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_double kernel generator.
42 #include "../nb_kernel.h"
43 #include "types/simple.h"
44 #include "gromacs/math/vec.h"
47 #include "gromacs/simd/math_x86_avx_256_double.h"
48 #include "kernelutil_x86_avx_256_double.h"
51 * Gromacs nonbonded kernel: nb_kernel_ElecEwSw_VdwLJSw_GeomW3P1_VF_avx_256_double
52 * Electrostatics interaction: Ewald
53 * VdW interaction: LennardJones
54 * Geometry: Water3-Particle
55 * Calculate force/pot: PotentialAndForce
58 nb_kernel_ElecEwSw_VdwLJSw_GeomW3P1_VF_avx_256_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,C,D refer to j loop unrolling done with AVX, e.g. for the four 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 jnrlistA,jnrlistB,jnrlistC,jnrlistD;
76 int jnrlistE,jnrlistF,jnrlistG,jnrlistH;
77 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
78 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
80 real *shiftvec,*fshift,*x,*f;
81 real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
83 __m256d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
84 real * vdwioffsetptr0;
85 __m256d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
86 real * vdwioffsetptr1;
87 __m256d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
88 real * vdwioffsetptr2;
89 __m256d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
90 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
91 __m256d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
92 __m256d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
93 __m256d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
94 __m256d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
95 __m256d velec,felec,velecsum,facel,crf,krf,krf2;
98 __m256d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
101 __m256d one_sixth = _mm256_set1_pd(1.0/6.0);
102 __m256d one_twelfth = _mm256_set1_pd(1.0/12.0);
104 __m256d ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
105 __m256d beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
107 __m256d rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
108 real rswitch_scalar,d_scalar;
109 __m256d dummy_mask,cutoff_mask;
110 __m128 tmpmask0,tmpmask1;
111 __m256d signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
112 __m256d one = _mm256_set1_pd(1.0);
113 __m256d two = _mm256_set1_pd(2.0);
119 jindex = nlist->jindex;
121 shiftidx = nlist->shift;
123 shiftvec = fr->shift_vec[0];
124 fshift = fr->fshift[0];
125 facel = _mm256_set1_pd(fr->epsfac);
126 charge = mdatoms->chargeA;
127 nvdwtype = fr->ntype;
129 vdwtype = mdatoms->typeA;
131 sh_ewald = _mm256_set1_pd(fr->ic->sh_ewald);
132 beta = _mm256_set1_pd(fr->ic->ewaldcoeff_q);
133 beta2 = _mm256_mul_pd(beta,beta);
134 beta3 = _mm256_mul_pd(beta,beta2);
136 ewtab = fr->ic->tabq_coul_FDV0;
137 ewtabscale = _mm256_set1_pd(fr->ic->tabq_scale);
138 ewtabhalfspace = _mm256_set1_pd(0.5/fr->ic->tabq_scale);
140 /* Setup water-specific parameters */
141 inr = nlist->iinr[0];
142 iq0 = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+0]));
143 iq1 = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
144 iq2 = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
145 vdwioffsetptr0 = vdwparam+2*nvdwtype*vdwtype[inr+0];
147 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
148 rcutoff_scalar = fr->rcoulomb;
149 rcutoff = _mm256_set1_pd(rcutoff_scalar);
150 rcutoff2 = _mm256_mul_pd(rcutoff,rcutoff);
152 rswitch_scalar = fr->rcoulomb_switch;
153 rswitch = _mm256_set1_pd(rswitch_scalar);
154 /* Setup switch parameters */
155 d_scalar = rcutoff_scalar-rswitch_scalar;
156 d = _mm256_set1_pd(d_scalar);
157 swV3 = _mm256_set1_pd(-10.0/(d_scalar*d_scalar*d_scalar));
158 swV4 = _mm256_set1_pd( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
159 swV5 = _mm256_set1_pd( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
160 swF2 = _mm256_set1_pd(-30.0/(d_scalar*d_scalar*d_scalar));
161 swF3 = _mm256_set1_pd( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
162 swF4 = _mm256_set1_pd(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
164 /* Avoid stupid compiler warnings */
165 jnrA = jnrB = jnrC = jnrD = 0;
174 for(iidx=0;iidx<4*DIM;iidx++)
179 /* Start outer loop over neighborlists */
180 for(iidx=0; iidx<nri; iidx++)
182 /* Load shift vector for this list */
183 i_shift_offset = DIM*shiftidx[iidx];
185 /* Load limits for loop over neighbors */
186 j_index_start = jindex[iidx];
187 j_index_end = jindex[iidx+1];
189 /* Get outer coordinate index */
191 i_coord_offset = DIM*inr;
193 /* Load i particle coords and add shift vector */
194 gmx_mm256_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
195 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
197 fix0 = _mm256_setzero_pd();
198 fiy0 = _mm256_setzero_pd();
199 fiz0 = _mm256_setzero_pd();
200 fix1 = _mm256_setzero_pd();
201 fiy1 = _mm256_setzero_pd();
202 fiz1 = _mm256_setzero_pd();
203 fix2 = _mm256_setzero_pd();
204 fiy2 = _mm256_setzero_pd();
205 fiz2 = _mm256_setzero_pd();
207 /* Reset potential sums */
208 velecsum = _mm256_setzero_pd();
209 vvdwsum = _mm256_setzero_pd();
211 /* Start inner kernel loop */
212 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
215 /* Get j neighbor index, and coordinate index */
220 j_coord_offsetA = DIM*jnrA;
221 j_coord_offsetB = DIM*jnrB;
222 j_coord_offsetC = DIM*jnrC;
223 j_coord_offsetD = DIM*jnrD;
225 /* load j atom coordinates */
226 gmx_mm256_load_1rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
227 x+j_coord_offsetC,x+j_coord_offsetD,
230 /* Calculate displacement vector */
231 dx00 = _mm256_sub_pd(ix0,jx0);
232 dy00 = _mm256_sub_pd(iy0,jy0);
233 dz00 = _mm256_sub_pd(iz0,jz0);
234 dx10 = _mm256_sub_pd(ix1,jx0);
235 dy10 = _mm256_sub_pd(iy1,jy0);
236 dz10 = _mm256_sub_pd(iz1,jz0);
237 dx20 = _mm256_sub_pd(ix2,jx0);
238 dy20 = _mm256_sub_pd(iy2,jy0);
239 dz20 = _mm256_sub_pd(iz2,jz0);
241 /* Calculate squared distance and things based on it */
242 rsq00 = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
243 rsq10 = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
244 rsq20 = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
246 rinv00 = gmx_mm256_invsqrt_pd(rsq00);
247 rinv10 = gmx_mm256_invsqrt_pd(rsq10);
248 rinv20 = gmx_mm256_invsqrt_pd(rsq20);
250 rinvsq00 = _mm256_mul_pd(rinv00,rinv00);
251 rinvsq10 = _mm256_mul_pd(rinv10,rinv10);
252 rinvsq20 = _mm256_mul_pd(rinv20,rinv20);
254 /* Load parameters for j particles */
255 jq0 = gmx_mm256_load_4real_swizzle_pd(charge+jnrA+0,charge+jnrB+0,
256 charge+jnrC+0,charge+jnrD+0);
257 vdwjidx0A = 2*vdwtype[jnrA+0];
258 vdwjidx0B = 2*vdwtype[jnrB+0];
259 vdwjidx0C = 2*vdwtype[jnrC+0];
