?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Removed simple.h from nb_kernel_sse4_1_XX
[gromacs.git] / src / gromacs / gmxlib / nonbonded / nb_kernel_sse4_1_single / nb_kernel_ElecCoul_VdwCSTab_GeomW4W4_sse4_1_single.c
blob9ebd9218cbf3f06524c14d2556e532d72398e95a
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 sse4_1_single 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_sse4_1_single.h"
49 #include "kernelutil_x86_sse4_1_single.h"
50
51 /*
52 * Gromacs nonbonded kernel: nb_kernel_ElecCoul_VdwCSTab_GeomW4W4_VF_sse4_1_single
53 * Electrostatics interaction: Coulomb
54 * VdW interaction: CubicSplineTable
55 * Geometry: Water4-Water4
56 * Calculate force/pot: PotentialAndForce
57 */
58 void
59 nb_kernel_ElecCoul_VdwCSTab_GeomW4W4_VF_sse4_1_single
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,C,D refer to j loop unrolling done with SSE, e.g. for the four 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,jnrC,jnrD;
76 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
77 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
78 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
79 real rcutoff_scalar;
80 real *shiftvec,*fshift,*x,*f;
81 real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
82 real scratch[4*DIM];
83 __m128 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
84 int vdwioffset0;
85 __m128 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
86 int vdwioffset1;
87 __m128 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
88 int vdwioffset2;
89 __m128 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
90 int vdwioffset3;
91 __m128 ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
92 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
93 __m128 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
94 int vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
95 __m128 jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
96 int vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
97 __m128 jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
98 int vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
99 __m128 jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
100 __m128 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
101 __m128 dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
102 __m128 dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
103 __m128 dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
104 __m128 dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
105 __m128 dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
106 __m128 dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
107 __m128 dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
108 __m128 dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
109 __m128 dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
110 __m128 velec,felec,velecsum,facel,crf,krf,krf2;
111 real *charge;
112 int nvdwtype;
113 __m128 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
114 int *vdwtype;
115 real *vdwparam;
116 __m128 one_sixth = _mm_set1_ps(1.0/6.0);
117 __m128 one_twelfth = _mm_set1_ps(1.0/12.0);
118 __m128i vfitab;
119 __m128i ifour = _mm_set1_epi32(4);
120 __m128 rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
121 real *vftab;
122 __m128 dummy_mask,cutoff_mask;
123 __m128 signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
124 __m128 one = _mm_set1_ps(1.0);
125 __m128 two = _mm_set1_ps(2.0);
126 x = xx[0];
127 f = ff[0];
128
129 nri = nlist->nri;
130 iinr = nlist->iinr;
131 jindex = nlist->jindex;
132 jjnr = nlist->jjnr;
133 shiftidx = nlist->shift;
134 gid = nlist->gid;
135 shiftvec = fr->shift_vec[0];
136 fshift = fr->fshift[0];
137 facel = _mm_set1_ps(fr->epsfac);
138 charge = mdatoms->chargeA;
139 nvdwtype = fr->ntype;
140 vdwparam = fr->nbfp;
141 vdwtype = mdatoms->typeA;
142
143 vftab = kernel_data->table_vdw->data;
144 vftabscale = _mm_set1_ps(kernel_data->table_vdw->scale);
145
146 /* Setup water-specific parameters */
147 inr = nlist->iinr[0];
148 iq1 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1]));
149 iq2 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2]));
150 iq3 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+3]));
151 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
152
153 jq1 = _mm_set1_ps(charge[inr+1]);
154 jq2 = _mm_set1_ps(charge[inr+2]);
155 jq3 = _mm_set1_ps(charge[inr+3]);
156 vdwjidx0A = 2*vdwtype[inr+0];
157 c6_00 = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A]);
158 c12_00 = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A+1]);
159 qq11 = _mm_mul_ps(iq1,jq1);
160 qq12 = _mm_mul_ps(iq1,jq2);
161 qq13 = _mm_mul_ps(iq1,jq3);
162 qq21 = _mm_mul_ps(iq2,jq1);
163 qq22 = _mm_mul_ps(iq2,jq2);
164 qq23 = _mm_mul_ps(iq2,jq3);
165 qq31 = _mm_mul_ps(iq3,jq1);
166 qq32 = _mm_mul_ps(iq3,jq2);
167 qq33 = _mm_mul_ps(iq3,jq3);
168
169 /* Avoid stupid compiler warnings */
170 jnrA = jnrB = jnrC = jnrD = 0;
171 j_coord_offsetA = 0;
172 j_coord_offsetB = 0;
173 j_coord_offsetC = 0;
174 j_coord_offsetD = 0;
175
176 outeriter = 0;
177 inneriter = 0;
178
179 for(iidx=0;iidx<4*DIM;iidx++)
180 {
181 scratch[iidx] = 0.0;
182 }
183
184 /* Start outer loop over neighborlists */
185 for(iidx=0; iidx<nri; iidx++)
186 {
187 /* Load shift vector for this list */
188 i_shift_offset = DIM*shiftidx[iidx];
189
190 /* Load limits for loop over neighbors */
191 j_index_start = jindex[iidx];
192 j_index_end = jindex[iidx+1];
193
194 /* Get outer coordinate index */
195 inr = iinr[iidx];
196 i_coord_offset = DIM*inr;
197
198 /* Load i particle coords and add shift vector */
199 gmx_mm_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
200 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
201
202 fix0 = _mm_setzero_ps();
203 fiy0 = _mm_setzero_ps();
204 fiz0 = _mm_setzero_ps();
205 fix1 = _mm_setzero_ps();
206 fiy1 = _mm_setzero_ps();
207 fiz1 = _mm_setzero_ps();
208 fix2 = _mm_setzero_ps();
209 fiy2 = _mm_setzero_ps();
210 fiz2 = _mm_setzero_ps();
211 fix3 = _mm_setzero_ps();
212 fiy3 = _mm_setzero_ps();
213 fiz3 = _mm_setzero_ps();
214
215 /* Reset potential sums */
216 velecsum = _mm_setzero_ps();
217 vvdwsum = _mm_setzero_ps();
218
219 /* Start inner kernel loop */
220 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
221 {
222
223 /* Get j neighbor index, and coordinate index */
224 jnrA = jjnr[jidx];
225 jnrB = jjnr[jidx+1];
226 jnrC = jjnr[jidx+2];
227 jnrD = jjnr[jidx+3];
228 j_coord_offsetA = DIM*jnrA;
229 j_coord_offsetB = DIM*jnrB;
230 j_coord_offsetC = DIM*jnrC;
231 j_coord_offsetD = DIM*jnrD;
232
233 /* load j atom coordinates */
234 gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
235 x+j_coord_offsetC,x+j_coord_offsetD,
236 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
237 &jy2,&jz2,&jx3,&jy3,&jz3);
238
239 /* Calculate displacement vector */
240 dx00 = _mm_sub_ps(ix0,jx0);
241 dy00 = _mm_sub_ps(iy0,jy0);
242 dz00 = _mm_sub_ps(iz0,jz0);
243 dx11 = _mm_sub_ps(ix1,jx1);
244 dy11 = _mm_sub_ps(iy1,jy1);
245 dz11 = _mm_sub_ps(iz1,jz1);
246 dx12 = _mm_sub_ps(ix1,jx2);
247 dy12 = _mm_sub_ps(iy1,jy2);
248 dz12 = _mm_sub_ps(iz1,jz2);
249 dx13 = _mm_sub_ps(ix1,jx3);
250 dy13 = _mm_sub_ps(iy1,jy3);
251 dz13 = _mm_sub_ps(iz1,jz3);
252 dx21 = _mm_sub_ps(ix2,jx1);
253 dy21 = _mm_sub_ps(iy2,jy1);
254 dz21 = _mm_sub_ps(iz2,jz1);
255 dx22 = _mm_sub_ps(ix2,jx2);
256 dy22 = _mm_sub_ps(iy2,jy2);
257 dz22 = _mm_sub_ps(iz2,jz2);
258 dx23 = _mm_sub_ps(ix2,jx3);
259 dy23 = _mm_sub_ps(iy2,jy3);
260 dz23 = _mm_sub_ps(iz2,jz3);
261 dx31 = _mm_sub_ps(ix3,jx1);
262 dy31 = _mm_sub_ps(iy3,jy1);
263 dz31 = _mm_sub_ps(iz3,jz1);
264 dx32 = _mm_sub_ps(ix3,jx2);
265 dy32 = _mm_sub_ps(iy3,jy2);
266 dz32 = _mm_sub_ps(iz3,jz2);
267 dx33 = _mm_sub_ps(ix3,jx3);
268 dy33 = _mm_sub_ps(iy3,jy3);
269 dz33 = _mm_sub_ps(iz3,jz3);
270
271 /* Calculate squared distance and things based on it */
272 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
273 rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
274 rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
275 rsq13 = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
276 rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
277 rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
278 rsq23 = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
279 rsq31 = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
280 rsq32 = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
281 rsq33 = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
282
283 rinv00 = gmx_mm_invsqrt_ps(rsq00);
284 rinv11 = gmx_mm_invsqrt_ps(rsq11);
285 rinv12 = gmx_mm_invsqrt_ps(rsq12);
286 rinv13 = gmx_mm_invsqrt_ps(rsq13);
287 rinv21 = gmx_mm_invsqrt_ps(rsq21);
288 rinv22 = gmx_mm_invsqrt_ps(rsq22);
289 rinv23 = gmx_mm_invsqrt_ps(rsq23);
290 rinv31 = gmx_mm_invsqrt_ps(rsq31);
291 rinv32 = gmx_mm_invsqrt_ps(rsq32);
292 rinv33 = gmx_mm_invsqrt_ps(rsq33);
293
294 rinvsq11 = _mm_mul_ps(rinv11,rinv11);
295 rinvsq12 = _mm_mul_ps(rinv12,rinv12);
296 rinvsq13 = _mm_mul_ps(rinv13,rinv13);
297 rinvsq21 = _mm_mul_ps(rinv21,rinv21);
298 rinvsq22 = _mm_mul_ps(rinv22,rinv22);
299 rinvsq23 = _mm_mul_ps(rinv23,rinv23);
300 rinvsq31 = _mm_mul_ps(rinv31,rinv31);
301 rinvsq32 = _mm_mul_ps(rinv32,rinv32);
302 rinvsq33 = _mm_mul_ps(rinv33,rinv33);
303
304 fjx0 = _mm_setzero_ps();
305 fjy0 = _mm_setzero_ps();
306 fjz0 = _mm_setzero_ps();
307 fjx1 = _mm_setzero_ps();
308 fjy1 = _mm_setzero_ps();
309 fjz1 = _mm_setzero_ps();
310 fjx2 = _mm_setzero_ps();
311 fjy2 = _mm_setzero_ps();
312 fjz2 = _mm_setzero_ps();
313 fjx3 = _mm_setzero_ps();
314 fjy3 = _mm_setzero_ps();
315 fjz3 = _mm_setzero_ps();
316
317 /**************************
