2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2012,2013, 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_double kernel generator.
44 #include "../nb_kernel.h"
45 #include "types/simple.h"
49 #include "gromacs/simd/math_x86_avx_128_fma_double.h"
50 #include "kernelutil_x86_avx_128_fma_double.h"
53 * Gromacs nonbonded kernel: nb_kernel_ElecCSTab_VdwNone_GeomW3W3_VF_avx_128_fma_double
54 * Electrostatics interaction: CubicSplineTable
55 * VdW interaction: None
56 * Geometry: Water3-Water3
57 * Calculate force/pot: PotentialAndForce
60 nb_kernel_ElecCSTab_VdwNone_GeomW3W3_VF_avx_128_fma_double
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 refer to j loop unrolling done with SSE double precision, e.g. for the two 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;
77 int j_coord_offsetA,j_coord_offsetB;
78 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
80 real *shiftvec,*fshift,*x,*f;
81 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
83 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
85 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
87 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
88 int vdwjidx0A,vdwjidx0B;
89 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
90 int vdwjidx1A,vdwjidx1B;
91 __m128d jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
92 int vdwjidx2A,vdwjidx2B;
93 __m128d jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
94 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
95 __m128d dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
96 __m128d dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
97 __m128d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
98 __m128d dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
99 __m128d dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
100 __m128d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
101 __m128d dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
102 __m128d dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
103 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
106 __m128i ifour = _mm_set1_epi32(4);
107 __m128d rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
109 __m128d dummy_mask,cutoff_mask;
110 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
111 __m128d one = _mm_set1_pd(1.0);
112 __m128d two = _mm_set1_pd(2.0);
118 jindex = nlist->jindex;
120 shiftidx = nlist->shift;
122 shiftvec = fr->shift_vec[0];
123 fshift = fr->fshift[0];
124 facel = _mm_set1_pd(fr->epsfac);
125 charge = mdatoms->chargeA;
127 vftab = kernel_data->table_elec->data;
128 vftabscale = _mm_set1_pd(kernel_data->table_elec->scale);
130 /* Setup water-specific parameters */
131 inr = nlist->iinr[0];
132 iq0 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
133 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
134 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
136 jq0 = _mm_set1_pd(charge[inr+0]);
137 jq1 = _mm_set1_pd(charge[inr+1]);
138 jq2 = _mm_set1_pd(charge[inr+2]);
139 qq00 = _mm_mul_pd(iq0,jq0);
140 qq01 = _mm_mul_pd(iq0,jq1);
141 qq02 = _mm_mul_pd(iq0,jq2);
142 qq10 = _mm_mul_pd(iq1,jq0);
143 qq11 = _mm_mul_pd(iq1,jq1);
144 qq12 = _mm_mul_pd(iq1,jq2);
145 qq20 = _mm_mul_pd(iq2,jq0);
146 qq21 = _mm_mul_pd(iq2,jq1);
147 qq22 = _mm_mul_pd(iq2,jq2);
149 /* Avoid stupid compiler warnings */
157 /* Start outer loop over neighborlists */
158 for(iidx=0; iidx<nri; iidx++)
160 /* Load shift vector for this list */
161 i_shift_offset = DIM*shiftidx[iidx];
163 /* Load limits for loop over neighbors */
164 j_index_start = jindex[iidx];
165 j_index_end = jindex[iidx+1];
167 /* Get outer coordinate index */
169 i_coord_offset = DIM*inr;
171 /* Load i particle coords and add shift vector */
172 gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
173 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
175 fix0 = _mm_setzero_pd();
176 fiy0 = _mm_setzero_pd();
177 fiz0 = _mm_setzero_pd();
178 fix1 = _mm_setzero_pd();
179 fiy1 = _mm_setzero_pd();
180 fiz1 = _mm_setzero_pd();
181 fix2 = _mm_setzero_pd();
182 fiy2 = _mm_setzero_pd();
183 fiz2 = _mm_setzero_pd();
185 /* Reset potential sums */
186 velecsum = _mm_setzero_pd();
188 /* Start inner kernel loop */
189 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
192 /* Get j neighbor index, and coordinate index */
195 j_coord_offsetA = DIM*jnrA;
196 j_coord_offsetB = DIM*jnrB;
198 /* load j atom coordinates */
199 gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
200 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
202 /* Calculate displacement vector */
203 dx00 = _mm_sub_pd(ix0,jx0);
204 dy00 = _mm_sub_pd(iy0,jy0);
205 dz00 = _mm_sub_pd(iz0,jz0);
206 dx01 = _mm_sub_pd(ix0,jx1);
207 dy01 = _mm_sub_pd(iy0,jy1);
208 dz01 = _mm_sub_pd(iz0,jz1);
209 dx02 = _mm_sub_pd(ix0,jx2);
210 dy02 = _mm_sub_pd(iy0,jy2);
211 dz02 = _mm_sub_pd(iz0,jz2);
212 dx10 = _mm_sub_pd(ix1,jx0);
213 dy10 = _mm_sub_pd(iy1,jy0);
214 dz10 = _mm_sub_pd(iz1,jz0);
215 dx11 = _mm_sub_pd(ix1,jx1);
216 dy11 = _mm_sub_pd(iy1,jy1);
217 dz11 = _mm_sub_pd(iz1,jz1);
218 dx12 = _mm_sub_pd(ix1,jx2);
219 dy12 = _mm_sub_pd(iy1,jy2);
220 dz12 = _mm_sub_pd(iz1,jz2);
221 dx20 = _mm_sub_pd(ix2,jx0);
222 dy20 = _mm_sub_pd(iy2,jy0);
223 dz20 = _mm_sub_pd(iz2,jz0);
224 dx21 = _mm_sub_pd(ix2,jx1);
225 dy21 = _mm_sub_pd(iy2,jy1);
226 dz21 = _mm_sub_pd(iz2,jz1);
227 dx22 = _mm_sub_pd(ix2,jx2);
228 dy22 = _mm_sub_pd(iy2,jy2);
229 dz22 = _mm_sub_pd(iz2,jz2);
231 /* Calculate squared distance and things based on it */
232 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
233 rsq01 = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
234 rsq02 = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
235 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
236 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
237 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
238 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
239 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
240 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
242 rinv00 = gmx_mm_invsqrt_pd(rsq00);
243 rinv01 = gmx_mm_invsqrt_pd(rsq01);
244 rinv02 = gmx_mm_invsqrt_pd(rsq02);
245 rinv10 = gmx_mm_invsqrt_pd(rsq10);
246 rinv11 = gmx_mm_invsqrt_pd(rsq11);
247 rinv12 = gmx_mm_invsqrt_pd(rsq12);
248 rinv20 = gmx_mm_invsqrt_pd(rsq20);
249 rinv21 = gmx_mm_invsqrt_pd(rsq21);
250 rinv22 = gmx_mm_invsqrt_pd(rsq22);
252 fjx0 = _mm_setzero_pd();
253 fjy0 = _mm_setzero_pd();
254 fjz0 = _mm_setzero_pd();
255 fjx1 = _mm_setzero_pd();
256 fjy1 = _mm_setzero_pd();
257 fjz1 = _mm_setzero_pd();
258 fjx2 = _mm_setzero_pd();
259 fjy2 = _mm_setzero_pd();
260 fjz2 = _mm_setzero_pd();
262 /**************************
263 * CALCULATE INTERACTIONS *
264 **************************/
266 r00 = _mm_mul_pd(rsq00,rinv00);
268 /* Calculate table index by multiplying r with table scale and truncate to integer */
269 rt = _mm_mul_pd(r00,vftabscale);
270 vfitab = _mm_cvttpd_epi32(rt);
272 vfeps = _mm_frcz_pd(rt);
274 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
276 twovfeps = _mm_add_pd(vfeps,vfeps);
277 vfitab = _mm_slli_epi32(vfitab,2);
279 /* CUBIC SPLINE TABLE ELECTROSTATICS */
280 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
281 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
282 GMX_MM_TRANSPOSE2_PD(Y,F);
283 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
284 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
285 GMX_MM_TRANSPOSE2_PD(G,H);
286 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
287 VV = _mm_macc_pd(vfeps,Fp,Y);
288 velec = _mm_mul_pd(qq00,VV);
289 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
290 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq00,FF),_mm_mul_pd(vftabscale,rinv00)));
292 /* Update potential sum for this i atom from the interaction with this j atom. */
293 velecsum = _mm_add_pd(velecsum,velec);
297 /* Update vectorial force */
298 fix0 = _mm_macc_pd(dx00,fscal,fix0);
299 fiy0 = _mm_macc_pd(dy00,fscal,fiy0);
300 fiz0 = _mm_macc_pd(dz00,fscal,fiz0);
302 fjx0 = _mm_macc_pd(dx00,fscal,fjx0);
303 fjy0 = _mm_macc_pd(dy00,fscal,fjy0);
304 fjz0 = _mm_macc_pd(dz00,fscal,fjz0);
306 /**************************
307 * CALCULATE INTERACTIONS *
308 **************************/
310 r01 = _mm_mul_pd(rsq01,rinv01);
312 /* Calculate table index by multiplying r with table scale and truncate to integer */
313 rt = _mm_mul_pd(r01,vftabscale);
314 vfitab = _mm_cvttpd_epi32(rt);
316 vfeps = _mm_frcz_pd(rt);
318 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
320 twovfeps = _mm_add_pd(vfeps,vfeps);
321 vfitab = _mm_slli_epi32(vfitab,2);
323 /* CUBIC SPLINE TABLE ELECTROSTATICS */
324 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
325 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
326 GMX_MM_TRANSPOSE2_PD(Y,F);
327 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
328 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
329 GMX_MM_TRANSPOSE2_PD(G,H);
330 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
331 VV = _mm_macc_pd(vfeps,Fp,Y);
332 velec = _mm_mul_pd(qq01,VV);
333 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
334 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq01,FF),_mm_mul_pd(vftabscale,rinv01)));
336 /* Update potential sum for this i atom from the interaction with this j atom. */
337 velecsum = _mm_add_pd(velecsum,velec);