260 vdwjidx0D = 2*vdwtype[jnrD+0];
262 fjx0 = _mm256_setzero_pd();
263 fjy0 = _mm256_setzero_pd();
264 fjz0 = _mm256_setzero_pd();
266 /**************************
267 * CALCULATE INTERACTIONS *
268 **************************/
270 if (gmx_mm256_any_lt(rsq00,rcutoff2))
273 r00 = _mm256_mul_pd(rsq00,rinv00);
275 /* Compute parameters for interactions between i and j atoms */
276 qq00 = _mm256_mul_pd(iq0,jq0);
277 gmx_mm256_load_4pair_swizzle_pd(vdwioffsetptr0+vdwjidx0A,
278 vdwioffsetptr0+vdwjidx0B,
279 vdwioffsetptr0+vdwjidx0C,
280 vdwioffsetptr0+vdwjidx0D,
283 /* EWALD ELECTROSTATICS */
285 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
286 ewrt = _mm256_mul_pd(r00,ewtabscale);
287 ewitab = _mm256_cvttpd_epi32(ewrt);
288 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
289 ewitab = _mm_slli_epi32(ewitab,2);
290 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
291 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
292 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
293 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
294 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
295 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
296 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
297 velec = _mm256_mul_pd(qq00,_mm256_sub_pd(rinv00,velec));
298 felec = _mm256_mul_pd(_mm256_mul_pd(qq00,rinv00),_mm256_sub_pd(rinvsq00,felec));
300 /* LENNARD-JONES DISPERSION/REPULSION */
302 rinvsix = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
303 vvdw6 = _mm256_mul_pd(c6_00,rinvsix);
304 vvdw12 = _mm256_mul_pd(c12_00,_mm256_mul_pd(rinvsix,rinvsix));
305 vvdw = _mm256_sub_pd( _mm256_mul_pd(vvdw12,one_twelfth) , _mm256_mul_pd(vvdw6,one_sixth) );
306 fvdw = _mm256_mul_pd(_mm256_sub_pd(vvdw12,vvdw6),rinvsq00);
308 d = _mm256_sub_pd(r00,rswitch);
309 d = _mm256_max_pd(d,_mm256_setzero_pd());
310 d2 = _mm256_mul_pd(d,d);
311 sw = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
313 dsw = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
315 /* Evaluate switch function */
316 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
317 felec = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv00,_mm256_mul_pd(velec,dsw)) );
318 fvdw = _mm256_sub_pd( _mm256_mul_pd(fvdw,sw) , _mm256_mul_pd(rinv00,_mm256_mul_pd(vvdw,dsw)) );
319 velec = _mm256_mul_pd(velec,sw);
320 vvdw = _mm256_mul_pd(vvdw,sw);
321 cutoff_mask = _mm256_cmp_pd(rsq00,rcutoff2,_CMP_LT_OQ);
323 /* Update potential sum for this i atom from the interaction with this j atom. */
324 velec = _mm256_and_pd(velec,cutoff_mask);
325 velecsum = _mm256_add_pd(velecsum,velec);
326 vvdw = _mm256_and_pd(vvdw,cutoff_mask);
327 vvdwsum = _mm256_add_pd(vvdwsum,vvdw);
329 fscal = _mm256_add_pd(felec,fvdw);
331 fscal = _mm256_and_pd(fscal,cutoff_mask);
333 /* Calculate temporary vectorial force */
334 tx = _mm256_mul_pd(fscal,dx00);
335 ty = _mm256_mul_pd(fscal,dy00);
336 tz = _mm256_mul_pd(fscal,dz00);
338 /* Update vectorial force */
339 fix0 = _mm256_add_pd(fix0,tx);
340 fiy0 = _mm256_add_pd(fiy0,ty);
341 fiz0 = _mm256_add_pd(fiz0,tz);
343 fjx0 = _mm256_add_pd(fjx0,tx);
344 fjy0 = _mm256_add_pd(fjy0,ty);
345 fjz0 = _mm256_add_pd(fjz0,tz);
349 /**************************
350 * CALCULATE INTERACTIONS *
351 **************************/
353 if (gmx_mm256_any_lt(rsq10,rcutoff2))
356 r10 = _mm256_mul_pd(rsq10,rinv10);
358 /* Compute parameters for interactions between i and j atoms */
359 qq10 = _mm256_mul_pd(iq1,jq0);
361 /* EWALD ELECTROSTATICS */
363 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
364 ewrt = _mm256_mul_pd(r10,ewtabscale);
365 ewitab = _mm256_cvttpd_epi32(ewrt);
366 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
367 ewitab = _mm_slli_epi32(ewitab,2);
368 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
369 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
370 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
371 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
372 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
373 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
374 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
375 velec = _mm256_mul_pd(qq10,_mm256_sub_pd(rinv10,velec));
376 felec = _mm256_mul_pd(_mm256_mul_pd(qq10,rinv10),_mm256_sub_pd(rinvsq10,felec));
378 d = _mm256_sub_pd(r10,rswitch);
379 d = _mm256_max_pd(d,_mm256_setzero_pd());
380 d2 = _mm256_mul_pd(d,d);
381 sw = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
383 dsw = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
385 /* Evaluate switch function */
386 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
387 felec = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv10,_mm256_mul_pd(velec,dsw)) );
388 velec = _mm256_mul_pd(velec,sw);
389 cutoff_mask = _mm256_cmp_pd(rsq10,rcutoff2,_CMP_LT_OQ);
391 /* Update potential sum for this i atom from the interaction with this j atom. */
392 velec = _mm256_and_pd(velec,cutoff_mask);
393 velecsum = _mm256_add_pd(velecsum,velec);
397 fscal = _mm256_and_pd(fscal,cutoff_mask);
399 /* Calculate temporary vectorial force */
400 tx = _mm256_mul_pd(fscal,dx10);
401 ty = _mm256_mul_pd(fscal,dy10);
402 tz = _mm256_mul_pd(fscal,dz10);
404 /* Update vectorial force */
405 fix1 = _mm256_add_pd(fix1,tx);
406 fiy1 = _mm256_add_pd(fiy1,ty);
407 fiz1 = _mm256_add_pd(fiz1,tz);
409 fjx0 = _mm256_add_pd(fjx0,tx);
410 fjy0 = _mm256_add_pd(fjy0,ty);
411 fjz0 = _mm256_add_pd(fjz0,tz);
415 /**************************
416 * CALCULATE INTERACTIONS *
417 **************************/
419 if (gmx_mm256_any_lt(rsq20,rcutoff2))
422 r20 = _mm256_mul_pd(rsq20,rinv20);
424 /* Compute parameters for interactions between i and j atoms */
425 qq20 = _mm256_mul_pd(iq2,jq0);
427 /* EWALD ELECTROSTATICS */
429 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
430 ewrt = _mm256_mul_pd(r20,ewtabscale);
431 ewitab = _mm256_cvttpd_epi32(ewrt);
432 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
433 ewitab = _mm_slli_epi32(ewitab,2);
434 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
435 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
436 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
437 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
438 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
439 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
440 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
441 velec = _mm256_mul_pd(qq20,_mm256_sub_pd(rinv20,velec));
442 felec = _mm256_mul_pd(_mm256_mul_pd(qq20,rinv20),_mm256_sub_pd(rinvsq20,felec));
444 d = _mm256_sub_pd(r20,rswitch);
445 d = _mm256_max_pd(d,_mm256_setzero_pd());
