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_256_single 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_avx_256_single.h"
48 #include "kernelutil_x86_avx_256_single.h"
51 * Gromacs nonbonded kernel: nb_kernel_ElecNone_VdwLJEw_GeomP1P1_VF_avx_256_single
52 * Electrostatics interaction: None
53 * VdW interaction: LJEwald
54 * Geometry: Particle-Particle
55 * Calculate force/pot: PotentialAndForce
58 nb_kernel_ElecNone_VdwLJEw_GeomP1P1_VF_avx_256_single
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,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight 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;
74 int jnrA,jnrB,jnrC,jnrD;
75 int jnrE,jnrF,jnrG,jnrH;
76 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
77 int jnrlistE,jnrlistF,jnrlistG,jnrlistH;
78 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
79 int j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
80 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
82 real *shiftvec,*fshift,*x,*f;
83 real *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
85 __m256 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
86 real * vdwioffsetptr0;
87 real * vdwgridioffsetptr0;
88 __m256 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
89 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
90 __m256 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
91 __m256 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
93 __m256 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
96 __m256 one_sixth = _mm256_set1_ps(1.0/6.0);
97 __m256 one_twelfth = _mm256_set1_ps(1.0/12.0);
100 __m256 ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
101 __m256 one_half = _mm256_set1_ps(0.5);
102 __m256 minus_one = _mm256_set1_ps(-1.0);
103 __m256 dummy_mask,cutoff_mask;
104 __m256 signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
105 __m256 one = _mm256_set1_ps(1.0);
106 __m256 two = _mm256_set1_ps(2.0);
112 jindex = nlist->jindex;
114 shiftidx = nlist->shift;
116 shiftvec = fr->shift_vec[0];
117 fshift = fr->fshift[0];
118 nvdwtype = fr->ntype;
120 vdwtype = mdatoms->typeA;
121 vdwgridparam = fr->ljpme_c6grid;
122 sh_lj_ewald = _mm256_set1_ps(fr->ic->sh_lj_ewald);
123 ewclj = _mm256_set1_ps(fr->ewaldcoeff_lj);
124 ewclj2 = _mm256_mul_ps(minus_one,_mm256_mul_ps(ewclj,ewclj));
126 /* Avoid stupid compiler warnings */
127 jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
140 for(iidx=0;iidx<4*DIM;iidx++)
145 /* Start outer loop over neighborlists */
146 for(iidx=0; iidx<nri; iidx++)
148 /* Load shift vector for this list */
149 i_shift_offset = DIM*shiftidx[iidx];
151 /* Load limits for loop over neighbors */
152 j_index_start = jindex[iidx];
153 j_index_end = jindex[iidx+1];
155 /* Get outer coordinate index */
157 i_coord_offset = DIM*inr;
159 /* Load i particle coords and add shift vector */
160 gmx_mm256_load_shift_and_1rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
162 fix0 = _mm256_setzero_ps();
163 fiy0 = _mm256_setzero_ps();
164 fiz0 = _mm256_setzero_ps();
166 /* Load parameters for i particles */
167 vdwioffsetptr0 = vdwparam+2*nvdwtype*vdwtype[inr+0];
168 vdwgridioffsetptr0 = vdwgridparam+2*nvdwtype*vdwtype[inr+0];
170 /* Reset potential sums */
171 vvdwsum = _mm256_setzero_ps();
173 /* Start inner kernel loop */
174 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
