| File: | gromacs/gmxlib/nonbonded/nb_kernel_sse4_1_single/nb_kernel_ElecEwSh_VdwLJSh_GeomW3W3_sse4_1_single.c |
| Location: | line 1570, column 5 |
| Description: | Value stored to 'jnrA' 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_ElecEwSh_VdwLJSh_GeomW3W3_VF_sse4_1_single |
| 54 | * Electrostatics interaction: Ewald |
| 55 | * VdW interaction: LennardJones |
| 56 | * Geometry: Water3-Water3 |
| 57 | * Calculate force/pot: PotentialAndForce |
| 58 | */ |
| 59 | void |
| 60 | nb_kernel_ElecEwSh_VdwLJSh_GeomW3W3_VF_sse4_1_single |
| 61 | (t_nblist * gmx_restrict nlist, |
| 62 | rvec * gmx_restrict xx, |
| 63 | rvec * gmx_restrict ff, |
| 64 | t_forcerec * gmx_restrict fr, |
| 65 | t_mdatoms * gmx_restrict mdatoms, |
| 66 | nb_kernel_data_t gmx_unused__attribute__ ((unused)) * gmx_restrict kernel_data, |
| 67 | t_nrnb * gmx_restrict nrnb) |
| 68 | { |
| 69 | /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or |
| 70 | * just 0 for non-waters. |
| 71 | * Suffixes A,B,C,D refer to j loop unrolling done with SSE, e.g. for the four different |
| 72 | * jnr indices corresponding to data put in the four positions in the SIMD register. |
| 73 | */ |
| 74 | int i_shift_offset,i_coord_offset,outeriter,inneriter; |
| 75 | int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx; |
| 76 | int jnrA,jnrB,jnrC,jnrD; |
| 77 | int jnrlistA,jnrlistB,jnrlistC,jnrlistD; |
| 78 | int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD; |
| 79 | int *iinr,*jindex,*jjnr,*shiftidx,*gid; |
| 80 | real rcutoff_scalar; |
| 81 | real *shiftvec,*fshift,*x,*f; |
| 82 | real *fjptrA,*fjptrB,*fjptrC,*fjptrD; |
| 83 | real scratch[4*DIM3]; |
| 84 | __m128 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall; |
| 85 | int vdwioffset0; |
| 86 | __m128 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0; |
| 87 | int vdwioffset1; |
| 88 | __m128 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1; |
| 89 | int vdwioffset2; |
| 90 | __m128 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2; |
| 91 | int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D; |
| 92 | __m128 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0; |
| 93 | int vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D; |
| 94 | __m128 jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1; |
| 95 | int vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D; |
| 96 | __m128 jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2; |
| 97 | __m128 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00; |
| 98 | __m128 dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01; |
| 99 | __m128 dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02; |
| 100 | __m128 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10; |
| 101 | __m128 dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11; |
| 102 | __m128 dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12; |
| 103 | __m128 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20; |
| 104 | __m128 dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21; |
| 105 | __m128 dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22; |
| 106 | __m128 velec,felec,velecsum,facel,crf,krf,krf2; |
| 107 | real *charge; |
| 108 | int nvdwtype; |
| 109 | __m128 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6; |
| 110 | int *vdwtype; |
| 111 | real *vdwparam; |
| 112 | __m128 one_sixth = _mm_set1_ps(1.0/6.0); |
| 113 | __m128 one_twelfth = _mm_set1_ps(1.0/12.0); |
| 114 | __m128i ewitab; |
| 115 | __m128 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV; |
| 116 | real *ewtab; |
| 117 | __m128 dummy_mask,cutoff_mask; |
| 118 | __m128 signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) ); |
| 119 | __m128 one = _mm_set1_ps(1.0); |
| 120 | __m128 two = _mm_set1_ps(2.0); |
| 121 | x = xx[0]; |
| 122 | f = ff[0]; |
| 123 | |
| 124 | nri = nlist->nri; |
| 125 | iinr = nlist->iinr; |
| 126 | jindex = nlist->jindex; |
| 127 | jjnr = nlist->jjnr; |
| 128 | shiftidx = nlist->shift; |
| 129 | gid = nlist->gid; |
| 130 | shiftvec = fr->shift_vec[0]; |
| 131 | fshift = fr->fshift[0]; |
| 132 | facel = _mm_set1_ps(fr->epsfac); |
| 133 | charge = mdatoms->chargeA; |
| 134 | nvdwtype = fr->ntype; |
| 135 | vdwparam = fr->nbfp; |
| 136 | vdwtype = mdatoms->typeA; |
| 137 | |
| 138 | sh_ewald = _mm_set1_ps(fr->ic->sh_ewald); |
| 139 | ewtab = fr->ic->tabq_coul_FDV0; |
| 140 | ewtabscale = _mm_set1_ps(fr->ic->tabq_scale); |
| 141 | ewtabhalfspace = _mm_set1_ps(0.5/fr->ic->tabq_scale); |
| 142 | |
| 143 | /* Setup water-specific parameters */ |
| 144 | inr = nlist->iinr[0]; |
| 145 | iq0 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+0])); |
| 146 | iq1 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1])); |
| 147 | iq2 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2])); |
| 148 | vdwioffset0 = 2*nvdwtype*vdwtype[inr+0]; |
| 149 | |
| 150 | jq0 = _mm_set1_ps(charge[inr+0]); |
| 151 | jq1 = _mm_set1_ps(charge[inr+1]); |
| 152 | jq2 = _mm_set1_ps(charge[inr+2]); |
| 153 | vdwjidx0A = 2*vdwtype[inr+0]; |
| 154 | qq00 = _mm_mul_ps(iq0,jq0); |
| 155 | c6_00 = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A]); |
| 156 | c12_00 = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A+1]); |
| 157 | qq01 = _mm_mul_ps(iq0,jq1); |
| 158 | qq02 = _mm_mul_ps(iq0,jq2); |
| 159 | qq10 = _mm_mul_ps(iq1,jq0); |
| 160 | qq11 = _mm_mul_ps(iq1,jq1); |
| 161 | qq12 = _mm_mul_ps(iq1,jq2); |
| 162 | qq20 = _mm_mul_ps(iq2,jq0); |
| 163 | qq21 = _mm_mul_ps(iq2,jq1); |
| 164 | qq22 = _mm_mul_ps(iq2,jq2); |
| 165 | |
| 166 | /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */ |
| 167 | rcutoff_scalar = fr->rcoulomb; |
| 168 | rcutoff = _mm_set1_ps(rcutoff_scalar); |
| 169 | rcutoff2 = _mm_mul_ps(rcutoff,rcutoff); |
| 170 | |
| 171 | sh_vdw_invrcut6 = _mm_set1_ps(fr->ic->sh_invrc6); |
| 172 | rvdw = _mm_set1_ps(fr->rvdw); |
| 173 | |
| 174 | /* Avoid stupid compiler warnings */ |
| 175 | jnrA = jnrB = jnrC = jnrD = 0; |
| 176 | j_coord_offsetA = 0; |
| 177 | j_coord_offsetB = 0; |
| 178 | j_coord_offsetC = 0; |
| 179 | j_coord_offsetD = 0; |
| 180 | |
| 181 | outeriter = 0; |
| 182 | inneriter = 0; |
| 183 | |
| 184 | for(iidx=0;iidx<4*DIM3;iidx++) |
| 185 | { |
| 186 | scratch[iidx] = 0.0; |
| 187 | } |
| 188 | |
| 189 | /* Start outer loop over neighborlists */ |
| 190 | for(iidx=0; iidx<nri; iidx++) |
| 191 | { |
| 192 | /* Load shift vector for this list */ |
| 193 | i_shift_offset = DIM3*shiftidx[iidx]; |
| 194 | |
| 195 | /* Load limits for loop over neighbors */ |
| 196 | j_index_start = jindex[iidx]; |
| 197 | j_index_end = jindex[iidx+1]; |
| 198 | |
| 199 | /* Get outer coordinate index */ |
| 200 | inr = iinr[iidx]; |
| 201 | i_coord_offset = DIM3*inr; |
| 202 | |
| 203 | /* Load i particle coords and add shift vector */ |
| 204 | gmx_mm_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset, |
| 205 | &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2); |
| 206 | |
| 207 | fix0 = _mm_setzero_ps(); |
| 208 | fiy0 = _mm_setzero_ps(); |
| 209 | fiz0 = _mm_setzero_ps(); |
| 210 | fix1 = _mm_setzero_ps(); |
| 211 | fiy1 = _mm_setzero_ps(); |
| 212 | fiz1 = _mm_setzero_ps(); |
| 213 | fix2 = _mm_setzero_ps(); |
| 214 | fiy2 = _mm_setzero_ps(); |
| 215 | fiz2 = _mm_setzero_ps(); |
| 216 | |
| 217 | /* Reset potential sums */ |
| 218 | velecsum = _mm_setzero_ps(); |
| 219 | vvdwsum = _mm_setzero_ps(); |
| 220 | |
| 221 | /* Start inner kernel loop */ |
| 222 | for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4) |
| 223 | { |
| 224 | |
| 225 | /* Get j neighbor index, and coordinate index */ |
| 226 | jnrA = jjnr[jidx]; |
| 227 | jnrB = jjnr[jidx+1]; |
| 228 | jnrC = jjnr[jidx+2]; |
| 229 | jnrD = jjnr[jidx+3]; |
| 230 | j_coord_offsetA = DIM3*jnrA; |
| 231 | j_coord_offsetB = DIM3*jnrB; |
| 232 | j_coord_offsetC = DIM3*jnrC; |
| 233 | j_coord_offsetD = DIM3*jnrD; |
| 234 | |
| 235 | /* load j atom coordinates */ |
| 236 | gmx_mm_load_3rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB, |
| 237 | x+j_coord_offsetC,x+j_coord_offsetD, |
| 238 | &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2); |
| 239 | |
| 240 | /* Calculate displacement vector */ |
| 241 | dx00 = _mm_sub_ps(ix0,jx0); |
| 242 | dy00 = _mm_sub_ps(iy0,jy0); |
| 243 | dz00 = _mm_sub_ps(iz0,jz0); |
| 244 | dx01 = _mm_sub_ps(ix0,jx1); |
| 245 | dy01 = _mm_sub_ps(iy0,jy1); |
| 246 | dz01 = _mm_sub_ps(iz0,jz1); |
| 247 | dx02 = _mm_sub_ps(ix0,jx2); |
| 248 | dy02 = _mm_sub_ps(iy0,jy2); |
| 249 | dz02 = _mm_sub_ps(iz0,jz2); |
| 250 | dx10 = _mm_sub_ps(ix1,jx0); |
| 251 | dy10 = _mm_sub_ps(iy1,jy0); |
| 252 | dz10 = _mm_sub_ps(iz1,jz0); |
| 253 | dx11 = _mm_sub_ps(ix1,jx1); |
| 254 | dy11 = _mm_sub_ps(iy1,jy1); |
| 255 | dz11 = _mm_sub_ps(iz1,jz1); |
| 256 | dx12 = _mm_sub_ps(ix1,jx2); |
| 257 | dy12 = _mm_sub_ps(iy1,jy2); |
| 258 | dz12 = _mm_sub_ps(iz1,jz2); |
| 259 | dx20 = _mm_sub_ps(ix2,jx0); |
| 260 | dy20 = _mm_sub_ps(iy2,jy0); |
| 261 | dz20 = _mm_sub_ps(iz2,jz0); |
| 262 | dx21 = _mm_sub_ps(ix2,jx1); |
| 263 | dy21 = _mm_sub_ps(iy2,jy1); |
| 264 | dz21 = _mm_sub_ps(iz2,jz1); |
| 265 | dx22 = _mm_sub_ps(ix2,jx2); |
| 266 | dy22 = _mm_sub_ps(iy2,jy2); |
| 267 | dz22 = _mm_sub_ps(iz2,jz2); |
| 268 | |
| 269 | /* Calculate squared distance and things based on it */ |
| 270 | rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00); |
| 271 | rsq01 = gmx_mm_calc_rsq_ps(dx01,dy01,dz01); |
| 272 | rsq02 = gmx_mm_calc_rsq_ps(dx02,dy02,dz02); |
| 273 | rsq10 = gmx_mm_calc_rsq_ps(dx10,dy10,dz10); |
| 274 | rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11); |
| 275 | rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12); |
| 276 | rsq20 = gmx_mm_calc_rsq_ps(dx20,dy20,dz20); |
| 277 | rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21); |
| 278 | rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22); |
| 279 | |
| 280 | rinv00 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq00); |
| 281 | rinv01 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq01); |
| 282 | rinv02 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq02); |
| 283 | rinv10 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq10); |
| 284 | rinv11 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq11); |
| 285 | rinv12 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq12); |
| 286 | rinv20 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq20); |
| 287 | rinv21 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq21); |
| 288 | rinv22 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq22); |
| 289 | |
| 290 | rinvsq00 = _mm_mul_ps(rinv00,rinv00); |
| 291 | rinvsq01 = _mm_mul_ps(rinv01,rinv01); |
| 292 | rinvsq02 = _mm_mul_ps(rinv02,rinv02); |
| 293 | rinvsq10 = _mm_mul_ps(rinv10,rinv10); |
| 294 | rinvsq11 = _mm_mul_ps(rinv11,rinv11); |
| 295 | rinvsq12 = _mm_mul_ps(rinv12,rinv12); |
| 296 | rinvsq20 = _mm_mul_ps(rinv20,rinv20); |
| 297 | rinvsq21 = _mm_mul_ps(rinv21,rinv21); |
| 298 | rinvsq22 = _mm_mul_ps(rinv22,rinv22); |
| 299 | |
| 300 | fjx0 = _mm_setzero_ps(); |
| 301 | fjy0 = _mm_setzero_ps(); |
| 302 | fjz0 = _mm_setzero_ps(); |
| 303 | fjx1 = _mm_setzero_ps(); |
| 304 | fjy1 = _mm_setzero_ps(); |
| 305 | fjz1 = _mm_setzero_ps(); |
| 306 | fjx2 = _mm_setzero_ps(); |
| 307 | fjy2 = _mm_setzero_ps(); |
| 308 | fjz2 = _mm_setzero_ps(); |
| 309 | |
| 310 | /************************** |
| 311 | * CALCULATE INTERACTIONS * |
| 312 | **************************/ |
| 313 | |
| 314 | if (gmx_mm_any_lt(rsq00,rcutoff2)) |
| 315 | { |
| 316 | |
| 317 | r00 = _mm_mul_ps(rsq00,rinv00); |
| 318 | |
| 319 | /* EWALD ELECTROSTATICS */ |
| 320 | |
| 321 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 322 | ewrt = _mm_mul_ps(r00,ewtabscale); |
| 323 | ewitab = _mm_cvttps_epi32(ewrt); |
| 324 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 325 | ewitab = _mm_slli_epi32(ewitab,2); |
| 326 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 327 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 328 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 329 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 330 | _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); |
| 331 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 332 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 333 | velec = _mm_mul_ps(qq00,_mm_sub_ps(_mm_sub_ps(rinv00,sh_ewald),velec)); |
| 334 | felec = _mm_mul_ps(_mm_mul_ps(qq00,rinv00),_mm_sub_ps(rinvsq00,felec)); |
| 335 | |
| 336 | /* LENNARD-JONES DISPERSION/REPULSION */ |
| 337 | |
| 338 | rinvsix = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00); |
| 339 | vvdw6 = _mm_mul_ps(c6_00,rinvsix); |
| 340 | vvdw12 = _mm_mul_ps(c12_00,_mm_mul_ps(rinvsix,rinvsix)); |
| 341 | vvdw = _mm_sub_ps(_mm_mul_ps( _mm_sub_ps(vvdw12 , _mm_mul_ps(c12_00,_mm_mul_ps(sh_vdw_invrcut6,sh_vdw_invrcut6))), one_twelfth) , |
