?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Removed simple.h from nb_kernel_sse2_XX
[gromacs.git] / src / gromacs / gmxlib / nonbonded / nb_kernel_sse2_single / nb_kernel_ElecGB_VdwLJ_GeomP1P1_sse2_single.c
blob18a7e4a2538e0205c71cf033985cf24118695175
1 /*
2 * This file is part of the GROMACS molecular simulation package.
3 *
4 * Copyright (c) 2012,2013,2014,2015, 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 sse2_single kernel generator.
37 */
38 #include "gmxpre.h"
39
40 #include "config.h"
41
42 #include <math.h>
43
44 #include "../nb_kernel.h"
45 #include "gromacs/math/vec.h"
46 #include "gromacs/legacyheaders/nrnb.h"
47
48 #include "gromacs/simd/math_x86_sse2_single.h"
49 #include "kernelutil_x86_sse2_single.h"
50
51 /*
52 * Gromacs nonbonded kernel: nb_kernel_ElecGB_VdwLJ_GeomP1P1_VF_sse2_single
53 * Electrostatics interaction: GeneralizedBorn
54 * VdW interaction: LennardJones
55 * Geometry: Particle-Particle
56 * Calculate force/pot: PotentialAndForce
57 */
58 void
59 nb_kernel_ElecGB_VdwLJ_GeomP1P1_VF_sse2_single
60 (t_nblist * gmx_restrict nlist,
61 rvec * gmx_restrict xx,
62 rvec * gmx_restrict ff,
63 t_forcerec * gmx_restrict fr,
64 t_mdatoms * gmx_restrict mdatoms,
65 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
66 t_nrnb * gmx_restrict nrnb)
67 {
68 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
69 * just 0 for non-waters.
70 * Suffixes A,B,C,D refer to j loop unrolling done with SSE, e.g. for the four different
71 * jnr indices corresponding to data put in the four positions in the SIMD register.
72 */
73 int i_shift_offset,i_coord_offset,outeriter,inneriter;
74 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
75 int jnrA,jnrB,jnrC,jnrD;
76 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
77 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
78 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
79 real rcutoff_scalar;
80 real *shiftvec,*fshift,*x,*f;
81 real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
82 real scratch[4*DIM];
83 __m128 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
84 int vdwioffset0;
85 __m128 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
86 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
87 __m128 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
88 __m128 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
89 __m128 velec,felec,velecsum,facel,crf,krf,krf2;
90 real *charge;
91 __m128i gbitab;
92 __m128 vgb,fgb,vgbsum,dvdasum,gbscale,gbtabscale,isaprod,gbqqfactor,gbinvepsdiff,gbeps,dvdatmp;
93 __m128 minushalf = _mm_set1_ps(-0.5);
94 real *invsqrta,*dvda,*gbtab;
95 int nvdwtype;
96 __m128 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
97 int *vdwtype;
98 real *vdwparam;
99 __m128 one_sixth = _mm_set1_ps(1.0/6.0);
100 __m128 one_twelfth = _mm_set1_ps(1.0/12.0);
101 __m128i vfitab;
102 __m128i ifour = _mm_set1_epi32(4);
103 __m128 rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
104 real *vftab;
105 __m128 dummy_mask,cutoff_mask;
106 __m128 signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
107 __m128 one = _mm_set1_ps(1.0);
108 __m128 two = _mm_set1_ps(2.0);
109 x = xx[0];
110 f = ff[0];
111
112 nri = nlist->nri;
113 iinr = nlist->iinr;
114 jindex = nlist->jindex;
115 jjnr = nlist->jjnr;
116 shiftidx = nlist->shift;
117 gid = nlist->gid;
118 shiftvec = fr->shift_vec[0];
119 fshift = fr->fshift[0];
120 facel = _mm_set1_ps(fr->epsfac);
121 charge = mdatoms->chargeA;
122 nvdwtype = fr->ntype;
123 vdwparam = fr->nbfp;
124 vdwtype = mdatoms->typeA;
125
126 invsqrta = fr->invsqrta;
127 dvda = fr->dvda;
