?  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_double / nb_kernel_ElecEwSh_VdwLJSh_GeomW4W4_avx_128_fma_double.c
blob565e87e473c25de030222c67bdf6036c3937d997
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_double 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_double.h"
48 #include "kernelutil_x86_avx_128_fma_double.h"
49
50 /*
51 * Gromacs nonbonded kernel: nb_kernel_ElecEwSh_VdwLJSh_GeomW4W4_VF_avx_128_fma_double
52 * Electrostatics interaction: Ewald
53 * VdW interaction: LennardJones
54 * Geometry: Water4-Water4
55 * Calculate force/pot: PotentialAndForce
56 */
57 void
58 nb_kernel_ElecEwSh_VdwLJSh_GeomW4W4_VF_avx_128_fma_double
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 refer to j loop unrolling done with SSE double precision, e.g. for the two 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;
75 int j_coord_offsetA,j_coord_offsetB;
76 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
77 real rcutoff_scalar;
78 real *shiftvec,*fshift,*x,*f;
79 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
80 int vdwioffset0;
81 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
82 int vdwioffset1;
83 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
84 int vdwioffset2;
85 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
86 int vdwioffset3;
87 __m128d ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
88 int vdwjidx0A,vdwjidx0B;
89 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
90 int vdwjidx1A,vdwjidx1B;
91 __m128d jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
92 int vdwjidx2A,vdwjidx2B;
93 __m128d jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
94 int vdwjidx3A,vdwjidx3B;
95 __m128d jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
96 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
97 __m128d dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
98 __m128d dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
99 __m128d dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
100 __m128d dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
101 __m128d dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
102 __m128d dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
103 __m128d dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
104 __m128d dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
105 __m128d dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
106 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
107 real *charge;
108 int nvdwtype;
109 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
110 int *vdwtype;
111 real *vdwparam;
112 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
113 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
114 __m128i ewitab;
115 __m128d ewtabscale,eweps,twoeweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
116 real *ewtab;
117 __m128d dummy_mask,cutoff_mask;
118 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
119 __m128d one = _mm_set1_pd(1.0);
120 __m128d two = _mm_set1_pd(2.0);
121 x = xx[0];
122 f = ff[0];
123
124 nri = nlist->nri;
125 iinr = nlist->iinr;
126 jindex = nlist->jindex;
127 jjnr = nlist->jjnr;
128 shiftidx = nlist->shift;
129 gid = nlist->gid;
130 shiftvec = fr->shift_vec[0];
131 fshift = fr->fshift[0];
132 facel = _mm_set1_pd(fr->epsfac);
133 charge = mdatoms->chargeA;
134 nvdwtype = fr->ntype;
135 vdwparam = fr->nbfp;
136 vdwtype = mdatoms->typeA;
137
138 sh_ewald = _mm_set1_pd(fr->ic->sh_ewald);
139 ewtab = fr->ic->tabq_coul_FDV0;
140 ewtabscale = _mm_set1_pd(fr->ic->tabq_scale);
141 ewtabhalfspace = _mm_set1_pd(0.5/fr->ic->tabq_scale);
142
143 /* Setup water-specific parameters */
144 inr = nlist->iinr[0];
145 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
146 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
147 iq3 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
148 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
149
150 jq1 = _mm_set1_pd(charge[inr+1]);
151 jq2 = _mm_set1_pd(charge[inr+2]);
152 jq3 = _mm_set1_pd(charge[inr+3]);
153 vdwjidx0A = 2*vdwtype[inr+0];
154 c6_00 = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
155 c12_00 = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
156 qq11 = _mm_mul_pd(iq1,jq1);
157 qq12 = _mm_mul_pd(iq1,jq2);
158 qq13 = _mm_mul_pd(iq1,jq3);
159 qq21 = _mm_mul_pd(iq2,jq1);
160 qq22 = _mm_mul_pd(iq2,jq2);
161 qq23 = _mm_mul_pd(iq2,jq3);
162 qq31 = _mm_mul_pd(iq3,jq1);
163 qq32 = _mm_mul_pd(iq3,jq2);
164 qq33 = _mm_mul_pd(iq3,jq3);
165
166 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
167 rcutoff_scalar = fr->rcoulomb;
168 rcutoff = _mm_set1_pd(rcutoff_scalar);
169 rcutoff2 = _mm_mul_pd(rcutoff,rcutoff);
170
171 sh_vdw_invrcut6 = _mm_set1_pd(fr->ic->sh_invrc6);
172 rvdw = _mm_set1_pd(fr->rvdw);
173
174 /* Avoid stupid compiler warnings */
175 jnrA = jnrB = 0;
176 j_coord_offsetA = 0;
177 j_coord_offsetB = 0;
178
179 outeriter = 0;
180 inneriter = 0;
181
182 /* Start outer loop over neighborlists */
183 for(iidx=0; iidx<nri; iidx++)
184 {
185 /* Load shift vector for this list */
186 i_shift_offset = DIM*shiftidx[iidx];
187
188 /* Load limits for loop over neighbors */
189 j_index_start = jindex[iidx];
190 j_index_end = jindex[iidx+1];
191
192 /* Get outer coordinate index */
193 inr = iinr[iidx];
194 i_coord_offset = DIM*inr;
195
196 /* Load i particle coords and add shift vector */
197 gmx_mm_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
198 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
199
200 fix0 = _mm_setzero_pd();
201 fiy0 = _mm_setzero_pd();
202 fiz0 = _mm_setzero_pd();
203 fix1 = _mm_setzero_pd();
204 fiy1 = _mm_setzero_pd();
205 fiz1 = _mm_setzero_pd();
206 fix2 = _mm_setzero_pd();
207 fiy2 = _mm_setzero_pd();
208 fiz2 = _mm_setzero_pd();
209 fix3 = _mm_setzero_pd();
210 fiy3 = _mm_setzero_pd();
211 fiz3 = _mm_setzero_pd();
212
213 /* Reset potential sums */
214 velecsum = _mm_setzero_pd();
215 vvdwsum = _mm_setzero_pd();
216
217 /* Start inner kernel loop */
218 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
219 {
220
221 /* Get j neighbor index, and coordinate index */
222 jnrA = jjnr[jidx];
223 jnrB = jjnr[jidx+1];
224 j_coord_offsetA = DIM*jnrA;
225 j_coord_offsetB = DIM*jnrB;
226
227 /* load j atom coordinates */
228 gmx_mm_load_4rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
229 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
230 &jy2,&jz2,&jx3,&jy3,&jz3);
231
232 /* Calculate displacement vector */
233 dx00 = _mm_sub_pd(ix0,jx0);
234 dy00 = _mm_sub_pd(iy0,jy0);
235 dz00 = _mm_sub_pd(iz0,jz0);
236 dx11 = _mm_sub_pd(ix1,jx1);
237 dy11 = _mm_sub_pd(iy1,jy1);
238 dz11 = _mm_sub_pd(iz1,jz1);
239 dx12 = _mm_sub_pd(ix1,jx2);
240 dy12 = _mm_sub_pd(iy1,jy2);
241 dz12 = _mm_sub_pd(iz1,jz2);
242 dx13 = _mm_sub_pd(ix1,jx3);
243 dy13 = _mm_sub_pd(iy1,jy3);
244 dz13 = _mm_sub_pd(iz1,jz3);
245 dx21 = _mm_sub_pd(ix2,jx1);
246 dy21 = _mm_sub_pd(iy2,jy1);
247 dz21 = _mm_sub_pd(iz2,jz1);
248 dx22 = _mm_sub_pd(ix2,jx2);
249 dy22 = _mm_sub_pd(iy2,jy2);
250 dz22 = _mm_sub_pd(iz2,jz2);
251 dx23 = _mm_sub_pd(ix2,jx3);
252 dy23 = _mm_sub_pd(iy2,jy3);
253 dz23 = _mm_sub_pd(iz2,jz3);
254 dx31 = _mm_sub_pd(ix3,jx1);
255 dy31 = _mm_sub_pd(iy3,jy1);
256 dz31 = _mm_sub_pd(iz3,jz1);
257 dx32 = _mm_sub_pd(ix3,jx2);
258 dy32 = _mm_sub_pd(iy3,jy2);
259 dz32 = _mm_sub_pd(iz3,jz2);
260 dx33 = _mm_sub_pd(ix3,jx3);
261 dy33 = _mm_sub_pd(iy3,jy3);
262 dz33 = _mm_sub_pd(iz3,jz3);
263
264 /* Calculate squared distance and things based on it */
265 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
266 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
267 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
268 rsq13 = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
269 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
270 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
271 rsq23 = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
272 rsq31 = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
273 rsq32 = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
274 rsq33 = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
275
276 rinv11 = gmx_mm_invsqrt_pd(rsq11);
277 rinv12 = gmx_mm_invsqrt_pd(rsq12);
278 rinv13 = gmx_mm_invsqrt_pd(rsq13);
279 rinv21 = gmx_mm_invsqrt_pd(rsq21);
280 rinv22 = gmx_mm_invsqrt_pd(rsq22);
281 rinv23 = gmx_mm_invsqrt_pd(rsq23);
282 rinv31 = gmx_mm_invsqrt_pd(rsq31);
283 rinv32 = gmx_mm_invsqrt_pd(rsq32);
284 rinv33 = gmx_mm_invsqrt_pd(rsq33);
285
286 rinvsq00 = gmx_mm_inv_pd(rsq00);
287 rinvsq11 = _mm_mul_pd(rinv11,rinv11);
288 rinvsq12 = _mm_mul_pd(rinv12,rinv12);
289 rinvsq13 = _mm_mul_pd(rinv13,rinv13);
290 rinvsq21 = _mm_mul_pd(rinv21,rinv21);
291 rinvsq22 = _mm_mul_pd(rinv22,rinv22);
292 rinvsq23 = _mm_mul_pd(rinv23,rinv23);
293 rinvsq31 = _mm_mul_pd(rinv31,rinv31);
294 rinvsq32 = _mm_mul_pd(rinv32,rinv32);
295 rinvsq33 = _mm_mul_pd(rinv33,rinv33);
296
297 fjx0 = _mm_setzero_pd();
298 fjy0 = _mm_setzero_pd();
299 fjz0 = _mm_setzero_pd();
300 fjx1 = _mm_setzero_pd();
301 fjy1 = _mm_setzero_pd();
302 fjz1 = _mm_setzero_pd();
303 fjx2 = _mm_setzero_pd();
304 fjy2 = _mm_setzero_pd();
305 fjz2 = _mm_setzero_pd();
306 fjx3 = _mm_setzero_pd();
307 fjy3 = _mm_setzero_pd();
308 fjz3 = _mm_setzero_pd();
309
310 /**************************
311 * CALCULATE INTERACTIONS *
312 **************************/
313
314 if (gmx_mm_any_lt(rsq00,rcutoff2))
315 {
316
317 /* LENNARD-JONES DISPERSION/REPULSION */
318
319 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
320 vvdw6 = _mm_mul_pd(c6_00,rinvsix);
321 vvdw12 = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
322 vvdw = _mm_msub_pd(_mm_nmacc_pd(c12_00,_mm_mul_pd(sh_vdw_invrcut6,sh_vdw_invrcut6),vvdw12),one_twelfth,
323 _mm_mul_pd(_mm_nmacc_pd( c6_00,sh_vdw_invrcut6,vvdw6),one_sixth));
324 fvdw = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
325
326 cutoff_mask = _mm_cmplt_pd(rsq00,rcutoff2);
327
328 /* Update potential sum for this i atom from the interaction with this j atom. */
329 vvdw = _mm_and_pd(vvdw,cutoff_mask);
330 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
331
332 fscal = fvdw;
333
334 fscal = _mm_and_pd(fscal,cutoff_mask);
335
336 /* Update vectorial force */
337 fix0 = _mm_macc_pd(dx00,fscal,fix0);
338 fiy0 = _mm_macc_pd(dy00,fscal,fiy0);
339 fiz0 = _mm_macc_pd(dz00,fscal,fiz0);
340
341 fjx0 = _mm_macc_pd(dx00,fscal,fjx0);
342 fjy0 = _mm_macc_pd(dy00,fscal,fjy0);
343 fjz0 = _mm_macc_pd(dz00,fscal,fjz0);
344
345 }
346
347 /**************************
348 * CALCULATE INTERACTIONS *
349 **************************/
350
351 if (gmx_mm_any_lt(rsq11,rcutoff2))
352 {
353
354 r11 = _mm_mul_pd(rsq11,rinv11);
355
356 /* EWALD ELECTROSTATICS */
357
358 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
359 ewrt = _mm_mul_pd(r11,ewtabscale);
360 ewitab = _mm_cvttpd_epi32(ewrt);
361 #ifdef __XOP__
362 eweps = _mm_frcz_pd(ewrt);
363 #else
