File: | gromacs/gmxlib/nonbonded/nb_kernel_sse4_1_single/nb_kernel_ElecCoul_VdwCSTab_GeomW4W4_sse4_1_single.c |
Location: | line 118, column 22 |
Description: | Value stored to 'one_twelfth' during its initialization is never read |
1 | /* |
2 | * This file is part of the GROMACS molecular simulation package. |
3 | * |
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. |
8 | * |
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. |
13 | * |
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. |
18 | * |
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. |
23 | * |
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. |
31 | * |
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. |
34 | */ |
35 | /* |
36 | * Note: this file was generated by the GROMACS sse4_1_single kernel generator. |
37 | */ |
38 | #ifdef HAVE_CONFIG_H1 |
39 | #include <config.h> |
40 | #endif |
41 | |
42 | #include <math.h> |
43 | |
44 | #include "../nb_kernel.h" |
45 | #include "types/simple.h" |
46 | #include "gromacs/math/vec.h" |
47 | #include "nrnb.h" |
48 | |
49 | #include "gromacs/simd/math_x86_sse4_1_single.h" |
50 | #include "kernelutil_x86_sse4_1_single.h" |
51 | |
52 | /* |
53 | * Gromacs nonbonded kernel: nb_kernel_ElecCoul_VdwCSTab_GeomW4W4_VF_sse4_1_single |
54 | * Electrostatics interaction: Coulomb |
55 | * VdW interaction: CubicSplineTable |
56 | * Geometry: Water4-Water4 |
57 | * Calculate force/pot: PotentialAndForce |
58 | */ |
59 | void |
60 | nb_kernel_ElecCoul_VdwCSTab_GeomW4W4_VF_sse4_1_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__attribute__ ((unused)) * gmx_restrict kernel_data, |
67 | t_nrnb * gmx_restrict nrnb) |
68 | { |
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. |
73 | */ |
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; |
80 | real rcutoff_scalar; |
81 | real *shiftvec,*fshift,*x,*f; |
82 | real *fjptrA,*fjptrB,*fjptrC,*fjptrD; |
83 | real scratch[4*DIM3]; |
84 | __m128 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall; |
85 | int vdwioffset0; |
86 | __m128 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0; |
87 | int vdwioffset1; |
88 | __m128 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1; |
89 | int vdwioffset2; |
90 | __m128 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2; |
91 | int vdwioffset3; |
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; |
112 | real *charge; |
113 | int nvdwtype; |
114 | __m128 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6; |
115 | int *vdwtype; |
116 | real *vdwparam; |
117 | __m128 one_sixth = _mm_set1_ps(1.0/6.0); |
118 | __m128 one_twelfth = _mm_set1_ps(1.0/12.0); |
Value stored to 'one_twelfth' during its initialization is never read | |
119 | __m128i vfitab; |
120 | __m128i ifour = _mm_set1_epi32(4); |
121 | __m128 rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF; |
122 | real *vftab; |
123 | __m128 dummy_mask,cutoff_mask; |
124 | __m128 signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) ); |
125 | __m128 one = _mm_set1_ps(1.0); |
126 | __m128 two = _mm_set1_ps(2.0); |
127 | x = xx[0]; |
128 | f = ff[0]; |
129 | |
130 | nri = nlist->nri; |
131 | iinr = nlist->iinr; |
132 | jindex = nlist->jindex; |
133 | jjnr = nlist->jjnr; |
134 | shiftidx = nlist->shift; |
135 | gid = nlist->gid; |
136 | shiftvec = fr->shift_vec[0]; |
137 | fshift = fr->fshift[0]; |
138 | facel = _mm_set1_ps(fr->epsfac); |
139 | charge = mdatoms->chargeA; |
140 | nvdwtype = fr->ntype; |
141 | vdwparam = fr->nbfp; |
142 | vdwtype = mdatoms->typeA; |
143 | |
144 | vftab = kernel_data->table_vdw->data; |
145 | vftabscale = _mm_set1_ps(kernel_data->table_vdw->scale); |
146 | |
147 | /* Setup water-specific parameters */ |
148 | inr = nlist->iinr[0]; |
149 | iq1 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1])); |
150 | iq2 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2])); |
151 | iq3 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+3])); |
152 | vdwioffset0 = 2*nvdwtype*vdwtype[inr+0]; |
153 | |
154 | jq1 = _mm_set1_ps(charge[inr+1]); |
155 | jq2 = _mm_set1_ps(charge[inr+2]); |
156 | jq3 = _mm_set1_ps(charge[inr+3]); |
157 | vdwjidx0A = 2*vdwtype[inr+0]; |
158 | c6_00 = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A]); |
159 | c12_00 = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A+1]); |
160 | qq11 = _mm_mul_ps(iq1,jq1); |
161 | qq12 = _mm_mul_ps(iq1,jq2); |
162 | qq13 = _mm_mul_ps(iq1,jq3); |
163 | qq21 = _mm_mul_ps(iq2,jq1); |
164 | qq22 = _mm_mul_ps(iq2,jq2); |
165 | qq23 = _mm_mul_ps(iq2,jq3); |
166 | qq31 = _mm_mul_ps(iq3,jq1); |
167 | qq32 = _mm_mul_ps(iq3,jq2); |
168 | qq33 = _mm_mul_ps(iq3,jq3); |
169 | |
170 | /* Avoid stupid compiler warnings */ |
171 | jnrA = jnrB = jnrC = jnrD = 0; |
172 | j_coord_offsetA = 0; |
173 | j_coord_offsetB = 0; |
174 | j_coord_offsetC = 0; |
175 | j_coord_offsetD = 0; |
176 | |
177 | outeriter = 0; |
178 | inneriter = 0; |
179 | |
180 | for(iidx=0;iidx<4*DIM3;iidx++) |
181 | { |
182 | scratch[iidx] = 0.0; |
183 | } |
184 | |
185 | /* Start outer loop over neighborlists */ |
186 | for(iidx=0; iidx<nri; iidx++) |
187 | { |
188 | /* Load shift vector for this list */ |
189 | i_shift_offset = DIM3*shiftidx[iidx]; |
190 | |
191 | /* Load limits for loop over neighbors */ |
192 | j_index_start = jindex[iidx]; |
193 | j_index_end = jindex[iidx+1]; |
194 | |
195 | /* Get outer coordinate index */ |
196 | inr = iinr[iidx]; |
197 | i_coord_offset = DIM3*inr; |
198 | |
199 | /* Load i particle coords and add shift vector */ |
200 | gmx_mm_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset, |
201 | &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3); |
202 | |
203 | fix0 = _mm_setzero_ps(); |
204 | fiy0 = _mm_setzero_ps(); |
205 | fiz0 = _mm_setzero_ps(); |
206 | fix1 = _mm_setzero_ps(); |
207 | fiy1 = _mm_setzero_ps(); |
208 | fiz1 = _mm_setzero_ps(); |
209 | fix2 = _mm_setzero_ps(); |
210 | fiy2 = _mm_setzero_ps(); |
211 | fiz2 = _mm_setzero_ps(); |
212 | fix3 = _mm_setzero_ps(); |
213 | fiy3 = _mm_setzero_ps(); |
214 | fiz3 = _mm_setzero_ps(); |
215 | |
216 | /* Reset potential sums */ |
217 | velecsum = _mm_setzero_ps(); |
218 | vvdwsum = _mm_setzero_ps(); |
219 | |
220 | /* Start inner kernel loop */ |
221 | for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4) |
222 | { |
223 | |
224 | /* Get j neighbor index, and coordinate index */ |
225 | jnrA = jjnr[jidx]; |
226 | jnrB = jjnr[jidx+1]; |
227 | jnrC = jjnr[jidx+2]; |
228 | jnrD = jjnr[jidx+3]; |
229 | j_coord_offsetA = DIM3*jnrA; |
230 | j_coord_offsetB = DIM3*jnrB; |
231 | j_coord_offsetC = DIM3*jnrC; |
232 | j_coord_offsetD = DIM3*jnrD; |
233 | |
234 | /* load j atom coordinates */ |
235 | gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB, |
236 | x+j_coord_offsetC,x+j_coord_offsetD, |
237 | &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2, |
238 | &jy2,&jz2,&jx3,&jy3,&jz3); |
239 | |
240 | /* Calculate displacement vector */ |
241 | dx00 = _mm_sub_ps(ix0,jx0); |
242 | dy00 = _mm_sub_ps(iy0,jy0); |
243 | dz00 = _mm_sub_ps(iz0,jz0); |
244 | dx11 = _mm_sub_ps(ix1,jx1); |
245 | dy11 = _mm_sub_ps(iy1,jy1); |
246 | dz11 = _mm_sub_ps(iz1,jz1); |
247 | dx12 = _mm_sub_ps(ix1,jx2); |
248 | dy12 = _mm_sub_ps(iy1,jy2); |
249 | dz12 = _mm_sub_ps(iz1,jz2); |
250 | dx13 = _mm_sub_ps(ix1,jx3); |
251 | dy13 = _mm_sub_ps(iy1,jy3); |
252 | dz13 = _mm_sub_ps(iz1,jz3); |
253 | dx21 = _mm_sub_ps(ix2,jx1); |
254 | dy21 = _mm_sub_ps(iy2,jy1); |
255 | dz21 = _mm_sub_ps(iz2,jz1); |
256 | dx22 = _mm_sub_ps(ix2,jx2); |
257 | dy22 = _mm_sub_ps(iy2,jy2); |
258 | dz22 = _mm_sub_ps(iz2,jz2); |
259 | dx23 = _mm_sub_ps(ix2,jx3); |
260 | dy23 = _mm_sub_ps(iy2,jy3); |
261 | dz23 = _mm_sub_ps(iz2,jz3); |
262 | dx31 = _mm_sub_ps(ix3,jx1); |
263 | dy31 = _mm_sub_ps(iy3,jy1); |
264 | dz31 = _mm_sub_ps(iz3,jz1); |
265 | dx32 = _mm_sub_ps(ix3,jx2); |
266 | dy32 = _mm_sub_ps(iy3,jy2); |
267 | dz32 = _mm_sub_ps(iz3,jz2); |
268 | dx33 = _mm_sub_ps(ix3,jx3); |
269 | dy33 = _mm_sub_ps(iy3,jy3); |
270 | dz33 = _mm_sub_ps(iz3,jz3); |
271 | |
272 | /* Calculate squared distance and things based on it */ |
273 | rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00); |
274 | rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11); |
275 | rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12); |
276 | rsq13 = gmx_mm_calc_rsq_ps(dx13,dy13,dz13); |
277 | rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21); |
278 | rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22); |
279 | rsq23 = gmx_mm_calc_rsq_ps(dx23,dy23,dz23); |
280 | rsq31 = gmx_mm_calc_rsq_ps(dx31,dy31,dz31); |
281 | rsq32 = gmx_mm_calc_rsq_ps(dx32,dy32,dz32); |
282 | rsq33 = gmx_mm_calc_rsq_ps(dx33,dy33,dz33); |
283 | |
284 | rinv00 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq00); |
285 | rinv11 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq11); |
286 | rinv12 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq12); |
287 | rinv13 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq13); |
288 | rinv21 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq21); |
289 | rinv22 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq22); |
290 | rinv23 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq23); |
291 | rinv31 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq31); |
292 | rinv32 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq32); |
293 | rinv33 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq33); |
294 | |
295 | rinvsq11 = _mm_mul_ps(rinv11,rinv11); |
296 | rinvsq12 = _mm_mul_ps(rinv12,rinv12); |
297 | rinvsq13 = _mm_mul_ps(rinv13,rinv13); |
298 | rinvsq21 = _mm_mul_ps(rinv21,rinv21); |
299 | rinvsq22 = _mm_mul_ps(rinv22,rinv22); |
300 | rinvsq23 = _mm_mul_ps(rinv23,rinv23); |
301 | rinvsq31 = _mm_mul_ps(rinv31,rinv31); |
302 | rinvsq32 = _mm_mul_ps(rinv32,rinv32); |
303 | rinvsq33 = _mm_mul_ps(rinv33,rinv33); |