128 gbtabscale = _mm_set1_ps(fr->gbtab.scale);
129 gbtab = fr->gbtab.data;
130 gbinvepsdiff = _mm_set1_ps((1.0/fr->epsilon_r) - (1.0/fr->gb_epsilon_solvent));
131
132 /* Avoid stupid compiler warnings */
133 jnrA = jnrB = jnrC = jnrD = 0;
134 j_coord_offsetA = 0;
135 j_coord_offsetB = 0;
136 j_coord_offsetC = 0;
137 j_coord_offsetD = 0;
138
139 outeriter = 0;
140 inneriter = 0;
141
142 for(iidx=0;iidx<4*DIM;iidx++)
143 {
144 scratch[iidx] = 0.0;
145 }
146
147 /* Start outer loop over neighborlists */
148 for(iidx=0; iidx<nri; iidx++)
149 {
150 /* Load shift vector for this list */
151 i_shift_offset = DIM*shiftidx[iidx];
152
153 /* Load limits for loop over neighbors */
154 j_index_start = jindex[iidx];
155 j_index_end = jindex[iidx+1];
156
157 /* Get outer coordinate index */
158 inr = iinr[iidx];
159 i_coord_offset = DIM*inr;
160
161 /* Load i particle coords and add shift vector */
162 gmx_mm_load_shift_and_1rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
163
164 fix0 = _mm_setzero_ps();
165 fiy0 = _mm_setzero_ps();
166 fiz0 = _mm_setzero_ps();
167
168 /* Load parameters for i particles */
169 iq0 = _mm_mul_ps(facel,_mm_load1_ps(charge+inr+0));
170 isai0 = _mm_load1_ps(invsqrta+inr+0);
171 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
172
173 /* Reset potential sums */
174 velecsum = _mm_setzero_ps();
175 vgbsum = _mm_setzero_ps();
176 vvdwsum = _mm_setzero_ps();
177 dvdasum = _mm_setzero_ps();
178
179 /* Start inner kernel loop */
180 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
181 {
182
183 /* Get j neighbor index, and coordinate index */
184 jnrA = jjnr[jidx];
185 jnrB = jjnr[jidx+1];
186 jnrC = jjnr[jidx+2];
187 jnrD = jjnr[jidx+3];
188 j_coord_offsetA = DIM*jnrA;
189 j_coord_offsetB = DIM*jnrB;
190 j_coord_offsetC = DIM*jnrC;
191 j_coord_offsetD = DIM*jnrD;
192
193 /* load j atom coordinates */
194 gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
195 x+j_coord_offsetC,x+j_coord_offsetD,
196 &jx0,&jy0,&jz0);
197
198 /* Calculate displacement vector */
199 dx00 = _mm_sub_ps(ix0,jx0);
200 dy00 = _mm_sub_ps(iy0,jy0);
201 dz00 = _mm_sub_ps(iz0,jz0);
202
203 /* Calculate squared distance and things based on it */
204 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
205
206 rinv00 = gmx_mm_invsqrt_ps(rsq00);
207
208 rinvsq00 = _mm_mul_ps(rinv00,rinv00);
209
210 /* Load parameters for j particles */
211 jq0 = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
212 charge+jnrC+0,charge+jnrD+0);
213 isaj0 = gmx_mm_load_4real_swizzle_ps(invsqrta+jnrA+0,invsqrta+jnrB+0,
214 invsqrta+jnrC+0,invsqrta+jnrD+0);
215 vdwjidx0A = 2*vdwtype[jnrA+0];
216 vdwjidx0B = 2*vdwtype[jnrB+0];
217 vdwjidx0C = 2*vdwtype[jnrC+0];
218 vdwjidx0D = 2*vdwtype[jnrD+0];
219
220 /**************************
221 * CALCULATE INTERACTIONS *
222 **************************/
223
224 r00 = _mm_mul_ps(rsq00,rinv00);
225
226 /* Compute parameters for interactions between i and j atoms */
227 qq00 = _mm_mul_ps(iq0,jq0);
228 gmx_mm_load_4pair_swizzle_ps(vdwparam+vdwioffset0+vdwjidx0A,
229 vdwparam+vdwioffset0+vdwjidx0B,
230 vdwparam+vdwioffset0+vdwjidx0C,
231 vdwparam+vdwioffset0+vdwjidx0D,
232 &c6_00,&c12_00);
233
234 /* GENERALIZED BORN AND COULOMB ELECTROSTATICS */
235 isaprod = _mm_mul_ps(isai0,isaj0);
236 gbqqfactor = _mm_xor_ps(signbit,_mm_mul_ps(qq00,_mm_mul_ps(isaprod,gbinvepsdiff)));
237 gbscale = _mm_mul_ps(isaprod,gbtabscale);
238
239 /* Calculate generalized born table index - this is a separate table from the normal one,
240 * but we use the same procedure by multiplying r with scale and truncating to integer.
