3 * This file is part of the GROMACS molecular simulation package.
5 * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
6 * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
7 * and including many others, as listed in the AUTHORS file in the
8 * top-level source directory and at http://www.gromacs.org.
10 * GROMACS is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU Lesser General Public License
12 * as published by the Free Software Foundation; either version 2.1
13 * of the License, or (at your option) any later version.
15 * GROMACS is distributed in the hope that it will be useful,
16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
18 * Lesser General Public License for more details.
20 * You should have received a copy of the GNU Lesser General Public
21 * License along with GROMACS; if not, see
22 * http://www.gnu.org/licenses, or write to the Free Software Foundation,
23 * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
25 * If you want to redistribute modifications to GROMACS, please
26 * consider that scientific software is very special. Version
27 * control is crucial - bugs must be traceable. We will be happy to
28 * consider code for inclusion in the official distribution, but
29 * derived work must not be called official GROMACS. Details are found
30 * in the README & COPYING files - if they are missing, get the
31 * official version at http://www.gromacs.org.
33 * To help us fund GROMACS development, we humbly ask that you cite
34 * the research papers on the package. Check out http://www.gromacs.org.
36 #error This file must be processed with the Gromacs pre-preprocessor
38 /* #if INCLUDE_HEADER */
43 #include "../nb_kernel.h"
44 #include "types/simple.h"
45 #include "gromacs/math/vec.h"
48 #include "gromacs/simd/math_x86_avx_256_double.h"
49 #include "kernelutil_x86_avx_256_double.h"
52 /* ## List of variables set by the generating script: */
54 /* ## Setttings that apply to the entire kernel: */
55 /* ## KERNEL_ELEC: String, choice for electrostatic interactions */
56 /* ## KERNEL_VDW: String, choice for van der Waals interactions */
57 /* ## KERNEL_NAME: String, name of this kernel */
58 /* ## KERNEL_VF: String telling if we calculate potential, force, or both */
59 /* ## GEOMETRY_I/GEOMETRY_J: String, name of each geometry, e.g. 'Water3' or '1Particle' */
61 /* ## Setttings that apply to particles in the outer (I) or inner (J) loops: */
62 /* ## PARTICLES_I[]/ Arrays with lists of i/j particles to use in kernel. It is */
63 /* ## PARTICLES_J[]: just [0] for particle geometry, but can be longer for water */
64 /* ## PARTICLES_ELEC_I[]/ Arrays with lists of i/j particle that have electrostatics */
65 /* ## PARTICLES_ELEC_J[]: interactions that should be calculated in this kernel. */
66 /* ## PARTICLES_VDW_I[]/ Arrays with the list of i/j particle that have VdW */
67 /* ## PARTICLES_VDW_J[]: interactions that should be calculated in this kernel. */
69 /* ## Setttings for pairs of interactions (e.g. 2nd i particle against 1st j particle) */
70 /* ## PAIRS_IJ[]: Array with (i,j) tuples of pairs for which interactions */
71 /* ## should be calculated in this kernel. Zero-charge particles */
72 /* ## do not have interactions with particles without vdw, and */
73 /* ## Vdw-only interactions are not evaluated in a no-vdw-kernel. */
74 /* ## INTERACTION_FLAGS[][]: 2D matrix, dimension e.g. 3*3 for water-water interactions. */
75 /* ## For each i-j pair, the element [I][J] is a list of strings */
76 /* ## defining properties/flags of this interaction. Examples */
77 /* ## include 'electrostatics'/'vdw' if that type of interaction */
78 /* ## should be evaluated, 'rsq'/'rinv'/'rinvsq' if those values */
79 /* ## are needed, and 'exactcutoff' or 'shift','switch' to */
80 /* ## decide if the force/potential should be modified. This way */
81 /* ## we only calculate values absolutely needed for each case. */
83 /* ## Calculate the size and offset for (merged/interleaved) table data */
86 * Gromacs nonbonded kernel: {KERNEL_NAME}
87 * Electrostatics interaction: {KERNEL_ELEC}
88 * VdW interaction: {KERNEL_VDW}
89 * Geometry: {GEOMETRY_I}-{GEOMETRY_J}
90 * Calculate force/pot: {KERNEL_VF}
94 (t_nblist * gmx_restrict nlist,
95 rvec * gmx_restrict xx,
96 rvec * gmx_restrict ff,
97 t_forcerec * gmx_restrict fr,
98 t_mdatoms * gmx_restrict mdatoms,
99 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
100 t_nrnb * gmx_restrict nrnb)
102 /* ## Not all variables are used for all kernels, but any optimizing compiler fixes that, */
103 /* ## so there is no point in going to extremes to exclude variables that are not needed. */
104 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
105 * just 0 for non-waters.
106 * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
107 * jnr indices corresponding to data put in the four positions in the SIMD register.
109 int i_shift_offset,i_coord_offset,outeriter,inneriter;
110 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
111 int jnrA,jnrB,jnrC,jnrD;
112 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
113 int jnrlistE,jnrlistF,jnrlistG,jnrlistH;
114 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
115 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
117 real *shiftvec,*fshift,*x,*f;
118 real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
120 __m256d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
121 /* #for I in PARTICLES_I */
122 real * vdwioffsetptr{I};
123 /* #if 'LJEwald' in KERNEL_VDW */
124 real * vdwgridioffsetptr{I};