304 | |
305 | fjx0 = _mm_setzero_ps(); |
306 | fjy0 = _mm_setzero_ps(); |
307 | fjz0 = _mm_setzero_ps(); |
308 | fjx1 = _mm_setzero_ps(); |
309 | fjy1 = _mm_setzero_ps(); |
310 | fjz1 = _mm_setzero_ps(); |
311 | fjx2 = _mm_setzero_ps(); |
312 | fjy2 = _mm_setzero_ps(); |
313 | fjz2 = _mm_setzero_ps(); |
314 | fjx3 = _mm_setzero_ps(); |
315 | fjy3 = _mm_setzero_ps(); |
316 | fjz3 = _mm_setzero_ps(); |
317 | |
318 | /************************** |
319 | * CALCULATE INTERACTIONS * |
320 | **************************/ |
321 | |
322 | r00 = _mm_mul_ps(rsq00,rinv00); |
323 | |
324 | /* Calculate table index by multiplying r with table scale and truncate to integer */ |
325 | rt = _mm_mul_ps(r00,vftabscale); |
326 | vfitab = _mm_cvttps_epi32(rt); |
327 | vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (rt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
328 | vfitab = _mm_slli_epi32(vfitab,3); |
329 | |
330 | /* CUBIC SPLINE TABLE DISPERSION */ |
331 | Y = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0)(__extension__ ({ __v4si __a = (__v4si)(vfitab); __a[(0) & 3];})) ); |
332 | F = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1)(__extension__ ({ __v4si __a = (__v4si)(vfitab); __a[(1) & 3];})) ); |
333 | G = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2)(__extension__ ({ __v4si __a = (__v4si)(vfitab); __a[(2) & 3];})) ); |
334 | H = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3)(__extension__ ({ __v4si __a = (__v4si)(vfitab); __a[(3) & 3];})) ); |
335 | _MM_TRANSPOSE4_PS(Y,F,G,H)do { __m128 tmp3, tmp2, tmp1, tmp0; tmp0 = _mm_unpacklo_ps((Y ), (F)); tmp2 = _mm_unpacklo_ps((G), (H)); tmp1 = _mm_unpackhi_ps ((Y), (F)); tmp3 = _mm_unpackhi_ps((G), (H)); (Y) = _mm_movelh_ps (tmp0, tmp2); (F) = _mm_movehl_ps(tmp2, tmp0); (G) = _mm_movelh_ps (tmp1, tmp3); (H) = _mm_movehl_ps(tmp3, tmp1); } while (0); |
336 | Heps = _mm_mul_ps(vfeps,H); |
337 | Fp = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps))); |
338 | VV = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp)); |
339 | vvdw6 = _mm_mul_ps(c6_00,VV); |
340 | FF = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps)))); |
341 | fvdw6 = _mm_mul_ps(c6_00,FF); |
342 | |
343 | /* CUBIC SPLINE TABLE REPULSION */ |
344 | vfitab = _mm_add_epi32(vfitab,ifour); |
345 | Y = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0)(__extension__ ({ __v4si __a = (__v4si)(vfitab); __a[(0) & 3];})) ); |
346 | F = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1)(__extension__ ({ __v4si __a = (__v4si)(vfitab); __a[(1) & 3];})) ); |
347 | G = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2)(__extension__ ({ __v4si __a = (__v4si)(vfitab); __a[(2) & 3];})) ); |
348 | H = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3)(__extension__ ({ __v4si __a = (__v4si)(vfitab); __a[(3) & 3];})) ); |
349 | _MM_TRANSPOSE4_PS(Y,F,G,H)do { __m128 tmp3, tmp2, tmp1, tmp0; tmp0 = _mm_unpacklo_ps((Y ), (F)); tmp2 = _mm_unpacklo_ps((G), (H)); tmp1 = _mm_unpackhi_ps ((Y), (F)); tmp3 = _mm_unpackhi_ps((G), (H)); (Y) = _mm_movelh_ps (tmp0, tmp2); (F) = _mm_movehl_ps(tmp2, tmp0); (G) = _mm_movelh_ps (tmp1, tmp3); (H) = _mm_movehl_ps(tmp3, tmp1); } while (0); |
350 | Heps = _mm_mul_ps(vfeps,H); |
351 | Fp = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps))); |
352 | VV = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp)); |
353 | vvdw12 = _mm_mul_ps(c12_00,VV); |
354 | FF = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps)))); |
355 | fvdw12 = _mm_mul_ps(c12_00,FF); |
356 | vvdw = _mm_add_ps(vvdw12,vvdw6); |
357 | fvdw = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00))); |
358 | |
359 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
360 | vvdwsum = _mm_add_ps(vvdwsum,vvdw); |
361 | |
362 | fscal = fvdw; |
363 | |
364 | /* Calculate temporary vectorial force */ |
365 | tx = _mm_mul_ps(fscal,dx00); |
366 | ty = _mm_mul_ps(fscal,dy00); |
367 | tz = _mm_mul_ps(fscal,dz00); |
368 | |
369 | /* Update vectorial force */ |
370 | fix0 = _mm_add_ps(fix0,tx); |
371 | fiy0 = _mm_add_ps(fiy0,ty); |
372 | fiz0 = _mm_add_ps(fiz0,tz); |
373 | |
374 | fjx0 = _mm_add_ps(fjx0,tx); |
375 | fjy0 = _mm_add_ps(fjy0,ty); |
376 | fjz0 = _mm_add_ps(fjz0,tz); |
377 | |
378 | /************************** |
379 | * CALCULATE INTERACTIONS * |
380 | **************************/ |
381 | |
382 | /* COULOMB ELECTROSTATICS */ |
383 | velec = _mm_mul_ps(qq11,rinv11); |
384 | felec = _mm_mul_ps(velec,rinvsq11); |
385 | |
386 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
387 | velecsum = _mm_add_ps(velecsum,velec); |
388 | |
389 | fscal = felec; |
390 | |
391 | /* Calculate temporary vectorial force */ |
392 | tx = _mm_mul_ps(fscal,dx11); |
393 | ty = _mm_mul_ps(fscal,dy11); |
394 | tz = _mm_mul_ps(fscal,dz11); |
395 | |
396 | /* Update vectorial force */ |
397 | fix1 = _mm_add_ps(fix1,tx); |
398 | fiy1 = _mm_add_ps(fiy1,ty); |
399 | fiz1 = _mm_add_ps(fiz1,tz); |
400 | |
401 | fjx1 = _mm_add_ps(fjx1,tx); |
402 | fjy1 = _mm_add_ps(fjy1,ty); |
403 | fjz1 = _mm_add_ps(fjz1,tz); |
404 | |
405 | /************************** |
406 | * CALCULATE INTERACTIONS * |
407 | **************************/ |
408 | |
409 | /* COULOMB ELECTROSTATICS */ |
410 | velec = _mm_mul_ps(qq12,rinv12); |
411 | felec = _mm_mul_ps(velec,rinvsq12); |
412 | |
413 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
414 | velecsum = _mm_add_ps(velecsum,velec); |
415 | |
416 | fscal = felec; |
417 | |
418 | /* Calculate temporary vectorial force */ |
419 | tx = _mm_mul_ps(fscal,dx12); |
420 | ty = _mm_mul_ps(fscal,dy12); |
421 | tz = _mm_mul_ps(fscal,dz12); |
422 | |
423 | /* Update vectorial force */ |
424 | fix1 = _mm_add_ps(fix1,tx); |
425 | fiy1 = _mm_add_ps(fiy1,ty); |
426 | fiz1 = _mm_add_ps(fiz1,tz); |
427 | |
428 | fjx2 = _mm_add_ps(fjx2,tx); |
429 | fjy2 = _mm_add_ps(fjy2,ty); |
430 | fjz2 = _mm_add_ps(fjz2,tz); |
431 | |
432 | /************************** |
433 | * CALCULATE INTERACTIONS * |
434 | **************************/ |
435 | |
436 | /* COULOMB ELECTROSTATICS */ |
437 | velec = _mm_mul_ps(qq13,rinv13); |
438 | felec = _mm_mul_ps(velec,rinvsq13); |
439 | |
440 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
441 | velecsum = _mm_add_ps(velecsum,velec); |
442 | |
443 | fscal = felec; |
444 | |
445 | /* Calculate temporary vectorial force */ |
446 | tx = _mm_mul_ps(fscal,dx13); |
447 | ty = _mm_mul_ps(fscal,dy13); |
448 | tz = _mm_mul_ps(fscal,dz13); |
449 | |
450 | /* Update vectorial force */ |
451 | fix1 = _mm_add_ps(fix1,tx); |
452 | fiy1 = _mm_add_ps(fiy1,ty); |
453 | fiz1 = _mm_add_ps(fiz1,tz); |
454 | |
455 | fjx3 = _mm_add_ps(fjx3,tx); |
456 | fjy3 = _mm_add_ps(fjy3,ty); |
457 | fjz3 = _mm_add_ps(fjz3,tz); |
458 | |
459 | /************************** |
460 | * CALCULATE INTERACTIONS * |
461 | **************************/ |
462 | |
463 | /* COULOMB ELECTROSTATICS */ |
464 | velec = _mm_mul_ps(qq21,rinv21); |
465 | felec = _mm_mul_ps(velec,rinvsq21); |
466 | |
467 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
468 | velecsum = _mm_add_ps(velecsum,velec); |
469 | |
470 | fscal = felec; |
471 | |
472 | /* Calculate temporary vectorial force */ |
473 | tx = _mm_mul_ps(fscal,dx21); |
474 | ty = _mm_mul_ps(fscal,dy21); |
475 | tz = _mm_mul_ps(fscal,dz21); |
476 | |
477 | /* Update vectorial force */ |
478 | fix2 = _mm_add_ps(fix2,tx); |
479 | fiy2 = _mm_add_ps(fiy2,ty); |
480 | fiz2 = _mm_add_ps(fiz2,tz); |
481 | |
482 | fjx1 = _mm_add_ps(fjx1,tx); |
483 | fjy1 = _mm_add_ps(fjy1,ty); |
484 | fjz1 = _mm_add_ps(fjz1,tz); |
485 | |
486 | /************************** |
487 | * CALCULATE INTERACTIONS * |
488 | **************************/ |
489 | |
490 | /* COULOMB ELECTROSTATICS */ |
491 | velec = _mm_mul_ps(qq22,rinv22); |
492 | felec = _mm_mul_ps(velec,rinvsq22); |
493 | |
494 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
495 | velecsum = _mm_add_ps(velecsum,velec); |
496 | |
497 | fscal = felec; |
498 | |
499 | /* Calculate temporary vectorial force */ |
500 | tx = _mm_mul_ps(fscal,dx22); |
501 | ty = _mm_mul_ps(fscal,dy22); |
502 | tz = _mm_mul_ps(fscal,dz22); |
503 | |
504 | /* Update vectorial force */ |
505 | fix2 = _mm_add_ps(fix2,tx); |
506 | fiy2 = _mm_add_ps(fiy2,ty); |
507 | fiz2 = _mm_add_ps(fiz2,tz); |
508 | |
509 | fjx2 = _mm_add_ps(fjx2,tx); |
510 | fjy2 = _mm_add_ps(fjy2,ty); |
511 | fjz2 = _mm_add_ps(fjz2,tz); |
512 | |
513 | /************************** |
514 | * CALCULATE INTERACTIONS * |
515 | **************************/ |
516 | |
517 | /* COULOMB ELECTROSTATICS */ |
518 | velec = _mm_mul_ps(qq23,rinv23); |
519 | felec = _mm_mul_ps(velec,rinvsq23); |
520 | |
521 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
522 | velecsum = _mm_add_ps(velecsum,velec); |
523 | |
524 | fscal = felec; |
525 | |
526 | /* Calculate temporary vectorial force */ |
527 | tx = _mm_mul_ps(fscal,dx23); |
528 | ty = _mm_mul_ps(fscal,dy23); |
529 | tz = _mm_mul_ps(fscal,dz23); |
530 | |
531 | /* Update vectorial force */ |
532 | fix2 = _mm_add_ps(fix2,tx); |
533 | fiy2 = _mm_add_ps(fiy2,ty); |
534 | fiz2 = _mm_add_ps(fiz2,tz); |
535 | |
536 | fjx3 = _mm_add_ps(fjx3,tx); |
537 | fjy3 = _mm_add_ps(fjy3,ty); |
538 | fjz3 = _mm_add_ps(fjz3,tz); |
539 | |
540 | /************************** |
541 | * CALCULATE INTERACTIONS * |
542 | **************************/ |
543 | |
544 | /* COULOMB ELECTROSTATICS */ |
545 | velec = _mm_mul_ps(qq31,rinv31); |
546 | felec = _mm_mul_ps(velec,rinvsq31); |
547 | |
548 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
549 | velecsum = _mm_add_ps(velecsum,velec); |
550 | |
551 | fscal = felec; |
552 | |
553 | /* Calculate temporary vectorial force */ |
554 | tx = _mm_mul_ps(fscal,dx31); |
555 | ty = _mm_mul_ps(fscal,dy31); |
556 | tz = _mm_mul_ps(fscal,dz31); |
557 | |
558 | /* Update vectorial force */ |
559 | fix3 = _mm_add_ps(fix3,tx); |
