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 */
45 #include "../nb_kernel.h"
46 #include "types/simple.h"
47 #include "gromacs/math/vec.h"
50 #include "gromacs/simd/math_x86_avx_128_fma_double.h"
51 #include "kernelutil_x86_avx_128_fma_double.h"
54 /* ## List of variables set by the generating script: */
56 /* ## Setttings that apply to the entire kernel: */
57 /* ## KERNEL_ELEC: String, choice for electrostatic interactions */
58 /* ## KERNEL_VDW: String, choice for van der Waals interactions */
59 /* ## KERNEL_NAME: String, name of this kernel */
60 /* ## KERNEL_VF: String telling if we calculate potential, force, or both */
61 /* ## GEOMETRY_I/GEOMETRY_J: String, name of each geometry, e.g. 'Water3' or '1Particle' */
63 /* ## Setttings that apply to particles in the outer (I) or inner (J) loops: */
64 /* ## PARTICLES_I[]/ Arrays with lists of i/j particles to use in kernel. It is */
65 /* ## PARTICLES_J[]: just [0] for particle geometry, but can be longer for water */
66 /* ## PARTICLES_ELEC_I[]/ Arrays with lists of i/j particle that have electrostatics */
67 /* ## PARTICLES_ELEC_J[]: interactions that should be calculated in this kernel. */
68 /* ## PARTICLES_VDW_I[]/ Arrays with the list of i/j particle that have VdW */
69 /* ## PARTICLES_VDW_J[]: interactions that should be calculated in this kernel. */
71 /* ## Setttings for pairs of interactions (e.g. 2nd i particle against 1st j particle) */
72 /* ## PAIRS_IJ[]: Array with (i,j) tuples of pairs for which interactions */
73 /* ## should be calculated in this kernel. Zero-charge particles */
74 /* ## do not have interactions with particles without vdw, and */
75 /* ## Vdw-only interactions are not evaluated in a no-vdw-kernel. */
76 /* ## INTERACTION_FLAGS[][]: 2D matrix, dimension e.g. 3*3 for water-water interactions. */
77 /* ## For each i-j pair, the element [I][J] is a list of strings */
78 /* ## defining properties/flags of this interaction. Examples */
79 /* ## include 'electrostatics'/'vdw' if that type of interaction */
80 /* ## should be evaluated, 'rsq'/'rinv'/'rinvsq' if those values */
81 /* ## are needed, and 'exactcutoff' or 'shift','switch' to */
82 /* ## decide if the force/potential should be modified. This way */
83 /* ## we only calculate values absolutely needed for each case. */
85 /* ## Calculate the size and offset for (merged/interleaved) table data */
88 * Gromacs nonbonded kernel: {KERNEL_NAME}
89 * Electrostatics interaction: {KERNEL_ELEC}
90 * VdW interaction: {KERNEL_VDW}
91 * Geometry: {GEOMETRY_I}-{GEOMETRY_J}
92 * Calculate force/pot: {KERNEL_VF}
96 (t_nblist * gmx_restrict nlist,
97 rvec * gmx_restrict xx,
98 rvec * gmx_restrict ff,
99 t_forcerec * gmx_restrict fr,
100 t_mdatoms * gmx_restrict mdatoms,
101 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
102 t_nrnb * gmx_restrict nrnb)
104 /* ## Not all variables are used for all kernels, but any optimizing compiler fixes that, */
105 /* ## so there is no point in going to extremes to exclude variables that are not needed. */
106 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
107 * just 0 for non-waters.
108 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
109 * jnr indices corresponding to data put in the four positions in the SIMD register.
111 int i_shift_offset,i_coord_offset,outeriter,inneriter;
112 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
114 int j_coord_offsetA,j_coord_offsetB;
115 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
117 real *shiftvec,*fshift,*x,*f;
118 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
119 /* #for I in PARTICLES_I */
121 __m128d ix{I},iy{I},iz{I},fix{I},fiy{I},fiz{I},iq{I},isai{I};
123 /* #for J in PARTICLES_J */
124 int vdwjidx{J}A,vdwjidx{J}B;
125 __m128d jx{J},jy{J},jz{J},fjx{J},fjy{J},fjz{J},jq{J},isaj{J};
127 /* #for I,J in PAIRS_IJ */
128 __m128d 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};
130 /* #if KERNEL_ELEC != 'None' */
131 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
134 /* #if 'GeneralizedBorn' in KERNEL_ELEC */
136 __m128d vgb,fgb,vgbsum,dvdasum,gbscale,gbtabscale,isaprod,gbqqfactor,gbinvepsdiff,dvdaj,gbeps,twogbeps,dvdatmp;
137 __m128d minushalf = _mm_set1_pd(-0.5);
138 real *invsqrta,*dvda,*gbtab;
140 /* #if KERNEL_VDW != 'None' */
142 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
145 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
146 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
148 /* #if 'Table' in KERNEL_ELEC or 'GeneralizedBorn' in KERNEL_ELEC or 'Table' in KERNEL_VDW */
150 __m128i ifour = _mm_set1_epi32(4);
151 __m128d rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
154 /* #if 'LJEwald' in KERNEL_VDW */
155 /* #for I,J in PAIRS_IJ */