126 __m256d ix{I},iy{I},iz{I},fix{I},fiy{I},fiz{I},iq{I},isai{I};
128 /* #for J in PARTICLES_J */
129 int vdwjidx{J}A,vdwjidx{J}B,vdwjidx{J}C,vdwjidx{J}D;
130 __m256d jx{J},jy{J},jz{J},fjx{J},fjy{J},fjz{J},jq{J},isaj{J};
132 /* #for I,J in PAIRS_IJ */
133 __m256d dx{I}{J},dy{I}{J},dz{I}{J},rsq{I}{J},rinv{I}{J},rinvsq{I}{J},r{I}{J},qq{I}{J},c6_{I}{J},c12_{I}{J};
135 /* #if KERNEL_ELEC != 'None' */
136 __m256d velec,felec,velecsum,facel,crf,krf,krf2;
139 /* #if 'GeneralizedBorn' in KERNEL_ELEC */
141 __m256d vgb,fgb,vgbsum,dvdasum,gbscale,gbtabscale,isaprod,gbqqfactor,gbinvepsdiff,gbeps,dvdatmp;
142 __m256d minushalf = _mm256_set1_pd(-0.5);
143 real *invsqrta,*dvda,*gbtab;
145 /* #if KERNEL_VDW != 'None' */
147 __m256d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
150 __m256d one_sixth = _mm256_set1_pd(1.0/6.0);
151 __m256d one_twelfth = _mm256_set1_pd(1.0/12.0);
153 /* #if 'Table' in KERNEL_ELEC or 'GeneralizedBorn' in KERNEL_ELEC or 'Table' in KERNEL_VDW */
155 __m128i ifour = _mm_set1_epi32(4);
156 __m256d rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
159 /* #if 'LJEwald' in KERNEL_VDW */
160 /* #for I,J in PAIRS_IJ */
161 __m256d c6grid_{I}{J};
164 __m256d ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
165 __m256d one_half = _mm256_set1_pd(0.5);
166 __m256d minus_one = _mm256_set1_pd(-1.0);
168 /* #if 'Ewald' in KERNEL_ELEC */
170 __m256d ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
171 __m256d beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
174 /* #if 'PotentialSwitch' in [KERNEL_MOD_ELEC,KERNEL_MOD_VDW] */
175 __m256d rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
176 real rswitch_scalar,d_scalar;
178 __m256d dummy_mask,cutoff_mask;
179 __m128 tmpmask0,tmpmask1;
180 __m256d signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
181 __m256d one = _mm256_set1_pd(1.0);
182 __m256d two = _mm256_set1_pd(2.0);
188 jindex = nlist->jindex;
190 shiftidx = nlist->shift;
192 shiftvec = fr->shift_vec[0];
193 fshift = fr->fshift[0];
194 /* #if KERNEL_ELEC != 'None' */
195 facel = _mm256_set1_pd(fr->epsfac);
196 charge = mdatoms->chargeA;
197 /* #if 'ReactionField' in KERNEL_ELEC */
198 krf = _mm256_set1_pd(fr->ic->k_rf);
199 krf2 = _mm256_set1_pd(fr->ic->k_rf*2.0);
200 crf = _mm256_set1_pd(fr->ic->c_rf);
203 /* #if KERNEL_VDW != 'None' */
204 nvdwtype = fr->ntype;
206 vdwtype = mdatoms->typeA;
208 /* #if 'LJEwald' in KERNEL_VDW */
209 vdwgridparam = fr->ljpme_c6grid;
210 sh_lj_ewald = _mm256_set1_pd(fr->ic->sh_lj_ewald);
211 ewclj = _mm256_set1_pd(fr->ewaldcoeff_lj);
212 ewclj2 = _mm256_mul_pd(minus_one,_mm256_mul_pd(ewclj,ewclj));
215 /* #if 'Table' in KERNEL_ELEC and 'Table' in KERNEL_VDW */
216 vftab = kernel_data->table_elec_vdw->data;
217 vftabscale = _mm256_set1_pd(kernel_data->table_elec_vdw->scale);
218 /* #elif 'Table' in KERNEL_ELEC */
219 vftab = kernel_data->table_elec->data;
220 vftabscale = _mm256_set1_pd(kernel_data->table_elec->scale);
221 /* #elif 'Table' in KERNEL_VDW */
222 vftab = kernel_data->table_vdw->data;
223 vftabscale = _mm256_set1_pd(kernel_data->table_vdw->scale);
226 /* #if 'Ewald' in KERNEL_ELEC */
227 sh_ewald = _mm256_set1_pd(fr->ic->sh_ewald);
228 beta = _mm256_set1_pd(fr->ic->ewaldcoeff_q);
229 beta2 = _mm256_mul_pd(beta,beta);
230 beta3 = _mm256_mul_pd(beta,beta2);
232 /* #if KERNEL_VF=='Force' and KERNEL_MOD_ELEC!='PotentialSwitch' */
233 ewtab = fr->ic->tabq_coul_F;
234 ewtabscale = _mm256_set1_pd(fr->ic->tabq_scale);
235 ewtabhalfspace = _mm256_set1_pd(0.5/fr->ic->tabq_scale);
237 ewtab = fr->ic->tabq_coul_FDV0;
238 ewtabscale = _mm256_set1_pd(fr->ic->tabq_scale);
239 ewtabhalfspace = _mm256_set1_pd(0.5/fr->ic->tabq_scale);
243 /* #if KERNEL_ELEC=='GeneralizedBorn' */
244 invsqrta = fr->invsqrta;
246 gbtabscale = _mm256_set1_pd(fr->gbtab.scale);
247 gbtab = fr->gbtab.data;
248 gbinvepsdiff = _mm256_set1_pd((1.0/fr->epsilon_r) - (1.0/fr->gb_epsilon_solvent));
251 /* #if 'Water' in GEOMETRY_I */
252 /* Setup water-specific parameters */
253 inr = nlist->iinr[0];
254 /* #for I in PARTICLES_ELEC_I */
255 iq{I} = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+{I}]));
257 /* #for I in PARTICLES_VDW_I */
258 vdwioffsetptr{I} = vdwparam+2*nvdwtype*vdwtype[inr+{I}];
259 /* #if 'LJEwald' in KERNEL_VDW */
260 vdwgridioffsetptr{I} = vdwgridparam+2*nvdwtype*vdwtype[inr+{I}];
265 /* #if 'Water' in GEOMETRY_J */
266 /* #for J in PARTICLES_ELEC_J */
267 jq{J} = _mm256_set1_pd(charge[inr+{J}]);
269 /* #for J in PARTICLES_VDW_J */
270 vdwjidx{J}A = 2*vdwtype[inr+{J}];
272 /* #for I,J in PAIRS_IJ */
273 /* #if 'electrostatics' in INTERACTION_FLAGS[I][J] */
274 qq{I}{J} = _mm256_mul_pd(iq{I},jq{J});
276 /* #if 'vdw' in INTERACTION_FLAGS[I][J] */
277 /* #if 'LJEwald' in KERNEL_VDW */
278 c6_{I}{J} = _mm256_set1_pd(vdwioffsetptr{I}[vdwjidx{J}A]);
279 c12_{I}{J} = _mm256_set1_pd(vdwioffsetptr{I}[vdwjidx{J}A+1]);
280 c6grid_{I}{J} = _mm256_set1_pd(vdwgridioffsetptr{I}[vdwjidx{J}A]);
282 c6_{I}{J} = _mm256_set1_pd(vdwioffsetptr{I}[vdwjidx{J}A]);
283 c12_{I}{J} = _mm256_set1_pd(vdwioffsetptr{I}[vdwjidx{J}A+1]);
289 /* #if KERNEL_MOD_ELEC!='None' or KERNEL_MOD_VDW!='None' */
290 /* #if KERNEL_ELEC!='None' */
291 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
292 rcutoff_scalar = fr->rcoulomb;
294 rcutoff_scalar = fr->rvdw;
296 rcutoff = _mm256_set1_pd(rcutoff_scalar);
297 rcutoff2 = _mm256_mul_pd(rcutoff,rcutoff);
300 /* #if KERNEL_MOD_VDW=='PotentialShift' */
301 sh_vdw_invrcut6 = _mm256_set1_pd(fr->ic->sh_invrc6);
302 rvdw = _mm256_set1_pd(fr->rvdw);
305 /* #if 'PotentialSwitch' in [KERNEL_MOD_ELEC,KERNEL_MOD_VDW] */