446 d2 = _mm256_mul_pd(d,d);
447 sw = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
449 dsw = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
451 /* Evaluate switch function */
452 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
453 felec = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv20,_mm256_mul_pd(velec,dsw)) );
454 velec = _mm256_mul_pd(velec,sw);
455 cutoff_mask = _mm256_cmp_pd(rsq20,rcutoff2,_CMP_LT_OQ);
457 /* Update potential sum for this i atom from the interaction with this j atom. */
458 velec = _mm256_and_pd(velec,cutoff_mask);
459 velecsum = _mm256_add_pd(velecsum,velec);
463 fscal = _mm256_and_pd(fscal,cutoff_mask);
465 /* Calculate temporary vectorial force */
466 tx = _mm256_mul_pd(fscal,dx20);
467 ty = _mm256_mul_pd(fscal,dy20);
468 tz = _mm256_mul_pd(fscal,dz20);
470 /* Update vectorial force */
471 fix2 = _mm256_add_pd(fix2,tx);
472 fiy2 = _mm256_add_pd(fiy2,ty);
473 fiz2 = _mm256_add_pd(fiz2,tz);
475 fjx0 = _mm256_add_pd(fjx0,tx);
476 fjy0 = _mm256_add_pd(fjy0,ty);
477 fjz0 = _mm256_add_pd(fjz0,tz);
481 fjptrA = f+j_coord_offsetA;
482 fjptrB = f+j_coord_offsetB;
483 fjptrC = f+j_coord_offsetC;
484 fjptrD = f+j_coord_offsetD;
486 gmx_mm256_decrement_1rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,fjx0,fjy0,fjz0);
488 /* Inner loop uses 216 flops */
494 /* Get j neighbor index, and coordinate index */
495 jnrlistA = jjnr[jidx];
496 jnrlistB = jjnr[jidx+1];
497 jnrlistC = jjnr[jidx+2];
498 jnrlistD = jjnr[jidx+3];
499 /* Sign of each element will be negative for non-real atoms.
500 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
501 * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
503 tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
505 tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
506 tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
507 dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
509 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
510 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
511 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
512 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
513 j_coord_offsetA = DIM*jnrA;
514 j_coord_offsetB = DIM*jnrB;
515 j_coord_offsetC = DIM*jnrC;
516 j_coord_offsetD = DIM*jnrD;
518 /* load j atom coordinates */
519 gmx_mm256_load_1rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
520 x+j_coord_offsetC,x+j_coord_offsetD,
523 /* Calculate displacement vector */
524 dx00 = _mm256_sub_pd(ix0,jx0);
525 dy00 = _mm256_sub_pd(iy0,jy0);
526 dz00 = _mm256_sub_pd(iz0,jz0);
527 dx10 = _mm256_sub_pd(ix1,jx0);
528 dy10 = _mm256_sub_pd(iy1,jy0);
529 dz10 = _mm256_sub_pd(iz1,jz0);
530 dx20 = _mm256_sub_pd(ix2,jx0);
531 dy20 = _mm256_sub_pd(iy2,jy0);
532 dz20 = _mm256_sub_pd(iz2,jz0);
534 /* Calculate squared distance and things based on it */
535 rsq00 = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
536 rsq10 = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
537 rsq20 = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
539 rinv00 = gmx_mm256_invsqrt_pd(rsq00);
540 rinv10 = gmx_mm256_invsqrt_pd(rsq10);
541 rinv20 = gmx_mm256_invsqrt_pd(rsq20);
543 rinvsq00 = _mm256_mul_pd(rinv00,rinv00);
544 rinvsq10 = _mm256_mul_pd(rinv10,rinv10);
545 rinvsq20 = _mm256_mul_pd(rinv20,rinv20);
547 /* Load parameters for j particles */
548 jq0 = gmx_mm256_load_4real_swizzle_pd(charge+jnrA+0,charge+jnrB+0,
549 charge+jnrC+0,charge+jnrD+0);
550 vdwjidx0A = 2*vdwtype[jnrA+0];
551 vdwjidx0B = 2*vdwtype[jnrB+0];
552 vdwjidx0C = 2*vdwtype[jnrC+0];
553 vdwjidx0D = 2*vdwtype[jnrD+0];
555 fjx0 = _mm256_setzero_pd();
556 fjy0 = _mm256_setzero_pd();
557 fjz0 = _mm256_setzero_pd();
559 /**************************
560 * CALCULATE INTERACTIONS *
561 **************************/
563 if (gmx_mm256_any_lt(rsq00,rcutoff2))
566 r00 = _mm256_mul_pd(rsq00,rinv00);
567 r00 = _mm256_andnot_pd(dummy_mask,r00);
569 /* Compute parameters for interactions between i and j atoms */
570 qq00 = _mm256_mul_pd(iq0,jq0);
571 gmx_mm256_load_4pair_swizzle_pd(vdwioffsetptr0+vdwjidx0A,
572 vdwioffsetptr0+vdwjidx0B,
573 vdwioffsetptr0+vdwjidx0C,
574 vdwioffsetptr0+vdwjidx0D,
577 /* EWALD ELECTROSTATICS */
579 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
580 ewrt = _mm256_mul_pd(r00,ewtabscale);
581 ewitab = _mm256_cvttpd_epi32(ewrt);
582 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
583 ewitab = _mm_slli_epi32(ewitab,2);
584 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
585 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
586 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
587 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
588 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
589 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
590 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
591 velec = _mm256_mul_pd(qq00,_mm256_sub_pd(rinv00,velec));
592 felec = _mm256_mul_pd(_mm256_mul_pd(qq00,rinv00),_mm256_sub_pd(rinvsq00,felec));
594 /* LENNARD-JONES DISPERSION/REPULSION */
596 rinvsix = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
597 vvdw6 = _mm256_mul_pd(c6_00,rinvsix);
598 vvdw12 = _mm256_mul_pd(c12_00,_mm256_mul_pd(rinvsix,rinvsix));
599 vvdw = _mm256_sub_pd( _mm256_mul_pd(vvdw12,one_twelfth) , _mm256_mul_pd(vvdw6,one_sixth) );
600 fvdw = _mm256_mul_pd(_mm256_sub_pd(vvdw12,vvdw6),rinvsq00);
602 d = _mm256_sub_pd(r00,rswitch);
603 d = _mm256_max_pd(d,_mm256_setzero_pd());
604 d2 = _mm256_mul_pd(d,d);
605 sw = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
607 dsw = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
609 /* Evaluate switch function */
610 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
611 felec = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv00,_mm256_mul_pd(velec,dsw)) );
612 fvdw = _mm256_sub_pd( _mm256_mul_pd(fvdw,sw) , _mm256_mul_pd(rinv00,_mm256_mul_pd(vvdw,dsw)) );
613 velec = _mm256_mul_pd(velec,sw);
614 vvdw = _mm256_mul_pd(vvdw,sw);
615 cutoff_mask = _mm256_cmp_pd(rsq00,rcutoff2,_CMP_LT_OQ);
617 /* Update potential sum for this i atom from the interaction with this j atom. */
618 velec = _mm256_and_pd(velec,cutoff_mask);
619 velec = _mm256_andnot_pd(dummy_mask,velec);
620 velecsum = _mm256_add_pd(velecsum,velec);
621 vvdw = _mm256_and_pd(vvdw,cutoff_mask);
622 vvdw = _mm256_andnot_pd(dummy_mask,vvdw);
623 vvdwsum = _mm256_add_pd(vvdwsum,vvdw);
625 fscal = _mm256_add_pd(felec,fvdw);
627 fscal = _mm256_and_pd(fscal,cutoff_mask);
629 fscal = _mm256_andnot_pd(dummy_mask,fscal);
631 /* Calculate temporary vectorial force */
632 tx = _mm256_mul_pd(fscal,dx00);
633 ty = _mm256_mul_pd(fscal,dy00);
