2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2012,2013, 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 sse2_single kernel generator.
44 #include "../nb_kernel.h"
45 #include "types/simple.h"
49 #include "gromacs/simd/math_x86_sse2_single.h"
50 #include "kernelutil_x86_sse2_single.h"
53 * Gromacs nonbonded kernel: nb_kernel_ElecEw_VdwLJ_GeomW4W4_VF_sse2_single
54 * Electrostatics interaction: Ewald
55 * VdW interaction: LennardJones
56 * Geometry: Water4-Water4
57 * Calculate force/pot: PotentialAndForce
60 nb_kernel_ElecEw_VdwLJ_GeomW4W4_VF_sse2_single
61 (t_nblist * gmx_restrict nlist,
62 rvec * gmx_restrict xx,
63 rvec * gmx_restrict ff,
64 t_forcerec * gmx_restrict fr,
65 t_mdatoms * gmx_restrict mdatoms,
66 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
67 t_nrnb * gmx_restrict nrnb)
69 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
70 * just 0 for non-waters.
71 * Suffixes A,B,C,D refer to j loop unrolling done with SSE, e.g. for the four different
72 * jnr indices corresponding to data put in the four positions in the SIMD register.
74 int i_shift_offset,i_coord_offset,outeriter,inneriter;
75 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
76 int jnrA,jnrB,jnrC,jnrD;
77 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
78 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
79 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
81 real *shiftvec,*fshift,*x,*f;
82 real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
84 __m128 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
86 __m128 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
88 __m128 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
90 __m128 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
92 __m128 ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
93 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
94 __m128 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
95 int vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
96 __m128 jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
97 int vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
98 __m128 jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
99 int vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
100 __m128 jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
101 __m128 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
102 __m128 dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
103 __m128 dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
104 __m128 dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
105 __m128 dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
106 __m128 dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
107 __m128 dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
108 __m128 dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
109 __m128 dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
110 __m128 dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
111 __m128 velec,felec,velecsum,facel,crf,krf,krf2;
114 __m128 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
117 __m128 one_sixth = _mm_set1_ps(1.0/6.0);
118 __m128 one_twelfth = _mm_set1_ps(1.0/12.0);
120 __m128 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
122 __m128 dummy_mask,cutoff_mask;
123 __m128 signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
124 __m128 one = _mm_set1_ps(1.0);
125 __m128 two = _mm_set1_ps(2.0);
131 jindex = nlist->jindex;
133 shiftidx = nlist->shift;
135 shiftvec = fr->shift_vec[0];
136 fshift = fr->fshift[0];
137 facel = _mm_set1_ps(fr->epsfac);
138 charge = mdatoms->chargeA;
139 nvdwtype = fr->ntype;
141 vdwtype = mdatoms->typeA;
143 sh_ewald = _mm_set1_ps(fr->ic->sh_ewald);
144 ewtab = fr->ic->tabq_coul_FDV0;
145 ewtabscale = _mm_set1_ps(fr->ic->tabq_scale);
146 ewtabhalfspace = _mm_set1_ps(0.5/fr->ic->tabq_scale);
148 /* Setup water-specific parameters */
149 inr = nlist->iinr[0];
150 iq1 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1]));
151 iq2 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2]));
152 iq3 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+3]));
153 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
155 jq1 = _mm_set1_ps(charge[inr+1]);
156 jq2 = _mm_set1_ps(charge[inr+2]);
157 jq3 = _mm_set1_ps(charge[inr+3]);
158 vdwjidx0A = 2*vdwtype[inr+0];
159 c6_00 = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A]);
160 c12_00 = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A+1]);
161 qq11 = _mm_mul_ps(iq1,jq1);
162 qq12 = _mm_mul_ps(iq1,jq2);
163 qq13 = _mm_mul_ps(iq1,jq3);
164 qq21 = _mm_mul_ps(iq2,jq1);
165 qq22 = _mm_mul_ps(iq2,jq2);
166 qq23 = _mm_mul_ps(iq2,jq3);
167 qq31 = _mm_mul_ps(iq3,jq1);
168 qq32 = _mm_mul_ps(iq3,jq2);
169 qq33 = _mm_mul_ps(iq3,jq3);
171 /* Avoid stupid compiler warnings */
172 jnrA = jnrB = jnrC = jnrD = 0;
181 for(iidx=0;iidx<4*DIM;iidx++)
186 /* Start outer loop over neighborlists */
187 for(iidx=0; iidx<nri; iidx++)
189 /* Load shift vector for this list */
190 i_shift_offset = DIM*shiftidx[iidx];
192 /* Load limits for loop over neighbors */
193 j_index_start = jindex[iidx];
194 j_index_end = jindex[iidx+1];
196 /* Get outer coordinate index */
198 i_coord_offset = DIM*inr;
200 /* Load i particle coords and add shift vector */
201 gmx_mm_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
202 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
204 fix0 = _mm_setzero_ps();
205 fiy0 = _mm_setzero_ps();
206 fiz0 = _mm_setzero_ps();
207 fix1 = _mm_setzero_ps();
208 fiy1 = _mm_setzero_ps();
209 fiz1 = _mm_setzero_ps();
210 fix2 = _mm_setzero_ps();
211 fiy2 = _mm_setzero_ps();
212 fiz2 = _mm_setzero_ps();
213 fix3 = _mm_setzero_ps();
214 fiy3 = _mm_setzero_ps();
215 fiz3 = _mm_setzero_ps();
217 /* Reset potential sums */
218 velecsum = _mm_setzero_ps();
219 vvdwsum = _mm_setzero_ps();
221 /* Start inner kernel loop */
222 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
225 /* Get j neighbor index, and coordinate index */
230 j_coord_offsetA = DIM*jnrA;
231 j_coord_offsetB = DIM*jnrB;
232 j_coord_offsetC = DIM*jnrC;
233 j_coord_offsetD = DIM*jnrD;
235 /* load j atom coordinates */
236 gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
237 x+j_coord_offsetC,x+j_coord_offsetD,
238 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
239 &jy2,&jz2,&jx3,&jy3,&jz3);
241 /* Calculate displacement vector */
242 dx00 = _mm_sub_ps(ix0,jx0);
243 dy00 = _mm_sub_ps(iy0,jy0);
244 dz00 = _mm_sub_ps(iz0,jz0);
245 dx11 = _mm_sub_ps(ix1,jx1);
246 dy11 = _mm_sub_ps(iy1,jy1);
247 dz11 = _mm_sub_ps(iz1,jz1);
248 dx12 = _mm_sub_ps(ix1,jx2);
249 dy12 = _mm_sub_ps(iy1,jy2);
250 dz12 = _mm_sub_ps(iz1,jz2);
251 dx13 = _mm_sub_ps(ix1,jx3);
252 dy13 = _mm_sub_ps(iy1,jy3);
253 dz13 = _mm_sub_ps(iz1,jz3);
254 dx21 = _mm_sub_ps(ix2,jx1);
255 dy21 = _mm_sub_ps(iy2,jy1);
256 dz21 = _mm_sub_ps(iz2,jz1);
257 dx22 = _mm_sub_ps(ix2,jx2);
258 dy22 = _mm_sub_ps(iy2,jy2);
259 dz22 = _mm_sub_ps(iz2,jz2);
260 dx23 = _mm_sub_ps(ix2,jx3);
261 dy23 = _mm_sub_ps(iy2,jy3);
262 dz23 = _mm_sub_ps(iz2,jz3);
263 dx31 = _mm_sub_ps(ix3,jx1);
264 dy31 = _mm_sub_ps(iy3,jy1);
265 dz31 = _mm_sub_ps(iz3,jz1);
266 dx32 = _mm_sub_ps(ix3,jx2);
267 dy32 = _mm_sub_ps(iy3,jy2);
268 dz32 = _mm_sub_ps(iz3,jz2);
269 dx33 = _mm_sub_ps(ix3,jx3);
270 dy33 = _mm_sub_ps(iy3,jy3);
271 dz33 = _mm_sub_ps(iz3,jz3);
273 /* Calculate squared distance and things based on it */
274 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
275 rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
276 rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
277 rsq13 = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
278 rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
279 rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
280 rsq23 = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
281 rsq31 = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
282 rsq32 = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
283 rsq33 = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
285 rinv11 = gmx_mm_invsqrt_ps(rsq11);
286 rinv12 = gmx_mm_invsqrt_ps(rsq12);
287 rinv13 = gmx_mm_invsqrt_ps(rsq13);
288 rinv21 = gmx_mm_invsqrt_ps(rsq21);
289 rinv22 = gmx_mm_invsqrt_ps(rsq22);
290 rinv23 = gmx_mm_invsqrt_ps(rsq23);
291 rinv31 = gmx_mm_invsqrt_ps(rsq31);
292 rinv32 = gmx_mm_invsqrt_ps(rsq32);
293 rinv33 = gmx_mm_invsqrt_ps(rsq33);
295 rinvsq00 = gmx_mm_inv_ps(rsq00);
296 rinvsq11 = _mm_mul_ps(rinv11,rinv11);
297 rinvsq12 = _mm_mul_ps(rinv12,rinv12);
298 rinvsq13 = _mm_mul_ps(rinv13,rinv13);
299 rinvsq21 = _mm_mul_ps(rinv21,rinv21);
300 rinvsq22 = _mm_mul_ps(rinv22,rinv22);
301 rinvsq23 = _mm_mul_ps(rinv23,rinv23);
302 rinvsq31 = _mm_mul_ps(rinv31,rinv31);
303 rinvsq32 = _mm_mul_ps(rinv32,rinv32);
304 rinvsq33 = _mm_mul_ps(rinv33,rinv33);
306 fjx0 = _mm_setzero_ps();
307 fjy0 = _mm_setzero_ps();
308 fjz0 = _mm_setzero_ps();
309 fjx1 = _mm_setzero_ps();
310 fjy1 = _mm_setzero_ps();
311 fjz1 = _mm_setzero_ps();
312 fjx2 = _mm_setzero_ps();
313 fjy2 = _mm_setzero_ps();
314 fjz2 = _mm_setzero_ps();
315 fjx3 = _mm_setzero_ps();
316 fjy3 = _mm_setzero_ps();
317 fjz3 = _mm_setzero_ps();
319 /**************************
320 * CALCULATE INTERACTIONS *
321 **************************/
323 /* LENNARD-JONES DISPERSION/REPULSION */
325 rinvsix = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00);
326 vvdw6 = _mm_mul_ps(c6_00,rinvsix);
327 vvdw12 = _mm_mul_ps(c12_00,_mm_mul_ps(rinvsix,rinvsix));
328 vvdw = _mm_sub_ps( _mm_mul_ps(vvdw12,one_twelfth) , _mm_mul_ps(vvdw6,one_sixth) );
329 fvdw = _mm_mul_ps(_mm_sub_ps(vvdw12,vvdw6),rinvsq00);
331 /* Update potential sum for this i atom from the interaction with this j atom. */
332 vvdwsum = _mm_add_ps(vvdwsum,vvdw);
336 /* Calculate temporary vectorial force */
337 tx = _mm_mul_ps(fscal,dx00);
338 ty = _mm_mul_ps(fscal,dy00);
339 tz = _mm_mul_ps(fscal,dz00);
341 /* Update vectorial force */
342 fix0 = _mm_add_ps(fix0,tx);
343 fiy0 = _mm_add_ps(fiy0,ty);
344 fiz0 = _mm_add_ps(fiz0,tz);
346 fjx0 = _mm_add_ps(fjx0,tx);
347 fjy0 = _mm_add_ps(fjy0,ty);
348 fjz0 = _mm_add_ps(fjz0,tz);
350 /**************************
351 * CALCULATE INTERACTIONS *
352 **************************/
354 r11 = _mm_mul_ps(rsq11,rinv11);
356 /* EWALD ELECTROSTATICS */
358 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
359 ewrt = _mm_mul_ps(r11,ewtabscale);
360 ewitab = _mm_cvttps_epi32(ewrt);
361 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
362 ewitab = _mm_slli_epi32(ewitab,2);
363 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
364 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
365 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
366 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
367 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
368 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
369 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
370 velec = _mm_mul_ps(qq11,_mm_sub_ps(rinv11,velec));
371 felec = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec));
373 /* Update potential sum for this i atom from the interaction with this j atom. */
374 velecsum = _mm_add_ps(velecsum,velec);
378 /* Calculate temporary vectorial force */
379 tx = _mm_mul_ps(fscal,dx11);
380 ty = _mm_mul_ps(fscal,dy11);
381 tz = _mm_mul_ps(fscal,dz11);
383 /* Update vectorial force */
384 fix1 = _mm_add_ps(fix1,tx);
385 fiy1 = _mm_add_ps(fiy1,ty);
386 fiz1 = _mm_add_ps(fiz1,tz);
388 fjx1 = _mm_add_ps(fjx1,tx);
389 fjy1 = _mm_add_ps(fjy1,ty);
390 fjz1 = _mm_add_ps(fjz1,tz);
392 /**************************
393 * CALCULATE INTERACTIONS *
394 **************************/
396 r12 = _mm_mul_ps(rsq12,rinv12);
398 /* EWALD ELECTROSTATICS */
400 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
401 ewrt = _mm_mul_ps(r12,ewtabscale);