306 /* #if KERNEL_MOD_ELEC=='PotentialSwitch' */
307 rswitch_scalar = fr->rcoulomb_switch;
308 rswitch = _mm256_set1_pd(rswitch_scalar);
310 rswitch_scalar = fr->rvdw_switch;
311 rswitch = _mm256_set1_pd(rswitch_scalar);
313 /* Setup switch parameters */
314 d_scalar = rcutoff_scalar-rswitch_scalar;
315 d = _mm256_set1_pd(d_scalar);
316 swV3 = _mm256_set1_pd(-10.0/(d_scalar*d_scalar*d_scalar));
317 swV4 = _mm256_set1_pd( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
318 swV5 = _mm256_set1_pd( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
319 /* #if 'Force' in KERNEL_VF */
320 swF2 = _mm256_set1_pd(-30.0/(d_scalar*d_scalar*d_scalar));
321 swF3 = _mm256_set1_pd( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
322 swF4 = _mm256_set1_pd(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
326 /* Avoid stupid compiler warnings */
327 jnrA = jnrB = jnrC = jnrD = 0;
333 /* ## Keep track of the floating point operations we issue for reporting! */
334 /* #define OUTERFLOPS 0 */
338 for(iidx=0;iidx<4*DIM;iidx++)
343 /* Start outer loop over neighborlists */
344 for(iidx=0; iidx<nri; iidx++)
346 /* Load shift vector for this list */
347 i_shift_offset = DIM*shiftidx[iidx];
349 /* Load limits for loop over neighbors */
350 j_index_start = jindex[iidx];
351 j_index_end = jindex[iidx+1];
353 /* Get outer coordinate index */
355 i_coord_offset = DIM*inr;
357 /* Load i particle coords and add shift vector */
358 /* #if GEOMETRY_I == 'Particle' */
359 gmx_mm256_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
360 /* #elif GEOMETRY_I == 'Water3' */
361 gmx_mm256_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
362 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
363 /* #elif GEOMETRY_I == 'Water4' */
364 /* #if 0 in PARTICLES_I */
365 gmx_mm256_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
366 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
368 gmx_mm256_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
369 &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
373 /* #if 'Force' in KERNEL_VF */
374 /* #for I in PARTICLES_I */
375 fix{I} = _mm256_setzero_pd();
376 fiy{I} = _mm256_setzero_pd();
377 fiz{I} = _mm256_setzero_pd();
381 /* ## For water we already preloaded parameters at the start of the kernel */
382 /* #if not 'Water' in GEOMETRY_I */
383 /* Load parameters for i particles */
384 /* #for I in PARTICLES_ELEC_I */
385 iq{I} = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+{I}]));
386 /* #define OUTERFLOPS OUTERFLOPS+1 */
387 /* #if KERNEL_ELEC=='GeneralizedBorn' */
388 isai{I} = _mm256_set1_pd(invsqrta[inr+{I}]);
391 /* #for I in PARTICLES_VDW_I */
392 vdwioffsetptr{I} = vdwparam+2*nvdwtype*vdwtype[inr+{I}];
393 /* #if 'LJEwald' in KERNEL_VDW */
394 vdwgridioffsetptr{I} = vdwgridparam+2*nvdwtype*vdwtype[inr+{I}];
399 /* #if 'Potential' in KERNEL_VF */
400 /* Reset potential sums */
401 /* #if KERNEL_ELEC != 'None' */
402 velecsum = _mm256_setzero_pd();
404 /* #if 'GeneralizedBorn' in KERNEL_ELEC */
405 vgbsum = _mm256_setzero_pd();
407 /* #if KERNEL_VDW != 'None' */
408 vvdwsum = _mm256_setzero_pd();
411 /* #if 'GeneralizedBorn' in KERNEL_ELEC and 'Force' in KERNEL_VF */
412 dvdasum = _mm256_setzero_pd();
415 /* #for ROUND in ['Loop','Epilogue'] */
417 /* #if ROUND =='Loop' */
418 /* Start inner kernel loop */
419 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
421 /* ## First round is normal loop (next statement resets indentation) */
428 /* ## Second round is epilogue */
430 /* #define INNERFLOPS 0 */
432 /* Get j neighbor index, and coordinate index */
433 /* #if ROUND =='Loop' */
439 jnrlistA = jjnr[jidx];
440 jnrlistB = jjnr[jidx+1];
441 jnrlistC = jjnr[jidx+2];
442 jnrlistD = jjnr[jidx+3];
443 /* Sign of each element will be negative for non-real atoms.
444 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
445 * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
447 tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
449 tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
450 tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
451 dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
453 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
454 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
455 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
456 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
458 j_coord_offsetA = DIM*jnrA;
459 j_coord_offsetB = DIM*jnrB;
460 j_coord_offsetC = DIM*jnrC;
461 j_coord_offsetD = DIM*jnrD;
463 /* load j atom coordinates */
464 /* #if GEOMETRY_J == 'Particle' */
465 gmx_mm256_load_1rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
466 x+j_coord_offsetC,x+j_coord_offsetD,
468 /* #elif GEOMETRY_J == 'Water3' */
469 gmx_mm256_load_3rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
470 x+j_coord_offsetC,x+j_coord_offsetD,
471 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
472 /* #elif GEOMETRY_J == 'Water4' */
473 /* #if 0 in PARTICLES_J */
474 gmx_mm256_load_4rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
475 x+j_coord_offsetC,x+j_coord_offsetD,
476 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
477 &jy2,&jz2,&jx3,&jy3,&jz3);
479 gmx_mm256_load_3rvec_4ptr_swizzle_pd(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
480 x+j_coord_offsetC+DIM,x+j_coord_offsetD+DIM,
481 &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
485 /* Calculate displacement vector */
486 /* #for I,J in PAIRS_IJ */
487 dx{I}{J} = _mm256_sub_pd(ix{I},jx{J});
488 dy{I}{J} = _mm256_sub_pd(iy{I},jy{J});
