2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2012,2013,2014,2015, 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 "gromacs/math/vec.h"
46 #include "gromacs/legacyheaders/nrnb.h"
48 #include "gromacs/simd/math_x86_sse2_double.h"
49 #include "kernelutil_x86_sse2_double.h"
52 * Gromacs nonbonded kernel: nb_kernel_ElecNone_VdwLJ_GeomP1P1_VF_sse2_double
53 * Electrostatics interaction: None
54 * VdW interaction: LennardJones
55 * Geometry: Particle-Particle
56 * Calculate force/pot: PotentialAndForce
59 nb_kernel_ElecNone_VdwLJ_GeomP1P1_VF_sse2_double
60 (t_nblist
* gmx_restrict nlist
,
61 rvec
* gmx_restrict xx
,
62 rvec
* gmx_restrict ff
,
63 t_forcerec
* gmx_restrict fr
,
64 t_mdatoms
* gmx_restrict mdatoms
,
65 nb_kernel_data_t gmx_unused
* gmx_restrict kernel_data
,
66 t_nrnb
* gmx_restrict nrnb
)
68 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
69 * just 0 for non-waters.
70 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
71 * jnr indices corresponding to data put in the four positions in the SIMD register.
73 int i_shift_offset
,i_coord_offset
,outeriter
,inneriter
;
74 int j_index_start
,j_index_end
,jidx
,nri
,inr
,ggid
,iidx
;
76 int j_coord_offsetA
,j_coord_offsetB
;
77 int *iinr
,*jindex
,*jjnr
,*shiftidx
,*gid
;
79 real
*shiftvec
,*fshift
,*x
,*f
;
80 __m128d tx
,ty
,tz
,fscal
,rcutoff
,rcutoff2
,jidxall
;
82 __m128d ix0
,iy0
,iz0
,fix0
,fiy0
,fiz0
,iq0
,isai0
;
83 int vdwjidx0A
,vdwjidx0B
;
84 __m128d jx0
,jy0
,jz0
,fjx0
,fjy0
,fjz0
,jq0
,isaj0
;
85 __m128d dx00
,dy00
,dz00
,rsq00
,rinv00
,rinvsq00
,r00
,qq00
,c6_00
,c12_00
;
87 __m128d rinvsix
,rvdw
,vvdw
,vvdw6
,vvdw12
,fvdw
,fvdw6
,fvdw12
,vvdwsum
,sh_vdw_invrcut6
;
90 __m128d one_sixth
= _mm_set1_pd(1.0/6.0);
91 __m128d one_twelfth
= _mm_set1_pd(1.0/12.0);
92 __m128d dummy_mask
,cutoff_mask
;
93 __m128d signbit
= gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
94 __m128d one
= _mm_set1_pd(1.0);
95 __m128d two
= _mm_set1_pd(2.0);
101 jindex
= nlist
->jindex
;
103 shiftidx
= nlist
->shift
;
105 shiftvec
= fr
->shift_vec
[0];
106 fshift
= fr
->fshift
[0];
107 nvdwtype
= fr
->ntype
;
109 vdwtype
= mdatoms
->typeA
;
111 /* Avoid stupid compiler warnings */
119 /* Start outer loop over neighborlists */
120 for(iidx
=0; iidx
<nri
; iidx
++)
122 /* Load shift vector for this list */
123 i_shift_offset
= DIM
*shiftidx
[iidx
];
125 /* Load limits for loop over neighbors */
126 j_index_start
= jindex
[iidx
];
127 j_index_end
= jindex
[iidx
+1];
129 /* Get outer coordinate index */
131 i_coord_offset
= DIM
*inr
;
133 /* Load i particle coords and add shift vector */
134 gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec
