File: | gromacs/gmxlib/nonbonded/nb_kernel_sse4_1_single/nb_kernel_ElecEwSw_VdwNone_GeomW4W4_sse4_1_single.c |
Location: | line 1682, column 22 |
Description: | Value stored to 'two' during its initialization is never read |
1 | /* |
2 | * This file is part of the GROMACS molecular simulation package. |
3 | * |
4 | * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by |
5 | * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl, |
6 | * and including many others, as listed in the AUTHORS file in the |
7 | * top-level source directory and at http://www.gromacs.org. |
8 | * |
9 | * GROMACS is free software; you can redistribute it and/or |
10 | * modify it under the terms of the GNU Lesser General Public License |
11 | * as published by the Free Software Foundation; either version 2.1 |
12 | * of the License, or (at your option) any later version. |
13 | * |
14 | * GROMACS is distributed in the hope that it will be useful, |
15 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
16 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
17 | * Lesser General Public License for more details. |
18 | * |
19 | * You should have received a copy of the GNU Lesser General Public |
20 | * License along with GROMACS; if not, see |
21 | * http://www.gnu.org/licenses, or write to the Free Software Foundation, |
22 | * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA. |
23 | * |
24 | * If you want to redistribute modifications to GROMACS, please |
25 | * consider that scientific software is very special. Version |
26 | * control is crucial - bugs must be traceable. We will be happy to |
27 | * consider code for inclusion in the official distribution, but |
28 | * derived work must not be called official GROMACS. Details are found |
29 | * in the README & COPYING files - if they are missing, get the |
30 | * official version at http://www.gromacs.org. |
31 | * |
32 | * To help us fund GROMACS development, we humbly ask that you cite |
33 | * the research papers on the package. Check out http://www.gromacs.org. |
34 | */ |
35 | /* |
36 | * Note: this file was generated by the GROMACS sse4_1_single kernel generator. |
37 | */ |
38 | #ifdef HAVE_CONFIG_H1 |
39 | #include <config.h> |
40 | #endif |
41 | |
42 | #include <math.h> |
43 | |
44 | #include "../nb_kernel.h" |
45 | #include "types/simple.h" |
46 | #include "gromacs/math/vec.h" |
47 | #include "nrnb.h" |
48 | |
49 | #include "gromacs/simd/math_x86_sse4_1_single.h" |
50 | #include "kernelutil_x86_sse4_1_single.h" |
51 | |
52 | /* |
53 | * Gromacs nonbonded kernel: nb_kernel_ElecEwSw_VdwNone_GeomW4W4_VF_sse4_1_single |
54 | * Electrostatics interaction: Ewald |
55 | * VdW interaction: None |
56 | * Geometry: Water4-Water4 |
57 | * Calculate force/pot: PotentialAndForce |
58 | */ |
59 | void |
60 | nb_kernel_ElecEwSw_VdwNone_GeomW4W4_VF_sse4_1_single |
61 | (t_nblist * gmx_restrict nlist, |
62 | rvec * gmx_restrict xx, |
63 | rvec * gmx_restrict ff, |
64 | t_forcerec * gmx_restrict fr, |
65 | t_mdatoms * gmx_restrict mdatoms, |
66 | nb_kernel_data_t gmx_unused__attribute__ ((unused)) * gmx_restrict kernel_data, |
67 | t_nrnb * gmx_restrict nrnb) |
68 | { |
69 | /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or |
70 | * just 0 for non-waters. |
71 | * Suffixes A,B,C,D refer to j loop unrolling done with SSE, e.g. for the four different |
72 | * jnr indices corresponding to data put in the four positions in the SIMD register. |
73 | */ |
74 | int i_shift_offset,i_coord_offset,outeriter,inneriter; |
75 | int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx; |
76 | int jnrA,jnrB,jnrC,jnrD; |
77 | int jnrlistA,jnrlistB,jnrlistC,jnrlistD; |
78 | int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD; |
79 | int *iinr,*jindex,*jjnr,*shiftidx,*gid; |
80 | real rcutoff_scalar; |
81 | real *shiftvec,*fshift,*x,*f; |
82 | real *fjptrA,*fjptrB,*fjptrC,*fjptrD; |
83 | real scratch[4*DIM3]; |
84 | __m128 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall; |
85 | int vdwioffset1; |
86 | __m128 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1; |
87 | int vdwioffset2; |
88 | __m128 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2; |
89 | int vdwioffset3; |
90 | __m128 ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3; |
91 | int vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D; |
92 | __m128 jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1; |
93 | int vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D; |
94 | __m128 jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2; |
95 | int vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D; |
96 | __m128 jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3; |
97 | __m128 dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11; |
98 | __m128 dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12; |
99 | __m128 dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13; |
100 | __m128 dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21; |
101 | __m128 dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22; |
102 | __m128 dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23; |
103 | __m128 dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31; |
104 | __m128 dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32; |
105 | __m128 dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33; |
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 rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw; |
112 | real rswitch_scalar,d_scalar; |
113 | __m128 dummy_mask,cutoff_mask; |
114 | __m128 signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) ); |
115 | __m128 one = _mm_set1_ps(1.0); |
116 | __m128 two = _mm_set1_ps(2.0); |
117 | x = xx[0]; |
118 | f = ff[0]; |
119 | |
120 | nri = nlist->nri; |
121 | iinr = nlist->iinr; |
122 | jindex = nlist->jindex; |
123 | jjnr = nlist->jjnr; |
124 | shiftidx = nlist->shift; |
125 | gid = nlist->gid; |
126 | shiftvec = fr->shift_vec[0]; |
127 | fshift = fr->fshift[0]; |
128 | facel = _mm_set1_ps(fr->epsfac); |
129 | charge = mdatoms->chargeA; |
130 | |
131 | sh_ewald = _mm_set1_ps(fr->ic->sh_ewald); |
132 | ewtab = fr->ic->tabq_coul_FDV0; |
133 | ewtabscale = _mm_set1_ps(fr->ic->tabq_scale); |
134 | ewtabhalfspace = _mm_set1_ps(0.5/fr->ic->tabq_scale); |
135 | |
136 | /* Setup water-specific parameters */ |
137 | inr = nlist->iinr[0]; |
138 | iq1 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1])); |
139 | iq2 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2])); |
140 | iq3 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+3])); |
141 | |
142 | jq1 = _mm_set1_ps(charge[inr+1]); |
143 | jq2 = _mm_set1_ps(charge[inr+2]); |
144 | jq3 = _mm_set1_ps(charge[inr+3]); |
145 | qq11 = _mm_mul_ps(iq1,jq1); |
146 | qq12 = _mm_mul_ps(iq1,jq2); |
147 | qq13 = _mm_mul_ps(iq1,jq3); |
148 | qq21 = _mm_mul_ps(iq2,jq1); |
149 | qq22 = _mm_mul_ps(iq2,jq2); |
150 | qq23 = _mm_mul_ps(iq2,jq3); |
151 | qq31 = _mm_mul_ps(iq3,jq1); |
152 | qq32 = _mm_mul_ps(iq3,jq2); |
153 | qq33 = _mm_mul_ps(iq3,jq3); |
154 | |
155 | /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */ |
156 | rcutoff_scalar = fr->rcoulomb; |
157 | rcutoff = _mm_set1_ps(rcutoff_scalar); |
158 | rcutoff2 = _mm_mul_ps(rcutoff,rcutoff); |
159 | |
160 | rswitch_scalar = fr->rcoulomb_switch; |
161 | rswitch = _mm_set1_ps(rswitch_scalar); |
162 | /* Setup switch parameters */ |
163 | d_scalar = rcutoff_scalar-rswitch_scalar; |
164 | d = _mm_set1_ps(d_scalar); |
165 | swV3 = _mm_set1_ps(-10.0/(d_scalar*d_scalar*d_scalar)); |
166 | swV4 = _mm_set1_ps( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar)); |
167 | swV5 = _mm_set1_ps( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar)); |
168 | swF2 = _mm_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar)); |
169 | swF3 = _mm_set1_ps( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar)); |
170 | swF4 = _mm_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar)); |
171 | |
172 | /* Avoid stupid compiler warnings */ |
173 | jnrA = jnrB = jnrC = jnrD = 0; |
174 | j_coord_offsetA = 0; |
175 | j_coord_offsetB = 0; |
176 | j_coord_offsetC = 0; |
177 | j_coord_offsetD = 0; |
178 | |
179 | outeriter = 0; |
180 | inneriter = 0; |
181 | |
182 | for(iidx=0;iidx<4*DIM3;iidx++) |
183 | { |
184 | scratch[iidx] = 0.0; |
185 | } |
186 | |
187 | /* Start outer loop over neighborlists */ |
188 | for(iidx=0; iidx<nri; iidx++) |
189 | { |
190 | /* Load shift vector for this list */ |
191 | i_shift_offset = DIM3*shiftidx[iidx]; |
192 | |
193 | /* Load limits for loop over neighbors */ |
194 | j_index_start = jindex[iidx]; |
195 | j_index_end = jindex[iidx+1]; |
196 | |
197 | /* Get outer coordinate index */ |
198 | inr = iinr[iidx]; |
199 | i_coord_offset = DIM3*inr; |
200 | |
201 | /* Load i particle coords and add shift vector */ |
202 | gmx_mm_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset+DIM3, |
203 | &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3); |
204 | |
205 | fix1 = _mm_setzero_ps(); |
206 | fiy1 = _mm_setzero_ps(); |
207 | fiz1 = _mm_setzero_ps(); |
208 | fix2 = _mm_setzero_ps(); |
209 | fiy2 = _mm_setzero_ps(); |
210 | fiz2 = _mm_setzero_ps(); |
211 | fix3 = _mm_setzero_ps(); |
212 | fiy3 = _mm_setzero_ps(); |
213 | fiz3 = _mm_setzero_ps(); |
214 | |
215 | /* Reset potential sums */ |
216 | velecsum = _mm_setzero_ps(); |
217 | |
218 | /* Start inner kernel loop */ |
219 | for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4) |
220 | { |
221 | |
222 | /* Get j neighbor index, and coordinate index */ |
223 | jnrA = jjnr[jidx]; |
224 | jnrB = jjnr[jidx+1]; |
225 | jnrC = jjnr[jidx+2]; |
226 | jnrD = jjnr[jidx+3]; |
227 | j_coord_offsetA = DIM3*jnrA; |
228 | j_coord_offsetB = DIM3*jnrB; |
229 | j_coord_offsetC = DIM3*jnrC; |
230 | j_coord_offsetD = DIM3*jnrD; |
231 | |
232 | /* load j atom coordinates */ |
233 | gmx_mm_load_3rvec_4ptr_swizzle_ps(x+j_coord_offsetA+DIM3,x+j_coord_offsetB+DIM3, |
234 | x+j_coord_offsetC+DIM3,x+j_coord_offsetD+DIM3, |
235 | &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3); |
236 | |
237 | /* Calculate displacement vector */ |
238 | dx11 = _mm_sub_ps(ix1,jx1); |
239 | dy11 = _mm_sub_ps(iy1,jy1); |
240 | dz11 = _mm_sub_ps(iz1,jz1); |
241 | dx12 = _mm_sub_ps(ix1,jx2); |
242 | dy12 = _mm_sub_ps(iy1,jy2); |
243 | dz12 = _mm_sub_ps(iz1,jz2); |
244 | dx13 = _mm_sub_ps(ix1,jx3); |
245 | dy13 = _mm_sub_ps(iy1,jy3); |
246 | dz13 = _mm_sub_ps(iz1,jz3); |
247 | dx21 = _mm_sub_ps(ix2,jx1); |
248 | dy21 = _mm_sub_ps(iy2,jy1); |
249 | dz21 = _mm_sub_ps(iz2,jz1); |
250 | dx22 = _mm_sub_ps(ix2,jx2); |
251 | dy22 = _mm_sub_ps(iy2,jy2); |
252 | dz22 = _mm_sub_ps(iz2,jz2); |
253 | dx23 = _mm_sub_ps(ix2,jx3); |
254 | dy23 = _mm_sub_ps(iy2,jy3); |
255 | dz23 = _mm_sub_ps(iz2,jz3); |
256 | dx31 = _mm_sub_ps(ix3,jx1); |
257 | dy31 = _mm_sub_ps(iy3,jy1); |
258 | dz31 = _mm_sub_ps(iz3,jz1); |
259 | dx32 = _mm_sub_ps(ix3,jx2); |
260 | dy32 = _mm_sub_ps(iy3,jy2); |
261 | dz32 = _mm_sub_ps(iz3,jz2); |
262 | dx33 = _mm_sub_ps(ix3,jx3); |
263 | dy33 = _mm_sub_ps(iy3,jy3); |
264 | dz33 = _mm_sub_ps(iz3,jz3); |
265 | |
266 | /* Calculate squared distance and things based on it */ |
267 | rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11); |
268 | rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12); |
269 | rsq13 = gmx_mm_calc_rsq_ps(dx13,dy13,dz13); |
270 | rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21); |
271 | rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22); |
272 | rsq23 = gmx_mm_calc_rsq_ps(dx23,dy23,dz23); |
273 | rsq31 = gmx_mm_calc_rsq_ps(dx31,dy31,dz31); |
274 | rsq32 = gmx_mm_calc_rsq_ps(dx32,dy32,dz32); |
275 | rsq33 = gmx_mm_calc_rsq_ps(dx33,dy33,dz33); |
276 | |
277 | rinv11 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq11); |
278 | rinv12 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq12); |
279 | rinv13 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq13); |
280 | rinv21 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq21); |
281 | rinv22 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq22); |
282 | rinv23 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq23); |
283 | rinv31 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq31); |
284 | rinv32 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq32); |
285 | rinv33 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq33); |
286 | |
287 | rinvsq11 = _mm_mul_ps(rinv11,rinv11); |
288 | rinvsq12 = _mm_mul_ps(rinv12,rinv12); |
289 | rinvsq13 = _mm_mul_ps(rinv13,rinv13); |
290 | rinvsq21 = _mm_mul_ps(rinv21,rinv21); |
291 | rinvsq22 = _mm_mul_ps(rinv22,rinv22); |
292 | rinvsq23 = _mm_mul_ps(rinv23,rinv23); |
293 | rinvsq31 = _mm_mul_ps(rinv31,rinv31); |
294 | rinvsq32 = _mm_mul_ps(rinv32,rinv32); |
295 | rinvsq33 = _mm_mul_ps(rinv33,rinv33); |
296 | |
297 | fjx1 = _mm_setzero_ps(); |
298 | fjy1 = _mm_setzero_ps(); |
299 | fjz1 = _mm_setzero_ps(); |
300 | fjx2 = _mm_setzero_ps(); |
301 | fjy2 = _mm_setzero_ps(); |
302 | fjz2 = _mm_setzero_ps(); |
303 | fjx3 = _mm_setzero_ps(); |
304 | fjy3 = _mm_setzero_ps(); |
305 | fjz3 = _mm_setzero_ps(); |
306 | |
307 | /************************** |
308 | * CALCULATE INTERACTIONS * |
309 | **************************/ |
310 | |
311 | if (gmx_mm_any_lt(rsq11,rcutoff2)) |
312 | { |
313 | |
314 | r11 = _mm_mul_ps(rsq11,rinv11); |
315 | |
316 | /* EWALD ELECTROSTATICS */ |
317 | |
318 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
319 | ewrt = _mm_mul_ps(r11,ewtabscale); |
320 | ewitab = _mm_cvttps_epi32(ewrt); |
321 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
322 | ewitab = _mm_slli_epi32(ewitab,2); |
323 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
324 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
325 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
326 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
327 | _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); |
328 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
329 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
330 | velec = _mm_mul_ps(qq11,_mm_sub_ps(rinv11,velec)); |
331 | felec = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec)); |
332 | |
333 | d = _mm_sub_ps(r11,rswitch); |
334 | d = _mm_max_ps(d,_mm_setzero_ps()); |
335 | d2 = _mm_mul_ps(d,d); |
336 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
337 | |
338 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
339 | |
340 | /* Evaluate switch function */ |
341 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
342 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv11,_mm_mul_ps(velec,dsw)) ); |
343 | velec = _mm_mul_ps(velec,sw); |
344 | cutoff_mask = _mm_cmplt_ps(rsq11,rcutoff2); |
345 | |
346 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
347 | velec = _mm_and_ps(velec,cutoff_mask); |
348 | velecsum = _mm_add_ps(velecsum,velec); |
349 | |
350 | fscal = felec; |
351 | |
352 | fscal = _mm_and_ps(fscal,cutoff_mask); |
