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.
44 #include "../nb_kernel.h"
45 #include "types/simple.h"
49 #include "gromacs/simd/math_x86_avx_128_fma_single.h"
50 #include "kernelutil_x86_avx_128_fma_single.h"
53 * Gromacs nonbonded kernel: nb_kernel_ElecEw_VdwLJEw_GeomP1P1_VF_avx_128_fma_single
54 * Electrostatics interaction: Ewald
55 * VdW interaction: LJEwald
56 * Geometry: Particle-Particle
57 * Calculate force/pot: PotentialAndForce
60 nb_kernel_ElecEw_VdwLJEw_GeomP1P1_VF_avx_128_fma_single
61 (t_nblist * gmx_restrict nlist,
62 rvec * gmx_restrict xx,
63 rvec * gmx_restrict ff,
64 t_forcerec * gmx_restrict fr,
65 t_mdatoms * gmx_restrict mdatoms,
66 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
67 t_nrnb * gmx_restrict nrnb)
69 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
70 * just 0 for non-waters.
71 * Suffixes A,B,C,D refer to j loop unrolling done with AVX_128, e.g. for the four different
72 * jnr indices corresponding to data put in the four positions in the SIMD register.
74 int i_shift_offset,i_coord_offset,outeriter,inneriter;
75 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
76 int jnrA,jnrB,jnrC,jnrD;
77 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
78 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
79 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
81 real *shiftvec,*fshift,*x,*f;
82 real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
84 __m128 fscal,rcutoff,rcutoff2,jidxall;
86 __m128 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
87 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
88 __m128 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
89 __m128 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
90 __m128 velec,felec,velecsum,facel,crf,krf,krf2;
93 __m128 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
96 __m128 one_sixth = _mm_set1_ps(1.0/6.0);
97 __m128 one_twelfth = _mm_set1_ps(1.0/12.0);
100 __m128 ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
101 __m128 one_half = _mm_set1_ps(0.5);
102 __m128 minus_one = _mm_set1_ps(-1.0);
104 __m128 ewtabscale,eweps,twoeweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
105 __m128 beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
107 __m128 dummy_mask,cutoff_mask;
108 __m128 signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
109 __m128 one = _mm_set1_ps(1.0);
110 __m128 two = _mm_set1_ps(2.0);
116 jindex = nlist->jindex;
118 shiftidx = nlist->shift;
120 shiftvec = fr->shift_vec[0];
121 fshift = fr->fshift[0];
122 facel = _mm_set1_ps(fr->epsfac);
123 charge = mdatoms->chargeA;
124 nvdwtype = fr->ntype;
126 vdwtype = mdatoms->typeA;
127 vdwgridparam = fr->ljpme_c6grid;
128 sh_lj_ewald = _mm_set1_ps(fr->ic->sh_lj_ewald);
129 ewclj = _mm_set1_ps(fr->ewaldcoeff_lj);
130 ewclj2 = _mm_mul_ps(minus_one,_mm_mul_ps(ewclj,ewclj));
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 /* Avoid stupid compiler warnings */
141 jnrA = jnrB = jnrC = jnrD = 0;
150 for(iidx=0;iidx<4*DIM;iidx++)
155 /* Start outer loop over neighborlists */
156 for(iidx=0; iidx<nri; iidx++)
158 /* Load shift vector for this list */
159 i_shift_offset = DIM*shiftidx[iidx];
161 /* Load limits for loop over neighbors */
162 j_index_start = jindex[iidx];
163 j_index_end = jindex[iidx+1];
165 /* Get outer coordinate index */
167 i_coord_offset = DIM*inr;
169 /* Load i particle coords and add shift vector */
170 gmx_mm_load_shift_and_1rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
172 fix0 = _mm_setzero_ps();
173 fiy0 = _mm_setzero_ps();
174 fiz0 = _mm_setzero_ps();
176 /* Load parameters for i particles */
177 iq0 = _mm_mul_ps(facel,_mm_load1_ps(charge+inr+0));
178 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
180 /* Reset potential sums */
181 velecsum = _mm_setzero_ps();
182 vvdwsum = _mm_setzero_ps();
