File: | gromacs/gmxlib/nonbonded/nb_kernel_sse4_1_single/nb_kernel_ElecEw_VdwNone_GeomW3W3_sse4_1_single.c |
Location: | line 129, column 5 |
Description: | Value stored to 'sh_ewald' 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_VdwNone_GeomW3W3_VF_sse4_1_single |
54 | * Electrostatics interaction: Ewald |
55 | * VdW interaction: None |
56 | * Geometry: Water3-Water3 |
57 | * Calculate force/pot: PotentialAndForce |
58 | */ |
59 | void |
60 | nb_kernel_ElecEw_VdwNone_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 | __m128i ewitab; |
109 | __m128 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV; |
110 | real *ewtab; |
111 | __m128 dummy_mask,cutoff_mask; |
112 | __m128 signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) ); |
113 | __m128 one = _mm_set1_ps(1.0); |
114 | __m128 two = _mm_set1_ps(2.0); |
115 | x = xx[0]; |
116 | f = ff[0]; |
117 | |
118 | nri = nlist->nri; |
119 | iinr = nlist->iinr; |
120 | jindex = nlist->jindex; |
121 | jjnr = nlist->jjnr; |
122 | shiftidx = nlist->shift; |
123 | gid = nlist->gid; |
124 | shiftvec = fr->shift_vec[0]; |
125 | fshift = fr->fshift[0]; |
126 | facel = _mm_set1_ps(fr->epsfac); |
127 | charge = mdatoms->chargeA; |
128 | |
129 | sh_ewald = _mm_set1_ps(fr->ic->sh_ewald); |
Value stored to 'sh_ewald' is never read | |
130 | ewtab = fr->ic->tabq_coul_FDV0; |
131 | ewtabscale = _mm_set1_ps(fr->ic->tabq_scale); |
132 | ewtabhalfspace = _mm_set1_ps(0.5/fr->ic->tabq_scale); |
133 | |
134 | /* Setup water-specific parameters */ |
135 | inr = nlist->iinr[0]; |
136 | iq0 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+0])); |
137 | iq1 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1])); |
138 | iq2 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2])); |
139 | |
140 | jq0 = _mm_set1_ps(charge[inr+0]); |
141 | jq1 = _mm_set1_ps(charge[inr+1]); |
142 | jq2 = _mm_set1_ps(charge[inr+2]); |
143 | qq00 = _mm_mul_ps(iq0,jq0); |
144 | qq01 = _mm_mul_ps(iq0,jq1); |
145 | qq02 = _mm_mul_ps(iq0,jq2); |
146 | qq10 = _mm_mul_ps(iq1,jq0); |
147 | qq11 = _mm_mul_ps(iq1,jq1); |
148 | qq12 = _mm_mul_ps(iq1,jq2); |
149 | qq20 = _mm_mul_ps(iq2,jq0); |
150 | qq21 = _mm_mul_ps(iq2,jq1); |
151 | qq22 = _mm_mul_ps(iq2,jq2); |
152 | |
153 | /* Avoid stupid compiler warnings */ |
154 | jnrA = jnrB = jnrC = jnrD = 0; |
155 | j_coord_offsetA = 0; |
156 | j_coord_offsetB = 0; |
157 | j_coord_offsetC = 0; |
158 | j_coord_offsetD = 0; |
159 | |
160 | outeriter = 0; |
161 | inneriter = 0; |
162 | |
163 | for(iidx=0;iidx<4*DIM3;iidx++) |
164 | { |
165 | scratch[iidx] = 0.0; |
166 | } |
167 | |
168 | /* Start outer loop over neighborlists */ |
169 | for(iidx=0; iidx<nri; iidx++) |
170 | { |
171 | /* Load shift vector for this list */ |
172 | i_shift_offset = DIM3*shiftidx[iidx]; |
173 | |
174 | /* Load limits for loop over neighbors */ |
175 | j_index_start = jindex[iidx]; |
176 | j_index_end = jindex[iidx+1]; |
177 | |
178 | /* Get outer coordinate index */ |
179 | inr = iinr[iidx]; |
180 | i_coord_offset = DIM3*inr; |
181 | |
182 | /* Load i particle coords and add shift vector */ |
183 | gmx_mm_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset, |
184 | &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2); |
185 | |
186 | fix0 = _mm_setzero_ps(); |
187 | fiy0 = _mm_setzero_ps(); |
188 | fiz0 = _mm_setzero_ps(); |
189 | fix1 = _mm_setzero_ps(); |
190 | fiy1 = _mm_setzero_ps(); |
191 | fiz1 = _mm_setzero_ps(); |
192 | fix2 = _mm_setzero_ps(); |
193 | fiy2 = _mm_setzero_ps(); |
194 | fiz2 = _mm_setzero_ps(); |
195 | |
196 | /* Reset potential sums */ |
197 | velecsum = _mm_setzero_ps(); |
198 | |
199 | /* Start inner kernel loop */ |
200 | for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4) |
201 | { |
202 | |
203 | /* Get j neighbor index, and coordinate index */ |
204 | jnrA = jjnr[jidx]; |
205 | jnrB = jjnr[jidx+1]; |
206 | jnrC = jjnr[jidx+2]; |
207 | jnrD = jjnr[jidx+3]; |
208 | j_coord_offsetA = DIM3*jnrA; |
209 | j_coord_offsetB = DIM3*jnrB; |
210 | j_coord_offsetC = DIM3*jnrC; |
211 | j_coord_offsetD = DIM3*jnrD; |
212 | |
213 | /* load j atom coordinates */ |
214 | gmx_mm_load_3rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB, |
215 | x+j_coord_offsetC,x+j_coord_offsetD, |
216 | &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2); |
217 | |
218 | /* Calculate displacement vector */ |
219 | dx00 = _mm_sub_ps(ix0,jx0); |
220 | dy00 = _mm_sub_ps(iy0,jy0); |
221 | dz00 = _mm_sub_ps(iz0,jz0); |
222 | dx01 = _mm_sub_ps(ix0,jx1); |
223 | dy01 = _mm_sub_ps(iy0,jy1); |
224 | dz01 = _mm_sub_ps(iz0,jz1); |
225 | dx02 = _mm_sub_ps(ix0,jx2); |
226 | dy02 = _mm_sub_ps(iy0,jy2); |
227 | dz02 = _mm_sub_ps(iz0,jz2); |
228 | dx10 = _mm_sub_ps(ix1,jx0); |
229 | dy10 = _mm_sub_ps(iy1,jy0); |
230 | dz10 = _mm_sub_ps(iz1,jz0); |
231 | dx11 = _mm_sub_ps(ix1,jx1); |
232 | dy11 = _mm_sub_ps(iy1,jy1); |
233 | dz11 = _mm_sub_ps(iz1,jz1); |
234 | dx12 = _mm_sub_ps(ix1,jx2); |
235 | dy12 = _mm_sub_ps(iy1,jy2); |
236 | dz12 = _mm_sub_ps(iz1,jz2); |
237 | dx20 = _mm_sub_ps(ix2,jx0); |
238 | dy20 = _mm_sub_ps(iy2,jy0); |
239 | dz20 = _mm_sub_ps(iz2,jz0); |
240 | dx21 = _mm_sub_ps(ix2,jx1); |
241 | dy21 = _mm_sub_ps(iy2,jy1); |
242 | dz21 = _mm_sub_ps(iz2,jz1); |
243 | dx22 = _mm_sub_ps(ix2,jx2); |
244 | dy22 = _mm_sub_ps(iy2,jy2); |
245 | dz22 = _mm_sub_ps(iz2,jz2); |
246 | |
247 | /* Calculate squared distance and things based on it */ |
248 | rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00); |
249 | rsq01 = gmx_mm_calc_rsq_ps(dx01,dy01,dz01); |
250 | rsq02 = gmx_mm_calc_rsq_ps(dx02,dy02,dz02); |
251 | rsq10 = gmx_mm_calc_rsq_ps(dx10,dy10,dz10); |
252 | rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11); |
253 | rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12); |
254 | rsq20 = gmx_mm_calc_rsq_ps(dx20,dy20,dz20); |
255 | rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21); |
256 | rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22); |
257 | |
258 | rinv00 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq00); |
259 | rinv01 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq01); |
260 | rinv02 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq02); |
261 | rinv10 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq10); |
262 | rinv11 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq11); |
263 | rinv12 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq12); |
264 | rinv20 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq20); |
265 | rinv21 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq21); |
266 | rinv22 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq22); |
267 | |
268 | rinvsq00 = _mm_mul_ps(rinv00,rinv00); |
269 | rinvsq01 = _mm_mul_ps(rinv01,rinv01); |
270 | rinvsq02 = _mm_mul_ps(rinv02,rinv02); |
271 | rinvsq10 = _mm_mul_ps(rinv10,rinv10); |
272 | rinvsq11 = _mm_mul_ps(rinv11,rinv11); |
273 | rinvsq12 = _mm_mul_ps(rinv12,rinv12); |
274 | rinvsq20 = _mm_mul_ps(rinv20,rinv20); |
275 | rinvsq21 = _mm_mul_ps(rinv21,rinv21); |
276 | rinvsq22 = _mm_mul_ps(rinv22,rinv22); |
277 | |
278 | fjx0 = _mm_setzero_ps(); |
279 | fjy0 = _mm_setzero_ps(); |
280 | fjz0 = _mm_setzero_ps(); |
281 | fjx1 = _mm_setzero_ps(); |
282 | fjy1 = _mm_setzero_ps(); |
283 | fjz1 = _mm_setzero_ps(); |
284 | fjx2 = _mm_setzero_ps(); |
285 | fjy2 = _mm_setzero_ps(); |
286 | fjz2 = _mm_setzero_ps(); |
287 | |
288 | /************************** |
289 | * CALCULATE INTERACTIONS * |
290 | **************************/ |
291 | |
292 | r00 = _mm_mul_ps(rsq00,rinv00); |
293 | |
294 | /* EWALD ELECTROSTATICS */ |
295 | |
296 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
297 | ewrt = _mm_mul_ps(r00,ewtabscale); |
298 | ewitab = _mm_cvttps_epi32(ewrt); |
299 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
300 | ewitab = _mm_slli_epi32(ewitab,2); |
301 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
302 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
303 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
304 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
305 | _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); |
306 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
307 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
308 | velec = _mm_mul_ps(qq00,_mm_sub_ps(rinv00,velec)); |
309 | felec = _mm_mul_ps(_mm_mul_ps(qq00,rinv00),_mm_sub_ps(rinvsq00,felec)); |
310 | |
311 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
312 | velecsum = _mm_add_ps(velecsum,velec); |
313 | |
314 | fscal = felec; |
315 | |
316 | /* Calculate temporary vectorial force */ |
317 | tx = _mm_mul_ps(fscal,dx00); |
318 | ty = _mm_mul_ps(fscal,dy00); |
319 | tz = _mm_mul_ps(fscal,dz00); |
320 | |
321 | /* Update vectorial force */ |
322 | fix0 = _mm_add_ps(fix0,tx); |
323 | fiy0 = _mm_add_ps(fiy0,ty); |
324 | fiz0 = _mm_add_ps(fiz0,tz); |
325 | |
326 | fjx0 = _mm_add_ps(fjx0,tx); |
327 | fjy0 = _mm_add_ps(fjy0,ty); |
328 | fjz0 = _mm_add_ps(fjz0,tz); |
329 | |
330 | /************************** |
331 | * CALCULATE INTERACTIONS * |
332 | **************************/ |
333 | |
334 | r01 = _mm_mul_ps(rsq01,rinv01); |
335 | |
336 | /* EWALD ELECTROSTATICS */ |
337 | |
338 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
339 | ewrt = _mm_mul_ps(r01,ewtabscale); |
340 | ewitab = _mm_cvttps_epi32(ewrt); |
341 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
342 | ewitab = _mm_slli_epi32(ewitab,2); |
343 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
344 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
345 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
346 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
347 | _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); |
348 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
349 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
350 | velec = _mm_mul_ps(qq01,_mm_sub_ps(rinv01,velec)); |
351 | felec = _mm_mul_ps(_mm_mul_ps(qq01,rinv01),_mm_sub_ps(rinvsq01,felec)); |
352 | |
353 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
354 | velecsum = _mm_add_ps(velecsum,velec); |
355 | |
356 | fscal = felec; |
357 | |
358 | /* Calculate temporary vectorial force */ |
359 | tx = _mm_mul_ps(fscal,dx01); |
360 | ty = _mm_mul_ps(fscal,dy01); |
361 | tz = _mm_mul_ps(fscal,dz01); |