364 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
365 #endif
366 twoeweps = _mm_add_pd(eweps,eweps);
367 ewitab = _mm_slli_epi32(ewitab,2);
368 ewtabF = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
369 ewtabD = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
370 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
371 ewtabV = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
372 ewtabFn = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
373 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
374 felec = _mm_macc_pd(eweps,ewtabD,ewtabF);
375 velec = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
376 velec = _mm_mul_pd(qq11,_mm_sub_pd(_mm_sub_pd(rinv11,sh_ewald),velec));
377 felec = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
378
379 cutoff_mask = _mm_cmplt_pd(rsq11,rcutoff2);
380
381 /* Update potential sum for this i atom from the interaction with this j atom. */
382 velec = _mm_and_pd(velec,cutoff_mask);
383 velecsum = _mm_add_pd(velecsum,velec);
384
385 fscal = felec;
386
387 fscal = _mm_and_pd(fscal,cutoff_mask);
388
389 /* Update vectorial force */
390 fix1 = _mm_macc_pd(dx11,fscal,fix1);
391 fiy1 = _mm_macc_pd(dy11,fscal,fiy1);
392 fiz1 = _mm_macc_pd(dz11,fscal,fiz1);
393
394 fjx1 = _mm_macc_pd(dx11,fscal,fjx1);
395 fjy1 = _mm_macc_pd(dy11,fscal,fjy1);
396 fjz1 = _mm_macc_pd(dz11,fscal,fjz1);
397
398 }
399
400 /**************************
401 * CALCULATE INTERACTIONS *
402 **************************/
403
404 if (gmx_mm_any_lt(rsq12,rcutoff2))
405 {
406
407 r12 = _mm_mul_pd(rsq12,rinv12);
408
409 /* EWALD ELECTROSTATICS */
410
411 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
412 ewrt = _mm_mul_pd(r12,ewtabscale);
413 ewitab = _mm_cvttpd_epi32(ewrt);
414 #ifdef __XOP__
415 eweps = _mm_frcz_pd(ewrt);
416 #else
417 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
418 #endif
419 twoeweps = _mm_add_pd(eweps,eweps);
420 ewitab = _mm_slli_epi32(ewitab,2);
421 ewtabF = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
422 ewtabD = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
423 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
424 ewtabV = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
425 ewtabFn = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
426 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
427 felec = _mm_macc_pd(eweps,ewtabD,ewtabF);
428 velec = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
429 velec = _mm_mul_pd(qq12,_mm_sub_pd(_mm_sub_pd(rinv12,sh_ewald),velec));
430 felec = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
431
432 cutoff_mask = _mm_cmplt_pd(rsq12,rcutoff2);
433
434 /* Update potential sum for this i atom from the interaction with this j atom. */
435 velec = _mm_and_pd(velec,cutoff_mask);
436 velecsum = _mm_add_pd(velecsum,velec);
437
438 fscal = felec;
439
440 fscal = _mm_and_pd(fscal,cutoff_mask);
441
442 /* Update vectorial force */
443 fix1 = _mm_macc_pd(dx12,fscal,fix1);
444 fiy1 = _mm_macc_pd(dy12,fscal,fiy1);
445 fiz1 = _mm_macc_pd(dz12,fscal,fiz1);
446
447 fjx2 = _mm_macc_pd(dx12,fscal,fjx2);
448 fjy2 = _mm_macc_pd(dy12,fscal,fjy2);
449 fjz2 = _mm_macc_pd(dz12,fscal,fjz2);
450
451 }
452
453 /**************************
454 * CALCULATE INTERACTIONS *
455 **************************/
456
457 if (gmx_mm_any_lt(rsq13,rcutoff2))
458 {
459
460 r13 = _mm_mul_pd(rsq13,rinv13);
461
462 /* EWALD ELECTROSTATICS */
463
464 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
465 ewrt = _mm_mul_pd(r13,ewtabscale);
466 ewitab = _mm_cvttpd_epi32(ewrt);
467 #ifdef __XOP__
468 eweps = _mm_frcz_pd(ewrt);
469 #else
470 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
471 #endif
472 twoeweps = _mm_add_pd(eweps,eweps);
473 ewitab = _mm_slli_epi32(ewitab,2);
474 ewtabF = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
475 ewtabD = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
476 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
477 ewtabV = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
478 ewtabFn = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
479 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
480 felec = _mm_macc_pd(eweps,ewtabD,ewtabF);
481 velec = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
482 velec = _mm_mul_pd(qq13,_mm_sub_pd(_mm_sub_pd(rinv13,sh_ewald),velec));
483 felec = _mm_mul_pd(_mm_mul_pd(qq13,rinv13),_mm_sub_pd(rinvsq13,felec));
484
485 cutoff_mask = _mm_cmplt_pd(rsq13,rcutoff2);
486
487 /* Update potential sum for this i atom from the interaction with this j atom. */
488 velec = _mm_and_pd(velec,cutoff_mask);
489 velecsum = _mm_add_pd(velecsum,velec);
490
491 fscal = felec;
492
493 fscal = _mm_and_pd(fscal,cutoff_mask);
494
495 /* Update vectorial force */
496 fix1 = _mm_macc_pd(dx13,fscal,fix1);
497 fiy1 = _mm_macc_pd(dy13,fscal,fiy1);
498 fiz1 = _mm_macc_pd(dz13,fscal,fiz1);
499
500 fjx3 = _mm_macc_pd(dx13,fscal,fjx3);
501 fjy3 = _mm_macc_pd(dy13,fscal,fjy3);
502 fjz3 = _mm_macc_pd(dz13,fscal,fjz3);
503
504 }
505
506 /**************************
507 * CALCULATE INTERACTIONS *
508 **************************/
509
510 if (gmx_mm_any_lt(rsq21,rcutoff2))
511 {
512
513 r21 = _mm_mul_pd(rsq21,rinv21);
514
515 /* EWALD ELECTROSTATICS */
516
517 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
518 ewrt = _mm_mul_pd(r21,ewtabscale);
519 ewitab = _mm_cvttpd_epi32(ewrt);
520 #ifdef __XOP__
521 eweps = _mm_frcz_pd(ewrt);
522 #else
523 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
524 #endif
525 twoeweps = _mm_add_pd(eweps,eweps);
526 ewitab = _mm_slli_epi32(ewitab,2);
527 ewtabF = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
528 ewtabD = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
529 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
530 ewtabV = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
531 ewtabFn = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
532 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
533 felec = _mm_macc_pd(eweps,ewtabD,ewtabF);
534 velec = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
535 velec = _mm_mul_pd(qq21,_mm_sub_pd(_mm_sub_pd(rinv21,sh_ewald),velec));
536 felec = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
537
538 cutoff_mask = _mm_cmplt_pd(rsq21,rcutoff2);
539
540 /* Update potential sum for this i atom from the interaction with this j atom. */
541 velec = _mm_and_pd(velec,cutoff_mask);
542 velecsum = _mm_add_pd(velecsum,velec);
543
544 fscal = felec;
545
546 fscal = _mm_and_pd(fscal,cutoff_mask);
547
548 /* Update vectorial force */
549 fix2 = _mm_macc_pd(dx21,fscal,fix2);
550 fiy2 = _mm_macc_pd(dy21,fscal,fiy2);
551 fiz2 = _mm_macc_pd(dz21,fscal,fiz2);
552
553 fjx1 = _mm_macc_pd(dx21,fscal,fjx1);
554 fjy1 = _mm_macc_pd(dy21,fscal,fjy1);
555 fjz1 = _mm_macc_pd(dz21,fscal,fjz1);
556
557 }
558
559 /**************************
560 * CALCULATE INTERACTIONS *
561 **************************/
562
563 if (gmx_mm_any_lt(rsq22,rcutoff2))
564 {
565
566 r22 = _mm_mul_pd(rsq22,rinv22);
567
568 /* EWALD ELECTROSTATICS */
569
570 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
571 ewrt = _mm_mul_pd(r22,ewtabscale);
572 ewitab = _mm_cvttpd_epi32(ewrt);
573 #ifdef __XOP__
574 eweps = _mm_frcz_pd(ewrt);
575 #else
576 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
577 #endif
578 twoeweps = _mm_add_pd(eweps,eweps);
579 ewitab = _mm_slli_epi32(ewitab,2);
580 ewtabF = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
581 ewtabD = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
582 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
583 ewtabV = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
584 ewtabFn = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
585 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
586 felec = _mm_macc_pd(eweps,ewtabD,ewtabF);
587 velec = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
588 velec = _mm_mul_pd(qq22,_mm_sub_pd(_mm_sub_pd(rinv22,sh_ewald),velec));
589 felec = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
590
591 cutoff_mask = _mm_cmplt_pd(rsq22,rcutoff2);
592
593 /* Update potential sum for this i atom from the interaction with this j atom. */
594 velec = _mm_and_pd(velec,cutoff_mask);
595 velecsum = _mm_add_pd(velecsum,velec);
596
597 fscal = felec;
598
599 fscal = _mm_and_pd(fscal,cutoff_mask);
600
601 /* Update vectorial force */
602 fix2 = _mm_macc_pd(dx22,fscal,fix2);
603 fiy2 = _mm_macc_pd(dy22,fscal,fiy2);
604 fiz2 = _mm_macc_pd(dz22,fscal,fiz2);
605
606 fjx2 = _mm_macc_pd(dx22,fscal,fjx2);
607 fjy2 = _mm_macc_pd(dy22,fscal,fjy2);
608 fjz2 = _mm_macc_pd(dz22,fscal,fjz2);
609
610 }
611
612 /**************************
613 * CALCULATE INTERACTIONS *
614 **************************/
615
616 if (gmx_mm_any_lt(rsq23,rcutoff2))
617 {
618
619 r23 = _mm_mul_pd(rsq23,rinv23);
620
621 /* EWALD ELECTROSTATICS */
622
623 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
624 ewrt = _mm_mul_pd(r23,ewtabscale);
625 ewitab = _mm_cvttpd_epi32(ewrt);
626 #ifdef __XOP__
627 eweps = _mm_frcz_pd(ewrt);
628 #else
629 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
630 #endif
631 twoeweps = _mm_add_pd(eweps,eweps);
632 ewitab = _mm_slli_epi32(ewitab,2);
633 ewtabF = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
634 ewtabD = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
635 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
636 ewtabV = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
637 ewtabFn = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
638 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
639 felec = _mm_macc_pd(eweps,ewtabD,ewtabF);
640 velec = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
641 velec = _mm_mul_pd(qq23,_mm_sub_pd(_mm_sub_pd(rinv23,sh_ewald),velec));
642 felec = _mm_mul_pd(_mm_mul_pd(qq23,rinv23),_mm_sub_pd(rinvsq23,felec));
643
644 cutoff_mask = _mm_cmplt_pd(rsq23,rcutoff2);
645
646 /* Update potential sum for this i atom from the interaction with this j atom. */
647 velec = _mm_and_pd(velec,cutoff_mask);
648 velecsum = _mm_add_pd(velecsum,velec);
649
650 fscal = felec;
651
652 fscal = _mm_and_pd(fscal,cutoff_mask);
653
654 /* Update vectorial force */
655 fix2 = _mm_macc_pd(dx23,fscal,fix2);
656 fiy2 = _mm_macc_pd(dy23,fscal,fiy2);
657 fiz2 = _mm_macc_pd(dz23,fscal,fiz2);
658
659 fjx3 = _mm_macc_pd(dx23,fscal,fjx3);
660 fjy3 = _mm_macc_pd(dy23,fscal,fjy3);
661 fjz3 = _mm_macc_pd(dz23,fscal,fjz3);
662
663 }
664
665 /**************************
666 * CALCULATE INTERACTIONS *
667 **************************/
668
669 if (gmx_mm_any_lt(rsq31,rcutoff2))
670 {
671
672 r31 = _mm_mul_pd(rsq31,rinv31);
673
674 /* EWALD ELECTROSTATICS */
675
676 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
677 ewrt = _mm_mul_pd(r31,ewtabscale);
678 ewitab = _mm_cvttpd_epi32(ewrt);
679 #ifdef __XOP__
680 eweps = _mm_frcz_pd(ewrt);
681 #else
682 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
683 #endif
