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_128_fma_single kernel generator.
42 #include "../nb_kernel.h"
43 #include "gromacs/legacyheaders/types/simple.h"
44 #include "gromacs/math/vec.h"
45 #include "gromacs/legacyheaders/nrnb.h"
47 #include "gromacs/simd/math_x86_avx_128_fma_single.h"
48 #include "kernelutil_x86_avx_128_fma_single.h"
51 * Gromacs nonbonded kernel: nb_kernel_ElecEwSw_VdwLJSw_GeomW4P1_VF_avx_128_fma_single
52 * Electrostatics interaction: Ewald
53 * VdW interaction: LennardJones
54 * Geometry: Water4-Particle
55 * Calculate force/pot: PotentialAndForce
58 nb_kernel_ElecEwSw_VdwLJSw_GeomW4P1_VF_avx_128_fma_single
59 (t_nblist * gmx_restrict nlist,
60 rvec * gmx_restrict xx,
61 rvec * gmx_restrict ff,
62 t_forcerec * gmx_restrict fr,
63 t_mdatoms * gmx_restrict mdatoms,
64 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
65 t_nrnb * gmx_restrict nrnb)
67 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
68 * just 0 for non-waters.
69 * Suffixes A,B,C,D refer to j loop unrolling done with AVX_128, 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 j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
77 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
79 real *shiftvec,*fshift,*x,*f;
80 real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
82 __m128 fscal,rcutoff,rcutoff2,jidxall;
84 __m128 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
86 __m128 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
88 __m128 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
90 __m128 ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
91 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
92 __m128 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
93 __m128 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
94 __m128 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
95 __m128 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
96 __m128 dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
97 __m128 velec,felec,velecsum,facel,crf,krf,krf2;
100 __m128 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
103 __m128 one_sixth = _mm_set1_ps(1.0/6.0);
104 __m128 one_twelfth = _mm_set1_ps(1.0/12.0);
106 __m128 ewtabscale,eweps,twoeweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
107 __m128 beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
109 __m128 rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
110 real rswitch_scalar,d_scalar;
111 __m128 dummy_mask,cutoff_mask;
112 __m128 signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
113 __m128 one = _mm_set1_ps(1.0);
114 __m128 two = _mm_set1_ps(2.0);
120 jindex = nlist->jindex;
122 shiftidx = nlist->shift;
124 shiftvec = fr->shift_vec[0];
125 fshift = fr->fshift[0];
126 facel = _mm_set1_ps(fr->epsfac);
127 charge = mdatoms->chargeA;
128 nvdwtype = fr->ntype;
130 vdwtype = mdatoms->typeA;
132 sh_ewald = _mm_set1_ps(fr->ic->sh_ewald);
133 beta = _mm_set1_ps(fr->ic->ewaldcoeff_q);
134 beta2 = _mm_mul_ps(beta,beta);
135 beta3 = _mm_mul_ps(beta,beta2);
136 ewtab = fr->ic->tabq_coul_FDV0;
137 ewtabscale = _mm_set1_ps(fr->ic->tabq_scale);
138 ewtabhalfspace = _mm_set1_ps(0.5/fr->ic->tabq_scale);
140 /* Setup water-specific parameters */
141 inr = nlist->iinr[0];
142 iq1 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1]));
143 iq2 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2]));
144 iq3 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+3]));
145 vdwioffset0 = 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 = _mm_set1_ps(rcutoff_scalar);
150 rcutoff2 = _mm_mul_ps(rcutoff,rcutoff);
152 rswitch_scalar = fr->rcoulomb_switch;
153 rswitch = _mm_set1_ps(rswitch_scalar);
154 /* Setup switch parameters */
155 d_scalar = rcutoff_scalar-rswitch_scalar;
156 d = _mm_set1_ps(d_scalar);
157 swV3 = _mm_set1_ps(-10.0/(d_scalar*d_scalar*d_scalar));
158 swV4 = _mm_set1_ps( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
159 swV5 = _mm_set1_ps( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
160 swF2 = _mm_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar));
161 swF3 = _mm_set1_ps( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
162 swF4 = _mm_set1_ps(-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_mm_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
195 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
197 fix0 = _mm_setzero_ps();
198 fiy0 = _mm_setzero_ps();
199 fiz0 = _mm_setzero_ps();
200 fix1 = _mm_setzero_ps();
201 fiy1 = _mm_setzero_ps();
202 fiz1 = _mm_setzero_ps();
203 fix2 = _mm_setzero_ps();
204 fiy2 = _mm_setzero_ps();
205 fiz2 = _mm_setzero_ps();
206 fix3 = _mm_setzero_ps();
207 fiy3 = _mm_setzero_ps();
208 fiz3 = _mm_setzero_ps();
210 /* Reset potential sums */
211 velecsum = _mm_setzero_ps();
212 vvdwsum = _mm_setzero_ps();
214 /* Start inner kernel loop */
215 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
218 /* Get j neighbor index, and coordinate index */
223 j_coord_offsetA = DIM*jnrA;
224 j_coord_offsetB = DIM*jnrB;
225 j_coord_offsetC = DIM*jnrC;
226 j_coord_offsetD = DIM*jnrD;
228 /* load j atom coordinates */
229 gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
230 x+j_coord_offsetC,x+j_coord_offsetD,
233 /* Calculate displacement vector */
234 dx00 = _mm_sub_ps(ix0,jx0);
235 dy00 = _mm_sub_ps(iy0,jy0);
236 dz00 = _mm_sub_ps(iz0,jz0);
237 dx10 = _mm_sub_ps(ix1,jx0);
238 dy10 = _mm_sub_ps(iy1,jy0);
239 dz10 = _mm_sub_ps(iz1,jz0);
240 dx20 = _mm_sub_ps(ix2,jx0);
241 dy20 = _mm_sub_ps(iy2,jy0);
242 dz20 = _mm_sub_ps(iz2,jz0);
243 dx30 = _mm_sub_ps(ix3,jx0);
244 dy30 = _mm_sub_ps(iy3,jy0);
245 dz30 = _mm_sub_ps(iz3,jz0);
247 /* Calculate squared distance and things based on it */
248 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
249 rsq10 = gmx_mm_calc_rsq_ps(dx10,dy10,dz10);
250 rsq20 = gmx_mm_calc_rsq_ps(dx20,dy20,dz20);
251 rsq30 = gmx_mm_calc_rsq_ps(dx30,dy30,dz30);
253 rinv00 = gmx_mm_invsqrt_ps(rsq00);
254 rinv10 = gmx_mm_invsqrt_ps(rsq10);
255 rinv20 = gmx_mm_invsqrt_ps(rsq20);
256 rinv30 = gmx_mm_invsqrt_ps(rsq30);
258 rinvsq00 = _mm_mul_ps(rinv00,rinv00);
259 rinvsq10 = _mm_mul_ps(rinv10,rinv10);
260 rinvsq20 = _mm_mul_ps(rinv20,rinv20);
261 rinvsq30 = _mm_mul_ps(rinv30,rinv30);
263 /* Load parameters for j particles */
264 jq0 = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
265 charge+jnrC+0,charge+jnrD+0);
266 vdwjidx0A = 2*vdwtype[jnrA+0];
267 vdwjidx0B = 2*vdwtype[jnrB+0];
268 vdwjidx0C = 2*vdwtype[jnrC+0];
269 vdwjidx0D = 2*vdwtype[jnrD+0];
271 fjx0 = _mm_setzero_ps();
272 fjy0 = _mm_setzero_ps();
273 fjz0 = _mm_setzero_ps();
275 /**************************
276 * CALCULATE INTERACTIONS *
277 **************************/
279 if (gmx_mm_any_lt(rsq00,rcutoff2))
282 r00 = _mm_mul_ps(rsq00,rinv00);
284 /* Compute parameters for interactions between i and j atoms */
285 gmx_mm_load_4pair_swizzle_ps(vdwparam+vdwioffset0+vdwjidx0A,
