File: | gromacs/gmxlib/nonbonded/nb_kernel_sse4_1_single/nb_kernel_ElecEw_VdwLJEw_GeomW3W3_sse4_1_single.c |
Location: | line 1419, column 5 |
Description: | Value stored to 'j_coord_offsetA' 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_ElecEw_VdwLJEw_GeomW3W3_VF_sse4_1_single |
54 | * Electrostatics interaction: Ewald |
55 | * VdW interaction: LJEwald |
56 | * Geometry: Water3-Water3 |
57 | * Calculate force/pot: PotentialAndForce |
58 | */ |
59 | void |
60 | nb_kernel_ElecEw_VdwLJEw_GeomW3W3_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 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 | __m128 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00; |
98 | __m128 dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01; |
99 | __m128 dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02; |
100 | __m128 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10; |
101 | __m128 dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11; |
102 | __m128 dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12; |
103 | __m128 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20; |
104 | __m128 dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21; |
105 | __m128 dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22; |
106 | __m128 velec,felec,velecsum,facel,crf,krf,krf2; |
107 | real *charge; |
108 | int nvdwtype; |
109 | __m128 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6; |
110 | int *vdwtype; |
111 | real *vdwparam; |
112 | __m128 one_sixth = _mm_set1_ps(1.0/6.0); |
113 | __m128 one_twelfth = _mm_set1_ps(1.0/12.0); |
114 | __m128 c6grid_00; |
115 | __m128 c6grid_01; |
116 | __m128 c6grid_02; |
117 | __m128 c6grid_10; |
118 | __m128 c6grid_11; |
119 | __m128 c6grid_12; |
120 | __m128 c6grid_20; |
121 | __m128 c6grid_21; |
122 | __m128 c6grid_22; |
123 | __m128 ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald; |
124 | real *vdwgridparam; |
125 | __m128 one_half = _mm_set1_ps(0.5); |
126 | __m128 minus_one = _mm_set1_ps(-1.0); |
127 | __m128i ewitab; |
128 | __m128 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV; |
129 | real *ewtab; |
130 | __m128 dummy_mask,cutoff_mask; |
131 | __m128 signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) ); |
132 | __m128 one = _mm_set1_ps(1.0); |
133 | __m128 two = _mm_set1_ps(2.0); |
134 | x = xx[0]; |
135 | f = ff[0]; |
136 | |
137 | nri = nlist->nri; |
138 | iinr = nlist->iinr; |
139 | jindex = nlist->jindex; |
140 | jjnr = nlist->jjnr; |
141 | shiftidx = nlist->shift; |
142 | gid = nlist->gid; |
143 | shiftvec = fr->shift_vec[0]; |
144 | fshift = fr->fshift[0]; |
145 | facel = _mm_set1_ps(fr->epsfac); |
146 | charge = mdatoms->chargeA; |
147 | nvdwtype = fr->ntype; |
148 | vdwparam = fr->nbfp; |
149 | vdwtype = mdatoms->typeA; |
150 | vdwgridparam = fr->ljpme_c6grid; |
151 | sh_lj_ewald = _mm_set1_ps(fr->ic->sh_lj_ewald); |
152 | ewclj = _mm_set1_ps(fr->ewaldcoeff_lj); |
153 | ewclj2 = _mm_mul_ps(minus_one,_mm_mul_ps(ewclj,ewclj)); |
154 | |
155 | sh_ewald = _mm_set1_ps(fr->ic->sh_ewald); |
156 | ewtab = fr->ic->tabq_coul_FDV0; |
157 | ewtabscale = _mm_set1_ps(fr->ic->tabq_scale); |
158 | ewtabhalfspace = _mm_set1_ps(0.5/fr->ic->tabq_scale); |
159 | |
160 | /* Setup water-specific parameters */ |
161 | inr = nlist->iinr[0]; |
162 | iq0 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+0])); |
163 | iq1 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1])); |
164 | iq2 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2])); |
165 | vdwioffset0 = 2*nvdwtype*vdwtype[inr+0]; |
166 | |
167 | jq0 = _mm_set1_ps(charge[inr+0]); |
168 | jq1 = _mm_set1_ps(charge[inr+1]); |
169 | jq2 = _mm_set1_ps(charge[inr+2]); |
170 | vdwjidx0A = 2*vdwtype[inr+0]; |
171 | qq00 = _mm_mul_ps(iq0,jq0); |
172 | c6_00 = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A]); |
173 | c12_00 = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A+1]); |
174 | c6grid_00 = _mm_set1_ps(vdwgridparam[vdwioffset0+vdwjidx0A]); |
175 | qq01 = _mm_mul_ps(iq0,jq1); |
176 | qq02 = _mm_mul_ps(iq0,jq2); |
177 | qq10 = _mm_mul_ps(iq1,jq0); |
178 | qq11 = _mm_mul_ps(iq1,jq1); |
179 | qq12 = _mm_mul_ps(iq1,jq2); |
180 | qq20 = _mm_mul_ps(iq2,jq0); |
181 | qq21 = _mm_mul_ps(iq2,jq1); |
182 | qq22 = _mm_mul_ps(iq2,jq2); |
183 | |
184 | /* Avoid stupid compiler warnings */ |
185 | jnrA = jnrB = jnrC = jnrD = 0; |
186 | j_coord_offsetA = 0; |
187 | j_coord_offsetB = 0; |
188 | j_coord_offsetC = 0; |
189 | j_coord_offsetD = 0; |
190 | |
191 | outeriter = 0; |
192 | inneriter = 0; |
193 | |
194 | for(iidx=0;iidx<4*DIM3;iidx++) |
195 | { |
196 | scratch[iidx] = 0.0; |
197 | } |
198 | |
199 | /* Start outer loop over neighborlists */ |
200 | for(iidx=0; iidx<nri; iidx++) |
201 | { |
202 | /* Load shift vector for this list */ |
203 | i_shift_offset = DIM3*shiftidx[iidx]; |
204 | |
205 | /* Load limits for loop over neighbors */ |
206 | j_index_start = jindex[iidx]; |
207 | j_index_end = jindex[iidx+1]; |
208 | |
209 | /* Get outer coordinate index */ |
210 | inr = iinr[iidx]; |
211 | i_coord_offset = DIM3*inr; |
212 | |
213 | /* Load i particle coords and add shift vector */ |
214 | gmx_mm_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset, |
215 | &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2); |
216 | |
217 | fix0 = _mm_setzero_ps(); |
218 | fiy0 = _mm_setzero_ps(); |
219 | fiz0 = _mm_setzero_ps(); |
220 | fix1 = _mm_setzero_ps(); |
221 | fiy1 = _mm_setzero_ps(); |
222 | fiz1 = _mm_setzero_ps(); |
223 | fix2 = _mm_setzero_ps(); |
224 | fiy2 = _mm_setzero_ps(); |
225 | fiz2 = _mm_setzero_ps(); |
226 | |
227 | /* Reset potential sums */ |
228 | velecsum = _mm_setzero_ps(); |
229 | vvdwsum = _mm_setzero_ps(); |
230 | |
231 | /* Start inner kernel loop */ |
232 | for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4) |
233 | { |
234 | |
235 | /* Get j neighbor index, and coordinate index */ |
236 | jnrA = jjnr[jidx]; |
237 | jnrB = jjnr[jidx+1]; |
238 | jnrC = jjnr[jidx+2]; |
239 | jnrD = jjnr[jidx+3]; |
240 | j_coord_offsetA = DIM3*jnrA; |
241 | j_coord_offsetB = DIM3*jnrB; |
242 | j_coord_offsetC = DIM3*jnrC; |
243 | j_coord_offsetD = DIM3*jnrD; |
244 | |
245 | /* load j atom coordinates */ |
246 | gmx_mm_load_3rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB, |
247 | x+j_coord_offsetC,x+j_coord_offsetD, |
248 | &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2); |
249 | |
250 | /* Calculate displacement vector */ |
251 | dx00 = _mm_sub_ps(ix0,jx0); |
252 | dy00 = _mm_sub_ps(iy0,jy0); |
253 | dz00 = _mm_sub_ps(iz0,jz0); |
254 | dx01 = _mm_sub_ps(ix0,jx1); |
255 | dy01 = _mm_sub_ps(iy0,jy1); |
256 | dz01 = _mm_sub_ps(iz0,jz1); |
257 | dx02 = _mm_sub_ps(ix0,jx2); |
258 | dy02 = _mm_sub_ps(iy0,jy2); |
259 | dz02 = _mm_sub_ps(iz0,jz2); |
260 | dx10 = _mm_sub_ps(ix1,jx0); |
261 | dy10 = _mm_sub_ps(iy1,jy0); |
262 | dz10 = _mm_sub_ps(iz1,jz0); |
263 | dx11 = _mm_sub_ps(ix1,jx1); |
264 | dy11 = _mm_sub_ps(iy1,jy1); |
265 | dz11 = _mm_sub_ps(iz1,jz1); |
266 | dx12 = _mm_sub_ps(ix1,jx2); |
267 | dy12 = _mm_sub_ps(iy1,jy2); |
268 | dz12 = _mm_sub_ps(iz1,jz2); |
269 | dx20 = _mm_sub_ps(ix2,jx0); |
270 | dy20 = _mm_sub_ps(iy2,jy0); |
271 | dz20 = _mm_sub_ps(iz2,jz0); |
272 | dx21 = _mm_sub_ps(ix2,jx1); |
273 | dy21 = _mm_sub_ps(iy2,jy1); |
274 | dz21 = _mm_sub_ps(iz2,jz1); |
275 | dx22 = _mm_sub_ps(ix2,jx2); |
276 | dy22 = _mm_sub_ps(iy2,jy2); |
277 | dz22 = _mm_sub_ps(iz2,jz2); |
278 | |
279 | /* Calculate squared distance and things based on it */ |
280 | rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00); |
281 | rsq01 = gmx_mm_calc_rsq_ps(dx01,dy01,dz01); |
282 | rsq02 = gmx_mm_calc_rsq_ps(dx02,dy02,dz02); |
283 | rsq10 = gmx_mm_calc_rsq_ps(dx10,dy10,dz10); |
284 | rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11); |
285 | rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12); |
286 | rsq20 = gmx_mm_calc_rsq_ps(dx20,dy20,dz20); |
287 | rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21); |
288 | rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22); |
289 | |
290 | rinv00 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq00); |
291 | rinv01 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq01); |
292 | rinv02 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq02); |
293 | rinv10 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq10); |
294 | rinv11 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq11); |
295 | rinv12 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq12); |
296 | rinv20 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq20); |
297 | rinv21 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq21); |
298 | rinv22 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq22); |
299 | |
300 | rinvsq00 = _mm_mul_ps(rinv00,rinv00); |
301 | rinvsq01 = _mm_mul_ps(rinv01,rinv01); |
302 | rinvsq02 = _mm_mul_ps(rinv02,rinv02); |
303 | rinvsq10 = _mm_mul_ps(rinv10,rinv10); |
304 | rinvsq11 = _mm_mul_ps(rinv11,rinv11); |
305 | rinvsq12 = _mm_mul_ps(rinv12,rinv12); |
306 | rinvsq20 = _mm_mul_ps(rinv20,rinv20); |
307 | rinvsq21 = _mm_mul_ps(rinv21,rinv21); |
308 | rinvsq22 = _mm_mul_ps(rinv22,rinv22); |
309 | |
310 | fjx0 = _mm_setzero_ps(); |
311 | fjy0 = _mm_setzero_ps(); |
312 | fjz0 = _mm_setzero_ps(); |
313 | fjx1 = _mm_setzero_ps(); |
314 | fjy1 = _mm_setzero_ps(); |
315 | fjz1 = _mm_setzero_ps(); |
316 | fjx2 = _mm_setzero_ps(); |
317 | fjy2 = _mm_setzero_ps(); |
318 | fjz2 = _mm_setzero_ps(); |
319 | |
320 | /************************** |
321 | * CALCULATE INTERACTIONS * |
322 | **************************/ |
323 | |
324 | r00 = _mm_mul_ps(rsq00,rinv00); |
325 | |
326 | /* EWALD ELECTROSTATICS */ |
327 | |
328 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
329 | ewrt = _mm_mul_ps(r00,ewtabscale); |
330 | ewitab = _mm_cvttps_epi32(ewrt); |
331 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
332 | ewitab = _mm_slli_epi32(ewitab,2); |
333 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
334 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
335 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
336 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
337 | _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn)do { __m128 tmp3, tmp2, tmp1, tmp0; tmp0 = _mm_unpacklo_ps((ewtabF ), (ewtabD)); tmp2 = _mm_unpacklo_ps((ewtabV), (ewtabFn)); tmp1 = _mm_unpackhi_ps((ewtabF), (ewtabD)); tmp3 = _mm_unpackhi_ps ((ewtabV), (ewtabFn)); (ewtabF) = _mm_movelh_ps(tmp0, tmp2); ( ewtabD) = _mm_movehl_ps(tmp2, tmp0); (ewtabV) = _mm_movelh_ps (tmp1, tmp3); (ewtabFn) = _mm_movehl_ps(tmp3, tmp1); } while ( 0); |
338 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
339 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
340 | velec = _mm_mul_ps(qq00,_mm_sub_ps(rinv00,velec)); |
341 | felec = _mm_mul_ps(_mm_mul_ps(qq00,rinv00),_mm_sub_ps(rinvsq00,felec)); |
342 | |
343 | /* Analytical LJ-PME */ |
344 | rinvsix = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00); |
345 | ewcljrsq = _mm_mul_ps(ewclj2,rsq00); |
346 | ewclj6 = _mm_mul_ps(ewclj2,_mm_mul_ps(ewclj2,ewclj2)); |
347 | exponent = gmx_simd_exp_rgmx_simd_exp_f(ewcljrsq); |
348 | /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */ |
349 | poly = _mm_mul_ps(exponent,_mm_add_ps(_mm_sub_ps(one,ewcljrsq),_mm_mul_ps(_mm_mul_ps(ewcljrsq,ewcljrsq),one_half))); |
350 | /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */ |
351 | vvdw6 = _mm_mul_ps(_mm_sub_ps(c6_00,_mm_mul_ps(c6grid_00,_mm_sub_ps(one,poly))),rinvsix); |
352 | vvdw12 = _mm_mul_ps(c12_00,_mm_mul_ps(rinvsix,rinvsix)); |
353 | vvdw = _mm_sub_ps(_mm_mul_ps(vvdw12,one_twelfth),_mm_mul_ps(vvdw6,one_sixth)); |
354 | /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */ |
355 | fvdw = _mm_mul_ps(_mm_sub_ps(vvdw12,_mm_sub_ps(vvdw6,_mm_mul_ps(_mm_mul_ps(c6grid_00,one_sixth),_mm_mul_ps(exponent,ewclj6)))),rinvsq00); |
356 | |
357 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
358 | velecsum = _mm_add_ps(velecsum,velec); |
359 | vvdwsum = _mm_add_ps(vvdwsum,vvdw); |
360 | |
361 | fscal = _mm_add_ps(felec,fvdw); |
362 | |
