?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Remove all unnecessary HAVE_CONFIG_H
[gromacs.git] / src / gromacs / gmxlib / nonbonded / nb_kernel_avx_128_fma_single / nb_kernel_ElecRF_VdwCSTab_GeomW3W3_avx_128_fma_single.c
blob11023cba224db152cb2509cf2e2418133d922875
1 /*
2 * This file is part of the GROMACS molecular simulation package.
3 *
4 * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
5 * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
6 * and including many others, as listed in the AUTHORS file in the
7 * top-level source directory and at http://www.gromacs.org.
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 avx_128_fma_single kernel generator.
37 */
38 #include "config.h"
39
40 #include <math.h>
41
42 #include "../nb_kernel.h"
43 #include "types/simple.h"
44 #include "gromacs/math/vec.h"
45 #include "nrnb.h"
46
47 #include "gromacs/simd/math_x86_avx_128_fma_single.h"
48 #include "kernelutil_x86_avx_128_fma_single.h"
49
50 /*
51 * Gromacs nonbonded kernel: nb_kernel_ElecRF_VdwCSTab_GeomW3W3_VF_avx_128_fma_single
52 * Electrostatics interaction: ReactionField
53 * VdW interaction: CubicSplineTable
54 * Geometry: Water3-Water3
55 * Calculate force/pot: PotentialAndForce
56 */
57 void
58 nb_kernel_ElecRF_VdwCSTab_GeomW3W3_VF_avx_128_fma_single
59 (t_nblist * gmx_restrict nlist,
60 rvec * gmx_restrict xx,
61 rvec * gmx_restrict ff,
62 t_forcerec * gmx_restrict fr,
63 t_mdatoms * gmx_restrict mdatoms,
64 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
65 t_nrnb * gmx_restrict nrnb)
66 {
67 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
68 * just 0 for non-waters.
69 * Suffixes A,B,C,D refer to j loop unrolling done with AVX_128, e.g. for the four different
70 * jnr indices corresponding to data put in the four positions in the SIMD register.
71 */
72 int i_shift_offset,i_coord_offset,outeriter,inneriter;
73 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
74 int jnrA,jnrB,jnrC,jnrD;
75 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
76 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
77 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
78 real rcutoff_scalar;
79 real *shiftvec,*fshift,*x,*f;
80 real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
81 real scratch[4*DIM];
82 __m128 fscal,rcutoff,rcutoff2,jidxall;
83 int vdwioffset0;
84 __m128 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
85 int vdwioffset1;
86 __m128 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
87 int vdwioffset2;
88 __m128 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
89 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
90 __m128 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
91 int vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
92 __m128 jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
93 int vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
94 __m128 jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
95 __m128 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
96 __m128 dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
97 __m128 dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
98 __m128 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
99 __m128 dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
100 __m128 dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
101 __m128 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
102 __m128 dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
103 __m128 dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
104 __m128 velec,felec,velecsum,facel,crf,krf,krf2;
105 real *charge;
106 int nvdwtype;
107 __m128 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
108 int *vdwtype;
109 real *vdwparam;
110 __m128 one_sixth = _mm_set1_ps(1.0/6.0);
111 __m128 one_twelfth = _mm_set1_ps(1.0/12.0);
112 __m128i vfitab;
113 __m128i ifour = _mm_set1_epi32(4);
114 __m128 rt,vfeps,twovfeps,vftabscale,Y,F,G,H,Fp,VV,FF;
115 real *vftab;
116 __m128 dummy_mask,cutoff_mask;
117 __m128 signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
118 __m128 one = _mm_set1_ps(1.0);
119 __m128 two = _mm_set1_ps(2.0);
120 x = xx[0];
121 f = ff[0];
122
123 nri = nlist->nri;
124 iinr = nlist->iinr;
125 jindex = nlist->jindex;
126 jjnr = nlist->jjnr;
127 shiftidx = nlist->shift;
128 gid = nlist->gid;
129 shiftvec = fr->shift_vec[0];
130 fshift = fr->fshift[0];
131 facel = _mm_set1_ps(fr->epsfac);
132 charge = mdatoms->chargeA;
133 krf = _mm_set1_ps(fr->ic->k_rf);
134 krf2 = _mm_set1_ps(fr->ic->k_rf*2.0);
135 crf = _mm_set1_ps(fr->ic->c_rf);
136 nvdwtype = fr->ntype;
137 vdwparam = fr->nbfp;
138 vdwtype = mdatoms->typeA;
139
140 vftab = kernel_data->table_vdw->data;
141 vftabscale = _mm_set1_ps(kernel_data->table_vdw->scale);
142
143 /* Setup water-specific parameters */
144 inr = nlist->iinr[0];
145 iq0 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+0]));
146 iq1 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1]));
147 iq2 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2]));
148 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
149
150 jq0 = _mm_set1_ps(charge[inr+0]);
151 jq1 = _mm_set1_ps(charge[inr+1]);
152 jq2 = _mm_set1_ps(charge[inr+2]);
153 vdwjidx0A = 2*vdwtype[inr+0];
154 qq00 = _mm_mul_ps(iq0,jq0);
155 c6_00 = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A]);
156 c12_00 = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A+1]);
157 qq01 = _mm_mul_ps(iq0,jq1);
158 qq02 = _mm_mul_ps(iq0,jq2);
159 qq10 = _mm_mul_ps(iq1,jq0);
160 qq11 = _mm_mul_ps(iq1,jq1);
161 qq12 = _mm_mul_ps(iq1,jq2);
162 qq20 = _mm_mul_ps(iq2,jq0);
163 qq21 = _mm_mul_ps(iq2,jq1);
164 qq22 = _mm_mul_ps(iq2,jq2);
165
166 /* Avoid stupid compiler warnings */
167 jnrA = jnrB = jnrC = jnrD = 0;
168 j_coord_offsetA = 0;
169 j_coord_offsetB = 0;
170 j_coord_offsetC = 0;
171 j_coord_offsetD = 0;
172
173 outeriter = 0;
174 inneriter = 0;
175
176 for(iidx=0;iidx<4*DIM;iidx++)
177 {
178 scratch[iidx] = 0.0;
179 }
180
181 /* Start outer loop over neighborlists */
182 for(iidx=0; iidx<nri; iidx++)
183 {
184 /* Load shift vector for this list */
185 i_shift_offset = DIM*shiftidx[iidx];
186
187 /* Load limits for loop over neighbors */
188 j_index_start = jindex[iidx];
189 j_index_end = jindex[iidx+1];
190
191 /* Get outer coordinate index */
192 inr = iinr[iidx];
193 i_coord_offset = DIM*inr;
194
195 /* Load i particle coords and add shift vector */
196 gmx_mm_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
197 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
198
199 fix0 = _mm_setzero_ps();
200 fiy0 = _mm_setzero_ps();
201 fiz0 = _mm_setzero_ps();
202 fix1 = _mm_setzero_ps();
203 fiy1 = _mm_setzero_ps();
204 fiz1 = _mm_setzero_ps();
205 fix2 = _mm_setzero_ps();
206 fiy2 = _mm_setzero_ps();
207 fiz2 = _mm_setzero_ps();
208
209 /* Reset potential sums */
210 velecsum = _mm_setzero_ps();
211 vvdwsum = _mm_setzero_ps();
212
213 /* Start inner kernel loop */
214 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
215 {
216
217 /* Get j neighbor index, and coordinate index */
218 jnrA = jjnr[jidx];
219 jnrB = jjnr[jidx+1];
220 jnrC = jjnr[jidx+2];
221 jnrD = jjnr[jidx+3];
222 j_coord_offsetA = DIM*jnrA;
223 j_coord_offsetB = DIM*jnrB;
224 j_coord_offsetC = DIM*jnrC;
225 j_coord_offsetD = DIM*jnrD;
226
227 /* load j atom coordinates */
228 gmx_mm_load_3rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
