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_VdwCSTab_GeomW3W3_VF_avx_128_fma_double
54 * Electrostatics interaction: CubicSplineTable
55 * VdW interaction: CubicSplineTable
56 * Geometry: Water3-Water3
57 * Calculate force/pot: PotentialAndForce
60 nb_kernel_ElecCSTab_VdwCSTab_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 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
109 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
110 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
112 __m128i ifour = _mm_set1_epi32(4);
113 __m128d rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
115 __m128d dummy_mask,cutoff_mask;
116 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
117 __m128d one = _mm_set1_pd(1.0);
118 __m128d two = _mm_set1_pd(2.0);
124 jindex = nlist->jindex;
126 shiftidx = nlist->shift;
128 shiftvec = fr->shift_vec[0];
129 fshift = fr->fshift[0];
130 facel = _mm_set1_pd(fr->epsfac);
131 charge = mdatoms->chargeA;
132 nvdwtype = fr->ntype;
134 vdwtype = mdatoms->typeA;
136 vftab = kernel_data->table_elec_vdw->data;
137 vftabscale = _mm_set1_pd(kernel_data->table_elec_vdw->scale);
139 /* Setup water-specific parameters */
140 inr = nlist->iinr[0];
141 iq0 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
142 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
143 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
144 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
146 jq0 = _mm_set1_pd(charge[inr+0]);
147 jq1 = _mm_set1_pd(charge[inr+1]);
148 jq2 = _mm_set1_pd(charge[inr+2]);
149 vdwjidx0A = 2*vdwtype[inr+0];
150 qq00 = _mm_mul_pd(iq0,jq0);
151 c6_00 = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
152 c12_00 = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
153 qq01 = _mm_mul_pd(iq0,jq1);
154 qq02 = _mm_mul_pd(iq0,jq2);
155 qq10 = _mm_mul_pd(iq1,jq0);
156 qq11 = _mm_mul_pd(iq1,jq1);
157 qq12 = _mm_mul_pd(iq1,jq2);
158 qq20 = _mm_mul_pd(iq2,jq0);
159 qq21 = _mm_mul_pd(iq2,jq1);
160 qq22 = _mm_mul_pd(iq2,jq2);
162 /* Avoid stupid compiler warnings */
170 /* Start outer loop over neighborlists */
171 for(iidx=0; iidx<nri; iidx++)
173 /* Load shift vector for this list */
174 i_shift_offset = DIM*shiftidx[iidx];
176 /* Load limits for loop over neighbors */
177 j_index_start = jindex[iidx];
178 j_index_end = jindex[iidx+1];
180 /* Get outer coordinate index */
182 i_coord_offset = DIM*inr;
184 /* Load i particle coords and add shift vector */
185 gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
186 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
188 fix0 = _mm_setzero_pd();
189 fiy0 = _mm_setzero_pd();
190 fiz0 = _mm_setzero_pd();
191 fix1 = _mm_setzero_pd();
192 fiy1 = _mm_setzero_pd();
193 fiz1 = _mm_setzero_pd();
194 fix2 = _mm_setzero_pd();
195 fiy2 = _mm_setzero_pd();
196 fiz2 = _mm_setzero_pd();
198 /* Reset potential sums */
199 velecsum = _mm_setzero_pd();
200 vvdwsum = _mm_setzero_pd();
202 /* Start inner kernel loop */
203 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
206 /* Get j neighbor index, and coordinate index */
209 j_coord_offsetA = DIM*jnrA;
210 j_coord_offsetB = DIM*jnrB;
212 /* load j atom coordinates */
213 gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
214 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
216 /* Calculate displacement vector */
217 dx00 = _mm_sub_pd(ix0,jx0);
218 dy00 = _mm_sub_pd(iy0,jy0);
219 dz00 = _mm_sub_pd(iz0,jz0);
220 dx01 = _mm_sub_pd(ix0,jx1);
221 dy01 = _mm_sub_pd(iy0,jy1);
222 dz01 = _mm_sub_pd(iz0,jz1);
223 dx02 = _mm_sub_pd(ix0,jx2);
224 dy02 = _mm_sub_pd(iy0,jy2);
225 dz02 = _mm_sub_pd(iz0,jz2);
226 dx10 = _mm_sub_pd(ix1,jx0);
227 dy10 = _mm_sub_pd(iy1,jy0);
228 dz10 = _mm_sub_pd(iz1,jz0);
229 dx11 = _mm_sub_pd(ix1,jx1);
230 dy11 = _mm_sub_pd(iy1,jy1);
231 dz11 = _mm_sub_pd(iz1,jz1);
232 dx12 = _mm_sub_pd(ix1,jx2);
233 dy12 = _mm_sub_pd(iy1,jy2);
234 dz12 = _mm_sub_pd(iz1,jz2);
235 dx20 = _mm_sub_pd(ix2,jx0);
236 dy20 = _mm_sub_pd(iy2,jy0);
237 dz20 = _mm_sub_pd(iz2,jz0);
238 dx21 = _mm_sub_pd(ix2,jx1);
239 dy21 = _mm_sub_pd(iy2,jy1);
240 dz21 = _mm_sub_pd(iz2,jz1);
241 dx22 = _mm_sub_pd(ix2,jx2);
242 dy22 = _mm_sub_pd(iy2,jy2);
243 dz22 = _mm_sub_pd(iz2,jz2);
245 /* Calculate squared distance and things based on it */
246 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
247 rsq01 = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
248 rsq02 = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
249 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
250 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
251 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
252 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
253 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
254 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
256 rinv00 = gmx_mm_invsqrt_pd(rsq00);
257 rinv01 = gmx_mm_invsqrt_pd(rsq01);
258 rinv02 = gmx_mm_invsqrt_pd(rsq02);
259 rinv10 = gmx_mm_invsqrt_pd(rsq10);
260 rinv11 = gmx_mm_invsqrt_pd(rsq11);
261 rinv12 = gmx_mm_invsqrt_pd(rsq12);
262 rinv20 = gmx_mm_invsqrt_pd(rsq20);
263 rinv21 = gmx_mm_invsqrt_pd(rsq21);
264 rinv22 = gmx_mm_invsqrt_pd(rsq22);
266 fjx0 = _mm_setzero_pd();
267 fjy0 = _mm_setzero_pd();
268 fjz0 = _mm_setzero_pd();
269 fjx1 = _mm_setzero_pd();
270 fjy1 = _mm_setzero_pd();
271 fjz1 = _mm_setzero_pd();
272 fjx2 = _mm_setzero_pd();
273 fjy2 = _mm_setzero_pd();
274 fjz2 = _mm_setzero_pd();
276 /**************************
277 * CALCULATE INTERACTIONS *
278 **************************/
280 r00 = _mm_mul_pd(rsq00,rinv00);
282 /* Calculate table index by multiplying r with table scale and truncate to integer */
283 rt = _mm_mul_pd(r00,vftabscale);
284 vfitab = _mm_cvttpd_epi32(rt);
286 vfeps = _mm_frcz_pd(rt);
288 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
290 twovfeps = _mm_add_pd(vfeps,vfeps);
291 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
293 /* CUBIC SPLINE TABLE ELECTROSTATICS */
294 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
295 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
296 GMX_MM_TRANSPOSE2_PD(Y,F);
297 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
298 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
299 GMX_MM_TRANSPOSE2_PD(G,H);
300 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
301 VV = _mm_macc_pd(vfeps,Fp,Y);
302 velec = _mm_mul_pd(qq00,VV);
303 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
304 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq00,FF),_mm_mul_pd(vftabscale,rinv00)));
306 /* CUBIC SPLINE TABLE DISPERSION */
307 vfitab = _mm_add_epi32(vfitab,ifour);
308 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
309 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
310 GMX_MM_TRANSPOSE2_PD(Y,F);
311 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
312 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
313 GMX_MM_TRANSPOSE2_PD(G,H);
314 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(H,vfeps,G),F);
315 VV = _mm_macc_pd(vfeps,Fp,Y);
316 vvdw6 = _mm_mul_pd(c6_00,VV);
317 FF = _mm_macc_pd(vfeps,_mm_macc_pd(twovfeps,H,G),Fp);
318 fvdw6 = _mm_mul_pd(c6_00,FF);
320 /* CUBIC SPLINE TABLE REPULSION */
321 vfitab = _mm_add_epi32(vfitab,ifour);
322 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
323 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
324 GMX_MM_TRANSPOSE2_PD(Y,F);
325 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
326 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
327 GMX_MM_TRANSPOSE2_PD(G,H);
328 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(H,vfeps,G),F);
329 VV = _mm_macc_pd(vfeps,Fp,Y);
330 vvdw12 = _mm_mul_pd(c12_00,VV);
331 FF = _mm_macc_pd(vfeps,_mm_macc_pd(twovfeps,H,G),Fp);
332 fvdw12 = _mm_mul_pd(c12_00,FF);
333 vvdw = _mm_add_pd(vvdw12,vvdw6);
334 fvdw = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
336 /* Update potential sum for this i atom from the interaction with this j atom. */
337 velecsum = _mm_add_pd(velecsum,velec);
338 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
340 fscal = _mm_add_pd(felec,fvdw);
342 /* Update vectorial force */
343 fix0 = _mm_macc_pd(dx00,fscal,fix0);
344 fiy0 = _mm_macc_pd(dy00,fscal,fiy0);
345 fiz0 = _mm_macc_pd(dz00,fscal,fiz0);
347 fjx0 = _mm_macc_pd(dx00,fscal,fjx0);
348 fjy0 = _mm_macc_pd(dy00,fscal,fjy0);
349 fjz0 = _mm_macc_pd(dz00,fscal,fjz0);
351 /**************************
352 * CALCULATE INTERACTIONS *
353 **************************/
355 r01 = _mm_mul_pd(rsq01,rinv01);
357 /* Calculate table index by multiplying r with table scale and truncate to integer */
358 rt = _mm_mul_pd(r01,vftabscale);
359 vfitab = _mm_cvttpd_epi32(rt);
361 vfeps = _mm_frcz_pd(rt);
363 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
365 twovfeps = _mm_add_pd(vfeps,vfeps);
366 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
368 /* CUBIC SPLINE TABLE ELECTROSTATICS */
369 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
370 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
371 GMX_MM_TRANSPOSE2_PD(Y,F);
372 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
373 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
374 GMX_MM_TRANSPOSE2_PD(G,H);
375 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
376 VV = _mm_macc_pd(vfeps,Fp,Y);
377 velec = _mm_mul_pd(qq01,VV);
378 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
379 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq01,FF),_mm_mul_pd(vftabscale,rinv01)));
381 /* Update potential sum for this i atom from the interaction with this j atom. */
382 velecsum = _mm_add_pd(velecsum,velec);
386 /* Update vectorial force */
387 fix0 = _mm_macc_pd(dx01,fscal,fix0);
388 fiy0 = _mm_macc_pd(dy01,fscal,fiy0);
389 fiz0 = _mm_macc_pd(dz01,fscal,fiz0);
391 fjx1 = _mm_macc_pd(dx01,fscal,fjx1);
392 fjy1 = _mm_macc_pd(dy01,fscal,fjy1);
393 fjz1 = _mm_macc_pd(dz01,fscal,fjz1);
