2 * This file is part of the GROMACS molecular simulation package.
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.
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.
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.
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.
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.
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.
36 * Note: this file was generated by the GROMACS avx_128_fma_double kernel generator.
42 #include "../nb_kernel.h"
43 #include "gromacs/legacyheaders/types/simple.h"
44 #include "gromacs/math/vec.h"
45 #include "gromacs/legacyheaders/nrnb.h"
47 #include "gromacs/simd/math_x86_avx_128_fma_double.h"
48 #include "kernelutil_x86_avx_128_fma_double.h"
51 * Gromacs nonbonded kernel: nb_kernel_ElecCSTab_VdwCSTab_GeomW4W4_VF_avx_128_fma_double
52 * Electrostatics interaction: CubicSplineTable
53 * VdW interaction: CubicSplineTable
54 * Geometry: Water4-Water4
55 * Calculate force/pot: PotentialAndForce
58 nb_kernel_ElecCSTab_VdwCSTab_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)
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.
72 int i_shift_offset,i_coord_offset,outeriter,inneriter;
73 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
75 int j_coord_offsetA,j_coord_offsetB;
76 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
78 real *shiftvec,*fshift,*x,*f;
79 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
81 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
83 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
85 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
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;
109 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
112 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
113 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
115 __m128i ifour = _mm_set1_epi32(4);
116 __m128d rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
118 __m128d dummy_mask,cutoff_mask;
119 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
120 __m128d one = _mm_set1_pd(1.0);
121 __m128d two = _mm_set1_pd(2.0);
127 jindex = nlist->jindex;
129 shiftidx = nlist->shift;
131 shiftvec = fr->shift_vec[0];
132 fshift = fr->fshift[0];
133 facel = _mm_set1_pd(fr->epsfac);
134 charge = mdatoms->chargeA;
135 nvdwtype = fr->ntype;
137 vdwtype = mdatoms->typeA;
139 vftab = kernel_data->table_elec_vdw->data;
140 vftabscale = _mm_set1_pd(kernel_data->table_elec_vdw->scale);
142 /* Setup water-specific parameters */
143 inr = nlist->iinr[0];
144 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
145 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
146 iq3 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
147 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
149 jq1 = _mm_set1_pd(charge[inr+1]);
150 jq2 = _mm_set1_pd(charge[inr+2]);
151 jq3 = _mm_set1_pd(charge[inr+3]);
152 vdwjidx0A = 2*vdwtype[inr+0];
153 c6_00 = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
154 c12_00 = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
155 qq11 = _mm_mul_pd(iq1,jq1);
156 qq12 = _mm_mul_pd(iq1,jq2);
157 qq13 = _mm_mul_pd(iq1,jq3);
158 qq21 = _mm_mul_pd(iq2,jq1);
159 qq22 = _mm_mul_pd(iq2,jq2);
160 qq23 = _mm_mul_pd(iq2,jq3);
161 qq31 = _mm_mul_pd(iq3,jq1);
162 qq32 = _mm_mul_pd(iq3,jq2);
163 qq33 = _mm_mul_pd(iq3,jq3);
165 /* Avoid stupid compiler warnings */
173 /* Start outer loop over neighborlists */
174 for(iidx=0; iidx<nri; iidx++)
176 /* Load shift vector for this list */
177 i_shift_offset = DIM*shiftidx[iidx];
179 /* Load limits for loop over neighbors */
180 j_index_start = jindex[iidx];
181 j_index_end = jindex[iidx+1];
183 /* Get outer coordinate index */
185 i_coord_offset = DIM*inr;
187 /* Load i particle coords and add shift vector */
188 gmx_mm_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
189 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
191 fix0 = _mm_setzero_pd();
192 fiy0 = _mm_setzero_pd();
193 fiz0 = _mm_setzero_pd();
194 fix1 = _mm_setzero_pd();
195 fiy1 = _mm_setzero_pd();
196 fiz1 = _mm_setzero_pd();
197 fix2 = _mm_setzero_pd();
198 fiy2 = _mm_setzero_pd();
199 fiz2 = _mm_setzero_pd();
200 fix3 = _mm_setzero_pd();
201 fiy3 = _mm_setzero_pd();
202 fiz3 = _mm_setzero_pd();
204 /* Reset potential sums */
205 velecsum = _mm_setzero_pd();
206 vvdwsum = _mm_setzero_pd();
208 /* Start inner kernel loop */
209 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
212 /* Get j neighbor index, and coordinate index */
215 j_coord_offsetA = DIM*jnrA;
216 j_coord_offsetB = DIM*jnrB;
218 /* load j atom coordinates */
219 gmx_mm_load_4rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
220 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
221 &jy2,&jz2,&jx3,&jy3,&jz3);
223 /* Calculate displacement vector */
224 dx00 = _mm_sub_pd(ix0,jx0);
225 dy00 = _mm_sub_pd(iy0,jy0);
226 dz00 = _mm_sub_pd(iz0,jz0);
227 dx11 = _mm_sub_pd(ix1,jx1);
228 dy11 = _mm_sub_pd(iy1,jy1);
229 dz11 = _mm_sub_pd(iz1,jz1);
230 dx12 = _mm_sub_pd(ix1,jx2);
231 dy12 = _mm_sub_pd(iy1,jy2);
232 dz12 = _mm_sub_pd(iz1,jz2);
233 dx13 = _mm_sub_pd(ix1,jx3);
234 dy13 = _mm_sub_pd(iy1,jy3);
235 dz13 = _mm_sub_pd(iz1,jz3);
236 dx21 = _mm_sub_pd(ix2,jx1);
237 dy21 = _mm_sub_pd(iy2,jy1);
238 dz21 = _mm_sub_pd(iz2,jz1);
239 dx22 = _mm_sub_pd(ix2,jx2);
240 dy22 = _mm_sub_pd(iy2,jy2);
241 dz22 = _mm_sub_pd(iz2,jz2);
242 dx23 = _mm_sub_pd(ix2,jx3);
243 dy23 = _mm_sub_pd(iy2,jy3);
244 dz23 = _mm_sub_pd(iz2,jz3);
245 dx31 = _mm_sub_pd(ix3,jx1);
246 dy31 = _mm_sub_pd(iy3,jy1);
247 dz31 = _mm_sub_pd(iz3,jz1);
248 dx32 = _mm_sub_pd(ix3,jx2);
249 dy32 = _mm_sub_pd(iy3,jy2);
250 dz32 = _mm_sub_pd(iz3,jz2);
251 dx33 = _mm_sub_pd(ix3,jx3);
252 dy33 = _mm_sub_pd(iy3,jy3);
253 dz33 = _mm_sub_pd(iz3,jz3);
255 /* Calculate squared distance and things based on it */
256 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
257 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
258 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
259 rsq13 = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
260 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
261 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
262 rsq23 = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
263 rsq31 = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
264 rsq32 = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
265 rsq33 = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
267 rinv00 = gmx_mm_invsqrt_pd(rsq00);
268 rinv11 = gmx_mm_invsqrt_pd(rsq11);
269 rinv12 = gmx_mm_invsqrt_pd(rsq12);
270 rinv13 = gmx_mm_invsqrt_pd(rsq13);
271 rinv21 = gmx_mm_invsqrt_pd(rsq21);
272 rinv22 = gmx_mm_invsqrt_pd(rsq22);
273 rinv23 = gmx_mm_invsqrt_pd(rsq23);
274 rinv31 = gmx_mm_invsqrt_pd(rsq31);
275 rinv32 = gmx_mm_invsqrt_pd(rsq32);
276 rinv33 = gmx_mm_invsqrt_pd(rsq33);
278 fjx0 = _mm_setzero_pd();
279 fjy0 = _mm_setzero_pd();
280 fjz0 = _mm_setzero_pd();
281 fjx1 = _mm_setzero_pd();
282 fjy1 = _mm_setzero_pd();
283 fjz1 = _mm_setzero_pd();
284 fjx2 = _mm_setzero_pd();
285 fjy2 = _mm_setzero_pd();
286 fjz2 = _mm_setzero_pd();
287 fjx3 = _mm_setzero_pd();
288 fjy3 = _mm_setzero_pd();
289 fjz3 = _mm_setzero_pd();
291 /**************************
292 * CALCULATE INTERACTIONS *
293 **************************/
295 r00 = _mm_mul_pd(rsq00,rinv00);
297 /* Calculate table index by multiplying r with table scale and truncate to integer */
298 rt = _mm_mul_pd(r00,vftabscale);
299 vfitab = _mm_cvttpd_epi32(rt);
301 vfeps = _mm_frcz_pd(rt);
303 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
305 twovfeps = _mm_add_pd(vfeps,vfeps);
306 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
308 /* CUBIC SPLINE TABLE DISPERSION */
309 vfitab = _mm_add_epi32(vfitab,ifour);
310 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
311 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
312 GMX_MM_TRANSPOSE2_PD(Y,F);
313 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
314 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
315 GMX_MM_TRANSPOSE2_PD(G,H);
316 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(H,vfeps,G),F);
317 VV = _mm_macc_pd(vfeps,Fp,Y);
318 vvdw6 = _mm_mul_pd(c6_00,VV);
319 FF = _mm_macc_pd(vfeps,_mm_macc_pd(twovfeps,H,G),Fp);
320 fvdw6 = _mm_mul_pd(c6_00,FF);
322 /* CUBIC SPLINE TABLE REPULSION */
323 vfitab = _mm_add_epi32(vfitab,ifour);
324 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
325 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
326 GMX_MM_TRANSPOSE2_PD(Y,F);
327 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
328 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
329 GMX_MM_TRANSPOSE2_PD(G,H);
330 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(H,vfeps,G),F);
331 VV = _mm_macc_pd(vfeps,Fp,Y);
332 vvdw12 = _mm_mul_pd(c12_00,VV);
333 FF = _mm_macc_pd(vfeps,_mm_macc_pd(twovfeps,H,G),Fp);
334 fvdw12 = _mm_mul_pd(c12_00,FF);
335 vvdw = _mm_add_pd(vvdw12,vvdw6);
336 fvdw = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
338 /* Update potential sum for this i atom from the interaction with this j atom. */
339 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
343 /* Update vectorial force */
344 fix0 = _mm_macc_pd(dx00,fscal,fix0);
345 fiy0 = _mm_macc_pd(dy00,fscal,fiy0);
346 fiz0 = _mm_macc_pd(dz00,fscal,fiz0);
348 fjx0 = _mm_macc_pd(dx00,fscal,fjx0);
349 fjy0 = _mm_macc_pd(dy00,fscal,fjy0);
350 fjz0 = _mm_macc_pd(dz00,fscal,fjz0);
352 /**************************
353 * CALCULATE INTERACTIONS *
354 **************************/
356 r11 = _mm_mul_pd(rsq11,rinv11);
358 /* Calculate table index by multiplying r with table scale and truncate to integer */
359 rt = _mm_mul_pd(r11,vftabscale);
360 vfitab = _mm_cvttpd_epi32(rt);
362 vfeps = _mm_frcz_pd(rt);
364 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
366 twovfeps = _mm_add_pd(vfeps,vfeps);
367 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
369 /* CUBIC SPLINE TABLE ELECTROSTATICS */
370 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
371 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
372 GMX_MM_TRANSPOSE2_PD(Y,F);
373 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
374 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
375 GMX_MM_TRANSPOSE2_PD(G,H);
376 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
377 VV = _mm_macc_pd(vfeps,Fp,Y);
378 velec = _mm_mul_pd(qq11,VV);
379 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
380 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
382 /* Update potential sum for this i atom from the interaction with this j atom. */
383 velecsum = _mm_add_pd(velecsum,velec);
387 /* Update vectorial force */
388 fix1 = _mm_macc_pd(dx11,fscal,fix1);
389 fiy1 = _mm_macc_pd(dy11,fscal,fiy1);
390 fiz1 = _mm_macc_pd(dz11,fscal,fiz1);
392 fjx1 = _mm_macc_pd(dx11,fscal,fjx1);
393 fjy1 = _mm_macc_pd(dy11,fscal,fjy1);
