?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Removed simple.h from nb_kernel_sse2_XX
[gromacs.git] / src / gromacs / gmxlib / nonbonded / nb_kernel_sse2_double / nb_kernel_ElecCSTab_VdwLJ_GeomW4W4_sse2_double.c
blob6b5a555c63cf065ca860fca1c057203c001b7cc6
1 /*
2 * This file is part of the GROMACS molecular simulation package.
3 *
4 * Copyright (c) 2012,2013,2014,2015, 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.
8 *
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.
13 *
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.
18 *
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.
23 *
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.
31 *
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.
34 */
35 /*
36 * Note: this file was generated by the GROMACS sse2_double kernel generator.
37 */
38 #include "gmxpre.h"
39
40 #include "config.h"
41
42 #include <math.h>
43
44 #include "../nb_kernel.h"
45 #include "gromacs/math/vec.h"
46 #include "gromacs/legacyheaders/nrnb.h"
47
48 #include "gromacs/simd/math_x86_sse2_double.h"
49 #include "kernelutil_x86_sse2_double.h"
50
51 /*
52 * Gromacs nonbonded kernel: nb_kernel_ElecCSTab_VdwLJ_GeomW4W4_VF_sse2_double
53 * Electrostatics interaction: CubicSplineTable
54 * VdW interaction: LennardJones
55 * Geometry: Water4-Water4
56 * Calculate force/pot: PotentialAndForce
57 */
58 void
59 nb_kernel_ElecCSTab_VdwLJ_GeomW4W4_VF_sse2_double
60 (t_nblist * gmx_restrict nlist,
61 rvec * gmx_restrict xx,
62 rvec * gmx_restrict ff,
63 t_forcerec * gmx_restrict fr,
64 t_mdatoms * gmx_restrict mdatoms,
65 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
66 t_nrnb * gmx_restrict nrnb)
67 {
68 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
69 * just 0 for non-waters.
70 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
71 * jnr indices corresponding to data put in the four positions in the SIMD register.
72 */
73 int i_shift_offset,i_coord_offset,outeriter,inneriter;
74 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
75 int jnrA,jnrB;
76 int j_coord_offsetA,j_coord_offsetB;
77 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
78 real rcutoff_scalar;
79 real *shiftvec,*fshift,*x,*f;
80 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
81 int vdwioffset0;
82 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
83 int vdwioffset1;
84 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
85 int vdwioffset2;
86 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
87 int vdwioffset3;
88 __m128d ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
89 int vdwjidx0A,vdwjidx0B;
90 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
91 int vdwjidx1A,vdwjidx1B;
92 __m128d jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
93 int vdwjidx2A,vdwjidx2B;
94 __m128d jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
95 int vdwjidx3A,vdwjidx3B;
96 __m128d jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
97 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
98 __m128d dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
99 __m128d dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
100 __m128d dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
101 __m128d dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
102 __m128d dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
103 __m128d dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
104 __m128d dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
105 __m128d dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
106 __m128d dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
107 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
108 real *charge;
109 int nvdwtype;
110 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
111 int *vdwtype;
112 real *vdwparam;
113 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
114 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
115 __m128i vfitab;
116 __m128i ifour = _mm_set1_epi32(4);
117 __m128d rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
118 real *vftab;
119 __m128d dummy_mask,cutoff_mask;
120 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
121 __m128d one = _mm_set1_pd(1.0);
122 __m128d two = _mm_set1_pd(2.0);
123 x = xx[0];
124 f = ff[0];
125
126 nri = nlist->nri;
127 iinr = nlist->iinr;
128 jindex = nlist->jindex;
129 jjnr = nlist->jjnr;
130 shiftidx = nlist->shift;
131 gid = nlist->gid;
132 shiftvec = fr->shift_vec[0];
133 fshift = fr->fshift[0];
134 facel = _mm_set1_pd(fr->epsfac);
135 charge = mdatoms->chargeA;
136 nvdwtype = fr->ntype;
137 vdwparam = fr->nbfp;
138 vdwtype = mdatoms->typeA;
139
140 vftab = kernel_data->table_elec->data;
141 vftabscale = _mm_set1_pd(kernel_data->table_elec->scale);
142
143 /* Setup water-specific parameters */
144 inr = nlist->iinr[0];
145 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
146 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
147 iq3 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
148 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
149
150 jq1 = _mm_set1_pd(charge[inr+1]);
151 jq2 = _mm_set1_pd(charge[inr+2]);
152 jq3 = _mm_set1_pd(charge[inr+3]);
153 vdwjidx0A = 2*vdwtype[inr+0];
154 c6_00 = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
155 c12_00 = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
156 qq11 = _mm_mul_pd(iq1,jq1);
157 qq12 = _mm_mul_pd(iq1,jq2);
158 qq13 = _mm_mul_pd(iq1,jq3);
159 qq21 = _mm_mul_pd(iq2,jq1);
160 qq22 = _mm_mul_pd(iq2,jq2);
161 qq23 = _mm_mul_pd(iq2,jq3);
162 qq31 = _mm_mul_pd(iq3,jq1);
163 qq32 = _mm_mul_pd(iq3,jq2);
164 qq33 = _mm_mul_pd(iq3,jq3);
165
166 /* Avoid stupid compiler warnings */
167 jnrA = jnrB = 0;
168 j_coord_offsetA = 0;
169 j_coord_offsetB = 0;
170
171 outeriter = 0;
172 inneriter = 0;
173
174 /* Start outer loop over neighborlists */
175 for(iidx=0; iidx<nri; iidx++)
176 {
177 /* Load shift vector for this list */
178 i_shift_offset = DIM*shiftidx[iidx];
179
180 /* Load limits for loop over neighbors */
181 j_index_start = jindex[iidx];
182 j_index_end = jindex[iidx+1];
183
184 /* Get outer coordinate index */
185 inr = iinr[iidx];
186 i_coord_offset = DIM*inr;
187
188 /* Load i particle coords and add shift vector */
189 gmx_mm_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
190 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
191
192 fix0 = _mm_setzero_pd();
193 fiy0 = _mm_setzero_pd();
194 fiz0 = _mm_setzero_pd();
195 fix1 = _mm_setzero_pd();
196 fiy1 = _mm_setzero_pd();
197 fiz1 = _mm_setzero_pd();
198 fix2 = _mm_setzero_pd();
199 fiy2 = _mm_setzero_pd();
200 fiz2 = _mm_setzero_pd();
201 fix3 = _mm_setzero_pd();
202 fiy3 = _mm_setzero_pd();
203 fiz3 = _mm_setzero_pd();
204
205 /* Reset potential sums */
206 velecsum = _mm_setzero_pd();
207 vvdwsum = _mm_setzero_pd();
208
209 /* Start inner kernel loop */
210 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
211 {
212
213 /* Get j neighbor index, and coordinate index */
214 jnrA = jjnr[jidx];
215 jnrB = jjnr[jidx+1];
216 j_coord_offsetA = DIM*jnrA;
217 j_coord_offsetB = DIM*jnrB;
218
219 /* load j atom coordinates */
220 gmx_mm_load_4rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
221 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
222 &jy2,&jz2,&jx3,&jy3,&jz3);
223
224 /* Calculate displacement vector */
225 dx00 = _mm_sub_pd(ix0,jx0);
226 dy00 = _mm_sub_pd(iy0,jy0);
227 dz00 = _mm_sub_pd(iz0,jz0);
228 dx11 = _mm_sub_pd(ix1,jx1);
229 dy11 = _mm_sub_pd(iy1,jy1);
230 dz11 = _mm_sub_pd(iz1,jz1);
231 dx12 = _mm_sub_pd(ix1,jx2);
232 dy12 = _mm_sub_pd(iy1,jy2);
233 dz12 = _mm_sub_pd(iz1,jz2);
234 dx13 = _mm_sub_pd(ix1,jx3);
235 dy13 = _mm_sub_pd(iy1,jy3);
236 dz13 = _mm_sub_pd(iz1,jz3);
237 dx21 = _mm_sub_pd(ix2,jx1);
238 dy21 = _mm_sub_pd(iy2,jy1);
239 dz21 = _mm_sub_pd(iz2,jz1);
240 dx22 = _mm_sub_pd(ix2,jx2);
241 dy22 = _mm_sub_pd(iy2,jy2);
242 dz22 = _mm_sub_pd(iz2,jz2);
243 dx23 = _mm_sub_pd(ix2,jx3);
244 dy23 = _mm_sub_pd(iy2,jy3);
245 dz23 = _mm_sub_pd(iz2,jz3);
246 dx31 = _mm_sub_pd(ix3,jx1);
247 dy31 = _mm_sub_pd(iy3,jy1);
248 dz31 = _mm_sub_pd(iz3,jz1);
249 dx32 = _mm_sub_pd(ix3,jx2);
250 dy32 = _mm_sub_pd(iy3,jy2);
251 dz32 = _mm_sub_pd(iz3,jz2);
252 dx33 = _mm_sub_pd(ix3,jx3);
253 dy33 = _mm_sub_pd(iy3,jy3);
254 dz33 = _mm_sub_pd(iz3,jz3);
255
256 /* Calculate squared distance and things based on it */
257 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
258 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
259 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
260 rsq13 = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
261 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
262 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
263 rsq23 = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
264 rsq31 = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
265 rsq32 = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
266 rsq33 = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
267
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);
277
278 rinvsq00 = gmx_mm_inv_pd(rsq00);
279
280 fjx0 = _mm_setzero_pd();
281 fjy0 = _mm_setzero_pd();
282 fjz0 = _mm_setzero_pd();
283 fjx1 = _mm_setzero_pd();
284 fjy1 = _mm_setzero_pd();
285 fjz1 = _mm_setzero_pd();
286 fjx2 = _mm_setzero_pd();
287 fjy2 = _mm_setzero_pd();
288 fjz2 = _mm_setzero_pd();
289 fjx3 = _mm_setzero_pd();
290 fjy3 = _mm_setzero_pd();
291 fjz3 = _mm_setzero_pd();
292
293 /**************************
294 * CALCULATE INTERACTIONS *
295 **************************/
296
297 /* LENNARD-JONES DISPERSION/REPULSION */
298
299 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
300 vvdw6 = _mm_mul_pd(c6_00,rinvsix);
301 vvdw12 = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
302 vvdw = _mm_sub_pd( _mm_mul_pd(vvdw12,one_twelfth) , _mm_mul_pd(vvdw6,one_sixth) );
303 fvdw = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
304
305 /* Update potential sum for this i atom from the interaction with this j atom. */
306 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
307
308 fscal = fvdw;
309
310 /* Calculate temporary vectorial force */
311 tx = _mm_mul_pd(fscal,dx00);
312 ty = _mm_mul_pd(fscal,dy00);
313 tz = _mm_mul_pd(fscal,dz00);
314
315 /* Update vectorial force */
316 fix0 = _mm_add_pd(fix0,tx);
317 fiy0 = _mm_add_pd(fiy0,ty);
318 fiz0 = _mm_add_pd(fiz0,tz);
319
320 fjx0 = _mm_add_pd(fjx0,tx);
321 fjy0 = _mm_add_pd(fjy0,ty);
322 fjz0 = _mm_add_pd(fjz0,tz);
323
324 /**************************
325 * CALCULATE INTERACTIONS *
326 **************************/
327
328 r11 = _mm_mul_pd(rsq11,rinv11);
329
330 /* Calculate table index by multiplying r with table scale and truncate to integer */
331 rt = _mm_mul_pd(r11,vftabscale);
332 vfitab = _mm_cvttpd_epi32(rt);
333 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
334 vfitab = _mm_slli_epi32(vfitab,2);
335
336 /* CUBIC SPLINE TABLE ELECTROSTATICS */
337 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
338 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
339 GMX_MM_TRANSPOSE2_PD(Y,F);
340 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
341 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
342 GMX_MM_TRANSPOSE2_PD(G,H);
343 Heps = _mm_mul_pd(vfeps,H);
344 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
345 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
346 velec = _mm_mul_pd(qq11,VV);
347 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
