| File: | gromacs/gmxlib/nonbonded/nb_kernel_sse4_1_single/nb_kernel_ElecEwSw_VdwLJSw_GeomW4W4_sse4_1_single.c |
| Location: | line 1905, column 5 |
| Description: | Value stored to 'd' is never read |
| 1 | /* |
| 2 | * This file is part of the GROMACS molecular simulation package. |
| 3 | * |
| 4 | * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by |
| 5 | * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl, |
| 6 | * and including many others, as listed in the AUTHORS file in the |
| 7 | * top-level source directory and at http://www.gromacs.org. |
| 8 | * |
| 9 | * GROMACS is free software; you can redistribute it and/or |
| 10 | * modify it under the terms of the GNU Lesser General Public License |
| 11 | * as published by the Free Software Foundation; either version 2.1 |
| 12 | * of the License, or (at your option) any later version. |
| 13 | * |
| 14 | * GROMACS is distributed in the hope that it will be useful, |
| 15 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 16 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 17 | * Lesser General Public License for more details. |
| 18 | * |
| 19 | * You should have received a copy of the GNU Lesser General Public |
| 20 | * License along with GROMACS; if not, see |
| 21 | * http://www.gnu.org/licenses, or write to the Free Software Foundation, |
| 22 | * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA. |
| 23 | * |
| 24 | * If you want to redistribute modifications to GROMACS, please |
| 25 | * consider that scientific software is very special. Version |
| 26 | * control is crucial - bugs must be traceable. We will be happy to |
| 27 | * consider code for inclusion in the official distribution, but |
| 28 | * derived work must not be called official GROMACS. Details are found |
| 29 | * in the README & COPYING files - if they are missing, get the |
| 30 | * official version at http://www.gromacs.org. |
| 31 | * |
| 32 | * To help us fund GROMACS development, we humbly ask that you cite |
| 33 | * the research papers on the package. Check out http://www.gromacs.org. |
| 34 | */ |
| 35 | /* |
| 36 | * Note: this file was generated by the GROMACS sse4_1_single kernel generator. |
| 37 | */ |
| 38 | #ifdef HAVE_CONFIG_H1 |
| 39 | #include <config.h> |
| 40 | #endif |
| 41 | |
| 42 | #include <math.h> |
| 43 | |
| 44 | #include "../nb_kernel.h" |
| 45 | #include "types/simple.h" |
| 46 | #include "gromacs/math/vec.h" |
| 47 | #include "nrnb.h" |
| 48 | |
| 49 | #include "gromacs/simd/math_x86_sse4_1_single.h" |
| 50 | #include "kernelutil_x86_sse4_1_single.h" |
| 51 | |
| 52 | /* |
| 53 | * Gromacs nonbonded kernel: nb_kernel_ElecEwSw_VdwLJSw_GeomW4W4_VF_sse4_1_single |
| 54 | * Electrostatics interaction: Ewald |
| 55 | * VdW interaction: LennardJones |
| 56 | * Geometry: Water4-Water4 |
| 57 | * Calculate force/pot: PotentialAndForce |
| 58 | */ |
| 59 | void |
| 60 | nb_kernel_ElecEwSw_VdwLJSw_GeomW4W4_VF_sse4_1_single |
| 61 | (t_nblist * gmx_restrict nlist, |
| 62 | rvec * gmx_restrict xx, |
| 63 | rvec * gmx_restrict ff, |
| 64 | t_forcerec * gmx_restrict fr, |
| 65 | t_mdatoms * gmx_restrict mdatoms, |
| 66 | nb_kernel_data_t gmx_unused__attribute__ ((unused)) * gmx_restrict kernel_data, |
| 67 | t_nrnb * gmx_restrict nrnb) |
| 68 | { |
| 69 | /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or |
| 70 | * just 0 for non-waters. |
| 71 | * Suffixes A,B,C,D refer to j loop unrolling done with SSE, e.g. for the four different |
| 72 | * jnr indices corresponding to data put in the four positions in the SIMD register. |
| 73 | */ |
| 74 | int i_shift_offset,i_coord_offset,outeriter,inneriter; |
| 75 | int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx; |
| 76 | int jnrA,jnrB,jnrC,jnrD; |
| 77 | int jnrlistA,jnrlistB,jnrlistC,jnrlistD; |
| 78 | int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD; |
| 79 | int *iinr,*jindex,*jjnr,*shiftidx,*gid; |
| 80 | real rcutoff_scalar; |
| 81 | real *shiftvec,*fshift,*x,*f; |
| 82 | real *fjptrA,*fjptrB,*fjptrC,*fjptrD; |
| 83 | real scratch[4*DIM3]; |
| 84 | __m128 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall; |
| 85 | int vdwioffset0; |
| 86 | __m128 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0; |
| 87 | int vdwioffset1; |
| 88 | __m128 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1; |
| 89 | int vdwioffset2; |
| 90 | __m128 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2; |
| 91 | int vdwioffset3; |
| 92 | __m128 ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3; |
| 93 | int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D; |
| 94 | __m128 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0; |
| 95 | int vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D; |
| 96 | __m128 jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1; |
| 97 | int vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D; |
| 98 | __m128 jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2; |
| 99 | int vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D; |
| 100 | __m128 jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3; |
| 101 | __m128 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00; |
| 102 | __m128 dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11; |
| 103 | __m128 dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12; |
| 104 | __m128 dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13; |
| 105 | __m128 dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21; |
| 106 | __m128 dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22; |
| 107 | __m128 dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23; |
| 108 | __m128 dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31; |
| 109 | __m128 dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32; |
| 110 | __m128 dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33; |
| 111 | __m128 velec,felec,velecsum,facel,crf,krf,krf2; |
| 112 | real *charge; |
| 113 | int nvdwtype; |
| 114 | __m128 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6; |
| 115 | int *vdwtype; |
| 116 | real *vdwparam; |
| 117 | __m128 one_sixth = _mm_set1_ps(1.0/6.0); |
| 118 | __m128 one_twelfth = _mm_set1_ps(1.0/12.0); |
| 119 | __m128i ewitab; |
| 120 | __m128 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV; |
| 121 | real *ewtab; |
| 122 | __m128 rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw; |
| 123 | real rswitch_scalar,d_scalar; |
| 124 | __m128 dummy_mask,cutoff_mask; |
| 125 | __m128 signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) ); |
| 126 | __m128 one = _mm_set1_ps(1.0); |
| 127 | __m128 two = _mm_set1_ps(2.0); |
| 128 | x = xx[0]; |
| 129 | f = ff[0]; |
| 130 | |
| 131 | nri = nlist->nri; |
| 132 | iinr = nlist->iinr; |
| 133 | jindex = nlist->jindex; |
| 134 | jjnr = nlist->jjnr; |
| 135 | shiftidx = nlist->shift; |
| 136 | gid = nlist->gid; |
| 137 | shiftvec = fr->shift_vec[0]; |
| 138 | fshift = fr->fshift[0]; |
| 139 | facel = _mm_set1_ps(fr->epsfac); |
| 140 | charge = mdatoms->chargeA; |
| 141 | nvdwtype = fr->ntype; |
| 142 | vdwparam = fr->nbfp; |
| 143 | vdwtype = mdatoms->typeA; |
| 144 | |
| 145 | sh_ewald = _mm_set1_ps(fr->ic->sh_ewald); |
| 146 | ewtab = fr->ic->tabq_coul_FDV0; |
| 147 | ewtabscale = _mm_set1_ps(fr->ic->tabq_scale); |
| 148 | ewtabhalfspace = _mm_set1_ps(0.5/fr->ic->tabq_scale); |
| 149 | |
| 150 | /* Setup water-specific parameters */ |
| 151 | inr = nlist->iinr[0]; |
| 152 | iq1 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1])); |
| 153 | iq2 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2])); |
| 154 | iq3 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+3])); |
| 155 | vdwioffset0 = 2*nvdwtype*vdwtype[inr+0]; |
| 156 | |
| 157 | jq1 = _mm_set1_ps(charge[inr+1]); |
| 158 | jq2 = _mm_set1_ps(charge[inr+2]); |
| 159 | jq3 = _mm_set1_ps(charge[inr+3]); |
| 160 | vdwjidx0A = 2*vdwtype[inr+0]; |
| 161 | c6_00 = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A]); |
| 162 | c12_00 = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A+1]); |
| 163 | qq11 = _mm_mul_ps(iq1,jq1); |
| 164 | qq12 = _mm_mul_ps(iq1,jq2); |
| 165 | qq13 = _mm_mul_ps(iq1,jq3); |
| 166 | qq21 = _mm_mul_ps(iq2,jq1); |
| 167 | qq22 = _mm_mul_ps(iq2,jq2); |
| 168 | qq23 = _mm_mul_ps(iq2,jq3); |
| 169 | qq31 = _mm_mul_ps(iq3,jq1); |
| 170 | qq32 = _mm_mul_ps(iq3,jq2); |
| 171 | qq33 = _mm_mul_ps(iq3,jq3); |
| 172 | |
| 173 | /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */ |
| 174 | rcutoff_scalar = fr->rcoulomb; |
| 175 | rcutoff = _mm_set1_ps(rcutoff_scalar); |
| 176 | rcutoff2 = _mm_mul_ps(rcutoff,rcutoff); |
| 177 | |
| 178 | rswitch_scalar = fr->rcoulomb_switch; |
| 179 | rswitch = _mm_set1_ps(rswitch_scalar); |
| 180 | /* Setup switch parameters */ |
| 181 | d_scalar = rcutoff_scalar-rswitch_scalar; |
| 182 | d = _mm_set1_ps(d_scalar); |
| 183 | swV3 = _mm_set1_ps(-10.0/(d_scalar*d_scalar*d_scalar)); |
| 184 | swV4 = _mm_set1_ps( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar)); |
| 185 | swV5 = _mm_set1_ps( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar)); |
| 186 | swF2 = _mm_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar)); |
| 187 | swF3 = _mm_set1_ps( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar)); |
| 188 | swF4 = _mm_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar)); |
| 189 | |
| 190 | /* Avoid stupid compiler warnings */ |
| 191 | jnrA = jnrB = jnrC = jnrD = 0; |
| 192 | j_coord_offsetA = 0; |
| 193 | j_coord_offsetB = 0; |
| 194 | j_coord_offsetC = 0; |
| 195 | j_coord_offsetD = 0; |
| 196 | |
| 197 | outeriter = 0; |
| 198 | inneriter = 0; |
| 199 | |
| 200 | for(iidx=0;iidx<4*DIM3;iidx++) |
| 201 | { |
| 202 | scratch[iidx] = 0.0; |
| 203 | } |
| 204 | |
| 205 | /* Start outer loop over neighborlists */ |
| 206 | for(iidx=0; iidx<nri; iidx++) |
| 207 | { |
| 208 | /* Load shift vector for this list */ |
| 209 | i_shift_offset = DIM3*shiftidx[iidx]; |
| 210 | |
| 211 | /* Load limits for loop over neighbors */ |
| 212 | j_index_start = jindex[iidx]; |
| 213 | j_index_end = jindex[iidx+1]; |
| 214 | |
| 215 | /* Get outer coordinate index */ |
| 216 | inr = iinr[iidx]; |
| 217 | i_coord_offset = DIM3*inr; |
| 218 | |
| 219 | /* Load i particle coords and add shift vector */ |
| 220 | gmx_mm_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset, |
| 221 | &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3); |
| 222 | |
| 223 | fix0 = _mm_setzero_ps(); |
| 224 | fiy0 = _mm_setzero_ps(); |
| 225 | fiz0 = _mm_setzero_ps(); |
| 226 | fix1 = _mm_setzero_ps(); |
| 227 | fiy1 = _mm_setzero_ps(); |
| 228 | fiz1 = _mm_setzero_ps(); |
| 229 | fix2 = _mm_setzero_ps(); |
| 230 | fiy2 = _mm_setzero_ps(); |
| 231 | fiz2 = _mm_setzero_ps(); |
| 232 | fix3 = _mm_setzero_ps(); |
| 233 | fiy3 = _mm_setzero_ps(); |
| 234 | fiz3 = _mm_setzero_ps(); |
| 235 | |
| 236 | /* Reset potential sums */ |
| 237 | velecsum = _mm_setzero_ps(); |
| 238 | vvdwsum = _mm_setzero_ps(); |
| 239 | |
| 240 | /* Start inner kernel loop */ |
| 241 | for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4) |
| 242 | { |
| 243 | |
| 244 | /* Get j neighbor index, and coordinate index */ |
| 245 | jnrA = jjnr[jidx]; |
| 246 | jnrB = jjnr[jidx+1]; |
| 247 | jnrC = jjnr[jidx+2]; |
| 248 | jnrD = jjnr[jidx+3]; |
| 249 | j_coord_offsetA = DIM3*jnrA; |
| 250 | j_coord_offsetB = DIM3*jnrB; |
| 251 | j_coord_offsetC = DIM3*jnrC; |
| 252 | j_coord_offsetD = DIM3*jnrD; |
| 253 | |
| 254 | /* load j atom coordinates */ |
| 255 | gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB, |
| 256 | x+j_coord_offsetC,x+j_coord_offsetD, |
| 257 | &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2, |
| 258 | &jy2,&jz2,&jx3,&jy3,&jz3); |
| 259 | |
| 260 | /* Calculate displacement vector */ |
| 261 | dx00 = _mm_sub_ps(ix0,jx0); |
| 262 | dy00 = _mm_sub_ps(iy0,jy0); |
| 263 | dz00 = _mm_sub_ps(iz0,jz0); |
| 264 | dx11 = _mm_sub_ps(ix1,jx1); |
| 265 | dy11 = _mm_sub_ps(iy1,jy1); |
| 266 | dz11 = _mm_sub_ps(iz1,jz1); |
| 267 | dx12 = _mm_sub_ps(ix1,jx2); |
| 268 | dy12 = _mm_sub_ps(iy1,jy2); |
| 269 | dz12 = _mm_sub_ps(iz1,jz2); |
| 270 | dx13 = _mm_sub_ps(ix1,jx3); |
| 271 | dy13 = _mm_sub_ps(iy1,jy3); |
| 272 | dz13 = _mm_sub_ps(iz1,jz3); |
| 273 | dx21 = _mm_sub_ps(ix2,jx1); |
| 274 | dy21 = _mm_sub_ps(iy2,jy1); |
| 275 | dz21 = _mm_sub_ps(iz2,jz1); |
| 276 | dx22 = _mm_sub_ps(ix2,jx2); |
| 277 | dy22 = _mm_sub_ps(iy2,jy2); |
| 278 | dz22 = _mm_sub_ps(iz2,jz2); |
| 279 | dx23 = _mm_sub_ps(ix2,jx3); |
| 280 | dy23 = _mm_sub_ps(iy2,jy3); |
| 281 | dz23 = _mm_sub_ps(iz2,jz3); |
| 282 | dx31 = _mm_sub_ps(ix3,jx1); |
| 283 | dy31 = _mm_sub_ps(iy3,jy1); |
| 284 | dz31 = _mm_sub_ps(iz3,jz1); |
| 285 | dx32 = _mm_sub_ps(ix3,jx2); |
| 286 | dy32 = _mm_sub_ps(iy3,jy2); |
| 287 | dz32 = _mm_sub_ps(iz3,jz2); |
| 288 | dx33 = _mm_sub_ps(ix3,jx3); |
| 289 | dy33 = _mm_sub_ps(iy3,jy3); |
| 290 | dz33 = _mm_sub_ps(iz3,jz3); |
| 291 | |
| 292 | /* Calculate squared distance and things based on it */ |
| 293 | rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00); |
| 294 | rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11); |
| 295 | rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12); |
| 296 | rsq13 = gmx_mm_calc_rsq_ps(dx13,dy13,dz13); |
| 297 | rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21); |
| 298 | rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22); |
| 299 | rsq23 = gmx_mm_calc_rsq_ps(dx23,dy23,dz23); |
| 300 | rsq31 = gmx_mm_calc_rsq_ps(dx31,dy31,dz31); |
| 301 | rsq32 = gmx_mm_calc_rsq_ps(dx32,dy32,dz32); |
| 302 | rsq33 = gmx_mm_calc_rsq_ps(dx33,dy33,dz33); |
| 303 | |
| 304 | rinv00 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq00); |
| 305 | rinv11 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq11); |
| 306 | rinv12 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq12); |
| 307 | rinv13 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq13); |
| 308 | rinv21 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq21); |
| 309 | rinv22 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq22); |
| 310 | rinv23 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq23); |
| 311 | rinv31 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq31); |
| 312 | rinv32 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq32); |
| 313 | rinv33 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq33); |
| 314 | |
| 315 | rinvsq00 = _mm_mul_ps(rinv00,rinv00); |
| 316 | rinvsq11 = _mm_mul_ps(rinv11,rinv11); |
| 317 | rinvsq12 = _mm_mul_ps(rinv12,rinv12); |
| 318 | rinvsq13 = _mm_mul_ps(rinv13,rinv13); |
| 319 | rinvsq21 = _mm_mul_ps(rinv21,rinv21); |
| 320 | rinvsq22 = _mm_mul_ps(rinv22,rinv22); |
| 321 | rinvsq23 = _mm_mul_ps(rinv23,rinv23); |
| 322 | rinvsq31 = _mm_mul_ps(rinv31,rinv31); |
| 323 | rinvsq32 = _mm_mul_ps(rinv32,rinv32); |
| 324 | rinvsq33 = _mm_mul_ps(rinv33,rinv33); |
| 325 | |
| 326 | fjx0 = _mm_setzero_ps(); |
| 327 | fjy0 = _mm_setzero_ps(); |
| 328 | fjz0 = _mm_setzero_ps(); |
| 329 | fjx1 = _mm_setzero_ps(); |
| 330 | fjy1 = _mm_setzero_ps(); |
| 331 | fjz1 = _mm_setzero_ps(); |
| 332 | fjx2 = _mm_setzero_ps(); |
| 333 | fjy2 = _mm_setzero_ps(); |
| 334 | fjz2 = _mm_setzero_ps(); |
| 335 | fjx3 = _mm_setzero_ps(); |
| 336 | fjy3 = _mm_setzero_ps(); |
| 337 | fjz3 = _mm_setzero_ps(); |
| 338 | |
| 339 | /************************** |
| 340 | * CALCULATE INTERACTIONS * |
| 341 | **************************/ |
| 342 | |
| 343 | if (gmx_mm_any_lt(rsq00,rcutoff2)) |
| 344 | { |
| 345 | |
| 346 | r00 = _mm_mul_ps(rsq00,rinv00); |
| 347 | |
| 348 | /* LENNARD-JONES DISPERSION/REPULSION */ |
| 349 | |
| 350 | rinvsix = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00); |
| 351 | vvdw6 = _mm_mul_ps(c6_00,rinvsix); |
| 352 | vvdw12 = _mm_mul_ps(c12_00,_mm_mul_ps(rinvsix,rinvsix)); |
| 353 | vvdw = _mm_sub_ps( _mm_mul_ps(vvdw12,one_twelfth) , _mm_mul_ps(vvdw6,one_sixth) ); |
| 354 | fvdw = _mm_mul_ps(_mm_sub_ps(vvdw12,vvdw6),rinvsq00); |
| 355 | |
| 356 | d = _mm_sub_ps(r00,rswitch); |
| 357 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 358 | d2 = _mm_mul_ps(d,d); |
| 359 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
| 360 | |
| 361 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 362 | |
| 363 | /* Evaluate switch function */ |
| 364 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 365 | fvdw = _mm_sub_ps( _mm_mul_ps(fvdw,sw) , _mm_mul_ps(rinv00,_mm_mul_ps(vvdw,dsw)) ); |
| 366 | vvdw = _mm_mul_ps(vvdw,sw); |
| 367 | cutoff_mask = _mm_cmplt_ps(rsq00,rcutoff2); |
| 368 | |
| 369 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 370 | vvdw = _mm_and_ps(vvdw,cutoff_mask); |
| 371 | vvdwsum = _mm_add_ps(vvdwsum,vvdw); |
| 372 | |
| 373 | fscal = fvdw; |
| 374 | |
| 375 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 376 | |
| 377 | /* Calculate temporary vectorial force */ |
| 378 | tx = _mm_mul_ps(fscal,dx00); |
| 379 | ty = _mm_mul_ps(fscal,dy00); |
| 380 | tz = _mm_mul_ps(fscal,dz00); |
| 381 | |
| 382 | /* Update vectorial force */ |
| 383 | fix0 = _mm_add_ps(fix0,tx); |
| 384 | fiy0 = _mm_add_ps(fiy0,ty); |
| 385 | fiz0 = _mm_add_ps(fiz0,tz); |
| 386 | |
| 387 | fjx0 = _mm_add_ps(fjx0,tx); |
| 388 | fjy0 = _mm_add_ps(fjy0,ty); |