362 | |
363 | /* Update vectorial force */ |
364 | fix0 = _mm_add_ps(fix0,tx); |
365 | fiy0 = _mm_add_ps(fiy0,ty); |
366 | fiz0 = _mm_add_ps(fiz0,tz); |
367 | |
368 | fjx1 = _mm_add_ps(fjx1,tx); |
369 | fjy1 = _mm_add_ps(fjy1,ty); |
370 | fjz1 = _mm_add_ps(fjz1,tz); |
371 | |
372 | /************************** |
373 | * CALCULATE INTERACTIONS * |
374 | **************************/ |
375 | |
376 | r02 = _mm_mul_ps(rsq02,rinv02); |
377 | |
378 | /* EWALD ELECTROSTATICS */ |
379 | |
380 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
381 | ewrt = _mm_mul_ps(r02,ewtabscale); |
382 | ewitab = _mm_cvttps_epi32(ewrt); |
383 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
384 | ewitab = _mm_slli_epi32(ewitab,2); |
385 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
386 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
387 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
388 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
389 | _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); |
390 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
391 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
392 | velec = _mm_mul_ps(qq02,_mm_sub_ps(rinv02,velec)); |
393 | felec = _mm_mul_ps(_mm_mul_ps(qq02,rinv02),_mm_sub_ps(rinvsq02,felec)); |
394 | |
395 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
396 | velecsum = _mm_add_ps(velecsum,velec); |
397 | |
398 | fscal = felec; |
399 | |
400 | /* Calculate temporary vectorial force */ |
401 | tx = _mm_mul_ps(fscal,dx02); |
402 | ty = _mm_mul_ps(fscal,dy02); |
403 | tz = _mm_mul_ps(fscal,dz02); |
404 | |
405 | /* Update vectorial force */ |
406 | fix0 = _mm_add_ps(fix0,tx); |
407 | fiy0 = _mm_add_ps(fiy0,ty); |
408 | fiz0 = _mm_add_ps(fiz0,tz); |
409 | |
410 | fjx2 = _mm_add_ps(fjx2,tx); |
411 | fjy2 = _mm_add_ps(fjy2,ty); |
412 | fjz2 = _mm_add_ps(fjz2,tz); |
413 | |
414 | /************************** |
415 | * CALCULATE INTERACTIONS * |
416 | **************************/ |
417 | |
418 | r10 = _mm_mul_ps(rsq10,rinv10); |
419 | |
420 | /* EWALD ELECTROSTATICS */ |
421 | |
422 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
423 | ewrt = _mm_mul_ps(r10,ewtabscale); |
424 | ewitab = _mm_cvttps_epi32(ewrt); |
425 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
426 | ewitab = _mm_slli_epi32(ewitab,2); |
427 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
428 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
429 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
430 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
431 | _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); |
432 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
433 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
434 | velec = _mm_mul_ps(qq10,_mm_sub_ps(rinv10,velec)); |
435 | felec = _mm_mul_ps(_mm_mul_ps(qq10,rinv10),_mm_sub_ps(rinvsq10,felec)); |
436 | |
437 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
438 | velecsum = _mm_add_ps(velecsum,velec); |
439 | |
440 | fscal = felec; |
441 | |
442 | /* Calculate temporary vectorial force */ |
443 | tx = _mm_mul_ps(fscal,dx10); |
444 | ty = _mm_mul_ps(fscal,dy10); |
445 | tz = _mm_mul_ps(fscal,dz10); |
446 | |
447 | /* Update vectorial force */ |
448 | fix1 = _mm_add_ps(fix1,tx); |
449 | fiy1 = _mm_add_ps(fiy1,ty); |
450 | fiz1 = _mm_add_ps(fiz1,tz); |
451 | |
452 | fjx0 = _mm_add_ps(fjx0,tx); |
453 | fjy0 = _mm_add_ps(fjy0,ty); |
454 | fjz0 = _mm_add_ps(fjz0,tz); |
455 | |
456 | /************************** |
457 | * CALCULATE INTERACTIONS * |
458 | **************************/ |
459 | |
460 | r11 = _mm_mul_ps(rsq11,rinv11); |
461 | |
462 | /* EWALD ELECTROSTATICS */ |
463 | |
464 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
465 | ewrt = _mm_mul_ps(r11,ewtabscale); |
466 | ewitab = _mm_cvttps_epi32(ewrt); |
467 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
468 | ewitab = _mm_slli_epi32(ewitab,2); |
469 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
470 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
471 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
472 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
473 | _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); |
474 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
475 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
476 | velec = _mm_mul_ps(qq11,_mm_sub_ps(rinv11,velec)); |
477 | felec = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec)); |
478 | |
479 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
480 | velecsum = _mm_add_ps(velecsum,velec); |
481 | |
482 | fscal = felec; |
483 | |
484 | /* Calculate temporary vectorial force */ |
485 | tx = _mm_mul_ps(fscal,dx11); |
486 | ty = _mm_mul_ps(fscal,dy11); |
487 | tz = _mm_mul_ps(fscal,dz11); |
488 | |
489 | /* Update vectorial force */ |
490 | fix1 = _mm_add_ps(fix1,tx); |
491 | fiy1 = _mm_add_ps(fiy1,ty); |
492 | fiz1 = _mm_add_ps(fiz1,tz); |
493 | |
494 | fjx1 = _mm_add_ps(fjx1,tx); |
495 | fjy1 = _mm_add_ps(fjy1,ty); |
496 | fjz1 = _mm_add_ps(fjz1,tz); |
497 | |
498 | /************************** |
499 | * CALCULATE INTERACTIONS * |
500 | **************************/ |
501 | |
502 | r12 = _mm_mul_ps(rsq12,rinv12); |
503 | |
504 | /* EWALD ELECTROSTATICS */ |
505 | |
506 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
507 | ewrt = _mm_mul_ps(r12,ewtabscale); |
508 | ewitab = _mm_cvttps_epi32(ewrt); |
509 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
510 | ewitab = _mm_slli_epi32(ewitab,2); |
511 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
512 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
513 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
514 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
515 | _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); |
516 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
517 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
518 | velec = _mm_mul_ps(qq12,_mm_sub_ps(rinv12,velec)); |
519 | felec = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec)); |
520 | |
521 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
522 | velecsum = _mm_add_ps(velecsum,velec); |
523 | |
524 | fscal = felec; |
525 | |
526 | /* Calculate temporary vectorial force */ |
527 | tx = _mm_mul_ps(fscal,dx12); |
528 | ty = _mm_mul_ps(fscal,dy12); |
529 | tz = _mm_mul_ps(fscal,dz12); |
530 | |
531 | /* Update vectorial force */ |
532 | fix1 = _mm_add_ps(fix1,tx); |
533 | fiy1 = _mm_add_ps(fiy1,ty); |
534 | fiz1 = _mm_add_ps(fiz1,tz); |
535 | |
536 | fjx2 = _mm_add_ps(fjx2,tx); |
537 | fjy2 = _mm_add_ps(fjy2,ty); |
538 | fjz2 = _mm_add_ps(fjz2,tz); |
539 | |
540 | /************************** |
541 | * CALCULATE INTERACTIONS * |
542 | **************************/ |
543 | |
544 | r20 = _mm_mul_ps(rsq20,rinv20); |
545 | |
546 | /* EWALD ELECTROSTATICS */ |
547 | |
548 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
549 | ewrt = _mm_mul_ps(r20,ewtabscale); |
550 | ewitab = _mm_cvttps_epi32(ewrt); |
551 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
552 | ewitab = _mm_slli_epi32(ewitab,2); |
553 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
554 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
555 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
556 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
557 | _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); |
558 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
559 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
560 | velec = _mm_mul_ps(qq20,_mm_sub_ps(rinv20,velec)); |
561 | felec = _mm_mul_ps(_mm_mul_ps(qq20,rinv20),_mm_sub_ps(rinvsq20,felec)); |
562 | |
563 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
564 | velecsum = _mm_add_ps(velecsum,velec); |
565 | |
566 | fscal = felec; |
567 | |
568 | /* Calculate temporary vectorial force */ |
569 | tx = _mm_mul_ps(fscal,dx20); |
570 | ty = _mm_mul_ps(fscal,dy20); |
571 | tz = _mm_mul_ps(fscal,dz20); |
572 | |
573 | /* Update vectorial force */ |
574 | fix2 = _mm_add_ps(fix2,tx); |
575 | fiy2 = _mm_add_ps(fiy2,ty); |
576 | fiz2 = _mm_add_ps(fiz2,tz); |
577 | |
578 | fjx0 = _mm_add_ps(fjx0,tx); |
579 | fjy0 = _mm_add_ps(fjy0,ty); |
580 | fjz0 = _mm_add_ps(fjz0,tz); |
581 | |
582 | /************************** |
583 | * CALCULATE INTERACTIONS * |
584 | **************************/ |
585 | |
586 | r21 = _mm_mul_ps(rsq21,rinv21); |
587 | |
588 | /* EWALD ELECTROSTATICS */ |
589 | |
590 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
591 | ewrt = _mm_mul_ps(r21,ewtabscale); |
592 | ewitab = _mm_cvttps_epi32(ewrt); |
593 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
594 | ewitab = _mm_slli_epi32(ewitab,2); |
595 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
596 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
597 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
598 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
599 | _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); |
600 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
601 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
602 | velec = _mm_mul_ps(qq21,_mm_sub_ps(rinv21,velec)); |
603 | felec = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec)); |
604 | |
605 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
606 | velecsum = _mm_add_ps(velecsum,velec); |
607 | |
608 | fscal = felec; |
609 | |
610 | /* Calculate temporary vectorial force */ |
611 | tx = _mm_mul_ps(fscal,dx21); |
612 | ty = _mm_mul_ps(fscal,dy21); |
613 | tz = _mm_mul_ps(fscal,dz21); |
614 | |
615 | /* Update vectorial force */ |
616 | fix2 = _mm_add_ps(fix2,tx); |
617 | fiy2 = _mm_add_ps(fiy2,ty); |
618 | fiz2 = _mm_add_ps(fiz2,tz); |
619 | |
620 | fjx1 = _mm_add_ps(fjx1,tx); |
621 | fjy1 = _mm_add_ps(fjy1,ty); |
622 | fjz1 = _mm_add_ps(fjz1,tz); |
623 | |
624 | /************************** |
625 | * CALCULATE INTERACTIONS * |
626 | **************************/ |
627 | |
628 | r22 = _mm_mul_ps(rsq22,rinv22); |
629 | |
630 | /* EWALD ELECTROSTATICS */ |
631 | |
632 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
633 | ewrt = _mm_mul_ps(r22,ewtabscale); |
634 | ewitab = _mm_cvttps_epi32(ewrt); |
635 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