489 dz{I}{J} = _mm256_sub_pd(iz{I},jz{J});
490 /* #define INNERFLOPS INNERFLOPS+3 */
493 /* Calculate squared distance and things based on it */
494 /* #for I,J in PAIRS_IJ */
495 rsq{I}{J} = gmx_mm256_calc_rsq_pd(dx{I}{J},dy{I}{J},dz{I}{J});
496 /* #define INNERFLOPS INNERFLOPS+5 */
499 /* #for I,J in PAIRS_IJ */
500 /* #if 'rinv' in INTERACTION_FLAGS[I][J] */
501 rinv{I}{J} = gmx_mm256_invsqrt_pd(rsq{I}{J});
502 /* #define INNERFLOPS INNERFLOPS+5 */
506 /* #for I,J in PAIRS_IJ */
507 /* #if 'rinvsq' in INTERACTION_FLAGS[I][J] */
508 /* # if 'rinv' not in INTERACTION_FLAGS[I][J] */
509 rinvsq{I}{J} = gmx_mm256_inv_pd(rsq{I}{J});
510 /* #define INNERFLOPS INNERFLOPS+4 */
512 rinvsq{I}{J} = _mm256_mul_pd(rinv{I}{J},rinv{I}{J});
513 /* #define INNERFLOPS INNERFLOPS+1 */
518 /* #if not 'Water' in GEOMETRY_J */
519 /* Load parameters for j particles */
520 /* #for J in PARTICLES_ELEC_J */
521 jq{J} = gmx_mm256_load_4real_swizzle_pd(charge+jnrA+{J},charge+jnrB+{J},
522 charge+jnrC+{J},charge+jnrD+{J});
523 /* #if KERNEL_ELEC=='GeneralizedBorn' */
524 isaj{J} = gmx_mm256_load_4real_swizzle_pd(invsqrta+jnrA+{J},invsqrta+jnrB+{J},
525 invsqrta+jnrC+{J},invsqrta+jnrD+{J});
528 /* #for J in PARTICLES_VDW_J */
529 vdwjidx{J}A = 2*vdwtype[jnrA+{J}];
530 vdwjidx{J}B = 2*vdwtype[jnrB+{J}];
531 vdwjidx{J}C = 2*vdwtype[jnrC+{J}];
532 vdwjidx{J}D = 2*vdwtype[jnrD+{J}];
536 /* #if 'Force' in KERNEL_VF and not 'Particle' in GEOMETRY_I */
537 /* #for J in PARTICLES_J */
538 fjx{J} = _mm256_setzero_pd();
539 fjy{J} = _mm256_setzero_pd();
540 fjz{J} = _mm256_setzero_pd();
544 /* #for I,J in PAIRS_IJ */
546 /**************************
547 * CALCULATE INTERACTIONS *
548 **************************/
550 /* ## Note special check for TIP4P-TIP4P. Since we are cutting of all hydrogen interactions we also cut the LJ-only O-O interaction */
551 /* #if 'exactcutoff' in INTERACTION_FLAGS[I][J] or (GEOMETRY_I=='Water4' and GEOMETRY_J=='Water4' and 'exactcutoff' in INTERACTION_FLAGS[1][1]) */
552 /* ## We always calculate rinv/rinvsq above to enable pipelineing in compilers (performance tested on x86) */
553 if (gmx_mm256_any_lt(rsq{I}{J},rcutoff2))
555 /* #if 0 ## this and the next two lines is a hack to maintain auto-indentation in template file */
558 /* #define INNERFLOPS INNERFLOPS+1 */
561 /* #if 'r' in INTERACTION_FLAGS[I][J] */
562 r{I}{J} = _mm256_mul_pd(rsq{I}{J},rinv{I}{J});
563 /* #if ROUND == 'Epilogue' */
564 r{I}{J} = _mm256_andnot_pd(dummy_mask,r{I}{J});
565 /* #define INNERFLOPS INNERFLOPS+1 */
567 /* #define INNERFLOPS INNERFLOPS+1 */
570 /* ## For water geometries we already loaded parameters at the start of the kernel */
571 /* #if not 'Water' in GEOMETRY_J */
572 /* Compute parameters for interactions between i and j atoms */
573 /* #if 'electrostatics' in INTERACTION_FLAGS[I][J] */
574 qq{I}{J} = _mm256_mul_pd(iq{I},jq{J});
575 /* #define INNERFLOPS INNERFLOPS+1 */
577 /* #if 'vdw' in INTERACTION_FLAGS[I][J] */
578 gmx_mm256_load_4pair_swizzle_pd(vdwioffsetptr{I}+vdwjidx{J}A,
579 vdwioffsetptr{I}+vdwjidx{J}B,
580 vdwioffsetptr{I}+vdwjidx{J}C,
581 vdwioffsetptr{I}+vdwjidx{J}D,
582 &c6_{I}{J},&c12_{I}{J});
584 /* #if 'LJEwald' in KERNEL_VDW */
585 c6grid_{I}{J} = gmx_mm256_load_4real_swizzle_pd(vdwgridioffsetptr{I}+vdwjidx{J}A,
586 vdwgridioffsetptr{I}+vdwjidx{J}B,
587 vdwgridioffsetptr{I}+vdwjidx{J}C,
588 vdwgridioffsetptr{I}+vdwjidx{J}D);
593 /* #if 'table' in INTERACTION_FLAGS[I][J] */
594 /* Calculate table index by multiplying r with table scale and truncate to integer */
595 rt = _mm256_mul_pd(r{I}{J},vftabscale);
596 vfitab = _mm256_cvttpd_epi32(rt);
597 vfeps = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
598 /* #define INNERFLOPS INNERFLOPS+4 */
599 /* #if 'Table' in KERNEL_ELEC and 'Table' in KERNEL_VDW */
600 /* ## 3 tables, 4 bytes per point: multiply index by 12 */
601 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
602 /* #elif 'Table' in KERNEL_ELEC */
603 /* ## 1 table, 4 bytes per point: multiply index by 4 */
604 vfitab = _mm_slli_epi32(vfitab,2);
605 /* #elif 'Table' in KERNEL_VDW */
606 /* ## 2 tables, 4 bytes per point: multiply index by 8 */
607 vfitab = _mm_slli_epi32(vfitab,3);
611 /* ## ELECTROSTATIC INTERACTIONS */
612 /* #if 'electrostatics' in INTERACTION_FLAGS[I][J] */
614 /* #if KERNEL_ELEC=='Coulomb' */
616 /* COULOMB ELECTROSTATICS */
617 velec = _mm256_mul_pd(qq{I}{J},rinv{I}{J});
618 /* #define INNERFLOPS INNERFLOPS+1 */
619 /* #if 'Force' in KERNEL_VF */
620 felec = _mm256_mul_pd(velec,rinvsq{I}{J});
621 /* #define INNERFLOPS INNERFLOPS+1 */
624 /* #elif KERNEL_ELEC=='ReactionField' */
626 /* REACTION-FIELD ELECTROSTATICS */
627 /* #if 'Potential' in KERNEL_VF */
628 velec = _mm256_mul_pd(qq{I}{J},_mm256_sub_pd(_mm256_add_pd(rinv{I}{J},_mm256_mul_pd(krf,rsq{I}{J})),crf));
629 /* #define INNERFLOPS INNERFLOPS+4 */
631 /* #if 'Force' in KERNEL_VF */
632 felec = _mm256_mul_pd(qq{I}{J},_mm256_sub_pd(_mm256_mul_pd(rinv{I}{J},rinvsq{I}{J}),krf2));
633 /* #define INNERFLOPS INNERFLOPS+3 */
636 /* #elif KERNEL_ELEC=='GeneralizedBorn' */
638 /* GENERALIZED BORN AND COULOMB ELECTROSTATICS */
639 isaprod = _mm256_mul_pd(isai{I},isaj{J});
640 gbqqfactor = _mm256_xor_pd(signbit,_mm256_mul_pd(qq{I}{J},_mm256_mul_pd(isaprod,gbinvepsdiff)));
641 gbscale = _mm256_mul_pd(isaprod,gbtabscale);
642 /* #define INNERFLOPS INNERFLOPS+5 */
644 /* Calculate generalized born table index - this is a separate table from the normal one,
645 * but we use the same procedure by multiplying r with scale and truncating to integer.