| 342 | _mm_mul_ps( _mm_sub_ps(vvdw6,_mm_mul_ps(c6_00,sh_vdw_invrcut6)),one_sixth)); |
| 343 | fvdw = _mm_mul_ps(_mm_sub_ps(vvdw12,vvdw6),rinvsq00); |
| 344 | |
| 345 | cutoff_mask = _mm_cmplt_ps(rsq00,rcutoff2); |
| 346 | |
| 347 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 348 | velec = _mm_and_ps(velec,cutoff_mask); |
| 349 | velecsum = _mm_add_ps(velecsum,velec); |
| 350 | vvdw = _mm_and_ps(vvdw,cutoff_mask); |
| 351 | vvdwsum = _mm_add_ps(vvdwsum,vvdw); |
| 352 | |
| 353 | fscal = _mm_add_ps(felec,fvdw); |
| 354 | |
| 355 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 356 | |
| 357 | /* Calculate temporary vectorial force */ |
| 358 | tx = _mm_mul_ps(fscal,dx00); |
| 359 | ty = _mm_mul_ps(fscal,dy00); |
| 360 | tz = _mm_mul_ps(fscal,dz00); |
| 361 | |
| 362 | /* Update vectorial force */ |
| 363 | fix0 = _mm_add_ps(fix0,tx); |
| 364 | fiy0 = _mm_add_ps(fiy0,ty); |
| 365 | fiz0 = _mm_add_ps(fiz0,tz); |
| 366 | |
| 367 | fjx0 = _mm_add_ps(fjx0,tx); |
| 368 | fjy0 = _mm_add_ps(fjy0,ty); |
| 369 | fjz0 = _mm_add_ps(fjz0,tz); |
| 370 | |
| 371 | } |
| 372 | |
| 373 | /************************** |
| 374 | * CALCULATE INTERACTIONS * |
| 375 | **************************/ |
| 376 | |
| 377 | if (gmx_mm_any_lt(rsq01,rcutoff2)) |
| 378 | { |
| 379 | |
| 380 | r01 = _mm_mul_ps(rsq01,rinv01); |
| 381 | |
| 382 | /* EWALD ELECTROSTATICS */ |
| 383 | |
| 384 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 385 | ewrt = _mm_mul_ps(r01,ewtabscale); |
| 386 | ewitab = _mm_cvttps_epi32(ewrt); |
| 387 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 388 | ewitab = _mm_slli_epi32(ewitab,2); |
| 389 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 390 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 391 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 392 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 393 | _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); |
| 394 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 395 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 396 | velec = _mm_mul_ps(qq01,_mm_sub_ps(_mm_sub_ps(rinv01,sh_ewald),velec)); |
| 397 | felec = _mm_mul_ps(_mm_mul_ps(qq01,rinv01),_mm_sub_ps(rinvsq01,felec)); |
| 398 | |
| 399 | cutoff_mask = _mm_cmplt_ps(rsq01,rcutoff2); |
| 400 | |
| 401 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 402 | velec = _mm_and_ps(velec,cutoff_mask); |
| 403 | velecsum = _mm_add_ps(velecsum,velec); |
| 404 | |
| 405 | fscal = felec; |
| 406 | |
| 407 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 408 | |
| 409 | /* Calculate temporary vectorial force */ |
| 410 | tx = _mm_mul_ps(fscal,dx01); |
| 411 | ty = _mm_mul_ps(fscal,dy01); |
| 412 | tz = _mm_mul_ps(fscal,dz01); |
| 413 | |
| 414 | /* Update vectorial force */ |
| 415 | fix0 = _mm_add_ps(fix0,tx); |
| 416 | fiy0 = _mm_add_ps(fiy0,ty); |
| 417 | fiz0 = _mm_add_ps(fiz0,tz); |
| 418 | |
| 419 | fjx1 = _mm_add_ps(fjx1,tx); |
| 420 | fjy1 = _mm_add_ps(fjy1,ty); |
| 421 | fjz1 = _mm_add_ps(fjz1,tz); |
| 422 | |
| 423 | } |
| 424 | |
| 425 | /************************** |
| 426 | * CALCULATE INTERACTIONS * |
| 427 | **************************/ |
| 428 | |
| 429 | if (gmx_mm_any_lt(rsq02,rcutoff2)) |
| 430 | { |
| 431 | |
| 432 | r02 = _mm_mul_ps(rsq02,rinv02); |
| 433 | |
| 434 | /* EWALD ELECTROSTATICS */ |
| 435 | |
| 436 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 437 | ewrt = _mm_mul_ps(r02,ewtabscale); |
| 438 | ewitab = _mm_cvttps_epi32(ewrt); |
| 439 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 440 | ewitab = _mm_slli_epi32(ewitab,2); |
| 441 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 442 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 443 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 444 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 445 | _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); |
| 446 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 447 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 448 | velec = _mm_mul_ps(qq02,_mm_sub_ps(_mm_sub_ps(rinv02,sh_ewald),velec)); |
| 449 | felec = _mm_mul_ps(_mm_mul_ps(qq02,rinv02),_mm_sub_ps(rinvsq02,felec)); |
| 450 | |
| 451 | cutoff_mask = _mm_cmplt_ps(rsq02,rcutoff2); |
| 452 | |
| 453 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 454 | velec = _mm_and_ps(velec,cutoff_mask); |
| 455 | velecsum = _mm_add_ps(velecsum,velec); |
| 456 | |
| 457 | fscal = felec; |
| 458 | |
| 459 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 460 | |
| 461 | /* Calculate temporary vectorial force */ |
| 462 | tx = _mm_mul_ps(fscal,dx02); |
| 463 | ty = _mm_mul_ps(fscal,dy02); |
| 464 | tz = _mm_mul_ps(fscal,dz02); |
| 465 | |
| 466 | /* Update vectorial force */ |
| 467 | fix0 = _mm_add_ps(fix0,tx); |
| 468 | fiy0 = _mm_add_ps(fiy0,ty); |
| 469 | fiz0 = _mm_add_ps(fiz0,tz); |
| 470 | |
| 471 | fjx2 = _mm_add_ps(fjx2,tx); |
| 472 | fjy2 = _mm_add_ps(fjy2,ty); |
| 473 | fjz2 = _mm_add_ps(fjz2,tz); |
| 474 | |
| 475 | } |
| 476 | |
| 477 | /************************** |
| 478 | * CALCULATE INTERACTIONS * |
| 479 | **************************/ |
| 480 | |
| 481 | if (gmx_mm_any_lt(rsq10,rcutoff2)) |
| 482 | { |
| 483 | |
| 484 | r10 = _mm_mul_ps(rsq10,rinv10); |
| 485 | |
| 486 | /* EWALD ELECTROSTATICS */ |
| 487 | |
| 488 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 489 | ewrt = _mm_mul_ps(r10,ewtabscale); |
| 490 | ewitab = _mm_cvttps_epi32(ewrt); |
| 491 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 492 | ewitab = _mm_slli_epi32(ewitab,2); |
| 493 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 494 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 495 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 496 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 497 | _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); |
| 498 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 499 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 500 | velec = _mm_mul_ps(qq10,_mm_sub_ps(_mm_sub_ps(rinv10,sh_ewald),velec)); |
| 501 | felec = _mm_mul_ps(_mm_mul_ps(qq10,rinv10),_mm_sub_ps(rinvsq10,felec)); |
| 502 | |
| 503 | cutoff_mask = _mm_cmplt_ps(rsq10,rcutoff2); |
| 504 | |
| 505 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 506 | velec = _mm_and_ps(velec,cutoff_mask); |
| 507 | velecsum = _mm_add_ps(velecsum,velec); |
| 508 | |
| 509 | fscal = felec; |
| 510 | |
| 511 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 512 | |
| 513 | /* Calculate temporary vectorial force */ |
| 514 | tx = _mm_mul_ps(fscal,dx10); |
| 515 | ty = _mm_mul_ps(fscal,dy10); |
| 516 | tz = _mm_mul_ps(fscal,dz10); |
| 517 | |
| 518 | /* Update vectorial force */ |
| 519 | fix1 = _mm_add_ps(fix1,tx); |
| 520 | fiy1 = _mm_add_ps(fiy1,ty); |
| 521 | fiz1 = _mm_add_ps(fiz1,tz); |
| 522 | |
| 523 | fjx0 = _mm_add_ps(fjx0,tx); |
| 524 | fjy0 = _mm_add_ps(fjy0,ty); |
| 525 | fjz0 = _mm_add_ps(fjz0,tz); |
| 526 | |
| 527 | } |
| 528 | |
| 529 | /************************** |
| 530 | * CALCULATE INTERACTIONS * |
| 531 | **************************/ |
| 532 | |
| 533 | if (gmx_mm_any_lt(rsq11,rcutoff2)) |
| 534 | { |
| 535 | |
| 536 | r11 = _mm_mul_ps(rsq11,rinv11); |
| 537 | |
| 538 | /* EWALD ELECTROSTATICS */ |
| 539 | |
| 540 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 541 | ewrt = _mm_mul_ps(r11,ewtabscale); |
| 542 | ewitab = _mm_cvttps_epi32(ewrt); |
| 543 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 544 | ewitab = _mm_slli_epi32(ewitab,2); |
| 545 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 546 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 547 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 548 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 549 | _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); |
| 550 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 551 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 552 | velec = _mm_mul_ps(qq11,_mm_sub_ps(_mm_sub_ps(rinv11,sh_ewald),velec)); |
| 553 | felec = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec)); |
| 554 | |
| 555 | cutoff_mask = _mm_cmplt_ps(rsq11,rcutoff2); |
| 556 | |
| 557 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 558 | velec = _mm_and_ps(velec,cutoff_mask); |
| 559 | velecsum = _mm_add_ps(velecsum,velec); |
| 560 | |
| 561 | fscal = felec; |
| 562 | |
| 563 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 564 | |
| 565 | /* Calculate temporary vectorial force */ |
| 566 | tx = _mm_mul_ps(fscal,dx11); |
| 567 | ty = _mm_mul_ps(fscal,dy11); |
| 568 | tz = _mm_mul_ps(fscal,dz11); |
| 569 | |
| 570 | /* Update vectorial force */ |
| 571 | fix1 = _mm_add_ps(fix1,tx); |
| 572 | fiy1 = _mm_add_ps(fiy1,ty); |
| 573 | fiz1 = _mm_add_ps(fiz1,tz); |
| 574 | |
| 575 | fjx1 = _mm_add_ps(fjx1,tx); |
| 576 | fjy1 = _mm_add_ps(fjy1,ty); |
| 577 | fjz1 = _mm_add_ps(fjz1,tz); |
| 578 | |
| 579 | } |
| 580 | |
| 581 | /************************** |
| 582 | * CALCULATE INTERACTIONS * |
| 583 | **************************/ |
| 584 | |
| 585 | if (gmx_mm_any_lt(rsq12,rcutoff2)) |
| 586 | { |
| 587 | |
| 588 | r12 = _mm_mul_ps(rsq12,rinv12); |
| 589 | |
| 590 | /* EWALD ELECTROSTATICS */ |
| 591 | |
| 592 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 593 | ewrt = _mm_mul_ps(r12,ewtabscale); |
| 594 | ewitab = _mm_cvttps_epi32(ewrt); |
| 595 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 596 | ewitab = _mm_slli_epi32(ewitab,2); |
| 597 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 598 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 599 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 600 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 601 | _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); |
| 602 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 603 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 604 | velec = _mm_mul_ps(qq12,_mm_sub_ps(_mm_sub_ps(rinv12,sh_ewald),velec)); |
| 605 | felec = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec)); |
| 606 | |
| 607 | cutoff_mask = _mm_cmplt_ps(rsq12,rcutoff2); |
| 608 | |
| 609 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 610 | velec = _mm_and_ps(velec,cutoff_mask); |
| 611 | velecsum = _mm_add_ps(velecsum,velec); |
| 612 | |
| 613 | fscal = felec; |
| 614 | |
| 615 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 616 | |
| 617 | /* Calculate temporary vectorial force */ |
| 618 | tx = _mm_mul_ps(fscal,dx12); |
| 619 | ty = _mm_mul_ps(fscal,dy12); |
| 620 | tz = _mm_mul_ps(fscal,dz12); |
| 621 | |
| 622 | /* Update vectorial force */ |
| 623 | fix1 = _mm_add_ps(fix1,tx); |
| 624 | fiy1 = _mm_add_ps(fiy1,ty); |
| 625 | fiz1 = _mm_add_ps(fiz1,tz); |
| 626 | |
| 627 | fjx2 = _mm_add_ps(fjx2,tx); |
| 628 | fjy2 = _mm_add_ps(fjy2,ty); |
| 629 | fjz2 = _mm_add_ps(fjz2,tz); |
| 630 | |
| 631 | } |
| 632 | |
| 633 | /************************** |
| 634 | * CALCULATE INTERACTIONS * |
| 635 | **************************/ |
| 636 | |
| 637 | if (gmx_mm_any_lt(rsq20,rcutoff2)) |
| 638 | { |
| 639 | |
| 640 | r20 = _mm_mul_ps(rsq20,rinv20); |
| 641 | |
| 642 | /* EWALD ELECTROSTATICS */ |
| 643 | |
| 644 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 645 | ewrt = _mm_mul_ps(r20,ewtabscale); |
| 646 | ewitab = _mm_cvttps_epi32(ewrt); |
| 647 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 648 | ewitab = _mm_slli_epi32(ewitab,2); |
| 649 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 650 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 651 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 652 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 653 | _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); |
| 654 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 655 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 656 | velec = _mm_mul_ps(qq20,_mm_sub_ps(_mm_sub_ps(rinv20,sh_ewald),velec)); |
| 657 | felec = _mm_mul_ps(_mm_mul_ps(qq20,rinv20),_mm_sub_ps(rinvsq20,felec)); |
| 658 | |
| 659 | cutoff_mask = _mm_cmplt_ps(rsq20,rcutoff2); |
| 660 | |
| 661 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 662 | velec = _mm_and_ps(velec,cutoff_mask); |
| 663 | velecsum = _mm_add_ps(velecsum,velec); |
| 664 | |
| 665 | fscal = felec; |
| 666 | |
| 667 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 668 | |
| 669 | /* Calculate temporary vectorial force */ |
| 670 | tx = _mm_mul_ps(fscal,dx20); |
| 671 | ty = _mm_mul_ps(fscal,dy20); |
| 672 | tz = _mm_mul_ps(fscal,dz20); |
| 673 | |
| 674 | /* Update vectorial force */ |
| 675 | fix2 = _mm_add_ps(fix2,tx); |
| 676 | fiy2 = _mm_add_ps(fiy2,ty); |
| 677 | fiz2 = _mm_add_ps(fiz2,tz); |
| 678 | |
| 679 | fjx0 = _mm_add_ps(fjx0,tx); |
| 680 | fjy0 = _mm_add_ps(fjy0,ty); |
| 681 | fjz0 = _mm_add_ps(fjz0,tz); |
| 682 | |
| 683 | } |
| 684 | |