156 __m128d c6grid_{I}{J};
159 __m128d ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
160 __m128d one_half = _mm_set1_pd(0.5);
161 __m128d minus_one = _mm_set1_pd(-1.0);
163 /* #if 'Ewald' in KERNEL_ELEC */
165 __m128d ewtabscale,eweps,twoeweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
168 /* #if 'PotentialSwitch' in [KERNEL_MOD_ELEC,KERNEL_MOD_VDW] */
169 __m128d rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
170 real rswitch_scalar,d_scalar;
172 __m128d dummy_mask,cutoff_mask;
173 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
174 __m128d one = _mm_set1_pd(1.0);
175 __m128d two = _mm_set1_pd(2.0);
181 jindex = nlist->jindex;
183 shiftidx = nlist->shift;
185 shiftvec = fr->shift_vec[0];
186 fshift = fr->fshift[0];
187 /* #if KERNEL_ELEC != 'None' */
188 facel = _mm_set1_pd(fr->epsfac);
189 charge = mdatoms->chargeA;
190 /* #if 'ReactionField' in KERNEL_ELEC */
191 krf = _mm_set1_pd(fr->ic->k_rf);
192 krf2 = _mm_set1_pd(fr->ic->k_rf*2.0);
193 crf = _mm_set1_pd(fr->ic->c_rf);
196 /* #if KERNEL_VDW != 'None' */
197 nvdwtype = fr->ntype;
199 vdwtype = mdatoms->typeA;
201 /* #if 'LJEwald' in KERNEL_VDW */
202 vdwgridparam = fr->ljpme_c6grid;
203 sh_lj_ewald = _mm_set1_pd(fr->ic->sh_lj_ewald);
204 ewclj = _mm_set1_pd(fr->ewaldcoeff_lj);
205 ewclj2 = _mm_mul_pd(minus_one,_mm_mul_pd(ewclj,ewclj));
208 /* #if 'Table' in KERNEL_ELEC and 'Table' in KERNEL_VDW */
209 vftab = kernel_data->table_elec_vdw->data;
210 vftabscale = _mm_set1_pd(kernel_data->table_elec_vdw->scale);
211 /* #elif 'Table' in KERNEL_ELEC */
212 vftab = kernel_data->table_elec->data;
213 vftabscale = _mm_set1_pd(kernel_data->table_elec->scale);
214 /* #elif 'Table' in KERNEL_VDW */
215 vftab = kernel_data->table_vdw->data;
216 vftabscale = _mm_set1_pd(kernel_data->table_vdw->scale);
219 /* #if 'Ewald' in KERNEL_ELEC */
220 sh_ewald = _mm_set1_pd(fr->ic->sh_ewald);
221 /* #if KERNEL_VF=='Force' and KERNEL_MOD_ELEC!='PotentialSwitch' */
222 ewtab = fr->ic->tabq_coul_F;
223 ewtabscale = _mm_set1_pd(fr->ic->tabq_scale);
224 ewtabhalfspace = _mm_set1_pd(0.5/fr->ic->tabq_scale);
226 ewtab = fr->ic->tabq_coul_FDV0;
227 ewtabscale = _mm_set1_pd(fr->ic->tabq_scale);
228 ewtabhalfspace = _mm_set1_pd(0.5/fr->ic->tabq_scale);
232 /* #if KERNEL_ELEC=='GeneralizedBorn' */
233 invsqrta = fr->invsqrta;
235 gbtabscale = _mm_set1_pd(fr->gbtab.scale);
236 gbtab = fr->gbtab.data;
237 gbinvepsdiff = _mm_set1_pd((1.0/fr->epsilon_r) - (1.0/fr->gb_epsilon_solvent));
240 /* #if 'Water' in GEOMETRY_I */
241 /* Setup water-specific parameters */
242 inr = nlist->iinr[0];
243 /* #for I in PARTICLES_ELEC_I */
244 iq{I} = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+{I}]));
246 /* #for I in PARTICLES_VDW_I */
247 vdwioffset{I} = 2*nvdwtype*vdwtype[inr+{I}];
251 /* #if 'Water' in GEOMETRY_J */
252 /* #for J in PARTICLES_ELEC_J */
253 jq{J} = _mm_set1_pd(charge[inr+{J}]);
255 /* #for J in PARTICLES_VDW_J */
256 vdwjidx{J}A = 2*vdwtype[inr+{J}];
258 /* #for I,J in PAIRS_IJ */
259 /* #if 'electrostatics' in INTERACTION_FLAGS[I][J] */
260 qq{I}{J} = _mm_mul_pd(iq{I},jq{J});
262 /* #if 'vdw' in INTERACTION_FLAGS[I][J] */
263 /* #if 'LJEwald' in KERNEL_VDW */
264 c6_{I}{J} = _mm_set1_pd(vdwparam[vdwioffset{I}+vdwjidx{J}A]);
265 c12_{I}{J} = _mm_set1_pd(vdwparam[vdwioffset{I}+vdwjidx{J}A+1]);
266 c6grid_{I}{J} = _mm_set1_pd(vdwgridparam[vdwioffset{I}+vdwjidx{J}A]);
268 c6_{I}{J} = _mm_set1_pd(vdwparam[vdwioffset{I}+vdwjidx{J}A]);
269 c12_{I}{J} = _mm_set1_pd(vdwparam[vdwioffset{I}+vdwjidx{J}A+1]);
275 /* #if KERNEL_MOD_ELEC!='None' or KERNEL_MOD_VDW!='None' */
276 /* #if KERNEL_ELEC!='None' */
277 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
278 rcutoff_scalar = fr->rcoulomb;
280 rcutoff_scalar = fr->rvdw;
282 rcutoff = _mm_set1_pd(rcutoff_scalar);
283 rcutoff2 = _mm_mul_pd(rcutoff,rcutoff);
286 /* #if KERNEL_MOD_VDW=='PotentialShift' */
287 sh_vdw_invrcut6 = _mm_set1_pd(fr->ic->sh_invrc6);
288 rvdw = _mm_set1_pd(fr->rvdw);
291 /* #if 'PotentialSwitch' in [KERNEL_MOD_ELEC,KERNEL_MOD_VDW] */
292 /* #if KERNEL_MOD_ELEC=='PotentialSwitch' */
293 rswitch_scalar = fr->rcoulomb_switch;
294 rswitch = _mm_set1_pd(rswitch_scalar);
296 rswitch_scalar = fr->rvdw_switch;
297 rswitch = _mm_set1_pd(rswitch_scalar);
299 /* Setup switch parameters */
300 d_scalar = rcutoff_scalar-rswitch_scalar;
301 d = _mm_set1_pd(d_scalar);
302 swV3 = _mm_set1_pd(-10.0/(d_scalar*d_scalar*d_scalar));
303 swV4 = _mm_set1_pd( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
304 swV5 = _mm_set1_pd( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
305 /* #if 'Force' in KERNEL_VF */
306 swF2 = _mm_set1_pd(-30.0/(d_scalar*d_scalar*d_scalar));
307 swF3 = _mm_set1_pd( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