353 | |
354 | /* Calculate temporary vectorial force */ |
355 | tx = _mm_mul_ps(fscal,dx11); |
356 | ty = _mm_mul_ps(fscal,dy11); |
357 | tz = _mm_mul_ps(fscal,dz11); |
358 | |
359 | /* Update vectorial force */ |
360 | fix1 = _mm_add_ps(fix1,tx); |
361 | fiy1 = _mm_add_ps(fiy1,ty); |
362 | fiz1 = _mm_add_ps(fiz1,tz); |
363 | |
364 | fjx1 = _mm_add_ps(fjx1,tx); |
365 | fjy1 = _mm_add_ps(fjy1,ty); |
366 | fjz1 = _mm_add_ps(fjz1,tz); |
367 | |
368 | } |
369 | |
370 | /************************** |
371 | * CALCULATE INTERACTIONS * |
372 | **************************/ |
373 | |
374 | if (gmx_mm_any_lt(rsq12,rcutoff2)) |
375 | { |
376 | |
377 | r12 = _mm_mul_ps(rsq12,rinv12); |
378 | |
379 | /* EWALD ELECTROSTATICS */ |
380 | |
381 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
382 | ewrt = _mm_mul_ps(r12,ewtabscale); |
383 | ewitab = _mm_cvttps_epi32(ewrt); |
384 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
385 | ewitab = _mm_slli_epi32(ewitab,2); |
386 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
387 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
388 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
389 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
390 | _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); |
391 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
392 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
393 | velec = _mm_mul_ps(qq12,_mm_sub_ps(rinv12,velec)); |
394 | felec = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec)); |
395 | |
396 | d = _mm_sub_ps(r12,rswitch); |
397 | d = _mm_max_ps(d,_mm_setzero_ps()); |
398 | d2 = _mm_mul_ps(d,d); |
399 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
400 | |
401 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
402 | |
403 | /* Evaluate switch function */ |
404 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
405 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv12,_mm_mul_ps(velec,dsw)) ); |
406 | velec = _mm_mul_ps(velec,sw); |
407 | cutoff_mask = _mm_cmplt_ps(rsq12,rcutoff2); |
408 | |
409 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
410 | velec = _mm_and_ps(velec,cutoff_mask); |
411 | velecsum = _mm_add_ps(velecsum,velec); |
412 | |
413 | fscal = felec; |
414 | |
415 | fscal = _mm_and_ps(fscal,cutoff_mask); |
416 | |
417 | /* Calculate temporary vectorial force */ |
418 | tx = _mm_mul_ps(fscal,dx12); |
419 | ty = _mm_mul_ps(fscal,dy12); |
420 | tz = _mm_mul_ps(fscal,dz12); |
421 | |
422 | /* Update vectorial force */ |
423 | fix1 = _mm_add_ps(fix1,tx); |
424 | fiy1 = _mm_add_ps(fiy1,ty); |
425 | fiz1 = _mm_add_ps(fiz1,tz); |
426 | |
427 | fjx2 = _mm_add_ps(fjx2,tx); |
428 | fjy2 = _mm_add_ps(fjy2,ty); |
429 | fjz2 = _mm_add_ps(fjz2,tz); |
430 | |
431 | } |
432 | |
433 | /************************** |
434 | * CALCULATE INTERACTIONS * |
435 | **************************/ |
436 | |
437 | if (gmx_mm_any_lt(rsq13,rcutoff2)) |
438 | { |
439 | |
440 | r13 = _mm_mul_ps(rsq13,rinv13); |
441 | |
442 | /* EWALD ELECTROSTATICS */ |
443 | |
444 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
445 | ewrt = _mm_mul_ps(r13,ewtabscale); |
446 | ewitab = _mm_cvttps_epi32(ewrt); |
447 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
448 | ewitab = _mm_slli_epi32(ewitab,2); |
449 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
450 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
451 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
452 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
453 | _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); |
454 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
455 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
456 | velec = _mm_mul_ps(qq13,_mm_sub_ps(rinv13,velec)); |
457 | felec = _mm_mul_ps(_mm_mul_ps(qq13,rinv13),_mm_sub_ps(rinvsq13,felec)); |
458 | |
459 | d = _mm_sub_ps(r13,rswitch); |
460 | d = _mm_max_ps(d,_mm_setzero_ps()); |
461 | d2 = _mm_mul_ps(d,d); |
462 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
463 | |
464 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
465 | |
466 | /* Evaluate switch function */ |
467 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
468 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv13,_mm_mul_ps(velec,dsw)) ); |
469 | velec = _mm_mul_ps(velec,sw); |
470 | cutoff_mask = _mm_cmplt_ps(rsq13,rcutoff2); |
471 | |
472 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
473 | velec = _mm_and_ps(velec,cutoff_mask); |
474 | velecsum = _mm_add_ps(velecsum,velec); |
475 | |
476 | fscal = felec; |
477 | |
478 | fscal = _mm_and_ps(fscal,cutoff_mask); |
479 | |
480 | /* Calculate temporary vectorial force */ |
481 | tx = _mm_mul_ps(fscal,dx13); |
482 | ty = _mm_mul_ps(fscal,dy13); |
483 | tz = _mm_mul_ps(fscal,dz13); |
484 | |
485 | /* Update vectorial force */ |
486 | fix1 = _mm_add_ps(fix1,tx); |
487 | fiy1 = _mm_add_ps(fiy1,ty); |
488 | fiz1 = _mm_add_ps(fiz1,tz); |
489 | |
490 | fjx3 = _mm_add_ps(fjx3,tx); |
491 | fjy3 = _mm_add_ps(fjy3,ty); |
492 | fjz3 = _mm_add_ps(fjz3,tz); |
493 | |
494 | } |
495 | |
496 | /************************** |
497 | * CALCULATE INTERACTIONS * |
498 | **************************/ |
499 | |
500 | if (gmx_mm_any_lt(rsq21,rcutoff2)) |
501 | { |
502 | |
503 | r21 = _mm_mul_ps(rsq21,rinv21); |
504 | |
505 | /* EWALD ELECTROSTATICS */ |
506 | |
507 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
508 | ewrt = _mm_mul_ps(r21,ewtabscale); |
509 | ewitab = _mm_cvttps_epi32(ewrt); |
510 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
511 | ewitab = _mm_slli_epi32(ewitab,2); |
512 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
513 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
514 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
515 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
516 | _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); |
517 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
518 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
519 | velec = _mm_mul_ps(qq21,_mm_sub_ps(rinv21,velec)); |
520 | felec = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec)); |
521 | |
522 | d = _mm_sub_ps(r21,rswitch); |
523 | d = _mm_max_ps(d,_mm_setzero_ps()); |
524 | d2 = _mm_mul_ps(d,d); |
525 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
526 | |
527 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
528 | |
529 | /* Evaluate switch function */ |
530 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
531 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv21,_mm_mul_ps(velec,dsw)) ); |
532 | velec = _mm_mul_ps(velec,sw); |
533 | cutoff_mask = _mm_cmplt_ps(rsq21,rcutoff2); |
534 | |
535 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
536 | velec = _mm_and_ps(velec,cutoff_mask); |
537 | velecsum = _mm_add_ps(velecsum,velec); |
538 | |
539 | fscal = felec; |
540 | |
541 | fscal = _mm_and_ps(fscal,cutoff_mask); |
542 | |
543 | /* Calculate temporary vectorial force */ |
544 | tx = _mm_mul_ps(fscal,dx21); |
545 | ty = _mm_mul_ps(fscal,dy21); |
546 | tz = _mm_mul_ps(fscal,dz21); |
547 | |
548 | /* Update vectorial force */ |
549 | fix2 = _mm_add_ps(fix2,tx); |
550 | fiy2 = _mm_add_ps(fiy2,ty); |
551 | fiz2 = _mm_add_ps(fiz2,tz); |
552 | |
553 | fjx1 = _mm_add_ps(fjx1,tx); |
554 | fjy1 = _mm_add_ps(fjy1,ty); |
555 | fjz1 = _mm_add_ps(fjz1,tz); |
556 | |
557 | } |
558 | |
559 | /************************** |
560 | * CALCULATE INTERACTIONS * |
561 | **************************/ |
562 | |
563 | if (gmx_mm_any_lt(rsq22,rcutoff2)) |
564 | { |
565 | |
566 | r22 = _mm_mul_ps(rsq22,rinv22); |
567 | |
568 | /* EWALD ELECTROSTATICS */ |
569 | |
570 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
571 | ewrt = _mm_mul_ps(r22,ewtabscale); |
572 | ewitab = _mm_cvttps_epi32(ewrt); |
573 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
574 | ewitab = _mm_slli_epi32(ewitab,2); |
575 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
576 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
577 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
578 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
579 | _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); |
580 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
581 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
582 | velec = _mm_mul_ps(qq22,_mm_sub_ps(rinv22,velec)); |
583 | felec = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec)); |
584 | |
585 | d = _mm_sub_ps(r22,rswitch); |
586 | d = _mm_max_ps(d,_mm_setzero_ps()); |
587 | d2 = _mm_mul_ps(d,d); |
588 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
589 | |
590 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
591 | |
592 | /* Evaluate switch function */ |
593 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
594 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv22,_mm_mul_ps(velec,dsw)) ); |
595 | velec = _mm_mul_ps(velec,sw); |
596 | cutoff_mask = _mm_cmplt_ps(rsq22,rcutoff2); |
597 | |
598 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
599 | velec = _mm_and_ps(velec,cutoff_mask); |
600 | velecsum = _mm_add_ps(velecsum,velec); |
601 | |
602 | fscal = felec; |
603 | |
604 | fscal = _mm_and_ps(fscal,cutoff_mask); |
605 | |
606 | /* Calculate temporary vectorial force */ |
607 | tx = _mm_mul_ps(fscal,dx22); |
608 | ty = _mm_mul_ps(fscal,dy22); |
609 | tz = _mm_mul_ps(fscal,dz22); |
610 | |
611 | /* Update vectorial force */ |
612 | fix2 = _mm_add_ps(fix2,tx); |
613 | fiy2 = _mm_add_ps(fiy2,ty); |
614 | fiz2 = _mm_add_ps(fiz2,tz); |
615 | |
616 | fjx2 = _mm_add_ps(fjx2,tx); |
617 | fjy2 = _mm_add_ps(fjy2,ty); |
618 | fjz2 = _mm_add_ps(fjz2,tz); |
619 | |
620 | } |
621 | |
622 | /************************** |
623 | * CALCULATE INTERACTIONS * |
624 | **************************/ |
625 | |
626 | if (gmx_mm_any_lt(rsq23,rcutoff2)) |
627 | { |
628 | |
629 | r23 = _mm_mul_ps(rsq23,rinv23); |
630 | |
631 | /* EWALD ELECTROSTATICS */ |
632 | |
633 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
634 | ewrt = _mm_mul_ps(r23,ewtabscale); |
635 | ewitab = _mm_cvttps_epi32(ewrt); |
636 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
637 | ewitab = _mm_slli_epi32(ewitab,2); |
638 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
639 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
640 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
641 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
642 | _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); |
643 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
644 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
645 | velec = _mm_mul_ps(qq23,_mm_sub_ps(rinv23,velec)); |
646 | felec = _mm_mul_ps(_mm_mul_ps(qq23,rinv23),_mm_sub_ps(rinvsq23,felec)); |
647 | |
648 | d = _mm_sub_ps(r23,rswitch); |
649 | d = _mm_max_ps(d,_mm_setzero_ps()); |
650 | d2 = _mm_mul_ps(d,d); |
651 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
652 | |
653 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
654 | |
655 | /* Evaluate switch function */ |
656 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
657 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv23,_mm_mul_ps(velec,dsw)) ); |
658 | velec = _mm_mul_ps(velec,sw); |
659 | cutoff_mask = _mm_cmplt_ps(rsq23,rcutoff2); |
660 | |
661 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
662 | velec = _mm_and_ps(velec,cutoff_mask); |
663 | velecsum = _mm_add_ps(velecsum,velec); |
664 | |
665 | fscal = felec; |
666 | |
667 | fscal = _mm_and_ps(fscal,cutoff_mask); |
668 | |
669 | /* Calculate temporary vectorial force */ |
670 | tx = _mm_mul_ps(fscal,dx23); |
671 | ty = _mm_mul_ps(fscal,dy23); |
672 | tz = _mm_mul_ps(fscal,dz23); |
673 | |
674 | /* Update vectorial force */ |
675 | fix2 = _mm_add_ps(fix2,tx); |
676 | fiy2 = _mm_add_ps(fiy2,ty); |
677 | fiz2 = _mm_add_ps(fiz2,tz); |
678 | |
679 | fjx3 = _mm_add_ps(fjx3,tx); |
680 | fjy3 = _mm_add_ps(fjy3,ty); |
681 | fjz3 = _mm_add_ps(fjz3,tz); |
682 | |
683 | } |
684 | |
685 | /************************** |
686 | * CALCULATE INTERACTIONS * |
687 | **************************/ |
688 | |
689 | if (gmx_mm_any_lt(rsq31,rcutoff2)) |
690 | { |
691 | |
692 | r31 = _mm_mul_ps(rsq31,rinv31); |
693 | |
694 | /* EWALD ELECTROSTATICS */ |
695 | |
696 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
697 | ewrt = _mm_mul_ps(r31,ewtabscale); |
698 | ewitab = _mm_cvttps_epi32(ewrt); |
699 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
700 | ewitab = _mm_slli_epi32(ewitab,2); |
701 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
702 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
703 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
704 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
705 | _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); |
706 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
707 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
708 | velec = _mm_mul_ps(qq31,_mm_sub_ps(rinv31,velec)); |
709 | felec = _mm_mul_ps(_mm_mul_ps(qq31,rinv31),_mm_sub_ps(rinvsq31,felec)); |
710 | |
711 | d = _mm_sub_ps(r31,rswitch); |
712 | d = _mm_max_ps(d,_mm_setzero_ps()); |
713 | d2 = _mm_mul_ps(d,d); |
714 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
715 | |
716 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
717 | |
718 | /* Evaluate switch function */ |
719 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
720 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv31,_mm_mul_ps(velec,dsw)) ); |
721 | velec = _mm_mul_ps(velec,sw); |
722 | cutoff_mask = _mm_cmplt_ps(rsq31,rcutoff2); |
723 | |
724 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
725 | velec = _mm_and_ps(velec,cutoff_mask); |
726 | velecsum = _mm_add_ps(velecsum,velec); |
727 | |
728 | fscal = felec; |
729 | |
730 | fscal = _mm_and_ps(fscal,cutoff_mask); |
731 | |
732 | /* Calculate temporary vectorial force */ |
733 | tx = _mm_mul_ps(fscal,dx31); |
734 | ty = _mm_mul_ps(fscal,dy31); |
735 | tz = _mm_mul_ps(fscal,dz31); |
736 | |
737 | /* Update vectorial force */ |
738 | fix3 = _mm_add_ps(fix3,tx); |
739 | fiy3 = _mm_add_ps(fiy3,ty); |
740 | fiz3 = _mm_add_ps(fiz3,tz); |
741 | |
742 | fjx1 = _mm_add_ps(fjx1,tx); |
743 | fjy1 = _mm_add_ps(fjy1,ty); |
744 | fjz1 = _mm_add_ps(fjz1,tz); |
745 | |
746 | } |
747 | |
748 | /************************** |
749 | * CALCULATE INTERACTIONS * |
750 | **************************/ |
751 | |
752 | if (gmx_mm_any_lt(rsq32,rcutoff2)) |
753 | { |
754 | |
755 | r32 = _mm_mul_ps(rsq32,rinv32); |
756 | |
757 | /* EWALD ELECTROSTATICS */ |
758 | |
759 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
760 | ewrt = _mm_mul_ps(r32,ewtabscale); |
761 | ewitab = _mm_cvttps_epi32(ewrt); |
762 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
763 | ewitab = _mm_slli_epi32(ewitab,2); |
764 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
765 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