684 twoeweps = _mm_add_pd(eweps,eweps);
685 ewitab = _mm_slli_epi32(ewitab,2);
686 ewtabF = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
687 ewtabD = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
688 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
689 ewtabV = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
690 ewtabFn = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
691 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
692 felec = _mm_macc_pd(eweps,ewtabD,ewtabF);
693 velec = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
694 velec = _mm_mul_pd(qq31,_mm_sub_pd(_mm_sub_pd(rinv31,sh_ewald),velec));
695 felec = _mm_mul_pd(_mm_mul_pd(qq31,rinv31),_mm_sub_pd(rinvsq31,felec));
696
697 cutoff_mask = _mm_cmplt_pd(rsq31,rcutoff2);
698
699 /* Update potential sum for this i atom from the interaction with this j atom. */
700 velec = _mm_and_pd(velec,cutoff_mask);
701 velecsum = _mm_add_pd(velecsum,velec);
702
703 fscal = felec;
704
705 fscal = _mm_and_pd(fscal,cutoff_mask);
706
707 /* Update vectorial force */
708 fix3 = _mm_macc_pd(dx31,fscal,fix3);
709 fiy3 = _mm_macc_pd(dy31,fscal,fiy3);
710 fiz3 = _mm_macc_pd(dz31,fscal,fiz3);
711
712 fjx1 = _mm_macc_pd(dx31,fscal,fjx1);
713 fjy1 = _mm_macc_pd(dy31,fscal,fjy1);
714 fjz1 = _mm_macc_pd(dz31,fscal,fjz1);
715
716 }
717
718 /**************************
719 * CALCULATE INTERACTIONS *
720 **************************/
721
722 if (gmx_mm_any_lt(rsq32,rcutoff2))
723 {
724
725 r32 = _mm_mul_pd(rsq32,rinv32);
726
727 /* EWALD ELECTROSTATICS */
728
729 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
730 ewrt = _mm_mul_pd(r32,ewtabscale);
731 ewitab = _mm_cvttpd_epi32(ewrt);
732 #ifdef __XOP__
733 eweps = _mm_frcz_pd(ewrt);
734 #else
735 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
736 #endif
737 twoeweps = _mm_add_pd(eweps,eweps);
738 ewitab = _mm_slli_epi32(ewitab,2);
739 ewtabF = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
740 ewtabD = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
741 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
742 ewtabV = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
743 ewtabFn = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
744 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
745 felec = _mm_macc_pd(eweps,ewtabD,ewtabF);
746 velec = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
747 velec = _mm_mul_pd(qq32,_mm_sub_pd(_mm_sub_pd(rinv32,sh_ewald),velec));
748 felec = _mm_mul_pd(_mm_mul_pd(qq32,rinv32),_mm_sub_pd(rinvsq32,felec));
749
750 cutoff_mask = _mm_cmplt_pd(rsq32,rcutoff2);
751
752 /* Update potential sum for this i atom from the interaction with this j atom. */
753 velec = _mm_and_pd(velec,cutoff_mask);
754 velecsum = _mm_add_pd(velecsum,velec);
755
756 fscal = felec;
757
758 fscal = _mm_and_pd(fscal,cutoff_mask);
759
760 /* Update vectorial force */
761 fix3 = _mm_macc_pd(dx32,fscal,fix3);
762 fiy3 = _mm_macc_pd(dy32,fscal,fiy3);
763 fiz3 = _mm_macc_pd(dz32,fscal,fiz3);
764
765 fjx2 = _mm_macc_pd(dx32,fscal,fjx2);
766 fjy2 = _mm_macc_pd(dy32,fscal,fjy2);
767 fjz2 = _mm_macc_pd(dz32,fscal,fjz2);
768
769 }
770
771 /**************************
772 * CALCULATE INTERACTIONS *
773 **************************/
774
775 if (gmx_mm_any_lt(rsq33,rcutoff2))
776 {
777
778 r33 = _mm_mul_pd(rsq33,rinv33);
779
780 /* EWALD ELECTROSTATICS */
781
782 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
783 ewrt = _mm_mul_pd(r33,ewtabscale);
784 ewitab = _mm_cvttpd_epi32(ewrt);
785 #ifdef __XOP__
786 eweps = _mm_frcz_pd(ewrt);
787 #else
788 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
789 #endif
790 twoeweps = _mm_add_pd(eweps,eweps);
791 ewitab = _mm_slli_epi32(ewitab,2);
792 ewtabF = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
793 ewtabD = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
794 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
795 ewtabV = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
796 ewtabFn = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
797 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
798 felec = _mm_macc_pd(eweps,ewtabD,ewtabF);
799 velec = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
800 velec = _mm_mul_pd(qq33,_mm_sub_pd(_mm_sub_pd(rinv33,sh_ewald),velec));
801 felec = _mm_mul_pd(_mm_mul_pd(qq33,rinv33),_mm_sub_pd(rinvsq33,felec));
802
803 cutoff_mask = _mm_cmplt_pd(rsq33,rcutoff2);
804
805 /* Update potential sum for this i atom from the interaction with this j atom. */
806 velec = _mm_and_pd(velec,cutoff_mask);
807 velecsum = _mm_add_pd(velecsum,velec);
808
809 fscal = felec;
810
811 fscal = _mm_and_pd(fscal,cutoff_mask);
812
813 /* Update vectorial force */
814 fix3 = _mm_macc_pd(dx33,fscal,fix3);
815 fiy3 = _mm_macc_pd(dy33,fscal,fiy3);
816 fiz3 = _mm_macc_pd(dz33,fscal,fiz3);
817
818 fjx3 = _mm_macc_pd(dx33,fscal,fjx3);
819 fjy3 = _mm_macc_pd(dy33,fscal,fjy3);
820 fjz3 = _mm_macc_pd(dz33,fscal,fjz3);
821
822 }
823
824 gmx_mm_decrement_4rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
825
826 /* Inner loop uses 488 flops */
827 }
828
829 if(jidx<j_index_end)
830 {
831
832 jnrA = jjnr[jidx];
833 j_coord_offsetA = DIM*jnrA;
834
835 /* load j atom coordinates */
836 gmx_mm_load_4rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
837 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
838 &jy2,&jz2,&jx3,&jy3,&jz3);
839
840 /* Calculate displacement vector */
841 dx00 = _mm_sub_pd(ix0,jx0);
842 dy00 = _mm_sub_pd(iy0,jy0);
843 dz00 = _mm_sub_pd(iz0,jz0);
844 dx11 = _mm_sub_pd(ix1,jx1);
845 dy11 = _mm_sub_pd(iy1,jy1);
846 dz11 = _mm_sub_pd(iz1,jz1);
847 dx12 = _mm_sub_pd(ix1,jx2);
848 dy12 = _mm_sub_pd(iy1,jy2);
849 dz12 = _mm_sub_pd(iz1,jz2);
850 dx13 = _mm_sub_pd(ix1,jx3);
851 dy13 = _mm_sub_pd(iy1,jy3);
852 dz13 = _mm_sub_pd(iz1,jz3);
853 dx21 = _mm_sub_pd(ix2,jx1);
854 dy21 = _mm_sub_pd(iy2,jy1);
855 dz21 = _mm_sub_pd(iz2,jz1);
856 dx22 = _mm_sub_pd(ix2,jx2);
857 dy22 = _mm_sub_pd(iy2,jy2);
858 dz22 = _mm_sub_pd(iz2,jz2);
859 dx23 = _mm_sub_pd(ix2,jx3);
860 dy23 = _mm_sub_pd(iy2,jy3);
861 dz23 = _mm_sub_pd(iz2,jz3);
862 dx31 = _mm_sub_pd(ix3,jx1);
863 dy31 = _mm_sub_pd(iy3,jy1);
864 dz31 = _mm_sub_pd(iz3,jz1);
865 dx32 = _mm_sub_pd(ix3,jx2);
866 dy32 = _mm_sub_pd(iy3,jy2);
867 dz32 = _mm_sub_pd(iz3,jz2);
868 dx33 = _mm_sub_pd(ix3,jx3);
869 dy33 = _mm_sub_pd(iy3,jy3);
870 dz33 = _mm_sub_pd(iz3,jz3);
871
872 /* Calculate squared distance and things based on it */
873 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
874 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
875 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
876 rsq13 = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
877 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
878 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
879 rsq23 = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
880 rsq31 = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
881 rsq32 = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
882 rsq33 = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
883
884 rinv11 = gmx_mm_invsqrt_pd(rsq11);
885 rinv12 = gmx_mm_invsqrt_pd(rsq12);
886 rinv13 = gmx_mm_invsqrt_pd(rsq13);
887 rinv21 = gmx_mm_invsqrt_pd(rsq21);
888 rinv22 = gmx_mm_invsqrt_pd(rsq22);
889 rinv23 = gmx_mm_invsqrt_pd(rsq23);
890 rinv31 = gmx_mm_invsqrt_pd(rsq31);
891 rinv32 = gmx_mm_invsqrt_pd(rsq32);
892 rinv33 = gmx_mm_invsqrt_pd(rsq33);
893
894 rinvsq00 = gmx_mm_inv_pd(rsq00);
895 rinvsq11 = _mm_mul_pd(rinv11,rinv11);
896 rinvsq12 = _mm_mul_pd(rinv12,rinv12);
897 rinvsq13 = _mm_mul_pd(rinv13,rinv13);
898 rinvsq21 = _mm_mul_pd(rinv21,rinv21);
899 rinvsq22 = _mm_mul_pd(rinv22,rinv22);
900 rinvsq23 = _mm_mul_pd(rinv23,rinv23);
901 rinvsq31 = _mm_mul_pd(rinv31,rinv31);
902 rinvsq32 = _mm_mul_pd(rinv32,rinv32);
903 rinvsq33 = _mm_mul_pd(rinv33,rinv33);
904
905 fjx0 = _mm_setzero_pd();
906 fjy0 = _mm_setzero_pd();
907 fjz0 = _mm_setzero_pd();
908 fjx1 = _mm_setzero_pd();
909 fjy1 = _mm_setzero_pd();
910 fjz1 = _mm_setzero_pd();
911 fjx2 = _mm_setzero_pd();
912 fjy2 = _mm_setzero_pd();
913 fjz2 = _mm_setzero_pd();
914 fjx3 = _mm_setzero_pd();
915 fjy3 = _mm_setzero_pd();
916 fjz3 = _mm_setzero_pd();
917
918 /**************************
919 * CALCULATE INTERACTIONS *
920 **************************/
921
922 if (gmx_mm_any_lt(rsq00,rcutoff2))
923 {
924
925 /* LENNARD-JONES DISPERSION/REPULSION */
926
927 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
928 vvdw6 = _mm_mul_pd(c6_00,rinvsix);
929 vvdw12 = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
930 vvdw = _mm_msub_pd(_mm_nmacc_pd(c12_00,_mm_mul_pd(sh_vdw_invrcut6,sh_vdw_invrcut6),vvdw12),one_twelfth,
931 _mm_mul_pd(_mm_nmacc_pd( c6_00,sh_vdw_invrcut6,vvdw6),one_sixth));
932 fvdw = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
933
934 cutoff_mask = _mm_cmplt_pd(rsq00,rcutoff2);
935
936 /* Update potential sum for this i atom from the interaction with this j atom. */
937 vvdw = _mm_and_pd(vvdw,cutoff_mask);
938 vvdw = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
939 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
940
941 fscal = fvdw;
942
943 fscal = _mm_and_pd(fscal,cutoff_mask);
944
945 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
946
947 /* Update vectorial force */
948 fix0 = _mm_macc_pd(dx00,fscal,fix0);
949 fiy0 = _mm_macc_pd(dy00,fscal,fiy0);
950 fiz0 = _mm_macc_pd(dz00,fscal,fiz0);
951
952 fjx0 = _mm_macc_pd(dx00,fscal,fjx0);
953 fjy0 = _mm_macc_pd(dy00,fscal,fjy0);
954 fjz0 = _mm_macc_pd(dz00,fscal,fjz0);
955
956 }
957
958 /**************************
959 * CALCULATE INTERACTIONS *
960 **************************/
961
962 if (gmx_mm_any_lt(rsq11,rcutoff2))
963 {
964
965 r11 = _mm_mul_pd(rsq11,rinv11);
966
967 /* EWALD ELECTROSTATICS */
968
969 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
970 ewrt = _mm_mul_pd(r11,ewtabscale);
971 ewitab = _mm_cvttpd_epi32(ewrt);
972 #ifdef __XOP__
973 eweps = _mm_frcz_pd(ewrt);
974 #else
975 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
976 #endif
977 twoeweps = _mm_add_pd(eweps,eweps);
978 ewitab = _mm_slli_epi32(ewitab,2);
979 ewtabF = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
980 ewtabD = _mm_setzero_pd();
981 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
982 ewtabV = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
983 ewtabFn = _mm_setzero_pd();
984 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
985 felec = _mm_macc_pd(eweps,ewtabD,ewtabF);
986 velec = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
987 velec = _mm_mul_pd(qq11,_mm_sub_pd(_mm_sub_pd(rinv11,sh_ewald),velec));
988 felec = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
989
990 cutoff_mask = _mm_cmplt_pd(rsq11,rcutoff2);
991
992 /* Update potential sum for this i atom from the interaction with this j atom. */
993 velec = _mm_and_pd(velec,cutoff_mask);
994 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
