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.
42 #include "../nb_kernel.h"
43 #include "types/simple.h"
44 #include "gromacs/math/vec.h"
47 #include "gromacs/simd/math_x86_sse4_1_double.h"
48 #include "kernelutil_x86_sse4_1_double.h"
51 * Gromacs nonbonded kernel: nb_kernel_ElecCoul_VdwNone_GeomP1P1_VF_sse4_1_double
52 * Electrostatics interaction: Coulomb
53 * VdW interaction: None
54 * Geometry: Particle-Particle
55 * Calculate force/pot: PotentialAndForce
58 nb_kernel_ElecCoul_VdwNone_GeomP1P1_VF_sse4_1_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;
105 /* Avoid stupid compiler warnings */
113 /* Start outer loop over neighborlists */
114 for(iidx=0; iidx<nri; iidx++)
116 /* Load shift vector for this list */
117 i_shift_offset = DIM*shiftidx[iidx];
119 /* Load limits for loop over neighbors */
120 j_index_start = jindex[iidx];
121 j_index_end = jindex[iidx+1];
123 /* Get outer coordinate index */
125 i_coord_offset = DIM*inr;
127 /* Load i particle coords and add shift vector */
128 gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
130 fix0 = _mm_setzero_pd();
131 fiy0 = _mm_setzero_pd();
132 fiz0 = _mm_setzero_pd();
134 /* Load parameters for i particles */
135 iq0 = _mm_mul_pd(facel,_mm_load1_pd(charge+inr+0));
137 /* Reset potential sums */
138 velecsum = _mm_setzero_pd();
140 /* Start inner kernel loop */
141 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
144 /* Get j neighbor index, and coordinate index */
147 j_coord_offsetA = DIM*jnrA;
148 j_coord_offsetB = DIM*jnrB;
150 /* load j atom coordinates */
151 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
154 /* Calculate displacement vector */
155 dx00 = _mm_sub_pd(ix0,jx0);
156 dy00 = _mm_sub_pd(iy0,jy0);
157 dz00 = _mm_sub_pd(iz0,jz0);
159 /* Calculate squared distance and things based on it */
160 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
162 rinv00 = gmx_mm_invsqrt_pd(rsq00);
164 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
166 /* Load parameters for j particles */
167 jq0 = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
169 /**************************
170 * CALCULATE INTERACTIONS *
171 **************************/
173 /* Compute parameters for interactions between i and j atoms */
174 qq00 = _mm_mul_pd(iq0,jq0);
176 /* COULOMB ELECTROSTATICS */
177 velec = _mm_mul_pd(qq00,rinv00);
178 felec = _mm_mul_pd(velec,rinvsq00);
180 /* Update potential sum for this i atom from the interaction with this j atom. */
181 velecsum = _mm_add_pd(velecsum,velec);
185 /* Calculate temporary vectorial force */
186 tx = _mm_mul_pd(fscal,dx00);
187 ty = _mm_mul_pd(fscal,dy00);
188 tz = _mm_mul_pd(fscal,dz00);
190 /* Update vectorial force */
191 fix0 = _mm_add_pd(fix0,tx);
192 fiy0 = _mm_add_pd(fiy0,ty);
193 fiz0 = _mm_add_pd(fiz0,tz);
195 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,tx,ty,tz);
197 /* Inner loop uses 28 flops */
204 j_coord_offsetA = DIM*jnrA;
206 /* load j atom coordinates */
207 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
210 /* Calculate displacement vector */
211 dx00 = _mm_sub_pd(ix0,jx0);
212 dy00 = _mm_sub_pd(iy0,jy0);
213 dz00 = _mm_sub_pd(iz0,jz0);
215 /* Calculate squared distance and things based on it */
216 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
218 rinv00 = gmx_mm_invsqrt_pd(rsq00);
220 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
222 /* Load parameters for j particles */
223 jq0 = _mm_load_sd(charge+jnrA+0);
225 /**************************
226 * CALCULATE INTERACTIONS *
227 **************************/
229 /* Compute parameters for interactions between i and j atoms */
230 qq00 = _mm_mul_pd(iq0,jq0);
232 /* COULOMB ELECTROSTATICS */
233 velec = _mm_mul_pd(qq00,rinv00);
234 felec = _mm_mul_pd(velec,rinvsq00);
236 /* Update potential sum for this i atom from the interaction with this j atom. */
237 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
238 velecsum = _mm_add_pd(velecsum,velec);
242 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
244 /* Calculate temporary vectorial force */
245 tx = _mm_mul_pd(fscal,dx00);
246 ty = _mm_mul_pd(fscal,dy00);
247 tz = _mm_mul_pd(fscal,dz00);
249 /* Update vectorial force */
250 fix0 = _mm_add_pd(fix0,tx);
251 fiy0 = _mm_add_pd(fiy0,ty);
252 fiz0 = _mm_add_pd(fiz0,tz);
254 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,tx,ty,tz);
256 /* Inner loop uses 28 flops */
259 /* End of innermost loop */
261 gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
262 f+i_coord_offset,fshift+i_shift_offset);
265 /* Update potential energies */
266 gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
268 /* Increment number of inner iterations */
269 inneriter += j_index_end - j_index_start;
271 /* Outer loop uses 8 flops */
274 /* Increment number of outer iterations */
277 /* Update outer/inner flops */
279 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VF,outeriter*8 + inneriter*28);
282 * Gromacs nonbonded kernel: nb_kernel_ElecCoul_VdwNone_GeomP1P1_F_sse4_1_double
283 * Electrostatics interaction: Coulomb
284 * VdW interaction: None
285 * Geometry: Particle-Particle
286 * Calculate force/pot: Force
289 nb_kernel_ElecCoul_VdwNone_GeomP1P1_F_sse4_1_double
290 (t_nblist * gmx_restrict nlist,
291 rvec * gmx_restrict xx,
292 rvec * gmx_restrict ff,
293 t_forcerec * gmx_restrict fr,
294 t_mdatoms * gmx_restrict mdatoms,
295 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
296 t_nrnb * gmx_restrict nrnb)
298 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
299 * just 0 for non-waters.
