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_ElecCSTab_VdwNone_GeomP1P1_VF_avx_128_fma_single
52 * Electrostatics interaction: CubicSplineTable
53 * VdW interaction: None
54 * Geometry: Particle-Particle
55 * Calculate force/pot: PotentialAndForce
58 nb_kernel_ElecCSTab_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;
91 __m128i ifour = _mm_set1_epi32(4);
92 __m128 rt,vfeps,twovfeps,vftabscale,Y,F,G,H,Fp,VV,FF;
94 __m128 dummy_mask,cutoff_mask;
95 __m128 signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
96 __m128 one = _mm_set1_ps(1.0);
97 __m128 two = _mm_set1_ps(2.0);
103 jindex = nlist->jindex;
105 shiftidx = nlist->shift;
107 shiftvec = fr->shift_vec[0];
108 fshift = fr->fshift[0];
109 facel = _mm_set1_ps(fr->epsfac);
110 charge = mdatoms->chargeA;
112 vftab = kernel_data->table_elec->data;
113 vftabscale = _mm_set1_ps(kernel_data->table_elec->scale);
115 /* Avoid stupid compiler warnings */
116 jnrA = jnrB = jnrC = jnrD = 0;
125 for(iidx=0;iidx<4*DIM;iidx++)
130 /* Start outer loop over neighborlists */
131 for(iidx=0; iidx<nri; iidx++)
133 /* Load shift vector for this list */
134 i_shift_offset = DIM*shiftidx[iidx];
136 /* Load limits for loop over neighbors */
137 j_index_start = jindex[iidx];
138 j_index_end = jindex[iidx+1];
140 /* Get outer coordinate index */
142 i_coord_offset = DIM*inr;
144 /* Load i particle coords and add shift vector */
145 gmx_mm_load_shift_and_1rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
147 fix0 = _mm_setzero_ps();
148 fiy0 = _mm_setzero_ps();
149 fiz0 = _mm_setzero_ps();
151 /* Load parameters for i particles */
152 iq0 = _mm_mul_ps(facel,_mm_load1_ps(charge+inr+0));
154 /* Reset potential sums */
155 velecsum = _mm_setzero_ps();
157 /* Start inner kernel loop */
158 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
161 /* Get j neighbor index, and coordinate index */
166 j_coord_offsetA = DIM*jnrA;
167 j_coord_offsetB = DIM*jnrB;
168 j_coord_offsetC = DIM*jnrC;
169 j_coord_offsetD = DIM*jnrD;
171 /* load j atom coordinates */
172 gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
173 x+j_coord_offsetC,x+j_coord_offsetD,
176 /* Calculate displacement vector */
177 dx00 = _mm_sub_ps(ix0,jx0);
178 dy00 = _mm_sub_ps(iy0,jy0);
179 dz00 = _mm_sub_ps(iz0,jz0);
181 /* Calculate squared distance and things based on it */
182 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
184 rinv00 = gmx_mm_invsqrt_ps(rsq00);
186 /* Load parameters for j particles */
187 jq0 = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
188 charge+jnrC+0,charge+jnrD+0);
190 /**************************
191 * CALCULATE INTERACTIONS *
192 **************************/
194 r00 = _mm_mul_ps(rsq00,rinv00);
196 /* Compute parameters for interactions between i and j atoms */
197 qq00 = _mm_mul_ps(iq0,jq0);
199 /* Calculate table index by multiplying r with table scale and truncate to integer */
200 rt = _mm_mul_ps(r00,vftabscale);
201 vfitab = _mm_cvttps_epi32(rt);
203 vfeps = _mm_frcz_ps(rt);
205 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
207 twovfeps = _mm_add_ps(vfeps,vfeps);
208 vfitab = _mm_slli_epi32(vfitab,2);
210 /* CUBIC SPLINE TABLE ELECTROSTATICS */
211 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
212 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
213 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
214 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
215 _MM_TRANSPOSE4_PS(Y,F,G,H);
216 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
217 VV = _mm_macc_ps(vfeps,Fp,Y);
218 velec = _mm_mul_ps(qq00,VV);
219 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
220 felec = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq00,FF),_mm_mul_ps(vftabscale,rinv00)));
222 /* Update potential sum for this i atom from the interaction with this j atom. */
223 velecsum = _mm_add_ps(velecsum,velec);
227 /* Update vectorial force */
228 fix0 = _mm_macc_ps(dx00,fscal,fix0);
229 fiy0 = _mm_macc_ps(dy00,fscal,fiy0);
230 fiz0 = _mm_macc_ps(dz00,fscal,fiz0);
232 fjptrA = f+j_coord_offsetA;
233 fjptrB = f+j_coord_offsetB;
234 fjptrC = f+j_coord_offsetC;
235 fjptrD = f+j_coord_offsetD;
236 gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
237 _mm_mul_ps(dx00,fscal),
238 _mm_mul_ps(dy00,fscal),
239 _mm_mul_ps(dz00,fscal));
241 /* Inner loop uses 46 flops */
247 /* Get j neighbor index, and coordinate index */
248 jnrlistA = jjnr[jidx];
249 jnrlistB = jjnr[jidx+1];
250 jnrlistC = jjnr[jidx+2];
251 jnrlistD = jjnr[jidx+3];
252 /* Sign of each element will be negative for non-real atoms.