363 | /* Calculate temporary vectorial force */ |
364 | tx = _mm_mul_ps(fscal,dx00); |
365 | ty = _mm_mul_ps(fscal,dy00); |
366 | tz = _mm_mul_ps(fscal,dz00); |
367 | |
368 | /* Update vectorial force */ |
369 | fix0 = _mm_add_ps(fix0,tx); |
370 | fiy0 = _mm_add_ps(fiy0,ty); |
371 | fiz0 = _mm_add_ps(fiz0,tz); |
372 | |
373 | fjx0 = _mm_add_ps(fjx0,tx); |
374 | fjy0 = _mm_add_ps(fjy0,ty); |
375 | fjz0 = _mm_add_ps(fjz0,tz); |
376 | |
377 | /************************** |
378 | * CALCULATE INTERACTIONS * |
379 | **************************/ |
380 | |
381 | r01 = _mm_mul_ps(rsq01,rinv01); |
382 | |
383 | /* EWALD ELECTROSTATICS */ |
384 | |
385 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
386 | ewrt = _mm_mul_ps(r01,ewtabscale); |
387 | ewitab = _mm_cvttps_epi32(ewrt); |
388 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
389 | ewitab = _mm_slli_epi32(ewitab,2); |
390 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
391 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
392 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
393 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
394 | _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn)do { __m128 tmp3, tmp2, tmp1, tmp0; tmp0 = _mm_unpacklo_ps((ewtabF ), (ewtabD)); tmp2 = _mm_unpacklo_ps((ewtabV), (ewtabFn)); tmp1 = _mm_unpackhi_ps((ewtabF), (ewtabD)); tmp3 = _mm_unpackhi_ps ((ewtabV), (ewtabFn)); (ewtabF) = _mm_movelh_ps(tmp0, tmp2); ( ewtabD) = _mm_movehl_ps(tmp2, tmp0); (ewtabV) = _mm_movelh_ps (tmp1, tmp3); (ewtabFn) = _mm_movehl_ps(tmp3, tmp1); } while ( 0); |
395 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
396 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
397 | velec = _mm_mul_ps(qq01,_mm_sub_ps(rinv01,velec)); |
398 | felec = _mm_mul_ps(_mm_mul_ps(qq01,rinv01),_mm_sub_ps(rinvsq01,felec)); |
399 | |
400 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
401 | velecsum = _mm_add_ps(velecsum,velec); |
402 | |
403 | fscal = felec; |
404 | |
405 | /* Calculate temporary vectorial force */ |
406 | tx = _mm_mul_ps(fscal,dx01); |
407 | ty = _mm_mul_ps(fscal,dy01); |
408 | tz = _mm_mul_ps(fscal,dz01); |
409 | |
410 | /* Update vectorial force */ |
411 | fix0 = _mm_add_ps(fix0,tx); |
412 | fiy0 = _mm_add_ps(fiy0,ty); |
413 | fiz0 = _mm_add_ps(fiz0,tz); |
414 | |
415 | fjx1 = _mm_add_ps(fjx1,tx); |
416 | fjy1 = _mm_add_ps(fjy1,ty); |
417 | fjz1 = _mm_add_ps(fjz1,tz); |
418 | |
419 | /************************** |
420 | * CALCULATE INTERACTIONS * |
421 | **************************/ |
422 | |
423 | r02 = _mm_mul_ps(rsq02,rinv02); |
424 | |
425 | /* EWALD ELECTROSTATICS */ |
426 | |
427 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
428 | ewrt = _mm_mul_ps(r02,ewtabscale); |
429 | ewitab = _mm_cvttps_epi32(ewrt); |
430 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
431 | ewitab = _mm_slli_epi32(ewitab,2); |
432 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
433 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
434 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
435 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
436 | _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn)do { __m128 tmp3, tmp2, tmp1, tmp0; tmp0 = _mm_unpacklo_ps((ewtabF ), (ewtabD)); tmp2 = _mm_unpacklo_ps((ewtabV), (ewtabFn)); tmp1 = _mm_unpackhi_ps((ewtabF), (ewtabD)); tmp3 = _mm_unpackhi_ps ((ewtabV), (ewtabFn)); (ewtabF) = _mm_movelh_ps(tmp0, tmp2); ( ewtabD) = _mm_movehl_ps(tmp2, tmp0); (ewtabV) = _mm_movelh_ps (tmp1, tmp3); (ewtabFn) = _mm_movehl_ps(tmp3, tmp1); } while ( 0); |
437 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
438 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
439 | velec = _mm_mul_ps(qq02,_mm_sub_ps(rinv02,velec)); |
440 | felec = _mm_mul_ps(_mm_mul_ps(qq02,rinv02),_mm_sub_ps(rinvsq02,felec)); |
441 | |
442 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
443 | velecsum = _mm_add_ps(velecsum,velec); |
444 | |
445 | fscal = felec; |
446 | |
447 | /* Calculate temporary vectorial force */ |
448 | tx = _mm_mul_ps(fscal,dx02); |
449 | ty = _mm_mul_ps(fscal,dy02); |
450 | tz = _mm_mul_ps(fscal,dz02); |
451 | |
452 | /* Update vectorial force */ |
453 | fix0 = _mm_add_ps(fix0,tx); |
454 | fiy0 = _mm_add_ps(fiy0,ty); |
455 | fiz0 = _mm_add_ps(fiz0,tz); |
456 | |
457 | fjx2 = _mm_add_ps(fjx2,tx); |
458 | fjy2 = _mm_add_ps(fjy2,ty); |
459 | fjz2 = _mm_add_ps(fjz2,tz); |
460 | |
461 | /************************** |
462 | * CALCULATE INTERACTIONS * |
463 | **************************/ |
464 | |
465 | r10 = _mm_mul_ps(rsq10,rinv10); |
466 | |
467 | /* EWALD ELECTROSTATICS */ |
468 | |
469 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
470 | ewrt = _mm_mul_ps(r10,ewtabscale); |
471 | ewitab = _mm_cvttps_epi32(ewrt); |
472 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
473 | ewitab = _mm_slli_epi32(ewitab,2); |
474 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
475 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
476 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
477 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
478 | _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn)do { __m128 tmp3, tmp2, tmp1, tmp0; tmp0 = _mm_unpacklo_ps((ewtabF ), (ewtabD)); tmp2 = _mm_unpacklo_ps((ewtabV), (ewtabFn)); tmp1 = _mm_unpackhi_ps((ewtabF), (ewtabD)); tmp3 = _mm_unpackhi_ps ((ewtabV), (ewtabFn)); (ewtabF) = _mm_movelh_ps(tmp0, tmp2); ( ewtabD) = _mm_movehl_ps(tmp2, tmp0); (ewtabV) = _mm_movelh_ps (tmp1, tmp3); (ewtabFn) = _mm_movehl_ps(tmp3, tmp1); } while ( 0); |
479 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
480 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
481 | velec = _mm_mul_ps(qq10,_mm_sub_ps(rinv10,velec)); |
482 | felec = _mm_mul_ps(_mm_mul_ps(qq10,rinv10),_mm_sub_ps(rinvsq10,felec)); |
483 | |
484 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
485 | velecsum = _mm_add_ps(velecsum,velec); |
486 | |
487 | fscal = felec; |
488 | |
489 | /* Calculate temporary vectorial force */ |
490 | tx = _mm_mul_ps(fscal,dx10); |
491 | ty = _mm_mul_ps(fscal,dy10); |
492 | tz = _mm_mul_ps(fscal,dz10); |
493 | |
494 | /* Update vectorial force */ |
495 | fix1 = _mm_add_ps(fix1,tx); |
496 | fiy1 = _mm_add_ps(fiy1,ty); |
497 | fiz1 = _mm_add_ps(fiz1,tz); |
498 | |
499 | fjx0 = _mm_add_ps(fjx0,tx); |
500 | fjy0 = _mm_add_ps(fjy0,ty); |
501 | fjz0 = _mm_add_ps(fjz0,tz); |
502 | |
503 | /************************** |
504 | * CALCULATE INTERACTIONS * |
505 | **************************/ |
506 | |
507 | r11 = _mm_mul_ps(rsq11,rinv11); |
508 | |
509 | /* EWALD ELECTROSTATICS */ |
510 | |
511 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
512 | ewrt = _mm_mul_ps(r11,ewtabscale); |
513 | ewitab = _mm_cvttps_epi32(ewrt); |
514 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
515 | ewitab = _mm_slli_epi32(ewitab,2); |
516 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
517 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
518 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
519 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
520 | _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn)do { __m128 tmp3, tmp2, tmp1, tmp0; tmp0 = _mm_unpacklo_ps((ewtabF ), (ewtabD)); tmp2 = _mm_unpacklo_ps((ewtabV), (ewtabFn)); tmp1 = _mm_unpackhi_ps((ewtabF), (ewtabD)); tmp3 = _mm_unpackhi_ps ((ewtabV), (ewtabFn)); (ewtabF) = _mm_movelh_ps(tmp0, tmp2); ( ewtabD) = _mm_movehl_ps(tmp2, tmp0); (ewtabV) = _mm_movelh_ps (tmp1, tmp3); (ewtabFn) = _mm_movehl_ps(tmp3, tmp1); } while ( 0); |
521 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
522 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
523 | velec = _mm_mul_ps(qq11,_mm_sub_ps(rinv11,velec)); |
524 | felec = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec)); |
525 | |
526 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
527 | velecsum = _mm_add_ps(velecsum,velec); |
528 | |
529 | fscal = felec; |
530 | |
531 | /* Calculate temporary vectorial force */ |
532 | tx = _mm_mul_ps(fscal,dx11); |
533 | ty = _mm_mul_ps(fscal,dy11); |
534 | tz = _mm_mul_ps(fscal,dz11); |
535 | |
536 | /* Update vectorial force */ |
537 | fix1 = _mm_add_ps(fix1,tx); |
538 | fiy1 = _mm_add_ps(fiy1,ty); |
539 | fiz1 = _mm_add_ps(fiz1,tz); |
540 | |
541 | fjx1 = _mm_add_ps(fjx1,tx); |
542 | fjy1 = _mm_add_ps(fjy1,ty); |
543 | fjz1 = _mm_add_ps(fjz1,tz); |
544 | |
545 | /************************** |
546 | * CALCULATE INTERACTIONS * |
547 | **************************/ |
548 | |
549 | r12 = _mm_mul_ps(rsq12,rinv12); |
550 | |
551 | /* EWALD ELECTROSTATICS */ |
552 | |
553 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
554 | ewrt = _mm_mul_ps(r12,ewtabscale); |
555 | ewitab = _mm_cvttps_epi32(ewrt); |
556 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
557 | ewitab = _mm_slli_epi32(ewitab,2); |
558 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
559 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
560 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
561 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
562 | _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn)do { __m128 tmp3, tmp2, tmp1, tmp0; tmp0 = _mm_unpacklo_ps((ewtabF ), (ewtabD)); tmp2 = _mm_unpacklo_ps((ewtabV), (ewtabFn)); tmp1 = _mm_unpackhi_ps((ewtabF), (ewtabD)); tmp3 = _mm_unpackhi_ps ((ewtabV), (ewtabFn)); (ewtabF) = _mm_movelh_ps(tmp0, tmp2); ( ewtabD) = _mm_movehl_ps(tmp2, tmp0); (ewtabV) = _mm_movelh_ps (tmp1, tmp3); (ewtabFn) = _mm_movehl_ps(tmp3, tmp1); } while ( 0); |
563 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
564 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
565 | velec = _mm_mul_ps(qq12,_mm_sub_ps(rinv12,velec)); |
566 | felec = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec)); |
567 | |
568 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
569 | velecsum = _mm_add_ps(velecsum,velec); |
570 | |
571 | fscal = felec; |
572 | |
573 | /* Calculate temporary vectorial force */ |
574 | tx = _mm_mul_ps(fscal,dx12); |
575 | ty = _mm_mul_ps(fscal,dy12); |
576 | tz = _mm_mul_ps(fscal,dz12); |
577 | |
578 | /* Update vectorial force */ |
579 | fix1 = _mm_add_ps(fix1,tx); |
580 | fiy1 = _mm_add_ps(fiy1,ty); |
581 | fiz1 = _mm_add_ps(fiz1,tz); |
582 | |
583 | fjx2 = _mm_add_ps(fjx2,tx); |
584 | fjy2 = _mm_add_ps(fjy2,ty); |
585 | fjz2 = _mm_add_ps(fjz2,tz); |
586 | |
587 | /************************** |
588 | * CALCULATE INTERACTIONS * |
589 | **************************/ |
590 | |
591 | r20 = _mm_mul_ps(rsq20,rinv20); |
592 | |
593 | /* EWALD ELECTROSTATICS */ |
594 | |
595 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
596 | ewrt = _mm_mul_ps(r20,ewtabscale); |
597 | ewitab = _mm_cvttps_epi32(ewrt); |
598 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
599 | ewitab = _mm_slli_epi32(ewitab,2); |
600 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
601 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
602 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
603 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
604 | _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn)do { __m128 tmp3, tmp2, tmp1, tmp0; tmp0 = _mm_unpacklo_ps((ewtabF ), (ewtabD)); tmp2 = _mm_unpacklo_ps((ewtabV), (ewtabFn)); tmp1 = _mm_unpackhi_ps((ewtabF), (ewtabD)); tmp3 = _mm_unpackhi_ps ((ewtabV), (ewtabFn)); (ewtabF) = _mm_movelh_ps(tmp0, tmp2); ( ewtabD) = _mm_movehl_ps(tmp2, tmp0); (ewtabV) = _mm_movelh_ps (tmp1, tmp3); (ewtabFn) = _mm_movehl_ps(tmp3, tmp1); } while ( 0); |
605 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
606 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
607 | velec = _mm_mul_ps(qq20,_mm_sub_ps(rinv20,velec)); |
608 | felec = _mm_mul_ps(_mm_mul_ps(qq20,rinv20),_mm_sub_ps(rinvsq20,felec)); |
609 | |
610 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
611 | velecsum = _mm_add_ps(velecsum,velec); |
612 | |
613 | fscal = felec; |
614 | |
615 | /* Calculate temporary vectorial force */ |
616 | tx = _mm_mul_ps(fscal,dx20); |
617 | ty = _mm_mul_ps(fscal,dy20); |
618 | tz = _mm_mul_ps(fscal,dz20); |
619 | |
620 | /* Update vectorial force */ |
