src/gromacs/mdlib/genborn_sse2_single.c

   1 /*
   2  *
   3  *                This source code is part of
   4  *
   5  *                 G   R   O   M   A   C   S
   6  *
   7  *          GROningen MAchine for Chemical Simulations
   8  *
   9  * Written by David van der Spoel, Erik Lindahl, Berk Hess, and others.
  10  * Copyright (c) 1991-2000, University of Groningen, The Netherlands.
  11  * Copyright (c) 2001-2008, The GROMACS development team,
  12  * check out http://www.gromacs.org for more information.
  13
  14  * This program is free software; you can redistribute it and/or
  15  * modify it under the terms of the GNU General Public License
  16  * as published by the Free Software Foundation; either version 2
  17  * of the License, or (at your option) any later version.
  18  *
  19  * If you want to redistribute modifications, please consider that
  20  * scientific software is very special. Version control is crucial -
  21  * bugs must be traceable. We will be happy to consider code for
  22  * inclusion in the official distribution, but derived work must not
  23  * be called official GROMACS. Details are found in the README & COPYING
  24  * files - if they are missing, get the official version at www.gromacs.org.
  25  *
  26  * To help us fund GROMACS development, we humbly ask that you cite
  27  * the papers on the package - you can find them in the top README file.
  28  *
  29  * For more info, check our website at http://www.gromacs.org
  30  *
  31  * And Hey:
  32  * Gallium Rubidium Oxygen Manganese Argon Carbon Silicon
  33  */
  34 #ifdef HAVE_CONFIG_H
  35 #include <config.h>
  36 #endif
  37
  38 #include <math.h>
  39 #include <string.h>
  40
  41 #include "typedefs.h"
  42 #include "smalloc.h"
  43 #include "genborn.h"
  44 #include "vec.h"
  45 #include "grompp.h"
  46 #include "pdbio.h"
  47 #include "names.h"
  48 #include "physics.h"
  49 #include "partdec.h"
  50 #include "domdec.h"
  51 #include "network.h"
  52 #include "gmx_fatal.h"
  53 #include "mtop_util.h"
  54 #include "genborn.h"
  55
  56 #include "gromacs/utility/gmxmpi.h"
  57
  58
  59 /* Only compile this file if SSE intrinsics are available */
  60 #if 0 && defined (GMX_X86_SSE2)
  61
  62 #include <gmx_sse2_single.h>
  63 #include <emmintrin.h>
  64
  65 #include "genborn_sse2_single.h"
  66
  67
  68 int
  69 calc_gb_rad_still_sse2_single(t_commrec *cr, t_forcerec *fr,
  70                               int natoms, gmx_localtop_t *top,
  71                               const t_atomtypes *atype, float *x, t_nblist *nl,
  72                               gmx_genborn_t *born)
  73 {
  74     int          i, k, n, ii, is3, ii3, nj0, nj1, offset;
  75     int          jnrA, jnrB, jnrC, jnrD, j3A, j3B, j3C, j3D;
  76     int          jnrE, jnrF, jnrG, jnrH, j3E, j3F, j3G, j3H;
  77     int          shift;
  78     int         *mdtype;
  79     real         shX, shY, shZ;
  80     int         *jjnr;
  81     real        *shiftvec;
  82
  83     float        gpi_ai, gpi2;
  84     float        factor;
  85     float       *gb_radius;
  86     float       *vsolv;
  87     float       *work;
  88     float       *dadx;
  89
  90     __m128       ix, iy, iz;
  91     __m128       jx, jy, jz;
  92     __m128       dx, dy, dz;
  93     __m128       tx, ty, tz;
  94     __m128       jxB, jyB, jzB;
  95     __m128       dxB, dyB, dzB;
  96     __m128       txB, tyB, tzB;
  97     __m128       rsq, rinv, rinv2, rinv4, rinv6;
  98     __m128       rsqB, rinvB, rinv2B, rinv4B, rinv6B;
  99     __m128       ratio, gpi, rai, raj, vai, vaj, rvdw;
 100     __m128       ratioB, rajB, vajB, rvdwB;
 101     __m128       ccf, dccf, theta, cosq, term, sinq, res, prod, prod_ai, tmp;
 102     __m128       ccfB, dccfB, thetaB, cosqB, termB, sinqB, resB, prodB;
 103     __m128       mask, icf4, icf6, mask_cmp;
 104     __m128       icf4B, icf6B, mask_cmpB;
 105
 106     __m128       mask1 = gmx_mm_castsi128_ps( _mm_set_epi32(0, 0, 0, 0xffffffff) );
 107     __m128       mask2 = gmx_mm_castsi128_ps( _mm_set_epi32(0, 0, 0xffffffff, 0xffffffff) );
 108     __m128       mask3 = gmx_mm_castsi128_ps( _mm_set_epi32(0, 0xffffffff, 0xffffffff, 0xffffffff) );
 109
 110     const __m128 half   = _mm_set1_ps(0.5f);
 111     const __m128 three  = _mm_set1_ps(3.0f);
 112     const __m128 one    = _mm_set1_ps(1.0f);
 113     const __m128 two    = _mm_set1_ps(2.0f);
 114     const __m128 zero   = _mm_set1_ps(0.0f);
 115     const __m128 four   = _mm_set1_ps(4.0f);
 116
 117     const __m128 still_p5inv  = _mm_set1_ps(STILL_P5INV);
 118     const __m128 still_pip5   = _mm_set1_ps(STILL_PIP5);
 119     const __m128 still_p4     = _mm_set1_ps(STILL_P4);
 120
 121     factor  = 0.5 * ONE_4PI_EPS0;
 122
 123     gb_radius = born->gb_radius;
 124     vsolv     = born->vsolv;
 125     work      = born->gpol_still_work;
 126     jjnr      = nl->jjnr;
 127     shiftvec  = fr->shift_vec[0];
 128     dadx      = fr->dadx;
 129
 130     jnrA = jnrB = jnrC = jnrD = 0;
 131     jx   = _mm_setzero_ps();
 132     jy   = _mm_setzero_ps();
 133     jz   = _mm_setzero_ps();
 134
 135     n = 0;
 136
 137     for (i = 0; i < natoms; i++)
 138     {
 139         work[i] = 0;
 140     }
 141
 142     for (i = 0; i < nl->nri; i++)
 143     {
 144         ii     = nl->iinr[i];
 145         ii3    = ii*3;
 146         is3    = 3*nl->shift[i];
 147         shX    = shiftvec[is3];
 148         shY    = shiftvec[is3+1];
 149         shZ    = shiftvec[is3+2];
 150         nj0    = nl->jindex[i];
 151         nj1    = nl->jindex[i+1];
 152
 153         ix     = _mm_set1_ps(shX+x[ii3+0]);
 154         iy     = _mm_set1_ps(shY+x[ii3+1]);
 155         iz     = _mm_set1_ps(shZ+x[ii3+2]);
 156
 157         offset = (nj1-nj0)%4;
 158
 159         /* Polarization energy for atom ai */
 160         gpi    = _mm_setzero_ps();
 161
 162         rai     = _mm_load1_ps(gb_radius+ii);
 163         prod_ai = _mm_set1_ps(STILL_P4*vsolv[ii]);
 164
 165         for (k = nj0; k < nj1-4-offset; k += 8)
 166         {
 167             jnrA        = jjnr[k];
 168             jnrB        = jjnr[k+1];
 169             jnrC        = jjnr[k+2];
 170             jnrD        = jjnr[k+3];
 171             jnrE        = jjnr[k+4];
 172             jnrF        = jjnr[k+5];
 173             jnrG        = jjnr[k+6];
 174             jnrH        = jjnr[k+7];
 175
 176             j3A         = 3*jnrA;
 177             j3B         = 3*jnrB;
 178             j3C         = 3*jnrC;
 179             j3D         = 3*jnrD;
 180             j3E         = 3*jnrE;
 181             j3F         = 3*jnrF;
 182             j3G         = 3*jnrG;
 183             j3H         = 3*jnrH;
 184
 185             GMX_MM_LOAD_1RVEC_4POINTERS_PS(x+j3A, x+j3B, x+j3C, x+j3D, jx, jy, jz);
 186             GMX_MM_LOAD_1RVEC_4POINTERS_PS(x+j3E, x+j3F, x+j3G, x+j3H, jxB, jyB, jzB);
 187
 188             GMX_MM_LOAD_4VALUES_PS(gb_radius+jnrA, gb_radius+jnrB, gb_radius+jnrC, gb_radius+jnrD, raj);
 189             GMX_MM_LOAD_4VALUES_PS(gb_radius+jnrE, gb_radius+jnrF, gb_radius+jnrG, gb_radius+jnrH, rajB);
 190             GMX_MM_LOAD_4VALUES_PS(vsolv+jnrA, vsolv+jnrB, vsolv+jnrC, vsolv+jnrD, vaj);
 191             GMX_MM_LOAD_4VALUES_PS(vsolv+jnrE, vsolv+jnrF, vsolv+jnrG, vsolv+jnrH, vajB);
 192
 193             dx          = _mm_sub_ps(ix, jx);
 194             dy          = _mm_sub_ps(iy, jy);
 195             dz          = _mm_sub_ps(iz, jz);
 196             dxB         = _mm_sub_ps(ix, jxB);
 197             dyB         = _mm_sub_ps(iy, jyB);
 198             dzB         = _mm_sub_ps(iz, jzB);
 199
 200             rsq         = gmx_mm_calc_rsq_ps(dx, dy, dz);
 201             rsqB        = gmx_mm_calc_rsq_ps(dxB, dyB, dzB);
 202             rinv        = gmx_mm_invsqrt_ps(rsq);
 203             rinvB       = gmx_mm_invsqrt_ps(rsqB);
 204             rinv2       = _mm_mul_ps(rinv, rinv);
 205             rinv2B      = _mm_mul_ps(rinvB, rinvB);
 206             rinv4       = _mm_mul_ps(rinv2, rinv2);
 207             rinv4B      = _mm_mul_ps(rinv2B, rinv2B);
 208             rinv6       = _mm_mul_ps(rinv4, rinv2);
 209             rinv6B      = _mm_mul_ps(rinv4B, rinv2B);
 210
 211             rvdw        = _mm_add_ps(rai, raj);
 212             rvdwB       = _mm_add_ps(rai, rajB);
 213             ratio       = _mm_mul_ps(rsq, gmx_mm_inv_ps( _mm_mul_ps(rvdw, rvdw)));
 214             ratioB      = _mm_mul_ps(rsqB, gmx_mm_inv_ps( _mm_mul_ps(rvdwB, rvdwB)));
 215
 216             mask_cmp    = _mm_cmple_ps(ratio, still_p5inv);
 217             mask_cmpB   = _mm_cmple_ps(ratioB, still_p5inv);
 218
 219             /* gmx_mm_sincos_ps() is quite expensive, so avoid calculating it if we can! */
 220             if (0 == _mm_movemask_ps(mask_cmp) )