308 swF4 = _mm_set1_pd(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
312 /* Avoid stupid compiler warnings */
317 /* ## Keep track of the floating point operations we issue for reporting! */
318 /* #define OUTERFLOPS 0 */
322 /* Start outer loop over neighborlists */
323 for(iidx=0; iidx<nri; iidx++)
325 /* Load shift vector for this list */
326 i_shift_offset = DIM*shiftidx[iidx];
328 /* Load limits for loop over neighbors */
329 j_index_start = jindex[iidx];
330 j_index_end = jindex[iidx+1];
332 /* Get outer coordinate index */
334 i_coord_offset = DIM*inr;
336 /* Load i particle coords and add shift vector */
337 /* #if GEOMETRY_I == 'Particle' */
338 gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
339 /* #elif GEOMETRY_I == 'Water3' */
340 gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
341 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
342 /* #elif GEOMETRY_I == 'Water4' */
343 /* #if 0 in PARTICLES_I */
344 gmx_mm_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
345 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
347 gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
348 &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
352 /* #if 'Force' in KERNEL_VF */
353 /* #for I in PARTICLES_I */
354 fix{I} = _mm_setzero_pd();
355 fiy{I} = _mm_setzero_pd();
356 fiz{I} = _mm_setzero_pd();
360 /* ## For water we already preloaded parameters at the start of the kernel */
361 /* #if not 'Water' in GEOMETRY_I */
362 /* Load parameters for i particles */
363 /* #for I in PARTICLES_ELEC_I */
364 iq{I} = _mm_mul_pd(facel,_mm_load1_pd(charge+inr+{I}));
365 /* #define OUTERFLOPS OUTERFLOPS+1 */
366 /* #if KERNEL_ELEC=='GeneralizedBorn' */
367 isai{I} = _mm_load1_pd(invsqrta+inr+{I});
370 /* #for I in PARTICLES_VDW_I */
371 vdwioffset{I} = 2*nvdwtype*vdwtype[inr+{I}];
375 /* #if 'Potential' in KERNEL_VF */
376 /* Reset potential sums */
377 /* #if KERNEL_ELEC != 'None' */
378 velecsum = _mm_setzero_pd();
380 /* #if 'GeneralizedBorn' in KERNEL_ELEC */
381 vgbsum = _mm_setzero_pd();
383 /* #if KERNEL_VDW != 'None' */
384 vvdwsum = _mm_setzero_pd();
387 /* #if 'GeneralizedBorn' in KERNEL_ELEC and 'Force' in KERNEL_VF */
388 dvdasum = _mm_setzero_pd();
391 /* #for ROUND in ['Loop','Epilogue'] */
393 /* #if ROUND =='Loop' */
394 /* Start inner kernel loop */
395 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
397 /* ## First round is normal loop (next statement resets indentation) */
404 /* ## Second round is epilogue */
406 /* #define INNERFLOPS 0 */
408 /* #if ROUND =='Loop' */
409 /* Get j neighbor index, and coordinate index */
412 j_coord_offsetA = DIM*jnrA;
413 j_coord_offsetB = DIM*jnrB;
415 /* load j atom coordinates */
416 /* #if GEOMETRY_J == 'Particle' */
417 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
419 /* #elif GEOMETRY_J == 'Water3' */
420 gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
421 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
422 /* #elif GEOMETRY_J == 'Water4' */
423 /* #if 0 in PARTICLES_J */
424 gmx_mm_load_4rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
425 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
426 &jy2,&jz2,&jx3,&jy3,&jz3);
428 gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
429 &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
434 j_coord_offsetA = DIM*jnrA;
436 /* load j atom coordinates */
437 /* #if GEOMETRY_J == 'Particle' */
438 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
440 /* #elif GEOMETRY_J == 'Water3' */
441 gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
442 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
443 /* #elif GEOMETRY_J == 'Water4' */
444 /* #if 0 in PARTICLES_J */
445 gmx_mm_load_4rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
446 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
447 &jy2,&jz2,&jx3,&jy3,&jz3);
449 gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA+DIM,
450 &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
455 /* Calculate displacement vector */
456 /* #for I,J in PAIRS_IJ */
457 dx{I}{J} = _mm_sub_pd(ix{I},jx{J});
458 dy{I}{J} = _mm_sub_pd(iy{I},jy{J});
459 dz{I}{J} = _mm_sub_pd(iz{I},jz{J});
460 /* #define INNERFLOPS INNERFLOPS+3 */
463 /* Calculate squared distance and things based on it */
464 /* #for I,J in PAIRS_IJ */
465 rsq{I}{J} = gmx_mm_calc_rsq_pd(dx{I}{J},dy{I}{J},dz{I}{J});
466 /* #define INNERFLOPS INNERFLOPS+5 */
469 /* #for I,J in PAIRS_IJ */
470 /* #if 'rinv' in INTERACTION_FLAGS[I][J] */
471 rinv{I}{J} = gmx_mm_invsqrt_pd(rsq{I}{J});
472 /* #define INNERFLOPS INNERFLOPS+5 */
476 /* #for I,J in PAIRS_IJ */
477 /* #if 'rinvsq' in INTERACTION_FLAGS[I][J] */