| 389 | fjz0 = _mm_add_ps(fjz0,tz); |
| 390 | |
| 391 | } |
| 392 | |
| 393 | /************************** |
| 394 | * CALCULATE INTERACTIONS * |
| 395 | **************************/ |
| 396 | |
| 397 | if (gmx_mm_any_lt(rsq11,rcutoff2)) |
| 398 | { |
| 399 | |
| 400 | r11 = _mm_mul_ps(rsq11,rinv11); |
| 401 | |
| 402 | /* EWALD ELECTROSTATICS */ |
| 403 | |
| 404 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 405 | ewrt = _mm_mul_ps(r11,ewtabscale); |
| 406 | ewitab = _mm_cvttps_epi32(ewrt); |
| 407 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 408 | ewitab = _mm_slli_epi32(ewitab,2); |
| 409 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 410 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 411 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 412 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 413 | _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); |
| 414 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 415 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 416 | velec = _mm_mul_ps(qq11,_mm_sub_ps(rinv11,velec)); |
| 417 | felec = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec)); |
| 418 | |
| 419 | d = _mm_sub_ps(r11,rswitch); |
| 420 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 421 | d2 = _mm_mul_ps(d,d); |
| 422 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
| 423 | |
| 424 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 425 | |
| 426 | /* Evaluate switch function */ |
| 427 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 428 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv11,_mm_mul_ps(velec,dsw)) ); |
| 429 | velec = _mm_mul_ps(velec,sw); |
| 430 | cutoff_mask = _mm_cmplt_ps(rsq11,rcutoff2); |
| 431 | |
| 432 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 433 | velec = _mm_and_ps(velec,cutoff_mask); |
| 434 | velecsum = _mm_add_ps(velecsum,velec); |
| 435 | |
| 436 | fscal = felec; |
| 437 | |
| 438 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 439 | |
| 440 | /* Calculate temporary vectorial force */ |
| 441 | tx = _mm_mul_ps(fscal,dx11); |
| 442 | ty = _mm_mul_ps(fscal,dy11); |
| 443 | tz = _mm_mul_ps(fscal,dz11); |
| 444 | |
| 445 | /* Update vectorial force */ |
| 446 | fix1 = _mm_add_ps(fix1,tx); |
| 447 | fiy1 = _mm_add_ps(fiy1,ty); |
| 448 | fiz1 = _mm_add_ps(fiz1,tz); |
| 449 | |
| 450 | fjx1 = _mm_add_ps(fjx1,tx); |
| 451 | fjy1 = _mm_add_ps(fjy1,ty); |
| 452 | fjz1 = _mm_add_ps(fjz1,tz); |
| 453 | |
| 454 | } |
| 455 | |
| 456 | /************************** |
| 457 | * CALCULATE INTERACTIONS * |
| 458 | **************************/ |
| 459 | |
| 460 | if (gmx_mm_any_lt(rsq12,rcutoff2)) |
| 461 | { |
| 462 | |
| 463 | r12 = _mm_mul_ps(rsq12,rinv12); |
| 464 | |
| 465 | /* EWALD ELECTROSTATICS */ |
| 466 | |
| 467 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 468 | ewrt = _mm_mul_ps(r12,ewtabscale); |
| 469 | ewitab = _mm_cvttps_epi32(ewrt); |
| 470 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 471 | ewitab = _mm_slli_epi32(ewitab,2); |
| 472 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 473 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 474 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 475 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 476 | _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); |
| 477 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 478 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 479 | velec = _mm_mul_ps(qq12,_mm_sub_ps(rinv12,velec)); |
| 480 | felec = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec)); |
| 481 | |
| 482 | d = _mm_sub_ps(r12,rswitch); |
| 483 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 484 | d2 = _mm_mul_ps(d,d); |
| 485 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
| 486 | |
| 487 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 488 | |
| 489 | /* Evaluate switch function */ |
| 490 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 491 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv12,_mm_mul_ps(velec,dsw)) ); |
| 492 | velec = _mm_mul_ps(velec,sw); |
| 493 | cutoff_mask = _mm_cmplt_ps(rsq12,rcutoff2); |
| 494 | |
| 495 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 496 | velec = _mm_and_ps(velec,cutoff_mask); |
| 497 | velecsum = _mm_add_ps(velecsum,velec); |
| 498 | |
| 499 | fscal = felec; |
| 500 | |
| 501 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 502 | |
| 503 | /* Calculate temporary vectorial force */ |
| 504 | tx = _mm_mul_ps(fscal,dx12); |
| 505 | ty = _mm_mul_ps(fscal,dy12); |
| 506 | tz = _mm_mul_ps(fscal,dz12); |
| 507 | |
| 508 | /* Update vectorial force */ |
| 509 | fix1 = _mm_add_ps(fix1,tx); |
| 510 | fiy1 = _mm_add_ps(fiy1,ty); |
| 511 | fiz1 = _mm_add_ps(fiz1,tz); |
| 512 | |
| 513 | fjx2 = _mm_add_ps(fjx2,tx); |
| 514 | fjy2 = _mm_add_ps(fjy2,ty); |
| 515 | fjz2 = _mm_add_ps(fjz2,tz); |
| 516 | |
| 517 | } |
| 518 | |
| 519 | /************************** |
| 520 | * CALCULATE INTERACTIONS * |
| 521 | **************************/ |
| 522 | |
| 523 | if (gmx_mm_any_lt(rsq13,rcutoff2)) |
| 524 | { |
| 525 | |
| 526 | r13 = _mm_mul_ps(rsq13,rinv13); |
| 527 | |
| 528 | /* EWALD ELECTROSTATICS */ |
| 529 | |
| 530 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 531 | ewrt = _mm_mul_ps(r13,ewtabscale); |
| 532 | ewitab = _mm_cvttps_epi32(ewrt); |
| 533 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 534 | ewitab = _mm_slli_epi32(ewitab,2); |
| 535 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 536 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 537 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 538 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 539 | _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); |
| 540 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 541 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 542 | velec = _mm_mul_ps(qq13,_mm_sub_ps(rinv13,velec)); |
| 543 | felec = _mm_mul_ps(_mm_mul_ps(qq13,rinv13),_mm_sub_ps(rinvsq13,felec)); |
| 544 | |
| 545 | d = _mm_sub_ps(r13,rswitch); |
| 546 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 547 | d2 = _mm_mul_ps(d,d); |
| 548 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
| 549 | |
| 550 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 551 | |
| 552 | /* Evaluate switch function */ |
| 553 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 554 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv13,_mm_mul_ps(velec,dsw)) ); |
| 555 | velec = _mm_mul_ps(velec,sw); |
| 556 | cutoff_mask = _mm_cmplt_ps(rsq13,rcutoff2); |
| 557 | |
| 558 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 559 | velec = _mm_and_ps(velec,cutoff_mask); |
| 560 | velecsum = _mm_add_ps(velecsum,velec); |
| 561 | |
| 562 | fscal = felec; |
| 563 | |
| 564 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 565 | |
| 566 | /* Calculate temporary vectorial force */ |
| 567 | tx = _mm_mul_ps(fscal,dx13); |
| 568 | ty = _mm_mul_ps(fscal,dy13); |
| 569 | tz = _mm_mul_ps(fscal,dz13); |
| 570 | |
| 571 | /* Update vectorial force */ |
| 572 | fix1 = _mm_add_ps(fix1,tx); |
| 573 | fiy1 = _mm_add_ps(fiy1,ty); |
| 574 | fiz1 = _mm_add_ps(fiz1,tz); |
| 575 | |
| 576 | fjx3 = _mm_add_ps(fjx3,tx); |
| 577 | fjy3 = _mm_add_ps(fjy3,ty); |
| 578 | fjz3 = _mm_add_ps(fjz3,tz); |
| 579 | |
| 580 | } |
| 581 | |
| 582 | /************************** |
| 583 | * CALCULATE INTERACTIONS * |
| 584 | **************************/ |
| 585 | |
| 586 | if (gmx_mm_any_lt(rsq21,rcutoff2)) |
| 587 | { |
| 588 | |
| 589 | r21 = _mm_mul_ps(rsq21,rinv21); |
| 590 | |
| 591 | /* EWALD ELECTROSTATICS */ |
| 592 | |
| 593 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 594 | ewrt = _mm_mul_ps(r21,ewtabscale); |
| 595 | ewitab = _mm_cvttps_epi32(ewrt); |
| 596 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 597 | ewitab = _mm_slli_epi32(ewitab,2); |
| 598 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 599 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 600 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 601 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 602 | _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); |
| 603 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 604 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 605 | velec = _mm_mul_ps(qq21,_mm_sub_ps(rinv21,velec)); |
| 606 | felec = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec)); |
| 607 | |
| 608 | d = _mm_sub_ps(r21,rswitch); |
| 609 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 610 | d2 = _mm_mul_ps(d,d); |
| 611 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
| 612 | |
| 613 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 614 | |
| 615 | /* Evaluate switch function */ |
| 616 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 617 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv21,_mm_mul_ps(velec,dsw)) ); |
| 618 | velec = _mm_mul_ps(velec,sw); |
| 619 | cutoff_mask = _mm_cmplt_ps(rsq21,rcutoff2); |
| 620 | |
| 621 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 622 | velec = _mm_and_ps(velec,cutoff_mask); |
| 623 | velecsum = _mm_add_ps(velecsum,velec); |
| 624 | |
| 625 | fscal = felec; |
| 626 | |
| 627 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 628 | |
| 629 | /* Calculate temporary vectorial force */ |
| 630 | tx = _mm_mul_ps(fscal,dx21); |
| 631 | ty = _mm_mul_ps(fscal,dy21); |
| 632 | tz = _mm_mul_ps(fscal,dz21); |
| 633 | |
| 634 | /* Update vectorial force */ |
| 635 | fix2 = _mm_add_ps(fix2,tx); |
| 636 | fiy2 = _mm_add_ps(fiy2,ty); |
| 637 | fiz2 = _mm_add_ps(fiz2,tz); |
| 638 | |
| 639 | fjx1 = _mm_add_ps(fjx1,tx); |
| 640 | fjy1 = _mm_add_ps(fjy1,ty); |
| 641 | fjz1 = _mm_add_ps(fjz1,tz); |
| 642 | |
| 643 | } |
| 644 | |
| 645 | /************************** |
| 646 | * CALCULATE INTERACTIONS * |
| 647 | **************************/ |
| 648 | |
| 649 | if (gmx_mm_any_lt(rsq22,rcutoff2)) |
| 650 | { |
| 651 | |
| 652 | r22 = _mm_mul_ps(rsq22,rinv22); |
| 653 | |
| 654 | /* EWALD ELECTROSTATICS */ |
| 655 | |
| 656 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 657 | ewrt = _mm_mul_ps(r22,ewtabscale); |
| 658 | ewitab = _mm_cvttps_epi32(ewrt); |
| 659 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 660 | ewitab = _mm_slli_epi32(ewitab,2); |
| 661 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 662 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 663 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 664 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 665 | _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); |
| 666 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 667 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 668 | velec = _mm_mul_ps(qq22,_mm_sub_ps(rinv22,velec)); |
| 669 | felec = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec)); |
| 670 | |
| 671 | d = _mm_sub_ps(r22,rswitch); |
| 672 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 673 | d2 = _mm_mul_ps(d,d); |
| 674 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
| 675 | |
| 676 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 677 | |
| 678 | /* Evaluate switch function */ |
| 679 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 680 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv22,_mm_mul_ps(velec,dsw)) ); |
| 681 | velec = _mm_mul_ps(velec,sw); |
| 682 | cutoff_mask = _mm_cmplt_ps(rsq22,rcutoff2); |
| 683 | |
| 684 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 685 | velec = _mm_and_ps(velec,cutoff_mask); |
| 686 | velecsum = _mm_add_ps(velecsum,velec); |
| 687 | |
| 688 | fscal = felec; |
| 689 | |
| 690 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 691 | |
| 692 | /* Calculate temporary vectorial force */ |
| 693 | tx = _mm_mul_ps(fscal,dx22); |
| 694 | ty = _mm_mul_ps(fscal,dy22); |
| 695 | tz = _mm_mul_ps(fscal,dz22); |
| 696 | |
| 697 | /* Update vectorial force */ |
| 698 | fix2 = _mm_add_ps(fix2,tx); |
| 699 | fiy2 = _mm_add_ps(fiy2,ty); |
| 700 | fiz2 = _mm_add_ps(fiz2,tz); |
| 701 | |
| 702 | fjx2 = _mm_add_ps(fjx2,tx); |
| 703 | fjy2 = _mm_add_ps(fjy2,ty); |
| 704 | fjz2 = _mm_add_ps(fjz2,tz); |
| 705 | |
| 706 | } |
| 707 | |
| 708 | /************************** |
| 709 | * CALCULATE INTERACTIONS * |
| 710 | **************************/ |
| 711 | |
| 712 | if (gmx_mm_any_lt(rsq23,rcutoff2)) |
| 713 | { |
| 714 | |
| 715 | r23 = _mm_mul_ps(rsq23,rinv23); |
| 716 | |
| 717 | /* EWALD ELECTROSTATICS */ |
| 718 | |
| 719 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 720 | ewrt = _mm_mul_ps(r23,ewtabscale); |
| 721 | ewitab = _mm_cvttps_epi32(ewrt); |
| 722 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 723 | ewitab = _mm_slli_epi32(ewitab,2); |
| 724 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 725 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 726 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 727 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 728 | _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); |
| 729 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 730 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 731 | velec = _mm_mul_ps(qq23,_mm_sub_ps(rinv23,velec)); |
| 732 | felec = _mm_mul_ps(_mm_mul_ps(qq23,rinv23),_mm_sub_ps(rinvsq23,felec)); |
| 733 | |
| 734 | d = _mm_sub_ps(r23,rswitch); |
| 735 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 736 | d2 = _mm_mul_ps(d,d); |
| 737 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
| 738 | |
| 739 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 740 | |
| 741 | /* Evaluate switch function */ |
| 742 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 743 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv23,_mm_mul_ps(velec,dsw)) ); |
| 744 | velec = _mm_mul_ps(velec,sw); |
| 745 | cutoff_mask = _mm_cmplt_ps(rsq23,rcutoff2); |
| 746 | |
| 747 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 748 | velec = _mm_and_ps(velec,cutoff_mask); |
| 749 | velecsum = _mm_add_ps(velecsum,velec); |
| 750 | |
| 751 | fscal = felec; |
| 752 | |
| 753 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 754 | |
| 755 | /* Calculate temporary vectorial force */ |
| 756 | tx = _mm_mul_ps(fscal,dx23); |
| 757 | ty = _mm_mul_ps(fscal,dy23); |
| 758 | tz = _mm_mul_ps(fscal,dz23); |
| 759 | |
| 760 | /* Update vectorial force */ |
| 761 | fix2 = _mm_add_ps(fix2,tx); |
| 762 | fiy2 = _mm_add_ps(fiy2,ty); |
| 763 | fiz2 = _mm_add_ps(fiz2,tz); |
| 764 | |
| 765 | fjx3 = _mm_add_ps(fjx3,tx); |
| 766 | fjy3 = _mm_add_ps(fjy3,ty); |
| 767 | fjz3 = _mm_add_ps(fjz3,tz); |
| 768 | |
| 769 | } |
| 770 | |
| 771 | /************************** |
| 772 | * CALCULATE INTERACTIONS * |
| 773 | **************************/ |
| 774 | |
| 775 | if (gmx_mm_any_lt(rsq31,rcutoff2)) |
| 776 | { |
| 777 | |
| 778 | r31 = _mm_mul_ps(rsq31,rinv31); |
| 779 | |
| 780 | /* EWALD ELECTROSTATICS */ |
| 781 | |
| 782 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 783 | ewrt = _mm_mul_ps(r31,ewtabscale); |
| 784 | ewitab = _mm_cvttps_epi32(ewrt); |
| 785 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 786 | ewitab = _mm_slli_epi32(ewitab,2); |
| 787 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 788 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 789 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 790 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 791 | _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); |
| 792 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 793 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 794 | velec = _mm_mul_ps(qq31,_mm_sub_ps(rinv31,velec)); |
| 795 | felec = _mm_mul_ps(_mm_mul_ps(qq31,rinv31),_mm_sub_ps(rinvsq31,felec)); |
| 796 | |
| 797 | d = _mm_sub_ps(r31,rswitch); |
| 798 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 799 | d2 = _mm_mul_ps(d,d); |
| 800 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
| 801 | |
| 802 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 803 | |
| 804 | /* Evaluate switch function */ |
| 805 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 806 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv31,_mm_mul_ps(velec,dsw)) ); |
| 807 | velec = _mm_mul_ps(velec,sw); |
| 808 | cutoff_mask = _mm_cmplt_ps(rsq31,rcutoff2); |
| 809 | |
| 810 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 811 | velec = _mm_and_ps(velec,cutoff_mask); |
| 812 | velecsum = _mm_add_ps(velecsum,velec); |
| 813 | |
| 814 | fscal = felec; |
| 815 | |
| 816 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 817 | |
| 818 | /* Calculate temporary vectorial force */ |
| 819 | tx = _mm_mul_ps(fscal,dx31); |
| 820 | ty = _mm_mul_ps(fscal,dy31); |
| 821 | tz = _mm_mul_ps(fscal,dz31); |
| 822 | |
| 823 | /* Update vectorial force */ |
| 824 | fix3 = _mm_add_ps(fix3,tx); |
| 825 | fiy3 = _mm_add_ps(fiy3,ty); |
| 826 | fiz3 = _mm_add_ps(fiz3,tz); |
| 827 | |
| 828 | fjx1 = _mm_add_ps(fjx1,tx); |
| 829 | fjy1 = _mm_add_ps(fjy1,ty); |
| 830 | fjz1 = _mm_add_ps(fjz1,tz); |
| 831 | |
| 832 | } |
| 833 | |
| 834 | /************************** |
| 835 | * CALCULATE INTERACTIONS * |
| 836 | **************************/ |
| 837 | |