286 vdwparam+vdwioffset0+vdwjidx0B,
287 vdwparam+vdwioffset0+vdwjidx0C,
288 vdwparam+vdwioffset0+vdwjidx0D,
291 /* LENNARD-JONES DISPERSION/REPULSION */
293 rinvsix = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00);
294 vvdw6 = _mm_mul_ps(c6_00,rinvsix);
295 vvdw12 = _mm_mul_ps(c12_00,_mm_mul_ps(rinvsix,rinvsix));
296 vvdw = _mm_msub_ps(vvdw12,one_twelfth,_mm_mul_ps(vvdw6,one_sixth));
297 fvdw = _mm_mul_ps(_mm_sub_ps(vvdw12,vvdw6),rinvsq00);
299 d = _mm_sub_ps(r00,rswitch);
300 d = _mm_max_ps(d,_mm_setzero_ps());
301 d2 = _mm_mul_ps(d,d);
302 sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
304 dsw = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
306 /* Evaluate switch function */
307 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
308 fvdw = _mm_msub_ps( fvdw,sw , _mm_mul_ps(rinv00,_mm_mul_ps(vvdw,dsw)) );
309 vvdw = _mm_mul_ps(vvdw,sw);
310 cutoff_mask = _mm_cmplt_ps(rsq00,rcutoff2);
312 /* Update potential sum for this i atom from the interaction with this j atom. */
313 vvdw = _mm_and_ps(vvdw,cutoff_mask);
314 vvdwsum = _mm_add_ps(vvdwsum,vvdw);
318 fscal = _mm_and_ps(fscal,cutoff_mask);
320 /* Update vectorial force */
321 fix0 = _mm_macc_ps(dx00,fscal,fix0);
322 fiy0 = _mm_macc_ps(dy00,fscal,fiy0);
323 fiz0 = _mm_macc_ps(dz00,fscal,fiz0);
325 fjx0 = _mm_macc_ps(dx00,fscal,fjx0);
326 fjy0 = _mm_macc_ps(dy00,fscal,fjy0);
327 fjz0 = _mm_macc_ps(dz00,fscal,fjz0);
331 /**************************
332 * CALCULATE INTERACTIONS *
333 **************************/
335 if (gmx_mm_any_lt(rsq10,rcutoff2))
338 r10 = _mm_mul_ps(rsq10,rinv10);
340 /* Compute parameters for interactions between i and j atoms */
341 qq10 = _mm_mul_ps(iq1,jq0);
343 /* EWALD ELECTROSTATICS */
345 /* Analytical PME correction */
346 zeta2 = _mm_mul_ps(beta2,rsq10);
347 rinv3 = _mm_mul_ps(rinvsq10,rinv10);
348 pmecorrF = gmx_mm_pmecorrF_ps(zeta2);
349 felec = _mm_macc_ps(pmecorrF,beta3,rinv3);
350 felec = _mm_mul_ps(qq10,felec);
351 pmecorrV = gmx_mm_pmecorrV_ps(zeta2);
352 velec = _mm_nmacc_ps(pmecorrV,beta,rinv10);
353 velec = _mm_mul_ps(qq10,velec);
355 d = _mm_sub_ps(r10,rswitch);
356 d = _mm_max_ps(d,_mm_setzero_ps());
357 d2 = _mm_mul_ps(d,d);
358 sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
360 dsw = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
362 /* Evaluate switch function */
363 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
364 felec = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv10,_mm_mul_ps(velec,dsw)) );
365 velec = _mm_mul_ps(velec,sw);
366 cutoff_mask = _mm_cmplt_ps(rsq10,rcutoff2);
368 /* Update potential sum for this i atom from the interaction with this j atom. */
369 velec = _mm_and_ps(velec,cutoff_mask);
370 velecsum = _mm_add_ps(velecsum,velec);
374 fscal = _mm_and_ps(fscal,cutoff_mask);
376 /* Update vectorial force */
377 fix1 = _mm_macc_ps(dx10,fscal,fix1);
378 fiy1 = _mm_macc_ps(dy10,fscal,fiy1);
379 fiz1 = _mm_macc_ps(dz10,fscal,fiz1);
381 fjx0 = _mm_macc_ps(dx10,fscal,fjx0);
382 fjy0 = _mm_macc_ps(dy10,fscal,fjy0);
383 fjz0 = _mm_macc_ps(dz10,fscal,fjz0);
387 /**************************
388 * CALCULATE INTERACTIONS *
389 **************************/
391 if (gmx_mm_any_lt(rsq20,rcutoff2))
394 r20 = _mm_mul_ps(rsq20,rinv20);
396 /* Compute parameters for interactions between i and j atoms */
397 qq20 = _mm_mul_ps(iq2,jq0);
399 /* EWALD ELECTROSTATICS */
401 /* Analytical PME correction */
402 zeta2 = _mm_mul_ps(beta2,rsq20);
403 rinv3 = _mm_mul_ps(rinvsq20,rinv20);
404 pmecorrF = gmx_mm_pmecorrF_ps(zeta2);
405 felec = _mm_macc_ps(pmecorrF,beta3,rinv3);
406 felec = _mm_mul_ps(qq20,felec);
407 pmecorrV = gmx_mm_pmecorrV_ps(zeta2);
408 velec = _mm_nmacc_ps(pmecorrV,beta,rinv20);
409 velec = _mm_mul_ps(qq20,velec);
411 d = _mm_sub_ps(r20,rswitch);
412 d = _mm_max_ps(d,_mm_setzero_ps());
413 d2 = _mm_mul_ps(d,d);
414 sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
416 dsw = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
418 /* Evaluate switch function */
419 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
420 felec = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv20,_mm_mul_ps(velec,dsw)) );
421 velec = _mm_mul_ps(velec,sw);
422 cutoff_mask = _mm_cmplt_ps(rsq20,rcutoff2);
424 /* Update potential sum for this i atom from the interaction with this j atom. */
425 velec = _mm_and_ps(velec,cutoff_mask);
426 velecsum = _mm_add_ps(velecsum,velec);
430 fscal = _mm_and_ps(fscal,cutoff_mask);
432 /* Update vectorial force */
433 fix2 = _mm_macc_ps(dx20,fscal,fix2);
434 fiy2 = _mm_macc_ps(dy20,fscal,fiy2);
435 fiz2 = _mm_macc_ps(dz20,fscal,fiz2);
437 fjx0 = _mm_macc_ps(dx20,fscal,fjx0);
438 fjy0 = _mm_macc_ps(dy20,fscal,fjy0);
439 fjz0 = _mm_macc_ps(dz20,fscal,fjz0);
443 /**************************
444 * CALCULATE INTERACTIONS *
445 **************************/
447 if (gmx_mm_any_lt(rsq30,rcutoff2))
450 r30 = _mm_mul_ps(rsq30,rinv30);
452 /* Compute parameters for interactions between i and j atoms */
453 qq30 = _mm_mul_ps(iq3,jq0);
455 /* EWALD ELECTROSTATICS */
457 /* Analytical PME correction */
458 zeta2 = _mm_mul_ps(beta2,rsq30);
459 rinv3 = _mm_mul_ps(rinvsq30,rinv30);
460 pmecorrF = gmx_mm_pmecorrF_ps(zeta2);
461 felec = _mm_macc_ps(pmecorrF,beta3,rinv3);
462 felec = _mm_mul_ps(qq30,felec);
463 pmecorrV = gmx_mm_pmecorrV_ps(zeta2);
464 velec = _mm_nmacc_ps(pmecorrV,beta,rinv30);
465 velec = _mm_mul_ps(qq30,velec);
467 d = _mm_sub_ps(r30,rswitch);
468 d = _mm_max_ps(d,_mm_setzero_ps());
469 d2 = _mm_mul_ps(d,d);
470 sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
472 dsw = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
474 /* Evaluate switch function */
475 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
476 felec = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv30,_mm_mul_ps(velec,dsw)) );
477 velec = _mm_mul_ps(velec,sw);
478 cutoff_mask = _mm_cmplt_ps(rsq30,rcutoff2);
480 /* Update potential sum for this i atom from the interaction with this j atom. */
481 velec = _mm_and_ps(velec,cutoff_mask);
482 velecsum = _mm_add_ps(velecsum,velec);
486 fscal = _mm_and_ps(fscal,cutoff_mask);
488 /* Update vectorial force */
489 fix3 = _mm_macc_ps(dx30,fscal,fix3);
490 fiy3 = _mm_macc_ps(dy30,fscal,fiy3);
491 fiz3 = _mm_macc_ps(dz30,fscal,fiz3);
493 fjx0 = _mm_macc_ps(dx30,fscal,fjx0);
494 fjy0 = _mm_macc_ps(dy30,fscal,fjy0);
495 fjz0 = _mm_macc_ps(dz30,fscal,fjz0);
499 fjptrA = f+j_coord_offsetA;
500 fjptrB = f+j_coord_offsetB;
501 fjptrC = f+j_coord_offsetC;
502 fjptrD = f+j_coord_offsetD;
504 gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjx0,fjy0,fjz0);
506 /* Inner loop uses 221 flops */
512 /* Get j neighbor index, and coordinate index */
513 jnrlistA = jjnr[jidx];
514 jnrlistB = jjnr[jidx+1];
515 jnrlistC = jjnr[jidx+2];
516 jnrlistD = jjnr[jidx+3];
517 /* Sign of each element will be negative for non-real atoms.