395 /**************************
396 * CALCULATE INTERACTIONS *
397 **************************/
399 r02 = _mm_mul_pd(rsq02,rinv02);
401 /* Calculate table index by multiplying r with table scale and truncate to integer */
402 rt = _mm_mul_pd(r02,vftabscale);
403 vfitab = _mm_cvttpd_epi32(rt);
405 vfeps = _mm_frcz_pd(rt);
407 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
409 twovfeps = _mm_add_pd(vfeps,vfeps);
410 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
412 /* CUBIC SPLINE TABLE ELECTROSTATICS */
413 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
414 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
415 GMX_MM_TRANSPOSE2_PD(Y,F);
416 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
417 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
418 GMX_MM_TRANSPOSE2_PD(G,H);
419 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
420 VV = _mm_macc_pd(vfeps,Fp,Y);
421 velec = _mm_mul_pd(qq02,VV);
422 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
423 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq02,FF),_mm_mul_pd(vftabscale,rinv02)));
425 /* Update potential sum for this i atom from the interaction with this j atom. */
426 velecsum = _mm_add_pd(velecsum,velec);
430 /* Update vectorial force */
431 fix0 = _mm_macc_pd(dx02,fscal,fix0);
432 fiy0 = _mm_macc_pd(dy02,fscal,fiy0);
433 fiz0 = _mm_macc_pd(dz02,fscal,fiz0);
435 fjx2 = _mm_macc_pd(dx02,fscal,fjx2);
436 fjy2 = _mm_macc_pd(dy02,fscal,fjy2);
437 fjz2 = _mm_macc_pd(dz02,fscal,fjz2);
439 /**************************
440 * CALCULATE INTERACTIONS *
441 **************************/
443 r10 = _mm_mul_pd(rsq10,rinv10);
445 /* Calculate table index by multiplying r with table scale and truncate to integer */
446 rt = _mm_mul_pd(r10,vftabscale);
447 vfitab = _mm_cvttpd_epi32(rt);
449 vfeps = _mm_frcz_pd(rt);
451 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
453 twovfeps = _mm_add_pd(vfeps,vfeps);
454 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
456 /* CUBIC SPLINE TABLE ELECTROSTATICS */
457 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
458 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
459 GMX_MM_TRANSPOSE2_PD(Y,F);
460 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
461 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
462 GMX_MM_TRANSPOSE2_PD(G,H);
463 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
464 VV = _mm_macc_pd(vfeps,Fp,Y);
465 velec = _mm_mul_pd(qq10,VV);
466 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
467 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq10,FF),_mm_mul_pd(vftabscale,rinv10)));
469 /* Update potential sum for this i atom from the interaction with this j atom. */
470 velecsum = _mm_add_pd(velecsum,velec);
474 /* Update vectorial force */
475 fix1 = _mm_macc_pd(dx10,fscal,fix1);
476 fiy1 = _mm_macc_pd(dy10,fscal,fiy1);
477 fiz1 = _mm_macc_pd(dz10,fscal,fiz1);
479 fjx0 = _mm_macc_pd(dx10,fscal,fjx0);
480 fjy0 = _mm_macc_pd(dy10,fscal,fjy0);
481 fjz0 = _mm_macc_pd(dz10,fscal,fjz0);
483 /**************************
484 * CALCULATE INTERACTIONS *
485 **************************/
487 r11 = _mm_mul_pd(rsq11,rinv11);
489 /* Calculate table index by multiplying r with table scale and truncate to integer */
490 rt = _mm_mul_pd(r11,vftabscale);
491 vfitab = _mm_cvttpd_epi32(rt);
493 vfeps = _mm_frcz_pd(rt);
495 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
497 twovfeps = _mm_add_pd(vfeps,vfeps);
498 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
500 /* CUBIC SPLINE TABLE ELECTROSTATICS */
501 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
502 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
503 GMX_MM_TRANSPOSE2_PD(Y,F);
504 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
505 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
506 GMX_MM_TRANSPOSE2_PD(G,H);
507 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
508 VV = _mm_macc_pd(vfeps,Fp,Y);
509 velec = _mm_mul_pd(qq11,VV);
510 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
511 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
513 /* Update potential sum for this i atom from the interaction with this j atom. */
514 velecsum = _mm_add_pd(velecsum,velec);
518 /* Update vectorial force */
519 fix1 = _mm_macc_pd(dx11,fscal,fix1);
520 fiy1 = _mm_macc_pd(dy11,fscal,fiy1);
521 fiz1 = _mm_macc_pd(dz11,fscal,fiz1);
523 fjx1 = _mm_macc_pd(dx11,fscal,fjx1);
524 fjy1 = _mm_macc_pd(dy11,fscal,fjy1);
525 fjz1 = _mm_macc_pd(dz11,fscal,fjz1);
527 /**************************
528 * CALCULATE INTERACTIONS *
529 **************************/
531 r12 = _mm_mul_pd(rsq12,rinv12);
533 /* Calculate table index by multiplying r with table scale and truncate to integer */
534 rt = _mm_mul_pd(r12,vftabscale);
535 vfitab = _mm_cvttpd_epi32(rt);
537 vfeps = _mm_frcz_pd(rt);
539 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
541 twovfeps = _mm_add_pd(vfeps,vfeps);
542 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
544 /* CUBIC SPLINE TABLE ELECTROSTATICS */
545 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
546 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
547 GMX_MM_TRANSPOSE2_PD(Y,F);
548 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
549 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
550 GMX_MM_TRANSPOSE2_PD(G,H);
551 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
552 VV = _mm_macc_pd(vfeps,Fp,Y);
553 velec = _mm_mul_pd(qq12,VV);
554 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
555 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
557 /* Update potential sum for this i atom from the interaction with this j atom. */
558 velecsum = _mm_add_pd(velecsum,velec);
562 /* Update vectorial force */
563 fix1 = _mm_macc_pd(dx12,fscal,fix1);
564 fiy1 = _mm_macc_pd(dy12,fscal,fiy1);
565 fiz1 = _mm_macc_pd(dz12,fscal,fiz1);
567 fjx2 = _mm_macc_pd(dx12,fscal,fjx2);
568 fjy2 = _mm_macc_pd(dy12,fscal,fjy2);
569 fjz2 = _mm_macc_pd(dz12,fscal,fjz2);
571 /**************************
572 * CALCULATE INTERACTIONS *
573 **************************/
575 r20 = _mm_mul_pd(rsq20,rinv20);
577 /* Calculate table index by multiplying r with table scale and truncate to integer */
578 rt = _mm_mul_pd(r20,vftabscale);
579 vfitab = _mm_cvttpd_epi32(rt);
581 vfeps = _mm_frcz_pd(rt);
583 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
585 twovfeps = _mm_add_pd(vfeps,vfeps);
586 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
588 /* CUBIC SPLINE TABLE ELECTROSTATICS */
589 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
590 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
591 GMX_MM_TRANSPOSE2_PD(Y,F);
592 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
593 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
594 GMX_MM_TRANSPOSE2_PD(G,H);
595 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
596 VV = _mm_macc_pd(vfeps,Fp,Y);
597 velec = _mm_mul_pd(qq20,VV);
598 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
599 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq20,FF),_mm_mul_pd(vftabscale,rinv20)));
601 /* Update potential sum for this i atom from the interaction with this j atom. */
602 velecsum = _mm_add_pd(velecsum,velec);
606 /* Update vectorial force */
607 fix2 = _mm_macc_pd(dx20,fscal,fix2);
608 fiy2 = _mm_macc_pd(dy20,fscal,fiy2);
609 fiz2 = _mm_macc_pd(dz20,fscal,fiz2);
611 fjx0 = _mm_macc_pd(dx20,fscal,fjx0);
612 fjy0 = _mm_macc_pd(dy20,fscal,fjy0);
613 fjz0 = _mm_macc_pd(dz20,fscal,fjz0);
615 /**************************
616 * CALCULATE INTERACTIONS *
617 **************************/
619 r21 = _mm_mul_pd(rsq21,rinv21);
621 /* Calculate table index by multiplying r with table scale and truncate to integer */
622 rt = _mm_mul_pd(r21,vftabscale);
623 vfitab = _mm_cvttpd_epi32(rt);
625 vfeps = _mm_frcz_pd(rt);
627 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
629 twovfeps = _mm_add_pd(vfeps,vfeps);
630 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
632 /* CUBIC SPLINE TABLE ELECTROSTATICS */
633 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
634 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
635 GMX_MM_TRANSPOSE2_PD(Y,F);
636 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
637 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
638 GMX_MM_TRANSPOSE2_PD(G,H);
639 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
640 VV = _mm_macc_pd(vfeps,Fp,Y);
641 velec = _mm_mul_pd(qq21,VV);
642 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
643 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
645 /* Update potential sum for this i atom from the interaction with this j atom. */
646 velecsum = _mm_add_pd(velecsum,velec);
650 /* Update vectorial force */
651 fix2 = _mm_macc_pd(dx21,fscal,fix2);
652 fiy2 = _mm_macc_pd(dy21,fscal,fiy2);
653 fiz2 = _mm_macc_pd(dz21,fscal,fiz2);
655 fjx1 = _mm_macc_pd(dx21,fscal,fjx1);
656 fjy1 = _mm_macc_pd(dy21,fscal,fjy1);
657 fjz1 = _mm_macc_pd(dz21,fscal,fjz1);
659 /**************************
660 * CALCULATE INTERACTIONS *
661 **************************/
663 r22 = _mm_mul_pd(rsq22,rinv22);
665 /* Calculate table index by multiplying r with table scale and truncate to integer */
666 rt = _mm_mul_pd(r22,vftabscale);
667 vfitab = _mm_cvttpd_epi32(rt);
669 vfeps = _mm_frcz_pd(rt);
671 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
673 twovfeps = _mm_add_pd(vfeps,vfeps);
674 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
676 /* CUBIC SPLINE TABLE ELECTROSTATICS */
677 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
678 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
679 GMX_MM_TRANSPOSE2_PD(Y,F);
680 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
681 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
682 GMX_MM_TRANSPOSE2_PD(G,H);
683 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
684 VV = _mm_macc_pd(vfeps,Fp,Y);
685 velec = _mm_mul_pd(qq22,VV);