995 velecsum = _mm_add_pd(velecsum,velec);
996
997 fscal = felec;
998
999 fscal = _mm_and_pd(fscal,cutoff_mask);
1000
1001 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1002
1003 /* Update vectorial force */
1004 fix1 = _mm_macc_pd(dx11,fscal,fix1);
1005 fiy1 = _mm_macc_pd(dy11,fscal,fiy1);
1006 fiz1 = _mm_macc_pd(dz11,fscal,fiz1);
1007
1008 fjx1 = _mm_macc_pd(dx11,fscal,fjx1);
1009 fjy1 = _mm_macc_pd(dy11,fscal,fjy1);
1010 fjz1 = _mm_macc_pd(dz11,fscal,fjz1);
1011
1012 }
1013
1014 /**************************
1015 * CALCULATE INTERACTIONS *
1016 **************************/
1017
1018 if (gmx_mm_any_lt(rsq12,rcutoff2))
1019 {
1020
1021 r12 = _mm_mul_pd(rsq12,rinv12);
1022
1023 /* EWALD ELECTROSTATICS */
1024
1025 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1026 ewrt = _mm_mul_pd(r12,ewtabscale);
1027 ewitab = _mm_cvttpd_epi32(ewrt);
1028 #ifdef __XOP__
1029 eweps = _mm_frcz_pd(ewrt);
1030 #else
1031 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1032 #endif
1033 twoeweps = _mm_add_pd(eweps,eweps);
1034 ewitab = _mm_slli_epi32(ewitab,2);
1035 ewtabF = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1036 ewtabD = _mm_setzero_pd();
1037 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1038 ewtabV = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
1039 ewtabFn = _mm_setzero_pd();
1040 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1041 felec = _mm_macc_pd(eweps,ewtabD,ewtabF);
1042 velec = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
1043 velec = _mm_mul_pd(qq12,_mm_sub_pd(_mm_sub_pd(rinv12,sh_ewald),velec));
1044 felec = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
1045
1046 cutoff_mask = _mm_cmplt_pd(rsq12,rcutoff2);
1047
1048 /* Update potential sum for this i atom from the interaction with this j atom. */
1049 velec = _mm_and_pd(velec,cutoff_mask);
1050 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1051 velecsum = _mm_add_pd(velecsum,velec);
1052
1053 fscal = felec;
1054
1055 fscal = _mm_and_pd(fscal,cutoff_mask);
1056
1057 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1058
1059 /* Update vectorial force */
1060 fix1 = _mm_macc_pd(dx12,fscal,fix1);
1061 fiy1 = _mm_macc_pd(dy12,fscal,fiy1);
1062 fiz1 = _mm_macc_pd(dz12,fscal,fiz1);
1063
1064 fjx2 = _mm_macc_pd(dx12,fscal,fjx2);
1065 fjy2 = _mm_macc_pd(dy12,fscal,fjy2);
1066 fjz2 = _mm_macc_pd(dz12,fscal,fjz2);
1067
1068 }
1069
1070 /**************************
1071 * CALCULATE INTERACTIONS *
1072 **************************/
1073
1074 if (gmx_mm_any_lt(rsq13,rcutoff2))
1075 {
1076
1077 r13 = _mm_mul_pd(rsq13,rinv13);
1078
1079 /* EWALD ELECTROSTATICS */
1080
1081 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1082 ewrt = _mm_mul_pd(r13,ewtabscale);
1083 ewitab = _mm_cvttpd_epi32(ewrt);
1084 #ifdef __XOP__
1085 eweps = _mm_frcz_pd(ewrt);
1086 #else
1087 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1088 #endif
1089 twoeweps = _mm_add_pd(eweps,eweps);
1090 ewitab = _mm_slli_epi32(ewitab,2);
1091 ewtabF = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1092 ewtabD = _mm_setzero_pd();
1093 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1094 ewtabV = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
1095 ewtabFn = _mm_setzero_pd();
1096 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1097 felec = _mm_macc_pd(eweps,ewtabD,ewtabF);
1098 velec = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
1099 velec = _mm_mul_pd(qq13,_mm_sub_pd(_mm_sub_pd(rinv13,sh_ewald),velec));
1100 felec = _mm_mul_pd(_mm_mul_pd(qq13,rinv13),_mm_sub_pd(rinvsq13,felec));
1101
1102 cutoff_mask = _mm_cmplt_pd(rsq13,rcutoff2);
1103
1104 /* Update potential sum for this i atom from the interaction with this j atom. */
1105 velec = _mm_and_pd(velec,cutoff_mask);
1106 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1107 velecsum = _mm_add_pd(velecsum,velec);
1108
1109 fscal = felec;
1110
1111 fscal = _mm_and_pd(fscal,cutoff_mask);
1112
1113 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1114
1115 /* Update vectorial force */
1116 fix1 = _mm_macc_pd(dx13,fscal,fix1);
1117 fiy1 = _mm_macc_pd(dy13,fscal,fiy1);
1118 fiz1 = _mm_macc_pd(dz13,fscal,fiz1);
1119
1120 fjx3 = _mm_macc_pd(dx13,fscal,fjx3);
1121 fjy3 = _mm_macc_pd(dy13,fscal,fjy3);
1122 fjz3 = _mm_macc_pd(dz13,fscal,fjz3);
1123
1124 }
1125
1126 /**************************
1127 * CALCULATE INTERACTIONS *
1128 **************************/
1129
1130 if (gmx_mm_any_lt(rsq21,rcutoff2))
1131 {
1132
1133 r21 = _mm_mul_pd(rsq21,rinv21);
1134
1135 /* EWALD ELECTROSTATICS */
1136
1137 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1138 ewrt = _mm_mul_pd(r21,ewtabscale);
1139 ewitab = _mm_cvttpd_epi32(ewrt);
1140 #ifdef __XOP__
1141 eweps = _mm_frcz_pd(ewrt);
1142 #else
1143 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1144 #endif
1145 twoeweps = _mm_add_pd(eweps,eweps);
1146 ewitab = _mm_slli_epi32(ewitab,2);
1147 ewtabF = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1148 ewtabD = _mm_setzero_pd();
1149 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1150 ewtabV = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
1151 ewtabFn = _mm_setzero_pd();
1152 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1153 felec = _mm_macc_pd(eweps,ewtabD,ewtabF);
1154 velec = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
1155 velec = _mm_mul_pd(qq21,_mm_sub_pd(_mm_sub_pd(rinv21,sh_ewald),velec));
1156 felec = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
1157
1158 cutoff_mask = _mm_cmplt_pd(rsq21,rcutoff2);
1159
1160 /* Update potential sum for this i atom from the interaction with this j atom. */
1161 velec = _mm_and_pd(velec,cutoff_mask);
1162 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1163 velecsum = _mm_add_pd(velecsum,velec);
1164
1165 fscal = felec;
1166
1167 fscal = _mm_and_pd(fscal,cutoff_mask);
1168
1169 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1170
1171 /* Update vectorial force */
1172 fix2 = _mm_macc_pd(dx21,fscal,fix2);
1173 fiy2 = _mm_macc_pd(dy21,fscal,fiy2);
1174 fiz2 = _mm_macc_pd(dz21,fscal,fiz2);
1175
1176 fjx1 = _mm_macc_pd(dx21,fscal,fjx1);
1177 fjy1 = _mm_macc_pd(dy21,fscal,fjy1);
1178 fjz1 = _mm_macc_pd(dz21,fscal,fjz1);
1179
1180 }
1181
1182 /**************************
1183 * CALCULATE INTERACTIONS *
1184 **************************/
1185
1186 if (gmx_mm_any_lt(rsq22,rcutoff2))
1187 {
1188
1189 r22 = _mm_mul_pd(rsq22,rinv22);
1190
1191 /* EWALD ELECTROSTATICS */
1192
1193 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1194 ewrt = _mm_mul_pd(r22,ewtabscale);
1195 ewitab = _mm_cvttpd_epi32(ewrt);
1196 #ifdef __XOP__
1197 eweps = _mm_frcz_pd(ewrt);
1198 #else
1199 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1200 #endif
1201 twoeweps = _mm_add_pd(eweps,eweps);
1202 ewitab = _mm_slli_epi32(ewitab,2);
1203 ewtabF = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1204 ewtabD = _mm_setzero_pd();
1205 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1206 ewtabV = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
1207 ewtabFn = _mm_setzero_pd();
1208 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1209 felec = _mm_macc_pd(eweps,ewtabD,ewtabF);
1210 velec = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
1211 velec = _mm_mul_pd(qq22,_mm_sub_pd(_mm_sub_pd(rinv22,sh_ewald),velec));
1212 felec = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
1213
1214 cutoff_mask = _mm_cmplt_pd(rsq22,rcutoff2);
1215
1216 /* Update potential sum for this i atom from the interaction with this j atom. */
1217 velec = _mm_and_pd(velec,cutoff_mask);
1218 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1219 velecsum = _mm_add_pd(velecsum,velec);
1220
1221 fscal = felec;
1222
1223 fscal = _mm_and_pd(fscal,cutoff_mask);
1224
1225 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1226
1227 /* Update vectorial force */
1228 fix2 = _mm_macc_pd(dx22,fscal,fix2);
1229 fiy2 = _mm_macc_pd(dy22,fscal,fiy2);
1230 fiz2 = _mm_macc_pd(dz22,fscal,fiz2);
1231
1232 fjx2 = _mm_macc_pd(dx22,fscal,fjx2);
1233 fjy2 = _mm_macc_pd(dy22,fscal,fjy2);
1234 fjz2 = _mm_macc_pd(dz22,fscal,fjz2);
1235
1236 }
1237
1238 /**************************
1239 * CALCULATE INTERACTIONS *
1240 **************************/
1241
1242 if (gmx_mm_any_lt(rsq23,rcutoff2))
1243 {
1244
1245 r23 = _mm_mul_pd(rsq23,rinv23);
1246
1247 /* EWALD ELECTROSTATICS */
1248
1249 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1250 ewrt = _mm_mul_pd(r23,ewtabscale);
1251 ewitab = _mm_cvttpd_epi32(ewrt);
1252 #ifdef __XOP__
1253 eweps = _mm_frcz_pd(ewrt);
1254 #else
1255 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1256 #endif
1257 twoeweps = _mm_add_pd(eweps,eweps);
1258 ewitab = _mm_slli_epi32(ewitab,2);
1259 ewtabF = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1260 ewtabD = _mm_setzero_pd();
1261 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1262 ewtabV = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
1263 ewtabFn = _mm_setzero_pd();
1264 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1265 felec = _mm_macc_pd(eweps,ewtabD,ewtabF);
1266 velec = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
1267 velec = _mm_mul_pd(qq23,_mm_sub_pd(_mm_sub_pd(rinv23,sh_ewald),velec));
1268 felec = _mm_mul_pd(_mm_mul_pd(qq23,rinv23),_mm_sub_pd(rinvsq23,felec));
1269
1270 cutoff_mask = _mm_cmplt_pd(rsq23,rcutoff2);
1271
1272 /* Update potential sum for this i atom from the interaction with this j atom. */
1273 velec = _mm_and_pd(velec,cutoff_mask);
1274 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1275 velecsum = _mm_add_pd(velecsum,velec);
1276
1277 fscal = felec;
1278
1279 fscal = _mm_and_pd(fscal,cutoff_mask);
1280
1281 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1282
1283 /* Update vectorial force */
1284 fix2 = _mm_macc_pd(dx23,fscal,fix2);
1285 fiy2 = _mm_macc_pd(dy23,fscal,fiy2);
1286 fiz2 = _mm_macc_pd(dz23,fscal,fiz2);
1287
1288 fjx3 = _mm_macc_pd(dx23,fscal,fjx3);
1289 fjy3 = _mm_macc_pd(dy23,fscal,fjy3);
1290 fjz3 = _mm_macc_pd(dz23,fscal,fjz3);
1291
1292 }
1293
1294 /**************************
1295 * CALCULATE INTERACTIONS *
1296 **************************/
1297
1298 if (gmx_mm_any_lt(rsq31,rcutoff2))
1299 {
1300
1301 r31 = _mm_mul_pd(rsq31,rinv31);
1302
1303 /* EWALD ELECTROSTATICS */
1304
1305 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1306 ewrt = _mm_mul_pd(r31,ewtabscale);
1307 ewitab = _mm_cvttpd_epi32(ewrt);
1308 #ifdef __XOP__
1309 eweps = _mm_frcz_pd(ewrt);
1310 #else
1311 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1312 #endif
1313 twoeweps = _mm_add_pd(eweps,eweps);
1314 ewitab = _mm_slli_epi32(ewitab,2);
1315 ewtabF = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1316 ewtabD = _mm_setzero_pd();
1317 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1318 ewtabV = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
1319 ewtabFn = _mm_setzero_pd();
