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 "gromacs/legacyheaders/types/simple.h"
44 #include "gromacs/math/vec.h"
45 #include "gromacs/legacyheaders/nrnb.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_ElecNone_VdwLJ_GeomP1P1_VF_sse4_1_double
52 * Electrostatics interaction: None
53 * VdW interaction: LennardJones
54 * Geometry: Particle-Particle
55 * Calculate force/pot: PotentialAndForce
58 nb_kernel_ElecNone_VdwLJ_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;
86 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
89 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
90 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
91 __m128d dummy_mask,cutoff_mask;
92 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
93 __m128d one = _mm_set1_pd(1.0);
94 __m128d two = _mm_set1_pd(2.0);
100 jindex = nlist->jindex;
102 shiftidx = nlist->shift;
104 shiftvec = fr->shift_vec[0];
105 fshift = fr->fshift[0];
106 nvdwtype = fr->ntype;
108 vdwtype = mdatoms->typeA;
110 /* Avoid stupid compiler warnings */
118 /* Start outer loop over neighborlists */
119 for(iidx=0; iidx<nri; iidx++)
121 /* Load shift vector for this list */
122 i_shift_offset = DIM*shiftidx[iidx];
124 /* Load limits for loop over neighbors */
125 j_index_start = jindex[iidx];
126 j_index_end = jindex[iidx+1];
128 /* Get outer coordinate index */
130 i_coord_offset = DIM*inr;
132 /* Load i particle coords and add shift vector */
133 gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
135 fix0 = _mm_setzero_pd();
136 fiy0 = _mm_setzero_pd();
137 fiz0 = _mm_setzero_pd();
139 /* Load parameters for i particles */
140 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
142 /* Reset potential sums */
143 vvdwsum = _mm_setzero_pd();
145 /* Start inner kernel loop */
146 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
149 /* Get j neighbor index, and coordinate index */
152 j_coord_offsetA = DIM*jnrA;
153 j_coord_offsetB = DIM*jnrB;
155 /* load j atom coordinates */
156 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
159 /* Calculate displacement vector */
160 dx00 = _mm_sub_pd(ix0,jx0);
161 dy00 = _mm_sub_pd(iy0,jy0);
162 dz00 = _mm_sub_pd(iz0,jz0);
164 /* Calculate squared distance and things based on it */
165 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
167 rinvsq00 = gmx_mm_inv_pd(rsq00);
169 /* Load parameters for j particles */
170 vdwjidx0A = 2*vdwtype[jnrA+0];
171 vdwjidx0B = 2*vdwtype[jnrB+0];
173 /**************************
174 * CALCULATE INTERACTIONS *
175 **************************/
177 /* Compute parameters for interactions between i and j atoms */
178 gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
179 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
181 /* LENNARD-JONES DISPERSION/REPULSION */
183 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
184 vvdw6 = _mm_mul_pd(c6_00,rinvsix);
185 vvdw12 = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
186 vvdw = _mm_sub_pd( _mm_mul_pd(vvdw12,one_twelfth) , _mm_mul_pd(vvdw6,one_sixth) );
187 fvdw = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
189 /* Update potential sum for this i atom from the interaction with this j atom. */
190 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
194 /* Calculate temporary vectorial force */
195 tx = _mm_mul_pd(fscal,dx00);