647 rt = _mm256_mul_pd(r{I}{J},gbscale);
648 gbitab = _mm256_cvttpd_epi32(rt);
649 gbeps = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
650 gbitab = _mm_slli_epi32(gbitab,2);
651 Y = _mm256_load_pd( gbtab + _mm_extract_epi32(gbitab,0) );
652 F = _mm256_load_pd( gbtab + _mm_extract_epi32(gbitab,1) );
653 G = _mm256_load_pd( gbtab + _mm_extract_epi32(gbitab,2) );
654 H = _mm256_load_pd( gbtab + _mm_extract_epi32(gbitab,3) );
655 GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
656 Heps = _mm256_mul_pd(gbeps,H);
657 Fp = _mm256_add_pd(F,_mm256_mul_pd(gbeps,_mm256_add_pd(G,Heps)));
658 VV = _mm256_add_pd(Y,_mm256_mul_pd(gbeps,Fp));
659 vgb = _mm256_mul_pd(gbqqfactor,VV);
660 /* #define INNERFLOPS INNERFLOPS+10 */
662 /* #if 'Force' in KERNEL_VF */
663 FF = _mm256_add_pd(Fp,_mm256_mul_pd(gbeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
664 fgb = _mm256_mul_pd(gbqqfactor,_mm256_mul_pd(FF,gbscale));
665 dvdatmp = _mm256_mul_pd(minushalf,_mm256_add_pd(vgb,_mm256_mul_pd(fgb,r{I}{J})));
666 /* #if ROUND == 'Epilogue' */
667 dvdatmp = _mm256_andnot_pd(dummy_mask,dvdatmp);
669 dvdasum = _mm256_add_pd(dvdasum,dvdatmp);
670 /* #if ROUND == 'Loop' */
676 /* The pointers to scratch make sure that this code with compilers that take gmx_restrict seriously (e.g. icc 13) really can't screw things up. */
677 fjptrA = (jnrlistA>=0) ? dvda+jnrA : scratch;
678 fjptrB = (jnrlistB>=0) ? dvda+jnrB : scratch;
679 fjptrC = (jnrlistC>=0) ? dvda+jnrC : scratch;
680 fjptrD = (jnrlistD>=0) ? dvda+jnrD : scratch;
682 gmx_mm256_increment_4real_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
683 _mm256_mul_pd(dvdatmp,_mm256_mul_pd(isaj{J},isaj{J})));
684 /* #define INNERFLOPS INNERFLOPS+12 */
686 velec = _mm256_mul_pd(qq{I}{J},rinv{I}{J});
687 /* #define INNERFLOPS INNERFLOPS+1 */
688 /* #if 'Force' in KERNEL_VF */
689 felec = _mm256_mul_pd(_mm256_sub_pd(_mm256_mul_pd(velec,rinv{I}{J}),fgb),rinv{I}{J});
690 /* #define INNERFLOPS INNERFLOPS+3 */
693 /* #elif KERNEL_ELEC=='Ewald' */
694 /* EWALD ELECTROSTATICS */
696 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
697 ewrt = _mm256_mul_pd(r{I}{J},ewtabscale);
698 ewitab = _mm256_cvttpd_epi32(ewrt);
699 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
700 /* #define INNERFLOPS INNERFLOPS+4 */
701 /* #if 'Potential' in KERNEL_VF or KERNEL_MOD_ELEC=='PotentialSwitch' */
702 ewitab = _mm_slli_epi32(ewitab,2);
703 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
704 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
705 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
706 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
707 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
708 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
709 /* #define INNERFLOPS INNERFLOPS+2 */
710 /* #if KERNEL_MOD_ELEC=='PotentialShift' */
711 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
712 velec = _mm256_mul_pd(qq{I}{J},_mm256_sub_pd(_mm256_sub_pd(rinv{I}{J},sh_ewald),velec));
713 /* #define INNERFLOPS INNERFLOPS+7 */
715 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
716 velec = _mm256_mul_pd(qq{I}{J},_mm256_sub_pd(rinv{I}{J},velec));
717 /* #define INNERFLOPS INNERFLOPS+6 */
719 /* #if 'Force' in KERNEL_VF */
720 felec = _mm256_mul_pd(_mm256_mul_pd(qq{I}{J},rinv{I}{J}),_mm256_sub_pd(rinvsq{I}{J},felec));
721 /* #define INNERFLOPS INNERFLOPS+3 */
723 /* #elif KERNEL_VF=='Force' */
724 gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
725 ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
727 felec = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
728 felec = _mm256_mul_pd(_mm256_mul_pd(qq{I}{J},rinv{I}{J}),_mm256_sub_pd(rinvsq{I}{J},felec));
729 /* #define INNERFLOPS INNERFLOPS+7 */
732 /* #elif KERNEL_ELEC=='CubicSplineTable' */
734 /* CUBIC SPLINE TABLE ELECTROSTATICS */
735 Y = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
736 F = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
737 G = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
738 H = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
739 GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
740 Heps = _mm256_mul_pd(vfeps,H);
741 Fp = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
742 /* #define INNERFLOPS INNERFLOPS+4 */
743 /* #if 'Potential' in KERNEL_VF */
744 VV = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
745 velec = _mm256_mul_pd(qq{I}{J},VV);
746 /* #define INNERFLOPS INNERFLOPS+3 */
748 /* #if 'Force' in KERNEL_VF */
749 FF = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
750 felec = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq{I}{J},FF),_mm256_mul_pd(vftabscale,rinv{I}{J})));
751 /* #define INNERFLOPS INNERFLOPS+7 */
754 /* ## End of check for electrostatics interaction forms */
756 /* ## END OF ELECTROSTATIC INTERACTION CHECK FOR PAIR I-J */
758 /* #if 'vdw' in INTERACTION_FLAGS[I][J] */
760 /* #if KERNEL_VDW=='LennardJones' */
762 /* LENNARD-JONES DISPERSION/REPULSION */
764 rinvsix = _mm256_mul_pd(_mm256_mul_pd(rinvsq{I}{J},rinvsq{I}{J}),rinvsq{I}{J});
765 /* #define INNERFLOPS INNERFLOPS+2 */
766 /* #if 'Potential' in KERNEL_VF or KERNEL_MOD_VDW=='PotentialSwitch' */
767 vvdw6 = _mm256_mul_pd(c6_{I}{J},rinvsix);
768 vvdw12 = _mm256_mul_pd(c12_{I}{J},_mm256_mul_pd(rinvsix,rinvsix));
