2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
5 * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
6 * and including many others, as listed in the AUTHORS file in the
7 * top-level source directory and at http://www.gromacs.org.
9 * GROMACS is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU Lesser General Public License
11 * as published by the Free Software Foundation; either version 2.1
12 * of the License, or (at your option) any later version.
14 * GROMACS is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
17 * Lesser General Public License for more details.
19 * You should have received a copy of the GNU Lesser General Public
20 * License along with GROMACS; if not, see
21 * http://www.gnu.org/licenses, or write to the Free Software Foundation,
22 * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
24 * If you want to redistribute modifications to GROMACS, please
25 * consider that scientific software is very special. Version
26 * control is crucial - bugs must be traceable. We will be happy to
27 * consider code for inclusion in the official distribution, but
28 * derived work must not be called official GROMACS. Details are found
29 * in the README & COPYING files - if they are missing, get the
30 * official version at http://www.gromacs.org.
32 * To help us fund GROMACS development, we humbly ask that you cite
33 * the research papers on the package. Check out http://www.gromacs.org.
36 * Note: this file was generated by the GROMACS avx_128_fma_single kernel generator.
42 #include "../nb_kernel.h"
43 #include "gromacs/legacyheaders/types/simple.h"
44 #include "gromacs/math/vec.h"
45 #include "gromacs/legacyheaders/nrnb.h"
47 #include "gromacs/simd/math_x86_avx_128_fma_single.h"
48 #include "kernelutil_x86_avx_128_fma_single.h"
51 * Gromacs nonbonded kernel: nb_kernel_ElecCSTab_VdwCSTab_GeomW4W4_VF_avx_128_fma_single
52 * Electrostatics interaction: CubicSplineTable
53 * VdW interaction: CubicSplineTable
54 * Geometry: Water4-Water4
55 * Calculate force/pot: PotentialAndForce
58 nb_kernel_ElecCSTab_VdwCSTab_GeomW4W4_VF_avx_128_fma_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 AVX_128, 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 fscal,rcutoff,rcutoff2,jidxall;
84 __m128 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
86 __m128 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
88 __m128 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
90 __m128 ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
91 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
92 __m128 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
93 int vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
94 __m128 jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
95 int vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
96 __m128 jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
97 int vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
98 __m128 jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
99 __m128 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
100 __m128 dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
101 __m128 dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
102 __m128 dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
103 __m128 dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
104 __m128 dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
105 __m128 dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
106 __m128 dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
107 __m128 dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
108 __m128 dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
109 __m128 velec,felec,velecsum,facel,crf,krf,krf2;
112 __m128 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
115 __m128 one_sixth = _mm_set1_ps(1.0/6.0);
116 __m128 one_twelfth = _mm_set1_ps(1.0/12.0);
118 __m128i ifour = _mm_set1_epi32(4);
119 __m128 rt,vfeps,twovfeps,vftabscale,Y,F,G,H,Fp,VV,FF;
121 __m128 dummy_mask,cutoff_mask;
122 __m128 signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
123 __m128 one = _mm_set1_ps(1.0);
124 __m128 two = _mm_set1_ps(2.0);
130 jindex = nlist->jindex;
132 shiftidx = nlist->shift;
134 shiftvec = fr->shift_vec[0];
135 fshift = fr->fshift[0];
136 facel = _mm_set1_ps(fr->epsfac);
137 charge = mdatoms->chargeA;
138 nvdwtype = fr->ntype;
140 vdwtype = mdatoms->typeA;
142 vftab = kernel_data->table_elec_vdw->data;
143 vftabscale = _mm_set1_ps(kernel_data->table_elec_vdw->scale);
145 /* Setup water-specific parameters */
146 inr = nlist->iinr[0];
147 iq1 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1]));
148 iq2 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2]));
149 iq3 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+3]));
150 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
152 jq1 = _mm_set1_ps(charge[inr+1]);
153 jq2 = _mm_set1_ps(charge[inr+2]);
154 jq3 = _mm_set1_ps(charge[inr+3]);
155 vdwjidx0A = 2*vdwtype[inr+0];
156 c6_00 = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A]);
157 c12_00 = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A+1]);
158 qq11 = _mm_mul_ps(iq1,jq1);
159 qq12 = _mm_mul_ps(iq1,jq2);
160 qq13 = _mm_mul_ps(iq1,jq3);
161 qq21 = _mm_mul_ps(iq2,jq1);
162 qq22 = _mm_mul_ps(iq2,jq2);
163 qq23 = _mm_mul_ps(iq2,jq3);
164 qq31 = _mm_mul_ps(iq3,jq1);
165 qq32 = _mm_mul_ps(iq3,jq2);
166 qq33 = _mm_mul_ps(iq3,jq3);
168 /* Avoid stupid compiler warnings */
169 jnrA = jnrB = jnrC = jnrD = 0;
178 for(iidx=0;iidx<4*DIM;iidx++)
183 /* Start outer loop over neighborlists */
184 for(iidx=0; iidx<nri; iidx++)
186 /* Load shift vector for this list */
187 i_shift_offset = DIM*shiftidx[iidx];
189 /* Load limits for loop over neighbors */
190 j_index_start = jindex[iidx];
191 j_index_end = jindex[iidx+1];
193 /* Get outer coordinate index */
195 i_coord_offset = DIM*inr;
197 /* Load i particle coords and add shift vector */
198 gmx_mm_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
199 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
201 fix0 = _mm_setzero_ps();
202 fiy0 = _mm_setzero_ps();
203 fiz0 = _mm_setzero_ps();
204 fix1 = _mm_setzero_ps();
205 fiy1 = _mm_setzero_ps();
206 fiz1 = _mm_setzero_ps();
207 fix2 = _mm_setzero_ps();
208 fiy2 = _mm_setzero_ps();
209 fiz2 = _mm_setzero_ps();
210 fix3 = _mm_setzero_ps();
211 fiy3 = _mm_setzero_ps();
212 fiz3 = _mm_setzero_ps();
214 /* Reset potential sums */
215 velecsum = _mm_setzero_ps();
216 vvdwsum = _mm_setzero_ps();
218 /* Start inner kernel loop */
219 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
222 /* Get j neighbor index, and coordinate index */
227 j_coord_offsetA = DIM*jnrA;
228 j_coord_offsetB = DIM*jnrB;
229 j_coord_offsetC = DIM*jnrC;
230 j_coord_offsetD = DIM*jnrD;
232 /* load j atom coordinates */
233 gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
234 x+j_coord_offsetC,x+j_coord_offsetD,
235 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
236 &jy2,&jz2,&jx3,&jy3,&jz3);
238 /* Calculate displacement vector */
239 dx00 = _mm_sub_ps(ix0,jx0);
240 dy00 = _mm_sub_ps(iy0,jy0);
241 dz00 = _mm_sub_ps(iz0,jz0);
242 dx11 = _mm_sub_ps(ix1,jx1);
243 dy11 = _mm_sub_ps(iy1,jy1);
244 dz11 = _mm_sub_ps(iz1,jz1);
245 dx12 = _mm_sub_ps(ix1,jx2);
246 dy12 = _mm_sub_ps(iy1,jy2);
247 dz12 = _mm_sub_ps(iz1,jz2);
248 dx13 = _mm_sub_ps(ix1,jx3);
249 dy13 = _mm_sub_ps(iy1,jy3);
250 dz13 = _mm_sub_ps(iz1,jz3);
251 dx21 = _mm_sub_ps(ix2,jx1);
252 dy21 = _mm_sub_ps(iy2,jy1);
253 dz21 = _mm_sub_ps(iz2,jz1);
254 dx22 = _mm_sub_ps(ix2,jx2);
255 dy22 = _mm_sub_ps(iy2,jy2);
256 dz22 = _mm_sub_ps(iz2,jz2);
257 dx23 = _mm_sub_ps(ix2,jx3);
258 dy23 = _mm_sub_ps(iy2,jy3);
259 dz23 = _mm_sub_ps(iz2,jz3);
260 dx31 = _mm_sub_ps(ix3,jx1);
261 dy31 = _mm_sub_ps(iy3,jy1);
262 dz31 = _mm_sub_ps(iz3,jz1);
263 dx32 = _mm_sub_ps(ix3,jx2);
264 dy32 = _mm_sub_ps(iy3,jy2);
265 dz32 = _mm_sub_ps(iz3,jz2);
266 dx33 = _mm_sub_ps(ix3,jx3);
267 dy33 = _mm_sub_ps(iy3,jy3);
268 dz33 = _mm_sub_ps(iz3,jz3);
270 /* Calculate squared distance and things based on it */
271 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
272 rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
273 rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
274 rsq13 = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
275 rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
276 rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
277 rsq23 = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
278 rsq31 = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
279 rsq32 = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
280 rsq33 = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
282 rinv00 = gmx_mm_invsqrt_ps(rsq00);
283 rinv11 = gmx_mm_invsqrt_ps(rsq11);
284 rinv12 = gmx_mm_invsqrt_ps(rsq12);
285 rinv13 = gmx_mm_invsqrt_ps(rsq13);
286 rinv21 = gmx_mm_invsqrt_ps(rsq21);
287 rinv22 = gmx_mm_invsqrt_ps(rsq22);
288 rinv23 = gmx_mm_invsqrt_ps(rsq23);
289 rinv31 = gmx_mm_invsqrt_ps(rsq31);
290 rinv32 = gmx_mm_invsqrt_ps(rsq32);
291 rinv33 = gmx_mm_invsqrt_ps(rsq33);
293 fjx0 = _mm_setzero_ps();
294 fjy0 = _mm_setzero_ps();
295 fjz0 = _mm_setzero_ps();
296 fjx1 = _mm_setzero_ps();
297 fjy1 = _mm_setzero_ps();
298 fjz1 = _mm_setzero_ps();
299 fjx2 = _mm_setzero_ps();
300 fjy2 = _mm_setzero_ps();
301 fjz2 = _mm_setzero_ps();
302 fjx3 = _mm_setzero_ps();
303 fjy3 = _mm_setzero_ps();
304 fjz3 = _mm_setzero_ps();
306 /**************************
307 * CALCULATE INTERACTIONS *
308 **************************/
310 r00 = _mm_mul_ps(rsq00,rinv00);
312 /* Calculate table index by multiplying r with table scale and truncate to integer */
313 rt = _mm_mul_ps(r00,vftabscale);
314 vfitab = _mm_cvttps_epi32(rt);
316 vfeps = _mm_frcz_ps(rt);
318 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
320 twovfeps = _mm_add_ps(vfeps,vfeps);
321 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
323 /* CUBIC SPLINE TABLE DISPERSION */
324 vfitab = _mm_add_epi32(vfitab,ifour);
325 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
326 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
327 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
328 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
329 _MM_TRANSPOSE4_PS(Y,F,G,H);
330 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
331 VV = _mm_macc_ps(vfeps,Fp,Y);
332 vvdw6 = _mm_mul_ps(c6_00,VV);
333 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
334 fvdw6 = _mm_mul_ps(c6_00,FF);
336 /* CUBIC SPLINE TABLE REPULSION */
337 vfitab = _mm_add_epi32(vfitab,ifour);
338 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
339 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
340 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
341 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
342 _MM_TRANSPOSE4_PS(Y,F,G,H);
343 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
344 VV = _mm_macc_ps(vfeps,Fp,Y);
345 vvdw12 = _mm_mul_ps(c12_00,VV);
346 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
347 fvdw12 = _mm_mul_ps(c12_00,FF);
348 vvdw = _mm_add_ps(vvdw12,vvdw6);
349 fvdw = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
351 /* Update potential sum for this i atom from the interaction with this j atom. */
352 vvdwsum = _mm_add_ps(vvdwsum,vvdw);
356 /* Update vectorial force */
357 fix0 = _mm_macc_ps(dx00,fscal,fix0);
358 fiy0 = _mm_macc_ps(dy00,fscal,fiy0);
359 fiz0 = _mm_macc_ps(dz00,fscal,fiz0);
361 fjx0 = _mm_macc_ps(dx00,fscal,fjx0);
362 fjy0 = _mm_macc_ps(dy00,fscal,fjy0);
363 fjz0 = _mm_macc_ps(dz00,fscal,fjz0);
365 /**************************
366 * CALCULATE INTERACTIONS *
367 **************************/
369 r11 = _mm_mul_ps(rsq11,rinv11);
371 /* Calculate table index by multiplying r with table scale and truncate to integer */
372 rt = _mm_mul_ps(r11,vftabscale);