 221             {
 222                 /* if ratio>still_p5inv for ALL elements */
 223                 ccf         = one;
 224                 dccf        = _mm_setzero_ps();
 225             }
 226             else
 227             {
 228                 ratio       = _mm_min_ps(ratio, still_p5inv);
 229                 theta       = _mm_mul_ps(ratio, still_pip5);
 230                 gmx_mm_sincos_ps(theta, &sinq, &cosq);
 231                 term        = _mm_mul_ps(half, _mm_sub_ps(one, cosq));
 232                 ccf         = _mm_mul_ps(term, term);
 233                 dccf        = _mm_mul_ps(_mm_mul_ps(two, term),
 234                                          _mm_mul_ps(sinq, theta));
 235             }
 236             if (0 == _mm_movemask_ps(mask_cmpB) )
 237             {
 238                 /* if ratio>still_p5inv for ALL elements */
 239                 ccfB        = one;
 240                 dccfB       = _mm_setzero_ps();
 241             }
 242             else
 243             {
 244                 ratioB      = _mm_min_ps(ratioB, still_p5inv);
 245                 thetaB      = _mm_mul_ps(ratioB, still_pip5);
 246                 gmx_mm_sincos_ps(thetaB, &sinqB, &cosqB);
 247                 termB       = _mm_mul_ps(half, _mm_sub_ps(one, cosqB));
 248                 ccfB        = _mm_mul_ps(termB, termB);
 249                 dccfB       = _mm_mul_ps(_mm_mul_ps(two, termB),
 250                                          _mm_mul_ps(sinqB, thetaB));
 251             }
 252
 253             prod        = _mm_mul_ps(still_p4, vaj);
 254             prodB       = _mm_mul_ps(still_p4, vajB);
 255             icf4        = _mm_mul_ps(ccf, rinv4);
 256             icf4B       = _mm_mul_ps(ccfB, rinv4B);
 257             icf6        = _mm_mul_ps( _mm_sub_ps( _mm_mul_ps(four, ccf), dccf), rinv6);
 258             icf6B       = _mm_mul_ps( _mm_sub_ps( _mm_mul_ps(four, ccfB), dccfB), rinv6B);
 259
 260             GMX_MM_INCREMENT_4VALUES_PS(work+jnrA, work+jnrB, work+jnrC, work+jnrD, _mm_mul_ps(prod_ai, icf4));
 261             GMX_MM_INCREMENT_4VALUES_PS(work+jnrE, work+jnrF, work+jnrG, work+jnrH, _mm_mul_ps(prod_ai, icf4B));
 262
 263             gpi           = _mm_add_ps(gpi, _mm_add_ps( _mm_mul_ps(prod, icf4), _mm_mul_ps(prodB, icf4B) ) );
 264
 265             _mm_store_ps(dadx, _mm_mul_ps(prod, icf6));
 266             dadx += 4;
 267             _mm_store_ps(dadx, _mm_mul_ps(prod_ai, icf6));
 268             dadx += 4;
 269             _mm_store_ps(dadx, _mm_mul_ps(prodB, icf6B));
 270             dadx += 4;
 271             _mm_store_ps(dadx, _mm_mul_ps(prod_ai, icf6B));
 272             dadx += 4;
 273         }
 274
 275         for (; k < nj1-offset; k += 4)
 276         {
 277             jnrA        = jjnr[k];
 278             jnrB        = jjnr[k+1];
 279             jnrC        = jjnr[k+2];
 280             jnrD        = jjnr[k+3];
 281
 282             j3A         = 3*jnrA;
 283             j3B         = 3*jnrB;
 284             j3C         = 3*jnrC;
 285             j3D         = 3*jnrD;
 286
 287             GMX_MM_LOAD_1RVEC_4POINTERS_PS(x+j3A, x+j3B, x+j3C, x+j3D, jx, jy, jz);
 288
 289             GMX_MM_LOAD_4VALUES_PS(gb_radius+jnrA, gb_radius+jnrB, gb_radius+jnrC, gb_radius+jnrD, raj);
 290             GMX_MM_LOAD_4VALUES_PS(vsolv+jnrA, vsolv+jnrB, vsolv+jnrC, vsolv+jnrD, vaj);
 291
 292             dx          = _mm_sub_ps(ix, jx);
 293             dy          = _mm_sub_ps(iy, jy);
 294             dz          = _mm_sub_ps(iz, jz);
 295
 296             rsq         = gmx_mm_calc_rsq_ps(dx, dy, dz);
 297             rinv        = gmx_mm_invsqrt_ps(rsq);
 298             rinv2       = _mm_mul_ps(rinv, rinv);
 299             rinv4       = _mm_mul_ps(rinv2, rinv2);
 300             rinv6       = _mm_mul_ps(rinv4, rinv2);
 301
 302             rvdw        = _mm_add_ps(rai, raj);
 303             ratio       = _mm_mul_ps(rsq, gmx_mm_inv_ps( _mm_mul_ps(rvdw, rvdw)));
 304
 305             mask_cmp    = _mm_cmple_ps(ratio, still_p5inv);
 306
 307             /* gmx_mm_sincos_ps() is quite expensive, so avoid calculating it if we can! */
 308             if (0 == _mm_movemask_ps(mask_cmp))
 309             {
 310                 /* if ratio>still_p5inv for ALL elements */
 311                 ccf         = one;
 312                 dccf        = _mm_setzero_ps();
 313             }
 314             else
 315             {
 316                 ratio       = _mm_min_ps(ratio, still_p5inv);
 317                 theta       = _mm_mul_ps(ratio, still_pip5);
 318                 gmx_mm_sincos_ps(theta, &sinq, &cosq);
 319                 term        = _mm_mul_ps(half, _mm_sub_ps(one, cosq));
 320                 ccf         = _mm_mul_ps(term, term);
 321                 dccf        = _mm_mul_ps(_mm_mul_ps(two, term),
 322                                          _mm_mul_ps(sinq, theta));
 323             }
 324
 325             prod        = _mm_mul_ps(still_p4, vaj);
 326             icf4        = _mm_mul_ps(ccf, rinv4);
 327             icf6        = _mm_mul_ps( _mm_sub_ps( _mm_mul_ps(four, ccf), dccf), rinv6);
 328
 329             GMX_MM_INCREMENT_4VALUES_PS(work+jnrA, work+jnrB, work+jnrC, work+jnrD, _mm_mul_ps(prod_ai, icf4));
 330
 331             gpi           = _mm_add_ps(gpi, _mm_mul_ps(prod, icf4));
 332
 333             _mm_store_ps(dadx, _mm_mul_ps(prod, icf6));
 334             dadx += 4;
 335             _mm_store_ps(dadx, _mm_mul_ps(prod_ai, icf6));
 336             dadx += 4;
 337         }
 338
 339         if (offset != 0)
 340         {
 341             if (offset == 1)
 342             {
 343                 jnrA        = jjnr[k];
 344                 j3A         = 3*jnrA;
 345                 GMX_MM_LOAD_1RVEC_1POINTER_PS(x+j3A, jx, jy, jz);
 346                 GMX_MM_LOAD_1VALUE_PS(gb_radius+jnrA, raj);
 347                 GMX_MM_LOAD_1VALUE_PS(vsolv+jnrA, vaj);
 348                 mask        = mask1;
 349             }
 350             else if (offset == 2)
 351             {
 352                 jnrA        = jjnr[k];
 353                 jnrB        = jjnr[k+1];
 354                 j3A         = 3*jnrA;
 355                 j3B         = 3*jnrB;
 356                 GMX_MM_LOAD_1RVEC_2POINTERS_PS(x+j3A, x+j3B, jx, jy, jz);
 357                 GMX_MM_LOAD_2VALUES_PS(gb_radius+jnrA, gb_radius+jnrB, raj);
 358                 GMX_MM_LOAD_2VALUES_PS(vsolv+jnrA, vsolv+jnrB, vaj);
 359                 mask        = mask2;
 360             }
 361             else
 362             {
 363                 /* offset must be 3 */
 364                 jnrA        = jjnr[k];
 365                 jnrB        = jjnr[k+1];
 366                 jnrC        = jjnr[k+2];
 367                 j3A         = 3*jnrA;
 368                 j3B         = 3*jnrB;
 369                 j3C         = 3*jnrC;
 370                 GMX_MM_LOAD_1RVEC_3POINTERS_PS(x+j3A, x+j3B, x+j3C, jx, jy, jz);
 371                 GMX_MM_LOAD_3VALUES_PS(gb_radius+jnrA, gb_radius+jnrB, gb_radius+jnrC, raj);
 372                 GMX_MM_LOAD_3VALUES_PS(vsolv+jnrA, vsolv+jnrB, vsolv+jnrC, vaj);
 373                 mask        = mask3;
 374             }
 375
 376             dx          = _mm_sub_ps(ix, jx);
 377             dy          = _mm_sub_ps(iy, jy);
 378             dz          = _mm_sub_ps(iz, jz);
 379
 380             rsq         = gmx_mm_calc_rsq_ps(dx, dy, dz);
 381             rinv        = gmx_mm_invsqrt_ps(rsq);
 382             rinv2       = _mm_mul_ps(rinv, rinv);
 383             rinv4       = _mm_mul_ps(rinv2, rinv2);
 384             rinv6       = _mm_mul_ps(rinv4, rinv2);
 385
 386             rvdw        = _mm_add_ps(rai, raj);
 387             ratio       = _mm_mul_ps(rsq, gmx_mm_inv_ps( _mm_mul_ps(rvdw, rvdw)));
 388
 389             mask_cmp    = _mm_cmple_ps(ratio, still_p5inv);
 390
 391             if (0 == _mm_movemask_ps(mask_cmp))
 392             {
 393                 /* if ratio>still_p5inv for ALL elements */
 394                 ccf         = one;
 395                 dccf        = _mm_setzero_ps();
 396             }
 397             else
 398             {
 399                 ratio       = _mm_min_ps(ratio, still_p5inv);
 400                 theta       = _mm_mul_ps(ratio, still_pip5);
 401                 gmx_mm_sincos_ps(theta, &sinq, &cosq);
 402                 term        = _mm_mul_ps(half, _mm_sub_ps(one, cosq));
 403                 ccf         = _mm_mul_ps(term, term);
 404                 dccf        = _mm_mul_ps(_mm_mul_ps(two, term),
 405                                          _mm_mul_ps(sinq, theta));
 406             }
 407
 408             prod        = _mm_mul_ps(still_p4, vaj);
 409             icf4        = _mm_mul_ps(ccf, rinv4);