318 * CALCULATE INTERACTIONS *
319 **************************/
320
321 r00 = _mm_mul_ps(rsq00,rinv00);
322
323 /* Calculate table index by multiplying r with table scale and truncate to integer */
324 rt = _mm_mul_ps(r00,vftabscale);
325 vfitab = _mm_cvttps_epi32(rt);
326 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
327 vfitab = _mm_slli_epi32(vfitab,3);
328
329 /* CUBIC SPLINE TABLE DISPERSION */
330 Y = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
331 F = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
332 G = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
333 H = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
334 _MM_TRANSPOSE4_PS(Y,F,G,H);
335 Heps = _mm_mul_ps(vfeps,H);
336 Fp = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
337 VV = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
338 vvdw6 = _mm_mul_ps(c6_00,VV);
339 FF = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
340 fvdw6 = _mm_mul_ps(c6_00,FF);
341
342 /* CUBIC SPLINE TABLE REPULSION */
343 vfitab = _mm_add_epi32(vfitab,ifour);
344 Y = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
345 F = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
346 G = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
347 H = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
348 _MM_TRANSPOSE4_PS(Y,F,G,H);
349 Heps = _mm_mul_ps(vfeps,H);
350 Fp = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
351 VV = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
352 vvdw12 = _mm_mul_ps(c12_00,VV);
353 FF = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
354 fvdw12 = _mm_mul_ps(c12_00,FF);
355 vvdw = _mm_add_ps(vvdw12,vvdw6);
356 fvdw = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
357
358 /* Update potential sum for this i atom from the interaction with this j atom. */
359 vvdwsum = _mm_add_ps(vvdwsum,vvdw);
360
361 fscal = fvdw;
362
363 /* Calculate temporary vectorial force */
364 tx = _mm_mul_ps(fscal,dx00);
365 ty = _mm_mul_ps(fscal,dy00);
366 tz = _mm_mul_ps(fscal,dz00);
367
368 /* Update vectorial force */
369 fix0 = _mm_add_ps(fix0,tx);
370 fiy0 = _mm_add_ps(fiy0,ty);
371 fiz0 = _mm_add_ps(fiz0,tz);
372
373 fjx0 = _mm_add_ps(fjx0,tx);
374 fjy0 = _mm_add_ps(fjy0,ty);
375 fjz0 = _mm_add_ps(fjz0,tz);
376
377 /**************************
378 * CALCULATE INTERACTIONS *
379 **************************/
380
381 /* COULOMB ELECTROSTATICS */
382 velec = _mm_mul_ps(qq11,rinv11);
383 felec = _mm_mul_ps(velec,rinvsq11);
384
385 /* Update potential sum for this i atom from the interaction with this j atom. */
386 velecsum = _mm_add_ps(velecsum,velec);
387
388 fscal = felec;
389
390 /* Calculate temporary vectorial force */
391 tx = _mm_mul_ps(fscal,dx11);
392 ty = _mm_mul_ps(fscal,dy11);
393 tz = _mm_mul_ps(fscal,dz11);
394
395 /* Update vectorial force */
396 fix1 = _mm_add_ps(fix1,tx);
397 fiy1 = _mm_add_ps(fiy1,ty);
398 fiz1 = _mm_add_ps(fiz1,tz);
399
400 fjx1 = _mm_add_ps(fjx1,tx);
401 fjy1 = _mm_add_ps(fjy1,ty);
402 fjz1 = _mm_add_ps(fjz1,tz);
403
404 /**************************
405 * CALCULATE INTERACTIONS *
406 **************************/
407
408 /* COULOMB ELECTROSTATICS */
409 velec = _mm_mul_ps(qq12,rinv12);
410 felec = _mm_mul_ps(velec,rinvsq12);
411
412 /* Update potential sum for this i atom from the interaction with this j atom. */
413 velecsum = _mm_add_ps(velecsum,velec);
414
415 fscal = felec;
416
417 /* Calculate temporary vectorial force */
418 tx = _mm_mul_ps(fscal,dx12);
419 ty = _mm_mul_ps(fscal,dy12);
420 tz = _mm_mul_ps(fscal,dz12);
421
422 /* Update vectorial force */
423 fix1 = _mm_add_ps(fix1,tx);
424 fiy1 = _mm_add_ps(fiy1,ty);
425 fiz1 = _mm_add_ps(fiz1,tz);
426
427 fjx2 = _mm_add_ps(fjx2,tx);
428 fjy2 = _mm_add_ps(fjy2,ty);
429 fjz2 = _mm_add_ps(fjz2,tz);
430
431 /**************************
432 * CALCULATE INTERACTIONS *
433 **************************/
434
435 /* COULOMB ELECTROSTATICS */
436 velec = _mm_mul_ps(qq13,rinv13);
437 felec = _mm_mul_ps(velec,rinvsq13);
438
439 /* Update potential sum for this i atom from the interaction with this j atom. */
440 velecsum = _mm_add_ps(velecsum,velec);
441
442 fscal = felec;
443
444 /* Calculate temporary vectorial force */
445 tx = _mm_mul_ps(fscal,dx13);
446 ty = _mm_mul_ps(fscal,dy13);
447 tz = _mm_mul_ps(fscal,dz13);
448
449 /* Update vectorial force */
450 fix1 = _mm_add_ps(fix1,tx);
451 fiy1 = _mm_add_ps(fiy1,ty);
452 fiz1 = _mm_add_ps(fiz1,tz);
453
454 fjx3 = _mm_add_ps(fjx3,tx);
455 fjy3 = _mm_add_ps(fjy3,ty);
456 fjz3 = _mm_add_ps(fjz3,tz);
457
458 /**************************
459 * CALCULATE INTERACTIONS *
460 **************************/
461
462 /* COULOMB ELECTROSTATICS */
463 velec = _mm_mul_ps(qq21,rinv21);
464 felec = _mm_mul_ps(velec,rinvsq21);
465
466 /* Update potential sum for this i atom from the interaction with this j atom. */
467 velecsum = _mm_add_ps(velecsum,velec);
468
469 fscal = felec;
470
471 /* Calculate temporary vectorial force */
472 tx = _mm_mul_ps(fscal,dx21);
473 ty = _mm_mul_ps(fscal,dy21);
474 tz = _mm_mul_ps(fscal,dz21);
475
476 /* Update vectorial force */
477 fix2 = _mm_add_ps(fix2,tx);
478 fiy2 = _mm_add_ps(fiy2,ty);
479 fiz2 = _mm_add_ps(fiz2,tz);
480
481 fjx1 = _mm_add_ps(fjx1,tx);
482 fjy1 = _mm_add_ps(fjy1,ty);
483 fjz1 = _mm_add_ps(fjz1,tz);
484
485 /**************************
486 * CALCULATE INTERACTIONS *
487 **************************/
488
489 /* COULOMB ELECTROSTATICS */
490 velec = _mm_mul_ps(qq22,rinv22);
491 felec = _mm_mul_ps(velec,rinvsq22);
492
493 /* Update potential sum for this i atom from the interaction with this j atom. */
494 velecsum = _mm_add_ps(velecsum,velec);
495
496 fscal = felec;
497
498 /* Calculate temporary vectorial force */
499 tx = _mm_mul_ps(fscal,dx22);
500 ty = _mm_mul_ps(fscal,dy22);
501 tz = _mm_mul_ps(fscal,dz22);
502
503 /* Update vectorial force */
504 fix2 = _mm_add_ps(fix2,tx);
505 fiy2 = _mm_add_ps(fiy2,ty);
506 fiz2 = _mm_add_ps(fiz2,tz);
507
508 fjx2 = _mm_add_ps(fjx2,tx);
509 fjy2 = _mm_add_ps(fjy2,ty);
510 fjz2 = _mm_add_ps(fjz2,tz);
511
512 /**************************
513 * CALCULATE INTERACTIONS *
514 **************************/
515
516 /* COULOMB ELECTROSTATICS */
517 velec = _mm_mul_ps(qq23,rinv23);
518 felec = _mm_mul_ps(velec,rinvsq23);
519
520 /* Update potential sum for this i atom from the interaction with this j atom. */
521 velecsum = _mm_add_ps(velecsum,velec);
522
523 fscal = felec;
524
525 /* Calculate temporary vectorial force */
526 tx = _mm_mul_ps(fscal,dx23);
527 ty = _mm_mul_ps(fscal,dy23);
528 tz = _mm_mul_ps(fscal,dz23);
529
530 /* Update vectorial force */
531 fix2 = _mm_add_ps(fix2,tx);
532 fiy2 = _mm_add_ps(fiy2,ty);
533 fiz2 = _mm_add_ps(fiz2,tz);
534
535 fjx3 = _mm_add_ps(fjx3,tx);
536 fjy3 = _mm_add_ps(fjy3,ty);
537 fjz3 = _mm_add_ps(fjz3,tz);
538
539 /**************************
540 * CALCULATE INTERACTIONS *
541 **************************/
542
543 /* COULOMB ELECTROSTATICS */
544 velec = _mm_mul_ps(qq31,rinv31);
545 felec = _mm_mul_ps(velec,rinvsq31);
546
547 /* Update potential sum for this i atom from the interaction with this j atom. */
548 velecsum = _mm_add_ps(velecsum,velec);
549
550 fscal = felec;
551
552 /* Calculate temporary vectorial force */
553 tx = _mm_mul_ps(fscal,dx31);
554 ty = _mm_mul_ps(fscal,dy31);
555 tz = _mm_mul_ps(fscal,dz31);
556
557 /* Update vectorial force */
558 fix3 = _mm_add_ps(fix3,tx);
559 fiy3 = _mm_add_ps(fiy3,ty);
560 fiz3 = _mm_add_ps(fiz3,tz);
561
562 fjx1 = _mm_add_ps(fjx1,tx);
563 fjy1 = _mm_add_ps(fjy1,ty);
564 fjz1 = _mm_add_ps(fjz1,tz);
565
566 /**************************
567 * CALCULATE INTERACTIONS *
568 **************************/
569
570 /* COULOMB ELECTROSTATICS */
571 velec = _mm_mul_ps(qq32,rinv32);
572 felec = _mm_mul_ps(velec,rinvsq32);
573
574 /* Update potential sum for this i atom from the interaction with this j atom. */
575 velecsum = _mm_add_ps(velecsum,velec);
576
577 fscal = felec;
578
579 /* Calculate temporary vectorial force */
580 tx = _mm_mul_ps(fscal,dx32);
581 ty = _mm_mul_ps(fscal,dy32);
582 tz = _mm_mul_ps(fscal,dz32);
583
584 /* Update vectorial force */
585 fix3 = _mm_add_ps(fix3,tx);
586 fiy3 = _mm_add_ps(fiy3,ty);
587 fiz3 = _mm_add_ps(fiz3,tz);
588
589 fjx2 = _mm_add_ps(fjx2,tx);
590 fjy2 = _mm_add_ps(fjy2,ty);
591 fjz2 = _mm_add_ps(fjz2,tz);
592
593 /**************************
594 * CALCULATE INTERACTIONS *
595 **************************/
596
597 /* COULOMB ELECTROSTATICS */
598 velec = _mm_mul_ps(qq33,rinv33);
599 felec = _mm_mul_ps(velec,rinvsq33);
600
601 /* Update potential sum for this i atom from the interaction with this j atom. */
602 velecsum = _mm_add_ps(velecsum,velec);
603
604 fscal = felec;
605
606 /* Calculate temporary vectorial force */
607 tx = _mm_mul_ps(fscal,dx33);
608 ty = _mm_mul_ps(fscal,dy33);
609 tz = _mm_mul_ps(fscal,dz33);
610
611 /* Update vectorial force */
612 fix3 = _mm_add_ps(fix3,tx);
613 fiy3 = _mm_add_ps(fiy3,ty);
614 fiz3 = _mm_add_ps(fiz3,tz);
615
616 fjx3 = _mm_add_ps(fjx3,tx);
617 fjy3 = _mm_add_ps(fjy3,ty);
618 fjz3 = _mm_add_ps(fjz3,tz);
619
620 fjptrA = f+j_coord_offsetA;
621 fjptrB = f+j_coord_offsetB;
622 fjptrC = f+j_coord_offsetC;
623 fjptrD = f+j_coord_offsetD;
624
625 gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
626 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
627 fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
628
629 /* Inner loop uses 311 flops */
630 }
631
632 if(jidx<j_index_end)
633 {
634
635 /* Get j neighbor index, and coordinate index */
636 jnrlistA = jjnr[jidx];
637 jnrlistB = jjnr[jidx+1];
638 jnrlistC = jjnr[jidx+2];
639 jnrlistD = jjnr[jidx+3];
640 /* Sign of each element will be negative for non-real atoms.
