?  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_VdwCSTab_GeomW3W3_sse2_double.c
blobd6f8849a14c3f48242450d49ff1ccb9946ca5022
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_VdwCSTab_GeomW3W3_VF_sse2_double
53 * Electrostatics interaction: CubicSplineTable
54 * VdW interaction: CubicSplineTable
55 * Geometry: Water3-Water3
56 * Calculate force/pot: PotentialAndForce
57 */
58 void
59 nb_kernel_ElecCSTab_VdwCSTab_GeomW3W3_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 vdwjidx0A,vdwjidx0B;
88 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
89 int vdwjidx1A,vdwjidx1B;
90 __m128d jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
91 int vdwjidx2A,vdwjidx2B;
92 __m128d jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
93 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
94 __m128d dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
95 __m128d dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
96 __m128d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
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 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
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 velec,felec,velecsum,facel,crf,krf,krf2;
103 real *charge;
104 int nvdwtype;
105 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
106 int *vdwtype;
107 real *vdwparam;
108 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
109 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
110 __m128i vfitab;
111 __m128i ifour = _mm_set1_epi32(4);
112 __m128d rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
113 real *vftab;
114 __m128d dummy_mask,cutoff_mask;
115 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
116 __m128d one = _mm_set1_pd(1.0);
117 __m128d two = _mm_set1_pd(2.0);
118 x = xx[0];
119 f = ff[0];
120
121 nri = nlist->nri;
122 iinr = nlist->iinr;
123 jindex = nlist->jindex;
124 jjnr = nlist->jjnr;
125 shiftidx = nlist->shift;
126 gid = nlist->gid;
127 shiftvec = fr->shift_vec[0];
128 fshift = fr->fshift[0];
129 facel = _mm_set1_pd(fr->epsfac);
130 charge = mdatoms->chargeA;
131 nvdwtype = fr->ntype;
132 vdwparam = fr->nbfp;
133 vdwtype = mdatoms->typeA;
134
135 vftab = kernel_data->table_elec_vdw->data;
136 vftabscale = _mm_set1_pd(kernel_data->table_elec_vdw->scale);
137
138 /* Setup water-specific parameters */
139 inr = nlist->iinr[0];
140 iq0 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
141 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
142 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
143 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
144
145 jq0 = _mm_set1_pd(charge[inr+0]);
146 jq1 = _mm_set1_pd(charge[inr+1]);
147 jq2 = _mm_set1_pd(charge[inr+2]);
148 vdwjidx0A = 2*vdwtype[inr+0];
149 qq00 = _mm_mul_pd(iq0,jq0);
150 c6_00 = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
151 c12_00 = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
152 qq01 = _mm_mul_pd(iq0,jq1);
153 qq02 = _mm_mul_pd(iq0,jq2);
154 qq10 = _mm_mul_pd(iq1,jq0);
155 qq11 = _mm_mul_pd(iq1,jq1);
156 qq12 = _mm_mul_pd(iq1,jq2);
157 qq20 = _mm_mul_pd(iq2,jq0);
158 qq21 = _mm_mul_pd(iq2,jq1);
159 qq22 = _mm_mul_pd(iq2,jq2);
160
161 /* Avoid stupid compiler warnings */
162 jnrA = jnrB = 0;
163 j_coord_offsetA = 0;
164 j_coord_offsetB = 0;
165
166 outeriter = 0;
167 inneriter = 0;
168
169 /* Start outer loop over neighborlists */
170 for(iidx=0; iidx<nri; iidx++)
171 {
172 /* Load shift vector for this list */
173 i_shift_offset = DIM*shiftidx[iidx];
174
175 /* Load limits for loop over neighbors */
176 j_index_start = jindex[iidx];
177 j_index_end = jindex[iidx+1];
178
179 /* Get outer coordinate index */
180 inr = iinr[iidx];
181 i_coord_offset = DIM*inr;
182
183 /* Load i particle coords and add shift vector */
184 gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
185 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
186
187 fix0 = _mm_setzero_pd();
188 fiy0 = _mm_setzero_pd();
189 fiz0 = _mm_setzero_pd();
190 fix1 = _mm_setzero_pd();
191 fiy1 = _mm_setzero_pd();
192 fiz1 = _mm_setzero_pd();
193 fix2 = _mm_setzero_pd();
194 fiy2 = _mm_setzero_pd();
195 fiz2 = _mm_setzero_pd();
196
197 /* Reset potential sums */
198 velecsum = _mm_setzero_pd();
199 vvdwsum = _mm_setzero_pd();
200
201 /* Start inner kernel loop */
202 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
203 {
204
205 /* Get j neighbor index, and coordinate index */
206 jnrA = jjnr[jidx];
207 jnrB = jjnr[jidx+1];
208 j_coord_offsetA = DIM*jnrA;
209 j_coord_offsetB = DIM*jnrB;
210
211 /* load j atom coordinates */
212 gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
213 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
214
215 /* Calculate displacement vector */
216 dx00 = _mm_sub_pd(ix0,jx0);
217 dy00 = _mm_sub_pd(iy0,jy0);
218 dz00 = _mm_sub_pd(iz0,jz0);
219 dx01 = _mm_sub_pd(ix0,jx1);
220 dy01 = _mm_sub_pd(iy0,jy1);
221 dz01 = _mm_sub_pd(iz0,jz1);
222 dx02 = _mm_sub_pd(ix0,jx2);
223 dy02 = _mm_sub_pd(iy0,jy2);
224 dz02 = _mm_sub_pd(iz0,jz2);
225 dx10 = _mm_sub_pd(ix1,jx0);
226 dy10 = _mm_sub_pd(iy1,jy0);
227 dz10 = _mm_sub_pd(iz1,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 dx20 = _mm_sub_pd(ix2,jx0);
235 dy20 = _mm_sub_pd(iy2,jy0);
236 dz20 = _mm_sub_pd(iz2,jz0);
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
244 /* Calculate squared distance and things based on it */
245 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
246 rsq01 = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
247 rsq02 = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
248 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
249 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
250 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
251 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
252 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
253 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
254
255 rinv00 = gmx_mm_invsqrt_pd(rsq00);
256 rinv01 = gmx_mm_invsqrt_pd(rsq01);
257 rinv02 = gmx_mm_invsqrt_pd(rsq02);
258 rinv10 = gmx_mm_invsqrt_pd(rsq10);
259 rinv11 = gmx_mm_invsqrt_pd(rsq11);
260 rinv12 = gmx_mm_invsqrt_pd(rsq12);
261 rinv20 = gmx_mm_invsqrt_pd(rsq20);
262 rinv21 = gmx_mm_invsqrt_pd(rsq21);
263 rinv22 = gmx_mm_invsqrt_pd(rsq22);
264
265 fjx0 = _mm_setzero_pd();
266 fjy0 = _mm_setzero_pd();
267 fjz0 = _mm_setzero_pd();
268 fjx1 = _mm_setzero_pd();
269 fjy1 = _mm_setzero_pd();
270 fjz1 = _mm_setzero_pd();
271 fjx2 = _mm_setzero_pd();
272 fjy2 = _mm_setzero_pd();
273 fjz2 = _mm_setzero_pd();
274
275 /**************************
276 * CALCULATE INTERACTIONS *
277 **************************/
278
279 r00 = _mm_mul_pd(rsq00,rinv00);
280
281 /* Calculate table index by multiplying r with table scale and truncate to integer */
282 rt = _mm_mul_pd(r00,vftabscale);
283 vfitab = _mm_cvttpd_epi32(rt);
284 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
285 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
286
287 /* CUBIC SPLINE TABLE ELECTROSTATICS */
288 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
289 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
290 GMX_MM_TRANSPOSE2_PD(Y,F);
291 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
292 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
293 GMX_MM_TRANSPOSE2_PD(G,H);
294 Heps = _mm_mul_pd(vfeps,H);
295 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
296 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
297 velec = _mm_mul_pd(qq00,VV);
298 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
299 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq00,FF),_mm_mul_pd(vftabscale,rinv00)));
300
301 /* CUBIC SPLINE TABLE DISPERSION */
302 vfitab = _mm_add_epi32(vfitab,ifour);
303 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
304 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
305 GMX_MM_TRANSPOSE2_PD(Y,F);
306 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
307 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
308 GMX_MM_TRANSPOSE2_PD(G,H);
309 Heps = _mm_mul_pd(vfeps,H);
310 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
311 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
312 vvdw6 = _mm_mul_pd(c6_00,VV);
313 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
314 fvdw6 = _mm_mul_pd(c6_00,FF);
315
316 /* CUBIC SPLINE TABLE REPULSION */
317 vfitab = _mm_add_epi32(vfitab,ifour);
318 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
319 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
320 GMX_MM_TRANSPOSE2_PD(Y,F);
321 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
322 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
323 GMX_MM_TRANSPOSE2_PD(G,H);
324 Heps = _mm_mul_pd(vfeps,H);
325 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
326 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
327 vvdw12 = _mm_mul_pd(c12_00,VV);
328 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
329 fvdw12 = _mm_mul_pd(c12_00,FF);
330 vvdw = _mm_add_pd(vvdw12,vvdw6);
331 fvdw = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
332
333 /* Update potential sum for this i atom from the interaction with this j atom. */
334 velecsum = _mm_add_pd(velecsum,velec);
335 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
336
337 fscal = _mm_add_pd(felec,fvdw);
338
339 /* Calculate temporary vectorial force */
340 tx = _mm_mul_pd(fscal,dx00);
341 ty = _mm_mul_pd(fscal,dy00);
342 tz = _mm_mul_pd(fscal,dz00);
343
344 /* Update vectorial force */
345 fix0 = _mm_add_pd(fix0,tx);
346 fiy0 = _mm_add_pd(fiy0,ty);
347 fiz0 = _mm_add_pd(fiz0,tz);
348
349 fjx0 = _mm_add_pd(fjx0,tx);
350 fjy0 = _mm_add_pd(fjy0,ty);
351 fjz0 = _mm_add_pd(fjz0,tz);
352
353 /**************************
354 * CALCULATE INTERACTIONS *
355 **************************/
356
357 r01 = _mm_mul_pd(rsq01,rinv01);
358
359 /* Calculate table index by multiplying r with table scale and truncate to integer */
360 rt = _mm_mul_pd(r01,vftabscale);
361 vfitab = _mm_cvttpd_epi32(rt);
362 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
363 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
364
365 /* CUBIC SPLINE TABLE ELECTROSTATICS */
366 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
367 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
368 GMX_MM_TRANSPOSE2_PD(Y,F);
369 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
370 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
371 GMX_MM_TRANSPOSE2_PD(G,H);
372 Heps = _mm_mul_pd(vfeps,H);
373 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
374 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
375 velec = _mm_mul_pd(qq01,VV);
376 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
377 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq01,FF),_mm_mul_pd(vftabscale,rinv01)));
378
379 /* Update potential sum for this i atom from the interaction with this j atom. */