229 x+j_coord_offsetC,x+j_coord_offsetD,
230 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
231
232 /* Calculate displacement vector */
233 dx00 = _mm_sub_ps(ix0,jx0);
234 dy00 = _mm_sub_ps(iy0,jy0);
235 dz00 = _mm_sub_ps(iz0,jz0);
236 dx01 = _mm_sub_ps(ix0,jx1);
237 dy01 = _mm_sub_ps(iy0,jy1);
238 dz01 = _mm_sub_ps(iz0,jz1);
239 dx02 = _mm_sub_ps(ix0,jx2);
240 dy02 = _mm_sub_ps(iy0,jy2);
241 dz02 = _mm_sub_ps(iz0,jz2);
242 dx10 = _mm_sub_ps(ix1,jx0);
243 dy10 = _mm_sub_ps(iy1,jy0);
244 dz10 = _mm_sub_ps(iz1,jz0);
245 dx11 = _mm_sub_ps(ix1,jx1);
246 dy11 = _mm_sub_ps(iy1,jy1);
247 dz11 = _mm_sub_ps(iz1,jz1);
248 dx12 = _mm_sub_ps(ix1,jx2);
249 dy12 = _mm_sub_ps(iy1,jy2);
250 dz12 = _mm_sub_ps(iz1,jz2);
251 dx20 = _mm_sub_ps(ix2,jx0);
252 dy20 = _mm_sub_ps(iy2,jy0);
253 dz20 = _mm_sub_ps(iz2,jz0);
254 dx21 = _mm_sub_ps(ix2,jx1);
255 dy21 = _mm_sub_ps(iy2,jy1);
256 dz21 = _mm_sub_ps(iz2,jz1);
257 dx22 = _mm_sub_ps(ix2,jx2);
258 dy22 = _mm_sub_ps(iy2,jy2);
259 dz22 = _mm_sub_ps(iz2,jz2);
260
261 /* Calculate squared distance and things based on it */
262 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
263 rsq01 = gmx_mm_calc_rsq_ps(dx01,dy01,dz01);
264 rsq02 = gmx_mm_calc_rsq_ps(dx02,dy02,dz02);
265 rsq10 = gmx_mm_calc_rsq_ps(dx10,dy10,dz10);
266 rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
267 rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
268 rsq20 = gmx_mm_calc_rsq_ps(dx20,dy20,dz20);
269 rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
270 rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
271
272 rinv00 = gmx_mm_invsqrt_ps(rsq00);
273 rinv01 = gmx_mm_invsqrt_ps(rsq01);
274 rinv02 = gmx_mm_invsqrt_ps(rsq02);
275 rinv10 = gmx_mm_invsqrt_ps(rsq10);
276 rinv11 = gmx_mm_invsqrt_ps(rsq11);
277 rinv12 = gmx_mm_invsqrt_ps(rsq12);
278 rinv20 = gmx_mm_invsqrt_ps(rsq20);
279 rinv21 = gmx_mm_invsqrt_ps(rsq21);
280 rinv22 = gmx_mm_invsqrt_ps(rsq22);
281
282 rinvsq00 = _mm_mul_ps(rinv00,rinv00);
283 rinvsq01 = _mm_mul_ps(rinv01,rinv01);
284 rinvsq02 = _mm_mul_ps(rinv02,rinv02);
285 rinvsq10 = _mm_mul_ps(rinv10,rinv10);
286 rinvsq11 = _mm_mul_ps(rinv11,rinv11);
287 rinvsq12 = _mm_mul_ps(rinv12,rinv12);
288 rinvsq20 = _mm_mul_ps(rinv20,rinv20);
289 rinvsq21 = _mm_mul_ps(rinv21,rinv21);
290 rinvsq22 = _mm_mul_ps(rinv22,rinv22);
291
292 fjx0 = _mm_setzero_ps();
293 fjy0 = _mm_setzero_ps();
294 fjz0 = _mm_setzero_ps();
295 fjx1 = _mm_setzero_ps();
296 fjy1 = _mm_setzero_ps();
297 fjz1 = _mm_setzero_ps();
298 fjx2 = _mm_setzero_ps();
299 fjy2 = _mm_setzero_ps();
300 fjz2 = _mm_setzero_ps();
301
302 /**************************
303 * CALCULATE INTERACTIONS *
304 **************************/
305
306 r00 = _mm_mul_ps(rsq00,rinv00);
307
308 /* Calculate table index by multiplying r with table scale and truncate to integer */
309 rt = _mm_mul_ps(r00,vftabscale);
310 vfitab = _mm_cvttps_epi32(rt);
311 #ifdef __XOP__
312 vfeps = _mm_frcz_ps(rt);
313 #else
314 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
315 #endif
316 twovfeps = _mm_add_ps(vfeps,vfeps);
317 vfitab = _mm_slli_epi32(vfitab,3);
318
319 /* REACTION-FIELD ELECTROSTATICS */
320 velec = _mm_mul_ps(qq00,_mm_sub_ps(_mm_macc_ps(krf,rsq00,rinv00),crf));
321 felec = _mm_mul_ps(qq00,_mm_msub_ps(rinv00,rinvsq00,krf2));
322
323 /* CUBIC SPLINE TABLE DISPERSION */
324 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
325 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
326 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
327 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
328 _MM_TRANSPOSE4_PS(Y,F,G,H);
329 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
330 VV = _mm_macc_ps(vfeps,Fp,Y);
331 vvdw6 = _mm_mul_ps(c6_00,VV);
332 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
333 fvdw6 = _mm_mul_ps(c6_00,FF);
334
335 /* CUBIC SPLINE TABLE REPULSION */
336 vfitab = _mm_add_epi32(vfitab,ifour);
337 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
338 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
339 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
340 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
341 _MM_TRANSPOSE4_PS(Y,F,G,H);
342 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
343 VV = _mm_macc_ps(vfeps,Fp,Y);
344 vvdw12 = _mm_mul_ps(c12_00,VV);
345 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
346 fvdw12 = _mm_mul_ps(c12_00,FF);
347 vvdw = _mm_add_ps(vvdw12,vvdw6);
348 fvdw = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
349
350 /* Update potential sum for this i atom from the interaction with this j atom. */
351 velecsum = _mm_add_ps(velecsum,velec);
352 vvdwsum = _mm_add_ps(vvdwsum,vvdw);
353
354 fscal = _mm_add_ps(felec,fvdw);
355
356 /* Update vectorial force */
357 fix0 = _mm_macc_ps(dx00,fscal,fix0);
358 fiy0 = _mm_macc_ps(dy00,fscal,fiy0);
359 fiz0 = _mm_macc_ps(dz00,fscal,fiz0);
360
361 fjx0 = _mm_macc_ps(dx00,fscal,fjx0);
362 fjy0 = _mm_macc_ps(dy00,fscal,fjy0);
363 fjz0 = _mm_macc_ps(dz00,fscal,fjz0);
364
365 /**************************
366 * CALCULATE INTERACTIONS *
367 **************************/
368
369 /* REACTION-FIELD ELECTROSTATICS */
370 velec = _mm_mul_ps(qq01,_mm_sub_ps(_mm_macc_ps(krf,rsq01,rinv01),crf));
371 felec = _mm_mul_ps(qq01,_mm_msub_ps(rinv01,rinvsq01,krf2));
372
373 /* Update potential sum for this i atom from the interaction with this j atom. */
374 velecsum = _mm_add_ps(velecsum,velec);
375
376 fscal = felec;
377
378 /* Update vectorial force */
379 fix0 = _mm_macc_ps(dx01,fscal,fix0);
380 fiy0 = _mm_macc_ps(dy01,fscal,fiy0);
381 fiz0 = _mm_macc_ps(dz01,fscal,fiz0);
382
383 fjx1 = _mm_macc_ps(dx01,fscal,fjx1);
384 fjy1 = _mm_macc_ps(dy01,fscal,fjy1);
385 fjz1 = _mm_macc_ps(dz01,fscal,fjz1);
386
387 /**************************
388 * CALCULATE INTERACTIONS *
389 **************************/
390
391 /* REACTION-FIELD ELECTROSTATICS */
392 velec = _mm_mul_ps(qq02,_mm_sub_ps(_mm_macc_ps(krf,rsq02,rinv02),crf));
393 felec = _mm_mul_ps(qq02,_mm_msub_ps(rinv02,rinvsq02,krf2));
394
395 /* Update potential sum for this i atom from the interaction with this j atom. */
396 velecsum = _mm_add_ps(velecsum,velec);
397
398 fscal = felec;
399
400 /* Update vectorial force */
401 fix0 = _mm_macc_ps(dx02,fscal,fix0);
402 fiy0 = _mm_macc_ps(dy02,fscal,fiy0);
403 fiz0 = _mm_macc_ps(dz02,fscal,fiz0);
404
405 fjx2 = _mm_macc_ps(dx02,fscal,fjx2);
406 fjy2 = _mm_macc_ps(dy02,fscal,fjy2);
407 fjz2 = _mm_macc_ps(dz02,fscal,fjz2);
408
409 /**************************
410 * CALCULATE INTERACTIONS *
411 **************************/
412
413 /* REACTION-FIELD ELECTROSTATICS */
414 velec = _mm_mul_ps(qq10,_mm_sub_ps(_mm_macc_ps(krf,rsq10,rinv10),crf));
415 felec = _mm_mul_ps(qq10,_mm_msub_ps(rinv10,rinvsq10,krf2));
416
417 /* Update potential sum for this i atom from the interaction with this j atom. */
418 velecsum = _mm_add_ps(velecsum,velec);
419
420 fscal = felec;
421
422 /* Update vectorial force */
423 fix1 = _mm_macc_ps(dx10,fscal,fix1);
424 fiy1 = _mm_macc_ps(dy10,fscal,fiy1);
425 fiz1 = _mm_macc_ps(dz10,fscal,fiz1);
426
427 fjx0 = _mm_macc_ps(dx10,fscal,fjx0);
428 fjy0 = _mm_macc_ps(dy10,fscal,fjy0);
429 fjz0 = _mm_macc_ps(dz10,fscal,fjz0);
430
431 /**************************
432 * CALCULATE INTERACTIONS *
433 **************************/
434
435 /* REACTION-FIELD ELECTROSTATICS */
436 velec = _mm_mul_ps(qq11,_mm_sub_ps(_mm_macc_ps(krf,rsq11,rinv11),crf));