686 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
687 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
689 /* Update potential sum for this i atom from the interaction with this j atom. */
690 velecsum = _mm_add_pd(velecsum,velec);
694 /* Update vectorial force */
695 fix2 = _mm_macc_pd(dx22,fscal,fix2);
696 fiy2 = _mm_macc_pd(dy22,fscal,fiy2);
697 fiz2 = _mm_macc_pd(dz22,fscal,fiz2);
699 fjx2 = _mm_macc_pd(dx22,fscal,fjx2);
700 fjy2 = _mm_macc_pd(dy22,fscal,fjy2);
701 fjz2 = _mm_macc_pd(dz22,fscal,fjz2);
703 gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
705 /* Inner loop uses 444 flops */
712 j_coord_offsetA = DIM*jnrA;
714 /* load j atom coordinates */
715 gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
716 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
718 /* Calculate displacement vector */
719 dx00 = _mm_sub_pd(ix0,jx0);
720 dy00 = _mm_sub_pd(iy0,jy0);
721 dz00 = _mm_sub_pd(iz0,jz0);
722 dx01 = _mm_sub_pd(ix0,jx1);
723 dy01 = _mm_sub_pd(iy0,jy1);
724 dz01 = _mm_sub_pd(iz0,jz1);
725 dx02 = _mm_sub_pd(ix0,jx2);
726 dy02 = _mm_sub_pd(iy0,jy2);
727 dz02 = _mm_sub_pd(iz0,jz2);
728 dx10 = _mm_sub_pd(ix1,jx0);
729 dy10 = _mm_sub_pd(iy1,jy0);
730 dz10 = _mm_sub_pd(iz1,jz0);
731 dx11 = _mm_sub_pd(ix1,jx1);
732 dy11 = _mm_sub_pd(iy1,jy1);
733 dz11 = _mm_sub_pd(iz1,jz1);
734 dx12 = _mm_sub_pd(ix1,jx2);
735 dy12 = _mm_sub_pd(iy1,jy2);
736 dz12 = _mm_sub_pd(iz1,jz2);
737 dx20 = _mm_sub_pd(ix2,jx0);
738 dy20 = _mm_sub_pd(iy2,jy0);
739 dz20 = _mm_sub_pd(iz2,jz0);
740 dx21 = _mm_sub_pd(ix2,jx1);
741 dy21 = _mm_sub_pd(iy2,jy1);
742 dz21 = _mm_sub_pd(iz2,jz1);
743 dx22 = _mm_sub_pd(ix2,jx2);
744 dy22 = _mm_sub_pd(iy2,jy2);
745 dz22 = _mm_sub_pd(iz2,jz2);
747 /* Calculate squared distance and things based on it */
748 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
749 rsq01 = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
750 rsq02 = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
751 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
752 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
753 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
754 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
755 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
756 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
758 rinv00 = gmx_mm_invsqrt_pd(rsq00);
759 rinv01 = gmx_mm_invsqrt_pd(rsq01);
760 rinv02 = gmx_mm_invsqrt_pd(rsq02);
761 rinv10 = gmx_mm_invsqrt_pd(rsq10);
762 rinv11 = gmx_mm_invsqrt_pd(rsq11);
763 rinv12 = gmx_mm_invsqrt_pd(rsq12);
764 rinv20 = gmx_mm_invsqrt_pd(rsq20);
765 rinv21 = gmx_mm_invsqrt_pd(rsq21);
766 rinv22 = gmx_mm_invsqrt_pd(rsq22);
768 fjx0 = _mm_setzero_pd();
769 fjy0 = _mm_setzero_pd();
770 fjz0 = _mm_setzero_pd();
771 fjx1 = _mm_setzero_pd();
772 fjy1 = _mm_setzero_pd();
773 fjz1 = _mm_setzero_pd();
774 fjx2 = _mm_setzero_pd();
775 fjy2 = _mm_setzero_pd();
776 fjz2 = _mm_setzero_pd();
778 /**************************
779 * CALCULATE INTERACTIONS *
780 **************************/
782 r00 = _mm_mul_pd(rsq00,rinv00);
784 /* Calculate table index by multiplying r with table scale and truncate to integer */
785 rt = _mm_mul_pd(r00,vftabscale);
786 vfitab = _mm_cvttpd_epi32(rt);
788 vfeps = _mm_frcz_pd(rt);
790 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
792 twovfeps = _mm_add_pd(vfeps,vfeps);
793 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
795 /* CUBIC SPLINE TABLE ELECTROSTATICS */
796 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
797 F = _mm_setzero_pd();
798 GMX_MM_TRANSPOSE2_PD(Y,F);
799 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
800 H = _mm_setzero_pd();
801 GMX_MM_TRANSPOSE2_PD(G,H);
802 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
803 VV = _mm_macc_pd(vfeps,Fp,Y);
804 velec = _mm_mul_pd(qq00,VV);
805 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
806 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq00,FF),_mm_mul_pd(vftabscale,rinv00)));
808 /* CUBIC SPLINE TABLE DISPERSION */
809 vfitab = _mm_add_epi32(vfitab,ifour);
810 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
811 F = _mm_setzero_pd();
812 GMX_MM_TRANSPOSE2_PD(Y,F);
813 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
814 H = _mm_setzero_pd();
815 GMX_MM_TRANSPOSE2_PD(G,H);
816 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(H,vfeps,G),F);
817 VV = _mm_macc_pd(vfeps,Fp,Y);
818 vvdw6 = _mm_mul_pd(c6_00,VV);
819 FF = _mm_macc_pd(vfeps,_mm_macc_pd(twovfeps,H,G),Fp);
820 fvdw6 = _mm_mul_pd(c6_00,FF);
822 /* CUBIC SPLINE TABLE REPULSION */
823 vfitab = _mm_add_epi32(vfitab,ifour);
824 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
825 F = _mm_setzero_pd();
826 GMX_MM_TRANSPOSE2_PD(Y,F);
827 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
828 H = _mm_setzero_pd();
829 GMX_MM_TRANSPOSE2_PD(G,H);
830 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(H,vfeps,G),F);
831 VV = _mm_macc_pd(vfeps,Fp,Y);
832 vvdw12 = _mm_mul_pd(c12_00,VV);
833 FF = _mm_macc_pd(vfeps,_mm_macc_pd(twovfeps,H,G),Fp);
834 fvdw12 = _mm_mul_pd(c12_00,FF);
835 vvdw = _mm_add_pd(vvdw12,vvdw6);
836 fvdw = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
838 /* Update potential sum for this i atom from the interaction with this j atom. */
839 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
840 velecsum = _mm_add_pd(velecsum,velec);
841 vvdw = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
842 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
844 fscal = _mm_add_pd(felec,fvdw);
846 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
848 /* Update vectorial force */
849 fix0 = _mm_macc_pd(dx00,fscal,fix0);
850 fiy0 = _mm_macc_pd(dy00,fscal,fiy0);
851 fiz0 = _mm_macc_pd(dz00,fscal,fiz0);
853 fjx0 = _mm_macc_pd(dx00,fscal,fjx0);
854 fjy0 = _mm_macc_pd(dy00,fscal,fjy0);
855 fjz0 = _mm_macc_pd(dz00,fscal,fjz0);
857 /**************************
858 * CALCULATE INTERACTIONS *
859 **************************/
861 r01 = _mm_mul_pd(rsq01,rinv01);
863 /* Calculate table index by multiplying r with table scale and truncate to integer */
864 rt = _mm_mul_pd(r01,vftabscale);
865 vfitab = _mm_cvttpd_epi32(rt);
867 vfeps = _mm_frcz_pd(rt);
869 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
871 twovfeps = _mm_add_pd(vfeps,vfeps);
872 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
874 /* CUBIC SPLINE TABLE ELECTROSTATICS */
875 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
876 F = _mm_setzero_pd();
877 GMX_MM_TRANSPOSE2_PD(Y,F);
878 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
879 H = _mm_setzero_pd();
880 GMX_MM_TRANSPOSE2_PD(G,H);
881 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
882 VV = _mm_macc_pd(vfeps,Fp,Y);
883 velec = _mm_mul_pd(qq01,VV);
884 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
885 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq01,FF),_mm_mul_pd(vftabscale,rinv01)));
887 /* Update potential sum for this i atom from the interaction with this j atom. */
888 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
889 velecsum = _mm_add_pd(velecsum,velec);
893 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
895 /* Update vectorial force */
896 fix0 = _mm_macc_pd(dx01,fscal,fix0);
897 fiy0 = _mm_macc_pd(dy01,fscal,fiy0);
898 fiz0 = _mm_macc_pd(dz01,fscal,fiz0);
900 fjx1 = _mm_macc_pd(dx01,fscal,fjx1);
901 fjy1 = _mm_macc_pd(dy01,fscal,fjy1);
902 fjz1 = _mm_macc_pd(dz01,fscal,fjz1);
904 /**************************
905 * CALCULATE INTERACTIONS *
906 **************************/
908 r02 = _mm_mul_pd(rsq02,rinv02);
910 /* Calculate table index by multiplying r with table scale and truncate to integer */
911 rt = _mm_mul_pd(r02,vftabscale);
912 vfitab = _mm_cvttpd_epi32(rt);
914 vfeps = _mm_frcz_pd(rt);
916 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
918 twovfeps = _mm_add_pd(vfeps,vfeps);
919 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
921 /* CUBIC SPLINE TABLE ELECTROSTATICS */
922 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
923 F = _mm_setzero_pd();
924 GMX_MM_TRANSPOSE2_PD(Y,F);
925 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
926 H = _mm_setzero_pd();
927 GMX_MM_TRANSPOSE2_PD(G,H);
928 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
929 VV = _mm_macc_pd(vfeps,Fp,Y);
930 velec = _mm_mul_pd(qq02,VV);
931 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
932 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq02,FF),_mm_mul_pd(vftabscale,rinv02)));
934 /* Update potential sum for this i atom from the interaction with this j atom. */
935 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
936 velecsum = _mm_add_pd(velecsum,velec);
940 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
942 /* Update vectorial force */
943 fix0 = _mm_macc_pd(dx02,fscal,fix0);
944 fiy0 = _mm_macc_pd(dy02,fscal,fiy0);
945 fiz0 = _mm_macc_pd(dz02,fscal,fiz0);
947 fjx2 = _mm_macc_pd(dx02,fscal,fjx2);
948 fjy2 = _mm_macc_pd(dy02,fscal,fjy2);
949 fjz2 = _mm_macc_pd(dz02,fscal,fjz2);
951 /**************************
952 * CALCULATE INTERACTIONS *
953 **************************/
955 r10 = _mm_mul_pd(rsq10,rinv10);
957 /* Calculate table index by multiplying r with table scale and truncate to integer */
958 rt = _mm_mul_pd(r10,vftabscale);
959 vfitab = _mm_cvttpd_epi32(rt);
961 vfeps = _mm_frcz_pd(rt);
963 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
965 twovfeps = _mm_add_pd(vfeps,vfeps);
966 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
968 /* CUBIC SPLINE TABLE ELECTROSTATICS */
969 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
970 F = _mm_setzero_pd();
971 GMX_MM_TRANSPOSE2_PD(Y,F);
972 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
973 H = _mm_setzero_pd();
974 GMX_MM_TRANSPOSE2_PD(G,H);
975 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
976 VV = _mm_macc_pd(vfeps,Fp,Y);
977 velec = _mm_mul_pd(qq10,VV);