+i_shift_offset
,x
+i_coord_offset
,&ix0
,&iy0
,&iz0
);
136 fix0
= _mm_setzero_pd();
137 fiy0
= _mm_setzero_pd();
138 fiz0
= _mm_setzero_pd();
140 /* Load parameters for i particles */
141 vdwioffset0
= 2*nvdwtype
*vdwtype
[inr
+0];
143 /* Reset potential sums */
144 vvdwsum
= _mm_setzero_pd();
146 /* Start inner kernel loop */
147 for(jidx
=j_index_start
; jidx
<j_index_end
-1; jidx
+=2)
150 /* Get j neighbor index, and coordinate index */
153 j_coord_offsetA
= DIM
*jnrA
;
154 j_coord_offsetB
= DIM
*jnrB
;
156 /* load j atom coordinates */
157 gmx_mm_load_1rvec_2ptr_swizzle_pd(x
+j_coord_offsetA
,x
+j_coord_offsetB
,
160 /* Calculate displacement vector */
161 dx00
= _mm_sub_pd(ix0
,jx0
);
162 dy00
= _mm_sub_pd(iy0
,jy0
);
163 dz00
= _mm_sub_pd(iz0
,jz0
);
165 /* Calculate squared distance and things based on it */
166 rsq00
= gmx_mm_calc_rsq_pd(dx00
,dy00
,dz00
);
168 rinvsq00
= gmx_mm_inv_pd(rsq00
);
170 /* Load parameters for j particles */
171 vdwjidx0A
= 2*vdwtype
[jnrA
+0];
172 vdwjidx0B
= 2*vdwtype
[jnrB
+0];
174 /**************************
175 * CALCULATE INTERACTIONS *
176 **************************/
178 /* Compute parameters for interactions between i and j atoms */
179 gmx_mm_load_2pair_swizzle_pd(vdwparam
+vdwioffset0
+vdwjidx0A
,
180 vdwparam
+vdwioffset0
+vdwjidx0B
,&c6_00
,&c12_00
);
182 /* LENNARD-JONES DISPERSION/REPULSION */
184 rinvsix
= _mm_mul_pd(_mm_mul_pd(rinvsq00
,rinvsq00
),rinvsq00
);
185 vvdw6
= _mm_mul_pd(c6_00
,rinvsix
);
186 vvdw12
= _mm_mul_pd(c12_00
,_mm_mul_pd(rinvsix
,rinvsix
));
187 vvdw
= _mm_sub_pd( _mm_mul_pd(vvdw12
,one_twelfth
) , _mm_mul_pd(vvdw6
,one_sixth
) );
188 fvdw
= _mm_mul_pd(_mm_sub_pd(vvdw12
,vvdw6
),rinvsq00
);
190 /* Update potential sum for this i atom from the interaction with this j atom. */
191 vvdwsum
= _mm_add_pd(vvdwsum
,vvdw
);
195 /* Calculate temporary vectorial force */
196 tx
= _mm_mul_pd(fscal
,dx00
);
197 ty
= _mm_mul_pd(fscal
,dy00
);
198 tz
= _mm_mul_pd(fscal
,dz00
);
200 /* Update vectorial force */
201 fix0
= _mm_add_pd(fix0
,tx
);
202 fiy0
= _mm_add_pd(fiy0
,ty
);
203 fiz0
= _mm_add_pd(fiz0
,tz
);
205 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f
+j_coord_offsetA
,f
+j_coord_offsetB
,tx
,ty
,tz
);
207 /* Inner loop uses 32 flops */
214 j_coord_offsetA
= DIM
*jnrA
;
216 /* load j atom coordinates */
217 gmx_mm_load_1rvec_1ptr_swizzle_pd(x
+j_coord_offsetA
,
220 /* Calculate displacement vector */
221 dx00
= _mm_sub_pd(ix0
,jx0
);
222 dy00
= _mm_sub_pd(iy0
,jy0
);
223 dz00
= _mm_sub_pd(iz0
,jz0
);
225 /* Calculate squared distance and things based on it */
226 rsq00
= gmx_mm_calc_rsq_pd(dx00
,dy00
,dz00
);
228 rinvsq00
= gmx_mm_inv_pd(rsq00
