2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
5 * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
6 * and including many others, as listed in the AUTHORS file in the
7 * top-level source directory and at http://www.gromacs.org.
9 * GROMACS is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU Lesser General Public License
11 * as published by the Free Software Foundation; either version 2.1
12 * of the License, or (at your option) any later version.
14 * GROMACS is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
17 * Lesser General Public License for more details.
19 * You should have received a copy of the GNU Lesser General Public
20 * License along with GROMACS; if not, see
21 * http://www.gnu.org/licenses, or write to the Free Software Foundation,
22 * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
24 * If you want to redistribute modifications to GROMACS, please
25 * consider that scientific software is very special. Version
26 * control is crucial - bugs must be traceable. We will be happy to
27 * consider code for inclusion in the official distribution, but
28 * derived work must not be called official GROMACS. Details are found
29 * in the README & COPYING files - if they are missing, get the
30 * official version at http://www.gromacs.org.
32 * To help us fund GROMACS development, we humbly ask that you cite
33 * the research papers on the package. Check out http://www.gromacs.org.
36 * Note: this file was generated by the GROMACS avx_128_fma_single kernel generator.
42 #include "../nb_kernel.h"
43 #include "types/simple.h"
44 #include "gromacs/math/vec.h"
47 #include "gromacs/simd/math_x86_avx_128_fma_single.h"
48 #include "kernelutil_x86_avx_128_fma_single.h"
51 * Gromacs nonbonded kernel: nb_kernel_ElecRFCut_VdwNone_GeomP1P1_VF_avx_128_fma_single
52 * Electrostatics interaction: ReactionField
53 * VdW interaction: None
54 * Geometry: Particle-Particle
55 * Calculate force/pot: PotentialAndForce
58 nb_kernel_ElecRFCut_VdwNone_GeomP1P1_VF_avx_128_fma_single
59 (t_nblist * gmx_restrict nlist,
60 rvec * gmx_restrict xx,
61 rvec * gmx_restrict ff,
62 t_forcerec * gmx_restrict fr,
63 t_mdatoms * gmx_restrict mdatoms,
64 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
65 t_nrnb * gmx_restrict nrnb)
67 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
68 * just 0 for non-waters.
69 * Suffixes A,B,C,D refer to j loop unrolling done with AVX_128, e.g. for the four different
70 * jnr indices corresponding to data put in the four positions in the SIMD register.
72 int i_shift_offset,i_coord_offset,outeriter,inneriter;
73 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
74 int jnrA,jnrB,jnrC,jnrD;
75 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
76 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
77 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
79 real *shiftvec,*fshift,*x,*f;
80 real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
82 __m128 fscal,rcutoff,rcutoff2,jidxall;
84 __m128 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
85 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
86 __m128 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
87 __m128 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
88 __m128 velec,felec,velecsum,facel,crf,krf,krf2;
90 __m128 dummy_mask,cutoff_mask;
91 __m128 signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
92 __m128 one = _mm_set1_ps(1.0);
93 __m128 two = _mm_set1_ps(2.0);
99 jindex = nlist->jindex;
101 shiftidx = nlist->shift;
