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 sse4_1_single kernel generator.
42 #include "../nb_kernel.h"
43 #include "types/simple.h"
44 #include "gromacs/math/vec.h"
47 #include "gromacs/simd/math_x86_sse4_1_single.h"
48 #include "kernelutil_x86_sse4_1_single.h"
51 * Gromacs nonbonded kernel: nb_kernel_ElecEwSh_VdwNone_GeomW4W4_VF_sse4_1_single
52 * Electrostatics interaction: Ewald
53 * VdW interaction: None
54 * Geometry: Water4-Water4
55 * Calculate force/pot: PotentialAndForce
58 nb_kernel_ElecEwSh_VdwNone_GeomW4W4_VF_sse4_1_single
59 (t_nblist * gmx_restrict nlist,
60 rvec * gmx_restrict xx,
61 rvec * gmx_restrict ff,
62 t_forcerec * gmx_restrict fr,
63 t_mdatoms * gmx_restrict mdatoms,
64 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
65 t_nrnb * gmx_restrict nrnb)
67 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
68 * just 0 for non-waters.
69 * Suffixes A,B,C,D refer to j loop unrolling done with SSE, e.g. for the four 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;
74 int jnrA,jnrB,jnrC,jnrD;
75 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
76 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
77 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
79 real *shiftvec,*fshift,*x,*f;
80 real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
82 __m128 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
84 __m128 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
86 __m128 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
88 __m128 ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
89 int vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
90 __m128 jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
91 int vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
92 __m128 jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
93 int vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
94 __m128 jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
95 __m128 dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
96 __m128 dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
97 __m128 dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
98 __m128 dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
99 __m128 dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
100 __m128 dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
101 __m128 dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
102 __m128 dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
103 __m128 dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
104 __m128 velec,felec,velecsum,facel,crf,krf,krf2;
107 __m128 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
109 __m128 dummy_mask,cutoff_mask;
110 __m128 signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
111 __m128 one = _mm_set1_ps(1.0);
112 __m128 two = _mm_set1_ps(2.0);
118 jindex = nlist->jindex;
120 shiftidx = nlist->shift;
122 shiftvec = fr->shift_vec[0];
123 fshift = fr->fshift[0];
124 facel = _mm_set1_ps(fr->epsfac);
125 charge = mdatoms->chargeA;
127 sh_ewald = _mm_set1_ps(fr->ic->sh_ewald);
128 ewtab = fr->ic->tabq_coul_FDV0;
129 ewtabscale = _mm_set1_ps(fr->ic->tabq_scale);
130 ewtabhalfspace = _mm_set1_ps(0.5/fr->ic->tabq_scale);
132 /* Setup water-specific parameters */
133 inr = nlist->iinr[0];
134 iq1 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1]));
135 iq2 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2]));
136 iq3 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+3]));
138 jq1 = _mm_set1_ps(charge[inr+1]);
139 jq2 = _mm_set1_ps(charge[inr+2]);
140 jq3 = _mm_set1_ps(charge[inr+3]);
141 qq11 = _mm_mul_ps(iq1,jq1);
142 qq12 = _mm_mul_ps(iq1,jq2);
143 qq13 = _mm_mul_ps(iq1,jq3);
144 qq21 = _mm_mul_ps(iq2,jq1);
145 qq22 = _mm_mul_ps(iq2,jq2);
146 qq23 = _mm_mul_ps(iq2,jq3);
147 qq31 = _mm_mul_ps(iq3,jq1);
148 qq32 = _mm_mul_ps(iq3,jq2);
149 qq33 = _mm_mul_ps(iq3,jq3);
151 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
152 rcutoff_scalar = fr->rcoulomb;
153 rcutoff = _mm_set1_ps(rcutoff_scalar);
154 rcutoff2 = _mm_mul_ps(rcutoff,rcutoff);
156 /* Avoid stupid compiler warnings */
157 jnrA = jnrB = jnrC = jnrD = 0;
166 for(iidx=0;iidx<4*DIM;iidx++)
171 /* Start outer loop over neighborlists */
172 for(iidx=0; iidx<nri; iidx++)
174 /* Load shift vector for this list */
175 i_shift_offset = DIM*shiftidx[iidx];
177 /* Load limits for loop over neighbors */
178 j_index_start = jindex[iidx];
179 j_index_end = jindex[iidx+1];
181 /* Get outer coordinate index */
183 i_coord_offset = DIM*inr;
185 /* Load i particle coords and add shift vector */
186 gmx_mm_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
187 &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
189 fix1 = _mm_setzero_ps();
190 fiy1 = _mm_setzero_ps();
191 fiz1 = _mm_setzero_ps();
192 fix2 = _mm_setzero_ps();
193 fiy2 = _mm_setzero_ps();
194 fiz2 = _mm_setzero_ps();
195 fix3 = _mm_setzero_ps();
196 fiy3 = _mm_setzero_ps();
197 fiz3 = _mm_setzero_ps();
199 /* Reset potential sums */
200 velecsum = _mm_setzero_ps();
202 /* Start inner kernel loop */
203 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
206 /* Get j neighbor index, and coordinate index */
211 j_coord_offsetA = DIM*jnrA;
212 j_coord_offsetB = DIM*jnrB;
213 j_coord_offsetC = DIM*jnrC;
214 j_coord_offsetD = DIM*jnrD;
216 /* load j atom coordinates */
217 gmx_mm_load_3rvec_4ptr_swizzle_ps(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
218 x+j_coord_offsetC+DIM,x+j_coord_offsetD+DIM,
219 &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
221 /* Calculate displacement vector */
222 dx11 = _mm_sub_ps(ix1,jx1);
223 dy11 = _mm_sub_ps(iy1,jy1);
224 dz11 = _mm_sub_ps(iz1,jz1);
225 dx12 = _mm_sub_ps(ix1,jx2);
226 dy12 = _mm_sub_ps(iy1,jy2);
227 dz12 = _mm_sub_ps(iz1,jz2);
228 dx13 = _mm_sub_ps(ix1,jx3);
229 dy13 = _mm_sub_ps(iy1,jy3);
230 dz13 = _mm_sub_ps(iz1,jz3);
231 dx21 = _mm_sub_ps(ix2,jx1);
232 dy21 = _mm_sub_ps(iy2,jy1);
233 dz21 = _mm_sub_ps(iz2,jz1);
234 dx22 = _mm_sub_ps(ix2,jx2);
235 dy22 = _mm_sub_ps(iy2,jy2);
236 dz22 = _mm_sub_ps(iz2,jz2);
237 dx23 = _mm_sub_ps(ix2,jx3);
238 dy23 = _mm_sub_ps(iy2,jy3);
239 dz23 = _mm_sub_ps(iz2,jz3);
240 dx31 = _mm_sub_ps(ix3,jx1);
241 dy31 = _mm_sub_ps(iy3,jy1);
242 dz31 = _mm_sub_ps(iz3,jz1);
243 dx32 = _mm_sub_ps(ix3,jx2);
244 dy32 = _mm_sub_ps(iy3,jy2);
245 dz32 = _mm_sub_ps(iz3,jz2);
246 dx33 = _mm_sub_ps(ix3,jx3);
247 dy33 = _mm_sub_ps(iy3,jy3);
248 dz33 = _mm_sub_ps(iz3,jz3);
250 /* Calculate squared distance and things based on it */
251 rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
252 rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
253 rsq13 = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
254 rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
255 rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
256 rsq23 = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
257 rsq31 = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
258 rsq32 = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
259 rsq33 = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
261 rinv11 = gmx_mm_invsqrt_ps(rsq11);
262 rinv12 = gmx_mm_invsqrt_ps(rsq12);
263 rinv13 = gmx_mm_invsqrt_ps(rsq13);
264 rinv21 = gmx_mm_invsqrt_ps(rsq21);
265 rinv22 = gmx_mm_invsqrt_ps(rsq22);
266 rinv23 = gmx_mm_invsqrt_ps(rsq23);
267 rinv31 = gmx_mm_invsqrt_ps(rsq31);
268 rinv32 = gmx_mm_invsqrt_ps(rsq32);
269 rinv33 = gmx_mm_invsqrt_ps(rsq33);
271 rinvsq11 = _mm_mul_ps(rinv11,rinv11);
272 rinvsq12 = _mm_mul_ps(rinv12,rinv12);
273 rinvsq13 = _mm_mul_ps(rinv13,rinv13);
274 rinvsq21 = _mm_mul_ps(rinv21,rinv21);
275 rinvsq22 = _mm_mul_ps(rinv22,rinv22);
276 rinvsq23 = _mm_mul_ps(rinv23,rinv23);
277 rinvsq31 = _mm_mul_ps(rinv31,rinv31);
278 rinvsq32 = _mm_mul_ps(rinv32,rinv32);
279 rinvsq33 = _mm_mul_ps(rinv33,rinv33);
281 fjx1 = _mm_setzero_ps();
282 fjy1 = _mm_setzero_ps();
283 fjz1 = _mm_setzero_ps();
284 fjx2 = _mm_setzero_ps();
285 fjy2 = _mm_setzero_ps();
286 fjz2 = _mm_setzero_ps();
287 fjx3 = _mm_setzero_ps();
288 fjy3 = _mm_setzero_ps();
289 fjz3 = _mm_setzero_ps();
291 /**************************
292 * CALCULATE INTERACTIONS *
293 **************************/
295 if (gmx_mm_any_lt(rsq11,rcutoff2))
298 r11 = _mm_mul_ps(rsq11,rinv11);
300 /* EWALD ELECTROSTATICS */
302 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
303 ewrt = _mm_mul_ps(r11,ewtabscale);
304 ewitab = _mm_cvttps_epi32(ewrt);
305 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
306 ewitab = _mm_slli_epi32(ewitab,2);
307 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
308 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
309 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
310 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
311 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
312 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
313 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
314 velec = _mm_mul_ps(qq11,_mm_sub_ps(_mm_sub_ps(rinv11,sh_ewald),velec));
315 felec = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec));
317 cutoff_mask = _mm_cmplt_ps(rsq11,rcutoff2);
319 /* Update potential sum for this i atom from the interaction with this j atom. */
320 velec = _mm_and_ps(velec,cutoff_mask);
321 velecsum = _mm_add_ps(velecsum,velec);
325 fscal = _mm_and_ps(fscal,cutoff_mask);
327 /* Calculate temporary vectorial force */
328 tx = _mm_mul_ps(fscal,dx11);
329 ty = _mm_mul_ps(fscal,dy11);
330 tz = _mm_mul_ps(fscal,dz11);
332 /* Update vectorial force */
333 fix1 = _mm_add_ps(fix1,tx);
334 fiy1 = _mm_add_ps(fiy1,ty);
335 fiz1 = _mm_add_ps(fiz1,tz);
337 fjx1 = _mm_add_ps(fjx1,tx);
338 fjy1 = _mm_add_ps(fjy1,ty);
339 fjz1 = _mm_add_ps(fjz1,tz);
343 /**************************
344 * CALCULATE INTERACTIONS *
345 **************************/
347 if (gmx_mm_any_lt(rsq12,rcutoff2))
350 r12 = _mm_mul_ps(rsq12,rinv12);
352 /* EWALD ELECTROSTATICS */
354 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
355 ewrt = _mm_mul_ps(r12,ewtabscale);
356 ewitab = _mm_cvttps_epi32(ewrt);
357 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
358 ewitab = _mm_slli_epi32(ewitab,2);
359 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
360 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
361 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
362 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
363 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
364 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
365 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
366 velec = _mm_mul_ps(qq12,_mm_sub_ps(_mm_sub_ps(rinv12,sh_ewald),velec));
367 felec = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec));
369 cutoff_mask = _mm_cmplt_ps(rsq12,rcutoff2);
371 /* Update potential sum for this i atom from the interaction with this j atom. */
372 velec = _mm_and_ps(velec,cutoff_mask);
373 velecsum = _mm_add_ps(velecsum,velec);
377 fscal = _mm_and_ps(fscal,cutoff_mask);
379 /* Calculate temporary vectorial force */
380 tx = _mm_mul_ps(fscal,dx12);
381 ty = _mm_mul_ps(fscal,dy12);
382 tz = _mm_mul_ps(fscal,dz12);
384 /* Update vectorial force */
385 fix1 = _mm_add_ps(fix1,tx);
386 fiy1 = _mm_add_ps(fiy1,ty);
387 fiz1 = _mm_add_ps(fiz1,tz);
389 fjx2 = _mm_add_ps(fjx2,tx);
390 fjy2 = _mm_add_ps(fjy2,ty);
391 fjz2 = _mm_add_ps(fjz2,tz);
395 /**************************
396 * CALCULATE INTERACTIONS *
397 **************************/
399 if (gmx_mm_any_lt(rsq13,rcutoff2))
402 r13 = _mm_mul_ps(rsq13,rinv13);
404 /* EWALD ELECTROSTATICS */
406 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
407 ewrt = _mm_mul_ps(r13,ewtabscale);
