2 * Note: this file was generated by the Gromacs sse4_1_single kernel generator.
4 * This source code is part of
8 * Copyright (c) 2001-2012, The GROMACS Development Team
10 * Gromacs is a library for molecular simulation and trajectory analysis,
11 * written by Erik Lindahl, David van der Spoel, Berk Hess, and others - for
12 * a full list of developers and information, check out http://www.gromacs.org
14 * This program is free software; you can redistribute it and/or modify it under
15 * the terms of the GNU Lesser General Public License as published by the Free
16 * Software Foundation; either version 2 of the License, or (at your option) any
19 * To help fund GROMACS development, we humbly ask that you cite
20 * the papers people have written on it - you can find them on the website.
28 #include "../nb_kernel.h"
29 #include "types/simple.h"
33 #include "gmx_math_x86_sse4_1_single.h"
34 #include "kernelutil_x86_sse4_1_single.h"
37 * Gromacs nonbonded kernel: nb_kernel_ElecNone_VdwLJ_GeomP1P1_VF_sse4_1_single
38 * Electrostatics interaction: None
39 * VdW interaction: LennardJones
40 * Geometry: Particle-Particle
41 * Calculate force/pot: PotentialAndForce
44 nb_kernel_ElecNone_VdwLJ_GeomP1P1_VF_sse4_1_single
45 (t_nblist * gmx_restrict nlist,
46 rvec * gmx_restrict xx,
47 rvec * gmx_restrict ff,
48 t_forcerec * gmx_restrict fr,
49 t_mdatoms * gmx_restrict mdatoms,
50 nb_kernel_data_t * gmx_restrict kernel_data,
51 t_nrnb * gmx_restrict nrnb)
53 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
54 * just 0 for non-waters.
55 * Suffixes A,B,C,D refer to j loop unrolling done with SSE, e.g. for the four different
56 * jnr indices corresponding to data put in the four positions in the SIMD register.
58 int i_shift_offset,i_coord_offset,outeriter,inneriter;
59 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
60 int jnrA,jnrB,jnrC,jnrD;
61 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
62 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
63 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
65 real *shiftvec,*fshift,*x,*f;
66 real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
68 __m128 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
70 __m128 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
71 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
72 __m128 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
73 __m128 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
75 __m128 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
78 __m128 one_sixth = _mm_set1_ps(1.0/6.0);
79 __m128 one_twelfth = _mm_set1_ps(1.0/12.0);
80 __m128 dummy_mask,cutoff_mask;
81 __m128 signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
82 __m128 one = _mm_set1_ps(1.0);
83 __m128 two = _mm_set1_ps(2.0);
89 jindex = nlist->jindex;
91 shiftidx = nlist->shift;
93 shiftvec = fr->shift_vec[0];
94 fshift = fr->fshift[0];
97 vdwtype = mdatoms->typeA;
99 /* Avoid stupid compiler warnings */
100 jnrA = jnrB = jnrC = jnrD = 0;
109 for(iidx=0;iidx<4*DIM;iidx++)
114 /* Start outer loop over neighborlists */
115 for(iidx=0; iidx<nri; iidx++)