518 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
519 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
521 dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
522 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
523 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
524 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
525 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
526 j_coord_offsetA = DIM*jnrA;
527 j_coord_offsetB = DIM*jnrB;
528 j_coord_offsetC = DIM*jnrC;
529 j_coord_offsetD = DIM*jnrD;
531 /* load j atom coordinates */
532 gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
533 x+j_coord_offsetC,x+j_coord_offsetD,
536 /* Calculate displacement vector */
537 dx00 = _mm_sub_ps(ix0,jx0);
538 dy00 = _mm_sub_ps(iy0,jy0);
539 dz00 = _mm_sub_ps(iz0,jz0);
540 dx10 = _mm_sub_ps(ix1,jx0);
541 dy10 = _mm_sub_ps(iy1,jy0);
542 dz10 = _mm_sub_ps(iz1,jz0);
543 dx20 = _mm_sub_ps(ix2,jx0);
544 dy20 = _mm_sub_ps(iy2,jy0);
545 dz20 = _mm_sub_ps(iz2,jz0);
546 dx30 = _mm_sub_ps(ix3,jx0);
547 dy30 = _mm_sub_ps(iy3,jy0);
548 dz30 = _mm_sub_ps(iz3,jz0);
550 /* Calculate squared distance and things based on it */
551 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
552 rsq10 = gmx_mm_calc_rsq_ps(dx10,dy10,dz10);
553 rsq20 = gmx_mm_calc_rsq_ps(dx20,dy20,dz20);
554 rsq30 = gmx_mm_calc_rsq_ps(dx30,dy30,dz30);
556 rinv00 = gmx_mm_invsqrt_ps(rsq00);
557 rinv10 = gmx_mm_invsqrt_ps(rsq10);
558 rinv20 = gmx_mm_invsqrt_ps(rsq20);
559 rinv30 = gmx_mm_invsqrt_ps(rsq30);
561 rinvsq00 = _mm_mul_ps(rinv00,rinv00);
562 rinvsq10 = _mm_mul_ps(rinv10,rinv10);
563 rinvsq20 = _mm_mul_ps(rinv20,rinv20);
564 rinvsq30 = _mm_mul_ps(rinv30,rinv30);
566 /* Load parameters for j particles */
567 jq0 = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
568 charge+jnrC+0,charge+jnrD+0);
569 vdwjidx0A = 2*vdwtype[jnrA+0];
570 vdwjidx0B = 2*vdwtype[jnrB+0];
571 vdwjidx0C = 2*vdwtype[jnrC+0];
572 vdwjidx0D = 2*vdwtype[jnrD+0];
574 fjx0 = _mm_setzero_ps();
575 fjy0 = _mm_setzero_ps();
576 fjz0 = _mm_setzero_ps();
578 /**************************
579 * CALCULATE INTERACTIONS *
580 **************************/
582 if (gmx_mm_any_lt(rsq00,rcutoff2))
585 r00 = _mm_mul_ps(rsq00,rinv00);
586 r00 = _mm_andnot_ps(dummy_mask,r00);
588 /* Compute parameters for interactions between i and j atoms */
589 gmx_mm_load_4pair_swizzle_ps(vdwparam+vdwioffset0+vdwjidx0A,
590 vdwparam+vdwioffset0+vdwjidx0B,
591 vdwparam+vdwioffset0+vdwjidx0C,
592 vdwparam+vdwioffset0+vdwjidx0D,
595 /* LENNARD-JONES DISPERSION/REPULSION */
597 rinvsix = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00);
598 vvdw6 = _mm_mul_ps(c6_00,rinvsix);
599 vvdw12 = _mm_mul_ps(c12_00,_mm_mul_ps(rinvsix,rinvsix));
600 vvdw = _mm_msub_ps(vvdw12,one_twelfth,_mm_mul_ps(vvdw6,one_sixth));
601 fvdw = _mm_mul_ps(_mm_sub_ps(vvdw12,vvdw6),rinvsq00);
603 d = _mm_sub_ps(r00,rswitch);
604 d = _mm_max_ps(d,_mm_setzero_ps());
605 d2 = _mm_mul_ps(d,d);
606 sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
608 dsw = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
610 /* Evaluate switch function */
611 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
612 fvdw = _mm_msub_ps( fvdw,sw , _mm_mul_ps(rinv00,_mm_mul_ps(vvdw,dsw)) );
613 vvdw = _mm_mul_ps(vvdw,sw);
614 cutoff_mask = _mm_cmplt_ps(rsq00,rcutoff2);
616 /* Update potential sum for this i atom from the interaction with this j atom. */
617 vvdw = _mm_and_ps(vvdw,cutoff_mask);
618 vvdw = _mm_andnot_ps(dummy_mask,vvdw);
619 vvdwsum = _mm_add_ps(vvdwsum,vvdw);
623 fscal = _mm_and_ps(fscal,cutoff_mask);
625 fscal = _mm_andnot_ps(dummy_mask,fscal);
627 /* Update vectorial force */
628 fix0 = _mm_macc_ps(dx00,fscal,fix0);
629 fiy0 = _mm_macc_ps(dy00,fscal,fiy0);
630 fiz0 = _mm_macc_ps(dz00,fscal,fiz0);
632 fjx0 = _mm_macc_ps(dx00,fscal,fjx0);
633 fjy0 = _mm_macc_ps(dy00,fscal,fjy0);
634 fjz0 = _mm_macc_ps(dz00,fscal,fjz0);
638 /**************************
639 * CALCULATE INTERACTIONS *
640 **************************/
642 if (gmx_mm_any_lt(rsq10,rcutoff2))
645 r10 = _mm_mul_ps(rsq10,rinv10);
646 r10 = _mm_andnot_ps(dummy_mask,r10);
648 /* Compute parameters for interactions between i and j atoms */
649 qq10 = _mm_mul_ps(iq1,jq0);
651 /* EWALD ELECTROSTATICS */
653 /* Analytical PME correction */
654 zeta2 = _mm_mul_ps(beta2,rsq10);
655 rinv3 = _mm_mul_ps(rinvsq10,rinv10);
656 pmecorrF = gmx_mm_pmecorrF_ps(zeta2);
657 felec = _mm_macc_ps(pmecorrF,beta3,rinv3);
658 felec = _mm_mul_ps(qq10,felec);
659 pmecorrV = gmx_mm_pmecorrV_ps(zeta2);
660 velec = _mm_nmacc_ps(pmecorrV,beta,rinv10);
661 velec = _mm_mul_ps(qq10,velec);
663 d = _mm_sub_ps(r10,rswitch);
664 d = _mm_max_ps(d,_mm_setzero_ps());
665 d2 = _mm_mul_ps(d,d);
666 sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
668 dsw = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
670 /* Evaluate switch function */
671 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
672 felec = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv10,_mm_mul_ps(velec,dsw)) );
673 velec = _mm_mul_ps(velec,sw);
674 cutoff_mask = _mm_cmplt_ps(rsq10,rcutoff2);
676 /* Update potential sum for this i atom from the interaction with this j atom. */
677 velec = _mm_and_ps(velec,cutoff_mask);
678 velec = _mm_andnot_ps(dummy_mask,velec);
679 velecsum = _mm_add_ps(velecsum,velec);
683 fscal = _mm_and_ps(fscal,cutoff_mask);
685 fscal = _mm_andnot_ps(dummy_mask,fscal);
