src/gromacs/nbnxm/kerneldispatch.cpp

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014,2015,2016,2017,2018,2019, 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 #include "gmxpre.h"
  37
  38 #include "gromacs/gmxlib/nrnb.h"
  39 #include "gromacs/gmxlib/nonbonded/nb_free_energy.h"
  40 #include "gromacs/gmxlib/nonbonded/nb_kernel.h"
  41 #include "gromacs/gmxlib/nonbonded/nonbonded.h"
  42 #include "gromacs/math/vectypes.h"
  43 #include "gromacs/mdlib/force.h"
  44 #include "gromacs/mdlib/force_flags.h"
  45 #include "gromacs/mdlib/gmx_omp_nthreads.h"
  46 #include "gromacs/mdtypes/enerdata.h"
  47 #include "gromacs/mdtypes/inputrec.h"
  48 #include "gromacs/mdtypes/interaction_const.h"
  49 #include "gromacs/mdtypes/md_enums.h"
  50 #include "gromacs/mdtypes/mdatom.h"
  51 #include "gromacs/nbnxm/gpu_data_mgmt.h"
  52 #include "gromacs/nbnxm/nbnxm.h"
  53 #include "gromacs/nbnxm/nbnxm_simd.h"
  54 #include "gromacs/nbnxm/kernels_reference/kernel_gpu_ref.h"
  55 #include "gromacs/simd/simd.h"
  56 #include "gromacs/utility/gmxassert.h"
  57 #include "gromacs/utility/real.h"
  58
  59 #include "kernel_common.h"
  60 #define INCLUDE_KERNELFUNCTION_TABLES
  61 #include "gromacs/nbnxm/kernels_reference/kernel_ref.h"
  62 #ifdef GMX_NBNXN_SIMD_2XNN
  63 #include "gromacs/nbnxm/kernels_simd_2xmm/kernels.h"
  64 #endif
  65 #ifdef GMX_NBNXN_SIMD_4XN
  66 #include "gromacs/nbnxm/kernels_simd_4xm/kernels.h"
  67 #endif
  68 #undef INCLUDE_FUNCTION_TABLES
  69
  70 /*! \brief Clears the energy group output buffers
  71  *
  72  * \param[in,out] out  nbnxn kernel output struct
  73  */
  74 static void clearGroupEnergies(nbnxn_atomdata_output_t *out)
  75 {
  76     std::fill(out->Vvdw.begin(), out->Vvdw.end(), 0.0_real);
  77     std::fill(out->Vc.begin(), out->Vc.end(), 0.0_real);
  78     std::fill(out->VSvdw.begin(), out->VSvdw.end(), 0.0_real);
  79     std::fill(out->VSc.begin(), out->VSc.end(), 0.0_real);
  80 }
  81
  82 /*! \brief Reduce the group-pair energy buffers produced by a SIMD kernel
  83  * to single terms in the output buffers.
  84  *
  85  * The SIMD kernels produce a large number of energy buffer in SIMD registers
  86  * to avoid scattered reads and writes.
