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.
44 #include "../nb_kernel.h"
45 #include "types/simple.h"
46 #include "gromacs/math/vec.h"
49 #include "gromacs/simd/math_x86_avx_256_double.h"
50 #include "kernelutil_x86_avx_256_double.h"
53 * Gromacs nonbonded kernel: nb_kernel_ElecEwSw_VdwNone_GeomW3P1_VF_avx_256_double
54 * Electrostatics interaction: Ewald
55 * VdW interaction: None
56 * Geometry: Water3-Particle
57 * Calculate force/pot: PotentialAndForce
60 nb_kernel_ElecEwSw_VdwNone_GeomW3P1_VF_avx_256_double
61 (t_nblist * gmx_restrict nlist,
62 rvec * gmx_restrict xx,
63 rvec * gmx_restrict ff,
64 t_forcerec * gmx_restrict fr,
65 t_mdatoms * gmx_restrict mdatoms,
66 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
67 t_nrnb * gmx_restrict nrnb)
69 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
70 * just 0 for non-waters.
71 * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
72 * jnr indices corresponding to data put in the four positions in the SIMD register.
74 int i_shift_offset,i_coord_offset,outeriter,inneriter;
75 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
76 int jnrA,jnrB,jnrC,jnrD;
77 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
78 int jnrlistE,jnrlistF,jnrlistG,jnrlistH;
79 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
80 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
82 real *shiftvec,*fshift,*x,*f;
83 real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
85 __m256d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
86 real * vdwioffsetptr0;
87 __m256d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
88 real * vdwioffsetptr1;
89 __m256d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
90 real * vdwioffsetptr2;
91 __m256d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
92 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
93 __m256d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
94 __m256d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
95 __m256d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
96 __m256d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
97 __m256d velec,felec,velecsum,facel,crf,krf,krf2;
100 __m256d ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
101 __m256d beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
103 __m256d rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
104 real rswitch_scalar,d_scalar;
105 __m256d dummy_mask,cutoff_mask;
106 __m128 tmpmask0,tmpmask1;
107 __m256d signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
108 __m256d one = _mm256_set1_pd(1.0);
109 __m256d two = _mm256_set1_pd(2.0);
115 jindex = nlist->jindex;
117 shiftidx = nlist->shift;
119 shiftvec = fr->shift_vec[0];
120 fshift = fr->fshift[0];
121 facel = _mm256_set1_pd(fr->epsfac);
122 charge = mdatoms->chargeA;
124 sh_ewald = _mm256_set1_pd(fr->ic->sh_ewald);
125 beta = _mm256_set1_pd(fr->ic->ewaldcoeff_q);
126 beta2 = _mm256_mul_pd(beta,beta);
127 beta3 = _mm256_mul_pd(beta,beta2);
129 ewtab = fr->ic->tabq_coul_FDV0;
130 ewtabscale = _mm256_set1_pd(fr->ic->tabq_scale);
131 ewtabhalfspace = _mm256_set1_pd(0.5/fr->ic->tabq_scale);
133 /* Setup water-specific parameters */
134 inr = nlist->iinr[0];
135 iq0 = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+0]));
136 iq1 = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
137 iq2 = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
139 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
140 rcutoff_scalar = fr->rcoulomb;
141 rcutoff = _mm256_set1_pd(rcutoff_scalar);
142 rcutoff2 = _mm256_mul_pd(rcutoff,rcutoff);
144 rswitch_scalar = fr->rcoulomb_switch;
145 rswitch = _mm256_set1_pd(rswitch_scalar);
146 /* Setup switch parameters */
147 d_scalar = rcutoff_scalar-rswitch_scalar;
148 d = _mm256_set1_pd(d_scalar);
149 swV3 = _mm256_set1_pd(-10.0/(d_scalar*d_scalar*d_scalar));
150 swV4 = _mm256_set1_pd( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
151 swV5 = _mm256_set1_pd( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
152 swF2 = _mm256_set1_pd(-30.0/(d_scalar*d_scalar*d_scalar));
153 swF3 = _mm256_set1_pd( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
154 swF4 = _mm256_set1_pd(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
156 /* Avoid stupid compiler warnings */
157 jnrA = jnrB = jnrC = jnrD = 0;
166 for(iidx=0;iidx<4*DIM;iidx++)
171 /* Start outer loop over neighborlists */
172 for(iidx=0; iidx<nri; iidx++)
174 /* Load shift vector for this list */
175 i_shift_offset = DIM*shiftidx[iidx];
177 /* Load limits for loop over neighbors */
178 j_index_start = jindex[iidx];
179 j_index_end = jindex[iidx+1];
181 /* Get outer coordinate index */
183 i_coord_offset = DIM*inr;
185 /* Load i particle coords and add shift vector */
186 gmx_mm256_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
187 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
189 fix0 = _mm256_setzero_pd();
190 fiy0 = _mm256_setzero_pd();
191 fiz0 = _mm256_setzero_pd();
192 fix1 = _mm256_setzero_pd();
193 fiy1 = _mm256_setzero_pd();
194 fiz1 = _mm256_setzero_pd();
195 fix2 = _mm256_setzero_pd();
196 fiy2 = _mm256_setzero_pd();
197 fiz2 = _mm256_setzero_pd();
199 /* Reset potential sums */
200 velecsum = _mm256_setzero_pd();
202 /* Start inner kernel loop */
203 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
206 /* Get j neighbor index, and coordinate index */
211 j_coord_offsetA = DIM*jnrA;
212 j_coord_offsetB = DIM*jnrB;
213 j_coord_offsetC = DIM*jnrC;
214 j_coord_offsetD = DIM*jnrD;
216 /* load j atom coordinates */
217 gmx_mm256_load_1rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
218 x+j_coord_offsetC,x+j_coord_offsetD,
221 /* Calculate displacement vector */
222 dx00 = _mm256_sub_pd(ix0,jx0);
223 dy00 = _mm256_sub_pd(iy0,jy0);
224 dz00 = _mm256_sub_pd(iz0,jz0);
225 dx10 = _mm256_sub_pd(ix1,jx0);
226 dy10 = _mm256_sub_pd(iy1,jy0);
227 dz10 = _mm256_sub_pd(iz1,jz0);
228 dx20 = _mm256_sub_pd(ix2,jx0);
229 dy20 = _mm256_sub_pd(iy2,jy0);
230 dz20 = _mm256_sub_pd(iz2,jz0);
232 /* Calculate squared distance and things based on it */
233 rsq00 = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
234 rsq10 = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