437 felec = _mm_mul_ps(qq11,_mm_msub_ps(rinv11,rinvsq11,krf2));
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 /* Update vectorial force */
445 fix1 = _mm_macc_ps(dx11,fscal,fix1);
446 fiy1 = _mm_macc_ps(dy11,fscal,fiy1);
447 fiz1 = _mm_macc_ps(dz11,fscal,fiz1);
448
449 fjx1 = _mm_macc_ps(dx11,fscal,fjx1);
450 fjy1 = _mm_macc_ps(dy11,fscal,fjy1);
451 fjz1 = _mm_macc_ps(dz11,fscal,fjz1);
452
453 /**************************
454 * CALCULATE INTERACTIONS *
455 **************************/
456
457 /* REACTION-FIELD ELECTROSTATICS */
458 velec = _mm_mul_ps(qq12,_mm_sub_ps(_mm_macc_ps(krf,rsq12,rinv12),crf));
459 felec = _mm_mul_ps(qq12,_mm_msub_ps(rinv12,rinvsq12,krf2));
460
461 /* Update potential sum for this i atom from the interaction with this j atom. */
462 velecsum = _mm_add_ps(velecsum,velec);
463
464 fscal = felec;
465
466 /* Update vectorial force */
467 fix1 = _mm_macc_ps(dx12,fscal,fix1);
468 fiy1 = _mm_macc_ps(dy12,fscal,fiy1);
469 fiz1 = _mm_macc_ps(dz12,fscal,fiz1);
470
471 fjx2 = _mm_macc_ps(dx12,fscal,fjx2);
472 fjy2 = _mm_macc_ps(dy12,fscal,fjy2);
473 fjz2 = _mm_macc_ps(dz12,fscal,fjz2);
474
475 /**************************
476 * CALCULATE INTERACTIONS *
477 **************************/
478
479 /* REACTION-FIELD ELECTROSTATICS */
480 velec = _mm_mul_ps(qq20,_mm_sub_ps(_mm_macc_ps(krf,rsq20,rinv20),crf));
481 felec = _mm_mul_ps(qq20,_mm_msub_ps(rinv20,rinvsq20,krf2));
482
483 /* Update potential sum for this i atom from the interaction with this j atom. */
484 velecsum = _mm_add_ps(velecsum,velec);
485
486 fscal = felec;
487
488 /* Update vectorial force */
489 fix2 = _mm_macc_ps(dx20,fscal,fix2);
490 fiy2 = _mm_macc_ps(dy20,fscal,fiy2);
491 fiz2 = _mm_macc_ps(dz20,fscal,fiz2);
492
493 fjx0 = _mm_macc_ps(dx20,fscal,fjx0);
494 fjy0 = _mm_macc_ps(dy20,fscal,fjy0);
495 fjz0 = _mm_macc_ps(dz20,fscal,fjz0);
496
497 /**************************
498 * CALCULATE INTERACTIONS *
499 **************************/
500
501 /* REACTION-FIELD ELECTROSTATICS */
502 velec = _mm_mul_ps(qq21,_mm_sub_ps(_mm_macc_ps(krf,rsq21,rinv21),crf));
503 felec = _mm_mul_ps(qq21,_mm_msub_ps(rinv21,rinvsq21,krf2));
504
505 /* Update potential sum for this i atom from the interaction with this j atom. */
506 velecsum = _mm_add_ps(velecsum,velec);
507
508 fscal = felec;
509
510 /* Update vectorial force */
511 fix2 = _mm_macc_ps(dx21,fscal,fix2);
512 fiy2 = _mm_macc_ps(dy21,fscal,fiy2);
513 fiz2 = _mm_macc_ps(dz21,fscal,fiz2);
514
515 fjx1 = _mm_macc_ps(dx21,fscal,fjx1);
516 fjy1 = _mm_macc_ps(dy21,fscal,fjy1);
517 fjz1 = _mm_macc_ps(dz21,fscal,fjz1);
518
519 /**************************
520 * CALCULATE INTERACTIONS *
521 **************************/
522
523 /* REACTION-FIELD ELECTROSTATICS */
524 velec = _mm_mul_ps(qq22,_mm_sub_ps(_mm_macc_ps(krf,rsq22,rinv22),crf));
525 felec = _mm_mul_ps(qq22,_mm_msub_ps(rinv22,rinvsq22,krf2));
526
527 /* Update potential sum for this i atom from the interaction with this j atom. */
528 velecsum = _mm_add_ps(velecsum,velec);
529
530 fscal = felec;
531
532 /* Update vectorial force */
533 fix2 = _mm_macc_ps(dx22,fscal,fix2);
534 fiy2 = _mm_macc_ps(dy22,fscal,fiy2);
535 fiz2 = _mm_macc_ps(dz22,fscal,fiz2);
536
537 fjx2 = _mm_macc_ps(dx22,fscal,fjx2);
538 fjy2 = _mm_macc_ps(dy22,fscal,fjy2);
539 fjz2 = _mm_macc_ps(dz22,fscal,fjz2);
540
541 fjptrA = f+j_coord_offsetA;
542 fjptrB = f+j_coord_offsetB;
543 fjptrC = f+j_coord_offsetC;
544 fjptrD = f+j_coord_offsetD;
545
546 gmx_mm_decrement_3rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
547 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
548
549 /* Inner loop uses 350 flops */
550 }
551
552 if(jidx<j_index_end)
553 {
554
555 /* Get j neighbor index, and coordinate index */
556 jnrlistA = jjnr[jidx];
557 jnrlistB = jjnr[jidx+1];
558 jnrlistC = jjnr[jidx+2];
559 jnrlistD = jjnr[jidx+3];
560 /* Sign of each element will be negative for non-real atoms.
561 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
562 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
563 */
564 dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
565 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
566 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
567 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
568 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
569 j_coord_offsetA = DIM*jnrA;
570 j_coord_offsetB = DIM*jnrB;
571 j_coord_offsetC = DIM*jnrC;
572 j_coord_offsetD = DIM*jnrD;
573
574 /* load j atom coordinates */
575 gmx_mm_load_3rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
576 x+j_coord_offsetC,x+j_coord_offsetD,
577 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
578
579 /* Calculate displacement vector */
580 dx00 = _mm_sub_ps(ix0,jx0);
581 dy00 = _mm_sub_ps(iy0,jy0);
582 dz00 = _mm_sub_ps(iz0,jz0);
583 dx01 = _mm_sub_ps(ix0,jx1);
584 dy01 = _mm_sub_ps(iy0,jy1);
585 dz01 = _mm_sub_ps(iz0,jz1);
586 dx02 = _mm_sub_ps(ix0,jx2);
587 dy02 = _mm_sub_ps(iy0,jy2);
588 dz02 = _mm_sub_ps(iz0,jz2);
589 dx10 = _mm_sub_ps(ix1,jx0);
590 dy10 = _mm_sub_ps(iy1,jy0);
591 dz10 = _mm_sub_ps(iz1,jz0);
592 dx11 = _mm_sub_ps(ix1,jx1);
593 dy11 = _mm_sub_ps(iy1,jy1);
594 dz11 = _mm_sub_ps(iz1,jz1);
595 dx12 = _mm_sub_ps(ix1,jx2);
596 dy12 = _mm_sub_ps(iy1,jy2);
597 dz12 = _mm_sub_ps(iz1,jz2);
598 dx20 = _mm_sub_ps(ix2,jx0);
599 dy20 = _mm_sub_ps(iy2,jy0);
600 dz20 = _mm_sub_ps(iz2,jz0);
601 dx21 = _mm_sub_ps(ix2,jx1);
602 dy21 = _mm_sub_ps(iy2,jy1);
603 dz21 = _mm_sub_ps(iz2,jz1);
604 dx22 = _mm_sub_ps(ix2,jx2);
605 dy22 = _mm_sub_ps(iy2,jy2);
606 dz22 = _mm_sub_ps(iz2,jz2);
607
608 /* Calculate squared distance and things based on it */
609 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
610 rsq01 = gmx_mm_calc_rsq_ps(dx01,dy01,dz01);
611 rsq02 = gmx_mm_calc_rsq_ps(dx02,dy02,dz02);
612 rsq10 = gmx_mm_calc_rsq_ps(dx10,dy10,dz10);
613 rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
614 rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
615 rsq20 = gmx_mm_calc_rsq_ps(dx20,dy20,dz20);
616 rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
617 rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
618
619 rinv00 = gmx_mm_invsqrt_ps(rsq00);
620 rinv01 = gmx_mm_invsqrt_ps(rsq01);
621 rinv02 = gmx_mm_invsqrt_ps(rsq02);
622 rinv10 = gmx_mm_invsqrt_ps(rsq10);
623 rinv11 = gmx_mm_invsqrt_ps(rsq11);
624 rinv12 = gmx_mm_invsqrt_ps(rsq12);
625 rinv20 = gmx_mm_invsqrt_ps(rsq20);
626 rinv21 = gmx_mm_invsqrt_ps(rsq21);
627 rinv22 = gmx_mm_invsqrt_ps(rsq22);
628
629 rinvsq00 = _mm_mul_ps(rinv00,rinv00);
630 rinvsq01 = _mm_mul_ps(rinv01,rinv01);
631 rinvsq02 = _mm_mul_ps(rinv02,rinv02);
632 rinvsq10 = _mm_mul_ps(rinv10,rinv10);
633 rinvsq11 = _mm_mul_ps(rinv11,rinv11);
634 rinvsq12 = _mm_mul_ps(rinv12,rinv12);
635 rinvsq20 = _mm_mul_ps(rinv20,rinv20);
636 rinvsq21 = _mm_mul_ps(rinv21,rinv21);
637 rinvsq22 = _mm_mul_ps(rinv22,rinv22);
638
639 fjx0 = _mm_setzero_ps();
640 fjy0 = _mm_setzero_ps();
641 fjz0 = _mm_setzero_ps();
642 fjx1 = _mm_setzero_ps();
643 fjy1 = _mm_setzero_ps();
644 fjz1 = _mm_setzero_ps();
645 fjx2 = _mm_setzero_ps();
646 fjy2 = _mm_setzero_ps();
647 fjz2 = _mm_setzero_ps();
648
649 /**************************
650 * CALCULATE INTERACTIONS *
651 **************************/
652
653 r00 = _mm_mul_ps(rsq00,rinv00);
654 r00 = _mm_andnot_ps(dummy_mask,r00);
655
656 /* Calculate table index by multiplying r with table scale and truncate to integer */
