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.
42 #include "../nb_kernel.h"
43 #include "types/simple.h"
44 #include "gromacs/math/vec.h"
47 #include "gromacs/simd/math_x86_sse2_double.h"
48 #include "kernelutil_x86_sse2_double.h"
51 * Gromacs nonbonded kernel: nb_kernel_ElecCSTab_VdwCSTab_GeomW4W4_VF_sse2_double
52 * Electrostatics interaction: CubicSplineTable
53 * VdW interaction: CubicSplineTable
54 * Geometry: Water4-Water4
55 * Calculate force/pot: PotentialAndForce
58 nb_kernel_ElecCSTab_VdwCSTab_GeomW4W4_VF_sse2_double
59 (t_nblist * gmx_restrict nlist,
60 rvec * gmx_restrict xx,
61 rvec * gmx_restrict ff,
62 t_forcerec * gmx_restrict fr,
63 t_mdatoms * gmx_restrict mdatoms,
64 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
65 t_nrnb * gmx_restrict nrnb)
67 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
68 * just 0 for non-waters.
69 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
70 * jnr indices corresponding to data put in the four positions in the SIMD register.
72 int i_shift_offset,i_coord_offset,outeriter,inneriter;
73 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
75 int j_coord_offsetA,j_coord_offsetB;
76 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
78 real *shiftvec,*fshift,*x,*f;
79 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
81 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
83 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
85 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
87 __m128d ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
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 int vdwjidx3A,vdwjidx3B;
95 __m128d jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
96 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
97 __m128d dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
98 __m128d dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
99 __m128d dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
100 __m128d dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
101 __m128d dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
102 __m128d dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
103 __m128d dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
104 __m128d dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
105 __m128d dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
106 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
109 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
112 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
113 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
115 __m128i ifour = _mm_set1_epi32(4);
116 __m128d rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
118 __m128d dummy_mask,cutoff_mask;
119 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
120 __m128d one = _mm_set1_pd(1.0);
121 __m128d two = _mm_set1_pd(2.0);
127 jindex = nlist->jindex;
129 shiftidx = nlist->shift;
131 shiftvec = fr->shift_vec[0];
132 fshift = fr->fshift[0];
133 facel = _mm_set1_pd(fr->epsfac);
134 charge = mdatoms->chargeA;
135 nvdwtype = fr->ntype;
137 vdwtype = mdatoms->typeA;
139 vftab = kernel_data->table_elec_vdw->data;
140 vftabscale = _mm_set1_pd(kernel_data->table_elec_vdw->scale);
142 /* Setup water-specific parameters */
143 inr = nlist->iinr[0];
144 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
145 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
146 iq3 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
147 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
149 jq1 = _mm_set1_pd(charge[inr+1]);
150 jq2 = _mm_set1_pd(charge[inr+2]);
151 jq3 = _mm_set1_pd(charge[inr+3]);
152 vdwjidx0A = 2*vdwtype[inr+0];
153 c6_00 = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
154 c12_00 = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
155 qq11 = _mm_mul_pd(iq1,jq1);
156 qq12 = _mm_mul_pd(iq1,jq2);
157 qq13 = _mm_mul_pd(iq1,jq3);
158 qq21 = _mm_mul_pd(iq2,jq1);
159 qq22 = _mm_mul_pd(iq2,jq2);
160 qq23 = _mm_mul_pd(iq2,jq3);
161 qq31 = _mm_mul_pd(iq3,jq1);
162 qq32 = _mm_mul_pd(iq3,jq2);
163 qq33 = _mm_mul_pd(iq3,jq3);
165 /* Avoid stupid compiler warnings */
173 /* Start outer loop over neighborlists */
174 for(iidx=0; iidx<nri; iidx++)
176 /* Load shift vector for this list */
177 i_shift_offset = DIM*shiftidx[iidx];
179 /* Load limits for loop over neighbors */
180 j_index_start = jindex[iidx];
181 j_index_end = jindex[iidx+1];
183 /* Get outer coordinate index */
185 i_coord_offset = DIM*inr;
187 /* Load i particle coords and add shift vector */
188 gmx_mm_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
189 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
191 fix0 = _mm_setzero_pd();
192 fiy0 = _mm_setzero_pd();
193 fiz0 = _mm_setzero_pd();
194 fix1 = _mm_setzero_pd();
195 fiy1 = _mm_setzero_pd();
196 fiz1 = _mm_setzero_pd();
197 fix2 = _mm_setzero_pd();
198 fiy2 = _mm_setzero_pd();
199 fiz2 = _mm_setzero_pd();
200 fix3 = _mm_setzero_pd();
201 fiy3 = _mm_setzero_pd();
202 fiz3 = _mm_setzero_pd();
204 /* Reset potential sums */
205 velecsum = _mm_setzero_pd();
206 vvdwsum = _mm_setzero_pd();
208 /* Start inner kernel loop */
209 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
212 /* Get j neighbor index, and coordinate index */
215 j_coord_offsetA = DIM*jnrA;
216 j_coord_offsetB = DIM*jnrB;
218 /* load j atom coordinates */
219 gmx_mm_load_4rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
220 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
221 &jy2,&jz2,&jx3,&jy3,&jz3);
223 /* Calculate displacement vector */
224 dx00 = _mm_sub_pd(ix0,jx0);
225 dy00 = _mm_sub_pd(iy0,jy0);
226 dz00 = _mm_sub_pd(iz0,jz0);
227 dx11 = _mm_sub_pd(ix1,jx1);
228 dy11 = _mm_sub_pd(iy1,jy1);
229 dz11 = _mm_sub_pd(iz1,jz1);
230 dx12 = _mm_sub_pd(ix1,jx2);
231 dy12 = _mm_sub_pd(iy1,jy2);
232 dz12 = _mm_sub_pd(iz1,jz2);
233 dx13 = _mm_sub_pd(ix1,jx3);
234 dy13 = _mm_sub_pd(iy1,jy3);
235 dz13 = _mm_sub_pd(iz1,jz3);
236 dx21 = _mm_sub_pd(ix2,jx1);
237 dy21 = _mm_sub_pd(iy2,jy1);
238 dz21 = _mm_sub_pd(iz2,jz1);
239 dx22 = _mm_sub_pd(ix2,jx2);
240 dy22 = _mm_sub_pd(iy2,jy2);
241 dz22 = _mm_sub_pd(iz2,jz2);
242 dx23 = _mm_sub_pd(ix2,jx3);
243 dy23 = _mm_sub_pd(iy2,jy3);
244 dz23 = _mm_sub_pd(iz2,jz3);
245 dx31 = _mm_sub_pd(ix3,jx1);
246 dy31 = _mm_sub_pd(iy3,jy1);
247 dz31 = _mm_sub_pd(iz3,jz1);
248 dx32 = _mm_sub_pd(ix3,jx2);
249 dy32 = _mm_sub_pd(iy3,jy2);
250 dz32 = _mm_sub_pd(iz3,jz2);
251 dx33 = _mm_sub_pd(ix3,jx3);
252 dy33 = _mm_sub_pd(iy3,jy3);
253 dz33 = _mm_sub_pd(iz3,jz3);
255 /* Calculate squared distance and things based on it */
256 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
257 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
258 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
259 rsq13 = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
260 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
261 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
262 rsq23 = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
263 rsq31 = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
264 rsq32 = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
265 rsq33 = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
267 rinv00 = gmx_mm_invsqrt_pd(rsq00);
268 rinv11 = gmx_mm_invsqrt_pd(rsq11);
269 rinv12 = gmx_mm_invsqrt_pd(rsq12);
270 rinv13 = gmx_mm_invsqrt_pd(rsq13);
271 rinv21 = gmx_mm_invsqrt_pd(rsq21);
272 rinv22 = gmx_mm_invsqrt_pd(rsq22);
273 rinv23 = gmx_mm_invsqrt_pd(rsq23);
274 rinv31 = gmx_mm_invsqrt_pd(rsq31);
275 rinv32 = gmx_mm_invsqrt_pd(rsq32);
276 rinv33 = gmx_mm_invsqrt_pd(rsq33);
278 fjx0 = _mm_setzero_pd();
279 fjy0 = _mm_setzero_pd();
280 fjz0 = _mm_setzero_pd();
281 fjx1 = _mm_setzero_pd();
282 fjy1 = _mm_setzero_pd();
283 fjz1 = _mm_setzero_pd();
284 fjx2 = _mm_setzero_pd();
285 fjy2 = _mm_setzero_pd();
286 fjz2 = _mm_setzero_pd();
287 fjx3 = _mm_setzero_pd();
288 fjy3 = _mm_setzero_pd();
289 fjz3 = _mm_setzero_pd();
291 /**************************
292 * CALCULATE INTERACTIONS *
293 **************************/
295 r00 = _mm_mul_pd(rsq00,rinv00);
297 /* Calculate table index by multiplying r with table scale and truncate to integer */
298 rt = _mm_mul_pd(r00,vftabscale);
299 vfitab = _mm_cvttpd_epi32(rt);
300 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
301 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
303 /* CUBIC SPLINE TABLE DISPERSION */
304 vfitab = _mm_add_epi32(vfitab,ifour);
305 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
306 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
307 GMX_MM_TRANSPOSE2_PD(Y,F);
308 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
309 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
310 GMX_MM_TRANSPOSE2_PD(G,H);
311 Heps = _mm_mul_pd(vfeps,H);
312 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
313 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
314 vvdw6 = _mm_mul_pd(c6_00,VV);
315 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
316 fvdw6 = _mm_mul_pd(c6_00,FF);
318 /* CUBIC SPLINE TABLE REPULSION */
319 vfitab = _mm_add_epi32(vfitab,ifour);
320 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
321 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
322 GMX_MM_TRANSPOSE2_PD(Y,F);
323 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
324 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
325 GMX_MM_TRANSPOSE2_PD(G,H);
326 Heps = _mm_mul_pd(vfeps,H);
327 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
328 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
329 vvdw12 = _mm_mul_pd(c12_00,VV);
330 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
331 fvdw12 = _mm_mul_pd(c12_00,FF);
332 vvdw = _mm_add_pd(vvdw12,vvdw6);
333 fvdw = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
335 /* Update potential sum for this i atom from the interaction with this j atom. */
336 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
340 /* Calculate temporary vectorial force */
341 tx = _mm_mul_pd(fscal,dx00);
342 ty = _mm_mul_pd(fscal,dy00);
343 tz = _mm_mul_pd(fscal,dz00);
345 /* Update vectorial force */
346 fix0 = _mm_add_pd(fix0,tx);
347 fiy0 = _mm_add_pd(fiy0,ty);
348 fiz0 = _mm_add_pd(fiz0,tz);
350 fjx0 = _mm_add_pd(fjx0,tx);
351 fjy0 = _mm_add_pd(fjy0,ty);
352 fjz0 = _mm_add_pd(fjz0,tz);
354 /**************************
355 * CALCULATE INTERACTIONS *
356 **************************/
358 r11 = _mm_mul_pd(rsq11,rinv11);
360 /* Calculate table index by multiplying r with table scale and truncate to integer */
361 rt = _mm_mul_pd(r11,vftabscale);
362 vfitab = _mm_cvttpd_epi32(rt);
363 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
364 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
366 /* CUBIC SPLINE TABLE ELECTROSTATICS */
367 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
368 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
369 GMX_MM_TRANSPOSE2_PD(Y,F);
370 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
371 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
372 GMX_MM_TRANSPOSE2_PD(G,H);
373 Heps = _mm_mul_pd(vfeps,H);
374 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
375 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
376 velec = _mm_mul_pd(qq11,VV);
377 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
378 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
380 /* Update potential sum for this i atom from the interaction with this j atom. */