636 | ewitab = _mm_slli_epi32(ewitab,2); |
637 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
638 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
639 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
640 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
641 | _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); |
642 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
643 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
644 | velec = _mm_mul_ps(qq22,_mm_sub_ps(rinv22,velec)); |
645 | felec = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec)); |
646 | |
647 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
648 | velecsum = _mm_add_ps(velecsum,velec); |
649 | |
650 | fscal = felec; |
651 | |
652 | /* Calculate temporary vectorial force */ |
653 | tx = _mm_mul_ps(fscal,dx22); |
654 | ty = _mm_mul_ps(fscal,dy22); |
655 | tz = _mm_mul_ps(fscal,dz22); |
656 | |
657 | /* Update vectorial force */ |
658 | fix2 = _mm_add_ps(fix2,tx); |
659 | fiy2 = _mm_add_ps(fiy2,ty); |
660 | fiz2 = _mm_add_ps(fiz2,tz); |
661 | |
662 | fjx2 = _mm_add_ps(fjx2,tx); |
663 | fjy2 = _mm_add_ps(fjy2,ty); |
664 | fjz2 = _mm_add_ps(fjz2,tz); |
665 | |
666 | fjptrA = f+j_coord_offsetA; |
667 | fjptrB = f+j_coord_offsetB; |
668 | fjptrC = f+j_coord_offsetC; |
669 | fjptrD = f+j_coord_offsetD; |
670 | |
671 | gmx_mm_decrement_3rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD, |
672 | fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2); |
673 | |
674 | /* Inner loop uses 369 flops */ |
675 | } |
676 | |
677 | if(jidx<j_index_end) |
678 | { |
679 | |
680 | /* Get j neighbor index, and coordinate index */ |
681 | jnrlistA = jjnr[jidx]; |
682 | jnrlistB = jjnr[jidx+1]; |
683 | jnrlistC = jjnr[jidx+2]; |
684 | jnrlistD = jjnr[jidx+3]; |
685 | /* Sign of each element will be negative for non-real atoms. |
686 | * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones, |
687 | * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries. |
688 | */ |
689 | dummy_mask = gmx_mm_castsi128_ps_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())); |
690 | jnrA = (jnrlistA>=0) ? jnrlistA : 0; |
691 | jnrB = (jnrlistB>=0) ? jnrlistB : 0; |
692 | jnrC = (jnrlistC>=0) ? jnrlistC : 0; |
693 | jnrD = (jnrlistD>=0) ? jnrlistD : 0; |
694 | j_coord_offsetA = DIM3*jnrA; |
695 | j_coord_offsetB = DIM3*jnrB; |
696 | j_coord_offsetC = DIM3*jnrC; |
697 | j_coord_offsetD = DIM3*jnrD; |
698 | |
699 | /* load j atom coordinates */ |
700 | gmx_mm_load_3rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB, |
701 | x+j_coord_offsetC,x+j_coord_offsetD, |
702 | &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2); |
703 | |
704 | /* Calculate displacement vector */ |
705 | dx00 = _mm_sub_ps(ix0,jx0); |
706 | dy00 = _mm_sub_ps(iy0,jy0); |
707 | dz00 = _mm_sub_ps(iz0,jz0); |
708 | dx01 = _mm_sub_ps(ix0,jx1); |
709 | dy01 = _mm_sub_ps(iy0,jy1); |
710 | dz01 = _mm_sub_ps(iz0,jz1); |
711 | dx02 = _mm_sub_ps(ix0,jx2); |
712 | dy02 = _mm_sub_ps(iy0,jy2); |
713 | dz02 = _mm_sub_ps(iz0,jz2); |
714 | dx10 = _mm_sub_ps(ix1,jx0); |
715 | dy10 = _mm_sub_ps(iy1,jy0); |
716 | dz10 = _mm_sub_ps(iz1,jz0); |
717 | dx11 = _mm_sub_ps(ix1,jx1); |
718 | dy11 = _mm_sub_ps(iy1,jy1); |
719 | dz11 = _mm_sub_ps(iz1,jz1); |
720 | dx12 = _mm_sub_ps(ix1,jx2); |
721 | dy12 = _mm_sub_ps(iy1,jy2); |
722 | dz12 = _mm_sub_ps(iz1,jz2); |
723 | dx20 = _mm_sub_ps(ix2,jx0); |
724 | dy20 = _mm_sub_ps(iy2,jy0); |
725 | dz20 = _mm_sub_ps(iz2,jz0); |
726 | dx21 = _mm_sub_ps(ix2,jx1); |
727 | dy21 = _mm_sub_ps(iy2,jy1); |
728 | dz21 = _mm_sub_ps(iz2,jz1); |
729 | dx22 = _mm_sub_ps(ix2,jx2); |
730 | dy22 = _mm_sub_ps(iy2,jy2); |
731 | dz22 = _mm_sub_ps(iz2,jz2); |
732 | |
733 | /* Calculate squared distance and things based on it */ |
734 | rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00); |
735 | rsq01 = gmx_mm_calc_rsq_ps(dx01,dy01,dz01); |
736 | rsq02 = gmx_mm_calc_rsq_ps(dx02,dy02,dz02); |
737 | rsq10 = gmx_mm_calc_rsq_ps(dx10,dy10,dz10); |
738 | rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11); |
739 | rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12); |
740 | rsq20 = gmx_mm_calc_rsq_ps(dx20,dy20,dz20); |
741 | rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21); |
742 | rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22); |
743 | |
744 | rinv00 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq00); |
745 | rinv01 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq01); |
746 | rinv02 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq02); |
747 | rinv10 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq10); |
748 | rinv11 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq11); |
749 | rinv12 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq12); |
750 | rinv20 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq20); |
751 | rinv21 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq21); |
752 | rinv22 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq22); |
753 | |
754 | rinvsq00 = _mm_mul_ps(rinv00,rinv00); |
755 | rinvsq01 = _mm_mul_ps(rinv01,rinv01); |
756 | rinvsq02 = _mm_mul_ps(rinv02,rinv02); |
757 | rinvsq10 = _mm_mul_ps(rinv10,rinv10); |
758 | rinvsq11 = _mm_mul_ps(rinv11,rinv11); |
759 | rinvsq12 = _mm_mul_ps(rinv12,rinv12); |
760 | rinvsq20 = _mm_mul_ps(rinv20,rinv20); |
761 | rinvsq21 = _mm_mul_ps(rinv21,rinv21); |
762 | rinvsq22 = _mm_mul_ps(rinv22,rinv22); |
763 | |
764 | fjx0 = _mm_setzero_ps(); |
765 | fjy0 = _mm_setzero_ps(); |
766 | fjz0 = _mm_setzero_ps(); |
767 | fjx1 = _mm_setzero_ps(); |
768 | fjy1 = _mm_setzero_ps(); |
769 | fjz1 = _mm_setzero_ps(); |
770 | fjx2 = _mm_setzero_ps(); |
771 | fjy2 = _mm_setzero_ps(); |
772 | fjz2 = _mm_setzero_ps(); |
773 | |
774 | /************************** |
775 | * CALCULATE INTERACTIONS * |
776 | **************************/ |
777 | |
778 | r00 = _mm_mul_ps(rsq00,rinv00); |
779 | r00 = _mm_andnot_ps(dummy_mask,r00); |
780 | |
781 | /* EWALD ELECTROSTATICS */ |
782 | |
783 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
784 | ewrt = _mm_mul_ps(r00,ewtabscale); |
785 | ewitab = _mm_cvttps_epi32(ewrt); |
786 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
787 | ewitab = _mm_slli_epi32(ewitab,2); |
788 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
789 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
790 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
791 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
792 | _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); |
793 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
794 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
795 | velec = _mm_mul_ps(qq00,_mm_sub_ps(rinv00,velec)); |
796 | felec = _mm_mul_ps(_mm_mul_ps(qq00,rinv00),_mm_sub_ps(rinvsq00,felec)); |
797 | |
798 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
799 | velec = _mm_andnot_ps(dummy_mask,velec); |
800 | velecsum = _mm_add_ps(velecsum,velec); |
801 | |
802 | fscal = felec; |
803 | |
804 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
805 | |
806 | /* Calculate temporary vectorial force */ |
807 | tx = _mm_mul_ps(fscal,dx00); |
808 | ty = _mm_mul_ps(fscal,dy00); |
809 | tz = _mm_mul_ps(fscal,dz00); |
810 | |
811 | /* Update vectorial force */ |
812 | fix0 = _mm_add_ps(fix0,tx); |
813 | fiy0 = _mm_add_ps(fiy0,ty); |
814 | fiz0 = _mm_add_ps(fiz0,tz); |
815 | |
816 | fjx0 = _mm_add_ps(fjx0,tx); |
817 | fjy0 = _mm_add_ps(fjy0,ty); |
818 | fjz0 = _mm_add_ps(fjz0,tz); |
819 | |
820 | /************************** |
821 | * CALCULATE INTERACTIONS * |
822 | **************************/ |
823 | |
824 | r01 = _mm_mul_ps(rsq01,rinv01); |
825 | r01 = _mm_andnot_ps(dummy_mask,r01); |
826 | |
827 | /* EWALD ELECTROSTATICS */ |
828 | |
829 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
830 | ewrt = _mm_mul_ps(r01,ewtabscale); |
831 | ewitab = _mm_cvttps_epi32(ewrt); |
832 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
833 | ewitab = _mm_slli_epi32(ewitab,2); |
834 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
835 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
836 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
837 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
838 | _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); |
839 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
840 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
841 | velec = _mm_mul_ps(qq01,_mm_sub_ps(rinv01,velec)); |
842 | felec = _mm_mul_ps(_mm_mul_ps(qq01,rinv01),_mm_sub_ps(rinvsq01,felec)); |
843 | |
844 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
845 | velec = _mm_andnot_ps(dummy_mask,velec); |
846 | velecsum = _mm_add_ps(velecsum,velec); |
847 | |
848 | fscal = felec; |
849 | |
850 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
851 | |
852 | /* Calculate temporary vectorial force */ |
853 | tx = _mm_mul_ps(fscal,dx01); |
854 | ty = _mm_mul_ps(fscal,dy01); |
855 | tz = _mm_mul_ps(fscal,dz01); |
856 | |
857 | /* Update vectorial force */ |
858 | fix0 = _mm_add_ps(fix0,tx); |
859 | fiy0 = _mm_add_ps(fiy0,ty); |
860 | fiz0 = _mm_add_ps(fiz0,tz); |
861 | |
862 | fjx1 = _mm_add_ps(fjx1,tx); |
863 | fjy1 = _mm_add_ps(fjy1,ty); |
864 | fjz1 = _mm_add_ps(fjz1,tz); |
865 | |
866 | /************************** |
867 | * CALCULATE INTERACTIONS * |
868 | **************************/ |
869 | |
870 | r02 = _mm_mul_ps(rsq02,rinv02); |
871 | r02 = _mm_andnot_ps(dummy_mask,r02); |
872 | |
873 | /* EWALD ELECTROSTATICS */ |
874 | |
875 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
876 | ewrt = _mm_mul_ps(r02,ewtabscale); |
877 | ewitab = _mm_cvttps_epi32(ewrt); |
878 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
879 | ewitab = _mm_slli_epi32(ewitab,2); |
880 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
881 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
882 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
883 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
884 | _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); |
885 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
886 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
887 | velec = _mm_mul_ps(qq02,_mm_sub_ps(rinv02,velec)); |
888 | felec = _mm_mul_ps(_mm_mul_ps(qq02,rinv02),_mm_sub_ps(rinvsq02,felec)); |