687 /* Update vectorial force */
688 fix1 = _mm_macc_ps(dx10,fscal,fix1);
689 fiy1 = _mm_macc_ps(dy10,fscal,fiy1);
690 fiz1 = _mm_macc_ps(dz10,fscal,fiz1);
692 fjx0 = _mm_macc_ps(dx10,fscal,fjx0);
693 fjy0 = _mm_macc_ps(dy10,fscal,fjy0);
694 fjz0 = _mm_macc_ps(dz10,fscal,fjz0);
698 /**************************
699 * CALCULATE INTERACTIONS *
700 **************************/
702 if (gmx_mm_any_lt(rsq20,rcutoff2))
705 r20 = _mm_mul_ps(rsq20,rinv20);
706 r20 = _mm_andnot_ps(dummy_mask,r20);
708 /* Compute parameters for interactions between i and j atoms */
709 qq20 = _mm_mul_ps(iq2,jq0);
711 /* EWALD ELECTROSTATICS */
713 /* Analytical PME correction */
714 zeta2 = _mm_mul_ps(beta2,rsq20);
715 rinv3 = _mm_mul_ps(rinvsq20,rinv20);
716 pmecorrF = gmx_mm_pmecorrF_ps(zeta2);
717 felec = _mm_macc_ps(pmecorrF,beta3,rinv3);
718 felec = _mm_mul_ps(qq20,felec);
719 pmecorrV = gmx_mm_pmecorrV_ps(zeta2);
720 velec = _mm_nmacc_ps(pmecorrV,beta,rinv20);
721 velec = _mm_mul_ps(qq20,velec);
723 d = _mm_sub_ps(r20,rswitch);
724 d = _mm_max_ps(d,_mm_setzero_ps());
725 d2 = _mm_mul_ps(d,d);
726 sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
728 dsw = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
730 /* Evaluate switch function */
731 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
732 felec = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv20,_mm_mul_ps(velec,dsw)) );
733 velec = _mm_mul_ps(velec,sw);
734 cutoff_mask = _mm_cmplt_ps(rsq20,rcutoff2);
736 /* Update potential sum for this i atom from the interaction with this j atom. */
737 velec = _mm_and_ps(velec,cutoff_mask);
738 velec = _mm_andnot_ps(dummy_mask,velec);
739 velecsum = _mm_add_ps(velecsum,velec);
743 fscal = _mm_and_ps(fscal,cutoff_mask);
745 fscal = _mm_andnot_ps(dummy_mask,fscal);
747 /* Update vectorial force */
748 fix2 = _mm_macc_ps(dx20,fscal,fix2);
749 fiy2 = _mm_macc_ps(dy20,fscal,fiy2);
750 fiz2 = _mm_macc_ps(dz20,fscal,fiz2);
752 fjx0 = _mm_macc_ps(dx20,fscal,fjx0);
753 fjy0 = _mm_macc_ps(dy20,fscal,fjy0);
754 fjz0 = _mm_macc_ps(dz20,fscal,fjz0);
758 /**************************
759 * CALCULATE INTERACTIONS *
760 **************************/
762 if (gmx_mm_any_lt(rsq30,rcutoff2))
765 r30 = _mm_mul_ps(rsq30,rinv30);
766 r30 = _mm_andnot_ps(dummy_mask,r30);
768 /* Compute parameters for interactions between i and j atoms */
769 qq30 = _mm_mul_ps(iq3,jq0);
771 /* EWALD ELECTROSTATICS */
773 /* Analytical PME correction */
774 zeta2 = _mm_mul_ps(beta2,rsq30);
775 rinv3 = _mm_mul_ps(rinvsq30,rinv30);
776 pmecorrF = gmx_mm_pmecorrF_ps(zeta2);
777 felec = _mm_macc_ps(pmecorrF,beta3,rinv3);
778 felec = _mm_mul_ps(qq30,felec);
779 pmecorrV = gmx_mm_pmecorrV_ps(zeta2);
780 velec = _mm_nmacc_ps(pmecorrV,beta,rinv30);
781 velec = _mm_mul_ps(qq30,velec);
783 d = _mm_sub_ps(r30,rswitch);
784 d = _mm_max_ps(d,_mm_setzero_ps());
785 d2 = _mm_mul_ps(d,d);
786 sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
788 dsw = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
790 /* Evaluate switch function */
791 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
792 felec = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv30,_mm_mul_ps(velec,dsw)) );
793 velec = _mm_mul_ps(velec,sw);
794 cutoff_mask = _mm_cmplt_ps(rsq30,rcutoff2);
796 /* Update potential sum for this i atom from the interaction with this j atom. */
797 velec = _mm_and_ps(velec,cutoff_mask);
798 velec = _mm_andnot_ps(dummy_mask,velec);
799 velecsum = _mm_add_ps(velecsum,velec);
803 fscal = _mm_and_ps(fscal,cutoff_mask);
805 fscal = _mm_andnot_ps(dummy_mask,fscal);
807 /* Update vectorial force */
808 fix3 = _mm_macc_ps(dx30,fscal,fix3);
809 fiy3 = _mm_macc_ps(dy30,fscal,fiy3);
810 fiz3 = _mm_macc_ps(dz30,fscal,fiz3);
812 fjx0 = _mm_macc_ps(dx30,fscal,fjx0);
813 fjy0 = _mm_macc_ps(dy30,fscal,fjy0);
814 fjz0 = _mm_macc_ps(dz30,fscal,fjz0);
818 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
819 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
820 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
821 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
823 gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjx0,fjy0,fjz0);
825 /* Inner loop uses 225 flops */
828 /* End of innermost loop */
830 gmx_mm_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
831 f+i_coord_offset,fshift+i_shift_offset);
834 /* Update potential energies */
835 gmx_mm_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
836 gmx_mm_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
838 /* Increment number of inner iterations */
839 inneriter += j_index_end - j_index_start;
841 /* Outer loop uses 26 flops */
844 /* Increment number of outer iterations */
847 /* Update outer/inner flops */
849 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_VF,outeriter*26 + inneriter*225);
852 * Gromacs nonbonded kernel: nb_kernel_ElecEwSw_VdwLJSw_GeomW4P1_F_avx_128_fma_single
853 * Electrostatics interaction: Ewald
854 * VdW interaction: LennardJones
855 * Geometry: Water4-Particle
856 * Calculate force/pot: Force
859 nb_kernel_ElecEwSw_VdwLJSw_GeomW4P1_F_avx_128_fma_single
860 (t_nblist * gmx_restrict nlist,
861 rvec * gmx_restrict xx,
862 rvec * gmx_restrict ff,
863 t_forcerec * gmx_restrict fr,
864 t_mdatoms * gmx_restrict mdatoms,
865 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
866 t_nrnb * gmx_restrict nrnb)
868 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
869 * just 0 for non-waters.