103 shiftvec = fr->shift_vec[0];
104 fshift = fr->fshift[0];
105 facel = _mm_set1_ps(fr->epsfac);
106 charge = mdatoms->chargeA;
107 krf = _mm_set1_ps(fr->ic->k_rf);
108 krf2 = _mm_set1_ps(fr->ic->k_rf*2.0);
109 crf = _mm_set1_ps(fr->ic->c_rf);
111 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
112 rcutoff_scalar = fr->rcoulomb;
113 rcutoff = _mm_set1_ps(rcutoff_scalar);
114 rcutoff2 = _mm_mul_ps(rcutoff,rcutoff);
116 /* Avoid stupid compiler warnings */
117 jnrA = jnrB = jnrC = jnrD = 0;
126 for(iidx=0;iidx<4*DIM;iidx++)
131 /* Start outer loop over neighborlists */
132 for(iidx=0; iidx<nri; iidx++)
134 /* Load shift vector for this list */
135 i_shift_offset = DIM*shiftidx[iidx];
137 /* Load limits for loop over neighbors */
138 j_index_start = jindex[iidx];
139 j_index_end = jindex[iidx+1];
141 /* Get outer coordinate index */
143 i_coord_offset = DIM*inr;
145 /* Load i particle coords and add shift vector */
146 gmx_mm_load_shift_and_1rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
148 fix0 = _mm_setzero_ps();
149 fiy0 = _mm_setzero_ps();
150 fiz0 = _mm_setzero_ps();
152 /* Load parameters for i particles */
153 iq0 = _mm_mul_ps(facel,_mm_load1_ps(charge+inr+0));
155 /* Reset potential sums */
156 velecsum = _mm_setzero_ps();
158 /* Start inner kernel loop */
159 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
162 /* Get j neighbor index, and coordinate index */
167 j_coord_offsetA = DIM*jnrA;
168 j_coord_offsetB = DIM*jnrB;
169 j_coord_offsetC = DIM*jnrC;
170 j_coord_offsetD = DIM*jnrD;
172 /* load j atom coordinates */
173 gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
174 x+j_coord_offsetC,x+j_coord_offsetD,
177 /* Calculate displacement vector */
178 dx00 = _mm_sub_ps(ix0,jx0);
179 dy00 = _mm_sub_ps(iy0,jy0);
180 dz00 = _mm_sub_ps(iz0,jz0);
182 /* Calculate squared distance and things based on it */
183 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
185 rinv00 = gmx_mm_invsqrt_ps(rsq00);
187 rinvsq00 = _mm_mul_ps(rinv00,rinv00);
189 /* Load parameters for j particles */
190 jq0 = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
191 charge+jnrC+0,charge+jnrD+0);
193 /**************************
194 * CALCULATE INTERACTIONS *
195 **************************/
197 if (gmx_mm_any_lt(rsq00,rcutoff2))
200 /* Compute parameters for interactions between i and j atoms */
201 qq00 = _mm_mul_ps(iq0,jq0);
203 /* REACTION-FIELD ELECTROSTATICS */
204 velec = _mm_mul_ps(qq00,_mm_sub_ps(_mm_macc_ps(krf,rsq00,rinv00),crf));
205 felec = _mm_mul_ps(qq00,_mm_msub_ps(rinv00,rinvsq00,krf2));
207 cutoff_mask = _mm_cmplt_ps(rsq00,rcutoff2);
209 /* Update potential sum for this i atom from the interaction with this j atom. */
210 velec = _mm_and_ps(velec,cutoff_mask);
211 velecsum = _mm_add_ps(velecsum,velec);
215 fscal = _mm_and_ps(fscal,cutoff_mask);
217 /* Update vectorial force */
218 fix0 = _mm_macc_ps(dx00,fscal,fix0);
219 fiy0 = _mm_macc_ps(dy00,fscal,fiy0);
220 fiz0 = _mm_macc_ps(dz00,fscal,fiz0);
222 fjptrA = f+j_coord_offsetA;
223 fjptrB = f+j_coord_offsetB;
224 fjptrC = f+j_coord_offsetC;
225 fjptrD = f+j_coord_offsetD;
226 gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
227 _mm_mul_ps(dx00,fscal),
228 _mm_mul_ps(dy00,fscal),
229 _mm_mul_ps(dz00,fscal));
233 /* Inner loop uses 39 flops */
239 /* Get j neighbor index, and coordinate index */
240 jnrlistA = jjnr[jidx];
241 jnrlistB = jjnr[jidx+1];
242 jnrlistC = jjnr[jidx+2];
243 jnrlistD = jjnr[jidx+3];
244 /* Sign of each element will be negative for non-real atoms.