408 ewitab = _mm_cvttps_epi32(ewrt);
409 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
410 ewitab = _mm_slli_epi32(ewitab,2);
411 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
412 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
413 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
414 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
415 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
416 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
417 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
418 velec = _mm_mul_ps(qq13,_mm_sub_ps(_mm_sub_ps(rinv13,sh_ewald),velec));
419 felec = _mm_mul_ps(_mm_mul_ps(qq13,rinv13),_mm_sub_ps(rinvsq13,felec));
421 cutoff_mask = _mm_cmplt_ps(rsq13,rcutoff2);
423 /* Update potential sum for this i atom from the interaction with this j atom. */
424 velec = _mm_and_ps(velec,cutoff_mask);
425 velecsum = _mm_add_ps(velecsum,velec);
429 fscal = _mm_and_ps(fscal,cutoff_mask);
431 /* Calculate temporary vectorial force */
432 tx = _mm_mul_ps(fscal,dx13);
433 ty = _mm_mul_ps(fscal,dy13);
434 tz = _mm_mul_ps(fscal,dz13);
436 /* Update vectorial force */
437 fix1 = _mm_add_ps(fix1,tx);
438 fiy1 = _mm_add_ps(fiy1,ty);
439 fiz1 = _mm_add_ps(fiz1,tz);
441 fjx3 = _mm_add_ps(fjx3,tx);
442 fjy3 = _mm_add_ps(fjy3,ty);
443 fjz3 = _mm_add_ps(fjz3,tz);
447 /**************************
448 * CALCULATE INTERACTIONS *
449 **************************/
451 if (gmx_mm_any_lt(rsq21,rcutoff2))
454 r21 = _mm_mul_ps(rsq21,rinv21);
456 /* EWALD ELECTROSTATICS */
458 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
459 ewrt = _mm_mul_ps(r21,ewtabscale);
460 ewitab = _mm_cvttps_epi32(ewrt);
461 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
462 ewitab = _mm_slli_epi32(ewitab,2);
463 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
464 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
465 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
466 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
467 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
468 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
469 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
470 velec = _mm_mul_ps(qq21,_mm_sub_ps(_mm_sub_ps(rinv21,sh_ewald),velec));
471 felec = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec));
473 cutoff_mask = _mm_cmplt_ps(rsq21,rcutoff2);
475 /* Update potential sum for this i atom from the interaction with this j atom. */
476 velec = _mm_and_ps(velec,cutoff_mask);
477 velecsum = _mm_add_ps(velecsum,velec);
481 fscal = _mm_and_ps(fscal,cutoff_mask);
483 /* Calculate temporary vectorial force */
484 tx = _mm_mul_ps(fscal,dx21);
485 ty = _mm_mul_ps(fscal,dy21);
486 tz = _mm_mul_ps(fscal,dz21);
488 /* Update vectorial force */
489 fix2 = _mm_add_ps(fix2,tx);
490 fiy2 = _mm_add_ps(fiy2,ty);
491 fiz2 = _mm_add_ps(fiz2,tz);
493 fjx1 = _mm_add_ps(fjx1,tx);
494 fjy1 = _mm_add_ps(fjy1,ty);
495 fjz1 = _mm_add_ps(fjz1,tz);
499 /**************************
500 * CALCULATE INTERACTIONS *
501 **************************/
503 if (gmx_mm_any_lt(rsq22,rcutoff2))
506 r22 = _mm_mul_ps(rsq22,rinv22);
508 /* EWALD ELECTROSTATICS */
510 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
511 ewrt = _mm_mul_ps(r22,ewtabscale);
512 ewitab = _mm_cvttps_epi32(ewrt);
513 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
514 ewitab = _mm_slli_epi32(ewitab,2);
515 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
516 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
517 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
518 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
519 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
520 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
521 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
522 velec = _mm_mul_ps(qq22,_mm_sub_ps(_mm_sub_ps(rinv22,sh_ewald),velec));
523 felec = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec));
525 cutoff_mask = _mm_cmplt_ps(rsq22,rcutoff2);
527 /* Update potential sum for this i atom from the interaction with this j atom. */
528 velec = _mm_and_ps(velec,cutoff_mask);
529 velecsum = _mm_add_ps(velecsum,velec);
533 fscal = _mm_and_ps(fscal,cutoff_mask);
535 /* Calculate temporary vectorial force */
536 tx = _mm_mul_ps(fscal,dx22);
537 ty = _mm_mul_ps(fscal,dy22);
538 tz = _mm_mul_ps(fscal,dz22);
540 /* Update vectorial force */
541 fix2 = _mm_add_ps(fix2,tx);
542 fiy2 = _mm_add_ps(fiy2,ty);
543 fiz2 = _mm_add_ps(fiz2,tz);
545 fjx2 = _mm_add_ps(fjx2,tx);
546 fjy2 = _mm_add_ps(fjy2,ty);
547 fjz2 = _mm_add_ps(fjz2,tz);
551 /**************************
552 * CALCULATE INTERACTIONS *
553 **************************/
555 if (gmx_mm_any_lt(rsq23,rcutoff2))
558 r23 = _mm_mul_ps(rsq23,rinv23);
560 /* EWALD ELECTROSTATICS */
562 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
563 ewrt = _mm_mul_ps(r23,ewtabscale);
564 ewitab = _mm_cvttps_epi32(ewrt);
565 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
566 ewitab = _mm_slli_epi32(ewitab,2);
567 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
568 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
569 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
570 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
571 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
572 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
573 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
574 velec = _mm_mul_ps(qq23,_mm_sub_ps(_mm_sub_ps(rinv23,sh_ewald),velec));
575 felec = _mm_mul_ps(_mm_mul_ps(qq23,rinv23),_mm_sub_ps(rinvsq23,felec));
577 cutoff_mask = _mm_cmplt_ps(rsq23,rcutoff2);
579 /* Update potential sum for this i atom from the interaction with this j atom. */
580 velec = _mm_and_ps(velec,cutoff_mask);
581 velecsum = _mm_add_ps(velecsum,velec);
585 fscal = _mm_and_ps(fscal,cutoff_mask);
587 /* Calculate temporary vectorial force */
588 tx = _mm_mul_ps(fscal,dx23);
589 ty = _mm_mul_ps(fscal,dy23);
590 tz = _mm_mul_ps(fscal,dz23);
592 /* Update vectorial force */
593 fix2 = _mm_add_ps(fix2,tx);
594 fiy2 = _mm_add_ps(fiy2,ty);
595 fiz2 = _mm_add_ps(fiz2,tz);
597 fjx3 = _mm_add_ps(fjx3,tx);
598 fjy3 = _mm_add_ps(fjy3,ty);
599 fjz3 = _mm_add_ps(fjz3,tz);
603 /**************************
604 * CALCULATE INTERACTIONS *
605 **************************/
607 if (gmx_mm_any_lt(rsq31,rcutoff2))
610 r31 = _mm_mul_ps(rsq31,rinv31);
612 /* EWALD ELECTROSTATICS */
614 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
615 ewrt = _mm_mul_ps(r31,ewtabscale);
616 ewitab = _mm_cvttps_epi32(ewrt);
617 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
618 ewitab = _mm_slli_epi32(ewitab,2);
619 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
620 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
621 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
622 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
623 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
624 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
625 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
626 velec = _mm_mul_ps(qq31,_mm_sub_ps(_mm_sub_ps(rinv31,sh_ewald),velec));
627 felec = _mm_mul_ps(_mm_mul_ps(qq31,rinv31),_mm_sub_ps(rinvsq31,felec));
629 cutoff_mask = _mm_cmplt_ps(rsq31,rcutoff2);
631 /* Update potential sum for this i atom from the interaction with this j atom. */
632 velec = _mm_and_ps(velec,cutoff_mask);
633 velecsum = _mm_add_ps(velecsum,velec);
637 fscal = _mm_and_ps(fscal,cutoff_mask);
639 /* Calculate temporary vectorial force */
640 tx = _mm_mul_ps(fscal,dx31);
641 ty = _mm_mul_ps(fscal,dy31);
642 tz = _mm_mul_ps(fscal,dz31);
644 /* Update vectorial force */
645 fix3 = _mm_add_ps(fix3,tx);
646 fiy3 = _mm_add_ps(fiy3,ty);
647 fiz3 = _mm_add_ps(fiz3,tz);
649 fjx1 = _mm_add_ps(fjx1,tx);
650 fjy1 = _mm_add_ps(fjy1,ty);
651 fjz1 = _mm_add_ps(fjz1,tz);
655 /**************************
656 * CALCULATE INTERACTIONS *
657 **************************/
659 if (gmx_mm_any_lt(rsq32,rcutoff2))
662 r32 = _mm_mul_ps(rsq32,rinv32);
664 /* EWALD ELECTROSTATICS */
666 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
667 ewrt = _mm_mul_ps(r32,ewtabscale);
668 ewitab = _mm_cvttps_epi32(ewrt);
669 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
670 ewitab = _mm_slli_epi32(ewitab,2);
671 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
672 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
673 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
674 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
675 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
676 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
677 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
678 velec = _mm_mul_ps(qq32,_mm_sub_ps(_mm_sub_ps(rinv32,sh_ewald),velec));
679 felec = _mm_mul_ps(_mm_mul_ps(qq32,rinv32),_mm_sub_ps(rinvsq32,felec));
681 cutoff_mask = _mm_cmplt_ps(rsq32,rcutoff2);
683 /* Update potential sum for this i atom from the interaction with this j atom. */
684 velec = _mm_and_ps(velec,cutoff_mask);
685 velecsum = _mm_add_ps(velecsum,velec);
689 fscal = _mm_and_ps(fscal,cutoff_mask);
691 /* Calculate temporary vectorial force */
692 tx = _mm_mul_ps(fscal,dx32);
693 ty = _mm_mul_ps(fscal,dy32);
694 tz = _mm_mul_ps(fscal,dz32);
696 /* Update vectorial force */
697 fix3 = _mm_add_ps(fix3,tx);
698 fiy3 = _mm_add_ps(fiy3,ty);
699 fiz3 = _mm_add_ps(fiz3,tz);
701 fjx2 = _mm_add_ps(fjx2,tx);
702 fjy2 = _mm_add_ps(fjy2,ty);
703 fjz2 = _mm_add_ps(fjz2,tz);
707 /**************************
708 * CALCULATE INTERACTIONS *
709 **************************/
711 if (gmx_mm_any_lt(rsq33,rcutoff2))
714 r33 = _mm_mul_ps(rsq33,rinv33);
716 /* EWALD ELECTROSTATICS */
718 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
719 ewrt = _mm_mul_ps(r33,ewtabscale);
720 ewitab = _mm_cvttps_epi32(ewrt);
721 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
722 ewitab = _mm_slli_epi32(ewitab,2);
723 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
724 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
725 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
726 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
727 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
728 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
729 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
730 velec = _mm_mul_ps(qq33,_mm_sub_ps(_mm_sub_ps(rinv33,sh_ewald),velec));
731 felec = _mm_mul_ps(_mm_mul_ps(qq33,rinv33),_mm_sub_ps(rinvsq33,felec));
733 cutoff_mask = _mm_cmplt_ps(rsq33,rcutoff2);
735 /* Update potential sum for this i atom from the interaction with this j atom. */
736 velec = _mm_and_ps(velec,cutoff_mask);
737 velecsum = _mm_add_ps(velecsum,velec);
741 fscal = _mm_and_ps(fscal,cutoff_mask);
743 /* Calculate temporary vectorial force */
744 tx = _mm_mul_ps(fscal,dx33);
745 ty = _mm_mul_ps(fscal,dy33);
746 tz = _mm_mul_ps(fscal,dz33);
748 /* Update vectorial force */
749 fix3 = _mm_add_ps(fix3,tx);
750 fiy3 = _mm_add_ps(fiy3,ty);
751 fiz3 = _mm_add_ps(fiz3,tz);
753 fjx3 = _mm_add_ps(fjx3,tx);
754 fjy3 = _mm_add_ps(fjy3,ty);
755 fjz3 = _mm_add_ps(fjz3,tz);
759 fjptrA = f+j_coord_offsetA;
760 fjptrB = f+j_coord_offsetB;
761 fjptrC = f+j_coord_offsetC;
762 fjptrD = f+j_coord_offsetD;
764 gmx_mm_decrement_3rvec_4ptr_swizzle_ps(fjptrA+DIM,fjptrB+DIM,fjptrC+DIM,fjptrD+DIM,
765 fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
767 /* Inner loop uses 414 flops */
773 /* Get j neighbor index, and coordinate index */
774 jnrlistA = jjnr[jidx];
775 jnrlistB = jjnr[jidx+1];
776 jnrlistC = jjnr[jidx+2];
777 jnrlistD = jjnr[jidx+3];
778 /* Sign of each element will be negative for non-real atoms.