348 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
349
350 /* Update potential sum for this i atom from the interaction with this j atom. */
351 velecsum = _mm_add_pd(velecsum,velec);
352
353 fscal = felec;
354
355 /* Calculate temporary vectorial force */
356 tx = _mm_mul_pd(fscal,dx11);
357 ty = _mm_mul_pd(fscal,dy11);
358 tz = _mm_mul_pd(fscal,dz11);
359
360 /* Update vectorial force */
361 fix1 = _mm_add_pd(fix1,tx);
362 fiy1 = _mm_add_pd(fiy1,ty);
363 fiz1 = _mm_add_pd(fiz1,tz);
364
365 fjx1 = _mm_add_pd(fjx1,tx);
366 fjy1 = _mm_add_pd(fjy1,ty);
367 fjz1 = _mm_add_pd(fjz1,tz);
368
369 /**************************
370 * CALCULATE INTERACTIONS *
371 **************************/
372
373 r12 = _mm_mul_pd(rsq12,rinv12);
374
375 /* Calculate table index by multiplying r with table scale and truncate to integer */
376 rt = _mm_mul_pd(r12,vftabscale);
377 vfitab = _mm_cvttpd_epi32(rt);
378 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
379 vfitab = _mm_slli_epi32(vfitab,2);
380
381 /* CUBIC SPLINE TABLE ELECTROSTATICS */
382 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
383 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
384 GMX_MM_TRANSPOSE2_PD(Y,F);
385 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
386 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
387 GMX_MM_TRANSPOSE2_PD(G,H);
388 Heps = _mm_mul_pd(vfeps,H);
389 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
390 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
391 velec = _mm_mul_pd(qq12,VV);
392 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
393 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
394
395 /* Update potential sum for this i atom from the interaction with this j atom. */
396 velecsum = _mm_add_pd(velecsum,velec);
397
398 fscal = felec;
399
400 /* Calculate temporary vectorial force */
401 tx = _mm_mul_pd(fscal,dx12);
402 ty = _mm_mul_pd(fscal,dy12);
403 tz = _mm_mul_pd(fscal,dz12);
404
405 /* Update vectorial force */
406 fix1 = _mm_add_pd(fix1,tx);
407 fiy1 = _mm_add_pd(fiy1,ty);
408 fiz1 = _mm_add_pd(fiz1,tz);
409
410 fjx2 = _mm_add_pd(fjx2,tx);
411 fjy2 = _mm_add_pd(fjy2,ty);
412 fjz2 = _mm_add_pd(fjz2,tz);
413
414 /**************************
415 * CALCULATE INTERACTIONS *
416 **************************/
417
418 r13 = _mm_mul_pd(rsq13,rinv13);
419
420 /* Calculate table index by multiplying r with table scale and truncate to integer */
421 rt = _mm_mul_pd(r13,vftabscale);
422 vfitab = _mm_cvttpd_epi32(rt);
423 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
424 vfitab = _mm_slli_epi32(vfitab,2);
425
426 /* CUBIC SPLINE TABLE ELECTROSTATICS */
427 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
428 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
429 GMX_MM_TRANSPOSE2_PD(Y,F);
430 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
431 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
432 GMX_MM_TRANSPOSE2_PD(G,H);
433 Heps = _mm_mul_pd(vfeps,H);
434 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
435 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
436 velec = _mm_mul_pd(qq13,VV);
437 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
438 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq13,FF),_mm_mul_pd(vftabscale,rinv13)));
439
440 /* Update potential sum for this i atom from the interaction with this j atom. */
441 velecsum = _mm_add_pd(velecsum,velec);
442
443 fscal = felec;
444
445 /* Calculate temporary vectorial force */
446 tx = _mm_mul_pd(fscal,dx13);
447 ty = _mm_mul_pd(fscal,dy13);
448 tz = _mm_mul_pd(fscal,dz13);
449
450 /* Update vectorial force */
451 fix1 = _mm_add_pd(fix1,tx);
452 fiy1 = _mm_add_pd(fiy1,ty);
453 fiz1 = _mm_add_pd(fiz1,tz);
454
455 fjx3 = _mm_add_pd(fjx3,tx);
456 fjy3 = _mm_add_pd(fjy3,ty);
457 fjz3 = _mm_add_pd(fjz3,tz);
458
459 /**************************
460 * CALCULATE INTERACTIONS *
461 **************************/
462
463 r21 = _mm_mul_pd(rsq21,rinv21);
464
465 /* Calculate table index by multiplying r with table scale and truncate to integer */
466 rt = _mm_mul_pd(r21,vftabscale);
467 vfitab = _mm_cvttpd_epi32(rt);
468 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
469 vfitab = _mm_slli_epi32(vfitab,2);
470
471 /* CUBIC SPLINE TABLE ELECTROSTATICS */
472 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
473 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
474 GMX_MM_TRANSPOSE2_PD(Y,F);
475 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
476 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
477 GMX_MM_TRANSPOSE2_PD(G,H);
478 Heps = _mm_mul_pd(vfeps,H);
479 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
480 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
481 velec = _mm_mul_pd(qq21,VV);
482 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
483 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
484
485 /* Update potential sum for this i atom from the interaction with this j atom. */
486 velecsum = _mm_add_pd(velecsum,velec);
487
488 fscal = felec;
489
490 /* Calculate temporary vectorial force */
491 tx = _mm_mul_pd(fscal,dx21);
492 ty = _mm_mul_pd(fscal,dy21);
493 tz = _mm_mul_pd(fscal,dz21);
494
495 /* Update vectorial force */
496 fix2 = _mm_add_pd(fix2,tx);
497 fiy2 = _mm_add_pd(fiy2,ty);
498 fiz2 = _mm_add_pd(fiz2,tz);
499
500 fjx1 = _mm_add_pd(fjx1,tx);
501 fjy1 = _mm_add_pd(fjy1,ty);
502 fjz1 = _mm_add_pd(fjz1,tz);
503
504 /**************************
505 * CALCULATE INTERACTIONS *
506 **************************/
507
508 r22 = _mm_mul_pd(rsq22,rinv22);
509
510 /* Calculate table index by multiplying r with table scale and truncate to integer */
511 rt = _mm_mul_pd(r22,vftabscale);
512 vfitab = _mm_cvttpd_epi32(rt);
513 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
514 vfitab = _mm_slli_epi32(vfitab,2);
515
516 /* CUBIC SPLINE TABLE ELECTROSTATICS */
517 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
518 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
519 GMX_MM_TRANSPOSE2_PD(Y,F);
520 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
521 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
522 GMX_MM_TRANSPOSE2_PD(G,H);
523 Heps = _mm_mul_pd(vfeps,H);
524 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
525 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
526 velec = _mm_mul_pd(qq22,VV);
527 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
528 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
529
530 /* Update potential sum for this i atom from the interaction with this j atom. */
531 velecsum = _mm_add_pd(velecsum,velec);
532
533 fscal = felec;
534
535 /* Calculate temporary vectorial force */
536 tx = _mm_mul_pd(fscal,dx22);
537 ty = _mm_mul_pd(fscal,dy22);
538 tz = _mm_mul_pd(fscal,dz22);
539
540 /* Update vectorial force */
541 fix2 = _mm_add_pd(fix2,tx);
542 fiy2 = _mm_add_pd(fiy2,ty);
543 fiz2 = _mm_add_pd(fiz2,tz);
544
545 fjx2 = _mm_add_pd(fjx2,tx);
546 fjy2 = _mm_add_pd(fjy2,ty);
547 fjz2 = _mm_add_pd(fjz2,tz);
548
549 /**************************
550 * CALCULATE INTERACTIONS *
551 **************************/
552
553 r23 = _mm_mul_pd(rsq23,rinv23);
554
555 /* Calculate table index by multiplying r with table scale and truncate to integer */
556 rt = _mm_mul_pd(r23,vftabscale);
557 vfitab = _mm_cvttpd_epi32(rt);
558 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
559 vfitab = _mm_slli_epi32(vfitab,2);
560
561 /* CUBIC SPLINE TABLE ELECTROSTATICS */
562 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
563 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
564 GMX_MM_TRANSPOSE2_PD(Y,F);
565 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
566 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
567 GMX_MM_TRANSPOSE2_PD(G,H);
568 Heps = _mm_mul_pd(vfeps,H);
569 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
570 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
571 velec = _mm_mul_pd(qq23,VV);
572 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
573 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq23,FF),_mm_mul_pd(vftabscale,rinv23)));
574
575 /* Update potential sum for this i atom from the interaction with this j atom. */
576 velecsum = _mm_add_pd(velecsum,velec);
577
578 fscal = felec;
579
580 /* Calculate temporary vectorial force */
581 tx = _mm_mul_pd(fscal,dx23);
582 ty = _mm_mul_pd(fscal,dy23);
583 tz = _mm_mul_pd(fscal,dz23);
584
585 /* Update vectorial force */
586 fix2 = _mm_add_pd(fix2,tx);
587 fiy2 = _mm_add_pd(fiy2,ty);
588 fiz2 = _mm_add_pd(fiz2,tz);
589
590 fjx3 = _mm_add_pd(fjx3,tx);
591 fjy3 = _mm_add_pd(fjy3,ty);
592 fjz3 = _mm_add_pd(fjz3,tz);
593
594 /**************************
595 * CALCULATE INTERACTIONS *
596 **************************/
597
598 r31 = _mm_mul_pd(rsq31,rinv31);
599
600 /* Calculate table index by multiplying r with table scale and truncate to integer */
601 rt = _mm_mul_pd(r31,vftabscale);
602 vfitab = _mm_cvttpd_epi32(rt);
603 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
604 vfitab = _mm_slli_epi32(vfitab,2);
605
606 /* CUBIC SPLINE TABLE ELECTROSTATICS */
607 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
608 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
609 GMX_MM_TRANSPOSE2_PD(Y,F);
610 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
611 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
612 GMX_MM_TRANSPOSE2_PD(G,H);
613 Heps = _mm_mul_pd(vfeps,H);
614 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
615 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
616 velec = _mm_mul_pd(qq31,VV);
617 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
618 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq31,FF),_mm_mul_pd(vftabscale,rinv31)));
619
620 /* Update potential sum for this i atom from the interaction with this j atom. */
621 velecsum = _mm_add_pd(velecsum,velec);
622
623 fscal = felec;
624
625 /* Calculate temporary vectorial force */
626 tx = _mm_mul_pd(fscal,dx31);
627 ty = _mm_mul_pd(fscal,dy31);
628 tz = _mm_mul_pd(fscal,dz31);
629
630 /* Update vectorial force */
631 fix3 = _mm_add_pd(fix3,tx);
632 fiy3 = _mm_add_pd(fiy3,ty);
633 fiz3 = _mm_add_pd(fiz3,tz);
634
635 fjx1 = _mm_add_pd(fjx1,tx);
636 fjy1 = _mm_add_pd(fjy1,ty);
637 fjz1 = _mm_add_pd(fjz1,tz);
638
639 /**************************
640 * CALCULATE INTERACTIONS *
641 **************************/
642
643 r32 = _mm_mul_pd(rsq32,rinv32);
644
645 /* Calculate table index by multiplying r with table scale and truncate to integer */
646 rt = _mm_mul_pd(r32,vftabscale);
647 vfitab = _mm_cvttpd_epi32(rt);
648 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
649 vfitab = _mm_slli_epi32(vfitab,2);
650
651 /* CUBIC SPLINE TABLE ELECTROSTATICS */
652 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
653 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
654 GMX_MM_TRANSPOSE2_PD(Y,F);
655 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
656 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
657 GMX_MM_TRANSPOSE2_PD(G,H);
658 Heps = _mm_mul_pd(vfeps,H);
659 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
660 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
661 velec = _mm_mul_pd(qq32,VV);
662 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
663 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq32,FF),_mm_mul_pd(vftabscale,rinv32)));
664
665 /* Update potential sum for this i atom from the interaction with this j atom. */
