2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
5 * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
6 * and including many others, as listed in the AUTHORS file in the
7 * top-level source directory and at http://www.gromacs.org.
9 * GROMACS is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU Lesser General Public License
11 * as published by the Free Software Foundation; either version 2.1
12 * of the License, or (at your option) any later version.
14 * GROMACS is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
17 * Lesser General Public License for more details.
19 * You should have received a copy of the GNU Lesser General Public
20 * License along with GROMACS; if not, see
21 * http://www.gnu.org/licenses, or write to the Free Software Foundation,
22 * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
24 * If you want to redistribute modifications to GROMACS, please
25 * consider that scientific software is very special. Version
26 * control is crucial - bugs must be traceable. We will be happy to
27 * consider code for inclusion in the official distribution, but
28 * derived work must not be called official GROMACS. Details are found
29 * in the README & COPYING files - if they are missing, get the
30 * official version at http://www.gromacs.org.
32 * To help us fund GROMACS development, we humbly ask that you cite
33 * the research papers on the package. Check out http://www.gromacs.org.
36 * Note: this file was generated by the GROMACS sse2_double kernel generator.
44 #include "../nb_kernel.h"
45 #include "types/simple.h"
46 #include "gromacs/math/vec.h"
49 #include "gromacs/simd/math_x86_sse2_double.h"
50 #include "kernelutil_x86_sse2_double.h"
53 * Gromacs nonbonded kernel: nb_kernel_ElecCSTab_VdwLJ_GeomW4W4_VF_sse2_double
54 * Electrostatics interaction: CubicSplineTable
55 * VdW interaction: LennardJones
56 * Geometry: Water4-Water4
57 * Calculate force/pot: PotentialAndForce
60 nb_kernel_ElecCSTab_VdwLJ_GeomW4W4_VF_sse2_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;
89 __m128d ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
90 int vdwjidx0A,vdwjidx0B;
91 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
92 int vdwjidx1A,vdwjidx1B;
93 __m128d jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
94 int vdwjidx2A,vdwjidx2B;
95 __m128d jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
96 int vdwjidx3A,vdwjidx3B;
97 __m128d jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
98 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
99 __m128d dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
100 __m128d dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
101 __m128d dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
102 __m128d dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
103 __m128d dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
104 __m128d dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
105 __m128d dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
106 __m128d dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
107 __m128d dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
108 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
111 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
114 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
115 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
117 __m128i ifour = _mm_set1_epi32(4);
118 __m128d rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
120 __m128d dummy_mask,cutoff_mask;
121 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
122 __m128d one = _mm_set1_pd(1.0);
123 __m128d two = _mm_set1_pd(2.0);
129 jindex = nlist->jindex;
131 shiftidx = nlist->shift;
133 shiftvec = fr->shift_vec[0];
134 fshift = fr->fshift[0];
135 facel = _mm_set1_pd(fr->epsfac);
136 charge = mdatoms->chargeA;
137 nvdwtype = fr->ntype;
139 vdwtype = mdatoms->typeA;
141 vftab = kernel_data->table_elec->data;
142 vftabscale = _mm_set1_pd(kernel_data->table_elec->scale);
144 /* Setup water-specific parameters */
145 inr = nlist->iinr[0];
146 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
147 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
148 iq3 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
149 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
151 jq1 = _mm_set1_pd(charge[inr+1]);
152 jq2 = _mm_set1_pd(charge[inr+2]);
153 jq3 = _mm_set1_pd(charge[inr+3]);
154 vdwjidx0A = 2*vdwtype[inr+0];
155 c6_00 = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
156 c12_00 = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
157 qq11 = _mm_mul_pd(iq1,jq1);
158 qq12 = _mm_mul_pd(iq1,jq2);
159 qq13 = _mm_mul_pd(iq1,jq3);
160 qq21 = _mm_mul_pd(iq2,jq1);
161 qq22 = _mm_mul_pd(iq2,jq2);
162 qq23 = _mm_mul_pd(iq2,jq3);
163 qq31 = _mm_mul_pd(iq3,jq1);
164 qq32 = _mm_mul_pd(iq3,jq2);
165 qq33 = _mm_mul_pd(iq3,jq3);
167 /* Avoid stupid compiler warnings */
175 /* Start outer loop over neighborlists */
176 for(iidx=0; iidx<nri; iidx++)
178 /* Load shift vector for this list */
179 i_shift_offset = DIM*shiftidx[iidx];
181 /* Load limits for loop over neighbors */
182 j_index_start = jindex[iidx];
183 j_index_end = jindex[iidx+1];
185 /* Get outer coordinate index */
187 i_coord_offset = DIM*inr;
189 /* Load i particle coords and add shift vector */
190 gmx_mm_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
191 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
193 fix0 = _mm_setzero_pd();
194 fiy0 = _mm_setzero_pd();
195 fiz0 = _mm_setzero_pd();
196 fix1 = _mm_setzero_pd();
197 fiy1 = _mm_setzero_pd();
198 fiz1 = _mm_setzero_pd();
199 fix2 = _mm_setzero_pd();
200 fiy2 = _mm_setzero_pd();
201 fiz2 = _mm_setzero_pd();
202 fix3 = _mm_setzero_pd();
203 fiy3 = _mm_setzero_pd();
204 fiz3 = _mm_setzero_pd();
206 /* Reset potential sums */
207 velecsum = _mm_setzero_pd();
208 vvdwsum = _mm_setzero_pd();
210 /* Start inner kernel loop */
211 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
214 /* Get j neighbor index, and coordinate index */
217 j_coord_offsetA = DIM*jnrA;
218 j_coord_offsetB = DIM*jnrB;
220 /* load j atom coordinates */
221 gmx_mm_load_4rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
222 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
223 &jy2,&jz2,&jx3,&jy3,&jz3);
225 /* Calculate displacement vector */
226 dx00 = _mm_sub_pd(ix0,jx0);
227 dy00 = _mm_sub_pd(iy0,jy0);
228 dz00 = _mm_sub_pd(iz0,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 dx13 = _mm_sub_pd(ix1,jx3);
236 dy13 = _mm_sub_pd(iy1,jy3);
237 dz13 = _mm_sub_pd(iz1,jz3);
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);
244 dx23 = _mm_sub_pd(ix2,jx3);
245 dy23 = _mm_sub_pd(iy2,jy3);
246 dz23 = _mm_sub_pd(iz2,jz3);
247 dx31 = _mm_sub_pd(ix3,jx1);
248 dy31 = _mm_sub_pd(iy3,jy1);
249 dz31 = _mm_sub_pd(iz3,jz1);
250 dx32 = _mm_sub_pd(ix3,jx2);
251 dy32 = _mm_sub_pd(iy3,jy2);
252 dz32 = _mm_sub_pd(iz3,jz2);
253 dx33 = _mm_sub_pd(ix3,jx3);
254 dy33 = _mm_sub_pd(iy3,jy3);
255 dz33 = _mm_sub_pd(iz3,jz3);
257 /* Calculate squared distance and things based on it */
258 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
259 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
260 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
261 rsq13 = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
262 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
263 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
264 rsq23 = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
265 rsq31 = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
266 rsq32 = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
267 rsq33 = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
269 rinv11 = gmx_mm_invsqrt_pd(rsq11);
270 rinv12 = gmx_mm_invsqrt_pd(rsq12);
271 rinv13 = gmx_mm_invsqrt_pd(rsq13);
272 rinv21 = gmx_mm_invsqrt_pd(rsq21);
273 rinv22 = gmx_mm_invsqrt_pd(rsq22);
274 rinv23 = gmx_mm_invsqrt_pd(rsq23);
275 rinv31 = gmx_mm_invsqrt_pd(rsq31);
276 rinv32 = gmx_mm_invsqrt_pd(rsq32);
277 rinv33 = gmx_mm_invsqrt_pd(rsq33);
279 rinvsq00 = gmx_mm_inv_pd(rsq00);
281 fjx0 = _mm_setzero_pd();
282 fjy0 = _mm_setzero_pd();
283 fjz0 = _mm_setzero_pd();
284 fjx1 = _mm_setzero_pd();
285 fjy1 = _mm_setzero_pd();
286 fjz1 = _mm_setzero_pd();
287 fjx2 = _mm_setzero_pd();
288 fjy2 = _mm_setzero_pd();
289 fjz2 = _mm_setzero_pd();
290 fjx3 = _mm_setzero_pd();
291 fjy3 = _mm_setzero_pd();
292 fjz3 = _mm_setzero_pd();
294 /**************************
295 * CALCULATE INTERACTIONS *
296 **************************/
298 /* LENNARD-JONES DISPERSION/REPULSION */
300 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
301 vvdw6 = _mm_mul_pd(c6_00,rinvsix);
302 vvdw12 = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
303 vvdw = _mm_sub_pd( _mm_mul_pd(vvdw12,one_twelfth) , _mm_mul_pd(vvdw6,one_sixth) );
304 fvdw = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
306 /* Update potential sum for this i atom from the interaction with this j atom. */
307 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
311 /* Calculate temporary vectorial force */
312 tx = _mm_mul_pd(fscal,dx00);
313 ty = _mm_mul_pd(fscal,dy00);
314 tz = _mm_mul_pd(fscal,dz00);
316 /* Update vectorial force */
317 fix0 = _mm_add_pd(fix0,tx);
318 fiy0 = _mm_add_pd(fiy0,ty);
319 fiz0 = _mm_add_pd(fiz0,tz);
321 fjx0 = _mm_add_pd(fjx0,tx);
322 fjy0 = _mm_add_pd(fjy0,ty);
323 fjz0 = _mm_add_pd(fjz0,tz);
325 /**************************
326 * CALCULATE INTERACTIONS *
327 **************************/
329 r11 = _mm_mul_pd(rsq11,rinv11);
331 /* Calculate table index by multiplying r with table scale and truncate to integer */
332 rt = _mm_mul_pd(r11,vftabscale);
333 vfitab = _mm_cvttpd_epi32(rt);
334 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
335 vfitab = _mm_slli_epi32(vfitab,2);
337 /* CUBIC SPLINE TABLE ELECTROSTATICS */
338 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
339 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
340 GMX_MM_TRANSPOSE2_PD(Y,F);
341 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
342 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
343 GMX_MM_TRANSPOSE2_PD(G,H);
344 Heps = _mm_mul_pd(vfeps,H);
345 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
346 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
347 velec = _mm_mul_pd(qq11,VV);
348 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
349 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
351 /* Update potential sum for this i atom from the interaction with this j atom. */
352 velecsum = _mm_add_pd(velecsum,velec);
356 /* Calculate temporary vectorial force */
357 tx = _mm_mul_pd(fscal,dx11);
358 ty = _mm_mul_pd(fscal,dy11);
359 tz = _mm_mul_pd(fscal,dz11);
361 /* Update vectorial force */
362 fix1 = _mm_add_pd(fix1,tx);
363 fiy1 = _mm_add_pd(fiy1,ty);