380 velecsum = _mm_add_pd(velecsum,velec);
381
382 fscal = felec;
383
384 /* Calculate temporary vectorial force */
385 tx = _mm_mul_pd(fscal,dx01);
386 ty = _mm_mul_pd(fscal,dy01);
387 tz = _mm_mul_pd(fscal,dz01);
388
389 /* Update vectorial force */
390 fix0 = _mm_add_pd(fix0,tx);
391 fiy0 = _mm_add_pd(fiy0,ty);
392 fiz0 = _mm_add_pd(fiz0,tz);
393
394 fjx1 = _mm_add_pd(fjx1,tx);
395 fjy1 = _mm_add_pd(fjy1,ty);
396 fjz1 = _mm_add_pd(fjz1,tz);
397
398 /**************************
399 * CALCULATE INTERACTIONS *
400 **************************/
401
402 r02 = _mm_mul_pd(rsq02,rinv02);
403
404 /* Calculate table index by multiplying r with table scale and truncate to integer */
405 rt = _mm_mul_pd(r02,vftabscale);
406 vfitab = _mm_cvttpd_epi32(rt);
407 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
408 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
409
410 /* CUBIC SPLINE TABLE ELECTROSTATICS */
411 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
412 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
413 GMX_MM_TRANSPOSE2_PD(Y,F);
414 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
415 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
416 GMX_MM_TRANSPOSE2_PD(G,H);
417 Heps = _mm_mul_pd(vfeps,H);
418 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
419 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
420 velec = _mm_mul_pd(qq02,VV);
421 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
422 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq02,FF),_mm_mul_pd(vftabscale,rinv02)));
423
424 /* Update potential sum for this i atom from the interaction with this j atom. */
425 velecsum = _mm_add_pd(velecsum,velec);
426
427 fscal = felec;
428
429 /* Calculate temporary vectorial force */
430 tx = _mm_mul_pd(fscal,dx02);
431 ty = _mm_mul_pd(fscal,dy02);
432 tz = _mm_mul_pd(fscal,dz02);
433
434 /* Update vectorial force */
435 fix0 = _mm_add_pd(fix0,tx);
436 fiy0 = _mm_add_pd(fiy0,ty);
437 fiz0 = _mm_add_pd(fiz0,tz);
438
439 fjx2 = _mm_add_pd(fjx2,tx);
440 fjy2 = _mm_add_pd(fjy2,ty);
441 fjz2 = _mm_add_pd(fjz2,tz);
442
443 /**************************
444 * CALCULATE INTERACTIONS *
445 **************************/
446
447 r10 = _mm_mul_pd(rsq10,rinv10);
448
449 /* Calculate table index by multiplying r with table scale and truncate to integer */
450 rt = _mm_mul_pd(r10,vftabscale);
451 vfitab = _mm_cvttpd_epi32(rt);
452 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
453 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
454
455 /* CUBIC SPLINE TABLE ELECTROSTATICS */
456 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
457 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
458 GMX_MM_TRANSPOSE2_PD(Y,F);
459 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
460 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
461 GMX_MM_TRANSPOSE2_PD(G,H);
462 Heps = _mm_mul_pd(vfeps,H);
463 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
464 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
465 velec = _mm_mul_pd(qq10,VV);
466 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
467 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq10,FF),_mm_mul_pd(vftabscale,rinv10)));
468
469 /* Update potential sum for this i atom from the interaction with this j atom. */
470 velecsum = _mm_add_pd(velecsum,velec);
471
472 fscal = felec;
473
474 /* Calculate temporary vectorial force */
475 tx = _mm_mul_pd(fscal,dx10);
476 ty = _mm_mul_pd(fscal,dy10);
477 tz = _mm_mul_pd(fscal,dz10);
478
479 /* Update vectorial force */
480 fix1 = _mm_add_pd(fix1,tx);
481 fiy1 = _mm_add_pd(fiy1,ty);
482 fiz1 = _mm_add_pd(fiz1,tz);
483
484 fjx0 = _mm_add_pd(fjx0,tx);
485 fjy0 = _mm_add_pd(fjy0,ty);
486 fjz0 = _mm_add_pd(fjz0,tz);
487
488 /**************************
489 * CALCULATE INTERACTIONS *
490 **************************/
491
492 r11 = _mm_mul_pd(rsq11,rinv11);
493
494 /* Calculate table index by multiplying r with table scale and truncate to integer */
495 rt = _mm_mul_pd(r11,vftabscale);
496 vfitab = _mm_cvttpd_epi32(rt);
497 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
498 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
499
500 /* CUBIC SPLINE TABLE ELECTROSTATICS */
501 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
502 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
503 GMX_MM_TRANSPOSE2_PD(Y,F);
504 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
505 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
506 GMX_MM_TRANSPOSE2_PD(G,H);
507 Heps = _mm_mul_pd(vfeps,H);
508 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
509 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
510 velec = _mm_mul_pd(qq11,VV);
511 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
512 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
513
514 /* Update potential sum for this i atom from the interaction with this j atom. */
515 velecsum = _mm_add_pd(velecsum,velec);
516
517 fscal = felec;
518
519 /* Calculate temporary vectorial force */
520 tx = _mm_mul_pd(fscal,dx11);
521 ty = _mm_mul_pd(fscal,dy11);
522 tz = _mm_mul_pd(fscal,dz11);
523
524 /* Update vectorial force */
525 fix1 = _mm_add_pd(fix1,tx);
526 fiy1 = _mm_add_pd(fiy1,ty);
527 fiz1 = _mm_add_pd(fiz1,tz);
528
529 fjx1 = _mm_add_pd(fjx1,tx);
530 fjy1 = _mm_add_pd(fjy1,ty);
531 fjz1 = _mm_add_pd(fjz1,tz);
532
533 /**************************
534 * CALCULATE INTERACTIONS *
535 **************************/
536
537 r12 = _mm_mul_pd(rsq12,rinv12);
538
539 /* Calculate table index by multiplying r with table scale and truncate to integer */
540 rt = _mm_mul_pd(r12,vftabscale);
541 vfitab = _mm_cvttpd_epi32(rt);
542 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
543 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
544
545 /* CUBIC SPLINE TABLE ELECTROSTATICS */
546 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
547 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
548 GMX_MM_TRANSPOSE2_PD(Y,F);
549 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
550 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
551 GMX_MM_TRANSPOSE2_PD(G,H);
552 Heps = _mm_mul_pd(vfeps,H);
553 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
554 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
555 velec = _mm_mul_pd(qq12,VV);
556 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
557 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
558
559 /* Update potential sum for this i atom from the interaction with this j atom. */
560 velecsum = _mm_add_pd(velecsum,velec);
561
562 fscal = felec;
563
564 /* Calculate temporary vectorial force */
565 tx = _mm_mul_pd(fscal,dx12);
566 ty = _mm_mul_pd(fscal,dy12);
567 tz = _mm_mul_pd(fscal,dz12);
568
569 /* Update vectorial force */
570 fix1 = _mm_add_pd(fix1,tx);
571 fiy1 = _mm_add_pd(fiy1,ty);
572 fiz1 = _mm_add_pd(fiz1,tz);
573
574 fjx2 = _mm_add_pd(fjx2,tx);
575 fjy2 = _mm_add_pd(fjy2,ty);
576 fjz2 = _mm_add_pd(fjz2,tz);
577
578 /**************************
579 * CALCULATE INTERACTIONS *
580 **************************/
581
582 r20 = _mm_mul_pd(rsq20,rinv20);
583
584 /* Calculate table index by multiplying r with table scale and truncate to integer */
585 rt = _mm_mul_pd(r20,vftabscale);
586 vfitab = _mm_cvttpd_epi32(rt);
587 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
588 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
589
590 /* CUBIC SPLINE TABLE ELECTROSTATICS */
591 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
592 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
593 GMX_MM_TRANSPOSE2_PD(Y,F);
594 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
595 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
596 GMX_MM_TRANSPOSE2_PD(G,H);
597 Heps = _mm_mul_pd(vfeps,H);
598 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
599 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
600 velec = _mm_mul_pd(qq20,VV);
601 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
602 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq20,FF),_mm_mul_pd(vftabscale,rinv20)));
603
604 /* Update potential sum for this i atom from the interaction with this j atom. */
605 velecsum = _mm_add_pd(velecsum,velec);
606
607 fscal = felec;
608
609 /* Calculate temporary vectorial force */
610 tx = _mm_mul_pd(fscal,dx20);
611 ty = _mm_mul_pd(fscal,dy20);
612 tz = _mm_mul_pd(fscal,dz20);
613
614 /* Update vectorial force */
615 fix2 = _mm_add_pd(fix2,tx);
616 fiy2 = _mm_add_pd(fiy2,ty);
617 fiz2 = _mm_add_pd(fiz2,tz);
618
619 fjx0 = _mm_add_pd(fjx0,tx);
620 fjy0 = _mm_add_pd(fjy0,ty);
621 fjz0 = _mm_add_pd(fjz0,tz);
622
623 /**************************
624 * CALCULATE INTERACTIONS *
625 **************************/
626
627 r21 = _mm_mul_pd(rsq21,rinv21);
628
629 /* Calculate table index by multiplying r with table scale and truncate to integer */
630 rt = _mm_mul_pd(r21,vftabscale);
631 vfitab = _mm_cvttpd_epi32(rt);
632 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
633 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
634
635 /* CUBIC SPLINE TABLE ELECTROSTATICS */
636 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
637 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
638 GMX_MM_TRANSPOSE2_PD(Y,F);
639 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
640 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
641 GMX_MM_TRANSPOSE2_PD(G,H);
642 Heps = _mm_mul_pd(vfeps,H);
643 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
644 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
645 velec = _mm_mul_pd(qq21,VV);
646 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
647 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
648
649 /* Update potential sum for this i atom from the interaction with this j atom. */
650 velecsum = _mm_add_pd(velecsum,velec);
651
652 fscal = felec;
653
654 /* Calculate temporary vectorial force */
655 tx = _mm_mul_pd(fscal,dx21);
656 ty = _mm_mul_pd(fscal,dy21);
657 tz = _mm_mul_pd(fscal,dz21);
658
659 /* Update vectorial force */
660 fix2 = _mm_add_pd(fix2,tx);
661 fiy2 = _mm_add_pd(fiy2,ty);
662 fiz2 = _mm_add_pd(fiz2,tz);
663
664 fjx1 = _mm_add_pd(fjx1,tx);
665 fjy1 = _mm_add_pd(fjy1,ty);
666 fjz1 = _mm_add_pd(fjz1,tz);
667
668 /**************************
669 * CALCULATE INTERACTIONS *
670 **************************/
671
672 r22 = _mm_mul_pd(rsq22,rinv22);
673
674 /* Calculate table index by multiplying r with table scale and truncate to integer */
675 rt = _mm_mul_pd(r22,vftabscale);
676 vfitab = _mm_cvttpd_epi32(rt);
677 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
678 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
679
680 /* CUBIC SPLINE TABLE ELECTROSTATICS */
681 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
682 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
683 GMX_MM_TRANSPOSE2_PD(Y,F);
684 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
