File: | gromacs/gmxlib/nonbonded/nb_kernel_sse4_1_single/nb_kernel_ElecEwSw_VdwLJSw_GeomW3W3_sse4_1_single.c |
Location: | line 189, column 5 |
Description: | Value stored to 'j_coord_offsetC' 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_VdwLJSw_GeomW3W3_VF_sse4_1_single |
54 | * Electrostatics interaction: Ewald |
55 | * VdW interaction: LennardJones |
56 | * Geometry: Water3-Water3 |
57 | * Calculate force/pot: PotentialAndForce |
58 | */ |
59 | void |
60 | nb_kernel_ElecEwSw_VdwLJSw_GeomW3W3_VF_sse4_1_single |
61 | (t_nblist * gmx_restrict nlist, |
62 | rvec * gmx_restrict xx, |
63 | rvec * gmx_restrict ff, |
64 | t_forcerec * gmx_restrict fr, |
65 | t_mdatoms * gmx_restrict mdatoms, |
66 | nb_kernel_data_t gmx_unused__attribute__ ((unused)) * gmx_restrict kernel_data, |
67 | t_nrnb * gmx_restrict nrnb) |
68 | { |
69 | /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or |
70 | * just 0 for non-waters. |
71 | * Suffixes A,B,C,D refer to j loop unrolling done with SSE, e.g. for the four different |
72 | * jnr indices corresponding to data put in the four positions in the SIMD register. |
73 | */ |
74 | int i_shift_offset,i_coord_offset,outeriter,inneriter; |
75 | int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx; |
76 | int jnrA,jnrB,jnrC,jnrD; |
77 | int jnrlistA,jnrlistB,jnrlistC,jnrlistD; |
78 | int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD; |
79 | int *iinr,*jindex,*jjnr,*shiftidx,*gid; |
80 | real rcutoff_scalar; |
81 | real *shiftvec,*fshift,*x,*f; |
82 | real *fjptrA,*fjptrB,*fjptrC,*fjptrD; |
83 | real scratch[4*DIM3]; |
84 | __m128 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall; |
85 | int vdwioffset0; |
86 | __m128 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0; |
87 | int vdwioffset1; |
88 | __m128 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1; |
89 | int vdwioffset2; |
90 | __m128 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2; |
91 | int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D; |
92 | __m128 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0; |
93 | int vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D; |
94 | __m128 jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1; |
95 | int vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D; |
96 | __m128 jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2; |
97 | __m128 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00; |
98 | __m128 dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01; |
99 | __m128 dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02; |
100 | __m128 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10; |
101 | __m128 dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11; |
102 | __m128 dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12; |
103 | __m128 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20; |
104 | __m128 dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21; |
105 | __m128 dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22; |
106 | __m128 velec,felec,velecsum,facel,crf,krf,krf2; |
107 | real *charge; |
108 | int nvdwtype; |
109 | __m128 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6; |
110 | int *vdwtype; |
111 | real *vdwparam; |
112 | __m128 one_sixth = _mm_set1_ps(1.0/6.0); |
113 | __m128 one_twelfth = _mm_set1_ps(1.0/12.0); |
114 | __m128i ewitab; |
115 | __m128 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV; |
116 | real *ewtab; |
117 | __m128 rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw; |
118 | real rswitch_scalar,d_scalar; |
119 | __m128 dummy_mask,cutoff_mask; |
120 | __m128 signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) ); |
121 | __m128 one = _mm_set1_ps(1.0); |
122 | __m128 two = _mm_set1_ps(2.0); |
123 | x = xx[0]; |
124 | f = ff[0]; |
125 | |
126 | nri = nlist->nri; |
127 | iinr = nlist->iinr; |
128 | jindex = nlist->jindex; |
129 | jjnr = nlist->jjnr; |
130 | shiftidx = nlist->shift; |
131 | gid = nlist->gid; |
132 | shiftvec = fr->shift_vec[0]; |
133 | fshift = fr->fshift[0]; |
134 | facel = _mm_set1_ps(fr->epsfac); |
135 | charge = mdatoms->chargeA; |
136 | nvdwtype = fr->ntype; |
137 | vdwparam = fr->nbfp; |
138 | vdwtype = mdatoms->typeA; |
139 | |
140 | sh_ewald = _mm_set1_ps(fr->ic->sh_ewald); |
141 | ewtab = fr->ic->tabq_coul_FDV0; |
142 | ewtabscale = _mm_set1_ps(fr->ic->tabq_scale); |
143 | ewtabhalfspace = _mm_set1_ps(0.5/fr->ic->tabq_scale); |
144 | |
145 | /* Setup water-specific parameters */ |
146 | inr = nlist->iinr[0]; |
147 | iq0 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+0])); |
148 | iq1 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1])); |
149 | iq2 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2])); |
150 | vdwioffset0 = 2*nvdwtype*vdwtype[inr+0]; |
151 | |
152 | jq0 = _mm_set1_ps(charge[inr+0]); |
153 | jq1 = _mm_set1_ps(charge[inr+1]); |
154 | jq2 = _mm_set1_ps(charge[inr+2]); |
155 | vdwjidx0A = 2*vdwtype[inr+0]; |
156 | qq00 = _mm_mul_ps(iq0,jq0); |
157 | c6_00 = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A]); |
158 | c12_00 = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A+1]); |
159 | qq01 = _mm_mul_ps(iq0,jq1); |
160 | qq02 = _mm_mul_ps(iq0,jq2); |
161 | qq10 = _mm_mul_ps(iq1,jq0); |
162 | qq11 = _mm_mul_ps(iq1,jq1); |
163 | qq12 = _mm_mul_ps(iq1,jq2); |
164 | qq20 = _mm_mul_ps(iq2,jq0); |
165 | qq21 = _mm_mul_ps(iq2,jq1); |
166 | qq22 = _mm_mul_ps(iq2,jq2); |
167 | |
168 | /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */ |
169 | rcutoff_scalar = fr->rcoulomb; |
170 | rcutoff = _mm_set1_ps(rcutoff_scalar); |
171 | rcutoff2 = _mm_mul_ps(rcutoff,rcutoff); |
172 | |
173 | rswitch_scalar = fr->rcoulomb_switch; |
174 | rswitch = _mm_set1_ps(rswitch_scalar); |
175 | /* Setup switch parameters */ |
176 | d_scalar = rcutoff_scalar-rswitch_scalar; |
177 | d = _mm_set1_ps(d_scalar); |
178 | swV3 = _mm_set1_ps(-10.0/(d_scalar*d_scalar*d_scalar)); |
179 | swV4 = _mm_set1_ps( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar)); |
180 | swV5 = _mm_set1_ps( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar)); |
181 | swF2 = _mm_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar)); |
182 | swF3 = _mm_set1_ps( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar)); |
183 | swF4 = _mm_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar)); |
184 | |
185 | /* Avoid stupid compiler warnings */ |
186 | jnrA = jnrB = jnrC = jnrD = 0; |
187 | j_coord_offsetA = 0; |
188 | j_coord_offsetB = 0; |
189 | j_coord_offsetC = 0; |
Value stored to 'j_coord_offsetC' is never read | |
190 | j_coord_offsetD = 0; |
191 | |
192 | outeriter = 0; |
193 | inneriter = 0; |
194 | |
195 | for(iidx=0;iidx<4*DIM3;iidx++) |
196 | { |
197 | scratch[iidx] = 0.0; |
198 | } |
199 | |
200 | /* Start outer loop over neighborlists */ |
201 | for(iidx=0; iidx<nri; iidx++) |
202 | { |
203 | /* Load shift vector for this list */ |
204 | i_shift_offset = DIM3*shiftidx[iidx]; |
205 | |
206 | /* Load limits for loop over neighbors */ |
207 | j_index_start = jindex[iidx]; |
208 | j_index_end = jindex[iidx+1]; |
209 | |
210 | /* Get outer coordinate index */ |
211 | inr = iinr[iidx]; |
212 | i_coord_offset = DIM3*inr; |
213 | |
214 | /* Load i particle coords and add shift vector */ |
215 | gmx_mm_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset, |
216 | &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2); |
217 | |
218 | fix0 = _mm_setzero_ps(); |
219 | fiy0 = _mm_setzero_ps(); |
220 | fiz0 = _mm_setzero_ps(); |
221 | fix1 = _mm_setzero_ps(); |
222 | fiy1 = _mm_setzero_ps(); |
223 | fiz1 = _mm_setzero_ps(); |
224 | fix2 = _mm_setzero_ps(); |
225 | fiy2 = _mm_setzero_ps(); |
226 | fiz2 = _mm_setzero_ps(); |
227 | |
228 | /* Reset potential sums */ |
229 | velecsum = _mm_setzero_ps(); |
230 | vvdwsum = _mm_setzero_ps(); |
231 | |
232 | /* Start inner kernel loop */ |
233 | for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4) |
234 | { |
235 | |
236 | /* Get j neighbor index, and coordinate index */ |
237 | jnrA = jjnr[jidx]; |
238 | jnrB = jjnr[jidx+1]; |
239 | jnrC = jjnr[jidx+2]; |
240 | jnrD = jjnr[jidx+3]; |
241 | j_coord_offsetA = DIM3*jnrA; |
242 | j_coord_offsetB = DIM3*jnrB; |
243 | j_coord_offsetC = DIM3*jnrC; |
244 | j_coord_offsetD = DIM3*jnrD; |
245 | |
246 | /* load j atom coordinates */ |
247 | gmx_mm_load_3rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB, |
248 | x+j_coord_offsetC,x+j_coord_offsetD, |
249 | &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2); |
250 | |
251 | /* Calculate displacement vector */ |
252 | dx00 = _mm_sub_ps(ix0,jx0); |
253 | dy00 = _mm_sub_ps(iy0,jy0); |
254 | dz00 = _mm_sub_ps(iz0,jz0); |
255 | dx01 = _mm_sub_ps(ix0,jx1); |
256 | dy01 = _mm_sub_ps(iy0,jy1); |
257 | dz01 = _mm_sub_ps(iz0,jz1); |
258 | dx02 = _mm_sub_ps(ix0,jx2); |
259 | dy02 = _mm_sub_ps(iy0,jy2); |
260 | dz02 = _mm_sub_ps(iz0,jz2); |
261 | dx10 = _mm_sub_ps(ix1,jx0); |
262 | dy10 = _mm_sub_ps(iy1,jy0); |
263 | dz10 = _mm_sub_ps(iz1,jz0); |
264 | dx11 = _mm_sub_ps(ix1,jx1); |
265 | dy11 = _mm_sub_ps(iy1,jy1); |
266 | dz11 = _mm_sub_ps(iz1,jz1); |
267 | dx12 = _mm_sub_ps(ix1,jx2); |
268 | dy12 = _mm_sub_ps(iy1,jy2); |
269 | dz12 = _mm_sub_ps(iz1,jz2); |
270 | dx20 = _mm_sub_ps(ix2,jx0); |
271 | dy20 = _mm_sub_ps(iy2,jy0); |
272 | dz20 = _mm_sub_ps(iz2,jz0); |
273 | dx21 = _mm_sub_ps(ix2,jx1); |
274 | dy21 = _mm_sub_ps(iy2,jy1); |
275 | dz21 = _mm_sub_ps(iz2,jz1); |
276 | dx22 = _mm_sub_ps(ix2,jx2); |
277 | dy22 = _mm_sub_ps(iy2,jy2); |
278 | dz22 = _mm_sub_ps(iz2,jz2); |
279 | |
280 | /* Calculate squared distance and things based on it */ |
281 | rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00); |
282 | rsq01 = gmx_mm_calc_rsq_ps(dx01,dy01,dz01); |
283 | rsq02 = gmx_mm_calc_rsq_ps(dx02,dy02,dz02); |
284 | rsq10 = gmx_mm_calc_rsq_ps(dx10,dy10,dz10); |
285 | rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11); |
286 | rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12); |
287 | rsq20 = gmx_mm_calc_rsq_ps(dx20,dy20,dz20); |
288 | rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21); |
289 | rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22); |
290 | |
291 | rinv00 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq00); |
292 | rinv01 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq01); |
293 | rinv02 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq02); |
294 | rinv10 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq10); |
295 | rinv11 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq11); |
296 | rinv12 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq12); |
297 | rinv20 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq20); |
298 | rinv21 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq21); |
299 | rinv22 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq22); |
300 | |
301 | rinvsq00 = _mm_mul_ps(rinv00,rinv00); |
302 | rinvsq01 = _mm_mul_ps(rinv01,rinv01); |
303 | rinvsq02 = _mm_mul_ps(rinv02,rinv02); |
304 | rinvsq10 = _mm_mul_ps(rinv10,rinv10); |
305 | rinvsq11 = _mm_mul_ps(rinv11,rinv11); |
306 | rinvsq12 = _mm_mul_ps(rinv12,rinv12); |
307 | rinvsq20 = _mm_mul_ps(rinv20,rinv20); |
308 | rinvsq21 = _mm_mul_ps(rinv21,rinv21); |
309 | rinvsq22 = _mm_mul_ps(rinv22,rinv22); |
310 | |
311 | fjx0 = _mm_setzero_ps(); |
312 | fjy0 = _mm_setzero_ps(); |
313 | fjz0 = _mm_setzero_ps(); |
314 | fjx1 = _mm_setzero_ps(); |
315 | fjy1 = _mm_setzero_ps(); |
316 | fjz1 = _mm_setzero_ps(); |
317 | fjx2 = _mm_setzero_ps(); |
318 | fjy2 = _mm_setzero_ps(); |
319 | fjz2 = _mm_setzero_ps(); |
320 | |
321 | /************************** |
322 | * CALCULATE INTERACTIONS * |
323 | **************************/ |
324 | |
325 | if (gmx_mm_any_lt(rsq00,rcutoff2)) |
326 | { |
327 | |
328 | r00 = _mm_mul_ps(rsq00,rinv00); |
329 | |
330 | /* EWALD ELECTROSTATICS */ |
331 | |
332 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
333 | ewrt = _mm_mul_ps(r00,ewtabscale); |
334 | ewitab = _mm_cvttps_epi32(ewrt); |
335 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
336 | ewitab = _mm_slli_epi32(ewitab,2); |
337 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
338 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
339 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
340 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
341 | _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); |
342 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
343 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
344 | velec = _mm_mul_ps(qq00,_mm_sub_ps(rinv00,velec)); |
345 | felec = _mm_mul_ps(_mm_mul_ps(qq00,rinv00),_mm_sub_ps(rinvsq00,felec)); |
346 | |
347 | /* LENNARD-JONES DISPERSION/REPULSION */ |
348 | |
349 | rinvsix = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00); |
350 | vvdw6 = _mm_mul_ps(c6_00,rinvsix); |
351 | vvdw12 = _mm_mul_ps(c12_00,_mm_mul_ps(rinvsix,rinvsix)); |
352 | vvdw = _mm_sub_ps( _mm_mul_ps(vvdw12,one_twelfth) , _mm_mul_ps(vvdw6,one_sixth) ); |
353 | fvdw = _mm_mul_ps(_mm_sub_ps(vvdw12,vvdw6),rinvsq00); |
354 | |
355 | d = _mm_sub_ps(r00,rswitch); |
356 | d = _mm_max_ps(d,_mm_setzero_ps()); |
357 | d2 = _mm_mul_ps(d,d); |
358 | 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))))))); |
359 | |
360 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
361 | |
362 | /* Evaluate switch function */ |
363 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
364 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv00,_mm_mul_ps(velec,dsw)) ); |
365 | fvdw = _mm_sub_ps( _mm_mul_ps(fvdw,sw) , _mm_mul_ps(rinv00,_mm_mul_ps(vvdw,dsw)) ); |
366 | velec = _mm_mul_ps(velec,sw); |
367 | vvdw = _mm_mul_ps(vvdw,sw); |
368 | cutoff_mask = _mm_cmplt_ps(rsq00,rcutoff2); |
369 | |
370 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
371 | velec = _mm_and_ps(velec,cutoff_mask); |
372 | velecsum = _mm_add_ps(velecsum,velec); |
373 | vvdw = _mm_and_ps(vvdw,cutoff_mask); |
374 | vvdwsum = _mm_add_ps(vvdwsum,vvdw); |
375 | |
376 | fscal = _mm_add_ps(felec,fvdw); |
377 | |
378 | fscal = _mm_and_ps(fscal,cutoff_mask); |
379 | |
380 | /* Calculate temporary vectorial force */ |
381 | tx = _mm_mul_ps(fscal,dx00); |
382 | ty = _mm_mul_ps(fscal,dy00); |
383 | tz = _mm_mul_ps(fscal,dz00); |
384 | |
385 | /* Update vectorial force */ |
386 | fix0 = _mm_add_ps(fix0,tx); |
387 | fiy0 = _mm_add_ps(fiy0,ty); |
388 | fiz0 = _mm_add_ps(fiz0,tz); |
389 | |
390 | fjx0 = _mm_add_ps(fjx0,tx); |
391 | fjy0 = _mm_add_ps(fjy0,ty); |
392 | fjz0 = _mm_add_ps(fjz0,tz); |
393 | |
394 | } |
395 | |
396 | /************************** |
397 | * CALCULATE INTERACTIONS * |
398 | **************************/ |
399 | |
400 | if (gmx_mm_any_lt(rsq01,rcutoff2)) |
401 | { |
402 | |
403 | r01 = _mm_mul_ps(rsq01,rinv01); |
404 | |
405 | /* EWALD ELECTROSTATICS */ |
406 | |
407 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
408 | ewrt = _mm_mul_ps(r01,ewtabscale); |
409 | ewitab = _mm_cvttps_epi32(ewrt); |
410 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
411 | ewitab = _mm_slli_epi32(ewitab,2); |
412 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
413 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