235 rsq20 = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
237 rinv00 = gmx_mm256_invsqrt_pd(rsq00);
238 rinv10 = gmx_mm256_invsqrt_pd(rsq10);
239 rinv20 = gmx_mm256_invsqrt_pd(rsq20);
241 rinvsq00 = _mm256_mul_pd(rinv00,rinv00);
242 rinvsq10 = _mm256_mul_pd(rinv10,rinv10);
243 rinvsq20 = _mm256_mul_pd(rinv20,rinv20);
245 /* Load parameters for j particles */
246 jq0 = gmx_mm256_load_4real_swizzle_pd(charge+jnrA+0,charge+jnrB+0,
247 charge+jnrC+0,charge+jnrD+0);
249 fjx0 = _mm256_setzero_pd();
250 fjy0 = _mm256_setzero_pd();
251 fjz0 = _mm256_setzero_pd();
253 /**************************
254 * CALCULATE INTERACTIONS *
255 **************************/
257 if (gmx_mm256_any_lt(rsq00,rcutoff2))
260 r00 = _mm256_mul_pd(rsq00,rinv00);
262 /* Compute parameters for interactions between i and j atoms */
263 qq00 = _mm256_mul_pd(iq0,jq0);
265 /* EWALD ELECTROSTATICS */
267 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
268 ewrt = _mm256_mul_pd(r00,ewtabscale);
269 ewitab = _mm256_cvttpd_epi32(ewrt);
270 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
271 ewitab = _mm_slli_epi32(ewitab,2);
272 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
273 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
274 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
275 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
276 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
277 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
278 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
279 velec = _mm256_mul_pd(qq00,_mm256_sub_pd(rinv00,velec));
280 felec = _mm256_mul_pd(_mm256_mul_pd(qq00,rinv00),_mm256_sub_pd(rinvsq00,felec));
282 d = _mm256_sub_pd(r00,rswitch);
283 d = _mm256_max_pd(d,_mm256_setzero_pd());
284 d2 = _mm256_mul_pd(d,d);
285 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)))))));
287 dsw = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
289 /* Evaluate switch function */
290 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
291 felec = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv00,_mm256_mul_pd(velec,dsw)) );
292 velec = _mm256_mul_pd(velec,sw);
293 cutoff_mask = _mm256_cmp_pd(rsq00,rcutoff2,_CMP_LT_OQ);
295 /* Update potential sum for this i atom from the interaction with this j atom. */
296 velec = _mm256_and_pd(velec,cutoff_mask);
297 velecsum = _mm256_add_pd(velecsum,velec);
301 fscal = _mm256_and_pd(fscal,cutoff_mask);
303 /* Calculate temporary vectorial force */
304 tx = _mm256_mul_pd(fscal,dx00);
305 ty = _mm256_mul_pd(fscal,dy00);
306 tz = _mm256_mul_pd(fscal,dz00);
308 /* Update vectorial force */
309 fix0 = _mm256_add_pd(fix0,tx);
310 fiy0 = _mm256_add_pd(fiy0,ty);
311 fiz0 = _mm256_add_pd(fiz0,tz);
313 fjx0 = _mm256_add_pd(fjx0,tx);
314 fjy0 = _mm256_add_pd(fjy0,ty);
315 fjz0 = _mm256_add_pd(fjz0,tz);
319 /**************************
320 * CALCULATE INTERACTIONS *
321 **************************/
323 if (gmx_mm256_any_lt(rsq10,rcutoff2))
326 r10 = _mm256_mul_pd(rsq10,rinv10);
328 /* Compute parameters for interactions between i and j atoms */
329 qq10 = _mm256_mul_pd(iq1,jq0);
331 /* EWALD ELECTROSTATICS */
333 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
334 ewrt = _mm256_mul_pd(r10,ewtabscale);
335 ewitab = _mm256_cvttpd_epi32(ewrt);
336 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
337 ewitab = _mm_slli_epi32(ewitab,2);
338 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
339 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
340 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
341 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
342 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
343 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
344 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
345 velec = _mm256_mul_pd(qq10,_mm256_sub_pd(rinv10,velec));
346 felec = _mm256_mul_pd(_mm256_mul_pd(qq10,rinv10),_mm256_sub_pd(rinvsq10,felec));
348 d = _mm256_sub_pd(r10,rswitch);
349 d = _mm256_max_pd(d,_mm256_setzero_pd());
350 d2 = _mm256_mul_pd(d,d);
351 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)))))));
353 dsw = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
355 /* Evaluate switch function */
356 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
357 felec = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv10,_mm256_mul_pd(velec,dsw)) );
358 velec = _mm256_mul_pd(velec,sw);
359 cutoff_mask = _mm256_cmp_pd(rsq10,rcutoff2,_CMP_LT_OQ);
361 /* Update potential sum for this i atom from the interaction with this j atom. */
362 velec = _mm256_and_pd(velec,cutoff_mask);
363 velecsum = _mm256_add_pd(velecsum,velec);
367 fscal = _mm256_and_pd(fscal,cutoff_mask);
369 /* Calculate temporary vectorial force */
370 tx = _mm256_mul_pd(fscal,dx10);
371 ty = _mm256_mul_pd(fscal,dy10);
372 tz = _mm256_mul_pd(fscal,dz10);
374 /* Update vectorial force */
375 fix1 = _mm256_add_pd(fix1,tx);
376 fiy1 = _mm256_add_pd(fiy1,ty);
377 fiz1 = _mm256_add_pd(fiz1,tz);
379 fjx0 = _mm256_add_pd(fjx0,tx);
380 fjy0 = _mm256_add_pd(fjy0,ty);
381 fjz0 = _mm256_add_pd(fjz0,tz);
385 /**************************
386 * CALCULATE INTERACTIONS *
387 **************************/
389 if (gmx_mm256_any_lt(rsq20,rcutoff2))
392 r20 = _mm256_mul_pd(rsq20,rinv20);
394 /* Compute parameters for interactions between i and j atoms */
395 qq20 = _mm256_mul_pd(iq2,jq0);
397 /* EWALD ELECTROSTATICS */
399 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
400 ewrt = _mm256_mul_pd(r20,ewtabscale);
401 ewitab = _mm256_cvttpd_epi32(ewrt);
402 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
403 ewitab = _mm_slli_epi32(ewitab,2);
404 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
405 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
406 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
407 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
408 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
409 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
410 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
411 velec = _mm256_mul_pd(qq20,_mm256_sub_pd(rinv20,velec));
412 felec = _mm256_mul_pd(_mm256_mul_pd(qq20,rinv20),_mm256_sub_pd(rinvsq20,felec));