117 /* Load shift vector for this list */
118 i_shift_offset = DIM*shiftidx[iidx];
120 /* Load limits for loop over neighbors */
121 j_index_start = jindex[iidx];
122 j_index_end = jindex[iidx+1];
124 /* Get outer coordinate index */
126 i_coord_offset = DIM*inr;
128 /* Load i particle coords and add shift vector */
129 gmx_mm_load_shift_and_1rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
131 fix0 = _mm_setzero_ps();
132 fiy0 = _mm_setzero_ps();
133 fiz0 = _mm_setzero_ps();
135 /* Load parameters for i particles */
136 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
138 /* Reset potential sums */
139 vvdwsum = _mm_setzero_ps();
141 /* Start inner kernel loop */
142 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
145 /* Get j neighbor index, and coordinate index */
150 j_coord_offsetA = DIM*jnrA;
151 j_coord_offsetB = DIM*jnrB;
152 j_coord_offsetC = DIM*jnrC;
153 j_coord_offsetD = DIM*jnrD;
155 /* load j atom coordinates */
156 gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
157 x+j_coord_offsetC,x+j_coord_offsetD,
160 /* Calculate displacement vector */
161 dx00 = _mm_sub_ps(ix0,jx0);
162 dy00 = _mm_sub_ps(iy0,jy0);
163 dz00 = _mm_sub_ps(iz0,jz0);
165 /* Calculate squared distance and things based on it */
166 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
168 rinvsq00 = gmx_mm_inv_ps(rsq00);
170 /* Load parameters for j particles */
171 vdwjidx0A = 2*vdwtype[jnrA+0];
172 vdwjidx0B = 2*vdwtype[jnrB+0];
173 vdwjidx0C = 2*vdwtype[jnrC+0];
174 vdwjidx0D = 2*vdwtype[jnrD+0];
176 /**************************
177 * CALCULATE INTERACTIONS *
178 **************************/
180 /* Compute parameters for interactions between i and j atoms */
181 gmx_mm_load_4pair_swizzle_ps(vdwparam+vdwioffset0+vdwjidx0A,
182 vdwparam+vdwioffset0+vdwjidx0B,
183 vdwparam+vdwioffset0+vdwjidx0C,
184 vdwparam+vdwioffset0+vdwjidx0D,
187 /* LENNARD-JONES DISPERSION/REPULSION */
189 rinvsix = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00);
190 vvdw6 = _mm_mul_ps(c6_00,rinvsix);
191 vvdw12 = _mm_mul_ps(c12_00,_mm_mul_ps(rinvsix,rinvsix));
192 vvdw = _mm_sub_ps( _mm_mul_ps(vvdw12,one_twelfth) , _mm_mul_ps(vvdw6,one_sixth) );
193 fvdw = _mm_mul_ps(_mm_sub_ps(vvdw12,vvdw6),rinvsq00);
195 /* Update potential sum for this i atom from the interaction with this j atom. */
196 vvdwsum = _mm_add_ps(vvdwsum,vvdw);
200 /* Calculate temporary vectorial force */
201 tx = _mm_mul_ps(fscal,dx00);
202 ty = _mm_mul_ps(fscal,dy00);
203 tz = _mm_mul_ps(fscal,dz00);
205 /* Update vectorial force */
206 fix0 = _mm_add_ps(fix0,tx);
207 fiy0 = _mm_add_ps(fiy0,ty);
208 fiz0 = _mm_add_ps(fiz0,tz);
210 fjptrA = f+j_coord_offsetA;
211 fjptrB = f+j_coord_offsetB;
212 fjptrC = f+j_coord_offsetC;
213 fjptrD = f+j_coord_offsetD;
214 gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,tx,ty,tz);
216 /* Inner loop uses 32 flops */
222 /* Get j neighbor index, and coordinate index */
223 jnrlistA = jjnr[jidx];
224 jnrlistB = jjnr[jidx+1];
225 jnrlistC = jjnr[jidx+2];
226 jnrlistD = jjnr[jidx+3];
227 /* Sign of each element will be negative for non-real atoms.