184 /* Start inner kernel loop */
185 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
188 /* Get j neighbor index, and coordinate index */
193 j_coord_offsetA = DIM*jnrA;
194 j_coord_offsetB = DIM*jnrB;
195 j_coord_offsetC = DIM*jnrC;
196 j_coord_offsetD = DIM*jnrD;
198 /* load j atom coordinates */
199 gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
200 x+j_coord_offsetC,x+j_coord_offsetD,
203 /* Calculate displacement vector */
204 dx00 = _mm_sub_ps(ix0,jx0);
205 dy00 = _mm_sub_ps(iy0,jy0);
206 dz00 = _mm_sub_ps(iz0,jz0);
208 /* Calculate squared distance and things based on it */
209 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
211 rinv00 = gmx_mm_invsqrt_ps(rsq00);
213 rinvsq00 = _mm_mul_ps(rinv00,rinv00);
215 /* Load parameters for j particles */
216 jq0 = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
217 charge+jnrC+0,charge+jnrD+0);
218 vdwjidx0A = 2*vdwtype[jnrA+0];
219 vdwjidx0B = 2*vdwtype[jnrB+0];
220 vdwjidx0C = 2*vdwtype[jnrC+0];
221 vdwjidx0D = 2*vdwtype[jnrD+0];
223 /**************************
224 * CALCULATE INTERACTIONS *
225 **************************/
227 r00 = _mm_mul_ps(rsq00,rinv00);
229 /* Compute parameters for interactions between i and j atoms */
230 qq00 = _mm_mul_ps(iq0,jq0);
231 gmx_mm_load_4pair_swizzle_ps(vdwparam+vdwioffset0+vdwjidx0A,
232 vdwparam+vdwioffset0+vdwjidx0B,
233 vdwparam+vdwioffset0+vdwjidx0C,
234 vdwparam+vdwioffset0+vdwjidx0D,
237 c6grid_00 = gmx_mm_load_4real_swizzle_ps(vdwgridparam+vdwioffset0+vdwjidx0A,
238 vdwgridparam+vdwioffset0+vdwjidx0B,
239 vdwgridparam+vdwioffset0+vdwjidx0C,
240 vdwgridparam+vdwioffset0+vdwjidx0D);
242 /* EWALD ELECTROSTATICS */
244 /* Analytical PME correction */
245 zeta2 = _mm_mul_ps(beta2,rsq00);
246 rinv3 = _mm_mul_ps(rinvsq00,rinv00);
247 pmecorrF = gmx_mm_pmecorrF_ps(zeta2);
248 felec = _mm_macc_ps(pmecorrF,beta3,rinv3);
249 felec = _mm_mul_ps(qq00,felec);
250 pmecorrV = gmx_mm_pmecorrV_ps(zeta2);
251 velec = _mm_nmacc_ps(pmecorrV,beta,rinv00);
252 velec = _mm_mul_ps(qq00,velec);
254 /* Analytical LJ-PME */
255 rinvsix = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00);
256 ewcljrsq = _mm_mul_ps(ewclj2,rsq00);
257 ewclj6 = _mm_mul_ps(ewclj2,_mm_mul_ps(ewclj2,ewclj2));
258 exponent = gmx_simd_exp_r(ewcljrsq);
259 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
260 poly = _mm_mul_ps(exponent,_mm_macc_ps(_mm_mul_ps(ewcljrsq,ewcljrsq),one_half,_mm_sub_ps(one,ewcljrsq)));
261 /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
262 vvdw6 = _mm_mul_ps(_mm_macc_ps(-c6grid_00,_mm_sub_ps(one,poly),c6_00),rinvsix);
263 vvdw12 = _mm_mul_ps(c12_00,_mm_mul_ps(rinvsix,rinvsix));
264 vvdw = _mm_msub_ps(vvdw12,one_twelfth,_mm_mul_ps(vvdw6,one_sixth));
265 /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
266 fvdw = _mm_mul_ps(_mm_add_ps(vvdw12,_mm_msub_ps(_mm_mul_ps(c6grid_00,one_sixth),_mm_mul_ps(exponent,ewclj6),vvdw6)),rinvsq00);
268 /* Update potential sum for this i atom from the interaction with this j atom. */
269 velecsum = _mm_add_ps(velecsum,velec);
270 vvdwsum = _mm_add_ps(vvdwsum,vvdw);
272 fscal = _mm_add_ps(felec,fvdw);
274 /* Update vectorial force */
275 fix0 = _mm_macc_ps(dx00,fscal,fix0);
276 fiy0 = _mm_macc_ps(dy00,fscal,fiy0);
277 fiz0 = _mm_macc_ps(dz00,fscal,fiz0);
279 fjptrA = f+j_coord_offsetA;
280 fjptrB = f+j_coord_offsetB;
281 fjptrC = f+j_coord_offsetC;
282 fjptrD = f+j_coord_offsetD;
283 gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
284 _mm_mul_ps(dx00,fscal),
285 _mm_mul_ps(dy00,fscal),
286 _mm_mul_ps(dz00,fscal));
288 /* Inner loop uses 53 flops */
294 /* Get j neighbor index, and coordinate index */
295 jnrlistA = jjnr[jidx];
296 jnrlistB = jjnr[jidx+1];
297 jnrlistC = jjnr[jidx+2];
298 jnrlistD = jjnr[jidx+3];
299 /* Sign of each element will be negative for non-real atoms.