1320 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1321 felec = _mm_macc_pd(eweps,ewtabD,ewtabF);
1322 velec = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
1323 velec = _mm_mul_pd(qq31,_mm_sub_pd(_mm_sub_pd(rinv31,sh_ewald),velec));
1324 felec = _mm_mul_pd(_mm_mul_pd(qq31,rinv31),_mm_sub_pd(rinvsq31,felec));
1325
1326 cutoff_mask = _mm_cmplt_pd(rsq31,rcutoff2);
1327
1328 /* Update potential sum for this i atom from the interaction with this j atom. */
1329 velec = _mm_and_pd(velec,cutoff_mask);
1330 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1331 velecsum = _mm_add_pd(velecsum,velec);
1332
1333 fscal = felec;
1334
1335 fscal = _mm_and_pd(fscal,cutoff_mask);
1336
1337 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1338
1339 /* Update vectorial force */
1340 fix3 = _mm_macc_pd(dx31,fscal,fix3);
1341 fiy3 = _mm_macc_pd(dy31,fscal,fiy3);
1342 fiz3 = _mm_macc_pd(dz31,fscal,fiz3);
1343
1344 fjx1 = _mm_macc_pd(dx31,fscal,fjx1);
1345 fjy1 = _mm_macc_pd(dy31,fscal,fjy1);
1346 fjz1 = _mm_macc_pd(dz31,fscal,fjz1);
1347
1348 }
1349
1350 /**************************
1351 * CALCULATE INTERACTIONS *
1352 **************************/
1353
1354 if (gmx_mm_any_lt(rsq32,rcutoff2))
1355 {
1356
1357 r32 = _mm_mul_pd(rsq32,rinv32);
1358
1359 /* EWALD ELECTROSTATICS */
1360
1361 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1362 ewrt = _mm_mul_pd(r32,ewtabscale);
1363 ewitab = _mm_cvttpd_epi32(ewrt);
1364 #ifdef __XOP__
1365 eweps = _mm_frcz_pd(ewrt);
1366 #else
1367 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1368 #endif
1369 twoeweps = _mm_add_pd(eweps,eweps);
1370 ewitab = _mm_slli_epi32(ewitab,2);
1371 ewtabF = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1372 ewtabD = _mm_setzero_pd();
1373 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1374 ewtabV = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
1375 ewtabFn = _mm_setzero_pd();
1376 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1377 felec = _mm_macc_pd(eweps,ewtabD,ewtabF);
1378 velec = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
1379 velec = _mm_mul_pd(qq32,_mm_sub_pd(_mm_sub_pd(rinv32,sh_ewald),velec));
1380 felec = _mm_mul_pd(_mm_mul_pd(qq32,rinv32),_mm_sub_pd(rinvsq32,felec));
1381
1382 cutoff_mask = _mm_cmplt_pd(rsq32,rcutoff2);
1383
1384 /* Update potential sum for this i atom from the interaction with this j atom. */
1385 velec = _mm_and_pd(velec,cutoff_mask);
1386 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1387 velecsum = _mm_add_pd(velecsum,velec);
1388
1389 fscal = felec;
1390
1391 fscal = _mm_and_pd(fscal,cutoff_mask);
1392
1393 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1394
1395 /* Update vectorial force */
1396 fix3 = _mm_macc_pd(dx32,fscal,fix3);
1397 fiy3 = _mm_macc_pd(dy32,fscal,fiy3);
1398 fiz3 = _mm_macc_pd(dz32,fscal,fiz3);
1399
1400 fjx2 = _mm_macc_pd(dx32,fscal,fjx2);
1401 fjy2 = _mm_macc_pd(dy32,fscal,fjy2);
1402 fjz2 = _mm_macc_pd(dz32,fscal,fjz2);
1403
1404 }
1405
1406 /**************************
1407 * CALCULATE INTERACTIONS *
1408 **************************/
1409
1410 if (gmx_mm_any_lt(rsq33,rcutoff2))
1411 {
1412
1413 r33 = _mm_mul_pd(rsq33,rinv33);
1414
1415 /* EWALD ELECTROSTATICS */
1416
1417 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1418 ewrt = _mm_mul_pd(r33,ewtabscale);
1419 ewitab = _mm_cvttpd_epi32(ewrt);
1420 #ifdef __XOP__
1421 eweps = _mm_frcz_pd(ewrt);
1422 #else
1423 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1424 #endif
1425 twoeweps = _mm_add_pd(eweps,eweps);
1426 ewitab = _mm_slli_epi32(ewitab,2);
1427 ewtabF = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1428 ewtabD = _mm_setzero_pd();
1429 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1430 ewtabV = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
1431 ewtabFn = _mm_setzero_pd();
1432 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1433 felec = _mm_macc_pd(eweps,ewtabD,ewtabF);
1434 velec = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
1435 velec = _mm_mul_pd(qq33,_mm_sub_pd(_mm_sub_pd(rinv33,sh_ewald),velec));
1436 felec = _mm_mul_pd(_mm_mul_pd(qq33,rinv33),_mm_sub_pd(rinvsq33,felec));
1437
1438 cutoff_mask = _mm_cmplt_pd(rsq33,rcutoff2);
1439
1440 /* Update potential sum for this i atom from the interaction with this j atom. */
1441 velec = _mm_and_pd(velec,cutoff_mask);
1442 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1443 velecsum = _mm_add_pd(velecsum,velec);
1444
1445 fscal = felec;
1446
1447 fscal = _mm_and_pd(fscal,cutoff_mask);
1448
1449 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1450
1451 /* Update vectorial force */
1452 fix3 = _mm_macc_pd(dx33,fscal,fix3);
1453 fiy3 = _mm_macc_pd(dy33,fscal,fiy3);
1454 fiz3 = _mm_macc_pd(dz33,fscal,fiz3);
1455
1456 fjx3 = _mm_macc_pd(dx33,fscal,fjx3);
1457 fjy3 = _mm_macc_pd(dy33,fscal,fjy3);
1458 fjz3 = _mm_macc_pd(dz33,fscal,fjz3);
1459
1460 }
1461
1462 gmx_mm_decrement_4rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1463
1464 /* Inner loop uses 488 flops */
1465 }
1466
1467 /* End of innermost loop */
1468
1469 gmx_mm_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1470 f+i_coord_offset,fshift+i_shift_offset);
1471
1472 ggid = gid[iidx];
1473 /* Update potential energies */
1474 gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1475 gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1476
1477 /* Increment number of inner iterations */
1478 inneriter += j_index_end - j_index_start;
1479
1480 /* Outer loop uses 26 flops */
1481 }
1482
1483 /* Increment number of outer iterations */
1484 outeriter += nri;
1485
1486 /* Update outer/inner flops */
1487
1488 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*488);
1489 }
1490 /*
1491 * Gromacs nonbonded kernel: nb_kernel_ElecEwSh_VdwLJSh_GeomW4W4_F_avx_128_fma_double
1492 * Electrostatics interaction: Ewald
1493 * VdW interaction: LennardJones
1494 * Geometry: Water4-Water4
1495 * Calculate force/pot: Force
1496 */
1497 void
1498 nb_kernel_ElecEwSh_VdwLJSh_GeomW4W4_F_avx_128_fma_double
1499 (t_nblist * gmx_restrict nlist,
1500 rvec * gmx_restrict xx,
1501 rvec * gmx_restrict ff,
1502 t_forcerec * gmx_restrict fr,
1503 t_mdatoms * gmx_restrict mdatoms,
1504 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1505 t_nrnb * gmx_restrict nrnb)
1506 {
1507 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1508 * just 0 for non-waters.
1509 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
1510 * jnr indices corresponding to data put in the four positions in the SIMD register.
1511 */
1512 int i_shift_offset,i_coord_offset,outeriter,inneriter;
1513 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1514 int jnrA,jnrB;
1515 int j_coord_offsetA,j_coord_offsetB;
1516 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
1517 real rcutoff_scalar;
1518 real *shiftvec,*fshift,*x,*f;
1519 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1520 int vdwioffset0;
1521 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1522 int vdwioffset1;
1523 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1524 int vdwioffset2;
1525 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1526 int vdwioffset3;
1527 __m128d ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1528 int vdwjidx0A,vdwjidx0B;
1529 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1530 int vdwjidx1A,vdwjidx1B;
1531 __m128d jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1532 int vdwjidx2A,vdwjidx2B;
1533 __m128d jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1534 int vdwjidx3A,vdwjidx3B;
1535 __m128d jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1536 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1537 __m128d dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1538 __m128d dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1539 __m128d dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1540 __m128d dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1541 __m128d dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1542 __m128d dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1543 __m128d dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1544 __m128d dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1545 __m128d dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1546 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
1547 real *charge;
1548 int nvdwtype;
1549 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1550 int *vdwtype;
1551 real *vdwparam;
1552 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
1553 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
1554 __m128i ewitab;
1555 __m128d ewtabscale,eweps,twoeweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1556 real *ewtab;
1557 __m128d dummy_mask,cutoff_mask;
1558 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
1559 __m128d one = _mm_set1_pd(1.0);
1560 __m128d two = _mm_set1_pd(2.0);
1561 x = xx[0];
1562 f = ff[0];
1563
1564 nri = nlist->nri;
1565 iinr = nlist->iinr;
1566 jindex = nlist->jindex;
1567 jjnr = nlist->jjnr;
1568 shiftidx = nlist->shift;
1569 gid = nlist->gid;
1570 shiftvec = fr->shift_vec[0];
1571 fshift = fr->fshift[0];
1572 facel = _mm_set1_pd(fr->epsfac);
1573 charge = mdatoms->chargeA;
1574 nvdwtype = fr->ntype;
1575 vdwparam = fr->nbfp;
1576 vdwtype = mdatoms->typeA;
1577
1578 sh_ewald = _mm_set1_pd(fr->ic->sh_ewald);
1579 ewtab = fr->ic->tabq_coul_F;
1580 ewtabscale = _mm_set1_pd(fr->ic->tabq_scale);
1581 ewtabhalfspace = _mm_set1_pd(0.5/fr->ic->tabq_scale);
1582
1583 /* Setup water-specific parameters */
1584 inr = nlist->iinr[0];
1585 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
1586 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
1587 iq3 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
1588 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
1589
1590 jq1 = _mm_set1_pd(charge[inr+1]);
1591 jq2 = _mm_set1_pd(charge[inr+2]);
1592 jq3 = _mm_set1_pd(charge[inr+3]);
1593 vdwjidx0A = 2*vdwtype[inr+0];
1594 c6_00 = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
1595 c12_00 = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
1596 qq11 = _mm_mul_pd(iq1,jq1);
1597 qq12 = _mm_mul_pd(iq1,jq2);
1598 qq13 = _mm_mul_pd(iq1,jq3);
1599 qq21 = _mm_mul_pd(iq2,jq1);
1600 qq22 = _mm_mul_pd(iq2,jq2);
1601 qq23 = _mm_mul_pd(iq2,jq3);
1602 qq31 = _mm_mul_pd(iq3,jq1);
1603 qq32 = _mm_mul_pd(iq3,jq2);
1604 qq33 = _mm_mul_pd(iq3,jq3);
1605
1606 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
1607 rcutoff_scalar = fr->rcoulomb;
1608 rcutoff = _mm_set1_pd(rcutoff_scalar);
1609 rcutoff2 = _mm_mul_pd(rcutoff,rcutoff);
1610
1611 sh_vdw_invrcut6 = _mm_set1_pd(fr->ic->sh_invrc6);
1612 rvdw = _mm_set1_pd(fr->rvdw);
1613
1614 /* Avoid stupid compiler warnings */
1615 jnrA = jnrB = 0;
1616 j_coord_offsetA = 0;
1617 j_coord_offsetB = 0;
1618
1619 outeriter = 0;
1620 inneriter = 0;
1621
1622 /* Start outer loop over neighborlists */