766 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
767 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
768 | _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); |
769 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
770 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
771 | velec = _mm_mul_ps(qq32,_mm_sub_ps(rinv32,velec)); |
772 | felec = _mm_mul_ps(_mm_mul_ps(qq32,rinv32),_mm_sub_ps(rinvsq32,felec)); |
773 | |
774 | d = _mm_sub_ps(r32,rswitch); |
775 | d = _mm_max_ps(d,_mm_setzero_ps()); |
776 | d2 = _mm_mul_ps(d,d); |
777 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
778 | |
779 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
780 | |
781 | /* Evaluate switch function */ |
782 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
783 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv32,_mm_mul_ps(velec,dsw)) ); |
784 | velec = _mm_mul_ps(velec,sw); |
785 | cutoff_mask = _mm_cmplt_ps(rsq32,rcutoff2); |
786 | |
787 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
788 | velec = _mm_and_ps(velec,cutoff_mask); |
789 | velecsum = _mm_add_ps(velecsum,velec); |
790 | |
791 | fscal = felec; |
792 | |
793 | fscal = _mm_and_ps(fscal,cutoff_mask); |
794 | |
795 | /* Calculate temporary vectorial force */ |
796 | tx = _mm_mul_ps(fscal,dx32); |
797 | ty = _mm_mul_ps(fscal,dy32); |
798 | tz = _mm_mul_ps(fscal,dz32); |
799 | |
800 | /* Update vectorial force */ |
801 | fix3 = _mm_add_ps(fix3,tx); |
802 | fiy3 = _mm_add_ps(fiy3,ty); |
803 | fiz3 = _mm_add_ps(fiz3,tz); |
804 | |
805 | fjx2 = _mm_add_ps(fjx2,tx); |
806 | fjy2 = _mm_add_ps(fjy2,ty); |
807 | fjz2 = _mm_add_ps(fjz2,tz); |
808 | |
809 | } |
810 | |
811 | /************************** |
812 | * CALCULATE INTERACTIONS * |
813 | **************************/ |
814 | |
815 | if (gmx_mm_any_lt(rsq33,rcutoff2)) |
816 | { |
817 | |
818 | r33 = _mm_mul_ps(rsq33,rinv33); |
819 | |
820 | /* EWALD ELECTROSTATICS */ |
821 | |
822 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
823 | ewrt = _mm_mul_ps(r33,ewtabscale); |
824 | ewitab = _mm_cvttps_epi32(ewrt); |
825 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
826 | ewitab = _mm_slli_epi32(ewitab,2); |
827 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
828 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
829 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
830 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
831 | _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); |
832 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
833 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
834 | velec = _mm_mul_ps(qq33,_mm_sub_ps(rinv33,velec)); |
835 | felec = _mm_mul_ps(_mm_mul_ps(qq33,rinv33),_mm_sub_ps(rinvsq33,felec)); |
836 | |
837 | d = _mm_sub_ps(r33,rswitch); |
838 | d = _mm_max_ps(d,_mm_setzero_ps()); |
839 | d2 = _mm_mul_ps(d,d); |
840 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
841 | |
842 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
843 | |
844 | /* Evaluate switch function */ |
845 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
846 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv33,_mm_mul_ps(velec,dsw)) ); |
847 | velec = _mm_mul_ps(velec,sw); |
848 | cutoff_mask = _mm_cmplt_ps(rsq33,rcutoff2); |
849 | |
850 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
851 | velec = _mm_and_ps(velec,cutoff_mask); |
852 | velecsum = _mm_add_ps(velecsum,velec); |
853 | |
854 | fscal = felec; |
855 | |
856 | fscal = _mm_and_ps(fscal,cutoff_mask); |
857 | |
858 | /* Calculate temporary vectorial force */ |
859 | tx = _mm_mul_ps(fscal,dx33); |
860 | ty = _mm_mul_ps(fscal,dy33); |
861 | tz = _mm_mul_ps(fscal,dz33); |
862 | |
863 | /* Update vectorial force */ |
864 | fix3 = _mm_add_ps(fix3,tx); |
865 | fiy3 = _mm_add_ps(fiy3,ty); |
866 | fiz3 = _mm_add_ps(fiz3,tz); |
867 | |
868 | fjx3 = _mm_add_ps(fjx3,tx); |
869 | fjy3 = _mm_add_ps(fjy3,ty); |
870 | fjz3 = _mm_add_ps(fjz3,tz); |
871 | |
872 | } |
873 | |
874 | fjptrA = f+j_coord_offsetA; |
875 | fjptrB = f+j_coord_offsetB; |
876 | fjptrC = f+j_coord_offsetC; |
877 | fjptrD = f+j_coord_offsetD; |
878 | |
879 | gmx_mm_decrement_3rvec_4ptr_swizzle_ps(fjptrA+DIM3,fjptrB+DIM3,fjptrC+DIM3,fjptrD+DIM3, |
880 | fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3); |
881 | |
882 | /* Inner loop uses 585 flops */ |
883 | } |
884 | |
885 | if(jidx<j_index_end) |
886 | { |
887 | |
888 | /* Get j neighbor index, and coordinate index */ |
889 | jnrlistA = jjnr[jidx]; |
890 | jnrlistB = jjnr[jidx+1]; |
891 | jnrlistC = jjnr[jidx+2]; |
892 | jnrlistD = jjnr[jidx+3]; |
893 | /* Sign of each element will be negative for non-real atoms. |
894 | * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones, |
895 | * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries. |
896 | */ |
897 | dummy_mask = gmx_mm_castsi128_ps_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())); |
898 | jnrA = (jnrlistA>=0) ? jnrlistA : 0; |
899 | jnrB = (jnrlistB>=0) ? jnrlistB : 0; |
900 | jnrC = (jnrlistC>=0) ? jnrlistC : 0; |
901 | jnrD = (jnrlistD>=0) ? jnrlistD : 0; |
902 | j_coord_offsetA = DIM3*jnrA; |
903 | j_coord_offsetB = DIM3*jnrB; |
904 | j_coord_offsetC = DIM3*jnrC; |
905 | j_coord_offsetD = DIM3*jnrD; |
906 | |
907 | /* load j atom coordinates */ |
908 | gmx_mm_load_3rvec_4ptr_swizzle_ps(x+j_coord_offsetA+DIM3,x+j_coord_offsetB+DIM3, |
909 | x+j_coord_offsetC+DIM3,x+j_coord_offsetD+DIM3, |
910 | &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3); |
911 | |
912 | /* Calculate displacement vector */ |
913 | dx11 = _mm_sub_ps(ix1,jx1); |
914 | dy11 = _mm_sub_ps(iy1,jy1); |
915 | dz11 = _mm_sub_ps(iz1,jz1); |
916 | dx12 = _mm_sub_ps(ix1,jx2); |
917 | dy12 = _mm_sub_ps(iy1,jy2); |
918 | dz12 = _mm_sub_ps(iz1,jz2); |
919 | dx13 = _mm_sub_ps(ix1,jx3); |
920 | dy13 = _mm_sub_ps(iy1,jy3); |
921 | dz13 = _mm_sub_ps(iz1,jz3); |
922 | dx21 = _mm_sub_ps(ix2,jx1); |
923 | dy21 = _mm_sub_ps(iy2,jy1); |
924 | dz21 = _mm_sub_ps(iz2,jz1); |
925 | dx22 = _mm_sub_ps(ix2,jx2); |
926 | dy22 = _mm_sub_ps(iy2,jy2); |
927 | dz22 = _mm_sub_ps(iz2,jz2); |
928 | dx23 = _mm_sub_ps(ix2,jx3); |
929 | dy23 = _mm_sub_ps(iy2,jy3); |
930 | dz23 = _mm_sub_ps(iz2,jz3); |
931 | dx31 = _mm_sub_ps(ix3,jx1); |
932 | dy31 = _mm_sub_ps(iy3,jy1); |
933 | dz31 = _mm_sub_ps(iz3,jz1); |
934 | dx32 = _mm_sub_ps(ix3,jx2); |
935 | dy32 = _mm_sub_ps(iy3,jy2); |
936 | dz32 = _mm_sub_ps(iz3,jz2); |
937 | dx33 = _mm_sub_ps(ix3,jx3); |
938 | dy33 = _mm_sub_ps(iy3,jy3); |
939 | dz33 = _mm_sub_ps(iz3,jz3); |
940 | |
941 | /* Calculate squared distance and things based on it */ |
942 | rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11); |
943 | rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12); |
944 | rsq13 = gmx_mm_calc_rsq_ps(dx13,dy13,dz13); |
945 | rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21); |
946 | rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22); |
947 | rsq23 = gmx_mm_calc_rsq_ps(dx23,dy23,dz23); |
948 | rsq31 = gmx_mm_calc_rsq_ps(dx31,dy31,dz31); |
949 | rsq32 = gmx_mm_calc_rsq_ps(dx32,dy32,dz32); |
950 | rsq33 = gmx_mm_calc_rsq_ps(dx33,dy33,dz33); |
951 | |
952 | rinv11 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq11); |
953 | rinv12 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq12); |
954 | rinv13 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq13); |
955 | rinv21 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq21); |
956 | rinv22 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq22); |
957 | rinv23 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq23); |
958 | rinv31 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq31); |
959 | rinv32 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq32); |
960 | rinv33 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq33); |
961 | |
962 | rinvsq11 = _mm_mul_ps(rinv11,rinv11); |
963 | rinvsq12 = _mm_mul_ps(rinv12,rinv12); |
964 | rinvsq13 = _mm_mul_ps(rinv13,rinv13); |
965 | rinvsq21 = _mm_mul_ps(rinv21,rinv21); |
966 | rinvsq22 = _mm_mul_ps(rinv22,rinv22); |
967 | rinvsq23 = _mm_mul_ps(rinv23,rinv23); |
968 | rinvsq31 = _mm_mul_ps(rinv31,rinv31); |
969 | rinvsq32 = _mm_mul_ps(rinv32,rinv32); |
970 | rinvsq33 = _mm_mul_ps(rinv33,rinv33); |
971 | |
972 | fjx1 = _mm_setzero_ps(); |
973 | fjy1 = _mm_setzero_ps(); |
974 | fjz1 = _mm_setzero_ps(); |
975 | fjx2 = _mm_setzero_ps(); |
976 | fjy2 = _mm_setzero_ps(); |
977 | fjz2 = _mm_setzero_ps(); |
978 | fjx3 = _mm_setzero_ps(); |
979 | fjy3 = _mm_setzero_ps(); |
980 | fjz3 = _mm_setzero_ps(); |
981 | |
982 | /************************** |
983 | * CALCULATE INTERACTIONS * |
984 | **************************/ |
985 | |
986 | if (gmx_mm_any_lt(rsq11,rcutoff2)) |
987 | { |
988 | |
989 | r11 = _mm_mul_ps(rsq11,rinv11); |
990 | r11 = _mm_andnot_ps(dummy_mask,r11); |
991 | |
992 | /* EWALD ELECTROSTATICS */ |
993 | |
994 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
995 | ewrt = _mm_mul_ps(r11,ewtabscale); |
996 | ewitab = _mm_cvttps_epi32(ewrt); |
997 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
998 | ewitab = _mm_slli_epi32(ewitab,2); |
999 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
1000 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
1001 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
1002 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
1003 | _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); |
1004 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
1005 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
1006 | velec = _mm_mul_ps(qq11,_mm_sub_ps(rinv11,velec)); |
1007 | felec = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec)); |
1008 | |
1009 | d = _mm_sub_ps(r11,rswitch); |
1010 | d = _mm_max_ps(d,_mm_setzero_ps()); |
1011 | d2 = _mm_mul_ps(d,d); |
1012 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
1013 | |
1014 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
1015 | |
1016 | /* Evaluate switch function */ |
1017 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
1018 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv11,_mm_mul_ps(velec,dsw)) ); |
1019 | velec = _mm_mul_ps(velec,sw); |
1020 | cutoff_mask = _mm_cmplt_ps(rsq11,rcutoff2); |
1021 | |
1022 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
1023 | velec = _mm_and_ps(velec,cutoff_mask); |
1024 | velec = _mm_andnot_ps(dummy_mask,velec); |
1025 | velecsum = _mm_add_ps(velecsum,velec); |
1026 | |
1027 | fscal = felec; |
1028 | |
1029 | fscal = _mm_and_ps(fscal,cutoff_mask); |
1030 | |
1031 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
1032 | |
1033 | /* Calculate temporary vectorial force */ |
1034 | tx = _mm_mul_ps(fscal,dx11); |
1035 | ty = _mm_mul_ps(fscal,dy11); |
1036 | tz = _mm_mul_ps(fscal,dz11); |
1037 | |
1038 | /* Update vectorial force */ |
1039 | fix1 = _mm_add_ps(fix1,tx); |
1040 | fiy1 = _mm_add_ps(fiy1,ty); |
1041 | fiz1 = _mm_add_ps(fiz1,tz); |
1042 | |
1043 | fjx1 = _mm_add_ps(fjx1,tx); |
1044 | fjy1 = _mm_add_ps(fjy1,ty); |
1045 | fjz1 = _mm_add_ps(fjz1,tz); |
1046 | |
1047 | } |
1048 | |
1049 | /************************** |
1050 | * CALCULATE INTERACTIONS * |
1051 | **************************/ |
1052 | |
1053 | if (gmx_mm_any_lt(rsq12,rcutoff2)) |
1054 | { |
1055 | |
1056 | r12 = _mm_mul_ps(rsq12,rinv12); |
1057 | r12 = _mm_andnot_ps(dummy_mask,r12); |
1058 | |
1059 | /* EWALD ELECTROSTATICS */ |
1060 | |
1061 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1062 | ewrt = _mm_mul_ps(r12,ewtabscale); |
1063 | ewitab = _mm_cvttps_epi32(ewrt); |
1064 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1065 | ewitab = _mm_slli_epi32(ewitab,2); |
1066 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
1067 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
1068 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
1069 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
1070 | _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); |
1071 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
1072 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
1073 | velec = _mm_mul_ps(qq12,_mm_sub_ps(rinv12,velec)); |
1074 | felec = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec)); |
1075 | |
1076 | d = _mm_sub_ps(r12,rswitch); |
1077 | d = _mm_max_ps(d,_mm_setzero_ps()); |
1078 | d2 = _mm_mul_ps(d,d); |
1079 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
1080 | |
1081 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
1082 | |
1083 | /* Evaluate switch function */ |
1084 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
1085 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv12,_mm_mul_ps(velec,dsw)) ); |
1086 | velec = _mm_mul_ps(velec,sw); |
1087 | cutoff_mask = _mm_cmplt_ps(rsq12,rcutoff2); |
1088 | |
1089 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
1090 | velec = _mm_and_ps(velec,cutoff_mask); |
1091 | velec = _mm_andnot_ps(dummy_mask,velec); |
1092 | velecsum = _mm_add_ps(velecsum,velec); |
1093 | |
1094 | fscal = felec; |
1095 | |
1096 | fscal = _mm_and_ps(fscal,cutoff_mask); |
1097 | |
1098 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
1099 | |
1100 | /* Calculate temporary vectorial force */ |
1101 | tx = _mm_mul_ps(fscal,dx12); |
1102 | ty = _mm_mul_ps(fscal,dy12); |
1103 | tz = _mm_mul_ps(fscal,dz12); |
1104 | |
1105 | /* Update vectorial force */ |
1106 | fix1 = _mm_add_ps(fix1,tx); |
1107 | fiy1 = _mm_add_ps(fiy1,ty); |
1108 | fiz1 = _mm_add_ps(fiz1,tz); |
1109 | |
1110 | fjx2 = _mm_add_ps(fjx2,tx); |
1111 | fjy2 = _mm_add_ps(fjy2,ty); |
1112 | fjz2 = _mm_add_ps(fjz2,tz); |
1113 | |
1114 | } |
1115 | |
1116 | /************************** |
1117 | * CALCULATE INTERACTIONS * |
1118 | **************************/ |
1119 | |
1120 | if (gmx_mm_any_lt(rsq13,rcutoff2)) |
1121 | { |
1122 | |
1123 | r13 = _mm_mul_ps(rsq13,rinv13); |
1124 | r13 = _mm_andnot_ps(dummy_mask,r13); |
1125 | |
1126 | /* EWALD ELECTROSTATICS */ |
1127 | |
1128 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1129 | ewrt = _mm_mul_ps(r13,ewtabscale); |
1130 | ewitab = _mm_cvttps_epi32(ewrt); |
1131 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1132 | ewitab = _mm_slli_epi32(ewitab,2); |
1133 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
1134 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
1135 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
1136 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
1137 | _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); |
1138 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
1139 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
1140 | velec = _mm_mul_ps(qq13,_mm_sub_ps(rinv13,velec)); |
1141 | felec = _mm_mul_ps(_mm_mul_ps(qq13,rinv13),_mm_sub_ps(rinvsq13,felec)); |
1142 | |
1143 | d = _mm_sub_ps(r13,rswitch); |
1144 | d = _mm_max_ps(d,_mm_setzero_ps()); |
1145 | d2 = _mm_mul_ps(d,d); |
1146 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