634 tz = _mm256_mul_pd(fscal,dz00);
636 /* Update vectorial force */
637 fix0 = _mm256_add_pd(fix0,tx);
638 fiy0 = _mm256_add_pd(fiy0,ty);
639 fiz0 = _mm256_add_pd(fiz0,tz);
641 fjx0 = _mm256_add_pd(fjx0,tx);
642 fjy0 = _mm256_add_pd(fjy0,ty);
643 fjz0 = _mm256_add_pd(fjz0,tz);
647 /**************************
648 * CALCULATE INTERACTIONS *
649 **************************/
651 if (gmx_mm256_any_lt(rsq10,rcutoff2))
654 r10 = _mm256_mul_pd(rsq10,rinv10);
655 r10 = _mm256_andnot_pd(dummy_mask,r10);
657 /* Compute parameters for interactions between i and j atoms */
658 qq10 = _mm256_mul_pd(iq1,jq0);
660 /* EWALD ELECTROSTATICS */
662 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
663 ewrt = _mm256_mul_pd(r10,ewtabscale);
664 ewitab = _mm256_cvttpd_epi32(ewrt);
665 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
666 ewitab = _mm_slli_epi32(ewitab,2);
667 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
668 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
669 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
670 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
671 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
672 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
673 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
674 velec = _mm256_mul_pd(qq10,_mm256_sub_pd(rinv10,velec));
675 felec = _mm256_mul_pd(_mm256_mul_pd(qq10,rinv10),_mm256_sub_pd(rinvsq10,felec));
677 d = _mm256_sub_pd(r10,rswitch);
678 d = _mm256_max_pd(d,_mm256_setzero_pd());
679 d2 = _mm256_mul_pd(d,d);
680 sw = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
682 dsw = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
684 /* Evaluate switch function */
685 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
686 felec = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv10,_mm256_mul_pd(velec,dsw)) );
687 velec = _mm256_mul_pd(velec,sw);
688 cutoff_mask = _mm256_cmp_pd(rsq10,rcutoff2,_CMP_LT_OQ);
690 /* Update potential sum for this i atom from the interaction with this j atom. */
691 velec = _mm256_and_pd(velec,cutoff_mask);
692 velec = _mm256_andnot_pd(dummy_mask,velec);
693 velecsum = _mm256_add_pd(velecsum,velec);
697 fscal = _mm256_and_pd(fscal,cutoff_mask);
699 fscal = _mm256_andnot_pd(dummy_mask,fscal);
701 /* Calculate temporary vectorial force */
702 tx = _mm256_mul_pd(fscal,dx10);
703 ty = _mm256_mul_pd(fscal,dy10);
704 tz = _mm256_mul_pd(fscal,dz10);
706 /* Update vectorial force */
707 fix1 = _mm256_add_pd(fix1,tx);
708 fiy1 = _mm256_add_pd(fiy1,ty);
709 fiz1 = _mm256_add_pd(fiz1,tz);
711 fjx0 = _mm256_add_pd(fjx0,tx);
712 fjy0 = _mm256_add_pd(fjy0,ty);
713 fjz0 = _mm256_add_pd(fjz0,tz);
717 /**************************
718 * CALCULATE INTERACTIONS *
719 **************************/
721 if (gmx_mm256_any_lt(rsq20,rcutoff2))
724 r20 = _mm256_mul_pd(rsq20,rinv20);
725 r20 = _mm256_andnot_pd(dummy_mask,r20);
727 /* Compute parameters for interactions between i and j atoms */
728 qq20 = _mm256_mul_pd(iq2,jq0);
730 /* EWALD ELECTROSTATICS */
732 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
733 ewrt = _mm256_mul_pd(r20,ewtabscale);
734 ewitab = _mm256_cvttpd_epi32(ewrt);
735 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
736 ewitab = _mm_slli_epi32(ewitab,2);
737 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
738 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
739 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
740 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
741 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
742 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
743 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
744 velec = _mm256_mul_pd(qq20,_mm256_sub_pd(rinv20,velec));
745 felec = _mm256_mul_pd(_mm256_mul_pd(qq20,rinv20),_mm256_sub_pd(rinvsq20,felec));
747 d = _mm256_sub_pd(r20,rswitch);
748 d = _mm256_max_pd(d,_mm256_setzero_pd());
749 d2 = _mm256_mul_pd(d,d);
750 sw = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
752 dsw = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
754 /* Evaluate switch function */
755 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
756 felec = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv20,_mm256_mul_pd(velec,dsw)) );
757 velec = _mm256_mul_pd(velec,sw);
758 cutoff_mask = _mm256_cmp_pd(rsq20,rcutoff2,_CMP_LT_OQ);
760 /* Update potential sum for this i atom from the interaction with this j atom. */
761 velec = _mm256_and_pd(velec,cutoff_mask);
762 velec = _mm256_andnot_pd(dummy_mask,velec);
763 velecsum = _mm256_add_pd(velecsum,velec);
767 fscal = _mm256_and_pd(fscal,cutoff_mask);
769 fscal = _mm256_andnot_pd(dummy_mask,fscal);
771 /* Calculate temporary vectorial force */
772 tx = _mm256_mul_pd(fscal,dx20);
773 ty = _mm256_mul_pd(fscal,dy20);
774 tz = _mm256_mul_pd(fscal,dz20);
776 /* Update vectorial force */
777 fix2 = _mm256_add_pd(fix2,tx);
778 fiy2 = _mm256_add_pd(fiy2,ty);
779 fiz2 = _mm256_add_pd(fiz2,tz);
781 fjx0 = _mm256_add_pd(fjx0,tx);
782 fjy0 = _mm256_add_pd(fjy0,ty);
783 fjz0 = _mm256_add_pd(fjz0,tz);
787 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
788 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
789 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
790 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
792 gmx_mm256_decrement_1rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,fjx0,fjy0,fjz0);
794 /* Inner loop uses 219 flops */
797 /* End of innermost loop */
799 gmx_mm256_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
800 f+i_coord_offset,fshift+i_shift_offset);
803 /* Update potential energies */
804 gmx_mm256_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
805 gmx_mm256_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
807 /* Increment number of inner iterations */
808 inneriter += j_index_end - j_index_start;
810 /* Outer loop uses 20 flops */
813 /* Increment number of outer iterations */
816 /* Update outer/inner flops */
818 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*219);
821 * Gromacs nonbonded kernel: nb_kernel_ElecEwSw_VdwLJSw_GeomW3P1_F_avx_256_double
822 * Electrostatics interaction: Ewald
823 * VdW interaction: LennardJones
824 * Geometry: Water3-Particle
825 * Calculate force/pot: Force
828 nb_kernel_ElecEwSw_VdwLJSw_GeomW3P1_F_avx_256_double
829 (t_nblist * gmx_restrict nlist,
830 rvec * gmx_restrict xx,
831 rvec * gmx_restrict ff,
832 t_forcerec * gmx_restrict fr,
833 t_mdatoms * gmx_restrict mdatoms,
834 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
835 t_nrnb * gmx_restrict nrnb)
837 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
838 * just 0 for non-waters.
