2 * Note: this file was generated by the Gromacs sse2_double kernel generator.
4 * This source code is part of
8 * Copyright (c) 2001-2012, The GROMACS Development Team
10 * Gromacs is a library for molecular simulation and trajectory analysis,
11 * written by Erik Lindahl, David van der Spoel, Berk Hess, and others - for
12 * a full list of developers and information, check out http://www.gromacs.org
14 * This program is free software; you can redistribute it and/or modify it under
15 * the terms of the GNU Lesser General Public License as published by the Free
16 * Software Foundation; either version 2 of the License, or (at your option) any
19 * To help fund GROMACS development, we humbly ask that you cite
20 * the papers people have written on it - you can find them on the website.
28 #include "../nb_kernel.h"
29 #include "types/simple.h"
33 #include "gmx_math_x86_sse2_double.h"
34 #include "kernelutil_x86_sse2_double.h"
37 * Gromacs nonbonded kernel: nb_kernel_ElecEw_VdwNone_GeomW4W4_VF_sse2_double
38 * Electrostatics interaction: Ewald
39 * VdW interaction: None
40 * Geometry: Water4-Water4
41 * Calculate force/pot: PotentialAndForce
44 nb_kernel_ElecEw_VdwNone_GeomW4W4_VF_sse2_double
45 (t_nblist * gmx_restrict nlist,
46 rvec * gmx_restrict xx,
47 rvec * gmx_restrict ff,
48 t_forcerec * gmx_restrict fr,
49 t_mdatoms * gmx_restrict mdatoms,
50 nb_kernel_data_t * gmx_restrict kernel_data,
51 t_nrnb * gmx_restrict nrnb)
53 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
54 * just 0 for non-waters.
55 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
56 * jnr indices corresponding to data put in the four positions in the SIMD register.
58 int i_shift_offset,i_coord_offset,outeriter,inneriter;
59 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
61 int j_coord_offsetA,j_coord_offsetB;
62 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
64 real *shiftvec,*fshift,*x,*f;
65 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
67 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
69 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
71 __m128d ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
72 int vdwjidx1A,vdwjidx1B;
73 __m128d jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
74 int vdwjidx2A,vdwjidx2B;
75 __m128d jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
76 int vdwjidx3A,vdwjidx3B;
77 __m128d jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
78 __m128d dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
79 __m128d dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
80 __m128d dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
81 __m128d dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
82 __m128d dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
83 __m128d dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
84 __m128d dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
85 __m128d dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
86 __m128d dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
87 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
90 __m128d ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
92 __m128d dummy_mask,cutoff_mask;
93 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
94 __m128d one = _mm_set1_pd(1.0);
95 __m128d two = _mm_set1_pd(2.0);
101 jindex = nlist->jindex;
103 shiftidx = nlist->shift;
105 shiftvec = fr->shift_vec[0];
106 fshift = fr->fshift[0];
107 facel = _mm_set1_pd(fr->epsfac);
108 charge = mdatoms->chargeA;
110 sh_ewald = _mm_set1_pd(fr->ic->sh_ewald);
111 ewtab = fr->ic->tabq_coul_FDV0;
112 ewtabscale = _mm_set1_pd(fr->ic->tabq_scale);
113 ewtabhalfspace = _mm_set1_pd(0.5/fr->ic->tabq_scale);
115 /* Setup water-specific parameters */
116 inr = nlist->iinr[0];
117 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
118 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
119 iq3 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
121 jq1 = _mm_set1_pd(charge[inr+1]);
122 jq2 = _mm_set1_pd(charge[inr+2]);
123 jq3 = _mm_set1_pd(charge[inr+3]);
124 qq11 = _mm_mul_pd(iq1,jq1);
125 qq12 = _mm_mul_pd(iq1,jq2);
126 qq13 = _mm_mul_pd(iq1,jq3);
127 qq21 = _mm_mul_pd(iq2,jq1);
128 qq22 = _mm_mul_pd(iq2,jq2);
129 qq23 = _mm_mul_pd(iq2,jq3);
130 qq31 = _mm_mul_pd(iq3,jq1);
131 qq32 = _mm_mul_pd(iq3,jq2);
132 qq33 = _mm_mul_pd(iq3,jq3);
134 /* Avoid stupid compiler warnings */
142 /* Start outer loop over neighborlists */
143 for(iidx=0; iidx<nri; iidx++)
145 /* Load shift vector for this list */
146 i_shift_offset = DIM*shiftidx[iidx];
148 /* Load limits for loop over neighbors */
149 j_index_start = jindex[iidx];
150 j_index_end = jindex[iidx+1];
152 /* Get outer coordinate index */
154 i_coord_offset = DIM*inr;
156 /* Load i particle coords and add shift vector */
157 gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
158 &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
160 fix1 = _mm_setzero_pd();
161 fiy1 = _mm_setzero_pd();
162 fiz1 = _mm_setzero_pd();
163 fix2 = _mm_setzero_pd();
164 fiy2 = _mm_setzero_pd();
165 fiz2 = _mm_setzero_pd();
166 fix3 = _mm_setzero_pd();
167 fiy3 = _mm_setzero_pd();
168 fiz3 = _mm_setzero_pd();
170 /* Reset potential sums */
171 velecsum = _mm_setzero_pd();
173 /* Start inner kernel loop */
174 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
177 /* Get j neighbor index, and coordinate index */
180 j_coord_offsetA = DIM*jnrA;
181 j_coord_offsetB = DIM*jnrB;
183 /* load j atom coordinates */
184 gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
185 &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
187 /* Calculate displacement vector */
188 dx11 = _mm_sub_pd(ix1,jx1);
189 dy11 = _mm_sub_pd(iy1,jy1);
190 dz11 = _mm_sub_pd(iz1,jz1);
191 dx12 = _mm_sub_pd(ix1,jx2);
192 dy12 = _mm_sub_pd(iy1,jy2);
193 dz12 = _mm_sub_pd(iz1,jz2);
194 dx13 = _mm_sub_pd(ix1,jx3);
195 dy13 = _mm_sub_pd(iy1,jy3);
196 dz13 = _mm_sub_pd(iz1,jz3);
197 dx21 = _mm_sub_pd(ix2,jx1);
198 dy21 = _mm_sub_pd(iy2,jy1);
199 dz21 = _mm_sub_pd(iz2,jz1);
200 dx22 = _mm_sub_pd(ix2,jx2);
201 dy22 = _mm_sub_pd(iy2,jy2);
202 dz22 = _mm_sub_pd(iz2,jz2);
203 dx23 = _mm_sub_pd(ix2,jx3);
204 dy23 = _mm_sub_pd(iy2,jy3);
205 dz23 = _mm_sub_pd(iz2,jz3);
206 dx31 = _mm_sub_pd(ix3,jx1);
207 dy31 = _mm_sub_pd(iy3,jy1);
208 dz31 = _mm_sub_pd(iz3,jz1);
209 dx32 = _mm_sub_pd(ix3,jx2);
210 dy32 = _mm_sub_pd(iy3,jy2);
211 dz32 = _mm_sub_pd(iz3,jz2);
212 dx33 = _mm_sub_pd(ix3,jx3);
213 dy33 = _mm_sub_pd(iy3,jy3);
214 dz33 = _mm_sub_pd(iz3,jz3);
216 /* Calculate squared distance and things based on it */
217 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
218 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
219 rsq13 = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
220 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
221 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
222 rsq23 = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
223 rsq31 = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
224 rsq32 = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
225 rsq33 = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
227 rinv11 = gmx_mm_invsqrt_pd(rsq11);
228 rinv12 = gmx_mm_invsqrt_pd(rsq12);
229 rinv13 = gmx_mm_invsqrt_pd(rsq13);
230 rinv21 = gmx_mm_invsqrt_pd(rsq21);
231 rinv22 = gmx_mm_invsqrt_pd(rsq22);
232 rinv23 = gmx_mm_invsqrt_pd(rsq23);
233 rinv31 = gmx_mm_invsqrt_pd(rsq31);
234 rinv32 = gmx_mm_invsqrt_pd(rsq32);
235 rinv33 = gmx_mm_invsqrt_pd(rsq33);
237 rinvsq11 = _mm_mul_pd(rinv11,rinv11);
238 rinvsq12 = _mm_mul_pd(rinv12,rinv12);
239 rinvsq13 = _mm_mul_pd(rinv13,rinv13);
240 rinvsq21 = _mm_mul_pd(rinv21,rinv21);
241 rinvsq22 = _mm_mul_pd(rinv22,rinv22);
242 rinvsq23 = _mm_mul_pd(rinv23,rinv23);
243 rinvsq31 = _mm_mul_pd(rinv31,rinv31);
244 rinvsq32 = _mm_mul_pd(rinv32,rinv32);
245 rinvsq33 = _mm_mul_pd(rinv33,rinv33);
247 fjx1 = _mm_setzero_pd();
248 fjy1 = _mm_setzero_pd();
249 fjz1 = _mm_setzero_pd();
250 fjx2 = _mm_setzero_pd();
251 fjy2 = _mm_setzero_pd();
252 fjz2 = _mm_setzero_pd();
253 fjx3 = _mm_setzero_pd();
254 fjy3 = _mm_setzero_pd();
255 fjz3 = _mm_setzero_pd();
257 /**************************
258 * CALCULATE INTERACTIONS *
259 **************************/
261 r11 = _mm_mul_pd(rsq11,rinv11);
263 /* EWALD ELECTROSTATICS */
265 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
266 ewrt = _mm_mul_pd(r11,ewtabscale);
267 ewitab = _mm_cvttpd_epi32(ewrt);