228 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
229 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
231 dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
232 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
233 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
234 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
235 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
236 j_coord_offsetA = DIM*jnrA;
237 j_coord_offsetB = DIM*jnrB;
238 j_coord_offsetC = DIM*jnrC;
239 j_coord_offsetD = DIM*jnrD;
241 /* load j atom coordinates */
242 gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
243 x+j_coord_offsetC,x+j_coord_offsetD,
246 /* Calculate displacement vector */
247 dx00 = _mm_sub_ps(ix0,jx0);
248 dy00 = _mm_sub_ps(iy0,jy0);
249 dz00 = _mm_sub_ps(iz0,jz0);
251 /* Calculate squared distance and things based on it */
252 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
254 rinvsq00 = gmx_mm_inv_ps(rsq00);
256 /* Load parameters for j particles */
257 vdwjidx0A = 2*vdwtype[jnrA+0];
258 vdwjidx0B = 2*vdwtype[jnrB+0];
259 vdwjidx0C = 2*vdwtype[jnrC+0];
260 vdwjidx0D = 2*vdwtype[jnrD+0];
262 /**************************
263 * CALCULATE INTERACTIONS *
264 **************************/
266 /* Compute parameters for interactions between i and j atoms */
267 gmx_mm_load_4pair_swizzle_ps(vdwparam+vdwioffset0+vdwjidx0A,
268 vdwparam+vdwioffset0+vdwjidx0B,
269 vdwparam+vdwioffset0+vdwjidx0C,
270 vdwparam+vdwioffset0+vdwjidx0D,
273 /* LENNARD-JONES DISPERSION/REPULSION */
275 rinvsix = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00);
276 vvdw6 = _mm_mul_ps(c6_00,rinvsix);
277 vvdw12 = _mm_mul_ps(c12_00,_mm_mul_ps(rinvsix,rinvsix));
278 vvdw = _mm_sub_ps( _mm_mul_ps(vvdw12,one_twelfth) , _mm_mul_ps(vvdw6,one_sixth) );
279 fvdw = _mm_mul_ps(_mm_sub_ps(vvdw12,vvdw6),rinvsq00);
281 /* Update potential sum for this i atom from the interaction with this j atom. */
282 vvdw = _mm_andnot_ps(dummy_mask,vvdw);
283 vvdwsum = _mm_add_ps(vvdwsum,vvdw);
287 fscal = _mm_andnot_ps(dummy_mask,fscal);
289 /* Calculate temporary vectorial force */
290 tx = _mm_mul_ps(fscal,dx00);
291 ty = _mm_mul_ps(fscal,dy00);
292 tz = _mm_mul_ps(fscal,dz00);
294 /* Update vectorial force */
295 fix0 = _mm_add_ps(fix0,tx);
296 fiy0 = _mm_add_ps(fiy0,ty);
297 fiz0 = _mm_add_ps(fiz0,tz);
299 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
300 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
301 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
302 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
303 gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,tx,ty,tz);
305 /* Inner loop uses 32 flops */
308 /* End of innermost loop */
310 gmx_mm_update_iforce_1atom_swizzle_ps(fix0,fiy0,fiz0,
311 f+i_coord_offset,fshift+i_shift_offset);
314 /* Update potential energies */
315 gmx_mm_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
317 /* Increment number of inner iterations */
318 inneriter += j_index_end - j_index_start;
320 /* Outer loop uses 7 flops */
323 /* Increment number of outer iterations */
326 /* Update outer/inner flops */
328 inc_nrnb(nrnb,eNR_NBKERNEL_VDW_VF,outeriter*7 + inneriter*32);
331 * Gromacs nonbonded kernel: nb_kernel_ElecNone_VdwLJ_GeomP1P1_F_sse4_1_single
332 * Electrostatics interaction: None
333 * VdW interaction: LennardJones
334 * Geometry: Particle-Particle
335 * Calculate force/pot: Force
338 nb_kernel_ElecNone_VdwLJ_GeomP1P1_F_sse4_1_single
339 (t_nblist * gmx_restrict nlist,
340 rvec * gmx_restrict xx,
341 rvec * gmx_restrict ff,
342 t_forcerec * gmx_restrict fr,
343 t_mdatoms * gmx_restrict mdatoms,
344 nb_kernel_data_t * gmx_restrict kernel_data,
345 t_nrnb * gmx_restrict nrnb)
347 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
348 * just 0 for non-waters.
349 * Suffixes A,B,C,D refer to j loop unrolling done with SSE, e.g. for the four different
350 * jnr indices corresponding to data put in the four positions in the SIMD register.