1147 | |
1148 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
1149 | |
1150 | /* Evaluate switch function */ |
1151 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
1152 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv13,_mm_mul_ps(velec,dsw)) ); |
1153 | velec = _mm_mul_ps(velec,sw); |
1154 | cutoff_mask = _mm_cmplt_ps(rsq13,rcutoff2); |
1155 | |
1156 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
1157 | velec = _mm_and_ps(velec,cutoff_mask); |
1158 | velec = _mm_andnot_ps(dummy_mask,velec); |
1159 | velecsum = _mm_add_ps(velecsum,velec); |
1160 | |
1161 | fscal = felec; |
1162 | |
1163 | fscal = _mm_and_ps(fscal,cutoff_mask); |
1164 | |
1165 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
1166 | |
1167 | /* Calculate temporary vectorial force */ |
1168 | tx = _mm_mul_ps(fscal,dx13); |
1169 | ty = _mm_mul_ps(fscal,dy13); |
1170 | tz = _mm_mul_ps(fscal,dz13); |
1171 | |
1172 | /* Update vectorial force */ |
1173 | fix1 = _mm_add_ps(fix1,tx); |
1174 | fiy1 = _mm_add_ps(fiy1,ty); |
1175 | fiz1 = _mm_add_ps(fiz1,tz); |
1176 | |
1177 | fjx3 = _mm_add_ps(fjx3,tx); |
1178 | fjy3 = _mm_add_ps(fjy3,ty); |
1179 | fjz3 = _mm_add_ps(fjz3,tz); |
1180 | |
1181 | } |
1182 | |
1183 | /************************** |
1184 | * CALCULATE INTERACTIONS * |
1185 | **************************/ |
1186 | |
1187 | if (gmx_mm_any_lt(rsq21,rcutoff2)) |
1188 | { |
1189 | |
1190 | r21 = _mm_mul_ps(rsq21,rinv21); |
1191 | r21 = _mm_andnot_ps(dummy_mask,r21); |
1192 | |
1193 | /* EWALD ELECTROSTATICS */ |
1194 | |
1195 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1196 | ewrt = _mm_mul_ps(r21,ewtabscale); |
1197 | ewitab = _mm_cvttps_epi32(ewrt); |
1198 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1199 | ewitab = _mm_slli_epi32(ewitab,2); |
1200 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
1201 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
1202 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
1203 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
1204 | _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); |
1205 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
1206 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
1207 | velec = _mm_mul_ps(qq21,_mm_sub_ps(rinv21,velec)); |
1208 | felec = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec)); |
1209 | |
1210 | d = _mm_sub_ps(r21,rswitch); |
1211 | d = _mm_max_ps(d,_mm_setzero_ps()); |
1212 | d2 = _mm_mul_ps(d,d); |
1213 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
1214 | |
1215 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
1216 | |
1217 | /* Evaluate switch function */ |
1218 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
1219 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv21,_mm_mul_ps(velec,dsw)) ); |
1220 | velec = _mm_mul_ps(velec,sw); |
1221 | cutoff_mask = _mm_cmplt_ps(rsq21,rcutoff2); |
1222 | |
1223 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
1224 | velec = _mm_and_ps(velec,cutoff_mask); |
1225 | velec = _mm_andnot_ps(dummy_mask,velec); |
1226 | velecsum = _mm_add_ps(velecsum,velec); |
1227 | |
1228 | fscal = felec; |
1229 | |
1230 | fscal = _mm_and_ps(fscal,cutoff_mask); |
1231 | |
1232 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
1233 | |
1234 | /* Calculate temporary vectorial force */ |
1235 | tx = _mm_mul_ps(fscal,dx21); |
1236 | ty = _mm_mul_ps(fscal,dy21); |
1237 | tz = _mm_mul_ps(fscal,dz21); |
1238 | |
1239 | /* Update vectorial force */ |
1240 | fix2 = _mm_add_ps(fix2,tx); |
1241 | fiy2 = _mm_add_ps(fiy2,ty); |
1242 | fiz2 = _mm_add_ps(fiz2,tz); |
1243 | |
1244 | fjx1 = _mm_add_ps(fjx1,tx); |
1245 | fjy1 = _mm_add_ps(fjy1,ty); |
1246 | fjz1 = _mm_add_ps(fjz1,tz); |
1247 | |
1248 | } |
1249 | |
1250 | /************************** |
1251 | * CALCULATE INTERACTIONS * |
1252 | **************************/ |
1253 | |
1254 | if (gmx_mm_any_lt(rsq22,rcutoff2)) |
1255 | { |
1256 | |
1257 | r22 = _mm_mul_ps(rsq22,rinv22); |
1258 | r22 = _mm_andnot_ps(dummy_mask,r22); |
1259 | |
1260 | /* EWALD ELECTROSTATICS */ |
1261 | |
1262 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1263 | ewrt = _mm_mul_ps(r22,ewtabscale); |
1264 | ewitab = _mm_cvttps_epi32(ewrt); |
1265 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1266 | ewitab = _mm_slli_epi32(ewitab,2); |
1267 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
1268 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
1269 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
1270 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
1271 | _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); |
1272 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
1273 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
1274 | velec = _mm_mul_ps(qq22,_mm_sub_ps(rinv22,velec)); |
1275 | felec = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec)); |
1276 | |
1277 | d = _mm_sub_ps(r22,rswitch); |
1278 | d = _mm_max_ps(d,_mm_setzero_ps()); |
1279 | d2 = _mm_mul_ps(d,d); |
1280 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
1281 | |
1282 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
1283 | |
1284 | /* Evaluate switch function */ |
1285 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
1286 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv22,_mm_mul_ps(velec,dsw)) ); |
1287 | velec = _mm_mul_ps(velec,sw); |
1288 | cutoff_mask = _mm_cmplt_ps(rsq22,rcutoff2); |
1289 | |
1290 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
1291 | velec = _mm_and_ps(velec,cutoff_mask); |
1292 | velec = _mm_andnot_ps(dummy_mask,velec); |
1293 | velecsum = _mm_add_ps(velecsum,velec); |
1294 | |
1295 | fscal = felec; |
1296 | |
1297 | fscal = _mm_and_ps(fscal,cutoff_mask); |
1298 | |
1299 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
1300 | |
1301 | /* Calculate temporary vectorial force */ |
1302 | tx = _mm_mul_ps(fscal,dx22); |
1303 | ty = _mm_mul_ps(fscal,dy22); |
1304 | tz = _mm_mul_ps(fscal,dz22); |
1305 | |
1306 | /* Update vectorial force */ |
1307 | fix2 = _mm_add_ps(fix2,tx); |
1308 | fiy2 = _mm_add_ps(fiy2,ty); |
1309 | fiz2 = _mm_add_ps(fiz2,tz); |
1310 | |
1311 | fjx2 = _mm_add_ps(fjx2,tx); |
1312 | fjy2 = _mm_add_ps(fjy2,ty); |
1313 | fjz2 = _mm_add_ps(fjz2,tz); |
1314 | |
1315 | } |
1316 | |
1317 | /************************** |
1318 | * CALCULATE INTERACTIONS * |
1319 | **************************/ |
1320 | |
1321 | if (gmx_mm_any_lt(rsq23,rcutoff2)) |
1322 | { |
1323 | |
1324 | r23 = _mm_mul_ps(rsq23,rinv23); |
1325 | r23 = _mm_andnot_ps(dummy_mask,r23); |
1326 | |
1327 | /* EWALD ELECTROSTATICS */ |
1328 | |
1329 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1330 | ewrt = _mm_mul_ps(r23,ewtabscale); |
1331 | ewitab = _mm_cvttps_epi32(ewrt); |
1332 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1333 | ewitab = _mm_slli_epi32(ewitab,2); |
1334 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
1335 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
1336 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
1337 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
1338 | _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); |
1339 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
1340 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
1341 | velec = _mm_mul_ps(qq23,_mm_sub_ps(rinv23,velec)); |
1342 | felec = _mm_mul_ps(_mm_mul_ps(qq23,rinv23),_mm_sub_ps(rinvsq23,felec)); |
1343 | |
1344 | d = _mm_sub_ps(r23,rswitch); |
1345 | d = _mm_max_ps(d,_mm_setzero_ps()); |
1346 | d2 = _mm_mul_ps(d,d); |
1347 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
1348 | |
1349 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
1350 | |
1351 | /* Evaluate switch function */ |
1352 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
1353 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv23,_mm_mul_ps(velec,dsw)) ); |
1354 | velec = _mm_mul_ps(velec,sw); |
1355 | cutoff_mask = _mm_cmplt_ps(rsq23,rcutoff2); |
1356 | |
1357 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
1358 | velec = _mm_and_ps(velec,cutoff_mask); |
1359 | velec = _mm_andnot_ps(dummy_mask,velec); |
1360 | velecsum = _mm_add_ps(velecsum,velec); |
1361 | |
1362 | fscal = felec; |
1363 | |
1364 | fscal = _mm_and_ps(fscal,cutoff_mask); |
1365 | |
1366 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
1367 | |
1368 | /* Calculate temporary vectorial force */ |
1369 | tx = _mm_mul_ps(fscal,dx23); |
1370 | ty = _mm_mul_ps(fscal,dy23); |
1371 | tz = _mm_mul_ps(fscal,dz23); |
1372 | |
1373 | /* Update vectorial force */ |
1374 | fix2 = _mm_add_ps(fix2,tx); |
1375 | fiy2 = _mm_add_ps(fiy2,ty); |
1376 | fiz2 = _mm_add_ps(fiz2,tz); |
1377 | |
1378 | fjx3 = _mm_add_ps(fjx3,tx); |
1379 | fjy3 = _mm_add_ps(fjy3,ty); |
1380 | fjz3 = _mm_add_ps(fjz3,tz); |
1381 | |
1382 | } |
1383 | |
1384 | /************************** |
1385 | * CALCULATE INTERACTIONS * |
1386 | **************************/ |
1387 | |
1388 | if (gmx_mm_any_lt(rsq31,rcutoff2)) |
1389 | { |
1390 | |
1391 | r31 = _mm_mul_ps(rsq31,rinv31); |
1392 | r31 = _mm_andnot_ps(dummy_mask,r31); |
1393 | |
1394 | /* EWALD ELECTROSTATICS */ |
1395 | |
1396 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1397 | ewrt = _mm_mul_ps(r31,ewtabscale); |
1398 | ewitab = _mm_cvttps_epi32(ewrt); |
1399 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1400 | ewitab = _mm_slli_epi32(ewitab,2); |
1401 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
1402 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
1403 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
1404 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
1405 | _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); |
1406 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
1407 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
1408 | velec = _mm_mul_ps(qq31,_mm_sub_ps(rinv31,velec)); |
1409 | felec = _mm_mul_ps(_mm_mul_ps(qq31,rinv31),_mm_sub_ps(rinvsq31,felec)); |
1410 | |
1411 | d = _mm_sub_ps(r31,rswitch); |
1412 | d = _mm_max_ps(d,_mm_setzero_ps()); |
1413 | d2 = _mm_mul_ps(d,d); |
1414 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
1415 | |
1416 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
1417 | |
1418 | /* Evaluate switch function */ |
1419 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
1420 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv31,_mm_mul_ps(velec,dsw)) ); |
1421 | velec = _mm_mul_ps(velec,sw); |
1422 | cutoff_mask = _mm_cmplt_ps(rsq31,rcutoff2); |
1423 | |
1424 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
1425 | velec = _mm_and_ps(velec,cutoff_mask); |
1426 | velec = _mm_andnot_ps(dummy_mask,velec); |
1427 | velecsum = _mm_add_ps(velecsum,velec); |
1428 | |
1429 | fscal = felec; |
1430 | |
1431 | fscal = _mm_and_ps(fscal,cutoff_mask); |
1432 | |
1433 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
1434 | |
1435 | /* Calculate temporary vectorial force */ |
1436 | tx = _mm_mul_ps(fscal,dx31); |
1437 | ty = _mm_mul_ps(fscal,dy31); |
1438 | tz = _mm_mul_ps(fscal,dz31); |
1439 | |
1440 | /* Update vectorial force */ |
1441 | fix3 = _mm_add_ps(fix3,tx); |
1442 | fiy3 = _mm_add_ps(fiy3,ty); |
1443 | fiz3 = _mm_add_ps(fiz3,tz); |
1444 | |
1445 | fjx1 = _mm_add_ps(fjx1,tx); |
1446 | fjy1 = _mm_add_ps(fjy1,ty); |
1447 | fjz1 = _mm_add_ps(fjz1,tz); |
1448 | |
1449 | } |
1450 | |
1451 | /************************** |
1452 | * CALCULATE INTERACTIONS * |
1453 | **************************/ |
1454 | |
1455 | if (gmx_mm_any_lt(rsq32,rcutoff2)) |
1456 | { |
1457 | |
1458 | r32 = _mm_mul_ps(rsq32,rinv32); |
1459 | r32 = _mm_andnot_ps(dummy_mask,r32); |
1460 | |
1461 | /* EWALD ELECTROSTATICS */ |
1462 | |
1463 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1464 | ewrt = _mm_mul_ps(r32,ewtabscale); |
1465 | ewitab = _mm_cvttps_epi32(ewrt); |
1466 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1467 | ewitab = _mm_slli_epi32(ewitab,2); |
1468 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
1469 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
1470 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
1471 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
1472 | _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); |
1473 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
1474 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
1475 | velec = _mm_mul_ps(qq32,_mm_sub_ps(rinv32,velec)); |
1476 | felec = _mm_mul_ps(_mm_mul_ps(qq32,rinv32),_mm_sub_ps(rinvsq32,felec)); |
1477 | |
1478 | d = _mm_sub_ps(r32,rswitch); |
1479 | d = _mm_max_ps(d,_mm_setzero_ps()); |
1480 | d2 = _mm_mul_ps(d,d); |
1481 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
1482 | |
1483 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
1484 | |
1485 | /* Evaluate switch function */ |
1486 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
1487 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv32,_mm_mul_ps(velec,dsw)) ); |
1488 | velec = _mm_mul_ps(velec,sw); |
1489 | cutoff_mask = _mm_cmplt_ps(rsq32,rcutoff2); |
1490 | |
1491 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
1492 | velec = _mm_and_ps(velec,cutoff_mask); |
1493 | velec = _mm_andnot_ps(dummy_mask,velec); |
1494 | velecsum = _mm_add_ps(velecsum,velec); |
1495 | |
1496 | fscal = felec; |
1497 | |
1498 | fscal = _mm_and_ps(fscal,cutoff_mask); |
1499 | |
1500 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
1501 | |
1502 | /* Calculate temporary vectorial force */ |
1503 | tx = _mm_mul_ps(fscal,dx32); |
1504 | ty = _mm_mul_ps(fscal,dy32); |
1505 | tz = _mm_mul_ps(fscal,dz32); |
1506 | |
1507 | /* Update vectorial force */ |
1508 | fix3 = _mm_add_ps(fix3,tx); |
1509 | fiy3 = _mm_add_ps(fiy3,ty); |
1510 | fiz3 = _mm_add_ps(fiz3,tz); |
1511 | |
1512 | fjx2 = _mm_add_ps(fjx2,tx); |
1513 | fjy2 = _mm_add_ps(fjy2,ty); |
1514 | fjz2 = _mm_add_ps(fjz2,tz); |
1515 | |
1516 | } |
1517 | |
1518 | /************************** |
1519 | * CALCULATE INTERACTIONS * |
1520 | **************************/ |
1521 | |
1522 | if (gmx_mm_any_lt(rsq33,rcutoff2)) |
1523 | { |
1524 | |
1525 | r33 = _mm_mul_ps(rsq33,rinv33); |
1526 | r33 = _mm_andnot_ps(dummy_mask,r33); |
1527 | |
1528 | /* EWALD ELECTROSTATICS */ |
1529 | |
1530 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1531 | ewrt = _mm_mul_ps(r33,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 | ewitab = _mm_slli_epi32(ewitab,2); |
1535 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
1536 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
1537 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
1538 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
1539 | _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); |
1540 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
1541 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
1542 | velec = _mm_mul_ps(qq33,_mm_sub_ps(rinv33,velec)); |
1543 | felec = _mm_mul_ps(_mm_mul_ps(qq33,rinv33),_mm_sub_ps(rinvsq33,felec)); |