341 /* Update vectorial force */
342 fix0 = _mm_macc_pd(dx01,fscal,fix0);
343 fiy0 = _mm_macc_pd(dy01,fscal,fiy0);
344 fiz0 = _mm_macc_pd(dz01,fscal,fiz0);
346 fjx1 = _mm_macc_pd(dx01,fscal,fjx1);
347 fjy1 = _mm_macc_pd(dy01,fscal,fjy1);
348 fjz1 = _mm_macc_pd(dz01,fscal,fjz1);
350 /**************************
351 * CALCULATE INTERACTIONS *
352 **************************/
354 r02 = _mm_mul_pd(rsq02,rinv02);
356 /* Calculate table index by multiplying r with table scale and truncate to integer */
357 rt = _mm_mul_pd(r02,vftabscale);
358 vfitab = _mm_cvttpd_epi32(rt);
360 vfeps = _mm_frcz_pd(rt);
362 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
364 twovfeps = _mm_add_pd(vfeps,vfeps);
365 vfitab = _mm_slli_epi32(vfitab,2);
367 /* CUBIC SPLINE TABLE ELECTROSTATICS */
368 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
369 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
370 GMX_MM_TRANSPOSE2_PD(Y,F);
371 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
372 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
373 GMX_MM_TRANSPOSE2_PD(G,H);
374 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
375 VV = _mm_macc_pd(vfeps,Fp,Y);
376 velec = _mm_mul_pd(qq02,VV);
377 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
378 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq02,FF),_mm_mul_pd(vftabscale,rinv02)));
380 /* Update potential sum for this i atom from the interaction with this j atom. */
381 velecsum = _mm_add_pd(velecsum,velec);
385 /* Update vectorial force */
386 fix0 = _mm_macc_pd(dx02,fscal,fix0);
387 fiy0 = _mm_macc_pd(dy02,fscal,fiy0);
388 fiz0 = _mm_macc_pd(dz02,fscal,fiz0);
390 fjx2 = _mm_macc_pd(dx02,fscal,fjx2);
391 fjy2 = _mm_macc_pd(dy02,fscal,fjy2);
392 fjz2 = _mm_macc_pd(dz02,fscal,fjz2);
394 /**************************
395 * CALCULATE INTERACTIONS *
396 **************************/
398 r10 = _mm_mul_pd(rsq10,rinv10);
400 /* Calculate table index by multiplying r with table scale and truncate to integer */
401 rt = _mm_mul_pd(r10,vftabscale);
402 vfitab = _mm_cvttpd_epi32(rt);
404 vfeps = _mm_frcz_pd(rt);
406 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
408 twovfeps = _mm_add_pd(vfeps,vfeps);
409 vfitab = _mm_slli_epi32(vfitab,2);
411 /* CUBIC SPLINE TABLE ELECTROSTATICS */
412 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
413 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
414 GMX_MM_TRANSPOSE2_PD(Y,F);
415 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
416 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
417 GMX_MM_TRANSPOSE2_PD(G,H);
418 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
419 VV = _mm_macc_pd(vfeps,Fp,Y);
420 velec = _mm_mul_pd(qq10,VV);
421 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
422 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq10,FF),_mm_mul_pd(vftabscale,rinv10)));
424 /* Update potential sum for this i atom from the interaction with this j atom. */
425 velecsum = _mm_add_pd(velecsum,velec);
429 /* Update vectorial force */
430 fix1 = _mm_macc_pd(dx10,fscal,fix1);
431 fiy1 = _mm_macc_pd(dy10,fscal,fiy1);
432 fiz1 = _mm_macc_pd(dz10,fscal,fiz1);
434 fjx0 = _mm_macc_pd(dx10,fscal,fjx0);
435 fjy0 = _mm_macc_pd(dy10,fscal,fjy0);
436 fjz0 = _mm_macc_pd(dz10,fscal,fjz0);
438 /**************************
439 * CALCULATE INTERACTIONS *
440 **************************/
442 r11 = _mm_mul_pd(rsq11,rinv11);
444 /* Calculate table index by multiplying r with table scale and truncate to integer */
445 rt = _mm_mul_pd(r11,vftabscale);
446 vfitab = _mm_cvttpd_epi32(rt);
448 vfeps = _mm_frcz_pd(rt);
450 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
452 twovfeps = _mm_add_pd(vfeps,vfeps);
453 vfitab = _mm_slli_epi32(vfitab,2);
455 /* CUBIC SPLINE TABLE ELECTROSTATICS */
456 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
457 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
458 GMX_MM_TRANSPOSE2_PD(Y,F);
459 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
460 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
461 GMX_MM_TRANSPOSE2_PD(G,H);
462 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
463 VV = _mm_macc_pd(vfeps,Fp,Y);
464 velec = _mm_mul_pd(qq11,VV);
465 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
466 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
468 /* Update potential sum for this i atom from the interaction with this j atom. */
469 velecsum = _mm_add_pd(velecsum,velec);
473 /* Update vectorial force */
474 fix1 = _mm_macc_pd(dx11,fscal,fix1);
475 fiy1 = _mm_macc_pd(dy11,fscal,fiy1);
476 fiz1 = _mm_macc_pd(dz11,fscal,fiz1);
478 fjx1 = _mm_macc_pd(dx11,fscal,fjx1);
479 fjy1 = _mm_macc_pd(dy11,fscal,fjy1);
480 fjz1 = _mm_macc_pd(dz11,fscal,fjz1);
482 /**************************
483 * CALCULATE INTERACTIONS *
484 **************************/
486 r12 = _mm_mul_pd(rsq12,rinv12);
488 /* Calculate table index by multiplying r with table scale and truncate to integer */
489 rt = _mm_mul_pd(r12,vftabscale);
490 vfitab = _mm_cvttpd_epi32(rt);
492 vfeps = _mm_frcz_pd(rt);
494 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
496 twovfeps = _mm_add_pd(vfeps,vfeps);
497 vfitab = _mm_slli_epi32(vfitab,2);
499 /* CUBIC SPLINE TABLE ELECTROSTATICS */
500 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
501 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
502 GMX_MM_TRANSPOSE2_PD(Y,F);
503 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
504 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
505 GMX_MM_TRANSPOSE2_PD(G,H);
506 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
507 VV = _mm_macc_pd(vfeps,Fp,Y);
508 velec = _mm_mul_pd(qq12,VV);
509 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
510 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
512 /* Update potential sum for this i atom from the interaction with this j atom. */
513 velecsum = _mm_add_pd(velecsum,velec);
517 /* Update vectorial force */
518 fix1 = _mm_macc_pd(dx12,fscal,fix1);
519 fiy1 = _mm_macc_pd(dy12,fscal,fiy1);
520 fiz1 = _mm_macc_pd(dz12,fscal,fiz1);
522 fjx2 = _mm_macc_pd(dx12,fscal,fjx2);
523 fjy2 = _mm_macc_pd(dy12,fscal,fjy2);
524 fjz2 = _mm_macc_pd(dz12,fscal,fjz2);
526 /**************************
527 * CALCULATE INTERACTIONS *
528 **************************/
530 r20 = _mm_mul_pd(rsq20,rinv20);
532 /* Calculate table index by multiplying r with table scale and truncate to integer */
533 rt = _mm_mul_pd(r20,vftabscale);
534 vfitab = _mm_cvttpd_epi32(rt);
536 vfeps = _mm_frcz_pd(rt);
538 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
540 twovfeps = _mm_add_pd(vfeps,vfeps);
541 vfitab = _mm_slli_epi32(vfitab,2);
543 /* CUBIC SPLINE TABLE ELECTROSTATICS */
544 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
545 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
546 GMX_MM_TRANSPOSE2_PD(Y,F);
547 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
548 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
549 GMX_MM_TRANSPOSE2_PD(G,H);
550 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
551 VV = _mm_macc_pd(vfeps,Fp,Y);
552 velec = _mm_mul_pd(qq20,VV);
553 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
554 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq20,FF),_mm_mul_pd(vftabscale,rinv20)));
556 /* Update potential sum for this i atom from the interaction with this j atom. */
557 velecsum = _mm_add_pd(velecsum,velec);
561 /* Update vectorial force */
562 fix2 = _mm_macc_pd(dx20,fscal,fix2);
563 fiy2 = _mm_macc_pd(dy20,fscal,fiy2);
564 fiz2 = _mm_macc_pd(dz20,fscal,fiz2);
566 fjx0 = _mm_macc_pd(dx20,fscal,fjx0);
567 fjy0 = _mm_macc_pd(dy20,fscal,fjy0);
568 fjz0 = _mm_macc_pd(dz20,fscal,fjz0);
570 /**************************
571 * CALCULATE INTERACTIONS *
572 **************************/
574 r21 = _mm_mul_pd(rsq21,rinv21);
576 /* Calculate table index by multiplying r with table scale and truncate to integer */
577 rt = _mm_mul_pd(r21,vftabscale);
578 vfitab = _mm_cvttpd_epi32(rt);
580 vfeps = _mm_frcz_pd(rt);
582 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
584 twovfeps = _mm_add_pd(vfeps,vfeps);
585 vfitab = _mm_slli_epi32(vfitab,2);
587 /* CUBIC SPLINE TABLE ELECTROSTATICS */
588 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
589 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
590 GMX_MM_TRANSPOSE2_PD(Y,F);
591 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
592 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
593 GMX_MM_TRANSPOSE2_PD(G,H);
594 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
595 VV = _mm_macc_pd(vfeps,Fp,Y);
596 velec = _mm_mul_pd(qq21,VV);
597 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
598 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
600 /* Update potential sum for this i atom from the interaction with this j atom. */
601 velecsum = _mm_add_pd(velecsum,velec);
605 /* Update vectorial force */
606 fix2 = _mm_macc_pd(dx21,fscal,fix2);
607 fiy2 = _mm_macc_pd(dy21,fscal,fiy2);
608 fiz2 = _mm_macc_pd(dz21,fscal,fiz2);
610 fjx1 = _mm_macc_pd(dx21,fscal,fjx1);
611 fjy1 = _mm_macc_pd(dy21,fscal,fjy1);
612 fjz1 = _mm_macc_pd(dz21,fscal,fjz1);
614 /**************************
615 * CALCULATE INTERACTIONS *
616 **************************/
618 r22 = _mm_mul_pd(rsq22,rinv22);
620 /* Calculate table index by multiplying r with table scale and truncate to integer */
621 rt = _mm_mul_pd(r22,vftabscale);