414 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
415 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
416 | _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); |
417 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
418 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
419 | velec = _mm_mul_ps(qq01,_mm_sub_ps(rinv01,velec)); |
420 | felec = _mm_mul_ps(_mm_mul_ps(qq01,rinv01),_mm_sub_ps(rinvsq01,felec)); |
421 | |
422 | d = _mm_sub_ps(r01,rswitch); |
423 | d = _mm_max_ps(d,_mm_setzero_ps()); |
424 | d2 = _mm_mul_ps(d,d); |
425 | 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))))))); |
426 | |
427 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
428 | |
429 | /* Evaluate switch function */ |
430 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
431 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv01,_mm_mul_ps(velec,dsw)) ); |
432 | velec = _mm_mul_ps(velec,sw); |
433 | cutoff_mask = _mm_cmplt_ps(rsq01,rcutoff2); |
434 | |
435 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
436 | velec = _mm_and_ps(velec,cutoff_mask); |
437 | velecsum = _mm_add_ps(velecsum,velec); |
438 | |
439 | fscal = felec; |
440 | |
441 | fscal = _mm_and_ps(fscal,cutoff_mask); |
442 | |
443 | /* Calculate temporary vectorial force */ |
444 | tx = _mm_mul_ps(fscal,dx01); |
445 | ty = _mm_mul_ps(fscal,dy01); |
446 | tz = _mm_mul_ps(fscal,dz01); |
447 | |
448 | /* Update vectorial force */ |
449 | fix0 = _mm_add_ps(fix0,tx); |
450 | fiy0 = _mm_add_ps(fiy0,ty); |
451 | fiz0 = _mm_add_ps(fiz0,tz); |
452 | |
453 | fjx1 = _mm_add_ps(fjx1,tx); |
454 | fjy1 = _mm_add_ps(fjy1,ty); |
455 | fjz1 = _mm_add_ps(fjz1,tz); |
456 | |
457 | } |
458 | |
459 | /************************** |
460 | * CALCULATE INTERACTIONS * |
461 | **************************/ |
462 | |
463 | if (gmx_mm_any_lt(rsq02,rcutoff2)) |
464 | { |
465 | |
466 | r02 = _mm_mul_ps(rsq02,rinv02); |
467 | |
468 | /* EWALD ELECTROSTATICS */ |
469 | |
470 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
471 | ewrt = _mm_mul_ps(r02,ewtabscale); |
472 | ewitab = _mm_cvttps_epi32(ewrt); |
473 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
474 | ewitab = _mm_slli_epi32(ewitab,2); |
475 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
476 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
477 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
478 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
479 | _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); |
480 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
481 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
482 | velec = _mm_mul_ps(qq02,_mm_sub_ps(rinv02,velec)); |
483 | felec = _mm_mul_ps(_mm_mul_ps(qq02,rinv02),_mm_sub_ps(rinvsq02,felec)); |
484 | |
485 | d = _mm_sub_ps(r02,rswitch); |
486 | d = _mm_max_ps(d,_mm_setzero_ps()); |
487 | d2 = _mm_mul_ps(d,d); |
488 | 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))))))); |
489 | |
490 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
491 | |
492 | /* Evaluate switch function */ |
493 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
494 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv02,_mm_mul_ps(velec,dsw)) ); |
495 | velec = _mm_mul_ps(velec,sw); |
496 | cutoff_mask = _mm_cmplt_ps(rsq02,rcutoff2); |
497 | |
498 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
499 | velec = _mm_and_ps(velec,cutoff_mask); |
500 | velecsum = _mm_add_ps(velecsum,velec); |
501 | |
502 | fscal = felec; |
503 | |
504 | fscal = _mm_and_ps(fscal,cutoff_mask); |
505 | |
506 | /* Calculate temporary vectorial force */ |
507 | tx = _mm_mul_ps(fscal,dx02); |
508 | ty = _mm_mul_ps(fscal,dy02); |
509 | tz = _mm_mul_ps(fscal,dz02); |
510 | |
511 | /* Update vectorial force */ |
512 | fix0 = _mm_add_ps(fix0,tx); |
513 | fiy0 = _mm_add_ps(fiy0,ty); |
514 | fiz0 = _mm_add_ps(fiz0,tz); |
515 | |
516 | fjx2 = _mm_add_ps(fjx2,tx); |
517 | fjy2 = _mm_add_ps(fjy2,ty); |
518 | fjz2 = _mm_add_ps(fjz2,tz); |
519 | |
520 | } |
521 | |
522 | /************************** |
523 | * CALCULATE INTERACTIONS * |
524 | **************************/ |
525 | |
526 | if (gmx_mm_any_lt(rsq10,rcutoff2)) |
527 | { |
528 | |
529 | r10 = _mm_mul_ps(rsq10,rinv10); |
530 | |
531 | /* EWALD ELECTROSTATICS */ |
532 | |
533 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
534 | ewrt = _mm_mul_ps(r10,ewtabscale); |
535 | ewitab = _mm_cvttps_epi32(ewrt); |
536 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
537 | ewitab = _mm_slli_epi32(ewitab,2); |
538 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
539 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
540 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
541 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
542 | _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); |
543 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
544 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
545 | velec = _mm_mul_ps(qq10,_mm_sub_ps(rinv10,velec)); |
546 | felec = _mm_mul_ps(_mm_mul_ps(qq10,rinv10),_mm_sub_ps(rinvsq10,felec)); |
547 | |
548 | d = _mm_sub_ps(r10,rswitch); |
549 | d = _mm_max_ps(d,_mm_setzero_ps()); |
550 | d2 = _mm_mul_ps(d,d); |
551 | 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))))))); |
552 | |
553 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
554 | |
555 | /* Evaluate switch function */ |
556 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
557 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv10,_mm_mul_ps(velec,dsw)) ); |
558 | velec = _mm_mul_ps(velec,sw); |
559 | cutoff_mask = _mm_cmplt_ps(rsq10,rcutoff2); |
560 | |
561 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
562 | velec = _mm_and_ps(velec,cutoff_mask); |
563 | velecsum = _mm_add_ps(velecsum,velec); |
564 | |
565 | fscal = felec; |
566 | |
567 | fscal = _mm_and_ps(fscal,cutoff_mask); |
568 | |
569 | /* Calculate temporary vectorial force */ |
570 | tx = _mm_mul_ps(fscal,dx10); |
571 | ty = _mm_mul_ps(fscal,dy10); |
572 | tz = _mm_mul_ps(fscal,dz10); |
573 | |
574 | /* Update vectorial force */ |
575 | fix1 = _mm_add_ps(fix1,tx); |
576 | fiy1 = _mm_add_ps(fiy1,ty); |
577 | fiz1 = _mm_add_ps(fiz1,tz); |
578 | |
579 | fjx0 = _mm_add_ps(fjx0,tx); |
580 | fjy0 = _mm_add_ps(fjy0,ty); |
581 | fjz0 = _mm_add_ps(fjz0,tz); |
582 | |
583 | } |
584 | |
585 | /************************** |
586 | * CALCULATE INTERACTIONS * |
587 | **************************/ |
588 | |
589 | if (gmx_mm_any_lt(rsq11,rcutoff2)) |
590 | { |
591 | |
592 | r11 = _mm_mul_ps(rsq11,rinv11); |
593 | |
594 | /* EWALD ELECTROSTATICS */ |
595 | |
596 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
597 | ewrt = _mm_mul_ps(r11,ewtabscale); |
598 | ewitab = _mm_cvttps_epi32(ewrt); |
599 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
600 | ewitab = _mm_slli_epi32(ewitab,2); |
601 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
602 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
603 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
604 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
605 | _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); |
606 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
607 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
608 | velec = _mm_mul_ps(qq11,_mm_sub_ps(rinv11,velec)); |
609 | felec = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec)); |
610 | |
611 | d = _mm_sub_ps(r11,rswitch); |
612 | d = _mm_max_ps(d,_mm_setzero_ps()); |
613 | d2 = _mm_mul_ps(d,d); |
614 | 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))))))); |
615 | |
616 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
617 | |
618 | /* Evaluate switch function */ |
619 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
620 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv11,_mm_mul_ps(velec,dsw)) ); |
621 | velec = _mm_mul_ps(velec,sw); |
622 | cutoff_mask = _mm_cmplt_ps(rsq11,rcutoff2); |
623 | |
624 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
625 | velec = _mm_and_ps(velec,cutoff_mask); |
626 | velecsum = _mm_add_ps(velecsum,velec); |
627 | |
628 | fscal = felec; |
629 | |
630 | fscal = _mm_and_ps(fscal,cutoff_mask); |
631 | |
632 | /* Calculate temporary vectorial force */ |
633 | tx = _mm_mul_ps(fscal,dx11); |
634 | ty = _mm_mul_ps(fscal,dy11); |
635 | tz = _mm_mul_ps(fscal,dz11); |
636 | |
637 | /* Update vectorial force */ |
638 | fix1 = _mm_add_ps(fix1,tx); |
639 | fiy1 = _mm_add_ps(fiy1,ty); |
640 | fiz1 = _mm_add_ps(fiz1,tz); |
641 | |
642 | fjx1 = _mm_add_ps(fjx1,tx); |
643 | fjy1 = _mm_add_ps(fjy1,ty); |
644 | fjz1 = _mm_add_ps(fjz1,tz); |
645 | |
646 | } |
647 | |
648 | /************************** |
649 | * CALCULATE INTERACTIONS * |
650 | **************************/ |
651 | |
652 | if (gmx_mm_any_lt(rsq12,rcutoff2)) |
653 | { |
654 | |
655 | r12 = _mm_mul_ps(rsq12,rinv12); |
656 | |
657 | /* EWALD ELECTROSTATICS */ |
658 | |
659 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
660 | ewrt = _mm_mul_ps(r12,ewtabscale); |
661 | ewitab = _mm_cvttps_epi32(ewrt); |
662 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
663 | ewitab = _mm_slli_epi32(ewitab,2); |
664 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
665 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
666 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
667 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
668 | _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); |
669 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
670 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
671 | velec = _mm_mul_ps(qq12,_mm_sub_ps(rinv12,velec)); |
672 | felec = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec)); |
673 | |
674 | d = _mm_sub_ps(r12,rswitch); |
675 | d = _mm_max_ps(d,_mm_setzero_ps()); |
676 | d2 = _mm_mul_ps(d,d); |
677 | 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))))))); |
678 | |
679 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
680 | |
681 | /* Evaluate switch function */ |
682 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
683 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv12,_mm_mul_ps(velec,dsw)) ); |
684 | velec = _mm_mul_ps(velec,sw); |
685 | cutoff_mask = _mm_cmplt_ps(rsq12,rcutoff2); |
686 | |
687 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
688 | velec = _mm_and_ps(velec,cutoff_mask); |
689 | velecsum = _mm_add_ps(velecsum,velec); |
690 | |
691 | fscal = felec; |
692 | |
693 | fscal = _mm_and_ps(fscal,cutoff_mask); |
694 | |
695 | /* Calculate temporary vectorial force */ |
696 | tx = _mm_mul_ps(fscal,dx12); |
697 | ty = _mm_mul_ps(fscal,dy12); |
698 | tz = _mm_mul_ps(fscal,dz12); |
699 | |
700 | /* Update vectorial force */ |
701 | fix1 = _mm_add_ps(fix1,tx); |
702 | fiy1 = _mm_add_ps(fiy1,ty); |
703 | fiz1 = _mm_add_ps(fiz1,tz); |
704 | |
705 | fjx2 = _mm_add_ps(fjx2,tx); |
706 | fjy2 = _mm_add_ps(fjy2,ty); |
707 | fjz2 = _mm_add_ps(fjz2,tz); |
708 | |
709 | } |
710 | |
711 | /************************** |
712 | * CALCULATE INTERACTIONS * |
713 | **************************/ |
714 | |
715 | if (gmx_mm_any_lt(rsq20,rcutoff2)) |
716 | { |
717 | |
718 | r20 = _mm_mul_ps(rsq20,rinv20); |
719 | |
720 | /* EWALD ELECTROSTATICS */ |
721 | |
722 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
723 | ewrt = _mm_mul_ps(r20,ewtabscale); |
724 | ewitab = _mm_cvttps_epi32(ewrt); |
725 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
726 | ewitab = _mm_slli_epi32(ewitab,2); |
727 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
728 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
729 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
730 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
731 | _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); |
732 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
733 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
734 | velec = _mm_mul_ps(qq20,_mm_sub_ps(rinv20,velec)); |
735 | felec = _mm_mul_ps(_mm_mul_ps(qq20,rinv20),_mm_sub_ps(rinvsq20,felec)); |
736 | |
737 | d = _mm_sub_ps(r20,rswitch); |
738 | d = _mm_max_ps(d,_mm_setzero_ps()); |
739 | d2 = _mm_mul_ps(d,d); |
740 | 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))))))); |
741 | |
742 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
743 | |
744 | /* Evaluate switch function */ |
745 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
746 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv20,_mm_mul_ps(velec,dsw)) ); |
747 | velec = _mm_mul_ps(velec,sw); |
748 | cutoff_mask = _mm_cmplt_ps(rsq20,rcutoff2); |
749 | |
750 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
751 | velec = _mm_and_ps(velec,cutoff_mask); |
752 | velecsum = _mm_add_ps(velecsum,velec); |
753 | |
754 | fscal = felec; |
755 | |
756 | fscal = _mm_and_ps(fscal,cutoff_mask); |
757 | |
758 | /* Calculate temporary vectorial force */ |
759 | tx = _mm_mul_ps(fscal,dx20); |
760 | ty = _mm_mul_ps(fscal,dy20); |
761 | tz = _mm_mul_ps(fscal,dz20); |
762 | |
763 | /* Update vectorial force */ |
764 | fix2 = _mm_add_ps(fix2,tx); |
765 | fiy2 = _mm_add_ps(fiy2,ty); |
766 | fiz2 = _mm_add_ps(fiz2,tz); |
767 | |
768 | fjx0 = _mm_add_ps(fjx0,tx); |
769 | fjy0 = _mm_add_ps(fjy0,ty); |
770 | fjz0 = _mm_add_ps(fjz0,tz); |
771 | |
772 | } |
773 | |
774 | /************************** |
775 | * CALCULATE INTERACTIONS * |
776 | **************************/ |
777 | |
778 | if (gmx_mm_any_lt(rsq21,rcutoff2)) |
779 | { |
780 | |
781 | r21 = _mm_mul_ps(rsq21,rinv21); |
782 | |
783 | /* EWALD ELECTROSTATICS */ |
784 | |
785 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
786 | ewrt = _mm_mul_ps(r21,ewtabscale); |
787 | ewitab = _mm_cvttps_epi32(ewrt); |
788 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
789 | ewitab = _mm_slli_epi32(ewitab,2); |
790 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
791 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
792 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
793 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
794 | _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); |
795 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
796 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
797 | velec = _mm_mul_ps(qq21,_mm_sub_ps(rinv21,velec)); |
798 | felec = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec)); |
799 | |
800 | d = _mm_sub_ps(r21,rswitch); |
801 | d = _mm_max_ps(d,_mm_setzero_ps()); |
802 | d2 = _mm_mul_ps(d,d); |
803 | 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))))))); |
804 | |
805 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
806 | |
807 | /* Evaluate switch function */ |
808 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
809 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv21,_mm_mul_ps(velec,dsw)) ); |
810 | velec = _mm_mul_ps(velec,sw); |
811 | cutoff_mask = _mm_cmplt_ps(rsq21,rcutoff2); |
812 | |
813 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
814 | velec = _mm_and_ps(velec,cutoff_mask); |
815 | velecsum = _mm_add_ps(velecsum,velec); |
816 | |
817 | fscal = felec; |
818 | |
819 | fscal = _mm_and_ps(fscal,cutoff_mask); |
820 | |