870 * Suffixes A,B,C,D refer to j loop unrolling done with AVX_128, e.g. for the four different
871 * jnr indices corresponding to data put in the four positions in the SIMD register.
873 int i_shift_offset,i_coord_offset,outeriter,inneriter;
874 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
875 int jnrA,jnrB,jnrC,jnrD;
876 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
877 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
878 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
880 real *shiftvec,*fshift,*x,*f;
881 real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
883 __m128 fscal,rcutoff,rcutoff2,jidxall;
885 __m128 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
887 __m128 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
889 __m128 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
891 __m128 ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
892 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
893 __m128 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
894 __m128 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
895 __m128 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
896 __m128 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
897 __m128 dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
898 __m128 velec,felec,velecsum,facel,crf,krf,krf2;
901 __m128 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
904 __m128 one_sixth = _mm_set1_ps(1.0/6.0);
905 __m128 one_twelfth = _mm_set1_ps(1.0/12.0);
907 __m128 ewtabscale,eweps,twoeweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
908 __m128 beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
910 __m128 rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
911 real rswitch_scalar,d_scalar;
912 __m128 dummy_mask,cutoff_mask;
913 __m128 signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
914 __m128 one = _mm_set1_ps(1.0);
915 __m128 two = _mm_set1_ps(2.0);
921 jindex = nlist->jindex;
923 shiftidx = nlist->shift;
925 shiftvec = fr->shift_vec[0];
926 fshift = fr->fshift[0];
927 facel = _mm_set1_ps(fr->epsfac);
928 charge = mdatoms->chargeA;
929 nvdwtype = fr->ntype;
931 vdwtype = mdatoms->typeA;
933 sh_ewald = _mm_set1_ps(fr->ic->sh_ewald);
934 beta = _mm_set1_ps(fr->ic->ewaldcoeff_q);
935 beta2 = _mm_mul_ps(beta,beta);
936 beta3 = _mm_mul_ps(beta,beta2);
937 ewtab = fr->ic->tabq_coul_FDV0;
938 ewtabscale = _mm_set1_ps(fr->ic->tabq_scale);
939 ewtabhalfspace = _mm_set1_ps(0.5/fr->ic->tabq_scale);
941 /* Setup water-specific parameters */
942 inr = nlist->iinr[0];
943 iq1 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1]));
944 iq2 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2]));
945 iq3 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+3]));
946 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
948 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
949 rcutoff_scalar = fr->rcoulomb;
950 rcutoff = _mm_set1_ps(rcutoff_scalar);
951 rcutoff2 = _mm_mul_ps(rcutoff,rcutoff);
953 rswitch_scalar = fr->rcoulomb_switch;
954 rswitch = _mm_set1_ps(rswitch_scalar);
955 /* Setup switch parameters */
956 d_scalar = rcutoff_scalar-rswitch_scalar;
957 d = _mm_set1_ps(d_scalar);
958 swV3 = _mm_set1_ps(-10.0/(d_scalar*d_scalar*d_scalar));
959 swV4 = _mm_set1_ps( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
960 swV5 = _mm_set1_ps( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
961 swF2 = _mm_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar));
962 swF3 = _mm_set1_ps( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
963 swF4 = _mm_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
965 /* Avoid stupid compiler warnings */
966 jnrA = jnrB = jnrC = jnrD = 0;
975 for(iidx=0;iidx<4*DIM;iidx++)
980 /* Start outer loop over neighborlists */
981 for(iidx=0; iidx<nri; iidx++)
983 /* Load shift vector for this list */
984 i_shift_offset = DIM*shiftidx[iidx];
986 /* Load limits for loop over neighbors */
987 j_index_start = jindex[iidx];
988 j_index_end = jindex[iidx+1];
990 /* Get outer coordinate index */
992 i_coord_offset = DIM*inr;
994 /* Load i particle coords and add shift vector */
995 gmx_mm_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
996 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
998 fix0 = _mm_setzero_ps();
999 fiy0 = _mm_setzero_ps();
1000 fiz0 = _mm_setzero_ps();
1001 fix1 = _mm_setzero_ps();
1002 fiy1 = _mm_setzero_ps();
1003 fiz1 = _mm_setzero_ps();
1004 fix2 = _mm_setzero_ps();
1005 fiy2 = _mm_setzero_ps();
1006 fiz2 = _mm_setzero_ps();
1007 fix3 = _mm_setzero_ps();
1008 fiy3 = _mm_setzero_ps();
1009 fiz3 = _mm_setzero_ps();
1011 /* Start inner kernel loop */
1012 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
1015 /* Get j neighbor index, and coordinate index */
1017 jnrB = jjnr[jidx+1];
1018 jnrC = jjnr[jidx+2];
1019 jnrD = jjnr[jidx+3];
1020 j_coord_offsetA = DIM*jnrA;
1021 j_coord_offsetB = DIM*jnrB;
1022 j_coord_offsetC = DIM*jnrC;
1023 j_coord_offsetD = DIM*jnrD;
1025 /* load j atom coordinates */
1026 gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1027 x+j_coord_offsetC,x+j_coord_offsetD,
1030 /* Calculate displacement vector */
1031 dx00 = _mm_sub_ps(ix0,jx0);
1032 dy00 = _mm_sub_ps(iy0,jy0);
1033 dz00 = _mm_sub_ps(iz0,jz0);
1034 dx10 = _mm_sub_ps(ix1,jx0);
1035 dy10 = _mm_sub_ps(iy1,jy0);
1036 dz10 = _mm_sub_ps(iz1,jz0);
1037 dx20 = _mm_sub_ps(ix2,jx0);
1038 dy20 = _mm_sub_ps(iy2,jy0);
1039 dz20 = _mm_sub_ps(iz2,jz0);
1040 dx30 = _mm_sub_ps(ix3,jx0);
1041 dy30 = _mm_sub_ps(iy3,jy0);
1042 dz30 = _mm_sub_ps(iz3,jz0);
