| File: | gromacs/gmxlib/nonbonded/nb_kernel_sse4_1_single/nb_kernel_ElecEwSw_VdwLJSw_GeomW4P1_sse4_1_single.c |
| Location: | line 1033, 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_GeomW4P1_VF_sse4_1_single |
| 54 | * Electrostatics interaction: Ewald |
| 55 | * VdW interaction: LennardJones |
| 56 | * Geometry: Water4-Particle |
| 57 | * Calculate force/pot: PotentialAndForce |
| 58 | */ |
| 59 | void |
| 60 | nb_kernel_ElecEwSw_VdwLJSw_GeomW4P1_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 vdwioffset3; |
| 92 | __m128 ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3; |
| 93 | int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D; |
| 94 | __m128 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0; |
| 95 | __m128 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00; |
| 96 | __m128 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10; |
| 97 | __m128 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20; |
| 98 | __m128 dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30; |
| 99 | __m128 velec,felec,velecsum,facel,crf,krf,krf2; |
| 100 | real *charge; |
| 101 | int nvdwtype; |
| 102 | __m128 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6; |
| 103 | int *vdwtype; |
| 104 | real *vdwparam; |
| 105 | __m128 one_sixth = _mm_set1_ps(1.0/6.0); |
| 106 | __m128 one_twelfth = _mm_set1_ps(1.0/12.0); |
| 107 | __m128i ewitab; |
| 108 | __m128 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV; |
| 109 | real *ewtab; |
| 110 | __m128 rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw; |
| 111 | real rswitch_scalar,d_scalar; |
| 112 | __m128 dummy_mask,cutoff_mask; |
| 113 | __m128 signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) ); |
| 114 | __m128 one = _mm_set1_ps(1.0); |
| 115 | __m128 two = _mm_set1_ps(2.0); |
| 116 | x = xx[0]; |
| 117 | f = ff[0]; |
| 118 | |
| 119 | nri = nlist->nri; |
| 120 | iinr = nlist->iinr; |
| 121 | jindex = nlist->jindex; |
| 122 | jjnr = nlist->jjnr; |
| 123 | shiftidx = nlist->shift; |
| 124 | gid = nlist->gid; |
| 125 | shiftvec = fr->shift_vec[0]; |
| 126 | fshift = fr->fshift[0]; |
| 127 | facel = _mm_set1_ps(fr->epsfac); |
| 128 | charge = mdatoms->chargeA; |
| 129 | nvdwtype = fr->ntype; |
| 130 | vdwparam = fr->nbfp; |
| 131 | vdwtype = mdatoms->typeA; |
| 132 | |
| 133 | sh_ewald = _mm_set1_ps(fr->ic->sh_ewald); |
| 134 | ewtab = fr->ic->tabq_coul_FDV0; |
| 135 | ewtabscale = _mm_set1_ps(fr->ic->tabq_scale); |
| 136 | ewtabhalfspace = _mm_set1_ps(0.5/fr->ic->tabq_scale); |
| 137 | |
| 138 | /* Setup water-specific parameters */ |
| 139 | inr = nlist->iinr[0]; |
| 140 | iq1 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1])); |
| 141 | iq2 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2])); |
| 142 | iq3 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+3])); |
| 143 | vdwioffset0 = 2*nvdwtype*vdwtype[inr+0]; |
| 144 | |
| 145 | /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */ |
| 146 | rcutoff_scalar = fr->rcoulomb; |
| 147 | rcutoff = _mm_set1_ps(rcutoff_scalar); |
| 148 | rcutoff2 = _mm_mul_ps(rcutoff,rcutoff); |
| 149 | |
| 150 | rswitch_scalar = fr->rcoulomb_switch; |
| 151 | rswitch = _mm_set1_ps(rswitch_scalar); |
| 152 | /* Setup switch parameters */ |
| 153 | d_scalar = rcutoff_scalar-rswitch_scalar; |
| 154 | d = _mm_set1_ps(d_scalar); |
| 155 | swV3 = _mm_set1_ps(-10.0/(d_scalar*d_scalar*d_scalar)); |
| 156 | swV4 = _mm_set1_ps( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar)); |
| 157 | swV5 = _mm_set1_ps( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar)); |
| 158 | swF2 = _mm_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar)); |
| 159 | swF3 = _mm_set1_ps( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar)); |
| 160 | swF4 = _mm_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar)); |
| 161 | |
| 162 | /* Avoid stupid compiler warnings */ |
| 163 | jnrA = jnrB = jnrC = jnrD = 0; |
| 164 | j_coord_offsetA = 0; |
| 165 | j_coord_offsetB = 0; |
| 166 | j_coord_offsetC = 0; |
| 167 | j_coord_offsetD = 0; |
| 168 | |
| 169 | outeriter = 0; |
| 170 | inneriter = 0; |
| 171 | |
| 172 | for(iidx=0;iidx<4*DIM3;iidx++) |
| 173 | { |
| 174 | scratch[iidx] = 0.0; |
| 175 | } |
| 176 | |
| 177 | /* Start outer loop over neighborlists */ |
| 178 | for(iidx=0; iidx<nri; iidx++) |
| 179 | { |
| 180 | /* Load shift vector for this list */ |
| 181 | i_shift_offset = DIM3*shiftidx[iidx]; |
| 182 | |
| 183 | /* Load limits for loop over neighbors */ |
| 184 | j_index_start = jindex[iidx]; |
| 185 | j_index_end = jindex[iidx+1]; |
| 186 | |
| 187 | /* Get outer coordinate index */ |
| 188 | inr = iinr[iidx]; |
| 189 | i_coord_offset = DIM3*inr; |
| 190 | |
| 191 | /* Load i particle coords and add shift vector */ |
| 192 | gmx_mm_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset, |
| 193 | &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3); |
| 194 | |
| 195 | fix0 = _mm_setzero_ps(); |
| 196 | fiy0 = _mm_setzero_ps(); |
| 197 | fiz0 = _mm_setzero_ps(); |
| 198 | fix1 = _mm_setzero_ps(); |
| 199 | fiy1 = _mm_setzero_ps(); |
| 200 | fiz1 = _mm_setzero_ps(); |
| 201 | fix2 = _mm_setzero_ps(); |
| 202 | fiy2 = _mm_setzero_ps(); |
| 203 | fiz2 = _mm_setzero_ps(); |
| 204 | fix3 = _mm_setzero_ps(); |
| 205 | fiy3 = _mm_setzero_ps(); |
| 206 | fiz3 = _mm_setzero_ps(); |
| 207 | |
| 208 | /* Reset potential sums */ |
| 209 | velecsum = _mm_setzero_ps(); |
| 210 | vvdwsum = _mm_setzero_ps(); |
| 211 | |
| 212 | /* Start inner kernel loop */ |
| 213 | for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4) |
| 214 | { |
| 215 | |
| 216 | /* Get j neighbor index, and coordinate index */ |
| 217 | jnrA = jjnr[jidx]; |
| 218 | jnrB = jjnr[jidx+1]; |
| 219 | jnrC = jjnr[jidx+2]; |
| 220 | jnrD = jjnr[jidx+3]; |
| 221 | j_coord_offsetA = DIM3*jnrA; |
| 222 | j_coord_offsetB = DIM3*jnrB; |
| 223 | j_coord_offsetC = DIM3*jnrC; |
| 224 | j_coord_offsetD = DIM3*jnrD; |
| 225 | |
| 226 | /* load j atom coordinates */ |
| 227 | gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB, |
| 228 | x+j_coord_offsetC,x+j_coord_offsetD, |
| 229 | &jx0,&jy0,&jz0); |
| 230 | |
| 231 | /* Calculate displacement vector */ |
| 232 | dx00 = _mm_sub_ps(ix0,jx0); |
| 233 | dy00 = _mm_sub_ps(iy0,jy0); |
| 234 | dz00 = _mm_sub_ps(iz0,jz0); |
| 235 | dx10 = _mm_sub_ps(ix1,jx0); |
| 236 | dy10 = _mm_sub_ps(iy1,jy0); |
| 237 | dz10 = _mm_sub_ps(iz1,jz0); |
| 238 | dx20 = _mm_sub_ps(ix2,jx0); |
| 239 | dy20 = _mm_sub_ps(iy2,jy0); |
| 240 | dz20 = _mm_sub_ps(iz2,jz0); |
| 241 | dx30 = _mm_sub_ps(ix3,jx0); |
| 242 | dy30 = _mm_sub_ps(iy3,jy0); |
| 243 | dz30 = _mm_sub_ps(iz3,jz0); |
| 244 | |
| 245 | /* Calculate squared distance and things based on it */ |
| 246 | rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00); |
| 247 | rsq10 = gmx_mm_calc_rsq_ps(dx10,dy10,dz10); |
| 248 | rsq20 = gmx_mm_calc_rsq_ps(dx20,dy20,dz20); |
| 249 | rsq30 = gmx_mm_calc_rsq_ps(dx30,dy30,dz30); |
| 250 | |
| 251 | rinv00 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq00); |
| 252 | rinv10 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq10); |
| 253 | rinv20 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq20); |
| 254 | rinv30 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq30); |
| 255 | |
| 256 | rinvsq00 = _mm_mul_ps(rinv00,rinv00); |
| 257 | rinvsq10 = _mm_mul_ps(rinv10,rinv10); |
| 258 | rinvsq20 = _mm_mul_ps(rinv20,rinv20); |
| 259 | rinvsq30 = _mm_mul_ps(rinv30,rinv30); |
| 260 | |
| 261 | /* Load parameters for j particles */ |
| 262 | jq0 = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0, |
| 263 | charge+jnrC+0,charge+jnrD+0); |
| 264 | vdwjidx0A = 2*vdwtype[jnrA+0]; |
| 265 | vdwjidx0B = 2*vdwtype[jnrB+0]; |
| 266 | vdwjidx0C = 2*vdwtype[jnrC+0]; |
| 267 | vdwjidx0D = 2*vdwtype[jnrD+0]; |
| 268 | |
| 269 | fjx0 = _mm_setzero_ps(); |
| 270 | fjy0 = _mm_setzero_ps(); |
| 271 | fjz0 = _mm_setzero_ps(); |
| 272 | |
| 273 | /************************** |
| 274 | * CALCULATE INTERACTIONS * |
| 275 | **************************/ |
| 276 | |
| 277 | if (gmx_mm_any_lt(rsq00,rcutoff2)) |
| 278 | { |
| 279 | |
| 280 | r00 = _mm_mul_ps(rsq00,rinv00); |
| 281 | |
| 282 | /* Compute parameters for interactions between i and j atoms */ |
| 283 | gmx_mm_load_4pair_swizzle_ps(vdwparam+vdwioffset0+vdwjidx0A, |
