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_ElecRF_VdwNone_GeomP1P1_VF_avx_128_fma_double
52 * Electrostatics interaction: ReactionField
53 * VdW interaction: None
54 * Geometry: Particle-Particle
55 * Calculate force/pot: PotentialAndForce
58 nb_kernel_ElecRF_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;
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 /* Avoid stupid compiler warnings */
116 /* Start outer loop over neighborlists */
117 for(iidx=0; iidx<nri; iidx++)
119 /* Load shift vector for this list */
120 i_shift_offset = DIM*shiftidx[iidx];
122 /* Load limits for loop over neighbors */
123 j_index_start = jindex[iidx];
124 j_index_end = jindex[iidx+1];
126 /* Get outer coordinate index */
128 i_coord_offset = DIM*inr;
130 /* Load i particle coords and add shift vector */
131 gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
133 fix0 = _mm_setzero_pd();
134 fiy0 = _mm_setzero_pd();
135 fiz0 = _mm_setzero_pd();
137 /* Load parameters for i particles */
138 iq0 = _mm_mul_pd(facel,_mm_load1_pd(charge+inr+0));
140 /* Reset potential sums */
141 velecsum = _mm_setzero_pd();
143 /* Start inner kernel loop */
144 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
147 /* Get j neighbor index, and coordinate index */
150 j_coord_offsetA = DIM*jnrA;
151 j_coord_offsetB = DIM*jnrB;
153 /* load j atom coordinates */
154 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
157 /* Calculate displacement vector */
158 dx00 = _mm_sub_pd(ix0,jx0);
159 dy00 = _mm_sub_pd(iy0,jy0);
160 dz00 = _mm_sub_pd(iz0,jz0);
162 /* Calculate squared distance and things based on it */
163 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
165 rinv00 = gmx_mm_invsqrt_pd(rsq00);
167 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
169 /* Load parameters for j particles */
170 jq0 = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
172 /**************************
173 * CALCULATE INTERACTIONS *
174 **************************/
176 /* Compute parameters for interactions between i and j atoms */
177 qq00 = _mm_mul_pd(iq0,jq0);
179 /* REACTION-FIELD ELECTROSTATICS */
180 velec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_macc_pd(krf,rsq00,rinv00),crf));
181 felec = _mm_mul_pd(qq00,_mm_msub_pd(rinv00,rinvsq00,krf2));
183 /* Update potential sum for this i atom from the interaction with this j atom. */
184 velecsum = _mm_add_pd(velecsum,velec);
188 /* Update vectorial force */
189 fix0 = _mm_macc_pd(dx00,fscal,fix0);
190 fiy0 = _mm_macc_pd(dy00,fscal,fiy0);
191 fiz0 = _mm_macc_pd(dz00,fscal,fiz0);
193 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,
194 _mm_mul_pd(dx00,fscal),
195 _mm_mul_pd(dy00,fscal),
196 _mm_mul_pd(dz00,fscal));
198 /* Inner loop uses 35 flops */
205 j_coord_offsetA = DIM*jnrA;
207 /* load j atom coordinates */
208 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
211 /* Calculate displacement vector */
212 dx00 = _mm_sub_pd(ix0,jx0);
213 dy00 = _mm_sub_pd(iy0,jy0);
214 dz00 = _mm_sub_pd(iz0,jz0);
216 /* Calculate squared distance and things based on it */
217 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
219 rinv00 = gmx_mm_invsqrt_pd(rsq00);
221 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
223 /* Load parameters for j particles */
224 jq0 = _mm_load_sd(charge+jnrA+0);
226 /**************************
227 * CALCULATE INTERACTIONS *
228 **************************/
230 /* Compute parameters for interactions between i and j atoms */
231 qq00 = _mm_mul_pd(iq0,jq0);
233 /* REACTION-FIELD ELECTROSTATICS */
234 velec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_macc_pd(krf,rsq00,rinv00),crf));
235 felec = _mm_mul_pd(qq00,_mm_msub_pd(rinv00,rinvsq00,krf2));
237 /* Update potential sum for this i atom from the interaction with this j atom. */
238 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
239 velecsum = _mm_add_pd(velecsum,velec);
243 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
245 /* Update vectorial force */
246 fix0 = _mm_macc_pd(dx00,fscal,fix0);
247 fiy0 = _mm_macc_pd(dy00,fscal,fiy0);
248 fiz0 = _mm_macc_pd(dz00,fscal,fiz0);
250 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,
251 _mm_mul_pd(dx00,fscal),
252 _mm_mul_pd(dy00,fscal),
253 _mm_mul_pd(dz00,fscal));
255 /* Inner loop uses 35 flops */
258 /* End of innermost loop */
260 gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
261 f+i_coord_offset,fshift+i_shift_offset);
264 /* Update potential energies */
265 gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
267 /* Increment number of inner iterations */
268 inneriter += j_index_end - j_index_start;
270 /* Outer loop uses 8 flops */
273 /* Increment number of outer iterations */
276 /* Update outer/inner flops */
278 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VF,outeriter*8 + inneriter*35);
281 * Gromacs nonbonded kernel: nb_kernel_ElecRF_VdwNone_GeomP1P1_F_avx_128_fma_double
282 * Electrostatics interaction: ReactionField
283 * VdW interaction: None
284 * Geometry: Particle-Particle
285 * Calculate force/pot: Force
288 nb_kernel_ElecRF_VdwNone_GeomP1P1_F_avx_128_fma_double
289 (t_nblist * gmx_restrict nlist,
290 rvec * gmx_restrict xx,
291 rvec * gmx_restrict ff,
292 t_forcerec * gmx_restrict fr,
293 t_mdatoms * gmx_restrict mdatoms,
294 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
295 t_nrnb * gmx_restrict nrnb)
297 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
298 * just 0 for non-waters.