414 d = _mm256_sub_pd(r20,rswitch);
415 d = _mm256_max_pd(d,_mm256_setzero_pd());
416 d2 = _mm256_mul_pd(d,d);
417 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)))))));
419 dsw = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
421 /* Evaluate switch function */
422 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
423 felec = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv20,_mm256_mul_pd(velec,dsw)) );
424 velec = _mm256_mul_pd(velec,sw);
425 cutoff_mask = _mm256_cmp_pd(rsq20,rcutoff2,_CMP_LT_OQ);
427 /* Update potential sum for this i atom from the interaction with this j atom. */
428 velec = _mm256_and_pd(velec,cutoff_mask);
429 velecsum = _mm256_add_pd(velecsum,velec);
433 fscal = _mm256_and_pd(fscal,cutoff_mask);
435 /* Calculate temporary vectorial force */
436 tx = _mm256_mul_pd(fscal,dx20);
437 ty = _mm256_mul_pd(fscal,dy20);
438 tz = _mm256_mul_pd(fscal,dz20);
440 /* Update vectorial force */
441 fix2 = _mm256_add_pd(fix2,tx);
442 fiy2 = _mm256_add_pd(fiy2,ty);
443 fiz2 = _mm256_add_pd(fiz2,tz);
445 fjx0 = _mm256_add_pd(fjx0,tx);
446 fjy0 = _mm256_add_pd(fjy0,ty);
447 fjz0 = _mm256_add_pd(fjz0,tz);
451 fjptrA = f+j_coord_offsetA;
452 fjptrB = f+j_coord_offsetB;
453 fjptrC = f+j_coord_offsetC;
454 fjptrD = f+j_coord_offsetD;
456 gmx_mm256_decrement_1rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,fjx0,fjy0,fjz0);
458 /* Inner loop uses 198 flops */
464 /* Get j neighbor index, and coordinate index */
465 jnrlistA = jjnr[jidx];
466 jnrlistB = jjnr[jidx+1];
467 jnrlistC = jjnr[jidx+2];
468 jnrlistD = jjnr[jidx+3];
469 /* Sign of each element will be negative for non-real atoms.
470 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
471 * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
473 tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
475 tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
476 tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
477 dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
479 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
480 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
481 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
482 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
483 j_coord_offsetA = DIM*jnrA;
484 j_coord_offsetB = DIM*jnrB;
485 j_coord_offsetC = DIM*jnrC;
486 j_coord_offsetD = DIM*jnrD;
488 /* load j atom coordinates */
489 gmx_mm256_load_1rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
490 x+j_coord_offsetC,x+j_coord_offsetD,
493 /* Calculate displacement vector */
494 dx00 = _mm256_sub_pd(ix0,jx0);
495 dy00 = _mm256_sub_pd(iy0,jy0);
496 dz00 = _mm256_sub_pd(iz0,jz0);
497 dx10 = _mm256_sub_pd(ix1,jx0);
498 dy10 = _mm256_sub_pd(iy1,jy0);
499 dz10 = _mm256_sub_pd(iz1,jz0);
500 dx20 = _mm256_sub_pd(ix2,jx0);
501 dy20 = _mm256_sub_pd(iy2,jy0);
502 dz20 = _mm256_sub_pd(iz2,jz0);
504 /* Calculate squared distance and things based on it */
505 rsq00 = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
506 rsq10 = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
507 rsq20 = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
509 rinv00 = gmx_mm256_invsqrt_pd(rsq00);
510 rinv10 = gmx_mm256_invsqrt_pd(rsq10);
511 rinv20 = gmx_mm256_invsqrt_pd(rsq20);
513 rinvsq00 = _mm256_mul_pd(rinv00,rinv00);
514 rinvsq10 = _mm256_mul_pd(rinv10,rinv10);
515 rinvsq20 = _mm256_mul_pd(rinv20,rinv20);
517 /* Load parameters for j particles */
518 jq0 = gmx_mm256_load_4real_swizzle_pd(charge+jnrA+0,charge+jnrB+0,
519 charge+jnrC+0,charge+jnrD+0);
521 fjx0 = _mm256_setzero_pd();
522 fjy0 = _mm256_setzero_pd();
523 fjz0 = _mm256_setzero_pd();
525 /**************************
526 * CALCULATE INTERACTIONS *
527 **************************/
529 if (gmx_mm256_any_lt(rsq00,rcutoff2))
532 r00 = _mm256_mul_pd(rsq00,rinv00);
533 r00 = _mm256_andnot_pd(dummy_mask,r00);
535 /* Compute parameters for interactions between i and j atoms */
536 qq00 = _mm256_mul_pd(iq0,jq0);
538 /* EWALD ELECTROSTATICS */
540 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
541 ewrt = _mm256_mul_pd(r00,ewtabscale);
542 ewitab = _mm256_cvttpd_epi32(ewrt);
543 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
544 ewitab = _mm_slli_epi32(ewitab,2);
545 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
546 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
547 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
548 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
549 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
550 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
551 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
552 velec = _mm256_mul_pd(qq00,_mm256_sub_pd(rinv00,velec));
553 felec = _mm256_mul_pd(_mm256_mul_pd(qq00,rinv00),_mm256_sub_pd(rinvsq00,felec));
555 d = _mm256_sub_pd(r00,rswitch);
556 d = _mm256_max_pd(d,_mm256_setzero_pd());
557 d2 = _mm256_mul_pd(d,d);
558 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)))))));
560 dsw = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
562 /* Evaluate switch function */
563 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
564 felec = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv00,_mm256_mul_pd(velec,dsw)) );
565 velec = _mm256_mul_pd(velec,sw);
566 cutoff_mask = _mm256_cmp_pd(rsq00,rcutoff2,_CMP_LT_OQ);
568 /* Update potential sum for this i atom from the interaction with this j atom. */
569 velec = _mm256_and_pd(velec,cutoff_mask);
570 velec = _mm256_andnot_pd(dummy_mask,velec);
571 velecsum = _mm256_add_pd(velecsum,velec);
575 fscal = _mm256_and_pd(fscal,cutoff_mask);
577 fscal = _mm256_andnot_pd(dummy_mask,fscal);
579 /* Calculate temporary vectorial force */
580 tx = _mm256_mul_pd(fscal,dx00);
581 ty = _mm256_mul_pd(fscal,dy00);
582 tz = _mm256_mul_pd(fscal,dz00);
584 /* Update vectorial force */
585 fix0 = _mm256_add_pd(fix0,tx);
586 fiy0 = _mm256_add_pd(fiy0,ty);
587 fiz0 = _mm256_add_pd(fiz0,tz);
589 fjx0 = _mm256_add_pd(fjx0,tx);
590 fjy0 = _mm256_add_pd(fjy0,ty);
591 fjz0 = _mm256_add_pd(fjz0,tz);
595 /**************************
596 * CALCULATE INTERACTIONS *
597 **************************/
599 if (gmx_mm256_any_lt(rsq10,rcutoff2))
