?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Fix segmentation fault in minimize
[gromacs.git] / src / gromacs / gmxlib / nonbonded / nb_kernel_avx_256_double / nb_kernel_ElecNone_VdwLJEwSh_GeomP1P1_avx_256_double.cpp
blob16c927c00bae8ac391e07b245e79f4bcfcbcd2d9
1 /*
2 * This file is part of the GROMACS molecular simulation package.
3 *
4 * Copyright (c) 2012,2013,2014,2015,2017,2018, 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.
8 *
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.
13 *
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.
18 *
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.
23 *
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.
31 *
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.
34 */
35 /*
36 * Note: this file was generated by the GROMACS avx_256_double kernel generator.
37 */
38 #include "gmxpre.h"
39
40 #include "config.h"
41
42 #include <math.h>
43
44 #include "../nb_kernel.h"
45 #include "gromacs/gmxlib/nrnb.h"
46
47 #include "kernelutil_x86_avx_256_double.h"
48
49 /*
50 * Gromacs nonbonded kernel: nb_kernel_ElecNone_VdwLJEwSh_GeomP1P1_VF_avx_256_double
51 * Electrostatics interaction: None
52 * VdW interaction: LJEwald
53 * Geometry: Particle-Particle
54 * Calculate force/pot: PotentialAndForce
55 */
56 void
57 nb_kernel_ElecNone_VdwLJEwSh_GeomP1P1_VF_avx_256_double
58 (t_nblist * gmx_restrict nlist,
59 rvec * gmx_restrict xx,
60 rvec * gmx_restrict ff,
61 struct t_forcerec * gmx_restrict fr,
62 t_mdatoms * gmx_restrict mdatoms,
63 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
64 t_nrnb * gmx_restrict nrnb)
65 {
66 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
67 * just 0 for non-waters.
68 * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
69 * jnr indices corresponding to data put in the four positions in the SIMD register.
70 */
71 int i_shift_offset,i_coord_offset,outeriter,inneriter;
72 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
73 int jnrA,jnrB,jnrC,jnrD;
74 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
75 int jnrlistE,jnrlistF,jnrlistG,jnrlistH;
76 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
77 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
78 real rcutoff_scalar;
79 real *shiftvec,*fshift,*x,*f;
80 real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
81 real scratch[4*DIM];
82 __m256d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
83 real * vdwioffsetptr0;
84 real * vdwgridioffsetptr0;
85 __m256d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
86 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
87 __m256d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
88 __m256d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
89 int nvdwtype;
90 __m256d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
91 int *vdwtype;
92 real *vdwparam;
93 __m256d one_sixth = _mm256_set1_pd(1.0/6.0);
94 __m256d one_twelfth = _mm256_set1_pd(1.0/12.0);
95 __m256d c6grid_00;
96 real *vdwgridparam;
97 __m256d ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
98 __m256d one_half = _mm256_set1_pd(0.5);
99 __m256d minus_one = _mm256_set1_pd(-1.0);