253 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
254 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
256 dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
257 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
258 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
259 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
260 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
261 j_coord_offsetA = DIM*jnrA;
262 j_coord_offsetB = DIM*jnrB;
263 j_coord_offsetC = DIM*jnrC;
264 j_coord_offsetD = DIM*jnrD;
266 /* load j atom coordinates */
267 gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
268 x+j_coord_offsetC,x+j_coord_offsetD,
271 /* Calculate displacement vector */
272 dx00 = _mm_sub_ps(ix0,jx0);
273 dy00 = _mm_sub_ps(iy0,jy0);
274 dz00 = _mm_sub_ps(iz0,jz0);
276 /* Calculate squared distance and things based on it */
277 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
279 rinv00 = gmx_mm_invsqrt_ps(rsq00);
281 /* Load parameters for j particles */
282 jq0 = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
283 charge+jnrC+0,charge+jnrD+0);
285 /**************************
286 * CALCULATE INTERACTIONS *
287 **************************/
289 r00 = _mm_mul_ps(rsq00,rinv00);
290 r00 = _mm_andnot_ps(dummy_mask,r00);
292 /* Compute parameters for interactions between i and j atoms */
293 qq00 = _mm_mul_ps(iq0,jq0);
295 /* Calculate table index by multiplying r with table scale and truncate to integer */
296 rt = _mm_mul_ps(r00,vftabscale);
297 vfitab = _mm_cvttps_epi32(rt);
299 vfeps = _mm_frcz_ps(rt);
301 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
303 twovfeps = _mm_add_ps(vfeps,vfeps);
304 vfitab = _mm_slli_epi32(vfitab,2);
306 /* CUBIC SPLINE TABLE ELECTROSTATICS */
307 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
308 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
309 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
310 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
311 _MM_TRANSPOSE4_PS(Y,F,G,H);
312 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
313 VV = _mm_macc_ps(vfeps,Fp,Y);
314 velec = _mm_mul_ps(qq00,VV);
315 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
316 felec = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq00,FF),_mm_mul_ps(vftabscale,rinv00)));
318 /* Update potential sum for this i atom from the interaction with this j atom. */
319 velec = _mm_andnot_ps(dummy_mask,velec);
320 velecsum = _mm_add_ps(velecsum,velec);
324 fscal = _mm_andnot_ps(dummy_mask,fscal);
326 /* Update vectorial force */
327 fix0 = _mm_macc_ps(dx00,fscal,fix0);
328 fiy0 = _mm_macc_ps(dy00,fscal,fiy0);
329 fiz0 = _mm_macc_ps(dz00,fscal,fiz0);
331 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
332 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
333 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
334 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
335 gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
336 _mm_mul_ps(dx00,fscal),
337 _mm_mul_ps(dy00,fscal),
338 _mm_mul_ps(dz00,fscal));
340 /* Inner loop uses 47 flops */
343 /* End of innermost loop */
345 gmx_mm_update_iforce_1atom_swizzle_ps(fix0,fiy0,fiz0,
346 f+i_coord_offset,fshift+i_shift_offset);
349 /* Update potential energies */
350 gmx_mm_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
352 /* Increment number of inner iterations */
353 inneriter += j_index_end - j_index_start;
355 /* Outer loop uses 8 flops */
358 /* Increment number of outer iterations */
361 /* Update outer/inner flops */
363 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VF,outeriter*8 + inneriter*47);
366 * Gromacs nonbonded kernel: nb_kernel_ElecCSTab_VdwNone_GeomP1P1_F_avx_128_fma_single
367 * Electrostatics interaction: CubicSplineTable
368 * VdW interaction: None
369 * Geometry: Particle-Particle
370 * Calculate force/pot: Force
373 nb_kernel_ElecCSTab_VdwNone_GeomP1P1_F_avx_128_fma_single
374 (t_nblist * gmx_restrict nlist,
375 rvec * gmx_restrict xx,
376 rvec * gmx_restrict ff,
377 t_forcerec * gmx_restrict fr,
378 t_mdatoms * gmx_restrict mdatoms,
379 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
380 t_nrnb * gmx_restrict nrnb)
382 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
383 * just 0 for non-waters.