394 fjz1 = _mm_macc_pd(dz11,fscal,fjz1);
396 /**************************
397 * CALCULATE INTERACTIONS *
398 **************************/
400 r12 = _mm_mul_pd(rsq12,rinv12);
402 /* Calculate table index by multiplying r with table scale and truncate to integer */
403 rt = _mm_mul_pd(r12,vftabscale);
404 vfitab = _mm_cvttpd_epi32(rt);
406 vfeps = _mm_frcz_pd(rt);
408 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
410 twovfeps = _mm_add_pd(vfeps,vfeps);
411 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
413 /* CUBIC SPLINE TABLE ELECTROSTATICS */
414 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
415 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
416 GMX_MM_TRANSPOSE2_PD(Y,F);
417 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
418 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
419 GMX_MM_TRANSPOSE2_PD(G,H);
420 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
421 VV = _mm_macc_pd(vfeps,Fp,Y);
422 velec = _mm_mul_pd(qq12,VV);
423 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
424 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
426 /* Update potential sum for this i atom from the interaction with this j atom. */
427 velecsum = _mm_add_pd(velecsum,velec);
431 /* Update vectorial force */
432 fix1 = _mm_macc_pd(dx12,fscal,fix1);
433 fiy1 = _mm_macc_pd(dy12,fscal,fiy1);
434 fiz1 = _mm_macc_pd(dz12,fscal,fiz1);
436 fjx2 = _mm_macc_pd(dx12,fscal,fjx2);
437 fjy2 = _mm_macc_pd(dy12,fscal,fjy2);
438 fjz2 = _mm_macc_pd(dz12,fscal,fjz2);
440 /**************************
441 * CALCULATE INTERACTIONS *
442 **************************/
444 r13 = _mm_mul_pd(rsq13,rinv13);
446 /* Calculate table index by multiplying r with table scale and truncate to integer */
447 rt = _mm_mul_pd(r13,vftabscale);
448 vfitab = _mm_cvttpd_epi32(rt);
450 vfeps = _mm_frcz_pd(rt);
452 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
454 twovfeps = _mm_add_pd(vfeps,vfeps);
455 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
457 /* CUBIC SPLINE TABLE ELECTROSTATICS */
458 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
459 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
460 GMX_MM_TRANSPOSE2_PD(Y,F);
461 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
462 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
463 GMX_MM_TRANSPOSE2_PD(G,H);
464 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
465 VV = _mm_macc_pd(vfeps,Fp,Y);
466 velec = _mm_mul_pd(qq13,VV);
467 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
468 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq13,FF),_mm_mul_pd(vftabscale,rinv13)));
470 /* Update potential sum for this i atom from the interaction with this j atom. */
471 velecsum = _mm_add_pd(velecsum,velec);
475 /* Update vectorial force */
476 fix1 = _mm_macc_pd(dx13,fscal,fix1);
477 fiy1 = _mm_macc_pd(dy13,fscal,fiy1);
478 fiz1 = _mm_macc_pd(dz13,fscal,fiz1);
480 fjx3 = _mm_macc_pd(dx13,fscal,fjx3);
481 fjy3 = _mm_macc_pd(dy13,fscal,fjy3);
482 fjz3 = _mm_macc_pd(dz13,fscal,fjz3);
484 /**************************
485 * CALCULATE INTERACTIONS *
486 **************************/
488 r21 = _mm_mul_pd(rsq21,rinv21);
490 /* Calculate table index by multiplying r with table scale and truncate to integer */
491 rt = _mm_mul_pd(r21,vftabscale);
492 vfitab = _mm_cvttpd_epi32(rt);
494 vfeps = _mm_frcz_pd(rt);
496 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
498 twovfeps = _mm_add_pd(vfeps,vfeps);
499 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
501 /* CUBIC SPLINE TABLE ELECTROSTATICS */
502 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
503 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
504 GMX_MM_TRANSPOSE2_PD(Y,F);
505 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
506 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
507 GMX_MM_TRANSPOSE2_PD(G,H);
508 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
509 VV = _mm_macc_pd(vfeps,Fp,Y);
510 velec = _mm_mul_pd(qq21,VV);
511 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
512 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
514 /* Update potential sum for this i atom from the interaction with this j atom. */
515 velecsum = _mm_add_pd(velecsum,velec);
519 /* Update vectorial force */
520 fix2 = _mm_macc_pd(dx21,fscal,fix2);
521 fiy2 = _mm_macc_pd(dy21,fscal,fiy2);
522 fiz2 = _mm_macc_pd(dz21,fscal,fiz2);
524 fjx1 = _mm_macc_pd(dx21,fscal,fjx1);
525 fjy1 = _mm_macc_pd(dy21,fscal,fjy1);
526 fjz1 = _mm_macc_pd(dz21,fscal,fjz1);
528 /**************************
529 * CALCULATE INTERACTIONS *
530 **************************/
532 r22 = _mm_mul_pd(rsq22,rinv22);
534 /* Calculate table index by multiplying r with table scale and truncate to integer */
535 rt = _mm_mul_pd(r22,vftabscale);
536 vfitab = _mm_cvttpd_epi32(rt);
538 vfeps = _mm_frcz_pd(rt);
540 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
542 twovfeps = _mm_add_pd(vfeps,vfeps);
543 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
545 /* CUBIC SPLINE TABLE ELECTROSTATICS */
546 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
547 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
548 GMX_MM_TRANSPOSE2_PD(Y,F);
549 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
550 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
551 GMX_MM_TRANSPOSE2_PD(G,H);
552 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
553 VV = _mm_macc_pd(vfeps,Fp,Y);
554 velec = _mm_mul_pd(qq22,VV);
555 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
556 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
558 /* Update potential sum for this i atom from the interaction with this j atom. */
559 velecsum = _mm_add_pd(velecsum,velec);
563 /* Update vectorial force */
564 fix2 = _mm_macc_pd(dx22,fscal,fix2);
565 fiy2 = _mm_macc_pd(dy22,fscal,fiy2);
566 fiz2 = _mm_macc_pd(dz22,fscal,fiz2);
568 fjx2 = _mm_macc_pd(dx22,fscal,fjx2);
569 fjy2 = _mm_macc_pd(dy22,fscal,fjy2);
570 fjz2 = _mm_macc_pd(dz22,fscal,fjz2);
572 /**************************
573 * CALCULATE INTERACTIONS *
574 **************************/
576 r23 = _mm_mul_pd(rsq23,rinv23);
578 /* Calculate table index by multiplying r with table scale and truncate to integer */
579 rt = _mm_mul_pd(r23,vftabscale);
580 vfitab = _mm_cvttpd_epi32(rt);
582 vfeps = _mm_frcz_pd(rt);
584 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
586 twovfeps = _mm_add_pd(vfeps,vfeps);
587 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
589 /* CUBIC SPLINE TABLE ELECTROSTATICS */
590 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
591 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
592 GMX_MM_TRANSPOSE2_PD(Y,F);
593 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
594 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
595 GMX_MM_TRANSPOSE2_PD(G,H);
596 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
597 VV = _mm_macc_pd(vfeps,Fp,Y);
598 velec = _mm_mul_pd(qq23,VV);
599 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
600 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq23,FF),_mm_mul_pd(vftabscale,rinv23)));
602 /* Update potential sum for this i atom from the interaction with this j atom. */
603 velecsum = _mm_add_pd(velecsum,velec);
607 /* Update vectorial force */
608 fix2 = _mm_macc_pd(dx23,fscal,fix2);
609 fiy2 = _mm_macc_pd(dy23,fscal,fiy2);
610 fiz2 = _mm_macc_pd(dz23,fscal,fiz2);
612 fjx3 = _mm_macc_pd(dx23,fscal,fjx3);
613 fjy3 = _mm_macc_pd(dy23,fscal,fjy3);
614 fjz3 = _mm_macc_pd(dz23,fscal,fjz3);
616 /**************************
617 * CALCULATE INTERACTIONS *
618 **************************/
620 r31 = _mm_mul_pd(rsq31,rinv31);
622 /* Calculate table index by multiplying r with table scale and truncate to integer */
623 rt = _mm_mul_pd(r31,vftabscale);
624 vfitab = _mm_cvttpd_epi32(rt);
626 vfeps = _mm_frcz_pd(rt);
628 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
630 twovfeps = _mm_add_pd(vfeps,vfeps);
631 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
633 /* CUBIC SPLINE TABLE ELECTROSTATICS */
634 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
635 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
636 GMX_MM_TRANSPOSE2_PD(Y,F);
637 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
638 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
639 GMX_MM_TRANSPOSE2_PD(G,H);
640 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
641 VV = _mm_macc_pd(vfeps,Fp,Y);
642 velec = _mm_mul_pd(qq31,VV);
643 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
644 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq31,FF),_mm_mul_pd(vftabscale,rinv31)));
646 /* Update potential sum for this i atom from the interaction with this j atom. */
647 velecsum = _mm_add_pd(velecsum,velec);
651 /* Update vectorial force */
652 fix3 = _mm_macc_pd(dx31,fscal,fix3);
653 fiy3 = _mm_macc_pd(dy31,fscal,fiy3);
654 fiz3 = _mm_macc_pd(dz31,fscal,fiz3);
656 fjx1 = _mm_macc_pd(dx31,fscal,fjx1);
657 fjy1 = _mm_macc_pd(dy31,fscal,fjy1);
658 fjz1 = _mm_macc_pd(dz31,fscal,fjz1);
660 /**************************
661 * CALCULATE INTERACTIONS *
662 **************************/
664 r32 = _mm_mul_pd(rsq32,rinv32);
666 /* Calculate table index by multiplying r with table scale and truncate to integer */
667 rt = _mm_mul_pd(r32,vftabscale);
668 vfitab = _mm_cvttpd_epi32(rt);
670 vfeps = _mm_frcz_pd(rt);
672 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
674 twovfeps = _mm_add_pd(vfeps,vfeps);
675 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
677 /* CUBIC SPLINE TABLE ELECTROSTATICS */
678 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
679 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
680 GMX_MM_TRANSPOSE2_PD(Y,F);
681 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
682 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
683 GMX_MM_TRANSPOSE2_PD(G,H);
684 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
685 VV = _mm_macc_pd(vfeps,Fp,Y);
686 velec = _mm_mul_pd(qq32,VV);
687 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
688 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq32,FF),_mm_mul_pd(vftabscale,rinv32)));
690 /* Update potential sum for this i atom from the interaction with this j atom. */
691 velecsum = _mm_add_pd(velecsum,velec);
695 /* Update vectorial force */
696 fix3 = _mm_macc_pd(dx32,fscal,fix3);
697 fiy3 = _mm_macc_pd(dy32,fscal,fiy3);
698 fiz3 = _mm_macc_pd(dz32,fscal,fiz3);
700 fjx2 = _mm_macc_pd(dx32,fscal,fjx2);
701 fjy2 = _mm_macc_pd(dy32,fscal,fjy2);
702 fjz2 = _mm_macc_pd(dz32,fscal,fjz2);
704 /**************************
705 * CALCULATE INTERACTIONS *
706 **************************/
708 r33 = _mm_mul_pd(rsq33,rinv33);
710 /* Calculate table index by multiplying r with table scale and truncate to integer */
711 rt = _mm_mul_pd(r33,vftabscale);
712 vfitab = _mm_cvttpd_epi32(rt);
714 vfeps = _mm_frcz_pd(rt);