373 vfitab = _mm_cvttps_epi32(rt);
375 vfeps = _mm_frcz_ps(rt);
377 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
379 twovfeps = _mm_add_ps(vfeps,vfeps);
380 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
382 /* CUBIC SPLINE TABLE ELECTROSTATICS */
383 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
384 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
385 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
386 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
387 _MM_TRANSPOSE4_PS(Y,F,G,H);
388 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
389 VV = _mm_macc_ps(vfeps,Fp,Y);
390 velec = _mm_mul_ps(qq11,VV);
391 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
392 felec = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq11,FF),_mm_mul_ps(vftabscale,rinv11)));
394 /* Update potential sum for this i atom from the interaction with this j atom. */
395 velecsum = _mm_add_ps(velecsum,velec);
399 /* Update vectorial force */
400 fix1 = _mm_macc_ps(dx11,fscal,fix1);
401 fiy1 = _mm_macc_ps(dy11,fscal,fiy1);
402 fiz1 = _mm_macc_ps(dz11,fscal,fiz1);
404 fjx1 = _mm_macc_ps(dx11,fscal,fjx1);
405 fjy1 = _mm_macc_ps(dy11,fscal,fjy1);
406 fjz1 = _mm_macc_ps(dz11,fscal,fjz1);
408 /**************************
409 * CALCULATE INTERACTIONS *
410 **************************/
412 r12 = _mm_mul_ps(rsq12,rinv12);
414 /* Calculate table index by multiplying r with table scale and truncate to integer */
415 rt = _mm_mul_ps(r12,vftabscale);
416 vfitab = _mm_cvttps_epi32(rt);
418 vfeps = _mm_frcz_ps(rt);
420 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
422 twovfeps = _mm_add_ps(vfeps,vfeps);
423 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
425 /* CUBIC SPLINE TABLE ELECTROSTATICS */
426 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
427 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
428 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
429 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
430 _MM_TRANSPOSE4_PS(Y,F,G,H);
431 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
432 VV = _mm_macc_ps(vfeps,Fp,Y);
433 velec = _mm_mul_ps(qq12,VV);
434 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
435 felec = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq12,FF),_mm_mul_ps(vftabscale,rinv12)));
437 /* Update potential sum for this i atom from the interaction with this j atom. */
438 velecsum = _mm_add_ps(velecsum,velec);
442 /* Update vectorial force */
443 fix1 = _mm_macc_ps(dx12,fscal,fix1);
444 fiy1 = _mm_macc_ps(dy12,fscal,fiy1);
445 fiz1 = _mm_macc_ps(dz12,fscal,fiz1);
447 fjx2 = _mm_macc_ps(dx12,fscal,fjx2);
448 fjy2 = _mm_macc_ps(dy12,fscal,fjy2);
449 fjz2 = _mm_macc_ps(dz12,fscal,fjz2);
451 /**************************
452 * CALCULATE INTERACTIONS *
453 **************************/
455 r13 = _mm_mul_ps(rsq13,rinv13);
457 /* Calculate table index by multiplying r with table scale and truncate to integer */
458 rt = _mm_mul_ps(r13,vftabscale);
459 vfitab = _mm_cvttps_epi32(rt);
461 vfeps = _mm_frcz_ps(rt);
463 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
465 twovfeps = _mm_add_ps(vfeps,vfeps);
466 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
468 /* CUBIC SPLINE TABLE ELECTROSTATICS */
469 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
470 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
471 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
472 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
473 _MM_TRANSPOSE4_PS(Y,F,G,H);
474 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
475 VV = _mm_macc_ps(vfeps,Fp,Y);
476 velec = _mm_mul_ps(qq13,VV);
477 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
478 felec = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq13,FF),_mm_mul_ps(vftabscale,rinv13)));
480 /* Update potential sum for this i atom from the interaction with this j atom. */
481 velecsum = _mm_add_ps(velecsum,velec);
485 /* Update vectorial force */
486 fix1 = _mm_macc_ps(dx13,fscal,fix1);
487 fiy1 = _mm_macc_ps(dy13,fscal,fiy1);
488 fiz1 = _mm_macc_ps(dz13,fscal,fiz1);
490 fjx3 = _mm_macc_ps(dx13,fscal,fjx3);
491 fjy3 = _mm_macc_ps(dy13,fscal,fjy3);
492 fjz3 = _mm_macc_ps(dz13,fscal,fjz3);
494 /**************************
495 * CALCULATE INTERACTIONS *
496 **************************/
498 r21 = _mm_mul_ps(rsq21,rinv21);
500 /* Calculate table index by multiplying r with table scale and truncate to integer */
501 rt = _mm_mul_ps(r21,vftabscale);
502 vfitab = _mm_cvttps_epi32(rt);
504 vfeps = _mm_frcz_ps(rt);
506 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
508 twovfeps = _mm_add_ps(vfeps,vfeps);
509 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
511 /* CUBIC SPLINE TABLE ELECTROSTATICS */
512 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
513 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
514 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
515 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
516 _MM_TRANSPOSE4_PS(Y,F,G,H);
517 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
518 VV = _mm_macc_ps(vfeps,Fp,Y);
519 velec = _mm_mul_ps(qq21,VV);
520 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
521 felec = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq21,FF),_mm_mul_ps(vftabscale,rinv21)));
523 /* Update potential sum for this i atom from the interaction with this j atom. */
524 velecsum = _mm_add_ps(velecsum,velec);
528 /* Update vectorial force */
529 fix2 = _mm_macc_ps(dx21,fscal,fix2);
530 fiy2 = _mm_macc_ps(dy21,fscal,fiy2);
531 fiz2 = _mm_macc_ps(dz21,fscal,fiz2);
533 fjx1 = _mm_macc_ps(dx21,fscal,fjx1);
534 fjy1 = _mm_macc_ps(dy21,fscal,fjy1);
535 fjz1 = _mm_macc_ps(dz21,fscal,fjz1);
537 /**************************
538 * CALCULATE INTERACTIONS *
539 **************************/
541 r22 = _mm_mul_ps(rsq22,rinv22);
543 /* Calculate table index by multiplying r with table scale and truncate to integer */
544 rt = _mm_mul_ps(r22,vftabscale);
545 vfitab = _mm_cvttps_epi32(rt);
547 vfeps = _mm_frcz_ps(rt);
549 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
551 twovfeps = _mm_add_ps(vfeps,vfeps);
552 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
554 /* CUBIC SPLINE TABLE ELECTROSTATICS */
555 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
556 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
557 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
558 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
559 _MM_TRANSPOSE4_PS(Y,F,G,H);
560 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
561 VV = _mm_macc_ps(vfeps,Fp,Y);
562 velec = _mm_mul_ps(qq22,VV);
563 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
564 felec = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq22,FF),_mm_mul_ps(vftabscale,rinv22)));
566 /* Update potential sum for this i atom from the interaction with this j atom. */
567 velecsum = _mm_add_ps(velecsum,velec);
571 /* Update vectorial force */
572 fix2 = _mm_macc_ps(dx22,fscal,fix2);
573 fiy2 = _mm_macc_ps(dy22,fscal,fiy2);
574 fiz2 = _mm_macc_ps(dz22,fscal,fiz2);
576 fjx2 = _mm_macc_ps(dx22,fscal,fjx2);
577 fjy2 = _mm_macc_ps(dy22,fscal,fjy2);
578 fjz2 = _mm_macc_ps(dz22,fscal,fjz2);
580 /**************************
581 * CALCULATE INTERACTIONS *
582 **************************/
584 r23 = _mm_mul_ps(rsq23,rinv23);
586 /* Calculate table index by multiplying r with table scale and truncate to integer */
587 rt = _mm_mul_ps(r23,vftabscale);
588 vfitab = _mm_cvttps_epi32(rt);
590 vfeps = _mm_frcz_ps(rt);
592 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
594 twovfeps = _mm_add_ps(vfeps,vfeps);
595 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
597 /* CUBIC SPLINE TABLE ELECTROSTATICS */
598 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
599 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
600 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
601 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
602 _MM_TRANSPOSE4_PS(Y,F,G,H);
603 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
604 VV = _mm_macc_ps(vfeps,Fp,Y);
605 velec = _mm_mul_ps(qq23,VV);
606 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
607 felec = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq23,FF),_mm_mul_ps(vftabscale,rinv23)));
609 /* Update potential sum for this i atom from the interaction with this j atom. */
610 velecsum = _mm_add_ps(velecsum,velec);
614 /* Update vectorial force */
615 fix2 = _mm_macc_ps(dx23,fscal,fix2);
616 fiy2 = _mm_macc_ps(dy23,fscal,fiy2);
617 fiz2 = _mm_macc_ps(dz23,fscal,fiz2);
619 fjx3 = _mm_macc_ps(dx23,fscal,fjx3);
620 fjy3 = _mm_macc_ps(dy23,fscal,fjy3);
621 fjz3 = _mm_macc_ps(dz23,fscal,fjz3);
623 /**************************
624 * CALCULATE INTERACTIONS *
625 **************************/
627 r31 = _mm_mul_ps(rsq31,rinv31);
629 /* Calculate table index by multiplying r with table scale and truncate to integer */
630 rt = _mm_mul_ps(r31,vftabscale);
631 vfitab = _mm_cvttps_epi32(rt);
633 vfeps = _mm_frcz_ps(rt);
635 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
637 twovfeps = _mm_add_ps(vfeps,vfeps);
638 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
640 /* CUBIC SPLINE TABLE ELECTROSTATICS */
641 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
642 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
643 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
644 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
645 _MM_TRANSPOSE4_PS(Y,F,G,H);
646 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
647 VV = _mm_macc_ps(vfeps,Fp,Y);
648 velec = _mm_mul_ps(qq31,VV);
649 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
650 felec = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq31,FF),_mm_mul_ps(vftabscale,rinv31)));
652 /* Update potential sum for this i atom from the interaction with this j atom. */
653 velecsum = _mm_add_ps(velecsum,velec);
657 /* Update vectorial force */
658 fix3 = _mm_macc_ps(dx31,fscal,fix3);
659 fiy3 = _mm_macc_ps(dy31,fscal,fiy3);
660 fiz3 = _mm_macc_ps(dz31,fscal,fiz3);
662 fjx1 = _mm_macc_ps(dx31,fscal,fjx1);
663 fjy1 = _mm_macc_ps(dy31,fscal,fjy1);
664 fjz1 = _mm_macc_ps(dz31,fscal,fjz1);
666 /**************************
667 * CALCULATE INTERACTIONS *
668 **************************/
670 r32 = _mm_mul_ps(rsq32,rinv32);
672 /* Calculate table index by multiplying r with table scale and truncate to integer */
673 rt = _mm_mul_ps(r32,vftabscale);
674 vfitab = _mm_cvttps_epi32(rt);
676 vfeps = _mm_frcz_ps(rt);
678 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
680 twovfeps = _mm_add_ps(vfeps,vfeps);
681 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
683 /* CUBIC SPLINE TABLE ELECTROSTATICS */
684 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
685 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
686 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
687 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
688 _MM_TRANSPOSE4_PS(Y,F,G,H);
689 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
690 VV = _mm_macc_ps(vfeps,Fp,Y);
691 velec = _mm_mul_ps(qq32,VV);
692 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
693 felec = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq32,FF),_mm_mul_ps(vftabscale,rinv32)));
695 /* Update potential sum for this i atom from the interaction with this j atom. */
696 velecsum = _mm_add_ps(velecsum,velec);
700 /* Update vectorial force */
701 fix3 = _mm_macc_ps(dx32,fscal,fix3);
702 fiy3 = _mm_macc_ps(dy32,fscal,fiy3);
703 fiz3 = _mm_macc_ps(dz32,fscal,fiz3);
705 fjx2 = _mm_macc_ps(dx32,fscal,fjx2);
706 fjy2 = _mm_macc_ps(dy32,fscal,fjy2);
707 fjz2 = _mm_macc_ps(dz32,fscal,fjz2);
709 /**************************
710 * CALCULATE INTERACTIONS *
711 **************************/
713 r33 = _mm_mul_ps(rsq33,rinv33);