602 r10 = _mm256_mul_pd(rsq10,rinv10);
603 r10 = _mm256_andnot_pd(dummy_mask,r10);
605 /* Compute parameters for interactions between i and j atoms */
606 qq10 = _mm256_mul_pd(iq1,jq0);
608 /* EWALD ELECTROSTATICS */
610 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
611 ewrt = _mm256_mul_pd(r10,ewtabscale);
612 ewitab = _mm256_cvttpd_epi32(ewrt);
613 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
614 ewitab = _mm_slli_epi32(ewitab,2);
615 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
616 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
617 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
618 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
619 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
620 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
621 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
622 velec = _mm256_mul_pd(qq10,_mm256_sub_pd(rinv10,velec));
623 felec = _mm256_mul_pd(_mm256_mul_pd(qq10,rinv10),_mm256_sub_pd(rinvsq10,felec));
625 d = _mm256_sub_pd(r10,rswitch);
626 d = _mm256_max_pd(d,_mm256_setzero_pd());
627 d2 = _mm256_mul_pd(d,d);
628 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)))))));
630 dsw = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
632 /* Evaluate switch function */
633 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
634 felec = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv10,_mm256_mul_pd(velec,dsw)) );
635 velec = _mm256_mul_pd(velec,sw);
636 cutoff_mask = _mm256_cmp_pd(rsq10,rcutoff2,_CMP_LT_OQ);
638 /* Update potential sum for this i atom from the interaction with this j atom. */
639 velec = _mm256_and_pd(velec,cutoff_mask);
640 velec = _mm256_andnot_pd(dummy_mask,velec);
641 velecsum = _mm256_add_pd(velecsum,velec);
645 fscal = _mm256_and_pd(fscal,cutoff_mask);
647 fscal = _mm256_andnot_pd(dummy_mask,fscal);
649 /* Calculate temporary vectorial force */
650 tx = _mm256_mul_pd(fscal,dx10);
651 ty = _mm256_mul_pd(fscal,dy10);
652 tz = _mm256_mul_pd(fscal,dz10);
654 /* Update vectorial force */
655 fix1 = _mm256_add_pd(fix1,tx);
656 fiy1 = _mm256_add_pd(fiy1,ty);
657 fiz1 = _mm256_add_pd(fiz1,tz);
659 fjx0 = _mm256_add_pd(fjx0,tx);
660 fjy0 = _mm256_add_pd(fjy0,ty);
661 fjz0 = _mm256_add_pd(fjz0,tz);
665 /**************************
666 * CALCULATE INTERACTIONS *
667 **************************/
669 if (gmx_mm256_any_lt(rsq20,rcutoff2))
672 r20 = _mm256_mul_pd(rsq20,rinv20);
673 r20 = _mm256_andnot_pd(dummy_mask,r20);
675 /* Compute parameters for interactions between i and j atoms */
676 qq20 = _mm256_mul_pd(iq2,jq0);
678 /* EWALD ELECTROSTATICS */
680 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
681 ewrt = _mm256_mul_pd(r20,ewtabscale);
682 ewitab = _mm256_cvttpd_epi32(ewrt);
683 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
684 ewitab = _mm_slli_epi32(ewitab,2);
685 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
686 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
687 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
688 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
689 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
690 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
691 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
692 velec = _mm256_mul_pd(qq20,_mm256_sub_pd(rinv20,velec));
693 felec = _mm256_mul_pd(_mm256_mul_pd(qq20,rinv20),_mm256_sub_pd(rinvsq20,felec));
695 d = _mm256_sub_pd(r20,rswitch);
696 d = _mm256_max_pd(d,_mm256_setzero_pd());
697 d2 = _mm256_mul_pd(d,d);
698 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)))))));
700 dsw = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
702 /* Evaluate switch function */
703 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
704 felec = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv20,_mm256_mul_pd(velec,dsw)) );
705 velec = _mm256_mul_pd(velec,sw);
706 cutoff_mask = _mm256_cmp_pd(rsq20,rcutoff2,_CMP_LT_OQ);
708 /* Update potential sum for this i atom from the interaction with this j atom. */
709 velec = _mm256_and_pd(velec,cutoff_mask);
710 velec = _mm256_andnot_pd(dummy_mask,velec);
711 velecsum = _mm256_add_pd(velecsum,velec);
715 fscal = _mm256_and_pd(fscal,cutoff_mask);
717 fscal = _mm256_andnot_pd(dummy_mask,fscal);
719 /* Calculate temporary vectorial force */
720 tx = _mm256_mul_pd(fscal,dx20);
721 ty = _mm256_mul_pd(fscal,dy20);
722 tz = _mm256_mul_pd(fscal,dz20);
724 /* Update vectorial force */
725 fix2 = _mm256_add_pd(fix2,tx);
726 fiy2 = _mm256_add_pd(fiy2,ty);
727 fiz2 = _mm256_add_pd(fiz2,tz);
729 fjx0 = _mm256_add_pd(fjx0,tx);
730 fjy0 = _mm256_add_pd(fjy0,ty);
731 fjz0 = _mm256_add_pd(fjz0,tz);
735 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
736 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
737 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
738 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
740 gmx_mm256_decrement_1rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,fjx0,fjy0,fjz0);
742 /* Inner loop uses 201 flops */
745 /* End of innermost loop */
747 gmx_mm256_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
748 f+i_coord_offset,fshift+i_shift_offset);
751 /* Update potential energies */
752 gmx_mm256_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
754 /* Increment number of inner iterations */
755 inneriter += j_index_end - j_index_start;
757 /* Outer loop uses 19 flops */
760 /* Increment number of outer iterations */
763 /* Update outer/inner flops */
765 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3_VF,outeriter*19 + inneriter*201);
768 * Gromacs nonbonded kernel: nb_kernel_ElecEwSw_VdwNone_GeomW3P1_F_avx_256_double
769 * Electrostatics interaction: Ewald
770 * VdW interaction: None
771 * Geometry: Water3-Particle
772 * Calculate force/pot: Force
775 nb_kernel_ElecEwSw_VdwNone_GeomW3P1_F_avx_256_double
776 (t_nblist * gmx_restrict nlist,
777 rvec * gmx_restrict xx,
778 rvec * gmx_restrict ff,
779 t_forcerec * gmx_restrict fr,
780 t_mdatoms * gmx_restrict mdatoms,
781 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
782 t_nrnb * gmx_restrict nrnb)
784 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
785 * just 0 for non-waters.