685 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
686 GMX_MM_TRANSPOSE2_PD(G,H);
687 Heps = _mm_mul_pd(vfeps,H);
688 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
689 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
690 velec = _mm_mul_pd(qq22,VV);
691 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
692 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
693
694 /* Update potential sum for this i atom from the interaction with this j atom. */
695 velecsum = _mm_add_pd(velecsum,velec);
696
697 fscal = felec;
698
699 /* Calculate temporary vectorial force */
700 tx = _mm_mul_pd(fscal,dx22);
701 ty = _mm_mul_pd(fscal,dy22);
702 tz = _mm_mul_pd(fscal,dz22);
703
704 /* Update vectorial force */
705 fix2 = _mm_add_pd(fix2,tx);
706 fiy2 = _mm_add_pd(fiy2,ty);
707 fiz2 = _mm_add_pd(fiz2,tz);
708
709 fjx2 = _mm_add_pd(fjx2,tx);
710 fjy2 = _mm_add_pd(fjy2,ty);
711 fjz2 = _mm_add_pd(fjz2,tz);
712
713 gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
714
715 /* Inner loop uses 417 flops */
716 }
717
718 if(jidx<j_index_end)
719 {
720
721 jnrA = jjnr[jidx];
722 j_coord_offsetA = DIM*jnrA;
723
724 /* load j atom coordinates */
725 gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
726 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
727
728 /* Calculate displacement vector */
729 dx00 = _mm_sub_pd(ix0,jx0);
730 dy00 = _mm_sub_pd(iy0,jy0);
731 dz00 = _mm_sub_pd(iz0,jz0);
732 dx01 = _mm_sub_pd(ix0,jx1);
733 dy01 = _mm_sub_pd(iy0,jy1);
734 dz01 = _mm_sub_pd(iz0,jz1);
735 dx02 = _mm_sub_pd(ix0,jx2);
736 dy02 = _mm_sub_pd(iy0,jy2);
737 dz02 = _mm_sub_pd(iz0,jz2);
738 dx10 = _mm_sub_pd(ix1,jx0);
739 dy10 = _mm_sub_pd(iy1,jy0);
740 dz10 = _mm_sub_pd(iz1,jz0);
741 dx11 = _mm_sub_pd(ix1,jx1);
742 dy11 = _mm_sub_pd(iy1,jy1);
743 dz11 = _mm_sub_pd(iz1,jz1);
744 dx12 = _mm_sub_pd(ix1,jx2);
745 dy12 = _mm_sub_pd(iy1,jy2);
746 dz12 = _mm_sub_pd(iz1,jz2);
747 dx20 = _mm_sub_pd(ix2,jx0);
748 dy20 = _mm_sub_pd(iy2,jy0);
749 dz20 = _mm_sub_pd(iz2,jz0);
750 dx21 = _mm_sub_pd(ix2,jx1);
751 dy21 = _mm_sub_pd(iy2,jy1);
752 dz21 = _mm_sub_pd(iz2,jz1);
753 dx22 = _mm_sub_pd(ix2,jx2);
754 dy22 = _mm_sub_pd(iy2,jy2);
755 dz22 = _mm_sub_pd(iz2,jz2);
756
757 /* Calculate squared distance and things based on it */
758 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
759 rsq01 = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
760 rsq02 = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
761 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
762 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
763 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
764 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
765 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
766 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
767
768 rinv00 = gmx_mm_invsqrt_pd(rsq00);
769 rinv01 = gmx_mm_invsqrt_pd(rsq01);
770 rinv02 = gmx_mm_invsqrt_pd(rsq02);
771 rinv10 = gmx_mm_invsqrt_pd(rsq10);
772 rinv11 = gmx_mm_invsqrt_pd(rsq11);
773 rinv12 = gmx_mm_invsqrt_pd(rsq12);
774 rinv20 = gmx_mm_invsqrt_pd(rsq20);
775 rinv21 = gmx_mm_invsqrt_pd(rsq21);
776 rinv22 = gmx_mm_invsqrt_pd(rsq22);
777
778 fjx0 = _mm_setzero_pd();
779 fjy0 = _mm_setzero_pd();
780 fjz0 = _mm_setzero_pd();
781 fjx1 = _mm_setzero_pd();
782 fjy1 = _mm_setzero_pd();
783 fjz1 = _mm_setzero_pd();
784 fjx2 = _mm_setzero_pd();
785 fjy2 = _mm_setzero_pd();
786 fjz2 = _mm_setzero_pd();
787
788 /**************************
789 * CALCULATE INTERACTIONS *
790 **************************/
791
792 r00 = _mm_mul_pd(rsq00,rinv00);
793
794 /* Calculate table index by multiplying r with table scale and truncate to integer */
795 rt = _mm_mul_pd(r00,vftabscale);
796 vfitab = _mm_cvttpd_epi32(rt);
797 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
798 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
799
800 /* CUBIC SPLINE TABLE ELECTROSTATICS */
801 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
802 F = _mm_setzero_pd();
803 GMX_MM_TRANSPOSE2_PD(Y,F);
804 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
805 H = _mm_setzero_pd();
806 GMX_MM_TRANSPOSE2_PD(G,H);
807 Heps = _mm_mul_pd(vfeps,H);
808 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
809 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
810 velec = _mm_mul_pd(qq00,VV);
811 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
812 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq00,FF),_mm_mul_pd(vftabscale,rinv00)));
813
814 /* CUBIC SPLINE TABLE DISPERSION */
815 vfitab = _mm_add_epi32(vfitab,ifour);
816 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
817 F = _mm_setzero_pd();
818 GMX_MM_TRANSPOSE2_PD(Y,F);
819 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
820 H = _mm_setzero_pd();
821 GMX_MM_TRANSPOSE2_PD(G,H);
822 Heps = _mm_mul_pd(vfeps,H);
823 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
824 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
825 vvdw6 = _mm_mul_pd(c6_00,VV);
826 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
827 fvdw6 = _mm_mul_pd(c6_00,FF);
828
829 /* CUBIC SPLINE TABLE REPULSION */
830 vfitab = _mm_add_epi32(vfitab,ifour);
831 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
832 F = _mm_setzero_pd();
833 GMX_MM_TRANSPOSE2_PD(Y,F);
834 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
835 H = _mm_setzero_pd();
836 GMX_MM_TRANSPOSE2_PD(G,H);
837 Heps = _mm_mul_pd(vfeps,H);
838 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
839 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
840 vvdw12 = _mm_mul_pd(c12_00,VV);
841 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
842 fvdw12 = _mm_mul_pd(c12_00,FF);
843 vvdw = _mm_add_pd(vvdw12,vvdw6);
844 fvdw = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
845
846 /* Update potential sum for this i atom from the interaction with this j atom. */
847 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
848 velecsum = _mm_add_pd(velecsum,velec);
849 vvdw = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
850 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
851
852 fscal = _mm_add_pd(felec,fvdw);
853
854 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
855
856 /* Calculate temporary vectorial force */
857 tx = _mm_mul_pd(fscal,dx00);
858 ty = _mm_mul_pd(fscal,dy00);
859 tz = _mm_mul_pd(fscal,dz00);
860
861 /* Update vectorial force */
862 fix0 = _mm_add_pd(fix0,tx);
863 fiy0 = _mm_add_pd(fiy0,ty);
864 fiz0 = _mm_add_pd(fiz0,tz);
865
866 fjx0 = _mm_add_pd(fjx0,tx);
867 fjy0 = _mm_add_pd(fjy0,ty);
868 fjz0 = _mm_add_pd(fjz0,tz);
869
870 /**************************
871 * CALCULATE INTERACTIONS *
872 **************************/
873
874 r01 = _mm_mul_pd(rsq01,rinv01);
875
876 /* Calculate table index by multiplying r with table scale and truncate to integer */
877 rt = _mm_mul_pd(r01,vftabscale);
878 vfitab = _mm_cvttpd_epi32(rt);
879 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
880 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
881
882 /* CUBIC SPLINE TABLE ELECTROSTATICS */
883 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
884 F = _mm_setzero_pd();
885 GMX_MM_TRANSPOSE2_PD(Y,F);
886 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
887 H = _mm_setzero_pd();
888 GMX_MM_TRANSPOSE2_PD(G,H);
889 Heps = _mm_mul_pd(vfeps,H);
890 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
891 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
892 velec = _mm_mul_pd(qq01,VV);
893 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
894 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq01,FF),_mm_mul_pd(vftabscale,rinv01)));
895
896 /* Update potential sum for this i atom from the interaction with this j atom. */
897 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
898 velecsum = _mm_add_pd(velecsum,velec);
899
900 fscal = felec;
901
902 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
903
904 /* Calculate temporary vectorial force */
905 tx = _mm_mul_pd(fscal,dx01);
906 ty = _mm_mul_pd(fscal,dy01);
907 tz = _mm_mul_pd(fscal,dz01);
908
909 /* Update vectorial force */
910 fix0 = _mm_add_pd(fix0,tx);
911 fiy0 = _mm_add_pd(fiy0,ty);
912 fiz0 = _mm_add_pd(fiz0,tz);
913
914 fjx1 = _mm_add_pd(fjx1,tx);
915 fjy1 = _mm_add_pd(fjy1,ty);
916 fjz1 = _mm_add_pd(fjz1,tz);
917
918 /**************************
919 * CALCULATE INTERACTIONS *
920 **************************/
921
922 r02 = _mm_mul_pd(rsq02,rinv02);
923
924 /* Calculate table index by multiplying r with table scale and truncate to integer */
925 rt = _mm_mul_pd(r02,vftabscale);
926 vfitab = _mm_cvttpd_epi32(rt);
927 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
928 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
929
930 /* CUBIC SPLINE TABLE ELECTROSTATICS */
931 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
932 F = _mm_setzero_pd();
933 GMX_MM_TRANSPOSE2_PD(Y,F);
934 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
935 H = _mm_setzero_pd();
936 GMX_MM_TRANSPOSE2_PD(G,H);
937 Heps = _mm_mul_pd(vfeps,H);
938 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
939 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
940 velec = _mm_mul_pd(qq02,VV);
941 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
942 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq02,FF),_mm_mul_pd(vftabscale,rinv02)));
943
944 /* Update potential sum for this i atom from the interaction with this j atom. */
945 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
946 velecsum = _mm_add_pd(velecsum,velec);
947
948 fscal = felec;
949
950 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
951
952 /* Calculate temporary vectorial force */
953 tx = _mm_mul_pd(fscal,dx02);
954 ty = _mm_mul_pd(fscal,dy02);
955 tz = _mm_mul_pd(fscal,dz02);
956
957 /* Update vectorial force */
958 fix0 = _mm_add_pd(fix0,tx);
959 fiy0 = _mm_add_pd(fiy0,ty);
960 fiz0 = _mm_add_pd(fiz0,tz);
961
962 fjx2 = _mm_add_pd(fjx2,tx);
963 fjy2 = _mm_add_pd(fjy2,ty);
964 fjz2 = _mm_add_pd(fjz2,tz);
965
966 /**************************
967 * CALCULATE INTERACTIONS *
968 **************************/
969
970 r10 = _mm_mul_pd(rsq10,rinv10);
971
972 /* Calculate table index by multiplying r with table scale and truncate to integer */
973 rt = _mm_mul_pd(r10,vftabscale);
974 vfitab = _mm_cvttpd_epi32(rt);
975 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
976 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
977
978 /* CUBIC SPLINE TABLE ELECTROSTATICS */
979 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
980 F = _mm_setzero_pd();
981 GMX_MM_TRANSPOSE2_PD(Y,F);
982 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
983 H = _mm_setzero_pd();
984 GMX_MM_TRANSPOSE2_PD(G,H);
985 Heps = _mm_mul_pd(vfeps,H);
986 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
987 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
988 velec = _mm_mul_pd(qq10,VV);
989 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