621 | fix2 = _mm_add_ps(fix2,tx); |
622 | fiy2 = _mm_add_ps(fiy2,ty); |
623 | fiz2 = _mm_add_ps(fiz2,tz); |
624 | |
625 | fjx0 = _mm_add_ps(fjx0,tx); |
626 | fjy0 = _mm_add_ps(fjy0,ty); |
627 | fjz0 = _mm_add_ps(fjz0,tz); |
628 | |
629 | /************************** |
630 | * CALCULATE INTERACTIONS * |
631 | **************************/ |
632 | |
633 | r21 = _mm_mul_ps(rsq21,rinv21); |
634 | |
635 | /* EWALD ELECTROSTATICS */ |
636 | |
637 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
638 | ewrt = _mm_mul_ps(r21,ewtabscale); |
639 | ewitab = _mm_cvttps_epi32(ewrt); |
640 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
641 | ewitab = _mm_slli_epi32(ewitab,2); |
642 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
643 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
644 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
645 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
646 | _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn)do { __m128 tmp3, tmp2, tmp1, tmp0; tmp0 = _mm_unpacklo_ps((ewtabF ), (ewtabD)); tmp2 = _mm_unpacklo_ps((ewtabV), (ewtabFn)); tmp1 = _mm_unpackhi_ps((ewtabF), (ewtabD)); tmp3 = _mm_unpackhi_ps ((ewtabV), (ewtabFn)); (ewtabF) = _mm_movelh_ps(tmp0, tmp2); ( ewtabD) = _mm_movehl_ps(tmp2, tmp0); (ewtabV) = _mm_movelh_ps (tmp1, tmp3); (ewtabFn) = _mm_movehl_ps(tmp3, tmp1); } while ( 0); |
647 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
648 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
649 | velec = _mm_mul_ps(qq21,_mm_sub_ps(rinv21,velec)); |
650 | felec = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec)); |
651 | |
652 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
653 | velecsum = _mm_add_ps(velecsum,velec); |
654 | |
655 | fscal = felec; |
656 | |
657 | /* Calculate temporary vectorial force */ |
658 | tx = _mm_mul_ps(fscal,dx21); |
659 | ty = _mm_mul_ps(fscal,dy21); |
660 | tz = _mm_mul_ps(fscal,dz21); |
661 | |
662 | /* Update vectorial force */ |
663 | fix2 = _mm_add_ps(fix2,tx); |
664 | fiy2 = _mm_add_ps(fiy2,ty); |
665 | fiz2 = _mm_add_ps(fiz2,tz); |
666 | |
667 | fjx1 = _mm_add_ps(fjx1,tx); |
668 | fjy1 = _mm_add_ps(fjy1,ty); |
669 | fjz1 = _mm_add_ps(fjz1,tz); |
670 | |
671 | /************************** |
672 | * CALCULATE INTERACTIONS * |
673 | **************************/ |
674 | |
675 | r22 = _mm_mul_ps(rsq22,rinv22); |
676 | |
677 | /* EWALD ELECTROSTATICS */ |
678 | |
679 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
680 | ewrt = _mm_mul_ps(r22,ewtabscale); |
681 | ewitab = _mm_cvttps_epi32(ewrt); |
682 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
683 | ewitab = _mm_slli_epi32(ewitab,2); |
684 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
685 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
686 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
687 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
688 | _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn)do { __m128 tmp3, tmp2, tmp1, tmp0; tmp0 = _mm_unpacklo_ps((ewtabF ), (ewtabD)); tmp2 = _mm_unpacklo_ps((ewtabV), (ewtabFn)); tmp1 = _mm_unpackhi_ps((ewtabF), (ewtabD)); tmp3 = _mm_unpackhi_ps ((ewtabV), (ewtabFn)); (ewtabF) = _mm_movelh_ps(tmp0, tmp2); ( ewtabD) = _mm_movehl_ps(tmp2, tmp0); (ewtabV) = _mm_movelh_ps (tmp1, tmp3); (ewtabFn) = _mm_movehl_ps(tmp3, tmp1); } while ( 0); |
689 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
690 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
691 | velec = _mm_mul_ps(qq22,_mm_sub_ps(rinv22,velec)); |
692 | felec = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec)); |
693 | |
694 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
695 | velecsum = _mm_add_ps(velecsum,velec); |
696 | |
697 | fscal = felec; |
698 | |
699 | /* Calculate temporary vectorial force */ |
700 | tx = _mm_mul_ps(fscal,dx22); |
701 | ty = _mm_mul_ps(fscal,dy22); |
702 | tz = _mm_mul_ps(fscal,dz22); |
703 | |
704 | /* Update vectorial force */ |
705 | fix2 = _mm_add_ps(fix2,tx); |
706 | fiy2 = _mm_add_ps(fiy2,ty); |
707 | fiz2 = _mm_add_ps(fiz2,tz); |
708 | |
709 | fjx2 = _mm_add_ps(fjx2,tx); |
710 | fjy2 = _mm_add_ps(fjy2,ty); |
711 | fjz2 = _mm_add_ps(fjz2,tz); |
712 | |
713 | fjptrA = f+j_coord_offsetA; |
714 | fjptrB = f+j_coord_offsetB; |
715 | fjptrC = f+j_coord_offsetC; |
716 | fjptrD = f+j_coord_offsetD; |
717 | |
718 | gmx_mm_decrement_3rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD, |
719 | fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2); |
720 | |
721 | /* Inner loop uses 397 flops */ |
722 | } |
723 | |
724 | if(jidx<j_index_end) |
725 | { |
726 | |
727 | /* Get j neighbor index, and coordinate index */ |
728 | jnrlistA = jjnr[jidx]; |
729 | jnrlistB = jjnr[jidx+1]; |
730 | jnrlistC = jjnr[jidx+2]; |
731 | jnrlistD = jjnr[jidx+3]; |
732 | /* Sign of each element will be negative for non-real atoms. |
733 | * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones, |
734 | * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries. |
735 | */ |
736 | dummy_mask = gmx_mm_castsi128_ps_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())); |
737 | jnrA = (jnrlistA>=0) ? jnrlistA : 0; |
738 | jnrB = (jnrlistB>=0) ? jnrlistB : 0; |
739 | jnrC = (jnrlistC>=0) ? jnrlistC : 0; |
740 | jnrD = (jnrlistD>=0) ? jnrlistD : 0; |
741 | j_coord_offsetA = DIM3*jnrA; |
742 | j_coord_offsetB = DIM3*jnrB; |
743 | j_coord_offsetC = DIM3*jnrC; |
744 | j_coord_offsetD = DIM3*jnrD; |
745 | |
746 | /* load j atom coordinates */ |
747 | gmx_mm_load_3rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB, |
748 | x+j_coord_offsetC,x+j_coord_offsetD, |
749 | &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2); |
750 | |
751 | /* Calculate displacement vector */ |
752 | dx00 = _mm_sub_ps(ix0,jx0); |
753 | dy00 = _mm_sub_ps(iy0,jy0); |
754 | dz00 = _mm_sub_ps(iz0,jz0); |
755 | dx01 = _mm_sub_ps(ix0,jx1); |
756 | dy01 = _mm_sub_ps(iy0,jy1); |
757 | dz01 = _mm_sub_ps(iz0,jz1); |
758 | dx02 = _mm_sub_ps(ix0,jx2); |
759 | dy02 = _mm_sub_ps(iy0,jy2); |
760 | dz02 = _mm_sub_ps(iz0,jz2); |
761 | dx10 = _mm_sub_ps(ix1,jx0); |
762 | dy10 = _mm_sub_ps(iy1,jy0); |
763 | dz10 = _mm_sub_ps(iz1,jz0); |
764 | dx11 = _mm_sub_ps(ix1,jx1); |
765 | dy11 = _mm_sub_ps(iy1,jy1); |
766 | dz11 = _mm_sub_ps(iz1,jz1); |
767 | dx12 = _mm_sub_ps(ix1,jx2); |
768 | dy12 = _mm_sub_ps(iy1,jy2); |
769 | dz12 = _mm_sub_ps(iz1,jz2); |
770 | dx20 = _mm_sub_ps(ix2,jx0); |
771 | dy20 = _mm_sub_ps(iy2,jy0); |
772 | dz20 = _mm_sub_ps(iz2,jz0); |
773 | dx21 = _mm_sub_ps(ix2,jx1); |
774 | dy21 = _mm_sub_ps(iy2,jy1); |
775 | dz21 = _mm_sub_ps(iz2,jz1); |
776 | dx22 = _mm_sub_ps(ix2,jx2); |
777 | dy22 = _mm_sub_ps(iy2,jy2); |
778 | dz22 = _mm_sub_ps(iz2,jz2); |
779 | |
780 | /* Calculate squared distance and things based on it */ |
781 | rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00); |
782 | rsq01 = gmx_mm_calc_rsq_ps(dx01,dy01,dz01); |
783 | rsq02 = gmx_mm_calc_rsq_ps(dx02,dy02,dz02); |
784 | rsq10 = gmx_mm_calc_rsq_ps(dx10,dy10,dz10); |
785 | rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11); |
786 | rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12); |
787 | rsq20 = gmx_mm_calc_rsq_ps(dx20,dy20,dz20); |
788 | rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21); |
789 | rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22); |
790 | |
791 | rinv00 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq00); |
792 | rinv01 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq01); |
793 | rinv02 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq02); |
794 | rinv10 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq10); |
795 | rinv11 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq11); |
796 | rinv12 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq12); |
797 | rinv20 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq20); |
798 | rinv21 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq21); |
799 | rinv22 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq22); |
800 | |
801 | rinvsq00 = _mm_mul_ps(rinv00,rinv00); |
802 | rinvsq01 = _mm_mul_ps(rinv01,rinv01); |
803 | rinvsq02 = _mm_mul_ps(rinv02,rinv02); |
804 | rinvsq10 = _mm_mul_ps(rinv10,rinv10); |
805 | rinvsq11 = _mm_mul_ps(rinv11,rinv11); |
806 | rinvsq12 = _mm_mul_ps(rinv12,rinv12); |
807 | rinvsq20 = _mm_mul_ps(rinv20,rinv20); |
808 | rinvsq21 = _mm_mul_ps(rinv21,rinv21); |
809 | rinvsq22 = _mm_mul_ps(rinv22,rinv22); |
810 | |
811 | fjx0 = _mm_setzero_ps(); |
812 | fjy0 = _mm_setzero_ps(); |
813 | fjz0 = _mm_setzero_ps(); |
814 | fjx1 = _mm_setzero_ps(); |
815 | fjy1 = _mm_setzero_ps(); |
816 | fjz1 = _mm_setzero_ps(); |
817 | fjx2 = _mm_setzero_ps(); |
818 | fjy2 = _mm_setzero_ps(); |
819 | fjz2 = _mm_setzero_ps(); |
820 | |
821 | /************************** |
822 | * CALCULATE INTERACTIONS * |
823 | **************************/ |
824 | |
825 | r00 = _mm_mul_ps(rsq00,rinv00); |
826 | r00 = _mm_andnot_ps(dummy_mask,r00); |
827 | |
828 | /* EWALD ELECTROSTATICS */ |
829 | |
830 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
831 | ewrt = _mm_mul_ps(r00,ewtabscale); |
832 | ewitab = _mm_cvttps_epi32(ewrt); |
833 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
834 | ewitab = _mm_slli_epi32(ewitab,2); |
835 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
836 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
837 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
838 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
839 | _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn)do { __m128 tmp3, tmp2, tmp1, tmp0; tmp0 = _mm_unpacklo_ps((ewtabF ), (ewtabD)); tmp2 = _mm_unpacklo_ps((ewtabV), (ewtabFn)); tmp1 = _mm_unpackhi_ps((ewtabF), (ewtabD)); tmp3 = _mm_unpackhi_ps ((ewtabV), (ewtabFn)); (ewtabF) = _mm_movelh_ps(tmp0, tmp2); ( ewtabD) = _mm_movehl_ps(tmp2, tmp0); (ewtabV) = _mm_movelh_ps (tmp1, tmp3); (ewtabFn) = _mm_movehl_ps(tmp3, tmp1); } while ( 0); |
840 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
841 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
842 | velec = _mm_mul_ps(qq00,_mm_sub_ps(rinv00,velec)); |
843 | felec = _mm_mul_ps(_mm_mul_ps(qq00,rinv00),_mm_sub_ps(rinvsq00,felec)); |
844 | |
845 | /* Analytical LJ-PME */ |
846 | rinvsix = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00); |
847 | ewcljrsq = _mm_mul_ps(ewclj2,rsq00); |
848 | ewclj6 = _mm_mul_ps(ewclj2,_mm_mul_ps(ewclj2,ewclj2)); |
849 | exponent = gmx_simd_exp_rgmx_simd_exp_f(ewcljrsq); |
850 | /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */ |
851 | poly = _mm_mul_ps(exponent,_mm_add_ps(_mm_sub_ps(one,ewcljrsq),_mm_mul_ps(_mm_mul_ps(ewcljrsq,ewcljrsq),one_half))); |
852 | /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */ |
853 | vvdw6 = _mm_mul_ps(_mm_sub_ps(c6_00,_mm_mul_ps(c6grid_00,_mm_sub_ps(one,poly))),rinvsix); |
854 | vvdw12 = _mm_mul_ps(c12_00,_mm_mul_ps(rinvsix,rinvsix)); |
855 | vvdw = _mm_sub_ps(_mm_mul_ps(vvdw12,one_twelfth),_mm_mul_ps(vvdw6,one_sixth)); |
856 | /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */ |
857 | fvdw = _mm_mul_ps(_mm_sub_ps(vvdw12,_mm_sub_ps(vvdw6,_mm_mul_ps(_mm_mul_ps(c6grid_00,one_sixth),_mm_mul_ps(exponent,ewclj6)))),rinvsq00); |
858 | |
859 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
860 | velec = _mm_andnot_ps(dummy_mask,velec); |
861 | velecsum = _mm_add_ps(velecsum,velec); |
862 | vvdw = _mm_andnot_ps(dummy_mask,vvdw); |
863 | vvdwsum = _mm_add_ps(vvdwsum,vvdw); |
864 | |
865 | fscal = _mm_add_ps(felec,fvdw); |
866 | |
867 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
868 | |
869 | /* Calculate temporary vectorial force */ |
870 | tx = _mm_mul_ps(fscal,dx00); |
871 | ty = _mm_mul_ps(fscal,dy00); |
872 | tz = _mm_mul_ps(fscal,dz00); |
873 | |
874 | /* Update vectorial force */ |
875 | fix0 = _mm_add_ps(fix0,tx); |
876 | fiy0 = _mm_add_ps(fiy0,ty); |
877 | fiz0 = _mm_add_ps(fiz0,tz); |
878 | |
879 | fjx0 = _mm_add_ps(fjx0,tx); |
880 | fjy0 = _mm_add_ps(fjy0,ty); |
881 | fjz0 = _mm_add_ps(fjz0,tz); |
882 | |
883 | /************************** |
884 | * CALCULATE INTERACTIONS * |
885 | **************************/ |
886 | |