  87  *
  88  * \tparam        unrollj         The unroll size for j-particles in the SIMD kernel
  89  * \param[in]     numGroups       The number of energy groups
  90  * \param[in]     numGroups_2log  Log2 of numGroups, rounded up
  91  * \param[in,out] out             Struct with energy buffers
  92  */
  93 template <int unrollj> static void
  94 reduceGroupEnergySimdBuffers(int                       numGroups,
  95                              int                       numGroups_2log,
  96                              nbnxn_atomdata_output_t  *out)
  97 {
  98     const int                 unrollj_half     = unrollj/2;
  99     /* Energies are stored in SIMD registers with size 2^numGroups_2log */
 100     const int                 numGroupsStorage = (1 << numGroups_2log);
 101
 102     const real * gmx_restrict vVdwSimd     = out->VSvdw.data();
 103     const real * gmx_restrict vCoulombSimd = out->VSc.data();
 104     real * gmx_restrict       vVdw         = out->Vvdw.data();
 105     real * gmx_restrict       vCoulomb     = out->Vc.data();
 106
 107     /* The size of the SIMD energy group buffer array is:
 108      * numGroups*numGroups*numGroupsStorage*unrollj_half*simd_width
 109      */
 110     for (int i = 0; i < numGroups; i++)
 111     {
 112         for (int j1 = 0; j1 < numGroups; j1++)
 113         {
 114             for (int j0 = 0; j0 < numGroups; j0++)
 115             {
 116                 int c = ((i*numGroups + j1)*numGroupsStorage + j0)*unrollj_half*unrollj;
 117                 for (int s = 0; s < unrollj_half; s++)
 118                 {
 119                     vVdw    [i*numGroups + j0] += vVdwSimd    [c + 0];
 120                     vVdw    [i*numGroups + j1] += vVdwSimd    [c + 1];
 121                     vCoulomb[i*numGroups + j0] += vCoulombSimd[c + 0];
 122                     vCoulomb[i*numGroups + j1] += vCoulombSimd[c + 1];
 123                     c                          += unrollj + 2;
 124                 }
 125             }
 126         }
 127     }
 128 }
 129
 130 /*! \brief Dispatches the non-bonded N versus M atom cluster CPU kernels.
 131  *
 132  * OpenMP parallelization is performed within this function.
 133  * Energy reduction, but not force and shift force reduction, is performed
 134  * within this function.
 135  *
 136  * \param[in]     pairlistSet   Pairlists with local or non-local interactions to compute
 137  * \param[in]     kernelSetup   The non-bonded kernel setup
 138  * \param[in,out] nbat          The atomdata for the interactions
 139  * \param[in]     ic            Non-bonded interaction constants
 140  * \param[in]     shiftVectors  The PBC shift vectors
 141  * \param[in]     forceFlags    Flags that tell what to compute
 142  * \param[in]     clearF        Enum that tells if to clear the force output buffer
 143  * \param[out]    fshift        Shift force output buffer
 144  * \param[out]    vCoulomb      Output buffer for Coulomb energies
 145  * \param[out]    vVdw          Output buffer for Van der Waals energies
 146  */
 147 static void
 148 nbnxn_kernel_cpu(const nbnxn_pairlist_set_t     &pairlistSet,
 149                  const Nbnxm::KernelSetup       &kernelSetup,
 150                  nbnxn_atomdata_t               *nbat,
 151                  const interaction_const_t      &ic,
 152                  rvec                           *shiftVectors,
 153                  int                             forceFlags,
 154                  int                             clearF,
 155                  real                           *fshift,
 156                  real                           *vCoulomb,
 157                  real                           *vVdw)
 158 {
 159
 160     int                      coulkt;
 161     if (EEL_RF(ic.eeltype) || ic.eeltype == eelCUT)
 162     {
 163         coulkt = coulktRF;
 164     }
 165     else
 166     {
 167         if (kernelSetup.ewaldExclusionType == Nbnxm::EwaldExclusionType::Table)
 168         {
 169             if (ic.rcoulomb == ic.rvdw)
 170             {
 171                 coulkt = coulktTAB;
 172             }
 173             else
 174             {
 175                 coulkt = coulktTAB_TWIN;
 176             }
 177         }
 178         else