| 284 | vdwparam+vdwioffset0+vdwjidx0B, |
| 285 | vdwparam+vdwioffset0+vdwjidx0C, |
| 286 | vdwparam+vdwioffset0+vdwjidx0D, |
| 287 | &c6_00,&c12_00); |
| 288 | |
| 289 | /* LENNARD-JONES DISPERSION/REPULSION */ |
| 290 | |
| 291 | rinvsix = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00); |
| 292 | vvdw6 = _mm_mul_ps(c6_00,rinvsix); |
| 293 | vvdw12 = _mm_mul_ps(c12_00,_mm_mul_ps(rinvsix,rinvsix)); |
| 294 | vvdw = _mm_sub_ps( _mm_mul_ps(vvdw12,one_twelfth) , _mm_mul_ps(vvdw6,one_sixth) ); |
| 295 | fvdw = _mm_mul_ps(_mm_sub_ps(vvdw12,vvdw6),rinvsq00); |
| 296 | |
| 297 | d = _mm_sub_ps(r00,rswitch); |
| 298 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 299 | d2 = _mm_mul_ps(d,d); |
| 300 | 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))))))); |
| 301 | |
| 302 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 303 | |
| 304 | /* Evaluate switch function */ |
| 305 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 306 | fvdw = _mm_sub_ps( _mm_mul_ps(fvdw,sw) , _mm_mul_ps(rinv00,_mm_mul_ps(vvdw,dsw)) ); |
| 307 | vvdw = _mm_mul_ps(vvdw,sw); |
| 308 | cutoff_mask = _mm_cmplt_ps(rsq00,rcutoff2); |
| 309 | |
| 310 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 311 | vvdw = _mm_and_ps(vvdw,cutoff_mask); |
| 312 | vvdwsum = _mm_add_ps(vvdwsum,vvdw); |
| 313 | |
| 314 | fscal = fvdw; |
| 315 | |
| 316 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 317 | |
| 318 | /* Calculate temporary vectorial force */ |
| 319 | tx = _mm_mul_ps(fscal,dx00); |
| 320 | ty = _mm_mul_ps(fscal,dy00); |
| 321 | tz = _mm_mul_ps(fscal,dz00); |
| 322 | |
| 323 | /* Update vectorial force */ |
| 324 | fix0 = _mm_add_ps(fix0,tx); |
| 325 | fiy0 = _mm_add_ps(fiy0,ty); |
| 326 | fiz0 = _mm_add_ps(fiz0,tz); |
| 327 | |
| 328 | fjx0 = _mm_add_ps(fjx0,tx); |
| 329 | fjy0 = _mm_add_ps(fjy0,ty); |
| 330 | fjz0 = _mm_add_ps(fjz0,tz); |
| 331 | |
| 332 | } |
| 333 | |
| 334 | /************************** |
| 335 | * CALCULATE INTERACTIONS * |
| 336 | **************************/ |
| 337 | |
| 338 | if (gmx_mm_any_lt(rsq10,rcutoff2)) |
| 339 | { |
| 340 | |
| 341 | r10 = _mm_mul_ps(rsq10,rinv10); |
| 342 | |
| 343 | /* Compute parameters for interactions between i and j atoms */ |
| 344 | qq10 = _mm_mul_ps(iq1,jq0); |
| 345 | |
| 346 | /* EWALD ELECTROSTATICS */ |
| 347 | |
| 348 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 349 | ewrt = _mm_mul_ps(r10,ewtabscale); |
| 350 | ewitab = _mm_cvttps_epi32(ewrt); |
| 351 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 352 | ewitab = _mm_slli_epi32(ewitab,2); |
| 353 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 354 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 355 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 356 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 357 | _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); |
| 358 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 359 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 360 | velec = _mm_mul_ps(qq10,_mm_sub_ps(rinv10,velec)); |
| 361 | felec = _mm_mul_ps(_mm_mul_ps(qq10,rinv10),_mm_sub_ps(rinvsq10,felec)); |
| 362 | |
| 363 | d = _mm_sub_ps(r10,rswitch); |
| 364 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 365 | d2 = _mm_mul_ps(d,d); |
| 366 | 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))))))); |
| 367 | |
| 368 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 369 | |
| 370 | /* Evaluate switch function */ |
| 371 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 372 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv10,_mm_mul_ps(velec,dsw)) ); |
| 373 | velec = _mm_mul_ps(velec,sw); |
| 374 | cutoff_mask = _mm_cmplt_ps(rsq10,rcutoff2); |
| 375 | |
| 376 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 377 | velec = _mm_and_ps(velec,cutoff_mask); |
| 378 | velecsum = _mm_add_ps(velecsum,velec); |
| 379 | |
| 380 | fscal = felec; |
| 381 | |
| 382 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 383 | |
| 384 | /* Calculate temporary vectorial force */ |
| 385 | tx = _mm_mul_ps(fscal,dx10); |
| 386 | ty = _mm_mul_ps(fscal,dy10); |
| 387 | tz = _mm_mul_ps(fscal,dz10); |
| 388 | |
| 389 | /* Update vectorial force */ |
| 390 | fix1 = _mm_add_ps(fix1,tx); |
| 391 | fiy1 = _mm_add_ps(fiy1,ty); |
| 392 | fiz1 = _mm_add_ps(fiz1,tz); |
| 393 | |
| 394 | fjx0 = _mm_add_ps(fjx0,tx); |
| 395 | fjy0 = _mm_add_ps(fjy0,ty); |
| 396 | fjz0 = _mm_add_ps(fjz0,tz); |
| 397 | |
| 398 | } |
| 399 | |
| 400 | /************************** |
| 401 | * CALCULATE INTERACTIONS * |
| 402 | **************************/ |
| 403 | |
| 404 | if (gmx_mm_any_lt(rsq20,rcutoff2)) |
| 405 | { |
| 406 | |
| 407 | r20 = _mm_mul_ps(rsq20,rinv20); |
| 408 | |
| 409 | /* Compute parameters for interactions between i and j atoms */ |
| 410 | qq20 = _mm_mul_ps(iq2,jq0); |
| 411 | |
| 412 | /* EWALD ELECTROSTATICS */ |
| 413 | |
| 414 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 415 | ewrt = _mm_mul_ps(r20,ewtabscale); |
| 416 | ewitab = _mm_cvttps_epi32(ewrt); |
| 417 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 418 | ewitab = _mm_slli_epi32(ewitab,2); |
| 419 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 420 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 421 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 422 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 423 | _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); |
| 424 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 425 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 426 | velec = _mm_mul_ps(qq20,_mm_sub_ps(rinv20,velec)); |
| 427 | felec = _mm_mul_ps(_mm_mul_ps(qq20,rinv20),_mm_sub_ps(rinvsq20,felec)); |
| 428 | |
| 429 | d = _mm_sub_ps(r20,rswitch); |
| 430 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 431 | d2 = _mm_mul_ps(d,d); |
| 432 | 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))))))); |
| 433 | |
| 434 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 435 | |
| 436 | /* Evaluate switch function */ |
| 437 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 438 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv20,_mm_mul_ps(velec,dsw)) ); |
| 439 | velec = _mm_mul_ps(velec,sw); |
| 440 | cutoff_mask = _mm_cmplt_ps(rsq20,rcutoff2); |
| 441 | |
| 442 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 443 | velec = _mm_and_ps(velec,cutoff_mask); |
| 444 | velecsum = _mm_add_ps(velecsum,velec); |
| 445 | |
| 446 | fscal = felec; |
| 447 | |
| 448 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 449 | |
| 450 | /* Calculate temporary vectorial force */ |
| 451 | tx = _mm_mul_ps(fscal,dx20); |
| 452 | ty = _mm_mul_ps(fscal,dy20); |
| 453 | tz = _mm_mul_ps(fscal,dz20); |
| 454 | |
| 455 | /* Update vectorial force */ |
| 456 | fix2 = _mm_add_ps(fix2,tx); |
| 457 | fiy2 = _mm_add_ps(fiy2,ty); |
| 458 | fiz2 = _mm_add_ps(fiz2,tz); |
| 459 | |
| 460 | fjx0 = _mm_add_ps(fjx0,tx); |
| 461 | fjy0 = _mm_add_ps(fjy0,ty); |
| 462 | fjz0 = _mm_add_ps(fjz0,tz); |
| 463 | |
| 464 | } |
| 465 | |
| 466 | /************************** |
| 467 | * CALCULATE INTERACTIONS * |
| 468 | **************************/ |
| 469 | |
| 470 | if (gmx_mm_any_lt(rsq30,rcutoff2)) |
| 471 | { |
| 472 | |
| 473 | r30 = _mm_mul_ps(rsq30,rinv30); |
| 474 | |
| 475 | /* Compute parameters for interactions between i and j atoms */ |
| 476 | qq30 = _mm_mul_ps(iq3,jq0); |
| 477 | |
| 478 | /* EWALD ELECTROSTATICS */ |
| 479 | |
| 480 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 481 | ewrt = _mm_mul_ps(r30,ewtabscale); |
| 482 | ewitab = _mm_cvttps_epi32(ewrt); |
| 483 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 484 | ewitab = _mm_slli_epi32(ewitab,2); |
| 485 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 486 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 487 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 488 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 489 | _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); |
| 490 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 491 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 492 | velec = _mm_mul_ps(qq30,_mm_sub_ps(rinv30,velec)); |
| 493 | felec = _mm_mul_ps(_mm_mul_ps(qq30,rinv30),_mm_sub_ps(rinvsq30,felec)); |