990 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq10,FF),_mm_mul_pd(vftabscale,rinv10)));
991
992 /* Update potential sum for this i atom from the interaction with this j atom. */
993 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
994 velecsum = _mm_add_pd(velecsum,velec);
995
996 fscal = felec;
997
998 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
999
1000 /* Calculate temporary vectorial force */
1001 tx = _mm_mul_pd(fscal,dx10);
1002 ty = _mm_mul_pd(fscal,dy10);
1003 tz = _mm_mul_pd(fscal,dz10);
1004
1005 /* Update vectorial force */
1006 fix1 = _mm_add_pd(fix1,tx);
1007 fiy1 = _mm_add_pd(fiy1,ty);
1008 fiz1 = _mm_add_pd(fiz1,tz);
1009
1010 fjx0 = _mm_add_pd(fjx0,tx);
1011 fjy0 = _mm_add_pd(fjy0,ty);
1012 fjz0 = _mm_add_pd(fjz0,tz);
1013
1014 /**************************
1015 * CALCULATE INTERACTIONS *
1016 **************************/
1017
1018 r11 = _mm_mul_pd(rsq11,rinv11);
1019
1020 /* Calculate table index by multiplying r with table scale and truncate to integer */
1021 rt = _mm_mul_pd(r11,vftabscale);
1022 vfitab = _mm_cvttpd_epi32(rt);
1023 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1024 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1025
1026 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1027 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1028 F = _mm_setzero_pd();
1029 GMX_MM_TRANSPOSE2_PD(Y,F);
1030 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1031 H = _mm_setzero_pd();
1032 GMX_MM_TRANSPOSE2_PD(G,H);
1033 Heps = _mm_mul_pd(vfeps,H);
1034 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1035 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1036 velec = _mm_mul_pd(qq11,VV);
1037 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1038 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
1039
1040 /* Update potential sum for this i atom from the interaction with this j atom. */
1041 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1042 velecsum = _mm_add_pd(velecsum,velec);
1043
1044 fscal = felec;
1045
1046 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1047
1048 /* Calculate temporary vectorial force */
1049 tx = _mm_mul_pd(fscal,dx11);
1050 ty = _mm_mul_pd(fscal,dy11);
1051 tz = _mm_mul_pd(fscal,dz11);
1052
1053 /* Update vectorial force */
1054 fix1 = _mm_add_pd(fix1,tx);
1055 fiy1 = _mm_add_pd(fiy1,ty);
1056 fiz1 = _mm_add_pd(fiz1,tz);
1057
1058 fjx1 = _mm_add_pd(fjx1,tx);
1059 fjy1 = _mm_add_pd(fjy1,ty);
1060 fjz1 = _mm_add_pd(fjz1,tz);
1061
1062 /**************************
1063 * CALCULATE INTERACTIONS *
1064 **************************/
1065
1066 r12 = _mm_mul_pd(rsq12,rinv12);
1067
1068 /* Calculate table index by multiplying r with table scale and truncate to integer */
1069 rt = _mm_mul_pd(r12,vftabscale);
1070 vfitab = _mm_cvttpd_epi32(rt);
1071 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1072 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1073
1074 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1075 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1076 F = _mm_setzero_pd();
1077 GMX_MM_TRANSPOSE2_PD(Y,F);
1078 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1079 H = _mm_setzero_pd();
1080 GMX_MM_TRANSPOSE2_PD(G,H);
1081 Heps = _mm_mul_pd(vfeps,H);
1082 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1083 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1084 velec = _mm_mul_pd(qq12,VV);
1085 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1086 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
1087
1088 /* Update potential sum for this i atom from the interaction with this j atom. */
1089 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1090 velecsum = _mm_add_pd(velecsum,velec);
1091
1092 fscal = felec;
1093
1094 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1095
1096 /* Calculate temporary vectorial force */
1097 tx = _mm_mul_pd(fscal,dx12);
1098 ty = _mm_mul_pd(fscal,dy12);
1099 tz = _mm_mul_pd(fscal,dz12);
1100
1101 /* Update vectorial force */
1102 fix1 = _mm_add_pd(fix1,tx);
1103 fiy1 = _mm_add_pd(fiy1,ty);
1104 fiz1 = _mm_add_pd(fiz1,tz);
1105
1106 fjx2 = _mm_add_pd(fjx2,tx);
1107 fjy2 = _mm_add_pd(fjy2,ty);
1108 fjz2 = _mm_add_pd(fjz2,tz);
1109
1110 /**************************
1111 * CALCULATE INTERACTIONS *
1112 **************************/
1113
1114 r20 = _mm_mul_pd(rsq20,rinv20);
1115
1116 /* Calculate table index by multiplying r with table scale and truncate to integer */
1117 rt = _mm_mul_pd(r20,vftabscale);
1118 vfitab = _mm_cvttpd_epi32(rt);
1119 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1120 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1121
1122 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1123 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1124 F = _mm_setzero_pd();
1125 GMX_MM_TRANSPOSE2_PD(Y,F);
1126 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1127 H = _mm_setzero_pd();
1128 GMX_MM_TRANSPOSE2_PD(G,H);
1129 Heps = _mm_mul_pd(vfeps,H);
1130 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1131 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1132 velec = _mm_mul_pd(qq20,VV);
1133 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1134 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq20,FF),_mm_mul_pd(vftabscale,rinv20)));
1135
1136 /* Update potential sum for this i atom from the interaction with this j atom. */
1137 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1138 velecsum = _mm_add_pd(velecsum,velec);
1139
1140 fscal = felec;
1141
1142 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1143
1144 /* Calculate temporary vectorial force */
1145 tx = _mm_mul_pd(fscal,dx20);
1146 ty = _mm_mul_pd(fscal,dy20);
1147 tz = _mm_mul_pd(fscal,dz20);
1148
1149 /* Update vectorial force */
1150 fix2 = _mm_add_pd(fix2,tx);
1151 fiy2 = _mm_add_pd(fiy2,ty);
1152 fiz2 = _mm_add_pd(fiz2,tz);
1153
1154 fjx0 = _mm_add_pd(fjx0,tx);
1155 fjy0 = _mm_add_pd(fjy0,ty);
1156 fjz0 = _mm_add_pd(fjz0,tz);
1157
1158 /**************************
1159 * CALCULATE INTERACTIONS *
1160 **************************/
1161
1162 r21 = _mm_mul_pd(rsq21,rinv21);
1163
1164 /* Calculate table index by multiplying r with table scale and truncate to integer */
1165 rt = _mm_mul_pd(r21,vftabscale);
1166 vfitab = _mm_cvttpd_epi32(rt);
1167 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1168 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1169
1170 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1171 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1172 F = _mm_setzero_pd();
1173 GMX_MM_TRANSPOSE2_PD(Y,F);
1174 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1175 H = _mm_setzero_pd();
1176 GMX_MM_TRANSPOSE2_PD(G,H);
1177 Heps = _mm_mul_pd(vfeps,H);
1178 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1179 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1180 velec = _mm_mul_pd(qq21,VV);
1181 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1182 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
1183
1184 /* Update potential sum for this i atom from the interaction with this j atom. */
1185 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1186 velecsum = _mm_add_pd(velecsum,velec);
1187
1188 fscal = felec;
1189
1190 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1191
1192 /* Calculate temporary vectorial force */
1193 tx = _mm_mul_pd(fscal,dx21);
1194 ty = _mm_mul_pd(fscal,dy21);
1195 tz = _mm_mul_pd(fscal,dz21);
1196
1197 /* Update vectorial force */
1198 fix2 = _mm_add_pd(fix2,tx);
1199 fiy2 = _mm_add_pd(fiy2,ty);
1200 fiz2 = _mm_add_pd(fiz2,tz);
1201
1202 fjx1 = _mm_add_pd(fjx1,tx);
1203 fjy1 = _mm_add_pd(fjy1,ty);
1204 fjz1 = _mm_add_pd(fjz1,tz);
1205
1206 /**************************
1207 * CALCULATE INTERACTIONS *
1208 **************************/
1209
1210 r22 = _mm_mul_pd(rsq22,rinv22);
1211
1212 /* Calculate table index by multiplying r with table scale and truncate to integer */
1213 rt = _mm_mul_pd(r22,vftabscale);
1214 vfitab = _mm_cvttpd_epi32(rt);
1215 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1216 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1217
1218 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1219 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1220 F = _mm_setzero_pd();
1221 GMX_MM_TRANSPOSE2_PD(Y,F);
1222 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1223 H = _mm_setzero_pd();
1224 GMX_MM_TRANSPOSE2_PD(G,H);
1225 Heps = _mm_mul_pd(vfeps,H);
1226 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1227 VV = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1228 velec = _mm_mul_pd(qq22,VV);
1229 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1230 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
1231
1232 /* Update potential sum for this i atom from the interaction with this j atom. */
1233 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1234 velecsum = _mm_add_pd(velecsum,velec);
1235
1236 fscal = felec;
1237
1238 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1239
1240 /* Calculate temporary vectorial force */
1241 tx = _mm_mul_pd(fscal,dx22);
1242 ty = _mm_mul_pd(fscal,dy22);
1243 tz = _mm_mul_pd(fscal,dz22);
1244
1245 /* Update vectorial force */
1246 fix2 = _mm_add_pd(fix2,tx);
1247 fiy2 = _mm_add_pd(fiy2,ty);
1248 fiz2 = _mm_add_pd(fiz2,tz);
1249
1250 fjx2 = _mm_add_pd(fjx2,tx);
1251 fjy2 = _mm_add_pd(fjy2,ty);
1252 fjz2 = _mm_add_pd(fjz2,tz);
1253
1254 gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1255
1256 /* Inner loop uses 417 flops */
1257 }
1258
1259 /* End of innermost loop */
1260
1261 gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1262 f+i_coord_offset,fshift+i_shift_offset);
1263
1264 ggid = gid[iidx];
1265 /* Update potential energies */
1266 gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1267 gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1268
1269 /* Increment number of inner iterations */
1270 inneriter += j_index_end - j_index_start;
1271
1272 /* Outer loop uses 20 flops */
1273 }
1274
1275 /* Increment number of outer iterations */
1276 outeriter += nri;
1277
1278 /* Update outer/inner flops */
1279
1280 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_VF,outeriter*20 + inneriter*417);
1281 }
1282 /*
1283 * Gromacs nonbonded kernel: nb_kernel_ElecCSTab_VdwCSTab_GeomW3W3_F_sse2_double
1284 * Electrostatics interaction: CubicSplineTable
1285 * VdW interaction: CubicSplineTable
1286 * Geometry: Water3-Water3
1287 * Calculate force/pot: Force
1288 */
1289 void
1290 nb_kernel_ElecCSTab_VdwCSTab_GeomW3W3_F_sse2_double
1291 (t_nblist * gmx_restrict nlist,
1292 rvec * gmx_restrict xx,
1293 rvec * gmx_restrict ff,
1294 t_forcerec * gmx_restrict fr,
1295 t_mdatoms * gmx_restrict mdatoms,
1296 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1297 t_nrnb * gmx_restrict nrnb)
1298 {
1299 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1300 * just 0 for non-waters.