 179         {
 180             if (ic.rcoulomb == ic.rvdw)
 181             {
 182                 coulkt = coulktEWALD;
 183             }
 184             else
 185             {
 186                 coulkt = coulktEWALD_TWIN;
 187             }
 188         }
 189     }
 190
 191     const nbnxn_atomdata_t::Params &nbatParams = nbat->params();
 192
 193     int vdwkt = 0;
 194     if (ic.vdwtype == evdwCUT)
 195     {
 196         switch (ic.vdw_modifier)
 197         {
 198             case eintmodNONE:
 199             case eintmodPOTSHIFT:
 200                 switch (nbatParams.comb_rule)
 201                 {
 202                     case ljcrGEOM: vdwkt = vdwktLJCUT_COMBGEOM; break;
 203                     case ljcrLB:   vdwkt = vdwktLJCUT_COMBLB;   break;
 204                     case ljcrNONE: vdwkt = vdwktLJCUT_COMBNONE; break;
 205                     default:
 206                         GMX_RELEASE_ASSERT(false, "Unknown combination rule");
 207                 }
 208                 break;
 209             case eintmodFORCESWITCH:
 210                 vdwkt = vdwktLJFORCESWITCH;
 211                 break;
 212             case eintmodPOTSWITCH:
 213                 vdwkt = vdwktLJPOTSWITCH;
 214                 break;
 215             default:
 216                 GMX_RELEASE_ASSERT(false, "Unsupported VdW interaction modifier");
 217         }
 218     }
 219     else if (ic.vdwtype == evdwPME)
 220     {
 221         if (ic.ljpme_comb_rule == eljpmeGEOM)
 222         {
 223             vdwkt = vdwktLJEWALDCOMBGEOM;
 224         }
 225         else
 226         {
 227             vdwkt = vdwktLJEWALDCOMBLB;
 228             /* At setup we (should have) selected the C reference kernel */
 229             GMX_RELEASE_ASSERT(kernelSetup.kernelType == Nbnxm::KernelType::Cpu4x4_PlainC, "Only the C reference nbnxn SIMD kernel supports LJ-PME with LB combination rules");
 230         }
 231     }
 232     else
 233     {
 234         GMX_RELEASE_ASSERT(false, "Unsupported VdW interaction type");
 235     }
 236
 237     int                        nnbl = pairlistSet.nnbl;
 238     NbnxnPairlistCpu * const * nbl  = pairlistSet.nbl;
 239
 240     int gmx_unused             nthreads = gmx_omp_nthreads_get(emntNonbonded);
 241 #pragma omp parallel for schedule(static) num_threads(nthreads)
 242     for (int nb = 0; nb < nnbl; nb++)
 243     {
 244         // Presently, the kernels do not call C++ code that can throw,
 245         // so no need for a try/catch pair in this OpenMP region.
 246         nbnxn_atomdata_output_t *out = &nbat->out[nb];
 247
 248         if (clearF == enbvClearFYes)
 249         {
 250             clear_f(nbat, nb, out->f.data());
 251         }
 252
 253         real *fshift_p;
 254         if ((forceFlags & GMX_FORCE_VIRIAL) && nnbl == 1)
 255         {
 256             fshift_p = fshift;
 257         }
 258         else
 259         {
 260             fshift_p = out->fshift.data();
 261
 262             if (clearF == enbvClearFYes)
 263             {
 264                 clear_fshift(fshift_p);
 265             }
 266         }
 267
 268         if (!(forceFlags & GMX_FORCE_ENERGY))
 269         {
 270             /* Don't calculate energies */
 271             switch (kernelSetup.kernelType)
 272             {
 273                 case Nbnxm::KernelType::Cpu4x4_PlainC:
 274                     nbnxn_kernel_noener_ref[coulkt][vdwkt](nbl[nb], nbat,
 275                                                            &ic,
 276                                                            shiftVectors,
 277                                                            out->f.data(),
 278                                                            fshift_p);
 279                     break;
 280 #ifdef GMX_NBNXN_SIMD_2XNN
 281                 case Nbnxm::KernelType::Cpu4xN_Simd_2xNN:
 282                     nbnxm_kernel_noener_simd_2xmm[coulkt][vdwkt](nbl[nb], nbat,
 283                                                                  &ic,
 284                                                                  shiftVectors,
 285                                                                  out->f.data(),
 286                                                                  fshift_p);