786 * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
787 * jnr indices corresponding to data put in the four positions in the SIMD register.
789 int i_shift_offset,i_coord_offset,outeriter,inneriter;
790 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
791 int jnrA,jnrB,jnrC,jnrD;
792 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
793 int jnrlistE,jnrlistF,jnrlistG,jnrlistH;
794 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
795 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
797 real *shiftvec,*fshift,*x,*f;
798 real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
800 __m256d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
801 real * vdwioffsetptr0;
802 __m256d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
803 real * vdwioffsetptr1;
804 __m256d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
805 real * vdwioffsetptr2;
806 __m256d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
807 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
808 __m256d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
809 __m256d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
810 __m256d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
811 __m256d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
812 __m256d velec,felec,velecsum,facel,crf,krf,krf2;
815 __m256d ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
816 __m256d beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
818 __m256d rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
819 real rswitch_scalar,d_scalar;
820 __m256d dummy_mask,cutoff_mask;
821 __m128 tmpmask0,tmpmask1;
822 __m256d signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
823 __m256d one = _mm256_set1_pd(1.0);
824 __m256d two = _mm256_set1_pd(2.0);
830 jindex = nlist->jindex;
832 shiftidx = nlist->shift;
834 shiftvec = fr->shift_vec[0];
835 fshift = fr->fshift[0];
836 facel = _mm256_set1_pd(fr->epsfac);
837 charge = mdatoms->chargeA;
839 sh_ewald = _mm256_set1_pd(fr->ic->sh_ewald);
840 beta = _mm256_set1_pd(fr->ic->ewaldcoeff_q);
841 beta2 = _mm256_mul_pd(beta,beta);
842 beta3 = _mm256_mul_pd(beta,beta2);
844 ewtab = fr->ic->tabq_coul_FDV0;
845 ewtabscale = _mm256_set1_pd(fr->ic->tabq_scale);
846 ewtabhalfspace = _mm256_set1_pd(0.5/fr->ic->tabq_scale);
848 /* Setup water-specific parameters */
849 inr = nlist->iinr[0];
850 iq0 = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+0]));
851 iq1 = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
852 iq2 = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
854 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
855 rcutoff_scalar = fr->rcoulomb;
856 rcutoff = _mm256_set1_pd(rcutoff_scalar);
857 rcutoff2 = _mm256_mul_pd(rcutoff,rcutoff);
859 rswitch_scalar = fr->rcoulomb_switch;
860 rswitch = _mm256_set1_pd(rswitch_scalar);
861 /* Setup switch parameters */
862 d_scalar = rcutoff_scalar-rswitch_scalar;
863 d = _mm256_set1_pd(d_scalar);
864 swV3 = _mm256_set1_pd(-10.0/(d_scalar*d_scalar*d_scalar));
865 swV4 = _mm256_set1_pd( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
866 swV5 = _mm256_set1_pd( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
867 swF2 = _mm256_set1_pd(-30.0/(d_scalar*d_scalar*d_scalar));
868 swF3 = _mm256_set1_pd( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
869 swF4 = _mm256_set1_pd(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
871 /* Avoid stupid compiler warnings */
872 jnrA = jnrB = jnrC = jnrD = 0;
881 for(iidx=0;iidx<4*DIM;iidx++)
886 /* Start outer loop over neighborlists */
887 for(iidx=0; iidx<nri; iidx++)
889 /* Load shift vector for this list */
890 i_shift_offset = DIM*shiftidx[iidx];
892 /* Load limits for loop over neighbors */
893 j_index_start = jindex[iidx];
894 j_index_end = jindex[iidx+1];
896 /* Get outer coordinate index */
898 i_coord_offset = DIM*inr;
900 /* Load i particle coords and add shift vector */
901 gmx_mm256_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
902 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
904 fix0 = _mm256_setzero_pd();
905 fiy0 = _mm256_setzero_pd();
906 fiz0 = _mm256_setzero_pd();
907 fix1 = _mm256_setzero_pd();
908 fiy1 = _mm256_setzero_pd();
909 fiz1 = _mm256_setzero_pd();
910 fix2 = _mm256_setzero_pd();
911 fiy2 = _mm256_setzero_pd();
912 fiz2 = _mm256_setzero_pd();
914 /* Start inner kernel loop */
915 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
918 /* Get j neighbor index, and coordinate index */
923 j_coord_offsetA = DIM*jnrA;
924 j_coord_offsetB = DIM*jnrB;
925 j_coord_offsetC = DIM*jnrC;
926 j_coord_offsetD = DIM*jnrD;
928 /* load j atom coordinates */
929 gmx_mm256_load_1rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
930 x+j_coord_offsetC,x+j_coord_offsetD,
933 /* Calculate displacement vector */
934 dx00 = _mm256_sub_pd(ix0,jx0);
935 dy00 = _mm256_sub_pd(iy0,jy0);
936 dz00 = _mm256_sub_pd(iz0,jz0);
937 dx10 = _mm256_sub_pd(ix1,jx0);
938 dy10 = _mm256_sub_pd(iy1,jy0);
939 dz10 = _mm256_sub_pd(iz1,jz0);
940 dx20 = _mm256_sub_pd(ix2,jx0);
941 dy20 = _mm256_sub_pd(iy2,jy0);
942 dz20 = _mm256_sub_pd(iz2,jz0);
944 /* Calculate squared distance and things based on it */
945 rsq00 = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
946 rsq10 = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
947 rsq20 = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
949 rinv00 = gmx_mm256_invsqrt_pd(rsq00);
950 rinv10 = gmx_mm256_invsqrt_pd(rsq10);
951 rinv20 = gmx_mm256_invsqrt_pd(rsq20);