779 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
780 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
782 dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
783 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
784 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
785 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
786 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
787 j_coord_offsetA = DIM*jnrA;
788 j_coord_offsetB = DIM*jnrB;
789 j_coord_offsetC = DIM*jnrC;
790 j_coord_offsetD = DIM*jnrD;
792 /* load j atom coordinates */
793 gmx_mm_load_3rvec_4ptr_swizzle_ps(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
794 x+j_coord_offsetC+DIM,x+j_coord_offsetD+DIM,
795 &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
797 /* Calculate displacement vector */
798 dx11 = _mm_sub_ps(ix1,jx1);
799 dy11 = _mm_sub_ps(iy1,jy1);
800 dz11 = _mm_sub_ps(iz1,jz1);
801 dx12 = _mm_sub_ps(ix1,jx2);
802 dy12 = _mm_sub_ps(iy1,jy2);
803 dz12 = _mm_sub_ps(iz1,jz2);
804 dx13 = _mm_sub_ps(ix1,jx3);
805 dy13 = _mm_sub_ps(iy1,jy3);
806 dz13 = _mm_sub_ps(iz1,jz3);
807 dx21 = _mm_sub_ps(ix2,jx1);
808 dy21 = _mm_sub_ps(iy2,jy1);
809 dz21 = _mm_sub_ps(iz2,jz1);
810 dx22 = _mm_sub_ps(ix2,jx2);
811 dy22 = _mm_sub_ps(iy2,jy2);
812 dz22 = _mm_sub_ps(iz2,jz2);
813 dx23 = _mm_sub_ps(ix2,jx3);
814 dy23 = _mm_sub_ps(iy2,jy3);
815 dz23 = _mm_sub_ps(iz2,jz3);
816 dx31 = _mm_sub_ps(ix3,jx1);
817 dy31 = _mm_sub_ps(iy3,jy1);
818 dz31 = _mm_sub_ps(iz3,jz1);
819 dx32 = _mm_sub_ps(ix3,jx2);
820 dy32 = _mm_sub_ps(iy3,jy2);
821 dz32 = _mm_sub_ps(iz3,jz2);
822 dx33 = _mm_sub_ps(ix3,jx3);
823 dy33 = _mm_sub_ps(iy3,jy3);
824 dz33 = _mm_sub_ps(iz3,jz3);
826 /* Calculate squared distance and things based on it */
827 rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
828 rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
829 rsq13 = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
830 rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
831 rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
832 rsq23 = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
833 rsq31 = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
834 rsq32 = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
835 rsq33 = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
837 rinv11 = gmx_mm_invsqrt_ps(rsq11);
838 rinv12 = gmx_mm_invsqrt_ps(rsq12);
839 rinv13 = gmx_mm_invsqrt_ps(rsq13);
840 rinv21 = gmx_mm_invsqrt_ps(rsq21);
841 rinv22 = gmx_mm_invsqrt_ps(rsq22);
842 rinv23 = gmx_mm_invsqrt_ps(rsq23);
843 rinv31 = gmx_mm_invsqrt_ps(rsq31);
844 rinv32 = gmx_mm_invsqrt_ps(rsq32);
845 rinv33 = gmx_mm_invsqrt_ps(rsq33);
847 rinvsq11 = _mm_mul_ps(rinv11,rinv11);
848 rinvsq12 = _mm_mul_ps(rinv12,rinv12);
849 rinvsq13 = _mm_mul_ps(rinv13,rinv13);
850 rinvsq21 = _mm_mul_ps(rinv21,rinv21);
851 rinvsq22 = _mm_mul_ps(rinv22,rinv22);
852 rinvsq23 = _mm_mul_ps(rinv23,rinv23);
853 rinvsq31 = _mm_mul_ps(rinv31,rinv31);
854 rinvsq32 = _mm_mul_ps(rinv32,rinv32);
855 rinvsq33 = _mm_mul_ps(rinv33,rinv33);
857 fjx1 = _mm_setzero_ps();
858 fjy1 = _mm_setzero_ps();
859 fjz1 = _mm_setzero_ps();
860 fjx2 = _mm_setzero_ps();
861 fjy2 = _mm_setzero_ps();
862 fjz2 = _mm_setzero_ps();
863 fjx3 = _mm_setzero_ps();
864 fjy3 = _mm_setzero_ps();
865 fjz3 = _mm_setzero_ps();
867 /**************************
868 * CALCULATE INTERACTIONS *
869 **************************/
871 if (gmx_mm_any_lt(rsq11,rcutoff2))
874 r11 = _mm_mul_ps(rsq11,rinv11);
875 r11 = _mm_andnot_ps(dummy_mask,r11);
877 /* EWALD ELECTROSTATICS */
879 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
880 ewrt = _mm_mul_ps(r11,ewtabscale);
881 ewitab = _mm_cvttps_epi32(ewrt);
882 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
883 ewitab = _mm_slli_epi32(ewitab,2);
884 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
885 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
886 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
887 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
888 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
889 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
890 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
891 velec = _mm_mul_ps(qq11,_mm_sub_ps(_mm_sub_ps(rinv11,sh_ewald),velec));
892 felec = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec));
894 cutoff_mask = _mm_cmplt_ps(rsq11,rcutoff2);
896 /* Update potential sum for this i atom from the interaction with this j atom. */
897 velec = _mm_and_ps(velec,cutoff_mask);
898 velec = _mm_andnot_ps(dummy_mask,velec);
899 velecsum = _mm_add_ps(velecsum,velec);
903 fscal = _mm_and_ps(fscal,cutoff_mask);
905 fscal = _mm_andnot_ps(dummy_mask,fscal);
907 /* Calculate temporary vectorial force */
908 tx = _mm_mul_ps(fscal,dx11);
909 ty = _mm_mul_ps(fscal,dy11);
910 tz = _mm_mul_ps(fscal,dz11);
912 /* Update vectorial force */
913 fix1 = _mm_add_ps(fix1,tx);
914 fiy1 = _mm_add_ps(fiy1,ty);
915 fiz1 = _mm_add_ps(fiz1,tz);
917 fjx1 = _mm_add_ps(fjx1,tx);
918 fjy1 = _mm_add_ps(fjy1,ty);
919 fjz1 = _mm_add_ps(fjz1,tz);
923 /**************************
924 * CALCULATE INTERACTIONS *
925 **************************/
927 if (gmx_mm_any_lt(rsq12,rcutoff2))
930 r12 = _mm_mul_ps(rsq12,rinv12);
931 r12 = _mm_andnot_ps(dummy_mask,r12);
933 /* EWALD ELECTROSTATICS */
935 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
936 ewrt = _mm_mul_ps(r12,ewtabscale);
937 ewitab = _mm_cvttps_epi32(ewrt);
938 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
939 ewitab = _mm_slli_epi32(ewitab,2);
940 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
941 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
942 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
943 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
944 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
945 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
946 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
947 velec = _mm_mul_ps(qq12,_mm_sub_ps(_mm_sub_ps(rinv12,sh_ewald),velec));
948 felec = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec));
950 cutoff_mask = _mm_cmplt_ps(rsq12,rcutoff2);
952 /* Update potential sum for this i atom from the interaction with this j atom. */
953 velec = _mm_and_ps(velec,cutoff_mask);
954 velec = _mm_andnot_ps(dummy_mask,velec);
955 velecsum = _mm_add_ps(velecsum,velec);
959 fscal = _mm_and_ps(fscal,cutoff_mask);
961 fscal = _mm_andnot_ps(dummy_mask,fscal);
963 /* Calculate temporary vectorial force */
964 tx = _mm_mul_ps(fscal,dx12);
965 ty = _mm_mul_ps(fscal,dy12);
966 tz = _mm_mul_ps(fscal,dz12);
968 /* Update vectorial force */
969 fix1 = _mm_add_ps(fix1,tx);
970 fiy1 = _mm_add_ps(fiy1,ty);
971 fiz1 = _mm_add_ps(fiz1,tz);
973 fjx2 = _mm_add_ps(fjx2,tx);
974 fjy2 = _mm_add_ps(fjy2,ty);
975 fjz2 = _mm_add_ps(fjz2,tz);
979 /**************************
980 * CALCULATE INTERACTIONS *
981 **************************/
983 if (gmx_mm_any_lt(rsq13,rcutoff2))
986 r13 = _mm_mul_ps(rsq13,rinv13);
987 r13 = _mm_andnot_ps(dummy_mask,r13);
989 /* EWALD ELECTROSTATICS */
991 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
992 ewrt = _mm_mul_ps(r13,ewtabscale);
993 ewitab = _mm_cvttps_epi32(ewrt);
994 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
995 ewitab = _mm_slli_epi32(ewitab,2);
996 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
997 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
998 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
999 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1000 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1001 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1002 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1003 velec = _mm_mul_ps(qq13,_mm_sub_ps(_mm_sub_ps(rinv13,sh_ewald),velec));
1004 felec = _mm_mul_ps(_mm_mul_ps(qq13,rinv13),_mm_sub_ps(rinvsq13,felec));
1006 cutoff_mask = _mm_cmplt_ps(rsq13,rcutoff2);
1008 /* Update potential sum for this i atom from the interaction with this j atom. */
1009 velec = _mm_and_ps(velec,cutoff_mask);
1010 velec = _mm_andnot_ps(dummy_mask,velec);
1011 velecsum = _mm_add_ps(velecsum,velec);
1015 fscal = _mm_and_ps(fscal,cutoff_mask);
1017 fscal = _mm_andnot_ps(dummy_mask,fscal);
1019 /* Calculate temporary vectorial force */
1020 tx = _mm_mul_ps(fscal,dx13);
1021 ty = _mm_mul_ps(fscal,dy13);
1022 tz = _mm_mul_ps(fscal,dz13);
1024 /* Update vectorial force */
1025 fix1 = _mm_add_ps(fix1,tx);
1026 fiy1 = _mm_add_ps(fiy1,ty);
1027 fiz1 = _mm_add_ps(fiz1,tz);
1029 fjx3 = _mm_add_ps(fjx3,tx);
1030 fjy3 = _mm_add_ps(fjy3,ty);
1031 fjz3 = _mm_add_ps(fjz3,tz);
1035 /**************************
1036 * CALCULATE INTERACTIONS *
1037 **************************/
1039 if (gmx_mm_any_lt(rsq21,rcutoff2))
1042 r21 = _mm_mul_ps(rsq21,rinv21);
1043 r21 = _mm_andnot_ps(dummy_mask,r21);
1045 /* EWALD ELECTROSTATICS */
1047 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1048 ewrt = _mm_mul_ps(r21,ewtabscale);
1049 ewitab = _mm_cvttps_epi32(ewrt);
1050 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1051 ewitab = _mm_slli_epi32(ewitab,2);