241 */
242 rt = _mm_mul_ps(r00,gbscale);
243 gbitab = _mm_cvttps_epi32(rt);
244 gbeps = _mm_sub_ps(rt,_mm_cvtepi32_ps(gbitab));
245 gbitab = _mm_slli_epi32(gbitab,2);
246
247 Y = _mm_load_ps( gbtab + gmx_mm_extract_epi32(gbitab,0) );
248 F = _mm_load_ps( gbtab + gmx_mm_extract_epi32(gbitab,1) );
249 G = _mm_load_ps( gbtab + gmx_mm_extract_epi32(gbitab,2) );
250 H = _mm_load_ps( gbtab + gmx_mm_extract_epi32(gbitab,3) );
251 _MM_TRANSPOSE4_PS(Y,F,G,H);
252 Heps = _mm_mul_ps(gbeps,H);
253 Fp = _mm_add_ps(F,_mm_mul_ps(gbeps,_mm_add_ps(G,Heps)));
254 VV = _mm_add_ps(Y,_mm_mul_ps(gbeps,Fp));
255 vgb = _mm_mul_ps(gbqqfactor,VV);
256
257 FF = _mm_add_ps(Fp,_mm_mul_ps(gbeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
258 fgb = _mm_mul_ps(gbqqfactor,_mm_mul_ps(FF,gbscale));
259 dvdatmp = _mm_mul_ps(minushalf,_mm_add_ps(vgb,_mm_mul_ps(fgb,r00)));
260 dvdasum = _mm_add_ps(dvdasum,dvdatmp);
261 fjptrA = dvda+jnrA;
262 fjptrB = dvda+jnrB;
263 fjptrC = dvda+jnrC;
264 fjptrD = dvda+jnrD;
265 gmx_mm_increment_4real_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,_mm_mul_ps(dvdatmp,_mm_mul_ps(isaj0,isaj0)));
266 velec = _mm_mul_ps(qq00,rinv00);
267 felec = _mm_mul_ps(_mm_sub_ps(_mm_mul_ps(velec,rinv00),fgb),rinv00);
268
269 /* LENNARD-JONES DISPERSION/REPULSION */
270
271 rinvsix = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00);
272 vvdw6 = _mm_mul_ps(c6_00,rinvsix);
273 vvdw12 = _mm_mul_ps(c12_00,_mm_mul_ps(rinvsix,rinvsix));
274 vvdw = _mm_sub_ps( _mm_mul_ps(vvdw12,one_twelfth) , _mm_mul_ps(vvdw6,one_sixth) );
275 fvdw = _mm_mul_ps(_mm_sub_ps(vvdw12,vvdw6),rinvsq00);
276
277 /* Update potential sum for this i atom from the interaction with this j atom. */
278 velecsum = _mm_add_ps(velecsum,velec);
279 vgbsum = _mm_add_ps(vgbsum,vgb);
280 vvdwsum = _mm_add_ps(vvdwsum,vvdw);
281
282 fscal = _mm_add_ps(felec,fvdw);
283
284 /* Calculate temporary vectorial force */
285 tx = _mm_mul_ps(fscal,dx00);
286 ty = _mm_mul_ps(fscal,dy00);
287 tz = _mm_mul_ps(fscal,dz00);
288
289 /* Update vectorial force */
290 fix0 = _mm_add_ps(fix0,tx);
291 fiy0 = _mm_add_ps(fiy0,ty);
292 fiz0 = _mm_add_ps(fiz0,tz);
293
294 fjptrA = f+j_coord_offsetA;
295 fjptrB = f+j_coord_offsetB;
296 fjptrC = f+j_coord_offsetC;
297 fjptrD = f+j_coord_offsetD;
298 gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,tx,ty,tz);
299
300 /* Inner loop uses 71 flops */
301 }
302
303 if(jidx<j_index_end)
304 {
305
306 /* Get j neighbor index, and coordinate index */
307 jnrlistA = jjnr[jidx];
308 jnrlistB = jjnr[jidx+1];
309 jnrlistC = jjnr[jidx+2];
310 jnrlistD = jjnr[jidx+3];
311 /* Sign of each element will be negative for non-real atoms.
312 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
313 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
314 */
315 dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
316 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
317 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
318 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
319 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
320 j_coord_offsetA = DIM*jnrA;
321 j_coord_offsetB = DIM*jnrB;
322 j_coord_offsetC = DIM*jnrC;
323 j_coord_offsetD = DIM*jnrD;
324
325 /* load j atom coordinates */
326 gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
327 x+j_coord_offsetC,x+j_coord_offsetD,
328 &jx0,&jy0,&jz0);
329
330 /* Calculate displacement vector */
331 dx00 = _mm_sub_ps(ix0,jx0);
332 dy00 = _mm_sub_ps(iy0,jy0);
333 dz00 = _mm_sub_ps(iz0,jz0);
334
335 /* Calculate squared distance and things based on it */
336 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
337
338 rinv00 = gmx_mm_invsqrt_ps(rsq00);
339
340 rinvsq00 = _mm_mul_ps(rinv00,rinv00);
341
342 /* Load parameters for j particles */
343 jq0 = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
344 charge+jnrC+0,charge+jnrD+0);
345 isaj0 = gmx_mm_load_4real_swizzle_ps(invsqrta+jnrA+0,invsqrta+jnrB+0,
346 invsqrta+jnrC+0,invsqrta+jnrD+0);
347 vdwjidx0A = 2*vdwtype[jnrA+0];
348 vdwjidx0B = 2*vdwtype[jnrB+0];
349 vdwjidx0C = 2*vdwtype[jnrC+0];
350 vdwjidx0D = 2*vdwtype[jnrD+0];
351
352 /**************************
353 * CALCULATE INTERACTIONS *
354 **************************/
355
356 r00 = _mm_mul_ps(rsq00,rinv00);
357 r00 = _mm_andnot_ps(dummy_mask,r00);
358
359 /* Compute parameters for interactions between i and j atoms */
360 qq00 = _mm_mul_ps(iq0,jq0);
361 gmx_mm_load_4pair_swizzle_ps(vdwparam+vdwioffset0+vdwjidx0A,
362 vdwparam+vdwioffset0+vdwjidx0B,
363 vdwparam+vdwioffset0+vdwjidx0C,
364 vdwparam+vdwioffset0+vdwjidx0D,
365 &c6_00,&c12_00);
366
367 /* GENERALIZED BORN AND COULOMB ELECTROSTATICS */
368 isaprod = _mm_mul_ps(isai0,isaj0);
369 gbqqfactor = _mm_xor_ps(signbit,_mm_mul_ps(qq00,_mm_mul_ps(isaprod,gbinvepsdiff)));
370 gbscale = _mm_mul_ps(isaprod,gbtabscale);
371
372 /* Calculate generalized born table index - this is a separate table from the normal one,
373 * but we use the same procedure by multiplying r with scale and truncating to integer.