| 685 | /************************** |
| 686 | * CALCULATE INTERACTIONS * |
| 687 | **************************/ |
| 688 | |
| 689 | if (gmx_mm_any_lt(rsq21,rcutoff2)) |
| 690 | { |
| 691 | |
| 692 | r21 = _mm_mul_ps(rsq21,rinv21); |
| 693 | |
| 694 | /* EWALD ELECTROSTATICS */ |
| 695 | |
| 696 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 697 | ewrt = _mm_mul_ps(r21,ewtabscale); |
| 698 | ewitab = _mm_cvttps_epi32(ewrt); |
| 699 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 700 | ewitab = _mm_slli_epi32(ewitab,2); |
| 701 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 702 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 703 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 704 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 705 | _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn)do { __m128 tmp3, tmp2, tmp1, tmp0; tmp0 = _mm_unpacklo_ps((ewtabF ), (ewtabD)); tmp2 = _mm_unpacklo_ps((ewtabV), (ewtabFn)); tmp1 = _mm_unpackhi_ps((ewtabF), (ewtabD)); tmp3 = _mm_unpackhi_ps ((ewtabV), (ewtabFn)); (ewtabF) = _mm_movelh_ps(tmp0, tmp2); ( ewtabD) = _mm_movehl_ps(tmp2, tmp0); (ewtabV) = _mm_movelh_ps (tmp1, tmp3); (ewtabFn) = _mm_movehl_ps(tmp3, tmp1); } while ( 0); |
| 706 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 707 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 708 | velec = _mm_mul_ps(qq21,_mm_sub_ps(_mm_sub_ps(rinv21,sh_ewald),velec)); |
| 709 | felec = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec)); |
| 710 | |
| 711 | cutoff_mask = _mm_cmplt_ps(rsq21,rcutoff2); |
| 712 | |
| 713 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 714 | velec = _mm_and_ps(velec,cutoff_mask); |
| 715 | velecsum = _mm_add_ps(velecsum,velec); |
| 716 | |
| 717 | fscal = felec; |
| 718 | |
| 719 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 720 | |
| 721 | /* Calculate temporary vectorial force */ |
| 722 | tx = _mm_mul_ps(fscal,dx21); |
| 723 | ty = _mm_mul_ps(fscal,dy21); |
| 724 | tz = _mm_mul_ps(fscal,dz21); |
| 725 | |
| 726 | /* Update vectorial force */ |
| 727 | fix2 = _mm_add_ps(fix2,tx); |
| 728 | fiy2 = _mm_add_ps(fiy2,ty); |
| 729 | fiz2 = _mm_add_ps(fiz2,tz); |
| 730 | |
| 731 | fjx1 = _mm_add_ps(fjx1,tx); |
| 732 | fjy1 = _mm_add_ps(fjy1,ty); |
| 733 | fjz1 = _mm_add_ps(fjz1,tz); |
| 734 | |
| 735 | } |
| 736 | |
| 737 | /************************** |
| 738 | * CALCULATE INTERACTIONS * |
| 739 | **************************/ |
| 740 | |
| 741 | if (gmx_mm_any_lt(rsq22,rcutoff2)) |
| 742 | { |
| 743 | |
| 744 | r22 = _mm_mul_ps(rsq22,rinv22); |
| 745 | |
| 746 | /* EWALD ELECTROSTATICS */ |
| 747 | |
| 748 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 749 | ewrt = _mm_mul_ps(r22,ewtabscale); |
| 750 | ewitab = _mm_cvttps_epi32(ewrt); |
| 751 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 752 | ewitab = _mm_slli_epi32(ewitab,2); |
| 753 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 754 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 755 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 756 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 757 | _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); |
| 758 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 759 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 760 | velec = _mm_mul_ps(qq22,_mm_sub_ps(_mm_sub_ps(rinv22,sh_ewald),velec)); |
| 761 | felec = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec)); |
| 762 | |
| 763 | cutoff_mask = _mm_cmplt_ps(rsq22,rcutoff2); |
| 764 | |
| 765 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 766 | velec = _mm_and_ps(velec,cutoff_mask); |
| 767 | velecsum = _mm_add_ps(velecsum,velec); |
| 768 | |
| 769 | fscal = felec; |
| 770 | |
| 771 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 772 | |
| 773 | /* Calculate temporary vectorial force */ |
| 774 | tx = _mm_mul_ps(fscal,dx22); |
| 775 | ty = _mm_mul_ps(fscal,dy22); |
| 776 | tz = _mm_mul_ps(fscal,dz22); |
| 777 | |
| 778 | /* Update vectorial force */ |
| 779 | fix2 = _mm_add_ps(fix2,tx); |
| 780 | fiy2 = _mm_add_ps(fiy2,ty); |
| 781 | fiz2 = _mm_add_ps(fiz2,tz); |
| 782 | |
| 783 | fjx2 = _mm_add_ps(fjx2,tx); |
| 784 | fjy2 = _mm_add_ps(fjy2,ty); |
| 785 | fjz2 = _mm_add_ps(fjz2,tz); |
| 786 | |
| 787 | } |
| 788 | |
| 789 | fjptrA = f+j_coord_offsetA; |
| 790 | fjptrB = f+j_coord_offsetB; |
| 791 | fjptrC = f+j_coord_offsetC; |
| 792 | fjptrD = f+j_coord_offsetD; |
| 793 | |
| 794 | gmx_mm_decrement_3rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD, |
| 795 | fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2); |
| 796 | |
| 797 | /* Inner loop uses 432 flops */ |
| 798 | } |
| 799 | |
| 800 | if(jidx<j_index_end) |
| 801 | { |
| 802 | |
| 803 | /* Get j neighbor index, and coordinate index */ |
| 804 | jnrlistA = jjnr[jidx]; |
| 805 | jnrlistB = jjnr[jidx+1]; |
| 806 | jnrlistC = jjnr[jidx+2]; |
| 807 | jnrlistD = jjnr[jidx+3]; |
| 808 | /* Sign of each element will be negative for non-real atoms. |
| 809 | * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones, |
| 810 | * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries. |
| 811 | */ |
| 812 | dummy_mask = gmx_mm_castsi128_ps_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())); |
| 813 | jnrA = (jnrlistA>=0) ? jnrlistA : 0; |
| 814 | jnrB = (jnrlistB>=0) ? jnrlistB : 0; |
| 815 | jnrC = (jnrlistC>=0) ? jnrlistC : 0; |
| 816 | jnrD = (jnrlistD>=0) ? jnrlistD : 0; |
| 817 | j_coord_offsetA = DIM3*jnrA; |
| 818 | j_coord_offsetB = DIM3*jnrB; |
| 819 | j_coord_offsetC = DIM3*jnrC; |
| 820 | j_coord_offsetD = DIM3*jnrD; |
| 821 | |
| 822 | /* load j atom coordinates */ |
| 823 | gmx_mm_load_3rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB, |
| 824 | x+j_coord_offsetC,x+j_coord_offsetD, |
| 825 | &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2); |
| 826 | |
| 827 | /* Calculate displacement vector */ |
| 828 | dx00 = _mm_sub_ps(ix0,jx0); |
| 829 | dy00 = _mm_sub_ps(iy0,jy0); |
| 830 | dz00 = _mm_sub_ps(iz0,jz0); |
| 831 | dx01 = _mm_sub_ps(ix0,jx1); |
| 832 | dy01 = _mm_sub_ps(iy0,jy1); |
| 833 | dz01 = _mm_sub_ps(iz0,jz1); |
| 834 | dx02 = _mm_sub_ps(ix0,jx2); |
| 835 | dy02 = _mm_sub_ps(iy0,jy2); |
| 836 | dz02 = _mm_sub_ps(iz0,jz2); |
| 837 | dx10 = _mm_sub_ps(ix1,jx0); |
| 838 | dy10 = _mm_sub_ps(iy1,jy0); |
| 839 | dz10 = _mm_sub_ps(iz1,jz0); |
| 840 | dx11 = _mm_sub_ps(ix1,jx1); |
| 841 | dy11 = _mm_sub_ps(iy1,jy1); |
| 842 | dz11 = _mm_sub_ps(iz1,jz1); |
| 843 | dx12 = _mm_sub_ps(ix1,jx2); |
| 844 | dy12 = _mm_sub_ps(iy1,jy2); |
| 845 | dz12 = _mm_sub_ps(iz1,jz2); |
| 846 | dx20 = _mm_sub_ps(ix2,jx0); |
| 847 | dy20 = _mm_sub_ps(iy2,jy0); |
| 848 | dz20 = _mm_sub_ps(iz2,jz0); |
| 849 | dx21 = _mm_sub_ps(ix2,jx1); |
| 850 | dy21 = _mm_sub_ps(iy2,jy1); |
| 851 | dz21 = _mm_sub_ps(iz2,jz1); |
| 852 | dx22 = _mm_sub_ps(ix2,jx2); |
| 853 | dy22 = _mm_sub_ps(iy2,jy2); |
| 854 | dz22 = _mm_sub_ps(iz2,jz2); |
| 855 | |
| 856 | /* Calculate squared distance and things based on it */ |
| 857 | rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00); |
| 858 | rsq01 = gmx_mm_calc_rsq_ps(dx01,dy01,dz01); |
| 859 | rsq02 = gmx_mm_calc_rsq_ps(dx02,dy02,dz02); |
| 860 | rsq10 = gmx_mm_calc_rsq_ps(dx10,dy10,dz10); |
| 861 | rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11); |
| 862 | rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12); |
| 863 | rsq20 = gmx_mm_calc_rsq_ps(dx20,dy20,dz20); |
| 864 | rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21); |
| 865 | rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22); |
| 866 | |
| 867 | rinv00 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq00); |
| 868 | rinv01 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq01); |
| 869 | rinv02 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq02); |
| 870 | rinv10 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq10); |
| 871 | rinv11 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq11); |
| 872 | rinv12 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq12); |
| 873 | rinv20 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq20); |
| 874 | rinv21 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq21); |
| 875 | rinv22 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq22); |
| 876 | |
| 877 | rinvsq00 = _mm_mul_ps(rinv00,rinv00); |
| 878 | rinvsq01 = _mm_mul_ps(rinv01,rinv01); |
| 879 | rinvsq02 = _mm_mul_ps(rinv02,rinv02); |
| 880 | rinvsq10 = _mm_mul_ps(rinv10,rinv10); |
| 881 | rinvsq11 = _mm_mul_ps(rinv11,rinv11); |
| 882 | rinvsq12 = _mm_mul_ps(rinv12,rinv12); |
| 883 | rinvsq20 = _mm_mul_ps(rinv20,rinv20); |
| 884 | rinvsq21 = _mm_mul_ps(rinv21,rinv21); |
| 885 | rinvsq22 = _mm_mul_ps(rinv22,rinv22); |
| 886 | |
| 887 | fjx0 = _mm_setzero_ps(); |
| 888 | fjy0 = _mm_setzero_ps(); |
| 889 | fjz0 = _mm_setzero_ps(); |
| 890 | fjx1 = _mm_setzero_ps(); |
| 891 | fjy1 = _mm_setzero_ps(); |
| 892 | fjz1 = _mm_setzero_ps(); |
| 893 | fjx2 = _mm_setzero_ps(); |
| 894 | fjy2 = _mm_setzero_ps(); |
| 895 | fjz2 = _mm_setzero_ps(); |
| 896 | |
| 897 | /************************** |
| 898 | * CALCULATE INTERACTIONS * |
| 899 | **************************/ |
| 900 | |
| 901 | if (gmx_mm_any_lt(rsq00,rcutoff2)) |
| 902 | { |
| 903 | |
| 904 | r00 = _mm_mul_ps(rsq00,rinv00); |
| 905 | r00 = _mm_andnot_ps(dummy_mask,r00); |
| 906 | |
| 907 | /* EWALD ELECTROSTATICS */ |
| 908 | |
| 909 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 910 | ewrt = _mm_mul_ps(r00,ewtabscale); |
| 911 | ewitab = _mm_cvttps_epi32(ewrt); |
| 912 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 913 | ewitab = _mm_slli_epi32(ewitab,2); |
| 914 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 915 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 916 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 917 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 918 | _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); |
| 919 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 920 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 921 | velec = _mm_mul_ps(qq00,_mm_sub_ps(_mm_sub_ps(rinv00,sh_ewald),velec)); |
| 922 | felec = _mm_mul_ps(_mm_mul_ps(qq00,rinv00),_mm_sub_ps(rinvsq00,felec)); |
| 923 | |
| 924 | /* LENNARD-JONES DISPERSION/REPULSION */ |
| 925 | |
| 926 | rinvsix = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00); |
| 927 | vvdw6 = _mm_mul_ps(c6_00,rinvsix); |
| 928 | vvdw12 = _mm_mul_ps(c12_00,_mm_mul_ps(rinvsix,rinvsix)); |
| 929 | vvdw = _mm_sub_ps(_mm_mul_ps( _mm_sub_ps(vvdw12 , _mm_mul_ps(c12_00,_mm_mul_ps(sh_vdw_invrcut6,sh_vdw_invrcut6))), one_twelfth) , |
| 930 | _mm_mul_ps( _mm_sub_ps(vvdw6,_mm_mul_ps(c6_00,sh_vdw_invrcut6)),one_sixth)); |
| 931 | fvdw = _mm_mul_ps(_mm_sub_ps(vvdw12,vvdw6),rinvsq00); |
| 932 | |
| 933 | cutoff_mask = _mm_cmplt_ps(rsq00,rcutoff2); |
| 934 | |
| 935 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 936 | velec = _mm_and_ps(velec,cutoff_mask); |
| 937 | velec = _mm_andnot_ps(dummy_mask,velec); |
| 938 | velecsum = _mm_add_ps(velecsum,velec); |
| 939 | vvdw = _mm_and_ps(vvdw,cutoff_mask); |
| 940 | vvdw = _mm_andnot_ps(dummy_mask,vvdw); |
| 941 | vvdwsum = _mm_add_ps(vvdwsum,vvdw); |
| 942 | |
| 943 | fscal = _mm_add_ps(felec,fvdw); |
| 944 | |
| 945 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 946 | |
| 947 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 948 | |
| 949 | /* Calculate temporary vectorial force */ |
| 950 | tx = _mm_mul_ps(fscal,dx00); |
| 951 | ty = _mm_mul_ps(fscal,dy00); |
| 952 | tz = _mm_mul_ps(fscal,dz00); |
| 953 | |
| 954 | /* Update vectorial force */ |
| 955 | fix0 = _mm_add_ps(fix0,tx); |
| 956 | fiy0 = _mm_add_ps(fiy0,ty); |
| 957 | fiz0 = _mm_add_ps(fiz0,tz); |
| 958 | |
| 959 | fjx0 = _mm_add_ps(fjx0,tx); |
| 960 | fjy0 = _mm_add_ps(fjy0,ty); |
| 961 | fjz0 = _mm_add_ps(fjz0,tz); |
| 962 | |
| 963 | } |
| 964 | |
| 965 | /************************** |
| 966 | * CALCULATE INTERACTIONS * |
| 967 | **************************/ |
| 968 | |
| 969 | if (gmx_mm_any_lt(rsq01,rcutoff2)) |
| 970 | { |
| 971 | |
| 972 | r01 = _mm_mul_ps(rsq01,rinv01); |
| 973 | r01 = _mm_andnot_ps(dummy_mask,r01); |
| 974 | |
| 975 | /* EWALD ELECTROSTATICS */ |
| 976 | |
| 977 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 978 | ewrt = _mm_mul_ps(r01,ewtabscale); |
| 979 | ewitab = _mm_cvttps_epi32(ewrt); |
| 980 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 981 | ewitab = _mm_slli_epi32(ewitab,2); |
| 982 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 983 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 984 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 985 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 986 | _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); |