715 /* Calculate table index by multiplying r with table scale and truncate to integer */
716 rt = _mm_mul_ps(r33,vftabscale);
717 vfitab = _mm_cvttps_epi32(rt);
719 vfeps = _mm_frcz_ps(rt);
721 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
723 twovfeps = _mm_add_ps(vfeps,vfeps);
724 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
726 /* CUBIC SPLINE TABLE ELECTROSTATICS */
727 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
728 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
729 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
730 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
731 _MM_TRANSPOSE4_PS(Y,F,G,H);
732 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
733 VV = _mm_macc_ps(vfeps,Fp,Y);
734 velec = _mm_mul_ps(qq33,VV);
735 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
736 felec = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq33,FF),_mm_mul_ps(vftabscale,rinv33)));
738 /* Update potential sum for this i atom from the interaction with this j atom. */
739 velecsum = _mm_add_ps(velecsum,velec);
743 /* Update vectorial force */
744 fix3 = _mm_macc_ps(dx33,fscal,fix3);
745 fiy3 = _mm_macc_ps(dy33,fscal,fiy3);
746 fiz3 = _mm_macc_ps(dz33,fscal,fiz3);
748 fjx3 = _mm_macc_ps(dx33,fscal,fjx3);
749 fjy3 = _mm_macc_ps(dy33,fscal,fjy3);
750 fjz3 = _mm_macc_ps(dz33,fscal,fjz3);
752 fjptrA = f+j_coord_offsetA;
753 fjptrB = f+j_coord_offsetB;
754 fjptrC = f+j_coord_offsetC;
755 fjptrD = f+j_coord_offsetD;
757 gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
758 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
759 fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
761 /* Inner loop uses 476 flops */
767 /* Get j neighbor index, and coordinate index */
768 jnrlistA = jjnr[jidx];
769 jnrlistB = jjnr[jidx+1];
770 jnrlistC = jjnr[jidx+2];
771 jnrlistD = jjnr[jidx+3];
772 /* Sign of each element will be negative for non-real atoms.
773 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
774 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
776 dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
777 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
778 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
779 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
780 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
781 j_coord_offsetA = DIM*jnrA;
782 j_coord_offsetB = DIM*jnrB;
783 j_coord_offsetC = DIM*jnrC;
784 j_coord_offsetD = DIM*jnrD;
786 /* load j atom coordinates */
787 gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
788 x+j_coord_offsetC,x+j_coord_offsetD,
789 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
790 &jy2,&jz2,&jx3,&jy3,&jz3);
792 /* Calculate displacement vector */
793 dx00 = _mm_sub_ps(ix0,jx0);
794 dy00 = _mm_sub_ps(iy0,jy0);
795 dz00 = _mm_sub_ps(iz0,jz0);
796 dx11 = _mm_sub_ps(ix1,jx1);
797 dy11 = _mm_sub_ps(iy1,jy1);
798 dz11 = _mm_sub_ps(iz1,jz1);
799 dx12 = _mm_sub_ps(ix1,jx2);
800 dy12 = _mm_sub_ps(iy1,jy2);
801 dz12 = _mm_sub_ps(iz1,jz2);
802 dx13 = _mm_sub_ps(ix1,jx3);
803 dy13 = _mm_sub_ps(iy1,jy3);
804 dz13 = _mm_sub_ps(iz1,jz3);
805 dx21 = _mm_sub_ps(ix2,jx1);
806 dy21 = _mm_sub_ps(iy2,jy1);
807 dz21 = _mm_sub_ps(iz2,jz1);
808 dx22 = _mm_sub_ps(ix2,jx2);
809 dy22 = _mm_sub_ps(iy2,jy2);
810 dz22 = _mm_sub_ps(iz2,jz2);
811 dx23 = _mm_sub_ps(ix2,jx3);
812 dy23 = _mm_sub_ps(iy2,jy3);
813 dz23 = _mm_sub_ps(iz2,jz3);
814 dx31 = _mm_sub_ps(ix3,jx1);
815 dy31 = _mm_sub_ps(iy3,jy1);
816 dz31 = _mm_sub_ps(iz3,jz1);
817 dx32 = _mm_sub_ps(ix3,jx2);
818 dy32 = _mm_sub_ps(iy3,jy2);
819 dz32 = _mm_sub_ps(iz3,jz2);
820 dx33 = _mm_sub_ps(ix3,jx3);
821 dy33 = _mm_sub_ps(iy3,jy3);
822 dz33 = _mm_sub_ps(iz3,jz3);
824 /* Calculate squared distance and things based on it */
825 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
826 rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
827 rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
828 rsq13 = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
829 rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
830 rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
831 rsq23 = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
832 rsq31 = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
833 rsq32 = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
834 rsq33 = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
836 rinv00 = gmx_mm_invsqrt_ps(rsq00);
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 fjx0 = _mm_setzero_ps();
848 fjy0 = _mm_setzero_ps();
849 fjz0 = _mm_setzero_ps();
850 fjx1 = _mm_setzero_ps();
851 fjy1 = _mm_setzero_ps();
852 fjz1 = _mm_setzero_ps();
853 fjx2 = _mm_setzero_ps();
854 fjy2 = _mm_setzero_ps();
855 fjz2 = _mm_setzero_ps();
856 fjx3 = _mm_setzero_ps();
857 fjy3 = _mm_setzero_ps();
858 fjz3 = _mm_setzero_ps();
860 /**************************
861 * CALCULATE INTERACTIONS *
862 **************************/
864 r00 = _mm_mul_ps(rsq00,rinv00);
865 r00 = _mm_andnot_ps(dummy_mask,r00);
867 /* Calculate table index by multiplying r with table scale and truncate to integer */
868 rt = _mm_mul_ps(r00,vftabscale);
869 vfitab = _mm_cvttps_epi32(rt);
871 vfeps = _mm_frcz_ps(rt);
873 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
875 twovfeps = _mm_add_ps(vfeps,vfeps);
876 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
878 /* CUBIC SPLINE TABLE DISPERSION */
879 vfitab = _mm_add_epi32(vfitab,ifour);
880 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
881 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
882 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
883 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
884 _MM_TRANSPOSE4_PS(Y,F,G,H);
885 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
886 VV = _mm_macc_ps(vfeps,Fp,Y);
887 vvdw6 = _mm_mul_ps(c6_00,VV);
888 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
889 fvdw6 = _mm_mul_ps(c6_00,FF);
891 /* CUBIC SPLINE TABLE REPULSION */
892 vfitab = _mm_add_epi32(vfitab,ifour);
893 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
894 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
895 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
896 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
897 _MM_TRANSPOSE4_PS(Y,F,G,H);
898 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
899 VV = _mm_macc_ps(vfeps,Fp,Y);
900 vvdw12 = _mm_mul_ps(c12_00,VV);
901 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
902 fvdw12 = _mm_mul_ps(c12_00,FF);
903 vvdw = _mm_add_ps(vvdw12,vvdw6);
904 fvdw = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
906 /* Update potential sum for this i atom from the interaction with this j atom. */
907 vvdw = _mm_andnot_ps(dummy_mask,vvdw);
908 vvdwsum = _mm_add_ps(vvdwsum,vvdw);
912 fscal = _mm_andnot_ps(dummy_mask,fscal);
914 /* Update vectorial force */
915 fix0 = _mm_macc_ps(dx00,fscal,fix0);
916 fiy0 = _mm_macc_ps(dy00,fscal,fiy0);
917 fiz0 = _mm_macc_ps(dz00,fscal,fiz0);
919 fjx0 = _mm_macc_ps(dx00,fscal,fjx0);
920 fjy0 = _mm_macc_ps(dy00,fscal,fjy0);
921 fjz0 = _mm_macc_ps(dz00,fscal,fjz0);
923 /**************************
924 * CALCULATE INTERACTIONS *
925 **************************/
927 r11 = _mm_mul_ps(rsq11,rinv11);
928 r11 = _mm_andnot_ps(dummy_mask,r11);
930 /* Calculate table index by multiplying r with table scale and truncate to integer */
931 rt = _mm_mul_ps(r11,vftabscale);
932 vfitab = _mm_cvttps_epi32(rt);
934 vfeps = _mm_frcz_ps(rt);
936 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
938 twovfeps = _mm_add_ps(vfeps,vfeps);
939 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
941 /* CUBIC SPLINE TABLE ELECTROSTATICS */
942 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
943 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
944 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
945 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
946 _MM_TRANSPOSE4_PS(Y,F,G,H);
947 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
948 VV = _mm_macc_ps(vfeps,Fp,Y);
949 velec = _mm_mul_ps(qq11,VV);
950 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
951 felec = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq11,FF),_mm_mul_ps(vftabscale,rinv11)));
953 /* Update potential sum for this i atom from the interaction with this j atom. */
954 velec = _mm_andnot_ps(dummy_mask,velec);
955 velecsum = _mm_add_ps(velecsum,velec);
959 fscal = _mm_andnot_ps(dummy_mask,fscal);
961 /* Update vectorial force */
962 fix1 = _mm_macc_ps(dx11,fscal,fix1);
963 fiy1 = _mm_macc_ps(dy11,fscal,fiy1);
964 fiz1 = _mm_macc_ps(dz11,fscal,fiz1);
966 fjx1 = _mm_macc_ps(dx11,fscal,fjx1);
967 fjy1 = _mm_macc_ps(dy11,fscal,fjy1);
968 fjz1 = _mm_macc_ps(dz11,fscal,fjz1);
970 /**************************
971 * CALCULATE INTERACTIONS *
972 **************************/
974 r12 = _mm_mul_ps(rsq12,rinv12);
975 r12 = _mm_andnot_ps(dummy_mask,r12);
977 /* Calculate table index by multiplying r with table scale and truncate to integer */
978 rt = _mm_mul_ps(r12,vftabscale);
979 vfitab = _mm_cvttps_epi32(rt);
981 vfeps = _mm_frcz_ps(rt);
983 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
985 twovfeps = _mm_add_ps(vfeps,vfeps);
986 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
988 /* CUBIC SPLINE TABLE ELECTROSTATICS */
989 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
990 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
991 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
992 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
993 _MM_TRANSPOSE4_PS(Y,F,G,H);
994 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
995 VV = _mm_macc_ps(vfeps,Fp,Y);
996 velec = _mm_mul_ps(qq12,VV);
997 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
998 felec = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq12,FF),_mm_mul_ps(vftabscale,rinv12)));
1000 /* Update potential sum for this i atom from the interaction with this j atom. */
1001 velec = _mm_andnot_ps(dummy_mask,velec);
1002 velecsum = _mm_add_ps(velecsum,velec);
1006 fscal = _mm_andnot_ps(dummy_mask,fscal);
1008 /* Update vectorial force */
1009 fix1 = _mm_macc_ps(dx12,fscal,fix1);
1010 fiy1 = _mm_macc_ps(dy12,fscal,fiy1);
1011 fiz1 = _mm_macc_ps(dz12,fscal,fiz1);
1013 fjx2 = _mm_macc_ps(dx12,fscal,fjx2);
1014 fjy2 = _mm_macc_ps(dy12,fscal,fjy2);
1015 fjz2 = _mm_macc_ps(dz12,fscal,fjz2);
1017 /**************************
1018 * CALCULATE INTERACTIONS *
1019 **************************/
1021 r13 = _mm_mul_ps(rsq13,rinv13);
1022 r13 = _mm_andnot_ps(dummy_mask,r13);
1024 /* Calculate table index by multiplying r with table scale and truncate to integer */
1025 rt = _mm_mul_ps(r13,vftabscale);
1026 vfitab = _mm_cvttps_epi32(rt);
1028 vfeps = _mm_frcz_ps(rt);
1030 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1032 twovfeps = _mm_add_ps(vfeps,vfeps);
1033 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1035 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1036 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
1037 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
1038 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
1039 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
1040 _MM_TRANSPOSE4_PS(Y,F,G,H);
1041 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
1042 VV = _mm_macc_ps(vfeps,Fp,Y);
1043 velec = _mm_mul_ps(qq13,VV);
1044 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
1045 felec = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq13,FF),_mm_mul_ps(vftabscale,rinv13)));
1047 /* Update potential sum for this i atom from the interaction with this j atom. */
1048 velec = _mm_andnot_ps(dummy_mask,velec);
1049 velecsum = _mm_add_ps(velecsum,velec);
1053 fscal = _mm_andnot_ps(dummy_mask,fscal);
1055 /* Update vectorial force */
1056 fix1 = _mm_macc_ps(dx13,fscal,fix1);
1057 fiy1 = _mm_macc_ps(dy13,fscal,fiy1);
1058 fiz1 = _mm_macc_ps(dz13,fscal,fiz1);