100 __m256d dummy_mask,cutoff_mask;
101 __m128 tmpmask0,tmpmask1;
102 __m256d signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
103 __m256d one = _mm256_set1_pd(1.0);
104 __m256d two = _mm256_set1_pd(2.0);
105 x = xx[0];
106 f = ff[0];
107
108 nri = nlist->nri;
109 iinr = nlist->iinr;
110 jindex = nlist->jindex;
111 jjnr = nlist->jjnr;
112 shiftidx = nlist->shift;
113 gid = nlist->gid;
114 shiftvec = fr->shift_vec[0];
115 fshift = fr->fshift[0];
116 nvdwtype = fr->ntype;
117 vdwparam = fr->nbfp;
118 vdwtype = mdatoms->typeA;
119 vdwgridparam = fr->ljpme_c6grid;
120 sh_lj_ewald = _mm256_set1_pd(fr->ic->sh_lj_ewald);
121 ewclj = _mm256_set1_pd(fr->ic->ewaldcoeff_lj);
122 ewclj2 = _mm256_mul_pd(minus_one,_mm256_mul_pd(ewclj,ewclj));
123
124 rcutoff_scalar = fr->ic->rvdw;
125 rcutoff = _mm256_set1_pd(rcutoff_scalar);
126 rcutoff2 = _mm256_mul_pd(rcutoff,rcutoff);
127
128 sh_vdw_invrcut6 = _mm256_set1_pd(fr->ic->sh_invrc6);
129 rvdw = _mm256_set1_pd(fr->ic->rvdw);
130
131 /* Avoid stupid compiler warnings */
132 jnrA = jnrB = jnrC = jnrD = 0;
133 j_coord_offsetA = 0;
134 j_coord_offsetB = 0;
135 j_coord_offsetC = 0;
136 j_coord_offsetD = 0;
137
138 outeriter = 0;
139 inneriter = 0;
140
141 for(iidx=0;iidx<4*DIM;iidx++)
142 {
143 scratch[iidx] = 0.0;
144 }
145
146 /* Start outer loop over neighborlists */
147 for(iidx=0; iidx<nri; iidx++)
148 {
149 /* Load shift vector for this list */
150 i_shift_offset = DIM*shiftidx[iidx];
151
152 /* Load limits for loop over neighbors */
153 j_index_start = jindex[iidx];
154 j_index_end = jindex[iidx+1];
155
156 /* Get outer coordinate index */
157 inr = iinr[iidx];
158 i_coord_offset = DIM*inr;
159
160 /* Load i particle coords and add shift vector */
161 gmx_mm256_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
162
163 fix0 = _mm256_setzero_pd();
164 fiy0 = _mm256_setzero_pd();
165 fiz0 = _mm256_setzero_pd();
166
167 /* Load parameters for i particles */
168 vdwioffsetptr0 = vdwparam+2*nvdwtype*vdwtype[inr+0];
169 vdwgridioffsetptr0 = vdwgridparam+2*nvdwtype*vdwtype[inr+0];
170
171 /* Reset potential sums */
172 vvdwsum = _mm256_setzero_pd();
173
174 /* Start inner kernel loop */
175 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
176 {
177
178 /* Get j neighbor index, and coordinate index */
179 jnrA = jjnr[jidx];
180 jnrB = jjnr[jidx+1];
181 jnrC = jjnr[jidx+2];
182 jnrD = jjnr[jidx+3];
183 j_coord_offsetA = DIM*jnrA;
184 j_coord_offsetB = DIM*jnrB;
185 j_coord_offsetC = DIM*jnrC;
186 j_coord_offsetD = DIM*jnrD;
187
188 /* load j atom coordinates */
189 gmx_mm256_load_1rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
190 x+j_coord_offsetC,x+j_coord_offsetD,
191 &jx0,&jy0,&jz0);
192
193 /* Calculate displacement vector */
194 dx00 = _mm256_sub_pd(ix0,jx0);
195 dy00 = _mm256_sub_pd(iy0,jy0);
196 dz00 = _mm256_sub_pd(iz0,jz0);
197
198 /* Calculate squared distance and things based on it */
199 rsq00 = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
200
201 rinv00 = avx256_invsqrt_d(rsq00);
202
203 rinvsq00 = _mm256_mul_pd(rinv00,rinv00);
204
205 /* Load parameters for j particles */