| 494 | |
| 495 | d = _mm_sub_ps(r30,rswitch); |
| 496 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 497 | d2 = _mm_mul_ps(d,d); |
| 498 | 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))))))); |
| 499 | |
| 500 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 501 | |
| 502 | /* Evaluate switch function */ |
| 503 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 504 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv30,_mm_mul_ps(velec,dsw)) ); |
| 505 | velec = _mm_mul_ps(velec,sw); |
| 506 | cutoff_mask = _mm_cmplt_ps(rsq30,rcutoff2); |
| 507 | |
| 508 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 509 | velec = _mm_and_ps(velec,cutoff_mask); |
| 510 | velecsum = _mm_add_ps(velecsum,velec); |
| 511 | |
| 512 | fscal = felec; |
| 513 | |
| 514 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 515 | |
| 516 | /* Calculate temporary vectorial force */ |
| 517 | tx = _mm_mul_ps(fscal,dx30); |
| 518 | ty = _mm_mul_ps(fscal,dy30); |
| 519 | tz = _mm_mul_ps(fscal,dz30); |
| 520 | |
| 521 | /* Update vectorial force */ |
| 522 | fix3 = _mm_add_ps(fix3,tx); |
| 523 | fiy3 = _mm_add_ps(fiy3,ty); |
| 524 | fiz3 = _mm_add_ps(fiz3,tz); |
| 525 | |
| 526 | fjx0 = _mm_add_ps(fjx0,tx); |
| 527 | fjy0 = _mm_add_ps(fjy0,ty); |
| 528 | fjz0 = _mm_add_ps(fjz0,tz); |
| 529 | |
| 530 | } |
| 531 | |
| 532 | fjptrA = f+j_coord_offsetA; |
| 533 | fjptrB = f+j_coord_offsetB; |
| 534 | fjptrC = f+j_coord_offsetC; |
| 535 | fjptrD = f+j_coord_offsetD; |
| 536 | |
| 537 | gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjx0,fjy0,fjz0); |
| 538 | |
| 539 | /* Inner loop uses 254 flops */ |
| 540 | } |
| 541 | |
| 542 | if(jidx<j_index_end) |
| 543 | { |
| 544 | |
| 545 | /* Get j neighbor index, and coordinate index */ |
| 546 | jnrlistA = jjnr[jidx]; |
| 547 | jnrlistB = jjnr[jidx+1]; |
| 548 | jnrlistC = jjnr[jidx+2]; |
| 549 | jnrlistD = jjnr[jidx+3]; |
| 550 | /* Sign of each element will be negative for non-real atoms. |
| 551 | * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones, |
| 552 | * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries. |
| 553 | */ |
| 554 | dummy_mask = gmx_mm_castsi128_ps_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())); |
| 555 | jnrA = (jnrlistA>=0) ? jnrlistA : 0; |
| 556 | jnrB = (jnrlistB>=0) ? jnrlistB : 0; |
| 557 | jnrC = (jnrlistC>=0) ? jnrlistC : 0; |
| 558 | jnrD = (jnrlistD>=0) ? jnrlistD : 0; |
| 559 | j_coord_offsetA = DIM3*jnrA; |
| 560 | j_coord_offsetB = DIM3*jnrB; |
| 561 | j_coord_offsetC = DIM3*jnrC; |
| 562 | j_coord_offsetD = DIM3*jnrD; |
| 563 | |
| 564 | /* load j atom coordinates */ |
| 565 | gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB, |
| 566 | x+j_coord_offsetC,x+j_coord_offsetD, |
| 567 | &jx0,&jy0,&jz0); |
| 568 | |
| 569 | /* Calculate displacement vector */ |
| 570 | dx00 = _mm_sub_ps(ix0,jx0); |
| 571 | dy00 = _mm_sub_ps(iy0,jy0); |
| 572 | dz00 = _mm_sub_ps(iz0,jz0); |
| 573 | dx10 = _mm_sub_ps(ix1,jx0); |
| 574 | dy10 = _mm_sub_ps(iy1,jy0); |
| 575 | dz10 = _mm_sub_ps(iz1,jz0); |
| 576 | dx20 = _mm_sub_ps(ix2,jx0); |
| 577 | dy20 = _mm_sub_ps(iy2,jy0); |
| 578 | dz20 = _mm_sub_ps(iz2,jz0); |
| 579 | dx30 = _mm_sub_ps(ix3,jx0); |
| 580 | dy30 = _mm_sub_ps(iy3,jy0); |
| 581 | dz30 = _mm_sub_ps(iz3,jz0); |
| 582 | |
| 583 | /* Calculate squared distance and things based on it */ |
| 584 | rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00); |
| 585 | rsq10 = gmx_mm_calc_rsq_ps(dx10,dy10,dz10); |
| 586 | rsq20 = gmx_mm_calc_rsq_ps(dx20,dy20,dz20); |
| 587 | rsq30 = gmx_mm_calc_rsq_ps(dx30,dy30,dz30); |
| 588 | |
| 589 | rinv00 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq00); |
| 590 | rinv10 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq10); |
| 591 | rinv20 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq20); |
| 592 | rinv30 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq30); |
| 593 | |
| 594 | rinvsq00 = _mm_mul_ps(rinv00,rinv00); |
| 595 | rinvsq10 = _mm_mul_ps(rinv10,rinv10); |
| 596 | rinvsq20 = _mm_mul_ps(rinv20,rinv20); |
| 597 | rinvsq30 = _mm_mul_ps(rinv30,rinv30); |
| 598 | |
| 599 | /* Load parameters for j particles */ |
| 600 | jq0 = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0, |
| 601 | charge+jnrC+0,charge+jnrD+0); |
| 602 | vdwjidx0A = 2*vdwtype[jnrA+0]; |
| 603 | vdwjidx0B = 2*vdwtype[jnrB+0]; |
| 604 | vdwjidx0C = 2*vdwtype[jnrC+0]; |
| 605 | vdwjidx0D = 2*vdwtype[jnrD+0]; |
| 606 | |
| 607 | fjx0 = _mm_setzero_ps(); |
| 608 | fjy0 = _mm_setzero_ps(); |
| 609 | fjz0 = _mm_setzero_ps(); |
| 610 | |
| 611 | /************************** |
| 612 | * CALCULATE INTERACTIONS * |
| 613 | **************************/ |
| 614 | |
| 615 | if (gmx_mm_any_lt(rsq00,rcutoff2)) |
| 616 | { |
| 617 | |
| 618 | r00 = _mm_mul_ps(rsq00,rinv00); |
| 619 | r00 = _mm_andnot_ps(dummy_mask,r00); |
| 620 | |
| 621 | /* Compute parameters for interactions between i and j atoms */ |
| 622 | gmx_mm_load_4pair_swizzle_ps(vdwparam+vdwioffset0+vdwjidx0A, |
| 623 | vdwparam+vdwioffset0+vdwjidx0B, |
| 624 | vdwparam+vdwioffset0+vdwjidx0C, |
| 625 | vdwparam+vdwioffset0+vdwjidx0D, |
| 626 | &c6_00,&c12_00); |
| 627 | |
| 628 | /* LENNARD-JONES DISPERSION/REPULSION */ |
| 629 | |
| 630 | rinvsix = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00); |
| 631 | vvdw6 = _mm_mul_ps(c6_00,rinvsix); |
| 632 | vvdw12 = _mm_mul_ps(c12_00,_mm_mul_ps(rinvsix,rinvsix)); |
| 633 | vvdw = _mm_sub_ps( _mm_mul_ps(vvdw12,one_twelfth) , _mm_mul_ps(vvdw6,one_sixth) ); |
| 634 | fvdw = _mm_mul_ps(_mm_sub_ps(vvdw12,vvdw6),rinvsq00); |
| 635 | |
| 636 | d = _mm_sub_ps(r00,rswitch); |
| 637 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 638 | d2 = _mm_mul_ps(d,d); |
| 639 | 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))))))); |
| 640 | |
| 641 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 642 | |
| 643 | /* Evaluate switch function */ |
| 644 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 645 | fvdw = _mm_sub_ps( _mm_mul_ps(fvdw,sw) , _mm_mul_ps(rinv00,_mm_mul_ps(vvdw,dsw)) ); |
| 646 | vvdw = _mm_mul_ps(vvdw,sw); |
| 647 | cutoff_mask = _mm_cmplt_ps(rsq00,rcutoff2); |
| 648 | |
| 649 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 650 | vvdw = _mm_and_ps(vvdw,cutoff_mask); |
| 651 | vvdw = _mm_andnot_ps(dummy_mask,vvdw); |
| 652 | vvdwsum = _mm_add_ps(vvdwsum,vvdw); |
| 653 | |
| 654 | fscal = fvdw; |
| 655 | |
| 656 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 657 | |
| 658 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 659 | |
| 660 | /* Calculate temporary vectorial force */ |
| 661 | tx = _mm_mul_ps(fscal,dx00); |
| 662 | ty = _mm_mul_ps(fscal,dy00); |
| 663 | tz = _mm_mul_ps(fscal,dz00); |
| 664 | |
| 665 | /* Update vectorial force */ |
| 666 | fix0 = _mm_add_ps(fix0,tx); |
| 667 | fiy0 = _mm_add_ps(fiy0,ty); |
| 668 | fiz0 = _mm_add_ps(fiz0,tz); |
| 669 | |
| 670 | fjx0 = _mm_add_ps(fjx0,tx); |
| 671 | fjy0 = _mm_add_ps(fjy0,ty); |
| 672 | fjz0 = _mm_add_ps(fjz0,tz); |
| 673 | |
| 674 | } |
| 675 | |
| 676 | /************************** |
| 677 | * CALCULATE INTERACTIONS * |
| 678 | **************************/ |
| 679 | |
| 680 | if (gmx_mm_any_lt(rsq10,rcutoff2)) |
| 681 | { |
| 682 | |
| 683 | r10 = _mm_mul_ps(rsq10,rinv10); |
| 684 | r10 = _mm_andnot_ps(dummy_mask,r10); |
| 685 | |
| 686 | /* Compute parameters for interactions between i and j atoms */ |
| 687 | qq10 = _mm_mul_ps(iq1,jq0); |
| 688 | |
| 689 | /* EWALD ELECTROSTATICS */ |
| 690 | |
| 691 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 692 | ewrt = _mm_mul_ps(r10,ewtabscale); |
| 693 | ewitab = _mm_cvttps_epi32(ewrt); |
| 694 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 695 | ewitab = _mm_slli_epi32(ewitab,2); |
| 696 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 697 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 698 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 699 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 700 | _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); |
| 701 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 702 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 703 | velec = _mm_mul_ps(qq10,_mm_sub_ps(rinv10,velec)); |
| 704 | felec = _mm_mul_ps(_mm_mul_ps(qq10,rinv10),_mm_sub_ps(rinvsq10,felec)); |