889 | |
890 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
891 | velec = _mm_andnot_ps(dummy_mask,velec); |
892 | velecsum = _mm_add_ps(velecsum,velec); |
893 | |
894 | fscal = felec; |
895 | |
896 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
897 | |
898 | /* Calculate temporary vectorial force */ |
899 | tx = _mm_mul_ps(fscal,dx02); |
900 | ty = _mm_mul_ps(fscal,dy02); |
901 | tz = _mm_mul_ps(fscal,dz02); |
902 | |
903 | /* Update vectorial force */ |
904 | fix0 = _mm_add_ps(fix0,tx); |
905 | fiy0 = _mm_add_ps(fiy0,ty); |
906 | fiz0 = _mm_add_ps(fiz0,tz); |
907 | |
908 | fjx2 = _mm_add_ps(fjx2,tx); |
909 | fjy2 = _mm_add_ps(fjy2,ty); |
910 | fjz2 = _mm_add_ps(fjz2,tz); |
911 | |
912 | /************************** |
913 | * CALCULATE INTERACTIONS * |
914 | **************************/ |
915 | |
916 | r10 = _mm_mul_ps(rsq10,rinv10); |
917 | r10 = _mm_andnot_ps(dummy_mask,r10); |
918 | |
919 | /* EWALD ELECTROSTATICS */ |
920 | |
921 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
922 | ewrt = _mm_mul_ps(r10,ewtabscale); |
923 | ewitab = _mm_cvttps_epi32(ewrt); |
924 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
925 | ewitab = _mm_slli_epi32(ewitab,2); |
926 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
927 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
928 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
929 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
930 | _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); |
931 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
932 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
933 | velec = _mm_mul_ps(qq10,_mm_sub_ps(rinv10,velec)); |
934 | felec = _mm_mul_ps(_mm_mul_ps(qq10,rinv10),_mm_sub_ps(rinvsq10,felec)); |
935 | |
936 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
937 | velec = _mm_andnot_ps(dummy_mask,velec); |
938 | velecsum = _mm_add_ps(velecsum,velec); |
939 | |
940 | fscal = felec; |
941 | |
942 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
943 | |
944 | /* Calculate temporary vectorial force */ |
945 | tx = _mm_mul_ps(fscal,dx10); |
946 | ty = _mm_mul_ps(fscal,dy10); |
947 | tz = _mm_mul_ps(fscal,dz10); |
948 | |
949 | /* Update vectorial force */ |
950 | fix1 = _mm_add_ps(fix1,tx); |
951 | fiy1 = _mm_add_ps(fiy1,ty); |
952 | fiz1 = _mm_add_ps(fiz1,tz); |
953 | |
954 | fjx0 = _mm_add_ps(fjx0,tx); |
955 | fjy0 = _mm_add_ps(fjy0,ty); |
956 | fjz0 = _mm_add_ps(fjz0,tz); |
957 | |
958 | /************************** |
959 | * CALCULATE INTERACTIONS * |
960 | **************************/ |
961 | |
962 | r11 = _mm_mul_ps(rsq11,rinv11); |
963 | r11 = _mm_andnot_ps(dummy_mask,r11); |
964 | |
965 | /* EWALD ELECTROSTATICS */ |
966 | |
967 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
968 | ewrt = _mm_mul_ps(r11,ewtabscale); |
969 | ewitab = _mm_cvttps_epi32(ewrt); |
970 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
971 | ewitab = _mm_slli_epi32(ewitab,2); |
972 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
973 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
974 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
975 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
976 | _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); |
977 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
978 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
979 | velec = _mm_mul_ps(qq11,_mm_sub_ps(rinv11,velec)); |
980 | felec = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec)); |
981 | |
982 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
983 | velec = _mm_andnot_ps(dummy_mask,velec); |
984 | velecsum = _mm_add_ps(velecsum,velec); |
985 | |
986 | fscal = felec; |
987 | |
988 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
989 | |
990 | /* Calculate temporary vectorial force */ |
991 | tx = _mm_mul_ps(fscal,dx11); |
992 | ty = _mm_mul_ps(fscal,dy11); |
993 | tz = _mm_mul_ps(fscal,dz11); |
994 | |
995 | /* Update vectorial force */ |
996 | fix1 = _mm_add_ps(fix1,tx); |
997 | fiy1 = _mm_add_ps(fiy1,ty); |
998 | fiz1 = _mm_add_ps(fiz1,tz); |
999 | |
1000 | fjx1 = _mm_add_ps(fjx1,tx); |
1001 | fjy1 = _mm_add_ps(fjy1,ty); |
1002 | fjz1 = _mm_add_ps(fjz1,tz); |
1003 | |
1004 | /************************** |
1005 | * CALCULATE INTERACTIONS * |
1006 | **************************/ |
1007 | |
1008 | r12 = _mm_mul_ps(rsq12,rinv12); |
1009 | r12 = _mm_andnot_ps(dummy_mask,r12); |
1010 | |
1011 | /* EWALD ELECTROSTATICS */ |
1012 | |
1013 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1014 | ewrt = _mm_mul_ps(r12,ewtabscale); |
1015 | ewitab = _mm_cvttps_epi32(ewrt); |
1016 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1017 | ewitab = _mm_slli_epi32(ewitab,2); |
1018 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
1019 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
1020 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
1021 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
1022 | _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); |
1023 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
1024 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
1025 | velec = _mm_mul_ps(qq12,_mm_sub_ps(rinv12,velec)); |
1026 | felec = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec)); |
1027 | |
1028 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
1029 | velec = _mm_andnot_ps(dummy_mask,velec); |
1030 | velecsum = _mm_add_ps(velecsum,velec); |
1031 | |
1032 | fscal = felec; |
1033 | |
1034 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
1035 | |
1036 | /* Calculate temporary vectorial force */ |
1037 | tx = _mm_mul_ps(fscal,dx12); |
1038 | ty = _mm_mul_ps(fscal,dy12); |
1039 | tz = _mm_mul_ps(fscal,dz12); |
1040 | |
1041 | /* Update vectorial force */ |
1042 | fix1 = _mm_add_ps(fix1,tx); |
1043 | fiy1 = _mm_add_ps(fiy1,ty); |
1044 | fiz1 = _mm_add_ps(fiz1,tz); |
1045 | |
1046 | fjx2 = _mm_add_ps(fjx2,tx); |
1047 | fjy2 = _mm_add_ps(fjy2,ty); |
1048 | fjz2 = _mm_add_ps(fjz2,tz); |
1049 | |
1050 | /************************** |
1051 | * CALCULATE INTERACTIONS * |
1052 | **************************/ |
1053 | |
1054 | r20 = _mm_mul_ps(rsq20,rinv20); |
1055 | r20 = _mm_andnot_ps(dummy_mask,r20); |
1056 | |
1057 | /* EWALD ELECTROSTATICS */ |
1058 | |
1059 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1060 | ewrt = _mm_mul_ps(r20,ewtabscale); |
1061 | ewitab = _mm_cvttps_epi32(ewrt); |
1062 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1063 | ewitab = _mm_slli_epi32(ewitab,2); |
1064 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
1065 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
1066 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
1067 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
1068 | _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); |
1069 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
1070 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
1071 | velec = _mm_mul_ps(qq20,_mm_sub_ps(rinv20,velec)); |
1072 | felec = _mm_mul_ps(_mm_mul_ps(qq20,rinv20),_mm_sub_ps(rinvsq20,felec)); |
1073 | |
1074 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
1075 | velec = _mm_andnot_ps(dummy_mask,velec); |
1076 | velecsum = _mm_add_ps(velecsum,velec); |
1077 | |
1078 | fscal = felec; |
1079 | |
1080 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
1081 | |
1082 | /* Calculate temporary vectorial force */ |
1083 | tx = _mm_mul_ps(fscal,dx20); |
1084 | ty = _mm_mul_ps(fscal,dy20); |
1085 | tz = _mm_mul_ps(fscal,dz20); |
1086 | |
1087 | /* Update vectorial force */ |
1088 | fix2 = _mm_add_ps(fix2,tx); |
1089 | fiy2 = _mm_add_ps(fiy2,ty); |
1090 | fiz2 = _mm_add_ps(fiz2,tz); |
1091 | |
1092 | fjx0 = _mm_add_ps(fjx0,tx); |
1093 | fjy0 = _mm_add_ps(fjy0,ty); |
1094 | fjz0 = _mm_add_ps(fjz0,tz); |
1095 | |
1096 | /************************** |
1097 | * CALCULATE INTERACTIONS * |
1098 | **************************/ |
1099 | |
1100 | r21 = _mm_mul_ps(rsq21,rinv21); |
1101 | r21 = _mm_andnot_ps(dummy_mask,r21); |
1102 | |
1103 | /* EWALD ELECTROSTATICS */ |
1104 | |
1105 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1106 | ewrt = _mm_mul_ps(r21,ewtabscale); |
1107 | ewitab = _mm_cvttps_epi32(ewrt); |
1108 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1109 | ewitab = _mm_slli_epi32(ewitab,2); |
1110 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
1111 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
1112 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
1113 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
1114 | _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); |
1115 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
1116 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
1117 | velec = _mm_mul_ps(qq21,_mm_sub_ps(rinv21,velec)); |
1118 | felec = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec)); |
1119 | |
1120 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
1121 | velec = _mm_andnot_ps(dummy_mask,velec); |
1122 | velecsum = _mm_add_ps(velecsum,velec); |
1123 | |
1124 | fscal = felec; |
1125 | |
1126 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
1127 | |
1128 | /* Calculate temporary vectorial force */ |
1129 | tx = _mm_mul_ps(fscal,dx21); |
1130 | ty = _mm_mul_ps(fscal,dy21); |
1131 | tz = _mm_mul_ps(fscal,dz21); |
1132 | |
1133 | /* Update vectorial force */ |
1134 | fix2 = _mm_add_ps(fix2,tx); |
1135 | fiy2 = _mm_add_ps(fiy2,ty); |
1136 | fiz2 = _mm_add_ps(fiz2,tz); |
1137 | |
1138 | fjx1 = _mm_add_ps(fjx1,tx); |
1139 | fjy1 = _mm_add_ps(fjy1,ty); |
1140 | fjz1 = _mm_add_ps(fjz1,tz); |
1141 | |
1142 | /************************** |
1143 | * CALCULATE INTERACTIONS * |
1144 | **************************/ |
1145 | |
1146 | r22 = _mm_mul_ps(rsq22,rinv22); |
1147 | r22 = _mm_andnot_ps(dummy_mask,r22); |
1148 | |
1149 | /* EWALD ELECTROSTATICS */ |
1150 | |
1151 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1152 | ewrt = _mm_mul_ps(r22,ewtabscale); |
1153 | ewitab = _mm_cvttps_epi32(ewrt); |
1154 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1155 | ewitab = _mm_slli_epi32(ewitab,2); |
1156 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
1157 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
1158 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