1301 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
1302 * jnr indices corresponding to data put in the four positions in the SIMD register.
1303 */
1304 int i_shift_offset,i_coord_offset,outeriter,inneriter;
1305 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1306 int jnrA,jnrB;
1307 int j_coord_offsetA,j_coord_offsetB;
1308 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
1309 real rcutoff_scalar;
1310 real *shiftvec,*fshift,*x,*f;
1311 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1312 int vdwioffset0;
1313 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1314 int vdwioffset1;
1315 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1316 int vdwioffset2;
1317 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1318 int vdwjidx0A,vdwjidx0B;
1319 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1320 int vdwjidx1A,vdwjidx1B;
1321 __m128d jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1322 int vdwjidx2A,vdwjidx2B;
1323 __m128d jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1324 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1325 __m128d dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1326 __m128d dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1327 __m128d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1328 __m128d dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1329 __m128d dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1330 __m128d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1331 __m128d dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1332 __m128d dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1333 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
1334 real *charge;
1335 int nvdwtype;
1336 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1337 int *vdwtype;
1338 real *vdwparam;
1339 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
1340 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
1341 __m128i vfitab;
1342 __m128i ifour = _mm_set1_epi32(4);
1343 __m128d rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
1344 real *vftab;
1345 __m128d dummy_mask,cutoff_mask;
1346 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
1347 __m128d one = _mm_set1_pd(1.0);
1348 __m128d two = _mm_set1_pd(2.0);
1349 x = xx[0];
1350 f = ff[0];
1351
1352 nri = nlist->nri;
1353 iinr = nlist->iinr;
1354 jindex = nlist->jindex;
1355 jjnr = nlist->jjnr;
1356 shiftidx = nlist->shift;
1357 gid = nlist->gid;
1358 shiftvec = fr->shift_vec[0];
1359 fshift = fr->fshift[0];
1360 facel = _mm_set1_pd(fr->epsfac);
1361 charge = mdatoms->chargeA;
1362 nvdwtype = fr->ntype;
1363 vdwparam = fr->nbfp;
1364 vdwtype = mdatoms->typeA;
1365
1366 vftab = kernel_data->table_elec_vdw->data;
1367 vftabscale = _mm_set1_pd(kernel_data->table_elec_vdw->scale);
1368
1369 /* Setup water-specific parameters */
1370 inr = nlist->iinr[0];
1371 iq0 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
1372 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
1373 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
1374 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
1375
1376 jq0 = _mm_set1_pd(charge[inr+0]);
1377 jq1 = _mm_set1_pd(charge[inr+1]);
1378 jq2 = _mm_set1_pd(charge[inr+2]);
1379 vdwjidx0A = 2*vdwtype[inr+0];
1380 qq00 = _mm_mul_pd(iq0,jq0);
1381 c6_00 = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
1382 c12_00 = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
1383 qq01 = _mm_mul_pd(iq0,jq1);
1384 qq02 = _mm_mul_pd(iq0,jq2);
1385 qq10 = _mm_mul_pd(iq1,jq0);
1386 qq11 = _mm_mul_pd(iq1,jq1);
1387 qq12 = _mm_mul_pd(iq1,jq2);
1388 qq20 = _mm_mul_pd(iq2,jq0);
1389 qq21 = _mm_mul_pd(iq2,jq1);
1390 qq22 = _mm_mul_pd(iq2,jq2);
1391
1392 /* Avoid stupid compiler warnings */
1393 jnrA = jnrB = 0;
1394 j_coord_offsetA = 0;
1395 j_coord_offsetB = 0;
1396
1397 outeriter = 0;
1398 inneriter = 0;
1399
1400 /* Start outer loop over neighborlists */
1401 for(iidx=0; iidx<nri; iidx++)
1402 {
1403 /* Load shift vector for this list */
1404 i_shift_offset = DIM*shiftidx[iidx];
1405
1406 /* Load limits for loop over neighbors */
1407 j_index_start = jindex[iidx];
1408 j_index_end = jindex[iidx+1];
1409
1410 /* Get outer coordinate index */
1411 inr = iinr[iidx];
1412 i_coord_offset = DIM*inr;
1413
1414 /* Load i particle coords and add shift vector */
1415 gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1416 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1417
1418 fix0 = _mm_setzero_pd();
1419 fiy0 = _mm_setzero_pd();
1420 fiz0 = _mm_setzero_pd();
1421 fix1 = _mm_setzero_pd();
1422 fiy1 = _mm_setzero_pd();
1423 fiz1 = _mm_setzero_pd();
1424 fix2 = _mm_setzero_pd();
1425 fiy2 = _mm_setzero_pd();
1426 fiz2 = _mm_setzero_pd();
1427
1428 /* Start inner kernel loop */
1429 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
1430 {
1431
1432 /* Get j neighbor index, and coordinate index */
1433 jnrA = jjnr[jidx];
1434 jnrB = jjnr[jidx+1];
1435 j_coord_offsetA = DIM*jnrA;
1436 j_coord_offsetB = DIM*jnrB;
1437
1438 /* load j atom coordinates */
1439 gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1440 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1441
1442 /* Calculate displacement vector */
1443 dx00 = _mm_sub_pd(ix0,jx0);
1444 dy00 = _mm_sub_pd(iy0,jy0);
1445 dz00 = _mm_sub_pd(iz0,jz0);
1446 dx01 = _mm_sub_pd(ix0,jx1);
1447 dy01 = _mm_sub_pd(iy0,jy1);
1448 dz01 = _mm_sub_pd(iz0,jz1);
1449 dx02 = _mm_sub_pd(ix0,jx2);
1450 dy02 = _mm_sub_pd(iy0,jy2);
1451 dz02 = _mm_sub_pd(iz0,jz2);
1452 dx10 = _mm_sub_pd(ix1,jx0);
1453 dy10 = _mm_sub_pd(iy1,jy0);
1454 dz10 = _mm_sub_pd(iz1,jz0);
1455 dx11 = _mm_sub_pd(ix1,jx1);
1456 dy11 = _mm_sub_pd(iy1,jy1);
1457 dz11 = _mm_sub_pd(iz1,jz1);
1458 dx12 = _mm_sub_pd(ix1,jx2);
1459 dy12 = _mm_sub_pd(iy1,jy2);
1460 dz12 = _mm_sub_pd(iz1,jz2);
1461 dx20 = _mm_sub_pd(ix2,jx0);
1462 dy20 = _mm_sub_pd(iy2,jy0);
1463 dz20 = _mm_sub_pd(iz2,jz0);
1464 dx21 = _mm_sub_pd(ix2,jx1);
1465 dy21 = _mm_sub_pd(iy2,jy1);
1466 dz21 = _mm_sub_pd(iz2,jz1);
1467 dx22 = _mm_sub_pd(ix2,jx2);
1468 dy22 = _mm_sub_pd(iy2,jy2);
1469 dz22 = _mm_sub_pd(iz2,jz2);
1470
1471 /* Calculate squared distance and things based on it */
1472 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1473 rsq01 = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
1474 rsq02 = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
1475 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
1476 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1477 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1478 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
1479 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1480 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1481
1482 rinv00 = gmx_mm_invsqrt_pd(rsq00);
1483 rinv01 = gmx_mm_invsqrt_pd(rsq01);
1484 rinv02 = gmx_mm_invsqrt_pd(rsq02);
1485 rinv10 = gmx_mm_invsqrt_pd(rsq10);
1486 rinv11 = gmx_mm_invsqrt_pd(rsq11);
1487 rinv12 = gmx_mm_invsqrt_pd(rsq12);
1488 rinv20 = gmx_mm_invsqrt_pd(rsq20);
1489 rinv21 = gmx_mm_invsqrt_pd(rsq21);
1490 rinv22 = gmx_mm_invsqrt_pd(rsq22);
1491
1492 fjx0 = _mm_setzero_pd();
1493 fjy0 = _mm_setzero_pd();
1494 fjz0 = _mm_setzero_pd();
1495 fjx1 = _mm_setzero_pd();
1496 fjy1 = _mm_setzero_pd();
1497 fjz1 = _mm_setzero_pd();
1498 fjx2 = _mm_setzero_pd();
1499 fjy2 = _mm_setzero_pd();
1500 fjz2 = _mm_setzero_pd();
1501
1502 /**************************
1503 * CALCULATE INTERACTIONS *
1504 **************************/
1505
1506 r00 = _mm_mul_pd(rsq00,rinv00);
1507
1508 /* Calculate table index by multiplying r with table scale and truncate to integer */
1509 rt = _mm_mul_pd(r00,vftabscale);
1510 vfitab = _mm_cvttpd_epi32(rt);
1511 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1512 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1513
1514 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1515 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1516 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1517 GMX_MM_TRANSPOSE2_PD(Y,F);
1518 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1519 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1520 GMX_MM_TRANSPOSE2_PD(G,H);
1521 Heps = _mm_mul_pd(vfeps,H);
1522 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1523 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1524 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq00,FF),_mm_mul_pd(vftabscale,rinv00)));
1525
1526 /* CUBIC SPLINE TABLE DISPERSION */
1527 vfitab = _mm_add_epi32(vfitab,ifour);
1528 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1529 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1530 GMX_MM_TRANSPOSE2_PD(Y,F);
1531 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1532 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1533 GMX_MM_TRANSPOSE2_PD(G,H);
1534 Heps = _mm_mul_pd(vfeps,H);
1535 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1536 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1537 fvdw6 = _mm_mul_pd(c6_00,FF);
1538
1539 /* CUBIC SPLINE TABLE REPULSION */
1540 vfitab = _mm_add_epi32(vfitab,ifour);
1541 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1542 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1543 GMX_MM_TRANSPOSE2_PD(Y,F);
1544 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1545 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1546 GMX_MM_TRANSPOSE2_PD(G,H);
1547 Heps = _mm_mul_pd(vfeps,H);
1548 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1549 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1550 fvdw12 = _mm_mul_pd(c12_00,FF);
1551 fvdw = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
1552
1553 fscal = _mm_add_pd(felec,fvdw);
1554
1555 /* Calculate temporary vectorial force */
1556 tx = _mm_mul_pd(fscal,dx00);
1557 ty = _mm_mul_pd(fscal,dy00);