641 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
642 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
643 */
644 dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
645 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
646 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
647 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
648 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
649 j_coord_offsetA = DIM*jnrA;
650 j_coord_offsetB = DIM*jnrB;
651 j_coord_offsetC = DIM*jnrC;
652 j_coord_offsetD = DIM*jnrD;
653
654 /* load j atom coordinates */
655 gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
656 x+j_coord_offsetC,x+j_coord_offsetD,
657 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
658 &jy2,&jz2,&jx3,&jy3,&jz3);
659
660 /* Calculate displacement vector */
661 dx00 = _mm_sub_ps(ix0,jx0);
662 dy00 = _mm_sub_ps(iy0,jy0);
663 dz00 = _mm_sub_ps(iz0,jz0);
664 dx11 = _mm_sub_ps(ix1,jx1);
665 dy11 = _mm_sub_ps(iy1,jy1);
666 dz11 = _mm_sub_ps(iz1,jz1);
667 dx12 = _mm_sub_ps(ix1,jx2);
668 dy12 = _mm_sub_ps(iy1,jy2);
669 dz12 = _mm_sub_ps(iz1,jz2);
670 dx13 = _mm_sub_ps(ix1,jx3);
671 dy13 = _mm_sub_ps(iy1,jy3);
672 dz13 = _mm_sub_ps(iz1,jz3);
673 dx21 = _mm_sub_ps(ix2,jx1);
674 dy21 = _mm_sub_ps(iy2,jy1);
675 dz21 = _mm_sub_ps(iz2,jz1);
676 dx22 = _mm_sub_ps(ix2,jx2);
677 dy22 = _mm_sub_ps(iy2,jy2);
678 dz22 = _mm_sub_ps(iz2,jz2);
679 dx23 = _mm_sub_ps(ix2,jx3);
680 dy23 = _mm_sub_ps(iy2,jy3);
681 dz23 = _mm_sub_ps(iz2,jz3);
682 dx31 = _mm_sub_ps(ix3,jx1);
683 dy31 = _mm_sub_ps(iy3,jy1);
684 dz31 = _mm_sub_ps(iz3,jz1);
685 dx32 = _mm_sub_ps(ix3,jx2);
686 dy32 = _mm_sub_ps(iy3,jy2);
687 dz32 = _mm_sub_ps(iz3,jz2);
688 dx33 = _mm_sub_ps(ix3,jx3);
689 dy33 = _mm_sub_ps(iy3,jy3);
690 dz33 = _mm_sub_ps(iz3,jz3);
691
692 /* Calculate squared distance and things based on it */
693 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
694 rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
695 rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
696 rsq13 = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
697 rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
698 rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
699 rsq23 = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
700 rsq31 = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
701 rsq32 = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
702 rsq33 = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
703
704 rinv00 = gmx_mm_invsqrt_ps(rsq00);
705 rinv11 = gmx_mm_invsqrt_ps(rsq11);
706 rinv12 = gmx_mm_invsqrt_ps(rsq12);
707 rinv13 = gmx_mm_invsqrt_ps(rsq13);
708 rinv21 = gmx_mm_invsqrt_ps(rsq21);
709 rinv22 = gmx_mm_invsqrt_ps(rsq22);
710 rinv23 = gmx_mm_invsqrt_ps(rsq23);
711 rinv31 = gmx_mm_invsqrt_ps(rsq31);
712 rinv32 = gmx_mm_invsqrt_ps(rsq32);
713 rinv33 = gmx_mm_invsqrt_ps(rsq33);
714
715 rinvsq11 = _mm_mul_ps(rinv11,rinv11);
716 rinvsq12 = _mm_mul_ps(rinv12,rinv12);
717 rinvsq13 = _mm_mul_ps(rinv13,rinv13);
718 rinvsq21 = _mm_mul_ps(rinv21,rinv21);
719 rinvsq22 = _mm_mul_ps(rinv22,rinv22);
720 rinvsq23 = _mm_mul_ps(rinv23,rinv23);
721 rinvsq31 = _mm_mul_ps(rinv31,rinv31);
722 rinvsq32 = _mm_mul_ps(rinv32,rinv32);
723 rinvsq33 = _mm_mul_ps(rinv33,rinv33);
724
725 fjx0 = _mm_setzero_ps();
726 fjy0 = _mm_setzero_ps();
727 fjz0 = _mm_setzero_ps();
728 fjx1 = _mm_setzero_ps();
729 fjy1 = _mm_setzero_ps();
730 fjz1 = _mm_setzero_ps();
731 fjx2 = _mm_setzero_ps();
732 fjy2 = _mm_setzero_ps();
733 fjz2 = _mm_setzero_ps();
734 fjx3 = _mm_setzero_ps();
735 fjy3 = _mm_setzero_ps();
736 fjz3 = _mm_setzero_ps();
737
738 /**************************
739 * CALCULATE INTERACTIONS *
740 **************************/
741
742 r00 = _mm_mul_ps(rsq00,rinv00);
743 r00 = _mm_andnot_ps(dummy_mask,r00);
744
745 /* Calculate table index by multiplying r with table scale and truncate to integer */
746 rt = _mm_mul_ps(r00,vftabscale);
747 vfitab = _mm_cvttps_epi32(rt);
748 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
749 vfitab = _mm_slli_epi32(vfitab,3);
750
751 /* CUBIC SPLINE TABLE DISPERSION */
752 Y = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
753 F = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
754 G = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
755 H = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
756 _MM_TRANSPOSE4_PS(Y,F,G,H);
757 Heps = _mm_mul_ps(vfeps,H);
758 Fp = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
759 VV = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
760 vvdw6 = _mm_mul_ps(c6_00,VV);
761 FF = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
762 fvdw6 = _mm_mul_ps(c6_00,FF);
763
764 /* CUBIC SPLINE TABLE REPULSION */
765 vfitab = _mm_add_epi32(vfitab,ifour);
766 Y = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
767 F = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
768 G = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
769 H = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
770 _MM_TRANSPOSE4_PS(Y,F,G,H);
771 Heps = _mm_mul_ps(vfeps,H);
772 Fp = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
773 VV = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
774 vvdw12 = _mm_mul_ps(c12_00,VV);
775 FF = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
776 fvdw12 = _mm_mul_ps(c12_00,FF);
777 vvdw = _mm_add_ps(vvdw12,vvdw6);
778 fvdw = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
779
780 /* Update potential sum for this i atom from the interaction with this j atom. */
781 vvdw = _mm_andnot_ps(dummy_mask,vvdw);
782 vvdwsum = _mm_add_ps(vvdwsum,vvdw);
783
784 fscal = fvdw;
785
786 fscal = _mm_andnot_ps(dummy_mask,fscal);
787
788 /* Calculate temporary vectorial force */
789 tx = _mm_mul_ps(fscal,dx00);
790 ty = _mm_mul_ps(fscal,dy00);
791 tz = _mm_mul_ps(fscal,dz00);
792
793 /* Update vectorial force */
794 fix0 = _mm_add_ps(fix0,tx);
795 fiy0 = _mm_add_ps(fiy0,ty);
796 fiz0 = _mm_add_ps(fiz0,tz);
797
798 fjx0 = _mm_add_ps(fjx0,tx);
799 fjy0 = _mm_add_ps(fjy0,ty);
800 fjz0 = _mm_add_ps(fjz0,tz);
801
802 /**************************
803 * CALCULATE INTERACTIONS *
804 **************************/
805
806 /* COULOMB ELECTROSTATICS */
807 velec = _mm_mul_ps(qq11,rinv11);
808 felec = _mm_mul_ps(velec,rinvsq11);
809
810 /* Update potential sum for this i atom from the interaction with this j atom. */
811 velec = _mm_andnot_ps(dummy_mask,velec);
812 velecsum = _mm_add_ps(velecsum,velec);
813
814 fscal = felec;
815
816 fscal = _mm_andnot_ps(dummy_mask,fscal);
817
818 /* Calculate temporary vectorial force */
819 tx = _mm_mul_ps(fscal,dx11);
820 ty = _mm_mul_ps(fscal,dy11);
821 tz = _mm_mul_ps(fscal,dz11);
822
823 /* Update vectorial force */
824 fix1 = _mm_add_ps(fix1,tx);
825 fiy1 = _mm_add_ps(fiy1,ty);
826 fiz1 = _mm_add_ps(fiz1,tz);
827
828 fjx1 = _mm_add_ps(fjx1,tx);
829 fjy1 = _mm_add_ps(fjy1,ty);
830 fjz1 = _mm_add_ps(fjz1,tz);
831
832 /**************************
833 * CALCULATE INTERACTIONS *
834 **************************/
835
836 /* COULOMB ELECTROSTATICS */
837 velec = _mm_mul_ps(qq12,rinv12);
838 felec = _mm_mul_ps(velec,rinvsq12);
839
840 /* Update potential sum for this i atom from the interaction with this j atom. */
841 velec = _mm_andnot_ps(dummy_mask,velec);
842 velecsum = _mm_add_ps(velecsum,velec);
843
844 fscal = felec;
845
846 fscal = _mm_andnot_ps(dummy_mask,fscal);
847
848 /* Calculate temporary vectorial force */
849 tx = _mm_mul_ps(fscal,dx12);
850 ty = _mm_mul_ps(fscal,dy12);
851 tz = _mm_mul_ps(fscal,dz12);
852
853 /* Update vectorial force */
854 fix1 = _mm_add_ps(fix1,tx);
855 fiy1 = _mm_add_ps(fiy1,ty);
856 fiz1 = _mm_add_ps(fiz1,tz);
857
858 fjx2 = _mm_add_ps(fjx2,tx);
859 fjy2 = _mm_add_ps(fjy2,ty);
860 fjz2 = _mm_add_ps(fjz2,tz);
861
862 /**************************
863 * CALCULATE INTERACTIONS *