769 /* #define INNERFLOPS INNERFLOPS+3 */
770 /* #if KERNEL_MOD_VDW=='PotentialShift' */
771 vvdw = _mm256_sub_pd(_mm256_mul_pd( _mm256_sub_pd(vvdw12 , _mm256_mul_pd(c12_{I}{J},_mm256_mul_pd(sh_vdw_invrcut6,sh_vdw_invrcut6))), one_twelfth) ,
772 _mm256_mul_pd( _mm256_sub_pd(vvdw6,_mm256_mul_pd(c6_{I}{J},sh_vdw_invrcut6)),one_sixth));
773 /* #define INNERFLOPS INNERFLOPS+8 */
775 vvdw = _mm256_sub_pd( _mm256_mul_pd(vvdw12,one_twelfth) , _mm256_mul_pd(vvdw6,one_sixth) );
776 /* #define INNERFLOPS INNERFLOPS+3 */
778 /* ## Check for force inside potential check, i.e. this means we already did the potential part */
779 /* #if 'Force' in KERNEL_VF */
780 fvdw = _mm256_mul_pd(_mm256_sub_pd(vvdw12,vvdw6),rinvsq{I}{J});
781 /* #define INNERFLOPS INNERFLOPS+2 */
783 /* #elif KERNEL_VF=='Force' */
784 /* ## Force-only LennardJones makes it possible to save 1 flop (they do add up...) */
785 fvdw = _mm256_mul_pd(_mm256_sub_pd(_mm256_mul_pd(c12_{I}{J},rinvsix),c6_{I}{J}),_mm256_mul_pd(rinvsix,rinvsq{I}{J}));
786 /* #define INNERFLOPS INNERFLOPS+4 */
789 /* #elif KERNEL_VDW=='CubicSplineTable' */
791 /* CUBIC SPLINE TABLE DISPERSION */
792 /* #if 'Table' in KERNEL_ELEC */
793 vfitab = _mm_add_epi32(vfitab,ifour);
795 Y = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
796 F = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
797 G = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
798 H = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
799 GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
800 Heps = _mm256_mul_pd(vfeps,H);
801 Fp = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
802 /* #define INNERFLOPS INNERFLOPS+4 */
803 /* #if 'Potential' in KERNEL_VF */
804 VV = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
805 vvdw6 = _mm256_mul_pd(c6_{I}{J},VV);
806 /* #define INNERFLOPS INNERFLOPS+3 */
808 /* #if 'Force' in KERNEL_VF */
809 FF = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
810 fvdw6 = _mm256_mul_pd(c6_{I}{J},FF);
811 /* #define INNERFLOPS INNERFLOPS+4 */
814 /* CUBIC SPLINE TABLE REPULSION */
815 vfitab = _mm_add_epi32(vfitab,ifour);
816 Y = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
817 F = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
818 G = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
819 H = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
820 GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
821 Heps = _mm256_mul_pd(vfeps,H);
822 Fp = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
823 /* #define INNERFLOPS INNERFLOPS+4 */
824 /* #if 'Potential' in KERNEL_VF */
825 VV = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
826 vvdw12 = _mm256_mul_pd(c12_{I}{J},VV);
827 /* #define INNERFLOPS INNERFLOPS+3 */
829 /* #if 'Force' in KERNEL_VF */
830 FF = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
831 fvdw12 = _mm256_mul_pd(c12_{I}{J},FF);
832 /* #define INNERFLOPS INNERFLOPS+5 */
834 /* #if 'Potential' in KERNEL_VF */
835 vvdw = _mm256_add_pd(vvdw12,vvdw6);
836 /* #define INNERFLOPS INNERFLOPS+1 */
838 /* #if 'Force' in KERNEL_VF */
839 fvdw = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_add_pd(fvdw6,fvdw12),_mm256_mul_pd(vftabscale,rinv{I}{J})));
840 /* #define INNERFLOPS INNERFLOPS+4 */
843 /* #elif KERNEL_VDW=='LJEwald' */
845 /* Analytical LJ-PME */
846 rinvsix = _mm256_mul_pd(_mm256_mul_pd(rinvsq{I}{J},rinvsq{I}{J}),rinvsq{I}{J});
847 ewcljrsq = _mm256_mul_pd(ewclj2,rsq{I}{J});
848 ewclj6 = _mm256_mul_pd(ewclj2,_mm256_mul_pd(ewclj2,ewclj2));
849 exponent = gmx_simd_exp_d(ewcljrsq);
850 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
851 poly = _mm256_mul_pd(exponent,_mm256_add_pd(_mm256_sub_pd(one,ewcljrsq),_mm256_mul_pd(_mm256_mul_pd(ewcljrsq,ewcljrsq),one_half)));
852 /* #define INNERFLOPS INNERFLOPS+11 */
853 /* #if 'Potential' in KERNEL_VF or KERNEL_MOD_VDW=='PotentialSwitch' */
854 /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
855 vvdw6 = _mm256_mul_pd(_mm256_sub_pd(c6_{I}{J},_mm256_mul_pd(c6grid_{I}{J},_mm256_sub_pd(one,poly))),rinvsix);
856 vvdw12 = _mm256_mul_pd(c12_{I}{J},_mm256_mul_pd(rinvsix,rinvsix));
857 /* #define INNERFLOPS INNERFLOPS+6 */
858 /* #if KERNEL_MOD_VDW=='PotentialShift' */
859 vvdw = _mm256_sub_pd(_mm256_mul_pd( _mm256_sub_pd(vvdw12 , _mm256_mul_pd(c12_{I}{J},_mm256_mul_pd(sh_vdw_invrcut6,sh_vdw_invrcut6))), one_twelfth) ,
860 _mm256_mul_pd( _mm256_sub_pd(vvdw6,_mm256_add_pd(_mm256_mul_pd(c6_{I}{J},sh_vdw_invrcut6),_mm256_mul_pd(c6grid_{I}{J},sh_lj_ewald))),one_sixth));
861 /* #define INNERFLOPS INNERFLOPS+10 */
863 vvdw = _mm256_sub_pd(_mm256_mul_pd(vvdw12,one_twelfth),_mm256_mul_pd(vvdw6,one_sixth));
864 /* #define INNERFLOPS INNERFLOPS+6 */
866 /* ## Check for force inside potential check, i.e. this means we already did the potential part */
867 /* #if 'Force' in KERNEL_VF */
868 /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
869 fvdw = _mm256_mul_pd(_mm256_sub_pd(vvdw12,_mm256_sub_pd(vvdw6,_mm256_mul_pd(_mm256_mul_pd(c6grid_{I}{J},one_sixth),_mm256_mul_pd(exponent,ewclj6)))),rinvsq{I}{J});
870 /* #define INNERFLOPS INNERFLOPS+6 */
872 /* #elif KERNEL_VF=='Force' */
873 /* f6A = 6 * C6grid * (1 - poly) */
874 f6A = _mm256_mul_pd(c6grid_{I}{J},_mm256_sub_pd(one,poly));
875 /* f6B = C6grid * exponent * beta^6 */
876 f6B = _mm256_mul_pd(_mm256_mul_pd(c6grid_{I}{J},one_sixth),_mm256_mul_pd(exponent,ewclj6));