177 /* Get j neighbor index, and coordinate index */
186 j_coord_offsetA = DIM*jnrA;
187 j_coord_offsetB = DIM*jnrB;
188 j_coord_offsetC = DIM*jnrC;
189 j_coord_offsetD = DIM*jnrD;
190 j_coord_offsetE = DIM*jnrE;
191 j_coord_offsetF = DIM*jnrF;
192 j_coord_offsetG = DIM*jnrG;
193 j_coord_offsetH = DIM*jnrH;
195 /* load j atom coordinates */
196 gmx_mm256_load_1rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
197 x+j_coord_offsetC,x+j_coord_offsetD,
198 x+j_coord_offsetE,x+j_coord_offsetF,
199 x+j_coord_offsetG,x+j_coord_offsetH,
202 /* Calculate displacement vector */
203 dx00 = _mm256_sub_ps(ix0,jx0);
204 dy00 = _mm256_sub_ps(iy0,jy0);
205 dz00 = _mm256_sub_ps(iz0,jz0);
207 /* Calculate squared distance and things based on it */
208 rsq00 = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
210 rinv00 = gmx_mm256_invsqrt_ps(rsq00);
212 rinvsq00 = _mm256_mul_ps(rinv00,rinv00);
214 /* Load parameters for j particles */
215 vdwjidx0A = 2*vdwtype[jnrA+0];
216 vdwjidx0B = 2*vdwtype[jnrB+0];
217 vdwjidx0C = 2*vdwtype[jnrC+0];
218 vdwjidx0D = 2*vdwtype[jnrD+0];
219 vdwjidx0E = 2*vdwtype[jnrE+0];
220 vdwjidx0F = 2*vdwtype[jnrF+0];
221 vdwjidx0G = 2*vdwtype[jnrG+0];
222 vdwjidx0H = 2*vdwtype[jnrH+0];
224 /**************************
225 * CALCULATE INTERACTIONS *
226 **************************/
228 r00 = _mm256_mul_ps(rsq00,rinv00);
230 /* Compute parameters for interactions between i and j atoms */
231 gmx_mm256_load_8pair_swizzle_ps(vdwioffsetptr0+vdwjidx0A,
232 vdwioffsetptr0+vdwjidx0B,
233 vdwioffsetptr0+vdwjidx0C,
234 vdwioffsetptr0+vdwjidx0D,
235 vdwioffsetptr0+vdwjidx0E,
236 vdwioffsetptr0+vdwjidx0F,
237 vdwioffsetptr0+vdwjidx0G,
238 vdwioffsetptr0+vdwjidx0H,
241 c6grid_00 = gmx_mm256_load_8real_swizzle_ps(vdwgridioffsetptr0+vdwjidx0A,
242 vdwgridioffsetptr0+vdwjidx0B,
243 vdwgridioffsetptr0+vdwjidx0C,
244 vdwgridioffsetptr0+vdwjidx0D,
245 vdwgridioffsetptr0+vdwjidx0E,
246 vdwgridioffsetptr0+vdwjidx0F,
247 vdwgridioffsetptr0+vdwjidx0G,
248 vdwgridioffsetptr0+vdwjidx0H);
250 /* Analytical LJ-PME */
251 rinvsix = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
252 ewcljrsq = _mm256_mul_ps(ewclj2,rsq00);
253 ewclj6 = _mm256_mul_ps(ewclj2,_mm256_mul_ps(ewclj2,ewclj2));
254 exponent = gmx_simd_exp_r(ewcljrsq);
255 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
256 poly = _mm256_mul_ps(exponent,_mm256_add_ps(_mm256_sub_ps(one,ewcljrsq),_mm256_mul_ps(_mm256_mul_ps(ewcljrsq,ewcljrsq),one_half)));
257 /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
258 vvdw6 = _mm256_mul_ps(_mm256_sub_ps(c6_00,_mm256_mul_ps(c6grid_00,_mm256_sub_ps(one,poly))),rinvsix);
259 vvdw12 = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
260 vvdw = _mm256_sub_ps(_mm256_mul_ps(vvdw12,one_twelfth),_mm256_mul_ps(vvdw6,one_sixth));
261 /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
262 fvdw = _mm256_mul_ps(_mm256_sub_ps(vvdw12,_mm256_sub_ps(vvdw6,_mm256_mul_ps(_mm256_mul_ps(c6grid_00,one_sixth),_mm256_mul_ps(exponent,ewclj6)))),rinvsq00);
264 /* Update potential sum for this i atom from the interaction with this j atom. */
265 vvdwsum = _mm256_add_ps(vvdwsum,vvdw);
269 /* Calculate temporary vectorial force */
270 tx = _mm256_mul_ps(fscal,dx00);
271 ty = _mm256_mul_ps(fscal,dy00);
272 tz = _mm256_mul_ps(fscal,dz00);
274 /* Update vectorial force */
275 fix0 = _mm256_add_ps(fix0,tx);