 410             icf6        = _mm_mul_ps( _mm_sub_ps( _mm_mul_ps(four, ccf), dccf), rinv6);
 411
 412             gpi           = _mm_add_ps(gpi, _mm_mul_ps(prod, icf4));
 413
 414             _mm_store_ps(dadx, _mm_mul_ps(prod, icf6));
 415             dadx += 4;
 416             _mm_store_ps(dadx, _mm_mul_ps(prod_ai, icf6));
 417             dadx += 4;
 418
 419             tmp = _mm_mul_ps(prod_ai, icf4);
 420
 421             if (offset == 1)
 422             {
 423                 GMX_MM_INCREMENT_1VALUE_PS(work+jnrA, tmp);
 424             }
 425             else if (offset == 2)
 426             {
 427                 GMX_MM_INCREMENT_2VALUES_PS(work+jnrA, work+jnrB, tmp);
 428             }
 429             else
 430             {
 431                 /* offset must be 3 */
 432                 GMX_MM_INCREMENT_3VALUES_PS(work+jnrA, work+jnrB, work+jnrC, tmp);
 433             }
 434         }
 435         GMX_MM_UPDATE_1POT_PS(gpi, work+ii);
 436     }
 437
 438     /* Sum up the polarization energy from other nodes */
 439     if (PARTDECOMP(cr))
 440     {
 441         gmx_sum(natoms, work, cr);
 442     }
 443     else if (DOMAINDECOMP(cr))
 444     {
 445         dd_atom_sum_real(cr->dd, work);
 446     }
 447
 448     /* Compute the radii */
 449     for (i = 0; i < fr->natoms_force; i++) /* PELA born->nr */
 450     {
 451         if (born->use[i] != 0)
 452         {
 453             gpi_ai           = born->gpol[i] + work[i]; /* add gpi to the initial pol energy gpi_ai*/
 454             gpi2             = gpi_ai * gpi_ai;
 455             born->bRad[i]    = factor*gmx_invsqrt(gpi2);
 456             fr->invsqrta[i]  = gmx_invsqrt(born->bRad[i]);
 457         }
 458     }
 459
 460     /* Extra (local) communication required for DD */
 461     if (DOMAINDECOMP(cr))
 462     {
 463         dd_atom_spread_real(cr->dd, born->bRad);
 464         dd_atom_spread_real(cr->dd, fr->invsqrta);
 465     }
 466
 467     return 0;
 468 }
 469
 470
 471 int
 472 calc_gb_rad_hct_obc_sse2_single(t_commrec *cr, t_forcerec * fr, int natoms, gmx_localtop_t *top,
 473                                 const t_atomtypes *atype, float *x, t_nblist *nl, gmx_genborn_t *born, t_mdatoms *md, int gb_algorithm)
 474 {
 475     int          i, ai, k, n, ii, ii3, is3, nj0, nj1, at0, at1, offset;
 476     int          jnrA, jnrB, jnrC, jnrD;
 477     int          j3A, j3B, j3C, j3D;
 478     int          jnrE, jnrF, jnrG, jnrH;
 479     int          j3E, j3F, j3G, j3H;
 480     float        shX, shY, shZ;
 481     float        rr, rr_inv, rr_inv2, sum_tmp, sum, sum2, sum3, gbr;
 482     float        sum_ai2, sum_ai3, tsum, tchain, doffset;
 483     float       *obc_param;
 484     float       *gb_radius;
 485     float       *work;
 486     int       *  jjnr;
 487     float       *dadx;
 488     float       *shiftvec;
 489     float        min_rad, rad;
 490
 491     __m128       ix, iy, iz, jx, jy, jz;
 492     __m128       dx, dy, dz, t1, t2, t3, t4;
 493     __m128       rsq, rinv, r;
 494     __m128       rai, rai_inv, raj, raj_inv, rai_inv2, sk, sk2, lij, dlij, duij;
 495     __m128       uij, lij2, uij2, lij3, uij3, diff2;
 496     __m128       lij_inv, sk2_inv, prod, log_term, tmp, tmp_sum;
 497     __m128       sum_ai, tmp_ai, sk_ai, sk_aj, sk2_ai, sk2_aj, sk2_rinv;
 498     __m128       dadx1, dadx2;
 499     __m128       logterm;
 500     __m128       mask;
 501     __m128       obc_mask1, obc_mask2, obc_mask3;
 502     __m128       jxB, jyB, jzB, t1B, t2B, t3B, t4B;
 503     __m128       dxB, dyB, dzB, rsqB, rinvB, rB;
 504     __m128       rajB, raj_invB, rai_inv2B, sk2B, lijB, dlijB, duijB;
 505     __m128       uijB, lij2B, uij2B, lij3B, uij3B, diff2B;
 506     __m128       lij_invB, sk2_invB, prodB;
 507     __m128       sk_ajB, sk2_ajB, sk2_rinvB;
 508     __m128       dadx1B, dadx2B;
 509     __m128       logtermB;
 510     __m128       obc_mask1B, obc_mask2B, obc_mask3B;
 511
 512     __m128       mask1 = gmx_mm_castsi128_ps( _mm_set_epi32(0, 0, 0, 0xffffffff) );
 513     __m128       mask2 = gmx_mm_castsi128_ps( _mm_set_epi32(0, 0, 0xffffffff, 0xffffffff) );
 514     __m128       mask3 = gmx_mm_castsi128_ps( _mm_set_epi32(0, 0xffffffff, 0xffffffff, 0xffffffff) );
 515
 516     __m128       oneeighth   = _mm_set1_ps(0.125);
 517     __m128       onefourth   = _mm_set1_ps(0.25);
 518
 519     const __m128 half  = _mm_set1_ps(0.5f);
 520     const __m128 three = _mm_set1_ps(3.0f);
 521     const __m128 one   = _mm_set1_ps(1.0f);
 522     const __m128 two   = _mm_set1_ps(2.0f);
 523     const __m128 zero  = _mm_set1_ps(0.0f);
 524     const __m128 neg   = _mm_set1_ps(-1.0f);
 525
 526     /* Set the dielectric offset */
 527     doffset   = born->gb_doffset;
 528     gb_radius = born->gb_radius;
 529     obc_param = born->param;
 530     work      = born->gpol_hct_work;
 531     jjnr      = nl->jjnr;
 532     dadx      = fr->dadx;
 533     shiftvec  = fr->shift_vec[0];
 534
 535     jx        = _mm_setzero_ps();
 536     jy        = _mm_setzero_ps();
 537     jz        = _mm_setzero_ps();
 538
 539     jnrA = jnrB = jnrC = jnrD = 0;
 540
 541     for (i = 0; i < born->nr; i++)
 542     {
 543         work[i] = 0;
 544     }
 545
 546     for (i = 0; i < nl->nri; i++)
 547     {
 548         ii     = nl->iinr[i];
 549         ii3    = ii*3;
 550         is3    = 3*nl->shift[i];
 551         shX    = shiftvec[is3];
 552         shY    = shiftvec[is3+1];
 553         shZ    = shiftvec[is3+2];
 554         nj0    = nl->jindex[i];
 555         nj1    = nl->jindex[i+1];
 556
 557         ix     = _mm_set1_ps(shX+x[ii3+0]);
 558         iy     = _mm_set1_ps(shY+x[ii3+1]);
 559         iz     = _mm_set1_ps(shZ+x[ii3+2]);
 560
 561         offset = (nj1-nj0)%4;
 562
 563         rai     = _mm_load1_ps(gb_radius+ii);
 564         rai_inv = gmx_mm_inv_ps(rai);
 565
 566         sum_ai = _mm_setzero_ps();
 567
 568         sk_ai  = _mm_load1_ps(born->param+ii);
 569         sk2_ai = _mm_mul_ps(sk_ai, sk_ai);
 570
 571         for (k = nj0; k < nj1-4-offset; k += 8)
 572         {
 573             jnrA        = jjnr[k];
 574             jnrB        = jjnr[k+1];
 575             jnrC        = jjnr[k+2];
 576             jnrD        = jjnr[k+3];
 577             jnrE        = jjnr[k+4];
 578             jnrF        = jjnr[k+5];
 579             jnrG        = jjnr[k+6];
 580             jnrH        = jjnr[k+7];
 581
 582             j3A         = 3*jnrA;
 583             j3B         = 3*jnrB;
 584             j3C         = 3*jnrC;
 585             j3D         = 3*jnrD;
 586             j3E         = 3*jnrE;
 587             j3F         = 3*jnrF;
 588             j3G         = 3*jnrG;
 589             j3H         = 3*jnrH;
 590
 591             GMX_MM_LOAD_1RVEC_4POINTERS_PS(x+j3A, x+j3B, x+j3C, x+j3D, jx, jy, jz);
 592             GMX_MM_LOAD_1RVEC_4POINTERS_PS(x+j3E, x+j3F, x+j3G, x+j3H, jxB, jyB, jzB);
 593             GMX_MM_LOAD_4VALUES_PS(gb_radius+jnrA, gb_radius+jnrB, gb_radius+jnrC, gb_radius+jnrD, raj);
 594             GMX_MM_LOAD_4VALUES_PS(gb_radius+jnrE, gb_radius+jnrF, gb_radius+jnrG, gb_radius+jnrH, rajB);
 595             GMX_MM_LOAD_4VALUES_PS(obc_param+jnrA, obc_param+jnrB, obc_param+jnrC, obc_param+jnrD, sk_aj);
 596             GMX_MM_LOAD_4VALUES_PS(obc_param+jnrE, obc_param+jnrF, obc_param+jnrG, obc_param+jnrH, sk_ajB);
 597
 598             dx    = _mm_sub_ps(ix, jx);
 599             dy    = _mm_sub_ps(iy, jy);
 600             dz    = _mm_sub_ps(iz, jz);
 601             dxB   = _mm_sub_ps(ix, jxB);
 602             dyB   = _mm_sub_ps(iy, jyB);
 603             dzB   = _mm_sub_ps(iz, jzB);
 604
 605             rsq         = gmx_mm_calc_rsq_ps(dx, dy, dz);
 606             rsqB        = gmx_mm_calc_rsq_ps(dxB, dyB, dzB);
 607
 608             rinv        = gmx_mm_invsqrt_ps(rsq);
 609             r           = _mm_mul_ps(rsq, rinv);
 610             rinvB       = gmx_mm_invsqrt_ps(rsqB);
 611             rB          = _mm_mul_ps(rsqB, rinvB);
 612
 613             /* Compute raj_inv aj1-4 */
 614             raj_inv     = gmx_mm_inv_ps(raj);
 615             raj_invB    = gmx_mm_inv_ps(rajB);
 616
 617             /* Evaluate influence of atom aj -> ai */
 618             t1            = _mm_add_ps(r, sk_aj);
 619             t2            = _mm_sub_ps(r, sk_aj);
 620             t3            = _mm_sub_ps(sk_aj, r);
 621             t1B           = _mm_add_ps(rB, sk_ajB);
 622             t2B           = _mm_sub_ps(rB, sk_ajB);
 623             t3B           = _mm_sub_ps(sk_ajB, rB);
 624             obc_mask1     = _mm_cmplt_ps(rai, t1);
 625             obc_mask2     = _mm_cmplt_ps(rai, t2);
 626             obc_mask3     = _mm_cmplt_ps(rai, t3);
 627             obc_mask1B    = _mm_cmplt_ps(rai, t1B);
 628             obc_mask2B    = _mm_cmplt_ps(rai, t2B);