622 vfitab = _mm_cvttpd_epi32(rt);
624 vfeps = _mm_frcz_pd(rt);
626 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
628 twovfeps = _mm_add_pd(vfeps,vfeps);
629 vfitab = _mm_slli_epi32(vfitab,2);
631 /* CUBIC SPLINE TABLE ELECTROSTATICS */
632 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
633 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
634 GMX_MM_TRANSPOSE2_PD(Y,F);
635 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
636 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
637 GMX_MM_TRANSPOSE2_PD(G,H);
638 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
639 VV = _mm_macc_pd(vfeps,Fp,Y);
640 velec = _mm_mul_pd(qq22,VV);
641 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
642 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
644 /* Update potential sum for this i atom from the interaction with this j atom. */
645 velecsum = _mm_add_pd(velecsum,velec);
649 /* Update vectorial force */
650 fix2 = _mm_macc_pd(dx22,fscal,fix2);
651 fiy2 = _mm_macc_pd(dy22,fscal,fiy2);
652 fiz2 = _mm_macc_pd(dz22,fscal,fiz2);
654 fjx2 = _mm_macc_pd(dx22,fscal,fjx2);
655 fjy2 = _mm_macc_pd(dy22,fscal,fjy2);
656 fjz2 = _mm_macc_pd(dz22,fscal,fjz2);
658 gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
660 /* Inner loop uses 414 flops */
667 j_coord_offsetA = DIM*jnrA;
669 /* load j atom coordinates */
670 gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
671 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
673 /* Calculate displacement vector */
674 dx00 = _mm_sub_pd(ix0,jx0);
675 dy00 = _mm_sub_pd(iy0,jy0);
676 dz00 = _mm_sub_pd(iz0,jz0);
677 dx01 = _mm_sub_pd(ix0,jx1);
678 dy01 = _mm_sub_pd(iy0,jy1);
679 dz01 = _mm_sub_pd(iz0,jz1);
680 dx02 = _mm_sub_pd(ix0,jx2);
681 dy02 = _mm_sub_pd(iy0,jy2);
682 dz02 = _mm_sub_pd(iz0,jz2);
683 dx10 = _mm_sub_pd(ix1,jx0);
684 dy10 = _mm_sub_pd(iy1,jy0);
685 dz10 = _mm_sub_pd(iz1,jz0);
686 dx11 = _mm_sub_pd(ix1,jx1);
687 dy11 = _mm_sub_pd(iy1,jy1);
688 dz11 = _mm_sub_pd(iz1,jz1);
689 dx12 = _mm_sub_pd(ix1,jx2);
690 dy12 = _mm_sub_pd(iy1,jy2);
691 dz12 = _mm_sub_pd(iz1,jz2);
692 dx20 = _mm_sub_pd(ix2,jx0);
693 dy20 = _mm_sub_pd(iy2,jy0);
694 dz20 = _mm_sub_pd(iz2,jz0);
695 dx21 = _mm_sub_pd(ix2,jx1);
696 dy21 = _mm_sub_pd(iy2,jy1);
697 dz21 = _mm_sub_pd(iz2,jz1);
698 dx22 = _mm_sub_pd(ix2,jx2);
699 dy22 = _mm_sub_pd(iy2,jy2);
700 dz22 = _mm_sub_pd(iz2,jz2);
702 /* Calculate squared distance and things based on it */
703 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
704 rsq01 = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
705 rsq02 = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
706 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
707 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
708 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
709 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
710 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
711 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
713 rinv00 = gmx_mm_invsqrt_pd(rsq00);
714 rinv01 = gmx_mm_invsqrt_pd(rsq01);
715 rinv02 = gmx_mm_invsqrt_pd(rsq02);
716 rinv10 = gmx_mm_invsqrt_pd(rsq10);
717 rinv11 = gmx_mm_invsqrt_pd(rsq11);
718 rinv12 = gmx_mm_invsqrt_pd(rsq12);
719 rinv20 = gmx_mm_invsqrt_pd(rsq20);
720 rinv21 = gmx_mm_invsqrt_pd(rsq21);
721 rinv22 = gmx_mm_invsqrt_pd(rsq22);
723 fjx0 = _mm_setzero_pd();
724 fjy0 = _mm_setzero_pd();
725 fjz0 = _mm_setzero_pd();
726 fjx1 = _mm_setzero_pd();
727 fjy1 = _mm_setzero_pd();
728 fjz1 = _mm_setzero_pd();
729 fjx2 = _mm_setzero_pd();
730 fjy2 = _mm_setzero_pd();
731 fjz2 = _mm_setzero_pd();
733 /**************************
734 * CALCULATE INTERACTIONS *
735 **************************/
737 r00 = _mm_mul_pd(rsq00,rinv00);
739 /* Calculate table index by multiplying r with table scale and truncate to integer */
740 rt = _mm_mul_pd(r00,vftabscale);
741 vfitab = _mm_cvttpd_epi32(rt);
743 vfeps = _mm_frcz_pd(rt);
745 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
747 twovfeps = _mm_add_pd(vfeps,vfeps);
748 vfitab = _mm_slli_epi32(vfitab,2);
750 /* CUBIC SPLINE TABLE ELECTROSTATICS */
751 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
752 F = _mm_setzero_pd();
753 GMX_MM_TRANSPOSE2_PD(Y,F);
754 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
755 H = _mm_setzero_pd();
756 GMX_MM_TRANSPOSE2_PD(G,H);
757 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
758 VV = _mm_macc_pd(vfeps,Fp,Y);
759 velec = _mm_mul_pd(qq00,VV);
760 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
761 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq00,FF),_mm_mul_pd(vftabscale,rinv00)));
763 /* Update potential sum for this i atom from the interaction with this j atom. */
764 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
765 velecsum = _mm_add_pd(velecsum,velec);
769 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
771 /* Update vectorial force */
772 fix0 = _mm_macc_pd(dx00,fscal,fix0);
773 fiy0 = _mm_macc_pd(dy00,fscal,fiy0);
774 fiz0 = _mm_macc_pd(dz00,fscal,fiz0);
776 fjx0 = _mm_macc_pd(dx00,fscal,fjx0);
777 fjy0 = _mm_macc_pd(dy00,fscal,fjy0);
778 fjz0 = _mm_macc_pd(dz00,fscal,fjz0);
780 /**************************
781 * CALCULATE INTERACTIONS *
782 **************************/
784 r01 = _mm_mul_pd(rsq01,rinv01);
786 /* Calculate table index by multiplying r with table scale and truncate to integer */
787 rt = _mm_mul_pd(r01,vftabscale);
788 vfitab = _mm_cvttpd_epi32(rt);
790 vfeps = _mm_frcz_pd(rt);
792 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
794 twovfeps = _mm_add_pd(vfeps,vfeps);
795 vfitab = _mm_slli_epi32(vfitab,2);
797 /* CUBIC SPLINE TABLE ELECTROSTATICS */
798 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
799 F = _mm_setzero_pd();
800 GMX_MM_TRANSPOSE2_PD(Y,F);
801 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
802 H = _mm_setzero_pd();
803 GMX_MM_TRANSPOSE2_PD(G,H);
804 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
805 VV = _mm_macc_pd(vfeps,Fp,Y);
806 velec = _mm_mul_pd(qq01,VV);
807 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
808 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq01,FF),_mm_mul_pd(vftabscale,rinv01)));
810 /* Update potential sum for this i atom from the interaction with this j atom. */
811 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
812 velecsum = _mm_add_pd(velecsum,velec);
816 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
818 /* Update vectorial force */
819 fix0 = _mm_macc_pd(dx01,fscal,fix0);
820 fiy0 = _mm_macc_pd(dy01,fscal,fiy0);
821 fiz0 = _mm_macc_pd(dz01,fscal,fiz0);
823 fjx1 = _mm_macc_pd(dx01,fscal,fjx1);
824 fjy1 = _mm_macc_pd(dy01,fscal,fjy1);
825 fjz1 = _mm_macc_pd(dz01,fscal,fjz1);
827 /**************************
828 * CALCULATE INTERACTIONS *
829 **************************/
831 r02 = _mm_mul_pd(rsq02,rinv02);
833 /* Calculate table index by multiplying r with table scale and truncate to integer */
834 rt = _mm_mul_pd(r02,vftabscale);
835 vfitab = _mm_cvttpd_epi32(rt);
837 vfeps = _mm_frcz_pd(rt);
839 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
841 twovfeps = _mm_add_pd(vfeps,vfeps);
842 vfitab = _mm_slli_epi32(vfitab,2);
844 /* CUBIC SPLINE TABLE ELECTROSTATICS */
845 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
846 F = _mm_setzero_pd();
847 GMX_MM_TRANSPOSE2_PD(Y,F);
848 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
849 H = _mm_setzero_pd();
850 GMX_MM_TRANSPOSE2_PD(G,H);
851 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
852 VV = _mm_macc_pd(vfeps,Fp,Y);
853 velec = _mm_mul_pd(qq02,VV);
854 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
855 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq02,FF),_mm_mul_pd(vftabscale,rinv02)));
857 /* Update potential sum for this i atom from the interaction with this j atom. */
858 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
859 velecsum = _mm_add_pd(velecsum,velec);
863 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
865 /* Update vectorial force */
866 fix0 = _mm_macc_pd(dx02,fscal,fix0);
867 fiy0 = _mm_macc_pd(dy02,fscal,fiy0);
868 fiz0 = _mm_macc_pd(dz02,fscal,fiz0);
870 fjx2 = _mm_macc_pd(dx02,fscal,fjx2);
871 fjy2 = _mm_macc_pd(dy02,fscal,fjy2);
872 fjz2 = _mm_macc_pd(dz02,fscal,fjz2);
874 /**************************
875 * CALCULATE INTERACTIONS *
876 **************************/
878 r10 = _mm_mul_pd(rsq10,rinv10);
880 /* Calculate table index by multiplying r with table scale and truncate to integer */
881 rt = _mm_mul_pd(r10,vftabscale);
882 vfitab = _mm_cvttpd_epi32(rt);
884 vfeps = _mm_frcz_pd(rt);
886 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
888 twovfeps = _mm_add_pd(vfeps,vfeps);
889 vfitab = _mm_slli_epi32(vfitab,2);
891 /* CUBIC SPLINE TABLE ELECTROSTATICS */
892 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
893 F = _mm_setzero_pd();
894 GMX_MM_TRANSPOSE2_PD(Y,F);
895 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
896 H = _mm_setzero_pd();
897 GMX_MM_TRANSPOSE2_PD(G,H);
898 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
899 VV = _mm_macc_pd(vfeps,Fp,Y);
900 velec = _mm_mul_pd(qq10,VV);
901 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
902 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq10,FF),_mm_mul_pd(vftabscale,rinv10)));
904 /* Update potential sum for this i atom from the interaction with this j atom. */
905 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