| 987 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 988 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 989 | velec = _mm_mul_ps(qq01,_mm_sub_ps(_mm_sub_ps(rinv01,sh_ewald),velec)); |
| 990 | felec = _mm_mul_ps(_mm_mul_ps(qq01,rinv01),_mm_sub_ps(rinvsq01,felec)); |
| 991 | |
| 992 | cutoff_mask = _mm_cmplt_ps(rsq01,rcutoff2); |
| 993 | |
| 994 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 995 | velec = _mm_and_ps(velec,cutoff_mask); |
| 996 | velec = _mm_andnot_ps(dummy_mask,velec); |
| 997 | velecsum = _mm_add_ps(velecsum,velec); |
| 998 | |
| 999 | fscal = felec; |
| 1000 | |
| 1001 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 1002 | |
| 1003 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 1004 | |
| 1005 | /* Calculate temporary vectorial force */ |
| 1006 | tx = _mm_mul_ps(fscal,dx01); |
| 1007 | ty = _mm_mul_ps(fscal,dy01); |
| 1008 | tz = _mm_mul_ps(fscal,dz01); |
| 1009 | |
| 1010 | /* Update vectorial force */ |
| 1011 | fix0 = _mm_add_ps(fix0,tx); |
| 1012 | fiy0 = _mm_add_ps(fiy0,ty); |
| 1013 | fiz0 = _mm_add_ps(fiz0,tz); |
| 1014 | |
| 1015 | fjx1 = _mm_add_ps(fjx1,tx); |
| 1016 | fjy1 = _mm_add_ps(fjy1,ty); |
| 1017 | fjz1 = _mm_add_ps(fjz1,tz); |
| 1018 | |
| 1019 | } |
| 1020 | |
| 1021 | /************************** |
| 1022 | * CALCULATE INTERACTIONS * |
| 1023 | **************************/ |
| 1024 | |
| 1025 | if (gmx_mm_any_lt(rsq02,rcutoff2)) |
| 1026 | { |
| 1027 | |
| 1028 | r02 = _mm_mul_ps(rsq02,rinv02); |
| 1029 | r02 = _mm_andnot_ps(dummy_mask,r02); |
| 1030 | |
| 1031 | /* EWALD ELECTROSTATICS */ |
| 1032 | |
| 1033 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 1034 | ewrt = _mm_mul_ps(r02,ewtabscale); |
| 1035 | ewitab = _mm_cvttps_epi32(ewrt); |
| 1036 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 1037 | ewitab = _mm_slli_epi32(ewitab,2); |
| 1038 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 1039 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 1040 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 1041 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 1042 | _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); |
| 1043 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 1044 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 1045 | velec = _mm_mul_ps(qq02,_mm_sub_ps(_mm_sub_ps(rinv02,sh_ewald),velec)); |
| 1046 | felec = _mm_mul_ps(_mm_mul_ps(qq02,rinv02),_mm_sub_ps(rinvsq02,felec)); |
| 1047 | |
| 1048 | cutoff_mask = _mm_cmplt_ps(rsq02,rcutoff2); |
| 1049 | |
| 1050 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 1051 | velec = _mm_and_ps(velec,cutoff_mask); |
| 1052 | velec = _mm_andnot_ps(dummy_mask,velec); |
| 1053 | velecsum = _mm_add_ps(velecsum,velec); |
| 1054 | |
| 1055 | fscal = felec; |
| 1056 | |
| 1057 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 1058 | |
| 1059 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 1060 | |
| 1061 | /* Calculate temporary vectorial force */ |
| 1062 | tx = _mm_mul_ps(fscal,dx02); |
| 1063 | ty = _mm_mul_ps(fscal,dy02); |
| 1064 | tz = _mm_mul_ps(fscal,dz02); |
| 1065 | |
| 1066 | /* Update vectorial force */ |
| 1067 | fix0 = _mm_add_ps(fix0,tx); |
| 1068 | fiy0 = _mm_add_ps(fiy0,ty); |
| 1069 | fiz0 = _mm_add_ps(fiz0,tz); |
| 1070 | |
| 1071 | fjx2 = _mm_add_ps(fjx2,tx); |
| 1072 | fjy2 = _mm_add_ps(fjy2,ty); |
| 1073 | fjz2 = _mm_add_ps(fjz2,tz); |
| 1074 | |
| 1075 | } |
| 1076 | |
| 1077 | /************************** |
| 1078 | * CALCULATE INTERACTIONS * |
| 1079 | **************************/ |
| 1080 | |
| 1081 | if (gmx_mm_any_lt(rsq10,rcutoff2)) |
| 1082 | { |
| 1083 | |
| 1084 | r10 = _mm_mul_ps(rsq10,rinv10); |
| 1085 | r10 = _mm_andnot_ps(dummy_mask,r10); |
| 1086 | |
| 1087 | /* EWALD ELECTROSTATICS */ |
| 1088 | |
| 1089 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 1090 | ewrt = _mm_mul_ps(r10,ewtabscale); |
| 1091 | ewitab = _mm_cvttps_epi32(ewrt); |
| 1092 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 1093 | ewitab = _mm_slli_epi32(ewitab,2); |
| 1094 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 1095 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 1096 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 1097 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 1098 | _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); |
| 1099 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 1100 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 1101 | velec = _mm_mul_ps(qq10,_mm_sub_ps(_mm_sub_ps(rinv10,sh_ewald),velec)); |
| 1102 | felec = _mm_mul_ps(_mm_mul_ps(qq10,rinv10),_mm_sub_ps(rinvsq10,felec)); |
| 1103 | |
| 1104 | cutoff_mask = _mm_cmplt_ps(rsq10,rcutoff2); |
| 1105 | |
| 1106 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 1107 | velec = _mm_and_ps(velec,cutoff_mask); |
| 1108 | velec = _mm_andnot_ps(dummy_mask,velec); |
| 1109 | velecsum = _mm_add_ps(velecsum,velec); |
| 1110 | |
| 1111 | fscal = felec; |
| 1112 | |
| 1113 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 1114 | |
| 1115 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 1116 | |
| 1117 | /* Calculate temporary vectorial force */ |
| 1118 | tx = _mm_mul_ps(fscal,dx10); |
| 1119 | ty = _mm_mul_ps(fscal,dy10); |
| 1120 | tz = _mm_mul_ps(fscal,dz10); |
| 1121 | |
| 1122 | /* Update vectorial force */ |
| 1123 | fix1 = _mm_add_ps(fix1,tx); |
| 1124 | fiy1 = _mm_add_ps(fiy1,ty); |
| 1125 | fiz1 = _mm_add_ps(fiz1,tz); |
| 1126 | |
| 1127 | fjx0 = _mm_add_ps(fjx0,tx); |
| 1128 | fjy0 = _mm_add_ps(fjy0,ty); |
| 1129 | fjz0 = _mm_add_ps(fjz0,tz); |
| 1130 | |
| 1131 | } |
| 1132 | |
| 1133 | /************************** |
| 1134 | * CALCULATE INTERACTIONS * |
| 1135 | **************************/ |
| 1136 | |
| 1137 | if (gmx_mm_any_lt(rsq11,rcutoff2)) |
| 1138 | { |
| 1139 | |
| 1140 | r11 = _mm_mul_ps(rsq11,rinv11); |
| 1141 | r11 = _mm_andnot_ps(dummy_mask,r11); |
| 1142 | |
| 1143 | /* EWALD ELECTROSTATICS */ |
| 1144 | |
| 1145 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 1146 | ewrt = _mm_mul_ps(r11,ewtabscale); |
| 1147 | ewitab = _mm_cvttps_epi32(ewrt); |
| 1148 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 1149 | ewitab = _mm_slli_epi32(ewitab,2); |
| 1150 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 1151 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 1152 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 1153 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 1154 | _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); |
| 1155 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 1156 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 1157 | velec = _mm_mul_ps(qq11,_mm_sub_ps(_mm_sub_ps(rinv11,sh_ewald),velec)); |
| 1158 | felec = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec)); |
| 1159 | |
| 1160 | cutoff_mask = _mm_cmplt_ps(rsq11,rcutoff2); |
| 1161 | |
| 1162 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 1163 | velec = _mm_and_ps(velec,cutoff_mask); |
| 1164 | velec = _mm_andnot_ps(dummy_mask,velec); |
| 1165 | velecsum = _mm_add_ps(velecsum,velec); |
| 1166 | |
| 1167 | fscal = felec; |
| 1168 | |
| 1169 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 1170 | |
| 1171 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 1172 | |
| 1173 | /* Calculate temporary vectorial force */ |
| 1174 | tx = _mm_mul_ps(fscal,dx11); |
| 1175 | ty = _mm_mul_ps(fscal,dy11); |
| 1176 | tz = _mm_mul_ps(fscal,dz11); |
| 1177 | |
| 1178 | /* Update vectorial force */ |
| 1179 | fix1 = _mm_add_ps(fix1,tx); |
| 1180 | fiy1 = _mm_add_ps(fiy1,ty); |
| 1181 | fiz1 = _mm_add_ps(fiz1,tz); |
| 1182 | |
| 1183 | fjx1 = _mm_add_ps(fjx1,tx); |
| 1184 | fjy1 = _mm_add_ps(fjy1,ty); |
| 1185 | fjz1 = _mm_add_ps(fjz1,tz); |
| 1186 | |
| 1187 | } |
| 1188 | |
| 1189 | /************************** |
| 1190 | * CALCULATE INTERACTIONS * |
| 1191 | **************************/ |
| 1192 | |
| 1193 | if (gmx_mm_any_lt(rsq12,rcutoff2)) |
| 1194 | { |
| 1195 | |
| 1196 | r12 = _mm_mul_ps(rsq12,rinv12); |
| 1197 | r12 = _mm_andnot_ps(dummy_mask,r12); |
| 1198 | |
| 1199 | /* EWALD ELECTROSTATICS */ |
| 1200 | |
| 1201 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 1202 | ewrt = _mm_mul_ps(r12,ewtabscale); |
| 1203 | ewitab = _mm_cvttps_epi32(ewrt); |
| 1204 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 1205 | ewitab = _mm_slli_epi32(ewitab,2); |
| 1206 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 1207 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 1208 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 1209 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 1210 | _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); |
| 1211 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 1212 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 1213 | velec = _mm_mul_ps(qq12,_mm_sub_ps(_mm_sub_ps(rinv12,sh_ewald),velec)); |
| 1214 | felec = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec)); |
| 1215 | |
| 1216 | cutoff_mask = _mm_cmplt_ps(rsq12,rcutoff2); |
| 1217 | |
| 1218 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 1219 | velec = _mm_and_ps(velec,cutoff_mask); |
| 1220 | velec = _mm_andnot_ps(dummy_mask,velec); |
| 1221 | velecsum = _mm_add_ps(velecsum,velec); |
| 1222 | |
| 1223 | fscal = felec; |
| 1224 | |
| 1225 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 1226 | |
| 1227 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 1228 | |
| 1229 | /* Calculate temporary vectorial force */ |
| 1230 | tx = _mm_mul_ps(fscal,dx12); |
| 1231 | ty = _mm_mul_ps(fscal,dy12); |
| 1232 | tz = _mm_mul_ps(fscal,dz12); |
| 1233 | |
| 1234 | /* Update vectorial force */ |
| 1235 | fix1 = _mm_add_ps(fix1,tx); |
| 1236 | fiy1 = _mm_add_ps(fiy1,ty); |
| 1237 | fiz1 = _mm_add_ps(fiz1,tz); |
| 1238 | |
| 1239 | fjx2 = _mm_add_ps(fjx2,tx); |
| 1240 | fjy2 = _mm_add_ps(fjy2,ty); |
| 1241 | fjz2 = _mm_add_ps(fjz2,tz); |
| 1242 | |
| 1243 | } |
| 1244 | |
| 1245 | /************************** |
| 1246 | * CALCULATE INTERACTIONS * |
| 1247 | **************************/ |
| 1248 | |
| 1249 | if (gmx_mm_any_lt(rsq20,rcutoff2)) |
| 1250 | { |
| 1251 | |
| 1252 | r20 = _mm_mul_ps(rsq20,rinv20); |
| 1253 | r20 = _mm_andnot_ps(dummy_mask,r20); |
| 1254 | |
| 1255 | /* EWALD ELECTROSTATICS */ |
| 1256 | |
| 1257 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 1258 | ewrt = _mm_mul_ps(r20,ewtabscale); |
| 1259 | ewitab = _mm_cvttps_epi32(ewrt); |
| 1260 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 1261 | ewitab = _mm_slli_epi32(ewitab,2); |
| 1262 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 1263 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 1264 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 1265 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 1266 | _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); |
| 1267 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 1268 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 1269 | velec = _mm_mul_ps(qq20,_mm_sub_ps(_mm_sub_ps(rinv20,sh_ewald),velec)); |
| 1270 | felec = _mm_mul_ps(_mm_mul_ps(qq20,rinv20),_mm_sub_ps(rinvsq20,felec)); |
| 1271 | |
| 1272 | cutoff_mask = _mm_cmplt_ps(rsq20,rcutoff2); |
| 1273 | |
| 1274 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 1275 | velec = _mm_and_ps(velec,cutoff_mask); |
| 1276 | velec = _mm_andnot_ps(dummy_mask,velec); |
| 1277 | velecsum = _mm_add_ps(velecsum,velec); |
| 1278 | |
| 1279 | fscal = felec; |
| 1280 | |
| 1281 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 1282 | |
| 1283 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 1284 | |
| 1285 | /* Calculate temporary vectorial force */ |
| 1286 | tx = _mm_mul_ps(fscal,dx20); |
| 1287 | ty = _mm_mul_ps(fscal,dy20); |
| 1288 | tz = _mm_mul_ps(fscal,dz20); |
| 1289 | |
| 1290 | /* Update vectorial force */ |
| 1291 | fix2 = _mm_add_ps(fix2,tx); |
| 1292 | fiy2 = _mm_add_ps(fiy2,ty); |
| 1293 | fiz2 = _mm_add_ps(fiz2,tz); |
| 1294 | |
| 1295 | fjx0 = _mm_add_ps(fjx0,tx); |
| 1296 | fjy0 = _mm_add_ps(fjy0,ty); |
| 1297 | fjz0 = _mm_add_ps(fjz0,tz); |
| 1298 | |
| 1299 | } |
| 1300 | |
| 1301 | /************************** |
| 1302 | * CALCULATE INTERACTIONS * |
| 1303 | **************************/ |
| 1304 | |
| 1305 | if (gmx_mm_any_lt(rsq21,rcutoff2)) |
| 1306 | { |
| 1307 | |
| 1308 | r21 = _mm_mul_ps(rsq21,rinv21); |
| 1309 | r21 = _mm_andnot_ps(dummy_mask,r21); |
| 1310 | |
| 1311 | /* EWALD ELECTROSTATICS */ |
| 1312 | |
| 1313 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 1314 | ewrt = _mm_mul_ps(r21,ewtabscale); |
| 1315 | ewitab = _mm_cvttps_epi32(ewrt); |
| 1316 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 1317 | ewitab = _mm_slli_epi32(ewitab,2); |
| 1318 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 1319 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 1320 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 1321 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 1322 | _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); |