206 vdwjidx0A = 2*vdwtype[jnrA+0];
207 vdwjidx0B = 2*vdwtype[jnrB+0];
208 vdwjidx0C = 2*vdwtype[jnrC+0];
209 vdwjidx0D = 2*vdwtype[jnrD+0];
210
211 /**************************
212 * CALCULATE INTERACTIONS *
213 **************************/
214
215 if (gmx_mm256_any_lt(rsq00,rcutoff2))
216 {
217
218 r00 = _mm256_mul_pd(rsq00,rinv00);
219
220 /* Compute parameters for interactions between i and j atoms */
221 gmx_mm256_load_4pair_swizzle_pd(vdwioffsetptr0+vdwjidx0A,
222 vdwioffsetptr0+vdwjidx0B,
223 vdwioffsetptr0+vdwjidx0C,
224 vdwioffsetptr0+vdwjidx0D,
225 &c6_00,&c12_00);
226
227 c6grid_00 = gmx_mm256_load_4real_swizzle_pd(vdwgridioffsetptr0+vdwjidx0A,
228 vdwgridioffsetptr0+vdwjidx0B,
229 vdwgridioffsetptr0+vdwjidx0C,
230 vdwgridioffsetptr0+vdwjidx0D);
231
232 /* Analytical LJ-PME */
233 rinvsix = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
234 ewcljrsq = _mm256_mul_pd(ewclj2,rsq00);
235 ewclj6 = _mm256_mul_pd(ewclj2,_mm256_mul_pd(ewclj2,ewclj2));
236 exponent = avx256_exp_d(ewcljrsq);
237 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
238 poly = _mm256_mul_pd(exponent,_mm256_add_pd(_mm256_sub_pd(one,ewcljrsq),_mm256_mul_pd(_mm256_mul_pd(ewcljrsq,ewcljrsq),one_half)));
239 /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
240 vvdw6 = _mm256_mul_pd(_mm256_sub_pd(c6_00,_mm256_mul_pd(c6grid_00,_mm256_sub_pd(one,poly))),rinvsix);
241 vvdw12 = _mm256_mul_pd(c12_00,_mm256_mul_pd(rinvsix,rinvsix));
242 vvdw = _mm256_sub_pd(_mm256_mul_pd( _mm256_sub_pd(vvdw12 , _mm256_mul_pd(c12_00,_mm256_mul_pd(sh_vdw_invrcut6,sh_vdw_invrcut6))), one_twelfth) ,
243 _mm256_mul_pd( _mm256_sub_pd(vvdw6,_mm256_add_pd(_mm256_mul_pd(c6_00,sh_vdw_invrcut6),_mm256_mul_pd(c6grid_00,sh_lj_ewald))),one_sixth));
244 /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
245 fvdw = _mm256_mul_pd(_mm256_sub_pd(vvdw12,_mm256_sub_pd(vvdw6,_mm256_mul_pd(_mm256_mul_pd(c6grid_00,one_sixth),_mm256_mul_pd(exponent,ewclj6)))),rinvsq00);
246
247 cutoff_mask = _mm256_cmp_pd(rsq00,rcutoff2,_CMP_LT_OQ);
248
249 /* Update potential sum for this i atom from the interaction with this j atom. */
250 vvdw = _mm256_and_pd(vvdw,cutoff_mask);
251 vvdwsum = _mm256_add_pd(vvdwsum,vvdw);
252
253 fscal = fvdw;
254
255 fscal = _mm256_and_pd(fscal,cutoff_mask);
256
257 /* Calculate temporary vectorial force */
258 tx = _mm256_mul_pd(fscal,dx00);
259 ty = _mm256_mul_pd(fscal,dy00);
260 tz = _mm256_mul_pd(fscal,dz00);
261
262 /* Update vectorial force */
263 fix0 = _mm256_add_pd(fix0,tx);
264 fiy0 = _mm256_add_pd(fiy0,ty);
265 fiz0 = _mm256_add_pd(fiz0,tz);
266
267 fjptrA = f+j_coord_offsetA;
268 fjptrB = f+j_coord_offsetB;
269 fjptrC = f+j_coord_offsetC;
270 fjptrD = f+j_coord_offsetD;
271 gmx_mm256_decrement_1rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,tx,ty,tz);
272
273 }
274
275 /* Inner loop uses 62 flops */
276 }
277
278 if(jidx<j_index_end)
279 {
280
281 /* Get j neighbor index, and coordinate index */
282 jnrlistA = jjnr[jidx];
283 jnrlistB = jjnr[jidx+1];
284 jnrlistC = jjnr[jidx+2];
285 jnrlistD = jjnr[jidx+3];
286 /* Sign of each element will be negative for non-real atoms.