978 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
979 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq10,FF),_mm_mul_pd(vftabscale,rinv10)));
981 /* Update potential sum for this i atom from the interaction with this j atom. */
982 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
983 velecsum = _mm_add_pd(velecsum,velec);
987 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
989 /* Update vectorial force */
990 fix1 = _mm_macc_pd(dx10,fscal,fix1);
991 fiy1 = _mm_macc_pd(dy10,fscal,fiy1);
992 fiz1 = _mm_macc_pd(dz10,fscal,fiz1);
994 fjx0 = _mm_macc_pd(dx10,fscal,fjx0);
995 fjy0 = _mm_macc_pd(dy10,fscal,fjy0);
996 fjz0 = _mm_macc_pd(dz10,fscal,fjz0);
998 /**************************
999 * CALCULATE INTERACTIONS *
1000 **************************/
1002 r11 = _mm_mul_pd(rsq11,rinv11);
1004 /* Calculate table index by multiplying r with table scale and truncate to integer */
1005 rt = _mm_mul_pd(r11,vftabscale);
1006 vfitab = _mm_cvttpd_epi32(rt);
1008 vfeps = _mm_frcz_pd(rt);
1010 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1012 twovfeps = _mm_add_pd(vfeps,vfeps);
1013 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1015 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1016 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1017 F = _mm_setzero_pd();
1018 GMX_MM_TRANSPOSE2_PD(Y,F);
1019 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1020 H = _mm_setzero_pd();
1021 GMX_MM_TRANSPOSE2_PD(G,H);
1022 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1023 VV = _mm_macc_pd(vfeps,Fp,Y);
1024 velec = _mm_mul_pd(qq11,VV);
1025 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1026 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
1028 /* Update potential sum for this i atom from the interaction with this j atom. */
1029 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1030 velecsum = _mm_add_pd(velecsum,velec);
1034 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1036 /* Update vectorial force */
1037 fix1 = _mm_macc_pd(dx11,fscal,fix1);
1038 fiy1 = _mm_macc_pd(dy11,fscal,fiy1);
1039 fiz1 = _mm_macc_pd(dz11,fscal,fiz1);
1041 fjx1 = _mm_macc_pd(dx11,fscal,fjx1);
1042 fjy1 = _mm_macc_pd(dy11,fscal,fjy1);
1043 fjz1 = _mm_macc_pd(dz11,fscal,fjz1);
1045 /**************************
1046 * CALCULATE INTERACTIONS *
1047 **************************/
1049 r12 = _mm_mul_pd(rsq12,rinv12);
1051 /* Calculate table index by multiplying r with table scale and truncate to integer */
1052 rt = _mm_mul_pd(r12,vftabscale);
1053 vfitab = _mm_cvttpd_epi32(rt);
1055 vfeps = _mm_frcz_pd(rt);
1057 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1059 twovfeps = _mm_add_pd(vfeps,vfeps);
1060 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1062 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1063 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1064 F = _mm_setzero_pd();
1065 GMX_MM_TRANSPOSE2_PD(Y,F);
1066 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1067 H = _mm_setzero_pd();
1068 GMX_MM_TRANSPOSE2_PD(G,H);
1069 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1070 VV = _mm_macc_pd(vfeps,Fp,Y);
1071 velec = _mm_mul_pd(qq12,VV);
1072 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1073 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
1075 /* Update potential sum for this i atom from the interaction with this j atom. */
1076 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1077 velecsum = _mm_add_pd(velecsum,velec);
1081 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1083 /* Update vectorial force */
1084 fix1 = _mm_macc_pd(dx12,fscal,fix1);
1085 fiy1 = _mm_macc_pd(dy12,fscal,fiy1);
1086 fiz1 = _mm_macc_pd(dz12,fscal,fiz1);
1088 fjx2 = _mm_macc_pd(dx12,fscal,fjx2);
1089 fjy2 = _mm_macc_pd(dy12,fscal,fjy2);
1090 fjz2 = _mm_macc_pd(dz12,fscal,fjz2);
1092 /**************************
1093 * CALCULATE INTERACTIONS *
1094 **************************/
1096 r20 = _mm_mul_pd(rsq20,rinv20);
1098 /* Calculate table index by multiplying r with table scale and truncate to integer */
1099 rt = _mm_mul_pd(r20,vftabscale);
1100 vfitab = _mm_cvttpd_epi32(rt);
1102 vfeps = _mm_frcz_pd(rt);
1104 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1106 twovfeps = _mm_add_pd(vfeps,vfeps);
1107 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1109 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1110 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1111 F = _mm_setzero_pd();
1112 GMX_MM_TRANSPOSE2_PD(Y,F);
1113 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1114 H = _mm_setzero_pd();
1115 GMX_MM_TRANSPOSE2_PD(G,H);
1116 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1117 VV = _mm_macc_pd(vfeps,Fp,Y);
1118 velec = _mm_mul_pd(qq20,VV);
1119 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1120 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq20,FF),_mm_mul_pd(vftabscale,rinv20)));
1122 /* Update potential sum for this i atom from the interaction with this j atom. */
1123 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1124 velecsum = _mm_add_pd(velecsum,velec);
1128 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1130 /* Update vectorial force */
1131 fix2 = _mm_macc_pd(dx20,fscal,fix2);
1132 fiy2 = _mm_macc_pd(dy20,fscal,fiy2);
1133 fiz2 = _mm_macc_pd(dz20,fscal,fiz2);
1135 fjx0 = _mm_macc_pd(dx20,fscal,fjx0);
1136 fjy0 = _mm_macc_pd(dy20,fscal,fjy0);
1137 fjz0 = _mm_macc_pd(dz20,fscal,fjz0);
1139 /**************************
1140 * CALCULATE INTERACTIONS *
1141 **************************/
1143 r21 = _mm_mul_pd(rsq21,rinv21);
1145 /* Calculate table index by multiplying r with table scale and truncate to integer */
1146 rt = _mm_mul_pd(r21,vftabscale);
1147 vfitab = _mm_cvttpd_epi32(rt);
1149 vfeps = _mm_frcz_pd(rt);
1151 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1153 twovfeps = _mm_add_pd(vfeps,vfeps);
1154 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1156 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1157 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1158 F = _mm_setzero_pd();
1159 GMX_MM_TRANSPOSE2_PD(Y,F);
1160 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1161 H = _mm_setzero_pd();
1162 GMX_MM_TRANSPOSE2_PD(G,H);
1163 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1164 VV = _mm_macc_pd(vfeps,Fp,Y);
1165 velec = _mm_mul_pd(qq21,VV);
1166 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1167 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
1169 /* Update potential sum for this i atom from the interaction with this j atom. */
1170 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1171 velecsum = _mm_add_pd(velecsum,velec);
1175 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1177 /* Update vectorial force */
1178 fix2 = _mm_macc_pd(dx21,fscal,fix2);
1179 fiy2 = _mm_macc_pd(dy21,fscal,fiy2);
1180 fiz2 = _mm_macc_pd(dz21,fscal,fiz2);
1182 fjx1 = _mm_macc_pd(dx21,fscal,fjx1);
1183 fjy1 = _mm_macc_pd(dy21,fscal,fjy1);
1184 fjz1 = _mm_macc_pd(dz21,fscal,fjz1);
1186 /**************************
1187 * CALCULATE INTERACTIONS *
1188 **************************/
1190 r22 = _mm_mul_pd(rsq22,rinv22);
1192 /* Calculate table index by multiplying r with table scale and truncate to integer */
1193 rt = _mm_mul_pd(r22,vftabscale);
1194 vfitab = _mm_cvttpd_epi32(rt);
1196 vfeps = _mm_frcz_pd(rt);
1198 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1200 twovfeps = _mm_add_pd(vfeps,vfeps);
1201 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1203 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1204 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1205 F = _mm_setzero_pd();
1206 GMX_MM_TRANSPOSE2_PD(Y,F);
1207 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1208 H = _mm_setzero_pd();
1209 GMX_MM_TRANSPOSE2_PD(G,H);
1210 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1211 VV = _mm_macc_pd(vfeps,Fp,Y);
1212 velec = _mm_mul_pd(qq22,VV);
1213 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1214 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
1216 /* Update potential sum for this i atom from the interaction with this j atom. */
1217 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1218 velecsum = _mm_add_pd(velecsum,velec);
1222 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1224 /* Update vectorial force */
1225 fix2 = _mm_macc_pd(dx22,fscal,fix2);
1226 fiy2 = _mm_macc_pd(dy22,fscal,fiy2);
1227 fiz2 = _mm_macc_pd(dz22,fscal,fiz2);
1229 fjx2 = _mm_macc_pd(dx22,fscal,fjx2);
1230 fjy2 = _mm_macc_pd(dy22,fscal,fjy2);
1231 fjz2 = _mm_macc_pd(dz22,fscal,fjz2);
1233 gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1235 /* Inner loop uses 444 flops */
1238 /* End of innermost loop */
1240 gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1241 f+i_coord_offset,fshift+i_shift_offset);
1244 /* Update potential energies */
1245 gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1246 gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1248 /* Increment number of inner iterations */
1249 inneriter += j_index_end - j_index_start;
1251 /* Outer loop uses 20 flops */
1254 /* Increment number of outer iterations */
1257 /* Update outer/inner flops */
1259 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_VF,outeriter*20 + inneriter*444);
1262 * Gromacs nonbonded kernel: nb_kernel_ElecCSTab_VdwCSTab_GeomW3W3_F_avx_128_fma_double
1263 * Electrostatics interaction: CubicSplineTable
1264 * VdW interaction: CubicSplineTable
1265 * Geometry: Water3-Water3
1266 * Calculate force/pot: Force
1269 nb_kernel_ElecCSTab_VdwCSTab_GeomW3W3_F_avx_128_fma_double
1270 (t_nblist * gmx_restrict nlist,
1271 rvec * gmx_restrict xx,
1272 rvec * gmx_restrict ff,
1273 t_forcerec * gmx_restrict fr,
1274 t_mdatoms * gmx_restrict mdatoms,
1275 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1276 t_nrnb * gmx_restrict nrnb)
1278 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1279 * just 0 for non-waters.
