?  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_single / nb_kernel_ElecEw_VdwCSTab_GeomW3W3_sse2_single.c
blob9481430dfc7ef9d141a48d86056ffa0d9f74d0f5
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_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_sse2_single.h"
49 #include "kernelutil_x86_sse2_single.h"
50
51 /*
52 * Gromacs nonbonded kernel: nb_kernel_ElecEw_VdwCSTab_GeomW3W3_VF_sse2_single
53 * Electrostatics interaction: Ewald
54 * VdW interaction: CubicSplineTable
55 * Geometry: Water3-Water3
56 * Calculate force/pot: PotentialAndForce
57 */
58 void
59 nb_kernel_ElecEw_VdwCSTab_GeomW3W3_VF_sse2_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 vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
91 __m128 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
92 int vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
93 __m128 jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
94 int vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
95 __m128 jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
96 __m128 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
97 __m128 dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
98 __m128 dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
99 __m128 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
100 __m128 dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
101 __m128 dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
102 __m128 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
103 __m128 dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
104 __m128 dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
105 __m128 velec,felec,velecsum,facel,crf,krf,krf2;
106 real *charge;
107 int nvdwtype;
108 __m128 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
109 int *vdwtype;
110 real *vdwparam;
111 __m128 one_sixth = _mm_set1_ps(1.0/6.0);
112 __m128 one_twelfth = _mm_set1_ps(1.0/12.0);
113 __m128i vfitab;
114 __m128i ifour = _mm_set1_epi32(4);
115 __m128 rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
116 real *vftab;
117 __m128i ewitab;
118 __m128 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
119 real *ewtab;
120 __m128 dummy_mask,cutoff_mask;
121 __m128 signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
122 __m128 one = _mm_set1_ps(1.0);
123 __m128 two = _mm_set1_ps(2.0);
124 x = xx[0];
125 f = ff[0];
126
127 nri = nlist->nri;
128 iinr = nlist->iinr;
129 jindex = nlist->jindex;
130 jjnr = nlist->jjnr;
131 shiftidx = nlist->shift;
132 gid = nlist->gid;
133 shiftvec = fr->shift_vec[0];
134 fshift = fr->fshift[0];
135 facel = _mm_set1_ps(fr->epsfac);
136 charge = mdatoms->chargeA;
137 nvdwtype = fr->ntype;
138 vdwparam = fr->nbfp;
139 vdwtype = mdatoms->typeA;
140
141 vftab = kernel_data->table_vdw->data;
142 vftabscale = _mm_set1_ps(kernel_data->table_vdw->scale);
143
144 sh_ewald = _mm_set1_ps(fr->ic->sh_ewald);
145 ewtab = fr->ic->tabq_coul_FDV0;
146 ewtabscale = _mm_set1_ps(fr->ic->tabq_scale);
147 ewtabhalfspace = _mm_set1_ps(0.5/fr->ic->tabq_scale);
148
149 /* Setup water-specific parameters */
150 inr = nlist->iinr[0];
151 iq0 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+0]));
152 iq1 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1]));
153 iq2 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2]));
154 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
155
156 jq0 = _mm_set1_ps(charge[inr+0]);
157 jq1 = _mm_set1_ps(charge[inr+1]);
158 jq2 = _mm_set1_ps(charge[inr+2]);
159 vdwjidx0A = 2*vdwtype[inr+0];
160 qq00 = _mm_mul_ps(iq0,jq0);
161 c6_00 = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A]);
162 c12_00 = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A+1]);
163 qq01 = _mm_mul_ps(iq0,jq1);
164 qq02 = _mm_mul_ps(iq0,jq2);
165 qq10 = _mm_mul_ps(iq1,jq0);
166 qq11 = _mm_mul_ps(iq1,jq1);
167 qq12 = _mm_mul_ps(iq1,jq2);
168 qq20 = _mm_mul_ps(iq2,jq0);
169 qq21 = _mm_mul_ps(iq2,jq1);
170 qq22 = _mm_mul_ps(iq2,jq2);
171
172 /* Avoid stupid compiler warnings */
173 jnrA = jnrB = jnrC = jnrD = 0;
174 j_coord_offsetA = 0;
175 j_coord_offsetB = 0;
176 j_coord_offsetC = 0;
177 j_coord_offsetD = 0;
178
179 outeriter = 0;
180 inneriter = 0;
181
182 for(iidx=0;iidx<4*DIM;iidx++)
183 {
184 scratch[iidx] = 0.0;
185 }
186
187 /* Start outer loop over neighborlists */
188 for(iidx=0; iidx<nri; iidx++)
189 {
190 /* Load shift vector for this list */
191 i_shift_offset = DIM*shiftidx[iidx];
192
193 /* Load limits for loop over neighbors */
194 j_index_start = jindex[iidx];
195 j_index_end = jindex[iidx+1];
196
197 /* Get outer coordinate index */
198 inr = iinr[iidx];
199 i_coord_offset = DIM*inr;
200
201 /* Load i particle coords and add shift vector */
202 gmx_mm_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
203 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
204
205 fix0 = _mm_setzero_ps();
206 fiy0 = _mm_setzero_ps();
207 fiz0 = _mm_setzero_ps();
208 fix1 = _mm_setzero_ps();
209 fiy1 = _mm_setzero_ps();
210 fiz1 = _mm_setzero_ps();
211 fix2 = _mm_setzero_ps();
212 fiy2 = _mm_setzero_ps();
213 fiz2 = _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_3rvec_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,&jy2,&jz2);
237
238 /* Calculate displacement vector */
239 dx00 = _mm_sub_ps(ix0,jx0);
240 dy00 = _mm_sub_ps(iy0,jy0);
241 dz00 = _mm_sub_ps(iz0,jz0);
242 dx01 = _mm_sub_ps(ix0,jx1);
243 dy01 = _mm_sub_ps(iy0,jy1);
244 dz01 = _mm_sub_ps(iz0,jz1);
245 dx02 = _mm_sub_ps(ix0,jx2);
246 dy02 = _mm_sub_ps(iy0,jy2);
247 dz02 = _mm_sub_ps(iz0,jz2);
248 dx10 = _mm_sub_ps(ix1,jx0);
249 dy10 = _mm_sub_ps(iy1,jy0);
250 dz10 = _mm_sub_ps(iz1,jz0);
251 dx11 = _mm_sub_ps(ix1,jx1);
252 dy11 = _mm_sub_ps(iy1,jy1);
253 dz11 = _mm_sub_ps(iz1,jz1);
254 dx12 = _mm_sub_ps(ix1,jx2);
255 dy12 = _mm_sub_ps(iy1,jy2);
256 dz12 = _mm_sub_ps(iz1,jz2);
257 dx20 = _mm_sub_ps(ix2,jx0);
258 dy20 = _mm_sub_ps(iy2,jy0);
259 dz20 = _mm_sub_ps(iz2,jz0);
260 dx21 = _mm_sub_ps(ix2,jx1);
261 dy21 = _mm_sub_ps(iy2,jy1);
262 dz21 = _mm_sub_ps(iz2,jz1);
263 dx22 = _mm_sub_ps(ix2,jx2);
264 dy22 = _mm_sub_ps(iy2,jy2);
265 dz22 = _mm_sub_ps(iz2,jz2);
266
267 /* Calculate squared distance and things based on it */
268 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
269 rsq01 = gmx_mm_calc_rsq_ps(dx01,dy01,dz01);
270 rsq02 = gmx_mm_calc_rsq_ps(dx02,dy02,dz02);
271 rsq10 = gmx_mm_calc_rsq_ps(dx10,dy10,dz10);
272 rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
273 rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
274 rsq20 = gmx_mm_calc_rsq_ps(dx20,dy20,dz20);
275 rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
276 rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
277
278 rinv00 = gmx_mm_invsqrt_ps(rsq00);
279 rinv01 = gmx_mm_invsqrt_ps(rsq01);
280 rinv02 = gmx_mm_invsqrt_ps(rsq02);
281 rinv10 = gmx_mm_invsqrt_ps(rsq10);
282 rinv11 = gmx_mm_invsqrt_ps(rsq11);
283 rinv12 = gmx_mm_invsqrt_ps(rsq12);
284 rinv20 = gmx_mm_invsqrt_ps(rsq20);
285 rinv21 = gmx_mm_invsqrt_ps(rsq21);
286 rinv22 = gmx_mm_invsqrt_ps(rsq22);
287
288 rinvsq00 = _mm_mul_ps(rinv00,rinv00);
289 rinvsq01 = _mm_mul_ps(rinv01,rinv01);
290 rinvsq02 = _mm_mul_ps(rinv02,rinv02);
291 rinvsq10 = _mm_mul_ps(rinv10,rinv10);
292 rinvsq11 = _mm_mul_ps(rinv11,rinv11);
293 rinvsq12 = _mm_mul_ps(rinv12,rinv12);
294 rinvsq20 = _mm_mul_ps(rinv20,rinv20);
295 rinvsq21 = _mm_mul_ps(rinv21,rinv21);
296 rinvsq22 = _mm_mul_ps(rinv22,rinv22);
297
298 fjx0 = _mm_setzero_ps();
299 fjy0 = _mm_setzero_ps();
300 fjz0 = _mm_setzero_ps();
301 fjx1 = _mm_setzero_ps();
302 fjy1 = _mm_setzero_ps();
303 fjz1 = _mm_setzero_ps();
304 fjx2 = _mm_setzero_ps();
305 fjy2 = _mm_setzero_ps();
306 fjz2 = _mm_setzero_ps();
307
308 /**************************
309 * CALCULATE INTERACTIONS *
310 **************************/
311
312 r00 = _mm_mul_ps(rsq00,rinv00);
313
314 /* Calculate table index by multiplying r with table scale and truncate to integer */
315 rt = _mm_mul_ps(r00,vftabscale);
316 vfitab = _mm_cvttps_epi32(rt);
317 vfeps = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
318 vfitab = _mm_slli_epi32(vfitab,3);
319
320 /* EWALD ELECTROSTATICS */
321
322 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
323 ewrt = _mm_mul_ps(r00,ewtabscale);
324 ewitab = _mm_cvttps_epi32(ewrt);
325 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
326 ewitab = _mm_slli_epi32(ewitab,2);
327 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
328 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
329 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
330 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
331 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
332 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
333 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
334 velec = _mm_mul_ps(qq00,_mm_sub_ps(rinv00,velec));
335 felec = _mm_mul_ps(_mm_mul_ps(qq00,rinv00),_mm_sub_ps(rinvsq00,felec));
336
337 /* CUBIC SPLINE TABLE DISPERSION */
338 Y = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
339 F = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
340 G = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
341 H = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
342 _MM_TRANSPOSE4_PS(Y,F,G,H);
343 Heps = _mm_mul_ps(vfeps,H);
344 Fp = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
345 VV = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
346 vvdw6 = _mm_mul_ps(c6_00,VV);
347 FF = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
348 fvdw6 = _mm_mul_ps(c6_00,FF);
349
350 /* CUBIC SPLINE TABLE REPULSION */
351 vfitab = _mm_add_epi32(vfitab,ifour);
352 Y = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
353 F = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
354 G = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
355 H = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
356 _MM_TRANSPOSE4_PS(Y,F,G,H);
357 Heps = _mm_mul_ps(vfeps,H);
358 Fp = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
359 VV = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
360 vvdw12 = _mm_mul_ps(c12_00,VV);
361 FF = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
362 fvdw12 = _mm_mul_ps(c12_00,FF);
363 vvdw = _mm_add_ps(vvdw12,vvdw6);
364 fvdw = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
365
366 /* Update potential sum for this i atom from the interaction with this j atom. */
367 velecsum = _mm_add_ps(velecsum,velec);
368 vvdwsum = _mm_add_ps(vvdwsum,vvdw);
369
370 fscal = _mm_add_ps(felec,fvdw);
371
372 /* Calculate temporary vectorial force */
373 tx = _mm_mul_ps(fscal,dx00);
374 ty = _mm_mul_ps(fscal,dy00);
375 tz = _mm_mul_ps(fscal,dz00);
376
377 /* Update vectorial force */
378 fix0 = _mm_add_ps(fix0,tx);
379 fiy0 = _mm_add_ps(fiy0,ty);
380 fiz0 = _mm_add_ps(fiz0,tz);
381
382 fjx0 = _mm_add_ps(fjx0,tx);
383 fjy0 = _mm_add_ps(fjy0,ty);
384 fjz0 = _mm_add_ps(fjz0,tz);
385
386 /**************************
387 * CALCULATE INTERACTIONS *
388 **************************/
389
390 r01 = _mm_mul_ps(rsq01,rinv01);
391
392 /* EWALD ELECTROSTATICS */
393
394 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
395 ewrt = _mm_mul_ps(r01,ewtabscale);
396 ewitab = _mm_cvttps_epi32(ewrt);
397 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
398 ewitab = _mm_slli_epi32(ewitab,2);
399 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
400 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
401 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
402 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
403 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
