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 sse2_double kernel generator.
44 #include "../nb_kernel.h"
45 #include "types/simple.h"
49 #include "gromacs/simd/math_x86_sse2_double.h"
50 #include "kernelutil_x86_sse2_double.h"
53 * Gromacs nonbonded kernel: nb_kernel_ElecNone_VdwLJ_GeomP1P1_VF_sse2_double
54 * Electrostatics interaction: None
55 * VdW interaction: LennardJones
56 * Geometry: Particle-Particle
57 * Calculate force/pot: PotentialAndForce
60 nb_kernel_ElecNone_VdwLJ_GeomP1P1_VF_sse2_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;
88 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
91 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
92 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
93 __m128d dummy_mask,cutoff_mask;
94 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
95 __m128d one = _mm_set1_pd(1.0);
96 __m128d two = _mm_set1_pd(2.0);
102 jindex = nlist->jindex;
104 shiftidx = nlist->shift;
106 shiftvec = fr->shift_vec[0];
107 fshift = fr->fshift[0];
108 nvdwtype = fr->ntype;
110 vdwtype = mdatoms->typeA;
112 /* Avoid stupid compiler warnings */
120 /* Start outer loop over neighborlists */
121 for(iidx=0; iidx<nri; iidx++)
123 /* Load shift vector for this list */
124 i_shift_offset = DIM*shiftidx[iidx];
126 /* Load limits for loop over neighbors */
127 j_index_start = jindex[iidx];
128 j_index_end = jindex[iidx+1];
130 /* Get outer coordinate index */
132 i_coord_offset = DIM*inr;
134 /* Load i particle coords and add shift vector */
135 gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
137 fix0 = _mm_setzero_pd();
138 fiy0 = _mm_setzero_pd();
139 fiz0 = _mm_setzero_pd();
141 /* Load parameters for i particles */
142 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
144 /* Reset potential sums */
145 vvdwsum = _mm_setzero_pd();
147 /* Start inner kernel loop */
148 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
151 /* Get j neighbor index, and coordinate index */
154 j_coord_offsetA = DIM*jnrA;
155 j_coord_offsetB = DIM*jnrB;
157 /* load j atom coordinates */
158 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
161 /* Calculate displacement vector */
162 dx00 = _mm_sub_pd(ix0,jx0);
163 dy00 = _mm_sub_pd(iy0,jy0);
164 dz00 = _mm_sub_pd(iz0,jz0);
166 /* Calculate squared distance and things based on it */
167 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
169 rinvsq00 = gmx_mm_inv_pd(rsq00);
171 /* Load parameters for j particles */
172 vdwjidx0A = 2*vdwtype[jnrA+0];
173 vdwjidx0B = 2*vdwtype[jnrB+0];
175 /**************************
176 * CALCULATE INTERACTIONS *
177 **************************/
179 /* Compute parameters for interactions between i and j atoms */
180 gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
181 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
183 /* LENNARD-JONES DISPERSION/REPULSION */
185 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
186 vvdw6 = _mm_mul_pd(c6_00,rinvsix);
187 vvdw12 = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
188 vvdw = _mm_sub_pd( _mm_mul_pd(vvdw12,one_twelfth) , _mm_mul_pd(vvdw6,one_sixth) );
189 fvdw = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
191 /* Update potential sum for this i atom from the interaction with this j atom. */
192 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
196 /* Calculate temporary vectorial force */