1060 fjx3 = _mm_macc_ps(dx13,fscal,fjx3);
1061 fjy3 = _mm_macc_ps(dy13,fscal,fjy3);
1062 fjz3 = _mm_macc_ps(dz13,fscal,fjz3);
1064 /**************************
1065 * CALCULATE INTERACTIONS *
1066 **************************/
1068 r21 = _mm_mul_ps(rsq21,rinv21);
1069 r21 = _mm_andnot_ps(dummy_mask,r21);
1071 /* Calculate table index by multiplying r with table scale and truncate to integer */
1072 rt = _mm_mul_ps(r21,vftabscale);
1073 vfitab = _mm_cvttps_epi32(rt);
1075 vfeps = _mm_frcz_ps(rt);
1077 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1079 twovfeps = _mm_add_ps(vfeps,vfeps);
1080 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1082 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1083 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
1084 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
1085 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
1086 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
1087 _MM_TRANSPOSE4_PS(Y,F,G,H);
1088 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
1089 VV = _mm_macc_ps(vfeps,Fp,Y);
1090 velec = _mm_mul_ps(qq21,VV);
1091 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
1092 felec = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq21,FF),_mm_mul_ps(vftabscale,rinv21)));
1094 /* Update potential sum for this i atom from the interaction with this j atom. */
1095 velec = _mm_andnot_ps(dummy_mask,velec);
1096 velecsum = _mm_add_ps(velecsum,velec);
1100 fscal = _mm_andnot_ps(dummy_mask,fscal);
1102 /* Update vectorial force */
1103 fix2 = _mm_macc_ps(dx21,fscal,fix2);
1104 fiy2 = _mm_macc_ps(dy21,fscal,fiy2);
1105 fiz2 = _mm_macc_ps(dz21,fscal,fiz2);
1107 fjx1 = _mm_macc_ps(dx21,fscal,fjx1);
1108 fjy1 = _mm_macc_ps(dy21,fscal,fjy1);
1109 fjz1 = _mm_macc_ps(dz21,fscal,fjz1);
1111 /**************************
1112 * CALCULATE INTERACTIONS *
1113 **************************/
1115 r22 = _mm_mul_ps(rsq22,rinv22);
1116 r22 = _mm_andnot_ps(dummy_mask,r22);
1118 /* Calculate table index by multiplying r with table scale and truncate to integer */
1119 rt = _mm_mul_ps(r22,vftabscale);
1120 vfitab = _mm_cvttps_epi32(rt);
1122 vfeps = _mm_frcz_ps(rt);
1124 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1126 twovfeps = _mm_add_ps(vfeps,vfeps);
1127 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1129 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1130 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
1131 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
1132 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
1133 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
1134 _MM_TRANSPOSE4_PS(Y,F,G,H);
1135 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
1136 VV = _mm_macc_ps(vfeps,Fp,Y);
1137 velec = _mm_mul_ps(qq22,VV);
1138 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
1139 felec = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq22,FF),_mm_mul_ps(vftabscale,rinv22)));
1141 /* Update potential sum for this i atom from the interaction with this j atom. */
1142 velec = _mm_andnot_ps(dummy_mask,velec);
1143 velecsum = _mm_add_ps(velecsum,velec);
1147 fscal = _mm_andnot_ps(dummy_mask,fscal);
1149 /* Update vectorial force */
1150 fix2 = _mm_macc_ps(dx22,fscal,fix2);
1151 fiy2 = _mm_macc_ps(dy22,fscal,fiy2);
1152 fiz2 = _mm_macc_ps(dz22,fscal,fiz2);
1154 fjx2 = _mm_macc_ps(dx22,fscal,fjx2);
1155 fjy2 = _mm_macc_ps(dy22,fscal,fjy2);
1156 fjz2 = _mm_macc_ps(dz22,fscal,fjz2);
1158 /**************************
1159 * CALCULATE INTERACTIONS *
1160 **************************/
1162 r23 = _mm_mul_ps(rsq23,rinv23);
1163 r23 = _mm_andnot_ps(dummy_mask,r23);
1165 /* Calculate table index by multiplying r with table scale and truncate to integer */
1166 rt = _mm_mul_ps(r23,vftabscale);
1167 vfitab = _mm_cvttps_epi32(rt);
1169 vfeps = _mm_frcz_ps(rt);
1171 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1173 twovfeps = _mm_add_ps(vfeps,vfeps);
1174 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1176 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1177 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
1178 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
1179 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
1180 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
1181 _MM_TRANSPOSE4_PS(Y,F,G,H);
1182 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
1183 VV = _mm_macc_ps(vfeps,Fp,Y);
1184 velec = _mm_mul_ps(qq23,VV);
1185 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
1186 felec = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq23,FF),_mm_mul_ps(vftabscale,rinv23)));
1188 /* Update potential sum for this i atom from the interaction with this j atom. */
1189 velec = _mm_andnot_ps(dummy_mask,velec);
1190 velecsum = _mm_add_ps(velecsum,velec);
1194 fscal = _mm_andnot_ps(dummy_mask,fscal);
1196 /* Update vectorial force */
1197 fix2 = _mm_macc_ps(dx23,fscal,fix2);
1198 fiy2 = _mm_macc_ps(dy23,fscal,fiy2);
1199 fiz2 = _mm_macc_ps(dz23,fscal,fiz2);
1201 fjx3 = _mm_macc_ps(dx23,fscal,fjx3);
1202 fjy3 = _mm_macc_ps(dy23,fscal,fjy3);
1203 fjz3 = _mm_macc_ps(dz23,fscal,fjz3);
1205 /**************************
1206 * CALCULATE INTERACTIONS *
1207 **************************/
1209 r31 = _mm_mul_ps(rsq31,rinv31);
1210 r31 = _mm_andnot_ps(dummy_mask,r31);
1212 /* Calculate table index by multiplying r with table scale and truncate to integer */
1213 rt = _mm_mul_ps(r31,vftabscale);
1214 vfitab = _mm_cvttps_epi32(rt);
1216 vfeps = _mm_frcz_ps(rt);
1218 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1220 twovfeps = _mm_add_ps(vfeps,vfeps);
1221 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1223 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1224 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
1225 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
1226 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
1227 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
1228 _MM_TRANSPOSE4_PS(Y,F,G,H);
1229 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
1230 VV = _mm_macc_ps(vfeps,Fp,Y);
1231 velec = _mm_mul_ps(qq31,VV);
1232 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
1233 felec = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq31,FF),_mm_mul_ps(vftabscale,rinv31)));
1235 /* Update potential sum for this i atom from the interaction with this j atom. */
1236 velec = _mm_andnot_ps(dummy_mask,velec);
1237 velecsum = _mm_add_ps(velecsum,velec);
1241 fscal = _mm_andnot_ps(dummy_mask,fscal);
1243 /* Update vectorial force */
1244 fix3 = _mm_macc_ps(dx31,fscal,fix3);
1245 fiy3 = _mm_macc_ps(dy31,fscal,fiy3);
1246 fiz3 = _mm_macc_ps(dz31,fscal,fiz3);
1248 fjx1 = _mm_macc_ps(dx31,fscal,fjx1);
1249 fjy1 = _mm_macc_ps(dy31,fscal,fjy1);
1250 fjz1 = _mm_macc_ps(dz31,fscal,fjz1);
1252 /**************************
1253 * CALCULATE INTERACTIONS *
1254 **************************/
1256 r32 = _mm_mul_ps(rsq32,rinv32);
1257 r32 = _mm_andnot_ps(dummy_mask,r32);
1259 /* Calculate table index by multiplying r with table scale and truncate to integer */
1260 rt = _mm_mul_ps(r32,vftabscale);
1261 vfitab = _mm_cvttps_epi32(rt);
1263 vfeps = _mm_frcz_ps(rt);
1265 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1267 twovfeps = _mm_add_ps(vfeps,vfeps);
1268 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1270 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1271 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
1272 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
1273 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
1274 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
1275 _MM_TRANSPOSE4_PS(Y,F,G,H);
1276 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
1277 VV = _mm_macc_ps(vfeps,Fp,Y);
1278 velec = _mm_mul_ps(qq32,VV);
1279 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
1280 felec = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq32,FF),_mm_mul_ps(vftabscale,rinv32)));
1282 /* Update potential sum for this i atom from the interaction with this j atom. */
1283 velec = _mm_andnot_ps(dummy_mask,velec);
1284 velecsum = _mm_add_ps(velecsum,velec);
1288 fscal = _mm_andnot_ps(dummy_mask,fscal);
1290 /* Update vectorial force */
1291 fix3 = _mm_macc_ps(dx32,fscal,fix3);
1292 fiy3 = _mm_macc_ps(dy32,fscal,fiy3);
1293 fiz3 = _mm_macc_ps(dz32,fscal,fiz3);
1295 fjx2 = _mm_macc_ps(dx32,fscal,fjx2);
1296 fjy2 = _mm_macc_ps(dy32,fscal,fjy2);
1297 fjz2 = _mm_macc_ps(dz32,fscal,fjz2);
1299 /**************************
1300 * CALCULATE INTERACTIONS *
1301 **************************/
1303 r33 = _mm_mul_ps(rsq33,rinv33);
1304 r33 = _mm_andnot_ps(dummy_mask,r33);
1306 /* Calculate table index by multiplying r with table scale and truncate to integer */
1307 rt = _mm_mul_ps(r33,vftabscale);
1308 vfitab = _mm_cvttps_epi32(rt);
1310 vfeps = _mm_frcz_ps(rt);
1312 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1314 twovfeps = _mm_add_ps(vfeps,vfeps);
1315 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1317 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1318 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
1319 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
1320 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
1321 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
1322 _MM_TRANSPOSE4_PS(Y,F,G,H);
1323 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
1324 VV = _mm_macc_ps(vfeps,Fp,Y);
1325 velec = _mm_mul_ps(qq33,VV);
1326 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
1327 felec = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq33,FF),_mm_mul_ps(vftabscale,rinv33)));
1329 /* Update potential sum for this i atom from the interaction with this j atom. */
1330 velec = _mm_andnot_ps(dummy_mask,velec);
1331 velecsum = _mm_add_ps(velecsum,velec);
1335 fscal = _mm_andnot_ps(dummy_mask,fscal);
1337 /* Update vectorial force */
1338 fix3 = _mm_macc_ps(dx33,fscal,fix3);
1339 fiy3 = _mm_macc_ps(dy33,fscal,fiy3);
1340 fiz3 = _mm_macc_ps(dz33,fscal,fiz3);
1342 fjx3 = _mm_macc_ps(dx33,fscal,fjx3);
1343 fjy3 = _mm_macc_ps(dy33,fscal,fjy3);
1344 fjz3 = _mm_macc_ps(dz33,fscal,fjz3);
1346 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1347 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1348 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1349 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1351 gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
1352 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1353 fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1355 /* Inner loop uses 486 flops */
1358 /* End of innermost loop */
1360 gmx_mm_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1361 f+i_coord_offset,fshift+i_shift_offset);
1364 /* Update potential energies */
1365 gmx_mm_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
1366 gmx_mm_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
1368 /* Increment number of inner iterations */
1369 inneriter += j_index_end - j_index_start;
1371 /* Outer loop uses 26 flops */
1374 /* Increment number of outer iterations */
1377 /* Update outer/inner flops */
1379 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*486);
1382 * Gromacs nonbonded kernel: nb_kernel_ElecCSTab_VdwCSTab_GeomW4W4_F_avx_128_fma_single
1383 * Electrostatics interaction: CubicSplineTable
1384 * VdW interaction: CubicSplineTable
1385 * Geometry: Water4-Water4
1386 * Calculate force/pot: Force
1389 nb_kernel_ElecCSTab_VdwCSTab_GeomW4W4_F_avx_128_fma_single
1390 (t_nblist * gmx_restrict nlist,
1391 rvec * gmx_restrict xx,
1392 rvec * gmx_restrict ff,
1393 t_forcerec * gmx_restrict fr,
1394 t_mdatoms * gmx_restrict mdatoms,
1395 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1396 t_nrnb * gmx_restrict nrnb)
1398 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1399 * just 0 for non-waters.