716 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
718 twovfeps = _mm_add_pd(vfeps,vfeps);
719 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
721 /* CUBIC SPLINE TABLE ELECTROSTATICS */
722 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
723 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
724 GMX_MM_TRANSPOSE2_PD(Y,F);
725 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
726 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
727 GMX_MM_TRANSPOSE2_PD(G,H);
728 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
729 VV = _mm_macc_pd(vfeps,Fp,Y);
730 velec = _mm_mul_pd(qq33,VV);
731 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
732 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq33,FF),_mm_mul_pd(vftabscale,rinv33)));
734 /* Update potential sum for this i atom from the interaction with this j atom. */
735 velecsum = _mm_add_pd(velecsum,velec);
739 /* Update vectorial force */
740 fix3 = _mm_macc_pd(dx33,fscal,fix3);
741 fiy3 = _mm_macc_pd(dy33,fscal,fiy3);
742 fiz3 = _mm_macc_pd(dz33,fscal,fiz3);
744 fjx3 = _mm_macc_pd(dx33,fscal,fjx3);
745 fjy3 = _mm_macc_pd(dy33,fscal,fjy3);
746 fjz3 = _mm_macc_pd(dz33,fscal,fjz3);
748 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);
750 /* Inner loop uses 476 flops */
757 j_coord_offsetA = DIM*jnrA;
759 /* load j atom coordinates */
760 gmx_mm_load_4rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
761 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
762 &jy2,&jz2,&jx3,&jy3,&jz3);
764 /* Calculate displacement vector */
765 dx00 = _mm_sub_pd(ix0,jx0);
766 dy00 = _mm_sub_pd(iy0,jy0);
767 dz00 = _mm_sub_pd(iz0,jz0);
768 dx11 = _mm_sub_pd(ix1,jx1);
769 dy11 = _mm_sub_pd(iy1,jy1);
770 dz11 = _mm_sub_pd(iz1,jz1);
771 dx12 = _mm_sub_pd(ix1,jx2);
772 dy12 = _mm_sub_pd(iy1,jy2);
773 dz12 = _mm_sub_pd(iz1,jz2);
774 dx13 = _mm_sub_pd(ix1,jx3);
775 dy13 = _mm_sub_pd(iy1,jy3);
776 dz13 = _mm_sub_pd(iz1,jz3);
777 dx21 = _mm_sub_pd(ix2,jx1);
778 dy21 = _mm_sub_pd(iy2,jy1);
779 dz21 = _mm_sub_pd(iz2,jz1);
780 dx22 = _mm_sub_pd(ix2,jx2);
781 dy22 = _mm_sub_pd(iy2,jy2);
782 dz22 = _mm_sub_pd(iz2,jz2);
783 dx23 = _mm_sub_pd(ix2,jx3);
784 dy23 = _mm_sub_pd(iy2,jy3);
785 dz23 = _mm_sub_pd(iz2,jz3);
786 dx31 = _mm_sub_pd(ix3,jx1);
787 dy31 = _mm_sub_pd(iy3,jy1);
788 dz31 = _mm_sub_pd(iz3,jz1);
789 dx32 = _mm_sub_pd(ix3,jx2);
790 dy32 = _mm_sub_pd(iy3,jy2);
791 dz32 = _mm_sub_pd(iz3,jz2);
792 dx33 = _mm_sub_pd(ix3,jx3);
793 dy33 = _mm_sub_pd(iy3,jy3);
794 dz33 = _mm_sub_pd(iz3,jz3);
796 /* Calculate squared distance and things based on it */
797 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
798 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
799 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
800 rsq13 = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
801 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
802 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
803 rsq23 = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
804 rsq31 = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
805 rsq32 = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
806 rsq33 = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
808 rinv00 = gmx_mm_invsqrt_pd(rsq00);
809 rinv11 = gmx_mm_invsqrt_pd(rsq11);
810 rinv12 = gmx_mm_invsqrt_pd(rsq12);
811 rinv13 = gmx_mm_invsqrt_pd(rsq13);
812 rinv21 = gmx_mm_invsqrt_pd(rsq21);
813 rinv22 = gmx_mm_invsqrt_pd(rsq22);
814 rinv23 = gmx_mm_invsqrt_pd(rsq23);
815 rinv31 = gmx_mm_invsqrt_pd(rsq31);
816 rinv32 = gmx_mm_invsqrt_pd(rsq32);
817 rinv33 = gmx_mm_invsqrt_pd(rsq33);
819 fjx0 = _mm_setzero_pd();
820 fjy0 = _mm_setzero_pd();
821 fjz0 = _mm_setzero_pd();
822 fjx1 = _mm_setzero_pd();
823 fjy1 = _mm_setzero_pd();
824 fjz1 = _mm_setzero_pd();
825 fjx2 = _mm_setzero_pd();
826 fjy2 = _mm_setzero_pd();
827 fjz2 = _mm_setzero_pd();
828 fjx3 = _mm_setzero_pd();
829 fjy3 = _mm_setzero_pd();
830 fjz3 = _mm_setzero_pd();
832 /**************************
833 * CALCULATE INTERACTIONS *
834 **************************/
836 r00 = _mm_mul_pd(rsq00,rinv00);
838 /* Calculate table index by multiplying r with table scale and truncate to integer */
839 rt = _mm_mul_pd(r00,vftabscale);
840 vfitab = _mm_cvttpd_epi32(rt);
842 vfeps = _mm_frcz_pd(rt);
844 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
846 twovfeps = _mm_add_pd(vfeps,vfeps);
847 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
849 /* CUBIC SPLINE TABLE DISPERSION */
850 vfitab = _mm_add_epi32(vfitab,ifour);
851 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
852 F = _mm_setzero_pd();
853 GMX_MM_TRANSPOSE2_PD(Y,F);
854 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
855 H = _mm_setzero_pd();
856 GMX_MM_TRANSPOSE2_PD(G,H);
857 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(H,vfeps,G),F);
858 VV = _mm_macc_pd(vfeps,Fp,Y);
859 vvdw6 = _mm_mul_pd(c6_00,VV);
860 FF = _mm_macc_pd(vfeps,_mm_macc_pd(twovfeps,H,G),Fp);
861 fvdw6 = _mm_mul_pd(c6_00,FF);
863 /* CUBIC SPLINE TABLE REPULSION */
864 vfitab = _mm_add_epi32(vfitab,ifour);
865 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
866 F = _mm_setzero_pd();
867 GMX_MM_TRANSPOSE2_PD(Y,F);
868 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
869 H = _mm_setzero_pd();
870 GMX_MM_TRANSPOSE2_PD(G,H);
871 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(H,vfeps,G),F);
872 VV = _mm_macc_pd(vfeps,Fp,Y);
873 vvdw12 = _mm_mul_pd(c12_00,VV);
874 FF = _mm_macc_pd(vfeps,_mm_macc_pd(twovfeps,H,G),Fp);
875 fvdw12 = _mm_mul_pd(c12_00,FF);
876 vvdw = _mm_add_pd(vvdw12,vvdw6);
877 fvdw = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
879 /* Update potential sum for this i atom from the interaction with this j atom. */
880 vvdw = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
881 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
885 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
887 /* Update vectorial force */
888 fix0 = _mm_macc_pd(dx00,fscal,fix0);
889 fiy0 = _mm_macc_pd(dy00,fscal,fiy0);
890 fiz0 = _mm_macc_pd(dz00,fscal,fiz0);
892 fjx0 = _mm_macc_pd(dx00,fscal,fjx0);
893 fjy0 = _mm_macc_pd(dy00,fscal,fjy0);
894 fjz0 = _mm_macc_pd(dz00,fscal,fjz0);
896 /**************************
897 * CALCULATE INTERACTIONS *
898 **************************/
900 r11 = _mm_mul_pd(rsq11,rinv11);
902 /* Calculate table index by multiplying r with table scale and truncate to integer */
903 rt = _mm_mul_pd(r11,vftabscale);
904 vfitab = _mm_cvttpd_epi32(rt);
906 vfeps = _mm_frcz_pd(rt);
908 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
910 twovfeps = _mm_add_pd(vfeps,vfeps);
911 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
913 /* CUBIC SPLINE TABLE ELECTROSTATICS */
914 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
915 F = _mm_setzero_pd();
916 GMX_MM_TRANSPOSE2_PD(Y,F);
917 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
918 H = _mm_setzero_pd();
919 GMX_MM_TRANSPOSE2_PD(G,H);
920 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
921 VV = _mm_macc_pd(vfeps,Fp,Y);
922 velec = _mm_mul_pd(qq11,VV);
923 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
924 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
926 /* Update potential sum for this i atom from the interaction with this j atom. */
927 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
928 velecsum = _mm_add_pd(velecsum,velec);
932 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
934 /* Update vectorial force */
935 fix1 = _mm_macc_pd(dx11,fscal,fix1);
936 fiy1 = _mm_macc_pd(dy11,fscal,fiy1);
937 fiz1 = _mm_macc_pd(dz11,fscal,fiz1);
939 fjx1 = _mm_macc_pd(dx11,fscal,fjx1);
940 fjy1 = _mm_macc_pd(dy11,fscal,fjy1);
941 fjz1 = _mm_macc_pd(dz11,fscal,fjz1);
943 /**************************
944 * CALCULATE INTERACTIONS *
945 **************************/
947 r12 = _mm_mul_pd(rsq12,rinv12);
949 /* Calculate table index by multiplying r with table scale and truncate to integer */
950 rt = _mm_mul_pd(r12,vftabscale);
951 vfitab = _mm_cvttpd_epi32(rt);
953 vfeps = _mm_frcz_pd(rt);
955 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
957 twovfeps = _mm_add_pd(vfeps,vfeps);
958 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
960 /* CUBIC SPLINE TABLE ELECTROSTATICS */
961 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
962 F = _mm_setzero_pd();
963 GMX_MM_TRANSPOSE2_PD(Y,F);
964 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
965 H = _mm_setzero_pd();
966 GMX_MM_TRANSPOSE2_PD(G,H);
967 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
968 VV = _mm_macc_pd(vfeps,Fp,Y);
969 velec = _mm_mul_pd(qq12,VV);
970 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
971 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
973 /* Update potential sum for this i atom from the interaction with this j atom. */
974 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
975 velecsum = _mm_add_pd(velecsum,velec);
979 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
981 /* Update vectorial force */
982 fix1 = _mm_macc_pd(dx12,fscal,fix1);
983 fiy1 = _mm_macc_pd(dy12,fscal,fiy1);
984 fiz1 = _mm_macc_pd(dz12,fscal,fiz1);
986 fjx2 = _mm_macc_pd(dx12,fscal,fjx2);
987 fjy2 = _mm_macc_pd(dy12,fscal,fjy2);
988 fjz2 = _mm_macc_pd(dz12,fscal,fjz2);
990 /**************************
991 * CALCULATE INTERACTIONS *
992 **************************/
994 r13 = _mm_mul_pd(rsq13,rinv13);
996 /* Calculate table index by multiplying r with table scale and truncate to integer */
997 rt = _mm_mul_pd(r13,vftabscale);
998 vfitab = _mm_cvttpd_epi32(rt);
1000 vfeps = _mm_frcz_pd(rt);
1002 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1004 twovfeps = _mm_add_pd(vfeps,vfeps);
1005 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1007 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1008 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1009 F = _mm_setzero_pd();
1010 GMX_MM_TRANSPOSE2_PD(Y,F);
1011 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1012 H = _mm_setzero_pd();
1013 GMX_MM_TRANSPOSE2_PD(G,H);
1014 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1015 VV = _mm_macc_pd(vfeps,Fp,Y);
1016 velec = _mm_mul_pd(qq13,VV);
1017 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1018 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq13,FF),_mm_mul_pd(vftabscale,rinv13)));
1020 /* Update potential sum for this i atom from the interaction with this j atom. */
1021 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1022 velecsum = _mm_add_pd(velecsum,velec);
1026 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1028 /* Update vectorial force */
1029 fix1 = _mm_macc_pd(dx13,fscal,fix1);
1030 fiy1 = _mm_macc_pd(dy13,fscal,fiy1);
1031 fiz1 = _mm_macc_pd(dz13,fscal,fiz1);
1033 fjx3 = _mm_macc_pd(dx13,fscal,fjx3);
1034 fjy3 = _mm_macc_pd(dy13,fscal,fjy3);
1035 fjz3 = _mm_macc_pd(dz13,fscal,fjz3);
1037 /**************************
1038 * CALCULATE INTERACTIONS *