268 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
269 ewitab = _mm_slli_epi32(ewitab,2);
270 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
271 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
272 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
273 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
274 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
275 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
276 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
277 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
278 velec = _mm_mul_pd(qq11,_mm_sub_pd(rinv11,velec));
279 felec = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
281 /* Update potential sum for this i atom from the interaction with this j atom. */
282 velecsum = _mm_add_pd(velecsum,velec);
286 /* Calculate temporary vectorial force */
287 tx = _mm_mul_pd(fscal,dx11);
288 ty = _mm_mul_pd(fscal,dy11);
289 tz = _mm_mul_pd(fscal,dz11);
291 /* Update vectorial force */
292 fix1 = _mm_add_pd(fix1,tx);
293 fiy1 = _mm_add_pd(fiy1,ty);
294 fiz1 = _mm_add_pd(fiz1,tz);
296 fjx1 = _mm_add_pd(fjx1,tx);
297 fjy1 = _mm_add_pd(fjy1,ty);
298 fjz1 = _mm_add_pd(fjz1,tz);
300 /**************************
301 * CALCULATE INTERACTIONS *
302 **************************/
304 r12 = _mm_mul_pd(rsq12,rinv12);
306 /* EWALD ELECTROSTATICS */
308 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
309 ewrt = _mm_mul_pd(r12,ewtabscale);
310 ewitab = _mm_cvttpd_epi32(ewrt);
311 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
312 ewitab = _mm_slli_epi32(ewitab,2);
313 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
314 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
315 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
316 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
317 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
318 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
319 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
320 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
321 velec = _mm_mul_pd(qq12,_mm_sub_pd(rinv12,velec));
322 felec = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
324 /* Update potential sum for this i atom from the interaction with this j atom. */
325 velecsum = _mm_add_pd(velecsum,velec);
329 /* Calculate temporary vectorial force */
330 tx = _mm_mul_pd(fscal,dx12);
331 ty = _mm_mul_pd(fscal,dy12);
332 tz = _mm_mul_pd(fscal,dz12);
334 /* Update vectorial force */
335 fix1 = _mm_add_pd(fix1,tx);
336 fiy1 = _mm_add_pd(fiy1,ty);
337 fiz1 = _mm_add_pd(fiz1,tz);
339 fjx2 = _mm_add_pd(fjx2,tx);
340 fjy2 = _mm_add_pd(fjy2,ty);
341 fjz2 = _mm_add_pd(fjz2,tz);
343 /**************************
344 * CALCULATE INTERACTIONS *
345 **************************/
347 r13 = _mm_mul_pd(rsq13,rinv13);
349 /* EWALD ELECTROSTATICS */
351 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
352 ewrt = _mm_mul_pd(r13,ewtabscale);
353 ewitab = _mm_cvttpd_epi32(ewrt);
354 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
355 ewitab = _mm_slli_epi32(ewitab,2);
356 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
357 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
358 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
359 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
360 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
361 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
362 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
363 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
364 velec = _mm_mul_pd(qq13,_mm_sub_pd(rinv13,velec));
365 felec = _mm_mul_pd(_mm_mul_pd(qq13,rinv13),_mm_sub_pd(rinvsq13,felec));
367 /* Update potential sum for this i atom from the interaction with this j atom. */
368 velecsum = _mm_add_pd(velecsum,velec);
372 /* Calculate temporary vectorial force */
373 tx = _mm_mul_pd(fscal,dx13);
374 ty = _mm_mul_pd(fscal,dy13);
375 tz = _mm_mul_pd(fscal,dz13);
377 /* Update vectorial force */
378 fix1 = _mm_add_pd(fix1,tx);
379 fiy1 = _mm_add_pd(fiy1,ty);
380 fiz1 = _mm_add_pd(fiz1,tz);
382 fjx3 = _mm_add_pd(fjx3,tx);
383 fjy3 = _mm_add_pd(fjy3,ty);
384 fjz3 = _mm_add_pd(fjz3,tz);
386 /**************************
387 * CALCULATE INTERACTIONS *
388 **************************/
390 r21 = _mm_mul_pd(rsq21,rinv21);
392 /* EWALD ELECTROSTATICS */
394 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
395 ewrt = _mm_mul_pd(r21,ewtabscale);
396 ewitab = _mm_cvttpd_epi32(ewrt);
397 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
398 ewitab = _mm_slli_epi32(ewitab,2);
399 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
400 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
401 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
402 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
403 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
404 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
405 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
406 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
407 velec = _mm_mul_pd(qq21,_mm_sub_pd(rinv21,velec));
408 felec = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
410 /* Update potential sum for this i atom from the interaction with this j atom. */
411 velecsum = _mm_add_pd(velecsum,velec);
415 /* Calculate temporary vectorial force */
416 tx = _mm_mul_pd(fscal,dx21);
417 ty = _mm_mul_pd(fscal,dy21);
418 tz = _mm_mul_pd(fscal,dz21);
420 /* Update vectorial force */
421 fix2 = _mm_add_pd(fix2,tx);
422 fiy2 = _mm_add_pd(fiy2,ty);
423 fiz2 = _mm_add_pd(fiz2,tz);
425 fjx1 = _mm_add_pd(fjx1,tx);
426 fjy1 = _mm_add_pd(fjy1,ty);
427 fjz1 = _mm_add_pd(fjz1,tz);
429 /**************************
430 * CALCULATE INTERACTIONS *
431 **************************/
433 r22 = _mm_mul_pd(rsq22,rinv22);
435 /* EWALD ELECTROSTATICS */
437 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
438 ewrt = _mm_mul_pd(r22,ewtabscale);
439 ewitab = _mm_cvttpd_epi32(ewrt);
440 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
441 ewitab = _mm_slli_epi32(ewitab,2);
442 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
443 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
444 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
445 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
446 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
447 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
448 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
449 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
450 velec = _mm_mul_pd(qq22,_mm_sub_pd(rinv22,velec));
451 felec = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
453 /* Update potential sum for this i atom from the interaction with this j atom. */
454 velecsum = _mm_add_pd(velecsum,velec);
458 /* Calculate temporary vectorial force */
459 tx = _mm_mul_pd(fscal,dx22);
460 ty = _mm_mul_pd(fscal,dy22);
461 tz = _mm_mul_pd(fscal,dz22);
463 /* Update vectorial force */
464 fix2 = _mm_add_pd(fix2,tx);
465 fiy2 = _mm_add_pd(fiy2,ty);
466 fiz2 = _mm_add_pd(fiz2,tz);
468 fjx2 = _mm_add_pd(fjx2,tx);
469 fjy2 = _mm_add_pd(fjy2,ty);
470 fjz2 = _mm_add_pd(fjz2,tz);
472 /**************************
473 * CALCULATE INTERACTIONS *
474 **************************/
476 r23 = _mm_mul_pd(rsq23,rinv23);
478 /* EWALD ELECTROSTATICS */
480 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
481 ewrt = _mm_mul_pd(r23,ewtabscale);
482 ewitab = _mm_cvttpd_epi32(ewrt);
483 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
484 ewitab = _mm_slli_epi32(ewitab,2);
485 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
486 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
487 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
488 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
489 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
490 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
491 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
492 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
493 velec = _mm_mul_pd(qq23,_mm_sub_pd(rinv23,velec));
494 felec = _mm_mul_pd(_mm_mul_pd(qq23,rinv23),_mm_sub_pd(rinvsq23,felec));
496 /* Update potential sum for this i atom from the interaction with this j atom. */
497 velecsum = _mm_add_pd(velecsum,velec);
501 /* Calculate temporary vectorial force */
502 tx = _mm_mul_pd(fscal,dx23);
503 ty = _mm_mul_pd(fscal,dy23);
504 tz = _mm_mul_pd(fscal,dz23);
506 /* Update vectorial force */
507 fix2 = _mm_add_pd(fix2,tx);