384 * Suffixes A,B,C,D refer to j loop unrolling done with AVX_128, e.g. for the four different
385 * jnr indices corresponding to data put in the four positions in the SIMD register.
387 int i_shift_offset,i_coord_offset,outeriter,inneriter;
388 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
389 int jnrA,jnrB,jnrC,jnrD;
390 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
391 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
392 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
394 real *shiftvec,*fshift,*x,*f;
395 real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
397 __m128 fscal,rcutoff,rcutoff2,jidxall;
399 __m128 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
400 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
401 __m128 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
402 __m128 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
403 __m128 velec,felec,velecsum,facel,crf,krf,krf2;
406 __m128i ifour = _mm_set1_epi32(4);
407 __m128 rt,vfeps,twovfeps,vftabscale,Y,F,G,H,Fp,VV,FF;
409 __m128 dummy_mask,cutoff_mask;
410 __m128 signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
411 __m128 one = _mm_set1_ps(1.0);
412 __m128 two = _mm_set1_ps(2.0);
418 jindex = nlist->jindex;
420 shiftidx = nlist->shift;
422 shiftvec = fr->shift_vec[0];
423 fshift = fr->fshift[0];
424 facel = _mm_set1_ps(fr->epsfac);
425 charge = mdatoms->chargeA;
427 vftab = kernel_data->table_elec->data;
428 vftabscale = _mm_set1_ps(kernel_data->table_elec->scale);
430 /* Avoid stupid compiler warnings */
431 jnrA = jnrB = jnrC = jnrD = 0;
440 for(iidx=0;iidx<4*DIM;iidx++)
445 /* Start outer loop over neighborlists */
446 for(iidx=0; iidx<nri; iidx++)
448 /* Load shift vector for this list */
449 i_shift_offset = DIM*shiftidx[iidx];
451 /* Load limits for loop over neighbors */
452 j_index_start = jindex[iidx];
453 j_index_end = jindex[iidx+1];
455 /* Get outer coordinate index */
457 i_coord_offset = DIM*inr;
459 /* Load i particle coords and add shift vector */
460 gmx_mm_load_shift_and_1rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
462 fix0 = _mm_setzero_ps();
463 fiy0 = _mm_setzero_ps();
464 fiz0 = _mm_setzero_ps();
466 /* Load parameters for i particles */
467 iq0 = _mm_mul_ps(facel,_mm_load1_ps(charge+inr+0));
469 /* Start inner kernel loop */
470 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
473 /* Get j neighbor index, and coordinate index */
478 j_coord_offsetA = DIM*jnrA;
479 j_coord_offsetB = DIM*jnrB;
480 j_coord_offsetC = DIM*jnrC;
481 j_coord_offsetD = DIM*jnrD;
483 /* load j atom coordinates */
484 gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
485 x+j_coord_offsetC,x+j_coord_offsetD,
488 /* Calculate displacement vector */
489 dx00 = _mm_sub_ps(ix0,jx0);
490 dy00 = _mm_sub_ps(iy0,jy0);
491 dz00 = _mm_sub_ps(iz0,jz0);
493 /* Calculate squared distance and things based on it */
494 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
496 rinv00 = gmx_mm_invsqrt_ps(rsq00);
498 /* Load parameters for j particles */
499 jq0 = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
500 charge+jnrC+0,charge+jnrD+0);
502 /**************************
503 * CALCULATE INTERACTIONS *
504 **************************/
506 r00 = _mm_mul_ps(rsq00,rinv00);
508 /* Compute parameters for interactions between i and j atoms */
509 qq00 = _mm_mul_ps(iq0,jq0);
511 /* Calculate table index by multiplying r with table scale and truncate to integer */
512 rt = _mm_mul_ps(r00,vftabscale);
513 vfitab = _mm_cvttps_epi32(rt);
515 vfeps = _mm_frcz_ps(rt);
517 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
519 twovfeps = _mm_add_ps(vfeps,vfeps);
520 vfitab = _mm_slli_epi32(vfitab,2);
522 /* CUBIC SPLINE TABLE ELECTROSTATICS */
523 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
524 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
525 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
526 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
527 _MM_TRANSPOSE4_PS(Y,F,G,H);
528 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
529 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