1039 **************************/
1041 r21 = _mm_mul_pd(rsq21,rinv21);
1043 /* Calculate table index by multiplying r with table scale and truncate to integer */
1044 rt = _mm_mul_pd(r21,vftabscale);
1045 vfitab = _mm_cvttpd_epi32(rt);
1047 vfeps = _mm_frcz_pd(rt);
1049 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1051 twovfeps = _mm_add_pd(vfeps,vfeps);
1052 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1054 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1055 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1056 F = _mm_setzero_pd();
1057 GMX_MM_TRANSPOSE2_PD(Y,F);
1058 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1059 H = _mm_setzero_pd();
1060 GMX_MM_TRANSPOSE2_PD(G,H);
1061 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1062 VV = _mm_macc_pd(vfeps,Fp,Y);
1063 velec = _mm_mul_pd(qq21,VV);
1064 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1065 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
1067 /* Update potential sum for this i atom from the interaction with this j atom. */
1068 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1069 velecsum = _mm_add_pd(velecsum,velec);
1073 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1075 /* Update vectorial force */
1076 fix2 = _mm_macc_pd(dx21,fscal,fix2);
1077 fiy2 = _mm_macc_pd(dy21,fscal,fiy2);
1078 fiz2 = _mm_macc_pd(dz21,fscal,fiz2);
1080 fjx1 = _mm_macc_pd(dx21,fscal,fjx1);
1081 fjy1 = _mm_macc_pd(dy21,fscal,fjy1);
1082 fjz1 = _mm_macc_pd(dz21,fscal,fjz1);
1084 /**************************
1085 * CALCULATE INTERACTIONS *
1086 **************************/
1088 r22 = _mm_mul_pd(rsq22,rinv22);
1090 /* Calculate table index by multiplying r with table scale and truncate to integer */
1091 rt = _mm_mul_pd(r22,vftabscale);
1092 vfitab = _mm_cvttpd_epi32(rt);
1094 vfeps = _mm_frcz_pd(rt);
1096 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1098 twovfeps = _mm_add_pd(vfeps,vfeps);
1099 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1101 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1102 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1103 F = _mm_setzero_pd();
1104 GMX_MM_TRANSPOSE2_PD(Y,F);
1105 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1106 H = _mm_setzero_pd();
1107 GMX_MM_TRANSPOSE2_PD(G,H);
1108 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1109 VV = _mm_macc_pd(vfeps,Fp,Y);
1110 velec = _mm_mul_pd(qq22,VV);
1111 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1112 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
1114 /* Update potential sum for this i atom from the interaction with this j atom. */
1115 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1116 velecsum = _mm_add_pd(velecsum,velec);
1120 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1122 /* Update vectorial force */
1123 fix2 = _mm_macc_pd(dx22,fscal,fix2);
1124 fiy2 = _mm_macc_pd(dy22,fscal,fiy2);
1125 fiz2 = _mm_macc_pd(dz22,fscal,fiz2);
1127 fjx2 = _mm_macc_pd(dx22,fscal,fjx2);
1128 fjy2 = _mm_macc_pd(dy22,fscal,fjy2);
1129 fjz2 = _mm_macc_pd(dz22,fscal,fjz2);
1131 /**************************
1132 * CALCULATE INTERACTIONS *
1133 **************************/
1135 r23 = _mm_mul_pd(rsq23,rinv23);
1137 /* Calculate table index by multiplying r with table scale and truncate to integer */
1138 rt = _mm_mul_pd(r23,vftabscale);
1139 vfitab = _mm_cvttpd_epi32(rt);
1141 vfeps = _mm_frcz_pd(rt);
1143 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1145 twovfeps = _mm_add_pd(vfeps,vfeps);
1146 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1148 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1149 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1150 F = _mm_setzero_pd();
1151 GMX_MM_TRANSPOSE2_PD(Y,F);
1152 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1153 H = _mm_setzero_pd();
1154 GMX_MM_TRANSPOSE2_PD(G,H);
1155 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1156 VV = _mm_macc_pd(vfeps,Fp,Y);
1157 velec = _mm_mul_pd(qq23,VV);
1158 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1159 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq23,FF),_mm_mul_pd(vftabscale,rinv23)));
1161 /* Update potential sum for this i atom from the interaction with this j atom. */
1162 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1163 velecsum = _mm_add_pd(velecsum,velec);
1167 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1169 /* Update vectorial force */
1170 fix2 = _mm_macc_pd(dx23,fscal,fix2);
1171 fiy2 = _mm_macc_pd(dy23,fscal,fiy2);
1172 fiz2 = _mm_macc_pd(dz23,fscal,fiz2);
1174 fjx3 = _mm_macc_pd(dx23,fscal,fjx3);
1175 fjy3 = _mm_macc_pd(dy23,fscal,fjy3);
1176 fjz3 = _mm_macc_pd(dz23,fscal,fjz3);
1178 /**************************
1179 * CALCULATE INTERACTIONS *
1180 **************************/
1182 r31 = _mm_mul_pd(rsq31,rinv31);
1184 /* Calculate table index by multiplying r with table scale and truncate to integer */
1185 rt = _mm_mul_pd(r31,vftabscale);
1186 vfitab = _mm_cvttpd_epi32(rt);
1188 vfeps = _mm_frcz_pd(rt);
1190 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1192 twovfeps = _mm_add_pd(vfeps,vfeps);
1193 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1195 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1196 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1197 F = _mm_setzero_pd();
1198 GMX_MM_TRANSPOSE2_PD(Y,F);
1199 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1200 H = _mm_setzero_pd();
1201 GMX_MM_TRANSPOSE2_PD(G,H);
1202 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1203 VV = _mm_macc_pd(vfeps,Fp,Y);
1204 velec = _mm_mul_pd(qq31,VV);
1205 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1206 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq31,FF),_mm_mul_pd(vftabscale,rinv31)));
1208 /* Update potential sum for this i atom from the interaction with this j atom. */
1209 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1210 velecsum = _mm_add_pd(velecsum,velec);
1214 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1216 /* Update vectorial force */
1217 fix3 = _mm_macc_pd(dx31,fscal,fix3);
1218 fiy3 = _mm_macc_pd(dy31,fscal,fiy3);
1219 fiz3 = _mm_macc_pd(dz31,fscal,fiz3);
1221 fjx1 = _mm_macc_pd(dx31,fscal,fjx1);
1222 fjy1 = _mm_macc_pd(dy31,fscal,fjy1);
1223 fjz1 = _mm_macc_pd(dz31,fscal,fjz1);
1225 /**************************
1226 * CALCULATE INTERACTIONS *
1227 **************************/
1229 r32 = _mm_mul_pd(rsq32,rinv32);
1231 /* Calculate table index by multiplying r with table scale and truncate to integer */
1232 rt = _mm_mul_pd(r32,vftabscale);
1233 vfitab = _mm_cvttpd_epi32(rt);
1235 vfeps = _mm_frcz_pd(rt);
1237 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1239 twovfeps = _mm_add_pd(vfeps,vfeps);
1240 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1242 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1243 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1244 F = _mm_setzero_pd();
1245 GMX_MM_TRANSPOSE2_PD(Y,F);
1246 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1247 H = _mm_setzero_pd();
1248 GMX_MM_TRANSPOSE2_PD(G,H);
1249 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1250 VV = _mm_macc_pd(vfeps,Fp,Y);
1251 velec = _mm_mul_pd(qq32,VV);
1252 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1253 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq32,FF),_mm_mul_pd(vftabscale,rinv32)));
1255 /* Update potential sum for this i atom from the interaction with this j atom. */
1256 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1257 velecsum = _mm_add_pd(velecsum,velec);
1261 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1263 /* Update vectorial force */
1264 fix3 = _mm_macc_pd(dx32,fscal,fix3);
1265 fiy3 = _mm_macc_pd(dy32,fscal,fiy3);
1266 fiz3 = _mm_macc_pd(dz32,fscal,fiz3);
1268 fjx2 = _mm_macc_pd(dx32,fscal,fjx2);
1269 fjy2 = _mm_macc_pd(dy32,fscal,fjy2);
1270 fjz2 = _mm_macc_pd(dz32,fscal,fjz2);
1272 /**************************
1273 * CALCULATE INTERACTIONS *
1274 **************************/
1276 r33 = _mm_mul_pd(rsq33,rinv33);
1278 /* Calculate table index by multiplying r with table scale and truncate to integer */
1279 rt = _mm_mul_pd(r33,vftabscale);
1280 vfitab = _mm_cvttpd_epi32(rt);
1282 vfeps = _mm_frcz_pd(rt);
1284 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1286 twovfeps = _mm_add_pd(vfeps,vfeps);
1287 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1289 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1290 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1291 F = _mm_setzero_pd();
1292 GMX_MM_TRANSPOSE2_PD(Y,F);
1293 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1294 H = _mm_setzero_pd();
1295 GMX_MM_TRANSPOSE2_PD(G,H);
1296 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1297 VV = _mm_macc_pd(vfeps,Fp,Y);
1298 velec = _mm_mul_pd(qq33,VV);
1299 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1300 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq33,FF),_mm_mul_pd(vftabscale,rinv33)));
1302 /* Update potential sum for this i atom from the interaction with this j atom. */
1303 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1304 velecsum = _mm_add_pd(velecsum,velec);
1308 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1310 /* Update vectorial force */
1311 fix3 = _mm_macc_pd(dx33,fscal,fix3);
1312 fiy3 = _mm_macc_pd(dy33,fscal,fiy3);
1313 fiz3 = _mm_macc_pd(dz33,fscal,fiz3);
1315 fjx3 = _mm_macc_pd(dx33,fscal,fjx3);
1316 fjy3 = _mm_macc_pd(dy33,fscal,fjy3);
1317 fjz3 = _mm_macc_pd(dz33,fscal,fjz3);
1319 gmx_mm_decrement_4rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1321 /* Inner loop uses 476 flops */
1324 /* End of innermost loop */
1326 gmx_mm_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1327 f+i_coord_offset,fshift+i_shift_offset);
1330 /* Update potential energies */
1331 gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1332 gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1334 /* Increment number of inner iterations */
1335 inneriter += j_index_end - j_index_start;
1337 /* Outer loop uses 26 flops */
1340 /* Increment number of outer iterations */
1343 /* Update outer/inner flops */
1345 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*476);
1348 * Gromacs nonbonded kernel: nb_kernel_ElecCSTab_VdwCSTab_GeomW4W4_F_avx_128_fma_double
1349 * Electrostatics interaction: CubicSplineTable
1350 * VdW interaction: CubicSplineTable
1351 * Geometry: Water4-Water4
1352 * Calculate force/pot: Force
1355 nb_kernel_ElecCSTab_VdwCSTab_GeomW4W4_F_avx_128_fma_double
1356 (t_nblist * gmx_restrict nlist,
1357 rvec * gmx_restrict xx,
1358 rvec * gmx_restrict ff,
1359 t_forcerec * gmx_restrict fr,
1360 t_mdatoms * gmx_restrict mdatoms,
1361 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1362 t_nrnb * gmx_restrict nrnb)
1364 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1365 * just 0 for non-waters.