508 fiy2 = _mm_add_pd(fiy2,ty);
509 fiz2 = _mm_add_pd(fiz2,tz);
511 fjx3 = _mm_add_pd(fjx3,tx);
512 fjy3 = _mm_add_pd(fjy3,ty);
513 fjz3 = _mm_add_pd(fjz3,tz);
515 /**************************
516 * CALCULATE INTERACTIONS *
517 **************************/
519 r31 = _mm_mul_pd(rsq31,rinv31);
521 /* EWALD ELECTROSTATICS */
523 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
524 ewrt = _mm_mul_pd(r31,ewtabscale);
525 ewitab = _mm_cvttpd_epi32(ewrt);
526 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
527 ewitab = _mm_slli_epi32(ewitab,2);
528 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
529 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
530 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
531 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
532 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
533 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
534 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
535 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
536 velec = _mm_mul_pd(qq31,_mm_sub_pd(rinv31,velec));
537 felec = _mm_mul_pd(_mm_mul_pd(qq31,rinv31),_mm_sub_pd(rinvsq31,felec));
539 /* Update potential sum for this i atom from the interaction with this j atom. */
540 velecsum = _mm_add_pd(velecsum,velec);
544 /* Calculate temporary vectorial force */
545 tx = _mm_mul_pd(fscal,dx31);
546 ty = _mm_mul_pd(fscal,dy31);
547 tz = _mm_mul_pd(fscal,dz31);
549 /* Update vectorial force */
550 fix3 = _mm_add_pd(fix3,tx);
551 fiy3 = _mm_add_pd(fiy3,ty);
552 fiz3 = _mm_add_pd(fiz3,tz);
554 fjx1 = _mm_add_pd(fjx1,tx);
555 fjy1 = _mm_add_pd(fjy1,ty);
556 fjz1 = _mm_add_pd(fjz1,tz);
558 /**************************
559 * CALCULATE INTERACTIONS *
560 **************************/
562 r32 = _mm_mul_pd(rsq32,rinv32);
564 /* EWALD ELECTROSTATICS */
566 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
567 ewrt = _mm_mul_pd(r32,ewtabscale);
568 ewitab = _mm_cvttpd_epi32(ewrt);
569 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
570 ewitab = _mm_slli_epi32(ewitab,2);
571 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
572 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
573 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
574 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
575 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
576 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
577 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
578 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
579 velec = _mm_mul_pd(qq32,_mm_sub_pd(rinv32,velec));
580 felec = _mm_mul_pd(_mm_mul_pd(qq32,rinv32),_mm_sub_pd(rinvsq32,felec));
582 /* Update potential sum for this i atom from the interaction with this j atom. */
583 velecsum = _mm_add_pd(velecsum,velec);
587 /* Calculate temporary vectorial force */
588 tx = _mm_mul_pd(fscal,dx32);
589 ty = _mm_mul_pd(fscal,dy32);
590 tz = _mm_mul_pd(fscal,dz32);
592 /* Update vectorial force */
593 fix3 = _mm_add_pd(fix3,tx);
594 fiy3 = _mm_add_pd(fiy3,ty);
595 fiz3 = _mm_add_pd(fiz3,tz);
597 fjx2 = _mm_add_pd(fjx2,tx);
598 fjy2 = _mm_add_pd(fjy2,ty);
599 fjz2 = _mm_add_pd(fjz2,tz);
601 /**************************
602 * CALCULATE INTERACTIONS *
603 **************************/
605 r33 = _mm_mul_pd(rsq33,rinv33);
607 /* EWALD ELECTROSTATICS */
609 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
610 ewrt = _mm_mul_pd(r33,ewtabscale);
611 ewitab = _mm_cvttpd_epi32(ewrt);
612 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
613 ewitab = _mm_slli_epi32(ewitab,2);
614 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
615 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
616 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
617 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
618 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
619 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
620 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
621 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
622 velec = _mm_mul_pd(qq33,_mm_sub_pd(rinv33,velec));
623 felec = _mm_mul_pd(_mm_mul_pd(qq33,rinv33),_mm_sub_pd(rinvsq33,felec));
625 /* Update potential sum for this i atom from the interaction with this j atom. */
626 velecsum = _mm_add_pd(velecsum,velec);
630 /* Calculate temporary vectorial force */
631 tx = _mm_mul_pd(fscal,dx33);
632 ty = _mm_mul_pd(fscal,dy33);
633 tz = _mm_mul_pd(fscal,dz33);
635 /* Update vectorial force */
636 fix3 = _mm_add_pd(fix3,tx);
637 fiy3 = _mm_add_pd(fiy3,ty);
638 fiz3 = _mm_add_pd(fiz3,tz);
640 fjx3 = _mm_add_pd(fjx3,tx);
641 fjy3 = _mm_add_pd(fjy3,ty);
642 fjz3 = _mm_add_pd(fjz3,tz);
644 gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA+DIM,f+j_coord_offsetB+DIM,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
646 /* Inner loop uses 369 flops */
653 j_coord_offsetA = DIM*jnrA;
655 /* load j atom coordinates */
656 gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA+DIM,
657 &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
659 /* Calculate displacement vector */
660 dx11 = _mm_sub_pd(ix1,jx1);
661 dy11 = _mm_sub_pd(iy1,jy1);
662 dz11 = _mm_sub_pd(iz1,jz1);
663 dx12 = _mm_sub_pd(ix1,jx2);
664 dy12 = _mm_sub_pd(iy1,jy2);
665 dz12 = _mm_sub_pd(iz1,jz2);
666 dx13 = _mm_sub_pd(ix1,jx3);
667 dy13 = _mm_sub_pd(iy1,jy3);
668 dz13 = _mm_sub_pd(iz1,jz3);
669 dx21 = _mm_sub_pd(ix2,jx1);
670 dy21 = _mm_sub_pd(iy2,jy1);
671 dz21 = _mm_sub_pd(iz2,jz1);
672 dx22 = _mm_sub_pd(ix2,jx2);
673 dy22 = _mm_sub_pd(iy2,jy2);
674 dz22 = _mm_sub_pd(iz2,jz2);
675 dx23 = _mm_sub_pd(ix2,jx3);
676 dy23 = _mm_sub_pd(iy2,jy3);
677 dz23 = _mm_sub_pd(iz2,jz3);
678 dx31 = _mm_sub_pd(ix3,jx1);
679 dy31 = _mm_sub_pd(iy3,jy1);
680 dz31 = _mm_sub_pd(iz3,jz1);
681 dx32 = _mm_sub_pd(ix3,jx2);
682 dy32 = _mm_sub_pd(iy3,jy2);
683 dz32 = _mm_sub_pd(iz3,jz2);
684 dx33 = _mm_sub_pd(ix3,jx3);
685 dy33 = _mm_sub_pd(iy3,jy3);
686 dz33 = _mm_sub_pd(iz3,jz3);
688 /* Calculate squared distance and things based on it */
689 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
690 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
691 rsq13 = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
692 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
693 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
694 rsq23 = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
695 rsq31 = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
696 rsq32 = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
697 rsq33 = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
699 rinv11 = gmx_mm_invsqrt_pd(rsq11);
700 rinv12 = gmx_mm_invsqrt_pd(rsq12);
701 rinv13 = gmx_mm_invsqrt_pd(rsq13);
702 rinv21 = gmx_mm_invsqrt_pd(rsq21);
703 rinv22 = gmx_mm_invsqrt_pd(rsq22);
704 rinv23 = gmx_mm_invsqrt_pd(rsq23);
705 rinv31 = gmx_mm_invsqrt_pd(rsq31);
706 rinv32 = gmx_mm_invsqrt_pd(rsq32);
707 rinv33 = gmx_mm_invsqrt_pd(rsq33);
709 rinvsq11 = _mm_mul_pd(rinv11,rinv11);
710 rinvsq12 = _mm_mul_pd(rinv12,rinv12);
711 rinvsq13 = _mm_mul_pd(rinv13,rinv13);
712 rinvsq21 = _mm_mul_pd(rinv21,rinv21);
713 rinvsq22 = _mm_mul_pd(rinv22,rinv22);
714 rinvsq23 = _mm_mul_pd(rinv23,rinv23);
715 rinvsq31 = _mm_mul_pd(rinv31,rinv31);
716 rinvsq32 = _mm_mul_pd(rinv32,rinv32);
717 rinvsq33 = _mm_mul_pd(rinv33,rinv33);
719 fjx1 = _mm_setzero_pd();
720 fjy1 = _mm_setzero_pd();
721 fjz1 = _mm_setzero_pd();
722 fjx2 = _mm_setzero_pd();
723 fjy2 = _mm_setzero_pd();
724 fjz2 = _mm_setzero_pd();
725 fjx3 = _mm_setzero_pd();
726 fjy3 = _mm_setzero_pd();
727 fjz3 = _mm_setzero_pd();
729 /**************************
730 * CALCULATE INTERACTIONS *
731 **************************/
733 r11 = _mm_mul_pd(rsq11,rinv11);
735 /* EWALD ELECTROSTATICS */
737 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
738 ewrt = _mm_mul_pd(r11,ewtabscale);
739 ewitab = _mm_cvttpd_epi32(ewrt);
740 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
741 ewitab = _mm_slli_epi32(ewitab,2);
742 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
743 ewtabD = _mm_setzero_pd();
744 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
745 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
746 ewtabFn = _mm_setzero_pd();
747 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
748 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
749 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
750 velec = _mm_mul_pd(qq11,_mm_sub_pd(rinv11,velec));
751 felec = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