404 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
405 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
406 velec = _mm_mul_ps(qq01,_mm_sub_ps(rinv01,velec));
407 felec = _mm_mul_ps(_mm_mul_ps(qq01,rinv01),_mm_sub_ps(rinvsq01,felec));
408
409 /* Update potential sum for this i atom from the interaction with this j atom. */
410 velecsum = _mm_add_ps(velecsum,velec);
411
412 fscal = felec;
413
414 /* Calculate temporary vectorial force */
415 tx = _mm_mul_ps(fscal,dx01);
416 ty = _mm_mul_ps(fscal,dy01);
417 tz = _mm_mul_ps(fscal,dz01);
418
419 /* Update vectorial force */
420 fix0 = _mm_add_ps(fix0,tx);
421 fiy0 = _mm_add_ps(fiy0,ty);
422 fiz0 = _mm_add_ps(fiz0,tz);
423
424 fjx1 = _mm_add_ps(fjx1,tx);
425 fjy1 = _mm_add_ps(fjy1,ty);
426 fjz1 = _mm_add_ps(fjz1,tz);
427
428 /**************************
429 * CALCULATE INTERACTIONS *
430 **************************/
431
432 r02 = _mm_mul_ps(rsq02,rinv02);
433
434 /* EWALD ELECTROSTATICS */
435
436 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
437 ewrt = _mm_mul_ps(r02,ewtabscale);
438 ewitab = _mm_cvttps_epi32(ewrt);
439 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
440 ewitab = _mm_slli_epi32(ewitab,2);
441 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
442 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
443 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
444 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
445 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
446 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
447 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
448 velec = _mm_mul_ps(qq02,_mm_sub_ps(rinv02,velec));
449 felec = _mm_mul_ps(_mm_mul_ps(qq02,rinv02),_mm_sub_ps(rinvsq02,felec));
450
451 /* Update potential sum for this i atom from the interaction with this j atom. */
452 velecsum = _mm_add_ps(velecsum,velec);
453
454 fscal = felec;
455
456 /* Calculate temporary vectorial force */
457 tx = _mm_mul_ps(fscal,dx02);
458 ty = _mm_mul_ps(fscal,dy02);
459 tz = _mm_mul_ps(fscal,dz02);
460
461 /* Update vectorial force */
462 fix0 = _mm_add_ps(fix0,tx);
463 fiy0 = _mm_add_ps(fiy0,ty);
464 fiz0 = _mm_add_ps(fiz0,tz);
465
466 fjx2 = _mm_add_ps(fjx2,tx);
467 fjy2 = _mm_add_ps(fjy2,ty);
468 fjz2 = _mm_add_ps(fjz2,tz);
469
470 /**************************
471 * CALCULATE INTERACTIONS *
472 **************************/
473
474 r10 = _mm_mul_ps(rsq10,rinv10);
475
476 /* EWALD ELECTROSTATICS */
477
478 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
479 ewrt = _mm_mul_ps(r10,ewtabscale);
480 ewitab = _mm_cvttps_epi32(ewrt);
481 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
482 ewitab = _mm_slli_epi32(ewitab,2);
483 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
484 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
485 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
486 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
487 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
488 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
489 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
490 velec = _mm_mul_ps(qq10,_mm_sub_ps(rinv10,velec));
491 felec = _mm_mul_ps(_mm_mul_ps(qq10,rinv10),_mm_sub_ps(rinvsq10,felec));
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,dx10);
500 ty = _mm_mul_ps(fscal,dy10);
501 tz = _mm_mul_ps(fscal,dz10);
502
503 /* Update vectorial force */
504 fix1 = _mm_add_ps(fix1,tx);
505 fiy1 = _mm_add_ps(fiy1,ty);
506 fiz1 = _mm_add_ps(fiz1,tz);
507
508 fjx0 = _mm_add_ps(fjx0,tx);
509 fjy0 = _mm_add_ps(fjy0,ty);
510 fjz0 = _mm_add_ps(fjz0,tz);
511
512 /**************************
513 * CALCULATE INTERACTIONS *
514 **************************/
515
516 r11 = _mm_mul_ps(rsq11,rinv11);
517
518 /* EWALD ELECTROSTATICS */
519
520 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
521 ewrt = _mm_mul_ps(r11,ewtabscale);
522 ewitab = _mm_cvttps_epi32(ewrt);
523 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
524 ewitab = _mm_slli_epi32(ewitab,2);
525 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
526 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
527 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
528 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
529 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
530 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
531 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
532 velec = _mm_mul_ps(qq11,_mm_sub_ps(rinv11,velec));
533 felec = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec));
534
535 /* Update potential sum for this i atom from the interaction with this j atom. */
536 velecsum = _mm_add_ps(velecsum,velec);
537
538 fscal = felec;
539
540 /* Calculate temporary vectorial force */
541 tx = _mm_mul_ps(fscal,dx11);
542 ty = _mm_mul_ps(fscal,dy11);
543 tz = _mm_mul_ps(fscal,dz11);
544
545 /* Update vectorial force */
546 fix1 = _mm_add_ps(fix1,tx);
547 fiy1 = _mm_add_ps(fiy1,ty);
548 fiz1 = _mm_add_ps(fiz1,tz);
549
550 fjx1 = _mm_add_ps(fjx1,tx);
551 fjy1 = _mm_add_ps(fjy1,ty);
552 fjz1 = _mm_add_ps(fjz1,tz);
553
554 /**************************
555 * CALCULATE INTERACTIONS *
556 **************************/
557
558 r12 = _mm_mul_ps(rsq12,rinv12);
559
560 /* EWALD ELECTROSTATICS */
561
562 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
563 ewrt = _mm_mul_ps(r12,ewtabscale);
564 ewitab = _mm_cvttps_epi32(ewrt);
565 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
566 ewitab = _mm_slli_epi32(ewitab,2);
567 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
568 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
569 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
570 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
571 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
572 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
573 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
574 velec = _mm_mul_ps(qq12,_mm_sub_ps(rinv12,velec));
575 felec = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec));
576
577 /* Update potential sum for this i atom from the interaction with this j atom. */
578 velecsum = _mm_add_ps(velecsum,velec);
579
580 fscal = felec;
581
582 /* Calculate temporary vectorial force */
583 tx = _mm_mul_ps(fscal,dx12);
584 ty = _mm_mul_ps(fscal,dy12);
585 tz = _mm_mul_ps(fscal,dz12);
586
587 /* Update vectorial force */
588 fix1 = _mm_add_ps(fix1,tx);
589 fiy1 = _mm_add_ps(fiy1,ty);
590 fiz1 = _mm_add_ps(fiz1,tz);
591
592 fjx2 = _mm_add_ps(fjx2,tx);
593 fjy2 = _mm_add_ps(fjy2,ty);
594 fjz2 = _mm_add_ps(fjz2,tz);
595
596 /**************************
597 * CALCULATE INTERACTIONS *
598 **************************/
599
600 r20 = _mm_mul_ps(rsq20,rinv20);
601
602 /* EWALD ELECTROSTATICS */
603
604 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
605 ewrt = _mm_mul_ps(r20,ewtabscale);
606 ewitab = _mm_cvttps_epi32(ewrt);
607 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
608 ewitab = _mm_slli_epi32(ewitab,2);
609 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
610 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
611 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
612 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
613 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
614 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
615 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
616 velec = _mm_mul_ps(qq20,_mm_sub_ps(rinv20,velec));
617 felec = _mm_mul_ps(_mm_mul_ps(qq20,rinv20),_mm_sub_ps(rinvsq20,felec));
618
619 /* Update potential sum for this i atom from the interaction with this j atom. */
620 velecsum = _mm_add_ps(velecsum,velec);
621
622 fscal = felec;
623
624 /* Calculate temporary vectorial force */
625 tx = _mm_mul_ps(fscal,dx20);
626 ty = _mm_mul_ps(fscal,dy20);
627 tz = _mm_mul_ps(fscal,dz20);
628
629 /* Update vectorial force */
630 fix2 = _mm_add_ps(fix2,tx);
631 fiy2 = _mm_add_ps(fiy2,ty);
632 fiz2 = _mm_add_ps(fiz2,tz);
633
634 fjx0 = _mm_add_ps(fjx0,tx);
635 fjy0 = _mm_add_ps(fjy0,ty);
636 fjz0 = _mm_add_ps(fjz0,tz);
637
638 /**************************
639 * CALCULATE INTERACTIONS *
640 **************************/
641
642 r21 = _mm_mul_ps(rsq21,rinv21);
643
644 /* EWALD ELECTROSTATICS */
645
646 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
647 ewrt = _mm_mul_ps(r21,ewtabscale);
648 ewitab = _mm_cvttps_epi32(ewrt);
649 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
650 ewitab = _mm_slli_epi32(ewitab,2);
651 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
652 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
653 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
654 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
655 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
656 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
657 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
658 velec = _mm_mul_ps(qq21,_mm_sub_ps(rinv21,velec));
659 felec = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec));
660
661 /* Update potential sum for this i atom from the interaction with this j atom. */
662 velecsum = _mm_add_ps(velecsum,velec);
663
664 fscal = felec;
665
666 /* Calculate temporary vectorial force */
667 tx = _mm_mul_ps(fscal,dx21);
668 ty = _mm_mul_ps(fscal,dy21);
669 tz = _mm_mul_ps(fscal,dz21);
670
671 /* Update vectorial force */
672 fix2 = _mm_add_ps(fix2,tx);
673 fiy2 = _mm_add_ps(fiy2,ty);
674 fiz2 = _mm_add_ps(fiz2,tz);
675
676 fjx1 = _mm_add_ps(fjx1,tx);
677 fjy1 = _mm_add_ps(fjy1,ty);
678 fjz1 = _mm_add_ps(fjz1,tz);
679
680 /**************************
681 * CALCULATE INTERACTIONS *
682 **************************/
683
684 r22 = _mm_mul_ps(rsq22,rinv22);
685
686 /* EWALD ELECTROSTATICS */
687
688 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
689 ewrt = _mm_mul_ps(r22,ewtabscale);
690 ewitab = _mm_cvttps_epi32(ewrt);
691 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
692 ewitab = _mm_slli_epi32(ewitab,2);
693 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
694 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
695 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
696 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
697 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
698 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
699 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
700 velec = _mm_mul_ps(qq22,_mm_sub_ps(rinv22,velec));
701 felec = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec));
702
703 /* Update potential sum for this i atom from the interaction with this j atom. */
704 velecsum = _mm_add_ps(velecsum,velec);
705
706 fscal = felec;
707
708 /* Calculate temporary vectorial force */
709 tx = _mm_mul_ps(fscal,dx22);
710 ty = _mm_mul_ps(fscal,dy22);
711 tz = _mm_mul_ps(fscal,dz22);
712
713 /* Update vectorial force */
714 fix2 = _mm_add_ps(fix2,tx);
715 fiy2 = _mm_add_ps(fiy2,ty);
716 fiz2 = _mm_add_ps(fiz2,tz);
717
718 fjx2 = _mm_add_ps(fjx2,tx);
719 fjy2 = _mm_add_ps(fjy2,ty);
720 fjz2 = _mm_add_ps(fjz2,tz);
721
722 fjptrA = f+j_coord_offsetA;
723 fjptrB = f+j_coord_offsetB;
724 fjptrC = f+j_coord_offsetC;
725 fjptrD = f+j_coord_offsetD;
726
727 gmx_mm_decrement_3rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
728 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
729
730 /* Inner loop uses 403 flops */
731 }
732
733 if(jidx<j_index_end)
734 {
735
736 /* Get j neighbor index, and coordinate index */
737 jnrlistA = jjnr[jidx];
738 jnrlistB = jjnr[jidx+1];
739 jnrlistC = jjnr[jidx+2];
740 jnrlistD = jjnr[jidx+3];
741 /* Sign of each element will be negative for non-real atoms.