276 fiy0 = _mm256_add_ps(fiy0,ty);
277 fiz0 = _mm256_add_ps(fiz0,tz);
279 fjptrA = f+j_coord_offsetA;
280 fjptrB = f+j_coord_offsetB;
281 fjptrC = f+j_coord_offsetC;
282 fjptrD = f+j_coord_offsetD;
283 fjptrE = f+j_coord_offsetE;
284 fjptrF = f+j_coord_offsetF;
285 fjptrG = f+j_coord_offsetG;
286 fjptrH = f+j_coord_offsetH;
287 gmx_mm256_decrement_1rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,tx,ty,tz);
289 /* Inner loop uses 51 flops */
295 /* Get j neighbor index, and coordinate index */
296 jnrlistA = jjnr[jidx];
297 jnrlistB = jjnr[jidx+1];
298 jnrlistC = jjnr[jidx+2];
299 jnrlistD = jjnr[jidx+3];
300 jnrlistE = jjnr[jidx+4];
301 jnrlistF = jjnr[jidx+5];
302 jnrlistG = jjnr[jidx+6];
303 jnrlistH = jjnr[jidx+7];
304 /* Sign of each element will be negative for non-real atoms.
305 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
306 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
308 dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
309 gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
311 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
312 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
313 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
314 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
315 jnrE = (jnrlistE>=0) ? jnrlistE : 0;
316 jnrF = (jnrlistF>=0) ? jnrlistF : 0;
317 jnrG = (jnrlistG>=0) ? jnrlistG : 0;
318 jnrH = (jnrlistH>=0) ? jnrlistH : 0;
319 j_coord_offsetA = DIM*jnrA;
320 j_coord_offsetB = DIM*jnrB;
321 j_coord_offsetC = DIM*jnrC;
322 j_coord_offsetD = DIM*jnrD;
323 j_coord_offsetE = DIM*jnrE;
324 j_coord_offsetF = DIM*jnrF;
325 j_coord_offsetG = DIM*jnrG;
326 j_coord_offsetH = DIM*jnrH;
328 /* load j atom coordinates */
329 gmx_mm256_load_1rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
330 x+j_coord_offsetC,x+j_coord_offsetD,
331 x+j_coord_offsetE,x+j_coord_offsetF,
332 x+j_coord_offsetG,x+j_coord_offsetH,
335 /* Calculate displacement vector */
336 dx00 = _mm256_sub_ps(ix0,jx0);
337 dy00 = _mm256_sub_ps(iy0,jy0);
338 dz00 = _mm256_sub_ps(iz0,jz0);
340 /* Calculate squared distance and things based on it */
341 rsq00 = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
343 rinv00 = gmx_mm256_invsqrt_ps(rsq00);
345 rinvsq00 = _mm256_mul_ps(rinv00,rinv00);
347 /* Load parameters for j particles */
348 vdwjidx0A = 2*vdwtype[jnrA+0];
349 vdwjidx0B = 2*vdwtype[jnrB+0];
350 vdwjidx0C = 2*vdwtype[jnrC+0];
351 vdwjidx0D = 2*vdwtype[jnrD+0];
352 vdwjidx0E = 2*vdwtype[jnrE+0];
353 vdwjidx0F = 2*vdwtype[jnrF+0];
354 vdwjidx0G = 2*vdwtype[jnrG+0];
355 vdwjidx0H = 2*vdwtype[jnrH+0];
357 /**************************
358 * CALCULATE INTERACTIONS *
359 **************************/
361 r00 = _mm256_mul_ps(rsq00,rinv00);
362 r00 = _mm256_andnot_ps(dummy_mask,r00);
364 /* Compute parameters for interactions between i and j atoms */
365 gmx_mm256_load_8pair_swizzle_ps(vdwioffsetptr0+vdwjidx0A,
366 vdwioffsetptr0+vdwjidx0B,
367 vdwioffsetptr0+vdwjidx0C,
368 vdwioffsetptr0+vdwjidx0D,
369 vdwioffsetptr0+vdwjidx0E,
370 vdwioffsetptr0+vdwjidx0F,
371 vdwioffsetptr0+vdwjidx0G,
372 vdwioffsetptr0+vdwjidx0H,
375 c6grid_00 = gmx_mm256_load_8real_swizzle_ps(vdwgridioffsetptr0+vdwjidx0A,
376 vdwgridioffsetptr0+vdwjidx0B,
377 vdwgridioffsetptr0+vdwjidx0C,
378 vdwgridioffsetptr0+vdwjidx0D,
379 vdwgridioffsetptr0+vdwjidx0E,