1280 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
1281 * jnr indices corresponding to data put in the four positions in the SIMD register.
1283 int i_shift_offset,i_coord_offset,outeriter,inneriter;
1284 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1286 int j_coord_offsetA,j_coord_offsetB;
1287 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
1288 real rcutoff_scalar;
1289 real *shiftvec,*fshift,*x,*f;
1290 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1292 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1294 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1296 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1297 int vdwjidx0A,vdwjidx0B;
1298 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1299 int vdwjidx1A,vdwjidx1B;
1300 __m128d jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1301 int vdwjidx2A,vdwjidx2B;
1302 __m128d jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1303 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1304 __m128d dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1305 __m128d dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1306 __m128d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1307 __m128d dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1308 __m128d dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1309 __m128d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1310 __m128d dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1311 __m128d dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1312 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
1315 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1318 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
1319 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
1321 __m128i ifour = _mm_set1_epi32(4);
1322 __m128d rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
1324 __m128d dummy_mask,cutoff_mask;
1325 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
1326 __m128d one = _mm_set1_pd(1.0);
1327 __m128d two = _mm_set1_pd(2.0);
1333 jindex = nlist->jindex;
1335 shiftidx = nlist->shift;
1337 shiftvec = fr->shift_vec[0];
1338 fshift = fr->fshift[0];
1339 facel = _mm_set1_pd(fr->epsfac);
1340 charge = mdatoms->chargeA;
1341 nvdwtype = fr->ntype;
1342 vdwparam = fr->nbfp;
1343 vdwtype = mdatoms->typeA;
1345 vftab = kernel_data->table_elec_vdw->data;
1346 vftabscale = _mm_set1_pd(kernel_data->table_elec_vdw->scale);
1348 /* Setup water-specific parameters */
1349 inr = nlist->iinr[0];
1350 iq0 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
1351 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
1352 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
1353 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
1355 jq0 = _mm_set1_pd(charge[inr+0]);
1356 jq1 = _mm_set1_pd(charge[inr+1]);
1357 jq2 = _mm_set1_pd(charge[inr+2]);
1358 vdwjidx0A = 2*vdwtype[inr+0];
1359 qq00 = _mm_mul_pd(iq0,jq0);
1360 c6_00 = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
1361 c12_00 = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
1362 qq01 = _mm_mul_pd(iq0,jq1);
1363 qq02 = _mm_mul_pd(iq0,jq2);
1364 qq10 = _mm_mul_pd(iq1,jq0);
1365 qq11 = _mm_mul_pd(iq1,jq1);
1366 qq12 = _mm_mul_pd(iq1,jq2);
1367 qq20 = _mm_mul_pd(iq2,jq0);
1368 qq21 = _mm_mul_pd(iq2,jq1);
1369 qq22 = _mm_mul_pd(iq2,jq2);
1371 /* Avoid stupid compiler warnings */
1373 j_coord_offsetA = 0;
1374 j_coord_offsetB = 0;
1379 /* Start outer loop over neighborlists */
1380 for(iidx=0; iidx<nri; iidx++)
1382 /* Load shift vector for this list */
1383 i_shift_offset = DIM*shiftidx[iidx];
1385 /* Load limits for loop over neighbors */
1386 j_index_start = jindex[iidx];
1387 j_index_end = jindex[iidx+1];
1389 /* Get outer coordinate index */
1391 i_coord_offset = DIM*inr;
1393 /* Load i particle coords and add shift vector */
1394 gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1395 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1397 fix0 = _mm_setzero_pd();
1398 fiy0 = _mm_setzero_pd();
1399 fiz0 = _mm_setzero_pd();
1400 fix1 = _mm_setzero_pd();
1401 fiy1 = _mm_setzero_pd();
1402 fiz1 = _mm_setzero_pd();
1403 fix2 = _mm_setzero_pd();
1404 fiy2 = _mm_setzero_pd();
1405 fiz2 = _mm_setzero_pd();
1407 /* Start inner kernel loop */
1408 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
1411 /* Get j neighbor index, and coordinate index */
1413 jnrB = jjnr[jidx+1];
1414 j_coord_offsetA = DIM*jnrA;
1415 j_coord_offsetB = DIM*jnrB;
1417 /* load j atom coordinates */
1418 gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1419 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1421 /* Calculate displacement vector */
1422 dx00 = _mm_sub_pd(ix0,jx0);
1423 dy00 = _mm_sub_pd(iy0,jy0);
1424 dz00 = _mm_sub_pd(iz0,jz0);
1425 dx01 = _mm_sub_pd(ix0,jx1);
1426 dy01 = _mm_sub_pd(iy0,jy1);
1427 dz01 = _mm_sub_pd(iz0,jz1);
1428 dx02 = _mm_sub_pd(ix0,jx2);
1429 dy02 = _mm_sub_pd(iy0,jy2);
1430 dz02 = _mm_sub_pd(iz0,jz2);
1431 dx10 = _mm_sub_pd(ix1,jx0);
1432 dy10 = _mm_sub_pd(iy1,jy0);
1433 dz10 = _mm_sub_pd(iz1,jz0);
1434 dx11 = _mm_sub_pd(ix1,jx1);
1435 dy11 = _mm_sub_pd(iy1,jy1);
1436 dz11 = _mm_sub_pd(iz1,jz1);
1437 dx12 = _mm_sub_pd(ix1,jx2);
1438 dy12 = _mm_sub_pd(iy1,jy2);
1439 dz12 = _mm_sub_pd(iz1,jz2);
1440 dx20 = _mm_sub_pd(ix2,jx0);
1441 dy20 = _mm_sub_pd(iy2,jy0);
1442 dz20 = _mm_sub_pd(iz2,jz0);
1443 dx21 = _mm_sub_pd(ix2,jx1);
1444 dy21 = _mm_sub_pd(iy2,jy1);
1445 dz21 = _mm_sub_pd(iz2,jz1);
1446 dx22 = _mm_sub_pd(ix2,jx2);
1447 dy22 = _mm_sub_pd(iy2,jy2);
1448 dz22 = _mm_sub_pd(iz2,jz2);
1450 /* Calculate squared distance and things based on it */
1451 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1452 rsq01 = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
1453 rsq02 = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
1454 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
1455 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1456 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1457 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
1458 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1459 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1461 rinv00 = gmx_mm_invsqrt_pd(rsq00);
1462 rinv01 = gmx_mm_invsqrt_pd(rsq01);
1463 rinv02 = gmx_mm_invsqrt_pd(rsq02);
1464 rinv10 = gmx_mm_invsqrt_pd(rsq10);
1465 rinv11 = gmx_mm_invsqrt_pd(rsq11);
1466 rinv12 = gmx_mm_invsqrt_pd(rsq12);
1467 rinv20 = gmx_mm_invsqrt_pd(rsq20);
1468 rinv21 = gmx_mm_invsqrt_pd(rsq21);
1469 rinv22 = gmx_mm_invsqrt_pd(rsq22);
1471 fjx0 = _mm_setzero_pd();
1472 fjy0 = _mm_setzero_pd();
1473 fjz0 = _mm_setzero_pd();
1474 fjx1 = _mm_setzero_pd();
1475 fjy1 = _mm_setzero_pd();
1476 fjz1 = _mm_setzero_pd();
1477 fjx2 = _mm_setzero_pd();
1478 fjy2 = _mm_setzero_pd();
1479 fjz2 = _mm_setzero_pd();
1481 /**************************
1482 * CALCULATE INTERACTIONS *
1483 **************************/
1485 r00 = _mm_mul_pd(rsq00,rinv00);
1487 /* Calculate table index by multiplying r with table scale and truncate to integer */
1488 rt = _mm_mul_pd(r00,vftabscale);
1489 vfitab = _mm_cvttpd_epi32(rt);
1491 vfeps = _mm_frcz_pd(rt);
1493 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1495 twovfeps = _mm_add_pd(vfeps,vfeps);
1496 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1498 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1499 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1500 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1501 GMX_MM_TRANSPOSE2_PD(Y,F);
1502 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1503 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1504 GMX_MM_TRANSPOSE2_PD(G,H);
1505 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1506 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1507 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq00,FF),_mm_mul_pd(vftabscale,rinv00)));
1509 /* CUBIC SPLINE TABLE DISPERSION */
1510 vfitab = _mm_add_epi32(vfitab,ifour);
1511 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1512 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1513 GMX_MM_TRANSPOSE2_PD(Y,F);
1514 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1515 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1516 GMX_MM_TRANSPOSE2_PD(G,H);
1517 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(H,vfeps,G),F);
1518 FF = _mm_macc_pd(vfeps,_mm_macc_pd(twovfeps,H,G),Fp);
1519 fvdw6 = _mm_mul_pd(c6_00,FF);
1521 /* CUBIC SPLINE TABLE REPULSION */
1522 vfitab = _mm_add_epi32(vfitab,ifour);
1523 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1524 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1525 GMX_MM_TRANSPOSE2_PD(Y,F);
1526 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1527 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1528 GMX_MM_TRANSPOSE2_PD(G,H);
1529 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(H,vfeps,G),F);
1530 FF = _mm_macc_pd(vfeps,_mm_macc_pd(twovfeps,H,G),Fp);
1531 fvdw12 = _mm_mul_pd(c12_00,FF);
1532 fvdw = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