657 rt = _mm_mul_ps(r00,vftabscale);
658 vfitab = _mm_cvttps_epi32(rt);
659 #ifdef __XOP__
660 vfeps = _mm_frcz_ps(rt);
661 #else
662 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
663 #endif
664 twovfeps = _mm_add_ps(vfeps,vfeps);
665 vfitab = _mm_slli_epi32(vfitab,3);
666
667 /* REACTION-FIELD ELECTROSTATICS */
668 velec = _mm_mul_ps(qq00,_mm_sub_ps(_mm_macc_ps(krf,rsq00,rinv00),crf));
669 felec = _mm_mul_ps(qq00,_mm_msub_ps(rinv00,rinvsq00,krf2));
670
671 /* CUBIC SPLINE TABLE DISPERSION */
672 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
673 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
674 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
675 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
676 _MM_TRANSPOSE4_PS(Y,F,G,H);
677 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
678 VV = _mm_macc_ps(vfeps,Fp,Y);
679 vvdw6 = _mm_mul_ps(c6_00,VV);
680 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
681 fvdw6 = _mm_mul_ps(c6_00,FF);
682
683 /* CUBIC SPLINE TABLE REPULSION */
684 vfitab = _mm_add_epi32(vfitab,ifour);
685 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
686 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
687 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
688 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
689 _MM_TRANSPOSE4_PS(Y,F,G,H);
690 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
691 VV = _mm_macc_ps(vfeps,Fp,Y);
692 vvdw12 = _mm_mul_ps(c12_00,VV);
693 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
694 fvdw12 = _mm_mul_ps(c12_00,FF);
695 vvdw = _mm_add_ps(vvdw12,vvdw6);
696 fvdw = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
697
698 /* Update potential sum for this i atom from the interaction with this j atom. */
699 velec = _mm_andnot_ps(dummy_mask,velec);
700 velecsum = _mm_add_ps(velecsum,velec);
701 vvdw = _mm_andnot_ps(dummy_mask,vvdw);
702 vvdwsum = _mm_add_ps(vvdwsum,vvdw);
703
704 fscal = _mm_add_ps(felec,fvdw);
705
706 fscal = _mm_andnot_ps(dummy_mask,fscal);
707
708 /* Update vectorial force */
709 fix0 = _mm_macc_ps(dx00,fscal,fix0);
710 fiy0 = _mm_macc_ps(dy00,fscal,fiy0);
711 fiz0 = _mm_macc_ps(dz00,fscal,fiz0);
712
713 fjx0 = _mm_macc_ps(dx00,fscal,fjx0);
714 fjy0 = _mm_macc_ps(dy00,fscal,fjy0);
715 fjz0 = _mm_macc_ps(dz00,fscal,fjz0);
716
717 /**************************
718 * CALCULATE INTERACTIONS *
719 **************************/
720
721 /* REACTION-FIELD ELECTROSTATICS */
722 velec = _mm_mul_ps(qq01,_mm_sub_ps(_mm_macc_ps(krf,rsq01,rinv01),crf));
723 felec = _mm_mul_ps(qq01,_mm_msub_ps(rinv01,rinvsq01,krf2));
724
725 /* Update potential sum for this i atom from the interaction with this j atom. */
726 velec = _mm_andnot_ps(dummy_mask,velec);
727 velecsum = _mm_add_ps(velecsum,velec);
728
729 fscal = felec;
730
731 fscal = _mm_andnot_ps(dummy_mask,fscal);
732
733 /* Update vectorial force */
734 fix0 = _mm_macc_ps(dx01,fscal,fix0);
735 fiy0 = _mm_macc_ps(dy01,fscal,fiy0);
736 fiz0 = _mm_macc_ps(dz01,fscal,fiz0);
737
738 fjx1 = _mm_macc_ps(dx01,fscal,fjx1);
739 fjy1 = _mm_macc_ps(dy01,fscal,fjy1);
740 fjz1 = _mm_macc_ps(dz01,fscal,fjz1);
741
742 /**************************
743 * CALCULATE INTERACTIONS *
744 **************************/
745
746 /* REACTION-FIELD ELECTROSTATICS */
747 velec = _mm_mul_ps(qq02,_mm_sub_ps(_mm_macc_ps(krf,rsq02,rinv02),crf));
748 felec = _mm_mul_ps(qq02,_mm_msub_ps(rinv02,rinvsq02,krf2));
749
750 /* Update potential sum for this i atom from the interaction with this j atom. */
751 velec = _mm_andnot_ps(dummy_mask,velec);
752 velecsum = _mm_add_ps(velecsum,velec);
753
754 fscal = felec;
755
756 fscal = _mm_andnot_ps(dummy_mask,fscal);
757
758 /* Update vectorial force */
759 fix0 = _mm_macc_ps(dx02,fscal,fix0);
760 fiy0 = _mm_macc_ps(dy02,fscal,fiy0);
761 fiz0 = _mm_macc_ps(dz02,fscal,fiz0);
762
763 fjx2 = _mm_macc_ps(dx02,fscal,fjx2);
764 fjy2 = _mm_macc_ps(dy02,fscal,fjy2);
765 fjz2 = _mm_macc_ps(dz02,fscal,fjz2);
766
767 /**************************
768 * CALCULATE INTERACTIONS *
769 **************************/
770
771 /* REACTION-FIELD ELECTROSTATICS */
772 velec = _mm_mul_ps(qq10,_mm_sub_ps(_mm_macc_ps(krf,rsq10,rinv10),crf));
773 felec = _mm_mul_ps(qq10,_mm_msub_ps(rinv10,rinvsq10,krf2));
774
775 /* Update potential sum for this i atom from the interaction with this j atom. */
776 velec = _mm_andnot_ps(dummy_mask,velec);
777 velecsum = _mm_add_ps(velecsum,velec);
778
779 fscal = felec;
780
781 fscal = _mm_andnot_ps(dummy_mask,fscal);
782
783 /* Update vectorial force */
784 fix1 = _mm_macc_ps(dx10,fscal,fix1);
785 fiy1 = _mm_macc_ps(dy10,fscal,fiy1);
786 fiz1 = _mm_macc_ps(dz10,fscal,fiz1);
787
788 fjx0 = _mm_macc_ps(dx10,fscal,fjx0);
789 fjy0 = _mm_macc_ps(dy10,fscal,fjy0);
790 fjz0 = _mm_macc_ps(dz10,fscal,fjz0);
791
792 /**************************
793 * CALCULATE INTERACTIONS *
794 **************************/
795
796 /* REACTION-FIELD ELECTROSTATICS */
797 velec = _mm_mul_ps(qq11,_mm_sub_ps(_mm_macc_ps(krf,rsq11,rinv11),crf));
798 felec = _mm_mul_ps(qq11,_mm_msub_ps(rinv11,rinvsq11,krf2));
799
800 /* Update potential sum for this i atom from the interaction with this j atom. */
801 velec = _mm_andnot_ps(dummy_mask,velec);
802 velecsum = _mm_add_ps(velecsum,velec);
803
804 fscal = felec;
805
806 fscal = _mm_andnot_ps(dummy_mask,fscal);
807
808 /* Update vectorial force */
809 fix1 = _mm_macc_ps(dx11,fscal,fix1);
810 fiy1 = _mm_macc_ps(dy11,fscal,fiy1);
811 fiz1 = _mm_macc_ps(dz11,fscal,fiz1);
812
813 fjx1 = _mm_macc_ps(dx11,fscal,fjx1);
814 fjy1 = _mm_macc_ps(dy11,fscal,fjy1);
815 fjz1 = _mm_macc_ps(dz11,fscal,fjz1);
816
817 /**************************
818 * CALCULATE INTERACTIONS *
819 **************************/
820
821 /* REACTION-FIELD ELECTROSTATICS */
822 velec = _mm_mul_ps(qq12,_mm_sub_ps(_mm_macc_ps(krf,rsq12,rinv12),crf));
823 felec = _mm_mul_ps(qq12,_mm_msub_ps(rinv12,rinvsq12,krf2));
824
825 /* Update potential sum for this i atom from the interaction with this j atom. */
826 velec = _mm_andnot_ps(dummy_mask,velec);
827 velecsum = _mm_add_ps(velecsum,velec);
828
829 fscal = felec;
830
831 fscal = _mm_andnot_ps(dummy_mask,fscal);
832
833 /* Update vectorial force */
834 fix1 = _mm_macc_ps(dx12,fscal,fix1);
835 fiy1 = _mm_macc_ps(dy12,fscal,fiy1);
836 fiz1 = _mm_macc_ps(dz12,fscal,fiz1);
837
838 fjx2 = _mm_macc_ps(dx12,fscal,fjx2);
839 fjy2 = _mm_macc_ps(dy12,fscal,fjy2);
840 fjz2 = _mm_macc_ps(dz12,fscal,fjz2);
841
842 /**************************
843 * CALCULATE INTERACTIONS *
844 **************************/
845
846 /* REACTION-FIELD ELECTROSTATICS */
847 velec = _mm_mul_ps(qq20,_mm_sub_ps(_mm_macc_ps(krf,rsq20,rinv20),crf));
848 felec = _mm_mul_ps(qq20,_mm_msub_ps(rinv20,rinvsq20,krf2));
849
850 /* Update potential sum for this i atom from the interaction with this j atom. */
851 velec = _mm_andnot_ps(dummy_mask,velec);
852 velecsum = _mm_add_ps(velecsum,velec);
853
854 fscal = felec;
855
856 fscal = _mm_andnot_ps(dummy_mask,fscal);
857
858 /* Update vectorial force */
859 fix2 = _mm_macc_ps(dx20,fscal,fix2);
860 fiy2 = _mm_macc_ps(dy20,fscal,fiy2);
861 fiz2 = _mm_macc_ps(dz20,fscal,fiz2);
862
863 fjx0 = _mm_macc_ps(dx20,fscal,fjx0);
864 fjy0 = _mm_macc_ps(dy20,fscal,fjy0);
865 fjz0 = _mm_macc_ps(dz20,fscal,fjz0);
866
867 /**************************
868 * CALCULATE INTERACTIONS *
869 **************************/
870
871 /* REACTION-FIELD ELECTROSTATICS */