381 velecsum = _mm_add_pd(velecsum,velec);
385 /* Calculate temporary vectorial force */
386 tx = _mm_mul_pd(fscal,dx11);
387 ty = _mm_mul_pd(fscal,dy11);
388 tz = _mm_mul_pd(fscal,dz11);
390 /* Update vectorial force */
391 fix1 = _mm_add_pd(fix1,tx);
392 fiy1 = _mm_add_pd(fiy1,ty);
393 fiz1 = _mm_add_pd(fiz1,tz);
395 fjx1 = _mm_add_pd(fjx1,tx);
396 fjy1 = _mm_add_pd(fjy1,ty);
397 fjz1 = _mm_add_pd(fjz1,tz);
399 /**************************
400 * CALCULATE INTERACTIONS *
401 **************************/
403 r12 = _mm_mul_pd(rsq12,rinv12);
405 /* Calculate table index by multiplying r with table scale and truncate to integer */
406 rt = _mm_mul_pd(r12,vftabscale);
407 vfitab = _mm_cvttpd_epi32(rt);
408 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
409 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
411 /* CUBIC SPLINE TABLE ELECTROSTATICS */
412 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
413 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
414 GMX_MM_TRANSPOSE2_PD(Y,F);
415 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
416 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
417 GMX_MM_TRANSPOSE2_PD(G,H);
418 Heps = _mm_mul_pd(vfeps,H);
419 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
420 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
421 velec = _mm_mul_pd(qq12,VV);
422 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
423 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
425 /* Update potential sum for this i atom from the interaction with this j atom. */
426 velecsum = _mm_add_pd(velecsum,velec);
430 /* Calculate temporary vectorial force */
431 tx = _mm_mul_pd(fscal,dx12);
432 ty = _mm_mul_pd(fscal,dy12);
433 tz = _mm_mul_pd(fscal,dz12);
435 /* Update vectorial force */
436 fix1 = _mm_add_pd(fix1,tx);
437 fiy1 = _mm_add_pd(fiy1,ty);
438 fiz1 = _mm_add_pd(fiz1,tz);
440 fjx2 = _mm_add_pd(fjx2,tx);
441 fjy2 = _mm_add_pd(fjy2,ty);
442 fjz2 = _mm_add_pd(fjz2,tz);
444 /**************************
445 * CALCULATE INTERACTIONS *
446 **************************/
448 r13 = _mm_mul_pd(rsq13,rinv13);
450 /* Calculate table index by multiplying r with table scale and truncate to integer */
451 rt = _mm_mul_pd(r13,vftabscale);
452 vfitab = _mm_cvttpd_epi32(rt);
453 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
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 + gmx_mm_extract_epi32(vfitab,0) );
458 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
459 GMX_MM_TRANSPOSE2_PD(Y,F);
460 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
461 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
462 GMX_MM_TRANSPOSE2_PD(G,H);
463 Heps = _mm_mul_pd(vfeps,H);
464 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
465 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
466 velec = _mm_mul_pd(qq13,VV);
467 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
468 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq13,FF),_mm_mul_pd(vftabscale,rinv13)));
470 /* Update potential sum for this i atom from the interaction with this j atom. */
471 velecsum = _mm_add_pd(velecsum,velec);
475 /* Calculate temporary vectorial force */
476 tx = _mm_mul_pd(fscal,dx13);
477 ty = _mm_mul_pd(fscal,dy13);
478 tz = _mm_mul_pd(fscal,dz13);
480 /* Update vectorial force */
481 fix1 = _mm_add_pd(fix1,tx);
482 fiy1 = _mm_add_pd(fiy1,ty);
483 fiz1 = _mm_add_pd(fiz1,tz);
485 fjx3 = _mm_add_pd(fjx3,tx);
486 fjy3 = _mm_add_pd(fjy3,ty);
487 fjz3 = _mm_add_pd(fjz3,tz);
489 /**************************
490 * CALCULATE INTERACTIONS *
491 **************************/
493 r21 = _mm_mul_pd(rsq21,rinv21);
495 /* Calculate table index by multiplying r with table scale and truncate to integer */
496 rt = _mm_mul_pd(r21,vftabscale);
497 vfitab = _mm_cvttpd_epi32(rt);
498 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
499 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
501 /* CUBIC SPLINE TABLE ELECTROSTATICS */
502 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
503 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
504 GMX_MM_TRANSPOSE2_PD(Y,F);
505 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
506 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
507 GMX_MM_TRANSPOSE2_PD(G,H);
508 Heps = _mm_mul_pd(vfeps,H);
509 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
510 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
511 velec = _mm_mul_pd(qq21,VV);
512 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
513 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
515 /* Update potential sum for this i atom from the interaction with this j atom. */
516 velecsum = _mm_add_pd(velecsum,velec);
520 /* Calculate temporary vectorial force */
521 tx = _mm_mul_pd(fscal,dx21);
522 ty = _mm_mul_pd(fscal,dy21);
523 tz = _mm_mul_pd(fscal,dz21);
525 /* Update vectorial force */
526 fix2 = _mm_add_pd(fix2,tx);
527 fiy2 = _mm_add_pd(fiy2,ty);
528 fiz2 = _mm_add_pd(fiz2,tz);
530 fjx1 = _mm_add_pd(fjx1,tx);
531 fjy1 = _mm_add_pd(fjy1,ty);
532 fjz1 = _mm_add_pd(fjz1,tz);
534 /**************************
535 * CALCULATE INTERACTIONS *
536 **************************/
538 r22 = _mm_mul_pd(rsq22,rinv22);
540 /* Calculate table index by multiplying r with table scale and truncate to integer */
541 rt = _mm_mul_pd(r22,vftabscale);
542 vfitab = _mm_cvttpd_epi32(rt);
543 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
544 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
546 /* CUBIC SPLINE TABLE ELECTROSTATICS */
547 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
548 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
549 GMX_MM_TRANSPOSE2_PD(Y,F);
550 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
551 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
552 GMX_MM_TRANSPOSE2_PD(G,H);
553 Heps = _mm_mul_pd(vfeps,H);
554 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
555 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
556 velec = _mm_mul_pd(qq22,VV);
557 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
558 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
560 /* Update potential sum for this i atom from the interaction with this j atom. */
561 velecsum = _mm_add_pd(velecsum,velec);
565 /* Calculate temporary vectorial force */
566 tx = _mm_mul_pd(fscal,dx22);
567 ty = _mm_mul_pd(fscal,dy22);
568 tz = _mm_mul_pd(fscal,dz22);
570 /* Update vectorial force */
571 fix2 = _mm_add_pd(fix2,tx);
572 fiy2 = _mm_add_pd(fiy2,ty);
573 fiz2 = _mm_add_pd(fiz2,tz);
575 fjx2 = _mm_add_pd(fjx2,tx);
576 fjy2 = _mm_add_pd(fjy2,ty);
577 fjz2 = _mm_add_pd(fjz2,tz);
579 /**************************
580 * CALCULATE INTERACTIONS *
581 **************************/
583 r23 = _mm_mul_pd(rsq23,rinv23);
585 /* Calculate table index by multiplying r with table scale and truncate to integer */
586 rt = _mm_mul_pd(r23,vftabscale);
587 vfitab = _mm_cvttpd_epi32(rt);
588 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
589 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
591 /* CUBIC SPLINE TABLE ELECTROSTATICS */
592 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
593 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
594 GMX_MM_TRANSPOSE2_PD(Y,F);
595 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
596 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
597 GMX_MM_TRANSPOSE2_PD(G,H);
598 Heps = _mm_mul_pd(vfeps,H);
599 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
600 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
601 velec = _mm_mul_pd(qq23,VV);
602 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
603 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq23,FF),_mm_mul_pd(vftabscale,rinv23)));
605 /* Update potential sum for this i atom from the interaction with this j atom. */
606 velecsum = _mm_add_pd(velecsum,velec);
610 /* Calculate temporary vectorial force */
611 tx = _mm_mul_pd(fscal,dx23);
612 ty = _mm_mul_pd(fscal,dy23);
613 tz = _mm_mul_pd(fscal,dz23);
615 /* Update vectorial force */
616 fix2 = _mm_add_pd(fix2,tx);
617 fiy2 = _mm_add_pd(fiy2,ty);
618 fiz2 = _mm_add_pd(fiz2,tz);
620 fjx3 = _mm_add_pd(fjx3,tx);
621 fjy3 = _mm_add_pd(fjy3,ty);
622 fjz3 = _mm_add_pd(fjz3,tz);
624 /**************************
625 * CALCULATE INTERACTIONS *
626 **************************/
628 r31 = _mm_mul_pd(rsq31,rinv31);
630 /* Calculate table index by multiplying r with table scale and truncate to integer */
631 rt = _mm_mul_pd(r31,vftabscale);
632 vfitab = _mm_cvttpd_epi32(rt);
633 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
634 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
636 /* CUBIC SPLINE TABLE ELECTROSTATICS */
637 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
638 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
639 GMX_MM_TRANSPOSE2_PD(Y,F);
640 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
641 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
642 GMX_MM_TRANSPOSE2_PD(G,H);
643 Heps = _mm_mul_pd(vfeps,H);
644 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
645 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
646 velec = _mm_mul_pd(qq31,VV);
647 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
648 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq31,FF),_mm_mul_pd(vftabscale,rinv31)));
650 /* Update potential sum for this i atom from the interaction with this j atom. */
651 velecsum = _mm_add_pd(velecsum,velec);
655 /* Calculate temporary vectorial force */
656 tx = _mm_mul_pd(fscal,dx31);
657 ty = _mm_mul_pd(fscal,dy31);
658 tz = _mm_mul_pd(fscal,dz31);
660 /* Update vectorial force */
661 fix3 = _mm_add_pd(fix3,tx);
662 fiy3 = _mm_add_pd(fiy3,ty);
663 fiz3 = _mm_add_pd(fiz3,tz);
665 fjx1 = _mm_add_pd(fjx1,tx);
666 fjy1 = _mm_add_pd(fjy1,ty);
667 fjz1 = _mm_add_pd(fjz1,tz);
669 /**************************
670 * CALCULATE INTERACTIONS *
671 **************************/
673 r32 = _mm_mul_pd(rsq32,rinv32);
675 /* Calculate table index by multiplying r with table scale and truncate to integer */
676 rt = _mm_mul_pd(r32,vftabscale);
677 vfitab = _mm_cvttpd_epi32(rt);
678 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
679 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
681 /* CUBIC SPLINE TABLE ELECTROSTATICS */
682 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
683 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
684 GMX_MM_TRANSPOSE2_PD(Y,F);
685 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
686 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
687 GMX_MM_TRANSPOSE2_PD(G,H);
688 Heps = _mm_mul_pd(vfeps,H);
689 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
690 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
691 velec = _mm_mul_pd(qq32,VV);
692 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
693 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq32,FF),_mm_mul_pd(vftabscale,rinv32)));
695 /* Update potential sum for this i atom from the interaction with this j atom. */
696 velecsum = _mm_add_pd(velecsum,velec);
700 /* Calculate temporary vectorial force */
701 tx = _mm_mul_pd(fscal,dx32);
702 ty = _mm_mul_pd(fscal,dy32);
703 tz = _mm_mul_pd(fscal,dz32);
705 /* Update vectorial force */
706 fix3 = _mm_add_pd(fix3,tx);
707 fiy3 = _mm_add_pd(fiy3,ty);
708 fiz3 = _mm_add_pd(fiz3,tz);
710 fjx2 = _mm_add_pd(fjx2,tx);
711 fjy2 = _mm_add_pd(fjy2,ty);
712 fjz2 = _mm_add_pd(fjz2,tz);
714 /**************************