1159 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
1160 | _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); |
1161 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
1162 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
1163 | velec = _mm_mul_ps(qq22,_mm_sub_ps(rinv22,velec)); |
1164 | felec = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec)); |
1165 | |
1166 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
1167 | velec = _mm_andnot_ps(dummy_mask,velec); |
1168 | velecsum = _mm_add_ps(velecsum,velec); |
1169 | |
1170 | fscal = felec; |
1171 | |
1172 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
1173 | |
1174 | /* Calculate temporary vectorial force */ |
1175 | tx = _mm_mul_ps(fscal,dx22); |
1176 | ty = _mm_mul_ps(fscal,dy22); |
1177 | tz = _mm_mul_ps(fscal,dz22); |
1178 | |
1179 | /* Update vectorial force */ |
1180 | fix2 = _mm_add_ps(fix2,tx); |
1181 | fiy2 = _mm_add_ps(fiy2,ty); |
1182 | fiz2 = _mm_add_ps(fiz2,tz); |
1183 | |
1184 | fjx2 = _mm_add_ps(fjx2,tx); |
1185 | fjy2 = _mm_add_ps(fjy2,ty); |
1186 | fjz2 = _mm_add_ps(fjz2,tz); |
1187 | |
1188 | fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch; |
1189 | fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch; |
1190 | fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch; |
1191 | fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch; |
1192 | |
1193 | gmx_mm_decrement_3rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD, |
1194 | fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2); |
1195 | |
1196 | /* Inner loop uses 378 flops */ |
1197 | } |
1198 | |
1199 | /* End of innermost loop */ |
1200 | |
1201 | gmx_mm_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2, |
1202 | f+i_coord_offset,fshift+i_shift_offset); |
1203 | |
1204 | ggid = gid[iidx]; |
1205 | /* Update potential energies */ |
1206 | gmx_mm_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid); |
1207 | |
1208 | /* Increment number of inner iterations */ |
1209 | inneriter += j_index_end - j_index_start; |
1210 | |
1211 | /* Outer loop uses 19 flops */ |
1212 | } |
1213 | |
1214 | /* Increment number of outer iterations */ |
1215 | outeriter += nri; |
1216 | |
1217 | /* Update outer/inner flops */ |
1218 | |
1219 | inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3W3_VF,outeriter*19 + inneriter*378)(nrnb)->n[eNR_NBKERNEL_ELEC_W3W3_VF] += outeriter*19 + inneriter *378; |
1220 | } |
1221 | /* |
1222 | * Gromacs nonbonded kernel: nb_kernel_ElecEw_VdwNone_GeomW3W3_F_sse4_1_single |
1223 | * Electrostatics interaction: Ewald |
1224 | * VdW interaction: None |
1225 | * Geometry: Water3-Water3 |
1226 | * Calculate force/pot: Force |
1227 | */ |
1228 | void |
1229 | nb_kernel_ElecEw_VdwNone_GeomW3W3_F_sse4_1_single |
1230 | (t_nblist * gmx_restrict nlist, |
1231 | rvec * gmx_restrict xx, |
1232 | rvec * gmx_restrict ff, |
1233 | t_forcerec * gmx_restrict fr, |
1234 | t_mdatoms * gmx_restrict mdatoms, |
1235 | nb_kernel_data_t gmx_unused__attribute__ ((unused)) * gmx_restrict kernel_data, |
1236 | t_nrnb * gmx_restrict nrnb) |
1237 | { |
1238 | /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or |
1239 | * just 0 for non-waters. |
1240 | * Suffixes A,B,C,D refer to j loop unrolling done with SSE, e.g. for the four different |
1241 | * jnr indices corresponding to data put in the four positions in the SIMD register. |
1242 | */ |
1243 | int i_shift_offset,i_coord_offset,outeriter,inneriter; |
1244 | int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx; |
1245 | int jnrA,jnrB,jnrC,jnrD; |
1246 | int jnrlistA,jnrlistB,jnrlistC,jnrlistD; |
1247 | int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD; |
1248 | int *iinr,*jindex,*jjnr,*shiftidx,*gid; |
1249 | real rcutoff_scalar; |
1250 | real *shiftvec,*fshift,*x,*f; |
1251 | real *fjptrA,*fjptrB,*fjptrC,*fjptrD; |
1252 | real scratch[4*DIM3]; |
1253 | __m128 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall; |
1254 | int vdwioffset0; |
1255 | __m128 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0; |
1256 | int vdwioffset1; |
1257 | __m128 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1; |
1258 | int vdwioffset2; |
1259 | __m128 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2; |
1260 | int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D; |
1261 | __m128 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0; |
1262 | int vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D; |
1263 | __m128 jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1; |
1264 | int vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D; |
1265 | __m128 jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2; |
1266 | __m128 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00; |
1267 | __m128 dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01; |
1268 | __m128 dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02; |
1269 | __m128 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10; |
1270 | __m128 dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11; |
1271 | __m128 dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12; |
1272 | __m128 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20; |
1273 | __m128 dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21; |
1274 | __m128 dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22; |
1275 | __m128 velec,felec,velecsum,facel,crf,krf,krf2; |
1276 | real *charge; |
1277 | __m128i ewitab; |
1278 | __m128 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV; |
1279 | real *ewtab; |
1280 | __m128 dummy_mask,cutoff_mask; |
1281 | __m128 signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) ); |
1282 | __m128 one = _mm_set1_ps(1.0); |
1283 | __m128 two = _mm_set1_ps(2.0); |
1284 | x = xx[0]; |
1285 | f = ff[0]; |
1286 | |
1287 | nri = nlist->nri; |
1288 | iinr = nlist->iinr; |
1289 | jindex = nlist->jindex; |
1290 | jjnr = nlist->jjnr; |
1291 | shiftidx = nlist->shift; |
1292 | gid = nlist->gid; |
1293 | shiftvec = fr->shift_vec[0]; |
1294 | fshift = fr->fshift[0]; |
1295 | facel = _mm_set1_ps(fr->epsfac); |
1296 | charge = mdatoms->chargeA; |
1297 | |
1298 | sh_ewald = _mm_set1_ps(fr->ic->sh_ewald); |
1299 | ewtab = fr->ic->tabq_coul_F; |
1300 | ewtabscale = _mm_set1_ps(fr->ic->tabq_scale); |
1301 | ewtabhalfspace = _mm_set1_ps(0.5/fr->ic->tabq_scale); |
1302 | |
1303 | /* Setup water-specific parameters */ |
1304 | inr = nlist->iinr[0]; |
1305 | iq0 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+0])); |
1306 | iq1 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1])); |
1307 | iq2 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2])); |
1308 | |
1309 | jq0 = _mm_set1_ps(charge[inr+0]); |
1310 | jq1 = _mm_set1_ps(charge[inr+1]); |
1311 | jq2 = _mm_set1_ps(charge[inr+2]); |
1312 | qq00 = _mm_mul_ps(iq0,jq0); |
1313 | qq01 = _mm_mul_ps(iq0,jq1); |
1314 | qq02 = _mm_mul_ps(iq0,jq2); |
1315 | qq10 = _mm_mul_ps(iq1,jq0); |
1316 | qq11 = _mm_mul_ps(iq1,jq1); |
1317 | qq12 = _mm_mul_ps(iq1,jq2); |
1318 | qq20 = _mm_mul_ps(iq2,jq0); |
1319 | qq21 = _mm_mul_ps(iq2,jq1); |
1320 | qq22 = _mm_mul_ps(iq2,jq2); |
1321 | |
1322 | /* Avoid stupid compiler warnings */ |
1323 | jnrA = jnrB = jnrC = jnrD = 0; |
1324 | j_coord_offsetA = 0; |
1325 | j_coord_offsetB = 0; |
1326 | j_coord_offsetC = 0; |
1327 | j_coord_offsetD = 0; |
1328 | |
1329 | outeriter = 0; |
1330 | inneriter = 0; |
1331 | |
1332 | for(iidx=0;iidx<4*DIM3;iidx++) |
1333 | { |
1334 | scratch[iidx] = 0.0; |
1335 | } |
1336 | |
1337 | /* Start outer loop over neighborlists */ |
1338 | for(iidx=0; iidx<nri; iidx++) |
1339 | { |
1340 | /* Load shift vector for this list */ |
1341 | i_shift_offset = DIM3*shiftidx[iidx]; |
1342 | |
1343 | /* Load limits for loop over neighbors */ |
1344 | j_index_start = jindex[iidx]; |
1345 | j_index_end = jindex[iidx+1]; |
1346 | |
1347 | /* Get outer coordinate index */ |
1348 | inr = iinr[iidx]; |
1349 | i_coord_offset = DIM3*inr; |
1350 | |
1351 | /* Load i particle coords and add shift vector */ |
1352 | gmx_mm_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset, |
1353 | &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2); |
1354 | |
1355 | fix0 = _mm_setzero_ps(); |
1356 | fiy0 = _mm_setzero_ps(); |
1357 | fiz0 = _mm_setzero_ps(); |
1358 | fix1 = _mm_setzero_ps(); |
1359 | fiy1 = _mm_setzero_ps(); |
1360 | fiz1 = _mm_setzero_ps(); |
1361 | fix2 = _mm_setzero_ps(); |
1362 | fiy2 = _mm_setzero_ps(); |
1363 | fiz2 = _mm_setzero_ps(); |
1364 | |
1365 | /* Start inner kernel loop */ |
1366 | for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4) |
1367 | { |
1368 | |
1369 | /* Get j neighbor index, and coordinate index */ |
1370 | jnrA = jjnr[jidx]; |
1371 | jnrB = jjnr[jidx+1]; |
1372 | jnrC = jjnr[jidx+2]; |
1373 | jnrD = jjnr[jidx+3]; |
1374 | j_coord_offsetA = DIM3*jnrA; |
1375 | j_coord_offsetB = DIM3*jnrB; |
1376 | j_coord_offsetC = DIM3*jnrC; |
1377 | j_coord_offsetD = DIM3*jnrD; |
1378 | |
1379 | /* load j atom coordinates */ |
1380 | gmx_mm_load_3rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB, |
1381 | x+j_coord_offsetC,x+j_coord_offsetD, |
1382 | &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2); |
1383 | |
1384 | /* Calculate displacement vector */ |
1385 | dx00 = _mm_sub_ps(ix0,jx0); |
1386 | dy00 = _mm_sub_ps(iy0,jy0); |
1387 | dz00 = _mm_sub_ps(iz0,jz0); |
1388 | dx01 = _mm_sub_ps(ix0,jx1); |
1389 | dy01 = _mm_sub_ps(iy0,jy1); |
1390 | dz01 = _mm_sub_ps(iz0,jz1); |
1391 | dx02 = _mm_sub_ps(ix0,jx2); |
1392 | dy02 = _mm_sub_ps(iy0,jy2); |
1393 | dz02 = _mm_sub_ps(iz0,jz2); |
1394 | dx10 = _mm_sub_ps(ix1,jx0); |
1395 | dy10 = _mm_sub_ps(iy1,jy0); |
1396 | dz10 = _mm_sub_ps(iz1,jz0); |
1397 | dx11 = _mm_sub_ps(ix1,jx1); |
1398 | dy11 = _mm_sub_ps(iy1,jy1); |
1399 | dz11 = _mm_sub_ps(iz1,jz1); |
1400 | dx12 = _mm_sub_ps(ix1,jx2); |
1401 | dy12 = _mm_sub_ps(iy1,jy2); |
1402 | dz12 = _mm_sub_ps(iz1,jz2); |
1403 | dx20 = _mm_sub_ps(ix2,jx0); |
1404 | dy20 = _mm_sub_ps(iy2,jy0); |
1405 | dz20 = _mm_sub_ps(iz2,jz0); |
1406 | dx21 = _mm_sub_ps(ix2,jx1); |
1407 | dy21 = _mm_sub_ps(iy2,jy1); |
1408 | dz21 = _mm_sub_ps(iz2,jz1); |
1409 | dx22 = _mm_sub_ps(ix2,jx2); |
1410 | dy22 = _mm_sub_ps(iy2,jy2); |
1411 | dz22 = _mm_sub_ps(iz2,jz2); |
1412 | |