| 838 | if (gmx_mm_any_lt(rsq32,rcutoff2)) |
| 839 | { |
| 840 | |
| 841 | r32 = _mm_mul_ps(rsq32,rinv32); |
| 842 | |
| 843 | /* EWALD ELECTROSTATICS */ |
| 844 | |
| 845 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 846 | ewrt = _mm_mul_ps(r32,ewtabscale); |
| 847 | ewitab = _mm_cvttps_epi32(ewrt); |
| 848 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 849 | ewitab = _mm_slli_epi32(ewitab,2); |
| 850 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 851 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 852 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 853 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 854 | _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); |
| 855 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 856 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 857 | velec = _mm_mul_ps(qq32,_mm_sub_ps(rinv32,velec)); |
| 858 | felec = _mm_mul_ps(_mm_mul_ps(qq32,rinv32),_mm_sub_ps(rinvsq32,felec)); |
| 859 | |
| 860 | d = _mm_sub_ps(r32,rswitch); |
| 861 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 862 | d2 = _mm_mul_ps(d,d); |
| 863 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
| 864 | |
| 865 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 866 | |
| 867 | /* Evaluate switch function */ |
| 868 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 869 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv32,_mm_mul_ps(velec,dsw)) ); |
| 870 | velec = _mm_mul_ps(velec,sw); |
| 871 | cutoff_mask = _mm_cmplt_ps(rsq32,rcutoff2); |
| 872 | |
| 873 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 874 | velec = _mm_and_ps(velec,cutoff_mask); |
| 875 | velecsum = _mm_add_ps(velecsum,velec); |
| 876 | |
| 877 | fscal = felec; |
| 878 | |
| 879 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 880 | |
| 881 | /* Calculate temporary vectorial force */ |
| 882 | tx = _mm_mul_ps(fscal,dx32); |
| 883 | ty = _mm_mul_ps(fscal,dy32); |
| 884 | tz = _mm_mul_ps(fscal,dz32); |
| 885 | |
| 886 | /* Update vectorial force */ |
| 887 | fix3 = _mm_add_ps(fix3,tx); |
| 888 | fiy3 = _mm_add_ps(fiy3,ty); |
| 889 | fiz3 = _mm_add_ps(fiz3,tz); |
| 890 | |
| 891 | fjx2 = _mm_add_ps(fjx2,tx); |
| 892 | fjy2 = _mm_add_ps(fjy2,ty); |
| 893 | fjz2 = _mm_add_ps(fjz2,tz); |
| 894 | |
| 895 | } |
| 896 | |
| 897 | /************************** |
| 898 | * CALCULATE INTERACTIONS * |
| 899 | **************************/ |
| 900 | |
| 901 | if (gmx_mm_any_lt(rsq33,rcutoff2)) |
| 902 | { |
| 903 | |
| 904 | r33 = _mm_mul_ps(rsq33,rinv33); |
| 905 | |
| 906 | /* EWALD ELECTROSTATICS */ |
| 907 | |
| 908 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 909 | ewrt = _mm_mul_ps(r33,ewtabscale); |
| 910 | ewitab = _mm_cvttps_epi32(ewrt); |
| 911 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 912 | ewitab = _mm_slli_epi32(ewitab,2); |
| 913 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 914 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 915 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 916 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 917 | _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); |
| 918 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 919 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 920 | velec = _mm_mul_ps(qq33,_mm_sub_ps(rinv33,velec)); |
| 921 | felec = _mm_mul_ps(_mm_mul_ps(qq33,rinv33),_mm_sub_ps(rinvsq33,felec)); |
| 922 | |
| 923 | d = _mm_sub_ps(r33,rswitch); |
| 924 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 925 | d2 = _mm_mul_ps(d,d); |
| 926 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
| 927 | |
| 928 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 929 | |
| 930 | /* Evaluate switch function */ |
| 931 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 932 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv33,_mm_mul_ps(velec,dsw)) ); |
| 933 | velec = _mm_mul_ps(velec,sw); |
| 934 | cutoff_mask = _mm_cmplt_ps(rsq33,rcutoff2); |
| 935 | |
| 936 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 937 | velec = _mm_and_ps(velec,cutoff_mask); |
| 938 | velecsum = _mm_add_ps(velecsum,velec); |
| 939 | |
| 940 | fscal = felec; |
| 941 | |
| 942 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 943 | |
| 944 | /* Calculate temporary vectorial force */ |
| 945 | tx = _mm_mul_ps(fscal,dx33); |
| 946 | ty = _mm_mul_ps(fscal,dy33); |
| 947 | tz = _mm_mul_ps(fscal,dz33); |
| 948 | |
| 949 | /* Update vectorial force */ |
| 950 | fix3 = _mm_add_ps(fix3,tx); |
| 951 | fiy3 = _mm_add_ps(fiy3,ty); |
| 952 | fiz3 = _mm_add_ps(fiz3,tz); |
| 953 | |
| 954 | fjx3 = _mm_add_ps(fjx3,tx); |
| 955 | fjy3 = _mm_add_ps(fjy3,ty); |
| 956 | fjz3 = _mm_add_ps(fjz3,tz); |
| 957 | |
| 958 | } |
| 959 | |
| 960 | fjptrA = f+j_coord_offsetA; |
| 961 | fjptrB = f+j_coord_offsetB; |
| 962 | fjptrC = f+j_coord_offsetC; |
| 963 | fjptrD = f+j_coord_offsetD; |
| 964 | |
| 965 | gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD, |
| 966 | fjx0,fjy0,fjz0,fjx1,fjy1,fjz1, |
| 967 | fjx2,fjy2,fjz2,fjx3,fjy3,fjz3); |
| 968 | |
| 969 | /* Inner loop uses 647 flops */ |
| 970 | } |
| 971 | |
| 972 | if(jidx<j_index_end) |
| 973 | { |
| 974 | |
| 975 | /* Get j neighbor index, and coordinate index */ |
| 976 | jnrlistA = jjnr[jidx]; |
| 977 | jnrlistB = jjnr[jidx+1]; |
| 978 | jnrlistC = jjnr[jidx+2]; |
| 979 | jnrlistD = jjnr[jidx+3]; |
| 980 | /* Sign of each element will be negative for non-real atoms. |
| 981 | * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones, |
| 982 | * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries. |
| 983 | */ |
| 984 | dummy_mask = gmx_mm_castsi128_ps_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())); |
| 985 | jnrA = (jnrlistA>=0) ? jnrlistA : 0; |
| 986 | jnrB = (jnrlistB>=0) ? jnrlistB : 0; |
| 987 | jnrC = (jnrlistC>=0) ? jnrlistC : 0; |
| 988 | jnrD = (jnrlistD>=0) ? jnrlistD : 0; |
| 989 | j_coord_offsetA = DIM3*jnrA; |
| 990 | j_coord_offsetB = DIM3*jnrB; |
| 991 | j_coord_offsetC = DIM3*jnrC; |
| 992 | j_coord_offsetD = DIM3*jnrD; |
| 993 | |
| 994 | /* load j atom coordinates */ |
| 995 | gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB, |
| 996 | x+j_coord_offsetC,x+j_coord_offsetD, |
| 997 | &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2, |
| 998 | &jy2,&jz2,&jx3,&jy3,&jz3); |
| 999 | |
| 1000 | /* Calculate displacement vector */ |
| 1001 | dx00 = _mm_sub_ps(ix0,jx0); |
| 1002 | dy00 = _mm_sub_ps(iy0,jy0); |
| 1003 | dz00 = _mm_sub_ps(iz0,jz0); |
| 1004 | dx11 = _mm_sub_ps(ix1,jx1); |
| 1005 | dy11 = _mm_sub_ps(iy1,jy1); |
| 1006 | dz11 = _mm_sub_ps(iz1,jz1); |
| 1007 | dx12 = _mm_sub_ps(ix1,jx2); |
| 1008 | dy12 = _mm_sub_ps(iy1,jy2); |
| 1009 | dz12 = _mm_sub_ps(iz1,jz2); |
| 1010 | dx13 = _mm_sub_ps(ix1,jx3); |
| 1011 | dy13 = _mm_sub_ps(iy1,jy3); |
| 1012 | dz13 = _mm_sub_ps(iz1,jz3); |
| 1013 | dx21 = _mm_sub_ps(ix2,jx1); |
| 1014 | dy21 = _mm_sub_ps(iy2,jy1); |
| 1015 | dz21 = _mm_sub_ps(iz2,jz1); |
| 1016 | dx22 = _mm_sub_ps(ix2,jx2); |
| 1017 | dy22 = _mm_sub_ps(iy2,jy2); |
| 1018 | dz22 = _mm_sub_ps(iz2,jz2); |
| 1019 | dx23 = _mm_sub_ps(ix2,jx3); |
| 1020 | dy23 = _mm_sub_ps(iy2,jy3); |
| 1021 | dz23 = _mm_sub_ps(iz2,jz3); |
| 1022 | dx31 = _mm_sub_ps(ix3,jx1); |
| 1023 | dy31 = _mm_sub_ps(iy3,jy1); |
| 1024 | dz31 = _mm_sub_ps(iz3,jz1); |
| 1025 | dx32 = _mm_sub_ps(ix3,jx2); |
| 1026 | dy32 = _mm_sub_ps(iy3,jy2); |
| 1027 | dz32 = _mm_sub_ps(iz3,jz2); |
| 1028 | dx33 = _mm_sub_ps(ix3,jx3); |
| 1029 | dy33 = _mm_sub_ps(iy3,jy3); |
| 1030 | dz33 = _mm_sub_ps(iz3,jz3); |
| 1031 | |
| 1032 | /* Calculate squared distance and things based on it */ |
| 1033 | rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00); |
| 1034 | rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11); |
| 1035 | rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12); |
| 1036 | rsq13 = gmx_mm_calc_rsq_ps(dx13,dy13,dz13); |
| 1037 | rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21); |
| 1038 | rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22); |
| 1039 | rsq23 = gmx_mm_calc_rsq_ps(dx23,dy23,dz23); |
| 1040 | rsq31 = gmx_mm_calc_rsq_ps(dx31,dy31,dz31); |
| 1041 | rsq32 = gmx_mm_calc_rsq_ps(dx32,dy32,dz32); |
| 1042 | rsq33 = gmx_mm_calc_rsq_ps(dx33,dy33,dz33); |
| 1043 | |
| 1044 | rinv00 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq00); |
| 1045 | rinv11 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq11); |
| 1046 | rinv12 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq12); |
| 1047 | rinv13 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq13); |
| 1048 | rinv21 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq21); |
| 1049 | rinv22 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq22); |
| 1050 | rinv23 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq23); |
| 1051 | rinv31 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq31); |
| 1052 | rinv32 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq32); |
| 1053 | rinv33 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq33); |
| 1054 | |
| 1055 | rinvsq00 = _mm_mul_ps(rinv00,rinv00); |
| 1056 | rinvsq11 = _mm_mul_ps(rinv11,rinv11); |
| 1057 | rinvsq12 = _mm_mul_ps(rinv12,rinv12); |
| 1058 | rinvsq13 = _mm_mul_ps(rinv13,rinv13); |
| 1059 | rinvsq21 = _mm_mul_ps(rinv21,rinv21); |
| 1060 | rinvsq22 = _mm_mul_ps(rinv22,rinv22); |
| 1061 | rinvsq23 = _mm_mul_ps(rinv23,rinv23); |
| 1062 | rinvsq31 = _mm_mul_ps(rinv31,rinv31); |
| 1063 | rinvsq32 = _mm_mul_ps(rinv32,rinv32); |
| 1064 | rinvsq33 = _mm_mul_ps(rinv33,rinv33); |
| 1065 | |
| 1066 | fjx0 = _mm_setzero_ps(); |
| 1067 | fjy0 = _mm_setzero_ps(); |
| 1068 | fjz0 = _mm_setzero_ps(); |
| 1069 | fjx1 = _mm_setzero_ps(); |
| 1070 | fjy1 = _mm_setzero_ps(); |
| 1071 | fjz1 = _mm_setzero_ps(); |
| 1072 | fjx2 = _mm_setzero_ps(); |
| 1073 | fjy2 = _mm_setzero_ps(); |
| 1074 | fjz2 = _mm_setzero_ps(); |
| 1075 | fjx3 = _mm_setzero_ps(); |
| 1076 | fjy3 = _mm_setzero_ps(); |
| 1077 | fjz3 = _mm_setzero_ps(); |
| 1078 | |
| 1079 | /************************** |
| 1080 | * CALCULATE INTERACTIONS * |
| 1081 | **************************/ |
| 1082 | |
| 1083 | if (gmx_mm_any_lt(rsq00,rcutoff2)) |
| 1084 | { |
| 1085 | |
| 1086 | r00 = _mm_mul_ps(rsq00,rinv00); |
| 1087 | r00 = _mm_andnot_ps(dummy_mask,r00); |
| 1088 | |
| 1089 | /* LENNARD-JONES DISPERSION/REPULSION */ |
| 1090 | |
| 1091 | rinvsix = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00); |
| 1092 | vvdw6 = _mm_mul_ps(c6_00,rinvsix); |
| 1093 | vvdw12 = _mm_mul_ps(c12_00,_mm_mul_ps(rinvsix,rinvsix)); |
| 1094 | vvdw = _mm_sub_ps( _mm_mul_ps(vvdw12,one_twelfth) , _mm_mul_ps(vvdw6,one_sixth) ); |
| 1095 | fvdw = _mm_mul_ps(_mm_sub_ps(vvdw12,vvdw6),rinvsq00); |
| 1096 | |
| 1097 | d = _mm_sub_ps(r00,rswitch); |
| 1098 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 1099 | d2 = _mm_mul_ps(d,d); |
| 1100 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
| 1101 | |
| 1102 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 1103 | |
| 1104 | /* Evaluate switch function */ |
| 1105 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 1106 | fvdw = _mm_sub_ps( _mm_mul_ps(fvdw,sw) , _mm_mul_ps(rinv00,_mm_mul_ps(vvdw,dsw)) ); |
| 1107 | vvdw = _mm_mul_ps(vvdw,sw); |
| 1108 | cutoff_mask = _mm_cmplt_ps(rsq00,rcutoff2); |
| 1109 | |
| 1110 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 1111 | vvdw = _mm_and_ps(vvdw,cutoff_mask); |
| 1112 | vvdw = _mm_andnot_ps(dummy_mask,vvdw); |
| 1113 | vvdwsum = _mm_add_ps(vvdwsum,vvdw); |
| 1114 | |
| 1115 | fscal = fvdw; |
| 1116 | |
| 1117 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 1118 | |
| 1119 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 1120 | |
| 1121 | /* Calculate temporary vectorial force */ |
| 1122 | tx = _mm_mul_ps(fscal,dx00); |
| 1123 | ty = _mm_mul_ps(fscal,dy00); |
| 1124 | tz = _mm_mul_ps(fscal,dz00); |
| 1125 | |
| 1126 | /* Update vectorial force */ |
| 1127 | fix0 = _mm_add_ps(fix0,tx); |
| 1128 | fiy0 = _mm_add_ps(fiy0,ty); |
| 1129 | fiz0 = _mm_add_ps(fiz0,tz); |
| 1130 | |
| 1131 | fjx0 = _mm_add_ps(fjx0,tx); |
| 1132 | fjy0 = _mm_add_ps(fjy0,ty); |
| 1133 | fjz0 = _mm_add_ps(fjz0,tz); |
| 1134 | |
| 1135 | } |
| 1136 | |
| 1137 | /************************** |
| 1138 | * CALCULATE INTERACTIONS * |
| 1139 | **************************/ |
| 1140 | |
| 1141 | if (gmx_mm_any_lt(rsq11,rcutoff2)) |
| 1142 | { |
| 1143 | |
| 1144 | r11 = _mm_mul_ps(rsq11,rinv11); |
| 1145 | r11 = _mm_andnot_ps(dummy_mask,r11); |
| 1146 | |
| 1147 | /* EWALD ELECTROSTATICS */ |
| 1148 | |
| 1149 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 1150 | ewrt = _mm_mul_ps(r11,ewtabscale); |
| 1151 | ewitab = _mm_cvttps_epi32(ewrt); |
| 1152 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 1153 | ewitab = _mm_slli_epi32(ewitab,2); |
| 1154 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 1155 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 1156 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 1157 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 1158 | _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); |
| 1159 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 1160 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 1161 | velec = _mm_mul_ps(qq11,_mm_sub_ps(rinv11,velec)); |
| 1162 | felec = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec)); |
| 1163 | |
| 1164 | d = _mm_sub_ps(r11,rswitch); |
| 1165 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 1166 | d2 = _mm_mul_ps(d,d); |
| 1167 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
| 1168 | |
| 1169 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 1170 | |
| 1171 | /* Evaluate switch function */ |
| 1172 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 1173 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv11,_mm_mul_ps(velec,dsw)) ); |
| 1174 | velec = _mm_mul_ps(velec,sw); |
| 1175 | cutoff_mask = _mm_cmplt_ps(rsq11,rcutoff2); |
| 1176 | |
| 1177 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 1178 | velec = _mm_and_ps(velec,cutoff_mask); |
| 1179 | velec = _mm_andnot_ps(dummy_mask,velec); |
| 1180 | velecsum = _mm_add_ps(velecsum,velec); |
| 1181 | |
| 1182 | fscal = felec; |
| 1183 | |
| 1184 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 1185 | |
| 1186 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 1187 | |
| 1188 | /* Calculate temporary vectorial force */ |
| 1189 | tx = _mm_mul_ps(fscal,dx11); |
| 1190 | ty = _mm_mul_ps(fscal,dy11); |
| 1191 | tz = _mm_mul_ps(fscal,dz11); |
| 1192 | |
| 1193 | /* Update vectorial force */ |
| 1194 | fix1 = _mm_add_ps(fix1,tx); |
| 1195 | fiy1 = _mm_add_ps(fiy1,ty); |
| 1196 | fiz1 = _mm_add_ps(fiz1,tz); |
| 1197 | |
| 1198 | fjx1 = _mm_add_ps(fjx1,tx); |
| 1199 | fjy1 = _mm_add_ps(fjy1,ty); |
| 1200 | fjz1 = _mm_add_ps(fjz1,tz); |
| 1201 | |
| 1202 | } |
| 1203 | |
| 1204 | /************************** |
| 1205 | * CALCULATE INTERACTIONS * |
| 1206 | **************************/ |
| 1207 | |
| 1208 | if (gmx_mm_any_lt(rsq12,rcutoff2)) |
| 1209 | { |
| 1210 | |
| 1211 | r12 = _mm_mul_ps(rsq12,rinv12); |
| 1212 | r12 = _mm_andnot_ps(dummy_mask,r12); |
| 1213 | |
| 1214 | /* EWALD ELECTROSTATICS */ |
| 1215 | |
| 1216 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 1217 | ewrt = _mm_mul_ps(r12,ewtabscale); |
| 1218 | ewitab = _mm_cvttps_epi32(ewrt); |
| 1219 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 1220 | ewitab = _mm_slli_epi32(ewitab,2); |
| 1221 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 1222 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 1223 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 1224 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 1225 | _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); |
| 1226 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 1227 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 1228 | velec = _mm_mul_ps(qq12,_mm_sub_ps(rinv12,velec)); |
| 1229 | felec = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec)); |
| 1230 | |
| 1231 | d = _mm_sub_ps(r12,rswitch); |
| 1232 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 1233 | d2 = _mm_mul_ps(d,d); |
| 1234 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
| 1235 | |
| 1236 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 1237 | |
| 1238 | /* Evaluate switch function */ |
| 1239 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 1240 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv12,_mm_mul_ps(velec,dsw)) ); |
| 1241 | velec = _mm_mul_ps(velec,sw); |
| 1242 | cutoff_mask = _mm_cmplt_ps(rsq12,rcutoff2); |
| 1243 | |
| 1244 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 1245 | velec = _mm_and_ps(velec,cutoff_mask); |
| 1246 | velec = _mm_andnot_ps(dummy_mask,velec); |
| 1247 | velecsum = _mm_add_ps(velecsum,velec); |
| 1248 | |
| 1249 | fscal = felec; |
| 1250 | |
| 1251 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 1252 | |
| 1253 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 1254 | |
| 1255 | /* Calculate temporary vectorial force */ |
| 1256 | tx = _mm_mul_ps(fscal,dx12); |
| 1257 | ty = _mm_mul_ps(fscal,dy12); |
| 1258 | tz = _mm_mul_ps(fscal,dz12); |
| 1259 | |
| 1260 | /* Update vectorial force */ |