 629             obc_mask3B    = _mm_cmplt_ps(rai, t3B);
 630
 631             uij           = gmx_mm_inv_ps(t1);
 632             lij           = _mm_or_ps(   _mm_and_ps(obc_mask2, gmx_mm_inv_ps(t2)),
 633                                          _mm_andnot_ps(obc_mask2, rai_inv));
 634             dlij          = _mm_and_ps(one, obc_mask2);
 635             uij2          = _mm_mul_ps(uij, uij);
 636             uij3          = _mm_mul_ps(uij2, uij);
 637             lij2          = _mm_mul_ps(lij, lij);
 638             lij3          = _mm_mul_ps(lij2, lij);
 639
 640             uijB          = gmx_mm_inv_ps(t1B);
 641             lijB          = _mm_or_ps(   _mm_and_ps(obc_mask2B, gmx_mm_inv_ps(t2B)),
 642                                          _mm_andnot_ps(obc_mask2B, rai_inv));
 643             dlijB         = _mm_and_ps(one, obc_mask2B);
 644             uij2B         = _mm_mul_ps(uijB, uijB);
 645             uij3B         = _mm_mul_ps(uij2B, uijB);
 646             lij2B         = _mm_mul_ps(lijB, lijB);
 647             lij3B         = _mm_mul_ps(lij2B, lijB);
 648
 649             diff2         = _mm_sub_ps(uij2, lij2);
 650             lij_inv       = gmx_mm_invsqrt_ps(lij2);
 651             sk2_aj        = _mm_mul_ps(sk_aj, sk_aj);
 652             sk2_rinv      = _mm_mul_ps(sk2_aj, rinv);
 653             prod          = _mm_mul_ps(onefourth, sk2_rinv);
 654
 655             diff2B        = _mm_sub_ps(uij2B, lij2B);
 656             lij_invB      = gmx_mm_invsqrt_ps(lij2B);
 657             sk2_ajB       = _mm_mul_ps(sk_ajB, sk_ajB);
 658             sk2_rinvB     = _mm_mul_ps(sk2_ajB, rinvB);
 659             prodB         = _mm_mul_ps(onefourth, sk2_rinvB);
 660
 661             logterm       = gmx_mm_log_ps(_mm_mul_ps(uij, lij_inv));
 662             logtermB      = gmx_mm_log_ps(_mm_mul_ps(uijB, lij_invB));
 663
 664             t1            = _mm_sub_ps(lij, uij);
 665             t2            = _mm_mul_ps(diff2,
 666                                        _mm_sub_ps(_mm_mul_ps(onefourth, r),
 667                                                   prod));
 668             t3            = _mm_mul_ps(half, _mm_mul_ps(rinv, logterm));
 669             t1            = _mm_add_ps(t1, _mm_add_ps(t2, t3));
 670             t4            = _mm_mul_ps(two, _mm_sub_ps(rai_inv, lij));
 671             t4            = _mm_and_ps(t4, obc_mask3);
 672             t1            = _mm_mul_ps(half, _mm_add_ps(t1, t4));
 673
 674             t1B           = _mm_sub_ps(lijB, uijB);
 675             t2B           = _mm_mul_ps(diff2B,
 676                                        _mm_sub_ps(_mm_mul_ps(onefourth, rB),
 677                                                   prodB));
 678             t3B           = _mm_mul_ps(half, _mm_mul_ps(rinvB, logtermB));
 679             t1B           = _mm_add_ps(t1B, _mm_add_ps(t2B, t3B));
 680             t4B           = _mm_mul_ps(two, _mm_sub_ps(rai_inv, lijB));
 681             t4B           = _mm_and_ps(t4B, obc_mask3B);
 682             t1B           = _mm_mul_ps(half, _mm_add_ps(t1B, t4B));
 683
 684             sum_ai        = _mm_add_ps(sum_ai, _mm_add_ps( _mm_and_ps(t1, obc_mask1), _mm_and_ps(t1B, obc_mask1B) ));
 685
 686             t1            = _mm_add_ps(_mm_mul_ps(half, lij2),
 687                                        _mm_mul_ps(prod, lij3));
 688             t1            = _mm_sub_ps(t1,
 689                                        _mm_mul_ps(onefourth,
 690                                                   _mm_add_ps(_mm_mul_ps(lij, rinv),
 691                                                              _mm_mul_ps(lij3, r))));
 692             t2            = _mm_mul_ps(onefourth,
 693                                        _mm_add_ps(_mm_mul_ps(uij, rinv),
 694                                                   _mm_mul_ps(uij3, r)));
 695             t2            = _mm_sub_ps(t2,
 696                                        _mm_add_ps(_mm_mul_ps(half, uij2),
 697                                                   _mm_mul_ps(prod, uij3)));
 698             t3            = _mm_mul_ps(_mm_mul_ps(onefourth, logterm),
 699                                        _mm_mul_ps(rinv, rinv));
 700             t3            = _mm_sub_ps(t3,
 701                                        _mm_mul_ps(_mm_mul_ps(diff2, oneeighth),
 702                                                   _mm_add_ps(one,
 703                                                              _mm_mul_ps(sk2_rinv, rinv))));
 704             t1            = _mm_mul_ps(rinv,
 705                                        _mm_add_ps(_mm_mul_ps(dlij, t1),
 706                                                   _mm_add_ps(t2, t3)));
 707
 708
 709
 710             t1B           = _mm_add_ps(_mm_mul_ps(half, lij2B),
 711                                        _mm_mul_ps(prodB, lij3B));
 712             t1B           = _mm_sub_ps(t1B,
 713                                        _mm_mul_ps(onefourth,
 714                                                   _mm_add_ps(_mm_mul_ps(lijB, rinvB),
 715                                                              _mm_mul_ps(lij3B, rB))));
 716             t2B           = _mm_mul_ps(onefourth,
 717                                        _mm_add_ps(_mm_mul_ps(uijB, rinvB),
 718                                                   _mm_mul_ps(uij3B, rB)));
 719             t2B           = _mm_sub_ps(t2B,
 720                                        _mm_add_ps(_mm_mul_ps(half, uij2B),
 721                                                   _mm_mul_ps(prodB, uij3B)));
 722             t3B           = _mm_mul_ps(_mm_mul_ps(onefourth, logtermB),
 723                                        _mm_mul_ps(rinvB, rinvB));
 724             t3B           = _mm_sub_ps(t3B,
 725                                        _mm_mul_ps(_mm_mul_ps(diff2B, oneeighth),
 726                                                   _mm_add_ps(one,
 727                                                              _mm_mul_ps(sk2_rinvB, rinvB))));
 728             t1B           = _mm_mul_ps(rinvB,
 729                                        _mm_add_ps(_mm_mul_ps(dlijB, t1B),
 730                                                   _mm_add_ps(t2B, t3B)));
 731
 732             dadx1         = _mm_and_ps(t1, obc_mask1);
 733             dadx1B        = _mm_and_ps(t1B, obc_mask1B);
 734
 735
 736             /* Evaluate influence of atom ai -> aj */
 737             t1            = _mm_add_ps(r, sk_ai);
 738             t2            = _mm_sub_ps(r, sk_ai);
 739             t3            = _mm_sub_ps(sk_ai, r);
 740             t1B           = _mm_add_ps(rB, sk_ai);
 741             t2B           = _mm_sub_ps(rB, sk_ai);
 742             t3B           = _mm_sub_ps(sk_ai, rB);
 743             obc_mask1     = _mm_cmplt_ps(raj, t1);
 744             obc_mask2     = _mm_cmplt_ps(raj, t2);
 745             obc_mask3     = _mm_cmplt_ps(raj, t3);
 746             obc_mask1B    = _mm_cmplt_ps(rajB, t1B);
 747             obc_mask2B    = _mm_cmplt_ps(rajB, t2B);
 748             obc_mask3B    = _mm_cmplt_ps(rajB, t3B);
 749
 750             uij           = gmx_mm_inv_ps(t1);
 751             lij           = _mm_or_ps(   _mm_and_ps(obc_mask2, gmx_mm_inv_ps(t2)),
 752                                          _mm_andnot_ps(obc_mask2, raj_inv));
 753             dlij          = _mm_and_ps(one, obc_mask2);
 754             uij2          = _mm_mul_ps(uij, uij);
 755             uij3          = _mm_mul_ps(uij2, uij);
 756             lij2          = _mm_mul_ps(lij, lij);
 757             lij3          = _mm_mul_ps(lij2, lij);
 758
 759             uijB          = gmx_mm_inv_ps(t1B);
 760             lijB          = _mm_or_ps(   _mm_and_ps(obc_mask2B, gmx_mm_inv_ps(t2B)),
 761                                          _mm_andnot_ps(obc_mask2B, raj_invB));
 762             dlijB         = _mm_and_ps(one, obc_mask2B);
 763             uij2B         = _mm_mul_ps(uijB, uijB);
 764             uij3B         = _mm_mul_ps(uij2B, uijB);
 765             lij2B         = _mm_mul_ps(lijB, lijB);
 766             lij3B         = _mm_mul_ps(lij2B, lijB);
 767
 768             diff2         = _mm_sub_ps(uij2, lij2);
 769             lij_inv       = gmx_mm_invsqrt_ps(lij2);
 770             sk2_rinv      = _mm_mul_ps(sk2_ai, rinv);
 771             prod          = _mm_mul_ps(onefourth, sk2_rinv);
 772
 773             diff2B        = _mm_sub_ps(uij2B, lij2B);
 774             lij_invB      = gmx_mm_invsqrt_ps(lij2B);
 775             sk2_rinvB     = _mm_mul_ps(sk2_ai, rinvB);
 776             prodB         = _mm_mul_ps(onefourth, sk2_rinvB);
 777
 778             logterm       = gmx_mm_log_ps(_mm_mul_ps(uij, lij_inv));