1413 | /* Calculate squared distance and things based on it */ |
1414 | rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00); |
1415 | rsq01 = gmx_mm_calc_rsq_ps(dx01,dy01,dz01); |
1416 | rsq02 = gmx_mm_calc_rsq_ps(dx02,dy02,dz02); |
1417 | rsq10 = gmx_mm_calc_rsq_ps(dx10,dy10,dz10); |
1418 | rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11); |
1419 | rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12); |
1420 | rsq20 = gmx_mm_calc_rsq_ps(dx20,dy20,dz20); |
1421 | rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21); |
1422 | rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22); |
1423 | |
1424 | rinv00 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq00); |
1425 | rinv01 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq01); |
1426 | rinv02 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq02); |
1427 | rinv10 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq10); |
1428 | rinv11 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq11); |
1429 | rinv12 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq12); |
1430 | rinv20 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq20); |
1431 | rinv21 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq21); |
1432 | rinv22 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq22); |
1433 | |
1434 | rinvsq00 = _mm_mul_ps(rinv00,rinv00); |
1435 | rinvsq01 = _mm_mul_ps(rinv01,rinv01); |
1436 | rinvsq02 = _mm_mul_ps(rinv02,rinv02); |
1437 | rinvsq10 = _mm_mul_ps(rinv10,rinv10); |
1438 | rinvsq11 = _mm_mul_ps(rinv11,rinv11); |
1439 | rinvsq12 = _mm_mul_ps(rinv12,rinv12); |
1440 | rinvsq20 = _mm_mul_ps(rinv20,rinv20); |
1441 | rinvsq21 = _mm_mul_ps(rinv21,rinv21); |
1442 | rinvsq22 = _mm_mul_ps(rinv22,rinv22); |
1443 | |
1444 | fjx0 = _mm_setzero_ps(); |
1445 | fjy0 = _mm_setzero_ps(); |
1446 | fjz0 = _mm_setzero_ps(); |
1447 | fjx1 = _mm_setzero_ps(); |
1448 | fjy1 = _mm_setzero_ps(); |
1449 | fjz1 = _mm_setzero_ps(); |
1450 | fjx2 = _mm_setzero_ps(); |
1451 | fjy2 = _mm_setzero_ps(); |
1452 | fjz2 = _mm_setzero_ps(); |
1453 | |
1454 | /************************** |
1455 | * CALCULATE INTERACTIONS * |
1456 | **************************/ |
1457 | |
1458 | r00 = _mm_mul_ps(rsq00,rinv00); |
1459 | |
1460 | /* EWALD ELECTROSTATICS */ |
1461 | |
1462 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1463 | ewrt = _mm_mul_ps(r00,ewtabscale); |
1464 | ewitab = _mm_cvttps_epi32(ewrt); |
1465 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1466 | 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];})), |
1467 | 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];})), |
1468 | &ewtabF,&ewtabFn); |
1469 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
1470 | felec = _mm_mul_ps(_mm_mul_ps(qq00,rinv00),_mm_sub_ps(rinvsq00,felec)); |
1471 | |
1472 | fscal = felec; |
1473 | |
1474 | /* Calculate temporary vectorial force */ |
1475 | tx = _mm_mul_ps(fscal,dx00); |
1476 | ty = _mm_mul_ps(fscal,dy00); |
1477 | tz = _mm_mul_ps(fscal,dz00); |
1478 | |
1479 | /* Update vectorial force */ |
1480 | fix0 = _mm_add_ps(fix0,tx); |
1481 | fiy0 = _mm_add_ps(fiy0,ty); |
1482 | fiz0 = _mm_add_ps(fiz0,tz); |
1483 | |
1484 | fjx0 = _mm_add_ps(fjx0,tx); |
1485 | fjy0 = _mm_add_ps(fjy0,ty); |
1486 | fjz0 = _mm_add_ps(fjz0,tz); |
1487 | |
1488 | /************************** |
1489 | * CALCULATE INTERACTIONS * |
1490 | **************************/ |
1491 | |
1492 | r01 = _mm_mul_ps(rsq01,rinv01); |
1493 | |
1494 | /* EWALD ELECTROSTATICS */ |
1495 | |
1496 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1497 | ewrt = _mm_mul_ps(r01,ewtabscale); |
1498 | ewitab = _mm_cvttps_epi32(ewrt); |
1499 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1500 | 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];})), |
1501 | 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];})), |
1502 | &ewtabF,&ewtabFn); |
1503 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
1504 | felec = _mm_mul_ps(_mm_mul_ps(qq01,rinv01),_mm_sub_ps(rinvsq01,felec)); |
1505 | |
1506 | fscal = felec; |
1507 | |
1508 | /* Calculate temporary vectorial force */ |
1509 | tx = _mm_mul_ps(fscal,dx01); |
1510 | ty = _mm_mul_ps(fscal,dy01); |
1511 | tz = _mm_mul_ps(fscal,dz01); |
1512 | |
1513 | /* Update vectorial force */ |
1514 | fix0 = _mm_add_ps(fix0,tx); |
1515 | fiy0 = _mm_add_ps(fiy0,ty); |
1516 | fiz0 = _mm_add_ps(fiz0,tz); |
1517 | |
1518 | fjx1 = _mm_add_ps(fjx1,tx); |
1519 | fjy1 = _mm_add_ps(fjy1,ty); |
1520 | fjz1 = _mm_add_ps(fjz1,tz); |
1521 | |
1522 | /************************** |
1523 | * CALCULATE INTERACTIONS * |
1524 | **************************/ |
1525 | |
1526 | r02 = _mm_mul_ps(rsq02,rinv02); |
1527 | |
1528 | /* EWALD ELECTROSTATICS */ |
1529 | |
1530 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1531 | ewrt = _mm_mul_ps(r02,ewtabscale); |
1532 | ewitab = _mm_cvttps_epi32(ewrt); |
1533 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1534 | 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];})), |
1535 | 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];})), |
1536 | &ewtabF,&ewtabFn); |
1537 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
1538 | felec = _mm_mul_ps(_mm_mul_ps(qq02,rinv02),_mm_sub_ps(rinvsq02,felec)); |
1539 | |
1540 | fscal = felec; |
1541 | |
1542 | /* Calculate temporary vectorial force */ |
1543 | tx = _mm_mul_ps(fscal,dx02); |
1544 | ty = _mm_mul_ps(fscal,dy02); |
1545 | tz = _mm_mul_ps(fscal,dz02); |
1546 | |
1547 | /* Update vectorial force */ |
1548 | fix0 = _mm_add_ps(fix0,tx); |
1549 | fiy0 = _mm_add_ps(fiy0,ty); |
1550 | fiz0 = _mm_add_ps(fiz0,tz); |
1551 | |
1552 | fjx2 = _mm_add_ps(fjx2,tx); |
1553 | fjy2 = _mm_add_ps(fjy2,ty); |
1554 | fjz2 = _mm_add_ps(fjz2,tz); |
1555 | |
1556 | /************************** |
1557 | * CALCULATE INTERACTIONS * |
1558 | **************************/ |
1559 | |
1560 | r10 = _mm_mul_ps(rsq10,rinv10); |
1561 | |
1562 | /* EWALD ELECTROSTATICS */ |
1563 | |
1564 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1565 | ewrt = _mm_mul_ps(r10,ewtabscale); |
1566 | ewitab = _mm_cvttps_epi32(ewrt); |
1567 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1568 | 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];})), |
1569 | 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];})), |
1570 | &ewtabF,&ewtabFn); |
1571 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
1572 | felec = _mm_mul_ps(_mm_mul_ps(qq10,rinv10),_mm_sub_ps(rinvsq10,felec)); |
1573 | |
1574 | fscal = felec; |
1575 | |
1576 | /* Calculate temporary vectorial force */ |
1577 | tx = _mm_mul_ps(fscal,dx10); |
1578 | ty = _mm_mul_ps(fscal,dy10); |
1579 | tz = _mm_mul_ps(fscal,dz10); |
1580 | |
1581 | /* Update vectorial force */ |
1582 | fix1 = _mm_add_ps(fix1,tx); |
1583 | fiy1 = _mm_add_ps(fiy1,ty); |
1584 | fiz1 = _mm_add_ps(fiz1,tz); |
1585 | |
1586 | fjx0 = _mm_add_ps(fjx0,tx); |
1587 | fjy0 = _mm_add_ps(fjy0,ty); |
1588 | fjz0 = _mm_add_ps(fjz0,tz); |
1589 | |
1590 | /************************** |
1591 | * CALCULATE INTERACTIONS * |
1592 | **************************/ |
1593 | |
1594 | r11 = _mm_mul_ps(rsq11,rinv11); |
1595 | |
1596 | /* EWALD ELECTROSTATICS */ |
1597 | |
1598 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1599 | ewrt = _mm_mul_ps(r11,ewtabscale); |
1600 | ewitab = _mm_cvttps_epi32(ewrt); |
1601 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1602 | 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];})), |
1603 | 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];})), |
1604 | &ewtabF,&ewtabFn); |
1605 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
1606 | felec = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec)); |
1607 | |
1608 | fscal = felec; |
1609 | |
1610 | /* Calculate temporary vectorial force */ |
1611 | tx = _mm_mul_ps(fscal,dx11); |
1612 | ty = _mm_mul_ps(fscal,dy11); |
1613 | tz = _mm_mul_ps(fscal,dz11); |
1614 | |
1615 | /* Update vectorial force */ |
1616 | fix1 = _mm_add_ps(fix1,tx); |
1617 | fiy1 = _mm_add_ps(fiy1,ty); |
1618 | fiz1 = _mm_add_ps(fiz1,tz); |
1619 | |
1620 | fjx1 = _mm_add_ps(fjx1,tx); |
1621 | fjy1 = _mm_add_ps(fjy1,ty); |
1622 | fjz1 = _mm_add_ps(fjz1,tz); |
1623 | |
1624 | /************************** |
1625 | * CALCULATE INTERACTIONS * |
1626 | **************************/ |
1627 | |
1628 | r12 = _mm_mul_ps(rsq12,rinv12); |
1629 | |
1630 | /* EWALD ELECTROSTATICS */ |
1631 | |
1632 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1633 | ewrt = _mm_mul_ps(r12,ewtabscale); |
1634 | ewitab = _mm_cvttps_epi32(ewrt); |
1635 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1636 | 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];})), |
1637 | 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];})), |
1638 | &ewtabF,&ewtabFn); |
1639 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
1640 | felec = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec)); |
1641 | |
1642 | fscal = felec; |
1643 | |
1644 | /* Calculate temporary vectorial force */ |
1645 | tx = _mm_mul_ps(fscal,dx12); |
1646 | ty = _mm_mul_ps(fscal,dy12); |
1647 | tz = _mm_mul_ps(fscal,dz12); |
1648 | |
1649 | /* Update vectorial force */ |
1650 | fix1 = _mm_add_ps(fix1,tx); |
1651 | fiy1 = _mm_add_ps(fiy1,ty); |
1652 | fiz1 = _mm_add_ps(fiz1,tz); |
1653 | |
1654 | fjx2 = _mm_add_ps(fjx2,tx); |
1655 | fjy2 = _mm_add_ps(fjy2,ty); |
1656 | fjz2 = _mm_add_ps(fjz2,tz); |
1657 | |
1658 | /************************** |
1659 | * CALCULATE INTERACTIONS * |
1660 | **************************/ |
1661 | |
1662 | r20 = _mm_mul_ps(rsq20,rinv20); |
1663 | |
1664 | /* EWALD ELECTROSTATICS */ |
1665 | |
1666 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1667 | ewrt = _mm_mul_ps(r20,ewtabscale); |
1668 | ewitab = _mm_cvttps_epi32(ewrt); |
1669 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1670 | 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];})), |
1671 | 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];})), |
1672 | &ewtabF,&ewtabFn); |
1673 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
1674 | felec = _mm_mul_ps(_mm_mul_ps(qq20,rinv20),_mm_sub_ps(rinvsq20,felec)); |
1675 | |
1676 | fscal = felec; |
1677 | |
1678 | /* Calculate temporary vectorial force */ |
1679 | tx = _mm_mul_ps(fscal,dx20); |
1680 | ty = _mm_mul_ps(fscal,dy20); |
1681 | tz = _mm_mul_ps(fscal,dz20); |