1544 | |
1545 | d = _mm_sub_ps(r33,rswitch); |
1546 | d = _mm_max_ps(d,_mm_setzero_ps()); |
1547 | d2 = _mm_mul_ps(d,d); |
1548 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
1549 | |
1550 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
1551 | |
1552 | /* Evaluate switch function */ |
1553 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
1554 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv33,_mm_mul_ps(velec,dsw)) ); |
1555 | velec = _mm_mul_ps(velec,sw); |
1556 | cutoff_mask = _mm_cmplt_ps(rsq33,rcutoff2); |
1557 | |
1558 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
1559 | velec = _mm_and_ps(velec,cutoff_mask); |
1560 | velec = _mm_andnot_ps(dummy_mask,velec); |
1561 | velecsum = _mm_add_ps(velecsum,velec); |
1562 | |
1563 | fscal = felec; |
1564 | |
1565 | fscal = _mm_and_ps(fscal,cutoff_mask); |
1566 | |
1567 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
1568 | |
1569 | /* Calculate temporary vectorial force */ |
1570 | tx = _mm_mul_ps(fscal,dx33); |
1571 | ty = _mm_mul_ps(fscal,dy33); |
1572 | tz = _mm_mul_ps(fscal,dz33); |
1573 | |
1574 | /* Update vectorial force */ |
1575 | fix3 = _mm_add_ps(fix3,tx); |
1576 | fiy3 = _mm_add_ps(fiy3,ty); |
1577 | fiz3 = _mm_add_ps(fiz3,tz); |
1578 | |
1579 | fjx3 = _mm_add_ps(fjx3,tx); |
1580 | fjy3 = _mm_add_ps(fjy3,ty); |
1581 | fjz3 = _mm_add_ps(fjz3,tz); |
1582 | |
1583 | } |
1584 | |
1585 | fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch; |
1586 | fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch; |
1587 | fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch; |
1588 | fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch; |
1589 | |
1590 | gmx_mm_decrement_3rvec_4ptr_swizzle_ps(fjptrA+DIM3,fjptrB+DIM3,fjptrC+DIM3,fjptrD+DIM3, |
1591 | fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3); |
1592 | |
1593 | /* Inner loop uses 594 flops */ |
1594 | } |
1595 | |
1596 | /* End of innermost loop */ |
1597 | |
1598 | gmx_mm_update_iforce_3atom_swizzle_ps(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3, |
1599 | f+i_coord_offset+DIM3,fshift+i_shift_offset); |
1600 | |
1601 | ggid = gid[iidx]; |
1602 | /* Update potential energies */ |
1603 | gmx_mm_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid); |
1604 | |
1605 | /* Increment number of inner iterations */ |
1606 | inneriter += j_index_end - j_index_start; |
1607 | |
1608 | /* Outer loop uses 19 flops */ |
1609 | } |
1610 | |
1611 | /* Increment number of outer iterations */ |
1612 | outeriter += nri; |
1613 | |
1614 | /* Update outer/inner flops */ |
1615 | |
1616 | inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4W4_VF,outeriter*19 + inneriter*594)(nrnb)->n[eNR_NBKERNEL_ELEC_W4W4_VF] += outeriter*19 + inneriter *594; |
1617 | } |
1618 | /* |
1619 | * Gromacs nonbonded kernel: nb_kernel_ElecEwSw_VdwNone_GeomW4W4_F_sse4_1_single |
1620 | * Electrostatics interaction: Ewald |
1621 | * VdW interaction: None |
1622 | * Geometry: Water4-Water4 |
1623 | * Calculate force/pot: Force |
1624 | */ |
1625 | void |
1626 | nb_kernel_ElecEwSw_VdwNone_GeomW4W4_F_sse4_1_single |
1627 | (t_nblist * gmx_restrict nlist, |
1628 | rvec * gmx_restrict xx, |
1629 | rvec * gmx_restrict ff, |
1630 | t_forcerec * gmx_restrict fr, |
1631 | t_mdatoms * gmx_restrict mdatoms, |
1632 | nb_kernel_data_t gmx_unused__attribute__ ((unused)) * gmx_restrict kernel_data, |
1633 | t_nrnb * gmx_restrict nrnb) |
1634 | { |
1635 | /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or |
1636 | * just 0 for non-waters. |
1637 | * Suffixes A,B,C,D refer to j loop unrolling done with SSE, e.g. for the four different |
1638 | * jnr indices corresponding to data put in the four positions in the SIMD register. |
1639 | */ |
1640 | int i_shift_offset,i_coord_offset,outeriter,inneriter; |
1641 | int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx; |
1642 | int jnrA,jnrB,jnrC,jnrD; |
1643 | int jnrlistA,jnrlistB,jnrlistC,jnrlistD; |
1644 | int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD; |
1645 | int *iinr,*jindex,*jjnr,*shiftidx,*gid; |
1646 | real rcutoff_scalar; |
1647 | real *shiftvec,*fshift,*x,*f; |
1648 | real *fjptrA,*fjptrB,*fjptrC,*fjptrD; |
1649 | real scratch[4*DIM3]; |
1650 | __m128 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall; |
1651 | int vdwioffset1; |
1652 | __m128 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1; |
1653 | int vdwioffset2; |
1654 | __m128 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2; |
1655 | int vdwioffset3; |
1656 | __m128 ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3; |
1657 | int vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D; |
1658 | __m128 jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1; |
1659 | int vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D; |
1660 | __m128 jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2; |
1661 | int vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D; |
1662 | __m128 jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3; |
1663 | __m128 dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11; |
1664 | __m128 dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12; |
1665 | __m128 dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13; |
1666 | __m128 dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21; |
1667 | __m128 dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22; |
1668 | __m128 dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23; |
1669 | __m128 dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31; |
1670 | __m128 dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32; |
1671 | __m128 dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33; |
1672 | __m128 velec,felec,velecsum,facel,crf,krf,krf2; |
1673 | real *charge; |
1674 | __m128i ewitab; |
1675 | __m128 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV; |
1676 | real *ewtab; |
1677 | __m128 rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw; |
1678 | real rswitch_scalar,d_scalar; |
1679 | __m128 dummy_mask,cutoff_mask; |
1680 | __m128 signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) ); |
1681 | __m128 one = _mm_set1_ps(1.0); |
1682 | __m128 two = _mm_set1_ps(2.0); |
Value stored to 'two' during its initialization is never read | |
1683 | x = xx[0]; |
1684 | f = ff[0]; |
1685 | |
1686 | nri = nlist->nri; |
1687 | iinr = nlist->iinr; |
1688 | jindex = nlist->jindex; |
1689 | jjnr = nlist->jjnr; |
1690 | shiftidx = nlist->shift; |
1691 | gid = nlist->gid; |
1692 | shiftvec = fr->shift_vec[0]; |
1693 | fshift = fr->fshift[0]; |
1694 | facel = _mm_set1_ps(fr->epsfac); |
1695 | charge = mdatoms->chargeA; |
1696 | |
1697 | sh_ewald = _mm_set1_ps(fr->ic->sh_ewald); |
1698 | ewtab = fr->ic->tabq_coul_FDV0; |
1699 | ewtabscale = _mm_set1_ps(fr->ic->tabq_scale); |
1700 | ewtabhalfspace = _mm_set1_ps(0.5/fr->ic->tabq_scale); |
1701 | |
1702 | /* Setup water-specific parameters */ |
1703 | inr = nlist->iinr[0]; |
1704 | iq1 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1])); |
1705 | iq2 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2])); |
1706 | iq3 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+3])); |
1707 | |
1708 | jq1 = _mm_set1_ps(charge[inr+1]); |
1709 | jq2 = _mm_set1_ps(charge[inr+2]); |
1710 | jq3 = _mm_set1_ps(charge[inr+3]); |
1711 | qq11 = _mm_mul_ps(iq1,jq1); |
1712 | qq12 = _mm_mul_ps(iq1,jq2); |
1713 | qq13 = _mm_mul_ps(iq1,jq3); |
1714 | qq21 = _mm_mul_ps(iq2,jq1); |
1715 | qq22 = _mm_mul_ps(iq2,jq2); |
1716 | qq23 = _mm_mul_ps(iq2,jq3); |
1717 | qq31 = _mm_mul_ps(iq3,jq1); |
1718 | qq32 = _mm_mul_ps(iq3,jq2); |
1719 | qq33 = _mm_mul_ps(iq3,jq3); |
1720 | |
1721 | /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */ |
1722 | rcutoff_scalar = fr->rcoulomb; |
1723 | rcutoff = _mm_set1_ps(rcutoff_scalar); |
1724 | rcutoff2 = _mm_mul_ps(rcutoff,rcutoff); |
1725 | |
1726 | rswitch_scalar = fr->rcoulomb_switch; |
1727 | rswitch = _mm_set1_ps(rswitch_scalar); |
1728 | /* Setup switch parameters */ |
1729 | d_scalar = rcutoff_scalar-rswitch_scalar; |
1730 | d = _mm_set1_ps(d_scalar); |
1731 | swV3 = _mm_set1_ps(-10.0/(d_scalar*d_scalar*d_scalar)); |
1732 | swV4 = _mm_set1_ps( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar)); |
1733 | swV5 = _mm_set1_ps( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar)); |
1734 | swF2 = _mm_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar)); |
1735 | swF3 = _mm_set1_ps( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar)); |
1736 | swF4 = _mm_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar)); |
1737 | |
1738 | /* Avoid stupid compiler warnings */ |
1739 | jnrA = jnrB = jnrC = jnrD = 0; |
1740 | j_coord_offsetA = 0; |
1741 | j_coord_offsetB = 0; |
1742 | j_coord_offsetC = 0; |
1743 | j_coord_offsetD = 0; |
1744 | |
1745 | outeriter = 0; |
1746 | inneriter = 0; |
1747 | |
1748 | for(iidx=0;iidx<4*DIM3;iidx++) |
1749 | { |
1750 | scratch[iidx] = 0.0; |
1751 | } |
1752 | |
1753 | /* Start outer loop over neighborlists */ |
1754 | for(iidx=0; iidx<nri; iidx++) |
1755 | { |
1756 | /* Load shift vector for this list */ |
1757 | i_shift_offset = DIM3*shiftidx[iidx]; |
1758 | |
1759 | /* Load limits for loop over neighbors */ |
1760 | j_index_start = jindex[iidx]; |
1761 | j_index_end = jindex[iidx+1]; |
1762 | |
1763 | /* Get outer coordinate index */ |
1764 | inr = iinr[iidx]; |
1765 | i_coord_offset = DIM3*inr; |
1766 | |
1767 | /* Load i particle coords and add shift vector */ |
1768 | gmx_mm_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset+DIM3, |
1769 | &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3); |
1770 | |
1771 | fix1 = _mm_setzero_ps(); |
1772 | fiy1 = _mm_setzero_ps(); |
1773 | fiz1 = _mm_setzero_ps(); |
1774 | fix2 = _mm_setzero_ps(); |
1775 | fiy2 = _mm_setzero_ps(); |
1776 | fiz2 = _mm_setzero_ps(); |
1777 | fix3 = _mm_setzero_ps(); |
1778 | fiy3 = _mm_setzero_ps(); |
1779 | fiz3 = _mm_setzero_ps(); |
1780 | |
1781 | /* Start inner kernel loop */ |
1782 | for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4) |
1783 | { |
1784 | |
1785 | /* Get j neighbor index, and coordinate index */ |
1786 | jnrA = jjnr[jidx]; |
1787 | jnrB = jjnr[jidx+1]; |
1788 | jnrC = jjnr[jidx+2]; |
1789 | jnrD = jjnr[jidx+3]; |
1790 | j_coord_offsetA = DIM3*jnrA; |
1791 | j_coord_offsetB = DIM3*jnrB; |
1792 | j_coord_offsetC = DIM3*jnrC; |
1793 | j_coord_offsetD = DIM3*jnrD; |
1794 | |
1795 | /* load j atom coordinates */ |
1796 | gmx_mm_load_3rvec_4ptr_swizzle_ps(x+j_coord_offsetA+DIM3,x+j_coord_offsetB+DIM3, |
1797 | x+j_coord_offsetC+DIM3,x+j_coord_offsetD+DIM3, |
1798 | &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3); |
1799 | |
1800 | /* Calculate displacement vector */ |
1801 | dx11 = _mm_sub_ps(ix1,jx1); |
1802 | dy11 = _mm_sub_ps(iy1,jy1); |
1803 | dz11 = _mm_sub_ps(iz1,jz1); |
1804 | dx12 = _mm_sub_ps(ix1,jx2); |
1805 | dy12 = _mm_sub_ps(iy1,jy2); |
1806 | dz12 = _mm_sub_ps(iz1,jz2); |
1807 | dx13 = _mm_sub_ps(ix1,jx3); |
1808 | dy13 = _mm_sub_ps(iy1,jy3); |
1809 | dz13 = _mm_sub_ps(iz1,jz3); |
1810 | dx21 = _mm_sub_ps(ix2,jx1); |
1811 | dy21 = _mm_sub_ps(iy2,jy1); |
1812 | dz21 = _mm_sub_ps(iz2,jz1); |
1813 | dx22 = _mm_sub_ps(ix2,jx2); |
1814 | dy22 = _mm_sub_ps(iy2,jy2); |
1815 | dz22 = _mm_sub_ps(iz2,jz2); |
1816 | dx23 = _mm_sub_ps(ix2,jx3); |
1817 | dy23 = _mm_sub_ps(iy2,jy3); |
1818 | dz23 = _mm_sub_ps(iz2,jz3); |
1819 | dx31 = _mm_sub_ps(ix3,jx1); |
1820 | dy31 = _mm_sub_ps(iy3,jy1); |
1821 | dz31 = _mm_sub_ps(iz3,jz1); |
1822 | dx32 = _mm_sub_ps(ix3,jx2); |
1823 | dy32 = _mm_sub_ps(iy3,jy2); |
1824 | dz32 = _mm_sub_ps(iz3,jz2); |
1825 | dx33 = _mm_sub_ps(ix3,jx3); |
1826 | dy33 = _mm_sub_ps(iy3,jy3); |
1827 | dz33 = _mm_sub_ps(iz3,jz3); |
1828 | |
1829 | /* Calculate squared distance and things based on it */ |
1830 | rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11); |
1831 | rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12); |
1832 | rsq13 = gmx_mm_calc_rsq_ps(dx13,dy13,dz13); |
1833 | rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21); |
1834 | rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22); |
1835 | rsq23 = gmx_mm_calc_rsq_ps(dx23,dy23,dz23); |
1836 | rsq31 = gmx_mm_calc_rsq_ps(dx31,dy31,dz31); |
1837 | rsq32 = gmx_mm_calc_rsq_ps(dx32,dy32,dz32); |
1838 | rsq33 = gmx_mm_calc_rsq_ps(dx33,dy33,dz33); |
1839 | |
1840 | rinv11 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq11); |
1841 | rinv12 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq12); |
1842 | rinv13 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq13); |
1843 | rinv21 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq21); |
1844 | rinv22 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq22); |
1845 | rinv23 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq23); |
1846 | rinv31 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq31); |
1847 | rinv32 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq32); |
1848 | rinv33 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq33); |
1849 | |
1850 | rinvsq11 = _mm_mul_ps(rinv11,rinv11); |
1851 | rinvsq12 = _mm_mul_ps(rinv12,rinv12); |
1852 | rinvsq13 = _mm_mul_ps(rinv13,rinv13); |
1853 | rinvsq21 = _mm_mul_ps(rinv21,rinv21); |
1854 | rinvsq22 = _mm_mul_ps(rinv22,rinv22); |
1855 | rinvsq23 = _mm_mul_ps(rinv23,rinv23); |
1856 | rinvsq31 = _mm_mul_ps(rinv31,rinv31); |
1857 | rinvsq32 = _mm_mul_ps(rinv32,rinv32); |
1858 | rinvsq33 = _mm_mul_ps(rinv33,rinv33); |
1859 | |
1860 | fjx1 = _mm_setzero_ps(); |
1861 | fjy1 = _mm_setzero_ps(); |
1862 | fjz1 = _mm_setzero_ps(); |
1863 | fjx2 = _mm_setzero_ps(); |
1864 | fjy2 = _mm_setzero_ps(); |
1865 | fjz2 = _mm_setzero_ps(); |
1866 | fjx3 = _mm_setzero_ps(); |
1867 | fjy3 = _mm_setzero_ps(); |
1868 | fjz3 = _mm_setzero_ps(); |
1869 | |
1870 | /************************** |
1871 | * CALCULATE INTERACTIONS * |
1872 | **************************/ |
1873 | |
1874 | if (gmx_mm_any_lt(rsq11,rcutoff2)) |
1875 | { |
1876 | |
1877 | r11 = _mm_mul_ps(rsq11,rinv11); |
1878 | |
1879 | /* EWALD ELECTROSTATICS */ |
1880 | |
1881 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1882 | ewrt = _mm_mul_ps(r11,ewtabscale); |
1883 | ewitab = _mm_cvttps_epi32(ewrt); |
1884 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1885 | ewitab = _mm_slli_epi32(ewitab,2); |
1886 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
1887 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
1888 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
1889 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
1890 | _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); |
1891 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
1892 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
1893 | velec = _mm_mul_ps(qq11,_mm_sub_ps(rinv11,velec)); |
1894 | felec = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec)); |
1895 | |
1896 | d = _mm_sub_ps(r11,rswitch); |