872 velec = _mm_mul_ps(qq21,_mm_sub_ps(_mm_macc_ps(krf,rsq21,rinv21),crf));
873 felec = _mm_mul_ps(qq21,_mm_msub_ps(rinv21,rinvsq21,krf2));
874
875 /* Update potential sum for this i atom from the interaction with this j atom. */
876 velec = _mm_andnot_ps(dummy_mask,velec);
877 velecsum = _mm_add_ps(velecsum,velec);
878
879 fscal = felec;
880
881 fscal = _mm_andnot_ps(dummy_mask,fscal);
882
883 /* Update vectorial force */
884 fix2 = _mm_macc_ps(dx21,fscal,fix2);
885 fiy2 = _mm_macc_ps(dy21,fscal,fiy2);
886 fiz2 = _mm_macc_ps(dz21,fscal,fiz2);
887
888 fjx1 = _mm_macc_ps(dx21,fscal,fjx1);
889 fjy1 = _mm_macc_ps(dy21,fscal,fjy1);
890 fjz1 = _mm_macc_ps(dz21,fscal,fjz1);
891
892 /**************************
893 * CALCULATE INTERACTIONS *
894 **************************/
895
896 /* REACTION-FIELD ELECTROSTATICS */
897 velec = _mm_mul_ps(qq22,_mm_sub_ps(_mm_macc_ps(krf,rsq22,rinv22),crf));
898 felec = _mm_mul_ps(qq22,_mm_msub_ps(rinv22,rinvsq22,krf2));
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 /* Update vectorial force */
909 fix2 = _mm_macc_ps(dx22,fscal,fix2);
910 fiy2 = _mm_macc_ps(dy22,fscal,fiy2);
911 fiz2 = _mm_macc_ps(dz22,fscal,fiz2);
912
913 fjx2 = _mm_macc_ps(dx22,fscal,fjx2);
914 fjy2 = _mm_macc_ps(dy22,fscal,fjy2);
915 fjz2 = _mm_macc_ps(dz22,fscal,fjz2);
916
917 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
918 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
919 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
920 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
921
922 gmx_mm_decrement_3rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
923 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
924
925 /* Inner loop uses 351 flops */
926 }
927
928 /* End of innermost loop */
929
930 gmx_mm_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
931 f+i_coord_offset,fshift+i_shift_offset);
932
933 ggid = gid[iidx];
934 /* Update potential energies */
935 gmx_mm_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
936 gmx_mm_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
937
938 /* Increment number of inner iterations */
939 inneriter += j_index_end - j_index_start;
940
941 /* Outer loop uses 20 flops */
942 }
943
944 /* Increment number of outer iterations */
945 outeriter += nri;
946
947 /* Update outer/inner flops */
948
949 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_VF,outeriter*20 + inneriter*351);
950 }
951 /*
952 * Gromacs nonbonded kernel: nb_kernel_ElecRF_VdwCSTab_GeomW3W3_F_avx_128_fma_single
953 * Electrostatics interaction: ReactionField
954 * VdW interaction: CubicSplineTable
955 * Geometry: Water3-Water3
956 * Calculate force/pot: Force
957 */
958 void
959 nb_kernel_ElecRF_VdwCSTab_GeomW3W3_F_avx_128_fma_single
960 (t_nblist * gmx_restrict nlist,
961 rvec * gmx_restrict xx,
962 rvec * gmx_restrict ff,
963 t_forcerec * gmx_restrict fr,
964 t_mdatoms * gmx_restrict mdatoms,
965 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
966 t_nrnb * gmx_restrict nrnb)
967 {
968 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
969 * just 0 for non-waters.
970 * Suffixes A,B,C,D refer to j loop unrolling done with AVX_128, e.g. for the four different
971 * jnr indices corresponding to data put in the four positions in the SIMD register.
972 */
973 int i_shift_offset,i_coord_offset,outeriter,inneriter;
974 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
975 int jnrA,jnrB,jnrC,jnrD;
976 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
977 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
978 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
979 real rcutoff_scalar;
980 real *shiftvec,*fshift,*x,*f;
981 real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
982 real scratch[4*DIM];
983 __m128 fscal,rcutoff,rcutoff2,jidxall;
984 int vdwioffset0;
985 __m128 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
986 int vdwioffset1;
987 __m128 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
988 int vdwioffset2;
989 __m128 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
990 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
991 __m128 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
992 int vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
993 __m128 jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
994 int vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
995 __m128 jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
996 __m128 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
997 __m128 dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
998 __m128 dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
999 __m128 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1000 __m128 dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1001 __m128 dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1002 __m128 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1003 __m128 dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1004 __m128 dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1005 __m128 velec,felec,velecsum,facel,crf,krf,krf2;
1006 real *charge;
1007 int nvdwtype;
1008 __m128 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1009 int *vdwtype;
1010 real *vdwparam;
1011 __m128 one_sixth = _mm_set1_ps(1.0/6.0);
1012 __m128 one_twelfth = _mm_set1_ps(1.0/12.0);
1013 __m128i vfitab;
1014 __m128i ifour = _mm_set1_epi32(4);
1015 __m128 rt,vfeps,twovfeps,vftabscale,Y,F,G,H,Fp,VV,FF;
1016 real *vftab;
1017 __m128 dummy_mask,cutoff_mask;
1018 __m128 signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
1019 __m128 one = _mm_set1_ps(1.0);
1020 __m128 two = _mm_set1_ps(2.0);
1021 x = xx[0];
1022 f = ff[0];
1023
1024 nri = nlist->nri;
1025 iinr = nlist->iinr;
1026 jindex = nlist->jindex;
1027 jjnr = nlist->jjnr;
1028 shiftidx = nlist->shift;
1029 gid = nlist->gid;
1030 shiftvec = fr->shift_vec[0];
1031 fshift = fr->fshift[0];
1032 facel = _mm_set1_ps(fr->epsfac);
1033 charge = mdatoms->chargeA;
1034 krf = _mm_set1_ps(fr->ic->k_rf);
1035 krf2 = _mm_set1_ps(fr->ic->k_rf*2.0);
1036 crf = _mm_set1_ps(fr->ic->c_rf);
1037 nvdwtype = fr->ntype;
1038 vdwparam = fr->nbfp;
1039 vdwtype = mdatoms->typeA;
1040
1041 vftab = kernel_data->table_vdw->data;
1042 vftabscale = _mm_set1_ps(kernel_data->table_vdw->scale);
1043
1044 /* Setup water-specific parameters */
1045 inr = nlist->iinr[0];
1046 iq0 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+0]));
1047 iq1 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1]));
1048 iq2 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2]));
1049 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
1050
1051 jq0 = _mm_set1_ps(charge[inr+0]);
1052 jq1 = _mm_set1_ps(charge[inr+1]);
1053 jq2 = _mm_set1_ps(charge[inr+2]);
1054 vdwjidx0A = 2*vdwtype[inr+0];
1055 qq00 = _mm_mul_ps(iq0,jq0);
1056 c6_00 = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A]);
1057 c12_00 = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A+1]);
1058 qq01 = _mm_mul_ps(iq0,jq1);
1059 qq02 = _mm_mul_ps(iq0,jq2);
1060 qq10 = _mm_mul_ps(iq1,jq0);
1061 qq11 = _mm_mul_ps(iq1,jq1);
1062 qq12 = _mm_mul_ps(iq1,jq2);
1063 qq20 = _mm_mul_ps(iq2,jq0);
1064 qq21 = _mm_mul_ps(iq2,jq1);
1065 qq22 = _mm_mul_ps(iq2,jq2);
1066
1067 /* Avoid stupid compiler warnings */
1068 jnrA = jnrB = jnrC = jnrD = 0;
1069 j_coord_offsetA = 0;
1070 j_coord_offsetB = 0;
1071 j_coord_offsetC = 0;
1072 j_coord_offsetD = 0;
1073
1074 outeriter = 0;