364 fiz1 = _mm_add_pd(fiz1,tz);
366 fjx1 = _mm_add_pd(fjx1,tx);
367 fjy1 = _mm_add_pd(fjy1,ty);
368 fjz1 = _mm_add_pd(fjz1,tz);
370 /**************************
371 * CALCULATE INTERACTIONS *
372 **************************/
374 r12 = _mm_mul_pd(rsq12,rinv12);
376 /* Calculate table index by multiplying r with table scale and truncate to integer */
377 rt = _mm_mul_pd(r12,vftabscale);
378 vfitab = _mm_cvttpd_epi32(rt);
379 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
380 vfitab = _mm_slli_epi32(vfitab,2);
382 /* CUBIC SPLINE TABLE ELECTROSTATICS */
383 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
384 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
385 GMX_MM_TRANSPOSE2_PD(Y,F);
386 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
387 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
388 GMX_MM_TRANSPOSE2_PD(G,H);
389 Heps = _mm_mul_pd(vfeps,H);
390 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
391 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
392 velec = _mm_mul_pd(qq12,VV);
393 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
394 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
396 /* Update potential sum for this i atom from the interaction with this j atom. */
397 velecsum = _mm_add_pd(velecsum,velec);
401 /* Calculate temporary vectorial force */
402 tx = _mm_mul_pd(fscal,dx12);
403 ty = _mm_mul_pd(fscal,dy12);
404 tz = _mm_mul_pd(fscal,dz12);
406 /* Update vectorial force */
407 fix1 = _mm_add_pd(fix1,tx);
408 fiy1 = _mm_add_pd(fiy1,ty);
409 fiz1 = _mm_add_pd(fiz1,tz);
411 fjx2 = _mm_add_pd(fjx2,tx);
412 fjy2 = _mm_add_pd(fjy2,ty);
413 fjz2 = _mm_add_pd(fjz2,tz);
415 /**************************
416 * CALCULATE INTERACTIONS *
417 **************************/
419 r13 = _mm_mul_pd(rsq13,rinv13);
421 /* Calculate table index by multiplying r with table scale and truncate to integer */
422 rt = _mm_mul_pd(r13,vftabscale);
423 vfitab = _mm_cvttpd_epi32(rt);
424 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
425 vfitab = _mm_slli_epi32(vfitab,2);
427 /* CUBIC SPLINE TABLE ELECTROSTATICS */
428 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
429 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
430 GMX_MM_TRANSPOSE2_PD(Y,F);
431 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
432 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
433 GMX_MM_TRANSPOSE2_PD(G,H);
434 Heps = _mm_mul_pd(vfeps,H);
435 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
436 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
437 velec = _mm_mul_pd(qq13,VV);
438 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
439 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq13,FF),_mm_mul_pd(vftabscale,rinv13)));
441 /* Update potential sum for this i atom from the interaction with this j atom. */
442 velecsum = _mm_add_pd(velecsum,velec);
446 /* Calculate temporary vectorial force */
447 tx = _mm_mul_pd(fscal,dx13);
448 ty = _mm_mul_pd(fscal,dy13);
449 tz = _mm_mul_pd(fscal,dz13);
451 /* Update vectorial force */
452 fix1 = _mm_add_pd(fix1,tx);
453 fiy1 = _mm_add_pd(fiy1,ty);
454 fiz1 = _mm_add_pd(fiz1,tz);
456 fjx3 = _mm_add_pd(fjx3,tx);
457 fjy3 = _mm_add_pd(fjy3,ty);
458 fjz3 = _mm_add_pd(fjz3,tz);
460 /**************************
461 * CALCULATE INTERACTIONS *
462 **************************/
464 r21 = _mm_mul_pd(rsq21,rinv21);
466 /* Calculate table index by multiplying r with table scale and truncate to integer */
467 rt = _mm_mul_pd(r21,vftabscale);
468 vfitab = _mm_cvttpd_epi32(rt);
469 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
470 vfitab = _mm_slli_epi32(vfitab,2);
472 /* CUBIC SPLINE TABLE ELECTROSTATICS */
473 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
474 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
475 GMX_MM_TRANSPOSE2_PD(Y,F);
476 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
477 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
478 GMX_MM_TRANSPOSE2_PD(G,H);
479 Heps = _mm_mul_pd(vfeps,H);
480 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
481 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
482 velec = _mm_mul_pd(qq21,VV);
483 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
484 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
486 /* Update potential sum for this i atom from the interaction with this j atom. */
487 velecsum = _mm_add_pd(velecsum,velec);
491 /* Calculate temporary vectorial force */
492 tx = _mm_mul_pd(fscal,dx21);
493 ty = _mm_mul_pd(fscal,dy21);
494 tz = _mm_mul_pd(fscal,dz21);
496 /* Update vectorial force */
497 fix2 = _mm_add_pd(fix2,tx);
498 fiy2 = _mm_add_pd(fiy2,ty);
499 fiz2 = _mm_add_pd(fiz2,tz);
501 fjx1 = _mm_add_pd(fjx1,tx);
502 fjy1 = _mm_add_pd(fjy1,ty);
503 fjz1 = _mm_add_pd(fjz1,tz);
505 /**************************
506 * CALCULATE INTERACTIONS *
507 **************************/
509 r22 = _mm_mul_pd(rsq22,rinv22);
511 /* Calculate table index by multiplying r with table scale and truncate to integer */
512 rt = _mm_mul_pd(r22,vftabscale);
513 vfitab = _mm_cvttpd_epi32(rt);
514 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
515 vfitab = _mm_slli_epi32(vfitab,2);
517 /* CUBIC SPLINE TABLE ELECTROSTATICS */
518 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
519 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
520 GMX_MM_TRANSPOSE2_PD(Y,F);
521 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
522 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
523 GMX_MM_TRANSPOSE2_PD(G,H);
524 Heps = _mm_mul_pd(vfeps,H);
525 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
526 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
527 velec = _mm_mul_pd(qq22,VV);
528 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
529 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
531 /* Update potential sum for this i atom from the interaction with this j atom. */
532 velecsum = _mm_add_pd(velecsum,velec);
536 /* Calculate temporary vectorial force */
537 tx = _mm_mul_pd(fscal,dx22);
538 ty = _mm_mul_pd(fscal,dy22);
539 tz = _mm_mul_pd(fscal,dz22);
541 /* Update vectorial force */
542 fix2 = _mm_add_pd(fix2,tx);
543 fiy2 = _mm_add_pd(fiy2,ty);
544 fiz2 = _mm_add_pd(fiz2,tz);
546 fjx2 = _mm_add_pd(fjx2,tx);
547 fjy2 = _mm_add_pd(fjy2,ty);
548 fjz2 = _mm_add_pd(fjz2,tz);
550 /**************************
551 * CALCULATE INTERACTIONS *
552 **************************/
554 r23 = _mm_mul_pd(rsq23,rinv23);
556 /* Calculate table index by multiplying r with table scale and truncate to integer */
557 rt = _mm_mul_pd(r23,vftabscale);
558 vfitab = _mm_cvttpd_epi32(rt);
559 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
560 vfitab = _mm_slli_epi32(vfitab,2);
562 /* CUBIC SPLINE TABLE ELECTROSTATICS */
563 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
564 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
565 GMX_MM_TRANSPOSE2_PD(Y,F);
566 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
567 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
568 GMX_MM_TRANSPOSE2_PD(G,H);
569 Heps = _mm_mul_pd(vfeps,H);
570 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
571 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
572 velec = _mm_mul_pd(qq23,VV);
573 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
574 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq23,FF),_mm_mul_pd(vftabscale,rinv23)));
576 /* Update potential sum for this i atom from the interaction with this j atom. */
577 velecsum = _mm_add_pd(velecsum,velec);
581 /* Calculate temporary vectorial force */
582 tx = _mm_mul_pd(fscal,dx23);
583 ty = _mm_mul_pd(fscal,dy23);
584 tz = _mm_mul_pd(fscal,dz23);
586 /* Update vectorial force */
587 fix2 = _mm_add_pd(fix2,tx);
588 fiy2 = _mm_add_pd(fiy2,ty);
589 fiz2 = _mm_add_pd(fiz2,tz);
591 fjx3 = _mm_add_pd(fjx3,tx);
592 fjy3 = _mm_add_pd(fjy3,ty);
593 fjz3 = _mm_add_pd(fjz3,tz);
595 /**************************
596 * CALCULATE INTERACTIONS *
597 **************************/
599 r31 = _mm_mul_pd(rsq31,rinv31);
601 /* Calculate table index by multiplying r with table scale and truncate to integer */
602 rt = _mm_mul_pd(r31,vftabscale);
603 vfitab = _mm_cvttpd_epi32(rt);
604 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
605 vfitab = _mm_slli_epi32(vfitab,2);
607 /* CUBIC SPLINE TABLE ELECTROSTATICS */
608 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
609 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
610 GMX_MM_TRANSPOSE2_PD(Y,F);
611 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
612 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
613 GMX_MM_TRANSPOSE2_PD(G,H);
614 Heps = _mm_mul_pd(vfeps,H);
615 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
616 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
617 velec = _mm_mul_pd(qq31,VV);
618 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
619 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq31,FF),_mm_mul_pd(vftabscale,rinv31)));
621 /* Update potential sum for this i atom from the interaction with this j atom. */
622 velecsum = _mm_add_pd(velecsum,velec);
626 /* Calculate temporary vectorial force */
627 tx = _mm_mul_pd(fscal,dx31);
628 ty = _mm_mul_pd(fscal,dy31);
629 tz = _mm_mul_pd(fscal,dz31);
631 /* Update vectorial force */
632 fix3 = _mm_add_pd(fix3,tx);
633 fiy3 = _mm_add_pd(fiy3,ty);
634 fiz3 = _mm_add_pd(fiz3,tz);
636 fjx1 = _mm_add_pd(fjx1,tx);
637 fjy1 = _mm_add_pd(fjy1,ty);
638 fjz1 = _mm_add_pd(fjz1,tz);
640 /**************************
641 * CALCULATE INTERACTIONS *
642 **************************/
644 r32 = _mm_mul_pd(rsq32,rinv32);
646 /* Calculate table index by multiplying r with table scale and truncate to integer */
647 rt = _mm_mul_pd(r32,vftabscale);
648 vfitab = _mm_cvttpd_epi32(rt);
649 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
650 vfitab = _mm_slli_epi32(vfitab,2);
652 /* CUBIC SPLINE TABLE ELECTROSTATICS */
653 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
654 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
655 GMX_MM_TRANSPOSE2_PD(Y,F);
656 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
657 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
658 GMX_MM_TRANSPOSE2_PD(G,H);
659 Heps = _mm_mul_pd(vfeps,H);
660 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
661 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
662 velec = _mm_mul_pd(qq32,VV);
663 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
664 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq32,FF),_mm_mul_pd(vftabscale,rinv32)));
666 /* Update potential sum for this i atom from the interaction with this j atom. */
667 velecsum = _mm_add_pd(velecsum,velec);
671 /* Calculate temporary vectorial force */
672 tx = _mm_mul_pd(fscal,dx32);
673 ty = _mm_mul_pd(fscal,dy32);
674 tz = _mm_mul_pd(fscal,dz32);
676 /* Update vectorial force */
677 fix3 = _mm_add_pd(fix3,tx);