1400 * Suffixes A,B,C,D refer to j loop unrolling done with AVX_128, e.g. for the four different
1401 * jnr indices corresponding to data put in the four positions in the SIMD register.
1403 int i_shift_offset,i_coord_offset,outeriter,inneriter;
1404 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1405 int jnrA,jnrB,jnrC,jnrD;
1406 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1407 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1408 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
1409 real rcutoff_scalar;
1410 real *shiftvec,*fshift,*x,*f;
1411 real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
1412 real scratch[4*DIM];
1413 __m128 fscal,rcutoff,rcutoff2,jidxall;
1415 __m128 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1417 __m128 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1419 __m128 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1421 __m128 ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1422 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
1423 __m128 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1424 int vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
1425 __m128 jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1426 int vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
1427 __m128 jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1428 int vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
1429 __m128 jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1430 __m128 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1431 __m128 dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1432 __m128 dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1433 __m128 dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1434 __m128 dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1435 __m128 dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1436 __m128 dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1437 __m128 dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1438 __m128 dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1439 __m128 dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1440 __m128 velec,felec,velecsum,facel,crf,krf,krf2;
1443 __m128 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1446 __m128 one_sixth = _mm_set1_ps(1.0/6.0);
1447 __m128 one_twelfth = _mm_set1_ps(1.0/12.0);
1449 __m128i ifour = _mm_set1_epi32(4);
1450 __m128 rt,vfeps,twovfeps,vftabscale,Y,F,G,H,Fp,VV,FF;
1452 __m128 dummy_mask,cutoff_mask;
1453 __m128 signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
1454 __m128 one = _mm_set1_ps(1.0);
1455 __m128 two = _mm_set1_ps(2.0);
1461 jindex = nlist->jindex;
1463 shiftidx = nlist->shift;
1465 shiftvec = fr->shift_vec[0];
1466 fshift = fr->fshift[0];
1467 facel = _mm_set1_ps(fr->epsfac);
1468 charge = mdatoms->chargeA;
1469 nvdwtype = fr->ntype;
1470 vdwparam = fr->nbfp;
1471 vdwtype = mdatoms->typeA;
1473 vftab = kernel_data->table_elec_vdw->data;
1474 vftabscale = _mm_set1_ps(kernel_data->table_elec_vdw->scale);
1476 /* Setup water-specific parameters */
1477 inr = nlist->iinr[0];
1478 iq1 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1]));
1479 iq2 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2]));
1480 iq3 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+3]));
1481 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
1483 jq1 = _mm_set1_ps(charge[inr+1]);
1484 jq2 = _mm_set1_ps(charge[inr+2]);
1485 jq3 = _mm_set1_ps(charge[inr+3]);
1486 vdwjidx0A = 2*vdwtype[inr+0];
1487 c6_00 = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A]);
1488 c12_00 = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A+1]);
1489 qq11 = _mm_mul_ps(iq1,jq1);
1490 qq12 = _mm_mul_ps(iq1,jq2);
1491 qq13 = _mm_mul_ps(iq1,jq3);
1492 qq21 = _mm_mul_ps(iq2,jq1);
1493 qq22 = _mm_mul_ps(iq2,jq2);
1494 qq23 = _mm_mul_ps(iq2,jq3);
1495 qq31 = _mm_mul_ps(iq3,jq1);
1496 qq32 = _mm_mul_ps(iq3,jq2);
1497 qq33 = _mm_mul_ps(iq3,jq3);
1499 /* Avoid stupid compiler warnings */
1500 jnrA = jnrB = jnrC = jnrD = 0;
1501 j_coord_offsetA = 0;
1502 j_coord_offsetB = 0;
1503 j_coord_offsetC = 0;
1504 j_coord_offsetD = 0;
1509 for(iidx=0;iidx<4*DIM;iidx++)
1511 scratch[iidx] = 0.0;
1514 /* Start outer loop over neighborlists */
1515 for(iidx=0; iidx<nri; iidx++)
1517 /* Load shift vector for this list */
1518 i_shift_offset = DIM*shiftidx[iidx];
1520 /* Load limits for loop over neighbors */
1521 j_index_start = jindex[iidx];
1522 j_index_end = jindex[iidx+1];
1524 /* Get outer coordinate index */
1526 i_coord_offset = DIM*inr;
1528 /* Load i particle coords and add shift vector */
1529 gmx_mm_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1530 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1532 fix0 = _mm_setzero_ps();
1533 fiy0 = _mm_setzero_ps();
1534 fiz0 = _mm_setzero_ps();
1535 fix1 = _mm_setzero_ps();
1536 fiy1 = _mm_setzero_ps();
1537 fiz1 = _mm_setzero_ps();
1538 fix2 = _mm_setzero_ps();
1539 fiy2 = _mm_setzero_ps();
1540 fiz2 = _mm_setzero_ps();
1541 fix3 = _mm_setzero_ps();
1542 fiy3 = _mm_setzero_ps();
1543 fiz3 = _mm_setzero_ps();
1545 /* Start inner kernel loop */
1546 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
1549 /* Get j neighbor index, and coordinate index */
1551 jnrB = jjnr[jidx+1];
1552 jnrC = jjnr[jidx+2];
1553 jnrD = jjnr[jidx+3];
1554 j_coord_offsetA = DIM*jnrA;
1555 j_coord_offsetB = DIM*jnrB;
1556 j_coord_offsetC = DIM*jnrC;
1557 j_coord_offsetD = DIM*jnrD;
1559 /* load j atom coordinates */
1560 gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1561 x+j_coord_offsetC,x+j_coord_offsetD,
1562 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1563 &jy2,&jz2,&jx3,&jy3,&jz3);
1565 /* Calculate displacement vector */
1566 dx00 = _mm_sub_ps(ix0,jx0);
1567 dy00 = _mm_sub_ps(iy0,jy0);
1568 dz00 = _mm_sub_ps(iz0,jz0);
1569 dx11 = _mm_sub_ps(ix1,jx1);
1570 dy11 = _mm_sub_ps(iy1,jy1);
1571 dz11 = _mm_sub_ps(iz1,jz1);
1572 dx12 = _mm_sub_ps(ix1,jx2);
1573 dy12 = _mm_sub_ps(iy1,jy2);
1574 dz12 = _mm_sub_ps(iz1,jz2);
1575 dx13 = _mm_sub_ps(ix1,jx3);
1576 dy13 = _mm_sub_ps(iy1,jy3);
1577 dz13 = _mm_sub_ps(iz1,jz3);
1578 dx21 = _mm_sub_ps(ix2,jx1);
1579 dy21 = _mm_sub_ps(iy2,jy1);
1580 dz21 = _mm_sub_ps(iz2,jz1);
1581 dx22 = _mm_sub_ps(ix2,jx2);
1582 dy22 = _mm_sub_ps(iy2,jy2);
1583 dz22 = _mm_sub_ps(iz2,jz2);
1584 dx23 = _mm_sub_ps(ix2,jx3);
1585 dy23 = _mm_sub_ps(iy2,jy3);
1586 dz23 = _mm_sub_ps(iz2,jz3);
1587 dx31 = _mm_sub_ps(ix3,jx1);
1588 dy31 = _mm_sub_ps(iy3,jy1);
1589 dz31 = _mm_sub_ps(iz3,jz1);
1590 dx32 = _mm_sub_ps(ix3,jx2);
1591 dy32 = _mm_sub_ps(iy3,jy2);
1592 dz32 = _mm_sub_ps(iz3,jz2);
1593 dx33 = _mm_sub_ps(ix3,jx3);
1594 dy33 = _mm_sub_ps(iy3,jy3);
1595 dz33 = _mm_sub_ps(iz3,jz3);
1597 /* Calculate squared distance and things based on it */
1598 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
1599 rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
1600 rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
1601 rsq13 = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
1602 rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
1603 rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
1604 rsq23 = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
1605 rsq31 = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
1606 rsq32 = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
1607 rsq33 = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
1609 rinv00 = gmx_mm_invsqrt_ps(rsq00);
1610 rinv11 = gmx_mm_invsqrt_ps(rsq11);
1611 rinv12 = gmx_mm_invsqrt_ps(rsq12);
1612 rinv13 = gmx_mm_invsqrt_ps(rsq13);
1613 rinv21 = gmx_mm_invsqrt_ps(rsq21);
1614 rinv22 = gmx_mm_invsqrt_ps(rsq22);
1615 rinv23 = gmx_mm_invsqrt_ps(rsq23);
1616 rinv31 = gmx_mm_invsqrt_ps(rsq31);
1617 rinv32 = gmx_mm_invsqrt_ps(rsq32);
1618 rinv33 = gmx_mm_invsqrt_ps(rsq33);
1620 fjx0 = _mm_setzero_ps();
1621 fjy0 = _mm_setzero_ps();
1622 fjz0 = _mm_setzero_ps();
1623 fjx1 = _mm_setzero_ps();
1624 fjy1 = _mm_setzero_ps();
1625 fjz1 = _mm_setzero_ps();
1626 fjx2 = _mm_setzero_ps();
1627 fjy2 = _mm_setzero_ps();
1628 fjz2 = _mm_setzero_ps();
1629 fjx3 = _mm_setzero_ps();
1630 fjy3 = _mm_setzero_ps();
1631 fjz3 = _mm_setzero_ps();
1633 /**************************
1634 * CALCULATE INTERACTIONS *
1635 **************************/
1637 r00 = _mm_mul_ps(rsq00,rinv00);
1639 /* Calculate table index by multiplying r with table scale and truncate to integer */
1640 rt = _mm_mul_ps(r00,vftabscale);
1641 vfitab = _mm_cvttps_epi32(rt);
1643 vfeps = _mm_frcz_ps(rt);
1645 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1647 twovfeps = _mm_add_ps(vfeps,vfeps);
1648 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1650 /* CUBIC SPLINE TABLE DISPERSION */
1651 vfitab = _mm_add_epi32(vfitab,ifour);
1652 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
1653 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
1654 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
1655 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
1656 _MM_TRANSPOSE4_PS(Y,F,G,H);
1657 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
1658 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
1659 fvdw6 = _mm_mul_ps(c6_00,FF);
1661 /* CUBIC SPLINE TABLE REPULSION */
1662 vfitab = _mm_add_epi32(vfitab,ifour);
1663 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
1664 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
1665 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
1666 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
1667 _MM_TRANSPOSE4_PS(Y,F,G,H);
1668 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
1669 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
1670 fvdw12 = _mm_mul_ps(c12_00,FF);
1671 fvdw = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
1675 /* Update vectorial force */
1676 fix0 = _mm_macc_ps(dx00,fscal,fix0);
1677 fiy0 = _mm_macc_ps(dy00,fscal,fiy0);
1678 fiz0 = _mm_macc_ps(dz00,fscal,fiz0);
1680 fjx0 = _mm_macc_ps(dx00,fscal,fjx0);
1681 fjy0 = _mm_macc_ps(dy00,fscal,fjy0);
1682 fjz0 = _mm_macc_ps(dz00,fscal,fjz0);
1684 /**************************
1685 * CALCULATE INTERACTIONS *
1686 **************************/
1688 r11 = _mm_mul_ps(rsq11,rinv11);
1690 /* Calculate table index by multiplying r with table scale and truncate to integer */
1691 rt = _mm_mul_ps(r11,vftabscale);
1692 vfitab = _mm_cvttps_epi32(rt);
1694 vfeps = _mm_frcz_ps(rt);