478 /* # if 'rinv' not in INTERACTION_FLAGS[I][J] */
479 rinvsq{I}{J} = gmx_mm_inv_pd(rsq{I}{J});
480 /* #define INNERFLOPS INNERFLOPS+4 */
482 rinvsq{I}{J} = _mm_mul_pd(rinv{I}{J},rinv{I}{J});
483 /* #define INNERFLOPS INNERFLOPS+1 */
488 /* #if not 'Water' in GEOMETRY_J */
489 /* Load parameters for j particles */
490 /* #for J in PARTICLES_ELEC_J */
491 /* #if ROUND =='Loop' */
492 jq{J} = gmx_mm_load_2real_swizzle_pd(charge+jnrA+{J},charge+jnrB+{J});
494 jq{J} = _mm_load_sd(charge+jnrA+{J});
496 /* #if KERNEL_ELEC=='GeneralizedBorn' */
497 /* #if ROUND =='Loop' */
498 isaj{J} = gmx_mm_load_2real_swizzle_pd(invsqrta+jnrA+{J},invsqrta+jnrB+{J});
500 isaj{J} = _mm_load_sd(invsqrta+jnrA+{J});
504 /* #for J in PARTICLES_VDW_J */
505 vdwjidx{J}A = 2*vdwtype[jnrA+{J}];
506 /* #if ROUND =='Loop' */
507 vdwjidx{J}B = 2*vdwtype[jnrB+{J}];
512 /* #if 'Force' in KERNEL_VF and not 'Particle' in GEOMETRY_I */
513 /* #for J in PARTICLES_J */
514 fjx{J} = _mm_setzero_pd();
515 fjy{J} = _mm_setzero_pd();
516 fjz{J} = _mm_setzero_pd();
520 /* #for I,J in PAIRS_IJ */
522 /**************************
523 * CALCULATE INTERACTIONS *
524 **************************/
526 /* #if 'exactcutoff' in INTERACTION_FLAGS[I][J] */
527 /* ## We always calculate rinv/rinvsq above to enable pipelineing in compilers (performance tested on x86) */
528 if (gmx_mm_any_lt(rsq{I}{J},rcutoff2))
530 /* #if 0 ## this and the next two lines is a hack to maintain auto-indentation in template file */
533 /* #define INNERFLOPS INNERFLOPS+1 */
536 /* #if 'r' in INTERACTION_FLAGS[I][J] */
537 r{I}{J} = _mm_mul_pd(rsq{I}{J},rinv{I}{J});
538 /* #define INNERFLOPS INNERFLOPS+1 */
541 /* ## For water geometries we already loaded parameters at the start of the kernel */
542 /* #if not 'Water' in GEOMETRY_J */
543 /* Compute parameters for interactions between i and j atoms */
544 /* #if 'electrostatics' in INTERACTION_FLAGS[I][J] */
545 qq{I}{J} = _mm_mul_pd(iq{I},jq{J});
546 /* #define INNERFLOPS INNERFLOPS+1 */
548 /* #if 'vdw' in INTERACTION_FLAGS[I][J] */
549 /* #if ROUND == 'Loop' */
550 gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset{I}+vdwjidx{J}A,
551 vdwparam+vdwioffset{I}+vdwjidx{J}B,&c6_{I}{J},&c12_{I}{J});
552 /* #if 'LJEwald' in KERNEL_VDW */
553 c6grid_{I}{J} = gmx_mm_load_2real_swizzle_pd(vdwgridparam+vdwioffset{I}+vdwjidx{J}A,
554 vdwgridparam+vdwioffset{I}+vdwjidx{J}B);
557 gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset{I}+vdwjidx{J}A,&c6_{I}{J},&c12_{I}{J});
558 /* #if 'LJEwald' in KERNEL_VDW */
559 c6grid_{I}{J} = gmx_mm_load_1real_pd(vdwgridparam+vdwioffset{I}+vdwjidx{J}A);
565 /* #if 'table' in INTERACTION_FLAGS[I][J] */
566 /* Calculate table index by multiplying r with table scale and truncate to integer */
567 rt = _mm_mul_pd(r{I}{J},vftabscale);
568 vfitab = _mm_cvttpd_epi32(rt);
570 vfeps = _mm_frcz_pd(rt);
572 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
574 twovfeps = _mm_add_pd(vfeps,vfeps);
575 /* #define INNERFLOPS INNERFLOPS+4 */
576 /* #if 'Table' in KERNEL_ELEC and 'Table' in KERNEL_VDW */
577 /* ## 3 tables, 4 data per point: multiply index by 12 */
578 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
579 /* #elif 'Table' in KERNEL_ELEC */
580 /* ## 1 table, 4 data per point: multiply index by 4 */
581 vfitab = _mm_slli_epi32(vfitab,2);
582 /* #elif 'Table' in KERNEL_VDW */
583 /* ## 2 tables, 4 data per point: multiply index by 8 */
584 vfitab = _mm_slli_epi32(vfitab,3);
588 /* ## ELECTROSTATIC INTERACTIONS */
589 /* #if 'electrostatics' in INTERACTION_FLAGS[I][J] */
591 /* #if KERNEL_ELEC=='Coulomb' */
593 /* COULOMB ELECTROSTATICS */
594 velec = _mm_mul_pd(qq{I}{J},rinv{I}{J});
595 /* #define INNERFLOPS INNERFLOPS+1 */
596 /* #if 'Force' in KERNEL_VF */
597 felec = _mm_mul_pd(velec,rinvsq{I}{J});
598 /* #define INNERFLOPS INNERFLOPS+2 */
601 /* #elif KERNEL_ELEC=='ReactionField' */
603 /* REACTION-FIELD ELECTROSTATICS */
604 /* #if 'Potential' in KERNEL_VF */
605 velec = _mm_mul_pd(qq{I}{J},_mm_sub_pd(_mm_macc_pd(krf,rsq{I}{J},rinv{I}{J}),crf));
606 /* #define INNERFLOPS INNERFLOPS+4 */
608 /* #if 'Force' in KERNEL_VF */
609 felec = _mm_mul_pd(qq{I}{J},_mm_msub_pd(rinv{I}{J},rinvsq{I}{J},krf2));
610 /* #define INNERFLOPS INNERFLOPS+3 */
613 /* #elif KERNEL_ELEC=='GeneralizedBorn' */
615 /* GENERALIZED BORN AND COULOMB ELECTROSTATICS */
616 isaprod = _mm_mul_pd(isai{I},isaj{J});
617 gbqqfactor = _mm_xor_pd(signbit,_mm_mul_pd(qq{I}{J},_mm_mul_pd(isaprod,gbinvepsdiff)));
618 gbscale = _mm_mul_pd(isaprod,gbtabscale);
619 /* #define INNERFLOPS INNERFLOPS+5 */
621 /* Calculate generalized born table index - this is a separate table from the normal one,
622 * but we use the same procedure by multiplying r with scale and truncating to integer.