299 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
300 * jnr indices corresponding to data put in the four positions in the SIMD register.
302 int i_shift_offset,i_coord_offset,outeriter,inneriter;
303 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
305 int j_coord_offsetA,j_coord_offsetB;
306 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
308 real *shiftvec,*fshift,*x,*f;
309 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
311 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
312 int vdwjidx0A,vdwjidx0B;
313 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
314 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
315 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
317 __m128d dummy_mask,cutoff_mask;
318 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
319 __m128d one = _mm_set1_pd(1.0);
320 __m128d two = _mm_set1_pd(2.0);
326 jindex = nlist->jindex;
328 shiftidx = nlist->shift;
330 shiftvec = fr->shift_vec[0];
331 fshift = fr->fshift[0];
332 facel = _mm_set1_pd(fr->epsfac);
333 charge = mdatoms->chargeA;
334 krf = _mm_set1_pd(fr->ic->k_rf);
335 krf2 = _mm_set1_pd(fr->ic->k_rf*2.0);
336 crf = _mm_set1_pd(fr->ic->c_rf);
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 /* REACTION-FIELD ELECTROSTATICS */
407 felec = _mm_mul_pd(qq00,_mm_msub_pd(rinv00,rinvsq00,krf2));
411 /* Update vectorial force */
412 fix0 = _mm_macc_pd(dx00,fscal,fix0);
413 fiy0 = _mm_macc_pd(dy00,fscal,fiy0);
414 fiz0 = _mm_macc_pd(dz00,fscal,fiz0);
416 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,
417 _mm_mul_pd(dx00,fscal),
418 _mm_mul_pd(dy00,fscal),
419 _mm_mul_pd(dz00,fscal));
421 /* Inner loop uses 30 flops */
428 j_coord_offsetA = DIM*jnrA;
430 /* load j atom coordinates */
431 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
434 /* Calculate displacement vector */
435 dx00 = _mm_sub_pd(ix0,jx0);
436 dy00 = _mm_sub_pd(iy0,jy0);
437 dz00 = _mm_sub_pd(iz0,jz0);
439 /* Calculate squared distance and things based on it */
440 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
442 rinv00 = gmx_mm_invsqrt_pd(rsq00);
444 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
446 /* Load parameters for j particles */
447 jq0 = _mm_load_sd(charge+jnrA+0);
449 /**************************
450 * CALCULATE INTERACTIONS *
451 **************************/
453 /* Compute parameters for interactions between i and j atoms */
454 qq00 = _mm_mul_pd(iq0,jq0);
456 /* REACTION-FIELD ELECTROSTATICS */
457 felec = _mm_mul_pd(qq00,_mm_msub_pd(rinv00,rinvsq00,krf2));
461 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
463 /* Update vectorial force */
464 fix0 = _mm_macc_pd(dx00,fscal,fix0);
465 fiy0 = _mm_macc_pd(dy00,fscal,fiy0);
466 fiz0 = _mm_macc_pd(dz00,fscal,fiz0);
468 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,
469 _mm_mul_pd(dx00,fscal),
470 _mm_mul_pd(dy00,fscal),
471 _mm_mul_pd(dz00,fscal));
473 /* Inner loop uses 30 flops */
476 /* End of innermost loop */
478 gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
479 f+i_coord_offset,fshift+i_shift_offset);
481 /* Increment number of inner iterations */
482 inneriter += j_index_end - j_index_start;
484 /* Outer loop uses 7 flops */
487 /* Increment number of outer iterations */
490 /* Update outer/inner flops */
492 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_F,outeriter*7 + inneriter*30);