197 tx = _mm_mul_pd(fscal,dx00);
198 ty = _mm_mul_pd(fscal,dy00);
199 tz = _mm_mul_pd(fscal,dz00);
201 /* Update vectorial force */
202 fix0 = _mm_add_pd(fix0,tx);
203 fiy0 = _mm_add_pd(fiy0,ty);
204 fiz0 = _mm_add_pd(fiz0,tz);
206 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,tx,ty,tz);
208 /* Inner loop uses 32 flops */
215 j_coord_offsetA = DIM*jnrA;
217 /* load j atom coordinates */
218 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
221 /* Calculate displacement vector */
222 dx00 = _mm_sub_pd(ix0,jx0);
223 dy00 = _mm_sub_pd(iy0,jy0);
224 dz00 = _mm_sub_pd(iz0,jz0);
226 /* Calculate squared distance and things based on it */
227 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
229 rinvsq00 = gmx_mm_inv_pd(rsq00);
231 /* Load parameters for j particles */
232 vdwjidx0A = 2*vdwtype[jnrA+0];
234 /**************************
235 * CALCULATE INTERACTIONS *
236 **************************/
238 /* Compute parameters for interactions between i and j atoms */
239 gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
241 /* LENNARD-JONES DISPERSION/REPULSION */
243 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
244 vvdw6 = _mm_mul_pd(c6_00,rinvsix);
245 vvdw12 = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
246 vvdw = _mm_sub_pd( _mm_mul_pd(vvdw12,one_twelfth) , _mm_mul_pd(vvdw6,one_sixth) );
247 fvdw = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
249 /* Update potential sum for this i atom from the interaction with this j atom. */
250 vvdw = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
251 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
255 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
257 /* Calculate temporary vectorial force */
258 tx = _mm_mul_pd(fscal,dx00);
259 ty = _mm_mul_pd(fscal,dy00);
260 tz = _mm_mul_pd(fscal,dz00);
262 /* Update vectorial force */
263 fix0 = _mm_add_pd(fix0,tx);
264 fiy0 = _mm_add_pd(fiy0,ty);
265 fiz0 = _mm_add_pd(fiz0,tz);
267 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,tx,ty,tz);
269 /* Inner loop uses 32 flops */
272 /* End of innermost loop */
274 gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
275 f+i_coord_offset,fshift+i_shift_offset);
278 /* Update potential energies */
279 gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
281 /* Increment number of inner iterations */
282 inneriter += j_index_end - j_index_start;
284 /* Outer loop uses 7 flops */
287 /* Increment number of outer iterations */
290 /* Update outer/inner flops */
292 inc_nrnb(nrnb,eNR_NBKERNEL_VDW_VF,outeriter*7 + inneriter*32);
295 * Gromacs nonbonded kernel: nb_kernel_ElecNone_VdwLJ_GeomP1P1_F_sse2_double
296 * Electrostatics interaction: None
297 * VdW interaction: LennardJones
298 * Geometry: Particle-Particle
299 * Calculate force/pot: Force
302 nb_kernel_ElecNone_VdwLJ_GeomP1P1_F_sse2_double
303 (t_nblist * gmx_restrict nlist,
304 rvec * gmx_restrict xx,
305 rvec * gmx_restrict ff,
306 t_forcerec * gmx_restrict fr,
307 t_mdatoms * gmx_restrict mdatoms,
308 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
309 t_nrnb * gmx_restrict nrnb)
311 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
312 * just 0 for non-waters.
313 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
314 * jnr indices corresponding to data put in the four positions in the SIMD register.