624 rt = _mm_mul_pd(r{I}{J},gbscale);
625 gbitab = _mm_cvttpd_epi32(rt);
627 gbeps = _mm_frcz_pd(rt);
629 gbeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
631 gbitab = _mm_slli_epi32(gbitab,2);
633 Y = _mm_load_pd( gbtab + _mm_extract_epi32(gbitab,0) );
634 /* #if ROUND == 'Loop' */
635 F = _mm_load_pd( gbtab + _mm_extract_epi32(gbitab,1) );
637 F = _mm_setzero_pd();
639 GMX_MM_TRANSPOSE2_PD(Y,F);
640 G = _mm_load_pd( gbtab + _mm_extract_epi32(gbitab,0) +2);
641 /* #if ROUND == 'Loop' */
642 H = _mm_load_pd( gbtab + _mm_extract_epi32(gbitab,1) +2);
644 H = _mm_setzero_pd();
646 GMX_MM_TRANSPOSE2_PD(G,H);
647 Fp = _mm_macc_pd(gbeps,_mm_macc_pd(gbeps,H,G),F);
648 VV = _mm_macc_pd(gbeps,Fp,Y);
649 vgb = _mm_mul_pd(gbqqfactor,VV);
650 /* #define INNERFLOPS INNERFLOPS+10 */
652 /* #if 'Force' in KERNEL_VF */
653 twogbeps = _mm_add_pd(gbeps,gbeps);
654 FF = _mm_macc_pd(_mm_macc_pd(twogbeps,H,G),gbeps,Fp);
655 fgb = _mm_mul_pd(gbqqfactor,_mm_mul_pd(FF,gbscale));
656 dvdatmp = _mm_mul_pd(minushalf,_mm_macc_pd(fgb,r{I}{J},vgb));
657 /* #if ROUND == 'Epilogue' */
658 dvdatmp = _mm_unpacklo_pd(dvdatmp,_mm_setzero_pd());
660 dvdasum = _mm_add_pd(dvdasum,dvdatmp);
661 /* #if ROUND == 'Loop' */
662 gmx_mm_increment_2real_swizzle_pd(dvda+jnrA,dvda+jnrB,_mm_mul_pd(dvdatmp,_mm_mul_pd(isaj{J},isaj{J})));
664 gmx_mm_increment_1real_pd(dvda+jnrA,_mm_mul_pd(dvdatmp,_mm_mul_pd(isaj{J},isaj{J})));
666 /* #define INNERFLOPS INNERFLOPS+13 */
668 velec = _mm_mul_pd(qq{I}{J},rinv{I}{J});
669 /* #define INNERFLOPS INNERFLOPS+1 */
670 /* #if 'Force' in KERNEL_VF */
671 felec = _mm_mul_pd(_mm_msub_pd(velec,rinv{I}{J},fgb),rinv{I}{J});
672 /* #define INNERFLOPS INNERFLOPS+3 */
675 /* #elif KERNEL_ELEC=='Ewald' */
676 /* EWALD ELECTROSTATICS */
678 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
679 ewrt = _mm_mul_pd(r{I}{J},ewtabscale);
680 ewitab = _mm_cvttpd_epi32(ewrt);
682 eweps = _mm_frcz_pd(ewrt);
684 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
686 twoeweps = _mm_add_pd(eweps,eweps);
687 /* #define INNERFLOPS INNERFLOPS+4 */
688 /* #if 'Potential' in KERNEL_VF or KERNEL_MOD_ELEC=='PotentialSwitch' */
689 ewitab = _mm_slli_epi32(ewitab,2);
690 ewtabF = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
691 /* #if ROUND == 'Loop' */
692 ewtabD = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
694 ewtabD = _mm_setzero_pd();
696 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
697 ewtabV = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
698 /* #if ROUND == 'Loop' */
699 ewtabFn = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
701 ewtabFn = _mm_setzero_pd();
703 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
704 felec = _mm_macc_pd(eweps,ewtabD,ewtabF);
705 /* #define INNERFLOPS INNERFLOPS+2 */
706 /* #if KERNEL_MOD_ELEC=='PotentialShift' */
707 velec = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
708 velec = _mm_mul_pd(qq{I}{J},_mm_sub_pd(_mm_sub_pd(rinv{I}{J},sh_ewald),velec));
709 /* #define INNERFLOPS INNERFLOPS+7 */
711 velec = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
712 velec = _mm_mul_pd(qq{I}{J},_mm_sub_pd(rinv{I}{J},velec));
713 /* #define INNERFLOPS INNERFLOPS+6 */
715 /* #if 'Force' in KERNEL_VF */
716 felec = _mm_mul_pd(_mm_mul_pd(qq{I}{J},rinv{I}{J}),_mm_sub_pd(rinvsq{I}{J},felec));
717 /* #define INNERFLOPS INNERFLOPS+3 */
719 /* #elif KERNEL_VF=='Force' */
720 /* #if ROUND == 'Loop' */
721 gmx_mm_load_2pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),ewtab+_mm_extract_epi32(ewitab,1),
724 gmx_mm_load_1pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
726 felec = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
727 felec = _mm_mul_pd(_mm_mul_pd(qq{I}{J},rinv{I}{J}),_mm_sub_pd(rinvsq{I}{J},felec));
728 /* #define INNERFLOPS INNERFLOPS+7 */
731 /* #elif KERNEL_ELEC=='CubicSplineTable' */
733 /* CUBIC SPLINE TABLE ELECTROSTATICS */
734 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
735 /* #if ROUND == 'Loop' */
736 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
738 F = _mm_setzero_pd();
740 GMX_MM_TRANSPOSE2_PD(Y,F);
741 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
742 /* #if ROUND == 'Loop' */
743 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
745 H = _mm_setzero_pd();
747 GMX_MM_TRANSPOSE2_PD(G,H);
748 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
749 /* #define INNERFLOPS INNERFLOPS+4 */
750 /* #if 'Potential' in KERNEL_VF */
751 VV = _mm_macc_pd(vfeps,Fp,Y);
752 velec = _mm_mul_pd(qq{I}{J},VV);
753 /* #define INNERFLOPS INNERFLOPS+3 */
755 /* #if 'Force' in KERNEL_VF */
756 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
757 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq{I}{J},FF),_mm_mul_pd(vftabscale,rinv{I}{J})));
758 /* #define INNERFLOPS INNERFLOPS+7 */
761 /* ## End of check for electrostatics interaction forms */
763 /* ## END OF ELECTROSTATIC INTERACTION CHECK FOR PAIR I-J */