245 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
246 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
248 dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
249 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
250 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
251 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
252 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
253 j_coord_offsetA = DIM*jnrA;
254 j_coord_offsetB = DIM*jnrB;
255 j_coord_offsetC = DIM*jnrC;
256 j_coord_offsetD = DIM*jnrD;
258 /* load j atom coordinates */
259 gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
260 x+j_coord_offsetC,x+j_coord_offsetD,
263 /* Calculate displacement vector */
264 dx00 = _mm_sub_ps(ix0,jx0);
265 dy00 = _mm_sub_ps(iy0,jy0);
266 dz00 = _mm_sub_ps(iz0,jz0);
268 /* Calculate squared distance and things based on it */
269 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
271 rinv00 = gmx_mm_invsqrt_ps(rsq00);
273 rinvsq00 = _mm_mul_ps(rinv00,rinv00);
275 /* Load parameters for j particles */
276 jq0 = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
277 charge+jnrC+0,charge+jnrD+0);
279 /**************************
280 * CALCULATE INTERACTIONS *
281 **************************/
283 if (gmx_mm_any_lt(rsq00,rcutoff2))
286 /* Compute parameters for interactions between i and j atoms */
287 qq00 = _mm_mul_ps(iq0,jq0);
289 /* REACTION-FIELD ELECTROSTATICS */
290 velec = _mm_mul_ps(qq00,_mm_sub_ps(_mm_macc_ps(krf,rsq00,rinv00),crf));
291 felec = _mm_mul_ps(qq00,_mm_msub_ps(rinv00,rinvsq00,krf2));
293 cutoff_mask = _mm_cmplt_ps(rsq00,rcutoff2);
295 /* Update potential sum for this i atom from the interaction with this j atom. */
296 velec = _mm_and_ps(velec,cutoff_mask);
297 velec = _mm_andnot_ps(dummy_mask,velec);
298 velecsum = _mm_add_ps(velecsum,velec);
302 fscal = _mm_and_ps(fscal,cutoff_mask);
304 fscal = _mm_andnot_ps(dummy_mask,fscal);
306 /* Update vectorial force */
307 fix0 = _mm_macc_ps(dx00,fscal,fix0);
308 fiy0 = _mm_macc_ps(dy00,fscal,fiy0);
309 fiz0 = _mm_macc_ps(dz00,fscal,fiz0);
311 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
312 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
313 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
314 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
315 gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
316 _mm_mul_ps(dx00,fscal),
317 _mm_mul_ps(dy00,fscal),
318 _mm_mul_ps(dz00,fscal));
322 /* Inner loop uses 39 flops */
325 /* End of innermost loop */
327 gmx_mm_update_iforce_1atom_swizzle_ps(fix0,fiy0,fiz0,
328 f+i_coord_offset,fshift+i_shift_offset);
331 /* Update potential energies */
332 gmx_mm_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
334 /* Increment number of inner iterations */
335 inneriter += j_index_end - j_index_start;
337 /* Outer loop uses 8 flops */
340 /* Increment number of outer iterations */
343 /* Update outer/inner flops */
345 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VF,outeriter*8 + inneriter*39);
348 * Gromacs nonbonded kernel: nb_kernel_ElecRFCut_VdwNone_GeomP1P1_F_avx_128_fma_single
349 * Electrostatics interaction: ReactionField
350 * VdW interaction: None
351 * Geometry: Particle-Particle
352 * Calculate force/pot: Force
355 nb_kernel_ElecRFCut_VdwNone_GeomP1P1_F_avx_128_fma_single
356 (t_nblist * gmx_restrict nlist,
357 rvec * gmx_restrict xx,
358 rvec * gmx_restrict ff,
359 t_forcerec * gmx_restrict fr,
360 t_mdatoms * gmx_restrict mdatoms,
361 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
362 t_nrnb * gmx_restrict nrnb)
364 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
365 * just 0 for non-waters.