1696 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1698 twovfeps = _mm_add_ps(vfeps,vfeps);
1699 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1701 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1702 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
1703 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
1704 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
1705 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
1706 _MM_TRANSPOSE4_PS(Y,F,G,H);
1707 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
1708 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
1709 felec = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq11,FF),_mm_mul_ps(vftabscale,rinv11)));
1713 /* Update vectorial force */
1714 fix1 = _mm_macc_ps(dx11,fscal,fix1);
1715 fiy1 = _mm_macc_ps(dy11,fscal,fiy1);
1716 fiz1 = _mm_macc_ps(dz11,fscal,fiz1);
1718 fjx1 = _mm_macc_ps(dx11,fscal,fjx1);
1719 fjy1 = _mm_macc_ps(dy11,fscal,fjy1);
1720 fjz1 = _mm_macc_ps(dz11,fscal,fjz1);
1722 /**************************
1723 * CALCULATE INTERACTIONS *
1724 **************************/
1726 r12 = _mm_mul_ps(rsq12,rinv12);
1728 /* Calculate table index by multiplying r with table scale and truncate to integer */
1729 rt = _mm_mul_ps(r12,vftabscale);
1730 vfitab = _mm_cvttps_epi32(rt);
1732 vfeps = _mm_frcz_ps(rt);
1734 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1736 twovfeps = _mm_add_ps(vfeps,vfeps);
1737 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1739 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1740 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
1741 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
1742 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
1743 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
1744 _MM_TRANSPOSE4_PS(Y,F,G,H);
1745 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
1746 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
1747 felec = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq12,FF),_mm_mul_ps(vftabscale,rinv12)));
1751 /* Update vectorial force */
1752 fix1 = _mm_macc_ps(dx12,fscal,fix1);
1753 fiy1 = _mm_macc_ps(dy12,fscal,fiy1);
1754 fiz1 = _mm_macc_ps(dz12,fscal,fiz1);
1756 fjx2 = _mm_macc_ps(dx12,fscal,fjx2);
1757 fjy2 = _mm_macc_ps(dy12,fscal,fjy2);
1758 fjz2 = _mm_macc_ps(dz12,fscal,fjz2);
1760 /**************************
1761 * CALCULATE INTERACTIONS *
1762 **************************/
1764 r13 = _mm_mul_ps(rsq13,rinv13);
1766 /* Calculate table index by multiplying r with table scale and truncate to integer */
1767 rt = _mm_mul_ps(r13,vftabscale);
1768 vfitab = _mm_cvttps_epi32(rt);
1770 vfeps = _mm_frcz_ps(rt);
1772 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1774 twovfeps = _mm_add_ps(vfeps,vfeps);
1775 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1777 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1778 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
1779 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
1780 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
1781 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
1782 _MM_TRANSPOSE4_PS(Y,F,G,H);
1783 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
1784 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
1785 felec = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq13,FF),_mm_mul_ps(vftabscale,rinv13)));
1789 /* Update vectorial force */
1790 fix1 = _mm_macc_ps(dx13,fscal,fix1);
1791 fiy1 = _mm_macc_ps(dy13,fscal,fiy1);
1792 fiz1 = _mm_macc_ps(dz13,fscal,fiz1);
1794 fjx3 = _mm_macc_ps(dx13,fscal,fjx3);
1795 fjy3 = _mm_macc_ps(dy13,fscal,fjy3);
1796 fjz3 = _mm_macc_ps(dz13,fscal,fjz3);
1798 /**************************
1799 * CALCULATE INTERACTIONS *
1800 **************************/
1802 r21 = _mm_mul_ps(rsq21,rinv21);
1804 /* Calculate table index by multiplying r with table scale and truncate to integer */
1805 rt = _mm_mul_ps(r21,vftabscale);
1806 vfitab = _mm_cvttps_epi32(rt);
1808 vfeps = _mm_frcz_ps(rt);
1810 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1812 twovfeps = _mm_add_ps(vfeps,vfeps);
1813 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1815 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1816 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
1817 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
1818 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
1819 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
1820 _MM_TRANSPOSE4_PS(Y,F,G,H);
1821 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
1822 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
1823 felec = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq21,FF),_mm_mul_ps(vftabscale,rinv21)));
1827 /* Update vectorial force */
1828 fix2 = _mm_macc_ps(dx21,fscal,fix2);
1829 fiy2 = _mm_macc_ps(dy21,fscal,fiy2);
1830 fiz2 = _mm_macc_ps(dz21,fscal,fiz2);
1832 fjx1 = _mm_macc_ps(dx21,fscal,fjx1);
1833 fjy1 = _mm_macc_ps(dy21,fscal,fjy1);
1834 fjz1 = _mm_macc_ps(dz21,fscal,fjz1);
1836 /**************************
1837 * CALCULATE INTERACTIONS *
1838 **************************/
1840 r22 = _mm_mul_ps(rsq22,rinv22);
1842 /* Calculate table index by multiplying r with table scale and truncate to integer */
1843 rt = _mm_mul_ps(r22,vftabscale);
1844 vfitab = _mm_cvttps_epi32(rt);
1846 vfeps = _mm_frcz_ps(rt);
1848 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1850 twovfeps = _mm_add_ps(vfeps,vfeps);
1851 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1853 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1854 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
1855 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
1856 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
1857 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
1858 _MM_TRANSPOSE4_PS(Y,F,G,H);
1859 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
1860 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
1861 felec = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq22,FF),_mm_mul_ps(vftabscale,rinv22)));
1865 /* Update vectorial force */
1866 fix2 = _mm_macc_ps(dx22,fscal,fix2);
1867 fiy2 = _mm_macc_ps(dy22,fscal,fiy2);
1868 fiz2 = _mm_macc_ps(dz22,fscal,fiz2);
1870 fjx2 = _mm_macc_ps(dx22,fscal,fjx2);
1871 fjy2 = _mm_macc_ps(dy22,fscal,fjy2);
1872 fjz2 = _mm_macc_ps(dz22,fscal,fjz2);
1874 /**************************
1875 * CALCULATE INTERACTIONS *
1876 **************************/
1878 r23 = _mm_mul_ps(rsq23,rinv23);
1880 /* Calculate table index by multiplying r with table scale and truncate to integer */
1881 rt = _mm_mul_ps(r23,vftabscale);
1882 vfitab = _mm_cvttps_epi32(rt);
1884 vfeps = _mm_frcz_ps(rt);
1886 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1888 twovfeps = _mm_add_ps(vfeps,vfeps);
1889 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1891 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1892 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
1893 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
1894 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
1895 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
1896 _MM_TRANSPOSE4_PS(Y,F,G,H);
1897 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
1898 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
1899 felec = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq23,FF),_mm_mul_ps(vftabscale,rinv23)));
1903 /* Update vectorial force */
1904 fix2 = _mm_macc_ps(dx23,fscal,fix2);
1905 fiy2 = _mm_macc_ps(dy23,fscal,fiy2);
1906 fiz2 = _mm_macc_ps(dz23,fscal,fiz2);
1908 fjx3 = _mm_macc_ps(dx23,fscal,fjx3);
1909 fjy3 = _mm_macc_ps(dy23,fscal,fjy3);
1910 fjz3 = _mm_macc_ps(dz23,fscal,fjz3);
1912 /**************************
1913 * CALCULATE INTERACTIONS *
1914 **************************/
1916 r31 = _mm_mul_ps(rsq31,rinv31);
1918 /* Calculate table index by multiplying r with table scale and truncate to integer */
1919 rt = _mm_mul_ps(r31,vftabscale);
1920 vfitab = _mm_cvttps_epi32(rt);
1922 vfeps = _mm_frcz_ps(rt);
1924 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1926 twovfeps = _mm_add_ps(vfeps,vfeps);
1927 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1929 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1930 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
1931 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
1932 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
1933 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
1934 _MM_TRANSPOSE4_PS(Y,F,G,H);
1935 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
1936 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
1937 felec = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq31,FF),_mm_mul_ps(vftabscale,rinv31)));
1941 /* Update vectorial force */
1942 fix3 = _mm_macc_ps(dx31,fscal,fix3);
1943 fiy3 = _mm_macc_ps(dy31,fscal,fiy3);
1944 fiz3 = _mm_macc_ps(dz31,fscal,fiz3);
1946 fjx1 = _mm_macc_ps(dx31,fscal,fjx1);
1947 fjy1 = _mm_macc_ps(dy31,fscal,fjy1);
1948 fjz1 = _mm_macc_ps(dz31,fscal,fjz1);
1950 /**************************
1951 * CALCULATE INTERACTIONS *
1952 **************************/
1954 r32 = _mm_mul_ps(rsq32,rinv32);
1956 /* Calculate table index by multiplying r with table scale and truncate to integer */
1957 rt = _mm_mul_ps(r32,vftabscale);
1958 vfitab = _mm_cvttps_epi32(rt);
1960 vfeps = _mm_frcz_ps(rt);
1962 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1964 twovfeps = _mm_add_ps(vfeps,vfeps);
1965 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1967 /* CUBIC SPLINE TABLE ELECTROSTATICS */
1968 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
1969 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
1970 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
1971 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
1972 _MM_TRANSPOSE4_PS(Y,F,G,H);
1973 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
1974 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
1975 felec = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq32,FF),_mm_mul_ps(vftabscale,rinv32)));
1979 /* Update vectorial force */
1980 fix3 = _mm_macc_ps(dx32,fscal,fix3);
1981 fiy3 = _mm_macc_ps(dy32,fscal,fiy3);
1982 fiz3 = _mm_macc_ps(dz32,fscal,fiz3);
1984 fjx2 = _mm_macc_ps(dx32,fscal,fjx2);
1985 fjy2 = _mm_macc_ps(dy32,fscal,fjy2);
1986 fjz2 = _mm_macc_ps(dz32,fscal,fjz2);
1988 /**************************
1989 * CALCULATE INTERACTIONS *
1990 **************************/