| 705 | |
| 706 | d = _mm_sub_ps(r10,rswitch); |
| 707 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 708 | d2 = _mm_mul_ps(d,d); |
| 709 | 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))))))); |
| 710 | |
| 711 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 712 | |
| 713 | /* Evaluate switch function */ |
| 714 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 715 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv10,_mm_mul_ps(velec,dsw)) ); |
| 716 | velec = _mm_mul_ps(velec,sw); |
| 717 | cutoff_mask = _mm_cmplt_ps(rsq10,rcutoff2); |
| 718 | |
| 719 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 720 | velec = _mm_and_ps(velec,cutoff_mask); |
| 721 | velec = _mm_andnot_ps(dummy_mask,velec); |
| 722 | velecsum = _mm_add_ps(velecsum,velec); |
| 723 | |
| 724 | fscal = felec; |
| 725 | |
| 726 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 727 | |
| 728 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 729 | |
| 730 | /* Calculate temporary vectorial force */ |
| 731 | tx = _mm_mul_ps(fscal,dx10); |
| 732 | ty = _mm_mul_ps(fscal,dy10); |
| 733 | tz = _mm_mul_ps(fscal,dz10); |
| 734 | |
| 735 | /* Update vectorial force */ |
| 736 | fix1 = _mm_add_ps(fix1,tx); |
| 737 | fiy1 = _mm_add_ps(fiy1,ty); |
| 738 | fiz1 = _mm_add_ps(fiz1,tz); |
| 739 | |
| 740 | fjx0 = _mm_add_ps(fjx0,tx); |
| 741 | fjy0 = _mm_add_ps(fjy0,ty); |
| 742 | fjz0 = _mm_add_ps(fjz0,tz); |
| 743 | |
| 744 | } |
| 745 | |
| 746 | /************************** |
| 747 | * CALCULATE INTERACTIONS * |
| 748 | **************************/ |
| 749 | |
| 750 | if (gmx_mm_any_lt(rsq20,rcutoff2)) |
| 751 | { |
| 752 | |
| 753 | r20 = _mm_mul_ps(rsq20,rinv20); |
| 754 | r20 = _mm_andnot_ps(dummy_mask,r20); |
| 755 | |
| 756 | /* Compute parameters for interactions between i and j atoms */ |
| 757 | qq20 = _mm_mul_ps(iq2,jq0); |
| 758 | |
| 759 | /* EWALD ELECTROSTATICS */ |
| 760 | |
| 761 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 762 | ewrt = _mm_mul_ps(r20,ewtabscale); |
| 763 | ewitab = _mm_cvttps_epi32(ewrt); |
| 764 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 765 | ewitab = _mm_slli_epi32(ewitab,2); |
| 766 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 767 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 768 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 769 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 770 | _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); |
| 771 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 772 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 773 | velec = _mm_mul_ps(qq20,_mm_sub_ps(rinv20,velec)); |
| 774 | felec = _mm_mul_ps(_mm_mul_ps(qq20,rinv20),_mm_sub_ps(rinvsq20,felec)); |
| 775 | |
| 776 | d = _mm_sub_ps(r20,rswitch); |
| 777 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 778 | d2 = _mm_mul_ps(d,d); |
| 779 | 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))))))); |
| 780 | |
| 781 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 782 | |
| 783 | /* Evaluate switch function */ |
| 784 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 785 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv20,_mm_mul_ps(velec,dsw)) ); |
| 786 | velec = _mm_mul_ps(velec,sw); |
| 787 | cutoff_mask = _mm_cmplt_ps(rsq20,rcutoff2); |
| 788 | |
| 789 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 790 | velec = _mm_and_ps(velec,cutoff_mask); |
| 791 | velec = _mm_andnot_ps(dummy_mask,velec); |
| 792 | velecsum = _mm_add_ps(velecsum,velec); |
| 793 | |
| 794 | fscal = felec; |
| 795 | |
| 796 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 797 | |
| 798 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 799 | |
| 800 | /* Calculate temporary vectorial force */ |
| 801 | tx = _mm_mul_ps(fscal,dx20); |
| 802 | ty = _mm_mul_ps(fscal,dy20); |
| 803 | tz = _mm_mul_ps(fscal,dz20); |
| 804 | |
| 805 | /* Update vectorial force */ |
| 806 | fix2 = _mm_add_ps(fix2,tx); |
| 807 | fiy2 = _mm_add_ps(fiy2,ty); |
| 808 | fiz2 = _mm_add_ps(fiz2,tz); |
| 809 | |
| 810 | fjx0 = _mm_add_ps(fjx0,tx); |
| 811 | fjy0 = _mm_add_ps(fjy0,ty); |
| 812 | fjz0 = _mm_add_ps(fjz0,tz); |
| 813 | |
| 814 | } |
| 815 | |
| 816 | /************************** |
| 817 | * CALCULATE INTERACTIONS * |
| 818 | **************************/ |
| 819 | |
| 820 | if (gmx_mm_any_lt(rsq30,rcutoff2)) |
| 821 | { |
| 822 | |
| 823 | r30 = _mm_mul_ps(rsq30,rinv30); |
| 824 | r30 = _mm_andnot_ps(dummy_mask,r30); |
| 825 | |
| 826 | /* Compute parameters for interactions between i and j atoms */ |
| 827 | qq30 = _mm_mul_ps(iq3,jq0); |
| 828 | |
| 829 | /* EWALD ELECTROSTATICS */ |
| 830 | |
| 831 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 832 | ewrt = _mm_mul_ps(r30,ewtabscale); |
| 833 | ewitab = _mm_cvttps_epi32(ewrt); |
| 834 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 835 | ewitab = _mm_slli_epi32(ewitab,2); |
| 836 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 837 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 838 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 839 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 840 | _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); |
| 841 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 842 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 843 | velec = _mm_mul_ps(qq30,_mm_sub_ps(rinv30,velec)); |
| 844 | felec = _mm_mul_ps(_mm_mul_ps(qq30,rinv30),_mm_sub_ps(rinvsq30,felec)); |
| 845 | |
| 846 | d = _mm_sub_ps(r30,rswitch); |
| 847 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 848 | d2 = _mm_mul_ps(d,d); |
| 849 | 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))))))); |
| 850 | |
| 851 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 852 | |
| 853 | /* Evaluate switch function */ |
| 854 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 855 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv30,_mm_mul_ps(velec,dsw)) ); |
| 856 | velec = _mm_mul_ps(velec,sw); |
| 857 | cutoff_mask = _mm_cmplt_ps(rsq30,rcutoff2); |
| 858 | |
| 859 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 860 | velec = _mm_and_ps(velec,cutoff_mask); |
| 861 | velec = _mm_andnot_ps(dummy_mask,velec); |
| 862 | velecsum = _mm_add_ps(velecsum,velec); |
| 863 | |
| 864 | fscal = felec; |
| 865 | |
| 866 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 867 | |
| 868 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 869 | |
| 870 | /* Calculate temporary vectorial force */ |
| 871 | tx = _mm_mul_ps(fscal,dx30); |
| 872 | ty = _mm_mul_ps(fscal,dy30); |
| 873 | tz = _mm_mul_ps(fscal,dz30); |
| 874 | |
| 875 | /* Update vectorial force */ |
| 876 | fix3 = _mm_add_ps(fix3,tx); |
| 877 | fiy3 = _mm_add_ps(fiy3,ty); |
| 878 | fiz3 = _mm_add_ps(fiz3,tz); |
| 879 | |
| 880 | fjx0 = _mm_add_ps(fjx0,tx); |
| 881 | fjy0 = _mm_add_ps(fjy0,ty); |
| 882 | fjz0 = _mm_add_ps(fjz0,tz); |
| 883 | |
| 884 | } |
| 885 | |
| 886 | fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch; |
| 887 | fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch; |
| 888 | fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch; |
| 889 | fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch; |
| 890 | |
| 891 | gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjx0,fjy0,fjz0); |
| 892 | |
| 893 | /* Inner loop uses 258 flops */ |
| 894 | } |
| 895 | |
| 896 | /* End of innermost loop */ |
| 897 | |
| 898 | gmx_mm_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3, |
| 899 | f+i_coord_offset,fshift+i_shift_offset); |
| 900 | |
| 901 | ggid = gid[iidx]; |
| 902 | /* Update potential energies */ |
| 903 | gmx_mm_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid); |
| 904 | gmx_mm_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid); |
| 905 | |
| 906 | /* Increment number of inner iterations */ |
| 907 | inneriter += j_index_end - j_index_start; |
| 908 | |
| 909 | /* Outer loop uses 26 flops */ |
| 910 | } |
| 911 | |
| 912 | /* Increment number of outer iterations */ |
| 913 | outeriter += nri; |
| 914 | |
| 915 | /* Update outer/inner flops */ |
| 916 | |
| 917 | inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_VF,outeriter*26 + inneriter*258)(nrnb)->n[eNR_NBKERNEL_ELEC_VDW_W4_VF] += outeriter*26 + inneriter *258; |