887 | r01 = _mm_mul_ps(rsq01,rinv01); |
888 | r01 = _mm_andnot_ps(dummy_mask,r01); |
889 | |
890 | /* EWALD ELECTROSTATICS */ |
891 | |
892 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
893 | ewrt = _mm_mul_ps(r01,ewtabscale); |
894 | ewitab = _mm_cvttps_epi32(ewrt); |
895 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
896 | ewitab = _mm_slli_epi32(ewitab,2); |
897 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
898 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
899 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
900 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
901 | _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn)do { __m128 tmp3, tmp2, tmp1, tmp0; tmp0 = _mm_unpacklo_ps((ewtabF ), (ewtabD)); tmp2 = _mm_unpacklo_ps((ewtabV), (ewtabFn)); tmp1 = _mm_unpackhi_ps((ewtabF), (ewtabD)); tmp3 = _mm_unpackhi_ps ((ewtabV), (ewtabFn)); (ewtabF) = _mm_movelh_ps(tmp0, tmp2); ( ewtabD) = _mm_movehl_ps(tmp2, tmp0); (ewtabV) = _mm_movelh_ps (tmp1, tmp3); (ewtabFn) = _mm_movehl_ps(tmp3, tmp1); } while ( 0); |
902 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
903 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
904 | velec = _mm_mul_ps(qq01,_mm_sub_ps(rinv01,velec)); |
905 | felec = _mm_mul_ps(_mm_mul_ps(qq01,rinv01),_mm_sub_ps(rinvsq01,felec)); |
906 | |
907 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
908 | velec = _mm_andnot_ps(dummy_mask,velec); |
909 | velecsum = _mm_add_ps(velecsum,velec); |
910 | |
911 | fscal = felec; |
912 | |
913 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
914 | |
915 | /* Calculate temporary vectorial force */ |
916 | tx = _mm_mul_ps(fscal,dx01); |
917 | ty = _mm_mul_ps(fscal,dy01); |
918 | tz = _mm_mul_ps(fscal,dz01); |
919 | |
920 | /* Update vectorial force */ |
921 | fix0 = _mm_add_ps(fix0,tx); |
922 | fiy0 = _mm_add_ps(fiy0,ty); |
923 | fiz0 = _mm_add_ps(fiz0,tz); |
924 | |
925 | fjx1 = _mm_add_ps(fjx1,tx); |
926 | fjy1 = _mm_add_ps(fjy1,ty); |
927 | fjz1 = _mm_add_ps(fjz1,tz); |
928 | |
929 | /************************** |
930 | * CALCULATE INTERACTIONS * |
931 | **************************/ |
932 | |
933 | r02 = _mm_mul_ps(rsq02,rinv02); |
934 | r02 = _mm_andnot_ps(dummy_mask,r02); |
935 | |
936 | /* EWALD ELECTROSTATICS */ |
937 | |
938 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
939 | ewrt = _mm_mul_ps(r02,ewtabscale); |
940 | ewitab = _mm_cvttps_epi32(ewrt); |
941 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
942 | ewitab = _mm_slli_epi32(ewitab,2); |
943 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
944 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
945 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
946 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
947 | _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn)do { __m128 tmp3, tmp2, tmp1, tmp0; tmp0 = _mm_unpacklo_ps((ewtabF ), (ewtabD)); tmp2 = _mm_unpacklo_ps((ewtabV), (ewtabFn)); tmp1 = _mm_unpackhi_ps((ewtabF), (ewtabD)); tmp3 = _mm_unpackhi_ps ((ewtabV), (ewtabFn)); (ewtabF) = _mm_movelh_ps(tmp0, tmp2); ( ewtabD) = _mm_movehl_ps(tmp2, tmp0); (ewtabV) = _mm_movelh_ps (tmp1, tmp3); (ewtabFn) = _mm_movehl_ps(tmp3, tmp1); } while ( 0); |
948 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
949 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
950 | velec = _mm_mul_ps(qq02,_mm_sub_ps(rinv02,velec)); |
951 | felec = _mm_mul_ps(_mm_mul_ps(qq02,rinv02),_mm_sub_ps(rinvsq02,felec)); |
952 | |
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); |
956 | |
957 | fscal = felec; |
958 | |
959 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
960 | |
961 | /* Calculate temporary vectorial force */ |
962 | tx = _mm_mul_ps(fscal,dx02); |
963 | ty = _mm_mul_ps(fscal,dy02); |
964 | tz = _mm_mul_ps(fscal,dz02); |
965 | |
966 | /* Update vectorial force */ |
967 | fix0 = _mm_add_ps(fix0,tx); |
968 | fiy0 = _mm_add_ps(fiy0,ty); |
969 | fiz0 = _mm_add_ps(fiz0,tz); |
970 | |
971 | fjx2 = _mm_add_ps(fjx2,tx); |
972 | fjy2 = _mm_add_ps(fjy2,ty); |
973 | fjz2 = _mm_add_ps(fjz2,tz); |
974 | |
975 | /************************** |
976 | * CALCULATE INTERACTIONS * |
977 | **************************/ |
978 | |
979 | r10 = _mm_mul_ps(rsq10,rinv10); |
980 | r10 = _mm_andnot_ps(dummy_mask,r10); |
981 | |
982 | /* EWALD ELECTROSTATICS */ |
983 | |
984 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
985 | ewrt = _mm_mul_ps(r10,ewtabscale); |
986 | ewitab = _mm_cvttps_epi32(ewrt); |
987 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
988 | ewitab = _mm_slli_epi32(ewitab,2); |
989 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
990 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
991 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
992 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
993 | _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn)do { __m128 tmp3, tmp2, tmp1, tmp0; tmp0 = _mm_unpacklo_ps((ewtabF ), (ewtabD)); tmp2 = _mm_unpacklo_ps((ewtabV), (ewtabFn)); tmp1 = _mm_unpackhi_ps((ewtabF), (ewtabD)); tmp3 = _mm_unpackhi_ps ((ewtabV), (ewtabFn)); (ewtabF) = _mm_movelh_ps(tmp0, tmp2); ( ewtabD) = _mm_movehl_ps(tmp2, tmp0); (ewtabV) = _mm_movelh_ps (tmp1, tmp3); (ewtabFn) = _mm_movehl_ps(tmp3, tmp1); } while ( 0); |
994 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
995 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
996 | velec = _mm_mul_ps(qq10,_mm_sub_ps(rinv10,velec)); |
997 | felec = _mm_mul_ps(_mm_mul_ps(qq10,rinv10),_mm_sub_ps(rinvsq10,felec)); |
998 | |
999 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
1000 | velec = _mm_andnot_ps(dummy_mask,velec); |
1001 | velecsum = _mm_add_ps(velecsum,velec); |
1002 | |
1003 | fscal = felec; |
1004 | |
1005 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
1006 | |
1007 | /* Calculate temporary vectorial force */ |
1008 | tx = _mm_mul_ps(fscal,dx10); |
1009 | ty = _mm_mul_ps(fscal,dy10); |
1010 | tz = _mm_mul_ps(fscal,dz10); |
1011 | |
1012 | /* Update vectorial force */ |
1013 | fix1 = _mm_add_ps(fix1,tx); |
1014 | fiy1 = _mm_add_ps(fiy1,ty); |
1015 | fiz1 = _mm_add_ps(fiz1,tz); |
1016 | |
1017 | fjx0 = _mm_add_ps(fjx0,tx); |
1018 | fjy0 = _mm_add_ps(fjy0,ty); |
1019 | fjz0 = _mm_add_ps(fjz0,tz); |
1020 | |
1021 | /************************** |
1022 | * CALCULATE INTERACTIONS * |
1023 | **************************/ |
1024 | |
1025 | r11 = _mm_mul_ps(rsq11,rinv11); |
1026 | r11 = _mm_andnot_ps(dummy_mask,r11); |
1027 | |
1028 | /* EWALD ELECTROSTATICS */ |
1029 | |
1030 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1031 | ewrt = _mm_mul_ps(r11,ewtabscale); |
1032 | ewitab = _mm_cvttps_epi32(ewrt); |
1033 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1034 | ewitab = _mm_slli_epi32(ewitab,2); |
1035 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
1036 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
1037 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
1038 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
1039 | _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn)do { __m128 tmp3, tmp2, tmp1, tmp0; tmp0 = _mm_unpacklo_ps((ewtabF ), (ewtabD)); tmp2 = _mm_unpacklo_ps((ewtabV), (ewtabFn)); tmp1 = _mm_unpackhi_ps((ewtabF), (ewtabD)); tmp3 = _mm_unpackhi_ps ((ewtabV), (ewtabFn)); (ewtabF) = _mm_movelh_ps(tmp0, tmp2); ( ewtabD) = _mm_movehl_ps(tmp2, tmp0); (ewtabV) = _mm_movelh_ps (tmp1, tmp3); (ewtabFn) = _mm_movehl_ps(tmp3, tmp1); } while ( 0); |
1040 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
1041 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
1042 | velec = _mm_mul_ps(qq11,_mm_sub_ps(rinv11,velec)); |
1043 | felec = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec)); |
1044 | |
1045 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
1046 | velec = _mm_andnot_ps(dummy_mask,velec); |
1047 | velecsum = _mm_add_ps(velecsum,velec); |
1048 | |
1049 | fscal = felec; |
1050 | |
1051 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
1052 | |
1053 | /* Calculate temporary vectorial force */ |
1054 | tx = _mm_mul_ps(fscal,dx11); |
1055 | ty = _mm_mul_ps(fscal,dy11); |
1056 | tz = _mm_mul_ps(fscal,dz11); |
1057 | |
1058 | /* Update vectorial force */ |
1059 | fix1 = _mm_add_ps(fix1,tx); |
1060 | fiy1 = _mm_add_ps(fiy1,ty); |
1061 | fiz1 = _mm_add_ps(fiz1,tz); |
1062 | |
1063 | fjx1 = _mm_add_ps(fjx1,tx); |
1064 | fjy1 = _mm_add_ps(fjy1,ty); |
1065 | fjz1 = _mm_add_ps(fjz1,tz); |
1066 | |
1067 | /************************** |
1068 | * CALCULATE INTERACTIONS * |
1069 | **************************/ |
1070 | |
1071 | r12 = _mm_mul_ps(rsq12,rinv12); |
1072 | r12 = _mm_andnot_ps(dummy_mask,r12); |
1073 | |
1074 | /* EWALD ELECTROSTATICS */ |
1075 | |
1076 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1077 | ewrt = _mm_mul_ps(r12,ewtabscale); |
1078 | ewitab = _mm_cvttps_epi32(ewrt); |
1079 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1080 | ewitab = _mm_slli_epi32(ewitab,2); |
1081 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
1082 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
1083 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
1084 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
1085 | _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn)do { __m128 tmp3, tmp2, tmp1, tmp0; tmp0 = _mm_unpacklo_ps((ewtabF ), (ewtabD)); tmp2 = _mm_unpacklo_ps((ewtabV), (ewtabFn)); tmp1 = _mm_unpackhi_ps((ewtabF), (ewtabD)); tmp3 = _mm_unpackhi_ps ((ewtabV), (ewtabFn)); (ewtabF) = _mm_movelh_ps(tmp0, tmp2); ( ewtabD) = _mm_movehl_ps(tmp2, tmp0); (ewtabV) = _mm_movelh_ps (tmp1, tmp3); (ewtabFn) = _mm_movehl_ps(tmp3, tmp1); } while ( 0); |
1086 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
1087 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
1088 | velec = _mm_mul_ps(qq12,_mm_sub_ps(rinv12,velec)); |
1089 | felec = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec)); |
1090 | |
1091 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
1092 | velec = _mm_andnot_ps(dummy_mask,velec); |
1093 | velecsum = _mm_add_ps(velecsum,velec); |
1094 | |
1095 | fscal = felec; |
1096 | |
1097 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
1098 | |
1099 | /* Calculate temporary vectorial force */ |
1100 | tx = _mm_mul_ps(fscal,dx12); |
1101 | ty = _mm_mul_ps(fscal,dy12); |
1102 | tz = _mm_mul_ps(fscal,dz12); |
1103 | |
1104 | /* Update vectorial force */ |
1105 | fix1 = _mm_add_ps(fix1,tx); |
1106 | fiy1 = _mm_add_ps(fiy1,ty); |
1107 | fiz1 = _mm_add_ps(fiz1,tz); |
1108 | |
1109 | fjx2 = _mm_add_ps(fjx2,tx); |
1110 | fjy2 = _mm_add_ps(fjy2,ty); |
1111 | fjz2 = _mm_add_ps(fjz2,tz); |
1112 | |
1113 | /************************** |
1114 | * CALCULATE INTERACTIONS * |
1115 | **************************/ |
1116 | |
1117 | r20 = _mm_mul_ps(rsq20,rinv20); |
1118 | r20 = _mm_andnot_ps(dummy_mask,r20); |
1119 | |
1120 | /* EWALD ELECTROSTATICS */ |
1121 | |
1122 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1123 | ewrt = _mm_mul_ps(r20,ewtabscale); |
1124 | ewitab = _mm_cvttps_epi32(ewrt); |
1125 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1126 | ewitab = _mm_slli_epi32(ewitab,2); |
1127 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
1128 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
1129 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
1130 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
1131 | _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn)do { __m128 tmp3, tmp2, tmp1, tmp0; tmp0 = _mm_unpacklo_ps((ewtabF ), (ewtabD)); tmp2 = _mm_unpacklo_ps((ewtabV), (ewtabFn)); tmp1 = _mm_unpackhi_ps((ewtabF), (ewtabD)); tmp3 = _mm_unpackhi_ps ((ewtabV), (ewtabFn)); (ewtabF) = _mm_movelh_ps(tmp0, tmp2); ( ewtabD) = _mm_movehl_ps(tmp2, tmp0); (ewtabV) = _mm_movelh_ps (tmp1, tmp3); (ewtabFn) = _mm_movehl_ps(tmp3, tmp1); } while ( 0); |
1132 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
1133 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
1134 | velec = _mm_mul_ps(qq20,_mm_sub_ps(rinv20,velec)); |
1135 | felec = _mm_mul_ps(_mm_mul_ps(qq20,rinv20),_mm_sub_ps(rinvsq20,felec)); |
1136 | |