300 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
301 * jnr indices corresponding to data put in the four positions in the SIMD register.
303 int i_shift_offset,i_coord_offset,outeriter,inneriter;
304 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
306 int j_coord_offsetA,j_coord_offsetB;
307 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
309 real *shiftvec,*fshift,*x,*f;
310 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
312 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
313 int vdwjidx0A,vdwjidx0B;
314 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
315 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
316 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
318 __m128d dummy_mask,cutoff_mask;
319 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
320 __m128d one = _mm_set1_pd(1.0);
321 __m128d two = _mm_set1_pd(2.0);
327 jindex = nlist->jindex;
329 shiftidx = nlist->shift;
331 shiftvec = fr->shift_vec[0];
332 fshift = fr->fshift[0];
333 facel = _mm_set1_pd(fr->epsfac);
334 charge = mdatoms->chargeA;
336 /* Avoid stupid compiler warnings */
344 /* Start outer loop over neighborlists */
345 for(iidx=0; iidx<nri; iidx++)
347 /* Load shift vector for this list */
348 i_shift_offset = DIM*shiftidx[iidx];
350 /* Load limits for loop over neighbors */
351 j_index_start = jindex[iidx];
352 j_index_end = jindex[iidx+1];
354 /* Get outer coordinate index */
356 i_coord_offset = DIM*inr;
358 /* Load i particle coords and add shift vector */
359 gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
361 fix0 = _mm_setzero_pd();
362 fiy0 = _mm_setzero_pd();
363 fiz0 = _mm_setzero_pd();
365 /* Load parameters for i particles */
366 iq0 = _mm_mul_pd(facel,_mm_load1_pd(charge+inr+0));
368 /* Start inner kernel loop */
369 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
372 /* Get j neighbor index, and coordinate index */
375 j_coord_offsetA = DIM*jnrA;
376 j_coord_offsetB = DIM*jnrB;
378 /* load j atom coordinates */
379 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
382 /* Calculate displacement vector */
383 dx00 = _mm_sub_pd(ix0,jx0);
384 dy00 = _mm_sub_pd(iy0,jy0);
385 dz00 = _mm_sub_pd(iz0,jz0);
387 /* Calculate squared distance and things based on it */
388 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
390 rinv00 = gmx_mm_invsqrt_pd(rsq00);
392 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
394 /* Load parameters for j particles */
395 jq0 = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
397 /**************************
398 * CALCULATE INTERACTIONS *
399 **************************/
401 /* Compute parameters for interactions between i and j atoms */
402 qq00 = _mm_mul_pd(iq0,jq0);
404 /* COULOMB ELECTROSTATICS */
405 velec = _mm_mul_pd(qq00,rinv00);
406 felec = _mm_mul_pd(velec,rinvsq00);
410 /* Calculate temporary vectorial force */
411 tx = _mm_mul_pd(fscal,dx00);
412 ty = _mm_mul_pd(fscal,dy00);
413 tz = _mm_mul_pd(fscal,dz00);
415 /* Update vectorial force */
416 fix0 = _mm_add_pd(fix0,tx);
417 fiy0 = _mm_add_pd(fiy0,ty);
418 fiz0 = _mm_add_pd(fiz0,tz);
420 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,tx,ty,tz);
422 /* Inner loop uses 27 flops */
429 j_coord_offsetA = DIM*jnrA;
431 /* load j atom coordinates */
432 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
435 /* Calculate displacement vector */
436 dx00 = _mm_sub_pd(ix0,jx0);
437 dy00 = _mm_sub_pd(iy0,jy0);
438 dz00 = _mm_sub_pd(iz0,jz0);
440 /* Calculate squared distance and things based on it */
441 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
443 rinv00 = gmx_mm_invsqrt_pd(rsq00);
445 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
447 /* Load parameters for j particles */
448 jq0 = _mm_load_sd(charge+jnrA+0);
450 /**************************
451 * CALCULATE INTERACTIONS *
452 **************************/
454 /* Compute parameters for interactions between i and j atoms */
455 qq00 = _mm_mul_pd(iq0,jq0);
457 /* COULOMB ELECTROSTATICS */
458 velec = _mm_mul_pd(qq00,rinv00);
459 felec = _mm_mul_pd(velec,rinvsq00);
463 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
465 /* Calculate temporary vectorial force */
466 tx = _mm_mul_pd(fscal,dx00);
467 ty = _mm_mul_pd(fscal,dy00);
468 tz = _mm_mul_pd(fscal,dz00);
470 /* Update vectorial force */
471 fix0 = _mm_add_pd(fix0,tx);
472 fiy0 = _mm_add_pd(fiy0,ty);
473 fiz0 = _mm_add_pd(fiz0,tz);
475 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,tx,ty,tz);
477 /* Inner loop uses 27 flops */
480 /* End of innermost loop */
482 gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
483 f+i_coord_offset,fshift+i_shift_offset);
485 /* Increment number of inner iterations */
486 inneriter += j_index_end - j_index_start;
488 /* Outer loop uses 7 flops */
491 /* Increment number of outer iterations */
494 /* Update outer/inner flops */
496 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_F,outeriter*7 + inneriter*27);