1558 tz = _mm_mul_pd(fscal,dz00);
1559
1560 /* Update vectorial force */
1561 fix0 = _mm_add_pd(fix0,tx);
1562 fiy0 = _mm_add_pd(fiy0,ty);
1563 fiz0 = _mm_add_pd(fiz0,tz);
1564
1565 fjx0 = _mm_add_pd(fjx0,tx);
1566 fjy0 = _mm_add_pd(fjy0,ty);
1567 fjz0 = _mm_add_pd(fjz0,tz);
1568
1569 /**************************
1570 * CALCULATE INTERACTIONS *
1571 **************************/
1572
1573 r01 = _mm_mul_pd(rsq01,rinv01);
1574
1575 /* Calculate table index by multiplying r with table scale and truncate to integer */
1576 rt = _mm_mul_pd(r01,vftabscale);
1577 vfitab = _mm_cvttpd_epi32(rt);
1578 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1579 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1580
1581 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1582 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1583 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1584 GMX_MM_TRANSPOSE2_PD(Y,F);
1585 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1586 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1587 GMX_MM_TRANSPOSE2_PD(G,H);
1588 Heps = _mm_mul_pd(vfeps,H);
1589 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1590 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1591 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq01,FF),_mm_mul_pd(vftabscale,rinv01)));
1592
1593 fscal = felec;
1594
1595 /* Calculate temporary vectorial force */
1596 tx = _mm_mul_pd(fscal,dx01);
1597 ty = _mm_mul_pd(fscal,dy01);
1598 tz = _mm_mul_pd(fscal,dz01);
1599
1600 /* Update vectorial force */
1601 fix0 = _mm_add_pd(fix0,tx);
1602 fiy0 = _mm_add_pd(fiy0,ty);
1603 fiz0 = _mm_add_pd(fiz0,tz);
1604
1605 fjx1 = _mm_add_pd(fjx1,tx);
1606 fjy1 = _mm_add_pd(fjy1,ty);
1607 fjz1 = _mm_add_pd(fjz1,tz);
1608
1609 /**************************
1610 * CALCULATE INTERACTIONS *
1611 **************************/
1612
1613 r02 = _mm_mul_pd(rsq02,rinv02);
1614
1615 /* Calculate table index by multiplying r with table scale and truncate to integer */
1616 rt = _mm_mul_pd(r02,vftabscale);
1617 vfitab = _mm_cvttpd_epi32(rt);
1618 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1619 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1620
1621 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1622 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1623 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1624 GMX_MM_TRANSPOSE2_PD(Y,F);
1625 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1626 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1627 GMX_MM_TRANSPOSE2_PD(G,H);
1628 Heps = _mm_mul_pd(vfeps,H);
1629 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1630 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1631 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq02,FF),_mm_mul_pd(vftabscale,rinv02)));
1632
1633 fscal = felec;
1634
1635 /* Calculate temporary vectorial force */
1636 tx = _mm_mul_pd(fscal,dx02);
1637 ty = _mm_mul_pd(fscal,dy02);
1638 tz = _mm_mul_pd(fscal,dz02);
1639
1640 /* Update vectorial force */
1641 fix0 = _mm_add_pd(fix0,tx);
1642 fiy0 = _mm_add_pd(fiy0,ty);
1643 fiz0 = _mm_add_pd(fiz0,tz);
1644
1645 fjx2 = _mm_add_pd(fjx2,tx);
1646 fjy2 = _mm_add_pd(fjy2,ty);
1647 fjz2 = _mm_add_pd(fjz2,tz);
1648
1649 /**************************
1650 * CALCULATE INTERACTIONS *
1651 **************************/
1652
1653 r10 = _mm_mul_pd(rsq10,rinv10);
1654
1655 /* Calculate table index by multiplying r with table scale and truncate to integer */
1656 rt = _mm_mul_pd(r10,vftabscale);
1657 vfitab = _mm_cvttpd_epi32(rt);
1658 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1659 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1660
1661 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1662 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1663 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1664 GMX_MM_TRANSPOSE2_PD(Y,F);
1665 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1666 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1667 GMX_MM_TRANSPOSE2_PD(G,H);
1668 Heps = _mm_mul_pd(vfeps,H);
1669 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1670 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1671 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq10,FF),_mm_mul_pd(vftabscale,rinv10)));
1672
1673 fscal = felec;
1674
1675 /* Calculate temporary vectorial force */
1676 tx = _mm_mul_pd(fscal,dx10);
1677 ty = _mm_mul_pd(fscal,dy10);
1678 tz = _mm_mul_pd(fscal,dz10);
1679
1680 /* Update vectorial force */
1681 fix1 = _mm_add_pd(fix1,tx);
1682 fiy1 = _mm_add_pd(fiy1,ty);
1683 fiz1 = _mm_add_pd(fiz1,tz);
1684
1685 fjx0 = _mm_add_pd(fjx0,tx);
1686 fjy0 = _mm_add_pd(fjy0,ty);
1687 fjz0 = _mm_add_pd(fjz0,tz);
1688
1689 /**************************
1690 * CALCULATE INTERACTIONS *
1691 **************************/
1692
1693 r11 = _mm_mul_pd(rsq11,rinv11);
1694
1695 /* Calculate table index by multiplying r with table scale and truncate to integer */
1696 rt = _mm_mul_pd(r11,vftabscale);
1697 vfitab = _mm_cvttpd_epi32(rt);
1698 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1699 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1700
1701 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1702 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1703 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1704 GMX_MM_TRANSPOSE2_PD(Y,F);
1705 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1706 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1707 GMX_MM_TRANSPOSE2_PD(G,H);
1708 Heps = _mm_mul_pd(vfeps,H);
1709 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1710 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1711 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
1712
1713 fscal = felec;
1714
1715 /* Calculate temporary vectorial force */
1716 tx = _mm_mul_pd(fscal,dx11);
1717 ty = _mm_mul_pd(fscal,dy11);
1718 tz = _mm_mul_pd(fscal,dz11);
1719
1720 /* Update vectorial force */
1721 fix1 = _mm_add_pd(fix1,tx);
1722 fiy1 = _mm_add_pd(fiy1,ty);
1723 fiz1 = _mm_add_pd(fiz1,tz);
1724
1725 fjx1 = _mm_add_pd(fjx1,tx);
1726 fjy1 = _mm_add_pd(fjy1,ty);
1727 fjz1 = _mm_add_pd(fjz1,tz);
1728
1729 /**************************
1730 * CALCULATE INTERACTIONS *
1731 **************************/
1732
1733 r12 = _mm_mul_pd(rsq12,rinv12);
1734
1735 /* Calculate table index by multiplying r with table scale and truncate to integer */
1736 rt = _mm_mul_pd(r12,vftabscale);
1737 vfitab = _mm_cvttpd_epi32(rt);
1738 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1739 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1740
1741 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1742 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1743 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1744 GMX_MM_TRANSPOSE2_PD(Y,F);
1745 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1746 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1747 GMX_MM_TRANSPOSE2_PD(G,H);
1748 Heps = _mm_mul_pd(vfeps,H);
1749 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1750 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1751 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
1752
1753 fscal = felec;
1754
1755 /* Calculate temporary vectorial force */
1756 tx = _mm_mul_pd(fscal,dx12);
1757 ty = _mm_mul_pd(fscal,dy12);
1758 tz = _mm_mul_pd(fscal,dz12);
1759
1760 /* Update vectorial force */
1761 fix1 = _mm_add_pd(fix1,tx);
1762 fiy1 = _mm_add_pd(fiy1,ty);
1763 fiz1 = _mm_add_pd(fiz1,tz);
1764
1765 fjx2 = _mm_add_pd(fjx2,tx);
1766 fjy2 = _mm_add_pd(fjy2,ty);
1767 fjz2 = _mm_add_pd(fjz2,tz);
1768
1769 /**************************
1770 * CALCULATE INTERACTIONS *
1771 **************************/
1772
1773 r20 = _mm_mul_pd(rsq20,rinv20);
1774
1775 /* Calculate table index by multiplying r with table scale and truncate to integer */
1776 rt = _mm_mul_pd(r20,vftabscale);
1777 vfitab = _mm_cvttpd_epi32(rt);
1778 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1779 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1780
1781 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1782 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1783 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1784 GMX_MM_TRANSPOSE2_PD(Y,F);
1785 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1786 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1787 GMX_MM_TRANSPOSE2_PD(G,H);
1788 Heps = _mm_mul_pd(vfeps,H);
1789 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1790 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1791 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq20,FF),_mm_mul_pd(vftabscale,rinv20)));
1792
1793 fscal = felec;
1794
1795 /* Calculate temporary vectorial force */
1796 tx = _mm_mul_pd(fscal,dx20);
1797 ty = _mm_mul_pd(fscal,dy20);
1798 tz = _mm_mul_pd(fscal,dz20);
1799
1800 /* Update vectorial force */
1801 fix2 = _mm_add_pd(fix2,tx);
1802 fiy2 = _mm_add_pd(fiy2,ty);
1803 fiz2 = _mm_add_pd(fiz2,tz);
1804
1805 fjx0 = _mm_add_pd(fjx0,tx);
1806 fjy0 = _mm_add_pd(fjy0,ty);
1807 fjz0 = _mm_add_pd(fjz0,tz);
1808
1809 /**************************
1810 * CALCULATE INTERACTIONS *
1811 **************************/
1812
1813 r21 = _mm_mul_pd(rsq21,rinv21);
1814
1815 /* Calculate table index by multiplying r with table scale and truncate to integer */
1816 rt = _mm_mul_pd(r21,vftabscale);
1817 vfitab = _mm_cvttpd_epi32(rt);
1818 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1819 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1820
1821 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1822 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1823 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1824 GMX_MM_TRANSPOSE2_PD(Y,F);
1825 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1826 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1827 GMX_MM_TRANSPOSE2_PD(G,H);