864 **************************/
865
866 /* COULOMB ELECTROSTATICS */
867 velec = _mm_mul_ps(qq13,rinv13);
868 felec = _mm_mul_ps(velec,rinvsq13);
869
870 /* Update potential sum for this i atom from the interaction with this j atom. */
871 velec = _mm_andnot_ps(dummy_mask,velec);
872 velecsum = _mm_add_ps(velecsum,velec);
873
874 fscal = felec;
875
876 fscal = _mm_andnot_ps(dummy_mask,fscal);
877
878 /* Calculate temporary vectorial force */
879 tx = _mm_mul_ps(fscal,dx13);
880 ty = _mm_mul_ps(fscal,dy13);
881 tz = _mm_mul_ps(fscal,dz13);
882
883 /* Update vectorial force */
884 fix1 = _mm_add_ps(fix1,tx);
885 fiy1 = _mm_add_ps(fiy1,ty);
886 fiz1 = _mm_add_ps(fiz1,tz);
887
888 fjx3 = _mm_add_ps(fjx3,tx);
889 fjy3 = _mm_add_ps(fjy3,ty);
890 fjz3 = _mm_add_ps(fjz3,tz);
891
892 /**************************
893 * CALCULATE INTERACTIONS *
894 **************************/
895
896 /* COULOMB ELECTROSTATICS */
897 velec = _mm_mul_ps(qq21,rinv21);
898 felec = _mm_mul_ps(velec,rinvsq21);
899
900 /* Update potential sum for this i atom from the interaction with this j atom. */
901 velec = _mm_andnot_ps(dummy_mask,velec);
902 velecsum = _mm_add_ps(velecsum,velec);
903
904 fscal = felec;
905
906 fscal = _mm_andnot_ps(dummy_mask,fscal);
907
908 /* Calculate temporary vectorial force */
909 tx = _mm_mul_ps(fscal,dx21);
910 ty = _mm_mul_ps(fscal,dy21);
911 tz = _mm_mul_ps(fscal,dz21);
912
913 /* Update vectorial force */
914 fix2 = _mm_add_ps(fix2,tx);
915 fiy2 = _mm_add_ps(fiy2,ty);
916 fiz2 = _mm_add_ps(fiz2,tz);
917
918 fjx1 = _mm_add_ps(fjx1,tx);
919 fjy1 = _mm_add_ps(fjy1,ty);
920 fjz1 = _mm_add_ps(fjz1,tz);
921
922 /**************************
923 * CALCULATE INTERACTIONS *
924 **************************/
925
926 /* COULOMB ELECTROSTATICS */
927 velec = _mm_mul_ps(qq22,rinv22);
928 felec = _mm_mul_ps(velec,rinvsq22);
929
930 /* Update potential sum for this i atom from the interaction with this j atom. */
931 velec = _mm_andnot_ps(dummy_mask,velec);
932 velecsum = _mm_add_ps(velecsum,velec);
933
934 fscal = felec;
935
936 fscal = _mm_andnot_ps(dummy_mask,fscal);
937
938 /* Calculate temporary vectorial force */
939 tx = _mm_mul_ps(fscal,dx22);
940 ty = _mm_mul_ps(fscal,dy22);
941 tz = _mm_mul_ps(fscal,dz22);
942
943 /* Update vectorial force */
944 fix2 = _mm_add_ps(fix2,tx);
945 fiy2 = _mm_add_ps(fiy2,ty);
946 fiz2 = _mm_add_ps(fiz2,tz);
947
948 fjx2 = _mm_add_ps(fjx2,tx);
949 fjy2 = _mm_add_ps(fjy2,ty);
950 fjz2 = _mm_add_ps(fjz2,tz);
951
952 /**************************
953 * CALCULATE INTERACTIONS *
954 **************************/
955
956 /* COULOMB ELECTROSTATICS */
957 velec = _mm_mul_ps(qq23,rinv23);
958 felec = _mm_mul_ps(velec,rinvsq23);
959
960 /* Update potential sum for this i atom from the interaction with this j atom. */
961 velec = _mm_andnot_ps(dummy_mask,velec);
962 velecsum = _mm_add_ps(velecsum,velec);
963
964 fscal = felec;
965
966 fscal = _mm_andnot_ps(dummy_mask,fscal);
967
968 /* Calculate temporary vectorial force */
969 tx = _mm_mul_ps(fscal,dx23);
970 ty = _mm_mul_ps(fscal,dy23);
971 tz = _mm_mul_ps(fscal,dz23);
972
973 /* Update vectorial force */
974 fix2 = _mm_add_ps(fix2,tx);
975 fiy2 = _mm_add_ps(fiy2,ty);
976 fiz2 = _mm_add_ps(fiz2,tz);
977
978 fjx3 = _mm_add_ps(fjx3,tx);
979 fjy3 = _mm_add_ps(fjy3,ty);
980 fjz3 = _mm_add_ps(fjz3,tz);
981
982 /**************************
983 * CALCULATE INTERACTIONS *
984 **************************/
985
986 /* COULOMB ELECTROSTATICS */
987 velec = _mm_mul_ps(qq31,rinv31);
988 felec = _mm_mul_ps(velec,rinvsq31);
989
990 /* Update potential sum for this i atom from the interaction with this j atom. */
991 velec = _mm_andnot_ps(dummy_mask,velec);
992 velecsum = _mm_add_ps(velecsum,velec);
993
994 fscal = felec;
995
996 fscal = _mm_andnot_ps(dummy_mask,fscal);
997
998 /* Calculate temporary vectorial force */
999 tx = _mm_mul_ps(fscal,dx31);
1000 ty = _mm_mul_ps(fscal,dy31);
1001 tz = _mm_mul_ps(fscal,dz31);
1002
1003 /* Update vectorial force */
1004 fix3 = _mm_add_ps(fix3,tx);
1005 fiy3 = _mm_add_ps(fiy3,ty);
1006 fiz3 = _mm_add_ps(fiz3,tz);
1007
1008 fjx1 = _mm_add_ps(fjx1,tx);
1009 fjy1 = _mm_add_ps(fjy1,ty);
1010 fjz1 = _mm_add_ps(fjz1,tz);
1011
1012 /**************************
1013 * CALCULATE INTERACTIONS *
1014 **************************/
1015
1016 /* COULOMB ELECTROSTATICS */
1017 velec = _mm_mul_ps(qq32,rinv32);
1018 felec = _mm_mul_ps(velec,rinvsq32);
1019
1020 /* Update potential sum for this i atom from the interaction with this j atom. */
1021 velec = _mm_andnot_ps(dummy_mask,velec);
1022 velecsum = _mm_add_ps(velecsum,velec);
1023
1024 fscal = felec;
1025
1026 fscal = _mm_andnot_ps(dummy_mask,fscal);
1027
1028 /* Calculate temporary vectorial force */
1029 tx = _mm_mul_ps(fscal,dx32);
1030 ty = _mm_mul_ps(fscal,dy32);
1031 tz = _mm_mul_ps(fscal,dz32);
1032
1033 /* Update vectorial force */
1034 fix3 = _mm_add_ps(fix3,tx);
1035 fiy3 = _mm_add_ps(fiy3,ty);
1036 fiz3 = _mm_add_ps(fiz3,tz);
1037
1038 fjx2 = _mm_add_ps(fjx2,tx);
1039 fjy2 = _mm_add_ps(fjy2,ty);
1040 fjz2 = _mm_add_ps(fjz2,tz);
1041
1042 /**************************
1043 * CALCULATE INTERACTIONS *
1044 **************************/
1045
1046 /* COULOMB ELECTROSTATICS */
1047 velec = _mm_mul_ps(qq33,rinv33);
1048 felec = _mm_mul_ps(velec,rinvsq33);
1049
1050 /* Update potential sum for this i atom from the interaction with this j atom. */
1051 velec = _mm_andnot_ps(dummy_mask,velec);
1052 velecsum = _mm_add_ps(velecsum,velec);
1053
1054 fscal = felec;
1055
1056 fscal = _mm_andnot_ps(dummy_mask,fscal);
1057
1058 /* Calculate temporary vectorial force */
1059 tx = _mm_mul_ps(fscal,dx33);
1060 ty = _mm_mul_ps(fscal,dy33);
1061 tz = _mm_mul_ps(fscal,dz33);
1062
1063 /* Update vectorial force */
1064 fix3 = _mm_add_ps(fix3,tx);
1065 fiy3 = _mm_add_ps(fiy3,ty);
1066 fiz3 = _mm_add_ps(fiz3,tz);
1067
1068 fjx3 = _mm_add_ps(fjx3,tx);
1069 fjy3 = _mm_add_ps(fjy3,ty);
1070 fjz3 = _mm_add_ps(fjz3,tz);
1071
1072 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1073 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1074 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1075 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1076
1077 gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
1078 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1079 fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1080
1081 /* Inner loop uses 312 flops */
1082 }
1083
1084 /* End of innermost loop */
1085
1086 gmx_mm_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1087 f+i_coord_offset,fshift+i_shift_offset);
1088
1089 ggid = gid[iidx];
1090 /* Update potential energies */
1091 gmx_mm_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
1092 gmx_mm_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
1093
1094 /* Increment number of inner iterations */
1095 inneriter += j_index_end - j_index_start;
1096
1097 /* Outer loop uses 26 flops */
1098 }
1099
1100 /* Increment number of outer iterations */
1101 outeriter += nri;
1102
1103 /* Update outer/inner flops */
1104
1105 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*312);
1106 }
1107 /*
1108 * Gromacs nonbonded kernel: nb_kernel_ElecCoul_VdwCSTab_GeomW4W4_F_sse4_1_single
1109 * Electrostatics interaction: Coulomb
1110 * VdW interaction: CubicSplineTable
1111 * Geometry: Water4-Water4
1112 * Calculate force/pot: Force
1113 */
1114 void
1115 nb_kernel_ElecCoul_VdwCSTab_GeomW4W4_F_sse4_1_single
1116 (t_nblist * gmx_restrict nlist,
1117 rvec * gmx_restrict xx,
1118 rvec * gmx_restrict ff,
1119 t_forcerec * gmx_restrict fr,
1120 t_mdatoms * gmx_restrict mdatoms,
1121 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1122 t_nrnb * gmx_restrict nrnb)
1123 {
1124 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1125 * just 0 for non-waters.