1052 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1053 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1054 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1055 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1056 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1057 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1058 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1059 velec = _mm_mul_ps(qq21,_mm_sub_ps(_mm_sub_ps(rinv21,sh_ewald),velec));
1060 felec = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec));
1062 cutoff_mask = _mm_cmplt_ps(rsq21,rcutoff2);
1064 /* Update potential sum for this i atom from the interaction with this j atom. */
1065 velec = _mm_and_ps(velec,cutoff_mask);
1066 velec = _mm_andnot_ps(dummy_mask,velec);
1067 velecsum = _mm_add_ps(velecsum,velec);
1071 fscal = _mm_and_ps(fscal,cutoff_mask);
1073 fscal = _mm_andnot_ps(dummy_mask,fscal);
1075 /* Calculate temporary vectorial force */
1076 tx = _mm_mul_ps(fscal,dx21);
1077 ty = _mm_mul_ps(fscal,dy21);
1078 tz = _mm_mul_ps(fscal,dz21);
1080 /* Update vectorial force */
1081 fix2 = _mm_add_ps(fix2,tx);
1082 fiy2 = _mm_add_ps(fiy2,ty);
1083 fiz2 = _mm_add_ps(fiz2,tz);
1085 fjx1 = _mm_add_ps(fjx1,tx);
1086 fjy1 = _mm_add_ps(fjy1,ty);
1087 fjz1 = _mm_add_ps(fjz1,tz);
1091 /**************************
1092 * CALCULATE INTERACTIONS *
1093 **************************/
1095 if (gmx_mm_any_lt(rsq22,rcutoff2))
1098 r22 = _mm_mul_ps(rsq22,rinv22);
1099 r22 = _mm_andnot_ps(dummy_mask,r22);
1101 /* EWALD ELECTROSTATICS */
1103 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1104 ewrt = _mm_mul_ps(r22,ewtabscale);
1105 ewitab = _mm_cvttps_epi32(ewrt);
1106 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1107 ewitab = _mm_slli_epi32(ewitab,2);
1108 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1109 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1110 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1111 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1112 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1113 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1114 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1115 velec = _mm_mul_ps(qq22,_mm_sub_ps(_mm_sub_ps(rinv22,sh_ewald),velec));
1116 felec = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec));
1118 cutoff_mask = _mm_cmplt_ps(rsq22,rcutoff2);
1120 /* Update potential sum for this i atom from the interaction with this j atom. */
1121 velec = _mm_and_ps(velec,cutoff_mask);
1122 velec = _mm_andnot_ps(dummy_mask,velec);
1123 velecsum = _mm_add_ps(velecsum,velec);
1127 fscal = _mm_and_ps(fscal,cutoff_mask);
1129 fscal = _mm_andnot_ps(dummy_mask,fscal);
1131 /* Calculate temporary vectorial force */
1132 tx = _mm_mul_ps(fscal,dx22);
1133 ty = _mm_mul_ps(fscal,dy22);
1134 tz = _mm_mul_ps(fscal,dz22);
1136 /* Update vectorial force */
1137 fix2 = _mm_add_ps(fix2,tx);
1138 fiy2 = _mm_add_ps(fiy2,ty);
1139 fiz2 = _mm_add_ps(fiz2,tz);
1141 fjx2 = _mm_add_ps(fjx2,tx);
1142 fjy2 = _mm_add_ps(fjy2,ty);
1143 fjz2 = _mm_add_ps(fjz2,tz);
1147 /**************************
1148 * CALCULATE INTERACTIONS *
1149 **************************/
1151 if (gmx_mm_any_lt(rsq23,rcutoff2))
1154 r23 = _mm_mul_ps(rsq23,rinv23);
1155 r23 = _mm_andnot_ps(dummy_mask,r23);
1157 /* EWALD ELECTROSTATICS */
1159 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1160 ewrt = _mm_mul_ps(r23,ewtabscale);
1161 ewitab = _mm_cvttps_epi32(ewrt);
1162 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1163 ewitab = _mm_slli_epi32(ewitab,2);
1164 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1165 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1166 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1167 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1168 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1169 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1170 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1171 velec = _mm_mul_ps(qq23,_mm_sub_ps(_mm_sub_ps(rinv23,sh_ewald),velec));
1172 felec = _mm_mul_ps(_mm_mul_ps(qq23,rinv23),_mm_sub_ps(rinvsq23,felec));
1174 cutoff_mask = _mm_cmplt_ps(rsq23,rcutoff2);
1176 /* Update potential sum for this i atom from the interaction with this j atom. */
1177 velec = _mm_and_ps(velec,cutoff_mask);
1178 velec = _mm_andnot_ps(dummy_mask,velec);
1179 velecsum = _mm_add_ps(velecsum,velec);
1183 fscal = _mm_and_ps(fscal,cutoff_mask);
1185 fscal = _mm_andnot_ps(dummy_mask,fscal);
1187 /* Calculate temporary vectorial force */
1188 tx = _mm_mul_ps(fscal,dx23);
1189 ty = _mm_mul_ps(fscal,dy23);
1190 tz = _mm_mul_ps(fscal,dz23);
1192 /* Update vectorial force */
1193 fix2 = _mm_add_ps(fix2,tx);
1194 fiy2 = _mm_add_ps(fiy2,ty);
1195 fiz2 = _mm_add_ps(fiz2,tz);
1197 fjx3 = _mm_add_ps(fjx3,tx);
1198 fjy3 = _mm_add_ps(fjy3,ty);
1199 fjz3 = _mm_add_ps(fjz3,tz);
1203 /**************************
1204 * CALCULATE INTERACTIONS *
1205 **************************/
1207 if (gmx_mm_any_lt(rsq31,rcutoff2))
1210 r31 = _mm_mul_ps(rsq31,rinv31);
1211 r31 = _mm_andnot_ps(dummy_mask,r31);
1213 /* EWALD ELECTROSTATICS */
1215 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1216 ewrt = _mm_mul_ps(r31,ewtabscale);
1217 ewitab = _mm_cvttps_epi32(ewrt);
1218 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1219 ewitab = _mm_slli_epi32(ewitab,2);
1220 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1221 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1222 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1223 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1224 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1225 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1226 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1227 velec = _mm_mul_ps(qq31,_mm_sub_ps(_mm_sub_ps(rinv31,sh_ewald),velec));
1228 felec = _mm_mul_ps(_mm_mul_ps(qq31,rinv31),_mm_sub_ps(rinvsq31,felec));
1230 cutoff_mask = _mm_cmplt_ps(rsq31,rcutoff2);
1232 /* Update potential sum for this i atom from the interaction with this j atom. */
1233 velec = _mm_and_ps(velec,cutoff_mask);
1234 velec = _mm_andnot_ps(dummy_mask,velec);
1235 velecsum = _mm_add_ps(velecsum,velec);
1239 fscal = _mm_and_ps(fscal,cutoff_mask);
1241 fscal = _mm_andnot_ps(dummy_mask,fscal);
1243 /* Calculate temporary vectorial force */
1244 tx = _mm_mul_ps(fscal,dx31);
1245 ty = _mm_mul_ps(fscal,dy31);
1246 tz = _mm_mul_ps(fscal,dz31);
1248 /* Update vectorial force */
1249 fix3 = _mm_add_ps(fix3,tx);
1250 fiy3 = _mm_add_ps(fiy3,ty);
1251 fiz3 = _mm_add_ps(fiz3,tz);
1253 fjx1 = _mm_add_ps(fjx1,tx);
1254 fjy1 = _mm_add_ps(fjy1,ty);
1255 fjz1 = _mm_add_ps(fjz1,tz);
1259 /**************************
1260 * CALCULATE INTERACTIONS *
1261 **************************/
1263 if (gmx_mm_any_lt(rsq32,rcutoff2))
1266 r32 = _mm_mul_ps(rsq32,rinv32);
1267 r32 = _mm_andnot_ps(dummy_mask,r32);
1269 /* EWALD ELECTROSTATICS */
1271 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1272 ewrt = _mm_mul_ps(r32,ewtabscale);
1273 ewitab = _mm_cvttps_epi32(ewrt);
1274 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1275 ewitab = _mm_slli_epi32(ewitab,2);
1276 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1277 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1278 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1279 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1280 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1281 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1282 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1283 velec = _mm_mul_ps(qq32,_mm_sub_ps(_mm_sub_ps(rinv32,sh_ewald),velec));
1284 felec = _mm_mul_ps(_mm_mul_ps(qq32,rinv32),_mm_sub_ps(rinvsq32,felec));
1286 cutoff_mask = _mm_cmplt_ps(rsq32,rcutoff2);
1288 /* Update potential sum for this i atom from the interaction with this j atom. */
1289 velec = _mm_and_ps(velec,cutoff_mask);
1290 velec = _mm_andnot_ps(dummy_mask,velec);
1291 velecsum = _mm_add_ps(velecsum,velec);
1295 fscal = _mm_and_ps(fscal,cutoff_mask);
1297 fscal = _mm_andnot_ps(dummy_mask,fscal);
1299 /* Calculate temporary vectorial force */
1300 tx = _mm_mul_ps(fscal,dx32);
1301 ty = _mm_mul_ps(fscal,dy32);
1302 tz = _mm_mul_ps(fscal,dz32);
1304 /* Update vectorial force */
1305 fix3 = _mm_add_ps(fix3,tx);
1306 fiy3 = _mm_add_ps(fiy3,ty);
1307 fiz3 = _mm_add_ps(fiz3,tz);
1309 fjx2 = _mm_add_ps(fjx2,tx);
1310 fjy2 = _mm_add_ps(fjy2,ty);
1311 fjz2 = _mm_add_ps(fjz2,tz);
1315 /**************************
1316 * CALCULATE INTERACTIONS *
1317 **************************/
1319 if (gmx_mm_any_lt(rsq33,rcutoff2))
1322 r33 = _mm_mul_ps(rsq33,rinv33);
1323 r33 = _mm_andnot_ps(dummy_mask,r33);
1325 /* EWALD ELECTROSTATICS */
1327 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1328 ewrt = _mm_mul_ps(r33,ewtabscale);
1329 ewitab = _mm_cvttps_epi32(ewrt);
1330 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1331 ewitab = _mm_slli_epi32(ewitab,2);