| 918 | } |
| 919 | /* |
| 920 | * Gromacs nonbonded kernel: nb_kernel_ElecEwSw_VdwLJSw_GeomW4P1_F_sse4_1_single |
| 921 | * Electrostatics interaction: Ewald |
| 922 | * VdW interaction: LennardJones |
| 923 | * Geometry: Water4-Particle |
| 924 | * Calculate force/pot: Force |
| 925 | */ |
| 926 | void |
| 927 | nb_kernel_ElecEwSw_VdwLJSw_GeomW4P1_F_sse4_1_single |
| 928 | (t_nblist * gmx_restrict nlist, |
| 929 | rvec * gmx_restrict xx, |
| 930 | rvec * gmx_restrict ff, |
| 931 | t_forcerec * gmx_restrict fr, |
| 932 | t_mdatoms * gmx_restrict mdatoms, |
| 933 | nb_kernel_data_t gmx_unused__attribute__ ((unused)) * gmx_restrict kernel_data, |
| 934 | t_nrnb * gmx_restrict nrnb) |
| 935 | { |
| 936 | /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or |
| 937 | * just 0 for non-waters. |
| 938 | * Suffixes A,B,C,D refer to j loop unrolling done with SSE, e.g. for the four different |
| 939 | * jnr indices corresponding to data put in the four positions in the SIMD register. |
| 940 | */ |
| 941 | int i_shift_offset,i_coord_offset,outeriter,inneriter; |
| 942 | int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx; |
| 943 | int jnrA,jnrB,jnrC,jnrD; |
| 944 | int jnrlistA,jnrlistB,jnrlistC,jnrlistD; |
| 945 | int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD; |
| 946 | int *iinr,*jindex,*jjnr,*shiftidx,*gid; |
| 947 | real rcutoff_scalar; |
| 948 | real *shiftvec,*fshift,*x,*f; |
| 949 | real *fjptrA,*fjptrB,*fjptrC,*fjptrD; |
| 950 | real scratch[4*DIM3]; |
| 951 | __m128 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall; |
| 952 | int vdwioffset0; |
| 953 | __m128 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0; |
| 954 | int vdwioffset1; |
| 955 | __m128 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1; |
| 956 | int vdwioffset2; |
| 957 | __m128 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2; |
| 958 | int vdwioffset3; |
| 959 | __m128 ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3; |
| 960 | int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D; |
| 961 | __m128 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0; |
| 962 | __m128 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00; |
| 963 | __m128 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10; |
| 964 | __m128 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20; |
| 965 | __m128 dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30; |
| 966 | __m128 velec,felec,velecsum,facel,crf,krf,krf2; |
| 967 | real *charge; |
| 968 | int nvdwtype; |
| 969 | __m128 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6; |
| 970 | int *vdwtype; |
| 971 | real *vdwparam; |
| 972 | __m128 one_sixth = _mm_set1_ps(1.0/6.0); |
| 973 | __m128 one_twelfth = _mm_set1_ps(1.0/12.0); |
| 974 | __m128i ewitab; |
| 975 | __m128 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV; |
| 976 | real *ewtab; |
| 977 | __m128 rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw; |
| 978 | real rswitch_scalar,d_scalar; |
| 979 | __m128 dummy_mask,cutoff_mask; |
| 980 | __m128 signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) ); |
| 981 | __m128 one = _mm_set1_ps(1.0); |
| 982 | __m128 two = _mm_set1_ps(2.0); |
| 983 | x = xx[0]; |
| 984 | f = ff[0]; |
| 985 | |
| 986 | nri = nlist->nri; |
| 987 | iinr = nlist->iinr; |
| 988 | jindex = nlist->jindex; |
| 989 | jjnr = nlist->jjnr; |
| 990 | shiftidx = nlist->shift; |
| 991 | gid = nlist->gid; |
| 992 | shiftvec = fr->shift_vec[0]; |
| 993 | fshift = fr->fshift[0]; |
| 994 | facel = _mm_set1_ps(fr->epsfac); |
| 995 | charge = mdatoms->chargeA; |
| 996 | nvdwtype = fr->ntype; |
| 997 | vdwparam = fr->nbfp; |
| 998 | vdwtype = mdatoms->typeA; |
| 999 | |
| 1000 | sh_ewald = _mm_set1_ps(fr->ic->sh_ewald); |
| 1001 | ewtab = fr->ic->tabq_coul_FDV0; |
| 1002 | ewtabscale = _mm_set1_ps(fr->ic->tabq_scale); |
| 1003 | ewtabhalfspace = _mm_set1_ps(0.5/fr->ic->tabq_scale); |
| 1004 | |
| 1005 | /* Setup water-specific parameters */ |
| 1006 | inr = nlist->iinr[0]; |
| 1007 | iq1 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1])); |
| 1008 | iq2 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2])); |
| 1009 | iq3 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+3])); |
| 1010 | vdwioffset0 = 2*nvdwtype*vdwtype[inr+0]; |
| 1011 | |
| 1012 | /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */ |
| 1013 | rcutoff_scalar = fr->rcoulomb; |
| 1014 | rcutoff = _mm_set1_ps(rcutoff_scalar); |
| 1015 | rcutoff2 = _mm_mul_ps(rcutoff,rcutoff); |
| 1016 | |
| 1017 | rswitch_scalar = fr->rcoulomb_switch; |
| 1018 | rswitch = _mm_set1_ps(rswitch_scalar); |
| 1019 | /* Setup switch parameters */ |
| 1020 | d_scalar = rcutoff_scalar-rswitch_scalar; |
| 1021 | d = _mm_set1_ps(d_scalar); |
| 1022 | swV3 = _mm_set1_ps(-10.0/(d_scalar*d_scalar*d_scalar)); |
| 1023 | swV4 = _mm_set1_ps( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar)); |
| 1024 | swV5 = _mm_set1_ps( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar)); |
| 1025 | swF2 = _mm_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar)); |
| 1026 | swF3 = _mm_set1_ps( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar)); |
| 1027 | swF4 = _mm_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar)); |
| 1028 | |
| 1029 | /* Avoid stupid compiler warnings */ |
| 1030 | jnrA = jnrB = jnrC = jnrD = 0; |
| 1031 | j_coord_offsetA = 0; |
| 1032 | j_coord_offsetB = 0; |
| 1033 | j_coord_offsetC = 0; |
Value stored to 'j_coord_offsetC' is never read | |
| 1034 | j_coord_offsetD = 0; |
| 1035 | |
| 1036 | outeriter = 0; |
| 1037 | inneriter = 0; |
| 1038 | |
| 1039 | for(iidx=0;iidx<4*DIM3;iidx++) |
| 1040 | { |
| 1041 | scratch[iidx] = 0.0; |
| 1042 | } |
| 1043 | |
| 1044 | /* Start outer loop over neighborlists */ |
| 1045 | for(iidx=0; iidx<nri; iidx++) |
| 1046 | { |
| 1047 | /* Load shift vector for this list */ |
| 1048 | i_shift_offset = DIM3*shiftidx[iidx]; |
| 1049 | |
| 1050 | /* Load limits for loop over neighbors */ |
| 1051 | j_index_start = jindex[iidx]; |
| 1052 | j_index_end = jindex[iidx+1]; |
| 1053 | |
| 1054 | /* Get outer coordinate index */ |
| 1055 | inr = iinr[iidx]; |
| 1056 | i_coord_offset = DIM3*inr; |
| 1057 | |
| 1058 | /* Load i particle coords and add shift vector */ |
| 1059 | gmx_mm_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset, |
| 1060 | &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3); |
| 1061 | |
| 1062 | fix0 = _mm_setzero_ps(); |
| 1063 | fiy0 = _mm_setzero_ps(); |
| 1064 | fiz0 = _mm_setzero_ps(); |
| 1065 | fix1 = _mm_setzero_ps(); |
| 1066 | fiy1 = _mm_setzero_ps(); |
| 1067 | fiz1 = _mm_setzero_ps(); |
| 1068 | fix2 = _mm_setzero_ps(); |
| 1069 | fiy2 = _mm_setzero_ps(); |
| 1070 | fiz2 = _mm_setzero_ps(); |
| 1071 | fix3 = _mm_setzero_ps(); |
| 1072 | fiy3 = _mm_setzero_ps(); |
| 1073 | fiz3 = _mm_setzero_ps(); |
| 1074 | |
| 1075 | /* Start inner kernel loop */ |
| 1076 | for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4) |
| 1077 | { |
| 1078 | |
| 1079 | /* Get j neighbor index, and coordinate index */ |
| 1080 | jnrA = jjnr[jidx]; |
| 1081 | jnrB = jjnr[jidx+1]; |
| 1082 | jnrC = jjnr[jidx+2]; |
| 1083 | jnrD = jjnr[jidx+3]; |
| 1084 | j_coord_offsetA = DIM3*jnrA; |
| 1085 | j_coord_offsetB = DIM3*jnrB; |
| 1086 | j_coord_offsetC = DIM3*jnrC; |
| 1087 | j_coord_offsetD = DIM3*jnrD; |
| 1088 | |
| 1089 | /* load j atom coordinates */ |
| 1090 | gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB, |
| 1091 | x+j_coord_offsetC,x+j_coord_offsetD, |
| 1092 | &jx0,&jy0,&jz0); |
| 1093 | |
| 1094 | /* Calculate displacement vector */ |
| 1095 | dx00 = _mm_sub_ps(ix0,jx0); |
| 1096 | dy00 = _mm_sub_ps(iy0,jy0); |
| 1097 | dz00 = _mm_sub_ps(iz0,jz0); |
| 1098 | dx10 = _mm_sub_ps(ix1,jx0); |
| 1099 | dy10 = _mm_sub_ps(iy1,jy0); |
| 1100 | dz10 = _mm_sub_ps(iz1,jz0); |
| 1101 | dx20 = _mm_sub_ps(ix2,jx0); |
| 1102 | dy20 = _mm_sub_ps(iy2,jy0); |
| 1103 | dz20 = _mm_sub_ps(iz2,jz0); |
| 1104 | dx30 = _mm_sub_ps(ix3,jx0); |
| 1105 | dy30 = _mm_sub_ps(iy3,jy0); |
| 1106 | dz30 = _mm_sub_ps(iz3,jz0); |
| 1107 | |
| 1108 | /* Calculate squared distance and things based on it */ |
| 1109 | rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00); |
| 1110 | rsq10 = gmx_mm_calc_rsq_ps(dx10,dy10,dz10); |