287 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
288 * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
289 */
290 tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
291
292 tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
293 tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
294 dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
295
296 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
297 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
298 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
299 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
300 j_coord_offsetA = DIM*jnrA;
301 j_coord_offsetB = DIM*jnrB;
302 j_coord_offsetC = DIM*jnrC;
303 j_coord_offsetD = DIM*jnrD;
304
305 /* load j atom coordinates */
306 gmx_mm256_load_1rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
307 x+j_coord_offsetC,x+j_coord_offsetD,
308 &jx0,&jy0,&jz0);
309
310 /* Calculate displacement vector */
311 dx00 = _mm256_sub_pd(ix0,jx0);
312 dy00 = _mm256_sub_pd(iy0,jy0);
313 dz00 = _mm256_sub_pd(iz0,jz0);
314
315 /* Calculate squared distance and things based on it */
316 rsq00 = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
317
318 rinv00 = avx256_invsqrt_d(rsq00);
319
320 rinvsq00 = _mm256_mul_pd(rinv00,rinv00);
321
322 /* Load parameters for j particles */
323 vdwjidx0A = 2*vdwtype[jnrA+0];
324 vdwjidx0B = 2*vdwtype[jnrB+0];
325 vdwjidx0C = 2*vdwtype[jnrC+0];
326 vdwjidx0D = 2*vdwtype[jnrD+0];
327
328 /**************************
329 * CALCULATE INTERACTIONS *
330 **************************/
331
332 if (gmx_mm256_any_lt(rsq00,rcutoff2))
333 {
334
335 r00 = _mm256_mul_pd(rsq00,rinv00);
336 r00 = _mm256_andnot_pd(dummy_mask,r00);
337
338 /* Compute parameters for interactions between i and j atoms */
339 gmx_mm256_load_4pair_swizzle_pd(vdwioffsetptr0+vdwjidx0A,
340 vdwioffsetptr0+vdwjidx0B,
341 vdwioffsetptr0+vdwjidx0C,
342 vdwioffsetptr0+vdwjidx0D,
343 &c6_00,&c12_00);
344
345 c6grid_00 = gmx_mm256_load_4real_swizzle_pd(vdwgridioffsetptr0+vdwjidx0A,
346 vdwgridioffsetptr0+vdwjidx0B,
347 vdwgridioffsetptr0+vdwjidx0C,
348 vdwgridioffsetptr0+vdwjidx0D);
349
350 /* Analytical LJ-PME */
351 rinvsix = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
352 ewcljrsq = _mm256_mul_pd(ewclj2,rsq00);
353 ewclj6 = _mm256_mul_pd(ewclj2,_mm256_mul_pd(ewclj2,ewclj2));
354 exponent = avx256_exp_d(ewcljrsq);
355 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
356 poly = _mm256_mul_pd(exponent,_mm256_add_pd(_mm256_sub_pd(one,ewcljrsq),_mm256_mul_pd(_mm256_mul_pd(ewcljrsq,ewcljrsq),one_half)));
357 /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
358 vvdw6 = _mm256_mul_pd(_mm256_sub_pd(c6_00,_mm256_mul_pd(c6grid_00,_mm256_sub_pd(one,poly))),rinvsix);
359 vvdw12 = _mm256_mul_pd(c12_00,_mm256_mul_pd(rinvsix,rinvsix));
360 vvdw = _mm256_sub_pd(_mm256_mul_pd( _mm256_sub_pd(vvdw12 , _mm256_mul_pd(c12_00,_mm256_mul_pd(sh_vdw_invrcut6,sh_vdw_invrcut6))), one_twelfth) ,
361 _mm256_mul_pd( _mm256_sub_pd(vvdw6,_mm256_add_pd(_mm256_mul_pd(c6_00,sh_vdw_invrcut6),_mm256_mul_pd(c6grid_00,sh_lj_ewald))),one_sixth));
362 /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