 287                     break;
 288 #endif
 289 #ifdef GMX_NBNXN_SIMD_4XN
 290                 case Nbnxm::KernelType::Cpu4xN_Simd_4xN:
 291                     nbnxm_kernel_noener_simd_4xm[coulkt][vdwkt](nbl[nb], nbat,
 292                                                                 &ic,
 293                                                                 shiftVectors,
 294                                                                 out->f.data(),
 295                                                                 fshift_p);
 296                     break;
 297 #endif
 298                 default:
 299                     GMX_RELEASE_ASSERT(false, "Unsupported kernel architecture");
 300             }
 301         }
 302         else if (out->Vvdw.size() == 1)
 303         {
 304             /* A single energy group (pair) */
 305             out->Vvdw[0] = 0;
 306             out->Vc[0]   = 0;
 307
 308             switch (kernelSetup.kernelType)
 309             {
 310                 case Nbnxm::KernelType::Cpu4x4_PlainC:
 311                     nbnxn_kernel_ener_ref[coulkt][vdwkt](nbl[nb], nbat,
 312                                                          &ic,
 313                                                          shiftVectors,
 314                                                          out->f.data(),
 315                                                          fshift_p,
 316                                                          out->Vvdw.data(),
 317                                                          out->Vc.data());
 318                     break;
 319 #ifdef GMX_NBNXN_SIMD_2XNN
 320                 case Nbnxm::KernelType::Cpu4xN_Simd_2xNN:
 321                     nbnxm_kernel_ener_simd_2xmm[coulkt][vdwkt](nbl[nb], nbat,
 322                                                                &ic,
 323                                                                shiftVectors,
 324                                                                out->f.data(),
 325                                                                fshift_p,
 326                                                                out->Vvdw.data(),
 327                                                                out->Vc.data());
 328                     break;
 329 #endif
 330 #ifdef GMX_NBNXN_SIMD_4XN
 331                 case Nbnxm::KernelType::Cpu4xN_Simd_4xN:
 332                     nbnxm_kernel_ener_simd_4xm[coulkt][vdwkt](nbl[nb], nbat,
 333                                                               &ic,
 334                                                               shiftVectors,
 335                                                               out->f.data(),
 336                                                               fshift_p,
 337                                                               out->Vvdw.data(),
 338                                                               out->Vc.data());
 339                     break;
 340 #endif
 341                 default:
 342                     GMX_RELEASE_ASSERT(false, "Unsupported kernel architecture");
 343             }
 344         }
 345         else
 346         {
 347             /* Calculate energy group contributions */
 348             clearGroupEnergies(out);
 349
 350             int unrollj = 0;
 351
 352             switch (kernelSetup.kernelType)
 353             {
 354                 case Nbnxm::KernelType::Cpu4x4_PlainC:
 355                     unrollj = c_nbnxnCpuIClusterSize;
 356                     nbnxn_kernel_energrp_ref[coulkt][vdwkt](nbl[nb], nbat,
 357                                                             &ic,
 358                                                             shiftVectors,
 359                                                             out->f.data(),
 360                                                             fshift_p,
 361                                                             out->Vvdw.data(),
 362                                                             out->Vc.data());
 363                     break;
 364 #ifdef GMX_NBNXN_SIMD_2XNN
 365                 case Nbnxm::KernelType::Cpu4xN_Simd_2xNN:
 366                     unrollj = GMX_SIMD_REAL_WIDTH/2;
 367                     nbnxm_kernel_energrp_simd_2xmm[coulkt][vdwkt](nbl[nb], nbat,
 368                                                                   &ic,
 369                                                                   shiftVectors,
 370                                                                   out->f.data(),
 371                                                                   fshift_p,
 372                                                                   out->VSvdw.data(),
 373                                                                   out->VSc.data());
 374                     break;
 375 #endif
 376 #ifdef GMX_NBNXN_SIMD_4XN
 377                 case Nbnxm::KernelType::Cpu4xN_Simd_4xN:
 378                     unrollj = GMX_SIMD_REAL_WIDTH;
 379                     nbnxm_kernel_energrp_simd_4xm[coulkt][vdwkt](nbl[nb], nbat,
 380                                                                  &ic,
 381                                                                  shiftVectors,
 382                                                                  out->f.data(),
 383                                                                  fshift_p,
 384                                                                  out->VSvdw.data(),
 385                                                                  out->VSc.data());
 386                     break;
 387 #endif
 388                 default:
 389                     GMX_RELEASE_ASSERT(false, "Unsupported kernel architecture");
 390             }
 391
 392             if (kernelSetup.kernelType != Nbnxm::KernelType::Cpu4x4_PlainC)
 393             {
 394                 switch (unrollj)
 395                 {
 396                     case 2:
 397                         reduceGroupEnergySimdBuffers<2>(nbatParams.nenergrp,
 398                                                         nbatParams.neg_2log,
 399                                                         out);
 400                         break;
 401                     case 4:
 402                         reduceGroupEnergySimdBuffers<4>(nbatParams.nenergrp,
 403                                                         nbatParams.neg_2log,
 404                                                         out);
 405                         break;
 406                     case 8:
 407                         reduceGroupEnergySimdBuffers<8>(nbatParams.nenergrp,
 408                                                         nbatParams.neg_2log,
 409                                                         out);
 410                         break;
 411                     default:
 412                         GMX_RELEASE_ASSERT(false, "Unsupported j-unroll size");
 413                 }
 414             }
 415         }
 416     }
 417
 418     if (forceFlags & GMX_FORCE_ENERGY)
 419     {
 420         reduce_energies_over_lists(nbat, nnbl, vVdw, vCoulomb);
 421     }
 422 }
 423
 424 static void accountFlops(t_nrnb                           *nrnb,
 425                          const nbnxn_pairlist_set_t       &pairlistSet,
 426                          const nonbonded_verlet_t         &nbv,
 427                          const interaction_const_t        &ic,
 428                          const int                         forceFlags)
 429 {
 430     const bool usingGpuKernels = nbv.useGpu();
 431
 432     int        enr_nbnxn_kernel_ljc;
 433     if (EEL_RF(ic.eeltype) || ic.eeltype == eelCUT)
 434     {
 435         enr_nbnxn_kernel_ljc = eNR_NBNXN_LJ_RF;
 436     }
 437     else if ((!usingGpuKernels && nbv.kernelSetup().ewaldExclusionType == Nbnxm::EwaldExclusionType::Analytical) ||
 438              (usingGpuKernels && Nbnxm::gpu_is_kernel_ewald_analytical(nbv.gpu_nbv)))
 439     {
 440         enr_nbnxn_kernel_ljc = eNR_NBNXN_LJ_EWALD;
 441     }
 442     else
 443     {
 444         enr_nbnxn_kernel_ljc = eNR_NBNXN_LJ_TAB;
 445     }
 446     int enr_nbnxn_kernel_lj = eNR_NBNXN_LJ;
 447     if (forceFlags & GMX_FORCE_ENERGY)
 448     {
 449         /* In eNR_??? the nbnxn F+E kernels are always the F kernel + 1 */
 450         enr_nbnxn_kernel_ljc += 1;
 451         enr_nbnxn_kernel_lj  += 1;