1126 * Suffixes A,B,C,D refer to j loop unrolling done with SSE, e.g. for the four different
1127 * jnr indices corresponding to data put in the four positions in the SIMD register.
1128 */
1129 int i_shift_offset,i_coord_offset,outeriter,inneriter;
1130 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1131 int jnrA,jnrB,jnrC,jnrD;
1132 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1133 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1134 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
1135 real rcutoff_scalar;
1136 real *shiftvec,*fshift,*x,*f;
1137 real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
1138 real scratch[4*DIM];
1139 __m128 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1140 int vdwioffset0;
1141 __m128 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1142 int vdwioffset1;
1143 __m128 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1144 int vdwioffset2;
1145 __m128 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1146 int vdwioffset3;
1147 __m128 ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1148 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
1149 __m128 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1150 int vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
1151 __m128 jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1152 int vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
1153 __m128 jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1154 int vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
1155 __m128 jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1156 __m128 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1157 __m128 dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1158 __m128 dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1159 __m128 dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1160 __m128 dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1161 __m128 dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1162 __m128 dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1163 __m128 dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1164 __m128 dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1165 __m128 dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1166 __m128 velec,felec,velecsum,facel,crf,krf,krf2;
1167 real *charge;
1168 int nvdwtype;
1169 __m128 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1170 int *vdwtype;
1171 real *vdwparam;
1172 __m128 one_sixth = _mm_set1_ps(1.0/6.0);
1173 __m128 one_twelfth = _mm_set1_ps(1.0/12.0);
1174 __m128i vfitab;
1175 __m128i ifour = _mm_set1_epi32(4);
1176 __m128 rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
1177 real *vftab;
1178 __m128 dummy_mask,cutoff_mask;
1179 __m128 signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
1180 __m128 one = _mm_set1_ps(1.0);
1181 __m128 two = _mm_set1_ps(2.0);
1182 x = xx[0];
1183 f = ff[0];
1184
1185 nri = nlist->nri;
1186 iinr = nlist->iinr;
1187 jindex = nlist->jindex;
1188 jjnr = nlist->jjnr;
1189 shiftidx = nlist->shift;
1190 gid = nlist->gid;
1191 shiftvec = fr->shift_vec[0];
1192 fshift = fr->fshift[0];
1193 facel = _mm_set1_ps(fr->epsfac);
1194 charge = mdatoms->chargeA;
1195 nvdwtype = fr->ntype;
1196 vdwparam = fr->nbfp;
1197 vdwtype = mdatoms->typeA;
1198
1199 vftab = kernel_data->table_vdw->data;
1200 vftabscale = _mm_set1_ps(kernel_data->table_vdw->scale);
1201
1202 /* Setup water-specific parameters */
1203 inr = nlist->iinr[0];
1204 iq1 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1]));
1205 iq2 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2]));
1206 iq3 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+3]));
1207 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
1208
1209 jq1 = _mm_set1_ps(charge[inr+1]);
1210 jq2 = _mm_set1_ps(charge[inr+2]);
1211 jq3 = _mm_set1_ps(charge[inr+3]);
1212 vdwjidx0A = 2*vdwtype[inr+0];
1213 c6_00 = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A]);
1214 c12_00 = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A+1]);
1215 qq11 = _mm_mul_ps(iq1,jq1);
1216 qq12 = _mm_mul_ps(iq1,jq2);
1217 qq13 = _mm_mul_ps(iq1,jq3);
1218 qq21 = _mm_mul_ps(iq2,jq1);
1219 qq22 = _mm_mul_ps(iq2,jq2);
1220 qq23 = _mm_mul_ps(iq2,jq3);
1221 qq31 = _mm_mul_ps(iq3,jq1);
1222 qq32 = _mm_mul_ps(iq3,jq2);
1223 qq33 = _mm_mul_ps(iq3,jq3);
1224
1225 /* Avoid stupid compiler warnings */
1226 jnrA = jnrB = jnrC = jnrD = 0;
1227 j_coord_offsetA = 0;
1228 j_coord_offsetB = 0;
1229 j_coord_offsetC = 0;
1230 j_coord_offsetD = 0;
1231
1232 outeriter = 0;
1233 inneriter = 0;
1234
1235 for(iidx=0;iidx<4*DIM;iidx++)
1236 {
1237 scratch[iidx] = 0.0;
1238 }
1239
1240 /* Start outer loop over neighborlists */
1241 for(iidx=0; iidx<nri; iidx++)
1242 {
1243 /* Load shift vector for this list */
1244 i_shift_offset = DIM*shiftidx[iidx];
1245
1246 /* Load limits for loop over neighbors */
1247 j_index_start = jindex[iidx];
1248 j_index_end = jindex[iidx+1];
1249
1250 /* Get outer coordinate index */
1251 inr = iinr[iidx];
1252 i_coord_offset = DIM*inr;
1253
1254 /* Load i particle coords and add shift vector */
1255 gmx_mm_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1256 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1257
1258 fix0 = _mm_setzero_ps();
1259 fiy0 = _mm_setzero_ps();
1260 fiz0 = _mm_setzero_ps();
1261 fix1 = _mm_setzero_ps();
1262 fiy1 = _mm_setzero_ps();
1263 fiz1 = _mm_setzero_ps();
1264 fix2 = _mm_setzero_ps();
1265 fiy2 = _mm_setzero_ps();
1266 fiz2 = _mm_setzero_ps();
1267 fix3 = _mm_setzero_ps();
1268 fiy3 = _mm_setzero_ps();
1269 fiz3 = _mm_setzero_ps();
1270
1271 /* Start inner kernel loop */
1272 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
1273 {
1274
1275 /* Get j neighbor index, and coordinate index */
1276 jnrA = jjnr[jidx];
1277 jnrB = jjnr[jidx+1];
1278 jnrC = jjnr[jidx+2];
1279 jnrD = jjnr[jidx+3];
1280 j_coord_offsetA = DIM*jnrA;
1281 j_coord_offsetB = DIM*jnrB;
1282 j_coord_offsetC = DIM*jnrC;
1283 j_coord_offsetD = DIM*jnrD;
1284
1285 /* load j atom coordinates */
1286 gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1287 x+j_coord_offsetC,x+j_coord_offsetD,
1288 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1289 &jy2,&jz2,&jx3,&jy3,&jz3);
1290
1291 /* Calculate displacement vector */
1292 dx00 = _mm_sub_ps(ix0,jx0);
1293 dy00 = _mm_sub_ps(iy0,jy0);
1294 dz00 = _mm_sub_ps(iz0,jz0);
1295 dx11 = _mm_sub_ps(ix1,jx1);
1296 dy11 = _mm_sub_ps(iy1,jy1);
1297 dz11 = _mm_sub_ps(iz1,jz1);
1298 dx12 = _mm_sub_ps(ix1,jx2);
1299 dy12 = _mm_sub_ps(iy1,jy2);
1300 dz12 = _mm_sub_ps(iz1,jz2);
1301 dx13 = _mm_sub_ps(ix1,jx3);
1302 dy13 = _mm_sub_ps(iy1,jy3);
1303 dz13 = _mm_sub_ps(iz1,jz3);
1304 dx21 = _mm_sub_ps(ix2,jx1);
1305 dy21 = _mm_sub_ps(iy2,jy1);
1306 dz21 = _mm_sub_ps(iz2,jz1);
1307 dx22 = _mm_sub_ps(ix2,jx2);
1308 dy22 = _mm_sub_ps(iy2,jy2);
1309 dz22 = _mm_sub_ps(iz2,jz2);
1310 dx23 = _mm_sub_ps(ix2,jx3);
1311 dy23 = _mm_sub_ps(iy2,jy3);
1312 dz23 = _mm_sub_ps(iz2,jz3);
1313 dx31 = _mm_sub_ps(ix3,jx1);
1314 dy31 = _mm_sub_ps(iy3,jy1);
1315 dz31 = _mm_sub_ps(iz3,jz1);
1316 dx32 = _mm_sub_ps(ix3,jx2);
1317 dy32 = _mm_sub_ps(iy3,jy2);
1318 dz32 = _mm_sub_ps(iz3,jz2);
1319 dx33 = _mm_sub_ps(ix3,jx3);
1320 dy33 = _mm_sub_ps(iy3,jy3);
1321 dz33 = _mm_sub_ps(iz3,jz3);
1322
1323 /* Calculate squared distance and things based on it */
1324 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
1325 rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
1326 rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
1327 rsq13 = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
1328 rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