363 fvdw = _mm256_mul_pd(_mm256_sub_pd(vvdw12,_mm256_sub_pd(vvdw6,_mm256_mul_pd(_mm256_mul_pd(c6grid_00,one_sixth),_mm256_mul_pd(exponent,ewclj6)))),rinvsq00);
364
365 cutoff_mask = _mm256_cmp_pd(rsq00,rcutoff2,_CMP_LT_OQ);
366
367 /* Update potential sum for this i atom from the interaction with this j atom. */
368 vvdw = _mm256_and_pd(vvdw,cutoff_mask);
369 vvdw = _mm256_andnot_pd(dummy_mask,vvdw);
370 vvdwsum = _mm256_add_pd(vvdwsum,vvdw);
371
372 fscal = fvdw;
373
374 fscal = _mm256_and_pd(fscal,cutoff_mask);
375
376 fscal = _mm256_andnot_pd(dummy_mask,fscal);
377
378 /* Calculate temporary vectorial force */
379 tx = _mm256_mul_pd(fscal,dx00);
380 ty = _mm256_mul_pd(fscal,dy00);
381 tz = _mm256_mul_pd(fscal,dz00);
382
383 /* Update vectorial force */
384 fix0 = _mm256_add_pd(fix0,tx);
385 fiy0 = _mm256_add_pd(fiy0,ty);
386 fiz0 = _mm256_add_pd(fiz0,tz);
387
388 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
389 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
390 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
391 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
392 gmx_mm256_decrement_1rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,tx,ty,tz);
393
394 }
395
396 /* Inner loop uses 63 flops */
397 }
398
399 /* End of innermost loop */
400
401 gmx_mm256_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
402 f+i_coord_offset,fshift+i_shift_offset);
403
404 ggid = gid[iidx];
405 /* Update potential energies */
406 gmx_mm256_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
407
408 /* Increment number of inner iterations */
409 inneriter += j_index_end - j_index_start;
410
411 /* Outer loop uses 7 flops */
412 }
413
414 /* Increment number of outer iterations */
415 outeriter += nri;
416
417 /* Update outer/inner flops */
418
419 inc_nrnb(nrnb,eNR_NBKERNEL_VDW_VF,outeriter*7 + inneriter*63);
420 }
421 /*
422 * Gromacs nonbonded kernel: nb_kernel_ElecNone_VdwLJEwSh_GeomP1P1_F_avx_256_double
423 * Electrostatics interaction: None
424 * VdW interaction: LJEwald
425 * Geometry: Particle-Particle
426 * Calculate force/pot: Force
427 */
428 void
429 nb_kernel_ElecNone_VdwLJEwSh_GeomP1P1_F_avx_256_double
430 (t_nblist * gmx_restrict nlist,
431 rvec * gmx_restrict xx,
432 rvec * gmx_restrict ff,
433 struct t_forcerec * gmx_restrict fr,
434 t_mdatoms * gmx_restrict mdatoms,
435 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
436 t_nrnb * gmx_restrict nrnb)
437 {
438 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
439 * just 0 for non-waters.
440 * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
441 * jnr indices corresponding to data put in the four positions in the SIMD register.
442 */
443 int i_shift_offset,i_coord_offset,outeriter,inneriter;
444 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
445 int jnrA,jnrB,jnrC,jnrD;
446 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
447 int jnrlistE,jnrlistF,jnrlistG,jnrlistH;
448 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
449 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
450 real rcutoff_scalar;
451 real *shiftvec,*fshift,*x,*f;
452 real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
453 real scratch[4*DIM];
454 __m256d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