300 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
301 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
303 dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
304 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
305 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
306 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
307 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
308 j_coord_offsetA = DIM*jnrA;
309 j_coord_offsetB = DIM*jnrB;
310 j_coord_offsetC = DIM*jnrC;
311 j_coord_offsetD = DIM*jnrD;
313 /* load j atom coordinates */
314 gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
315 x+j_coord_offsetC,x+j_coord_offsetD,
318 /* Calculate displacement vector */
319 dx00 = _mm_sub_ps(ix0,jx0);
320 dy00 = _mm_sub_ps(iy0,jy0);
321 dz00 = _mm_sub_ps(iz0,jz0);
323 /* Calculate squared distance and things based on it */
324 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
326 rinv00 = gmx_mm_invsqrt_ps(rsq00);
328 rinvsq00 = _mm_mul_ps(rinv00,rinv00);
330 /* Load parameters for j particles */
331 jq0 = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
332 charge+jnrC+0,charge+jnrD+0);
333 vdwjidx0A = 2*vdwtype[jnrA+0];
334 vdwjidx0B = 2*vdwtype[jnrB+0];
335 vdwjidx0C = 2*vdwtype[jnrC+0];
336 vdwjidx0D = 2*vdwtype[jnrD+0];
338 /**************************
339 * CALCULATE INTERACTIONS *
340 **************************/
342 r00 = _mm_mul_ps(rsq00,rinv00);
343 r00 = _mm_andnot_ps(dummy_mask,r00);
345 /* Compute parameters for interactions between i and j atoms */
346 qq00 = _mm_mul_ps(iq0,jq0);
347 gmx_mm_load_4pair_swizzle_ps(vdwparam+vdwioffset0+vdwjidx0A,
348 vdwparam+vdwioffset0+vdwjidx0B,
349 vdwparam+vdwioffset0+vdwjidx0C,
350 vdwparam+vdwioffset0+vdwjidx0D,
353 c6grid_00 = gmx_mm_load_4real_swizzle_ps(vdwgridparam+vdwioffset0+vdwjidx0A,
354 vdwgridparam+vdwioffset0+vdwjidx0B,
355 vdwgridparam+vdwioffset0+vdwjidx0C,
356 vdwgridparam+vdwioffset0+vdwjidx0D);
358 /* EWALD ELECTROSTATICS */
360 /* Analytical PME correction */
361 zeta2 = _mm_mul_ps(beta2,rsq00);
362 rinv3 = _mm_mul_ps(rinvsq00,rinv00);
363 pmecorrF = gmx_mm_pmecorrF_ps(zeta2);
364 felec = _mm_macc_ps(pmecorrF,beta3,rinv3);
365 felec = _mm_mul_ps(qq00,felec);
366 pmecorrV = gmx_mm_pmecorrV_ps(zeta2);
367 velec = _mm_nmacc_ps(pmecorrV,beta,rinv00);
368 velec = _mm_mul_ps(qq00,velec);
370 /* Analytical LJ-PME */
371 rinvsix = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00);
372 ewcljrsq = _mm_mul_ps(ewclj2,rsq00);
373 ewclj6 = _mm_mul_ps(ewclj2,_mm_mul_ps(ewclj2,ewclj2));
374 exponent = gmx_simd_exp_r(ewcljrsq);
375 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
376 poly = _mm_mul_ps(exponent,_mm_macc_ps(_mm_mul_ps(ewcljrsq,ewcljrsq),one_half,_mm_sub_ps(one,ewcljrsq)));