1329 rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
1330 rsq23 = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
1331 rsq31 = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
1332 rsq32 = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
1333 rsq33 = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
1334
1335 rinv00 = gmx_mm_invsqrt_ps(rsq00);
1336 rinv11 = gmx_mm_invsqrt_ps(rsq11);
1337 rinv12 = gmx_mm_invsqrt_ps(rsq12);
1338 rinv13 = gmx_mm_invsqrt_ps(rsq13);
1339 rinv21 = gmx_mm_invsqrt_ps(rsq21);
1340 rinv22 = gmx_mm_invsqrt_ps(rsq22);
1341 rinv23 = gmx_mm_invsqrt_ps(rsq23);
1342 rinv31 = gmx_mm_invsqrt_ps(rsq31);
1343 rinv32 = gmx_mm_invsqrt_ps(rsq32);
1344 rinv33 = gmx_mm_invsqrt_ps(rsq33);
1345
1346 rinvsq11 = _mm_mul_ps(rinv11,rinv11);
1347 rinvsq12 = _mm_mul_ps(rinv12,rinv12);
1348 rinvsq13 = _mm_mul_ps(rinv13,rinv13);
1349 rinvsq21 = _mm_mul_ps(rinv21,rinv21);
1350 rinvsq22 = _mm_mul_ps(rinv22,rinv22);
1351 rinvsq23 = _mm_mul_ps(rinv23,rinv23);
1352 rinvsq31 = _mm_mul_ps(rinv31,rinv31);
1353 rinvsq32 = _mm_mul_ps(rinv32,rinv32);
1354 rinvsq33 = _mm_mul_ps(rinv33,rinv33);
1355
1356 fjx0 = _mm_setzero_ps();
1357 fjy0 = _mm_setzero_ps();
1358 fjz0 = _mm_setzero_ps();
1359 fjx1 = _mm_setzero_ps();
1360 fjy1 = _mm_setzero_ps();
1361 fjz1 = _mm_setzero_ps();
1362 fjx2 = _mm_setzero_ps();
1363 fjy2 = _mm_setzero_ps();
1364 fjz2 = _mm_setzero_ps();
1365 fjx3 = _mm_setzero_ps();
1366 fjy3 = _mm_setzero_ps();
1367 fjz3 = _mm_setzero_ps();
1368
1369 /**************************
1370 * CALCULATE INTERACTIONS *
1371 **************************/
1372
1373 r00 = _mm_mul_ps(rsq00,rinv00);
1374
1375 /* Calculate table index by multiplying r with table scale and truncate to integer */
1376 rt = _mm_mul_ps(r00,vftabscale);
1377 vfitab = _mm_cvttps_epi32(rt);
1378 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1379 vfitab = _mm_slli_epi32(vfitab,3);
1380
1381 /* CUBIC SPLINE TABLE DISPERSION */
1382 Y = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1383 F = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1384 G = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1385 H = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1386 _MM_TRANSPOSE4_PS(Y,F,G,H);
1387 Heps = _mm_mul_ps(vfeps,H);
1388 Fp = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1389 FF = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1390 fvdw6 = _mm_mul_ps(c6_00,FF);
1391
1392 /* CUBIC SPLINE TABLE REPULSION */
1393 vfitab = _mm_add_epi32(vfitab,ifour);
1394 Y = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1395 F = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1396 G = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1397 H = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1398 _MM_TRANSPOSE4_PS(Y,F,G,H);
1399 Heps = _mm_mul_ps(vfeps,H);
1400 Fp = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1401 FF = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1402 fvdw12 = _mm_mul_ps(c12_00,FF);
1403 fvdw = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
1404
1405 fscal = fvdw;
1406
1407 /* Calculate temporary vectorial force */
1408 tx = _mm_mul_ps(fscal,dx00);
1409 ty = _mm_mul_ps(fscal,dy00);
1410 tz = _mm_mul_ps(fscal,dz00);
1411
1412 /* Update vectorial force */
1413 fix0 = _mm_add_ps(fix0,tx);
1414 fiy0 = _mm_add_ps(fiy0,ty);
1415 fiz0 = _mm_add_ps(fiz0,tz);
1416
1417 fjx0 = _mm_add_ps(fjx0,tx);
1418 fjy0 = _mm_add_ps(fjy0,ty);
1419 fjz0 = _mm_add_ps(fjz0,tz);
1420
1421 /**************************
1422 * CALCULATE INTERACTIONS *
1423 **************************/
1424
1425 /* COULOMB ELECTROSTATICS */
1426 velec = _mm_mul_ps(qq11,rinv11);
1427 felec = _mm_mul_ps(velec,rinvsq11);
1428
1429 fscal = felec;
1430
1431 /* Calculate temporary vectorial force */
1432 tx = _mm_mul_ps(fscal,dx11);
1433 ty = _mm_mul_ps(fscal,dy11);
1434 tz = _mm_mul_ps(fscal,dz11);
1435
1436 /* Update vectorial force */
1437 fix1 = _mm_add_ps(fix1,tx);
1438 fiy1 = _mm_add_ps(fiy1,ty);
1439 fiz1 = _mm_add_ps(fiz1,tz);
1440
1441 fjx1 = _mm_add_ps(fjx1,tx);
1442 fjy1 = _mm_add_ps(fjy1,ty);
1443 fjz1 = _mm_add_ps(fjz1,tz);
1444
1445 /**************************
1446 * CALCULATE INTERACTIONS *
1447 **************************/
1448
1449 /* COULOMB ELECTROSTATICS */
1450 velec = _mm_mul_ps(qq12,rinv12);
1451 felec = _mm_mul_ps(velec,rinvsq12);
1452
1453 fscal = felec;
1454
1455 /* Calculate temporary vectorial force */
1456 tx = _mm_mul_ps(fscal,dx12);
1457 ty = _mm_mul_ps(fscal,dy12);
1458 tz = _mm_mul_ps(fscal,dz12);
1459
1460 /* Update vectorial force */
1461 fix1 = _mm_add_ps(fix1,tx);
1462 fiy1 = _mm_add_ps(fiy1,ty);
1463 fiz1 = _mm_add_ps(fiz1,tz);
1464
1465 fjx2 = _mm_add_ps(fjx2,tx);
1466 fjy2 = _mm_add_ps(fjy2,ty);
1467 fjz2 = _mm_add_ps(fjz2,tz);
1468
1469 /**************************
1470 * CALCULATE INTERACTIONS *
1471 **************************/
1472
1473 /* COULOMB ELECTROSTATICS */
1474 velec = _mm_mul_ps(qq13,rinv13);
1475 felec = _mm_mul_ps(velec,rinvsq13);
1476
1477 fscal = felec;
1478
1479 /* Calculate temporary vectorial force */
1480 tx = _mm_mul_ps(fscal,dx13);
1481 ty = _mm_mul_ps(fscal,dy13);
1482 tz = _mm_mul_ps(fscal,dz13);
1483
1484 /* Update vectorial force */
1485 fix1 = _mm_add_ps(fix1,tx);
1486 fiy1 = _mm_add_ps(fiy1,ty);
1487 fiz1 = _mm_add_ps(fiz1,tz);
1488
1489 fjx3 = _mm_add_ps(fjx3,tx);
1490 fjy3 = _mm_add_ps(fjy3,ty);
1491 fjz3 = _mm_add_ps(fjz3,tz);
1492
1493 /**************************
1494 * CALCULATE INTERACTIONS *
1495 **************************/
1496
1497 /* COULOMB ELECTROSTATICS */
1498 velec = _mm_mul_ps(qq21,rinv21);
1499 felec = _mm_mul_ps(velec,rinvsq21);
1500
1501 fscal = felec;
1502
1503 /* Calculate temporary vectorial force */
1504 tx = _mm_mul_ps(fscal,dx21);
1505 ty = _mm_mul_ps(fscal,dy21);
1506 tz = _mm_mul_ps(fscal,dz21);
1507
1508 /* Update vectorial force */
1509 fix2 = _mm_add_ps(fix2,tx);
1510 fiy2 = _mm_add_ps(fiy2,ty);
1511 fiz2 = _mm_add_ps(fiz2,tz);
1512
1513 fjx1 = _mm_add_ps(fjx1,tx);
1514 fjy1 = _mm_add_ps(fjy1,ty);
1515 fjz1 = _mm_add_ps(fjz1,tz);
1516
1517 /**************************
1518 * CALCULATE INTERACTIONS *
1519 **************************/
1520
1521 /* COULOMB ELECTROSTATICS */
1522 velec = _mm_mul_ps(qq22,rinv22);
1523 felec = _mm_mul_ps(velec,rinvsq22);
1524
1525 fscal = felec;
1526
1527 /* Calculate temporary vectorial force */
1528 tx = _mm_mul_ps(fscal,dx22);
1529 ty = _mm_mul_ps(fscal,dy22);
1530 tz = _mm_mul_ps(fscal,dz22);
1531
1532 /* Update vectorial force */
1533 fix2 = _mm_add_ps(fix2,tx);
1534 fiy2 = _mm_add_ps(fiy2,ty);
1535 fiz2 = _mm_add_ps(fiz2,tz);
1536
1537 fjx2 = _mm_add_ps(fjx2,tx);
1538 fjy2 = _mm_add_ps(fjy2,ty);
1539 fjz2 = _mm_add_ps(fjz2,tz);
1540
1541 /**************************
1542 * CALCULATE INTERACTIONS *
1543 **************************/
1544
1545 /* COULOMB ELECTROSTATICS */
1546 velec = _mm_mul_ps(qq23,rinv23);
1547 felec = _mm_mul_ps(velec,rinvsq23);
1548
1549 fscal = felec;
1550
1551 /* Calculate temporary vectorial force */
1552 tx = _mm_mul_ps(fscal,dx23);
1553 ty = _mm_mul_ps(fscal,dy23);
1554 tz = _mm_mul_ps(fscal,dz23);
1555
1556 /* Update vectorial force */
1557 fix2 = _mm_add_ps(fix2,tx);
1558 fiy2 = _mm_add_ps(fiy2,ty);
1559 fiz2 = _mm_add_ps(fiz2,tz);
1560
1561 fjx3 = _mm_add_ps(fjx3,tx);
1562 fjy3 = _mm_add_ps(fjy3,ty);
1563 fjz3 = _mm_add_ps(fjz3,tz);
1564
1565 /**************************
1566 * CALCULATE INTERACTIONS *
1567 **************************/
1568
1569 /* COULOMB ELECTROSTATICS */
1570 velec = _mm_mul_ps(qq31,rinv31);
1571 felec = _mm_mul_ps(velec,rinvsq31);
1572
1573 fscal = felec;
1574
1575 /* Calculate temporary vectorial force */
1576 tx = _mm_mul_ps(fscal,dx31);