| 1261 | fix1 = _mm_add_ps(fix1,tx); |
| 1262 | fiy1 = _mm_add_ps(fiy1,ty); |
| 1263 | fiz1 = _mm_add_ps(fiz1,tz); |
| 1264 | |
| 1265 | fjx2 = _mm_add_ps(fjx2,tx); |
| 1266 | fjy2 = _mm_add_ps(fjy2,ty); |
| 1267 | fjz2 = _mm_add_ps(fjz2,tz); |
| 1268 | |
| 1269 | } |
| 1270 | |
| 1271 | /************************** |
| 1272 | * CALCULATE INTERACTIONS * |
| 1273 | **************************/ |
| 1274 | |
| 1275 | if (gmx_mm_any_lt(rsq13,rcutoff2)) |
| 1276 | { |
| 1277 | |
| 1278 | r13 = _mm_mul_ps(rsq13,rinv13); |
| 1279 | r13 = _mm_andnot_ps(dummy_mask,r13); |
| 1280 | |
| 1281 | /* EWALD ELECTROSTATICS */ |
| 1282 | |
| 1283 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 1284 | ewrt = _mm_mul_ps(r13,ewtabscale); |
| 1285 | ewitab = _mm_cvttps_epi32(ewrt); |
| 1286 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 1287 | ewitab = _mm_slli_epi32(ewitab,2); |
| 1288 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 1289 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 1290 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 1291 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 1292 | _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); |
| 1293 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 1294 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 1295 | velec = _mm_mul_ps(qq13,_mm_sub_ps(rinv13,velec)); |
| 1296 | felec = _mm_mul_ps(_mm_mul_ps(qq13,rinv13),_mm_sub_ps(rinvsq13,felec)); |
| 1297 | |
| 1298 | d = _mm_sub_ps(r13,rswitch); |
| 1299 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 1300 | d2 = _mm_mul_ps(d,d); |
| 1301 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
| 1302 | |
| 1303 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 1304 | |
| 1305 | /* Evaluate switch function */ |
| 1306 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 1307 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv13,_mm_mul_ps(velec,dsw)) ); |
| 1308 | velec = _mm_mul_ps(velec,sw); |
| 1309 | cutoff_mask = _mm_cmplt_ps(rsq13,rcutoff2); |
| 1310 | |
| 1311 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 1312 | velec = _mm_and_ps(velec,cutoff_mask); |
| 1313 | velec = _mm_andnot_ps(dummy_mask,velec); |
| 1314 | velecsum = _mm_add_ps(velecsum,velec); |
| 1315 | |
| 1316 | fscal = felec; |
| 1317 | |
| 1318 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 1319 | |
| 1320 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 1321 | |
| 1322 | /* Calculate temporary vectorial force */ |
| 1323 | tx = _mm_mul_ps(fscal,dx13); |
| 1324 | ty = _mm_mul_ps(fscal,dy13); |
| 1325 | tz = _mm_mul_ps(fscal,dz13); |
| 1326 | |
| 1327 | /* Update vectorial force */ |
| 1328 | fix1 = _mm_add_ps(fix1,tx); |
| 1329 | fiy1 = _mm_add_ps(fiy1,ty); |
| 1330 | fiz1 = _mm_add_ps(fiz1,tz); |
| 1331 | |
| 1332 | fjx3 = _mm_add_ps(fjx3,tx); |
| 1333 | fjy3 = _mm_add_ps(fjy3,ty); |
| 1334 | fjz3 = _mm_add_ps(fjz3,tz); |
| 1335 | |
| 1336 | } |
| 1337 | |
| 1338 | /************************** |
| 1339 | * CALCULATE INTERACTIONS * |
| 1340 | **************************/ |
| 1341 | |
| 1342 | if (gmx_mm_any_lt(rsq21,rcutoff2)) |
| 1343 | { |
| 1344 | |
| 1345 | r21 = _mm_mul_ps(rsq21,rinv21); |
| 1346 | r21 = _mm_andnot_ps(dummy_mask,r21); |
| 1347 | |
| 1348 | /* EWALD ELECTROSTATICS */ |
| 1349 | |
| 1350 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 1351 | ewrt = _mm_mul_ps(r21,ewtabscale); |
| 1352 | ewitab = _mm_cvttps_epi32(ewrt); |
| 1353 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 1354 | ewitab = _mm_slli_epi32(ewitab,2); |
| 1355 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 1356 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 1357 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 1358 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 1359 | _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); |
| 1360 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 1361 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 1362 | velec = _mm_mul_ps(qq21,_mm_sub_ps(rinv21,velec)); |
| 1363 | felec = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec)); |
| 1364 | |
| 1365 | d = _mm_sub_ps(r21,rswitch); |
| 1366 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 1367 | d2 = _mm_mul_ps(d,d); |
| 1368 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
| 1369 | |
| 1370 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 1371 | |
| 1372 | /* Evaluate switch function */ |
| 1373 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 1374 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv21,_mm_mul_ps(velec,dsw)) ); |
| 1375 | velec = _mm_mul_ps(velec,sw); |
| 1376 | cutoff_mask = _mm_cmplt_ps(rsq21,rcutoff2); |
| 1377 | |
| 1378 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 1379 | velec = _mm_and_ps(velec,cutoff_mask); |
| 1380 | velec = _mm_andnot_ps(dummy_mask,velec); |
| 1381 | velecsum = _mm_add_ps(velecsum,velec); |
| 1382 | |
| 1383 | fscal = felec; |
| 1384 | |
| 1385 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 1386 | |
| 1387 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 1388 | |
| 1389 | /* Calculate temporary vectorial force */ |
| 1390 | tx = _mm_mul_ps(fscal,dx21); |
| 1391 | ty = _mm_mul_ps(fscal,dy21); |
| 1392 | tz = _mm_mul_ps(fscal,dz21); |
| 1393 | |
| 1394 | /* Update vectorial force */ |
| 1395 | fix2 = _mm_add_ps(fix2,tx); |
| 1396 | fiy2 = _mm_add_ps(fiy2,ty); |
| 1397 | fiz2 = _mm_add_ps(fiz2,tz); |
| 1398 | |
| 1399 | fjx1 = _mm_add_ps(fjx1,tx); |
| 1400 | fjy1 = _mm_add_ps(fjy1,ty); |
| 1401 | fjz1 = _mm_add_ps(fjz1,tz); |
| 1402 | |
| 1403 | } |
| 1404 | |
| 1405 | /************************** |
| 1406 | * CALCULATE INTERACTIONS * |
| 1407 | **************************/ |
| 1408 | |
| 1409 | if (gmx_mm_any_lt(rsq22,rcutoff2)) |
| 1410 | { |
| 1411 | |
| 1412 | r22 = _mm_mul_ps(rsq22,rinv22); |
| 1413 | r22 = _mm_andnot_ps(dummy_mask,r22); |
| 1414 | |
| 1415 | /* EWALD ELECTROSTATICS */ |
| 1416 | |
| 1417 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 1418 | ewrt = _mm_mul_ps(r22,ewtabscale); |
| 1419 | ewitab = _mm_cvttps_epi32(ewrt); |
| 1420 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 1421 | ewitab = _mm_slli_epi32(ewitab,2); |
| 1422 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 1423 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 1424 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 1425 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 1426 | _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); |
| 1427 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 1428 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 1429 | velec = _mm_mul_ps(qq22,_mm_sub_ps(rinv22,velec)); |
| 1430 | felec = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec)); |
| 1431 | |
| 1432 | d = _mm_sub_ps(r22,rswitch); |
| 1433 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 1434 | d2 = _mm_mul_ps(d,d); |
| 1435 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
| 1436 | |
| 1437 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 1438 | |
| 1439 | /* Evaluate switch function */ |
| 1440 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 1441 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv22,_mm_mul_ps(velec,dsw)) ); |
| 1442 | velec = _mm_mul_ps(velec,sw); |
| 1443 | cutoff_mask = _mm_cmplt_ps(rsq22,rcutoff2); |
| 1444 | |
| 1445 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 1446 | velec = _mm_and_ps(velec,cutoff_mask); |
| 1447 | velec = _mm_andnot_ps(dummy_mask,velec); |
| 1448 | velecsum = _mm_add_ps(velecsum,velec); |
| 1449 | |
| 1450 | fscal = felec; |
| 1451 | |
| 1452 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 1453 | |
| 1454 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 1455 | |
| 1456 | /* Calculate temporary vectorial force */ |
| 1457 | tx = _mm_mul_ps(fscal,dx22); |
| 1458 | ty = _mm_mul_ps(fscal,dy22); |
| 1459 | tz = _mm_mul_ps(fscal,dz22); |
| 1460 | |
| 1461 | /* Update vectorial force */ |
| 1462 | fix2 = _mm_add_ps(fix2,tx); |
| 1463 | fiy2 = _mm_add_ps(fiy2,ty); |
| 1464 | fiz2 = _mm_add_ps(fiz2,tz); |
| 1465 | |
| 1466 | fjx2 = _mm_add_ps(fjx2,tx); |
| 1467 | fjy2 = _mm_add_ps(fjy2,ty); |
| 1468 | fjz2 = _mm_add_ps(fjz2,tz); |
| 1469 | |
| 1470 | } |
| 1471 | |
| 1472 | /************************** |
| 1473 | * CALCULATE INTERACTIONS * |
| 1474 | **************************/ |
| 1475 | |
| 1476 | if (gmx_mm_any_lt(rsq23,rcutoff2)) |
| 1477 | { |
| 1478 | |
| 1479 | r23 = _mm_mul_ps(rsq23,rinv23); |
| 1480 | r23 = _mm_andnot_ps(dummy_mask,r23); |
| 1481 | |
| 1482 | /* EWALD ELECTROSTATICS */ |
| 1483 | |
| 1484 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 1485 | ewrt = _mm_mul_ps(r23,ewtabscale); |
| 1486 | ewitab = _mm_cvttps_epi32(ewrt); |
| 1487 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 1488 | ewitab = _mm_slli_epi32(ewitab,2); |
| 1489 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 1490 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 1491 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 1492 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 1493 | _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); |
| 1494 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 1495 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 1496 | velec = _mm_mul_ps(qq23,_mm_sub_ps(rinv23,velec)); |
| 1497 | felec = _mm_mul_ps(_mm_mul_ps(qq23,rinv23),_mm_sub_ps(rinvsq23,felec)); |
| 1498 | |
| 1499 | d = _mm_sub_ps(r23,rswitch); |
| 1500 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 1501 | d2 = _mm_mul_ps(d,d); |
| 1502 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
| 1503 | |
| 1504 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 1505 | |
| 1506 | /* Evaluate switch function */ |
| 1507 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 1508 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv23,_mm_mul_ps(velec,dsw)) ); |
| 1509 | velec = _mm_mul_ps(velec,sw); |
| 1510 | cutoff_mask = _mm_cmplt_ps(rsq23,rcutoff2); |
| 1511 | |
| 1512 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 1513 | velec = _mm_and_ps(velec,cutoff_mask); |
| 1514 | velec = _mm_andnot_ps(dummy_mask,velec); |
| 1515 | velecsum = _mm_add_ps(velecsum,velec); |
| 1516 | |
| 1517 | fscal = felec; |
| 1518 | |
| 1519 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 1520 | |
| 1521 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 1522 | |
| 1523 | /* Calculate temporary vectorial force */ |
| 1524 | tx = _mm_mul_ps(fscal,dx23); |
| 1525 | ty = _mm_mul_ps(fscal,dy23); |
| 1526 | tz = _mm_mul_ps(fscal,dz23); |
| 1527 | |
| 1528 | /* Update vectorial force */ |
| 1529 | fix2 = _mm_add_ps(fix2,tx); |
| 1530 | fiy2 = _mm_add_ps(fiy2,ty); |
| 1531 | fiz2 = _mm_add_ps(fiz2,tz); |
| 1532 | |
| 1533 | fjx3 = _mm_add_ps(fjx3,tx); |
| 1534 | fjy3 = _mm_add_ps(fjy3,ty); |
| 1535 | fjz3 = _mm_add_ps(fjz3,tz); |
| 1536 | |
| 1537 | } |
| 1538 | |
| 1539 | /************************** |
| 1540 | * CALCULATE INTERACTIONS * |
| 1541 | **************************/ |
| 1542 | |
| 1543 | if (gmx_mm_any_lt(rsq31,rcutoff2)) |
| 1544 | { |
| 1545 | |
| 1546 | r31 = _mm_mul_ps(rsq31,rinv31); |
| 1547 | r31 = _mm_andnot_ps(dummy_mask,r31); |
| 1548 | |
| 1549 | /* EWALD ELECTROSTATICS */ |
| 1550 | |
| 1551 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 1552 | ewrt = _mm_mul_ps(r31,ewtabscale); |
| 1553 | ewitab = _mm_cvttps_epi32(ewrt); |
| 1554 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 1555 | ewitab = _mm_slli_epi32(ewitab,2); |
| 1556 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 1557 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 1558 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 1559 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 1560 | _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); |
| 1561 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 1562 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 1563 | velec = _mm_mul_ps(qq31,_mm_sub_ps(rinv31,velec)); |
| 1564 | felec = _mm_mul_ps(_mm_mul_ps(qq31,rinv31),_mm_sub_ps(rinvsq31,felec)); |
| 1565 | |
| 1566 | d = _mm_sub_ps(r31,rswitch); |
| 1567 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 1568 | d2 = _mm_mul_ps(d,d); |
| 1569 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
| 1570 | |
| 1571 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 1572 | |
| 1573 | /* Evaluate switch function */ |
| 1574 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 1575 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv31,_mm_mul_ps(velec,dsw)) ); |
| 1576 | velec = _mm_mul_ps(velec,sw); |
| 1577 | cutoff_mask = _mm_cmplt_ps(rsq31,rcutoff2); |
| 1578 | |
| 1579 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 1580 | velec = _mm_and_ps(velec,cutoff_mask); |
| 1581 | velec = _mm_andnot_ps(dummy_mask,velec); |
| 1582 | velecsum = _mm_add_ps(velecsum,velec); |
| 1583 | |
| 1584 | fscal = felec; |
| 1585 | |
| 1586 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 1587 | |
| 1588 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 1589 | |
| 1590 | /* Calculate temporary vectorial force */ |
| 1591 | tx = _mm_mul_ps(fscal,dx31); |
| 1592 | ty = _mm_mul_ps(fscal,dy31); |
| 1593 | tz = _mm_mul_ps(fscal,dz31); |
| 1594 | |
| 1595 | /* Update vectorial force */ |
| 1596 | fix3 = _mm_add_ps(fix3,tx); |
| 1597 | fiy3 = _mm_add_ps(fiy3,ty); |
| 1598 | fiz3 = _mm_add_ps(fiz3,tz); |
| 1599 | |
| 1600 | fjx1 = _mm_add_ps(fjx1,tx); |
| 1601 | fjy1 = _mm_add_ps(fjy1,ty); |
| 1602 | fjz1 = _mm_add_ps(fjz1,tz); |
| 1603 | |
| 1604 | } |
| 1605 | |
| 1606 | /************************** |
| 1607 | * CALCULATE INTERACTIONS * |
| 1608 | **************************/ |
| 1609 | |
| 1610 | if (gmx_mm_any_lt(rsq32,rcutoff2)) |
| 1611 | { |
| 1612 | |
| 1613 | r32 = _mm_mul_ps(rsq32,rinv32); |
| 1614 | r32 = _mm_andnot_ps(dummy_mask,r32); |
| 1615 | |
| 1616 | /* EWALD ELECTROSTATICS */ |
| 1617 | |
| 1618 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 1619 | ewrt = _mm_mul_ps(r32,ewtabscale); |
| 1620 | ewitab = _mm_cvttps_epi32(ewrt); |
| 1621 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 1622 | ewitab = _mm_slli_epi32(ewitab,2); |
| 1623 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 1624 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 1625 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 1626 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 1627 | _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); |
| 1628 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 1629 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 1630 | velec = _mm_mul_ps(qq32,_mm_sub_ps(rinv32,velec)); |
| 1631 | felec = _mm_mul_ps(_mm_mul_ps(qq32,rinv32),_mm_sub_ps(rinvsq32,felec)); |
| 1632 | |
| 1633 | d = _mm_sub_ps(r32,rswitch); |
| 1634 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 1635 | d2 = _mm_mul_ps(d,d); |
| 1636 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
| 1637 | |
| 1638 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 1639 | |
| 1640 | /* Evaluate switch function */ |
| 1641 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 1642 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv32,_mm_mul_ps(velec,dsw)) ); |
| 1643 | velec = _mm_mul_ps(velec,sw); |
| 1644 | cutoff_mask = _mm_cmplt_ps(rsq32,rcutoff2); |
| 1645 | |
| 1646 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 1647 | velec = _mm_and_ps(velec,cutoff_mask); |
| 1648 | velec = _mm_andnot_ps(dummy_mask,velec); |
| 1649 | velecsum = _mm_add_ps(velecsum,velec); |
| 1650 | |
| 1651 | fscal = felec; |
| 1652 | |
| 1653 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 1654 | |
| 1655 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 1656 | |
| 1657 | /* Calculate temporary vectorial force */ |
| 1658 | tx = _mm_mul_ps(fscal,dx32); |
| 1659 | ty = _mm_mul_ps(fscal,dy32); |
| 1660 | tz = _mm_mul_ps(fscal,dz32); |
| 1661 | |
| 1662 | /* Update vectorial force */ |
| 1663 | fix3 = _mm_add_ps(fix3,tx); |
| 1664 | fiy3 = _mm_add_ps(fiy3,ty); |
| 1665 | fiz3 = _mm_add_ps(fiz3,tz); |
| 1666 | |
| 1667 | fjx2 = _mm_add_ps(fjx2,tx); |
| 1668 | fjy2 = _mm_add_ps(fjy2,ty); |
| 1669 | fjz2 = _mm_add_ps(fjz2,tz); |
| 1670 | |
| 1671 | } |
| 1672 | |
| 1673 | /************************** |
| 1674 | * CALCULATE INTERACTIONS * |
| 1675 | **************************/ |
| 1676 | |
| 1677 | if (gmx_mm_any_lt(rsq33,rcutoff2)) |
| 1678 | { |
| 1679 | |
| 1680 | r33 = _mm_mul_ps(rsq33,rinv33); |
| 1681 | r33 = _mm_andnot_ps(dummy_mask,r33); |
| 1682 | |
| 1683 | /* EWALD ELECTROSTATICS */ |
| 1684 | |
| 1685 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 1686 | ewrt = _mm_mul_ps(r33,ewtabscale); |
| 1687 | ewitab = _mm_cvttps_epi32(ewrt); |
| 1688 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 1689 | ewitab = _mm_slli_epi32(ewitab,2); |
| 1690 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 1691 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 1692 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 1693 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 1694 | _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); |