765 /* #if 'vdw' in INTERACTION_FLAGS[I][J] */
767 /* #if KERNEL_VDW=='LennardJones' */
769 /* LENNARD-JONES DISPERSION/REPULSION */
771 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq{I}{J},rinvsq{I}{J}),rinvsq{I}{J});
772 /* #define INNERFLOPS INNERFLOPS+2 */
773 /* #if 'Potential' in KERNEL_VF or KERNEL_MOD_VDW=='PotentialSwitch' */
774 vvdw6 = _mm_mul_pd(c6_{I}{J},rinvsix);
775 vvdw12 = _mm_mul_pd(c12_{I}{J},_mm_mul_pd(rinvsix,rinvsix));
776 /* #define INNERFLOPS INNERFLOPS+3 */
777 /* #if KERNEL_MOD_VDW=='PotentialShift' */
778 vvdw = _mm_msub_pd(_mm_nmacc_pd(c12_{I}{J},_mm_mul_pd(sh_vdw_invrcut6,sh_vdw_invrcut6),vvdw12),one_twelfth,
779 _mm_mul_pd(_mm_nmacc_pd( c6_{I}{J},sh_vdw_invrcut6,vvdw6),one_sixth));
780 /* #define INNERFLOPS INNERFLOPS+8 */
782 vvdw = _mm_msub_pd( vvdw12,one_twelfth, _mm_mul_pd(vvdw6,one_sixth) );
783 /* #define INNERFLOPS INNERFLOPS+3 */
785 /* ## Check for force inside potential check, i.e. this means we already did the potential part */
786 /* #if 'Force' in KERNEL_VF */
787 fvdw = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq{I}{J});
788 /* #define INNERFLOPS INNERFLOPS+2 */
790 /* #elif KERNEL_VF=='Force' */
791 /* ## Force-only LennardJones makes it possible to save 1 flop (they do add up...) */
792 fvdw = _mm_mul_pd(_mm_msub_pd(c12_{I}{J},rinvsix,c6_{I}{J}),_mm_mul_pd(rinvsix,rinvsq{I}{J}));
793 /* #define INNERFLOPS INNERFLOPS+4 */
796 /* #elif KERNEL_VDW=='CubicSplineTable' */
798 /* CUBIC SPLINE TABLE DISPERSION */
799 /* #if 'Table' in KERNEL_ELEC */
800 vfitab = _mm_add_epi32(vfitab,ifour);
802 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
803 /* #if ROUND == 'Loop' */
804 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
806 F = _mm_setzero_pd();
808 GMX_MM_TRANSPOSE2_PD(Y,F);
809 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
810 /* #if ROUND == 'Loop' */
811 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
813 H = _mm_setzero_pd();
815 GMX_MM_TRANSPOSE2_PD(G,H);
816 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(H,vfeps,G),F);
817 /* #define INNERFLOPS INNERFLOPS+4 */
818 /* #if 'Potential' in KERNEL_VF */
819 VV = _mm_macc_pd(vfeps,Fp,Y);
820 vvdw6 = _mm_mul_pd(c6_{I}{J},VV);
821 /* #define INNERFLOPS INNERFLOPS+3 */
823 /* #if 'Force' in KERNEL_VF */
824 FF = _mm_macc_pd(vfeps,_mm_macc_pd(twovfeps,H,G),Fp);
825 fvdw6 = _mm_mul_pd(c6_{I}{J},FF);
826 /* #define INNERFLOPS INNERFLOPS+4 */
829 /* CUBIC SPLINE TABLE REPULSION */
830 vfitab = _mm_add_epi32(vfitab,ifour);
831 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
832 /* #if ROUND == 'Loop' */
833 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
835 F = _mm_setzero_pd();
837 GMX_MM_TRANSPOSE2_PD(Y,F);
838 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
839 /* #if ROUND == 'Loop' */
840 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
842 H = _mm_setzero_pd();
844 GMX_MM_TRANSPOSE2_PD(G,H);
845 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(H,vfeps,G),F);
846 /* #define INNERFLOPS INNERFLOPS+4 */
847 /* #if 'Potential' in KERNEL_VF */
848 VV = _mm_macc_pd(vfeps,Fp,Y);
849 vvdw12 = _mm_mul_pd(c12_{I}{J},VV);
850 /* #define INNERFLOPS INNERFLOPS+3 */
852 /* #if 'Force' in KERNEL_VF */
853 FF = _mm_macc_pd(vfeps,_mm_macc_pd(twovfeps,H,G),Fp);
854 fvdw12 = _mm_mul_pd(c12_{I}{J},FF);
855 /* #define INNERFLOPS INNERFLOPS+5 */
857 /* #if 'Potential' in KERNEL_VF */
858 vvdw = _mm_add_pd(vvdw12,vvdw6);
859 /* #define INNERFLOPS INNERFLOPS+1 */
861 /* #if 'Force' in KERNEL_VF */
862 fvdw = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv{I}{J})));
863 /* #define INNERFLOPS INNERFLOPS+4 */
866 /* #elif KERNEL_VDW=='LJEwald' */
868 /* Analytical LJ-PME */
869 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq{I}{J},rinvsq{I}{J}),rinvsq{I}{J});
870 ewcljrsq = _mm_mul_pd(ewclj2,rsq{I}{J});
871 ewclj6 = _mm_mul_pd(ewclj2,_mm_mul_pd(ewclj2,ewclj2));
872 exponent = gmx_simd_exp_d(ewcljrsq);
873 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
874 poly = _mm_mul_pd(exponent,_mm_macc_pd(_mm_mul_pd(ewcljrsq,ewcljrsq),one_half,_mm_sub_pd(one,ewcljrsq)));
875 /* #define INNERFLOPS INNERFLOPS+10 */
876 /* #if 'Potential' in KERNEL_VF or KERNEL_MOD_VDW=='PotentialSwitch' */
877 /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
878 vvdw6 = _mm_mul_pd(_mm_macc_pd(-c6grid_{I}{J},_mm_sub_pd(one,poly),c6_{I}{J}),rinvsix);
879 vvdw12 = _mm_mul_pd(c12_{I}{J},_mm_mul_pd(rinvsix,rinvsix));
880 /* #define INNERFLOPS INNERFLOPS+5 */
881 /* #if KERNEL_MOD_VDW=='PotentialShift' */
882 vvdw = _mm_msub_pd(_mm_nmacc_pd(c12_{I}{J},_mm_mul_pd(sh_vdw_invrcut6,sh_vdw_invrcut6),vvdw12),one_twelfth,
883 _mm_mul_pd(_mm_sub_pd(vvdw6,_mm_macc_pd(c6grid_{I}{J},sh_lj_ewald,_mm_mul_pd(c6_{I}{J},sh_vdw_invrcut6))),one_sixth));
884 /* #define INNERFLOPS INNERFLOPS+6 */
886 vvdw = _mm_msub_pd(vvdw12,one_twelfth,_mm_mul_pd(vvdw6,one_sixth));
887 /* #define INNERFLOPS INNERFLOPS+2 */