742 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
743 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
744 */
745 dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
746 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
747 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
748 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
749 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
750 j_coord_offsetA = DIM*jnrA;
751 j_coord_offsetB = DIM*jnrB;
752 j_coord_offsetC = DIM*jnrC;
753 j_coord_offsetD = DIM*jnrD;
754
755 /* load j atom coordinates */
756 gmx_mm_load_3rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
757 x+j_coord_offsetC,x+j_coord_offsetD,
758 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
759
760 /* Calculate displacement vector */
761 dx00 = _mm_sub_ps(ix0,jx0);
762 dy00 = _mm_sub_ps(iy0,jy0);
763 dz00 = _mm_sub_ps(iz0,jz0);
764 dx01 = _mm_sub_ps(ix0,jx1);
765 dy01 = _mm_sub_ps(iy0,jy1);
766 dz01 = _mm_sub_ps(iz0,jz1);
767 dx02 = _mm_sub_ps(ix0,jx2);
768 dy02 = _mm_sub_ps(iy0,jy2);
769 dz02 = _mm_sub_ps(iz0,jz2);
770 dx10 = _mm_sub_ps(ix1,jx0);
771 dy10 = _mm_sub_ps(iy1,jy0);
772 dz10 = _mm_sub_ps(iz1,jz0);
773 dx11 = _mm_sub_ps(ix1,jx1);
774 dy11 = _mm_sub_ps(iy1,jy1);
775 dz11 = _mm_sub_ps(iz1,jz1);
776 dx12 = _mm_sub_ps(ix1,jx2);
777 dy12 = _mm_sub_ps(iy1,jy2);
778 dz12 = _mm_sub_ps(iz1,jz2);
779 dx20 = _mm_sub_ps(ix2,jx0);
780 dy20 = _mm_sub_ps(iy2,jy0);
781 dz20 = _mm_sub_ps(iz2,jz0);
782 dx21 = _mm_sub_ps(ix2,jx1);
783 dy21 = _mm_sub_ps(iy2,jy1);
784 dz21 = _mm_sub_ps(iz2,jz1);
785 dx22 = _mm_sub_ps(ix2,jx2);
786 dy22 = _mm_sub_ps(iy2,jy2);
787 dz22 = _mm_sub_ps(iz2,jz2);
788
789 /* Calculate squared distance and things based on it */
790 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
791 rsq01 = gmx_mm_calc_rsq_ps(dx01,dy01,dz01);
792 rsq02 = gmx_mm_calc_rsq_ps(dx02,dy02,dz02);
793 rsq10 = gmx_mm_calc_rsq_ps(dx10,dy10,dz10);
794 rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
795 rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
796 rsq20 = gmx_mm_calc_rsq_ps(dx20,dy20,dz20);
797 rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
798 rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
799
800 rinv00 = gmx_mm_invsqrt_ps(rsq00);
801 rinv01 = gmx_mm_invsqrt_ps(rsq01);
802 rinv02 = gmx_mm_invsqrt_ps(rsq02);
803 rinv10 = gmx_mm_invsqrt_ps(rsq10);
804 rinv11 = gmx_mm_invsqrt_ps(rsq11);
805 rinv12 = gmx_mm_invsqrt_ps(rsq12);
806 rinv20 = gmx_mm_invsqrt_ps(rsq20);
807 rinv21 = gmx_mm_invsqrt_ps(rsq21);
808 rinv22 = gmx_mm_invsqrt_ps(rsq22);
809
810 rinvsq00 = _mm_mul_ps(rinv00,rinv00);
811 rinvsq01 = _mm_mul_ps(rinv01,rinv01);
812 rinvsq02 = _mm_mul_ps(rinv02,rinv02);
813 rinvsq10 = _mm_mul_ps(rinv10,rinv10);
814 rinvsq11 = _mm_mul_ps(rinv11,rinv11);
815 rinvsq12 = _mm_mul_ps(rinv12,rinv12);
816 rinvsq20 = _mm_mul_ps(rinv20,rinv20);
817 rinvsq21 = _mm_mul_ps(rinv21,rinv21);
818 rinvsq22 = _mm_mul_ps(rinv22,rinv22);
819
820 fjx0 = _mm_setzero_ps();
821 fjy0 = _mm_setzero_ps();
822 fjz0 = _mm_setzero_ps();
823 fjx1 = _mm_setzero_ps();
824 fjy1 = _mm_setzero_ps();
825 fjz1 = _mm_setzero_ps();
826 fjx2 = _mm_setzero_ps();
827 fjy2 = _mm_setzero_ps();
828 fjz2 = _mm_setzero_ps();
829
830 /**************************
831 * CALCULATE INTERACTIONS *
832 **************************/
833
834 r00 = _mm_mul_ps(rsq00,rinv00);
835 r00 = _mm_andnot_ps(dummy_mask,r00);
836
837 /* Calculate table index by multiplying r with table scale and truncate to integer */
838 rt = _mm_mul_ps(r00,vftabscale);
839 vfitab = _mm_cvttps_epi32(rt);
840 vfeps = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
841 vfitab = _mm_slli_epi32(vfitab,3);
842
843 /* EWALD ELECTROSTATICS */
844
845 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
846 ewrt = _mm_mul_ps(r00,ewtabscale);
847 ewitab = _mm_cvttps_epi32(ewrt);
848 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
849 ewitab = _mm_slli_epi32(ewitab,2);
850 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
851 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
852 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
853 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
854 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
855 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
856 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
857 velec = _mm_mul_ps(qq00,_mm_sub_ps(rinv00,velec));
858 felec = _mm_mul_ps(_mm_mul_ps(qq00,rinv00),_mm_sub_ps(rinvsq00,felec));
859
860 /* CUBIC SPLINE TABLE DISPERSION */
861 Y = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
862 F = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
863 G = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
864 H = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
865 _MM_TRANSPOSE4_PS(Y,F,G,H);
866 Heps = _mm_mul_ps(vfeps,H);
867 Fp = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
868 VV = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
869 vvdw6 = _mm_mul_ps(c6_00,VV);
870 FF = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
871 fvdw6 = _mm_mul_ps(c6_00,FF);
872
873 /* CUBIC SPLINE TABLE REPULSION */
874 vfitab = _mm_add_epi32(vfitab,ifour);
875 Y = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
876 F = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
877 G = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
878 H = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
879 _MM_TRANSPOSE4_PS(Y,F,G,H);
880 Heps = _mm_mul_ps(vfeps,H);
881 Fp = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
882 VV = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
883 vvdw12 = _mm_mul_ps(c12_00,VV);
884 FF = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
885 fvdw12 = _mm_mul_ps(c12_00,FF);
886 vvdw = _mm_add_ps(vvdw12,vvdw6);
887 fvdw = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
888
889 /* Update potential sum for this i atom from the interaction with this j atom. */
890 velec = _mm_andnot_ps(dummy_mask,velec);
891 velecsum = _mm_add_ps(velecsum,velec);
892 vvdw = _mm_andnot_ps(dummy_mask,vvdw);
893 vvdwsum = _mm_add_ps(vvdwsum,vvdw);
894
895 fscal = _mm_add_ps(felec,fvdw);
896
897 fscal = _mm_andnot_ps(dummy_mask,fscal);
898
899 /* Calculate temporary vectorial force */
900 tx = _mm_mul_ps(fscal,dx00);
901 ty = _mm_mul_ps(fscal,dy00);
902 tz = _mm_mul_ps(fscal,dz00);
903
904 /* Update vectorial force */
905 fix0 = _mm_add_ps(fix0,tx);
906 fiy0 = _mm_add_ps(fiy0,ty);
907 fiz0 = _mm_add_ps(fiz0,tz);
908
909 fjx0 = _mm_add_ps(fjx0,tx);
910 fjy0 = _mm_add_ps(fjy0,ty);
911 fjz0 = _mm_add_ps(fjz0,tz);
912
913 /**************************
914 * CALCULATE INTERACTIONS *
915 **************************/
916
917 r01 = _mm_mul_ps(rsq01,rinv01);
918 r01 = _mm_andnot_ps(dummy_mask,r01);
919
920 /* EWALD ELECTROSTATICS */
921
922 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
923 ewrt = _mm_mul_ps(r01,ewtabscale);
924 ewitab = _mm_cvttps_epi32(ewrt);
925 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
926 ewitab = _mm_slli_epi32(ewitab,2);
927 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
928 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
929 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
930 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
931 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
932 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
933 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
934 velec = _mm_mul_ps(qq01,_mm_sub_ps(rinv01,velec));
935 felec = _mm_mul_ps(_mm_mul_ps(qq01,rinv01),_mm_sub_ps(rinvsq01,felec));
936
937 /* Update potential sum for this i atom from the interaction with this j atom. */
938 velec = _mm_andnot_ps(dummy_mask,velec);
939 velecsum = _mm_add_ps(velecsum,velec);
940
941 fscal = felec;
942
943 fscal = _mm_andnot_ps(dummy_mask,fscal);
944
945 /* Calculate temporary vectorial force */
946 tx = _mm_mul_ps(fscal,dx01);
947 ty = _mm_mul_ps(fscal,dy01);
948 tz = _mm_mul_ps(fscal,dz01);
949
950 /* Update vectorial force */
951 fix0 = _mm_add_ps(fix0,tx);
952 fiy0 = _mm_add_ps(fiy0,ty);
953 fiz0 = _mm_add_ps(fiz0,tz);
954
955 fjx1 = _mm_add_ps(fjx1,tx);
956 fjy1 = _mm_add_ps(fjy1,ty);
957 fjz1 = _mm_add_ps(fjz1,tz);
958
959 /**************************
960 * CALCULATE INTERACTIONS *
961 **************************/
962
963 r02 = _mm_mul_ps(rsq02,rinv02);
964 r02 = _mm_andnot_ps(dummy_mask,r02);
965
966 /* EWALD ELECTROSTATICS */
967
968 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
969 ewrt = _mm_mul_ps(r02,ewtabscale);
970 ewitab = _mm_cvttps_epi32(ewrt);
971 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
972 ewitab = _mm_slli_epi32(ewitab,2);
973 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
974 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
975 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
976 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
977 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
978 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
979 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
980 velec = _mm_mul_ps(qq02,_mm_sub_ps(rinv02,velec));
981 felec = _mm_mul_ps(_mm_mul_ps(qq02,rinv02),_mm_sub_ps(rinvsq02,felec));
982
983 /* Update potential sum for this i atom from the interaction with this j atom. */
984 velec = _mm_andnot_ps(dummy_mask,velec);
985 velecsum = _mm_add_ps(velecsum,velec);
986
987 fscal = felec;
988
989 fscal = _mm_andnot_ps(dummy_mask,fscal);
990
991 /* Calculate temporary vectorial force */
992 tx = _mm_mul_ps(fscal,dx02);
993 ty = _mm_mul_ps(fscal,dy02);
994 tz = _mm_mul_ps(fscal,dz02);
995
996 /* Update vectorial force */
997 fix0 = _mm_add_ps(fix0,tx);
998 fiy0 = _mm_add_ps(fiy0,ty);
999 fiz0 = _mm_add_ps(fiz0,tz);
1000
1001 fjx2 = _mm_add_ps(fjx2,tx);
1002 fjy2 = _mm_add_ps(fjy2,ty);
1003 fjz2 = _mm_add_ps(fjz2,tz);
1004
1005 /**************************
1006 * CALCULATE INTERACTIONS *
1007 **************************/
1008
1009 r10 = _mm_mul_ps(rsq10,rinv10);
1010 r10 = _mm_andnot_ps(dummy_mask,r10);
1011
1012 /* EWALD ELECTROSTATICS */
1013
1014 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1015 ewrt = _mm_mul_ps(r10,ewtabscale);
1016 ewitab = _mm_cvttps_epi32(ewrt);
1017 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1018 ewitab = _mm_slli_epi32(ewitab,2);
1019 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1020 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1021 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1022 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1023 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1024 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1025 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1026 velec = _mm_mul_ps(qq10,_mm_sub_ps(rinv10,velec));