1828 Heps = _mm_mul_pd(vfeps,H);
1829 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1830 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1831 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
1832
1833 fscal = felec;
1834
1835 /* Calculate temporary vectorial force */
1836 tx = _mm_mul_pd(fscal,dx21);
1837 ty = _mm_mul_pd(fscal,dy21);
1838 tz = _mm_mul_pd(fscal,dz21);
1839
1840 /* Update vectorial force */
1841 fix2 = _mm_add_pd(fix2,tx);
1842 fiy2 = _mm_add_pd(fiy2,ty);
1843 fiz2 = _mm_add_pd(fiz2,tz);
1844
1845 fjx1 = _mm_add_pd(fjx1,tx);
1846 fjy1 = _mm_add_pd(fjy1,ty);
1847 fjz1 = _mm_add_pd(fjz1,tz);
1848
1849 /**************************
1850 * CALCULATE INTERACTIONS *
1851 **************************/
1852
1853 r22 = _mm_mul_pd(rsq22,rinv22);
1854
1855 /* Calculate table index by multiplying r with table scale and truncate to integer */
1856 rt = _mm_mul_pd(r22,vftabscale);
1857 vfitab = _mm_cvttpd_epi32(rt);
1858 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1859 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1860
1861 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1862 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1863 F = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1864 GMX_MM_TRANSPOSE2_PD(Y,F);
1865 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1866 H = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1867 GMX_MM_TRANSPOSE2_PD(G,H);
1868 Heps = _mm_mul_pd(vfeps,H);
1869 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1870 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1871 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
1872
1873 fscal = felec;
1874
1875 /* Calculate temporary vectorial force */
1876 tx = _mm_mul_pd(fscal,dx22);
1877 ty = _mm_mul_pd(fscal,dy22);
1878 tz = _mm_mul_pd(fscal,dz22);
1879
1880 /* Update vectorial force */
1881 fix2 = _mm_add_pd(fix2,tx);
1882 fiy2 = _mm_add_pd(fiy2,ty);
1883 fiz2 = _mm_add_pd(fiz2,tz);
1884
1885 fjx2 = _mm_add_pd(fjx2,tx);
1886 fjy2 = _mm_add_pd(fjy2,ty);
1887 fjz2 = _mm_add_pd(fjz2,tz);
1888
1889 gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1890
1891 /* Inner loop uses 373 flops */
1892 }
1893
1894 if(jidx<j_index_end)
1895 {
1896
1897 jnrA = jjnr[jidx];
1898 j_coord_offsetA = DIM*jnrA;
1899
1900 /* load j atom coordinates */
1901 gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
1902 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1903
1904 /* Calculate displacement vector */
1905 dx00 = _mm_sub_pd(ix0,jx0);
1906 dy00 = _mm_sub_pd(iy0,jy0);
1907 dz00 = _mm_sub_pd(iz0,jz0);
1908 dx01 = _mm_sub_pd(ix0,jx1);
1909 dy01 = _mm_sub_pd(iy0,jy1);
1910 dz01 = _mm_sub_pd(iz0,jz1);
1911 dx02 = _mm_sub_pd(ix0,jx2);
1912 dy02 = _mm_sub_pd(iy0,jy2);
1913 dz02 = _mm_sub_pd(iz0,jz2);
1914 dx10 = _mm_sub_pd(ix1,jx0);
1915 dy10 = _mm_sub_pd(iy1,jy0);
1916 dz10 = _mm_sub_pd(iz1,jz0);
1917 dx11 = _mm_sub_pd(ix1,jx1);
1918 dy11 = _mm_sub_pd(iy1,jy1);
1919 dz11 = _mm_sub_pd(iz1,jz1);
1920 dx12 = _mm_sub_pd(ix1,jx2);
1921 dy12 = _mm_sub_pd(iy1,jy2);
1922 dz12 = _mm_sub_pd(iz1,jz2);
1923 dx20 = _mm_sub_pd(ix2,jx0);
1924 dy20 = _mm_sub_pd(iy2,jy0);
1925 dz20 = _mm_sub_pd(iz2,jz0);
1926 dx21 = _mm_sub_pd(ix2,jx1);
1927 dy21 = _mm_sub_pd(iy2,jy1);
1928 dz21 = _mm_sub_pd(iz2,jz1);
1929 dx22 = _mm_sub_pd(ix2,jx2);
1930 dy22 = _mm_sub_pd(iy2,jy2);
1931 dz22 = _mm_sub_pd(iz2,jz2);
1932
1933 /* Calculate squared distance and things based on it */
1934 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1935 rsq01 = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
1936 rsq02 = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
1937 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
1938 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1939 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1940 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
1941 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1942 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1943
1944 rinv00 = gmx_mm_invsqrt_pd(rsq00);
1945 rinv01 = gmx_mm_invsqrt_pd(rsq01);
1946 rinv02 = gmx_mm_invsqrt_pd(rsq02);
1947 rinv10 = gmx_mm_invsqrt_pd(rsq10);
1948 rinv11 = gmx_mm_invsqrt_pd(rsq11);
1949 rinv12 = gmx_mm_invsqrt_pd(rsq12);
1950 rinv20 = gmx_mm_invsqrt_pd(rsq20);
1951 rinv21 = gmx_mm_invsqrt_pd(rsq21);
1952 rinv22 = gmx_mm_invsqrt_pd(rsq22);
1953
1954 fjx0 = _mm_setzero_pd();
1955 fjy0 = _mm_setzero_pd();
1956 fjz0 = _mm_setzero_pd();
1957 fjx1 = _mm_setzero_pd();
1958 fjy1 = _mm_setzero_pd();
1959 fjz1 = _mm_setzero_pd();
1960 fjx2 = _mm_setzero_pd();
1961 fjy2 = _mm_setzero_pd();
1962 fjz2 = _mm_setzero_pd();
1963
1964 /**************************
1965 * CALCULATE INTERACTIONS *
1966 **************************/
1967
1968 r00 = _mm_mul_pd(rsq00,rinv00);
1969
1970 /* Calculate table index by multiplying r with table scale and truncate to integer */
1971 rt = _mm_mul_pd(r00,vftabscale);
1972 vfitab = _mm_cvttpd_epi32(rt);
1973 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1974 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1975
1976 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1977 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1978 F = _mm_setzero_pd();
1979 GMX_MM_TRANSPOSE2_PD(Y,F);
1980 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1981 H = _mm_setzero_pd();
1982 GMX_MM_TRANSPOSE2_PD(G,H);
1983 Heps = _mm_mul_pd(vfeps,H);
1984 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1985 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1986 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq00,FF),_mm_mul_pd(vftabscale,rinv00)));
1987
1988 /* CUBIC SPLINE TABLE DISPERSION */
1989 vfitab = _mm_add_epi32(vfitab,ifour);
1990 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1991 F = _mm_setzero_pd();
1992 GMX_MM_TRANSPOSE2_PD(Y,F);
1993 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1994 H = _mm_setzero_pd();
1995 GMX_MM_TRANSPOSE2_PD(G,H);
1996 Heps = _mm_mul_pd(vfeps,H);
1997 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1998 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1999 fvdw6 = _mm_mul_pd(c6_00,FF);
2000
2001 /* CUBIC SPLINE TABLE REPULSION */
2002 vfitab = _mm_add_epi32(vfitab,ifour);
2003 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2004 F = _mm_setzero_pd();
2005 GMX_MM_TRANSPOSE2_PD(Y,F);
2006 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2007 H = _mm_setzero_pd();
2008 GMX_MM_TRANSPOSE2_PD(G,H);
2009 Heps = _mm_mul_pd(vfeps,H);
2010 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2011 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2012 fvdw12 = _mm_mul_pd(c12_00,FF);
2013 fvdw = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
2014
2015 fscal = _mm_add_pd(felec,fvdw);
2016
2017 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2018
2019 /* Calculate temporary vectorial force */
2020 tx = _mm_mul_pd(fscal,dx00);
2021 ty = _mm_mul_pd(fscal,dy00);
2022 tz = _mm_mul_pd(fscal,dz00);
2023
2024 /* Update vectorial force */
2025 fix0 = _mm_add_pd(fix0,tx);
2026 fiy0 = _mm_add_pd(fiy0,ty);
2027 fiz0 = _mm_add_pd(fiz0,tz);
2028
2029 fjx0 = _mm_add_pd(fjx0,tx);
2030 fjy0 = _mm_add_pd(fjy0,ty);
2031 fjz0 = _mm_add_pd(fjz0,tz);
2032
2033 /**************************
2034 * CALCULATE INTERACTIONS *
2035 **************************/
2036
2037 r01 = _mm_mul_pd(rsq01,rinv01);
2038
2039 /* Calculate table index by multiplying r with table scale and truncate to integer */
2040 rt = _mm_mul_pd(r01,vftabscale);
2041 vfitab = _mm_cvttpd_epi32(rt);
2042 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2043 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2044
2045 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2046 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2047 F = _mm_setzero_pd();
2048 GMX_MM_TRANSPOSE2_PD(Y,F);
2049 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2050 H = _mm_setzero_pd();
2051 GMX_MM_TRANSPOSE2_PD(G,H);
2052 Heps = _mm_mul_pd(vfeps,H);
2053 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2054 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2055 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq01,FF),_mm_mul_pd(vftabscale,rinv01)));
2056
2057 fscal = felec;
2058
2059 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2060
2061 /* Calculate temporary vectorial force */
2062 tx = _mm_mul_pd(fscal,dx01);
2063 ty = _mm_mul_pd(fscal,dy01);
2064 tz = _mm_mul_pd(fscal,dz01);
2065
2066 /* Update vectorial force */
2067 fix0 = _mm_add_pd(fix0,tx);
2068 fiy0 = _mm_add_pd(fiy0,ty);
2069 fiz0 = _mm_add_pd(fiz0,tz);
2070
2071 fjx1 = _mm_add_pd(fjx1,tx);
2072 fjy1 = _mm_add_pd(fjy1,ty);
2073 fjz1 = _mm_add_pd(fjz1,tz);
2074
2075 /**************************
2076 * CALCULATE INTERACTIONS *
2077 **************************/
2078
2079 r02 = _mm_mul_pd(rsq02,rinv02);
2080
2081 /* Calculate table index by multiplying r with table scale and truncate to integer */
2082 rt = _mm_mul_pd(r02,vftabscale);
2083 vfitab = _mm_cvttpd_epi32(rt);
2084 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2085 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2086
2087 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2088 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2089 F = _mm_setzero_pd();
2090 GMX_MM_TRANSPOSE2_PD(Y,F);
2091 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2092 H = _mm_setzero_pd();
2093 GMX_MM_TRANSPOSE2_PD(G,H);
2094 Heps = _mm_mul_pd(vfeps,H);
2095 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2096 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2097 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq02,FF),_mm_mul_pd(vftabscale,rinv02)));
2098
2099 fscal = felec;
2100
2101 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2102