);
230 /* Load parameters for j particles */
231 vdwjidx0A
= 2*vdwtype
[jnrA
+0];
233 /**************************
234 * CALCULATE INTERACTIONS *
235 **************************/
237 /* Compute parameters for interactions between i and j atoms */
238 gmx_mm_load_1pair_swizzle_pd(vdwparam
+vdwioffset0
+vdwjidx0A
,&c6_00
,&c12_00
);
240 /* LENNARD-JONES DISPERSION/REPULSION */
242 rinvsix
= _mm_mul_pd(_mm_mul_pd(rinvsq00
,rinvsq00
),rinvsq00
);
243 vvdw6
= _mm_mul_pd(c6_00
,rinvsix
);
244 vvdw12
= _mm_mul_pd(c12_00
,_mm_mul_pd(rinvsix
,rinvsix
));
245 vvdw
= _mm_sub_pd( _mm_mul_pd(vvdw12
,one_twelfth
) , _mm_mul_pd(vvdw6
,one_sixth
) );
246 fvdw
= _mm_mul_pd(_mm_sub_pd(vvdw12
,vvdw6
),rinvsq00
);
248 /* Update potential sum for this i atom from the interaction with this j atom. */
249 vvdw
= _mm_unpacklo_pd(vvdw
,_mm_setzero_pd());
250 vvdwsum
= _mm_add_pd(vvdwsum
,vvdw
);
254 fscal
= _mm_unpacklo_pd(fscal
,_mm_setzero_pd());
256 /* Calculate temporary vectorial force */
257 tx
= _mm_mul_pd(fscal
,dx00
);
258 ty
= _mm_mul_pd(fscal
,dy00
);
259 tz
= _mm_mul_pd(fscal
,dz00
);
261 /* Update vectorial force */
262 fix0
= _mm_add_pd(fix0
,tx
);
263 fiy0
= _mm_add_pd(fiy0
,ty
);
264 fiz0
= _mm_add_pd(fiz0
,tz
);
266 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f
+j_coord_offsetA
,tx
,ty
,tz
);
268 /* Inner loop uses 32 flops */
271 /* End of innermost loop */
273 gmx_mm_update_iforce_1atom_swizzle_pd(fix0
,fiy0
,fiz0
,
274 f
+i_coord_offset
,fshift
+i_shift_offset
);
277 /* Update potential energies */
278 gmx_mm_update_1pot_pd(vvdwsum
,kernel_data
->energygrp_vdw
+ggid
);
280 /* Increment number of inner iterations */
281 inneriter
+= j_index_end
- j_index_start
;
283 /* Outer loop uses 7 flops */
286 /* Increment number of outer iterations */
289 /* Update outer/inner flops */
291 inc_nrnb(nrnb
,eNR_NBKERNEL_VDW_VF
,outeriter
*7 + inneriter
*32);
294 * Gromacs nonbonded kernel: nb_kernel_ElecNone_VdwLJ_GeomP1P1_F_sse2_double
295 * Electrostatics interaction: None
296 * VdW interaction: LennardJones
297 * Geometry: Particle-Particle
298 * Calculate force/pot: Force
301 nb_kernel_ElecNone_VdwLJ_GeomP1P1_F_sse2_double
302 (t_nblist
* gmx_restrict nlist
,
303 rvec
* gmx_restrict xx
,
304 rvec
* gmx_restrict ff
,
305 t_forcerec
* gmx_restrict fr
,
306 t_mdatoms
* gmx_restrict mdatoms
,
307 nb_kernel_data_t gmx_unused
* gmx_restrict kernel_data
,
308 t_nrnb
* gmx_restrict nrnb
)
310 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
311 * just 0 for non-waters.
312 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
313 * jnr indices corresponding to data put in the four positions in the SIMD register.
