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 avx_128_fma_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_avx_128_fma_double.h"
48 #include "kernelutil_x86_avx_128_fma_double.h"
51 * Gromacs nonbonded kernel: nb_kernel_ElecCoul_VdwNone_GeomP1P1_VF_avx_128_fma_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_avx_128_fma_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 /* Update vectorial force */
186 fix0 = _mm_macc_pd(dx00,fscal,fix0);
187 fiy0 = _mm_macc_pd(dy00,fscal,fiy0);
188 fiz0 = _mm_macc_pd(dz00,fscal,fiz0);
190 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,
191 _mm_mul_pd(dx00,fscal),
192 _mm_mul_pd(dy00,fscal),
193 _mm_mul_pd(dz00,fscal));
195 /* Inner loop uses 31 flops */
202 j_coord_offsetA = DIM*jnrA;
204 /* load j atom coordinates */
205 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
208 /* Calculate displacement vector */
209 dx00 = _mm_sub_pd(ix0,jx0);
210 dy00 = _mm_sub_pd(iy0,jy0);
211 dz00 = _mm_sub_pd(iz0,jz0);
213 /* Calculate squared distance and things based on it */
214 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
216 rinv00 = gmx_mm_invsqrt_pd(rsq00);
218 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
220 /* Load parameters for j particles */
221 jq0 = _mm_load_sd(charge+jnrA+0);
223 /**************************
224 * CALCULATE INTERACTIONS *
225 **************************/
227 /* Compute parameters for interactions between i and j atoms */
228 qq00 = _mm_mul_pd(iq0,jq0);
230 /* COULOMB ELECTROSTATICS */
231 velec = _mm_mul_pd(qq00,rinv00);
232 felec = _mm_mul_pd(velec,rinvsq00);
234 /* Update potential sum for this i atom from the interaction with this j atom. */
235 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
236 velecsum = _mm_add_pd(velecsum,velec);
240 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
242 /* Update vectorial force */
243 fix0 = _mm_macc_pd(dx00,fscal,fix0);
244 fiy0 = _mm_macc_pd(dy00,fscal,fiy0);
245 fiz0 = _mm_macc_pd(dz00,fscal,fiz0);
247 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,
248 _mm_mul_pd(dx00,fscal),
249 _mm_mul_pd(dy00,fscal),
250 _mm_mul_pd(dz00,fscal));
252 /* Inner loop uses 31 flops */
255 /* End of innermost loop */
257 gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
258 f+i_coord_offset,fshift+i_shift_offset);
261 /* Update potential energies */
262 gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
264 /* Increment number of inner iterations */
265 inneriter += j_index_end - j_index_start;
267 /* Outer loop uses 8 flops */
270 /* Increment number of outer iterations */
273 /* Update outer/inner flops */
275 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VF,outeriter*8 + inneriter*31);
278 * Gromacs nonbonded kernel: nb_kernel_ElecCoul_VdwNone_GeomP1P1_F_avx_128_fma_double
279 * Electrostatics interaction: Coulomb
280 * VdW interaction: None
281 * Geometry: Particle-Particle
282 * Calculate force/pot: Force
285 nb_kernel_ElecCoul_VdwNone_GeomP1P1_F_avx_128_fma_double
286 (t_nblist * gmx_restrict nlist,
287 rvec * gmx_restrict xx,
288 rvec * gmx_restrict ff,
289 t_forcerec * gmx_restrict fr,
290 t_mdatoms * gmx_restrict mdatoms,
291 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
292 t_nrnb * gmx_restrict nrnb)
294 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
295 * just 0 for non-waters.
296 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
297 * jnr indices corresponding to data put in the four positions in the SIMD register.