366 * Suffixes A,B,C,D refer to j loop unrolling done with AVX_128, e.g. for the four different
367 * jnr indices corresponding to data put in the four positions in the SIMD register.
369 int i_shift_offset,i_coord_offset,outeriter,inneriter;
370 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
371 int jnrA,jnrB,jnrC,jnrD;
372 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
373 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
374 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
376 real *shiftvec,*fshift,*x,*f;
377 real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
379 __m128 fscal,rcutoff,rcutoff2,jidxall;
381 __m128 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
382 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
383 __m128 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
384 __m128 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
385 __m128 velec,felec,velecsum,facel,crf,krf,krf2;
387 __m128 dummy_mask,cutoff_mask;
388 __m128 signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
389 __m128 one = _mm_set1_ps(1.0);
390 __m128 two = _mm_set1_ps(2.0);
396 jindex = nlist->jindex;
398 shiftidx = nlist->shift;
400 shiftvec = fr->shift_vec[0];
401 fshift = fr->fshift[0];
402 facel = _mm_set1_ps(fr->epsfac);
403 charge = mdatoms->chargeA;
404 krf = _mm_set1_ps(fr->ic->k_rf);
405 krf2 = _mm_set1_ps(fr->ic->k_rf*2.0);
406 crf = _mm_set1_ps(fr->ic->c_rf);
408 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
409 rcutoff_scalar = fr->rcoulomb;
410 rcutoff = _mm_set1_ps(rcutoff_scalar);
411 rcutoff2 = _mm_mul_ps(rcutoff,rcutoff);
413 /* Avoid stupid compiler warnings */
414 jnrA = jnrB = jnrC = jnrD = 0;
423 for(iidx=0;iidx<4*DIM;iidx++)
428 /* Start outer loop over neighborlists */
429 for(iidx=0; iidx<nri; iidx++)
431 /* Load shift vector for this list */
432 i_shift_offset = DIM*shiftidx[iidx];
434 /* Load limits for loop over neighbors */
435 j_index_start = jindex[iidx];
436 j_index_end = jindex[iidx+1];
438 /* Get outer coordinate index */
440 i_coord_offset = DIM*inr;
442 /* Load i particle coords and add shift vector */
443 gmx_mm_load_shift_and_1rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
445 fix0 = _mm_setzero_ps();
446 fiy0 = _mm_setzero_ps();
447 fiz0 = _mm_setzero_ps();
449 /* Load parameters for i particles */
450 iq0 = _mm_mul_ps(facel,_mm_load1_ps(charge+inr+0));
452 /* Start inner kernel loop */
453 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
456 /* Get j neighbor index, and coordinate index */
461 j_coord_offsetA = DIM*jnrA;
462 j_coord_offsetB = DIM*jnrB;
463 j_coord_offsetC = DIM*jnrC;
464 j_coord_offsetD = DIM*jnrD;
466 /* load j atom coordinates */
467 gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
468 x+j_coord_offsetC,x+j_coord_offsetD,
471 /* Calculate displacement vector */
472 dx00 = _mm_sub_ps(ix0,jx0);
473 dy00 = _mm_sub_ps(iy0,jy0);
474 dz00 = _mm_sub_ps(iz0,jz0);
476 /* Calculate squared distance and things based on it */
477 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
479 rinv00 = gmx_mm_invsqrt_ps(rsq00);
481 rinvsq00 = _mm_mul_ps(rinv00,rinv00);
483 /* Load parameters for j particles */
484 jq0 = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
485 charge+jnrC+0,charge+jnrD+0);
487 /**************************
488 * CALCULATE INTERACTIONS *
489 **************************/
491 if (gmx_mm_any_lt(rsq00,rcutoff2))
494 /* Compute parameters for interactions between i and j atoms */
495 qq00 = _mm_mul_ps(iq0,jq0);