196 ty = _mm_mul_pd(fscal,dy00);
197 tz = _mm_mul_pd(fscal,dz00);
199 /* Update vectorial force */
200 fix0 = _mm_add_pd(fix0,tx);
201 fiy0 = _mm_add_pd(fiy0,ty);
202 fiz0 = _mm_add_pd(fiz0,tz);
204 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,tx,ty,tz);
206 /* Inner loop uses 32 flops */
213 j_coord_offsetA = DIM*jnrA;
215 /* load j atom coordinates */
216 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
219 /* Calculate displacement vector */
220 dx00 = _mm_sub_pd(ix0,jx0);
221 dy00 = _mm_sub_pd(iy0,jy0);
222 dz00 = _mm_sub_pd(iz0,jz0);
224 /* Calculate squared distance and things based on it */
225 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
227 rinvsq00 = gmx_mm_inv_pd(rsq00);
229 /* Load parameters for j particles */
230 vdwjidx0A = 2*vdwtype[jnrA+0];
232 /**************************
233 * CALCULATE INTERACTIONS *
234 **************************/
236 /* Compute parameters for interactions between i and j atoms */
237 gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
239 /* LENNARD-JONES DISPERSION/REPULSION */
241 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
242 vvdw6 = _mm_mul_pd(c6_00,rinvsix);
243 vvdw12 = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
244 vvdw = _mm_sub_pd( _mm_mul_pd(vvdw12,one_twelfth) , _mm_mul_pd(vvdw6,one_sixth) );
245 fvdw = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
247 /* Update potential sum for this i atom from the interaction with this j atom. */
248 vvdw = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
249 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
253 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
255 /* Calculate temporary vectorial force */
256 tx = _mm_mul_pd(fscal,dx00);
257 ty = _mm_mul_pd(fscal,dy00);
258 tz = _mm_mul_pd(fscal,dz00);
260 /* Update vectorial force */
261 fix0 = _mm_add_pd(fix0,tx);
262 fiy0 = _mm_add_pd(fiy0,ty);
263 fiz0 = _mm_add_pd(fiz0,tz);
265 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,tx,ty,tz);
267 /* Inner loop uses 32 flops */
270 /* End of innermost loop */
272 gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
273 f+i_coord_offset,fshift+i_shift_offset);
276 /* Update potential energies */
277 gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
279 /* Increment number of inner iterations */
280 inneriter += j_index_end - j_index_start;
282 /* Outer loop uses 7 flops */
285 /* Increment number of outer iterations */
288 /* Update outer/inner flops */
290 inc_nrnb(nrnb,eNR_NBKERNEL_VDW_VF,outeriter*7 + inneriter*32);
293 * Gromacs nonbonded kernel: nb_kernel_ElecNone_VdwLJ_GeomP1P1_F_sse4_1_double
294 * Electrostatics interaction: None
295 * VdW interaction: LennardJones
296 * Geometry: Particle-Particle
297 * Calculate force/pot: Force
300 nb_kernel_ElecNone_VdwLJ_GeomP1P1_F_sse4_1_double
301 (t_nblist * gmx_restrict nlist,
302 rvec * gmx_restrict xx,
303 rvec * gmx_restrict ff,
304 t_forcerec * gmx_restrict fr,
305 t_mdatoms * gmx_restrict mdatoms,
306 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
307 t_nrnb * gmx_restrict nrnb)
309 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
310 * just 0 for non-waters.
311 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
312 * jnr indices corresponding to data put in the four positions in the SIMD register.