678 fiy3 = _mm_add_pd(fiy3,ty);
679 fiz3 = _mm_add_pd(fiz3,tz);
681 fjx2 = _mm_add_pd(fjx2,tx);
682 fjy2 = _mm_add_pd(fjy2,ty);
683 fjz2 = _mm_add_pd(fjz2,tz);
685 /**************************
686 * CALCULATE INTERACTIONS *
687 **************************/
689 r33 = _mm_mul_pd(rsq33,rinv33);
691 /* Calculate table index by multiplying r with table scale and truncate to integer */
692 rt = _mm_mul_pd(r33,vftabscale);
693 vfitab = _mm_cvttpd_epi32(rt);
694 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
695 vfitab = _mm_slli_epi32(vfitab,2);
697 /* CUBIC SPLINE TABLE ELECTROSTATICS */
698 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
699 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
700 GMX_MM_TRANSPOSE2_PD(Y,F);
701 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
702 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
703 GMX_MM_TRANSPOSE2_PD(G,H);
704 Heps = _mm_mul_pd(vfeps,H);
705 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
706 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
707 velec = _mm_mul_pd(qq33,VV);
708 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
709 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq33,FF),_mm_mul_pd(vftabscale,rinv33)));
711 /* Update potential sum for this i atom from the interaction with this j atom. */
712 velecsum = _mm_add_pd(velecsum,velec);
716 /* Calculate temporary vectorial force */
717 tx = _mm_mul_pd(fscal,dx33);
718 ty = _mm_mul_pd(fscal,dy33);
719 tz = _mm_mul_pd(fscal,dz33);
721 /* Update vectorial force */
722 fix3 = _mm_add_pd(fix3,tx);
723 fiy3 = _mm_add_pd(fiy3,ty);
724 fiz3 = _mm_add_pd(fiz3,tz);
726 fjx3 = _mm_add_pd(fjx3,tx);
727 fjy3 = _mm_add_pd(fjy3,ty);
728 fjz3 = _mm_add_pd(fjz3,tz);
730 gmx_mm_decrement_4rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
732 /* Inner loop uses 422 flops */
739 j_coord_offsetA = DIM*jnrA;
741 /* load j atom coordinates */
742 gmx_mm_load_4rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
743 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
744 &jy2,&jz2,&jx3,&jy3,&jz3);
746 /* Calculate displacement vector */
747 dx00 = _mm_sub_pd(ix0,jx0);
748 dy00 = _mm_sub_pd(iy0,jy0);
749 dz00 = _mm_sub_pd(iz0,jz0);
750 dx11 = _mm_sub_pd(ix1,jx1);
751 dy11 = _mm_sub_pd(iy1,jy1);
752 dz11 = _mm_sub_pd(iz1,jz1);
753 dx12 = _mm_sub_pd(ix1,jx2);
754 dy12 = _mm_sub_pd(iy1,jy2);
755 dz12 = _mm_sub_pd(iz1,jz2);
756 dx13 = _mm_sub_pd(ix1,jx3);
757 dy13 = _mm_sub_pd(iy1,jy3);
758 dz13 = _mm_sub_pd(iz1,jz3);
759 dx21 = _mm_sub_pd(ix2,jx1);
760 dy21 = _mm_sub_pd(iy2,jy1);
761 dz21 = _mm_sub_pd(iz2,jz1);
762 dx22 = _mm_sub_pd(ix2,jx2);
763 dy22 = _mm_sub_pd(iy2,jy2);
764 dz22 = _mm_sub_pd(iz2,jz2);
765 dx23 = _mm_sub_pd(ix2,jx3);
766 dy23 = _mm_sub_pd(iy2,jy3);
767 dz23 = _mm_sub_pd(iz2,jz3);
768 dx31 = _mm_sub_pd(ix3,jx1);
769 dy31 = _mm_sub_pd(iy3,jy1);
770 dz31 = _mm_sub_pd(iz3,jz1);
771 dx32 = _mm_sub_pd(ix3,jx2);
772 dy32 = _mm_sub_pd(iy3,jy2);
773 dz32 = _mm_sub_pd(iz3,jz2);
774 dx33 = _mm_sub_pd(ix3,jx3);
775 dy33 = _mm_sub_pd(iy3,jy3);
776 dz33 = _mm_sub_pd(iz3,jz3);
778 /* Calculate squared distance and things based on it */
779 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
780 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
781 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
782 rsq13 = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
783 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
784 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
785 rsq23 = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
786 rsq31 = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
787 rsq32 = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
788 rsq33 = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
790 rinv11 = gmx_mm_invsqrt_pd(rsq11);
791 rinv12 = gmx_mm_invsqrt_pd(rsq12);
792 rinv13 = gmx_mm_invsqrt_pd(rsq13);
793 rinv21 = gmx_mm_invsqrt_pd(rsq21);
794 rinv22 = gmx_mm_invsqrt_pd(rsq22);
795 rinv23 = gmx_mm_invsqrt_pd(rsq23);
796 rinv31 = gmx_mm_invsqrt_pd(rsq31);
797 rinv32 = gmx_mm_invsqrt_pd(rsq32);
798 rinv33 = gmx_mm_invsqrt_pd(rsq33);
800 rinvsq00 = gmx_mm_inv_pd(rsq00);
802 fjx0 = _mm_setzero_pd();
803 fjy0 = _mm_setzero_pd();
804 fjz0 = _mm_setzero_pd();
805 fjx1 = _mm_setzero_pd();
806 fjy1 = _mm_setzero_pd();
807 fjz1 = _mm_setzero_pd();
808 fjx2 = _mm_setzero_pd();
809 fjy2 = _mm_setzero_pd();
810 fjz2 = _mm_setzero_pd();
811 fjx3 = _mm_setzero_pd();
812 fjy3 = _mm_setzero_pd();
813 fjz3 = _mm_setzero_pd();
815 /**************************
816 * CALCULATE INTERACTIONS *
817 **************************/
819 /* LENNARD-JONES DISPERSION/REPULSION */
821 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
822 vvdw6 = _mm_mul_pd(c6_00,rinvsix);
823 vvdw12 = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
824 vvdw = _mm_sub_pd( _mm_mul_pd(vvdw12,one_twelfth) , _mm_mul_pd(vvdw6,one_sixth) );
825 fvdw = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
827 /* Update potential sum for this i atom from the interaction with this j atom. */
828 vvdw = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
829 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
833 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
835 /* Calculate temporary vectorial force */
836 tx = _mm_mul_pd(fscal,dx00);
837 ty = _mm_mul_pd(fscal,dy00);
838 tz = _mm_mul_pd(fscal,dz00);
840 /* Update vectorial force */
841 fix0 = _mm_add_pd(fix0,tx);
842 fiy0 = _mm_add_pd(fiy0,ty);
843 fiz0 = _mm_add_pd(fiz0,tz);
845 fjx0 = _mm_add_pd(fjx0,tx);
846 fjy0 = _mm_add_pd(fjy0,ty);
847 fjz0 = _mm_add_pd(fjz0,tz);
849 /**************************
850 * CALCULATE INTERACTIONS *
851 **************************/
853 r11 = _mm_mul_pd(rsq11,rinv11);
855 /* Calculate table index by multiplying r with table scale and truncate to integer */
856 rt = _mm_mul_pd(r11,vftabscale);
857 vfitab = _mm_cvttpd_epi32(rt);
858 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
859 vfitab = _mm_slli_epi32(vfitab,2);
861 /* CUBIC SPLINE TABLE ELECTROSTATICS */
862 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
863 F = _mm_setzero_pd();
864 GMX_MM_TRANSPOSE2_PD(Y,F);
865 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
866 H = _mm_setzero_pd();
867 GMX_MM_TRANSPOSE2_PD(G,H);
868 Heps = _mm_mul_pd(vfeps,H);
869 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
870 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
871 velec = _mm_mul_pd(qq11,VV);
872 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
873 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
875 /* Update potential sum for this i atom from the interaction with this j atom. */
876 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
877 velecsum = _mm_add_pd(velecsum,velec);
881 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
883 /* Calculate temporary vectorial force */
884 tx = _mm_mul_pd(fscal,dx11);
885 ty = _mm_mul_pd(fscal,dy11);
886 tz = _mm_mul_pd(fscal,dz11);
888 /* Update vectorial force */
889 fix1 = _mm_add_pd(fix1,tx);
890 fiy1 = _mm_add_pd(fiy1,ty);
891 fiz1 = _mm_add_pd(fiz1,tz);
893 fjx1 = _mm_add_pd(fjx1,tx);
894 fjy1 = _mm_add_pd(fjy1,ty);
895 fjz1 = _mm_add_pd(fjz1,tz);
897 /**************************
898 * CALCULATE INTERACTIONS *
899 **************************/
901 r12 = _mm_mul_pd(rsq12,rinv12);
903 /* Calculate table index by multiplying r with table scale and truncate to integer */
904 rt = _mm_mul_pd(r12,vftabscale);
905 vfitab = _mm_cvttpd_epi32(rt);
906 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
907 vfitab = _mm_slli_epi32(vfitab,2);
909 /* CUBIC SPLINE TABLE ELECTROSTATICS */
910 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
911 F = _mm_setzero_pd();
912 GMX_MM_TRANSPOSE2_PD(Y,F);
913 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
914 H = _mm_setzero_pd();
915 GMX_MM_TRANSPOSE2_PD(G,H);
916 Heps = _mm_mul_pd(vfeps,H);
917 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
918 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
919 velec = _mm_mul_pd(qq12,VV);
920 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
921 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
923 /* Update potential sum for this i atom from the interaction with this j atom. */
924 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
925 velecsum = _mm_add_pd(velecsum,velec);
929 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
931 /* Calculate temporary vectorial force */
932 tx = _mm_mul_pd(fscal,dx12);
933 ty = _mm_mul_pd(fscal,dy12);
934 tz = _mm_mul_pd(fscal,dz12);
936 /* Update vectorial force */
937 fix1 = _mm_add_pd(fix1,tx);
938 fiy1 = _mm_add_pd(fiy1,ty);
939 fiz1 = _mm_add_pd(fiz1,tz);
941 fjx2 = _mm_add_pd(fjx2,tx);
942 fjy2 = _mm_add_pd(fjy2,ty);
943 fjz2 = _mm_add_pd(fjz2,tz);
945 /**************************
946 * CALCULATE INTERACTIONS *
947 **************************/
949 r13 = _mm_mul_pd(rsq13,rinv13);
951 /* Calculate table index by multiplying r with table scale and truncate to integer */
952 rt = _mm_mul_pd(r13,vftabscale);
953 vfitab = _mm_cvttpd_epi32(rt);
954 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
955 vfitab = _mm_slli_epi32(vfitab,2);
957 /* CUBIC SPLINE TABLE ELECTROSTATICS */
958 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
959 F = _mm_setzero_pd();
960 GMX_MM_TRANSPOSE2_PD(Y,F);
961 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
962 H = _mm_setzero_pd();
963 GMX_MM_TRANSPOSE2_PD(G,H);
964 Heps = _mm_mul_pd(vfeps,H);
965 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
966 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
967 velec = _mm_mul_pd(qq13,VV);
968 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
969 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq13,FF),_mm_mul_pd(vftabscale,rinv13)));
971 /* Update potential sum for this i atom from the interaction with this j atom. */
972 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
973 velecsum = _mm_add_pd(velecsum,velec);
977 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
979 /* Calculate temporary vectorial force */
980 tx = _mm_mul_pd(fscal,dx13);
981 ty = _mm_mul_pd(fscal,dy13);
982 tz = _mm_mul_pd(fscal,dz13);
984 /* Update vectorial force */
985 fix1 = _mm_add_pd(fix1,tx);
986 fiy1 = _mm_add_pd(fiy1,ty);
987 fiz1 = _mm_add_pd(fiz1,tz);
989 fjx3 = _mm_add_pd(fjx3,tx);
990 fjy3 = _mm_add_pd(fjy3,ty);
991 fjz3 = _mm_add_pd(fjz3,tz);
993 /**************************
994 * CALCULATE INTERACTIONS *
995 **************************/
997 r21 = _mm_mul_pd(rsq21,rinv21);
999 /* Calculate table index by multiplying r with table scale and truncate to integer */
1000 rt = _mm_mul_pd(r21,vftabscale);
1001 vfitab = _mm_cvttpd_epi32(rt);
1002 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1003 vfitab = _mm_slli_epi32(vfitab,2);