402 ewitab = _mm_cvttps_epi32(ewrt);
403 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
404 ewitab = _mm_slli_epi32(ewitab,2);
405 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
406 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
407 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
408 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
409 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
410 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
411 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
412 velec = _mm_mul_ps(qq12,_mm_sub_ps(rinv12,velec));
413 felec = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec));
415 /* Update potential sum for this i atom from the interaction with this j atom. */
416 velecsum = _mm_add_ps(velecsum,velec);
420 /* Calculate temporary vectorial force */
421 tx = _mm_mul_ps(fscal,dx12);
422 ty = _mm_mul_ps(fscal,dy12);
423 tz = _mm_mul_ps(fscal,dz12);
425 /* Update vectorial force */
426 fix1 = _mm_add_ps(fix1,tx);
427 fiy1 = _mm_add_ps(fiy1,ty);
428 fiz1 = _mm_add_ps(fiz1,tz);
430 fjx2 = _mm_add_ps(fjx2,tx);
431 fjy2 = _mm_add_ps(fjy2,ty);
432 fjz2 = _mm_add_ps(fjz2,tz);
434 /**************************
435 * CALCULATE INTERACTIONS *
436 **************************/
438 r13 = _mm_mul_ps(rsq13,rinv13);
440 /* EWALD ELECTROSTATICS */
442 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
443 ewrt = _mm_mul_ps(r13,ewtabscale);
444 ewitab = _mm_cvttps_epi32(ewrt);
445 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
446 ewitab = _mm_slli_epi32(ewitab,2);
447 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
448 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
449 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
450 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
451 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
452 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
453 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
454 velec = _mm_mul_ps(qq13,_mm_sub_ps(rinv13,velec));
455 felec = _mm_mul_ps(_mm_mul_ps(qq13,rinv13),_mm_sub_ps(rinvsq13,felec));
457 /* Update potential sum for this i atom from the interaction with this j atom. */
458 velecsum = _mm_add_ps(velecsum,velec);
462 /* Calculate temporary vectorial force */
463 tx = _mm_mul_ps(fscal,dx13);
464 ty = _mm_mul_ps(fscal,dy13);
465 tz = _mm_mul_ps(fscal,dz13);
467 /* Update vectorial force */
468 fix1 = _mm_add_ps(fix1,tx);
469 fiy1 = _mm_add_ps(fiy1,ty);
470 fiz1 = _mm_add_ps(fiz1,tz);
472 fjx3 = _mm_add_ps(fjx3,tx);
473 fjy3 = _mm_add_ps(fjy3,ty);
474 fjz3 = _mm_add_ps(fjz3,tz);
476 /**************************
477 * CALCULATE INTERACTIONS *
478 **************************/
480 r21 = _mm_mul_ps(rsq21,rinv21);
482 /* EWALD ELECTROSTATICS */
484 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
485 ewrt = _mm_mul_ps(r21,ewtabscale);
486 ewitab = _mm_cvttps_epi32(ewrt);
487 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
488 ewitab = _mm_slli_epi32(ewitab,2);
489 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
490 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
491 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
492 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
493 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
494 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
495 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
496 velec = _mm_mul_ps(qq21,_mm_sub_ps(rinv21,velec));
497 felec = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec));
499 /* Update potential sum for this i atom from the interaction with this j atom. */
500 velecsum = _mm_add_ps(velecsum,velec);
504 /* Calculate temporary vectorial force */
505 tx = _mm_mul_ps(fscal,dx21);
506 ty = _mm_mul_ps(fscal,dy21);
507 tz = _mm_mul_ps(fscal,dz21);
509 /* Update vectorial force */
510 fix2 = _mm_add_ps(fix2,tx);
511 fiy2 = _mm_add_ps(fiy2,ty);
512 fiz2 = _mm_add_ps(fiz2,tz);
514 fjx1 = _mm_add_ps(fjx1,tx);
515 fjy1 = _mm_add_ps(fjy1,ty);
516 fjz1 = _mm_add_ps(fjz1,tz);
518 /**************************
519 * CALCULATE INTERACTIONS *
520 **************************/
522 r22 = _mm_mul_ps(rsq22,rinv22);
524 /* EWALD ELECTROSTATICS */
526 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
527 ewrt = _mm_mul_ps(r22,ewtabscale);
528 ewitab = _mm_cvttps_epi32(ewrt);
529 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
530 ewitab = _mm_slli_epi32(ewitab,2);
531 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
532 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
533 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
534 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
535 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
536 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
537 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
538 velec = _mm_mul_ps(qq22,_mm_sub_ps(rinv22,velec));
539 felec = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec));
541 /* Update potential sum for this i atom from the interaction with this j atom. */
542 velecsum = _mm_add_ps(velecsum,velec);
546 /* Calculate temporary vectorial force */
547 tx = _mm_mul_ps(fscal,dx22);
548 ty = _mm_mul_ps(fscal,dy22);
549 tz = _mm_mul_ps(fscal,dz22);
551 /* Update vectorial force */
552 fix2 = _mm_add_ps(fix2,tx);
553 fiy2 = _mm_add_ps(fiy2,ty);
554 fiz2 = _mm_add_ps(fiz2,tz);
556 fjx2 = _mm_add_ps(fjx2,tx);
557 fjy2 = _mm_add_ps(fjy2,ty);
558 fjz2 = _mm_add_ps(fjz2,tz);
560 /**************************
561 * CALCULATE INTERACTIONS *
562 **************************/
564 r23 = _mm_mul_ps(rsq23,rinv23);
566 /* EWALD ELECTROSTATICS */
568 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
569 ewrt = _mm_mul_ps(r23,ewtabscale);
570 ewitab = _mm_cvttps_epi32(ewrt);
571 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
572 ewitab = _mm_slli_epi32(ewitab,2);
573 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
574 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
575 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
576 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
577 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
578 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
579 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
580 velec = _mm_mul_ps(qq23,_mm_sub_ps(rinv23,velec));
581 felec = _mm_mul_ps(_mm_mul_ps(qq23,rinv23),_mm_sub_ps(rinvsq23,felec));
583 /* Update potential sum for this i atom from the interaction with this j atom. */
584 velecsum = _mm_add_ps(velecsum,velec);
588 /* Calculate temporary vectorial force */
589 tx = _mm_mul_ps(fscal,dx23);
590 ty = _mm_mul_ps(fscal,dy23);
591 tz = _mm_mul_ps(fscal,dz23);
593 /* Update vectorial force */
594 fix2 = _mm_add_ps(fix2,tx);
595 fiy2 = _mm_add_ps(fiy2,ty);
596 fiz2 = _mm_add_ps(fiz2,tz);
598 fjx3 = _mm_add_ps(fjx3,tx);
599 fjy3 = _mm_add_ps(fjy3,ty);
600 fjz3 = _mm_add_ps(fjz3,tz);
602 /**************************
603 * CALCULATE INTERACTIONS *
604 **************************/
606 r31 = _mm_mul_ps(rsq31,rinv31);
608 /* EWALD ELECTROSTATICS */
610 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
611 ewrt = _mm_mul_ps(r31,ewtabscale);
612 ewitab = _mm_cvttps_epi32(ewrt);
613 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
614 ewitab = _mm_slli_epi32(ewitab,2);
615 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
616 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
617 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
618 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
619 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
620 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
621 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
622 velec = _mm_mul_ps(qq31,_mm_sub_ps(rinv31,velec));
623 felec = _mm_mul_ps(_mm_mul_ps(qq31,rinv31),_mm_sub_ps(rinvsq31,felec));
625 /* Update potential sum for this i atom from the interaction with this j atom. */
626 velecsum = _mm_add_ps(velecsum,velec);
630 /* Calculate temporary vectorial force */
631 tx = _mm_mul_ps(fscal,dx31);
632 ty = _mm_mul_ps(fscal,dy31);
633 tz = _mm_mul_ps(fscal,dz31);
635 /* Update vectorial force */
636 fix3 = _mm_add_ps(fix3,tx);
637 fiy3 = _mm_add_ps(fiy3,ty);
638 fiz3 = _mm_add_ps(fiz3,tz);
640 fjx1 = _mm_add_ps(fjx1,tx);
641 fjy1 = _mm_add_ps(fjy1,ty);
642 fjz1 = _mm_add_ps(fjz1,tz);
644 /**************************
645 * CALCULATE INTERACTIONS *
646 **************************/
648 r32 = _mm_mul_ps(rsq32,rinv32);
650 /* EWALD ELECTROSTATICS */
652 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
653 ewrt = _mm_mul_ps(r32,ewtabscale);
654 ewitab = _mm_cvttps_epi32(ewrt);
655 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
656 ewitab = _mm_slli_epi32(ewitab,2);
657 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
658 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
659 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
660 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
661 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
662 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
663 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
664 velec = _mm_mul_ps(qq32,_mm_sub_ps(rinv32,velec));
665 felec = _mm_mul_ps(_mm_mul_ps(qq32,rinv32),_mm_sub_ps(rinvsq32,felec));
667 /* Update potential sum for this i atom from the interaction with this j atom. */
668 velecsum = _mm_add_ps(velecsum,velec);
672 /* Calculate temporary vectorial force */
673 tx = _mm_mul_ps(fscal,dx32);
674 ty = _mm_mul_ps(fscal,dy32);
675 tz = _mm_mul_ps(fscal,dz32);
677 /* Update vectorial force */
678 fix3 = _mm_add_ps(fix3,tx);
679 fiy3 = _mm_add_ps(fiy3,ty);
680 fiz3 = _mm_add_ps(fiz3,tz);
682 fjx2 = _mm_add_ps(fjx2,tx);
683 fjy2 = _mm_add_ps(fjy2,ty);
684 fjz2 = _mm_add_ps(fjz2,tz);
686 /**************************
687 * CALCULATE INTERACTIONS *
688 **************************/
690 r33 = _mm_mul_ps(rsq33,rinv33);
692 /* EWALD ELECTROSTATICS */
694 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
695 ewrt = _mm_mul_ps(r33,ewtabscale);
696 ewitab = _mm_cvttps_epi32(ewrt);
697 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
698 ewitab = _mm_slli_epi32(ewitab,2);
699 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
700 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
701 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
702 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
703 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
704 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
705 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
706 velec = _mm_mul_ps(qq33,_mm_sub_ps(rinv33,velec));
707 felec = _mm_mul_ps(_mm_mul_ps(qq33,rinv33),_mm_sub_ps(rinvsq33,felec));
709 /* Update potential sum for this i atom from the interaction with this j atom. */
710 velecsum = _mm_add_ps(velecsum,velec);
714 /* Calculate temporary vectorial force */
715 tx = _mm_mul_ps(fscal,dx33);
716 ty = _mm_mul_ps(fscal,dy33);
717 tz = _mm_mul_ps(fscal,dz33);
719 /* Update vectorial force */
720 fix3 = _mm_add_ps(fix3,tx);
721 fiy3 = _mm_add_ps(fiy3,ty);
722 fiz3 = _mm_add_ps(fiz3,tz);
724 fjx3 = _mm_add_ps(fjx3,tx);
725 fjy3 = _mm_add_ps(fjy3,ty);
726 fjz3 = _mm_add_ps(fjz3,tz);
728 fjptrA = f+j_coord_offsetA;
729 fjptrB = f+j_coord_offsetB;
730 fjptrC = f+j_coord_offsetC;
731 fjptrD = f+j_coord_offsetD;
733 gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
734 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
735 fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
737 /* Inner loop uses 404 flops */
743 /* Get j neighbor index, and coordinate index */
744 jnrlistA = jjnr[jidx];
745 jnrlistB = jjnr[jidx+1];
746 jnrlistC = jjnr[jidx+2];
747 jnrlistD = jjnr[jidx+3];
748 /* Sign of each element will be negative for non-real atoms.