715 * CALCULATE INTERACTIONS *
716 **************************/
718 r33 = _mm_mul_pd(rsq33,rinv33);
720 /* Calculate table index by multiplying r with table scale and truncate to integer */
721 rt = _mm_mul_pd(r33,vftabscale);
722 vfitab = _mm_cvttpd_epi32(rt);
723 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
724 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
726 /* CUBIC SPLINE TABLE ELECTROSTATICS */
727 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
728 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
729 GMX_MM_TRANSPOSE2_PD(Y,F);
730 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
731 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
732 GMX_MM_TRANSPOSE2_PD(G,H);
733 Heps = _mm_mul_pd(vfeps,H);
734 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
735 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
736 velec = _mm_mul_pd(qq33,VV);
737 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
738 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq33,FF),_mm_mul_pd(vftabscale,rinv33)));
740 /* Update potential sum for this i atom from the interaction with this j atom. */
741 velecsum = _mm_add_pd(velecsum,velec);
745 /* Calculate temporary vectorial force */
746 tx = _mm_mul_pd(fscal,dx33);
747 ty = _mm_mul_pd(fscal,dy33);
748 tz = _mm_mul_pd(fscal,dz33);
750 /* Update vectorial force */
751 fix3 = _mm_add_pd(fix3,tx);
752 fiy3 = _mm_add_pd(fiy3,ty);
753 fiz3 = _mm_add_pd(fiz3,tz);
755 fjx3 = _mm_add_pd(fjx3,tx);
756 fjy3 = _mm_add_pd(fjy3,ty);
757 fjz3 = _mm_add_pd(fjz3,tz);
759 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);
761 /* Inner loop uses 446 flops */
768 j_coord_offsetA = DIM*jnrA;
770 /* load j atom coordinates */
771 gmx_mm_load_4rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
772 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
773 &jy2,&jz2,&jx3,&jy3,&jz3);
775 /* Calculate displacement vector */
776 dx00 = _mm_sub_pd(ix0,jx0);
777 dy00 = _mm_sub_pd(iy0,jy0);
778 dz00 = _mm_sub_pd(iz0,jz0);
779 dx11 = _mm_sub_pd(ix1,jx1);
780 dy11 = _mm_sub_pd(iy1,jy1);
781 dz11 = _mm_sub_pd(iz1,jz1);
782 dx12 = _mm_sub_pd(ix1,jx2);
783 dy12 = _mm_sub_pd(iy1,jy2);
784 dz12 = _mm_sub_pd(iz1,jz2);
785 dx13 = _mm_sub_pd(ix1,jx3);
786 dy13 = _mm_sub_pd(iy1,jy3);
787 dz13 = _mm_sub_pd(iz1,jz3);
788 dx21 = _mm_sub_pd(ix2,jx1);
789 dy21 = _mm_sub_pd(iy2,jy1);
790 dz21 = _mm_sub_pd(iz2,jz1);
791 dx22 = _mm_sub_pd(ix2,jx2);
792 dy22 = _mm_sub_pd(iy2,jy2);
793 dz22 = _mm_sub_pd(iz2,jz2);
794 dx23 = _mm_sub_pd(ix2,jx3);
795 dy23 = _mm_sub_pd(iy2,jy3);
796 dz23 = _mm_sub_pd(iz2,jz3);
797 dx31 = _mm_sub_pd(ix3,jx1);
798 dy31 = _mm_sub_pd(iy3,jy1);
799 dz31 = _mm_sub_pd(iz3,jz1);
800 dx32 = _mm_sub_pd(ix3,jx2);
801 dy32 = _mm_sub_pd(iy3,jy2);
802 dz32 = _mm_sub_pd(iz3,jz2);
803 dx33 = _mm_sub_pd(ix3,jx3);
804 dy33 = _mm_sub_pd(iy3,jy3);
805 dz33 = _mm_sub_pd(iz3,jz3);
807 /* Calculate squared distance and things based on it */
808 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
809 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
810 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
811 rsq13 = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
812 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
813 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
814 rsq23 = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
815 rsq31 = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
816 rsq32 = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
817 rsq33 = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
819 rinv00 = gmx_mm_invsqrt_pd(rsq00);
820 rinv11 = gmx_mm_invsqrt_pd(rsq11);
821 rinv12 = gmx_mm_invsqrt_pd(rsq12);
822 rinv13 = gmx_mm_invsqrt_pd(rsq13);
823 rinv21 = gmx_mm_invsqrt_pd(rsq21);
824 rinv22 = gmx_mm_invsqrt_pd(rsq22);
825 rinv23 = gmx_mm_invsqrt_pd(rsq23);
826 rinv31 = gmx_mm_invsqrt_pd(rsq31);
827 rinv32 = gmx_mm_invsqrt_pd(rsq32);
828 rinv33 = gmx_mm_invsqrt_pd(rsq33);
830 fjx0 = _mm_setzero_pd();
831 fjy0 = _mm_setzero_pd();
832 fjz0 = _mm_setzero_pd();
833 fjx1 = _mm_setzero_pd();
834 fjy1 = _mm_setzero_pd();
835 fjz1 = _mm_setzero_pd();
836 fjx2 = _mm_setzero_pd();
837 fjy2 = _mm_setzero_pd();
838 fjz2 = _mm_setzero_pd();
839 fjx3 = _mm_setzero_pd();
840 fjy3 = _mm_setzero_pd();
841 fjz3 = _mm_setzero_pd();
843 /**************************
844 * CALCULATE INTERACTIONS *
845 **************************/
847 r00 = _mm_mul_pd(rsq00,rinv00);
849 /* Calculate table index by multiplying r with table scale and truncate to integer */
850 rt = _mm_mul_pd(r00,vftabscale);
851 vfitab = _mm_cvttpd_epi32(rt);
852 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
853 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
855 /* CUBIC SPLINE TABLE DISPERSION */
856 vfitab = _mm_add_epi32(vfitab,ifour);
857 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
858 F = _mm_setzero_pd();
859 GMX_MM_TRANSPOSE2_PD(Y,F);
860 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
861 H = _mm_setzero_pd();
862 GMX_MM_TRANSPOSE2_PD(G,H);
863 Heps = _mm_mul_pd(vfeps,H);
864 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
865 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
866 vvdw6 = _mm_mul_pd(c6_00,VV);
867 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
868 fvdw6 = _mm_mul_pd(c6_00,FF);
870 /* CUBIC SPLINE TABLE REPULSION */
871 vfitab = _mm_add_epi32(vfitab,ifour);
872 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
873 F = _mm_setzero_pd();
874 GMX_MM_TRANSPOSE2_PD(Y,F);
875 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
876 H = _mm_setzero_pd();
877 GMX_MM_TRANSPOSE2_PD(G,H);
878 Heps = _mm_mul_pd(vfeps,H);
879 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
880 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
881 vvdw12 = _mm_mul_pd(c12_00,VV);
882 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
883 fvdw12 = _mm_mul_pd(c12_00,FF);
884 vvdw = _mm_add_pd(vvdw12,vvdw6);
885 fvdw = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
887 /* Update potential sum for this i atom from the interaction with this j atom. */
888 vvdw = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
889 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
893 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
895 /* Calculate temporary vectorial force */
896 tx = _mm_mul_pd(fscal,dx00);
897 ty = _mm_mul_pd(fscal,dy00);
898 tz = _mm_mul_pd(fscal,dz00);
900 /* Update vectorial force */
901 fix0 = _mm_add_pd(fix0,tx);
902 fiy0 = _mm_add_pd(fiy0,ty);
903 fiz0 = _mm_add_pd(fiz0,tz);
905 fjx0 = _mm_add_pd(fjx0,tx);
906 fjy0 = _mm_add_pd(fjy0,ty);
907 fjz0 = _mm_add_pd(fjz0,tz);
909 /**************************
910 * CALCULATE INTERACTIONS *
911 **************************/
913 r11 = _mm_mul_pd(rsq11,rinv11);
915 /* Calculate table index by multiplying r with table scale and truncate to integer */
916 rt = _mm_mul_pd(r11,vftabscale);
917 vfitab = _mm_cvttpd_epi32(rt);
918 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
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 + gmx_mm_extract_epi32(vfitab,0) );
923 F = _mm_setzero_pd();
924 GMX_MM_TRANSPOSE2_PD(Y,F);
925 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
926 H = _mm_setzero_pd();
927 GMX_MM_TRANSPOSE2_PD(G,H);
928 Heps = _mm_mul_pd(vfeps,H);
929 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
930 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
931 velec = _mm_mul_pd(qq11,VV);
932 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
933 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
935 /* Update potential sum for this i atom from the interaction with this j atom. */
936 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
937 velecsum = _mm_add_pd(velecsum,velec);
941 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
943 /* Calculate temporary vectorial force */
944 tx = _mm_mul_pd(fscal,dx11);
945 ty = _mm_mul_pd(fscal,dy11);
946 tz = _mm_mul_pd(fscal,dz11);
948 /* Update vectorial force */
949 fix1 = _mm_add_pd(fix1,tx);
950 fiy1 = _mm_add_pd(fiy1,ty);
951 fiz1 = _mm_add_pd(fiz1,tz);
953 fjx1 = _mm_add_pd(fjx1,tx);
954 fjy1 = _mm_add_pd(fjy1,ty);
955 fjz1 = _mm_add_pd(fjz1,tz);
957 /**************************
958 * CALCULATE INTERACTIONS *
959 **************************/
961 r12 = _mm_mul_pd(rsq12,rinv12);
963 /* Calculate table index by multiplying r with table scale and truncate to integer */
964 rt = _mm_mul_pd(r12,vftabscale);
965 vfitab = _mm_cvttpd_epi32(rt);
966 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
967 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
969 /* CUBIC SPLINE TABLE ELECTROSTATICS */
970 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
971 F = _mm_setzero_pd();
972 GMX_MM_TRANSPOSE2_PD(Y,F);
973 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
974 H = _mm_setzero_pd();
975 GMX_MM_TRANSPOSE2_PD(G,H);
976 Heps = _mm_mul_pd(vfeps,H);
977 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
978 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
979 velec = _mm_mul_pd(qq12,VV);
980 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
981 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
983 /* Update potential sum for this i atom from the interaction with this j atom. */
984 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
985 velecsum = _mm_add_pd(velecsum,velec);
989 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
991 /* Calculate temporary vectorial force */
992 tx = _mm_mul_pd(fscal,dx12);
993 ty = _mm_mul_pd(fscal,dy12);
994 tz = _mm_mul_pd(fscal,dz12);
996 /* Update vectorial force */
997 fix1 = _mm_add_pd(fix1,tx);
998 fiy1 = _mm_add_pd(fiy1,ty);
999 fiz1 = _mm_add_pd(fiz1,tz);
1001 fjx2 = _mm_add_pd(fjx2,tx);
1002 fjy2 = _mm_add_pd(fjy2,ty);
1003 fjz2 = _mm_add_pd(fjz2,tz);
1005 /**************************
1006 * CALCULATE INTERACTIONS *
1007 **************************/
1009 r13 = _mm_mul_pd(rsq13,rinv13);
1011 /* Calculate table index by multiplying r with table scale and truncate to integer */
1012 rt = _mm_mul_pd(r13,vftabscale);
1013 vfitab = _mm_cvttpd_epi32(rt);
1014 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1015 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1017 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1018 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1019 F = _mm_setzero_pd();
1020 GMX_MM_TRANSPOSE2_PD(Y,F);
1021 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1022 H = _mm_setzero_pd();
1023 GMX_MM_TRANSPOSE2_PD(G,H);
1024 Heps = _mm_mul_pd(vfeps,H);
1025 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1026 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1027 velec = _mm_mul_pd(qq13,VV);
1028 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1029 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq13,FF),_mm_mul_pd(vftabscale,rinv13)));
1031 /* Update potential sum for this i atom from the interaction with this j atom. */
1032 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1033 velecsum = _mm_add_pd(velecsum,velec);
1037 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1039 /* Calculate temporary vectorial force */
1040 tx = _mm_mul_pd(fscal,dx13);
1041 ty = _mm_mul_pd(fscal,dy13);
1042 tz = _mm_mul_pd(fscal,dz13);
1044 /* Update vectorial force */
1045 fix1 = _mm_add_pd(fix1,tx);
1046 fiy1 = _mm_add_pd(fiy1,ty);
1047 fiz1 = _mm_add_pd(fiz1,tz);
1049 fjx3 = _mm_add_pd(fjx3,tx);
1050 fjy3 = _mm_add_pd(fjy3,ty);