1005 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1006 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1007 F = _mm_setzero_pd();
1008 GMX_MM_TRANSPOSE2_PD(Y,F);
1009 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1010 H = _mm_setzero_pd();
1011 GMX_MM_TRANSPOSE2_PD(G,H);
1012 Heps = _mm_mul_pd(vfeps,H);
1013 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1014 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1015 velec = _mm_mul_pd(qq21,VV);
1016 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1017 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
1019 /* Update potential sum for this i atom from the interaction with this j atom. */
1020 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1021 velecsum = _mm_add_pd(velecsum,velec);
1025 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1027 /* Calculate temporary vectorial force */
1028 tx = _mm_mul_pd(fscal,dx21);
1029 ty = _mm_mul_pd(fscal,dy21);
1030 tz = _mm_mul_pd(fscal,dz21);
1032 /* Update vectorial force */
1033 fix2 = _mm_add_pd(fix2,tx);
1034 fiy2 = _mm_add_pd(fiy2,ty);
1035 fiz2 = _mm_add_pd(fiz2,tz);
1037 fjx1 = _mm_add_pd(fjx1,tx);
1038 fjy1 = _mm_add_pd(fjy1,ty);
1039 fjz1 = _mm_add_pd(fjz1,tz);
1041 /**************************
1042 * CALCULATE INTERACTIONS *
1043 **************************/
1045 r22 = _mm_mul_pd(rsq22,rinv22);
1047 /* Calculate table index by multiplying r with table scale and truncate to integer */
1048 rt = _mm_mul_pd(r22,vftabscale);
1049 vfitab = _mm_cvttpd_epi32(rt);
1050 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1051 vfitab = _mm_slli_epi32(vfitab,2);
1053 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1054 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1055 F = _mm_setzero_pd();
1056 GMX_MM_TRANSPOSE2_PD(Y,F);
1057 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1058 H = _mm_setzero_pd();
1059 GMX_MM_TRANSPOSE2_PD(G,H);
1060 Heps = _mm_mul_pd(vfeps,H);
1061 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1062 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1063 velec = _mm_mul_pd(qq22,VV);
1064 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1065 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
1067 /* Update potential sum for this i atom from the interaction with this j atom. */
1068 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1069 velecsum = _mm_add_pd(velecsum,velec);
1073 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1075 /* Calculate temporary vectorial force */
1076 tx = _mm_mul_pd(fscal,dx22);
1077 ty = _mm_mul_pd(fscal,dy22);
1078 tz = _mm_mul_pd(fscal,dz22);
1080 /* Update vectorial force */
1081 fix2 = _mm_add_pd(fix2,tx);
1082 fiy2 = _mm_add_pd(fiy2,ty);
1083 fiz2 = _mm_add_pd(fiz2,tz);
1085 fjx2 = _mm_add_pd(fjx2,tx);
1086 fjy2 = _mm_add_pd(fjy2,ty);
1087 fjz2 = _mm_add_pd(fjz2,tz);
1089 /**************************
1090 * CALCULATE INTERACTIONS *
1091 **************************/
1093 r23 = _mm_mul_pd(rsq23,rinv23);
1095 /* Calculate table index by multiplying r with table scale and truncate to integer */
1096 rt = _mm_mul_pd(r23,vftabscale);
1097 vfitab = _mm_cvttpd_epi32(rt);
1098 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1099 vfitab = _mm_slli_epi32(vfitab,2);
1101 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1102 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1103 F = _mm_setzero_pd();
1104 GMX_MM_TRANSPOSE2_PD(Y,F);
1105 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1106 H = _mm_setzero_pd();
1107 GMX_MM_TRANSPOSE2_PD(G,H);
1108 Heps = _mm_mul_pd(vfeps,H);
1109 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1110 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1111 velec = _mm_mul_pd(qq23,VV);
1112 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1113 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq23,FF),_mm_mul_pd(vftabscale,rinv23)));
1115 /* Update potential sum for this i atom from the interaction with this j atom. */
1116 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1117 velecsum = _mm_add_pd(velecsum,velec);
1121 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1123 /* Calculate temporary vectorial force */
1124 tx = _mm_mul_pd(fscal,dx23);
1125 ty = _mm_mul_pd(fscal,dy23);
1126 tz = _mm_mul_pd(fscal,dz23);
1128 /* Update vectorial force */
1129 fix2 = _mm_add_pd(fix2,tx);
1130 fiy2 = _mm_add_pd(fiy2,ty);
1131 fiz2 = _mm_add_pd(fiz2,tz);
1133 fjx3 = _mm_add_pd(fjx3,tx);
1134 fjy3 = _mm_add_pd(fjy3,ty);
1135 fjz3 = _mm_add_pd(fjz3,tz);
1137 /**************************
1138 * CALCULATE INTERACTIONS *
1139 **************************/
1141 r31 = _mm_mul_pd(rsq31,rinv31);
1143 /* Calculate table index by multiplying r with table scale and truncate to integer */
1144 rt = _mm_mul_pd(r31,vftabscale);
1145 vfitab = _mm_cvttpd_epi32(rt);
1146 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1147 vfitab = _mm_slli_epi32(vfitab,2);
1149 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1150 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1151 F = _mm_setzero_pd();
1152 GMX_MM_TRANSPOSE2_PD(Y,F);
1153 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1154 H = _mm_setzero_pd();
1155 GMX_MM_TRANSPOSE2_PD(G,H);
1156 Heps = _mm_mul_pd(vfeps,H);
1157 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1158 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1159 velec = _mm_mul_pd(qq31,VV);
1160 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1161 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq31,FF),_mm_mul_pd(vftabscale,rinv31)));
1163 /* Update potential sum for this i atom from the interaction with this j atom. */
1164 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1165 velecsum = _mm_add_pd(velecsum,velec);
1169 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1171 /* Calculate temporary vectorial force */
1172 tx = _mm_mul_pd(fscal,dx31);
1173 ty = _mm_mul_pd(fscal,dy31);
1174 tz = _mm_mul_pd(fscal,dz31);
1176 /* Update vectorial force */
1177 fix3 = _mm_add_pd(fix3,tx);
1178 fiy3 = _mm_add_pd(fiy3,ty);
1179 fiz3 = _mm_add_pd(fiz3,tz);
1181 fjx1 = _mm_add_pd(fjx1,tx);
1182 fjy1 = _mm_add_pd(fjy1,ty);
1183 fjz1 = _mm_add_pd(fjz1,tz);
1185 /**************************
1186 * CALCULATE INTERACTIONS *
1187 **************************/
1189 r32 = _mm_mul_pd(rsq32,rinv32);
1191 /* Calculate table index by multiplying r with table scale and truncate to integer */
1192 rt = _mm_mul_pd(r32,vftabscale);
1193 vfitab = _mm_cvttpd_epi32(rt);
1194 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1195 vfitab = _mm_slli_epi32(vfitab,2);
1197 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1198 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1199 F = _mm_setzero_pd();
1200 GMX_MM_TRANSPOSE2_PD(Y,F);
1201 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1202 H = _mm_setzero_pd();
1203 GMX_MM_TRANSPOSE2_PD(G,H);
1204 Heps = _mm_mul_pd(vfeps,H);
1205 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1206 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1207 velec = _mm_mul_pd(qq32,VV);
1208 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1209 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq32,FF),_mm_mul_pd(vftabscale,rinv32)));
1211 /* Update potential sum for this i atom from the interaction with this j atom. */
1212 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1213 velecsum = _mm_add_pd(velecsum,velec);
1217 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1219 /* Calculate temporary vectorial force */
1220 tx = _mm_mul_pd(fscal,dx32);
1221 ty = _mm_mul_pd(fscal,dy32);
1222 tz = _mm_mul_pd(fscal,dz32);
1224 /* Update vectorial force */
1225 fix3 = _mm_add_pd(fix3,tx);
1226 fiy3 = _mm_add_pd(fiy3,ty);
1227 fiz3 = _mm_add_pd(fiz3,tz);
1229 fjx2 = _mm_add_pd(fjx2,tx);
1230 fjy2 = _mm_add_pd(fjy2,ty);
1231 fjz2 = _mm_add_pd(fjz2,tz);
1233 /**************************
1234 * CALCULATE INTERACTIONS *
1235 **************************/
1237 r33 = _mm_mul_pd(rsq33,rinv33);
1239 /* Calculate table index by multiplying r with table scale and truncate to integer */
1240 rt = _mm_mul_pd(r33,vftabscale);
1241 vfitab = _mm_cvttpd_epi32(rt);
1242 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1243 vfitab = _mm_slli_epi32(vfitab,2);
1245 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1246 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1247 F = _mm_setzero_pd();
1248 GMX_MM_TRANSPOSE2_PD(Y,F);
1249 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1250 H = _mm_setzero_pd();
1251 GMX_MM_TRANSPOSE2_PD(G,H);
1252 Heps = _mm_mul_pd(vfeps,H);
1253 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1254 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1255 velec = _mm_mul_pd(qq33,VV);
1256 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1257 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq33,FF),_mm_mul_pd(vftabscale,rinv33)));
1259 /* Update potential sum for this i atom from the interaction with this j atom. */
1260 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1261 velecsum = _mm_add_pd(velecsum,velec);
1265 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1267 /* Calculate temporary vectorial force */
1268 tx = _mm_mul_pd(fscal,dx33);
1269 ty = _mm_mul_pd(fscal,dy33);
1270 tz = _mm_mul_pd(fscal,dz33);
1272 /* Update vectorial force */
1273 fix3 = _mm_add_pd(fix3,tx);
1274 fiy3 = _mm_add_pd(fiy3,ty);
1275 fiz3 = _mm_add_pd(fiz3,tz);
1277 fjx3 = _mm_add_pd(fjx3,tx);
1278 fjy3 = _mm_add_pd(fjy3,ty);
1279 fjz3 = _mm_add_pd(fjz3,tz);
1281 gmx_mm_decrement_4rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1283 /* Inner loop uses 422 flops */
1286 /* End of innermost loop */
1288 gmx_mm_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1289 f+i_coord_offset,fshift+i_shift_offset);
1292 /* Update potential energies */
1293 gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1294 gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1296 /* Increment number of inner iterations */
1297 inneriter += j_index_end - j_index_start;
1299 /* Outer loop uses 26 flops */
1302 /* Increment number of outer iterations */
1305 /* Update outer/inner flops */
1307 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*422);
1310 * Gromacs nonbonded kernel: nb_kernel_ElecCSTab_VdwLJ_GeomW4W4_F_sse2_double
1311 * Electrostatics interaction: CubicSplineTable
1312 * VdW interaction: LennardJones
1313 * Geometry: Water4-Water4
1314 * Calculate force/pot: Force
1317 nb_kernel_ElecCSTab_VdwLJ_GeomW4W4_F_sse2_double
1318 (t_nblist * gmx_restrict nlist,
1319 rvec * gmx_restrict xx,
1320 rvec * gmx_restrict ff,
1321 t_forcerec * gmx_restrict fr,
1322 t_mdatoms * gmx_restrict mdatoms,
1323 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1324 t_nrnb * gmx_restrict nrnb)
1326 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1327 * just 0 for non-waters.