749 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
750 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
752 dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
753 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
754 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
755 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
756 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
757 j_coord_offsetA = DIM*jnrA;
758 j_coord_offsetB = DIM*jnrB;
759 j_coord_offsetC = DIM*jnrC;
760 j_coord_offsetD = DIM*jnrD;
762 /* load j atom coordinates */
763 gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
764 x+j_coord_offsetC,x+j_coord_offsetD,
765 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
766 &jy2,&jz2,&jx3,&jy3,&jz3);
768 /* Calculate displacement vector */
769 dx00 = _mm_sub_ps(ix0,jx0);
770 dy00 = _mm_sub_ps(iy0,jy0);
771 dz00 = _mm_sub_ps(iz0,jz0);
772 dx11 = _mm_sub_ps(ix1,jx1);
773 dy11 = _mm_sub_ps(iy1,jy1);
774 dz11 = _mm_sub_ps(iz1,jz1);
775 dx12 = _mm_sub_ps(ix1,jx2);
776 dy12 = _mm_sub_ps(iy1,jy2);
777 dz12 = _mm_sub_ps(iz1,jz2);
778 dx13 = _mm_sub_ps(ix1,jx3);
779 dy13 = _mm_sub_ps(iy1,jy3);
780 dz13 = _mm_sub_ps(iz1,jz3);
781 dx21 = _mm_sub_ps(ix2,jx1);
782 dy21 = _mm_sub_ps(iy2,jy1);
783 dz21 = _mm_sub_ps(iz2,jz1);
784 dx22 = _mm_sub_ps(ix2,jx2);
785 dy22 = _mm_sub_ps(iy2,jy2);
786 dz22 = _mm_sub_ps(iz2,jz2);
787 dx23 = _mm_sub_ps(ix2,jx3);
788 dy23 = _mm_sub_ps(iy2,jy3);
789 dz23 = _mm_sub_ps(iz2,jz3);
790 dx31 = _mm_sub_ps(ix3,jx1);
791 dy31 = _mm_sub_ps(iy3,jy1);
792 dz31 = _mm_sub_ps(iz3,jz1);
793 dx32 = _mm_sub_ps(ix3,jx2);
794 dy32 = _mm_sub_ps(iy3,jy2);
795 dz32 = _mm_sub_ps(iz3,jz2);
796 dx33 = _mm_sub_ps(ix3,jx3);
797 dy33 = _mm_sub_ps(iy3,jy3);
798 dz33 = _mm_sub_ps(iz3,jz3);
800 /* Calculate squared distance and things based on it */
801 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
802 rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
803 rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
804 rsq13 = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
805 rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
806 rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
807 rsq23 = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
808 rsq31 = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
809 rsq32 = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
810 rsq33 = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
812 rinv11 = gmx_mm_invsqrt_ps(rsq11);
813 rinv12 = gmx_mm_invsqrt_ps(rsq12);
814 rinv13 = gmx_mm_invsqrt_ps(rsq13);
815 rinv21 = gmx_mm_invsqrt_ps(rsq21);
816 rinv22 = gmx_mm_invsqrt_ps(rsq22);
817 rinv23 = gmx_mm_invsqrt_ps(rsq23);
818 rinv31 = gmx_mm_invsqrt_ps(rsq31);
819 rinv32 = gmx_mm_invsqrt_ps(rsq32);
820 rinv33 = gmx_mm_invsqrt_ps(rsq33);
822 rinvsq00 = gmx_mm_inv_ps(rsq00);
823 rinvsq11 = _mm_mul_ps(rinv11,rinv11);
824 rinvsq12 = _mm_mul_ps(rinv12,rinv12);
825 rinvsq13 = _mm_mul_ps(rinv13,rinv13);
826 rinvsq21 = _mm_mul_ps(rinv21,rinv21);
827 rinvsq22 = _mm_mul_ps(rinv22,rinv22);
828 rinvsq23 = _mm_mul_ps(rinv23,rinv23);
829 rinvsq31 = _mm_mul_ps(rinv31,rinv31);
830 rinvsq32 = _mm_mul_ps(rinv32,rinv32);
831 rinvsq33 = _mm_mul_ps(rinv33,rinv33);
833 fjx0 = _mm_setzero_ps();
834 fjy0 = _mm_setzero_ps();
835 fjz0 = _mm_setzero_ps();
836 fjx1 = _mm_setzero_ps();
837 fjy1 = _mm_setzero_ps();
838 fjz1 = _mm_setzero_ps();
839 fjx2 = _mm_setzero_ps();
840 fjy2 = _mm_setzero_ps();
841 fjz2 = _mm_setzero_ps();
842 fjx3 = _mm_setzero_ps();
843 fjy3 = _mm_setzero_ps();
844 fjz3 = _mm_setzero_ps();
846 /**************************
847 * CALCULATE INTERACTIONS *
848 **************************/
850 /* LENNARD-JONES DISPERSION/REPULSION */
852 rinvsix = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00);
853 vvdw6 = _mm_mul_ps(c6_00,rinvsix);
854 vvdw12 = _mm_mul_ps(c12_00,_mm_mul_ps(rinvsix,rinvsix));
855 vvdw = _mm_sub_ps( _mm_mul_ps(vvdw12,one_twelfth) , _mm_mul_ps(vvdw6,one_sixth) );
856 fvdw = _mm_mul_ps(_mm_sub_ps(vvdw12,vvdw6),rinvsq00);
858 /* Update potential sum for this i atom from the interaction with this j atom. */
859 vvdw = _mm_andnot_ps(dummy_mask,vvdw);
860 vvdwsum = _mm_add_ps(vvdwsum,vvdw);
864 fscal = _mm_andnot_ps(dummy_mask,fscal);
866 /* Calculate temporary vectorial force */
867 tx = _mm_mul_ps(fscal,dx00);
868 ty = _mm_mul_ps(fscal,dy00);
869 tz = _mm_mul_ps(fscal,dz00);
871 /* Update vectorial force */
872 fix0 = _mm_add_ps(fix0,tx);
873 fiy0 = _mm_add_ps(fiy0,ty);
874 fiz0 = _mm_add_ps(fiz0,tz);
876 fjx0 = _mm_add_ps(fjx0,tx);
877 fjy0 = _mm_add_ps(fjy0,ty);
878 fjz0 = _mm_add_ps(fjz0,tz);
880 /**************************
881 * CALCULATE INTERACTIONS *
882 **************************/
884 r11 = _mm_mul_ps(rsq11,rinv11);
885 r11 = _mm_andnot_ps(dummy_mask,r11);
887 /* EWALD ELECTROSTATICS */
889 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
890 ewrt = _mm_mul_ps(r11,ewtabscale);
891 ewitab = _mm_cvttps_epi32(ewrt);
892 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
893 ewitab = _mm_slli_epi32(ewitab,2);
894 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
895 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
896 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
897 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
898 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
899 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
900 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
901 velec = _mm_mul_ps(qq11,_mm_sub_ps(rinv11,velec));
902 felec = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec));
904 /* Update potential sum for this i atom from the interaction with this j atom. */
905 velec = _mm_andnot_ps(dummy_mask,velec);
906 velecsum = _mm_add_ps(velecsum,velec);
910 fscal = _mm_andnot_ps(dummy_mask,fscal);
912 /* Calculate temporary vectorial force */
913 tx = _mm_mul_ps(fscal,dx11);
914 ty = _mm_mul_ps(fscal,dy11);
915 tz = _mm_mul_ps(fscal,dz11);
917 /* Update vectorial force */
918 fix1 = _mm_add_ps(fix1,tx);
919 fiy1 = _mm_add_ps(fiy1,ty);
920 fiz1 = _mm_add_ps(fiz1,tz);
922 fjx1 = _mm_add_ps(fjx1,tx);
923 fjy1 = _mm_add_ps(fjy1,ty);
924 fjz1 = _mm_add_ps(fjz1,tz);
926 /**************************
927 * CALCULATE INTERACTIONS *
928 **************************/
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_cvtepi32_ps(ewitab));
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(rinv12,velec));
948 felec = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec));
950 /* Update potential sum for this i atom from the interaction with this j atom. */
951 velec = _mm_andnot_ps(dummy_mask,velec);
952 velecsum = _mm_add_ps(velecsum,velec);
956 fscal = _mm_andnot_ps(dummy_mask,fscal);
958 /* Calculate temporary vectorial force */
959 tx = _mm_mul_ps(fscal,dx12);
960 ty = _mm_mul_ps(fscal,dy12);
961 tz = _mm_mul_ps(fscal,dz12);
963 /* Update vectorial force */
964 fix1 = _mm_add_ps(fix1,tx);
965 fiy1 = _mm_add_ps(fiy1,ty);
966 fiz1 = _mm_add_ps(fiz1,tz);
968 fjx2 = _mm_add_ps(fjx2,tx);
969 fjy2 = _mm_add_ps(fjy2,ty);
970 fjz2 = _mm_add_ps(fjz2,tz);
972 /**************************
973 * CALCULATE INTERACTIONS *
974 **************************/
976 r13 = _mm_mul_ps(rsq13,rinv13);
977 r13 = _mm_andnot_ps(dummy_mask,r13);
979 /* EWALD ELECTROSTATICS */
981 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
982 ewrt = _mm_mul_ps(r13,ewtabscale);
983 ewitab = _mm_cvttps_epi32(ewrt);
984 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
985 ewitab = _mm_slli_epi32(ewitab,2);
986 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
987 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
988 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
989 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
990 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
991 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
992 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
993 velec = _mm_mul_ps(qq13,_mm_sub_ps(rinv13,velec));
994 felec = _mm_mul_ps(_mm_mul_ps(qq13,rinv13),_mm_sub_ps(rinvsq13,felec));
996 /* Update potential sum for this i atom from the interaction with this j atom. */
997 velec = _mm_andnot_ps(dummy_mask,velec);
998 velecsum = _mm_add_ps(velecsum,velec);
1002 fscal = _mm_andnot_ps(dummy_mask,fscal);
1004 /* Calculate temporary vectorial force */
1005 tx = _mm_mul_ps(fscal,dx13);
1006 ty = _mm_mul_ps(fscal,dy13);
1007 tz = _mm_mul_ps(fscal,dz13);
1009 /* Update vectorial force */
1010 fix1 = _mm_add_ps(fix1,tx);
1011 fiy1 = _mm_add_ps(fiy1,ty);
1012 fiz1 = _mm_add_ps(fiz1,tz);
1014 fjx3 = _mm_add_ps(fjx3,tx);
1015 fjy3 = _mm_add_ps(fjy3,ty);
1016 fjz3 = _mm_add_ps(fjz3,tz);
1018 /**************************
1019 * CALCULATE INTERACTIONS *
1020 **************************/
1022 r21 = _mm_mul_ps(rsq21,rinv21);
1023 r21 = _mm_andnot_ps(dummy_mask,r21);
1025 /* EWALD ELECTROSTATICS */
1027 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1028 ewrt = _mm_mul_ps(r21,ewtabscale);