889 /* ## Check for force inside potential check, i.e. this means we already did the potential part */
890 /* #if 'Force' in KERNEL_VF */
891 /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
892 fvdw = _mm_mul_pd(_mm_add_pd(vvdw12,_mm_msub_pd(_mm_mul_pd(c6grid_{I}{J},one_sixth),_mm_mul_pd(exponent,ewclj6),vvdw6)),rinvsq{I}{J});
893 /* #define INNERFLOPS INNERFLOPS+5 */
895 /* #elif KERNEL_VF=='Force' */
896 /* f6A = 6 * C6grid * (1 - poly) */
897 f6A = _mm_mul_pd(c6grid_{I}{J},_mm_sub_pd(one,poly));
898 /* f6B = C6grid * exponent * beta^6 */
899 f6B = _mm_mul_pd(_mm_mul_pd(c6grid_{I}{J},one_sixth),_mm_mul_pd(exponent,ewclj6));
900 /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
901 fvdw = _mm_mul_pd(_mm_macc_pd(_mm_msub_pd(c12_{I}{J},rinvsix,_mm_sub_pd(c6_{I}{J},f6A)),rinvsix,f6B),rinvsq{I}{J});
902 /* #define INNERFLOPS INNERFLOPS+10 */
905 /* ## End of check for vdw interaction forms */
907 /* ## END OF VDW INTERACTION CHECK FOR PAIR I-J */
909 /* #if 'switch' in INTERACTION_FLAGS[I][J] */
910 d = _mm_sub_pd(r{I}{J},rswitch);
911 d = _mm_max_pd(d,_mm_setzero_pd());
912 d2 = _mm_mul_pd(d,d);
913 sw = _mm_add_pd(one,_mm_mul_pd(d2,_mm_mul_pd(d,_mm_macc_pd(d,_mm_macc_pd(d,swV5,swV4),swV3))));
914 /* #define INNERFLOPS INNERFLOPS+10 */
916 /* #if 'Force' in KERNEL_VF */
917 dsw = _mm_mul_pd(d2,_mm_macc_pd(d,_mm_macc_pd(d,swF4,swF3),swF2));
918 /* #define INNERFLOPS INNERFLOPS+5 */
921 /* Evaluate switch function */
922 /* #if 'Force' in KERNEL_VF */
923 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
924 /* #if 'electrostatics' in INTERACTION_FLAGS[I][J] and KERNEL_MOD_ELEC=='PotentialSwitch' */
925 felec = _mm_msub_pd( felec,sw , _mm_mul_pd(rinv{I}{J},_mm_mul_pd(velec,dsw)) );
926 /* #define INNERFLOPS INNERFLOPS+4 */
928 /* #if 'vdw' in INTERACTION_FLAGS[I][J] and KERNEL_MOD_VDW=='PotentialSwitch' */
929 fvdw = _mm_msub_pd( fvdw,sw , _mm_mul_pd(rinv{I}{J},_mm_mul_pd(vvdw,dsw)) );
930 /* #define INNERFLOPS INNERFLOPS+4 */
933 /* #if 'Potential' in KERNEL_VF */
934 /* #if 'electrostatics' in INTERACTION_FLAGS[I][J] and KERNEL_MOD_ELEC=='PotentialSwitch' */
935 velec = _mm_mul_pd(velec,sw);
936 /* #define INNERFLOPS INNERFLOPS+1 */
938 /* #if 'vdw' in INTERACTION_FLAGS[I][J] and KERNEL_MOD_VDW=='PotentialSwitch' */
939 vvdw = _mm_mul_pd(vvdw,sw);
940 /* #define INNERFLOPS INNERFLOPS+1 */
944 /* #if 'exactcutoff' in INTERACTION_FLAGS[I][J] */
945 cutoff_mask = _mm_cmplt_pd(rsq{I}{J},rcutoff2);
946 /* #define INNERFLOPS INNERFLOPS+1 */
949 /* #if 'Potential' in KERNEL_VF */
950 /* Update potential sum for this i atom from the interaction with this j atom. */
951 /* #if 'electrostatics' in INTERACTION_FLAGS[I][J] */
952 /* #if 'exactcutoff' in INTERACTION_FLAGS[I][J] */
953 velec = _mm_and_pd(velec,cutoff_mask);
954 /* #define INNERFLOPS INNERFLOPS+1 */
956 /* #if ROUND == 'Epilogue' */
957 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
959 velecsum = _mm_add_pd(velecsum,velec);
960 /* #define INNERFLOPS INNERFLOPS+1 */
961 /* #if KERNEL_ELEC=='GeneralizedBorn' */
962 /* #if 'exactcutoff' in INTERACTION_FLAGS[I][J] */
963 vgb = _mm_and_pd(vgb,cutoff_mask);
964 /* #define INNERFLOPS INNERFLOPS+1 */
966 /* #if ROUND == 'Epilogue' */
967 vgb = _mm_unpacklo_pd(vgb,_mm_setzero_pd());
969 vgbsum = _mm_add_pd(vgbsum,vgb);
970 /* #define INNERFLOPS INNERFLOPS+1 */
973 /* #if 'vdw' in INTERACTION_FLAGS[I][J] */
974 /* #if 'exactcutoff' in INTERACTION_FLAGS[I][J] */
975 vvdw = _mm_and_pd(vvdw,cutoff_mask);
976 /* #define INNERFLOPS INNERFLOPS+1 */
978 /* #if ROUND == 'Epilogue' */
979 vvdw = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
981 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
982 /* #define INNERFLOPS INNERFLOPS+1 */
986 /* #if 'Force' in KERNEL_VF */
988 /* #if 'electrostatics' in INTERACTION_FLAGS[I][J] and 'vdw' in INTERACTION_FLAGS[I][J] */
989 fscal = _mm_add_pd(felec,fvdw);
990 /* #define INNERFLOPS INNERFLOPS+1 */
991 /* #elif 'electrostatics' in INTERACTION_FLAGS[I][J] */
993 /* #elif 'vdw' in INTERACTION_FLAGS[I][J] */
997 /* #if 'exactcutoff' in INTERACTION_FLAGS[I][J] */
998 fscal = _mm_and_pd(fscal,cutoff_mask);
999 /* #define INNERFLOPS INNERFLOPS+1 */
1002 /* #if ROUND == 'Epilogue' */
1003 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1006 /* ## Construction of vectorial force built into FMA instructions now */
1007 /* #define INNERFLOPS INNERFLOPS+3 */
1009 /* Update vectorial force */
1010 fix{I} = _mm_macc_pd(dx{I}{J},fscal,fix{I});
1011 fiy{I} = _mm_macc_pd(dy{I}{J},fscal,fiy{I});
1012 fiz{I} = _mm_macc_pd(dz{I}{J},fscal,fiz{I});
1013 /* #define INNERFLOPS INNERFLOPS+6 */
1015 /* #if GEOMETRY_I == 'Particle' */
1016 /* #if ROUND == 'Loop' */
1017 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,
1018 _mm_mul_pd(dx{I}{J},fscal),
1019 _mm_mul_pd(dy{I}{J},fscal),
1020 _mm_mul_pd(dz{I}{J},fscal));
1022 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,