380 vdwgridioffsetptr0+vdwjidx0F,
381 vdwgridioffsetptr0+vdwjidx0G,
382 vdwgridioffsetptr0+vdwjidx0H);
384 /* Analytical LJ-PME */
385 rinvsix = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
386 ewcljrsq = _mm256_mul_ps(ewclj2,rsq00);
387 ewclj6 = _mm256_mul_ps(ewclj2,_mm256_mul_ps(ewclj2,ewclj2));
388 exponent = gmx_simd_exp_r(ewcljrsq);
389 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
390 poly = _mm256_mul_ps(exponent,_mm256_add_ps(_mm256_sub_ps(one,ewcljrsq),_mm256_mul_ps(_mm256_mul_ps(ewcljrsq,ewcljrsq),one_half)));
391 /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
392 vvdw6 = _mm256_mul_ps(_mm256_sub_ps(c6_00,_mm256_mul_ps(c6grid_00,_mm256_sub_ps(one,poly))),rinvsix);
393 vvdw12 = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
394 vvdw = _mm256_sub_ps(_mm256_mul_ps(vvdw12,one_twelfth),_mm256_mul_ps(vvdw6,one_sixth));
395 /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
396 fvdw = _mm256_mul_ps(_mm256_sub_ps(vvdw12,_mm256_sub_ps(vvdw6,_mm256_mul_ps(_mm256_mul_ps(c6grid_00,one_sixth),_mm256_mul_ps(exponent,ewclj6)))),rinvsq00);
398 /* Update potential sum for this i atom from the interaction with this j atom. */
399 vvdw = _mm256_andnot_ps(dummy_mask,vvdw);
400 vvdwsum = _mm256_add_ps(vvdwsum,vvdw);
404 fscal = _mm256_andnot_ps(dummy_mask,fscal);
406 /* Calculate temporary vectorial force */
407 tx = _mm256_mul_ps(fscal,dx00);
408 ty = _mm256_mul_ps(fscal,dy00);
409 tz = _mm256_mul_ps(fscal,dz00);
411 /* Update vectorial force */
412 fix0 = _mm256_add_ps(fix0,tx);
413 fiy0 = _mm256_add_ps(fiy0,ty);
414 fiz0 = _mm256_add_ps(fiz0,tz);
416 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
417 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
418 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
419 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
420 fjptrE = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
421 fjptrF = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
422 fjptrG = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
423 fjptrH = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
424 gmx_mm256_decrement_1rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,tx,ty,tz);
426 /* Inner loop uses 52 flops */
429 /* End of innermost loop */
431 gmx_mm256_update_iforce_1atom_swizzle_ps(fix0,fiy0,fiz0,
432 f+i_coord_offset,fshift+i_shift_offset);
435 /* Update potential energies */
436 gmx_mm256_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
438 /* Increment number of inner iterations */
439 inneriter += j_index_end - j_index_start;
441 /* Outer loop uses 7 flops */
444 /* Increment number of outer iterations */
447 /* Update outer/inner flops */
449 inc_nrnb(nrnb,eNR_NBKERNEL_VDW_VF,outeriter*7 + inneriter*52);
452 * Gromacs nonbonded kernel: nb_kernel_ElecNone_VdwLJEw_GeomP1P1_F_avx_256_single
453 * Electrostatics interaction: None
454 * VdW interaction: LJEwald
455 * Geometry: Particle-Particle
456 * Calculate force/pot: Force
459 nb_kernel_ElecNone_VdwLJEw_GeomP1P1_F_avx_256_single
460 (t_nblist * gmx_restrict nlist,
461 rvec * gmx_restrict xx,
462 rvec * gmx_restrict ff,
463 t_forcerec * gmx_restrict fr,
464 t_mdatoms * gmx_restrict mdatoms,
465 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
466 t_nrnb * gmx_restrict nrnb)
468 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
469 * just 0 for non-waters.