839 * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
840 * jnr indices corresponding to data put in the four positions in the SIMD register.
842 int i_shift_offset,i_coord_offset,outeriter,inneriter;
843 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
844 int jnrA,jnrB,jnrC,jnrD;
845 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
846 int jnrlistE,jnrlistF,jnrlistG,jnrlistH;
847 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
848 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
850 real *shiftvec,*fshift,*x,*f;
851 real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
853 __m256d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
854 real * vdwioffsetptr0;
855 __m256d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
856 real * vdwioffsetptr1;
857 __m256d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
858 real * vdwioffsetptr2;
859 __m256d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
860 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
861 __m256d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
862 __m256d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
863 __m256d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
864 __m256d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
865 __m256d velec,felec,velecsum,facel,crf,krf,krf2;
868 __m256d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
871 __m256d one_sixth = _mm256_set1_pd(1.0/6.0);
872 __m256d one_twelfth = _mm256_set1_pd(1.0/12.0);
874 __m256d ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
875 __m256d beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
877 __m256d rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
878 real rswitch_scalar,d_scalar;
879 __m256d dummy_mask,cutoff_mask;
880 __m128 tmpmask0,tmpmask1;
881 __m256d signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
882 __m256d one = _mm256_set1_pd(1.0);
883 __m256d two = _mm256_set1_pd(2.0);
889 jindex = nlist->jindex;
891 shiftidx = nlist->shift;
893 shiftvec = fr->shift_vec[0];
894 fshift = fr->fshift[0];
895 facel = _mm256_set1_pd(fr->epsfac);
896 charge = mdatoms->chargeA;
897 nvdwtype = fr->ntype;
899 vdwtype = mdatoms->typeA;
901 sh_ewald = _mm256_set1_pd(fr->ic->sh_ewald);
902 beta = _mm256_set1_pd(fr->ic->ewaldcoeff_q);
903 beta2 = _mm256_mul_pd(beta,beta);
904 beta3 = _mm256_mul_pd(beta,beta2);
906 ewtab = fr->ic->tabq_coul_FDV0;
907 ewtabscale = _mm256_set1_pd(fr->ic->tabq_scale);
908 ewtabhalfspace = _mm256_set1_pd(0.5/fr->ic->tabq_scale);
910 /* Setup water-specific parameters */
911 inr = nlist->iinr[0];
912 iq0 = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+0]));
913 iq1 = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
914 iq2 = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
915 vdwioffsetptr0 = vdwparam+2*nvdwtype*vdwtype[inr+0];
917 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
918 rcutoff_scalar = fr->rcoulomb;
919 rcutoff = _mm256_set1_pd(rcutoff_scalar);
920 rcutoff2 = _mm256_mul_pd(rcutoff,rcutoff);
922 rswitch_scalar = fr->rcoulomb_switch;
923 rswitch = _mm256_set1_pd(rswitch_scalar);
924 /* Setup switch parameters */
925 d_scalar = rcutoff_scalar-rswitch_scalar;
926 d = _mm256_set1_pd(d_scalar);
927 swV3 = _mm256_set1_pd(-10.0/(d_scalar*d_scalar*d_scalar));
928 swV4 = _mm256_set1_pd( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
929 swV5 = _mm256_set1_pd( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
930 swF2 = _mm256_set1_pd(-30.0/(d_scalar*d_scalar*d_scalar));
931 swF3 = _mm256_set1_pd( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
932 swF4 = _mm256_set1_pd(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
934 /* Avoid stupid compiler warnings */
935 jnrA = jnrB = jnrC = jnrD = 0;
944 for(iidx=0;iidx<4*DIM;iidx++)
949 /* Start outer loop over neighborlists */
950 for(iidx=0; iidx<nri; iidx++)
952 /* Load shift vector for this list */
953 i_shift_offset = DIM*shiftidx[iidx];
955 /* Load limits for loop over neighbors */
956 j_index_start = jindex[iidx];
957 j_index_end = jindex[iidx+1];
959 /* Get outer coordinate index */
961 i_coord_offset = DIM*inr;
963 /* Load i particle coords and add shift vector */
964 gmx_mm256_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
965 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
967 fix0 = _mm256_setzero_pd();
968 fiy0 = _mm256_setzero_pd();
969 fiz0 = _mm256_setzero_pd();
970 fix1 = _mm256_setzero_pd();
971 fiy1 = _mm256_setzero_pd();
972 fiz1 = _mm256_setzero_pd();
973 fix2 = _mm256_setzero_pd();
974 fiy2 = _mm256_setzero_pd();
975 fiz2 = _mm256_setzero_pd();
977 /* Start inner kernel loop */
978 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
981 /* Get j neighbor index, and coordinate index */
986 j_coord_offsetA = DIM*jnrA;
987 j_coord_offsetB = DIM*jnrB;
988 j_coord_offsetC = DIM*jnrC;
989 j_coord_offsetD = DIM*jnrD;
991 /* load j atom coordinates */
992 gmx_mm256_load_1rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
993 x+j_coord_offsetC,x+j_coord_offsetD,
996 /* Calculate displacement vector */
997 dx00 = _mm256_sub_pd(ix0,jx0);
998 dy00 = _mm256_sub_pd(iy0,jy0);
999 dz00 = _mm256_sub_pd(iz0,jz0);
1000 dx10 = _mm256_sub_pd(ix1,jx0);
1001 dy10 = _mm256_sub_pd(iy1,jy0);
1002 dz10 = _mm256_sub_pd(iz1,jz0);
1003 dx20 = _mm256_sub_pd(ix2,jx0);
1004 dy20 = _mm256_sub_pd(iy2,jy0);
1005 dz20 = _mm256_sub_pd(iz2,jz0);
1007 /* Calculate squared distance and things based on it */
1008 rsq00 = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
1009 rsq10 = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
1010 rsq20 = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
1012 rinv00 = gmx_mm256_invsqrt_pd(rsq00);
1013 rinv10 = gmx_mm256_invsqrt_pd(rsq10);
1014 rinv20 = gmx_mm256_invsqrt_pd(rsq20);
1016 rinvsq00 = _mm256_mul_pd(rinv00,rinv00);
1017 rinvsq10 = _mm256_mul_pd(rinv10,rinv10);
1018 rinvsq20 = _mm256_mul_pd(rinv20,rinv20);
1020 /* Load parameters for j particles */