299 int i_shift_offset,i_coord_offset,outeriter,inneriter;
300 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
302 int j_coord_offsetA,j_coord_offsetB;
303 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
305 real *shiftvec,*fshift,*x,*f;
306 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
308 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
309 int vdwjidx0A,vdwjidx0B;
310 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
311 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
312 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
314 __m128d dummy_mask,cutoff_mask;
315 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
316 __m128d one = _mm_set1_pd(1.0);
317 __m128d two = _mm_set1_pd(2.0);
323 jindex = nlist->jindex;
325 shiftidx = nlist->shift;
327 shiftvec = fr->shift_vec[0];
328 fshift = fr->fshift[0];
329 facel = _mm_set1_pd(fr->epsfac);
330 charge = mdatoms->chargeA;
332 /* Avoid stupid compiler warnings */
340 /* Start outer loop over neighborlists */
341 for(iidx=0; iidx<nri; iidx++)
343 /* Load shift vector for this list */
344 i_shift_offset = DIM*shiftidx[iidx];
346 /* Load limits for loop over neighbors */
347 j_index_start = jindex[iidx];
348 j_index_end = jindex[iidx+1];
350 /* Get outer coordinate index */
352 i_coord_offset = DIM*inr;
354 /* Load i particle coords and add shift vector */
355 gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
357 fix0 = _mm_setzero_pd();
358 fiy0 = _mm_setzero_pd();
359 fiz0 = _mm_setzero_pd();
361 /* Load parameters for i particles */
362 iq0 = _mm_mul_pd(facel,_mm_load1_pd(charge+inr+0));
364 /* Start inner kernel loop */
365 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
368 /* Get j neighbor index, and coordinate index */
371 j_coord_offsetA = DIM*jnrA;
372 j_coord_offsetB = DIM*jnrB;
374 /* load j atom coordinates */
375 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
378 /* Calculate displacement vector */
379 dx00 = _mm_sub_pd(ix0,jx0);
380 dy00 = _mm_sub_pd(iy0,jy0);
381 dz00 = _mm_sub_pd(iz0,jz0);
383 /* Calculate squared distance and things based on it */
384 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
386 rinv00 = gmx_mm_invsqrt_pd(rsq00);
388 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
390 /* Load parameters for j particles */
391 jq0 = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
393 /**************************
394 * CALCULATE INTERACTIONS *
395 **************************/
397 /* Compute parameters for interactions between i and j atoms */
398 qq00 = _mm_mul_pd(iq0,jq0);
400 /* COULOMB ELECTROSTATICS */
401 velec = _mm_mul_pd(qq00,rinv00);
402 felec = _mm_mul_pd(velec,rinvsq00);
406 /* Update vectorial force */
407 fix0 = _mm_macc_pd(dx00,fscal,fix0);
408 fiy0 = _mm_macc_pd(dy00,fscal,fiy0);
409 fiz0 = _mm_macc_pd(dz00,fscal,fiz0);
411 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,
412 _mm_mul_pd(dx00,fscal),
413 _mm_mul_pd(dy00,fscal),
414 _mm_mul_pd(dz00,fscal));
416 /* Inner loop uses 30 flops */
423 j_coord_offsetA = DIM*jnrA;
425 /* load j atom coordinates */
426 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
429 /* Calculate displacement vector */
430 dx00 = _mm_sub_pd(ix0,jx0);
431 dy00 = _mm_sub_pd(iy0,jy0);
432 dz00 = _mm_sub_pd(iz0,jz0);
434 /* Calculate squared distance and things based on it */
435 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
437 rinv00 = gmx_mm_invsqrt_pd(rsq00);
439 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
441 /* Load parameters for j particles */
442 jq0 = _mm_load_sd(charge+jnrA+0);
444 /**************************
445 * CALCULATE INTERACTIONS *
446 **************************/
448 /* Compute parameters for interactions between i and j atoms */
449 qq00 = _mm_mul_pd(iq0,jq0);
451 /* COULOMB ELECTROSTATICS */
452 velec = _mm_mul_pd(qq00,rinv00);
453 felec = _mm_mul_pd(velec,rinvsq00);
457 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
459 /* Update vectorial force */
460 fix0 = _mm_macc_pd(dx00,fscal,fix0);
461 fiy0 = _mm_macc_pd(dy00,fscal,fiy0);
462 fiz0 = _mm_macc_pd(dz00,fscal,fiz0);
464 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,
465 _mm_mul_pd(dx00,fscal),
466 _mm_mul_pd(dy00,fscal),
467 _mm_mul_pd(dz00,fscal));
469 /* Inner loop uses 30 flops */
472 /* End of innermost loop */
474 gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
475 f+i_coord_offset,fshift+i_shift_offset);
477 /* Increment number of inner iterations */
478 inneriter += j_index_end - j_index_start;
480 /* Outer loop uses 7 flops */
483 /* Increment number of outer iterations */
486 /* Update outer/inner flops */
488 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_F,outeriter*7 + inneriter*30);