1897 | d = _mm_max_ps(d,_mm_setzero_ps()); |
1898 | d2 = _mm_mul_ps(d,d); |
1899 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
1900 | |
1901 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
1902 | |
1903 | /* Evaluate switch function */ |
1904 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
1905 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv11,_mm_mul_ps(velec,dsw)) ); |
1906 | cutoff_mask = _mm_cmplt_ps(rsq11,rcutoff2); |
1907 | |
1908 | fscal = felec; |
1909 | |
1910 | fscal = _mm_and_ps(fscal,cutoff_mask); |
1911 | |
1912 | /* Calculate temporary vectorial force */ |
1913 | tx = _mm_mul_ps(fscal,dx11); |
1914 | ty = _mm_mul_ps(fscal,dy11); |
1915 | tz = _mm_mul_ps(fscal,dz11); |
1916 | |
1917 | /* Update vectorial force */ |
1918 | fix1 = _mm_add_ps(fix1,tx); |
1919 | fiy1 = _mm_add_ps(fiy1,ty); |
1920 | fiz1 = _mm_add_ps(fiz1,tz); |
1921 | |
1922 | fjx1 = _mm_add_ps(fjx1,tx); |
1923 | fjy1 = _mm_add_ps(fjy1,ty); |
1924 | fjz1 = _mm_add_ps(fjz1,tz); |
1925 | |
1926 | } |
1927 | |
1928 | /************************** |
1929 | * CALCULATE INTERACTIONS * |
1930 | **************************/ |
1931 | |
1932 | if (gmx_mm_any_lt(rsq12,rcutoff2)) |
1933 | { |
1934 | |
1935 | r12 = _mm_mul_ps(rsq12,rinv12); |
1936 | |
1937 | /* EWALD ELECTROSTATICS */ |
1938 | |
1939 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1940 | ewrt = _mm_mul_ps(r12,ewtabscale); |
1941 | ewitab = _mm_cvttps_epi32(ewrt); |
1942 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1943 | ewitab = _mm_slli_epi32(ewitab,2); |
1944 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
1945 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
1946 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
1947 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
1948 | _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); |
1949 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
1950 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
1951 | velec = _mm_mul_ps(qq12,_mm_sub_ps(rinv12,velec)); |
1952 | felec = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec)); |
1953 | |
1954 | d = _mm_sub_ps(r12,rswitch); |
1955 | d = _mm_max_ps(d,_mm_setzero_ps()); |
1956 | d2 = _mm_mul_ps(d,d); |
1957 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
1958 | |
1959 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
1960 | |
1961 | /* Evaluate switch function */ |
1962 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
1963 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv12,_mm_mul_ps(velec,dsw)) ); |
1964 | cutoff_mask = _mm_cmplt_ps(rsq12,rcutoff2); |
1965 | |
1966 | fscal = felec; |
1967 | |
1968 | fscal = _mm_and_ps(fscal,cutoff_mask); |
1969 | |
1970 | /* Calculate temporary vectorial force */ |
1971 | tx = _mm_mul_ps(fscal,dx12); |
1972 | ty = _mm_mul_ps(fscal,dy12); |
1973 | tz = _mm_mul_ps(fscal,dz12); |
1974 | |
1975 | /* Update vectorial force */ |
1976 | fix1 = _mm_add_ps(fix1,tx); |
1977 | fiy1 = _mm_add_ps(fiy1,ty); |
1978 | fiz1 = _mm_add_ps(fiz1,tz); |
1979 | |
1980 | fjx2 = _mm_add_ps(fjx2,tx); |
1981 | fjy2 = _mm_add_ps(fjy2,ty); |
1982 | fjz2 = _mm_add_ps(fjz2,tz); |
1983 | |
1984 | } |
1985 | |
1986 | /************************** |
1987 | * CALCULATE INTERACTIONS * |
1988 | **************************/ |
1989 | |
1990 | if (gmx_mm_any_lt(rsq13,rcutoff2)) |
1991 | { |
1992 | |
1993 | r13 = _mm_mul_ps(rsq13,rinv13); |
1994 | |
1995 | /* EWALD ELECTROSTATICS */ |
1996 | |
1997 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1998 | ewrt = _mm_mul_ps(r13,ewtabscale); |
1999 | ewitab = _mm_cvttps_epi32(ewrt); |
2000 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
2001 | ewitab = _mm_slli_epi32(ewitab,2); |
2002 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
2003 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
2004 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
2005 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
2006 | _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); |
2007 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
2008 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
2009 | velec = _mm_mul_ps(qq13,_mm_sub_ps(rinv13,velec)); |
2010 | felec = _mm_mul_ps(_mm_mul_ps(qq13,rinv13),_mm_sub_ps(rinvsq13,felec)); |
2011 | |
2012 | d = _mm_sub_ps(r13,rswitch); |
2013 | d = _mm_max_ps(d,_mm_setzero_ps()); |
2014 | d2 = _mm_mul_ps(d,d); |
2015 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
2016 | |
2017 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
2018 | |
2019 | /* Evaluate switch function */ |
2020 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
2021 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv13,_mm_mul_ps(velec,dsw)) ); |
2022 | cutoff_mask = _mm_cmplt_ps(rsq13,rcutoff2); |
2023 | |
2024 | fscal = felec; |
2025 | |
2026 | fscal = _mm_and_ps(fscal,cutoff_mask); |
2027 | |
2028 | /* Calculate temporary vectorial force */ |
2029 | tx = _mm_mul_ps(fscal,dx13); |
2030 | ty = _mm_mul_ps(fscal,dy13); |
2031 | tz = _mm_mul_ps(fscal,dz13); |
2032 | |
2033 | /* Update vectorial force */ |
2034 | fix1 = _mm_add_ps(fix1,tx); |
2035 | fiy1 = _mm_add_ps(fiy1,ty); |
2036 | fiz1 = _mm_add_ps(fiz1,tz); |
2037 | |
2038 | fjx3 = _mm_add_ps(fjx3,tx); |
2039 | fjy3 = _mm_add_ps(fjy3,ty); |
2040 | fjz3 = _mm_add_ps(fjz3,tz); |
2041 | |
2042 | } |
2043 | |
2044 | /************************** |
2045 | * CALCULATE INTERACTIONS * |
2046 | **************************/ |
2047 | |
2048 | if (gmx_mm_any_lt(rsq21,rcutoff2)) |
2049 | { |
2050 | |
2051 | r21 = _mm_mul_ps(rsq21,rinv21); |
2052 | |
2053 | /* EWALD ELECTROSTATICS */ |
2054 | |
2055 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
2056 | ewrt = _mm_mul_ps(r21,ewtabscale); |
2057 | ewitab = _mm_cvttps_epi32(ewrt); |
2058 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
2059 | ewitab = _mm_slli_epi32(ewitab,2); |
2060 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
2061 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
2062 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
2063 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
2064 | _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); |
2065 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
2066 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
2067 | velec = _mm_mul_ps(qq21,_mm_sub_ps(rinv21,velec)); |
2068 | felec = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec)); |
2069 | |
2070 | d = _mm_sub_ps(r21,rswitch); |
2071 | d = _mm_max_ps(d,_mm_setzero_ps()); |
2072 | d2 = _mm_mul_ps(d,d); |
2073 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
2074 | |
2075 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
2076 | |
2077 | /* Evaluate switch function */ |
2078 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
2079 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv21,_mm_mul_ps(velec,dsw)) ); |
2080 | cutoff_mask = _mm_cmplt_ps(rsq21,rcutoff2); |
2081 | |
2082 | fscal = felec; |
2083 | |
2084 | fscal = _mm_and_ps(fscal,cutoff_mask); |
2085 | |
2086 | /* Calculate temporary vectorial force */ |
2087 | tx = _mm_mul_ps(fscal,dx21); |
2088 | ty = _mm_mul_ps(fscal,dy21); |
2089 | tz = _mm_mul_ps(fscal,dz21); |
2090 | |
2091 | /* Update vectorial force */ |
2092 | fix2 = _mm_add_ps(fix2,tx); |
2093 | fiy2 = _mm_add_ps(fiy2,ty); |
2094 | fiz2 = _mm_add_ps(fiz2,tz); |
2095 | |
2096 | fjx1 = _mm_add_ps(fjx1,tx); |
2097 | fjy1 = _mm_add_ps(fjy1,ty); |
2098 | fjz1 = _mm_add_ps(fjz1,tz); |
2099 | |
2100 | } |
2101 | |
2102 | /************************** |
2103 | * CALCULATE INTERACTIONS * |
2104 | **************************/ |
2105 | |
2106 | if (gmx_mm_any_lt(rsq22,rcutoff2)) |
2107 | { |
2108 | |
2109 | r22 = _mm_mul_ps(rsq22,rinv22); |
2110 | |
2111 | /* EWALD ELECTROSTATICS */ |
2112 | |
2113 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
2114 | ewrt = _mm_mul_ps(r22,ewtabscale); |
2115 | ewitab = _mm_cvttps_epi32(ewrt); |
2116 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
2117 | ewitab = _mm_slli_epi32(ewitab,2); |
2118 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
2119 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
2120 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
2121 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
2122 | _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); |
2123 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
2124 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
2125 | velec = _mm_mul_ps(qq22,_mm_sub_ps(rinv22,velec)); |
2126 | felec = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec)); |
2127 | |
2128 | d = _mm_sub_ps(r22,rswitch); |
2129 | d = _mm_max_ps(d,_mm_setzero_ps()); |
2130 | d2 = _mm_mul_ps(d,d); |
2131 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
2132 | |
2133 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
2134 | |
2135 | /* Evaluate switch function */ |
2136 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
2137 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv22,_mm_mul_ps(velec,dsw)) ); |
2138 | cutoff_mask = _mm_cmplt_ps(rsq22,rcutoff2); |
2139 | |
2140 | fscal = felec; |
2141 | |
2142 | fscal = _mm_and_ps(fscal,cutoff_mask); |
2143 | |
2144 | /* Calculate temporary vectorial force */ |
2145 | tx = _mm_mul_ps(fscal,dx22); |
2146 | ty = _mm_mul_ps(fscal,dy22); |
2147 | tz = _mm_mul_ps(fscal,dz22); |
2148 | |
2149 | /* Update vectorial force */ |
2150 | fix2 = _mm_add_ps(fix2,tx); |
2151 | fiy2 = _mm_add_ps(fiy2,ty); |
2152 | fiz2 = _mm_add_ps(fiz2,tz); |
2153 | |
2154 | fjx2 = _mm_add_ps(fjx2,tx); |
2155 | fjy2 = _mm_add_ps(fjy2,ty); |
2156 | fjz2 = _mm_add_ps(fjz2,tz); |
2157 | |
2158 | } |
2159 | |
2160 | /************************** |
2161 | * CALCULATE INTERACTIONS * |
2162 | **************************/ |
2163 | |
2164 | if (gmx_mm_any_lt(rsq23,rcutoff2)) |
2165 | { |
2166 | |
2167 | r23 = _mm_mul_ps(rsq23,rinv23); |
2168 | |
2169 | /* EWALD ELECTROSTATICS */ |
2170 | |
2171 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
2172 | ewrt = _mm_mul_ps(r23,ewtabscale); |
2173 | ewitab = _mm_cvttps_epi32(ewrt); |
2174 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
2175 | ewitab = _mm_slli_epi32(ewitab,2); |
2176 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
2177 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
2178 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
2179 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
2180 | _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); |
2181 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
2182 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
2183 | velec = _mm_mul_ps(qq23,_mm_sub_ps(rinv23,velec)); |
2184 | felec = _mm_mul_ps(_mm_mul_ps(qq23,rinv23),_mm_sub_ps(rinvsq23,felec)); |
2185 | |
2186 | d = _mm_sub_ps(r23,rswitch); |
2187 | d = _mm_max_ps(d,_mm_setzero_ps()); |
2188 | d2 = _mm_mul_ps(d,d); |
2189 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
2190 | |
2191 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
2192 | |
2193 | /* Evaluate switch function */ |
2194 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
2195 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv23,_mm_mul_ps(velec,dsw)) ); |
2196 | cutoff_mask = _mm_cmplt_ps(rsq23,rcutoff2); |
2197 | |
2198 | fscal = felec; |
2199 | |
2200 | fscal = _mm_and_ps(fscal,cutoff_mask); |
2201 | |
2202 | /* Calculate temporary vectorial force */ |
2203 | tx = _mm_mul_ps(fscal,dx23); |
2204 | ty = _mm_mul_ps(fscal,dy23); |
2205 | tz = _mm_mul_ps(fscal,dz23); |
2206 | |
2207 | /* Update vectorial force */ |
2208 | fix2 = _mm_add_ps(fix2,tx); |
2209 | fiy2 = _mm_add_ps(fiy2,ty); |
2210 | fiz2 = _mm_add_ps(fiz2,tz); |
2211 | |
2212 | fjx3 = _mm_add_ps(fjx3,tx); |
2213 | fjy3 = _mm_add_ps(fjy3,ty); |
2214 | fjz3 = _mm_add_ps(fjz3,tz); |
2215 | |
2216 | } |
2217 | |
2218 | /************************** |
2219 | * CALCULATE INTERACTIONS * |
2220 | **************************/ |
2221 | |
2222 | if (gmx_mm_any_lt(rsq31,rcutoff2)) |
2223 | { |
2224 | |
2225 | r31 = _mm_mul_ps(rsq31,rinv31); |
2226 | |
2227 | /* EWALD ELECTROSTATICS */ |
2228 | |
2229 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
2230 | ewrt = _mm_mul_ps(r31,ewtabscale); |
2231 | ewitab = _mm_cvttps_epi32(ewrt); |
2232 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
2233 | ewitab = _mm_slli_epi32(ewitab,2); |
2234 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
2235 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
2236 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
2237 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
2238 | _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); |
2239 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
2240 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
2241 | velec = _mm_mul_ps(qq31,_mm_sub_ps(rinv31,velec)); |
2242 | felec = _mm_mul_ps(_mm_mul_ps(qq31,rinv31),_mm_sub_ps(rinvsq31,felec)); |
2243 | |
2244 | d = _mm_sub_ps(r31,rswitch); |
2245 | d = _mm_max_ps(d,_mm_setzero_ps()); |
2246 | d2 = _mm_mul_ps(d,d); |
2247 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
2248 | |
2249 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
2250 | |
2251 | /* Evaluate switch function */ |
2252 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
2253 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv31,_mm_mul_ps(velec,dsw)) ); |
2254 | cutoff_mask = _mm_cmplt_ps(rsq31,rcutoff2); |
2255 | |
2256 | fscal = felec; |
2257 | |
2258 | fscal = _mm_and_ps(fscal,cutoff_mask); |
2259 | |
2260 | /* Calculate temporary vectorial force */ |
2261 | tx = _mm_mul_ps(fscal,dx31); |
2262 | ty = _mm_mul_ps(fscal,dy31); |
2263 | tz = _mm_mul_ps(fscal,dz31); |
2264 | |
2265 | /* Update vectorial force */ |
2266 | fix3 = _mm_add_ps(fix3,tx); |
2267 | fiy3 = _mm_add_ps(fiy3,ty); |
2268 | fiz3 = _mm_add_ps(fiz3,tz); |
2269 | |
2270 | fjx1 = _mm_add_ps(fjx1,tx); |
2271 | fjy1 = _mm_add_ps(fjy1,ty); |
2272 | fjz1 = _mm_add_ps(fjz1,tz); |
2273 | |
2274 | } |
2275 | |
2276 | /************************** |
2277 | * CALCULATE INTERACTIONS * |
2278 | **************************/ |
2279 | |
2280 | if (gmx_mm_any_lt(rsq32,rcutoff2)) |
2281 | { |
2282 | |
2283 | r32 = _mm_mul_ps(rsq32,rinv32); |
2284 | |
2285 | /* EWALD ELECTROSTATICS */ |
2286 | |
2287 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
2288 | ewrt = _mm_mul_ps(r32,ewtabscale); |
2289 | ewitab = _mm_cvttps_epi32(ewrt); |
2290 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
2291 | ewitab = _mm_slli_epi32(ewitab,2); |
2292 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
2293 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
2294 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
2295 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
2296 | _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); |
2297 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