953 rinvsq00 = _mm256_mul_pd(rinv00,rinv00);
954 rinvsq10 = _mm256_mul_pd(rinv10,rinv10);
955 rinvsq20 = _mm256_mul_pd(rinv20,rinv20);
957 /* Load parameters for j particles */
958 jq0 = gmx_mm256_load_4real_swizzle_pd(charge+jnrA+0,charge+jnrB+0,
959 charge+jnrC+0,charge+jnrD+0);
961 fjx0 = _mm256_setzero_pd();
962 fjy0 = _mm256_setzero_pd();
963 fjz0 = _mm256_setzero_pd();
965 /**************************
966 * CALCULATE INTERACTIONS *
967 **************************/
969 if (gmx_mm256_any_lt(rsq00,rcutoff2))
972 r00 = _mm256_mul_pd(rsq00,rinv00);
974 /* Compute parameters for interactions between i and j atoms */
975 qq00 = _mm256_mul_pd(iq0,jq0);
977 /* EWALD ELECTROSTATICS */
979 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
980 ewrt = _mm256_mul_pd(r00,ewtabscale);
981 ewitab = _mm256_cvttpd_epi32(ewrt);
982 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
983 ewitab = _mm_slli_epi32(ewitab,2);
984 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
985 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
986 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
987 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
988 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
989 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
990 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
991 velec = _mm256_mul_pd(qq00,_mm256_sub_pd(rinv00,velec));
992 felec = _mm256_mul_pd(_mm256_mul_pd(qq00,rinv00),_mm256_sub_pd(rinvsq00,felec));
994 d = _mm256_sub_pd(r00,rswitch);
995 d = _mm256_max_pd(d,_mm256_setzero_pd());
996 d2 = _mm256_mul_pd(d,d);
997 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)))))));
999 dsw = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
1001 /* Evaluate switch function */
1002 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1003 felec = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv00,_mm256_mul_pd(velec,dsw)) );
1004 cutoff_mask = _mm256_cmp_pd(rsq00,rcutoff2,_CMP_LT_OQ);
1008 fscal = _mm256_and_pd(fscal,cutoff_mask);
1010 /* Calculate temporary vectorial force */
1011 tx = _mm256_mul_pd(fscal,dx00);
1012 ty = _mm256_mul_pd(fscal,dy00);
1013 tz = _mm256_mul_pd(fscal,dz00);
1015 /* Update vectorial force */
1016 fix0 = _mm256_add_pd(fix0,tx);
1017 fiy0 = _mm256_add_pd(fiy0,ty);
1018 fiz0 = _mm256_add_pd(fiz0,tz);
1020 fjx0 = _mm256_add_pd(fjx0,tx);
1021 fjy0 = _mm256_add_pd(fjy0,ty);
1022 fjz0 = _mm256_add_pd(fjz0,tz);
1026 /**************************
1027 * CALCULATE INTERACTIONS *
1028 **************************/
1030 if (gmx_mm256_any_lt(rsq10,rcutoff2))
1033 r10 = _mm256_mul_pd(rsq10,rinv10);
1035 /* Compute parameters for interactions between i and j atoms */
1036 qq10 = _mm256_mul_pd(iq1,jq0);
1038 /* EWALD ELECTROSTATICS */
1040 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1041 ewrt = _mm256_mul_pd(r10,ewtabscale);
1042 ewitab = _mm256_cvttpd_epi32(ewrt);
1043 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1044 ewitab = _mm_slli_epi32(ewitab,2);
1045 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1046 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1047 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1048 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1049 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1050 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1051 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1052 velec = _mm256_mul_pd(qq10,_mm256_sub_pd(rinv10,velec));
1053 felec = _mm256_mul_pd(_mm256_mul_pd(qq10,rinv10),_mm256_sub_pd(rinvsq10,felec));
1055 d = _mm256_sub_pd(r10,rswitch);
1056 d = _mm256_max_pd(d,_mm256_setzero_pd());
1057 d2 = _mm256_mul_pd(d,d);
1058 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)))))));
1060 dsw = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
1062 /* Evaluate switch function */
1063 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1064 felec = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv10,_mm256_mul_pd(velec,dsw)) );
1065 cutoff_mask = _mm256_cmp_pd(rsq10,rcutoff2,_CMP_LT_OQ);
1069 fscal = _mm256_and_pd(fscal,cutoff_mask);
1071 /* Calculate temporary vectorial force */
1072 tx = _mm256_mul_pd(fscal,dx10);
1073 ty = _mm256_mul_pd(fscal,dy10);
1074 tz = _mm256_mul_pd(fscal,dz10);
1076 /* Update vectorial force */
1077 fix1 = _mm256_add_pd(fix1,tx);
1078 fiy1 = _mm256_add_pd(fiy1,ty);
1079 fiz1 = _mm256_add_pd(fiz1,tz);
1081 fjx0 = _mm256_add_pd(fjx0,tx);
1082 fjy0 = _mm256_add_pd(fjy0,ty);
1083 fjz0 = _mm256_add_pd(fjz0,tz);
1087 /**************************
1088 * CALCULATE INTERACTIONS *
1089 **************************/
1091 if (gmx_mm256_any_lt(rsq20,rcutoff2))
1094 r20 = _mm256_mul_pd(rsq20,rinv20);
1096 /* Compute parameters for interactions between i and j atoms */
1097 qq20 = _mm256_mul_pd(iq2,jq0);
1099 /* EWALD ELECTROSTATICS */
1101 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1102 ewrt = _mm256_mul_pd(r20,ewtabscale);
1103 ewitab = _mm256_cvttpd_epi32(ewrt);
1104 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1105 ewitab = _mm_slli_epi32(ewitab,2);
1106 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1107 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1108 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1109 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1110 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1111 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1112 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1113 velec = _mm256_mul_pd(qq20,_mm256_sub_pd(rinv20,velec));