| 1695 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 1696 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 1697 | velec = _mm_mul_ps(qq33,_mm_sub_ps(rinv33,velec)); |
| 1698 | felec = _mm_mul_ps(_mm_mul_ps(qq33,rinv33),_mm_sub_ps(rinvsq33,felec)); |
| 1699 | |
| 1700 | d = _mm_sub_ps(r33,rswitch); |
| 1701 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 1702 | d2 = _mm_mul_ps(d,d); |
| 1703 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
| 1704 | |
| 1705 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 1706 | |
| 1707 | /* Evaluate switch function */ |
| 1708 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 1709 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv33,_mm_mul_ps(velec,dsw)) ); |
| 1710 | velec = _mm_mul_ps(velec,sw); |
| 1711 | cutoff_mask = _mm_cmplt_ps(rsq33,rcutoff2); |
| 1712 | |
| 1713 | /* Update potential sum for this i atom from the interaction with this j atom. */ |
| 1714 | velec = _mm_and_ps(velec,cutoff_mask); |
| 1715 | velec = _mm_andnot_ps(dummy_mask,velec); |
| 1716 | velecsum = _mm_add_ps(velecsum,velec); |
| 1717 | |
| 1718 | fscal = felec; |
| 1719 | |
| 1720 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 1721 | |
| 1722 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 1723 | |
| 1724 | /* Calculate temporary vectorial force */ |
| 1725 | tx = _mm_mul_ps(fscal,dx33); |
| 1726 | ty = _mm_mul_ps(fscal,dy33); |
| 1727 | tz = _mm_mul_ps(fscal,dz33); |
| 1728 | |
| 1729 | /* Update vectorial force */ |
| 1730 | fix3 = _mm_add_ps(fix3,tx); |
| 1731 | fiy3 = _mm_add_ps(fiy3,ty); |
| 1732 | fiz3 = _mm_add_ps(fiz3,tz); |
| 1733 | |
| 1734 | fjx3 = _mm_add_ps(fjx3,tx); |
| 1735 | fjy3 = _mm_add_ps(fjy3,ty); |
| 1736 | fjz3 = _mm_add_ps(fjz3,tz); |
| 1737 | |
| 1738 | } |
| 1739 | |
| 1740 | fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch; |
| 1741 | fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch; |
| 1742 | fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch; |
| 1743 | fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch; |
| 1744 | |
| 1745 | gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD, |
| 1746 | fjx0,fjy0,fjz0,fjx1,fjy1,fjz1, |
| 1747 | fjx2,fjy2,fjz2,fjx3,fjy3,fjz3); |
| 1748 | |
| 1749 | /* Inner loop uses 657 flops */ |
| 1750 | } |
| 1751 | |
| 1752 | /* End of innermost loop */ |
| 1753 | |
| 1754 | gmx_mm_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3, |
| 1755 | f+i_coord_offset,fshift+i_shift_offset); |
| 1756 | |
| 1757 | ggid = gid[iidx]; |
| 1758 | /* Update potential energies */ |
| 1759 | gmx_mm_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid); |
| 1760 | gmx_mm_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid); |
| 1761 | |
| 1762 | /* Increment number of inner iterations */ |
| 1763 | inneriter += j_index_end - j_index_start; |
| 1764 | |
| 1765 | /* Outer loop uses 26 flops */ |
| 1766 | } |
| 1767 | |
| 1768 | /* Increment number of outer iterations */ |
| 1769 | outeriter += nri; |
| 1770 | |
| 1771 | /* Update outer/inner flops */ |
| 1772 | |
| 1773 | inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*657)(nrnb)->n[eNR_NBKERNEL_ELEC_VDW_W4W4_VF] += outeriter*26 + inneriter*657; |
| 1774 | } |
| 1775 | /* |
| 1776 | * Gromacs nonbonded kernel: nb_kernel_ElecEwSw_VdwLJSw_GeomW4W4_F_sse4_1_single |
| 1777 | * Electrostatics interaction: Ewald |
| 1778 | * VdW interaction: LennardJones |
| 1779 | * Geometry: Water4-Water4 |
| 1780 | * Calculate force/pot: Force |
| 1781 | */ |
| 1782 | void |
| 1783 | nb_kernel_ElecEwSw_VdwLJSw_GeomW4W4_F_sse4_1_single |
| 1784 | (t_nblist * gmx_restrict nlist, |
| 1785 | rvec * gmx_restrict xx, |
| 1786 | rvec * gmx_restrict ff, |
| 1787 | t_forcerec * gmx_restrict fr, |
| 1788 | t_mdatoms * gmx_restrict mdatoms, |
| 1789 | nb_kernel_data_t gmx_unused__attribute__ ((unused)) * gmx_restrict kernel_data, |
| 1790 | t_nrnb * gmx_restrict nrnb) |
| 1791 | { |
| 1792 | /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or |
| 1793 | * just 0 for non-waters. |
| 1794 | * Suffixes A,B,C,D refer to j loop unrolling done with SSE, e.g. for the four different |
| 1795 | * jnr indices corresponding to data put in the four positions in the SIMD register. |
| 1796 | */ |
| 1797 | int i_shift_offset,i_coord_offset,outeriter,inneriter; |
| 1798 | int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx; |
| 1799 | int jnrA,jnrB,jnrC,jnrD; |
| 1800 | int jnrlistA,jnrlistB,jnrlistC,jnrlistD; |
| 1801 | int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD; |
| 1802 | int *iinr,*jindex,*jjnr,*shiftidx,*gid; |
| 1803 | real rcutoff_scalar; |
| 1804 | real *shiftvec,*fshift,*x,*f; |
| 1805 | real *fjptrA,*fjptrB,*fjptrC,*fjptrD; |
| 1806 | real scratch[4*DIM3]; |
| 1807 | __m128 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall; |
| 1808 | int vdwioffset0; |
| 1809 | __m128 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0; |
| 1810 | int vdwioffset1; |
| 1811 | __m128 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1; |
| 1812 | int vdwioffset2; |
| 1813 | __m128 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2; |
| 1814 | int vdwioffset3; |
| 1815 | __m128 ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3; |
| 1816 | int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D; |
| 1817 | __m128 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0; |
| 1818 | int vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D; |
| 1819 | __m128 jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1; |
| 1820 | int vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D; |
| 1821 | __m128 jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2; |
| 1822 | int vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D; |
| 1823 | __m128 jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3; |
| 1824 | __m128 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00; |
| 1825 | __m128 dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11; |
| 1826 | __m128 dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12; |
| 1827 | __m128 dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13; |
| 1828 | __m128 dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21; |
| 1829 | __m128 dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22; |
| 1830 | __m128 dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23; |
| 1831 | __m128 dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31; |
| 1832 | __m128 dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32; |
| 1833 | __m128 dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33; |
| 1834 | __m128 velec,felec,velecsum,facel,crf,krf,krf2; |
| 1835 | real *charge; |
| 1836 | int nvdwtype; |
| 1837 | __m128 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6; |
| 1838 | int *vdwtype; |
| 1839 | real *vdwparam; |
| 1840 | __m128 one_sixth = _mm_set1_ps(1.0/6.0); |
| 1841 | __m128 one_twelfth = _mm_set1_ps(1.0/12.0); |
| 1842 | __m128i ewitab; |
| 1843 | __m128 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV; |
| 1844 | real *ewtab; |
| 1845 | __m128 rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw; |
| 1846 | real rswitch_scalar,d_scalar; |
| 1847 | __m128 dummy_mask,cutoff_mask; |
| 1848 | __m128 signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) ); |
| 1849 | __m128 one = _mm_set1_ps(1.0); |
| 1850 | __m128 two = _mm_set1_ps(2.0); |
| 1851 | x = xx[0]; |
| 1852 | f = ff[0]; |
| 1853 | |
| 1854 | nri = nlist->nri; |
| 1855 | iinr = nlist->iinr; |
| 1856 | jindex = nlist->jindex; |
| 1857 | jjnr = nlist->jjnr; |
| 1858 | shiftidx = nlist->shift; |
| 1859 | gid = nlist->gid; |
| 1860 | shiftvec = fr->shift_vec[0]; |
| 1861 | fshift = fr->fshift[0]; |
| 1862 | facel = _mm_set1_ps(fr->epsfac); |
| 1863 | charge = mdatoms->chargeA; |
| 1864 | nvdwtype = fr->ntype; |
| 1865 | vdwparam = fr->nbfp; |
| 1866 | vdwtype = mdatoms->typeA; |
| 1867 | |
| 1868 | sh_ewald = _mm_set1_ps(fr->ic->sh_ewald); |
| 1869 | ewtab = fr->ic->tabq_coul_FDV0; |
| 1870 | ewtabscale = _mm_set1_ps(fr->ic->tabq_scale); |
| 1871 | ewtabhalfspace = _mm_set1_ps(0.5/fr->ic->tabq_scale); |
| 1872 | |
| 1873 | /* Setup water-specific parameters */ |
| 1874 | inr = nlist->iinr[0]; |
| 1875 | iq1 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1])); |
| 1876 | iq2 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2])); |
| 1877 | iq3 = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+3])); |
| 1878 | vdwioffset0 = 2*nvdwtype*vdwtype[inr+0]; |
| 1879 | |
| 1880 | jq1 = _mm_set1_ps(charge[inr+1]); |
| 1881 | jq2 = _mm_set1_ps(charge[inr+2]); |
| 1882 | jq3 = _mm_set1_ps(charge[inr+3]); |
| 1883 | vdwjidx0A = 2*vdwtype[inr+0]; |
| 1884 | c6_00 = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A]); |
| 1885 | c12_00 = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A+1]); |
| 1886 | qq11 = _mm_mul_ps(iq1,jq1); |
| 1887 | qq12 = _mm_mul_ps(iq1,jq2); |
| 1888 | qq13 = _mm_mul_ps(iq1,jq3); |
| 1889 | qq21 = _mm_mul_ps(iq2,jq1); |
| 1890 | qq22 = _mm_mul_ps(iq2,jq2); |
| 1891 | qq23 = _mm_mul_ps(iq2,jq3); |
| 1892 | qq31 = _mm_mul_ps(iq3,jq1); |
| 1893 | qq32 = _mm_mul_ps(iq3,jq2); |
| 1894 | qq33 = _mm_mul_ps(iq3,jq3); |
| 1895 | |
| 1896 | /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */ |
| 1897 | rcutoff_scalar = fr->rcoulomb; |
| 1898 | rcutoff = _mm_set1_ps(rcutoff_scalar); |
| 1899 | rcutoff2 = _mm_mul_ps(rcutoff,rcutoff); |
| 1900 | |
| 1901 | rswitch_scalar = fr->rcoulomb_switch; |
| 1902 | rswitch = _mm_set1_ps(rswitch_scalar); |
| 1903 | /* Setup switch parameters */ |
| 1904 | d_scalar = rcutoff_scalar-rswitch_scalar; |
| 1905 | d = _mm_set1_ps(d_scalar); |
Value stored to 'd' is never read | |
| 1906 | swV3 = _mm_set1_ps(-10.0/(d_scalar*d_scalar*d_scalar)); |
| 1907 | swV4 = _mm_set1_ps( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar)); |
| 1908 | swV5 = _mm_set1_ps( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar)); |
| 1909 | swF2 = _mm_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar)); |
| 1910 | swF3 = _mm_set1_ps( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar)); |
| 1911 | swF4 = _mm_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar)); |
| 1912 | |
| 1913 | /* Avoid stupid compiler warnings */ |
| 1914 | jnrA = jnrB = jnrC = jnrD = 0; |
| 1915 | j_coord_offsetA = 0; |
| 1916 | j_coord_offsetB = 0; |
| 1917 | j_coord_offsetC = 0; |
| 1918 | j_coord_offsetD = 0; |
| 1919 | |
| 1920 | outeriter = 0; |
| 1921 | inneriter = 0; |
| 1922 | |
| 1923 | for(iidx=0;iidx<4*DIM3;iidx++) |
| 1924 | { |
| 1925 | scratch[iidx] = 0.0; |
| 1926 | } |
| 1927 | |
| 1928 | /* Start outer loop over neighborlists */ |
| 1929 | for(iidx=0; iidx<nri; iidx++) |
| 1930 | { |
| 1931 | /* Load shift vector for this list */ |
| 1932 | i_shift_offset = DIM3*shiftidx[iidx]; |
| 1933 | |
| 1934 | /* Load limits for loop over neighbors */ |
| 1935 | j_index_start = jindex[iidx]; |
| 1936 | j_index_end = jindex[iidx+1]; |
| 1937 | |
| 1938 | /* Get outer coordinate index */ |
| 1939 | inr = iinr[iidx]; |
| 1940 | i_coord_offset = DIM3*inr; |
| 1941 | |
| 1942 | /* Load i particle coords and add shift vector */ |
| 1943 | gmx_mm_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset, |
| 1944 | &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3); |
| 1945 | |
| 1946 | fix0 = _mm_setzero_ps(); |
| 1947 | fiy0 = _mm_setzero_ps(); |
| 1948 | fiz0 = _mm_setzero_ps(); |
| 1949 | fix1 = _mm_setzero_ps(); |
| 1950 | fiy1 = _mm_setzero_ps(); |
| 1951 | fiz1 = _mm_setzero_ps(); |
| 1952 | fix2 = _mm_setzero_ps(); |
| 1953 | fiy2 = _mm_setzero_ps(); |
| 1954 | fiz2 = _mm_setzero_ps(); |
| 1955 | fix3 = _mm_setzero_ps(); |
| 1956 | fiy3 = _mm_setzero_ps(); |
| 1957 | fiz3 = _mm_setzero_ps(); |
| 1958 | |
| 1959 | /* Start inner kernel loop */ |
| 1960 | for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4) |
| 1961 | { |
| 1962 | |
| 1963 | /* Get j neighbor index, and coordinate index */ |
| 1964 | jnrA = jjnr[jidx]; |
| 1965 | jnrB = jjnr[jidx+1]; |
| 1966 | jnrC = jjnr[jidx+2]; |
| 1967 | jnrD = jjnr[jidx+3]; |
| 1968 | j_coord_offsetA = DIM3*jnrA; |
| 1969 | j_coord_offsetB = DIM3*jnrB; |
| 1970 | j_coord_offsetC = DIM3*jnrC; |
| 1971 | j_coord_offsetD = DIM3*jnrD; |
| 1972 | |
| 1973 | /* load j atom coordinates */ |
| 1974 | gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB, |
| 1975 | x+j_coord_offsetC,x+j_coord_offsetD, |
| 1976 | &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2, |
| 1977 | &jy2,&jz2,&jx3,&jy3,&jz3); |
| 1978 | |
| 1979 | /* Calculate displacement vector */ |
| 1980 | dx00 = _mm_sub_ps(ix0,jx0); |
| 1981 | dy00 = _mm_sub_ps(iy0,jy0); |
| 1982 | dz00 = _mm_sub_ps(iz0,jz0); |
| 1983 | dx11 = _mm_sub_ps(ix1,jx1); |
| 1984 | dy11 = _mm_sub_ps(iy1,jy1); |
| 1985 | dz11 = _mm_sub_ps(iz1,jz1); |
| 1986 | dx12 = _mm_sub_ps(ix1,jx2); |
| 1987 | dy12 = _mm_sub_ps(iy1,jy2); |
| 1988 | dz12 = _mm_sub_ps(iz1,jz2); |
| 1989 | dx13 = _mm_sub_ps(ix1,jx3); |
| 1990 | dy13 = _mm_sub_ps(iy1,jy3); |
| 1991 | dz13 = _mm_sub_ps(iz1,jz3); |
| 1992 | dx21 = _mm_sub_ps(ix2,jx1); |
| 1993 | dy21 = _mm_sub_ps(iy2,jy1); |
| 1994 | dz21 = _mm_sub_ps(iz2,jz1); |
| 1995 | dx22 = _mm_sub_ps(ix2,jx2); |
| 1996 | dy22 = _mm_sub_ps(iy2,jy2); |
| 1997 | dz22 = _mm_sub_ps(iz2,jz2); |
| 1998 | dx23 = _mm_sub_ps(ix2,jx3); |
| 1999 | dy23 = _mm_sub_ps(iy2,jy3); |
| 2000 | dz23 = _mm_sub_ps(iz2,jz3); |
| 2001 | dx31 = _mm_sub_ps(ix3,jx1); |
| 2002 | dy31 = _mm_sub_ps(iy3,jy1); |
| 2003 | dz31 = _mm_sub_ps(iz3,jz1); |
| 2004 | dx32 = _mm_sub_ps(ix3,jx2); |
| 2005 | dy32 = _mm_sub_ps(iy3,jy2); |
| 2006 | dz32 = _mm_sub_ps(iz3,jz2); |
| 2007 | dx33 = _mm_sub_ps(ix3,jx3); |
| 2008 | dy33 = _mm_sub_ps(iy3,jy3); |
| 2009 | dz33 = _mm_sub_ps(iz3,jz3); |
| 2010 | |
| 2011 | /* Calculate squared distance and things based on it */ |
| 2012 | rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00); |
| 2013 | rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11); |
| 2014 | rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12); |
| 2015 | rsq13 = gmx_mm_calc_rsq_ps(dx13,dy13,dz13); |
| 2016 | rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21); |
| 2017 | rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22); |
| 2018 | rsq23 = gmx_mm_calc_rsq_ps(dx23,dy23,dz23); |
| 2019 | rsq31 = gmx_mm_calc_rsq_ps(dx31,dy31,dz31); |
| 2020 | rsq32 = gmx_mm_calc_rsq_ps(dx32,dy32,dz32); |
| 2021 | rsq33 = gmx_mm_calc_rsq_ps(dx33,dy33,dz33); |
| 2022 | |
| 2023 | rinv00 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq00); |
| 2024 | rinv11 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq11); |
| 2025 | rinv12 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq12); |
| 2026 | rinv13 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq13); |
| 2027 | rinv21 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq21); |
| 2028 | rinv22 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq22); |
| 2029 | rinv23 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq23); |
| 2030 | rinv31 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq31); |
| 2031 | rinv32 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq32); |
| 2032 | rinv33 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq33); |
| 2033 | |
| 2034 | rinvsq00 = _mm_mul_ps(rinv00,rinv00); |
| 2035 | rinvsq11 = _mm_mul_ps(rinv11,rinv11); |
| 2036 | rinvsq12 = _mm_mul_ps(rinv12,rinv12); |
| 2037 | rinvsq13 = _mm_mul_ps(rinv13,rinv13); |
| 2038 | rinvsq21 = _mm_mul_ps(rinv21,rinv21); |
| 2039 | rinvsq22 = _mm_mul_ps(rinv22,rinv22); |
| 2040 | rinvsq23 = _mm_mul_ps(rinv23,rinv23); |
| 2041 | rinvsq31 = _mm_mul_ps(rinv31,rinv31); |
| 2042 | rinvsq32 = _mm_mul_ps(rinv32,rinv32); |
| 2043 | rinvsq33 = _mm_mul_ps(rinv33,rinv33); |
| 2044 | |
| 2045 | fjx0 = _mm_setzero_ps(); |
| 2046 | fjy0 = _mm_setzero_ps(); |
| 2047 | fjz0 = _mm_setzero_ps(); |
| 2048 | fjx1 = _mm_setzero_ps(); |
| 2049 | fjy1 = _mm_setzero_ps(); |
| 2050 | fjz1 = _mm_setzero_ps(); |
| 2051 | fjx2 = _mm_setzero_ps(); |
| 2052 | fjy2 = _mm_setzero_ps(); |
| 2053 | fjz2 = _mm_setzero_ps(); |
| 2054 | fjx3 = _mm_setzero_ps(); |
| 2055 | fjy3 = _mm_setzero_ps(); |
| 2056 | fjz3 = _mm_setzero_ps(); |
| 2057 | |
| 2058 | /************************** |
| 2059 | * CALCULATE INTERACTIONS * |
| 2060 | **************************/ |
| 2061 | |
| 2062 | if (gmx_mm_any_lt(rsq00,rcutoff2)) |
| 2063 | { |
| 2064 | |
| 2065 | r00 = _mm_mul_ps(rsq00,rinv00); |
| 2066 | |
| 2067 | /* LENNARD-JONES DISPERSION/REPULSION */ |
| 2068 | |
| 2069 | rinvsix = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00); |
| 2070 | vvdw6 = _mm_mul_ps(c6_00,rinvsix); |
| 2071 | vvdw12 = _mm_mul_ps(c12_00,_mm_mul_ps(rinvsix,rinvsix)); |
| 2072 | vvdw = _mm_sub_ps( _mm_mul_ps(vvdw12,one_twelfth) , _mm_mul_ps(vvdw6,one_sixth) ); |
| 2073 | fvdw = _mm_mul_ps(_mm_sub_ps(vvdw12,vvdw6),rinvsq00); |
| 2074 | |
| 2075 | d = _mm_sub_ps(r00,rswitch); |
| 2076 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 2077 | d2 = _mm_mul_ps(d,d); |
| 2078 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
| 2079 | |
| 2080 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 2081 | |
| 2082 | /* Evaluate switch function */ |