1366 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
1367 * jnr indices corresponding to data put in the four positions in the SIMD register.
1369 int i_shift_offset,i_coord_offset,outeriter,inneriter;
1370 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1372 int j_coord_offsetA,j_coord_offsetB;
1373 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
1374 real rcutoff_scalar;
1375 real *shiftvec,*fshift,*x,*f;
1376 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1378 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1380 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1382 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1384 __m128d ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1385 int vdwjidx0A,vdwjidx0B;
1386 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1387 int vdwjidx1A,vdwjidx1B;
1388 __m128d jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1389 int vdwjidx2A,vdwjidx2B;
1390 __m128d jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1391 int vdwjidx3A,vdwjidx3B;
1392 __m128d jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1393 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1394 __m128d dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1395 __m128d dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1396 __m128d dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1397 __m128d dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1398 __m128d dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1399 __m128d dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1400 __m128d dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1401 __m128d dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1402 __m128d dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1403 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
1406 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1409 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
1410 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
1412 __m128i ifour = _mm_set1_epi32(4);
1413 __m128d rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
1415 __m128d dummy_mask,cutoff_mask;
1416 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
1417 __m128d one = _mm_set1_pd(1.0);
1418 __m128d two = _mm_set1_pd(2.0);
1424 jindex = nlist->jindex;
1426 shiftidx = nlist->shift;
1428 shiftvec = fr->shift_vec[0];
1429 fshift = fr->fshift[0];
1430 facel = _mm_set1_pd(fr->epsfac);
1431 charge = mdatoms->chargeA;
1432 nvdwtype = fr->ntype;
1433 vdwparam = fr->nbfp;
1434 vdwtype = mdatoms->typeA;
1436 vftab = kernel_data->table_elec_vdw->data;
1437 vftabscale = _mm_set1_pd(kernel_data->table_elec_vdw->scale);
1439 /* Setup water-specific parameters */
1440 inr = nlist->iinr[0];
1441 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
1442 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
1443 iq3 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
1444 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
1446 jq1 = _mm_set1_pd(charge[inr+1]);
1447 jq2 = _mm_set1_pd(charge[inr+2]);
1448 jq3 = _mm_set1_pd(charge[inr+3]);
1449 vdwjidx0A = 2*vdwtype[inr+0];
1450 c6_00 = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
1451 c12_00 = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
1452 qq11 = _mm_mul_pd(iq1,jq1);
1453 qq12 = _mm_mul_pd(iq1,jq2);
1454 qq13 = _mm_mul_pd(iq1,jq3);
1455 qq21 = _mm_mul_pd(iq2,jq1);
1456 qq22 = _mm_mul_pd(iq2,jq2);
1457 qq23 = _mm_mul_pd(iq2,jq3);
1458 qq31 = _mm_mul_pd(iq3,jq1);
1459 qq32 = _mm_mul_pd(iq3,jq2);
1460 qq33 = _mm_mul_pd(iq3,jq3);
1462 /* Avoid stupid compiler warnings */
1464 j_coord_offsetA = 0;
1465 j_coord_offsetB = 0;
1470 /* Start outer loop over neighborlists */
1471 for(iidx=0; iidx<nri; iidx++)
1473 /* Load shift vector for this list */
1474 i_shift_offset = DIM*shiftidx[iidx];
1476 /* Load limits for loop over neighbors */
1477 j_index_start = jindex[iidx];
1478 j_index_end = jindex[iidx+1];
1480 /* Get outer coordinate index */
1482 i_coord_offset = DIM*inr;
1484 /* Load i particle coords and add shift vector */
1485 gmx_mm_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1486 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1488 fix0 = _mm_setzero_pd();
1489 fiy0 = _mm_setzero_pd();
1490 fiz0 = _mm_setzero_pd();
1491 fix1 = _mm_setzero_pd();
1492 fiy1 = _mm_setzero_pd();
1493 fiz1 = _mm_setzero_pd();
1494 fix2 = _mm_setzero_pd();
1495 fiy2 = _mm_setzero_pd();
1496 fiz2 = _mm_setzero_pd();
1497 fix3 = _mm_setzero_pd();
1498 fiy3 = _mm_setzero_pd();
1499 fiz3 = _mm_setzero_pd();
1501 /* Start inner kernel loop */
1502 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
1505 /* Get j neighbor index, and coordinate index */
1507 jnrB = jjnr[jidx+1];
1508 j_coord_offsetA = DIM*jnrA;
1509 j_coord_offsetB = DIM*jnrB;
1511 /* load j atom coordinates */
1512 gmx_mm_load_4rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1513 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1514 &jy2,&jz2,&jx3,&jy3,&jz3);
1516 /* Calculate displacement vector */
1517 dx00 = _mm_sub_pd(ix0,jx0);
1518 dy00 = _mm_sub_pd(iy0,jy0);
1519 dz00 = _mm_sub_pd(iz0,jz0);
1520 dx11 = _mm_sub_pd(ix1,jx1);
1521 dy11 = _mm_sub_pd(iy1,jy1);
1522 dz11 = _mm_sub_pd(iz1,jz1);
1523 dx12 = _mm_sub_pd(ix1,jx2);
1524 dy12 = _mm_sub_pd(iy1,jy2);
1525 dz12 = _mm_sub_pd(iz1,jz2);
1526 dx13 = _mm_sub_pd(ix1,jx3);
1527 dy13 = _mm_sub_pd(iy1,jy3);
1528 dz13 = _mm_sub_pd(iz1,jz3);
1529 dx21 = _mm_sub_pd(ix2,jx1);
1530 dy21 = _mm_sub_pd(iy2,jy1);
1531 dz21 = _mm_sub_pd(iz2,jz1);
1532 dx22 = _mm_sub_pd(ix2,jx2);
1533 dy22 = _mm_sub_pd(iy2,jy2);
1534 dz22 = _mm_sub_pd(iz2,jz2);
1535 dx23 = _mm_sub_pd(ix2,jx3);
1536 dy23 = _mm_sub_pd(iy2,jy3);
1537 dz23 = _mm_sub_pd(iz2,jz3);
1538 dx31 = _mm_sub_pd(ix3,jx1);
1539 dy31 = _mm_sub_pd(iy3,jy1);
1540 dz31 = _mm_sub_pd(iz3,jz1);
1541 dx32 = _mm_sub_pd(ix3,jx2);
1542 dy32 = _mm_sub_pd(iy3,jy2);
1543 dz32 = _mm_sub_pd(iz3,jz2);
1544 dx33 = _mm_sub_pd(ix3,jx3);
1545 dy33 = _mm_sub_pd(iy3,jy3);
1546 dz33 = _mm_sub_pd(iz3,jz3);
1548 /* Calculate squared distance and things based on it */
1549 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1550 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1551 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1552 rsq13 = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
1553 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1554 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1555 rsq23 = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
1556 rsq31 = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
1557 rsq32 = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
1558 rsq33 = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
1560 rinv00 = gmx_mm_invsqrt_pd(rsq00);
1561 rinv11 = gmx_mm_invsqrt_pd(rsq11);
1562 rinv12 = gmx_mm_invsqrt_pd(rsq12);
1563 rinv13 = gmx_mm_invsqrt_pd(rsq13);
1564 rinv21 = gmx_mm_invsqrt_pd(rsq21);
1565 rinv22 = gmx_mm_invsqrt_pd(rsq22);
1566 rinv23 = gmx_mm_invsqrt_pd(rsq23);
1567 rinv31 = gmx_mm_invsqrt_pd(rsq31);
1568 rinv32 = gmx_mm_invsqrt_pd(rsq32);
1569 rinv33 = gmx_mm_invsqrt_pd(rsq33);
1571 fjx0 = _mm_setzero_pd();
1572 fjy0 = _mm_setzero_pd();
1573 fjz0 = _mm_setzero_pd();
1574 fjx1 = _mm_setzero_pd();
1575 fjy1 = _mm_setzero_pd();
1576 fjz1 = _mm_setzero_pd();
1577 fjx2 = _mm_setzero_pd();
1578 fjy2 = _mm_setzero_pd();
1579 fjz2 = _mm_setzero_pd();
1580 fjx3 = _mm_setzero_pd();
1581 fjy3 = _mm_setzero_pd();
1582 fjz3 = _mm_setzero_pd();
1584 /**************************
1585 * CALCULATE INTERACTIONS *
1586 **************************/
1588 r00 = _mm_mul_pd(rsq00,rinv00);
1590 /* Calculate table index by multiplying r with table scale and truncate to integer */
1591 rt = _mm_mul_pd(r00,vftabscale);
1592 vfitab = _mm_cvttpd_epi32(rt);
1594 vfeps = _mm_frcz_pd(rt);
1596 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1598 twovfeps = _mm_add_pd(vfeps,vfeps);
1599 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1601 /* CUBIC SPLINE TABLE DISPERSION */
1602 vfitab = _mm_add_epi32(vfitab,ifour);
1603 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1604 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1605 GMX_MM_TRANSPOSE2_PD(Y,F);
1606 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1607 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1608 GMX_MM_TRANSPOSE2_PD(G,H);
1609 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(H,vfeps,G),F);
1610 FF = _mm_macc_pd(vfeps,_mm_macc_pd(twovfeps,H,G),Fp);
1611 fvdw6 = _mm_mul_pd(c6_00,FF);
1613 /* CUBIC SPLINE TABLE REPULSION */
1614 vfitab = _mm_add_epi32(vfitab,ifour);
1615 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1616 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1617 GMX_MM_TRANSPOSE2_PD(Y,F);
1618 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1619 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1620 GMX_MM_TRANSPOSE2_PD(G,H);
1621 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(H,vfeps,G),F);
1622 FF = _mm_macc_pd(vfeps,_mm_macc_pd(twovfeps,H,G),Fp);
1623 fvdw12 = _mm_mul_pd(c12_00,FF);
1624 fvdw = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
1628 /* Update vectorial force */
1629 fix0 = _mm_macc_pd(dx00,fscal,fix0);
1630 fiy0 = _mm_macc_pd(dy00,fscal,fiy0);
1631 fiz0 = _mm_macc_pd(dz00,fscal,fiz0);
1633 fjx0 = _mm_macc_pd(dx00,fscal,fjx0);
1634 fjy0 = _mm_macc_pd(dy00,fscal,fjy0);
1635 fjz0 = _mm_macc_pd(dz00,fscal,fjz0);
1637 /**************************
1638 * CALCULATE INTERACTIONS *
1639 **************************/
1641 r11 = _mm_mul_pd(rsq11,rinv11);
1643 /* Calculate table index by multiplying r with table scale and truncate to integer */