1075 inneriter = 0;
1076
1077 for(iidx=0;iidx<4*DIM;iidx++)
1078 {
1079 scratch[iidx] = 0.0;
1080 }
1081
1082 /* Start outer loop over neighborlists */
1083 for(iidx=0; iidx<nri; iidx++)
1084 {
1085 /* Load shift vector for this list */
1086 i_shift_offset = DIM*shiftidx[iidx];
1087
1088 /* Load limits for loop over neighbors */
1089 j_index_start = jindex[iidx];
1090 j_index_end = jindex[iidx+1];
1091
1092 /* Get outer coordinate index */
1093 inr = iinr[iidx];
1094 i_coord_offset = DIM*inr;
1095
1096 /* Load i particle coords and add shift vector */
1097 gmx_mm_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1098 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1099
1100 fix0 = _mm_setzero_ps();
1101 fiy0 = _mm_setzero_ps();
1102 fiz0 = _mm_setzero_ps();
1103 fix1 = _mm_setzero_ps();
1104 fiy1 = _mm_setzero_ps();
1105 fiz1 = _mm_setzero_ps();
1106 fix2 = _mm_setzero_ps();
1107 fiy2 = _mm_setzero_ps();
1108 fiz2 = _mm_setzero_ps();
1109
1110 /* Start inner kernel loop */
1111 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
1112 {
1113
1114 /* Get j neighbor index, and coordinate index */
1115 jnrA = jjnr[jidx];
1116 jnrB = jjnr[jidx+1];
1117 jnrC = jjnr[jidx+2];
1118 jnrD = jjnr[jidx+3];
1119 j_coord_offsetA = DIM*jnrA;
1120 j_coord_offsetB = DIM*jnrB;
1121 j_coord_offsetC = DIM*jnrC;
1122 j_coord_offsetD = DIM*jnrD;
1123
1124 /* load j atom coordinates */
1125 gmx_mm_load_3rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1126 x+j_coord_offsetC,x+j_coord_offsetD,
1127 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1128
1129 /* Calculate displacement vector */
1130 dx00 = _mm_sub_ps(ix0,jx0);
1131 dy00 = _mm_sub_ps(iy0,jy0);
1132 dz00 = _mm_sub_ps(iz0,jz0);
1133 dx01 = _mm_sub_ps(ix0,jx1);
1134 dy01 = _mm_sub_ps(iy0,jy1);
1135 dz01 = _mm_sub_ps(iz0,jz1);
1136 dx02 = _mm_sub_ps(ix0,jx2);
1137 dy02 = _mm_sub_ps(iy0,jy2);
1138 dz02 = _mm_sub_ps(iz0,jz2);
1139 dx10 = _mm_sub_ps(ix1,jx0);
1140 dy10 = _mm_sub_ps(iy1,jy0);
1141 dz10 = _mm_sub_ps(iz1,jz0);
1142 dx11 = _mm_sub_ps(ix1,jx1);
1143 dy11 = _mm_sub_ps(iy1,jy1);
1144 dz11 = _mm_sub_ps(iz1,jz1);
1145 dx12 = _mm_sub_ps(ix1,jx2);
1146 dy12 = _mm_sub_ps(iy1,jy2);
1147 dz12 = _mm_sub_ps(iz1,jz2);
1148 dx20 = _mm_sub_ps(ix2,jx0);
1149 dy20 = _mm_sub_ps(iy2,jy0);
1150 dz20 = _mm_sub_ps(iz2,jz0);
1151 dx21 = _mm_sub_ps(ix2,jx1);
1152 dy21 = _mm_sub_ps(iy2,jy1);
1153 dz21 = _mm_sub_ps(iz2,jz1);
1154 dx22 = _mm_sub_ps(ix2,jx2);
1155 dy22 = _mm_sub_ps(iy2,jy2);
1156 dz22 = _mm_sub_ps(iz2,jz2);
1157
1158 /* Calculate squared distance and things based on it */
1159 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
1160 rsq01 = gmx_mm_calc_rsq_ps(dx01,dy01,dz01);
1161 rsq02 = gmx_mm_calc_rsq_ps(dx02,dy02,dz02);
1162 rsq10 = gmx_mm_calc_rsq_ps(dx10,dy10,dz10);
1163 rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
1164 rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
1165 rsq20 = gmx_mm_calc_rsq_ps(dx20,dy20,dz20);
1166 rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
1167 rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
1168
1169 rinv00 = gmx_mm_invsqrt_ps(rsq00);
1170 rinv01 = gmx_mm_invsqrt_ps(rsq01);
1171 rinv02 = gmx_mm_invsqrt_ps(rsq02);
1172 rinv10 = gmx_mm_invsqrt_ps(rsq10);
1173 rinv11 = gmx_mm_invsqrt_ps(rsq11);
1174 rinv12 = gmx_mm_invsqrt_ps(rsq12);
1175 rinv20 = gmx_mm_invsqrt_ps(rsq20);
1176 rinv21 = gmx_mm_invsqrt_ps(rsq21);
1177 rinv22 = gmx_mm_invsqrt_ps(rsq22);
1178
1179 rinvsq00 = _mm_mul_ps(rinv00,rinv00);
1180 rinvsq01 = _mm_mul_ps(rinv01,rinv01);
1181 rinvsq02 = _mm_mul_ps(rinv02,rinv02);
1182 rinvsq10 = _mm_mul_ps(rinv10,rinv10);
1183 rinvsq11 = _mm_mul_ps(rinv11,rinv11);
1184 rinvsq12 = _mm_mul_ps(rinv12,rinv12);
1185 rinvsq20 = _mm_mul_ps(rinv20,rinv20);
1186 rinvsq21 = _mm_mul_ps(rinv21,rinv21);
1187 rinvsq22 = _mm_mul_ps(rinv22,rinv22);
1188
1189 fjx0 = _mm_setzero_ps();
1190 fjy0 = _mm_setzero_ps();
1191 fjz0 = _mm_setzero_ps();
1192 fjx1 = _mm_setzero_ps();
1193 fjy1 = _mm_setzero_ps();
1194 fjz1 = _mm_setzero_ps();
1195 fjx2 = _mm_setzero_ps();
1196 fjy2 = _mm_setzero_ps();
1197 fjz2 = _mm_setzero_ps();
1198
1199 /**************************
1200 * CALCULATE INTERACTIONS *
1201 **************************/
1202
1203 r00 = _mm_mul_ps(rsq00,rinv00);
1204
1205 /* Calculate table index by multiplying r with table scale and truncate to integer */
1206 rt = _mm_mul_ps(r00,vftabscale);
1207 vfitab = _mm_cvttps_epi32(rt);
1208 #ifdef __XOP__
1209 vfeps = _mm_frcz_ps(rt);
1210 #else
1211 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1212 #endif
1213 twovfeps = _mm_add_ps(vfeps,vfeps);
1214 vfitab = _mm_slli_epi32(vfitab,3);
1215
1216 /* REACTION-FIELD ELECTROSTATICS */
1217 felec = _mm_mul_ps(qq00,_mm_msub_ps(rinv00,rinvsq00,krf2));
1218
1219 /* CUBIC SPLINE TABLE DISPERSION */
1220 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
1221 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
1222 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
1223 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
1224 _MM_TRANSPOSE4_PS(Y,F,G,H);
1225 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
1226 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
1227 fvdw6 = _mm_mul_ps(c6_00,FF);
1228
1229 /* CUBIC SPLINE TABLE REPULSION */
1230 vfitab = _mm_add_epi32(vfitab,ifour);
1231 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
1232 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
1233 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
1234 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
1235 _MM_TRANSPOSE4_PS(Y,F,G,H);
1236 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
1237 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
1238 fvdw12 = _mm_mul_ps(c12_00,FF);
1239 fvdw = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
1240
1241 fscal = _mm_add_ps(felec,fvdw);
1242
1243 /* Update vectorial force */
1244 fix0 = _mm_macc_ps(dx00,fscal,fix0);
1245 fiy0 = _mm_macc_ps(dy00,fscal,fiy0);
1246 fiz0 = _mm_macc_ps(dz00,fscal,fiz0);
1247
1248 fjx0 = _mm_macc_ps(dx00,fscal,fjx0);
1249 fjy0 = _mm_macc_ps(dy00,fscal,fjy0);
1250 fjz0 = _mm_macc_ps(dz00,fscal,fjz0);
1251
1252 /**************************
1253 * CALCULATE INTERACTIONS *
1254 **************************/
1255
1256 /* REACTION-FIELD ELECTROSTATICS */
1257 felec = _mm_mul_ps(qq01,_mm_msub_ps(rinv01,rinvsq01,krf2));
1258
1259 fscal = felec;
1260
1261 /* Update vectorial force */
1262 fix0 = _mm_macc_ps(dx01,fscal,fix0);
1263 fiy0 = _mm_macc_ps(dy01,fscal,fiy0);
1264 fiz0 = _mm_macc_ps(dz01,fscal,fiz0);
1265
1266 fjx1 = _mm_macc_ps(dx01,fscal,fjx1);
1267 fjy1 = _mm_macc_ps(dy01,fscal,fjy1);
1268 fjz1 = _mm_macc_ps(dz01,fscal,fjz1);
1269
1270 /**************************
1271 * CALCULATE INTERACTIONS *
1272 **************************/
1273
1274 /* REACTION-FIELD ELECTROSTATICS */
1275 felec = _mm_mul_ps(qq02,_mm_msub_ps(rinv02,rinvsq02,krf2));
1276
1277 fscal = felec;
1278
1279 /* Update vectorial force */
1280 fix0 = _mm_macc_ps(dx02,fscal,fix0);
1281 fiy0 = _mm_macc_ps(dy02,fscal,fiy0);
1282 fiz0 = _mm_macc_ps(dz02,fscal,fiz0);
1283
1284 fjx2 = _mm_macc_ps(dx02,fscal,fjx2);
1285 fjy2 = _mm_macc_ps(dy02,fscal,fjy2);
1286 fjz2 = _mm_macc_ps(dz02,fscal,fjz2);
1287
1288 /**************************
1289 * CALCULATE INTERACTIONS *
1290 **************************/
1291
1292 /* REACTION-FIELD ELECTROSTATICS */