470 * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
471 * jnr indices corresponding to data put in the four positions in the SIMD register.
473 int i_shift_offset,i_coord_offset,outeriter,inneriter;
474 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
475 int jnrA,jnrB,jnrC,jnrD;
476 int jnrE,jnrF,jnrG,jnrH;
477 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
478 int jnrlistE,jnrlistF,jnrlistG,jnrlistH;
479 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
480 int j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
481 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
483 real *shiftvec,*fshift,*x,*f;
484 real *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
486 __m256 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
487 real * vdwioffsetptr0;
488 real * vdwgridioffsetptr0;
489 __m256 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
490 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
491 __m256 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
492 __m256 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
494 __m256 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
497 __m256 one_sixth = _mm256_set1_ps(1.0/6.0);
498 __m256 one_twelfth = _mm256_set1_ps(1.0/12.0);
501 __m256 ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
502 __m256 one_half = _mm256_set1_ps(0.5);
503 __m256 minus_one = _mm256_set1_ps(-1.0);
504 __m256 dummy_mask,cutoff_mask;
505 __m256 signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
506 __m256 one = _mm256_set1_ps(1.0);
507 __m256 two = _mm256_set1_ps(2.0);
513 jindex = nlist->jindex;
515 shiftidx = nlist->shift;
517 shiftvec = fr->shift_vec[0];
518 fshift = fr->fshift[0];
519 nvdwtype = fr->ntype;
521 vdwtype = mdatoms->typeA;
522 vdwgridparam = fr->ljpme_c6grid;
523 sh_lj_ewald = _mm256_set1_ps(fr->ic->sh_lj_ewald);
524 ewclj = _mm256_set1_ps(fr->ewaldcoeff_lj);
525 ewclj2 = _mm256_mul_ps(minus_one,_mm256_mul_ps(ewclj,ewclj));
527 /* Avoid stupid compiler warnings */
528 jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
541 for(iidx=0;iidx<4*DIM;iidx++)
546 /* Start outer loop over neighborlists */
547 for(iidx=0; iidx<nri; iidx++)
549 /* Load shift vector for this list */
550 i_shift_offset = DIM*shiftidx[iidx];
552 /* Load limits for loop over neighbors */
553 j_index_start = jindex[iidx];
554 j_index_end = jindex[iidx+1];
556 /* Get outer coordinate index */
558 i_coord_offset = DIM*inr;
560 /* Load i particle coords and add shift vector */
561 gmx_mm256_load_shift_and_1rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
563 fix0 = _mm256_setzero_ps();
564 fiy0 = _mm256_setzero_ps();
565 fiz0 = _mm256_setzero_ps();
567 /* Load parameters for i particles */
568 vdwioffsetptr0 = vdwparam+2*nvdwtype*vdwtype[inr+0];
569 vdwgridioffsetptr0 = vdwgridparam+2*nvdwtype*vdwtype[inr+0];
571 /* Start inner kernel loop */
572 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