1027 felec = _mm_mul_ps(_mm_mul_ps(qq10,rinv10),_mm_sub_ps(rinvsq10,felec));
1028
1029 /* Update potential sum for this i atom from the interaction with this j atom. */
1030 velec = _mm_andnot_ps(dummy_mask,velec);
1031 velecsum = _mm_add_ps(velecsum,velec);
1032
1033 fscal = felec;
1034
1035 fscal = _mm_andnot_ps(dummy_mask,fscal);
1036
1037 /* Calculate temporary vectorial force */
1038 tx = _mm_mul_ps(fscal,dx10);
1039 ty = _mm_mul_ps(fscal,dy10);
1040 tz = _mm_mul_ps(fscal,dz10);
1041
1042 /* Update vectorial force */
1043 fix1 = _mm_add_ps(fix1,tx);
1044 fiy1 = _mm_add_ps(fiy1,ty);
1045 fiz1 = _mm_add_ps(fiz1,tz);
1046
1047 fjx0 = _mm_add_ps(fjx0,tx);
1048 fjy0 = _mm_add_ps(fjy0,ty);
1049 fjz0 = _mm_add_ps(fjz0,tz);
1050
1051 /**************************
1052 * CALCULATE INTERACTIONS *
1053 **************************/
1054
1055 r11 = _mm_mul_ps(rsq11,rinv11);
1056 r11 = _mm_andnot_ps(dummy_mask,r11);
1057
1058 /* EWALD ELECTROSTATICS */
1059
1060 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1061 ewrt = _mm_mul_ps(r11,ewtabscale);
1062 ewitab = _mm_cvttps_epi32(ewrt);
1063 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1064 ewitab = _mm_slli_epi32(ewitab,2);
1065 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1066 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1067 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1068 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1069 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1070 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1071 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1072 velec = _mm_mul_ps(qq11,_mm_sub_ps(rinv11,velec));
1073 felec = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec));
1074
1075 /* Update potential sum for this i atom from the interaction with this j atom. */
1076 velec = _mm_andnot_ps(dummy_mask,velec);
1077 velecsum = _mm_add_ps(velecsum,velec);
1078
1079 fscal = felec;
1080
1081 fscal = _mm_andnot_ps(dummy_mask,fscal);
1082
1083 /* Calculate temporary vectorial force */
1084 tx = _mm_mul_ps(fscal,dx11);
1085 ty = _mm_mul_ps(fscal,dy11);
1086 tz = _mm_mul_ps(fscal,dz11);
1087
1088 /* Update vectorial force */
1089 fix1 = _mm_add_ps(fix1,tx);
1090 fiy1 = _mm_add_ps(fiy1,ty);
1091 fiz1 = _mm_add_ps(fiz1,tz);
1092
1093 fjx1 = _mm_add_ps(fjx1,tx);
1094 fjy1 = _mm_add_ps(fjy1,ty);
1095 fjz1 = _mm_add_ps(fjz1,tz);
1096
1097 /**************************
1098 * CALCULATE INTERACTIONS *
1099 **************************/
1100
1101 r12 = _mm_mul_ps(rsq12,rinv12);
1102 r12 = _mm_andnot_ps(dummy_mask,r12);
1103
1104 /* EWALD ELECTROSTATICS */
1105
1106 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1107 ewrt = _mm_mul_ps(r12,ewtabscale);
1108 ewitab = _mm_cvttps_epi32(ewrt);
1109 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1110 ewitab = _mm_slli_epi32(ewitab,2);
1111 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1112 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1113 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1114 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1115 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1116 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1117 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1118 velec = _mm_mul_ps(qq12,_mm_sub_ps(rinv12,velec));
1119 felec = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec));
1120
1121 /* Update potential sum for this i atom from the interaction with this j atom. */
1122 velec = _mm_andnot_ps(dummy_mask,velec);
1123 velecsum = _mm_add_ps(velecsum,velec);
1124
1125 fscal = felec;
1126
1127 fscal = _mm_andnot_ps(dummy_mask,fscal);
1128
1129 /* Calculate temporary vectorial force */
1130 tx = _mm_mul_ps(fscal,dx12);
1131 ty = _mm_mul_ps(fscal,dy12);
1132 tz = _mm_mul_ps(fscal,dz12);
1133
1134 /* Update vectorial force */
1135 fix1 = _mm_add_ps(fix1,tx);
1136 fiy1 = _mm_add_ps(fiy1,ty);
1137 fiz1 = _mm_add_ps(fiz1,tz);
1138
1139 fjx2 = _mm_add_ps(fjx2,tx);
1140 fjy2 = _mm_add_ps(fjy2,ty);
1141 fjz2 = _mm_add_ps(fjz2,tz);
1142
1143 /**************************
1144 * CALCULATE INTERACTIONS *
1145 **************************/
1146
1147 r20 = _mm_mul_ps(rsq20,rinv20);
1148 r20 = _mm_andnot_ps(dummy_mask,r20);
1149
1150 /* EWALD ELECTROSTATICS */
1151
1152 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1153 ewrt = _mm_mul_ps(r20,ewtabscale);
1154 ewitab = _mm_cvttps_epi32(ewrt);
1155 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1156 ewitab = _mm_slli_epi32(ewitab,2);
1157 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1158 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1159 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1160 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1161 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1162 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1163 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1164 velec = _mm_mul_ps(qq20,_mm_sub_ps(rinv20,velec));
1165 felec = _mm_mul_ps(_mm_mul_ps(qq20,rinv20),_mm_sub_ps(rinvsq20,felec));
1166
1167 /* Update potential sum for this i atom from the interaction with this j atom. */
1168 velec = _mm_andnot_ps(dummy_mask,velec);
1169 velecsum = _mm_add_ps(velecsum,velec);
1170
1171 fscal = felec;
1172
1173 fscal = _mm_andnot_ps(dummy_mask,fscal);
1174
1175 /* Calculate temporary vectorial force */
1176 tx = _mm_mul_ps(fscal,dx20);
1177 ty = _mm_mul_ps(fscal,dy20);
1178 tz = _mm_mul_ps(fscal,dz20);
1179
1180 /* Update vectorial force */
1181 fix2 = _mm_add_ps(fix2,tx);
1182 fiy2 = _mm_add_ps(fiy2,ty);
1183 fiz2 = _mm_add_ps(fiz2,tz);
1184
1185 fjx0 = _mm_add_ps(fjx0,tx);
1186 fjy0 = _mm_add_ps(fjy0,ty);
1187 fjz0 = _mm_add_ps(fjz0,tz);
1188
1189 /**************************
1190 * CALCULATE INTERACTIONS *
1191 **************************/
1192
1193 r21 = _mm_mul_ps(rsq21,rinv21);
1194 r21 = _mm_andnot_ps(dummy_mask,r21);
1195
1196 /* EWALD ELECTROSTATICS */
1197
1198 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1199 ewrt = _mm_mul_ps(r21,ewtabscale);
1200 ewitab = _mm_cvttps_epi32(ewrt);
1201 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1202 ewitab = _mm_slli_epi32(ewitab,2);
1203 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1204 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1205 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1206 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1207 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1208 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1209 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1210 velec = _mm_mul_ps(qq21,_mm_sub_ps(rinv21,velec));
1211 felec = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec));
1212
1213 /* Update potential sum for this i atom from the interaction with this j atom. */
1214 velec = _mm_andnot_ps(dummy_mask,velec);
1215 velecsum = _mm_add_ps(velecsum,velec);
1216
1217 fscal = felec;
1218
1219 fscal = _mm_andnot_ps(dummy_mask,fscal);
1220
1221 /* Calculate temporary vectorial force */
1222 tx = _mm_mul_ps(fscal,dx21);
1223 ty = _mm_mul_ps(fscal,dy21);
1224 tz = _mm_mul_ps(fscal,dz21);
1225
1226 /* Update vectorial force */
1227 fix2 = _mm_add_ps(fix2,tx);
1228 fiy2 = _mm_add_ps(fiy2,ty);
1229 fiz2 = _mm_add_ps(fiz2,tz);
1230
1231 fjx1 = _mm_add_ps(fjx1,tx);
1232 fjy1 = _mm_add_ps(fjy1,ty);
1233 fjz1 = _mm_add_ps(fjz1,tz);
1234
1235 /**************************
1236 * CALCULATE INTERACTIONS *
1237 **************************/
1238
1239 r22 = _mm_mul_ps(rsq22,rinv22);
1240 r22 = _mm_andnot_ps(dummy_mask,r22);
1241
1242 /* EWALD ELECTROSTATICS */
1243
1244 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1245 ewrt = _mm_mul_ps(r22,ewtabscale);
1246 ewitab = _mm_cvttps_epi32(ewrt);
1247 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1248 ewitab = _mm_slli_epi32(ewitab,2);
1249 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1250 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1251 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1252 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1253 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1254 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1255 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1256 velec = _mm_mul_ps(qq22,_mm_sub_ps(rinv22,velec));
1257 felec = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec));
1258
1259 /* Update potential sum for this i atom from the interaction with this j atom. */
1260 velec = _mm_andnot_ps(dummy_mask,velec);
1261 velecsum = _mm_add_ps(velecsum,velec);
1262
1263 fscal = felec;
1264
1265 fscal = _mm_andnot_ps(dummy_mask,fscal);
1266
1267 /* Calculate temporary vectorial force */
1268 tx = _mm_mul_ps(fscal,dx22);
1269 ty = _mm_mul_ps(fscal,dy22);
1270 tz = _mm_mul_ps(fscal,dz22);
1271
1272 /* Update vectorial force */
1273 fix2 = _mm_add_ps(fix2,tx);
1274 fiy2 = _mm_add_ps(fiy2,ty);
1275 fiz2 = _mm_add_ps(fiz2,tz);
1276
1277 fjx2 = _mm_add_ps(fjx2,tx);
1278 fjy2 = _mm_add_ps(fjy2,ty);
1279 fjz2 = _mm_add_ps(fjz2,tz);
1280
1281 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1282 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1283 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1284 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1285
1286 gmx_mm_decrement_3rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
1287 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1288
1289 /* Inner loop uses 412 flops */
1290 }
1291
1292 /* End of innermost loop */
1293
1294 gmx_mm_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1295 f+i_coord_offset,fshift+i_shift_offset);
1296
1297 ggid = gid[iidx];
1298 /* Update potential energies */
1299 gmx_mm_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
1300 gmx_mm_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
1301
1302 /* Increment number of inner iterations */
1303 inneriter += j_index_end - j_index_start;
1304
1305 /* Outer loop uses 20 flops */
1306 }
1307
1308 /* Increment number of outer iterations */
1309 outeriter += nri;
1310
1311 /* Update outer/inner flops */
1312
1313 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_VF,outeriter*20 + inneriter*412);
1314 }
1315 /*
1316 * Gromacs nonbonded kernel: nb_kernel_ElecEw_VdwCSTab_GeomW3W3_F_sse2_single
1317 * Electrostatics interaction: Ewald
1318 * VdW interaction: CubicSplineTable
1319 * Geometry: Water3-Water3
1320 * Calculate force/pot: Force
1321 */
1322 void
1323 nb_kernel_ElecEw_VdwCSTab_GeomW3W3_F_sse2_single
1324 (t_nblist * gmx_restrict nlist,
1325 rvec * gmx_restrict xx,
1326 rvec * gmx_restrict ff,
1327 t_forcerec * gmx_restrict fr,
1328 t_mdatoms * gmx_restrict mdatoms,
1329 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1330 t_nrnb * gmx_restrict nrnb)
1331 {
1332 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1333 * just 0 for non-waters.