1534 fscal = _mm_add_pd(felec,fvdw);
1536 /* Update vectorial force */
1537 fix0 = _mm_macc_pd(dx00,fscal,fix0);
1538 fiy0 = _mm_macc_pd(dy00,fscal,fiy0);
1539 fiz0 = _mm_macc_pd(dz00,fscal,fiz0);
1541 fjx0 = _mm_macc_pd(dx00,fscal,fjx0);
1542 fjy0 = _mm_macc_pd(dy00,fscal,fjy0);
1543 fjz0 = _mm_macc_pd(dz00,fscal,fjz0);
1545 /**************************
1546 * CALCULATE INTERACTIONS *
1547 **************************/
1549 r01 = _mm_mul_pd(rsq01,rinv01);
1551 /* Calculate table index by multiplying r with table scale and truncate to integer */
1552 rt = _mm_mul_pd(r01,vftabscale);
1553 vfitab = _mm_cvttpd_epi32(rt);
1555 vfeps = _mm_frcz_pd(rt);
1557 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1559 twovfeps = _mm_add_pd(vfeps,vfeps);
1560 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1562 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1563 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1564 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1565 GMX_MM_TRANSPOSE2_PD(Y,F);
1566 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1567 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1568 GMX_MM_TRANSPOSE2_PD(G,H);
1569 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1570 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1571 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq01,FF),_mm_mul_pd(vftabscale,rinv01)));
1575 /* Update vectorial force */
1576 fix0 = _mm_macc_pd(dx01,fscal,fix0);
1577 fiy0 = _mm_macc_pd(dy01,fscal,fiy0);
1578 fiz0 = _mm_macc_pd(dz01,fscal,fiz0);
1580 fjx1 = _mm_macc_pd(dx01,fscal,fjx1);
1581 fjy1 = _mm_macc_pd(dy01,fscal,fjy1);
1582 fjz1 = _mm_macc_pd(dz01,fscal,fjz1);
1584 /**************************
1585 * CALCULATE INTERACTIONS *
1586 **************************/
1588 r02 = _mm_mul_pd(rsq02,rinv02);
1590 /* Calculate table index by multiplying r with table scale and truncate to integer */
1591 rt = _mm_mul_pd(r02,vftabscale);
1592 vfitab = _mm_cvttpd_epi32(rt);
1594 vfeps = _mm_frcz_pd(rt);
1596 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1598 twovfeps = _mm_add_pd(vfeps,vfeps);
1599 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1601 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1602 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1603 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1604 GMX_MM_TRANSPOSE2_PD(Y,F);
1605 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1606 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1607 GMX_MM_TRANSPOSE2_PD(G,H);
1608 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1609 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1610 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq02,FF),_mm_mul_pd(vftabscale,rinv02)));
1614 /* Update vectorial force */
1615 fix0 = _mm_macc_pd(dx02,fscal,fix0);
1616 fiy0 = _mm_macc_pd(dy02,fscal,fiy0);
1617 fiz0 = _mm_macc_pd(dz02,fscal,fiz0);
1619 fjx2 = _mm_macc_pd(dx02,fscal,fjx2);
1620 fjy2 = _mm_macc_pd(dy02,fscal,fjy2);
1621 fjz2 = _mm_macc_pd(dz02,fscal,fjz2);
1623 /**************************
1624 * CALCULATE INTERACTIONS *
1625 **************************/
1627 r10 = _mm_mul_pd(rsq10,rinv10);
1629 /* Calculate table index by multiplying r with table scale and truncate to integer */
1630 rt = _mm_mul_pd(r10,vftabscale);
1631 vfitab = _mm_cvttpd_epi32(rt);
1633 vfeps = _mm_frcz_pd(rt);
1635 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1637 twovfeps = _mm_add_pd(vfeps,vfeps);
1638 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1640 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1641 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1642 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1643 GMX_MM_TRANSPOSE2_PD(Y,F);
1644 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1645 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1646 GMX_MM_TRANSPOSE2_PD(G,H);
1647 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1648 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1649 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq10,FF),_mm_mul_pd(vftabscale,rinv10)));
1653 /* Update vectorial force */
1654 fix1 = _mm_macc_pd(dx10,fscal,fix1);
1655 fiy1 = _mm_macc_pd(dy10,fscal,fiy1);
1656 fiz1 = _mm_macc_pd(dz10,fscal,fiz1);
1658 fjx0 = _mm_macc_pd(dx10,fscal,fjx0);
1659 fjy0 = _mm_macc_pd(dy10,fscal,fjy0);
1660 fjz0 = _mm_macc_pd(dz10,fscal,fjz0);
1662 /**************************
1663 * CALCULATE INTERACTIONS *
1664 **************************/
1666 r11 = _mm_mul_pd(rsq11,rinv11);
1668 /* Calculate table index by multiplying r with table scale and truncate to integer */
1669 rt = _mm_mul_pd(r11,vftabscale);
1670 vfitab = _mm_cvttpd_epi32(rt);
1672 vfeps = _mm_frcz_pd(rt);
1674 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1676 twovfeps = _mm_add_pd(vfeps,vfeps);
1677 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1679 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1680 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1681 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1682 GMX_MM_TRANSPOSE2_PD(Y,F);
1683 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1684 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1685 GMX_MM_TRANSPOSE2_PD(G,H);
1686 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1687 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1688 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
1692 /* Update vectorial force */
1693 fix1 = _mm_macc_pd(dx11,fscal,fix1);
1694 fiy1 = _mm_macc_pd(dy11,fscal,fiy1);
1695 fiz1 = _mm_macc_pd(dz11,fscal,fiz1);
1697 fjx1 = _mm_macc_pd(dx11,fscal,fjx1);
1698 fjy1 = _mm_macc_pd(dy11,fscal,fjy1);
1699 fjz1 = _mm_macc_pd(dz11,fscal,fjz1);
1701 /**************************
1702 * CALCULATE INTERACTIONS *
1703 **************************/
1705 r12 = _mm_mul_pd(rsq12,rinv12);
1707 /* Calculate table index by multiplying r with table scale and truncate to integer */
1708 rt = _mm_mul_pd(r12,vftabscale);
1709 vfitab = _mm_cvttpd_epi32(rt);
1711 vfeps = _mm_frcz_pd(rt);
1713 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1715 twovfeps = _mm_add_pd(vfeps,vfeps);
1716 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1718 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1719 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1720 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1721 GMX_MM_TRANSPOSE2_PD(Y,F);
1722 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1723 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1724 GMX_MM_TRANSPOSE2_PD(G,H);
1725 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1726 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1727 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
1731 /* Update vectorial force */
1732 fix1 = _mm_macc_pd(dx12,fscal,fix1);
1733 fiy1 = _mm_macc_pd(dy12,fscal,fiy1);
1734 fiz1 = _mm_macc_pd(dz12,fscal,fiz1);
1736 fjx2 = _mm_macc_pd(dx12,fscal,fjx2);
1737 fjy2 = _mm_macc_pd(dy12,fscal,fjy2);
1738 fjz2 = _mm_macc_pd(dz12,fscal,fjz2);
1740 /**************************
1741 * CALCULATE INTERACTIONS *
1742 **************************/
1744 r20 = _mm_mul_pd(rsq20,rinv20);
1746 /* Calculate table index by multiplying r with table scale and truncate to integer */
1747 rt = _mm_mul_pd(r20,vftabscale);
1748 vfitab = _mm_cvttpd_epi32(rt);
1750 vfeps = _mm_frcz_pd(rt);
1752 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1754 twovfeps = _mm_add_pd(vfeps,vfeps);
1755 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1757 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1758 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1759 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1760 GMX_MM_TRANSPOSE2_PD(Y,F);
1761 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1762 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1763 GMX_MM_TRANSPOSE2_PD(G,H);
1764 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1765 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1766 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq20,FF),_mm_mul_pd(vftabscale,rinv20)));
1770 /* Update vectorial force */
1771 fix2 = _mm_macc_pd(dx20,fscal,fix2);
1772 fiy2 = _mm_macc_pd(dy20,fscal,fiy2);
1773 fiz2 = _mm_macc_pd(dz20,fscal,fiz2);
1775 fjx0 = _mm_macc_pd(dx20,fscal,fjx0);
1776 fjy0 = _mm_macc_pd(dy20,fscal,fjy0);
1777 fjz0 = _mm_macc_pd(dz20,fscal,fjz0);
1779 /**************************
1780 * CALCULATE INTERACTIONS *
1781 **************************/
1783 r21 = _mm_mul_pd(rsq21,rinv21);
1785 /* Calculate table index by multiplying r with table scale and truncate to integer */
1786 rt = _mm_mul_pd(r21,vftabscale);
1787 vfitab = _mm_cvttpd_epi32(rt);
1789 vfeps = _mm_frcz_pd(rt);
1791 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1793 twovfeps = _mm_add_pd(vfeps,vfeps);
1794 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1796 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1797 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1798 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1799 GMX_MM_TRANSPOSE2_PD(Y,F);