1023 _mm_mul_pd(dx{I}{J},fscal),
1024 _mm_mul_pd(dy{I}{J},fscal),
1025 _mm_mul_pd(dz{I}{J},fscal));
1027 /* #define INNERFLOPS INNERFLOPS+3 */
1029 fjx{J} = _mm_macc_pd(dx{I}{J},fscal,fjx{J});
1030 fjy{J} = _mm_macc_pd(dy{I}{J},fscal,fjy{J});
1031 fjz{J} = _mm_macc_pd(dz{I}{J},fscal,fjz{J});
1032 /* #define INNERFLOPS INNERFLOPS+3 */
1037 /* #if 'exactcutoff' in INTERACTION_FLAGS[I][J] */
1038 /* #if 0 ## This and next two lines is a hack to maintain indentation in template file */
1043 /* ## End of check for the interaction being outside the cutoff */
1046 /* ## End of loop over i-j interaction pairs */
1048 /* #if 'Water' in GEOMETRY_I and GEOMETRY_J == 'Particle' */
1049 /* #if ROUND == 'Loop' */
1050 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
1052 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1054 /* #define INNERFLOPS INNERFLOPS+3 */
1055 /* #elif GEOMETRY_J == 'Water3' */
1056 /* #if ROUND == 'Loop' */
1057 gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1059 gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1061 /* #define INNERFLOPS INNERFLOPS+9 */
1062 /* #elif GEOMETRY_J == 'Water4' */
1063 /* #if 0 in PARTICLES_J */
1064 /* #if ROUND == 'Loop' */
1065 gmx_mm_decrement_4rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1067 gmx_mm_decrement_4rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1069 /* #define INNERFLOPS INNERFLOPS+12 */
1071 /* #if ROUND == 'Loop' */
1072 gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA+DIM,f+j_coord_offsetB+DIM,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1074 gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA+DIM,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1076 /* #define INNERFLOPS INNERFLOPS+9 */
1080 /* Inner loop uses {INNERFLOPS} flops */
1085 /* End of innermost loop */
1087 /* #if 'Force' in KERNEL_VF */
1088 /* #if GEOMETRY_I == 'Particle' */
1089 gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
1090 f+i_coord_offset,fshift+i_shift_offset);
1091 /* #define OUTERFLOPS OUTERFLOPS+6 */
1092 /* #elif GEOMETRY_I == 'Water3' */
1093 gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1094 f+i_coord_offset,fshift+i_shift_offset);
1095 /* #define OUTERFLOPS OUTERFLOPS+18 */
1096 /* #elif GEOMETRY_I == 'Water4' */
1097 /* #if 0 in PARTICLES_I */
1098 gmx_mm_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1099 f+i_coord_offset,fshift+i_shift_offset);
1100 /* #define OUTERFLOPS OUTERFLOPS+24 */
1102 gmx_mm_update_iforce_3atom_swizzle_pd(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1103 f+i_coord_offset+DIM,fshift+i_shift_offset);
1104 /* #define OUTERFLOPS OUTERFLOPS+18 */
1109 /* #if 'Potential' in KERNEL_VF */
1111 /* Update potential energies */
1112 /* #if KERNEL_ELEC != 'None' */
1113 gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1114 /* #define OUTERFLOPS OUTERFLOPS+1 */
1116 /* #if 'GeneralizedBorn' in KERNEL_ELEC */
1117 gmx_mm_update_1pot_pd(vgbsum,kernel_data->energygrp_polarization+ggid);
1118 /* #define OUTERFLOPS OUTERFLOPS+1 */
1120 /* #if KERNEL_VDW != 'None' */
1121 gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1122 /* #define OUTERFLOPS OUTERFLOPS+1 */
1125 /* #if 'GeneralizedBorn' in KERNEL_ELEC and 'Force' in KERNEL_VF */
1126 dvdasum = _mm_mul_pd(dvdasum, _mm_mul_pd(isai{I},isai{I}));
1127 gmx_mm_update_1pot_pd(dvdasum,dvda+inr);
1130 /* Increment number of inner iterations */
1131 inneriter += j_index_end - j_index_start;
1133 /* Outer loop uses {OUTERFLOPS} flops */
1136 /* Increment number of outer iterations */
1139 /* Update outer/inner flops */
1140 /* ## NB: This is not important, it just affects the flopcount. However, since our preprocessor is */
1141 /* ## primitive and replaces aggressively even in strings inside these directives, we need to */
1142 /* ## assemble the main part of the name (containing KERNEL/ELEC/VDW) directly in the source. */
1143 /* #if GEOMETRY_I == 'Water3' */
1144 /* #define ISUFFIX '_W3' */
1145 /* #elif GEOMETRY_I == 'Water4' */
1146 /* #define ISUFFIX '_W4' */
1148 /* #define ISUFFIX '' */
1150 /* #if GEOMETRY_J == 'Water3' */
1151 /* #define JSUFFIX 'W3' */
1152 /* #elif GEOMETRY_J == 'Water4' */
1153 /* #define JSUFFIX 'W4' */
1155 /* #define JSUFFIX '' */
1157 /* #if 'PotentialAndForce' in KERNEL_VF */
1158 /* #define VFSUFFIX '_VF' */
1159 /* #elif 'Potential' in KERNEL_VF */
1160 /* #define VFSUFFIX '_V' */
1162 /* #define VFSUFFIX '_F' */
1165 /* #if KERNEL_ELEC != 'None' and KERNEL_VDW != 'None' */
1166 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW{ISUFFIX}{JSUFFIX}{VFSUFFIX},outeriter*{OUTERFLOPS} + inneriter*{INNERFLOPS});
1167 /* #elif KERNEL_ELEC != 'None' */
1168 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC{ISUFFIX}{JSUFFIX}{VFSUFFIX},outeriter*{OUTERFLOPS} + inneriter*{INNERFLOPS});
1170 inc_nrnb(nrnb,eNR_NBKERNEL_VDW{ISUFFIX}{JSUFFIX}{VFSUFFIX},outeriter*{OUTERFLOPS} + inneriter*{INNERFLOPS});