1114 felec = _mm256_mul_pd(_mm256_mul_pd(qq20,rinv20),_mm256_sub_pd(rinvsq20,felec));
1116 d = _mm256_sub_pd(r20,rswitch);
1117 d = _mm256_max_pd(d,_mm256_setzero_pd());
1118 d2 = _mm256_mul_pd(d,d);
1119 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)))))));
1121 dsw = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
1123 /* Evaluate switch function */
1124 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1125 felec = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv20,_mm256_mul_pd(velec,dsw)) );
1126 cutoff_mask = _mm256_cmp_pd(rsq20,rcutoff2,_CMP_LT_OQ);
1130 fscal = _mm256_and_pd(fscal,cutoff_mask);
1132 /* Calculate temporary vectorial force */
1133 tx = _mm256_mul_pd(fscal,dx20);
1134 ty = _mm256_mul_pd(fscal,dy20);
1135 tz = _mm256_mul_pd(fscal,dz20);
1137 /* Update vectorial force */
1138 fix2 = _mm256_add_pd(fix2,tx);
1139 fiy2 = _mm256_add_pd(fiy2,ty);
1140 fiz2 = _mm256_add_pd(fiz2,tz);
1142 fjx0 = _mm256_add_pd(fjx0,tx);
1143 fjy0 = _mm256_add_pd(fjy0,ty);
1144 fjz0 = _mm256_add_pd(fjz0,tz);
1148 fjptrA = f+j_coord_offsetA;
1149 fjptrB = f+j_coord_offsetB;
1150 fjptrC = f+j_coord_offsetC;
1151 fjptrD = f+j_coord_offsetD;
1153 gmx_mm256_decrement_1rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,fjx0,fjy0,fjz0);
1155 /* Inner loop uses 189 flops */
1158 if(jidx<j_index_end)
1161 /* Get j neighbor index, and coordinate index */
1162 jnrlistA = jjnr[jidx];
1163 jnrlistB = jjnr[jidx+1];
1164 jnrlistC = jjnr[jidx+2];
1165 jnrlistD = jjnr[jidx+3];
1166 /* Sign of each element will be negative for non-real atoms.
1167 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1168 * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
1170 tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
1172 tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
1173 tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
1174 dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
1176 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
1177 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
1178 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
1179 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
1180 j_coord_offsetA = DIM*jnrA;
1181 j_coord_offsetB = DIM*jnrB;
1182 j_coord_offsetC = DIM*jnrC;
1183 j_coord_offsetD = DIM*jnrD;
1185 /* load j atom coordinates */
1186 gmx_mm256_load_1rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1187 x+j_coord_offsetC,x+j_coord_offsetD,
1190 /* Calculate displacement vector */
1191 dx00 = _mm256_sub_pd(ix0,jx0);
1192 dy00 = _mm256_sub_pd(iy0,jy0);
1193 dz00 = _mm256_sub_pd(iz0,jz0);
1194 dx10 = _mm256_sub_pd(ix1,jx0);
1195 dy10 = _mm256_sub_pd(iy1,jy0);
1196 dz10 = _mm256_sub_pd(iz1,jz0);
1197 dx20 = _mm256_sub_pd(ix2,jx0);
1198 dy20 = _mm256_sub_pd(iy2,jy0);
1199 dz20 = _mm256_sub_pd(iz2,jz0);
1201 /* Calculate squared distance and things based on it */
1202 rsq00 = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
1203 rsq10 = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
1204 rsq20 = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
1206 rinv00 = gmx_mm256_invsqrt_pd(rsq00);
1207 rinv10 = gmx_mm256_invsqrt_pd(rsq10);
1208 rinv20 = gmx_mm256_invsqrt_pd(rsq20);
1210 rinvsq00 = _mm256_mul_pd(rinv00,rinv00);
1211 rinvsq10 = _mm256_mul_pd(rinv10,rinv10);
1212 rinvsq20 = _mm256_mul_pd(rinv20,rinv20);
1214 /* Load parameters for j particles */
1215 jq0 = gmx_mm256_load_4real_swizzle_pd(charge+jnrA+0,charge+jnrB+0,
1216 charge+jnrC+0,charge+jnrD+0);
1218 fjx0 = _mm256_setzero_pd();
1219 fjy0 = _mm256_setzero_pd();
1220 fjz0 = _mm256_setzero_pd();
1222 /**************************
1223 * CALCULATE INTERACTIONS *
1224 **************************/
1226 if (gmx_mm256_any_lt(rsq00,rcutoff2))
1229 r00 = _mm256_mul_pd(rsq00,rinv00);
1230 r00 = _mm256_andnot_pd(dummy_mask,r00);
1232 /* Compute parameters for interactions between i and j atoms */
1233 qq00 = _mm256_mul_pd(iq0,jq0);
1235 /* EWALD ELECTROSTATICS */
1237 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1238 ewrt = _mm256_mul_pd(r00,ewtabscale);
1239 ewitab = _mm256_cvttpd_epi32(ewrt);
1240 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1241 ewitab = _mm_slli_epi32(ewitab,2);
1242 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1243 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1244 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1245 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1246 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1247 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1248 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1249 velec = _mm256_mul_pd(qq00,_mm256_sub_pd(rinv00,velec));
1250 felec = _mm256_mul_pd(_mm256_mul_pd(qq00,rinv00),_mm256_sub_pd(rinvsq00,felec));
1252 d = _mm256_sub_pd(r00,rswitch);
1253 d = _mm256_max_pd(d,_mm256_setzero_pd());
1254 d2 = _mm256_mul_pd(d,d);
1255 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)))))));
1257 dsw = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
1259 /* Evaluate switch function */
1260 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1261 felec = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv00,_mm256_mul_pd(velec,dsw)) );
1262 cutoff_mask = _mm256_cmp_pd(rsq00,rcutoff2,_CMP_LT_OQ);
1266 fscal = _mm256_and_pd(fscal,cutoff_mask);
1268 fscal = _mm256_andnot_pd(dummy_mask,fscal);
1270 /* Calculate temporary vectorial force */
1271 tx = _mm256_mul_pd(fscal,dx00);