906 velecsum = _mm_add_pd(velecsum,velec);
910 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
912 /* Update vectorial force */
913 fix1 = _mm_macc_pd(dx10,fscal,fix1);
914 fiy1 = _mm_macc_pd(dy10,fscal,fiy1);
915 fiz1 = _mm_macc_pd(dz10,fscal,fiz1);
917 fjx0 = _mm_macc_pd(dx10,fscal,fjx0);
918 fjy0 = _mm_macc_pd(dy10,fscal,fjy0);
919 fjz0 = _mm_macc_pd(dz10,fscal,fjz0);
921 /**************************
922 * CALCULATE INTERACTIONS *
923 **************************/
925 r11 = _mm_mul_pd(rsq11,rinv11);
927 /* Calculate table index by multiplying r with table scale and truncate to integer */
928 rt = _mm_mul_pd(r11,vftabscale);
929 vfitab = _mm_cvttpd_epi32(rt);
931 vfeps = _mm_frcz_pd(rt);
933 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
935 twovfeps = _mm_add_pd(vfeps,vfeps);
936 vfitab = _mm_slli_epi32(vfitab,2);
938 /* CUBIC SPLINE TABLE ELECTROSTATICS */
939 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
940 F = _mm_setzero_pd();
941 GMX_MM_TRANSPOSE2_PD(Y,F);
942 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
943 H = _mm_setzero_pd();
944 GMX_MM_TRANSPOSE2_PD(G,H);
945 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
946 VV = _mm_macc_pd(vfeps,Fp,Y);
947 velec = _mm_mul_pd(qq11,VV);
948 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
949 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
951 /* Update potential sum for this i atom from the interaction with this j atom. */
952 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
953 velecsum = _mm_add_pd(velecsum,velec);
957 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
959 /* Update vectorial force */
960 fix1 = _mm_macc_pd(dx11,fscal,fix1);
961 fiy1 = _mm_macc_pd(dy11,fscal,fiy1);
962 fiz1 = _mm_macc_pd(dz11,fscal,fiz1);
964 fjx1 = _mm_macc_pd(dx11,fscal,fjx1);
965 fjy1 = _mm_macc_pd(dy11,fscal,fjy1);
966 fjz1 = _mm_macc_pd(dz11,fscal,fjz1);
968 /**************************
969 * CALCULATE INTERACTIONS *
970 **************************/
972 r12 = _mm_mul_pd(rsq12,rinv12);
974 /* Calculate table index by multiplying r with table scale and truncate to integer */
975 rt = _mm_mul_pd(r12,vftabscale);
976 vfitab = _mm_cvttpd_epi32(rt);
978 vfeps = _mm_frcz_pd(rt);
980 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
982 twovfeps = _mm_add_pd(vfeps,vfeps);
983 vfitab = _mm_slli_epi32(vfitab,2);
985 /* CUBIC SPLINE TABLE ELECTROSTATICS */
986 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
987 F = _mm_setzero_pd();
988 GMX_MM_TRANSPOSE2_PD(Y,F);
989 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
990 H = _mm_setzero_pd();
991 GMX_MM_TRANSPOSE2_PD(G,H);
992 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
993 VV = _mm_macc_pd(vfeps,Fp,Y);
994 velec = _mm_mul_pd(qq12,VV);
995 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
996 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
998 /* Update potential sum for this i atom from the interaction with this j atom. */
999 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1000 velecsum = _mm_add_pd(velecsum,velec);
1004 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1006 /* Update vectorial force */
1007 fix1 = _mm_macc_pd(dx12,fscal,fix1);
1008 fiy1 = _mm_macc_pd(dy12,fscal,fiy1);
1009 fiz1 = _mm_macc_pd(dz12,fscal,fiz1);
1011 fjx2 = _mm_macc_pd(dx12,fscal,fjx2);
1012 fjy2 = _mm_macc_pd(dy12,fscal,fjy2);
1013 fjz2 = _mm_macc_pd(dz12,fscal,fjz2);
1015 /**************************
1016 * CALCULATE INTERACTIONS *
1017 **************************/
1019 r20 = _mm_mul_pd(rsq20,rinv20);
1021 /* Calculate table index by multiplying r with table scale and truncate to integer */
1022 rt = _mm_mul_pd(r20,vftabscale);
1023 vfitab = _mm_cvttpd_epi32(rt);
1025 vfeps = _mm_frcz_pd(rt);
1027 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1029 twovfeps = _mm_add_pd(vfeps,vfeps);
1030 vfitab = _mm_slli_epi32(vfitab,2);
1032 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1033 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1034 F = _mm_setzero_pd();
1035 GMX_MM_TRANSPOSE2_PD(Y,F);
1036 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1037 H = _mm_setzero_pd();
1038 GMX_MM_TRANSPOSE2_PD(G,H);
1039 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1040 VV = _mm_macc_pd(vfeps,Fp,Y);
1041 velec = _mm_mul_pd(qq20,VV);
1042 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1043 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq20,FF),_mm_mul_pd(vftabscale,rinv20)));
1045 /* Update potential sum for this i atom from the interaction with this j atom. */
1046 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1047 velecsum = _mm_add_pd(velecsum,velec);
1051 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1053 /* Update vectorial force */
1054 fix2 = _mm_macc_pd(dx20,fscal,fix2);
1055 fiy2 = _mm_macc_pd(dy20,fscal,fiy2);
1056 fiz2 = _mm_macc_pd(dz20,fscal,fiz2);
1058 fjx0 = _mm_macc_pd(dx20,fscal,fjx0);
1059 fjy0 = _mm_macc_pd(dy20,fscal,fjy0);
1060 fjz0 = _mm_macc_pd(dz20,fscal,fjz0);
1062 /**************************
1063 * CALCULATE INTERACTIONS *
1064 **************************/
1066 r21 = _mm_mul_pd(rsq21,rinv21);
1068 /* Calculate table index by multiplying r with table scale and truncate to integer */
1069 rt = _mm_mul_pd(r21,vftabscale);
1070 vfitab = _mm_cvttpd_epi32(rt);
1072 vfeps = _mm_frcz_pd(rt);
1074 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1076 twovfeps = _mm_add_pd(vfeps,vfeps);
1077 vfitab = _mm_slli_epi32(vfitab,2);
1079 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1080 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1081 F = _mm_setzero_pd();
1082 GMX_MM_TRANSPOSE2_PD(Y,F);
1083 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1084 H = _mm_setzero_pd();
1085 GMX_MM_TRANSPOSE2_PD(G,H);
1086 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1087 VV = _mm_macc_pd(vfeps,Fp,Y);
1088 velec = _mm_mul_pd(qq21,VV);
1089 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1090 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
1092 /* Update potential sum for this i atom from the interaction with this j atom. */
1093 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1094 velecsum = _mm_add_pd(velecsum,velec);
1098 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1100 /* Update vectorial force */
1101 fix2 = _mm_macc_pd(dx21,fscal,fix2);
1102 fiy2 = _mm_macc_pd(dy21,fscal,fiy2);
1103 fiz2 = _mm_macc_pd(dz21,fscal,fiz2);
1105 fjx1 = _mm_macc_pd(dx21,fscal,fjx1);
1106 fjy1 = _mm_macc_pd(dy21,fscal,fjy1);
1107 fjz1 = _mm_macc_pd(dz21,fscal,fjz1);
1109 /**************************
1110 * CALCULATE INTERACTIONS *
1111 **************************/
1113 r22 = _mm_mul_pd(rsq22,rinv22);
1115 /* Calculate table index by multiplying r with table scale and truncate to integer */
1116 rt = _mm_mul_pd(r22,vftabscale);
1117 vfitab = _mm_cvttpd_epi32(rt);
1119 vfeps = _mm_frcz_pd(rt);
1121 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1123 twovfeps = _mm_add_pd(vfeps,vfeps);
1124 vfitab = _mm_slli_epi32(vfitab,2);
1126 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1127 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1128 F = _mm_setzero_pd();
1129 GMX_MM_TRANSPOSE2_PD(Y,F);
1130 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1131 H = _mm_setzero_pd();
1132 GMX_MM_TRANSPOSE2_PD(G,H);
1133 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1134 VV = _mm_macc_pd(vfeps,Fp,Y);
1135 velec = _mm_mul_pd(qq22,VV);
1136 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1137 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
1139 /* Update potential sum for this i atom from the interaction with this j atom. */
1140 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1141 velecsum = _mm_add_pd(velecsum,velec);
1145 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1147 /* Update vectorial force */
1148 fix2 = _mm_macc_pd(dx22,fscal,fix2);
1149 fiy2 = _mm_macc_pd(dy22,fscal,fiy2);
1150 fiz2 = _mm_macc_pd(dz22,fscal,fiz2);
1152 fjx2 = _mm_macc_pd(dx22,fscal,fjx2);
1153 fjy2 = _mm_macc_pd(dy22,fscal,fjy2);
1154 fjz2 = _mm_macc_pd(dz22,fscal,fjz2);
1156 gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1158 /* Inner loop uses 414 flops */
1161 /* End of innermost loop */
1163 gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1164 f+i_coord_offset,fshift+i_shift_offset);
1167 /* Update potential energies */
1168 gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1170 /* Increment number of inner iterations */
1171 inneriter += j_index_end - j_index_start;
1173 /* Outer loop uses 19 flops */
1176 /* Increment number of outer iterations */
1179 /* Update outer/inner flops */
1181 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3W3_VF,outeriter*19 + inneriter*414);
1184 * Gromacs nonbonded kernel: nb_kernel_ElecCSTab_VdwNone_GeomW3W3_F_avx_128_fma_double
1185 * Electrostatics interaction: CubicSplineTable
1186 * VdW interaction: None
1187 * Geometry: Water3-Water3
1188 * Calculate force/pot: Force
1191 nb_kernel_ElecCSTab_VdwNone_GeomW3W3_F_avx_128_fma_double
1192 (t_nblist * gmx_restrict nlist,
1193 rvec * gmx_restrict xx,
1194 rvec * gmx_restrict ff,
1195 t_forcerec * gmx_restrict fr,
1196 t_mdatoms * gmx_restrict mdatoms,
1197 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1198 t_nrnb * gmx_restrict nrnb)
1200 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1201 * just 0 for non-waters.