374 */
375 rt = _mm_mul_ps(r00,gbscale);
376 gbitab = _mm_cvttps_epi32(rt);
377 gbeps = _mm_sub_ps(rt,_mm_cvtepi32_ps(gbitab));
378 gbitab = _mm_slli_epi32(gbitab,2);
379
380 Y = _mm_load_ps( gbtab + gmx_mm_extract_epi32(gbitab,0) );
381 F = _mm_load_ps( gbtab + gmx_mm_extract_epi32(gbitab,1) );
382 G = _mm_load_ps( gbtab + gmx_mm_extract_epi32(gbitab,2) );
383 H = _mm_load_ps( gbtab + gmx_mm_extract_epi32(gbitab,3) );
384 _MM_TRANSPOSE4_PS(Y,F,G,H);
385 Heps = _mm_mul_ps(gbeps,H);
386 Fp = _mm_add_ps(F,_mm_mul_ps(gbeps,_mm_add_ps(G,Heps)));
387 VV = _mm_add_ps(Y,_mm_mul_ps(gbeps,Fp));
388 vgb = _mm_mul_ps(gbqqfactor,VV);
389
390 FF = _mm_add_ps(Fp,_mm_mul_ps(gbeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
391 fgb = _mm_mul_ps(gbqqfactor,_mm_mul_ps(FF,gbscale));
392 dvdatmp = _mm_mul_ps(minushalf,_mm_add_ps(vgb,_mm_mul_ps(fgb,r00)));
393 dvdatmp = _mm_andnot_ps(dummy_mask,dvdatmp);
394 dvdasum = _mm_add_ps(dvdasum,dvdatmp);
395 /* The pointers to scratch make sure that this code with compilers that take gmx_restrict seriously (e.g. icc 13) really can't screw things up. */
396 fjptrA = (jnrlistA>=0) ? dvda+jnrA : scratch;
397 fjptrB = (jnrlistB>=0) ? dvda+jnrB : scratch;
398 fjptrC = (jnrlistC>=0) ? dvda+jnrC : scratch;
399 fjptrD = (jnrlistD>=0) ? dvda+jnrD : scratch;
400 gmx_mm_increment_4real_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,_mm_mul_ps(dvdatmp,_mm_mul_ps(isaj0,isaj0)));
401 velec = _mm_mul_ps(qq00,rinv00);
402 felec = _mm_mul_ps(_mm_sub_ps(_mm_mul_ps(velec,rinv00),fgb),rinv00);
403
404 /* LENNARD-JONES DISPERSION/REPULSION */
405
406 rinvsix = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00);
407 vvdw6 = _mm_mul_ps(c6_00,rinvsix);
408 vvdw12 = _mm_mul_ps(c12_00,_mm_mul_ps(rinvsix,rinvsix));
409 vvdw = _mm_sub_ps( _mm_mul_ps(vvdw12,one_twelfth) , _mm_mul_ps(vvdw6,one_sixth) );
410 fvdw = _mm_mul_ps(_mm_sub_ps(vvdw12,vvdw6),rinvsq00);
411
412 /* Update potential sum for this i atom from the interaction with this j atom. */
413 velec = _mm_andnot_ps(dummy_mask,velec);
414 velecsum = _mm_add_ps(velecsum,velec);
415 vgb = _mm_andnot_ps(dummy_mask,vgb);
416 vgbsum = _mm_add_ps(vgbsum,vgb);
417 vvdw = _mm_andnot_ps(dummy_mask,vvdw);
418 vvdwsum = _mm_add_ps(vvdwsum,vvdw);
419
420 fscal = _mm_add_ps(felec,fvdw);
421
422 fscal = _mm_andnot_ps(dummy_mask,fscal);
423
424 /* Calculate temporary vectorial force */
425 tx = _mm_mul_ps(fscal,dx00);
426 ty = _mm_mul_ps(fscal,dy00);
427 tz = _mm_mul_ps(fscal,dz00);
428
429 /* Update vectorial force */
430 fix0 = _mm_add_ps(fix0,tx);
431 fiy0 = _mm_add_ps(fiy0,ty);
432 fiz0 = _mm_add_ps(fiz0,tz);
433
434 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
435 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
436 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
437 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
438 gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,tx,ty,tz);
439
440 /* Inner loop uses 72 flops */
441 }
442
443 /* End of innermost loop */
444
445 gmx_mm_update_iforce_1atom_swizzle_ps(fix0,fiy0,fiz0,
446 f+i_coord_offset,fshift+i_shift_offset);
447
448 ggid = gid[iidx];
449 /* Update potential energies */
450 gmx_mm_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
451 gmx_mm_update_1pot_ps(vgbsum,kernel_data->energygrp_polarization+ggid);
452 gmx_mm_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