1021 jq0 = gmx_mm256_load_4real_swizzle_pd(charge+jnrA+0,charge+jnrB+0,
1022 charge+jnrC+0,charge+jnrD+0);
1023 vdwjidx0A = 2*vdwtype[jnrA+0];
1024 vdwjidx0B = 2*vdwtype[jnrB+0];
1025 vdwjidx0C = 2*vdwtype[jnrC+0];
1026 vdwjidx0D = 2*vdwtype[jnrD+0];
1028 fjx0 = _mm256_setzero_pd();
1029 fjy0 = _mm256_setzero_pd();
1030 fjz0 = _mm256_setzero_pd();
1032 /**************************
1033 * CALCULATE INTERACTIONS *
1034 **************************/
1036 if (gmx_mm256_any_lt(rsq00,rcutoff2))
1039 r00 = _mm256_mul_pd(rsq00,rinv00);
1041 /* Compute parameters for interactions between i and j atoms */
1042 qq00 = _mm256_mul_pd(iq0,jq0);
1043 gmx_mm256_load_4pair_swizzle_pd(vdwioffsetptr0+vdwjidx0A,
1044 vdwioffsetptr0+vdwjidx0B,
1045 vdwioffsetptr0+vdwjidx0C,
1046 vdwioffsetptr0+vdwjidx0D,
1049 /* EWALD ELECTROSTATICS */
1051 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1052 ewrt = _mm256_mul_pd(r00,ewtabscale);
1053 ewitab = _mm256_cvttpd_epi32(ewrt);
1054 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1055 ewitab = _mm_slli_epi32(ewitab,2);
1056 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1057 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1058 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1059 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1060 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1061 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1062 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1063 velec = _mm256_mul_pd(qq00,_mm256_sub_pd(rinv00,velec));
1064 felec = _mm256_mul_pd(_mm256_mul_pd(qq00,rinv00),_mm256_sub_pd(rinvsq00,felec));
1066 /* LENNARD-JONES DISPERSION/REPULSION */
1068 rinvsix = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1069 vvdw6 = _mm256_mul_pd(c6_00,rinvsix);
1070 vvdw12 = _mm256_mul_pd(c12_00,_mm256_mul_pd(rinvsix,rinvsix));
1071 vvdw = _mm256_sub_pd( _mm256_mul_pd(vvdw12,one_twelfth) , _mm256_mul_pd(vvdw6,one_sixth) );
1072 fvdw = _mm256_mul_pd(_mm256_sub_pd(vvdw12,vvdw6),rinvsq00);
1074 d = _mm256_sub_pd(r00,rswitch);
1075 d = _mm256_max_pd(d,_mm256_setzero_pd());
1076 d2 = _mm256_mul_pd(d,d);
1077 sw = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
1079 dsw = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
1081 /* Evaluate switch function */
1082 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1083 felec = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv00,_mm256_mul_pd(velec,dsw)) );
1084 fvdw = _mm256_sub_pd( _mm256_mul_pd(fvdw,sw) , _mm256_mul_pd(rinv00,_mm256_mul_pd(vvdw,dsw)) );
1085 cutoff_mask = _mm256_cmp_pd(rsq00,rcutoff2,_CMP_LT_OQ);
1087 fscal = _mm256_add_pd(felec,fvdw);
1089 fscal = _mm256_and_pd(fscal,cutoff_mask);
1091 /* Calculate temporary vectorial force */
1092 tx = _mm256_mul_pd(fscal,dx00);
1093 ty = _mm256_mul_pd(fscal,dy00);
1094 tz = _mm256_mul_pd(fscal,dz00);
1096 /* Update vectorial force */
1097 fix0 = _mm256_add_pd(fix0,tx);
1098 fiy0 = _mm256_add_pd(fiy0,ty);
1099 fiz0 = _mm256_add_pd(fiz0,tz);
1101 fjx0 = _mm256_add_pd(fjx0,tx);
1102 fjy0 = _mm256_add_pd(fjy0,ty);
1103 fjz0 = _mm256_add_pd(fjz0,tz);
1107 /**************************
1108 * CALCULATE INTERACTIONS *
1109 **************************/
1111 if (gmx_mm256_any_lt(rsq10,rcutoff2))
1114 r10 = _mm256_mul_pd(rsq10,rinv10);
1116 /* Compute parameters for interactions between i and j atoms */
1117 qq10 = _mm256_mul_pd(iq1,jq0);
1119 /* EWALD ELECTROSTATICS */
1121 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1122 ewrt = _mm256_mul_pd(r10,ewtabscale);
1123 ewitab = _mm256_cvttpd_epi32(ewrt);
1124 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1125 ewitab = _mm_slli_epi32(ewitab,2);
1126 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1127 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1128 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1129 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1130 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1131 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1132 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1133 velec = _mm256_mul_pd(qq10,_mm256_sub_pd(rinv10,velec));
1134 felec = _mm256_mul_pd(_mm256_mul_pd(qq10,rinv10),_mm256_sub_pd(rinvsq10,felec));
1136 d = _mm256_sub_pd(r10,rswitch);
1137 d = _mm256_max_pd(d,_mm256_setzero_pd());
1138 d2 = _mm256_mul_pd(d,d);
1139 sw = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
1141 dsw = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
1143 /* Evaluate switch function */
1144 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1145 felec = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv10,_mm256_mul_pd(velec,dsw)) );
1146 cutoff_mask = _mm256_cmp_pd(rsq10,rcutoff2,_CMP_LT_OQ);
1150 fscal = _mm256_and_pd(fscal,cutoff_mask);
1152 /* Calculate temporary vectorial force */
1153 tx = _mm256_mul_pd(fscal,dx10);
1154 ty = _mm256_mul_pd(fscal,dy10);
1155 tz = _mm256_mul_pd(fscal,dz10);
1157 /* Update vectorial force */
1158 fix1 = _mm256_add_pd(fix1,tx);
1159 fiy1 = _mm256_add_pd(fiy1,ty);
1160 fiz1 = _mm256_add_pd(fiz1,tz);
1162 fjx0 = _mm256_add_pd(fjx0,tx);
1163 fjy0 = _mm256_add_pd(fjy0,ty);
1164 fjz0 = _mm256_add_pd(fjz0,tz);
1168 /**************************
1169 * CALCULATE INTERACTIONS *
1170 **************************/
1172 if (gmx_mm256_any_lt(rsq20,rcutoff2))
1175 r20 = _mm256_mul_pd(rsq20,rinv20);
1177 /* Compute parameters for interactions between i and j atoms */
1178 qq20 = _mm256_mul_pd(iq2,jq0);
1180 /* EWALD ELECTROSTATICS */
1182 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1183 ewrt = _mm256_mul_pd(r20,ewtabscale);
1184 ewitab = _mm256_cvttpd_epi32(ewrt);
1185 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1186 ewitab = _mm_slli_epi32(ewitab,2);
1187 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1188 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1189 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1190 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1191 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1192 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1193 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1194 velec = _mm256_mul_pd(qq20,_mm256_sub_pd(rinv20,velec));