753 /* Update potential sum for this i atom from the interaction with this j atom. */
754 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
755 velecsum = _mm_add_pd(velecsum,velec);
759 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
761 /* Calculate temporary vectorial force */
762 tx = _mm_mul_pd(fscal,dx11);
763 ty = _mm_mul_pd(fscal,dy11);
764 tz = _mm_mul_pd(fscal,dz11);
766 /* Update vectorial force */
767 fix1 = _mm_add_pd(fix1,tx);
768 fiy1 = _mm_add_pd(fiy1,ty);
769 fiz1 = _mm_add_pd(fiz1,tz);
771 fjx1 = _mm_add_pd(fjx1,tx);
772 fjy1 = _mm_add_pd(fjy1,ty);
773 fjz1 = _mm_add_pd(fjz1,tz);
775 /**************************
776 * CALCULATE INTERACTIONS *
777 **************************/
779 r12 = _mm_mul_pd(rsq12,rinv12);
781 /* EWALD ELECTROSTATICS */
783 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
784 ewrt = _mm_mul_pd(r12,ewtabscale);
785 ewitab = _mm_cvttpd_epi32(ewrt);
786 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
787 ewitab = _mm_slli_epi32(ewitab,2);
788 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
789 ewtabD = _mm_setzero_pd();
790 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
791 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
792 ewtabFn = _mm_setzero_pd();
793 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
794 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
795 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
796 velec = _mm_mul_pd(qq12,_mm_sub_pd(rinv12,velec));
797 felec = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
799 /* Update potential sum for this i atom from the interaction with this j atom. */
800 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
801 velecsum = _mm_add_pd(velecsum,velec);
805 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
807 /* Calculate temporary vectorial force */
808 tx = _mm_mul_pd(fscal,dx12);
809 ty = _mm_mul_pd(fscal,dy12);
810 tz = _mm_mul_pd(fscal,dz12);
812 /* Update vectorial force */
813 fix1 = _mm_add_pd(fix1,tx);
814 fiy1 = _mm_add_pd(fiy1,ty);
815 fiz1 = _mm_add_pd(fiz1,tz);
817 fjx2 = _mm_add_pd(fjx2,tx);
818 fjy2 = _mm_add_pd(fjy2,ty);
819 fjz2 = _mm_add_pd(fjz2,tz);
821 /**************************
822 * CALCULATE INTERACTIONS *
823 **************************/
825 r13 = _mm_mul_pd(rsq13,rinv13);
827 /* EWALD ELECTROSTATICS */
829 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
830 ewrt = _mm_mul_pd(r13,ewtabscale);
831 ewitab = _mm_cvttpd_epi32(ewrt);
832 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
833 ewitab = _mm_slli_epi32(ewitab,2);
834 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
835 ewtabD = _mm_setzero_pd();
836 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
837 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
838 ewtabFn = _mm_setzero_pd();
839 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
840 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
841 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
842 velec = _mm_mul_pd(qq13,_mm_sub_pd(rinv13,velec));
843 felec = _mm_mul_pd(_mm_mul_pd(qq13,rinv13),_mm_sub_pd(rinvsq13,felec));
845 /* Update potential sum for this i atom from the interaction with this j atom. */
846 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
847 velecsum = _mm_add_pd(velecsum,velec);
851 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
853 /* Calculate temporary vectorial force */
854 tx = _mm_mul_pd(fscal,dx13);
855 ty = _mm_mul_pd(fscal,dy13);
856 tz = _mm_mul_pd(fscal,dz13);
858 /* Update vectorial force */
859 fix1 = _mm_add_pd(fix1,tx);
860 fiy1 = _mm_add_pd(fiy1,ty);
861 fiz1 = _mm_add_pd(fiz1,tz);
863 fjx3 = _mm_add_pd(fjx3,tx);
864 fjy3 = _mm_add_pd(fjy3,ty);
865 fjz3 = _mm_add_pd(fjz3,tz);
867 /**************************
868 * CALCULATE INTERACTIONS *
869 **************************/
871 r21 = _mm_mul_pd(rsq21,rinv21);
873 /* EWALD ELECTROSTATICS */
875 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
876 ewrt = _mm_mul_pd(r21,ewtabscale);
877 ewitab = _mm_cvttpd_epi32(ewrt);
878 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
879 ewitab = _mm_slli_epi32(ewitab,2);
880 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
881 ewtabD = _mm_setzero_pd();
882 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
883 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
884 ewtabFn = _mm_setzero_pd();
885 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
886 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
887 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
888 velec = _mm_mul_pd(qq21,_mm_sub_pd(rinv21,velec));
889 felec = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
891 /* Update potential sum for this i atom from the interaction with this j atom. */
892 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
893 velecsum = _mm_add_pd(velecsum,velec);
897 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
899 /* Calculate temporary vectorial force */
900 tx = _mm_mul_pd(fscal,dx21);
901 ty = _mm_mul_pd(fscal,dy21);
902 tz = _mm_mul_pd(fscal,dz21);
904 /* Update vectorial force */
905 fix2 = _mm_add_pd(fix2,tx);
906 fiy2 = _mm_add_pd(fiy2,ty);
907 fiz2 = _mm_add_pd(fiz2,tz);
909 fjx1 = _mm_add_pd(fjx1,tx);
910 fjy1 = _mm_add_pd(fjy1,ty);
911 fjz1 = _mm_add_pd(fjz1,tz);
913 /**************************
914 * CALCULATE INTERACTIONS *
915 **************************/
917 r22 = _mm_mul_pd(rsq22,rinv22);
919 /* EWALD ELECTROSTATICS */
921 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
922 ewrt = _mm_mul_pd(r22,ewtabscale);
923 ewitab = _mm_cvttpd_epi32(ewrt);
924 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
925 ewitab = _mm_slli_epi32(ewitab,2);
926 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
927 ewtabD = _mm_setzero_pd();
928 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
929 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
930 ewtabFn = _mm_setzero_pd();
931 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
932 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
933 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
934 velec = _mm_mul_pd(qq22,_mm_sub_pd(rinv22,velec));
935 felec = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
937 /* Update potential sum for this i atom from the interaction with this j atom. */
938 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
939 velecsum = _mm_add_pd(velecsum,velec);
943 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
945 /* Calculate temporary vectorial force */
946 tx = _mm_mul_pd(fscal,dx22);
947 ty = _mm_mul_pd(fscal,dy22);
948 tz = _mm_mul_pd(fscal,dz22);
950 /* Update vectorial force */
951 fix2 = _mm_add_pd(fix2,tx);
952 fiy2 = _mm_add_pd(fiy2,ty);
953 fiz2 = _mm_add_pd(fiz2,tz);
955 fjx2 = _mm_add_pd(fjx2,tx);
956 fjy2 = _mm_add_pd(fjy2,ty);
957 fjz2 = _mm_add_pd(fjz2,tz);
959 /**************************
960 * CALCULATE INTERACTIONS *
961 **************************/
963 r23 = _mm_mul_pd(rsq23,rinv23);
965 /* EWALD ELECTROSTATICS */
967 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
968 ewrt = _mm_mul_pd(r23,ewtabscale);
969 ewitab = _mm_cvttpd_epi32(ewrt);
970 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
971 ewitab = _mm_slli_epi32(ewitab,2);
972 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
973 ewtabD = _mm_setzero_pd();
974 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
975 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
976 ewtabFn = _mm_setzero_pd();
977 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
978 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
979 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
980 velec = _mm_mul_pd(qq23,_mm_sub_pd(rinv23,velec));
981 felec = _mm_mul_pd(_mm_mul_pd(qq23,rinv23),_mm_sub_pd(rinvsq23,felec));
983 /* Update potential sum for this i atom from the interaction with this j atom. */
984 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
985 velecsum = _mm_add_pd(velecsum,velec);
989 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
991 /* Calculate temporary vectorial force */
992 tx = _mm_mul_pd(fscal,dx23);
993 ty = _mm_mul_pd(fscal,dy23);
994 tz = _mm_mul_pd(fscal,dz23);
996 /* Update vectorial force */
997 fix2 = _mm_add_pd(fix2,tx);
998 fiy2 = _mm_add_pd(fiy2,ty);
999 fiz2 = _mm_add_pd(fiz2,tz);
1001 fjx3 = _mm_add_pd(fjx3,tx);
1002 fjy3 = _mm_add_pd(fjy3,ty);
1003 fjz3 = _mm_add_pd(fjz3,tz);
1005 /**************************
1006 * CALCULATE INTERACTIONS *