575 /* Get j neighbor index, and coordinate index */
584 j_coord_offsetA = DIM*jnrA;
585 j_coord_offsetB = DIM*jnrB;
586 j_coord_offsetC = DIM*jnrC;
587 j_coord_offsetD = DIM*jnrD;
588 j_coord_offsetE = DIM*jnrE;
589 j_coord_offsetF = DIM*jnrF;
590 j_coord_offsetG = DIM*jnrG;
591 j_coord_offsetH = DIM*jnrH;
593 /* load j atom coordinates */
594 gmx_mm256_load_1rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
595 x+j_coord_offsetC,x+j_coord_offsetD,
596 x+j_coord_offsetE,x+j_coord_offsetF,
597 x+j_coord_offsetG,x+j_coord_offsetH,
600 /* Calculate displacement vector */
601 dx00 = _mm256_sub_ps(ix0,jx0);
602 dy00 = _mm256_sub_ps(iy0,jy0);
603 dz00 = _mm256_sub_ps(iz0,jz0);
605 /* Calculate squared distance and things based on it */
606 rsq00 = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
608 rinv00 = gmx_mm256_invsqrt_ps(rsq00);
610 rinvsq00 = _mm256_mul_ps(rinv00,rinv00);
612 /* Load parameters for j particles */
613 vdwjidx0A = 2*vdwtype[jnrA+0];
614 vdwjidx0B = 2*vdwtype[jnrB+0];
615 vdwjidx0C = 2*vdwtype[jnrC+0];
616 vdwjidx0D = 2*vdwtype[jnrD+0];
617 vdwjidx0E = 2*vdwtype[jnrE+0];
618 vdwjidx0F = 2*vdwtype[jnrF+0];
619 vdwjidx0G = 2*vdwtype[jnrG+0];
620 vdwjidx0H = 2*vdwtype[jnrH+0];
622 /**************************
623 * CALCULATE INTERACTIONS *
624 **************************/
626 r00 = _mm256_mul_ps(rsq00,rinv00);
628 /* Compute parameters for interactions between i and j atoms */
629 gmx_mm256_load_8pair_swizzle_ps(vdwioffsetptr0+vdwjidx0A,
630 vdwioffsetptr0+vdwjidx0B,
631 vdwioffsetptr0+vdwjidx0C,
632 vdwioffsetptr0+vdwjidx0D,
633 vdwioffsetptr0+vdwjidx0E,
634 vdwioffsetptr0+vdwjidx0F,
635 vdwioffsetptr0+vdwjidx0G,
636 vdwioffsetptr0+vdwjidx0H,
639 c6grid_00 = gmx_mm256_load_8real_swizzle_ps(vdwgridioffsetptr0+vdwjidx0A,
640 vdwgridioffsetptr0+vdwjidx0B,
641 vdwgridioffsetptr0+vdwjidx0C,
642 vdwgridioffsetptr0+vdwjidx0D,
643 vdwgridioffsetptr0+vdwjidx0E,
644 vdwgridioffsetptr0+vdwjidx0F,
645 vdwgridioffsetptr0+vdwjidx0G,
646 vdwgridioffsetptr0+vdwjidx0H);
648 /* Analytical LJ-PME */
649 rinvsix = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
650 ewcljrsq = _mm256_mul_ps(ewclj2,rsq00);
651 ewclj6 = _mm256_mul_ps(ewclj2,_mm256_mul_ps(ewclj2,ewclj2));
652 exponent = gmx_simd_exp_r(ewcljrsq);
653 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
654 poly = _mm256_mul_ps(exponent,_mm256_add_ps(_mm256_sub_ps(one,ewcljrsq),_mm256_mul_ps(_mm256_mul_ps(ewcljrsq,ewcljrsq),one_half)));
655 /* f6A = 6 * C6grid * (1 - poly) */
656 f6A = _mm256_mul_ps(c6grid_00,_mm256_sub_ps(one,poly));
657 /* f6B = C6grid * exponent * beta^6 */
658 f6B = _mm256_mul_ps(_mm256_mul_ps(c6grid_00,one_sixth),_mm256_mul_ps(exponent,ewclj6));
659 /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
660 fvdw = _mm256_mul_ps(_mm256_add_ps(_mm256_mul_ps(_mm256_sub_ps(_mm256_mul_ps(c12_00,rinvsix),_mm256_sub_ps(c6_00,f6A)),rinvsix),f6B),rinvsq00);
664 /* Calculate temporary vectorial force */