316 int i_shift_offset,i_coord_offset,outeriter,inneriter;
317 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
319 int j_coord_offsetA,j_coord_offsetB;
320 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
322 real *shiftvec,*fshift,*x,*f;
323 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
325 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
326 int vdwjidx0A,vdwjidx0B;
327 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
328 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
330 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
333 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
334 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
335 __m128d dummy_mask,cutoff_mask;
336 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
337 __m128d one = _mm_set1_pd(1.0);
338 __m128d two = _mm_set1_pd(2.0);
344 jindex = nlist->jindex;
346 shiftidx = nlist->shift;
348 shiftvec = fr->shift_vec[0];
349 fshift = fr->fshift[0];
350 nvdwtype = fr->ntype;
352 vdwtype = mdatoms->typeA;
354 /* Avoid stupid compiler warnings */
362 /* Start outer loop over neighborlists */
363 for(iidx=0; iidx<nri; iidx++)
365 /* Load shift vector for this list */
366 i_shift_offset = DIM*shiftidx[iidx];
368 /* Load limits for loop over neighbors */
369 j_index_start = jindex[iidx];
370 j_index_end = jindex[iidx+1];
372 /* Get outer coordinate index */
374 i_coord_offset = DIM*inr;
376 /* Load i particle coords and add shift vector */
377 gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
379 fix0 = _mm_setzero_pd();
380 fiy0 = _mm_setzero_pd();
381 fiz0 = _mm_setzero_pd();
383 /* Load parameters for i particles */
384 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
386 /* Start inner kernel loop */
387 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
390 /* Get j neighbor index, and coordinate index */
393 j_coord_offsetA = DIM*jnrA;
394 j_coord_offsetB = DIM*jnrB;
396 /* load j atom coordinates */
397 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
400 /* Calculate displacement vector */
401 dx00 = _mm_sub_pd(ix0,jx0);
402 dy00 = _mm_sub_pd(iy0,jy0);
403 dz00 = _mm_sub_pd(iz0,jz0);
405 /* Calculate squared distance and things based on it */
406 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
408 rinvsq00 = gmx_mm_inv_pd(rsq00);
410 /* Load parameters for j particles */
411 vdwjidx0A = 2*vdwtype[jnrA+0];
412 vdwjidx0B = 2*vdwtype[jnrB+0];
414 /**************************
415 * CALCULATE INTERACTIONS *
416 **************************/
418 /* Compute parameters for interactions between i and j atoms */
419 gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
420 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
422 /* LENNARD-JONES DISPERSION/REPULSION */
424 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
425 fvdw = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
429 /* Calculate temporary vectorial force */
430 tx = _mm_mul_pd(fscal,dx00);
431 ty = _mm_mul_pd(fscal,dy00);
432 tz = _mm_mul_pd(fscal,dz00);
434 /* Update vectorial force */
435 fix0 = _mm_add_pd(fix0,tx);
436 fiy0 = _mm_add_pd(fiy0,ty);
437 fiz0 = _mm_add_pd(fiz0,tz);
439 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,tx,ty,tz);
441 /* Inner loop uses 27 flops */
448 j_coord_offsetA = DIM*jnrA;
450 /* load j atom coordinates */
451 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
454 /* Calculate displacement vector */
455 dx00 = _mm_sub_pd(ix0,jx0);
456 dy00 = _mm_sub_pd(iy0,jy0);
457 dz00 = _mm_sub_pd(iz0,jz0);
459 /* Calculate squared distance and things based on it */
460 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
462 rinvsq00 = gmx_mm_inv_pd(rsq00);
464 /* Load parameters for j particles */
465 vdwjidx0A = 2*vdwtype[jnrA+0];
467 /**************************
468 * CALCULATE INTERACTIONS *
469 **************************/
471 /* Compute parameters for interactions between i and j atoms */
472 gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
474 /* LENNARD-JONES DISPERSION/REPULSION */
476 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
477 fvdw = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
481 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
483 /* Calculate temporary vectorial force */
484 tx = _mm_mul_pd(fscal,dx00);
485 ty = _mm_mul_pd(fscal,dy00);
486 tz = _mm_mul_pd(fscal,dz00);
488 /* Update vectorial force */
489 fix0 = _mm_add_pd(fix0,tx);
490 fiy0 = _mm_add_pd(fiy0,ty);
491 fiz0 = _mm_add_pd(fiz0,tz);
493 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,tx,ty,tz);
495 /* Inner loop uses 27 flops */
498 /* End of innermost loop */
500 gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
501 f+i_coord_offset,fshift+i_shift_offset);
503 /* Increment number of inner iterations */
504 inneriter += j_index_end - j_index_start;
506 /* Outer loop uses 6 flops */
509 /* Increment number of outer iterations */
512 /* Update outer/inner flops */
514 inc_nrnb(nrnb,eNR_NBKERNEL_VDW_F,outeriter*6 + inneriter*27);