1029 ewitab = _mm_cvttps_epi32(ewrt);
1030 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1031 ewitab = _mm_slli_epi32(ewitab,2);
1032 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1033 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1034 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1035 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1036 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1037 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1038 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1039 velec = _mm_mul_ps(qq21,_mm_sub_ps(rinv21,velec));
1040 felec = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec));
1042 /* Update potential sum for this i atom from the interaction with this j atom. */
1043 velec = _mm_andnot_ps(dummy_mask,velec);
1044 velecsum = _mm_add_ps(velecsum,velec);
1048 fscal = _mm_andnot_ps(dummy_mask,fscal);
1050 /* Calculate temporary vectorial force */
1051 tx = _mm_mul_ps(fscal,dx21);
1052 ty = _mm_mul_ps(fscal,dy21);
1053 tz = _mm_mul_ps(fscal,dz21);
1055 /* Update vectorial force */
1056 fix2 = _mm_add_ps(fix2,tx);
1057 fiy2 = _mm_add_ps(fiy2,ty);
1058 fiz2 = _mm_add_ps(fiz2,tz);
1060 fjx1 = _mm_add_ps(fjx1,tx);
1061 fjy1 = _mm_add_ps(fjy1,ty);
1062 fjz1 = _mm_add_ps(fjz1,tz);
1064 /**************************
1065 * CALCULATE INTERACTIONS *
1066 **************************/
1068 r22 = _mm_mul_ps(rsq22,rinv22);
1069 r22 = _mm_andnot_ps(dummy_mask,r22);
1071 /* EWALD ELECTROSTATICS */
1073 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1074 ewrt = _mm_mul_ps(r22,ewtabscale);
1075 ewitab = _mm_cvttps_epi32(ewrt);
1076 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1077 ewitab = _mm_slli_epi32(ewitab,2);
1078 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1079 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1080 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1081 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1082 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1083 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1084 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1085 velec = _mm_mul_ps(qq22,_mm_sub_ps(rinv22,velec));
1086 felec = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec));
1088 /* Update potential sum for this i atom from the interaction with this j atom. */
1089 velec = _mm_andnot_ps(dummy_mask,velec);
1090 velecsum = _mm_add_ps(velecsum,velec);
1094 fscal = _mm_andnot_ps(dummy_mask,fscal);
1096 /* Calculate temporary vectorial force */
1097 tx = _mm_mul_ps(fscal,dx22);
1098 ty = _mm_mul_ps(fscal,dy22);
1099 tz = _mm_mul_ps(fscal,dz22);
1101 /* Update vectorial force */
1102 fix2 = _mm_add_ps(fix2,tx);
1103 fiy2 = _mm_add_ps(fiy2,ty);
1104 fiz2 = _mm_add_ps(fiz2,tz);
1106 fjx2 = _mm_add_ps(fjx2,tx);
1107 fjy2 = _mm_add_ps(fjy2,ty);
1108 fjz2 = _mm_add_ps(fjz2,tz);
1110 /**************************
1111 * CALCULATE INTERACTIONS *
1112 **************************/
1114 r23 = _mm_mul_ps(rsq23,rinv23);
1115 r23 = _mm_andnot_ps(dummy_mask,r23);
1117 /* EWALD ELECTROSTATICS */
1119 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1120 ewrt = _mm_mul_ps(r23,ewtabscale);
1121 ewitab = _mm_cvttps_epi32(ewrt);
1122 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1123 ewitab = _mm_slli_epi32(ewitab,2);
1124 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1125 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1126 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1127 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1128 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1129 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1130 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1131 velec = _mm_mul_ps(qq23,_mm_sub_ps(rinv23,velec));
1132 felec = _mm_mul_ps(_mm_mul_ps(qq23,rinv23),_mm_sub_ps(rinvsq23,felec));
1134 /* Update potential sum for this i atom from the interaction with this j atom. */
1135 velec = _mm_andnot_ps(dummy_mask,velec);
1136 velecsum = _mm_add_ps(velecsum,velec);
1140 fscal = _mm_andnot_ps(dummy_mask,fscal);
1142 /* Calculate temporary vectorial force */
1143 tx = _mm_mul_ps(fscal,dx23);
1144 ty = _mm_mul_ps(fscal,dy23);
1145 tz = _mm_mul_ps(fscal,dz23);
1147 /* Update vectorial force */
1148 fix2 = _mm_add_ps(fix2,tx);
1149 fiy2 = _mm_add_ps(fiy2,ty);
1150 fiz2 = _mm_add_ps(fiz2,tz);
1152 fjx3 = _mm_add_ps(fjx3,tx);
1153 fjy3 = _mm_add_ps(fjy3,ty);
1154 fjz3 = _mm_add_ps(fjz3,tz);
1156 /**************************
1157 * CALCULATE INTERACTIONS *
1158 **************************/
1160 r31 = _mm_mul_ps(rsq31,rinv31);
1161 r31 = _mm_andnot_ps(dummy_mask,r31);
1163 /* EWALD ELECTROSTATICS */
1165 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1166 ewrt = _mm_mul_ps(r31,ewtabscale);
1167 ewitab = _mm_cvttps_epi32(ewrt);
1168 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1169 ewitab = _mm_slli_epi32(ewitab,2);
1170 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1171 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1172 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1173 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1174 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1175 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1176 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1177 velec = _mm_mul_ps(qq31,_mm_sub_ps(rinv31,velec));
1178 felec = _mm_mul_ps(_mm_mul_ps(qq31,rinv31),_mm_sub_ps(rinvsq31,felec));
1180 /* Update potential sum for this i atom from the interaction with this j atom. */
1181 velec = _mm_andnot_ps(dummy_mask,velec);
1182 velecsum = _mm_add_ps(velecsum,velec);
1186 fscal = _mm_andnot_ps(dummy_mask,fscal);
1188 /* Calculate temporary vectorial force */
1189 tx = _mm_mul_ps(fscal,dx31);
1190 ty = _mm_mul_ps(fscal,dy31);
1191 tz = _mm_mul_ps(fscal,dz31);
1193 /* Update vectorial force */
1194 fix3 = _mm_add_ps(fix3,tx);
1195 fiy3 = _mm_add_ps(fiy3,ty);
1196 fiz3 = _mm_add_ps(fiz3,tz);
1198 fjx1 = _mm_add_ps(fjx1,tx);
1199 fjy1 = _mm_add_ps(fjy1,ty);
1200 fjz1 = _mm_add_ps(fjz1,tz);
1202 /**************************
1203 * CALCULATE INTERACTIONS *
1204 **************************/
1206 r32 = _mm_mul_ps(rsq32,rinv32);
1207 r32 = _mm_andnot_ps(dummy_mask,r32);
1209 /* EWALD ELECTROSTATICS */
1211 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1212 ewrt = _mm_mul_ps(r32,ewtabscale);
1213 ewitab = _mm_cvttps_epi32(ewrt);
1214 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1215 ewitab = _mm_slli_epi32(ewitab,2);
1216 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1217 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1218 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1219 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1220 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1221 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1222 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1223 velec = _mm_mul_ps(qq32,_mm_sub_ps(rinv32,velec));
1224 felec = _mm_mul_ps(_mm_mul_ps(qq32,rinv32),_mm_sub_ps(rinvsq32,felec));
1226 /* Update potential sum for this i atom from the interaction with this j atom. */
1227 velec = _mm_andnot_ps(dummy_mask,velec);
1228 velecsum = _mm_add_ps(velecsum,velec);
1232 fscal = _mm_andnot_ps(dummy_mask,fscal);
1234 /* Calculate temporary vectorial force */
1235 tx = _mm_mul_ps(fscal,dx32);
1236 ty = _mm_mul_ps(fscal,dy32);
1237 tz = _mm_mul_ps(fscal,dz32);
1239 /* Update vectorial force */
1240 fix3 = _mm_add_ps(fix3,tx);
1241 fiy3 = _mm_add_ps(fiy3,ty);
1242 fiz3 = _mm_add_ps(fiz3,tz);
1244 fjx2 = _mm_add_ps(fjx2,tx);
1245 fjy2 = _mm_add_ps(fjy2,ty);
1246 fjz2 = _mm_add_ps(fjz2,tz);
1248 /**************************
1249 * CALCULATE INTERACTIONS *
1250 **************************/
1252 r33 = _mm_mul_ps(rsq33,rinv33);
1253 r33 = _mm_andnot_ps(dummy_mask,r33);
1255 /* EWALD ELECTROSTATICS */
1257 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1258 ewrt = _mm_mul_ps(r33,ewtabscale);
1259 ewitab = _mm_cvttps_epi32(ewrt);
1260 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1261 ewitab = _mm_slli_epi32(ewitab,2);
1262 ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1263 ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1264 ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1265 ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1266 _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1267 felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1268 velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1269 velec = _mm_mul_ps(qq33,_mm_sub_ps(rinv33,velec));
1270 felec = _mm_mul_ps(_mm_mul_ps(qq33,rinv33),_mm_sub_ps(rinvsq33,felec));
1272 /* Update potential sum for this i atom from the interaction with this j atom. */
1273 velec = _mm_andnot_ps(dummy_mask,velec);
1274 velecsum = _mm_add_ps(velecsum,velec);
1278 fscal = _mm_andnot_ps(dummy_mask,fscal);
1280 /* Calculate temporary vectorial force */
1281 tx = _mm_mul_ps(fscal,dx33);
1282 ty = _mm_mul_ps(fscal,dy33);