1334 * Suffixes A,B,C,D refer to j loop unrolling done with SSE, e.g. for the four different
1335 * jnr indices corresponding to data put in the four positions in the SIMD register.
1336 */
1337 int i_shift_offset,i_coord_offset,outeriter,inneriter;
1338 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1339 int jnrA,jnrB,jnrC,jnrD;
1340 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1341 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1342 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
1343 real rcutoff_scalar;
1344 real *shiftvec,*fshift,*x,*f;
1345 real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
1346 real scratch[4*DIM];
1347 __m128 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1348 int vdwioffset0;
1349 __m128 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1350 int vdwioffset1;
1351 __m128 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1352 int vdwioffset2;
1353 __m128 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1354 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
1355 __m128 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1356 int vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
1357 __m128 jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1358 int vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
1359 __m128 jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1360 __m128 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1361 __m128 dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1362 __m128 dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1363 __m128 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1364 __m128 dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1365 __m128 dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1366 __m128 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1367 __m128 dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1368 __m128 dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1369 __m128 velec,felec,velecsum,facel,crf,krf,krf2;
1370 real *charge;
1371 int nvdwtype;
1372 __m128 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1373 int *vdwtype;
1374 real *vdwparam;
1375 __m128 one_sixth = _mm_set1_ps(1.0/6.0);
1376 __m128 one_twelfth = _mm_set1_ps(1.0/12.0);
1377 __m128i vfitab;
1378 __m128i ifour = _mm_set1_epi32(4);
1379 __m128 rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
1380 real *vftab;
1381 __m128i ewitab;
1382 __m128 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1383 real *ewtab;
1384 __m128 dummy_mask,cutoff_mask;
1385 __m128 signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
1386 __m128 one = _mm_set1_ps(1.0);
1387 __m128 two = _mm_set1_ps(2.0);
1388 x = xx[0];
1389 f = ff[0];
1390
1391 nri = nlist->nri;
1392 iinr = nlist->iinr;
1393 jindex = nlist->jindex;
1394 jjnr = nlist->jjnr;
1395 shiftidx = nlist->shift;
1396 gid = nlist->gid;
1397 shiftvec = fr->shift_vec[0];
1398 fshift = fr->fshift[0];
1399 facel = _mm_set1_ps(fr->epsfac);
1400 charge = mdatoms->chargeA;
1401 nvdwtype = fr->ntype;
1402 vdwparam = fr->nbfp;
1403 vdwtype = mdatoms->typeA;
1404
1405 vftab = kernel_data->table_vdw->data;
1406 vftabscale = _mm_set1_ps(kernel_data->table_vdw->scale);
1407
1408 sh_ewald = _mm_set1_ps(fr->ic->sh_ewald);
1409 ewtab = fr->ic->tabq_coul_F;
1410 ewtabscale = _mm_set1_ps(fr->ic->tabq_scale);
1411 ewtabhalfspace = _mm_set1_ps(0.5/fr->ic->tabq_scale);
1412
1413 /* Setup water-specific parameters */
1414 inr = nlist->iinr[0];
1415 iq0 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+0]));
1416 iq1 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1]));
1417 iq2 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2]));
1418 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
1419
1420 jq0 = _mm_set1_ps(charge[inr+0]);
1421 jq1 = _mm_set1_ps(charge[inr+1]);
1422 jq2 = _mm_set1_ps(charge[inr+2]);
1423 vdwjidx0A = 2*vdwtype[inr+0];
1424 qq00 = _mm_mul_ps(iq0,jq0);
1425 c6_00 = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A]);
1426 c12_00 = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A+1]);
1427 qq01 = _mm_mul_ps(iq0,jq1);
1428 qq02 = _mm_mul_ps(iq0,jq2);
1429 qq10 = _mm_mul_ps(iq1,jq0);
1430 qq11 = _mm_mul_ps(iq1,jq1);
1431 qq12 = _mm_mul_ps(iq1,jq2);
1432 qq20 = _mm_mul_ps(iq2,jq0);
1433 qq21 = _mm_mul_ps(iq2,jq1);
1434 qq22 = _mm_mul_ps(iq2,jq2);
1435
1436 /* Avoid stupid compiler warnings */
1437 jnrA = jnrB = jnrC = jnrD = 0;
1438 j_coord_offsetA = 0;
1439 j_coord_offsetB = 0;
1440 j_coord_offsetC = 0;
1441 j_coord_offsetD = 0;
1442
1443 outeriter = 0;
1444 inneriter = 0;
1445
1446 for(iidx=0;iidx<4*DIM;iidx++)
1447 {
1448 scratch[iidx] = 0.0;
1449 }
1450
1451 /* Start outer loop over neighborlists */
1452 for(iidx=0; iidx<nri; iidx++)
1453 {
1454 /* Load shift vector for this list */
1455 i_shift_offset = DIM*shiftidx[iidx];
1456
1457 /* Load limits for loop over neighbors */
1458 j_index_start = jindex[iidx];
1459 j_index_end = jindex[iidx+1];
1460
1461 /* Get outer coordinate index */
1462 inr = iinr[iidx];
1463 i_coord_offset = DIM*inr;
1464
1465 /* Load i particle coords and add shift vector */
1466 gmx_mm_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1467 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1468
1469 fix0 = _mm_setzero_ps();
1470 fiy0 = _mm_setzero_ps();
1471 fiz0 = _mm_setzero_ps();
1472 fix1 = _mm_setzero_ps();
1473 fiy1 = _mm_setzero_ps();
1474 fiz1 = _mm_setzero_ps();
1475 fix2 = _mm_setzero_ps();
1476 fiy2 = _mm_setzero_ps();
1477 fiz2 = _mm_setzero_ps();
1478
1479 /* Start inner kernel loop */
1480 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
1481 {
1482
1483 /* Get j neighbor index, and coordinate index */
1484 jnrA = jjnr[jidx];
1485 jnrB = jjnr[jidx+1];
1486 jnrC = jjnr[jidx+2];
1487 jnrD = jjnr[jidx+3];
1488 j_coord_offsetA = DIM*jnrA;
1489 j_coord_offsetB = DIM*jnrB;
1490 j_coord_offsetC = DIM*jnrC;
1491 j_coord_offsetD = DIM*jnrD;
1492
1493 /* load j atom coordinates */
1494 gmx_mm_load_3rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1495 x+j_coord_offsetC,x+j_coord_offsetD,
1496 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1497
1498 /* Calculate displacement vector */
1499 dx00 = _mm_sub_ps(ix0,jx0);
1500 dy00 = _mm_sub_ps(iy0,jy0);
1501 dz00 = _mm_sub_ps(iz0,jz0);
1502 dx01 = _mm_sub_ps(ix0,jx1);
1503 dy01 = _mm_sub_ps(iy0,jy1);
1504 dz01 = _mm_sub_ps(iz0,jz1);
1505 dx02 = _mm_sub_ps(ix0,jx2);
1506 dy02 = _mm_sub_ps(iy0,jy2);
1507 dz02 = _mm_sub_ps(iz0,jz2);
1508 dx10 = _mm_sub_ps(ix1,jx0);
1509 dy10 = _mm_sub_ps(iy1,jy0);
1510 dz10 = _mm_sub_ps(iz1,jz0);
1511 dx11 = _mm_sub_ps(ix1,jx1);
1512 dy11 = _mm_sub_ps(iy1,jy1);
1513 dz11 = _mm_sub_ps(iz1,jz1);
1514 dx12 = _mm_sub_ps(ix1,jx2);
1515 dy12 = _mm_sub_ps(iy1,jy2);
1516 dz12 = _mm_sub_ps(iz1,jz2);
1517 dx20 = _mm_sub_ps(ix2,jx0);
1518 dy20 = _mm_sub_ps(iy2,jy0);
1519 dz20 = _mm_sub_ps(iz2,jz0);
1520 dx21 = _mm_sub_ps(ix2,jx1);
1521 dy21 = _mm_sub_ps(iy2,jy1);
1522 dz21 = _mm_sub_ps(iz2,jz1);
1523 dx22 = _mm_sub_ps(ix2,jx2);
1524 dy22 = _mm_sub_ps(iy2,jy2);
1525 dz22 = _mm_sub_ps(iz2,jz2);
1526
1527 /* Calculate squared distance and things based on it */
1528 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
1529 rsq01 = gmx_mm_calc_rsq_ps(dx01,dy01,dz01);
1530 rsq02 = gmx_mm_calc_rsq_ps(dx02,dy02,dz02);
1531 rsq10 = gmx_mm_calc_rsq_ps(dx10,dy10,dz10);
1532 rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
1533 rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
1534 rsq20 = gmx_mm_calc_rsq_ps(dx20,dy20,dz20);
1535 rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
1536 rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
1537
1538 rinv00 = gmx_mm_invsqrt_ps(rsq00);
1539 rinv01 = gmx_mm_invsqrt_ps(rsq01);
1540 rinv02 = gmx_mm_invsqrt_ps(rsq02);
1541 rinv10 = gmx_mm_invsqrt_ps(rsq10);
1542 rinv11 = gmx_mm_invsqrt_ps(rsq11);
1543 rinv12 = gmx_mm_invsqrt_ps(rsq12);
1544 rinv20 = gmx_mm_invsqrt_ps(rsq20);
1545 rinv21 = gmx_mm_invsqrt_ps(rsq21);
1546 rinv22 = gmx_mm_invsqrt_ps(rsq22);
1547
1548 rinvsq00 = _mm_mul_ps(rinv00,rinv00);
1549 rinvsq01 = _mm_mul_ps(rinv01,rinv01);
1550 rinvsq02 = _mm_mul_ps(rinv02,rinv02);
1551 rinvsq10 = _mm_mul_ps(rinv10,rinv10);
1552 rinvsq11 = _mm_mul_ps(rinv11,rinv11);
1553 rinvsq12 = _mm_mul_ps(rinv12,rinv12);
1554 rinvsq20 = _mm_mul_ps(rinv20,rinv20);
1555 rinvsq21 = _mm_mul_ps(rinv21,rinv21);
1556 rinvsq22 = _mm_mul_ps(rinv22,rinv22);
1557
1558 fjx0 = _mm_setzero_ps();
1559 fjy0 = _mm_setzero_ps();
1560 fjz0 = _mm_setzero_ps();
1561 fjx1 = _mm_setzero_ps();
1562 fjy1 = _mm_setzero_ps();
1563 fjz1 = _mm_setzero_ps();
1564 fjx2 = _mm_setzero_ps();
1565 fjy2 = _mm_setzero_ps();
1566 fjz2 = _mm_setzero_ps();
1567
1568 /**************************
1569 * CALCULATE INTERACTIONS *
1570 **************************/
1571
1572 r00 = _mm_mul_ps(rsq00,rinv00);
1573
1574 /* Calculate table index by multiplying r with table scale and truncate to integer */
1575 rt = _mm_mul_ps(r00,vftabscale);
1576 vfitab = _mm_cvttps_epi32(rt);
1577 vfeps = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1578 vfitab = _mm_slli_epi32(vfitab,3);
1579
1580 /* EWALD ELECTROSTATICS */
1581
1582 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1583 ewrt = _mm_mul_ps(r00,ewtabscale);
1584 ewitab = _mm_cvttps_epi32(ewrt);
1585 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1586 gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1587 ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
1588 &ewtabF,&ewtabFn);
1589 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1590 felec = _mm_mul_ps(_mm_mul_ps(qq00,rinv00),_mm_sub_ps(rinvsq00,felec));
1591
1592 /* CUBIC SPLINE TABLE DISPERSION */
1593 Y = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1594 F = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1595 G = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1596 H = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1597 _MM_TRANSPOSE4_PS(Y,F,G,H);
1598 Heps = _mm_mul_ps(vfeps,H);
1599 Fp = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1600 FF = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1601 fvdw6 = _mm_mul_ps(c6_00,FF);
1602
1603 /* CUBIC SPLINE TABLE REPULSION */
1604 vfitab = _mm_add_epi32(vfitab,ifour);