560 | fiy3 = _mm_add_ps(fiy3,ty); |
561 | fiz3 = _mm_add_ps(fiz3,tz); |
562 | |
563 | fjx1 = _mm_add_ps(fjx1,tx); |
564 | fjy1 = _mm_add_ps(fjy1,ty); |
565 | fjz1 = _mm_add_ps(fjz1,tz); |
566 | |
567 | /************************** |
568 | * CALCULATE INTERACTIONS * |
569 | **************************/ |
570 | |
571 | /* COULOMB ELECTROSTATICS */ |
572 | velec = _mm_mul_ps(qq32,rinv32); |
573 | felec = _mm_mul_ps(velec,rinvsq32); |
574 | |
575 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
576 | velecsum = _mm_add_ps(velecsum,velec); |
577 | |
578 | fscal = felec; |
579 | |
580 | /* Calculate temporary vectorial force */ |
581 | tx = _mm_mul_ps(fscal,dx32); |
582 | ty = _mm_mul_ps(fscal,dy32); |
583 | tz = _mm_mul_ps(fscal,dz32); |
584 | |
585 | /* Update vectorial force */ |
586 | fix3 = _mm_add_ps(fix3,tx); |
587 | fiy3 = _mm_add_ps(fiy3,ty); |
588 | fiz3 = _mm_add_ps(fiz3,tz); |
589 | |
590 | fjx2 = _mm_add_ps(fjx2,tx); |
591 | fjy2 = _mm_add_ps(fjy2,ty); |
592 | fjz2 = _mm_add_ps(fjz2,tz); |
593 | |
594 | /************************** |
595 | * CALCULATE INTERACTIONS * |
596 | **************************/ |
597 | |
598 | /* COULOMB ELECTROSTATICS */ |
599 | velec = _mm_mul_ps(qq33,rinv33); |
600 | felec = _mm_mul_ps(velec,rinvsq33); |
601 | |
602 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
603 | velecsum = _mm_add_ps(velecsum,velec); |
604 | |
605 | fscal = felec; |
606 | |
607 | /* Calculate temporary vectorial force */ |
608 | tx = _mm_mul_ps(fscal,dx33); |
609 | ty = _mm_mul_ps(fscal,dy33); |
610 | tz = _mm_mul_ps(fscal,dz33); |
611 | |
612 | /* Update vectorial force */ |
613 | fix3 = _mm_add_ps(fix3,tx); |
614 | fiy3 = _mm_add_ps(fiy3,ty); |
615 | fiz3 = _mm_add_ps(fiz3,tz); |
616 | |
617 | fjx3 = _mm_add_ps(fjx3,tx); |
618 | fjy3 = _mm_add_ps(fjy3,ty); |
619 | fjz3 = _mm_add_ps(fjz3,tz); |
620 | |
621 | fjptrA = f+j_coord_offsetA; |
622 | fjptrB = f+j_coord_offsetB; |
623 | fjptrC = f+j_coord_offsetC; |
624 | fjptrD = f+j_coord_offsetD; |
625 | |
626 | gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD, |
627 | fjx0,fjy0,fjz0,fjx1,fjy1,fjz1, |
628 | fjx2,fjy2,fjz2,fjx3,fjy3,fjz3); |
629 | |
630 | /* Inner loop uses 311 flops */ |
631 | } |
632 | |
633 | if(jidx<j_index_end) |
634 | { |
635 | |
636 | /* Get j neighbor index, and coordinate index */ |
637 | jnrlistA = jjnr[jidx]; |
638 | jnrlistB = jjnr[jidx+1]; |
639 | jnrlistC = jjnr[jidx+2]; |
640 | jnrlistD = jjnr[jidx+3]; |
641 | /* Sign of each element will be negative for non-real atoms. |
642 | * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones, |
643 | * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries. |
644 | */ |
645 | dummy_mask = gmx_mm_castsi128_ps_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())); |
646 | jnrA = (jnrlistA>=0) ? jnrlistA : 0; |
647 | jnrB = (jnrlistB>=0) ? jnrlistB : 0; |
648 | jnrC = (jnrlistC>=0) ? jnrlistC : 0; |
649 | jnrD = (jnrlistD>=0) ? jnrlistD : 0; |
650 | j_coord_offsetA = DIM3*jnrA; |
651 | j_coord_offsetB = DIM3*jnrB; |
652 | j_coord_offsetC = DIM3*jnrC; |
653 | j_coord_offsetD = DIM3*jnrD; |
654 | |
655 | /* load j atom coordinates */ |
656 | gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB, |
657 | x+j_coord_offsetC,x+j_coord_offsetD, |
658 | &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2, |
659 | &jy2,&jz2,&jx3,&jy3,&jz3); |
660 | |
661 | /* Calculate displacement vector */ |
662 | dx00 = _mm_sub_ps(ix0,jx0); |
663 | dy00 = _mm_sub_ps(iy0,jy0); |
664 | dz00 = _mm_sub_ps(iz0,jz0); |
665 | dx11 = _mm_sub_ps(ix1,jx1); |
666 | dy11 = _mm_sub_ps(iy1,jy1); |
667 | dz11 = _mm_sub_ps(iz1,jz1); |
668 | dx12 = _mm_sub_ps(ix1,jx2); |
669 | dy12 = _mm_sub_ps(iy1,jy2); |
670 | dz12 = _mm_sub_ps(iz1,jz2); |
671 | dx13 = _mm_sub_ps(ix1,jx3); |
672 | dy13 = _mm_sub_ps(iy1,jy3); |
673 | dz13 = _mm_sub_ps(iz1,jz3); |
674 | dx21 = _mm_sub_ps(ix2,jx1); |
675 | dy21 = _mm_sub_ps(iy2,jy1); |
676 | dz21 = _mm_sub_ps(iz2,jz1); |
677 | dx22 = _mm_sub_ps(ix2,jx2); |
678 | dy22 = _mm_sub_ps(iy2,jy2); |
679 | dz22 = _mm_sub_ps(iz2,jz2); |
680 | dx23 = _mm_sub_ps(ix2,jx3); |
681 | dy23 = _mm_sub_ps(iy2,jy3); |
682 | dz23 = _mm_sub_ps(iz2,jz3); |
683 | dx31 = _mm_sub_ps(ix3,jx1); |
684 | dy31 = _mm_sub_ps(iy3,jy1); |
685 | dz31 = _mm_sub_ps(iz3,jz1); |
686 | dx32 = _mm_sub_ps(ix3,jx2); |
687 | dy32 = _mm_sub_ps(iy3,jy2); |
688 | dz32 = _mm_sub_ps(iz3,jz2); |
689 | dx33 = _mm_sub_ps(ix3,jx3); |
690 | dy33 = _mm_sub_ps(iy3,jy3); |
691 | dz33 = _mm_sub_ps(iz3,jz3); |
692 | |
693 | /* Calculate squared distance and things based on it */ |
694 | rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00); |
695 | rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11); |
696 | rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12); |
697 | rsq13 = gmx_mm_calc_rsq_ps(dx13,dy13,dz13); |
698 | rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21); |
699 | rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22); |
700 | rsq23 = gmx_mm_calc_rsq_ps(dx23,dy23,dz23); |
701 | rsq31 = gmx_mm_calc_rsq_ps(dx31,dy31,dz31); |
702 | rsq32 = gmx_mm_calc_rsq_ps(dx32,dy32,dz32); |
703 | rsq33 = gmx_mm_calc_rsq_ps(dx33,dy33,dz33); |
704 | |
705 | rinv00 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq00); |
706 | rinv11 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq11); |
707 | rinv12 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq12); |
708 | rinv13 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq13); |
709 | rinv21 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq21); |
710 | rinv22 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq22); |
711 | rinv23 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq23); |
712 | rinv31 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq31); |
713 | rinv32 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq32); |
714 | rinv33 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq33); |
715 | |
716 | rinvsq11 = _mm_mul_ps(rinv11,rinv11); |
717 | rinvsq12 = _mm_mul_ps(rinv12,rinv12); |
718 | rinvsq13 = _mm_mul_ps(rinv13,rinv13); |
719 | rinvsq21 = _mm_mul_ps(rinv21,rinv21); |
720 | rinvsq22 = _mm_mul_ps(rinv22,rinv22); |
721 | rinvsq23 = _mm_mul_ps(rinv23,rinv23); |
722 | rinvsq31 = _mm_mul_ps(rinv31,rinv31); |
723 | rinvsq32 = _mm_mul_ps(rinv32,rinv32); |
724 | rinvsq33 = _mm_mul_ps(rinv33,rinv33); |
725 | |
726 | fjx0 = _mm_setzero_ps(); |
727 | fjy0 = _mm_setzero_ps(); |
728 | fjz0 = _mm_setzero_ps(); |
729 | fjx1 = _mm_setzero_ps(); |
730 | fjy1 = _mm_setzero_ps(); |
731 | fjz1 = _mm_setzero_ps(); |
732 | fjx2 = _mm_setzero_ps(); |
733 | fjy2 = _mm_setzero_ps(); |
734 | fjz2 = _mm_setzero_ps(); |
735 | fjx3 = _mm_setzero_ps(); |
736 | fjy3 = _mm_setzero_ps(); |
737 | fjz3 = _mm_setzero_ps(); |
738 | |
739 | /************************** |
740 | * CALCULATE INTERACTIONS * |
741 | **************************/ |
742 | |
743 | r00 = _mm_mul_ps(rsq00,rinv00); |
744 | r00 = _mm_andnot_ps(dummy_mask,r00); |
745 | |
746 | /* Calculate table index by multiplying r with table scale and truncate to integer */ |
747 | rt = _mm_mul_ps(r00,vftabscale); |
748 | vfitab = _mm_cvttps_epi32(rt); |
749 | vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (rt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
750 | vfitab = _mm_slli_epi32(vfitab,3); |
751 | |
752 | /* CUBIC SPLINE TABLE DISPERSION */ |
753 | Y = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0)(__extension__ ({ __v4si __a = (__v4si)(vfitab); __a[(0) & 3];})) ); |
754 | F = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1)(__extension__ ({ __v4si __a = (__v4si)(vfitab); __a[(1) & 3];})) ); |
755 | G = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2)(__extension__ ({ __v4si __a = (__v4si)(vfitab); __a[(2) & 3];})) ); |
756 | H = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3)(__extension__ ({ __v4si __a = (__v4si)(vfitab); __a[(3) & 3];})) ); |
757 | _MM_TRANSPOSE4_PS(Y,F,G,H)do { __m128 tmp3, tmp2, tmp1, tmp0; tmp0 = _mm_unpacklo_ps((Y ), (F)); tmp2 = _mm_unpacklo_ps((G), (H)); tmp1 = _mm_unpackhi_ps ((Y), (F)); tmp3 = _mm_unpackhi_ps((G), (H)); (Y) = _mm_movelh_ps (tmp0, tmp2); (F) = _mm_movehl_ps(tmp2, tmp0); (G) = _mm_movelh_ps (tmp1, tmp3); (H) = _mm_movehl_ps(tmp3, tmp1); } while (0); |
758 | Heps = _mm_mul_ps(vfeps,H); |
759 | Fp = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps))); |
760 | VV = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp)); |
761 | vvdw6 = _mm_mul_ps(c6_00,VV); |
762 | FF = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps)))); |
763 | fvdw6 = _mm_mul_ps(c6_00,FF); |
764 | |
765 | /* CUBIC SPLINE TABLE REPULSION */ |
766 | vfitab = _mm_add_epi32(vfitab,ifour); |
767 | Y = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0)(__extension__ ({ __v4si __a = (__v4si)(vfitab); __a[(0) & 3];})) ); |
768 | F = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1)(__extension__ ({ __v4si __a = (__v4si)(vfitab); __a[(1) & 3];})) ); |
769 | G = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2)(__extension__ ({ __v4si __a = (__v4si)(vfitab); __a[(2) & 3];})) ); |
770 | H = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3)(__extension__ ({ __v4si __a = (__v4si)(vfitab); __a[(3) & 3];})) ); |
771 | _MM_TRANSPOSE4_PS(Y,F,G,H)do { __m128 tmp3, tmp2, tmp1, tmp0; tmp0 = _mm_unpacklo_ps((Y ), (F)); tmp2 = _mm_unpacklo_ps((G), (H)); tmp1 = _mm_unpackhi_ps ((Y), (F)); tmp3 = _mm_unpackhi_ps((G), (H)); (Y) = _mm_movelh_ps (tmp0, tmp2); (F) = _mm_movehl_ps(tmp2, tmp0); (G) = _mm_movelh_ps (tmp1, tmp3); (H) = _mm_movehl_ps(tmp3, tmp1); } while (0); |
772 | Heps = _mm_mul_ps(vfeps,H); |
773 | Fp = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps))); |
774 | VV = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp)); |
775 | vvdw12 = _mm_mul_ps(c12_00,VV); |
776 | FF = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps)))); |