666 velecsum = _mm_add_pd(velecsum,velec);
667
668 fscal = felec;
669
670 /* Calculate temporary vectorial force */
671 tx = _mm_mul_pd(fscal,dx32);
672 ty = _mm_mul_pd(fscal,dy32);
673 tz = _mm_mul_pd(fscal,dz32);
674
675 /* Update vectorial force */
676 fix3 = _mm_add_pd(fix3,tx);
677 fiy3 = _mm_add_pd(fiy3,ty);
678 fiz3 = _mm_add_pd(fiz3,tz);
679
680 fjx2 = _mm_add_pd(fjx2,tx);
681 fjy2 = _mm_add_pd(fjy2,ty);
682 fjz2 = _mm_add_pd(fjz2,tz);
683
684 /**************************
685 * CALCULATE INTERACTIONS *
686 **************************/
687
688 r33 = _mm_mul_pd(rsq33,rinv33);
689
690 /* Calculate table index by multiplying r with table scale and truncate to integer */
691 rt = _mm_mul_pd(r33,vftabscale);
692 vfitab = _mm_cvttpd_epi32(rt);
693 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
694 vfitab = _mm_slli_epi32(vfitab,2);
695
696 /* CUBIC SPLINE TABLE ELECTROSTATICS */
697 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
698 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
699 GMX_MM_TRANSPOSE2_PD(Y,F);
700 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
701 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
702 GMX_MM_TRANSPOSE2_PD(G,H);
703 Heps = _mm_mul_pd(vfeps,H);
704 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
705 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
706 velec = _mm_mul_pd(qq33,VV);
707 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
708 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq33,FF),_mm_mul_pd(vftabscale,rinv33)));
709
710 /* Update potential sum for this i atom from the interaction with this j atom. */
711 velecsum = _mm_add_pd(velecsum,velec);
712
713 fscal = felec;
714
715 /* Calculate temporary vectorial force */
716 tx = _mm_mul_pd(fscal,dx33);
717 ty = _mm_mul_pd(fscal,dy33);
718 tz = _mm_mul_pd(fscal,dz33);
719
720 /* Update vectorial force */
721 fix3 = _mm_add_pd(fix3,tx);
722 fiy3 = _mm_add_pd(fiy3,ty);
723 fiz3 = _mm_add_pd(fiz3,tz);
724
725 fjx3 = _mm_add_pd(fjx3,tx);
726 fjy3 = _mm_add_pd(fjy3,ty);
727 fjz3 = _mm_add_pd(fjz3,tz);
728
729 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);
730
731 /* Inner loop uses 422 flops */
732 }
733
734 if(jidx<j_index_end)
735 {
736
737 jnrA = jjnr[jidx];
738 j_coord_offsetA = DIM*jnrA;
739
740 /* load j atom coordinates */
741 gmx_mm_load_4rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
742 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
743 &jy2,&jz2,&jx3,&jy3,&jz3);
744
745 /* Calculate displacement vector */
746 dx00 = _mm_sub_pd(ix0,jx0);
747 dy00 = _mm_sub_pd(iy0,jy0);
748 dz00 = _mm_sub_pd(iz0,jz0);
749 dx11 = _mm_sub_pd(ix1,jx1);
750 dy11 = _mm_sub_pd(iy1,jy1);
751 dz11 = _mm_sub_pd(iz1,jz1);
752 dx12 = _mm_sub_pd(ix1,jx2);
753 dy12 = _mm_sub_pd(iy1,jy2);
754 dz12 = _mm_sub_pd(iz1,jz2);
755 dx13 = _mm_sub_pd(ix1,jx3);
756 dy13 = _mm_sub_pd(iy1,jy3);
757 dz13 = _mm_sub_pd(iz1,jz3);
758 dx21 = _mm_sub_pd(ix2,jx1);
759 dy21 = _mm_sub_pd(iy2,jy1);
760 dz21 = _mm_sub_pd(iz2,jz1);
761 dx22 = _mm_sub_pd(ix2,jx2);
762 dy22 = _mm_sub_pd(iy2,jy2);
763 dz22 = _mm_sub_pd(iz2,jz2);
764 dx23 = _mm_sub_pd(ix2,jx3);
765 dy23 = _mm_sub_pd(iy2,jy3);
766 dz23 = _mm_sub_pd(iz2,jz3);
767 dx31 = _mm_sub_pd(ix3,jx1);
768 dy31 = _mm_sub_pd(iy3,jy1);
769 dz31 = _mm_sub_pd(iz3,jz1);
770 dx32 = _mm_sub_pd(ix3,jx2);
771 dy32 = _mm_sub_pd(iy3,jy2);
772 dz32 = _mm_sub_pd(iz3,jz2);
773 dx33 = _mm_sub_pd(ix3,jx3);
774 dy33 = _mm_sub_pd(iy3,jy3);
775 dz33 = _mm_sub_pd(iz3,jz3);
776
777 /* Calculate squared distance and things based on it */
778 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
779 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
780 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
781 rsq13 = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
782 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
783 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
784 rsq23 = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
785 rsq31 = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
786 rsq32 = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
787 rsq33 = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
788
789 rinv11 = gmx_mm_invsqrt_pd(rsq11);
790 rinv12 = gmx_mm_invsqrt_pd(rsq12);
791 rinv13 = gmx_mm_invsqrt_pd(rsq13);
792 rinv21 = gmx_mm_invsqrt_pd(rsq21);
793 rinv22 = gmx_mm_invsqrt_pd(rsq22);
794 rinv23 = gmx_mm_invsqrt_pd(rsq23);
795 rinv31 = gmx_mm_invsqrt_pd(rsq31);
796 rinv32 = gmx_mm_invsqrt_pd(rsq32);
797 rinv33 = gmx_mm_invsqrt_pd(rsq33);
798
799 rinvsq00 = gmx_mm_inv_pd(rsq00);
800
801 fjx0 = _mm_setzero_pd();
802 fjy0 = _mm_setzero_pd();
803 fjz0 = _mm_setzero_pd();
804 fjx1 = _mm_setzero_pd();
805 fjy1 = _mm_setzero_pd();
806 fjz1 = _mm_setzero_pd();
807 fjx2 = _mm_setzero_pd();
808 fjy2 = _mm_setzero_pd();
809 fjz2 = _mm_setzero_pd();
810 fjx3 = _mm_setzero_pd();
811 fjy3 = _mm_setzero_pd();
812 fjz3 = _mm_setzero_pd();
813
814 /**************************
815 * CALCULATE INTERACTIONS *
816 **************************/
817
818 /* LENNARD-JONES DISPERSION/REPULSION */
819
820 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
821 vvdw6 = _mm_mul_pd(c6_00,rinvsix);
822 vvdw12 = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
823 vvdw = _mm_sub_pd( _mm_mul_pd(vvdw12,one_twelfth) , _mm_mul_pd(vvdw6,one_sixth) );
824 fvdw = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
825
826 /* Update potential sum for this i atom from the interaction with this j atom. */
827 vvdw = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
828 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
829
830 fscal = fvdw;
831
832 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
833
834 /* Calculate temporary vectorial force */
835 tx = _mm_mul_pd(fscal,dx00);
836 ty = _mm_mul_pd(fscal,dy00);
837 tz = _mm_mul_pd(fscal,dz00);
838
839 /* Update vectorial force */
840 fix0 = _mm_add_pd(fix0,tx);
841 fiy0 = _mm_add_pd(fiy0,ty);
842 fiz0 = _mm_add_pd(fiz0,tz);
843
844 fjx0 = _mm_add_pd(fjx0,tx);
845 fjy0 = _mm_add_pd(fjy0,ty);
846 fjz0 = _mm_add_pd(fjz0,tz);
847
848 /**************************
849 * CALCULATE INTERACTIONS *
850 **************************/
851
852 r11 = _mm_mul_pd(rsq11,rinv11);
853
854 /* Calculate table index by multiplying r with table scale and truncate to integer */
855 rt = _mm_mul_pd(r11,vftabscale);
856 vfitab = _mm_cvttpd_epi32(rt);
857 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
858 vfitab = _mm_slli_epi32(vfitab,2);
859
860 /* CUBIC SPLINE TABLE ELECTROSTATICS */
861 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
862 F = _mm_setzero_pd();
863 GMX_MM_TRANSPOSE2_PD(Y,F);
864 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
865 H = _mm_setzero_pd();
866 GMX_MM_TRANSPOSE2_PD(G,H);
867 Heps = _mm_mul_pd(vfeps,H);
868 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
869 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
870 velec = _mm_mul_pd(qq11,VV);
871 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
872 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
873
874 /* Update potential sum for this i atom from the interaction with this j atom. */
875 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
876 velecsum = _mm_add_pd(velecsum,velec);
877
878 fscal = felec;
879
880 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
881
882 /* Calculate temporary vectorial force */
883 tx = _mm_mul_pd(fscal,dx11);
884 ty = _mm_mul_pd(fscal,dy11);
885 tz = _mm_mul_pd(fscal,dz11);
886
887 /* Update vectorial force */
888 fix1 = _mm_add_pd(fix1,tx);
889 fiy1 = _mm_add_pd(fiy1,ty);
890 fiz1 = _mm_add_pd(fiz1,tz);
891
892 fjx1 = _mm_add_pd(fjx1,tx);
893 fjy1 = _mm_add_pd(fjy1,ty);
894 fjz1 = _mm_add_pd(fjz1,tz);
895
896 /**************************
897 * CALCULATE INTERACTIONS *
898 **************************/
899
900 r12 = _mm_mul_pd(rsq12,rinv12);
901
902 /* Calculate table index by multiplying r with table scale and truncate to integer */
903 rt = _mm_mul_pd(r12,vftabscale);
904 vfitab = _mm_cvttpd_epi32(rt);
905 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
906 vfitab = _mm_slli_epi32(vfitab,2);
907
908 /* CUBIC SPLINE TABLE ELECTROSTATICS */
909 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
910 F = _mm_setzero_pd();
911 GMX_MM_TRANSPOSE2_PD(Y,F);
912 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
913 H = _mm_setzero_pd();
914 GMX_MM_TRANSPOSE2_PD(G,H);
915 Heps = _mm_mul_pd(vfeps,H);
916 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
917 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
918 velec = _mm_mul_pd(qq12,VV);
919 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
920 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
921
922 /* Update potential sum for this i atom from the interaction with this j atom. */
923 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
924 velecsum = _mm_add_pd(velecsum,velec);
925
926 fscal = felec;
927
928 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
929
930 /* Calculate temporary vectorial force */
931 tx = _mm_mul_pd(fscal,dx12);
932 ty = _mm_mul_pd(fscal,dy12);
933 tz = _mm_mul_pd(fscal,dz12);
934
935 /* Update vectorial force */
936 fix1 = _mm_add_pd(fix1,tx);
937 fiy1 = _mm_add_pd(fiy1,ty);
938 fiz1 = _mm_add_pd(fiz1,tz);
939
940 fjx2 = _mm_add_pd(fjx2,tx);
941 fjy2 = _mm_add_pd(fjy2,ty);
942 fjz2 = _mm_add_pd(fjz2,tz);
943
944 /**************************
945 * CALCULATE INTERACTIONS *
946 **************************/
947
948 r13 = _mm_mul_pd(rsq13,rinv13);
949
950 /* Calculate table index by multiplying r with table scale and truncate to integer */
951 rt = _mm_mul_pd(r13,vftabscale);
952 vfitab = _mm_cvttpd_epi32(rt);
953 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
954 vfitab = _mm_slli_epi32(vfitab,2);
955
956 /* CUBIC SPLINE TABLE ELECTROSTATICS */
957 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
958 F = _mm_setzero_pd();
959 GMX_MM_TRANSPOSE2_PD(Y,F);
960 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
961 H = _mm_setzero_pd();
962 GMX_MM_TRANSPOSE2_PD(G,H);
963 Heps = _mm_mul_pd(vfeps,H);
964 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
965 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
966 velec = _mm_mul_pd(qq13,VV);
967 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
968 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq13,FF),_mm_mul_pd(vftabscale,rinv13)));
969
970 /* Update potential sum for this i atom from the interaction with this j atom. */
971 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
972 velecsum = _mm_add_pd(velecsum,velec);
973
974 fscal = felec;
975
976 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
977
978 /* Calculate temporary vectorial force */
979 tx = _mm_mul_pd(fscal,dx13);
980 ty = _mm_mul_pd(fscal,dy13);
981 tz = _mm_mul_pd(fscal,dz13);
982
983 /* Update vectorial force */
984 fix1 = _mm_add_pd(fix1,tx);
985 fiy1 = _mm_add_pd(fiy1,ty);
986 fiz1 = _mm_add_pd(fiz1,tz);
987
988 fjx3 = _mm_add_pd(fjx3,tx);
989 fjy3 = _mm_add_pd(fjy3,ty);
990 fjz3 = _mm_add_pd(fjz3,tz);
991
992 /**************************
993 * CALCULATE INTERACTIONS *
994 **************************/
995
996 r21 = _mm_mul_pd(rsq21,rinv21);
997
998 /* Calculate table index by multiplying r with table scale and truncate to integer */
999 rt = _mm_mul_pd(r21,vftabscale);
1000 vfitab = _mm_cvttpd_epi32(rt);
1001 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1002 vfitab = _mm_slli_epi32(vfitab,2);