1293 felec = _mm_mul_ps(qq10,_mm_msub_ps(rinv10,rinvsq10,krf2));
1294
1295 fscal = felec;
1296
1297 /* Update vectorial force */
1298 fix1 = _mm_macc_ps(dx10,fscal,fix1);
1299 fiy1 = _mm_macc_ps(dy10,fscal,fiy1);
1300 fiz1 = _mm_macc_ps(dz10,fscal,fiz1);
1301
1302 fjx0 = _mm_macc_ps(dx10,fscal,fjx0);
1303 fjy0 = _mm_macc_ps(dy10,fscal,fjy0);
1304 fjz0 = _mm_macc_ps(dz10,fscal,fjz0);
1305
1306 /**************************
1307 * CALCULATE INTERACTIONS *
1308 **************************/
1309
1310 /* REACTION-FIELD ELECTROSTATICS */
1311 felec = _mm_mul_ps(qq11,_mm_msub_ps(rinv11,rinvsq11,krf2));
1312
1313 fscal = felec;
1314
1315 /* Update vectorial force */
1316 fix1 = _mm_macc_ps(dx11,fscal,fix1);
1317 fiy1 = _mm_macc_ps(dy11,fscal,fiy1);
1318 fiz1 = _mm_macc_ps(dz11,fscal,fiz1);
1319
1320 fjx1 = _mm_macc_ps(dx11,fscal,fjx1);
1321 fjy1 = _mm_macc_ps(dy11,fscal,fjy1);
1322 fjz1 = _mm_macc_ps(dz11,fscal,fjz1);
1323
1324 /**************************
1325 * CALCULATE INTERACTIONS *
1326 **************************/
1327
1328 /* REACTION-FIELD ELECTROSTATICS */
1329 felec = _mm_mul_ps(qq12,_mm_msub_ps(rinv12,rinvsq12,krf2));
1330
1331 fscal = felec;
1332
1333 /* Update vectorial force */
1334 fix1 = _mm_macc_ps(dx12,fscal,fix1);
1335 fiy1 = _mm_macc_ps(dy12,fscal,fiy1);
1336 fiz1 = _mm_macc_ps(dz12,fscal,fiz1);
1337
1338 fjx2 = _mm_macc_ps(dx12,fscal,fjx2);
1339 fjy2 = _mm_macc_ps(dy12,fscal,fjy2);
1340 fjz2 = _mm_macc_ps(dz12,fscal,fjz2);
1341
1342 /**************************
1343 * CALCULATE INTERACTIONS *
1344 **************************/
1345
1346 /* REACTION-FIELD ELECTROSTATICS */
1347 felec = _mm_mul_ps(qq20,_mm_msub_ps(rinv20,rinvsq20,krf2));
1348
1349 fscal = felec;
1350
1351 /* Update vectorial force */
1352 fix2 = _mm_macc_ps(dx20,fscal,fix2);
1353 fiy2 = _mm_macc_ps(dy20,fscal,fiy2);
1354 fiz2 = _mm_macc_ps(dz20,fscal,fiz2);
1355
1356 fjx0 = _mm_macc_ps(dx20,fscal,fjx0);
1357 fjy0 = _mm_macc_ps(dy20,fscal,fjy0);
1358 fjz0 = _mm_macc_ps(dz20,fscal,fjz0);
1359
1360 /**************************
1361 * CALCULATE INTERACTIONS *
1362 **************************/
1363
1364 /* REACTION-FIELD ELECTROSTATICS */
1365 felec = _mm_mul_ps(qq21,_mm_msub_ps(rinv21,rinvsq21,krf2));
1366
1367 fscal = felec;
1368
1369 /* Update vectorial force */
1370 fix2 = _mm_macc_ps(dx21,fscal,fix2);
1371 fiy2 = _mm_macc_ps(dy21,fscal,fiy2);
1372 fiz2 = _mm_macc_ps(dz21,fscal,fiz2);
1373
1374 fjx1 = _mm_macc_ps(dx21,fscal,fjx1);
1375 fjy1 = _mm_macc_ps(dy21,fscal,fjy1);
1376 fjz1 = _mm_macc_ps(dz21,fscal,fjz1);
1377
1378 /**************************
1379 * CALCULATE INTERACTIONS *
1380 **************************/
1381
1382 /* REACTION-FIELD ELECTROSTATICS */
1383 felec = _mm_mul_ps(qq22,_mm_msub_ps(rinv22,rinvsq22,krf2));
1384
1385 fscal = felec;
1386
1387 /* Update vectorial force */
1388 fix2 = _mm_macc_ps(dx22,fscal,fix2);
1389 fiy2 = _mm_macc_ps(dy22,fscal,fiy2);
1390 fiz2 = _mm_macc_ps(dz22,fscal,fiz2);
1391
1392 fjx2 = _mm_macc_ps(dx22,fscal,fjx2);
1393 fjy2 = _mm_macc_ps(dy22,fscal,fjy2);
1394 fjz2 = _mm_macc_ps(dz22,fscal,fjz2);
1395
1396 fjptrA = f+j_coord_offsetA;
1397 fjptrB = f+j_coord_offsetB;
1398 fjptrC = f+j_coord_offsetC;
1399 fjptrD = f+j_coord_offsetD;
1400
1401 gmx_mm_decrement_3rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
1402 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1403
1404 /* Inner loop uses 297 flops */
1405 }
1406
1407 if(jidx<j_index_end)
1408 {
1409
1410 /* Get j neighbor index, and coordinate index */
1411 jnrlistA = jjnr[jidx];
1412 jnrlistB = jjnr[jidx+1];
1413 jnrlistC = jjnr[jidx+2];
1414 jnrlistD = jjnr[jidx+3];
1415 /* Sign of each element will be negative for non-real atoms.
1416 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1417 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1418 */
1419 dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
1420 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
1421 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
1422 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
1423 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
1424 j_coord_offsetA = DIM*jnrA;
1425 j_coord_offsetB = DIM*jnrB;
1426 j_coord_offsetC = DIM*jnrC;
1427 j_coord_offsetD = DIM*jnrD;
1428
1429 /* load j atom coordinates */
1430 gmx_mm_load_3rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1431 x+j_coord_offsetC,x+j_coord_offsetD,
1432 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1433
1434 /* Calculate displacement vector */
1435 dx00 = _mm_sub_ps(ix0,jx0);
1436 dy00 = _mm_sub_ps(iy0,jy0);
1437 dz00 = _mm_sub_ps(iz0,jz0);
1438 dx01 = _mm_sub_ps(ix0,jx1);
1439 dy01 = _mm_sub_ps(iy0,jy1);
1440 dz01 = _mm_sub_ps(iz0,jz1);
1441 dx02 = _mm_sub_ps(ix0,jx2);
1442 dy02 = _mm_sub_ps(iy0,jy2);
1443 dz02 = _mm_sub_ps(iz0,jz2);
1444 dx10 = _mm_sub_ps(ix1,jx0);
1445 dy10 = _mm_sub_ps(iy1,jy0);
1446 dz10 = _mm_sub_ps(iz1,jz0);
1447 dx11 = _mm_sub_ps(ix1,jx1);
1448 dy11 = _mm_sub_ps(iy1,jy1);
1449 dz11 = _mm_sub_ps(iz1,jz1);
1450 dx12 = _mm_sub_ps(ix1,jx2);
1451 dy12 = _mm_sub_ps(iy1,jy2);
1452 dz12 = _mm_sub_ps(iz1,jz2);
1453 dx20 = _mm_sub_ps(ix2,jx0);
1454 dy20 = _mm_sub_ps(iy2,jy0);
1455 dz20 = _mm_sub_ps(iz2,jz0);
1456 dx21 = _mm_sub_ps(ix2,jx1);
1457 dy21 = _mm_sub_ps(iy2,jy1);
1458 dz21 = _mm_sub_ps(iz2,jz1);
1459 dx22 = _mm_sub_ps(ix2,jx2);
1460 dy22 = _mm_sub_ps(iy2,jy2);
1461 dz22 = _mm_sub_ps(iz2,jz2);
1462
1463 /* Calculate squared distance and things based on it */
1464 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
1465 rsq01 = gmx_mm_calc_rsq_ps(dx01,dy01,dz01);
1466 rsq02 = gmx_mm_calc_rsq_ps(dx02,dy02,dz02);
1467 rsq10 = gmx_mm_calc_rsq_ps(dx10,dy10,dz10);
1468 rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
1469 rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
1470 rsq20 = gmx_mm_calc_rsq_ps(dx20,dy20,dz20);
1471 rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
1472 rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
1473
1474 rinv00 = gmx_mm_invsqrt_ps(rsq00);
1475 rinv01 = gmx_mm_invsqrt_ps(rsq01);
1476 rinv02 = gmx_mm_invsqrt_ps(rsq02);
1477 rinv10 = gmx_mm_invsqrt_ps(rsq10);
1478 rinv11 = gmx_mm_invsqrt_ps(rsq11);
1479 rinv12 = gmx_mm_invsqrt_ps(rsq12);
1480 rinv20 = gmx_mm_invsqrt_ps(rsq20);
1481 rinv21 = gmx_mm_invsqrt_ps(rsq21);
1482 rinv22 = gmx_mm_invsqrt_ps(rsq22);
1483
1484 rinvsq00 = _mm_mul_ps(rinv00,rinv00);
1485 rinvsq01 = _mm_mul_ps(rinv01,rinv01);
1486 rinvsq02 = _mm_mul_ps(rinv02,rinv02);
1487 rinvsq10 = _mm_mul_ps(rinv10,rinv10);
1488 rinvsq11 = _mm_mul_ps(rinv11,rinv11);
1489 rinvsq12 = _mm_mul_ps(rinv12,rinv12);
1490 rinvsq20 = _mm_mul_ps(rinv20,rinv20);
1491 rinvsq21 = _mm_mul_ps(rinv21,rinv21);
1492 rinvsq22 = _mm_mul_ps(rinv22,rinv22);
1493
1494 fjx0 = _mm_setzero_ps();
1495 fjy0 = _mm_setzero_ps();
1496 fjz0 = _mm_setzero_ps();
1497 fjx1 = _mm_setzero_ps();
1498 fjy1 = _mm_setzero_ps();
1499 fjz1 = _mm_setzero_ps();
1500 fjx2 = _mm_setzero_ps();
1501 fjy2 = _mm_setzero_ps();
1502 fjz2 = _mm_setzero_ps();
1503
1504 /**************************
1505 * CALCULATE INTERACTIONS *
1506 **************************/
1507
1508 r00 = _mm_mul_ps(rsq00,rinv00);
1509 r00 = _mm_andnot_ps(dummy_mask,r00);
1510
1511 /* Calculate table index by multiplying r with table scale and truncate to integer */
1512 rt = _mm_mul_ps(r00,vftabscale);
1513 vfitab = _mm_cvttps_epi32(rt);
1514 #ifdef __XOP__
1515 vfeps = _mm_frcz_ps(rt);
1516 #else