 779             logtermB      = gmx_mm_log_ps(_mm_mul_ps(uijB, lij_invB));
 780
 781             t1            = _mm_sub_ps(lij, uij);
 782             t2            = _mm_mul_ps(diff2,
 783                                        _mm_sub_ps(_mm_mul_ps(onefourth, r),
 784                                                   prod));
 785             t3            = _mm_mul_ps(half, _mm_mul_ps(rinv, logterm));
 786             t1            = _mm_add_ps(t1, _mm_add_ps(t2, t3));
 787             t4            = _mm_mul_ps(two, _mm_sub_ps(raj_inv, lij));
 788             t4            = _mm_and_ps(t4, obc_mask3);
 789             t1            = _mm_mul_ps(half, _mm_add_ps(t1, t4));
 790
 791             t1B           = _mm_sub_ps(lijB, uijB);
 792             t2B           = _mm_mul_ps(diff2B,
 793                                        _mm_sub_ps(_mm_mul_ps(onefourth, rB),
 794                                                   prodB));
 795             t3B           = _mm_mul_ps(half, _mm_mul_ps(rinvB, logtermB));
 796             t1B           = _mm_add_ps(t1B, _mm_add_ps(t2B, t3B));
 797             t4B           = _mm_mul_ps(two, _mm_sub_ps(raj_invB, lijB));
 798             t4B           = _mm_and_ps(t4B, obc_mask3B);
 799             t1B           = _mm_mul_ps(half, _mm_add_ps(t1B, t4B));
 800
 801             GMX_MM_INCREMENT_4VALUES_PS(work+jnrA, work+jnrB, work+jnrC, work+jnrD, _mm_and_ps(t1, obc_mask1));
 802             GMX_MM_INCREMENT_4VALUES_PS(work+jnrE, work+jnrF, work+jnrG, work+jnrH, _mm_and_ps(t1B, obc_mask1B));
 803
 804             t1            = _mm_add_ps(_mm_mul_ps(half, lij2),
 805                                        _mm_mul_ps(prod, lij3));
 806             t1            = _mm_sub_ps(t1,
 807                                        _mm_mul_ps(onefourth,
 808                                                   _mm_add_ps(_mm_mul_ps(lij, rinv),
 809                                                              _mm_mul_ps(lij3, r))));
 810             t2            = _mm_mul_ps(onefourth,
 811                                        _mm_add_ps(_mm_mul_ps(uij, rinv),
 812                                                   _mm_mul_ps(uij3, r)));
 813             t2            = _mm_sub_ps(t2,
 814                                        _mm_add_ps(_mm_mul_ps(half, uij2),
 815                                                   _mm_mul_ps(prod, uij3)));
 816             t3            = _mm_mul_ps(_mm_mul_ps(onefourth, logterm),
 817                                        _mm_mul_ps(rinv, rinv));
 818             t3            = _mm_sub_ps(t3,
 819                                        _mm_mul_ps(_mm_mul_ps(diff2, oneeighth),
 820                                                   _mm_add_ps(one,
 821                                                              _mm_mul_ps(sk2_rinv, rinv))));
 822             t1            = _mm_mul_ps(rinv,
 823                                        _mm_add_ps(_mm_mul_ps(dlij, t1),
 824                                                   _mm_add_ps(t2, t3)));
 825
 826
 827             t1B           = _mm_add_ps(_mm_mul_ps(half, lij2B),
 828                                        _mm_mul_ps(prodB, lij3B));
 829             t1B           = _mm_sub_ps(t1B,
 830                                        _mm_mul_ps(onefourth,
 831                                                   _mm_add_ps(_mm_mul_ps(lijB, rinvB),
 832                                                              _mm_mul_ps(lij3B, rB))));
 833             t2B           = _mm_mul_ps(onefourth,
 834                                        _mm_add_ps(_mm_mul_ps(uijB, rinvB),
 835                                                   _mm_mul_ps(uij3B, rB)));
 836             t2B           = _mm_sub_ps(t2B,
 837                                        _mm_add_ps(_mm_mul_ps(half, uij2B),
 838                                                   _mm_mul_ps(prodB, uij3B)));
 839             t3B           = _mm_mul_ps(_mm_mul_ps(onefourth, logtermB),
 840                                        _mm_mul_ps(rinvB, rinvB));
 841             t3B           = _mm_sub_ps(t3B,
 842                                        _mm_mul_ps(_mm_mul_ps(diff2B, oneeighth),
 843                                                   _mm_add_ps(one,
 844                                                              _mm_mul_ps(sk2_rinvB, rinvB))));
 845             t1B           = _mm_mul_ps(rinvB,
 846                                        _mm_add_ps(_mm_mul_ps(dlijB, t1B),
 847                                                   _mm_add_ps(t2B, t3B)));
 848
 849
 850             dadx2         = _mm_and_ps(t1, obc_mask1);
 851             dadx2B        = _mm_and_ps(t1B, obc_mask1B);
 852
 853             _mm_store_ps(dadx, dadx1);
 854             dadx += 4;
 855             _mm_store_ps(dadx, dadx2);
 856             dadx += 4;
 857             _mm_store_ps(dadx, dadx1B);
 858             dadx += 4;
 859             _mm_store_ps(dadx, dadx2B);
 860             dadx += 4;
 861
 862         } /* end normal inner loop */
 863
 864         for (; k < nj1-offset; k += 4)
 865         {
 866             jnrA        = jjnr[k];
 867             jnrB        = jjnr[k+1];
 868             jnrC        = jjnr[k+2];
 869             jnrD        = jjnr[k+3];
 870
 871             j3A         = 3*jnrA;
 872             j3B         = 3*jnrB;
 873             j3C         = 3*jnrC;
 874             j3D         = 3*jnrD;
 875
 876             GMX_MM_LOAD_1RVEC_4POINTERS_PS(x+j3A, x+j3B, x+j3C, x+j3D, jx, jy, jz);
 877             GMX_MM_LOAD_4VALUES_PS(gb_radius+jnrA, gb_radius+jnrB, gb_radius+jnrC, gb_radius+jnrD, raj);
 878             GMX_MM_LOAD_4VALUES_PS(obc_param+jnrA, obc_param+jnrB, obc_param+jnrC, obc_param+jnrD, sk_aj);
 879
 880             dx    = _mm_sub_ps(ix, jx);
 881             dy    = _mm_sub_ps(iy, jy);
 882             dz    = _mm_sub_ps(iz, jz);
 883
 884             rsq         = gmx_mm_calc_rsq_ps(dx, dy, dz);
 885
 886             rinv        = gmx_mm_invsqrt_ps(rsq);
 887             r           = _mm_mul_ps(rsq, rinv);
 888
 889             /* Compute raj_inv aj1-4 */
 890             raj_inv     = gmx_mm_inv_ps(raj);
 891
 892             /* Evaluate influence of atom aj -> ai */
 893             t1            = _mm_add_ps(r, sk_aj);
 894             obc_mask1     = _mm_cmplt_ps(rai, t1);
 895
 896             if (_mm_movemask_ps(obc_mask1))
 897             {
 898                 /* If any of the elements has rai<dr+sk, this is executed */
 899                 t2            = _mm_sub_ps(r, sk_aj);
 900                 t3            = _mm_sub_ps(sk_aj, r);
 901
 902                 obc_mask2     = _mm_cmplt_ps(rai, t2);
 903                 obc_mask3     = _mm_cmplt_ps(rai, t3);
 904
 905                 uij           = gmx_mm_inv_ps(t1);
 906                 lij           = _mm_or_ps(   _mm_and_ps(obc_mask2, gmx_mm_inv_ps(t2)),
 907                                              _mm_andnot_ps(obc_mask2, rai_inv));
 908                 dlij          = _mm_and_ps(one, obc_mask2);
 909                 uij2          = _mm_mul_ps(uij, uij);
 910                 uij3          = _mm_mul_ps(uij2, uij);
 911                 lij2          = _mm_mul_ps(lij, lij);
 912                 lij3          = _mm_mul_ps(lij2, lij);
 913                 diff2         = _mm_sub_ps(uij2, lij2);
 914                 lij_inv       = gmx_mm_invsqrt_ps(lij2);
 915                 sk2_aj        = _mm_mul_ps(sk_aj, sk_aj);
 916                 sk2_rinv      = _mm_mul_ps(sk2_aj, rinv);
 917                 prod          = _mm_mul_ps(onefourth, sk2_rinv);
 918                 logterm       = gmx_mm_log_ps(_mm_mul_ps(uij, lij_inv));
 919                 t1            = _mm_sub_ps(lij, uij);
 920                 t2            = _mm_mul_ps(diff2,
 921                                            _mm_sub_ps(_mm_mul_ps(onefourth, r),
 922                                                       prod));
 923                 t3            = _mm_mul_ps(half, _mm_mul_ps(rinv, logterm));
 924                 t1            = _mm_add_ps(t1, _mm_add_ps(t2, t3));
 925                 t4            = _mm_mul_ps(two, _mm_sub_ps(rai_inv, lij));
 926                 t4            = _mm_and_ps(t4, obc_mask3);
 927                 t1            = _mm_mul_ps(half, _mm_add_ps(t1, t4));
 928                 sum_ai        = _mm_add_ps(sum_ai, _mm_and_ps(t1, obc_mask1));
 929                 t1            = _mm_add_ps(_mm_mul_ps(half, lij2),
 930                                            _mm_mul_ps(prod, lij3));
 931                 t1            = _mm_sub_ps(t1,
 932                                            _mm_mul_ps(onefourth,
 933                                                       _mm_add_ps(_mm_mul_ps(lij, rinv),