1202 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
1203 * jnr indices corresponding to data put in the four positions in the SIMD register.
1205 int i_shift_offset,i_coord_offset,outeriter,inneriter;
1206 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1208 int j_coord_offsetA,j_coord_offsetB;
1209 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
1210 real rcutoff_scalar;
1211 real *shiftvec,*fshift,*x,*f;
1212 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1214 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1216 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1218 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1219 int vdwjidx0A,vdwjidx0B;
1220 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1221 int vdwjidx1A,vdwjidx1B;
1222 __m128d jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1223 int vdwjidx2A,vdwjidx2B;
1224 __m128d jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1225 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1226 __m128d dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1227 __m128d dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1228 __m128d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1229 __m128d dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1230 __m128d dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1231 __m128d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1232 __m128d dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1233 __m128d dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1234 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
1237 __m128i ifour = _mm_set1_epi32(4);
1238 __m128d rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
1240 __m128d dummy_mask,cutoff_mask;
1241 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
1242 __m128d one = _mm_set1_pd(1.0);
1243 __m128d two = _mm_set1_pd(2.0);
1249 jindex = nlist->jindex;
1251 shiftidx = nlist->shift;
1253 shiftvec = fr->shift_vec[0];
1254 fshift = fr->fshift[0];
1255 facel = _mm_set1_pd(fr->epsfac);
1256 charge = mdatoms->chargeA;
1258 vftab = kernel_data->table_elec->data;
1259 vftabscale = _mm_set1_pd(kernel_data->table_elec->scale);
1261 /* Setup water-specific parameters */
1262 inr = nlist->iinr[0];
1263 iq0 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
1264 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
1265 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
1267 jq0 = _mm_set1_pd(charge[inr+0]);
1268 jq1 = _mm_set1_pd(charge[inr+1]);
1269 jq2 = _mm_set1_pd(charge[inr+2]);
1270 qq00 = _mm_mul_pd(iq0,jq0);
1271 qq01 = _mm_mul_pd(iq0,jq1);
1272 qq02 = _mm_mul_pd(iq0,jq2);
1273 qq10 = _mm_mul_pd(iq1,jq0);
1274 qq11 = _mm_mul_pd(iq1,jq1);
1275 qq12 = _mm_mul_pd(iq1,jq2);
1276 qq20 = _mm_mul_pd(iq2,jq0);
1277 qq21 = _mm_mul_pd(iq2,jq1);
1278 qq22 = _mm_mul_pd(iq2,jq2);
1280 /* Avoid stupid compiler warnings */
1282 j_coord_offsetA = 0;
1283 j_coord_offsetB = 0;
1288 /* Start outer loop over neighborlists */
1289 for(iidx=0; iidx<nri; iidx++)
1291 /* Load shift vector for this list */
1292 i_shift_offset = DIM*shiftidx[iidx];
1294 /* Load limits for loop over neighbors */
1295 j_index_start = jindex[iidx];
1296 j_index_end = jindex[iidx+1];
1298 /* Get outer coordinate index */
1300 i_coord_offset = DIM*inr;
1302 /* Load i particle coords and add shift vector */
1303 gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1304 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1306 fix0 = _mm_setzero_pd();
1307 fiy0 = _mm_setzero_pd();
1308 fiz0 = _mm_setzero_pd();
1309 fix1 = _mm_setzero_pd();
1310 fiy1 = _mm_setzero_pd();
1311 fiz1 = _mm_setzero_pd();
1312 fix2 = _mm_setzero_pd();
1313 fiy2 = _mm_setzero_pd();
1314 fiz2 = _mm_setzero_pd();
1316 /* Start inner kernel loop */
1317 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
1320 /* Get j neighbor index, and coordinate index */
1322 jnrB = jjnr[jidx+1];
1323 j_coord_offsetA = DIM*jnrA;
1324 j_coord_offsetB = DIM*jnrB;
1326 /* load j atom coordinates */
1327 gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1328 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1330 /* Calculate displacement vector */
1331 dx00 = _mm_sub_pd(ix0,jx0);
1332 dy00 = _mm_sub_pd(iy0,jy0);
1333 dz00 = _mm_sub_pd(iz0,jz0);
1334 dx01 = _mm_sub_pd(ix0,jx1);
1335 dy01 = _mm_sub_pd(iy0,jy1);
1336 dz01 = _mm_sub_pd(iz0,jz1);
1337 dx02 = _mm_sub_pd(ix0,jx2);
1338 dy02 = _mm_sub_pd(iy0,jy2);
1339 dz02 = _mm_sub_pd(iz0,jz2);
1340 dx10 = _mm_sub_pd(ix1,jx0);
1341 dy10 = _mm_sub_pd(iy1,jy0);
1342 dz10 = _mm_sub_pd(iz1,jz0);
1343 dx11 = _mm_sub_pd(ix1,jx1);
1344 dy11 = _mm_sub_pd(iy1,jy1);
1345 dz11 = _mm_sub_pd(iz1,jz1);
1346 dx12 = _mm_sub_pd(ix1,jx2);
1347 dy12 = _mm_sub_pd(iy1,jy2);
1348 dz12 = _mm_sub_pd(iz1,jz2);
1349 dx20 = _mm_sub_pd(ix2,jx0);
1350 dy20 = _mm_sub_pd(iy2,jy0);
1351 dz20 = _mm_sub_pd(iz2,jz0);
1352 dx21 = _mm_sub_pd(ix2,jx1);
1353 dy21 = _mm_sub_pd(iy2,jy1);
1354 dz21 = _mm_sub_pd(iz2,jz1);
1355 dx22 = _mm_sub_pd(ix2,jx2);
1356 dy22 = _mm_sub_pd(iy2,jy2);
1357 dz22 = _mm_sub_pd(iz2,jz2);
1359 /* Calculate squared distance and things based on it */
1360 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1361 rsq01 = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
1362 rsq02 = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
1363 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
1364 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1365 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1366 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
1367 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1368 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1370 rinv00 = gmx_mm_invsqrt_pd(rsq00);
1371 rinv01 = gmx_mm_invsqrt_pd(rsq01);
1372 rinv02 = gmx_mm_invsqrt_pd(rsq02);
1373 rinv10 = gmx_mm_invsqrt_pd(rsq10);
1374 rinv11 = gmx_mm_invsqrt_pd(rsq11);
1375 rinv12 = gmx_mm_invsqrt_pd(rsq12);
1376 rinv20 = gmx_mm_invsqrt_pd(rsq20);
1377 rinv21 = gmx_mm_invsqrt_pd(rsq21);
1378 rinv22 = gmx_mm_invsqrt_pd(rsq22);
1380 fjx0 = _mm_setzero_pd();
1381 fjy0 = _mm_setzero_pd();
1382 fjz0 = _mm_setzero_pd();
1383 fjx1 = _mm_setzero_pd();
1384 fjy1 = _mm_setzero_pd();
1385 fjz1 = _mm_setzero_pd();
1386 fjx2 = _mm_setzero_pd();
1387 fjy2 = _mm_setzero_pd();
1388 fjz2 = _mm_setzero_pd();
1390 /**************************
1391 * CALCULATE INTERACTIONS *
1392 **************************/
1394 r00 = _mm_mul_pd(rsq00,rinv00);
1396 /* Calculate table index by multiplying r with table scale and truncate to integer */
1397 rt = _mm_mul_pd(r00,vftabscale);
1398 vfitab = _mm_cvttpd_epi32(rt);
1400 vfeps = _mm_frcz_pd(rt);
1402 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1404 twovfeps = _mm_add_pd(vfeps,vfeps);
1405 vfitab = _mm_slli_epi32(vfitab,2);
1407 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1408 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1409 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1410 GMX_MM_TRANSPOSE2_PD(Y,F);
1411 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1412 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1413 GMX_MM_TRANSPOSE2_PD(G,H);
1414 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1415 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1416 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq00,FF),_mm_mul_pd(vftabscale,rinv00)));
1420 /* Update vectorial force */
1421 fix0 = _mm_macc_pd(dx00,fscal,fix0);
1422 fiy0 = _mm_macc_pd(dy00,fscal,fiy0);
1423 fiz0 = _mm_macc_pd(dz00,fscal,fiz0);
1425 fjx0 = _mm_macc_pd(dx00,fscal,fjx0);
1426 fjy0 = _mm_macc_pd(dy00,fscal,fjy0);
1427 fjz0 = _mm_macc_pd(dz00,fscal,fjz0);
1429 /**************************
1430 * CALCULATE INTERACTIONS *
1431 **************************/
1433 r01 = _mm_mul_pd(rsq01,rinv01);
1435 /* Calculate table index by multiplying r with table scale and truncate to integer */
1436 rt = _mm_mul_pd(r01,vftabscale);
1437 vfitab = _mm_cvttpd_epi32(rt);
1439 vfeps = _mm_frcz_pd(rt);
1441 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1443 twovfeps = _mm_add_pd(vfeps,vfeps);
1444 vfitab = _mm_slli_epi32(vfitab,2);
1446 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1447 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1448 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1449 GMX_MM_TRANSPOSE2_PD(Y,F);