1044 /* Calculate squared distance and things based on it */
1045 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
1046 rsq10 = gmx_mm_calc_rsq_ps(dx10,dy10,dz10);
1047 rsq20 = gmx_mm_calc_rsq_ps(dx20,dy20,dz20);
1048 rsq30 = gmx_mm_calc_rsq_ps(dx30,dy30,dz30);
1050 rinv00 = gmx_mm_invsqrt_ps(rsq00);
1051 rinv10 = gmx_mm_invsqrt_ps(rsq10);
1052 rinv20 = gmx_mm_invsqrt_ps(rsq20);
1053 rinv30 = gmx_mm_invsqrt_ps(rsq30);
1055 rinvsq00 = _mm_mul_ps(rinv00,rinv00);
1056 rinvsq10 = _mm_mul_ps(rinv10,rinv10);
1057 rinvsq20 = _mm_mul_ps(rinv20,rinv20);
1058 rinvsq30 = _mm_mul_ps(rinv30,rinv30);
1060 /* Load parameters for j particles */
1061 jq0 = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
1062 charge+jnrC+0,charge+jnrD+0);
1063 vdwjidx0A = 2*vdwtype[jnrA+0];
1064 vdwjidx0B = 2*vdwtype[jnrB+0];
1065 vdwjidx0C = 2*vdwtype[jnrC+0];
1066 vdwjidx0D = 2*vdwtype[jnrD+0];
1068 fjx0 = _mm_setzero_ps();
1069 fjy0 = _mm_setzero_ps();
1070 fjz0 = _mm_setzero_ps();
1072 /**************************
1073 * CALCULATE INTERACTIONS *
1074 **************************/
1076 if (gmx_mm_any_lt(rsq00,rcutoff2))
1079 r00 = _mm_mul_ps(rsq00,rinv00);
1081 /* Compute parameters for interactions between i and j atoms */
1082 gmx_mm_load_4pair_swizzle_ps(vdwparam+vdwioffset0+vdwjidx0A,
1083 vdwparam+vdwioffset0+vdwjidx0B,
1084 vdwparam+vdwioffset0+vdwjidx0C,
1085 vdwparam+vdwioffset0+vdwjidx0D,
1088 /* LENNARD-JONES DISPERSION/REPULSION */
1090 rinvsix = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00);
1091 vvdw6 = _mm_mul_ps(c6_00,rinvsix);
1092 vvdw12 = _mm_mul_ps(c12_00,_mm_mul_ps(rinvsix,rinvsix));
1093 vvdw = _mm_msub_ps(vvdw12,one_twelfth,_mm_mul_ps(vvdw6,one_sixth));
1094 fvdw = _mm_mul_ps(_mm_sub_ps(vvdw12,vvdw6),rinvsq00);
1096 d = _mm_sub_ps(r00,rswitch);
1097 d = _mm_max_ps(d,_mm_setzero_ps());
1098 d2 = _mm_mul_ps(d,d);
1099 sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
1101 dsw = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
1103 /* Evaluate switch function */
1104 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1105 fvdw = _mm_msub_ps( fvdw,sw , _mm_mul_ps(rinv00,_mm_mul_ps(vvdw,dsw)) );
1106 cutoff_mask = _mm_cmplt_ps(rsq00,rcutoff2);
1110 fscal = _mm_and_ps(fscal,cutoff_mask);
1112 /* Update vectorial force */
1113 fix0 = _mm_macc_ps(dx00,fscal,fix0);
1114 fiy0 = _mm_macc_ps(dy00,fscal,fiy0);
1115 fiz0 = _mm_macc_ps(dz00,fscal,fiz0);
1117 fjx0 = _mm_macc_ps(dx00,fscal,fjx0);
1118 fjy0 = _mm_macc_ps(dy00,fscal,fjy0);
1119 fjz0 = _mm_macc_ps(dz00,fscal,fjz0);
1123 /**************************
1124 * CALCULATE INTERACTIONS *
1125 **************************/
1127 if (gmx_mm_any_lt(rsq10,rcutoff2))
1130 r10 = _mm_mul_ps(rsq10,rinv10);
1132 /* Compute parameters for interactions between i and j atoms */
1133 qq10 = _mm_mul_ps(iq1,jq0);
1135 /* EWALD ELECTROSTATICS */
1137 /* Analytical PME correction */
1138 zeta2 = _mm_mul_ps(beta2,rsq10);
1139 rinv3 = _mm_mul_ps(rinvsq10,rinv10);
1140 pmecorrF = gmx_mm_pmecorrF_ps(zeta2);
1141 felec = _mm_macc_ps(pmecorrF,beta3,rinv3);
1142 felec = _mm_mul_ps(qq10,felec);
1143 pmecorrV = gmx_mm_pmecorrV_ps(zeta2);
1144 velec = _mm_nmacc_ps(pmecorrV,beta,rinv10);
1145 velec = _mm_mul_ps(qq10,velec);
1147 d = _mm_sub_ps(r10,rswitch);
1148 d = _mm_max_ps(d,_mm_setzero_ps());
1149 d2 = _mm_mul_ps(d,d);
1150 sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
1152 dsw = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
1154 /* Evaluate switch function */
1155 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1156 felec = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv10,_mm_mul_ps(velec,dsw)) );
1157 cutoff_mask = _mm_cmplt_ps(rsq10,rcutoff2);
1161 fscal = _mm_and_ps(fscal,cutoff_mask);
1163 /* Update vectorial force */
1164 fix1 = _mm_macc_ps(dx10,fscal,fix1);
1165 fiy1 = _mm_macc_ps(dy10,fscal,fiy1);
1166 fiz1 = _mm_macc_ps(dz10,fscal,fiz1);
1168 fjx0 = _mm_macc_ps(dx10,fscal,fjx0);
1169 fjy0 = _mm_macc_ps(dy10,fscal,fjy0);
1170 fjz0 = _mm_macc_ps(dz10,fscal,fjz0);
1174 /**************************
1175 * CALCULATE INTERACTIONS *
1176 **************************/
1178 if (gmx_mm_any_lt(rsq20,rcutoff2))
1181 r20 = _mm_mul_ps(rsq20,rinv20);
1183 /* Compute parameters for interactions between i and j atoms */
1184 qq20 = _mm_mul_ps(iq2,jq0);
1186 /* EWALD ELECTROSTATICS */
1188 /* Analytical PME correction */
1189 zeta2 = _mm_mul_ps(beta2,rsq20);
1190 rinv3 = _mm_mul_ps(rinvsq20,rinv20);
1191 pmecorrF = gmx_mm_pmecorrF_ps(zeta2);
1192 felec = _mm_macc_ps(pmecorrF,beta3,rinv3);
1193 felec = _mm_mul_ps(qq20,felec);
1194 pmecorrV = gmx_mm_pmecorrV_ps(zeta2);
1195 velec = _mm_nmacc_ps(pmecorrV,beta,rinv20);
1196 velec = _mm_mul_ps(qq20,velec);
1198 d = _mm_sub_ps(r20,rswitch);
1199 d = _mm_max_ps(d,_mm_setzero_ps());
1200 d2 = _mm_mul_ps(d,d);
1201 sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
1203 dsw = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
1205 /* Evaluate switch function */
1206 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1207 felec = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv20,_mm_mul_ps(velec,dsw)) );
1208 cutoff_mask = _mm_cmplt_ps(rsq20,rcutoff2);
1212 fscal = _mm_and_ps(fscal,cutoff_mask);
1214 /* Update vectorial force */
1215 fix2 = _mm_macc_ps(dx20,fscal,fix2);
1216 fiy2 = _mm_macc_ps(dy20,fscal,fiy2);
1217 fiz2 = _mm_macc_ps(dz20,fscal,fiz2);
1219 fjx0 = _mm_macc_ps(dx20,fscal,fjx0);
1220 fjy0 = _mm_macc_ps(dy20,fscal,fjy0);
1221 fjz0 = _mm_macc_ps(dz20,fscal,fjz0);
1225 /**************************
1226 * CALCULATE INTERACTIONS *