1328 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
1329 * jnr indices corresponding to data put in the four positions in the SIMD register.
1331 int i_shift_offset,i_coord_offset,outeriter,inneriter;
1332 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1334 int j_coord_offsetA,j_coord_offsetB;
1335 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
1336 real rcutoff_scalar;
1337 real *shiftvec,*fshift,*x,*f;
1338 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1340 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1342 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1344 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1346 __m128d ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1347 int vdwjidx0A,vdwjidx0B;
1348 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1349 int vdwjidx1A,vdwjidx1B;
1350 __m128d jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1351 int vdwjidx2A,vdwjidx2B;
1352 __m128d jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1353 int vdwjidx3A,vdwjidx3B;
1354 __m128d jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1355 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1356 __m128d dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1357 __m128d dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1358 __m128d dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1359 __m128d dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1360 __m128d dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1361 __m128d dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1362 __m128d dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1363 __m128d dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1364 __m128d dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1365 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
1368 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1371 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
1372 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
1374 __m128i ifour = _mm_set1_epi32(4);
1375 __m128d rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
1377 __m128d dummy_mask,cutoff_mask;
1378 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
1379 __m128d one = _mm_set1_pd(1.0);
1380 __m128d two = _mm_set1_pd(2.0);
1386 jindex = nlist->jindex;
1388 shiftidx = nlist->shift;
1390 shiftvec = fr->shift_vec[0];
1391 fshift = fr->fshift[0];
1392 facel = _mm_set1_pd(fr->epsfac);
1393 charge = mdatoms->chargeA;
1394 nvdwtype = fr->ntype;
1395 vdwparam = fr->nbfp;
1396 vdwtype = mdatoms->typeA;
1398 vftab = kernel_data->table_elec->data;
1399 vftabscale = _mm_set1_pd(kernel_data->table_elec->scale);
1401 /* Setup water-specific parameters */
1402 inr = nlist->iinr[0];
1403 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
1404 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
1405 iq3 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
1406 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
1408 jq1 = _mm_set1_pd(charge[inr+1]);
1409 jq2 = _mm_set1_pd(charge[inr+2]);
1410 jq3 = _mm_set1_pd(charge[inr+3]);
1411 vdwjidx0A = 2*vdwtype[inr+0];
1412 c6_00 = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
1413 c12_00 = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
1414 qq11 = _mm_mul_pd(iq1,jq1);
1415 qq12 = _mm_mul_pd(iq1,jq2);
1416 qq13 = _mm_mul_pd(iq1,jq3);
1417 qq21 = _mm_mul_pd(iq2,jq1);
1418 qq22 = _mm_mul_pd(iq2,jq2);
1419 qq23 = _mm_mul_pd(iq2,jq3);
1420 qq31 = _mm_mul_pd(iq3,jq1);
1421 qq32 = _mm_mul_pd(iq3,jq2);
1422 qq33 = _mm_mul_pd(iq3,jq3);
1424 /* Avoid stupid compiler warnings */
1426 j_coord_offsetA = 0;
1427 j_coord_offsetB = 0;
1432 /* Start outer loop over neighborlists */
1433 for(iidx=0; iidx<nri; iidx++)
1435 /* Load shift vector for this list */
1436 i_shift_offset = DIM*shiftidx[iidx];
1438 /* Load limits for loop over neighbors */
1439 j_index_start = jindex[iidx];
1440 j_index_end = jindex[iidx+1];
1442 /* Get outer coordinate index */
1444 i_coord_offset = DIM*inr;
1446 /* Load i particle coords and add shift vector */
1447 gmx_mm_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1448 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1450 fix0 = _mm_setzero_pd();
1451 fiy0 = _mm_setzero_pd();
1452 fiz0 = _mm_setzero_pd();
1453 fix1 = _mm_setzero_pd();
1454 fiy1 = _mm_setzero_pd();
1455 fiz1 = _mm_setzero_pd();
1456 fix2 = _mm_setzero_pd();
1457 fiy2 = _mm_setzero_pd();
1458 fiz2 = _mm_setzero_pd();
1459 fix3 = _mm_setzero_pd();
1460 fiy3 = _mm_setzero_pd();
1461 fiz3 = _mm_setzero_pd();
1463 /* Start inner kernel loop */
1464 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
1467 /* Get j neighbor index, and coordinate index */
1469 jnrB = jjnr[jidx+1];
1470 j_coord_offsetA = DIM*jnrA;
1471 j_coord_offsetB = DIM*jnrB;
1473 /* load j atom coordinates */
1474 gmx_mm_load_4rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1475 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1476 &jy2,&jz2,&jx3,&jy3,&jz3);
1478 /* Calculate displacement vector */
1479 dx00 = _mm_sub_pd(ix0,jx0);
1480 dy00 = _mm_sub_pd(iy0,jy0);
1481 dz00 = _mm_sub_pd(iz0,jz0);
1482 dx11 = _mm_sub_pd(ix1,jx1);
1483 dy11 = _mm_sub_pd(iy1,jy1);
1484 dz11 = _mm_sub_pd(iz1,jz1);
1485 dx12 = _mm_sub_pd(ix1,jx2);
1486 dy12 = _mm_sub_pd(iy1,jy2);
1487 dz12 = _mm_sub_pd(iz1,jz2);
1488 dx13 = _mm_sub_pd(ix1,jx3);
1489 dy13 = _mm_sub_pd(iy1,jy3);
1490 dz13 = _mm_sub_pd(iz1,jz3);
1491 dx21 = _mm_sub_pd(ix2,jx1);
1492 dy21 = _mm_sub_pd(iy2,jy1);
1493 dz21 = _mm_sub_pd(iz2,jz1);
1494 dx22 = _mm_sub_pd(ix2,jx2);
1495 dy22 = _mm_sub_pd(iy2,jy2);
1496 dz22 = _mm_sub_pd(iz2,jz2);
1497 dx23 = _mm_sub_pd(ix2,jx3);
1498 dy23 = _mm_sub_pd(iy2,jy3);
1499 dz23 = _mm_sub_pd(iz2,jz3);
1500 dx31 = _mm_sub_pd(ix3,jx1);
1501 dy31 = _mm_sub_pd(iy3,jy1);
1502 dz31 = _mm_sub_pd(iz3,jz1);
1503 dx32 = _mm_sub_pd(ix3,jx2);
1504 dy32 = _mm_sub_pd(iy3,jy2);
1505 dz32 = _mm_sub_pd(iz3,jz2);
1506 dx33 = _mm_sub_pd(ix3,jx3);
1507 dy33 = _mm_sub_pd(iy3,jy3);
1508 dz33 = _mm_sub_pd(iz3,jz3);
1510 /* Calculate squared distance and things based on it */
1511 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1512 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1513 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1514 rsq13 = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
1515 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1516 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1517 rsq23 = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
1518 rsq31 = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
1519 rsq32 = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
1520 rsq33 = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
1522 rinv11 = gmx_mm_invsqrt_pd(rsq11);
1523 rinv12 = gmx_mm_invsqrt_pd(rsq12);
1524 rinv13 = gmx_mm_invsqrt_pd(rsq13);
1525 rinv21 = gmx_mm_invsqrt_pd(rsq21);
1526 rinv22 = gmx_mm_invsqrt_pd(rsq22);
1527 rinv23 = gmx_mm_invsqrt_pd(rsq23);
1528 rinv31 = gmx_mm_invsqrt_pd(rsq31);
1529 rinv32 = gmx_mm_invsqrt_pd(rsq32);
1530 rinv33 = gmx_mm_invsqrt_pd(rsq33);
1532 rinvsq00 = gmx_mm_inv_pd(rsq00);
1534 fjx0 = _mm_setzero_pd();
1535 fjy0 = _mm_setzero_pd();
1536 fjz0 = _mm_setzero_pd();
1537 fjx1 = _mm_setzero_pd();
1538 fjy1 = _mm_setzero_pd();
1539 fjz1 = _mm_setzero_pd();
1540 fjx2 = _mm_setzero_pd();
1541 fjy2 = _mm_setzero_pd();
1542 fjz2 = _mm_setzero_pd();
1543 fjx3 = _mm_setzero_pd();
1544 fjy3 = _mm_setzero_pd();
1545 fjz3 = _mm_setzero_pd();
1547 /**************************
1548 * CALCULATE INTERACTIONS *
1549 **************************/
1551 /* LENNARD-JONES DISPERSION/REPULSION */
1553 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1554 fvdw = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
1558 /* Calculate temporary vectorial force */
1559 tx = _mm_mul_pd(fscal,dx00);
1560 ty = _mm_mul_pd(fscal,dy00);
1561 tz = _mm_mul_pd(fscal,dz00);
1563 /* Update vectorial force */
1564 fix0 = _mm_add_pd(fix0,tx);
1565 fiy0 = _mm_add_pd(fiy0,ty);
1566 fiz0 = _mm_add_pd(fiz0,tz);
1568 fjx0 = _mm_add_pd(fjx0,tx);
1569 fjy0 = _mm_add_pd(fjy0,ty);
1570 fjz0 = _mm_add_pd(fjz0,tz);
1572 /**************************
1573 * CALCULATE INTERACTIONS *
1574 **************************/
1576 r11 = _mm_mul_pd(rsq11,rinv11);
1578 /* Calculate table index by multiplying r with table scale and truncate to integer */
1579 rt = _mm_mul_pd(r11,vftabscale);
1580 vfitab = _mm_cvttpd_epi32(rt);
1581 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1582 vfitab = _mm_slli_epi32(vfitab,2);
1584 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1585 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1586 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1587 GMX_MM_TRANSPOSE2_PD(Y,F);
1588 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1589 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1590 GMX_MM_TRANSPOSE2_PD(G,H);
1591 Heps = _mm_mul_pd(vfeps,H);
1592 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1593 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1594 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
1598 /* Calculate temporary vectorial force */
1599 tx = _mm_mul_pd(fscal,dx11);
1600 ty = _mm_mul_pd(fscal,dy11);