 934                                                                  _mm_mul_ps(lij3, r))));
 935                 t2            = _mm_mul_ps(onefourth,
 936                                            _mm_add_ps(_mm_mul_ps(uij, rinv),
 937                                                       _mm_mul_ps(uij3, r)));
 938                 t2            = _mm_sub_ps(t2,
 939                                            _mm_add_ps(_mm_mul_ps(half, uij2),
 940                                                       _mm_mul_ps(prod, uij3)));
 941                 t3            = _mm_mul_ps(_mm_mul_ps(onefourth, logterm),
 942                                            _mm_mul_ps(rinv, rinv));
 943                 t3            = _mm_sub_ps(t3,
 944                                            _mm_mul_ps(_mm_mul_ps(diff2, oneeighth),
 945                                                       _mm_add_ps(one,
 946                                                                  _mm_mul_ps(sk2_rinv, rinv))));
 947                 t1            = _mm_mul_ps(rinv,
 948                                            _mm_add_ps(_mm_mul_ps(dlij, t1),
 949                                                       _mm_add_ps(t2, t3)));
 950
 951                 dadx1         = _mm_and_ps(t1, obc_mask1);
 952             }
 953             else
 954             {
 955                 dadx1         = _mm_setzero_ps();
 956             }
 957
 958             /* Evaluate influence of atom ai -> aj */
 959             t1            = _mm_add_ps(r, sk_ai);
 960             obc_mask1     = _mm_cmplt_ps(raj, t1);
 961
 962             if (_mm_movemask_ps(obc_mask1))
 963             {
 964                 t2            = _mm_sub_ps(r, sk_ai);
 965                 t3            = _mm_sub_ps(sk_ai, r);
 966                 obc_mask2     = _mm_cmplt_ps(raj, t2);
 967                 obc_mask3     = _mm_cmplt_ps(raj, t3);
 968
 969                 uij           = gmx_mm_inv_ps(t1);
 970                 lij           = _mm_or_ps(   _mm_and_ps(obc_mask2, gmx_mm_inv_ps(t2)),
 971                                              _mm_andnot_ps(obc_mask2, raj_inv));
 972                 dlij          = _mm_and_ps(one, obc_mask2);
 973                 uij2          = _mm_mul_ps(uij, uij);
 974                 uij3          = _mm_mul_ps(uij2, uij);
 975                 lij2          = _mm_mul_ps(lij, lij);
 976                 lij3          = _mm_mul_ps(lij2, lij);
 977                 diff2         = _mm_sub_ps(uij2, lij2);
 978                 lij_inv       = gmx_mm_invsqrt_ps(lij2);
 979                 sk2_rinv      = _mm_mul_ps(sk2_ai, rinv);
 980                 prod          = _mm_mul_ps(onefourth, sk2_rinv);
 981                 logterm       = gmx_mm_log_ps(_mm_mul_ps(uij, lij_inv));
 982                 t1            = _mm_sub_ps(lij, uij);
 983                 t2            = _mm_mul_ps(diff2,
 984                                            _mm_sub_ps(_mm_mul_ps(onefourth, r),
 985                                                       prod));
 986                 t3            = _mm_mul_ps(half, _mm_mul_ps(rinv, logterm));
 987                 t1            = _mm_add_ps(t1, _mm_add_ps(t2, t3));
 988                 t4            = _mm_mul_ps(two, _mm_sub_ps(raj_inv, lij));
 989                 t4            = _mm_and_ps(t4, obc_mask3);
 990                 t1            = _mm_mul_ps(half, _mm_add_ps(t1, t4));
 991
 992                 GMX_MM_INCREMENT_4VALUES_PS(work+jnrA, work+jnrB, work+jnrC, work+jnrD, _mm_and_ps(t1, obc_mask1));
 993
 994                 t1            = _mm_add_ps(_mm_mul_ps(half, lij2),
 995                                            _mm_mul_ps(prod, lij3));
 996                 t1            = _mm_sub_ps(t1,
 997                                            _mm_mul_ps(onefourth,
 998                                                       _mm_add_ps(_mm_mul_ps(lij, rinv),
 999                                                                  _mm_mul_ps(lij3, r))));
1000                 t2            = _mm_mul_ps(onefourth,
1001                                            _mm_add_ps(_mm_mul_ps(uij, rinv),
1002                                                       _mm_mul_ps(uij3, r)));
1003                 t2            = _mm_sub_ps(t2,
1004                                            _mm_add_ps(_mm_mul_ps(half, uij2),
1005                                                       _mm_mul_ps(prod, uij3)));
1006                 t3            = _mm_mul_ps(_mm_mul_ps(onefourth, logterm),
1007                                            _mm_mul_ps(rinv, rinv));
1008                 t3            = _mm_sub_ps(t3,
1009                                            _mm_mul_ps(_mm_mul_ps(diff2, oneeighth),
1010                                                       _mm_add_ps(one,
1011                                                                  _mm_mul_ps(sk2_rinv, rinv))));
1012                 t1            = _mm_mul_ps(rinv,
1013                                            _mm_add_ps(_mm_mul_ps(dlij, t1),
1014                                                       _mm_add_ps(t2, t3)));
1015                 dadx2         = _mm_and_ps(t1, obc_mask1);
1016             }
1017             else
1018             {
1019                 dadx2         = _mm_setzero_ps();
1020             }
1021
1022             _mm_store_ps(dadx, dadx1);
1023             dadx += 4;
1024             _mm_store_ps(dadx, dadx2);
1025             dadx += 4;
1026         } /* end normal inner loop */
1027
1028         if (offset != 0)
1029         {
1030             if (offset == 1)
1031             {
1032                 jnrA        = jjnr[k];
1033                 j3A         = 3*jnrA;
1034                 GMX_MM_LOAD_1RVEC_1POINTER_PS(x+j3A, jx, jy, jz);
1035                 GMX_MM_LOAD_1VALUE_PS(gb_radius+jnrA, raj);
1036                 GMX_MM_LOAD_1VALUE_PS(obc_param+jnrA, sk_aj);
1037                 mask        = mask1;
1038             }
1039             else if (offset == 2)
1040             {
1041                 jnrA        = jjnr[k];
1042                 jnrB        = jjnr[k+1];
1043                 j3A         = 3*jnrA;
1044                 j3B         = 3*jnrB;
1045                 GMX_MM_LOAD_1RVEC_2POINTERS_PS(x+j3A, x+j3B, jx, jy, jz);
1046                 GMX_MM_LOAD_2VALUES_PS(gb_radius+jnrA, gb_radius+jnrB, raj);
1047                 GMX_MM_LOAD_2VALUES_PS(obc_param+jnrA, obc_param+jnrB, sk_aj);
1048                 mask        = mask2;
1049             }
1050             else
1051             {
1052                 /* offset must be 3 */
1053                 jnrA        = jjnr[k];
1054                 jnrB        = jjnr[k+1];
1055                 jnrC        = jjnr[k+2];
1056                 j3A         = 3*jnrA;
1057                 j3B         = 3*jnrB;
1058                 j3C         = 3*jnrC;
1059                 GMX_MM_LOAD_1RVEC_3POINTERS_PS(x+j3A, x+j3B, x+j3C, jx, jy, jz);
1060                 GMX_MM_LOAD_3VALUES_PS(gb_radius+jnrA, gb_radius+jnrB, gb_radius+jnrC, raj);
1061                 GMX_MM_LOAD_3VALUES_PS(obc_param+jnrA, obc_param+jnrB, obc_param+jnrC, sk_aj);
1062                 mask        = mask3;
1063             }
1064
1065             dx    = _mm_sub_ps(ix, jx);
1066             dy    = _mm_sub_ps(iy, jy);
1067             dz    = _mm_sub_ps(iz, jz);
1068
1069             rsq         = gmx_mm_calc_rsq_ps(dx, dy, dz);
1070
1071             rinv        = gmx_mm_invsqrt_ps(rsq);
1072             r           = _mm_mul_ps(rsq, rinv);
1073
1074             /* Compute raj_inv aj1-4 */
1075             raj_inv     = gmx_mm_inv_ps(raj);
1076
1077             /* Evaluate influence of atom aj -> ai */
1078             t1            = _mm_add_ps(r, sk_aj);
1079             obc_mask1     = _mm_cmplt_ps(rai, t1);
1080             obc_mask1     = _mm_and_ps(obc_mask1, mask);
1081
1082             if (_mm_movemask_ps(obc_mask1))
1083             {
1084                 t2            = _mm_sub_ps(r, sk_aj);
1085                 t3            = _mm_sub_ps(sk_aj, r);
1086                 obc_mask2     = _mm_cmplt_ps(rai, t2);
1087                 obc_mask3     = _mm_cmplt_ps(rai, t3);
1088
1089                 uij           = gmx_mm_inv_ps(t1);
1090                 lij           = _mm_or_ps(   _mm_and_ps(obc_mask2, gmx_mm_inv_ps(t2)),
1091                                              _mm_andnot_ps(obc_mask2, rai_inv));
1092                 dlij           = _mm_and_ps(one, obc_mask2);
1093                 uij2           = _mm_mul_ps(uij, uij);
1094                 uij3           = _mm_mul_ps(uij2, uij);
1095                 lij2           = _mm_mul_ps(lij, lij);
1096                 lij3           = _mm_mul_ps(lij2, lij);
1097                 diff2          = _mm_sub_ps(uij2, lij2);