1003
1004 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1005 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1006 F = _mm_setzero_pd();
1007 GMX_MM_TRANSPOSE2_PD(Y,F);
1008 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1009 H = _mm_setzero_pd();
1010 GMX_MM_TRANSPOSE2_PD(G,H);
1011 Heps = _mm_mul_pd(vfeps,H);
1012 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1013 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1014 velec = _mm_mul_pd(qq21,VV);
1015 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1016 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
1017
1018 /* Update potential sum for this i atom from the interaction with this j atom. */
1019 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1020 velecsum = _mm_add_pd(velecsum,velec);
1021
1022 fscal = felec;
1023
1024 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1025
1026 /* Calculate temporary vectorial force */
1027 tx = _mm_mul_pd(fscal,dx21);
1028 ty = _mm_mul_pd(fscal,dy21);
1029 tz = _mm_mul_pd(fscal,dz21);
1030
1031 /* Update vectorial force */
1032 fix2 = _mm_add_pd(fix2,tx);
1033 fiy2 = _mm_add_pd(fiy2,ty);
1034 fiz2 = _mm_add_pd(fiz2,tz);
1035
1036 fjx1 = _mm_add_pd(fjx1,tx);
1037 fjy1 = _mm_add_pd(fjy1,ty);
1038 fjz1 = _mm_add_pd(fjz1,tz);
1039
1040 /**************************
1041 * CALCULATE INTERACTIONS *
1042 **************************/
1043
1044 r22 = _mm_mul_pd(rsq22,rinv22);
1045
1046 /* Calculate table index by multiplying r with table scale and truncate to integer */
1047 rt = _mm_mul_pd(r22,vftabscale);
1048 vfitab = _mm_cvttpd_epi32(rt);
1049 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1050 vfitab = _mm_slli_epi32(vfitab,2);
1051
1052 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1053 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1054 F = _mm_setzero_pd();
1055 GMX_MM_TRANSPOSE2_PD(Y,F);
1056 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1057 H = _mm_setzero_pd();
1058 GMX_MM_TRANSPOSE2_PD(G,H);
1059 Heps = _mm_mul_pd(vfeps,H);
1060 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1061 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1062 velec = _mm_mul_pd(qq22,VV);
1063 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1064 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
1065
1066 /* Update potential sum for this i atom from the interaction with this j atom. */
1067 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1068 velecsum = _mm_add_pd(velecsum,velec);
1069
1070 fscal = felec;
1071
1072 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1073
1074 /* Calculate temporary vectorial force */
1075 tx = _mm_mul_pd(fscal,dx22);
1076 ty = _mm_mul_pd(fscal,dy22);
1077 tz = _mm_mul_pd(fscal,dz22);
1078
1079 /* Update vectorial force */
1080 fix2 = _mm_add_pd(fix2,tx);
1081 fiy2 = _mm_add_pd(fiy2,ty);
1082 fiz2 = _mm_add_pd(fiz2,tz);
1083
1084 fjx2 = _mm_add_pd(fjx2,tx);
1085 fjy2 = _mm_add_pd(fjy2,ty);
1086 fjz2 = _mm_add_pd(fjz2,tz);
1087
1088 /**************************
1089 * CALCULATE INTERACTIONS *
1090 **************************/
1091
1092 r23 = _mm_mul_pd(rsq23,rinv23);
1093
1094 /* Calculate table index by multiplying r with table scale and truncate to integer */
1095 rt = _mm_mul_pd(r23,vftabscale);
1096 vfitab = _mm_cvttpd_epi32(rt);
1097 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1098 vfitab = _mm_slli_epi32(vfitab,2);
1099
1100 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1101 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1102 F = _mm_setzero_pd();
1103 GMX_MM_TRANSPOSE2_PD(Y,F);
1104 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1105 H = _mm_setzero_pd();
1106 GMX_MM_TRANSPOSE2_PD(G,H);
1107 Heps = _mm_mul_pd(vfeps,H);
1108 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1109 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1110 velec = _mm_mul_pd(qq23,VV);
1111 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1112 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq23,FF),_mm_mul_pd(vftabscale,rinv23)));
1113
1114 /* Update potential sum for this i atom from the interaction with this j atom. */
1115 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1116 velecsum = _mm_add_pd(velecsum,velec);
1117
1118 fscal = felec;
1119
1120 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1121
1122 /* Calculate temporary vectorial force */
1123 tx = _mm_mul_pd(fscal,dx23);
1124 ty = _mm_mul_pd(fscal,dy23);
1125 tz = _mm_mul_pd(fscal,dz23);
1126
1127 /* Update vectorial force */
1128 fix2 = _mm_add_pd(fix2,tx);
1129 fiy2 = _mm_add_pd(fiy2,ty);
1130 fiz2 = _mm_add_pd(fiz2,tz);
1131
1132 fjx3 = _mm_add_pd(fjx3,tx);
1133 fjy3 = _mm_add_pd(fjy3,ty);
1134 fjz3 = _mm_add_pd(fjz3,tz);
1135
1136 /**************************
1137 * CALCULATE INTERACTIONS *
1138 **************************/
1139
1140 r31 = _mm_mul_pd(rsq31,rinv31);
1141
1142 /* Calculate table index by multiplying r with table scale and truncate to integer */
1143 rt = _mm_mul_pd(r31,vftabscale);
1144 vfitab = _mm_cvttpd_epi32(rt);
1145 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1146 vfitab = _mm_slli_epi32(vfitab,2);
1147
1148 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1149 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1150 F = _mm_setzero_pd();
1151 GMX_MM_TRANSPOSE2_PD(Y,F);
1152 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1153 H = _mm_setzero_pd();
1154 GMX_MM_TRANSPOSE2_PD(G,H);
1155 Heps = _mm_mul_pd(vfeps,H);
1156 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1157 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1158 velec = _mm_mul_pd(qq31,VV);
1159 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1160 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq31,FF),_mm_mul_pd(vftabscale,rinv31)));
1161
1162 /* Update potential sum for this i atom from the interaction with this j atom. */
1163 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1164 velecsum = _mm_add_pd(velecsum,velec);
1165
1166 fscal = felec;
1167
1168 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1169
1170 /* Calculate temporary vectorial force */
1171 tx = _mm_mul_pd(fscal,dx31);
1172 ty = _mm_mul_pd(fscal,dy31);
1173 tz = _mm_mul_pd(fscal,dz31);
1174
1175 /* Update vectorial force */
1176 fix3 = _mm_add_pd(fix3,tx);
1177 fiy3 = _mm_add_pd(fiy3,ty);
1178 fiz3 = _mm_add_pd(fiz3,tz);
1179
1180 fjx1 = _mm_add_pd(fjx1,tx);
1181 fjy1 = _mm_add_pd(fjy1,ty);
1182 fjz1 = _mm_add_pd(fjz1,tz);
1183
1184 /**************************
1185 * CALCULATE INTERACTIONS *
1186 **************************/
1187
1188 r32 = _mm_mul_pd(rsq32,rinv32);
1189
1190 /* Calculate table index by multiplying r with table scale and truncate to integer */
1191 rt = _mm_mul_pd(r32,vftabscale);
1192 vfitab = _mm_cvttpd_epi32(rt);
1193 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1194 vfitab = _mm_slli_epi32(vfitab,2);
1195
1196 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1197 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1198 F = _mm_setzero_pd();
1199 GMX_MM_TRANSPOSE2_PD(Y,F);
1200 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1201 H = _mm_setzero_pd();
1202 GMX_MM_TRANSPOSE2_PD(G,H);
1203 Heps = _mm_mul_pd(vfeps,H);
1204 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1205 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1206 velec = _mm_mul_pd(qq32,VV);
1207 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1208 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq32,FF),_mm_mul_pd(vftabscale,rinv32)));
1209
1210 /* Update potential sum for this i atom from the interaction with this j atom. */
1211 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1212 velecsum = _mm_add_pd(velecsum,velec);
1213
1214 fscal = felec;
1215
1216 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1217
1218 /* Calculate temporary vectorial force */
1219 tx = _mm_mul_pd(fscal,dx32);
1220 ty = _mm_mul_pd(fscal,dy32);
1221 tz = _mm_mul_pd(fscal,dz32);
1222
1223 /* Update vectorial force */
1224 fix3 = _mm_add_pd(fix3,tx);
1225 fiy3 = _mm_add_pd(fiy3,ty);
1226 fiz3 = _mm_add_pd(fiz3,tz);
1227
1228 fjx2 = _mm_add_pd(fjx2,tx);
1229 fjy2 = _mm_add_pd(fjy2,ty);
1230 fjz2 = _mm_add_pd(fjz2,tz);
1231
1232 /**************************
1233 * CALCULATE INTERACTIONS *
1234 **************************/
1235
1236 r33 = _mm_mul_pd(rsq33,rinv33);
1237
1238 /* Calculate table index by multiplying r with table scale and truncate to integer */
1239 rt = _mm_mul_pd(r33,vftabscale);
1240 vfitab = _mm_cvttpd_epi32(rt);
1241 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1242 vfitab = _mm_slli_epi32(vfitab,2);
1243
1244 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1245 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1246 F = _mm_setzero_pd();
1247 GMX_MM_TRANSPOSE2_PD(Y,F);
1248 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1249 H = _mm_setzero_pd();
1250 GMX_MM_TRANSPOSE2_PD(G,H);
1251 Heps = _mm_mul_pd(vfeps,H);
1252 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1253 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1254 velec = _mm_mul_pd(qq33,VV);
1255 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1256 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq33,FF),_mm_mul_pd(vftabscale,rinv33)));
1257
1258 /* Update potential sum for this i atom from the interaction with this j atom. */
1259 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1260 velecsum = _mm_add_pd(velecsum,velec);
1261
1262 fscal = felec;
1263
1264 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1265
1266 /* Calculate temporary vectorial force */
1267 tx = _mm_mul_pd(fscal,dx33);
1268 ty = _mm_mul_pd(fscal,dy33);
1269 tz = _mm_mul_pd(fscal,dz33);
1270
1271 /* Update vectorial force */
1272 fix3 = _mm_add_pd(fix3,tx);
1273 fiy3 = _mm_add_pd(fiy3,ty);
1274 fiz3 = _mm_add_pd(fiz3,tz);
1275
1276 fjx3 = _mm_add_pd(fjx3,tx);
1277 fjy3 = _mm_add_pd(fjy3,ty);
1278 fjz3 = _mm_add_pd(fjz3,tz);
1279
1280 gmx_mm_decrement_4rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1281
1282 /* Inner loop uses 422 flops */
1283 }
1284
1285 /* End of innermost loop */
1286
1287 gmx_mm_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1288 f+i_coord_offset,fshift+i_shift_offset);
1289
1290 ggid = gid[iidx];
1291 /* Update potential energies */
1292 gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1293 gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1294
1295 /* Increment number of inner iterations */
1296 inneriter += j_index_end - j_index_start;
1297
1298 /* Outer loop uses 26 flops */
1299 }
1300
1301 /* Increment number of outer iterations */
1302 outeriter += nri;
1303
1304 /* Update outer/inner flops */
1305
1306 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*422);
1307 }
1308 /*
1309 * Gromacs nonbonded kernel: nb_kernel_ElecCSTab_VdwLJ_GeomW4W4_F_sse2_double
1310 * Electrostatics interaction: CubicSplineTable
1311 * VdW interaction: LennardJones
1312 * Geometry: Water4-Water4
1313 * Calculate force/pot: Force
1314 */
1315 void
1316 nb_kernel_ElecCSTab_VdwLJ_GeomW4W4_F_sse2_double
1317 (t_nblist * gmx_restrict nlist,
1318 rvec * gmx_restrict xx,
1319 rvec * gmx_restrict ff,
1320 t_forcerec * gmx_restrict fr,
1321 t_mdatoms * gmx_restrict mdatoms,
1322 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1323 t_nrnb * gmx_restrict nrnb)
1324 {
1325 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1326 * just 0 for non-waters.