1601 tz = _mm_mul_pd(fscal,dz11);
1603 /* Update vectorial force */
1604 fix1 = _mm_add_pd(fix1,tx);
1605 fiy1 = _mm_add_pd(fiy1,ty);
1606 fiz1 = _mm_add_pd(fiz1,tz);
1608 fjx1 = _mm_add_pd(fjx1,tx);
1609 fjy1 = _mm_add_pd(fjy1,ty);
1610 fjz1 = _mm_add_pd(fjz1,tz);
1612 /**************************
1613 * CALCULATE INTERACTIONS *
1614 **************************/
1616 r12 = _mm_mul_pd(rsq12,rinv12);
1618 /* Calculate table index by multiplying r with table scale and truncate to integer */
1619 rt = _mm_mul_pd(r12,vftabscale);
1620 vfitab = _mm_cvttpd_epi32(rt);
1621 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1622 vfitab = _mm_slli_epi32(vfitab,2);
1624 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1625 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1626 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1627 GMX_MM_TRANSPOSE2_PD(Y,F);
1628 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1629 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1630 GMX_MM_TRANSPOSE2_PD(G,H);
1631 Heps = _mm_mul_pd(vfeps,H);
1632 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1633 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1634 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
1638 /* Calculate temporary vectorial force */
1639 tx = _mm_mul_pd(fscal,dx12);
1640 ty = _mm_mul_pd(fscal,dy12);
1641 tz = _mm_mul_pd(fscal,dz12);
1643 /* Update vectorial force */
1644 fix1 = _mm_add_pd(fix1,tx);
1645 fiy1 = _mm_add_pd(fiy1,ty);
1646 fiz1 = _mm_add_pd(fiz1,tz);
1648 fjx2 = _mm_add_pd(fjx2,tx);
1649 fjy2 = _mm_add_pd(fjy2,ty);
1650 fjz2 = _mm_add_pd(fjz2,tz);
1652 /**************************
1653 * CALCULATE INTERACTIONS *
1654 **************************/
1656 r13 = _mm_mul_pd(rsq13,rinv13);
1658 /* Calculate table index by multiplying r with table scale and truncate to integer */
1659 rt = _mm_mul_pd(r13,vftabscale);
1660 vfitab = _mm_cvttpd_epi32(rt);
1661 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1662 vfitab = _mm_slli_epi32(vfitab,2);
1664 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1665 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1666 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1667 GMX_MM_TRANSPOSE2_PD(Y,F);
1668 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1669 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1670 GMX_MM_TRANSPOSE2_PD(G,H);
1671 Heps = _mm_mul_pd(vfeps,H);
1672 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1673 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1674 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq13,FF),_mm_mul_pd(vftabscale,rinv13)));
1678 /* Calculate temporary vectorial force */
1679 tx = _mm_mul_pd(fscal,dx13);
1680 ty = _mm_mul_pd(fscal,dy13);
1681 tz = _mm_mul_pd(fscal,dz13);
1683 /* Update vectorial force */
1684 fix1 = _mm_add_pd(fix1,tx);
1685 fiy1 = _mm_add_pd(fiy1,ty);
1686 fiz1 = _mm_add_pd(fiz1,tz);
1688 fjx3 = _mm_add_pd(fjx3,tx);
1689 fjy3 = _mm_add_pd(fjy3,ty);
1690 fjz3 = _mm_add_pd(fjz3,tz);
1692 /**************************
1693 * CALCULATE INTERACTIONS *
1694 **************************/
1696 r21 = _mm_mul_pd(rsq21,rinv21);
1698 /* Calculate table index by multiplying r with table scale and truncate to integer */
1699 rt = _mm_mul_pd(r21,vftabscale);
1700 vfitab = _mm_cvttpd_epi32(rt);
1701 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1702 vfitab = _mm_slli_epi32(vfitab,2);
1704 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1705 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1706 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1707 GMX_MM_TRANSPOSE2_PD(Y,F);
1708 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1709 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1710 GMX_MM_TRANSPOSE2_PD(G,H);
1711 Heps = _mm_mul_pd(vfeps,H);
1712 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1713 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1714 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
1718 /* Calculate temporary vectorial force */
1719 tx = _mm_mul_pd(fscal,dx21);
1720 ty = _mm_mul_pd(fscal,dy21);
1721 tz = _mm_mul_pd(fscal,dz21);
1723 /* Update vectorial force */
1724 fix2 = _mm_add_pd(fix2,tx);
1725 fiy2 = _mm_add_pd(fiy2,ty);
1726 fiz2 = _mm_add_pd(fiz2,tz);
1728 fjx1 = _mm_add_pd(fjx1,tx);
1729 fjy1 = _mm_add_pd(fjy1,ty);
1730 fjz1 = _mm_add_pd(fjz1,tz);
1732 /**************************
1733 * CALCULATE INTERACTIONS *
1734 **************************/
1736 r22 = _mm_mul_pd(rsq22,rinv22);
1738 /* Calculate table index by multiplying r with table scale and truncate to integer */
1739 rt = _mm_mul_pd(r22,vftabscale);
1740 vfitab = _mm_cvttpd_epi32(rt);
1741 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1742 vfitab = _mm_slli_epi32(vfitab,2);
1744 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1745 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1746 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1747 GMX_MM_TRANSPOSE2_PD(Y,F);
1748 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1749 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1750 GMX_MM_TRANSPOSE2_PD(G,H);
1751 Heps = _mm_mul_pd(vfeps,H);
1752 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1753 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1754 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
1758 /* Calculate temporary vectorial force */
1759 tx = _mm_mul_pd(fscal,dx22);
1760 ty = _mm_mul_pd(fscal,dy22);
1761 tz = _mm_mul_pd(fscal,dz22);
1763 /* Update vectorial force */
1764 fix2 = _mm_add_pd(fix2,tx);
1765 fiy2 = _mm_add_pd(fiy2,ty);
1766 fiz2 = _mm_add_pd(fiz2,tz);
1768 fjx2 = _mm_add_pd(fjx2,tx);
1769 fjy2 = _mm_add_pd(fjy2,ty);
1770 fjz2 = _mm_add_pd(fjz2,tz);
1772 /**************************
1773 * CALCULATE INTERACTIONS *
1774 **************************/
1776 r23 = _mm_mul_pd(rsq23,rinv23);
1778 /* Calculate table index by multiplying r with table scale and truncate to integer */
1779 rt = _mm_mul_pd(r23,vftabscale);
1780 vfitab = _mm_cvttpd_epi32(rt);
1781 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1782 vfitab = _mm_slli_epi32(vfitab,2);
1784 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1785 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1786 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1787 GMX_MM_TRANSPOSE2_PD(Y,F);
1788 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1789 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1790 GMX_MM_TRANSPOSE2_PD(G,H);
1791 Heps = _mm_mul_pd(vfeps,H);
1792 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1793 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1794 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq23,FF),_mm_mul_pd(vftabscale,rinv23)));
1798 /* Calculate temporary vectorial force */
1799 tx = _mm_mul_pd(fscal,dx23);
1800 ty = _mm_mul_pd(fscal,dy23);
1801 tz = _mm_mul_pd(fscal,dz23);
1803 /* Update vectorial force */
1804 fix2 = _mm_add_pd(fix2,tx);
1805 fiy2 = _mm_add_pd(fiy2,ty);
1806 fiz2 = _mm_add_pd(fiz2,tz);
1808 fjx3 = _mm_add_pd(fjx3,tx);
1809 fjy3 = _mm_add_pd(fjy3,ty);
1810 fjz3 = _mm_add_pd(fjz3,tz);
1812 /**************************
1813 * CALCULATE INTERACTIONS *
1814 **************************/
1816 r31 = _mm_mul_pd(rsq31,rinv31);
1818 /* Calculate table index by multiplying r with table scale and truncate to integer */
1819 rt = _mm_mul_pd(r31,vftabscale);
1820 vfitab = _mm_cvttpd_epi32(rt);
1821 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1822 vfitab = _mm_slli_epi32(vfitab,2);
1824 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1825 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1826 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1827 GMX_MM_TRANSPOSE2_PD(Y,F);
1828 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1829 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1830 GMX_MM_TRANSPOSE2_PD(G,H);
1831 Heps = _mm_mul_pd(vfeps,H);
1832 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1833 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1834 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq31,FF),_mm_mul_pd(vftabscale,rinv31)));
1838 /* Calculate temporary vectorial force */
1839 tx = _mm_mul_pd(fscal,dx31);
1840 ty = _mm_mul_pd(fscal,dy31);
1841 tz = _mm_mul_pd(fscal,dz31);
1843 /* Update vectorial force */
1844 fix3 = _mm_add_pd(fix3,tx);
1845 fiy3 = _mm_add_pd(fiy3,ty);
1846 fiz3 = _mm_add_pd(fiz3,tz);
1848 fjx1 = _mm_add_pd(fjx1,tx);
1849 fjy1 = _mm_add_pd(fjy1,ty);
1850 fjz1 = _mm_add_pd(fjz1,tz);
1852 /**************************
1853 * CALCULATE INTERACTIONS *
1854 **************************/
1856 r32 = _mm_mul_pd(rsq32,rinv32);
1858 /* Calculate table index by multiplying r with table scale and truncate to integer */
1859 rt = _mm_mul_pd(r32,vftabscale);
1860 vfitab = _mm_cvttpd_epi32(rt);
1861 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1862 vfitab = _mm_slli_epi32(vfitab,2);
1864 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1865 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1866 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1867 GMX_MM_TRANSPOSE2_PD(Y,F);
1868 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1869 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1870 GMX_MM_TRANSPOSE2_PD(G,H);