1098                 lij_inv        = gmx_mm_invsqrt_ps(lij2);
1099                 sk2_aj         = _mm_mul_ps(sk_aj, sk_aj);
1100                 sk2_rinv       = _mm_mul_ps(sk2_aj, rinv);
1101                 prod           = _mm_mul_ps(onefourth, sk2_rinv);
1102                 logterm        = gmx_mm_log_ps(_mm_mul_ps(uij, lij_inv));
1103                 t1             = _mm_sub_ps(lij, uij);
1104                 t2             = _mm_mul_ps(diff2,
1105                                             _mm_sub_ps(_mm_mul_ps(onefourth, r),
1106                                                        prod));
1107                 t3            = _mm_mul_ps(half, _mm_mul_ps(rinv, logterm));
1108                 t1            = _mm_add_ps(t1, _mm_add_ps(t2, t3));
1109                 t4            = _mm_mul_ps(two, _mm_sub_ps(rai_inv, lij));
1110                 t4            = _mm_and_ps(t4, obc_mask3);
1111                 t1            = _mm_mul_ps(half, _mm_add_ps(t1, t4));
1112                 sum_ai        = _mm_add_ps(sum_ai, _mm_and_ps(t1, obc_mask1));
1113                 t1            = _mm_add_ps(_mm_mul_ps(half, lij2),
1114                                            _mm_mul_ps(prod, lij3));
1115                 t1            = _mm_sub_ps(t1,
1116                                            _mm_mul_ps(onefourth,
1117                                                       _mm_add_ps(_mm_mul_ps(lij, rinv),
1118                                                                  _mm_mul_ps(lij3, r))));
1119                 t2            = _mm_mul_ps(onefourth,
1120                                            _mm_add_ps(_mm_mul_ps(uij, rinv),
1121                                                       _mm_mul_ps(uij3, r)));
1122                 t2            = _mm_sub_ps(t2,
1123                                            _mm_add_ps(_mm_mul_ps(half, uij2),
1124                                                       _mm_mul_ps(prod, uij3)));
1125                 t3            = _mm_mul_ps(_mm_mul_ps(onefourth, logterm),
1126                                            _mm_mul_ps(rinv, rinv));
1127                 t3            = _mm_sub_ps(t3,
1128                                            _mm_mul_ps(_mm_mul_ps(diff2, oneeighth),
1129                                                       _mm_add_ps(one,
1130                                                                  _mm_mul_ps(sk2_rinv, rinv))));
1131                 t1            = _mm_mul_ps(rinv,
1132                                            _mm_add_ps(_mm_mul_ps(dlij, t1),
1133                                                       _mm_add_ps(t2, t3)));
1134                 dadx1         = _mm_and_ps(t1, obc_mask1);
1135             }
1136             else
1137             {
1138                 dadx1         = _mm_setzero_ps();
1139             }
1140
1141             /* Evaluate influence of atom ai -> aj */
1142             t1            = _mm_add_ps(r, sk_ai);
1143             obc_mask1     = _mm_cmplt_ps(raj, t1);
1144             obc_mask1     = _mm_and_ps(obc_mask1, mask);
1145
1146             if (_mm_movemask_ps(obc_mask1))
1147             {
1148                 t2            = _mm_sub_ps(r, sk_ai);
1149                 t3            = _mm_sub_ps(sk_ai, r);
1150                 obc_mask2     = _mm_cmplt_ps(raj, t2);
1151                 obc_mask3     = _mm_cmplt_ps(raj, t3);
1152
1153                 uij           = gmx_mm_inv_ps(t1);
1154                 lij           = _mm_or_ps(_mm_and_ps(obc_mask2, gmx_mm_inv_ps(t2)),
1155                                           _mm_andnot_ps(obc_mask2, raj_inv));
1156                 dlij          = _mm_and_ps(one, obc_mask2);
1157                 uij2          = _mm_mul_ps(uij, uij);
1158                 uij3          = _mm_mul_ps(uij2, uij);
1159                 lij2          = _mm_mul_ps(lij, lij);
1160                 lij3          = _mm_mul_ps(lij2, lij);
1161                 diff2         = _mm_sub_ps(uij2, lij2);
1162                 lij_inv       = gmx_mm_invsqrt_ps(lij2);
1163                 sk2_rinv      = _mm_mul_ps(sk2_ai, rinv);
1164                 prod          = _mm_mul_ps(onefourth, sk2_rinv);
1165                 logterm       = gmx_mm_log_ps(_mm_mul_ps(uij, lij_inv));
1166                 t1            = _mm_sub_ps(lij, uij);
1167                 t2            = _mm_mul_ps(diff2,
1168                                            _mm_sub_ps(_mm_mul_ps(onefourth, r),
1169                                                       prod));
1170                 t3            = _mm_mul_ps(half, _mm_mul_ps(rinv, logterm));
1171                 t1            = _mm_add_ps(t1, _mm_add_ps(t2, t3));
1172                 t4            = _mm_mul_ps(two, _mm_sub_ps(raj_inv, lij));
1173                 t4            = _mm_and_ps(t4, obc_mask3);
1174                 t1            = _mm_mul_ps(half, _mm_add_ps(t1, t4));
1175
1176                 tmp           = _mm_and_ps(t1, obc_mask1);
1177
1178                 t1            = _mm_add_ps(_mm_mul_ps(half, lij2),
1179                                            _mm_mul_ps(prod, lij3));
1180                 t1            = _mm_sub_ps(t1,
1181                                            _mm_mul_ps(onefourth,
1182                                                       _mm_add_ps(_mm_mul_ps(lij, rinv),
1183                                                                  _mm_mul_ps(lij3, r))));
1184                 t2            = _mm_mul_ps(onefourth,
1185                                            _mm_add_ps(_mm_mul_ps(uij, rinv),
1186                                                       _mm_mul_ps(uij3, r)));
1187                 t2            = _mm_sub_ps(t2,
1188                                            _mm_add_ps(_mm_mul_ps(half, uij2),
1189                                                       _mm_mul_ps(prod, uij3)));
1190                 t3            = _mm_mul_ps(_mm_mul_ps(onefourth, logterm),
1191                                            _mm_mul_ps(rinv, rinv));
1192                 t3            = _mm_sub_ps(t3,
1193                                            _mm_mul_ps(_mm_mul_ps(diff2, oneeighth),
1194                                                       _mm_add_ps(one,
1195                                                                  _mm_mul_ps(sk2_rinv, rinv))));
1196                 t1            = _mm_mul_ps(rinv,
1197                                            _mm_add_ps(_mm_mul_ps(dlij, t1),
1198                                                       _mm_add_ps(t2, t3)));
1199                 dadx2         = _mm_and_ps(t1, obc_mask1);
1200             }
1201             else
1202             {
1203                 dadx2         = _mm_setzero_ps();
1204                 tmp           = _mm_setzero_ps();
1205             }
1206
1207             _mm_store_ps(dadx, dadx1);
1208             dadx += 4;
1209             _mm_store_ps(dadx, dadx2);
1210             dadx += 4;
1211
1212             if (offset == 1)
1213             {
1214                 GMX_MM_INCREMENT_1VALUE_PS(work+jnrA, tmp);
1215             }
1216             else if (offset == 2)
1217             {
1218                 GMX_MM_INCREMENT_2VALUES_PS(work+jnrA, work+jnrB, tmp);
1219             }
1220             else
1221             {
1222                 /* offset must be 3 */
1223                 GMX_MM_INCREMENT_3VALUES_PS(work+jnrA, work+jnrB, work+jnrC, tmp);
1224             }
1225
1226         }
1227         GMX_MM_UPDATE_1POT_PS(sum_ai, work+ii);
1228
1229     }
1230
1231     /* Parallel summations */
1232     if (PARTDECOMP(cr))
1233     {
1234         gmx_sum(natoms, work, cr);
1235     }
1236     else if (DOMAINDECOMP(cr))
1237     {
1238         dd_atom_sum_real(cr->dd, work);
1239     }
1240
1241     if (gb_algorithm == egbHCT)
1242     {
1243         /* HCT */
1244         for (i = 0; i < fr->natoms_force; i++) /* PELA born->nr */
1245         {
1246             if (born->use[i] != 0)
1247             {
1248                 rr      = top->atomtypes.gb_radius[md->typeA[i]]-doffset;
1249                 sum     = 1.0/rr - work[i];
1250                 min_rad = rr + doffset;
1251                 rad     = 1.0/sum;
1252
1253                 born->bRad[i]   = rad > min_rad ? rad : min_rad;
1254                 fr->invsqrta[i] = gmx_invsqrt(born->bRad[i]);
1255             }
1256         }
1257
1258         /* Extra communication required for DD */
1259         if (DOMAINDECOMP(cr))
1260         {
1261             dd_atom_spread_real(cr->dd, born->bRad);
1262             dd_atom_spread_real(cr->dd, fr->invsqrta);
1263         }
1264     }
1265     else
1266     {
1267         /* OBC */
1268         for (i = 0; i < fr->natoms_force; i++) /* PELA born->nr */
1269         {
1270             if (born->use[i] != 0)