2298 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
2299 | velec = _mm_mul_ps(qq32,_mm_sub_ps(rinv32,velec)); |
2300 | felec = _mm_mul_ps(_mm_mul_ps(qq32,rinv32),_mm_sub_ps(rinvsq32,felec)); |
2301 | |
2302 | d = _mm_sub_ps(r32,rswitch); |
2303 | d = _mm_max_ps(d,_mm_setzero_ps()); |
2304 | d2 = _mm_mul_ps(d,d); |
2305 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
2306 | |
2307 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
2308 | |
2309 | /* Evaluate switch function */ |
2310 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
2311 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv32,_mm_mul_ps(velec,dsw)) ); |
2312 | cutoff_mask = _mm_cmplt_ps(rsq32,rcutoff2); |
2313 | |
2314 | fscal = felec; |
2315 | |
2316 | fscal = _mm_and_ps(fscal,cutoff_mask); |
2317 | |
2318 | /* Calculate temporary vectorial force */ |
2319 | tx = _mm_mul_ps(fscal,dx32); |
2320 | ty = _mm_mul_ps(fscal,dy32); |
2321 | tz = _mm_mul_ps(fscal,dz32); |
2322 | |
2323 | /* Update vectorial force */ |
2324 | fix3 = _mm_add_ps(fix3,tx); |
2325 | fiy3 = _mm_add_ps(fiy3,ty); |
2326 | fiz3 = _mm_add_ps(fiz3,tz); |
2327 | |
2328 | fjx2 = _mm_add_ps(fjx2,tx); |
2329 | fjy2 = _mm_add_ps(fjy2,ty); |
2330 | fjz2 = _mm_add_ps(fjz2,tz); |
2331 | |
2332 | } |
2333 | |
2334 | /************************** |
2335 | * CALCULATE INTERACTIONS * |
2336 | **************************/ |
2337 | |
2338 | if (gmx_mm_any_lt(rsq33,rcutoff2)) |
2339 | { |
2340 | |
2341 | r33 = _mm_mul_ps(rsq33,rinv33); |
2342 | |
2343 | /* EWALD ELECTROSTATICS */ |
2344 | |
2345 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
2346 | ewrt = _mm_mul_ps(r33,ewtabscale); |
2347 | ewitab = _mm_cvttps_epi32(ewrt); |
2348 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
2349 | ewitab = _mm_slli_epi32(ewitab,2); |
2350 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
2351 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
2352 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
2353 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
2354 | _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); |
2355 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
2356 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
2357 | velec = _mm_mul_ps(qq33,_mm_sub_ps(rinv33,velec)); |
2358 | felec = _mm_mul_ps(_mm_mul_ps(qq33,rinv33),_mm_sub_ps(rinvsq33,felec)); |
2359 | |
2360 | d = _mm_sub_ps(r33,rswitch); |
2361 | d = _mm_max_ps(d,_mm_setzero_ps()); |
2362 | d2 = _mm_mul_ps(d,d); |
2363 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
2364 | |
2365 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
2366 | |
2367 | /* Evaluate switch function */ |
2368 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
2369 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv33,_mm_mul_ps(velec,dsw)) ); |
2370 | cutoff_mask = _mm_cmplt_ps(rsq33,rcutoff2); |
2371 | |
2372 | fscal = felec; |
2373 | |
2374 | fscal = _mm_and_ps(fscal,cutoff_mask); |
2375 | |
2376 | /* Calculate temporary vectorial force */ |
2377 | tx = _mm_mul_ps(fscal,dx33); |
2378 | ty = _mm_mul_ps(fscal,dy33); |
2379 | tz = _mm_mul_ps(fscal,dz33); |
2380 | |
2381 | /* Update vectorial force */ |
2382 | fix3 = _mm_add_ps(fix3,tx); |
2383 | fiy3 = _mm_add_ps(fiy3,ty); |
2384 | fiz3 = _mm_add_ps(fiz3,tz); |
2385 | |
2386 | fjx3 = _mm_add_ps(fjx3,tx); |
2387 | fjy3 = _mm_add_ps(fjy3,ty); |
2388 | fjz3 = _mm_add_ps(fjz3,tz); |
2389 | |
2390 | } |
2391 | |
2392 | fjptrA = f+j_coord_offsetA; |
2393 | fjptrB = f+j_coord_offsetB; |
2394 | fjptrC = f+j_coord_offsetC; |
2395 | fjptrD = f+j_coord_offsetD; |
2396 | |
2397 | gmx_mm_decrement_3rvec_4ptr_swizzle_ps(fjptrA+DIM3,fjptrB+DIM3,fjptrC+DIM3,fjptrD+DIM3, |
2398 | fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3); |
2399 | |
2400 | /* Inner loop uses 558 flops */ |
2401 | } |
2402 | |
2403 | if(jidx<j_index_end) |
2404 | { |
2405 | |
2406 | /* Get j neighbor index, and coordinate index */ |
2407 | jnrlistA = jjnr[jidx]; |
2408 | jnrlistB = jjnr[jidx+1]; |
2409 | jnrlistC = jjnr[jidx+2]; |
2410 | jnrlistD = jjnr[jidx+3]; |
2411 | /* Sign of each element will be negative for non-real atoms. |
2412 | * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones, |
2413 | * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries. |
2414 | */ |
2415 | dummy_mask = gmx_mm_castsi128_ps_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())); |
2416 | jnrA = (jnrlistA>=0) ? jnrlistA : 0; |
2417 | jnrB = (jnrlistB>=0) ? jnrlistB : 0; |
2418 | jnrC = (jnrlistC>=0) ? jnrlistC : 0; |
2419 | jnrD = (jnrlistD>=0) ? jnrlistD : 0; |
2420 | j_coord_offsetA = DIM3*jnrA; |
2421 | j_coord_offsetB = DIM3*jnrB; |
2422 | j_coord_offsetC = DIM3*jnrC; |
2423 | j_coord_offsetD = DIM3*jnrD; |
2424 | |
2425 | /* load j atom coordinates */ |
2426 | gmx_mm_load_3rvec_4ptr_swizzle_ps(x+j_coord_offsetA+DIM3,x+j_coord_offsetB+DIM3, |
2427 | x+j_coord_offsetC+DIM3,x+j_coord_offsetD+DIM3, |
2428 | &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3); |
2429 | |
2430 | /* Calculate displacement vector */ |
2431 | dx11 = _mm_sub_ps(ix1,jx1); |
2432 | dy11 = _mm_sub_ps(iy1,jy1); |
2433 | dz11 = _mm_sub_ps(iz1,jz1); |
2434 | dx12 = _mm_sub_ps(ix1,jx2); |
2435 | dy12 = _mm_sub_ps(iy1,jy2); |
2436 | dz12 = _mm_sub_ps(iz1,jz2); |
2437 | dx13 = _mm_sub_ps(ix1,jx3); |
2438 | dy13 = _mm_sub_ps(iy1,jy3); |
2439 | dz13 = _mm_sub_ps(iz1,jz3); |
2440 | dx21 = _mm_sub_ps(ix2,jx1); |
2441 | dy21 = _mm_sub_ps(iy2,jy1); |
2442 | dz21 = _mm_sub_ps(iz2,jz1); |
2443 | dx22 = _mm_sub_ps(ix2,jx2); |
2444 | dy22 = _mm_sub_ps(iy2,jy2); |
2445 | dz22 = _mm_sub_ps(iz2,jz2); |
2446 | dx23 = _mm_sub_ps(ix2,jx3); |
2447 | dy23 = _mm_sub_ps(iy2,jy3); |
2448 | dz23 = _mm_sub_ps(iz2,jz3); |
2449 | dx31 = _mm_sub_ps(ix3,jx1); |
2450 | dy31 = _mm_sub_ps(iy3,jy1); |
2451 | dz31 = _mm_sub_ps(iz3,jz1); |
2452 | dx32 = _mm_sub_ps(ix3,jx2); |
2453 | dy32 = _mm_sub_ps(iy3,jy2); |
2454 | dz32 = _mm_sub_ps(iz3,jz2); |
2455 | dx33 = _mm_sub_ps(ix3,jx3); |
2456 | dy33 = _mm_sub_ps(iy3,jy3); |
2457 | dz33 = _mm_sub_ps(iz3,jz3); |
2458 | |
2459 | /* Calculate squared distance and things based on it */ |
2460 | rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11); |
2461 | rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12); |
2462 | rsq13 = gmx_mm_calc_rsq_ps(dx13,dy13,dz13); |
2463 | rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21); |
2464 | rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22); |
2465 | rsq23 = gmx_mm_calc_rsq_ps(dx23,dy23,dz23); |
2466 | rsq31 = gmx_mm_calc_rsq_ps(dx31,dy31,dz31); |
2467 | rsq32 = gmx_mm_calc_rsq_ps(dx32,dy32,dz32); |
2468 | rsq33 = gmx_mm_calc_rsq_ps(dx33,dy33,dz33); |
2469 | |
2470 | rinv11 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq11); |
2471 | rinv12 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq12); |
2472 | rinv13 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq13); |
2473 | rinv21 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq21); |
2474 | rinv22 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq22); |
2475 | rinv23 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq23); |
2476 | rinv31 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq31); |
2477 | rinv32 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq32); |
2478 | rinv33 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq33); |
2479 | |
2480 | rinvsq11 = _mm_mul_ps(rinv11,rinv11); |
2481 | rinvsq12 = _mm_mul_ps(rinv12,rinv12); |
2482 | rinvsq13 = _mm_mul_ps(rinv13,rinv13); |
2483 | rinvsq21 = _mm_mul_ps(rinv21,rinv21); |
2484 | rinvsq22 = _mm_mul_ps(rinv22,rinv22); |
2485 | rinvsq23 = _mm_mul_ps(rinv23,rinv23); |
2486 | rinvsq31 = _mm_mul_ps(rinv31,rinv31); |
2487 | rinvsq32 = _mm_mul_ps(rinv32,rinv32); |
2488 | rinvsq33 = _mm_mul_ps(rinv33,rinv33); |
2489 | |
2490 | fjx1 = _mm_setzero_ps(); |
2491 | fjy1 = _mm_setzero_ps(); |
2492 | fjz1 = _mm_setzero_ps(); |
2493 | fjx2 = _mm_setzero_ps(); |
2494 | fjy2 = _mm_setzero_ps(); |
2495 | fjz2 = _mm_setzero_ps(); |
2496 | fjx3 = _mm_setzero_ps(); |
2497 | fjy3 = _mm_setzero_ps(); |
2498 | fjz3 = _mm_setzero_ps(); |
2499 | |
2500 | /************************** |
2501 | * CALCULATE INTERACTIONS * |
2502 | **************************/ |
2503 | |
2504 | if (gmx_mm_any_lt(rsq11,rcutoff2)) |
2505 | { |
2506 | |
2507 | r11 = _mm_mul_ps(rsq11,rinv11); |
2508 | r11 = _mm_andnot_ps(dummy_mask,r11); |
2509 | |
2510 | /* EWALD ELECTROSTATICS */ |
2511 | |
2512 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
2513 | ewrt = _mm_mul_ps(r11,ewtabscale); |
2514 | ewitab = _mm_cvttps_epi32(ewrt); |
2515 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
2516 | ewitab = _mm_slli_epi32(ewitab,2); |
2517 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
2518 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
2519 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
2520 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
2521 | _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); |
2522 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
2523 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
2524 | velec = _mm_mul_ps(qq11,_mm_sub_ps(rinv11,velec)); |
2525 | felec = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec)); |
2526 | |
2527 | d = _mm_sub_ps(r11,rswitch); |
2528 | d = _mm_max_ps(d,_mm_setzero_ps()); |
2529 | d2 = _mm_mul_ps(d,d); |
2530 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
2531 | |
2532 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
2533 | |
2534 | /* Evaluate switch function */ |
2535 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
2536 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv11,_mm_mul_ps(velec,dsw)) ); |
2537 | cutoff_mask = _mm_cmplt_ps(rsq11,rcutoff2); |
2538 | |
2539 | fscal = felec; |
2540 | |
2541 | fscal = _mm_and_ps(fscal,cutoff_mask); |
2542 | |
2543 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
2544 | |
2545 | /* Calculate temporary vectorial force */ |
2546 | tx = _mm_mul_ps(fscal,dx11); |
2547 | ty = _mm_mul_ps(fscal,dy11); |
2548 | tz = _mm_mul_ps(fscal,dz11); |
2549 | |
2550 | /* Update vectorial force */ |
2551 | fix1 = _mm_add_ps(fix1,tx); |
2552 | fiy1 = _mm_add_ps(fiy1,ty); |
2553 | fiz1 = _mm_add_ps(fiz1,tz); |
2554 | |
2555 | fjx1 = _mm_add_ps(fjx1,tx); |
2556 | fjy1 = _mm_add_ps(fjy1,ty); |
2557 | fjz1 = _mm_add_ps(fjz1,tz); |
2558 | |
2559 | } |
2560 | |
2561 | /************************** |
2562 | * CALCULATE INTERACTIONS * |
2563 | **************************/ |
2564 | |
2565 | if (gmx_mm_any_lt(rsq12,rcutoff2)) |
2566 | { |
2567 | |
2568 | r12 = _mm_mul_ps(rsq12,rinv12); |
2569 | r12 = _mm_andnot_ps(dummy_mask,r12); |
2570 | |
2571 | /* EWALD ELECTROSTATICS */ |
2572 | |
2573 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
2574 | ewrt = _mm_mul_ps(r12,ewtabscale); |
2575 | ewitab = _mm_cvttps_epi32(ewrt); |
2576 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
2577 | ewitab = _mm_slli_epi32(ewitab,2); |
2578 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
2579 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
2580 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
2581 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
2582 | _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); |
2583 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
2584 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
2585 | velec = _mm_mul_ps(qq12,_mm_sub_ps(rinv12,velec)); |
2586 | felec = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec)); |
2587 | |
2588 | d = _mm_sub_ps(r12,rswitch); |
2589 | d = _mm_max_ps(d,_mm_setzero_ps()); |
2590 | d2 = _mm_mul_ps(d,d); |
2591 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
2592 | |
2593 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
2594 | |
2595 | /* Evaluate switch function */ |
2596 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
2597 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv12,_mm_mul_ps(velec,dsw)) ); |
2598 | cutoff_mask = _mm_cmplt_ps(rsq12,rcutoff2); |
2599 | |
2600 | fscal = felec; |
2601 | |
2602 | fscal = _mm_and_ps(fscal,cutoff_mask); |
2603 | |
2604 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
2605 | |
2606 | /* Calculate temporary vectorial force */ |
2607 | tx = _mm_mul_ps(fscal,dx12); |
2608 | ty = _mm_mul_ps(fscal,dy12); |
2609 | tz = _mm_mul_ps(fscal,dz12); |
2610 | |
2611 | /* Update vectorial force */ |
2612 | fix1 = _mm_add_ps(fix1,tx); |
2613 | fiy1 = _mm_add_ps(fiy1,ty); |
2614 | fiz1 = _mm_add_ps(fiz1,tz); |
2615 | |
2616 | fjx2 = _mm_add_ps(fjx2,tx); |
2617 | fjy2 = _mm_add_ps(fjy2,ty); |
2618 | fjz2 = _mm_add_ps(fjz2,tz); |
2619 | |
2620 | } |
2621 | |
2622 | /************************** |
2623 | * CALCULATE INTERACTIONS * |
2624 | **************************/ |
2625 | |
2626 | if (gmx_mm_any_lt(rsq13,rcutoff2)) |
2627 | { |
2628 | |
2629 | r13 = _mm_mul_ps(rsq13,rinv13); |
2630 | r13 = _mm_andnot_ps(dummy_mask,r13); |
2631 | |
2632 | /* EWALD ELECTROSTATICS */ |
2633 | |
2634 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
2635 | ewrt = _mm_mul_ps(r13,ewtabscale); |
2636 | ewitab = _mm_cvttps_epi32(ewrt); |
2637 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
2638 | ewitab = _mm_slli_epi32(ewitab,2); |
2639 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
2640 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
2641 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
2642 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
2643 | _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); |
2644 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
2645 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
2646 | velec = _mm_mul_ps(qq13,_mm_sub_ps(rinv13,velec)); |
2647 | felec = _mm_mul_ps(_mm_mul_ps(qq13,rinv13),_mm_sub_ps(rinvsq13,felec)); |
2648 | |
2649 | d = _mm_sub_ps(r13,rswitch); |
2650 | d = _mm_max_ps(d,_mm_setzero_ps()); |
2651 | d2 = _mm_mul_ps(d,d); |
2652 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
2653 | |
2654 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
2655 | |
2656 | /* Evaluate switch function */ |
2657 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
2658 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv13,_mm_mul_ps(velec,dsw)) ); |
2659 | cutoff_mask = _mm_cmplt_ps(rsq13,rcutoff2); |
2660 | |
2661 | fscal = felec; |
2662 | |
2663 | fscal = _mm_and_ps(fscal,cutoff_mask); |
2664 | |
2665 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
2666 | |
2667 | /* Calculate temporary vectorial force */ |
2668 | tx = _mm_mul_ps(fscal,dx13); |
2669 | ty = _mm_mul_ps(fscal,dy13); |
2670 | tz = _mm_mul_ps(fscal,dz13); |
2671 | |
2672 | /* Update vectorial force */ |
2673 | fix1 = _mm_add_ps(fix1,tx); |
2674 | fiy1 = _mm_add_ps(fiy1,ty); |
2675 | fiz1 = _mm_add_ps(fiz1,tz); |
2676 | |
2677 | fjx3 = _mm_add_ps(fjx3,tx); |
2678 | fjy3 = _mm_add_ps(fjy3,ty); |
2679 | fjz3 = _mm_add_ps(fjz3,tz); |