455 real * vdwioffsetptr0;
456 real * vdwgridioffsetptr0;
457 __m256d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
458 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
459 __m256d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
460 __m256d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
461 int nvdwtype;
462 __m256d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
463 int *vdwtype;
464 real *vdwparam;
465 __m256d one_sixth = _mm256_set1_pd(1.0/6.0);
466 __m256d one_twelfth = _mm256_set1_pd(1.0/12.0);
467 __m256d c6grid_00;
468 real *vdwgridparam;
469 __m256d ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
470 __m256d one_half = _mm256_set1_pd(0.5);
471 __m256d minus_one = _mm256_set1_pd(-1.0);
472 __m256d dummy_mask,cutoff_mask;
473 __m128 tmpmask0,tmpmask1;
474 __m256d signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
475 __m256d one = _mm256_set1_pd(1.0);
476 __m256d two = _mm256_set1_pd(2.0);
477 x = xx[0];
478 f = ff[0];
479
480 nri = nlist->nri;
481 iinr = nlist->iinr;
482 jindex = nlist->jindex;
483 jjnr = nlist->jjnr;
484 shiftidx = nlist->shift;
485 gid = nlist->gid;
486 shiftvec = fr->shift_vec[0];
487 fshift = fr->fshift[0];
488 nvdwtype = fr->ntype;
489 vdwparam = fr->nbfp;
490 vdwtype = mdatoms->typeA;
491 vdwgridparam = fr->ljpme_c6grid;
492 sh_lj_ewald = _mm256_set1_pd(fr->ic->sh_lj_ewald);
493 ewclj = _mm256_set1_pd(fr->ic->ewaldcoeff_lj);
494 ewclj2 = _mm256_mul_pd(minus_one,_mm256_mul_pd(ewclj,ewclj));
495
496 rcutoff_scalar = fr->ic->rvdw;
497 rcutoff = _mm256_set1_pd(rcutoff_scalar);
498 rcutoff2 = _mm256_mul_pd(rcutoff,rcutoff);
499
500 sh_vdw_invrcut6 = _mm256_set1_pd(fr->ic->sh_invrc6);
501 rvdw = _mm256_set1_pd(fr->ic->rvdw);
502
503 /* Avoid stupid compiler warnings */
504 jnrA = jnrB = jnrC = jnrD = 0;
505 j_coord_offsetA = 0;
506 j_coord_offsetB = 0;
507 j_coord_offsetC = 0;
508 j_coord_offsetD = 0;
509
510 outeriter = 0;
511 inneriter = 0;
512
513 for(iidx=0;iidx<4*DIM;iidx++)
514 {
515 scratch[iidx] = 0.0;
516 }
517
518 /* Start outer loop over neighborlists */
519 for(iidx=0; iidx<nri; iidx++)
520 {
521 /* Load shift vector for this list */
522 i_shift_offset = DIM*shiftidx[iidx];
523
524 /* Load limits for loop over neighbors */
525 j_index_start = jindex[iidx];
526 j_index_end = jindex[iidx+1];
527
528 /* Get outer coordinate index */
529 inr = iinr[iidx];
530 i_coord_offset = DIM*inr;
531
532 /* Load i particle coords and add shift vector */
533 gmx_mm256_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
534
535 fix0 = _mm256_setzero_pd();
536 fiy0 = _mm256_setzero_pd();
537 fiz0 = _mm256_setzero_pd();
538
539 /* Load parameters for i particles */
540 vdwioffsetptr0 = vdwparam+2*nvdwtype*vdwtype[inr+0];
541 vdwgridioffsetptr0 = vdwgridparam+2*nvdwtype*vdwtype[inr+0];
542
543 /* Start inner kernel loop */
544 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
545 {
546
547 /* Get j neighbor index, and coordinate index */
548 jnrA = jjnr[jidx];
549 jnrB = jjnr[jidx+1];
550 jnrC = jjnr[jidx+2];
551 jnrD = jjnr[jidx+3];
552 j_coord_offsetA = DIM*jnrA;
553 j_coord_offsetB = DIM*jnrB;
554 j_coord_offsetC = DIM*jnrC;
555 j_coord_offsetD = DIM*jnrD;
556