1332 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1333 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1334 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1335 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1336 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1337 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1338 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1339 velec = _mm_mul_ps(qq33,_mm_sub_ps(_mm_sub_ps(rinv33,sh_ewald),velec));
1340 felec = _mm_mul_ps(_mm_mul_ps(qq33,rinv33),_mm_sub_ps(rinvsq33,felec));
1342 cutoff_mask = _mm_cmplt_ps(rsq33,rcutoff2);
1344 /* Update potential sum for this i atom from the interaction with this j atom. */
1345 velec = _mm_and_ps(velec,cutoff_mask);
1346 velec = _mm_andnot_ps(dummy_mask,velec);
1347 velecsum = _mm_add_ps(velecsum,velec);
1351 fscal = _mm_and_ps(fscal,cutoff_mask);
1353 fscal = _mm_andnot_ps(dummy_mask,fscal);
1355 /* Calculate temporary vectorial force */
1356 tx = _mm_mul_ps(fscal,dx33);
1357 ty = _mm_mul_ps(fscal,dy33);
1358 tz = _mm_mul_ps(fscal,dz33);
1360 /* Update vectorial force */
1361 fix3 = _mm_add_ps(fix3,tx);
1362 fiy3 = _mm_add_ps(fiy3,ty);
1363 fiz3 = _mm_add_ps(fiz3,tz);
1365 fjx3 = _mm_add_ps(fjx3,tx);
1366 fjy3 = _mm_add_ps(fjy3,ty);
1367 fjz3 = _mm_add_ps(fjz3,tz);
1371 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1372 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1373 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1374 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1376 gmx_mm_decrement_3rvec_4ptr_swizzle_ps(fjptrA+DIM,fjptrB+DIM,fjptrC+DIM,fjptrD+DIM,
1377 fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1379 /* Inner loop uses 423 flops */
1382 /* End of innermost loop */
1384 gmx_mm_update_iforce_3atom_swizzle_ps(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1385 f+i_coord_offset+DIM,fshift+i_shift_offset);
1388 /* Update potential energies */
1389 gmx_mm_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
1391 /* Increment number of inner iterations */
1392 inneriter += j_index_end - j_index_start;
1394 /* Outer loop uses 19 flops */
1397 /* Increment number of outer iterations */
1400 /* Update outer/inner flops */
1402 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4W4_VF,outeriter*19 + inneriter*423);
1405 * Gromacs nonbonded kernel: nb_kernel_ElecEwSh_VdwNone_GeomW4W4_F_sse4_1_single
1406 * Electrostatics interaction: Ewald
1407 * VdW interaction: None
1408 * Geometry: Water4-Water4
1409 * Calculate force/pot: Force
1412 nb_kernel_ElecEwSh_VdwNone_GeomW4W4_F_sse4_1_single
1413 (t_nblist * gmx_restrict nlist,
1414 rvec * gmx_restrict xx,
1415 rvec * gmx_restrict ff,
1416 t_forcerec * gmx_restrict fr,
1417 t_mdatoms * gmx_restrict mdatoms,
1418 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1419 t_nrnb * gmx_restrict nrnb)
1421 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1422 * just 0 for non-waters.
1423 * Suffixes A,B,C,D refer to j loop unrolling done with SSE, e.g. for the four different
1424 * jnr indices corresponding to data put in the four positions in the SIMD register.
1426 int i_shift_offset,i_coord_offset,outeriter,inneriter;
1427 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1428 int jnrA,jnrB,jnrC,jnrD;
1429 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1430 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1431 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
1432 real rcutoff_scalar;
1433 real *shiftvec,*fshift,*x,*f;
1434 real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
1435 real scratch[4*DIM];
1436 __m128 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1438 __m128 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1440 __m128 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1442 __m128 ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1443 int vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
1444 __m128 jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1445 int vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
1446 __m128 jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1447 int vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
1448 __m128 jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1449 __m128 dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1450 __m128 dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1451 __m128 dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1452 __m128 dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1453 __m128 dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1454 __m128 dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1455 __m128 dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1456 __m128 dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1457 __m128 dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1458 __m128 velec,felec,velecsum,facel,crf,krf,krf2;
1461 __m128 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1463 __m128 dummy_mask,cutoff_mask;
1464 __m128 signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
1465 __m128 one = _mm_set1_ps(1.0);
1466 __m128 two = _mm_set1_ps(2.0);
1472 jindex = nlist->jindex;
1474 shiftidx = nlist->shift;
1476 shiftvec = fr->shift_vec[0];
1477 fshift = fr->fshift[0];
1478 facel = _mm_set1_ps(fr->epsfac);
1479 charge = mdatoms->chargeA;
1481 sh_ewald = _mm_set1_ps(fr->ic->sh_ewald);
1482 ewtab = fr->ic->tabq_coul_F;
1483 ewtabscale = _mm_set1_ps(fr->ic->tabq_scale);
1484 ewtabhalfspace = _mm_set1_ps(0.5/fr->ic->tabq_scale);
1486 /* Setup water-specific parameters */
1487 inr = nlist->iinr[0];
1488 iq1 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1]));
1489 iq2 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2]));
1490 iq3 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+3]));
1492 jq1 = _mm_set1_ps(charge[inr+1]);
1493 jq2 = _mm_set1_ps(charge[inr+2]);
1494 jq3 = _mm_set1_ps(charge[inr+3]);
1495 qq11 = _mm_mul_ps(iq1,jq1);
1496 qq12 = _mm_mul_ps(iq1,jq2);
1497 qq13 = _mm_mul_ps(iq1,jq3);
1498 qq21 = _mm_mul_ps(iq2,jq1);
1499 qq22 = _mm_mul_ps(iq2,jq2);
1500 qq23 = _mm_mul_ps(iq2,jq3);
1501 qq31 = _mm_mul_ps(iq3,jq1);
1502 qq32 = _mm_mul_ps(iq3,jq2);
1503 qq33 = _mm_mul_ps(iq3,jq3);
1505 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
1506 rcutoff_scalar = fr->rcoulomb;
1507 rcutoff = _mm_set1_ps(rcutoff_scalar);
1508 rcutoff2 = _mm_mul_ps(rcutoff,rcutoff);
1510 /* Avoid stupid compiler warnings */
1511 jnrA = jnrB = jnrC = jnrD = 0;
1512 j_coord_offsetA = 0;
1513 j_coord_offsetB = 0;
1514 j_coord_offsetC = 0;
1515 j_coord_offsetD = 0;
1520 for(iidx=0;iidx<4*DIM;iidx++)
1522 scratch[iidx] = 0.0;
1525 /* Start outer loop over neighborlists */
1526 for(iidx=0; iidx<nri; iidx++)
1528 /* Load shift vector for this list */
1529 i_shift_offset = DIM*shiftidx[iidx];
1531 /* Load limits for loop over neighbors */
1532 j_index_start = jindex[iidx];
1533 j_index_end = jindex[iidx+1];
1535 /* Get outer coordinate index */
1537 i_coord_offset = DIM*inr;
1539 /* Load i particle coords and add shift vector */
1540 gmx_mm_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
1541 &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1543 fix1 = _mm_setzero_ps();
1544 fiy1 = _mm_setzero_ps();
1545 fiz1 = _mm_setzero_ps();
1546 fix2 = _mm_setzero_ps();
1547 fiy2 = _mm_setzero_ps();
1548 fiz2 = _mm_setzero_ps();
1549 fix3 = _mm_setzero_ps();
1550 fiy3 = _mm_setzero_ps();
1551 fiz3 = _mm_setzero_ps();
1553 /* Start inner kernel loop */
1554 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
1557 /* Get j neighbor index, and coordinate index */
1559 jnrB = jjnr[jidx+1];
1560 jnrC = jjnr[jidx+2];
1561 jnrD = jjnr[jidx+3];
1562 j_coord_offsetA = DIM*jnrA;
1563 j_coord_offsetB = DIM*jnrB;
1564 j_coord_offsetC = DIM*jnrC;
1565 j_coord_offsetD = DIM*jnrD;
1567 /* load j atom coordinates */
1568 gmx_mm_load_3rvec_4ptr_swizzle_ps(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
1569 x+j_coord_offsetC+DIM,x+j_coord_offsetD+DIM,
1570 &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
1572 /* Calculate displacement vector */
1573 dx11 = _mm_sub_ps(ix1,jx1);
1574 dy11 = _mm_sub_ps(iy1,jy1);
1575 dz11 = _mm_sub_ps(iz1,jz1);
1576 dx12 = _mm_sub_ps(ix1,jx2);
1577 dy12 = _mm_sub_ps(iy1,jy2);
1578 dz12 = _mm_sub_ps(iz1,jz2);
1579 dx13 = _mm_sub_ps(ix1,jx3);
1580 dy13 = _mm_sub_ps(iy1,jy3);
1581 dz13 = _mm_sub_ps(iz1,jz3);
1582 dx21 = _mm_sub_ps(ix2,jx1);
1583 dy21 = _mm_sub_ps(iy2,jy1);
1584 dz21 = _mm_sub_ps(iz2,jz1);
1585 dx22 = _mm_sub_ps(ix2,jx2);
1586 dy22 = _mm_sub_ps(iy2,jy2);
1587 dz22 = _mm_sub_ps(iz2,jz2);
1588 dx23 = _mm_sub_ps(ix2,jx3);
1589 dy23 = _mm_sub_ps(iy2,jy3);
1590 dz23 = _mm_sub_ps(iz2,jz3);
1591 dx31 = _mm_sub_ps(ix3,jx1);
1592 dy31 = _mm_sub_ps(iy3,jy1);
1593 dz31 = _mm_sub_ps(iz3,jz1);
1594 dx32 = _mm_sub_ps(ix3,jx2);
1595 dy32 = _mm_sub_ps(iy3,jy2);
1596 dz32 = _mm_sub_ps(iz3,jz2);
1597 dx33 = _mm_sub_ps(ix3,jx3);
1598 dy33 = _mm_sub_ps(iy3,jy3);
1599 dz33 = _mm_sub_ps(iz3,jz3);
1601 /* Calculate squared distance and things based on it */
1602 rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
1603 rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
1604 rsq13 = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
1605 rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