777 | fvdw12 = _mm_mul_ps(c12_00,FF); |
778 | vvdw = _mm_add_ps(vvdw12,vvdw6); |
779 | fvdw = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00))); |
780 | |
781 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
782 | vvdw = _mm_andnot_ps(dummy_mask,vvdw); |
783 | vvdwsum = _mm_add_ps(vvdwsum,vvdw); |
784 | |
785 | fscal = fvdw; |
786 | |
787 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
788 | |
789 | /* Calculate temporary vectorial force */ |
790 | tx = _mm_mul_ps(fscal,dx00); |
791 | ty = _mm_mul_ps(fscal,dy00); |
792 | tz = _mm_mul_ps(fscal,dz00); |
793 | |
794 | /* Update vectorial force */ |
795 | fix0 = _mm_add_ps(fix0,tx); |
796 | fiy0 = _mm_add_ps(fiy0,ty); |
797 | fiz0 = _mm_add_ps(fiz0,tz); |
798 | |
799 | fjx0 = _mm_add_ps(fjx0,tx); |
800 | fjy0 = _mm_add_ps(fjy0,ty); |
801 | fjz0 = _mm_add_ps(fjz0,tz); |
802 | |
803 | /************************** |
804 | * CALCULATE INTERACTIONS * |
805 | **************************/ |
806 | |
807 | /* COULOMB ELECTROSTATICS */ |
808 | velec = _mm_mul_ps(qq11,rinv11); |
809 | felec = _mm_mul_ps(velec,rinvsq11); |
810 | |
811 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
812 | velec = _mm_andnot_ps(dummy_mask,velec); |
813 | velecsum = _mm_add_ps(velecsum,velec); |
814 | |
815 | fscal = felec; |
816 | |
817 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
818 | |
819 | /* Calculate temporary vectorial force */ |
820 | tx = _mm_mul_ps(fscal,dx11); |
821 | ty = _mm_mul_ps(fscal,dy11); |
822 | tz = _mm_mul_ps(fscal,dz11); |
823 | |
824 | /* Update vectorial force */ |
825 | fix1 = _mm_add_ps(fix1,tx); |
826 | fiy1 = _mm_add_ps(fiy1,ty); |
827 | fiz1 = _mm_add_ps(fiz1,tz); |
828 | |
829 | fjx1 = _mm_add_ps(fjx1,tx); |
830 | fjy1 = _mm_add_ps(fjy1,ty); |
831 | fjz1 = _mm_add_ps(fjz1,tz); |
832 | |
833 | /************************** |
834 | * CALCULATE INTERACTIONS * |
835 | **************************/ |
836 | |
837 | /* COULOMB ELECTROSTATICS */ |
838 | velec = _mm_mul_ps(qq12,rinv12); |
839 | felec = _mm_mul_ps(velec,rinvsq12); |
840 | |
841 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
842 | velec = _mm_andnot_ps(dummy_mask,velec); |
843 | velecsum = _mm_add_ps(velecsum,velec); |
844 | |
845 | fscal = felec; |
846 | |
847 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
848 | |
849 | /* Calculate temporary vectorial force */ |
850 | tx = _mm_mul_ps(fscal,dx12); |
851 | ty = _mm_mul_ps(fscal,dy12); |
852 | tz = _mm_mul_ps(fscal,dz12); |
853 | |
854 | /* Update vectorial force */ |
855 | fix1 = _mm_add_ps(fix1,tx); |
856 | fiy1 = _mm_add_ps(fiy1,ty); |
857 | fiz1 = _mm_add_ps(fiz1,tz); |
858 | |
859 | fjx2 = _mm_add_ps(fjx2,tx); |
860 | fjy2 = _mm_add_ps(fjy2,ty); |
861 | fjz2 = _mm_add_ps(fjz2,tz); |
862 | |
863 | /************************** |
864 | * CALCULATE INTERACTIONS * |
865 | **************************/ |
866 | |
867 | /* COULOMB ELECTROSTATICS */ |
868 | velec = _mm_mul_ps(qq13,rinv13); |
869 | felec = _mm_mul_ps(velec,rinvsq13); |
870 | |
871 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
872 | velec = _mm_andnot_ps(dummy_mask,velec); |
873 | velecsum = _mm_add_ps(velecsum,velec); |
874 | |
875 | fscal = felec; |
876 | |
877 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
878 | |
879 | /* Calculate temporary vectorial force */ |
880 | tx = _mm_mul_ps(fscal,dx13); |
881 | ty = _mm_mul_ps(fscal,dy13); |
882 | tz = _mm_mul_ps(fscal,dz13); |
883 | |
884 | /* Update vectorial force */ |
885 | fix1 = _mm_add_ps(fix1,tx); |
886 | fiy1 = _mm_add_ps(fiy1,ty); |
887 | fiz1 = _mm_add_ps(fiz1,tz); |
888 | |
889 | fjx3 = _mm_add_ps(fjx3,tx); |
890 | fjy3 = _mm_add_ps(fjy3,ty); |
891 | fjz3 = _mm_add_ps(fjz3,tz); |
892 | |
893 | /************************** |
894 | * CALCULATE INTERACTIONS * |
895 | **************************/ |
896 | |
897 | /* COULOMB ELECTROSTATICS */ |
898 | velec = _mm_mul_ps(qq21,rinv21); |
899 | felec = _mm_mul_ps(velec,rinvsq21); |
900 | |
901 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
902 | velec = _mm_andnot_ps(dummy_mask,velec); |
903 | velecsum = _mm_add_ps(velecsum,velec); |
904 | |
905 | fscal = felec; |
906 | |
907 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
908 | |
909 | /* Calculate temporary vectorial force */ |
910 | tx = _mm_mul_ps(fscal,dx21); |
911 | ty = _mm_mul_ps(fscal,dy21); |
912 | tz = _mm_mul_ps(fscal,dz21); |
913 | |
914 | /* Update vectorial force */ |
915 | fix2 = _mm_add_ps(fix2,tx); |
916 | fiy2 = _mm_add_ps(fiy2,ty); |
917 | fiz2 = _mm_add_ps(fiz2,tz); |
918 | |
919 | fjx1 = _mm_add_ps(fjx1,tx); |
920 | fjy1 = _mm_add_ps(fjy1,ty); |
921 | fjz1 = _mm_add_ps(fjz1,tz); |
922 | |
923 | /************************** |
924 | * CALCULATE INTERACTIONS * |
925 | **************************/ |
926 | |
927 | /* COULOMB ELECTROSTATICS */ |
928 | velec = _mm_mul_ps(qq22,rinv22); |
929 | felec = _mm_mul_ps(velec,rinvsq22); |
930 | |
931 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
932 | velec = _mm_andnot_ps(dummy_mask,velec); |
933 | velecsum = _mm_add_ps(velecsum,velec); |
934 | |
935 | fscal = felec; |
936 | |
937 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
938 | |
939 | /* Calculate temporary vectorial force */ |
940 | tx = _mm_mul_ps(fscal,dx22); |
941 | ty = _mm_mul_ps(fscal,dy22); |
942 | tz = _mm_mul_ps(fscal,dz22); |
943 | |
944 | /* Update vectorial force */ |
945 | fix2 = _mm_add_ps(fix2,tx); |
946 | fiy2 = _mm_add_ps(fiy2,ty); |
947 | fiz2 = _mm_add_ps(fiz2,tz); |
948 | |
949 | fjx2 = _mm_add_ps(fjx2,tx); |
950 | fjy2 = _mm_add_ps(fjy2,ty); |
951 | fjz2 = _mm_add_ps(fjz2,tz); |
952 | |
953 | /************************** |
954 | * CALCULATE INTERACTIONS * |
955 | **************************/ |
956 | |
957 | /* COULOMB ELECTROSTATICS */ |
958 | velec = _mm_mul_ps(qq23,rinv23); |
959 | felec = _mm_mul_ps(velec,rinvsq23); |
960 | |
961 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
962 | velec = _mm_andnot_ps(dummy_mask,velec); |
963 | velecsum = _mm_add_ps(velecsum,velec); |
964 | |
965 | fscal = felec; |
966 | |
967 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
968 | |
969 | /* Calculate temporary vectorial force */ |
970 | tx = _mm_mul_ps(fscal,dx23); |
971 | ty = _mm_mul_ps(fscal,dy23); |
972 | tz = _mm_mul_ps(fscal,dz23); |
973 | |
974 | /* Update vectorial force */ |
975 | fix2 = _mm_add_ps(fix2,tx); |
976 | fiy2 = _mm_add_ps(fiy2,ty); |
977 | fiz2 = _mm_add_ps(fiz2,tz); |
978 | |
979 | fjx3 = _mm_add_ps(fjx3,tx); |
980 | fjy3 = _mm_add_ps(fjy3,ty); |
981 | fjz3 = _mm_add_ps(fjz3,tz); |
982 | |
983 | /************************** |
984 | * CALCULATE INTERACTIONS * |
985 | **************************/ |
986 | |
987 | /* COULOMB ELECTROSTATICS */ |
988 | velec = _mm_mul_ps(qq31,rinv31); |
989 | felec = _mm_mul_ps(velec,rinvsq31); |
990 | |
991 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
992 | velec = _mm_andnot_ps(dummy_mask,velec); |
993 | velecsum = _mm_add_ps(velecsum,velec); |
994 | |
995 | fscal = felec; |
996 | |
997 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
998 | |
999 | /* Calculate temporary vectorial force */ |
1000 | tx = _mm_mul_ps(fscal,dx31); |
1001 | ty = _mm_mul_ps(fscal,dy31); |
1002 | tz = _mm_mul_ps(fscal,dz31); |
1003 | |
1004 | /* Update vectorial force */ |
1005 | fix3 = _mm_add_ps(fix3,tx); |
1006 | fiy3 = _mm_add_ps(fiy3,ty); |
1007 | fiz3 = _mm_add_ps(fiz3,tz); |
1008 | |
1009 | fjx1 = _mm_add_ps(fjx1,tx); |
1010 | fjy1 = _mm_add_ps(fjy1,ty); |
1011 | fjz1 = _mm_add_ps(fjz1,tz); |
1012 | |
1013 | /************************** |
1014 | * CALCULATE INTERACTIONS * |
1015 | **************************/ |
1016 | |
1017 | /* COULOMB ELECTROSTATICS */ |
1018 | velec = _mm_mul_ps(qq32,rinv32); |
1019 | felec = _mm_mul_ps(velec,rinvsq32); |
1020 | |
1021 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
1022 | velec = _mm_andnot_ps(dummy_mask,velec); |
1023 | velecsum = _mm_add_ps(velecsum,velec); |
1024 | |
1025 | fscal = felec; |
1026 | |
1027 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
1028 | |
1029 | /* Calculate temporary vectorial force */ |
1030 | tx = _mm_mul_ps(fscal,dx32); |
1031 | ty = _mm_mul_ps(fscal,dy32); |
1032 | tz = _mm_mul_ps(fscal,dz32); |
1033 | |
1034 | /* Update vectorial force */ |
1035 | fix3 = _mm_add_ps(fix3,tx); |
1036 | fiy3 = _mm_add_ps(fiy3,ty); |
1037 | fiz3 = _mm_add_ps(fiz3,tz); |
1038 | |
1039 | fjx2 = _mm_add_ps(fjx2,tx); |
1040 | fjy2 = _mm_add_ps(fjy2,ty); |
1041 | fjz2 = _mm_add_ps(fjz2,tz); |
1042 | |
1043 | /************************** |
1044 | * CALCULATE INTERACTIONS * |
1045 | **************************/ |
1046 | |
1047 | /* COULOMB ELECTROSTATICS */ |
1048 | velec = _mm_mul_ps(qq33,rinv33); |
1049 | felec = _mm_mul_ps(velec,rinvsq33); |
1050 | |
1051 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
1052 | velec = _mm_andnot_ps(dummy_mask,velec); |
1053 | velecsum = _mm_add_ps(velecsum,velec); |
1054 | |
1055 | fscal = felec; |
1056 | |
1057 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
1058 | |
1059 | /* Calculate temporary vectorial force */ |
1060 | tx = _mm_mul_ps(fscal,dx33); |
1061 | ty = _mm_mul_ps(fscal,dy33); |
1062 | tz = _mm_mul_ps(fscal,dz33); |
1063 | |
1064 | /* Update vectorial force */ |
1065 | fix3 = _mm_add_ps(fix3,tx); |
1066 | fiy3 = _mm_add_ps(fiy3,ty); |
1067 | fiz3 = _mm_add_ps(fiz3,tz); |
1068 | |
1069 | fjx3 = _mm_add_ps(fjx3,tx); |
1070 | fjy3 = _mm_add_ps(fjy3,ty); |
1071 | fjz3 = _mm_add_ps(fjz3,tz); |
1072 | |
1073 | fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch; |
1074 | fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch; |
1075 | fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch; |
1076 | fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch; |
1077 | |
1078 | gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD, |
1079 | fjx0,fjy0,fjz0,fjx1,fjy1,fjz1, |