| 1323 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 1324 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 1325 | velec = _mm_mul_ps(qq21,_mm_sub_ps(_mm_sub_ps(rinv21,sh_ewald),velec)); |
| 1326 | felec = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec)); |
| 1327 | |
| 1328 | cutoff_mask = _mm_cmplt_ps(rsq21,rcutoff2); |
| 1329 | |
| 1330 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 1331 | velec = _mm_and_ps(velec,cutoff_mask); |
| 1332 | velec = _mm_andnot_ps(dummy_mask,velec); |
| 1333 | velecsum = _mm_add_ps(velecsum,velec); |
| 1334 | |
| 1335 | fscal = felec; |
| 1336 | |
| 1337 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 1338 | |
| 1339 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 1340 | |
| 1341 | /* Calculate temporary vectorial force */ |
| 1342 | tx = _mm_mul_ps(fscal,dx21); |
| 1343 | ty = _mm_mul_ps(fscal,dy21); |
| 1344 | tz = _mm_mul_ps(fscal,dz21); |
| 1345 | |
| 1346 | /* Update vectorial force */ |
| 1347 | fix2 = _mm_add_ps(fix2,tx); |
| 1348 | fiy2 = _mm_add_ps(fiy2,ty); |
| 1349 | fiz2 = _mm_add_ps(fiz2,tz); |
| 1350 | |
| 1351 | fjx1 = _mm_add_ps(fjx1,tx); |
| 1352 | fjy1 = _mm_add_ps(fjy1,ty); |
| 1353 | fjz1 = _mm_add_ps(fjz1,tz); |
| 1354 | |
| 1355 | } |
| 1356 | |
| 1357 | /************************** |
| 1358 | * CALCULATE INTERACTIONS * |
| 1359 | **************************/ |
| 1360 | |
| 1361 | if (gmx_mm_any_lt(rsq22,rcutoff2)) |
| 1362 | { |
| 1363 | |
| 1364 | r22 = _mm_mul_ps(rsq22,rinv22); |
| 1365 | r22 = _mm_andnot_ps(dummy_mask,r22); |
| 1366 | |
| 1367 | /* EWALD ELECTROSTATICS */ |
| 1368 | |
| 1369 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 1370 | ewrt = _mm_mul_ps(r22,ewtabscale); |
| 1371 | ewitab = _mm_cvttps_epi32(ewrt); |
| 1372 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 1373 | ewitab = _mm_slli_epi32(ewitab,2); |
| 1374 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 1375 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 1376 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 1377 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 1378 | _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); |
| 1379 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 1380 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 1381 | velec = _mm_mul_ps(qq22,_mm_sub_ps(_mm_sub_ps(rinv22,sh_ewald),velec)); |
| 1382 | felec = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec)); |
| 1383 | |
| 1384 | cutoff_mask = _mm_cmplt_ps(rsq22,rcutoff2); |
| 1385 | |
| 1386 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 1387 | velec = _mm_and_ps(velec,cutoff_mask); |
| 1388 | velec = _mm_andnot_ps(dummy_mask,velec); |
| 1389 | velecsum = _mm_add_ps(velecsum,velec); |
| 1390 | |
| 1391 | fscal = felec; |
| 1392 | |
| 1393 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 1394 | |
| 1395 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 1396 | |
| 1397 | /* Calculate temporary vectorial force */ |
| 1398 | tx = _mm_mul_ps(fscal,dx22); |
| 1399 | ty = _mm_mul_ps(fscal,dy22); |
| 1400 | tz = _mm_mul_ps(fscal,dz22); |
| 1401 | |
| 1402 | /* Update vectorial force */ |
| 1403 | fix2 = _mm_add_ps(fix2,tx); |
| 1404 | fiy2 = _mm_add_ps(fiy2,ty); |
| 1405 | fiz2 = _mm_add_ps(fiz2,tz); |
| 1406 | |
| 1407 | fjx2 = _mm_add_ps(fjx2,tx); |
| 1408 | fjy2 = _mm_add_ps(fjy2,ty); |
| 1409 | fjz2 = _mm_add_ps(fjz2,tz); |
| 1410 | |
| 1411 | } |
| 1412 | |
| 1413 | fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch; |
| 1414 | fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch; |
| 1415 | fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch; |
| 1416 | fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch; |
| 1417 | |
| 1418 | gmx_mm_decrement_3rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD, |
| 1419 | fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2); |
| 1420 | |
| 1421 | /* Inner loop uses 441 flops */ |
| 1422 | } |
| 1423 | |
| 1424 | /* End of innermost loop */ |
| 1425 | |
| 1426 | gmx_mm_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2, |
| 1427 | f+i_coord_offset,fshift+i_shift_offset); |
| 1428 | |
| 1429 | ggid = gid[iidx]; |
| 1430 | /* Update potential energies */ |
| 1431 | gmx_mm_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid); |
| 1432 | gmx_mm_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid); |
| 1433 | |
| 1434 | /* Increment number of inner iterations */ |
| 1435 | inneriter += j_index_end - j_index_start; |
| 1436 | |
| 1437 | /* Outer loop uses 20 flops */ |
| 1438 | } |
| 1439 | |
| 1440 | /* Increment number of outer iterations */ |
| 1441 | outeriter += nri; |
| 1442 | |
| 1443 | /* Update outer/inner flops */ |
| 1444 | |
| 1445 | inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_VF,outeriter*20 + inneriter*441)(nrnb)->n[eNR_NBKERNEL_ELEC_VDW_W3W3_VF] += outeriter*20 + inneriter*441; |
| 1446 | } |
| 1447 | /* |
| 1448 | * Gromacs nonbonded kernel: nb_kernel_ElecEwSh_VdwLJSh_GeomW3W3_F_sse4_1_single |
| 1449 | * Electrostatics interaction: Ewald |
| 1450 | * VdW interaction: LennardJones |
| 1451 | * Geometry: Water3-Water3 |
| 1452 | * Calculate force/pot: Force |
| 1453 | */ |
| 1454 | void |
| 1455 | nb_kernel_ElecEwSh_VdwLJSh_GeomW3W3_F_sse4_1_single |
| 1456 | (t_nblist * gmx_restrict nlist, |
| 1457 | rvec * gmx_restrict xx, |
| 1458 | rvec * gmx_restrict ff, |
| 1459 | t_forcerec * gmx_restrict fr, |
| 1460 | t_mdatoms * gmx_restrict mdatoms, |
| 1461 | nb_kernel_data_t gmx_unused__attribute__ ((unused)) * gmx_restrict kernel_data, |
| 1462 | t_nrnb * gmx_restrict nrnb) |
| 1463 | { |
| 1464 | /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or |
| 1465 | * just 0 for non-waters. |
| 1466 | * Suffixes A,B,C,D refer to j loop unrolling done with SSE, e.g. for the four different |
| 1467 | * jnr indices corresponding to data put in the four positions in the SIMD register. |
| 1468 | */ |
| 1469 | int i_shift_offset,i_coord_offset,outeriter,inneriter; |
| 1470 | int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx; |
| 1471 | int jnrA,jnrB,jnrC,jnrD; |
| 1472 | int jnrlistA,jnrlistB,jnrlistC,jnrlistD; |
| 1473 | int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD; |
| 1474 | int *iinr,*jindex,*jjnr,*shiftidx,*gid; |
| 1475 | real rcutoff_scalar; |
| 1476 | real *shiftvec,*fshift,*x,*f; |
| 1477 | real *fjptrA,*fjptrB,*fjptrC,*fjptrD; |
| 1478 | real scratch[4*DIM3]; |
| 1479 | __m128 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall; |
| 1480 | int vdwioffset0; |
| 1481 | __m128 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0; |
| 1482 | int vdwioffset1; |
| 1483 | __m128 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1; |
| 1484 | int vdwioffset2; |
| 1485 | __m128 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2; |
| 1486 | int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D; |
| 1487 | __m128 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0; |
| 1488 | int vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D; |
| 1489 | __m128 jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1; |
| 1490 | int vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D; |
| 1491 | __m128 jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2; |
| 1492 | __m128 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00; |
| 1493 | __m128 dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01; |
| 1494 | __m128 dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02; |
| 1495 | __m128 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10; |
| 1496 | __m128 dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11; |
| 1497 | __m128 dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12; |
| 1498 | __m128 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20; |
| 1499 | __m128 dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21; |
| 1500 | __m128 dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22; |
| 1501 | __m128 velec,felec,velecsum,facel,crf,krf,krf2; |
| 1502 | real *charge; |
| 1503 | int nvdwtype; |
| 1504 | __m128 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6; |
| 1505 | int *vdwtype; |
| 1506 | real *vdwparam; |
| 1507 | __m128 one_sixth = _mm_set1_ps(1.0/6.0); |
| 1508 | __m128 one_twelfth = _mm_set1_ps(1.0/12.0); |
| 1509 | __m128i ewitab; |
| 1510 | __m128 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV; |
| 1511 | real *ewtab; |
| 1512 | __m128 dummy_mask,cutoff_mask; |
| 1513 | __m128 signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) ); |
| 1514 | __m128 one = _mm_set1_ps(1.0); |
| 1515 | __m128 two = _mm_set1_ps(2.0); |
| 1516 | x = xx[0]; |
| 1517 | f = ff[0]; |
| 1518 | |
| 1519 | nri = nlist->nri; |
| 1520 | iinr = nlist->iinr; |
| 1521 | jindex = nlist->jindex; |
| 1522 | jjnr = nlist->jjnr; |
| 1523 | shiftidx = nlist->shift; |
| 1524 | gid = nlist->gid; |
| 1525 | shiftvec = fr->shift_vec[0]; |
| 1526 | fshift = fr->fshift[0]; |
| 1527 | facel = _mm_set1_ps(fr->epsfac); |
| 1528 | charge = mdatoms->chargeA; |
| 1529 | nvdwtype = fr->ntype; |
| 1530 | vdwparam = fr->nbfp; |
| 1531 | vdwtype = mdatoms->typeA; |
| 1532 | |
| 1533 | sh_ewald = _mm_set1_ps(fr->ic->sh_ewald); |
| 1534 | ewtab = fr->ic->tabq_coul_F; |
| 1535 | ewtabscale = _mm_set1_ps(fr->ic->tabq_scale); |
| 1536 | ewtabhalfspace = _mm_set1_ps(0.5/fr->ic->tabq_scale); |
| 1537 | |
| 1538 | /* Setup water-specific parameters */ |
| 1539 | inr = nlist->iinr[0]; |
| 1540 | iq0 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+0])); |
| 1541 | iq1 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1])); |
| 1542 | iq2 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2])); |
| 1543 | vdwioffset0 = 2*nvdwtype*vdwtype[inr+0]; |
| 1544 | |
| 1545 | jq0 = _mm_set1_ps(charge[inr+0]); |
| 1546 | jq1 = _mm_set1_ps(charge[inr+1]); |
| 1547 | jq2 = _mm_set1_ps(charge[inr+2]); |
| 1548 | vdwjidx0A = 2*vdwtype[inr+0]; |
| 1549 | qq00 = _mm_mul_ps(iq0,jq0); |
| 1550 | c6_00 = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A]); |
| 1551 | c12_00 = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A+1]); |
| 1552 | qq01 = _mm_mul_ps(iq0,jq1); |
| 1553 | qq02 = _mm_mul_ps(iq0,jq2); |
| 1554 | qq10 = _mm_mul_ps(iq1,jq0); |
| 1555 | qq11 = _mm_mul_ps(iq1,jq1); |
| 1556 | qq12 = _mm_mul_ps(iq1,jq2); |
| 1557 | qq20 = _mm_mul_ps(iq2,jq0); |
| 1558 | qq21 = _mm_mul_ps(iq2,jq1); |
| 1559 | qq22 = _mm_mul_ps(iq2,jq2); |
| 1560 | |
| 1561 | /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */ |
| 1562 | rcutoff_scalar = fr->rcoulomb; |
| 1563 | rcutoff = _mm_set1_ps(rcutoff_scalar); |
| 1564 | rcutoff2 = _mm_mul_ps(rcutoff,rcutoff); |
| 1565 | |
| 1566 | sh_vdw_invrcut6 = _mm_set1_ps(fr->ic->sh_invrc6); |
| 1567 | rvdw = _mm_set1_ps(fr->rvdw); |
| 1568 | |
| 1569 | /* Avoid stupid compiler warnings */ |
| 1570 | jnrA = jnrB = jnrC = jnrD = 0; |
Value stored to 'jnrA' is never read | |
| 1571 | j_coord_offsetA = 0; |
| 1572 | j_coord_offsetB = 0; |
| 1573 | j_coord_offsetC = 0; |
| 1574 | j_coord_offsetD = 0; |
| 1575 | |
| 1576 | outeriter = 0; |
| 1577 | inneriter = 0; |
| 1578 | |
| 1579 | for(iidx=0;iidx<4*DIM3;iidx++) |
| 1580 | { |
| 1581 | scratch[iidx] = 0.0; |
| 1582 | } |
| 1583 | |
| 1584 | /* Start outer loop over neighborlists */ |
| 1585 | for(iidx=0; iidx<nri; iidx++) |
| 1586 | { |
| 1587 | /* Load shift vector for this list */ |
| 1588 | i_shift_offset = DIM3*shiftidx[iidx]; |
| 1589 | |
| 1590 | /* Load limits for loop over neighbors */ |
| 1591 | j_index_start = jindex[iidx]; |
| 1592 | j_index_end = jindex[iidx+1]; |
| 1593 | |
| 1594 | /* Get outer coordinate index */ |
| 1595 | inr = iinr[iidx]; |
| 1596 | i_coord_offset = DIM3*inr; |
| 1597 | |
| 1598 | /* Load i particle coords and add shift vector */ |
| 1599 | gmx_mm_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset, |
| 1600 | &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2); |
| 1601 | |
| 1602 | fix0 = _mm_setzero_ps(); |
| 1603 | fiy0 = _mm_setzero_ps(); |
| 1604 | fiz0 = _mm_setzero_ps(); |
| 1605 | fix1 = _mm_setzero_ps(); |
| 1606 | fiy1 = _mm_setzero_ps(); |
| 1607 | fiz1 = _mm_setzero_ps(); |
| 1608 | fix2 = _mm_setzero_ps(); |
| 1609 | fiy2 = _mm_setzero_ps(); |
| 1610 | fiz2 = _mm_setzero_ps(); |
| 1611 | |
| 1612 | /* Start inner kernel loop */ |
| 1613 | for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4) |
| 1614 | { |
| 1615 | |
| 1616 | /* Get j neighbor index, and coordinate index */ |
| 1617 | jnrA = jjnr[jidx]; |
| 1618 | jnrB = jjnr[jidx+1]; |
| 1619 | jnrC = jjnr[jidx+2]; |
| 1620 | jnrD = jjnr[jidx+3]; |
| 1621 | j_coord_offsetA = DIM3*jnrA; |
| 1622 | j_coord_offsetB = DIM3*jnrB; |
| 1623 | j_coord_offsetC = DIM3*jnrC; |
| 1624 | j_coord_offsetD = DIM3*jnrD; |