1623 for(iidx=0; iidx<nri; iidx++)
1624 {
1625 /* Load shift vector for this list */
1626 i_shift_offset = DIM*shiftidx[iidx];
1627
1628 /* Load limits for loop over neighbors */
1629 j_index_start = jindex[iidx];
1630 j_index_end = jindex[iidx+1];
1631
1632 /* Get outer coordinate index */
1633 inr = iinr[iidx];
1634 i_coord_offset = DIM*inr;
1635
1636 /* Load i particle coords and add shift vector */
1637 gmx_mm_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1638 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1639
1640 fix0 = _mm_setzero_pd();
1641 fiy0 = _mm_setzero_pd();
1642 fiz0 = _mm_setzero_pd();
1643 fix1 = _mm_setzero_pd();
1644 fiy1 = _mm_setzero_pd();
1645 fiz1 = _mm_setzero_pd();
1646 fix2 = _mm_setzero_pd();
1647 fiy2 = _mm_setzero_pd();
1648 fiz2 = _mm_setzero_pd();
1649 fix3 = _mm_setzero_pd();
1650 fiy3 = _mm_setzero_pd();
1651 fiz3 = _mm_setzero_pd();
1652
1653 /* Start inner kernel loop */
1654 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
1655 {
1656
1657 /* Get j neighbor index, and coordinate index */
1658 jnrA = jjnr[jidx];
1659 jnrB = jjnr[jidx+1];
1660 j_coord_offsetA = DIM*jnrA;
1661 j_coord_offsetB = DIM*jnrB;
1662
1663 /* load j atom coordinates */
1664 gmx_mm_load_4rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1665 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1666 &jy2,&jz2,&jx3,&jy3,&jz3);
1667
1668 /* Calculate displacement vector */
1669 dx00 = _mm_sub_pd(ix0,jx0);
1670 dy00 = _mm_sub_pd(iy0,jy0);
1671 dz00 = _mm_sub_pd(iz0,jz0);
1672 dx11 = _mm_sub_pd(ix1,jx1);
1673 dy11 = _mm_sub_pd(iy1,jy1);
1674 dz11 = _mm_sub_pd(iz1,jz1);
1675 dx12 = _mm_sub_pd(ix1,jx2);
1676 dy12 = _mm_sub_pd(iy1,jy2);
1677 dz12 = _mm_sub_pd(iz1,jz2);
1678 dx13 = _mm_sub_pd(ix1,jx3);
1679 dy13 = _mm_sub_pd(iy1,jy3);
1680 dz13 = _mm_sub_pd(iz1,jz3);
1681 dx21 = _mm_sub_pd(ix2,jx1);
1682 dy21 = _mm_sub_pd(iy2,jy1);
1683 dz21 = _mm_sub_pd(iz2,jz1);
1684 dx22 = _mm_sub_pd(ix2,jx2);
1685 dy22 = _mm_sub_pd(iy2,jy2);
1686 dz22 = _mm_sub_pd(iz2,jz2);
1687 dx23 = _mm_sub_pd(ix2,jx3);
1688 dy23 = _mm_sub_pd(iy2,jy3);
1689 dz23 = _mm_sub_pd(iz2,jz3);
1690 dx31 = _mm_sub_pd(ix3,jx1);
1691 dy31 = _mm_sub_pd(iy3,jy1);
1692 dz31 = _mm_sub_pd(iz3,jz1);
1693 dx32 = _mm_sub_pd(ix3,jx2);
1694 dy32 = _mm_sub_pd(iy3,jy2);
1695 dz32 = _mm_sub_pd(iz3,jz2);
1696 dx33 = _mm_sub_pd(ix3,jx3);
1697 dy33 = _mm_sub_pd(iy3,jy3);
1698 dz33 = _mm_sub_pd(iz3,jz3);
1699
1700 /* Calculate squared distance and things based on it */
1701 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1702 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1703 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1704 rsq13 = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
1705 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1706 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1707 rsq23 = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
1708 rsq31 = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
1709 rsq32 = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
1710 rsq33 = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
1711
1712 rinv11 = gmx_mm_invsqrt_pd(rsq11);
1713 rinv12 = gmx_mm_invsqrt_pd(rsq12);
1714 rinv13 = gmx_mm_invsqrt_pd(rsq13);
1715 rinv21 = gmx_mm_invsqrt_pd(rsq21);
1716 rinv22 = gmx_mm_invsqrt_pd(rsq22);
1717 rinv23 = gmx_mm_invsqrt_pd(rsq23);
1718 rinv31 = gmx_mm_invsqrt_pd(rsq31);
1719 rinv32 = gmx_mm_invsqrt_pd(rsq32);
1720 rinv33 = gmx_mm_invsqrt_pd(rsq33);
1721
1722 rinvsq00 = gmx_mm_inv_pd(rsq00);
1723 rinvsq11 = _mm_mul_pd(rinv11,rinv11);
1724 rinvsq12 = _mm_mul_pd(rinv12,rinv12);
1725 rinvsq13 = _mm_mul_pd(rinv13,rinv13);
1726 rinvsq21 = _mm_mul_pd(rinv21,rinv21);
1727 rinvsq22 = _mm_mul_pd(rinv22,rinv22);
1728 rinvsq23 = _mm_mul_pd(rinv23,rinv23);
1729 rinvsq31 = _mm_mul_pd(rinv31,rinv31);
1730 rinvsq32 = _mm_mul_pd(rinv32,rinv32);
1731 rinvsq33 = _mm_mul_pd(rinv33,rinv33);
1732
1733 fjx0 = _mm_setzero_pd();
1734 fjy0 = _mm_setzero_pd();
1735 fjz0 = _mm_setzero_pd();
1736 fjx1 = _mm_setzero_pd();
1737 fjy1 = _mm_setzero_pd();
1738 fjz1 = _mm_setzero_pd();
1739 fjx2 = _mm_setzero_pd();
1740 fjy2 = _mm_setzero_pd();
1741 fjz2 = _mm_setzero_pd();
1742 fjx3 = _mm_setzero_pd();
1743 fjy3 = _mm_setzero_pd();
1744 fjz3 = _mm_setzero_pd();
1745
1746 /**************************
1747 * CALCULATE INTERACTIONS *
1748 **************************/
1749
1750 if (gmx_mm_any_lt(rsq00,rcutoff2))
1751 {
1752
1753 /* LENNARD-JONES DISPERSION/REPULSION */
1754
1755 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1756 fvdw = _mm_mul_pd(_mm_msub_pd(c12_00,rinvsix,c6_00),_mm_mul_pd(rinvsix,rinvsq00));
1757
1758 cutoff_mask = _mm_cmplt_pd(rsq00,rcutoff2);
1759
1760 fscal = fvdw;
1761
1762 fscal = _mm_and_pd(fscal,cutoff_mask);
1763
1764 /* Update vectorial force */
1765 fix0 = _mm_macc_pd(dx00,fscal,fix0);
1766 fiy0 = _mm_macc_pd(dy00,fscal,fiy0);
1767 fiz0 = _mm_macc_pd(dz00,fscal,fiz0);
1768
1769 fjx0 = _mm_macc_pd(dx00,fscal,fjx0);
1770 fjy0 = _mm_macc_pd(dy00,fscal,fjy0);
1771 fjz0 = _mm_macc_pd(dz00,fscal,fjz0);
1772
1773 }
1774
1775 /**************************
1776 * CALCULATE INTERACTIONS *
1777 **************************/
1778
1779 if (gmx_mm_any_lt(rsq11,rcutoff2))
1780 {
1781
1782 r11 = _mm_mul_pd(rsq11,rinv11);
1783
1784 /* EWALD ELECTROSTATICS */
1785
1786 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1787 ewrt = _mm_mul_pd(r11,ewtabscale);
1788 ewitab = _mm_cvttpd_epi32(ewrt);
1789 #ifdef __XOP__
1790 eweps = _mm_frcz_pd(ewrt);
1791 #else
1792 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1793 #endif
1794 twoeweps = _mm_add_pd(eweps,eweps);
1795 gmx_mm_load_2pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),ewtab+_mm_extract_epi32(ewitab,1),
1796 &ewtabF,&ewtabFn);
1797 felec = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
1798 felec = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
1799
1800 cutoff_mask = _mm_cmplt_pd(rsq11,rcutoff2);
1801
1802 fscal = felec;
1803
1804 fscal = _mm_and_pd(fscal,cutoff_mask);
1805
1806 /* Update vectorial force */
1807 fix1 = _mm_macc_pd(dx11,fscal,fix1);
1808 fiy1 = _mm_macc_pd(dy11,fscal,fiy1);
1809 fiz1 = _mm_macc_pd(dz11,fscal,fiz1);
1810
1811 fjx1 = _mm_macc_pd(dx11,fscal,fjx1);
1812 fjy1 = _mm_macc_pd(dy11,fscal,fjy1);
1813 fjz1 = _mm_macc_pd(dz11,fscal,fjz1);
1814
1815 }
1816
1817 /**************************
1818 * CALCULATE INTERACTIONS *
1819 **************************/
1820
1821 if (gmx_mm_any_lt(rsq12,rcutoff2))
1822 {
1823
1824 r12 = _mm_mul_pd(rsq12,rinv12);
1825
1826 /* EWALD ELECTROSTATICS */
1827
1828 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1829 ewrt = _mm_mul_pd(r12,ewtabscale);
1830 ewitab = _mm_cvttpd_epi32(ewrt);
1831 #ifdef __XOP__
1832 eweps = _mm_frcz_pd(ewrt);
1833 #else
1834 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1835 #endif
1836 twoeweps = _mm_add_pd(eweps,eweps);
1837 gmx_mm_load_2pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),ewtab+_mm_extract_epi32(ewitab,1),
1838 &ewtabF,&ewtabFn);
1839 felec = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
1840 felec = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
1841
1842 cutoff_mask = _mm_cmplt_pd(rsq12,rcutoff2);
1843
1844 fscal = felec;
1845
1846 fscal = _mm_and_pd(fscal,cutoff_mask);
1847
1848 /* Update vectorial force */
1849 fix1 = _mm_macc_pd(dx12,fscal,fix1);
1850 fiy1 = _mm_macc_pd(dy12,fscal,fiy1);
1851 fiz1 = _mm_macc_pd(dz12,fscal,fiz1);
1852
1853 fjx2 = _mm_macc_pd(dx12,fscal,fjx2);
1854 fjy2 = _mm_macc_pd(dy12,fscal,fjy2);
1855 fjz2 = _mm_macc_pd(dz12,fscal,fjz2);
1856
1857 }
1858
1859 /**************************
1860 * CALCULATE INTERACTIONS *
1861 **************************/
1862
1863 if (gmx_mm_any_lt(rsq13,rcutoff2))
1864 {
1865
1866 r13 = _mm_mul_pd(rsq13,rinv13);
1867
1868 /* EWALD ELECTROSTATICS */
1869
1870 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1871 ewrt = _mm_mul_pd(r13,ewtabscale);
1872 ewitab = _mm_cvttpd_epi32(ewrt);
1873 #ifdef __XOP__
1874 eweps = _mm_frcz_pd(ewrt);
1875 #else
1876 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1877 #endif
1878 twoeweps = _mm_add_pd(eweps,eweps);
1879 gmx_mm_load_2pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),ewtab+_mm_extract_epi32(ewitab,1),
1880 &ewtabF,&ewtabFn);
1881 felec = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
1882 felec = _mm_mul_pd(_mm_mul_pd(qq13,rinv13),_mm_sub_pd(rinvsq13,felec));
1883
1884 cutoff_mask = _mm_cmplt_pd(rsq13,rcutoff2);
1885
1886 fscal = felec;
1887
1888 fscal = _mm_and_pd(fscal,cutoff_mask);
1889
1890 /* Update vectorial force */
1891 fix1 = _mm_macc_pd(dx13,fscal,fix1);
1892 fiy1 = _mm_macc_pd(dy13,fscal,fiy1);
1893 fiz1 = _mm_macc_pd(dz13,fscal,fiz1);
1894
1895 fjx3 = _mm_macc_pd(dx13,fscal,fjx3);
1896 fjy3 = _mm_macc_pd(dy13,fscal,fjy3);
1897 fjz3 = _mm_macc_pd(dz13,fscal,fjz3);
1898
1899 }
1900
1901 /**************************
1902 * CALCULATE INTERACTIONS *
1903 **************************/
1904
1905 if (gmx_mm_any_lt(rsq21,rcutoff2))
1906 {
1907
1908 r21 = _mm_mul_pd(rsq21,rinv21);
1909
1910 /* EWALD ELECTROSTATICS */
1911
1912 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1913 ewrt = _mm_mul_pd(r21,ewtabscale);
1914 ewitab = _mm_cvttpd_epi32(ewrt);
1915 #ifdef __XOP__
1916 eweps = _mm_frcz_pd(ewrt);
1917 #else
1918 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1919 #endif
1920 twoeweps = _mm_add_pd(eweps,eweps);
1921 gmx_mm_load_2pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),ewtab+_mm_extract_epi32(ewitab,1),
1922 &ewtabF,&ewtabFn);
1923 felec = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
1924 felec = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
1925
1926 cutoff_mask = _mm_cmplt_pd(rsq21,rcutoff2);
1927
1928 fscal = felec;
1929
1930 fscal = _mm_and_pd(fscal,cutoff_mask);
1931
1932 /* Update vectorial force */
1933 fix2 = _mm_macc_pd(dx21,fscal,fix2);
1934 fiy2 = _mm_macc_pd(dy21,fscal,fiy2);
1935 fiz2 = _mm_macc_pd(dz21,fscal,fiz2);
1936
1937 fjx1 = _mm_macc_pd(dx21,fscal,fjx1);
1938 fjy1 = _mm_macc_pd(dy21,fscal,fjy1);
1939 fjz1 = _mm_macc_pd(dz21,fscal,fjz1);
1940
1941 }
1942
1943 /**************************
1944 * CALCULATE INTERACTIONS *
1945 **************************/
1946
1947 if (gmx_mm_any_lt(rsq22,rcutoff2))
1948 {
1949
1950 r22 = _mm_mul_pd(rsq22,rinv22);
1951
1952 /* EWALD ELECTROSTATICS */
1953
1954 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1955 ewrt = _mm_mul_pd(r22,ewtabscale);
1956 ewitab = _mm_cvttpd_epi32(ewrt);
1957 #ifdef __XOP__
1958 eweps = _mm_frcz_pd(ewrt);
1959 #else
1960 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1961 #endif
1962 twoeweps = _mm_add_pd(eweps,eweps);
1963 gmx_mm_load_2pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),ewtab+_mm_extract_epi32(ewitab,1),
1964 &ewtabF,&ewtabFn);
1965 felec = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