377 /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
378 vvdw6 = _mm_mul_ps(_mm_macc_ps(-c6grid_00,_mm_sub_ps(one,poly),c6_00),rinvsix);
379 vvdw12 = _mm_mul_ps(c12_00,_mm_mul_ps(rinvsix,rinvsix));
380 vvdw = _mm_msub_ps(vvdw12,one_twelfth,_mm_mul_ps(vvdw6,one_sixth));
381 /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
382 fvdw = _mm_mul_ps(_mm_add_ps(vvdw12,_mm_msub_ps(_mm_mul_ps(c6grid_00,one_sixth),_mm_mul_ps(exponent,ewclj6),vvdw6)),rinvsq00);
384 /* Update potential sum for this i atom from the interaction with this j atom. */
385 velec = _mm_andnot_ps(dummy_mask,velec);
386 velecsum = _mm_add_ps(velecsum,velec);
387 vvdw = _mm_andnot_ps(dummy_mask,vvdw);
388 vvdwsum = _mm_add_ps(vvdwsum,vvdw);
390 fscal = _mm_add_ps(felec,fvdw);
392 fscal = _mm_andnot_ps(dummy_mask,fscal);
394 /* Update vectorial force */
395 fix0 = _mm_macc_ps(dx00,fscal,fix0);
396 fiy0 = _mm_macc_ps(dy00,fscal,fiy0);
397 fiz0 = _mm_macc_ps(dz00,fscal,fiz0);
399 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
400 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
401 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
402 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
403 gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
404 _mm_mul_ps(dx00,fscal),
405 _mm_mul_ps(dy00,fscal),
406 _mm_mul_ps(dz00,fscal));
408 /* Inner loop uses 54 flops */
411 /* End of innermost loop */
413 gmx_mm_update_iforce_1atom_swizzle_ps(fix0,fiy0,fiz0,
414 f+i_coord_offset,fshift+i_shift_offset);
417 /* Update potential energies */
418 gmx_mm_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
419 gmx_mm_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
421 /* Increment number of inner iterations */
422 inneriter += j_index_end - j_index_start;
424 /* Outer loop uses 9 flops */
427 /* Increment number of outer iterations */
430 /* Update outer/inner flops */
432 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_VF,outeriter*9 + inneriter*54);
435 * Gromacs nonbonded kernel: nb_kernel_ElecEw_VdwLJEw_GeomP1P1_F_avx_128_fma_single
436 * Electrostatics interaction: Ewald
437 * VdW interaction: LJEwald
438 * Geometry: Particle-Particle
439 * Calculate force/pot: Force
442 nb_kernel_ElecEw_VdwLJEw_GeomP1P1_F_avx_128_fma_single
443 (t_nblist * gmx_restrict nlist,
444 rvec * gmx_restrict xx,
445 rvec * gmx_restrict ff,
446 t_forcerec * gmx_restrict fr,
447 t_mdatoms * gmx_restrict mdatoms,
448 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
449 t_nrnb * gmx_restrict nrnb)
451 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
452 * just 0 for non-waters.
453 * Suffixes A,B,C,D refer to j loop unrolling done with AVX_128, e.g. for the four different
454 * jnr indices corresponding to data put in the four positions in the SIMD register.