 452     }
 453
 454     inc_nrnb(nrnb, enr_nbnxn_kernel_ljc,
 455              pairlistSet.natpair_ljq);
 456     inc_nrnb(nrnb, enr_nbnxn_kernel_lj,
 457              pairlistSet.natpair_lj);
 458     /* The Coulomb-only kernels are offset -eNR_NBNXN_LJ_RF+eNR_NBNXN_RF */
 459     inc_nrnb(nrnb, enr_nbnxn_kernel_ljc-eNR_NBNXN_LJ_RF+eNR_NBNXN_RF,
 460              pairlistSet.natpair_q);
 461
 462     const bool calcEnergy = ((forceFlags & GMX_FORCE_ENERGY) != 0);
 463     if (ic.vdw_modifier == eintmodFORCESWITCH)
 464     {
 465         /* We add up the switch cost separately */
 466         inc_nrnb(nrnb, eNR_NBNXN_ADD_LJ_FSW + (calcEnergy ? 1 : 0),
 467                  pairlistSet.natpair_ljq + pairlistSet.natpair_lj);
 468     }
 469     if (ic.vdw_modifier == eintmodPOTSWITCH)
 470     {
 471         /* We add up the switch cost separately */
 472         inc_nrnb(nrnb, eNR_NBNXN_ADD_LJ_PSW + (calcEnergy ? 1 : 0),
 473                  pairlistSet.natpair_ljq + pairlistSet.natpair_lj);
 474     }
 475     if (ic.vdwtype == evdwPME)
 476     {
 477         /* We add up the LJ Ewald cost separately */
 478         inc_nrnb(nrnb, eNR_NBNXN_ADD_LJ_EWALD + (calcEnergy ? 1 : 0),
 479                  pairlistSet.natpair_ljq + pairlistSet.natpair_lj);
 480     }
 481 }
 482
 483 void
 484 nonbonded_verlet_t::dispatchNonbondedKernel(Nbnxm::InteractionLocality iLocality,
 485                                             const interaction_const_t &ic,
 486                                             int                        forceFlags,
 487                                             int                        clearF,
 488                                             t_forcerec                *fr,
 489                                             gmx_enerdata_t            *enerd,
 490                                             t_nrnb                    *nrnb)
 491 {
 492     const nbnxn_pairlist_set_t &pairlistSet = pairlistSets().pairlistSet(iLocality);
 493
 494     switch (kernelSetup().kernelType)
 495     {
 496         case Nbnxm::KernelType::Cpu4x4_PlainC:
 497         case Nbnxm::KernelType::Cpu4xN_Simd_4xN:
 498         case Nbnxm::KernelType::Cpu4xN_Simd_2xNN:
 499             nbnxn_kernel_cpu(pairlistSet,
 500                              kernelSetup(),
 501                              nbat,
 502                              ic,
 503                              fr->shift_vec,
 504                              forceFlags,
 505                              clearF,
 506                              fr->fshift[0],
 507                              enerd->grpp.ener[egCOULSR],
 508                              fr->bBHAM ?
 509                              enerd->grpp.ener[egBHAMSR] :
 510                              enerd->grpp.ener[egLJSR]);
 511             break;
 512
 513         case Nbnxm::KernelType::Gpu8x8x8:
 514             Nbnxm::gpu_launch_kernel(gpu_nbv, forceFlags, iLocality);
 515             break;
 516
 517         case Nbnxm::KernelType::Cpu8x8x8_PlainC:
 518             nbnxn_kernel_gpu_ref(pairlistSet.nblGpu[0],
 519                                  nbat, &ic,
 520                                  fr->shift_vec,
 521                                  forceFlags,
 522                                  clearF,
 523                                  nbat->out[0].f,
 524                                  fr->fshift[0],
 525                                  enerd->grpp.ener[egCOULSR],
 526                                  fr->bBHAM ?
 527                                  enerd->grpp.ener[egBHAMSR] :
 528                                  enerd->grpp.ener[egLJSR]);
 529             break;
 530
 531         default:
 532             GMX_RELEASE_ASSERT(false, "Invalid nonbonded kernel type passed!");
 533
 534     }
 535
 536     accountFlops(nrnb, pairlistSet, *this, ic, forceFlags);
 537 }
 538
 539 void
 540 nonbonded_verlet_t::dispatchFreeEnergyKernel(Nbnxm::InteractionLocality  iLocality,
 541                                              t_forcerec                 *fr,
 542                                              rvec                        x[],
 543                                              rvec                        f[],