| 1111 | rsq20 = gmx_mm_calc_rsq_ps(dx20,dy20,dz20); |
| 1112 | rsq30 = gmx_mm_calc_rsq_ps(dx30,dy30,dz30); |
| 1113 | |
| 1114 | rinv00 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq00); |
| 1115 | rinv10 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq10); |
| 1116 | rinv20 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq20); |
| 1117 | rinv30 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq30); |
| 1118 | |
| 1119 | rinvsq00 = _mm_mul_ps(rinv00,rinv00); |
| 1120 | rinvsq10 = _mm_mul_ps(rinv10,rinv10); |
| 1121 | rinvsq20 = _mm_mul_ps(rinv20,rinv20); |
| 1122 | rinvsq30 = _mm_mul_ps(rinv30,rinv30); |
| 1123 | |
| 1124 | /* Load parameters for j particles */ |
| 1125 | jq0 = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0, |
| 1126 | charge+jnrC+0,charge+jnrD+0); |
| 1127 | vdwjidx0A = 2*vdwtype[jnrA+0]; |
| 1128 | vdwjidx0B = 2*vdwtype[jnrB+0]; |
| 1129 | vdwjidx0C = 2*vdwtype[jnrC+0]; |
| 1130 | vdwjidx0D = 2*vdwtype[jnrD+0]; |
| 1131 | |
| 1132 | fjx0 = _mm_setzero_ps(); |
| 1133 | fjy0 = _mm_setzero_ps(); |
| 1134 | fjz0 = _mm_setzero_ps(); |
| 1135 | |
| 1136 | /************************** |
| 1137 | * CALCULATE INTERACTIONS * |
| 1138 | **************************/ |
| 1139 | |
| 1140 | if (gmx_mm_any_lt(rsq00,rcutoff2)) |
| 1141 | { |
| 1142 | |
| 1143 | r00 = _mm_mul_ps(rsq00,rinv00); |
| 1144 | |
| 1145 | /* Compute parameters for interactions between i and j atoms */ |
| 1146 | gmx_mm_load_4pair_swizzle_ps(vdwparam+vdwioffset0+vdwjidx0A, |
| 1147 | vdwparam+vdwioffset0+vdwjidx0B, |
| 1148 | vdwparam+vdwioffset0+vdwjidx0C, |
| 1149 | vdwparam+vdwioffset0+vdwjidx0D, |
| 1150 | &c6_00,&c12_00); |
| 1151 | |
| 1152 | /* LENNARD-JONES DISPERSION/REPULSION */ |
| 1153 | |
| 1154 | rinvsix = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00); |
| 1155 | vvdw6 = _mm_mul_ps(c6_00,rinvsix); |
| 1156 | vvdw12 = _mm_mul_ps(c12_00,_mm_mul_ps(rinvsix,rinvsix)); |
| 1157 | vvdw = _mm_sub_ps( _mm_mul_ps(vvdw12,one_twelfth) , _mm_mul_ps(vvdw6,one_sixth) ); |
| 1158 | fvdw = _mm_mul_ps(_mm_sub_ps(vvdw12,vvdw6),rinvsq00); |
| 1159 | |
| 1160 | d = _mm_sub_ps(r00,rswitch); |
| 1161 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 1162 | d2 = _mm_mul_ps(d,d); |
| 1163 | 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))))))); |
| 1164 | |
| 1165 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 1166 | |
| 1167 | /* Evaluate switch function */ |
| 1168 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 1169 | fvdw = _mm_sub_ps( _mm_mul_ps(fvdw,sw) , _mm_mul_ps(rinv00,_mm_mul_ps(vvdw,dsw)) ); |
| 1170 | cutoff_mask = _mm_cmplt_ps(rsq00,rcutoff2); |
| 1171 | |
| 1172 | fscal = fvdw; |
| 1173 | |
| 1174 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 1175 | |
| 1176 | /* Calculate temporary vectorial force */ |
| 1177 | tx = _mm_mul_ps(fscal,dx00); |
| 1178 | ty = _mm_mul_ps(fscal,dy00); |
| 1179 | tz = _mm_mul_ps(fscal,dz00); |
| 1180 | |
| 1181 | /* Update vectorial force */ |
| 1182 | fix0 = _mm_add_ps(fix0,tx); |
| 1183 | fiy0 = _mm_add_ps(fiy0,ty); |
| 1184 | fiz0 = _mm_add_ps(fiz0,tz); |
| 1185 | |
| 1186 | fjx0 = _mm_add_ps(fjx0,tx); |
| 1187 | fjy0 = _mm_add_ps(fjy0,ty); |
| 1188 | fjz0 = _mm_add_ps(fjz0,tz); |
| 1189 | |
| 1190 | } |
| 1191 | |
| 1192 | /************************** |
| 1193 | * CALCULATE INTERACTIONS * |
| 1194 | **************************/ |
| 1195 | |
| 1196 | if (gmx_mm_any_lt(rsq10,rcutoff2)) |
| 1197 | { |
| 1198 | |
| 1199 | r10 = _mm_mul_ps(rsq10,rinv10); |
| 1200 | |
| 1201 | /* Compute parameters for interactions between i and j atoms */ |
| 1202 | qq10 = _mm_mul_ps(iq1,jq0); |
| 1203 | |
| 1204 | /* EWALD ELECTROSTATICS */ |
| 1205 | |
| 1206 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 1207 | ewrt = _mm_mul_ps(r10,ewtabscale); |
| 1208 | ewitab = _mm_cvttps_epi32(ewrt); |
| 1209 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 1210 | ewitab = _mm_slli_epi32(ewitab,2); |
| 1211 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 1212 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 1213 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 1214 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 1215 | _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); |
| 1216 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 1217 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 1218 | velec = _mm_mul_ps(qq10,_mm_sub_ps(rinv10,velec)); |
| 1219 | felec = _mm_mul_ps(_mm_mul_ps(qq10,rinv10),_mm_sub_ps(rinvsq10,felec)); |
| 1220 | |
| 1221 | d = _mm_sub_ps(r10,rswitch); |
| 1222 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 1223 | d2 = _mm_mul_ps(d,d); |
| 1224 | 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))))))); |
| 1225 | |
| 1226 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 1227 | |
| 1228 | /* Evaluate switch function */ |
| 1229 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 1230 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv10,_mm_mul_ps(velec,dsw)) ); |
| 1231 | cutoff_mask = _mm_cmplt_ps(rsq10,rcutoff2); |
| 1232 | |
| 1233 | fscal = felec; |
| 1234 | |
| 1235 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 1236 | |
| 1237 | /* Calculate temporary vectorial force */ |
| 1238 | tx = _mm_mul_ps(fscal,dx10); |
| 1239 | ty = _mm_mul_ps(fscal,dy10); |
| 1240 | tz = _mm_mul_ps(fscal,dz10); |
| 1241 | |
| 1242 | /* Update vectorial force */ |
| 1243 | fix1 = _mm_add_ps(fix1,tx); |
| 1244 | fiy1 = _mm_add_ps(fiy1,ty); |
| 1245 | fiz1 = _mm_add_ps(fiz1,tz); |
| 1246 | |
| 1247 | fjx0 = _mm_add_ps(fjx0,tx); |
| 1248 | fjy0 = _mm_add_ps(fjy0,ty); |
| 1249 | fjz0 = _mm_add_ps(fjz0,tz); |
| 1250 | |
| 1251 | } |
| 1252 | |
| 1253 | /************************** |
| 1254 | * CALCULATE INTERACTIONS * |
| 1255 | **************************/ |
| 1256 | |
| 1257 | if (gmx_mm_any_lt(rsq20,rcutoff2)) |
| 1258 | { |
| 1259 | |
| 1260 | r20 = _mm_mul_ps(rsq20,rinv20); |
| 1261 | |
| 1262 | /* Compute parameters for interactions between i and j atoms */ |
| 1263 | qq20 = _mm_mul_ps(iq2,jq0); |
| 1264 | |
| 1265 | /* EWALD ELECTROSTATICS */ |
| 1266 | |
| 1267 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 1268 | ewrt = _mm_mul_ps(r20,ewtabscale); |
| 1269 | ewitab = _mm_cvttps_epi32(ewrt); |
| 1270 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 1271 | ewitab = _mm_slli_epi32(ewitab,2); |
| 1272 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 1273 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 1274 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 1275 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 1276 | _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); |
| 1277 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 1278 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 1279 | velec = _mm_mul_ps(qq20,_mm_sub_ps(rinv20,velec)); |
| 1280 | felec = _mm_mul_ps(_mm_mul_ps(qq20,rinv20),_mm_sub_ps(rinvsq20,felec)); |
| 1281 | |
| 1282 | d = _mm_sub_ps(r20,rswitch); |
| 1283 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 1284 | d2 = _mm_mul_ps(d,d); |
| 1285 | 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))))))); |
| 1286 | |
| 1287 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 1288 | |
| 1289 | /* Evaluate switch function */ |
| 1290 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 1291 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv20,_mm_mul_ps(velec,dsw)) ); |
| 1292 | cutoff_mask = _mm_cmplt_ps(rsq20,rcutoff2); |
| 1293 | |
| 1294 | fscal = felec; |
| 1295 | |
| 1296 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 1297 | |
| 1298 | /* Calculate temporary vectorial force */ |
| 1299 | tx = _mm_mul_ps(fscal,dx20); |
| 1300 | ty = _mm_mul_ps(fscal,dy20); |
| 1301 | tz = _mm_mul_ps(fscal,dz20); |
| 1302 | |
| 1303 | /* Update vectorial force */ |
| 1304 | fix2 = _mm_add_ps(fix2,tx); |
| 1305 | fiy2 = _mm_add_ps(fiy2,ty); |
| 1306 | fiz2 = _mm_add_ps(fiz2,tz); |
| 1307 | |
| 1308 | fjx0 = _mm_add_ps(fjx0,tx); |
| 1309 | fjy0 = _mm_add_ps(fjy0,ty); |
| 1310 | fjz0 = _mm_add_ps(fjz0,tz); |
| 1311 | |
| 1312 | } |
| 1313 | |
| 1314 | /************************** |
| 1315 | * CALCULATE INTERACTIONS * |
| 1316 | **************************/ |
| 1317 | |
| 1318 | if (gmx_mm_any_lt(rsq30,rcutoff2)) |
| 1319 | { |