1080 | fjx2,fjy2,fjz2,fjx3,fjy3,fjz3); |
1081 | |
1082 | /* Inner loop uses 312 flops */ |
1083 | } |
1084 | |
1085 | /* End of innermost loop */ |
1086 | |
1087 | gmx_mm_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3, |
1088 | f+i_coord_offset,fshift+i_shift_offset); |
1089 | |
1090 | ggid = gid[iidx]; |
1091 | /* Update potential energies */ |
1092 | gmx_mm_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid); |
1093 | gmx_mm_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid); |
1094 | |
1095 | /* Increment number of inner iterations */ |
1096 | inneriter += j_index_end - j_index_start; |
1097 | |
1098 | /* Outer loop uses 26 flops */ |
1099 | } |
1100 | |
1101 | /* Increment number of outer iterations */ |
1102 | outeriter += nri; |
1103 | |
1104 | /* Update outer/inner flops */ |
1105 | |
1106 | inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*312)(nrnb)->n[eNR_NBKERNEL_ELEC_VDW_W4W4_VF] += outeriter*26 + inneriter*312; |
1107 | } |
1108 | /* |
1109 | * Gromacs nonbonded kernel: nb_kernel_ElecCoul_VdwCSTab_GeomW4W4_F_sse4_1_single |
1110 | * Electrostatics interaction: Coulomb |
1111 | * VdW interaction: CubicSplineTable |
1112 | * Geometry: Water4-Water4 |
1113 | * Calculate force/pot: Force |
1114 | */ |
1115 | void |
1116 | nb_kernel_ElecCoul_VdwCSTab_GeomW4W4_F_sse4_1_single |
1117 | (t_nblist * gmx_restrict nlist, |
1118 | rvec * gmx_restrict xx, |
1119 | rvec * gmx_restrict ff, |
1120 | t_forcerec * gmx_restrict fr, |
1121 | t_mdatoms * gmx_restrict mdatoms, |
1122 | nb_kernel_data_t gmx_unused__attribute__ ((unused)) * gmx_restrict kernel_data, |
1123 | t_nrnb * gmx_restrict nrnb) |
1124 | { |
1125 | /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or |
1126 | * just 0 for non-waters. |
1127 | * Suffixes A,B,C,D refer to j loop unrolling done with SSE, e.g. for the four different |
1128 | * jnr indices corresponding to data put in the four positions in the SIMD register. |
1129 | */ |
1130 | int i_shift_offset,i_coord_offset,outeriter,inneriter; |
1131 | int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx; |
1132 | int jnrA,jnrB,jnrC,jnrD; |
1133 | int jnrlistA,jnrlistB,jnrlistC,jnrlistD; |
1134 | int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD; |
1135 | int *iinr,*jindex,*jjnr,*shiftidx,*gid; |
1136 | real rcutoff_scalar; |
1137 | real *shiftvec,*fshift,*x,*f; |
1138 | real *fjptrA,*fjptrB,*fjptrC,*fjptrD; |
1139 | real scratch[4*DIM3]; |
1140 | __m128 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall; |
1141 | int vdwioffset0; |
1142 | __m128 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0; |
1143 | int vdwioffset1; |
1144 | __m128 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1; |
1145 | int vdwioffset2; |
1146 | __m128 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2; |
1147 | int vdwioffset3; |
1148 | __m128 ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3; |
1149 | int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D; |
1150 | __m128 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0; |
1151 | int vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D; |
1152 | __m128 jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1; |
1153 | int vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D; |
1154 | __m128 jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2; |
1155 | int vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D; |
1156 | __m128 jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3; |
1157 | __m128 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00; |
1158 | __m128 dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11; |
1159 | __m128 dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12; |
1160 | __m128 dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13; |
1161 | __m128 dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21; |
1162 | __m128 dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22; |
1163 | __m128 dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23; |
1164 | __m128 dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31; |
1165 | __m128 dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32; |
1166 | __m128 dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33; |
1167 | __m128 velec,felec,velecsum,facel,crf,krf,krf2; |
1168 | real *charge; |
1169 | int nvdwtype; |
1170 | __m128 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6; |
1171 | int *vdwtype; |
1172 | real *vdwparam; |
1173 | __m128 one_sixth = _mm_set1_ps(1.0/6.0); |
1174 | __m128 one_twelfth = _mm_set1_ps(1.0/12.0); |
1175 | __m128i vfitab; |
1176 | __m128i ifour = _mm_set1_epi32(4); |
1177 | __m128 rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF; |
1178 | real *vftab; |
1179 | __m128 dummy_mask,cutoff_mask; |
1180 | __m128 signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) ); |
1181 | __m128 one = _mm_set1_ps(1.0); |
1182 | __m128 two = _mm_set1_ps(2.0); |
1183 | x = xx[0]; |
1184 | f = ff[0]; |
1185 | |
1186 | nri = nlist->nri; |
1187 | iinr = nlist->iinr; |
1188 | jindex = nlist->jindex; |
1189 | jjnr = nlist->jjnr; |
1190 | shiftidx = nlist->shift; |
1191 | gid = nlist->gid; |
1192 | shiftvec = fr->shift_vec[0]; |
1193 | fshift = fr->fshift[0]; |
1194 | facel = _mm_set1_ps(fr->epsfac); |
1195 | charge = mdatoms->chargeA; |
1196 | nvdwtype = fr->ntype; |
1197 | vdwparam = fr->nbfp; |
1198 | vdwtype = mdatoms->typeA; |
1199 | |
1200 | vftab = kernel_data->table_vdw->data; |
1201 | vftabscale = _mm_set1_ps(kernel_data->table_vdw->scale); |
1202 | |
1203 | /* Setup water-specific parameters */ |
1204 | inr = nlist->iinr[0]; |
1205 | iq1 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1])); |
1206 | iq2 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2])); |
1207 | iq3 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+3])); |
1208 | vdwioffset0 = 2*nvdwtype*vdwtype[inr+0]; |
1209 | |
1210 | jq1 = _mm_set1_ps(charge[inr+1]); |
1211 | jq2 = _mm_set1_ps(charge[inr+2]); |
1212 | jq3 = _mm_set1_ps(charge[inr+3]); |
1213 | vdwjidx0A = 2*vdwtype[inr+0]; |
1214 | c6_00 = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A]); |
1215 | c12_00 = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A+1]); |
1216 | qq11 = _mm_mul_ps(iq1,jq1); |
1217 | qq12 = _mm_mul_ps(iq1,jq2); |
1218 | qq13 = _mm_mul_ps(iq1,jq3); |
1219 | qq21 = _mm_mul_ps(iq2,jq1); |
1220 | qq22 = _mm_mul_ps(iq2,jq2); |
1221 | qq23 = _mm_mul_ps(iq2,jq3); |
1222 | qq31 = _mm_mul_ps(iq3,jq1); |
1223 | qq32 = _mm_mul_ps(iq3,jq2); |
1224 | qq33 = _mm_mul_ps(iq3,jq3); |
1225 | |
1226 | /* Avoid stupid compiler warnings */ |
1227 | jnrA = jnrB = jnrC = jnrD = 0; |
1228 | j_coord_offsetA = 0; |
1229 | j_coord_offsetB = 0; |
1230 | j_coord_offsetC = 0; |
1231 | j_coord_offsetD = 0; |
1232 | |
1233 | outeriter = 0; |
1234 | inneriter = 0; |
1235 | |
1236 | for(iidx=0;iidx<4*DIM3;iidx++) |
1237 | { |
1238 | scratch[iidx] = 0.0; |
1239 | } |
1240 | |
1241 | /* Start outer loop over neighborlists */ |
1242 | for(iidx=0; iidx<nri; iidx++) |
1243 | { |
1244 | /* Load shift vector for this list */ |
1245 | i_shift_offset = DIM3*shiftidx[iidx]; |
1246 | |
1247 | /* Load limits for loop over neighbors */ |
1248 | j_index_start = jindex[iidx]; |
1249 | j_index_end = jindex[iidx+1]; |
1250 | |
1251 | /* Get outer coordinate index */ |
1252 | inr = iinr[iidx]; |
1253 | i_coord_offset = DIM3*inr; |
1254 | |
1255 | /* Load i particle coords and add shift vector */ |
1256 | gmx_mm_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset, |
1257 | &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3); |
1258 | |
1259 | fix0 = _mm_setzero_ps(); |
1260 | fiy0 = _mm_setzero_ps(); |
1261 | fiz0 = _mm_setzero_ps(); |
1262 | fix1 = _mm_setzero_ps(); |
1263 | fiy1 = _mm_setzero_ps(); |
1264 | fiz1 = _mm_setzero_ps(); |
1265 | fix2 = _mm_setzero_ps(); |
1266 | fiy2 = _mm_setzero_ps(); |
1267 | fiz2 = _mm_setzero_ps(); |
1268 | fix3 = _mm_setzero_ps(); |
1269 | fiy3 = _mm_setzero_ps(); |
1270 | fiz3 = _mm_setzero_ps(); |
1271 | |
1272 | /* Start inner kernel loop */ |
1273 | for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4) |
1274 | { |
1275 | |
1276 | /* Get j neighbor index, and coordinate index */ |
1277 | jnrA = jjnr[jidx]; |
1278 | jnrB = jjnr[jidx+1]; |
1279 | jnrC = jjnr[jidx+2]; |
1280 | jnrD = jjnr[jidx+3]; |
1281 | j_coord_offsetA = DIM3*jnrA; |
1282 | j_coord_offsetB = DIM3*jnrB; |
1283 | j_coord_offsetC = DIM3*jnrC; |
1284 | j_coord_offsetD = DIM3*jnrD; |
1285 | |
1286 | /* load j atom coordinates */ |
1287 | gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB, |
1288 | x+j_coord_offsetC,x+j_coord_offsetD, |
1289 | &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2, |
1290 | &jy2,&jz2,&jx3,&jy3,&jz3); |
1291 | |
1292 | /* Calculate displacement vector */ |
1293 | dx00 = _mm_sub_ps(ix0,jx0); |
1294 | dy00 = _mm_sub_ps(iy0,jy0); |
1295 | dz00 = _mm_sub_ps(iz0,jz0); |
1296 | dx11 = _mm_sub_ps(ix1,jx1); |
1297 | dy11 = _mm_sub_ps(iy1,jy1); |
1298 | dz11 = _mm_sub_ps(iz1,jz1); |
1299 | dx12 = _mm_sub_ps(ix1,jx2); |
1300 | dy12 = _mm_sub_ps(iy1,jy2); |
1301 | dz12 = _mm_sub_ps(iz1,jz2); |
1302 | dx13 = _mm_sub_ps(ix1,jx3); |
1303 | dy13 = _mm_sub_ps(iy1,jy3); |
1304 | dz13 = _mm_sub_ps(iz1,jz3); |
1305 | dx21 = _mm_sub_ps(ix2,jx1); |
1306 | dy21 = _mm_sub_ps(iy2,jy1); |
1307 | dz21 = _mm_sub_ps(iz2,jz1); |
1308 | dx22 = _mm_sub_ps(ix2,jx2); |
1309 | dy22 = _mm_sub_ps(iy2,jy2); |
1310 | dz22 = _mm_sub_ps(iz2,jz2); |
1311 | dx23 = _mm_sub_ps(ix2,jx3); |
1312 | dy23 = _mm_sub_ps(iy2,jy3); |
1313 | dz23 = _mm_sub_ps(iz2,jz3); |
1314 | dx31 = _mm_sub_ps(ix3,jx1); |
1315 | dy31 = _mm_sub_ps(iy3,jy1); |
1316 | dz31 = _mm_sub_ps(iz3,jz1); |
1317 | dx32 = _mm_sub_ps(ix3,jx2); |
1318 | dy32 = _mm_sub_ps(iy3,jy2); |
1319 | dz32 = _mm_sub_ps(iz3,jz2); |
1320 | dx33 = _mm_sub_ps(ix3,jx3); |
1321 | dy33 = _mm_sub_ps(iy3,jy3); |