497 /* REACTION-FIELD ELECTROSTATICS */
498 felec = _mm_mul_ps(qq00,_mm_msub_ps(rinv00,rinvsq00,krf2));
500 cutoff_mask = _mm_cmplt_ps(rsq00,rcutoff2);
504 fscal = _mm_and_ps(fscal,cutoff_mask);
506 /* Update vectorial force */
507 fix0 = _mm_macc_ps(dx00,fscal,fix0);
508 fiy0 = _mm_macc_ps(dy00,fscal,fiy0);
509 fiz0 = _mm_macc_ps(dz00,fscal,fiz0);
511 fjptrA = f+j_coord_offsetA;
512 fjptrB = f+j_coord_offsetB;
513 fjptrC = f+j_coord_offsetC;
514 fjptrD = f+j_coord_offsetD;
515 gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
516 _mm_mul_ps(dx00,fscal),
517 _mm_mul_ps(dy00,fscal),
518 _mm_mul_ps(dz00,fscal));
522 /* Inner loop uses 33 flops */
528 /* Get j neighbor index, and coordinate index */
529 jnrlistA = jjnr[jidx];
530 jnrlistB = jjnr[jidx+1];
531 jnrlistC = jjnr[jidx+2];
532 jnrlistD = jjnr[jidx+3];
533 /* Sign of each element will be negative for non-real atoms.
534 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
535 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
537 dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
538 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
539 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
540 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
541 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
542 j_coord_offsetA = DIM*jnrA;
543 j_coord_offsetB = DIM*jnrB;
544 j_coord_offsetC = DIM*jnrC;
545 j_coord_offsetD = DIM*jnrD;
547 /* load j atom coordinates */
548 gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
549 x+j_coord_offsetC,x+j_coord_offsetD,
552 /* Calculate displacement vector */
553 dx00 = _mm_sub_ps(ix0,jx0);
554 dy00 = _mm_sub_ps(iy0,jy0);
555 dz00 = _mm_sub_ps(iz0,jz0);
557 /* Calculate squared distance and things based on it */
558 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
560 rinv00 = gmx_mm_invsqrt_ps(rsq00);
562 rinvsq00 = _mm_mul_ps(rinv00,rinv00);
564 /* Load parameters for j particles */
565 jq0 = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
566 charge+jnrC+0,charge+jnrD+0);
568 /**************************
569 * CALCULATE INTERACTIONS *
570 **************************/
572 if (gmx_mm_any_lt(rsq00,rcutoff2))
575 /* Compute parameters for interactions between i and j atoms */
576 qq00 = _mm_mul_ps(iq0,jq0);
578 /* REACTION-FIELD ELECTROSTATICS */
579 felec = _mm_mul_ps(qq00,_mm_msub_ps(rinv00,rinvsq00,krf2));
581 cutoff_mask = _mm_cmplt_ps(rsq00,rcutoff2);
585 fscal = _mm_and_ps(fscal,cutoff_mask);
587 fscal = _mm_andnot_ps(dummy_mask,fscal);
589 /* Update vectorial force */
590 fix0 = _mm_macc_ps(dx00,fscal,fix0);
591 fiy0 = _mm_macc_ps(dy00,fscal,fiy0);
592 fiz0 = _mm_macc_ps(dz00,fscal,fiz0);
594 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
595 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
596 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
597 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
598 gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
599 _mm_mul_ps(dx00,fscal),
600 _mm_mul_ps(dy00,fscal),
601 _mm_mul_ps(dz00,fscal));
605 /* Inner loop uses 33 flops */
608 /* End of innermost loop */
610 gmx_mm_update_iforce_1atom_swizzle_ps(fix0,fiy0,fiz0,
611 f+i_coord_offset,fshift+i_shift_offset);
613 /* Increment number of inner iterations */
614 inneriter += j_index_end - j_index_start;
616 /* Outer loop uses 7 flops */
619 /* Increment number of outer iterations */
622 /* Update outer/inner flops */
624 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_F,outeriter*7 + inneriter*33);