1195 felec = _mm256_mul_pd(_mm256_mul_pd(qq20,rinv20),_mm256_sub_pd(rinvsq20,felec));
1197 d = _mm256_sub_pd(r20,rswitch);
1198 d = _mm256_max_pd(d,_mm256_setzero_pd());
1199 d2 = _mm256_mul_pd(d,d);
1200 sw = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
1202 dsw = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
1204 /* Evaluate switch function */
1205 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1206 felec = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv20,_mm256_mul_pd(velec,dsw)) );
1207 cutoff_mask = _mm256_cmp_pd(rsq20,rcutoff2,_CMP_LT_OQ);
1211 fscal = _mm256_and_pd(fscal,cutoff_mask);
1213 /* Calculate temporary vectorial force */
1214 tx = _mm256_mul_pd(fscal,dx20);
1215 ty = _mm256_mul_pd(fscal,dy20);
1216 tz = _mm256_mul_pd(fscal,dz20);
1218 /* Update vectorial force */
1219 fix2 = _mm256_add_pd(fix2,tx);
1220 fiy2 = _mm256_add_pd(fiy2,ty);
1221 fiz2 = _mm256_add_pd(fiz2,tz);
1223 fjx0 = _mm256_add_pd(fjx0,tx);
1224 fjy0 = _mm256_add_pd(fjy0,ty);
1225 fjz0 = _mm256_add_pd(fjz0,tz);
1229 fjptrA = f+j_coord_offsetA;
1230 fjptrB = f+j_coord_offsetB;
1231 fjptrC = f+j_coord_offsetC;
1232 fjptrD = f+j_coord_offsetD;
1234 gmx_mm256_decrement_1rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,fjx0,fjy0,fjz0);
1236 /* Inner loop uses 204 flops */
1239 if(jidx<j_index_end)
1242 /* Get j neighbor index, and coordinate index */
1243 jnrlistA = jjnr[jidx];
1244 jnrlistB = jjnr[jidx+1];
1245 jnrlistC = jjnr[jidx+2];
1246 jnrlistD = jjnr[jidx+3];
1247 /* Sign of each element will be negative for non-real atoms.
1248 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1249 * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
1251 tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
1253 tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
1254 tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
1255 dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
1257 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
1258 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
1259 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
1260 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
1261 j_coord_offsetA = DIM*jnrA;
1262 j_coord_offsetB = DIM*jnrB;
1263 j_coord_offsetC = DIM*jnrC;
1264 j_coord_offsetD = DIM*jnrD;
1266 /* load j atom coordinates */
1267 gmx_mm256_load_1rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1268 x+j_coord_offsetC,x+j_coord_offsetD,
1271 /* Calculate displacement vector */
1272 dx00 = _mm256_sub_pd(ix0,jx0);
1273 dy00 = _mm256_sub_pd(iy0,jy0);
1274 dz00 = _mm256_sub_pd(iz0,jz0);
1275 dx10 = _mm256_sub_pd(ix1,jx0);
1276 dy10 = _mm256_sub_pd(iy1,jy0);
1277 dz10 = _mm256_sub_pd(iz1,jz0);
1278 dx20 = _mm256_sub_pd(ix2,jx0);
1279 dy20 = _mm256_sub_pd(iy2,jy0);
1280 dz20 = _mm256_sub_pd(iz2,jz0);
1282 /* Calculate squared distance and things based on it */
1283 rsq00 = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
1284 rsq10 = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
1285 rsq20 = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
1287 rinv00 = gmx_mm256_invsqrt_pd(rsq00);
1288 rinv10 = gmx_mm256_invsqrt_pd(rsq10);
1289 rinv20 = gmx_mm256_invsqrt_pd(rsq20);
1291 rinvsq00 = _mm256_mul_pd(rinv00,rinv00);
1292 rinvsq10 = _mm256_mul_pd(rinv10,rinv10);
1293 rinvsq20 = _mm256_mul_pd(rinv20,rinv20);
1295 /* Load parameters for j particles */
1296 jq0 = gmx_mm256_load_4real_swizzle_pd(charge+jnrA+0,charge+jnrB+0,
1297 charge+jnrC+0,charge+jnrD+0);
1298 vdwjidx0A = 2*vdwtype[jnrA+0];
1299 vdwjidx0B = 2*vdwtype[jnrB+0];
1300 vdwjidx0C = 2*vdwtype[jnrC+0];
1301 vdwjidx0D = 2*vdwtype[jnrD+0];
1303 fjx0 = _mm256_setzero_pd();
1304 fjy0 = _mm256_setzero_pd();
1305 fjz0 = _mm256_setzero_pd();
1307 /**************************
1308 * CALCULATE INTERACTIONS *
1309 **************************/
1311 if (gmx_mm256_any_lt(rsq00,rcutoff2))
1314 r00 = _mm256_mul_pd(rsq00,rinv00);
1315 r00 = _mm256_andnot_pd(dummy_mask,r00);
1317 /* Compute parameters for interactions between i and j atoms */
1318 qq00 = _mm256_mul_pd(iq0,jq0);
1319 gmx_mm256_load_4pair_swizzle_pd(vdwioffsetptr0+vdwjidx0A,
1320 vdwioffsetptr0+vdwjidx0B,
1321 vdwioffsetptr0+vdwjidx0C,
1322 vdwioffsetptr0+vdwjidx0D,
1325 /* EWALD ELECTROSTATICS */
1327 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1328 ewrt = _mm256_mul_pd(r00,ewtabscale);
1329 ewitab = _mm256_cvttpd_epi32(ewrt);
1330 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1331 ewitab = _mm_slli_epi32(ewitab,2);
1332 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1333 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1334 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1335 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1336 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1337 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1338 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1339 velec = _mm256_mul_pd(qq00,_mm256_sub_pd(rinv00,velec));
1340 felec = _mm256_mul_pd(_mm256_mul_pd(qq00,rinv00),_mm256_sub_pd(rinvsq00,felec));
1342 /* LENNARD-JONES DISPERSION/REPULSION */
1344 rinvsix = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1345 vvdw6 = _mm256_mul_pd(c6_00,rinvsix);
1346 vvdw12 = _mm256_mul_pd(c12_00,_mm256_mul_pd(rinvsix,rinvsix));
1347 vvdw = _mm256_sub_pd( _mm256_mul_pd(vvdw12,one_twelfth) , _mm256_mul_pd(vvdw6,one_sixth) );
1348 fvdw = _mm256_mul_pd(_mm256_sub_pd(vvdw12,vvdw6),rinvsq00);
1350 d = _mm256_sub_pd(r00,rswitch);
1351 d = _mm256_max_pd(d,_mm256_setzero_pd());
1352 d2 = _mm256_mul_pd(d,d);
1353 sw = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
1355 dsw = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
1357 /* Evaluate switch function */