1577 ty = _mm_mul_ps(fscal,dy31);
1578 tz = _mm_mul_ps(fscal,dz31);
1579
1580 /* Update vectorial force */
1581 fix3 = _mm_add_ps(fix3,tx);
1582 fiy3 = _mm_add_ps(fiy3,ty);
1583 fiz3 = _mm_add_ps(fiz3,tz);
1584
1585 fjx1 = _mm_add_ps(fjx1,tx);
1586 fjy1 = _mm_add_ps(fjy1,ty);
1587 fjz1 = _mm_add_ps(fjz1,tz);
1588
1589 /**************************
1590 * CALCULATE INTERACTIONS *
1591 **************************/
1592
1593 /* COULOMB ELECTROSTATICS */
1594 velec = _mm_mul_ps(qq32,rinv32);
1595 felec = _mm_mul_ps(velec,rinvsq32);
1596
1597 fscal = felec;
1598
1599 /* Calculate temporary vectorial force */
1600 tx = _mm_mul_ps(fscal,dx32);
1601 ty = _mm_mul_ps(fscal,dy32);
1602 tz = _mm_mul_ps(fscal,dz32);
1603
1604 /* Update vectorial force */
1605 fix3 = _mm_add_ps(fix3,tx);
1606 fiy3 = _mm_add_ps(fiy3,ty);
1607 fiz3 = _mm_add_ps(fiz3,tz);
1608
1609 fjx2 = _mm_add_ps(fjx2,tx);
1610 fjy2 = _mm_add_ps(fjy2,ty);
1611 fjz2 = _mm_add_ps(fjz2,tz);
1612
1613 /**************************
1614 * CALCULATE INTERACTIONS *
1615 **************************/
1616
1617 /* COULOMB ELECTROSTATICS */
1618 velec = _mm_mul_ps(qq33,rinv33);
1619 felec = _mm_mul_ps(velec,rinvsq33);
1620
1621 fscal = felec;
1622
1623 /* Calculate temporary vectorial force */
1624 tx = _mm_mul_ps(fscal,dx33);
1625 ty = _mm_mul_ps(fscal,dy33);
1626 tz = _mm_mul_ps(fscal,dz33);
1627
1628 /* Update vectorial force */
1629 fix3 = _mm_add_ps(fix3,tx);
1630 fiy3 = _mm_add_ps(fiy3,ty);
1631 fiz3 = _mm_add_ps(fiz3,tz);
1632
1633 fjx3 = _mm_add_ps(fjx3,tx);
1634 fjy3 = _mm_add_ps(fjy3,ty);
1635 fjz3 = _mm_add_ps(fjz3,tz);
1636
1637 fjptrA = f+j_coord_offsetA;
1638 fjptrB = f+j_coord_offsetB;
1639 fjptrC = f+j_coord_offsetC;
1640 fjptrD = f+j_coord_offsetD;
1641
1642 gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
1643 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1644 fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1645
1646 /* Inner loop uses 294 flops */
1647 }
1648
1649 if(jidx<j_index_end)
1650 {
1651
1652 /* Get j neighbor index, and coordinate index */
1653 jnrlistA = jjnr[jidx];
1654 jnrlistB = jjnr[jidx+1];
1655 jnrlistC = jjnr[jidx+2];
1656 jnrlistD = jjnr[jidx+3];
1657 /* Sign of each element will be negative for non-real atoms.
1658 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1659 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1660 */
1661 dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
1662 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
1663 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
1664 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
1665 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
1666 j_coord_offsetA = DIM*jnrA;
1667 j_coord_offsetB = DIM*jnrB;
1668 j_coord_offsetC = DIM*jnrC;
1669 j_coord_offsetD = DIM*jnrD;
1670
1671 /* load j atom coordinates */
1672 gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1673 x+j_coord_offsetC,x+j_coord_offsetD,
1674 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1675 &jy2,&jz2,&jx3,&jy3,&jz3);
1676
1677 /* Calculate displacement vector */
1678 dx00 = _mm_sub_ps(ix0,jx0);
1679 dy00 = _mm_sub_ps(iy0,jy0);
1680 dz00 = _mm_sub_ps(iz0,jz0);
1681 dx11 = _mm_sub_ps(ix1,jx1);
1682 dy11 = _mm_sub_ps(iy1,jy1);
1683 dz11 = _mm_sub_ps(iz1,jz1);
1684 dx12 = _mm_sub_ps(ix1,jx2);
1685 dy12 = _mm_sub_ps(iy1,jy2);
1686 dz12 = _mm_sub_ps(iz1,jz2);
1687 dx13 = _mm_sub_ps(ix1,jx3);
1688 dy13 = _mm_sub_ps(iy1,jy3);
1689 dz13 = _mm_sub_ps(iz1,jz3);
1690 dx21 = _mm_sub_ps(ix2,jx1);
1691 dy21 = _mm_sub_ps(iy2,jy1);
1692 dz21 = _mm_sub_ps(iz2,jz1);
1693 dx22 = _mm_sub_ps(ix2,jx2);
1694 dy22 = _mm_sub_ps(iy2,jy2);
1695 dz22 = _mm_sub_ps(iz2,jz2);
1696 dx23 = _mm_sub_ps(ix2,jx3);
1697 dy23 = _mm_sub_ps(iy2,jy3);
1698 dz23 = _mm_sub_ps(iz2,jz3);
1699 dx31 = _mm_sub_ps(ix3,jx1);
1700 dy31 = _mm_sub_ps(iy3,jy1);
1701 dz31 = _mm_sub_ps(iz3,jz1);
1702 dx32 = _mm_sub_ps(ix3,jx2);
1703 dy32 = _mm_sub_ps(iy3,jy2);
1704 dz32 = _mm_sub_ps(iz3,jz2);
1705 dx33 = _mm_sub_ps(ix3,jx3);
1706 dy33 = _mm_sub_ps(iy3,jy3);
1707 dz33 = _mm_sub_ps(iz3,jz3);
1708
1709 /* Calculate squared distance and things based on it */
1710 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
1711 rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
1712 rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
1713 rsq13 = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
1714 rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
1715 rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
1716 rsq23 = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
1717 rsq31 = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
1718 rsq32 = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
1719 rsq33 = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
1720
1721 rinv00 = gmx_mm_invsqrt_ps(rsq00);
1722 rinv11 = gmx_mm_invsqrt_ps(rsq11);
1723 rinv12 = gmx_mm_invsqrt_ps(rsq12);
1724 rinv13 = gmx_mm_invsqrt_ps(rsq13);
1725 rinv21 = gmx_mm_invsqrt_ps(rsq21);
1726 rinv22 = gmx_mm_invsqrt_ps(rsq22);
1727 rinv23 = gmx_mm_invsqrt_ps(rsq23);
1728 rinv31 = gmx_mm_invsqrt_ps(rsq31);
1729 rinv32 = gmx_mm_invsqrt_ps(rsq32);
1730 rinv33 = gmx_mm_invsqrt_ps(rsq33);
1731
1732 rinvsq11 = _mm_mul_ps(rinv11,rinv11);
1733 rinvsq12 = _mm_mul_ps(rinv12,rinv12);
1734 rinvsq13 = _mm_mul_ps(rinv13,rinv13);
1735 rinvsq21 = _mm_mul_ps(rinv21,rinv21);
1736 rinvsq22 = _mm_mul_ps(rinv22,rinv22);
1737 rinvsq23 = _mm_mul_ps(rinv23,rinv23);
1738 rinvsq31 = _mm_mul_ps(rinv31,rinv31);
1739 rinvsq32 = _mm_mul_ps(rinv32,rinv32);
1740 rinvsq33 = _mm_mul_ps(rinv33,rinv33);
1741
1742 fjx0 = _mm_setzero_ps();
1743 fjy0 = _mm_setzero_ps();
1744 fjz0 = _mm_setzero_ps();
1745 fjx1 = _mm_setzero_ps();
1746 fjy1 = _mm_setzero_ps();
1747 fjz1 = _mm_setzero_ps();
1748 fjx2 = _mm_setzero_ps();
1749 fjy2 = _mm_setzero_ps();
1750 fjz2 = _mm_setzero_ps();
1751 fjx3 = _mm_setzero_ps();
1752 fjy3 = _mm_setzero_ps();
1753 fjz3 = _mm_setzero_ps();
1754
1755 /**************************
1756 * CALCULATE INTERACTIONS *
1757 **************************/
1758
1759 r00 = _mm_mul_ps(rsq00,rinv00);
1760 r00 = _mm_andnot_ps(dummy_mask,r00);
1761
1762 /* Calculate table index by multiplying r with table scale and truncate to integer */
1763 rt = _mm_mul_ps(r00,vftabscale);
1764 vfitab = _mm_cvttps_epi32(rt);
1765 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1766 vfitab = _mm_slli_epi32(vfitab,3);
1767
1768 /* CUBIC SPLINE TABLE DISPERSION */
1769 Y = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1770 F = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1771 G = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1772 H = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1773 _MM_TRANSPOSE4_PS(Y,F,G,H);
1774 Heps = _mm_mul_ps(vfeps,H);
1775 Fp = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1776 FF = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1777 fvdw6 = _mm_mul_ps(c6_00,FF);
1778
1779 /* CUBIC SPLINE TABLE REPULSION */
1780 vfitab = _mm_add_epi32(vfitab,ifour);
1781 Y = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1782 F = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1783 G = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1784 H = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1785 _MM_TRANSPOSE4_PS(Y,F,G,H);
1786 Heps = _mm_mul_ps(vfeps,H);
1787 Fp = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1788 FF = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1789 fvdw12 = _mm_mul_ps(c12_00,FF);
1790 fvdw = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
1791
1792 fscal = fvdw;
1793
1794 fscal = _mm_andnot_ps(dummy_mask,fscal);
1795