1327 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
1328 * jnr indices corresponding to data put in the four positions in the SIMD register.
1329 */
1330 int i_shift_offset,i_coord_offset,outeriter,inneriter;
1331 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1332 int jnrA,jnrB;
1333 int j_coord_offsetA,j_coord_offsetB;
1334 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
1335 real rcutoff_scalar;
1336 real *shiftvec,*fshift,*x,*f;
1337 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1338 int vdwioffset0;
1339 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1340 int vdwioffset1;
1341 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1342 int vdwioffset2;
1343 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1344 int vdwioffset3;
1345 __m128d ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1346 int vdwjidx0A,vdwjidx0B;
1347 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1348 int vdwjidx1A,vdwjidx1B;
1349 __m128d jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1350 int vdwjidx2A,vdwjidx2B;
1351 __m128d jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1352 int vdwjidx3A,vdwjidx3B;
1353 __m128d jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1354 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1355 __m128d dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1356 __m128d dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1357 __m128d dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1358 __m128d dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1359 __m128d dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1360 __m128d dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1361 __m128d dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1362 __m128d dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1363 __m128d dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1364 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
1365 real *charge;
1366 int nvdwtype;
1367 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1368 int *vdwtype;
1369 real *vdwparam;
1370 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
1371 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
1372 __m128i vfitab;
1373 __m128i ifour = _mm_set1_epi32(4);
1374 __m128d rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
1375 real *vftab;
1376 __m128d dummy_mask,cutoff_mask;
1377 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
1378 __m128d one = _mm_set1_pd(1.0);
1379 __m128d two = _mm_set1_pd(2.0);
1380 x = xx[0];
1381 f = ff[0];
1382
1383 nri = nlist->nri;
1384 iinr = nlist->iinr;
1385 jindex = nlist->jindex;
1386 jjnr = nlist->jjnr;
1387 shiftidx = nlist->shift;
1388 gid = nlist->gid;
1389 shiftvec = fr->shift_vec[0];
1390 fshift = fr->fshift[0];
1391 facel = _mm_set1_pd(fr->epsfac);
1392 charge = mdatoms->chargeA;
1393 nvdwtype = fr->ntype;
1394 vdwparam = fr->nbfp;
1395 vdwtype = mdatoms->typeA;
1396
1397 vftab = kernel_data->table_elec->data;
1398 vftabscale = _mm_set1_pd(kernel_data->table_elec->scale);
1399
1400 /* Setup water-specific parameters */
1401 inr = nlist->iinr[0];
1402 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
1403 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
1404 iq3 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
1405 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
1406
1407 jq1 = _mm_set1_pd(charge[inr+1]);
1408 jq2 = _mm_set1_pd(charge[inr+2]);
1409 jq3 = _mm_set1_pd(charge[inr+3]);
1410 vdwjidx0A = 2*vdwtype[inr+0];
1411 c6_00 = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
1412 c12_00 = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
1413 qq11 = _mm_mul_pd(iq1,jq1);
1414 qq12 = _mm_mul_pd(iq1,jq2);
1415 qq13 = _mm_mul_pd(iq1,jq3);
1416 qq21 = _mm_mul_pd(iq2,jq1);
1417 qq22 = _mm_mul_pd(iq2,jq2);
1418 qq23 = _mm_mul_pd(iq2,jq3);
1419 qq31 = _mm_mul_pd(iq3,jq1);
1420 qq32 = _mm_mul_pd(iq3,jq2);
1421 qq33 = _mm_mul_pd(iq3,jq3);
1422
1423 /* Avoid stupid compiler warnings */
1424 jnrA = jnrB = 0;
1425 j_coord_offsetA = 0;
1426 j_coord_offsetB = 0;
1427
1428 outeriter = 0;
1429 inneriter = 0;
1430
1431 /* Start outer loop over neighborlists */
1432 for(iidx=0; iidx<nri; iidx++)
1433 {
1434 /* Load shift vector for this list */
1435 i_shift_offset = DIM*shiftidx[iidx];
1436
1437 /* Load limits for loop over neighbors */
1438 j_index_start = jindex[iidx];
1439 j_index_end = jindex[iidx+1];
1440
1441 /* Get outer coordinate index */
1442 inr = iinr[iidx];
1443 i_coord_offset = DIM*inr;
1444
1445 /* Load i particle coords and add shift vector */
1446 gmx_mm_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1447 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1448
1449 fix0 = _mm_setzero_pd();
1450 fiy0 = _mm_setzero_pd();
1451 fiz0 = _mm_setzero_pd();
1452 fix1 = _mm_setzero_pd();
1453 fiy1 = _mm_setzero_pd();
1454 fiz1 = _mm_setzero_pd();
1455 fix2 = _mm_setzero_pd();
1456 fiy2 = _mm_setzero_pd();
1457 fiz2 = _mm_setzero_pd();
1458 fix3 = _mm_setzero_pd();
1459 fiy3 = _mm_setzero_pd();
1460 fiz3 = _mm_setzero_pd();
1461
1462 /* Start inner kernel loop */
1463 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
1464 {
1465
1466 /* Get j neighbor index, and coordinate index */
1467 jnrA = jjnr[jidx];
1468 jnrB = jjnr[jidx+1];
1469 j_coord_offsetA = DIM*jnrA;
1470 j_coord_offsetB = DIM*jnrB;
1471
1472 /* load j atom coordinates */
1473 gmx_mm_load_4rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1474 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1475 &jy2,&jz2,&jx3,&jy3,&jz3);
1476
1477 /* Calculate displacement vector */
1478 dx00 = _mm_sub_pd(ix0,jx0);
1479 dy00 = _mm_sub_pd(iy0,jy0);
1480 dz00 = _mm_sub_pd(iz0,jz0);
1481 dx11 = _mm_sub_pd(ix1,jx1);
1482 dy11 = _mm_sub_pd(iy1,jy1);
1483 dz11 = _mm_sub_pd(iz1,jz1);
1484 dx12 = _mm_sub_pd(ix1,jx2);
1485 dy12 = _mm_sub_pd(iy1,jy2);
1486 dz12 = _mm_sub_pd(iz1,jz2);
1487 dx13 = _mm_sub_pd(ix1,jx3);
1488 dy13 = _mm_sub_pd(iy1,jy3);
1489 dz13 = _mm_sub_pd(iz1,jz3);
1490 dx21 = _mm_sub_pd(ix2,jx1);
1491 dy21 = _mm_sub_pd(iy2,jy1);
1492 dz21 = _mm_sub_pd(iz2,jz1);
1493 dx22 = _mm_sub_pd(ix2,jx2);
1494 dy22 = _mm_sub_pd(iy2,jy2);
1495 dz22 = _mm_sub_pd(iz2,jz2);
1496 dx23 = _mm_sub_pd(ix2,jx3);
1497 dy23 = _mm_sub_pd(iy2,jy3);
1498 dz23 = _mm_sub_pd(iz2,jz3);
1499 dx31 = _mm_sub_pd(ix3,jx1);
1500 dy31 = _mm_sub_pd(iy3,jy1);
1501 dz31 = _mm_sub_pd(iz3,jz1);
1502 dx32 = _mm_sub_pd(ix3,jx2);
1503 dy32 = _mm_sub_pd(iy3,jy2);
1504 dz32 = _mm_sub_pd(iz3,jz2);
1505 dx33 = _mm_sub_pd(ix3,jx3);
1506 dy33 = _mm_sub_pd(iy3,jy3);
1507 dz33 = _mm_sub_pd(iz3,jz3);
1508
1509 /* Calculate squared distance and things based on it */
1510 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1511 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1512 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1513 rsq13 = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
1514 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1515 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1516 rsq23 = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
1517 rsq31 = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
1518 rsq32 = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
1519 rsq33 = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
1520
1521 rinv11 = gmx_mm_invsqrt_pd(rsq11);
1522 rinv12 = gmx_mm_invsqrt_pd(rsq12);
1523 rinv13 = gmx_mm_invsqrt_pd(rsq13);
1524 rinv21 = gmx_mm_invsqrt_pd(rsq21);
1525 rinv22 = gmx_mm_invsqrt_pd(rsq22);
1526 rinv23 = gmx_mm_invsqrt_pd(rsq23);
1527 rinv31 = gmx_mm_invsqrt_pd(rsq31);
1528 rinv32 = gmx_mm_invsqrt_pd(rsq32);
1529 rinv33 = gmx_mm_invsqrt_pd(rsq33);
1530
1531 rinvsq00 = gmx_mm_inv_pd(rsq00);
1532
1533 fjx0 = _mm_setzero_pd();
1534 fjy0 = _mm_setzero_pd();
1535 fjz0 = _mm_setzero_pd();
1536 fjx1 = _mm_setzero_pd();
1537 fjy1 = _mm_setzero_pd();
1538 fjz1 = _mm_setzero_pd();
1539 fjx2 = _mm_setzero_pd();
1540 fjy2 = _mm_setzero_pd();
1541 fjz2 = _mm_setzero_pd();
1542 fjx3 = _mm_setzero_pd();
1543 fjy3 = _mm_setzero_pd();
1544 fjz3 = _mm_setzero_pd();
1545
1546 /**************************
1547 * CALCULATE INTERACTIONS *
1548 **************************/
1549
1550 /* LENNARD-JONES DISPERSION/REPULSION */
1551
1552 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1553 fvdw = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
1554
1555 fscal = fvdw;
1556
1557 /* Calculate temporary vectorial force */
1558 tx = _mm_mul_pd(fscal,dx00);
1559 ty = _mm_mul_pd(fscal,dy00);
1560 tz = _mm_mul_pd(fscal,dz00);
1561
1562 /* Update vectorial force */
1563 fix0 = _mm_add_pd(fix0,tx);
1564 fiy0 = _mm_add_pd(fiy0,ty);
1565 fiz0 = _mm_add_pd(fiz0,tz);
1566
1567 fjx0 = _mm_add_pd(fjx0,tx);
1568 fjy0 = _mm_add_pd(fjy0,ty);
1569 fjz0 = _mm_add_pd(fjz0,tz);
1570
1571 /**************************
1572 * CALCULATE INTERACTIONS *
1573 **************************/
1574
1575 r11 = _mm_mul_pd(rsq11,rinv11);
1576
1577 /* Calculate table index by multiplying r with table scale and truncate to integer */
1578 rt = _mm_mul_pd(r11,vftabscale);
1579 vfitab = _mm_cvttpd_epi32(rt);
1580 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1581 vfitab = _mm_slli_epi32(vfitab,2);
1582
1583 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1584 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1585 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1586 GMX_MM_TRANSPOSE2_PD(Y,F);
1587 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1588 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1589 GMX_MM_TRANSPOSE2_PD(G,H);
1590 Heps = _mm_mul_pd(vfeps,H);
1591 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1592 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1593 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