453 dvdasum = _mm_mul_ps(dvdasum, _mm_mul_ps(isai0,isai0));
454 gmx_mm_update_1pot_ps(dvdasum,dvda+inr);
455
456 /* Increment number of inner iterations */
457 inneriter += j_index_end - j_index_start;
458
459 /* Outer loop uses 10 flops */
460 }
461
462 /* Increment number of outer iterations */
463 outeriter += nri;
464
465 /* Update outer/inner flops */
466
467 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_VF,outeriter*10 + inneriter*72);
468 }
469 /*
470 * Gromacs nonbonded kernel: nb_kernel_ElecGB_VdwLJ_GeomP1P1_F_sse2_single
471 * Electrostatics interaction: GeneralizedBorn
472 * VdW interaction: LennardJones
473 * Geometry: Particle-Particle
474 * Calculate force/pot: Force
475 */
476 void
477 nb_kernel_ElecGB_VdwLJ_GeomP1P1_F_sse2_single
478 (t_nblist * gmx_restrict nlist,
479 rvec * gmx_restrict xx,
480 rvec * gmx_restrict ff,
481 t_forcerec * gmx_restrict fr,
482 t_mdatoms * gmx_restrict mdatoms,
483 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
484 t_nrnb * gmx_restrict nrnb)
485 {
486 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
487 * just 0 for non-waters.
488 * Suffixes A,B,C,D refer to j loop unrolling done with SSE, e.g. for the four different
489 * jnr indices corresponding to data put in the four positions in the SIMD register.
490 */
491 int i_shift_offset,i_coord_offset,outeriter,inneriter;
492 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
493 int jnrA,jnrB,jnrC,jnrD;
494 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
495 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
496 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
497 real rcutoff_scalar;
498 real *shiftvec,*fshift,*x,*f;
499 real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
500 real scratch[4*DIM];
501 __m128 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
502 int vdwioffset0;
503 __m128 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
504 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
505 __m128 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
506 __m128 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
507 __m128 velec,felec,velecsum,facel,crf,krf,krf2;
508 real *charge;
509 __m128i gbitab;
510 __m128 vgb,fgb,vgbsum,dvdasum,gbscale,gbtabscale,isaprod,gbqqfactor,gbinvepsdiff,gbeps,dvdatmp;
511 __m128 minushalf = _mm_set1_ps(-0.5);
512 real *invsqrta,*dvda,*gbtab;
513 int nvdwtype;
514 __m128 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
515 int *vdwtype;
516 real *vdwparam;
517 __m128 one_sixth = _mm_set1_ps(1.0/6.0);
518 __m128 one_twelfth = _mm_set1_ps(1.0/12.0);
519 __m128i vfitab;
520 __m128i ifour = _mm_set1_epi32(4);
521 __m128 rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
522 real *vftab;
523 __m128 dummy_mask,cutoff_mask;
524 __m128 signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
525 __m128 one = _mm_set1_ps(1.0);
526 __m128 two = _mm_set1_ps(2.0);
527 x = xx[0];
528 f = ff[0];
529
530 nri = nlist->nri;
531 iinr = nlist->iinr;
532 jindex = nlist->jindex;
533 jjnr = nlist->jjnr;
534 shiftidx = nlist->shift;
535 gid = nlist->gid;
536 shiftvec = fr->shift_vec[0];
537 fshift = fr->fshift[0];
538 facel = _mm_set1_ps(fr->epsfac);
539 charge = mdatoms->chargeA;
540 nvdwtype = fr->ntype;
541 vdwparam = fr->nbfp;
542 vdwtype = mdatoms->typeA;
543
544 invsqrta = fr->invsqrta;
545 dvda = fr->dvda;
546 gbtabscale = _mm_set1_ps(fr->gbtab.scale);
547 gbtab = fr->gbtab.data;