1871 Heps = _mm_mul_pd(vfeps,H);
1872 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1873 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1874 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq32,FF),_mm_mul_pd(vftabscale,rinv32)));
1878 /* Calculate temporary vectorial force */
1879 tx = _mm_mul_pd(fscal,dx32);
1880 ty = _mm_mul_pd(fscal,dy32);
1881 tz = _mm_mul_pd(fscal,dz32);
1883 /* Update vectorial force */
1884 fix3 = _mm_add_pd(fix3,tx);
1885 fiy3 = _mm_add_pd(fiy3,ty);
1886 fiz3 = _mm_add_pd(fiz3,tz);
1888 fjx2 = _mm_add_pd(fjx2,tx);
1889 fjy2 = _mm_add_pd(fjy2,ty);
1890 fjz2 = _mm_add_pd(fjz2,tz);
1892 /**************************
1893 * CALCULATE INTERACTIONS *
1894 **************************/
1896 r33 = _mm_mul_pd(rsq33,rinv33);
1898 /* Calculate table index by multiplying r with table scale and truncate to integer */
1899 rt = _mm_mul_pd(r33,vftabscale);
1900 vfitab = _mm_cvttpd_epi32(rt);
1901 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1902 vfitab = _mm_slli_epi32(vfitab,2);
1904 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1905 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1906 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1907 GMX_MM_TRANSPOSE2_PD(Y,F);
1908 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1909 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1910 GMX_MM_TRANSPOSE2_PD(G,H);
1911 Heps = _mm_mul_pd(vfeps,H);
1912 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1913 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1914 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq33,FF),_mm_mul_pd(vftabscale,rinv33)));
1918 /* Calculate temporary vectorial force */
1919 tx = _mm_mul_pd(fscal,dx33);
1920 ty = _mm_mul_pd(fscal,dy33);
1921 tz = _mm_mul_pd(fscal,dz33);
1923 /* Update vectorial force */
1924 fix3 = _mm_add_pd(fix3,tx);
1925 fiy3 = _mm_add_pd(fiy3,ty);
1926 fiz3 = _mm_add_pd(fiz3,tz);
1928 fjx3 = _mm_add_pd(fjx3,tx);
1929 fjy3 = _mm_add_pd(fjy3,ty);
1930 fjz3 = _mm_add_pd(fjz3,tz);
1932 gmx_mm_decrement_4rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1934 /* Inner loop uses 381 flops */
1937 if(jidx<j_index_end)
1941 j_coord_offsetA = DIM*jnrA;
1943 /* load j atom coordinates */
1944 gmx_mm_load_4rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
1945 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1946 &jy2,&jz2,&jx3,&jy3,&jz3);
1948 /* Calculate displacement vector */
1949 dx00 = _mm_sub_pd(ix0,jx0);
1950 dy00 = _mm_sub_pd(iy0,jy0);
1951 dz00 = _mm_sub_pd(iz0,jz0);
1952 dx11 = _mm_sub_pd(ix1,jx1);
1953 dy11 = _mm_sub_pd(iy1,jy1);
1954 dz11 = _mm_sub_pd(iz1,jz1);
1955 dx12 = _mm_sub_pd(ix1,jx2);
1956 dy12 = _mm_sub_pd(iy1,jy2);
1957 dz12 = _mm_sub_pd(iz1,jz2);
1958 dx13 = _mm_sub_pd(ix1,jx3);
1959 dy13 = _mm_sub_pd(iy1,jy3);
1960 dz13 = _mm_sub_pd(iz1,jz3);
1961 dx21 = _mm_sub_pd(ix2,jx1);
1962 dy21 = _mm_sub_pd(iy2,jy1);
1963 dz21 = _mm_sub_pd(iz2,jz1);
1964 dx22 = _mm_sub_pd(ix2,jx2);
1965 dy22 = _mm_sub_pd(iy2,jy2);
1966 dz22 = _mm_sub_pd(iz2,jz2);
1967 dx23 = _mm_sub_pd(ix2,jx3);
1968 dy23 = _mm_sub_pd(iy2,jy3);
1969 dz23 = _mm_sub_pd(iz2,jz3);
1970 dx31 = _mm_sub_pd(ix3,jx1);
1971 dy31 = _mm_sub_pd(iy3,jy1);
1972 dz31 = _mm_sub_pd(iz3,jz1);
1973 dx32 = _mm_sub_pd(ix3,jx2);
1974 dy32 = _mm_sub_pd(iy3,jy2);
1975 dz32 = _mm_sub_pd(iz3,jz2);
1976 dx33 = _mm_sub_pd(ix3,jx3);
1977 dy33 = _mm_sub_pd(iy3,jy3);
1978 dz33 = _mm_sub_pd(iz3,jz3);
1980 /* Calculate squared distance and things based on it */
1981 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1982 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1983 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1984 rsq13 = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
1985 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1986 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1987 rsq23 = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
1988 rsq31 = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
1989 rsq32 = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
1990 rsq33 = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
1992 rinv11 = gmx_mm_invsqrt_pd(rsq11);
1993 rinv12 = gmx_mm_invsqrt_pd(rsq12);
1994 rinv13 = gmx_mm_invsqrt_pd(rsq13);
1995 rinv21 = gmx_mm_invsqrt_pd(rsq21);
1996 rinv22 = gmx_mm_invsqrt_pd(rsq22);
1997 rinv23 = gmx_mm_invsqrt_pd(rsq23);
1998 rinv31 = gmx_mm_invsqrt_pd(rsq31);
1999 rinv32 = gmx_mm_invsqrt_pd(rsq32);
2000 rinv33 = gmx_mm_invsqrt_pd(rsq33);
2002 rinvsq00 = gmx_mm_inv_pd(rsq00);
2004 fjx0 = _mm_setzero_pd();
2005 fjy0 = _mm_setzero_pd();
2006 fjz0 = _mm_setzero_pd();
2007 fjx1 = _mm_setzero_pd();
2008 fjy1 = _mm_setzero_pd();
2009 fjz1 = _mm_setzero_pd();
2010 fjx2 = _mm_setzero_pd();
2011 fjy2 = _mm_setzero_pd();
2012 fjz2 = _mm_setzero_pd();
2013 fjx3 = _mm_setzero_pd();
2014 fjy3 = _mm_setzero_pd();
2015 fjz3 = _mm_setzero_pd();
2017 /**************************
2018 * CALCULATE INTERACTIONS *
2019 **************************/
2021 /* LENNARD-JONES DISPERSION/REPULSION */
2023 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
2024 fvdw = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
2028 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2030 /* Calculate temporary vectorial force */
2031 tx = _mm_mul_pd(fscal,dx00);
2032 ty = _mm_mul_pd(fscal,dy00);
2033 tz = _mm_mul_pd(fscal,dz00);
2035 /* Update vectorial force */
2036 fix0 = _mm_add_pd(fix0,tx);
2037 fiy0 = _mm_add_pd(fiy0,ty);
2038 fiz0 = _mm_add_pd(fiz0,tz);
2040 fjx0 = _mm_add_pd(fjx0,tx);
2041 fjy0 = _mm_add_pd(fjy0,ty);
2042 fjz0 = _mm_add_pd(fjz0,tz);
2044 /**************************
2045 * CALCULATE INTERACTIONS *
2046 **************************/
2048 r11 = _mm_mul_pd(rsq11,rinv11);
2050 /* Calculate table index by multiplying r with table scale and truncate to integer */
2051 rt = _mm_mul_pd(r11,vftabscale);
2052 vfitab = _mm_cvttpd_epi32(rt);
2053 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2054 vfitab = _mm_slli_epi32(vfitab,2);
2056 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2057 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2058 F = _mm_setzero_pd();
2059 GMX_MM_TRANSPOSE2_PD(Y,F);
2060 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2061 H = _mm_setzero_pd();
2062 GMX_MM_TRANSPOSE2_PD(G,H);
2063 Heps = _mm_mul_pd(vfeps,H);
2064 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2065 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2066 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
2070 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2072 /* Calculate temporary vectorial force */
2073 tx = _mm_mul_pd(fscal,dx11);
2074 ty = _mm_mul_pd(fscal,dy11);
2075 tz = _mm_mul_pd(fscal,dz11);
2077 /* Update vectorial force */
2078 fix1 = _mm_add_pd(fix1,tx);
2079 fiy1 = _mm_add_pd(fiy1,ty);
2080 fiz1 = _mm_add_pd(fiz1,tz);
2082 fjx1 = _mm_add_pd(fjx1,tx);
2083 fjy1 = _mm_add_pd(fjy1,ty);
2084 fjz1 = _mm_add_pd(fjz1,tz);
2086 /**************************
2087 * CALCULATE INTERACTIONS *
2088 **************************/
2090 r12 = _mm_mul_pd(rsq12,rinv12);
2092 /* Calculate table index by multiplying r with table scale and truncate to integer */
2093 rt = _mm_mul_pd(r12,vftabscale);
2094 vfitab = _mm_cvttpd_epi32(rt);
2095 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2096 vfitab = _mm_slli_epi32(vfitab,2);
2098 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2099 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2100 F = _mm_setzero_pd();
2101 GMX_MM_TRANSPOSE2_PD(Y,F);
2102 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2103 H = _mm_setzero_pd();
2104 GMX_MM_TRANSPOSE2_PD(G,H);
2105 Heps = _mm_mul_pd(vfeps,H);
2106 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2107 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2108 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
2112 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2114 /* Calculate temporary vectorial force */
2115 tx = _mm_mul_pd(fscal,dx12);
2116 ty = _mm_mul_pd(fscal,dy12);
2117 tz = _mm_mul_pd(fscal,dz12);
2119 /* Update vectorial force */
2120 fix1 = _mm_add_pd(fix1,tx);
2121 fiy1 = _mm_add_pd(fiy1,ty);
2122 fiz1 = _mm_add_pd(fiz1,tz);
2124 fjx2 = _mm_add_pd(fjx2,tx);
2125 fjy2 = _mm_add_pd(fjy2,ty);
2126 fjz2 = _mm_add_pd(fjz2,tz);
2128 /**************************
2129 * CALCULATE INTERACTIONS *
2130 **************************/
2132 r13 = _mm_mul_pd(rsq13,rinv13);
2134 /* Calculate table index by multiplying r with table scale and truncate to integer */
2135 rt = _mm_mul_pd(r13,vftabscale);
2136 vfitab = _mm_cvttpd_epi32(rt);
2137 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2138 vfitab = _mm_slli_epi32(vfitab,2);
2140 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2141 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2142 F = _mm_setzero_pd();
2143 GMX_MM_TRANSPOSE2_PD(Y,F);
2144 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2145 H = _mm_setzero_pd();
2146 GMX_MM_TRANSPOSE2_PD(G,H);
2147 Heps = _mm_mul_pd(vfeps,H);
2148 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2149 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2150 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq13,FF),_mm_mul_pd(vftabscale,rinv13)));