1594
1595 fscal = felec;
1596
1597 /* Calculate temporary vectorial force */
1598 tx = _mm_mul_pd(fscal,dx11);
1599 ty = _mm_mul_pd(fscal,dy11);
1600 tz = _mm_mul_pd(fscal,dz11);
1601
1602 /* Update vectorial force */
1603 fix1 = _mm_add_pd(fix1,tx);
1604 fiy1 = _mm_add_pd(fiy1,ty);
1605 fiz1 = _mm_add_pd(fiz1,tz);
1606
1607 fjx1 = _mm_add_pd(fjx1,tx);
1608 fjy1 = _mm_add_pd(fjy1,ty);
1609 fjz1 = _mm_add_pd(fjz1,tz);
1610
1611 /**************************
1612 * CALCULATE INTERACTIONS *
1613 **************************/
1614
1615 r12 = _mm_mul_pd(rsq12,rinv12);
1616
1617 /* Calculate table index by multiplying r with table scale and truncate to integer */
1618 rt = _mm_mul_pd(r12,vftabscale);
1619 vfitab = _mm_cvttpd_epi32(rt);
1620 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1621 vfitab = _mm_slli_epi32(vfitab,2);
1622
1623 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1624 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1625 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1626 GMX_MM_TRANSPOSE2_PD(Y,F);
1627 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1628 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1629 GMX_MM_TRANSPOSE2_PD(G,H);
1630 Heps = _mm_mul_pd(vfeps,H);
1631 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1632 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1633 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
1634
1635 fscal = felec;
1636
1637 /* Calculate temporary vectorial force */
1638 tx = _mm_mul_pd(fscal,dx12);
1639 ty = _mm_mul_pd(fscal,dy12);
1640 tz = _mm_mul_pd(fscal,dz12);
1641
1642 /* Update vectorial force */
1643 fix1 = _mm_add_pd(fix1,tx);
1644 fiy1 = _mm_add_pd(fiy1,ty);
1645 fiz1 = _mm_add_pd(fiz1,tz);
1646
1647 fjx2 = _mm_add_pd(fjx2,tx);
1648 fjy2 = _mm_add_pd(fjy2,ty);
1649 fjz2 = _mm_add_pd(fjz2,tz);
1650
1651 /**************************
1652 * CALCULATE INTERACTIONS *
1653 **************************/
1654
1655 r13 = _mm_mul_pd(rsq13,rinv13);
1656
1657 /* Calculate table index by multiplying r with table scale and truncate to integer */
1658 rt = _mm_mul_pd(r13,vftabscale);
1659 vfitab = _mm_cvttpd_epi32(rt);
1660 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1661 vfitab = _mm_slli_epi32(vfitab,2);
1662
1663 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1664 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1665 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1666 GMX_MM_TRANSPOSE2_PD(Y,F);
1667 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1668 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1669 GMX_MM_TRANSPOSE2_PD(G,H);
1670 Heps = _mm_mul_pd(vfeps,H);
1671 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1672 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1673 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq13,FF),_mm_mul_pd(vftabscale,rinv13)));
1674
1675 fscal = felec;
1676
1677 /* Calculate temporary vectorial force */
1678 tx = _mm_mul_pd(fscal,dx13);
1679 ty = _mm_mul_pd(fscal,dy13);
1680 tz = _mm_mul_pd(fscal,dz13);
1681
1682 /* Update vectorial force */
1683 fix1 = _mm_add_pd(fix1,tx);
1684 fiy1 = _mm_add_pd(fiy1,ty);
1685 fiz1 = _mm_add_pd(fiz1,tz);
1686
1687 fjx3 = _mm_add_pd(fjx3,tx);
1688 fjy3 = _mm_add_pd(fjy3,ty);
1689 fjz3 = _mm_add_pd(fjz3,tz);
1690
1691 /**************************
1692 * CALCULATE INTERACTIONS *
1693 **************************/
1694
1695 r21 = _mm_mul_pd(rsq21,rinv21);
1696
1697 /* Calculate table index by multiplying r with table scale and truncate to integer */
1698 rt = _mm_mul_pd(r21,vftabscale);
1699 vfitab = _mm_cvttpd_epi32(rt);
1700 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1701 vfitab = _mm_slli_epi32(vfitab,2);
1702
1703 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1704 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1705 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1706 GMX_MM_TRANSPOSE2_PD(Y,F);
1707 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1708 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1709 GMX_MM_TRANSPOSE2_PD(G,H);
1710 Heps = _mm_mul_pd(vfeps,H);
1711 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1712 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1713 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
1714
1715 fscal = felec;
1716
1717 /* Calculate temporary vectorial force */
1718 tx = _mm_mul_pd(fscal,dx21);
1719 ty = _mm_mul_pd(fscal,dy21);
1720 tz = _mm_mul_pd(fscal,dz21);
1721
1722 /* Update vectorial force */
1723 fix2 = _mm_add_pd(fix2,tx);
1724 fiy2 = _mm_add_pd(fiy2,ty);
1725 fiz2 = _mm_add_pd(fiz2,tz);
1726
1727 fjx1 = _mm_add_pd(fjx1,tx);
1728 fjy1 = _mm_add_pd(fjy1,ty);
1729 fjz1 = _mm_add_pd(fjz1,tz);
1730
1731 /**************************
1732 * CALCULATE INTERACTIONS *
1733 **************************/
1734
1735 r22 = _mm_mul_pd(rsq22,rinv22);
1736
1737 /* Calculate table index by multiplying r with table scale and truncate to integer */
1738 rt = _mm_mul_pd(r22,vftabscale);
1739 vfitab = _mm_cvttpd_epi32(rt);
1740 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1741 vfitab = _mm_slli_epi32(vfitab,2);
1742
1743 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1744 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1745 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1746 GMX_MM_TRANSPOSE2_PD(Y,F);
1747 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1748 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1749 GMX_MM_TRANSPOSE2_PD(G,H);
1750 Heps = _mm_mul_pd(vfeps,H);
1751 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1752 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1753 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
1754
1755 fscal = felec;
1756
1757 /* Calculate temporary vectorial force */
1758 tx = _mm_mul_pd(fscal,dx22);
1759 ty = _mm_mul_pd(fscal,dy22);
1760 tz = _mm_mul_pd(fscal,dz22);
1761
1762 /* Update vectorial force */
1763 fix2 = _mm_add_pd(fix2,tx);
1764 fiy2 = _mm_add_pd(fiy2,ty);
1765 fiz2 = _mm_add_pd(fiz2,tz);
1766
1767 fjx2 = _mm_add_pd(fjx2,tx);
1768 fjy2 = _mm_add_pd(fjy2,ty);
1769 fjz2 = _mm_add_pd(fjz2,tz);
1770
1771 /**************************
1772 * CALCULATE INTERACTIONS *
1773 **************************/
1774
1775 r23 = _mm_mul_pd(rsq23,rinv23);
1776
1777 /* Calculate table index by multiplying r with table scale and truncate to integer */
1778 rt = _mm_mul_pd(r23,vftabscale);
1779 vfitab = _mm_cvttpd_epi32(rt);
1780 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1781 vfitab = _mm_slli_epi32(vfitab,2);
1782
1783 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1784 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1785 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1786 GMX_MM_TRANSPOSE2_PD(Y,F);
1787 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1788 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1789 GMX_MM_TRANSPOSE2_PD(G,H);
1790 Heps = _mm_mul_pd(vfeps,H);
1791 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1792 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1793 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq23,FF),_mm_mul_pd(vftabscale,rinv23)));
1794
1795 fscal = felec;
1796
1797 /* Calculate temporary vectorial force */
1798 tx = _mm_mul_pd(fscal,dx23);
1799 ty = _mm_mul_pd(fscal,dy23);
1800 tz = _mm_mul_pd(fscal,dz23);
1801
1802 /* Update vectorial force */
1803 fix2 = _mm_add_pd(fix2,tx);
1804 fiy2 = _mm_add_pd(fiy2,ty);
1805 fiz2 = _mm_add_pd(fiz2,tz);
1806
1807 fjx3 = _mm_add_pd(fjx3,tx);
1808 fjy3 = _mm_add_pd(fjy3,ty);
1809 fjz3 = _mm_add_pd(fjz3,tz);
1810
1811 /**************************
1812 * CALCULATE INTERACTIONS *
1813 **************************/
1814
1815 r31 = _mm_mul_pd(rsq31,rinv31);
1816
1817 /* Calculate table index by multiplying r with table scale and truncate to integer */
1818 rt = _mm_mul_pd(r31,vftabscale);
1819 vfitab = _mm_cvttpd_epi32(rt);
1820 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1821 vfitab = _mm_slli_epi32(vfitab,2);
1822
1823 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1824 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1825 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1826 GMX_MM_TRANSPOSE2_PD(Y,F);
1827 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1828 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1829 GMX_MM_TRANSPOSE2_PD(G,H);
1830 Heps = _mm_mul_pd(vfeps,H);
1831 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1832 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1833 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq31,FF),_mm_mul_pd(vftabscale,rinv31)));
1834
1835 fscal = felec;
1836
1837 /* Calculate temporary vectorial force */
1838 tx = _mm_mul_pd(fscal,dx31);
1839 ty = _mm_mul_pd(fscal,dy31);
1840 tz = _mm_mul_pd(fscal,dz31);
1841
1842 /* Update vectorial force */
1843 fix3 = _mm_add_pd(fix3,tx);
1844 fiy3 = _mm_add_pd(fiy3,ty);
1845 fiz3 = _mm_add_pd(fiz3,tz);
1846
1847 fjx1 = _mm_add_pd(fjx1,tx);
1848 fjy1 = _mm_add_pd(fjy1,ty);
1849 fjz1 = _mm_add_pd(fjz1,tz);
1850
1851 /**************************
1852 * CALCULATE INTERACTIONS *
1853 **************************/
1854
1855 r32 = _mm_mul_pd(rsq32,rinv32);
1856
1857 /* Calculate table index by multiplying r with table scale and truncate to integer */
1858 rt = _mm_mul_pd(r32,vftabscale);
1859 vfitab = _mm_cvttpd_epi32(rt);
1860 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1861 vfitab = _mm_slli_epi32(vfitab,2);
1862
1863 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1864 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1865 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1866 GMX_MM_TRANSPOSE2_PD(Y,F);
1867 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1868 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1869 GMX_MM_TRANSPOSE2_PD(G,H);