1227 **************************/
1229 if (gmx_mm_any_lt(rsq30,rcutoff2))
1232 r30 = _mm_mul_ps(rsq30,rinv30);
1234 /* Compute parameters for interactions between i and j atoms */
1235 qq30 = _mm_mul_ps(iq3,jq0);
1237 /* EWALD ELECTROSTATICS */
1239 /* Analytical PME correction */
1240 zeta2 = _mm_mul_ps(beta2,rsq30);
1241 rinv3 = _mm_mul_ps(rinvsq30,rinv30);
1242 pmecorrF = gmx_mm_pmecorrF_ps(zeta2);
1243 felec = _mm_macc_ps(pmecorrF,beta3,rinv3);
1244 felec = _mm_mul_ps(qq30,felec);
1245 pmecorrV = gmx_mm_pmecorrV_ps(zeta2);
1246 velec = _mm_nmacc_ps(pmecorrV,beta,rinv30);
1247 velec = _mm_mul_ps(qq30,velec);
1249 d = _mm_sub_ps(r30,rswitch);
1250 d = _mm_max_ps(d,_mm_setzero_ps());
1251 d2 = _mm_mul_ps(d,d);
1252 sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
1254 dsw = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
1256 /* Evaluate switch function */
1257 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1258 felec = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv30,_mm_mul_ps(velec,dsw)) );
1259 cutoff_mask = _mm_cmplt_ps(rsq30,rcutoff2);
1263 fscal = _mm_and_ps(fscal,cutoff_mask);
1265 /* Update vectorial force */
1266 fix3 = _mm_macc_ps(dx30,fscal,fix3);
1267 fiy3 = _mm_macc_ps(dy30,fscal,fiy3);
1268 fiz3 = _mm_macc_ps(dz30,fscal,fiz3);
1270 fjx0 = _mm_macc_ps(dx30,fscal,fjx0);
1271 fjy0 = _mm_macc_ps(dy30,fscal,fjy0);
1272 fjz0 = _mm_macc_ps(dz30,fscal,fjz0);
1276 fjptrA = f+j_coord_offsetA;
1277 fjptrB = f+j_coord_offsetB;
1278 fjptrC = f+j_coord_offsetC;
1279 fjptrD = f+j_coord_offsetD;
1281 gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjx0,fjy0,fjz0);
1283 /* Inner loop uses 209 flops */
1286 if(jidx<j_index_end)
1289 /* Get j neighbor index, and coordinate index */
1290 jnrlistA = jjnr[jidx];
1291 jnrlistB = jjnr[jidx+1];
1292 jnrlistC = jjnr[jidx+2];
1293 jnrlistD = jjnr[jidx+3];
1294 /* Sign of each element will be negative for non-real atoms.
1295 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1296 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1298 dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
1299 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
1300 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
1301 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
1302 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
1303 j_coord_offsetA = DIM*jnrA;
1304 j_coord_offsetB = DIM*jnrB;
1305 j_coord_offsetC = DIM*jnrC;
1306 j_coord_offsetD = DIM*jnrD;
1308 /* load j atom coordinates */
1309 gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1310 x+j_coord_offsetC,x+j_coord_offsetD,
1313 /* Calculate displacement vector */
1314 dx00 = _mm_sub_ps(ix0,jx0);
1315 dy00 = _mm_sub_ps(iy0,jy0);
1316 dz00 = _mm_sub_ps(iz0,jz0);
1317 dx10 = _mm_sub_ps(ix1,jx0);
1318 dy10 = _mm_sub_ps(iy1,jy0);
1319 dz10 = _mm_sub_ps(iz1,jz0);
1320 dx20 = _mm_sub_ps(ix2,jx0);
1321 dy20 = _mm_sub_ps(iy2,jy0);
1322 dz20 = _mm_sub_ps(iz2,jz0);
1323 dx30 = _mm_sub_ps(ix3,jx0);
1324 dy30 = _mm_sub_ps(iy3,jy0);
1325 dz30 = _mm_sub_ps(iz3,jz0);
1327 /* Calculate squared distance and things based on it */
1328 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
1329 rsq10 = gmx_mm_calc_rsq_ps(dx10,dy10,dz10);
1330 rsq20 = gmx_mm_calc_rsq_ps(dx20,dy20,dz20);
1331 rsq30 = gmx_mm_calc_rsq_ps(dx30,dy30,dz30);
1333 rinv00 = gmx_mm_invsqrt_ps(rsq00);
1334 rinv10 = gmx_mm_invsqrt_ps(rsq10);
1335 rinv20 = gmx_mm_invsqrt_ps(rsq20);
1336 rinv30 = gmx_mm_invsqrt_ps(rsq30);
1338 rinvsq00 = _mm_mul_ps(rinv00,rinv00);
1339 rinvsq10 = _mm_mul_ps(rinv10,rinv10);
1340 rinvsq20 = _mm_mul_ps(rinv20,rinv20);
1341 rinvsq30 = _mm_mul_ps(rinv30,rinv30);
1343 /* Load parameters for j particles */
1344 jq0 = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
1345 charge+jnrC+0,charge+jnrD+0);
1346 vdwjidx0A = 2*vdwtype[jnrA+0];
1347 vdwjidx0B = 2*vdwtype[jnrB+0];
1348 vdwjidx0C = 2*vdwtype[jnrC+0];
1349 vdwjidx0D = 2*vdwtype[jnrD+0];
1351 fjx0 = _mm_setzero_ps();
1352 fjy0 = _mm_setzero_ps();
1353 fjz0 = _mm_setzero_ps();
1355 /**************************
1356 * CALCULATE INTERACTIONS *
1357 **************************/
1359 if (gmx_mm_any_lt(rsq00,rcutoff2))
1362 r00 = _mm_mul_ps(rsq00,rinv00);
1363 r00 = _mm_andnot_ps(dummy_mask,r00);
1365 /* Compute parameters for interactions between i and j atoms */
1366 gmx_mm_load_4pair_swizzle_ps(vdwparam+vdwioffset0+vdwjidx0A,
1367 vdwparam+vdwioffset0+vdwjidx0B,
1368 vdwparam+vdwioffset0+vdwjidx0C,
1369 vdwparam+vdwioffset0+vdwjidx0D,
1372 /* LENNARD-JONES DISPERSION/REPULSION */
1374 rinvsix = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00);
1375 vvdw6 = _mm_mul_ps(c6_00,rinvsix);
1376 vvdw12 = _mm_mul_ps(c12_00,_mm_mul_ps(rinvsix,rinvsix));
1377 vvdw = _mm_msub_ps(vvdw12,one_twelfth,_mm_mul_ps(vvdw6,one_sixth));
1378 fvdw = _mm_mul_ps(_mm_sub_ps(vvdw12,vvdw6),rinvsq00);
1380 d = _mm_sub_ps(r00,rswitch);
1381 d = _mm_max_ps(d,_mm_setzero_ps());
1382 d2 = _mm_mul_ps(d,d);
1383 sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
1385 dsw = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
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 fvdw = _mm_msub_ps( fvdw,sw , _mm_mul_ps(rinv00,_mm_mul_ps(vvdw,dsw)) );
1390 cutoff_mask = _mm_cmplt_ps(rsq00,rcutoff2);
1394 fscal = _mm_and_ps(fscal,cutoff_mask);
1396 fscal = _mm_andnot_ps(dummy_mask,fscal);
1398 /* Update vectorial force */
1399 fix0 = _mm_macc_ps(dx00,fscal,fix0);
1400 fiy0 = _mm_macc_ps(dy00,fscal,fiy0);
1401 fiz0 = _mm_macc_ps(dz00,fscal,fiz0);
1403 fjx0 = _mm_macc_ps(dx00,fscal,fjx0);
1404 fjy0 = _mm_macc_ps(dy00,fscal,fjy0);