821 | /* Calculate temporary vectorial force */ |
822 | tx = _mm_mul_ps(fscal,dx21); |
823 | ty = _mm_mul_ps(fscal,dy21); |
824 | tz = _mm_mul_ps(fscal,dz21); |
825 | |
826 | /* Update vectorial force */ |
827 | fix2 = _mm_add_ps(fix2,tx); |
828 | fiy2 = _mm_add_ps(fiy2,ty); |
829 | fiz2 = _mm_add_ps(fiz2,tz); |
830 | |
831 | fjx1 = _mm_add_ps(fjx1,tx); |
832 | fjy1 = _mm_add_ps(fjy1,ty); |
833 | fjz1 = _mm_add_ps(fjz1,tz); |
834 | |
835 | } |
836 | |
837 | /************************** |
838 | * CALCULATE INTERACTIONS * |
839 | **************************/ |
840 | |
841 | if (gmx_mm_any_lt(rsq22,rcutoff2)) |
842 | { |
843 | |
844 | r22 = _mm_mul_ps(rsq22,rinv22); |
845 | |
846 | /* EWALD ELECTROSTATICS */ |
847 | |
848 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
849 | ewrt = _mm_mul_ps(r22,ewtabscale); |
850 | ewitab = _mm_cvttps_epi32(ewrt); |
851 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
852 | ewitab = _mm_slli_epi32(ewitab,2); |
853 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
854 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
855 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
856 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
857 | _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); |
858 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
859 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
860 | velec = _mm_mul_ps(qq22,_mm_sub_ps(rinv22,velec)); |
861 | felec = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec)); |
862 | |
863 | d = _mm_sub_ps(r22,rswitch); |
864 | d = _mm_max_ps(d,_mm_setzero_ps()); |
865 | d2 = _mm_mul_ps(d,d); |
866 | 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))))))); |
867 | |
868 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
869 | |
870 | /* Evaluate switch function */ |
871 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
872 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv22,_mm_mul_ps(velec,dsw)) ); |
873 | velec = _mm_mul_ps(velec,sw); |
874 | cutoff_mask = _mm_cmplt_ps(rsq22,rcutoff2); |
875 | |
876 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
877 | velec = _mm_and_ps(velec,cutoff_mask); |
878 | velecsum = _mm_add_ps(velecsum,velec); |
879 | |
880 | fscal = felec; |
881 | |
882 | fscal = _mm_and_ps(fscal,cutoff_mask); |
883 | |
884 | /* Calculate temporary vectorial force */ |
885 | tx = _mm_mul_ps(fscal,dx22); |
886 | ty = _mm_mul_ps(fscal,dy22); |
887 | tz = _mm_mul_ps(fscal,dz22); |
888 | |
889 | /* Update vectorial force */ |
890 | fix2 = _mm_add_ps(fix2,tx); |
891 | fiy2 = _mm_add_ps(fiy2,ty); |
892 | fiz2 = _mm_add_ps(fiz2,tz); |
893 | |
894 | fjx2 = _mm_add_ps(fjx2,tx); |
895 | fjy2 = _mm_add_ps(fjy2,ty); |
896 | fjz2 = _mm_add_ps(fjz2,tz); |
897 | |
898 | } |
899 | |
900 | fjptrA = f+j_coord_offsetA; |
901 | fjptrB = f+j_coord_offsetB; |
902 | fjptrC = f+j_coord_offsetC; |
903 | fjptrD = f+j_coord_offsetD; |
904 | |
905 | gmx_mm_decrement_3rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD, |
906 | fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2); |
907 | |
908 | /* Inner loop uses 603 flops */ |
909 | } |
910 | |
911 | if(jidx<j_index_end) |
912 | { |
913 | |
914 | /* Get j neighbor index, and coordinate index */ |
915 | jnrlistA = jjnr[jidx]; |
916 | jnrlistB = jjnr[jidx+1]; |
917 | jnrlistC = jjnr[jidx+2]; |
918 | jnrlistD = jjnr[jidx+3]; |
919 | /* Sign of each element will be negative for non-real atoms. |
920 | * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones, |
921 | * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries. |
922 | */ |
923 | dummy_mask = gmx_mm_castsi128_ps_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())); |
924 | jnrA = (jnrlistA>=0) ? jnrlistA : 0; |
925 | jnrB = (jnrlistB>=0) ? jnrlistB : 0; |
926 | jnrC = (jnrlistC>=0) ? jnrlistC : 0; |
927 | jnrD = (jnrlistD>=0) ? jnrlistD : 0; |
928 | j_coord_offsetA = DIM3*jnrA; |
929 | j_coord_offsetB = DIM3*jnrB; |
930 | j_coord_offsetC = DIM3*jnrC; |
931 | j_coord_offsetD = DIM3*jnrD; |
932 | |
933 | /* load j atom coordinates */ |
934 | gmx_mm_load_3rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB, |
935 | x+j_coord_offsetC,x+j_coord_offsetD, |
936 | &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2); |
937 | |
938 | /* Calculate displacement vector */ |
939 | dx00 = _mm_sub_ps(ix0,jx0); |
940 | dy00 = _mm_sub_ps(iy0,jy0); |
941 | dz00 = _mm_sub_ps(iz0,jz0); |
942 | dx01 = _mm_sub_ps(ix0,jx1); |
943 | dy01 = _mm_sub_ps(iy0,jy1); |
944 | dz01 = _mm_sub_ps(iz0,jz1); |
945 | dx02 = _mm_sub_ps(ix0,jx2); |
946 | dy02 = _mm_sub_ps(iy0,jy2); |
947 | dz02 = _mm_sub_ps(iz0,jz2); |
948 | dx10 = _mm_sub_ps(ix1,jx0); |
949 | dy10 = _mm_sub_ps(iy1,jy0); |
950 | dz10 = _mm_sub_ps(iz1,jz0); |
951 | dx11 = _mm_sub_ps(ix1,jx1); |
952 | dy11 = _mm_sub_ps(iy1,jy1); |
953 | dz11 = _mm_sub_ps(iz1,jz1); |
954 | dx12 = _mm_sub_ps(ix1,jx2); |
955 | dy12 = _mm_sub_ps(iy1,jy2); |
956 | dz12 = _mm_sub_ps(iz1,jz2); |
957 | dx20 = _mm_sub_ps(ix2,jx0); |
958 | dy20 = _mm_sub_ps(iy2,jy0); |
959 | dz20 = _mm_sub_ps(iz2,jz0); |
960 | dx21 = _mm_sub_ps(ix2,jx1); |
961 | dy21 = _mm_sub_ps(iy2,jy1); |
962 | dz21 = _mm_sub_ps(iz2,jz1); |
963 | dx22 = _mm_sub_ps(ix2,jx2); |
964 | dy22 = _mm_sub_ps(iy2,jy2); |
965 | dz22 = _mm_sub_ps(iz2,jz2); |
966 | |
967 | /* Calculate squared distance and things based on it */ |
968 | rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00); |
969 | rsq01 = gmx_mm_calc_rsq_ps(dx01,dy01,dz01); |
970 | rsq02 = gmx_mm_calc_rsq_ps(dx02,dy02,dz02); |
971 | rsq10 = gmx_mm_calc_rsq_ps(dx10,dy10,dz10); |
972 | rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11); |
973 | rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12); |
974 | rsq20 = gmx_mm_calc_rsq_ps(dx20,dy20,dz20); |
975 | rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21); |
976 | rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22); |
977 | |
978 | rinv00 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq00); |
979 | rinv01 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq01); |
980 | rinv02 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq02); |
981 | rinv10 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq10); |
982 | rinv11 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq11); |
983 | rinv12 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq12); |
984 | rinv20 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq20); |
985 | rinv21 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq21); |
986 | rinv22 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq22); |
987 | |
988 | rinvsq00 = _mm_mul_ps(rinv00,rinv00); |
989 | rinvsq01 = _mm_mul_ps(rinv01,rinv01); |
990 | rinvsq02 = _mm_mul_ps(rinv02,rinv02); |
991 | rinvsq10 = _mm_mul_ps(rinv10,rinv10); |
992 | rinvsq11 = _mm_mul_ps(rinv11,rinv11); |
993 | rinvsq12 = _mm_mul_ps(rinv12,rinv12); |
994 | rinvsq20 = _mm_mul_ps(rinv20,rinv20); |
995 | rinvsq21 = _mm_mul_ps(rinv21,rinv21); |
996 | rinvsq22 = _mm_mul_ps(rinv22,rinv22); |
997 | |
998 | fjx0 = _mm_setzero_ps(); |
999 | fjy0 = _mm_setzero_ps(); |
1000 | fjz0 = _mm_setzero_ps(); |
1001 | fjx1 = _mm_setzero_ps(); |
1002 | fjy1 = _mm_setzero_ps(); |
1003 | fjz1 = _mm_setzero_ps(); |
1004 | fjx2 = _mm_setzero_ps(); |
1005 | fjy2 = _mm_setzero_ps(); |
1006 | fjz2 = _mm_setzero_ps(); |
1007 | |
1008 | /************************** |
1009 | * CALCULATE INTERACTIONS * |
1010 | **************************/ |
1011 | |
1012 | if (gmx_mm_any_lt(rsq00,rcutoff2)) |
1013 | { |
1014 | |
1015 | r00 = _mm_mul_ps(rsq00,rinv00); |
1016 | r00 = _mm_andnot_ps(dummy_mask,r00); |
1017 | |
1018 | /* EWALD ELECTROSTATICS */ |
1019 | |
1020 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1021 | ewrt = _mm_mul_ps(r00,ewtabscale); |
1022 | ewitab = _mm_cvttps_epi32(ewrt); |
1023 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1024 | ewitab = _mm_slli_epi32(ewitab,2); |
1025 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
1026 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
1027 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
1028 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
1029 | _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); |
1030 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
1031 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
1032 | velec = _mm_mul_ps(qq00,_mm_sub_ps(rinv00,velec)); |
1033 | felec = _mm_mul_ps(_mm_mul_ps(qq00,rinv00),_mm_sub_ps(rinvsq00,felec)); |
1034 | |
1035 | /* LENNARD-JONES DISPERSION/REPULSION */ |
1036 | |
1037 | rinvsix = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00); |
1038 | vvdw6 = _mm_mul_ps(c6_00,rinvsix); |
1039 | vvdw12 = _mm_mul_ps(c12_00,_mm_mul_ps(rinvsix,rinvsix)); |
1040 | vvdw = _mm_sub_ps( _mm_mul_ps(vvdw12,one_twelfth) , _mm_mul_ps(vvdw6,one_sixth) ); |
1041 | fvdw = _mm_mul_ps(_mm_sub_ps(vvdw12,vvdw6),rinvsq00); |
1042 | |
1043 | d = _mm_sub_ps(r00,rswitch); |
1044 | d = _mm_max_ps(d,_mm_setzero_ps()); |
1045 | d2 = _mm_mul_ps(d,d); |
1046 | 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))))))); |
1047 | |
1048 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
1049 | |
1050 | /* Evaluate switch function */ |
1051 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
1052 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv00,_mm_mul_ps(velec,dsw)) ); |
1053 | fvdw = _mm_sub_ps( _mm_mul_ps(fvdw,sw) , _mm_mul_ps(rinv00,_mm_mul_ps(vvdw,dsw)) ); |
1054 | velec = _mm_mul_ps(velec,sw); |
1055 | vvdw = _mm_mul_ps(vvdw,sw); |
1056 | cutoff_mask = _mm_cmplt_ps(rsq00,rcutoff2); |
1057 | |
1058 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
1059 | velec = _mm_and_ps(velec,cutoff_mask); |
1060 | velec = _mm_andnot_ps(dummy_mask,velec); |
1061 | velecsum = _mm_add_ps(velecsum,velec); |
1062 | vvdw = _mm_and_ps(vvdw,cutoff_mask); |
1063 | vvdw = _mm_andnot_ps(dummy_mask,vvdw); |
1064 | vvdwsum = _mm_add_ps(vvdwsum,vvdw); |
1065 | |
1066 | fscal = _mm_add_ps(felec,fvdw); |
1067 | |
1068 | fscal = _mm_and_ps(fscal,cutoff_mask); |
1069 | |
1070 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
1071 | |
1072 | /* Calculate temporary vectorial force */ |
1073 | tx = _mm_mul_ps(fscal,dx00); |
1074 | ty = _mm_mul_ps(fscal,dy00); |
1075 | tz = _mm_mul_ps(fscal,dz00); |
1076 | |
1077 | /* Update vectorial force */ |
1078 | fix0 = _mm_add_ps(fix0,tx); |
1079 | fiy0 = _mm_add_ps(fiy0,ty); |
1080 | fiz0 = _mm_add_ps(fiz0,tz); |
1081 | |
1082 | fjx0 = _mm_add_ps(fjx0,tx); |
1083 | fjy0 = _mm_add_ps(fjy0,ty); |
1084 | fjz0 = _mm_add_ps(fjz0,tz); |
1085 | |
1086 | } |
1087 | |
1088 | /************************** |
1089 | * CALCULATE INTERACTIONS * |
1090 | **************************/ |
1091 | |
1092 | if (gmx_mm_any_lt(rsq01,rcutoff2)) |
1093 | { |
1094 | |
1095 | r01 = _mm_mul_ps(rsq01,rinv01); |
1096 | r01 = _mm_andnot_ps(dummy_mask,r01); |
1097 | |
1098 | /* EWALD ELECTROSTATICS */ |
1099 | |
1100 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1101 | ewrt = _mm_mul_ps(r01,ewtabscale); |
1102 | ewitab = _mm_cvttps_epi32(ewrt); |
1103 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1104 | ewitab = _mm_slli_epi32(ewitab,2); |
1105 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
1106 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
1107 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
1108 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
1109 | _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); |
1110 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
1111 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
1112 | velec = _mm_mul_ps(qq01,_mm_sub_ps(rinv01,velec)); |
1113 | felec = _mm_mul_ps(_mm_mul_ps(qq01,rinv01),_mm_sub_ps(rinvsq01,felec)); |
1114 | |
1115 | d = _mm_sub_ps(r01,rswitch); |
1116 | d = _mm_max_ps(d,_mm_setzero_ps()); |
1117 | d2 = _mm_mul_ps(d,d); |
1118 | 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))))))); |
1119 | |
1120 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
1121 | |
1122 | /* Evaluate switch function */ |
1123 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
1124 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv01,_mm_mul_ps(velec,dsw)) ); |
1125 | velec = _mm_mul_ps(velec,sw); |
1126 | cutoff_mask = _mm_cmplt_ps(rsq01,rcutoff2); |
1127 | |
1128 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
1129 | velec = _mm_and_ps(velec,cutoff_mask); |
1130 | velec = _mm_andnot_ps(dummy_mask,velec); |
1131 | velecsum = _mm_add_ps(velecsum,velec); |
1132 | |
1133 | fscal = felec; |
1134 | |
1135 | fscal = _mm_and_ps(fscal,cutoff_mask); |
1136 | |
1137 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
1138 | |
1139 | /* Calculate temporary vectorial force */ |
1140 | tx = _mm_mul_ps(fscal,dx01); |
1141 | ty = _mm_mul_ps(fscal,dy01); |
1142 | tz = _mm_mul_ps(fscal,dz01); |
1143 | |
1144 | /* Update vectorial force */ |
1145 | fix0 = _mm_add_ps(fix0,tx); |
1146 | fiy0 = _mm_add_ps(fiy0,ty); |
1147 | fiz0 = _mm_add_ps(fiz0,tz); |
1148 | |
1149 | fjx1 = _mm_add_ps(fjx1,tx); |
1150 | fjy1 = _mm_add_ps(fjy1,ty); |
1151 | fjz1 = _mm_add_ps(fjz1,tz); |
1152 | |
1153 | } |
1154 | |
1155 | /************************** |
1156 | * CALCULATE INTERACTIONS * |
1157 | **************************/ |
1158 | |
1159 | if (gmx_mm_any_lt(rsq02,rcutoff2)) |
1160 | { |
1161 | |
1162 | r02 = _mm_mul_ps(rsq02,rinv02); |
1163 | r02 = _mm_andnot_ps(dummy_mask,r02); |
1164 | |
1165 | /* EWALD ELECTROSTATICS */ |
1166 | |
1167 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1168 | ewrt = _mm_mul_ps(r02,ewtabscale); |
1169 | ewitab = _mm_cvttps_epi32(ewrt); |
1170 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1171 | ewitab = _mm_slli_epi32(ewitab,2); |
1172 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
1173 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
1174 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
1175 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
1176 | _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); |
1177 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
1178 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
1179 | velec = _mm_mul_ps(qq02,_mm_sub_ps(rinv02,velec)); |
1180 | felec = _mm_mul_ps(_mm_mul_ps(qq02,rinv02),_mm_sub_ps(rinvsq02,felec)); |
1181 | |
1182 | d = _mm_sub_ps(r02,rswitch); |
1183 | d = _mm_max_ps(d,_mm_setzero_ps()); |
1184 | d2 = _mm_mul_ps(d,d); |
1185 | 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))))))); |
1186 | |
1187 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
1188 | |
1189 | /* Evaluate switch function */ |
1190 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
1191 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv02,_mm_mul_ps(velec,dsw)) ); |
1192 | velec = _mm_mul_ps(velec,sw); |
1193 | cutoff_mask = _mm_cmplt_ps(rsq02,rcutoff2); |
1194 | |
1195 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
1196 | velec = _mm_and_ps(velec,cutoff_mask); |
1197 | velec = _mm_andnot_ps(dummy_mask,velec); |
1198 | velecsum = _mm_add_ps(velecsum,velec); |
1199 | |
1200 | fscal = felec; |
1201 | |
1202 | fscal = _mm_and_ps(fscal,cutoff_mask); |
1203 | |
1204 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