1644 rt = _mm_mul_pd(r11,vftabscale);
1645 vfitab = _mm_cvttpd_epi32(rt);
1647 vfeps = _mm_frcz_pd(rt);
1649 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1651 twovfeps = _mm_add_pd(vfeps,vfeps);
1652 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1654 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1655 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1656 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1657 GMX_MM_TRANSPOSE2_PD(Y,F);
1658 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1659 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1660 GMX_MM_TRANSPOSE2_PD(G,H);
1661 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1662 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1663 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
1667 /* Update vectorial force */
1668 fix1 = _mm_macc_pd(dx11,fscal,fix1);
1669 fiy1 = _mm_macc_pd(dy11,fscal,fiy1);
1670 fiz1 = _mm_macc_pd(dz11,fscal,fiz1);
1672 fjx1 = _mm_macc_pd(dx11,fscal,fjx1);
1673 fjy1 = _mm_macc_pd(dy11,fscal,fjy1);
1674 fjz1 = _mm_macc_pd(dz11,fscal,fjz1);
1676 /**************************
1677 * CALCULATE INTERACTIONS *
1678 **************************/
1680 r12 = _mm_mul_pd(rsq12,rinv12);
1682 /* Calculate table index by multiplying r with table scale and truncate to integer */
1683 rt = _mm_mul_pd(r12,vftabscale);
1684 vfitab = _mm_cvttpd_epi32(rt);
1686 vfeps = _mm_frcz_pd(rt);
1688 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1690 twovfeps = _mm_add_pd(vfeps,vfeps);
1691 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1693 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1694 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1695 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1696 GMX_MM_TRANSPOSE2_PD(Y,F);
1697 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1698 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1699 GMX_MM_TRANSPOSE2_PD(G,H);
1700 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1701 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1702 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
1706 /* Update vectorial force */
1707 fix1 = _mm_macc_pd(dx12,fscal,fix1);
1708 fiy1 = _mm_macc_pd(dy12,fscal,fiy1);
1709 fiz1 = _mm_macc_pd(dz12,fscal,fiz1);
1711 fjx2 = _mm_macc_pd(dx12,fscal,fjx2);
1712 fjy2 = _mm_macc_pd(dy12,fscal,fjy2);
1713 fjz2 = _mm_macc_pd(dz12,fscal,fjz2);
1715 /**************************
1716 * CALCULATE INTERACTIONS *
1717 **************************/
1719 r13 = _mm_mul_pd(rsq13,rinv13);
1721 /* Calculate table index by multiplying r with table scale and truncate to integer */
1722 rt = _mm_mul_pd(r13,vftabscale);
1723 vfitab = _mm_cvttpd_epi32(rt);
1725 vfeps = _mm_frcz_pd(rt);
1727 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1729 twovfeps = _mm_add_pd(vfeps,vfeps);
1730 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1732 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1733 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1734 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1735 GMX_MM_TRANSPOSE2_PD(Y,F);
1736 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1737 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1738 GMX_MM_TRANSPOSE2_PD(G,H);
1739 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1740 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1741 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq13,FF),_mm_mul_pd(vftabscale,rinv13)));
1745 /* Update vectorial force */
1746 fix1 = _mm_macc_pd(dx13,fscal,fix1);
1747 fiy1 = _mm_macc_pd(dy13,fscal,fiy1);
1748 fiz1 = _mm_macc_pd(dz13,fscal,fiz1);
1750 fjx3 = _mm_macc_pd(dx13,fscal,fjx3);
1751 fjy3 = _mm_macc_pd(dy13,fscal,fjy3);
1752 fjz3 = _mm_macc_pd(dz13,fscal,fjz3);
1754 /**************************
1755 * CALCULATE INTERACTIONS *
1756 **************************/
1758 r21 = _mm_mul_pd(rsq21,rinv21);
1760 /* Calculate table index by multiplying r with table scale and truncate to integer */
1761 rt = _mm_mul_pd(r21,vftabscale);
1762 vfitab = _mm_cvttpd_epi32(rt);
1764 vfeps = _mm_frcz_pd(rt);
1766 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1768 twovfeps = _mm_add_pd(vfeps,vfeps);
1769 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1771 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1772 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1773 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1774 GMX_MM_TRANSPOSE2_PD(Y,F);
1775 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1776 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1777 GMX_MM_TRANSPOSE2_PD(G,H);
1778 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1779 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1780 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
1784 /* Update vectorial force */
1785 fix2 = _mm_macc_pd(dx21,fscal,fix2);
1786 fiy2 = _mm_macc_pd(dy21,fscal,fiy2);
1787 fiz2 = _mm_macc_pd(dz21,fscal,fiz2);
1789 fjx1 = _mm_macc_pd(dx21,fscal,fjx1);
1790 fjy1 = _mm_macc_pd(dy21,fscal,fjy1);
1791 fjz1 = _mm_macc_pd(dz21,fscal,fjz1);
1793 /**************************
1794 * CALCULATE INTERACTIONS *
1795 **************************/
1797 r22 = _mm_mul_pd(rsq22,rinv22);
1799 /* Calculate table index by multiplying r with table scale and truncate to integer */
1800 rt = _mm_mul_pd(r22,vftabscale);
1801 vfitab = _mm_cvttpd_epi32(rt);
1803 vfeps = _mm_frcz_pd(rt);
1805 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1807 twovfeps = _mm_add_pd(vfeps,vfeps);
1808 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1810 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1811 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1812 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1813 GMX_MM_TRANSPOSE2_PD(Y,F);
1814 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1815 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1816 GMX_MM_TRANSPOSE2_PD(G,H);
1817 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1818 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1819 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
1823 /* Update vectorial force */
1824 fix2 = _mm_macc_pd(dx22,fscal,fix2);
1825 fiy2 = _mm_macc_pd(dy22,fscal,fiy2);
1826 fiz2 = _mm_macc_pd(dz22,fscal,fiz2);
1828 fjx2 = _mm_macc_pd(dx22,fscal,fjx2);
1829 fjy2 = _mm_macc_pd(dy22,fscal,fjy2);
1830 fjz2 = _mm_macc_pd(dz22,fscal,fjz2);
1832 /**************************
1833 * CALCULATE INTERACTIONS *
1834 **************************/
1836 r23 = _mm_mul_pd(rsq23,rinv23);
1838 /* Calculate table index by multiplying r with table scale and truncate to integer */
1839 rt = _mm_mul_pd(r23,vftabscale);
1840 vfitab = _mm_cvttpd_epi32(rt);
1842 vfeps = _mm_frcz_pd(rt);
1844 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1846 twovfeps = _mm_add_pd(vfeps,vfeps);
1847 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1849 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1850 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1851 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1852 GMX_MM_TRANSPOSE2_PD(Y,F);
1853 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1854 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1855 GMX_MM_TRANSPOSE2_PD(G,H);
1856 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1857 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1858 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq23,FF),_mm_mul_pd(vftabscale,rinv23)));
1862 /* Update vectorial force */
1863 fix2 = _mm_macc_pd(dx23,fscal,fix2);
1864 fiy2 = _mm_macc_pd(dy23,fscal,fiy2);
1865 fiz2 = _mm_macc_pd(dz23,fscal,fiz2);
1867 fjx3 = _mm_macc_pd(dx23,fscal,fjx3);
1868 fjy3 = _mm_macc_pd(dy23,fscal,fjy3);
1869 fjz3 = _mm_macc_pd(dz23,fscal,fjz3);
1871 /**************************
1872 * CALCULATE INTERACTIONS *
1873 **************************/
1875 r31 = _mm_mul_pd(rsq31,rinv31);
1877 /* Calculate table index by multiplying r with table scale and truncate to integer */
1878 rt = _mm_mul_pd(r31,vftabscale);
1879 vfitab = _mm_cvttpd_epi32(rt);
1881 vfeps = _mm_frcz_pd(rt);
1883 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1885 twovfeps = _mm_add_pd(vfeps,vfeps);
1886 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1888 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1889 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1890 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1891 GMX_MM_TRANSPOSE2_PD(Y,F);
1892 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1893 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1894 GMX_MM_TRANSPOSE2_PD(G,H);
1895 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1896 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1897 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq31,FF),_mm_mul_pd(vftabscale,rinv31)));
1901 /* Update vectorial force */
1902 fix3 = _mm_macc_pd(dx31,fscal,fix3);
1903 fiy3 = _mm_macc_pd(dy31,fscal,fiy3);
1904 fiz3 = _mm_macc_pd(dz31,fscal,fiz3);
1906 fjx1 = _mm_macc_pd(dx31,fscal,fjx1);
1907 fjy1 = _mm_macc_pd(dy31,fscal,fjy1);
1908 fjz1 = _mm_macc_pd(dz31,fscal,fjz1);
1910 /**************************
1911 * CALCULATE INTERACTIONS *
1912 **************************/
1914 r32 = _mm_mul_pd(rsq32,rinv32);
1916 /* Calculate table index by multiplying r with table scale and truncate to integer */
1917 rt = _mm_mul_pd(r32,vftabscale);
1918 vfitab = _mm_cvttpd_epi32(rt);
1920 vfeps = _mm_frcz_pd(rt);
1922 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1924 twovfeps = _mm_add_pd(vfeps,vfeps);
1925 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1927 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1928 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1929 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1930 GMX_MM_TRANSPOSE2_PD(Y,F);