1137 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
1138 | velec = _mm_andnot_ps(dummy_mask,velec); |
1139 | velecsum = _mm_add_ps(velecsum,velec); |
1140 | |
1141 | fscal = felec; |
1142 | |
1143 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
1144 | |
1145 | /* Calculate temporary vectorial force */ |
1146 | tx = _mm_mul_ps(fscal,dx20); |
1147 | ty = _mm_mul_ps(fscal,dy20); |
1148 | tz = _mm_mul_ps(fscal,dz20); |
1149 | |
1150 | /* Update vectorial force */ |
1151 | fix2 = _mm_add_ps(fix2,tx); |
1152 | fiy2 = _mm_add_ps(fiy2,ty); |
1153 | fiz2 = _mm_add_ps(fiz2,tz); |
1154 | |
1155 | fjx0 = _mm_add_ps(fjx0,tx); |
1156 | fjy0 = _mm_add_ps(fjy0,ty); |
1157 | fjz0 = _mm_add_ps(fjz0,tz); |
1158 | |
1159 | /************************** |
1160 | * CALCULATE INTERACTIONS * |
1161 | **************************/ |
1162 | |
1163 | r21 = _mm_mul_ps(rsq21,rinv21); |
1164 | r21 = _mm_andnot_ps(dummy_mask,r21); |
1165 | |
1166 | /* EWALD ELECTROSTATICS */ |
1167 | |
1168 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1169 | ewrt = _mm_mul_ps(r21,ewtabscale); |
1170 | ewitab = _mm_cvttps_epi32(ewrt); |
1171 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1172 | ewitab = _mm_slli_epi32(ewitab,2); |
1173 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
1174 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
1175 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
1176 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
1177 | _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn)do { __m128 tmp3, tmp2, tmp1, tmp0; tmp0 = _mm_unpacklo_ps((ewtabF ), (ewtabD)); tmp2 = _mm_unpacklo_ps((ewtabV), (ewtabFn)); tmp1 = _mm_unpackhi_ps((ewtabF), (ewtabD)); tmp3 = _mm_unpackhi_ps ((ewtabV), (ewtabFn)); (ewtabF) = _mm_movelh_ps(tmp0, tmp2); ( ewtabD) = _mm_movehl_ps(tmp2, tmp0); (ewtabV) = _mm_movelh_ps (tmp1, tmp3); (ewtabFn) = _mm_movehl_ps(tmp3, tmp1); } while ( 0); |
1178 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
1179 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
1180 | velec = _mm_mul_ps(qq21,_mm_sub_ps(rinv21,velec)); |
1181 | felec = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec)); |
1182 | |
1183 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
1184 | velec = _mm_andnot_ps(dummy_mask,velec); |
1185 | velecsum = _mm_add_ps(velecsum,velec); |
1186 | |
1187 | fscal = felec; |
1188 | |
1189 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
1190 | |
1191 | /* Calculate temporary vectorial force */ |
1192 | tx = _mm_mul_ps(fscal,dx21); |
1193 | ty = _mm_mul_ps(fscal,dy21); |
1194 | tz = _mm_mul_ps(fscal,dz21); |
1195 | |
1196 | /* Update vectorial force */ |
1197 | fix2 = _mm_add_ps(fix2,tx); |
1198 | fiy2 = _mm_add_ps(fiy2,ty); |
1199 | fiz2 = _mm_add_ps(fiz2,tz); |
1200 | |
1201 | fjx1 = _mm_add_ps(fjx1,tx); |
1202 | fjy1 = _mm_add_ps(fjy1,ty); |
1203 | fjz1 = _mm_add_ps(fjz1,tz); |
1204 | |
1205 | /************************** |
1206 | * CALCULATE INTERACTIONS * |
1207 | **************************/ |
1208 | |
1209 | r22 = _mm_mul_ps(rsq22,rinv22); |
1210 | r22 = _mm_andnot_ps(dummy_mask,r22); |
1211 | |
1212 | /* EWALD ELECTROSTATICS */ |
1213 | |
1214 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1215 | ewrt = _mm_mul_ps(r22,ewtabscale); |
1216 | ewitab = _mm_cvttps_epi32(ewrt); |
1217 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1218 | ewitab = _mm_slli_epi32(ewitab,2); |
1219 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
1220 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
1221 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
1222 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
1223 | _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn)do { __m128 tmp3, tmp2, tmp1, tmp0; tmp0 = _mm_unpacklo_ps((ewtabF ), (ewtabD)); tmp2 = _mm_unpacklo_ps((ewtabV), (ewtabFn)); tmp1 = _mm_unpackhi_ps((ewtabF), (ewtabD)); tmp3 = _mm_unpackhi_ps ((ewtabV), (ewtabFn)); (ewtabF) = _mm_movelh_ps(tmp0, tmp2); ( ewtabD) = _mm_movehl_ps(tmp2, tmp0); (ewtabV) = _mm_movelh_ps (tmp1, tmp3); (ewtabFn) = _mm_movehl_ps(tmp3, tmp1); } while ( 0); |
1224 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
1225 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
1226 | velec = _mm_mul_ps(qq22,_mm_sub_ps(rinv22,velec)); |
1227 | felec = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec)); |
1228 | |
1229 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
1230 | velec = _mm_andnot_ps(dummy_mask,velec); |
1231 | velecsum = _mm_add_ps(velecsum,velec); |
1232 | |
1233 | fscal = felec; |
1234 | |
1235 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
1236 | |
1237 | /* Calculate temporary vectorial force */ |
1238 | tx = _mm_mul_ps(fscal,dx22); |
1239 | ty = _mm_mul_ps(fscal,dy22); |
1240 | tz = _mm_mul_ps(fscal,dz22); |
1241 | |
1242 | /* Update vectorial force */ |
1243 | fix2 = _mm_add_ps(fix2,tx); |
1244 | fiy2 = _mm_add_ps(fiy2,ty); |
1245 | fiz2 = _mm_add_ps(fiz2,tz); |
1246 | |
1247 | fjx2 = _mm_add_ps(fjx2,tx); |
1248 | fjy2 = _mm_add_ps(fjy2,ty); |
1249 | fjz2 = _mm_add_ps(fjz2,tz); |
1250 | |
1251 | fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch; |
1252 | fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch; |
1253 | fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch; |
1254 | fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch; |
1255 | |
1256 | gmx_mm_decrement_3rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD, |
1257 | fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2); |
1258 | |
1259 | /* Inner loop uses 406 flops */ |
1260 | } |
1261 | |
1262 | /* End of innermost loop */ |
1263 | |
1264 | gmx_mm_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2, |
1265 | f+i_coord_offset,fshift+i_shift_offset); |
1266 | |
1267 | ggid = gid[iidx]; |
1268 | /* Update potential energies */ |
1269 | gmx_mm_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid); |
1270 | gmx_mm_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid); |
1271 | |
1272 | /* Increment number of inner iterations */ |
1273 | inneriter += j_index_end - j_index_start; |
1274 | |
1275 | /* Outer loop uses 20 flops */ |
1276 | } |
1277 | |
1278 | /* Increment number of outer iterations */ |
1279 | outeriter += nri; |
1280 | |
1281 | /* Update outer/inner flops */ |
1282 | |
1283 | inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_VF,outeriter*20 + inneriter*406)(nrnb)->n[eNR_NBKERNEL_ELEC_VDW_W3W3_VF] += outeriter*20 + inneriter*406; |
1284 | } |
1285 | /* |
1286 | * Gromacs nonbonded kernel: nb_kernel_ElecEw_VdwLJEw_GeomW3W3_F_sse4_1_single |
1287 | * Electrostatics interaction: Ewald |
1288 | * VdW interaction: LJEwald |
1289 | * Geometry: Water3-Water3 |
1290 | * Calculate force/pot: Force |
1291 | */ |
1292 | void |
1293 | nb_kernel_ElecEw_VdwLJEw_GeomW3W3_F_sse4_1_single |
1294 | (t_nblist * gmx_restrict nlist, |
1295 | rvec * gmx_restrict xx, |
1296 | rvec * gmx_restrict ff, |
1297 | t_forcerec * gmx_restrict fr, |
1298 | t_mdatoms * gmx_restrict mdatoms, |
1299 | nb_kernel_data_t gmx_unused__attribute__ ((unused)) * gmx_restrict kernel_data, |
1300 | t_nrnb * gmx_restrict nrnb) |
1301 | { |
1302 | /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or |
1303 | * just 0 for non-waters. |
1304 | * Suffixes A,B,C,D refer to j loop unrolling done with SSE, e.g. for the four different |
1305 | * jnr indices corresponding to data put in the four positions in the SIMD register. |
1306 | */ |
1307 | int i_shift_offset,i_coord_offset,outeriter,inneriter; |
1308 | int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx; |
1309 | int jnrA,jnrB,jnrC,jnrD; |
1310 | int jnrlistA,jnrlistB,jnrlistC,jnrlistD; |
1311 | int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD; |
1312 | int *iinr,*jindex,*jjnr,*shiftidx,*gid; |
1313 | real rcutoff_scalar; |
1314 | real *shiftvec,*fshift,*x,*f; |
1315 | real *fjptrA,*fjptrB,*fjptrC,*fjptrD; |
1316 | real scratch[4*DIM3]; |
1317 | __m128 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall; |
1318 | int vdwioffset0; |
1319 | __m128 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0; |
1320 | int vdwioffset1; |
1321 | __m128 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1; |
1322 | int vdwioffset2; |
1323 | __m128 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2; |
1324 | int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D; |
1325 | __m128 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0; |
1326 | int vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D; |
1327 | __m128 jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1; |
1328 | int vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D; |
1329 | __m128 jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2; |
1330 | __m128 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00; |
1331 | __m128 dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01; |
1332 | __m128 dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02; |
1333 | __m128 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10; |
1334 | __m128 dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11; |
1335 | __m128 dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12; |
1336 | __m128 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20; |
1337 | __m128 dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21; |
1338 | __m128 dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22; |
1339 | __m128 velec,felec,velecsum,facel,crf,krf,krf2; |
1340 | real *charge; |
1341 | int nvdwtype; |
1342 | __m128 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6; |
1343 | int *vdwtype; |
1344 | real *vdwparam; |
1345 | __m128 one_sixth = _mm_set1_ps(1.0/6.0); |
1346 | __m128 one_twelfth = _mm_set1_ps(1.0/12.0); |
1347 | __m128 c6grid_00; |
1348 | __m128 c6grid_01; |
1349 | __m128 c6grid_02; |
1350 | __m128 c6grid_10; |
1351 | __m128 c6grid_11; |
1352 | __m128 c6grid_12; |
1353 | __m128 c6grid_20; |
1354 | __m128 c6grid_21; |
1355 | __m128 c6grid_22; |
1356 | __m128 ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald; |
1357 | real *vdwgridparam; |
1358 | __m128 one_half = _mm_set1_ps(0.5); |
1359 | __m128 minus_one = _mm_set1_ps(-1.0); |
1360 | __m128i ewitab; |
1361 | __m128 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV; |
1362 | real *ewtab; |
1363 | __m128 dummy_mask,cutoff_mask; |
1364 | __m128 signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) ); |
1365 | __m128 one = _mm_set1_ps(1.0); |
1366 | __m128 two = _mm_set1_ps(2.0); |
1367 | x = xx[0]; |
1368 | f = ff[0]; |
1369 | |
1370 | nri = nlist->nri; |
1371 | iinr = nlist->iinr; |
1372 | jindex = nlist->jindex; |
1373 | jjnr = nlist->jjnr; |
1374 | shiftidx = nlist->shift; |
1375 | gid = nlist->gid; |
1376 | shiftvec = fr->shift_vec[0]; |
1377 | fshift = fr->fshift[0]; |
1378 | facel = _mm_set1_ps(fr->epsfac); |
1379 | charge = mdatoms->chargeA; |
1380 | nvdwtype = fr->ntype; |
1381 | vdwparam = fr->nbfp; |
1382 | vdwtype = mdatoms->typeA; |
1383 | vdwgridparam = fr->ljpme_c6grid; |
1384 | sh_lj_ewald = _mm_set1_ps(fr->ic->sh_lj_ewald); |
1385 | ewclj = _mm_set1_ps(fr->ewaldcoeff_lj); |
1386 | ewclj2 = _mm_mul_ps(minus_one,_mm_mul_ps(ewclj,ewclj)); |
1387 | |
1388 | sh_ewald = _mm_set1_ps(fr->ic->sh_ewald); |
1389 | ewtab = fr->ic->tabq_coul_F; |
1390 | ewtabscale = _mm_set1_ps(fr->ic->tabq_scale); |
1391 | ewtabhalfspace = _mm_set1_ps(0.5/fr->ic->tabq_scale); |
1392 | |
1393 | /* Setup water-specific parameters */ |
1394 | inr = nlist->iinr[0]; |
1395 | iq0 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+0])); |
1396 | iq1 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1])); |
1397 | iq2 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2])); |
1398 | vdwioffset0 = 2*nvdwtype*vdwtype[inr+0]; |
1399 | |
1400 | jq0 = _mm_set1_ps(charge[inr+0]); |
1401 | jq1 = _mm_set1_ps(charge[inr+1]); |
1402 | jq2 = _mm_set1_ps(charge[inr+2]); |
1403 | vdwjidx0A = 2*vdwtype[inr+0]; |
1404 | qq00 = _mm_mul_ps(iq0,jq0); |
1405 | c6_00 = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A]); |
1406 | c12_00 = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A+1]); |
