vectorial force */
188 tx = _mm_mul_pd(fscal,dx00);
189 ty = _mm_mul_pd(fscal,dy00);
190 tz = _mm_mul_pd(fscal,dz00);
192 /* Update vectorial force */
193 fix0 = _mm_add_pd(fix0,tx);
194 fiy0 = _mm_add_pd(fiy0,ty);
195 fiz0 = _mm_add_pd(fiz0,tz);
197 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,tx,ty,tz);
199 /* Inner loop uses 28 flops */
206 j_coord_offsetA = DIM*jnrA;
208 /* load j atom coordinates */
209 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
212 /* Calculate displacement vector */
213 dx00 = _mm_sub_pd(ix0,jx0);
214 dy00 = _mm_sub_pd(iy0,jy0);
215 dz00 = _mm_sub_pd(iz0,jz0);
217 /* Calculate squared distance and things based on it */
218 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
220 rinv00 = gmx_mm_invsqrt_pd(rsq00);
222 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
224 /* Load parameters for j particles */
225 jq0 = _mm_load_sd(charge+jnrA+0);
227 /**************************
228 * CALCULATE INTERACTIONS *
229 **************************/
231 /* Compute parameters for interactions between i and j atoms */
232 qq00 = _mm_mul_pd(iq0,jq0);
234 /* COULOMB ELECTROSTATICS */
235 velec = _mm_mul_pd(qq00,rinv00);
236 felec = _mm_mul_pd(velec,rinvsq00);
238 /* Update potential sum for this i atom from the interaction with this j atom. */
239 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
240 velecsum = _mm_add_pd(velecsum,velec);
244 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
246 /* Calculate temporary vectorial force */
247 tx = _mm_mul_pd(fscal,dx00);
248 ty = _mm_mul_pd(fscal,dy00);
249 tz = _mm_mul_pd(fscal,dz00);
251 /* Update vectorial force */
252 fix0 = _mm_add_pd(fix0,tx);
253 fiy0 = _mm_add_pd(fiy0,ty);
254 fiz0 = _mm_add_pd(fiz0,tz);
256 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,tx,ty,tz);
258 /* Inner loop uses 28 flops */
261 /* End of innermost loop */
263 gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
264 f+i_coord_offset,fshift+i_shift_offset);
267 /* Update potential energies */
268 gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
270 /* Increment number of inner iterations */
271 inneriter += j_index_end - j_index_start;
273 /* Outer loop uses 8 flops */
276 /* Increment number of outer iterations */
279 /* Update outer/inner flops */
281 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VF,outeriter*8 + inneriter*28);
284 * Gromacs nonbonded kernel: nb_kernel_ElecCoul_VdwNone_GeomP1P1_F_sse4_1_double
285 * Electrostatics interaction: Coulomb
286 * VdW interaction: None
287 * Geometry: Particle-Particle
288 * Calculate force/pot: Force
291 nb_kernel_ElecCoul_VdwNone_GeomP1P1_F_sse4_1_double
292 (t_nblist * gmx_restrict nlist,
293 rvec * gmx_restrict xx,
294 rvec * gmx_restrict ff,
295 t_forcerec * gmx_restrict fr,
296 t_mdatoms * gmx_restrict mdatoms,
297 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
298 t_nrnb * gmx_restrict nrnb)
300 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
301 * just 0 for non-waters.
302 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
303 * jnr indices corresponding to data put in the four positions in the SIMD register.