315 int i_shift_offset
,i_coord_offset
,outeriter
,inneriter
;
316 int j_index_start
,j_index_end
,jidx
,nri
,inr
,ggid
,iidx
;
318 int j_coord_offsetA
,j_coord_offsetB
;
319 int *iinr
,*jindex
,*jjnr
,*shiftidx
,*gid
;
321 real
*shiftvec
,*fshift
,*x
,*f
;
322 __m128d tx
,ty
,tz
,fscal
,rcutoff
,rcutoff2
,jidxall
;
324 __m128d ix0
,iy0
,iz0
,fix0
,fiy0
,fiz0
,iq0
,isai0
;
325 int vdwjidx0A
,vdwjidx0B
;
326 __m128d jx0
,jy0
,jz0
,fjx0
,fjy0
,fjz0
,jq0
,isaj0
;
327 __m128d dx00
,dy00
,dz00
,rsq00
,rinv00
,rinvsq00
,r00
,qq00
,c6_00
,c12_00
;
329 __m128d rinvsix
,rvdw
,vvdw
,vvdw6
,vvdw12
,fvdw
,fvdw6
,fvdw12
,vvdwsum
,sh_vdw_invrcut6
;
332 __m128d one_sixth
= _mm_set1_pd(1.0/6.0);
333 __m128d one_twelfth
= _mm_set1_pd(1.0/12.0);
334 __m128d dummy_mask
,cutoff_mask
;
335 __m128d signbit
= gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
336 __m128d one
= _mm_set1_pd(1.0);
337 __m128d two
= _mm_set1_pd(2.0);
343 jindex
= nlist
->jindex
;
345 shiftidx
= nlist
->shift
;
347 shiftvec
= fr
->shift_vec
[0];
348 fshift
= fr
->fshift
[0];
349 nvdwtype
= fr
->ntype
;
351 vdwtype
= mdatoms
->typeA
;
353 /* Avoid stupid compiler warnings */
361 /* Start outer loop over neighborlists */
362 for(iidx
=0; iidx
<nri
; iidx
++)
364 /* Load shift vector for this list */
365 i_shift_offset
= DIM
*shiftidx
[iidx
];
367 /* Load limits for loop over neighbors */
368 j_index_start
= jindex
[iidx
];
369 j_index_end
= jindex
[iidx
+1];
371 /* Get outer coordinate index */
373 i_coord_offset
= DIM
*inr
;
375 /* Load i particle coords and add shift vector */
376 gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec
+i_shift_offset
,x
+i_coord_offset
,&ix0
,&iy0
,&iz0
);
378 fix0
= _mm_setzero_pd();
379 fiy0
= _mm_setzero_pd();
380 fiz0
= _mm_setzero_pd();
382 /* Load parameters for i particles */
383 vdwioffset0
= 2*nvdwtype
*vdwtype
[inr
+0];
385 /* Start inner kernel loop */
386 for(jidx
=j_index_start
; jidx
<j_index_end
-1; jidx
+=2)
389 /* Get j neighbor index, and coordinate index */
392 j_coord_offsetA
= DIM
*jnrA
;
393 j_coord_offsetB
= DIM
*jnrB
;
395 /* load j atom coordinates */
396 gmx_mm_load_1rvec_2ptr_swizzle_pd(x
+j_coord_offsetA
,x
+j_coord_offsetB
,
399 /* Calculate displacement vector */
400 dx00
= _mm_sub_pd(ix0
,jx0
);
401 dy00
= _mm_sub_pd(iy0
,jy0
);
402 dz00
= _mm_sub_pd(iz0
,jz0
);
404 /* Calculate squared distance and things based on it */
405 rsq00
= gmx_mm_calc_rsq_pd(dx00
,dy00
,dz00
);
407 rinvsq00
= gmx_mm_inv_pd(rsq00
);
409 /* Load parameters for j particles */
410 vdwjidx0A
= 2*vdwtype
[jnrA
+0];
411 vdwjidx0B
= 2*vdwtype
[jnrB
+0];
413 /**************************
414 * CALCULATE INTERACTIONS *
415 **************************/