1007 **************************/
1009 r31 = _mm_mul_pd(rsq31,rinv31);
1011 /* EWALD ELECTROSTATICS */
1013 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1014 ewrt = _mm_mul_pd(r31,ewtabscale);
1015 ewitab = _mm_cvttpd_epi32(ewrt);
1016 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1017 ewitab = _mm_slli_epi32(ewitab,2);
1018 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1019 ewtabD = _mm_setzero_pd();
1020 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1021 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1022 ewtabFn = _mm_setzero_pd();
1023 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1024 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1025 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1026 velec = _mm_mul_pd(qq31,_mm_sub_pd(rinv31,velec));
1027 felec = _mm_mul_pd(_mm_mul_pd(qq31,rinv31),_mm_sub_pd(rinvsq31,felec));
1029 /* Update potential sum for this i atom from the interaction with this j atom. */
1030 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1031 velecsum = _mm_add_pd(velecsum,velec);
1035 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1037 /* Calculate temporary vectorial force */
1038 tx = _mm_mul_pd(fscal,dx31);
1039 ty = _mm_mul_pd(fscal,dy31);
1040 tz = _mm_mul_pd(fscal,dz31);
1042 /* Update vectorial force */
1043 fix3 = _mm_add_pd(fix3,tx);
1044 fiy3 = _mm_add_pd(fiy3,ty);
1045 fiz3 = _mm_add_pd(fiz3,tz);
1047 fjx1 = _mm_add_pd(fjx1,tx);
1048 fjy1 = _mm_add_pd(fjy1,ty);
1049 fjz1 = _mm_add_pd(fjz1,tz);
1051 /**************************
1052 * CALCULATE INTERACTIONS *
1053 **************************/
1055 r32 = _mm_mul_pd(rsq32,rinv32);
1057 /* EWALD ELECTROSTATICS */
1059 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1060 ewrt = _mm_mul_pd(r32,ewtabscale);
1061 ewitab = _mm_cvttpd_epi32(ewrt);
1062 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1063 ewitab = _mm_slli_epi32(ewitab,2);
1064 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1065 ewtabD = _mm_setzero_pd();
1066 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1067 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1068 ewtabFn = _mm_setzero_pd();
1069 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1070 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1071 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1072 velec = _mm_mul_pd(qq32,_mm_sub_pd(rinv32,velec));
1073 felec = _mm_mul_pd(_mm_mul_pd(qq32,rinv32),_mm_sub_pd(rinvsq32,felec));
1075 /* Update potential sum for this i atom from the interaction with this j atom. */
1076 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1077 velecsum = _mm_add_pd(velecsum,velec);
1081 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1083 /* Calculate temporary vectorial force */
1084 tx = _mm_mul_pd(fscal,dx32);
1085 ty = _mm_mul_pd(fscal,dy32);
1086 tz = _mm_mul_pd(fscal,dz32);
1088 /* Update vectorial force */
1089 fix3 = _mm_add_pd(fix3,tx);
1090 fiy3 = _mm_add_pd(fiy3,ty);
1091 fiz3 = _mm_add_pd(fiz3,tz);
1093 fjx2 = _mm_add_pd(fjx2,tx);
1094 fjy2 = _mm_add_pd(fjy2,ty);
1095 fjz2 = _mm_add_pd(fjz2,tz);
1097 /**************************
1098 * CALCULATE INTERACTIONS *
1099 **************************/
1101 r33 = _mm_mul_pd(rsq33,rinv33);
1103 /* EWALD ELECTROSTATICS */
1105 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1106 ewrt = _mm_mul_pd(r33,ewtabscale);
1107 ewitab = _mm_cvttpd_epi32(ewrt);
1108 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1109 ewitab = _mm_slli_epi32(ewitab,2);
1110 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1111 ewtabD = _mm_setzero_pd();
1112 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1113 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1114 ewtabFn = _mm_setzero_pd();
1115 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1116 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1117 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1118 velec = _mm_mul_pd(qq33,_mm_sub_pd(rinv33,velec));
1119 felec = _mm_mul_pd(_mm_mul_pd(qq33,rinv33),_mm_sub_pd(rinvsq33,felec));
1121 /* Update potential sum for this i atom from the interaction with this j atom. */
1122 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1123 velecsum = _mm_add_pd(velecsum,velec);
1127 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1129 /* Calculate temporary vectorial force */
1130 tx = _mm_mul_pd(fscal,dx33);
1131 ty = _mm_mul_pd(fscal,dy33);
1132 tz = _mm_mul_pd(fscal,dz33);
1134 /* Update vectorial force */
1135 fix3 = _mm_add_pd(fix3,tx);
1136 fiy3 = _mm_add_pd(fiy3,ty);
1137 fiz3 = _mm_add_pd(fiz3,tz);
1139 fjx3 = _mm_add_pd(fjx3,tx);
1140 fjy3 = _mm_add_pd(fjy3,ty);
1141 fjz3 = _mm_add_pd(fjz3,tz);
1143 gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA+DIM,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1145 /* Inner loop uses 369 flops */
1148 /* End of innermost loop */
1150 gmx_mm_update_iforce_3atom_swizzle_pd(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1151 f+i_coord_offset+DIM,fshift+i_shift_offset);
1154 /* Update potential energies */
1155 gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1157 /* Increment number of inner iterations */
1158 inneriter += j_index_end - j_index_start;
1160 /* Outer loop uses 19 flops */
1163 /* Increment number of outer iterations */
1166 /* Update outer/inner flops */
1168 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4W4_VF,outeriter*19 + inneriter*369);
1171 * Gromacs nonbonded kernel: nb_kernel_ElecEw_VdwNone_GeomW4W4_F_sse2_double
1172 * Electrostatics interaction: Ewald
1173 * VdW interaction: None
1174 * Geometry: Water4-Water4
1175 * Calculate force/pot: Force
1178 nb_kernel_ElecEw_VdwNone_GeomW4W4_F_sse2_double
1179 (t_nblist * gmx_restrict nlist,
1180 rvec * gmx_restrict xx,
1181 rvec * gmx_restrict ff,
1182 t_forcerec * gmx_restrict fr,
1183 t_mdatoms * gmx_restrict mdatoms,
1184 nb_kernel_data_t * gmx_restrict kernel_data,
1185 t_nrnb * gmx_restrict nrnb)
1187 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1188 * just 0 for non-waters.
1189 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
1190 * jnr indices corresponding to data put in the four positions in the SIMD register.
1192 int i_shift_offset,i_coord_offset,outeriter,inneriter;
1193 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1195 int j_coord_offsetA,j_coord_offsetB;
1196 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
1197 real rcutoff_scalar;
1198 real *shiftvec,*fshift,*x,*f;
1199 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1201 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1203 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1205 __m128d ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1206 int vdwjidx1A,vdwjidx1B;
1207 __m128d jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1208 int vdwjidx2A,vdwjidx2B;
1209 __m128d jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1210 int vdwjidx3A,vdwjidx3B;
1211 __m128d jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1212 __m128d dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1213 __m128d dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1214 __m128d dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1215 __m128d dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1216 __m128d dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1217 __m128d dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1218 __m128d dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1219 __m128d dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1220 __m128d dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1221 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
1224 __m128d ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1226 __m128d dummy_mask,cutoff_mask;
1227 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
1228 __m128d one = _mm_set1_pd(1.0);
1229 __m128d two = _mm_set1_pd(2.0);
1235 jindex = nlist->jindex;
1237 shiftidx = nlist->shift;
1239 shiftvec = fr->shift_vec[0];
1240 fshift = fr->fshift[0];
1241 facel = _mm_set1_pd(fr->epsfac);
1242 charge = mdatoms->chargeA;
1244 sh_ewald = _mm_set1_pd(fr->ic->sh_ewald);
1245 ewtab = fr->ic->tabq_coul_F;
1246 ewtabscale = _mm_set1_pd(fr->ic->tabq_scale);
1247 ewtabhalfspace = _mm_set1_pd(0.5/fr->ic->tabq_scale);
1249 /* Setup water-specific parameters */
1250 inr = nlist->iinr[0];
1251 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
1252 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