2103 /* Calculate temporary vectorial force */
2104 tx = _mm_mul_pd(fscal,dx02);
2105 ty = _mm_mul_pd(fscal,dy02);
2106 tz = _mm_mul_pd(fscal,dz02);
2107
2108 /* Update vectorial force */
2109 fix0 = _mm_add_pd(fix0,tx);
2110 fiy0 = _mm_add_pd(fiy0,ty);
2111 fiz0 = _mm_add_pd(fiz0,tz);
2112
2113 fjx2 = _mm_add_pd(fjx2,tx);
2114 fjy2 = _mm_add_pd(fjy2,ty);
2115 fjz2 = _mm_add_pd(fjz2,tz);
2116
2117 /**************************
2118 * CALCULATE INTERACTIONS *
2119 **************************/
2120
2121 r10 = _mm_mul_pd(rsq10,rinv10);
2122
2123 /* Calculate table index by multiplying r with table scale and truncate to integer */
2124 rt = _mm_mul_pd(r10,vftabscale);
2125 vfitab = _mm_cvttpd_epi32(rt);
2126 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2127 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2128
2129 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2130 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2131 F = _mm_setzero_pd();
2132 GMX_MM_TRANSPOSE2_PD(Y,F);
2133 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2134 H = _mm_setzero_pd();
2135 GMX_MM_TRANSPOSE2_PD(G,H);
2136 Heps = _mm_mul_pd(vfeps,H);
2137 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2138 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2139 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq10,FF),_mm_mul_pd(vftabscale,rinv10)));
2140
2141 fscal = felec;
2142
2143 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2144
2145 /* Calculate temporary vectorial force */
2146 tx = _mm_mul_pd(fscal,dx10);
2147 ty = _mm_mul_pd(fscal,dy10);
2148 tz = _mm_mul_pd(fscal,dz10);
2149
2150 /* Update vectorial force */
2151 fix1 = _mm_add_pd(fix1,tx);
2152 fiy1 = _mm_add_pd(fiy1,ty);
2153 fiz1 = _mm_add_pd(fiz1,tz);
2154
2155 fjx0 = _mm_add_pd(fjx0,tx);
2156 fjy0 = _mm_add_pd(fjy0,ty);
2157 fjz0 = _mm_add_pd(fjz0,tz);
2158
2159 /**************************
2160 * CALCULATE INTERACTIONS *
2161 **************************/
2162
2163 r11 = _mm_mul_pd(rsq11,rinv11);
2164
2165 /* Calculate table index by multiplying r with table scale and truncate to integer */
2166 rt = _mm_mul_pd(r11,vftabscale);
2167 vfitab = _mm_cvttpd_epi32(rt);
2168 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2169 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2170
2171 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2172 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2173 F = _mm_setzero_pd();
2174 GMX_MM_TRANSPOSE2_PD(Y,F);
2175 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2176 H = _mm_setzero_pd();
2177 GMX_MM_TRANSPOSE2_PD(G,H);
2178 Heps = _mm_mul_pd(vfeps,H);
2179 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2180 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2181 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
2182
2183 fscal = felec;
2184
2185 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2186
2187 /* Calculate temporary vectorial force */
2188 tx = _mm_mul_pd(fscal,dx11);
2189 ty = _mm_mul_pd(fscal,dy11);
2190 tz = _mm_mul_pd(fscal,dz11);
2191
2192 /* Update vectorial force */
2193 fix1 = _mm_add_pd(fix1,tx);
2194 fiy1 = _mm_add_pd(fiy1,ty);
2195 fiz1 = _mm_add_pd(fiz1,tz);
2196
2197 fjx1 = _mm_add_pd(fjx1,tx);
2198 fjy1 = _mm_add_pd(fjy1,ty);
2199 fjz1 = _mm_add_pd(fjz1,tz);
2200
2201 /**************************
2202 * CALCULATE INTERACTIONS *
2203 **************************/
2204
2205 r12 = _mm_mul_pd(rsq12,rinv12);
2206
2207 /* Calculate table index by multiplying r with table scale and truncate to integer */
2208 rt = _mm_mul_pd(r12,vftabscale);
2209 vfitab = _mm_cvttpd_epi32(rt);
2210 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2211 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2212
2213 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2214 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2215 F = _mm_setzero_pd();
2216 GMX_MM_TRANSPOSE2_PD(Y,F);
2217 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2218 H = _mm_setzero_pd();
2219 GMX_MM_TRANSPOSE2_PD(G,H);
2220 Heps = _mm_mul_pd(vfeps,H);
2221 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2222 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2223 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
2224
2225 fscal = felec;
2226
2227 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2228
2229 /* Calculate temporary vectorial force */
2230 tx = _mm_mul_pd(fscal,dx12);
2231 ty = _mm_mul_pd(fscal,dy12);
2232 tz = _mm_mul_pd(fscal,dz12);
2233
2234 /* Update vectorial force */
2235 fix1 = _mm_add_pd(fix1,tx);
2236 fiy1 = _mm_add_pd(fiy1,ty);
2237 fiz1 = _mm_add_pd(fiz1,tz);
2238
2239 fjx2 = _mm_add_pd(fjx2,tx);
2240 fjy2 = _mm_add_pd(fjy2,ty);
2241 fjz2 = _mm_add_pd(fjz2,tz);
2242
2243 /**************************
2244 * CALCULATE INTERACTIONS *
2245 **************************/
2246
2247 r20 = _mm_mul_pd(rsq20,rinv20);
2248
2249 /* Calculate table index by multiplying r with table scale and truncate to integer */
2250 rt = _mm_mul_pd(r20,vftabscale);
2251 vfitab = _mm_cvttpd_epi32(rt);
2252 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2253 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2254
2255 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2256 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2257 F = _mm_setzero_pd();
2258 GMX_MM_TRANSPOSE2_PD(Y,F);
2259 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2260 H = _mm_setzero_pd();
2261 GMX_MM_TRANSPOSE2_PD(G,H);
2262 Heps = _mm_mul_pd(vfeps,H);
2263 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2264 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2265 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq20,FF),_mm_mul_pd(vftabscale,rinv20)));
2266
2267 fscal = felec;
2268
2269 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2270
2271 /* Calculate temporary vectorial force */
2272 tx = _mm_mul_pd(fscal,dx20);
2273 ty = _mm_mul_pd(fscal,dy20);
2274 tz = _mm_mul_pd(fscal,dz20);
2275
2276 /* Update vectorial force */
2277 fix2 = _mm_add_pd(fix2,tx);
2278 fiy2 = _mm_add_pd(fiy2,ty);
2279 fiz2 = _mm_add_pd(fiz2,tz);
2280
2281 fjx0 = _mm_add_pd(fjx0,tx);
2282 fjy0 = _mm_add_pd(fjy0,ty);
2283 fjz0 = _mm_add_pd(fjz0,tz);
2284
2285 /**************************
2286 * CALCULATE INTERACTIONS *
2287 **************************/
2288
2289 r21 = _mm_mul_pd(rsq21,rinv21);
2290
2291 /* Calculate table index by multiplying r with table scale and truncate to integer */
2292 rt = _mm_mul_pd(r21,vftabscale);
2293 vfitab = _mm_cvttpd_epi32(rt);
2294 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2295 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2296
2297 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2298 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2299 F = _mm_setzero_pd();
2300 GMX_MM_TRANSPOSE2_PD(Y,F);
2301 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2302 H = _mm_setzero_pd();
2303 GMX_MM_TRANSPOSE2_PD(G,H);
2304 Heps = _mm_mul_pd(vfeps,H);
2305 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2306 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2307 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
2308
2309 fscal = felec;
2310
2311 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2312
2313 /* Calculate temporary vectorial force */
2314 tx = _mm_mul_pd(fscal,dx21);
2315 ty = _mm_mul_pd(fscal,dy21);
2316 tz = _mm_mul_pd(fscal,dz21);
2317
2318 /* Update vectorial force */
2319 fix2 = _mm_add_pd(fix2,tx);
2320 fiy2 = _mm_add_pd(fiy2,ty);
2321 fiz2 = _mm_add_pd(fiz2,tz);
2322
2323 fjx1 = _mm_add_pd(fjx1,tx);
2324 fjy1 = _mm_add_pd(fjy1,ty);
2325 fjz1 = _mm_add_pd(fjz1,tz);
2326
2327 /**************************
2328 * CALCULATE INTERACTIONS *
2329 **************************/
2330
2331 r22 = _mm_mul_pd(rsq22,rinv22);
2332
2333 /* Calculate table index by multiplying r with table scale and truncate to integer */
2334 rt = _mm_mul_pd(r22,vftabscale);
2335 vfitab = _mm_cvttpd_epi32(rt);
2336 vfeps = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2337 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2338
2339 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2340 Y = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2341 F = _mm_setzero_pd();
2342 GMX_MM_TRANSPOSE2_PD(Y,F);
2343 G = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2344 H = _mm_setzero_pd();
2345 GMX_MM_TRANSPOSE2_PD(G,H);
2346 Heps = _mm_mul_pd(vfeps,H);
2347 Fp = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2348 FF = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2349 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
2350
2351 fscal = felec;
2352
2353 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2354
2355 /* Calculate temporary vectorial force */
2356 tx = _mm_mul_pd(fscal,dx22);
2357 ty = _mm_mul_pd(fscal,dy22);
2358 tz = _mm_mul_pd(fscal,dz22);
2359
2360 /* Update vectorial force */
2361 fix2 = _mm_add_pd(fix2,tx);
2362 fiy2 = _mm_add_pd(fiy2,ty);
2363 fiz2 = _mm_add_pd(fiz2,tz);
2364
2365 fjx2 = _mm_add_pd(fjx2,tx);
2366 fjy2 = _mm_add_pd(fjy2,ty);
2367 fjz2 = _mm_add_pd(fjz2,tz);
2368
2369 gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
2370
2371 /* Inner loop uses 373 flops */
2372 }
2373
2374 /* End of innermost loop */
2375
2376 gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
2377 f+i_coord_offset,fshift+i_shift_offset);
2378
2379 /* Increment number of inner iterations */
2380 inneriter += j_index_end - j_index_start;
2381
2382 /* Outer loop uses 18 flops */
2383 }
2384
2385 /* Increment number of outer iterations */
2386 outeriter += nri;
2387
2388 /* Update outer/inner flops */
2389
2390 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_F,outeriter*18 + inneriter*373);
2391 }
The ultimate molecular dynamics simulation package