1322 | dz33 = _mm_sub_ps(iz3,jz3); |
1323 | |
1324 | /* Calculate squared distance and things based on it */ |
1325 | rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00); |
1326 | rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11); |
1327 | rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12); |
1328 | rsq13 = gmx_mm_calc_rsq_ps(dx13,dy13,dz13); |
1329 | rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21); |
1330 | rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22); |
1331 | rsq23 = gmx_mm_calc_rsq_ps(dx23,dy23,dz23); |
1332 | rsq31 = gmx_mm_calc_rsq_ps(dx31,dy31,dz31); |
1333 | rsq32 = gmx_mm_calc_rsq_ps(dx32,dy32,dz32); |
1334 | rsq33 = gmx_mm_calc_rsq_ps(dx33,dy33,dz33); |
1335 | |
1336 | rinv00 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq00); |
1337 | rinv11 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq11); |
1338 | rinv12 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq12); |
1339 | rinv13 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq13); |
1340 | rinv21 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq21); |
1341 | rinv22 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq22); |
1342 | rinv23 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq23); |
1343 | rinv31 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq31); |
1344 | rinv32 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq32); |
1345 | rinv33 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq33); |
1346 | |
1347 | rinvsq11 = _mm_mul_ps(rinv11,rinv11); |
1348 | rinvsq12 = _mm_mul_ps(rinv12,rinv12); |
1349 | rinvsq13 = _mm_mul_ps(rinv13,rinv13); |
1350 | rinvsq21 = _mm_mul_ps(rinv21,rinv21); |
1351 | rinvsq22 = _mm_mul_ps(rinv22,rinv22); |
1352 | rinvsq23 = _mm_mul_ps(rinv23,rinv23); |
1353 | rinvsq31 = _mm_mul_ps(rinv31,rinv31); |
1354 | rinvsq32 = _mm_mul_ps(rinv32,rinv32); |
1355 | rinvsq33 = _mm_mul_ps(rinv33,rinv33); |
1356 | |
1357 | fjx0 = _mm_setzero_ps(); |
1358 | fjy0 = _mm_setzero_ps(); |
1359 | fjz0 = _mm_setzero_ps(); |
1360 | fjx1 = _mm_setzero_ps(); |
1361 | fjy1 = _mm_setzero_ps(); |
1362 | fjz1 = _mm_setzero_ps(); |
1363 | fjx2 = _mm_setzero_ps(); |
1364 | fjy2 = _mm_setzero_ps(); |
1365 | fjz2 = _mm_setzero_ps(); |
1366 | fjx3 = _mm_setzero_ps(); |
1367 | fjy3 = _mm_setzero_ps(); |
1368 | fjz3 = _mm_setzero_ps(); |
1369 | |
1370 | /************************** |
1371 | * CALCULATE INTERACTIONS * |
1372 | **************************/ |
1373 | |
1374 | r00 = _mm_mul_ps(rsq00,rinv00); |
1375 | |
1376 | /* Calculate table index by multiplying r with table scale and truncate to integer */ |
1377 | rt = _mm_mul_ps(r00,vftabscale); |
1378 | vfitab = _mm_cvttps_epi32(rt); |
1379 | vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (rt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1380 | vfitab = _mm_slli_epi32(vfitab,3); |
1381 | |
1382 | /* CUBIC SPLINE TABLE DISPERSION */ |
1383 | Y = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0)(__extension__ ({ __v4si __a = (__v4si)(vfitab); __a[(0) & 3];})) ); |
1384 | F = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1)(__extension__ ({ __v4si __a = (__v4si)(vfitab); __a[(1) & 3];})) ); |
1385 | G = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2)(__extension__ ({ __v4si __a = (__v4si)(vfitab); __a[(2) & 3];})) ); |
1386 | H = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3)(__extension__ ({ __v4si __a = (__v4si)(vfitab); __a[(3) & 3];})) ); |
1387 | _MM_TRANSPOSE4_PS(Y,F,G,H)do { __m128 tmp3, tmp2, tmp1, tmp0; tmp0 = _mm_unpacklo_ps((Y ), (F)); tmp2 = _mm_unpacklo_ps((G), (H)); tmp1 = _mm_unpackhi_ps ((Y), (F)); tmp3 = _mm_unpackhi_ps((G), (H)); (Y) = _mm_movelh_ps (tmp0, tmp2); (F) = _mm_movehl_ps(tmp2, tmp0); (G) = _mm_movelh_ps (tmp1, tmp3); (H) = _mm_movehl_ps(tmp3, tmp1); } while (0); |
1388 | Heps = _mm_mul_ps(vfeps,H); |
1389 | Fp = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps))); |
1390 | FF = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps)))); |
1391 | fvdw6 = _mm_mul_ps(c6_00,FF); |
1392 | |
1393 | /* CUBIC SPLINE TABLE REPULSION */ |
1394 | vfitab = _mm_add_epi32(vfitab,ifour); |
1395 | Y = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0)(__extension__ ({ __v4si __a = (__v4si)(vfitab); __a[(0) & 3];})) ); |
1396 | F = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1)(__extension__ ({ __v4si __a = (__v4si)(vfitab); __a[(1) & 3];})) ); |
1397 | G = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2)(__extension__ ({ __v4si __a = (__v4si)(vfitab); __a[(2) & 3];})) ); |
1398 | H = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3)(__extension__ ({ __v4si __a = (__v4si)(vfitab); __a[(3) & 3];})) ); |
1399 | _MM_TRANSPOSE4_PS(Y,F,G,H)do { __m128 tmp3, tmp2, tmp1, tmp0; tmp0 = _mm_unpacklo_ps((Y ), (F)); tmp2 = _mm_unpacklo_ps((G), (H)); tmp1 = _mm_unpackhi_ps ((Y), (F)); tmp3 = _mm_unpackhi_ps((G), (H)); (Y) = _mm_movelh_ps (tmp0, tmp2); (F) = _mm_movehl_ps(tmp2, tmp0); (G) = _mm_movelh_ps (tmp1, tmp3); (H) = _mm_movehl_ps(tmp3, tmp1); } while (0); |
1400 | Heps = _mm_mul_ps(vfeps,H); |
1401 | Fp = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps))); |
1402 | FF = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps)))); |
1403 | fvdw12 = _mm_mul_ps(c12_00,FF); |
1404 | fvdw = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00))); |
1405 | |
1406 | fscal = fvdw; |
1407 | |
1408 | /* Calculate temporary vectorial force */ |
1409 | tx = _mm_mul_ps(fscal,dx00); |
1410 | ty = _mm_mul_ps(fscal,dy00); |
1411 | tz = _mm_mul_ps(fscal,dz00); |
1412 | |
1413 | /* Update vectorial force */ |
1414 | fix0 = _mm_add_ps(fix0,tx); |
1415 | fiy0 = _mm_add_ps(fiy0,ty); |
1416 | fiz0 = _mm_add_ps(fiz0,tz); |
1417 | |
1418 | fjx0 = _mm_add_ps(fjx0,tx); |
1419 | fjy0 = _mm_add_ps(fjy0,ty); |
1420 | fjz0 = _mm_add_ps(fjz0,tz); |
1421 | |
1422 | /************************** |
1423 | * CALCULATE INTERACTIONS * |
1424 | **************************/ |
1425 | |
1426 | /* COULOMB ELECTROSTATICS */ |
1427 | velec = _mm_mul_ps(qq11,rinv11); |
1428 | felec = _mm_mul_ps(velec,rinvsq11); |
1429 | |
1430 | fscal = felec; |
1431 | |
1432 | /* Calculate temporary vectorial force */ |
1433 | tx = _mm_mul_ps(fscal,dx11); |
1434 | ty = _mm_mul_ps(fscal,dy11); |
1435 | tz = _mm_mul_ps(fscal,dz11); |
1436 | |
1437 | /* Update vectorial force */ |
1438 | fix1 = _mm_add_ps(fix1,tx); |
1439 | fiy1 = _mm_add_ps(fiy1,ty); |
1440 | fiz1 = _mm_add_ps(fiz1,tz); |
1441 | |
1442 | fjx1 = _mm_add_ps(fjx1,tx); |
1443 | fjy1 = _mm_add_ps(fjy1,ty); |
1444 | fjz1 = _mm_add_ps(fjz1,tz); |
1445 | |
1446 | /************************** |
1447 | * CALCULATE INTERACTIONS * |
1448 | **************************/ |
1449 | |
1450 | /* COULOMB ELECTROSTATICS */ |
1451 | velec = _mm_mul_ps(qq12,rinv12); |
1452 | felec = _mm_mul_ps(velec,rinvsq12); |
1453 | |
1454 | fscal = felec; |
1455 | |
1456 | /* Calculate temporary vectorial force */ |
1457 | tx = _mm_mul_ps(fscal,dx12); |
1458 | ty = _mm_mul_ps(fscal,dy12); |
1459 | tz = _mm_mul_ps(fscal,dz12); |
1460 | |
1461 | /* Update vectorial force */ |
1462 | fix1 = _mm_add_ps(fix1,tx); |
1463 | fiy1 = _mm_add_ps(fiy1,ty); |
1464 | fiz1 = _mm_add_ps(fiz1,tz); |
1465 | |
1466 | fjx2 = _mm_add_ps(fjx2,tx); |
1467 | fjy2 = _mm_add_ps(fjy2,ty); |
1468 | fjz2 = _mm_add_ps(fjz2,tz); |
1469 | |
1470 | /************************** |
1471 | * CALCULATE INTERACTIONS * |
1472 | **************************/ |
1473 | |
1474 | /* COULOMB ELECTROSTATICS */ |
1475 | velec = _mm_mul_ps(qq13,rinv13); |
1476 | felec = _mm_mul_ps(velec,rinvsq13); |
1477 | |
1478 | fscal = felec; |
1479 | |
1480 | /* Calculate temporary vectorial force */ |
1481 | tx = _mm_mul_ps(fscal,dx13); |
1482 | ty = _mm_mul_ps(fscal,dy13); |
1483 | tz = _mm_mul_ps(fscal,dz13); |
1484 | |
1485 | /* Update vectorial force */ |
1486 | fix1 = _mm_add_ps(fix1,tx); |
1487 | fiy1 = _mm_add_ps(fiy1,ty); |
1488 | fiz1 = _mm_add_ps(fiz1,tz); |
1489 | |
1490 | fjx3 = _mm_add_ps(fjx3,tx); |
1491 | fjy3 = _mm_add_ps(fjy3,ty); |
1492 | fjz3 = _mm_add_ps(fjz3,tz); |
1493 | |
1494 | /************************** |
1495 | * CALCULATE INTERACTIONS * |
1496 | **************************/ |
1497 | |
1498 | /* COULOMB ELECTROSTATICS */ |
1499 | velec = _mm_mul_ps(qq21,rinv21); |
1500 | felec = _mm_mul_ps(velec,rinvsq21); |
1501 | |
1502 | fscal = felec; |
1503 | |
1504 | /* Calculate temporary vectorial force */ |
1505 | tx = _mm_mul_ps(fscal,dx21); |
1506 | ty = _mm_mul_ps(fscal,dy21); |
1507 | tz = _mm_mul_ps(fscal,dz21); |
1508 | |
1509 | /* Update vectorial force */ |
1510 | fix2 = _mm_add_ps(fix2,tx); |
1511 | fiy2 = _mm_add_ps(fiy2,ty); |
1512 | fiz2 = _mm_add_ps(fiz2,tz); |
1513 | |
1514 | fjx1 = _mm_add_ps(fjx1,tx); |
1515 | fjy1 = _mm_add_ps(fjy1,ty); |
1516 | fjz1 = _mm_add_ps(fjz1,tz); |
1517 | |
1518 | /************************** |
1519 | * CALCULATE INTERACTIONS * |
1520 | **************************/ |
1521 | |
1522 | /* COULOMB ELECTROSTATICS */ |
1523 | velec = _mm_mul_ps(qq22,rinv22); |
1524 | felec = _mm_mul_ps(velec,rinvsq22); |
1525 | |
1526 | fscal = felec; |
1527 | |
1528 | /* Calculate temporary vectorial force */ |
1529 | tx = _mm_mul_ps(fscal,dx22); |
1530 | ty = _mm_mul_ps(fscal,dy22); |
1531 | tz = _mm_mul_ps(fscal,dz22); |
1532 | |
1533 | /* Update vectorial force */ |
1534 | fix2 = _mm_add_ps(fix2,tx); |
1535 | fiy2 = _mm_add_ps(fiy2,ty); |
1536 | fiz2 = _mm_add_ps(fiz2,tz); |
1537 | |
1538 | fjx2 = _mm_add_ps(fjx2,tx); |
1539 | fjy2 = _mm_add_ps(fjy2,ty); |
1540 | fjz2 = _mm_add_ps(fjz2,tz); |
1541 | |
1542 | /************************** |
1543 | * CALCULATE INTERACTIONS * |
1544 | **************************/ |
1545 | |
1546 | /* COULOMB ELECTROSTATICS */ |
1547 | velec = _mm_mul_ps(qq23,rinv23); |
1548 | felec = _mm_mul_ps(velec,rinvsq23); |
1549 | |
1550 | fscal = felec; |
1551 | |