 544                                              const t_mdatoms            &mdatoms,
 545                                              t_lambda                   *fepvals,
 546                                              real                       *lambda,
 547                                              gmx_enerdata_t             *enerd,
 548                                              const int                   forceFlags,
 549                                              t_nrnb                     *nrnb)
 550 {
 551     const gmx::ArrayRef<t_nblist const * const > nbl_fep = pairlistSets().pairlistSet(iLocality).nbl_fep;
 552
 553     /* When the first list is empty, all are empty and there is nothing to do */
 554     if (nbl_fep[0]->nrj == 0)
 555     {
 556         return;
 557     }
 558
 559     int donb_flags = 0;
 560     /* Add short-range interactions */
 561     donb_flags |= GMX_NONBONDED_DO_SR;
 562
 563     /* Currently all group scheme kernels always calculate (shift-)forces */
 564     if (forceFlags & GMX_FORCE_FORCES)
 565     {
 566         donb_flags |= GMX_NONBONDED_DO_FORCE;
 567     }
 568     if (forceFlags & GMX_FORCE_VIRIAL)
 569     {
 570         donb_flags |= GMX_NONBONDED_DO_SHIFTFORCE;
 571     }
 572     if (forceFlags & GMX_FORCE_ENERGY)
 573     {
 574         donb_flags |= GMX_NONBONDED_DO_POTENTIAL;
 575     }
 576
 577     nb_kernel_data_t kernel_data;
 578     real             dvdl_nb[efptNR] = { 0 };
 579     kernel_data.flags  = donb_flags;
 580     kernel_data.lambda = lambda;
 581     kernel_data.dvdl   = dvdl_nb;
 582
 583     kernel_data.energygrp_elec = enerd->grpp.ener[egCOULSR];
 584     kernel_data.energygrp_vdw  = enerd->grpp.ener[egLJSR];
 585
 586     GMX_ASSERT(gmx_omp_nthreads_get(emntNonbonded) == nbl_fep.ssize(), "Number of lists should be same as number of NB threads");
 587
 588 #pragma omp parallel for schedule(static) num_threads(nbl_fep.ssize())
 589     for (int th = 0; th < nbl_fep.ssize(); th++)
 590     {
 591         try
 592         {
 593             gmx_nb_free_energy_kernel(nbl_fep[th],
 594                                       x, f, fr, &mdatoms, &kernel_data, nrnb);
 595         }
 596         GMX_CATCH_ALL_AND_EXIT_WITH_FATAL_ERROR;
 597     }
 598
 599     if (fepvals->sc_alpha != 0)
 600     {
 601         enerd->dvdl_nonlin[efptVDW]  += dvdl_nb[efptVDW];
 602         enerd->dvdl_nonlin[efptCOUL] += dvdl_nb[efptCOUL];
 603     }
 604     else
 605     {
 606         enerd->dvdl_lin[efptVDW]  += dvdl_nb[efptVDW];
 607         enerd->dvdl_lin[efptCOUL] += dvdl_nb[efptCOUL];
 608     }
 609
 610     /* If we do foreign lambda and we have soft-core interactions
 611      * we have to recalculate the (non-linear) energies contributions.
 612      */
 613     if (fepvals->n_lambda > 0 && (forceFlags & GMX_FORCE_DHDL) && fepvals->sc_alpha != 0)
 614     {
 615         real lam_i[efptNR];
 616         kernel_data.flags          = (donb_flags & ~(GMX_NONBONDED_DO_FORCE | GMX_NONBONDED_DO_SHIFTFORCE)) | GMX_NONBONDED_DO_FOREIGNLAMBDA;
 617         kernel_data.lambda         = lam_i;
 618         kernel_data.energygrp_elec = enerd->foreign_grpp.ener[egCOULSR];
 619         kernel_data.energygrp_vdw  = enerd->foreign_grpp.ener[egLJSR];
 620         /* Note that we add to kernel_data.dvdl, but ignore the result */
 621
 622         for (int i = 0; i < enerd->n_lambda; i++)
 623         {
 624             for (int j = 0; j < efptNR; j++)
 625             {
 626                 lam_i[j] = (i == 0 ? lambda[j] : fepvals->all_lambda[j][i-1]);
 627             }
 628             reset_foreign_enerdata(enerd);
 629 #pragma omp parallel for schedule(static) num_threads(nbl_fep.ssize())
 630             for (int th = 0; th < nbl_fep.ssize(); th++)
 631             {
 632                 try
 633                 {
 634                     gmx_nb_free_energy_kernel(nbl_fep[th],
 635                                               x, f, fr, &mdatoms, &kernel_data, nrnb);
 636                 }
 637                 GMX_CATCH_ALL_AND_EXIT_WITH_FATAL_ERROR;
 638             }
 639
 640             sum_epot(&(enerd->foreign_grpp), enerd->foreign_term);
 641             enerd->enerpart_lambda[i] += enerd->foreign_term[F_EPOT];
 642         }
 643     }
 644 }