1272 ty = _mm256_mul_pd(fscal,dy00);
1273 tz = _mm256_mul_pd(fscal,dz00);
1275 /* Update vectorial force */
1276 fix0 = _mm256_add_pd(fix0,tx);
1277 fiy0 = _mm256_add_pd(fiy0,ty);
1278 fiz0 = _mm256_add_pd(fiz0,tz);
1280 fjx0 = _mm256_add_pd(fjx0,tx);
1281 fjy0 = _mm256_add_pd(fjy0,ty);
1282 fjz0 = _mm256_add_pd(fjz0,tz);
1286 /**************************
1287 * CALCULATE INTERACTIONS *
1288 **************************/
1290 if (gmx_mm256_any_lt(rsq10,rcutoff2))
1293 r10 = _mm256_mul_pd(rsq10,rinv10);
1294 r10 = _mm256_andnot_pd(dummy_mask,r10);
1296 /* Compute parameters for interactions between i and j atoms */
1297 qq10 = _mm256_mul_pd(iq1,jq0);
1299 /* EWALD ELECTROSTATICS */
1301 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1302 ewrt = _mm256_mul_pd(r10,ewtabscale);
1303 ewitab = _mm256_cvttpd_epi32(ewrt);
1304 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1305 ewitab = _mm_slli_epi32(ewitab,2);
1306 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1307 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1308 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1309 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1310 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1311 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1312 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1313 velec = _mm256_mul_pd(qq10,_mm256_sub_pd(rinv10,velec));
1314 felec = _mm256_mul_pd(_mm256_mul_pd(qq10,rinv10),_mm256_sub_pd(rinvsq10,felec));
1316 d = _mm256_sub_pd(r10,rswitch);
1317 d = _mm256_max_pd(d,_mm256_setzero_pd());
1318 d2 = _mm256_mul_pd(d,d);
1319 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)))))));
1321 dsw = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
1323 /* Evaluate switch function */
1324 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1325 felec = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv10,_mm256_mul_pd(velec,dsw)) );
1326 cutoff_mask = _mm256_cmp_pd(rsq10,rcutoff2,_CMP_LT_OQ);
1330 fscal = _mm256_and_pd(fscal,cutoff_mask);
1332 fscal = _mm256_andnot_pd(dummy_mask,fscal);
1334 /* Calculate temporary vectorial force */
1335 tx = _mm256_mul_pd(fscal,dx10);
1336 ty = _mm256_mul_pd(fscal,dy10);
1337 tz = _mm256_mul_pd(fscal,dz10);
1339 /* Update vectorial force */
1340 fix1 = _mm256_add_pd(fix1,tx);
1341 fiy1 = _mm256_add_pd(fiy1,ty);
1342 fiz1 = _mm256_add_pd(fiz1,tz);
1344 fjx0 = _mm256_add_pd(fjx0,tx);
1345 fjy0 = _mm256_add_pd(fjy0,ty);
1346 fjz0 = _mm256_add_pd(fjz0,tz);
1350 /**************************
1351 * CALCULATE INTERACTIONS *
1352 **************************/
1354 if (gmx_mm256_any_lt(rsq20,rcutoff2))
1357 r20 = _mm256_mul_pd(rsq20,rinv20);
1358 r20 = _mm256_andnot_pd(dummy_mask,r20);
1360 /* Compute parameters for interactions between i and j atoms */
1361 qq20 = _mm256_mul_pd(iq2,jq0);
1363 /* EWALD ELECTROSTATICS */
1365 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1366 ewrt = _mm256_mul_pd(r20,ewtabscale);
1367 ewitab = _mm256_cvttpd_epi32(ewrt);
1368 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1369 ewitab = _mm_slli_epi32(ewitab,2);
1370 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1371 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1372 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1373 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1374 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1375 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1376 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1377 velec = _mm256_mul_pd(qq20,_mm256_sub_pd(rinv20,velec));
1378 felec = _mm256_mul_pd(_mm256_mul_pd(qq20,rinv20),_mm256_sub_pd(rinvsq20,felec));
1380 d = _mm256_sub_pd(r20,rswitch);
1381 d = _mm256_max_pd(d,_mm256_setzero_pd());
1382 d2 = _mm256_mul_pd(d,d);
1383 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)))))));
1385 dsw = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
1387 /* Evaluate switch function */
1388 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1389 felec = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv20,_mm256_mul_pd(velec,dsw)) );
1390 cutoff_mask = _mm256_cmp_pd(rsq20,rcutoff2,_CMP_LT_OQ);
1394 fscal = _mm256_and_pd(fscal,cutoff_mask);
1396 fscal = _mm256_andnot_pd(dummy_mask,fscal);
1398 /* Calculate temporary vectorial force */
1399 tx = _mm256_mul_pd(fscal,dx20);
1400 ty = _mm256_mul_pd(fscal,dy20);
1401 tz = _mm256_mul_pd(fscal,dz20);
1403 /* Update vectorial force */
1404 fix2 = _mm256_add_pd(fix2,tx);
1405 fiy2 = _mm256_add_pd(fiy2,ty);
1406 fiz2 = _mm256_add_pd(fiz2,tz);
1408 fjx0 = _mm256_add_pd(fjx0,tx);
1409 fjy0 = _mm256_add_pd(fjy0,ty);
1410 fjz0 = _mm256_add_pd(fjz0,tz);
1414 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1415 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1416 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1417 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1419 gmx_mm256_decrement_1rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,fjx0,fjy0,fjz0);
1421 /* Inner loop uses 192 flops */
1424 /* End of innermost loop */
1426 gmx_mm256_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1427 f+i_coord_offset,fshift+i_shift_offset);
1429 /* Increment number of inner iterations */
1430 inneriter += j_index_end - j_index_start;
1432 /* Outer loop uses 18 flops */
1435 /* Increment number of outer iterations */
1438 /* Update outer/inner flops */
1440 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3_F,outeriter*18 + inneriter*192);