1283 tz = _mm_mul_ps(fscal,dz33);
1285 /* Update vectorial force */
1286 fix3 = _mm_add_ps(fix3,tx);
1287 fiy3 = _mm_add_ps(fiy3,ty);
1288 fiz3 = _mm_add_ps(fiz3,tz);
1290 fjx3 = _mm_add_ps(fjx3,tx);
1291 fjy3 = _mm_add_ps(fjy3,ty);
1292 fjz3 = _mm_add_ps(fjz3,tz);
1294 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1295 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1296 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1297 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1299 gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
1300 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1301 fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1303 /* Inner loop uses 413 flops */
1306 /* End of innermost loop */
1308 gmx_mm_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1309 f+i_coord_offset,fshift+i_shift_offset);
1312 /* Update potential energies */
1313 gmx_mm_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
1314 gmx_mm_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
1316 /* Increment number of inner iterations */
1317 inneriter += j_index_end - j_index_start;
1319 /* Outer loop uses 26 flops */
1322 /* Increment number of outer iterations */
1325 /* Update outer/inner flops */
1327 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*413);
1330 * Gromacs nonbonded kernel: nb_kernel_ElecEw_VdwLJ_GeomW4W4_F_sse2_single
1331 * Electrostatics interaction: Ewald
1332 * VdW interaction: LennardJones
1333 * Geometry: Water4-Water4
1334 * Calculate force/pot: Force
1337 nb_kernel_ElecEw_VdwLJ_GeomW4W4_F_sse2_single
1338 (t_nblist * gmx_restrict nlist,
1339 rvec * gmx_restrict xx,
1340 rvec * gmx_restrict ff,
1341 t_forcerec * gmx_restrict fr,
1342 t_mdatoms * gmx_restrict mdatoms,
1343 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1344 t_nrnb * gmx_restrict nrnb)
1346 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1347 * just 0 for non-waters.
1348 * Suffixes A,B,C,D refer to j loop unrolling done with SSE, e.g. for the four different
1349 * jnr indices corresponding to data put in the four positions in the SIMD register.
1351 int i_shift_offset,i_coord_offset,outeriter,inneriter;
1352 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1353 int jnrA,jnrB,jnrC,jnrD;
1354 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1355 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1356 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
1357 real rcutoff_scalar;
1358 real *shiftvec,*fshift,*x,*f;
1359 real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
1360 real scratch[4*DIM];
1361 __m128 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1363 __m128 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1365 __m128 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1367 __m128 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1369 __m128 ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1370 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
1371 __m128 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1372 int vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
1373 __m128 jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1374 int vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
1375 __m128 jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1376 int vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
1377 __m128 jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1378 __m128 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1379 __m128 dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1380 __m128 dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1381 __m128 dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1382 __m128 dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1383 __m128 dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1384 __m128 dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1385 __m128 dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1386 __m128 dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1387 __m128 dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1388 __m128 velec,felec,velecsum,facel,crf,krf,krf2;
1391 __m128 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1394 __m128 one_sixth = _mm_set1_ps(1.0/6.0);
1395 __m128 one_twelfth = _mm_set1_ps(1.0/12.0);
1397 __m128 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1399 __m128 dummy_mask,cutoff_mask;
1400 __m128 signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
1401 __m128 one = _mm_set1_ps(1.0);
1402 __m128 two = _mm_set1_ps(2.0);
1408 jindex = nlist->jindex;
1410 shiftidx = nlist->shift;
1412 shiftvec = fr->shift_vec[0];
1413 fshift = fr->fshift[0];
1414 facel = _mm_set1_ps(fr->epsfac);
1415 charge = mdatoms->chargeA;
1416 nvdwtype = fr->ntype;
1417 vdwparam = fr->nbfp;
1418 vdwtype = mdatoms->typeA;
1420 sh_ewald = _mm_set1_ps(fr->ic->sh_ewald);
1421 ewtab = fr->ic->tabq_coul_F;
1422 ewtabscale = _mm_set1_ps(fr->ic->tabq_scale);
1423 ewtabhalfspace = _mm_set1_ps(0.5/fr->ic->tabq_scale);
1425 /* Setup water-specific parameters */
1426 inr = nlist->iinr[0];
1427 iq1 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1]));
1428 iq2 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2]));
1429 iq3 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+3]));
1430 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
1432 jq1 = _mm_set1_ps(charge[inr+1]);
1433 jq2 = _mm_set1_ps(charge[inr+2]);
1434 jq3 = _mm_set1_ps(charge[inr+3]);
1435 vdwjidx0A = 2*vdwtype[inr+0];
1436 c6_00 = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A]);
1437 c12_00 = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A+1]);
1438 qq11 = _mm_mul_ps(iq1,jq1);
1439 qq12 = _mm_mul_ps(iq1,jq2);
1440 qq13 = _mm_mul_ps(iq1,jq3);
1441 qq21 = _mm_mul_ps(iq2,jq1);
1442 qq22 = _mm_mul_ps(iq2,jq2);
1443 qq23 = _mm_mul_ps(iq2,jq3);
1444 qq31 = _mm_mul_ps(iq3,jq1);
1445 qq32 = _mm_mul_ps(iq3,jq2);
1446 qq33 = _mm_mul_ps(iq3,jq3);
1448 /* Avoid stupid compiler warnings */
1449 jnrA = jnrB = jnrC = jnrD = 0;
1450 j_coord_offsetA = 0;
1451 j_coord_offsetB = 0;
1452 j_coord_offsetC = 0;
1453 j_coord_offsetD = 0;
1458 for(iidx=0;iidx<4*DIM;iidx++)
1460 scratch[iidx] = 0.0;
1463 /* Start outer loop over neighborlists */
1464 for(iidx=0; iidx<nri; iidx++)
1466 /* Load shift vector for this list */
1467 i_shift_offset = DIM*shiftidx[iidx];
1469 /* Load limits for loop over neighbors */
1470 j_index_start = jindex[iidx];
1471 j_index_end = jindex[iidx+1];
1473 /* Get outer coordinate index */
1475 i_coord_offset = DIM*inr;
1477 /* Load i particle coords and add shift vector */
1478 gmx_mm_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1479 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1481 fix0 = _mm_setzero_ps();
1482 fiy0 = _mm_setzero_ps();
1483 fiz0 = _mm_setzero_ps();
1484 fix1 = _mm_setzero_ps();
1485 fiy1 = _mm_setzero_ps();
1486 fiz1 = _mm_setzero_ps();
1487 fix2 = _mm_setzero_ps();
1488 fiy2 = _mm_setzero_ps();
1489 fiz2 = _mm_setzero_ps();
1490 fix3 = _mm_setzero_ps();
1491 fiy3 = _mm_setzero_ps();
1492 fiz3 = _mm_setzero_ps();
1494 /* Start inner kernel loop */
1495 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
1498 /* Get j neighbor index, and coordinate index */
1500 jnrB = jjnr[jidx+1];
1501 jnrC = jjnr[jidx+2];
1502 jnrD = jjnr[jidx+3];
1503 j_coord_offsetA = DIM*jnrA;
1504 j_coord_offsetB = DIM*jnrB;
1505 j_coord_offsetC = DIM*jnrC;
1506 j_coord_offsetD = DIM*jnrD;
1508 /* load j atom coordinates */
1509 gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1510 x+j_coord_offsetC,x+j_coord_offsetD,
1511 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1512 &jy2,&jz2,&jx3,&jy3,&jz3);
1514 /* Calculate displacement vector */
1515 dx00 = _mm_sub_ps(ix0,jx0);
1516 dy00 = _mm_sub_ps(iy0,jy0);
1517 dz00 = _mm_sub_ps(iz0,jz0);
1518 dx11 = _mm_sub_ps(ix1,jx1);
1519 dy11 = _mm_sub_ps(iy1,jy1);
1520 dz11 = _mm_sub_ps(iz1,jz1);
1521 dx12 = _mm_sub_ps(ix1,jx2);
1522 dy12 = _mm_sub_ps(iy1,jy2);
1523 dz12 = _mm_sub_ps(iz1,jz2);
1524 dx13 = _mm_sub_ps(ix1,jx3);
1525 dy13 = _mm_sub_ps(iy1,jy3);
1526 dz13 = _mm_sub_ps(iz1,jz3);
1527 dx21 = _mm_sub_ps(ix2,jx1);
1528 dy21 = _mm_sub_ps(iy2,jy1);
1529 dz21 = _mm_sub_ps(iz2,jz1);
1530 dx22 = _mm_sub_ps(ix2,jx2);
1531 dy22 = _mm_sub_ps(iy2,jy2);
1532 dz22 = _mm_sub_ps(iz2,jz2);
1533 dx23 = _mm_sub_ps(ix2,jx3);
1534 dy23 = _mm_sub_ps(iy2,jy3);
1535 dz23 = _mm_sub_ps(iz2,jz3);
1536 dx31 = _mm_sub_ps(ix3,jx1);
1537 dy31 = _mm_sub_ps(iy3,jy1);
1538 dz31 = _mm_sub_ps(iz3,jz1);
1539 dx32 = _mm_sub_ps(ix3,jx2);
1540 dy32 = _mm_sub_ps(iy3,jy2);
1541 dz32 = _mm_sub_ps(iz3,jz2);
1542 dx33 = _mm_sub_ps(ix3,jx3);
1543 dy33 = _mm_sub_ps(iy3,jy3);
1544 dz33 = _mm_sub_ps(iz3,jz3);
1546 /* Calculate squared distance and things based on it */
1547 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
1548 rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
1549 rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