1205 | |
1206 | /* Calculate temporary vectorial force */ |
1207 | tx = _mm_mul_ps(fscal,dx02); |
1208 | ty = _mm_mul_ps(fscal,dy02); |
1209 | tz = _mm_mul_ps(fscal,dz02); |
1210 | |
1211 | /* Update vectorial force */ |
1212 | fix0 = _mm_add_ps(fix0,tx); |
1213 | fiy0 = _mm_add_ps(fiy0,ty); |
1214 | fiz0 = _mm_add_ps(fiz0,tz); |
1215 | |
1216 | fjx2 = _mm_add_ps(fjx2,tx); |
1217 | fjy2 = _mm_add_ps(fjy2,ty); |
1218 | fjz2 = _mm_add_ps(fjz2,tz); |
1219 | |
1220 | } |
1221 | |
1222 | /************************** |
1223 | * CALCULATE INTERACTIONS * |
1224 | **************************/ |
1225 | |
1226 | if (gmx_mm_any_lt(rsq10,rcutoff2)) |
1227 | { |
1228 | |
1229 | r10 = _mm_mul_ps(rsq10,rinv10); |
1230 | r10 = _mm_andnot_ps(dummy_mask,r10); |
1231 | |
1232 | /* EWALD ELECTROSTATICS */ |
1233 | |
1234 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1235 | ewrt = _mm_mul_ps(r10,ewtabscale); |
1236 | ewitab = _mm_cvttps_epi32(ewrt); |
1237 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1238 | ewitab = _mm_slli_epi32(ewitab,2); |
1239 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
1240 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
1241 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
1242 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
1243 | _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); |
1244 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
1245 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
1246 | velec = _mm_mul_ps(qq10,_mm_sub_ps(rinv10,velec)); |
1247 | felec = _mm_mul_ps(_mm_mul_ps(qq10,rinv10),_mm_sub_ps(rinvsq10,felec)); |
1248 | |
1249 | d = _mm_sub_ps(r10,rswitch); |
1250 | d = _mm_max_ps(d,_mm_setzero_ps()); |
1251 | d2 = _mm_mul_ps(d,d); |
1252 | 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))))))); |
1253 | |
1254 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
1255 | |
1256 | /* Evaluate switch function */ |
1257 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
1258 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv10,_mm_mul_ps(velec,dsw)) ); |
1259 | velec = _mm_mul_ps(velec,sw); |
1260 | cutoff_mask = _mm_cmplt_ps(rsq10,rcutoff2); |
1261 | |
1262 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
1263 | velec = _mm_and_ps(velec,cutoff_mask); |
1264 | velec = _mm_andnot_ps(dummy_mask,velec); |
1265 | velecsum = _mm_add_ps(velecsum,velec); |
1266 | |
1267 | fscal = felec; |
1268 | |
1269 | fscal = _mm_and_ps(fscal,cutoff_mask); |
1270 | |
1271 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
1272 | |
1273 | /* Calculate temporary vectorial force */ |
1274 | tx = _mm_mul_ps(fscal,dx10); |
1275 | ty = _mm_mul_ps(fscal,dy10); |
1276 | tz = _mm_mul_ps(fscal,dz10); |
1277 | |
1278 | /* Update vectorial force */ |
1279 | fix1 = _mm_add_ps(fix1,tx); |
1280 | fiy1 = _mm_add_ps(fiy1,ty); |
1281 | fiz1 = _mm_add_ps(fiz1,tz); |
1282 | |
1283 | fjx0 = _mm_add_ps(fjx0,tx); |
1284 | fjy0 = _mm_add_ps(fjy0,ty); |
1285 | fjz0 = _mm_add_ps(fjz0,tz); |
1286 | |
1287 | } |
1288 | |
1289 | /************************** |
1290 | * CALCULATE INTERACTIONS * |
1291 | **************************/ |
1292 | |
1293 | if (gmx_mm_any_lt(rsq11,rcutoff2)) |
1294 | { |
1295 | |
1296 | r11 = _mm_mul_ps(rsq11,rinv11); |
1297 | r11 = _mm_andnot_ps(dummy_mask,r11); |
1298 | |
1299 | /* EWALD ELECTROSTATICS */ |
1300 | |
1301 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1302 | ewrt = _mm_mul_ps(r11,ewtabscale); |
1303 | ewitab = _mm_cvttps_epi32(ewrt); |
1304 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1305 | ewitab = _mm_slli_epi32(ewitab,2); |
1306 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
1307 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
1308 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
1309 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
1310 | _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); |
1311 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
1312 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
1313 | velec = _mm_mul_ps(qq11,_mm_sub_ps(rinv11,velec)); |
1314 | felec = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec)); |
1315 | |
1316 | d = _mm_sub_ps(r11,rswitch); |
1317 | d = _mm_max_ps(d,_mm_setzero_ps()); |
1318 | d2 = _mm_mul_ps(d,d); |
1319 | 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))))))); |
1320 | |
1321 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
1322 | |
1323 | /* Evaluate switch function */ |
1324 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
1325 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv11,_mm_mul_ps(velec,dsw)) ); |
1326 | velec = _mm_mul_ps(velec,sw); |
1327 | cutoff_mask = _mm_cmplt_ps(rsq11,rcutoff2); |
1328 | |
1329 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
1330 | velec = _mm_and_ps(velec,cutoff_mask); |
1331 | velec = _mm_andnot_ps(dummy_mask,velec); |
1332 | velecsum = _mm_add_ps(velecsum,velec); |
1333 | |
1334 | fscal = felec; |
1335 | |
1336 | fscal = _mm_and_ps(fscal,cutoff_mask); |
1337 | |
1338 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
1339 | |
1340 | /* Calculate temporary vectorial force */ |
1341 | tx = _mm_mul_ps(fscal,dx11); |
1342 | ty = _mm_mul_ps(fscal,dy11); |
1343 | tz = _mm_mul_ps(fscal,dz11); |
1344 | |
1345 | /* Update vectorial force */ |
1346 | fix1 = _mm_add_ps(fix1,tx); |
1347 | fiy1 = _mm_add_ps(fiy1,ty); |
1348 | fiz1 = _mm_add_ps(fiz1,tz); |
1349 | |
1350 | fjx1 = _mm_add_ps(fjx1,tx); |
1351 | fjy1 = _mm_add_ps(fjy1,ty); |
1352 | fjz1 = _mm_add_ps(fjz1,tz); |
1353 | |
1354 | } |
1355 | |
1356 | /************************** |
1357 | * CALCULATE INTERACTIONS * |
1358 | **************************/ |
1359 | |
1360 | if (gmx_mm_any_lt(rsq12,rcutoff2)) |
1361 | { |
1362 | |
1363 | r12 = _mm_mul_ps(rsq12,rinv12); |
1364 | r12 = _mm_andnot_ps(dummy_mask,r12); |
1365 | |
1366 | /* EWALD ELECTROSTATICS */ |
1367 | |
1368 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1369 | ewrt = _mm_mul_ps(r12,ewtabscale); |
1370 | ewitab = _mm_cvttps_epi32(ewrt); |
1371 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1372 | ewitab = _mm_slli_epi32(ewitab,2); |
1373 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
1374 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
1375 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
1376 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
1377 | _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); |
1378 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
1379 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
1380 | velec = _mm_mul_ps(qq12,_mm_sub_ps(rinv12,velec)); |
1381 | felec = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec)); |
1382 | |
1383 | d = _mm_sub_ps(r12,rswitch); |
1384 | d = _mm_max_ps(d,_mm_setzero_ps()); |
1385 | d2 = _mm_mul_ps(d,d); |
1386 | 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))))))); |
1387 | |
1388 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
1389 | |
1390 | /* Evaluate switch function */ |
1391 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
1392 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv12,_mm_mul_ps(velec,dsw)) ); |
1393 | velec = _mm_mul_ps(velec,sw); |
1394 | cutoff_mask = _mm_cmplt_ps(rsq12,rcutoff2); |
1395 | |
1396 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
1397 | velec = _mm_and_ps(velec,cutoff_mask); |
1398 | velec = _mm_andnot_ps(dummy_mask,velec); |
1399 | velecsum = _mm_add_ps(velecsum,velec); |
1400 | |
1401 | fscal = felec; |
1402 | |
1403 | fscal = _mm_and_ps(fscal,cutoff_mask); |
1404 | |
1405 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
1406 | |
1407 | /* Calculate temporary vectorial force */ |
1408 | tx = _mm_mul_ps(fscal,dx12); |
1409 | ty = _mm_mul_ps(fscal,dy12); |
1410 | tz = _mm_mul_ps(fscal,dz12); |
1411 | |
1412 | /* Update vectorial force */ |
1413 | fix1 = _mm_add_ps(fix1,tx); |
1414 | fiy1 = _mm_add_ps(fiy1,ty); |
1415 | fiz1 = _mm_add_ps(fiz1,tz); |
1416 | |
1417 | fjx2 = _mm_add_ps(fjx2,tx); |
1418 | fjy2 = _mm_add_ps(fjy2,ty); |
1419 | fjz2 = _mm_add_ps(fjz2,tz); |
1420 | |
1421 | } |
1422 | |
1423 | /************************** |
1424 | * CALCULATE INTERACTIONS * |
1425 | **************************/ |
1426 | |
1427 | if (gmx_mm_any_lt(rsq20,rcutoff2)) |
1428 | { |
1429 | |
1430 | r20 = _mm_mul_ps(rsq20,rinv20); |
1431 | r20 = _mm_andnot_ps(dummy_mask,r20); |
1432 | |
1433 | /* EWALD ELECTROSTATICS */ |
1434 | |
1435 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1436 | ewrt = _mm_mul_ps(r20,ewtabscale); |
1437 | ewitab = _mm_cvttps_epi32(ewrt); |
1438 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1439 | ewitab = _mm_slli_epi32(ewitab,2); |
1440 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
1441 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
1442 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
1443 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
1444 | _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); |
1445 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
1446 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
1447 | velec = _mm_mul_ps(qq20,_mm_sub_ps(rinv20,velec)); |
1448 | felec = _mm_mul_ps(_mm_mul_ps(qq20,rinv20),_mm_sub_ps(rinvsq20,felec)); |
1449 | |
1450 | d = _mm_sub_ps(r20,rswitch); |
1451 | d = _mm_max_ps(d,_mm_setzero_ps()); |
1452 | d2 = _mm_mul_ps(d,d); |
1453 | 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))))))); |
1454 | |
1455 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
1456 | |
1457 | /* Evaluate switch function */ |
1458 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
1459 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv20,_mm_mul_ps(velec,dsw)) ); |
1460 | velec = _mm_mul_ps(velec,sw); |
1461 | cutoff_mask = _mm_cmplt_ps(rsq20,rcutoff2); |
1462 | |
1463 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
1464 | velec = _mm_and_ps(velec,cutoff_mask); |
1465 | velec = _mm_andnot_ps(dummy_mask,velec); |
1466 | velecsum = _mm_add_ps(velecsum,velec); |
1467 | |
1468 | fscal = felec; |
1469 | |
1470 | fscal = _mm_and_ps(fscal,cutoff_mask); |
1471 | |
1472 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
1473 | |
1474 | /* Calculate temporary vectorial force */ |
1475 | tx = _mm_mul_ps(fscal,dx20); |
1476 | ty = _mm_mul_ps(fscal,dy20); |
1477 | tz = _mm_mul_ps(fscal,dz20); |
1478 | |
1479 | /* Update vectorial force */ |
1480 | fix2 = _mm_add_ps(fix2,tx); |
1481 | fiy2 = _mm_add_ps(fiy2,ty); |
1482 | fiz2 = _mm_add_ps(fiz2,tz); |
1483 | |
1484 | fjx0 = _mm_add_ps(fjx0,tx); |
1485 | fjy0 = _mm_add_ps(fjy0,ty); |
1486 | fjz0 = _mm_add_ps(fjz0,tz); |
1487 | |
1488 | } |
1489 | |
1490 | /************************** |
1491 | * CALCULATE INTERACTIONS * |
1492 | **************************/ |
1493 | |
1494 | if (gmx_mm_any_lt(rsq21,rcutoff2)) |
1495 | { |
1496 | |
1497 | r21 = _mm_mul_ps(rsq21,rinv21); |
1498 | r21 = _mm_andnot_ps(dummy_mask,r21); |
1499 | |
1500 | /* EWALD ELECTROSTATICS */ |
1501 | |
1502 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1503 | ewrt = _mm_mul_ps(r21,ewtabscale); |
1504 | ewitab = _mm_cvttps_epi32(ewrt); |
1505 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1506 | ewitab = _mm_slli_epi32(ewitab,2); |
1507 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
1508 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
1509 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
1510 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
1511 | _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); |
1512 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
1513 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
1514 | velec = _mm_mul_ps(qq21,_mm_sub_ps(rinv21,velec)); |
1515 | felec = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec)); |
1516 | |
1517 | d = _mm_sub_ps(r21,rswitch); |
1518 | d = _mm_max_ps(d,_mm_setzero_ps()); |
1519 | d2 = _mm_mul_ps(d,d); |
1520 | 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))))))); |
1521 | |
1522 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
1523 | |
1524 | /* Evaluate switch function */ |
1525 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
1526 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv21,_mm_mul_ps(velec,dsw)) ); |
1527 | velec = _mm_mul_ps(velec,sw); |
1528 | cutoff_mask = _mm_cmplt_ps(rsq21,rcutoff2); |
1529 | |
1530 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
1531 | velec = _mm_and_ps(velec,cutoff_mask); |
1532 | velec = _mm_andnot_ps(dummy_mask,velec); |
1533 | velecsum = _mm_add_ps(velecsum,velec); |
1534 | |
1535 | fscal = felec; |
1536 | |
1537 | fscal = _mm_and_ps(fscal,cutoff_mask); |
1538 | |
1539 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
1540 | |
1541 | /* Calculate temporary vectorial force */ |
1542 | tx = _mm_mul_ps(fscal,dx21); |
1543 | ty = _mm_mul_ps(fscal,dy21); |
1544 | tz = _mm_mul_ps(fscal,dz21); |
1545 | |
1546 | /* Update vectorial force */ |
1547 | fix2 = _mm_add_ps(fix2,tx); |
1548 | fiy2 = _mm_add_ps(fiy2,ty); |
1549 | fiz2 = _mm_add_ps(fiz2,tz); |
1550 | |
1551 | fjx1 = _mm_add_ps(fjx1,tx); |
1552 | fjy1 = _mm_add_ps(fjy1,ty); |
1553 | fjz1 = _mm_add_ps(fjz1,tz); |
1554 | |
1555 | } |
1556 | |
1557 | /************************** |
1558 | * CALCULATE INTERACTIONS * |
1559 | **************************/ |
1560 | |
1561 | if (gmx_mm_any_lt(rsq22,rcutoff2)) |
1562 | { |
1563 | |
1564 | r22 = _mm_mul_ps(rsq22,rinv22); |
1565 | r22 = _mm_andnot_ps(dummy_mask,r22); |
1566 | |
1567 | /* EWALD ELECTROSTATICS */ |
1568 | |
1569 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1570 | ewrt = _mm_mul_ps(r22,ewtabscale); |
1571 | ewitab = _mm_cvttps_epi32(ewrt); |
1572 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1573 | ewitab = _mm_slli_epi32(ewitab,2); |
1574 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
1575 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
1576 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
1577 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
1578 | _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); |
1579 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
1580 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
1581 | velec = _mm_mul_ps(qq22,_mm_sub_ps(rinv22,velec)); |
1582 | felec = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec)); |
1583 | |
1584 | d = _mm_sub_ps(r22,rswitch); |
1585 | d = _mm_max_ps(d,_mm_setzero_ps()); |
1586 | d2 = _mm_mul_ps(d,d); |
1587 | 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))))))); |
1588 | |
1589 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
1590 | |
1591 | /* Evaluate switch function */ |
1592 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
1593 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv22,_mm_mul_ps(velec,dsw)) ); |
1594 | velec = _mm_mul_ps(velec,sw); |
1595 | cutoff_mask = _mm_cmplt_ps(rsq22,rcutoff2); |
1596 | |
1597 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
1598 | velec = _mm_and_ps(velec,cutoff_mask); |
1599 | velec = _mm_andnot_ps(dummy_mask,velec); |
1600 | velecsum = _mm_add_ps(velecsum,velec); |
1601 | |
1602 | fscal = felec; |