| 2083 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 2084 | fvdw = _mm_sub_ps( _mm_mul_ps(fvdw,sw) , _mm_mul_ps(rinv00,_mm_mul_ps(vvdw,dsw)) ); |
| 2085 | cutoff_mask = _mm_cmplt_ps(rsq00,rcutoff2); |
| 2086 | |
| 2087 | fscal = fvdw; |
| 2088 | |
| 2089 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 2090 | |
| 2091 | /* Calculate temporary vectorial force */ |
| 2092 | tx = _mm_mul_ps(fscal,dx00); |
| 2093 | ty = _mm_mul_ps(fscal,dy00); |
| 2094 | tz = _mm_mul_ps(fscal,dz00); |
| 2095 | |
| 2096 | /* Update vectorial force */ |
| 2097 | fix0 = _mm_add_ps(fix0,tx); |
| 2098 | fiy0 = _mm_add_ps(fiy0,ty); |
| 2099 | fiz0 = _mm_add_ps(fiz0,tz); |
| 2100 | |
| 2101 | fjx0 = _mm_add_ps(fjx0,tx); |
| 2102 | fjy0 = _mm_add_ps(fjy0,ty); |
| 2103 | fjz0 = _mm_add_ps(fjz0,tz); |
| 2104 | |
| 2105 | } |
| 2106 | |
| 2107 | /************************** |
| 2108 | * CALCULATE INTERACTIONS * |
| 2109 | **************************/ |
| 2110 | |
| 2111 | if (gmx_mm_any_lt(rsq11,rcutoff2)) |
| 2112 | { |
| 2113 | |
| 2114 | r11 = _mm_mul_ps(rsq11,rinv11); |
| 2115 | |
| 2116 | /* EWALD ELECTROSTATICS */ |
| 2117 | |
| 2118 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 2119 | ewrt = _mm_mul_ps(r11,ewtabscale); |
| 2120 | ewitab = _mm_cvttps_epi32(ewrt); |
| 2121 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 2122 | ewitab = _mm_slli_epi32(ewitab,2); |
| 2123 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 2124 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 2125 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 2126 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 2127 | _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); |
| 2128 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 2129 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 2130 | velec = _mm_mul_ps(qq11,_mm_sub_ps(rinv11,velec)); |
| 2131 | felec = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec)); |
| 2132 | |
| 2133 | d = _mm_sub_ps(r11,rswitch); |
| 2134 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 2135 | d2 = _mm_mul_ps(d,d); |
| 2136 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
| 2137 | |
| 2138 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 2139 | |
| 2140 | /* Evaluate switch function */ |
| 2141 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 2142 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv11,_mm_mul_ps(velec,dsw)) ); |
| 2143 | cutoff_mask = _mm_cmplt_ps(rsq11,rcutoff2); |
| 2144 | |
| 2145 | fscal = felec; |
| 2146 | |
| 2147 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 2148 | |
| 2149 | /* Calculate temporary vectorial force */ |
| 2150 | tx = _mm_mul_ps(fscal,dx11); |
| 2151 | ty = _mm_mul_ps(fscal,dy11); |
| 2152 | tz = _mm_mul_ps(fscal,dz11); |
| 2153 | |
| 2154 | /* Update vectorial force */ |
| 2155 | fix1 = _mm_add_ps(fix1,tx); |
| 2156 | fiy1 = _mm_add_ps(fiy1,ty); |
| 2157 | fiz1 = _mm_add_ps(fiz1,tz); |
| 2158 | |
| 2159 | fjx1 = _mm_add_ps(fjx1,tx); |
| 2160 | fjy1 = _mm_add_ps(fjy1,ty); |
| 2161 | fjz1 = _mm_add_ps(fjz1,tz); |
| 2162 | |
| 2163 | } |
| 2164 | |
| 2165 | /************************** |
| 2166 | * CALCULATE INTERACTIONS * |
| 2167 | **************************/ |
| 2168 | |
| 2169 | if (gmx_mm_any_lt(rsq12,rcutoff2)) |
| 2170 | { |
| 2171 | |
| 2172 | r12 = _mm_mul_ps(rsq12,rinv12); |
| 2173 | |
| 2174 | /* EWALD ELECTROSTATICS */ |
| 2175 | |
| 2176 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 2177 | ewrt = _mm_mul_ps(r12,ewtabscale); |
| 2178 | ewitab = _mm_cvttps_epi32(ewrt); |
| 2179 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 2180 | ewitab = _mm_slli_epi32(ewitab,2); |
| 2181 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 2182 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 2183 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 2184 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 2185 | _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); |
| 2186 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 2187 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 2188 | velec = _mm_mul_ps(qq12,_mm_sub_ps(rinv12,velec)); |
| 2189 | felec = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec)); |
| 2190 | |
| 2191 | d = _mm_sub_ps(r12,rswitch); |
| 2192 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 2193 | d2 = _mm_mul_ps(d,d); |
| 2194 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
| 2195 | |
| 2196 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 2197 | |
| 2198 | /* Evaluate switch function */ |
| 2199 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 2200 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv12,_mm_mul_ps(velec,dsw)) ); |
| 2201 | cutoff_mask = _mm_cmplt_ps(rsq12,rcutoff2); |
| 2202 | |
| 2203 | fscal = felec; |
| 2204 | |
| 2205 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 2206 | |
| 2207 | /* Calculate temporary vectorial force */ |
| 2208 | tx = _mm_mul_ps(fscal,dx12); |
| 2209 | ty = _mm_mul_ps(fscal,dy12); |
| 2210 | tz = _mm_mul_ps(fscal,dz12); |
| 2211 | |
| 2212 | /* Update vectorial force */ |
| 2213 | fix1 = _mm_add_ps(fix1,tx); |
| 2214 | fiy1 = _mm_add_ps(fiy1,ty); |
| 2215 | fiz1 = _mm_add_ps(fiz1,tz); |
| 2216 | |
| 2217 | fjx2 = _mm_add_ps(fjx2,tx); |
| 2218 | fjy2 = _mm_add_ps(fjy2,ty); |
| 2219 | fjz2 = _mm_add_ps(fjz2,tz); |
| 2220 | |
| 2221 | } |
| 2222 | |
| 2223 | /************************** |
| 2224 | * CALCULATE INTERACTIONS * |
| 2225 | **************************/ |
| 2226 | |
| 2227 | if (gmx_mm_any_lt(rsq13,rcutoff2)) |
| 2228 | { |
| 2229 | |
| 2230 | r13 = _mm_mul_ps(rsq13,rinv13); |
| 2231 | |
| 2232 | /* EWALD ELECTROSTATICS */ |
| 2233 | |
| 2234 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 2235 | ewrt = _mm_mul_ps(r13,ewtabscale); |
| 2236 | ewitab = _mm_cvttps_epi32(ewrt); |
| 2237 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 2238 | ewitab = _mm_slli_epi32(ewitab,2); |
| 2239 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 2240 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 2241 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 2242 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 2243 | _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); |
| 2244 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 2245 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 2246 | velec = _mm_mul_ps(qq13,_mm_sub_ps(rinv13,velec)); |
| 2247 | felec = _mm_mul_ps(_mm_mul_ps(qq13,rinv13),_mm_sub_ps(rinvsq13,felec)); |
| 2248 | |
| 2249 | d = _mm_sub_ps(r13,rswitch); |
| 2250 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 2251 | d2 = _mm_mul_ps(d,d); |
| 2252 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
| 2253 | |
| 2254 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 2255 | |
| 2256 | /* Evaluate switch function */ |
| 2257 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 2258 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv13,_mm_mul_ps(velec,dsw)) ); |
| 2259 | cutoff_mask = _mm_cmplt_ps(rsq13,rcutoff2); |
| 2260 | |
| 2261 | fscal = felec; |
| 2262 | |
| 2263 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 2264 | |
| 2265 | /* Calculate temporary vectorial force */ |
| 2266 | tx = _mm_mul_ps(fscal,dx13); |
| 2267 | ty = _mm_mul_ps(fscal,dy13); |
| 2268 | tz = _mm_mul_ps(fscal,dz13); |
| 2269 | |
| 2270 | /* Update vectorial force */ |
| 2271 | fix1 = _mm_add_ps(fix1,tx); |
| 2272 | fiy1 = _mm_add_ps(fiy1,ty); |
| 2273 | fiz1 = _mm_add_ps(fiz1,tz); |
| 2274 | |
| 2275 | fjx3 = _mm_add_ps(fjx3,tx); |
| 2276 | fjy3 = _mm_add_ps(fjy3,ty); |
| 2277 | fjz3 = _mm_add_ps(fjz3,tz); |
| 2278 | |
| 2279 | } |
| 2280 | |
| 2281 | /************************** |
| 2282 | * CALCULATE INTERACTIONS * |
| 2283 | **************************/ |
| 2284 | |
| 2285 | if (gmx_mm_any_lt(rsq21,rcutoff2)) |
| 2286 | { |
| 2287 | |
| 2288 | r21 = _mm_mul_ps(rsq21,rinv21); |
| 2289 | |
| 2290 | /* EWALD ELECTROSTATICS */ |
| 2291 | |
| 2292 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 2293 | ewrt = _mm_mul_ps(r21,ewtabscale); |
| 2294 | ewitab = _mm_cvttps_epi32(ewrt); |
| 2295 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 2296 | ewitab = _mm_slli_epi32(ewitab,2); |
| 2297 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 2298 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 2299 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 2300 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 2301 | _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); |
| 2302 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 2303 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 2304 | velec = _mm_mul_ps(qq21,_mm_sub_ps(rinv21,velec)); |
| 2305 | felec = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec)); |
| 2306 | |
| 2307 | d = _mm_sub_ps(r21,rswitch); |
| 2308 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 2309 | d2 = _mm_mul_ps(d,d); |
| 2310 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
| 2311 | |
| 2312 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 2313 | |
| 2314 | /* Evaluate switch function */ |
| 2315 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 2316 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv21,_mm_mul_ps(velec,dsw)) ); |
| 2317 | cutoff_mask = _mm_cmplt_ps(rsq21,rcutoff2); |
| 2318 | |
| 2319 | fscal = felec; |
| 2320 | |
| 2321 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 2322 | |
| 2323 | /* Calculate temporary vectorial force */ |
| 2324 | tx = _mm_mul_ps(fscal,dx21); |
| 2325 | ty = _mm_mul_ps(fscal,dy21); |
| 2326 | tz = _mm_mul_ps(fscal,dz21); |
| 2327 | |
| 2328 | /* Update vectorial force */ |
| 2329 | fix2 = _mm_add_ps(fix2,tx); |
| 2330 | fiy2 = _mm_add_ps(fiy2,ty); |
| 2331 | fiz2 = _mm_add_ps(fiz2,tz); |
| 2332 | |
| 2333 | fjx1 = _mm_add_ps(fjx1,tx); |
| 2334 | fjy1 = _mm_add_ps(fjy1,ty); |
| 2335 | fjz1 = _mm_add_ps(fjz1,tz); |
| 2336 | |
| 2337 | } |
| 2338 | |
| 2339 | /************************** |
| 2340 | * CALCULATE INTERACTIONS * |
| 2341 | **************************/ |
| 2342 | |
| 2343 | if (gmx_mm_any_lt(rsq22,rcutoff2)) |
| 2344 | { |
| 2345 | |
| 2346 | r22 = _mm_mul_ps(rsq22,rinv22); |
| 2347 | |
| 2348 | /* EWALD ELECTROSTATICS */ |
| 2349 | |
| 2350 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 2351 | ewrt = _mm_mul_ps(r22,ewtabscale); |
| 2352 | ewitab = _mm_cvttps_epi32(ewrt); |
| 2353 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 2354 | ewitab = _mm_slli_epi32(ewitab,2); |
| 2355 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 2356 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 2357 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 2358 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 2359 | _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); |
| 2360 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 2361 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 2362 | velec = _mm_mul_ps(qq22,_mm_sub_ps(rinv22,velec)); |
| 2363 | felec = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec)); |
| 2364 | |
| 2365 | d = _mm_sub_ps(r22,rswitch); |
| 2366 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 2367 | d2 = _mm_mul_ps(d,d); |
| 2368 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
| 2369 | |
| 2370 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 2371 | |
| 2372 | /* Evaluate switch function */ |
| 2373 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 2374 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv22,_mm_mul_ps(velec,dsw)) ); |
| 2375 | cutoff_mask = _mm_cmplt_ps(rsq22,rcutoff2); |
| 2376 | |
| 2377 | fscal = felec; |
| 2378 | |
| 2379 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 2380 | |
| 2381 | /* Calculate temporary vectorial force */ |
| 2382 | tx = _mm_mul_ps(fscal,dx22); |
| 2383 | ty = _mm_mul_ps(fscal,dy22); |
| 2384 | tz = _mm_mul_ps(fscal,dz22); |
| 2385 | |
| 2386 | /* Update vectorial force */ |
| 2387 | fix2 = _mm_add_ps(fix2,tx); |
| 2388 | fiy2 = _mm_add_ps(fiy2,ty); |
| 2389 | fiz2 = _mm_add_ps(fiz2,tz); |
| 2390 | |
| 2391 | fjx2 = _mm_add_ps(fjx2,tx); |
| 2392 | fjy2 = _mm_add_ps(fjy2,ty); |
| 2393 | fjz2 = _mm_add_ps(fjz2,tz); |
| 2394 | |
| 2395 | } |
| 2396 | |
| 2397 | /************************** |
| 2398 | * CALCULATE INTERACTIONS * |
| 2399 | **************************/ |
| 2400 | |
| 2401 | if (gmx_mm_any_lt(rsq23,rcutoff2)) |
| 2402 | { |
| 2403 | |
| 2404 | r23 = _mm_mul_ps(rsq23,rinv23); |
| 2405 | |
| 2406 | /* EWALD ELECTROSTATICS */ |
| 2407 | |
| 2408 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 2409 | ewrt = _mm_mul_ps(r23,ewtabscale); |
| 2410 | ewitab = _mm_cvttps_epi32(ewrt); |
| 2411 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 2412 | ewitab = _mm_slli_epi32(ewitab,2); |
| 2413 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 2414 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 2415 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 2416 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 2417 | _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); |
| 2418 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 2419 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 2420 | velec = _mm_mul_ps(qq23,_mm_sub_ps(rinv23,velec)); |
| 2421 | felec = _mm_mul_ps(_mm_mul_ps(qq23,rinv23),_mm_sub_ps(rinvsq23,felec)); |
| 2422 | |
| 2423 | d = _mm_sub_ps(r23,rswitch); |
| 2424 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 2425 | d2 = _mm_mul_ps(d,d); |
| 2426 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
| 2427 | |
| 2428 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 2429 | |
| 2430 | /* Evaluate switch function */ |
| 2431 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 2432 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv23,_mm_mul_ps(velec,dsw)) ); |
| 2433 | cutoff_mask = _mm_cmplt_ps(rsq23,rcutoff2); |
| 2434 | |
| 2435 | fscal = felec; |
| 2436 | |
| 2437 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 2438 | |
| 2439 | /* Calculate temporary vectorial force */ |
| 2440 | tx = _mm_mul_ps(fscal,dx23); |
| 2441 | ty = _mm_mul_ps(fscal,dy23); |
| 2442 | tz = _mm_mul_ps(fscal,dz23); |
| 2443 | |
| 2444 | /* Update vectorial force */ |
| 2445 | fix2 = _mm_add_ps(fix2,tx); |
| 2446 | fiy2 = _mm_add_ps(fiy2,ty); |
| 2447 | fiz2 = _mm_add_ps(fiz2,tz); |
| 2448 | |
| 2449 | fjx3 = _mm_add_ps(fjx3,tx); |
| 2450 | fjy3 = _mm_add_ps(fjy3,ty); |
| 2451 | fjz3 = _mm_add_ps(fjz3,tz); |
| 2452 | |
| 2453 | } |
| 2454 | |
| 2455 | /************************** |
| 2456 | * CALCULATE INTERACTIONS * |
| 2457 | **************************/ |
| 2458 | |
| 2459 | if (gmx_mm_any_lt(rsq31,rcutoff2)) |
| 2460 | { |
| 2461 | |
| 2462 | r31 = _mm_mul_ps(rsq31,rinv31); |
| 2463 | |
| 2464 | /* EWALD ELECTROSTATICS */ |
| 2465 | |
| 2466 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 2467 | ewrt = _mm_mul_ps(r31,ewtabscale); |
| 2468 | ewitab = _mm_cvttps_epi32(ewrt); |
| 2469 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 2470 | ewitab = _mm_slli_epi32(ewitab,2); |
| 2471 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 2472 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 2473 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 2474 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 2475 | _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); |
| 2476 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 2477 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 2478 | velec = _mm_mul_ps(qq31,_mm_sub_ps(rinv31,velec)); |
| 2479 | felec = _mm_mul_ps(_mm_mul_ps(qq31,rinv31),_mm_sub_ps(rinvsq31,felec)); |
| 2480 | |
| 2481 | d = _mm_sub_ps(r31,rswitch); |
| 2482 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 2483 | d2 = _mm_mul_ps(d,d); |
| 2484 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
| 2485 | |
| 2486 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 2487 | |
| 2488 | /* Evaluate switch function */ |
| 2489 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 2490 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv31,_mm_mul_ps(velec,dsw)) ); |
| 2491 | cutoff_mask = _mm_cmplt_ps(rsq31,rcutoff2); |
| 2492 | |
| 2493 | fscal = felec; |
| 2494 | |
| 2495 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 2496 | |
| 2497 | /* Calculate temporary vectorial force */ |
| 2498 | tx = _mm_mul_ps(fscal,dx31); |
| 2499 | ty = _mm_mul_ps(fscal,dy31); |
| 2500 | tz = _mm_mul_ps(fscal,dz31); |
| 2501 | |
| 2502 | /* Update vectorial force */ |
| 2503 | fix3 = _mm_add_ps(fix3,tx); |
| 2504 | fiy3 = _mm_add_ps(fiy3,ty); |
| 2505 | fiz3 = _mm_add_ps(fiz3,tz); |
| 2506 | |
| 2507 | fjx1 = _mm_add_ps(fjx1,tx); |
| 2508 | fjy1 = _mm_add_ps(fjy1,ty); |
| 2509 | fjz1 = _mm_add_ps(fjz1,tz); |
| 2510 | |
| 2511 | } |
| 2512 | |
| 2513 | /************************** |
| 2514 | * CALCULATE INTERACTIONS * |
| 2515 | **************************/ |
| 2516 | |
| 2517 | if (gmx_mm_any_lt(rsq32,rcutoff2)) |
| 2518 | { |
| 2519 | |
| 2520 | r32 = _mm_mul_ps(rsq32,rinv32); |
| 2521 | |
| 2522 | /* EWALD ELECTROSTATICS */ |
| 2523 | |
| 2524 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 2525 | ewrt = _mm_mul_ps(r32,ewtabscale); |
| 2526 | ewitab = _mm_cvttps_epi32(ewrt); |