1517 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1518 #endif
1519 twovfeps = _mm_add_ps(vfeps,vfeps);
1520 vfitab = _mm_slli_epi32(vfitab,3);
1521
1522 /* REACTION-FIELD ELECTROSTATICS */
1523 felec = _mm_mul_ps(qq00,_mm_msub_ps(rinv00,rinvsq00,krf2));
1524
1525 /* CUBIC SPLINE TABLE DISPERSION */
1526 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
1527 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
1528 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
1529 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
1530 _MM_TRANSPOSE4_PS(Y,F,G,H);
1531 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
1532 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
1533 fvdw6 = _mm_mul_ps(c6_00,FF);
1534
1535 /* CUBIC SPLINE TABLE REPULSION */
1536 vfitab = _mm_add_epi32(vfitab,ifour);
1537 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
1538 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
1539 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
1540 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
1541 _MM_TRANSPOSE4_PS(Y,F,G,H);
1542 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
1543 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
1544 fvdw12 = _mm_mul_ps(c12_00,FF);
1545 fvdw = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
1546
1547 fscal = _mm_add_ps(felec,fvdw);
1548
1549 fscal = _mm_andnot_ps(dummy_mask,fscal);
1550
1551 /* Update vectorial force */
1552 fix0 = _mm_macc_ps(dx00,fscal,fix0);
1553 fiy0 = _mm_macc_ps(dy00,fscal,fiy0);
1554 fiz0 = _mm_macc_ps(dz00,fscal,fiz0);
1555
1556 fjx0 = _mm_macc_ps(dx00,fscal,fjx0);
1557 fjy0 = _mm_macc_ps(dy00,fscal,fjy0);
1558 fjz0 = _mm_macc_ps(dz00,fscal,fjz0);
1559
1560 /**************************
1561 * CALCULATE INTERACTIONS *
1562 **************************/
1563
1564 /* REACTION-FIELD ELECTROSTATICS */
1565 felec = _mm_mul_ps(qq01,_mm_msub_ps(rinv01,rinvsq01,krf2));
1566
1567 fscal = felec;
1568
1569 fscal = _mm_andnot_ps(dummy_mask,fscal);
1570
1571 /* Update vectorial force */
1572 fix0 = _mm_macc_ps(dx01,fscal,fix0);
1573 fiy0 = _mm_macc_ps(dy01,fscal,fiy0);
1574 fiz0 = _mm_macc_ps(dz01,fscal,fiz0);
1575
1576 fjx1 = _mm_macc_ps(dx01,fscal,fjx1);
1577 fjy1 = _mm_macc_ps(dy01,fscal,fjy1);
1578 fjz1 = _mm_macc_ps(dz01,fscal,fjz1);
1579
1580 /**************************
1581 * CALCULATE INTERACTIONS *
1582 **************************/
1583
1584 /* REACTION-FIELD ELECTROSTATICS */
1585 felec = _mm_mul_ps(qq02,_mm_msub_ps(rinv02,rinvsq02,krf2));
1586
1587 fscal = felec;
1588
1589 fscal = _mm_andnot_ps(dummy_mask,fscal);
1590
1591 /* Update vectorial force */
1592 fix0 = _mm_macc_ps(dx02,fscal,fix0);
1593 fiy0 = _mm_macc_ps(dy02,fscal,fiy0);
1594 fiz0 = _mm_macc_ps(dz02,fscal,fiz0);
1595
1596 fjx2 = _mm_macc_ps(dx02,fscal,fjx2);
1597 fjy2 = _mm_macc_ps(dy02,fscal,fjy2);
1598 fjz2 = _mm_macc_ps(dz02,fscal,fjz2);
1599
1600 /**************************
1601 * CALCULATE INTERACTIONS *
1602 **************************/
1603
1604 /* REACTION-FIELD ELECTROSTATICS */
1605 felec = _mm_mul_ps(qq10,_mm_msub_ps(rinv10,rinvsq10,krf2));
1606
1607 fscal = felec;
1608
1609 fscal = _mm_andnot_ps(dummy_mask,fscal);
1610
1611 /* Update vectorial force */
1612 fix1 = _mm_macc_ps(dx10,fscal,fix1);
1613 fiy1 = _mm_macc_ps(dy10,fscal,fiy1);
1614 fiz1 = _mm_macc_ps(dz10,fscal,fiz1);
1615
1616 fjx0 = _mm_macc_ps(dx10,fscal,fjx0);
1617 fjy0 = _mm_macc_ps(dy10,fscal,fjy0);
1618 fjz0 = _mm_macc_ps(dz10,fscal,fjz0);
1619
1620 /**************************
1621 * CALCULATE INTERACTIONS *
1622 **************************/
1623
1624 /* REACTION-FIELD ELECTROSTATICS */
1625 felec = _mm_mul_ps(qq11,_mm_msub_ps(rinv11,rinvsq11,krf2));
1626
1627 fscal = felec;
1628
1629 fscal = _mm_andnot_ps(dummy_mask,fscal);
1630
1631 /* Update vectorial force */
1632 fix1 = _mm_macc_ps(dx11,fscal,fix1);
1633 fiy1 = _mm_macc_ps(dy11,fscal,fiy1);
1634 fiz1 = _mm_macc_ps(dz11,fscal,fiz1);
1635
1636 fjx1 = _mm_macc_ps(dx11,fscal,fjx1);
1637 fjy1 = _mm_macc_ps(dy11,fscal,fjy1);
1638 fjz1 = _mm_macc_ps(dz11,fscal,fjz1);
1639
1640 /**************************
1641 * CALCULATE INTERACTIONS *
1642 **************************/
1643
1644 /* REACTION-FIELD ELECTROSTATICS */
1645 felec = _mm_mul_ps(qq12,_mm_msub_ps(rinv12,rinvsq12,krf2));
1646
1647 fscal = felec;
1648
1649 fscal = _mm_andnot_ps(dummy_mask,fscal);
1650
1651 /* Update vectorial force */
1652 fix1 = _mm_macc_ps(dx12,fscal,fix1);
1653 fiy1 = _mm_macc_ps(dy12,fscal,fiy1);
1654 fiz1 = _mm_macc_ps(dz12,fscal,fiz1);
1655
1656 fjx2 = _mm_macc_ps(dx12,fscal,fjx2);
1657 fjy2 = _mm_macc_ps(dy12,fscal,fjy2);
1658 fjz2 = _mm_macc_ps(dz12,fscal,fjz2);
1659
1660 /**************************
1661 * CALCULATE INTERACTIONS *
1662 **************************/
1663
1664 /* REACTION-FIELD ELECTROSTATICS */
1665 felec = _mm_mul_ps(qq20,_mm_msub_ps(rinv20,rinvsq20,krf2));
1666
1667 fscal = felec;
1668
1669 fscal = _mm_andnot_ps(dummy_mask,fscal);
1670
1671 /* Update vectorial force */
1672 fix2 = _mm_macc_ps(dx20,fscal,fix2);
1673 fiy2 = _mm_macc_ps(dy20,fscal,fiy2);
1674 fiz2 = _mm_macc_ps(dz20,fscal,fiz2);
1675
1676 fjx0 = _mm_macc_ps(dx20,fscal,fjx0);
1677 fjy0 = _mm_macc_ps(dy20,fscal,fjy0);
1678 fjz0 = _mm_macc_ps(dz20,fscal,fjz0);
1679
1680 /**************************
1681 * CALCULATE INTERACTIONS *
1682 **************************/
1683
1684 /* REACTION-FIELD ELECTROSTATICS */
1685 felec = _mm_mul_ps(qq21,_mm_msub_ps(rinv21,rinvsq21,krf2));
1686
1687 fscal = felec;
1688
1689 fscal = _mm_andnot_ps(dummy_mask,fscal);
1690
1691 /* Update vectorial force */
1692 fix2 = _mm_macc_ps(dx21,fscal,fix2);
1693 fiy2 = _mm_macc_ps(dy21,fscal,fiy2);
1694 fiz2 = _mm_macc_ps(dz21,fscal,fiz2);
1695
1696 fjx1 = _mm_macc_ps(dx21,fscal,fjx1);
1697 fjy1 = _mm_macc_ps(dy21,fscal,fjy1);
1698 fjz1 = _mm_macc_ps(dz21,fscal,fjz1);
1699
1700 /**************************
1701 * CALCULATE INTERACTIONS *
1702 **************************/
1703
1704 /* REACTION-FIELD ELECTROSTATICS */
1705 felec = _mm_mul_ps(qq22,_mm_msub_ps(rinv22,rinvsq22,krf2));
1706
1707 fscal = felec;
1708
1709 fscal = _mm_andnot_ps(dummy_mask,fscal);
1710
1711 /* Update vectorial force */
1712 fix2 = _mm_macc_ps(dx22,fscal,fix2);
1713 fiy2 = _mm_macc_ps(dy22,fscal,fiy2);
1714 fiz2 = _mm_macc_ps(dz22,fscal,fiz2);
1715
1716 fjx2 = _mm_macc_ps(dx22,fscal,fjx2);
1717 fjy2 = _mm_macc_ps(dy22,fscal,fjy2);
1718 fjz2 = _mm_macc_ps(dz22,fscal,fjz2);
1719
1720 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1721 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1722 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1723 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1724
1725 gmx_mm_decrement_3rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
1726 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1727
1728 /* Inner loop uses 298 flops */
1729 }
1730
1731 /* End of innermost loop */
1732
1733 gmx_mm_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1734 f+i_coord_offset,fshift+i_shift_offset);
1735
1736 /* Increment number of inner iterations */
1737 inneriter += j_index_end - j_index_start;
1738
1739 /* Outer loop uses 18 flops */
1740 }
1741
1742 /* Increment number of outer iterations */
1743 outeriter += nri;
1744
1745 /* Update outer/inner flops */
1746
1747 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_F,outeriter*18 + inneriter*298);
1748 }
The ultimate molecular dynamics simulation package