1450 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1451 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1452 GMX_MM_TRANSPOSE2_PD(G,H);
1453 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1454 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1455 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq01,FF),_mm_mul_pd(vftabscale,rinv01)));
1459 /* Update vectorial force */
1460 fix0 = _mm_macc_pd(dx01,fscal,fix0);
1461 fiy0 = _mm_macc_pd(dy01,fscal,fiy0);
1462 fiz0 = _mm_macc_pd(dz01,fscal,fiz0);
1464 fjx1 = _mm_macc_pd(dx01,fscal,fjx1);
1465 fjy1 = _mm_macc_pd(dy01,fscal,fjy1);
1466 fjz1 = _mm_macc_pd(dz01,fscal,fjz1);
1468 /**************************
1469 * CALCULATE INTERACTIONS *
1470 **************************/
1472 r02 = _mm_mul_pd(rsq02,rinv02);
1474 /* Calculate table index by multiplying r with table scale and truncate to integer */
1475 rt = _mm_mul_pd(r02,vftabscale);
1476 vfitab = _mm_cvttpd_epi32(rt);
1478 vfeps = _mm_frcz_pd(rt);
1480 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1482 twovfeps = _mm_add_pd(vfeps,vfeps);
1483 vfitab = _mm_slli_epi32(vfitab,2);
1485 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1486 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1487 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1488 GMX_MM_TRANSPOSE2_PD(Y,F);
1489 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1490 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1491 GMX_MM_TRANSPOSE2_PD(G,H);
1492 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1493 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1494 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq02,FF),_mm_mul_pd(vftabscale,rinv02)));
1498 /* Update vectorial force */
1499 fix0 = _mm_macc_pd(dx02,fscal,fix0);
1500 fiy0 = _mm_macc_pd(dy02,fscal,fiy0);
1501 fiz0 = _mm_macc_pd(dz02,fscal,fiz0);
1503 fjx2 = _mm_macc_pd(dx02,fscal,fjx2);
1504 fjy2 = _mm_macc_pd(dy02,fscal,fjy2);
1505 fjz2 = _mm_macc_pd(dz02,fscal,fjz2);
1507 /**************************
1508 * CALCULATE INTERACTIONS *
1509 **************************/
1511 r10 = _mm_mul_pd(rsq10,rinv10);
1513 /* Calculate table index by multiplying r with table scale and truncate to integer */
1514 rt = _mm_mul_pd(r10,vftabscale);
1515 vfitab = _mm_cvttpd_epi32(rt);
1517 vfeps = _mm_frcz_pd(rt);
1519 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1521 twovfeps = _mm_add_pd(vfeps,vfeps);
1522 vfitab = _mm_slli_epi32(vfitab,2);
1524 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1525 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1526 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1527 GMX_MM_TRANSPOSE2_PD(Y,F);
1528 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1529 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1530 GMX_MM_TRANSPOSE2_PD(G,H);
1531 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1532 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1533 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq10,FF),_mm_mul_pd(vftabscale,rinv10)));
1537 /* Update vectorial force */
1538 fix1 = _mm_macc_pd(dx10,fscal,fix1);
1539 fiy1 = _mm_macc_pd(dy10,fscal,fiy1);
1540 fiz1 = _mm_macc_pd(dz10,fscal,fiz1);
1542 fjx0 = _mm_macc_pd(dx10,fscal,fjx0);
1543 fjy0 = _mm_macc_pd(dy10,fscal,fjy0);
1544 fjz0 = _mm_macc_pd(dz10,fscal,fjz0);
1546 /**************************
1547 * CALCULATE INTERACTIONS *
1548 **************************/
1550 r11 = _mm_mul_pd(rsq11,rinv11);
1552 /* Calculate table index by multiplying r with table scale and truncate to integer */
1553 rt = _mm_mul_pd(r11,vftabscale);
1554 vfitab = _mm_cvttpd_epi32(rt);
1556 vfeps = _mm_frcz_pd(rt);
1558 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1560 twovfeps = _mm_add_pd(vfeps,vfeps);
1561 vfitab = _mm_slli_epi32(vfitab,2);
1563 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1564 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1565 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1566 GMX_MM_TRANSPOSE2_PD(Y,F);
1567 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1568 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1569 GMX_MM_TRANSPOSE2_PD(G,H);
1570 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1571 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1572 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
1576 /* Update vectorial force */
1577 fix1 = _mm_macc_pd(dx11,fscal,fix1);
1578 fiy1 = _mm_macc_pd(dy11,fscal,fiy1);
1579 fiz1 = _mm_macc_pd(dz11,fscal,fiz1);
1581 fjx1 = _mm_macc_pd(dx11,fscal,fjx1);
1582 fjy1 = _mm_macc_pd(dy11,fscal,fjy1);
1583 fjz1 = _mm_macc_pd(dz11,fscal,fjz1);
1585 /**************************
1586 * CALCULATE INTERACTIONS *
1587 **************************/
1589 r12 = _mm_mul_pd(rsq12,rinv12);
1591 /* Calculate table index by multiplying r with table scale and truncate to integer */
1592 rt = _mm_mul_pd(r12,vftabscale);
1593 vfitab = _mm_cvttpd_epi32(rt);
1595 vfeps = _mm_frcz_pd(rt);
1597 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1599 twovfeps = _mm_add_pd(vfeps,vfeps);
1600 vfitab = _mm_slli_epi32(vfitab,2);
1602 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1603 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1604 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1605 GMX_MM_TRANSPOSE2_PD(Y,F);
1606 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1607 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1608 GMX_MM_TRANSPOSE2_PD(G,H);
1609 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1610 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1611 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
1615 /* Update vectorial force */
1616 fix1 = _mm_macc_pd(dx12,fscal,fix1);
1617 fiy1 = _mm_macc_pd(dy12,fscal,fiy1);
1618 fiz1 = _mm_macc_pd(dz12,fscal,fiz1);
1620 fjx2 = _mm_macc_pd(dx12,fscal,fjx2);
1621 fjy2 = _mm_macc_pd(dy12,fscal,fjy2);
1622 fjz2 = _mm_macc_pd(dz12,fscal,fjz2);
1624 /**************************
1625 * CALCULATE INTERACTIONS *
1626 **************************/
1628 r20 = _mm_mul_pd(rsq20,rinv20);
1630 /* Calculate table index by multiplying r with table scale and truncate to integer */
1631 rt = _mm_mul_pd(r20,vftabscale);
1632 vfitab = _mm_cvttpd_epi32(rt);
1634 vfeps = _mm_frcz_pd(rt);
1636 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1638 twovfeps = _mm_add_pd(vfeps,vfeps);
1639 vfitab = _mm_slli_epi32(vfitab,2);
1641 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1642 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1643 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1644 GMX_MM_TRANSPOSE2_PD(Y,F);
1645 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1646 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1647 GMX_MM_TRANSPOSE2_PD(G,H);
1648 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1649 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1650 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq20,FF),_mm_mul_pd(vftabscale,rinv20)));
1654 /* Update vectorial force */
1655 fix2 = _mm_macc_pd(dx20,fscal,fix2);
1656 fiy2 = _mm_macc_pd(dy20,fscal,fiy2);
1657 fiz2 = _mm_macc_pd(dz20,fscal,fiz2);
1659 fjx0 = _mm_macc_pd(dx20,fscal,fjx0);
1660 fjy0 = _mm_macc_pd(dy20,fscal,fjy0);
1661 fjz0 = _mm_macc_pd(dz20,fscal,fjz0);
1663 /**************************
1664 * CALCULATE INTERACTIONS *
1665 **************************/
1667 r21 = _mm_mul_pd(rsq21,rinv21);
1669 /* Calculate table index by multiplying r with table scale and truncate to integer */
1670 rt = _mm_mul_pd(r21,vftabscale);
1671 vfitab = _mm_cvttpd_epi32(rt);
1673 vfeps = _mm_frcz_pd(rt);
1675 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1677 twovfeps = _mm_add_pd(vfeps,vfeps);
1678 vfitab = _mm_slli_epi32(vfitab,2);
1680 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1681 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1682 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1683 GMX_MM_TRANSPOSE2_PD(Y,F);
1684 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1685 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1686 GMX_MM_TRANSPOSE2_PD(G,H);
1687 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1688 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1689 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