1605 Y = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1606 F = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1607 G = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1608 H = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1609 _MM_TRANSPOSE4_PS(Y,F,G,H);
1610 Heps = _mm_mul_ps(vfeps,H);
1611 Fp = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1612 FF = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1613 fvdw12 = _mm_mul_ps(c12_00,FF);
1614 fvdw = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
1615
1616 fscal = _mm_add_ps(felec,fvdw);
1617
1618 /* Calculate temporary vectorial force */
1619 tx = _mm_mul_ps(fscal,dx00);
1620 ty = _mm_mul_ps(fscal,dy00);
1621 tz = _mm_mul_ps(fscal,dz00);
1622
1623 /* Update vectorial force */
1624 fix0 = _mm_add_ps(fix0,tx);
1625 fiy0 = _mm_add_ps(fiy0,ty);
1626 fiz0 = _mm_add_ps(fiz0,tz);
1627
1628 fjx0 = _mm_add_ps(fjx0,tx);
1629 fjy0 = _mm_add_ps(fjy0,ty);
1630 fjz0 = _mm_add_ps(fjz0,tz);
1631
1632 /**************************
1633 * CALCULATE INTERACTIONS *
1634 **************************/
1635
1636 r01 = _mm_mul_ps(rsq01,rinv01);
1637
1638 /* EWALD ELECTROSTATICS */
1639
1640 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1641 ewrt = _mm_mul_ps(r01,ewtabscale);
1642 ewitab = _mm_cvttps_epi32(ewrt);
1643 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1644 gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1645 ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
1646 &ewtabF,&ewtabFn);
1647 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1648 felec = _mm_mul_ps(_mm_mul_ps(qq01,rinv01),_mm_sub_ps(rinvsq01,felec));
1649
1650 fscal = felec;
1651
1652 /* Calculate temporary vectorial force */
1653 tx = _mm_mul_ps(fscal,dx01);
1654 ty = _mm_mul_ps(fscal,dy01);
1655 tz = _mm_mul_ps(fscal,dz01);
1656
1657 /* Update vectorial force */
1658 fix0 = _mm_add_ps(fix0,tx);
1659 fiy0 = _mm_add_ps(fiy0,ty);
1660 fiz0 = _mm_add_ps(fiz0,tz);
1661
1662 fjx1 = _mm_add_ps(fjx1,tx);
1663 fjy1 = _mm_add_ps(fjy1,ty);
1664 fjz1 = _mm_add_ps(fjz1,tz);
1665
1666 /**************************
1667 * CALCULATE INTERACTIONS *
1668 **************************/
1669
1670 r02 = _mm_mul_ps(rsq02,rinv02);
1671
1672 /* EWALD ELECTROSTATICS */
1673
1674 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1675 ewrt = _mm_mul_ps(r02,ewtabscale);
1676 ewitab = _mm_cvttps_epi32(ewrt);
1677 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1678 gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1679 ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
1680 &ewtabF,&ewtabFn);
1681 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1682 felec = _mm_mul_ps(_mm_mul_ps(qq02,rinv02),_mm_sub_ps(rinvsq02,felec));
1683
1684 fscal = felec;
1685
1686 /* Calculate temporary vectorial force */
1687 tx = _mm_mul_ps(fscal,dx02);
1688 ty = _mm_mul_ps(fscal,dy02);
1689 tz = _mm_mul_ps(fscal,dz02);
1690
1691 /* Update vectorial force */
1692 fix0 = _mm_add_ps(fix0,tx);
1693 fiy0 = _mm_add_ps(fiy0,ty);
1694 fiz0 = _mm_add_ps(fiz0,tz);
1695
1696 fjx2 = _mm_add_ps(fjx2,tx);
1697 fjy2 = _mm_add_ps(fjy2,ty);
1698 fjz2 = _mm_add_ps(fjz2,tz);
1699
1700 /**************************
1701 * CALCULATE INTERACTIONS *
1702 **************************/
1703
1704 r10 = _mm_mul_ps(rsq10,rinv10);
1705
1706 /* EWALD ELECTROSTATICS */
1707
1708 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1709 ewrt = _mm_mul_ps(r10,ewtabscale);
1710 ewitab = _mm_cvttps_epi32(ewrt);
1711 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1712 gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1713 ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
1714 &ewtabF,&ewtabFn);
1715 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1716 felec = _mm_mul_ps(_mm_mul_ps(qq10,rinv10),_mm_sub_ps(rinvsq10,felec));
1717
1718 fscal = felec;
1719
1720 /* Calculate temporary vectorial force */
1721 tx = _mm_mul_ps(fscal,dx10);
1722 ty = _mm_mul_ps(fscal,dy10);
1723 tz = _mm_mul_ps(fscal,dz10);
1724
1725 /* Update vectorial force */
1726 fix1 = _mm_add_ps(fix1,tx);
1727 fiy1 = _mm_add_ps(fiy1,ty);
1728 fiz1 = _mm_add_ps(fiz1,tz);
1729
1730 fjx0 = _mm_add_ps(fjx0,tx);
1731 fjy0 = _mm_add_ps(fjy0,ty);
1732 fjz0 = _mm_add_ps(fjz0,tz);
1733
1734 /**************************
1735 * CALCULATE INTERACTIONS *
1736 **************************/
1737
1738 r11 = _mm_mul_ps(rsq11,rinv11);
1739
1740 /* EWALD ELECTROSTATICS */
1741
1742 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1743 ewrt = _mm_mul_ps(r11,ewtabscale);
1744 ewitab = _mm_cvttps_epi32(ewrt);
1745 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1746 gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1747 ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
1748 &ewtabF,&ewtabFn);
1749 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1750 felec = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec));
1751
1752 fscal = felec;
1753
1754 /* Calculate temporary vectorial force */
1755 tx = _mm_mul_ps(fscal,dx11);
1756 ty = _mm_mul_ps(fscal,dy11);
1757 tz = _mm_mul_ps(fscal,dz11);
1758
1759 /* Update vectorial force */
1760 fix1 = _mm_add_ps(fix1,tx);
1761 fiy1 = _mm_add_ps(fiy1,ty);
1762 fiz1 = _mm_add_ps(fiz1,tz);
1763
1764 fjx1 = _mm_add_ps(fjx1,tx);
1765 fjy1 = _mm_add_ps(fjy1,ty);
1766 fjz1 = _mm_add_ps(fjz1,tz);
1767
1768 /**************************
1769 * CALCULATE INTERACTIONS *
1770 **************************/
1771
1772 r12 = _mm_mul_ps(rsq12,rinv12);
1773
1774 /* EWALD ELECTROSTATICS */
1775
1776 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1777 ewrt = _mm_mul_ps(r12,ewtabscale);
1778 ewitab = _mm_cvttps_epi32(ewrt);
1779 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1780 gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1781 ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
1782 &ewtabF,&ewtabFn);
1783 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1784 felec = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec));
1785
1786 fscal = felec;
1787
1788 /* Calculate temporary vectorial force */
1789 tx = _mm_mul_ps(fscal,dx12);
1790 ty = _mm_mul_ps(fscal,dy12);
1791 tz = _mm_mul_ps(fscal,dz12);
1792
1793 /* Update vectorial force */
1794 fix1 = _mm_add_ps(fix1,tx);
1795 fiy1 = _mm_add_ps(fiy1,ty);
1796 fiz1 = _mm_add_ps(fiz1,tz);
1797
1798 fjx2 = _mm_add_ps(fjx2,tx);
1799 fjy2 = _mm_add_ps(fjy2,ty);
1800 fjz2 = _mm_add_ps(fjz2,tz);
1801
1802 /**************************
1803 * CALCULATE INTERACTIONS *
1804 **************************/
1805
1806 r20 = _mm_mul_ps(rsq20,rinv20);
1807
1808 /* EWALD ELECTROSTATICS */
1809
1810 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1811 ewrt = _mm_mul_ps(r20,ewtabscale);
1812 ewitab = _mm_cvttps_epi32(ewrt);
1813 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1814 gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1815 ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
1816 &ewtabF,&ewtabFn);
1817 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1818 felec = _mm_mul_ps(_mm_mul_ps(qq20,rinv20),_mm_sub_ps(rinvsq20,felec));
1819
1820 fscal = felec;
1821
1822 /* Calculate temporary vectorial force */
1823 tx = _mm_mul_ps(fscal,dx20);
1824 ty = _mm_mul_ps(fscal,dy20);
1825 tz = _mm_mul_ps(fscal,dz20);
1826
1827 /* Update vectorial force */
1828 fix2 = _mm_add_ps(fix2,tx);
1829 fiy2 = _mm_add_ps(fiy2,ty);
1830 fiz2 = _mm_add_ps(fiz2,tz);
1831
1832 fjx0 = _mm_add_ps(fjx0,tx);
1833 fjy0 = _mm_add_ps(fjy0,ty);
1834 fjz0 = _mm_add_ps(fjz0,tz);
1835
1836 /**************************
1837 * CALCULATE INTERACTIONS *
1838 **************************/
1839
1840 r21 = _mm_mul_ps(rsq21,rinv21);
1841
1842 /* EWALD ELECTROSTATICS */
1843
1844 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1845 ewrt = _mm_mul_ps(r21,ewtabscale);
1846 ewitab = _mm_cvttps_epi32(ewrt);
1847 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1848 gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1849 ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
1850 &ewtabF,&ewtabFn);
1851 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1852 felec = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec));
1853
1854 fscal = felec;
1855
1856 /* Calculate temporary vectorial force */
1857 tx = _mm_mul_ps(fscal,dx21);
1858 ty = _mm_mul_ps(fscal,dy21);
1859 tz = _mm_mul_ps(fscal,dz21);
1860
1861 /* Update vectorial force */
1862 fix2 = _mm_add_ps(fix2,tx);
1863 fiy2 = _mm_add_ps(fiy2,ty);
1864 fiz2 = _mm_add_ps(fiz2,tz);
1865
1866 fjx1 = _mm_add_ps(fjx1,tx);
1867 fjy1 = _mm_add_ps(fjy1,ty);
1868 fjz1 = _mm_add_ps(fjz1,tz);
1869
1870 /**************************
1871 * CALCULATE INTERACTIONS *
1872 **************************/
1873
1874 r22 = _mm_mul_ps(rsq22,rinv22);
1875
1876 /* EWALD ELECTROSTATICS */
1877
1878 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1879 ewrt = _mm_mul_ps(r22,ewtabscale);
1880 ewitab = _mm_cvttps_epi32(ewrt);
1881 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1882 gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1883 ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
1884 &ewtabF,&ewtabFn);
1885 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1886 felec = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec));
1887
1888 fscal = felec;
1889
1890 /* Calculate temporary vectorial force */
1891 tx = _mm_mul_ps(fscal,dx22);
1892 ty = _mm_mul_ps(fscal,dy22);
1893 tz = _mm_mul_ps(fscal,dz22);
1894
1895 /* Update vectorial force */
1896 fix2 = _mm_add_ps(fix2,tx);
1897 fiy2 = _mm_add_ps(fiy2,ty);
1898 fiz2 = _mm_add_ps(fiz2,tz);
1899
1900 fjx2 = _mm_add_ps(fjx2,tx);
1901 fjy2 = _mm_add_ps(fjy2,ty);
1902 fjz2 = _mm_add_ps(fjz2,tz);
1903
1904 fjptrA = f+j_coord_offsetA;
1905 fjptrB = f+j_coord_offsetB;
1906 fjptrC = f+j_coord_offsetC;
1907 fjptrD = f+j_coord_offsetD;
1908
1909 gmx_mm_decrement_3rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
1910 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1911
1912 /* Inner loop uses 350 flops */
1913 }
1914
1915 if(jidx<j_index_end)
1916 {
1917
1918 /* Get j neighbor index, and coordinate index */
1919 jnrlistA = jjnr[jidx];
1920 jnrlistB = jjnr[jidx+1];
1921 jnrlistC = jjnr[jidx+2];
1922 jnrlistD = jjnr[jidx+3];
1923 /* Sign of each element will be negative for non-real atoms.