456 int i_shift_offset,i_coord_offset,outeriter,inneriter;
457 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
458 int jnrA,jnrB,jnrC,jnrD;
459 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
460 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
461 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
463 real *shiftvec,*fshift,*x,*f;
464 real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
466 __m128 fscal,rcutoff,rcutoff2,jidxall;
468 __m128 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
469 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
470 __m128 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
471 __m128 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
472 __m128 velec,felec,velecsum,facel,crf,krf,krf2;
475 __m128 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
478 __m128 one_sixth = _mm_set1_ps(1.0/6.0);
479 __m128 one_twelfth = _mm_set1_ps(1.0/12.0);
482 __m128 ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
483 __m128 one_half = _mm_set1_ps(0.5);
484 __m128 minus_one = _mm_set1_ps(-1.0);
486 __m128 ewtabscale,eweps,twoeweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
487 __m128 beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
489 __m128 dummy_mask,cutoff_mask;
490 __m128 signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
491 __m128 one = _mm_set1_ps(1.0);
492 __m128 two = _mm_set1_ps(2.0);
498 jindex = nlist->jindex;
500 shiftidx = nlist->shift;
502 shiftvec = fr->shift_vec[0];
503 fshift = fr->fshift[0];
504 facel = _mm_set1_ps(fr->epsfac);
505 charge = mdatoms->chargeA;
506 nvdwtype = fr->ntype;
508 vdwtype = mdatoms->typeA;
509 vdwgridparam = fr->ljpme_c6grid;
510 sh_lj_ewald = _mm_set1_ps(fr->ic->sh_lj_ewald);
511 ewclj = _mm_set1_ps(fr->ewaldcoeff_lj);
512 ewclj2 = _mm_mul_ps(minus_one,_mm_mul_ps(ewclj,ewclj));
514 sh_ewald = _mm_set1_ps(fr->ic->sh_ewald);
515 beta = _mm_set1_ps(fr->ic->ewaldcoeff_q);
516 beta2 = _mm_mul_ps(beta,beta);
517 beta3 = _mm_mul_ps(beta,beta2);
518 ewtab = fr->ic->tabq_coul_F;
519 ewtabscale = _mm_set1_ps(fr->ic->tabq_scale);
520 ewtabhalfspace = _mm_set1_ps(0.5/fr->ic->tabq_scale);
522 /* Avoid stupid compiler warnings */
523 jnrA = jnrB = jnrC = jnrD = 0;
532 for(iidx=0;iidx<4*DIM;iidx++)
537 /* Start outer loop over neighborlists */
538 for(iidx=0; iidx<nri; iidx++)
540 /* Load shift vector for this list */
541 i_shift_offset = DIM*shiftidx[iidx];
543 /* Load limits for loop over neighbors */
544 j_index_start = jindex[iidx];
545 j_index_end = jindex[iidx+1];
547 /* Get outer coordinate index */
549 i_coord_offset = DIM*inr;
551 /* Load i particle coords and add shift vector */
552 gmx_mm_load_shift_and_1rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
554 fix0 = _mm_setzero_ps();
555 fiy0 = _mm_setzero_ps();
556 fiz0 = _mm_setzero_ps();
558 /* Load parameters for i particles */
559 iq0 = _mm_mul_ps(facel,_mm_load1_ps(charge+inr+0));
560 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
562 /* Start inner kernel loop */
563 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
566 /* Get j neighbor index, and coordinate index */
571 j_coord_offsetA = DIM*jnrA;
572 j_coord_offsetB = DIM*jnrB;
573 j_coord_offsetC = DIM*jnrC;