1682 | |
1683 | /* Update vectorial force */ |
1684 | fix2 = _mm_add_ps(fix2,tx); |
1685 | fiy2 = _mm_add_ps(fiy2,ty); |
1686 | fiz2 = _mm_add_ps(fiz2,tz); |
1687 | |
1688 | fjx0 = _mm_add_ps(fjx0,tx); |
1689 | fjy0 = _mm_add_ps(fjy0,ty); |
1690 | fjz0 = _mm_add_ps(fjz0,tz); |
1691 | |
1692 | /************************** |
1693 | * CALCULATE INTERACTIONS * |
1694 | **************************/ |
1695 | |
1696 | r21 = _mm_mul_ps(rsq21,rinv21); |
1697 | |
1698 | /* EWALD ELECTROSTATICS */ |
1699 | |
1700 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1701 | ewrt = _mm_mul_ps(r21,ewtabscale); |
1702 | ewitab = _mm_cvttps_epi32(ewrt); |
1703 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1704 | 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];})), |
1705 | 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];})), |
1706 | &ewtabF,&ewtabFn); |
1707 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
1708 | felec = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec)); |
1709 | |
1710 | fscal = felec; |
1711 | |
1712 | /* Calculate temporary vectorial force */ |
1713 | tx = _mm_mul_ps(fscal,dx21); |
1714 | ty = _mm_mul_ps(fscal,dy21); |
1715 | tz = _mm_mul_ps(fscal,dz21); |
1716 | |
1717 | /* Update vectorial force */ |
1718 | fix2 = _mm_add_ps(fix2,tx); |
1719 | fiy2 = _mm_add_ps(fiy2,ty); |
1720 | fiz2 = _mm_add_ps(fiz2,tz); |
1721 | |
1722 | fjx1 = _mm_add_ps(fjx1,tx); |
1723 | fjy1 = _mm_add_ps(fjy1,ty); |
1724 | fjz1 = _mm_add_ps(fjz1,tz); |
1725 | |
1726 | /************************** |
1727 | * CALCULATE INTERACTIONS * |
1728 | **************************/ |
1729 | |
1730 | r22 = _mm_mul_ps(rsq22,rinv22); |
1731 | |
1732 | /* EWALD ELECTROSTATICS */ |
1733 | |
1734 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1735 | ewrt = _mm_mul_ps(r22,ewtabscale); |
1736 | ewitab = _mm_cvttps_epi32(ewrt); |
1737 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1738 | 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];})), |
1739 | 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];})), |
1740 | &ewtabF,&ewtabFn); |
1741 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
1742 | felec = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec)); |
1743 | |
1744 | fscal = felec; |
1745 | |
1746 | /* Calculate temporary vectorial force */ |
1747 | tx = _mm_mul_ps(fscal,dx22); |
1748 | ty = _mm_mul_ps(fscal,dy22); |
1749 | tz = _mm_mul_ps(fscal,dz22); |
1750 | |
1751 | /* Update vectorial force */ |
1752 | fix2 = _mm_add_ps(fix2,tx); |
1753 | fiy2 = _mm_add_ps(fiy2,ty); |
1754 | fiz2 = _mm_add_ps(fiz2,tz); |
1755 | |
1756 | fjx2 = _mm_add_ps(fjx2,tx); |
1757 | fjy2 = _mm_add_ps(fjy2,ty); |
1758 | fjz2 = _mm_add_ps(fjz2,tz); |
1759 | |
1760 | fjptrA = f+j_coord_offsetA; |
1761 | fjptrB = f+j_coord_offsetB; |
1762 | fjptrC = f+j_coord_offsetC; |
1763 | fjptrD = f+j_coord_offsetD; |
1764 | |
1765 | gmx_mm_decrement_3rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD, |
1766 | fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2); |
1767 | |
1768 | /* Inner loop uses 324 flops */ |
1769 | } |
1770 | |
1771 | if(jidx<j_index_end) |
1772 | { |
1773 | |
1774 | /* Get j neighbor index, and coordinate index */ |
1775 | jnrlistA = jjnr[jidx]; |
1776 | jnrlistB = jjnr[jidx+1]; |
1777 | jnrlistC = jjnr[jidx+2]; |
1778 | jnrlistD = jjnr[jidx+3]; |
1779 | /* Sign of each element will be negative for non-real atoms. |
1780 | * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones, |
1781 | * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries. |
1782 | */ |
1783 | dummy_mask = gmx_mm_castsi128_ps_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())); |
1784 | jnrA = (jnrlistA>=0) ? jnrlistA : 0; |
1785 | jnrB = (jnrlistB>=0) ? jnrlistB : 0; |
1786 | jnrC = (jnrlistC>=0) ? jnrlistC : 0; |
1787 | jnrD = (jnrlistD>=0) ? jnrlistD : 0; |
1788 | j_coord_offsetA = DIM3*jnrA; |
1789 | j_coord_offsetB = DIM3*jnrB; |
1790 | j_coord_offsetC = DIM3*jnrC; |
1791 | j_coord_offsetD = DIM3*jnrD; |
1792 | |
1793 | /* load j atom coordinates */ |
1794 | gmx_mm_load_3rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB, |
1795 | x+j_coord_offsetC,x+j_coord_offsetD, |
1796 | &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2); |
1797 | |
1798 | /* Calculate displacement vector */ |
1799 | dx00 = _mm_sub_ps(ix0,jx0); |
1800 | dy00 = _mm_sub_ps(iy0,jy0); |
1801 | dz00 = _mm_sub_ps(iz0,jz0); |
1802 | dx01 = _mm_sub_ps(ix0,jx1); |
1803 | dy01 = _mm_sub_ps(iy0,jy1); |
1804 | dz01 = _mm_sub_ps(iz0,jz1); |
1805 | dx02 = _mm_sub_ps(ix0,jx2); |
1806 | dy02 = _mm_sub_ps(iy0,jy2); |
1807 | dz02 = _mm_sub_ps(iz0,jz2); |
1808 | dx10 = _mm_sub_ps(ix1,jx0); |
1809 | dy10 = _mm_sub_ps(iy1,jy0); |
1810 | dz10 = _mm_sub_ps(iz1,jz0); |
1811 | dx11 = _mm_sub_ps(ix1,jx1); |
1812 | dy11 = _mm_sub_ps(iy1,jy1); |
1813 | dz11 = _mm_sub_ps(iz1,jz1); |
1814 | dx12 = _mm_sub_ps(ix1,jx2); |
1815 | dy12 = _mm_sub_ps(iy1,jy2); |
1816 | dz12 = _mm_sub_ps(iz1,jz2); |
1817 | dx20 = _mm_sub_ps(ix2,jx0); |
1818 | dy20 = _mm_sub_ps(iy2,jy0); |
1819 | dz20 = _mm_sub_ps(iz2,jz0); |
1820 | dx21 = _mm_sub_ps(ix2,jx1); |
1821 | dy21 = _mm_sub_ps(iy2,jy1); |
1822 | dz21 = _mm_sub_ps(iz2,jz1); |
1823 | dx22 = _mm_sub_ps(ix2,jx2); |
1824 | dy22 = _mm_sub_ps(iy2,jy2); |
1825 | dz22 = _mm_sub_ps(iz2,jz2); |
1826 | |
1827 | /* Calculate squared distance and things based on it */ |
1828 | rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00); |
1829 | rsq01 = gmx_mm_calc_rsq_ps(dx01,dy01,dz01); |
1830 | rsq02 = gmx_mm_calc_rsq_ps(dx02,dy02,dz02); |
1831 | rsq10 = gmx_mm_calc_rsq_ps(dx10,dy10,dz10); |
1832 | rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11); |
1833 | rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12); |
1834 | rsq20 = gmx_mm_calc_rsq_ps(dx20,dy20,dz20); |
1835 | rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21); |
1836 | rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22); |
1837 | |
1838 | rinv00 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq00); |
1839 | rinv01 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq01); |
1840 | rinv02 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq02); |
1841 | rinv10 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq10); |
1842 | rinv11 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq11); |
1843 | rinv12 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq12); |
1844 | rinv20 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq20); |
1845 | rinv21 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq21); |
1846 | rinv22 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq22); |
1847 | |
1848 | rinvsq00 = _mm_mul_ps(rinv00,rinv00); |
1849 | rinvsq01 = _mm_mul_ps(rinv01,rinv01); |
1850 | rinvsq02 = _mm_mul_ps(rinv02,rinv02); |
1851 | rinvsq10 = _mm_mul_ps(rinv10,rinv10); |
1852 | rinvsq11 = _mm_mul_ps(rinv11,rinv11); |
1853 | rinvsq12 = _mm_mul_ps(rinv12,rinv12); |
1854 | rinvsq20 = _mm_mul_ps(rinv20,rinv20); |
1855 | rinvsq21 = _mm_mul_ps(rinv21,rinv21); |
1856 | rinvsq22 = _mm_mul_ps(rinv22,rinv22); |
1857 | |
1858 | fjx0 = _mm_setzero_ps(); |
1859 | fjy0 = _mm_setzero_ps(); |
1860 | fjz0 = _mm_setzero_ps(); |
1861 | fjx1 = _mm_setzero_ps(); |
1862 | fjy1 = _mm_setzero_ps(); |
1863 | fjz1 = _mm_setzero_ps(); |
1864 | fjx2 = _mm_setzero_ps(); |
1865 | fjy2 = _mm_setzero_ps(); |
1866 | fjz2 = _mm_setzero_ps(); |
1867 | |
1868 | /************************** |
1869 | * CALCULATE INTERACTIONS * |
1870 | **************************/ |
1871 | |
1872 | r00 = _mm_mul_ps(rsq00,rinv00); |
1873 | r00 = _mm_andnot_ps(dummy_mask,r00); |
1874 | |
1875 | /* EWALD ELECTROSTATICS */ |
1876 | |
1877 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1878 | ewrt = _mm_mul_ps(r00,ewtabscale); |
1879 | ewitab = _mm_cvttps_epi32(ewrt); |
1880 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1881 | 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];})), |
1882 | 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];})), |
1883 | &ewtabF,&ewtabFn); |
1884 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
1885 | felec = _mm_mul_ps(_mm_mul_ps(qq00,rinv00),_mm_sub_ps(rinvsq00,felec)); |
1886 | |
1887 | fscal = felec; |
1888 | |
1889 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
1890 | |
1891 | /* Calculate temporary vectorial force */ |
1892 | tx = _mm_mul_ps(fscal,dx00); |
1893 | ty = _mm_mul_ps(fscal,dy00); |
1894 | tz = _mm_mul_ps(fscal,dz00); |
1895 | |
1896 | /* Update vectorial force */ |
1897 | fix0 = _mm_add_ps(fix0,tx); |
1898 | fiy0 = _mm_add_ps(fiy0,ty); |
1899 | fiz0 = _mm_add_ps(fiz0,tz); |
1900 | |
1901 | fjx0 = _mm_add_ps(fjx0,tx); |
1902 | fjy0 = _mm_add_ps(fjy0,ty); |
1903 | fjz0 = _mm_add_ps(fjz0,tz); |
1904 | |
1905 | /************************** |
1906 | * CALCULATE INTERACTIONS * |
1907 | **************************/ |
1908 | |
1909 | r01 = _mm_mul_ps(rsq01,rinv01); |
1910 | r01 = _mm_andnot_ps(dummy_mask,r01); |
1911 | |
1912 | /* EWALD ELECTROSTATICS */ |
1913 | |
1914 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1915 | ewrt = _mm_mul_ps(r01,ewtabscale); |
1916 | ewitab = _mm_cvttps_epi32(ewrt); |
1917 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1918 | 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];})), |
1919 | 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];})), |
1920 | &ewtabF,&ewtabFn); |
1921 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
1922 | felec = _mm_mul_ps(_mm_mul_ps(qq01,rinv01),_mm_sub_ps(rinvsq01,felec)); |
1923 | |
1924 | fscal = felec; |
1925 | |
1926 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
1927 | |
1928 | /* Calculate temporary vectorial force */ |
1929 | tx = _mm_mul_ps(fscal,dx01); |
1930 | ty = _mm_mul_ps(fscal,dy01); |
1931 | tz = _mm_mul_ps(fscal,dz01); |
1932 | |
1933 | /* Update vectorial force */ |
1934 | fix0 = _mm_add_ps(fix0,tx); |
1935 | fiy0 = _mm_add_ps(fiy0,ty); |
1936 | fiz0 = _mm_add_ps(fiz0,tz); |
1937 | |
1938 | fjx1 = _mm_add_ps(fjx1,tx); |
1939 | fjy1 = _mm_add_ps(fjy1,ty); |
1940 | fjz1 = _mm_add_ps(fjz1,tz); |
1941 | |
1942 | /************************** |
1943 | * CALCULATE INTERACTIONS * |
1944 | **************************/ |
1945 | |
1946 | r02 = _mm_mul_ps(rsq02,rinv02); |
1947 | r02 = _mm_andnot_ps(dummy_mask,r02); |
1948 | |