1992 r33 = _mm_mul_ps(rsq33,rinv33);
1994 /* Calculate table index by multiplying r with table scale and truncate to integer */
1995 rt = _mm_mul_ps(r33,vftabscale);
1996 vfitab = _mm_cvttps_epi32(rt);
1998 vfeps = _mm_frcz_ps(rt);
2000 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
2002 twovfeps = _mm_add_ps(vfeps,vfeps);
2003 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2005 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2006 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
2007 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
2008 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
2009 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
2010 _MM_TRANSPOSE4_PS(Y,F,G,H);
2011 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
2012 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
2013 felec = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq33,FF),_mm_mul_ps(vftabscale,rinv33)));
2017 /* Update vectorial force */
2018 fix3 = _mm_macc_ps(dx33,fscal,fix3);
2019 fiy3 = _mm_macc_ps(dy33,fscal,fiy3);
2020 fiz3 = _mm_macc_ps(dz33,fscal,fiz3);
2022 fjx3 = _mm_macc_ps(dx33,fscal,fjx3);
2023 fjy3 = _mm_macc_ps(dy33,fscal,fjy3);
2024 fjz3 = _mm_macc_ps(dz33,fscal,fjz3);
2026 fjptrA = f+j_coord_offsetA;
2027 fjptrB = f+j_coord_offsetB;
2028 fjptrC = f+j_coord_offsetC;
2029 fjptrD = f+j_coord_offsetD;
2031 gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
2032 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
2033 fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2035 /* Inner loop uses 432 flops */
2038 if(jidx<j_index_end)
2041 /* Get j neighbor index, and coordinate index */
2042 jnrlistA = jjnr[jidx];
2043 jnrlistB = jjnr[jidx+1];
2044 jnrlistC = jjnr[jidx+2];
2045 jnrlistD = jjnr[jidx+3];
2046 /* Sign of each element will be negative for non-real atoms.
2047 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
2048 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
2050 dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
2051 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
2052 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
2053 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
2054 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
2055 j_coord_offsetA = DIM*jnrA;
2056 j_coord_offsetB = DIM*jnrB;
2057 j_coord_offsetC = DIM*jnrC;
2058 j_coord_offsetD = DIM*jnrD;
2060 /* load j atom coordinates */
2061 gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
2062 x+j_coord_offsetC,x+j_coord_offsetD,
2063 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
2064 &jy2,&jz2,&jx3,&jy3,&jz3);
2066 /* Calculate displacement vector */
2067 dx00 = _mm_sub_ps(ix0,jx0);
2068 dy00 = _mm_sub_ps(iy0,jy0);
2069 dz00 = _mm_sub_ps(iz0,jz0);
2070 dx11 = _mm_sub_ps(ix1,jx1);
2071 dy11 = _mm_sub_ps(iy1,jy1);
2072 dz11 = _mm_sub_ps(iz1,jz1);
2073 dx12 = _mm_sub_ps(ix1,jx2);
2074 dy12 = _mm_sub_ps(iy1,jy2);
2075 dz12 = _mm_sub_ps(iz1,jz2);
2076 dx13 = _mm_sub_ps(ix1,jx3);
2077 dy13 = _mm_sub_ps(iy1,jy3);
2078 dz13 = _mm_sub_ps(iz1,jz3);
2079 dx21 = _mm_sub_ps(ix2,jx1);
2080 dy21 = _mm_sub_ps(iy2,jy1);
2081 dz21 = _mm_sub_ps(iz2,jz1);
2082 dx22 = _mm_sub_ps(ix2,jx2);
2083 dy22 = _mm_sub_ps(iy2,jy2);
2084 dz22 = _mm_sub_ps(iz2,jz2);
2085 dx23 = _mm_sub_ps(ix2,jx3);
2086 dy23 = _mm_sub_ps(iy2,jy3);
2087 dz23 = _mm_sub_ps(iz2,jz3);
2088 dx31 = _mm_sub_ps(ix3,jx1);
2089 dy31 = _mm_sub_ps(iy3,jy1);
2090 dz31 = _mm_sub_ps(iz3,jz1);
2091 dx32 = _mm_sub_ps(ix3,jx2);
2092 dy32 = _mm_sub_ps(iy3,jy2);
2093 dz32 = _mm_sub_ps(iz3,jz2);
2094 dx33 = _mm_sub_ps(ix3,jx3);
2095 dy33 = _mm_sub_ps(iy3,jy3);
2096 dz33 = _mm_sub_ps(iz3,jz3);
2098 /* Calculate squared distance and things based on it */
2099 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
2100 rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
2101 rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
2102 rsq13 = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
2103 rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
2104 rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
2105 rsq23 = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
2106 rsq31 = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
2107 rsq32 = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
2108 rsq33 = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
2110 rinv00 = gmx_mm_invsqrt_ps(rsq00);
2111 rinv11 = gmx_mm_invsqrt_ps(rsq11);
2112 rinv12 = gmx_mm_invsqrt_ps(rsq12);
2113 rinv13 = gmx_mm_invsqrt_ps(rsq13);
2114 rinv21 = gmx_mm_invsqrt_ps(rsq21);
2115 rinv22 = gmx_mm_invsqrt_ps(rsq22);
2116 rinv23 = gmx_mm_invsqrt_ps(rsq23);
2117 rinv31 = gmx_mm_invsqrt_ps(rsq31);
2118 rinv32 = gmx_mm_invsqrt_ps(rsq32);
2119 rinv33 = gmx_mm_invsqrt_ps(rsq33);
2121 fjx0 = _mm_setzero_ps();
2122 fjy0 = _mm_setzero_ps();
2123 fjz0 = _mm_setzero_ps();
2124 fjx1 = _mm_setzero_ps();
2125 fjy1 = _mm_setzero_ps();
2126 fjz1 = _mm_setzero_ps();
2127 fjx2 = _mm_setzero_ps();
2128 fjy2 = _mm_setzero_ps();
2129 fjz2 = _mm_setzero_ps();
2130 fjx3 = _mm_setzero_ps();
2131 fjy3 = _mm_setzero_ps();
2132 fjz3 = _mm_setzero_ps();
2134 /**************************
2135 * CALCULATE INTERACTIONS *
2136 **************************/
2138 r00 = _mm_mul_ps(rsq00,rinv00);
2139 r00 = _mm_andnot_ps(dummy_mask,r00);
2141 /* Calculate table index by multiplying r with table scale and truncate to integer */
2142 rt = _mm_mul_ps(r00,vftabscale);
2143 vfitab = _mm_cvttps_epi32(rt);
2145 vfeps = _mm_frcz_ps(rt);
2147 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
2149 twovfeps = _mm_add_ps(vfeps,vfeps);
2150 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2152 /* CUBIC SPLINE TABLE DISPERSION */
2153 vfitab = _mm_add_epi32(vfitab,ifour);
2154 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
2155 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
2156 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
2157 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
2158 _MM_TRANSPOSE4_PS(Y,F,G,H);
2159 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
2160 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
2161 fvdw6 = _mm_mul_ps(c6_00,FF);
2163 /* CUBIC SPLINE TABLE REPULSION */
2164 vfitab = _mm_add_epi32(vfitab,ifour);
2165 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
2166 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
2167 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
2168 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
2169 _MM_TRANSPOSE4_PS(Y,F,G,H);
2170 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
2171 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
2172 fvdw12 = _mm_mul_ps(c12_00,FF);
2173 fvdw = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
2177 fscal = _mm_andnot_ps(dummy_mask,fscal);
2179 /* Update vectorial force */
2180 fix0 = _mm_macc_ps(dx00,fscal,fix0);
2181 fiy0 = _mm_macc_ps(dy00,fscal,fiy0);
2182 fiz0 = _mm_macc_ps(dz00,fscal,fiz0);
2184 fjx0 = _mm_macc_ps(dx00,fscal,fjx0);
2185 fjy0 = _mm_macc_ps(dy00,fscal,fjy0);
2186 fjz0 = _mm_macc_ps(dz00,fscal,fjz0);
2188 /**************************
2189 * CALCULATE INTERACTIONS *
2190 **************************/
2192 r11 = _mm_mul_ps(rsq11,rinv11);
2193 r11 = _mm_andnot_ps(dummy_mask,r11);
2195 /* Calculate table index by multiplying r with table scale and truncate to integer */
2196 rt = _mm_mul_ps(r11,vftabscale);
2197 vfitab = _mm_cvttps_epi32(rt);
2199 vfeps = _mm_frcz_ps(rt);
2201 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
2203 twovfeps = _mm_add_ps(vfeps,vfeps);
2204 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2206 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2207 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
2208 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
2209 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
2210 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
2211 _MM_TRANSPOSE4_PS(Y,F,G,H);
2212 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
2213 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
2214 felec = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq11,FF),_mm_mul_ps(vftabscale,rinv11)));
2218 fscal = _mm_andnot_ps(dummy_mask,fscal);
2220 /* Update vectorial force */
2221 fix1 = _mm_macc_ps(dx11,fscal,fix1);
2222 fiy1 = _mm_macc_ps(dy11,fscal,fiy1);
2223 fiz1 = _mm_macc_ps(dz11,fscal,fiz1);
2225 fjx1 = _mm_macc_ps(dx11,fscal,fjx1);
2226 fjy1 = _mm_macc_ps(dy11,fscal,fjy1);
2227 fjz1 = _mm_macc_ps(dz11,fscal,fjz1);
2229 /**************************
2230 * CALCULATE INTERACTIONS *
2231 **************************/
2233 r12 = _mm_mul_ps(rsq12,rinv12);
2234 r12 = _mm_andnot_ps(dummy_mask,r12);
2236 /* Calculate table index by multiplying r with table scale and truncate to integer */
2237 rt = _mm_mul_ps(r12,vftabscale);
2238 vfitab = _mm_cvttps_epi32(rt);
2240 vfeps = _mm_frcz_ps(rt);
2242 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
2244 twovfeps = _mm_add_ps(vfeps,vfeps);
2245 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2247 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2248 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
2249 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
2250 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
2251 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
2252 _MM_TRANSPOSE4_PS(Y,F,G,H);
2253 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
2254 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
2255 felec = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq12,FF),_mm_mul_ps(vftabscale,rinv12)));
2259 fscal = _mm_andnot_ps(dummy_mask,fscal);
2261 /* Update vectorial force */
2262 fix1 = _mm_macc_ps(dx12,fscal,fix1);
2263 fiy1 = _mm_macc_ps(dy12,fscal,fiy1);
2264 fiz1 = _mm_macc_ps(dz12,fscal,fiz1);
2266 fjx2 = _mm_macc_ps(dx12,fscal,fjx2);
2267 fjy2 = _mm_macc_ps(dy12,fscal,fjy2);
2268 fjz2 = _mm_macc_ps(dz12,fscal,fjz2);
2270 /**************************
2271 * CALCULATE INTERACTIONS *
2272 **************************/
2274 r13 = _mm_mul_ps(rsq13,rinv13);
2275 r13 = _mm_andnot_ps(dummy_mask,r13);
2277 /* Calculate table index by multiplying r with table scale and truncate to integer */
2278 rt = _mm_mul_ps(r13,vftabscale);
2279 vfitab = _mm_cvttps_epi32(rt);
2281 vfeps = _mm_frcz_ps(rt);
2283 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
2285 twovfeps = _mm_add_ps(vfeps,vfeps);