1253 iq3 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
1255 jq1 = _mm_set1_pd(charge[inr+1]);
1256 jq2 = _mm_set1_pd(charge[inr+2]);
1257 jq3 = _mm_set1_pd(charge[inr+3]);
1258 qq11 = _mm_mul_pd(iq1,jq1);
1259 qq12 = _mm_mul_pd(iq1,jq2);
1260 qq13 = _mm_mul_pd(iq1,jq3);
1261 qq21 = _mm_mul_pd(iq2,jq1);
1262 qq22 = _mm_mul_pd(iq2,jq2);
1263 qq23 = _mm_mul_pd(iq2,jq3);
1264 qq31 = _mm_mul_pd(iq3,jq1);
1265 qq32 = _mm_mul_pd(iq3,jq2);
1266 qq33 = _mm_mul_pd(iq3,jq3);
1268 /* Avoid stupid compiler warnings */
1270 j_coord_offsetA = 0;
1271 j_coord_offsetB = 0;
1276 /* Start outer loop over neighborlists */
1277 for(iidx=0; iidx<nri; iidx++)
1279 /* Load shift vector for this list */
1280 i_shift_offset = DIM*shiftidx[iidx];
1282 /* Load limits for loop over neighbors */
1283 j_index_start = jindex[iidx];
1284 j_index_end = jindex[iidx+1];
1286 /* Get outer coordinate index */
1288 i_coord_offset = DIM*inr;
1290 /* Load i particle coords and add shift vector */
1291 gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
1292 &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1294 fix1 = _mm_setzero_pd();
1295 fiy1 = _mm_setzero_pd();
1296 fiz1 = _mm_setzero_pd();
1297 fix2 = _mm_setzero_pd();
1298 fiy2 = _mm_setzero_pd();
1299 fiz2 = _mm_setzero_pd();
1300 fix3 = _mm_setzero_pd();
1301 fiy3 = _mm_setzero_pd();
1302 fiz3 = _mm_setzero_pd();
1304 /* Start inner kernel loop */
1305 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
1308 /* Get j neighbor index, and coordinate index */
1310 jnrB = jjnr[jidx+1];
1311 j_coord_offsetA = DIM*jnrA;
1312 j_coord_offsetB = DIM*jnrB;
1314 /* load j atom coordinates */
1315 gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
1316 &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
1318 /* Calculate displacement vector */
1319 dx11 = _mm_sub_pd(ix1,jx1);
1320 dy11 = _mm_sub_pd(iy1,jy1);
1321 dz11 = _mm_sub_pd(iz1,jz1);
1322 dx12 = _mm_sub_pd(ix1,jx2);
1323 dy12 = _mm_sub_pd(iy1,jy2);
1324 dz12 = _mm_sub_pd(iz1,jz2);
1325 dx13 = _mm_sub_pd(ix1,jx3);
1326 dy13 = _mm_sub_pd(iy1,jy3);
1327 dz13 = _mm_sub_pd(iz1,jz3);
1328 dx21 = _mm_sub_pd(ix2,jx1);
1329 dy21 = _mm_sub_pd(iy2,jy1);
1330 dz21 = _mm_sub_pd(iz2,jz1);
1331 dx22 = _mm_sub_pd(ix2,jx2);
1332 dy22 = _mm_sub_pd(iy2,jy2);
1333 dz22 = _mm_sub_pd(iz2,jz2);
1334 dx23 = _mm_sub_pd(ix2,jx3);
1335 dy23 = _mm_sub_pd(iy2,jy3);
1336 dz23 = _mm_sub_pd(iz2,jz3);
1337 dx31 = _mm_sub_pd(ix3,jx1);
1338 dy31 = _mm_sub_pd(iy3,jy1);
1339 dz31 = _mm_sub_pd(iz3,jz1);
1340 dx32 = _mm_sub_pd(ix3,jx2);
1341 dy32 = _mm_sub_pd(iy3,jy2);
1342 dz32 = _mm_sub_pd(iz3,jz2);
1343 dx33 = _mm_sub_pd(ix3,jx3);
1344 dy33 = _mm_sub_pd(iy3,jy3);
1345 dz33 = _mm_sub_pd(iz3,jz3);
1347 /* Calculate squared distance and things based on it */
1348 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1349 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1350 rsq13 = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
1351 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1352 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1353 rsq23 = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
1354 rsq31 = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
1355 rsq32 = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
1356 rsq33 = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
1358 rinv11 = gmx_mm_invsqrt_pd(rsq11);
1359 rinv12 = gmx_mm_invsqrt_pd(rsq12);
1360 rinv13 = gmx_mm_invsqrt_pd(rsq13);
1361 rinv21 = gmx_mm_invsqrt_pd(rsq21);
1362 rinv22 = gmx_mm_invsqrt_pd(rsq22);
1363 rinv23 = gmx_mm_invsqrt_pd(rsq23);
1364 rinv31 = gmx_mm_invsqrt_pd(rsq31);
1365 rinv32 = gmx_mm_invsqrt_pd(rsq32);
1366 rinv33 = gmx_mm_invsqrt_pd(rsq33);
1368 rinvsq11 = _mm_mul_pd(rinv11,rinv11);
1369 rinvsq12 = _mm_mul_pd(rinv12,rinv12);
1370 rinvsq13 = _mm_mul_pd(rinv13,rinv13);
1371 rinvsq21 = _mm_mul_pd(rinv21,rinv21);
1372 rinvsq22 = _mm_mul_pd(rinv22,rinv22);
1373 rinvsq23 = _mm_mul_pd(rinv23,rinv23);
1374 rinvsq31 = _mm_mul_pd(rinv31,rinv31);
1375 rinvsq32 = _mm_mul_pd(rinv32,rinv32);
1376 rinvsq33 = _mm_mul_pd(rinv33,rinv33);
1378 fjx1 = _mm_setzero_pd();
1379 fjy1 = _mm_setzero_pd();
1380 fjz1 = _mm_setzero_pd();
1381 fjx2 = _mm_setzero_pd();
1382 fjy2 = _mm_setzero_pd();
1383 fjz2 = _mm_setzero_pd();
1384 fjx3 = _mm_setzero_pd();
1385 fjy3 = _mm_setzero_pd();
1386 fjz3 = _mm_setzero_pd();
1388 /**************************
1389 * CALCULATE INTERACTIONS *
1390 **************************/
1392 r11 = _mm_mul_pd(rsq11,rinv11);
1394 /* EWALD ELECTROSTATICS */
1396 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1397 ewrt = _mm_mul_pd(r11,ewtabscale);
1398 ewitab = _mm_cvttpd_epi32(ewrt);
1399 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1400 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1402 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1403 felec = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
1407 /* Calculate temporary vectorial force */
1408 tx = _mm_mul_pd(fscal,dx11);
1409 ty = _mm_mul_pd(fscal,dy11);
1410 tz = _mm_mul_pd(fscal,dz11);
1412 /* Update vectorial force */
1413 fix1 = _mm_add_pd(fix1,tx);
1414 fiy1 = _mm_add_pd(fiy1,ty);
1415 fiz1 = _mm_add_pd(fiz1,tz);
1417 fjx1 = _mm_add_pd(fjx1,tx);
1418 fjy1 = _mm_add_pd(fjy1,ty);
1419 fjz1 = _mm_add_pd(fjz1,tz);
1421 /**************************
1422 * CALCULATE INTERACTIONS *
1423 **************************/
1425 r12 = _mm_mul_pd(rsq12,rinv12);
1427 /* EWALD ELECTROSTATICS */
1429 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1430 ewrt = _mm_mul_pd(r12,ewtabscale);
1431 ewitab = _mm_cvttpd_epi32(ewrt);
1432 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1433 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1435 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1436 felec = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
1440 /* Calculate temporary vectorial force */
1441 tx = _mm_mul_pd(fscal,dx12);
1442 ty = _mm_mul_pd(fscal,dy12);
1443 tz = _mm_mul_pd(fscal,dz12);
1445 /* Update vectorial force */
1446 fix1 = _mm_add_pd(fix1,tx);
1447 fiy1 = _mm_add_pd(fiy1,ty);
1448 fiz1 = _mm_add_pd(fiz1,tz);
1450 fjx2 = _mm_add_pd(fjx2,tx);
1451 fjy2 = _mm_add_pd(fjy2,ty);
1452 fjz2 = _mm_add_pd(fjz2,tz);
1454 /**************************
1455 * CALCULATE INTERACTIONS *
1456 **************************/
1458 r13 = _mm_mul_pd(rsq13,rinv13);
1460 /* EWALD ELECTROSTATICS */
1462 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1463 ewrt = _mm_mul_pd(r13,ewtabscale);
1464 ewitab = _mm_cvttpd_epi32(ewrt);
1465 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1466 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1468 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1469 felec = _mm_mul_pd(_mm_mul_pd(qq13,rinv13),_mm_sub_pd(rinvsq13,felec));
1473 /* Calculate temporary vectorial force */
1474 tx = _mm_mul_pd(fscal,dx13);
1475 ty = _mm_mul_pd(fscal,dy13);
1476 tz = _mm_mul_pd(fscal,dz13);
1478 /* Update vectorial force */
1479 fix1 = _mm_add_pd(fix1,tx);
1480 fiy1 = _mm_add_pd(fiy1,ty);
1481 fiz1 = _mm_add_pd(fiz1,tz);
1483 fjx3 = _mm_add_pd(fjx3,tx);
1484 fjy3 = _mm_add_pd(fjy3,ty);
1485 fjz3 = _mm_add_pd(fjz3,tz);
1487 /**************************
1488 * CALCULATE INTERACTIONS *
1489 **************************/
1491 r21 = _mm_mul_pd(rsq21,rinv21);
1493 /* EWALD ELECTROSTATICS */
1495 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1496 ewrt = _mm_mul_pd(r21,ewtabscale);
1497 ewitab = _mm_cvttpd_epi32(ewrt);
1498 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1499 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1501 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1502 felec = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
1506 /* Calculate temporary vectorial force */
1507 tx = _mm_mul_pd(fscal,dx21);
1508 ty = _mm_mul_pd(fscal,dy21);
1509 tz = _mm_mul_pd(fscal,dz21);
1511 /* Update vectorial force */
1512 fix2 = _mm_add_pd(fix2,tx);
1513 fiy2 = _mm_add_pd(fiy2,ty);
1514 fiz2 = _mm_add_pd(fiz2,tz);
1516 fjx1 = _mm_add_pd(fjx1,tx);
1517 fjy1 = _mm_add_pd(fjy1,ty);
1518 fjz1 = _mm_add_pd(fjz1,tz);
1520 /**************************
1521 * CALCULATE INTERACTIONS *
1522 **************************/
1524 r22 = _mm_mul_pd(rsq22,rinv22);
1526 /* EWALD ELECTROSTATICS */
1528 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1529 ewrt = _mm_mul_pd(r22,ewtabscale);
1530 ewitab = _mm_cvttpd_epi32(ewrt);