417 /* Compute parameters for interactions between i and j atoms */
418 gmx_mm_load_2pair_swizzle_pd(vdwparam
+vdwioffset0
+vdwjidx0A
,
419 vdwparam
+vdwioffset0
+vdwjidx0B
,&c6_00
,&c12_00
);
421 /* LENNARD-JONES DISPERSION/REPULSION */
423 rinvsix
= _mm_mul_pd(_mm_mul_pd(rinvsq00
,rinvsq00
),rinvsq00
);
424 fvdw
= _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00
,rinvsix
),c6_00
),_mm_mul_pd(rinvsix
,rinvsq00
));
428 /* Calculate temporary vectorial force */
429 tx
= _mm_mul_pd(fscal
,dx00
);
430 ty
= _mm_mul_pd(fscal
,dy00
);
431 tz
= _mm_mul_pd(fscal
,dz00
);
433 /* Update vectorial force */
434 fix0
= _mm_add_pd(fix0
,tx
);
435 fiy0
= _mm_add_pd(fiy0
,ty
);
436 fiz0
= _mm_add_pd(fiz0
,tz
);
438 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f
+j_coord_offsetA
,f
+j_coord_offsetB
,tx
,ty
,tz
);
440 /* Inner loop uses 27 flops */
447 j_coord_offsetA
= DIM
*jnrA
;
449 /* load j atom coordinates */
450 gmx_mm_load_1rvec_1ptr_swizzle_pd(x
+j_coord_offsetA
,
453 /* Calculate displacement vector */
454 dx00
= _mm_sub_pd(ix0
,jx0
);
455 dy00
= _mm_sub_pd(iy0
,jy0
);
456 dz00
= _mm_sub_pd(iz0
,jz0
);
458 /* Calculate squared distance and things based on it */
459 rsq00
= gmx_mm_calc_rsq_pd(dx00
,dy00
,dz00
);
461 rinvsq00
= gmx_mm_inv_pd(rsq00
);
463 /* Load parameters for j particles */
464 vdwjidx0A
= 2*vdwtype
[jnrA
+0];
466 /**************************
467 * CALCULATE INTERACTIONS *
468 **************************/
470 /* Compute parameters for interactions between i and j atoms */
471 gmx_mm_load_1pair_swizzle_pd(vdwparam
+vdwioffset0
+vdwjidx0A
,&c6_00
,&c12_00
);
473 /* LENNARD-JONES DISPERSION/REPULSION */
475 rinvsix
= _mm_mul_pd(_mm_mul_pd(rinvsq00
,rinvsq00
),rinvsq00
);
476 fvdw
= _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00
,rinvsix
),c6_00
),_mm_mul_pd(rinvsix
,rinvsq00
));
480 fscal
= _mm_unpacklo_pd(fscal
,_mm_setzero_pd());
482 /* Calculate temporary vectorial force */
483 tx
= _mm_mul_pd(fscal
,dx00
);
484 ty
= _mm_mul_pd(fscal
,dy00
);
485 tz
= _mm_mul_pd(fscal
,dz00
);
487 /* Update vectorial force */
488 fix0
= _mm_add_pd(fix0
,tx
);
489 fiy0
= _mm_add_pd(fiy0
,ty
);
490 fiz0
= _mm_add_pd(fiz0
,tz
);
492 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f
+j_coord_offsetA
,tx
,ty
,tz
);
494 /* Inner loop uses 27 flops */
497 /* End of innermost loop */
499 gmx_mm_update_iforce_1atom_swizzle_pd(fix0
,fiy0
,fiz0
,
500 f
+i_coord_offset
,fshift
+i_shift_offset
);
502 /* Increment number of inner iterations */
503 inneriter
+= j_index_end
- j_index_start
;
505 /* Outer loop uses 6 flops */
508 /* Increment number of outer iterations */
511 /* Update outer/inner flops */
513 inc_nrnb(nrnb
,eNR_NBKERNEL_VDW_F
,outeriter
*6 + inneriter
*27);