1051 fjz3 = _mm_add_pd(fjz3,tz);
1053 /**************************
1054 * CALCULATE INTERACTIONS *
1055 **************************/
1057 r21 = _mm_mul_pd(rsq21,rinv21);
1059 /* Calculate table index by multiplying r with table scale and truncate to integer */
1060 rt = _mm_mul_pd(r21,vftabscale);
1061 vfitab = _mm_cvttpd_epi32(rt);
1062 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1063 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1065 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1066 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1067 F = _mm_setzero_pd();
1068 GMX_MM_TRANSPOSE2_PD(Y,F);
1069 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1070 H = _mm_setzero_pd();
1071 GMX_MM_TRANSPOSE2_PD(G,H);
1072 Heps = _mm_mul_pd(vfeps,H);
1073 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1074 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1075 velec = _mm_mul_pd(qq21,VV);
1076 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1077 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
1079 /* Update potential sum for this i atom from the interaction with this j atom. */
1080 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1081 velecsum = _mm_add_pd(velecsum,velec);
1085 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1087 /* Calculate temporary vectorial force */
1088 tx = _mm_mul_pd(fscal,dx21);
1089 ty = _mm_mul_pd(fscal,dy21);
1090 tz = _mm_mul_pd(fscal,dz21);
1092 /* Update vectorial force */
1093 fix2 = _mm_add_pd(fix2,tx);
1094 fiy2 = _mm_add_pd(fiy2,ty);
1095 fiz2 = _mm_add_pd(fiz2,tz);
1097 fjx1 = _mm_add_pd(fjx1,tx);
1098 fjy1 = _mm_add_pd(fjy1,ty);
1099 fjz1 = _mm_add_pd(fjz1,tz);
1101 /**************************
1102 * CALCULATE INTERACTIONS *
1103 **************************/
1105 r22 = _mm_mul_pd(rsq22,rinv22);
1107 /* Calculate table index by multiplying r with table scale and truncate to integer */
1108 rt = _mm_mul_pd(r22,vftabscale);
1109 vfitab = _mm_cvttpd_epi32(rt);
1110 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1111 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1113 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1114 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1115 F = _mm_setzero_pd();
1116 GMX_MM_TRANSPOSE2_PD(Y,F);
1117 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1118 H = _mm_setzero_pd();
1119 GMX_MM_TRANSPOSE2_PD(G,H);
1120 Heps = _mm_mul_pd(vfeps,H);
1121 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1122 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1123 velec = _mm_mul_pd(qq22,VV);
1124 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1125 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
1127 /* Update potential sum for this i atom from the interaction with this j atom. */
1128 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1129 velecsum = _mm_add_pd(velecsum,velec);
1133 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1135 /* Calculate temporary vectorial force */
1136 tx = _mm_mul_pd(fscal,dx22);
1137 ty = _mm_mul_pd(fscal,dy22);
1138 tz = _mm_mul_pd(fscal,dz22);
1140 /* Update vectorial force */
1141 fix2 = _mm_add_pd(fix2,tx);
1142 fiy2 = _mm_add_pd(fiy2,ty);
1143 fiz2 = _mm_add_pd(fiz2,tz);
1145 fjx2 = _mm_add_pd(fjx2,tx);
1146 fjy2 = _mm_add_pd(fjy2,ty);
1147 fjz2 = _mm_add_pd(fjz2,tz);
1149 /**************************
1150 * CALCULATE INTERACTIONS *
1151 **************************/
1153 r23 = _mm_mul_pd(rsq23,rinv23);
1155 /* Calculate table index by multiplying r with table scale and truncate to integer */
1156 rt = _mm_mul_pd(r23,vftabscale);
1157 vfitab = _mm_cvttpd_epi32(rt);
1158 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1159 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1161 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1162 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1163 F = _mm_setzero_pd();
1164 GMX_MM_TRANSPOSE2_PD(Y,F);
1165 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1166 H = _mm_setzero_pd();
1167 GMX_MM_TRANSPOSE2_PD(G,H);
1168 Heps = _mm_mul_pd(vfeps,H);
1169 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1170 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1171 velec = _mm_mul_pd(qq23,VV);
1172 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1173 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq23,FF),_mm_mul_pd(vftabscale,rinv23)));
1175 /* Update potential sum for this i atom from the interaction with this j atom. */
1176 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1177 velecsum = _mm_add_pd(velecsum,velec);
1181 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1183 /* Calculate temporary vectorial force */
1184 tx = _mm_mul_pd(fscal,dx23);
1185 ty = _mm_mul_pd(fscal,dy23);
1186 tz = _mm_mul_pd(fscal,dz23);
1188 /* Update vectorial force */
1189 fix2 = _mm_add_pd(fix2,tx);
1190 fiy2 = _mm_add_pd(fiy2,ty);
1191 fiz2 = _mm_add_pd(fiz2,tz);
1193 fjx3 = _mm_add_pd(fjx3,tx);
1194 fjy3 = _mm_add_pd(fjy3,ty);
1195 fjz3 = _mm_add_pd(fjz3,tz);
1197 /**************************
1198 * CALCULATE INTERACTIONS *
1199 **************************/
1201 r31 = _mm_mul_pd(rsq31,rinv31);
1203 /* Calculate table index by multiplying r with table scale and truncate to integer */
1204 rt = _mm_mul_pd(r31,vftabscale);
1205 vfitab = _mm_cvttpd_epi32(rt);
1206 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1207 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1209 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1210 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1211 F = _mm_setzero_pd();
1212 GMX_MM_TRANSPOSE2_PD(Y,F);
1213 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1214 H = _mm_setzero_pd();
1215 GMX_MM_TRANSPOSE2_PD(G,H);
1216 Heps = _mm_mul_pd(vfeps,H);
1217 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1218 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1219 velec = _mm_mul_pd(qq31,VV);
1220 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1221 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq31,FF),_mm_mul_pd(vftabscale,rinv31)));
1223 /* Update potential sum for this i atom from the interaction with this j atom. */
1224 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1225 velecsum = _mm_add_pd(velecsum,velec);
1229 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1231 /* Calculate temporary vectorial force */
1232 tx = _mm_mul_pd(fscal,dx31);
1233 ty = _mm_mul_pd(fscal,dy31);
1234 tz = _mm_mul_pd(fscal,dz31);
1236 /* Update vectorial force */
1237 fix3 = _mm_add_pd(fix3,tx);
1238 fiy3 = _mm_add_pd(fiy3,ty);
1239 fiz3 = _mm_add_pd(fiz3,tz);
1241 fjx1 = _mm_add_pd(fjx1,tx);
1242 fjy1 = _mm_add_pd(fjy1,ty);
1243 fjz1 = _mm_add_pd(fjz1,tz);
1245 /**************************
1246 * CALCULATE INTERACTIONS *
1247 **************************/
1249 r32 = _mm_mul_pd(rsq32,rinv32);
1251 /* Calculate table index by multiplying r with table scale and truncate to integer */
1252 rt = _mm_mul_pd(r32,vftabscale);
1253 vfitab = _mm_cvttpd_epi32(rt);
1254 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1255 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1257 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1258 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1259 F = _mm_setzero_pd();
1260 GMX_MM_TRANSPOSE2_PD(Y,F);
1261 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1262 H = _mm_setzero_pd();
1263 GMX_MM_TRANSPOSE2_PD(G,H);
1264 Heps = _mm_mul_pd(vfeps,H);
1265 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1266 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1267 velec = _mm_mul_pd(qq32,VV);
1268 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1269 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq32,FF),_mm_mul_pd(vftabscale,rinv32)));
1271 /* Update potential sum for this i atom from the interaction with this j atom. */
1272 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1273 velecsum = _mm_add_pd(velecsum,velec);
1277 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1279 /* Calculate temporary vectorial force */
1280 tx = _mm_mul_pd(fscal,dx32);
1281 ty = _mm_mul_pd(fscal,dy32);
1282 tz = _mm_mul_pd(fscal,dz32);
1284 /* Update vectorial force */
1285 fix3 = _mm_add_pd(fix3,tx);
1286 fiy3 = _mm_add_pd(fiy3,ty);
1287 fiz3 = _mm_add_pd(fiz3,tz);
1289 fjx2 = _mm_add_pd(fjx2,tx);
1290 fjy2 = _mm_add_pd(fjy2,ty);
1291 fjz2 = _mm_add_pd(fjz2,tz);
1293 /**************************
1294 * CALCULATE INTERACTIONS *
1295 **************************/
1297 r33 = _mm_mul_pd(rsq33,rinv33);
1299 /* Calculate table index by multiplying r with table scale and truncate to integer */
1300 rt = _mm_mul_pd(r33,vftabscale);
1301 vfitab = _mm_cvttpd_epi32(rt);
1302 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1303 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1305 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1306 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1307 F = _mm_setzero_pd();
1308 GMX_MM_TRANSPOSE2_PD(Y,F);
1309 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1310 H = _mm_setzero_pd();
1311 GMX_MM_TRANSPOSE2_PD(G,H);
1312 Heps = _mm_mul_pd(vfeps,H);
1313 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1314 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1315 velec = _mm_mul_pd(qq33,VV);
1316 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1317 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq33,FF),_mm_mul_pd(vftabscale,rinv33)));
1319 /* Update potential sum for this i atom from the interaction with this j atom. */
1320 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1321 velecsum = _mm_add_pd(velecsum,velec);
1325 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1327 /* Calculate temporary vectorial force */
1328 tx = _mm_mul_pd(fscal,dx33);
1329 ty = _mm_mul_pd(fscal,dy33);
1330 tz = _mm_mul_pd(fscal,dz33);
1332 /* Update vectorial force */
1333 fix3 = _mm_add_pd(fix3,tx);
1334 fiy3 = _mm_add_pd(fiy3,ty);
1335 fiz3 = _mm_add_pd(fiz3,tz);
1337 fjx3 = _mm_add_pd(fjx3,tx);
1338 fjy3 = _mm_add_pd(fjy3,ty);
1339 fjz3 = _mm_add_pd(fjz3,tz);
1341 gmx_mm_decrement_4rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1343 /* Inner loop uses 446 flops */
1346 /* End of innermost loop */
1348 gmx_mm_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1349 f+i_coord_offset,fshift+i_shift_offset);
1352 /* Update potential energies */
1353 gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1354 gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1356 /* Increment number of inner iterations */
1357 inneriter += j_index_end - j_index_start;
1359 /* Outer loop uses 26 flops */
1362 /* Increment number of outer iterations */
1365 /* Update outer/inner flops */
1367 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*446);
1370 * Gromacs nonbonded kernel: nb_kernel_ElecCSTab_VdwCSTab_GeomW4W4_F_sse2_double
1371 * Electrostatics interaction: CubicSplineTable
1372 * VdW interaction: CubicSplineTable
1373 * Geometry: Water4-Water4
1374 * Calculate force/pot: Force
1377 nb_kernel_ElecCSTab_VdwCSTab_GeomW4W4_F_sse2_double
1378 (t_nblist * gmx_restrict nlist,
1379 rvec * gmx_restrict xx,
1380 rvec * gmx_restrict ff,
1381 t_forcerec * gmx_restrict fr,
1382 t_mdatoms * gmx_restrict mdatoms,
1383 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1384 t_nrnb * gmx_restrict nrnb)
1386 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1387 * just 0 for non-waters.