1606 rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
1607 rsq23 = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
1608 rsq31 = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
1609 rsq32 = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
1610 rsq33 = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
1612 rinv11 = gmx_mm_invsqrt_ps(rsq11);
1613 rinv12 = gmx_mm_invsqrt_ps(rsq12);
1614 rinv13 = gmx_mm_invsqrt_ps(rsq13);
1615 rinv21 = gmx_mm_invsqrt_ps(rsq21);
1616 rinv22 = gmx_mm_invsqrt_ps(rsq22);
1617 rinv23 = gmx_mm_invsqrt_ps(rsq23);
1618 rinv31 = gmx_mm_invsqrt_ps(rsq31);
1619 rinv32 = gmx_mm_invsqrt_ps(rsq32);
1620 rinv33 = gmx_mm_invsqrt_ps(rsq33);
1622 rinvsq11 = _mm_mul_ps(rinv11,rinv11);
1623 rinvsq12 = _mm_mul_ps(rinv12,rinv12);
1624 rinvsq13 = _mm_mul_ps(rinv13,rinv13);
1625 rinvsq21 = _mm_mul_ps(rinv21,rinv21);
1626 rinvsq22 = _mm_mul_ps(rinv22,rinv22);
1627 rinvsq23 = _mm_mul_ps(rinv23,rinv23);
1628 rinvsq31 = _mm_mul_ps(rinv31,rinv31);
1629 rinvsq32 = _mm_mul_ps(rinv32,rinv32);
1630 rinvsq33 = _mm_mul_ps(rinv33,rinv33);
1632 fjx1 = _mm_setzero_ps();
1633 fjy1 = _mm_setzero_ps();
1634 fjz1 = _mm_setzero_ps();
1635 fjx2 = _mm_setzero_ps();
1636 fjy2 = _mm_setzero_ps();
1637 fjz2 = _mm_setzero_ps();
1638 fjx3 = _mm_setzero_ps();
1639 fjy3 = _mm_setzero_ps();
1640 fjz3 = _mm_setzero_ps();
1642 /**************************
1643 * CALCULATE INTERACTIONS *
1644 **************************/
1646 if (gmx_mm_any_lt(rsq11,rcutoff2))
1649 r11 = _mm_mul_ps(rsq11,rinv11);
1651 /* EWALD ELECTROSTATICS */
1653 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1654 ewrt = _mm_mul_ps(r11,ewtabscale);
1655 ewitab = _mm_cvttps_epi32(ewrt);
1656 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1657 gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
1658 ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
1660 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1661 felec = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec));
1663 cutoff_mask = _mm_cmplt_ps(rsq11,rcutoff2);
1667 fscal = _mm_and_ps(fscal,cutoff_mask);
1669 /* Calculate temporary vectorial force */
1670 tx = _mm_mul_ps(fscal,dx11);
1671 ty = _mm_mul_ps(fscal,dy11);
1672 tz = _mm_mul_ps(fscal,dz11);
1674 /* Update vectorial force */
1675 fix1 = _mm_add_ps(fix1,tx);
1676 fiy1 = _mm_add_ps(fiy1,ty);
1677 fiz1 = _mm_add_ps(fiz1,tz);
1679 fjx1 = _mm_add_ps(fjx1,tx);
1680 fjy1 = _mm_add_ps(fjy1,ty);
1681 fjz1 = _mm_add_ps(fjz1,tz);
1685 /**************************
1686 * CALCULATE INTERACTIONS *
1687 **************************/
1689 if (gmx_mm_any_lt(rsq12,rcutoff2))
1692 r12 = _mm_mul_ps(rsq12,rinv12);
1694 /* EWALD ELECTROSTATICS */
1696 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1697 ewrt = _mm_mul_ps(r12,ewtabscale);
1698 ewitab = _mm_cvttps_epi32(ewrt);
1699 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1700 gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
1701 ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
1703 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1704 felec = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec));
1706 cutoff_mask = _mm_cmplt_ps(rsq12,rcutoff2);
1710 fscal = _mm_and_ps(fscal,cutoff_mask);
1712 /* Calculate temporary vectorial force */
1713 tx = _mm_mul_ps(fscal,dx12);
1714 ty = _mm_mul_ps(fscal,dy12);
1715 tz = _mm_mul_ps(fscal,dz12);
1717 /* Update vectorial force */
1718 fix1 = _mm_add_ps(fix1,tx);
1719 fiy1 = _mm_add_ps(fiy1,ty);
1720 fiz1 = _mm_add_ps(fiz1,tz);
1722 fjx2 = _mm_add_ps(fjx2,tx);
1723 fjy2 = _mm_add_ps(fjy2,ty);
1724 fjz2 = _mm_add_ps(fjz2,tz);
1728 /**************************
1729 * CALCULATE INTERACTIONS *
1730 **************************/
1732 if (gmx_mm_any_lt(rsq13,rcutoff2))
1735 r13 = _mm_mul_ps(rsq13,rinv13);
1737 /* EWALD ELECTROSTATICS */
1739 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1740 ewrt = _mm_mul_ps(r13,ewtabscale);
1741 ewitab = _mm_cvttps_epi32(ewrt);
1742 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1743 gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
1744 ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
1746 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1747 felec = _mm_mul_ps(_mm_mul_ps(qq13,rinv13),_mm_sub_ps(rinvsq13,felec));
1749 cutoff_mask = _mm_cmplt_ps(rsq13,rcutoff2);
1753 fscal = _mm_and_ps(fscal,cutoff_mask);
1755 /* Calculate temporary vectorial force */
1756 tx = _mm_mul_ps(fscal,dx13);
1757 ty = _mm_mul_ps(fscal,dy13);
1758 tz = _mm_mul_ps(fscal,dz13);
1760 /* Update vectorial force */
1761 fix1 = _mm_add_ps(fix1,tx);
1762 fiy1 = _mm_add_ps(fiy1,ty);
1763 fiz1 = _mm_add_ps(fiz1,tz);
1765 fjx3 = _mm_add_ps(fjx3,tx);
1766 fjy3 = _mm_add_ps(fjy3,ty);
1767 fjz3 = _mm_add_ps(fjz3,tz);
1771 /**************************
1772 * CALCULATE INTERACTIONS *
1773 **************************/
1775 if (gmx_mm_any_lt(rsq21,rcutoff2))
1778 r21 = _mm_mul_ps(rsq21,rinv21);
1780 /* EWALD ELECTROSTATICS */
1782 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1783 ewrt = _mm_mul_ps(r21,ewtabscale);
1784 ewitab = _mm_cvttps_epi32(ewrt);
1785 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1786 gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
1787 ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
1789 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1790 felec = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec));
1792 cutoff_mask = _mm_cmplt_ps(rsq21,rcutoff2);
1796 fscal = _mm_and_ps(fscal,cutoff_mask);
1798 /* Calculate temporary vectorial force */
1799 tx = _mm_mul_ps(fscal,dx21);
1800 ty = _mm_mul_ps(fscal,dy21);
1801 tz = _mm_mul_ps(fscal,dz21);
1803 /* Update vectorial force */
1804 fix2 = _mm_add_ps(fix2,tx);
1805 fiy2 = _mm_add_ps(fiy2,ty);
1806 fiz2 = _mm_add_ps(fiz2,tz);
1808 fjx1 = _mm_add_ps(fjx1,tx);
1809 fjy1 = _mm_add_ps(fjy1,ty);
1810 fjz1 = _mm_add_ps(fjz1,tz);
1814 /**************************
1815 * CALCULATE INTERACTIONS *
1816 **************************/
1818 if (gmx_mm_any_lt(rsq22,rcutoff2))
1821 r22 = _mm_mul_ps(rsq22,rinv22);
1823 /* EWALD ELECTROSTATICS */
1825 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1826 ewrt = _mm_mul_ps(r22,ewtabscale);
1827 ewitab = _mm_cvttps_epi32(ewrt);
1828 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1829 gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
1830 ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
1832 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1833 felec = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec));
1835 cutoff_mask = _mm_cmplt_ps(rsq22,rcutoff2);
1839 fscal = _mm_and_ps(fscal,cutoff_mask);
1841 /* Calculate temporary vectorial force */
1842 tx = _mm_mul_ps(fscal,dx22);
1843 ty = _mm_mul_ps(fscal,dy22);
1844 tz = _mm_mul_ps(fscal,dz22);
1846 /* Update vectorial force */
1847 fix2 = _mm_add_ps(fix2,tx);
1848 fiy2 = _mm_add_ps(fiy2,ty);
1849 fiz2 = _mm_add_ps(fiz2,tz);
1851 fjx2 = _mm_add_ps(fjx2,tx);
1852 fjy2 = _mm_add_ps(fjy2,ty);
1853 fjz2 = _mm_add_ps(fjz2,tz);
1857 /**************************
1858 * CALCULATE INTERACTIONS *
1859 **************************/
1861 if (gmx_mm_any_lt(rsq23,rcutoff2))
1864 r23 = _mm_mul_ps(rsq23,rinv23);
1866 /* EWALD ELECTROSTATICS */
1868 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1869 ewrt = _mm_mul_ps(r23,ewtabscale);
1870 ewitab = _mm_cvttps_epi32(ewrt);
1871 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1872 gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
1873 ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
1875 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1876 felec = _mm_mul_ps(_mm_mul_ps(qq23,rinv23),_mm_sub_ps(rinvsq23,felec));
1878 cutoff_mask = _mm_cmplt_ps(rsq23,rcutoff2);
1882 fscal = _mm_and_ps(fscal,cutoff_mask);
1884 /* Calculate temporary vectorial force */
1885 tx = _mm_mul_ps(fscal,dx23);
1886 ty = _mm_mul_ps(fscal,dy23);
1887 tz = _mm_mul_ps(fscal,dz23);
1889 /* Update vectorial force */
1890 fix2 = _mm_add_ps(fix2,tx);
1891 fiy2 = _mm_add_ps(fiy2,ty);
1892 fiz2 = _mm_add_ps(fiz2,tz);
1894 fjx3 = _mm_add_ps(fjx3,tx);
1895 fjy3 = _mm_add_ps(fjy3,ty);
1896 fjz3 = _mm_add_ps(fjz3,tz);
1900 /**************************
1901 * CALCULATE INTERACTIONS *
1902 **************************/
1904 if (gmx_mm_any_lt(rsq31,rcutoff2))
1907 r31 = _mm_mul_ps(rsq31,rinv31);
1909 /* EWALD ELECTROSTATICS */
1911 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1912 ewrt = _mm_mul_ps(r31,ewtabscale);
1913 ewitab = _mm_cvttps_epi32(ewrt);
1914 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1915 gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
1916 ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
1918 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1919 felec = _mm_mul_ps(_mm_mul_ps(qq31,rinv31),_mm_sub_ps(rinvsq31,felec));
1921 cutoff_mask = _mm_cmplt_ps(rsq31,rcutoff2);
1925 fscal = _mm_and_ps(fscal,cutoff_mask);
1927 /* Calculate temporary vectorial force */
1928 tx = _mm_mul_ps(fscal,dx31);
1929 ty = _mm_mul_ps(fscal,dy31);
1930 tz = _mm_mul_ps(fscal,dz31);
1932 /* Update vectorial force */
1933 fix3 = _mm_add_ps(fix3,tx);
1934 fiy3 = _mm_add_ps(fiy3,ty);
1935 fiz3 = _mm_add_ps(fiz3,tz);