665 tx = _mm256_mul_ps(fscal,dx00);
666 ty = _mm256_mul_ps(fscal,dy00);
667 tz = _mm256_mul_ps(fscal,dz00);
669 /* Update vectorial force */
670 fix0 = _mm256_add_ps(fix0,tx);
671 fiy0 = _mm256_add_ps(fiy0,ty);
672 fiz0 = _mm256_add_ps(fiz0,tz);
674 fjptrA = f+j_coord_offsetA;
675 fjptrB = f+j_coord_offsetB;
676 fjptrC = f+j_coord_offsetC;
677 fjptrD = f+j_coord_offsetD;
678 fjptrE = f+j_coord_offsetE;
679 fjptrF = f+j_coord_offsetF;
680 fjptrG = f+j_coord_offsetG;
681 fjptrH = f+j_coord_offsetH;
682 gmx_mm256_decrement_1rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,tx,ty,tz);
684 /* Inner loop uses 46 flops */
690 /* Get j neighbor index, and coordinate index */
691 jnrlistA = jjnr[jidx];
692 jnrlistB = jjnr[jidx+1];
693 jnrlistC = jjnr[jidx+2];
694 jnrlistD = jjnr[jidx+3];
695 jnrlistE = jjnr[jidx+4];
696 jnrlistF = jjnr[jidx+5];
697 jnrlistG = jjnr[jidx+6];
698 jnrlistH = jjnr[jidx+7];
699 /* Sign of each element will be negative for non-real atoms.
700 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
701 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
703 dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
704 gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
706 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
707 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
708 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
709 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
710 jnrE = (jnrlistE>=0) ? jnrlistE : 0;
711 jnrF = (jnrlistF>=0) ? jnrlistF : 0;
712 jnrG = (jnrlistG>=0) ? jnrlistG : 0;
713 jnrH = (jnrlistH>=0) ? jnrlistH : 0;
714 j_coord_offsetA = DIM*jnrA;
715 j_coord_offsetB = DIM*jnrB;
716 j_coord_offsetC = DIM*jnrC;
717 j_coord_offsetD = DIM*jnrD;
718 j_coord_offsetE = DIM*jnrE;
719 j_coord_offsetF = DIM*jnrF;
720 j_coord_offsetG = DIM*jnrG;
721 j_coord_offsetH = DIM*jnrH;
723 /* load j atom coordinates */
724 gmx_mm256_load_1rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
725 x+j_coord_offsetC,x+j_coord_offsetD,
726 x+j_coord_offsetE,x+j_coord_offsetF,
727 x+j_coord_offsetG,x+j_coord_offsetH,
730 /* Calculate displacement vector */
731 dx00 = _mm256_sub_ps(ix0,jx0);
732 dy00 = _mm256_sub_ps(iy0,jy0);
733 dz00 = _mm256_sub_ps(iz0,jz0);
735 /* Calculate squared distance and things based on it */
736 rsq00 = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
738 rinv00 = gmx_mm256_invsqrt_ps(rsq00);
740 rinvsq00 = _mm256_mul_ps(rinv00,rinv00);
742 /* Load parameters for j particles */
743 vdwjidx0A = 2*vdwtype[jnrA+0];
744 vdwjidx0B = 2*vdwtype[jnrB+0];
745 vdwjidx0C = 2*vdwtype[jnrC+0];
746 vdwjidx0D = 2*vdwtype[jnrD+0];
747 vdwjidx0E = 2*vdwtype[jnrE+0];
748 vdwjidx0F = 2*vdwtype[jnrF+0];
749 vdwjidx0G = 2*vdwtype[jnrG+0];
750 vdwjidx0H = 2*vdwtype[jnrH+0];
752 /**************************
753 * CALCULATE INTERACTIONS *
754 **************************/
756 r00 = _mm256_mul_ps(rsq00,rinv00);
757 r00 = _mm256_andnot_ps(dummy_mask,r00);