1552 | /* Calculate temporary vectorial force */ |
1553 | tx = _mm_mul_ps(fscal,dx23); |
1554 | ty = _mm_mul_ps(fscal,dy23); |
1555 | tz = _mm_mul_ps(fscal,dz23); |
1556 | |
1557 | /* Update vectorial force */ |
1558 | fix2 = _mm_add_ps(fix2,tx); |
1559 | fiy2 = _mm_add_ps(fiy2,ty); |
1560 | fiz2 = _mm_add_ps(fiz2,tz); |
1561 | |
1562 | fjx3 = _mm_add_ps(fjx3,tx); |
1563 | fjy3 = _mm_add_ps(fjy3,ty); |
1564 | fjz3 = _mm_add_ps(fjz3,tz); |
1565 | |
1566 | /************************** |
1567 | * CALCULATE INTERACTIONS * |
1568 | **************************/ |
1569 | |
1570 | /* COULOMB ELECTROSTATICS */ |
1571 | velec = _mm_mul_ps(qq31,rinv31); |
1572 | felec = _mm_mul_ps(velec,rinvsq31); |
1573 | |
1574 | fscal = felec; |
1575 | |
1576 | /* Calculate temporary vectorial force */ |
1577 | tx = _mm_mul_ps(fscal,dx31); |
1578 | ty = _mm_mul_ps(fscal,dy31); |
1579 | tz = _mm_mul_ps(fscal,dz31); |
1580 | |
1581 | /* Update vectorial force */ |
1582 | fix3 = _mm_add_ps(fix3,tx); |
1583 | fiy3 = _mm_add_ps(fiy3,ty); |
1584 | fiz3 = _mm_add_ps(fiz3,tz); |
1585 | |
1586 | fjx1 = _mm_add_ps(fjx1,tx); |
1587 | fjy1 = _mm_add_ps(fjy1,ty); |
1588 | fjz1 = _mm_add_ps(fjz1,tz); |
1589 | |
1590 | /************************** |
1591 | * CALCULATE INTERACTIONS * |
1592 | **************************/ |
1593 | |
1594 | /* COULOMB ELECTROSTATICS */ |
1595 | velec = _mm_mul_ps(qq32,rinv32); |
1596 | felec = _mm_mul_ps(velec,rinvsq32); |
1597 | |
1598 | fscal = felec; |
1599 | |
1600 | /* Calculate temporary vectorial force */ |
1601 | tx = _mm_mul_ps(fscal,dx32); |
1602 | ty = _mm_mul_ps(fscal,dy32); |
1603 | tz = _mm_mul_ps(fscal,dz32); |
1604 | |
1605 | /* Update vectorial force */ |
1606 | fix3 = _mm_add_ps(fix3,tx); |
1607 | fiy3 = _mm_add_ps(fiy3,ty); |
1608 | fiz3 = _mm_add_ps(fiz3,tz); |
1609 | |
1610 | fjx2 = _mm_add_ps(fjx2,tx); |
1611 | fjy2 = _mm_add_ps(fjy2,ty); |
1612 | fjz2 = _mm_add_ps(fjz2,tz); |
1613 | |
1614 | /************************** |
1615 | * CALCULATE INTERACTIONS * |
1616 | **************************/ |
1617 | |
1618 | /* COULOMB ELECTROSTATICS */ |
1619 | velec = _mm_mul_ps(qq33,rinv33); |
1620 | felec = _mm_mul_ps(velec,rinvsq33); |
1621 | |
1622 | fscal = felec; |
1623 | |
1624 | /* Calculate temporary vectorial force */ |
1625 | tx = _mm_mul_ps(fscal,dx33); |
1626 | ty = _mm_mul_ps(fscal,dy33); |
1627 | tz = _mm_mul_ps(fscal,dz33); |
1628 | |
1629 | /* Update vectorial force */ |
1630 | fix3 = _mm_add_ps(fix3,tx); |
1631 | fiy3 = _mm_add_ps(fiy3,ty); |
1632 | fiz3 = _mm_add_ps(fiz3,tz); |
1633 | |
1634 | fjx3 = _mm_add_ps(fjx3,tx); |
1635 | fjy3 = _mm_add_ps(fjy3,ty); |
1636 | fjz3 = _mm_add_ps(fjz3,tz); |
1637 | |
1638 | fjptrA = f+j_coord_offsetA; |
1639 | fjptrB = f+j_coord_offsetB; |
1640 | fjptrC = f+j_coord_offsetC; |
1641 | fjptrD = f+j_coord_offsetD; |
1642 | |
1643 | gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD, |
1644 | fjx0,fjy0,fjz0,fjx1,fjy1,fjz1, |
1645 | fjx2,fjy2,fjz2,fjx3,fjy3,fjz3); |
1646 | |
1647 | /* Inner loop uses 294 flops */ |
1648 | } |
1649 | |
1650 | if(jidx<j_index_end) |
1651 | { |
1652 | |
1653 | /* Get j neighbor index, and coordinate index */ |
1654 | jnrlistA = jjnr[jidx]; |
1655 | jnrlistB = jjnr[jidx+1]; |
1656 | jnrlistC = jjnr[jidx+2]; |
1657 | jnrlistD = jjnr[jidx+3]; |
1658 | /* Sign of each element will be negative for non-real atoms. |
1659 | * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones, |
1660 | * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries. |
1661 | */ |
1662 | dummy_mask = gmx_mm_castsi128_ps_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())); |
1663 | jnrA = (jnrlistA>=0) ? jnrlistA : 0; |
1664 | jnrB = (jnrlistB>=0) ? jnrlistB : 0; |
1665 | jnrC = (jnrlistC>=0) ? jnrlistC : 0; |
1666 | jnrD = (jnrlistD>=0) ? jnrlistD : 0; |
1667 | j_coord_offsetA = DIM3*jnrA; |
1668 | j_coord_offsetB = DIM3*jnrB; |
1669 | j_coord_offsetC = DIM3*jnrC; |
1670 | j_coord_offsetD = DIM3*jnrD; |
1671 | |
1672 | /* load j atom coordinates */ |
1673 | gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB, |
1674 | x+j_coord_offsetC,x+j_coord_offsetD, |
1675 | &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2, |
1676 | &jy2,&jz2,&jx3,&jy3,&jz3); |
1677 | |
1678 | /* Calculate displacement vector */ |
1679 | dx00 = _mm_sub_ps(ix0,jx0); |
1680 | dy00 = _mm_sub_ps(iy0,jy0); |
1681 | dz00 = _mm_sub_ps(iz0,jz0); |
1682 | dx11 = _mm_sub_ps(ix1,jx1); |
1683 | dy11 = _mm_sub_ps(iy1,jy1); |
1684 | dz11 = _mm_sub_ps(iz1,jz1); |
1685 | dx12 = _mm_sub_ps(ix1,jx2); |
1686 | dy12 = _mm_sub_ps(iy1,jy2); |
1687 | dz12 = _mm_sub_ps(iz1,jz2); |
1688 | dx13 = _mm_sub_ps(ix1,jx3); |
1689 | dy13 = _mm_sub_ps(iy1,jy3); |
1690 | dz13 = _mm_sub_ps(iz1,jz3); |
1691 | dx21 = _mm_sub_ps(ix2,jx1); |
1692 | dy21 = _mm_sub_ps(iy2,jy1); |
1693 | dz21 = _mm_sub_ps(iz2,jz1); |
1694 | dx22 = _mm_sub_ps(ix2,jx2); |
1695 | dy22 = _mm_sub_ps(iy2,jy2); |
1696 | dz22 = _mm_sub_ps(iz2,jz2); |
1697 | dx23 = _mm_sub_ps(ix2,jx3); |
1698 | dy23 = _mm_sub_ps(iy2,jy3); |
1699 | dz23 = _mm_sub_ps(iz2,jz3); |
1700 | dx31 = _mm_sub_ps(ix3,jx1); |
1701 | dy31 = _mm_sub_ps(iy3,jy1); |
1702 | dz31 = _mm_sub_ps(iz3,jz1); |
1703 | dx32 = _mm_sub_ps(ix3,jx2); |
1704 | dy32 = _mm_sub_ps(iy3,jy2); |
1705 | dz32 = _mm_sub_ps(iz3,jz2); |
1706 | dx33 = _mm_sub_ps(ix3,jx3); |
1707 | dy33 = _mm_sub_ps(iy3,jy3); |
1708 | dz33 = _mm_sub_ps(iz3,jz3); |
1709 | |
1710 | /* Calculate squared distance and things based on it */ |
1711 | rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00); |
1712 | rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11); |
1713 | rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12); |
1714 | rsq13 = gmx_mm_calc_rsq_ps(dx13,dy13,dz13); |
1715 | rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21); |
1716 | rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22); |
1717 | rsq23 = gmx_mm_calc_rsq_ps(dx23,dy23,dz23); |
1718 | rsq31 = gmx_mm_calc_rsq_ps(dx31,dy31,dz31); |
1719 | rsq32 = gmx_mm_calc_rsq_ps(dx32,dy32,dz32); |
1720 | rsq33 = gmx_mm_calc_rsq_ps(dx33,dy33,dz33); |
1721 | |
1722 | rinv00 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq00); |
1723 | rinv11 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq11); |
1724 | rinv12 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq12); |
1725 | rinv13 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq13); |
1726 | rinv21 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq21); |
1727 | rinv22 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq22); |
1728 | rinv23 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq23); |
1729 | rinv31 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq31); |
1730 | rinv32 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq32); |
1731 | rinv33 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq33); |
1732 | |
1733 | rinvsq11 = _mm_mul_ps(rinv11,rinv11); |
1734 | rinvsq12 = _mm_mul_ps(rinv12,rinv12); |
1735 | rinvsq13 = _mm_mul_ps(rinv13,rinv13); |
1736 | rinvsq21 = _mm_mul_ps(rinv21,rinv21); |
1737 | rinvsq22 = _mm_mul_ps(rinv22,rinv22); |
1738 | rinvsq23 = _mm_mul_ps(rinv23,rinv23); |
1739 | rinvsq31 = _mm_mul_ps(rinv31,rinv31); |
1740 | rinvsq32 = _mm_mul_ps(rinv32,rinv32); |
1741 | rinvsq33 = _mm_mul_ps(rinv33,rinv33); |
1742 | |
1743 | fjx0 = _mm_setzero_ps(); |
1744 | fjy0 = _mm_setzero_ps(); |
1745 | fjz0 = _mm_setzero_ps(); |
1746 | fjx1 = _mm_setzero_ps(); |
1747 | fjy1 = _mm_setzero_ps(); |
1748 | fjz1 = _mm_setzero_ps(); |
1749 | fjx2 = _mm_setzero_ps(); |
1750 | fjy2 = _mm_setzero_ps(); |
1751 | fjz2 = _mm_setzero_ps(); |
1752 | fjx3 = _mm_setzero_ps(); |
1753 | fjy3 = _mm_setzero_ps(); |
1754 | fjz3 = _mm_setzero_ps(); |
1755 | |
1756 | /************************** |
1757 | * CALCULATE INTERACTIONS * |
1758 | **************************/ |
1759 | |
1760 | r00 = _mm_mul_ps(rsq00,rinv00); |
1761 | r00 = _mm_andnot_ps(dummy_mask,r00); |
1762 | |
1763 | /* Calculate table index by multiplying r with table scale and truncate to integer */ |
1764 | rt = _mm_mul_ps(r00,vftabscale); |
1765 | vfitab = _mm_cvttps_epi32(rt); |
1766 | vfeps = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (rt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1767 | vfitab = _mm_slli_epi32(vfitab,3); |
1768 | |
1769 | /* CUBIC SPLINE TABLE DISPERSION */ |
1770 | Y = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0)(__extension__ ({ __v4si __a = (__v4si)(vfitab); __a[(0) & 3];})) ); |
1771 | F = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1)(__extension__ ({ __v4si __a = (__v4si)(vfitab); __a[(1) & 3];})) ); |
1772 | G = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2)(__extension__ ({ __v4si __a = (__v4si)(vfitab); __a[(2) & 3];})) ); |
1773 | H = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3)(__extension__ ({ __v4si __a = (__v4si)(vfitab); __a[(3) & 3];})) ); |
1774 | _MM_TRANSPOSE4_PS(Y,F,G,H)do { __m128 tmp3, tmp2, tmp1, tmp0; tmp0 = _mm_unpacklo_ps((Y ), (F)); tmp2 = _mm_unpacklo_ps((G), (H)); tmp1 = _mm_unpackhi_ps ((Y), (F)); tmp3 = _mm_unpackhi_ps((G), (H)); (Y) = _mm_movelh_ps (tmp0, tmp2); (F) = _mm_movehl_ps(tmp2, tmp0); (G) = _mm_movelh_ps (tmp1, tmp3); (H) = _mm_movehl_ps(tmp3, tmp1); } while (0); |
1775 | Heps = _mm_mul_ps(vfeps,H); |
1776 | Fp = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps))); |
1777 | FF = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps)))); |
1778 | fvdw6 = _mm_mul_ps(c6_00,FF); |
1779 | |
1780 | /* CUBIC SPLINE TABLE REPULSION */ |
1781 | vfitab = _mm_add_epi32(vfitab,ifour); |
1782 | Y = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0)(__extension__ ({ __v4si __a = (__v4si)(vfitab); __a[(0) & 3];})) ); |
1783 | F = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1)(__extension__ ({ __v4si __a = (__v4si)(vfitab); __a[(1) & 3];})) ); |