1949 | /* EWALD ELECTROSTATICS */ |
1950 | |
1951 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1952 | ewrt = _mm_mul_ps(r02,ewtabscale); |
1953 | ewitab = _mm_cvttps_epi32(ewrt); |
1954 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1955 | 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];})), |
1956 | 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];})), |
1957 | &ewtabF,&ewtabFn); |
1958 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
1959 | felec = _mm_mul_ps(_mm_mul_ps(qq02,rinv02),_mm_sub_ps(rinvsq02,felec)); |
1960 | |
1961 | fscal = felec; |
1962 | |
1963 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
1964 | |
1965 | /* Calculate temporary vectorial force */ |
1966 | tx = _mm_mul_ps(fscal,dx02); |
1967 | ty = _mm_mul_ps(fscal,dy02); |
1968 | tz = _mm_mul_ps(fscal,dz02); |
1969 | |
1970 | /* Update vectorial force */ |
1971 | fix0 = _mm_add_ps(fix0,tx); |
1972 | fiy0 = _mm_add_ps(fiy0,ty); |
1973 | fiz0 = _mm_add_ps(fiz0,tz); |
1974 | |
1975 | fjx2 = _mm_add_ps(fjx2,tx); |
1976 | fjy2 = _mm_add_ps(fjy2,ty); |
1977 | fjz2 = _mm_add_ps(fjz2,tz); |
1978 | |
1979 | /************************** |
1980 | * CALCULATE INTERACTIONS * |
1981 | **************************/ |
1982 | |
1983 | r10 = _mm_mul_ps(rsq10,rinv10); |
1984 | r10 = _mm_andnot_ps(dummy_mask,r10); |
1985 | |
1986 | /* EWALD ELECTROSTATICS */ |
1987 | |
1988 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1989 | ewrt = _mm_mul_ps(r10,ewtabscale); |
1990 | ewitab = _mm_cvttps_epi32(ewrt); |
1991 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1992 | 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];})), |
1993 | 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];})), |
1994 | &ewtabF,&ewtabFn); |
1995 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
1996 | felec = _mm_mul_ps(_mm_mul_ps(qq10,rinv10),_mm_sub_ps(rinvsq10,felec)); |
1997 | |
1998 | fscal = felec; |
1999 | |
2000 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
2001 | |
2002 | /* Calculate temporary vectorial force */ |
2003 | tx = _mm_mul_ps(fscal,dx10); |
2004 | ty = _mm_mul_ps(fscal,dy10); |
2005 | tz = _mm_mul_ps(fscal,dz10); |
2006 | |
2007 | /* Update vectorial force */ |
2008 | fix1 = _mm_add_ps(fix1,tx); |
2009 | fiy1 = _mm_add_ps(fiy1,ty); |
2010 | fiz1 = _mm_add_ps(fiz1,tz); |
2011 | |
2012 | fjx0 = _mm_add_ps(fjx0,tx); |
2013 | fjy0 = _mm_add_ps(fjy0,ty); |
2014 | fjz0 = _mm_add_ps(fjz0,tz); |
2015 | |
2016 | /************************** |
2017 | * CALCULATE INTERACTIONS * |
2018 | **************************/ |
2019 | |
2020 | r11 = _mm_mul_ps(rsq11,rinv11); |
2021 | r11 = _mm_andnot_ps(dummy_mask,r11); |
2022 | |
2023 | /* EWALD ELECTROSTATICS */ |
2024 | |
2025 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
2026 | ewrt = _mm_mul_ps(r11,ewtabscale); |
2027 | ewitab = _mm_cvttps_epi32(ewrt); |
2028 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
2029 | 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];})), |
2030 | 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];})), |
2031 | &ewtabF,&ewtabFn); |
2032 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
2033 | felec = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec)); |
2034 | |
2035 | fscal = felec; |
2036 | |
2037 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
2038 | |
2039 | /* Calculate temporary vectorial force */ |
2040 | tx = _mm_mul_ps(fscal,dx11); |
2041 | ty = _mm_mul_ps(fscal,dy11); |
2042 | tz = _mm_mul_ps(fscal,dz11); |
2043 | |
2044 | /* Update vectorial force */ |
2045 | fix1 = _mm_add_ps(fix1,tx); |
2046 | fiy1 = _mm_add_ps(fiy1,ty); |
2047 | fiz1 = _mm_add_ps(fiz1,tz); |
2048 | |
2049 | fjx1 = _mm_add_ps(fjx1,tx); |
2050 | fjy1 = _mm_add_ps(fjy1,ty); |
2051 | fjz1 = _mm_add_ps(fjz1,tz); |
2052 | |
2053 | /************************** |
2054 | * CALCULATE INTERACTIONS * |
2055 | **************************/ |
2056 | |
2057 | r12 = _mm_mul_ps(rsq12,rinv12); |
2058 | r12 = _mm_andnot_ps(dummy_mask,r12); |
2059 | |
2060 | /* EWALD ELECTROSTATICS */ |
2061 | |
2062 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
2063 | ewrt = _mm_mul_ps(r12,ewtabscale); |
2064 | ewitab = _mm_cvttps_epi32(ewrt); |
2065 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
2066 | 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];})), |
2067 | 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];})), |
2068 | &ewtabF,&ewtabFn); |
2069 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
2070 | felec = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec)); |
2071 | |
2072 | fscal = felec; |
2073 | |
2074 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
2075 | |
2076 | /* Calculate temporary vectorial force */ |
2077 | tx = _mm_mul_ps(fscal,dx12); |
2078 | ty = _mm_mul_ps(fscal,dy12); |
2079 | tz = _mm_mul_ps(fscal,dz12); |
2080 | |
2081 | /* Update vectorial force */ |
2082 | fix1 = _mm_add_ps(fix1,tx); |
2083 | fiy1 = _mm_add_ps(fiy1,ty); |
2084 | fiz1 = _mm_add_ps(fiz1,tz); |
2085 | |
2086 | fjx2 = _mm_add_ps(fjx2,tx); |
2087 | fjy2 = _mm_add_ps(fjy2,ty); |
2088 | fjz2 = _mm_add_ps(fjz2,tz); |
2089 | |
2090 | /************************** |
2091 | * CALCULATE INTERACTIONS * |
2092 | **************************/ |
2093 | |
2094 | r20 = _mm_mul_ps(rsq20,rinv20); |
2095 | r20 = _mm_andnot_ps(dummy_mask,r20); |
2096 | |
2097 | /* EWALD ELECTROSTATICS */ |
2098 | |
2099 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
2100 | ewrt = _mm_mul_ps(r20,ewtabscale); |
2101 | ewitab = _mm_cvttps_epi32(ewrt); |
2102 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
2103 | 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];})), |
2104 | 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];})), |
2105 | &ewtabF,&ewtabFn); |
2106 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
2107 | felec = _mm_mul_ps(_mm_mul_ps(qq20,rinv20),_mm_sub_ps(rinvsq20,felec)); |
2108 | |
2109 | fscal = felec; |
2110 | |
2111 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
2112 | |
2113 | /* Calculate temporary vectorial force */ |
2114 | tx = _mm_mul_ps(fscal,dx20); |
2115 | ty = _mm_mul_ps(fscal,dy20); |
2116 | tz = _mm_mul_ps(fscal,dz20); |
2117 | |
2118 | /* Update vectorial force */ |
2119 | fix2 = _mm_add_ps(fix2,tx); |
2120 | fiy2 = _mm_add_ps(fiy2,ty); |
2121 | fiz2 = _mm_add_ps(fiz2,tz); |
2122 | |
2123 | fjx0 = _mm_add_ps(fjx0,tx); |
2124 | fjy0 = _mm_add_ps(fjy0,ty); |
2125 | fjz0 = _mm_add_ps(fjz0,tz); |
2126 | |
2127 | /************************** |
2128 | * CALCULATE INTERACTIONS * |
2129 | **************************/ |
2130 | |
2131 | r21 = _mm_mul_ps(rsq21,rinv21); |
2132 | r21 = _mm_andnot_ps(dummy_mask,r21); |
2133 | |
2134 | /* EWALD ELECTROSTATICS */ |
2135 | |
2136 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
2137 | ewrt = _mm_mul_ps(r21,ewtabscale); |
2138 | ewitab = _mm_cvttps_epi32(ewrt); |
2139 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
2140 | 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];})), |
2141 | 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];})), |
2142 | &ewtabF,&ewtabFn); |
2143 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
2144 | felec = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec)); |
2145 | |
2146 | fscal = felec; |
2147 | |
2148 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
2149 | |
2150 | /* Calculate temporary vectorial force */ |
2151 | tx = _mm_mul_ps(fscal,dx21); |
2152 | ty = _mm_mul_ps(fscal,dy21); |
2153 | tz = _mm_mul_ps(fscal,dz21); |
2154 | |
2155 | /* Update vectorial force */ |
2156 | fix2 = _mm_add_ps(fix2,tx); |
2157 | fiy2 = _mm_add_ps(fiy2,ty); |
2158 | fiz2 = _mm_add_ps(fiz2,tz); |
2159 | |
2160 | fjx1 = _mm_add_ps(fjx1,tx); |
2161 | fjy1 = _mm_add_ps(fjy1,ty); |
2162 | fjz1 = _mm_add_ps(fjz1,tz); |
2163 | |
2164 | /************************** |
2165 | * CALCULATE INTERACTIONS * |
2166 | **************************/ |
2167 | |
2168 | r22 = _mm_mul_ps(rsq22,rinv22); |
2169 | r22 = _mm_andnot_ps(dummy_mask,r22); |
2170 | |
2171 | /* EWALD ELECTROSTATICS */ |
2172 | |
2173 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
2174 | ewrt = _mm_mul_ps(r22,ewtabscale); |
2175 | ewitab = _mm_cvttps_epi32(ewrt); |
2176 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
2177 | 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];})), |
2178 | 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];})), |
2179 | &ewtabF,&ewtabFn); |
2180 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
2181 | felec = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec)); |
2182 | |
2183 | fscal = felec; |
2184 | |
2185 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
2186 | |
2187 | /* Calculate temporary vectorial force */ |
2188 | tx = _mm_mul_ps(fscal,dx22); |
2189 | ty = _mm_mul_ps(fscal,dy22); |
2190 | tz = _mm_mul_ps(fscal,dz22); |
2191 | |
2192 | /* Update vectorial force */ |
2193 | fix2 = _mm_add_ps(fix2,tx); |
2194 | fiy2 = _mm_add_ps(fiy2,ty); |
2195 | fiz2 = _mm_add_ps(fiz2,tz); |
2196 | |
2197 | fjx2 = _mm_add_ps(fjx2,tx); |
2198 | fjy2 = _mm_add_ps(fjy2,ty); |
2199 | fjz2 = _mm_add_ps(fjz2,tz); |
2200 | |
2201 | fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch; |
2202 | fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch; |
2203 | fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch; |
2204 | fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch; |
2205 | |
2206 | gmx_mm_decrement_3rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD, |
2207 | fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2); |
2208 | |
2209 | /* Inner loop uses 333 flops */ |
2210 | } |
2211 | |
2212 | /* End of innermost loop */ |
2213 | |
2214 | gmx_mm_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2, |
2215 | f+i_coord_offset,fshift+i_shift_offset); |
2216 | |
2217 | /* Increment number of inner iterations */ |
2218 | inneriter += j_index_end - j_index_start; |
2219 | |
2220 | /* Outer loop uses 18 flops */ |
2221 | } |
2222 | |
2223 | /* Increment number of outer iterations */ |
2224 | outeriter += nri; |
2225 | |
2226 | /* Update outer/inner flops */ |
2227 | |
2228 | inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3W3_F,outeriter*18 + inneriter*333)(nrnb)->n[eNR_NBKERNEL_ELEC_W3W3_F] += outeriter*18 + inneriter *333; |
2229 | } |