759 /* Compute parameters for interactions between i and j atoms */
760 gmx_mm256_load_8pair_swizzle_ps(vdwioffsetptr0+vdwjidx0A,
761 vdwioffsetptr0+vdwjidx0B,
762 vdwioffsetptr0+vdwjidx0C,
763 vdwioffsetptr0+vdwjidx0D,
764 vdwioffsetptr0+vdwjidx0E,
765 vdwioffsetptr0+vdwjidx0F,
766 vdwioffsetptr0+vdwjidx0G,
767 vdwioffsetptr0+vdwjidx0H,
770 c6grid_00 = gmx_mm256_load_8real_swizzle_ps(vdwgridioffsetptr0+vdwjidx0A,
771 vdwgridioffsetptr0+vdwjidx0B,
772 vdwgridioffsetptr0+vdwjidx0C,
773 vdwgridioffsetptr0+vdwjidx0D,
774 vdwgridioffsetptr0+vdwjidx0E,
775 vdwgridioffsetptr0+vdwjidx0F,
776 vdwgridioffsetptr0+vdwjidx0G,
777 vdwgridioffsetptr0+vdwjidx0H);
779 /* Analytical LJ-PME */
780 rinvsix = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
781 ewcljrsq = _mm256_mul_ps(ewclj2,rsq00);
782 ewclj6 = _mm256_mul_ps(ewclj2,_mm256_mul_ps(ewclj2,ewclj2));
783 exponent = gmx_simd_exp_r(ewcljrsq);
784 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
785 poly = _mm256_mul_ps(exponent,_mm256_add_ps(_mm256_sub_ps(one,ewcljrsq),_mm256_mul_ps(_mm256_mul_ps(ewcljrsq,ewcljrsq),one_half)));
786 /* f6A = 6 * C6grid * (1 - poly) */
787 f6A = _mm256_mul_ps(c6grid_00,_mm256_sub_ps(one,poly));
788 /* f6B = C6grid * exponent * beta^6 */
789 f6B = _mm256_mul_ps(_mm256_mul_ps(c6grid_00,one_sixth),_mm256_mul_ps(exponent,ewclj6));
790 /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
791 fvdw = _mm256_mul_ps(_mm256_add_ps(_mm256_mul_ps(_mm256_sub_ps(_mm256_mul_ps(c12_00,rinvsix),_mm256_sub_ps(c6_00,f6A)),rinvsix),f6B),rinvsq00);
795 fscal = _mm256_andnot_ps(dummy_mask,fscal);
797 /* Calculate temporary vectorial force */
798 tx = _mm256_mul_ps(fscal,dx00);
799 ty = _mm256_mul_ps(fscal,dy00);
800 tz = _mm256_mul_ps(fscal,dz00);
802 /* Update vectorial force */
803 fix0 = _mm256_add_ps(fix0,tx);
804 fiy0 = _mm256_add_ps(fiy0,ty);
805 fiz0 = _mm256_add_ps(fiz0,tz);
807 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
808 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
809 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
810 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
811 fjptrE = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
812 fjptrF = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
813 fjptrG = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
814 fjptrH = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
815 gmx_mm256_decrement_1rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,tx,ty,tz);
817 /* Inner loop uses 47 flops */
820 /* End of innermost loop */
822 gmx_mm256_update_iforce_1atom_swizzle_ps(fix0,fiy0,fiz0,
823 f+i_coord_offset,fshift+i_shift_offset);
825 /* Increment number of inner iterations */
826 inneriter += j_index_end - j_index_start;
828 /* Outer loop uses 6 flops */
831 /* Increment number of outer iterations */
834 /* Update outer/inner flops */
836 inc_nrnb(nrnb,eNR_NBKERNEL_VDW_F,outeriter*6 + inneriter*47);