| 1320 | |
| 1321 | r30 = _mm_mul_ps(rsq30,rinv30); |
| 1322 | |
| 1323 | /* Compute parameters for interactions between i and j atoms */ |
| 1324 | qq30 = _mm_mul_ps(iq3,jq0); |
| 1325 | |
| 1326 | /* EWALD ELECTROSTATICS */ |
| 1327 | |
| 1328 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 1329 | ewrt = _mm_mul_ps(r30,ewtabscale); |
| 1330 | ewitab = _mm_cvttps_epi32(ewrt); |
| 1331 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 1332 | ewitab = _mm_slli_epi32(ewitab,2); |
| 1333 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 1334 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 1335 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 1336 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 1337 | _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); |
| 1338 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 1339 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 1340 | velec = _mm_mul_ps(qq30,_mm_sub_ps(rinv30,velec)); |
| 1341 | felec = _mm_mul_ps(_mm_mul_ps(qq30,rinv30),_mm_sub_ps(rinvsq30,felec)); |
| 1342 | |
| 1343 | d = _mm_sub_ps(r30,rswitch); |
| 1344 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 1345 | d2 = _mm_mul_ps(d,d); |
| 1346 | 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))))))); |
| 1347 | |
| 1348 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 1349 | |
| 1350 | /* Evaluate switch function */ |
| 1351 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 1352 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv30,_mm_mul_ps(velec,dsw)) ); |
| 1353 | cutoff_mask = _mm_cmplt_ps(rsq30,rcutoff2); |
| 1354 | |
| 1355 | fscal = felec; |
| 1356 | |
| 1357 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 1358 | |
| 1359 | /* Calculate temporary vectorial force */ |
| 1360 | tx = _mm_mul_ps(fscal,dx30); |
| 1361 | ty = _mm_mul_ps(fscal,dy30); |
| 1362 | tz = _mm_mul_ps(fscal,dz30); |
| 1363 | |
| 1364 | /* Update vectorial force */ |
| 1365 | fix3 = _mm_add_ps(fix3,tx); |
| 1366 | fiy3 = _mm_add_ps(fiy3,ty); |
| 1367 | fiz3 = _mm_add_ps(fiz3,tz); |
| 1368 | |
| 1369 | fjx0 = _mm_add_ps(fjx0,tx); |
| 1370 | fjy0 = _mm_add_ps(fjy0,ty); |
| 1371 | fjz0 = _mm_add_ps(fjz0,tz); |
| 1372 | |
| 1373 | } |
| 1374 | |
| 1375 | fjptrA = f+j_coord_offsetA; |
| 1376 | fjptrB = f+j_coord_offsetB; |
| 1377 | fjptrC = f+j_coord_offsetC; |
| 1378 | fjptrD = f+j_coord_offsetD; |
| 1379 | |
| 1380 | gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjx0,fjy0,fjz0); |
| 1381 | |
| 1382 | /* Inner loop uses 242 flops */ |
| 1383 | } |
| 1384 | |
| 1385 | if(jidx<j_index_end) |
| 1386 | { |
| 1387 | |
| 1388 | /* Get j neighbor index, and coordinate index */ |
| 1389 | jnrlistA = jjnr[jidx]; |
| 1390 | jnrlistB = jjnr[jidx+1]; |
| 1391 | jnrlistC = jjnr[jidx+2]; |
| 1392 | jnrlistD = jjnr[jidx+3]; |
| 1393 | /* Sign of each element will be negative for non-real atoms. |
| 1394 | * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones, |
| 1395 | * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries. |
| 1396 | */ |
| 1397 | dummy_mask = gmx_mm_castsi128_ps_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())); |
| 1398 | jnrA = (jnrlistA>=0) ? jnrlistA : 0; |
| 1399 | jnrB = (jnrlistB>=0) ? jnrlistB : 0; |
| 1400 | jnrC = (jnrlistC>=0) ? jnrlistC : 0; |
| 1401 | jnrD = (jnrlistD>=0) ? jnrlistD : 0; |
| 1402 | j_coord_offsetA = DIM3*jnrA; |
| 1403 | j_coord_offsetB = DIM3*jnrB; |
| 1404 | j_coord_offsetC = DIM3*jnrC; |
| 1405 | j_coord_offsetD = DIM3*jnrD; |
| 1406 | |
| 1407 | /* load j atom coordinates */ |
| 1408 | gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB, |
| 1409 | x+j_coord_offsetC,x+j_coord_offsetD, |
| 1410 | &jx0,&jy0,&jz0); |
| 1411 | |
| 1412 | /* Calculate displacement vector */ |
| 1413 | dx00 = _mm_sub_ps(ix0,jx0); |
| 1414 | dy00 = _mm_sub_ps(iy0,jy0); |
| 1415 | dz00 = _mm_sub_ps(iz0,jz0); |
| 1416 | dx10 = _mm_sub_ps(ix1,jx0); |
| 1417 | dy10 = _mm_sub_ps(iy1,jy0); |
| 1418 | dz10 = _mm_sub_ps(iz1,jz0); |
| 1419 | dx20 = _mm_sub_ps(ix2,jx0); |
| 1420 | dy20 = _mm_sub_ps(iy2,jy0); |
| 1421 | dz20 = _mm_sub_ps(iz2,jz0); |
| 1422 | dx30 = _mm_sub_ps(ix3,jx0); |
| 1423 | dy30 = _mm_sub_ps(iy3,jy0); |
| 1424 | dz30 = _mm_sub_ps(iz3,jz0); |
| 1425 | |
| 1426 | /* Calculate squared distance and things based on it */ |
| 1427 | rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00); |
| 1428 | rsq10 = gmx_mm_calc_rsq_ps(dx10,dy10,dz10); |
| 1429 | rsq20 = gmx_mm_calc_rsq_ps(dx20,dy20,dz20); |
| 1430 | rsq30 = gmx_mm_calc_rsq_ps(dx30,dy30,dz30); |
| 1431 | |
| 1432 | rinv00 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq00); |
| 1433 | rinv10 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq10); |
| 1434 | rinv20 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq20); |
| 1435 | rinv30 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq30); |
| 1436 | |
| 1437 | rinvsq00 = _mm_mul_ps(rinv00,rinv00); |
| 1438 | rinvsq10 = _mm_mul_ps(rinv10,rinv10); |
| 1439 | rinvsq20 = _mm_mul_ps(rinv20,rinv20); |
| 1440 | rinvsq30 = _mm_mul_ps(rinv30,rinv30); |
| 1441 | |
| 1442 | /* Load parameters for j particles */ |
| 1443 | jq0 = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0, |
| 1444 | charge+jnrC+0,charge+jnrD+0); |
| 1445 | vdwjidx0A = 2*vdwtype[jnrA+0]; |
| 1446 | vdwjidx0B = 2*vdwtype[jnrB+0]; |
| 1447 | vdwjidx0C = 2*vdwtype[jnrC+0]; |
| 1448 | vdwjidx0D = 2*vdwtype[jnrD+0]; |
| 1449 | |
| 1450 | fjx0 = _mm_setzero_ps(); |
| 1451 | fjy0 = _mm_setzero_ps(); |
| 1452 | fjz0 = _mm_setzero_ps(); |
| 1453 | |
| 1454 | /************************** |
| 1455 | * CALCULATE INTERACTIONS * |
| 1456 | **************************/ |
| 1457 | |
| 1458 | if (gmx_mm_any_lt(rsq00,rcutoff2)) |
| 1459 | { |
| 1460 | |
| 1461 | r00 = _mm_mul_ps(rsq00,rinv00); |
| 1462 | r00 = _mm_andnot_ps(dummy_mask,r00); |
| 1463 | |
| 1464 | /* Compute parameters for interactions between i and j atoms */ |
| 1465 | gmx_mm_load_4pair_swizzle_ps(vdwparam+vdwioffset0+vdwjidx0A, |
| 1466 | vdwparam+vdwioffset0+vdwjidx0B, |
| 1467 | vdwparam+vdwioffset0+vdwjidx0C, |
| 1468 | vdwparam+vdwioffset0+vdwjidx0D, |
| 1469 | &c6_00,&c12_00); |
| 1470 | |
| 1471 | /* LENNARD-JONES DISPERSION/REPULSION */ |
| 1472 | |
| 1473 | rinvsix = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00); |
| 1474 | vvdw6 = _mm_mul_ps(c6_00,rinvsix); |
| 1475 | vvdw12 = _mm_mul_ps(c12_00,_mm_mul_ps(rinvsix,rinvsix)); |
| 1476 | vvdw = _mm_sub_ps( _mm_mul_ps(vvdw12,one_twelfth) , _mm_mul_ps(vvdw6,one_sixth) ); |
| 1477 | fvdw = _mm_mul_ps(_mm_sub_ps(vvdw12,vvdw6),rinvsq00); |
| 1478 | |
| 1479 | d = _mm_sub_ps(r00,rswitch); |
| 1480 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 1481 | d2 = _mm_mul_ps(d,d); |
| 1482 | 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))))))); |
| 1483 | |
| 1484 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 1485 | |
| 1486 | /* Evaluate switch function */ |
| 1487 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 1488 | fvdw = _mm_sub_ps( _mm_mul_ps(fvdw,sw) , _mm_mul_ps(rinv00,_mm_mul_ps(vvdw,dsw)) ); |
| 1489 | cutoff_mask = _mm_cmplt_ps(rsq00,rcutoff2); |
| 1490 | |
| 1491 | fscal = fvdw; |
| 1492 | |
| 1493 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 1494 | |
| 1495 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 1496 | |
| 1497 | /* Calculate temporary vectorial force */ |
| 1498 | tx = _mm_mul_ps(fscal,dx00); |
| 1499 | ty = _mm_mul_ps(fscal,dy00); |
| 1500 | tz = _mm_mul_ps(fscal,dz00); |
| 1501 | |
| 1502 | /* Update vectorial force */ |
| 1503 | fix0 = _mm_add_ps(fix0,tx); |
| 1504 | fiy0 = _mm_add_ps(fiy0,ty); |
| 1505 | fiz0 = _mm_add_ps(fiz0,tz); |
| 1506 | |
| 1507 | fjx0 = _mm_add_ps(fjx0,tx); |
| 1508 | fjy0 = _mm_add_ps(fjy0,ty); |
| 1509 | fjz0 = _mm_add_ps(fjz0,tz); |
| 1510 | |
| 1511 | } |
| 1512 | |
| 1513 | /************************** |
| 1514 | * CALCULATE INTERACTIONS * |
| 1515 | **************************/ |
| 1516 | |
| 1517 | if (gmx_mm_any_lt(rsq10,rcutoff2)) |
| 1518 | { |
| 1519 | |
| 1520 | r10 = _mm_mul_ps(rsq10,rinv10); |
| 1521 | r10 = _mm_andnot_ps(dummy_mask,r10); |
| 1522 | |
| 1523 | /* Compute parameters for interactions between i and j atoms */ |
| 1524 | qq10 = _mm_mul_ps(iq1,jq0); |
| 1525 | |
| 1526 | /* EWALD ELECTROSTATICS */ |
| 1527 | |