| 2527 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 2528 | ewitab = _mm_slli_epi32(ewitab,2); |
| 2529 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 2530 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 2531 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 2532 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 2533 | _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); |
| 2534 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 2535 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 2536 | velec = _mm_mul_ps(qq32,_mm_sub_ps(rinv32,velec)); |
| 2537 | felec = _mm_mul_ps(_mm_mul_ps(qq32,rinv32),_mm_sub_ps(rinvsq32,felec)); |
| 2538 | |
| 2539 | d = _mm_sub_ps(r32,rswitch); |
| 2540 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 2541 | d2 = _mm_mul_ps(d,d); |
| 2542 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
| 2543 | |
| 2544 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 2545 | |
| 2546 | /* Evaluate switch function */ |
| 2547 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 2548 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv32,_mm_mul_ps(velec,dsw)) ); |
| 2549 | cutoff_mask = _mm_cmplt_ps(rsq32,rcutoff2); |
| 2550 | |
| 2551 | fscal = felec; |
| 2552 | |
| 2553 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 2554 | |
| 2555 | /* Calculate temporary vectorial force */ |
| 2556 | tx = _mm_mul_ps(fscal,dx32); |
| 2557 | ty = _mm_mul_ps(fscal,dy32); |
| 2558 | tz = _mm_mul_ps(fscal,dz32); |
| 2559 | |
| 2560 | /* Update vectorial force */ |
| 2561 | fix3 = _mm_add_ps(fix3,tx); |
| 2562 | fiy3 = _mm_add_ps(fiy3,ty); |
| 2563 | fiz3 = _mm_add_ps(fiz3,tz); |
| 2564 | |
| 2565 | fjx2 = _mm_add_ps(fjx2,tx); |
| 2566 | fjy2 = _mm_add_ps(fjy2,ty); |
| 2567 | fjz2 = _mm_add_ps(fjz2,tz); |
| 2568 | |
| 2569 | } |
| 2570 | |
| 2571 | /************************** |
| 2572 | * CALCULATE INTERACTIONS * |
| 2573 | **************************/ |
| 2574 | |
| 2575 | if (gmx_mm_any_lt(rsq33,rcutoff2)) |
| 2576 | { |
| 2577 | |
| 2578 | r33 = _mm_mul_ps(rsq33,rinv33); |
| 2579 | |
| 2580 | /* EWALD ELECTROSTATICS */ |
| 2581 | |
| 2582 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 2583 | ewrt = _mm_mul_ps(r33,ewtabscale); |
| 2584 | ewitab = _mm_cvttps_epi32(ewrt); |
| 2585 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 2586 | ewitab = _mm_slli_epi32(ewitab,2); |
| 2587 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 2588 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 2589 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 2590 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 2591 | _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); |
| 2592 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 2593 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 2594 | velec = _mm_mul_ps(qq33,_mm_sub_ps(rinv33,velec)); |
| 2595 | felec = _mm_mul_ps(_mm_mul_ps(qq33,rinv33),_mm_sub_ps(rinvsq33,felec)); |
| 2596 | |
| 2597 | d = _mm_sub_ps(r33,rswitch); |
| 2598 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 2599 | d2 = _mm_mul_ps(d,d); |
| 2600 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
| 2601 | |
| 2602 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 2603 | |
| 2604 | /* Evaluate switch function */ |
| 2605 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 2606 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv33,_mm_mul_ps(velec,dsw)) ); |
| 2607 | cutoff_mask = _mm_cmplt_ps(rsq33,rcutoff2); |
| 2608 | |
| 2609 | fscal = felec; |
| 2610 | |
| 2611 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 2612 | |
| 2613 | /* Calculate temporary vectorial force */ |
| 2614 | tx = _mm_mul_ps(fscal,dx33); |
| 2615 | ty = _mm_mul_ps(fscal,dy33); |
| 2616 | tz = _mm_mul_ps(fscal,dz33); |
| 2617 | |
| 2618 | /* Update vectorial force */ |
| 2619 | fix3 = _mm_add_ps(fix3,tx); |
| 2620 | fiy3 = _mm_add_ps(fiy3,ty); |
| 2621 | fiz3 = _mm_add_ps(fiz3,tz); |
| 2622 | |
| 2623 | fjx3 = _mm_add_ps(fjx3,tx); |
| 2624 | fjy3 = _mm_add_ps(fjy3,ty); |
| 2625 | fjz3 = _mm_add_ps(fjz3,tz); |
| 2626 | |
| 2627 | } |
| 2628 | |
| 2629 | fjptrA = f+j_coord_offsetA; |
| 2630 | fjptrB = f+j_coord_offsetB; |
| 2631 | fjptrC = f+j_coord_offsetC; |
| 2632 | fjptrD = f+j_coord_offsetD; |
| 2633 | |
| 2634 | gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD, |
| 2635 | fjx0,fjy0,fjz0,fjx1,fjy1,fjz1, |
| 2636 | fjx2,fjy2,fjz2,fjx3,fjy3,fjz3); |
| 2637 | |
| 2638 | /* Inner loop uses 617 flops */ |
| 2639 | } |
| 2640 | |
| 2641 | if(jidx<j_index_end) |
| 2642 | { |
| 2643 | |
| 2644 | /* Get j neighbor index, and coordinate index */ |
| 2645 | jnrlistA = jjnr[jidx]; |
| 2646 | jnrlistB = jjnr[jidx+1]; |
| 2647 | jnrlistC = jjnr[jidx+2]; |
| 2648 | jnrlistD = jjnr[jidx+3]; |
| 2649 | /* Sign of each element will be negative for non-real atoms. |
| 2650 | * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones, |
| 2651 | * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries. |
| 2652 | */ |
| 2653 | dummy_mask = gmx_mm_castsi128_ps_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())); |
| 2654 | jnrA = (jnrlistA>=0) ? jnrlistA : 0; |
| 2655 | jnrB = (jnrlistB>=0) ? jnrlistB : 0; |
| 2656 | jnrC = (jnrlistC>=0) ? jnrlistC : 0; |
| 2657 | jnrD = (jnrlistD>=0) ? jnrlistD : 0; |
| 2658 | j_coord_offsetA = DIM3*jnrA; |
| 2659 | j_coord_offsetB = DIM3*jnrB; |
| 2660 | j_coord_offsetC = DIM3*jnrC; |
| 2661 | j_coord_offsetD = DIM3*jnrD; |
| 2662 | |
| 2663 | /* load j atom coordinates */ |
| 2664 | gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB, |
| 2665 | x+j_coord_offsetC,x+j_coord_offsetD, |
| 2666 | &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2, |
| 2667 | &jy2,&jz2,&jx3,&jy3,&jz3); |
| 2668 | |
| 2669 | /* Calculate displacement vector */ |
| 2670 | dx00 = _mm_sub_ps(ix0,jx0); |
| 2671 | dy00 = _mm_sub_ps(iy0,jy0); |
| 2672 | dz00 = _mm_sub_ps(iz0,jz0); |
| 2673 | dx11 = _mm_sub_ps(ix1,jx1); |
| 2674 | dy11 = _mm_sub_ps(iy1,jy1); |
| 2675 | dz11 = _mm_sub_ps(iz1,jz1); |
| 2676 | dx12 = _mm_sub_ps(ix1,jx2); |
| 2677 | dy12 = _mm_sub_ps(iy1,jy2); |
| 2678 | dz12 = _mm_sub_ps(iz1,jz2); |
| 2679 | dx13 = _mm_sub_ps(ix1,jx3); |
| 2680 | dy13 = _mm_sub_ps(iy1,jy3); |
| 2681 | dz13 = _mm_sub_ps(iz1,jz3); |
| 2682 | dx21 = _mm_sub_ps(ix2,jx1); |
| 2683 | dy21 = _mm_sub_ps(iy2,jy1); |
| 2684 | dz21 = _mm_sub_ps(iz2,jz1); |
| 2685 | dx22 = _mm_sub_ps(ix2,jx2); |
| 2686 | dy22 = _mm_sub_ps(iy2,jy2); |
| 2687 | dz22 = _mm_sub_ps(iz2,jz2); |
| 2688 | dx23 = _mm_sub_ps(ix2,jx3); |
| 2689 | dy23 = _mm_sub_ps(iy2,jy3); |
| 2690 | dz23 = _mm_sub_ps(iz2,jz3); |
| 2691 | dx31 = _mm_sub_ps(ix3,jx1); |
| 2692 | dy31 = _mm_sub_ps(iy3,jy1); |
| 2693 | dz31 = _mm_sub_ps(iz3,jz1); |
| 2694 | dx32 = _mm_sub_ps(ix3,jx2); |
| 2695 | dy32 = _mm_sub_ps(iy3,jy2); |
| 2696 | dz32 = _mm_sub_ps(iz3,jz2); |
| 2697 | dx33 = _mm_sub_ps(ix3,jx3); |
| 2698 | dy33 = _mm_sub_ps(iy3,jy3); |
| 2699 | dz33 = _mm_sub_ps(iz3,jz3); |
| 2700 | |
| 2701 | /* Calculate squared distance and things based on it */ |
| 2702 | rsq00 = gmx_mm_calc_rsq_ps(dx00,dy00,dz00); |
| 2703 | rsq11 = gmx_mm_calc_rsq_ps(dx11,dy11,dz11); |
| 2704 | rsq12 = gmx_mm_calc_rsq_ps(dx12,dy12,dz12); |
| 2705 | rsq13 = gmx_mm_calc_rsq_ps(dx13,dy13,dz13); |
| 2706 | rsq21 = gmx_mm_calc_rsq_ps(dx21,dy21,dz21); |
| 2707 | rsq22 = gmx_mm_calc_rsq_ps(dx22,dy22,dz22); |
| 2708 | rsq23 = gmx_mm_calc_rsq_ps(dx23,dy23,dz23); |
| 2709 | rsq31 = gmx_mm_calc_rsq_ps(dx31,dy31,dz31); |
| 2710 | rsq32 = gmx_mm_calc_rsq_ps(dx32,dy32,dz32); |
| 2711 | rsq33 = gmx_mm_calc_rsq_ps(dx33,dy33,dz33); |
| 2712 | |
| 2713 | rinv00 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq00); |
| 2714 | rinv11 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq11); |
| 2715 | rinv12 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq12); |
| 2716 | rinv13 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq13); |
| 2717 | rinv21 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq21); |
| 2718 | rinv22 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq22); |
| 2719 | rinv23 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq23); |
| 2720 | rinv31 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq31); |
| 2721 | rinv32 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq32); |
| 2722 | rinv33 = gmx_mm_invsqrt_psgmx_simd_invsqrt_f(rsq33); |
| 2723 | |
| 2724 | rinvsq00 = _mm_mul_ps(rinv00,rinv00); |
| 2725 | rinvsq11 = _mm_mul_ps(rinv11,rinv11); |
| 2726 | rinvsq12 = _mm_mul_ps(rinv12,rinv12); |
| 2727 | rinvsq13 = _mm_mul_ps(rinv13,rinv13); |
| 2728 | rinvsq21 = _mm_mul_ps(rinv21,rinv21); |
| 2729 | rinvsq22 = _mm_mul_ps(rinv22,rinv22); |
| 2730 | rinvsq23 = _mm_mul_ps(rinv23,rinv23); |
| 2731 | rinvsq31 = _mm_mul_ps(rinv31,rinv31); |
| 2732 | rinvsq32 = _mm_mul_ps(rinv32,rinv32); |
| 2733 | rinvsq33 = _mm_mul_ps(rinv33,rinv33); |
| 2734 | |
| 2735 | fjx0 = _mm_setzero_ps(); |
| 2736 | fjy0 = _mm_setzero_ps(); |
| 2737 | fjz0 = _mm_setzero_ps(); |
| 2738 | fjx1 = _mm_setzero_ps(); |
| 2739 | fjy1 = _mm_setzero_ps(); |
| 2740 | fjz1 = _mm_setzero_ps(); |
| 2741 | fjx2 = _mm_setzero_ps(); |
| 2742 | fjy2 = _mm_setzero_ps(); |
| 2743 | fjz2 = _mm_setzero_ps(); |
| 2744 | fjx3 = _mm_setzero_ps(); |
| 2745 | fjy3 = _mm_setzero_ps(); |
| 2746 | fjz3 = _mm_setzero_ps(); |
| 2747 | |
| 2748 | /************************** |
| 2749 | * CALCULATE INTERACTIONS * |
| 2750 | **************************/ |
| 2751 | |
| 2752 | if (gmx_mm_any_lt(rsq00,rcutoff2)) |
| 2753 | { |
| 2754 | |
| 2755 | r00 = _mm_mul_ps(rsq00,rinv00); |
| 2756 | r00 = _mm_andnot_ps(dummy_mask,r00); |
| 2757 | |
| 2758 | /* LENNARD-JONES DISPERSION/REPULSION */ |
| 2759 | |
| 2760 | rinvsix = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00); |
| 2761 | vvdw6 = _mm_mul_ps(c6_00,rinvsix); |
| 2762 | vvdw12 = _mm_mul_ps(c12_00,_mm_mul_ps(rinvsix,rinvsix)); |
| 2763 | vvdw = _mm_sub_ps( _mm_mul_ps(vvdw12,one_twelfth) , _mm_mul_ps(vvdw6,one_sixth) ); |
| 2764 | fvdw = _mm_mul_ps(_mm_sub_ps(vvdw12,vvdw6),rinvsq00); |
| 2765 | |
| 2766 | d = _mm_sub_ps(r00,rswitch); |
| 2767 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 2768 | d2 = _mm_mul_ps(d,d); |
| 2769 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
| 2770 | |
| 2771 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 2772 | |
| 2773 | /* Evaluate switch function */ |
| 2774 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 2775 | fvdw = _mm_sub_ps( _mm_mul_ps(fvdw,sw) , _mm_mul_ps(rinv00,_mm_mul_ps(vvdw,dsw)) ); |
| 2776 | cutoff_mask = _mm_cmplt_ps(rsq00,rcutoff2); |
| 2777 | |
| 2778 | fscal = fvdw; |
| 2779 | |
| 2780 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 2781 | |
| 2782 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 2783 | |
| 2784 | /* Calculate temporary vectorial force */ |
| 2785 | tx = _mm_mul_ps(fscal,dx00); |
| 2786 | ty = _mm_mul_ps(fscal,dy00); |
| 2787 | tz = _mm_mul_ps(fscal,dz00); |
| 2788 | |
| 2789 | /* Update vectorial force */ |
| 2790 | fix0 = _mm_add_ps(fix0,tx); |
| 2791 | fiy0 = _mm_add_ps(fiy0,ty); |
| 2792 | fiz0 = _mm_add_ps(fiz0,tz); |
| 2793 | |
| 2794 | fjx0 = _mm_add_ps(fjx0,tx); |
| 2795 | fjy0 = _mm_add_ps(fjy0,ty); |
| 2796 | fjz0 = _mm_add_ps(fjz0,tz); |
| 2797 | |
| 2798 | } |
| 2799 | |
| 2800 | /************************** |
| 2801 | * CALCULATE INTERACTIONS * |
| 2802 | **************************/ |
| 2803 | |
| 2804 | if (gmx_mm_any_lt(rsq11,rcutoff2)) |
| 2805 | { |
| 2806 | |
| 2807 | r11 = _mm_mul_ps(rsq11,rinv11); |
| 2808 | r11 = _mm_andnot_ps(dummy_mask,r11); |
| 2809 | |
| 2810 | /* EWALD ELECTROSTATICS */ |
| 2811 | |
| 2812 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 2813 | ewrt = _mm_mul_ps(r11,ewtabscale); |
| 2814 | ewitab = _mm_cvttps_epi32(ewrt); |
| 2815 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 2816 | ewitab = _mm_slli_epi32(ewitab,2); |
| 2817 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 2818 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 2819 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 2820 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 2821 | _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); |
| 2822 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 2823 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 2824 | velec = _mm_mul_ps(qq11,_mm_sub_ps(rinv11,velec)); |
| 2825 | felec = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec)); |
| 2826 | |
| 2827 | d = _mm_sub_ps(r11,rswitch); |
| 2828 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 2829 | d2 = _mm_mul_ps(d,d); |
| 2830 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
| 2831 | |
| 2832 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 2833 | |
| 2834 | /* Evaluate switch function */ |
| 2835 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 2836 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv11,_mm_mul_ps(velec,dsw)) ); |
| 2837 | cutoff_mask = _mm_cmplt_ps(rsq11,rcutoff2); |
| 2838 | |
| 2839 | fscal = felec; |
| 2840 | |
| 2841 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 2842 | |
| 2843 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 2844 | |
| 2845 | /* Calculate temporary vectorial force */ |
| 2846 | tx = _mm_mul_ps(fscal,dx11); |
| 2847 | ty = _mm_mul_ps(fscal,dy11); |
| 2848 | tz = _mm_mul_ps(fscal,dz11); |
| 2849 | |
| 2850 | /* Update vectorial force */ |
| 2851 | fix1 = _mm_add_ps(fix1,tx); |
| 2852 | fiy1 = _mm_add_ps(fiy1,ty); |
| 2853 | fiz1 = _mm_add_ps(fiz1,tz); |
| 2854 | |
| 2855 | fjx1 = _mm_add_ps(fjx1,tx); |
| 2856 | fjy1 = _mm_add_ps(fjy1,ty); |
| 2857 | fjz1 = _mm_add_ps(fjz1,tz); |
| 2858 | |
| 2859 | } |
| 2860 | |
| 2861 | /************************** |
| 2862 | * CALCULATE INTERACTIONS * |
| 2863 | **************************/ |
| 2864 | |
| 2865 | if (gmx_mm_any_lt(rsq12,rcutoff2)) |
| 2866 | { |
| 2867 | |
| 2868 | r12 = _mm_mul_ps(rsq12,rinv12); |
| 2869 | r12 = _mm_andnot_ps(dummy_mask,r12); |
| 2870 | |
| 2871 | /* EWALD ELECTROSTATICS */ |
| 2872 | |
| 2873 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 2874 | ewrt = _mm_mul_ps(r12,ewtabscale); |
| 2875 | ewitab = _mm_cvttps_epi32(ewrt); |
| 2876 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 2877 | ewitab = _mm_slli_epi32(ewitab,2); |
| 2878 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 2879 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 2880 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 2881 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 2882 | _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); |
| 2883 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 2884 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 2885 | velec = _mm_mul_ps(qq12,_mm_sub_ps(rinv12,velec)); |
| 2886 | felec = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec)); |
| 2887 | |
| 2888 | d = _mm_sub_ps(r12,rswitch); |
| 2889 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 2890 | d2 = _mm_mul_ps(d,d); |
| 2891 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
| 2892 | |
| 2893 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 2894 | |
| 2895 | /* Evaluate switch function */ |
| 2896 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 2897 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv12,_mm_mul_ps(velec,dsw)) ); |
| 2898 | cutoff_mask = _mm_cmplt_ps(rsq12,rcutoff2); |
| 2899 | |
| 2900 | fscal = felec; |
| 2901 | |
| 2902 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 2903 | |
| 2904 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 2905 | |
| 2906 | /* Calculate temporary vectorial force */ |
| 2907 | tx = _mm_mul_ps(fscal,dx12); |
| 2908 | ty = _mm_mul_ps(fscal,dy12); |
| 2909 | tz = _mm_mul_ps(fscal,dz12); |
| 2910 | |
| 2911 | /* Update vectorial force */ |
| 2912 | fix1 = _mm_add_ps(fix1,tx); |
| 2913 | fiy1 = _mm_add_ps(fiy1,ty); |
| 2914 | fiz1 = _mm_add_ps(fiz1,tz); |
| 2915 | |
| 2916 | fjx2 = _mm_add_ps(fjx2,tx); |
| 2917 | fjy2 = _mm_add_ps(fjy2,ty); |
| 2918 | fjz2 = _mm_add_ps(fjz2,tz); |
| 2919 | |
| 2920 | } |
| 2921 | |
| 2922 | /************************** |
| 2923 | * CALCULATE INTERACTIONS * |
| 2924 | **************************/ |
| 2925 | |
| 2926 | if (gmx_mm_any_lt(rsq13,rcutoff2)) |
| 2927 | { |
| 2928 | |
| 2929 | r13 = _mm_mul_ps(rsq13,rinv13); |
| 2930 | r13 = _mm_andnot_ps(dummy_mask,r13); |
| 2931 | |
| 2932 | /* EWALD ELECTROSTATICS */ |
| 2933 | |
| 2934 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 2935 | ewrt = _mm_mul_ps(r13,ewtabscale); |
| 2936 | ewitab = _mm_cvttps_epi32(ewrt); |
| 2937 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 2938 | ewitab = _mm_slli_epi32(ewitab,2); |
| 2939 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 2940 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 2941 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 2942 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 2943 | _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); |
| 2944 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 2945 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 2946 | velec = _mm_mul_ps(qq13,_mm_sub_ps(rinv13,velec)); |
| 2947 | felec = _mm_mul_ps(_mm_mul_ps(qq13,rinv13),_mm_sub_ps(rinvsq13,felec)); |
| 2948 | |