1531 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1532 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1534 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1535 felec = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
1539 /* Calculate temporary vectorial force */
1540 tx = _mm_mul_pd(fscal,dx22);
1541 ty = _mm_mul_pd(fscal,dy22);
1542 tz = _mm_mul_pd(fscal,dz22);
1544 /* Update vectorial force */
1545 fix2 = _mm_add_pd(fix2,tx);
1546 fiy2 = _mm_add_pd(fiy2,ty);
1547 fiz2 = _mm_add_pd(fiz2,tz);
1549 fjx2 = _mm_add_pd(fjx2,tx);
1550 fjy2 = _mm_add_pd(fjy2,ty);
1551 fjz2 = _mm_add_pd(fjz2,tz);
1553 /**************************
1554 * CALCULATE INTERACTIONS *
1555 **************************/
1557 r23 = _mm_mul_pd(rsq23,rinv23);
1559 /* EWALD ELECTROSTATICS */
1561 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1562 ewrt = _mm_mul_pd(r23,ewtabscale);
1563 ewitab = _mm_cvttpd_epi32(ewrt);
1564 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1565 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1567 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1568 felec = _mm_mul_pd(_mm_mul_pd(qq23,rinv23),_mm_sub_pd(rinvsq23,felec));
1572 /* Calculate temporary vectorial force */
1573 tx = _mm_mul_pd(fscal,dx23);
1574 ty = _mm_mul_pd(fscal,dy23);
1575 tz = _mm_mul_pd(fscal,dz23);
1577 /* Update vectorial force */
1578 fix2 = _mm_add_pd(fix2,tx);
1579 fiy2 = _mm_add_pd(fiy2,ty);
1580 fiz2 = _mm_add_pd(fiz2,tz);
1582 fjx3 = _mm_add_pd(fjx3,tx);
1583 fjy3 = _mm_add_pd(fjy3,ty);
1584 fjz3 = _mm_add_pd(fjz3,tz);
1586 /**************************
1587 * CALCULATE INTERACTIONS *
1588 **************************/
1590 r31 = _mm_mul_pd(rsq31,rinv31);
1592 /* EWALD ELECTROSTATICS */
1594 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1595 ewrt = _mm_mul_pd(r31,ewtabscale);
1596 ewitab = _mm_cvttpd_epi32(ewrt);
1597 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1598 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1600 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1601 felec = _mm_mul_pd(_mm_mul_pd(qq31,rinv31),_mm_sub_pd(rinvsq31,felec));
1605 /* Calculate temporary vectorial force */
1606 tx = _mm_mul_pd(fscal,dx31);
1607 ty = _mm_mul_pd(fscal,dy31);
1608 tz = _mm_mul_pd(fscal,dz31);
1610 /* Update vectorial force */
1611 fix3 = _mm_add_pd(fix3,tx);
1612 fiy3 = _mm_add_pd(fiy3,ty);
1613 fiz3 = _mm_add_pd(fiz3,tz);
1615 fjx1 = _mm_add_pd(fjx1,tx);
1616 fjy1 = _mm_add_pd(fjy1,ty);
1617 fjz1 = _mm_add_pd(fjz1,tz);
1619 /**************************
1620 * CALCULATE INTERACTIONS *
1621 **************************/
1623 r32 = _mm_mul_pd(rsq32,rinv32);
1625 /* EWALD ELECTROSTATICS */
1627 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1628 ewrt = _mm_mul_pd(r32,ewtabscale);
1629 ewitab = _mm_cvttpd_epi32(ewrt);
1630 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1631 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1633 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1634 felec = _mm_mul_pd(_mm_mul_pd(qq32,rinv32),_mm_sub_pd(rinvsq32,felec));
1638 /* Calculate temporary vectorial force */
1639 tx = _mm_mul_pd(fscal,dx32);
1640 ty = _mm_mul_pd(fscal,dy32);
1641 tz = _mm_mul_pd(fscal,dz32);
1643 /* Update vectorial force */
1644 fix3 = _mm_add_pd(fix3,tx);
1645 fiy3 = _mm_add_pd(fiy3,ty);
1646 fiz3 = _mm_add_pd(fiz3,tz);
1648 fjx2 = _mm_add_pd(fjx2,tx);
1649 fjy2 = _mm_add_pd(fjy2,ty);
1650 fjz2 = _mm_add_pd(fjz2,tz);
1652 /**************************
1653 * CALCULATE INTERACTIONS *
1654 **************************/
1656 r33 = _mm_mul_pd(rsq33,rinv33);
1658 /* EWALD ELECTROSTATICS */
1660 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1661 ewrt = _mm_mul_pd(r33,ewtabscale);
1662 ewitab = _mm_cvttpd_epi32(ewrt);
1663 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1664 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1666 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1667 felec = _mm_mul_pd(_mm_mul_pd(qq33,rinv33),_mm_sub_pd(rinvsq33,felec));
1671 /* Calculate temporary vectorial force */
1672 tx = _mm_mul_pd(fscal,dx33);
1673 ty = _mm_mul_pd(fscal,dy33);
1674 tz = _mm_mul_pd(fscal,dz33);
1676 /* Update vectorial force */
1677 fix3 = _mm_add_pd(fix3,tx);
1678 fiy3 = _mm_add_pd(fiy3,ty);
1679 fiz3 = _mm_add_pd(fiz3,tz);
1681 fjx3 = _mm_add_pd(fjx3,tx);
1682 fjy3 = _mm_add_pd(fjy3,ty);
1683 fjz3 = _mm_add_pd(fjz3,tz);
1685 gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA+DIM,f+j_coord_offsetB+DIM,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1687 /* Inner loop uses 324 flops */
1690 if(jidx<j_index_end)
1694 j_coord_offsetA = DIM*jnrA;
1696 /* load j atom coordinates */
1697 gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA+DIM,
1698 &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
1700 /* Calculate displacement vector */
1701 dx11 = _mm_sub_pd(ix1,jx1);
1702 dy11 = _mm_sub_pd(iy1,jy1);
1703 dz11 = _mm_sub_pd(iz1,jz1);
1704 dx12 = _mm_sub_pd(ix1,jx2);
1705 dy12 = _mm_sub_pd(iy1,jy2);
1706 dz12 = _mm_sub_pd(iz1,jz2);
1707 dx13 = _mm_sub_pd(ix1,jx3);
1708 dy13 = _mm_sub_pd(iy1,jy3);
1709 dz13 = _mm_sub_pd(iz1,jz3);
1710 dx21 = _mm_sub_pd(ix2,jx1);
1711 dy21 = _mm_sub_pd(iy2,jy1);
1712 dz21 = _mm_sub_pd(iz2,jz1);
1713 dx22 = _mm_sub_pd(ix2,jx2);
1714 dy22 = _mm_sub_pd(iy2,jy2);
1715 dz22 = _mm_sub_pd(iz2,jz2);
1716 dx23 = _mm_sub_pd(ix2,jx3);
1717 dy23 = _mm_sub_pd(iy2,jy3);
1718 dz23 = _mm_sub_pd(iz2,jz3);
1719 dx31 = _mm_sub_pd(ix3,jx1);
1720 dy31 = _mm_sub_pd(iy3,jy1);
1721 dz31 = _mm_sub_pd(iz3,jz1);
1722 dx32 = _mm_sub_pd(ix3,jx2);
1723 dy32 = _mm_sub_pd(iy3,jy2);
1724 dz32 = _mm_sub_pd(iz3,jz2);
1725 dx33 = _mm_sub_pd(ix3,jx3);
1726 dy33 = _mm_sub_pd(iy3,jy3);
1727 dz33 = _mm_sub_pd(iz3,jz3);
1729 /* Calculate squared distance and things based on it */
1730 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1731 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1732 rsq13 = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
1733 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1734 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1735 rsq23 = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
1736 rsq31 = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
1737 rsq32 = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
1738 rsq33 = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
1740 rinv11 = gmx_mm_invsqrt_pd(rsq11);
1741 rinv12 = gmx_mm_invsqrt_pd(rsq12);
1742 rinv13 = gmx_mm_invsqrt_pd(rsq13);
1743 rinv21 = gmx_mm_invsqrt_pd(rsq21);
1744 rinv22 = gmx_mm_invsqrt_pd(rsq22);
1745 rinv23 = gmx_mm_invsqrt_pd(rsq23);
1746 rinv31 = gmx_mm_invsqrt_pd(rsq31);
1747 rinv32 = gmx_mm_invsqrt_pd(rsq32);
1748 rinv33 = gmx_mm_invsqrt_pd(rsq33);
1750 rinvsq11 = _mm_mul_pd(rinv11,rinv11);
1751 rinvsq12 = _mm_mul_pd(rinv12,rinv12);
1752 rinvsq13 = _mm_mul_pd(rinv13,rinv13);
1753 rinvsq21 = _mm_mul_pd(rinv21,rinv21);
1754 rinvsq22 = _mm_mul_pd(rinv22,rinv22);
1755 rinvsq23 = _mm_mul_pd(rinv23,rinv23);
1756 rinvsq31 = _mm_mul_pd(rinv31,rinv31);
1757 rinvsq32 = _mm_mul_pd(rinv32,rinv32);
1758 rinvsq33 = _mm_mul_pd(rinv33,rinv33);
1760 fjx1 = _mm_setzero_pd();
1761 fjy1 = _mm_setzero_pd();
1762 fjz1 = _mm_setzero_pd();
1763 fjx2 = _mm_setzero_pd();
1764 fjy2 = _mm_setzero_pd();
1765 fjz2 = _mm_setzero_pd();
1766 fjx3 = _mm_setzero_pd();
1767 fjy3 = _mm_setzero_pd();
1768 fjz3 = _mm_setzero_pd();
1770 /**************************
1771 * CALCULATE INTERACTIONS *
1772 **************************/
1774 r11 = _mm_mul_pd(rsq11,rinv11);
1776 /* EWALD ELECTROSTATICS */
1778 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1779 ewrt = _mm_mul_pd(r11,ewtabscale);
1780 ewitab = _mm_cvttpd_epi32(ewrt);
1781 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1782 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1783 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1784 felec = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
1788 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1790 /* Calculate temporary vectorial force */
1791 tx = _mm_mul_pd(fscal,dx11);
1792 ty = _mm_mul_pd(fscal,dy11);
1793 tz = _mm_mul_pd(fscal,dz11);
1795 /* Update vectorial force */
1796 fix1 = _mm_add_pd(fix1,tx);
1797 fiy1 = _mm_add_pd(fiy1,ty);
1798 fiz1 = _mm_add_pd(fiz1,tz);
1800 fjx1 = _mm_add_pd(fjx1,tx);
1801 fjy1 = _mm_add_pd(fjy1,ty);
1802 fjz1 = _mm_add_pd(fjz1,tz);
1804 /**************************
1805 * CALCULATE INTERACTIONS *
1806 **************************/
1808 r12 = _mm_mul_pd(rsq12,rinv12);