1966 felec = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
1967
1968 cutoff_mask = _mm_cmplt_pd(rsq22,rcutoff2);
1969
1970 fscal = felec;
1971
1972 fscal = _mm_and_pd(fscal,cutoff_mask);
1973
1974 /* Update vectorial force */
1975 fix2 = _mm_macc_pd(dx22,fscal,fix2);
1976 fiy2 = _mm_macc_pd(dy22,fscal,fiy2);
1977 fiz2 = _mm_macc_pd(dz22,fscal,fiz2);
1978
1979 fjx2 = _mm_macc_pd(dx22,fscal,fjx2);
1980 fjy2 = _mm_macc_pd(dy22,fscal,fjy2);
1981 fjz2 = _mm_macc_pd(dz22,fscal,fjz2);
1982
1983 }
1984
1985 /**************************
1986 * CALCULATE INTERACTIONS *
1987 **************************/
1988
1989 if (gmx_mm_any_lt(rsq23,rcutoff2))
1990 {
1991
1992 r23 = _mm_mul_pd(rsq23,rinv23);
1993
1994 /* EWALD ELECTROSTATICS */
1995
1996 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1997 ewrt = _mm_mul_pd(r23,ewtabscale);
1998 ewitab = _mm_cvttpd_epi32(ewrt);
1999 #ifdef __XOP__
2000 eweps = _mm_frcz_pd(ewrt);
2001 #else
2002 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2003 #endif
2004 twoeweps = _mm_add_pd(eweps,eweps);
2005 gmx_mm_load_2pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),ewtab+_mm_extract_epi32(ewitab,1),
2006 &ewtabF,&ewtabFn);
2007 felec = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
2008 felec = _mm_mul_pd(_mm_mul_pd(qq23,rinv23),_mm_sub_pd(rinvsq23,felec));
2009
2010 cutoff_mask = _mm_cmplt_pd(rsq23,rcutoff2);
2011
2012 fscal = felec;
2013
2014 fscal = _mm_and_pd(fscal,cutoff_mask);
2015
2016 /* Update vectorial force */
2017 fix2 = _mm_macc_pd(dx23,fscal,fix2);
2018 fiy2 = _mm_macc_pd(dy23,fscal,fiy2);
2019 fiz2 = _mm_macc_pd(dz23,fscal,fiz2);
2020
2021 fjx3 = _mm_macc_pd(dx23,fscal,fjx3);
2022 fjy3 = _mm_macc_pd(dy23,fscal,fjy3);
2023 fjz3 = _mm_macc_pd(dz23,fscal,fjz3);
2024
2025 }
2026
2027 /**************************
2028 * CALCULATE INTERACTIONS *
2029 **************************/
2030
2031 if (gmx_mm_any_lt(rsq31,rcutoff2))
2032 {
2033
2034 r31 = _mm_mul_pd(rsq31,rinv31);
2035
2036 /* EWALD ELECTROSTATICS */
2037
2038 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2039 ewrt = _mm_mul_pd(r31,ewtabscale);
2040 ewitab = _mm_cvttpd_epi32(ewrt);
2041 #ifdef __XOP__
2042 eweps = _mm_frcz_pd(ewrt);
2043 #else
2044 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2045 #endif
2046 twoeweps = _mm_add_pd(eweps,eweps);
2047 gmx_mm_load_2pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),ewtab+_mm_extract_epi32(ewitab,1),
2048 &ewtabF,&ewtabFn);
2049 felec = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
2050 felec = _mm_mul_pd(_mm_mul_pd(qq31,rinv31),_mm_sub_pd(rinvsq31,felec));
2051
2052 cutoff_mask = _mm_cmplt_pd(rsq31,rcutoff2);
2053
2054 fscal = felec;
2055
2056 fscal = _mm_and_pd(fscal,cutoff_mask);
2057
2058 /* Update vectorial force */
2059 fix3 = _mm_macc_pd(dx31,fscal,fix3);
2060 fiy3 = _mm_macc_pd(dy31,fscal,fiy3);
2061 fiz3 = _mm_macc_pd(dz31,fscal,fiz3);
2062
2063 fjx1 = _mm_macc_pd(dx31,fscal,fjx1);
2064 fjy1 = _mm_macc_pd(dy31,fscal,fjy1);
2065 fjz1 = _mm_macc_pd(dz31,fscal,fjz1);
2066
2067 }
2068
2069 /**************************
2070 * CALCULATE INTERACTIONS *
2071 **************************/
2072
2073 if (gmx_mm_any_lt(rsq32,rcutoff2))
2074 {
2075
2076 r32 = _mm_mul_pd(rsq32,rinv32);
2077
2078 /* EWALD ELECTROSTATICS */
2079
2080 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2081 ewrt = _mm_mul_pd(r32,ewtabscale);
2082 ewitab = _mm_cvttpd_epi32(ewrt);
2083 #ifdef __XOP__
2084 eweps = _mm_frcz_pd(ewrt);
2085 #else
2086 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2087 #endif
2088 twoeweps = _mm_add_pd(eweps,eweps);
2089 gmx_mm_load_2pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),ewtab+_mm_extract_epi32(ewitab,1),
2090 &ewtabF,&ewtabFn);
2091 felec = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
2092 felec = _mm_mul_pd(_mm_mul_pd(qq32,rinv32),_mm_sub_pd(rinvsq32,felec));
2093
2094 cutoff_mask = _mm_cmplt_pd(rsq32,rcutoff2);
2095
2096 fscal = felec;
2097
2098 fscal = _mm_and_pd(fscal,cutoff_mask);
2099
2100 /* Update vectorial force */
2101 fix3 = _mm_macc_pd(dx32,fscal,fix3);
2102 fiy3 = _mm_macc_pd(dy32,fscal,fiy3);
2103 fiz3 = _mm_macc_pd(dz32,fscal,fiz3);
2104
2105 fjx2 = _mm_macc_pd(dx32,fscal,fjx2);
2106 fjy2 = _mm_macc_pd(dy32,fscal,fjy2);
2107 fjz2 = _mm_macc_pd(dz32,fscal,fjz2);
2108
2109 }
2110
2111 /**************************
2112 * CALCULATE INTERACTIONS *
2113 **************************/
2114
2115 if (gmx_mm_any_lt(rsq33,rcutoff2))
2116 {
2117
2118 r33 = _mm_mul_pd(rsq33,rinv33);
2119
2120 /* EWALD ELECTROSTATICS */
2121
2122 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2123 ewrt = _mm_mul_pd(r33,ewtabscale);
2124 ewitab = _mm_cvttpd_epi32(ewrt);
2125 #ifdef __XOP__
2126 eweps = _mm_frcz_pd(ewrt);
2127 #else
2128 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2129 #endif
2130 twoeweps = _mm_add_pd(eweps,eweps);
2131 gmx_mm_load_2pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),ewtab+_mm_extract_epi32(ewitab,1),
2132 &ewtabF,&ewtabFn);
2133 felec = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
2134 felec = _mm_mul_pd(_mm_mul_pd(qq33,rinv33),_mm_sub_pd(rinvsq33,felec));
2135
2136 cutoff_mask = _mm_cmplt_pd(rsq33,rcutoff2);
2137
2138 fscal = felec;
2139
2140 fscal = _mm_and_pd(fscal,cutoff_mask);
2141
2142 /* Update vectorial force */
2143 fix3 = _mm_macc_pd(dx33,fscal,fix3);
2144 fiy3 = _mm_macc_pd(dy33,fscal,fiy3);
2145 fiz3 = _mm_macc_pd(dz33,fscal,fiz3);
2146
2147 fjx3 = _mm_macc_pd(dx33,fscal,fjx3);
2148 fjy3 = _mm_macc_pd(dy33,fscal,fjy3);
2149 fjz3 = _mm_macc_pd(dz33,fscal,fjz3);
2150
2151 }
2152
2153 gmx_mm_decrement_4rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2154
2155 /* Inner loop uses 414 flops */
2156 }
2157
2158 if(jidx<j_index_end)
2159 {
2160
2161 jnrA = jjnr[jidx];
2162 j_coord_offsetA = DIM*jnrA;
2163
2164 /* load j atom coordinates */
2165 gmx_mm_load_4rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
2166 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
2167 &jy2,&jz2,&jx3,&jy3,&jz3);
2168
2169 /* Calculate displacement vector */
2170 dx00 = _mm_sub_pd(ix0,jx0);
2171 dy00 = _mm_sub_pd(iy0,jy0);
2172 dz00 = _mm_sub_pd(iz0,jz0);
2173 dx11 = _mm_sub_pd(ix1,jx1);
2174 dy11 = _mm_sub_pd(iy1,jy1);
2175 dz11 = _mm_sub_pd(iz1,jz1);
2176 dx12 = _mm_sub_pd(ix1,jx2);
2177 dy12 = _mm_sub_pd(iy1,jy2);
2178 dz12 = _mm_sub_pd(iz1,jz2);
2179 dx13 = _mm_sub_pd(ix1,jx3);
2180 dy13 = _mm_sub_pd(iy1,jy3);
2181 dz13 = _mm_sub_pd(iz1,jz3);
2182 dx21 = _mm_sub_pd(ix2,jx1);
2183 dy21 = _mm_sub_pd(iy2,jy1);
2184 dz21 = _mm_sub_pd(iz2,jz1);
2185 dx22 = _mm_sub_pd(ix2,jx2);
2186 dy22 = _mm_sub_pd(iy2,jy2);
2187 dz22 = _mm_sub_pd(iz2,jz2);
2188 dx23 = _mm_sub_pd(ix2,jx3);
2189 dy23 = _mm_sub_pd(iy2,jy3);
2190 dz23 = _mm_sub_pd(iz2,jz3);
2191 dx31 = _mm_sub_pd(ix3,jx1);
2192 dy31 = _mm_sub_pd(iy3,jy1);
2193 dz31 = _mm_sub_pd(iz3,jz1);
2194 dx32 = _mm_sub_pd(ix3,jx2);
2195 dy32 = _mm_sub_pd(iy3,jy2);
2196 dz32 = _mm_sub_pd(iz3,jz2);
2197 dx33 = _mm_sub_pd(ix3,jx3);
2198 dy33 = _mm_sub_pd(iy3,jy3);
2199 dz33 = _mm_sub_pd(iz3,jz3);
2200
2201 /* Calculate squared distance and things based on it */
2202 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
2203 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
2204 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
2205 rsq13 = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
2206 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
2207 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
2208 rsq23 = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
2209 rsq31 = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
2210 rsq32 = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
2211 rsq33 = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
2212
2213 rinv11 = gmx_mm_invsqrt_pd(rsq11);
2214 rinv12 = gmx_mm_invsqrt_pd(rsq12);
2215 rinv13 = gmx_mm_invsqrt_pd(rsq13);
2216 rinv21 = gmx_mm_invsqrt_pd(rsq21);
2217 rinv22 = gmx_mm_invsqrt_pd(rsq22);
2218 rinv23 = gmx_mm_invsqrt_pd(rsq23);
2219 rinv31 = gmx_mm_invsqrt_pd(rsq31);
2220 rinv32 = gmx_mm_invsqrt_pd(rsq32);
2221 rinv33 = gmx_mm_invsqrt_pd(rsq33);
2222
2223 rinvsq00 = gmx_mm_inv_pd(rsq00);
2224 rinvsq11 = _mm_mul_pd(rinv11,rinv11);
2225 rinvsq12 = _mm_mul_pd(rinv12,rinv12);
2226 rinvsq13 = _mm_mul_pd(rinv13,rinv13);
2227 rinvsq21 = _mm_mul_pd(rinv21,rinv21);
2228 rinvsq22 = _mm_mul_pd(rinv22,rinv22);
2229 rinvsq23 = _mm_mul_pd(rinv23,rinv23);
2230 rinvsq31 = _mm_mul_pd(rinv31,rinv31);
2231 rinvsq32 = _mm_mul_pd(rinv32,rinv32);
2232 rinvsq33 = _mm_mul_pd(rinv33,rinv33);
2233
2234 fjx0 = _mm_setzero_pd();
2235 fjy0 = _mm_setzero_pd();
2236 fjz0 = _mm_setzero_pd();
2237 fjx1 = _mm_setzero_pd();
2238 fjy1 = _mm_setzero_pd();
2239 fjz1 = _mm_setzero_pd();
2240 fjx2 = _mm_setzero_pd();
2241 fjy2 = _mm_setzero_pd();
2242 fjz2 = _mm_setzero_pd();
2243 fjx3 = _mm_setzero_pd();
2244 fjy3 = _mm_setzero_pd();
2245 fjz3 = _mm_setzero_pd();
2246
2247 /**************************
2248 * CALCULATE INTERACTIONS *
2249 **************************/
2250
2251 if (gmx_mm_any_lt(rsq00,rcutoff2))
2252 {
2253
2254 /* LENNARD-JONES DISPERSION/REPULSION */
2255
2256 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
2257 fvdw = _mm_mul_pd(_mm_msub_pd(c12_00,rinvsix,c6_00),_mm_mul_pd(rinvsix,rinvsq00));
2258
2259 cutoff_mask = _mm_cmplt_pd(rsq00,rcutoff2);
2260
2261 fscal = fvdw;
2262
2263 fscal = _mm_and_pd(fscal,cutoff_mask);
2264
2265 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2266
2267 /* Update vectorial force */
2268 fix0 = _mm_macc_pd(dx00,fscal,fix0);
2269 fiy0 = _mm_macc_pd(dy00,fscal,fiy0);
2270 fiz0 = _mm_macc_pd(dz00,fscal,fiz0);
2271
2272 fjx0 = _mm_macc_pd(dx00,fscal,fjx0);
2273 fjy0 = _mm_macc_pd(dy00,fscal,fjy0);
2274 fjz0 = _mm_macc_pd(dz00,fscal,fjz0);
2275
2276 }
2277
2278 /**************************
2279 * CALCULATE INTERACTIONS *
2280 **************************/
2281
2282 if (gmx_mm_any_lt(rsq11,rcutoff2))
2283 {
2284
2285 r11 = _mm_mul_pd(rsq11,rinv11);
2286
2287 /* EWALD ELECTROSTATICS */
2288
2289 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2290 ewrt = _mm_mul_pd(r11,ewtabscale);
2291 ewitab = _mm_cvttpd_epi32(ewrt);
2292 #ifdef __XOP__
2293 eweps = _mm_frcz_pd(ewrt);
2294 #else
2295 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2296 #endif
2297 twoeweps = _mm_add_pd(eweps,eweps);
2298 gmx_mm_load_1pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2299 felec = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
2300 felec = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
2301
2302 cutoff_mask = _mm_cmplt_pd(rsq11,rcutoff2);
2303
2304 fscal = felec;
2305
2306 fscal = _mm_and_pd(fscal,cutoff_mask);
2307
2308 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2309
2310 /* Update vectorial force */
2311 fix1 = _mm_macc_pd(dx11,fscal,fix1);
2312 fiy1 = _mm_macc_pd(dy11,fscal,fiy1);
2313 fiz1 = _mm_macc_pd(dz11,fscal,fiz1);
2314
2315 fjx1 = _mm_macc_pd(dx11,fscal,fjx1);