1405 fjz0 = _mm_macc_ps(dz00,fscal,fjz0);
1409 /**************************
1410 * CALCULATE INTERACTIONS *
1411 **************************/
1413 if (gmx_mm_any_lt(rsq10,rcutoff2))
1416 r10 = _mm_mul_ps(rsq10,rinv10);
1417 r10 = _mm_andnot_ps(dummy_mask,r10);
1419 /* Compute parameters for interactions between i and j atoms */
1420 qq10 = _mm_mul_ps(iq1,jq0);
1422 /* EWALD ELECTROSTATICS */
1424 /* Analytical PME correction */
1425 zeta2 = _mm_mul_ps(beta2,rsq10);
1426 rinv3 = _mm_mul_ps(rinvsq10,rinv10);
1427 pmecorrF = gmx_mm_pmecorrF_ps(zeta2);
1428 felec = _mm_macc_ps(pmecorrF,beta3,rinv3);
1429 felec = _mm_mul_ps(qq10,felec);
1430 pmecorrV = gmx_mm_pmecorrV_ps(zeta2);
1431 velec = _mm_nmacc_ps(pmecorrV,beta,rinv10);
1432 velec = _mm_mul_ps(qq10,velec);
1434 d = _mm_sub_ps(r10,rswitch);
1435 d = _mm_max_ps(d,_mm_setzero_ps());
1436 d2 = _mm_mul_ps(d,d);
1437 sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
1439 dsw = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
1441 /* Evaluate switch function */
1442 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1443 felec = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv10,_mm_mul_ps(velec,dsw)) );
1444 cutoff_mask = _mm_cmplt_ps(rsq10,rcutoff2);
1448 fscal = _mm_and_ps(fscal,cutoff_mask);
1450 fscal = _mm_andnot_ps(dummy_mask,fscal);
1452 /* Update vectorial force */
1453 fix1 = _mm_macc_ps(dx10,fscal,fix1);
1454 fiy1 = _mm_macc_ps(dy10,fscal,fiy1);
1455 fiz1 = _mm_macc_ps(dz10,fscal,fiz1);
1457 fjx0 = _mm_macc_ps(dx10,fscal,fjx0);
1458 fjy0 = _mm_macc_ps(dy10,fscal,fjy0);
1459 fjz0 = _mm_macc_ps(dz10,fscal,fjz0);
1463 /**************************
1464 * CALCULATE INTERACTIONS *
1465 **************************/
1467 if (gmx_mm_any_lt(rsq20,rcutoff2))
1470 r20 = _mm_mul_ps(rsq20,rinv20);
1471 r20 = _mm_andnot_ps(dummy_mask,r20);
1473 /* Compute parameters for interactions between i and j atoms */
1474 qq20 = _mm_mul_ps(iq2,jq0);
1476 /* EWALD ELECTROSTATICS */
1478 /* Analytical PME correction */
1479 zeta2 = _mm_mul_ps(beta2,rsq20);
1480 rinv3 = _mm_mul_ps(rinvsq20,rinv20);
1481 pmecorrF = gmx_mm_pmecorrF_ps(zeta2);
1482 felec = _mm_macc_ps(pmecorrF,beta3,rinv3);
1483 felec = _mm_mul_ps(qq20,felec);
1484 pmecorrV = gmx_mm_pmecorrV_ps(zeta2);
1485 velec = _mm_nmacc_ps(pmecorrV,beta,rinv20);
1486 velec = _mm_mul_ps(qq20,velec);
1488 d = _mm_sub_ps(r20,rswitch);
1489 d = _mm_max_ps(d,_mm_setzero_ps());
1490 d2 = _mm_mul_ps(d,d);
1491 sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
1493 dsw = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
1495 /* Evaluate switch function */
1496 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1497 felec = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv20,_mm_mul_ps(velec,dsw)) );
1498 cutoff_mask = _mm_cmplt_ps(rsq20,rcutoff2);
1502 fscal = _mm_and_ps(fscal,cutoff_mask);
1504 fscal = _mm_andnot_ps(dummy_mask,fscal);
1506 /* Update vectorial force */
1507 fix2 = _mm_macc_ps(dx20,fscal,fix2);
1508 fiy2 = _mm_macc_ps(dy20,fscal,fiy2);
1509 fiz2 = _mm_macc_ps(dz20,fscal,fiz2);
1511 fjx0 = _mm_macc_ps(dx20,fscal,fjx0);
1512 fjy0 = _mm_macc_ps(dy20,fscal,fjy0);
1513 fjz0 = _mm_macc_ps(dz20,fscal,fjz0);
1517 /**************************
1518 * CALCULATE INTERACTIONS *
1519 **************************/
1521 if (gmx_mm_any_lt(rsq30,rcutoff2))
1524 r30 = _mm_mul_ps(rsq30,rinv30);
1525 r30 = _mm_andnot_ps(dummy_mask,r30);
1527 /* Compute parameters for interactions between i and j atoms */
1528 qq30 = _mm_mul_ps(iq3,jq0);
1530 /* EWALD ELECTROSTATICS */
1532 /* Analytical PME correction */
1533 zeta2 = _mm_mul_ps(beta2,rsq30);
1534 rinv3 = _mm_mul_ps(rinvsq30,rinv30);
1535 pmecorrF = gmx_mm_pmecorrF_ps(zeta2);
1536 felec = _mm_macc_ps(pmecorrF,beta3,rinv3);
1537 felec = _mm_mul_ps(qq30,felec);
1538 pmecorrV = gmx_mm_pmecorrV_ps(zeta2);
1539 velec = _mm_nmacc_ps(pmecorrV,beta,rinv30);
1540 velec = _mm_mul_ps(qq30,velec);
1542 d = _mm_sub_ps(r30,rswitch);
1543 d = _mm_max_ps(d,_mm_setzero_ps());
1544 d2 = _mm_mul_ps(d,d);
1545 sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
1547 dsw = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
1549 /* Evaluate switch function */
1550 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1551 felec = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv30,_mm_mul_ps(velec,dsw)) );
1552 cutoff_mask = _mm_cmplt_ps(rsq30,rcutoff2);
1556 fscal = _mm_and_ps(fscal,cutoff_mask);
1558 fscal = _mm_andnot_ps(dummy_mask,fscal);
1560 /* Update vectorial force */
1561 fix3 = _mm_macc_ps(dx30,fscal,fix3);
1562 fiy3 = _mm_macc_ps(dy30,fscal,fiy3);
1563 fiz3 = _mm_macc_ps(dz30,fscal,fiz3);
1565 fjx0 = _mm_macc_ps(dx30,fscal,fjx0);
1566 fjy0 = _mm_macc_ps(dy30,fscal,fjy0);
1567 fjz0 = _mm_macc_ps(dz30,fscal,fjz0);
1571 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1572 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1573 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1574 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1576 gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjx0,fjy0,fjz0);
1578 /* Inner loop uses 213 flops */
1581 /* End of innermost loop */
1583 gmx_mm_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1584 f+i_coord_offset,fshift+i_shift_offset);
1586 /* Increment number of inner iterations */
1587 inneriter += j_index_end - j_index_start;
1589 /* Outer loop uses 24 flops */
1592 /* Increment number of outer iterations */
1595 /* Update outer/inner flops */
1597 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_F,outeriter*24 + inneriter*213);