2286 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2288 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2289 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
2290 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
2291 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
2292 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
2293 _MM_TRANSPOSE4_PS(Y,F,G,H);
2294 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
2295 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
2296 felec = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq13,FF),_mm_mul_ps(vftabscale,rinv13)));
2300 fscal = _mm_andnot_ps(dummy_mask,fscal);
2302 /* Update vectorial force */
2303 fix1 = _mm_macc_ps(dx13,fscal,fix1);
2304 fiy1 = _mm_macc_ps(dy13,fscal,fiy1);
2305 fiz1 = _mm_macc_ps(dz13,fscal,fiz1);
2307 fjx3 = _mm_macc_ps(dx13,fscal,fjx3);
2308 fjy3 = _mm_macc_ps(dy13,fscal,fjy3);
2309 fjz3 = _mm_macc_ps(dz13,fscal,fjz3);
2311 /**************************
2312 * CALCULATE INTERACTIONS *
2313 **************************/
2315 r21 = _mm_mul_ps(rsq21,rinv21);
2316 r21 = _mm_andnot_ps(dummy_mask,r21);
2318 /* Calculate table index by multiplying r with table scale and truncate to integer */
2319 rt = _mm_mul_ps(r21,vftabscale);
2320 vfitab = _mm_cvttps_epi32(rt);
2322 vfeps = _mm_frcz_ps(rt);
2324 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
2326 twovfeps = _mm_add_ps(vfeps,vfeps);
2327 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2329 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2330 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
2331 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
2332 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
2333 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
2334 _MM_TRANSPOSE4_PS(Y,F,G,H);
2335 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
2336 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
2337 felec = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq21,FF),_mm_mul_ps(vftabscale,rinv21)));
2341 fscal = _mm_andnot_ps(dummy_mask,fscal);
2343 /* Update vectorial force */
2344 fix2 = _mm_macc_ps(dx21,fscal,fix2);
2345 fiy2 = _mm_macc_ps(dy21,fscal,fiy2);
2346 fiz2 = _mm_macc_ps(dz21,fscal,fiz2);
2348 fjx1 = _mm_macc_ps(dx21,fscal,fjx1);
2349 fjy1 = _mm_macc_ps(dy21,fscal,fjy1);
2350 fjz1 = _mm_macc_ps(dz21,fscal,fjz1);
2352 /**************************
2353 * CALCULATE INTERACTIONS *
2354 **************************/
2356 r22 = _mm_mul_ps(rsq22,rinv22);
2357 r22 = _mm_andnot_ps(dummy_mask,r22);
2359 /* Calculate table index by multiplying r with table scale and truncate to integer */
2360 rt = _mm_mul_ps(r22,vftabscale);
2361 vfitab = _mm_cvttps_epi32(rt);
2363 vfeps = _mm_frcz_ps(rt);
2365 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
2367 twovfeps = _mm_add_ps(vfeps,vfeps);
2368 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2370 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2371 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
2372 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
2373 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
2374 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
2375 _MM_TRANSPOSE4_PS(Y,F,G,H);
2376 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
2377 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
2378 felec = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq22,FF),_mm_mul_ps(vftabscale,rinv22)));
2382 fscal = _mm_andnot_ps(dummy_mask,fscal);
2384 /* Update vectorial force */
2385 fix2 = _mm_macc_ps(dx22,fscal,fix2);
2386 fiy2 = _mm_macc_ps(dy22,fscal,fiy2);
2387 fiz2 = _mm_macc_ps(dz22,fscal,fiz2);
2389 fjx2 = _mm_macc_ps(dx22,fscal,fjx2);
2390 fjy2 = _mm_macc_ps(dy22,fscal,fjy2);
2391 fjz2 = _mm_macc_ps(dz22,fscal,fjz2);
2393 /**************************
2394 * CALCULATE INTERACTIONS *
2395 **************************/
2397 r23 = _mm_mul_ps(rsq23,rinv23);
2398 r23 = _mm_andnot_ps(dummy_mask,r23);
2400 /* Calculate table index by multiplying r with table scale and truncate to integer */
2401 rt = _mm_mul_ps(r23,vftabscale);
2402 vfitab = _mm_cvttps_epi32(rt);
2404 vfeps = _mm_frcz_ps(rt);
2406 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
2408 twovfeps = _mm_add_ps(vfeps,vfeps);
2409 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2411 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2412 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
2413 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
2414 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
2415 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
2416 _MM_TRANSPOSE4_PS(Y,F,G,H);
2417 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
2418 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
2419 felec = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq23,FF),_mm_mul_ps(vftabscale,rinv23)));
2423 fscal = _mm_andnot_ps(dummy_mask,fscal);
2425 /* Update vectorial force */
2426 fix2 = _mm_macc_ps(dx23,fscal,fix2);
2427 fiy2 = _mm_macc_ps(dy23,fscal,fiy2);
2428 fiz2 = _mm_macc_ps(dz23,fscal,fiz2);
2430 fjx3 = _mm_macc_ps(dx23,fscal,fjx3);
2431 fjy3 = _mm_macc_ps(dy23,fscal,fjy3);
2432 fjz3 = _mm_macc_ps(dz23,fscal,fjz3);
2434 /**************************
2435 * CALCULATE INTERACTIONS *
2436 **************************/
2438 r31 = _mm_mul_ps(rsq31,rinv31);
2439 r31 = _mm_andnot_ps(dummy_mask,r31);
2441 /* Calculate table index by multiplying r with table scale and truncate to integer */
2442 rt = _mm_mul_ps(r31,vftabscale);
2443 vfitab = _mm_cvttps_epi32(rt);
2445 vfeps = _mm_frcz_ps(rt);
2447 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
2449 twovfeps = _mm_add_ps(vfeps,vfeps);
2450 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2452 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2453 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
2454 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
2455 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
2456 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
2457 _MM_TRANSPOSE4_PS(Y,F,G,H);
2458 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
2459 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
2460 felec = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq31,FF),_mm_mul_ps(vftabscale,rinv31)));
2464 fscal = _mm_andnot_ps(dummy_mask,fscal);
2466 /* Update vectorial force */
2467 fix3 = _mm_macc_ps(dx31,fscal,fix3);
2468 fiy3 = _mm_macc_ps(dy31,fscal,fiy3);
2469 fiz3 = _mm_macc_ps(dz31,fscal,fiz3);
2471 fjx1 = _mm_macc_ps(dx31,fscal,fjx1);
2472 fjy1 = _mm_macc_ps(dy31,fscal,fjy1);
2473 fjz1 = _mm_macc_ps(dz31,fscal,fjz1);
2475 /**************************
2476 * CALCULATE INTERACTIONS *
2477 **************************/
2479 r32 = _mm_mul_ps(rsq32,rinv32);
2480 r32 = _mm_andnot_ps(dummy_mask,r32);
2482 /* Calculate table index by multiplying r with table scale and truncate to integer */
2483 rt = _mm_mul_ps(r32,vftabscale);
2484 vfitab = _mm_cvttps_epi32(rt);
2486 vfeps = _mm_frcz_ps(rt);
2488 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
2490 twovfeps = _mm_add_ps(vfeps,vfeps);
2491 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2493 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2494 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
2495 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
2496 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
2497 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
2498 _MM_TRANSPOSE4_PS(Y,F,G,H);
2499 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
2500 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
2501 felec = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq32,FF),_mm_mul_ps(vftabscale,rinv32)));
2505 fscal = _mm_andnot_ps(dummy_mask,fscal);
2507 /* Update vectorial force */
2508 fix3 = _mm_macc_ps(dx32,fscal,fix3);
2509 fiy3 = _mm_macc_ps(dy32,fscal,fiy3);
2510 fiz3 = _mm_macc_ps(dz32,fscal,fiz3);
2512 fjx2 = _mm_macc_ps(dx32,fscal,fjx2);
2513 fjy2 = _mm_macc_ps(dy32,fscal,fjy2);
2514 fjz2 = _mm_macc_ps(dz32,fscal,fjz2);
2516 /**************************
2517 * CALCULATE INTERACTIONS *
2518 **************************/
2520 r33 = _mm_mul_ps(rsq33,rinv33);
2521 r33 = _mm_andnot_ps(dummy_mask,r33);
2523 /* Calculate table index by multiplying r with table scale and truncate to integer */
2524 rt = _mm_mul_ps(r33,vftabscale);
2525 vfitab = _mm_cvttps_epi32(rt);
2527 vfeps = _mm_frcz_ps(rt);
2529 vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
2531 twovfeps = _mm_add_ps(vfeps,vfeps);
2532 vfitab = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2534 /* CUBIC SPLINE TABLE ELECTROSTATICS */
2535 Y = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
2536 F = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
2537 G = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
2538 H = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
2539 _MM_TRANSPOSE4_PS(Y,F,G,H);
2540 Fp = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
2541 FF = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
2542 felec = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq33,FF),_mm_mul_ps(vftabscale,rinv33)));
2546 fscal = _mm_andnot_ps(dummy_mask,fscal);
2548 /* Update vectorial force */
2549 fix3 = _mm_macc_ps(dx33,fscal,fix3);
2550 fiy3 = _mm_macc_ps(dy33,fscal,fiy3);
2551 fiz3 = _mm_macc_ps(dz33,fscal,fiz3);
2553 fjx3 = _mm_macc_ps(dx33,fscal,fjx3);
2554 fjy3 = _mm_macc_ps(dy33,fscal,fjy3);
2555 fjz3 = _mm_macc_ps(dz33,fscal,fjz3);
2557 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
2558 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
2559 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
2560 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
2562 gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
2563 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
2564 fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2566 /* Inner loop uses 442 flops */
2569 /* End of innermost loop */
2571 gmx_mm_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2572 f+i_coord_offset,fshift+i_shift_offset);
2574 /* Increment number of inner iterations */
2575 inneriter += j_index_end - j_index_start;
2577 /* Outer loop uses 24 flops */
2580 /* Increment number of outer iterations */
2583 /* Update outer/inner flops */
2585 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*442);