1931 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1932 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1933 GMX_MM_TRANSPOSE2_PD(G,H);
1934 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1935 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1936 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq32,FF),_mm_mul_pd(vftabscale,rinv32)));
1940 /* Update vectorial force */
1941 fix3 = _mm_macc_pd(dx32,fscal,fix3);
1942 fiy3 = _mm_macc_pd(dy32,fscal,fiy3);
1943 fiz3 = _mm_macc_pd(dz32,fscal,fiz3);
1945 fjx2 = _mm_macc_pd(dx32,fscal,fjx2);
1946 fjy2 = _mm_macc_pd(dy32,fscal,fjy2);
1947 fjz2 = _mm_macc_pd(dz32,fscal,fjz2);
1949 /**************************
1950 * CALCULATE INTERACTIONS *
1951 **************************/
1953 r33 = _mm_mul_pd(rsq33,rinv33);
1955 /* Calculate table index by multiplying r with table scale and truncate to integer */
1956 rt = _mm_mul_pd(r33,vftabscale);
1957 vfitab = _mm_cvttpd_epi32(rt);
1959 vfeps = _mm_frcz_pd(rt);
1961 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1963 twovfeps = _mm_add_pd(vfeps,vfeps);
1964 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1966 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1967 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1968 F = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1969 GMX_MM_TRANSPOSE2_PD(Y,F);
1970 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1971 H = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1972 GMX_MM_TRANSPOSE2_PD(G,H);
1973 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1974 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1975 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq33,FF),_mm_mul_pd(vftabscale,rinv33)));
1979 /* Update vectorial force */
1980 fix3 = _mm_macc_pd(dx33,fscal,fix3);
1981 fiy3 = _mm_macc_pd(dy33,fscal,fiy3);
1982 fiz3 = _mm_macc_pd(dz33,fscal,fiz3);
1984 fjx3 = _mm_macc_pd(dx33,fscal,fjx3);
1985 fjy3 = _mm_macc_pd(dy33,fscal,fjy3);
1986 fjz3 = _mm_macc_pd(dz33,fscal,fjz3);
1988 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);
1990 /* Inner loop uses 432 flops */
1993 if(jidx<j_index_end)
1997 j_coord_offsetA = DIM*jnrA;
1999 /* load j atom coordinates */
2000 gmx_mm_load_4rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
2001 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
2002 &jy2,&jz2,&jx3,&jy3,&jz3);
2004 /* Calculate displacement vector */
2005 dx00 = _mm_sub_pd(ix0,jx0);
2006 dy00 = _mm_sub_pd(iy0,jy0);
2007 dz00 = _mm_sub_pd(iz0,jz0);
2008 dx11 = _mm_sub_pd(ix1,jx1);
2009 dy11 = _mm_sub_pd(iy1,jy1);
2010 dz11 = _mm_sub_pd(iz1,jz1);
2011 dx12 = _mm_sub_pd(ix1,jx2);
2012 dy12 = _mm_sub_pd(iy1,jy2);
2013 dz12 = _mm_sub_pd(iz1,jz2);
2014 dx13 = _mm_sub_pd(ix1,jx3);
2015 dy13 = _mm_sub_pd(iy1,jy3);
2016 dz13 = _mm_sub_pd(iz1,jz3);
2017 dx21 = _mm_sub_pd(ix2,jx1);
2018 dy21 = _mm_sub_pd(iy2,jy1);
2019 dz21 = _mm_sub_pd(iz2,jz1);
2020 dx22 = _mm_sub_pd(ix2,jx2);
2021 dy22 = _mm_sub_pd(iy2,jy2);
2022 dz22 = _mm_sub_pd(iz2,jz2);
2023 dx23 = _mm_sub_pd(ix2,jx3);
2024 dy23 = _mm_sub_pd(iy2,jy3);
2025 dz23 = _mm_sub_pd(iz2,jz3);
2026 dx31 = _mm_sub_pd(ix3,jx1);
2027 dy31 = _mm_sub_pd(iy3,jy1);
2028 dz31 = _mm_sub_pd(iz3,jz1);
2029 dx32 = _mm_sub_pd(ix3,jx2);
2030 dy32 = _mm_sub_pd(iy3,jy2);
2031 dz32 = _mm_sub_pd(iz3,jz2);
2032 dx33 = _mm_sub_pd(ix3,jx3);
2033 dy33 = _mm_sub_pd(iy3,jy3);
2034 dz33 = _mm_sub_pd(iz3,jz3);
2036 /* Calculate squared distance and things based on it */
2037 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
2038 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
2039 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
2040 rsq13 = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
2041 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
2042 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
2043 rsq23 = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
2044 rsq31 = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
2045 rsq32 = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
2046 rsq33 = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
2048 rinv00 = gmx_mm_invsqrt_pd(rsq00);
2049 rinv11 = gmx_mm_invsqrt_pd(rsq11);
2050 rinv12 = gmx_mm_invsqrt_pd(rsq12);
2051 rinv13 = gmx_mm_invsqrt_pd(rsq13);
2052 rinv21 = gmx_mm_invsqrt_pd(rsq21);
2053 rinv22 = gmx_mm_invsqrt_pd(rsq22);
2054 rinv23 = gmx_mm_invsqrt_pd(rsq23);
2055 rinv31 = gmx_mm_invsqrt_pd(rsq31);
2056 rinv32 = gmx_mm_invsqrt_pd(rsq32);
2057 rinv33 = gmx_mm_invsqrt_pd(rsq33);
2059 fjx0 = _mm_setzero_pd();
2060 fjy0 = _mm_setzero_pd();
2061 fjz0 = _mm_setzero_pd();
2062 fjx1 = _mm_setzero_pd();
2063 fjy1 = _mm_setzero_pd();
2064 fjz1 = _mm_setzero_pd();
2065 fjx2 = _mm_setzero_pd();
2066 fjy2 = _mm_setzero_pd();
2067 fjz2 = _mm_setzero_pd();
2068 fjx3 = _mm_setzero_pd();
2069 fjy3 = _mm_setzero_pd();
2070 fjz3 = _mm_setzero_pd();
2072 /**************************
2073 * CALCULATE INTERACTIONS *
2074 **************************/
2076 r00 = _mm_mul_pd(rsq00,rinv00);
2078 /* Calculate table index by multiplying r with table scale and truncate to integer */
2079 rt = _mm_mul_pd(r00,vftabscale);
2080 vfitab = _mm_cvttpd_epi32(rt);
2082 vfeps = _mm_frcz_pd(rt);
2084 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2086 twovfeps = _mm_add_pd(vfeps,vfeps);
2087 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2089 /* CUBIC SPLINE TABLE DISPERSION */
2090 vfitab = _mm_add_epi32(vfitab,ifour);
2091 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2092 F = _mm_setzero_pd();
2093 GMX_MM_TRANSPOSE2_PD(Y,F);
2094 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
2095 H = _mm_setzero_pd();
2096 GMX_MM_TRANSPOSE2_PD(G,H);
2097 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(H,vfeps,G),F);
2098 FF = _mm_macc_pd(vfeps,_mm_macc_pd(twovfeps,H,G),Fp);
2099 fvdw6 = _mm_mul_pd(c6_00,FF);
2101 /* CUBIC SPLINE TABLE REPULSION */
2102 vfitab = _mm_add_epi32(vfitab,ifour);
2103 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2104 F = _mm_setzero_pd();
2105 GMX_MM_TRANSPOSE2_PD(Y,F);
2106 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
2107 H = _mm_setzero_pd();
2108 GMX_MM_TRANSPOSE2_PD(G,H);
2109 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(H,vfeps,G),F);
2110 FF = _mm_macc_pd(vfeps,_mm_macc_pd(twovfeps,H,G),Fp);
2111 fvdw12 = _mm_mul_pd(c12_00,FF);
2112 fvdw = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
2116 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2118 /* Update vectorial force */
2119 fix0 = _mm_macc_pd(dx00,fscal,fix0);
2120 fiy0 = _mm_macc_pd(dy00,fscal,fiy0);
2121 fiz0 = _mm_macc_pd(dz00,fscal,fiz0);
2123 fjx0 = _mm_macc_pd(dx00,fscal,fjx0);
2124 fjy0 = _mm_macc_pd(dy00,fscal,fjy0);
2125 fjz0 = _mm_macc_pd(dz00,fscal,fjz0);
2127 /**************************
2128 * CALCULATE INTERACTIONS *
2129 **************************/
2131 r11 = _mm_mul_pd(rsq11,rinv11);
2133 /* Calculate table index by multiplying r with table scale and truncate to integer */
2134 rt = _mm_mul_pd(r11,vftabscale);
2135 vfitab = _mm_cvttpd_epi32(rt);
2137 vfeps = _mm_frcz_pd(rt);
2139 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2141 twovfeps = _mm_add_pd(vfeps,vfeps);
2142 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2144 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2145 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2146 F = _mm_setzero_pd();
2147 GMX_MM_TRANSPOSE2_PD(Y,F);
2148 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
2149 H = _mm_setzero_pd();
2150 GMX_MM_TRANSPOSE2_PD(G,H);
2151 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
2152 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
2153 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
2157 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2159 /* Update vectorial force */
2160 fix1 = _mm_macc_pd(dx11,fscal,fix1);
2161 fiy1 = _mm_macc_pd(dy11,fscal,fiy1);
2162 fiz1 = _mm_macc_pd(dz11,fscal,fiz1);
2164 fjx1 = _mm_macc_pd(dx11,fscal,fjx1);
2165 fjy1 = _mm_macc_pd(dy11,fscal,fjy1);
2166 fjz1 = _mm_macc_pd(dz11,fscal,fjz1);
2168 /**************************
2169 * CALCULATE INTERACTIONS *
2170 **************************/
2172 r12 = _mm_mul_pd(rsq12,rinv12);
2174 /* Calculate table index by multiplying r with table scale and truncate to integer */
2175 rt = _mm_mul_pd(r12,vftabscale);
2176 vfitab = _mm_cvttpd_epi32(rt);
2178 vfeps = _mm_frcz_pd(rt);
2180 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2182 twovfeps = _mm_add_pd(vfeps,vfeps);
2183 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2185 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2186 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2187 F = _mm_setzero_pd();
2188 GMX_MM_TRANSPOSE2_PD(Y,F);
2189 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
2190 H = _mm_setzero_pd();
2191 GMX_MM_TRANSPOSE2_PD(G,H);
2192 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
2193 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
2194 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
2198 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2200 /* Update vectorial force */
2201 fix1 = _mm_macc_pd(dx12,fscal,fix1);
2202 fiy1 = _mm_macc_pd(dy12,fscal,fiy1);
2203 fiz1 = _mm_macc_pd(dz12,fscal,fiz1);
2205 fjx2 = _mm_macc_pd(dx12,fscal,fjx2);
2206 fjy2 = _mm_macc_pd(dy12,fscal,fjy2);
2207 fjz2 = _mm_macc_pd(dz12,fscal,fjz2);
2209 /**************************
2210 * CALCULATE INTERACTIONS *
2211 **************************/
2213 r13 = _mm_mul_pd(rsq13,rinv13);
2215 /* Calculate table index by multiplying r with table scale and truncate to integer */
2216 rt = _mm_mul_pd(r13,vftabscale);