548 gbinvepsdiff = _mm_set1_ps((1.0/fr->epsilon_r) - (1.0/fr->gb_epsilon_solvent));
549
550 /* Avoid stupid compiler warnings */
551 jnrA = jnrB = jnrC = jnrD = 0;
552 j_coord_offsetA = 0;
553 j_coord_offsetB = 0;
554 j_coord_offsetC = 0;
555 j_coord_offsetD = 0;
556
557 outeriter = 0;
558 inneriter = 0;
559
560 for(iidx=0;iidx<4*DIM;iidx++)
561 {
562 scratch[iidx] = 0.0;
563 }
564
565 /* Start outer loop over neighborlists */
566 for(iidx=0; iidx<nri; iidx++)
567 {
568 /* Load shift vector for this list */
569 i_shift_offset = DIM*shiftidx[iidx];
570
571 /* Load limits for loop over neighbors */
572 j_index_start = jindex[iidx];
573 j_index_end = jindex[iidx+1];
574
575 /* Get outer coordinate index */
576 inr = iinr[iidx];
577 i_coord_offset = DIM*inr;
578
579 /* Load i particle coords and add shift vector */
580 gmx_mm_load_shift_and_1rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
581
582 fix0 = _mm_setzero_ps();
583 fiy0 = _mm_setzero_ps();
584 fiz0 = _mm_setzero_ps();
585
586 /* Load parameters for i particles */
587 iq0 = _mm_mul_ps(facel,_mm_load1_ps(charge+inr+0));
588 isai0 = _mm_load1_ps(invsqrta+inr+0);
589 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
590
591 dvdasum = _mm_setzero_ps();
592
593 /* Start inner kernel loop */
594 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
595 {
596
597 /* Get j neighbor index, and coordinate index */
598 jnrA = jjnr[jidx];
599 jnrB = jjnr[jidx+1];
600 jnrC = jjnr[jidx+2];
601 jnrD = jjnr[jidx+3];
602 j_coord_offsetA = DIM*jnrA;
603 j_coord_offsetB = DIM*jnrB;
604 j_coord_offsetC = DIM*jnrC;
605 j_coord_offsetD = DIM*jnrD;
606
607 /* load j atom coordinates */
608 gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
609 x+j_coord_offsetC,x+j_coord_offsetD,
610 &jx0,&jy0,&jz0);
611
612 /* Calculate displacement vector */
613 dx00 = _mm_sub_ps(ix0,jx0);
614 dy00 = _mm_sub_ps(iy0,jy0);
615 dz00 = _mm_sub_ps(iz0,jz0);
616
617 /* Calculate squared distance and things based on it */
618 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
619
620 rinv00 = gmx_mm_invsqrt_ps(rsq00);
621
622 rinvsq00 = _mm_mul_ps(rinv00,rinv00);
623
624 /* Load parameters for j particles */
625 jq0 = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
626 charge+jnrC+0,charge+jnrD+0);
627 isaj0 = gmx_mm_load_4real_swizzle_ps(invsqrta+jnrA+0,invsqrta+jnrB+0,
628 invsqrta+jnrC+0,invsqrta+jnrD+0);
629 vdwjidx0A = 2*vdwtype[jnrA+0];
630 vdwjidx0B = 2*vdwtype[jnrB+0];
631 vdwjidx0C = 2*vdwtype[jnrC+0];
632 vdwjidx0D = 2*vdwtype[jnrD+0];
633
634 /**************************
635 * CALCULATE INTERACTIONS *
636 **************************/
637
638 r00 = _mm_mul_ps(rsq00,rinv00);
639
640 /* Compute parameters for interactions between i and j atoms */
641 qq00 = _mm_mul_ps(iq0,jq0);
642 gmx_mm_load_4pair_swizzle_ps(vdwparam+vdwioffset0+vdwjidx0A,
643 vdwparam+vdwioffset0+vdwjidx0B,
644 vdwparam+vdwioffset0+vdwjidx0C,
645 vdwparam+vdwioffset0+vdwjidx0D,
646 &c6_00,&c12_00);
647
648 /* GENERALIZED BORN AND COULOMB ELECTROSTATICS */
649 isaprod = _mm_mul_ps(isai0,isaj0);
650 gbqqfactor = _mm_xor_ps(signbit,_mm_mul_ps(qq00,_mm_mul_ps(isaprod,gbinvepsdiff)));
651 gbscale = _mm_mul_ps(isaprod,gbtabscale);
652
653 /* Calculate generalized born table index - this is a separate table from the normal one,
654 * but we use the same procedure by multiplying r with scale and truncating to integer.