1870 Heps = _mm_mul_pd(vfeps,H);
1871 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1872 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1873 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq32,FF),_mm_mul_pd(vftabscale,rinv32)));
1874
1875 fscal = felec;
1876
1877 /* Calculate temporary vectorial force */
1878 tx = _mm_mul_pd(fscal,dx32);
1879 ty = _mm_mul_pd(fscal,dy32);
1880 tz = _mm_mul_pd(fscal,dz32);
1881
1882 /* Update vectorial force */
1883 fix3 = _mm_add_pd(fix3,tx);
1884 fiy3 = _mm_add_pd(fiy3,ty);
1885 fiz3 = _mm_add_pd(fiz3,tz);
1886
1887 fjx2 = _mm_add_pd(fjx2,tx);
1888 fjy2 = _mm_add_pd(fjy2,ty);
1889 fjz2 = _mm_add_pd(fjz2,tz);
1890
1891 /**************************
1892 * CALCULATE INTERACTIONS *
1893 **************************/
1894
1895 r33 = _mm_mul_pd(rsq33,rinv33);
1896
1897 /* Calculate table index by multiplying r with table scale and truncate to integer */
1898 rt = _mm_mul_pd(r33,vftabscale);
1899 vfitab = _mm_cvttpd_epi32(rt);
1900 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1901 vfitab = _mm_slli_epi32(vfitab,2);
1902
1903 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1904 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1905 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1906 GMX_MM_TRANSPOSE2_PD(Y,F);
1907 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1908 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1909 GMX_MM_TRANSPOSE2_PD(G,H);
1910 Heps = _mm_mul_pd(vfeps,H);
1911 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1912 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1913 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq33,FF),_mm_mul_pd(vftabscale,rinv33)));
1914
1915 fscal = felec;
1916
1917 /* Calculate temporary vectorial force */
1918 tx = _mm_mul_pd(fscal,dx33);
1919 ty = _mm_mul_pd(fscal,dy33);
1920 tz = _mm_mul_pd(fscal,dz33);
1921
1922 /* Update vectorial force */
1923 fix3 = _mm_add_pd(fix3,tx);
1924 fiy3 = _mm_add_pd(fiy3,ty);
1925 fiz3 = _mm_add_pd(fiz3,tz);
1926
1927 fjx3 = _mm_add_pd(fjx3,tx);
1928 fjy3 = _mm_add_pd(fjy3,ty);
1929 fjz3 = _mm_add_pd(fjz3,tz);
1930
1931 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);
1932
1933 /* Inner loop uses 381 flops */
1934 }
1935
1936 if(jidx<j_index_end)
1937 {
1938
1939 jnrA = jjnr[jidx];
1940 j_coord_offsetA = DIM*jnrA;
1941
1942 /* load j atom coordinates */
1943 gmx_mm_load_4rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
1944 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1945 &jy2,&jz2,&jx3,&jy3,&jz3);
1946
1947 /* Calculate displacement vector */
1948 dx00 = _mm_sub_pd(ix0,jx0);
1949 dy00 = _mm_sub_pd(iy0,jy0);
1950 dz00 = _mm_sub_pd(iz0,jz0);
1951 dx11 = _mm_sub_pd(ix1,jx1);
1952 dy11 = _mm_sub_pd(iy1,jy1);
1953 dz11 = _mm_sub_pd(iz1,jz1);
1954 dx12 = _mm_sub_pd(ix1,jx2);
1955 dy12 = _mm_sub_pd(iy1,jy2);
1956 dz12 = _mm_sub_pd(iz1,jz2);
1957 dx13 = _mm_sub_pd(ix1,jx3);
1958 dy13 = _mm_sub_pd(iy1,jy3);
1959 dz13 = _mm_sub_pd(iz1,jz3);
1960 dx21 = _mm_sub_pd(ix2,jx1);
1961 dy21 = _mm_sub_pd(iy2,jy1);
1962 dz21 = _mm_sub_pd(iz2,jz1);
1963 dx22 = _mm_sub_pd(ix2,jx2);
1964 dy22 = _mm_sub_pd(iy2,jy2);
1965 dz22 = _mm_sub_pd(iz2,jz2);
1966 dx23 = _mm_sub_pd(ix2,jx3);
1967 dy23 = _mm_sub_pd(iy2,jy3);
1968 dz23 = _mm_sub_pd(iz2,jz3);
1969 dx31 = _mm_sub_pd(ix3,jx1);
1970 dy31 = _mm_sub_pd(iy3,jy1);
1971 dz31 = _mm_sub_pd(iz3,jz1);
1972 dx32 = _mm_sub_pd(ix3,jx2);
1973 dy32 = _mm_sub_pd(iy3,jy2);
1974 dz32 = _mm_sub_pd(iz3,jz2);
1975 dx33 = _mm_sub_pd(ix3,jx3);
1976 dy33 = _mm_sub_pd(iy3,jy3);
1977 dz33 = _mm_sub_pd(iz3,jz3);
1978
1979 /* Calculate squared distance and things based on it */
1980 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1981 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1982 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1983 rsq13 = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
1984 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1985 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1986 rsq23 = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
1987 rsq31 = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
1988 rsq32 = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
1989 rsq33 = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
1990
1991 rinv11 = gmx_mm_invsqrt_pd(rsq11);
1992 rinv12 = gmx_mm_invsqrt_pd(rsq12);
1993 rinv13 = gmx_mm_invsqrt_pd(rsq13);
1994 rinv21 = gmx_mm_invsqrt_pd(rsq21);
1995 rinv22 = gmx_mm_invsqrt_pd(rsq22);
1996 rinv23 = gmx_mm_invsqrt_pd(rsq23);
1997 rinv31 = gmx_mm_invsqrt_pd(rsq31);
1998 rinv32 = gmx_mm_invsqrt_pd(rsq32);
1999 rinv33 = gmx_mm_invsqrt_pd(rsq33);
2000
2001 rinvsq00 = gmx_mm_inv_pd(rsq00);
2002
2003 fjx0 = _mm_setzero_pd();
2004 fjy0 = _mm_setzero_pd();
2005 fjz0 = _mm_setzero_pd();
2006 fjx1 = _mm_setzero_pd();
2007 fjy1 = _mm_setzero_pd();
2008 fjz1 = _mm_setzero_pd();
2009 fjx2 = _mm_setzero_pd();
2010 fjy2 = _mm_setzero_pd();
2011 fjz2 = _mm_setzero_pd();
2012 fjx3 = _mm_setzero_pd();
2013 fjy3 = _mm_setzero_pd();
2014 fjz3 = _mm_setzero_pd();
2015
2016 /**************************
2017 * CALCULATE INTERACTIONS *
2018 **************************/
2019
2020 /* LENNARD-JONES DISPERSION/REPULSION */
2021
2022 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
2023 fvdw = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
2024
2025 fscal = fvdw;
2026
2027 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2028
2029 /* Calculate temporary vectorial force */
2030 tx = _mm_mul_pd(fscal,dx00);
2031 ty = _mm_mul_pd(fscal,dy00);
2032 tz = _mm_mul_pd(fscal,dz00);
2033
2034 /* Update vectorial force */
2035 fix0 = _mm_add_pd(fix0,tx);
2036 fiy0 = _mm_add_pd(fiy0,ty);
2037 fiz0 = _mm_add_pd(fiz0,tz);
2038
2039 fjx0 = _mm_add_pd(fjx0,tx);
2040 fjy0 = _mm_add_pd(fjy0,ty);
2041 fjz0 = _mm_add_pd(fjz0,tz);
2042
2043 /**************************
2044 * CALCULATE INTERACTIONS *
2045 **************************/
2046
2047 r11 = _mm_mul_pd(rsq11,rinv11);
2048
2049 /* Calculate table index by multiplying r with table scale and truncate to integer */
2050 rt = _mm_mul_pd(r11,vftabscale);
2051 vfitab = _mm_cvttpd_epi32(rt);
2052 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2053 vfitab = _mm_slli_epi32(vfitab,2);
2054
2055 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2056 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2057 F = _mm_setzero_pd();
2058 GMX_MM_TRANSPOSE2_PD(Y,F);
2059 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2060 H = _mm_setzero_pd();
2061 GMX_MM_TRANSPOSE2_PD(G,H);
2062 Heps = _mm_mul_pd(vfeps,H);
2063 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2064 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2065 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
2066
2067 fscal = felec;
2068
2069 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2070
2071 /* Calculate temporary vectorial force */
2072 tx = _mm_mul_pd(fscal,dx11);
2073 ty = _mm_mul_pd(fscal,dy11);
2074 tz = _mm_mul_pd(fscal,dz11);
2075
2076 /* Update vectorial force */
2077 fix1 = _mm_add_pd(fix1,tx);
2078 fiy1 = _mm_add_pd(fiy1,ty);
2079 fiz1 = _mm_add_pd(fiz1,tz);
2080
2081 fjx1 = _mm_add_pd(fjx1,tx);
2082 fjy1 = _mm_add_pd(fjy1,ty);
2083 fjz1 = _mm_add_pd(fjz1,tz);
2084
2085 /**************************
2086 * CALCULATE INTERACTIONS *
2087 **************************/
2088
2089 r12 = _mm_mul_pd(rsq12,rinv12);
2090
2091 /* Calculate table index by multiplying r with table scale and truncate to integer */
2092 rt = _mm_mul_pd(r12,vftabscale);
2093 vfitab = _mm_cvttpd_epi32(rt);
2094 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2095 vfitab = _mm_slli_epi32(vfitab,2);
2096
2097 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2098 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2099 F = _mm_setzero_pd();
2100 GMX_MM_TRANSPOSE2_PD(Y,F);
2101 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2102 H = _mm_setzero_pd();
2103 GMX_MM_TRANSPOSE2_PD(G,H);
2104 Heps = _mm_mul_pd(vfeps,H);
2105 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2106 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2107 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
2108
2109 fscal = felec;
2110
2111 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2112
2113 /* Calculate temporary vectorial force */
2114 tx = _mm_mul_pd(fscal,dx12);
2115 ty = _mm_mul_pd(fscal,dy12);
2116 tz = _mm_mul_pd(fscal,dz12);
2117
2118 /* Update vectorial force */
2119 fix1 = _mm_add_pd(fix1,tx);
2120 fiy1 = _mm_add_pd(fiy1,ty);
2121 fiz1 = _mm_add_pd(fiz1,tz);
2122
2123 fjx2 = _mm_add_pd(fjx2,tx);
2124 fjy2 = _mm_add_pd(fjy2,ty);
2125 fjz2 = _mm_add_pd(fjz2,tz);
2126
2127 /**************************
2128 * CALCULATE INTERACTIONS *
2129 **************************/
2130
2131 r13 = _mm_mul_pd(rsq13,rinv13);
2132
2133 /* Calculate table index by multiplying r with table scale and truncate to integer */
2134 rt = _mm_mul_pd(r13,vftabscale);
2135 vfitab = _mm_cvttpd_epi32(rt);
2136 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2137 vfitab = _mm_slli_epi32(vfitab,2);
2138
2139 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2140 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2141 F = _mm_setzero_pd();
2142 GMX_MM_TRANSPOSE2_PD(Y,F);
2143 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2144 H = _mm_setzero_pd();
2145 GMX_MM_TRANSPOSE2_PD(G,H);
2146 Heps = _mm_mul_pd(vfeps,H);
2147 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2148 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2149 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq13,FF),_mm_mul_pd(vftabscale,rinv13)));
2150
2151 fscal = felec;
2152