1800 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1801 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1802 GMX_MM_TRANSPOSE2_PD(G,H);
1803 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1804 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1805 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
1809 /* Update vectorial force */
1810 fix2 = _mm_macc_pd(dx21,fscal,fix2);
1811 fiy2 = _mm_macc_pd(dy21,fscal,fiy2);
1812 fiz2 = _mm_macc_pd(dz21,fscal,fiz2);
1814 fjx1 = _mm_macc_pd(dx21,fscal,fjx1);
1815 fjy1 = _mm_macc_pd(dy21,fscal,fjy1);
1816 fjz1 = _mm_macc_pd(dz21,fscal,fjz1);
1818 /**************************
1819 * CALCULATE INTERACTIONS *
1820 **************************/
1822 r22 = _mm_mul_pd(rsq22,rinv22);
1824 /* Calculate table index by multiplying r with table scale and truncate to integer */
1825 rt = _mm_mul_pd(r22,vftabscale);
1826 vfitab = _mm_cvttpd_epi32(rt);
1828 vfeps = _mm_frcz_pd(rt);
1830 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1832 twovfeps = _mm_add_pd(vfeps,vfeps);
1833 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1835 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1836 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1837 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1838 GMX_MM_TRANSPOSE2_PD(Y,F);
1839 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1840 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1841 GMX_MM_TRANSPOSE2_PD(G,H);
1842 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1843 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1844 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
1848 /* Update vectorial force */
1849 fix2 = _mm_macc_pd(dx22,fscal,fix2);
1850 fiy2 = _mm_macc_pd(dy22,fscal,fiy2);
1851 fiz2 = _mm_macc_pd(dz22,fscal,fiz2);
1853 fjx2 = _mm_macc_pd(dx22,fscal,fjx2);
1854 fjy2 = _mm_macc_pd(dy22,fscal,fjy2);
1855 fjz2 = _mm_macc_pd(dz22,fscal,fjz2);
1857 gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1859 /* Inner loop uses 400 flops */
1862 if(jidx<j_index_end)
1866 j_coord_offsetA = DIM*jnrA;
1868 /* load j atom coordinates */
1869 gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
1870 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1872 /* Calculate displacement vector */
1873 dx00 = _mm_sub_pd(ix0,jx0);
1874 dy00 = _mm_sub_pd(iy0,jy0);
1875 dz00 = _mm_sub_pd(iz0,jz0);
1876 dx01 = _mm_sub_pd(ix0,jx1);
1877 dy01 = _mm_sub_pd(iy0,jy1);
1878 dz01 = _mm_sub_pd(iz0,jz1);
1879 dx02 = _mm_sub_pd(ix0,jx2);
1880 dy02 = _mm_sub_pd(iy0,jy2);
1881 dz02 = _mm_sub_pd(iz0,jz2);
1882 dx10 = _mm_sub_pd(ix1,jx0);
1883 dy10 = _mm_sub_pd(iy1,jy0);
1884 dz10 = _mm_sub_pd(iz1,jz0);
1885 dx11 = _mm_sub_pd(ix1,jx1);
1886 dy11 = _mm_sub_pd(iy1,jy1);
1887 dz11 = _mm_sub_pd(iz1,jz1);
1888 dx12 = _mm_sub_pd(ix1,jx2);
1889 dy12 = _mm_sub_pd(iy1,jy2);
1890 dz12 = _mm_sub_pd(iz1,jz2);
1891 dx20 = _mm_sub_pd(ix2,jx0);
1892 dy20 = _mm_sub_pd(iy2,jy0);
1893 dz20 = _mm_sub_pd(iz2,jz0);
1894 dx21 = _mm_sub_pd(ix2,jx1);
1895 dy21 = _mm_sub_pd(iy2,jy1);
1896 dz21 = _mm_sub_pd(iz2,jz1);
1897 dx22 = _mm_sub_pd(ix2,jx2);
1898 dy22 = _mm_sub_pd(iy2,jy2);
1899 dz22 = _mm_sub_pd(iz2,jz2);
1901 /* Calculate squared distance and things based on it */
1902 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1903 rsq01 = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
1904 rsq02 = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
1905 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
1906 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1907 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1908 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
1909 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1910 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1912 rinv00 = gmx_mm_invsqrt_pd(rsq00);
1913 rinv01 = gmx_mm_invsqrt_pd(rsq01);
1914 rinv02 = gmx_mm_invsqrt_pd(rsq02);
1915 rinv10 = gmx_mm_invsqrt_pd(rsq10);
1916 rinv11 = gmx_mm_invsqrt_pd(rsq11);
1917 rinv12 = gmx_mm_invsqrt_pd(rsq12);
1918 rinv20 = gmx_mm_invsqrt_pd(rsq20);
1919 rinv21 = gmx_mm_invsqrt_pd(rsq21);
1920 rinv22 = gmx_mm_invsqrt_pd(rsq22);
1922 fjx0 = _mm_setzero_pd();
1923 fjy0 = _mm_setzero_pd();
1924 fjz0 = _mm_setzero_pd();
1925 fjx1 = _mm_setzero_pd();
1926 fjy1 = _mm_setzero_pd();
1927 fjz1 = _mm_setzero_pd();
1928 fjx2 = _mm_setzero_pd();
1929 fjy2 = _mm_setzero_pd();
1930 fjz2 = _mm_setzero_pd();
1932 /**************************
1933 * CALCULATE INTERACTIONS *
1934 **************************/
1936 r00 = _mm_mul_pd(rsq00,rinv00);
1938 /* Calculate table index by multiplying r with table scale and truncate to integer */
1939 rt = _mm_mul_pd(r00,vftabscale);
1940 vfitab = _mm_cvttpd_epi32(rt);
1942 vfeps = _mm_frcz_pd(rt);
1944 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1946 twovfeps = _mm_add_pd(vfeps,vfeps);
1947 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1949 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1950 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1951 F = _mm_setzero_pd();
1952 GMX_MM_TRANSPOSE2_PD(Y,F);
1953 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1954 H = _mm_setzero_pd();
1955 GMX_MM_TRANSPOSE2_PD(G,H);
1956 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1957 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1958 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq00,FF),_mm_mul_pd(vftabscale,rinv00)));
1960 /* CUBIC SPLINE TABLE DISPERSION */
1961 vfitab = _mm_add_epi32(vfitab,ifour);
1962 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1963 F = _mm_setzero_pd();
1964 GMX_MM_TRANSPOSE2_PD(Y,F);
1965 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1966 H = _mm_setzero_pd();
1967 GMX_MM_TRANSPOSE2_PD(G,H);
1968 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(H,vfeps,G),F);
1969 FF = _mm_macc_pd(vfeps,_mm_macc_pd(twovfeps,H,G),Fp);
1970 fvdw6 = _mm_mul_pd(c6_00,FF);
1972 /* CUBIC SPLINE TABLE REPULSION */
1973 vfitab = _mm_add_epi32(vfitab,ifour);
1974 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1975 F = _mm_setzero_pd();
1976 GMX_MM_TRANSPOSE2_PD(Y,F);
1977 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1978 H = _mm_setzero_pd();
1979 GMX_MM_TRANSPOSE2_PD(G,H);
1980 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(H,vfeps,G),F);
1981 FF = _mm_macc_pd(vfeps,_mm_macc_pd(twovfeps,H,G),Fp);
1982 fvdw12 = _mm_mul_pd(c12_00,FF);
1983 fvdw = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
1985 fscal = _mm_add_pd(felec,fvdw);
1987 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1989 /* Update vectorial force */
1990 fix0 = _mm_macc_pd(dx00,fscal,fix0);
1991 fiy0 = _mm_macc_pd(dy00,fscal,fiy0);
1992 fiz0 = _mm_macc_pd(dz00,fscal,fiz0);
1994 fjx0 = _mm_macc_pd(dx00,fscal,fjx0);
1995 fjy0 = _mm_macc_pd(dy00,fscal,fjy0);
1996 fjz0 = _mm_macc_pd(dz00,fscal,fjz0);
1998 /**************************
1999 * CALCULATE INTERACTIONS *
2000 **************************/
2002 r01 = _mm_mul_pd(rsq01,rinv01);
2004 /* Calculate table index by multiplying r with table scale and truncate to integer */
2005 rt = _mm_mul_pd(r01,vftabscale);
2006 vfitab = _mm_cvttpd_epi32(rt);
2008 vfeps = _mm_frcz_pd(rt);
2010 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2012 twovfeps = _mm_add_pd(vfeps,vfeps);
2013 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2015 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2016 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2017 F = _mm_setzero_pd();
2018 GMX_MM_TRANSPOSE2_PD(Y,F);
2019 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
2020 H = _mm_setzero_pd();
2021 GMX_MM_TRANSPOSE2_PD(G,H);
2022 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
2023 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
2024 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq01,FF),_mm_mul_pd(vftabscale,rinv01)));
2028 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2030 /* Update vectorial force */
2031 fix0 = _mm_macc_pd(dx01,fscal,fix0);
2032 fiy0 = _mm_macc_pd(dy01,fscal,fiy0);
2033 fiz0 = _mm_macc_pd(dz01,fscal,fiz0);
2035 fjx1 = _mm_macc_pd(dx01,fscal,fjx1);
2036 fjy1 = _mm_macc_pd(dy01,fscal,fjy1);
2037 fjz1 = _mm_macc_pd(dz01,fscal,fjz1);
2039 /**************************
2040 * CALCULATE INTERACTIONS *
2041 **************************/
2043 r02 = _mm_mul_pd(rsq02,rinv02);
2045 /* Calculate table index by multiplying r with table scale and truncate to integer */
2046 rt = _mm_mul_pd(r02,vftabscale);
2047 vfitab = _mm_cvttpd_epi32(rt);
2049 vfeps = _mm_frcz_pd(rt);
2051 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2053 twovfeps = _mm_add_pd(vfeps,vfeps);
2054 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2056 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2057 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2058 F = _mm_setzero_pd();
2059 GMX_MM_TRANSPOSE2_PD(Y,F);
2060 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
2061 H = _mm_setzero_pd();
2062 GMX_MM_TRANSPOSE2_PD(G,H);