1388 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
1389 * jnr indices corresponding to data put in the four positions in the SIMD register.
1391 int i_shift_offset,i_coord_offset,outeriter,inneriter;
1392 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1394 int j_coord_offsetA,j_coord_offsetB;
1395 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
1396 real rcutoff_scalar;
1397 real *shiftvec,*fshift,*x,*f;
1398 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1400 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1402 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1404 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1406 __m128d ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1407 int vdwjidx0A,vdwjidx0B;
1408 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1409 int vdwjidx1A,vdwjidx1B;
1410 __m128d jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1411 int vdwjidx2A,vdwjidx2B;
1412 __m128d jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1413 int vdwjidx3A,vdwjidx3B;
1414 __m128d jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1415 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1416 __m128d dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1417 __m128d dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1418 __m128d dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1419 __m128d dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1420 __m128d dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1421 __m128d dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1422 __m128d dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1423 __m128d dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1424 __m128d dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1425 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
1428 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1431 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
1432 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
1434 __m128i ifour = _mm_set1_epi32(4);
1435 __m128d rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
1437 __m128d dummy_mask,cutoff_mask;
1438 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
1439 __m128d one = _mm_set1_pd(1.0);
1440 __m128d two = _mm_set1_pd(2.0);
1446 jindex = nlist->jindex;
1448 shiftidx = nlist->shift;
1450 shiftvec = fr->shift_vec[0];
1451 fshift = fr->fshift[0];
1452 facel = _mm_set1_pd(fr->epsfac);
1453 charge = mdatoms->chargeA;
1454 nvdwtype = fr->ntype;
1455 vdwparam = fr->nbfp;
1456 vdwtype = mdatoms->typeA;
1458 vftab = kernel_data->table_elec_vdw->data;
1459 vftabscale = _mm_set1_pd(kernel_data->table_elec_vdw->scale);
1461 /* Setup water-specific parameters */
1462 inr = nlist->iinr[0];
1463 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
1464 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
1465 iq3 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
1466 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
1468 jq1 = _mm_set1_pd(charge[inr+1]);
1469 jq2 = _mm_set1_pd(charge[inr+2]);
1470 jq3 = _mm_set1_pd(charge[inr+3]);
1471 vdwjidx0A = 2*vdwtype[inr+0];
1472 c6_00 = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
1473 c12_00 = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
1474 qq11 = _mm_mul_pd(iq1,jq1);
1475 qq12 = _mm_mul_pd(iq1,jq2);
1476 qq13 = _mm_mul_pd(iq1,jq3);
1477 qq21 = _mm_mul_pd(iq2,jq1);
1478 qq22 = _mm_mul_pd(iq2,jq2);
1479 qq23 = _mm_mul_pd(iq2,jq3);
1480 qq31 = _mm_mul_pd(iq3,jq1);
1481 qq32 = _mm_mul_pd(iq3,jq2);
1482 qq33 = _mm_mul_pd(iq3,jq3);
1484 /* Avoid stupid compiler warnings */
1486 j_coord_offsetA = 0;
1487 j_coord_offsetB = 0;
1492 /* Start outer loop over neighborlists */
1493 for(iidx=0; iidx<nri; iidx++)
1495 /* Load shift vector for this list */
1496 i_shift_offset = DIM*shiftidx[iidx];
1498 /* Load limits for loop over neighbors */
1499 j_index_start = jindex[iidx];
1500 j_index_end = jindex[iidx+1];
1502 /* Get outer coordinate index */
1504 i_coord_offset = DIM*inr;
1506 /* Load i particle coords and add shift vector */
1507 gmx_mm_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1508 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1510 fix0 = _mm_setzero_pd();
1511 fiy0 = _mm_setzero_pd();
1512 fiz0 = _mm_setzero_pd();
1513 fix1 = _mm_setzero_pd();
1514 fiy1 = _mm_setzero_pd();
1515 fiz1 = _mm_setzero_pd();
1516 fix2 = _mm_setzero_pd();
1517 fiy2 = _mm_setzero_pd();
1518 fiz2 = _mm_setzero_pd();
1519 fix3 = _mm_setzero_pd();
1520 fiy3 = _mm_setzero_pd();
1521 fiz3 = _mm_setzero_pd();
1523 /* Start inner kernel loop */
1524 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
1527 /* Get j neighbor index, and coordinate index */
1529 jnrB = jjnr[jidx+1];
1530 j_coord_offsetA = DIM*jnrA;
1531 j_coord_offsetB = DIM*jnrB;
1533 /* load j atom coordinates */
1534 gmx_mm_load_4rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1535 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1536 &jy2,&jz2,&jx3,&jy3,&jz3);
1538 /* Calculate displacement vector */
1539 dx00 = _mm_sub_pd(ix0,jx0);
1540 dy00 = _mm_sub_pd(iy0,jy0);
1541 dz00 = _mm_sub_pd(iz0,jz0);
1542 dx11 = _mm_sub_pd(ix1,jx1);
1543 dy11 = _mm_sub_pd(iy1,jy1);
1544 dz11 = _mm_sub_pd(iz1,jz1);
1545 dx12 = _mm_sub_pd(ix1,jx2);
1546 dy12 = _mm_sub_pd(iy1,jy2);
1547 dz12 = _mm_sub_pd(iz1,jz2);
1548 dx13 = _mm_sub_pd(ix1,jx3);
1549 dy13 = _mm_sub_pd(iy1,jy3);
1550 dz13 = _mm_sub_pd(iz1,jz3);
1551 dx21 = _mm_sub_pd(ix2,jx1);
1552 dy21 = _mm_sub_pd(iy2,jy1);
1553 dz21 = _mm_sub_pd(iz2,jz1);
1554 dx22 = _mm_sub_pd(ix2,jx2);
1555 dy22 = _mm_sub_pd(iy2,jy2);
1556 dz22 = _mm_sub_pd(iz2,jz2);
1557 dx23 = _mm_sub_pd(ix2,jx3);
1558 dy23 = _mm_sub_pd(iy2,jy3);
1559 dz23 = _mm_sub_pd(iz2,jz3);
1560 dx31 = _mm_sub_pd(ix3,jx1);
1561 dy31 = _mm_sub_pd(iy3,jy1);
1562 dz31 = _mm_sub_pd(iz3,jz1);
1563 dx32 = _mm_sub_pd(ix3,jx2);
1564 dy32 = _mm_sub_pd(iy3,jy2);
1565 dz32 = _mm_sub_pd(iz3,jz2);
1566 dx33 = _mm_sub_pd(ix3,jx3);
1567 dy33 = _mm_sub_pd(iy3,jy3);
1568 dz33 = _mm_sub_pd(iz3,jz3);
1570 /* Calculate squared distance and things based on it */
1571 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1572 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1573 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1574 rsq13 = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
1575 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1576 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1577 rsq23 = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
1578 rsq31 = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
1579 rsq32 = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
1580 rsq33 = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
1582 rinv00 = gmx_mm_invsqrt_pd(rsq00);
1583 rinv11 = gmx_mm_invsqrt_pd(rsq11);
1584 rinv12 = gmx_mm_invsqrt_pd(rsq12);
1585 rinv13 = gmx_mm_invsqrt_pd(rsq13);
1586 rinv21 = gmx_mm_invsqrt_pd(rsq21);
1587 rinv22 = gmx_mm_invsqrt_pd(rsq22);
1588 rinv23 = gmx_mm_invsqrt_pd(rsq23);
1589 rinv31 = gmx_mm_invsqrt_pd(rsq31);
1590 rinv32 = gmx_mm_invsqrt_pd(rsq32);
1591 rinv33 = gmx_mm_invsqrt_pd(rsq33);
1593 fjx0 = _mm_setzero_pd();
1594 fjy0 = _mm_setzero_pd();
1595 fjz0 = _mm_setzero_pd();
1596 fjx1 = _mm_setzero_pd();
1597 fjy1 = _mm_setzero_pd();
1598 fjz1 = _mm_setzero_pd();
1599 fjx2 = _mm_setzero_pd();
1600 fjy2 = _mm_setzero_pd();
1601 fjz2 = _mm_setzero_pd();
1602 fjx3 = _mm_setzero_pd();
1603 fjy3 = _mm_setzero_pd();
1604 fjz3 = _mm_setzero_pd();
1606 /**************************
1607 * CALCULATE INTERACTIONS *
1608 **************************/
1610 r00 = _mm_mul_pd(rsq00,rinv00);
1612 /* Calculate table index by multiplying r with table scale and truncate to integer */
1613 rt = _mm_mul_pd(r00,vftabscale);
1614 vfitab = _mm_cvttpd_epi32(rt);
1615 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1616 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1618 /* CUBIC SPLINE TABLE DISPERSION */
1619 vfitab = _mm_add_epi32(vfitab,ifour);
1620 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1621 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1622 GMX_MM_TRANSPOSE2_PD(Y,F);
1623 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1624 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1625 GMX_MM_TRANSPOSE2_PD(G,H);
1626 Heps = _mm_mul_pd(vfeps,H);
1627 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1628 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1629 fvdw6 = _mm_mul_pd(c6_00,FF);
1631 /* CUBIC SPLINE TABLE REPULSION */
1632 vfitab = _mm_add_epi32(vfitab,ifour);
1633 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1634 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1635 GMX_MM_TRANSPOSE2_PD(Y,F);
1636 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1637 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1638 GMX_MM_TRANSPOSE2_PD(G,H);
1639 Heps = _mm_mul_pd(vfeps,H);
1640 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1641 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1642 fvdw12 = _mm_mul_pd(c12_00,FF);
1643 fvdw = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
1647 /* Calculate temporary vectorial force */
1648 tx = _mm_mul_pd(fscal,dx00);
1649 ty = _mm_mul_pd(fscal,dy00);
1650 tz = _mm_mul_pd(fscal,dz00);
1652 /* Update vectorial force */
1653 fix0 = _mm_add_pd(fix0,tx);
1654 fiy0 = _mm_add_pd(fiy0,ty);
1655 fiz0 = _mm_add_pd(fiz0,tz);
1657 fjx0 = _mm_add_pd(fjx0,tx);
1658 fjy0 = _mm_add_pd(fjy0,ty);
1659 fjz0 = _mm_add_pd(fjz0,tz);
1661 /**************************
1662 * CALCULATE INTERACTIONS *
1663 **************************/
1665 r11 = _mm_mul_pd(rsq11,rinv11);
1667 /* Calculate table index by multiplying r with table scale and truncate to integer */
1668 rt = _mm_mul_pd(r11,vftabscale);
1669 vfitab = _mm_cvttpd_epi32(rt);