1603 | |
1604 | fscal = _mm_and_ps(fscal,cutoff_mask); |
1605 | |
1606 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
1607 | |
1608 | /* Calculate temporary vectorial force */ |
1609 | tx = _mm_mul_ps(fscal,dx22); |
1610 | ty = _mm_mul_ps(fscal,dy22); |
1611 | tz = _mm_mul_ps(fscal,dz22); |
1612 | |
1613 | /* Update vectorial force */ |
1614 | fix2 = _mm_add_ps(fix2,tx); |
1615 | fiy2 = _mm_add_ps(fiy2,ty); |
1616 | fiz2 = _mm_add_ps(fiz2,tz); |
1617 | |
1618 | fjx2 = _mm_add_ps(fjx2,tx); |
1619 | fjy2 = _mm_add_ps(fjy2,ty); |
1620 | fjz2 = _mm_add_ps(fjz2,tz); |
1621 | |
1622 | } |
1623 | |
1624 | fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch; |
1625 | fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch; |
1626 | fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch; |
1627 | fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch; |
1628 | |
1629 | gmx_mm_decrement_3rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD, |
1630 | fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2); |
1631 | |
1632 | /* Inner loop uses 612 flops */ |
1633 | } |
1634 | |
1635 | /* End of innermost loop */ |
1636 | |
1637 | gmx_mm_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2, |
1638 | f+i_coord_offset,fshift+i_shift_offset); |
1639 | |
1640 | ggid = gid[iidx]; |
1641 | /* Update potential energies */ |
1642 | gmx_mm_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid); |
1643 | gmx_mm_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid); |
1644 | |
1645 | /* Increment number of inner iterations */ |
1646 | inneriter += j_index_end - j_index_start; |
1647 | |
1648 | /* Outer loop uses 20 flops */ |
1649 | } |
1650 | |
1651 | /* Increment number of outer iterations */ |
1652 | outeriter += nri; |
1653 | |
1654 | /* Update outer/inner flops */ |
1655 | |
1656 | inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_VF,outeriter*20 + inneriter*612)(nrnb)->n[eNR_NBKERNEL_ELEC_VDW_W3W3_VF] += outeriter*20 + inneriter*612; |
1657 | } |
1658 | /* |
1659 | * Gromacs nonbonded kernel: nb_kernel_ElecEwSw_VdwLJSw_GeomW3W3_F_sse4_1_single |
1660 | * Electrostatics interaction: Ewald |
1661 | * VdW interaction: LennardJones |
1662 | * Geometry: Water3-Water3 |
1663 | * Calculate force/pot: Force |
1664 | */ |
1665 | void |
1666 | nb_kernel_ElecEwSw_VdwLJSw_GeomW3W3_F_sse4_1_single |
1667 | (t_nblist * gmx_restrict nlist, |
1668 | rvec * gmx_restrict xx, |
1669 | rvec * gmx_restrict ff, |
1670 | t_forcerec * gmx_restrict fr, |
1671 | t_mdatoms * gmx_restrict mdatoms, |
1672 | nb_kernel_data_t gmx_unused__attribute__ ((unused)) * gmx_restrict kernel_data, |
1673 | t_nrnb * gmx_restrict nrnb) |
1674 | { |
1675 | /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or |
1676 | * just 0 for non-waters. |
1677 | * Suffixes A,B,C,D refer to j loop unrolling done with SSE, e.g. for the four different |
1678 | * jnr indices corresponding to data put in the four positions in the SIMD register. |
1679 | */ |
1680 | int i_shift_offset,i_coord_offset,outeriter,inneriter; |
1681 | int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx; |
1682 | int jnrA,jnrB,jnrC,jnrD; |
1683 | int jnrlistA,jnrlistB,jnrlistC,jnrlistD; |
1684 | int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD; |
1685 | int *iinr,*jindex,*jjnr,*shiftidx,*gid; |
1686 | real rcutoff_scalar; |
1687 | real *shiftvec,*fshift,*x,*f; |
1688 | real *fjptrA,*fjptrB,*fjptrC,*fjptrD; |
1689 | real scratch[4*DIM3]; |
1690 | __m128 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall; |
1691 | int vdwioffset0; |
1692 | __m128 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0; |
1693 | int vdwioffset1; |
1694 | __m128 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1; |
1695 | int vdwioffset2; |
1696 | __m128 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2; |
1697 | int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D; |
1698 | __m128 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0; |
1699 | int vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D; |
1700 | __m128 jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1; |
1701 | int vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D; |
1702 | __m128 jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2; |
1703 | __m128 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00; |
1704 | __m128 dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01; |
1705 | __m128 dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02; |
1706 | __m128 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10; |
1707 | __m128 dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11; |
1708 | __m128 dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12; |
1709 | __m128 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20; |
1710 | __m128 dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21; |
1711 | __m128 dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22; |
1712 | __m128 velec,felec,velecsum,facel,crf,krf,krf2; |
1713 | real *charge; |
1714 | int nvdwtype; |
1715 | __m128 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6; |
1716 | int *vdwtype; |
1717 | real *vdwparam; |
1718 | __m128 one_sixth = _mm_set1_ps(1.0/6.0); |
1719 | __m128 one_twelfth = _mm_set1_ps(1.0/12.0); |
1720 | __m128i ewitab; |
1721 | __m128 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV; |
1722 | real *ewtab; |
1723 | __m128 rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw; |
1724 | real rswitch_scalar,d_scalar; |
1725 | __m128 dummy_mask,cutoff_mask; |
1726 | __m128 signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) ); |
1727 | __m128 one = _mm_set1_ps(1.0); |
1728 | __m128 two = _mm_set1_ps(2.0); |
1729 | x = xx[0]; |
1730 | f = ff[0]; |
1731 | |
1732 | nri = nlist->nri; |
1733 | iinr = nlist->iinr; |
1734 | jindex = nlist->jindex; |
1735 | jjnr = nlist->jjnr; |
1736 | shiftidx = nlist->shift; |
1737 | gid = nlist->gid; |
1738 | shiftvec = fr->shift_vec[0]; |
1739 | fshift = fr->fshift[0]; |
1740 | facel = _mm_set1_ps(fr->epsfac); |
1741 | charge = mdatoms->chargeA; |
1742 | nvdwtype = fr->ntype; |
1743 | vdwparam = fr->nbfp; |
1744 | vdwtype = mdatoms->typeA; |
1745 | |
1746 | sh_ewald = _mm_set1_ps(fr->ic->sh_ewald); |
1747 | ewtab = fr->ic->tabq_coul_FDV0; |
1748 | ewtabscale = _mm_set1_ps(fr->ic->tabq_scale); |
1749 | ewtabhalfspace = _mm_set1_ps(0.5/fr->ic->tabq_scale); |
1750 | |
1751 | /* Setup water-specific parameters */ |
1752 | inr = nlist->iinr[0]; |
1753 | iq0 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+0])); |
1754 | iq1 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1])); |
1755 | iq2 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2])); |
1756 | vdwioffset0 = 2*nvdwtype*vdwtype[inr+0]; |
1757 | |
1758 | jq0 = _mm_set1_ps(charge[inr+0]); |
1759 | jq1 = _mm_set1_ps(charge[inr+1]); |
1760 | jq2 = _mm_set1_ps(charge[inr+2]); |
1761 | vdwjidx0A = 2*vdwtype[inr+0]; |
1762 | qq00 = _mm_mul_ps(iq0,jq0); |
1763 | c6_00 = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A]); |
1764 | c12_00 = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A+1]); |
1765 | qq01 = _mm_mul_ps(iq0,jq1); |
1766 | qq02 = _mm_mul_ps(iq0,jq2); |
1767 | qq10 = _mm_mul_ps(iq1,jq0); |
1768 | qq11 = _mm_mul_ps(iq1,jq1); |
1769 | qq12 = _mm_mul_ps(iq1,jq2); |
1770 | qq20 = _mm_mul_ps(iq2,jq0); |
1771 | qq21 = _mm_mul_ps(iq2,jq1); |
1772 | qq22 = _mm_mul_ps(iq2,jq2); |
1773 | |
1774 | /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */ |
1775 | rcutoff_scalar = fr->rcoulomb; |
1776 | rcutoff = _mm_set1_ps(rcutoff_scalar); |
1777 | rcutoff2 = _mm_mul_ps(rcutoff,rcutoff); |
1778 | |
1779 | rswitch_scalar = fr->rcoulomb_switch; |
1780 | rswitch = _mm_set1_ps(rswitch_scalar); |
1781 | /* Setup switch parameters */ |
1782 | d_scalar = rcutoff_scalar-rswitch_scalar; |
1783 | d = _mm_set1_ps(d_scalar); |
1784 | swV3 = _mm_set1_ps(-10.0/(d_scalar*d_scalar*d_scalar)); |
1785 | swV4 = _mm_set1_ps( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar)); |
1786 | swV5 = _mm_set1_ps( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar)); |
1787 | swF2 = _mm_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar)); |
1788 | swF3 = _mm_set1_ps( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar)); |
1789 | swF4 = _mm_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar)); |
1790 | |
1791 | /* Avoid stupid compiler warnings */ |
1792 | jnrA = jnrB = jnrC = jnrD = 0; |
1793 | j_coord_offsetA = 0; |
1794 | j_coord_offsetB = 0; |
1795 | j_coord_offsetC = 0; |
1796 | j_coord_offsetD = 0; |
1797 | |
1798 | outeriter = 0; |
1799 | inneriter = 0; |
1800 | |
1801 | for(iidx=0;iidx<4*DIM3;iidx++) |
1802 | { |
1803 | scratch[iidx] = 0.0; |
1804 | } |
1805 | |
1806 | /* Start outer loop over neighborlists */ |
1807 | for(iidx=0; iidx<nri; iidx++) |
1808 | { |
1809 | /* Load shift vector for this list */ |
1810 | i_shift_offset = DIM3*shiftidx[iidx]; |
1811 | |
1812 | /* Load limits for loop over neighbors */ |
1813 | j_index_start = jindex[iidx]; |
1814 | j_index_end = jindex[iidx+1]; |
1815 | |
1816 | /* Get outer coordinate index */ |
1817 | inr = iinr[iidx]; |
1818 | i_coord_offset = DIM3*inr; |
1819 | |
1820 | /* Load i particle coords and add shift vector */ |
1821 | gmx_mm_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset, |
1822 | &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2); |
1823 | |
1824 | fix0 = _mm_setzero_ps(); |
1825 | fiy0 = _mm_setzero_ps(); |
1826 | fiz0 = _mm_setzero_ps(); |
1827 | fix1 = _mm_setzero_ps(); |
1828 | fiy1 = _mm_setzero_ps(); |
1829 | fiz1 = _mm_setzero_ps(); |
1830 | fix2 = _mm_setzero_ps(); |
1831 | fiy2 = _mm_setzero_ps(); |
1832 | fiz2 = _mm_setzero_ps(); |
1833 | |
1834 | /* Start inner kernel loop */ |
1835 | for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4) |
1836 | { |
1837 | |
1838 | /* Get j neighbor index, and coordinate index */ |
1839 | jnrA = jjnr[jidx]; |
1840 | jnrB = jjnr[jidx+1]; |
1841 | jnrC = jjnr[jidx+2]; |
1842 | jnrD = jjnr[jidx+3]; |
1843 | j_coord_offsetA = DIM3*jnrA; |
1844 | j_coord_offsetB = DIM3*jnrB; |
1845 | j_coord_offsetC = DIM3*jnrC; |
1846 | j_coord_offsetD = DIM3*jnrD; |
1847 | |
1848 | /* load j atom coordinates */ |
1849 | gmx_mm_load_3rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB, |
1850 | x+j_coord_offsetC,x+j_coord_offsetD, |
1851 | &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2); |
1852 | |
1853 | /* Calculate displacement vector */ |
1854 | dx00 = _mm_sub_ps(ix0,jx0); |
1855 | dy00 = _mm_sub_ps(iy0,jy0); |
1856 | dz00 = _mm_sub_ps(iz0,jz0); |
1857 | dx01 = _mm_sub_ps(ix0,jx1); |
1858 | dy01 = _mm_sub_ps(iy0,jy1); |
1859 | dz01 = _mm_sub_ps(iz0,jz1); |
1860 | dx02 = _mm_sub_ps(ix0,jx2); |
1861 | dy02 = _mm_sub_ps(iy0,jy2); |
1862 | dz02 = _mm_sub_ps(iz0,jz2); |
1863 | dx10 = _mm_sub_ps(ix1,jx0); |
1864 | dy10 = _mm_sub_ps(iy1,jy0); |
1865 | dz10 = _mm_sub_ps(iz1,jz0); |
1866 | dx11 = _mm_sub_ps(ix1,jx1); |
1867 | dy11 = _mm_sub_ps(iy1,jy1); |
1868 | dz11 = _mm_sub_ps(iz1,jz1); |
1869 | dx12 = _mm_sub_ps(ix1,jx2); |
1870 | dy12 = _mm_sub_ps(iy1,jy2); |
1871 | dz12 = _mm_sub_ps(iz1,jz2); |
1872 | dx20 = _mm_sub_ps(ix2,jx0); |
1873 | dy20 = _mm_sub_ps(iy2,jy0); |
1874 | dz20 = _mm_sub_ps(iz2,jz0); |
1875 | dx21 = _mm_sub_ps(ix2,jx1); |
1876 | dy21 = _mm_sub_ps(iy2,jy1); |
1877 | dz21 = _mm_sub_ps(iz2,jz1); |
1878 | dx22 = _mm_sub_ps(ix2,jx2); |
1879 | dy22 = _mm_sub_ps(iy2,jy2); |
1880 | dz22 = _mm_sub_ps(iz2,jz2); |
1881 | |
1882 | /* Calculate squared distance and things based on it */ |
1883 | rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00); |
1884 | rsq01 = gmx_mm_calc_rsq_ps(dx01,dy01,dz01); |
1885 | rsq02 = gmx_mm_calc_rsq_ps(dx02,dy02,dz02); |
1886 | rsq10 = gmx_mm_calc_rsq_ps(dx10,dy10,dz10); |
1887 | rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11); |
1888 | rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12); |
1889 | rsq20 = gmx_mm_calc_rsq_ps(dx20,dy20,dz20); |
1890 | rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21); |
1891 | rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22); |
1892 | |
1893 | rinv00 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq00); |
1894 | rinv01 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq01); |
1895 | rinv02 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq02); |
1896 | rinv10 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq10); |
1897 | rinv11 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq11); |
1898 | rinv12 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq12); |
1899 | rinv20 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq20); |
1900 | rinv21 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq21); |
1901 | rinv22 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq22); |
1902 | |
1903 | rinvsq00 = _mm_mul_ps(rinv00,rinv00); |
1904 | rinvsq01 = _mm_mul_ps(rinv01,rinv01); |
1905 | rinvsq02 = _mm_mul_ps(rinv02,rinv02); |
1906 | rinvsq10 = _mm_mul_ps(rinv10,rinv10); |
1907 | rinvsq11 = _mm_mul_ps(rinv11,rinv11); |
1908 | rinvsq12 = _mm_mul_ps(rinv12,rinv12); |
1909 | rinvsq20 = _mm_mul_ps(rinv20,rinv20); |
1910 | rinvsq21 = _mm_mul_ps(rinv21,rinv21); |
1911 | rinvsq22 = _mm_mul_ps(rinv22,rinv22); |
1912 | |
1913 | fjx0 = _mm_setzero_ps(); |
1914 | fjy0 = _mm_setzero_ps(); |
1915 | fjz0 = _mm_setzero_ps(); |
1916 | fjx1 = _mm_setzero_ps(); |
1917 | fjy1 = _mm_setzero_ps(); |
1918 | fjz1 = _mm_setzero_ps(); |
1919 | fjx2 = _mm_setzero_ps(); |
1920 | fjy2 = _mm_setzero_ps(); |
1921 | fjz2 = _mm_setzero_ps(); |
1922 | |
1923 | /************************** |
1924 | * CALCULATE INTERACTIONS * |
1925 | **************************/ |
1926 | |
1927 | if (gmx_mm_any_lt(rsq00,rcutoff2)) |
1928 | { |
1929 | |
1930 | r00 = _mm_mul_ps(rsq00,rinv00); |
1931 | |
1932 | /* EWALD ELECTROSTATICS */ |
1933 | |
1934 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
1935 | ewrt = _mm_mul_ps(r00,ewtabscale); |
1936 | ewitab = _mm_cvttps_epi32(ewrt); |
1937 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
1938 | ewitab = _mm_slli_epi32(ewitab,2); |
1939 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
1940 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
1941 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
1942 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
1943 | _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); |
1944 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
1945 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
1946 | velec = _mm_mul_ps(qq00,_mm_sub_ps(rinv00,velec)); |
1947 | felec = _mm_mul_ps(_mm_mul_ps(qq00,rinv00),_mm_sub_ps(rinvsq00,felec)); |
1948 | |
1949 | /* LENNARD-JONES DISPERSION/REPULSION */ |
1950 | |
1951 | rinvsix = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00); |
1952 | vvdw6 = _mm_mul_ps(c6_00,rinvsix); |
1953 | vvdw12 = _mm_mul_ps(c12_00,_mm_mul_ps(rinvsix,rinvsix)); |
1954 | vvdw = _mm_sub_ps( _mm_mul_ps(vvdw12,one_twelfth) , _mm_mul_ps(vvdw6,one_sixth) ); |
1955 | fvdw = _mm_mul_ps(_mm_sub_ps(vvdw12,vvdw6),rinvsq00); |
1956 | |
1957 | d = _mm_sub_ps(r00,rswitch); |
1958 | d = _mm_max_ps(d,_mm_setzero_ps()); |
1959 | d2 = _mm_mul_ps(d,d); |