1784 | G = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2)(__extension__ ({ __v4si __a = (__v4si)(vfitab); __a[(2) & 3];})) ); |
1785 | H = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3)(__extension__ ({ __v4si __a = (__v4si)(vfitab); __a[(3) & 3];})) ); |
1786 | _MM_TRANSPOSE4_PS(Y,F,G,H)do { __m128 tmp3, tmp2, tmp1, tmp0; tmp0 = _mm_unpacklo_ps((Y ), (F)); tmp2 = _mm_unpacklo_ps((G), (H)); tmp1 = _mm_unpackhi_ps ((Y), (F)); tmp3 = _mm_unpackhi_ps((G), (H)); (Y) = _mm_movelh_ps (tmp0, tmp2); (F) = _mm_movehl_ps(tmp2, tmp0); (G) = _mm_movelh_ps (tmp1, tmp3); (H) = _mm_movehl_ps(tmp3, tmp1); } while (0); |
1787 | Heps = _mm_mul_ps(vfeps,H); |
1788 | Fp = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps))); |
1789 | FF = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps)))); |
1790 | fvdw12 = _mm_mul_ps(c12_00,FF); |
1791 | fvdw = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00))); |
1792 | |
1793 | fscal = fvdw; |
1794 | |
1795 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
1796 | |
1797 | /* Calculate temporary vectorial force */ |
1798 | tx = _mm_mul_ps(fscal,dx00); |
1799 | ty = _mm_mul_ps(fscal,dy00); |
1800 | tz = _mm_mul_ps(fscal,dz00); |
1801 | |
1802 | /* Update vectorial force */ |
1803 | fix0 = _mm_add_ps(fix0,tx); |
1804 | fiy0 = _mm_add_ps(fiy0,ty); |
1805 | fiz0 = _mm_add_ps(fiz0,tz); |
1806 | |
1807 | fjx0 = _mm_add_ps(fjx0,tx); |
1808 | fjy0 = _mm_add_ps(fjy0,ty); |
1809 | fjz0 = _mm_add_ps(fjz0,tz); |
1810 | |
1811 | /************************** |
1812 | * CALCULATE INTERACTIONS * |
1813 | **************************/ |
1814 | |
1815 | /* COULOMB ELECTROSTATICS */ |
1816 | velec = _mm_mul_ps(qq11,rinv11); |
1817 | felec = _mm_mul_ps(velec,rinvsq11); |
1818 | |
1819 | fscal = felec; |
1820 | |
1821 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
1822 | |
1823 | /* Calculate temporary vectorial force */ |
1824 | tx = _mm_mul_ps(fscal,dx11); |
1825 | ty = _mm_mul_ps(fscal,dy11); |
1826 | tz = _mm_mul_ps(fscal,dz11); |
1827 | |
1828 | /* Update vectorial force */ |
1829 | fix1 = _mm_add_ps(fix1,tx); |
1830 | fiy1 = _mm_add_ps(fiy1,ty); |
1831 | fiz1 = _mm_add_ps(fiz1,tz); |
1832 | |
1833 | fjx1 = _mm_add_ps(fjx1,tx); |
1834 | fjy1 = _mm_add_ps(fjy1,ty); |
1835 | fjz1 = _mm_add_ps(fjz1,tz); |
1836 | |
1837 | /************************** |
1838 | * CALCULATE INTERACTIONS * |
1839 | **************************/ |
1840 | |
1841 | /* COULOMB ELECTROSTATICS */ |
1842 | velec = _mm_mul_ps(qq12,rinv12); |
1843 | felec = _mm_mul_ps(velec,rinvsq12); |
1844 | |
1845 | fscal = felec; |
1846 | |
1847 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
1848 | |
1849 | /* Calculate temporary vectorial force */ |
1850 | tx = _mm_mul_ps(fscal,dx12); |
1851 | ty = _mm_mul_ps(fscal,dy12); |
1852 | tz = _mm_mul_ps(fscal,dz12); |
1853 | |
1854 | /* Update vectorial force */ |
1855 | fix1 = _mm_add_ps(fix1,tx); |
1856 | fiy1 = _mm_add_ps(fiy1,ty); |
1857 | fiz1 = _mm_add_ps(fiz1,tz); |
1858 | |
1859 | fjx2 = _mm_add_ps(fjx2,tx); |
1860 | fjy2 = _mm_add_ps(fjy2,ty); |
1861 | fjz2 = _mm_add_ps(fjz2,tz); |
1862 | |
1863 | /************************** |
1864 | * CALCULATE INTERACTIONS * |
1865 | **************************/ |
1866 | |
1867 | /* COULOMB ELECTROSTATICS */ |
1868 | velec = _mm_mul_ps(qq13,rinv13); |
1869 | felec = _mm_mul_ps(velec,rinvsq13); |
1870 | |
1871 | fscal = felec; |
1872 | |
1873 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
1874 | |
1875 | /* Calculate temporary vectorial force */ |
1876 | tx = _mm_mul_ps(fscal,dx13); |
1877 | ty = _mm_mul_ps(fscal,dy13); |
1878 | tz = _mm_mul_ps(fscal,dz13); |
1879 | |
1880 | /* Update vectorial force */ |
1881 | fix1 = _mm_add_ps(fix1,tx); |
1882 | fiy1 = _mm_add_ps(fiy1,ty); |
1883 | fiz1 = _mm_add_ps(fiz1,tz); |
1884 | |
1885 | fjx3 = _mm_add_ps(fjx3,tx); |
1886 | fjy3 = _mm_add_ps(fjy3,ty); |
1887 | fjz3 = _mm_add_ps(fjz3,tz); |
1888 | |
1889 | /************************** |
1890 | * CALCULATE INTERACTIONS * |
1891 | **************************/ |
1892 | |
1893 | /* COULOMB ELECTROSTATICS */ |
1894 | velec = _mm_mul_ps(qq21,rinv21); |
1895 | felec = _mm_mul_ps(velec,rinvsq21); |
1896 | |
1897 | fscal = felec; |
1898 | |
1899 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
1900 | |
1901 | /* Calculate temporary vectorial force */ |
1902 | tx = _mm_mul_ps(fscal,dx21); |
1903 | ty = _mm_mul_ps(fscal,dy21); |
1904 | tz = _mm_mul_ps(fscal,dz21); |
1905 | |
1906 | /* Update vectorial force */ |
1907 | fix2 = _mm_add_ps(fix2,tx); |
1908 | fiy2 = _mm_add_ps(fiy2,ty); |
1909 | fiz2 = _mm_add_ps(fiz2,tz); |
1910 | |
1911 | fjx1 = _mm_add_ps(fjx1,tx); |
1912 | fjy1 = _mm_add_ps(fjy1,ty); |
1913 | fjz1 = _mm_add_ps(fjz1,tz); |
1914 | |
1915 | /************************** |
1916 | * CALCULATE INTERACTIONS * |
1917 | **************************/ |
1918 | |
1919 | /* COULOMB ELECTROSTATICS */ |
1920 | velec = _mm_mul_ps(qq22,rinv22); |
1921 | felec = _mm_mul_ps(velec,rinvsq22); |
1922 | |
1923 | fscal = felec; |
1924 | |
1925 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
1926 | |
1927 | /* Calculate temporary vectorial force */ |
1928 | tx = _mm_mul_ps(fscal,dx22); |
1929 | ty = _mm_mul_ps(fscal,dy22); |
1930 | tz = _mm_mul_ps(fscal,dz22); |
1931 | |
1932 | /* Update vectorial force */ |
1933 | fix2 = _mm_add_ps(fix2,tx); |
1934 | fiy2 = _mm_add_ps(fiy2,ty); |
1935 | fiz2 = _mm_add_ps(fiz2,tz); |
1936 | |
1937 | fjx2 = _mm_add_ps(fjx2,tx); |
1938 | fjy2 = _mm_add_ps(fjy2,ty); |
1939 | fjz2 = _mm_add_ps(fjz2,tz); |
1940 | |
1941 | /************************** |
1942 | * CALCULATE INTERACTIONS * |
1943 | **************************/ |
1944 | |
1945 | /* COULOMB ELECTROSTATICS */ |
1946 | velec = _mm_mul_ps(qq23,rinv23); |
1947 | felec = _mm_mul_ps(velec,rinvsq23); |
1948 | |
1949 | fscal = felec; |
1950 | |
1951 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
1952 | |
1953 | /* Calculate temporary vectorial force */ |
1954 | tx = _mm_mul_ps(fscal,dx23); |
1955 | ty = _mm_mul_ps(fscal,dy23); |
1956 | tz = _mm_mul_ps(fscal,dz23); |
1957 | |
1958 | /* Update vectorial force */ |
1959 | fix2 = _mm_add_ps(fix2,tx); |
1960 | fiy2 = _mm_add_ps(fiy2,ty); |
1961 | fiz2 = _mm_add_ps(fiz2,tz); |
1962 | |
1963 | fjx3 = _mm_add_ps(fjx3,tx); |
1964 | fjy3 = _mm_add_ps(fjy3,ty); |
1965 | fjz3 = _mm_add_ps(fjz3,tz); |
1966 | |
1967 | /************************** |
1968 | * CALCULATE INTERACTIONS * |
1969 | **************************/ |
1970 | |
1971 | /* COULOMB ELECTROSTATICS */ |
1972 | velec = _mm_mul_ps(qq31,rinv31); |
1973 | felec = _mm_mul_ps(velec,rinvsq31); |
1974 | |
1975 | fscal = felec; |
1976 | |
1977 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
1978 | |
1979 | /* Calculate temporary vectorial force */ |
1980 | tx = _mm_mul_ps(fscal,dx31); |
1981 | ty = _mm_mul_ps(fscal,dy31); |
1982 | tz = _mm_mul_ps(fscal,dz31); |
1983 | |
1984 | /* Update vectorial force */ |
1985 | fix3 = _mm_add_ps(fix3,tx); |
1986 | fiy3 = _mm_add_ps(fiy3,ty); |
1987 | fiz3 = _mm_add_ps(fiz3,tz); |
1988 | |
1989 | fjx1 = _mm_add_ps(fjx1,tx); |
1990 | fjy1 = _mm_add_ps(fjy1,ty); |
1991 | fjz1 = _mm_add_ps(fjz1,tz); |
1992 | |
1993 | /************************** |
1994 | * CALCULATE INTERACTIONS * |
1995 | **************************/ |
1996 | |
1997 | /* COULOMB ELECTROSTATICS */ |
1998 | velec = _mm_mul_ps(qq32,rinv32); |
1999 | felec = _mm_mul_ps(velec,rinvsq32); |
2000 | |
2001 | fscal = felec; |
2002 | |
2003 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
2004 | |
2005 | /* Calculate temporary vectorial force */ |
2006 | tx = _mm_mul_ps(fscal,dx32); |
2007 | ty = _mm_mul_ps(fscal,dy32); |
2008 | tz = _mm_mul_ps(fscal,dz32); |
2009 | |
2010 | /* Update vectorial force */ |
2011 | fix3 = _mm_add_ps(fix3,tx); |
2012 | fiy3 = _mm_add_ps(fiy3,ty); |
2013 | fiz3 = _mm_add_ps(fiz3,tz); |
2014 | |
2015 | fjx2 = _mm_add_ps(fjx2,tx); |
2016 | fjy2 = _mm_add_ps(fjy2,ty); |
2017 | fjz2 = _mm_add_ps(fjz2,tz); |
2018 | |
2019 | /************************** |
2020 | * CALCULATE INTERACTIONS * |
2021 | **************************/ |
2022 | |
2023 | /* COULOMB ELECTROSTATICS */ |
2024 | velec = _mm_mul_ps(qq33,rinv33); |
2025 | felec = _mm_mul_ps(velec,rinvsq33); |
2026 | |
2027 | fscal = felec; |
2028 | |
2029 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
2030 | |
2031 | /* Calculate temporary vectorial force */ |
2032 | tx = _mm_mul_ps(fscal,dx33); |
2033 | ty = _mm_mul_ps(fscal,dy33); |
2034 | tz = _mm_mul_ps(fscal,dz33); |
2035 | |
2036 | /* Update vectorial force */ |
2037 | fix3 = _mm_add_ps(fix3,tx); |
2038 | fiy3 = _mm_add_ps(fiy3,ty); |
2039 | fiz3 = _mm_add_ps(fiz3,tz); |
2040 | |
2041 | fjx3 = _mm_add_ps(fjx3,tx); |
2042 | fjy3 = _mm_add_ps(fjy3,ty); |
2043 | fjz3 = _mm_add_ps(fjz3,tz); |
2044 | |
2045 | fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch; |
2046 | fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch; |
2047 | fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch; |
2048 | fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch; |
2049 | |
2050 | gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD, |
2051 | fjx0,fjy0,fjz0,fjx1,fjy1,fjz1, |
2052 | fjx2,fjy2,fjz2,fjx3,fjy3,fjz3); |
2053 | |
2054 | /* Inner loop uses 295 flops */ |
2055 | } |
2056 | |
2057 | /* End of innermost loop */ |
2058 | |
2059 | gmx_mm_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3, |
2060 | f+i_coord_offset,fshift+i_shift_offset); |
2061 | |
2062 | /* Increment number of inner iterations */ |
2063 | inneriter += j_index_end - j_index_start; |
2064 | |
2065 | /* Outer loop uses 24 flops */ |
2066 | } |
2067 | |
2068 | /* Increment number of outer iterations */ |
2069 | outeriter += nri; |
2070 | |
2071 | /* Update outer/inner flops */ |
2072 | |
2073 | inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*295)(nrnb)->n[eNR_NBKERNEL_ELEC_VDW_W4W4_F] += outeriter*24 + inneriter *295; |
2074 | } |