2316 fjy1 = _mm_macc_pd(dy11,fscal,fjy1);
2317 fjz1 = _mm_macc_pd(dz11,fscal,fjz1);
2318
2319 }
2320
2321 /**************************
2322 * CALCULATE INTERACTIONS *
2323 **************************/
2324
2325 if (gmx_mm_any_lt(rsq12,rcutoff2))
2326 {
2327
2328 r12 = _mm_mul_pd(rsq12,rinv12);
2329
2330 /* EWALD ELECTROSTATICS */
2331
2332 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2333 ewrt = _mm_mul_pd(r12,ewtabscale);
2334 ewitab = _mm_cvttpd_epi32(ewrt);
2335 #ifdef __XOP__
2336 eweps = _mm_frcz_pd(ewrt);
2337 #else
2338 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2339 #endif
2340 twoeweps = _mm_add_pd(eweps,eweps);
2341 gmx_mm_load_1pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2342 felec = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
2343 felec = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
2344
2345 cutoff_mask = _mm_cmplt_pd(rsq12,rcutoff2);
2346
2347 fscal = felec;
2348
2349 fscal = _mm_and_pd(fscal,cutoff_mask);
2350
2351 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2352
2353 /* Update vectorial force */
2354 fix1 = _mm_macc_pd(dx12,fscal,fix1);
2355 fiy1 = _mm_macc_pd(dy12,fscal,fiy1);
2356 fiz1 = _mm_macc_pd(dz12,fscal,fiz1);
2357
2358 fjx2 = _mm_macc_pd(dx12,fscal,fjx2);
2359 fjy2 = _mm_macc_pd(dy12,fscal,fjy2);
2360 fjz2 = _mm_macc_pd(dz12,fscal,fjz2);
2361
2362 }
2363
2364 /**************************
2365 * CALCULATE INTERACTIONS *
2366 **************************/
2367
2368 if (gmx_mm_any_lt(rsq13,rcutoff2))
2369 {
2370
2371 r13 = _mm_mul_pd(rsq13,rinv13);
2372
2373 /* EWALD ELECTROSTATICS */
2374
2375 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2376 ewrt = _mm_mul_pd(r13,ewtabscale);
2377 ewitab = _mm_cvttpd_epi32(ewrt);
2378 #ifdef __XOP__
2379 eweps = _mm_frcz_pd(ewrt);
2380 #else
2381 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2382 #endif
2383 twoeweps = _mm_add_pd(eweps,eweps);
2384 gmx_mm_load_1pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2385 felec = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
2386 felec = _mm_mul_pd(_mm_mul_pd(qq13,rinv13),_mm_sub_pd(rinvsq13,felec));
2387
2388 cutoff_mask = _mm_cmplt_pd(rsq13,rcutoff2);
2389
2390 fscal = felec;
2391
2392 fscal = _mm_and_pd(fscal,cutoff_mask);
2393
2394 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2395
2396 /* Update vectorial force */
2397 fix1 = _mm_macc_pd(dx13,fscal,fix1);
2398 fiy1 = _mm_macc_pd(dy13,fscal,fiy1);
2399 fiz1 = _mm_macc_pd(dz13,fscal,fiz1);
2400
2401 fjx3 = _mm_macc_pd(dx13,fscal,fjx3);
2402 fjy3 = _mm_macc_pd(dy13,fscal,fjy3);
2403 fjz3 = _mm_macc_pd(dz13,fscal,fjz3);
2404
2405 }
2406
2407 /**************************
2408 * CALCULATE INTERACTIONS *
2409 **************************/
2410
2411 if (gmx_mm_any_lt(rsq21,rcutoff2))
2412 {
2413
2414 r21 = _mm_mul_pd(rsq21,rinv21);
2415
2416 /* EWALD ELECTROSTATICS */
2417
2418 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2419 ewrt = _mm_mul_pd(r21,ewtabscale);
2420 ewitab = _mm_cvttpd_epi32(ewrt);
2421 #ifdef __XOP__
2422 eweps = _mm_frcz_pd(ewrt);
2423 #else
2424 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2425 #endif
2426 twoeweps = _mm_add_pd(eweps,eweps);
2427 gmx_mm_load_1pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2428 felec = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
2429 felec = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
2430
2431 cutoff_mask = _mm_cmplt_pd(rsq21,rcutoff2);
2432
2433 fscal = felec;
2434
2435 fscal = _mm_and_pd(fscal,cutoff_mask);
2436
2437 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2438
2439 /* Update vectorial force */
2440 fix2 = _mm_macc_pd(dx21,fscal,fix2);
2441 fiy2 = _mm_macc_pd(dy21,fscal,fiy2);
2442 fiz2 = _mm_macc_pd(dz21,fscal,fiz2);
2443
2444 fjx1 = _mm_macc_pd(dx21,fscal,fjx1);
2445 fjy1 = _mm_macc_pd(dy21,fscal,fjy1);
2446 fjz1 = _mm_macc_pd(dz21,fscal,fjz1);
2447
2448 }
2449
2450 /**************************
2451 * CALCULATE INTERACTIONS *
2452 **************************/
2453
2454 if (gmx_mm_any_lt(rsq22,rcutoff2))
2455 {
2456
2457 r22 = _mm_mul_pd(rsq22,rinv22);
2458
2459 /* EWALD ELECTROSTATICS */
2460
2461 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2462 ewrt = _mm_mul_pd(r22,ewtabscale);
2463 ewitab = _mm_cvttpd_epi32(ewrt);
2464 #ifdef __XOP__
2465 eweps = _mm_frcz_pd(ewrt);
2466 #else
2467 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2468 #endif
2469 twoeweps = _mm_add_pd(eweps,eweps);
2470 gmx_mm_load_1pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2471 felec = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
2472 felec = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
2473
2474 cutoff_mask = _mm_cmplt_pd(rsq22,rcutoff2);
2475
2476 fscal = felec;
2477
2478 fscal = _mm_and_pd(fscal,cutoff_mask);
2479
2480 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2481
2482 /* Update vectorial force */
2483 fix2 = _mm_macc_pd(dx22,fscal,fix2);
2484 fiy2 = _mm_macc_pd(dy22,fscal,fiy2);
2485 fiz2 = _mm_macc_pd(dz22,fscal,fiz2);
2486
2487 fjx2 = _mm_macc_pd(dx22,fscal,fjx2);
2488 fjy2 = _mm_macc_pd(dy22,fscal,fjy2);
2489 fjz2 = _mm_macc_pd(dz22,fscal,fjz2);
2490
2491 }
2492
2493 /**************************
2494 * CALCULATE INTERACTIONS *
2495 **************************/
2496
2497 if (gmx_mm_any_lt(rsq23,rcutoff2))
2498 {
2499
2500 r23 = _mm_mul_pd(rsq23,rinv23);
2501
2502 /* EWALD ELECTROSTATICS */
2503
2504 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2505 ewrt = _mm_mul_pd(r23,ewtabscale);
2506 ewitab = _mm_cvttpd_epi32(ewrt);
2507 #ifdef __XOP__
2508 eweps = _mm_frcz_pd(ewrt);
2509 #else
2510 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2511 #endif
2512 twoeweps = _mm_add_pd(eweps,eweps);
2513 gmx_mm_load_1pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2514 felec = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
2515 felec = _mm_mul_pd(_mm_mul_pd(qq23,rinv23),_mm_sub_pd(rinvsq23,felec));
2516
2517 cutoff_mask = _mm_cmplt_pd(rsq23,rcutoff2);
2518
2519 fscal = felec;
2520
2521 fscal = _mm_and_pd(fscal,cutoff_mask);
2522
2523 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2524
2525 /* Update vectorial force */
2526 fix2 = _mm_macc_pd(dx23,fscal,fix2);
2527 fiy2 = _mm_macc_pd(dy23,fscal,fiy2);
2528 fiz2 = _mm_macc_pd(dz23,fscal,fiz2);
2529
2530 fjx3 = _mm_macc_pd(dx23,fscal,fjx3);
2531 fjy3 = _mm_macc_pd(dy23,fscal,fjy3);
2532 fjz3 = _mm_macc_pd(dz23,fscal,fjz3);
2533
2534 }
2535
2536 /**************************
2537 * CALCULATE INTERACTIONS *
2538 **************************/
2539
2540 if (gmx_mm_any_lt(rsq31,rcutoff2))
2541 {
2542
2543 r31 = _mm_mul_pd(rsq31,rinv31);
2544
2545 /* EWALD ELECTROSTATICS */
2546
2547 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2548 ewrt = _mm_mul_pd(r31,ewtabscale);
2549 ewitab = _mm_cvttpd_epi32(ewrt);
2550 #ifdef __XOP__
2551 eweps = _mm_frcz_pd(ewrt);
2552 #else
2553 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2554 #endif
2555 twoeweps = _mm_add_pd(eweps,eweps);
2556 gmx_mm_load_1pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2557 felec = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
2558 felec = _mm_mul_pd(_mm_mul_pd(qq31,rinv31),_mm_sub_pd(rinvsq31,felec));
2559
2560 cutoff_mask = _mm_cmplt_pd(rsq31,rcutoff2);
2561
2562 fscal = felec;
2563
2564 fscal = _mm_and_pd(fscal,cutoff_mask);
2565
2566 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2567
2568 /* Update vectorial force */
2569 fix3 = _mm_macc_pd(dx31,fscal,fix3);
2570 fiy3 = _mm_macc_pd(dy31,fscal,fiy3);
2571 fiz3 = _mm_macc_pd(dz31,fscal,fiz3);
2572
2573 fjx1 = _mm_macc_pd(dx31,fscal,fjx1);
2574 fjy1 = _mm_macc_pd(dy31,fscal,fjy1);
2575 fjz1 = _mm_macc_pd(dz31,fscal,fjz1);
2576
2577 }
2578
2579 /**************************
2580 * CALCULATE INTERACTIONS *
2581 **************************/
2582
2583 if (gmx_mm_any_lt(rsq32,rcutoff2))
2584 {
2585
2586 r32 = _mm_mul_pd(rsq32,rinv32);
2587
2588 /* EWALD ELECTROSTATICS */
2589
2590 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2591 ewrt = _mm_mul_pd(r32,ewtabscale);
2592 ewitab = _mm_cvttpd_epi32(ewrt);
2593 #ifdef __XOP__
2594 eweps = _mm_frcz_pd(ewrt);
2595 #else
2596 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2597 #endif
2598 twoeweps = _mm_add_pd(eweps,eweps);
2599 gmx_mm_load_1pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2600 felec = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
2601 felec = _mm_mul_pd(_mm_mul_pd(qq32,rinv32),_mm_sub_pd(rinvsq32,felec));
2602
2603 cutoff_mask = _mm_cmplt_pd(rsq32,rcutoff2);
2604
2605 fscal = felec;
2606
2607 fscal = _mm_and_pd(fscal,cutoff_mask);
2608
2609 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2610
2611 /* Update vectorial force */
2612 fix3 = _mm_macc_pd(dx32,fscal,fix3);
2613 fiy3 = _mm_macc_pd(dy32,fscal,fiy3);
2614 fiz3 = _mm_macc_pd(dz32,fscal,fiz3);
2615
2616 fjx2 = _mm_macc_pd(dx32,fscal,fjx2);
2617 fjy2 = _mm_macc_pd(dy32,fscal,fjy2);
2618 fjz2 = _mm_macc_pd(dz32,fscal,fjz2);
2619
2620 }
2621
2622 /**************************
2623 * CALCULATE INTERACTIONS *
2624 **************************/
2625
2626 if (gmx_mm_any_lt(rsq33,rcutoff2))
2627 {
2628
2629 r33 = _mm_mul_pd(rsq33,rinv33);
2630
2631 /* EWALD ELECTROSTATICS */
2632
2633 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2634 ewrt = _mm_mul_pd(r33,ewtabscale);
2635 ewitab = _mm_cvttpd_epi32(ewrt);
2636 #ifdef __XOP__
2637 eweps = _mm_frcz_pd(ewrt);
2638 #else
2639 eweps = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2640 #endif
2641 twoeweps = _mm_add_pd(eweps,eweps);
2642 gmx_mm_load_1pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2643 felec = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
2644 felec = _mm_mul_pd(_mm_mul_pd(qq33,rinv33),_mm_sub_pd(rinvsq33,felec));
2645
2646 cutoff_mask = _mm_cmplt_pd(rsq33,rcutoff2);
2647
2648 fscal = felec;
2649
2650 fscal = _mm_and_pd(fscal,cutoff_mask);
2651
2652 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2653
2654 /* Update vectorial force */
2655 fix3 = _mm_macc_pd(dx33,fscal,fix3);
2656 fiy3 = _mm_macc_pd(dy33,fscal,fiy3);
2657 fiz3 = _mm_macc_pd(dz33,fscal,fiz3);
2658
2659 fjx3 = _mm_macc_pd(dx33,fscal,fjx3);
2660 fjy3 = _mm_macc_pd(dy33,fscal,fjy3);
2661 fjz3 = _mm_macc_pd(dz33,fscal,fjz3);
2662
2663 }
2664
2665 gmx_mm_decrement_4rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2666
2667 /* Inner loop uses 414 flops */
2668 }
2669
2670 /* End of innermost loop */
2671
2672 gmx_mm_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2673 f+i_coord_offset,fshift+i_shift_offset);
2674
2675 /* Increment number of inner iterations */
2676 inneriter += j_index_end - j_index_start;
2677
2678 /* Outer loop uses 24 flops */
2679 }
2680
2681 /* Increment number of outer iterations */
2682 outeriter += nri;
2683
2684 /* Update outer/inner flops */
2685
2686 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*414);
2687 }
The ultimate molecular dynamics simulation package