877 /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
878 fvdw = _mm256_mul_pd(_mm256_add_pd(_mm256_mul_pd(_mm256_sub_pd(_mm256_mul_pd(c12_{I}{J},rinvsix),_mm256_sub_pd(c6_{I}{J},f6A)),rinvsix),f6B),rinvsq{I}{J});
879 /* #define INNERFLOPS INNERFLOPS+11 */
882 /* ## End of check for vdw interaction forms */
884 /* ## END OF VDW INTERACTION CHECK FOR PAIR I-J */
886 /* #if 'switch' in INTERACTION_FLAGS[I][J] */
887 d = _mm256_sub_pd(r{I}{J},rswitch);
888 d = _mm256_max_pd(d,_mm256_setzero_pd());
889 d2 = _mm256_mul_pd(d,d);
890 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)))))));
891 /* #define INNERFLOPS INNERFLOPS+10 */
893 /* #if 'Force' in KERNEL_VF */
894 dsw = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
895 /* #define INNERFLOPS INNERFLOPS+5 */
898 /* Evaluate switch function */
899 /* #if 'Force' in KERNEL_VF */
900 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
901 /* #if 'electrostatics' in INTERACTION_FLAGS[I][J] and KERNEL_MOD_ELEC=='PotentialSwitch' */
902 felec = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv{I}{J},_mm256_mul_pd(velec,dsw)) );
903 /* #define INNERFLOPS INNERFLOPS+4 */
905 /* #if 'vdw' in INTERACTION_FLAGS[I][J] and KERNEL_MOD_VDW=='PotentialSwitch' */
906 fvdw = _mm256_sub_pd( _mm256_mul_pd(fvdw,sw) , _mm256_mul_pd(rinv{I}{J},_mm256_mul_pd(vvdw,dsw)) );
907 /* #define INNERFLOPS INNERFLOPS+4 */
910 /* #if 'Potential' in KERNEL_VF */
911 /* #if 'electrostatics' in INTERACTION_FLAGS[I][J] and KERNEL_MOD_ELEC=='PotentialSwitch' */
912 velec = _mm256_mul_pd(velec,sw);
913 /* #define INNERFLOPS INNERFLOPS+1 */
915 /* #if 'vdw' in INTERACTION_FLAGS[I][J] and KERNEL_MOD_VDW=='PotentialSwitch' */
916 vvdw = _mm256_mul_pd(vvdw,sw);
917 /* #define INNERFLOPS INNERFLOPS+1 */
921 /* ## Note special check for TIP4P-TIP4P. Since we are cutting of all hydrogen interactions we also cut the LJ-only O-O interaction */
922 /* #if 'exactcutoff' in INTERACTION_FLAGS[I][J] or (GEOMETRY_I=='Water4' and GEOMETRY_J=='Water4' and 'exactcutoff' in INTERACTION_FLAGS[1][1]) */
923 cutoff_mask = _mm256_cmp_pd(rsq{I}{J},rcutoff2,_CMP_LT_OQ);
924 /* #define INNERFLOPS INNERFLOPS+1 */
927 /* #if 'Potential' in KERNEL_VF */
928 /* Update potential sum for this i atom from the interaction with this j atom. */
929 /* #if 'electrostatics' in INTERACTION_FLAGS[I][J] */
930 /* #if 'exactcutoff' in INTERACTION_FLAGS[I][J] */
931 velec = _mm256_and_pd(velec,cutoff_mask);
932 /* #define INNERFLOPS INNERFLOPS+1 */
934 /* #if ROUND == 'Epilogue' */
935 velec = _mm256_andnot_pd(dummy_mask,velec);
937 velecsum = _mm256_add_pd(velecsum,velec);
938 /* #define INNERFLOPS INNERFLOPS+1 */
939 /* #if KERNEL_ELEC=='GeneralizedBorn' */
940 /* #if 'exactcutoff' in INTERACTION_FLAGS[I][J] */
941 vgb = _mm256_and_pd(vgb,cutoff_mask);
942 /* #define INNERFLOPS INNERFLOPS+1 */
944 /* #if ROUND == 'Epilogue' */
945 vgb = _mm256_andnot_pd(dummy_mask,vgb);
947 vgbsum = _mm256_add_pd(vgbsum,vgb);
948 /* #define INNERFLOPS INNERFLOPS+1 */
951 /* #if 'vdw' in INTERACTION_FLAGS[I][J] */
952 /* ## Note special check for TIP4P-TIP4P. Since we are cutting of all hydrogen interactions we also cut the LJ-only O-O interaction */
953 /* #if 'exactcutoff' in INTERACTION_FLAGS[I][J] or (GEOMETRY_I=='Water4' and GEOMETRY_J=='Water4' and 'exactcutoff' in INTERACTION_FLAGS[1][1]) */
954 vvdw = _mm256_and_pd(vvdw,cutoff_mask);
955 /* #define INNERFLOPS INNERFLOPS+1 */
957 /* #if ROUND == 'Epilogue' */
958 vvdw = _mm256_andnot_pd(dummy_mask,vvdw);
960 vvdwsum = _mm256_add_pd(vvdwsum,vvdw);
961 /* #define INNERFLOPS INNERFLOPS+1 */
965 /* #if 'Force' in KERNEL_VF */
967 /* #if 'electrostatics' in INTERACTION_FLAGS[I][J] and 'vdw' in INTERACTION_FLAGS[I][J] */
968 fscal = _mm256_add_pd(felec,fvdw);
969 /* #define INNERFLOPS INNERFLOPS+1 */
970 /* #elif 'electrostatics' in INTERACTION_FLAGS[I][J] */
972 /* #elif 'vdw' in INTERACTION_FLAGS[I][J] */
976 /* ## Note special check for TIP4P-TIP4P. Since we are cutting of all hydrogen interactions we also cut the LJ-only O-O interaction */
977 /* #if 'exactcutoff' in INTERACTION_FLAGS[I][J] or (GEOMETRY_I=='Water4' and GEOMETRY_J=='Water4' and 'exactcutoff' in INTERACTION_FLAGS[1][1]) */
978 fscal = _mm256_and_pd(fscal,cutoff_mask);
979 /* #define INNERFLOPS INNERFLOPS+1 */
982 /* #if ROUND == 'Epilogue' */
983 fscal = _mm256_andnot_pd(dummy_mask,fscal);
986 /* Calculate temporary vectorial force */
987 tx = _mm256_mul_pd(fscal,dx{I}{J});
988 ty = _mm256_mul_pd(fscal,dy{I}{J});
989 tz = _mm256_mul_pd(fscal,dz{I}{J});
991 /* Update vectorial force */
992 fix{I} = _mm256_add_pd(fix{I},tx);
993 fiy{I} = _mm256_add_pd(fiy{I},ty);
994 fiz{I} = _mm256_add_pd(fiz{I},tz);
995 /* #define INNERFLOPS INNERFLOPS+6 */
997 /* #if GEOMETRY_I == 'Particle' */
998 /* #if ROUND == 'Loop' */
999 fjptrA = f+j_coord_offsetA;
1000 fjptrB = f+j_coord_offsetB;
1001 fjptrC = f+j_coord_offsetC;
1002 fjptrD = f+j_coord_offsetD;
1004 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1005 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1006 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1007 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1009 gmx_mm256_decrement_1rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,tx,ty,tz);