574 j_coord_offsetD = DIM*jnrD;
576 /* load j atom coordinates */
577 gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
578 x+j_coord_offsetC,x+j_coord_offsetD,
581 /* Calculate displacement vector */
582 dx00 = _mm_sub_ps(ix0,jx0);
583 dy00 = _mm_sub_ps(iy0,jy0);
584 dz00 = _mm_sub_ps(iz0,jz0);
586 /* Calculate squared distance and things based on it */
587 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
589 rinv00 = gmx_mm_invsqrt_ps(rsq00);
591 rinvsq00 = _mm_mul_ps(rinv00,rinv00);
593 /* Load parameters for j particles */
594 jq0 = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
595 charge+jnrC+0,charge+jnrD+0);
596 vdwjidx0A = 2*vdwtype[jnrA+0];
597 vdwjidx0B = 2*vdwtype[jnrB+0];
598 vdwjidx0C = 2*vdwtype[jnrC+0];
599 vdwjidx0D = 2*vdwtype[jnrD+0];
601 /**************************
602 * CALCULATE INTERACTIONS *
603 **************************/
605 r00 = _mm_mul_ps(rsq00,rinv00);
607 /* Compute parameters for interactions between i and j atoms */
608 qq00 = _mm_mul_ps(iq0,jq0);
609 gmx_mm_load_4pair_swizzle_ps(vdwparam+vdwioffset0+vdwjidx0A,
610 vdwparam+vdwioffset0+vdwjidx0B,
611 vdwparam+vdwioffset0+vdwjidx0C,
612 vdwparam+vdwioffset0+vdwjidx0D,
615 c6grid_00 = gmx_mm_load_4real_swizzle_ps(vdwgridparam+vdwioffset0+vdwjidx0A,
616 vdwgridparam+vdwioffset0+vdwjidx0B,
617 vdwgridparam+vdwioffset0+vdwjidx0C,
618 vdwgridparam+vdwioffset0+vdwjidx0D);
620 /* EWALD ELECTROSTATICS */
622 /* Analytical PME correction */
623 zeta2 = _mm_mul_ps(beta2,rsq00);
624 rinv3 = _mm_mul_ps(rinvsq00,rinv00);
625 pmecorrF = gmx_mm_pmecorrF_ps(zeta2);
626 felec = _mm_macc_ps(pmecorrF,beta3,rinv3);
627 felec = _mm_mul_ps(qq00,felec);
629 /* Analytical LJ-PME */
630 rinvsix = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00);
631 ewcljrsq = _mm_mul_ps(ewclj2,rsq00);
632 ewclj6 = _mm_mul_ps(ewclj2,_mm_mul_ps(ewclj2,ewclj2));
633 exponent = gmx_simd_exp_r(ewcljrsq);
634 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
635 poly = _mm_mul_ps(exponent,_mm_macc_ps(_mm_mul_ps(ewcljrsq,ewcljrsq),one_half,_mm_sub_ps(one,ewcljrsq)));
636 /* f6A = 6 * C6grid * (1 - poly) */
637 f6A = _mm_mul_ps(c6grid_00,_mm_sub_ps(one,poly));
638 /* f6B = C6grid * exponent * beta^6 */
639 f6B = _mm_mul_ps(_mm_mul_ps(c6grid_00,one_sixth),_mm_mul_ps(exponent,ewclj6));
640 /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
641 fvdw = _mm_mul_ps(_mm_macc_ps(_mm_msub_ps(c12_00,rinvsix,_mm_sub_ps(c6_00,f6A)),rinvsix,f6B),rinvsq00);
643 fscal = _mm_add_ps(felec,fvdw);
645 /* Update vectorial force */
646 fix0 = _mm_macc_ps(dx00,fscal,fix0);
647 fiy0 = _mm_macc_ps(dy00,fscal,fiy0);
648 fiz0 = _mm_macc_ps(dz00,fscal,fiz0);
650 fjptrA = f+j_coord_offsetA;
651 fjptrB = f+j_coord_offsetB;
652 fjptrC = f+j_coord_offsetC;
653 fjptrD = f+j_coord_offsetD;
654 gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
655 _mm_mul_ps(dx00,fscal),
656 _mm_mul_ps(dy00,fscal),
657 _mm_mul_ps(dz00,fscal));
659 /* Inner loop uses 49 flops */
665 /* Get j neighbor index, and coordinate index */
666 jnrlistA = jjnr[jidx];
667 jnrlistB = jjnr[jidx+1];
668 jnrlistC = jjnr[jidx+2];
669 jnrlistD = jjnr[jidx+3];
670 /* Sign of each element will be negative for non-real atoms.