1550 rsq13 = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
1551 rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
1552 rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
1553 rsq23 = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
1554 rsq31 = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
1555 rsq32 = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
1556 rsq33 = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
1558 rinv11 = gmx_mm_invsqrt_ps(rsq11);
1559 rinv12 = gmx_mm_invsqrt_ps(rsq12);
1560 rinv13 = gmx_mm_invsqrt_ps(rsq13);
1561 rinv21 = gmx_mm_invsqrt_ps(rsq21);
1562 rinv22 = gmx_mm_invsqrt_ps(rsq22);
1563 rinv23 = gmx_mm_invsqrt_ps(rsq23);
1564 rinv31 = gmx_mm_invsqrt_ps(rsq31);
1565 rinv32 = gmx_mm_invsqrt_ps(rsq32);
1566 rinv33 = gmx_mm_invsqrt_ps(rsq33);
1568 rinvsq00 = gmx_mm_inv_ps(rsq00);
1569 rinvsq11 = _mm_mul_ps(rinv11,rinv11);
1570 rinvsq12 = _mm_mul_ps(rinv12,rinv12);
1571 rinvsq13 = _mm_mul_ps(rinv13,rinv13);
1572 rinvsq21 = _mm_mul_ps(rinv21,rinv21);
1573 rinvsq22 = _mm_mul_ps(rinv22,rinv22);
1574 rinvsq23 = _mm_mul_ps(rinv23,rinv23);
1575 rinvsq31 = _mm_mul_ps(rinv31,rinv31);
1576 rinvsq32 = _mm_mul_ps(rinv32,rinv32);
1577 rinvsq33 = _mm_mul_ps(rinv33,rinv33);
1579 fjx0 = _mm_setzero_ps();
1580 fjy0 = _mm_setzero_ps();
1581 fjz0 = _mm_setzero_ps();
1582 fjx1 = _mm_setzero_ps();
1583 fjy1 = _mm_setzero_ps();
1584 fjz1 = _mm_setzero_ps();
1585 fjx2 = _mm_setzero_ps();
1586 fjy2 = _mm_setzero_ps();
1587 fjz2 = _mm_setzero_ps();
1588 fjx3 = _mm_setzero_ps();
1589 fjy3 = _mm_setzero_ps();
1590 fjz3 = _mm_setzero_ps();
1592 /**************************
1593 * CALCULATE INTERACTIONS *
1594 **************************/
1596 /* LENNARD-JONES DISPERSION/REPULSION */
1598 rinvsix = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00);
1599 fvdw = _mm_mul_ps(_mm_sub_ps(_mm_mul_ps(c12_00,rinvsix),c6_00),_mm_mul_ps(rinvsix,rinvsq00));
1603 /* Calculate temporary vectorial force */
1604 tx = _mm_mul_ps(fscal,dx00);
1605 ty = _mm_mul_ps(fscal,dy00);
1606 tz = _mm_mul_ps(fscal,dz00);
1608 /* Update vectorial force */
1609 fix0 = _mm_add_ps(fix0,tx);
1610 fiy0 = _mm_add_ps(fiy0,ty);
1611 fiz0 = _mm_add_ps(fiz0,tz);
1613 fjx0 = _mm_add_ps(fjx0,tx);
1614 fjy0 = _mm_add_ps(fjy0,ty);
1615 fjz0 = _mm_add_ps(fjz0,tz);
1617 /**************************
1618 * CALCULATE INTERACTIONS *
1619 **************************/
1621 r11 = _mm_mul_ps(rsq11,rinv11);
1623 /* EWALD ELECTROSTATICS */
1625 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1626 ewrt = _mm_mul_ps(r11,ewtabscale);
1627 ewitab = _mm_cvttps_epi32(ewrt);
1628 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1629 gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1630 ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
1632 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1633 felec = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec));
1637 /* Calculate temporary vectorial force */
1638 tx = _mm_mul_ps(fscal,dx11);
1639 ty = _mm_mul_ps(fscal,dy11);
1640 tz = _mm_mul_ps(fscal,dz11);
1642 /* Update vectorial force */
1643 fix1 = _mm_add_ps(fix1,tx);
1644 fiy1 = _mm_add_ps(fiy1,ty);
1645 fiz1 = _mm_add_ps(fiz1,tz);
1647 fjx1 = _mm_add_ps(fjx1,tx);
1648 fjy1 = _mm_add_ps(fjy1,ty);
1649 fjz1 = _mm_add_ps(fjz1,tz);
1651 /**************************
1652 * CALCULATE INTERACTIONS *
1653 **************************/
1655 r12 = _mm_mul_ps(rsq12,rinv12);
1657 /* EWALD ELECTROSTATICS */
1659 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1660 ewrt = _mm_mul_ps(r12,ewtabscale);
1661 ewitab = _mm_cvttps_epi32(ewrt);
1662 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1663 gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1664 ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
1666 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1667 felec = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec));
1671 /* Calculate temporary vectorial force */
1672 tx = _mm_mul_ps(fscal,dx12);
1673 ty = _mm_mul_ps(fscal,dy12);
1674 tz = _mm_mul_ps(fscal,dz12);
1676 /* Update vectorial force */
1677 fix1 = _mm_add_ps(fix1,tx);
1678 fiy1 = _mm_add_ps(fiy1,ty);
1679 fiz1 = _mm_add_ps(fiz1,tz);
1681 fjx2 = _mm_add_ps(fjx2,tx);
1682 fjy2 = _mm_add_ps(fjy2,ty);
1683 fjz2 = _mm_add_ps(fjz2,tz);
1685 /**************************
1686 * CALCULATE INTERACTIONS *
1687 **************************/
1689 r13 = _mm_mul_ps(rsq13,rinv13);
1691 /* EWALD ELECTROSTATICS */
1693 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1694 ewrt = _mm_mul_ps(r13,ewtabscale);
1695 ewitab = _mm_cvttps_epi32(ewrt);
1696 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1697 gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1698 ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
1700 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1701 felec = _mm_mul_ps(_mm_mul_ps(qq13,rinv13),_mm_sub_ps(rinvsq13,felec));
1705 /* Calculate temporary vectorial force */
1706 tx = _mm_mul_ps(fscal,dx13);
1707 ty = _mm_mul_ps(fscal,dy13);
1708 tz = _mm_mul_ps(fscal,dz13);
1710 /* Update vectorial force */
1711 fix1 = _mm_add_ps(fix1,tx);
1712 fiy1 = _mm_add_ps(fiy1,ty);
1713 fiz1 = _mm_add_ps(fiz1,tz);
1715 fjx3 = _mm_add_ps(fjx3,tx);
1716 fjy3 = _mm_add_ps(fjy3,ty);
1717 fjz3 = _mm_add_ps(fjz3,tz);
1719 /**************************
1720 * CALCULATE INTERACTIONS *
1721 **************************/
1723 r21 = _mm_mul_ps(rsq21,rinv21);
1725 /* EWALD ELECTROSTATICS */
1727 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1728 ewrt = _mm_mul_ps(r21,ewtabscale);
1729 ewitab = _mm_cvttps_epi32(ewrt);
1730 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1731 gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1732 ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
1734 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1735 felec = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec));
1739 /* Calculate temporary vectorial force */
1740 tx = _mm_mul_ps(fscal,dx21);
1741 ty = _mm_mul_ps(fscal,dy21);
1742 tz = _mm_mul_ps(fscal,dz21);
1744 /* Update vectorial force */
1745 fix2 = _mm_add_ps(fix2,tx);
1746 fiy2 = _mm_add_ps(fiy2,ty);
1747 fiz2 = _mm_add_ps(fiz2,tz);
1749 fjx1 = _mm_add_ps(fjx1,tx);
1750 fjy1 = _mm_add_ps(fjy1,ty);
1751 fjz1 = _mm_add_ps(fjz1,tz);
1753 /**************************
1754 * CALCULATE INTERACTIONS *
1755 **************************/
1757 r22 = _mm_mul_ps(rsq22,rinv22);
1759 /* EWALD ELECTROSTATICS */
1761 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1762 ewrt = _mm_mul_ps(r22,ewtabscale);
1763 ewitab = _mm_cvttps_epi32(ewrt);
1764 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1765 gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1766 ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
1768 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1769 felec = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec));
1773 /* Calculate temporary vectorial force */
1774 tx = _mm_mul_ps(fscal,dx22);
1775 ty = _mm_mul_ps(fscal,dy22);
1776 tz = _mm_mul_ps(fscal,dz22);
1778 /* Update vectorial force */
1779 fix2 = _mm_add_ps(fix2,tx);
1780 fiy2 = _mm_add_ps(fiy2,ty);
1781 fiz2 = _mm_add_ps(fiz2,tz);
1783 fjx2 = _mm_add_ps(fjx2,tx);
1784 fjy2 = _mm_add_ps(fjy2,ty);
1785 fjz2 = _mm_add_ps(fjz2,tz);
1787 /**************************
1788 * CALCULATE INTERACTIONS *
1789 **************************/
1791 r23 = _mm_mul_ps(rsq23,rinv23);
1793 /* EWALD ELECTROSTATICS */
1795 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1796 ewrt = _mm_mul_ps(r23,ewtabscale);
1797 ewitab = _mm_cvttps_epi32(ewrt);
1798 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1799 gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1800 ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
1802 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1803 felec = _mm_mul_ps(_mm_mul_ps(qq23,rinv23),_mm_sub_ps(rinvsq23,felec));
1807 /* Calculate temporary vectorial force */
1808 tx = _mm_mul_ps(fscal,dx23);
1809 ty = _mm_mul_ps(fscal,dy23);
1810 tz = _mm_mul_ps(fscal,dz23);
1812 /* Update vectorial force */
1813 fix2 = _mm_add_ps(fix2,tx);
1814 fiy2 = _mm_add_ps(fiy2,ty);
1815 fiz2 = _mm_add_ps(fiz2,tz);
1817 fjx3 = _mm_add_ps(fjx3,tx);
1818 fjy3 = _mm_add_ps(fjy3,ty);
1819 fjz3 = _mm_add_ps(fjz3,tz);
1821 /**************************
1822 * CALCULATE INTERACTIONS *
1823 **************************/
1825 r31 = _mm_mul_ps(rsq31,rinv31);
1827 /* EWALD ELECTROSTATICS */
1829 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1830 ewrt = _mm_mul_ps(r31,ewtabscale);
1831 ewitab = _mm_cvttps_epi32(ewrt);
1832 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1833 gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1834 ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
1836 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1837 felec = _mm_mul_ps(_mm_mul_ps(qq31,rinv31),_mm_sub_ps(rinvsq31,felec));
1841 /* Calculate temporary vectorial force */
1842 tx = _mm_mul_ps(fscal,dx31);
1843 ty = _mm_mul_ps(fscal,dy31);