1407 | c6grid_00 = _mm_set1_ps(vdwgridparam[vdwioffset0+vdwjidx0A]); |
1408 | qq01 = _mm_mul_ps(iq0,jq1); |
1409 | qq02 = _mm_mul_ps(iq0,jq2); |
1410 | qq10 = _mm_mul_ps(iq1,jq0); |
1411 | qq11 = _mm_mul_ps(iq1,jq1); |
1412 | qq12 = _mm_mul_ps(iq1,jq2); |
1413 | qq20 = _mm_mul_ps(iq2,jq0); |
1414 | qq21 = _mm_mul_ps(iq2,jq1); |
1415 | qq22 = _mm_mul_ps(iq2,jq2); |
1416 | |
1417 | /* Avoid stupid compiler warnings */ |
1418 | jnrA = jnrB = jnrC = jnrD = 0; |
1419 | j_coord_offsetA = 0; |
Value stored to 'j_coord_offsetA' is never read | |
1420 | j_coord_offsetB = 0; |
1421 | j_coord_offsetC = 0; |
1422 | j_coord_offsetD = 0; |
1423 | |
1424 | outeriter = 0; |
1425 | inneriter = 0; |
1426 | |
1427 | for(iidx=0;iidx<4*DIM3;iidx++) |
1428 | { |
1429 | scratch[iidx] = 0.0; |
1430 | } |
1431 | |
1432 | /* Start outer loop over neighborlists */ |
1433 | for(iidx=0; iidx<nri; iidx++) |
1434 | { |
1435 | /* Load shift vector for this list */ |
1436 | i_shift_offset = DIM3*shiftidx[iidx]; |
1437 | |
1438 | /* Load limits for loop over neighbors */ |
1439 | j_index_start = jindex[iidx]; |
1440 | j_index_end = jindex[iidx+1]; |
1441 | |
1442 | /* Get outer coordinate index */ |
1443 | inr = iinr[iidx]; |
1444 | i_coord_offset = DIM3*inr; |
1445 | |
1446 | /* Load i particle coords and add shift vector */ |
1447 | gmx_mm_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset, |
1448 | &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2); |
1449 | |
1450 | fix0 = _mm_setzero_ps(); |
1451 | fiy0 = _mm_setzero_ps(); |
1452 | fiz0 = _mm_setzero_ps(); |
1453 | fix1 = _mm_setzero_ps(); |
1454 | fiy1 = _mm_setzero_ps(); |
1455 | fiz1 = _mm_setzero_ps(); |
1456 | fix2 = _mm_setzero_ps(); |
1457 | fiy2 = _mm_setzero_ps(); |
1458 | fiz2 = _mm_setzero_ps(); |
1459 | |
1460 | /* Start inner kernel loop */ |
1461 | for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4) |
1462 | { |
1463 | |
1464 | /* Get j neighbor index, and coordinate index */ |
1465 | jnrA = jjnr[jidx]; |
1466 | jnrB = jjnr[jidx+1]; |
1467 | jnrC = jjnr[jidx+2]; |
1468 | jnrD = jjnr[jidx+3]; |
1469 | j_coord_offsetA = DIM3*jnrA; |
1470 | j_coord_offsetB = DIM3*jnrB; |
1471 | j_coord_offsetC = DIM3*jnrC; |
1472 | j_coord_offsetD = DIM3*jnrD; |
1473 | |
1474 | /* load j atom coordinates */ |
1475 | gmx_mm_load_3rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB, |
1476 | x+j_coord_offsetC,x+j_coord_offsetD, |
1477 | &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2); |
1478 | |
1479 | /* Calculate displacement vector */ |
1480 | dx00 = _mm_sub_ps(ix0,jx0); |
1481 | dy00 = _mm_sub_ps(iy0,jy0); |
1482 | dz00 = _mm_sub_ps(iz0,jz0); |
1483 | dx01 = _mm_sub_ps(ix0,jx1); |
1484 | dy01 = _mm_sub_ps(iy0,jy1); |
1485 | dz01 = _mm_sub_ps(iz0,jz1); |
1486 | dx02 = _mm_sub_ps(ix0,jx2); |
1487 | dy02 = _mm_sub_ps(iy0,jy2); |
1488 | dz02 = _mm_sub_ps(iz0,jz2); |
1489 | dx10 = _mm_sub_ps(ix1,jx0); |
1490 | dy10 = _mm_sub_ps(iy1,jy0); |
1491 | dz10 = _mm_sub_ps(iz1,jz0); |
1492 | dx11 = _mm_sub_ps(ix1,jx1); |
1493 | dy11 = _mm_sub_ps(iy1,jy1); |
1494 | dz11 = _mm_sub_ps(iz1,jz1); |
1495 | dx12 = _mm_sub_ps(ix1,jx2); |
1496 | dy12 = _mm_sub_ps(iy1,jy2); |
1497 | dz12 = _mm_sub_ps(iz1,jz2); |
1498 | dx20 = _mm_sub_ps(ix2,jx0); |
1499 | dy20 = _mm_sub_ps(iy2,jy0); |
1500 | dz20 = _mm_sub_ps(iz2,jz0); |
1501 | dx21 = _mm_sub_ps(ix2,jx1); |
1502 | dy21 = _mm_sub_ps(iy2,jy1); |
1503 | dz21 = _mm_sub_ps(iz2,jz1); |
1504 | dx22 = _mm_sub_ps(ix2,jx2); |
1505 | dy22 = _mm_sub_ps(iy2,jy2); |
1506 | dz22 = _mm_sub_ps(iz2,jz2); |
1507 | |
1508 | /* Calculate squared distance and things based on it */ |
1509 | rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00); |
1510 | rsq01 = gmx_mm_calc_rsq_ps(dx01,dy01,dz01); |
1511 | rsq02 = gmx_mm_calc_rsq_ps(dx02,dy02,dz02); |
1512 | rsq10 = gmx_mm_calc_rsq_ps(dx10,dy10,dz10); |
1513 | rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11); |
1514 | rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12); |
1515 | rsq20 = gmx_mm_calc_rsq_ps(dx20,dy20,dz20); |
1516 | rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21); |
1517 | rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22); |
1518 | |
1519 | rinv00 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq00); |
1520 | rinv01 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq01); |
1521 | rinv02 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq02); |
1522 | rinv10 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq10); |
1523 | rinv11 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq11); |
1524 | rinv12 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq12); |
1525 | rinv20 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq20); |
1526 | rinv21 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq21); |
1527 | rinv22 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq22); |
1528 | |
1529 | rinvsq00 = _mm_mul_ps(rinv00,rinv00); |
1530 | rinvsq01 = _mm_mul_ps(rinv01,rinv01); |
1531 | rinvsq02 = _mm_mul_ps(rinv02,rinv02); |
1532 | rinvsq10 = _mm_mul_ps(rinv10,rinv10); |
1533 | rinvsq11 = _mm_mul_ps(rinv11,rinv11); |
1534 | rinvsq12 = _mm_mul_ps(rinv12,rinv12); |
1535 | rinvsq20 = _mm_mul_ps(rinv20,rinv20); |
1536 | rinvsq21 = _mm_mul_ps(rinv21,rinv21); |
1537 | rinvsq22 = _mm_mul_ps(rinv22,rinv22); |
1538 | |
1539 | fjx0 = _mm_setzero_ps(); |
1540 | fjy0 = _mm_setzero_ps(); |
1541 | fjz0 = _mm_setzero_ps(); |
1542 | fjx1 = _mm_setzero_ps(); |
1543 | fjy1 = _mm_setzero_ps(); |
1544 | fjz1 = _mm_setzero_ps(); |
1545 | fjx2 = _mm_setzero_ps(); |
1546 | fjy2 = _mm_setzero_ps(); |
1547 | fjz2 = _mm_setzero_ps(); |
1548 | |
1549 | /************************** |
1550 | * CALCULATE INTERACTIONS * |
1551 | **************************/ |
1552 | |
1553 | r00 = _mm_mul_ps(rsq00,rinv00); |
1554 | |
1555 | /* EWALD ELECTROSTATICS */ |
1556 | |
1557 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1558 | ewrt = _mm_mul_ps(r00,ewtabscale); |
1559 | ewitab = _mm_cvttps_epi32(ewrt); |
1560 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1561 | gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})), |
1562 | ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})), |
1563 | &ewtabF,&ewtabFn); |
1564 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
1565 | felec = _mm_mul_ps(_mm_mul_ps(qq00,rinv00),_mm_sub_ps(rinvsq00,felec)); |
1566 | |
1567 | /* Analytical LJ-PME */ |
1568 | rinvsix = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00); |
1569 | ewcljrsq = _mm_mul_ps(ewclj2,rsq00); |
1570 | ewclj6 = _mm_mul_ps(ewclj2,_mm_mul_ps(ewclj2,ewclj2)); |
1571 | exponent = gmx_simd_exp_rgmx_simd_exp_f(ewcljrsq); |
1572 | /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */ |
1573 | poly = _mm_mul_ps(exponent,_mm_add_ps(_mm_sub_ps(one,ewcljrsq),_mm_mul_ps(_mm_mul_ps(ewcljrsq,ewcljrsq),one_half))); |
1574 | /* f6A = 6 * C6grid * (1 - poly) */ |
1575 | f6A = _mm_mul_ps(c6grid_00,_mm_sub_ps(one,poly)); |
1576 | /* f6B = C6grid * exponent * beta^6 */ |
1577 | f6B = _mm_mul_ps(_mm_mul_ps(c6grid_00,one_sixth),_mm_mul_ps(exponent,ewclj6)); |
1578 | /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */ |
1579 | fvdw = _mm_mul_ps(_mm_add_ps(_mm_mul_ps(_mm_sub_ps(_mm_mul_ps(c12_00,rinvsix),_mm_sub_ps(c6_00,f6A)),rinvsix),f6B),rinvsq00); |
1580 | |
1581 | fscal = _mm_add_ps(felec,fvdw); |
1582 | |
1583 | /* Calculate temporary vectorial force */ |
1584 | tx = _mm_mul_ps(fscal,dx00); |
1585 | ty = _mm_mul_ps(fscal,dy00); |
1586 | tz = _mm_mul_ps(fscal,dz00); |
1587 | |
1588 | /* Update vectorial force */ |
1589 | fix0 = _mm_add_ps(fix0,tx); |
1590 | fiy0 = _mm_add_ps(fiy0,ty); |
1591 | fiz0 = _mm_add_ps(fiz0,tz); |
1592 | |
1593 | fjx0 = _mm_add_ps(fjx0,tx); |
1594 | fjy0 = _mm_add_ps(fjy0,ty); |
1595 | fjz0 = _mm_add_ps(fjz0,tz); |
1596 | |
1597 | /************************** |
1598 | * CALCULATE INTERACTIONS * |
1599 | **************************/ |
1600 | |
1601 | r01 = _mm_mul_ps(rsq01,rinv01); |
1602 | |
1603 | /* EWALD ELECTROSTATICS */ |
1604 | |
1605 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1606 | ewrt = _mm_mul_ps(r01,ewtabscale); |
1607 | ewitab = _mm_cvttps_epi32(ewrt); |
1608 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1609 | gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})), |
1610 | ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})), |
1611 | &ewtabF,&ewtabFn); |
1612 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
1613 | felec = _mm_mul_ps(_mm_mul_ps(qq01,rinv01),_mm_sub_ps(rinvsq01,felec)); |
1614 | |
1615 | fscal = felec; |
1616 | |
1617 | /* Calculate temporary vectorial force */ |
1618 | tx = _mm_mul_ps(fscal,dx01); |
1619 | ty = _mm_mul_ps(fscal,dy01); |
1620 | tz = _mm_mul_ps(fscal,dz01); |
1621 | |
1622 | /* Update vectorial force */ |
1623 | fix0 = _mm_add_ps(fix0,tx); |
1624 | fiy0 = _mm_add_ps(fiy0,ty); |
1625 | fiz0 = _mm_add_ps(fiz0,tz); |
1626 | |
1627 | fjx1 = _mm_add_ps(fjx1,tx); |
1628 | fjy1 = _mm_add_ps(fjy1,ty); |
1629 | fjz1 = _mm_add_ps(fjz1,tz); |
1630 | |
1631 | /************************** |
1632 | * CALCULATE INTERACTIONS * |
1633 | **************************/ |
1634 | |
1635 | r02 = _mm_mul_ps(rsq02,rinv02); |
1636 | |
1637 | /* EWALD ELECTROSTATICS */ |
1638 | |
1639 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1640 | ewrt = _mm_mul_ps(r02,ewtabscale); |
1641 | ewitab = _mm_cvttps_epi32(ewrt); |
1642 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1643 | gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})), |
1644 | ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})), |
1645 | &ewtabF,&ewtabFn); |
1646 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
1647 | felec = _mm_mul_ps(_mm_mul_ps(qq02,rinv02),_mm_sub_ps(rinvsq02,felec)); |
1648 | |
1649 | fscal = felec; |
1650 | |
1651 | /* Calculate temporary vectorial force */ |
1652 | tx = _mm_mul_ps(fscal,dx02); |
1653 | ty = _mm_mul_ps(fscal,dy02); |
1654 | tz = _mm_mul_ps(fscal,dz02); |
1655 | |
1656 | /* Update vectorial force */ |
1657 | fix0 = _mm_add_ps(fix0,tx); |
1658 | fiy0 = _mm_add_ps(fiy0,ty); |
1659 | fiz0 = _mm_add_ps(fiz0,tz); |
1660 | |
1661 | fjx2 = _mm_add_ps(fjx2,tx); |
1662 | fjy2 = _mm_add_ps(fjy2,ty); |
1663 | fjz2 = _mm_add_ps(fjz2,tz); |
1664 | |
1665 | /************************** |
1666 | * CALCULATE INTERACTIONS * |
1667 | **************************/ |
1668 | |
1669 | r10 = _mm_mul_ps(rsq10,rinv10); |
1670 | |
1671 | /* EWALD ELECTROSTATICS */ |
1672 | |
1673 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1674 | ewrt = _mm_mul_ps(r10,ewtabscale); |
1675 | ewitab = _mm_cvttps_epi32(ewrt); |
1676 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1677 | gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})), |
1678 | ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})), |
1679 | &ewtabF,&ewtabFn); |
1680 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
1681 | felec = _mm_mul_ps(_mm_mul_ps(qq10,rinv10),_mm_sub_ps(rinvsq10,felec)); |
1682 | |
1683 | fscal = felec; |
1684 | |
1685 | /* Calculate temporary vectorial force */ |
1686 | tx = _mm_mul_ps(fscal,dx10); |
1687 | ty = _mm_mul_ps(fscal,dy10); |
1688 | tz = _mm_mul_ps(fscal,dz10); |
1689 | |
1690 | /* Update vectorial force */ |
1691 | fix1 = _mm_add_ps(fix1,tx); |
1692 | fiy1 = _mm_add_ps(fiy1,ty); |
1693 | fiz1 = _mm_add_ps(fiz1,tz); |
1694 | |
1695 | fjx0 = _mm_add_ps(fjx0,tx); |
1696 | fjy0 = _mm_add_ps(fjy0,ty); |
1697 | fjz0 = _mm_add_ps(fjz0,tz); |
1698 | |
1699 | /************************** |
1700 | * CALCULATE INTERACTIONS * |
1701 | **************************/ |
1702 | |
1703 | r11 = _mm_mul_ps(rsq11,rinv11); |
1704 | |
1705 | /* EWALD ELECTROSTATICS */ |
1706 | |
1707 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1708 | ewrt = _mm_mul_ps(r11,ewtabscale); |
1709 | ewitab = _mm_cvttps_epi32(ewrt); |
1710 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1711 | gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})), |
1712 | ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})), |
1713 | &ewtabF,&ewtabFn); |
1714 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