1924 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1925 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1926 */
1927 dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
1928 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
1929 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
1930 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
1931 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
1932 j_coord_offsetA = DIM*jnrA;
1933 j_coord_offsetB = DIM*jnrB;
1934 j_coord_offsetC = DIM*jnrC;
1935 j_coord_offsetD = DIM*jnrD;
1936
1937 /* load j atom coordinates */
1938 gmx_mm_load_3rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1939 x+j_coord_offsetC,x+j_coord_offsetD,
1940 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1941
1942 /* Calculate displacement vector */
1943 dx00 = _mm_sub_ps(ix0,jx0);
1944 dy00 = _mm_sub_ps(iy0,jy0);
1945 dz00 = _mm_sub_ps(iz0,jz0);
1946 dx01 = _mm_sub_ps(ix0,jx1);
1947 dy01 = _mm_sub_ps(iy0,jy1);
1948 dz01 = _mm_sub_ps(iz0,jz1);
1949 dx02 = _mm_sub_ps(ix0,jx2);
1950 dy02 = _mm_sub_ps(iy0,jy2);
1951 dz02 = _mm_sub_ps(iz0,jz2);
1952 dx10 = _mm_sub_ps(ix1,jx0);
1953 dy10 = _mm_sub_ps(iy1,jy0);
1954 dz10 = _mm_sub_ps(iz1,jz0);
1955 dx11 = _mm_sub_ps(ix1,jx1);
1956 dy11 = _mm_sub_ps(iy1,jy1);
1957 dz11 = _mm_sub_ps(iz1,jz1);
1958 dx12 = _mm_sub_ps(ix1,jx2);
1959 dy12 = _mm_sub_ps(iy1,jy2);
1960 dz12 = _mm_sub_ps(iz1,jz2);
1961 dx20 = _mm_sub_ps(ix2,jx0);
1962 dy20 = _mm_sub_ps(iy2,jy0);
1963 dz20 = _mm_sub_ps(iz2,jz0);
1964 dx21 = _mm_sub_ps(ix2,jx1);
1965 dy21 = _mm_sub_ps(iy2,jy1);
1966 dz21 = _mm_sub_ps(iz2,jz1);
1967 dx22 = _mm_sub_ps(ix2,jx2);
1968 dy22 = _mm_sub_ps(iy2,jy2);
1969 dz22 = _mm_sub_ps(iz2,jz2);
1970
1971 /* Calculate squared distance and things based on it */
1972 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
1973 rsq01 = gmx_mm_calc_rsq_ps(dx01,dy01,dz01);
1974 rsq02 = gmx_mm_calc_rsq_ps(dx02,dy02,dz02);
1975 rsq10 = gmx_mm_calc_rsq_ps(dx10,dy10,dz10);
1976 rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
1977 rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
1978 rsq20 = gmx_mm_calc_rsq_ps(dx20,dy20,dz20);
1979 rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
1980 rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
1981
1982 rinv00 = gmx_mm_invsqrt_ps(rsq00);
1983 rinv01 = gmx_mm_invsqrt_ps(rsq01);
1984 rinv02 = gmx_mm_invsqrt_ps(rsq02);
1985 rinv10 = gmx_mm_invsqrt_ps(rsq10);
1986 rinv11 = gmx_mm_invsqrt_ps(rsq11);
1987 rinv12 = gmx_mm_invsqrt_ps(rsq12);
1988 rinv20 = gmx_mm_invsqrt_ps(rsq20);
1989 rinv21 = gmx_mm_invsqrt_ps(rsq21);
1990 rinv22 = gmx_mm_invsqrt_ps(rsq22);
1991
1992 rinvsq00 = _mm_mul_ps(rinv00,rinv00);
1993 rinvsq01 = _mm_mul_ps(rinv01,rinv01);
1994 rinvsq02 = _mm_mul_ps(rinv02,rinv02);
1995 rinvsq10 = _mm_mul_ps(rinv10,rinv10);
1996 rinvsq11 = _mm_mul_ps(rinv11,rinv11);
1997 rinvsq12 = _mm_mul_ps(rinv12,rinv12);
1998 rinvsq20 = _mm_mul_ps(rinv20,rinv20);
1999 rinvsq21 = _mm_mul_ps(rinv21,rinv21);
2000 rinvsq22 = _mm_mul_ps(rinv22,rinv22);
2001
2002 fjx0 = _mm_setzero_ps();
2003 fjy0 = _mm_setzero_ps();
2004 fjz0 = _mm_setzero_ps();
2005 fjx1 = _mm_setzero_ps();
2006 fjy1 = _mm_setzero_ps();
2007 fjz1 = _mm_setzero_ps();
2008 fjx2 = _mm_setzero_ps();
2009 fjy2 = _mm_setzero_ps();
2010 fjz2 = _mm_setzero_ps();
2011
2012 /**************************
2013 * CALCULATE INTERACTIONS *
2014 **************************/
2015
2016 r00 = _mm_mul_ps(rsq00,rinv00);
2017 r00 = _mm_andnot_ps(dummy_mask,r00);
2018
2019 /* Calculate table index by multiplying r with table scale and truncate to integer */
2020 rt = _mm_mul_ps(r00,vftabscale);
2021 vfitab = _mm_cvttps_epi32(rt);
2022 vfeps = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
2023 vfitab = _mm_slli_epi32(vfitab,3);
2024
2025 /* EWALD ELECTROSTATICS */
2026
2027 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2028 ewrt = _mm_mul_ps(r00,ewtabscale);
2029 ewitab = _mm_cvttps_epi32(ewrt);
2030 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
2031 gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
2032 ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
2033 &ewtabF,&ewtabFn);
2034 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2035 felec = _mm_mul_ps(_mm_mul_ps(qq00,rinv00),_mm_sub_ps(rinvsq00,felec));
2036
2037 /* CUBIC SPLINE TABLE DISPERSION */
2038 Y = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2039 F = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2040 G = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2041 H = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2042 _MM_TRANSPOSE4_PS(Y,F,G,H);
2043 Heps = _mm_mul_ps(vfeps,H);
2044 Fp = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2045 FF = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2046 fvdw6 = _mm_mul_ps(c6_00,FF);
2047
2048 /* CUBIC SPLINE TABLE REPULSION */
2049 vfitab = _mm_add_epi32(vfitab,ifour);
2050 Y = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2051 F = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2052 G = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2053 H = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2054 _MM_TRANSPOSE4_PS(Y,F,G,H);
2055 Heps = _mm_mul_ps(vfeps,H);
2056 Fp = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2057 FF = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2058 fvdw12 = _mm_mul_ps(c12_00,FF);
2059 fvdw = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
2060
2061 fscal = _mm_add_ps(felec,fvdw);
2062
2063 fscal = _mm_andnot_ps(dummy_mask,fscal);
2064
2065 /* Calculate temporary vectorial force */
2066 tx = _mm_mul_ps(fscal,dx00);
2067 ty = _mm_mul_ps(fscal,dy00);
2068 tz = _mm_mul_ps(fscal,dz00);
2069
2070 /* Update vectorial force */
2071 fix0 = _mm_add_ps(fix0,tx);
2072 fiy0 = _mm_add_ps(fiy0,ty);
2073 fiz0 = _mm_add_ps(fiz0,tz);
2074
2075 fjx0 = _mm_add_ps(fjx0,tx);
2076 fjy0 = _mm_add_ps(fjy0,ty);
2077 fjz0 = _mm_add_ps(fjz0,tz);
2078
2079 /**************************
2080 * CALCULATE INTERACTIONS *
2081 **************************/
2082
2083 r01 = _mm_mul_ps(rsq01,rinv01);
2084 r01 = _mm_andnot_ps(dummy_mask,r01);
2085
2086 /* EWALD ELECTROSTATICS */
2087
2088 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2089 ewrt = _mm_mul_ps(r01,ewtabscale);
2090 ewitab = _mm_cvttps_epi32(ewrt);
2091 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
2092 gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
2093 ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
2094 &ewtabF,&ewtabFn);
2095 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2096 felec = _mm_mul_ps(_mm_mul_ps(qq01,rinv01),_mm_sub_ps(rinvsq01,felec));
2097
2098 fscal = felec;
2099
2100 fscal = _mm_andnot_ps(dummy_mask,fscal);
2101
2102 /* Calculate temporary vectorial force */
2103 tx = _mm_mul_ps(fscal,dx01);
2104 ty = _mm_mul_ps(fscal,dy01);
2105 tz = _mm_mul_ps(fscal,dz01);
2106
2107 /* Update vectorial force */
2108 fix0 = _mm_add_ps(fix0,tx);
2109 fiy0 = _mm_add_ps(fiy0,ty);
2110 fiz0 = _mm_add_ps(fiz0,tz);
2111
2112 fjx1 = _mm_add_ps(fjx1,tx);
2113 fjy1 = _mm_add_ps(fjy1,ty);
2114 fjz1 = _mm_add_ps(fjz1,tz);
2115
2116 /**************************
2117 * CALCULATE INTERACTIONS *
2118 **************************/
2119
2120 r02 = _mm_mul_ps(rsq02,rinv02);
2121 r02 = _mm_andnot_ps(dummy_mask,r02);
2122
2123 /* EWALD ELECTROSTATICS */
2124
2125 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2126 ewrt = _mm_mul_ps(r02,ewtabscale);
2127 ewitab = _mm_cvttps_epi32(ewrt);
2128 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
2129 gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
2130 ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
2131 &ewtabF,&ewtabFn);
2132 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2133 felec = _mm_mul_ps(_mm_mul_ps(qq02,rinv02),_mm_sub_ps(rinvsq02,felec));
2134
2135 fscal = felec;
2136
2137 fscal = _mm_andnot_ps(dummy_mask,fscal);
2138
2139 /* Calculate temporary vectorial force */
2140 tx = _mm_mul_ps(fscal,dx02);
2141 ty = _mm_mul_ps(fscal,dy02);
2142 tz = _mm_mul_ps(fscal,dz02);
2143
2144 /* Update vectorial force */
2145 fix0 = _mm_add_ps(fix0,tx);
2146 fiy0 = _mm_add_ps(fiy0,ty);
2147 fiz0 = _mm_add_ps(fiz0,tz);