1844 tz = _mm_mul_ps(fscal,dz31);
1846 /* Update vectorial force */
1847 fix3 = _mm_add_ps(fix3,tx);
1848 fiy3 = _mm_add_ps(fiy3,ty);
1849 fiz3 = _mm_add_ps(fiz3,tz);
1851 fjx1 = _mm_add_ps(fjx1,tx);
1852 fjy1 = _mm_add_ps(fjy1,ty);
1853 fjz1 = _mm_add_ps(fjz1,tz);
1855 /**************************
1856 * CALCULATE INTERACTIONS *
1857 **************************/
1859 r32 = _mm_mul_ps(rsq32,rinv32);
1861 /* EWALD ELECTROSTATICS */
1863 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1864 ewrt = _mm_mul_ps(r32,ewtabscale);
1865 ewitab = _mm_cvttps_epi32(ewrt);
1866 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1867 gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1868 ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
1870 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1871 felec = _mm_mul_ps(_mm_mul_ps(qq32,rinv32),_mm_sub_ps(rinvsq32,felec));
1875 /* Calculate temporary vectorial force */
1876 tx = _mm_mul_ps(fscal,dx32);
1877 ty = _mm_mul_ps(fscal,dy32);
1878 tz = _mm_mul_ps(fscal,dz32);
1880 /* Update vectorial force */
1881 fix3 = _mm_add_ps(fix3,tx);
1882 fiy3 = _mm_add_ps(fiy3,ty);
1883 fiz3 = _mm_add_ps(fiz3,tz);
1885 fjx2 = _mm_add_ps(fjx2,tx);
1886 fjy2 = _mm_add_ps(fjy2,ty);
1887 fjz2 = _mm_add_ps(fjz2,tz);
1889 /**************************
1890 * CALCULATE INTERACTIONS *
1891 **************************/
1893 r33 = _mm_mul_ps(rsq33,rinv33);
1895 /* EWALD ELECTROSTATICS */
1897 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1898 ewrt = _mm_mul_ps(r33,ewtabscale);
1899 ewitab = _mm_cvttps_epi32(ewrt);
1900 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1901 gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1902 ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
1904 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1905 felec = _mm_mul_ps(_mm_mul_ps(qq33,rinv33),_mm_sub_ps(rinvsq33,felec));
1909 /* Calculate temporary vectorial force */
1910 tx = _mm_mul_ps(fscal,dx33);
1911 ty = _mm_mul_ps(fscal,dy33);
1912 tz = _mm_mul_ps(fscal,dz33);
1914 /* Update vectorial force */
1915 fix3 = _mm_add_ps(fix3,tx);
1916 fiy3 = _mm_add_ps(fiy3,ty);
1917 fiz3 = _mm_add_ps(fiz3,tz);
1919 fjx3 = _mm_add_ps(fjx3,tx);
1920 fjy3 = _mm_add_ps(fjy3,ty);
1921 fjz3 = _mm_add_ps(fjz3,tz);
1923 fjptrA = f+j_coord_offsetA;
1924 fjptrB = f+j_coord_offsetB;
1925 fjptrC = f+j_coord_offsetC;
1926 fjptrD = f+j_coord_offsetD;
1928 gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
1929 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1930 fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1932 /* Inner loop uses 354 flops */
1935 if(jidx<j_index_end)
1938 /* Get j neighbor index, and coordinate index */
1939 jnrlistA = jjnr[jidx];
1940 jnrlistB = jjnr[jidx+1];
1941 jnrlistC = jjnr[jidx+2];
1942 jnrlistD = jjnr[jidx+3];
1943 /* Sign of each element will be negative for non-real atoms.
1944 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1945 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1947 dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
1948 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
1949 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
1950 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
1951 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
1952 j_coord_offsetA = DIM*jnrA;
1953 j_coord_offsetB = DIM*jnrB;
1954 j_coord_offsetC = DIM*jnrC;
1955 j_coord_offsetD = DIM*jnrD;
1957 /* load j atom coordinates */
1958 gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1959 x+j_coord_offsetC,x+j_coord_offsetD,
1960 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1961 &jy2,&jz2,&jx3,&jy3,&jz3);
1963 /* Calculate displacement vector */
1964 dx00 = _mm_sub_ps(ix0,jx0);
1965 dy00 = _mm_sub_ps(iy0,jy0);
1966 dz00 = _mm_sub_ps(iz0,jz0);
1967 dx11 = _mm_sub_ps(ix1,jx1);
1968 dy11 = _mm_sub_ps(iy1,jy1);
1969 dz11 = _mm_sub_ps(iz1,jz1);
1970 dx12 = _mm_sub_ps(ix1,jx2);
1971 dy12 = _mm_sub_ps(iy1,jy2);
1972 dz12 = _mm_sub_ps(iz1,jz2);
1973 dx13 = _mm_sub_ps(ix1,jx3);
1974 dy13 = _mm_sub_ps(iy1,jy3);
1975 dz13 = _mm_sub_ps(iz1,jz3);
1976 dx21 = _mm_sub_ps(ix2,jx1);
1977 dy21 = _mm_sub_ps(iy2,jy1);
1978 dz21 = _mm_sub_ps(iz2,jz1);
1979 dx22 = _mm_sub_ps(ix2,jx2);
1980 dy22 = _mm_sub_ps(iy2,jy2);
1981 dz22 = _mm_sub_ps(iz2,jz2);
1982 dx23 = _mm_sub_ps(ix2,jx3);
1983 dy23 = _mm_sub_ps(iy2,jy3);
1984 dz23 = _mm_sub_ps(iz2,jz3);
1985 dx31 = _mm_sub_ps(ix3,jx1);
1986 dy31 = _mm_sub_ps(iy3,jy1);
1987 dz31 = _mm_sub_ps(iz3,jz1);
1988 dx32 = _mm_sub_ps(ix3,jx2);
1989 dy32 = _mm_sub_ps(iy3,jy2);
1990 dz32 = _mm_sub_ps(iz3,jz2);
1991 dx33 = _mm_sub_ps(ix3,jx3);
1992 dy33 = _mm_sub_ps(iy3,jy3);
1993 dz33 = _mm_sub_ps(iz3,jz3);
1995 /* Calculate squared distance and things based on it */
1996 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
1997 rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
1998 rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
1999 rsq13 = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
2000 rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
2001 rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
2002 rsq23 = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
2003 rsq31 = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
2004 rsq32 = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
2005 rsq33 = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
2007 rinv11 = gmx_mm_invsqrt_ps(rsq11);
2008 rinv12 = gmx_mm_invsqrt_ps(rsq12);
2009 rinv13 = gmx_mm_invsqrt_ps(rsq13);
2010 rinv21 = gmx_mm_invsqrt_ps(rsq21);
2011 rinv22 = gmx_mm_invsqrt_ps(rsq22);
2012 rinv23 = gmx_mm_invsqrt_ps(rsq23);
2013 rinv31 = gmx_mm_invsqrt_ps(rsq31);
2014 rinv32 = gmx_mm_invsqrt_ps(rsq32);
2015 rinv33 = gmx_mm_invsqrt_ps(rsq33);
2017 rinvsq00 = gmx_mm_inv_ps(rsq00);
2018 rinvsq11 = _mm_mul_ps(rinv11,rinv11);
2019 rinvsq12 = _mm_mul_ps(rinv12,rinv12);
2020 rinvsq13 = _mm_mul_ps(rinv13,rinv13);
2021 rinvsq21 = _mm_mul_ps(rinv21,rinv21);
2022 rinvsq22 = _mm_mul_ps(rinv22,rinv22);
2023 rinvsq23 = _mm_mul_ps(rinv23,rinv23);
2024 rinvsq31 = _mm_mul_ps(rinv31,rinv31);
2025 rinvsq32 = _mm_mul_ps(rinv32,rinv32);
2026 rinvsq33 = _mm_mul_ps(rinv33,rinv33);
2028 fjx0 = _mm_setzero_ps();
2029 fjy0 = _mm_setzero_ps();
2030 fjz0 = _mm_setzero_ps();
2031 fjx1 = _mm_setzero_ps();
2032 fjy1 = _mm_setzero_ps();
2033 fjz1 = _mm_setzero_ps();
2034 fjx2 = _mm_setzero_ps();
2035 fjy2 = _mm_setzero_ps();
2036 fjz2 = _mm_setzero_ps();
2037 fjx3 = _mm_setzero_ps();
2038 fjy3 = _mm_setzero_ps();
2039 fjz3 = _mm_setzero_ps();
2041 /**************************
2042 * CALCULATE INTERACTIONS *
2043 **************************/
2045 /* LENNARD-JONES DISPERSION/REPULSION */
2047 rinvsix = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00);
2048 fvdw = _mm_mul_ps(_mm_sub_ps(_mm_mul_ps(c12_00,rinvsix),c6_00),_mm_mul_ps(rinvsix,rinvsq00));
2052 fscal = _mm_andnot_ps(dummy_mask,fscal);
2054 /* Calculate temporary vectorial force */
2055 tx = _mm_mul_ps(fscal,dx00);
2056 ty = _mm_mul_ps(fscal,dy00);
2057 tz = _mm_mul_ps(fscal,dz00);
2059 /* Update vectorial force */
2060 fix0 = _mm_add_ps(fix0,tx);
2061 fiy0 = _mm_add_ps(fiy0,ty);
2062 fiz0 = _mm_add_ps(fiz0,tz);
2064 fjx0 = _mm_add_ps(fjx0,tx);
2065 fjy0 = _mm_add_ps(fjy0,ty);
2066 fjz0 = _mm_add_ps(fjz0,tz);
2068 /**************************
2069 * CALCULATE INTERACTIONS *
2070 **************************/
2072 r11 = _mm_mul_ps(rsq11,rinv11);
2073 r11 = _mm_andnot_ps(dummy_mask,r11);
2075 /* EWALD ELECTROSTATICS */
2077 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2078 ewrt = _mm_mul_ps(r11,ewtabscale);
2079 ewitab = _mm_cvttps_epi32(ewrt);
2080 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
2081 gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
2082 ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
2084 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2085 felec = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec));
2089 fscal = _mm_andnot_ps(dummy_mask,fscal);
2091 /* Calculate temporary vectorial force */
2092 tx = _mm_mul_ps(fscal,dx11);
2093 ty = _mm_mul_ps(fscal,dy11);
2094 tz = _mm_mul_ps(fscal,dz11);
2096 /* Update vectorial force */
2097 fix1 = _mm_add_ps(fix1,tx);
2098 fiy1 = _mm_add_ps(fiy1,ty);
2099 fiz1 = _mm_add_ps(fiz1,tz);
2101 fjx1 = _mm_add_ps(fjx1,tx);
2102 fjy1 = _mm_add_ps(fjy1,ty);
2103 fjz1 = _mm_add_ps(fjz1,tz);
2105 /**************************
2106 * CALCULATE INTERACTIONS *
2107 **************************/
2109 r12 = _mm_mul_ps(rsq12,rinv12);
2110 r12 = _mm_andnot_ps(dummy_mask,r12);
2112 /* EWALD ELECTROSTATICS */
2114 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2115 ewrt = _mm_mul_ps(r12,ewtabscale);
2116 ewitab = _mm_cvttps_epi32(ewrt);
2117 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
2118 gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
2119 ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
2121 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2122 felec = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec));