557 /* load j atom coordinates */
558 gmx_mm256_load_1rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
559 x+j_coord_offsetC,x+j_coord_offsetD,
560 &jx0,&jy0,&jz0);
561
562 /* Calculate displacement vector */
563 dx00 = _mm256_sub_pd(ix0,jx0);
564 dy00 = _mm256_sub_pd(iy0,jy0);
565 dz00 = _mm256_sub_pd(iz0,jz0);
566
567 /* Calculate squared distance and things based on it */
568 rsq00 = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
569
570 rinv00 = avx256_invsqrt_d(rsq00);
571
572 rinvsq00 = _mm256_mul_pd(rinv00,rinv00);
573
574 /* Load parameters for j particles */
575 vdwjidx0A = 2*vdwtype[jnrA+0];
576 vdwjidx0B = 2*vdwtype[jnrB+0];
577 vdwjidx0C = 2*vdwtype[jnrC+0];
578 vdwjidx0D = 2*vdwtype[jnrD+0];
579
580 /**************************
581 * CALCULATE INTERACTIONS *
582 **************************/
583
584 if (gmx_mm256_any_lt(rsq00,rcutoff2))
585 {
586
587 r00 = _mm256_mul_pd(rsq00,rinv00);
588
589 /* Compute parameters for interactions between i and j atoms */
590 gmx_mm256_load_4pair_swizzle_pd(vdwioffsetptr0+vdwjidx0A,
591 vdwioffsetptr0+vdwjidx0B,
592 vdwioffsetptr0+vdwjidx0C,
593 vdwioffsetptr0+vdwjidx0D,
594 &c6_00,&c12_00);
595
596 c6grid_00 = gmx_mm256_load_4real_swizzle_pd(vdwgridioffsetptr0+vdwjidx0A,
597 vdwgridioffsetptr0+vdwjidx0B,
598 vdwgridioffsetptr0+vdwjidx0C,
599 vdwgridioffsetptr0+vdwjidx0D);
600
601 /* Analytical LJ-PME */
602 rinvsix = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
603 ewcljrsq = _mm256_mul_pd(ewclj2,rsq00);
604 ewclj6 = _mm256_mul_pd(ewclj2,_mm256_mul_pd(ewclj2,ewclj2));
605 exponent = avx256_exp_d(ewcljrsq);
606 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
607 poly = _mm256_mul_pd(exponent,_mm256_add_pd(_mm256_sub_pd(one,ewcljrsq),_mm256_mul_pd(_mm256_mul_pd(ewcljrsq,ewcljrsq),one_half)));
608 /* f6A = 6 * C6grid * (1 - poly) */
609 f6A = _mm256_mul_pd(c6grid_00,_mm256_sub_pd(one,poly));
610 /* f6B = C6grid * exponent * beta^6 */
611 f6B = _mm256_mul_pd(_mm256_mul_pd(c6grid_00,one_sixth),_mm256_mul_pd(exponent,ewclj6));
612 /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
613 fvdw = _mm256_mul_pd(_mm256_add_pd(_mm256_mul_pd(_mm256_sub_pd(_mm256_mul_pd(c12_00,rinvsix),_mm256_sub_pd(c6_00,f6A)),rinvsix),f6B),rinvsq00);
614
615 cutoff_mask = _mm256_cmp_pd(rsq00,rcutoff2,_CMP_LT_OQ);
616
617 fscal = fvdw;
618
619 fscal = _mm256_and_pd(fscal,cutoff_mask);
620
621 /* Calculate temporary vectorial force */
622 tx = _mm256_mul_pd(fscal,dx00);
623 ty = _mm256_mul_pd(fscal,dy00);
624 tz = _mm256_mul_pd(fscal,dz00);
625
626 /* Update vectorial force */
627 fix0 = _mm256_add_pd(fix0,tx);
628 fiy0 = _mm256_add_pd(fiy0,ty);
629 fiz0 = _mm256_add_pd(fiz0,tz);
630
631 fjptrA = f+j_coord_offsetA;
632 fjptrB = f+j_coord_offsetB;
633 fjptrC = f+j_coord_offsetC;
634 fjptrD = f+j_coord_offsetD;
635 gmx_mm256_decrement_1rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,tx,ty,tz);
636
637 }
638
639 /* Inner loop uses 49 flops */
640 }
641
642 if(jidx<j_index_end)
643 {
644
645 /* Get j neighbor index, and coordinate index */
646 jnrlistA = jjnr[jidx];
647 jnrlistB = jjnr[jidx+1];
648 jnrlistC = jjnr[jidx+2];
649 jnrlistD = jjnr[jidx+3];
650 /* Sign of each element will be negative for non-real atoms.