1937 fjx1 = _mm_add_ps(fjx1,tx);
1938 fjy1 = _mm_add_ps(fjy1,ty);
1939 fjz1 = _mm_add_ps(fjz1,tz);
1943 /**************************
1944 * CALCULATE INTERACTIONS *
1945 **************************/
1947 if (gmx_mm_any_lt(rsq32,rcutoff2))
1950 r32 = _mm_mul_ps(rsq32,rinv32);
1952 /* EWALD ELECTROSTATICS */
1954 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1955 ewrt = _mm_mul_ps(r32,ewtabscale);
1956 ewitab = _mm_cvttps_epi32(ewrt);
1957 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1958 gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
1959 ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
1961 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1962 felec = _mm_mul_ps(_mm_mul_ps(qq32,rinv32),_mm_sub_ps(rinvsq32,felec));
1964 cutoff_mask = _mm_cmplt_ps(rsq32,rcutoff2);
1968 fscal = _mm_and_ps(fscal,cutoff_mask);
1970 /* Calculate temporary vectorial force */
1971 tx = _mm_mul_ps(fscal,dx32);
1972 ty = _mm_mul_ps(fscal,dy32);
1973 tz = _mm_mul_ps(fscal,dz32);
1975 /* Update vectorial force */
1976 fix3 = _mm_add_ps(fix3,tx);
1977 fiy3 = _mm_add_ps(fiy3,ty);
1978 fiz3 = _mm_add_ps(fiz3,tz);
1980 fjx2 = _mm_add_ps(fjx2,tx);
1981 fjy2 = _mm_add_ps(fjy2,ty);
1982 fjz2 = _mm_add_ps(fjz2,tz);
1986 /**************************
1987 * CALCULATE INTERACTIONS *
1988 **************************/
1990 if (gmx_mm_any_lt(rsq33,rcutoff2))
1993 r33 = _mm_mul_ps(rsq33,rinv33);
1995 /* EWALD ELECTROSTATICS */
1997 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1998 ewrt = _mm_mul_ps(r33,ewtabscale);
1999 ewitab = _mm_cvttps_epi32(ewrt);
2000 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
2001 gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
2002 ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
2004 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2005 felec = _mm_mul_ps(_mm_mul_ps(qq33,rinv33),_mm_sub_ps(rinvsq33,felec));
2007 cutoff_mask = _mm_cmplt_ps(rsq33,rcutoff2);
2011 fscal = _mm_and_ps(fscal,cutoff_mask);
2013 /* Calculate temporary vectorial force */
2014 tx = _mm_mul_ps(fscal,dx33);
2015 ty = _mm_mul_ps(fscal,dy33);
2016 tz = _mm_mul_ps(fscal,dz33);
2018 /* Update vectorial force */
2019 fix3 = _mm_add_ps(fix3,tx);
2020 fiy3 = _mm_add_ps(fiy3,ty);
2021 fiz3 = _mm_add_ps(fiz3,tz);
2023 fjx3 = _mm_add_ps(fjx3,tx);
2024 fjy3 = _mm_add_ps(fjy3,ty);
2025 fjz3 = _mm_add_ps(fjz3,tz);
2029 fjptrA = f+j_coord_offsetA;
2030 fjptrB = f+j_coord_offsetB;
2031 fjptrC = f+j_coord_offsetC;
2032 fjptrD = f+j_coord_offsetD;
2034 gmx_mm_decrement_3rvec_4ptr_swizzle_ps(fjptrA+DIM,fjptrB+DIM,fjptrC+DIM,fjptrD+DIM,
2035 fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2037 /* Inner loop uses 351 flops */
2040 if(jidx<j_index_end)
2043 /* Get j neighbor index, and coordinate index */
2044 jnrlistA = jjnr[jidx];
2045 jnrlistB = jjnr[jidx+1];
2046 jnrlistC = jjnr[jidx+2];
2047 jnrlistD = jjnr[jidx+3];
2048 /* Sign of each element will be negative for non-real atoms.
2049 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
2050 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
2052 dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
2053 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
2054 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
2055 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
2056 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
2057 j_coord_offsetA = DIM*jnrA;
2058 j_coord_offsetB = DIM*jnrB;
2059 j_coord_offsetC = DIM*jnrC;
2060 j_coord_offsetD = DIM*jnrD;
2062 /* load j atom coordinates */
2063 gmx_mm_load_3rvec_4ptr_swizzle_ps(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
2064 x+j_coord_offsetC+DIM,x+j_coord_offsetD+DIM,
2065 &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
2067 /* Calculate displacement vector */
2068 dx11 = _mm_sub_ps(ix1,jx1);
2069 dy11 = _mm_sub_ps(iy1,jy1);
2070 dz11 = _mm_sub_ps(iz1,jz1);
2071 dx12 = _mm_sub_ps(ix1,jx2);
2072 dy12 = _mm_sub_ps(iy1,jy2);
2073 dz12 = _mm_sub_ps(iz1,jz2);
2074 dx13 = _mm_sub_ps(ix1,jx3);
2075 dy13 = _mm_sub_ps(iy1,jy3);
2076 dz13 = _mm_sub_ps(iz1,jz3);
2077 dx21 = _mm_sub_ps(ix2,jx1);
2078 dy21 = _mm_sub_ps(iy2,jy1);
2079 dz21 = _mm_sub_ps(iz2,jz1);
2080 dx22 = _mm_sub_ps(ix2,jx2);
2081 dy22 = _mm_sub_ps(iy2,jy2);
2082 dz22 = _mm_sub_ps(iz2,jz2);
2083 dx23 = _mm_sub_ps(ix2,jx3);
2084 dy23 = _mm_sub_ps(iy2,jy3);
2085 dz23 = _mm_sub_ps(iz2,jz3);
2086 dx31 = _mm_sub_ps(ix3,jx1);
2087 dy31 = _mm_sub_ps(iy3,jy1);
2088 dz31 = _mm_sub_ps(iz3,jz1);
2089 dx32 = _mm_sub_ps(ix3,jx2);
2090 dy32 = _mm_sub_ps(iy3,jy2);
2091 dz32 = _mm_sub_ps(iz3,jz2);
2092 dx33 = _mm_sub_ps(ix3,jx3);
2093 dy33 = _mm_sub_ps(iy3,jy3);
2094 dz33 = _mm_sub_ps(iz3,jz3);
2096 /* Calculate squared distance and things based on it */
2097 rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
2098 rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
2099 rsq13 = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
2100 rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
2101 rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
2102 rsq23 = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
2103 rsq31 = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
2104 rsq32 = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
2105 rsq33 = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
2107 rinv11 = gmx_mm_invsqrt_ps(rsq11);
2108 rinv12 = gmx_mm_invsqrt_ps(rsq12);
2109 rinv13 = gmx_mm_invsqrt_ps(rsq13);
2110 rinv21 = gmx_mm_invsqrt_ps(rsq21);
2111 rinv22 = gmx_mm_invsqrt_ps(rsq22);
2112 rinv23 = gmx_mm_invsqrt_ps(rsq23);
2113 rinv31 = gmx_mm_invsqrt_ps(rsq31);
2114 rinv32 = gmx_mm_invsqrt_ps(rsq32);
2115 rinv33 = gmx_mm_invsqrt_ps(rsq33);
2117 rinvsq11 = _mm_mul_ps(rinv11,rinv11);
2118 rinvsq12 = _mm_mul_ps(rinv12,rinv12);
2119 rinvsq13 = _mm_mul_ps(rinv13,rinv13);
2120 rinvsq21 = _mm_mul_ps(rinv21,rinv21);
2121 rinvsq22 = _mm_mul_ps(rinv22,rinv22);
2122 rinvsq23 = _mm_mul_ps(rinv23,rinv23);
2123 rinvsq31 = _mm_mul_ps(rinv31,rinv31);
2124 rinvsq32 = _mm_mul_ps(rinv32,rinv32);
2125 rinvsq33 = _mm_mul_ps(rinv33,rinv33);
2127 fjx1 = _mm_setzero_ps();
2128 fjy1 = _mm_setzero_ps();
2129 fjz1 = _mm_setzero_ps();
2130 fjx2 = _mm_setzero_ps();
2131 fjy2 = _mm_setzero_ps();
2132 fjz2 = _mm_setzero_ps();
2133 fjx3 = _mm_setzero_ps();
2134 fjy3 = _mm_setzero_ps();
2135 fjz3 = _mm_setzero_ps();
2137 /**************************
2138 * CALCULATE INTERACTIONS *
2139 **************************/
2141 if (gmx_mm_any_lt(rsq11,rcutoff2))
2144 r11 = _mm_mul_ps(rsq11,rinv11);
2145 r11 = _mm_andnot_ps(dummy_mask,r11);
2147 /* EWALD ELECTROSTATICS */
2149 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2150 ewrt = _mm_mul_ps(r11,ewtabscale);
2151 ewitab = _mm_cvttps_epi32(ewrt);
2152 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
2153 gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
2154 ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
2156 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2157 felec = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec));
2159 cutoff_mask = _mm_cmplt_ps(rsq11,rcutoff2);
2163 fscal = _mm_and_ps(fscal,cutoff_mask);
2165 fscal = _mm_andnot_ps(dummy_mask,fscal);
2167 /* Calculate temporary vectorial force */
2168 tx = _mm_mul_ps(fscal,dx11);
2169 ty = _mm_mul_ps(fscal,dy11);
2170 tz = _mm_mul_ps(fscal,dz11);
2172 /* Update vectorial force */
2173 fix1 = _mm_add_ps(fix1,tx);
2174 fiy1 = _mm_add_ps(fiy1,ty);
2175 fiz1 = _mm_add_ps(fiz1,tz);
2177 fjx1 = _mm_add_ps(fjx1,tx);
2178 fjy1 = _mm_add_ps(fjy1,ty);
2179 fjz1 = _mm_add_ps(fjz1,tz);
2183 /**************************
2184 * CALCULATE INTERACTIONS *
2185 **************************/
2187 if (gmx_mm_any_lt(rsq12,rcutoff2))
2190 r12 = _mm_mul_ps(rsq12,rinv12);
2191 r12 = _mm_andnot_ps(dummy_mask,r12);
2193 /* EWALD ELECTROSTATICS */
2195 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2196 ewrt = _mm_mul_ps(r12,ewtabscale);
2197 ewitab = _mm_cvttps_epi32(ewrt);
2198 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
2199 gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
2200 ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
2202 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2203 felec = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec));
2205 cutoff_mask = _mm_cmplt_ps(rsq12,rcutoff2);
2209 fscal = _mm_and_ps(fscal,cutoff_mask);
2211 fscal = _mm_andnot_ps(dummy_mask,fscal);
2213 /* Calculate temporary vectorial force */
2214 tx = _mm_mul_ps(fscal,dx12);
2215 ty = _mm_mul_ps(fscal,dy12);
2216 tz = _mm_mul_ps(fscal,dz12);
2218 /* Update vectorial force */
2219 fix1 = _mm_add_ps(fix1,tx);
2220 fiy1 = _mm_add_ps(fiy1,ty);
2221 fiz1 = _mm_add_ps(fiz1,tz);
2223 fjx2 = _mm_add_ps(fjx2,tx);
2224 fjy2 = _mm_add_ps(fjy2,ty);
2225 fjz2 = _mm_add_ps(fjz2,tz);
2229 /**************************
2230 * CALCULATE INTERACTIONS *
2231 **************************/
2233 if (gmx_mm_any_lt(rsq13,rcutoff2))
2236 r13 = _mm_mul_ps(rsq13,rinv13);
2237 r13 = _mm_andnot_ps(dummy_mask,r13);
2239 /* EWALD ELECTROSTATICS */
2241 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2242 ewrt = _mm_mul_ps(r13,ewtabscale);