2217 vfitab = _mm_cvttpd_epi32(rt);
2219 vfeps = _mm_frcz_pd(rt);
2221 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2223 twovfeps = _mm_add_pd(vfeps,vfeps);
2224 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2226 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2227 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2228 F = _mm_setzero_pd();
2229 GMX_MM_TRANSPOSE2_PD(Y,F);
2230 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
2231 H = _mm_setzero_pd();
2232 GMX_MM_TRANSPOSE2_PD(G,H);
2233 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
2234 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
2235 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq13,FF),_mm_mul_pd(vftabscale,rinv13)));
2239 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2241 /* Update vectorial force */
2242 fix1 = _mm_macc_pd(dx13,fscal,fix1);
2243 fiy1 = _mm_macc_pd(dy13,fscal,fiy1);
2244 fiz1 = _mm_macc_pd(dz13,fscal,fiz1);
2246 fjx3 = _mm_macc_pd(dx13,fscal,fjx3);
2247 fjy3 = _mm_macc_pd(dy13,fscal,fjy3);
2248 fjz3 = _mm_macc_pd(dz13,fscal,fjz3);
2250 /**************************
2251 * CALCULATE INTERACTIONS *
2252 **************************/
2254 r21 = _mm_mul_pd(rsq21,rinv21);
2256 /* Calculate table index by multiplying r with table scale and truncate to integer */
2257 rt = _mm_mul_pd(r21,vftabscale);
2258 vfitab = _mm_cvttpd_epi32(rt);
2260 vfeps = _mm_frcz_pd(rt);
2262 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2264 twovfeps = _mm_add_pd(vfeps,vfeps);
2265 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2267 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2268 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2269 F = _mm_setzero_pd();
2270 GMX_MM_TRANSPOSE2_PD(Y,F);
2271 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
2272 H = _mm_setzero_pd();
2273 GMX_MM_TRANSPOSE2_PD(G,H);
2274 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
2275 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
2276 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
2280 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2282 /* Update vectorial force */
2283 fix2 = _mm_macc_pd(dx21,fscal,fix2);
2284 fiy2 = _mm_macc_pd(dy21,fscal,fiy2);
2285 fiz2 = _mm_macc_pd(dz21,fscal,fiz2);
2287 fjx1 = _mm_macc_pd(dx21,fscal,fjx1);
2288 fjy1 = _mm_macc_pd(dy21,fscal,fjy1);
2289 fjz1 = _mm_macc_pd(dz21,fscal,fjz1);
2291 /**************************
2292 * CALCULATE INTERACTIONS *
2293 **************************/
2295 r22 = _mm_mul_pd(rsq22,rinv22);
2297 /* Calculate table index by multiplying r with table scale and truncate to integer */
2298 rt = _mm_mul_pd(r22,vftabscale);
2299 vfitab = _mm_cvttpd_epi32(rt);
2301 vfeps = _mm_frcz_pd(rt);
2303 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2305 twovfeps = _mm_add_pd(vfeps,vfeps);
2306 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2308 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2309 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2310 F = _mm_setzero_pd();
2311 GMX_MM_TRANSPOSE2_PD(Y,F);
2312 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
2313 H = _mm_setzero_pd();
2314 GMX_MM_TRANSPOSE2_PD(G,H);
2315 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
2316 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
2317 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
2321 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2323 /* Update vectorial force */
2324 fix2 = _mm_macc_pd(dx22,fscal,fix2);
2325 fiy2 = _mm_macc_pd(dy22,fscal,fiy2);
2326 fiz2 = _mm_macc_pd(dz22,fscal,fiz2);
2328 fjx2 = _mm_macc_pd(dx22,fscal,fjx2);
2329 fjy2 = _mm_macc_pd(dy22,fscal,fjy2);
2330 fjz2 = _mm_macc_pd(dz22,fscal,fjz2);
2332 /**************************
2333 * CALCULATE INTERACTIONS *
2334 **************************/
2336 r23 = _mm_mul_pd(rsq23,rinv23);
2338 /* Calculate table index by multiplying r with table scale and truncate to integer */
2339 rt = _mm_mul_pd(r23,vftabscale);
2340 vfitab = _mm_cvttpd_epi32(rt);
2342 vfeps = _mm_frcz_pd(rt);
2344 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2346 twovfeps = _mm_add_pd(vfeps,vfeps);
2347 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2349 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2350 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2351 F = _mm_setzero_pd();
2352 GMX_MM_TRANSPOSE2_PD(Y,F);
2353 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
2354 H = _mm_setzero_pd();
2355 GMX_MM_TRANSPOSE2_PD(G,H);
2356 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
2357 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
2358 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq23,FF),_mm_mul_pd(vftabscale,rinv23)));
2362 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2364 /* Update vectorial force */
2365 fix2 = _mm_macc_pd(dx23,fscal,fix2);
2366 fiy2 = _mm_macc_pd(dy23,fscal,fiy2);
2367 fiz2 = _mm_macc_pd(dz23,fscal,fiz2);
2369 fjx3 = _mm_macc_pd(dx23,fscal,fjx3);
2370 fjy3 = _mm_macc_pd(dy23,fscal,fjy3);
2371 fjz3 = _mm_macc_pd(dz23,fscal,fjz3);
2373 /**************************
2374 * CALCULATE INTERACTIONS *
2375 **************************/
2377 r31 = _mm_mul_pd(rsq31,rinv31);
2379 /* Calculate table index by multiplying r with table scale and truncate to integer */
2380 rt = _mm_mul_pd(r31,vftabscale);
2381 vfitab = _mm_cvttpd_epi32(rt);
2383 vfeps = _mm_frcz_pd(rt);
2385 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2387 twovfeps = _mm_add_pd(vfeps,vfeps);
2388 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2390 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2391 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2392 F = _mm_setzero_pd();
2393 GMX_MM_TRANSPOSE2_PD(Y,F);
2394 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
2395 H = _mm_setzero_pd();
2396 GMX_MM_TRANSPOSE2_PD(G,H);
2397 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
2398 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
2399 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq31,FF),_mm_mul_pd(vftabscale,rinv31)));
2403 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2405 /* Update vectorial force */
2406 fix3 = _mm_macc_pd(dx31,fscal,fix3);
2407 fiy3 = _mm_macc_pd(dy31,fscal,fiy3);
2408 fiz3 = _mm_macc_pd(dz31,fscal,fiz3);
2410 fjx1 = _mm_macc_pd(dx31,fscal,fjx1);
2411 fjy1 = _mm_macc_pd(dy31,fscal,fjy1);
2412 fjz1 = _mm_macc_pd(dz31,fscal,fjz1);
2414 /**************************
2415 * CALCULATE INTERACTIONS *
2416 **************************/
2418 r32 = _mm_mul_pd(rsq32,rinv32);
2420 /* Calculate table index by multiplying r with table scale and truncate to integer */
2421 rt = _mm_mul_pd(r32,vftabscale);
2422 vfitab = _mm_cvttpd_epi32(rt);
2424 vfeps = _mm_frcz_pd(rt);
2426 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2428 twovfeps = _mm_add_pd(vfeps,vfeps);
2429 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2431 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2432 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2433 F = _mm_setzero_pd();
2434 GMX_MM_TRANSPOSE2_PD(Y,F);
2435 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
2436 H = _mm_setzero_pd();
2437 GMX_MM_TRANSPOSE2_PD(G,H);
2438 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
2439 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
2440 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq32,FF),_mm_mul_pd(vftabscale,rinv32)));
2444 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2446 /* Update vectorial force */
2447 fix3 = _mm_macc_pd(dx32,fscal,fix3);
2448 fiy3 = _mm_macc_pd(dy32,fscal,fiy3);
2449 fiz3 = _mm_macc_pd(dz32,fscal,fiz3);
2451 fjx2 = _mm_macc_pd(dx32,fscal,fjx2);
2452 fjy2 = _mm_macc_pd(dy32,fscal,fjy2);
2453 fjz2 = _mm_macc_pd(dz32,fscal,fjz2);
2455 /**************************
2456 * CALCULATE INTERACTIONS *
2457 **************************/
2459 r33 = _mm_mul_pd(rsq33,rinv33);
2461 /* Calculate table index by multiplying r with table scale and truncate to integer */
2462 rt = _mm_mul_pd(r33,vftabscale);
2463 vfitab = _mm_cvttpd_epi32(rt);
2465 vfeps = _mm_frcz_pd(rt);
2467 vfeps = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2469 twovfeps = _mm_add_pd(vfeps,vfeps);
2470 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2472 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2473 Y = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2474 F = _mm_setzero_pd();
2475 GMX_MM_TRANSPOSE2_PD(Y,F);
2476 G = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
2477 H = _mm_setzero_pd();
2478 GMX_MM_TRANSPOSE2_PD(G,H);
2479 Fp = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
2480 FF = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
2481 felec = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq33,FF),_mm_mul_pd(vftabscale,rinv33)));
2485 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2487 /* Update vectorial force */
2488 fix3 = _mm_macc_pd(dx33,fscal,fix3);
2489 fiy3 = _mm_macc_pd(dy33,fscal,fiy3);
2490 fiz3 = _mm_macc_pd(dz33,fscal,fiz3);
2492 fjx3 = _mm_macc_pd(dx33,fscal,fjx3);
2493 fjy3 = _mm_macc_pd(dy33,fscal,fjy3);
2494 fjz3 = _mm_macc_pd(dz33,fscal,fjz3);
2496 gmx_mm_decrement_4rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2498 /* Inner loop uses 432 flops */
2501 /* End of innermost loop */
2503 gmx_mm_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2504 f+i_coord_offset,fshift+i_shift_offset);
2506 /* Increment number of inner iterations */
2507 inneriter += j_index_end - j_index_start;
2509 /* Outer loop uses 24 flops */
2512 /* Increment number of outer iterations */
2515 /* Update outer/inner flops */
2517 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*432);