1358 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1359 felec = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv00,_mm256_mul_pd(velec,dsw)) );
1360 fvdw = _mm256_sub_pd( _mm256_mul_pd(fvdw,sw) , _mm256_mul_pd(rinv00,_mm256_mul_pd(vvdw,dsw)) );
1361 cutoff_mask = _mm256_cmp_pd(rsq00,rcutoff2,_CMP_LT_OQ);
1363 fscal = _mm256_add_pd(felec,fvdw);
1365 fscal = _mm256_and_pd(fscal,cutoff_mask);
1367 fscal = _mm256_andnot_pd(dummy_mask,fscal);
1369 /* Calculate temporary vectorial force */
1370 tx = _mm256_mul_pd(fscal,dx00);
1371 ty = _mm256_mul_pd(fscal,dy00);
1372 tz = _mm256_mul_pd(fscal,dz00);
1374 /* Update vectorial force */
1375 fix0 = _mm256_add_pd(fix0,tx);
1376 fiy0 = _mm256_add_pd(fiy0,ty);
1377 fiz0 = _mm256_add_pd(fiz0,tz);
1379 fjx0 = _mm256_add_pd(fjx0,tx);
1380 fjy0 = _mm256_add_pd(fjy0,ty);
1381 fjz0 = _mm256_add_pd(fjz0,tz);
1385 /**************************
1386 * CALCULATE INTERACTIONS *
1387 **************************/
1389 if (gmx_mm256_any_lt(rsq10,rcutoff2))
1392 r10 = _mm256_mul_pd(rsq10,rinv10);
1393 r10 = _mm256_andnot_pd(dummy_mask,r10);
1395 /* Compute parameters for interactions between i and j atoms */
1396 qq10 = _mm256_mul_pd(iq1,jq0);
1398 /* EWALD ELECTROSTATICS */
1400 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1401 ewrt = _mm256_mul_pd(r10,ewtabscale);
1402 ewitab = _mm256_cvttpd_epi32(ewrt);
1403 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1404 ewitab = _mm_slli_epi32(ewitab,2);
1405 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1406 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1407 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1408 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1409 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1410 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1411 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1412 velec = _mm256_mul_pd(qq10,_mm256_sub_pd(rinv10,velec));
1413 felec = _mm256_mul_pd(_mm256_mul_pd(qq10,rinv10),_mm256_sub_pd(rinvsq10,felec));
1415 d = _mm256_sub_pd(r10,rswitch);
1416 d = _mm256_max_pd(d,_mm256_setzero_pd());
1417 d2 = _mm256_mul_pd(d,d);
1418 sw = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
1420 dsw = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
1422 /* Evaluate switch function */
1423 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1424 felec = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv10,_mm256_mul_pd(velec,dsw)) );
1425 cutoff_mask = _mm256_cmp_pd(rsq10,rcutoff2,_CMP_LT_OQ);
1429 fscal = _mm256_and_pd(fscal,cutoff_mask);
1431 fscal = _mm256_andnot_pd(dummy_mask,fscal);
1433 /* Calculate temporary vectorial force */
1434 tx = _mm256_mul_pd(fscal,dx10);
1435 ty = _mm256_mul_pd(fscal,dy10);
1436 tz = _mm256_mul_pd(fscal,dz10);
1438 /* Update vectorial force */
1439 fix1 = _mm256_add_pd(fix1,tx);
1440 fiy1 = _mm256_add_pd(fiy1,ty);
1441 fiz1 = _mm256_add_pd(fiz1,tz);
1443 fjx0 = _mm256_add_pd(fjx0,tx);
1444 fjy0 = _mm256_add_pd(fjy0,ty);
1445 fjz0 = _mm256_add_pd(fjz0,tz);
1449 /**************************
1450 * CALCULATE INTERACTIONS *
1451 **************************/
1453 if (gmx_mm256_any_lt(rsq20,rcutoff2))
1456 r20 = _mm256_mul_pd(rsq20,rinv20);
1457 r20 = _mm256_andnot_pd(dummy_mask,r20);
1459 /* Compute parameters for interactions between i and j atoms */
1460 qq20 = _mm256_mul_pd(iq2,jq0);
1462 /* EWALD ELECTROSTATICS */
1464 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1465 ewrt = _mm256_mul_pd(r20,ewtabscale);
1466 ewitab = _mm256_cvttpd_epi32(ewrt);
1467 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1468 ewitab = _mm_slli_epi32(ewitab,2);
1469 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1470 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1471 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1472 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1473 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1474 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1475 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1476 velec = _mm256_mul_pd(qq20,_mm256_sub_pd(rinv20,velec));
1477 felec = _mm256_mul_pd(_mm256_mul_pd(qq20,rinv20),_mm256_sub_pd(rinvsq20,felec));
1479 d = _mm256_sub_pd(r20,rswitch);
1480 d = _mm256_max_pd(d,_mm256_setzero_pd());
1481 d2 = _mm256_mul_pd(d,d);
1482 sw = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
1484 dsw = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
1486 /* Evaluate switch function */
1487 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1488 felec = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv20,_mm256_mul_pd(velec,dsw)) );
1489 cutoff_mask = _mm256_cmp_pd(rsq20,rcutoff2,_CMP_LT_OQ);
1493 fscal = _mm256_and_pd(fscal,cutoff_mask);
1495 fscal = _mm256_andnot_pd(dummy_mask,fscal);
1497 /* Calculate temporary vectorial force */
1498 tx = _mm256_mul_pd(fscal,dx20);
1499 ty = _mm256_mul_pd(fscal,dy20);
1500 tz = _mm256_mul_pd(fscal,dz20);
1502 /* Update vectorial force */
1503 fix2 = _mm256_add_pd(fix2,tx);
1504 fiy2 = _mm256_add_pd(fiy2,ty);
1505 fiz2 = _mm256_add_pd(fiz2,tz);
1507 fjx0 = _mm256_add_pd(fjx0,tx);
1508 fjy0 = _mm256_add_pd(fjy0,ty);
1509 fjz0 = _mm256_add_pd(fjz0,tz);
1513 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1514 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1515 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1516 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1518 gmx_mm256_decrement_1rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,fjx0,fjy0,fjz0);
1520 /* Inner loop uses 207 flops */
1523 /* End of innermost loop */
1525 gmx_mm256_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1526 f+i_coord_offset,fshift+i_shift_offset);
1528 /* Increment number of inner iterations */
1529 inneriter += j_index_end - j_index_start;
1531 /* Outer loop uses 18 flops */
1534 /* Increment number of outer iterations */
1537 /* Update outer/inner flops */
1539 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_F,outeriter*18 + inneriter*207);