314 int i_shift_offset,i_coord_offset,outeriter,inneriter;
315 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
317 int j_coord_offsetA,j_coord_offsetB;
318 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
320 real *shiftvec,*fshift,*x,*f;
321 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
323 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
324 int vdwjidx0A,vdwjidx0B;
325 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
326 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
328 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
331 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
332 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
333 __m128d dummy_mask,cutoff_mask;
334 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
335 __m128d one = _mm_set1_pd(1.0);
336 __m128d two = _mm_set1_pd(2.0);
342 jindex = nlist->jindex;
344 shiftidx = nlist->shift;
346 shiftvec = fr->shift_vec[0];
347 fshift = fr->fshift[0];
348 nvdwtype = fr->ntype;
350 vdwtype = mdatoms->typeA;
352 /* Avoid stupid compiler warnings */
360 /* Start outer loop over neighborlists */
361 for(iidx=0; iidx<nri; iidx++)
363 /* Load shift vector for this list */
364 i_shift_offset = DIM*shiftidx[iidx];
366 /* Load limits for loop over neighbors */
367 j_index_start = jindex[iidx];
368 j_index_end = jindex[iidx+1];
370 /* Get outer coordinate index */
372 i_coord_offset = DIM*inr;
374 /* Load i particle coords and add shift vector */
375 gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
377 fix0 = _mm_setzero_pd();
378 fiy0 = _mm_setzero_pd();
379 fiz0 = _mm_setzero_pd();
381 /* Load parameters for i particles */
382 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
384 /* Start inner kernel loop */
385 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
388 /* Get j neighbor index, and coordinate index */
391 j_coord_offsetA = DIM*jnrA;
392 j_coord_offsetB = DIM*jnrB;
394 /* load j atom coordinates */
395 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
398 /* Calculate displacement vector */
399 dx00 = _mm_sub_pd(ix0,jx0);
400 dy00 = _mm_sub_pd(iy0,jy0);
401 dz00 = _mm_sub_pd(iz0,jz0);
403 /* Calculate squared distance and things based on it */
404 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
406 rinvsq00 = gmx_mm_inv_pd(rsq00);
408 /* Load parameters for j particles */
409 vdwjidx0A = 2*vdwtype[jnrA+0];
410 vdwjidx0B = 2*vdwtype[jnrB+0];
412 /**************************
413 * CALCULATE INTERACTIONS *
414 **************************/
416 /* Compute parameters for interactions between i and j atoms */
417 gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
418 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
420 /* LENNARD-JONES DISPERSION/REPULSION */
422 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
423 fvdw = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
427 /* Calculate temporary vectorial force */
428 tx = _mm_mul_pd(fscal,dx00);
429 ty = _mm_mul_pd(fscal,dy00);
430 tz = _mm_mul_pd(fscal,dz00);
432 /* Update vectorial force */
433 fix0 = _mm_add_pd(fix0,tx);
434 fiy0 = _mm_add_pd(fiy0,ty);
435 fiz0 = _mm_add_pd(fiz0,tz);
437 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,tx,ty,tz);
439 /* Inner loop uses 27 flops */
446 j_coord_offsetA = DIM*jnrA;
448 /* load j atom coordinates */
449 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
452 /* Calculate displacement vector */
453 dx00 = _mm_sub_pd(ix0,jx0);
454 dy00 = _mm_sub_pd(iy0,jy0);
455 dz00 = _mm_sub_pd(iz0,jz0);
457 /* Calculate squared distance and things based on it */
458 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
460 rinvsq00 = gmx_mm_inv_pd(rsq00);
462 /* Load parameters for j particles */
463 vdwjidx0A = 2*vdwtype[jnrA+0];
465 /**************************
466 * CALCULATE INTERACTIONS *
467 **************************/
469 /* Compute parameters for interactions between i and j atoms */
470 gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
472 /* LENNARD-JONES DISPERSION/REPULSION */
474 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
475 fvdw = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
479 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
481 /* Calculate temporary vectorial force */
482 tx = _mm_mul_pd(fscal,dx00);
483 ty = _mm_mul_pd(fscal,dy00);
484 tz = _mm_mul_pd(fscal,dz00);
486 /* Update vectorial force */
487 fix0 = _mm_add_pd(fix0,tx);
488 fiy0 = _mm_add_pd(fiy0,ty);
489 fiz0 = _mm_add_pd(fiz0,tz);
491 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,tx,ty,tz);
493 /* Inner loop uses 27 flops */
496 /* End of innermost loop */
498 gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
499 f+i_coord_offset,fshift+i_shift_offset);
501 /* Increment number of inner iterations */
502 inneriter += j_index_end - j_index_start;
504 /* Outer loop uses 6 flops */
507 /* Increment number of outer iterations */
510 /* Update outer/inner flops */
512 inc_nrnb(nrnb,eNR_NBKERNEL_VDW_F,outeriter*6 + inneriter*27);