1670 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1671 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1673 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1674 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1675 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1676 GMX_MM_TRANSPOSE2_PD(Y,F);
1677 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1678 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1679 GMX_MM_TRANSPOSE2_PD(G,H);
1680 Heps = _mm_mul_pd(vfeps,H);
1681 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1682 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1683 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
1687 /* Calculate temporary vectorial force */
1688 tx = _mm_mul_pd(fscal,dx11);
1689 ty = _mm_mul_pd(fscal,dy11);
1690 tz = _mm_mul_pd(fscal,dz11);
1692 /* Update vectorial force */
1693 fix1 = _mm_add_pd(fix1,tx);
1694 fiy1 = _mm_add_pd(fiy1,ty);
1695 fiz1 = _mm_add_pd(fiz1,tz);
1697 fjx1 = _mm_add_pd(fjx1,tx);
1698 fjy1 = _mm_add_pd(fjy1,ty);
1699 fjz1 = _mm_add_pd(fjz1,tz);
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);
1710 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1711 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1713 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1714 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1715 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1716 GMX_MM_TRANSPOSE2_PD(Y,F);
1717 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1718 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1719 GMX_MM_TRANSPOSE2_PD(G,H);
1720 Heps = _mm_mul_pd(vfeps,H);
1721 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1722 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1723 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
1727 /* Calculate temporary vectorial force */
1728 tx = _mm_mul_pd(fscal,dx12);
1729 ty = _mm_mul_pd(fscal,dy12);
1730 tz = _mm_mul_pd(fscal,dz12);
1732 /* Update vectorial force */
1733 fix1 = _mm_add_pd(fix1,tx);
1734 fiy1 = _mm_add_pd(fiy1,ty);
1735 fiz1 = _mm_add_pd(fiz1,tz);
1737 fjx2 = _mm_add_pd(fjx2,tx);
1738 fjy2 = _mm_add_pd(fjy2,ty);
1739 fjz2 = _mm_add_pd(fjz2,tz);
1741 /**************************
1742 * CALCULATE INTERACTIONS *
1743 **************************/
1745 r13 = _mm_mul_pd(rsq13,rinv13);
1747 /* Calculate table index by multiplying r with table scale and truncate to integer */
1748 rt = _mm_mul_pd(r13,vftabscale);
1749 vfitab = _mm_cvttpd_epi32(rt);
1750 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1751 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1753 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1754 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1755 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1756 GMX_MM_TRANSPOSE2_PD(Y,F);
1757 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1758 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1759 GMX_MM_TRANSPOSE2_PD(G,H);
1760 Heps = _mm_mul_pd(vfeps,H);
1761 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1762 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1763 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq13,FF),_mm_mul_pd(vftabscale,rinv13)));
1767 /* Calculate temporary vectorial force */
1768 tx = _mm_mul_pd(fscal,dx13);
1769 ty = _mm_mul_pd(fscal,dy13);
1770 tz = _mm_mul_pd(fscal,dz13);
1772 /* Update vectorial force */
1773 fix1 = _mm_add_pd(fix1,tx);
1774 fiy1 = _mm_add_pd(fiy1,ty);
1775 fiz1 = _mm_add_pd(fiz1,tz);
1777 fjx3 = _mm_add_pd(fjx3,tx);
1778 fjy3 = _mm_add_pd(fjy3,ty);
1779 fjz3 = _mm_add_pd(fjz3,tz);
1781 /**************************
1782 * CALCULATE INTERACTIONS *
1783 **************************/
1785 r21 = _mm_mul_pd(rsq21,rinv21);
1787 /* Calculate table index by multiplying r with table scale and truncate to integer */
1788 rt = _mm_mul_pd(r21,vftabscale);
1789 vfitab = _mm_cvttpd_epi32(rt);
1790 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1791 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1793 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1794 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1795 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1796 GMX_MM_TRANSPOSE2_PD(Y,F);
1797 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1798 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1799 GMX_MM_TRANSPOSE2_PD(G,H);
1800 Heps = _mm_mul_pd(vfeps,H);
1801 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1802 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1803 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
1807 /* Calculate temporary vectorial force */
1808 tx = _mm_mul_pd(fscal,dx21);
1809 ty = _mm_mul_pd(fscal,dy21);
1810 tz = _mm_mul_pd(fscal,dz21);
1812 /* Update vectorial force */
1813 fix2 = _mm_add_pd(fix2,tx);
1814 fiy2 = _mm_add_pd(fiy2,ty);
1815 fiz2 = _mm_add_pd(fiz2,tz);
1817 fjx1 = _mm_add_pd(fjx1,tx);
1818 fjy1 = _mm_add_pd(fjy1,ty);
1819 fjz1 = _mm_add_pd(fjz1,tz);
1821 /**************************
1822 * CALCULATE INTERACTIONS *
1823 **************************/
1825 r22 = _mm_mul_pd(rsq22,rinv22);
1827 /* Calculate table index by multiplying r with table scale and truncate to integer */
1828 rt = _mm_mul_pd(r22,vftabscale);
1829 vfitab = _mm_cvttpd_epi32(rt);
1830 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1831 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1833 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1834 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1835 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1836 GMX_MM_TRANSPOSE2_PD(Y,F);
1837 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1838 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1839 GMX_MM_TRANSPOSE2_PD(G,H);
1840 Heps = _mm_mul_pd(vfeps,H);
1841 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1842 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1843 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
1847 /* Calculate temporary vectorial force */
1848 tx = _mm_mul_pd(fscal,dx22);
1849 ty = _mm_mul_pd(fscal,dy22);
1850 tz = _mm_mul_pd(fscal,dz22);
1852 /* Update vectorial force */
1853 fix2 = _mm_add_pd(fix2,tx);
1854 fiy2 = _mm_add_pd(fiy2,ty);
1855 fiz2 = _mm_add_pd(fiz2,tz);
1857 fjx2 = _mm_add_pd(fjx2,tx);
1858 fjy2 = _mm_add_pd(fjy2,ty);
1859 fjz2 = _mm_add_pd(fjz2,tz);
1861 /**************************
1862 * CALCULATE INTERACTIONS *
1863 **************************/
1865 r23 = _mm_mul_pd(rsq23,rinv23);
1867 /* Calculate table index by multiplying r with table scale and truncate to integer */
1868 rt = _mm_mul_pd(r23,vftabscale);
1869 vfitab = _mm_cvttpd_epi32(rt);
1870 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1871 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1873 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1874 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1875 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1876 GMX_MM_TRANSPOSE2_PD(Y,F);
1877 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1878 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1879 GMX_MM_TRANSPOSE2_PD(G,H);
1880 Heps = _mm_mul_pd(vfeps,H);
1881 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1882 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1883 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq23,FF),_mm_mul_pd(vftabscale,rinv23)));
1887 /* Calculate temporary vectorial force */
1888 tx = _mm_mul_pd(fscal,dx23);
1889 ty = _mm_mul_pd(fscal,dy23);
1890 tz = _mm_mul_pd(fscal,dz23);
1892 /* Update vectorial force */
1893 fix2 = _mm_add_pd(fix2,tx);
1894 fiy2 = _mm_add_pd(fiy2,ty);
1895 fiz2 = _mm_add_pd(fiz2,tz);
1897 fjx3 = _mm_add_pd(fjx3,tx);
1898 fjy3 = _mm_add_pd(fjy3,ty);
1899 fjz3 = _mm_add_pd(fjz3,tz);
1901 /**************************
1902 * CALCULATE INTERACTIONS *
1903 **************************/
1905 r31 = _mm_mul_pd(rsq31,rinv31);
1907 /* Calculate table index by multiplying r with table scale and truncate to integer */
1908 rt = _mm_mul_pd(r31,vftabscale);
1909 vfitab = _mm_cvttpd_epi32(rt);
1910 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1911 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1913 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1914 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1915 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1916 GMX_MM_TRANSPOSE2_PD(Y,F);
1917 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1918 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1919 GMX_MM_TRANSPOSE2_PD(G,H);
1920 Heps = _mm_mul_pd(vfeps,H);
1921 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1922 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1923 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq31,FF),_mm_mul_pd(vftabscale,rinv31)));
1927 /* Calculate temporary vectorial force */
1928 tx = _mm_mul_pd(fscal,dx31);
1929 ty = _mm_mul_pd(fscal,dy31);
1930 tz = _mm_mul_pd(fscal,dz31);
1932 /* Update vectorial force */
1933 fix3 = _mm_add_pd(fix3,tx);
1934 fiy3 = _mm_add_pd(fiy3,ty);
1935 fiz3 = _mm_add_pd(fiz3,tz);
1937 fjx1 = _mm_add_pd(fjx1,tx);
1938 fjy1 = _mm_add_pd(fjy1,ty);
1939 fjz1 = _mm_add_pd(fjz1,tz);
1941 /**************************
1942 * CALCULATE INTERACTIONS *
1943 **************************/
1945 r32 = _mm_mul_pd(rsq32,rinv32);
1947 /* Calculate table index by multiplying r with table scale and truncate to integer */
1948 rt = _mm_mul_pd(r32,vftabscale);
1949 vfitab = _mm_cvttpd_epi32(rt);
1950 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1951 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1953 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1954 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1955 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1956 GMX_MM_TRANSPOSE2_PD(Y,F);
1957 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1958 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1959 GMX_MM_TRANSPOSE2_PD(G,H);
1960 Heps = _mm_mul_pd(vfeps,H);