2126 fscal = _mm_andnot_ps(dummy_mask,fscal);
2128 /* Calculate temporary vectorial force */
2129 tx = _mm_mul_ps(fscal,dx12);
2130 ty = _mm_mul_ps(fscal,dy12);
2131 tz = _mm_mul_ps(fscal,dz12);
2133 /* Update vectorial force */
2134 fix1 = _mm_add_ps(fix1,tx);
2135 fiy1 = _mm_add_ps(fiy1,ty);
2136 fiz1 = _mm_add_ps(fiz1,tz);
2138 fjx2 = _mm_add_ps(fjx2,tx);
2139 fjy2 = _mm_add_ps(fjy2,ty);
2140 fjz2 = _mm_add_ps(fjz2,tz);
2142 /**************************
2143 * CALCULATE INTERACTIONS *
2144 **************************/
2146 r13 = _mm_mul_ps(rsq13,rinv13);
2147 r13 = _mm_andnot_ps(dummy_mask,r13);
2149 /* EWALD ELECTROSTATICS */
2151 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2152 ewrt = _mm_mul_ps(r13,ewtabscale);
2153 ewitab = _mm_cvttps_epi32(ewrt);
2154 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
2155 gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
2156 ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
2158 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2159 felec = _mm_mul_ps(_mm_mul_ps(qq13,rinv13),_mm_sub_ps(rinvsq13,felec));
2163 fscal = _mm_andnot_ps(dummy_mask,fscal);
2165 /* Calculate temporary vectorial force */
2166 tx = _mm_mul_ps(fscal,dx13);
2167 ty = _mm_mul_ps(fscal,dy13);
2168 tz = _mm_mul_ps(fscal,dz13);
2170 /* Update vectorial force */
2171 fix1 = _mm_add_ps(fix1,tx);
2172 fiy1 = _mm_add_ps(fiy1,ty);
2173 fiz1 = _mm_add_ps(fiz1,tz);
2175 fjx3 = _mm_add_ps(fjx3,tx);
2176 fjy3 = _mm_add_ps(fjy3,ty);
2177 fjz3 = _mm_add_ps(fjz3,tz);
2179 /**************************
2180 * CALCULATE INTERACTIONS *
2181 **************************/
2183 r21 = _mm_mul_ps(rsq21,rinv21);
2184 r21 = _mm_andnot_ps(dummy_mask,r21);
2186 /* EWALD ELECTROSTATICS */
2188 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2189 ewrt = _mm_mul_ps(r21,ewtabscale);
2190 ewitab = _mm_cvttps_epi32(ewrt);
2191 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
2192 gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
2193 ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
2195 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2196 felec = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec));
2200 fscal = _mm_andnot_ps(dummy_mask,fscal);
2202 /* Calculate temporary vectorial force */
2203 tx = _mm_mul_ps(fscal,dx21);
2204 ty = _mm_mul_ps(fscal,dy21);
2205 tz = _mm_mul_ps(fscal,dz21);
2207 /* Update vectorial force */
2208 fix2 = _mm_add_ps(fix2,tx);
2209 fiy2 = _mm_add_ps(fiy2,ty);
2210 fiz2 = _mm_add_ps(fiz2,tz);
2212 fjx1 = _mm_add_ps(fjx1,tx);
2213 fjy1 = _mm_add_ps(fjy1,ty);
2214 fjz1 = _mm_add_ps(fjz1,tz);
2216 /**************************
2217 * CALCULATE INTERACTIONS *
2218 **************************/
2220 r22 = _mm_mul_ps(rsq22,rinv22);
2221 r22 = _mm_andnot_ps(dummy_mask,r22);
2223 /* EWALD ELECTROSTATICS */
2225 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2226 ewrt = _mm_mul_ps(r22,ewtabscale);
2227 ewitab = _mm_cvttps_epi32(ewrt);
2228 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
2229 gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
2230 ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
2232 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2233 felec = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec));
2237 fscal = _mm_andnot_ps(dummy_mask,fscal);
2239 /* Calculate temporary vectorial force */
2240 tx = _mm_mul_ps(fscal,dx22);
2241 ty = _mm_mul_ps(fscal,dy22);
2242 tz = _mm_mul_ps(fscal,dz22);
2244 /* Update vectorial force */
2245 fix2 = _mm_add_ps(fix2,tx);
2246 fiy2 = _mm_add_ps(fiy2,ty);
2247 fiz2 = _mm_add_ps(fiz2,tz);
2249 fjx2 = _mm_add_ps(fjx2,tx);
2250 fjy2 = _mm_add_ps(fjy2,ty);
2251 fjz2 = _mm_add_ps(fjz2,tz);
2253 /**************************
2254 * CALCULATE INTERACTIONS *
2255 **************************/
2257 r23 = _mm_mul_ps(rsq23,rinv23);
2258 r23 = _mm_andnot_ps(dummy_mask,r23);
2260 /* EWALD ELECTROSTATICS */
2262 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2263 ewrt = _mm_mul_ps(r23,ewtabscale);
2264 ewitab = _mm_cvttps_epi32(ewrt);
2265 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
2266 gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
2267 ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
2269 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2270 felec = _mm_mul_ps(_mm_mul_ps(qq23,rinv23),_mm_sub_ps(rinvsq23,felec));
2274 fscal = _mm_andnot_ps(dummy_mask,fscal);
2276 /* Calculate temporary vectorial force */
2277 tx = _mm_mul_ps(fscal,dx23);
2278 ty = _mm_mul_ps(fscal,dy23);
2279 tz = _mm_mul_ps(fscal,dz23);
2281 /* Update vectorial force */
2282 fix2 = _mm_add_ps(fix2,tx);
2283 fiy2 = _mm_add_ps(fiy2,ty);
2284 fiz2 = _mm_add_ps(fiz2,tz);
2286 fjx3 = _mm_add_ps(fjx3,tx);
2287 fjy3 = _mm_add_ps(fjy3,ty);
2288 fjz3 = _mm_add_ps(fjz3,tz);
2290 /**************************
2291 * CALCULATE INTERACTIONS *
2292 **************************/
2294 r31 = _mm_mul_ps(rsq31,rinv31);
2295 r31 = _mm_andnot_ps(dummy_mask,r31);
2297 /* EWALD ELECTROSTATICS */
2299 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2300 ewrt = _mm_mul_ps(r31,ewtabscale);
2301 ewitab = _mm_cvttps_epi32(ewrt);
2302 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
2303 gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
2304 ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
2306 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2307 felec = _mm_mul_ps(_mm_mul_ps(qq31,rinv31),_mm_sub_ps(rinvsq31,felec));
2311 fscal = _mm_andnot_ps(dummy_mask,fscal);
2313 /* Calculate temporary vectorial force */
2314 tx = _mm_mul_ps(fscal,dx31);
2315 ty = _mm_mul_ps(fscal,dy31);
2316 tz = _mm_mul_ps(fscal,dz31);
2318 /* Update vectorial force */
2319 fix3 = _mm_add_ps(fix3,tx);
2320 fiy3 = _mm_add_ps(fiy3,ty);
2321 fiz3 = _mm_add_ps(fiz3,tz);
2323 fjx1 = _mm_add_ps(fjx1,tx);
2324 fjy1 = _mm_add_ps(fjy1,ty);
2325 fjz1 = _mm_add_ps(fjz1,tz);
2327 /**************************
2328 * CALCULATE INTERACTIONS *
2329 **************************/
2331 r32 = _mm_mul_ps(rsq32,rinv32);
2332 r32 = _mm_andnot_ps(dummy_mask,r32);
2334 /* EWALD ELECTROSTATICS */
2336 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2337 ewrt = _mm_mul_ps(r32,ewtabscale);
2338 ewitab = _mm_cvttps_epi32(ewrt);
2339 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
2340 gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
2341 ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
2343 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2344 felec = _mm_mul_ps(_mm_mul_ps(qq32,rinv32),_mm_sub_ps(rinvsq32,felec));
2348 fscal = _mm_andnot_ps(dummy_mask,fscal);
2350 /* Calculate temporary vectorial force */
2351 tx = _mm_mul_ps(fscal,dx32);
2352 ty = _mm_mul_ps(fscal,dy32);
2353 tz = _mm_mul_ps(fscal,dz32);
2355 /* Update vectorial force */
2356 fix3 = _mm_add_ps(fix3,tx);
2357 fiy3 = _mm_add_ps(fiy3,ty);
2358 fiz3 = _mm_add_ps(fiz3,tz);
2360 fjx2 = _mm_add_ps(fjx2,tx);
2361 fjy2 = _mm_add_ps(fjy2,ty);
2362 fjz2 = _mm_add_ps(fjz2,tz);
2364 /**************************
2365 * CALCULATE INTERACTIONS *
2366 **************************/
2368 r33 = _mm_mul_ps(rsq33,rinv33);
2369 r33 = _mm_andnot_ps(dummy_mask,r33);
2371 /* EWALD ELECTROSTATICS */
2373 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2374 ewrt = _mm_mul_ps(r33,ewtabscale);
2375 ewitab = _mm_cvttps_epi32(ewrt);
2376 eweps = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
2377 gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
2378 ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
2380 felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2381 felec = _mm_mul_ps(_mm_mul_ps(qq33,rinv33),_mm_sub_ps(rinvsq33,felec));
2385 fscal = _mm_andnot_ps(dummy_mask,fscal);
2387 /* Calculate temporary vectorial force */
2388 tx = _mm_mul_ps(fscal,dx33);
2389 ty = _mm_mul_ps(fscal,dy33);
2390 tz = _mm_mul_ps(fscal,dz33);
2392 /* Update vectorial force */
2393 fix3 = _mm_add_ps(fix3,tx);
2394 fiy3 = _mm_add_ps(fiy3,ty);
2395 fiz3 = _mm_add_ps(fiz3,tz);
2397 fjx3 = _mm_add_ps(fjx3,tx);
2398 fjy3 = _mm_add_ps(fjy3,ty);
2399 fjz3 = _mm_add_ps(fjz3,tz);
2401 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
2402 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
2403 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
2404 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
2406 gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
2407 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
2408 fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2410 /* Inner loop uses 363 flops */
2413 /* End of innermost loop */
2415 gmx_mm_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2416 f+i_coord_offset,fshift+i_shift_offset);
2418 /* Increment number of inner iterations */
2419 inneriter += j_index_end - j_index_start;
2421 /* Outer loop uses 24 flops */
2424 /* Increment number of outer iterations */
2427 /* Update outer/inner flops */
2429 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*363);