2153 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2154
2155 /* Calculate temporary vectorial force */
2156 tx = _mm_mul_pd(fscal,dx13);
2157 ty = _mm_mul_pd(fscal,dy13);
2158 tz = _mm_mul_pd(fscal,dz13);
2159
2160 /* Update vectorial force */
2161 fix1 = _mm_add_pd(fix1,tx);
2162 fiy1 = _mm_add_pd(fiy1,ty);
2163 fiz1 = _mm_add_pd(fiz1,tz);
2164
2165 fjx3 = _mm_add_pd(fjx3,tx);
2166 fjy3 = _mm_add_pd(fjy3,ty);
2167 fjz3 = _mm_add_pd(fjz3,tz);
2168
2169 /**************************
2170 * CALCULATE INTERACTIONS *
2171 **************************/
2172
2173 r21 = _mm_mul_pd(rsq21,rinv21);
2174
2175 /* Calculate table index by multiplying r with table scale and truncate to integer */
2176 rt = _mm_mul_pd(r21,vftabscale);
2177 vfitab = _mm_cvttpd_epi32(rt);
2178 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2179 vfitab = _mm_slli_epi32(vfitab,2);
2180
2181 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2182 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2183 F = _mm_setzero_pd();
2184 GMX_MM_TRANSPOSE2_PD(Y,F);
2185 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2186 H = _mm_setzero_pd();
2187 GMX_MM_TRANSPOSE2_PD(G,H);
2188 Heps = _mm_mul_pd(vfeps,H);
2189 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2190 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2191 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
2192
2193 fscal = felec;
2194
2195 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2196
2197 /* Calculate temporary vectorial force */
2198 tx = _mm_mul_pd(fscal,dx21);
2199 ty = _mm_mul_pd(fscal,dy21);
2200 tz = _mm_mul_pd(fscal,dz21);
2201
2202 /* Update vectorial force */
2203 fix2 = _mm_add_pd(fix2,tx);
2204 fiy2 = _mm_add_pd(fiy2,ty);
2205 fiz2 = _mm_add_pd(fiz2,tz);
2206
2207 fjx1 = _mm_add_pd(fjx1,tx);
2208 fjy1 = _mm_add_pd(fjy1,ty);
2209 fjz1 = _mm_add_pd(fjz1,tz);
2210
2211 /**************************
2212 * CALCULATE INTERACTIONS *
2213 **************************/
2214
2215 r22 = _mm_mul_pd(rsq22,rinv22);
2216
2217 /* Calculate table index by multiplying r with table scale and truncate to integer */
2218 rt = _mm_mul_pd(r22,vftabscale);
2219 vfitab = _mm_cvttpd_epi32(rt);
2220 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2221 vfitab = _mm_slli_epi32(vfitab,2);
2222
2223 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2224 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2225 F = _mm_setzero_pd();
2226 GMX_MM_TRANSPOSE2_PD(Y,F);
2227 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2228 H = _mm_setzero_pd();
2229 GMX_MM_TRANSPOSE2_PD(G,H);
2230 Heps = _mm_mul_pd(vfeps,H);
2231 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2232 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2233 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
2234
2235 fscal = felec;
2236
2237 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2238
2239 /* Calculate temporary vectorial force */
2240 tx = _mm_mul_pd(fscal,dx22);
2241 ty = _mm_mul_pd(fscal,dy22);
2242 tz = _mm_mul_pd(fscal,dz22);
2243
2244 /* Update vectorial force */
2245 fix2 = _mm_add_pd(fix2,tx);
2246 fiy2 = _mm_add_pd(fiy2,ty);
2247 fiz2 = _mm_add_pd(fiz2,tz);
2248
2249 fjx2 = _mm_add_pd(fjx2,tx);
2250 fjy2 = _mm_add_pd(fjy2,ty);
2251 fjz2 = _mm_add_pd(fjz2,tz);
2252
2253 /**************************
2254 * CALCULATE INTERACTIONS *
2255 **************************/
2256
2257 r23 = _mm_mul_pd(rsq23,rinv23);
2258
2259 /* Calculate table index by multiplying r with table scale and truncate to integer */
2260 rt = _mm_mul_pd(r23,vftabscale);
2261 vfitab = _mm_cvttpd_epi32(rt);
2262 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2263 vfitab = _mm_slli_epi32(vfitab,2);
2264
2265 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2266 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2267 F = _mm_setzero_pd();
2268 GMX_MM_TRANSPOSE2_PD(Y,F);
2269 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2270 H = _mm_setzero_pd();
2271 GMX_MM_TRANSPOSE2_PD(G,H);
2272 Heps = _mm_mul_pd(vfeps,H);
2273 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2274 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2275 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq23,FF),_mm_mul_pd(vftabscale,rinv23)));
2276
2277 fscal = felec;
2278
2279 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2280
2281 /* Calculate temporary vectorial force */
2282 tx = _mm_mul_pd(fscal,dx23);
2283 ty = _mm_mul_pd(fscal,dy23);
2284 tz = _mm_mul_pd(fscal,dz23);
2285
2286 /* Update vectorial force */
2287 fix2 = _mm_add_pd(fix2,tx);
2288 fiy2 = _mm_add_pd(fiy2,ty);
2289 fiz2 = _mm_add_pd(fiz2,tz);
2290
2291 fjx3 = _mm_add_pd(fjx3,tx);
2292 fjy3 = _mm_add_pd(fjy3,ty);
2293 fjz3 = _mm_add_pd(fjz3,tz);
2294
2295 /**************************
2296 * CALCULATE INTERACTIONS *
2297 **************************/
2298
2299 r31 = _mm_mul_pd(rsq31,rinv31);
2300
2301 /* Calculate table index by multiplying r with table scale and truncate to integer */
2302 rt = _mm_mul_pd(r31,vftabscale);
2303 vfitab = _mm_cvttpd_epi32(rt);
2304 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2305 vfitab = _mm_slli_epi32(vfitab,2);
2306
2307 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2308 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2309 F = _mm_setzero_pd();
2310 GMX_MM_TRANSPOSE2_PD(Y,F);
2311 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2312 H = _mm_setzero_pd();
2313 GMX_MM_TRANSPOSE2_PD(G,H);
2314 Heps = _mm_mul_pd(vfeps,H);
2315 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2316 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2317 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq31,FF),_mm_mul_pd(vftabscale,rinv31)));
2318
2319 fscal = felec;
2320
2321 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2322
2323 /* Calculate temporary vectorial force */
2324 tx = _mm_mul_pd(fscal,dx31);
2325 ty = _mm_mul_pd(fscal,dy31);
2326 tz = _mm_mul_pd(fscal,dz31);
2327
2328 /* Update vectorial force */
2329 fix3 = _mm_add_pd(fix3,tx);
2330 fiy3 = _mm_add_pd(fiy3,ty);
2331 fiz3 = _mm_add_pd(fiz3,tz);
2332
2333 fjx1 = _mm_add_pd(fjx1,tx);
2334 fjy1 = _mm_add_pd(fjy1,ty);
2335 fjz1 = _mm_add_pd(fjz1,tz);
2336
2337 /**************************
2338 * CALCULATE INTERACTIONS *
2339 **************************/
2340
2341 r32 = _mm_mul_pd(rsq32,rinv32);
2342
2343 /* Calculate table index by multiplying r with table scale and truncate to integer */
2344 rt = _mm_mul_pd(r32,vftabscale);
2345 vfitab = _mm_cvttpd_epi32(rt);
2346 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2347 vfitab = _mm_slli_epi32(vfitab,2);
2348
2349 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2350 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2351 F = _mm_setzero_pd();
2352 GMX_MM_TRANSPOSE2_PD(Y,F);
2353 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2354 H = _mm_setzero_pd();
2355 GMX_MM_TRANSPOSE2_PD(G,H);
2356 Heps = _mm_mul_pd(vfeps,H);
2357 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2358 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2359 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq32,FF),_mm_mul_pd(vftabscale,rinv32)));
2360
2361 fscal = felec;
2362
2363 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2364
2365 /* Calculate temporary vectorial force */
2366 tx = _mm_mul_pd(fscal,dx32);
2367 ty = _mm_mul_pd(fscal,dy32);
2368 tz = _mm_mul_pd(fscal,dz32);
2369
2370 /* Update vectorial force */
2371 fix3 = _mm_add_pd(fix3,tx);
2372 fiy3 = _mm_add_pd(fiy3,ty);
2373 fiz3 = _mm_add_pd(fiz3,tz);
2374
2375 fjx2 = _mm_add_pd(fjx2,tx);
2376 fjy2 = _mm_add_pd(fjy2,ty);
2377 fjz2 = _mm_add_pd(fjz2,tz);
2378
2379 /**************************
2380 * CALCULATE INTERACTIONS *
2381 **************************/
2382
2383 r33 = _mm_mul_pd(rsq33,rinv33);
2384
2385 /* Calculate table index by multiplying r with table scale and truncate to integer */
2386 rt = _mm_mul_pd(r33,vftabscale);
2387 vfitab = _mm_cvttpd_epi32(rt);
2388 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2389 vfitab = _mm_slli_epi32(vfitab,2);
2390
2391 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2392 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2393 F = _mm_setzero_pd();
2394 GMX_MM_TRANSPOSE2_PD(Y,F);
2395 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2396 H = _mm_setzero_pd();
2397 GMX_MM_TRANSPOSE2_PD(G,H);
2398 Heps = _mm_mul_pd(vfeps,H);
2399 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2400 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2401 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq33,FF),_mm_mul_pd(vftabscale,rinv33)));
2402
2403 fscal = felec;
2404
2405 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2406
2407 /* Calculate temporary vectorial force */
2408 tx = _mm_mul_pd(fscal,dx33);
2409 ty = _mm_mul_pd(fscal,dy33);
2410 tz = _mm_mul_pd(fscal,dz33);
2411
2412 /* Update vectorial force */
2413 fix3 = _mm_add_pd(fix3,tx);
2414 fiy3 = _mm_add_pd(fiy3,ty);
2415 fiz3 = _mm_add_pd(fiz3,tz);
2416
2417 fjx3 = _mm_add_pd(fjx3,tx);
2418 fjy3 = _mm_add_pd(fjy3,ty);
2419 fjz3 = _mm_add_pd(fjz3,tz);
2420
2421 gmx_mm_decrement_4rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2422
2423 /* Inner loop uses 381 flops */
2424 }
2425
2426 /* End of innermost loop */
2427
2428 gmx_mm_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2429 f+i_coord_offset,fshift+i_shift_offset);
2430
2431 /* Increment number of inner iterations */
2432 inneriter += j_index_end - j_index_start;
2433
2434 /* Outer loop uses 24 flops */
2435 }
2436
2437 /* Increment number of outer iterations */
2438 outeriter += nri;
2439
2440 /* Update outer/inner flops */
2441
2442 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*381);
2443 }
The ultimate molecular dynamics simulation package