352 int i_shift_offset,i_coord_offset,outeriter,inneriter;
353 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
354 int jnrA,jnrB,jnrC,jnrD;
355 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
356 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
357 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
359 real *shiftvec,*fshift,*x,*f;
360 real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
362 __m128 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
364 __m128 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
365 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
366 __m128 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
367 __m128 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
369 __m128 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
372 __m128 one_sixth = _mm_set1_ps(1.0/6.0);
373 __m128 one_twelfth = _mm_set1_ps(1.0/12.0);
374 __m128 dummy_mask,cutoff_mask;
375 __m128 signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
376 __m128 one = _mm_set1_ps(1.0);
377 __m128 two = _mm_set1_ps(2.0);
383 jindex = nlist->jindex;
385 shiftidx = nlist->shift;
387 shiftvec = fr->shift_vec[0];
388 fshift = fr->fshift[0];
389 nvdwtype = fr->ntype;
391 vdwtype = mdatoms->typeA;
393 /* Avoid stupid compiler warnings */
394 jnrA = jnrB = jnrC = jnrD = 0;
403 for(iidx=0;iidx<4*DIM;iidx++)
408 /* Start outer loop over neighborlists */
409 for(iidx=0; iidx<nri; iidx++)
411 /* Load shift vector for this list */
412 i_shift_offset = DIM*shiftidx[iidx];
414 /* Load limits for loop over neighbors */
415 j_index_start = jindex[iidx];
416 j_index_end = jindex[iidx+1];
418 /* Get outer coordinate index */
420 i_coord_offset = DIM*inr;
422 /* Load i particle coords and add shift vector */
423 gmx_mm_load_shift_and_1rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
425 fix0 = _mm_setzero_ps();
426 fiy0 = _mm_setzero_ps();
427 fiz0 = _mm_setzero_ps();
429 /* Load parameters for i particles */
430 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
432 /* Start inner kernel loop */
433 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
436 /* Get j neighbor index, and coordinate index */
441 j_coord_offsetA = DIM*jnrA;
442 j_coord_offsetB = DIM*jnrB;
443 j_coord_offsetC = DIM*jnrC;
444 j_coord_offsetD = DIM*jnrD;
446 /* load j atom coordinates */
447 gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
448 x+j_coord_offsetC,x+j_coord_offsetD,
451 /* Calculate displacement vector */
452 dx00 = _mm_sub_ps(ix0,jx0);
453 dy00 = _mm_sub_ps(iy0,jy0);
454 dz00 = _mm_sub_ps(iz0,jz0);
456 /* Calculate squared distance and things based on it */
457 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
459 rinvsq00 = gmx_mm_inv_ps(rsq00);
461 /* Load parameters for j particles */
462 vdwjidx0A = 2*vdwtype[jnrA+0];
463 vdwjidx0B = 2*vdwtype[jnrB+0];
464 vdwjidx0C = 2*vdwtype[jnrC+0];
465 vdwjidx0D = 2*vdwtype[jnrD+0];
467 /**************************
468 * CALCULATE INTERACTIONS *
469 **************************/
471 /* Compute parameters for interactions between i and j atoms */
472 gmx_mm_load_4pair_swizzle_ps(vdwparam+vdwioffset0+vdwjidx0A,
473 vdwparam+vdwioffset0+vdwjidx0B,
474 vdwparam+vdwioffset0+vdwjidx0C,
475 vdwparam+vdwioffset0+vdwjidx0D,
478 /* LENNARD-JONES DISPERSION/REPULSION */
480 rinvsix = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00);
481 fvdw = _mm_mul_ps(_mm_sub_ps(_mm_mul_ps(c12_00,rinvsix),c6_00),_mm_mul_ps(rinvsix,rinvsq00));
485 /* Calculate temporary vectorial force */
486 tx = _mm_mul_ps(fscal,dx00);
487 ty = _mm_mul_ps(fscal,dy00);
488 tz = _mm_mul_ps(fscal,dz00);
490 /* Update vectorial force */
491 fix0 = _mm_add_ps(fix0,tx);
492 fiy0 = _mm_add_ps(fiy0,ty);
493 fiz0 = _mm_add_ps(fiz0,tz);
495 fjptrA = f+j_coord_offsetA;
496 fjptrB = f+j_coord_offsetB;
497 fjptrC = f+j_coord_offsetC;
498 fjptrD = f+j_coord_offsetD;
499 gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,tx,ty,tz);
501 /* Inner loop uses 27 flops */
507 /* Get j neighbor index, and coordinate index */
508 jnrlistA = jjnr[jidx];
509 jnrlistB = jjnr[jidx+1];
510 jnrlistC = jjnr[jidx+2];
511 jnrlistD = jjnr[jidx+3];
512 /* Sign of each element will be negative for non-real atoms.