671 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
672 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
674 dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
675 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
676 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
677 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
678 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
679 j_coord_offsetA = DIM*jnrA;
680 j_coord_offsetB = DIM*jnrB;
681 j_coord_offsetC = DIM*jnrC;
682 j_coord_offsetD = DIM*jnrD;
684 /* load j atom coordinates */
685 gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
686 x+j_coord_offsetC,x+j_coord_offsetD,
689 /* Calculate displacement vector */
690 dx00 = _mm_sub_ps(ix0,jx0);
691 dy00 = _mm_sub_ps(iy0,jy0);
692 dz00 = _mm_sub_ps(iz0,jz0);
694 /* Calculate squared distance and things based on it */
695 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
697 rinv00 = gmx_mm_invsqrt_ps(rsq00);
699 rinvsq00 = _mm_mul_ps(rinv00,rinv00);
701 /* Load parameters for j particles */
702 jq0 = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
703 charge+jnrC+0,charge+jnrD+0);
704 vdwjidx0A = 2*vdwtype[jnrA+0];
705 vdwjidx0B = 2*vdwtype[jnrB+0];
706 vdwjidx0C = 2*vdwtype[jnrC+0];
707 vdwjidx0D = 2*vdwtype[jnrD+0];
709 /**************************
710 * CALCULATE INTERACTIONS *
711 **************************/
713 r00 = _mm_mul_ps(rsq00,rinv00);
714 r00 = _mm_andnot_ps(dummy_mask,r00);
716 /* Compute parameters for interactions between i and j atoms */
717 qq00 = _mm_mul_ps(iq0,jq0);
718 gmx_mm_load_4pair_swizzle_ps(vdwparam+vdwioffset0+vdwjidx0A,
719 vdwparam+vdwioffset0+vdwjidx0B,
720 vdwparam+vdwioffset0+vdwjidx0C,
721 vdwparam+vdwioffset0+vdwjidx0D,
724 c6grid_00 = gmx_mm_load_4real_swizzle_ps(vdwgridparam+vdwioffset0+vdwjidx0A,
725 vdwgridparam+vdwioffset0+vdwjidx0B,
726 vdwgridparam+vdwioffset0+vdwjidx0C,
727 vdwgridparam+vdwioffset0+vdwjidx0D);
729 /* EWALD ELECTROSTATICS */
731 /* Analytical PME correction */
732 zeta2 = _mm_mul_ps(beta2,rsq00);
733 rinv3 = _mm_mul_ps(rinvsq00,rinv00);
734 pmecorrF = gmx_mm_pmecorrF_ps(zeta2);
735 felec = _mm_macc_ps(pmecorrF,beta3,rinv3);
736 felec = _mm_mul_ps(qq00,felec);
738 /* Analytical LJ-PME */
739 rinvsix = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00);
740 ewcljrsq = _mm_mul_ps(ewclj2,rsq00);
741 ewclj6 = _mm_mul_ps(ewclj2,_mm_mul_ps(ewclj2,ewclj2));
742 exponent = gmx_simd_exp_r(ewcljrsq);
743 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
744 poly = _mm_mul_ps(exponent,_mm_macc_ps(_mm_mul_ps(ewcljrsq,ewcljrsq),one_half,_mm_sub_ps(one,ewcljrsq)));
745 /* f6A = 6 * C6grid * (1 - poly) */
746 f6A = _mm_mul_ps(c6grid_00,_mm_sub_ps(one,poly));
747 /* f6B = C6grid * exponent * beta^6 */
748 f6B = _mm_mul_ps(_mm_mul_ps(c6grid_00,one_sixth),_mm_mul_ps(exponent,ewclj6));
749 /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
750 fvdw = _mm_mul_ps(_mm_macc_ps(_mm_msub_ps(c12_00,rinvsix,_mm_sub_ps(c6_00,f6A)),rinvsix,f6B),rinvsq00);
752 fscal = _mm_add_ps(felec,fvdw);
754 fscal = _mm_andnot_ps(dummy_mask,fscal);
756 /* Update vectorial force */
757 fix0 = _mm_macc_ps(dx00,fscal,fix0);
758 fiy0 = _mm_macc_ps(dy00,fscal,fiy0);
759 fiz0 = _mm_macc_ps(dz00,fscal,fiz0);
761 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
762 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
763 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
764 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
765 gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
766 _mm_mul_ps(dx00,fscal),
767 _mm_mul_ps(dy00,fscal),
768 _mm_mul_ps(dz00,fscal));
770 /* Inner loop uses 50 flops */
773 /* End of innermost loop */
775 gmx_mm_update_iforce_1atom_swizzle_ps(fix0,fiy0,fiz0,
776 f+i_coord_offset,fshift+i_shift_offset);
778 /* Increment number of inner iterations */
779 inneriter += j_index_end - j_index_start;
781 /* Outer loop uses 7 flops */
784 /* Increment number of outer iterations */
787 /* Update outer/inner flops */
789 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_F,outeriter*7 + inneriter*50);