1693 /* Update vectorial force */
1694 fix2 = _mm_macc_pd(dx21,fscal,fix2);
1695 fiy2 = _mm_macc_pd(dy21,fscal,fiy2);
1696 fiz2 = _mm_macc_pd(dz21,fscal,fiz2);
1698 fjx1 = _mm_macc_pd(dx21,fscal,fjx1);
1699 fjy1 = _mm_macc_pd(dy21,fscal,fjy1);
1700 fjz1 = _mm_macc_pd(dz21,fscal,fjz1);
1702 /**************************
1703 * CALCULATE INTERACTIONS *
1704 **************************/
1706 r22 = _mm_mul_pd(rsq22,rinv22);
1708 /* Calculate table index by multiplying r with table scale and truncate to integer */
1709 rt = _mm_mul_pd(r22,vftabscale);
1710 vfitab = _mm_cvttpd_epi32(rt);
1712 vfeps = _mm_frcz_pd(rt);
1714 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1716 twovfeps = _mm_add_pd(vfeps,vfeps);
1717 vfitab = _mm_slli_epi32(vfitab,2);
1719 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1720 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1721 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1722 GMX_MM_TRANSPOSE2_PD(Y,F);
1723 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1724 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1725 GMX_MM_TRANSPOSE2_PD(G,H);
1726 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1727 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1728 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
1732 /* Update vectorial force */
1733 fix2 = _mm_macc_pd(dx22,fscal,fix2);
1734 fiy2 = _mm_macc_pd(dy22,fscal,fiy2);
1735 fiz2 = _mm_macc_pd(dz22,fscal,fiz2);
1737 fjx2 = _mm_macc_pd(dx22,fscal,fjx2);
1738 fjy2 = _mm_macc_pd(dy22,fscal,fjy2);
1739 fjz2 = _mm_macc_pd(dz22,fscal,fjz2);
1741 gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1743 /* Inner loop uses 378 flops */
1746 if(jidx<j_index_end)
1750 j_coord_offsetA = DIM*jnrA;
1752 /* load j atom coordinates */
1753 gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
1754 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1756 /* Calculate displacement vector */
1757 dx00 = _mm_sub_pd(ix0,jx0);
1758 dy00 = _mm_sub_pd(iy0,jy0);
1759 dz00 = _mm_sub_pd(iz0,jz0);
1760 dx01 = _mm_sub_pd(ix0,jx1);
1761 dy01 = _mm_sub_pd(iy0,jy1);
1762 dz01 = _mm_sub_pd(iz0,jz1);
1763 dx02 = _mm_sub_pd(ix0,jx2);
1764 dy02 = _mm_sub_pd(iy0,jy2);
1765 dz02 = _mm_sub_pd(iz0,jz2);
1766 dx10 = _mm_sub_pd(ix1,jx0);
1767 dy10 = _mm_sub_pd(iy1,jy0);
1768 dz10 = _mm_sub_pd(iz1,jz0);
1769 dx11 = _mm_sub_pd(ix1,jx1);
1770 dy11 = _mm_sub_pd(iy1,jy1);
1771 dz11 = _mm_sub_pd(iz1,jz1);
1772 dx12 = _mm_sub_pd(ix1,jx2);
1773 dy12 = _mm_sub_pd(iy1,jy2);
1774 dz12 = _mm_sub_pd(iz1,jz2);
1775 dx20 = _mm_sub_pd(ix2,jx0);
1776 dy20 = _mm_sub_pd(iy2,jy0);
1777 dz20 = _mm_sub_pd(iz2,jz0);
1778 dx21 = _mm_sub_pd(ix2,jx1);
1779 dy21 = _mm_sub_pd(iy2,jy1);
1780 dz21 = _mm_sub_pd(iz2,jz1);
1781 dx22 = _mm_sub_pd(ix2,jx2);
1782 dy22 = _mm_sub_pd(iy2,jy2);
1783 dz22 = _mm_sub_pd(iz2,jz2);
1785 /* Calculate squared distance and things based on it */
1786 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1787 rsq01 = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
1788 rsq02 = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
1789 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
1790 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1791 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1792 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
1793 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1794 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1796 rinv00 = gmx_mm_invsqrt_pd(rsq00);
1797 rinv01 = gmx_mm_invsqrt_pd(rsq01);
1798 rinv02 = gmx_mm_invsqrt_pd(rsq02);
1799 rinv10 = gmx_mm_invsqrt_pd(rsq10);
1800 rinv11 = gmx_mm_invsqrt_pd(rsq11);
1801 rinv12 = gmx_mm_invsqrt_pd(rsq12);
1802 rinv20 = gmx_mm_invsqrt_pd(rsq20);
1803 rinv21 = gmx_mm_invsqrt_pd(rsq21);
1804 rinv22 = gmx_mm_invsqrt_pd(rsq22);
1806 fjx0 = _mm_setzero_pd();
1807 fjy0 = _mm_setzero_pd();
1808 fjz0 = _mm_setzero_pd();
1809 fjx1 = _mm_setzero_pd();
1810 fjy1 = _mm_setzero_pd();
1811 fjz1 = _mm_setzero_pd();
1812 fjx2 = _mm_setzero_pd();
1813 fjy2 = _mm_setzero_pd();
1814 fjz2 = _mm_setzero_pd();
1816 /**************************
1817 * CALCULATE INTERACTIONS *
1818 **************************/
1820 r00 = _mm_mul_pd(rsq00,rinv00);
1822 /* Calculate table index by multiplying r with table scale and truncate to integer */
1823 rt = _mm_mul_pd(r00,vftabscale);
1824 vfitab = _mm_cvttpd_epi32(rt);
1826 vfeps = _mm_frcz_pd(rt);
1828 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1830 twovfeps = _mm_add_pd(vfeps,vfeps);
1831 vfitab = _mm_slli_epi32(vfitab,2);
1833 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1834 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1835 F = _mm_setzero_pd();
1836 GMX_MM_TRANSPOSE2_PD(Y,F);
1837 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1838 H = _mm_setzero_pd();
1839 GMX_MM_TRANSPOSE2_PD(G,H);
1840 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1841 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1842 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq00,FF),_mm_mul_pd(vftabscale,rinv00)));
1846 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1848 /* Update vectorial force */
1849 fix0 = _mm_macc_pd(dx00,fscal,fix0);
1850 fiy0 = _mm_macc_pd(dy00,fscal,fiy0);
1851 fiz0 = _mm_macc_pd(dz00,fscal,fiz0);
1853 fjx0 = _mm_macc_pd(dx00,fscal,fjx0);
1854 fjy0 = _mm_macc_pd(dy00,fscal,fjy0);
1855 fjz0 = _mm_macc_pd(dz00,fscal,fjz0);
1857 /**************************
1858 * CALCULATE INTERACTIONS *
1859 **************************/
1861 r01 = _mm_mul_pd(rsq01,rinv01);
1863 /* Calculate table index by multiplying r with table scale and truncate to integer */
1864 rt = _mm_mul_pd(r01,vftabscale);
1865 vfitab = _mm_cvttpd_epi32(rt);
1867 vfeps = _mm_frcz_pd(rt);
1869 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1871 twovfeps = _mm_add_pd(vfeps,vfeps);
1872 vfitab = _mm_slli_epi32(vfitab,2);
1874 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1875 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1876 F = _mm_setzero_pd();
1877 GMX_MM_TRANSPOSE2_PD(Y,F);
1878 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1879 H = _mm_setzero_pd();
1880 GMX_MM_TRANSPOSE2_PD(G,H);
1881 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1882 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1883 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq01,FF),_mm_mul_pd(vftabscale,rinv01)));
1887 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1889 /* Update vectorial force */
1890 fix0 = _mm_macc_pd(dx01,fscal,fix0);
1891 fiy0 = _mm_macc_pd(dy01,fscal,fiy0);
1892 fiz0 = _mm_macc_pd(dz01,fscal,fiz0);
1894 fjx1 = _mm_macc_pd(dx01,fscal,fjx1);
1895 fjy1 = _mm_macc_pd(dy01,fscal,fjy1);
1896 fjz1 = _mm_macc_pd(dz01,fscal,fjz1);
1898 /**************************
1899 * CALCULATE INTERACTIONS *
1900 **************************/
1902 r02 = _mm_mul_pd(rsq02,rinv02);
1904 /* Calculate table index by multiplying r with table scale and truncate to integer */
1905 rt = _mm_mul_pd(r02,vftabscale);
1906 vfitab = _mm_cvttpd_epi32(rt);
1908 vfeps = _mm_frcz_pd(rt);
1910 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1912 twovfeps = _mm_add_pd(vfeps,vfeps);
1913 vfitab = _mm_slli_epi32(vfitab,2);
1915 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1916 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1917 F = _mm_setzero_pd();
1918 GMX_MM_TRANSPOSE2_PD(Y,F);
1919 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1920 H = _mm_setzero_pd();
1921 GMX_MM_TRANSPOSE2_PD(G,H);
1922 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1923 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1924 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq02,FF),_mm_mul_pd(vftabscale,rinv02)));
1928 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1930 /* Update vectorial force */
1931 fix0 = _mm_macc_pd(dx02,fscal,fix0);
1932 fiy0 = _mm_macc_pd(dy02,fscal,fiy0);
1933 fiz0 = _mm_macc_pd(dz02,fscal,fiz0);
1935 fjx2 = _mm_macc_pd(dx02,fscal,fjx2);
1936 fjy2 = _mm_macc_pd(dy02,fscal,fjy2);
1937 fjz2 = _mm_macc_pd(dz02,fscal,fjz2);
1939 /**************************
1940 * CALCULATE INTERACTIONS *
1941 **************************/
1943 r10 = _mm_mul_pd(rsq10,rinv10);