2063 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
2064 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
2065 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq02,FF),_mm_mul_pd(vftabscale,rinv02)));
2069 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2071 /* Update vectorial force */
2072 fix0 = _mm_macc_pd(dx02,fscal,fix0);
2073 fiy0 = _mm_macc_pd(dy02,fscal,fiy0);
2074 fiz0 = _mm_macc_pd(dz02,fscal,fiz0);
2076 fjx2 = _mm_macc_pd(dx02,fscal,fjx2);
2077 fjy2 = _mm_macc_pd(dy02,fscal,fjy2);
2078 fjz2 = _mm_macc_pd(dz02,fscal,fjz2);
2080 /**************************
2081 * CALCULATE INTERACTIONS *
2082 **************************/
2084 r10 = _mm_mul_pd(rsq10,rinv10);
2086 /* Calculate table index by multiplying r with table scale and truncate to integer */
2087 rt = _mm_mul_pd(r10,vftabscale);
2088 vfitab = _mm_cvttpd_epi32(rt);
2090 vfeps = _mm_frcz_pd(rt);
2092 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2094 twovfeps = _mm_add_pd(vfeps,vfeps);
2095 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2097 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2098 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2099 F = _mm_setzero_pd();
2100 GMX_MM_TRANSPOSE2_PD(Y,F);
2101 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
2102 H = _mm_setzero_pd();
2103 GMX_MM_TRANSPOSE2_PD(G,H);
2104 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
2105 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
2106 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq10,FF),_mm_mul_pd(vftabscale,rinv10)));
2110 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2112 /* Update vectorial force */
2113 fix1 = _mm_macc_pd(dx10,fscal,fix1);
2114 fiy1 = _mm_macc_pd(dy10,fscal,fiy1);
2115 fiz1 = _mm_macc_pd(dz10,fscal,fiz1);
2117 fjx0 = _mm_macc_pd(dx10,fscal,fjx0);
2118 fjy0 = _mm_macc_pd(dy10,fscal,fjy0);
2119 fjz0 = _mm_macc_pd(dz10,fscal,fjz0);
2121 /**************************
2122 * CALCULATE INTERACTIONS *
2123 **************************/
2125 r11 = _mm_mul_pd(rsq11,rinv11);
2127 /* Calculate table index by multiplying r with table scale and truncate to integer */
2128 rt = _mm_mul_pd(r11,vftabscale);
2129 vfitab = _mm_cvttpd_epi32(rt);
2131 vfeps = _mm_frcz_pd(rt);
2133 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2135 twovfeps = _mm_add_pd(vfeps,vfeps);
2136 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2138 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2139 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2140 F = _mm_setzero_pd();
2141 GMX_MM_TRANSPOSE2_PD(Y,F);
2142 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
2143 H = _mm_setzero_pd();
2144 GMX_MM_TRANSPOSE2_PD(G,H);
2145 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
2146 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
2147 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
2151 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2153 /* Update vectorial force */
2154 fix1 = _mm_macc_pd(dx11,fscal,fix1);
2155 fiy1 = _mm_macc_pd(dy11,fscal,fiy1);
2156 fiz1 = _mm_macc_pd(dz11,fscal,fiz1);
2158 fjx1 = _mm_macc_pd(dx11,fscal,fjx1);
2159 fjy1 = _mm_macc_pd(dy11,fscal,fjy1);
2160 fjz1 = _mm_macc_pd(dz11,fscal,fjz1);
2162 /**************************
2163 * CALCULATE INTERACTIONS *
2164 **************************/
2166 r12 = _mm_mul_pd(rsq12,rinv12);
2168 /* Calculate table index by multiplying r with table scale and truncate to integer */
2169 rt = _mm_mul_pd(r12,vftabscale);
2170 vfitab = _mm_cvttpd_epi32(rt);
2172 vfeps = _mm_frcz_pd(rt);
2174 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2176 twovfeps = _mm_add_pd(vfeps,vfeps);
2177 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2179 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2180 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2181 F = _mm_setzero_pd();
2182 GMX_MM_TRANSPOSE2_PD(Y,F);
2183 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
2184 H = _mm_setzero_pd();
2185 GMX_MM_TRANSPOSE2_PD(G,H);
2186 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
2187 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
2188 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
2192 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2194 /* Update vectorial force */
2195 fix1 = _mm_macc_pd(dx12,fscal,fix1);
2196 fiy1 = _mm_macc_pd(dy12,fscal,fiy1);
2197 fiz1 = _mm_macc_pd(dz12,fscal,fiz1);
2199 fjx2 = _mm_macc_pd(dx12,fscal,fjx2);
2200 fjy2 = _mm_macc_pd(dy12,fscal,fjy2);
2201 fjz2 = _mm_macc_pd(dz12,fscal,fjz2);
2203 /**************************
2204 * CALCULATE INTERACTIONS *
2205 **************************/
2207 r20 = _mm_mul_pd(rsq20,rinv20);
2209 /* Calculate table index by multiplying r with table scale and truncate to integer */
2210 rt = _mm_mul_pd(r20,vftabscale);
2211 vfitab = _mm_cvttpd_epi32(rt);
2213 vfeps = _mm_frcz_pd(rt);
2215 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2217 twovfeps = _mm_add_pd(vfeps,vfeps);
2218 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2220 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2221 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2222 F = _mm_setzero_pd();
2223 GMX_MM_TRANSPOSE2_PD(Y,F);
2224 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
2225 H = _mm_setzero_pd();
2226 GMX_MM_TRANSPOSE2_PD(G,H);
2227 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
2228 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
2229 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq20,FF),_mm_mul_pd(vftabscale,rinv20)));
2233 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2235 /* Update vectorial force */
2236 fix2 = _mm_macc_pd(dx20,fscal,fix2);
2237 fiy2 = _mm_macc_pd(dy20,fscal,fiy2);
2238 fiz2 = _mm_macc_pd(dz20,fscal,fiz2);
2240 fjx0 = _mm_macc_pd(dx20,fscal,fjx0);
2241 fjy0 = _mm_macc_pd(dy20,fscal,fjy0);
2242 fjz0 = _mm_macc_pd(dz20,fscal,fjz0);
2244 /**************************
2245 * CALCULATE INTERACTIONS *
2246 **************************/
2248 r21 = _mm_mul_pd(rsq21,rinv21);
2250 /* Calculate table index by multiplying r with table scale and truncate to integer */
2251 rt = _mm_mul_pd(r21,vftabscale);
2252 vfitab = _mm_cvttpd_epi32(rt);
2254 vfeps = _mm_frcz_pd(rt);
2256 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2258 twovfeps = _mm_add_pd(vfeps,vfeps);
2259 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2261 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2262 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2263 F = _mm_setzero_pd();
2264 GMX_MM_TRANSPOSE2_PD(Y,F);
2265 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
2266 H = _mm_setzero_pd();
2267 GMX_MM_TRANSPOSE2_PD(G,H);
2268 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
2269 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
2270 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
2274 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2276 /* Update vectorial force */
2277 fix2 = _mm_macc_pd(dx21,fscal,fix2);
2278 fiy2 = _mm_macc_pd(dy21,fscal,fiy2);
2279 fiz2 = _mm_macc_pd(dz21,fscal,fiz2);
2281 fjx1 = _mm_macc_pd(dx21,fscal,fjx1);
2282 fjy1 = _mm_macc_pd(dy21,fscal,fjy1);
2283 fjz1 = _mm_macc_pd(dz21,fscal,fjz1);
2285 /**************************
2286 * CALCULATE INTERACTIONS *
2287 **************************/
2289 r22 = _mm_mul_pd(rsq22,rinv22);
2291 /* Calculate table index by multiplying r with table scale and truncate to integer */
2292 rt = _mm_mul_pd(r22,vftabscale);
2293 vfitab = _mm_cvttpd_epi32(rt);
2295 vfeps = _mm_frcz_pd(rt);
2297 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2299 twovfeps = _mm_add_pd(vfeps,vfeps);
2300 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2302 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2303 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2304 F = _mm_setzero_pd();
2305 GMX_MM_TRANSPOSE2_PD(Y,F);
2306 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
2307 H = _mm_setzero_pd();
2308 GMX_MM_TRANSPOSE2_PD(G,H);
2309 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
2310 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
2311 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
2315 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2317 /* Update vectorial force */
2318 fix2 = _mm_macc_pd(dx22,fscal,fix2);
2319 fiy2 = _mm_macc_pd(dy22,fscal,fiy2);
2320 fiz2 = _mm_macc_pd(dz22,fscal,fiz2);
2322 fjx2 = _mm_macc_pd(dx22,fscal,fjx2);
2323 fjy2 = _mm_macc_pd(dy22,fscal,fjy2);
2324 fjz2 = _mm_macc_pd(dz22,fscal,fjz2);
2326 gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
2328 /* Inner loop uses 400 flops */
2331 /* End of innermost loop */
2333 gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
2334 f+i_coord_offset,fshift+i_shift_offset);
2336 /* Increment number of inner iterations */
2337 inneriter += j_index_end - j_index_start;
2339 /* Outer loop uses 18 flops */
2342 /* Increment number of outer iterations */
2345 /* Update outer/inner flops */
2347 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_F,outeriter*18 + inneriter*400);