1796 /* Calculate temporary vectorial force */
1797 tx = _mm_mul_ps(fscal,dx00);
1798 ty = _mm_mul_ps(fscal,dy00);
1799 tz = _mm_mul_ps(fscal,dz00);
1800
1801 /* Update vectorial force */
1802 fix0 = _mm_add_ps(fix0,tx);
1803 fiy0 = _mm_add_ps(fiy0,ty);
1804 fiz0 = _mm_add_ps(fiz0,tz);
1805
1806 fjx0 = _mm_add_ps(fjx0,tx);
1807 fjy0 = _mm_add_ps(fjy0,ty);
1808 fjz0 = _mm_add_ps(fjz0,tz);
1809
1810 /**************************
1811 * CALCULATE INTERACTIONS *
1812 **************************/
1813
1814 /* COULOMB ELECTROSTATICS */
1815 velec = _mm_mul_ps(qq11,rinv11);
1816 felec = _mm_mul_ps(velec,rinvsq11);
1817
1818 fscal = felec;
1819
1820 fscal = _mm_andnot_ps(dummy_mask,fscal);
1821
1822 /* Calculate temporary vectorial force */
1823 tx = _mm_mul_ps(fscal,dx11);
1824 ty = _mm_mul_ps(fscal,dy11);
1825 tz = _mm_mul_ps(fscal,dz11);
1826
1827 /* Update vectorial force */
1828 fix1 = _mm_add_ps(fix1,tx);
1829 fiy1 = _mm_add_ps(fiy1,ty);
1830 fiz1 = _mm_add_ps(fiz1,tz);
1831
1832 fjx1 = _mm_add_ps(fjx1,tx);
1833 fjy1 = _mm_add_ps(fjy1,ty);
1834 fjz1 = _mm_add_ps(fjz1,tz);
1835
1836 /**************************
1837 * CALCULATE INTERACTIONS *
1838 **************************/
1839
1840 /* COULOMB ELECTROSTATICS */
1841 velec = _mm_mul_ps(qq12,rinv12);
1842 felec = _mm_mul_ps(velec,rinvsq12);
1843
1844 fscal = felec;
1845
1846 fscal = _mm_andnot_ps(dummy_mask,fscal);
1847
1848 /* Calculate temporary vectorial force */
1849 tx = _mm_mul_ps(fscal,dx12);
1850 ty = _mm_mul_ps(fscal,dy12);
1851 tz = _mm_mul_ps(fscal,dz12);
1852
1853 /* Update vectorial force */
1854 fix1 = _mm_add_ps(fix1,tx);
1855 fiy1 = _mm_add_ps(fiy1,ty);
1856 fiz1 = _mm_add_ps(fiz1,tz);
1857
1858 fjx2 = _mm_add_ps(fjx2,tx);
1859 fjy2 = _mm_add_ps(fjy2,ty);
1860 fjz2 = _mm_add_ps(fjz2,tz);
1861
1862 /**************************
1863 * CALCULATE INTERACTIONS *
1864 **************************/
1865
1866 /* COULOMB ELECTROSTATICS */
1867 velec = _mm_mul_ps(qq13,rinv13);
1868 felec = _mm_mul_ps(velec,rinvsq13);
1869
1870 fscal = felec;
1871
1872 fscal = _mm_andnot_ps(dummy_mask,fscal);
1873
1874 /* Calculate temporary vectorial force */
1875 tx = _mm_mul_ps(fscal,dx13);
1876 ty = _mm_mul_ps(fscal,dy13);
1877 tz = _mm_mul_ps(fscal,dz13);
1878
1879 /* Update vectorial force */
1880 fix1 = _mm_add_ps(fix1,tx);
1881 fiy1 = _mm_add_ps(fiy1,ty);
1882 fiz1 = _mm_add_ps(fiz1,tz);
1883
1884 fjx3 = _mm_add_ps(fjx3,tx);
1885 fjy3 = _mm_add_ps(fjy3,ty);
1886 fjz3 = _mm_add_ps(fjz3,tz);
1887
1888 /**************************
1889 * CALCULATE INTERACTIONS *
1890 **************************/
1891
1892 /* COULOMB ELECTROSTATICS */
1893 velec = _mm_mul_ps(qq21,rinv21);
1894 felec = _mm_mul_ps(velec,rinvsq21);
1895
1896 fscal = felec;
1897
1898 fscal = _mm_andnot_ps(dummy_mask,fscal);
1899
1900 /* Calculate temporary vectorial force */
1901 tx = _mm_mul_ps(fscal,dx21);
1902 ty = _mm_mul_ps(fscal,dy21);
1903 tz = _mm_mul_ps(fscal,dz21);
1904
1905 /* Update vectorial force */
1906 fix2 = _mm_add_ps(fix2,tx);
1907 fiy2 = _mm_add_ps(fiy2,ty);
1908 fiz2 = _mm_add_ps(fiz2,tz);
1909
1910 fjx1 = _mm_add_ps(fjx1,tx);
1911 fjy1 = _mm_add_ps(fjy1,ty);
1912 fjz1 = _mm_add_ps(fjz1,tz);
1913
1914 /**************************
1915 * CALCULATE INTERACTIONS *
1916 **************************/
1917
1918 /* COULOMB ELECTROSTATICS */
1919 velec = _mm_mul_ps(qq22,rinv22);
1920 felec = _mm_mul_ps(velec,rinvsq22);
1921
1922 fscal = felec;
1923
1924 fscal = _mm_andnot_ps(dummy_mask,fscal);
1925
1926 /* Calculate temporary vectorial force */
1927 tx = _mm_mul_ps(fscal,dx22);
1928 ty = _mm_mul_ps(fscal,dy22);
1929 tz = _mm_mul_ps(fscal,dz22);
1930
1931 /* Update vectorial force */
1932 fix2 = _mm_add_ps(fix2,tx);
1933 fiy2 = _mm_add_ps(fiy2,ty);
1934 fiz2 = _mm_add_ps(fiz2,tz);
1935
1936 fjx2 = _mm_add_ps(fjx2,tx);
1937 fjy2 = _mm_add_ps(fjy2,ty);
1938 fjz2 = _mm_add_ps(fjz2,tz);
1939
1940 /**************************
1941 * CALCULATE INTERACTIONS *
1942 **************************/
1943
1944 /* COULOMB ELECTROSTATICS */
1945 velec = _mm_mul_ps(qq23,rinv23);
1946 felec = _mm_mul_ps(velec,rinvsq23);
1947
1948 fscal = felec;
1949
1950 fscal = _mm_andnot_ps(dummy_mask,fscal);
1951
1952 /* Calculate temporary vectorial force */
1953 tx = _mm_mul_ps(fscal,dx23);
1954 ty = _mm_mul_ps(fscal,dy23);
1955 tz = _mm_mul_ps(fscal,dz23);
1956
1957 /* Update vectorial force */
1958 fix2 = _mm_add_ps(fix2,tx);
1959 fiy2 = _mm_add_ps(fiy2,ty);
1960 fiz2 = _mm_add_ps(fiz2,tz);
1961
1962 fjx3 = _mm_add_ps(fjx3,tx);
1963 fjy3 = _mm_add_ps(fjy3,ty);
1964 fjz3 = _mm_add_ps(fjz3,tz);
1965
1966 /**************************
1967 * CALCULATE INTERACTIONS *
1968 **************************/
1969
1970 /* COULOMB ELECTROSTATICS */
1971 velec = _mm_mul_ps(qq31,rinv31);
1972 felec = _mm_mul_ps(velec,rinvsq31);
1973
1974 fscal = felec;
1975
1976 fscal = _mm_andnot_ps(dummy_mask,fscal);
1977
1978 /* Calculate temporary vectorial force */
1979 tx = _mm_mul_ps(fscal,dx31);
1980 ty = _mm_mul_ps(fscal,dy31);
1981 tz = _mm_mul_ps(fscal,dz31);
1982
1983 /* Update vectorial force */
1984 fix3 = _mm_add_ps(fix3,tx);
1985 fiy3 = _mm_add_ps(fiy3,ty);
1986 fiz3 = _mm_add_ps(fiz3,tz);
1987
1988 fjx1 = _mm_add_ps(fjx1,tx);
1989 fjy1 = _mm_add_ps(fjy1,ty);
1990 fjz1 = _mm_add_ps(fjz1,tz);
1991
1992 /**************************
1993 * CALCULATE INTERACTIONS *
1994 **************************/
1995
1996 /* COULOMB ELECTROSTATICS */
1997 velec = _mm_mul_ps(qq32,rinv32);
1998 felec = _mm_mul_ps(velec,rinvsq32);
1999
2000 fscal = felec;
2001
2002 fscal = _mm_andnot_ps(dummy_mask,fscal);
2003
2004 /* Calculate temporary vectorial force */
2005 tx = _mm_mul_ps(fscal,dx32);
2006 ty = _mm_mul_ps(fscal,dy32);
2007 tz = _mm_mul_ps(fscal,dz32);
2008
2009 /* Update vectorial force */
2010 fix3 = _mm_add_ps(fix3,tx);
2011 fiy3 = _mm_add_ps(fiy3,ty);
2012 fiz3 = _mm_add_ps(fiz3,tz);
2013
2014 fjx2 = _mm_add_ps(fjx2,tx);
2015 fjy2 = _mm_add_ps(fjy2,ty);
2016 fjz2 = _mm_add_ps(fjz2,tz);
2017
2018 /**************************
2019 * CALCULATE INTERACTIONS *
2020 **************************/
2021
2022 /* COULOMB ELECTROSTATICS */
2023 velec = _mm_mul_ps(qq33,rinv33);
2024 felec = _mm_mul_ps(velec,rinvsq33);
2025
2026 fscal = felec;
2027
2028 fscal = _mm_andnot_ps(dummy_mask,fscal);
2029
2030 /* Calculate temporary vectorial force */
2031 tx = _mm_mul_ps(fscal,dx33);
2032 ty = _mm_mul_ps(fscal,dy33);
2033 tz = _mm_mul_ps(fscal,dz33);
2034
2035 /* Update vectorial force */
2036 fix3 = _mm_add_ps(fix3,tx);
2037 fiy3 = _mm_add_ps(fiy3,ty);
2038 fiz3 = _mm_add_ps(fiz3,tz);
2039
2040 fjx3 = _mm_add_ps(fjx3,tx);
2041 fjy3 = _mm_add_ps(fjy3,ty);
2042 fjz3 = _mm_add_ps(fjz3,tz);
2043
2044 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
2045 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
2046 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
2047 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
2048
2049 gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
2050 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
2051 fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2052
2053 /* Inner loop uses 295 flops */
2054 }
2055
2056 /* End of innermost loop */
2057
2058 gmx_mm_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2059 f+i_coord_offset,fshift+i_shift_offset);
2060
2061 /* Increment number of inner iterations */
2062 inneriter += j_index_end - j_index_start;
2063
2064 /* Outer loop uses 24 flops */
2065 }
2066
2067 /* Increment number of outer iterations */
2068 outeriter += nri;
2069
2070 /* Update outer/inner flops */
2071
2072 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*295);
2073 }
The ultimate molecular dynamics simulation package