1945 /* Calculate table index by multiplying r with table scale and truncate to integer */
1946 rt = _mm_mul_pd(r10,vftabscale);
1947 vfitab = _mm_cvttpd_epi32(rt);
1949 vfeps = _mm_frcz_pd(rt);
1951 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1953 twovfeps = _mm_add_pd(vfeps,vfeps);
1954 vfitab = _mm_slli_epi32(vfitab,2);
1956 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1957 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1958 F = _mm_setzero_pd();
1959 GMX_MM_TRANSPOSE2_PD(Y,F);
1960 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1961 H = _mm_setzero_pd();
1962 GMX_MM_TRANSPOSE2_PD(G,H);
1963 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1964 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1965 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq10,FF),_mm_mul_pd(vftabscale,rinv10)));
1969 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1971 /* Update vectorial force */
1972 fix1 = _mm_macc_pd(dx10,fscal,fix1);
1973 fiy1 = _mm_macc_pd(dy10,fscal,fiy1);
1974 fiz1 = _mm_macc_pd(dz10,fscal,fiz1);
1976 fjx0 = _mm_macc_pd(dx10,fscal,fjx0);
1977 fjy0 = _mm_macc_pd(dy10,fscal,fjy0);
1978 fjz0 = _mm_macc_pd(dz10,fscal,fjz0);
1980 /**************************
1981 * CALCULATE INTERACTIONS *
1982 **************************/
1984 r11 = _mm_mul_pd(rsq11,rinv11);
1986 /* Calculate table index by multiplying r with table scale and truncate to integer */
1987 rt = _mm_mul_pd(r11,vftabscale);
1988 vfitab = _mm_cvttpd_epi32(rt);
1990 vfeps = _mm_frcz_pd(rt);
1992 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1994 twovfeps = _mm_add_pd(vfeps,vfeps);
1995 vfitab = _mm_slli_epi32(vfitab,2);
1997 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1998 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1999 F = _mm_setzero_pd();
2000 GMX_MM_TRANSPOSE2_PD(Y,F);
2001 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
2002 H = _mm_setzero_pd();
2003 GMX_MM_TRANSPOSE2_PD(G,H);
2004 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
2005 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
2006 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
2010 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2012 /* Update vectorial force */
2013 fix1 = _mm_macc_pd(dx11,fscal,fix1);
2014 fiy1 = _mm_macc_pd(dy11,fscal,fiy1);
2015 fiz1 = _mm_macc_pd(dz11,fscal,fiz1);
2017 fjx1 = _mm_macc_pd(dx11,fscal,fjx1);
2018 fjy1 = _mm_macc_pd(dy11,fscal,fjy1);
2019 fjz1 = _mm_macc_pd(dz11,fscal,fjz1);
2021 /**************************
2022 * CALCULATE INTERACTIONS *
2023 **************************/
2025 r12 = _mm_mul_pd(rsq12,rinv12);
2027 /* Calculate table index by multiplying r with table scale and truncate to integer */
2028 rt = _mm_mul_pd(r12,vftabscale);
2029 vfitab = _mm_cvttpd_epi32(rt);
2031 vfeps = _mm_frcz_pd(rt);
2033 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2035 twovfeps = _mm_add_pd(vfeps,vfeps);
2036 vfitab = _mm_slli_epi32(vfitab,2);
2038 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2039 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2040 F = _mm_setzero_pd();
2041 GMX_MM_TRANSPOSE2_PD(Y,F);
2042 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
2043 H = _mm_setzero_pd();
2044 GMX_MM_TRANSPOSE2_PD(G,H);
2045 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
2046 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
2047 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
2051 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2053 /* Update vectorial force */
2054 fix1 = _mm_macc_pd(dx12,fscal,fix1);
2055 fiy1 = _mm_macc_pd(dy12,fscal,fiy1);
2056 fiz1 = _mm_macc_pd(dz12,fscal,fiz1);
2058 fjx2 = _mm_macc_pd(dx12,fscal,fjx2);
2059 fjy2 = _mm_macc_pd(dy12,fscal,fjy2);
2060 fjz2 = _mm_macc_pd(dz12,fscal,fjz2);
2062 /**************************
2063 * CALCULATE INTERACTIONS *
2064 **************************/
2066 r20 = _mm_mul_pd(rsq20,rinv20);
2068 /* Calculate table index by multiplying r with table scale and truncate to integer */
2069 rt = _mm_mul_pd(r20,vftabscale);
2070 vfitab = _mm_cvttpd_epi32(rt);
2072 vfeps = _mm_frcz_pd(rt);
2074 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2076 twovfeps = _mm_add_pd(vfeps,vfeps);
2077 vfitab = _mm_slli_epi32(vfitab,2);
2079 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2080 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2081 F = _mm_setzero_pd();
2082 GMX_MM_TRANSPOSE2_PD(Y,F);
2083 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
2084 H = _mm_setzero_pd();
2085 GMX_MM_TRANSPOSE2_PD(G,H);
2086 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
2087 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
2088 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq20,FF),_mm_mul_pd(vftabscale,rinv20)));
2092 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2094 /* Update vectorial force */
2095 fix2 = _mm_macc_pd(dx20,fscal,fix2);
2096 fiy2 = _mm_macc_pd(dy20,fscal,fiy2);
2097 fiz2 = _mm_macc_pd(dz20,fscal,fiz2);
2099 fjx0 = _mm_macc_pd(dx20,fscal,fjx0);
2100 fjy0 = _mm_macc_pd(dy20,fscal,fjy0);
2101 fjz0 = _mm_macc_pd(dz20,fscal,fjz0);
2103 /**************************
2104 * CALCULATE INTERACTIONS *
2105 **************************/
2107 r21 = _mm_mul_pd(rsq21,rinv21);
2109 /* Calculate table index by multiplying r with table scale and truncate to integer */
2110 rt = _mm_mul_pd(r21,vftabscale);
2111 vfitab = _mm_cvttpd_epi32(rt);
2113 vfeps = _mm_frcz_pd(rt);
2115 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2117 twovfeps = _mm_add_pd(vfeps,vfeps);
2118 vfitab = _mm_slli_epi32(vfitab,2);
2120 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2121 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2122 F = _mm_setzero_pd();
2123 GMX_MM_TRANSPOSE2_PD(Y,F);
2124 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
2125 H = _mm_setzero_pd();
2126 GMX_MM_TRANSPOSE2_PD(G,H);
2127 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
2128 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
2129 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
2133 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2135 /* Update vectorial force */
2136 fix2 = _mm_macc_pd(dx21,fscal,fix2);
2137 fiy2 = _mm_macc_pd(dy21,fscal,fiy2);
2138 fiz2 = _mm_macc_pd(dz21,fscal,fiz2);
2140 fjx1 = _mm_macc_pd(dx21,fscal,fjx1);
2141 fjy1 = _mm_macc_pd(dy21,fscal,fjy1);
2142 fjz1 = _mm_macc_pd(dz21,fscal,fjz1);
2144 /**************************
2145 * CALCULATE INTERACTIONS *
2146 **************************/
2148 r22 = _mm_mul_pd(rsq22,rinv22);
2150 /* Calculate table index by multiplying r with table scale and truncate to integer */
2151 rt = _mm_mul_pd(r22,vftabscale);
2152 vfitab = _mm_cvttpd_epi32(rt);
2154 vfeps = _mm_frcz_pd(rt);
2156 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2158 twovfeps = _mm_add_pd(vfeps,vfeps);
2159 vfitab = _mm_slli_epi32(vfitab,2);
2161 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2162 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2163 F = _mm_setzero_pd();
2164 GMX_MM_TRANSPOSE2_PD(Y,F);
2165 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
2166 H = _mm_setzero_pd();
2167 GMX_MM_TRANSPOSE2_PD(G,H);
2168 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
2169 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
2170 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
2174 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2176 /* Update vectorial force */
2177 fix2 = _mm_macc_pd(dx22,fscal,fix2);
2178 fiy2 = _mm_macc_pd(dy22,fscal,fiy2);
2179 fiz2 = _mm_macc_pd(dz22,fscal,fiz2);
2181 fjx2 = _mm_macc_pd(dx22,fscal,fjx2);
2182 fjy2 = _mm_macc_pd(dy22,fscal,fjy2);
2183 fjz2 = _mm_macc_pd(dz22,fscal,fjz2);
2185 gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
2187 /* Inner loop uses 378 flops */
2190 /* End of innermost loop */
2192 gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
2193 f+i_coord_offset,fshift+i_shift_offset);
2195 /* Increment number of inner iterations */
2196 inneriter += j_index_end - j_index_start;
2198 /* Outer loop uses 18 flops */
2201 /* Increment number of outer iterations */
2204 /* Update outer/inner flops */
2206 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3W3_F,outeriter*18 + inneriter*378);