2243 ewitab = _mm_cvttps_epi32(ewrt);
2244 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
2245 gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
2246 ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
2248 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2249 felec = _mm_mul_ps(_mm_mul_ps(qq13,rinv13),_mm_sub_ps(rinvsq13,felec));
2251 cutoff_mask = _mm_cmplt_ps(rsq13,rcutoff2);
2255 fscal = _mm_and_ps(fscal,cutoff_mask);
2257 fscal = _mm_andnot_ps(dummy_mask,fscal);
2259 /* Calculate temporary vectorial force */
2260 tx = _mm_mul_ps(fscal,dx13);
2261 ty = _mm_mul_ps(fscal,dy13);
2262 tz = _mm_mul_ps(fscal,dz13);
2264 /* Update vectorial force */
2265 fix1 = _mm_add_ps(fix1,tx);
2266 fiy1 = _mm_add_ps(fiy1,ty);
2267 fiz1 = _mm_add_ps(fiz1,tz);
2269 fjx3 = _mm_add_ps(fjx3,tx);
2270 fjy3 = _mm_add_ps(fjy3,ty);
2271 fjz3 = _mm_add_ps(fjz3,tz);
2275 /**************************
2276 * CALCULATE INTERACTIONS *
2277 **************************/
2279 if (gmx_mm_any_lt(rsq21,rcutoff2))
2282 r21 = _mm_mul_ps(rsq21,rinv21);
2283 r21 = _mm_andnot_ps(dummy_mask,r21);
2285 /* EWALD ELECTROSTATICS */
2287 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2288 ewrt = _mm_mul_ps(r21,ewtabscale);
2289 ewitab = _mm_cvttps_epi32(ewrt);
2290 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
2291 gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
2292 ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
2294 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2295 felec = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec));
2297 cutoff_mask = _mm_cmplt_ps(rsq21,rcutoff2);
2301 fscal = _mm_and_ps(fscal,cutoff_mask);
2303 fscal = _mm_andnot_ps(dummy_mask,fscal);
2305 /* Calculate temporary vectorial force */
2306 tx = _mm_mul_ps(fscal,dx21);
2307 ty = _mm_mul_ps(fscal,dy21);
2308 tz = _mm_mul_ps(fscal,dz21);
2310 /* Update vectorial force */
2311 fix2 = _mm_add_ps(fix2,tx);
2312 fiy2 = _mm_add_ps(fiy2,ty);
2313 fiz2 = _mm_add_ps(fiz2,tz);
2315 fjx1 = _mm_add_ps(fjx1,tx);
2316 fjy1 = _mm_add_ps(fjy1,ty);
2317 fjz1 = _mm_add_ps(fjz1,tz);
2321 /**************************
2322 * CALCULATE INTERACTIONS *
2323 **************************/
2325 if (gmx_mm_any_lt(rsq22,rcutoff2))
2328 r22 = _mm_mul_ps(rsq22,rinv22);
2329 r22 = _mm_andnot_ps(dummy_mask,r22);
2331 /* EWALD ELECTROSTATICS */
2333 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2334 ewrt = _mm_mul_ps(r22,ewtabscale);
2335 ewitab = _mm_cvttps_epi32(ewrt);
2336 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
2337 gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
2338 ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
2340 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2341 felec = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec));
2343 cutoff_mask = _mm_cmplt_ps(rsq22,rcutoff2);
2347 fscal = _mm_and_ps(fscal,cutoff_mask);
2349 fscal = _mm_andnot_ps(dummy_mask,fscal);
2351 /* Calculate temporary vectorial force */
2352 tx = _mm_mul_ps(fscal,dx22);
2353 ty = _mm_mul_ps(fscal,dy22);
2354 tz = _mm_mul_ps(fscal,dz22);
2356 /* Update vectorial force */
2357 fix2 = _mm_add_ps(fix2,tx);
2358 fiy2 = _mm_add_ps(fiy2,ty);
2359 fiz2 = _mm_add_ps(fiz2,tz);
2361 fjx2 = _mm_add_ps(fjx2,tx);
2362 fjy2 = _mm_add_ps(fjy2,ty);
2363 fjz2 = _mm_add_ps(fjz2,tz);
2367 /**************************
2368 * CALCULATE INTERACTIONS *
2369 **************************/
2371 if (gmx_mm_any_lt(rsq23,rcutoff2))
2374 r23 = _mm_mul_ps(rsq23,rinv23);
2375 r23 = _mm_andnot_ps(dummy_mask,r23);
2377 /* EWALD ELECTROSTATICS */
2379 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2380 ewrt = _mm_mul_ps(r23,ewtabscale);
2381 ewitab = _mm_cvttps_epi32(ewrt);
2382 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
2383 gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
2384 ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
2386 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2387 felec = _mm_mul_ps(_mm_mul_ps(qq23,rinv23),_mm_sub_ps(rinvsq23,felec));
2389 cutoff_mask = _mm_cmplt_ps(rsq23,rcutoff2);
2393 fscal = _mm_and_ps(fscal,cutoff_mask);
2395 fscal = _mm_andnot_ps(dummy_mask,fscal);
2397 /* Calculate temporary vectorial force */
2398 tx = _mm_mul_ps(fscal,dx23);
2399 ty = _mm_mul_ps(fscal,dy23);
2400 tz = _mm_mul_ps(fscal,dz23);
2402 /* Update vectorial force */
2403 fix2 = _mm_add_ps(fix2,tx);
2404 fiy2 = _mm_add_ps(fiy2,ty);
2405 fiz2 = _mm_add_ps(fiz2,tz);
2407 fjx3 = _mm_add_ps(fjx3,tx);
2408 fjy3 = _mm_add_ps(fjy3,ty);
2409 fjz3 = _mm_add_ps(fjz3,tz);
2413 /**************************
2414 * CALCULATE INTERACTIONS *
2415 **************************/
2417 if (gmx_mm_any_lt(rsq31,rcutoff2))
2420 r31 = _mm_mul_ps(rsq31,rinv31);
2421 r31 = _mm_andnot_ps(dummy_mask,r31);
2423 /* EWALD ELECTROSTATICS */
2425 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2426 ewrt = _mm_mul_ps(r31,ewtabscale);
2427 ewitab = _mm_cvttps_epi32(ewrt);
2428 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
2429 gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
2430 ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
2432 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2433 felec = _mm_mul_ps(_mm_mul_ps(qq31,rinv31),_mm_sub_ps(rinvsq31,felec));
2435 cutoff_mask = _mm_cmplt_ps(rsq31,rcutoff2);
2439 fscal = _mm_and_ps(fscal,cutoff_mask);
2441 fscal = _mm_andnot_ps(dummy_mask,fscal);
2443 /* Calculate temporary vectorial force */
2444 tx = _mm_mul_ps(fscal,dx31);
2445 ty = _mm_mul_ps(fscal,dy31);
2446 tz = _mm_mul_ps(fscal,dz31);
2448 /* Update vectorial force */
2449 fix3 = _mm_add_ps(fix3,tx);
2450 fiy3 = _mm_add_ps(fiy3,ty);
2451 fiz3 = _mm_add_ps(fiz3,tz);
2453 fjx1 = _mm_add_ps(fjx1,tx);
2454 fjy1 = _mm_add_ps(fjy1,ty);
2455 fjz1 = _mm_add_ps(fjz1,tz);
2459 /**************************
2460 * CALCULATE INTERACTIONS *
2461 **************************/
2463 if (gmx_mm_any_lt(rsq32,rcutoff2))
2466 r32 = _mm_mul_ps(rsq32,rinv32);
2467 r32 = _mm_andnot_ps(dummy_mask,r32);
2469 /* EWALD ELECTROSTATICS */
2471 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2472 ewrt = _mm_mul_ps(r32,ewtabscale);
2473 ewitab = _mm_cvttps_epi32(ewrt);
2474 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
2475 gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
2476 ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
2478 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2479 felec = _mm_mul_ps(_mm_mul_ps(qq32,rinv32),_mm_sub_ps(rinvsq32,felec));
2481 cutoff_mask = _mm_cmplt_ps(rsq32,rcutoff2);
2485 fscal = _mm_and_ps(fscal,cutoff_mask);
2487 fscal = _mm_andnot_ps(dummy_mask,fscal);
2489 /* Calculate temporary vectorial force */
2490 tx = _mm_mul_ps(fscal,dx32);
2491 ty = _mm_mul_ps(fscal,dy32);
2492 tz = _mm_mul_ps(fscal,dz32);
2494 /* Update vectorial force */
2495 fix3 = _mm_add_ps(fix3,tx);
2496 fiy3 = _mm_add_ps(fiy3,ty);
2497 fiz3 = _mm_add_ps(fiz3,tz);
2499 fjx2 = _mm_add_ps(fjx2,tx);
2500 fjy2 = _mm_add_ps(fjy2,ty);
2501 fjz2 = _mm_add_ps(fjz2,tz);
2505 /**************************
2506 * CALCULATE INTERACTIONS *
2507 **************************/
2509 if (gmx_mm_any_lt(rsq33,rcutoff2))
2512 r33 = _mm_mul_ps(rsq33,rinv33);
2513 r33 = _mm_andnot_ps(dummy_mask,r33);
2515 /* EWALD ELECTROSTATICS */
2517 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2518 ewrt = _mm_mul_ps(r33,ewtabscale);
2519 ewitab = _mm_cvttps_epi32(ewrt);
2520 eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
2521 gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
2522 ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
2524 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2525 felec = _mm_mul_ps(_mm_mul_ps(qq33,rinv33),_mm_sub_ps(rinvsq33,felec));
2527 cutoff_mask = _mm_cmplt_ps(rsq33,rcutoff2);
2531 fscal = _mm_and_ps(fscal,cutoff_mask);
2533 fscal = _mm_andnot_ps(dummy_mask,fscal);
2535 /* Calculate temporary vectorial force */
2536 tx = _mm_mul_ps(fscal,dx33);
2537 ty = _mm_mul_ps(fscal,dy33);
2538 tz = _mm_mul_ps(fscal,dz33);
2540 /* Update vectorial force */
2541 fix3 = _mm_add_ps(fix3,tx);
2542 fiy3 = _mm_add_ps(fiy3,ty);
2543 fiz3 = _mm_add_ps(fiz3,tz);
2545 fjx3 = _mm_add_ps(fjx3,tx);
2546 fjy3 = _mm_add_ps(fjy3,ty);
2547 fjz3 = _mm_add_ps(fjz3,tz);
2551 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
2552 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
2553 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
2554 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
2556 gmx_mm_decrement_3rvec_4ptr_swizzle_ps(fjptrA+DIM,fjptrB+DIM,fjptrC+DIM,fjptrD+DIM,
2557 fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2559 /* Inner loop uses 360 flops */
2562 /* End of innermost loop */
2564 gmx_mm_update_iforce_3atom_swizzle_ps(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2565 f+i_coord_offset+DIM,fshift+i_shift_offset);
2567 /* Increment number of inner iterations */
2568 inneriter += j_index_end - j_index_start;
2570 /* Outer loop uses 18 flops */
2573 /* Increment number of outer iterations */
2576 /* Update outer/inner flops */
2578 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4W4_F,outeriter*18 + inneriter*360);