1960 | 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))))))); |
1961 | |
1962 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
1963 | |
1964 | /* Evaluate switch function */ |
1965 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
1966 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv00,_mm_mul_ps(velec,dsw)) ); |
1967 | fvdw = _mm_sub_ps( _mm_mul_ps(fvdw,sw) , _mm_mul_ps(rinv00,_mm_mul_ps(vvdw,dsw)) ); |
1968 | cutoff_mask = _mm_cmplt_ps(rsq00,rcutoff2); |
1969 | |
1970 | fscal = _mm_add_ps(felec,fvdw); |
1971 | |
1972 | fscal = _mm_and_ps(fscal,cutoff_mask); |
1973 | |
1974 | /* Calculate temporary vectorial force */ |
1975 | tx = _mm_mul_ps(fscal,dx00); |
1976 | ty = _mm_mul_ps(fscal,dy00); |
1977 | tz = _mm_mul_ps(fscal,dz00); |
1978 | |
1979 | /* Update vectorial force */ |
1980 | fix0 = _mm_add_ps(fix0,tx); |
1981 | fiy0 = _mm_add_ps(fiy0,ty); |
1982 | fiz0 = _mm_add_ps(fiz0,tz); |
1983 | |
1984 | fjx0 = _mm_add_ps(fjx0,tx); |
1985 | fjy0 = _mm_add_ps(fjy0,ty); |
1986 | fjz0 = _mm_add_ps(fjz0,tz); |
1987 | |
1988 | } |
1989 | |
1990 | /************************** |
1991 | * CALCULATE INTERACTIONS * |
1992 | **************************/ |
1993 | |
1994 | if (gmx_mm_any_lt(rsq01,rcutoff2)) |
1995 | { |
1996 | |
1997 | r01 = _mm_mul_ps(rsq01,rinv01); |
1998 | |
1999 | /* EWALD ELECTROSTATICS */ |
2000 | |
2001 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
2002 | ewrt = _mm_mul_ps(r01,ewtabscale); |
2003 | ewitab = _mm_cvttps_epi32(ewrt); |
2004 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
2005 | ewitab = _mm_slli_epi32(ewitab,2); |
2006 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
2007 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
2008 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
2009 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
2010 | _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); |
2011 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
2012 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
2013 | velec = _mm_mul_ps(qq01,_mm_sub_ps(rinv01,velec)); |
2014 | felec = _mm_mul_ps(_mm_mul_ps(qq01,rinv01),_mm_sub_ps(rinvsq01,felec)); |
2015 | |
2016 | d = _mm_sub_ps(r01,rswitch); |
2017 | d = _mm_max_ps(d,_mm_setzero_ps()); |
2018 | d2 = _mm_mul_ps(d,d); |
2019 | 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))))))); |
2020 | |
2021 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
2022 | |
2023 | /* Evaluate switch function */ |
2024 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
2025 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv01,_mm_mul_ps(velec,dsw)) ); |
2026 | cutoff_mask = _mm_cmplt_ps(rsq01,rcutoff2); |
2027 | |
2028 | fscal = felec; |
2029 | |
2030 | fscal = _mm_and_ps(fscal,cutoff_mask); |
2031 | |
2032 | /* Calculate temporary vectorial force */ |
2033 | tx = _mm_mul_ps(fscal,dx01); |
2034 | ty = _mm_mul_ps(fscal,dy01); |
2035 | tz = _mm_mul_ps(fscal,dz01); |
2036 | |
2037 | /* Update vectorial force */ |
2038 | fix0 = _mm_add_ps(fix0,tx); |
2039 | fiy0 = _mm_add_ps(fiy0,ty); |
2040 | fiz0 = _mm_add_ps(fiz0,tz); |
2041 | |
2042 | fjx1 = _mm_add_ps(fjx1,tx); |
2043 | fjy1 = _mm_add_ps(fjy1,ty); |
2044 | fjz1 = _mm_add_ps(fjz1,tz); |
2045 | |
2046 | } |
2047 | |
2048 | /************************** |
2049 | * CALCULATE INTERACTIONS * |
2050 | **************************/ |
2051 | |
2052 | if (gmx_mm_any_lt(rsq02,rcutoff2)) |
2053 | { |
2054 | |
2055 | r02 = _mm_mul_ps(rsq02,rinv02); |
2056 | |
2057 | /* EWALD ELECTROSTATICS */ |
2058 | |
2059 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
2060 | ewrt = _mm_mul_ps(r02,ewtabscale); |
2061 | ewitab = _mm_cvttps_epi32(ewrt); |
2062 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
2063 | ewitab = _mm_slli_epi32(ewitab,2); |
2064 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
2065 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
2066 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
2067 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
2068 | _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); |
2069 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
2070 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
2071 | velec = _mm_mul_ps(qq02,_mm_sub_ps(rinv02,velec)); |
2072 | felec = _mm_mul_ps(_mm_mul_ps(qq02,rinv02),_mm_sub_ps(rinvsq02,felec)); |
2073 | |
2074 | d = _mm_sub_ps(r02,rswitch); |
2075 | d = _mm_max_ps(d,_mm_setzero_ps()); |
2076 | d2 = _mm_mul_ps(d,d); |
2077 | 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))))))); |
2078 | |
2079 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
2080 | |
2081 | /* Evaluate switch function */ |
2082 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
2083 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv02,_mm_mul_ps(velec,dsw)) ); |
2084 | cutoff_mask = _mm_cmplt_ps(rsq02,rcutoff2); |
2085 | |
2086 | fscal = felec; |
2087 | |
2088 | fscal = _mm_and_ps(fscal,cutoff_mask); |
2089 | |
2090 | /* Calculate temporary vectorial force */ |
2091 | tx = _mm_mul_ps(fscal,dx02); |
2092 | ty = _mm_mul_ps(fscal,dy02); |
2093 | tz = _mm_mul_ps(fscal,dz02); |
2094 | |
2095 | /* Update vectorial force */ |
2096 | fix0 = _mm_add_ps(fix0,tx); |
2097 | fiy0 = _mm_add_ps(fiy0,ty); |
2098 | fiz0 = _mm_add_ps(fiz0,tz); |
2099 | |
2100 | fjx2 = _mm_add_ps(fjx2,tx); |
2101 | fjy2 = _mm_add_ps(fjy2,ty); |
2102 | fjz2 = _mm_add_ps(fjz2,tz); |
2103 | |
2104 | } |
2105 | |
2106 | /************************** |
2107 | * CALCULATE INTERACTIONS * |
2108 | **************************/ |
2109 | |
2110 | if (gmx_mm_any_lt(rsq10,rcutoff2)) |
2111 | { |
2112 | |
2113 | r10 = _mm_mul_ps(rsq10,rinv10); |
2114 | |
2115 | /* EWALD ELECTROSTATICS */ |
2116 | |
2117 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
2118 | ewrt = _mm_mul_ps(r10,ewtabscale); |
2119 | ewitab = _mm_cvttps_epi32(ewrt); |
2120 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
2121 | ewitab = _mm_slli_epi32(ewitab,2); |
2122 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
2123 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
2124 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
2125 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
2126 | _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); |
2127 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
2128 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
2129 | velec = _mm_mul_ps(qq10,_mm_sub_ps(rinv10,velec)); |
2130 | felec = _mm_mul_ps(_mm_mul_ps(qq10,rinv10),_mm_sub_ps(rinvsq10,felec)); |
2131 | |
2132 | d = _mm_sub_ps(r10,rswitch); |
2133 | d = _mm_max_ps(d,_mm_setzero_ps()); |
2134 | d2 = _mm_mul_ps(d,d); |
2135 | 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))))))); |
2136 | |
2137 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
2138 | |
2139 | /* Evaluate switch function */ |
2140 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
2141 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv10,_mm_mul_ps(velec,dsw)) ); |
2142 | cutoff_mask = _mm_cmplt_ps(rsq10,rcutoff2); |
2143 | |
2144 | fscal = felec; |
2145 | |
2146 | fscal = _mm_and_ps(fscal,cutoff_mask); |
2147 | |
2148 | /* Calculate temporary vectorial force */ |
2149 | tx = _mm_mul_ps(fscal,dx10); |
2150 | ty = _mm_mul_ps(fscal,dy10); |
2151 | tz = _mm_mul_ps(fscal,dz10); |
2152 | |
2153 | /* Update vectorial force */ |
2154 | fix1 = _mm_add_ps(fix1,tx); |
2155 | fiy1 = _mm_add_ps(fiy1,ty); |
2156 | fiz1 = _mm_add_ps(fiz1,tz); |
2157 | |
2158 | fjx0 = _mm_add_ps(fjx0,tx); |
2159 | fjy0 = _mm_add_ps(fjy0,ty); |
2160 | fjz0 = _mm_add_ps(fjz0,tz); |
2161 | |
2162 | } |
2163 | |
2164 | /************************** |
2165 | * CALCULATE INTERACTIONS * |
2166 | **************************/ |
2167 | |
2168 | if (gmx_mm_any_lt(rsq11,rcutoff2)) |
2169 | { |
2170 | |
2171 | r11 = _mm_mul_ps(rsq11,rinv11); |
2172 | |
2173 | /* EWALD ELECTROSTATICS */ |
2174 | |
2175 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
2176 | ewrt = _mm_mul_ps(r11,ewtabscale); |
2177 | ewitab = _mm_cvttps_epi32(ewrt); |
2178 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
2179 | ewitab = _mm_slli_epi32(ewitab,2); |
2180 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
2181 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
2182 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
2183 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
2184 | _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); |
2185 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
2186 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
2187 | velec = _mm_mul_ps(qq11,_mm_sub_ps(rinv11,velec)); |
2188 | felec = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec)); |
2189 | |
2190 | d = _mm_sub_ps(r11,rswitch); |
2191 | d = _mm_max_ps(d,_mm_setzero_ps()); |
2192 | d2 = _mm_mul_ps(d,d); |
2193 | 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))))))); |
2194 | |
2195 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
2196 | |
2197 | /* Evaluate switch function */ |
2198 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
2199 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv11,_mm_mul_ps(velec,dsw)) ); |
2200 | cutoff_mask = _mm_cmplt_ps(rsq11,rcutoff2); |
2201 | |
2202 | fscal = felec; |
2203 | |
2204 | fscal = _mm_and_ps(fscal,cutoff_mask); |
2205 | |
2206 | /* Calculate temporary vectorial force */ |
2207 | tx = _mm_mul_ps(fscal,dx11); |
2208 | ty = _mm_mul_ps(fscal,dy11); |
2209 | tz = _mm_mul_ps(fscal,dz11); |
2210 | |
2211 | /* Update vectorial force */ |
2212 | fix1 = _mm_add_ps(fix1,tx); |
2213 | fiy1 = _mm_add_ps(fiy1,ty); |
2214 | fiz1 = _mm_add_ps(fiz1,tz); |
2215 | |
2216 | fjx1 = _mm_add_ps(fjx1,tx); |
2217 | fjy1 = _mm_add_ps(fjy1,ty); |
2218 | fjz1 = _mm_add_ps(fjz1,tz); |
2219 | |
2220 | } |
2221 | |
2222 | /************************** |
2223 | * CALCULATE INTERACTIONS * |
2224 | **************************/ |
2225 | |
2226 | if (gmx_mm_any_lt(rsq12,rcutoff2)) |
2227 | { |
2228 | |
2229 | r12 = _mm_mul_ps(rsq12,rinv12); |
2230 | |
2231 | /* EWALD ELECTROSTATICS */ |
2232 | |
2233 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
2234 | ewrt = _mm_mul_ps(r12,ewtabscale); |
2235 | ewitab = _mm_cvttps_epi32(ewrt); |
2236 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
2237 | ewitab = _mm_slli_epi32(ewitab,2); |
2238 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
2239 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
2240 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
2241 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
2242 | _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); |
2243 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
2244 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
2245 | velec = _mm_mul_ps(qq12,_mm_sub_ps(rinv12,velec)); |
2246 | felec = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec)); |
2247 | |
2248 | d = _mm_sub_ps(r12,rswitch); |
2249 | d = _mm_max_ps(d,_mm_setzero_ps()); |
2250 | d2 = _mm_mul_ps(d,d); |
2251 | 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))))))); |
2252 | |
2253 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
2254 | |
2255 | /* Evaluate switch function */ |
2256 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
2257 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv12,_mm_mul_ps(velec,dsw)) ); |
2258 | cutoff_mask = _mm_cmplt_ps(rsq12,rcutoff2); |
2259 | |
2260 | fscal = felec; |
2261 | |
2262 | fscal = _mm_and_ps(fscal,cutoff_mask); |
2263 | |
2264 | /* Calculate temporary vectorial force */ |
2265 | tx = _mm_mul_ps(fscal,dx12); |
2266 | ty = _mm_mul_ps(fscal,dy12); |
2267 | tz = _mm_mul_ps(fscal,dz12); |
2268 | |
2269 | /* Update vectorial force */ |
2270 | fix1 = _mm_add_ps(fix1,tx); |
2271 | fiy1 = _mm_add_ps(fiy1,ty); |
2272 | fiz1 = _mm_add_ps(fiz1,tz); |
2273 | |
2274 | fjx2 = _mm_add_ps(fjx2,tx); |
2275 | fjy2 = _mm_add_ps(fjy2,ty); |
2276 | fjz2 = _mm_add_ps(fjz2,tz); |
2277 | |
2278 | } |
2279 | |
2280 | /************************** |
2281 | * CALCULATE INTERACTIONS * |
2282 | **************************/ |
2283 | |
2284 | if (gmx_mm_any_lt(rsq20,rcutoff2)) |
2285 | { |
2286 | |
2287 | r20 = _mm_mul_ps(rsq20,rinv20); |
2288 | |
2289 | /* EWALD ELECTROSTATICS */ |
2290 | |
2291 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
2292 | ewrt = _mm_mul_ps(r20,ewtabscale); |
2293 | ewitab = _mm_cvttps_epi32(ewrt); |
2294 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
2295 | ewitab = _mm_slli_epi32(ewitab,2); |
2296 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
2297 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
2298 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
2299 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
2300 | _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); |
2301 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
2302 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
2303 | velec = _mm_mul_ps(qq20,_mm_sub_ps(rinv20,velec)); |
2304 | felec = _mm_mul_ps(_mm_mul_ps(qq20,rinv20),_mm_sub_ps(rinvsq20,felec)); |
2305 | |
2306 | d = _mm_sub_ps(r20,rswitch); |
2307 | d = _mm_max_ps(d,_mm_setzero_ps()); |
2308 | d2 = _mm_mul_ps(d,d); |
2309 | 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))))))); |
2310 | |
2311 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
2312 | |
2313 | /* Evaluate switch function */ |
2314 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
2315 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv20,_mm_mul_ps(velec,dsw)) ); |
2316 | cutoff_mask = _mm_cmplt_ps(rsq20,rcutoff2); |
2317 | |
2318 | fscal = felec; |
2319 | |
2320 | fscal = _mm_and_ps(fscal,cutoff_mask); |
2321 | |
2322 | /* Calculate temporary vectorial force */ |
2323 | tx = _mm_mul_ps(fscal,dx20); |
2324 | ty = _mm_mul_ps(fscal,dy20); |
2325 | tz = _mm_mul_ps(fscal,dz20); |
2326 | |
2327 | /* Update vectorial force */ |
2328 | fix2 = _mm_add_ps(fix2,tx); |
2329 | fiy2 = _mm_add_ps(fiy2,ty); |
2330 | fiz2 = _mm_add_ps(fiz2,tz); |
2331 | |
2332 | fjx0 = _mm_add_ps(fjx0,tx); |
2333 | fjy0 = _mm_add_ps(fjy0,ty); |
2334 | fjz0 = _mm_add_ps(fjz0,tz); |
2335 | |
2336 | } |
2337 | |
2338 | /************************** |
2339 | * CALCULATE INTERACTIONS * |
2340 | **************************/ |
2341 | |
2342 | if (gmx_mm_any_lt(rsq21,rcutoff2)) |
2343 | { |
2344 | |
2345 | r21 = _mm_mul_ps(rsq21,rinv21); |
2346 | |
2347 | /* EWALD ELECTROSTATICS */ |
2348 | |
2349 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
2350 | ewrt = _mm_mul_ps(r21,ewtabscale); |
2351 | ewitab = _mm_cvttps_epi32(ewrt); |
2352 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
2353 | ewitab = _mm_slli_epi32(ewitab,2); |
2354 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
2355 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
2356 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
2357 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
2358 | _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); |
2359 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
2360 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
2361 | velec = _mm_mul_ps(qq21,_mm_sub_ps(rinv21,velec)); |