| 1528 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 1529 | ewrt = _mm_mul_ps(r10,ewtabscale); |
| 1530 | ewitab = _mm_cvttps_epi32(ewrt); |
| 1531 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 1532 | ewitab = _mm_slli_epi32(ewitab,2); |
| 1533 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 1534 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 1535 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 1536 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 1537 | _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); |
| 1538 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 1539 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 1540 | velec = _mm_mul_ps(qq10,_mm_sub_ps(rinv10,velec)); |
| 1541 | felec = _mm_mul_ps(_mm_mul_ps(qq10,rinv10),_mm_sub_ps(rinvsq10,felec)); |
| 1542 | |
| 1543 | d = _mm_sub_ps(r10,rswitch); |
| 1544 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 1545 | d2 = _mm_mul_ps(d,d); |
| 1546 | 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))))))); |
| 1547 | |
| 1548 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 1549 | |
| 1550 | /* Evaluate switch function */ |
| 1551 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 1552 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv10,_mm_mul_ps(velec,dsw)) ); |
| 1553 | cutoff_mask = _mm_cmplt_ps(rsq10,rcutoff2); |
| 1554 | |
| 1555 | fscal = felec; |
| 1556 | |
| 1557 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 1558 | |
| 1559 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 1560 | |
| 1561 | /* Calculate temporary vectorial force */ |
| 1562 | tx = _mm_mul_ps(fscal,dx10); |
| 1563 | ty = _mm_mul_ps(fscal,dy10); |
| 1564 | tz = _mm_mul_ps(fscal,dz10); |
| 1565 | |
| 1566 | /* Update vectorial force */ |
| 1567 | fix1 = _mm_add_ps(fix1,tx); |
| 1568 | fiy1 = _mm_add_ps(fiy1,ty); |
| 1569 | fiz1 = _mm_add_ps(fiz1,tz); |
| 1570 | |
| 1571 | fjx0 = _mm_add_ps(fjx0,tx); |
| 1572 | fjy0 = _mm_add_ps(fjy0,ty); |
| 1573 | fjz0 = _mm_add_ps(fjz0,tz); |
| 1574 | |
| 1575 | } |
| 1576 | |
| 1577 | /************************** |
| 1578 | * CALCULATE INTERACTIONS * |
| 1579 | **************************/ |
| 1580 | |
| 1581 | if (gmx_mm_any_lt(rsq20,rcutoff2)) |
| 1582 | { |
| 1583 | |
| 1584 | r20 = _mm_mul_ps(rsq20,rinv20); |
| 1585 | r20 = _mm_andnot_ps(dummy_mask,r20); |
| 1586 | |
| 1587 | /* Compute parameters for interactions between i and j atoms */ |
| 1588 | qq20 = _mm_mul_ps(iq2,jq0); |
| 1589 | |
| 1590 | /* EWALD ELECTROSTATICS */ |
| 1591 | |
| 1592 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 1593 | ewrt = _mm_mul_ps(r20,ewtabscale); |
| 1594 | ewitab = _mm_cvttps_epi32(ewrt); |
| 1595 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 1596 | ewitab = _mm_slli_epi32(ewitab,2); |
| 1597 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 1598 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 1599 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 1600 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 1601 | _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); |
| 1602 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 1603 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 1604 | velec = _mm_mul_ps(qq20,_mm_sub_ps(rinv20,velec)); |
| 1605 | felec = _mm_mul_ps(_mm_mul_ps(qq20,rinv20),_mm_sub_ps(rinvsq20,felec)); |
| 1606 | |
| 1607 | d = _mm_sub_ps(r20,rswitch); |
| 1608 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 1609 | d2 = _mm_mul_ps(d,d); |
| 1610 | 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))))))); |
| 1611 | |
| 1612 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 1613 | |
| 1614 | /* Evaluate switch function */ |
| 1615 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 1616 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv20,_mm_mul_ps(velec,dsw)) ); |
| 1617 | cutoff_mask = _mm_cmplt_ps(rsq20,rcutoff2); |
| 1618 | |
| 1619 | fscal = felec; |
| 1620 | |
| 1621 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 1622 | |
| 1623 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 1624 | |
| 1625 | /* Calculate temporary vectorial force */ |
| 1626 | tx = _mm_mul_ps(fscal,dx20); |
| 1627 | ty = _mm_mul_ps(fscal,dy20); |
| 1628 | tz = _mm_mul_ps(fscal,dz20); |
| 1629 | |
| 1630 | /* Update vectorial force */ |
| 1631 | fix2 = _mm_add_ps(fix2,tx); |
| 1632 | fiy2 = _mm_add_ps(fiy2,ty); |
| 1633 | fiz2 = _mm_add_ps(fiz2,tz); |
| 1634 | |
| 1635 | fjx0 = _mm_add_ps(fjx0,tx); |
| 1636 | fjy0 = _mm_add_ps(fjy0,ty); |
| 1637 | fjz0 = _mm_add_ps(fjz0,tz); |
| 1638 | |
| 1639 | } |
| 1640 | |
| 1641 | /************************** |
| 1642 | * CALCULATE INTERACTIONS * |
| 1643 | **************************/ |
| 1644 | |
| 1645 | if (gmx_mm_any_lt(rsq30,rcutoff2)) |
| 1646 | { |
| 1647 | |
| 1648 | r30 = _mm_mul_ps(rsq30,rinv30); |
| 1649 | r30 = _mm_andnot_ps(dummy_mask,r30); |
| 1650 | |
| 1651 | /* Compute parameters for interactions between i and j atoms */ |
| 1652 | qq30 = _mm_mul_ps(iq3,jq0); |
| 1653 | |
| 1654 | /* EWALD ELECTROSTATICS */ |
| 1655 | |
| 1656 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 1657 | ewrt = _mm_mul_ps(r30,ewtabscale); |
| 1658 | ewitab = _mm_cvttps_epi32(ewrt); |
| 1659 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 1660 | ewitab = _mm_slli_epi32(ewitab,2); |
| 1661 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 1662 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 1663 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 1664 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 1665 | _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); |
| 1666 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 1667 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 1668 | velec = _mm_mul_ps(qq30,_mm_sub_ps(rinv30,velec)); |
| 1669 | felec = _mm_mul_ps(_mm_mul_ps(qq30,rinv30),_mm_sub_ps(rinvsq30,felec)); |
| 1670 | |
| 1671 | d = _mm_sub_ps(r30,rswitch); |
| 1672 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 1673 | d2 = _mm_mul_ps(d,d); |
| 1674 | 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))))))); |
| 1675 | |
| 1676 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 1677 | |
| 1678 | /* Evaluate switch function */ |
| 1679 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 1680 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv30,_mm_mul_ps(velec,dsw)) ); |
| 1681 | cutoff_mask = _mm_cmplt_ps(rsq30,rcutoff2); |
| 1682 | |
| 1683 | fscal = felec; |
| 1684 | |
| 1685 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 1686 | |
| 1687 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 1688 | |
| 1689 | /* Calculate temporary vectorial force */ |
| 1690 | tx = _mm_mul_ps(fscal,dx30); |
| 1691 | ty = _mm_mul_ps(fscal,dy30); |
| 1692 | tz = _mm_mul_ps(fscal,dz30); |
| 1693 | |
| 1694 | /* Update vectorial force */ |
| 1695 | fix3 = _mm_add_ps(fix3,tx); |
| 1696 | fiy3 = _mm_add_ps(fiy3,ty); |
| 1697 | fiz3 = _mm_add_ps(fiz3,tz); |
| 1698 | |
| 1699 | fjx0 = _mm_add_ps(fjx0,tx); |
| 1700 | fjy0 = _mm_add_ps(fjy0,ty); |
| 1701 | fjz0 = _mm_add_ps(fjz0,tz); |
| 1702 | |
| 1703 | } |
| 1704 | |
| 1705 | fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch; |
| 1706 | fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch; |
| 1707 | fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch; |
| 1708 | fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch; |
| 1709 | |
| 1710 | gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjx0,fjy0,fjz0); |
| 1711 | |
| 1712 | /* Inner loop uses 246 flops */ |
| 1713 | } |
| 1714 | |
| 1715 | /* End of innermost loop */ |
| 1716 | |
| 1717 | gmx_mm_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3, |
| 1718 | f+i_coord_offset,fshift+i_shift_offset); |
| 1719 | |
| 1720 | /* Increment number of inner iterations */ |
| 1721 | inneriter += j_index_end - j_index_start; |
| 1722 | |
| 1723 | /* Outer loop uses 24 flops */ |
| 1724 | } |
| 1725 | |
| 1726 | /* Increment number of outer iterations */ |
| 1727 | outeriter += nri; |
| 1728 | |
| 1729 | /* Update outer/inner flops */ |
| 1730 | |
| 1731 | inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_F,outeriter*24 + inneriter*246)(nrnb)->n[eNR_NBKERNEL_ELEC_VDW_W4_F] += outeriter*24 + inneriter *246; |
| 1732 | } |