1810 /* EWALD ELECTROSTATICS */
1812 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1813 ewrt = _mm_mul_pd(r12,ewtabscale);
1814 ewitab = _mm_cvttpd_epi32(ewrt);
1815 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1816 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1817 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1818 felec = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
1822 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1824 /* Calculate temporary vectorial force */
1825 tx = _mm_mul_pd(fscal,dx12);
1826 ty = _mm_mul_pd(fscal,dy12);
1827 tz = _mm_mul_pd(fscal,dz12);
1829 /* Update vectorial force */
1830 fix1 = _mm_add_pd(fix1,tx);
1831 fiy1 = _mm_add_pd(fiy1,ty);
1832 fiz1 = _mm_add_pd(fiz1,tz);
1834 fjx2 = _mm_add_pd(fjx2,tx);
1835 fjy2 = _mm_add_pd(fjy2,ty);
1836 fjz2 = _mm_add_pd(fjz2,tz);
1838 /**************************
1839 * CALCULATE INTERACTIONS *
1840 **************************/
1842 r13 = _mm_mul_pd(rsq13,rinv13);
1844 /* EWALD ELECTROSTATICS */
1846 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1847 ewrt = _mm_mul_pd(r13,ewtabscale);
1848 ewitab = _mm_cvttpd_epi32(ewrt);
1849 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1850 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1851 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1852 felec = _mm_mul_pd(_mm_mul_pd(qq13,rinv13),_mm_sub_pd(rinvsq13,felec));
1856 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1858 /* Calculate temporary vectorial force */
1859 tx = _mm_mul_pd(fscal,dx13);
1860 ty = _mm_mul_pd(fscal,dy13);
1861 tz = _mm_mul_pd(fscal,dz13);
1863 /* Update vectorial force */
1864 fix1 = _mm_add_pd(fix1,tx);
1865 fiy1 = _mm_add_pd(fiy1,ty);
1866 fiz1 = _mm_add_pd(fiz1,tz);
1868 fjx3 = _mm_add_pd(fjx3,tx);
1869 fjy3 = _mm_add_pd(fjy3,ty);
1870 fjz3 = _mm_add_pd(fjz3,tz);
1872 /**************************
1873 * CALCULATE INTERACTIONS *
1874 **************************/
1876 r21 = _mm_mul_pd(rsq21,rinv21);
1878 /* EWALD ELECTROSTATICS */
1880 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1881 ewrt = _mm_mul_pd(r21,ewtabscale);
1882 ewitab = _mm_cvttpd_epi32(ewrt);
1883 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1884 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1885 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1886 felec = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
1890 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1892 /* Calculate temporary vectorial force */
1893 tx = _mm_mul_pd(fscal,dx21);
1894 ty = _mm_mul_pd(fscal,dy21);
1895 tz = _mm_mul_pd(fscal,dz21);
1897 /* Update vectorial force */
1898 fix2 = _mm_add_pd(fix2,tx);
1899 fiy2 = _mm_add_pd(fiy2,ty);
1900 fiz2 = _mm_add_pd(fiz2,tz);
1902 fjx1 = _mm_add_pd(fjx1,tx);
1903 fjy1 = _mm_add_pd(fjy1,ty);
1904 fjz1 = _mm_add_pd(fjz1,tz);
1906 /**************************
1907 * CALCULATE INTERACTIONS *
1908 **************************/
1910 r22 = _mm_mul_pd(rsq22,rinv22);
1912 /* EWALD ELECTROSTATICS */
1914 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1915 ewrt = _mm_mul_pd(r22,ewtabscale);
1916 ewitab = _mm_cvttpd_epi32(ewrt);
1917 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1918 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1919 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1920 felec = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
1924 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1926 /* Calculate temporary vectorial force */
1927 tx = _mm_mul_pd(fscal,dx22);
1928 ty = _mm_mul_pd(fscal,dy22);
1929 tz = _mm_mul_pd(fscal,dz22);
1931 /* Update vectorial force */
1932 fix2 = _mm_add_pd(fix2,tx);
1933 fiy2 = _mm_add_pd(fiy2,ty);
1934 fiz2 = _mm_add_pd(fiz2,tz);
1936 fjx2 = _mm_add_pd(fjx2,tx);
1937 fjy2 = _mm_add_pd(fjy2,ty);
1938 fjz2 = _mm_add_pd(fjz2,tz);
1940 /**************************
1941 * CALCULATE INTERACTIONS *
1942 **************************/
1944 r23 = _mm_mul_pd(rsq23,rinv23);
1946 /* EWALD ELECTROSTATICS */
1948 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1949 ewrt = _mm_mul_pd(r23,ewtabscale);
1950 ewitab = _mm_cvttpd_epi32(ewrt);
1951 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1952 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1953 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1954 felec = _mm_mul_pd(_mm_mul_pd(qq23,rinv23),_mm_sub_pd(rinvsq23,felec));
1958 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1960 /* Calculate temporary vectorial force */
1961 tx = _mm_mul_pd(fscal,dx23);
1962 ty = _mm_mul_pd(fscal,dy23);
1963 tz = _mm_mul_pd(fscal,dz23);
1965 /* Update vectorial force */
1966 fix2 = _mm_add_pd(fix2,tx);
1967 fiy2 = _mm_add_pd(fiy2,ty);
1968 fiz2 = _mm_add_pd(fiz2,tz);
1970 fjx3 = _mm_add_pd(fjx3,tx);
1971 fjy3 = _mm_add_pd(fjy3,ty);
1972 fjz3 = _mm_add_pd(fjz3,tz);
1974 /**************************
1975 * CALCULATE INTERACTIONS *
1976 **************************/
1978 r31 = _mm_mul_pd(rsq31,rinv31);
1980 /* EWALD ELECTROSTATICS */
1982 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1983 ewrt = _mm_mul_pd(r31,ewtabscale);
1984 ewitab = _mm_cvttpd_epi32(ewrt);
1985 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1986 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1987 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1988 felec = _mm_mul_pd(_mm_mul_pd(qq31,rinv31),_mm_sub_pd(rinvsq31,felec));
1992 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1994 /* Calculate temporary vectorial force */
1995 tx = _mm_mul_pd(fscal,dx31);
1996 ty = _mm_mul_pd(fscal,dy31);
1997 tz = _mm_mul_pd(fscal,dz31);
1999 /* Update vectorial force */
2000 fix3 = _mm_add_pd(fix3,tx);
2001 fiy3 = _mm_add_pd(fiy3,ty);
2002 fiz3 = _mm_add_pd(fiz3,tz);
2004 fjx1 = _mm_add_pd(fjx1,tx);
2005 fjy1 = _mm_add_pd(fjy1,ty);
2006 fjz1 = _mm_add_pd(fjz1,tz);
2008 /**************************
2009 * CALCULATE INTERACTIONS *
2010 **************************/
2012 r32 = _mm_mul_pd(rsq32,rinv32);
2014 /* EWALD ELECTROSTATICS */
2016 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2017 ewrt = _mm_mul_pd(r32,ewtabscale);
2018 ewitab = _mm_cvttpd_epi32(ewrt);
2019 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2020 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2021 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2022 felec = _mm_mul_pd(_mm_mul_pd(qq32,rinv32),_mm_sub_pd(rinvsq32,felec));
2026 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2028 /* Calculate temporary vectorial force */
2029 tx = _mm_mul_pd(fscal,dx32);
2030 ty = _mm_mul_pd(fscal,dy32);
2031 tz = _mm_mul_pd(fscal,dz32);
2033 /* Update vectorial force */
2034 fix3 = _mm_add_pd(fix3,tx);
2035 fiy3 = _mm_add_pd(fiy3,ty);
2036 fiz3 = _mm_add_pd(fiz3,tz);
2038 fjx2 = _mm_add_pd(fjx2,tx);
2039 fjy2 = _mm_add_pd(fjy2,ty);
2040 fjz2 = _mm_add_pd(fjz2,tz);
2042 /**************************
2043 * CALCULATE INTERACTIONS *
2044 **************************/
2046 r33 = _mm_mul_pd(rsq33,rinv33);
2048 /* EWALD ELECTROSTATICS */
2050 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2051 ewrt = _mm_mul_pd(r33,ewtabscale);
2052 ewitab = _mm_cvttpd_epi32(ewrt);
2053 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2054 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2055 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2056 felec = _mm_mul_pd(_mm_mul_pd(qq33,rinv33),_mm_sub_pd(rinvsq33,felec));
2060 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2062 /* Calculate temporary vectorial force */
2063 tx = _mm_mul_pd(fscal,dx33);
2064 ty = _mm_mul_pd(fscal,dy33);
2065 tz = _mm_mul_pd(fscal,dz33);
2067 /* Update vectorial force */
2068 fix3 = _mm_add_pd(fix3,tx);
2069 fiy3 = _mm_add_pd(fiy3,ty);
2070 fiz3 = _mm_add_pd(fiz3,tz);
2072 fjx3 = _mm_add_pd(fjx3,tx);
2073 fjy3 = _mm_add_pd(fjy3,ty);
2074 fjz3 = _mm_add_pd(fjz3,tz);
2076 gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA+DIM,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2078 /* Inner loop uses 324 flops */
2081 /* End of innermost loop */
2083 gmx_mm_update_iforce_3atom_swizzle_pd(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2084 f+i_coord_offset+DIM,fshift+i_shift_offset);
2086 /* Increment number of inner iterations */
2087 inneriter += j_index_end - j_index_start;
2089 /* Outer loop uses 18 flops */
2092 /* Increment number of outer iterations */
2095 /* Update outer/inner flops */
2097 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4W4_F,outeriter*18 + inneriter*324);