530 felec = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq00,FF),_mm_mul_ps(vftabscale,rinv00)));
534 /* Update vectorial force */
535 fix0 = _mm_macc_ps(dx00,fscal,fix0);
536 fiy0 = _mm_macc_ps(dy00,fscal,fiy0);
537 fiz0 = _mm_macc_ps(dz00,fscal,fiz0);
539 fjptrA = f+j_coord_offsetA;
540 fjptrB = f+j_coord_offsetB;
541 fjptrC = f+j_coord_offsetC;
542 fjptrD = f+j_coord_offsetD;
543 gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
544 _mm_mul_ps(dx00,fscal),
545 _mm_mul_ps(dy00,fscal),
546 _mm_mul_ps(dz00,fscal));
548 /* Inner loop uses 42 flops */
554 /* Get j neighbor index, and coordinate index */
555 jnrlistA = jjnr[jidx];
556 jnrlistB = jjnr[jidx+1];
557 jnrlistC = jjnr[jidx+2];
558 jnrlistD = jjnr[jidx+3];
559 /* Sign of each element will be negative for non-real atoms.
560 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
561 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
563 dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
564 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
565 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
566 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
567 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
568 j_coord_offsetA = DIM*jnrA;
569 j_coord_offsetB = DIM*jnrB;
570 j_coord_offsetC = DIM*jnrC;
571 j_coord_offsetD = DIM*jnrD;
573 /* load j atom coordinates */
574 gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
575 x+j_coord_offsetC,x+j_coord_offsetD,
578 /* Calculate displacement vector */
579 dx00 = _mm_sub_ps(ix0,jx0);
580 dy00 = _mm_sub_ps(iy0,jy0);
581 dz00 = _mm_sub_ps(iz0,jz0);
583 /* Calculate squared distance and things based on it */
584 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
586 rinv00 = gmx_mm_invsqrt_ps(rsq00);
588 /* Load parameters for j particles */
589 jq0 = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
590 charge+jnrC+0,charge+jnrD+0);
592 /**************************
593 * CALCULATE INTERACTIONS *
594 **************************/
596 r00 = _mm_mul_ps(rsq00,rinv00);
597 r00 = _mm_andnot_ps(dummy_mask,r00);
599 /* Compute parameters for interactions between i and j atoms */
600 qq00 = _mm_mul_ps(iq0,jq0);
602 /* Calculate table index by multiplying r with table scale and truncate to integer */
603 rt = _mm_mul_ps(r00,vftabscale);
604 vfitab = _mm_cvttps_epi32(rt);
606 vfeps = _mm_frcz_ps(rt);
608 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
610 twovfeps = _mm_add_ps(vfeps,vfeps);
611 vfitab = _mm_slli_epi32(vfitab,2);
613 /* CUBIC SPLINE TABLE ELECTROSTATICS */
614 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
615 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
616 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
617 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
618 _MM_TRANSPOSE4_PS(Y,F,G,H);
619 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
620 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
621 felec = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq00,FF),_mm_mul_ps(vftabscale,rinv00)));
625 fscal = _mm_andnot_ps(dummy_mask,fscal);
627 /* Update vectorial force */
628 fix0 = _mm_macc_ps(dx00,fscal,fix0);
629 fiy0 = _mm_macc_ps(dy00,fscal,fiy0);
630 fiz0 = _mm_macc_ps(dz00,fscal,fiz0);
632 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
633 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
634 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
635 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
636 gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
637 _mm_mul_ps(dx00,fscal),
638 _mm_mul_ps(dy00,fscal),
639 _mm_mul_ps(dz00,fscal));
641 /* Inner loop uses 43 flops */
644 /* End of innermost loop */
646 gmx_mm_update_iforce_1atom_swizzle_ps(fix0,fiy0,fiz0,
647 f+i_coord_offset,fshift+i_shift_offset);
649 /* Increment number of inner iterations */
650 inneriter += j_index_end - j_index_start;
652 /* Outer loop uses 7 flops */
655 /* Increment number of outer iterations */
658 /* Update outer/inner flops */
660 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_F,outeriter*7 + inneriter*43);