2680 | |
2681 | } |
2682 | |
2683 | /************************** |
2684 | * CALCULATE INTERACTIONS * |
2685 | **************************/ |
2686 | |
2687 | if (gmx_mm_any_lt(rsq21,rcutoff2)) |
2688 | { |
2689 | |
2690 | r21 = _mm_mul_ps(rsq21,rinv21); |
2691 | r21 = _mm_andnot_ps(dummy_mask,r21); |
2692 | |
2693 | /* EWALD ELECTROSTATICS */ |
2694 | |
2695 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
2696 | ewrt = _mm_mul_ps(r21,ewtabscale); |
2697 | ewitab = _mm_cvttps_epi32(ewrt); |
2698 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
2699 | ewitab = _mm_slli_epi32(ewitab,2); |
2700 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
2701 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
2702 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
2703 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
2704 | _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); |
2705 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
2706 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
2707 | velec = _mm_mul_ps(qq21,_mm_sub_ps(rinv21,velec)); |
2708 | felec = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec)); |
2709 | |
2710 | d = _mm_sub_ps(r21,rswitch); |
2711 | d = _mm_max_ps(d,_mm_setzero_ps()); |
2712 | d2 = _mm_mul_ps(d,d); |
2713 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
2714 | |
2715 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
2716 | |
2717 | /* Evaluate switch function */ |
2718 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
2719 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv21,_mm_mul_ps(velec,dsw)) ); |
2720 | cutoff_mask = _mm_cmplt_ps(rsq21,rcutoff2); |
2721 | |
2722 | fscal = felec; |
2723 | |
2724 | fscal = _mm_and_ps(fscal,cutoff_mask); |
2725 | |
2726 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
2727 | |
2728 | /* Calculate temporary vectorial force */ |
2729 | tx = _mm_mul_ps(fscal,dx21); |
2730 | ty = _mm_mul_ps(fscal,dy21); |
2731 | tz = _mm_mul_ps(fscal,dz21); |
2732 | |
2733 | /* Update vectorial force */ |
2734 | fix2 = _mm_add_ps(fix2,tx); |
2735 | fiy2 = _mm_add_ps(fiy2,ty); |
2736 | fiz2 = _mm_add_ps(fiz2,tz); |
2737 | |
2738 | fjx1 = _mm_add_ps(fjx1,tx); |
2739 | fjy1 = _mm_add_ps(fjy1,ty); |
2740 | fjz1 = _mm_add_ps(fjz1,tz); |
2741 | |
2742 | } |
2743 | |
2744 | /************************** |
2745 | * CALCULATE INTERACTIONS * |
2746 | **************************/ |
2747 | |
2748 | if (gmx_mm_any_lt(rsq22,rcutoff2)) |
2749 | { |
2750 | |
2751 | r22 = _mm_mul_ps(rsq22,rinv22); |
2752 | r22 = _mm_andnot_ps(dummy_mask,r22); |
2753 | |
2754 | /* EWALD ELECTROSTATICS */ |
2755 | |
2756 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
2757 | ewrt = _mm_mul_ps(r22,ewtabscale); |
2758 | ewitab = _mm_cvttps_epi32(ewrt); |
2759 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
2760 | ewitab = _mm_slli_epi32(ewitab,2); |
2761 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
2762 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
2763 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
2764 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
2765 | _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); |
2766 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
2767 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
2768 | velec = _mm_mul_ps(qq22,_mm_sub_ps(rinv22,velec)); |
2769 | felec = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec)); |
2770 | |
2771 | d = _mm_sub_ps(r22,rswitch); |
2772 | d = _mm_max_ps(d,_mm_setzero_ps()); |
2773 | d2 = _mm_mul_ps(d,d); |
2774 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
2775 | |
2776 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
2777 | |
2778 | /* Evaluate switch function */ |
2779 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
2780 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv22,_mm_mul_ps(velec,dsw)) ); |
2781 | cutoff_mask = _mm_cmplt_ps(rsq22,rcutoff2); |
2782 | |
2783 | fscal = felec; |
2784 | |
2785 | fscal = _mm_and_ps(fscal,cutoff_mask); |
2786 | |
2787 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
2788 | |
2789 | /* Calculate temporary vectorial force */ |
2790 | tx = _mm_mul_ps(fscal,dx22); |
2791 | ty = _mm_mul_ps(fscal,dy22); |
2792 | tz = _mm_mul_ps(fscal,dz22); |
2793 | |
2794 | /* Update vectorial force */ |
2795 | fix2 = _mm_add_ps(fix2,tx); |
2796 | fiy2 = _mm_add_ps(fiy2,ty); |
2797 | fiz2 = _mm_add_ps(fiz2,tz); |
2798 | |
2799 | fjx2 = _mm_add_ps(fjx2,tx); |
2800 | fjy2 = _mm_add_ps(fjy2,ty); |
2801 | fjz2 = _mm_add_ps(fjz2,tz); |
2802 | |
2803 | } |
2804 | |
2805 | /************************** |
2806 | * CALCULATE INTERACTIONS * |
2807 | **************************/ |
2808 | |
2809 | if (gmx_mm_any_lt(rsq23,rcutoff2)) |
2810 | { |
2811 | |
2812 | r23 = _mm_mul_ps(rsq23,rinv23); |
2813 | r23 = _mm_andnot_ps(dummy_mask,r23); |
2814 | |
2815 | /* EWALD ELECTROSTATICS */ |
2816 | |
2817 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
2818 | ewrt = _mm_mul_ps(r23,ewtabscale); |
2819 | ewitab = _mm_cvttps_epi32(ewrt); |
2820 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
2821 | ewitab = _mm_slli_epi32(ewitab,2); |
2822 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
2823 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
2824 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
2825 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
2826 | _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); |
2827 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
2828 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
2829 | velec = _mm_mul_ps(qq23,_mm_sub_ps(rinv23,velec)); |
2830 | felec = _mm_mul_ps(_mm_mul_ps(qq23,rinv23),_mm_sub_ps(rinvsq23,felec)); |
2831 | |
2832 | d = _mm_sub_ps(r23,rswitch); |
2833 | d = _mm_max_ps(d,_mm_setzero_ps()); |
2834 | d2 = _mm_mul_ps(d,d); |
2835 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
2836 | |
2837 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
2838 | |
2839 | /* Evaluate switch function */ |
2840 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
2841 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv23,_mm_mul_ps(velec,dsw)) ); |
2842 | cutoff_mask = _mm_cmplt_ps(rsq23,rcutoff2); |
2843 | |
2844 | fscal = felec; |
2845 | |
2846 | fscal = _mm_and_ps(fscal,cutoff_mask); |
2847 | |
2848 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
2849 | |
2850 | /* Calculate temporary vectorial force */ |
2851 | tx = _mm_mul_ps(fscal,dx23); |
2852 | ty = _mm_mul_ps(fscal,dy23); |
2853 | tz = _mm_mul_ps(fscal,dz23); |
2854 | |
2855 | /* Update vectorial force */ |
2856 | fix2 = _mm_add_ps(fix2,tx); |
2857 | fiy2 = _mm_add_ps(fiy2,ty); |
2858 | fiz2 = _mm_add_ps(fiz2,tz); |
2859 | |
2860 | fjx3 = _mm_add_ps(fjx3,tx); |
2861 | fjy3 = _mm_add_ps(fjy3,ty); |
2862 | fjz3 = _mm_add_ps(fjz3,tz); |
2863 | |
2864 | } |
2865 | |
2866 | /************************** |
2867 | * CALCULATE INTERACTIONS * |
2868 | **************************/ |
2869 | |
2870 | if (gmx_mm_any_lt(rsq31,rcutoff2)) |
2871 | { |
2872 | |
2873 | r31 = _mm_mul_ps(rsq31,rinv31); |
2874 | r31 = _mm_andnot_ps(dummy_mask,r31); |
2875 | |
2876 | /* EWALD ELECTROSTATICS */ |
2877 | |
2878 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
2879 | ewrt = _mm_mul_ps(r31,ewtabscale); |
2880 | ewitab = _mm_cvttps_epi32(ewrt); |
2881 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
2882 | ewitab = _mm_slli_epi32(ewitab,2); |
2883 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
2884 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
2885 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
2886 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
2887 | _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); |
2888 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
2889 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
2890 | velec = _mm_mul_ps(qq31,_mm_sub_ps(rinv31,velec)); |
2891 | felec = _mm_mul_ps(_mm_mul_ps(qq31,rinv31),_mm_sub_ps(rinvsq31,felec)); |
2892 | |
2893 | d = _mm_sub_ps(r31,rswitch); |
2894 | d = _mm_max_ps(d,_mm_setzero_ps()); |
2895 | d2 = _mm_mul_ps(d,d); |
2896 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
2897 | |
2898 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
2899 | |
2900 | /* Evaluate switch function */ |
2901 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
2902 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv31,_mm_mul_ps(velec,dsw)) ); |
2903 | cutoff_mask = _mm_cmplt_ps(rsq31,rcutoff2); |
2904 | |
2905 | fscal = felec; |
2906 | |
2907 | fscal = _mm_and_ps(fscal,cutoff_mask); |
2908 | |
2909 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
2910 | |
2911 | /* Calculate temporary vectorial force */ |
2912 | tx = _mm_mul_ps(fscal,dx31); |
2913 | ty = _mm_mul_ps(fscal,dy31); |
2914 | tz = _mm_mul_ps(fscal,dz31); |
2915 | |
2916 | /* Update vectorial force */ |
2917 | fix3 = _mm_add_ps(fix3,tx); |
2918 | fiy3 = _mm_add_ps(fiy3,ty); |
2919 | fiz3 = _mm_add_ps(fiz3,tz); |
2920 | |
2921 | fjx1 = _mm_add_ps(fjx1,tx); |
2922 | fjy1 = _mm_add_ps(fjy1,ty); |
2923 | fjz1 = _mm_add_ps(fjz1,tz); |
2924 | |
2925 | } |
2926 | |
2927 | /************************** |
2928 | * CALCULATE INTERACTIONS * |
2929 | **************************/ |
2930 | |
2931 | if (gmx_mm_any_lt(rsq32,rcutoff2)) |
2932 | { |
2933 | |
2934 | r32 = _mm_mul_ps(rsq32,rinv32); |
2935 | r32 = _mm_andnot_ps(dummy_mask,r32); |
2936 | |
2937 | /* EWALD ELECTROSTATICS */ |
2938 | |
2939 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
2940 | ewrt = _mm_mul_ps(r32,ewtabscale); |
2941 | ewitab = _mm_cvttps_epi32(ewrt); |
2942 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
2943 | ewitab = _mm_slli_epi32(ewitab,2); |
2944 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
2945 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
2946 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
2947 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
2948 | _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); |
2949 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
2950 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
2951 | velec = _mm_mul_ps(qq32,_mm_sub_ps(rinv32,velec)); |
2952 | felec = _mm_mul_ps(_mm_mul_ps(qq32,rinv32),_mm_sub_ps(rinvsq32,felec)); |
2953 | |
2954 | d = _mm_sub_ps(r32,rswitch); |
2955 | d = _mm_max_ps(d,_mm_setzero_ps()); |
2956 | d2 = _mm_mul_ps(d,d); |
2957 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
2958 | |
2959 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
2960 | |
2961 | /* Evaluate switch function */ |
2962 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
2963 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv32,_mm_mul_ps(velec,dsw)) ); |
2964 | cutoff_mask = _mm_cmplt_ps(rsq32,rcutoff2); |
2965 | |
2966 | fscal = felec; |
2967 | |
2968 | fscal = _mm_and_ps(fscal,cutoff_mask); |
2969 | |
2970 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
2971 | |
2972 | /* Calculate temporary vectorial force */ |
2973 | tx = _mm_mul_ps(fscal,dx32); |
2974 | ty = _mm_mul_ps(fscal,dy32); |
2975 | tz = _mm_mul_ps(fscal,dz32); |
2976 | |
2977 | /* Update vectorial force */ |
2978 | fix3 = _mm_add_ps(fix3,tx); |
2979 | fiy3 = _mm_add_ps(fiy3,ty); |
2980 | fiz3 = _mm_add_ps(fiz3,tz); |
2981 | |
2982 | fjx2 = _mm_add_ps(fjx2,tx); |
2983 | fjy2 = _mm_add_ps(fjy2,ty); |
2984 | fjz2 = _mm_add_ps(fjz2,tz); |
2985 | |
2986 | } |
2987 | |
2988 | /************************** |
2989 | * CALCULATE INTERACTIONS * |
2990 | **************************/ |
2991 | |
2992 | if (gmx_mm_any_lt(rsq33,rcutoff2)) |
2993 | { |
2994 | |
2995 | r33 = _mm_mul_ps(rsq33,rinv33); |
2996 | r33 = _mm_andnot_ps(dummy_mask,r33); |
2997 | |
2998 | /* EWALD ELECTROSTATICS */ |
2999 | |
3000 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
3001 | ewrt = _mm_mul_ps(r33,ewtabscale); |
3002 | ewitab = _mm_cvttps_epi32(ewrt); |
3003 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
3004 | ewitab = _mm_slli_epi32(ewitab,2); |
3005 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
3006 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
3007 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
3008 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
3009 | _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); |
3010 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
3011 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
3012 | velec = _mm_mul_ps(qq33,_mm_sub_ps(rinv33,velec)); |
3013 | felec = _mm_mul_ps(_mm_mul_ps(qq33,rinv33),_mm_sub_ps(rinvsq33,felec)); |
3014 | |
3015 | d = _mm_sub_ps(r33,rswitch); |
3016 | d = _mm_max_ps(d,_mm_setzero_ps()); |
3017 | d2 = _mm_mul_ps(d,d); |
3018 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
3019 | |
3020 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
3021 | |
3022 | /* Evaluate switch function */ |
3023 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
3024 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv33,_mm_mul_ps(velec,dsw)) ); |
3025 | cutoff_mask = _mm_cmplt_ps(rsq33,rcutoff2); |
3026 | |
3027 | fscal = felec; |
3028 | |
3029 | fscal = _mm_and_ps(fscal,cutoff_mask); |
3030 | |
3031 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
3032 | |
3033 | /* Calculate temporary vectorial force */ |
3034 | tx = _mm_mul_ps(fscal,dx33); |
3035 | ty = _mm_mul_ps(fscal,dy33); |
3036 | tz = _mm_mul_ps(fscal,dz33); |
3037 | |
3038 | /* Update vectorial force */ |
3039 | fix3 = _mm_add_ps(fix3,tx); |
3040 | fiy3 = _mm_add_ps(fiy3,ty); |
3041 | fiz3 = _mm_add_ps(fiz3,tz); |
3042 | |
3043 | fjx3 = _mm_add_ps(fjx3,tx); |
3044 | fjy3 = _mm_add_ps(fjy3,ty); |
3045 | fjz3 = _mm_add_ps(fjz3,tz); |
3046 | |
3047 | } |
3048 | |
3049 | fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch; |
3050 | fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch; |
3051 | fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch; |
3052 | fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch; |
3053 | |
3054 | gmx_mm_decrement_3rvec_4ptr_swizzle_ps(fjptrA+DIM3,fjptrB+DIM3,fjptrC+DIM3,fjptrD+DIM3, |
3055 | fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3); |
3056 | |
3057 | /* Inner loop uses 567 flops */ |
3058 | } |
3059 | |
3060 | /* End of innermost loop */ |
3061 | |
3062 | gmx_mm_update_iforce_3atom_swizzle_ps(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3, |
3063 | f+i_coord_offset+DIM3,fshift+i_shift_offset); |
3064 | |
3065 | /* Increment number of inner iterations */ |
3066 | inneriter += j_index_end - j_index_start; |
3067 | |
3068 | /* Outer loop uses 18 flops */ |
3069 | } |
3070 | |
3071 | /* Increment number of outer iterations */ |
3072 | outeriter += nri; |
3073 | |
3074 | /* Update outer/inner flops */ |
3075 | |
3076 | inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4W4_F,outeriter*18 + inneriter*567)(nrnb)->n[eNR_NBKERNEL_ELEC_W4W4_F] += outeriter*18 + inneriter *567; |
3077 | } |