1271             {
1272                 rr      = top->atomtypes.gb_radius[md->typeA[i]];
1273                 rr_inv2 = 1.0/rr;
1274                 rr      = rr-doffset;
1275                 rr_inv  = 1.0/rr;
1276                 sum     = rr * work[i];
1277                 sum2    = sum  * sum;
1278                 sum3    = sum2 * sum;
1279
1280                 tsum          = tanh(born->obc_alpha*sum-born->obc_beta*sum2+born->obc_gamma*sum3);
1281                 born->bRad[i] = rr_inv - tsum*rr_inv2;
1282                 born->bRad[i] = 1.0 / born->bRad[i];
1283
1284                 fr->invsqrta[i] = gmx_invsqrt(born->bRad[i]);
1285
1286                 tchain         = rr * (born->obc_alpha-2*born->obc_beta*sum+3*born->obc_gamma*sum2);
1287                 born->drobc[i] = (1.0-tsum*tsum)*tchain*rr_inv2;
1288             }
1289         }
1290         /* Extra (local) communication required for DD */
1291         if (DOMAINDECOMP(cr))
1292         {
1293             dd_atom_spread_real(cr->dd, born->bRad);
1294             dd_atom_spread_real(cr->dd, fr->invsqrta);
1295             dd_atom_spread_real(cr->dd, born->drobc);
1296         }
1297     }
1298
1299
1300
1301     return 0;
1302 }
1303
1304
1305
1306 float calc_gb_chainrule_sse2_single(int natoms, t_nblist *nl, float *dadx, float *dvda,
1307                                     float *x, float *f, float *fshift, float *shiftvec,
1308                                     int gb_algorithm, gmx_genborn_t *born, t_mdatoms *md)
1309 {
1310     int          i, k, n, ii, jnr, ii3, is3, nj0, nj1, offset, n0, n1;
1311     int          jnrA, jnrB, jnrC, jnrD;
1312     int          j3A, j3B, j3C, j3D;
1313     int          jnrE, jnrF, jnrG, jnrH;
1314     int          j3E, j3F, j3G, j3H;
1315     int       *  jjnr;
1316
1317     float        rbi, shX, shY, shZ;
1318     float       *rb;
1319
1320     __m128       ix, iy, iz;
1321     __m128       jx, jy, jz;
1322     __m128       jxB, jyB, jzB;
1323     __m128       fix, fiy, fiz;
1324     __m128       dx, dy, dz;
1325     __m128       tx, ty, tz;
1326     __m128       dxB, dyB, dzB;
1327     __m128       txB, tyB, tzB;
1328
1329     __m128       rbai, rbaj, rbajB, f_gb, f_gb_ai, f_gbB, f_gb_aiB;
1330     __m128       xmm1, xmm2, xmm3;
1331
1332     const __m128 two = _mm_set1_ps(2.0f);
1333
1334     rb     = born->work;
1335
1336     jjnr   = nl->jjnr;
1337
1338     /* Loop to get the proper form for the Born radius term, sse style */
1339     offset = natoms%4;
1340
1341     n0 = 0;
1342     n1 = natoms;
1343
1344     if (gb_algorithm == egbSTILL)
1345     {
1346         for (i = n0; i < n1; i++)
1347         {
1348             rbi   = born->bRad[i];
1349             rb[i] = (2 * rbi * rbi * dvda[i])/ONE_4PI_EPS0;
1350         }
1351     }
1352     else if (gb_algorithm == egbHCT)
1353     {
1354         for (i = n0; i < n1; i++)
1355         {
1356             rbi   = born->bRad[i];
1357             rb[i] = rbi * rbi * dvda[i];
1358         }
1359     }
1360     else if (gb_algorithm == egbOBC)
1361     {
1362         for (i = n0; i < n1; i++)
1363         {
1364             rbi   = born->bRad[i];
1365             rb[i] = rbi * rbi * born->drobc[i] * dvda[i];
1366         }
1367     }
1368
1369     jz = _mm_setzero_ps();
1370
1371     n = j3A = j3B = j3C = j3D = 0;
1372
1373     for (i = 0; i < nl->nri; i++)
1374     {
1375         ii     = nl->iinr[i];
1376         ii3    = ii*3;
1377         is3    = 3*nl->shift[i];
1378         shX    = shiftvec[is3];
1379         shY    = shiftvec[is3+1];
1380         shZ    = shiftvec[is3+2];
1381         nj0    = nl->jindex[i];
1382         nj1    = nl->jindex[i+1];
1383
1384         ix     = _mm_set1_ps(shX+x[ii3+0]);
1385         iy     = _mm_set1_ps(shY+x[ii3+1]);
1386         iz     = _mm_set1_ps(shZ+x[ii3+2]);
1387
1388         offset = (nj1-nj0)%4;
1389
1390         rbai   = _mm_load1_ps(rb+ii);
1391         fix    = _mm_setzero_ps();
1392         fiy    = _mm_setzero_ps();
1393         fiz    = _mm_setzero_ps();
1394
1395
1396         for (k = nj0; k < nj1-offset; k += 4)
1397         {
1398             jnrA        = jjnr[k];
1399             jnrB        = jjnr[k+1];
1400             jnrC        = jjnr[k+2];
1401             jnrD        = jjnr[k+3];
1402
1403             j3A         = 3*jnrA;
1404             j3B         = 3*jnrB;
1405             j3C         = 3*jnrC;
1406             j3D         = 3*jnrD;
1407
1408             GMX_MM_LOAD_1RVEC_4POINTERS_PS(x+j3A, x+j3B, x+j3C, x+j3D, jx, jy, jz);
1409
1410             dx          = _mm_sub_ps(ix, jx);
1411             dy          = _mm_sub_ps(iy, jy);
1412             dz          = _mm_sub_ps(iz, jz);
1413
1414             GMX_MM_LOAD_4VALUES_PS(rb+jnrA, rb+jnrB, rb+jnrC, rb+jnrD, rbaj);
1415
1416             /* load chain rule terms for j1-4 */
1417             f_gb        = _mm_load_ps(dadx);
1418             dadx       += 4;
1419             f_gb_ai     = _mm_load_ps(dadx);
1420             dadx       += 4;
1421
1422             /* calculate scalar force */
1423             f_gb    = _mm_mul_ps(f_gb, rbai);
1424             f_gb_ai = _mm_mul_ps(f_gb_ai, rbaj);
1425             f_gb    = _mm_add_ps(f_gb, f_gb_ai);
1426
1427             tx     = _mm_mul_ps(f_gb, dx);
1428             ty     = _mm_mul_ps(f_gb, dy);
1429             tz     = _mm_mul_ps(f_gb, dz);
1430
1431             fix    = _mm_add_ps(fix, tx);
1432             fiy    = _mm_add_ps(fiy, ty);
1433             fiz    = _mm_add_ps(fiz, tz);
1434
1435             GMX_MM_DECREMENT_1RVEC_4POINTERS_PS(f+j3A, f+j3B, f+j3C, f+j3D, tx, ty, tz);
1436         }
1437
1438         /*deal with odd elements */
1439         if (offset != 0)
1440         {
1441             if (offset == 1)
1442             {
1443                 jnrA        = jjnr[k];
1444                 j3A         = 3*jnrA;
1445                 GMX_MM_LOAD_1RVEC_1POINTER_PS(x+j3A, jx, jy, jz);
1446                 GMX_MM_LOAD_1VALUE_PS(rb+jnrA, rbaj);
1447             }
1448             else if (offset == 2)
1449             {
1450                 jnrA        = jjnr[k];
1451                 jnrB        = jjnr[k+1];
1452                 j3A         = 3*jnrA;
1453                 j3B         = 3*jnrB;
1454                 GMX_MM_LOAD_1RVEC_2POINTERS_PS(x+j3A, x+j3B, jx, jy, jz);
1455                 GMX_MM_LOAD_2VALUES_PS(rb+jnrA, rb+jnrB, rbaj);
1456             }
1457             else
1458             {
1459                 /* offset must be 3 */
1460                 jnrA        = jjnr[k];
1461                 jnrB        = jjnr[k+1];
1462                 jnrC        = jjnr[k+2];
1463                 j3A         = 3*jnrA;
1464                 j3B         = 3*jnrB;
1465                 j3C         = 3*jnrC;
1466                 GMX_MM_LOAD_1RVEC_3POINTERS_PS(x+j3A, x+j3B, x+j3C, jx, jy, jz);
1467                 GMX_MM_LOAD_3VALUES_PS(rb+jnrA, rb+jnrB, rb+jnrC, rbaj);
1468             }
1469
1470             dx          = _mm_sub_ps(ix, jx);
1471             dy          = _mm_sub_ps(iy, jy);
1472             dz          = _mm_sub_ps(iz, jz);
1473
1474             /* load chain rule terms for j1-4 */
1475             f_gb        = _mm_load_ps(dadx);
1476             dadx       += 4;
1477             f_gb_ai     = _mm_load_ps(dadx);
1478             dadx       += 4;
1479
1480             /* calculate scalar force */
1481             f_gb    = _mm_mul_ps(f_gb, rbai);
1482             f_gb_ai = _mm_mul_ps(f_gb_ai, rbaj);
1483             f_gb    = _mm_add_ps(f_gb, f_gb_ai);
1484
1485             tx     = _mm_mul_ps(f_gb, dx);
1486             ty     = _mm_mul_ps(f_gb, dy);
1487             tz     = _mm_mul_ps(f_gb, dz);
1488
1489             fix    = _mm_add_ps(fix, tx);
1490             fiy    = _mm_add_ps(fiy, ty);
1491             fiz    = _mm_add_ps(fiz, tz);
1492
1493             if (offset == 1)
1494             {
1495                 GMX_MM_DECREMENT_1RVEC_1POINTER_PS(f+j3A, tx, ty, tz);
1496             }
1497             else if (offset == 2)
1498             {
1499                 GMX_MM_DECREMENT_1RVEC_2POINTERS_PS(f+j3A, f+j3B, tx, ty, tz);
1500             }
1501             else
1502             {
1503                 /* offset must be 3 */
1504                 GMX_MM_DECREMENT_1RVEC_3POINTERS_PS(f+j3A, f+j3B, f+j3C, tx, ty, tz);
1505             }
1506         }
1507
1508         /* fix/fiy/fiz now contain four partial force terms, that all should be
1509          * added to the i particle forces and shift forces.
1510          */
1511         gmx_mm_update_iforce_1atom_ps(&fix, &fiy, &fiz, f+ii3, fshift+is3);
1512     }
1513
1514     return 0;
1515 }
1516
1517
1518 #else
1519 /* keep compiler happy */
1520 int genborn_sse_dummy;
1521
1522 #endif /* SSE intrinsics available */