| 1625 | |
| 1626 | /* load j atom coordinates */ |
| 1627 | gmx_mm_load_3rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB, |
| 1628 | x+j_coord_offsetC,x+j_coord_offsetD, |
| 1629 | &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2); |
| 1630 | |
| 1631 | /* Calculate displacement vector */ |
| 1632 | dx00 = _mm_sub_ps(ix0,jx0); |
| 1633 | dy00 = _mm_sub_ps(iy0,jy0); |
| 1634 | dz00 = _mm_sub_ps(iz0,jz0); |
| 1635 | dx01 = _mm_sub_ps(ix0,jx1); |
| 1636 | dy01 = _mm_sub_ps(iy0,jy1); |
| 1637 | dz01 = _mm_sub_ps(iz0,jz1); |
| 1638 | dx02 = _mm_sub_ps(ix0,jx2); |
| 1639 | dy02 = _mm_sub_ps(iy0,jy2); |
| 1640 | dz02 = _mm_sub_ps(iz0,jz2); |
| 1641 | dx10 = _mm_sub_ps(ix1,jx0); |
| 1642 | dy10 = _mm_sub_ps(iy1,jy0); |
| 1643 | dz10 = _mm_sub_ps(iz1,jz0); |
| 1644 | dx11 = _mm_sub_ps(ix1,jx1); |
| 1645 | dy11 = _mm_sub_ps(iy1,jy1); |
| 1646 | dz11 = _mm_sub_ps(iz1,jz1); |
| 1647 | dx12 = _mm_sub_ps(ix1,jx2); |
| 1648 | dy12 = _mm_sub_ps(iy1,jy2); |
| 1649 | dz12 = _mm_sub_ps(iz1,jz2); |
| 1650 | dx20 = _mm_sub_ps(ix2,jx0); |
| 1651 | dy20 = _mm_sub_ps(iy2,jy0); |
| 1652 | dz20 = _mm_sub_ps(iz2,jz0); |
| 1653 | dx21 = _mm_sub_ps(ix2,jx1); |
| 1654 | dy21 = _mm_sub_ps(iy2,jy1); |
| 1655 | dz21 = _mm_sub_ps(iz2,jz1); |
| 1656 | dx22 = _mm_sub_ps(ix2,jx2); |
| 1657 | dy22 = _mm_sub_ps(iy2,jy2); |
| 1658 | dz22 = _mm_sub_ps(iz2,jz2); |
| 1659 | |
| 1660 | /* Calculate squared distance and things based on it */ |
| 1661 | rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00); |
| 1662 | rsq01 = gmx_mm_calc_rsq_ps(dx01,dy01,dz01); |
| 1663 | rsq02 = gmx_mm_calc_rsq_ps(dx02,dy02,dz02); |
| 1664 | rsq10 = gmx_mm_calc_rsq_ps(dx10,dy10,dz10); |
| 1665 | rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11); |
| 1666 | rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12); |
| 1667 | rsq20 = gmx_mm_calc_rsq_ps(dx20,dy20,dz20); |
| 1668 | rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21); |
| 1669 | rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22); |
| 1670 | |
| 1671 | rinv00 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq00); |
| 1672 | rinv01 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq01); |
| 1673 | rinv02 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq02); |
| 1674 | rinv10 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq10); |
| 1675 | rinv11 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq11); |
| 1676 | rinv12 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq12); |
| 1677 | rinv20 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq20); |
| 1678 | rinv21 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq21); |
| 1679 | rinv22 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq22); |
| 1680 | |
| 1681 | rinvsq00 = _mm_mul_ps(rinv00,rinv00); |
| 1682 | rinvsq01 = _mm_mul_ps(rinv01,rinv01); |
| 1683 | rinvsq02 = _mm_mul_ps(rinv02,rinv02); |
| 1684 | rinvsq10 = _mm_mul_ps(rinv10,rinv10); |
| 1685 | rinvsq11 = _mm_mul_ps(rinv11,rinv11); |
| 1686 | rinvsq12 = _mm_mul_ps(rinv12,rinv12); |
| 1687 | rinvsq20 = _mm_mul_ps(rinv20,rinv20); |
| 1688 | rinvsq21 = _mm_mul_ps(rinv21,rinv21); |
| 1689 | rinvsq22 = _mm_mul_ps(rinv22,rinv22); |
| 1690 | |
| 1691 | fjx0 = _mm_setzero_ps(); |
| 1692 | fjy0 = _mm_setzero_ps(); |
| 1693 | fjz0 = _mm_setzero_ps(); |
| 1694 | fjx1 = _mm_setzero_ps(); |
| 1695 | fjy1 = _mm_setzero_ps(); |
| 1696 | fjz1 = _mm_setzero_ps(); |
| 1697 | fjx2 = _mm_setzero_ps(); |
| 1698 | fjy2 = _mm_setzero_ps(); |
| 1699 | fjz2 = _mm_setzero_ps(); |
| 1700 | |
| 1701 | /************************** |
| 1702 | * CALCULATE INTERACTIONS * |
| 1703 | **************************/ |
| 1704 | |
| 1705 | if (gmx_mm_any_lt(rsq00,rcutoff2)) |
| 1706 | { |
| 1707 | |
| 1708 | r00 = _mm_mul_ps(rsq00,rinv00); |
| 1709 | |
| 1710 | /* EWALD ELECTROSTATICS */ |
| 1711 | |
| 1712 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 1713 | ewrt = _mm_mul_ps(r00,ewtabscale); |
| 1714 | ewitab = _mm_cvttps_epi32(ewrt); |
| 1715 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 1716 | gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})), |
| 1717 | ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})), |
| 1718 | &ewtabF,&ewtabFn); |
| 1719 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
| 1720 | felec = _mm_mul_ps(_mm_mul_ps(qq00,rinv00),_mm_sub_ps(rinvsq00,felec)); |
| 1721 | |
| 1722 | /* LENNARD-JONES DISPERSION/REPULSION */ |
| 1723 | |
| 1724 | rinvsix = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00); |
| 1725 | fvdw = _mm_mul_ps(_mm_sub_ps(_mm_mul_ps(c12_00,rinvsix),c6_00),_mm_mul_ps(rinvsix,rinvsq00)); |
| 1726 | |
| 1727 | cutoff_mask = _mm_cmplt_ps(rsq00,rcutoff2); |
| 1728 | |
| 1729 | fscal = _mm_add_ps(felec,fvdw); |
| 1730 | |
| 1731 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 1732 | |
| 1733 | /* Calculate temporary vectorial force */ |
| 1734 | tx = _mm_mul_ps(fscal,dx00); |
| 1735 | ty = _mm_mul_ps(fscal,dy00); |
| 1736 | tz = _mm_mul_ps(fscal,dz00); |
| 1737 | |
| 1738 | /* Update vectorial force */ |
| 1739 | fix0 = _mm_add_ps(fix0,tx); |
| 1740 | fiy0 = _mm_add_ps(fiy0,ty); |
| 1741 | fiz0 = _mm_add_ps(fiz0,tz); |
| 1742 | |
| 1743 | fjx0 = _mm_add_ps(fjx0,tx); |
| 1744 | fjy0 = _mm_add_ps(fjy0,ty); |
| 1745 | fjz0 = _mm_add_ps(fjz0,tz); |
| 1746 | |
| 1747 | } |
| 1748 | |
| 1749 | /************************** |
| 1750 | * CALCULATE INTERACTIONS * |
| 1751 | **************************/ |
| 1752 | |
| 1753 | if (gmx_mm_any_lt(rsq01,rcutoff2)) |
| 1754 | { |
| 1755 | |
| 1756 | r01 = _mm_mul_ps(rsq01,rinv01); |
| 1757 | |
| 1758 | /* EWALD ELECTROSTATICS */ |
| 1759 | |
| 1760 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 1761 | ewrt = _mm_mul_ps(r01,ewtabscale); |
| 1762 | ewitab = _mm_cvttps_epi32(ewrt); |
| 1763 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 1764 | gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})), |
| 1765 | ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})), |
| 1766 | &ewtabF,&ewtabFn); |
| 1767 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
| 1768 | felec = _mm_mul_ps(_mm_mul_ps(qq01,rinv01),_mm_sub_ps(rinvsq01,felec)); |
| 1769 | |
| 1770 | cutoff_mask = _mm_cmplt_ps(rsq01,rcutoff2); |
| 1771 | |
| 1772 | fscal = felec; |
| 1773 | |
| 1774 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 1775 | |
| 1776 | /* Calculate temporary vectorial force */ |
| 1777 | tx = _mm_mul_ps(fscal,dx01); |
| 1778 | ty = _mm_mul_ps(fscal,dy01); |
| 1779 | tz = _mm_mul_ps(fscal,dz01); |
| 1780 | |
| 1781 | /* Update vectorial force */ |
| 1782 | fix0 = _mm_add_ps(fix0,tx); |
| 1783 | fiy0 = _mm_add_ps(fiy0,ty); |
| 1784 | fiz0 = _mm_add_ps(fiz0,tz); |
| 1785 | |
| 1786 | fjx1 = _mm_add_ps(fjx1,tx); |
| 1787 | fjy1 = _mm_add_ps(fjy1,ty); |
| 1788 | fjz1 = _mm_add_ps(fjz1,tz); |
| 1789 | |
| 1790 | } |
| 1791 | |
| 1792 | /************************** |
| 1793 | * CALCULATE INTERACTIONS * |
| 1794 | **************************/ |
| 1795 | |
| 1796 | if (gmx_mm_any_lt(rsq02,rcutoff2)) |
| 1797 | { |
| 1798 | |
| 1799 | r02 = _mm_mul_ps(rsq02,rinv02); |
| 1800 | |
| 1801 | /* EWALD ELECTROSTATICS */ |
| 1802 | |
| 1803 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 1804 | ewrt = _mm_mul_ps(r02,ewtabscale); |
| 1805 | ewitab = _mm_cvttps_epi32(ewrt); |
| 1806 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 1807 | gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})), |
| 1808 | ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})), |
| 1809 | &ewtabF,&ewtabFn); |
| 1810 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
| 1811 | felec = _mm_mul_ps(_mm_mul_ps(qq02,rinv02),_mm_sub_ps(rinvsq02,felec)); |
| 1812 | |
| 1813 | cutoff_mask = _mm_cmplt_ps(rsq02,rcutoff2); |
| 1814 | |
| 1815 | fscal = felec; |
| 1816 | |
| 1817 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 1818 | |
| 1819 | /* Calculate temporary vectorial force */ |
| 1820 | tx = _mm_mul_ps(fscal,dx02); |
| 1821 | ty = _mm_mul_ps(fscal,dy02); |
| 1822 | tz = _mm_mul_ps(fscal,dz02); |
| 1823 | |
| 1824 | /* Update vectorial force */ |
| 1825 | fix0 = _mm_add_ps(fix0,tx); |
| 1826 | fiy0 = _mm_add_ps(fiy0,ty); |
| 1827 | fiz0 = _mm_add_ps(fiz0,tz); |
| 1828 | |
| 1829 | fjx2 = _mm_add_ps(fjx2,tx); |
| 1830 | fjy2 = _mm_add_ps(fjy2,ty); |
| 1831 | fjz2 = _mm_add_ps(fjz2,tz); |
| 1832 | |
| 1833 | } |
| 1834 | |
| 1835 | /************************** |
| 1836 | * CALCULATE INTERACTIONS * |
| 1837 | **************************/ |
| 1838 | |
| 1839 | if (gmx_mm_any_lt(rsq10,rcutoff2)) |
| 1840 | { |
| 1841 | |
| 1842 | r10 = _mm_mul_ps(rsq10,rinv10); |
| 1843 | |
| 1844 | /* EWALD ELECTROSTATICS */ |
| 1845 | |
| 1846 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 1847 | ewrt = _mm_mul_ps(r10,ewtabscale); |
| 1848 | ewitab = _mm_cvttps_epi32(ewrt); |
| 1849 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 1850 | gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})), |
| 1851 | ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})), |
| 1852 | &ewtabF,&ewtabFn); |
| 1853 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
| 1854 | felec = _mm_mul_ps(_mm_mul_ps(qq10,rinv10),_mm_sub_ps(rinvsq10,felec)); |
| 1855 | |
| 1856 | cutoff_mask = _mm_cmplt_ps(rsq10,rcutoff2); |
| 1857 | |
| 1858 | fscal = felec; |
| 1859 | |
| 1860 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 1861 | |
| 1862 | /* Calculate temporary vectorial force */ |
| 1863 | tx = _mm_mul_ps(fscal,dx10); |
| 1864 | ty = _mm_mul_ps(fscal,dy10); |
| 1865 | tz = _mm_mul_ps(fscal,dz10); |
| 1866 | |
| 1867 | /* Update vectorial force */ |
| 1868 | fix1 = _mm_add_ps(fix1,tx); |
| 1869 | fiy1 = _mm_add_ps(fiy1,ty); |
| 1870 | fiz1 = _mm_add_ps(fiz1,tz); |
| 1871 | |
| 1872 | fjx0 = _mm_add_ps(fjx0,tx); |
| 1873 | fjy0 = _mm_add_ps(fjy0,ty); |
| 1874 | fjz0 = _mm_add_ps(fjz0,tz); |
| 1875 | |
| 1876 | } |
| 1877 | |
| 1878 | /************************** |
| 1879 | * CALCULATE INTERACTIONS * |
| 1880 | **************************/ |
| 1881 | |
| 1882 | if (gmx_mm_any_lt(rsq11,rcutoff2)) |
| 1883 | { |
| 1884 | |
| 1885 | r11 = _mm_mul_ps(rsq11,rinv11); |
| 1886 | |
| 1887 | /* EWALD ELECTROSTATICS */ |
| 1888 | |
| 1889 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 1890 | ewrt = _mm_mul_ps(r11,ewtabscale); |
| 1891 | ewitab = _mm_cvttps_epi32(ewrt); |
| 1892 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 1893 | gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})), |
| 1894 | ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})), |
| 1895 | &ewtabF,&ewtabFn); |
| 1896 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
| 1897 | felec = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec)); |
| 1898 | |
| 1899 | cutoff_mask = _mm_cmplt_ps(rsq11,rcutoff2); |
| 1900 | |
| 1901 | fscal = felec; |
| 1902 | |
| 1903 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 1904 | |
| 1905 | /* Calculate temporary vectorial force */ |
| 1906 | tx = _mm_mul_ps(fscal,dx11); |
| 1907 | ty = _mm_mul_ps(fscal,dy11); |
| 1908 | tz = _mm_mul_ps(fscal,dz11); |
| 1909 | |
| 1910 | /* Update vectorial force */ |
| 1911 | fix1 = _mm_add_ps(fix1,tx); |
| 1912 | fiy1 = _mm_add_ps(fiy1,ty); |
| 1913 | fiz1 = _mm_add_ps(fiz1,tz); |
| 1914 | |
| 1915 | fjx1 = _mm_add_ps(fjx1,tx); |
| 1916 | fjy1 = _mm_add_ps(fjy1,ty); |
| 1917 | fjz1 = _mm_add_ps(fjz1,tz); |
| 1918 | |
| 1919 | } |
| 1920 | |
| 1921 | /************************** |
| 1922 | * CALCULATE INTERACTIONS * |
| 1923 | **************************/ |
| 1924 | |
| 1925 | if (gmx_mm_any_lt(rsq12,rcutoff2)) |
| 1926 | { |
| 1927 | |
| 1928 | r12 = _mm_mul_ps(rsq12,rinv12); |
| 1929 | |
| 1930 | /* EWALD ELECTROSTATICS */ |
| 1931 | |
| 1932 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 1933 | ewrt = _mm_mul_ps(r12,ewtabscale); |
| 1934 | ewitab = _mm_cvttps_epi32(ewrt); |
| 1935 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 1936 | gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})), |
| 1937 | ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})), |
| 1938 | &ewtabF,&ewtabFn); |
| 1939 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
| 1940 | felec = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec)); |
| 1941 | |
| 1942 | cutoff_mask = _mm_cmplt_ps(rsq12,rcutoff2); |
| 1943 | |
| 1944 | fscal = felec; |
| 1945 | |
| 1946 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 1947 | |
| 1948 | /* Calculate temporary vectorial force */ |
| 1949 | tx = _mm_mul_ps(fscal,dx12); |
| 1950 | ty = _mm_mul_ps(fscal,dy12); |
| 1951 | tz = _mm_mul_ps(fscal,dz12); |
| 1952 | |
| 1953 | /* Update vectorial force */ |
| 1954 | fix1 = _mm_add_ps(fix1,tx); |
| 1955 | fiy1 = _mm_add_ps(fiy1,ty); |
| 1956 | fiz1 = _mm_add_ps(fiz1,tz); |
| 1957 | |
| 1958 | fjx2 = _mm_add_ps(fjx2,tx); |
| 1959 | fjy2 = _mm_add_ps(fjy2,ty); |
| 1960 | fjz2 = _mm_add_ps(fjz2,tz); |