1715 | felec = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec)); |
1716 | |
1717 | fscal = felec; |
1718 | |
1719 | /* Calculate temporary vectorial force */ |
1720 | tx = _mm_mul_ps(fscal,dx11); |
1721 | ty = _mm_mul_ps(fscal,dy11); |
1722 | tz = _mm_mul_ps(fscal,dz11); |
1723 | |
1724 | /* Update vectorial force */ |
1725 | fix1 = _mm_add_ps(fix1,tx); |
1726 | fiy1 = _mm_add_ps(fiy1,ty); |
1727 | fiz1 = _mm_add_ps(fiz1,tz); |
1728 | |
1729 | fjx1 = _mm_add_ps(fjx1,tx); |
1730 | fjy1 = _mm_add_ps(fjy1,ty); |
1731 | fjz1 = _mm_add_ps(fjz1,tz); |
1732 | |
1733 | /************************** |
1734 | * CALCULATE INTERACTIONS * |
1735 | **************************/ |
1736 | |
1737 | r12 = _mm_mul_ps(rsq12,rinv12); |
1738 | |
1739 | /* EWALD ELECTROSTATICS */ |
1740 | |
1741 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1742 | ewrt = _mm_mul_ps(r12,ewtabscale); |
1743 | ewitab = _mm_cvttps_epi32(ewrt); |
1744 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1745 | gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})), |
1746 | ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})), |
1747 | &ewtabF,&ewtabFn); |
1748 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
1749 | felec = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec)); |
1750 | |
1751 | fscal = felec; |
1752 | |
1753 | /* Calculate temporary vectorial force */ |
1754 | tx = _mm_mul_ps(fscal,dx12); |
1755 | ty = _mm_mul_ps(fscal,dy12); |
1756 | tz = _mm_mul_ps(fscal,dz12); |
1757 | |
1758 | /* Update vectorial force */ |
1759 | fix1 = _mm_add_ps(fix1,tx); |
1760 | fiy1 = _mm_add_ps(fiy1,ty); |
1761 | fiz1 = _mm_add_ps(fiz1,tz); |
1762 | |
1763 | fjx2 = _mm_add_ps(fjx2,tx); |
1764 | fjy2 = _mm_add_ps(fjy2,ty); |
1765 | fjz2 = _mm_add_ps(fjz2,tz); |
1766 | |
1767 | /************************** |
1768 | * CALCULATE INTERACTIONS * |
1769 | **************************/ |
1770 | |
1771 | r20 = _mm_mul_ps(rsq20,rinv20); |
1772 | |
1773 | /* EWALD ELECTROSTATICS */ |
1774 | |
1775 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1776 | ewrt = _mm_mul_ps(r20,ewtabscale); |
1777 | ewitab = _mm_cvttps_epi32(ewrt); |
1778 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1779 | gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})), |
1780 | ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})), |
1781 | &ewtabF,&ewtabFn); |
1782 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
1783 | felec = _mm_mul_ps(_mm_mul_ps(qq20,rinv20),_mm_sub_ps(rinvsq20,felec)); |
1784 | |
1785 | fscal = felec; |
1786 | |
1787 | /* Calculate temporary vectorial force */ |
1788 | tx = _mm_mul_ps(fscal,dx20); |
1789 | ty = _mm_mul_ps(fscal,dy20); |
1790 | tz = _mm_mul_ps(fscal,dz20); |
1791 | |
1792 | /* Update vectorial force */ |
1793 | fix2 = _mm_add_ps(fix2,tx); |
1794 | fiy2 = _mm_add_ps(fiy2,ty); |
1795 | fiz2 = _mm_add_ps(fiz2,tz); |
1796 | |
1797 | fjx0 = _mm_add_ps(fjx0,tx); |
1798 | fjy0 = _mm_add_ps(fjy0,ty); |
1799 | fjz0 = _mm_add_ps(fjz0,tz); |
1800 | |
1801 | /************************** |
1802 | * CALCULATE INTERACTIONS * |
1803 | **************************/ |
1804 | |
1805 | r21 = _mm_mul_ps(rsq21,rinv21); |
1806 | |
1807 | /* EWALD ELECTROSTATICS */ |
1808 | |
1809 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1810 | ewrt = _mm_mul_ps(r21,ewtabscale); |
1811 | ewitab = _mm_cvttps_epi32(ewrt); |
1812 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1813 | gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})), |
1814 | ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})), |
1815 | &ewtabF,&ewtabFn); |
1816 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
1817 | felec = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec)); |
1818 | |
1819 | fscal = felec; |
1820 | |
1821 | /* Calculate temporary vectorial force */ |
1822 | tx = _mm_mul_ps(fscal,dx21); |
1823 | ty = _mm_mul_ps(fscal,dy21); |
1824 | tz = _mm_mul_ps(fscal,dz21); |
1825 | |
1826 | /* Update vectorial force */ |
1827 | fix2 = _mm_add_ps(fix2,tx); |
1828 | fiy2 = _mm_add_ps(fiy2,ty); |
1829 | fiz2 = _mm_add_ps(fiz2,tz); |
1830 | |
1831 | fjx1 = _mm_add_ps(fjx1,tx); |
1832 | fjy1 = _mm_add_ps(fjy1,ty); |
1833 | fjz1 = _mm_add_ps(fjz1,tz); |
1834 | |
1835 | /************************** |
1836 | * CALCULATE INTERACTIONS * |
1837 | **************************/ |
1838 | |
1839 | r22 = _mm_mul_ps(rsq22,rinv22); |
1840 | |
1841 | /* EWALD ELECTROSTATICS */ |
1842 | |
1843 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1844 | ewrt = _mm_mul_ps(r22,ewtabscale); |
1845 | ewitab = _mm_cvttps_epi32(ewrt); |
1846 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1847 | gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})), |
1848 | ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})), |
1849 | &ewtabF,&ewtabFn); |
1850 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
1851 | felec = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec)); |
1852 | |
1853 | fscal = felec; |
1854 | |
1855 | /* Calculate temporary vectorial force */ |
1856 | tx = _mm_mul_ps(fscal,dx22); |
1857 | ty = _mm_mul_ps(fscal,dy22); |
1858 | tz = _mm_mul_ps(fscal,dz22); |
1859 | |
1860 | /* Update vectorial force */ |
1861 | fix2 = _mm_add_ps(fix2,tx); |
1862 | fiy2 = _mm_add_ps(fiy2,ty); |
1863 | fiz2 = _mm_add_ps(fiz2,tz); |
1864 | |
1865 | fjx2 = _mm_add_ps(fjx2,tx); |
1866 | fjy2 = _mm_add_ps(fjy2,ty); |
1867 | fjz2 = _mm_add_ps(fjz2,tz); |
1868 | |
1869 | fjptrA = f+j_coord_offsetA; |
1870 | fjptrB = f+j_coord_offsetB; |
1871 | fjptrC = f+j_coord_offsetC; |
1872 | fjptrD = f+j_coord_offsetD; |
1873 | |
1874 | gmx_mm_decrement_3rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD, |
1875 | fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2); |
1876 | |
1877 | /* Inner loop uses 347 flops */ |
1878 | } |
1879 | |
1880 | if(jidx<j_index_end) |
1881 | { |
1882 | |
1883 | /* Get j neighbor index, and coordinate index */ |
1884 | jnrlistA = jjnr[jidx]; |
1885 | jnrlistB = jjnr[jidx+1]; |
1886 | jnrlistC = jjnr[jidx+2]; |
1887 | jnrlistD = jjnr[jidx+3]; |
1888 | /* Sign of each element will be negative for non-real atoms. |
1889 | * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones, |
1890 | * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries. |
1891 | */ |
1892 | dummy_mask = gmx_mm_castsi128_ps_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())); |
1893 | jnrA = (jnrlistA>=0) ? jnrlistA : 0; |
1894 | jnrB = (jnrlistB>=0) ? jnrlistB : 0; |
1895 | jnrC = (jnrlistC>=0) ? jnrlistC : 0; |
1896 | jnrD = (jnrlistD>=0) ? jnrlistD : 0; |
1897 | j_coord_offsetA = DIM3*jnrA; |
1898 | j_coord_offsetB = DIM3*jnrB; |
1899 | j_coord_offsetC = DIM3*jnrC; |
1900 | j_coord_offsetD = DIM3*jnrD; |
1901 | |
1902 | /* load j atom coordinates */ |
1903 | gmx_mm_load_3rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB, |
1904 | x+j_coord_offsetC,x+j_coord_offsetD, |
1905 | &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2); |
1906 | |
1907 | /* Calculate displacement vector */ |
1908 | dx00 = _mm_sub_ps(ix0,jx0); |
1909 | dy00 = _mm_sub_ps(iy0,jy0); |
1910 | dz00 = _mm_sub_ps(iz0,jz0); |
1911 | dx01 = _mm_sub_ps(ix0,jx1); |
1912 | dy01 = _mm_sub_ps(iy0,jy1); |
1913 | dz01 = _mm_sub_ps(iz0,jz1); |
1914 | dx02 = _mm_sub_ps(ix0,jx2); |
1915 | dy02 = _mm_sub_ps(iy0,jy2); |
1916 | dz02 = _mm_sub_ps(iz0,jz2); |
1917 | dx10 = _mm_sub_ps(ix1,jx0); |
1918 | dy10 = _mm_sub_ps(iy1,jy0); |
1919 | dz10 = _mm_sub_ps(iz1,jz0); |
1920 | dx11 = _mm_sub_ps(ix1,jx1); |
1921 | dy11 = _mm_sub_ps(iy1,jy1); |
1922 | dz11 = _mm_sub_ps(iz1,jz1); |
1923 | dx12 = _mm_sub_ps(ix1,jx2); |
1924 | dy12 = _mm_sub_ps(iy1,jy2); |
1925 | dz12 = _mm_sub_ps(iz1,jz2); |
1926 | dx20 = _mm_sub_ps(ix2,jx0); |
1927 | dy20 = _mm_sub_ps(iy2,jy0); |
1928 | dz20 = _mm_sub_ps(iz2,jz0); |
1929 | dx21 = _mm_sub_ps(ix2,jx1); |
1930 | dy21 = _mm_sub_ps(iy2,jy1); |
1931 | dz21 = _mm_sub_ps(iz2,jz1); |
1932 | dx22 = _mm_sub_ps(ix2,jx2); |
1933 | dy22 = _mm_sub_ps(iy2,jy2); |
1934 | dz22 = _mm_sub_ps(iz2,jz2); |
1935 | |
1936 | /* Calculate squared distance and things based on it */ |
1937 | rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00); |
1938 | rsq01 = gmx_mm_calc_rsq_ps(dx01,dy01,dz01); |
1939 | rsq02 = gmx_mm_calc_rsq_ps(dx02,dy02,dz02); |
1940 | rsq10 = gmx_mm_calc_rsq_ps(dx10,dy10,dz10); |
1941 | rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11); |
1942 | rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12); |
1943 | rsq20 = gmx_mm_calc_rsq_ps(dx20,dy20,dz20); |
1944 | rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21); |
1945 | rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22); |
1946 | |
1947 | rinv00 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq00); |
1948 | rinv01 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq01); |
1949 | rinv02 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq02); |
1950 | rinv10 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq10); |
1951 | rinv11 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq11); |
1952 | rinv12 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq12); |
1953 | rinv20 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq20); |
1954 | rinv21 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq21); |
1955 | rinv22 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq22); |
1956 | |
1957 | rinvsq00 = _mm_mul_ps(rinv00,rinv00); |
1958 | rinvsq01 = _mm_mul_ps(rinv01,rinv01); |
1959 | rinvsq02 = _mm_mul_ps(rinv02,rinv02); |
1960 | rinvsq10 = _mm_mul_ps(rinv10,rinv10); |
1961 | rinvsq11 = _mm_mul_ps(rinv11,rinv11); |
1962 | rinvsq12 = _mm_mul_ps(rinv12,rinv12); |
1963 | rinvsq20 = _mm_mul_ps(rinv20,rinv20); |
1964 | rinvsq21 = _mm_mul_ps(rinv21,rinv21); |
1965 | rinvsq22 = _mm_mul_ps(rinv22,rinv22); |
1966 | |
1967 | fjx0 = _mm_setzero_ps(); |
1968 | fjy0 = _mm_setzero_ps(); |
1969 | fjz0 = _mm_setzero_ps(); |
1970 | fjx1 = _mm_setzero_ps(); |
1971 | fjy1 = _mm_setzero_ps(); |
1972 | fjz1 = _mm_setzero_ps(); |
1973 | fjx2 = _mm_setzero_ps(); |
1974 | fjy2 = _mm_setzero_ps(); |
1975 | fjz2 = _mm_setzero_ps(); |
1976 | |
1977 | /************************** |
1978 | * CALCULATE INTERACTIONS * |
1979 | **************************/ |
1980 | |
1981 | r00 = _mm_mul_ps(rsq00,rinv00); |
1982 | r00 = _mm_andnot_ps(dummy_mask,r00); |
1983 | |
1984 | /* EWALD ELECTROSTATICS */ |
1985 | |
1986 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1987 | ewrt = _mm_mul_ps(r00,ewtabscale); |
1988 | ewitab = _mm_cvttps_epi32(ewrt); |
1989 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1990 | gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})), |
1991 | ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})), |
1992 | &ewtabF,&ewtabFn); |
1993 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
1994 | felec = _mm_mul_ps(_mm_mul_ps(qq00,rinv00),_mm_sub_ps(rinvsq00,felec)); |
1995 | |
1996 | /* Analytical LJ-PME */ |
1997 | rinvsix = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00); |
1998 | ewcljrsq = _mm_mul_ps(ewclj2,rsq00); |
1999 | ewclj6 = _mm_mul_ps(ewclj2,_mm_mul_ps(ewclj2,ewclj2)); |
2000 | exponent = gmx_simd_exp_rgmx_simd_exp_f(ewcljrsq); |
2001 | /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */ |
2002 | poly = _mm_mul_ps(exponent,_mm_add_ps(_mm_sub_ps(one,ewcljrsq),_mm_mul_ps(_mm_mul_ps(ewcljrsq,ewcljrsq),one_half))); |
2003 | /* f6A = 6 * C6grid * (1 - poly) */ |
2004 | f6A = _mm_mul_ps(c6grid_00,_mm_sub_ps(one,poly)); |
2005 | /* f6B = C6grid * exponent * beta^6 */ |
2006 | f6B = _mm_mul_ps(_mm_mul_ps(c6grid_00,one_sixth),_mm_mul_ps(exponent,ewclj6)); |
2007 | /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */ |
2008 | fvdw = _mm_mul_ps(_mm_add_ps(_mm_mul_ps(_mm_sub_ps(_mm_mul_ps(c12_00,rinvsix),_mm_sub_ps(c6_00,f6A)),rinvsix),f6B),rinvsq00); |
2009 | |
2010 | fscal = _mm_add_ps(felec,fvdw); |
2011 | |
2012 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
2013 | |
2014 | /* Calculate temporary vectorial force */ |
2015 | tx = _mm_mul_ps(fscal,dx00); |
2016 | ty = _mm_mul_ps(fscal,dy00); |
2017 | tz = _mm_mul_ps(fscal,dz00); |
2018 | |