1961 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1962 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1963 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq32,FF),_mm_mul_pd(vftabscale,rinv32)));
1967 /* Calculate temporary vectorial force */
1968 tx = _mm_mul_pd(fscal,dx32);
1969 ty = _mm_mul_pd(fscal,dy32);
1970 tz = _mm_mul_pd(fscal,dz32);
1972 /* Update vectorial force */
1973 fix3 = _mm_add_pd(fix3,tx);
1974 fiy3 = _mm_add_pd(fiy3,ty);
1975 fiz3 = _mm_add_pd(fiz3,tz);
1977 fjx2 = _mm_add_pd(fjx2,tx);
1978 fjy2 = _mm_add_pd(fjy2,ty);
1979 fjz2 = _mm_add_pd(fjz2,tz);
1981 /**************************
1982 * CALCULATE INTERACTIONS *
1983 **************************/
1985 r33 = _mm_mul_pd(rsq33,rinv33);
1987 /* Calculate table index by multiplying r with table scale and truncate to integer */
1988 rt = _mm_mul_pd(r33,vftabscale);
1989 vfitab = _mm_cvttpd_epi32(rt);
1990 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1991 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1993 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1994 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1995 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1996 GMX_MM_TRANSPOSE2_PD(Y,F);
1997 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1998 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1999 GMX_MM_TRANSPOSE2_PD(G,H);
2000 Heps = _mm_mul_pd(vfeps,H);
2001 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2002 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2003 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq33,FF),_mm_mul_pd(vftabscale,rinv33)));
2007 /* Calculate temporary vectorial force */
2008 tx = _mm_mul_pd(fscal,dx33);
2009 ty = _mm_mul_pd(fscal,dy33);
2010 tz = _mm_mul_pd(fscal,dz33);
2012 /* Update vectorial force */
2013 fix3 = _mm_add_pd(fix3,tx);
2014 fiy3 = _mm_add_pd(fiy3,ty);
2015 fiz3 = _mm_add_pd(fiz3,tz);
2017 fjx3 = _mm_add_pd(fjx3,tx);
2018 fjy3 = _mm_add_pd(fjy3,ty);
2019 fjz3 = _mm_add_pd(fjz3,tz);
2021 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);
2023 /* Inner loop uses 402 flops */
2026 if(jidx<j_index_end)
2030 j_coord_offsetA = DIM*jnrA;
2032 /* load j atom coordinates */
2033 gmx_mm_load_4rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
2034 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
2035 &jy2,&jz2,&jx3,&jy3,&jz3);
2037 /* Calculate displacement vector */
2038 dx00 = _mm_sub_pd(ix0,jx0);
2039 dy00 = _mm_sub_pd(iy0,jy0);
2040 dz00 = _mm_sub_pd(iz0,jz0);
2041 dx11 = _mm_sub_pd(ix1,jx1);
2042 dy11 = _mm_sub_pd(iy1,jy1);
2043 dz11 = _mm_sub_pd(iz1,jz1);
2044 dx12 = _mm_sub_pd(ix1,jx2);
2045 dy12 = _mm_sub_pd(iy1,jy2);
2046 dz12 = _mm_sub_pd(iz1,jz2);
2047 dx13 = _mm_sub_pd(ix1,jx3);
2048 dy13 = _mm_sub_pd(iy1,jy3);
2049 dz13 = _mm_sub_pd(iz1,jz3);
2050 dx21 = _mm_sub_pd(ix2,jx1);
2051 dy21 = _mm_sub_pd(iy2,jy1);
2052 dz21 = _mm_sub_pd(iz2,jz1);
2053 dx22 = _mm_sub_pd(ix2,jx2);
2054 dy22 = _mm_sub_pd(iy2,jy2);
2055 dz22 = _mm_sub_pd(iz2,jz2);
2056 dx23 = _mm_sub_pd(ix2,jx3);
2057 dy23 = _mm_sub_pd(iy2,jy3);
2058 dz23 = _mm_sub_pd(iz2,jz3);
2059 dx31 = _mm_sub_pd(ix3,jx1);
2060 dy31 = _mm_sub_pd(iy3,jy1);
2061 dz31 = _mm_sub_pd(iz3,jz1);
2062 dx32 = _mm_sub_pd(ix3,jx2);
2063 dy32 = _mm_sub_pd(iy3,jy2);
2064 dz32 = _mm_sub_pd(iz3,jz2);
2065 dx33 = _mm_sub_pd(ix3,jx3);
2066 dy33 = _mm_sub_pd(iy3,jy3);
2067 dz33 = _mm_sub_pd(iz3,jz3);
2069 /* Calculate squared distance and things based on it */
2070 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
2071 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
2072 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
2073 rsq13 = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
2074 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
2075 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
2076 rsq23 = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
2077 rsq31 = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
2078 rsq32 = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
2079 rsq33 = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
2081 rinv00 = gmx_mm_invsqrt_pd(rsq00);
2082 rinv11 = gmx_mm_invsqrt_pd(rsq11);
2083 rinv12 = gmx_mm_invsqrt_pd(rsq12);
2084 rinv13 = gmx_mm_invsqrt_pd(rsq13);
2085 rinv21 = gmx_mm_invsqrt_pd(rsq21);
2086 rinv22 = gmx_mm_invsqrt_pd(rsq22);
2087 rinv23 = gmx_mm_invsqrt_pd(rsq23);
2088 rinv31 = gmx_mm_invsqrt_pd(rsq31);
2089 rinv32 = gmx_mm_invsqrt_pd(rsq32);
2090 rinv33 = gmx_mm_invsqrt_pd(rsq33);
2092 fjx0 = _mm_setzero_pd();
2093 fjy0 = _mm_setzero_pd();
2094 fjz0 = _mm_setzero_pd();
2095 fjx1 = _mm_setzero_pd();
2096 fjy1 = _mm_setzero_pd();
2097 fjz1 = _mm_setzero_pd();
2098 fjx2 = _mm_setzero_pd();
2099 fjy2 = _mm_setzero_pd();
2100 fjz2 = _mm_setzero_pd();
2101 fjx3 = _mm_setzero_pd();
2102 fjy3 = _mm_setzero_pd();
2103 fjz3 = _mm_setzero_pd();
2105 /**************************
2106 * CALCULATE INTERACTIONS *
2107 **************************/
2109 r00 = _mm_mul_pd(rsq00,rinv00);
2111 /* Calculate table index by multiplying r with table scale and truncate to integer */
2112 rt = _mm_mul_pd(r00,vftabscale);
2113 vfitab = _mm_cvttpd_epi32(rt);
2114 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2115 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2117 /* CUBIC SPLINE TABLE DISPERSION */
2118 vfitab = _mm_add_epi32(vfitab,ifour);
2119 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2120 F = _mm_setzero_pd();
2121 GMX_MM_TRANSPOSE2_PD(Y,F);
2122 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2123 H = _mm_setzero_pd();
2124 GMX_MM_TRANSPOSE2_PD(G,H);
2125 Heps = _mm_mul_pd(vfeps,H);
2126 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2127 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2128 fvdw6 = _mm_mul_pd(c6_00,FF);
2130 /* CUBIC SPLINE TABLE REPULSION */
2131 vfitab = _mm_add_epi32(vfitab,ifour);
2132 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2133 F = _mm_setzero_pd();
2134 GMX_MM_TRANSPOSE2_PD(Y,F);
2135 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2136 H = _mm_setzero_pd();
2137 GMX_MM_TRANSPOSE2_PD(G,H);
2138 Heps = _mm_mul_pd(vfeps,H);
2139 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2140 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2141 fvdw12 = _mm_mul_pd(c12_00,FF);
2142 fvdw = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
2146 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2148 /* Calculate temporary vectorial force */
2149 tx = _mm_mul_pd(fscal,dx00);
2150 ty = _mm_mul_pd(fscal,dy00);
2151 tz = _mm_mul_pd(fscal,dz00);
2153 /* Update vectorial force */
2154 fix0 = _mm_add_pd(fix0,tx);
2155 fiy0 = _mm_add_pd(fiy0,ty);
2156 fiz0 = _mm_add_pd(fiz0,tz);
2158 fjx0 = _mm_add_pd(fjx0,tx);
2159 fjy0 = _mm_add_pd(fjy0,ty);
2160 fjz0 = _mm_add_pd(fjz0,tz);
2162 /**************************
2163 * CALCULATE INTERACTIONS *
2164 **************************/
2166 r11 = _mm_mul_pd(rsq11,rinv11);
2168 /* Calculate table index by multiplying r with table scale and truncate to integer */
2169 rt = _mm_mul_pd(r11,vftabscale);
2170 vfitab = _mm_cvttpd_epi32(rt);
2171 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2172 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2174 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2175 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2176 F = _mm_setzero_pd();
2177 GMX_MM_TRANSPOSE2_PD(Y,F);
2178 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2179 H = _mm_setzero_pd();
2180 GMX_MM_TRANSPOSE2_PD(G,H);
2181 Heps = _mm_mul_pd(vfeps,H);
2182 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2183 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2184 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
2188 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2190 /* Calculate temporary vectorial force */
2191 tx = _mm_mul_pd(fscal,dx11);
2192 ty = _mm_mul_pd(fscal,dy11);
2193 tz = _mm_mul_pd(fscal,dz11);
2195 /* Update vectorial force */
2196 fix1 = _mm_add_pd(fix1,tx);
2197 fiy1 = _mm_add_pd(fiy1,ty);
2198 fiz1 = _mm_add_pd(fiz1,tz);
2200 fjx1 = _mm_add_pd(fjx1,tx);
2201 fjy1 = _mm_add_pd(fjy1,ty);
2202 fjz1 = _mm_add_pd(fjz1,tz);
2204 /**************************
2205 * CALCULATE INTERACTIONS *
2206 **************************/
2208 r12 = _mm_mul_pd(rsq12,rinv12);
2210 /* Calculate table index by multiplying r with table scale and truncate to integer */
2211 rt = _mm_mul_pd(r12,vftabscale);
2212 vfitab = _mm_cvttpd_epi32(rt);
2213 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2214 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2216 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2217 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2218 F = _mm_setzero_pd();
2219 GMX_MM_TRANSPOSE2_PD(Y,F);
2220 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2221 H = _mm_setzero_pd();
2222 GMX_MM_TRANSPOSE2_PD(G,H);
2223 Heps = _mm_mul_pd(vfeps,H);
2224 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2225 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2226 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
2230 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2232 /* Calculate temporary vectorial force */
2233 tx = _mm_mul_pd(fscal,dx12);
2234 ty = _mm_mul_pd(fscal,dy12);
2235 tz = _mm_mul_pd(fscal,dz12);
2237 /* Update vectorial force */
2238 fix1 = _mm_add_pd(fix1,tx);
2239 fiy1 = _mm_add_pd(fiy1,ty);
2240 fiz1 = _mm_add_pd(fiz1,tz);
2242 fjx2 = _mm_add_pd(fjx2,tx);
2243 fjy2 = _mm_add_pd(fjy2,ty);
2244 fjz2 = _mm_add_pd(fjz2,tz);
2246 /**************************
2247 * CALCULATE INTERACTIONS *
2248 **************************/
2250 r13 = _mm_mul_pd(rsq13,rinv13);
2252 /* Calculate table index by multiplying r with table scale and truncate to integer */
2253 rt = _mm_mul_pd(r13,vftabscale);
2254 vfitab = _mm_cvttpd_epi32(rt);