| 1961 | |
| 1962 | } |
| 1963 | |
| 1964 | /************************** |
| 1965 | * CALCULATE INTERACTIONS * |
| 1966 | **************************/ |
| 1967 | |
| 1968 | if (gmx_mm_any_lt(rsq20,rcutoff2)) |
| 1969 | { |
| 1970 | |
| 1971 | r20 = _mm_mul_ps(rsq20,rinv20); |
| 1972 | |
| 1973 | /* EWALD ELECTROSTATICS */ |
| 1974 | |
| 1975 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 1976 | ewrt = _mm_mul_ps(r20,ewtabscale); |
| 1977 | ewitab = _mm_cvttps_epi32(ewrt); |
| 1978 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 1979 | gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})), |
| 1980 | ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})), |
| 1981 | &ewtabF,&ewtabFn); |
| 1982 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
| 1983 | felec = _mm_mul_ps(_mm_mul_ps(qq20,rinv20),_mm_sub_ps(rinvsq20,felec)); |
| 1984 | |
| 1985 | cutoff_mask = _mm_cmplt_ps(rsq20,rcutoff2); |
| 1986 | |
| 1987 | fscal = felec; |
| 1988 | |
| 1989 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 1990 | |
| 1991 | /* Calculate temporary vectorial force */ |
| 1992 | tx = _mm_mul_ps(fscal,dx20); |
| 1993 | ty = _mm_mul_ps(fscal,dy20); |
| 1994 | tz = _mm_mul_ps(fscal,dz20); |
| 1995 | |
| 1996 | /* Update vectorial force */ |
| 1997 | fix2 = _mm_add_ps(fix2,tx); |
| 1998 | fiy2 = _mm_add_ps(fiy2,ty); |
| 1999 | fiz2 = _mm_add_ps(fiz2,tz); |
| 2000 | |
| 2001 | fjx0 = _mm_add_ps(fjx0,tx); |
| 2002 | fjy0 = _mm_add_ps(fjy0,ty); |
| 2003 | fjz0 = _mm_add_ps(fjz0,tz); |
| 2004 | |
| 2005 | } |
| 2006 | |
| 2007 | /************************** |
| 2008 | * CALCULATE INTERACTIONS * |
| 2009 | **************************/ |
| 2010 | |
| 2011 | if (gmx_mm_any_lt(rsq21,rcutoff2)) |
| 2012 | { |
| 2013 | |
| 2014 | r21 = _mm_mul_ps(rsq21,rinv21); |
| 2015 | |
| 2016 | /* EWALD ELECTROSTATICS */ |
| 2017 | |
| 2018 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 2019 | ewrt = _mm_mul_ps(r21,ewtabscale); |
| 2020 | ewitab = _mm_cvttps_epi32(ewrt); |
| 2021 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 2022 | gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})), |
| 2023 | ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})), |
| 2024 | &ewtabF,&ewtabFn); |
| 2025 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
| 2026 | felec = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec)); |
| 2027 | |
| 2028 | cutoff_mask = _mm_cmplt_ps(rsq21,rcutoff2); |
| 2029 | |
| 2030 | fscal = felec; |
| 2031 | |
| 2032 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 2033 | |
| 2034 | /* Calculate temporary vectorial force */ |
| 2035 | tx = _mm_mul_ps(fscal,dx21); |
| 2036 | ty = _mm_mul_ps(fscal,dy21); |
| 2037 | tz = _mm_mul_ps(fscal,dz21); |
| 2038 | |
| 2039 | /* Update vectorial force */ |
| 2040 | fix2 = _mm_add_ps(fix2,tx); |
| 2041 | fiy2 = _mm_add_ps(fiy2,ty); |
| 2042 | fiz2 = _mm_add_ps(fiz2,tz); |
| 2043 | |
| 2044 | fjx1 = _mm_add_ps(fjx1,tx); |
| 2045 | fjy1 = _mm_add_ps(fjy1,ty); |
| 2046 | fjz1 = _mm_add_ps(fjz1,tz); |
| 2047 | |
| 2048 | } |
| 2049 | |
| 2050 | /************************** |
| 2051 | * CALCULATE INTERACTIONS * |
| 2052 | **************************/ |
| 2053 | |
| 2054 | if (gmx_mm_any_lt(rsq22,rcutoff2)) |
| 2055 | { |
| 2056 | |
| 2057 | r22 = _mm_mul_ps(rsq22,rinv22); |
| 2058 | |
| 2059 | /* EWALD ELECTROSTATICS */ |
| 2060 | |
| 2061 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 2062 | ewrt = _mm_mul_ps(r22,ewtabscale); |
| 2063 | ewitab = _mm_cvttps_epi32(ewrt); |
| 2064 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 2065 | gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})), |
| 2066 | ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})), |
| 2067 | &ewtabF,&ewtabFn); |
| 2068 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
| 2069 | felec = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec)); |
| 2070 | |
| 2071 | cutoff_mask = _mm_cmplt_ps(rsq22,rcutoff2); |
| 2072 | |
| 2073 | fscal = felec; |
| 2074 | |
| 2075 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 2076 | |
| 2077 | /* Calculate temporary vectorial force */ |
| 2078 | tx = _mm_mul_ps(fscal,dx22); |
| 2079 | ty = _mm_mul_ps(fscal,dy22); |
| 2080 | tz = _mm_mul_ps(fscal,dz22); |
| 2081 | |
| 2082 | /* Update vectorial force */ |
| 2083 | fix2 = _mm_add_ps(fix2,tx); |
| 2084 | fiy2 = _mm_add_ps(fiy2,ty); |
| 2085 | fiz2 = _mm_add_ps(fiz2,tz); |
| 2086 | |
| 2087 | fjx2 = _mm_add_ps(fjx2,tx); |
| 2088 | fjy2 = _mm_add_ps(fjy2,ty); |
| 2089 | fjz2 = _mm_add_ps(fjz2,tz); |
| 2090 | |
| 2091 | } |
| 2092 | |
| 2093 | fjptrA = f+j_coord_offsetA; |
| 2094 | fjptrB = f+j_coord_offsetB; |
| 2095 | fjptrC = f+j_coord_offsetC; |
| 2096 | fjptrD = f+j_coord_offsetD; |
| 2097 | |
| 2098 | gmx_mm_decrement_3rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD, |
| 2099 | fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2); |
| 2100 | |
| 2101 | /* Inner loop uses 358 flops */ |
| 2102 | } |
| 2103 | |
| 2104 | if(jidx<j_index_end) |
| 2105 | { |
| 2106 | |
| 2107 | /* Get j neighbor index, and coordinate index */ |
| 2108 | jnrlistA = jjnr[jidx]; |
| 2109 | jnrlistB = jjnr[jidx+1]; |
| 2110 | jnrlistC = jjnr[jidx+2]; |
| 2111 | jnrlistD = jjnr[jidx+3]; |
| 2112 | /* Sign of each element will be negative for non-real atoms. |
| 2113 | * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones, |
| 2114 | * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries. |
| 2115 | */ |
| 2116 | dummy_mask = gmx_mm_castsi128_ps_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())); |
| 2117 | jnrA = (jnrlistA>=0) ? jnrlistA : 0; |
| 2118 | jnrB = (jnrlistB>=0) ? jnrlistB : 0; |
| 2119 | jnrC = (jnrlistC>=0) ? jnrlistC : 0; |
| 2120 | jnrD = (jnrlistD>=0) ? jnrlistD : 0; |
| 2121 | j_coord_offsetA = DIM3*jnrA; |
| 2122 | j_coord_offsetB = DIM3*jnrB; |
| 2123 | j_coord_offsetC = DIM3*jnrC; |
| 2124 | j_coord_offsetD = DIM3*jnrD; |
| 2125 | |
| 2126 | /* load j atom coordinates */ |
| 2127 | gmx_mm_load_3rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB, |
| 2128 | x+j_coord_offsetC,x+j_coord_offsetD, |
| 2129 | &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2); |
| 2130 | |
| 2131 | /* Calculate displacement vector */ |
| 2132 | dx00 = _mm_sub_ps(ix0,jx0); |
| 2133 | dy00 = _mm_sub_ps(iy0,jy0); |
| 2134 | dz00 = _mm_sub_ps(iz0,jz0); |
| 2135 | dx01 = _mm_sub_ps(ix0,jx1); |
| 2136 | dy01 = _mm_sub_ps(iy0,jy1); |
| 2137 | dz01 = _mm_sub_ps(iz0,jz1); |
| 2138 | dx02 = _mm_sub_ps(ix0,jx2); |
| 2139 | dy02 = _mm_sub_ps(iy0,jy2); |
| 2140 | dz02 = _mm_sub_ps(iz0,jz2); |
| 2141 | dx10 = _mm_sub_ps(ix1,jx0); |
| 2142 | dy10 = _mm_sub_ps(iy1,jy0); |
| 2143 | dz10 = _mm_sub_ps(iz1,jz0); |
| 2144 | dx11 = _mm_sub_ps(ix1,jx1); |
| 2145 | dy11 = _mm_sub_ps(iy1,jy1); |
| 2146 | dz11 = _mm_sub_ps(iz1,jz1); |
| 2147 | dx12 = _mm_sub_ps(ix1,jx2); |
| 2148 | dy12 = _mm_sub_ps(iy1,jy2); |
| 2149 | dz12 = _mm_sub_ps(iz1,jz2); |
| 2150 | dx20 = _mm_sub_ps(ix2,jx0); |
| 2151 | dy20 = _mm_sub_ps(iy2,jy0); |
| 2152 | dz20 = _mm_sub_ps(iz2,jz0); |
| 2153 | dx21 = _mm_sub_ps(ix2,jx1); |
| 2154 | dy21 = _mm_sub_ps(iy2,jy1); |
| 2155 | dz21 = _mm_sub_ps(iz2,jz1); |
| 2156 | dx22 = _mm_sub_ps(ix2,jx2); |
| 2157 | dy22 = _mm_sub_ps(iy2,jy2); |
| 2158 | dz22 = _mm_sub_ps(iz2,jz2); |
| 2159 | |
| 2160 | /* Calculate squared distance and things based on it */ |
| 2161 | rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00); |
| 2162 | rsq01 = gmx_mm_calc_rsq_ps(dx01,dy01,dz01); |
| 2163 | rsq02 = gmx_mm_calc_rsq_ps(dx02,dy02,dz02); |
| 2164 | rsq10 = gmx_mm_calc_rsq_ps(dx10,dy10,dz10); |
| 2165 | rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11); |
| 2166 | rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12); |
| 2167 | rsq20 = gmx_mm_calc_rsq_ps(dx20,dy20,dz20); |
| 2168 | rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21); |
| 2169 | rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22); |
| 2170 | |
| 2171 | rinv00 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq00); |
| 2172 | rinv01 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq01); |
| 2173 | rinv02 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq02); |
| 2174 | rinv10 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq10); |
| 2175 | rinv11 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq11); |
| 2176 | rinv12 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq12); |
| 2177 | rinv20 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq20); |
| 2178 | rinv21 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq21); |
| 2179 | rinv22 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq22); |
| 2180 | |
| 2181 | rinvsq00 = _mm_mul_ps(rinv00,rinv00); |
| 2182 | rinvsq01 = _mm_mul_ps(rinv01,rinv01); |
| 2183 | rinvsq02 = _mm_mul_ps(rinv02,rinv02); |
| 2184 | rinvsq10 = _mm_mul_ps(rinv10,rinv10); |
| 2185 | rinvsq11 = _mm_mul_ps(rinv11,rinv11); |
| 2186 | rinvsq12 = _mm_mul_ps(rinv12,rinv12); |
| 2187 | rinvsq20 = _mm_mul_ps(rinv20,rinv20); |
| 2188 | rinvsq21 = _mm_mul_ps(rinv21,rinv21); |
| 2189 | rinvsq22 = _mm_mul_ps(rinv22,rinv22); |
| 2190 | |
| 2191 | fjx0 = _mm_setzero_ps(); |
| 2192 | fjy0 = _mm_setzero_ps(); |
| 2193 | fjz0 = _mm_setzero_ps(); |
| 2194 | fjx1 = _mm_setzero_ps(); |
| 2195 | fjy1 = _mm_setzero_ps(); |
| 2196 | fjz1 = _mm_setzero_ps(); |
| 2197 | fjx2 = _mm_setzero_ps(); |
| 2198 | fjy2 = _mm_setzero_ps(); |
| 2199 | fjz2 = _mm_setzero_ps(); |
| 2200 | |
| 2201 | /************************** |
| 2202 | * CALCULATE INTERACTIONS * |
| 2203 | **************************/ |
| 2204 | |
| 2205 | if (gmx_mm_any_lt(rsq00,rcutoff2)) |
| 2206 | { |
| 2207 | |
| 2208 | r00 = _mm_mul_ps(rsq00,rinv00); |
| 2209 | r00 = _mm_andnot_ps(dummy_mask,r00); |
| 2210 | |
| 2211 | /* EWALD ELECTROSTATICS */ |
| 2212 | |
| 2213 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 2214 | ewrt = _mm_mul_ps(r00,ewtabscale); |
| 2215 | ewitab = _mm_cvttps_epi32(ewrt); |
| 2216 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 2217 | gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})), |
| 2218 | ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})), |
| 2219 | &ewtabF,&ewtabFn); |
| 2220 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
| 2221 | felec = _mm_mul_ps(_mm_mul_ps(qq00,rinv00),_mm_sub_ps(rinvsq00,felec)); |
| 2222 | |
| 2223 | /* LENNARD-JONES DISPERSION/REPULSION */ |
| 2224 | |
| 2225 | rinvsix = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00); |
| 2226 | fvdw = _mm_mul_ps(_mm_sub_ps(_mm_mul_ps(c12_00,rinvsix),c6_00),_mm_mul_ps(rinvsix,rinvsq00)); |
| 2227 | |
| 2228 | cutoff_mask = _mm_cmplt_ps(rsq00,rcutoff2); |
| 2229 | |
| 2230 | fscal = _mm_add_ps(felec,fvdw); |
| 2231 | |
| 2232 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 2233 | |
| 2234 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 2235 | |
| 2236 | /* Calculate temporary vectorial force */ |
| 2237 | tx = _mm_mul_ps(fscal,dx00); |
| 2238 | ty = _mm_mul_ps(fscal,dy00); |
| 2239 | tz = _mm_mul_ps(fscal,dz00); |
| 2240 | |
| 2241 | /* Update vectorial force */ |
| 2242 | fix0 = _mm_add_ps(fix0,tx); |
| 2243 | fiy0 = _mm_add_ps(fiy0,ty); |
| 2244 | fiz0 = _mm_add_ps(fiz0,tz); |
| 2245 | |
| 2246 | fjx0 = _mm_add_ps(fjx0,tx); |
| 2247 | fjy0 = _mm_add_ps(fjy0,ty); |
| 2248 | fjz0 = _mm_add_ps(fjz0,tz); |
| 2249 | |
| 2250 | } |
| 2251 | |
| 2252 | /************************** |
| 2253 | * CALCULATE INTERACTIONS * |
| 2254 | **************************/ |
| 2255 | |
| 2256 | if (gmx_mm_any_lt(rsq01,rcutoff2)) |
| 2257 | { |
| 2258 | |
| 2259 | r01 = _mm_mul_ps(rsq01,rinv01); |
| 2260 | r01 = _mm_andnot_ps(dummy_mask,r01); |
| 2261 | |
| 2262 | /* EWALD ELECTROSTATICS */ |
| 2263 | |
| 2264 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 2265 | ewrt = _mm_mul_ps(r01,ewtabscale); |
| 2266 | ewitab = _mm_cvttps_epi32(ewrt); |
| 2267 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 2268 | gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})), |
| 2269 | ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})), |
| 2270 | &ewtabF,&ewtabFn); |
| 2271 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
| 2272 | felec = _mm_mul_ps(_mm_mul_ps(qq01,rinv01),_mm_sub_ps(rinvsq01,felec)); |
| 2273 | |
| 2274 | cutoff_mask = _mm_cmplt_ps(rsq01,rcutoff2); |
| 2275 | |
| 2276 | fscal = felec; |
| 2277 | |
| 2278 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 2279 | |
| 2280 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 2281 | |
| 2282 | /* Calculate temporary vectorial force */ |
| 2283 | tx = _mm_mul_ps(fscal,dx01); |
| 2284 | ty = _mm_mul_ps(fscal,dy01); |
| 2285 | tz = _mm_mul_ps(fscal,dz01); |
| 2286 | |
| 2287 | /* Update vectorial force */ |
| 2288 | fix0 = _mm_add_ps(fix0,tx); |
| 2289 | fiy0 = _mm_add_ps(fiy0,ty); |
| 2290 | fiz0 = _mm_add_ps(fiz0,tz); |
| 2291 | |