651 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
652 * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
653 */
654 tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
655
656 tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
657 tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
658 dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
659
660 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
661 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
662 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
663 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
664 j_coord_offsetA = DIM*jnrA;
665 j_coord_offsetB = DIM*jnrB;
666 j_coord_offsetC = DIM*jnrC;
667 j_coord_offsetD = DIM*jnrD;
668
669 /* load j atom coordinates */
670 gmx_mm256_load_1rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
671 x+j_coord_offsetC,x+j_coord_offsetD,
672 &jx0,&jy0,&jz0);
673
674 /* Calculate displacement vector */
675 dx00 = _mm256_sub_pd(ix0,jx0);
676 dy00 = _mm256_sub_pd(iy0,jy0);
677 dz00 = _mm256_sub_pd(iz0,jz0);
678
679 /* Calculate squared distance and things based on it */
680 rsq00 = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
681
682 rinv00 = avx256_invsqrt_d(rsq00);
683
684 rinvsq00 = _mm256_mul_pd(rinv00,rinv00);
685
686 /* Load parameters for j particles */
687 vdwjidx0A = 2*vdwtype[jnrA+0];
688 vdwjidx0B = 2*vdwtype[jnrB+0];
689 vdwjidx0C = 2*vdwtype[jnrC+0];
690 vdwjidx0D = 2*vdwtype[jnrD+0];
691
692 /**************************
693 * CALCULATE INTERACTIONS *
694 **************************/
695
696 if (gmx_mm256_any_lt(rsq00,rcutoff2))
697 {
698
699 r00 = _mm256_mul_pd(rsq00,rinv00);
700 r00 = _mm256_andnot_pd(dummy_mask,r00);
701
702 /* Compute parameters for interactions between i and j atoms */
703 gmx_mm256_load_4pair_swizzle_pd(vdwioffsetptr0+vdwjidx0A,
704 vdwioffsetptr0+vdwjidx0B,
705 vdwioffsetptr0+vdwjidx0C,
706 vdwioffsetptr0+vdwjidx0D,
707 &c6_00,&c12_00);
708
709 c6grid_00 = gmx_mm256_load_4real_swizzle_pd(vdwgridioffsetptr0+vdwjidx0A,
710 vdwgridioffsetptr0+vdwjidx0B,
711 vdwgridioffsetptr0+vdwjidx0C,
712 vdwgridioffsetptr0+vdwjidx0D);
713
714 /* Analytical LJ-PME */
715 rinvsix = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
716 ewcljrsq = _mm256_mul_pd(ewclj2,rsq00);
717 ewclj6 = _mm256_mul_pd(ewclj2,_mm256_mul_pd(ewclj2,ewclj2));
718 exponent = avx256_exp_d(ewcljrsq);
719 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
720 poly = _mm256_mul_pd(exponent,_mm256_add_pd(_mm256_sub_pd(one,ewcljrsq),_mm256_mul_pd(_mm256_mul_pd(ewcljrsq,ewcljrsq),one_half)));
721 /* f6A = 6 * C6grid * (1 - poly) */
722 f6A = _mm256_mul_pd(c6grid_00,_mm256_sub_pd(one,poly));
723 /* f6B = C6grid * exponent * beta^6 */
724 f6B = _mm256_mul_pd(_mm256_mul_pd(c6grid_00,one_sixth),_mm256_mul_pd(exponent,ewclj6));
725 /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
726 fvdw = _mm256_mul_pd(_mm256_add_pd(_mm256_mul_pd(_mm256_sub_pd(_mm256_mul_pd(c12_00,rinvsix),_mm256_sub_pd(c6_00,f6A)),rinvsix),f6B),rinvsq00);
727
728 cutoff_mask = _mm256_cmp_pd(rsq00,rcutoff2,_CMP_LT_OQ);
729
730 fscal = fvdw;
731
732 fscal = _mm256_and_pd(fscal,cutoff_mask);
733
734 fscal = _mm256_andnot_pd(dummy_mask,fscal);
735
736 /* Calculate temporary vectorial force */
737 tx = _mm256_mul_pd(fscal,dx00);
738 ty = _mm256_mul_pd(fscal,dy00);
739 tz = _mm256_mul_pd(fscal,dz00);
740
741 /* Update vectorial force */
742 fix0 = _mm256_add_pd(fix0,tx);
743 fiy0 = _mm256_add_pd(fiy0,ty);
744 fiz0 = _mm256_add_pd(fiz0,tz);
745
746 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
747 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
748 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
749 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
750 gmx_mm256_decrement_1rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,tx,ty,tz);
751
752 }
753
754 /* Inner loop uses 50 flops */
755 }
756
757 /* End of innermost loop */
758
759 gmx_mm256_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
760 f+i_coord_offset,fshift+i_shift_offset);
761
762 /* Increment number of inner iterations */
763 inneriter += j_index_end - j_index_start;
764
765 /* Outer loop uses 6 flops */
766 }
767
768 /* Increment number of outer iterations */
769 outeriter += nri;
770
771 /* Update outer/inner flops */
772
773 inc_nrnb(nrnb,eNR_NBKERNEL_VDW_F,outeriter*6 + inneriter*50);
774 }
The ultimate molecular dynamics simulation package