655 */
656 rt = _mm_mul_ps(r00,gbscale);
657 gbitab = _mm_cvttps_epi32(rt);
658 gbeps = _mm_sub_ps(rt,_mm_cvtepi32_ps(gbitab));
659 gbitab = _mm_slli_epi32(gbitab,2);
660
661 Y = _mm_load_ps( gbtab + gmx_mm_extract_epi32(gbitab,0) );
662 F = _mm_load_ps( gbtab + gmx_mm_extract_epi32(gbitab,1) );
663 G = _mm_load_ps( gbtab + gmx_mm_extract_epi32(gbitab,2) );
664 H = _mm_load_ps( gbtab + gmx_mm_extract_epi32(gbitab,3) );
665 _MM_TRANSPOSE4_PS(Y,F,G,H);
666 Heps = _mm_mul_ps(gbeps,H);
667 Fp = _mm_add_ps(F,_mm_mul_ps(gbeps,_mm_add_ps(G,Heps)));
668 VV = _mm_add_ps(Y,_mm_mul_ps(gbeps,Fp));
669 vgb = _mm_mul_ps(gbqqfactor,VV);
670
671 FF = _mm_add_ps(Fp,_mm_mul_ps(gbeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
672 fgb = _mm_mul_ps(gbqqfactor,_mm_mul_ps(FF,gbscale));
673 dvdatmp = _mm_mul_ps(minushalf,_mm_add_ps(vgb,_mm_mul_ps(fgb,r00)));
674 dvdasum = _mm_add_ps(dvdasum,dvdatmp);
675 fjptrA = dvda+jnrA;
676 fjptrB = dvda+jnrB;
677 fjptrC = dvda+jnrC;
678 fjptrD = dvda+jnrD;
679 gmx_mm_increment_4real_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,_mm_mul_ps(dvdatmp,_mm_mul_ps(isaj0,isaj0)));
680 velec = _mm_mul_ps(qq00,rinv00);
681 felec = _mm_mul_ps(_mm_sub_ps(_mm_mul_ps(velec,rinv00),fgb),rinv00);
682
683 /* LENNARD-JONES DISPERSION/REPULSION */
684
685 rinvsix = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00);
686 fvdw = _mm_mul_ps(_mm_sub_ps(_mm_mul_ps(c12_00,rinvsix),c6_00),_mm_mul_ps(rinvsix,rinvsq00));
687
688 fscal = _mm_add_ps(felec,fvdw);
689
690 /* Calculate temporary vectorial force */
691 tx = _mm_mul_ps(fscal,dx00);
692 ty = _mm_mul_ps(fscal,dy00);
693 tz = _mm_mul_ps(fscal,dz00);
694
695 /* Update vectorial force */
696 fix0 = _mm_add_ps(fix0,tx);
697 fiy0 = _mm_add_ps(fiy0,ty);
698 fiz0 = _mm_add_ps(fiz0,tz);
699
700 fjptrA = f+j_coord_offsetA;
701 fjptrB = f+j_coord_offsetB;
702 fjptrC = f+j_coord_offsetC;
703 fjptrD = f+j_coord_offsetD;
704 gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,tx,ty,tz);
705
706 /* Inner loop uses 64 flops */
707 }
708
709 if(jidx<j_index_end)
710 {
711
712 /* Get j neighbor index, and coordinate index */
713 jnrlistA = jjnr[jidx];
714 jnrlistB = jjnr[jidx+1];
715 jnrlistC = jjnr[jidx+2];
716 jnrlistD = jjnr[jidx+3];
717 /* Sign of each element will be negative for non-real atoms.
718 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
719 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
720 */
721 dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
722 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
723 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
724 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
725 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
726 j_coord_offsetA = DIM*jnrA;
727 j_coord_offsetB = DIM*jnrB;
728 j_coord_offsetC = DIM*jnrC;
729 j_coord_offsetD = DIM*jnrD;
730
731 /* load j atom coordinates */
732 gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
733 x+j_coord_offsetC,x+j_coord_offsetD,
734 &jx0,&jy0,&jz0);
735
736 /* Calculate displacement vector */
737 dx00 = _mm_sub_ps(ix0,jx0);
738 dy00 = _mm_sub_ps(iy0,jy0);
739 dz00 = _mm_sub_ps(iz0,jz0);
740
741 /* Calculate squared distance and things based on it */
742 rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
743
744 rinv00 = gmx_mm_invsqrt_ps(rsq00);
745
746 rinvsq00 = _mm_mul_ps(rinv00,rinv00);
747
748 /* Load parameters for j particles */
749 jq0 = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
750 charge+jnrC+0,charge+jnrD+0);
751 isaj0 = gmx_mm_load_4real_swizzle_ps(invsqrta+jnrA+0,invsqrta+jnrB+0,
752 invsqrta+jnrC+0,invsqrta+jnrD+0);
753 vdwjidx0A = 2*vdwtype[jnrA+0];
754 vdwjidx0B = 2*vdwtype[jnrB+0];
755 vdwjidx0C = 2*vdwtype[jnrC+0];
756 vdwjidx0D = 2*vdwtype[jnrD+0];
757
758 /**************************
759 * CALCULATE INTERACTIONS *
760 **************************/
761
762 r00 = _mm_mul_ps(rsq00,rinv00);
763 r00 = _mm_andnot_ps(dummy_mask,r00);
764
765 /* Compute parameters for interactions between i and j atoms */
766 qq00 = _mm_mul_ps(iq0,jq0);
767 gmx_mm_load_4pair_swizzle_ps(vdwparam+vdwioffset0+vdwjidx0A,
768 vdwparam+vdwioffset0+vdwjidx0B,
769 vdwparam+vdwioffset0+vdwjidx0C,
770 vdwparam+vdwioffset0+vdwjidx0D,
771 &c6_00,&c12_00);
772
773 /* GENERALIZED BORN AND COULOMB ELECTROSTATICS */
774 isaprod = _mm_mul_ps(isai0,isaj0);
775 gbqqfactor = _mm_xor_ps(signbit,_mm_mul_ps(qq00,_mm_mul_ps(isaprod,gbinvepsdiff)));
776 gbscale = _mm_mul_ps(isaprod,gbtabscale);
777
778 /* Calculate generalized born table index - this is a separate table from the normal one,
779 * but we use the same procedure by multiplying r with scale and truncating to integer.