2362 | felec = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec)); |
2363 | |
2364 | d = _mm_sub_ps(r21,rswitch); |
2365 | d = _mm_max_ps(d,_mm_setzero_ps()); |
2366 | d2 = _mm_mul_ps(d,d); |
2367 | 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))))))); |
2368 | |
2369 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
2370 | |
2371 | /* Evaluate switch function */ |
2372 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
2373 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv21,_mm_mul_ps(velec,dsw)) ); |
2374 | cutoff_mask = _mm_cmplt_ps(rsq21,rcutoff2); |
2375 | |
2376 | fscal = felec; |
2377 | |
2378 | fscal = _mm_and_ps(fscal,cutoff_mask); |
2379 | |
2380 | /* Calculate temporary vectorial force */ |
2381 | tx = _mm_mul_ps(fscal,dx21); |
2382 | ty = _mm_mul_ps(fscal,dy21); |
2383 | tz = _mm_mul_ps(fscal,dz21); |
2384 | |
2385 | /* Update vectorial force */ |
2386 | fix2 = _mm_add_ps(fix2,tx); |
2387 | fiy2 = _mm_add_ps(fiy2,ty); |
2388 | fiz2 = _mm_add_ps(fiz2,tz); |
2389 | |
2390 | fjx1 = _mm_add_ps(fjx1,tx); |
2391 | fjy1 = _mm_add_ps(fjy1,ty); |
2392 | fjz1 = _mm_add_ps(fjz1,tz); |
2393 | |
2394 | } |
2395 | |
2396 | /************************** |
2397 | * CALCULATE INTERACTIONS * |
2398 | **************************/ |
2399 | |
2400 | if (gmx_mm_any_lt(rsq22,rcutoff2)) |
2401 | { |
2402 | |
2403 | r22 = _mm_mul_ps(rsq22,rinv22); |
2404 | |
2405 | /* EWALD ELECTROSTATICS */ |
2406 | |
2407 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
2408 | ewrt = _mm_mul_ps(r22,ewtabscale); |
2409 | ewitab = _mm_cvttps_epi32(ewrt); |
2410 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
2411 | ewitab = _mm_slli_epi32(ewitab,2); |
2412 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
2413 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
2414 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
2415 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
2416 | _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); |
2417 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
2418 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
2419 | velec = _mm_mul_ps(qq22,_mm_sub_ps(rinv22,velec)); |
2420 | felec = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec)); |
2421 | |
2422 | d = _mm_sub_ps(r22,rswitch); |
2423 | d = _mm_max_ps(d,_mm_setzero_ps()); |
2424 | d2 = _mm_mul_ps(d,d); |
2425 | 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))))))); |
2426 | |
2427 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
2428 | |
2429 | /* Evaluate switch function */ |
2430 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
2431 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv22,_mm_mul_ps(velec,dsw)) ); |
2432 | cutoff_mask = _mm_cmplt_ps(rsq22,rcutoff2); |
2433 | |
2434 | fscal = felec; |
2435 | |
2436 | fscal = _mm_and_ps(fscal,cutoff_mask); |
2437 | |
2438 | /* Calculate temporary vectorial force */ |
2439 | tx = _mm_mul_ps(fscal,dx22); |
2440 | ty = _mm_mul_ps(fscal,dy22); |
2441 | tz = _mm_mul_ps(fscal,dz22); |
2442 | |
2443 | /* Update vectorial force */ |
2444 | fix2 = _mm_add_ps(fix2,tx); |
2445 | fiy2 = _mm_add_ps(fiy2,ty); |
2446 | fiz2 = _mm_add_ps(fiz2,tz); |
2447 | |
2448 | fjx2 = _mm_add_ps(fjx2,tx); |
2449 | fjy2 = _mm_add_ps(fjy2,ty); |
2450 | fjz2 = _mm_add_ps(fjz2,tz); |
2451 | |
2452 | } |
2453 | |
2454 | fjptrA = f+j_coord_offsetA; |
2455 | fjptrB = f+j_coord_offsetB; |
2456 | fjptrC = f+j_coord_offsetC; |
2457 | fjptrD = f+j_coord_offsetD; |
2458 | |
2459 | gmx_mm_decrement_3rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD, |
2460 | fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2); |
2461 | |
2462 | /* Inner loop uses 573 flops */ |
2463 | } |
2464 | |
2465 | if(jidx<j_index_end) |
2466 | { |
2467 | |
2468 | /* Get j neighbor index, and coordinate index */ |
2469 | jnrlistA = jjnr[jidx]; |
2470 | jnrlistB = jjnr[jidx+1]; |
2471 | jnrlistC = jjnr[jidx+2]; |
2472 | jnrlistD = jjnr[jidx+3]; |
2473 | /* Sign of each element will be negative for non-real atoms. |
2474 | * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones, |
2475 | * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries. |
2476 | */ |
2477 | dummy_mask = gmx_mm_castsi128_ps_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())); |
2478 | jnrA = (jnrlistA>=0) ? jnrlistA : 0; |
2479 | jnrB = (jnrlistB>=0) ? jnrlistB : 0; |
2480 | jnrC = (jnrlistC>=0) ? jnrlistC : 0; |
2481 | jnrD = (jnrlistD>=0) ? jnrlistD : 0; |
2482 | j_coord_offsetA = DIM3*jnrA; |
2483 | j_coord_offsetB = DIM3*jnrB; |
2484 | j_coord_offsetC = DIM3*jnrC; |
2485 | j_coord_offsetD = DIM3*jnrD; |
2486 | |
2487 | /* load j atom coordinates */ |
2488 | gmx_mm_load_3rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB, |
2489 | x+j_coord_offsetC,x+j_coord_offsetD, |
2490 | &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2); |
2491 | |
2492 | /* Calculate displacement vector */ |
2493 | dx00 = _mm_sub_ps(ix0,jx0); |
2494 | dy00 = _mm_sub_ps(iy0,jy0); |
2495 | dz00 = _mm_sub_ps(iz0,jz0); |
2496 | dx01 = _mm_sub_ps(ix0,jx1); |
2497 | dy01 = _mm_sub_ps(iy0,jy1); |
2498 | dz01 = _mm_sub_ps(iz0,jz1); |
2499 | dx02 = _mm_sub_ps(ix0,jx2); |
2500 | dy02 = _mm_sub_ps(iy0,jy2); |
2501 | dz02 = _mm_sub_ps(iz0,jz2); |
2502 | dx10 = _mm_sub_ps(ix1,jx0); |
2503 | dy10 = _mm_sub_ps(iy1,jy0); |
2504 | dz10 = _mm_sub_ps(iz1,jz0); |
2505 | dx11 = _mm_sub_ps(ix1,jx1); |
2506 | dy11 = _mm_sub_ps(iy1,jy1); |
2507 | dz11 = _mm_sub_ps(iz1,jz1); |
2508 | dx12 = _mm_sub_ps(ix1,jx2); |
2509 | dy12 = _mm_sub_ps(iy1,jy2); |
2510 | dz12 = _mm_sub_ps(iz1,jz2); |
2511 | dx20 = _mm_sub_ps(ix2,jx0); |
2512 | dy20 = _mm_sub_ps(iy2,jy0); |
2513 | dz20 = _mm_sub_ps(iz2,jz0); |
2514 | dx21 = _mm_sub_ps(ix2,jx1); |
2515 | dy21 = _mm_sub_ps(iy2,jy1); |
2516 | dz21 = _mm_sub_ps(iz2,jz1); |
2517 | dx22 = _mm_sub_ps(ix2,jx2); |
2518 | dy22 = _mm_sub_ps(iy2,jy2); |
2519 | dz22 = _mm_sub_ps(iz2,jz2); |
2520 | |
2521 | /* Calculate squared distance and things based on it */ |
2522 | rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00); |
2523 | rsq01 = gmx_mm_calc_rsq_ps(dx01,dy01,dz01); |
2524 | rsq02 = gmx_mm_calc_rsq_ps(dx02,dy02,dz02); |
2525 | rsq10 = gmx_mm_calc_rsq_ps(dx10,dy10,dz10); |
2526 | rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11); |
2527 | rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12); |
2528 | rsq20 = gmx_mm_calc_rsq_ps(dx20,dy20,dz20); |
2529 | rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21); |
2530 | rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22); |
2531 | |
2532 | rinv00 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq00); |
2533 | rinv01 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq01); |
2534 | rinv02 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq02); |
2535 | rinv10 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq10); |
2536 | rinv11 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq11); |
2537 | rinv12 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq12); |
2538 | rinv20 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq20); |
2539 | rinv21 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq21); |
2540 | rinv22 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq22); |
2541 | |
2542 | rinvsq00 = _mm_mul_ps(rinv00,rinv00); |
2543 | rinvsq01 = _mm_mul_ps(rinv01,rinv01); |
2544 | rinvsq02 = _mm_mul_ps(rinv02,rinv02); |
2545 | rinvsq10 = _mm_mul_ps(rinv10,rinv10); |
2546 | rinvsq11 = _mm_mul_ps(rinv11,rinv11); |
2547 | rinvsq12 = _mm_mul_ps(rinv12,rinv12); |
2548 | rinvsq20 = _mm_mul_ps(rinv20,rinv20); |
2549 | rinvsq21 = _mm_mul_ps(rinv21,rinv21); |
2550 | rinvsq22 = _mm_mul_ps(rinv22,rinv22); |
2551 | |
2552 | fjx0 = _mm_setzero_ps(); |
2553 | fjy0 = _mm_setzero_ps(); |
2554 | fjz0 = _mm_setzero_ps(); |
2555 | fjx1 = _mm_setzero_ps(); |
2556 | fjy1 = _mm_setzero_ps(); |
2557 | fjz1 = _mm_setzero_ps(); |
2558 | fjx2 = _mm_setzero_ps(); |
2559 | fjy2 = _mm_setzero_ps(); |
2560 | fjz2 = _mm_setzero_ps(); |
2561 | |
2562 | /************************** |
2563 | * CALCULATE INTERACTIONS * |
2564 | **************************/ |
2565 | |
2566 | if (gmx_mm_any_lt(rsq00,rcutoff2)) |
2567 | { |
2568 | |
2569 | r00 = _mm_mul_ps(rsq00,rinv00); |
2570 | r00 = _mm_andnot_ps(dummy_mask,r00); |
2571 | |
2572 | /* EWALD ELECTROSTATICS */ |
2573 | |
2574 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
2575 | ewrt = _mm_mul_ps(r00,ewtabscale); |
2576 | ewitab = _mm_cvttps_epi32(ewrt); |
2577 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
2578 | ewitab = _mm_slli_epi32(ewitab,2); |
2579 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
2580 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
2581 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
2582 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
2583 | _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); |
2584 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
2585 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
2586 | velec = _mm_mul_ps(qq00,_mm_sub_ps(rinv00,velec)); |
2587 | felec = _mm_mul_ps(_mm_mul_ps(qq00,rinv00),_mm_sub_ps(rinvsq00,felec)); |
2588 | |
2589 | /* LENNARD-JONES DISPERSION/REPULSION */ |
2590 | |
2591 | rinvsix = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00); |
2592 | vvdw6 = _mm_mul_ps(c6_00,rinvsix); |
2593 | vvdw12 = _mm_mul_ps(c12_00,_mm_mul_ps(rinvsix,rinvsix)); |
2594 | vvdw = _mm_sub_ps( _mm_mul_ps(vvdw12,one_twelfth) , _mm_mul_ps(vvdw6,one_sixth) ); |
2595 | fvdw = _mm_mul_ps(_mm_sub_ps(vvdw12,vvdw6),rinvsq00); |
2596 | |
2597 | d = _mm_sub_ps(r00,rswitch); |
2598 | d = _mm_max_ps(d,_mm_setzero_ps()); |
2599 | d2 = _mm_mul_ps(d,d); |
2600 | 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))))))); |
2601 | |
2602 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
2603 | |
2604 | /* Evaluate switch function */ |
2605 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
2606 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv00,_mm_mul_ps(velec,dsw)) ); |
2607 | fvdw = _mm_sub_ps( _mm_mul_ps(fvdw,sw) , _mm_mul_ps(rinv00,_mm_mul_ps(vvdw,dsw)) ); |
2608 | cutoff_mask = _mm_cmplt_ps(rsq00,rcutoff2); |
2609 | |
2610 | fscal = _mm_add_ps(felec,fvdw); |
2611 | |
2612 | fscal = _mm_and_ps(fscal,cutoff_mask); |
2613 | |
2614 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
2615 | |
2616 | /* Calculate temporary vectorial force */ |
2617 | tx = _mm_mul_ps(fscal,dx00); |
2618 | ty = _mm_mul_ps(fscal,dy00); |
2619 | tz = _mm_mul_ps(fscal,dz00); |
2620 | |
2621 | /* Update vectorial force */ |
2622 | fix0 = _mm_add_ps(fix0,tx); |
2623 | fiy0 = _mm_add_ps(fiy0,ty); |
2624 | fiz0 = _mm_add_ps(fiz0,tz); |
2625 | |
2626 | fjx0 = _mm_add_ps(fjx0,tx); |
2627 | fjy0 = _mm_add_ps(fjy0,ty); |
2628 | fjz0 = _mm_add_ps(fjz0,tz); |
2629 | |
2630 | } |
2631 | |
2632 | /************************** |
2633 | * CALCULATE INTERACTIONS * |
2634 | **************************/ |
2635 | |
2636 | if (gmx_mm_any_lt(rsq01,rcutoff2)) |
2637 | { |
2638 | |
2639 | r01 = _mm_mul_ps(rsq01,rinv01); |
2640 | r01 = _mm_andnot_ps(dummy_mask,r01); |
2641 | |
2642 | /* EWALD ELECTROSTATICS */ |
2643 | |
2644 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
2645 | ewrt = _mm_mul_ps(r01,ewtabscale); |
2646 | ewitab = _mm_cvttps_epi32(ewrt); |
2647 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
2648 | ewitab = _mm_slli_epi32(ewitab,2); |
2649 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
2650 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
2651 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
2652 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
2653 | _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); |
2654 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
2655 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
2656 | velec = _mm_mul_ps(qq01,_mm_sub_ps(rinv01,velec)); |
2657 | felec = _mm_mul_ps(_mm_mul_ps(qq01,rinv01),_mm_sub_ps(rinvsq01,felec)); |
2658 | |
2659 | d = _mm_sub_ps(r01,rswitch); |
2660 | d = _mm_max_ps(d,_mm_setzero_ps()); |
2661 | d2 = _mm_mul_ps(d,d); |
2662 | 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))))))); |
2663 | |
2664 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
2665 | |
2666 | /* Evaluate switch function */ |
2667 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
2668 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv01,_mm_mul_ps(velec,dsw)) ); |
2669 | cutoff_mask = _mm_cmplt_ps(rsq01,rcutoff2); |
2670 | |
2671 | fscal = felec; |
2672 | |
2673 | fscal = _mm_and_ps(fscal,cutoff_mask); |
2674 | |
2675 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
2676 | |
2677 | /* Calculate temporary vectorial force */ |
2678 | tx = _mm_mul_ps(fscal,dx01); |
2679 | ty = _mm_mul_ps(fscal,dy01); |
2680 | tz = _mm_mul_ps(fscal,dz01); |
2681 | |
2682 | /* Update vectorial force */ |
2683 | fix0 = _mm_add_ps(fix0,tx); |
2684 | fiy0 = _mm_add_ps(fiy0,ty); |
2685 | fiz0 = _mm_add_ps(fiz0,tz); |
2686 | |
2687 | fjx1 = _mm_add_ps(fjx1,tx); |
2688 | fjy1 = _mm_add_ps(fjy1,ty); |
2689 | fjz1 = _mm_add_ps(fjz1,tz); |
2690 | |
2691 | } |
2692 | |
2693 | /************************** |
2694 | * CALCULATE INTERACTIONS * |
2695 | **************************/ |
2696 | |
2697 | if (gmx_mm_any_lt(rsq02,rcutoff2)) |
2698 | { |
2699 | |
2700 | r02 = _mm_mul_ps(rsq02,rinv02); |
2701 | r02 = _mm_andnot_ps(dummy_mask,r02); |
2702 | |
2703 | /* EWALD ELECTROSTATICS */ |
2704 | |
2705 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
2706 | ewrt = _mm_mul_ps(r02,ewtabscale); |
2707 | ewitab = _mm_cvttps_epi32(ewrt); |
2708 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
2709 | ewitab = _mm_slli_epi32(ewitab,2); |
2710 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
2711 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
2712 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
2713 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
2714 | _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); |
2715 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
2716 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
2717 | velec = _mm_mul_ps(qq02,_mm_sub_ps(rinv02,velec)); |
2718 | felec = _mm_mul_ps(_mm_mul_ps(qq02,rinv02),_mm_sub_ps(rinvsq02,felec)); |
2719 | |
2720 | d = _mm_sub_ps(r02,rswitch); |
2721 | d = _mm_max_ps(d,_mm_setzero_ps()); |
2722 | d2 = _mm_mul_ps(d,d); |
2723 | 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))))))); |
2724 | |
2725 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
2726 | |
2727 | /* Evaluate switch function */ |
2728 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
2729 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv02,_mm_mul_ps(velec,dsw)) ); |
2730 | cutoff_mask = _mm_cmplt_ps(rsq02,rcutoff2); |
2731 | |
2732 | fscal = felec; |
2733 | |
2734 | fscal = _mm_and_ps(fscal,cutoff_mask); |
2735 | |
2736 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
2737 | |
2738 | /* Calculate temporary vectorial force */ |
2739 | tx = _mm_mul_ps(fscal,dx02); |
2740 | ty = _mm_mul_ps(fscal,dy02); |
2741 | tz = _mm_mul_ps(fscal,dz02); |
2742 | |
2743 | /* Update vectorial force */ |