2154 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2156 /* Calculate temporary vectorial force */
2157 tx = _mm_mul_pd(fscal,dx13);
2158 ty = _mm_mul_pd(fscal,dy13);
2159 tz = _mm_mul_pd(fscal,dz13);
2161 /* Update vectorial force */
2162 fix1 = _mm_add_pd(fix1,tx);
2163 fiy1 = _mm_add_pd(fiy1,ty);
2164 fiz1 = _mm_add_pd(fiz1,tz);
2166 fjx3 = _mm_add_pd(fjx3,tx);
2167 fjy3 = _mm_add_pd(fjy3,ty);
2168 fjz3 = _mm_add_pd(fjz3,tz);
2170 /**************************
2171 * CALCULATE INTERACTIONS *
2172 **************************/
2174 r21 = _mm_mul_pd(rsq21,rinv21);
2176 /* Calculate table index by multiplying r with table scale and truncate to integer */
2177 rt = _mm_mul_pd(r21,vftabscale);
2178 vfitab = _mm_cvttpd_epi32(rt);
2179 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2180 vfitab = _mm_slli_epi32(vfitab,2);
2182 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2183 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2184 F = _mm_setzero_pd();
2185 GMX_MM_TRANSPOSE2_PD(Y,F);
2186 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2187 H = _mm_setzero_pd();
2188 GMX_MM_TRANSPOSE2_PD(G,H);
2189 Heps = _mm_mul_pd(vfeps,H);
2190 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2191 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2192 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
2196 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2198 /* Calculate temporary vectorial force */
2199 tx = _mm_mul_pd(fscal,dx21);
2200 ty = _mm_mul_pd(fscal,dy21);
2201 tz = _mm_mul_pd(fscal,dz21);
2203 /* Update vectorial force */
2204 fix2 = _mm_add_pd(fix2,tx);
2205 fiy2 = _mm_add_pd(fiy2,ty);
2206 fiz2 = _mm_add_pd(fiz2,tz);
2208 fjx1 = _mm_add_pd(fjx1,tx);
2209 fjy1 = _mm_add_pd(fjy1,ty);
2210 fjz1 = _mm_add_pd(fjz1,tz);
2212 /**************************
2213 * CALCULATE INTERACTIONS *
2214 **************************/
2216 r22 = _mm_mul_pd(rsq22,rinv22);
2218 /* Calculate table index by multiplying r with table scale and truncate to integer */
2219 rt = _mm_mul_pd(r22,vftabscale);
2220 vfitab = _mm_cvttpd_epi32(rt);
2221 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2222 vfitab = _mm_slli_epi32(vfitab,2);
2224 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2225 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2226 F = _mm_setzero_pd();
2227 GMX_MM_TRANSPOSE2_PD(Y,F);
2228 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2229 H = _mm_setzero_pd();
2230 GMX_MM_TRANSPOSE2_PD(G,H);
2231 Heps = _mm_mul_pd(vfeps,H);
2232 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2233 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2234 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
2238 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2240 /* Calculate temporary vectorial force */
2241 tx = _mm_mul_pd(fscal,dx22);
2242 ty = _mm_mul_pd(fscal,dy22);
2243 tz = _mm_mul_pd(fscal,dz22);
2245 /* Update vectorial force */
2246 fix2 = _mm_add_pd(fix2,tx);
2247 fiy2 = _mm_add_pd(fiy2,ty);
2248 fiz2 = _mm_add_pd(fiz2,tz);
2250 fjx2 = _mm_add_pd(fjx2,tx);
2251 fjy2 = _mm_add_pd(fjy2,ty);
2252 fjz2 = _mm_add_pd(fjz2,tz);
2254 /**************************
2255 * CALCULATE INTERACTIONS *
2256 **************************/
2258 r23 = _mm_mul_pd(rsq23,rinv23);
2260 /* Calculate table index by multiplying r with table scale and truncate to integer */
2261 rt = _mm_mul_pd(r23,vftabscale);
2262 vfitab = _mm_cvttpd_epi32(rt);
2263 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2264 vfitab = _mm_slli_epi32(vfitab,2);
2266 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2267 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2268 F = _mm_setzero_pd();
2269 GMX_MM_TRANSPOSE2_PD(Y,F);
2270 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2271 H = _mm_setzero_pd();
2272 GMX_MM_TRANSPOSE2_PD(G,H);
2273 Heps = _mm_mul_pd(vfeps,H);
2274 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2275 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2276 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq23,FF),_mm_mul_pd(vftabscale,rinv23)));
2280 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2282 /* Calculate temporary vectorial force */
2283 tx = _mm_mul_pd(fscal,dx23);
2284 ty = _mm_mul_pd(fscal,dy23);
2285 tz = _mm_mul_pd(fscal,dz23);
2287 /* Update vectorial force */
2288 fix2 = _mm_add_pd(fix2,tx);
2289 fiy2 = _mm_add_pd(fiy2,ty);
2290 fiz2 = _mm_add_pd(fiz2,tz);
2292 fjx3 = _mm_add_pd(fjx3,tx);
2293 fjy3 = _mm_add_pd(fjy3,ty);
2294 fjz3 = _mm_add_pd(fjz3,tz);
2296 /**************************
2297 * CALCULATE INTERACTIONS *
2298 **************************/
2300 r31 = _mm_mul_pd(rsq31,rinv31);
2302 /* Calculate table index by multiplying r with table scale and truncate to integer */
2303 rt = _mm_mul_pd(r31,vftabscale);
2304 vfitab = _mm_cvttpd_epi32(rt);
2305 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2306 vfitab = _mm_slli_epi32(vfitab,2);
2308 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2309 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2310 F = _mm_setzero_pd();
2311 GMX_MM_TRANSPOSE2_PD(Y,F);
2312 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2313 H = _mm_setzero_pd();
2314 GMX_MM_TRANSPOSE2_PD(G,H);
2315 Heps = _mm_mul_pd(vfeps,H);
2316 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2317 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2318 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq31,FF),_mm_mul_pd(vftabscale,rinv31)));
2322 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2324 /* Calculate temporary vectorial force */
2325 tx = _mm_mul_pd(fscal,dx31);
2326 ty = _mm_mul_pd(fscal,dy31);
2327 tz = _mm_mul_pd(fscal,dz31);
2329 /* Update vectorial force */
2330 fix3 = _mm_add_pd(fix3,tx);
2331 fiy3 = _mm_add_pd(fiy3,ty);
2332 fiz3 = _mm_add_pd(fiz3,tz);
2334 fjx1 = _mm_add_pd(fjx1,tx);
2335 fjy1 = _mm_add_pd(fjy1,ty);
2336 fjz1 = _mm_add_pd(fjz1,tz);
2338 /**************************
2339 * CALCULATE INTERACTIONS *
2340 **************************/
2342 r32 = _mm_mul_pd(rsq32,rinv32);
2344 /* Calculate table index by multiplying r with table scale and truncate to integer */
2345 rt = _mm_mul_pd(r32,vftabscale);
2346 vfitab = _mm_cvttpd_epi32(rt);
2347 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2348 vfitab = _mm_slli_epi32(vfitab,2);
2350 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2351 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2352 F = _mm_setzero_pd();
2353 GMX_MM_TRANSPOSE2_PD(Y,F);
2354 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2355 H = _mm_setzero_pd();
2356 GMX_MM_TRANSPOSE2_PD(G,H);
2357 Heps = _mm_mul_pd(vfeps,H);
2358 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2359 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2360 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq32,FF),_mm_mul_pd(vftabscale,rinv32)));
2364 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2366 /* Calculate temporary vectorial force */
2367 tx = _mm_mul_pd(fscal,dx32);
2368 ty = _mm_mul_pd(fscal,dy32);
2369 tz = _mm_mul_pd(fscal,dz32);
2371 /* Update vectorial force */
2372 fix3 = _mm_add_pd(fix3,tx);
2373 fiy3 = _mm_add_pd(fiy3,ty);
2374 fiz3 = _mm_add_pd(fiz3,tz);
2376 fjx2 = _mm_add_pd(fjx2,tx);
2377 fjy2 = _mm_add_pd(fjy2,ty);
2378 fjz2 = _mm_add_pd(fjz2,tz);
2380 /**************************
2381 * CALCULATE INTERACTIONS *
2382 **************************/
2384 r33 = _mm_mul_pd(rsq33,rinv33);
2386 /* Calculate table index by multiplying r with table scale and truncate to integer */
2387 rt = _mm_mul_pd(r33,vftabscale);
2388 vfitab = _mm_cvttpd_epi32(rt);
2389 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2390 vfitab = _mm_slli_epi32(vfitab,2);
2392 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2393 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2394 F = _mm_setzero_pd();
2395 GMX_MM_TRANSPOSE2_PD(Y,F);
2396 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2397 H = _mm_setzero_pd();
2398 GMX_MM_TRANSPOSE2_PD(G,H);
2399 Heps = _mm_mul_pd(vfeps,H);
2400 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2401 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2402 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq33,FF),_mm_mul_pd(vftabscale,rinv33)));
2406 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2408 /* Calculate temporary vectorial force */
2409 tx = _mm_mul_pd(fscal,dx33);
2410 ty = _mm_mul_pd(fscal,dy33);
2411 tz = _mm_mul_pd(fscal,dz33);
2413 /* Update vectorial force */
2414 fix3 = _mm_add_pd(fix3,tx);
2415 fiy3 = _mm_add_pd(fiy3,ty);
2416 fiz3 = _mm_add_pd(fiz3,tz);
2418 fjx3 = _mm_add_pd(fjx3,tx);
2419 fjy3 = _mm_add_pd(fjy3,ty);
2420 fjz3 = _mm_add_pd(fjz3,tz);
2422 gmx_mm_decrement_4rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2424 /* Inner loop uses 381 flops */
2427 /* End of innermost loop */
2429 gmx_mm_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2430 f+i_coord_offset,fshift+i_shift_offset);
2432 /* Increment number of inner iterations */
2433 inneriter += j_index_end - j_index_start;
2435 /* Outer loop uses 24 flops */
2438 /* Increment number of outer iterations */
2441 /* Update outer/inner flops */
2443 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*381);