| 2292 | fjx1 = _mm_add_ps(fjx1,tx); |
| 2293 | fjy1 = _mm_add_ps(fjy1,ty); |
| 2294 | fjz1 = _mm_add_ps(fjz1,tz); |
| 2295 | |
| 2296 | } |
| 2297 | |
| 2298 | /************************** |
| 2299 | * CALCULATE INTERACTIONS * |
| 2300 | **************************/ |
| 2301 | |
| 2302 | if (gmx_mm_any_lt(rsq02,rcutoff2)) |
| 2303 | { |
| 2304 | |
| 2305 | r02 = _mm_mul_ps(rsq02,rinv02); |
| 2306 | r02 = _mm_andnot_ps(dummy_mask,r02); |
| 2307 | |
| 2308 | /* EWALD ELECTROSTATICS */ |
| 2309 | |
| 2310 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 2311 | ewrt = _mm_mul_ps(r02,ewtabscale); |
| 2312 | ewitab = _mm_cvttps_epi32(ewrt); |
| 2313 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 2314 | gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})), |
| 2315 | ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})), |
| 2316 | &ewtabF,&ewtabFn); |
| 2317 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
| 2318 | felec = _mm_mul_ps(_mm_mul_ps(qq02,rinv02),_mm_sub_ps(rinvsq02,felec)); |
| 2319 | |
| 2320 | cutoff_mask = _mm_cmplt_ps(rsq02,rcutoff2); |
| 2321 | |
| 2322 | fscal = felec; |
| 2323 | |
| 2324 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 2325 | |
| 2326 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 2327 | |
| 2328 | /* Calculate temporary vectorial force */ |
| 2329 | tx = _mm_mul_ps(fscal,dx02); |
| 2330 | ty = _mm_mul_ps(fscal,dy02); |
| 2331 | tz = _mm_mul_ps(fscal,dz02); |
| 2332 | |
| 2333 | /* Update vectorial force */ |
| 2334 | fix0 = _mm_add_ps(fix0,tx); |
| 2335 | fiy0 = _mm_add_ps(fiy0,ty); |
| 2336 | fiz0 = _mm_add_ps(fiz0,tz); |
| 2337 | |
| 2338 | fjx2 = _mm_add_ps(fjx2,tx); |
| 2339 | fjy2 = _mm_add_ps(fjy2,ty); |
| 2340 | fjz2 = _mm_add_ps(fjz2,tz); |
| 2341 | |
| 2342 | } |
| 2343 | |
| 2344 | /************************** |
| 2345 | * CALCULATE INTERACTIONS * |
| 2346 | **************************/ |
| 2347 | |
| 2348 | if (gmx_mm_any_lt(rsq10,rcutoff2)) |
| 2349 | { |
| 2350 | |
| 2351 | r10 = _mm_mul_ps(rsq10,rinv10); |
| 2352 | r10 = _mm_andnot_ps(dummy_mask,r10); |
| 2353 | |
| 2354 | /* EWALD ELECTROSTATICS */ |
| 2355 | |
| 2356 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 2357 | ewrt = _mm_mul_ps(r10,ewtabscale); |
| 2358 | ewitab = _mm_cvttps_epi32(ewrt); |
| 2359 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 2360 | gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})), |
| 2361 | ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})), |
| 2362 | &ewtabF,&ewtabFn); |
| 2363 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
| 2364 | felec = _mm_mul_ps(_mm_mul_ps(qq10,rinv10),_mm_sub_ps(rinvsq10,felec)); |
| 2365 | |
| 2366 | cutoff_mask = _mm_cmplt_ps(rsq10,rcutoff2); |
| 2367 | |
| 2368 | fscal = felec; |
| 2369 | |
| 2370 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 2371 | |
| 2372 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 2373 | |
| 2374 | /* Calculate temporary vectorial force */ |
| 2375 | tx = _mm_mul_ps(fscal,dx10); |
| 2376 | ty = _mm_mul_ps(fscal,dy10); |
| 2377 | tz = _mm_mul_ps(fscal,dz10); |
| 2378 | |
| 2379 | /* Update vectorial force */ |
| 2380 | fix1 = _mm_add_ps(fix1,tx); |
| 2381 | fiy1 = _mm_add_ps(fiy1,ty); |
| 2382 | fiz1 = _mm_add_ps(fiz1,tz); |
| 2383 | |
| 2384 | fjx0 = _mm_add_ps(fjx0,tx); |
| 2385 | fjy0 = _mm_add_ps(fjy0,ty); |
| 2386 | fjz0 = _mm_add_ps(fjz0,tz); |
| 2387 | |
| 2388 | } |
| 2389 | |
| 2390 | /************************** |
| 2391 | * CALCULATE INTERACTIONS * |
| 2392 | **************************/ |
| 2393 | |
| 2394 | if (gmx_mm_any_lt(rsq11,rcutoff2)) |
| 2395 | { |
| 2396 | |
| 2397 | r11 = _mm_mul_ps(rsq11,rinv11); |
| 2398 | r11 = _mm_andnot_ps(dummy_mask,r11); |
| 2399 | |
| 2400 | /* EWALD ELECTROSTATICS */ |
| 2401 | |
| 2402 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 2403 | ewrt = _mm_mul_ps(r11,ewtabscale); |
| 2404 | ewitab = _mm_cvttps_epi32(ewrt); |
| 2405 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 2406 | gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})), |
| 2407 | ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})), |
| 2408 | &ewtabF,&ewtabFn); |
| 2409 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
| 2410 | felec = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec)); |
| 2411 | |
| 2412 | cutoff_mask = _mm_cmplt_ps(rsq11,rcutoff2); |
| 2413 | |
| 2414 | fscal = felec; |
| 2415 | |
| 2416 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 2417 | |
| 2418 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 2419 | |
| 2420 | /* Calculate temporary vectorial force */ |
| 2421 | tx = _mm_mul_ps(fscal,dx11); |
| 2422 | ty = _mm_mul_ps(fscal,dy11); |
| 2423 | tz = _mm_mul_ps(fscal,dz11); |
| 2424 | |
| 2425 | /* Update vectorial force */ |
| 2426 | fix1 = _mm_add_ps(fix1,tx); |
| 2427 | fiy1 = _mm_add_ps(fiy1,ty); |
| 2428 | fiz1 = _mm_add_ps(fiz1,tz); |
| 2429 | |
| 2430 | fjx1 = _mm_add_ps(fjx1,tx); |
| 2431 | fjy1 = _mm_add_ps(fjy1,ty); |
| 2432 | fjz1 = _mm_add_ps(fjz1,tz); |
| 2433 | |
| 2434 | } |
| 2435 | |
| 2436 | /************************** |
| 2437 | * CALCULATE INTERACTIONS * |
| 2438 | **************************/ |
| 2439 | |
| 2440 | if (gmx_mm_any_lt(rsq12,rcutoff2)) |
| 2441 | { |
| 2442 | |
| 2443 | r12 = _mm_mul_ps(rsq12,rinv12); |
| 2444 | r12 = _mm_andnot_ps(dummy_mask,r12); |
| 2445 | |
| 2446 | /* EWALD ELECTROSTATICS */ |
| 2447 | |
| 2448 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 2449 | ewrt = _mm_mul_ps(r12,ewtabscale); |
| 2450 | ewitab = _mm_cvttps_epi32(ewrt); |
| 2451 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 2452 | gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})), |
| 2453 | ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})), |
| 2454 | &ewtabF,&ewtabFn); |
| 2455 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
| 2456 | felec = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec)); |
| 2457 | |
| 2458 | cutoff_mask = _mm_cmplt_ps(rsq12,rcutoff2); |
| 2459 | |
| 2460 | fscal = felec; |
| 2461 | |
| 2462 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 2463 | |
| 2464 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 2465 | |
| 2466 | /* Calculate temporary vectorial force */ |
| 2467 | tx = _mm_mul_ps(fscal,dx12); |
| 2468 | ty = _mm_mul_ps(fscal,dy12); |
| 2469 | tz = _mm_mul_ps(fscal,dz12); |
| 2470 | |
| 2471 | /* Update vectorial force */ |
| 2472 | fix1 = _mm_add_ps(fix1,tx); |
| 2473 | fiy1 = _mm_add_ps(fiy1,ty); |
| 2474 | fiz1 = _mm_add_ps(fiz1,tz); |
| 2475 | |
| 2476 | fjx2 = _mm_add_ps(fjx2,tx); |
| 2477 | fjy2 = _mm_add_ps(fjy2,ty); |
| 2478 | fjz2 = _mm_add_ps(fjz2,tz); |
| 2479 | |
| 2480 | } |
| 2481 | |
| 2482 | /************************** |
| 2483 | * CALCULATE INTERACTIONS * |
| 2484 | **************************/ |
| 2485 | |
| 2486 | if (gmx_mm_any_lt(rsq20,rcutoff2)) |
| 2487 | { |
| 2488 | |
| 2489 | r20 = _mm_mul_ps(rsq20,rinv20); |
| 2490 | r20 = _mm_andnot_ps(dummy_mask,r20); |
| 2491 | |
| 2492 | /* EWALD ELECTROSTATICS */ |
| 2493 | |
| 2494 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 2495 | ewrt = _mm_mul_ps(r20,ewtabscale); |
| 2496 | ewitab = _mm_cvttps_epi32(ewrt); |
| 2497 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 2498 | gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})), |
| 2499 | ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})), |
| 2500 | &ewtabF,&ewtabFn); |
| 2501 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
| 2502 | felec = _mm_mul_ps(_mm_mul_ps(qq20,rinv20),_mm_sub_ps(rinvsq20,felec)); |
| 2503 | |
| 2504 | cutoff_mask = _mm_cmplt_ps(rsq20,rcutoff2); |
| 2505 | |
| 2506 | fscal = felec; |
| 2507 | |
| 2508 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 2509 | |
| 2510 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 2511 | |
| 2512 | /* Calculate temporary vectorial force */ |
| 2513 | tx = _mm_mul_ps(fscal,dx20); |
| 2514 | ty = _mm_mul_ps(fscal,dy20); |
| 2515 | tz = _mm_mul_ps(fscal,dz20); |
| 2516 | |
| 2517 | /* Update vectorial force */ |
| 2518 | fix2 = _mm_add_ps(fix2,tx); |
| 2519 | fiy2 = _mm_add_ps(fiy2,ty); |
| 2520 | fiz2 = _mm_add_ps(fiz2,tz); |
| 2521 | |
| 2522 | fjx0 = _mm_add_ps(fjx0,tx); |
| 2523 | fjy0 = _mm_add_ps(fjy0,ty); |
| 2524 | fjz0 = _mm_add_ps(fjz0,tz); |
| 2525 | |
| 2526 | } |
| 2527 | |
| 2528 | /************************** |
| 2529 | * CALCULATE INTERACTIONS * |
| 2530 | **************************/ |
| 2531 | |
| 2532 | if (gmx_mm_any_lt(rsq21,rcutoff2)) |
| 2533 | { |
| 2534 | |
| 2535 | r21 = _mm_mul_ps(rsq21,rinv21); |
| 2536 | r21 = _mm_andnot_ps(dummy_mask,r21); |
| 2537 | |
| 2538 | /* EWALD ELECTROSTATICS */ |
| 2539 | |
| 2540 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 2541 | ewrt = _mm_mul_ps(r21,ewtabscale); |
| 2542 | ewitab = _mm_cvttps_epi32(ewrt); |
| 2543 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 2544 | gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})), |
| 2545 | ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})), |
| 2546 | &ewtabF,&ewtabFn); |
| 2547 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
| 2548 | felec = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec)); |
| 2549 | |
| 2550 | cutoff_mask = _mm_cmplt_ps(rsq21,rcutoff2); |
| 2551 | |
| 2552 | fscal = felec; |
| 2553 | |
| 2554 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 2555 | |
| 2556 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 2557 | |
| 2558 | /* Calculate temporary vectorial force */ |
| 2559 | tx = _mm_mul_ps(fscal,dx21); |
| 2560 | ty = _mm_mul_ps(fscal,dy21); |
| 2561 | tz = _mm_mul_ps(fscal,dz21); |
| 2562 | |
| 2563 | /* Update vectorial force */ |
| 2564 | fix2 = _mm_add_ps(fix2,tx); |
| 2565 | fiy2 = _mm_add_ps(fiy2,ty); |
| 2566 | fiz2 = _mm_add_ps(fiz2,tz); |
| 2567 | |
| 2568 | fjx1 = _mm_add_ps(fjx1,tx); |
| 2569 | fjy1 = _mm_add_ps(fjy1,ty); |
| 2570 | fjz1 = _mm_add_ps(fjz1,tz); |
| 2571 | |
| 2572 | } |
| 2573 | |
| 2574 | /************************** |
| 2575 | * CALCULATE INTERACTIONS * |
| 2576 | **************************/ |
| 2577 | |
| 2578 | if (gmx_mm_any_lt(rsq22,rcutoff2)) |
| 2579 | { |
| 2580 | |
| 2581 | r22 = _mm_mul_ps(rsq22,rinv22); |
| 2582 | r22 = _mm_andnot_ps(dummy_mask,r22); |
| 2583 | |
| 2584 | /* EWALD ELECTROSTATICS */ |
| 2585 | |
| 2586 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 2587 | ewrt = _mm_mul_ps(r22,ewtabscale); |
| 2588 | ewitab = _mm_cvttps_epi32(ewrt); |
| 2589 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 2590 | gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})), |
| 2591 | ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})),ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})), |
| 2592 | &ewtabF,&ewtabFn); |
| 2593 | felec = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn)); |
| 2594 | felec = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec)); |
| 2595 | |
| 2596 | cutoff_mask = _mm_cmplt_ps(rsq22,rcutoff2); |
| 2597 | |
| 2598 | fscal = felec; |
| 2599 | |
| 2600 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 2601 | |
| 2602 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 2603 | |
| 2604 | /* Calculate temporary vectorial force */ |
| 2605 | tx = _mm_mul_ps(fscal,dx22); |
| 2606 | ty = _mm_mul_ps(fscal,dy22); |
| 2607 | tz = _mm_mul_ps(fscal,dz22); |
| 2608 | |
| 2609 | /* Update vectorial force */ |
| 2610 | fix2 = _mm_add_ps(fix2,tx); |
| 2611 | fiy2 = _mm_add_ps(fiy2,ty); |
| 2612 | fiz2 = _mm_add_ps(fiz2,tz); |
| 2613 | |
| 2614 | fjx2 = _mm_add_ps(fjx2,tx); |
| 2615 | fjy2 = _mm_add_ps(fjy2,ty); |
| 2616 | fjz2 = _mm_add_ps(fjz2,tz); |
| 2617 | |
| 2618 | } |
| 2619 | |
| 2620 | fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch; |
| 2621 | fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch; |
| 2622 | fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch; |
| 2623 | fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch; |
| 2624 | |
| 2625 | gmx_mm_decrement_3rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD, |
| 2626 | fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2); |
| 2627 | |
| 2628 | /* Inner loop uses 367 flops */ |
| 2629 | } |
| 2630 | |
| 2631 | /* End of innermost loop */ |
| 2632 | |
| 2633 | gmx_mm_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2, |
| 2634 | f+i_coord_offset,fshift+i_shift_offset); |
| 2635 | |
| 2636 | /* Increment number of inner iterations */ |
| 2637 | inneriter += j_index_end - j_index_start; |
| 2638 | |
| 2639 | /* Outer loop uses 18 flops */ |
| 2640 | } |
| 2641 | |
| 2642 | /* Increment number of outer iterations */ |
| 2643 | outeriter += nri; |
| 2644 | |
| 2645 | /* Update outer/inner flops */ |
| 2646 | |
| 2647 | inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_F,outeriter*18 + inneriter*367)(nrnb)->n[eNR_NBKERNEL_ELEC_VDW_W3W3_F] += outeriter*18 + inneriter *367; |
| 2648 | } |