780 */
781 rt = _mm_mul_ps(r00,gbscale);
782 gbitab = _mm_cvttps_epi32(rt);
783 gbeps = _mm_sub_ps(rt,_mm_cvtepi32_ps(gbitab));
784 gbitab = _mm_slli_epi32(gbitab,2);
785
786 Y = _mm_load_ps( gbtab + gmx_mm_extract_epi32(gbitab,0) );
787 F = _mm_load_ps( gbtab + gmx_mm_extract_epi32(gbitab,1) );
788 G = _mm_load_ps( gbtab + gmx_mm_extract_epi32(gbitab,2) );
789 H = _mm_load_ps( gbtab + gmx_mm_extract_epi32(gbitab,3) );
790 _MM_TRANSPOSE4_PS(Y,F,G,H);
791 Heps = _mm_mul_ps(gbeps,H);
792 Fp = _mm_add_ps(F,_mm_mul_ps(gbeps,_mm_add_ps(G,Heps)));
793 VV = _mm_add_ps(Y,_mm_mul_ps(gbeps,Fp));
794 vgb = _mm_mul_ps(gbqqfactor,VV);
795
796 FF = _mm_add_ps(Fp,_mm_mul_ps(gbeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
797 fgb = _mm_mul_ps(gbqqfactor,_mm_mul_ps(FF,gbscale));
798 dvdatmp = _mm_mul_ps(minushalf,_mm_add_ps(vgb,_mm_mul_ps(fgb,r00)));
799 dvdatmp = _mm_andnot_ps(dummy_mask,dvdatmp);
800 dvdasum = _mm_add_ps(dvdasum,dvdatmp);
801 /* The pointers to scratch make sure that this code with compilers that take gmx_restrict seriously (e.g. icc 13) really can't screw things up. */
802 fjptrA = (jnrlistA>=0) ? dvda+jnrA : scratch;
803 fjptrB = (jnrlistB>=0) ? dvda+jnrB : scratch;
804 fjptrC = (jnrlistC>=0) ? dvda+jnrC : scratch;
805 fjptrD = (jnrlistD>=0) ? dvda+jnrD : scratch;
806 gmx_mm_increment_4real_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,_mm_mul_ps(dvdatmp,_mm_mul_ps(isaj0,isaj0)));
807 velec = _mm_mul_ps(qq00,rinv00);
808 felec = _mm_mul_ps(_mm_sub_ps(_mm_mul_ps(velec,rinv00),fgb),rinv00);
809
810 /* LENNARD-JONES DISPERSION/REPULSION */
811
812 rinvsix = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00);
813 fvdw = _mm_mul_ps(_mm_sub_ps(_mm_mul_ps(c12_00,rinvsix),c6_00),_mm_mul_ps(rinvsix,rinvsq00));
814
815 fscal = _mm_add_ps(felec,fvdw);
816
817 fscal = _mm_andnot_ps(dummy_mask,fscal);
818
819 /* Calculate temporary vectorial force */
820 tx = _mm_mul_ps(fscal,dx00);
821 ty = _mm_mul_ps(fscal,dy00);
822 tz = _mm_mul_ps(fscal,dz00);
823
824 /* Update vectorial force */
825 fix0 = _mm_add_ps(fix0,tx);
826 fiy0 = _mm_add_ps(fiy0,ty);
827 fiz0 = _mm_add_ps(fiz0,tz);
828
829 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
830 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
831 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
832 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
833 gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,tx,ty,tz);
834
835 /* Inner loop uses 65 flops */
836 }
837
838 /* End of innermost loop */
839
840 gmx_mm_update_iforce_1atom_swizzle_ps(fix0,fiy0,fiz0,
841 f+i_coord_offset,fshift+i_shift_offset);
842
843 dvdasum = _mm_mul_ps(dvdasum, _mm_mul_ps(isai0,isai0));
844 gmx_mm_update_1pot_ps(dvdasum,dvda+inr);
845
846 /* Increment number of inner iterations */
847 inneriter += j_index_end - j_index_start;
848
849 /* Outer loop uses 7 flops */
850 }
851
852 /* Increment number of outer iterations */
853 outeriter += nri;
854
855 /* Update outer/inner flops */
856
857 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_F,outeriter*7 + inneriter*65);
858 }
The ultimate molecular dynamics simulation package