513 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
514 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
516 dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
517 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
518 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
519 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
520 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
521 j_coord_offsetA = DIM*jnrA;
522 j_coord_offsetB = DIM*jnrB;
523 j_coord_offsetC = DIM*jnrC;
524 j_coord_offsetD = DIM*jnrD;
526 /* load j atom coordinates */
527 gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
528 x+j_coord_offsetC,x+j_coord_offsetD,
531 /* Calculate displacement vector */
532 dx00 = _mm_sub_ps(ix0,jx0);
533 dy00 = _mm_sub_ps(iy0,jy0);
534 dz00 = _mm_sub_ps(iz0,jz0);
536 /* Calculate squared distance and things based on it */
537 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
539 rinvsq00 = gmx_mm_inv_ps(rsq00);
541 /* Load parameters for j particles */
542 vdwjidx0A = 2*vdwtype[jnrA+0];
543 vdwjidx0B = 2*vdwtype[jnrB+0];
544 vdwjidx0C = 2*vdwtype[jnrC+0];
545 vdwjidx0D = 2*vdwtype[jnrD+0];
547 /**************************
548 * CALCULATE INTERACTIONS *
549 **************************/
551 /* Compute parameters for interactions between i and j atoms */
552 gmx_mm_load_4pair_swizzle_ps(vdwparam+vdwioffset0+vdwjidx0A,
553 vdwparam+vdwioffset0+vdwjidx0B,
554 vdwparam+vdwioffset0+vdwjidx0C,
555 vdwparam+vdwioffset0+vdwjidx0D,
558 /* LENNARD-JONES DISPERSION/REPULSION */
560 rinvsix = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00);
561 fvdw = _mm_mul_ps(_mm_sub_ps(_mm_mul_ps(c12_00,rinvsix),c6_00),_mm_mul_ps(rinvsix,rinvsq00));
565 fscal = _mm_andnot_ps(dummy_mask,fscal);
567 /* Calculate temporary vectorial force */
568 tx = _mm_mul_ps(fscal,dx00);
569 ty = _mm_mul_ps(fscal,dy00);
570 tz = _mm_mul_ps(fscal,dz00);
572 /* Update vectorial force */
573 fix0 = _mm_add_ps(fix0,tx);
574 fiy0 = _mm_add_ps(fiy0,ty);
575 fiz0 = _mm_add_ps(fiz0,tz);
577 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
578 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
579 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
580 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
581 gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,tx,ty,tz);
583 /* Inner loop uses 27 flops */
586 /* End of innermost loop */
588 gmx_mm_update_iforce_1atom_swizzle_ps(fix0,fiy0,fiz0,
589 f+i_coord_offset,fshift+i_shift_offset);
591 /* Increment number of inner iterations */
592 inneriter += j_index_end - j_index_start;
594 /* Outer loop uses 6 flops */
597 /* Increment number of outer iterations */
600 /* Update outer/inner flops */
602 inc_nrnb(nrnb,eNR_NBKERNEL_VDW_F,outeriter*6 + inneriter*27);