2148
2149 fjx2 = _mm_add_ps(fjx2,tx);
2150 fjy2 = _mm_add_ps(fjy2,ty);
2151 fjz2 = _mm_add_ps(fjz2,tz);
2152
2153 /**************************
2154 * CALCULATE INTERACTIONS *
2155 **************************/
2156
2157 r10 = _mm_mul_ps(rsq10,rinv10);
2158 r10 = _mm_andnot_ps(dummy_mask,r10);
2159
2160 /* EWALD ELECTROSTATICS */
2161
2162 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2163 ewrt = _mm_mul_ps(r10,ewtabscale);
2164 ewitab = _mm_cvttps_epi32(ewrt);
2165 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
2166 gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
2167 ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
2168 &ewtabF,&ewtabFn);
2169 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2170 felec = _mm_mul_ps(_mm_mul_ps(qq10,rinv10),_mm_sub_ps(rinvsq10,felec));
2171
2172 fscal = felec;
2173
2174 fscal = _mm_andnot_ps(dummy_mask,fscal);
2175
2176 /* Calculate temporary vectorial force */
2177 tx = _mm_mul_ps(fscal,dx10);
2178 ty = _mm_mul_ps(fscal,dy10);
2179 tz = _mm_mul_ps(fscal,dz10);
2180
2181 /* Update vectorial force */
2182 fix1 = _mm_add_ps(fix1,tx);
2183 fiy1 = _mm_add_ps(fiy1,ty);
2184 fiz1 = _mm_add_ps(fiz1,tz);
2185
2186 fjx0 = _mm_add_ps(fjx0,tx);
2187 fjy0 = _mm_add_ps(fjy0,ty);
2188 fjz0 = _mm_add_ps(fjz0,tz);
2189
2190 /**************************
2191 * CALCULATE INTERACTIONS *
2192 **************************/
2193
2194 r11 = _mm_mul_ps(rsq11,rinv11);
2195 r11 = _mm_andnot_ps(dummy_mask,r11);
2196
2197 /* EWALD ELECTROSTATICS */
2198
2199 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2200 ewrt = _mm_mul_ps(r11,ewtabscale);
2201 ewitab = _mm_cvttps_epi32(ewrt);
2202 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
2203 gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
2204 ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
2205 &ewtabF,&ewtabFn);
2206 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2207 felec = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec));
2208
2209 fscal = felec;
2210
2211 fscal = _mm_andnot_ps(dummy_mask,fscal);
2212
2213 /* Calculate temporary vectorial force */
2214 tx = _mm_mul_ps(fscal,dx11);
2215 ty = _mm_mul_ps(fscal,dy11);
2216 tz = _mm_mul_ps(fscal,dz11);
2217
2218 /* Update vectorial force */
2219 fix1 = _mm_add_ps(fix1,tx);
2220 fiy1 = _mm_add_ps(fiy1,ty);
2221 fiz1 = _mm_add_ps(fiz1,tz);
2222
2223 fjx1 = _mm_add_ps(fjx1,tx);
2224 fjy1 = _mm_add_ps(fjy1,ty);
2225 fjz1 = _mm_add_ps(fjz1,tz);
2226
2227 /**************************
2228 * CALCULATE INTERACTIONS *
2229 **************************/
2230
2231 r12 = _mm_mul_ps(rsq12,rinv12);
2232 r12 = _mm_andnot_ps(dummy_mask,r12);
2233
2234 /* EWALD ELECTROSTATICS */
2235
2236 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2237 ewrt = _mm_mul_ps(r12,ewtabscale);
2238 ewitab = _mm_cvttps_epi32(ewrt);
2239 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
2240 gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
2241 ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
2242 &ewtabF,&ewtabFn);
2243 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2244 felec = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec));
2245
2246 fscal = felec;
2247
2248 fscal = _mm_andnot_ps(dummy_mask,fscal);
2249
2250 /* Calculate temporary vectorial force */
2251 tx = _mm_mul_ps(fscal,dx12);
2252 ty = _mm_mul_ps(fscal,dy12);
2253 tz = _mm_mul_ps(fscal,dz12);
2254
2255 /* Update vectorial force */
2256 fix1 = _mm_add_ps(fix1,tx);
2257 fiy1 = _mm_add_ps(fiy1,ty);
2258 fiz1 = _mm_add_ps(fiz1,tz);
2259
2260 fjx2 = _mm_add_ps(fjx2,tx);
2261 fjy2 = _mm_add_ps(fjy2,ty);
2262 fjz2 = _mm_add_ps(fjz2,tz);
2263
2264 /**************************
2265 * CALCULATE INTERACTIONS *
2266 **************************/
2267
2268 r20 = _mm_mul_ps(rsq20,rinv20);
2269 r20 = _mm_andnot_ps(dummy_mask,r20);
2270
2271 /* EWALD ELECTROSTATICS */
2272
2273 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2274 ewrt = _mm_mul_ps(r20,ewtabscale);
2275 ewitab = _mm_cvttps_epi32(ewrt);
2276 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
2277 gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
2278 ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
2279 &ewtabF,&ewtabFn);
2280 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2281 felec = _mm_mul_ps(_mm_mul_ps(qq20,rinv20),_mm_sub_ps(rinvsq20,felec));
2282
2283 fscal = felec;
2284
2285 fscal = _mm_andnot_ps(dummy_mask,fscal);
2286
2287 /* Calculate temporary vectorial force */
2288 tx = _mm_mul_ps(fscal,dx20);
2289 ty = _mm_mul_ps(fscal,dy20);
2290 tz = _mm_mul_ps(fscal,dz20);
2291
2292 /* Update vectorial force */
2293 fix2 = _mm_add_ps(fix2,tx);
2294 fiy2 = _mm_add_ps(fiy2,ty);
2295 fiz2 = _mm_add_ps(fiz2,tz);
2296
2297 fjx0 = _mm_add_ps(fjx0,tx);
2298 fjy0 = _mm_add_ps(fjy0,ty);
2299 fjz0 = _mm_add_ps(fjz0,tz);
2300
2301 /**************************
2302 * CALCULATE INTERACTIONS *
2303 **************************/
2304
2305 r21 = _mm_mul_ps(rsq21,rinv21);
2306 r21 = _mm_andnot_ps(dummy_mask,r21);
2307
2308 /* EWALD ELECTROSTATICS */
2309
2310 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2311 ewrt = _mm_mul_ps(r21,ewtabscale);
2312 ewitab = _mm_cvttps_epi32(ewrt);
2313 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
2314 gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
2315 ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
2316 &ewtabF,&ewtabFn);
2317 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2318 felec = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec));
2319
2320 fscal = felec;
2321
2322 fscal = _mm_andnot_ps(dummy_mask,fscal);
2323
2324 /* Calculate temporary vectorial force */
2325 tx = _mm_mul_ps(fscal,dx21);
2326 ty = _mm_mul_ps(fscal,dy21);
2327 tz = _mm_mul_ps(fscal,dz21);
2328
2329 /* Update vectorial force */
2330 fix2 = _mm_add_ps(fix2,tx);
2331 fiy2 = _mm_add_ps(fiy2,ty);
2332 fiz2 = _mm_add_ps(fiz2,tz);
2333
2334 fjx1 = _mm_add_ps(fjx1,tx);
2335 fjy1 = _mm_add_ps(fjy1,ty);
2336 fjz1 = _mm_add_ps(fjz1,tz);
2337
2338 /**************************
2339 * CALCULATE INTERACTIONS *
2340 **************************/
2341
2342 r22 = _mm_mul_ps(rsq22,rinv22);
2343 r22 = _mm_andnot_ps(dummy_mask,r22);
2344
2345 /* EWALD ELECTROSTATICS */
2346
2347 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2348 ewrt = _mm_mul_ps(r22,ewtabscale);
2349 ewitab = _mm_cvttps_epi32(ewrt);
2350 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
2351 gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
2352 ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
2353 &ewtabF,&ewtabFn);
2354 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2355 felec = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec));
2356
2357 fscal = felec;
2358
2359 fscal = _mm_andnot_ps(dummy_mask,fscal);
2360
2361 /* Calculate temporary vectorial force */
2362 tx = _mm_mul_ps(fscal,dx22);
2363 ty = _mm_mul_ps(fscal,dy22);
2364 tz = _mm_mul_ps(fscal,dz22);
2365
2366 /* Update vectorial force */
2367 fix2 = _mm_add_ps(fix2,tx);
2368 fiy2 = _mm_add_ps(fiy2,ty);
2369 fiz2 = _mm_add_ps(fiz2,tz);
2370
2371 fjx2 = _mm_add_ps(fjx2,tx);
2372 fjy2 = _mm_add_ps(fjy2,ty);
2373 fjz2 = _mm_add_ps(fjz2,tz);
2374
2375 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
2376 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
2377 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
2378 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
2379
2380 gmx_mm_decrement_3rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
2381 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
2382
2383 /* Inner loop uses 359 flops */
2384 }
2385
2386 /* End of innermost loop */
2387
2388 gmx_mm_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
2389 f+i_coord_offset,fshift+i_shift_offset);
2390
2391 /* Increment number of inner iterations */
2392 inneriter += j_index_end - j_index_start;
2393
2394 /* Outer loop uses 18 flops */
2395 }
2396
2397 /* Increment number of outer iterations */
2398 outeriter += nri;
2399
2400 /* Update outer/inner flops */
2401
2402 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_F,outeriter*18 + inneriter*359);
2403 }
The ultimate molecular dynamics simulation package