305 int i_shift_offset,i_coord_offset,outeriter,inneriter;
306 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
308 int j_coord_offsetA,j_coord_offsetB;
309 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
311 real *shiftvec,*fshift,*x,*f;
312 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
314 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
315 int vdwjidx0A,vdwjidx0B;
316 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
317 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
318 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
320 __m128d dummy_mask,cutoff_mask;
321 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
322 __m128d one = _mm_set1_pd(1.0);
323 __m128d two = _mm_set1_pd(2.0);
329 jindex = nlist->jindex;
331 shiftidx = nlist->shift;
333 shiftvec = fr->shift_vec[0];
334 fshift = fr->fshift[0];
335 facel = _mm_set1_pd(fr->epsfac);
336 charge = mdatoms->chargeA;
338 /* Avoid stupid compiler warnings */
346 /* Start outer loop over neighborlists */
347 for(iidx=0; iidx<nri; iidx++)
349 /* Load shift vector for this list */
350 i_shift_offset = DIM*shiftidx[iidx];
352 /* Load limits for loop over neighbors */
353 j_index_start = jindex[iidx];
354 j_index_end = jindex[iidx+1];
356 /* Get outer coordinate index */
358 i_coord_offset = DIM*inr;
360 /* Load i particle coords and add shift vector */
361 gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
363 fix0 = _mm_setzero_pd();
364 fiy0 = _mm_setzero_pd();
365 fiz0 = _mm_setzero_pd();
367 /* Load parameters for i particles */
368 iq0 = _mm_mul_pd(facel,_mm_load1_pd(charge+inr+0));
370 /* Start inner kernel loop */
371 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
374 /* Get j neighbor index, and coordinate index */
377 j_coord_offsetA = DIM*jnrA;
378 j_coord_offsetB = DIM*jnrB;
380 /* load j atom coordinates */
381 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
384 /* Calculate displacement vector */
385 dx00 = _mm_sub_pd(ix0,jx0);
386 dy00 = _mm_sub_pd(iy0,jy0);
387 dz00 = _mm_sub_pd(iz0,jz0);
389 /* Calculate squared distance and things based on it */
390 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
392 rinv00 = gmx_mm_invsqrt_pd(rsq00);
394 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
396 /* Load parameters for j particles */
397 jq0 = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
399 /**************************
400 * CALCULATE INTERACTIONS *
401 **************************/
403 /* Compute parameters for interactions between i and j atoms */
404 qq00 = _mm_mul_pd(iq0,jq0);
406 /* COULOMB ELECTROSTATICS */
407 velec = _mm_mul_pd(qq00,rinv00);
408 felec = _mm_mul_pd(velec,rinvsq00);
412 /* Calculate temporary vectorial force */
413 tx = _mm_mul_pd(fscal,dx00);
414 ty = _mm_mul_pd(fscal,dy00);
415 tz = _mm_mul_pd(fscal,dz00);
417 /* Update vectorial force */
418 fix0 = _mm_add_pd(fix0,tx);
419 fiy0 = _mm_add_pd(fiy0,ty);
420 fiz0 = _mm_add_pd(fiz0,tz);
422 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,tx,ty,tz);
424 /* Inner loop uses 27 flops */
431 j_coord_offsetA = DIM*jnrA;
433 /* load j atom coordinates */
434 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
437 /* Calculate displacement vector */
438 dx00 = _mm_sub_pd(ix0,jx0);
439 dy00 = _mm_sub_pd(iy0,jy0);
440 dz00 = _mm_sub_pd(iz0,jz0);
442 /* Calculate squared distance and things based on it */
443 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
445 rinv00 = gmx_mm_invsqrt_pd(rsq00);
447 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
449 /* Load parameters for j particles */
450 jq0 = _mm_load_sd(charge+jnrA+0);
452 /**************************
453 * CALCULATE INTERACTIONS *
454 **************************/
456 /* Compute parameters for interactions between i and j atoms */
457 qq00 = _mm_mul_pd(iq0,jq0);
459 /* COULOMB ELECTROSTATICS */
460 velec = _mm_mul_pd(qq00,rinv00);
461 felec = _mm_mul_pd(velec,rinvsq00);
465 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
467 /* Calculate temporary vectorial force */
468 tx = _mm_mul_pd(fscal,dx00);
469 ty = _mm_mul_pd(fscal,dy00);
470 tz = _mm_mul_pd(fscal,dz00);
472 /* Update vectorial force */
473 fix0 = _mm_add_pd(fix0,tx);
474 fiy0 = _mm_add_pd(fiy0,ty);
475 fiz0 = _mm_add_pd(fiz0,tz);
477 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,tx,ty,tz);
479 /* Inner loop uses 27 flops */
482 /* End of innermost loop */
484 gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
485 f+i_coord_offset,fshift+i_shift_offset);
487 /* Increment number of inner iterations */
488 inneriter += j_index_end - j_index_start;
490 /* Outer loop uses 7 flops */
493 /* Increment number of outer iterations */
496 /* Update outer/inner flops */
498 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_F,outeriter*7 + inneriter*27);