2255 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2256 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2258 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2259 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2260 F = _mm_setzero_pd();
2261 GMX_MM_TRANSPOSE2_PD(Y,F);
2262 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2263 H = _mm_setzero_pd();
2264 GMX_MM_TRANSPOSE2_PD(G,H);
2265 Heps = _mm_mul_pd(vfeps,H);
2266 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2267 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2268 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq13,FF),_mm_mul_pd(vftabscale,rinv13)));
2272 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2274 /* Calculate temporary vectorial force */
2275 tx = _mm_mul_pd(fscal,dx13);
2276 ty = _mm_mul_pd(fscal,dy13);
2277 tz = _mm_mul_pd(fscal,dz13);
2279 /* Update vectorial force */
2280 fix1 = _mm_add_pd(fix1,tx);
2281 fiy1 = _mm_add_pd(fiy1,ty);
2282 fiz1 = _mm_add_pd(fiz1,tz);
2284 fjx3 = _mm_add_pd(fjx3,tx);
2285 fjy3 = _mm_add_pd(fjy3,ty);
2286 fjz3 = _mm_add_pd(fjz3,tz);
2288 /**************************
2289 * CALCULATE INTERACTIONS *
2290 **************************/
2292 r21 = _mm_mul_pd(rsq21,rinv21);
2294 /* Calculate table index by multiplying r with table scale and truncate to integer */
2295 rt = _mm_mul_pd(r21,vftabscale);
2296 vfitab = _mm_cvttpd_epi32(rt);
2297 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2298 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2300 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2301 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2302 F = _mm_setzero_pd();
2303 GMX_MM_TRANSPOSE2_PD(Y,F);
2304 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2305 H = _mm_setzero_pd();
2306 GMX_MM_TRANSPOSE2_PD(G,H);
2307 Heps = _mm_mul_pd(vfeps,H);
2308 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2309 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2310 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
2314 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2316 /* Calculate temporary vectorial force */
2317 tx = _mm_mul_pd(fscal,dx21);
2318 ty = _mm_mul_pd(fscal,dy21);
2319 tz = _mm_mul_pd(fscal,dz21);
2321 /* Update vectorial force */
2322 fix2 = _mm_add_pd(fix2,tx);
2323 fiy2 = _mm_add_pd(fiy2,ty);
2324 fiz2 = _mm_add_pd(fiz2,tz);
2326 fjx1 = _mm_add_pd(fjx1,tx);
2327 fjy1 = _mm_add_pd(fjy1,ty);
2328 fjz1 = _mm_add_pd(fjz1,tz);
2330 /**************************
2331 * CALCULATE INTERACTIONS *
2332 **************************/
2334 r22 = _mm_mul_pd(rsq22,rinv22);
2336 /* Calculate table index by multiplying r with table scale and truncate to integer */
2337 rt = _mm_mul_pd(r22,vftabscale);
2338 vfitab = _mm_cvttpd_epi32(rt);
2339 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2340 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2342 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2343 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2344 F = _mm_setzero_pd();
2345 GMX_MM_TRANSPOSE2_PD(Y,F);
2346 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2347 H = _mm_setzero_pd();
2348 GMX_MM_TRANSPOSE2_PD(G,H);
2349 Heps = _mm_mul_pd(vfeps,H);
2350 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2351 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2352 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
2356 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2358 /* Calculate temporary vectorial force */
2359 tx = _mm_mul_pd(fscal,dx22);
2360 ty = _mm_mul_pd(fscal,dy22);
2361 tz = _mm_mul_pd(fscal,dz22);
2363 /* Update vectorial force */
2364 fix2 = _mm_add_pd(fix2,tx);
2365 fiy2 = _mm_add_pd(fiy2,ty);
2366 fiz2 = _mm_add_pd(fiz2,tz);
2368 fjx2 = _mm_add_pd(fjx2,tx);
2369 fjy2 = _mm_add_pd(fjy2,ty);
2370 fjz2 = _mm_add_pd(fjz2,tz);
2372 /**************************
2373 * CALCULATE INTERACTIONS *
2374 **************************/
2376 r23 = _mm_mul_pd(rsq23,rinv23);
2378 /* Calculate table index by multiplying r with table scale and truncate to integer */
2379 rt = _mm_mul_pd(r23,vftabscale);
2380 vfitab = _mm_cvttpd_epi32(rt);
2381 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2382 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2384 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2385 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2386 F = _mm_setzero_pd();
2387 GMX_MM_TRANSPOSE2_PD(Y,F);
2388 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2389 H = _mm_setzero_pd();
2390 GMX_MM_TRANSPOSE2_PD(G,H);
2391 Heps = _mm_mul_pd(vfeps,H);
2392 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2393 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2394 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq23,FF),_mm_mul_pd(vftabscale,rinv23)));
2398 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2400 /* Calculate temporary vectorial force */
2401 tx = _mm_mul_pd(fscal,dx23);
2402 ty = _mm_mul_pd(fscal,dy23);
2403 tz = _mm_mul_pd(fscal,dz23);
2405 /* Update vectorial force */
2406 fix2 = _mm_add_pd(fix2,tx);
2407 fiy2 = _mm_add_pd(fiy2,ty);
2408 fiz2 = _mm_add_pd(fiz2,tz);
2410 fjx3 = _mm_add_pd(fjx3,tx);
2411 fjy3 = _mm_add_pd(fjy3,ty);
2412 fjz3 = _mm_add_pd(fjz3,tz);
2414 /**************************
2415 * CALCULATE INTERACTIONS *
2416 **************************/
2418 r31 = _mm_mul_pd(rsq31,rinv31);
2420 /* Calculate table index by multiplying r with table scale and truncate to integer */
2421 rt = _mm_mul_pd(r31,vftabscale);
2422 vfitab = _mm_cvttpd_epi32(rt);
2423 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2424 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2426 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2427 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2428 F = _mm_setzero_pd();
2429 GMX_MM_TRANSPOSE2_PD(Y,F);
2430 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2431 H = _mm_setzero_pd();
2432 GMX_MM_TRANSPOSE2_PD(G,H);
2433 Heps = _mm_mul_pd(vfeps,H);
2434 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2435 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2436 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq31,FF),_mm_mul_pd(vftabscale,rinv31)));
2440 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2442 /* Calculate temporary vectorial force */
2443 tx = _mm_mul_pd(fscal,dx31);
2444 ty = _mm_mul_pd(fscal,dy31);
2445 tz = _mm_mul_pd(fscal,dz31);
2447 /* Update vectorial force */
2448 fix3 = _mm_add_pd(fix3,tx);
2449 fiy3 = _mm_add_pd(fiy3,ty);
2450 fiz3 = _mm_add_pd(fiz3,tz);
2452 fjx1 = _mm_add_pd(fjx1,tx);
2453 fjy1 = _mm_add_pd(fjy1,ty);
2454 fjz1 = _mm_add_pd(fjz1,tz);
2456 /**************************
2457 * CALCULATE INTERACTIONS *
2458 **************************/
2460 r32 = _mm_mul_pd(rsq32,rinv32);
2462 /* Calculate table index by multiplying r with table scale and truncate to integer */
2463 rt = _mm_mul_pd(r32,vftabscale);
2464 vfitab = _mm_cvttpd_epi32(rt);
2465 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2466 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2468 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2469 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2470 F = _mm_setzero_pd();
2471 GMX_MM_TRANSPOSE2_PD(Y,F);
2472 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2473 H = _mm_setzero_pd();
2474 GMX_MM_TRANSPOSE2_PD(G,H);
2475 Heps = _mm_mul_pd(vfeps,H);
2476 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2477 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2478 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq32,FF),_mm_mul_pd(vftabscale,rinv32)));
2482 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2484 /* Calculate temporary vectorial force */
2485 tx = _mm_mul_pd(fscal,dx32);
2486 ty = _mm_mul_pd(fscal,dy32);
2487 tz = _mm_mul_pd(fscal,dz32);
2489 /* Update vectorial force */
2490 fix3 = _mm_add_pd(fix3,tx);
2491 fiy3 = _mm_add_pd(fiy3,ty);
2492 fiz3 = _mm_add_pd(fiz3,tz);
2494 fjx2 = _mm_add_pd(fjx2,tx);
2495 fjy2 = _mm_add_pd(fjy2,ty);
2496 fjz2 = _mm_add_pd(fjz2,tz);
2498 /**************************
2499 * CALCULATE INTERACTIONS *
2500 **************************/
2502 r33 = _mm_mul_pd(rsq33,rinv33);
2504 /* Calculate table index by multiplying r with table scale and truncate to integer */
2505 rt = _mm_mul_pd(r33,vftabscale);
2506 vfitab = _mm_cvttpd_epi32(rt);
2507 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2508 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2510 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2511 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2512 F = _mm_setzero_pd();
2513 GMX_MM_TRANSPOSE2_PD(Y,F);
2514 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2515 H = _mm_setzero_pd();
2516 GMX_MM_TRANSPOSE2_PD(G,H);
2517 Heps = _mm_mul_pd(vfeps,H);
2518 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2519 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2520 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq33,FF),_mm_mul_pd(vftabscale,rinv33)));
2524 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2526 /* Calculate temporary vectorial force */
2527 tx = _mm_mul_pd(fscal,dx33);
2528 ty = _mm_mul_pd(fscal,dy33);
2529 tz = _mm_mul_pd(fscal,dz33);
2531 /* Update vectorial force */
2532 fix3 = _mm_add_pd(fix3,tx);
2533 fiy3 = _mm_add_pd(fiy3,ty);
2534 fiz3 = _mm_add_pd(fiz3,tz);
2536 fjx3 = _mm_add_pd(fjx3,tx);
2537 fjy3 = _mm_add_pd(fjy3,ty);
2538 fjz3 = _mm_add_pd(fjz3,tz);
2540 gmx_mm_decrement_4rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2542 /* Inner loop uses 402 flops */
2545 /* End of innermost loop */
2547 gmx_mm_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2548 f+i_coord_offset,fshift+i_shift_offset);
2550 /* Increment number of inner iterations */
2551 inneriter += j_index_end - j_index_start;
2553 /* Outer loop uses 24 flops */
2556 /* Increment number of outer iterations */
2559 /* Update outer/inner flops */
2561 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*402);