2019 | /* Update vectorial force */ |
2020 | fix0 = _mm_add_ps(fix0,tx); |
2021 | fiy0 = _mm_add_ps(fiy0,ty); |
2022 | fiz0 = _mm_add_ps(fiz0,tz); |
2023 | |
2024 | fjx0 = _mm_add_ps(fjx0,tx); |
2025 | fjy0 = _mm_add_ps(fjy0,ty); |
2026 | fjz0 = _mm_add_ps(fjz0,tz); |
2027 | |
2028 | /************************** |
2029 | * CALCULATE INTERACTIONS * |
2030 | **************************/ |
2031 | |
2032 | r01 = _mm_mul_ps(rsq01,rinv01); |
2033 | r01 = _mm_andnot_ps(dummy_mask,r01); |
2034 | |
2035 | /* EWALD ELECTROSTATICS */ |
2036 | |
2037 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
2038 | ewrt = _mm_mul_ps(r01,ewtabscale); |
2039 | ewitab = _mm_cvttps_epi32(ewrt); |
2040 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
2041 | gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})), |
2042 | ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})), |
2043 | &ewtabF,&ewtabFn); |
2044 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
2045 | felec = _mm_mul_ps(_mm_mul_ps(qq01,rinv01),_mm_sub_ps(rinvsq01,felec)); |
2046 | |
2047 | fscal = felec; |
2048 | |
2049 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
2050 | |
2051 | /* Calculate temporary vectorial force */ |
2052 | tx = _mm_mul_ps(fscal,dx01); |
2053 | ty = _mm_mul_ps(fscal,dy01); |
2054 | tz = _mm_mul_ps(fscal,dz01); |
2055 | |
2056 | /* Update vectorial force */ |
2057 | fix0 = _mm_add_ps(fix0,tx); |
2058 | fiy0 = _mm_add_ps(fiy0,ty); |
2059 | fiz0 = _mm_add_ps(fiz0,tz); |
2060 | |
2061 | fjx1 = _mm_add_ps(fjx1,tx); |
2062 | fjy1 = _mm_add_ps(fjy1,ty); |
2063 | fjz1 = _mm_add_ps(fjz1,tz); |
2064 | |
2065 | /************************** |
2066 | * CALCULATE INTERACTIONS * |
2067 | **************************/ |
2068 | |
2069 | r02 = _mm_mul_ps(rsq02,rinv02); |
2070 | r02 = _mm_andnot_ps(dummy_mask,r02); |
2071 | |
2072 | /* EWALD ELECTROSTATICS */ |
2073 | |
2074 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
2075 | ewrt = _mm_mul_ps(r02,ewtabscale); |
2076 | ewitab = _mm_cvttps_epi32(ewrt); |
2077 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
2078 | gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})), |
2079 | ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})), |
2080 | &ewtabF,&ewtabFn); |
2081 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
2082 | felec = _mm_mul_ps(_mm_mul_ps(qq02,rinv02),_mm_sub_ps(rinvsq02,felec)); |
2083 | |
2084 | fscal = felec; |
2085 | |
2086 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
2087 | |
2088 | /* Calculate temporary vectorial force */ |
2089 | tx = _mm_mul_ps(fscal,dx02); |
2090 | ty = _mm_mul_ps(fscal,dy02); |
2091 | tz = _mm_mul_ps(fscal,dz02); |
2092 | |
2093 | /* Update vectorial force */ |
2094 | fix0 = _mm_add_ps(fix0,tx); |
2095 | fiy0 = _mm_add_ps(fiy0,ty); |
2096 | fiz0 = _mm_add_ps(fiz0,tz); |
2097 | |
2098 | fjx2 = _mm_add_ps(fjx2,tx); |
2099 | fjy2 = _mm_add_ps(fjy2,ty); |
2100 | fjz2 = _mm_add_ps(fjz2,tz); |
2101 | |
2102 | /************************** |
2103 | * CALCULATE INTERACTIONS * |
2104 | **************************/ |
2105 | |
2106 | r10 = _mm_mul_ps(rsq10,rinv10); |
2107 | r10 = _mm_andnot_ps(dummy_mask,r10); |
2108 | |
2109 | /* EWALD ELECTROSTATICS */ |
2110 | |
2111 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
2112 | ewrt = _mm_mul_ps(r10,ewtabscale); |
2113 | ewitab = _mm_cvttps_epi32(ewrt); |
2114 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
2115 | gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})), |
2116 | ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})), |
2117 | &ewtabF,&ewtabFn); |
2118 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
2119 | felec = _mm_mul_ps(_mm_mul_ps(qq10,rinv10),_mm_sub_ps(rinvsq10,felec)); |
2120 | |
2121 | fscal = felec; |
2122 | |
2123 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
2124 | |
2125 | /* Calculate temporary vectorial force */ |
2126 | tx = _mm_mul_ps(fscal,dx10); |
2127 | ty = _mm_mul_ps(fscal,dy10); |
2128 | tz = _mm_mul_ps(fscal,dz10); |
2129 | |
2130 | /* Update vectorial force */ |
2131 | fix1 = _mm_add_ps(fix1,tx); |
2132 | fiy1 = _mm_add_ps(fiy1,ty); |
2133 | fiz1 = _mm_add_ps(fiz1,tz); |
2134 | |
2135 | fjx0 = _mm_add_ps(fjx0,tx); |
2136 | fjy0 = _mm_add_ps(fjy0,ty); |
2137 | fjz0 = _mm_add_ps(fjz0,tz); |
2138 | |
2139 | /************************** |
2140 | * CALCULATE INTERACTIONS * |
2141 | **************************/ |
2142 | |
2143 | r11 = _mm_mul_ps(rsq11,rinv11); |
2144 | r11 = _mm_andnot_ps(dummy_mask,r11); |
2145 | |
2146 | /* EWALD ELECTROSTATICS */ |
2147 | |
2148 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
2149 | ewrt = _mm_mul_ps(r11,ewtabscale); |
2150 | ewitab = _mm_cvttps_epi32(ewrt); |
2151 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
2152 | gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})), |
2153 | ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})), |
2154 | &ewtabF,&ewtabFn); |
2155 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
2156 | felec = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec)); |
2157 | |
2158 | fscal = felec; |
2159 | |
2160 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
2161 | |
2162 | /* Calculate temporary vectorial force */ |
2163 | tx = _mm_mul_ps(fscal,dx11); |
2164 | ty = _mm_mul_ps(fscal,dy11); |
2165 | tz = _mm_mul_ps(fscal,dz11); |
2166 | |
2167 | /* Update vectorial force */ |
2168 | fix1 = _mm_add_ps(fix1,tx); |
2169 | fiy1 = _mm_add_ps(fiy1,ty); |
2170 | fiz1 = _mm_add_ps(fiz1,tz); |
2171 | |
2172 | fjx1 = _mm_add_ps(fjx1,tx); |
2173 | fjy1 = _mm_add_ps(fjy1,ty); |
2174 | fjz1 = _mm_add_ps(fjz1,tz); |
2175 | |
2176 | /************************** |
2177 | * CALCULATE INTERACTIONS * |
2178 | **************************/ |
2179 | |
2180 | r12 = _mm_mul_ps(rsq12,rinv12); |
2181 | r12 = _mm_andnot_ps(dummy_mask,r12); |
2182 | |
2183 | /* EWALD ELECTROSTATICS */ |
2184 | |
2185 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
2186 | ewrt = _mm_mul_ps(r12,ewtabscale); |
2187 | ewitab = _mm_cvttps_epi32(ewrt); |
2188 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
2189 | gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})), |
2190 | ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})), |
2191 | &ewtabF,&ewtabFn); |
2192 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
2193 | felec = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec)); |
2194 | |
2195 | fscal = felec; |
2196 | |
2197 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
2198 | |
2199 | /* Calculate temporary vectorial force */ |
2200 | tx = _mm_mul_ps(fscal,dx12); |
2201 | ty = _mm_mul_ps(fscal,dy12); |
2202 | tz = _mm_mul_ps(fscal,dz12); |
2203 | |
2204 | /* Update vectorial force */ |
2205 | fix1 = _mm_add_ps(fix1,tx); |
2206 | fiy1 = _mm_add_ps(fiy1,ty); |
2207 | fiz1 = _mm_add_ps(fiz1,tz); |
2208 | |
2209 | fjx2 = _mm_add_ps(fjx2,tx); |
2210 | fjy2 = _mm_add_ps(fjy2,ty); |
2211 | fjz2 = _mm_add_ps(fjz2,tz); |
2212 | |
2213 | /************************** |
2214 | * CALCULATE INTERACTIONS * |
2215 | **************************/ |
2216 | |
2217 | r20 = _mm_mul_ps(rsq20,rinv20); |
2218 | r20 = _mm_andnot_ps(dummy_mask,r20); |
2219 | |
2220 | /* EWALD ELECTROSTATICS */ |
2221 | |
2222 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
2223 | ewrt = _mm_mul_ps(r20,ewtabscale); |
2224 | ewitab = _mm_cvttps_epi32(ewrt); |
2225 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
2226 | gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})), |
2227 | ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})), |
2228 | &ewtabF,&ewtabFn); |
2229 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
2230 | felec = _mm_mul_ps(_mm_mul_ps(qq20,rinv20),_mm_sub_ps(rinvsq20,felec)); |
2231 | |
2232 | fscal = felec; |
2233 | |
2234 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
2235 | |
2236 | /* Calculate temporary vectorial force */ |
2237 | tx = _mm_mul_ps(fscal,dx20); |
2238 | ty = _mm_mul_ps(fscal,dy20); |
2239 | tz = _mm_mul_ps(fscal,dz20); |
2240 | |
2241 | /* Update vectorial force */ |
2242 | fix2 = _mm_add_ps(fix2,tx); |
2243 | fiy2 = _mm_add_ps(fiy2,ty); |
2244 | fiz2 = _mm_add_ps(fiz2,tz); |
2245 | |
2246 | fjx0 = _mm_add_ps(fjx0,tx); |
2247 | fjy0 = _mm_add_ps(fjy0,ty); |
2248 | fjz0 = _mm_add_ps(fjz0,tz); |
2249 | |
2250 | /************************** |
2251 | * CALCULATE INTERACTIONS * |
2252 | **************************/ |
2253 | |
2254 | r21 = _mm_mul_ps(rsq21,rinv21); |
2255 | r21 = _mm_andnot_ps(dummy_mask,r21); |
2256 | |
2257 | /* EWALD ELECTROSTATICS */ |
2258 | |
2259 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
2260 | ewrt = _mm_mul_ps(r21,ewtabscale); |
2261 | ewitab = _mm_cvttps_epi32(ewrt); |
2262 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
2263 | gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})), |
2264 | ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})), |
2265 | &ewtabF,&ewtabFn); |
2266 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
2267 | felec = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec)); |
2268 | |
2269 | fscal = felec; |
2270 | |
2271 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
2272 | |
2273 | /* Calculate temporary vectorial force */ |
2274 | tx = _mm_mul_ps(fscal,dx21); |
2275 | ty = _mm_mul_ps(fscal,dy21); |
2276 | tz = _mm_mul_ps(fscal,dz21); |
2277 | |
2278 | /* Update vectorial force */ |
2279 | fix2 = _mm_add_ps(fix2,tx); |
2280 | fiy2 = _mm_add_ps(fiy2,ty); |
2281 | fiz2 = _mm_add_ps(fiz2,tz); |
2282 | |
2283 | fjx1 = _mm_add_ps(fjx1,tx); |
2284 | fjy1 = _mm_add_ps(fjy1,ty); |
2285 | fjz1 = _mm_add_ps(fjz1,tz); |
2286 | |
2287 | /************************** |
2288 | * CALCULATE INTERACTIONS * |
2289 | **************************/ |
2290 | |
2291 | r22 = _mm_mul_ps(rsq22,rinv22); |
2292 | r22 = _mm_andnot_ps(dummy_mask,r22); |
2293 | |
2294 | /* EWALD ELECTROSTATICS */ |
2295 | |
2296 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
2297 | ewrt = _mm_mul_ps(r22,ewtabscale); |
2298 | ewitab = _mm_cvttps_epi32(ewrt); |
2299 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
2300 | gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})), |
2301 | ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})), |
2302 | &ewtabF,&ewtabFn); |
2303 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
2304 | felec = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec)); |
2305 | |
2306 | fscal = felec; |
2307 | |
2308 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
2309 | |
2310 | /* Calculate temporary vectorial force */ |
2311 | tx = _mm_mul_ps(fscal,dx22); |
2312 | ty = _mm_mul_ps(fscal,dy22); |
2313 | tz = _mm_mul_ps(fscal,dz22); |
2314 | |
2315 | /* Update vectorial force */ |
2316 | fix2 = _mm_add_ps(fix2,tx); |
2317 | fiy2 = _mm_add_ps(fiy2,ty); |
2318 | fiz2 = _mm_add_ps(fiz2,tz); |
2319 | |
2320 | fjx2 = _mm_add_ps(fjx2,tx); |
2321 | fjy2 = _mm_add_ps(fjy2,ty); |
2322 | fjz2 = _mm_add_ps(fjz2,tz); |
2323 | |
2324 | fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch; |
2325 | fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch; |
2326 | fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch; |
2327 | fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch; |
2328 | |
2329 | gmx_mm_decrement_3rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD, |
2330 | fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2); |
2331 | |
2332 | /* Inner loop uses 356 flops */ |
2333 | } |
2334 | |
2335 | /* End of innermost loop */ |
2336 | |
2337 | gmx_mm_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2, |
2338 | f+i_coord_offset,fshift+i_shift_offset); |
2339 | |
2340 | /* Increment number of inner iterations */ |
2341 | inneriter += j_index_end - j_index_start; |
2342 | |
2343 | /* Outer loop uses 18 flops */ |
2344 | } |
2345 | |
2346 | /* Increment number of outer iterations */ |
2347 | outeriter += nri; |
2348 | |
2349 | /* Update outer/inner flops */ |
2350 | |
2351 | inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_F,outeriter*18 + inneriter*356)(nrnb)->n[eNR_NBKERNEL_ELEC_VDW_W3W3_F] += outeriter*18 + inneriter *356; |
2352 | } |