2744 | fix0 = _mm_add_ps(fix0,tx); |
2745 | fiy0 = _mm_add_ps(fiy0,ty); |
2746 | fiz0 = _mm_add_ps(fiz0,tz); |
2747 | |
2748 | fjx2 = _mm_add_ps(fjx2,tx); |
2749 | fjy2 = _mm_add_ps(fjy2,ty); |
2750 | fjz2 = _mm_add_ps(fjz2,tz); |
2751 | |
2752 | } |
2753 | |
2754 | /************************** |
2755 | * CALCULATE INTERACTIONS * |
2756 | **************************/ |
2757 | |
2758 | if (gmx_mm_any_lt(rsq10,rcutoff2)) |
2759 | { |
2760 | |
2761 | r10 = _mm_mul_ps(rsq10,rinv10); |
2762 | r10 = _mm_andnot_ps(dummy_mask,r10); |
2763 | |
2764 | /* EWALD ELECTROSTATICS */ |
2765 | |
2766 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
2767 | ewrt = _mm_mul_ps(r10,ewtabscale); |
2768 | ewitab = _mm_cvttps_epi32(ewrt); |
2769 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
2770 | ewitab = _mm_slli_epi32(ewitab,2); |
2771 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
2772 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
2773 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
2774 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
2775 | _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); |
2776 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
2777 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
2778 | velec = _mm_mul_ps(qq10,_mm_sub_ps(rinv10,velec)); |
2779 | felec = _mm_mul_ps(_mm_mul_ps(qq10,rinv10),_mm_sub_ps(rinvsq10,felec)); |
2780 | |
2781 | d = _mm_sub_ps(r10,rswitch); |
2782 | d = _mm_max_ps(d,_mm_setzero_ps()); |
2783 | d2 = _mm_mul_ps(d,d); |
2784 | 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))))))); |
2785 | |
2786 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
2787 | |
2788 | /* Evaluate switch function */ |
2789 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
2790 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv10,_mm_mul_ps(velec,dsw)) ); |
2791 | cutoff_mask = _mm_cmplt_ps(rsq10,rcutoff2); |
2792 | |
2793 | fscal = felec; |
2794 | |
2795 | fscal = _mm_and_ps(fscal,cutoff_mask); |
2796 | |
2797 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
2798 | |
2799 | /* Calculate temporary vectorial force */ |
2800 | tx = _mm_mul_ps(fscal,dx10); |
2801 | ty = _mm_mul_ps(fscal,dy10); |
2802 | tz = _mm_mul_ps(fscal,dz10); |
2803 | |
2804 | /* Update vectorial force */ |
2805 | fix1 = _mm_add_ps(fix1,tx); |
2806 | fiy1 = _mm_add_ps(fiy1,ty); |
2807 | fiz1 = _mm_add_ps(fiz1,tz); |
2808 | |
2809 | fjx0 = _mm_add_ps(fjx0,tx); |
2810 | fjy0 = _mm_add_ps(fjy0,ty); |
2811 | fjz0 = _mm_add_ps(fjz0,tz); |
2812 | |
2813 | } |
2814 | |
2815 | /************************** |
2816 | * CALCULATE INTERACTIONS * |
2817 | **************************/ |
2818 | |
2819 | if (gmx_mm_any_lt(rsq11,rcutoff2)) |
2820 | { |
2821 | |
2822 | r11 = _mm_mul_ps(rsq11,rinv11); |
2823 | r11 = _mm_andnot_ps(dummy_mask,r11); |
2824 | |
2825 | /* EWALD ELECTROSTATICS */ |
2826 | |
2827 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
2828 | ewrt = _mm_mul_ps(r11,ewtabscale); |
2829 | ewitab = _mm_cvttps_epi32(ewrt); |
2830 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
2831 | ewitab = _mm_slli_epi32(ewitab,2); |
2832 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
2833 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
2834 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
2835 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
2836 | _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); |
2837 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
2838 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
2839 | velec = _mm_mul_ps(qq11,_mm_sub_ps(rinv11,velec)); |
2840 | felec = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec)); |
2841 | |
2842 | d = _mm_sub_ps(r11,rswitch); |
2843 | d = _mm_max_ps(d,_mm_setzero_ps()); |
2844 | d2 = _mm_mul_ps(d,d); |
2845 | 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))))))); |
2846 | |
2847 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
2848 | |
2849 | /* Evaluate switch function */ |
2850 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
2851 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv11,_mm_mul_ps(velec,dsw)) ); |
2852 | cutoff_mask = _mm_cmplt_ps(rsq11,rcutoff2); |
2853 | |
2854 | fscal = felec; |
2855 | |
2856 | fscal = _mm_and_ps(fscal,cutoff_mask); |
2857 | |
2858 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
2859 | |
2860 | /* Calculate temporary vectorial force */ |
2861 | tx = _mm_mul_ps(fscal,dx11); |
2862 | ty = _mm_mul_ps(fscal,dy11); |
2863 | tz = _mm_mul_ps(fscal,dz11); |
2864 | |
2865 | /* Update vectorial force */ |
2866 | fix1 = _mm_add_ps(fix1,tx); |
2867 | fiy1 = _mm_add_ps(fiy1,ty); |
2868 | fiz1 = _mm_add_ps(fiz1,tz); |
2869 | |
2870 | fjx1 = _mm_add_ps(fjx1,tx); |
2871 | fjy1 = _mm_add_ps(fjy1,ty); |
2872 | fjz1 = _mm_add_ps(fjz1,tz); |
2873 | |
2874 | } |
2875 | |
2876 | /************************** |
2877 | * CALCULATE INTERACTIONS * |
2878 | **************************/ |
2879 | |
2880 | if (gmx_mm_any_lt(rsq12,rcutoff2)) |
2881 | { |
2882 | |
2883 | r12 = _mm_mul_ps(rsq12,rinv12); |
2884 | r12 = _mm_andnot_ps(dummy_mask,r12); |
2885 | |
2886 | /* EWALD ELECTROSTATICS */ |
2887 | |
2888 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
2889 | ewrt = _mm_mul_ps(r12,ewtabscale); |
2890 | ewitab = _mm_cvttps_epi32(ewrt); |
2891 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
2892 | ewitab = _mm_slli_epi32(ewitab,2); |
2893 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
2894 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
2895 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
2896 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
2897 | _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); |
2898 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
2899 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
2900 | velec = _mm_mul_ps(qq12,_mm_sub_ps(rinv12,velec)); |
2901 | felec = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec)); |
2902 | |
2903 | d = _mm_sub_ps(r12,rswitch); |
2904 | d = _mm_max_ps(d,_mm_setzero_ps()); |
2905 | d2 = _mm_mul_ps(d,d); |
2906 | 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))))))); |
2907 | |
2908 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
2909 | |
2910 | /* Evaluate switch function */ |
2911 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
2912 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv12,_mm_mul_ps(velec,dsw)) ); |
2913 | cutoff_mask = _mm_cmplt_ps(rsq12,rcutoff2); |
2914 | |
2915 | fscal = felec; |
2916 | |
2917 | fscal = _mm_and_ps(fscal,cutoff_mask); |
2918 | |
2919 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
2920 | |
2921 | /* Calculate temporary vectorial force */ |
2922 | tx = _mm_mul_ps(fscal,dx12); |
2923 | ty = _mm_mul_ps(fscal,dy12); |
2924 | tz = _mm_mul_ps(fscal,dz12); |
2925 | |
2926 | /* Update vectorial force */ |
2927 | fix1 = _mm_add_ps(fix1,tx); |
2928 | fiy1 = _mm_add_ps(fiy1,ty); |
2929 | fiz1 = _mm_add_ps(fiz1,tz); |
2930 | |
2931 | fjx2 = _mm_add_ps(fjx2,tx); |
2932 | fjy2 = _mm_add_ps(fjy2,ty); |
2933 | fjz2 = _mm_add_ps(fjz2,tz); |
2934 | |
2935 | } |
2936 | |
2937 | /************************** |
2938 | * CALCULATE INTERACTIONS * |
2939 | **************************/ |
2940 | |
2941 | if (gmx_mm_any_lt(rsq20,rcutoff2)) |
2942 | { |
2943 | |
2944 | r20 = _mm_mul_ps(rsq20,rinv20); |
2945 | r20 = _mm_andnot_ps(dummy_mask,r20); |
2946 | |
2947 | /* EWALD ELECTROSTATICS */ |
2948 | |
2949 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
2950 | ewrt = _mm_mul_ps(r20,ewtabscale); |
2951 | ewitab = _mm_cvttps_epi32(ewrt); |
2952 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
2953 | ewitab = _mm_slli_epi32(ewitab,2); |
2954 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
2955 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
2956 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
2957 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
2958 | _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); |
2959 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
2960 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
2961 | velec = _mm_mul_ps(qq20,_mm_sub_ps(rinv20,velec)); |
2962 | felec = _mm_mul_ps(_mm_mul_ps(qq20,rinv20),_mm_sub_ps(rinvsq20,felec)); |
2963 | |
2964 | d = _mm_sub_ps(r20,rswitch); |
2965 | d = _mm_max_ps(d,_mm_setzero_ps()); |
2966 | d2 = _mm_mul_ps(d,d); |
2967 | 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))))))); |
2968 | |
2969 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
2970 | |
2971 | /* Evaluate switch function */ |
2972 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
2973 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv20,_mm_mul_ps(velec,dsw)) ); |
2974 | cutoff_mask = _mm_cmplt_ps(rsq20,rcutoff2); |
2975 | |
2976 | fscal = felec; |
2977 | |
2978 | fscal = _mm_and_ps(fscal,cutoff_mask); |
2979 | |
2980 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
2981 | |
2982 | /* Calculate temporary vectorial force */ |
2983 | tx = _mm_mul_ps(fscal,dx20); |
2984 | ty = _mm_mul_ps(fscal,dy20); |
2985 | tz = _mm_mul_ps(fscal,dz20); |
2986 | |
2987 | /* Update vectorial force */ |
2988 | fix2 = _mm_add_ps(fix2,tx); |
2989 | fiy2 = _mm_add_ps(fiy2,ty); |
2990 | fiz2 = _mm_add_ps(fiz2,tz); |
2991 | |
2992 | fjx0 = _mm_add_ps(fjx0,tx); |
2993 | fjy0 = _mm_add_ps(fjy0,ty); |
2994 | fjz0 = _mm_add_ps(fjz0,tz); |
2995 | |
2996 | } |
2997 | |
2998 | /************************** |
2999 | * CALCULATE INTERACTIONS * |
3000 | **************************/ |
3001 | |
3002 | if (gmx_mm_any_lt(rsq21,rcutoff2)) |
3003 | { |
3004 | |
3005 | r21 = _mm_mul_ps(rsq21,rinv21); |
3006 | r21 = _mm_andnot_ps(dummy_mask,r21); |
3007 | |
3008 | /* EWALD ELECTROSTATICS */ |
3009 | |
3010 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
3011 | ewrt = _mm_mul_ps(r21,ewtabscale); |
3012 | ewitab = _mm_cvttps_epi32(ewrt); |
3013 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
3014 | ewitab = _mm_slli_epi32(ewitab,2); |
3015 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
3016 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
3017 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
3018 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
3019 | _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); |
3020 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
3021 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
3022 | velec = _mm_mul_ps(qq21,_mm_sub_ps(rinv21,velec)); |
3023 | felec = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec)); |
3024 | |
3025 | d = _mm_sub_ps(r21,rswitch); |
3026 | d = _mm_max_ps(d,_mm_setzero_ps()); |
3027 | d2 = _mm_mul_ps(d,d); |
3028 | 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))))))); |
3029 | |
3030 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
3031 | |
3032 | /* Evaluate switch function */ |
3033 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
3034 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv21,_mm_mul_ps(velec,dsw)) ); |
3035 | cutoff_mask = _mm_cmplt_ps(rsq21,rcutoff2); |
3036 | |
3037 | fscal = felec; |
3038 | |
3039 | fscal = _mm_and_ps(fscal,cutoff_mask); |
3040 | |
3041 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
3042 | |
3043 | /* Calculate temporary vectorial force */ |
3044 | tx = _mm_mul_ps(fscal,dx21); |
3045 | ty = _mm_mul_ps(fscal,dy21); |
3046 | tz = _mm_mul_ps(fscal,dz21); |
3047 | |
3048 | /* Update vectorial force */ |
3049 | fix2 = _mm_add_ps(fix2,tx); |
3050 | fiy2 = _mm_add_ps(fiy2,ty); |
3051 | fiz2 = _mm_add_ps(fiz2,tz); |
3052 | |
3053 | fjx1 = _mm_add_ps(fjx1,tx); |
3054 | fjy1 = _mm_add_ps(fjy1,ty); |
3055 | fjz1 = _mm_add_ps(fjz1,tz); |
3056 | |
3057 | } |
3058 | |
3059 | /************************** |
3060 | * CALCULATE INTERACTIONS * |
3061 | **************************/ |
3062 | |
3063 | if (gmx_mm_any_lt(rsq22,rcutoff2)) |
3064 | { |
3065 | |
3066 | r22 = _mm_mul_ps(rsq22,rinv22); |
3067 | r22 = _mm_andnot_ps(dummy_mask,r22); |
3068 | |
3069 | /* EWALD ELECTROSTATICS */ |
3070 | |
3071 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
3072 | ewrt = _mm_mul_ps(r22,ewtabscale); |
3073 | ewitab = _mm_cvttps_epi32(ewrt); |
3074 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
3075 | ewitab = _mm_slli_epi32(ewitab,2); |
3076 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
3077 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
3078 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
3079 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
3080 | _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); |
3081 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
3082 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
3083 | velec = _mm_mul_ps(qq22,_mm_sub_ps(rinv22,velec)); |
3084 | felec = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec)); |
3085 | |
3086 | d = _mm_sub_ps(r22,rswitch); |
3087 | d = _mm_max_ps(d,_mm_setzero_ps()); |
3088 | d2 = _mm_mul_ps(d,d); |
3089 | 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))))))); |
3090 | |
3091 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
3092 | |
3093 | /* Evaluate switch function */ |
3094 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
3095 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv22,_mm_mul_ps(velec,dsw)) ); |
3096 | cutoff_mask = _mm_cmplt_ps(rsq22,rcutoff2); |
3097 | |
3098 | fscal = felec; |
3099 | |
3100 | fscal = _mm_and_ps(fscal,cutoff_mask); |
3101 | |
3102 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
3103 | |
3104 | /* Calculate temporary vectorial force */ |
3105 | tx = _mm_mul_ps(fscal,dx22); |
3106 | ty = _mm_mul_ps(fscal,dy22); |
3107 | tz = _mm_mul_ps(fscal,dz22); |
3108 | |
3109 | /* Update vectorial force */ |
3110 | fix2 = _mm_add_ps(fix2,tx); |
3111 | fiy2 = _mm_add_ps(fiy2,ty); |
3112 | fiz2 = _mm_add_ps(fiz2,tz); |
3113 | |
3114 | fjx2 = _mm_add_ps(fjx2,tx); |
3115 | fjy2 = _mm_add_ps(fjy2,ty); |
3116 | fjz2 = _mm_add_ps(fjz2,tz); |
3117 | |
3118 | } |
3119 | |
3120 | fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch; |
3121 | fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch; |
3122 | fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch; |
3123 | fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch; |
3124 | |
3125 | gmx_mm_decrement_3rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD, |
3126 | fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2); |
3127 | |
3128 | /* Inner loop uses 582 flops */ |
3129 | } |
3130 | |
3131 | /* End of innermost loop */ |
3132 | |
3133 | gmx_mm_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2, |
3134 | f+i_coord_offset,fshift+i_shift_offset); |
3135 | |
3136 | /* Increment number of inner iterations */ |
3137 | inneriter += j_index_end - j_index_start; |
3138 | |
3139 | /* Outer loop uses 18 flops */ |
3140 | } |
3141 | |
3142 | /* Increment number of outer iterations */ |
3143 | outeriter += nri; |
3144 | |
3145 | /* Update outer/inner flops */ |
3146 | |
3147 | inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_F,outeriter*18 + inneriter*582)(nrnb)->n[eNR_NBKERNEL_ELEC_VDW_W3W3_F] += outeriter*18 + inneriter *582; |
3148 | } |