| 2949 | d = _mm_sub_ps(r13,rswitch); |
| 2950 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 2951 | d2 = _mm_mul_ps(d,d); |
| 2952 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
| 2953 | |
| 2954 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 2955 | |
| 2956 | /* Evaluate switch function */ |
| 2957 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 2958 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv13,_mm_mul_ps(velec,dsw)) ); |
| 2959 | cutoff_mask = _mm_cmplt_ps(rsq13,rcutoff2); |
| 2960 | |
| 2961 | fscal = felec; |
| 2962 | |
| 2963 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 2964 | |
| 2965 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 2966 | |
| 2967 | /* Calculate temporary vectorial force */ |
| 2968 | tx = _mm_mul_ps(fscal,dx13); |
| 2969 | ty = _mm_mul_ps(fscal,dy13); |
| 2970 | tz = _mm_mul_ps(fscal,dz13); |
| 2971 | |
| 2972 | /* Update vectorial force */ |
| 2973 | fix1 = _mm_add_ps(fix1,tx); |
| 2974 | fiy1 = _mm_add_ps(fiy1,ty); |
| 2975 | fiz1 = _mm_add_ps(fiz1,tz); |
| 2976 | |
| 2977 | fjx3 = _mm_add_ps(fjx3,tx); |
| 2978 | fjy3 = _mm_add_ps(fjy3,ty); |
| 2979 | fjz3 = _mm_add_ps(fjz3,tz); |
| 2980 | |
| 2981 | } |
| 2982 | |
| 2983 | /************************** |
| 2984 | * CALCULATE INTERACTIONS * |
| 2985 | **************************/ |
| 2986 | |
| 2987 | if (gmx_mm_any_lt(rsq21,rcutoff2)) |
| 2988 | { |
| 2989 | |
| 2990 | r21 = _mm_mul_ps(rsq21,rinv21); |
| 2991 | r21 = _mm_andnot_ps(dummy_mask,r21); |
| 2992 | |
| 2993 | /* EWALD ELECTROSTATICS */ |
| 2994 | |
| 2995 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 2996 | ewrt = _mm_mul_ps(r21,ewtabscale); |
| 2997 | ewitab = _mm_cvttps_epi32(ewrt); |
| 2998 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 2999 | ewitab = _mm_slli_epi32(ewitab,2); |
| 3000 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 3001 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 3002 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 3003 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 3004 | _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); |
| 3005 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 3006 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 3007 | velec = _mm_mul_ps(qq21,_mm_sub_ps(rinv21,velec)); |
| 3008 | felec = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec)); |
| 3009 | |
| 3010 | d = _mm_sub_ps(r21,rswitch); |
| 3011 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 3012 | d2 = _mm_mul_ps(d,d); |
| 3013 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
| 3014 | |
| 3015 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 3016 | |
| 3017 | /* Evaluate switch function */ |
| 3018 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 3019 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv21,_mm_mul_ps(velec,dsw)) ); |
| 3020 | cutoff_mask = _mm_cmplt_ps(rsq21,rcutoff2); |
| 3021 | |
| 3022 | fscal = felec; |
| 3023 | |
| 3024 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 3025 | |
| 3026 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 3027 | |
| 3028 | /* Calculate temporary vectorial force */ |
| 3029 | tx = _mm_mul_ps(fscal,dx21); |
| 3030 | ty = _mm_mul_ps(fscal,dy21); |
| 3031 | tz = _mm_mul_ps(fscal,dz21); |
| 3032 | |
| 3033 | /* Update vectorial force */ |
| 3034 | fix2 = _mm_add_ps(fix2,tx); |
| 3035 | fiy2 = _mm_add_ps(fiy2,ty); |
| 3036 | fiz2 = _mm_add_ps(fiz2,tz); |
| 3037 | |
| 3038 | fjx1 = _mm_add_ps(fjx1,tx); |
| 3039 | fjy1 = _mm_add_ps(fjy1,ty); |
| 3040 | fjz1 = _mm_add_ps(fjz1,tz); |
| 3041 | |
| 3042 | } |
| 3043 | |
| 3044 | /************************** |
| 3045 | * CALCULATE INTERACTIONS * |
| 3046 | **************************/ |
| 3047 | |
| 3048 | if (gmx_mm_any_lt(rsq22,rcutoff2)) |
| 3049 | { |
| 3050 | |
| 3051 | r22 = _mm_mul_ps(rsq22,rinv22); |
| 3052 | r22 = _mm_andnot_ps(dummy_mask,r22); |
| 3053 | |
| 3054 | /* EWALD ELECTROSTATICS */ |
| 3055 | |
| 3056 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 3057 | ewrt = _mm_mul_ps(r22,ewtabscale); |
| 3058 | ewitab = _mm_cvttps_epi32(ewrt); |
| 3059 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 3060 | ewitab = _mm_slli_epi32(ewitab,2); |
| 3061 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 3062 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 3063 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 3064 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 3065 | _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); |
| 3066 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 3067 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 3068 | velec = _mm_mul_ps(qq22,_mm_sub_ps(rinv22,velec)); |
| 3069 | felec = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec)); |
| 3070 | |
| 3071 | d = _mm_sub_ps(r22,rswitch); |
| 3072 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 3073 | d2 = _mm_mul_ps(d,d); |
| 3074 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
| 3075 | |
| 3076 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 3077 | |
| 3078 | /* Evaluate switch function */ |
| 3079 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 3080 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv22,_mm_mul_ps(velec,dsw)) ); |
| 3081 | cutoff_mask = _mm_cmplt_ps(rsq22,rcutoff2); |
| 3082 | |
| 3083 | fscal = felec; |
| 3084 | |
| 3085 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 3086 | |
| 3087 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 3088 | |
| 3089 | /* Calculate temporary vectorial force */ |
| 3090 | tx = _mm_mul_ps(fscal,dx22); |
| 3091 | ty = _mm_mul_ps(fscal,dy22); |
| 3092 | tz = _mm_mul_ps(fscal,dz22); |
| 3093 | |
| 3094 | /* Update vectorial force */ |
| 3095 | fix2 = _mm_add_ps(fix2,tx); |
| 3096 | fiy2 = _mm_add_ps(fiy2,ty); |
| 3097 | fiz2 = _mm_add_ps(fiz2,tz); |
| 3098 | |
| 3099 | fjx2 = _mm_add_ps(fjx2,tx); |
| 3100 | fjy2 = _mm_add_ps(fjy2,ty); |
| 3101 | fjz2 = _mm_add_ps(fjz2,tz); |
| 3102 | |
| 3103 | } |
| 3104 | |
| 3105 | /************************** |
| 3106 | * CALCULATE INTERACTIONS * |
| 3107 | **************************/ |
| 3108 | |
| 3109 | if (gmx_mm_any_lt(rsq23,rcutoff2)) |
| 3110 | { |
| 3111 | |
| 3112 | r23 = _mm_mul_ps(rsq23,rinv23); |
| 3113 | r23 = _mm_andnot_ps(dummy_mask,r23); |
| 3114 | |
| 3115 | /* EWALD ELECTROSTATICS */ |
| 3116 | |
| 3117 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 3118 | ewrt = _mm_mul_ps(r23,ewtabscale); |
| 3119 | ewitab = _mm_cvttps_epi32(ewrt); |
| 3120 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 3121 | ewitab = _mm_slli_epi32(ewitab,2); |
| 3122 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 3123 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 3124 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 3125 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 3126 | _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); |
| 3127 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 3128 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 3129 | velec = _mm_mul_ps(qq23,_mm_sub_ps(rinv23,velec)); |
| 3130 | felec = _mm_mul_ps(_mm_mul_ps(qq23,rinv23),_mm_sub_ps(rinvsq23,felec)); |
| 3131 | |
| 3132 | d = _mm_sub_ps(r23,rswitch); |
| 3133 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 3134 | d2 = _mm_mul_ps(d,d); |
| 3135 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
| 3136 | |
| 3137 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 3138 | |
| 3139 | /* Evaluate switch function */ |
| 3140 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 3141 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv23,_mm_mul_ps(velec,dsw)) ); |
| 3142 | cutoff_mask = _mm_cmplt_ps(rsq23,rcutoff2); |
| 3143 | |
| 3144 | fscal = felec; |
| 3145 | |
| 3146 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 3147 | |
| 3148 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 3149 | |
| 3150 | /* Calculate temporary vectorial force */ |
| 3151 | tx = _mm_mul_ps(fscal,dx23); |
| 3152 | ty = _mm_mul_ps(fscal,dy23); |
| 3153 | tz = _mm_mul_ps(fscal,dz23); |
| 3154 | |
| 3155 | /* Update vectorial force */ |
| 3156 | fix2 = _mm_add_ps(fix2,tx); |
| 3157 | fiy2 = _mm_add_ps(fiy2,ty); |
| 3158 | fiz2 = _mm_add_ps(fiz2,tz); |
| 3159 | |
| 3160 | fjx3 = _mm_add_ps(fjx3,tx); |
| 3161 | fjy3 = _mm_add_ps(fjy3,ty); |
| 3162 | fjz3 = _mm_add_ps(fjz3,tz); |
| 3163 | |
| 3164 | } |
| 3165 | |
| 3166 | /************************** |
| 3167 | * CALCULATE INTERACTIONS * |
| 3168 | **************************/ |
| 3169 | |
| 3170 | if (gmx_mm_any_lt(rsq31,rcutoff2)) |
| 3171 | { |
| 3172 | |
| 3173 | r31 = _mm_mul_ps(rsq31,rinv31); |
| 3174 | r31 = _mm_andnot_ps(dummy_mask,r31); |
| 3175 | |
| 3176 | /* EWALD ELECTROSTATICS */ |
| 3177 | |
| 3178 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 3179 | ewrt = _mm_mul_ps(r31,ewtabscale); |
| 3180 | ewitab = _mm_cvttps_epi32(ewrt); |
| 3181 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 3182 | ewitab = _mm_slli_epi32(ewitab,2); |
| 3183 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 3184 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 3185 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 3186 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 3187 | _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); |
| 3188 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 3189 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 3190 | velec = _mm_mul_ps(qq31,_mm_sub_ps(rinv31,velec)); |
| 3191 | felec = _mm_mul_ps(_mm_mul_ps(qq31,rinv31),_mm_sub_ps(rinvsq31,felec)); |
| 3192 | |
| 3193 | d = _mm_sub_ps(r31,rswitch); |
| 3194 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 3195 | d2 = _mm_mul_ps(d,d); |
| 3196 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
| 3197 | |
| 3198 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 3199 | |
| 3200 | /* Evaluate switch function */ |
| 3201 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 3202 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv31,_mm_mul_ps(velec,dsw)) ); |
| 3203 | cutoff_mask = _mm_cmplt_ps(rsq31,rcutoff2); |
| 3204 | |
| 3205 | fscal = felec; |
| 3206 | |
| 3207 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 3208 | |
| 3209 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 3210 | |
| 3211 | /* Calculate temporary vectorial force */ |
| 3212 | tx = _mm_mul_ps(fscal,dx31); |
| 3213 | ty = _mm_mul_ps(fscal,dy31); |
| 3214 | tz = _mm_mul_ps(fscal,dz31); |
| 3215 | |
| 3216 | /* Update vectorial force */ |
| 3217 | fix3 = _mm_add_ps(fix3,tx); |
| 3218 | fiy3 = _mm_add_ps(fiy3,ty); |
| 3219 | fiz3 = _mm_add_ps(fiz3,tz); |
| 3220 | |
| 3221 | fjx1 = _mm_add_ps(fjx1,tx); |
| 3222 | fjy1 = _mm_add_ps(fjy1,ty); |
| 3223 | fjz1 = _mm_add_ps(fjz1,tz); |
| 3224 | |
| 3225 | } |
| 3226 | |
| 3227 | /************************** |
| 3228 | * CALCULATE INTERACTIONS * |
| 3229 | **************************/ |
| 3230 | |
| 3231 | if (gmx_mm_any_lt(rsq32,rcutoff2)) |
| 3232 | { |
| 3233 | |
| 3234 | r32 = _mm_mul_ps(rsq32,rinv32); |
| 3235 | r32 = _mm_andnot_ps(dummy_mask,r32); |
| 3236 | |
| 3237 | /* EWALD ELECTROSTATICS */ |
| 3238 | |
| 3239 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 3240 | ewrt = _mm_mul_ps(r32,ewtabscale); |
| 3241 | ewitab = _mm_cvttps_epi32(ewrt); |
| 3242 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 3243 | ewitab = _mm_slli_epi32(ewitab,2); |
| 3244 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 3245 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 3246 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 3247 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 3248 | _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); |
| 3249 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 3250 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 3251 | velec = _mm_mul_ps(qq32,_mm_sub_ps(rinv32,velec)); |
| 3252 | felec = _mm_mul_ps(_mm_mul_ps(qq32,rinv32),_mm_sub_ps(rinvsq32,felec)); |
| 3253 | |
| 3254 | d = _mm_sub_ps(r32,rswitch); |
| 3255 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 3256 | d2 = _mm_mul_ps(d,d); |
| 3257 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
| 3258 | |
| 3259 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 3260 | |
| 3261 | /* Evaluate switch function */ |
| 3262 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 3263 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv32,_mm_mul_ps(velec,dsw)) ); |
| 3264 | cutoff_mask = _mm_cmplt_ps(rsq32,rcutoff2); |
| 3265 | |
| 3266 | fscal = felec; |
| 3267 | |
| 3268 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 3269 | |
| 3270 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 3271 | |
| 3272 | /* Calculate temporary vectorial force */ |
| 3273 | tx = _mm_mul_ps(fscal,dx32); |
| 3274 | ty = _mm_mul_ps(fscal,dy32); |
| 3275 | tz = _mm_mul_ps(fscal,dz32); |
| 3276 | |
| 3277 | /* Update vectorial force */ |
| 3278 | fix3 = _mm_add_ps(fix3,tx); |
| 3279 | fiy3 = _mm_add_ps(fiy3,ty); |
| 3280 | fiz3 = _mm_add_ps(fiz3,tz); |
| 3281 | |
| 3282 | fjx2 = _mm_add_ps(fjx2,tx); |
| 3283 | fjy2 = _mm_add_ps(fjy2,ty); |
| 3284 | fjz2 = _mm_add_ps(fjz2,tz); |
| 3285 | |
| 3286 | } |
| 3287 | |
| 3288 | /************************** |
| 3289 | * CALCULATE INTERACTIONS * |
| 3290 | **************************/ |
| 3291 | |
| 3292 | if (gmx_mm_any_lt(rsq33,rcutoff2)) |
| 3293 | { |
| 3294 | |
| 3295 | r33 = _mm_mul_ps(rsq33,rinv33); |
| 3296 | r33 = _mm_andnot_ps(dummy_mask,r33); |
| 3297 | |
| 3298 | /* EWALD ELECTROSTATICS */ |
| 3299 | |
| 3300 | /* Calculate Ewald table index by multiplying r with scale and truncate to integer */ |
| 3301 | ewrt = _mm_mul_ps(r33,ewtabscale); |
| 3302 | ewitab = _mm_cvttps_epi32(ewrt); |
| 3303 | eweps = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR)__extension__ ({ __m128 __X = (ewrt); (__m128) __builtin_ia32_roundps ((__v4sf)__X, ((0x00 | 0x01))); })); |
| 3304 | ewitab = _mm_slli_epi32(ewitab,2); |
| 3305 | ewtabF = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(0) & 3];})) ); |
| 3306 | ewtabD = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(1) & 3];})) ); |
| 3307 | ewtabV = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(2) & 3];})) ); |
| 3308 | ewtabFn = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3)(__extension__ ({ __v4si __a = (__v4si)(ewitab); __a[(3) & 3];})) ); |
| 3309 | _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); |
| 3310 | felec = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD)); |
| 3311 | velec = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec))); |
| 3312 | velec = _mm_mul_ps(qq33,_mm_sub_ps(rinv33,velec)); |
| 3313 | felec = _mm_mul_ps(_mm_mul_ps(qq33,rinv33),_mm_sub_ps(rinvsq33,felec)); |
| 3314 | |
| 3315 | d = _mm_sub_ps(r33,rswitch); |
| 3316 | d = _mm_max_ps(d,_mm_setzero_ps()); |
| 3317 | d2 = _mm_mul_ps(d,d); |
| 3318 | sw = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5))))))); |
| 3319 | |
| 3320 | dsw = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4))))); |
| 3321 | |
| 3322 | /* Evaluate switch function */ |
| 3323 | /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */ |
| 3324 | felec = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv33,_mm_mul_ps(velec,dsw)) ); |
| 3325 | cutoff_mask = _mm_cmplt_ps(rsq33,rcutoff2); |
| 3326 | |
| 3327 | fscal = felec; |
| 3328 | |
| 3329 | fscal = _mm_and_ps(fscal,cutoff_mask); |
| 3330 | |
| 3331 | fscal = _mm_andnot_ps(dummy_mask,fscal); |
| 3332 | |
| 3333 | /* Calculate temporary vectorial force */ |
| 3334 | tx = _mm_mul_ps(fscal,dx33); |
| 3335 | ty = _mm_mul_ps(fscal,dy33); |
| 3336 | tz = _mm_mul_ps(fscal,dz33); |
| 3337 | |
| 3338 | /* Update vectorial force */ |
| 3339 | fix3 = _mm_add_ps(fix3,tx); |
| 3340 | fiy3 = _mm_add_ps(fiy3,ty); |
| 3341 | fiz3 = _mm_add_ps(fiz3,tz); |
| 3342 | |
| 3343 | fjx3 = _mm_add_ps(fjx3,tx); |
| 3344 | fjy3 = _mm_add_ps(fjy3,ty); |
| 3345 | fjz3 = _mm_add_ps(fjz3,tz); |
| 3346 | |
| 3347 | } |
| 3348 | |
| 3349 | fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch; |
| 3350 | fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch; |
| 3351 | fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch; |
| 3352 | fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch; |
| 3353 | |
| 3354 | gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD, |
| 3355 | fjx0,fjy0,fjz0,fjx1,fjy1,fjz1, |
| 3356 | fjx2,fjy2,fjz2,fjx3,fjy3,fjz3); |
| 3357 | |
| 3358 | /* Inner loop uses 627 flops */ |
| 3359 | } |
| 3360 | |
| 3361 | /* End of innermost loop */ |
| 3362 | |
| 3363 | gmx_mm_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3, |
| 3364 | f+i_coord_offset,fshift+i_shift_offset); |
| 3365 | |
| 3366 | /* Increment number of inner iterations */ |
| 3367 | inneriter += j_index_end - j_index_start; |
| 3368 | |
| 3369 | /* Outer loop uses 24 flops */ |
| 3370 | } |
| 3371 | |
| 3372 | /* Increment number of outer iterations */ |
| 3373 | outeriter += nri; |
| 3374 | |
| 3375 | /* Update outer/inner flops */ |
| 3376 | |
| 3377 | inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*627)(nrnb)->n[eNR_NBKERNEL_ELEC_VDW_W4W4_F] += outeriter*24 + inneriter *627; |
| 3378 | } |