1010 /* #define INNERFLOPS INNERFLOPS+3 */
1012 fjx{J} = _mm256_add_pd(fjx{J},tx);
1013 fjy{J} = _mm256_add_pd(fjy{J},ty);
1014 fjz{J} = _mm256_add_pd(fjz{J},tz);
1015 /* #define INNERFLOPS INNERFLOPS+3 */
1020 /* ## Note special check for TIP4P-TIP4P. Since we are cutting of all hydrogen interactions we also cut the LJ-only O-O interaction */
1021 /* #if 'exactcutoff' in INTERACTION_FLAGS[I][J] or (GEOMETRY_I=='Water4' and GEOMETRY_J=='Water4' and 'exactcutoff' in INTERACTION_FLAGS[1][1]) */
1022 /* #if 0 ## This and next two lines is a hack to maintain indentation in template file */
1027 /* ## End of check for the interaction being outside the cutoff */
1030 /* ## End of loop over i-j interaction pairs */
1032 /* #if GEOMETRY_I != 'Particle' */
1033 /* #if ROUND == 'Loop' */
1034 fjptrA = f+j_coord_offsetA;
1035 fjptrB = f+j_coord_offsetB;
1036 fjptrC = f+j_coord_offsetC;
1037 fjptrD = f+j_coord_offsetD;
1039 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1040 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1041 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1042 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1046 /* #if 'Water' in GEOMETRY_I and GEOMETRY_J == 'Particle' */
1047 gmx_mm256_decrement_1rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,fjx0,fjy0,fjz0);
1048 /* #define INNERFLOPS INNERFLOPS+3 */
1049 /* #elif GEOMETRY_J == 'Water3' */
1050 gmx_mm256_decrement_3rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1051 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1052 /* #define INNERFLOPS INNERFLOPS+9 */
1053 /* #elif GEOMETRY_J == 'Water4' */
1054 /* #if 0 in PARTICLES_J */
1055 gmx_mm256_decrement_4rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1056 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1057 fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1058 /* #define INNERFLOPS INNERFLOPS+12 */
1060 gmx_mm256_decrement_3rvec_4ptr_swizzle_pd(fjptrA+DIM,fjptrB+DIM,fjptrC+DIM,fjptrD+DIM,
1061 fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1062 /* #define INNERFLOPS INNERFLOPS+9 */
1066 /* Inner loop uses {INNERFLOPS} flops */
1071 /* End of innermost loop */
1073 /* #if 'Force' in KERNEL_VF */
1074 /* #if GEOMETRY_I == 'Particle' */
1075 gmx_mm256_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
1076 f+i_coord_offset,fshift+i_shift_offset);
1077 /* #define OUTERFLOPS OUTERFLOPS+6 */
1078 /* #elif GEOMETRY_I == 'Water3' */
1079 gmx_mm256_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1080 f+i_coord_offset,fshift+i_shift_offset);
1081 /* #define OUTERFLOPS OUTERFLOPS+18 */
1082 /* #elif GEOMETRY_I == 'Water4' */
1083 /* #if 0 in PARTICLES_I */
1084 gmx_mm256_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1085 f+i_coord_offset,fshift+i_shift_offset);
1086 /* #define OUTERFLOPS OUTERFLOPS+24 */
1088 gmx_mm256_update_iforce_3atom_swizzle_pd(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1089 f+i_coord_offset+DIM,fshift+i_shift_offset);
1090 /* #define OUTERFLOPS OUTERFLOPS+18 */
1095 /* #if 'Potential' in KERNEL_VF */
1097 /* Update potential energies */
1098 /* #if KERNEL_ELEC != 'None' */
1099 gmx_mm256_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1100 /* #define OUTERFLOPS OUTERFLOPS+1 */
1102 /* #if 'GeneralizedBorn' in KERNEL_ELEC */
1103 gmx_mm256_update_1pot_pd(vgbsum,kernel_data->energygrp_polarization+ggid);
1104 /* #define OUTERFLOPS OUTERFLOPS+1 */
1106 /* #if KERNEL_VDW != 'None' */
1107 gmx_mm256_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1108 /* #define OUTERFLOPS OUTERFLOPS+1 */
1111 /* #if 'GeneralizedBorn' in KERNEL_ELEC and 'Force' in KERNEL_VF */
1112 dvdasum = _mm256_mul_pd(dvdasum, _mm256_mul_pd(isai{I},isai{I}));
1113 gmx_mm256_update_1pot_pd(dvdasum,dvda+inr);
1116 /* Increment number of inner iterations */
1117 inneriter += j_index_end - j_index_start;
1119 /* Outer loop uses {OUTERFLOPS} flops */
1122 /* Increment number of outer iterations */
1125 /* Update outer/inner flops */
1126 /* ## NB: This is not important, it just affects the flopcount. However, since our preprocessor is */
1127 /* ## primitive and replaces aggressively even in strings inside these directives, we need to */
1128 /* ## assemble the main part of the name (containing KERNEL/ELEC/VDW) directly in the source. */
1129 /* #if GEOMETRY_I == 'Water3' */
1130 /* #define ISUFFIX '_W3' */
1131 /* #elif GEOMETRY_I == 'Water4' */
1132 /* #define ISUFFIX '_W4' */
1134 /* #define ISUFFIX '' */
1136 /* #if GEOMETRY_J == 'Water3' */
1137 /* #define JSUFFIX 'W3' */
1138 /* #elif GEOMETRY_J == 'Water4' */
1139 /* #define JSUFFIX 'W4' */
1141 /* #define JSUFFIX '' */
1143 /* #if 'PotentialAndForce' in KERNEL_VF */
1144 /* #define VFSUFFIX '_VF' */
1145 /* #elif 'Potential' in KERNEL_VF */
1146 /* #define VFSUFFIX '_V' */
1148 /* #define VFSUFFIX '_F' */
1151 /* #if KERNEL_ELEC != 'None' and KERNEL_VDW != 'None' */
1152 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW{ISUFFIX}{JSUFFIX}{VFSUFFIX},outeriter*{OUTERFLOPS} + inneriter*{INNERFLOPS});
1153 /* #elif KERNEL_ELEC != 'None' */
1154 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC{ISUFFIX}{JSUFFIX}{VFSUFFIX},outeriter*{OUTERFLOPS} + inneriter*{INNERFLOPS});
1156 inc_nrnb(nrnb,eNR_NBKERNEL_VDW{ISUFFIX}{JSUFFIX}{VFSUFFIX},outeriter*{OUTERFLOPS} + inneriter*{INNERFLOPS});